diff options
Diffstat (limited to 'kernel/drivers/s390/block')
24 files changed, 21797 insertions, 0 deletions
diff --git a/kernel/drivers/s390/block/Kconfig b/kernel/drivers/s390/block/Kconfig new file mode 100644 index 000000000..4a3b62326 --- /dev/null +++ b/kernel/drivers/s390/block/Kconfig @@ -0,0 +1,90 @@ +comment "S/390 block device drivers" + depends on S390 && BLOCK + +config BLK_DEV_XPRAM + def_tristate m + prompt "XPRAM disk support" + depends on S390 && BLOCK + help + Select this option if you want to use your expanded storage on S/390 + or zSeries as a disk. This is useful as a _fast_ swap device if you + want to access more than 2G of memory when running in 31 bit mode. + This option is also available as a module which will be called + xpram. If unsure, say "N". + +config DCSSBLK + def_tristate m + prompt "DCSSBLK support" + depends on S390 && BLOCK + help + Support for dcss block device + +config DASD + def_tristate y + prompt "Support for DASD devices" + depends on CCW && BLOCK + select IOSCHED_DEADLINE + help + Enable this option if you want to access DASDs directly utilizing + S/390s channel subsystem commands. This is necessary for running + natively on a single image or an LPAR. + +config DASD_PROFILE + def_bool y + prompt "Profiling support for dasd devices" + depends on DASD + help + Enable this option if you want to see profiling information + in /proc/dasd/statistics. + +config DASD_ECKD + def_tristate y + prompt "Support for ECKD Disks" + depends on DASD + help + ECKD devices are the most commonly used devices. You should enable + this option unless you are very sure to have no ECKD device. + +config DASD_FBA + def_tristate y + prompt "Support for FBA Disks" + depends on DASD + help + Select this option to be able to access FBA devices. It is safe to + say "Y". + +config DASD_DIAG + def_tristate y + prompt "Support for DIAG access to Disks" + depends on DASD + help + Select this option if you want to use Diagnose250 command to access + Disks under VM. If you are not running under VM or unsure what it is, + say "N". + +config DASD_EER + def_bool y + prompt "Extended error reporting (EER)" + depends on DASD + help + This driver provides a character device interface to the + DASD extended error reporting. This is only needed if you want to + use applications written for the EER facility. + +config SCM_BLOCK + def_tristate m + prompt "Support for Storage Class Memory" + depends on S390 && BLOCK && EADM_SCH && SCM_BUS + help + Block device driver for Storage Class Memory (SCM). This driver + provides a block device interface for each available SCM increment. + + To compile this driver as a module, choose M here: the + module will be called scm_block. + +config SCM_BLOCK_CLUSTER_WRITE + def_bool y + prompt "SCM force cluster writes" + depends on SCM_BLOCK + help + Force writes to Storage Class Memory (SCM) to be in done in clusters. diff --git a/kernel/drivers/s390/block/Makefile b/kernel/drivers/s390/block/Makefile new file mode 100644 index 000000000..c2f4e673e --- /dev/null +++ b/kernel/drivers/s390/block/Makefile @@ -0,0 +1,25 @@ +# +# S/390 block devices +# + +dasd_eckd_mod-objs := dasd_eckd.o dasd_3990_erp.o dasd_alias.o +dasd_fba_mod-objs := dasd_fba.o +dasd_diag_mod-objs := dasd_diag.o +dasd_mod-objs := dasd.o dasd_ioctl.o dasd_proc.o dasd_devmap.o \ + dasd_genhd.o dasd_erp.o +ifdef CONFIG_DASD_EER +dasd_mod-objs += dasd_eer.o +endif + +obj-$(CONFIG_DASD) += dasd_mod.o +obj-$(CONFIG_DASD_DIAG) += dasd_diag_mod.o +obj-$(CONFIG_DASD_ECKD) += dasd_eckd_mod.o +obj-$(CONFIG_DASD_FBA) += dasd_fba_mod.o +obj-$(CONFIG_BLK_DEV_XPRAM) += xpram.o +obj-$(CONFIG_DCSSBLK) += dcssblk.o + +scm_block-objs := scm_drv.o scm_blk.o +ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE +scm_block-objs += scm_blk_cluster.o +endif +obj-$(CONFIG_SCM_BLOCK) += scm_block.o diff --git a/kernel/drivers/s390/block/dasd.c b/kernel/drivers/s390/block/dasd.c new file mode 100644 index 000000000..57fd66357 --- /dev/null +++ b/kernel/drivers/s390/block/dasd.c @@ -0,0 +1,3879 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2009 + */ + +#define KMSG_COMPONENT "dasd" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include <linux/kmod.h> +#include <linux/init.h> +#include <linux/interrupt.h> +#include <linux/ctype.h> +#include <linux/major.h> +#include <linux/slab.h> +#include <linux/hdreg.h> +#include <linux/async.h> +#include <linux/mutex.h> +#include <linux/debugfs.h> +#include <linux/seq_file.h> +#include <linux/vmalloc.h> + +#include <asm/ccwdev.h> +#include <asm/ebcdic.h> +#include <asm/idals.h> +#include <asm/itcw.h> +#include <asm/diag.h> + +/* This is ugly... */ +#define PRINTK_HEADER "dasd:" + +#include "dasd_int.h" +/* + * SECTION: Constant definitions to be used within this file + */ +#define DASD_CHANQ_MAX_SIZE 4 + +/* + * SECTION: exported variables of dasd.c + */ +debug_info_t *dasd_debug_area; +EXPORT_SYMBOL(dasd_debug_area); +static struct dentry *dasd_debugfs_root_entry; +struct dasd_discipline *dasd_diag_discipline_pointer; +EXPORT_SYMBOL(dasd_diag_discipline_pointer); +void dasd_int_handler(struct ccw_device *, unsigned long, struct irb *); + +MODULE_AUTHOR("Holger Smolinski <Holger.Smolinski@de.ibm.com>"); +MODULE_DESCRIPTION("Linux on S/390 DASD device driver," + " Copyright IBM Corp. 2000"); +MODULE_SUPPORTED_DEVICE("dasd"); +MODULE_LICENSE("GPL"); + +/* + * SECTION: prototypes for static functions of dasd.c + */ +static int dasd_alloc_queue(struct dasd_block *); +static void dasd_setup_queue(struct dasd_block *); +static void dasd_free_queue(struct dasd_block *); +static void dasd_flush_request_queue(struct dasd_block *); +static int dasd_flush_block_queue(struct dasd_block *); +static void dasd_device_tasklet(struct dasd_device *); +static void dasd_block_tasklet(struct dasd_block *); +static void do_kick_device(struct work_struct *); +static void do_restore_device(struct work_struct *); +static void do_reload_device(struct work_struct *); +static void dasd_return_cqr_cb(struct dasd_ccw_req *, void *); +static void dasd_device_timeout(unsigned long); +static void dasd_block_timeout(unsigned long); +static void __dasd_process_erp(struct dasd_device *, struct dasd_ccw_req *); +static void dasd_profile_init(struct dasd_profile *, struct dentry *); +static void dasd_profile_exit(struct dasd_profile *); + +/* + * SECTION: Operations on the device structure. + */ +static wait_queue_head_t dasd_init_waitq; +static wait_queue_head_t dasd_flush_wq; +static wait_queue_head_t generic_waitq; +static wait_queue_head_t shutdown_waitq; + +/* + * Allocate memory for a new device structure. + */ +struct dasd_device *dasd_alloc_device(void) +{ + struct dasd_device *device; + + device = kzalloc(sizeof(struct dasd_device), GFP_ATOMIC); + if (!device) + return ERR_PTR(-ENOMEM); + + /* Get two pages for normal block device operations. */ + device->ccw_mem = (void *) __get_free_pages(GFP_ATOMIC | GFP_DMA, 1); + if (!device->ccw_mem) { + kfree(device); + return ERR_PTR(-ENOMEM); + } + /* Get one page for error recovery. */ + device->erp_mem = (void *) get_zeroed_page(GFP_ATOMIC | GFP_DMA); + if (!device->erp_mem) { + free_pages((unsigned long) device->ccw_mem, 1); + kfree(device); + return ERR_PTR(-ENOMEM); + } + + dasd_init_chunklist(&device->ccw_chunks, device->ccw_mem, PAGE_SIZE*2); + dasd_init_chunklist(&device->erp_chunks, device->erp_mem, PAGE_SIZE); + spin_lock_init(&device->mem_lock); + atomic_set(&device->tasklet_scheduled, 0); + tasklet_init(&device->tasklet, + (void (*)(unsigned long)) dasd_device_tasklet, + (unsigned long) device); + INIT_LIST_HEAD(&device->ccw_queue); + init_timer(&device->timer); + device->timer.function = dasd_device_timeout; + device->timer.data = (unsigned long) device; + INIT_WORK(&device->kick_work, do_kick_device); + INIT_WORK(&device->restore_device, do_restore_device); + INIT_WORK(&device->reload_device, do_reload_device); + device->state = DASD_STATE_NEW; + device->target = DASD_STATE_NEW; + mutex_init(&device->state_mutex); + spin_lock_init(&device->profile.lock); + return device; +} + +/* + * Free memory of a device structure. + */ +void dasd_free_device(struct dasd_device *device) +{ + kfree(device->private); + free_page((unsigned long) device->erp_mem); + free_pages((unsigned long) device->ccw_mem, 1); + kfree(device); +} + +/* + * Allocate memory for a new device structure. + */ +struct dasd_block *dasd_alloc_block(void) +{ + struct dasd_block *block; + + block = kzalloc(sizeof(*block), GFP_ATOMIC); + if (!block) + return ERR_PTR(-ENOMEM); + /* open_count = 0 means device online but not in use */ + atomic_set(&block->open_count, -1); + + spin_lock_init(&block->request_queue_lock); + atomic_set(&block->tasklet_scheduled, 0); + tasklet_init(&block->tasklet, + (void (*)(unsigned long)) dasd_block_tasklet, + (unsigned long) block); + INIT_LIST_HEAD(&block->ccw_queue); + spin_lock_init(&block->queue_lock); + init_timer(&block->timer); + block->timer.function = dasd_block_timeout; + block->timer.data = (unsigned long) block; + spin_lock_init(&block->profile.lock); + + return block; +} +EXPORT_SYMBOL_GPL(dasd_alloc_block); + +/* + * Free memory of a device structure. + */ +void dasd_free_block(struct dasd_block *block) +{ + kfree(block); +} +EXPORT_SYMBOL_GPL(dasd_free_block); + +/* + * Make a new device known to the system. + */ +static int dasd_state_new_to_known(struct dasd_device *device) +{ + int rc; + + /* + * As long as the device is not in state DASD_STATE_NEW we want to + * keep the reference count > 0. + */ + dasd_get_device(device); + + if (device->block) { + rc = dasd_alloc_queue(device->block); + if (rc) { + dasd_put_device(device); + return rc; + } + } + device->state = DASD_STATE_KNOWN; + return 0; +} + +/* + * Let the system forget about a device. + */ +static int dasd_state_known_to_new(struct dasd_device *device) +{ + /* Disable extended error reporting for this device. */ + dasd_eer_disable(device); + /* Forget the discipline information. */ + if (device->discipline) { + if (device->discipline->uncheck_device) + device->discipline->uncheck_device(device); + module_put(device->discipline->owner); + } + device->discipline = NULL; + if (device->base_discipline) + module_put(device->base_discipline->owner); + device->base_discipline = NULL; + device->state = DASD_STATE_NEW; + + if (device->block) + dasd_free_queue(device->block); + + /* Give up reference we took in dasd_state_new_to_known. */ + dasd_put_device(device); + return 0; +} + +static struct dentry *dasd_debugfs_setup(const char *name, + struct dentry *base_dentry) +{ + struct dentry *pde; + + if (!base_dentry) + return NULL; + pde = debugfs_create_dir(name, base_dentry); + if (!pde || IS_ERR(pde)) + return NULL; + return pde; +} + +/* + * Request the irq line for the device. + */ +static int dasd_state_known_to_basic(struct dasd_device *device) +{ + struct dasd_block *block = device->block; + int rc = 0; + + /* Allocate and register gendisk structure. */ + if (block) { + rc = dasd_gendisk_alloc(block); + if (rc) + return rc; + block->debugfs_dentry = + dasd_debugfs_setup(block->gdp->disk_name, + dasd_debugfs_root_entry); + dasd_profile_init(&block->profile, block->debugfs_dentry); + if (dasd_global_profile_level == DASD_PROFILE_ON) + dasd_profile_on(&device->block->profile); + } + device->debugfs_dentry = + dasd_debugfs_setup(dev_name(&device->cdev->dev), + dasd_debugfs_root_entry); + dasd_profile_init(&device->profile, device->debugfs_dentry); + + /* register 'device' debug area, used for all DBF_DEV_XXX calls */ + device->debug_area = debug_register(dev_name(&device->cdev->dev), 4, 1, + 8 * sizeof(long)); + debug_register_view(device->debug_area, &debug_sprintf_view); + debug_set_level(device->debug_area, DBF_WARNING); + DBF_DEV_EVENT(DBF_EMERG, device, "%s", "debug area created"); + + device->state = DASD_STATE_BASIC; + + return rc; +} + +/* + * Release the irq line for the device. Terminate any running i/o. + */ +static int dasd_state_basic_to_known(struct dasd_device *device) +{ + int rc; + + if (device->discipline->basic_to_known) { + rc = device->discipline->basic_to_known(device); + if (rc) + return rc; + } + + if (device->block) { + dasd_profile_exit(&device->block->profile); + debugfs_remove(device->block->debugfs_dentry); + dasd_gendisk_free(device->block); + dasd_block_clear_timer(device->block); + } + rc = dasd_flush_device_queue(device); + if (rc) + return rc; + dasd_device_clear_timer(device); + dasd_profile_exit(&device->profile); + debugfs_remove(device->debugfs_dentry); + DBF_DEV_EVENT(DBF_EMERG, device, "%p debug area deleted", device); + if (device->debug_area != NULL) { + debug_unregister(device->debug_area); + device->debug_area = NULL; + } + device->state = DASD_STATE_KNOWN; + return 0; +} + +/* + * Do the initial analysis. The do_analysis function may return + * -EAGAIN in which case the device keeps the state DASD_STATE_BASIC + * until the discipline decides to continue the startup sequence + * by calling the function dasd_change_state. The eckd disciplines + * uses this to start a ccw that detects the format. The completion + * interrupt for this detection ccw uses the kernel event daemon to + * trigger the call to dasd_change_state. All this is done in the + * discipline code, see dasd_eckd.c. + * After the analysis ccw is done (do_analysis returned 0) the block + * device is setup. + * In case the analysis returns an error, the device setup is stopped + * (a fake disk was already added to allow formatting). + */ +static int dasd_state_basic_to_ready(struct dasd_device *device) +{ + int rc; + struct dasd_block *block; + + rc = 0; + block = device->block; + /* make disk known with correct capacity */ + if (block) { + if (block->base->discipline->do_analysis != NULL) + rc = block->base->discipline->do_analysis(block); + if (rc) { + if (rc != -EAGAIN) { + device->state = DASD_STATE_UNFMT; + goto out; + } + return rc; + } + dasd_setup_queue(block); + set_capacity(block->gdp, + block->blocks << block->s2b_shift); + device->state = DASD_STATE_READY; + rc = dasd_scan_partitions(block); + if (rc) { + device->state = DASD_STATE_BASIC; + return rc; + } + } else { + device->state = DASD_STATE_READY; + } +out: + if (device->discipline->basic_to_ready) + rc = device->discipline->basic_to_ready(device); + return rc; +} + +static inline +int _wait_for_empty_queues(struct dasd_device *device) +{ + if (device->block) + return list_empty(&device->ccw_queue) && + list_empty(&device->block->ccw_queue); + else + return list_empty(&device->ccw_queue); +} + +/* + * Remove device from block device layer. Destroy dirty buffers. + * Forget format information. Check if the target level is basic + * and if it is create fake disk for formatting. + */ +static int dasd_state_ready_to_basic(struct dasd_device *device) +{ + int rc; + + device->state = DASD_STATE_BASIC; + if (device->block) { + struct dasd_block *block = device->block; + rc = dasd_flush_block_queue(block); + if (rc) { + device->state = DASD_STATE_READY; + return rc; + } + dasd_flush_request_queue(block); + dasd_destroy_partitions(block); + block->blocks = 0; + block->bp_block = 0; + block->s2b_shift = 0; + } + return 0; +} + +/* + * Back to basic. + */ +static int dasd_state_unfmt_to_basic(struct dasd_device *device) +{ + device->state = DASD_STATE_BASIC; + return 0; +} + +/* + * Make the device online and schedule the bottom half to start + * the requeueing of requests from the linux request queue to the + * ccw queue. + */ +static int +dasd_state_ready_to_online(struct dasd_device * device) +{ + struct gendisk *disk; + struct disk_part_iter piter; + struct hd_struct *part; + + device->state = DASD_STATE_ONLINE; + if (device->block) { + dasd_schedule_block_bh(device->block); + if ((device->features & DASD_FEATURE_USERAW)) { + disk = device->block->gdp; + kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE); + return 0; + } + disk = device->block->bdev->bd_disk; + disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); + while ((part = disk_part_iter_next(&piter))) + kobject_uevent(&part_to_dev(part)->kobj, KOBJ_CHANGE); + disk_part_iter_exit(&piter); + } + return 0; +} + +/* + * Stop the requeueing of requests again. + */ +static int dasd_state_online_to_ready(struct dasd_device *device) +{ + int rc; + struct gendisk *disk; + struct disk_part_iter piter; + struct hd_struct *part; + + if (device->discipline->online_to_ready) { + rc = device->discipline->online_to_ready(device); + if (rc) + return rc; + } + + device->state = DASD_STATE_READY; + if (device->block && !(device->features & DASD_FEATURE_USERAW)) { + disk = device->block->bdev->bd_disk; + disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0); + while ((part = disk_part_iter_next(&piter))) + kobject_uevent(&part_to_dev(part)->kobj, KOBJ_CHANGE); + disk_part_iter_exit(&piter); + } + return 0; +} + +/* + * Device startup state changes. + */ +static int dasd_increase_state(struct dasd_device *device) +{ + int rc; + + rc = 0; + if (device->state == DASD_STATE_NEW && + device->target >= DASD_STATE_KNOWN) + rc = dasd_state_new_to_known(device); + + if (!rc && + device->state == DASD_STATE_KNOWN && + device->target >= DASD_STATE_BASIC) + rc = dasd_state_known_to_basic(device); + + if (!rc && + device->state == DASD_STATE_BASIC && + device->target >= DASD_STATE_READY) + rc = dasd_state_basic_to_ready(device); + + if (!rc && + device->state == DASD_STATE_UNFMT && + device->target > DASD_STATE_UNFMT) + rc = -EPERM; + + if (!rc && + device->state == DASD_STATE_READY && + device->target >= DASD_STATE_ONLINE) + rc = dasd_state_ready_to_online(device); + + return rc; +} + +/* + * Device shutdown state changes. + */ +static int dasd_decrease_state(struct dasd_device *device) +{ + int rc; + + rc = 0; + if (device->state == DASD_STATE_ONLINE && + device->target <= DASD_STATE_READY) + rc = dasd_state_online_to_ready(device); + + if (!rc && + device->state == DASD_STATE_READY && + device->target <= DASD_STATE_BASIC) + rc = dasd_state_ready_to_basic(device); + + if (!rc && + device->state == DASD_STATE_UNFMT && + device->target <= DASD_STATE_BASIC) + rc = dasd_state_unfmt_to_basic(device); + + if (!rc && + device->state == DASD_STATE_BASIC && + device->target <= DASD_STATE_KNOWN) + rc = dasd_state_basic_to_known(device); + + if (!rc && + device->state == DASD_STATE_KNOWN && + device->target <= DASD_STATE_NEW) + rc = dasd_state_known_to_new(device); + + return rc; +} + +/* + * This is the main startup/shutdown routine. + */ +static void dasd_change_state(struct dasd_device *device) +{ + int rc; + + if (device->state == device->target) + /* Already where we want to go today... */ + return; + if (device->state < device->target) + rc = dasd_increase_state(device); + else + rc = dasd_decrease_state(device); + if (rc == -EAGAIN) + return; + if (rc) + device->target = device->state; + + /* let user-space know that the device status changed */ + kobject_uevent(&device->cdev->dev.kobj, KOBJ_CHANGE); + + if (device->state == device->target) + wake_up(&dasd_init_waitq); +} + +/* + * Kick starter for devices that did not complete the startup/shutdown + * procedure or were sleeping because of a pending state. + * dasd_kick_device will schedule a call do do_kick_device to the kernel + * event daemon. + */ +static void do_kick_device(struct work_struct *work) +{ + struct dasd_device *device = container_of(work, struct dasd_device, kick_work); + mutex_lock(&device->state_mutex); + dasd_change_state(device); + mutex_unlock(&device->state_mutex); + dasd_schedule_device_bh(device); + dasd_put_device(device); +} + +void dasd_kick_device(struct dasd_device *device) +{ + dasd_get_device(device); + /* queue call to dasd_kick_device to the kernel event daemon. */ + if (!schedule_work(&device->kick_work)) + dasd_put_device(device); +} +EXPORT_SYMBOL(dasd_kick_device); + +/* + * dasd_reload_device will schedule a call do do_reload_device to the kernel + * event daemon. + */ +static void do_reload_device(struct work_struct *work) +{ + struct dasd_device *device = container_of(work, struct dasd_device, + reload_device); + device->discipline->reload(device); + dasd_put_device(device); +} + +void dasd_reload_device(struct dasd_device *device) +{ + dasd_get_device(device); + /* queue call to dasd_reload_device to the kernel event daemon. */ + if (!schedule_work(&device->reload_device)) + dasd_put_device(device); +} +EXPORT_SYMBOL(dasd_reload_device); + +/* + * dasd_restore_device will schedule a call do do_restore_device to the kernel + * event daemon. + */ +static void do_restore_device(struct work_struct *work) +{ + struct dasd_device *device = container_of(work, struct dasd_device, + restore_device); + device->cdev->drv->restore(device->cdev); + dasd_put_device(device); +} + +void dasd_restore_device(struct dasd_device *device) +{ + dasd_get_device(device); + /* queue call to dasd_restore_device to the kernel event daemon. */ + if (!schedule_work(&device->restore_device)) + dasd_put_device(device); +} + +/* + * Set the target state for a device and starts the state change. + */ +void dasd_set_target_state(struct dasd_device *device, int target) +{ + dasd_get_device(device); + mutex_lock(&device->state_mutex); + /* If we are in probeonly mode stop at DASD_STATE_READY. */ + if (dasd_probeonly && target > DASD_STATE_READY) + target = DASD_STATE_READY; + if (device->target != target) { + if (device->state == target) + wake_up(&dasd_init_waitq); + device->target = target; + } + if (device->state != device->target) + dasd_change_state(device); + mutex_unlock(&device->state_mutex); + dasd_put_device(device); +} +EXPORT_SYMBOL(dasd_set_target_state); + +/* + * Enable devices with device numbers in [from..to]. + */ +static inline int _wait_for_device(struct dasd_device *device) +{ + return (device->state == device->target); +} + +void dasd_enable_device(struct dasd_device *device) +{ + dasd_set_target_state(device, DASD_STATE_ONLINE); + if (device->state <= DASD_STATE_KNOWN) + /* No discipline for device found. */ + dasd_set_target_state(device, DASD_STATE_NEW); + /* Now wait for the devices to come up. */ + wait_event(dasd_init_waitq, _wait_for_device(device)); + + dasd_reload_device(device); + if (device->discipline->kick_validate) + device->discipline->kick_validate(device); +} +EXPORT_SYMBOL(dasd_enable_device); + +/* + * SECTION: device operation (interrupt handler, start i/o, term i/o ...) + */ + +unsigned int dasd_global_profile_level = DASD_PROFILE_OFF; + +#ifdef CONFIG_DASD_PROFILE +struct dasd_profile dasd_global_profile = { + .lock = __SPIN_LOCK_UNLOCKED(dasd_global_profile.lock), +}; +static struct dentry *dasd_debugfs_global_entry; + +/* + * Add profiling information for cqr before execution. + */ +static void dasd_profile_start(struct dasd_block *block, + struct dasd_ccw_req *cqr, + struct request *req) +{ + struct list_head *l; + unsigned int counter; + struct dasd_device *device; + + /* count the length of the chanq for statistics */ + counter = 0; + if (dasd_global_profile_level || block->profile.data) + list_for_each(l, &block->ccw_queue) + if (++counter >= 31) + break; + + spin_lock(&dasd_global_profile.lock); + if (dasd_global_profile.data) { + dasd_global_profile.data->dasd_io_nr_req[counter]++; + if (rq_data_dir(req) == READ) + dasd_global_profile.data->dasd_read_nr_req[counter]++; + } + spin_unlock(&dasd_global_profile.lock); + + spin_lock(&block->profile.lock); + if (block->profile.data) { + block->profile.data->dasd_io_nr_req[counter]++; + if (rq_data_dir(req) == READ) + block->profile.data->dasd_read_nr_req[counter]++; + } + spin_unlock(&block->profile.lock); + + /* + * We count the request for the start device, even though it may run on + * some other device due to error recovery. This way we make sure that + * we count each request only once. + */ + device = cqr->startdev; + if (device->profile.data) { + counter = 1; /* request is not yet queued on the start device */ + list_for_each(l, &device->ccw_queue) + if (++counter >= 31) + break; + } + spin_lock(&device->profile.lock); + if (device->profile.data) { + device->profile.data->dasd_io_nr_req[counter]++; + if (rq_data_dir(req) == READ) + device->profile.data->dasd_read_nr_req[counter]++; + } + spin_unlock(&device->profile.lock); +} + +/* + * Add profiling information for cqr after execution. + */ + +#define dasd_profile_counter(value, index) \ +{ \ + for (index = 0; index < 31 && value >> (2+index); index++) \ + ; \ +} + +static void dasd_profile_end_add_data(struct dasd_profile_info *data, + int is_alias, + int is_tpm, + int is_read, + long sectors, + int sectors_ind, + int tottime_ind, + int tottimeps_ind, + int strtime_ind, + int irqtime_ind, + int irqtimeps_ind, + int endtime_ind) +{ + /* in case of an overflow, reset the whole profile */ + if (data->dasd_io_reqs == UINT_MAX) { + memset(data, 0, sizeof(*data)); + getnstimeofday(&data->starttod); + } + data->dasd_io_reqs++; + data->dasd_io_sects += sectors; + if (is_alias) + data->dasd_io_alias++; + if (is_tpm) + data->dasd_io_tpm++; + + data->dasd_io_secs[sectors_ind]++; + data->dasd_io_times[tottime_ind]++; + data->dasd_io_timps[tottimeps_ind]++; + data->dasd_io_time1[strtime_ind]++; + data->dasd_io_time2[irqtime_ind]++; + data->dasd_io_time2ps[irqtimeps_ind]++; + data->dasd_io_time3[endtime_ind]++; + + if (is_read) { + data->dasd_read_reqs++; + data->dasd_read_sects += sectors; + if (is_alias) + data->dasd_read_alias++; + if (is_tpm) + data->dasd_read_tpm++; + data->dasd_read_secs[sectors_ind]++; + data->dasd_read_times[tottime_ind]++; + data->dasd_read_time1[strtime_ind]++; + data->dasd_read_time2[irqtime_ind]++; + data->dasd_read_time3[endtime_ind]++; + } +} + +static void dasd_profile_end(struct dasd_block *block, + struct dasd_ccw_req *cqr, + struct request *req) +{ + long strtime, irqtime, endtime, tottime; /* in microseconds */ + long tottimeps, sectors; + struct dasd_device *device; + int sectors_ind, tottime_ind, tottimeps_ind, strtime_ind; + int irqtime_ind, irqtimeps_ind, endtime_ind; + + device = cqr->startdev; + if (!(dasd_global_profile_level || + block->profile.data || + device->profile.data)) + return; + + sectors = blk_rq_sectors(req); + if (!cqr->buildclk || !cqr->startclk || + !cqr->stopclk || !cqr->endclk || + !sectors) + return; + + strtime = ((cqr->startclk - cqr->buildclk) >> 12); + irqtime = ((cqr->stopclk - cqr->startclk) >> 12); + endtime = ((cqr->endclk - cqr->stopclk) >> 12); + tottime = ((cqr->endclk - cqr->buildclk) >> 12); + tottimeps = tottime / sectors; + + dasd_profile_counter(sectors, sectors_ind); + dasd_profile_counter(tottime, tottime_ind); + dasd_profile_counter(tottimeps, tottimeps_ind); + dasd_profile_counter(strtime, strtime_ind); + dasd_profile_counter(irqtime, irqtime_ind); + dasd_profile_counter(irqtime / sectors, irqtimeps_ind); + dasd_profile_counter(endtime, endtime_ind); + + spin_lock(&dasd_global_profile.lock); + if (dasd_global_profile.data) { + dasd_profile_end_add_data(dasd_global_profile.data, + cqr->startdev != block->base, + cqr->cpmode == 1, + rq_data_dir(req) == READ, + sectors, sectors_ind, tottime_ind, + tottimeps_ind, strtime_ind, + irqtime_ind, irqtimeps_ind, + endtime_ind); + } + spin_unlock(&dasd_global_profile.lock); + + spin_lock(&block->profile.lock); + if (block->profile.data) + dasd_profile_end_add_data(block->profile.data, + cqr->startdev != block->base, + cqr->cpmode == 1, + rq_data_dir(req) == READ, + sectors, sectors_ind, tottime_ind, + tottimeps_ind, strtime_ind, + irqtime_ind, irqtimeps_ind, + endtime_ind); + spin_unlock(&block->profile.lock); + + spin_lock(&device->profile.lock); + if (device->profile.data) + dasd_profile_end_add_data(device->profile.data, + cqr->startdev != block->base, + cqr->cpmode == 1, + rq_data_dir(req) == READ, + sectors, sectors_ind, tottime_ind, + tottimeps_ind, strtime_ind, + irqtime_ind, irqtimeps_ind, + endtime_ind); + spin_unlock(&device->profile.lock); +} + +void dasd_profile_reset(struct dasd_profile *profile) +{ + struct dasd_profile_info *data; + + spin_lock_bh(&profile->lock); + data = profile->data; + if (!data) { + spin_unlock_bh(&profile->lock); + return; + } + memset(data, 0, sizeof(*data)); + getnstimeofday(&data->starttod); + spin_unlock_bh(&profile->lock); +} + +int dasd_profile_on(struct dasd_profile *profile) +{ + struct dasd_profile_info *data; + + data = kzalloc(sizeof(*data), GFP_KERNEL); + if (!data) + return -ENOMEM; + spin_lock_bh(&profile->lock); + if (profile->data) { + spin_unlock_bh(&profile->lock); + kfree(data); + return 0; + } + getnstimeofday(&data->starttod); + profile->data = data; + spin_unlock_bh(&profile->lock); + return 0; +} + +void dasd_profile_off(struct dasd_profile *profile) +{ + spin_lock_bh(&profile->lock); + kfree(profile->data); + profile->data = NULL; + spin_unlock_bh(&profile->lock); +} + +char *dasd_get_user_string(const char __user *user_buf, size_t user_len) +{ + char *buffer; + + buffer = vmalloc(user_len + 1); + if (buffer == NULL) + return ERR_PTR(-ENOMEM); + if (copy_from_user(buffer, user_buf, user_len) != 0) { + vfree(buffer); + return ERR_PTR(-EFAULT); + } + /* got the string, now strip linefeed. */ + if (buffer[user_len - 1] == '\n') + buffer[user_len - 1] = 0; + else + buffer[user_len] = 0; + return buffer; +} + +static ssize_t dasd_stats_write(struct file *file, + const char __user *user_buf, + size_t user_len, loff_t *pos) +{ + char *buffer, *str; + int rc; + struct seq_file *m = (struct seq_file *)file->private_data; + struct dasd_profile *prof = m->private; + + if (user_len > 65536) + user_len = 65536; + buffer = dasd_get_user_string(user_buf, user_len); + if (IS_ERR(buffer)) + return PTR_ERR(buffer); + + str = skip_spaces(buffer); + rc = user_len; + if (strncmp(str, "reset", 5) == 0) { + dasd_profile_reset(prof); + } else if (strncmp(str, "on", 2) == 0) { + rc = dasd_profile_on(prof); + if (rc) + goto out; + rc = user_len; + if (prof == &dasd_global_profile) { + dasd_profile_reset(prof); + dasd_global_profile_level = DASD_PROFILE_GLOBAL_ONLY; + } + } else if (strncmp(str, "off", 3) == 0) { + if (prof == &dasd_global_profile) + dasd_global_profile_level = DASD_PROFILE_OFF; + dasd_profile_off(prof); + } else + rc = -EINVAL; +out: + vfree(buffer); + return rc; +} + +static void dasd_stats_array(struct seq_file *m, unsigned int *array) +{ + int i; + + for (i = 0; i < 32; i++) + seq_printf(m, "%u ", array[i]); + seq_putc(m, '\n'); +} + +static void dasd_stats_seq_print(struct seq_file *m, + struct dasd_profile_info *data) +{ + seq_printf(m, "start_time %ld.%09ld\n", + data->starttod.tv_sec, data->starttod.tv_nsec); + seq_printf(m, "total_requests %u\n", data->dasd_io_reqs); + seq_printf(m, "total_sectors %u\n", data->dasd_io_sects); + seq_printf(m, "total_pav %u\n", data->dasd_io_alias); + seq_printf(m, "total_hpf %u\n", data->dasd_io_tpm); + seq_puts(m, "histogram_sectors "); + dasd_stats_array(m, data->dasd_io_secs); + seq_puts(m, "histogram_io_times "); + dasd_stats_array(m, data->dasd_io_times); + seq_puts(m, "histogram_io_times_weighted "); + dasd_stats_array(m, data->dasd_io_timps); + seq_puts(m, "histogram_time_build_to_ssch "); + dasd_stats_array(m, data->dasd_io_time1); + seq_puts(m, "histogram_time_ssch_to_irq "); + dasd_stats_array(m, data->dasd_io_time2); + seq_puts(m, "histogram_time_ssch_to_irq_weighted "); + dasd_stats_array(m, data->dasd_io_time2ps); + seq_puts(m, "histogram_time_irq_to_end "); + dasd_stats_array(m, data->dasd_io_time3); + seq_puts(m, "histogram_ccw_queue_length "); + dasd_stats_array(m, data->dasd_io_nr_req); + seq_printf(m, "total_read_requests %u\n", data->dasd_read_reqs); + seq_printf(m, "total_read_sectors %u\n", data->dasd_read_sects); + seq_printf(m, "total_read_pav %u\n", data->dasd_read_alias); + seq_printf(m, "total_read_hpf %u\n", data->dasd_read_tpm); + seq_puts(m, "histogram_read_sectors "); + dasd_stats_array(m, data->dasd_read_secs); + seq_puts(m, "histogram_read_times "); + dasd_stats_array(m, data->dasd_read_times); + seq_puts(m, "histogram_read_time_build_to_ssch "); + dasd_stats_array(m, data->dasd_read_time1); + seq_puts(m, "histogram_read_time_ssch_to_irq "); + dasd_stats_array(m, data->dasd_read_time2); + seq_puts(m, "histogram_read_time_irq_to_end "); + dasd_stats_array(m, data->dasd_read_time3); + seq_puts(m, "histogram_read_ccw_queue_length "); + dasd_stats_array(m, data->dasd_read_nr_req); +} + +static int dasd_stats_show(struct seq_file *m, void *v) +{ + struct dasd_profile *profile; + struct dasd_profile_info *data; + + profile = m->private; + spin_lock_bh(&profile->lock); + data = profile->data; + if (!data) { + spin_unlock_bh(&profile->lock); + seq_puts(m, "disabled\n"); + return 0; + } + dasd_stats_seq_print(m, data); + spin_unlock_bh(&profile->lock); + return 0; +} + +static int dasd_stats_open(struct inode *inode, struct file *file) +{ + struct dasd_profile *profile = inode->i_private; + return single_open(file, dasd_stats_show, profile); +} + +static const struct file_operations dasd_stats_raw_fops = { + .owner = THIS_MODULE, + .open = dasd_stats_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, + .write = dasd_stats_write, +}; + +static void dasd_profile_init(struct dasd_profile *profile, + struct dentry *base_dentry) +{ + umode_t mode; + struct dentry *pde; + + if (!base_dentry) + return; + profile->dentry = NULL; + profile->data = NULL; + mode = (S_IRUSR | S_IWUSR | S_IFREG); + pde = debugfs_create_file("statistics", mode, base_dentry, + profile, &dasd_stats_raw_fops); + if (pde && !IS_ERR(pde)) + profile->dentry = pde; + return; +} + +static void dasd_profile_exit(struct dasd_profile *profile) +{ + dasd_profile_off(profile); + debugfs_remove(profile->dentry); + profile->dentry = NULL; +} + +static void dasd_statistics_removeroot(void) +{ + dasd_global_profile_level = DASD_PROFILE_OFF; + dasd_profile_exit(&dasd_global_profile); + debugfs_remove(dasd_debugfs_global_entry); + debugfs_remove(dasd_debugfs_root_entry); +} + +static void dasd_statistics_createroot(void) +{ + struct dentry *pde; + + dasd_debugfs_root_entry = NULL; + pde = debugfs_create_dir("dasd", NULL); + if (!pde || IS_ERR(pde)) + goto error; + dasd_debugfs_root_entry = pde; + pde = debugfs_create_dir("global", dasd_debugfs_root_entry); + if (!pde || IS_ERR(pde)) + goto error; + dasd_debugfs_global_entry = pde; + dasd_profile_init(&dasd_global_profile, dasd_debugfs_global_entry); + return; + +error: + DBF_EVENT(DBF_ERR, "%s", + "Creation of the dasd debugfs interface failed"); + dasd_statistics_removeroot(); + return; +} + +#else +#define dasd_profile_start(block, cqr, req) do {} while (0) +#define dasd_profile_end(block, cqr, req) do {} while (0) + +static void dasd_statistics_createroot(void) +{ + return; +} + +static void dasd_statistics_removeroot(void) +{ + return; +} + +int dasd_stats_generic_show(struct seq_file *m, void *v) +{ + seq_puts(m, "Statistics are not activated in this kernel\n"); + return 0; +} + +static void dasd_profile_init(struct dasd_profile *profile, + struct dentry *base_dentry) +{ + return; +} + +static void dasd_profile_exit(struct dasd_profile *profile) +{ + return; +} + +int dasd_profile_on(struct dasd_profile *profile) +{ + return 0; +} + +#endif /* CONFIG_DASD_PROFILE */ + +/* + * Allocate memory for a channel program with 'cplength' channel + * command words and 'datasize' additional space. There are two + * variantes: 1) dasd_kmalloc_request uses kmalloc to get the needed + * memory and 2) dasd_smalloc_request uses the static ccw memory + * that gets allocated for each device. + */ +struct dasd_ccw_req *dasd_kmalloc_request(int magic, int cplength, + int datasize, + struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + + /* Sanity checks */ + BUG_ON(datasize > PAGE_SIZE || + (cplength*sizeof(struct ccw1)) > PAGE_SIZE); + + cqr = kzalloc(sizeof(struct dasd_ccw_req), GFP_ATOMIC); + if (cqr == NULL) + return ERR_PTR(-ENOMEM); + cqr->cpaddr = NULL; + if (cplength > 0) { + cqr->cpaddr = kcalloc(cplength, sizeof(struct ccw1), + GFP_ATOMIC | GFP_DMA); + if (cqr->cpaddr == NULL) { + kfree(cqr); + return ERR_PTR(-ENOMEM); + } + } + cqr->data = NULL; + if (datasize > 0) { + cqr->data = kzalloc(datasize, GFP_ATOMIC | GFP_DMA); + if (cqr->data == NULL) { + kfree(cqr->cpaddr); + kfree(cqr); + return ERR_PTR(-ENOMEM); + } + } + cqr->magic = magic; + set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + dasd_get_device(device); + return cqr; +} +EXPORT_SYMBOL(dasd_kmalloc_request); + +struct dasd_ccw_req *dasd_smalloc_request(int magic, int cplength, + int datasize, + struct dasd_device *device) +{ + unsigned long flags; + struct dasd_ccw_req *cqr; + char *data; + int size; + + size = (sizeof(struct dasd_ccw_req) + 7L) & -8L; + if (cplength > 0) + size += cplength * sizeof(struct ccw1); + if (datasize > 0) + size += datasize; + spin_lock_irqsave(&device->mem_lock, flags); + cqr = (struct dasd_ccw_req *) + dasd_alloc_chunk(&device->ccw_chunks, size); + spin_unlock_irqrestore(&device->mem_lock, flags); + if (cqr == NULL) + return ERR_PTR(-ENOMEM); + memset(cqr, 0, sizeof(struct dasd_ccw_req)); + data = (char *) cqr + ((sizeof(struct dasd_ccw_req) + 7L) & -8L); + cqr->cpaddr = NULL; + if (cplength > 0) { + cqr->cpaddr = (struct ccw1 *) data; + data += cplength*sizeof(struct ccw1); + memset(cqr->cpaddr, 0, cplength*sizeof(struct ccw1)); + } + cqr->data = NULL; + if (datasize > 0) { + cqr->data = data; + memset(cqr->data, 0, datasize); + } + cqr->magic = magic; + set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + dasd_get_device(device); + return cqr; +} +EXPORT_SYMBOL(dasd_smalloc_request); + +/* + * Free memory of a channel program. This function needs to free all the + * idal lists that might have been created by dasd_set_cda and the + * struct dasd_ccw_req itself. + */ +void dasd_kfree_request(struct dasd_ccw_req *cqr, struct dasd_device *device) +{ + struct ccw1 *ccw; + + /* Clear any idals used for the request. */ + ccw = cqr->cpaddr; + do { + clear_normalized_cda(ccw); + } while (ccw++->flags & (CCW_FLAG_CC | CCW_FLAG_DC)); + kfree(cqr->cpaddr); + kfree(cqr->data); + kfree(cqr); + dasd_put_device(device); +} +EXPORT_SYMBOL(dasd_kfree_request); + +void dasd_sfree_request(struct dasd_ccw_req *cqr, struct dasd_device *device) +{ + unsigned long flags; + + spin_lock_irqsave(&device->mem_lock, flags); + dasd_free_chunk(&device->ccw_chunks, cqr); + spin_unlock_irqrestore(&device->mem_lock, flags); + dasd_put_device(device); +} +EXPORT_SYMBOL(dasd_sfree_request); + +/* + * Check discipline magic in cqr. + */ +static inline int dasd_check_cqr(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + + if (cqr == NULL) + return -EINVAL; + device = cqr->startdev; + if (strncmp((char *) &cqr->magic, device->discipline->ebcname, 4)) { + DBF_DEV_EVENT(DBF_WARNING, device, + " dasd_ccw_req 0x%08x magic doesn't match" + " discipline 0x%08x", + cqr->magic, + *(unsigned int *) device->discipline->name); + return -EINVAL; + } + return 0; +} + +/* + * Terminate the current i/o and set the request to clear_pending. + * Timer keeps device runnig. + * ccw_device_clear can fail if the i/o subsystem + * is in a bad mood. + */ +int dasd_term_IO(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + int retries, rc; + char errorstring[ERRORLENGTH]; + + /* Check the cqr */ + rc = dasd_check_cqr(cqr); + if (rc) + return rc; + retries = 0; + device = (struct dasd_device *) cqr->startdev; + while ((retries < 5) && (cqr->status == DASD_CQR_IN_IO)) { + rc = ccw_device_clear(device->cdev, (long) cqr); + switch (rc) { + case 0: /* termination successful */ + cqr->status = DASD_CQR_CLEAR_PENDING; + cqr->stopclk = get_tod_clock(); + cqr->starttime = 0; + DBF_DEV_EVENT(DBF_DEBUG, device, + "terminate cqr %p successful", + cqr); + break; + case -ENODEV: + DBF_DEV_EVENT(DBF_ERR, device, "%s", + "device gone, retry"); + break; + case -EIO: + DBF_DEV_EVENT(DBF_ERR, device, "%s", + "I/O error, retry"); + break; + case -EINVAL: + /* + * device not valid so no I/O could be running + * handle CQR as termination successful + */ + cqr->status = DASD_CQR_CLEARED; + cqr->stopclk = get_tod_clock(); + cqr->starttime = 0; + /* no retries for invalid devices */ + cqr->retries = -1; + DBF_DEV_EVENT(DBF_ERR, device, "%s", + "EINVAL, handle as terminated"); + /* fake rc to success */ + rc = 0; + break; + case -EBUSY: + DBF_DEV_EVENT(DBF_ERR, device, "%s", + "device busy, retry later"); + break; + default: + /* internal error 10 - unknown rc*/ + snprintf(errorstring, ERRORLENGTH, "10 %d", rc); + dev_err(&device->cdev->dev, "An error occurred in the " + "DASD device driver, reason=%s\n", errorstring); + BUG(); + break; + } + retries++; + } + dasd_schedule_device_bh(device); + return rc; +} +EXPORT_SYMBOL(dasd_term_IO); + +/* + * Start the i/o. This start_IO can fail if the channel is really busy. + * In that case set up a timer to start the request later. + */ +int dasd_start_IO(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + int rc; + char errorstring[ERRORLENGTH]; + + /* Check the cqr */ + rc = dasd_check_cqr(cqr); + if (rc) { + cqr->intrc = rc; + return rc; + } + device = (struct dasd_device *) cqr->startdev; + if (((cqr->block && + test_bit(DASD_FLAG_LOCK_STOLEN, &cqr->block->base->flags)) || + test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags)) && + !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) { + DBF_DEV_EVENT(DBF_DEBUG, device, "start_IO: return request %p " + "because of stolen lock", cqr); + cqr->status = DASD_CQR_ERROR; + cqr->intrc = -EPERM; + return -EPERM; + } + if (cqr->retries < 0) { + /* internal error 14 - start_IO run out of retries */ + sprintf(errorstring, "14 %p", cqr); + dev_err(&device->cdev->dev, "An error occurred in the DASD " + "device driver, reason=%s\n", errorstring); + cqr->status = DASD_CQR_ERROR; + return -EIO; + } + cqr->startclk = get_tod_clock(); + cqr->starttime = jiffies; + cqr->retries--; + if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) { + cqr->lpm &= device->path_data.opm; + if (!cqr->lpm) + cqr->lpm = device->path_data.opm; + } + if (cqr->cpmode == 1) { + rc = ccw_device_tm_start(device->cdev, cqr->cpaddr, + (long) cqr, cqr->lpm); + } else { + rc = ccw_device_start(device->cdev, cqr->cpaddr, + (long) cqr, cqr->lpm, 0); + } + switch (rc) { + case 0: + cqr->status = DASD_CQR_IN_IO; + break; + case -EBUSY: + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "start_IO: device busy, retry later"); + break; + case -ETIMEDOUT: + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "start_IO: request timeout, retry later"); + break; + case -EACCES: + /* -EACCES indicates that the request used only a subset of the + * available paths and all these paths are gone. If the lpm of + * this request was only a subset of the opm (e.g. the ppm) then + * we just do a retry with all available paths. + * If we already use the full opm, something is amiss, and we + * need a full path verification. + */ + if (test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) { + DBF_DEV_EVENT(DBF_WARNING, device, + "start_IO: selected paths gone (%x)", + cqr->lpm); + } else if (cqr->lpm != device->path_data.opm) { + cqr->lpm = device->path_data.opm; + DBF_DEV_EVENT(DBF_DEBUG, device, "%s", + "start_IO: selected paths gone," + " retry on all paths"); + } else { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "start_IO: all paths in opm gone," + " do path verification"); + dasd_generic_last_path_gone(device); + device->path_data.opm = 0; + device->path_data.ppm = 0; + device->path_data.npm = 0; + device->path_data.tbvpm = + ccw_device_get_path_mask(device->cdev); + } + break; + case -ENODEV: + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "start_IO: -ENODEV device gone, retry"); + break; + case -EIO: + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "start_IO: -EIO device gone, retry"); + break; + case -EINVAL: + /* most likely caused in power management context */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "start_IO: -EINVAL device currently " + "not accessible"); + break; + default: + /* internal error 11 - unknown rc */ + snprintf(errorstring, ERRORLENGTH, "11 %d", rc); + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", errorstring); + BUG(); + break; + } + cqr->intrc = rc; + return rc; +} +EXPORT_SYMBOL(dasd_start_IO); + +/* + * Timeout function for dasd devices. This is used for different purposes + * 1) missing interrupt handler for normal operation + * 2) delayed start of request where start_IO failed with -EBUSY + * 3) timeout for missing state change interrupts + * The head of the ccw queue will have status DASD_CQR_IN_IO for 1), + * DASD_CQR_QUEUED for 2) and 3). + */ +static void dasd_device_timeout(unsigned long ptr) +{ + unsigned long flags; + struct dasd_device *device; + + device = (struct dasd_device *) ptr; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + /* re-activate request queue */ + dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + dasd_schedule_device_bh(device); +} + +/* + * Setup timeout for a device in jiffies. + */ +void dasd_device_set_timer(struct dasd_device *device, int expires) +{ + if (expires == 0) + del_timer(&device->timer); + else + mod_timer(&device->timer, jiffies + expires); +} +EXPORT_SYMBOL(dasd_device_set_timer); + +/* + * Clear timeout for a device. + */ +void dasd_device_clear_timer(struct dasd_device *device) +{ + del_timer(&device->timer); +} +EXPORT_SYMBOL(dasd_device_clear_timer); + +static void dasd_handle_killed_request(struct ccw_device *cdev, + unsigned long intparm) +{ + struct dasd_ccw_req *cqr; + struct dasd_device *device; + + if (!intparm) + return; + cqr = (struct dasd_ccw_req *) intparm; + if (cqr->status != DASD_CQR_IN_IO) { + DBF_EVENT_DEVID(DBF_DEBUG, cdev, + "invalid status in handle_killed_request: " + "%02x", cqr->status); + return; + } + + device = dasd_device_from_cdev_locked(cdev); + if (IS_ERR(device)) { + DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s", + "unable to get device from cdev"); + return; + } + + if (!cqr->startdev || + device != cqr->startdev || + strncmp(cqr->startdev->discipline->ebcname, + (char *) &cqr->magic, 4)) { + DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s", + "invalid device in request"); + dasd_put_device(device); + return; + } + + /* Schedule request to be retried. */ + cqr->status = DASD_CQR_QUEUED; + + dasd_device_clear_timer(device); + dasd_schedule_device_bh(device); + dasd_put_device(device); +} + +void dasd_generic_handle_state_change(struct dasd_device *device) +{ + /* First of all start sense subsystem status request. */ + dasd_eer_snss(device); + + dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING); + dasd_schedule_device_bh(device); + if (device->block) + dasd_schedule_block_bh(device->block); +} +EXPORT_SYMBOL_GPL(dasd_generic_handle_state_change); + +/* + * Interrupt handler for "normal" ssch-io based dasd devices. + */ +void dasd_int_handler(struct ccw_device *cdev, unsigned long intparm, + struct irb *irb) +{ + struct dasd_ccw_req *cqr, *next; + struct dasd_device *device; + unsigned long long now; + int expires; + + if (IS_ERR(irb)) { + switch (PTR_ERR(irb)) { + case -EIO: + break; + case -ETIMEDOUT: + DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: " + "request timed out\n", __func__); + break; + default: + DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: " + "unknown error %ld\n", __func__, + PTR_ERR(irb)); + } + dasd_handle_killed_request(cdev, intparm); + return; + } + + now = get_tod_clock(); + cqr = (struct dasd_ccw_req *) intparm; + /* check for conditions that should be handled immediately */ + if (!cqr || + !(scsw_dstat(&irb->scsw) == (DEV_STAT_CHN_END | DEV_STAT_DEV_END) && + scsw_cstat(&irb->scsw) == 0)) { + if (cqr) + memcpy(&cqr->irb, irb, sizeof(*irb)); + device = dasd_device_from_cdev_locked(cdev); + if (IS_ERR(device)) + return; + /* ignore unsolicited interrupts for DIAG discipline */ + if (device->discipline == dasd_diag_discipline_pointer) { + dasd_put_device(device); + return; + } + device->discipline->dump_sense_dbf(device, irb, "int"); + if (device->features & DASD_FEATURE_ERPLOG) + device->discipline->dump_sense(device, cqr, irb); + device->discipline->check_for_device_change(device, cqr, irb); + dasd_put_device(device); + } + + /* check for for attention message */ + if (scsw_dstat(&irb->scsw) & DEV_STAT_ATTENTION) { + device = dasd_device_from_cdev_locked(cdev); + device->discipline->check_attention(device, irb->esw.esw1.lpum); + dasd_put_device(device); + } + + if (!cqr) + return; + + device = (struct dasd_device *) cqr->startdev; + if (!device || + strncmp(device->discipline->ebcname, (char *) &cqr->magic, 4)) { + DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s", + "invalid device in request"); + return; + } + + /* Check for clear pending */ + if (cqr->status == DASD_CQR_CLEAR_PENDING && + scsw_fctl(&irb->scsw) & SCSW_FCTL_CLEAR_FUNC) { + cqr->status = DASD_CQR_CLEARED; + dasd_device_clear_timer(device); + wake_up(&dasd_flush_wq); + dasd_schedule_device_bh(device); + return; + } + + /* check status - the request might have been killed by dyn detach */ + if (cqr->status != DASD_CQR_IN_IO) { + DBF_DEV_EVENT(DBF_DEBUG, device, "invalid status: bus_id %s, " + "status %02x", dev_name(&cdev->dev), cqr->status); + return; + } + + next = NULL; + expires = 0; + if (scsw_dstat(&irb->scsw) == (DEV_STAT_CHN_END | DEV_STAT_DEV_END) && + scsw_cstat(&irb->scsw) == 0) { + /* request was completed successfully */ + cqr->status = DASD_CQR_SUCCESS; + cqr->stopclk = now; + /* Start first request on queue if possible -> fast_io. */ + if (cqr->devlist.next != &device->ccw_queue) { + next = list_entry(cqr->devlist.next, + struct dasd_ccw_req, devlist); + } + } else { /* error */ + /* + * If we don't want complex ERP for this request, then just + * reset this and retry it in the fastpath + */ + if (!test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags) && + cqr->retries > 0) { + if (cqr->lpm == device->path_data.opm) + DBF_DEV_EVENT(DBF_DEBUG, device, + "default ERP in fastpath " + "(%i retries left)", + cqr->retries); + if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) + cqr->lpm = device->path_data.opm; + cqr->status = DASD_CQR_QUEUED; + next = cqr; + } else + cqr->status = DASD_CQR_ERROR; + } + if (next && (next->status == DASD_CQR_QUEUED) && + (!device->stopped)) { + if (device->discipline->start_IO(next) == 0) + expires = next->expires; + } + if (expires != 0) + dasd_device_set_timer(device, expires); + else + dasd_device_clear_timer(device); + dasd_schedule_device_bh(device); +} +EXPORT_SYMBOL(dasd_int_handler); + +enum uc_todo dasd_generic_uc_handler(struct ccw_device *cdev, struct irb *irb) +{ + struct dasd_device *device; + + device = dasd_device_from_cdev_locked(cdev); + + if (IS_ERR(device)) + goto out; + if (test_bit(DASD_FLAG_OFFLINE, &device->flags) || + device->state != device->target || + !device->discipline->check_for_device_change){ + dasd_put_device(device); + goto out; + } + if (device->discipline->dump_sense_dbf) + device->discipline->dump_sense_dbf(device, irb, "uc"); + device->discipline->check_for_device_change(device, NULL, irb); + dasd_put_device(device); +out: + return UC_TODO_RETRY; +} +EXPORT_SYMBOL_GPL(dasd_generic_uc_handler); + +/* + * If we have an error on a dasd_block layer request then we cancel + * and return all further requests from the same dasd_block as well. + */ +static void __dasd_device_recovery(struct dasd_device *device, + struct dasd_ccw_req *ref_cqr) +{ + struct list_head *l, *n; + struct dasd_ccw_req *cqr; + + /* + * only requeue request that came from the dasd_block layer + */ + if (!ref_cqr->block) + return; + + list_for_each_safe(l, n, &device->ccw_queue) { + cqr = list_entry(l, struct dasd_ccw_req, devlist); + if (cqr->status == DASD_CQR_QUEUED && + ref_cqr->block == cqr->block) { + cqr->status = DASD_CQR_CLEARED; + } + } +}; + +/* + * Remove those ccw requests from the queue that need to be returned + * to the upper layer. + */ +static void __dasd_device_process_ccw_queue(struct dasd_device *device, + struct list_head *final_queue) +{ + struct list_head *l, *n; + struct dasd_ccw_req *cqr; + + /* Process request with final status. */ + list_for_each_safe(l, n, &device->ccw_queue) { + cqr = list_entry(l, struct dasd_ccw_req, devlist); + + /* Skip any non-final request. */ + if (cqr->status == DASD_CQR_QUEUED || + cqr->status == DASD_CQR_IN_IO || + cqr->status == DASD_CQR_CLEAR_PENDING) + continue; + if (cqr->status == DASD_CQR_ERROR) { + __dasd_device_recovery(device, cqr); + } + /* Rechain finished requests to final queue */ + list_move_tail(&cqr->devlist, final_queue); + } +} + +/* + * the cqrs from the final queue are returned to the upper layer + * by setting a dasd_block state and calling the callback function + */ +static void __dasd_device_process_final_queue(struct dasd_device *device, + struct list_head *final_queue) +{ + struct list_head *l, *n; + struct dasd_ccw_req *cqr; + struct dasd_block *block; + void (*callback)(struct dasd_ccw_req *, void *data); + void *callback_data; + char errorstring[ERRORLENGTH]; + + list_for_each_safe(l, n, final_queue) { + cqr = list_entry(l, struct dasd_ccw_req, devlist); + list_del_init(&cqr->devlist); + block = cqr->block; + callback = cqr->callback; + callback_data = cqr->callback_data; + if (block) + spin_lock_bh(&block->queue_lock); + switch (cqr->status) { + case DASD_CQR_SUCCESS: + cqr->status = DASD_CQR_DONE; + break; + case DASD_CQR_ERROR: + cqr->status = DASD_CQR_NEED_ERP; + break; + case DASD_CQR_CLEARED: + cqr->status = DASD_CQR_TERMINATED; + break; + default: + /* internal error 12 - wrong cqr status*/ + snprintf(errorstring, ERRORLENGTH, "12 %p %x02", cqr, cqr->status); + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", errorstring); + BUG(); + } + if (cqr->callback != NULL) + (callback)(cqr, callback_data); + if (block) + spin_unlock_bh(&block->queue_lock); + } +} + +/* + * Take a look at the first request on the ccw queue and check + * if it reached its expire time. If so, terminate the IO. + */ +static void __dasd_device_check_expire(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + + if (list_empty(&device->ccw_queue)) + return; + cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist); + if ((cqr->status == DASD_CQR_IN_IO && cqr->expires != 0) && + (time_after_eq(jiffies, cqr->expires + cqr->starttime))) { + if (test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) { + /* + * IO in safe offline processing should not + * run out of retries + */ + cqr->retries++; + } + if (device->discipline->term_IO(cqr) != 0) { + /* Hmpf, try again in 5 sec */ + dev_err(&device->cdev->dev, + "cqr %p timed out (%lus) but cannot be " + "ended, retrying in 5 s\n", + cqr, (cqr->expires/HZ)); + cqr->expires += 5*HZ; + dasd_device_set_timer(device, 5*HZ); + } else { + dev_err(&device->cdev->dev, + "cqr %p timed out (%lus), %i retries " + "remaining\n", cqr, (cqr->expires/HZ), + cqr->retries); + } + } +} + +/* + * Take a look at the first request on the ccw queue and check + * if it needs to be started. + */ +static void __dasd_device_start_head(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + int rc; + + if (list_empty(&device->ccw_queue)) + return; + cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist); + if (cqr->status != DASD_CQR_QUEUED) + return; + /* when device is stopped, return request to previous layer + * exception: only the disconnect or unresumed bits are set and the + * cqr is a path verification request + */ + if (device->stopped && + !(!(device->stopped & ~(DASD_STOPPED_DC_WAIT | DASD_UNRESUMED_PM)) + && test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags))) { + cqr->intrc = -EAGAIN; + cqr->status = DASD_CQR_CLEARED; + dasd_schedule_device_bh(device); + return; + } + + rc = device->discipline->start_IO(cqr); + if (rc == 0) + dasd_device_set_timer(device, cqr->expires); + else if (rc == -EACCES) { + dasd_schedule_device_bh(device); + } else + /* Hmpf, try again in 1/2 sec */ + dasd_device_set_timer(device, 50); +} + +static void __dasd_device_check_path_events(struct dasd_device *device) +{ + int rc; + + if (device->path_data.tbvpm) { + if (device->stopped & ~(DASD_STOPPED_DC_WAIT | + DASD_UNRESUMED_PM)) + return; + rc = device->discipline->verify_path( + device, device->path_data.tbvpm); + if (rc) + dasd_device_set_timer(device, 50); + else + device->path_data.tbvpm = 0; + } +}; + +/* + * Go through all request on the dasd_device request queue, + * terminate them on the cdev if necessary, and return them to the + * submitting layer via callback. + * Note: + * Make sure that all 'submitting layers' still exist when + * this function is called!. In other words, when 'device' is a base + * device then all block layer requests must have been removed before + * via dasd_flush_block_queue. + */ +int dasd_flush_device_queue(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr, *n; + int rc; + struct list_head flush_queue; + + INIT_LIST_HEAD(&flush_queue); + spin_lock_irq(get_ccwdev_lock(device->cdev)); + rc = 0; + list_for_each_entry_safe(cqr, n, &device->ccw_queue, devlist) { + /* Check status and move request to flush_queue */ + switch (cqr->status) { + case DASD_CQR_IN_IO: + rc = device->discipline->term_IO(cqr); + if (rc) { + /* unable to terminate requeust */ + dev_err(&device->cdev->dev, + "Flushing the DASD request queue " + "failed for request %p\n", cqr); + /* stop flush processing */ + goto finished; + } + break; + case DASD_CQR_QUEUED: + cqr->stopclk = get_tod_clock(); + cqr->status = DASD_CQR_CLEARED; + break; + default: /* no need to modify the others */ + break; + } + list_move_tail(&cqr->devlist, &flush_queue); + } +finished: + spin_unlock_irq(get_ccwdev_lock(device->cdev)); + /* + * After this point all requests must be in state CLEAR_PENDING, + * CLEARED, SUCCESS or ERROR. Now wait for CLEAR_PENDING to become + * one of the others. + */ + list_for_each_entry_safe(cqr, n, &flush_queue, devlist) + wait_event(dasd_flush_wq, + (cqr->status != DASD_CQR_CLEAR_PENDING)); + /* + * Now set each request back to TERMINATED, DONE or NEED_ERP + * and call the callback function of flushed requests + */ + __dasd_device_process_final_queue(device, &flush_queue); + return rc; +} +EXPORT_SYMBOL_GPL(dasd_flush_device_queue); + +/* + * Acquire the device lock and process queues for the device. + */ +static void dasd_device_tasklet(struct dasd_device *device) +{ + struct list_head final_queue; + + atomic_set (&device->tasklet_scheduled, 0); + INIT_LIST_HEAD(&final_queue); + spin_lock_irq(get_ccwdev_lock(device->cdev)); + /* Check expire time of first request on the ccw queue. */ + __dasd_device_check_expire(device); + /* find final requests on ccw queue */ + __dasd_device_process_ccw_queue(device, &final_queue); + __dasd_device_check_path_events(device); + spin_unlock_irq(get_ccwdev_lock(device->cdev)); + /* Now call the callback function of requests with final status */ + __dasd_device_process_final_queue(device, &final_queue); + spin_lock_irq(get_ccwdev_lock(device->cdev)); + /* Now check if the head of the ccw queue needs to be started. */ + __dasd_device_start_head(device); + spin_unlock_irq(get_ccwdev_lock(device->cdev)); + if (waitqueue_active(&shutdown_waitq)) + wake_up(&shutdown_waitq); + dasd_put_device(device); +} + +/* + * Schedules a call to dasd_tasklet over the device tasklet. + */ +void dasd_schedule_device_bh(struct dasd_device *device) +{ + /* Protect against rescheduling. */ + if (atomic_cmpxchg (&device->tasklet_scheduled, 0, 1) != 0) + return; + dasd_get_device(device); + tasklet_hi_schedule(&device->tasklet); +} +EXPORT_SYMBOL(dasd_schedule_device_bh); + +void dasd_device_set_stop_bits(struct dasd_device *device, int bits) +{ + device->stopped |= bits; +} +EXPORT_SYMBOL_GPL(dasd_device_set_stop_bits); + +void dasd_device_remove_stop_bits(struct dasd_device *device, int bits) +{ + device->stopped &= ~bits; + if (!device->stopped) + wake_up(&generic_waitq); +} +EXPORT_SYMBOL_GPL(dasd_device_remove_stop_bits); + +/* + * Queue a request to the head of the device ccw_queue. + * Start the I/O if possible. + */ +void dasd_add_request_head(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + unsigned long flags; + + device = cqr->startdev; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + cqr->status = DASD_CQR_QUEUED; + list_add(&cqr->devlist, &device->ccw_queue); + /* let the bh start the request to keep them in order */ + dasd_schedule_device_bh(device); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); +} +EXPORT_SYMBOL(dasd_add_request_head); + +/* + * Queue a request to the tail of the device ccw_queue. + * Start the I/O if possible. + */ +void dasd_add_request_tail(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + unsigned long flags; + + device = cqr->startdev; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + cqr->status = DASD_CQR_QUEUED; + list_add_tail(&cqr->devlist, &device->ccw_queue); + /* let the bh start the request to keep them in order */ + dasd_schedule_device_bh(device); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); +} +EXPORT_SYMBOL(dasd_add_request_tail); + +/* + * Wakeup helper for the 'sleep_on' functions. + */ +void dasd_wakeup_cb(struct dasd_ccw_req *cqr, void *data) +{ + spin_lock_irq(get_ccwdev_lock(cqr->startdev->cdev)); + cqr->callback_data = DASD_SLEEPON_END_TAG; + spin_unlock_irq(get_ccwdev_lock(cqr->startdev->cdev)); + wake_up(&generic_waitq); +} +EXPORT_SYMBOL_GPL(dasd_wakeup_cb); + +static inline int _wait_for_wakeup(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + int rc; + + device = cqr->startdev; + spin_lock_irq(get_ccwdev_lock(device->cdev)); + rc = (cqr->callback_data == DASD_SLEEPON_END_TAG); + spin_unlock_irq(get_ccwdev_lock(device->cdev)); + return rc; +} + +/* + * checks if error recovery is necessary, returns 1 if yes, 0 otherwise. + */ +static int __dasd_sleep_on_erp(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + dasd_erp_fn_t erp_fn; + + if (cqr->status == DASD_CQR_FILLED) + return 0; + device = cqr->startdev; + if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) { + if (cqr->status == DASD_CQR_TERMINATED) { + device->discipline->handle_terminated_request(cqr); + return 1; + } + if (cqr->status == DASD_CQR_NEED_ERP) { + erp_fn = device->discipline->erp_action(cqr); + erp_fn(cqr); + return 1; + } + if (cqr->status == DASD_CQR_FAILED) + dasd_log_sense(cqr, &cqr->irb); + if (cqr->refers) { + __dasd_process_erp(device, cqr); + return 1; + } + } + return 0; +} + +static int __dasd_sleep_on_loop_condition(struct dasd_ccw_req *cqr) +{ + if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) { + if (cqr->refers) /* erp is not done yet */ + return 1; + return ((cqr->status != DASD_CQR_DONE) && + (cqr->status != DASD_CQR_FAILED)); + } else + return (cqr->status == DASD_CQR_FILLED); +} + +static int _dasd_sleep_on(struct dasd_ccw_req *maincqr, int interruptible) +{ + struct dasd_device *device; + int rc; + struct list_head ccw_queue; + struct dasd_ccw_req *cqr; + + INIT_LIST_HEAD(&ccw_queue); + maincqr->status = DASD_CQR_FILLED; + device = maincqr->startdev; + list_add(&maincqr->blocklist, &ccw_queue); + for (cqr = maincqr; __dasd_sleep_on_loop_condition(cqr); + cqr = list_first_entry(&ccw_queue, + struct dasd_ccw_req, blocklist)) { + + if (__dasd_sleep_on_erp(cqr)) + continue; + if (cqr->status != DASD_CQR_FILLED) /* could be failed */ + continue; + if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) && + !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -EPERM; + continue; + } + /* Non-temporary stop condition will trigger fail fast */ + if (device->stopped & ~DASD_STOPPED_PENDING && + test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) && + (!dasd_eer_enabled(device))) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -ENOLINK; + continue; + } + /* + * Don't try to start requests if device is stopped + * except path verification requests + */ + if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) { + if (interruptible) { + rc = wait_event_interruptible( + generic_waitq, !(device->stopped)); + if (rc == -ERESTARTSYS) { + cqr->status = DASD_CQR_FAILED; + maincqr->intrc = rc; + continue; + } + } else + wait_event(generic_waitq, !(device->stopped)); + } + if (!cqr->callback) + cqr->callback = dasd_wakeup_cb; + + cqr->callback_data = DASD_SLEEPON_START_TAG; + dasd_add_request_tail(cqr); + if (interruptible) { + rc = wait_event_interruptible( + generic_waitq, _wait_for_wakeup(cqr)); + if (rc == -ERESTARTSYS) { + dasd_cancel_req(cqr); + /* wait (non-interruptible) for final status */ + wait_event(generic_waitq, + _wait_for_wakeup(cqr)); + cqr->status = DASD_CQR_FAILED; + maincqr->intrc = rc; + continue; + } + } else + wait_event(generic_waitq, _wait_for_wakeup(cqr)); + } + + maincqr->endclk = get_tod_clock(); + if ((maincqr->status != DASD_CQR_DONE) && + (maincqr->intrc != -ERESTARTSYS)) + dasd_log_sense(maincqr, &maincqr->irb); + if (maincqr->status == DASD_CQR_DONE) + rc = 0; + else if (maincqr->intrc) + rc = maincqr->intrc; + else + rc = -EIO; + return rc; +} + +static inline int _wait_for_wakeup_queue(struct list_head *ccw_queue) +{ + struct dasd_ccw_req *cqr; + + list_for_each_entry(cqr, ccw_queue, blocklist) { + if (cqr->callback_data != DASD_SLEEPON_END_TAG) + return 0; + } + + return 1; +} + +static int _dasd_sleep_on_queue(struct list_head *ccw_queue, int interruptible) +{ + struct dasd_device *device; + struct dasd_ccw_req *cqr, *n; + int rc; + +retry: + list_for_each_entry_safe(cqr, n, ccw_queue, blocklist) { + device = cqr->startdev; + if (cqr->status != DASD_CQR_FILLED) /*could be failed*/ + continue; + + if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) && + !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -EPERM; + continue; + } + /*Non-temporary stop condition will trigger fail fast*/ + if (device->stopped & ~DASD_STOPPED_PENDING && + test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) && + !dasd_eer_enabled(device)) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -EAGAIN; + continue; + } + + /*Don't try to start requests if device is stopped*/ + if (interruptible) { + rc = wait_event_interruptible( + generic_waitq, !device->stopped); + if (rc == -ERESTARTSYS) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = rc; + continue; + } + } else + wait_event(generic_waitq, !(device->stopped)); + + if (!cqr->callback) + cqr->callback = dasd_wakeup_cb; + cqr->callback_data = DASD_SLEEPON_START_TAG; + dasd_add_request_tail(cqr); + } + + wait_event(generic_waitq, _wait_for_wakeup_queue(ccw_queue)); + + rc = 0; + list_for_each_entry_safe(cqr, n, ccw_queue, blocklist) { + /* + * for alias devices simplify error recovery and + * return to upper layer + * do not skip ERP requests + */ + if (cqr->startdev != cqr->basedev && !cqr->refers && + (cqr->status == DASD_CQR_TERMINATED || + cqr->status == DASD_CQR_NEED_ERP)) + return -EAGAIN; + + /* normal recovery for basedev IO */ + if (__dasd_sleep_on_erp(cqr)) + /* handle erp first */ + goto retry; + } + + return 0; +} + +/* + * Queue a request to the tail of the device ccw_queue and wait for + * it's completion. + */ +int dasd_sleep_on(struct dasd_ccw_req *cqr) +{ + return _dasd_sleep_on(cqr, 0); +} +EXPORT_SYMBOL(dasd_sleep_on); + +/* + * Start requests from a ccw_queue and wait for their completion. + */ +int dasd_sleep_on_queue(struct list_head *ccw_queue) +{ + return _dasd_sleep_on_queue(ccw_queue, 0); +} +EXPORT_SYMBOL(dasd_sleep_on_queue); + +/* + * Queue a request to the tail of the device ccw_queue and wait + * interruptible for it's completion. + */ +int dasd_sleep_on_interruptible(struct dasd_ccw_req *cqr) +{ + return _dasd_sleep_on(cqr, 1); +} +EXPORT_SYMBOL(dasd_sleep_on_interruptible); + +/* + * Whoa nelly now it gets really hairy. For some functions (e.g. steal lock + * for eckd devices) the currently running request has to be terminated + * and be put back to status queued, before the special request is added + * to the head of the queue. Then the special request is waited on normally. + */ +static inline int _dasd_term_running_cqr(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + int rc; + + if (list_empty(&device->ccw_queue)) + return 0; + cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist); + rc = device->discipline->term_IO(cqr); + if (!rc) + /* + * CQR terminated because a more important request is pending. + * Undo decreasing of retry counter because this is + * not an error case. + */ + cqr->retries++; + return rc; +} + +int dasd_sleep_on_immediatly(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + int rc; + + device = cqr->startdev; + if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) && + !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -EPERM; + return -EIO; + } + spin_lock_irq(get_ccwdev_lock(device->cdev)); + rc = _dasd_term_running_cqr(device); + if (rc) { + spin_unlock_irq(get_ccwdev_lock(device->cdev)); + return rc; + } + cqr->callback = dasd_wakeup_cb; + cqr->callback_data = DASD_SLEEPON_START_TAG; + cqr->status = DASD_CQR_QUEUED; + /* + * add new request as second + * first the terminated cqr needs to be finished + */ + list_add(&cqr->devlist, device->ccw_queue.next); + + /* let the bh start the request to keep them in order */ + dasd_schedule_device_bh(device); + + spin_unlock_irq(get_ccwdev_lock(device->cdev)); + + wait_event(generic_waitq, _wait_for_wakeup(cqr)); + + if (cqr->status == DASD_CQR_DONE) + rc = 0; + else if (cqr->intrc) + rc = cqr->intrc; + else + rc = -EIO; + + /* kick tasklets */ + dasd_schedule_device_bh(device); + if (device->block) + dasd_schedule_block_bh(device->block); + + return rc; +} +EXPORT_SYMBOL(dasd_sleep_on_immediatly); + +/* + * Cancels a request that was started with dasd_sleep_on_req. + * This is useful to timeout requests. The request will be + * terminated if it is currently in i/o. + * Returns 0 if request termination was successful + * negative error code if termination failed + * Cancellation of a request is an asynchronous operation! The calling + * function has to wait until the request is properly returned via callback. + */ +int dasd_cancel_req(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device = cqr->startdev; + unsigned long flags; + int rc; + + rc = 0; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + switch (cqr->status) { + case DASD_CQR_QUEUED: + /* request was not started - just set to cleared */ + cqr->status = DASD_CQR_CLEARED; + if (cqr->callback_data == DASD_SLEEPON_START_TAG) + cqr->callback_data = DASD_SLEEPON_END_TAG; + break; + case DASD_CQR_IN_IO: + /* request in IO - terminate IO and release again */ + rc = device->discipline->term_IO(cqr); + if (rc) { + dev_err(&device->cdev->dev, + "Cancelling request %p failed with rc=%d\n", + cqr, rc); + } else { + cqr->stopclk = get_tod_clock(); + } + break; + default: /* already finished or clear pending - do nothing */ + break; + } + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + dasd_schedule_device_bh(device); + return rc; +} +EXPORT_SYMBOL(dasd_cancel_req); + +/* + * SECTION: Operations of the dasd_block layer. + */ + +/* + * Timeout function for dasd_block. This is used when the block layer + * is waiting for something that may not come reliably, (e.g. a state + * change interrupt) + */ +static void dasd_block_timeout(unsigned long ptr) +{ + unsigned long flags; + struct dasd_block *block; + + block = (struct dasd_block *) ptr; + spin_lock_irqsave(get_ccwdev_lock(block->base->cdev), flags); + /* re-activate request queue */ + dasd_device_remove_stop_bits(block->base, DASD_STOPPED_PENDING); + spin_unlock_irqrestore(get_ccwdev_lock(block->base->cdev), flags); + dasd_schedule_block_bh(block); +} + +/* + * Setup timeout for a dasd_block in jiffies. + */ +void dasd_block_set_timer(struct dasd_block *block, int expires) +{ + if (expires == 0) + del_timer(&block->timer); + else + mod_timer(&block->timer, jiffies + expires); +} +EXPORT_SYMBOL(dasd_block_set_timer); + +/* + * Clear timeout for a dasd_block. + */ +void dasd_block_clear_timer(struct dasd_block *block) +{ + del_timer(&block->timer); +} +EXPORT_SYMBOL(dasd_block_clear_timer); + +/* + * Process finished error recovery ccw. + */ +static void __dasd_process_erp(struct dasd_device *device, + struct dasd_ccw_req *cqr) +{ + dasd_erp_fn_t erp_fn; + + if (cqr->status == DASD_CQR_DONE) + DBF_DEV_EVENT(DBF_NOTICE, device, "%s", "ERP successful"); + else + dev_err(&device->cdev->dev, "ERP failed for the DASD\n"); + erp_fn = device->discipline->erp_postaction(cqr); + erp_fn(cqr); +} + +/* + * Fetch requests from the block device queue. + */ +static void __dasd_process_request_queue(struct dasd_block *block) +{ + struct request_queue *queue; + struct request *req; + struct dasd_ccw_req *cqr; + struct dasd_device *basedev; + unsigned long flags; + queue = block->request_queue; + basedev = block->base; + /* No queue ? Then there is nothing to do. */ + if (queue == NULL) + return; + + /* + * We requeue request from the block device queue to the ccw + * queue only in two states. In state DASD_STATE_READY the + * partition detection is done and we need to requeue requests + * for that. State DASD_STATE_ONLINE is normal block device + * operation. + */ + if (basedev->state < DASD_STATE_READY) { + while ((req = blk_fetch_request(block->request_queue))) + __blk_end_request_all(req, -EIO); + return; + } + + /* if device ist stopped do not fetch new requests */ + if (basedev->stopped) + return; + + /* Now we try to fetch requests from the request queue */ + while ((req = blk_peek_request(queue))) { + if (basedev->features & DASD_FEATURE_READONLY && + rq_data_dir(req) == WRITE) { + DBF_DEV_EVENT(DBF_ERR, basedev, + "Rejecting write request %p", + req); + blk_start_request(req); + __blk_end_request_all(req, -EIO); + continue; + } + if (test_bit(DASD_FLAG_ABORTALL, &basedev->flags) && + (basedev->features & DASD_FEATURE_FAILFAST || + blk_noretry_request(req))) { + DBF_DEV_EVENT(DBF_ERR, basedev, + "Rejecting failfast request %p", + req); + blk_start_request(req); + __blk_end_request_all(req, -ETIMEDOUT); + continue; + } + cqr = basedev->discipline->build_cp(basedev, block, req); + if (IS_ERR(cqr)) { + if (PTR_ERR(cqr) == -EBUSY) + break; /* normal end condition */ + if (PTR_ERR(cqr) == -ENOMEM) + break; /* terminate request queue loop */ + if (PTR_ERR(cqr) == -EAGAIN) { + /* + * The current request cannot be build right + * now, we have to try later. If this request + * is the head-of-queue we stop the device + * for 1/2 second. + */ + if (!list_empty(&block->ccw_queue)) + break; + spin_lock_irqsave( + get_ccwdev_lock(basedev->cdev), flags); + dasd_device_set_stop_bits(basedev, + DASD_STOPPED_PENDING); + spin_unlock_irqrestore( + get_ccwdev_lock(basedev->cdev), flags); + dasd_block_set_timer(block, HZ/2); + break; + } + DBF_DEV_EVENT(DBF_ERR, basedev, + "CCW creation failed (rc=%ld) " + "on request %p", + PTR_ERR(cqr), req); + blk_start_request(req); + __blk_end_request_all(req, -EIO); + continue; + } + /* + * Note: callback is set to dasd_return_cqr_cb in + * __dasd_block_start_head to cover erp requests as well + */ + cqr->callback_data = (void *) req; + cqr->status = DASD_CQR_FILLED; + req->completion_data = cqr; + blk_start_request(req); + list_add_tail(&cqr->blocklist, &block->ccw_queue); + INIT_LIST_HEAD(&cqr->devlist); + dasd_profile_start(block, cqr, req); + } +} + +static void __dasd_cleanup_cqr(struct dasd_ccw_req *cqr) +{ + struct request *req; + int status; + int error = 0; + + req = (struct request *) cqr->callback_data; + dasd_profile_end(cqr->block, cqr, req); + status = cqr->block->base->discipline->free_cp(cqr, req); + if (status < 0) + error = status; + else if (status == 0) { + if (cqr->intrc == -EPERM) + error = -EBADE; + else if (cqr->intrc == -ENOLINK || + cqr->intrc == -ETIMEDOUT) + error = cqr->intrc; + else + error = -EIO; + } + __blk_end_request_all(req, error); +} + +/* + * Process ccw request queue. + */ +static void __dasd_process_block_ccw_queue(struct dasd_block *block, + struct list_head *final_queue) +{ + struct list_head *l, *n; + struct dasd_ccw_req *cqr; + dasd_erp_fn_t erp_fn; + unsigned long flags; + struct dasd_device *base = block->base; + +restart: + /* Process request with final status. */ + list_for_each_safe(l, n, &block->ccw_queue) { + cqr = list_entry(l, struct dasd_ccw_req, blocklist); + if (cqr->status != DASD_CQR_DONE && + cqr->status != DASD_CQR_FAILED && + cqr->status != DASD_CQR_NEED_ERP && + cqr->status != DASD_CQR_TERMINATED) + continue; + + if (cqr->status == DASD_CQR_TERMINATED) { + base->discipline->handle_terminated_request(cqr); + goto restart; + } + + /* Process requests that may be recovered */ + if (cqr->status == DASD_CQR_NEED_ERP) { + erp_fn = base->discipline->erp_action(cqr); + if (IS_ERR(erp_fn(cqr))) + continue; + goto restart; + } + + /* log sense for fatal error */ + if (cqr->status == DASD_CQR_FAILED) { + dasd_log_sense(cqr, &cqr->irb); + } + + /* First of all call extended error reporting. */ + if (dasd_eer_enabled(base) && + cqr->status == DASD_CQR_FAILED) { + dasd_eer_write(base, cqr, DASD_EER_FATALERROR); + + /* restart request */ + cqr->status = DASD_CQR_FILLED; + cqr->retries = 255; + spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags); + dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE); + spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), + flags); + goto restart; + } + + /* Process finished ERP request. */ + if (cqr->refers) { + __dasd_process_erp(base, cqr); + goto restart; + } + + /* Rechain finished requests to final queue */ + cqr->endclk = get_tod_clock(); + list_move_tail(&cqr->blocklist, final_queue); + } +} + +static void dasd_return_cqr_cb(struct dasd_ccw_req *cqr, void *data) +{ + dasd_schedule_block_bh(cqr->block); +} + +static void __dasd_block_start_head(struct dasd_block *block) +{ + struct dasd_ccw_req *cqr; + + if (list_empty(&block->ccw_queue)) + return; + /* We allways begin with the first requests on the queue, as some + * of previously started requests have to be enqueued on a + * dasd_device again for error recovery. + */ + list_for_each_entry(cqr, &block->ccw_queue, blocklist) { + if (cqr->status != DASD_CQR_FILLED) + continue; + if (test_bit(DASD_FLAG_LOCK_STOLEN, &block->base->flags) && + !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -EPERM; + dasd_schedule_block_bh(block); + continue; + } + /* Non-temporary stop condition will trigger fail fast */ + if (block->base->stopped & ~DASD_STOPPED_PENDING && + test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) && + (!dasd_eer_enabled(block->base))) { + cqr->status = DASD_CQR_FAILED; + cqr->intrc = -ENOLINK; + dasd_schedule_block_bh(block); + continue; + } + /* Don't try to start requests if device is stopped */ + if (block->base->stopped) + return; + + /* just a fail safe check, should not happen */ + if (!cqr->startdev) + cqr->startdev = block->base; + + /* make sure that the requests we submit find their way back */ + cqr->callback = dasd_return_cqr_cb; + + dasd_add_request_tail(cqr); + } +} + +/* + * Central dasd_block layer routine. Takes requests from the generic + * block layer request queue, creates ccw requests, enqueues them on + * a dasd_device and processes ccw requests that have been returned. + */ +static void dasd_block_tasklet(struct dasd_block *block) +{ + struct list_head final_queue; + struct list_head *l, *n; + struct dasd_ccw_req *cqr; + + atomic_set(&block->tasklet_scheduled, 0); + INIT_LIST_HEAD(&final_queue); + spin_lock(&block->queue_lock); + /* Finish off requests on ccw queue */ + __dasd_process_block_ccw_queue(block, &final_queue); + spin_unlock(&block->queue_lock); + /* Now call the callback function of requests with final status */ + spin_lock_irq(&block->request_queue_lock); + list_for_each_safe(l, n, &final_queue) { + cqr = list_entry(l, struct dasd_ccw_req, blocklist); + list_del_init(&cqr->blocklist); + __dasd_cleanup_cqr(cqr); + } + spin_lock(&block->queue_lock); + /* Get new request from the block device request queue */ + __dasd_process_request_queue(block); + /* Now check if the head of the ccw queue needs to be started. */ + __dasd_block_start_head(block); + spin_unlock(&block->queue_lock); + spin_unlock_irq(&block->request_queue_lock); + if (waitqueue_active(&shutdown_waitq)) + wake_up(&shutdown_waitq); + dasd_put_device(block->base); +} + +static void _dasd_wake_block_flush_cb(struct dasd_ccw_req *cqr, void *data) +{ + wake_up(&dasd_flush_wq); +} + +/* + * Requeue a request back to the block request queue + * only works for block requests + */ +static int _dasd_requeue_request(struct dasd_ccw_req *cqr) +{ + struct dasd_block *block = cqr->block; + struct request *req; + unsigned long flags; + + if (!block) + return -EINVAL; + spin_lock_irqsave(&block->queue_lock, flags); + req = (struct request *) cqr->callback_data; + blk_requeue_request(block->request_queue, req); + spin_unlock_irqrestore(&block->queue_lock, flags); + + return 0; +} + +/* + * Go through all request on the dasd_block request queue, cancel them + * on the respective dasd_device, and return them to the generic + * block layer. + */ +static int dasd_flush_block_queue(struct dasd_block *block) +{ + struct dasd_ccw_req *cqr, *n; + int rc, i; + struct list_head flush_queue; + + INIT_LIST_HEAD(&flush_queue); + spin_lock_bh(&block->queue_lock); + rc = 0; +restart: + list_for_each_entry_safe(cqr, n, &block->ccw_queue, blocklist) { + /* if this request currently owned by a dasd_device cancel it */ + if (cqr->status >= DASD_CQR_QUEUED) + rc = dasd_cancel_req(cqr); + if (rc < 0) + break; + /* Rechain request (including erp chain) so it won't be + * touched by the dasd_block_tasklet anymore. + * Replace the callback so we notice when the request + * is returned from the dasd_device layer. + */ + cqr->callback = _dasd_wake_block_flush_cb; + for (i = 0; cqr != NULL; cqr = cqr->refers, i++) + list_move_tail(&cqr->blocklist, &flush_queue); + if (i > 1) + /* moved more than one request - need to restart */ + goto restart; + } + spin_unlock_bh(&block->queue_lock); + /* Now call the callback function of flushed requests */ +restart_cb: + list_for_each_entry_safe(cqr, n, &flush_queue, blocklist) { + wait_event(dasd_flush_wq, (cqr->status < DASD_CQR_QUEUED)); + /* Process finished ERP request. */ + if (cqr->refers) { + spin_lock_bh(&block->queue_lock); + __dasd_process_erp(block->base, cqr); + spin_unlock_bh(&block->queue_lock); + /* restart list_for_xx loop since dasd_process_erp + * might remove multiple elements */ + goto restart_cb; + } + /* call the callback function */ + spin_lock_irq(&block->request_queue_lock); + cqr->endclk = get_tod_clock(); + list_del_init(&cqr->blocklist); + __dasd_cleanup_cqr(cqr); + spin_unlock_irq(&block->request_queue_lock); + } + return rc; +} + +/* + * Schedules a call to dasd_tasklet over the device tasklet. + */ +void dasd_schedule_block_bh(struct dasd_block *block) +{ + /* Protect against rescheduling. */ + if (atomic_cmpxchg(&block->tasklet_scheduled, 0, 1) != 0) + return; + /* life cycle of block is bound to it's base device */ + dasd_get_device(block->base); + tasklet_hi_schedule(&block->tasklet); +} +EXPORT_SYMBOL(dasd_schedule_block_bh); + + +/* + * SECTION: external block device operations + * (request queue handling, open, release, etc.) + */ + +/* + * Dasd request queue function. Called from ll_rw_blk.c + */ +static void do_dasd_request(struct request_queue *queue) +{ + struct dasd_block *block; + + block = queue->queuedata; + spin_lock(&block->queue_lock); + /* Get new request from the block device request queue */ + __dasd_process_request_queue(block); + /* Now check if the head of the ccw queue needs to be started. */ + __dasd_block_start_head(block); + spin_unlock(&block->queue_lock); +} + +/* + * Block timeout callback, called from the block layer + * + * request_queue lock is held on entry. + * + * Return values: + * BLK_EH_RESET_TIMER if the request should be left running + * BLK_EH_NOT_HANDLED if the request is handled or terminated + * by the driver. + */ +enum blk_eh_timer_return dasd_times_out(struct request *req) +{ + struct dasd_ccw_req *cqr = req->completion_data; + struct dasd_block *block = req->q->queuedata; + struct dasd_device *device; + int rc = 0; + + if (!cqr) + return BLK_EH_NOT_HANDLED; + + device = cqr->startdev ? cqr->startdev : block->base; + if (!device->blk_timeout) + return BLK_EH_RESET_TIMER; + DBF_DEV_EVENT(DBF_WARNING, device, + " dasd_times_out cqr %p status %x", + cqr, cqr->status); + + spin_lock(&block->queue_lock); + spin_lock(get_ccwdev_lock(device->cdev)); + cqr->retries = -1; + cqr->intrc = -ETIMEDOUT; + if (cqr->status >= DASD_CQR_QUEUED) { + spin_unlock(get_ccwdev_lock(device->cdev)); + rc = dasd_cancel_req(cqr); + } else if (cqr->status == DASD_CQR_FILLED || + cqr->status == DASD_CQR_NEED_ERP) { + cqr->status = DASD_CQR_TERMINATED; + spin_unlock(get_ccwdev_lock(device->cdev)); + } else if (cqr->status == DASD_CQR_IN_ERP) { + struct dasd_ccw_req *searchcqr, *nextcqr, *tmpcqr; + + list_for_each_entry_safe(searchcqr, nextcqr, + &block->ccw_queue, blocklist) { + tmpcqr = searchcqr; + while (tmpcqr->refers) + tmpcqr = tmpcqr->refers; + if (tmpcqr != cqr) + continue; + /* searchcqr is an ERP request for cqr */ + searchcqr->retries = -1; + searchcqr->intrc = -ETIMEDOUT; + if (searchcqr->status >= DASD_CQR_QUEUED) { + spin_unlock(get_ccwdev_lock(device->cdev)); + rc = dasd_cancel_req(searchcqr); + spin_lock(get_ccwdev_lock(device->cdev)); + } else if ((searchcqr->status == DASD_CQR_FILLED) || + (searchcqr->status == DASD_CQR_NEED_ERP)) { + searchcqr->status = DASD_CQR_TERMINATED; + rc = 0; + } else if (searchcqr->status == DASD_CQR_IN_ERP) { + /* + * Shouldn't happen; most recent ERP + * request is at the front of queue + */ + continue; + } + break; + } + spin_unlock(get_ccwdev_lock(device->cdev)); + } + dasd_schedule_block_bh(block); + spin_unlock(&block->queue_lock); + + return rc ? BLK_EH_RESET_TIMER : BLK_EH_NOT_HANDLED; +} + +/* + * Allocate and initialize request queue and default I/O scheduler. + */ +static int dasd_alloc_queue(struct dasd_block *block) +{ + block->request_queue = blk_init_queue(do_dasd_request, + &block->request_queue_lock); + if (block->request_queue == NULL) + return -ENOMEM; + + block->request_queue->queuedata = block; + + return 0; +} + +/* + * Allocate and initialize request queue. + */ +static void dasd_setup_queue(struct dasd_block *block) +{ + int max; + + if (block->base->features & DASD_FEATURE_USERAW) { + /* + * the max_blocks value for raw_track access is 256 + * it is higher than the native ECKD value because we + * only need one ccw per track + * so the max_hw_sectors are + * 2048 x 512B = 1024kB = 16 tracks + */ + max = 2048; + } else { + max = block->base->discipline->max_blocks << block->s2b_shift; + } + blk_queue_logical_block_size(block->request_queue, + block->bp_block); + blk_queue_max_hw_sectors(block->request_queue, max); + blk_queue_max_segments(block->request_queue, -1L); + /* with page sized segments we can translate each segement into + * one idaw/tidaw + */ + blk_queue_max_segment_size(block->request_queue, PAGE_SIZE); + blk_queue_segment_boundary(block->request_queue, PAGE_SIZE - 1); +} + +/* + * Deactivate and free request queue. + */ +static void dasd_free_queue(struct dasd_block *block) +{ + if (block->request_queue) { + blk_cleanup_queue(block->request_queue); + block->request_queue = NULL; + } +} + +/* + * Flush request on the request queue. + */ +static void dasd_flush_request_queue(struct dasd_block *block) +{ + struct request *req; + + if (!block->request_queue) + return; + + spin_lock_irq(&block->request_queue_lock); + while ((req = blk_fetch_request(block->request_queue))) + __blk_end_request_all(req, -EIO); + spin_unlock_irq(&block->request_queue_lock); +} + +static int dasd_open(struct block_device *bdev, fmode_t mode) +{ + struct dasd_device *base; + int rc; + + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + + atomic_inc(&base->block->open_count); + if (test_bit(DASD_FLAG_OFFLINE, &base->flags)) { + rc = -ENODEV; + goto unlock; + } + + if (!try_module_get(base->discipline->owner)) { + rc = -EINVAL; + goto unlock; + } + + if (dasd_probeonly) { + dev_info(&base->cdev->dev, + "Accessing the DASD failed because it is in " + "probeonly mode\n"); + rc = -EPERM; + goto out; + } + + if (base->state <= DASD_STATE_BASIC) { + DBF_DEV_EVENT(DBF_ERR, base, " %s", + " Cannot open unrecognized device"); + rc = -ENODEV; + goto out; + } + + if ((mode & FMODE_WRITE) && + (test_bit(DASD_FLAG_DEVICE_RO, &base->flags) || + (base->features & DASD_FEATURE_READONLY))) { + rc = -EROFS; + goto out; + } + + dasd_put_device(base); + return 0; + +out: + module_put(base->discipline->owner); +unlock: + atomic_dec(&base->block->open_count); + dasd_put_device(base); + return rc; +} + +static void dasd_release(struct gendisk *disk, fmode_t mode) +{ + struct dasd_device *base = dasd_device_from_gendisk(disk); + if (base) { + atomic_dec(&base->block->open_count); + module_put(base->discipline->owner); + dasd_put_device(base); + } +} + +/* + * Return disk geometry. + */ +static int dasd_getgeo(struct block_device *bdev, struct hd_geometry *geo) +{ + struct dasd_device *base; + + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + + if (!base->discipline || + !base->discipline->fill_geometry) { + dasd_put_device(base); + return -EINVAL; + } + base->discipline->fill_geometry(base->block, geo); + geo->start = get_start_sect(bdev) >> base->block->s2b_shift; + dasd_put_device(base); + return 0; +} + +const struct block_device_operations +dasd_device_operations = { + .owner = THIS_MODULE, + .open = dasd_open, + .release = dasd_release, + .ioctl = dasd_ioctl, + .compat_ioctl = dasd_ioctl, + .getgeo = dasd_getgeo, +}; + +/******************************************************************************* + * end of block device operations + */ + +static void +dasd_exit(void) +{ +#ifdef CONFIG_PROC_FS + dasd_proc_exit(); +#endif + dasd_eer_exit(); + if (dasd_page_cache != NULL) { + kmem_cache_destroy(dasd_page_cache); + dasd_page_cache = NULL; + } + dasd_gendisk_exit(); + dasd_devmap_exit(); + if (dasd_debug_area != NULL) { + debug_unregister(dasd_debug_area); + dasd_debug_area = NULL; + } + dasd_statistics_removeroot(); +} + +/* + * SECTION: common functions for ccw_driver use + */ + +/* + * Is the device read-only? + * Note that this function does not report the setting of the + * readonly device attribute, but how it is configured in z/VM. + */ +int dasd_device_is_ro(struct dasd_device *device) +{ + struct ccw_dev_id dev_id; + struct diag210 diag_data; + int rc; + + if (!MACHINE_IS_VM) + return 0; + ccw_device_get_id(device->cdev, &dev_id); + memset(&diag_data, 0, sizeof(diag_data)); + diag_data.vrdcdvno = dev_id.devno; + diag_data.vrdclen = sizeof(diag_data); + rc = diag210(&diag_data); + if (rc == 0 || rc == 2) { + return diag_data.vrdcvfla & 0x80; + } else { + DBF_EVENT(DBF_WARNING, "diag210 failed for dev=%04x with rc=%d", + dev_id.devno, rc); + return 0; + } +} +EXPORT_SYMBOL_GPL(dasd_device_is_ro); + +static void dasd_generic_auto_online(void *data, async_cookie_t cookie) +{ + struct ccw_device *cdev = data; + int ret; + + ret = ccw_device_set_online(cdev); + if (ret) + pr_warn("%s: Setting the DASD online failed with rc=%d\n", + dev_name(&cdev->dev), ret); +} + +/* + * Initial attempt at a probe function. this can be simplified once + * the other detection code is gone. + */ +int dasd_generic_probe(struct ccw_device *cdev, + struct dasd_discipline *discipline) +{ + int ret; + + ret = dasd_add_sysfs_files(cdev); + if (ret) { + DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s", + "dasd_generic_probe: could not add " + "sysfs entries"); + return ret; + } + cdev->handler = &dasd_int_handler; + + /* + * Automatically online either all dasd devices (dasd_autodetect) + * or all devices specified with dasd= parameters during + * initial probe. + */ + if ((dasd_get_feature(cdev, DASD_FEATURE_INITIAL_ONLINE) > 0 ) || + (dasd_autodetect && dasd_busid_known(dev_name(&cdev->dev)) != 0)) + async_schedule(dasd_generic_auto_online, cdev); + return 0; +} +EXPORT_SYMBOL_GPL(dasd_generic_probe); + +/* + * This will one day be called from a global not_oper handler. + * It is also used by driver_unregister during module unload. + */ +void dasd_generic_remove(struct ccw_device *cdev) +{ + struct dasd_device *device; + struct dasd_block *block; + + cdev->handler = NULL; + + device = dasd_device_from_cdev(cdev); + if (IS_ERR(device)) { + dasd_remove_sysfs_files(cdev); + return; + } + if (test_and_set_bit(DASD_FLAG_OFFLINE, &device->flags) && + !test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) { + /* Already doing offline processing */ + dasd_put_device(device); + dasd_remove_sysfs_files(cdev); + return; + } + /* + * This device is removed unconditionally. Set offline + * flag to prevent dasd_open from opening it while it is + * no quite down yet. + */ + dasd_set_target_state(device, DASD_STATE_NEW); + /* dasd_delete_device destroys the device reference. */ + block = device->block; + dasd_delete_device(device); + /* + * life cycle of block is bound to device, so delete it after + * device was safely removed + */ + if (block) + dasd_free_block(block); + + dasd_remove_sysfs_files(cdev); +} +EXPORT_SYMBOL_GPL(dasd_generic_remove); + +/* + * Activate a device. This is called from dasd_{eckd,fba}_probe() when either + * the device is detected for the first time and is supposed to be used + * or the user has started activation through sysfs. + */ +int dasd_generic_set_online(struct ccw_device *cdev, + struct dasd_discipline *base_discipline) +{ + struct dasd_discipline *discipline; + struct dasd_device *device; + int rc; + + /* first online clears initial online feature flag */ + dasd_set_feature(cdev, DASD_FEATURE_INITIAL_ONLINE, 0); + device = dasd_create_device(cdev); + if (IS_ERR(device)) + return PTR_ERR(device); + + discipline = base_discipline; + if (device->features & DASD_FEATURE_USEDIAG) { + if (!dasd_diag_discipline_pointer) { + pr_warn("%s Setting the DASD online failed because of missing DIAG discipline\n", + dev_name(&cdev->dev)); + dasd_delete_device(device); + return -ENODEV; + } + discipline = dasd_diag_discipline_pointer; + } + if (!try_module_get(base_discipline->owner)) { + dasd_delete_device(device); + return -EINVAL; + } + if (!try_module_get(discipline->owner)) { + module_put(base_discipline->owner); + dasd_delete_device(device); + return -EINVAL; + } + device->base_discipline = base_discipline; + device->discipline = discipline; + + /* check_device will allocate block device if necessary */ + rc = discipline->check_device(device); + if (rc) { + pr_warn("%s Setting the DASD online with discipline %s failed with rc=%i\n", + dev_name(&cdev->dev), discipline->name, rc); + module_put(discipline->owner); + module_put(base_discipline->owner); + dasd_delete_device(device); + return rc; + } + + dasd_set_target_state(device, DASD_STATE_ONLINE); + if (device->state <= DASD_STATE_KNOWN) { + pr_warn("%s Setting the DASD online failed because of a missing discipline\n", + dev_name(&cdev->dev)); + rc = -ENODEV; + dasd_set_target_state(device, DASD_STATE_NEW); + if (device->block) + dasd_free_block(device->block); + dasd_delete_device(device); + } else + pr_debug("dasd_generic device %s found\n", + dev_name(&cdev->dev)); + + wait_event(dasd_init_waitq, _wait_for_device(device)); + + dasd_put_device(device); + return rc; +} +EXPORT_SYMBOL_GPL(dasd_generic_set_online); + +int dasd_generic_set_offline(struct ccw_device *cdev) +{ + struct dasd_device *device; + struct dasd_block *block; + int max_count, open_count, rc; + + rc = 0; + device = dasd_device_from_cdev(cdev); + if (IS_ERR(device)) + return PTR_ERR(device); + + /* + * We must make sure that this device is currently not in use. + * The open_count is increased for every opener, that includes + * the blkdev_get in dasd_scan_partitions. We are only interested + * in the other openers. + */ + if (device->block) { + max_count = device->block->bdev ? 0 : -1; + open_count = atomic_read(&device->block->open_count); + if (open_count > max_count) { + if (open_count > 0) + pr_warn("%s: The DASD cannot be set offline with open count %i\n", + dev_name(&cdev->dev), open_count); + else + pr_warn("%s: The DASD cannot be set offline while it is in use\n", + dev_name(&cdev->dev)); + clear_bit(DASD_FLAG_OFFLINE, &device->flags); + dasd_put_device(device); + return -EBUSY; + } + } + + if (test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) { + /* + * safe offline already running + * could only be called by normal offline so safe_offline flag + * needs to be removed to run normal offline and kill all I/O + */ + if (test_and_set_bit(DASD_FLAG_OFFLINE, &device->flags)) { + /* Already doing normal offline processing */ + dasd_put_device(device); + return -EBUSY; + } else + clear_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags); + + } else + if (test_bit(DASD_FLAG_OFFLINE, &device->flags)) { + /* Already doing offline processing */ + dasd_put_device(device); + return -EBUSY; + } + + /* + * if safe_offline called set safe_offline_running flag and + * clear safe_offline so that a call to normal offline + * can overrun safe_offline processing + */ + if (test_and_clear_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags) && + !test_and_set_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) { + /* + * If we want to set the device safe offline all IO operations + * should be finished before continuing the offline process + * so sync bdev first and then wait for our queues to become + * empty + */ + /* sync blockdev and partitions */ + rc = fsync_bdev(device->block->bdev); + if (rc != 0) + goto interrupted; + + /* schedule device tasklet and wait for completion */ + dasd_schedule_device_bh(device); + rc = wait_event_interruptible(shutdown_waitq, + _wait_for_empty_queues(device)); + if (rc != 0) + goto interrupted; + } + + set_bit(DASD_FLAG_OFFLINE, &device->flags); + dasd_set_target_state(device, DASD_STATE_NEW); + /* dasd_delete_device destroys the device reference. */ + block = device->block; + dasd_delete_device(device); + /* + * life cycle of block is bound to device, so delete it after + * device was safely removed + */ + if (block) + dasd_free_block(block); + return 0; + +interrupted: + /* interrupted by signal */ + clear_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags); + clear_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags); + clear_bit(DASD_FLAG_OFFLINE, &device->flags); + dasd_put_device(device); + return rc; +} +EXPORT_SYMBOL_GPL(dasd_generic_set_offline); + +int dasd_generic_last_path_gone(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + + dev_warn(&device->cdev->dev, "No operational channel path is left " + "for the device\n"); + DBF_DEV_EVENT(DBF_WARNING, device, "%s", "last path gone"); + /* First of all call extended error reporting. */ + dasd_eer_write(device, NULL, DASD_EER_NOPATH); + + if (device->state < DASD_STATE_BASIC) + return 0; + /* Device is active. We want to keep it. */ + list_for_each_entry(cqr, &device->ccw_queue, devlist) + if ((cqr->status == DASD_CQR_IN_IO) || + (cqr->status == DASD_CQR_CLEAR_PENDING)) { + cqr->status = DASD_CQR_QUEUED; + cqr->retries++; + } + dasd_device_set_stop_bits(device, DASD_STOPPED_DC_WAIT); + dasd_device_clear_timer(device); + dasd_schedule_device_bh(device); + return 1; +} +EXPORT_SYMBOL_GPL(dasd_generic_last_path_gone); + +int dasd_generic_path_operational(struct dasd_device *device) +{ + dev_info(&device->cdev->dev, "A channel path to the device has become " + "operational\n"); + DBF_DEV_EVENT(DBF_WARNING, device, "%s", "path operational"); + dasd_device_remove_stop_bits(device, DASD_STOPPED_DC_WAIT); + if (device->stopped & DASD_UNRESUMED_PM) { + dasd_device_remove_stop_bits(device, DASD_UNRESUMED_PM); + dasd_restore_device(device); + return 1; + } + dasd_schedule_device_bh(device); + if (device->block) + dasd_schedule_block_bh(device->block); + + if (!device->stopped) + wake_up(&generic_waitq); + + return 1; +} +EXPORT_SYMBOL_GPL(dasd_generic_path_operational); + +int dasd_generic_notify(struct ccw_device *cdev, int event) +{ + struct dasd_device *device; + int ret; + + device = dasd_device_from_cdev_locked(cdev); + if (IS_ERR(device)) + return 0; + ret = 0; + switch (event) { + case CIO_GONE: + case CIO_BOXED: + case CIO_NO_PATH: + device->path_data.opm = 0; + device->path_data.ppm = 0; + device->path_data.npm = 0; + ret = dasd_generic_last_path_gone(device); + break; + case CIO_OPER: + ret = 1; + if (device->path_data.opm) + ret = dasd_generic_path_operational(device); + break; + } + dasd_put_device(device); + return ret; +} +EXPORT_SYMBOL_GPL(dasd_generic_notify); + +void dasd_generic_path_event(struct ccw_device *cdev, int *path_event) +{ + int chp; + __u8 oldopm, eventlpm; + struct dasd_device *device; + + device = dasd_device_from_cdev_locked(cdev); + if (IS_ERR(device)) + return; + for (chp = 0; chp < 8; chp++) { + eventlpm = 0x80 >> chp; + if (path_event[chp] & PE_PATH_GONE) { + oldopm = device->path_data.opm; + device->path_data.opm &= ~eventlpm; + device->path_data.ppm &= ~eventlpm; + device->path_data.npm &= ~eventlpm; + if (oldopm && !device->path_data.opm) { + dev_warn(&device->cdev->dev, + "No verified channel paths remain " + "for the device\n"); + DBF_DEV_EVENT(DBF_WARNING, device, + "%s", "last verified path gone"); + dasd_eer_write(device, NULL, DASD_EER_NOPATH); + dasd_device_set_stop_bits(device, + DASD_STOPPED_DC_WAIT); + } + } + if (path_event[chp] & PE_PATH_AVAILABLE) { + device->path_data.opm &= ~eventlpm; + device->path_data.ppm &= ~eventlpm; + device->path_data.npm &= ~eventlpm; + device->path_data.tbvpm |= eventlpm; + dasd_schedule_device_bh(device); + } + if (path_event[chp] & PE_PATHGROUP_ESTABLISHED) { + if (!(device->path_data.opm & eventlpm) && + !(device->path_data.tbvpm & eventlpm)) { + /* + * we can not establish a pathgroup on an + * unavailable path, so trigger a path + * verification first + */ + device->path_data.tbvpm |= eventlpm; + dasd_schedule_device_bh(device); + } + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Pathgroup re-established\n"); + if (device->discipline->kick_validate) + device->discipline->kick_validate(device); + } + } + dasd_put_device(device); +} +EXPORT_SYMBOL_GPL(dasd_generic_path_event); + +int dasd_generic_verify_path(struct dasd_device *device, __u8 lpm) +{ + if (!device->path_data.opm && lpm) { + device->path_data.opm = lpm; + dasd_generic_path_operational(device); + } else + device->path_data.opm |= lpm; + return 0; +} +EXPORT_SYMBOL_GPL(dasd_generic_verify_path); + + +int dasd_generic_pm_freeze(struct ccw_device *cdev) +{ + struct dasd_device *device = dasd_device_from_cdev(cdev); + struct list_head freeze_queue; + struct dasd_ccw_req *cqr, *n; + struct dasd_ccw_req *refers; + int rc; + + if (IS_ERR(device)) + return PTR_ERR(device); + + /* mark device as suspended */ + set_bit(DASD_FLAG_SUSPENDED, &device->flags); + + if (device->discipline->freeze) + rc = device->discipline->freeze(device); + + /* disallow new I/O */ + dasd_device_set_stop_bits(device, DASD_STOPPED_PM); + + /* clear active requests and requeue them to block layer if possible */ + INIT_LIST_HEAD(&freeze_queue); + spin_lock_irq(get_ccwdev_lock(cdev)); + rc = 0; + list_for_each_entry_safe(cqr, n, &device->ccw_queue, devlist) { + /* Check status and move request to flush_queue */ + if (cqr->status == DASD_CQR_IN_IO) { + rc = device->discipline->term_IO(cqr); + if (rc) { + /* unable to terminate requeust */ + dev_err(&device->cdev->dev, + "Unable to terminate request %p " + "on suspend\n", cqr); + spin_unlock_irq(get_ccwdev_lock(cdev)); + dasd_put_device(device); + return rc; + } + } + list_move_tail(&cqr->devlist, &freeze_queue); + } + spin_unlock_irq(get_ccwdev_lock(cdev)); + + list_for_each_entry_safe(cqr, n, &freeze_queue, devlist) { + wait_event(dasd_flush_wq, + (cqr->status != DASD_CQR_CLEAR_PENDING)); + if (cqr->status == DASD_CQR_CLEARED) + cqr->status = DASD_CQR_QUEUED; + + /* requeue requests to blocklayer will only work for + block device requests */ + if (_dasd_requeue_request(cqr)) + continue; + + /* remove requests from device and block queue */ + list_del_init(&cqr->devlist); + while (cqr->refers != NULL) { + refers = cqr->refers; + /* remove the request from the block queue */ + list_del(&cqr->blocklist); + /* free the finished erp request */ + dasd_free_erp_request(cqr, cqr->memdev); + cqr = refers; + } + if (cqr->block) + list_del_init(&cqr->blocklist); + cqr->block->base->discipline->free_cp( + cqr, (struct request *) cqr->callback_data); + } + + /* + * if requests remain then they are internal request + * and go back to the device queue + */ + if (!list_empty(&freeze_queue)) { + /* move freeze_queue to start of the ccw_queue */ + spin_lock_irq(get_ccwdev_lock(cdev)); + list_splice_tail(&freeze_queue, &device->ccw_queue); + spin_unlock_irq(get_ccwdev_lock(cdev)); + } + dasd_put_device(device); + return rc; +} +EXPORT_SYMBOL_GPL(dasd_generic_pm_freeze); + +int dasd_generic_restore_device(struct ccw_device *cdev) +{ + struct dasd_device *device = dasd_device_from_cdev(cdev); + int rc = 0; + + if (IS_ERR(device)) + return PTR_ERR(device); + + /* allow new IO again */ + dasd_device_remove_stop_bits(device, + (DASD_STOPPED_PM | DASD_UNRESUMED_PM)); + + dasd_schedule_device_bh(device); + + /* + * call discipline restore function + * if device is stopped do nothing e.g. for disconnected devices + */ + if (device->discipline->restore && !(device->stopped)) + rc = device->discipline->restore(device); + if (rc || device->stopped) + /* + * if the resume failed for the DASD we put it in + * an UNRESUMED stop state + */ + device->stopped |= DASD_UNRESUMED_PM; + + if (device->block) + dasd_schedule_block_bh(device->block); + + clear_bit(DASD_FLAG_SUSPENDED, &device->flags); + dasd_put_device(device); + return 0; +} +EXPORT_SYMBOL_GPL(dasd_generic_restore_device); + +static struct dasd_ccw_req *dasd_generic_build_rdc(struct dasd_device *device, + void *rdc_buffer, + int rdc_buffer_size, + int magic) +{ + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + unsigned long *idaw; + + cqr = dasd_smalloc_request(magic, 1 /* RDC */, rdc_buffer_size, device); + + if (IS_ERR(cqr)) { + /* internal error 13 - Allocating the RDC request failed*/ + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", "13"); + return cqr; + } + + ccw = cqr->cpaddr; + ccw->cmd_code = CCW_CMD_RDC; + if (idal_is_needed(rdc_buffer, rdc_buffer_size)) { + idaw = (unsigned long *) (cqr->data); + ccw->cda = (__u32)(addr_t) idaw; + ccw->flags = CCW_FLAG_IDA; + idaw = idal_create_words(idaw, rdc_buffer, rdc_buffer_size); + } else { + ccw->cda = (__u32)(addr_t) rdc_buffer; + ccw->flags = 0; + } + + ccw->count = rdc_buffer_size; + cqr->startdev = device; + cqr->memdev = device; + cqr->expires = 10*HZ; + cqr->retries = 256; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +} + + +int dasd_generic_read_dev_chars(struct dasd_device *device, int magic, + void *rdc_buffer, int rdc_buffer_size) +{ + int ret; + struct dasd_ccw_req *cqr; + + cqr = dasd_generic_build_rdc(device, rdc_buffer, rdc_buffer_size, + magic); + if (IS_ERR(cqr)) + return PTR_ERR(cqr); + + ret = dasd_sleep_on(cqr); + dasd_sfree_request(cqr, cqr->memdev); + return ret; +} +EXPORT_SYMBOL_GPL(dasd_generic_read_dev_chars); + +/* + * In command mode and transport mode we need to look for sense + * data in different places. The sense data itself is allways + * an array of 32 bytes, so we can unify the sense data access + * for both modes. + */ +char *dasd_get_sense(struct irb *irb) +{ + struct tsb *tsb = NULL; + char *sense = NULL; + + if (scsw_is_tm(&irb->scsw) && (irb->scsw.tm.fcxs == 0x01)) { + if (irb->scsw.tm.tcw) + tsb = tcw_get_tsb((struct tcw *)(unsigned long) + irb->scsw.tm.tcw); + if (tsb && tsb->length == 64 && tsb->flags) + switch (tsb->flags & 0x07) { + case 1: /* tsa_iostat */ + sense = tsb->tsa.iostat.sense; + break; + case 2: /* tsa_ddpc */ + sense = tsb->tsa.ddpc.sense; + break; + default: + /* currently we don't use interrogate data */ + break; + } + } else if (irb->esw.esw0.erw.cons) { + sense = irb->ecw; + } + return sense; +} +EXPORT_SYMBOL_GPL(dasd_get_sense); + +void dasd_generic_shutdown(struct ccw_device *cdev) +{ + struct dasd_device *device; + + device = dasd_device_from_cdev(cdev); + if (IS_ERR(device)) + return; + + if (device->block) + dasd_schedule_block_bh(device->block); + + dasd_schedule_device_bh(device); + + wait_event(shutdown_waitq, _wait_for_empty_queues(device)); +} +EXPORT_SYMBOL_GPL(dasd_generic_shutdown); + +static int __init dasd_init(void) +{ + int rc; + + init_waitqueue_head(&dasd_init_waitq); + init_waitqueue_head(&dasd_flush_wq); + init_waitqueue_head(&generic_waitq); + init_waitqueue_head(&shutdown_waitq); + + /* register 'common' DASD debug area, used for all DBF_XXX calls */ + dasd_debug_area = debug_register("dasd", 1, 1, 8 * sizeof(long)); + if (dasd_debug_area == NULL) { + rc = -ENOMEM; + goto failed; + } + debug_register_view(dasd_debug_area, &debug_sprintf_view); + debug_set_level(dasd_debug_area, DBF_WARNING); + + DBF_EVENT(DBF_EMERG, "%s", "debug area created"); + + dasd_diag_discipline_pointer = NULL; + + dasd_statistics_createroot(); + + rc = dasd_devmap_init(); + if (rc) + goto failed; + rc = dasd_gendisk_init(); + if (rc) + goto failed; + rc = dasd_parse(); + if (rc) + goto failed; + rc = dasd_eer_init(); + if (rc) + goto failed; +#ifdef CONFIG_PROC_FS + rc = dasd_proc_init(); + if (rc) + goto failed; +#endif + + return 0; +failed: + pr_info("The DASD device driver could not be initialized\n"); + dasd_exit(); + return rc; +} + +module_init(dasd_init); +module_exit(dasd_exit); diff --git a/kernel/drivers/s390/block/dasd_3990_erp.c b/kernel/drivers/s390/block/dasd_3990_erp.c new file mode 100644 index 000000000..d26134713 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_3990_erp.c @@ -0,0 +1,2771 @@ +/* + * Author(s)......: Horst Hummel <Horst.Hummel@de.ibm.com> + * Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 2000, 2001 + * + */ + +#define KMSG_COMPONENT "dasd-eckd" + +#include <linux/timer.h> +#include <asm/idals.h> + +#define PRINTK_HEADER "dasd_erp(3990): " + +#include "dasd_int.h" +#include "dasd_eckd.h" + + +struct DCTL_data { + unsigned char subcommand; /* e.g Inhibit Write, Enable Write,... */ + unsigned char modifier; /* Subcommand modifier */ + unsigned short res; /* reserved */ +} __attribute__ ((packed)); + +/* + ***************************************************************************** + * SECTION ERP HANDLING + ***************************************************************************** + */ +/* + ***************************************************************************** + * 24 and 32 byte sense ERP functions + ***************************************************************************** + */ + +/* + * DASD_3990_ERP_CLEANUP + * + * DESCRIPTION + * Removes the already build but not necessary ERP request and sets + * the status of the original cqr / erp to the given (final) status + * + * PARAMETER + * erp request to be blocked + * final_status either DASD_CQR_DONE or DASD_CQR_FAILED + * + * RETURN VALUES + * cqr original cqr + */ +static struct dasd_ccw_req * +dasd_3990_erp_cleanup(struct dasd_ccw_req * erp, char final_status) +{ + struct dasd_ccw_req *cqr = erp->refers; + + dasd_free_erp_request(erp, erp->memdev); + cqr->status = final_status; + return cqr; + +} /* end dasd_3990_erp_cleanup */ + +/* + * DASD_3990_ERP_BLOCK_QUEUE + * + * DESCRIPTION + * Block the given device request queue to prevent from further + * processing until the started timer has expired or an related + * interrupt was received. + */ +static void dasd_3990_erp_block_queue(struct dasd_ccw_req *erp, int expires) +{ + + struct dasd_device *device = erp->startdev; + unsigned long flags; + + DBF_DEV_EVENT(DBF_INFO, device, + "blocking request queue for %is", expires/HZ); + + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dasd_device_set_stop_bits(device, DASD_STOPPED_PENDING); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + erp->status = DASD_CQR_FILLED; + if (erp->block) + dasd_block_set_timer(erp->block, expires); + else + dasd_device_set_timer(device, expires); +} + +/* + * DASD_3990_ERP_INT_REQ + * + * DESCRIPTION + * Handles 'Intervention Required' error. + * This means either device offline or not installed. + * + * PARAMETER + * erp current erp + * RETURN VALUES + * erp modified erp + */ +static struct dasd_ccw_req * +dasd_3990_erp_int_req(struct dasd_ccw_req * erp) +{ + + struct dasd_device *device = erp->startdev; + + /* first time set initial retry counter and erp_function */ + /* and retry once without blocking queue */ + /* (this enables easier enqueing of the cqr) */ + if (erp->function != dasd_3990_erp_int_req) { + + erp->retries = 256; + erp->function = dasd_3990_erp_int_req; + + } else { + + /* issue a message and wait for 'device ready' interrupt */ + dev_err(&device->cdev->dev, + "is offline or not installed - " + "INTERVENTION REQUIRED!!\n"); + + dasd_3990_erp_block_queue(erp, 60*HZ); + } + + return erp; + +} /* end dasd_3990_erp_int_req */ + +/* + * DASD_3990_ERP_ALTERNATE_PATH + * + * DESCRIPTION + * Repeat the operation on a different channel path. + * If all alternate paths have been tried, the request is posted with a + * permanent error. + * + * PARAMETER + * erp pointer to the current ERP + * + * RETURN VALUES + * erp modified pointer to the ERP + */ +static void +dasd_3990_erp_alternate_path(struct dasd_ccw_req * erp) +{ + struct dasd_device *device = erp->startdev; + __u8 opm; + unsigned long flags; + + /* try alternate valid path */ + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + opm = ccw_device_get_path_mask(device->cdev); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + if (erp->lpm == 0) + erp->lpm = device->path_data.opm & + ~(erp->irb.esw.esw0.sublog.lpum); + else + erp->lpm &= ~(erp->irb.esw.esw0.sublog.lpum); + + if ((erp->lpm & opm) != 0x00) { + + DBF_DEV_EVENT(DBF_WARNING, device, + "try alternate lpm=%x (lpum=%x / opm=%x)", + erp->lpm, erp->irb.esw.esw0.sublog.lpum, opm); + + /* reset status to submit the request again... */ + erp->status = DASD_CQR_FILLED; + erp->retries = 10; + } else { + dev_err(&device->cdev->dev, + "The DASD cannot be reached on any path (lpum=%x" + "/opm=%x)\n", erp->irb.esw.esw0.sublog.lpum, opm); + + /* post request with permanent error */ + erp->status = DASD_CQR_FAILED; + } +} /* end dasd_3990_erp_alternate_path */ + +/* + * DASD_3990_ERP_DCTL + * + * DESCRIPTION + * Setup cqr to do the Diagnostic Control (DCTL) command with an + * Inhibit Write subcommand (0x20) and the given modifier. + * + * PARAMETER + * erp pointer to the current (failed) ERP + * modifier subcommand modifier + * + * RETURN VALUES + * dctl_cqr pointer to NEW dctl_cqr + * + */ +static struct dasd_ccw_req * +dasd_3990_erp_DCTL(struct dasd_ccw_req * erp, char modifier) +{ + + struct dasd_device *device = erp->startdev; + struct DCTL_data *DCTL_data; + struct ccw1 *ccw; + struct dasd_ccw_req *dctl_cqr; + + dctl_cqr = dasd_alloc_erp_request((char *) &erp->magic, 1, + sizeof(struct DCTL_data), + device); + if (IS_ERR(dctl_cqr)) { + dev_err(&device->cdev->dev, + "Unable to allocate DCTL-CQR\n"); + erp->status = DASD_CQR_FAILED; + return erp; + } + + DCTL_data = dctl_cqr->data; + + DCTL_data->subcommand = 0x02; /* Inhibit Write */ + DCTL_data->modifier = modifier; + + ccw = dctl_cqr->cpaddr; + memset(ccw, 0, sizeof(struct ccw1)); + ccw->cmd_code = CCW_CMD_DCTL; + ccw->count = 4; + ccw->cda = (__u32)(addr_t) DCTL_data; + dctl_cqr->flags = erp->flags; + dctl_cqr->function = dasd_3990_erp_DCTL; + dctl_cqr->refers = erp; + dctl_cqr->startdev = device; + dctl_cqr->memdev = device; + dctl_cqr->magic = erp->magic; + dctl_cqr->expires = 5 * 60 * HZ; + dctl_cqr->retries = 2; + + dctl_cqr->buildclk = get_tod_clock(); + + dctl_cqr->status = DASD_CQR_FILLED; + + return dctl_cqr; + +} /* end dasd_3990_erp_DCTL */ + +/* + * DASD_3990_ERP_ACTION_1 + * + * DESCRIPTION + * Setup ERP to do the ERP action 1 (see Reference manual). + * Repeat the operation on a different channel path. + * As deviation from the recommended recovery action, we reset the path mask + * after we have tried each path and go through all paths a second time. + * This will cover situations where only one path at a time is actually down, + * but all paths fail and recover just with the same sequence and timing as + * we try to use them (flapping links). + * If all alternate paths have been tried twice, the request is posted with + * a permanent error. + * + * PARAMETER + * erp pointer to the current ERP + * + * RETURN VALUES + * erp pointer to the ERP + * + */ +static struct dasd_ccw_req *dasd_3990_erp_action_1_sec(struct dasd_ccw_req *erp) +{ + erp->function = dasd_3990_erp_action_1_sec; + dasd_3990_erp_alternate_path(erp); + return erp; +} + +static struct dasd_ccw_req *dasd_3990_erp_action_1(struct dasd_ccw_req *erp) +{ + erp->function = dasd_3990_erp_action_1; + dasd_3990_erp_alternate_path(erp); + if (erp->status == DASD_CQR_FAILED && + !test_bit(DASD_CQR_VERIFY_PATH, &erp->flags)) { + erp->status = DASD_CQR_FILLED; + erp->retries = 10; + erp->lpm = erp->startdev->path_data.opm; + erp->function = dasd_3990_erp_action_1_sec; + } + return erp; +} /* end dasd_3990_erp_action_1(b) */ + +/* + * DASD_3990_ERP_ACTION_4 + * + * DESCRIPTION + * Setup ERP to do the ERP action 4 (see Reference manual). + * Set the current request to PENDING to block the CQR queue for that device + * until the state change interrupt appears. + * Use a timer (20 seconds) to retry the cqr if the interrupt is still + * missing. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the current ERP + * + * RETURN VALUES + * erp pointer to the ERP + * + */ +static struct dasd_ccw_req * +dasd_3990_erp_action_4(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + /* first time set initial retry counter and erp_function */ + /* and retry once without waiting for state change pending */ + /* interrupt (this enables easier enqueing of the cqr) */ + if (erp->function != dasd_3990_erp_action_4) { + + DBF_DEV_EVENT(DBF_INFO, device, "%s", + "dasd_3990_erp_action_4: first time retry"); + + erp->retries = 256; + erp->function = dasd_3990_erp_action_4; + + } else { + if (sense && (sense[25] == 0x1D)) { /* state change pending */ + + DBF_DEV_EVENT(DBF_INFO, device, + "waiting for state change pending " + "interrupt, %d retries left", + erp->retries); + + dasd_3990_erp_block_queue(erp, 30*HZ); + + } else if (sense && (sense[25] == 0x1E)) { /* busy */ + DBF_DEV_EVENT(DBF_INFO, device, + "busy - redriving request later, " + "%d retries left", + erp->retries); + dasd_3990_erp_block_queue(erp, HZ); + } else { + /* no state change pending - retry */ + DBF_DEV_EVENT(DBF_INFO, device, + "redriving request immediately, " + "%d retries left", + erp->retries); + erp->status = DASD_CQR_FILLED; + } + } + + return erp; + +} /* end dasd_3990_erp_action_4 */ + +/* + ***************************************************************************** + * 24 byte sense ERP functions (only) + ***************************************************************************** + */ + +/* + * DASD_3990_ERP_ACTION_5 + * + * DESCRIPTION + * Setup ERP to do the ERP action 5 (see Reference manual). + * NOTE: Further handling is done in xxx_further_erp after the retries. + * + * PARAMETER + * erp pointer to the current ERP + * + * RETURN VALUES + * erp pointer to the ERP + * + */ +static struct dasd_ccw_req * +dasd_3990_erp_action_5(struct dasd_ccw_req * erp) +{ + + /* first of all retry */ + erp->retries = 10; + erp->function = dasd_3990_erp_action_5; + + return erp; + +} /* end dasd_3990_erp_action_5 */ + +/* + * DASD_3990_HANDLE_ENV_DATA + * + * DESCRIPTION + * Handles 24 byte 'Environmental data present'. + * Does a analysis of the sense data (message Format) + * and prints the error messages. + * + * PARAMETER + * sense current sense data + * + * RETURN VALUES + * void + */ +static void +dasd_3990_handle_env_data(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + char msg_format = (sense[7] & 0xF0); + char msg_no = (sense[7] & 0x0F); + char errorstring[ERRORLENGTH]; + + switch (msg_format) { + case 0x00: /* Format 0 - Program or System Checks */ + + if (sense[1] & 0x10) { /* check message to operator bit */ + + switch (msg_no) { + case 0x00: /* No Message */ + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Invalid Command\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Invalid Command " + "Sequence\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 0 - CCW Count less than " + "required\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Invalid Parameter\n"); + break; + case 0x05: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Diagnostic of Special" + " Command Violates File Mask\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Channel Returned with " + "Incorrect retry CCW\n"); + break; + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Reset Notification\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Storage Path Restart\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Channel requested " + "... %02x\n", sense[8]); + break; + case 0x0B: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Invalid Defective/" + "Alternate Track Pointer\n"); + break; + case 0x0C: + dev_warn(&device->cdev->dev, + "FORMAT 0 - DPS Installation " + "Check\n"); + break; + case 0x0E: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Command Invalid on " + "Secondary Address\n"); + break; + case 0x0F: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Status Not As " + "Required: reason %02x\n", + sense[8]); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Reserved\n"); + } + } else { + switch (msg_no) { + case 0x00: /* No Message */ + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Device Error " + "Source\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Reserved\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Device Fenced - " + "device = %02x\n", sense[4]); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Data Pinned for " + "Device\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 0 - Reserved\n"); + } + } + break; + + case 0x10: /* Format 1 - Device Equipment Checks */ + switch (msg_no) { + case 0x00: /* No Message */ + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Device Status 1 not as " + "expected\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Index missing\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Interruption cannot be " + "reset\n"); + break; + case 0x05: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Device did not respond to " + "selection\n"); + break; + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Device check-2 error or Set " + "Sector is not complete\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Head address does not " + "compare\n"); + break; + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Device status 1 not valid\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Device not ready\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Track physical address did " + "not compare\n"); + break; + case 0x0B: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Missing device address bit\n"); + break; + case 0x0C: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Drive motor switch is off\n"); + break; + case 0x0D: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Seek incomplete\n"); + break; + case 0x0E: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Cylinder address did not " + "compare\n"); + break; + case 0x0F: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Offset active cannot be " + "reset\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 1 - Reserved\n"); + } + break; + + case 0x20: /* Format 2 - 3990 Equipment Checks */ + switch (msg_no) { + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 2 - 3990 check-2 error\n"); + break; + case 0x0E: + dev_warn(&device->cdev->dev, + "FORMAT 2 - Support facility errors\n"); + break; + case 0x0F: + dev_warn(&device->cdev->dev, + "FORMAT 2 - Microcode detected error " + "%02x\n", + sense[8]); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 2 - Reserved\n"); + } + break; + + case 0x30: /* Format 3 - 3990 Control Checks */ + switch (msg_no) { + case 0x0F: + dev_warn(&device->cdev->dev, + "FORMAT 3 - Allegiance terminated\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 3 - Reserved\n"); + } + break; + + case 0x40: /* Format 4 - Data Checks */ + switch (msg_no) { + case 0x00: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Home address area error\n"); + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Count area error\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Key area error\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Data area error\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No sync byte in home address " + "area\n"); + break; + case 0x05: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No sync byte in count address " + "area\n"); + break; + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No sync byte in key area\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No sync byte in data area\n"); + break; + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Home address area error; " + "offset active\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Count area error; offset " + "active\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Key area error; offset " + "active\n"); + break; + case 0x0B: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Data area error; " + "offset active\n"); + break; + case 0x0C: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No sync byte in home " + "address area; offset active\n"); + break; + case 0x0D: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No syn byte in count " + "address area; offset active\n"); + break; + case 0x0E: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No sync byte in key area; " + "offset active\n"); + break; + case 0x0F: + dev_warn(&device->cdev->dev, + "FORMAT 4 - No syn byte in data area; " + "offset active\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 4 - Reserved\n"); + } + break; + + case 0x50: /* Format 5 - Data Check with displacement information */ + switch (msg_no) { + case 0x00: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the " + "home address area\n"); + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the count " + "area\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the key area\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the data " + "area\n"); + break; + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the " + "home address area; offset active\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the count area; " + "offset active\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the key area; " + "offset active\n"); + break; + case 0x0B: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Data Check in the data area; " + "offset active\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 5 - Reserved\n"); + } + break; + + case 0x60: /* Format 6 - Usage Statistics/Overrun Errors */ + switch (msg_no) { + case 0x00: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel A\n"); + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel B\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel C\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel D\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel E\n"); + break; + case 0x05: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel F\n"); + break; + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel G\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Overrun on channel H\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 6 - Reserved\n"); + } + break; + + case 0x70: /* Format 7 - Device Connection Control Checks */ + switch (msg_no) { + case 0x00: + dev_warn(&device->cdev->dev, + "FORMAT 7 - RCC initiated by a connection " + "check alert\n"); + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 7 - RCC 1 sequence not " + "successful\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT 7 - RCC 1 and RCC 2 sequences not " + "successful\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Invalid tag-in during " + "selection sequence\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 7 - extra RCC required\n"); + break; + case 0x05: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Invalid DCC selection " + "response or timeout\n"); + break; + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Missing end operation; device " + "transfer complete\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Missing end operation; device " + "transfer incomplete\n"); + break; + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Invalid tag-in for an " + "immediate command sequence\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Invalid tag-in for an " + "extended command sequence\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 7 - 3990 microcode time out when " + "stopping selection\n"); + break; + case 0x0B: + dev_warn(&device->cdev->dev, + "FORMAT 7 - No response to selection " + "after a poll interruption\n"); + break; + case 0x0C: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Permanent path error (DASD " + "controller not available)\n"); + break; + case 0x0D: + dev_warn(&device->cdev->dev, + "FORMAT 7 - DASD controller not available" + " on disconnected command chain\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 7 - Reserved\n"); + } + break; + + case 0x80: /* Format 8 - Additional Device Equipment Checks */ + switch (msg_no) { + case 0x00: /* No Message */ + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT 8 - Error correction code " + "hardware fault\n"); + break; + case 0x03: + dev_warn(&device->cdev->dev, + "FORMAT 8 - Unexpected end operation " + "response code\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT 8 - End operation with transfer " + "count not zero\n"); + break; + case 0x05: + dev_warn(&device->cdev->dev, + "FORMAT 8 - End operation with transfer " + "count zero\n"); + break; + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT 8 - DPS checks after a system " + "reset or selective reset\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 8 - DPS cannot be filled\n"); + break; + case 0x08: + dev_warn(&device->cdev->dev, + "FORMAT 8 - Short busy time-out during " + "device selection\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT 8 - DASD controller failed to " + "set or reset the long busy latch\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 8 - No interruption from device " + "during a command chain\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 8 - Reserved\n"); + } + break; + + case 0x90: /* Format 9 - Device Read, Write, and Seek Checks */ + switch (msg_no) { + case 0x00: + break; /* No Message */ + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT 9 - Device check-2 error\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT 9 - Head address did not " + "compare\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT 9 - Track physical address did " + "not compare while oriented\n"); + break; + case 0x0E: + dev_warn(&device->cdev->dev, + "FORMAT 9 - Cylinder address did not " + "compare\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT 9 - Reserved\n"); + } + break; + + case 0xF0: /* Format F - Cache Storage Checks */ + switch (msg_no) { + case 0x00: + dev_warn(&device->cdev->dev, + "FORMAT F - Operation Terminated\n"); + break; + case 0x01: + dev_warn(&device->cdev->dev, + "FORMAT F - Subsystem Processing Error\n"); + break; + case 0x02: + dev_warn(&device->cdev->dev, + "FORMAT F - Cache or nonvolatile storage " + "equipment failure\n"); + break; + case 0x04: + dev_warn(&device->cdev->dev, + "FORMAT F - Caching terminated\n"); + break; + case 0x06: + dev_warn(&device->cdev->dev, + "FORMAT F - Cache fast write access not " + "authorized\n"); + break; + case 0x07: + dev_warn(&device->cdev->dev, + "FORMAT F - Track format incorrect\n"); + break; + case 0x09: + dev_warn(&device->cdev->dev, + "FORMAT F - Caching reinitiated\n"); + break; + case 0x0A: + dev_warn(&device->cdev->dev, + "FORMAT F - Nonvolatile storage " + "terminated\n"); + break; + case 0x0B: + dev_warn(&device->cdev->dev, + "FORMAT F - Volume is suspended duplex\n"); + /* call extended error reporting (EER) */ + dasd_eer_write(device, erp->refers, + DASD_EER_PPRCSUSPEND); + break; + case 0x0C: + dev_warn(&device->cdev->dev, + "FORMAT F - Subsystem status cannot be " + "determined\n"); + break; + case 0x0D: + dev_warn(&device->cdev->dev, + "FORMAT F - Caching status reset to " + "default\n"); + break; + case 0x0E: + dev_warn(&device->cdev->dev, + "FORMAT F - DASD Fast Write inhibited\n"); + break; + default: + dev_warn(&device->cdev->dev, + "FORMAT D - Reserved\n"); + } + break; + + default: /* unknown message format - should not happen + internal error 03 - unknown message format */ + snprintf(errorstring, ERRORLENGTH, "03 %x02", msg_format); + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", errorstring); + break; + } /* end switch message format */ + +} /* end dasd_3990_handle_env_data */ + +/* + * DASD_3990_ERP_COM_REJ + * + * DESCRIPTION + * Handles 24 byte 'Command Reject' error. + * + * PARAMETER + * erp current erp_head + * sense current sense data + * + * RETURN VALUES + * erp 'new' erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_com_rej(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_com_rej; + + /* env data present (ACTION 10 - retry should work) */ + if (sense[2] & SNS2_ENV_DATA_PRESENT) { + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Command Reject - environmental data present"); + + dasd_3990_handle_env_data(erp, sense); + + erp->retries = 5; + + } else if (sense[1] & SNS1_WRITE_INHIBITED) { + dev_err(&device->cdev->dev, "An I/O request was rejected" + " because writing is inhibited\n"); + erp = dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + } else { + /* fatal error - set status to FAILED + internal error 09 - Command Reject */ + dev_err(&device->cdev->dev, "An error occurred in the DASD " + "device driver, reason=%s\n", "09"); + + erp = dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + } + + return erp; + +} /* end dasd_3990_erp_com_rej */ + +/* + * DASD_3990_ERP_BUS_OUT + * + * DESCRIPTION + * Handles 24 byte 'Bus Out Parity Check' error. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_bus_out(struct dasd_ccw_req * erp) +{ + + struct dasd_device *device = erp->startdev; + + /* first time set initial retry counter and erp_function */ + /* and retry once without blocking queue */ + /* (this enables easier enqueing of the cqr) */ + if (erp->function != dasd_3990_erp_bus_out) { + erp->retries = 256; + erp->function = dasd_3990_erp_bus_out; + + } else { + + /* issue a message and wait for 'device ready' interrupt */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "bus out parity error or BOPC requested by " + "channel"); + + dasd_3990_erp_block_queue(erp, 60*HZ); + + } + + return erp; + +} /* end dasd_3990_erp_bus_out */ + +/* + * DASD_3990_ERP_EQUIP_CHECK + * + * DESCRIPTION + * Handles 24 byte 'Equipment Check' error. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_equip_check(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_equip_check; + + if (sense[1] & SNS1_WRITE_INHIBITED) { + dev_info(&device->cdev->dev, + "Write inhibited path encountered\n"); + + /* vary path offline + internal error 04 - Path should be varied off-line.*/ + dev_err(&device->cdev->dev, "An error occurred in the DASD " + "device driver, reason=%s\n", "04"); + + erp = dasd_3990_erp_action_1(erp); + + } else if (sense[2] & SNS2_ENV_DATA_PRESENT) { + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Equipment Check - " "environmental data present"); + + dasd_3990_handle_env_data(erp, sense); + + erp = dasd_3990_erp_action_4(erp, sense); + + } else if (sense[1] & SNS1_PERM_ERR) { + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Equipment Check - retry exhausted or " + "undesirable"); + + erp = dasd_3990_erp_action_1(erp); + + } else { + /* all other equipment checks - Action 5 */ + /* rest is done when retries == 0 */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Equipment check or processing error"); + + erp = dasd_3990_erp_action_5(erp); + } + return erp; + +} /* end dasd_3990_erp_equip_check */ + +/* + * DASD_3990_ERP_DATA_CHECK + * + * DESCRIPTION + * Handles 24 byte 'Data Check' error. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_data_check(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_data_check; + + if (sense[2] & SNS2_CORRECTABLE) { /* correctable data check */ + + /* issue message that the data has been corrected */ + dev_emerg(&device->cdev->dev, + "Data recovered during retry with PCI " + "fetch mode active\n"); + + /* not possible to handle this situation in Linux */ + panic("No way to inform application about the possibly " + "incorrect data"); + + } else if (sense[2] & SNS2_ENV_DATA_PRESENT) { + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Uncorrectable data check recovered secondary " + "addr of duplex pair"); + + erp = dasd_3990_erp_action_4(erp, sense); + + } else if (sense[1] & SNS1_PERM_ERR) { + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Uncorrectable data check with internal " + "retry exhausted"); + + erp = dasd_3990_erp_action_1(erp); + + } else { + /* all other data checks */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Uncorrectable data check with retry count " + "exhausted..."); + + erp = dasd_3990_erp_action_5(erp); + } + + return erp; + +} /* end dasd_3990_erp_data_check */ + +/* + * DASD_3990_ERP_OVERRUN + * + * DESCRIPTION + * Handles 24 byte 'Overrun' error. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_overrun(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_overrun; + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Overrun - service overrun or overrun" + " error requested by channel"); + + erp = dasd_3990_erp_action_5(erp); + + return erp; + +} /* end dasd_3990_erp_overrun */ + +/* + * DASD_3990_ERP_INV_FORMAT + * + * DESCRIPTION + * Handles 24 byte 'Invalid Track Format' error. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_inv_format(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_inv_format; + + if (sense[2] & SNS2_ENV_DATA_PRESENT) { + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Track format error when destaging or " + "staging data"); + + dasd_3990_handle_env_data(erp, sense); + + erp = dasd_3990_erp_action_4(erp, sense); + + } else { + /* internal error 06 - The track format is not valid*/ + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", "06"); + + erp = dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + } + + return erp; + +} /* end dasd_3990_erp_inv_format */ + +/* + * DASD_3990_ERP_EOC + * + * DESCRIPTION + * Handles 24 byte 'End-of-Cylinder' error. + * + * PARAMETER + * erp already added default erp + * RETURN VALUES + * erp pointer to original (failed) cqr. + */ +static struct dasd_ccw_req * +dasd_3990_erp_EOC(struct dasd_ccw_req * default_erp, char *sense) +{ + + struct dasd_device *device = default_erp->startdev; + + dev_err(&device->cdev->dev, + "The cylinder data for accessing the DASD is inconsistent\n"); + + /* implement action 7 - BUG */ + return dasd_3990_erp_cleanup(default_erp, DASD_CQR_FAILED); + +} /* end dasd_3990_erp_EOC */ + +/* + * DASD_3990_ERP_ENV_DATA + * + * DESCRIPTION + * Handles 24 byte 'Environmental-Data Present' error. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_env_data(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_env_data; + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", "Environmental data present"); + + dasd_3990_handle_env_data(erp, sense); + + /* don't retry on disabled interface */ + if (sense[7] != 0x0F) { + erp = dasd_3990_erp_action_4(erp, sense); + } else { + erp->status = DASD_CQR_FILLED; + } + + return erp; + +} /* end dasd_3990_erp_env_data */ + +/* + * DASD_3990_ERP_NO_REC + * + * DESCRIPTION + * Handles 24 byte 'No Record Found' error. + * + * PARAMETER + * erp already added default ERP + * + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_no_rec(struct dasd_ccw_req * default_erp, char *sense) +{ + + struct dasd_device *device = default_erp->startdev; + + dev_err(&device->cdev->dev, + "The specified record was not found\n"); + + return dasd_3990_erp_cleanup(default_erp, DASD_CQR_FAILED); + +} /* end dasd_3990_erp_no_rec */ + +/* + * DASD_3990_ERP_FILE_PROT + * + * DESCRIPTION + * Handles 24 byte 'File Protected' error. + * Note: Seek related recovery is not implemented because + * wee don't use the seek command yet. + * + * PARAMETER + * erp current erp_head + * RETURN VALUES + * erp new erp_head - pointer to new ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_file_prot(struct dasd_ccw_req * erp) +{ + + struct dasd_device *device = erp->startdev; + + dev_err(&device->cdev->dev, "Accessing the DASD failed because of " + "a hardware error\n"); + + return dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + +} /* end dasd_3990_erp_file_prot */ + +/* + * DASD_3990_ERP_INSPECT_ALIAS + * + * DESCRIPTION + * Checks if the original request was started on an alias device. + * If yes, it modifies the original and the erp request so that + * the erp request can be started on a base device. + * + * PARAMETER + * erp pointer to the currently created default ERP + * + * RETURN VALUES + * erp pointer to the modified ERP, or NULL + */ + +static struct dasd_ccw_req *dasd_3990_erp_inspect_alias( + struct dasd_ccw_req *erp) +{ + struct dasd_ccw_req *cqr = erp->refers; + char *sense; + + if (cqr->block && + (cqr->block->base != cqr->startdev)) { + + sense = dasd_get_sense(&erp->refers->irb); + /* + * dynamic pav may have changed base alias mapping + */ + if (!test_bit(DASD_FLAG_OFFLINE, &cqr->startdev->flags) && sense + && (sense[0] == 0x10) && (sense[7] == 0x0F) + && (sense[8] == 0x67)) { + /* + * remove device from alias handling to prevent new + * requests from being scheduled on the + * wrong alias device + */ + dasd_alias_remove_device(cqr->startdev); + + /* schedule worker to reload device */ + dasd_reload_device(cqr->startdev); + } + + if (cqr->startdev->features & DASD_FEATURE_ERPLOG) { + DBF_DEV_EVENT(DBF_ERR, cqr->startdev, + "ERP on alias device for request %p," + " recover on base device %s", cqr, + dev_name(&cqr->block->base->cdev->dev)); + } + dasd_eckd_reset_ccw_to_base_io(cqr); + erp->startdev = cqr->block->base; + erp->function = dasd_3990_erp_inspect_alias; + return erp; + } else + return NULL; +} + + +/* + * DASD_3990_ERP_INSPECT_24 + * + * DESCRIPTION + * Does a detailed inspection of the 24 byte sense data + * and sets up a related error recovery action. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the currently created default ERP + * + * RETURN VALUES + * erp pointer to the (addtitional) ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_inspect_24(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_ccw_req *erp_filled = NULL; + + /* Check sense for .... */ + /* 'Command Reject' */ + if ((erp_filled == NULL) && (sense[0] & SNS0_CMD_REJECT)) { + erp_filled = dasd_3990_erp_com_rej(erp, sense); + } + /* 'Intervention Required' */ + if ((erp_filled == NULL) && (sense[0] & SNS0_INTERVENTION_REQ)) { + erp_filled = dasd_3990_erp_int_req(erp); + } + /* 'Bus Out Parity Check' */ + if ((erp_filled == NULL) && (sense[0] & SNS0_BUS_OUT_CHECK)) { + erp_filled = dasd_3990_erp_bus_out(erp); + } + /* 'Equipment Check' */ + if ((erp_filled == NULL) && (sense[0] & SNS0_EQUIPMENT_CHECK)) { + erp_filled = dasd_3990_erp_equip_check(erp, sense); + } + /* 'Data Check' */ + if ((erp_filled == NULL) && (sense[0] & SNS0_DATA_CHECK)) { + erp_filled = dasd_3990_erp_data_check(erp, sense); + } + /* 'Overrun' */ + if ((erp_filled == NULL) && (sense[0] & SNS0_OVERRUN)) { + erp_filled = dasd_3990_erp_overrun(erp, sense); + } + /* 'Invalid Track Format' */ + if ((erp_filled == NULL) && (sense[1] & SNS1_INV_TRACK_FORMAT)) { + erp_filled = dasd_3990_erp_inv_format(erp, sense); + } + /* 'End-of-Cylinder' */ + if ((erp_filled == NULL) && (sense[1] & SNS1_EOC)) { + erp_filled = dasd_3990_erp_EOC(erp, sense); + } + /* 'Environmental Data' */ + if ((erp_filled == NULL) && (sense[2] & SNS2_ENV_DATA_PRESENT)) { + erp_filled = dasd_3990_erp_env_data(erp, sense); + } + /* 'No Record Found' */ + if ((erp_filled == NULL) && (sense[1] & SNS1_NO_REC_FOUND)) { + erp_filled = dasd_3990_erp_no_rec(erp, sense); + } + /* 'File Protected' */ + if ((erp_filled == NULL) && (sense[1] & SNS1_FILE_PROTECTED)) { + erp_filled = dasd_3990_erp_file_prot(erp); + } + /* other (unknown) error - do default ERP */ + if (erp_filled == NULL) { + + erp_filled = erp; + } + + return erp_filled; + +} /* END dasd_3990_erp_inspect_24 */ + +/* + ***************************************************************************** + * 32 byte sense ERP functions (only) + ***************************************************************************** + */ + +/* + * DASD_3990_ERPACTION_10_32 + * + * DESCRIPTION + * Handles 32 byte 'Action 10' of Single Program Action Codes. + * Just retry and if retry doesn't work, return with error. + * + * PARAMETER + * erp current erp_head + * sense current sense data + * RETURN VALUES + * erp modified erp_head + */ +static struct dasd_ccw_req * +dasd_3990_erp_action_10_32(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->retries = 256; + erp->function = dasd_3990_erp_action_10_32; + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", "Perform logging requested"); + + return erp; + +} /* end dasd_3990_erp_action_10_32 */ + +/* + * DASD_3990_ERP_ACTION_1B_32 + * + * DESCRIPTION + * Handles 32 byte 'Action 1B' of Single Program Action Codes. + * A write operation could not be finished because of an unexpected + * condition. + * The already created 'default erp' is used to get the link to + * the erp chain, but it can not be used for this recovery + * action because it contains no DE/LO data space. + * + * PARAMETER + * default_erp already added default erp. + * sense current sense data + * + * RETURN VALUES + * erp new erp or + * default_erp in case of imprecise ending or error + */ +static struct dasd_ccw_req * +dasd_3990_erp_action_1B_32(struct dasd_ccw_req * default_erp, char *sense) +{ + + struct dasd_device *device = default_erp->startdev; + __u32 cpa = 0; + struct dasd_ccw_req *cqr; + struct dasd_ccw_req *erp; + struct DE_eckd_data *DE_data; + struct PFX_eckd_data *PFX_data; + char *LO_data; /* LO_eckd_data_t */ + struct ccw1 *ccw, *oldccw; + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Write not finished because of unexpected condition"); + + default_erp->function = dasd_3990_erp_action_1B_32; + + /* determine the original cqr */ + cqr = default_erp; + + while (cqr->refers != NULL) { + cqr = cqr->refers; + } + + if (scsw_is_tm(&cqr->irb.scsw)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "32 bit sense, action 1B is not defined" + " in transport mode - just retry"); + return default_erp; + } + + /* for imprecise ending just do default erp */ + if (sense[1] & 0x01) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Imprecise ending is set - just retry"); + + return default_erp; + } + + /* determine the address of the CCW to be restarted */ + /* Imprecise ending is not set -> addr from IRB-SCSW */ + cpa = default_erp->refers->irb.scsw.cmd.cpa; + + if (cpa == 0) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Unable to determine address of the CCW " + "to be restarted"); + + return dasd_3990_erp_cleanup(default_erp, DASD_CQR_FAILED); + } + + /* Build new ERP request including DE/LO */ + erp = dasd_alloc_erp_request((char *) &cqr->magic, + 2 + 1,/* DE/LO + TIC */ + sizeof(struct DE_eckd_data) + + sizeof(struct LO_eckd_data), device); + + if (IS_ERR(erp)) { + /* internal error 01 - Unable to allocate ERP */ + dev_err(&device->cdev->dev, "An error occurred in the DASD " + "device driver, reason=%s\n", "01"); + return dasd_3990_erp_cleanup(default_erp, DASD_CQR_FAILED); + } + + /* use original DE */ + DE_data = erp->data; + oldccw = cqr->cpaddr; + if (oldccw->cmd_code == DASD_ECKD_CCW_PFX) { + PFX_data = cqr->data; + memcpy(DE_data, &PFX_data->define_extent, + sizeof(struct DE_eckd_data)); + } else + memcpy(DE_data, cqr->data, sizeof(struct DE_eckd_data)); + + /* create LO */ + LO_data = erp->data + sizeof(struct DE_eckd_data); + + if ((sense[3] == 0x01) && (LO_data[1] & 0x01)) { + /* should not */ + return dasd_3990_erp_cleanup(default_erp, DASD_CQR_FAILED); + } + + if ((sense[7] & 0x3F) == 0x01) { + /* operation code is WRITE DATA -> data area orientation */ + LO_data[0] = 0x81; + + } else if ((sense[7] & 0x3F) == 0x03) { + /* operation code is FORMAT WRITE -> index orientation */ + LO_data[0] = 0xC3; + + } else { + LO_data[0] = sense[7]; /* operation */ + } + + LO_data[1] = sense[8]; /* auxiliary */ + LO_data[2] = sense[9]; + LO_data[3] = sense[3]; /* count */ + LO_data[4] = sense[29]; /* seek_addr.cyl */ + LO_data[5] = sense[30]; /* seek_addr.cyl 2nd byte */ + LO_data[7] = sense[31]; /* seek_addr.head 2nd byte */ + + memcpy(&(LO_data[8]), &(sense[11]), 8); + + /* create DE ccw */ + ccw = erp->cpaddr; + memset(ccw, 0, sizeof(struct ccw1)); + ccw->cmd_code = DASD_ECKD_CCW_DEFINE_EXTENT; + ccw->flags = CCW_FLAG_CC; + ccw->count = 16; + ccw->cda = (__u32)(addr_t) DE_data; + + /* create LO ccw */ + ccw++; + memset(ccw, 0, sizeof(struct ccw1)); + ccw->cmd_code = DASD_ECKD_CCW_LOCATE_RECORD; + ccw->flags = CCW_FLAG_CC; + ccw->count = 16; + ccw->cda = (__u32)(addr_t) LO_data; + + /* TIC to the failed ccw */ + ccw++; + ccw->cmd_code = CCW_CMD_TIC; + ccw->cda = cpa; + + /* fill erp related fields */ + erp->flags = default_erp->flags; + erp->function = dasd_3990_erp_action_1B_32; + erp->refers = default_erp->refers; + erp->startdev = device; + erp->memdev = device; + erp->magic = default_erp->magic; + erp->expires = default_erp->expires; + erp->retries = 256; + erp->buildclk = get_tod_clock(); + erp->status = DASD_CQR_FILLED; + + /* remove the default erp */ + dasd_free_erp_request(default_erp, device); + + return erp; + +} /* end dasd_3990_erp_action_1B_32 */ + +/* + * DASD_3990_UPDATE_1B + * + * DESCRIPTION + * Handles the update to the 32 byte 'Action 1B' of Single Program + * Action Codes in case the first action was not successful. + * The already created 'previous_erp' is the currently not successful + * ERP. + * + * PARAMETER + * previous_erp already created previous erp. + * sense current sense data + * RETURN VALUES + * erp modified erp + */ +static struct dasd_ccw_req * +dasd_3990_update_1B(struct dasd_ccw_req * previous_erp, char *sense) +{ + + struct dasd_device *device = previous_erp->startdev; + __u32 cpa = 0; + struct dasd_ccw_req *cqr; + struct dasd_ccw_req *erp; + char *LO_data; /* struct LO_eckd_data */ + struct ccw1 *ccw; + + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Write not finished because of unexpected condition" + " - follow on"); + + /* determine the original cqr */ + cqr = previous_erp; + + while (cqr->refers != NULL) { + cqr = cqr->refers; + } + + if (scsw_is_tm(&cqr->irb.scsw)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "32 bit sense, action 1B, update," + " in transport mode - just retry"); + return previous_erp; + } + + /* for imprecise ending just do default erp */ + if (sense[1] & 0x01) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Imprecise ending is set - just retry"); + + previous_erp->status = DASD_CQR_FILLED; + + return previous_erp; + } + + /* determine the address of the CCW to be restarted */ + /* Imprecise ending is not set -> addr from IRB-SCSW */ + cpa = previous_erp->irb.scsw.cmd.cpa; + + if (cpa == 0) { + /* internal error 02 - + Unable to determine address of the CCW to be restarted */ + dev_err(&device->cdev->dev, "An error occurred in the DASD " + "device driver, reason=%s\n", "02"); + + previous_erp->status = DASD_CQR_FAILED; + + return previous_erp; + } + + erp = previous_erp; + + /* update the LO with the new returned sense data */ + LO_data = erp->data + sizeof(struct DE_eckd_data); + + if ((sense[3] == 0x01) && (LO_data[1] & 0x01)) { + /* should not happen */ + previous_erp->status = DASD_CQR_FAILED; + + return previous_erp; + } + + if ((sense[7] & 0x3F) == 0x01) { + /* operation code is WRITE DATA -> data area orientation */ + LO_data[0] = 0x81; + + } else if ((sense[7] & 0x3F) == 0x03) { + /* operation code is FORMAT WRITE -> index orientation */ + LO_data[0] = 0xC3; + + } else { + LO_data[0] = sense[7]; /* operation */ + } + + LO_data[1] = sense[8]; /* auxiliary */ + LO_data[2] = sense[9]; + LO_data[3] = sense[3]; /* count */ + LO_data[4] = sense[29]; /* seek_addr.cyl */ + LO_data[5] = sense[30]; /* seek_addr.cyl 2nd byte */ + LO_data[7] = sense[31]; /* seek_addr.head 2nd byte */ + + memcpy(&(LO_data[8]), &(sense[11]), 8); + + /* TIC to the failed ccw */ + ccw = erp->cpaddr; /* addr of DE ccw */ + ccw++; /* addr of LE ccw */ + ccw++; /* addr of TIC ccw */ + ccw->cda = cpa; + + erp->status = DASD_CQR_FILLED; + + return erp; + +} /* end dasd_3990_update_1B */ + +/* + * DASD_3990_ERP_COMPOUND_RETRY + * + * DESCRIPTION + * Handles the compound ERP action retry code. + * NOTE: At least one retry is done even if zero is specified + * by the sense data. This makes enqueueing of the request + * easier. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the currently created ERP + * + * RETURN VALUES + * erp modified ERP pointer + * + */ +static void +dasd_3990_erp_compound_retry(struct dasd_ccw_req * erp, char *sense) +{ + + switch (sense[25] & 0x03) { + case 0x00: /* no not retry */ + erp->retries = 1; + break; + + case 0x01: /* retry 2 times */ + erp->retries = 2; + break; + + case 0x02: /* retry 10 times */ + erp->retries = 10; + break; + + case 0x03: /* retry 256 times */ + erp->retries = 256; + break; + + default: + BUG(); + } + + erp->function = dasd_3990_erp_compound_retry; + +} /* end dasd_3990_erp_compound_retry */ + +/* + * DASD_3990_ERP_COMPOUND_PATH + * + * DESCRIPTION + * Handles the compound ERP action for retry on alternate + * channel path. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the currently created ERP + * + * RETURN VALUES + * erp modified ERP pointer + * + */ +static void +dasd_3990_erp_compound_path(struct dasd_ccw_req * erp, char *sense) +{ + if (sense[25] & DASD_SENSE_BIT_3) { + dasd_3990_erp_alternate_path(erp); + + if (erp->status == DASD_CQR_FAILED && + !test_bit(DASD_CQR_VERIFY_PATH, &erp->flags)) { + /* reset the lpm and the status to be able to + * try further actions. */ + erp->lpm = erp->startdev->path_data.opm; + erp->status = DASD_CQR_NEED_ERP; + } + } + + erp->function = dasd_3990_erp_compound_path; + +} /* end dasd_3990_erp_compound_path */ + +/* + * DASD_3990_ERP_COMPOUND_CODE + * + * DESCRIPTION + * Handles the compound ERP action for retry code. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the currently created ERP + * + * RETURN VALUES + * erp NEW ERP pointer + * + */ +static struct dasd_ccw_req * +dasd_3990_erp_compound_code(struct dasd_ccw_req * erp, char *sense) +{ + + if (sense[25] & DASD_SENSE_BIT_2) { + + switch (sense[28]) { + case 0x17: + /* issue a Diagnostic Control command with an + * Inhibit Write subcommand and controller modifier */ + erp = dasd_3990_erp_DCTL(erp, 0x20); + break; + + case 0x25: + /* wait for 5 seconds and retry again */ + erp->retries = 1; + + dasd_3990_erp_block_queue (erp, 5*HZ); + break; + + default: + /* should not happen - continue */ + break; + } + } + + erp->function = dasd_3990_erp_compound_code; + + return erp; + +} /* end dasd_3990_erp_compound_code */ + +/* + * DASD_3990_ERP_COMPOUND_CONFIG + * + * DESCRIPTION + * Handles the compound ERP action for configruation + * dependent error. + * Note: duplex handling is not implemented (yet). + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the currently created ERP + * + * RETURN VALUES + * erp modified ERP pointer + * + */ +static void +dasd_3990_erp_compound_config(struct dasd_ccw_req * erp, char *sense) +{ + + if ((sense[25] & DASD_SENSE_BIT_1) && (sense[26] & DASD_SENSE_BIT_2)) { + + /* set to suspended duplex state then restart + internal error 05 - Set device to suspended duplex state + should be done */ + struct dasd_device *device = erp->startdev; + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", "05"); + + } + + erp->function = dasd_3990_erp_compound_config; + +} /* end dasd_3990_erp_compound_config */ + +/* + * DASD_3990_ERP_COMPOUND + * + * DESCRIPTION + * Does the further compound program action if + * compound retry was not successful. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the current (failed) ERP + * + * RETURN VALUES + * erp (additional) ERP pointer + * + */ +static struct dasd_ccw_req * +dasd_3990_erp_compound(struct dasd_ccw_req * erp, char *sense) +{ + + if ((erp->function == dasd_3990_erp_compound_retry) && + (erp->status == DASD_CQR_NEED_ERP)) { + + dasd_3990_erp_compound_path(erp, sense); + } + + if ((erp->function == dasd_3990_erp_compound_path) && + (erp->status == DASD_CQR_NEED_ERP)) { + + erp = dasd_3990_erp_compound_code(erp, sense); + } + + if ((erp->function == dasd_3990_erp_compound_code) && + (erp->status == DASD_CQR_NEED_ERP)) { + + dasd_3990_erp_compound_config(erp, sense); + } + + /* if no compound action ERP specified, the request failed */ + if (erp->status == DASD_CQR_NEED_ERP) + erp->status = DASD_CQR_FAILED; + + return erp; + +} /* end dasd_3990_erp_compound */ + +/* + *DASD_3990_ERP_HANDLE_SIM + * + *DESCRIPTION + * inspects the SIM SENSE data and starts an appropriate action + * + * PARAMETER + * sense sense data of the actual error + * + * RETURN VALUES + * none + */ +void +dasd_3990_erp_handle_sim(struct dasd_device *device, char *sense) +{ + /* print message according to log or message to operator mode */ + if ((sense[24] & DASD_SIM_MSG_TO_OP) || (sense[1] & 0x10)) { + /* print SIM SRC from RefCode */ + dev_err(&device->cdev->dev, "SIM - SRC: " + "%02x%02x%02x%02x\n", sense[22], + sense[23], sense[11], sense[12]); + } else if (sense[24] & DASD_SIM_LOG) { + /* print SIM SRC Refcode */ + dev_warn(&device->cdev->dev, "log SIM - SRC: " + "%02x%02x%02x%02x\n", sense[22], + sense[23], sense[11], sense[12]); + } +} + +/* + * DASD_3990_ERP_INSPECT_32 + * + * DESCRIPTION + * Does a detailed inspection of the 32 byte sense data + * and sets up a related error recovery action. + * + * PARAMETER + * sense sense data of the actual error + * erp pointer to the currently created default ERP + * + * RETURN VALUES + * erp_filled pointer to the ERP + * + */ +static struct dasd_ccw_req * +dasd_3990_erp_inspect_32(struct dasd_ccw_req * erp, char *sense) +{ + + struct dasd_device *device = erp->startdev; + + erp->function = dasd_3990_erp_inspect_32; + + /* check for SIM sense data */ + if ((sense[6] & DASD_SIM_SENSE) == DASD_SIM_SENSE) + dasd_3990_erp_handle_sim(device, sense); + + if (sense[25] & DASD_SENSE_BIT_0) { + + /* compound program action codes (byte25 bit 0 == '1') */ + dasd_3990_erp_compound_retry(erp, sense); + + } else { + + /* single program action codes (byte25 bit 0 == '0') */ + switch (sense[25]) { + + case 0x00: /* success - use default ERP for retries */ + DBF_DEV_EVENT(DBF_DEBUG, device, "%s", + "ERP called for successful request" + " - just retry"); + break; + + case 0x01: /* fatal error */ + dev_err(&device->cdev->dev, + "ERP failed for the DASD\n"); + + erp = dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + break; + + case 0x02: /* intervention required */ + case 0x03: /* intervention required during dual copy */ + erp = dasd_3990_erp_int_req(erp); + break; + + case 0x0F: /* length mismatch during update write command + internal error 08 - update write command error*/ + dev_err(&device->cdev->dev, "An error occurred in the " + "DASD device driver, reason=%s\n", "08"); + + erp = dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + break; + + case 0x10: /* logging required for other channel program */ + erp = dasd_3990_erp_action_10_32(erp, sense); + break; + + case 0x15: /* next track outside defined extend + internal error 07 - The next track is not + within the defined storage extent */ + dev_err(&device->cdev->dev, + "An error occurred in the DASD device driver, " + "reason=%s\n", "07"); + + erp = dasd_3990_erp_cleanup(erp, DASD_CQR_FAILED); + break; + + case 0x1B: /* unexpected condition during write */ + + erp = dasd_3990_erp_action_1B_32(erp, sense); + break; + + case 0x1C: /* invalid data */ + dev_emerg(&device->cdev->dev, + "Data recovered during retry with PCI " + "fetch mode active\n"); + + /* not possible to handle this situation in Linux */ + panic + ("Invalid data - No way to inform application " + "about the possibly incorrect data"); + break; + + case 0x1D: /* state-change pending */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "A State change pending condition exists " + "for the subsystem or device"); + + erp = dasd_3990_erp_action_4(erp, sense); + break; + + case 0x1E: /* busy */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Busy condition exists " + "for the subsystem or device"); + erp = dasd_3990_erp_action_4(erp, sense); + break; + + default: /* all others errors - default erp */ + break; + } + } + + return erp; + +} /* end dasd_3990_erp_inspect_32 */ + +/* + ***************************************************************************** + * main ERP control functions (24 and 32 byte sense) + ***************************************************************************** + */ + +/* + * DASD_3990_ERP_CONTROL_CHECK + * + * DESCRIPTION + * Does a generic inspection if a control check occurred and sets up + * the related error recovery procedure + * + * PARAMETER + * erp pointer to the currently created default ERP + * + * RETURN VALUES + * erp_filled pointer to the erp + */ + +static struct dasd_ccw_req * +dasd_3990_erp_control_check(struct dasd_ccw_req *erp) +{ + struct dasd_device *device = erp->startdev; + + if (scsw_cstat(&erp->refers->irb.scsw) & (SCHN_STAT_INTF_CTRL_CHK + | SCHN_STAT_CHN_CTRL_CHK)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "channel or interface control check"); + erp = dasd_3990_erp_action_4(erp, NULL); + } + return erp; +} + +/* + * DASD_3990_ERP_INSPECT + * + * DESCRIPTION + * Does a detailed inspection for sense data by calling either + * the 24-byte or the 32-byte inspection routine. + * + * PARAMETER + * erp pointer to the currently created default ERP + * RETURN VALUES + * erp_new contens was possibly modified + */ +static struct dasd_ccw_req * +dasd_3990_erp_inspect(struct dasd_ccw_req *erp) +{ + + struct dasd_ccw_req *erp_new = NULL; + char *sense; + + /* if this problem occurred on an alias retry on base */ + erp_new = dasd_3990_erp_inspect_alias(erp); + if (erp_new) + return erp_new; + + /* sense data are located in the refers record of the + * already set up new ERP ! + * check if concurrent sens is available + */ + sense = dasd_get_sense(&erp->refers->irb); + if (!sense) + erp_new = dasd_3990_erp_control_check(erp); + /* distinguish between 24 and 32 byte sense data */ + else if (sense[27] & DASD_SENSE_BIT_0) { + + /* inspect the 24 byte sense data */ + erp_new = dasd_3990_erp_inspect_24(erp, sense); + + } else { + + /* inspect the 32 byte sense data */ + erp_new = dasd_3990_erp_inspect_32(erp, sense); + + } /* end distinguish between 24 and 32 byte sense data */ + + return erp_new; +} + +/* + * DASD_3990_ERP_ADD_ERP + * + * DESCRIPTION + * This function adds an additional request block (ERP) to the head of + * the given cqr (or erp). + * For a command mode cqr the erp is initialized as an default erp + * (retry TIC). + * For transport mode we make a copy of the original TCW (points to + * the original TCCB, TIDALs, etc.) but give it a fresh + * TSB so the original sense data will not be changed. + * + * PARAMETER + * cqr head of the current ERP-chain (or single cqr if + * first error) + * RETURN VALUES + * erp pointer to new ERP-chain head + */ +static struct dasd_ccw_req *dasd_3990_erp_add_erp(struct dasd_ccw_req *cqr) +{ + + struct dasd_device *device = cqr->startdev; + struct ccw1 *ccw; + struct dasd_ccw_req *erp; + int cplength, datasize; + struct tcw *tcw; + struct tsb *tsb; + + if (cqr->cpmode == 1) { + cplength = 0; + /* TCW needs to be 64 byte aligned, so leave enough room */ + datasize = 64 + sizeof(struct tcw) + sizeof(struct tsb); + } else { + cplength = 2; + datasize = 0; + } + + /* allocate additional request block */ + erp = dasd_alloc_erp_request((char *) &cqr->magic, + cplength, datasize, device); + if (IS_ERR(erp)) { + if (cqr->retries <= 0) { + DBF_DEV_EVENT(DBF_ERR, device, "%s", + "Unable to allocate ERP request"); + cqr->status = DASD_CQR_FAILED; + cqr->stopclk = get_tod_clock(); + } else { + DBF_DEV_EVENT(DBF_ERR, device, + "Unable to allocate ERP request " + "(%i retries left)", + cqr->retries); + dasd_block_set_timer(device->block, (HZ << 3)); + } + return erp; + } + + ccw = cqr->cpaddr; + if (cqr->cpmode == 1) { + /* make a shallow copy of the original tcw but set new tsb */ + erp->cpmode = 1; + erp->cpaddr = PTR_ALIGN(erp->data, 64); + tcw = erp->cpaddr; + tsb = (struct tsb *) &tcw[1]; + *tcw = *((struct tcw *)cqr->cpaddr); + tcw->tsb = (long)tsb; + } else if (ccw->cmd_code == DASD_ECKD_CCW_PSF) { + /* PSF cannot be chained from NOOP/TIC */ + erp->cpaddr = cqr->cpaddr; + } else { + /* initialize request with default TIC to current ERP/CQR */ + ccw = erp->cpaddr; + ccw->cmd_code = CCW_CMD_NOOP; + ccw->flags = CCW_FLAG_CC; + ccw++; + ccw->cmd_code = CCW_CMD_TIC; + ccw->cda = (long)(cqr->cpaddr); + } + + erp->flags = cqr->flags; + erp->function = dasd_3990_erp_add_erp; + erp->refers = cqr; + erp->startdev = device; + erp->memdev = device; + erp->block = cqr->block; + erp->magic = cqr->magic; + erp->expires = cqr->expires; + erp->retries = 256; + erp->buildclk = get_tod_clock(); + erp->status = DASD_CQR_FILLED; + + return erp; +} + +/* + * DASD_3990_ERP_ADDITIONAL_ERP + * + * DESCRIPTION + * An additional ERP is needed to handle the current error. + * Add ERP to the head of the ERP-chain containing the ERP processing + * determined based on the sense data. + * + * PARAMETER + * cqr head of the current ERP-chain (or single cqr if + * first error) + * + * RETURN VALUES + * erp pointer to new ERP-chain head + */ +static struct dasd_ccw_req * +dasd_3990_erp_additional_erp(struct dasd_ccw_req * cqr) +{ + + struct dasd_ccw_req *erp = NULL; + + /* add erp and initialize with default TIC */ + erp = dasd_3990_erp_add_erp(cqr); + + if (IS_ERR(erp)) + return erp; + + /* inspect sense, determine specific ERP if possible */ + if (erp != cqr) { + + erp = dasd_3990_erp_inspect(erp); + } + + return erp; + +} /* end dasd_3990_erp_additional_erp */ + +/* + * DASD_3990_ERP_ERROR_MATCH + * + * DESCRIPTION + * Check if the device status of the given cqr is the same. + * This means that the failed CCW and the relevant sense data + * must match. + * I don't distinguish between 24 and 32 byte sense because in case of + * 24 byte sense byte 25 and 27 is set as well. + * + * PARAMETER + * cqr1 first cqr, which will be compared with the + * cqr2 second cqr. + * + * RETURN VALUES + * match 'boolean' for match found + * returns 1 if match found, otherwise 0. + */ +static int dasd_3990_erp_error_match(struct dasd_ccw_req *cqr1, + struct dasd_ccw_req *cqr2) +{ + char *sense1, *sense2; + + if (cqr1->startdev != cqr2->startdev) + return 0; + + sense1 = dasd_get_sense(&cqr1->irb); + sense2 = dasd_get_sense(&cqr2->irb); + + /* one request has sense data, the other not -> no match, return 0 */ + if (!sense1 != !sense2) + return 0; + /* no sense data in both cases -> check cstat for IFCC */ + if (!sense1 && !sense2) { + if ((scsw_cstat(&cqr1->irb.scsw) & (SCHN_STAT_INTF_CTRL_CHK | + SCHN_STAT_CHN_CTRL_CHK)) == + (scsw_cstat(&cqr2->irb.scsw) & (SCHN_STAT_INTF_CTRL_CHK | + SCHN_STAT_CHN_CTRL_CHK))) + return 1; /* match with ifcc*/ + } + /* check sense data; byte 0-2,25,27 */ + if (!(sense1 && sense2 && + (memcmp(sense1, sense2, 3) == 0) && + (sense1[27] == sense2[27]) && + (sense1[25] == sense2[25]))) { + + return 0; /* sense doesn't match */ + } + + return 1; /* match */ + +} /* end dasd_3990_erp_error_match */ + +/* + * DASD_3990_ERP_IN_ERP + * + * DESCRIPTION + * check if the current error already happened before. + * quick exit if current cqr is not an ERP (cqr->refers=NULL) + * + * PARAMETER + * cqr failed cqr (either original cqr or already an erp) + * + * RETURN VALUES + * erp erp-pointer to the already defined error + * recovery procedure OR + * NULL if a 'new' error occurred. + */ +static struct dasd_ccw_req * +dasd_3990_erp_in_erp(struct dasd_ccw_req *cqr) +{ + + struct dasd_ccw_req *erp_head = cqr, /* save erp chain head */ + *erp_match = NULL; /* save erp chain head */ + int match = 0; /* 'boolean' for matching error found */ + + if (cqr->refers == NULL) { /* return if not in erp */ + return NULL; + } + + /* check the erp/cqr chain for current error */ + do { + match = dasd_3990_erp_error_match(erp_head, cqr->refers); + erp_match = cqr; /* save possible matching erp */ + cqr = cqr->refers; /* check next erp/cqr in queue */ + + } while ((cqr->refers != NULL) && (!match)); + + if (!match) { + return NULL; /* no match was found */ + } + + return erp_match; /* return address of matching erp */ + +} /* END dasd_3990_erp_in_erp */ + +/* + * DASD_3990_ERP_FURTHER_ERP (24 & 32 byte sense) + * + * DESCRIPTION + * No retry is left for the current ERP. Check what has to be done + * with the ERP. + * - do further defined ERP action or + * - wait for interrupt or + * - exit with permanent error + * + * PARAMETER + * erp ERP which is in progress with no retry left + * + * RETURN VALUES + * erp modified/additional ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_further_erp(struct dasd_ccw_req *erp) +{ + + struct dasd_device *device = erp->startdev; + char *sense = dasd_get_sense(&erp->irb); + + /* check for 24 byte sense ERP */ + if ((erp->function == dasd_3990_erp_bus_out) || + (erp->function == dasd_3990_erp_action_1) || + (erp->function == dasd_3990_erp_action_4)) { + + erp = dasd_3990_erp_action_1(erp); + + } else if (erp->function == dasd_3990_erp_action_1_sec) { + erp = dasd_3990_erp_action_1_sec(erp); + } else if (erp->function == dasd_3990_erp_action_5) { + + /* retries have not been successful */ + /* prepare erp for retry on different channel path */ + erp = dasd_3990_erp_action_1(erp); + + if (sense && !(sense[2] & DASD_SENSE_BIT_0)) { + + /* issue a Diagnostic Control command with an + * Inhibit Write subcommand */ + + switch (sense[25]) { + case 0x17: + case 0x57:{ /* controller */ + erp = dasd_3990_erp_DCTL(erp, 0x20); + break; + } + case 0x18: + case 0x58:{ /* channel path */ + erp = dasd_3990_erp_DCTL(erp, 0x40); + break; + } + case 0x19: + case 0x59:{ /* storage director */ + erp = dasd_3990_erp_DCTL(erp, 0x80); + break; + } + default: + DBF_DEV_EVENT(DBF_WARNING, device, + "invalid subcommand modifier 0x%x " + "for Diagnostic Control Command", + sense[25]); + } + } + + /* check for 32 byte sense ERP */ + } else if (sense && + ((erp->function == dasd_3990_erp_compound_retry) || + (erp->function == dasd_3990_erp_compound_path) || + (erp->function == dasd_3990_erp_compound_code) || + (erp->function == dasd_3990_erp_compound_config))) { + + erp = dasd_3990_erp_compound(erp, sense); + + } else { + /* + * No retry left and no additional special handling + * necessary + */ + dev_err(&device->cdev->dev, + "ERP %p has run out of retries and failed\n", erp); + + erp->status = DASD_CQR_FAILED; + } + + return erp; + +} /* end dasd_3990_erp_further_erp */ + +/* + * DASD_3990_ERP_HANDLE_MATCH_ERP + * + * DESCRIPTION + * An error occurred again and an ERP has been detected which is already + * used to handle this error (e.g. retries). + * All prior ERP's are asumed to be successful and therefore removed + * from queue. + * If retry counter of matching erp is already 0, it is checked if further + * action is needed (besides retry) or if the ERP has failed. + * + * PARAMETER + * erp_head first ERP in ERP-chain + * erp ERP that handles the actual error. + * (matching erp) + * + * RETURN VALUES + * erp modified/additional ERP + */ +static struct dasd_ccw_req * +dasd_3990_erp_handle_match_erp(struct dasd_ccw_req *erp_head, + struct dasd_ccw_req *erp) +{ + + struct dasd_device *device = erp_head->startdev; + struct dasd_ccw_req *erp_done = erp_head; /* finished req */ + struct dasd_ccw_req *erp_free = NULL; /* req to be freed */ + + /* loop over successful ERPs and remove them from chanq */ + while (erp_done != erp) { + + if (erp_done == NULL) /* end of chain reached */ + panic(PRINTK_HEADER "Programming error in ERP! The " + "original request was lost\n"); + + /* remove the request from the device queue */ + list_del(&erp_done->blocklist); + + erp_free = erp_done; + erp_done = erp_done->refers; + + /* free the finished erp request */ + dasd_free_erp_request(erp_free, erp_free->memdev); + + } /* end while */ + + if (erp->retries > 0) { + + char *sense = dasd_get_sense(&erp->refers->irb); + + /* check for special retries */ + if (sense && erp->function == dasd_3990_erp_action_4) { + + erp = dasd_3990_erp_action_4(erp, sense); + + } else if (sense && + erp->function == dasd_3990_erp_action_1B_32) { + + erp = dasd_3990_update_1B(erp, sense); + + } else if (sense && erp->function == dasd_3990_erp_int_req) { + + erp = dasd_3990_erp_int_req(erp); + + } else { + /* simple retry */ + DBF_DEV_EVENT(DBF_DEBUG, device, + "%i retries left for erp %p", + erp->retries, erp); + + /* handle the request again... */ + erp->status = DASD_CQR_FILLED; + } + + } else { + /* no retry left - check for further necessary action */ + /* if no further actions, handle rest as permanent error */ + erp = dasd_3990_erp_further_erp(erp); + } + + return erp; + +} /* end dasd_3990_erp_handle_match_erp */ + +/* + * DASD_3990_ERP_ACTION + * + * DESCRIPTION + * control routine for 3990 erp actions. + * Has to be called with the queue lock (namely the s390_irq_lock) acquired. + * + * PARAMETER + * cqr failed cqr (either original cqr or already an erp) + * + * RETURN VALUES + * erp erp-pointer to the head of the ERP action chain. + * This means: + * - either a ptr to an additional ERP cqr or + * - the original given cqr (which's status might + * be modified) + */ +struct dasd_ccw_req * +dasd_3990_erp_action(struct dasd_ccw_req * cqr) +{ + struct dasd_ccw_req *erp = NULL; + struct dasd_device *device = cqr->startdev; + struct dasd_ccw_req *temp_erp = NULL; + + if (device->features & DASD_FEATURE_ERPLOG) { + /* print current erp_chain */ + dev_err(&device->cdev->dev, + "ERP chain at BEGINNING of ERP-ACTION\n"); + for (temp_erp = cqr; + temp_erp != NULL; temp_erp = temp_erp->refers) { + + dev_err(&device->cdev->dev, + "ERP %p (%02x) refers to %p\n", + temp_erp, temp_erp->status, + temp_erp->refers); + } + } + + /* double-check if current erp/cqr was successful */ + if ((scsw_cstat(&cqr->irb.scsw) == 0x00) && + (scsw_dstat(&cqr->irb.scsw) == + (DEV_STAT_CHN_END | DEV_STAT_DEV_END))) { + + DBF_DEV_EVENT(DBF_DEBUG, device, + "ERP called for successful request %p" + " - NO ERP necessary", cqr); + + cqr->status = DASD_CQR_DONE; + + return cqr; + } + + /* check if error happened before */ + erp = dasd_3990_erp_in_erp(cqr); + + if (erp == NULL) { + /* no matching erp found - set up erp */ + erp = dasd_3990_erp_additional_erp(cqr); + if (IS_ERR(erp)) + return erp; + } else { + /* matching erp found - set all leading erp's to DONE */ + erp = dasd_3990_erp_handle_match_erp(cqr, erp); + } + + if (device->features & DASD_FEATURE_ERPLOG) { + /* print current erp_chain */ + dev_err(&device->cdev->dev, + "ERP chain at END of ERP-ACTION\n"); + for (temp_erp = erp; + temp_erp != NULL; temp_erp = temp_erp->refers) { + + dev_err(&device->cdev->dev, + "ERP %p (%02x) refers to %p\n", + temp_erp, temp_erp->status, + temp_erp->refers); + } + } + + /* enqueue ERP request if it's a new one */ + if (list_empty(&erp->blocklist)) { + cqr->status = DASD_CQR_IN_ERP; + /* add erp request before the cqr */ + list_add_tail(&erp->blocklist, &cqr->blocklist); + } + + + + return erp; + +} /* end dasd_3990_erp_action */ diff --git a/kernel/drivers/s390/block/dasd_alias.c b/kernel/drivers/s390/block/dasd_alias.c new file mode 100644 index 000000000..a2597e683 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_alias.c @@ -0,0 +1,990 @@ +/* + * PAV alias management for the DASD ECKD discipline + * + * Copyright IBM Corp. 2007 + * Author(s): Stefan Weinhuber <wein@de.ibm.com> + */ + +#define KMSG_COMPONENT "dasd-eckd" + +#include <linux/list.h> +#include <linux/slab.h> +#include <asm/ebcdic.h> +#include "dasd_int.h" +#include "dasd_eckd.h" + +#ifdef PRINTK_HEADER +#undef PRINTK_HEADER +#endif /* PRINTK_HEADER */ +#define PRINTK_HEADER "dasd(eckd):" + + +/* + * General concept of alias management: + * - PAV and DASD alias management is specific to the eckd discipline. + * - A device is connected to an lcu as long as the device exists. + * dasd_alias_make_device_known_to_lcu will be called wenn the + * device is checked by the eckd discipline and + * dasd_alias_disconnect_device_from_lcu will be called + * before the device is deleted. + * - The dasd_alias_add_device / dasd_alias_remove_device + * functions mark the point when a device is 'ready for service'. + * - A summary unit check is a rare occasion, but it is mandatory to + * support it. It requires some complex recovery actions before the + * devices can be used again (see dasd_alias_handle_summary_unit_check). + * - dasd_alias_get_start_dev will find an alias device that can be used + * instead of the base device and does some (very simple) load balancing. + * This is the function that gets called for each I/O, so when improving + * something, this function should get faster or better, the rest has just + * to be correct. + */ + + +static void summary_unit_check_handling_work(struct work_struct *); +static void lcu_update_work(struct work_struct *); +static int _schedule_lcu_update(struct alias_lcu *, struct dasd_device *); + +static struct alias_root aliastree = { + .serverlist = LIST_HEAD_INIT(aliastree.serverlist), + .lock = __SPIN_LOCK_UNLOCKED(aliastree.lock), +}; + +static struct alias_server *_find_server(struct dasd_uid *uid) +{ + struct alias_server *pos; + list_for_each_entry(pos, &aliastree.serverlist, server) { + if (!strncmp(pos->uid.vendor, uid->vendor, + sizeof(uid->vendor)) + && !strncmp(pos->uid.serial, uid->serial, + sizeof(uid->serial))) + return pos; + }; + return NULL; +} + +static struct alias_lcu *_find_lcu(struct alias_server *server, + struct dasd_uid *uid) +{ + struct alias_lcu *pos; + list_for_each_entry(pos, &server->lculist, lcu) { + if (pos->uid.ssid == uid->ssid) + return pos; + }; + return NULL; +} + +static struct alias_pav_group *_find_group(struct alias_lcu *lcu, + struct dasd_uid *uid) +{ + struct alias_pav_group *pos; + __u8 search_unit_addr; + + /* for hyper pav there is only one group */ + if (lcu->pav == HYPER_PAV) { + if (list_empty(&lcu->grouplist)) + return NULL; + else + return list_first_entry(&lcu->grouplist, + struct alias_pav_group, group); + } + + /* for base pav we have to find the group that matches the base */ + if (uid->type == UA_BASE_DEVICE) + search_unit_addr = uid->real_unit_addr; + else + search_unit_addr = uid->base_unit_addr; + list_for_each_entry(pos, &lcu->grouplist, group) { + if (pos->uid.base_unit_addr == search_unit_addr && + !strncmp(pos->uid.vduit, uid->vduit, sizeof(uid->vduit))) + return pos; + }; + return NULL; +} + +static struct alias_server *_allocate_server(struct dasd_uid *uid) +{ + struct alias_server *server; + + server = kzalloc(sizeof(*server), GFP_KERNEL); + if (!server) + return ERR_PTR(-ENOMEM); + memcpy(server->uid.vendor, uid->vendor, sizeof(uid->vendor)); + memcpy(server->uid.serial, uid->serial, sizeof(uid->serial)); + INIT_LIST_HEAD(&server->server); + INIT_LIST_HEAD(&server->lculist); + return server; +} + +static void _free_server(struct alias_server *server) +{ + kfree(server); +} + +static struct alias_lcu *_allocate_lcu(struct dasd_uid *uid) +{ + struct alias_lcu *lcu; + + lcu = kzalloc(sizeof(*lcu), GFP_KERNEL); + if (!lcu) + return ERR_PTR(-ENOMEM); + lcu->uac = kzalloc(sizeof(*(lcu->uac)), GFP_KERNEL | GFP_DMA); + if (!lcu->uac) + goto out_err1; + lcu->rsu_cqr = kzalloc(sizeof(*lcu->rsu_cqr), GFP_KERNEL | GFP_DMA); + if (!lcu->rsu_cqr) + goto out_err2; + lcu->rsu_cqr->cpaddr = kzalloc(sizeof(struct ccw1), + GFP_KERNEL | GFP_DMA); + if (!lcu->rsu_cqr->cpaddr) + goto out_err3; + lcu->rsu_cqr->data = kzalloc(16, GFP_KERNEL | GFP_DMA); + if (!lcu->rsu_cqr->data) + goto out_err4; + + memcpy(lcu->uid.vendor, uid->vendor, sizeof(uid->vendor)); + memcpy(lcu->uid.serial, uid->serial, sizeof(uid->serial)); + lcu->uid.ssid = uid->ssid; + lcu->pav = NO_PAV; + lcu->flags = NEED_UAC_UPDATE | UPDATE_PENDING; + INIT_LIST_HEAD(&lcu->lcu); + INIT_LIST_HEAD(&lcu->inactive_devices); + INIT_LIST_HEAD(&lcu->active_devices); + INIT_LIST_HEAD(&lcu->grouplist); + INIT_WORK(&lcu->suc_data.worker, summary_unit_check_handling_work); + INIT_DELAYED_WORK(&lcu->ruac_data.dwork, lcu_update_work); + spin_lock_init(&lcu->lock); + init_completion(&lcu->lcu_setup); + return lcu; + +out_err4: + kfree(lcu->rsu_cqr->cpaddr); +out_err3: + kfree(lcu->rsu_cqr); +out_err2: + kfree(lcu->uac); +out_err1: + kfree(lcu); + return ERR_PTR(-ENOMEM); +} + +static void _free_lcu(struct alias_lcu *lcu) +{ + kfree(lcu->rsu_cqr->data); + kfree(lcu->rsu_cqr->cpaddr); + kfree(lcu->rsu_cqr); + kfree(lcu->uac); + kfree(lcu); +} + +/* + * This is the function that will allocate all the server and lcu data, + * so this function must be called first for a new device. + * If the return value is 1, the lcu was already known before, if it + * is 0, this is a new lcu. + * Negative return code indicates that something went wrong (e.g. -ENOMEM) + */ +int dasd_alias_make_device_known_to_lcu(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + unsigned long flags; + struct alias_server *server, *newserver; + struct alias_lcu *lcu, *newlcu; + struct dasd_uid uid; + + private = (struct dasd_eckd_private *) device->private; + + device->discipline->get_uid(device, &uid); + spin_lock_irqsave(&aliastree.lock, flags); + server = _find_server(&uid); + if (!server) { + spin_unlock_irqrestore(&aliastree.lock, flags); + newserver = _allocate_server(&uid); + if (IS_ERR(newserver)) + return PTR_ERR(newserver); + spin_lock_irqsave(&aliastree.lock, flags); + server = _find_server(&uid); + if (!server) { + list_add(&newserver->server, &aliastree.serverlist); + server = newserver; + } else { + /* someone was faster */ + _free_server(newserver); + } + } + + lcu = _find_lcu(server, &uid); + if (!lcu) { + spin_unlock_irqrestore(&aliastree.lock, flags); + newlcu = _allocate_lcu(&uid); + if (IS_ERR(newlcu)) + return PTR_ERR(newlcu); + spin_lock_irqsave(&aliastree.lock, flags); + lcu = _find_lcu(server, &uid); + if (!lcu) { + list_add(&newlcu->lcu, &server->lculist); + lcu = newlcu; + } else { + /* someone was faster */ + _free_lcu(newlcu); + } + } + spin_lock(&lcu->lock); + list_add(&device->alias_list, &lcu->inactive_devices); + private->lcu = lcu; + spin_unlock(&lcu->lock); + spin_unlock_irqrestore(&aliastree.lock, flags); + + return 0; +} + +/* + * This function removes a device from the scope of alias management. + * The complicated part is to make sure that it is not in use by + * any of the workers. If necessary cancel the work. + */ +void dasd_alias_disconnect_device_from_lcu(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + unsigned long flags; + struct alias_lcu *lcu; + struct alias_server *server; + int was_pending; + struct dasd_uid uid; + + private = (struct dasd_eckd_private *) device->private; + lcu = private->lcu; + /* nothing to do if already disconnected */ + if (!lcu) + return; + device->discipline->get_uid(device, &uid); + spin_lock_irqsave(&lcu->lock, flags); + list_del_init(&device->alias_list); + /* make sure that the workers don't use this device */ + if (device == lcu->suc_data.device) { + spin_unlock_irqrestore(&lcu->lock, flags); + cancel_work_sync(&lcu->suc_data.worker); + spin_lock_irqsave(&lcu->lock, flags); + if (device == lcu->suc_data.device) + lcu->suc_data.device = NULL; + } + was_pending = 0; + if (device == lcu->ruac_data.device) { + spin_unlock_irqrestore(&lcu->lock, flags); + was_pending = 1; + cancel_delayed_work_sync(&lcu->ruac_data.dwork); + spin_lock_irqsave(&lcu->lock, flags); + if (device == lcu->ruac_data.device) + lcu->ruac_data.device = NULL; + } + private->lcu = NULL; + spin_unlock_irqrestore(&lcu->lock, flags); + + spin_lock_irqsave(&aliastree.lock, flags); + spin_lock(&lcu->lock); + if (list_empty(&lcu->grouplist) && + list_empty(&lcu->active_devices) && + list_empty(&lcu->inactive_devices)) { + list_del(&lcu->lcu); + spin_unlock(&lcu->lock); + _free_lcu(lcu); + lcu = NULL; + } else { + if (was_pending) + _schedule_lcu_update(lcu, NULL); + spin_unlock(&lcu->lock); + } + server = _find_server(&uid); + if (server && list_empty(&server->lculist)) { + list_del(&server->server); + _free_server(server); + } + spin_unlock_irqrestore(&aliastree.lock, flags); +} + +/* + * This function assumes that the unit address configuration stored + * in the lcu is up to date and will update the device uid before + * adding it to a pav group. + */ + +static int _add_device_to_lcu(struct alias_lcu *lcu, + struct dasd_device *device, + struct dasd_device *pos) +{ + + struct dasd_eckd_private *private; + struct alias_pav_group *group; + struct dasd_uid uid; + unsigned long flags; + + private = (struct dasd_eckd_private *) device->private; + + /* only lock if not already locked */ + if (device != pos) + spin_lock_irqsave_nested(get_ccwdev_lock(device->cdev), flags, + CDEV_NESTED_SECOND); + private->uid.type = lcu->uac->unit[private->uid.real_unit_addr].ua_type; + private->uid.base_unit_addr = + lcu->uac->unit[private->uid.real_unit_addr].base_ua; + uid = private->uid; + + if (device != pos) + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + + /* if we have no PAV anyway, we don't need to bother with PAV groups */ + if (lcu->pav == NO_PAV) { + list_move(&device->alias_list, &lcu->active_devices); + return 0; + } + + group = _find_group(lcu, &uid); + if (!group) { + group = kzalloc(sizeof(*group), GFP_ATOMIC); + if (!group) + return -ENOMEM; + memcpy(group->uid.vendor, uid.vendor, sizeof(uid.vendor)); + memcpy(group->uid.serial, uid.serial, sizeof(uid.serial)); + group->uid.ssid = uid.ssid; + if (uid.type == UA_BASE_DEVICE) + group->uid.base_unit_addr = uid.real_unit_addr; + else + group->uid.base_unit_addr = uid.base_unit_addr; + memcpy(group->uid.vduit, uid.vduit, sizeof(uid.vduit)); + INIT_LIST_HEAD(&group->group); + INIT_LIST_HEAD(&group->baselist); + INIT_LIST_HEAD(&group->aliaslist); + list_add(&group->group, &lcu->grouplist); + } + if (uid.type == UA_BASE_DEVICE) + list_move(&device->alias_list, &group->baselist); + else + list_move(&device->alias_list, &group->aliaslist); + private->pavgroup = group; + return 0; +}; + +static void _remove_device_from_lcu(struct alias_lcu *lcu, + struct dasd_device *device) +{ + struct dasd_eckd_private *private; + struct alias_pav_group *group; + + private = (struct dasd_eckd_private *) device->private; + list_move(&device->alias_list, &lcu->inactive_devices); + group = private->pavgroup; + if (!group) + return; + private->pavgroup = NULL; + if (list_empty(&group->baselist) && list_empty(&group->aliaslist)) { + list_del(&group->group); + kfree(group); + return; + } + if (group->next == device) + group->next = NULL; +}; + +static int +suborder_not_supported(struct dasd_ccw_req *cqr) +{ + char *sense; + char reason; + char msg_format; + char msg_no; + + sense = dasd_get_sense(&cqr->irb); + if (!sense) + return 0; + + reason = sense[0]; + msg_format = (sense[7] & 0xF0); + msg_no = (sense[7] & 0x0F); + + /* command reject, Format 0 MSG 4 - invalid parameter */ + if ((reason == 0x80) && (msg_format == 0x00) && (msg_no == 0x04)) + return 1; + + return 0; +} + +static int read_unit_address_configuration(struct dasd_device *device, + struct alias_lcu *lcu) +{ + struct dasd_psf_prssd_data *prssdp; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int rc; + unsigned long flags; + + cqr = dasd_kmalloc_request(DASD_ECKD_MAGIC, 1 /* PSF */ + 1 /* RSSD */, + (sizeof(struct dasd_psf_prssd_data)), + device); + if (IS_ERR(cqr)) + return PTR_ERR(cqr); + cqr->startdev = device; + cqr->memdev = device; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + cqr->retries = 10; + cqr->expires = 20 * HZ; + + /* Prepare for Read Subsystem Data */ + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + memset(prssdp, 0, sizeof(struct dasd_psf_prssd_data)); + prssdp->order = PSF_ORDER_PRSSD; + prssdp->suborder = 0x0e; /* Read unit address configuration */ + /* all other bytes of prssdp must be zero */ + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->count = sizeof(struct dasd_psf_prssd_data); + ccw->flags |= CCW_FLAG_CC; + ccw->cda = (__u32)(addr_t) prssdp; + + /* Read Subsystem Data - feature codes */ + memset(lcu->uac, 0, sizeof(*(lcu->uac))); + + ccw++; + ccw->cmd_code = DASD_ECKD_CCW_RSSD; + ccw->count = sizeof(*(lcu->uac)); + ccw->cda = (__u32)(addr_t) lcu->uac; + + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + /* need to unset flag here to detect race with summary unit check */ + spin_lock_irqsave(&lcu->lock, flags); + lcu->flags &= ~NEED_UAC_UPDATE; + spin_unlock_irqrestore(&lcu->lock, flags); + + do { + rc = dasd_sleep_on(cqr); + if (rc && suborder_not_supported(cqr)) + return -EOPNOTSUPP; + } while (rc && (cqr->retries > 0)); + if (rc) { + spin_lock_irqsave(&lcu->lock, flags); + lcu->flags |= NEED_UAC_UPDATE; + spin_unlock_irqrestore(&lcu->lock, flags); + } + dasd_kfree_request(cqr, cqr->memdev); + return rc; +} + +static int _lcu_update(struct dasd_device *refdev, struct alias_lcu *lcu) +{ + unsigned long flags; + struct alias_pav_group *pavgroup, *tempgroup; + struct dasd_device *device, *tempdev; + int i, rc; + struct dasd_eckd_private *private; + + spin_lock_irqsave(&lcu->lock, flags); + list_for_each_entry_safe(pavgroup, tempgroup, &lcu->grouplist, group) { + list_for_each_entry_safe(device, tempdev, &pavgroup->baselist, + alias_list) { + list_move(&device->alias_list, &lcu->active_devices); + private = (struct dasd_eckd_private *) device->private; + private->pavgroup = NULL; + } + list_for_each_entry_safe(device, tempdev, &pavgroup->aliaslist, + alias_list) { + list_move(&device->alias_list, &lcu->active_devices); + private = (struct dasd_eckd_private *) device->private; + private->pavgroup = NULL; + } + list_del(&pavgroup->group); + kfree(pavgroup); + } + spin_unlock_irqrestore(&lcu->lock, flags); + + rc = read_unit_address_configuration(refdev, lcu); + if (rc) + return rc; + + /* need to take cdev lock before lcu lock */ + spin_lock_irqsave_nested(get_ccwdev_lock(refdev->cdev), flags, + CDEV_NESTED_FIRST); + spin_lock(&lcu->lock); + lcu->pav = NO_PAV; + for (i = 0; i < MAX_DEVICES_PER_LCU; ++i) { + switch (lcu->uac->unit[i].ua_type) { + case UA_BASE_PAV_ALIAS: + lcu->pav = BASE_PAV; + break; + case UA_HYPER_PAV_ALIAS: + lcu->pav = HYPER_PAV; + break; + } + if (lcu->pav != NO_PAV) + break; + } + + list_for_each_entry_safe(device, tempdev, &lcu->active_devices, + alias_list) { + _add_device_to_lcu(lcu, device, refdev); + } + spin_unlock(&lcu->lock); + spin_unlock_irqrestore(get_ccwdev_lock(refdev->cdev), flags); + return 0; +} + +static void lcu_update_work(struct work_struct *work) +{ + struct alias_lcu *lcu; + struct read_uac_work_data *ruac_data; + struct dasd_device *device; + unsigned long flags; + int rc; + + ruac_data = container_of(work, struct read_uac_work_data, dwork.work); + lcu = container_of(ruac_data, struct alias_lcu, ruac_data); + device = ruac_data->device; + rc = _lcu_update(device, lcu); + /* + * Need to check flags again, as there could have been another + * prepare_update or a new device a new device while we were still + * processing the data + */ + spin_lock_irqsave(&lcu->lock, flags); + if ((rc && (rc != -EOPNOTSUPP)) || (lcu->flags & NEED_UAC_UPDATE)) { + DBF_DEV_EVENT(DBF_WARNING, device, "could not update" + " alias data in lcu (rc = %d), retry later", rc); + schedule_delayed_work(&lcu->ruac_data.dwork, 30*HZ); + } else { + lcu->ruac_data.device = NULL; + lcu->flags &= ~UPDATE_PENDING; + } + spin_unlock_irqrestore(&lcu->lock, flags); +} + +static int _schedule_lcu_update(struct alias_lcu *lcu, + struct dasd_device *device) +{ + struct dasd_device *usedev = NULL; + struct alias_pav_group *group; + + lcu->flags |= NEED_UAC_UPDATE; + if (lcu->ruac_data.device) { + /* already scheduled or running */ + return 0; + } + if (device && !list_empty(&device->alias_list)) + usedev = device; + + if (!usedev && !list_empty(&lcu->grouplist)) { + group = list_first_entry(&lcu->grouplist, + struct alias_pav_group, group); + if (!list_empty(&group->baselist)) + usedev = list_first_entry(&group->baselist, + struct dasd_device, + alias_list); + else if (!list_empty(&group->aliaslist)) + usedev = list_first_entry(&group->aliaslist, + struct dasd_device, + alias_list); + } + if (!usedev && !list_empty(&lcu->active_devices)) { + usedev = list_first_entry(&lcu->active_devices, + struct dasd_device, alias_list); + } + /* + * if we haven't found a proper device yet, give up for now, the next + * device that will be set active will trigger an lcu update + */ + if (!usedev) + return -EINVAL; + lcu->ruac_data.device = usedev; + schedule_delayed_work(&lcu->ruac_data.dwork, 0); + return 0; +} + +int dasd_alias_add_device(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + struct alias_lcu *lcu; + unsigned long flags; + int rc; + + private = (struct dasd_eckd_private *) device->private; + lcu = private->lcu; + rc = 0; + + /* need to take cdev lock before lcu lock */ + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + spin_lock(&lcu->lock); + if (!(lcu->flags & UPDATE_PENDING)) { + rc = _add_device_to_lcu(lcu, device, device); + if (rc) + lcu->flags |= UPDATE_PENDING; + } + if (lcu->flags & UPDATE_PENDING) { + list_move(&device->alias_list, &lcu->active_devices); + _schedule_lcu_update(lcu, device); + } + spin_unlock(&lcu->lock); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + return rc; +} + +int dasd_alias_update_add_device(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + private = (struct dasd_eckd_private *) device->private; + private->lcu->flags |= UPDATE_PENDING; + return dasd_alias_add_device(device); +} + +int dasd_alias_remove_device(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + struct alias_lcu *lcu; + unsigned long flags; + + private = (struct dasd_eckd_private *) device->private; + lcu = private->lcu; + /* nothing to do if already removed */ + if (!lcu) + return 0; + spin_lock_irqsave(&lcu->lock, flags); + _remove_device_from_lcu(lcu, device); + spin_unlock_irqrestore(&lcu->lock, flags); + return 0; +} + +struct dasd_device *dasd_alias_get_start_dev(struct dasd_device *base_device) +{ + + struct dasd_device *alias_device; + struct alias_pav_group *group; + struct alias_lcu *lcu; + struct dasd_eckd_private *private, *alias_priv; + unsigned long flags; + + private = (struct dasd_eckd_private *) base_device->private; + group = private->pavgroup; + lcu = private->lcu; + if (!group || !lcu) + return NULL; + if (lcu->pav == NO_PAV || + lcu->flags & (NEED_UAC_UPDATE | UPDATE_PENDING)) + return NULL; + if (unlikely(!(private->features.feature[8] & 0x01))) { + /* + * PAV enabled but prefix not, very unlikely + * seems to be a lost pathgroup + * use base device to do IO + */ + DBF_DEV_EVENT(DBF_ERR, base_device, "%s", + "Prefix not enabled with PAV enabled\n"); + return NULL; + } + + spin_lock_irqsave(&lcu->lock, flags); + alias_device = group->next; + if (!alias_device) { + if (list_empty(&group->aliaslist)) { + spin_unlock_irqrestore(&lcu->lock, flags); + return NULL; + } else { + alias_device = list_first_entry(&group->aliaslist, + struct dasd_device, + alias_list); + } + } + if (list_is_last(&alias_device->alias_list, &group->aliaslist)) + group->next = list_first_entry(&group->aliaslist, + struct dasd_device, alias_list); + else + group->next = list_first_entry(&alias_device->alias_list, + struct dasd_device, alias_list); + spin_unlock_irqrestore(&lcu->lock, flags); + alias_priv = (struct dasd_eckd_private *) alias_device->private; + if ((alias_priv->count < private->count) && !alias_device->stopped) + return alias_device; + else + return NULL; +} + +/* + * Summary unit check handling depends on the way alias devices + * are handled so it is done here rather then in dasd_eckd.c + */ +static int reset_summary_unit_check(struct alias_lcu *lcu, + struct dasd_device *device, + char reason) +{ + struct dasd_ccw_req *cqr; + int rc = 0; + struct ccw1 *ccw; + + cqr = lcu->rsu_cqr; + strncpy((char *) &cqr->magic, "ECKD", 4); + ASCEBC((char *) &cqr->magic, 4); + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_RSCK; + ccw->flags = 0 ; + ccw->count = 16; + ccw->cda = (__u32)(addr_t) cqr->data; + ((char *)cqr->data)[0] = reason; + + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + cqr->retries = 255; /* set retry counter to enable basic ERP */ + cqr->startdev = device; + cqr->memdev = device; + cqr->block = NULL; + cqr->expires = 5 * HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + rc = dasd_sleep_on_immediatly(cqr); + return rc; +} + +static void _restart_all_base_devices_on_lcu(struct alias_lcu *lcu) +{ + struct alias_pav_group *pavgroup; + struct dasd_device *device; + struct dasd_eckd_private *private; + unsigned long flags; + + /* active and inactive list can contain alias as well as base devices */ + list_for_each_entry(device, &lcu->active_devices, alias_list) { + private = (struct dasd_eckd_private *) device->private; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + if (private->uid.type != UA_BASE_DEVICE) { + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), + flags); + continue; + } + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + dasd_schedule_block_bh(device->block); + dasd_schedule_device_bh(device); + } + list_for_each_entry(device, &lcu->inactive_devices, alias_list) { + private = (struct dasd_eckd_private *) device->private; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + if (private->uid.type != UA_BASE_DEVICE) { + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), + flags); + continue; + } + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + dasd_schedule_block_bh(device->block); + dasd_schedule_device_bh(device); + } + list_for_each_entry(pavgroup, &lcu->grouplist, group) { + list_for_each_entry(device, &pavgroup->baselist, alias_list) { + dasd_schedule_block_bh(device->block); + dasd_schedule_device_bh(device); + } + } +} + +static void flush_all_alias_devices_on_lcu(struct alias_lcu *lcu) +{ + struct alias_pav_group *pavgroup; + struct dasd_device *device, *temp; + struct dasd_eckd_private *private; + int rc; + unsigned long flags; + LIST_HEAD(active); + + /* + * Problem here ist that dasd_flush_device_queue may wait + * for termination of a request to complete. We can't keep + * the lcu lock during that time, so we must assume that + * the lists may have changed. + * Idea: first gather all active alias devices in a separate list, + * then flush the first element of this list unlocked, and afterwards + * check if it is still on the list before moving it to the + * active_devices list. + */ + + spin_lock_irqsave(&lcu->lock, flags); + list_for_each_entry_safe(device, temp, &lcu->active_devices, + alias_list) { + private = (struct dasd_eckd_private *) device->private; + if (private->uid.type == UA_BASE_DEVICE) + continue; + list_move(&device->alias_list, &active); + } + + list_for_each_entry(pavgroup, &lcu->grouplist, group) { + list_splice_init(&pavgroup->aliaslist, &active); + } + while (!list_empty(&active)) { + device = list_first_entry(&active, struct dasd_device, + alias_list); + spin_unlock_irqrestore(&lcu->lock, flags); + rc = dasd_flush_device_queue(device); + spin_lock_irqsave(&lcu->lock, flags); + /* + * only move device around if it wasn't moved away while we + * were waiting for the flush + */ + if (device == list_first_entry(&active, + struct dasd_device, alias_list)) + list_move(&device->alias_list, &lcu->active_devices); + } + spin_unlock_irqrestore(&lcu->lock, flags); +} + +static void __stop_device_on_lcu(struct dasd_device *device, + struct dasd_device *pos) +{ + /* If pos == device then device is already locked! */ + if (pos == device) { + dasd_device_set_stop_bits(pos, DASD_STOPPED_SU); + return; + } + spin_lock(get_ccwdev_lock(pos->cdev)); + dasd_device_set_stop_bits(pos, DASD_STOPPED_SU); + spin_unlock(get_ccwdev_lock(pos->cdev)); +} + +/* + * This function is called in interrupt context, so the + * cdev lock for device is already locked! + */ +static void _stop_all_devices_on_lcu(struct alias_lcu *lcu, + struct dasd_device *device) +{ + struct alias_pav_group *pavgroup; + struct dasd_device *pos; + + list_for_each_entry(pos, &lcu->active_devices, alias_list) + __stop_device_on_lcu(device, pos); + list_for_each_entry(pos, &lcu->inactive_devices, alias_list) + __stop_device_on_lcu(device, pos); + list_for_each_entry(pavgroup, &lcu->grouplist, group) { + list_for_each_entry(pos, &pavgroup->baselist, alias_list) + __stop_device_on_lcu(device, pos); + list_for_each_entry(pos, &pavgroup->aliaslist, alias_list) + __stop_device_on_lcu(device, pos); + } +} + +static void _unstop_all_devices_on_lcu(struct alias_lcu *lcu) +{ + struct alias_pav_group *pavgroup; + struct dasd_device *device; + unsigned long flags; + + list_for_each_entry(device, &lcu->active_devices, alias_list) { + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dasd_device_remove_stop_bits(device, DASD_STOPPED_SU); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + } + + list_for_each_entry(device, &lcu->inactive_devices, alias_list) { + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dasd_device_remove_stop_bits(device, DASD_STOPPED_SU); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + } + + list_for_each_entry(pavgroup, &lcu->grouplist, group) { + list_for_each_entry(device, &pavgroup->baselist, alias_list) { + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dasd_device_remove_stop_bits(device, DASD_STOPPED_SU); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), + flags); + } + list_for_each_entry(device, &pavgroup->aliaslist, alias_list) { + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dasd_device_remove_stop_bits(device, DASD_STOPPED_SU); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), + flags); + } + } +} + +static void summary_unit_check_handling_work(struct work_struct *work) +{ + struct alias_lcu *lcu; + struct summary_unit_check_work_data *suc_data; + unsigned long flags; + struct dasd_device *device; + + suc_data = container_of(work, struct summary_unit_check_work_data, + worker); + lcu = container_of(suc_data, struct alias_lcu, suc_data); + device = suc_data->device; + + /* 1. flush alias devices */ + flush_all_alias_devices_on_lcu(lcu); + + /* 2. reset summary unit check */ + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dasd_device_remove_stop_bits(device, + (DASD_STOPPED_SU | DASD_STOPPED_PENDING)); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + reset_summary_unit_check(lcu, device, suc_data->reason); + + spin_lock_irqsave(&lcu->lock, flags); + _unstop_all_devices_on_lcu(lcu); + _restart_all_base_devices_on_lcu(lcu); + /* 3. read new alias configuration */ + _schedule_lcu_update(lcu, device); + lcu->suc_data.device = NULL; + spin_unlock_irqrestore(&lcu->lock, flags); +} + +/* + * note: this will be called from int handler context (cdev locked) + */ +void dasd_alias_handle_summary_unit_check(struct dasd_device *device, + struct irb *irb) +{ + struct alias_lcu *lcu; + char reason; + struct dasd_eckd_private *private; + char *sense; + + private = (struct dasd_eckd_private *) device->private; + + sense = dasd_get_sense(irb); + if (sense) { + reason = sense[8]; + DBF_DEV_EVENT(DBF_NOTICE, device, "%s %x", + "eckd handle summary unit check: reason", reason); + } else { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "eckd handle summary unit check:" + " no reason code available"); + return; + } + + lcu = private->lcu; + if (!lcu) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "device not ready to handle summary" + " unit check (no lcu structure)"); + return; + } + spin_lock(&lcu->lock); + _stop_all_devices_on_lcu(lcu, device); + /* prepare for lcu_update */ + private->lcu->flags |= NEED_UAC_UPDATE | UPDATE_PENDING; + /* If this device is about to be removed just return and wait for + * the next interrupt on a different device + */ + if (list_empty(&device->alias_list)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "device is in offline processing," + " don't do summary unit check handling"); + spin_unlock(&lcu->lock); + return; + } + if (lcu->suc_data.device) { + /* already scheduled or running */ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "previous instance of summary unit check worker" + " still pending"); + spin_unlock(&lcu->lock); + return ; + } + lcu->suc_data.reason = reason; + lcu->suc_data.device = device; + spin_unlock(&lcu->lock); + schedule_work(&lcu->suc_data.worker); +}; diff --git a/kernel/drivers/s390/block/dasd_devmap.c b/kernel/drivers/s390/block/dasd_devmap.c new file mode 100644 index 000000000..8286f7424 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_devmap.c @@ -0,0 +1,1553 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999,2001 + * + * Device mapping and dasd= parameter parsing functions. All devmap + * functions may not be called from interrupt context. In particular + * dasd_get_device is a no-no from interrupt context. + * + */ + +#define KMSG_COMPONENT "dasd" + +#include <linux/ctype.h> +#include <linux/init.h> +#include <linux/module.h> +#include <linux/slab.h> + +#include <asm/debug.h> +#include <asm/uaccess.h> +#include <asm/ipl.h> + +/* This is ugly... */ +#define PRINTK_HEADER "dasd_devmap:" +#define DASD_BUS_ID_SIZE 20 + +#include "dasd_int.h" + +struct kmem_cache *dasd_page_cache; +EXPORT_SYMBOL_GPL(dasd_page_cache); + +/* + * dasd_devmap_t is used to store the features and the relation + * between device number and device index. To find a dasd_devmap_t + * that corresponds to a device number of a device index each + * dasd_devmap_t is added to two linked lists, one to search by + * the device number and one to search by the device index. As + * soon as big minor numbers are available the device index list + * can be removed since the device number will then be identical + * to the device index. + */ +struct dasd_devmap { + struct list_head list; + char bus_id[DASD_BUS_ID_SIZE]; + unsigned int devindex; + unsigned short features; + struct dasd_device *device; +}; + +/* + * Parameter parsing functions for dasd= parameter. The syntax is: + * <devno> : (0x)?[0-9a-fA-F]+ + * <busid> : [0-0a-f]\.[0-9a-f]\.(0x)?[0-9a-fA-F]+ + * <feature> : ro + * <feature_list> : \(<feature>(:<feature>)*\) + * <devno-range> : <devno>(-<devno>)?<feature_list>? + * <busid-range> : <busid>(-<busid>)?<feature_list>? + * <devices> : <devno-range>|<busid-range> + * <dasd_module> : dasd_diag_mod|dasd_eckd_mod|dasd_fba_mod + * + * <dasd> : autodetect|probeonly|<devices>(,<devices>)* + */ + +int dasd_probeonly = 0; /* is true, when probeonly mode is active */ +int dasd_autodetect = 0; /* is true, when autodetection is active */ +int dasd_nopav = 0; /* is true, when PAV is disabled */ +EXPORT_SYMBOL_GPL(dasd_nopav); +int dasd_nofcx; /* disable High Performance Ficon */ +EXPORT_SYMBOL_GPL(dasd_nofcx); + +/* + * char *dasd[] is intended to hold the ranges supplied by the dasd= statement + * it is named 'dasd' to directly be filled by insmod with the comma separated + * strings when running as a module. + */ +static char *dasd[256]; +module_param_array(dasd, charp, NULL, S_IRUGO); + +/* + * Single spinlock to protect devmap and servermap structures and lists. + */ +static DEFINE_SPINLOCK(dasd_devmap_lock); + +/* + * Hash lists for devmap structures. + */ +static struct list_head dasd_hashlists[256]; +int dasd_max_devindex; + +static struct dasd_devmap *dasd_add_busid(const char *, int); + +static inline int +dasd_hash_busid(const char *bus_id) +{ + int hash, i; + + hash = 0; + for (i = 0; (i < DASD_BUS_ID_SIZE) && *bus_id; i++, bus_id++) + hash += *bus_id; + return hash & 0xff; +} + +#ifndef MODULE +/* + * The parameter parsing functions for builtin-drivers are called + * before kmalloc works. Store the pointers to the parameters strings + * into dasd[] for later processing. + */ +static int __init +dasd_call_setup(char *str) +{ + static int count = 0; + + if (count < 256) + dasd[count++] = str; + return 1; +} + +__setup ("dasd=", dasd_call_setup); +#endif /* #ifndef MODULE */ + +#define DASD_IPLDEV "ipldev" + +/* + * Read a device busid/devno from a string. + */ +static int + +dasd_busid(char **str, int *id0, int *id1, int *devno) +{ + int val, old_style; + + /* Interpret ipldev busid */ + if (strncmp(DASD_IPLDEV, *str, strlen(DASD_IPLDEV)) == 0) { + if (ipl_info.type != IPL_TYPE_CCW) { + pr_err("The IPL device is not a CCW device\n"); + return -EINVAL; + } + *id0 = 0; + *id1 = ipl_info.data.ccw.dev_id.ssid; + *devno = ipl_info.data.ccw.dev_id.devno; + *str += strlen(DASD_IPLDEV); + + return 0; + } + /* check for leading '0x' */ + old_style = 0; + if ((*str)[0] == '0' && (*str)[1] == 'x') { + *str += 2; + old_style = 1; + } + if (!isxdigit((*str)[0])) /* We require at least one hex digit */ + return -EINVAL; + val = simple_strtoul(*str, str, 16); + if (old_style || (*str)[0] != '.') { + *id0 = *id1 = 0; + if (val < 0 || val > 0xffff) + return -EINVAL; + *devno = val; + return 0; + } + /* New style x.y.z busid */ + if (val < 0 || val > 0xff) + return -EINVAL; + *id0 = val; + (*str)++; + if (!isxdigit((*str)[0])) /* We require at least one hex digit */ + return -EINVAL; + val = simple_strtoul(*str, str, 16); + if (val < 0 || val > 0xff || (*str)++[0] != '.') + return -EINVAL; + *id1 = val; + if (!isxdigit((*str)[0])) /* We require at least one hex digit */ + return -EINVAL; + val = simple_strtoul(*str, str, 16); + if (val < 0 || val > 0xffff) + return -EINVAL; + *devno = val; + return 0; +} + +/* + * Read colon separated list of dasd features. Currently there is + * only one: "ro" for read-only devices. The default feature set + * is empty (value 0). + */ +static int +dasd_feature_list(char *str, char **endp) +{ + int features, len, rc; + + rc = 0; + if (*str != '(') { + *endp = str; + return DASD_FEATURE_DEFAULT; + } + str++; + features = 0; + + while (1) { + for (len = 0; + str[len] && str[len] != ':' && str[len] != ')'; len++); + if (len == 2 && !strncmp(str, "ro", 2)) + features |= DASD_FEATURE_READONLY; + else if (len == 4 && !strncmp(str, "diag", 4)) + features |= DASD_FEATURE_USEDIAG; + else if (len == 3 && !strncmp(str, "raw", 3)) + features |= DASD_FEATURE_USERAW; + else if (len == 6 && !strncmp(str, "erplog", 6)) + features |= DASD_FEATURE_ERPLOG; + else if (len == 8 && !strncmp(str, "failfast", 8)) + features |= DASD_FEATURE_FAILFAST; + else { + pr_warning("%*s is not a supported device option\n", + len, str); + rc = -EINVAL; + } + str += len; + if (*str != ':') + break; + str++; + } + if (*str != ')') { + pr_warning("A closing parenthesis ')' is missing in the " + "dasd= parameter\n"); + rc = -EINVAL; + } else + str++; + *endp = str; + if (rc != 0) + return rc; + return features; +} + +/* + * Try to match the first element on the comma separated parse string + * with one of the known keywords. If a keyword is found, take the approprate + * action and return a pointer to the residual string. If the first element + * could not be matched to any keyword then return an error code. + */ +static char * +dasd_parse_keyword( char *parsestring ) { + + char *nextcomma, *residual_str; + int length; + + nextcomma = strchr(parsestring,','); + if (nextcomma) { + length = nextcomma - parsestring; + residual_str = nextcomma + 1; + } else { + length = strlen(parsestring); + residual_str = parsestring + length; + } + if (strncmp("autodetect", parsestring, length) == 0) { + dasd_autodetect = 1; + pr_info("The autodetection mode has been activated\n"); + return residual_str; + } + if (strncmp("probeonly", parsestring, length) == 0) { + dasd_probeonly = 1; + pr_info("The probeonly mode has been activated\n"); + return residual_str; + } + if (strncmp("nopav", parsestring, length) == 0) { + if (MACHINE_IS_VM) + pr_info("'nopav' is not supported on z/VM\n"); + else { + dasd_nopav = 1; + pr_info("PAV support has be deactivated\n"); + } + return residual_str; + } + if (strncmp("nofcx", parsestring, length) == 0) { + dasd_nofcx = 1; + pr_info("High Performance FICON support has been " + "deactivated\n"); + return residual_str; + } + if (strncmp("fixedbuffers", parsestring, length) == 0) { + if (dasd_page_cache) + return residual_str; + dasd_page_cache = + kmem_cache_create("dasd_page_cache", PAGE_SIZE, + PAGE_SIZE, SLAB_CACHE_DMA, + NULL); + if (!dasd_page_cache) + DBF_EVENT(DBF_WARNING, "%s", "Failed to create slab, " + "fixed buffer mode disabled."); + else + DBF_EVENT(DBF_INFO, "%s", + "turning on fixed buffer mode"); + return residual_str; + } + return ERR_PTR(-EINVAL); +} + +/* + * Try to interprete the first element on the comma separated parse string + * as a device number or a range of devices. If the interpretation is + * successful, create the matching dasd_devmap entries and return a pointer + * to the residual string. + * If interpretation fails or in case of an error, return an error code. + */ +static char * +dasd_parse_range( char *parsestring ) { + + struct dasd_devmap *devmap; + int from, from_id0, from_id1; + int to, to_id0, to_id1; + int features, rc; + char bus_id[DASD_BUS_ID_SIZE+1], *str; + + str = parsestring; + rc = dasd_busid(&str, &from_id0, &from_id1, &from); + if (rc == 0) { + to = from; + to_id0 = from_id0; + to_id1 = from_id1; + if (*str == '-') { + str++; + rc = dasd_busid(&str, &to_id0, &to_id1, &to); + } + } + if (rc == 0 && + (from_id0 != to_id0 || from_id1 != to_id1 || from > to)) + rc = -EINVAL; + if (rc) { + pr_err("%s is not a valid device range\n", parsestring); + return ERR_PTR(rc); + } + features = dasd_feature_list(str, &str); + if (features < 0) + return ERR_PTR(-EINVAL); + /* each device in dasd= parameter should be set initially online */ + features |= DASD_FEATURE_INITIAL_ONLINE; + while (from <= to) { + sprintf(bus_id, "%01x.%01x.%04x", + from_id0, from_id1, from++); + devmap = dasd_add_busid(bus_id, features); + if (IS_ERR(devmap)) + return (char *)devmap; + } + if (*str == ',') + return str + 1; + if (*str == '\0') + return str; + pr_warning("The dasd= parameter value %s has an invalid ending\n", + str); + return ERR_PTR(-EINVAL); +} + +static char * +dasd_parse_next_element( char *parsestring ) { + char * residual_str; + residual_str = dasd_parse_keyword(parsestring); + if (!IS_ERR(residual_str)) + return residual_str; + residual_str = dasd_parse_range(parsestring); + return residual_str; +} + +/* + * Parse parameters stored in dasd[] + * The 'dasd=...' parameter allows to specify a comma separated list of + * keywords and device ranges. When the dasd driver is build into the kernel, + * the complete list will be stored as one element of the dasd[] array. + * When the dasd driver is build as a module, then the list is broken into + * it's elements and each dasd[] entry contains one element. + */ +int +dasd_parse(void) +{ + int rc, i; + char *parsestring; + + rc = 0; + for (i = 0; i < 256; i++) { + if (dasd[i] == NULL) + break; + parsestring = dasd[i]; + /* loop over the comma separated list in the parsestring */ + while (*parsestring) { + parsestring = dasd_parse_next_element(parsestring); + if(IS_ERR(parsestring)) { + rc = PTR_ERR(parsestring); + break; + } + } + if (rc) { + DBF_EVENT(DBF_ALERT, "%s", "invalid range found"); + break; + } + } + return rc; +} + +/* + * Add a devmap for the device specified by busid. It is possible that + * the devmap already exists (dasd= parameter). The order of the devices + * added through this function will define the kdevs for the individual + * devices. + */ +static struct dasd_devmap * +dasd_add_busid(const char *bus_id, int features) +{ + struct dasd_devmap *devmap, *new, *tmp; + int hash; + + new = kzalloc(sizeof(struct dasd_devmap), GFP_KERNEL); + if (!new) + return ERR_PTR(-ENOMEM); + spin_lock(&dasd_devmap_lock); + devmap = NULL; + hash = dasd_hash_busid(bus_id); + list_for_each_entry(tmp, &dasd_hashlists[hash], list) + if (strncmp(tmp->bus_id, bus_id, DASD_BUS_ID_SIZE) == 0) { + devmap = tmp; + break; + } + if (!devmap) { + /* This bus_id is new. */ + new->devindex = dasd_max_devindex++; + strncpy(new->bus_id, bus_id, DASD_BUS_ID_SIZE); + new->features = features; + new->device = NULL; + list_add(&new->list, &dasd_hashlists[hash]); + devmap = new; + new = NULL; + } + spin_unlock(&dasd_devmap_lock); + kfree(new); + return devmap; +} + +/* + * Find devmap for device with given bus_id. + */ +static struct dasd_devmap * +dasd_find_busid(const char *bus_id) +{ + struct dasd_devmap *devmap, *tmp; + int hash; + + spin_lock(&dasd_devmap_lock); + devmap = ERR_PTR(-ENODEV); + hash = dasd_hash_busid(bus_id); + list_for_each_entry(tmp, &dasd_hashlists[hash], list) { + if (strncmp(tmp->bus_id, bus_id, DASD_BUS_ID_SIZE) == 0) { + devmap = tmp; + break; + } + } + spin_unlock(&dasd_devmap_lock); + return devmap; +} + +/* + * Check if busid has been added to the list of dasd ranges. + */ +int +dasd_busid_known(const char *bus_id) +{ + return IS_ERR(dasd_find_busid(bus_id)) ? -ENOENT : 0; +} + +/* + * Forget all about the device numbers added so far. + * This may only be called at module unload or system shutdown. + */ +static void +dasd_forget_ranges(void) +{ + struct dasd_devmap *devmap, *n; + int i; + + spin_lock(&dasd_devmap_lock); + for (i = 0; i < 256; i++) { + list_for_each_entry_safe(devmap, n, &dasd_hashlists[i], list) { + BUG_ON(devmap->device != NULL); + list_del(&devmap->list); + kfree(devmap); + } + } + spin_unlock(&dasd_devmap_lock); +} + +/* + * Find the device struct by its device index. + */ +struct dasd_device * +dasd_device_from_devindex(int devindex) +{ + struct dasd_devmap *devmap, *tmp; + struct dasd_device *device; + int i; + + spin_lock(&dasd_devmap_lock); + devmap = NULL; + for (i = 0; (i < 256) && !devmap; i++) + list_for_each_entry(tmp, &dasd_hashlists[i], list) + if (tmp->devindex == devindex) { + /* Found the devmap for the device. */ + devmap = tmp; + break; + } + if (devmap && devmap->device) { + device = devmap->device; + dasd_get_device(device); + } else + device = ERR_PTR(-ENODEV); + spin_unlock(&dasd_devmap_lock); + return device; +} + +/* + * Return devmap for cdev. If no devmap exists yet, create one and + * connect it to the cdev. + */ +static struct dasd_devmap * +dasd_devmap_from_cdev(struct ccw_device *cdev) +{ + struct dasd_devmap *devmap; + + devmap = dasd_find_busid(dev_name(&cdev->dev)); + if (IS_ERR(devmap)) + devmap = dasd_add_busid(dev_name(&cdev->dev), + DASD_FEATURE_DEFAULT); + return devmap; +} + +/* + * Create a dasd device structure for cdev. + */ +struct dasd_device * +dasd_create_device(struct ccw_device *cdev) +{ + struct dasd_devmap *devmap; + struct dasd_device *device; + unsigned long flags; + int rc; + + devmap = dasd_devmap_from_cdev(cdev); + if (IS_ERR(devmap)) + return (void *) devmap; + + device = dasd_alloc_device(); + if (IS_ERR(device)) + return device; + atomic_set(&device->ref_count, 3); + + spin_lock(&dasd_devmap_lock); + if (!devmap->device) { + devmap->device = device; + device->devindex = devmap->devindex; + device->features = devmap->features; + get_device(&cdev->dev); + device->cdev = cdev; + rc = 0; + } else + /* Someone else was faster. */ + rc = -EBUSY; + spin_unlock(&dasd_devmap_lock); + + if (rc) { + dasd_free_device(device); + return ERR_PTR(rc); + } + + spin_lock_irqsave(get_ccwdev_lock(cdev), flags); + dev_set_drvdata(&cdev->dev, device); + spin_unlock_irqrestore(get_ccwdev_lock(cdev), flags); + + return device; +} + +/* + * Wait queue for dasd_delete_device waits. + */ +static DECLARE_WAIT_QUEUE_HEAD(dasd_delete_wq); + +/* + * Remove a dasd device structure. The passed referenced + * is destroyed. + */ +void +dasd_delete_device(struct dasd_device *device) +{ + struct ccw_device *cdev; + struct dasd_devmap *devmap; + unsigned long flags; + + /* First remove device pointer from devmap. */ + devmap = dasd_find_busid(dev_name(&device->cdev->dev)); + BUG_ON(IS_ERR(devmap)); + spin_lock(&dasd_devmap_lock); + if (devmap->device != device) { + spin_unlock(&dasd_devmap_lock); + dasd_put_device(device); + return; + } + devmap->device = NULL; + spin_unlock(&dasd_devmap_lock); + + /* Disconnect dasd_device structure from ccw_device structure. */ + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + dev_set_drvdata(&device->cdev->dev, NULL); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + + /* + * Drop ref_count by 3, one for the devmap reference, one for + * the cdev reference and one for the passed reference. + */ + atomic_sub(3, &device->ref_count); + + /* Wait for reference counter to drop to zero. */ + wait_event(dasd_delete_wq, atomic_read(&device->ref_count) == 0); + + /* Disconnect dasd_device structure from ccw_device structure. */ + cdev = device->cdev; + device->cdev = NULL; + + /* Put ccw_device structure. */ + put_device(&cdev->dev); + + /* Now the device structure can be freed. */ + dasd_free_device(device); +} + +/* + * Reference counter dropped to zero. Wake up waiter + * in dasd_delete_device. + */ +void +dasd_put_device_wake(struct dasd_device *device) +{ + wake_up(&dasd_delete_wq); +} +EXPORT_SYMBOL_GPL(dasd_put_device_wake); + +/* + * Return dasd_device structure associated with cdev. + * This function needs to be called with the ccw device + * lock held. It can be used from interrupt context. + */ +struct dasd_device * +dasd_device_from_cdev_locked(struct ccw_device *cdev) +{ + struct dasd_device *device = dev_get_drvdata(&cdev->dev); + + if (!device) + return ERR_PTR(-ENODEV); + dasd_get_device(device); + return device; +} + +/* + * Return dasd_device structure associated with cdev. + */ +struct dasd_device * +dasd_device_from_cdev(struct ccw_device *cdev) +{ + struct dasd_device *device; + unsigned long flags; + + spin_lock_irqsave(get_ccwdev_lock(cdev), flags); + device = dasd_device_from_cdev_locked(cdev); + spin_unlock_irqrestore(get_ccwdev_lock(cdev), flags); + return device; +} + +void dasd_add_link_to_gendisk(struct gendisk *gdp, struct dasd_device *device) +{ + struct dasd_devmap *devmap; + + devmap = dasd_find_busid(dev_name(&device->cdev->dev)); + if (IS_ERR(devmap)) + return; + spin_lock(&dasd_devmap_lock); + gdp->private_data = devmap; + spin_unlock(&dasd_devmap_lock); +} + +struct dasd_device *dasd_device_from_gendisk(struct gendisk *gdp) +{ + struct dasd_device *device; + struct dasd_devmap *devmap; + + if (!gdp->private_data) + return NULL; + device = NULL; + spin_lock(&dasd_devmap_lock); + devmap = gdp->private_data; + if (devmap && devmap->device) { + device = devmap->device; + dasd_get_device(device); + } + spin_unlock(&dasd_devmap_lock); + return device; +} + +/* + * SECTION: files in sysfs + */ + +/* + * failfast controls the behaviour, if no path is available + */ +static ssize_t dasd_ff_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct dasd_devmap *devmap; + int ff_flag; + + devmap = dasd_find_busid(dev_name(dev)); + if (!IS_ERR(devmap)) + ff_flag = (devmap->features & DASD_FEATURE_FAILFAST) != 0; + else + ff_flag = (DASD_FEATURE_DEFAULT & DASD_FEATURE_FAILFAST) != 0; + return snprintf(buf, PAGE_SIZE, ff_flag ? "1\n" : "0\n"); +} + +static ssize_t dasd_ff_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + int val; + char *endp; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + val = simple_strtoul(buf, &endp, 0); + if (((endp + 1) < (buf + count)) || (val > 1)) + return -EINVAL; + + spin_lock(&dasd_devmap_lock); + if (val) + devmap->features |= DASD_FEATURE_FAILFAST; + else + devmap->features &= ~DASD_FEATURE_FAILFAST; + if (devmap->device) + devmap->device->features = devmap->features; + spin_unlock(&dasd_devmap_lock); + return count; +} + +static DEVICE_ATTR(failfast, 0644, dasd_ff_show, dasd_ff_store); + +/* + * readonly controls the readonly status of a dasd + */ +static ssize_t +dasd_ro_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_devmap *devmap; + int ro_flag; + + devmap = dasd_find_busid(dev_name(dev)); + if (!IS_ERR(devmap)) + ro_flag = (devmap->features & DASD_FEATURE_READONLY) != 0; + else + ro_flag = (DASD_FEATURE_DEFAULT & DASD_FEATURE_READONLY) != 0; + return snprintf(buf, PAGE_SIZE, ro_flag ? "1\n" : "0\n"); +} + +static ssize_t +dasd_ro_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + struct dasd_device *device; + int val; + char *endp; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + val = simple_strtoul(buf, &endp, 0); + if (((endp + 1) < (buf + count)) || (val > 1)) + return -EINVAL; + + spin_lock(&dasd_devmap_lock); + if (val) + devmap->features |= DASD_FEATURE_READONLY; + else + devmap->features &= ~DASD_FEATURE_READONLY; + device = devmap->device; + if (device) { + device->features = devmap->features; + val = val || test_bit(DASD_FLAG_DEVICE_RO, &device->flags); + } + spin_unlock(&dasd_devmap_lock); + if (device && device->block && device->block->gdp) + set_disk_ro(device->block->gdp, val); + return count; +} + +static DEVICE_ATTR(readonly, 0644, dasd_ro_show, dasd_ro_store); +/* + * erplog controls the logging of ERP related data + * (e.g. failing channel programs). + */ +static ssize_t +dasd_erplog_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_devmap *devmap; + int erplog; + + devmap = dasd_find_busid(dev_name(dev)); + if (!IS_ERR(devmap)) + erplog = (devmap->features & DASD_FEATURE_ERPLOG) != 0; + else + erplog = (DASD_FEATURE_DEFAULT & DASD_FEATURE_ERPLOG) != 0; + return snprintf(buf, PAGE_SIZE, erplog ? "1\n" : "0\n"); +} + +static ssize_t +dasd_erplog_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + int val; + char *endp; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + val = simple_strtoul(buf, &endp, 0); + if (((endp + 1) < (buf + count)) || (val > 1)) + return -EINVAL; + + spin_lock(&dasd_devmap_lock); + if (val) + devmap->features |= DASD_FEATURE_ERPLOG; + else + devmap->features &= ~DASD_FEATURE_ERPLOG; + if (devmap->device) + devmap->device->features = devmap->features; + spin_unlock(&dasd_devmap_lock); + return count; +} + +static DEVICE_ATTR(erplog, 0644, dasd_erplog_show, dasd_erplog_store); + +/* + * use_diag controls whether the driver should use diag rather than ssch + * to talk to the device + */ +static ssize_t +dasd_use_diag_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_devmap *devmap; + int use_diag; + + devmap = dasd_find_busid(dev_name(dev)); + if (!IS_ERR(devmap)) + use_diag = (devmap->features & DASD_FEATURE_USEDIAG) != 0; + else + use_diag = (DASD_FEATURE_DEFAULT & DASD_FEATURE_USEDIAG) != 0; + return sprintf(buf, use_diag ? "1\n" : "0\n"); +} + +static ssize_t +dasd_use_diag_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + ssize_t rc; + int val; + char *endp; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + val = simple_strtoul(buf, &endp, 0); + if (((endp + 1) < (buf + count)) || (val > 1)) + return -EINVAL; + + spin_lock(&dasd_devmap_lock); + /* Changing diag discipline flag is only allowed in offline state. */ + rc = count; + if (!devmap->device && !(devmap->features & DASD_FEATURE_USERAW)) { + if (val) + devmap->features |= DASD_FEATURE_USEDIAG; + else + devmap->features &= ~DASD_FEATURE_USEDIAG; + } else + rc = -EPERM; + spin_unlock(&dasd_devmap_lock); + return rc; +} + +static DEVICE_ATTR(use_diag, 0644, dasd_use_diag_show, dasd_use_diag_store); + +/* + * use_raw controls whether the driver should give access to raw eckd data or + * operate in standard mode + */ +static ssize_t +dasd_use_raw_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_devmap *devmap; + int use_raw; + + devmap = dasd_find_busid(dev_name(dev)); + if (!IS_ERR(devmap)) + use_raw = (devmap->features & DASD_FEATURE_USERAW) != 0; + else + use_raw = (DASD_FEATURE_DEFAULT & DASD_FEATURE_USERAW) != 0; + return sprintf(buf, use_raw ? "1\n" : "0\n"); +} + +static ssize_t +dasd_use_raw_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + ssize_t rc; + unsigned long val; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + if ((kstrtoul(buf, 10, &val) != 0) || val > 1) + return -EINVAL; + + spin_lock(&dasd_devmap_lock); + /* Changing diag discipline flag is only allowed in offline state. */ + rc = count; + if (!devmap->device && !(devmap->features & DASD_FEATURE_USEDIAG)) { + if (val) + devmap->features |= DASD_FEATURE_USERAW; + else + devmap->features &= ~DASD_FEATURE_USERAW; + } else + rc = -EPERM; + spin_unlock(&dasd_devmap_lock); + return rc; +} + +static DEVICE_ATTR(raw_track_access, 0644, dasd_use_raw_show, + dasd_use_raw_store); + +static ssize_t +dasd_safe_offline_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct ccw_device *cdev = to_ccwdev(dev); + struct dasd_device *device; + int rc; + + device = dasd_device_from_cdev(cdev); + if (IS_ERR(device)) { + rc = PTR_ERR(device); + goto out; + } + + if (test_bit(DASD_FLAG_OFFLINE, &device->flags) || + test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) { + /* Already doing offline processing */ + dasd_put_device(device); + rc = -EBUSY; + goto out; + } + + set_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags); + dasd_put_device(device); + + rc = ccw_device_set_offline(cdev); + +out: + return rc ? rc : count; +} + +static DEVICE_ATTR(safe_offline, 0200, NULL, dasd_safe_offline_store); + +static ssize_t +dasd_discipline_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct dasd_device *device; + ssize_t len; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + goto out; + else if (!device->discipline) { + dasd_put_device(device); + goto out; + } else { + len = snprintf(buf, PAGE_SIZE, "%s\n", + device->discipline->name); + dasd_put_device(device); + return len; + } +out: + len = snprintf(buf, PAGE_SIZE, "none\n"); + return len; +} + +static DEVICE_ATTR(discipline, 0444, dasd_discipline_show, NULL); + +static ssize_t +dasd_device_status_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct dasd_device *device; + ssize_t len; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (!IS_ERR(device)) { + switch (device->state) { + case DASD_STATE_NEW: + len = snprintf(buf, PAGE_SIZE, "new\n"); + break; + case DASD_STATE_KNOWN: + len = snprintf(buf, PAGE_SIZE, "detected\n"); + break; + case DASD_STATE_BASIC: + len = snprintf(buf, PAGE_SIZE, "basic\n"); + break; + case DASD_STATE_UNFMT: + len = snprintf(buf, PAGE_SIZE, "unformatted\n"); + break; + case DASD_STATE_READY: + len = snprintf(buf, PAGE_SIZE, "ready\n"); + break; + case DASD_STATE_ONLINE: + len = snprintf(buf, PAGE_SIZE, "online\n"); + break; + default: + len = snprintf(buf, PAGE_SIZE, "no stat\n"); + break; + } + dasd_put_device(device); + } else + len = snprintf(buf, PAGE_SIZE, "unknown\n"); + return len; +} + +static DEVICE_ATTR(status, 0444, dasd_device_status_show, NULL); + +static ssize_t dasd_alias_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct dasd_device *device; + struct dasd_uid uid; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return sprintf(buf, "0\n"); + + if (device->discipline && device->discipline->get_uid && + !device->discipline->get_uid(device, &uid)) { + if (uid.type == UA_BASE_PAV_ALIAS || + uid.type == UA_HYPER_PAV_ALIAS) { + dasd_put_device(device); + return sprintf(buf, "1\n"); + } + } + dasd_put_device(device); + + return sprintf(buf, "0\n"); +} + +static DEVICE_ATTR(alias, 0444, dasd_alias_show, NULL); + +static ssize_t dasd_vendor_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct dasd_device *device; + struct dasd_uid uid; + char *vendor; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + vendor = ""; + if (IS_ERR(device)) + return snprintf(buf, PAGE_SIZE, "%s\n", vendor); + + if (device->discipline && device->discipline->get_uid && + !device->discipline->get_uid(device, &uid)) + vendor = uid.vendor; + + dasd_put_device(device); + + return snprintf(buf, PAGE_SIZE, "%s\n", vendor); +} + +static DEVICE_ATTR(vendor, 0444, dasd_vendor_show, NULL); + +#define UID_STRLEN ( /* vendor */ 3 + 1 + /* serial */ 14 + 1 +\ + /* SSID */ 4 + 1 + /* unit addr */ 2 + 1 +\ + /* vduit */ 32 + 1) + +static ssize_t +dasd_uid_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_device *device; + struct dasd_uid uid; + char uid_string[UID_STRLEN]; + char ua_string[3]; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + uid_string[0] = 0; + if (IS_ERR(device)) + return snprintf(buf, PAGE_SIZE, "%s\n", uid_string); + + if (device->discipline && device->discipline->get_uid && + !device->discipline->get_uid(device, &uid)) { + switch (uid.type) { + case UA_BASE_DEVICE: + snprintf(ua_string, sizeof(ua_string), "%02x", + uid.real_unit_addr); + break; + case UA_BASE_PAV_ALIAS: + snprintf(ua_string, sizeof(ua_string), "%02x", + uid.base_unit_addr); + break; + case UA_HYPER_PAV_ALIAS: + snprintf(ua_string, sizeof(ua_string), "xx"); + break; + default: + /* should not happen, treat like base device */ + snprintf(ua_string, sizeof(ua_string), "%02x", + uid.real_unit_addr); + break; + } + + if (strlen(uid.vduit) > 0) + snprintf(uid_string, sizeof(uid_string), + "%s.%s.%04x.%s.%s", + uid.vendor, uid.serial, uid.ssid, ua_string, + uid.vduit); + else + snprintf(uid_string, sizeof(uid_string), + "%s.%s.%04x.%s", + uid.vendor, uid.serial, uid.ssid, ua_string); + } + dasd_put_device(device); + + return snprintf(buf, PAGE_SIZE, "%s\n", uid_string); +} +static DEVICE_ATTR(uid, 0444, dasd_uid_show, NULL); + +/* + * extended error-reporting + */ +static ssize_t +dasd_eer_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_devmap *devmap; + int eer_flag; + + devmap = dasd_find_busid(dev_name(dev)); + if (!IS_ERR(devmap) && devmap->device) + eer_flag = dasd_eer_enabled(devmap->device); + else + eer_flag = 0; + return snprintf(buf, PAGE_SIZE, eer_flag ? "1\n" : "0\n"); +} + +static ssize_t +dasd_eer_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + int val, rc; + char *endp; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + if (!devmap->device) + return -ENODEV; + + val = simple_strtoul(buf, &endp, 0); + if (((endp + 1) < (buf + count)) || (val > 1)) + return -EINVAL; + + if (val) { + rc = dasd_eer_enable(devmap->device); + if (rc) + return rc; + } else + dasd_eer_disable(devmap->device); + return count; +} + +static DEVICE_ATTR(eer_enabled, 0644, dasd_eer_show, dasd_eer_store); + +/* + * expiration time for default requests + */ +static ssize_t +dasd_expires_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_device *device; + int len; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return -ENODEV; + len = snprintf(buf, PAGE_SIZE, "%lu\n", device->default_expires); + dasd_put_device(device); + return len; +} + +static ssize_t +dasd_expires_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_device *device; + unsigned long val; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return -ENODEV; + + if ((kstrtoul(buf, 10, &val) != 0) || + (val > DASD_EXPIRES_MAX) || val == 0) { + dasd_put_device(device); + return -EINVAL; + } + + if (val) + device->default_expires = val; + + dasd_put_device(device); + return count; +} + +static DEVICE_ATTR(expires, 0644, dasd_expires_show, dasd_expires_store); + +static ssize_t +dasd_retries_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dasd_device *device; + int len; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return -ENODEV; + len = snprintf(buf, PAGE_SIZE, "%lu\n", device->default_retries); + dasd_put_device(device); + return len; +} + +static ssize_t +dasd_retries_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_device *device; + unsigned long val; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return -ENODEV; + + if ((kstrtoul(buf, 10, &val) != 0) || + (val > DASD_RETRIES_MAX)) { + dasd_put_device(device); + return -EINVAL; + } + + if (val) + device->default_retries = val; + + dasd_put_device(device); + return count; +} + +static DEVICE_ATTR(retries, 0644, dasd_retries_show, dasd_retries_store); + +static ssize_t +dasd_timeout_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct dasd_device *device; + int len; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return -ENODEV; + len = snprintf(buf, PAGE_SIZE, "%lu\n", device->blk_timeout); + dasd_put_device(device); + return len; +} + +static ssize_t +dasd_timeout_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_device *device; + struct request_queue *q; + unsigned long val, flags; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device) || !device->block) + return -ENODEV; + + if ((kstrtoul(buf, 10, &val) != 0) || + val > UINT_MAX / HZ) { + dasd_put_device(device); + return -EINVAL; + } + q = device->block->request_queue; + if (!q) { + dasd_put_device(device); + return -ENODEV; + } + spin_lock_irqsave(&device->block->request_queue_lock, flags); + if (!val) + blk_queue_rq_timed_out(q, NULL); + else + blk_queue_rq_timed_out(q, dasd_times_out); + + device->blk_timeout = val; + + blk_queue_rq_timeout(q, device->blk_timeout * HZ); + spin_unlock_irqrestore(&device->block->request_queue_lock, flags); + + dasd_put_device(device); + return count; +} + +static DEVICE_ATTR(timeout, 0644, + dasd_timeout_show, dasd_timeout_store); + +static ssize_t dasd_reservation_policy_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct dasd_devmap *devmap; + int rc = 0; + + devmap = dasd_find_busid(dev_name(dev)); + if (IS_ERR(devmap)) { + rc = snprintf(buf, PAGE_SIZE, "ignore\n"); + } else { + spin_lock(&dasd_devmap_lock); + if (devmap->features & DASD_FEATURE_FAILONSLCK) + rc = snprintf(buf, PAGE_SIZE, "fail\n"); + else + rc = snprintf(buf, PAGE_SIZE, "ignore\n"); + spin_unlock(&dasd_devmap_lock); + } + return rc; +} + +static ssize_t dasd_reservation_policy_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_devmap *devmap; + int rc; + + devmap = dasd_devmap_from_cdev(to_ccwdev(dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + rc = 0; + spin_lock(&dasd_devmap_lock); + if (sysfs_streq("ignore", buf)) + devmap->features &= ~DASD_FEATURE_FAILONSLCK; + else if (sysfs_streq("fail", buf)) + devmap->features |= DASD_FEATURE_FAILONSLCK; + else + rc = -EINVAL; + if (devmap->device) + devmap->device->features = devmap->features; + spin_unlock(&dasd_devmap_lock); + if (rc) + return rc; + else + return count; +} + +static DEVICE_ATTR(reservation_policy, 0644, + dasd_reservation_policy_show, dasd_reservation_policy_store); + +static ssize_t dasd_reservation_state_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct dasd_device *device; + int rc = 0; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return snprintf(buf, PAGE_SIZE, "none\n"); + + if (test_bit(DASD_FLAG_IS_RESERVED, &device->flags)) + rc = snprintf(buf, PAGE_SIZE, "reserved\n"); + else if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags)) + rc = snprintf(buf, PAGE_SIZE, "lost\n"); + else + rc = snprintf(buf, PAGE_SIZE, "none\n"); + dasd_put_device(device); + return rc; +} + +static ssize_t dasd_reservation_state_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct dasd_device *device; + int rc = 0; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return -ENODEV; + if (sysfs_streq("reset", buf)) + clear_bit(DASD_FLAG_LOCK_STOLEN, &device->flags); + else + rc = -EINVAL; + dasd_put_device(device); + + if (rc) + return rc; + else + return count; +} + +static DEVICE_ATTR(last_known_reservation_state, 0644, + dasd_reservation_state_show, dasd_reservation_state_store); + +static ssize_t dasd_pm_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct dasd_device *device; + u8 opm, nppm, cablepm, cuirpm, hpfpm; + + device = dasd_device_from_cdev(to_ccwdev(dev)); + if (IS_ERR(device)) + return sprintf(buf, "0\n"); + + opm = device->path_data.opm; + nppm = device->path_data.npm; + cablepm = device->path_data.cablepm; + cuirpm = device->path_data.cuirpm; + hpfpm = device->path_data.hpfpm; + dasd_put_device(device); + + return sprintf(buf, "%02x %02x %02x %02x %02x\n", opm, nppm, + cablepm, cuirpm, hpfpm); +} + +static DEVICE_ATTR(path_masks, 0444, dasd_pm_show, NULL); + +static struct attribute * dasd_attrs[] = { + &dev_attr_readonly.attr, + &dev_attr_discipline.attr, + &dev_attr_status.attr, + &dev_attr_alias.attr, + &dev_attr_vendor.attr, + &dev_attr_uid.attr, + &dev_attr_use_diag.attr, + &dev_attr_raw_track_access.attr, + &dev_attr_eer_enabled.attr, + &dev_attr_erplog.attr, + &dev_attr_failfast.attr, + &dev_attr_expires.attr, + &dev_attr_retries.attr, + &dev_attr_timeout.attr, + &dev_attr_reservation_policy.attr, + &dev_attr_last_known_reservation_state.attr, + &dev_attr_safe_offline.attr, + &dev_attr_path_masks.attr, + NULL, +}; + +static struct attribute_group dasd_attr_group = { + .attrs = dasd_attrs, +}; + +/* + * Return value of the specified feature. + */ +int +dasd_get_feature(struct ccw_device *cdev, int feature) +{ + struct dasd_devmap *devmap; + + devmap = dasd_find_busid(dev_name(&cdev->dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + return ((devmap->features & feature) != 0); +} + +/* + * Set / reset given feature. + * Flag indicates whether to set (!=0) or the reset (=0) the feature. + */ +int +dasd_set_feature(struct ccw_device *cdev, int feature, int flag) +{ + struct dasd_devmap *devmap; + + devmap = dasd_find_busid(dev_name(&cdev->dev)); + if (IS_ERR(devmap)) + return PTR_ERR(devmap); + + spin_lock(&dasd_devmap_lock); + if (flag) + devmap->features |= feature; + else + devmap->features &= ~feature; + if (devmap->device) + devmap->device->features = devmap->features; + spin_unlock(&dasd_devmap_lock); + return 0; +} + + +int +dasd_add_sysfs_files(struct ccw_device *cdev) +{ + return sysfs_create_group(&cdev->dev.kobj, &dasd_attr_group); +} + +void +dasd_remove_sysfs_files(struct ccw_device *cdev) +{ + sysfs_remove_group(&cdev->dev.kobj, &dasd_attr_group); +} + + +int +dasd_devmap_init(void) +{ + int i; + + /* Initialize devmap structures. */ + dasd_max_devindex = 0; + for (i = 0; i < 256; i++) + INIT_LIST_HEAD(&dasd_hashlists[i]); + return 0; +} + +void +dasd_devmap_exit(void) +{ + dasd_forget_ranges(); +} diff --git a/kernel/drivers/s390/block/dasd_diag.c b/kernel/drivers/s390/block/dasd_diag.c new file mode 100644 index 000000000..c062f1620 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_diag.c @@ -0,0 +1,663 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Based on.......: linux/drivers/s390/block/mdisk.c + * ...............: by Hartmunt Penner <hpenner@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2000 + * + */ + +#define KMSG_COMPONENT "dasd" + +#include <linux/kernel_stat.h> +#include <linux/stddef.h> +#include <linux/kernel.h> +#include <linux/slab.h> +#include <linux/hdreg.h> +#include <linux/bio.h> +#include <linux/module.h> +#include <linux/init.h> +#include <linux/jiffies.h> + +#include <asm/dasd.h> +#include <asm/debug.h> +#include <asm/ebcdic.h> +#include <asm/io.h> +#include <asm/irq.h> +#include <asm/vtoc.h> +#include <asm/diag.h> + +#include "dasd_int.h" +#include "dasd_diag.h" + +#define PRINTK_HEADER "dasd(diag):" + +MODULE_LICENSE("GPL"); + +/* The maximum number of blocks per request (max_blocks) is dependent on the + * amount of storage that is available in the static I/O buffer for each + * device. Currently each device gets 2 pages. We want to fit two requests + * into the available memory so that we can immediately start the next if one + * finishes. */ +#define DIAG_MAX_BLOCKS (((2 * PAGE_SIZE - sizeof(struct dasd_ccw_req) - \ + sizeof(struct dasd_diag_req)) / \ + sizeof(struct dasd_diag_bio)) / 2) +#define DIAG_MAX_RETRIES 32 +#define DIAG_TIMEOUT 50 + +static struct dasd_discipline dasd_diag_discipline; + +struct dasd_diag_private { + struct dasd_diag_characteristics rdc_data; + struct dasd_diag_rw_io iob; + struct dasd_diag_init_io iib; + blocknum_t pt_block; + struct ccw_dev_id dev_id; +}; + +struct dasd_diag_req { + unsigned int block_count; + struct dasd_diag_bio bio[0]; +}; + +static const u8 DASD_DIAG_CMS1[] = { 0xc3, 0xd4, 0xe2, 0xf1 };/* EBCDIC CMS1 */ + +/* Perform DIAG250 call with block I/O parameter list iob (input and output) + * and function code cmd. + * In case of an exception return 3. Otherwise return result of bitwise OR of + * resulting condition code and DIAG return code. */ +static inline int dia250(void *iob, int cmd) +{ + register unsigned long reg2 asm ("2") = (unsigned long) iob; + typedef union { + struct dasd_diag_init_io init_io; + struct dasd_diag_rw_io rw_io; + } addr_type; + int rc; + + rc = 3; + asm volatile( + " diag 2,%2,0x250\n" + "0: ipm %0\n" + " srl %0,28\n" + " or %0,3\n" + "1:\n" + EX_TABLE(0b,1b) + : "+d" (rc), "=m" (*(addr_type *) iob) + : "d" (cmd), "d" (reg2), "m" (*(addr_type *) iob) + : "3", "cc"); + return rc; +} + +/* Initialize block I/O to DIAG device using the specified blocksize and + * block offset. On success, return zero and set end_block to contain the + * number of blocks on the device minus the specified offset. Return non-zero + * otherwise. */ +static inline int +mdsk_init_io(struct dasd_device *device, unsigned int blocksize, + blocknum_t offset, blocknum_t *end_block) +{ + struct dasd_diag_private *private; + struct dasd_diag_init_io *iib; + int rc; + + private = (struct dasd_diag_private *) device->private; + iib = &private->iib; + memset(iib, 0, sizeof (struct dasd_diag_init_io)); + + iib->dev_nr = private->dev_id.devno; + iib->block_size = blocksize; + iib->offset = offset; + iib->flaga = DASD_DIAG_FLAGA_DEFAULT; + + rc = dia250(iib, INIT_BIO); + + if ((rc & 3) == 0 && end_block) + *end_block = iib->end_block; + + return rc; +} + +/* Remove block I/O environment for device. Return zero on success, non-zero + * otherwise. */ +static inline int +mdsk_term_io(struct dasd_device * device) +{ + struct dasd_diag_private *private; + struct dasd_diag_init_io *iib; + int rc; + + private = (struct dasd_diag_private *) device->private; + iib = &private->iib; + memset(iib, 0, sizeof (struct dasd_diag_init_io)); + iib->dev_nr = private->dev_id.devno; + rc = dia250(iib, TERM_BIO); + return rc; +} + +/* Error recovery for failed DIAG requests - try to reestablish the DIAG + * environment. */ +static void +dasd_diag_erp(struct dasd_device *device) +{ + int rc; + + mdsk_term_io(device); + rc = mdsk_init_io(device, device->block->bp_block, 0, NULL); + if (rc == 4) { + if (!(test_and_set_bit(DASD_FLAG_DEVICE_RO, &device->flags))) + pr_warning("%s: The access mode of a DIAG device " + "changed to read-only\n", + dev_name(&device->cdev->dev)); + rc = 0; + } + if (rc) + pr_warning("%s: DIAG ERP failed with " + "rc=%d\n", dev_name(&device->cdev->dev), rc); +} + +/* Start a given request at the device. Return zero on success, non-zero + * otherwise. */ +static int +dasd_start_diag(struct dasd_ccw_req * cqr) +{ + struct dasd_device *device; + struct dasd_diag_private *private; + struct dasd_diag_req *dreq; + int rc; + + device = cqr->startdev; + if (cqr->retries < 0) { + DBF_DEV_EVENT(DBF_ERR, device, "DIAG start_IO: request %p " + "- no retry left)", cqr); + cqr->status = DASD_CQR_ERROR; + return -EIO; + } + private = (struct dasd_diag_private *) device->private; + dreq = (struct dasd_diag_req *) cqr->data; + + private->iob.dev_nr = private->dev_id.devno; + private->iob.key = 0; + private->iob.flags = DASD_DIAG_RWFLAG_ASYNC; + private->iob.block_count = dreq->block_count; + private->iob.interrupt_params = (addr_t) cqr; + private->iob.bio_list = dreq->bio; + private->iob.flaga = DASD_DIAG_FLAGA_DEFAULT; + + cqr->startclk = get_tod_clock(); + cqr->starttime = jiffies; + cqr->retries--; + + rc = dia250(&private->iob, RW_BIO); + switch (rc) { + case 0: /* Synchronous I/O finished successfully */ + cqr->stopclk = get_tod_clock(); + cqr->status = DASD_CQR_SUCCESS; + /* Indicate to calling function that only a dasd_schedule_bh() + and no timer is needed */ + rc = -EACCES; + break; + case 8: /* Asynchronous I/O was started */ + cqr->status = DASD_CQR_IN_IO; + rc = 0; + break; + default: /* Error condition */ + cqr->status = DASD_CQR_QUEUED; + DBF_DEV_EVENT(DBF_WARNING, device, "dia250 returned rc=%d", rc); + dasd_diag_erp(device); + rc = -EIO; + break; + } + cqr->intrc = rc; + return rc; +} + +/* Terminate given request at the device. */ +static int +dasd_diag_term_IO(struct dasd_ccw_req * cqr) +{ + struct dasd_device *device; + + device = cqr->startdev; + mdsk_term_io(device); + mdsk_init_io(device, device->block->bp_block, 0, NULL); + cqr->status = DASD_CQR_CLEAR_PENDING; + cqr->stopclk = get_tod_clock(); + dasd_schedule_device_bh(device); + return 0; +} + +/* Handle external interruption. */ +static void dasd_ext_handler(struct ext_code ext_code, + unsigned int param32, unsigned long param64) +{ + struct dasd_ccw_req *cqr, *next; + struct dasd_device *device; + unsigned long long expires; + unsigned long flags; + addr_t ip; + int rc; + + switch (ext_code.subcode >> 8) { + case DASD_DIAG_CODE_31BIT: + ip = (addr_t) param32; + break; + case DASD_DIAG_CODE_64BIT: + ip = (addr_t) param64; + break; + default: + return; + } + inc_irq_stat(IRQEXT_DSD); + if (!ip) { /* no intparm: unsolicited interrupt */ + DBF_EVENT(DBF_NOTICE, "%s", "caught unsolicited " + "interrupt"); + return; + } + cqr = (struct dasd_ccw_req *) ip; + device = (struct dasd_device *) cqr->startdev; + if (strncmp(device->discipline->ebcname, (char *) &cqr->magic, 4)) { + DBF_DEV_EVENT(DBF_WARNING, device, + " magic number of dasd_ccw_req 0x%08X doesn't" + " match discipline 0x%08X", + cqr->magic, *(int *) (&device->discipline->name)); + return; + } + + /* get irq lock to modify request queue */ + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + + /* Check for a pending clear operation */ + if (cqr->status == DASD_CQR_CLEAR_PENDING) { + cqr->status = DASD_CQR_CLEARED; + dasd_device_clear_timer(device); + dasd_schedule_device_bh(device); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + return; + } + + cqr->stopclk = get_tod_clock(); + + expires = 0; + if ((ext_code.subcode & 0xff) == 0) { + cqr->status = DASD_CQR_SUCCESS; + /* Start first request on queue if possible -> fast_io. */ + if (!list_empty(&device->ccw_queue)) { + next = list_entry(device->ccw_queue.next, + struct dasd_ccw_req, devlist); + if (next->status == DASD_CQR_QUEUED) { + rc = dasd_start_diag(next); + if (rc == 0) + expires = next->expires; + } + } + } else { + cqr->status = DASD_CQR_QUEUED; + DBF_DEV_EVENT(DBF_DEBUG, device, "interrupt status for " + "request %p was %d (%d retries left)", cqr, + ext_code.subcode & 0xff, cqr->retries); + dasd_diag_erp(device); + } + + if (expires != 0) + dasd_device_set_timer(device, expires); + else + dasd_device_clear_timer(device); + dasd_schedule_device_bh(device); + + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); +} + +/* Check whether device can be controlled by DIAG discipline. Return zero on + * success, non-zero otherwise. */ +static int +dasd_diag_check_device(struct dasd_device *device) +{ + struct dasd_block *block; + struct dasd_diag_private *private; + struct dasd_diag_characteristics *rdc_data; + struct dasd_diag_bio bio; + struct vtoc_cms_label *label; + blocknum_t end_block; + unsigned int sb, bsize; + int rc; + + private = (struct dasd_diag_private *) device->private; + if (private == NULL) { + private = kzalloc(sizeof(struct dasd_diag_private),GFP_KERNEL); + if (private == NULL) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Allocating memory for private DASD data " + "failed\n"); + return -ENOMEM; + } + ccw_device_get_id(device->cdev, &private->dev_id); + device->private = (void *) private; + } + block = dasd_alloc_block(); + if (IS_ERR(block)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "could not allocate dasd block structure"); + device->private = NULL; + kfree(private); + return PTR_ERR(block); + } + device->block = block; + block->base = device; + + /* Read Device Characteristics */ + rdc_data = (void *) &(private->rdc_data); + rdc_data->dev_nr = private->dev_id.devno; + rdc_data->rdc_len = sizeof (struct dasd_diag_characteristics); + + rc = diag210((struct diag210 *) rdc_data); + if (rc) { + DBF_DEV_EVENT(DBF_WARNING, device, "failed to retrieve device " + "information (rc=%d)", rc); + rc = -EOPNOTSUPP; + goto out; + } + + device->default_expires = DIAG_TIMEOUT; + device->default_retries = DIAG_MAX_RETRIES; + + /* Figure out position of label block */ + switch (private->rdc_data.vdev_class) { + case DEV_CLASS_FBA: + private->pt_block = 1; + break; + case DEV_CLASS_ECKD: + private->pt_block = 2; + break; + default: + pr_warning("%s: Device type %d is not supported " + "in DIAG mode\n", dev_name(&device->cdev->dev), + private->rdc_data.vdev_class); + rc = -EOPNOTSUPP; + goto out; + } + + DBF_DEV_EVENT(DBF_INFO, device, + "%04X: %04X on real %04X/%02X", + rdc_data->dev_nr, + rdc_data->vdev_type, + rdc_data->rdev_type, rdc_data->rdev_model); + + /* terminate all outstanding operations */ + mdsk_term_io(device); + + /* figure out blocksize of device */ + label = (struct vtoc_cms_label *) get_zeroed_page(GFP_KERNEL); + if (label == NULL) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "No memory to allocate initialization request"); + rc = -ENOMEM; + goto out; + } + rc = 0; + end_block = 0; + /* try all sizes - needed for ECKD devices */ + for (bsize = 512; bsize <= PAGE_SIZE; bsize <<= 1) { + mdsk_init_io(device, bsize, 0, &end_block); + memset(&bio, 0, sizeof (struct dasd_diag_bio)); + bio.type = MDSK_READ_REQ; + bio.block_number = private->pt_block + 1; + bio.buffer = label; + memset(&private->iob, 0, sizeof (struct dasd_diag_rw_io)); + private->iob.dev_nr = rdc_data->dev_nr; + private->iob.key = 0; + private->iob.flags = 0; /* do synchronous io */ + private->iob.block_count = 1; + private->iob.interrupt_params = 0; + private->iob.bio_list = &bio; + private->iob.flaga = DASD_DIAG_FLAGA_DEFAULT; + rc = dia250(&private->iob, RW_BIO); + if (rc == 3) { + pr_warning("%s: A 64-bit DIAG call failed\n", + dev_name(&device->cdev->dev)); + rc = -EOPNOTSUPP; + goto out_label; + } + mdsk_term_io(device); + if (rc == 0) + break; + } + if (bsize > PAGE_SIZE) { + pr_warning("%s: Accessing the DASD failed because of an " + "incorrect format (rc=%d)\n", + dev_name(&device->cdev->dev), rc); + rc = -EIO; + goto out_label; + } + /* check for label block */ + if (memcmp(label->label_id, DASD_DIAG_CMS1, + sizeof(DASD_DIAG_CMS1)) == 0) { + /* get formatted blocksize from label block */ + bsize = (unsigned int) label->block_size; + block->blocks = (unsigned long) label->block_count; + } else + block->blocks = end_block; + block->bp_block = bsize; + block->s2b_shift = 0; /* bits to shift 512 to get a block */ + for (sb = 512; sb < bsize; sb = sb << 1) + block->s2b_shift++; + rc = mdsk_init_io(device, block->bp_block, 0, NULL); + if (rc && (rc != 4)) { + pr_warning("%s: DIAG initialization failed with rc=%d\n", + dev_name(&device->cdev->dev), rc); + rc = -EIO; + } else { + if (rc == 4) + set_bit(DASD_FLAG_DEVICE_RO, &device->flags); + pr_info("%s: New DASD with %ld byte/block, total size %ld " + "KB%s\n", dev_name(&device->cdev->dev), + (unsigned long) block->bp_block, + (unsigned long) (block->blocks << + block->s2b_shift) >> 1, + (rc == 4) ? ", read-only device" : ""); + rc = 0; + } +out_label: + free_page((long) label); +out: + if (rc) { + device->block = NULL; + dasd_free_block(block); + device->private = NULL; + kfree(private); + } + return rc; +} + +/* Fill in virtual disk geometry for device. Return zero on success, non-zero + * otherwise. */ +static int +dasd_diag_fill_geometry(struct dasd_block *block, struct hd_geometry *geo) +{ + if (dasd_check_blocksize(block->bp_block) != 0) + return -EINVAL; + geo->cylinders = (block->blocks << block->s2b_shift) >> 10; + geo->heads = 16; + geo->sectors = 128 >> block->s2b_shift; + return 0; +} + +static dasd_erp_fn_t +dasd_diag_erp_action(struct dasd_ccw_req * cqr) +{ + return dasd_default_erp_action; +} + +static dasd_erp_fn_t +dasd_diag_erp_postaction(struct dasd_ccw_req * cqr) +{ + return dasd_default_erp_postaction; +} + +/* Create DASD request from block device request. Return pointer to new + * request on success, ERR_PTR otherwise. */ +static struct dasd_ccw_req *dasd_diag_build_cp(struct dasd_device *memdev, + struct dasd_block *block, + struct request *req) +{ + struct dasd_ccw_req *cqr; + struct dasd_diag_req *dreq; + struct dasd_diag_bio *dbio; + struct req_iterator iter; + struct bio_vec bv; + char *dst; + unsigned int count, datasize; + sector_t recid, first_rec, last_rec; + unsigned int blksize, off; + unsigned char rw_cmd; + + if (rq_data_dir(req) == READ) + rw_cmd = MDSK_READ_REQ; + else if (rq_data_dir(req) == WRITE) + rw_cmd = MDSK_WRITE_REQ; + else + return ERR_PTR(-EINVAL); + blksize = block->bp_block; + /* Calculate record id of first and last block. */ + first_rec = blk_rq_pos(req) >> block->s2b_shift; + last_rec = + (blk_rq_pos(req) + blk_rq_sectors(req) - 1) >> block->s2b_shift; + /* Check struct bio and count the number of blocks for the request. */ + count = 0; + rq_for_each_segment(bv, req, iter) { + if (bv.bv_len & (blksize - 1)) + /* Fba can only do full blocks. */ + return ERR_PTR(-EINVAL); + count += bv.bv_len >> (block->s2b_shift + 9); + } + /* Paranoia. */ + if (count != last_rec - first_rec + 1) + return ERR_PTR(-EINVAL); + /* Build the request */ + datasize = sizeof(struct dasd_diag_req) + + count*sizeof(struct dasd_diag_bio); + cqr = dasd_smalloc_request(DASD_DIAG_MAGIC, 0, datasize, memdev); + if (IS_ERR(cqr)) + return cqr; + + dreq = (struct dasd_diag_req *) cqr->data; + dreq->block_count = count; + dbio = dreq->bio; + recid = first_rec; + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + for (off = 0; off < bv.bv_len; off += blksize) { + memset(dbio, 0, sizeof (struct dasd_diag_bio)); + dbio->type = rw_cmd; + dbio->block_number = recid + 1; + dbio->buffer = dst; + dbio++; + dst += blksize; + recid++; + } + } + cqr->retries = memdev->default_retries; + cqr->buildclk = get_tod_clock(); + if (blk_noretry_request(req) || + block->base->features & DASD_FEATURE_FAILFAST) + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->startdev = memdev; + cqr->memdev = memdev; + cqr->block = block; + cqr->expires = memdev->default_expires * HZ; + cqr->status = DASD_CQR_FILLED; + return cqr; +} + +/* Release DASD request. Return non-zero if request was successful, zero + * otherwise. */ +static int +dasd_diag_free_cp(struct dasd_ccw_req *cqr, struct request *req) +{ + int status; + + status = cqr->status == DASD_CQR_DONE; + dasd_sfree_request(cqr, cqr->memdev); + return status; +} + +static void dasd_diag_handle_terminated_request(struct dasd_ccw_req *cqr) +{ + if (cqr->retries < 0) + cqr->status = DASD_CQR_FAILED; + else + cqr->status = DASD_CQR_FILLED; +}; + +/* Fill in IOCTL data for device. */ +static int +dasd_diag_fill_info(struct dasd_device * device, + struct dasd_information2_t * info) +{ + struct dasd_diag_private *private; + + private = (struct dasd_diag_private *) device->private; + info->label_block = (unsigned int) private->pt_block; + info->FBA_layout = 1; + info->format = DASD_FORMAT_LDL; + info->characteristics_size = sizeof (struct dasd_diag_characteristics); + memcpy(info->characteristics, + &((struct dasd_diag_private *) device->private)->rdc_data, + sizeof (struct dasd_diag_characteristics)); + info->confdata_size = 0; + return 0; +} + +static void +dasd_diag_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req, + struct irb *stat) +{ + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "dump sense not available for DIAG data"); +} + +static struct dasd_discipline dasd_diag_discipline = { + .owner = THIS_MODULE, + .name = "DIAG", + .ebcname = "DIAG", + .max_blocks = DIAG_MAX_BLOCKS, + .check_device = dasd_diag_check_device, + .verify_path = dasd_generic_verify_path, + .fill_geometry = dasd_diag_fill_geometry, + .start_IO = dasd_start_diag, + .term_IO = dasd_diag_term_IO, + .handle_terminated_request = dasd_diag_handle_terminated_request, + .erp_action = dasd_diag_erp_action, + .erp_postaction = dasd_diag_erp_postaction, + .build_cp = dasd_diag_build_cp, + .free_cp = dasd_diag_free_cp, + .dump_sense = dasd_diag_dump_sense, + .fill_info = dasd_diag_fill_info, +}; + +static int __init +dasd_diag_init(void) +{ + if (!MACHINE_IS_VM) { + pr_info("Discipline %s cannot be used without z/VM\n", + dasd_diag_discipline.name); + return -ENODEV; + } + ASCEBC(dasd_diag_discipline.ebcname, 4); + + irq_subclass_register(IRQ_SUBCLASS_SERVICE_SIGNAL); + register_external_irq(EXT_IRQ_CP_SERVICE, dasd_ext_handler); + dasd_diag_discipline_pointer = &dasd_diag_discipline; + return 0; +} + +static void __exit +dasd_diag_cleanup(void) +{ + unregister_external_irq(EXT_IRQ_CP_SERVICE, dasd_ext_handler); + irq_subclass_unregister(IRQ_SUBCLASS_SERVICE_SIGNAL); + dasd_diag_discipline_pointer = NULL; +} + +module_init(dasd_diag_init); +module_exit(dasd_diag_cleanup); diff --git a/kernel/drivers/s390/block/dasd_diag.h b/kernel/drivers/s390/block/dasd_diag.h new file mode 100644 index 000000000..e84a5468d --- /dev/null +++ b/kernel/drivers/s390/block/dasd_diag.h @@ -0,0 +1,80 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Based on.......: linux/drivers/s390/block/mdisk.h + * ...............: by Hartmunt Penner <hpenner@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2000 + * + */ + +#define MDSK_WRITE_REQ 0x01 +#define MDSK_READ_REQ 0x02 + +#define INIT_BIO 0x00 +#define RW_BIO 0x01 +#define TERM_BIO 0x02 + +#define DEV_CLASS_FBA 0x01 +#define DEV_CLASS_ECKD 0x04 + +#define DASD_DIAG_CODE_31BIT 0x03 +#define DASD_DIAG_CODE_64BIT 0x07 + +#define DASD_DIAG_RWFLAG_ASYNC 0x02 +#define DASD_DIAG_RWFLAG_NOCACHE 0x01 + +#define DASD_DIAG_FLAGA_FORMAT_64BIT 0x80 + +struct dasd_diag_characteristics { + u16 dev_nr; + u16 rdc_len; + u8 vdev_class; + u8 vdev_type; + u8 vdev_status; + u8 vdev_flags; + u8 rdev_class; + u8 rdev_type; + u8 rdev_model; + u8 rdev_features; +} __attribute__ ((packed, aligned(4))); + +#define DASD_DIAG_FLAGA_DEFAULT DASD_DIAG_FLAGA_FORMAT_64BIT + +typedef u64 blocknum_t; +typedef s64 sblocknum_t; + +struct dasd_diag_bio { + u8 type; + u8 status; + u8 spare1[2]; + u32 alet; + blocknum_t block_number; + void *buffer; +} __attribute__ ((packed, aligned(8))); + +struct dasd_diag_init_io { + u16 dev_nr; + u8 flaga; + u8 spare1[21]; + u32 block_size; + u8 spare2[4]; + blocknum_t offset; + sblocknum_t start_block; + blocknum_t end_block; + u8 spare3[8]; +} __attribute__ ((packed, aligned(8))); + +struct dasd_diag_rw_io { + u16 dev_nr; + u8 flaga; + u8 spare1[21]; + u8 key; + u8 flags; + u8 spare2[2]; + u32 block_count; + u32 alet; + u8 spare3[4]; + u64 interrupt_params; + struct dasd_diag_bio *bio_list; + u8 spare4[8]; +} __attribute__ ((packed, aligned(8))); diff --git a/kernel/drivers/s390/block/dasd_eckd.c b/kernel/drivers/s390/block/dasd_eckd.c new file mode 100644 index 000000000..6215f6455 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_eckd.c @@ -0,0 +1,4950 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2009 + * EMC Symmetrix ioctl Copyright EMC Corporation, 2008 + * Author.........: Nigel Hislop <hislop_nigel@emc.com> + */ + +#define KMSG_COMPONENT "dasd-eckd" + +#include <linux/stddef.h> +#include <linux/kernel.h> +#include <linux/slab.h> +#include <linux/hdreg.h> /* HDIO_GETGEO */ +#include <linux/bio.h> +#include <linux/module.h> +#include <linux/compat.h> +#include <linux/init.h> + +#include <asm/css_chars.h> +#include <asm/debug.h> +#include <asm/idals.h> +#include <asm/ebcdic.h> +#include <asm/io.h> +#include <asm/uaccess.h> +#include <asm/cio.h> +#include <asm/ccwdev.h> +#include <asm/itcw.h> +#include <asm/schid.h> +#include <asm/chpid.h> + +#include "dasd_int.h" +#include "dasd_eckd.h" + +#ifdef PRINTK_HEADER +#undef PRINTK_HEADER +#endif /* PRINTK_HEADER */ +#define PRINTK_HEADER "dasd(eckd):" + +#define ECKD_C0(i) (i->home_bytes) +#define ECKD_F(i) (i->formula) +#define ECKD_F1(i) (ECKD_F(i)==0x01?(i->factors.f_0x01.f1):\ + (i->factors.f_0x02.f1)) +#define ECKD_F2(i) (ECKD_F(i)==0x01?(i->factors.f_0x01.f2):\ + (i->factors.f_0x02.f2)) +#define ECKD_F3(i) (ECKD_F(i)==0x01?(i->factors.f_0x01.f3):\ + (i->factors.f_0x02.f3)) +#define ECKD_F4(i) (ECKD_F(i)==0x02?(i->factors.f_0x02.f4):0) +#define ECKD_F5(i) (ECKD_F(i)==0x02?(i->factors.f_0x02.f5):0) +#define ECKD_F6(i) (i->factor6) +#define ECKD_F7(i) (i->factor7) +#define ECKD_F8(i) (i->factor8) + +/* + * raw track access always map to 64k in memory + * so it maps to 16 blocks of 4k per track + */ +#define DASD_RAW_BLOCK_PER_TRACK 16 +#define DASD_RAW_BLOCKSIZE 4096 +/* 64k are 128 x 512 byte sectors */ +#define DASD_RAW_SECTORS_PER_TRACK 128 + +MODULE_LICENSE("GPL"); + +static struct dasd_discipline dasd_eckd_discipline; + +/* The ccw bus type uses this table to find devices that it sends to + * dasd_eckd_probe */ +static struct ccw_device_id dasd_eckd_ids[] = { + { CCW_DEVICE_DEVTYPE (0x3990, 0, 0x3390, 0), .driver_info = 0x1}, + { CCW_DEVICE_DEVTYPE (0x2105, 0, 0x3390, 0), .driver_info = 0x2}, + { CCW_DEVICE_DEVTYPE (0x3880, 0, 0x3380, 0), .driver_info = 0x3}, + { CCW_DEVICE_DEVTYPE (0x3990, 0, 0x3380, 0), .driver_info = 0x4}, + { CCW_DEVICE_DEVTYPE (0x2105, 0, 0x3380, 0), .driver_info = 0x5}, + { CCW_DEVICE_DEVTYPE (0x9343, 0, 0x9345, 0), .driver_info = 0x6}, + { CCW_DEVICE_DEVTYPE (0x2107, 0, 0x3390, 0), .driver_info = 0x7}, + { CCW_DEVICE_DEVTYPE (0x2107, 0, 0x3380, 0), .driver_info = 0x8}, + { CCW_DEVICE_DEVTYPE (0x1750, 0, 0x3390, 0), .driver_info = 0x9}, + { CCW_DEVICE_DEVTYPE (0x1750, 0, 0x3380, 0), .driver_info = 0xa}, + { /* end of list */ }, +}; + +MODULE_DEVICE_TABLE(ccw, dasd_eckd_ids); + +static struct ccw_driver dasd_eckd_driver; /* see below */ + +static void *rawpadpage; + +#define INIT_CQR_OK 0 +#define INIT_CQR_UNFORMATTED 1 +#define INIT_CQR_ERROR 2 + +/* emergency request for reserve/release */ +static struct { + struct dasd_ccw_req cqr; + struct ccw1 ccw; + char data[32]; +} *dasd_reserve_req; +static DEFINE_MUTEX(dasd_reserve_mutex); + +/* definitions for the path verification worker */ +struct path_verification_work_data { + struct work_struct worker; + struct dasd_device *device; + struct dasd_ccw_req cqr; + struct ccw1 ccw; + __u8 rcd_buffer[DASD_ECKD_RCD_DATA_SIZE]; + int isglobal; + __u8 tbvpm; +}; +static struct path_verification_work_data *path_verification_worker; +static DEFINE_MUTEX(dasd_path_verification_mutex); + +struct check_attention_work_data { + struct work_struct worker; + struct dasd_device *device; + __u8 lpum; +}; + +/* initial attempt at a probe function. this can be simplified once + * the other detection code is gone */ +static int +dasd_eckd_probe (struct ccw_device *cdev) +{ + int ret; + + /* set ECKD specific ccw-device options */ + ret = ccw_device_set_options(cdev, CCWDEV_ALLOW_FORCE | + CCWDEV_DO_PATHGROUP | CCWDEV_DO_MULTIPATH); + if (ret) { + DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s", + "dasd_eckd_probe: could not set " + "ccw-device options"); + return ret; + } + ret = dasd_generic_probe(cdev, &dasd_eckd_discipline); + return ret; +} + +static int +dasd_eckd_set_online(struct ccw_device *cdev) +{ + return dasd_generic_set_online(cdev, &dasd_eckd_discipline); +} + +static const int sizes_trk0[] = { 28, 148, 84 }; +#define LABEL_SIZE 140 + +/* head and record addresses of count_area read in analysis ccw */ +static const int count_area_head[] = { 0, 0, 0, 0, 2 }; +static const int count_area_rec[] = { 1, 2, 3, 4, 1 }; + +static inline unsigned int +round_up_multiple(unsigned int no, unsigned int mult) +{ + int rem = no % mult; + return (rem ? no - rem + mult : no); +} + +static inline unsigned int +ceil_quot(unsigned int d1, unsigned int d2) +{ + return (d1 + (d2 - 1)) / d2; +} + +static unsigned int +recs_per_track(struct dasd_eckd_characteristics * rdc, + unsigned int kl, unsigned int dl) +{ + int dn, kn; + + switch (rdc->dev_type) { + case 0x3380: + if (kl) + return 1499 / (15 + 7 + ceil_quot(kl + 12, 32) + + ceil_quot(dl + 12, 32)); + else + return 1499 / (15 + ceil_quot(dl + 12, 32)); + case 0x3390: + dn = ceil_quot(dl + 6, 232) + 1; + if (kl) { + kn = ceil_quot(kl + 6, 232) + 1; + return 1729 / (10 + 9 + ceil_quot(kl + 6 * kn, 34) + + 9 + ceil_quot(dl + 6 * dn, 34)); + } else + return 1729 / (10 + 9 + ceil_quot(dl + 6 * dn, 34)); + case 0x9345: + dn = ceil_quot(dl + 6, 232) + 1; + if (kl) { + kn = ceil_quot(kl + 6, 232) + 1; + return 1420 / (18 + 7 + ceil_quot(kl + 6 * kn, 34) + + ceil_quot(dl + 6 * dn, 34)); + } else + return 1420 / (18 + 7 + ceil_quot(dl + 6 * dn, 34)); + } + return 0; +} + +static void set_ch_t(struct ch_t *geo, __u32 cyl, __u8 head) +{ + geo->cyl = (__u16) cyl; + geo->head = cyl >> 16; + geo->head <<= 4; + geo->head |= head; +} + +static int +check_XRC (struct ccw1 *de_ccw, + struct DE_eckd_data *data, + struct dasd_device *device) +{ + struct dasd_eckd_private *private; + int rc; + + private = (struct dasd_eckd_private *) device->private; + if (!private->rdc_data.facilities.XRC_supported) + return 0; + + /* switch on System Time Stamp - needed for XRC Support */ + data->ga_extended |= 0x08; /* switch on 'Time Stamp Valid' */ + data->ga_extended |= 0x02; /* switch on 'Extended Parameter' */ + + rc = get_sync_clock(&data->ep_sys_time); + /* Ignore return code if sync clock is switched off. */ + if (rc == -EOPNOTSUPP || rc == -EACCES) + rc = 0; + + de_ccw->count = sizeof(struct DE_eckd_data); + de_ccw->flags |= CCW_FLAG_SLI; + return rc; +} + +static int +define_extent(struct ccw1 *ccw, struct DE_eckd_data *data, unsigned int trk, + unsigned int totrk, int cmd, struct dasd_device *device) +{ + struct dasd_eckd_private *private; + u32 begcyl, endcyl; + u16 heads, beghead, endhead; + int rc = 0; + + private = (struct dasd_eckd_private *) device->private; + + ccw->cmd_code = DASD_ECKD_CCW_DEFINE_EXTENT; + ccw->flags = 0; + ccw->count = 16; + ccw->cda = (__u32) __pa(data); + + memset(data, 0, sizeof(struct DE_eckd_data)); + switch (cmd) { + case DASD_ECKD_CCW_READ_HOME_ADDRESS: + case DASD_ECKD_CCW_READ_RECORD_ZERO: + case DASD_ECKD_CCW_READ: + case DASD_ECKD_CCW_READ_MT: + case DASD_ECKD_CCW_READ_CKD: + case DASD_ECKD_CCW_READ_CKD_MT: + case DASD_ECKD_CCW_READ_KD: + case DASD_ECKD_CCW_READ_KD_MT: + case DASD_ECKD_CCW_READ_COUNT: + data->mask.perm = 0x1; + data->attributes.operation = private->attrib.operation; + break; + case DASD_ECKD_CCW_WRITE: + case DASD_ECKD_CCW_WRITE_MT: + case DASD_ECKD_CCW_WRITE_KD: + case DASD_ECKD_CCW_WRITE_KD_MT: + data->mask.perm = 0x02; + data->attributes.operation = private->attrib.operation; + rc = check_XRC (ccw, data, device); + break; + case DASD_ECKD_CCW_WRITE_CKD: + case DASD_ECKD_CCW_WRITE_CKD_MT: + data->attributes.operation = DASD_BYPASS_CACHE; + rc = check_XRC (ccw, data, device); + break; + case DASD_ECKD_CCW_ERASE: + case DASD_ECKD_CCW_WRITE_HOME_ADDRESS: + case DASD_ECKD_CCW_WRITE_RECORD_ZERO: + data->mask.perm = 0x3; + data->mask.auth = 0x1; + data->attributes.operation = DASD_BYPASS_CACHE; + rc = check_XRC (ccw, data, device); + break; + default: + dev_err(&device->cdev->dev, + "0x%x is not a known command\n", cmd); + break; + } + + data->attributes.mode = 0x3; /* ECKD */ + + if ((private->rdc_data.cu_type == 0x2105 || + private->rdc_data.cu_type == 0x2107 || + private->rdc_data.cu_type == 0x1750) + && !(private->uses_cdl && trk < 2)) + data->ga_extended |= 0x40; /* Regular Data Format Mode */ + + heads = private->rdc_data.trk_per_cyl; + begcyl = trk / heads; + beghead = trk % heads; + endcyl = totrk / heads; + endhead = totrk % heads; + + /* check for sequential prestage - enhance cylinder range */ + if (data->attributes.operation == DASD_SEQ_PRESTAGE || + data->attributes.operation == DASD_SEQ_ACCESS) { + + if (endcyl + private->attrib.nr_cyl < private->real_cyl) + endcyl += private->attrib.nr_cyl; + else + endcyl = (private->real_cyl - 1); + } + + set_ch_t(&data->beg_ext, begcyl, beghead); + set_ch_t(&data->end_ext, endcyl, endhead); + return rc; +} + +static int check_XRC_on_prefix(struct PFX_eckd_data *pfxdata, + struct dasd_device *device) +{ + struct dasd_eckd_private *private; + int rc; + + private = (struct dasd_eckd_private *) device->private; + if (!private->rdc_data.facilities.XRC_supported) + return 0; + + /* switch on System Time Stamp - needed for XRC Support */ + pfxdata->define_extent.ga_extended |= 0x08; /* 'Time Stamp Valid' */ + pfxdata->define_extent.ga_extended |= 0x02; /* 'Extended Parameter' */ + pfxdata->validity.time_stamp = 1; /* 'Time Stamp Valid' */ + + rc = get_sync_clock(&pfxdata->define_extent.ep_sys_time); + /* Ignore return code if sync clock is switched off. */ + if (rc == -EOPNOTSUPP || rc == -EACCES) + rc = 0; + return rc; +} + +static void fill_LRE_data(struct LRE_eckd_data *data, unsigned int trk, + unsigned int rec_on_trk, int count, int cmd, + struct dasd_device *device, unsigned int reclen, + unsigned int tlf) +{ + struct dasd_eckd_private *private; + int sector; + int dn, d; + + private = (struct dasd_eckd_private *) device->private; + + memset(data, 0, sizeof(*data)); + sector = 0; + if (rec_on_trk) { + switch (private->rdc_data.dev_type) { + case 0x3390: + dn = ceil_quot(reclen + 6, 232); + d = 9 + ceil_quot(reclen + 6 * (dn + 1), 34); + sector = (49 + (rec_on_trk - 1) * (10 + d)) / 8; + break; + case 0x3380: + d = 7 + ceil_quot(reclen + 12, 32); + sector = (39 + (rec_on_trk - 1) * (8 + d)) / 7; + break; + } + } + data->sector = sector; + /* note: meaning of count depends on the operation + * for record based I/O it's the number of records, but for + * track based I/O it's the number of tracks + */ + data->count = count; + switch (cmd) { + case DASD_ECKD_CCW_WRITE_HOME_ADDRESS: + data->operation.orientation = 0x3; + data->operation.operation = 0x03; + break; + case DASD_ECKD_CCW_READ_HOME_ADDRESS: + data->operation.orientation = 0x3; + data->operation.operation = 0x16; + break; + case DASD_ECKD_CCW_WRITE_RECORD_ZERO: + data->operation.orientation = 0x1; + data->operation.operation = 0x03; + data->count++; + break; + case DASD_ECKD_CCW_READ_RECORD_ZERO: + data->operation.orientation = 0x3; + data->operation.operation = 0x16; + data->count++; + break; + case DASD_ECKD_CCW_WRITE: + case DASD_ECKD_CCW_WRITE_MT: + case DASD_ECKD_CCW_WRITE_KD: + case DASD_ECKD_CCW_WRITE_KD_MT: + data->auxiliary.length_valid = 0x1; + data->length = reclen; + data->operation.operation = 0x01; + break; + case DASD_ECKD_CCW_WRITE_CKD: + case DASD_ECKD_CCW_WRITE_CKD_MT: + data->auxiliary.length_valid = 0x1; + data->length = reclen; + data->operation.operation = 0x03; + break; + case DASD_ECKD_CCW_WRITE_FULL_TRACK: + data->operation.orientation = 0x0; + data->operation.operation = 0x3F; + data->extended_operation = 0x11; + data->length = 0; + data->extended_parameter_length = 0x02; + if (data->count > 8) { + data->extended_parameter[0] = 0xFF; + data->extended_parameter[1] = 0xFF; + data->extended_parameter[1] <<= (16 - count); + } else { + data->extended_parameter[0] = 0xFF; + data->extended_parameter[0] <<= (8 - count); + data->extended_parameter[1] = 0x00; + } + data->sector = 0xFF; + break; + case DASD_ECKD_CCW_WRITE_TRACK_DATA: + data->auxiliary.length_valid = 0x1; + data->length = reclen; /* not tlf, as one might think */ + data->operation.operation = 0x3F; + data->extended_operation = 0x23; + break; + case DASD_ECKD_CCW_READ: + case DASD_ECKD_CCW_READ_MT: + case DASD_ECKD_CCW_READ_KD: + case DASD_ECKD_CCW_READ_KD_MT: + data->auxiliary.length_valid = 0x1; + data->length = reclen; + data->operation.operation = 0x06; + break; + case DASD_ECKD_CCW_READ_CKD: + case DASD_ECKD_CCW_READ_CKD_MT: + data->auxiliary.length_valid = 0x1; + data->length = reclen; + data->operation.operation = 0x16; + break; + case DASD_ECKD_CCW_READ_COUNT: + data->operation.operation = 0x06; + break; + case DASD_ECKD_CCW_READ_TRACK: + data->operation.orientation = 0x1; + data->operation.operation = 0x0C; + data->extended_parameter_length = 0; + data->sector = 0xFF; + break; + case DASD_ECKD_CCW_READ_TRACK_DATA: + data->auxiliary.length_valid = 0x1; + data->length = tlf; + data->operation.operation = 0x0C; + break; + case DASD_ECKD_CCW_ERASE: + data->length = reclen; + data->auxiliary.length_valid = 0x1; + data->operation.operation = 0x0b; + break; + default: + DBF_DEV_EVENT(DBF_ERR, device, + "fill LRE unknown opcode 0x%x", cmd); + BUG(); + } + set_ch_t(&data->seek_addr, + trk / private->rdc_data.trk_per_cyl, + trk % private->rdc_data.trk_per_cyl); + data->search_arg.cyl = data->seek_addr.cyl; + data->search_arg.head = data->seek_addr.head; + data->search_arg.record = rec_on_trk; +} + +static int prefix_LRE(struct ccw1 *ccw, struct PFX_eckd_data *pfxdata, + unsigned int trk, unsigned int totrk, int cmd, + struct dasd_device *basedev, struct dasd_device *startdev, + unsigned char format, unsigned int rec_on_trk, int count, + unsigned int blksize, unsigned int tlf) +{ + struct dasd_eckd_private *basepriv, *startpriv; + struct DE_eckd_data *dedata; + struct LRE_eckd_data *lredata; + u32 begcyl, endcyl; + u16 heads, beghead, endhead; + int rc = 0; + + basepriv = (struct dasd_eckd_private *) basedev->private; + startpriv = (struct dasd_eckd_private *) startdev->private; + dedata = &pfxdata->define_extent; + lredata = &pfxdata->locate_record; + + ccw->cmd_code = DASD_ECKD_CCW_PFX; + ccw->flags = 0; + if (cmd == DASD_ECKD_CCW_WRITE_FULL_TRACK) { + ccw->count = sizeof(*pfxdata) + 2; + ccw->cda = (__u32) __pa(pfxdata); + memset(pfxdata, 0, sizeof(*pfxdata) + 2); + } else { + ccw->count = sizeof(*pfxdata); + ccw->cda = (__u32) __pa(pfxdata); + memset(pfxdata, 0, sizeof(*pfxdata)); + } + + /* prefix data */ + if (format > 1) { + DBF_DEV_EVENT(DBF_ERR, basedev, + "PFX LRE unknown format 0x%x", format); + BUG(); + return -EINVAL; + } + pfxdata->format = format; + pfxdata->base_address = basepriv->ned->unit_addr; + pfxdata->base_lss = basepriv->ned->ID; + pfxdata->validity.define_extent = 1; + + /* private uid is kept up to date, conf_data may be outdated */ + if (startpriv->uid.type != UA_BASE_DEVICE) { + pfxdata->validity.verify_base = 1; + if (startpriv->uid.type == UA_HYPER_PAV_ALIAS) + pfxdata->validity.hyper_pav = 1; + } + + /* define extend data (mostly)*/ + switch (cmd) { + case DASD_ECKD_CCW_READ_HOME_ADDRESS: + case DASD_ECKD_CCW_READ_RECORD_ZERO: + case DASD_ECKD_CCW_READ: + case DASD_ECKD_CCW_READ_MT: + case DASD_ECKD_CCW_READ_CKD: + case DASD_ECKD_CCW_READ_CKD_MT: + case DASD_ECKD_CCW_READ_KD: + case DASD_ECKD_CCW_READ_KD_MT: + case DASD_ECKD_CCW_READ_COUNT: + dedata->mask.perm = 0x1; + dedata->attributes.operation = basepriv->attrib.operation; + break; + case DASD_ECKD_CCW_READ_TRACK: + case DASD_ECKD_CCW_READ_TRACK_DATA: + dedata->mask.perm = 0x1; + dedata->attributes.operation = basepriv->attrib.operation; + dedata->blk_size = 0; + break; + case DASD_ECKD_CCW_WRITE: + case DASD_ECKD_CCW_WRITE_MT: + case DASD_ECKD_CCW_WRITE_KD: + case DASD_ECKD_CCW_WRITE_KD_MT: + dedata->mask.perm = 0x02; + dedata->attributes.operation = basepriv->attrib.operation; + rc = check_XRC_on_prefix(pfxdata, basedev); + break; + case DASD_ECKD_CCW_WRITE_CKD: + case DASD_ECKD_CCW_WRITE_CKD_MT: + dedata->attributes.operation = DASD_BYPASS_CACHE; + rc = check_XRC_on_prefix(pfxdata, basedev); + break; + case DASD_ECKD_CCW_ERASE: + case DASD_ECKD_CCW_WRITE_HOME_ADDRESS: + case DASD_ECKD_CCW_WRITE_RECORD_ZERO: + dedata->mask.perm = 0x3; + dedata->mask.auth = 0x1; + dedata->attributes.operation = DASD_BYPASS_CACHE; + rc = check_XRC_on_prefix(pfxdata, basedev); + break; + case DASD_ECKD_CCW_WRITE_FULL_TRACK: + dedata->mask.perm = 0x03; + dedata->attributes.operation = basepriv->attrib.operation; + dedata->blk_size = 0; + break; + case DASD_ECKD_CCW_WRITE_TRACK_DATA: + dedata->mask.perm = 0x02; + dedata->attributes.operation = basepriv->attrib.operation; + dedata->blk_size = blksize; + rc = check_XRC_on_prefix(pfxdata, basedev); + break; + default: + DBF_DEV_EVENT(DBF_ERR, basedev, + "PFX LRE unknown opcode 0x%x", cmd); + BUG(); + return -EINVAL; + } + + dedata->attributes.mode = 0x3; /* ECKD */ + + if ((basepriv->rdc_data.cu_type == 0x2105 || + basepriv->rdc_data.cu_type == 0x2107 || + basepriv->rdc_data.cu_type == 0x1750) + && !(basepriv->uses_cdl && trk < 2)) + dedata->ga_extended |= 0x40; /* Regular Data Format Mode */ + + heads = basepriv->rdc_data.trk_per_cyl; + begcyl = trk / heads; + beghead = trk % heads; + endcyl = totrk / heads; + endhead = totrk % heads; + + /* check for sequential prestage - enhance cylinder range */ + if (dedata->attributes.operation == DASD_SEQ_PRESTAGE || + dedata->attributes.operation == DASD_SEQ_ACCESS) { + + if (endcyl + basepriv->attrib.nr_cyl < basepriv->real_cyl) + endcyl += basepriv->attrib.nr_cyl; + else + endcyl = (basepriv->real_cyl - 1); + } + + set_ch_t(&dedata->beg_ext, begcyl, beghead); + set_ch_t(&dedata->end_ext, endcyl, endhead); + + if (format == 1) { + fill_LRE_data(lredata, trk, rec_on_trk, count, cmd, + basedev, blksize, tlf); + } + + return rc; +} + +static int prefix(struct ccw1 *ccw, struct PFX_eckd_data *pfxdata, + unsigned int trk, unsigned int totrk, int cmd, + struct dasd_device *basedev, struct dasd_device *startdev) +{ + return prefix_LRE(ccw, pfxdata, trk, totrk, cmd, basedev, startdev, + 0, 0, 0, 0, 0); +} + +static void +locate_record(struct ccw1 *ccw, struct LO_eckd_data *data, unsigned int trk, + unsigned int rec_on_trk, int no_rec, int cmd, + struct dasd_device * device, int reclen) +{ + struct dasd_eckd_private *private; + int sector; + int dn, d; + + private = (struct dasd_eckd_private *) device->private; + + DBF_DEV_EVENT(DBF_INFO, device, + "Locate: trk %d, rec %d, no_rec %d, cmd %d, reclen %d", + trk, rec_on_trk, no_rec, cmd, reclen); + + ccw->cmd_code = DASD_ECKD_CCW_LOCATE_RECORD; + ccw->flags = 0; + ccw->count = 16; + ccw->cda = (__u32) __pa(data); + + memset(data, 0, sizeof(struct LO_eckd_data)); + sector = 0; + if (rec_on_trk) { + switch (private->rdc_data.dev_type) { + case 0x3390: + dn = ceil_quot(reclen + 6, 232); + d = 9 + ceil_quot(reclen + 6 * (dn + 1), 34); + sector = (49 + (rec_on_trk - 1) * (10 + d)) / 8; + break; + case 0x3380: + d = 7 + ceil_quot(reclen + 12, 32); + sector = (39 + (rec_on_trk - 1) * (8 + d)) / 7; + break; + } + } + data->sector = sector; + data->count = no_rec; + switch (cmd) { + case DASD_ECKD_CCW_WRITE_HOME_ADDRESS: + data->operation.orientation = 0x3; + data->operation.operation = 0x03; + break; + case DASD_ECKD_CCW_READ_HOME_ADDRESS: + data->operation.orientation = 0x3; + data->operation.operation = 0x16; + break; + case DASD_ECKD_CCW_WRITE_RECORD_ZERO: + data->operation.orientation = 0x1; + data->operation.operation = 0x03; + data->count++; + break; + case DASD_ECKD_CCW_READ_RECORD_ZERO: + data->operation.orientation = 0x3; + data->operation.operation = 0x16; + data->count++; + break; + case DASD_ECKD_CCW_WRITE: + case DASD_ECKD_CCW_WRITE_MT: + case DASD_ECKD_CCW_WRITE_KD: + case DASD_ECKD_CCW_WRITE_KD_MT: + data->auxiliary.last_bytes_used = 0x1; + data->length = reclen; + data->operation.operation = 0x01; + break; + case DASD_ECKD_CCW_WRITE_CKD: + case DASD_ECKD_CCW_WRITE_CKD_MT: + data->auxiliary.last_bytes_used = 0x1; + data->length = reclen; + data->operation.operation = 0x03; + break; + case DASD_ECKD_CCW_READ: + case DASD_ECKD_CCW_READ_MT: + case DASD_ECKD_CCW_READ_KD: + case DASD_ECKD_CCW_READ_KD_MT: + data->auxiliary.last_bytes_used = 0x1; + data->length = reclen; + data->operation.operation = 0x06; + break; + case DASD_ECKD_CCW_READ_CKD: + case DASD_ECKD_CCW_READ_CKD_MT: + data->auxiliary.last_bytes_used = 0x1; + data->length = reclen; + data->operation.operation = 0x16; + break; + case DASD_ECKD_CCW_READ_COUNT: + data->operation.operation = 0x06; + break; + case DASD_ECKD_CCW_ERASE: + data->length = reclen; + data->auxiliary.last_bytes_used = 0x1; + data->operation.operation = 0x0b; + break; + default: + DBF_DEV_EVENT(DBF_ERR, device, "unknown locate record " + "opcode 0x%x", cmd); + } + set_ch_t(&data->seek_addr, + trk / private->rdc_data.trk_per_cyl, + trk % private->rdc_data.trk_per_cyl); + data->search_arg.cyl = data->seek_addr.cyl; + data->search_arg.head = data->seek_addr.head; + data->search_arg.record = rec_on_trk; +} + +/* + * Returns 1 if the block is one of the special blocks that needs + * to get read/written with the KD variant of the command. + * That is DASD_ECKD_READ_KD_MT instead of DASD_ECKD_READ_MT and + * DASD_ECKD_WRITE_KD_MT instead of DASD_ECKD_WRITE_MT. + * Luckily the KD variants differ only by one bit (0x08) from the + * normal variant. So don't wonder about code like: + * if (dasd_eckd_cdl_special(blk_per_trk, recid)) + * ccw->cmd_code |= 0x8; + */ +static inline int +dasd_eckd_cdl_special(int blk_per_trk, int recid) +{ + if (recid < 3) + return 1; + if (recid < blk_per_trk) + return 0; + if (recid < 2 * blk_per_trk) + return 1; + return 0; +} + +/* + * Returns the record size for the special blocks of the cdl format. + * Only returns something useful if dasd_eckd_cdl_special is true + * for the recid. + */ +static inline int +dasd_eckd_cdl_reclen(int recid) +{ + if (recid < 3) + return sizes_trk0[recid]; + return LABEL_SIZE; +} +/* create unique id from private structure. */ +static void create_uid(struct dasd_eckd_private *private) +{ + int count; + struct dasd_uid *uid; + + uid = &private->uid; + memset(uid, 0, sizeof(struct dasd_uid)); + memcpy(uid->vendor, private->ned->HDA_manufacturer, + sizeof(uid->vendor) - 1); + EBCASC(uid->vendor, sizeof(uid->vendor) - 1); + memcpy(uid->serial, private->ned->HDA_location, + sizeof(uid->serial) - 1); + EBCASC(uid->serial, sizeof(uid->serial) - 1); + uid->ssid = private->gneq->subsystemID; + uid->real_unit_addr = private->ned->unit_addr; + if (private->sneq) { + uid->type = private->sneq->sua_flags; + if (uid->type == UA_BASE_PAV_ALIAS) + uid->base_unit_addr = private->sneq->base_unit_addr; + } else { + uid->type = UA_BASE_DEVICE; + } + if (private->vdsneq) { + for (count = 0; count < 16; count++) { + sprintf(uid->vduit+2*count, "%02x", + private->vdsneq->uit[count]); + } + } +} + +/* + * Generate device unique id that specifies the physical device. + */ +static int dasd_eckd_generate_uid(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + unsigned long flags; + + private = (struct dasd_eckd_private *) device->private; + if (!private) + return -ENODEV; + if (!private->ned || !private->gneq) + return -ENODEV; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + create_uid(private); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + return 0; +} + +static int dasd_eckd_get_uid(struct dasd_device *device, struct dasd_uid *uid) +{ + struct dasd_eckd_private *private; + unsigned long flags; + + if (device->private) { + private = (struct dasd_eckd_private *)device->private; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + *uid = private->uid; + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + return 0; + } + return -EINVAL; +} + +/* + * compare device UID with data of a given dasd_eckd_private structure + * return 0 for match + */ +static int dasd_eckd_compare_path_uid(struct dasd_device *device, + struct dasd_eckd_private *private) +{ + struct dasd_uid device_uid; + + create_uid(private); + dasd_eckd_get_uid(device, &device_uid); + + return memcmp(&device_uid, &private->uid, sizeof(struct dasd_uid)); +} + +static void dasd_eckd_fill_rcd_cqr(struct dasd_device *device, + struct dasd_ccw_req *cqr, + __u8 *rcd_buffer, + __u8 lpm) +{ + struct ccw1 *ccw; + /* + * buffer has to start with EBCDIC "V1.0" to show + * support for virtual device SNEQ + */ + rcd_buffer[0] = 0xE5; + rcd_buffer[1] = 0xF1; + rcd_buffer[2] = 0x4B; + rcd_buffer[3] = 0xF0; + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_RCD; + ccw->flags = 0; + ccw->cda = (__u32)(addr_t)rcd_buffer; + ccw->count = DASD_ECKD_RCD_DATA_SIZE; + cqr->magic = DASD_ECKD_MAGIC; + + cqr->startdev = device; + cqr->memdev = device; + cqr->block = NULL; + cqr->expires = 10*HZ; + cqr->lpm = lpm; + cqr->retries = 256; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + set_bit(DASD_CQR_VERIFY_PATH, &cqr->flags); +} + +/* + * Wakeup helper for read_conf + * if the cqr is not done and needs some error recovery + * the buffer has to be re-initialized with the EBCDIC "V1.0" + * to show support for virtual device SNEQ + */ +static void read_conf_cb(struct dasd_ccw_req *cqr, void *data) +{ + struct ccw1 *ccw; + __u8 *rcd_buffer; + + if (cqr->status != DASD_CQR_DONE) { + ccw = cqr->cpaddr; + rcd_buffer = (__u8 *)((addr_t) ccw->cda); + memset(rcd_buffer, 0, sizeof(*rcd_buffer)); + + rcd_buffer[0] = 0xE5; + rcd_buffer[1] = 0xF1; + rcd_buffer[2] = 0x4B; + rcd_buffer[3] = 0xF0; + } + dasd_wakeup_cb(cqr, data); +} + +static int dasd_eckd_read_conf_immediately(struct dasd_device *device, + struct dasd_ccw_req *cqr, + __u8 *rcd_buffer, + __u8 lpm) +{ + struct ciw *ciw; + int rc; + /* + * sanity check: scan for RCD command in extended SenseID data + * some devices do not support RCD + */ + ciw = ccw_device_get_ciw(device->cdev, CIW_TYPE_RCD); + if (!ciw || ciw->cmd != DASD_ECKD_CCW_RCD) + return -EOPNOTSUPP; + + dasd_eckd_fill_rcd_cqr(device, cqr, rcd_buffer, lpm); + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + set_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags); + cqr->retries = 5; + cqr->callback = read_conf_cb; + rc = dasd_sleep_on_immediatly(cqr); + return rc; +} + +static int dasd_eckd_read_conf_lpm(struct dasd_device *device, + void **rcd_buffer, + int *rcd_buffer_size, __u8 lpm) +{ + struct ciw *ciw; + char *rcd_buf = NULL; + int ret; + struct dasd_ccw_req *cqr; + + /* + * sanity check: scan for RCD command in extended SenseID data + * some devices do not support RCD + */ + ciw = ccw_device_get_ciw(device->cdev, CIW_TYPE_RCD); + if (!ciw || ciw->cmd != DASD_ECKD_CCW_RCD) { + ret = -EOPNOTSUPP; + goto out_error; + } + rcd_buf = kzalloc(DASD_ECKD_RCD_DATA_SIZE, GFP_KERNEL | GFP_DMA); + if (!rcd_buf) { + ret = -ENOMEM; + goto out_error; + } + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1 /* RCD */, + 0, /* use rcd_buf as data ara */ + device); + if (IS_ERR(cqr)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Could not allocate RCD request"); + ret = -ENOMEM; + goto out_error; + } + dasd_eckd_fill_rcd_cqr(device, cqr, rcd_buf, lpm); + cqr->callback = read_conf_cb; + ret = dasd_sleep_on(cqr); + /* + * on success we update the user input parms + */ + dasd_sfree_request(cqr, cqr->memdev); + if (ret) + goto out_error; + + *rcd_buffer_size = DASD_ECKD_RCD_DATA_SIZE; + *rcd_buffer = rcd_buf; + return 0; +out_error: + kfree(rcd_buf); + *rcd_buffer = NULL; + *rcd_buffer_size = 0; + return ret; +} + +static int dasd_eckd_identify_conf_parts(struct dasd_eckd_private *private) +{ + + struct dasd_sneq *sneq; + int i, count; + + private->ned = NULL; + private->sneq = NULL; + private->vdsneq = NULL; + private->gneq = NULL; + count = private->conf_len / sizeof(struct dasd_sneq); + sneq = (struct dasd_sneq *)private->conf_data; + for (i = 0; i < count; ++i) { + if (sneq->flags.identifier == 1 && sneq->format == 1) + private->sneq = sneq; + else if (sneq->flags.identifier == 1 && sneq->format == 4) + private->vdsneq = (struct vd_sneq *)sneq; + else if (sneq->flags.identifier == 2) + private->gneq = (struct dasd_gneq *)sneq; + else if (sneq->flags.identifier == 3 && sneq->res1 == 1) + private->ned = (struct dasd_ned *)sneq; + sneq++; + } + if (!private->ned || !private->gneq) { + private->ned = NULL; + private->sneq = NULL; + private->vdsneq = NULL; + private->gneq = NULL; + return -EINVAL; + } + return 0; + +}; + +static unsigned char dasd_eckd_path_access(void *conf_data, int conf_len) +{ + struct dasd_gneq *gneq; + int i, count, found; + + count = conf_len / sizeof(*gneq); + gneq = (struct dasd_gneq *)conf_data; + found = 0; + for (i = 0; i < count; ++i) { + if (gneq->flags.identifier == 2) { + found = 1; + break; + } + gneq++; + } + if (found) + return ((char *)gneq)[18] & 0x07; + else + return 0; +} + +static int dasd_eckd_read_conf(struct dasd_device *device) +{ + void *conf_data; + int conf_len, conf_data_saved; + int rc, path_err; + __u8 lpm, opm; + struct dasd_eckd_private *private, path_private; + struct dasd_path *path_data; + struct dasd_uid *uid; + char print_path_uid[60], print_device_uid[60]; + + private = (struct dasd_eckd_private *) device->private; + path_data = &device->path_data; + opm = ccw_device_get_path_mask(device->cdev); + conf_data_saved = 0; + path_err = 0; + /* get configuration data per operational path */ + for (lpm = 0x80; lpm; lpm>>= 1) { + if (!(lpm & opm)) + continue; + rc = dasd_eckd_read_conf_lpm(device, &conf_data, + &conf_len, lpm); + if (rc && rc != -EOPNOTSUPP) { /* -EOPNOTSUPP is ok */ + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, + "Read configuration data returned " + "error %d", rc); + return rc; + } + if (conf_data == NULL) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", + "No configuration data " + "retrieved"); + /* no further analysis possible */ + path_data->opm |= lpm; + continue; /* no error */ + } + /* save first valid configuration data */ + if (!conf_data_saved) { + kfree(private->conf_data); + private->conf_data = conf_data; + private->conf_len = conf_len; + if (dasd_eckd_identify_conf_parts(private)) { + private->conf_data = NULL; + private->conf_len = 0; + kfree(conf_data); + continue; + } + /* + * build device UID that other path data + * can be compared to it + */ + dasd_eckd_generate_uid(device); + conf_data_saved++; + } else { + path_private.conf_data = conf_data; + path_private.conf_len = DASD_ECKD_RCD_DATA_SIZE; + if (dasd_eckd_identify_conf_parts( + &path_private)) { + path_private.conf_data = NULL; + path_private.conf_len = 0; + kfree(conf_data); + continue; + } + + if (dasd_eckd_compare_path_uid( + device, &path_private)) { + uid = &path_private.uid; + if (strlen(uid->vduit) > 0) + snprintf(print_path_uid, + sizeof(print_path_uid), + "%s.%s.%04x.%02x.%s", + uid->vendor, uid->serial, + uid->ssid, uid->real_unit_addr, + uid->vduit); + else + snprintf(print_path_uid, + sizeof(print_path_uid), + "%s.%s.%04x.%02x", + uid->vendor, uid->serial, + uid->ssid, + uid->real_unit_addr); + uid = &private->uid; + if (strlen(uid->vduit) > 0) + snprintf(print_device_uid, + sizeof(print_device_uid), + "%s.%s.%04x.%02x.%s", + uid->vendor, uid->serial, + uid->ssid, uid->real_unit_addr, + uid->vduit); + else + snprintf(print_device_uid, + sizeof(print_device_uid), + "%s.%s.%04x.%02x", + uid->vendor, uid->serial, + uid->ssid, + uid->real_unit_addr); + dev_err(&device->cdev->dev, + "Not all channel paths lead to " + "the same device, path %02X leads to " + "device %s instead of %s\n", lpm, + print_path_uid, print_device_uid); + path_err = -EINVAL; + path_data->cablepm |= lpm; + continue; + } + + path_private.conf_data = NULL; + path_private.conf_len = 0; + } + switch (dasd_eckd_path_access(conf_data, conf_len)) { + case 0x02: + path_data->npm |= lpm; + break; + case 0x03: + path_data->ppm |= lpm; + break; + } + path_data->opm |= lpm; + /* + * if the path is used + * it should not be in one of the negative lists + */ + path_data->cablepm &= ~lpm; + path_data->hpfpm &= ~lpm; + path_data->cuirpm &= ~lpm; + + if (conf_data != private->conf_data) + kfree(conf_data); + } + + return path_err; +} + +static int verify_fcx_max_data(struct dasd_device *device, __u8 lpm) +{ + struct dasd_eckd_private *private; + int mdc; + u32 fcx_max_data; + + private = (struct dasd_eckd_private *) device->private; + if (private->fcx_max_data) { + mdc = ccw_device_get_mdc(device->cdev, lpm); + if ((mdc < 0)) { + dev_warn(&device->cdev->dev, + "Detecting the maximum data size for zHPF " + "requests failed (rc=%d) for a new path %x\n", + mdc, lpm); + return mdc; + } + fcx_max_data = mdc * FCX_MAX_DATA_FACTOR; + if (fcx_max_data < private->fcx_max_data) { + dev_warn(&device->cdev->dev, + "The maximum data size for zHPF requests %u " + "on a new path %x is below the active maximum " + "%u\n", fcx_max_data, lpm, + private->fcx_max_data); + return -EACCES; + } + } + return 0; +} + +static int rebuild_device_uid(struct dasd_device *device, + struct path_verification_work_data *data) +{ + struct dasd_eckd_private *private; + struct dasd_path *path_data; + __u8 lpm, opm; + int rc; + + rc = -ENODEV; + private = (struct dasd_eckd_private *) device->private; + path_data = &device->path_data; + opm = device->path_data.opm; + + for (lpm = 0x80; lpm; lpm >>= 1) { + if (!(lpm & opm)) + continue; + memset(&data->rcd_buffer, 0, sizeof(data->rcd_buffer)); + memset(&data->cqr, 0, sizeof(data->cqr)); + data->cqr.cpaddr = &data->ccw; + rc = dasd_eckd_read_conf_immediately(device, &data->cqr, + data->rcd_buffer, + lpm); + + if (rc) { + if (rc == -EOPNOTSUPP) /* -EOPNOTSUPP is ok */ + continue; + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, + "Read configuration data " + "returned error %d", rc); + break; + } + memcpy(private->conf_data, data->rcd_buffer, + DASD_ECKD_RCD_DATA_SIZE); + if (dasd_eckd_identify_conf_parts(private)) { + rc = -ENODEV; + } else /* first valid path is enough */ + break; + } + + if (!rc) + rc = dasd_eckd_generate_uid(device); + + return rc; +} + +static void do_path_verification_work(struct work_struct *work) +{ + struct path_verification_work_data *data; + struct dasd_device *device; + struct dasd_eckd_private path_private; + struct dasd_uid *uid; + __u8 path_rcd_buf[DASD_ECKD_RCD_DATA_SIZE]; + __u8 lpm, opm, npm, ppm, epm, hpfpm, cablepm; + unsigned long flags; + char print_uid[60]; + int rc; + + data = container_of(work, struct path_verification_work_data, worker); + device = data->device; + + /* delay path verification until device was resumed */ + if (test_bit(DASD_FLAG_SUSPENDED, &device->flags)) { + schedule_work(work); + return; + } + + opm = 0; + npm = 0; + ppm = 0; + epm = 0; + hpfpm = 0; + cablepm = 0; + + for (lpm = 0x80; lpm; lpm >>= 1) { + if (!(lpm & data->tbvpm)) + continue; + memset(&data->rcd_buffer, 0, sizeof(data->rcd_buffer)); + memset(&data->cqr, 0, sizeof(data->cqr)); + data->cqr.cpaddr = &data->ccw; + rc = dasd_eckd_read_conf_immediately(device, &data->cqr, + data->rcd_buffer, + lpm); + if (!rc) { + switch (dasd_eckd_path_access(data->rcd_buffer, + DASD_ECKD_RCD_DATA_SIZE) + ) { + case 0x02: + npm |= lpm; + break; + case 0x03: + ppm |= lpm; + break; + } + opm |= lpm; + } else if (rc == -EOPNOTSUPP) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", + "path verification: No configuration " + "data retrieved"); + opm |= lpm; + } else if (rc == -EAGAIN) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", + "path verification: device is stopped," + " try again later"); + epm |= lpm; + } else { + dev_warn(&device->cdev->dev, + "Reading device feature codes failed " + "(rc=%d) for new path %x\n", rc, lpm); + continue; + } + if (verify_fcx_max_data(device, lpm)) { + opm &= ~lpm; + npm &= ~lpm; + ppm &= ~lpm; + hpfpm |= lpm; + continue; + } + + /* + * save conf_data for comparison after + * rebuild_device_uid may have changed + * the original data + */ + memcpy(&path_rcd_buf, data->rcd_buffer, + DASD_ECKD_RCD_DATA_SIZE); + path_private.conf_data = (void *) &path_rcd_buf; + path_private.conf_len = DASD_ECKD_RCD_DATA_SIZE; + if (dasd_eckd_identify_conf_parts(&path_private)) { + path_private.conf_data = NULL; + path_private.conf_len = 0; + continue; + } + + /* + * compare path UID with device UID only if at least + * one valid path is left + * in other case the device UID may have changed and + * the first working path UID will be used as device UID + */ + if (device->path_data.opm && + dasd_eckd_compare_path_uid(device, &path_private)) { + /* + * the comparison was not successful + * rebuild the device UID with at least one + * known path in case a z/VM hyperswap command + * has changed the device + * + * after this compare again + * + * if either the rebuild or the recompare fails + * the path can not be used + */ + if (rebuild_device_uid(device, data) || + dasd_eckd_compare_path_uid( + device, &path_private)) { + uid = &path_private.uid; + if (strlen(uid->vduit) > 0) + snprintf(print_uid, sizeof(print_uid), + "%s.%s.%04x.%02x.%s", + uid->vendor, uid->serial, + uid->ssid, uid->real_unit_addr, + uid->vduit); + else + snprintf(print_uid, sizeof(print_uid), + "%s.%s.%04x.%02x", + uid->vendor, uid->serial, + uid->ssid, + uid->real_unit_addr); + dev_err(&device->cdev->dev, + "The newly added channel path %02X " + "will not be used because it leads " + "to a different device %s\n", + lpm, print_uid); + opm &= ~lpm; + npm &= ~lpm; + ppm &= ~lpm; + cablepm |= lpm; + continue; + } + } + + /* + * There is a small chance that a path is lost again between + * above path verification and the following modification of + * the device opm mask. We could avoid that race here by using + * yet another path mask, but we rather deal with this unlikely + * situation in dasd_start_IO. + */ + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + if (!device->path_data.opm && opm) { + device->path_data.opm = opm; + device->path_data.cablepm &= ~opm; + device->path_data.cuirpm &= ~opm; + device->path_data.hpfpm &= ~opm; + dasd_generic_path_operational(device); + } else { + device->path_data.opm |= opm; + device->path_data.cablepm &= ~opm; + device->path_data.cuirpm &= ~opm; + device->path_data.hpfpm &= ~opm; + } + device->path_data.npm |= npm; + device->path_data.ppm |= ppm; + device->path_data.tbvpm |= epm; + device->path_data.cablepm |= cablepm; + device->path_data.hpfpm |= hpfpm; + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + } + + dasd_put_device(device); + if (data->isglobal) + mutex_unlock(&dasd_path_verification_mutex); + else + kfree(data); +} + +static int dasd_eckd_verify_path(struct dasd_device *device, __u8 lpm) +{ + struct path_verification_work_data *data; + + data = kmalloc(sizeof(*data), GFP_ATOMIC | GFP_DMA); + if (!data) { + if (mutex_trylock(&dasd_path_verification_mutex)) { + data = path_verification_worker; + data->isglobal = 1; + } else + return -ENOMEM; + } else { + memset(data, 0, sizeof(*data)); + data->isglobal = 0; + } + INIT_WORK(&data->worker, do_path_verification_work); + dasd_get_device(device); + data->device = device; + data->tbvpm = lpm; + schedule_work(&data->worker); + return 0; +} + +static int dasd_eckd_read_features(struct dasd_device *device) +{ + struct dasd_psf_prssd_data *prssdp; + struct dasd_rssd_features *features; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int rc; + struct dasd_eckd_private *private; + + private = (struct dasd_eckd_private *) device->private; + memset(&private->features, 0, sizeof(struct dasd_rssd_features)); + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1 /* PSF */ + 1 /* RSSD */, + (sizeof(struct dasd_psf_prssd_data) + + sizeof(struct dasd_rssd_features)), + device); + if (IS_ERR(cqr)) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", "Could not " + "allocate initialization request"); + return PTR_ERR(cqr); + } + cqr->startdev = device; + cqr->memdev = device; + cqr->block = NULL; + cqr->retries = 256; + cqr->expires = 10 * HZ; + + /* Prepare for Read Subsystem Data */ + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + memset(prssdp, 0, sizeof(struct dasd_psf_prssd_data)); + prssdp->order = PSF_ORDER_PRSSD; + prssdp->suborder = 0x41; /* Read Feature Codes */ + /* all other bytes of prssdp must be zero */ + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->count = sizeof(struct dasd_psf_prssd_data); + ccw->flags |= CCW_FLAG_CC; + ccw->cda = (__u32)(addr_t) prssdp; + + /* Read Subsystem Data - feature codes */ + features = (struct dasd_rssd_features *) (prssdp + 1); + memset(features, 0, sizeof(struct dasd_rssd_features)); + + ccw++; + ccw->cmd_code = DASD_ECKD_CCW_RSSD; + ccw->count = sizeof(struct dasd_rssd_features); + ccw->cda = (__u32)(addr_t) features; + + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + rc = dasd_sleep_on(cqr); + if (rc == 0) { + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + features = (struct dasd_rssd_features *) (prssdp + 1); + memcpy(&private->features, features, + sizeof(struct dasd_rssd_features)); + } else + dev_warn(&device->cdev->dev, "Reading device feature codes" + " failed with rc=%d\n", rc); + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + + +/* + * Build CP for Perform Subsystem Function - SSC. + */ +static struct dasd_ccw_req *dasd_eckd_build_psf_ssc(struct dasd_device *device, + int enable_pav) +{ + struct dasd_ccw_req *cqr; + struct dasd_psf_ssc_data *psf_ssc_data; + struct ccw1 *ccw; + + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1 /* PSF */ , + sizeof(struct dasd_psf_ssc_data), + device); + + if (IS_ERR(cqr)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Could not allocate PSF-SSC request"); + return cqr; + } + psf_ssc_data = (struct dasd_psf_ssc_data *)cqr->data; + psf_ssc_data->order = PSF_ORDER_SSC; + psf_ssc_data->suborder = 0xc0; + if (enable_pav) { + psf_ssc_data->suborder |= 0x08; + psf_ssc_data->reserved[0] = 0x88; + } + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->cda = (__u32)(addr_t)psf_ssc_data; + ccw->count = 66; + + cqr->startdev = device; + cqr->memdev = device; + cqr->block = NULL; + cqr->retries = 256; + cqr->expires = 10*HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +} + +/* + * Perform Subsystem Function. + * It is necessary to trigger CIO for channel revalidation since this + * call might change behaviour of DASD devices. + */ +static int +dasd_eckd_psf_ssc(struct dasd_device *device, int enable_pav, + unsigned long flags) +{ + struct dasd_ccw_req *cqr; + int rc; + + cqr = dasd_eckd_build_psf_ssc(device, enable_pav); + if (IS_ERR(cqr)) + return PTR_ERR(cqr); + + /* + * set flags e.g. turn on failfast, to prevent blocking + * the calling function should handle failed requests + */ + cqr->flags |= flags; + + rc = dasd_sleep_on(cqr); + if (!rc) + /* trigger CIO to reprobe devices */ + css_schedule_reprobe(); + else if (cqr->intrc == -EAGAIN) + rc = -EAGAIN; + + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * Valide storage server of current device. + */ +static int dasd_eckd_validate_server(struct dasd_device *device, + unsigned long flags) +{ + int rc; + struct dasd_eckd_private *private; + int enable_pav; + + private = (struct dasd_eckd_private *) device->private; + if (private->uid.type == UA_BASE_PAV_ALIAS || + private->uid.type == UA_HYPER_PAV_ALIAS) + return 0; + if (dasd_nopav || MACHINE_IS_VM) + enable_pav = 0; + else + enable_pav = 1; + rc = dasd_eckd_psf_ssc(device, enable_pav, flags); + + /* may be requested feature is not available on server, + * therefore just report error and go ahead */ + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "PSF-SSC for SSID %04x " + "returned rc=%d", private->uid.ssid, rc); + return rc; +} + +/* + * worker to do a validate server in case of a lost pathgroup + */ +static void dasd_eckd_do_validate_server(struct work_struct *work) +{ + struct dasd_device *device = container_of(work, struct dasd_device, + kick_validate); + unsigned long flags = 0; + + set_bit(DASD_CQR_FLAGS_FAILFAST, &flags); + if (dasd_eckd_validate_server(device, flags) + == -EAGAIN) { + /* schedule worker again if failed */ + schedule_work(&device->kick_validate); + return; + } + + dasd_put_device(device); +} + +static void dasd_eckd_kick_validate_server(struct dasd_device *device) +{ + dasd_get_device(device); + /* exit if device not online or in offline processing */ + if (test_bit(DASD_FLAG_OFFLINE, &device->flags) || + device->state < DASD_STATE_ONLINE) { + dasd_put_device(device); + return; + } + /* queue call to do_validate_server to the kernel event daemon. */ + if (!schedule_work(&device->kick_validate)) + dasd_put_device(device); +} + +static u32 get_fcx_max_data(struct dasd_device *device) +{ + int tpm, mdc; + int fcx_in_css, fcx_in_gneq, fcx_in_features; + struct dasd_eckd_private *private; + + if (dasd_nofcx) + return 0; + /* is transport mode supported? */ + private = (struct dasd_eckd_private *) device->private; + fcx_in_css = css_general_characteristics.fcx; + fcx_in_gneq = private->gneq->reserved2[7] & 0x04; + fcx_in_features = private->features.feature[40] & 0x80; + tpm = fcx_in_css && fcx_in_gneq && fcx_in_features; + + if (!tpm) + return 0; + + mdc = ccw_device_get_mdc(device->cdev, 0); + if (mdc < 0) { + dev_warn(&device->cdev->dev, "Detecting the maximum supported" + " data size for zHPF requests failed\n"); + return 0; + } else + return mdc * FCX_MAX_DATA_FACTOR; +} + +/* + * Check device characteristics. + * If the device is accessible using ECKD discipline, the device is enabled. + */ +static int +dasd_eckd_check_characteristics(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + struct dasd_block *block; + struct dasd_uid temp_uid; + int rc, i; + int readonly; + unsigned long value; + + /* setup work queue for validate server*/ + INIT_WORK(&device->kick_validate, dasd_eckd_do_validate_server); + + if (!ccw_device_is_pathgroup(device->cdev)) { + dev_warn(&device->cdev->dev, + "A channel path group could not be established\n"); + return -EIO; + } + if (!ccw_device_is_multipath(device->cdev)) { + dev_info(&device->cdev->dev, + "The DASD is not operating in multipath mode\n"); + } + private = (struct dasd_eckd_private *) device->private; + if (!private) { + private = kzalloc(sizeof(*private), GFP_KERNEL | GFP_DMA); + if (!private) { + dev_warn(&device->cdev->dev, + "Allocating memory for private DASD data " + "failed\n"); + return -ENOMEM; + } + device->private = (void *) private; + } else { + memset(private, 0, sizeof(*private)); + } + /* Invalidate status of initial analysis. */ + private->init_cqr_status = -1; + /* Set default cache operations. */ + private->attrib.operation = DASD_NORMAL_CACHE; + private->attrib.nr_cyl = 0; + + /* Read Configuration Data */ + rc = dasd_eckd_read_conf(device); + if (rc) + goto out_err1; + + /* set default timeout */ + device->default_expires = DASD_EXPIRES; + /* set default retry count */ + device->default_retries = DASD_RETRIES; + + if (private->gneq) { + value = 1; + for (i = 0; i < private->gneq->timeout.value; i++) + value = 10 * value; + value = value * private->gneq->timeout.number; + /* do not accept useless values */ + if (value != 0 && value <= DASD_EXPIRES_MAX) + device->default_expires = value; + } + + dasd_eckd_get_uid(device, &temp_uid); + if (temp_uid.type == UA_BASE_DEVICE) { + block = dasd_alloc_block(); + if (IS_ERR(block)) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", + "could not allocate dasd " + "block structure"); + rc = PTR_ERR(block); + goto out_err1; + } + device->block = block; + block->base = device; + } + + /* register lcu with alias handling, enable PAV */ + rc = dasd_alias_make_device_known_to_lcu(device); + if (rc) + goto out_err2; + + dasd_eckd_validate_server(device, 0); + + /* device may report different configuration data after LCU setup */ + rc = dasd_eckd_read_conf(device); + if (rc) + goto out_err3; + + /* Read Feature Codes */ + dasd_eckd_read_features(device); + + /* Read Device Characteristics */ + rc = dasd_generic_read_dev_chars(device, DASD_ECKD_MAGIC, + &private->rdc_data, 64); + if (rc) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, + "Read device characteristic failed, rc=%d", rc); + goto out_err3; + } + + if ((device->features & DASD_FEATURE_USERAW) && + !(private->rdc_data.facilities.RT_in_LR)) { + dev_err(&device->cdev->dev, "The storage server does not " + "support raw-track access\n"); + rc = -EINVAL; + goto out_err3; + } + + /* find the valid cylinder size */ + if (private->rdc_data.no_cyl == LV_COMPAT_CYL && + private->rdc_data.long_no_cyl) + private->real_cyl = private->rdc_data.long_no_cyl; + else + private->real_cyl = private->rdc_data.no_cyl; + + private->fcx_max_data = get_fcx_max_data(device); + + readonly = dasd_device_is_ro(device); + if (readonly) + set_bit(DASD_FLAG_DEVICE_RO, &device->flags); + + dev_info(&device->cdev->dev, "New DASD %04X/%02X (CU %04X/%02X) " + "with %d cylinders, %d heads, %d sectors%s\n", + private->rdc_data.dev_type, + private->rdc_data.dev_model, + private->rdc_data.cu_type, + private->rdc_data.cu_model.model, + private->real_cyl, + private->rdc_data.trk_per_cyl, + private->rdc_data.sec_per_trk, + readonly ? ", read-only device" : ""); + return 0; + +out_err3: + dasd_alias_disconnect_device_from_lcu(device); +out_err2: + dasd_free_block(device->block); + device->block = NULL; +out_err1: + kfree(private->conf_data); + kfree(device->private); + device->private = NULL; + return rc; +} + +static void dasd_eckd_uncheck_device(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + + private = (struct dasd_eckd_private *) device->private; + dasd_alias_disconnect_device_from_lcu(device); + private->ned = NULL; + private->sneq = NULL; + private->vdsneq = NULL; + private->gneq = NULL; + private->conf_len = 0; + kfree(private->conf_data); + private->conf_data = NULL; +} + +static struct dasd_ccw_req * +dasd_eckd_analysis_ccw(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + struct eckd_count *count_data; + struct LO_eckd_data *LO_data; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int cplength, datasize; + int i; + + private = (struct dasd_eckd_private *) device->private; + + cplength = 8; + datasize = sizeof(struct DE_eckd_data) + 2*sizeof(struct LO_eckd_data); + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, cplength, datasize, device); + if (IS_ERR(cqr)) + return cqr; + ccw = cqr->cpaddr; + /* Define extent for the first 3 tracks. */ + define_extent(ccw++, cqr->data, 0, 2, + DASD_ECKD_CCW_READ_COUNT, device); + LO_data = cqr->data + sizeof(struct DE_eckd_data); + /* Locate record for the first 4 records on track 0. */ + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, LO_data++, 0, 0, 4, + DASD_ECKD_CCW_READ_COUNT, device, 0); + + count_data = private->count_area; + for (i = 0; i < 4; i++) { + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = DASD_ECKD_CCW_READ_COUNT; + ccw->flags = 0; + ccw->count = 8; + ccw->cda = (__u32)(addr_t) count_data; + ccw++; + count_data++; + } + + /* Locate record for the first record on track 2. */ + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, LO_data++, 2, 0, 1, + DASD_ECKD_CCW_READ_COUNT, device, 0); + /* Read count ccw. */ + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = DASD_ECKD_CCW_READ_COUNT; + ccw->flags = 0; + ccw->count = 8; + ccw->cda = (__u32)(addr_t) count_data; + + cqr->block = NULL; + cqr->startdev = device; + cqr->memdev = device; + cqr->retries = 255; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +} + +/* differentiate between 'no record found' and any other error */ +static int dasd_eckd_analysis_evaluation(struct dasd_ccw_req *init_cqr) +{ + char *sense; + if (init_cqr->status == DASD_CQR_DONE) + return INIT_CQR_OK; + else if (init_cqr->status == DASD_CQR_NEED_ERP || + init_cqr->status == DASD_CQR_FAILED) { + sense = dasd_get_sense(&init_cqr->irb); + if (sense && (sense[1] & SNS1_NO_REC_FOUND)) + return INIT_CQR_UNFORMATTED; + else + return INIT_CQR_ERROR; + } else + return INIT_CQR_ERROR; +} + +/* + * This is the callback function for the init_analysis cqr. It saves + * the status of the initial analysis ccw before it frees it and kicks + * the device to continue the startup sequence. This will call + * dasd_eckd_do_analysis again (if the devices has not been marked + * for deletion in the meantime). + */ +static void dasd_eckd_analysis_callback(struct dasd_ccw_req *init_cqr, + void *data) +{ + struct dasd_eckd_private *private; + struct dasd_device *device; + + device = init_cqr->startdev; + private = (struct dasd_eckd_private *) device->private; + private->init_cqr_status = dasd_eckd_analysis_evaluation(init_cqr); + dasd_sfree_request(init_cqr, device); + dasd_kick_device(device); +} + +static int dasd_eckd_start_analysis(struct dasd_block *block) +{ + struct dasd_ccw_req *init_cqr; + + init_cqr = dasd_eckd_analysis_ccw(block->base); + if (IS_ERR(init_cqr)) + return PTR_ERR(init_cqr); + init_cqr->callback = dasd_eckd_analysis_callback; + init_cqr->callback_data = NULL; + init_cqr->expires = 5*HZ; + /* first try without ERP, so we can later handle unformatted + * devices as special case + */ + clear_bit(DASD_CQR_FLAGS_USE_ERP, &init_cqr->flags); + init_cqr->retries = 0; + dasd_add_request_head(init_cqr); + return -EAGAIN; +} + +static int dasd_eckd_end_analysis(struct dasd_block *block) +{ + struct dasd_device *device; + struct dasd_eckd_private *private; + struct eckd_count *count_area; + unsigned int sb, blk_per_trk; + int status, i; + struct dasd_ccw_req *init_cqr; + + device = block->base; + private = (struct dasd_eckd_private *) device->private; + status = private->init_cqr_status; + private->init_cqr_status = -1; + if (status == INIT_CQR_ERROR) { + /* try again, this time with full ERP */ + init_cqr = dasd_eckd_analysis_ccw(device); + dasd_sleep_on(init_cqr); + status = dasd_eckd_analysis_evaluation(init_cqr); + dasd_sfree_request(init_cqr, device); + } + + if (device->features & DASD_FEATURE_USERAW) { + block->bp_block = DASD_RAW_BLOCKSIZE; + blk_per_trk = DASD_RAW_BLOCK_PER_TRACK; + block->s2b_shift = 3; + goto raw; + } + + if (status == INIT_CQR_UNFORMATTED) { + dev_warn(&device->cdev->dev, "The DASD is not formatted\n"); + return -EMEDIUMTYPE; + } else if (status == INIT_CQR_ERROR) { + dev_err(&device->cdev->dev, + "Detecting the DASD disk layout failed because " + "of an I/O error\n"); + return -EIO; + } + + private->uses_cdl = 1; + /* Check Track 0 for Compatible Disk Layout */ + count_area = NULL; + for (i = 0; i < 3; i++) { + if (private->count_area[i].kl != 4 || + private->count_area[i].dl != dasd_eckd_cdl_reclen(i) - 4 || + private->count_area[i].cyl != 0 || + private->count_area[i].head != count_area_head[i] || + private->count_area[i].record != count_area_rec[i]) { + private->uses_cdl = 0; + break; + } + } + if (i == 3) + count_area = &private->count_area[4]; + + if (private->uses_cdl == 0) { + for (i = 0; i < 5; i++) { + if ((private->count_area[i].kl != 0) || + (private->count_area[i].dl != + private->count_area[0].dl) || + private->count_area[i].cyl != 0 || + private->count_area[i].head != count_area_head[i] || + private->count_area[i].record != count_area_rec[i]) + break; + } + if (i == 5) + count_area = &private->count_area[0]; + } else { + if (private->count_area[3].record == 1) + dev_warn(&device->cdev->dev, + "Track 0 has no records following the VTOC\n"); + } + + if (count_area != NULL && count_area->kl == 0) { + /* we found notthing violating our disk layout */ + if (dasd_check_blocksize(count_area->dl) == 0) + block->bp_block = count_area->dl; + } + if (block->bp_block == 0) { + dev_warn(&device->cdev->dev, + "The disk layout of the DASD is not supported\n"); + return -EMEDIUMTYPE; + } + block->s2b_shift = 0; /* bits to shift 512 to get a block */ + for (sb = 512; sb < block->bp_block; sb = sb << 1) + block->s2b_shift++; + + blk_per_trk = recs_per_track(&private->rdc_data, 0, block->bp_block); + +raw: + block->blocks = (private->real_cyl * + private->rdc_data.trk_per_cyl * + blk_per_trk); + + dev_info(&device->cdev->dev, + "DASD with %d KB/block, %d KB total size, %d KB/track, " + "%s\n", (block->bp_block >> 10), + ((private->real_cyl * + private->rdc_data.trk_per_cyl * + blk_per_trk * (block->bp_block >> 9)) >> 1), + ((blk_per_trk * block->bp_block) >> 10), + private->uses_cdl ? + "compatible disk layout" : "linux disk layout"); + + return 0; +} + +static int dasd_eckd_do_analysis(struct dasd_block *block) +{ + struct dasd_eckd_private *private; + + private = (struct dasd_eckd_private *) block->base->private; + if (private->init_cqr_status < 0) + return dasd_eckd_start_analysis(block); + else + return dasd_eckd_end_analysis(block); +} + +static int dasd_eckd_basic_to_ready(struct dasd_device *device) +{ + return dasd_alias_add_device(device); +}; + +static int dasd_eckd_online_to_ready(struct dasd_device *device) +{ + cancel_work_sync(&device->reload_device); + cancel_work_sync(&device->kick_validate); + return 0; +}; + +static int dasd_eckd_basic_to_known(struct dasd_device *device) +{ + return dasd_alias_remove_device(device); +}; + +static int +dasd_eckd_fill_geometry(struct dasd_block *block, struct hd_geometry *geo) +{ + struct dasd_eckd_private *private; + + private = (struct dasd_eckd_private *) block->base->private; + if (dasd_check_blocksize(block->bp_block) == 0) { + geo->sectors = recs_per_track(&private->rdc_data, + 0, block->bp_block); + } + geo->cylinders = private->rdc_data.no_cyl; + geo->heads = private->rdc_data.trk_per_cyl; + return 0; +} + +static struct dasd_ccw_req * +dasd_eckd_build_format(struct dasd_device *base, + struct format_data_t *fdata, + int enable_pav) +{ + struct dasd_eckd_private *base_priv; + struct dasd_eckd_private *start_priv; + struct dasd_device *startdev = NULL; + struct dasd_ccw_req *fcp; + struct eckd_count *ect; + struct ch_t address; + struct ccw1 *ccw; + void *data; + int rpt; + int cplength, datasize; + int i, j; + int intensity = 0; + int r0_perm; + int nr_tracks; + int use_prefix; + + if (enable_pav) + startdev = dasd_alias_get_start_dev(base); + + if (!startdev) + startdev = base; + + start_priv = (struct dasd_eckd_private *) startdev->private; + base_priv = (struct dasd_eckd_private *) base->private; + + rpt = recs_per_track(&base_priv->rdc_data, 0, fdata->blksize); + + nr_tracks = fdata->stop_unit - fdata->start_unit + 1; + + /* + * fdata->intensity is a bit string that tells us what to do: + * Bit 0: write record zero + * Bit 1: write home address, currently not supported + * Bit 2: invalidate tracks + * Bit 3: use OS/390 compatible disk layout (cdl) + * Bit 4: do not allow storage subsystem to modify record zero + * Only some bit combinations do make sense. + */ + if (fdata->intensity & 0x10) { + r0_perm = 0; + intensity = fdata->intensity & ~0x10; + } else { + r0_perm = 1; + intensity = fdata->intensity; + } + + use_prefix = base_priv->features.feature[8] & 0x01; + + switch (intensity) { + case 0x00: /* Normal format */ + case 0x08: /* Normal format, use cdl. */ + cplength = 2 + (rpt*nr_tracks); + if (use_prefix) + datasize = sizeof(struct PFX_eckd_data) + + sizeof(struct LO_eckd_data) + + rpt * nr_tracks * sizeof(struct eckd_count); + else + datasize = sizeof(struct DE_eckd_data) + + sizeof(struct LO_eckd_data) + + rpt * nr_tracks * sizeof(struct eckd_count); + break; + case 0x01: /* Write record zero and format track. */ + case 0x09: /* Write record zero and format track, use cdl. */ + cplength = 2 + rpt * nr_tracks; + if (use_prefix) + datasize = sizeof(struct PFX_eckd_data) + + sizeof(struct LO_eckd_data) + + sizeof(struct eckd_count) + + rpt * nr_tracks * sizeof(struct eckd_count); + else + datasize = sizeof(struct DE_eckd_data) + + sizeof(struct LO_eckd_data) + + sizeof(struct eckd_count) + + rpt * nr_tracks * sizeof(struct eckd_count); + break; + case 0x04: /* Invalidate track. */ + case 0x0c: /* Invalidate track, use cdl. */ + cplength = 3; + if (use_prefix) + datasize = sizeof(struct PFX_eckd_data) + + sizeof(struct LO_eckd_data) + + sizeof(struct eckd_count); + else + datasize = sizeof(struct DE_eckd_data) + + sizeof(struct LO_eckd_data) + + sizeof(struct eckd_count); + break; + default: + dev_warn(&startdev->cdev->dev, + "An I/O control call used incorrect flags 0x%x\n", + fdata->intensity); + return ERR_PTR(-EINVAL); + } + /* Allocate the format ccw request. */ + fcp = dasd_smalloc_request(DASD_ECKD_MAGIC, cplength, + datasize, startdev); + if (IS_ERR(fcp)) + return fcp; + + start_priv->count++; + data = fcp->data; + ccw = fcp->cpaddr; + + switch (intensity & ~0x08) { + case 0x00: /* Normal format. */ + if (use_prefix) { + prefix(ccw++, (struct PFX_eckd_data *) data, + fdata->start_unit, fdata->stop_unit, + DASD_ECKD_CCW_WRITE_CKD, base, startdev); + /* grant subsystem permission to format R0 */ + if (r0_perm) + ((struct PFX_eckd_data *)data) + ->define_extent.ga_extended |= 0x04; + data += sizeof(struct PFX_eckd_data); + } else { + define_extent(ccw++, (struct DE_eckd_data *) data, + fdata->start_unit, fdata->stop_unit, + DASD_ECKD_CCW_WRITE_CKD, startdev); + /* grant subsystem permission to format R0 */ + if (r0_perm) + ((struct DE_eckd_data *) data) + ->ga_extended |= 0x04; + data += sizeof(struct DE_eckd_data); + } + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, (struct LO_eckd_data *) data, + fdata->start_unit, 0, rpt*nr_tracks, + DASD_ECKD_CCW_WRITE_CKD, base, + fdata->blksize); + data += sizeof(struct LO_eckd_data); + break; + case 0x01: /* Write record zero + format track. */ + if (use_prefix) { + prefix(ccw++, (struct PFX_eckd_data *) data, + fdata->start_unit, fdata->stop_unit, + DASD_ECKD_CCW_WRITE_RECORD_ZERO, + base, startdev); + data += sizeof(struct PFX_eckd_data); + } else { + define_extent(ccw++, (struct DE_eckd_data *) data, + fdata->start_unit, fdata->stop_unit, + DASD_ECKD_CCW_WRITE_RECORD_ZERO, startdev); + data += sizeof(struct DE_eckd_data); + } + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, (struct LO_eckd_data *) data, + fdata->start_unit, 0, rpt * nr_tracks + 1, + DASD_ECKD_CCW_WRITE_RECORD_ZERO, base, + base->block->bp_block); + data += sizeof(struct LO_eckd_data); + break; + case 0x04: /* Invalidate track. */ + if (use_prefix) { + prefix(ccw++, (struct PFX_eckd_data *) data, + fdata->start_unit, fdata->stop_unit, + DASD_ECKD_CCW_WRITE_CKD, base, startdev); + data += sizeof(struct PFX_eckd_data); + } else { + define_extent(ccw++, (struct DE_eckd_data *) data, + fdata->start_unit, fdata->stop_unit, + DASD_ECKD_CCW_WRITE_CKD, startdev); + data += sizeof(struct DE_eckd_data); + } + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, (struct LO_eckd_data *) data, + fdata->start_unit, 0, 1, + DASD_ECKD_CCW_WRITE_CKD, base, 8); + data += sizeof(struct LO_eckd_data); + break; + } + + for (j = 0; j < nr_tracks; j++) { + /* calculate cylinder and head for the current track */ + set_ch_t(&address, + (fdata->start_unit + j) / + base_priv->rdc_data.trk_per_cyl, + (fdata->start_unit + j) % + base_priv->rdc_data.trk_per_cyl); + if (intensity & 0x01) { /* write record zero */ + ect = (struct eckd_count *) data; + data += sizeof(struct eckd_count); + ect->cyl = address.cyl; + ect->head = address.head; + ect->record = 0; + ect->kl = 0; + ect->dl = 8; + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = DASD_ECKD_CCW_WRITE_RECORD_ZERO; + ccw->flags = CCW_FLAG_SLI; + ccw->count = 8; + ccw->cda = (__u32)(addr_t) ect; + ccw++; + } + if ((intensity & ~0x08) & 0x04) { /* erase track */ + ect = (struct eckd_count *) data; + data += sizeof(struct eckd_count); + ect->cyl = address.cyl; + ect->head = address.head; + ect->record = 1; + ect->kl = 0; + ect->dl = 0; + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = DASD_ECKD_CCW_WRITE_CKD; + ccw->flags = CCW_FLAG_SLI; + ccw->count = 8; + ccw->cda = (__u32)(addr_t) ect; + } else { /* write remaining records */ + for (i = 0; i < rpt; i++) { + ect = (struct eckd_count *) data; + data += sizeof(struct eckd_count); + ect->cyl = address.cyl; + ect->head = address.head; + ect->record = i + 1; + ect->kl = 0; + ect->dl = fdata->blksize; + /* + * Check for special tracks 0-1 + * when formatting CDL + */ + if ((intensity & 0x08) && + fdata->start_unit == 0) { + if (i < 3) { + ect->kl = 4; + ect->dl = sizes_trk0[i] - 4; + } + } + if ((intensity & 0x08) && + fdata->start_unit == 1) { + ect->kl = 44; + ect->dl = LABEL_SIZE - 44; + } + ccw[-1].flags |= CCW_FLAG_CC; + if (i != 0 || j == 0) + ccw->cmd_code = + DASD_ECKD_CCW_WRITE_CKD; + else + ccw->cmd_code = + DASD_ECKD_CCW_WRITE_CKD_MT; + ccw->flags = CCW_FLAG_SLI; + ccw->count = 8; + ccw->cda = (__u32)(addr_t) ect; + ccw++; + } + } + } + + fcp->startdev = startdev; + fcp->memdev = startdev; + fcp->basedev = base; + fcp->retries = 256; + fcp->expires = startdev->default_expires * HZ; + fcp->buildclk = get_tod_clock(); + fcp->status = DASD_CQR_FILLED; + + return fcp; +} + +static int +dasd_eckd_format_device(struct dasd_device *base, + struct format_data_t *fdata, + int enable_pav) +{ + struct dasd_ccw_req *cqr, *n; + struct dasd_block *block; + struct dasd_eckd_private *private; + struct list_head format_queue; + struct dasd_device *device; + int old_stop, format_step; + int step, rc = 0, sleep_rc; + + block = base->block; + private = (struct dasd_eckd_private *) base->private; + + /* Sanity checks. */ + if (fdata->start_unit >= + (private->real_cyl * private->rdc_data.trk_per_cyl)) { + dev_warn(&base->cdev->dev, + "Start track number %u used in formatting is too big\n", + fdata->start_unit); + return -EINVAL; + } + if (fdata->stop_unit >= + (private->real_cyl * private->rdc_data.trk_per_cyl)) { + dev_warn(&base->cdev->dev, + "Stop track number %u used in formatting is too big\n", + fdata->stop_unit); + return -EINVAL; + } + if (fdata->start_unit > fdata->stop_unit) { + dev_warn(&base->cdev->dev, + "Start track %u used in formatting exceeds end track\n", + fdata->start_unit); + return -EINVAL; + } + if (dasd_check_blocksize(fdata->blksize) != 0) { + dev_warn(&base->cdev->dev, + "The DASD cannot be formatted with block size %u\n", + fdata->blksize); + return -EINVAL; + } + + INIT_LIST_HEAD(&format_queue); + + old_stop = fdata->stop_unit; + while (fdata->start_unit <= 1) { + fdata->stop_unit = fdata->start_unit; + cqr = dasd_eckd_build_format(base, fdata, enable_pav); + list_add(&cqr->blocklist, &format_queue); + + fdata->stop_unit = old_stop; + fdata->start_unit++; + + if (fdata->start_unit > fdata->stop_unit) + goto sleep; + } + +retry: + format_step = 255 / recs_per_track(&private->rdc_data, 0, + fdata->blksize); + while (fdata->start_unit <= old_stop) { + step = fdata->stop_unit - fdata->start_unit + 1; + if (step > format_step) + fdata->stop_unit = fdata->start_unit + format_step - 1; + + cqr = dasd_eckd_build_format(base, fdata, enable_pav); + if (IS_ERR(cqr)) { + if (PTR_ERR(cqr) == -ENOMEM) { + /* + * not enough memory available + * go to out and start requests + * retry after first requests were finished + */ + fdata->stop_unit = old_stop; + goto sleep; + } else + return PTR_ERR(cqr); + } + list_add(&cqr->blocklist, &format_queue); + + fdata->start_unit = fdata->stop_unit + 1; + fdata->stop_unit = old_stop; + } + +sleep: + sleep_rc = dasd_sleep_on_queue(&format_queue); + + list_for_each_entry_safe(cqr, n, &format_queue, blocklist) { + device = cqr->startdev; + private = (struct dasd_eckd_private *) device->private; + if (cqr->status == DASD_CQR_FAILED) + rc = -EIO; + list_del_init(&cqr->blocklist); + dasd_sfree_request(cqr, device); + private->count--; + } + + if (sleep_rc) + return sleep_rc; + + /* + * in case of ENOMEM we need to retry after + * first requests are finished + */ + if (fdata->start_unit <= fdata->stop_unit) + goto retry; + + return rc; +} + +static void dasd_eckd_handle_terminated_request(struct dasd_ccw_req *cqr) +{ + if (cqr->retries < 0) { + cqr->status = DASD_CQR_FAILED; + return; + } + cqr->status = DASD_CQR_FILLED; + if (cqr->block && (cqr->startdev != cqr->block->base)) { + dasd_eckd_reset_ccw_to_base_io(cqr); + cqr->startdev = cqr->block->base; + cqr->lpm = cqr->block->base->path_data.opm; + } +}; + +static dasd_erp_fn_t +dasd_eckd_erp_action(struct dasd_ccw_req * cqr) +{ + struct dasd_device *device = (struct dasd_device *) cqr->startdev; + struct ccw_device *cdev = device->cdev; + + switch (cdev->id.cu_type) { + case 0x3990: + case 0x2105: + case 0x2107: + case 0x1750: + return dasd_3990_erp_action; + case 0x9343: + case 0x3880: + default: + return dasd_default_erp_action; + } +} + +static dasd_erp_fn_t +dasd_eckd_erp_postaction(struct dasd_ccw_req * cqr) +{ + return dasd_default_erp_postaction; +} + +static void dasd_eckd_check_for_device_change(struct dasd_device *device, + struct dasd_ccw_req *cqr, + struct irb *irb) +{ + char mask; + char *sense = NULL; + struct dasd_eckd_private *private; + + private = (struct dasd_eckd_private *) device->private; + /* first of all check for state change pending interrupt */ + mask = DEV_STAT_ATTENTION | DEV_STAT_DEV_END | DEV_STAT_UNIT_EXCEP; + if ((scsw_dstat(&irb->scsw) & mask) == mask) { + /* + * for alias only, not in offline processing + * and only if not suspended + */ + if (!device->block && private->lcu && + device->state == DASD_STATE_ONLINE && + !test_bit(DASD_FLAG_OFFLINE, &device->flags) && + !test_bit(DASD_FLAG_SUSPENDED, &device->flags)) { + /* + * the state change could be caused by an alias + * reassignment remove device from alias handling + * to prevent new requests from being scheduled on + * the wrong alias device + */ + dasd_alias_remove_device(device); + + /* schedule worker to reload device */ + dasd_reload_device(device); + } + dasd_generic_handle_state_change(device); + return; + } + + sense = dasd_get_sense(irb); + if (!sense) + return; + + /* summary unit check */ + if ((sense[27] & DASD_SENSE_BIT_0) && (sense[7] == 0x0D) && + (scsw_dstat(&irb->scsw) & DEV_STAT_UNIT_CHECK)) { + dasd_alias_handle_summary_unit_check(device, irb); + return; + } + + /* service information message SIM */ + if (!cqr && !(sense[27] & DASD_SENSE_BIT_0) && + ((sense[6] & DASD_SIM_SENSE) == DASD_SIM_SENSE)) { + dasd_3990_erp_handle_sim(device, sense); + return; + } + + /* loss of device reservation is handled via base devices only + * as alias devices may be used with several bases + */ + if (device->block && (sense[27] & DASD_SENSE_BIT_0) && + (sense[7] == 0x3F) && + (scsw_dstat(&irb->scsw) & DEV_STAT_UNIT_CHECK) && + test_bit(DASD_FLAG_IS_RESERVED, &device->flags)) { + if (device->features & DASD_FEATURE_FAILONSLCK) + set_bit(DASD_FLAG_LOCK_STOLEN, &device->flags); + clear_bit(DASD_FLAG_IS_RESERVED, &device->flags); + dev_err(&device->cdev->dev, + "The device reservation was lost\n"); + } +} + +static struct dasd_ccw_req *dasd_eckd_build_cp_cmd_single( + struct dasd_device *startdev, + struct dasd_block *block, + struct request *req, + sector_t first_rec, + sector_t last_rec, + sector_t first_trk, + sector_t last_trk, + unsigned int first_offs, + unsigned int last_offs, + unsigned int blk_per_trk, + unsigned int blksize) +{ + struct dasd_eckd_private *private; + unsigned long *idaws; + struct LO_eckd_data *LO_data; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + struct req_iterator iter; + struct bio_vec bv; + char *dst; + unsigned int off; + int count, cidaw, cplength, datasize; + sector_t recid; + unsigned char cmd, rcmd; + int use_prefix; + struct dasd_device *basedev; + + basedev = block->base; + private = (struct dasd_eckd_private *) basedev->private; + if (rq_data_dir(req) == READ) + cmd = DASD_ECKD_CCW_READ_MT; + else if (rq_data_dir(req) == WRITE) + cmd = DASD_ECKD_CCW_WRITE_MT; + else + return ERR_PTR(-EINVAL); + + /* Check struct bio and count the number of blocks for the request. */ + count = 0; + cidaw = 0; + rq_for_each_segment(bv, req, iter) { + if (bv.bv_len & (blksize - 1)) + /* Eckd can only do full blocks. */ + return ERR_PTR(-EINVAL); + count += bv.bv_len >> (block->s2b_shift + 9); + if (idal_is_needed (page_address(bv.bv_page), bv.bv_len)) + cidaw += bv.bv_len >> (block->s2b_shift + 9); + } + /* Paranoia. */ + if (count != last_rec - first_rec + 1) + return ERR_PTR(-EINVAL); + + /* use the prefix command if available */ + use_prefix = private->features.feature[8] & 0x01; + if (use_prefix) { + /* 1x prefix + number of blocks */ + cplength = 2 + count; + /* 1x prefix + cidaws*sizeof(long) */ + datasize = sizeof(struct PFX_eckd_data) + + sizeof(struct LO_eckd_data) + + cidaw * sizeof(unsigned long); + } else { + /* 1x define extent + 1x locate record + number of blocks */ + cplength = 2 + count; + /* 1x define extent + 1x locate record + cidaws*sizeof(long) */ + datasize = sizeof(struct DE_eckd_data) + + sizeof(struct LO_eckd_data) + + cidaw * sizeof(unsigned long); + } + /* Find out the number of additional locate record ccws for cdl. */ + if (private->uses_cdl && first_rec < 2*blk_per_trk) { + if (last_rec >= 2*blk_per_trk) + count = 2*blk_per_trk - first_rec; + cplength += count; + datasize += count*sizeof(struct LO_eckd_data); + } + /* Allocate the ccw request. */ + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, cplength, datasize, + startdev); + if (IS_ERR(cqr)) + return cqr; + ccw = cqr->cpaddr; + /* First ccw is define extent or prefix. */ + if (use_prefix) { + if (prefix(ccw++, cqr->data, first_trk, + last_trk, cmd, basedev, startdev) == -EAGAIN) { + /* Clock not in sync and XRC is enabled. + * Try again later. + */ + dasd_sfree_request(cqr, startdev); + return ERR_PTR(-EAGAIN); + } + idaws = (unsigned long *) (cqr->data + + sizeof(struct PFX_eckd_data)); + } else { + if (define_extent(ccw++, cqr->data, first_trk, + last_trk, cmd, basedev) == -EAGAIN) { + /* Clock not in sync and XRC is enabled. + * Try again later. + */ + dasd_sfree_request(cqr, startdev); + return ERR_PTR(-EAGAIN); + } + idaws = (unsigned long *) (cqr->data + + sizeof(struct DE_eckd_data)); + } + /* Build locate_record+read/write/ccws. */ + LO_data = (struct LO_eckd_data *) (idaws + cidaw); + recid = first_rec; + if (private->uses_cdl == 0 || recid > 2*blk_per_trk) { + /* Only standard blocks so there is just one locate record. */ + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, LO_data++, first_trk, first_offs + 1, + last_rec - recid + 1, cmd, basedev, blksize); + } + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + if (dasd_page_cache) { + char *copy = kmem_cache_alloc(dasd_page_cache, + GFP_DMA | __GFP_NOWARN); + if (copy && rq_data_dir(req) == WRITE) + memcpy(copy + bv.bv_offset, dst, bv.bv_len); + if (copy) + dst = copy + bv.bv_offset; + } + for (off = 0; off < bv.bv_len; off += blksize) { + sector_t trkid = recid; + unsigned int recoffs = sector_div(trkid, blk_per_trk); + rcmd = cmd; + count = blksize; + /* Locate record for cdl special block ? */ + if (private->uses_cdl && recid < 2*blk_per_trk) { + if (dasd_eckd_cdl_special(blk_per_trk, recid)){ + rcmd |= 0x8; + count = dasd_eckd_cdl_reclen(recid); + if (count < blksize && + rq_data_dir(req) == READ) + memset(dst + count, 0xe5, + blksize - count); + } + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, LO_data++, + trkid, recoffs + 1, + 1, rcmd, basedev, count); + } + /* Locate record for standard blocks ? */ + if (private->uses_cdl && recid == 2*blk_per_trk) { + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, LO_data++, + trkid, recoffs + 1, + last_rec - recid + 1, + cmd, basedev, count); + } + /* Read/write ccw. */ + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = rcmd; + ccw->count = count; + if (idal_is_needed(dst, blksize)) { + ccw->cda = (__u32)(addr_t) idaws; + ccw->flags = CCW_FLAG_IDA; + idaws = idal_create_words(idaws, dst, blksize); + } else { + ccw->cda = (__u32)(addr_t) dst; + ccw->flags = 0; + } + ccw++; + dst += blksize; + recid++; + } + } + if (blk_noretry_request(req) || + block->base->features & DASD_FEATURE_FAILFAST) + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->startdev = startdev; + cqr->memdev = startdev; + cqr->block = block; + cqr->expires = startdev->default_expires * HZ; /* default 5 minutes */ + cqr->lpm = startdev->path_data.ppm; + cqr->retries = startdev->default_retries; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +} + +static struct dasd_ccw_req *dasd_eckd_build_cp_cmd_track( + struct dasd_device *startdev, + struct dasd_block *block, + struct request *req, + sector_t first_rec, + sector_t last_rec, + sector_t first_trk, + sector_t last_trk, + unsigned int first_offs, + unsigned int last_offs, + unsigned int blk_per_trk, + unsigned int blksize) +{ + unsigned long *idaws; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + struct req_iterator iter; + struct bio_vec bv; + char *dst, *idaw_dst; + unsigned int cidaw, cplength, datasize; + unsigned int tlf; + sector_t recid; + unsigned char cmd; + struct dasd_device *basedev; + unsigned int trkcount, count, count_to_trk_end; + unsigned int idaw_len, seg_len, part_len, len_to_track_end; + unsigned char new_track, end_idaw; + sector_t trkid; + unsigned int recoffs; + + basedev = block->base; + if (rq_data_dir(req) == READ) + cmd = DASD_ECKD_CCW_READ_TRACK_DATA; + else if (rq_data_dir(req) == WRITE) + cmd = DASD_ECKD_CCW_WRITE_TRACK_DATA; + else + return ERR_PTR(-EINVAL); + + /* Track based I/O needs IDAWs for each page, and not just for + * 64 bit addresses. We need additional idals for pages + * that get filled from two tracks, so we use the number + * of records as upper limit. + */ + cidaw = last_rec - first_rec + 1; + trkcount = last_trk - first_trk + 1; + + /* 1x prefix + one read/write ccw per track */ + cplength = 1 + trkcount; + + /* on 31-bit we need space for two 32 bit addresses per page + * on 64-bit one 64 bit address + */ + datasize = sizeof(struct PFX_eckd_data) + + cidaw * sizeof(unsigned long long); + + /* Allocate the ccw request. */ + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, cplength, datasize, + startdev); + if (IS_ERR(cqr)) + return cqr; + ccw = cqr->cpaddr; + /* transfer length factor: how many bytes to read from the last track */ + if (first_trk == last_trk) + tlf = last_offs - first_offs + 1; + else + tlf = last_offs + 1; + tlf *= blksize; + + if (prefix_LRE(ccw++, cqr->data, first_trk, + last_trk, cmd, basedev, startdev, + 1 /* format */, first_offs + 1, + trkcount, blksize, + tlf) == -EAGAIN) { + /* Clock not in sync and XRC is enabled. + * Try again later. + */ + dasd_sfree_request(cqr, startdev); + return ERR_PTR(-EAGAIN); + } + + /* + * The translation of request into ccw programs must meet the + * following conditions: + * - all idaws but the first and the last must address full pages + * (or 2K blocks on 31-bit) + * - the scope of a ccw and it's idal ends with the track boundaries + */ + idaws = (unsigned long *) (cqr->data + sizeof(struct PFX_eckd_data)); + recid = first_rec; + new_track = 1; + end_idaw = 0; + len_to_track_end = 0; + idaw_dst = NULL; + idaw_len = 0; + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + seg_len = bv.bv_len; + while (seg_len) { + if (new_track) { + trkid = recid; + recoffs = sector_div(trkid, blk_per_trk); + count_to_trk_end = blk_per_trk - recoffs; + count = min((last_rec - recid + 1), + (sector_t)count_to_trk_end); + len_to_track_end = count * blksize; + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = cmd; + ccw->count = len_to_track_end; + ccw->cda = (__u32)(addr_t)idaws; + ccw->flags = CCW_FLAG_IDA; + ccw++; + recid += count; + new_track = 0; + /* first idaw for a ccw may start anywhere */ + if (!idaw_dst) + idaw_dst = dst; + } + /* If we start a new idaw, we must make sure that it + * starts on an IDA_BLOCK_SIZE boundary. + * If we continue an idaw, we must make sure that the + * current segment begins where the so far accumulated + * idaw ends + */ + if (!idaw_dst) { + if (__pa(dst) & (IDA_BLOCK_SIZE-1)) { + dasd_sfree_request(cqr, startdev); + return ERR_PTR(-ERANGE); + } else + idaw_dst = dst; + } + if ((idaw_dst + idaw_len) != dst) { + dasd_sfree_request(cqr, startdev); + return ERR_PTR(-ERANGE); + } + part_len = min(seg_len, len_to_track_end); + seg_len -= part_len; + dst += part_len; + idaw_len += part_len; + len_to_track_end -= part_len; + /* collected memory area ends on an IDA_BLOCK border, + * -> create an idaw + * idal_create_words will handle cases where idaw_len + * is larger then IDA_BLOCK_SIZE + */ + if (!(__pa(idaw_dst + idaw_len) & (IDA_BLOCK_SIZE-1))) + end_idaw = 1; + /* We also need to end the idaw at track end */ + if (!len_to_track_end) { + new_track = 1; + end_idaw = 1; + } + if (end_idaw) { + idaws = idal_create_words(idaws, idaw_dst, + idaw_len); + idaw_dst = NULL; + idaw_len = 0; + end_idaw = 0; + } + } + } + + if (blk_noretry_request(req) || + block->base->features & DASD_FEATURE_FAILFAST) + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->startdev = startdev; + cqr->memdev = startdev; + cqr->block = block; + cqr->expires = startdev->default_expires * HZ; /* default 5 minutes */ + cqr->lpm = startdev->path_data.ppm; + cqr->retries = startdev->default_retries; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +} + +static int prepare_itcw(struct itcw *itcw, + unsigned int trk, unsigned int totrk, int cmd, + struct dasd_device *basedev, + struct dasd_device *startdev, + unsigned int rec_on_trk, int count, + unsigned int blksize, + unsigned int total_data_size, + unsigned int tlf, + unsigned int blk_per_trk) +{ + struct PFX_eckd_data pfxdata; + struct dasd_eckd_private *basepriv, *startpriv; + struct DE_eckd_data *dedata; + struct LRE_eckd_data *lredata; + struct dcw *dcw; + + u32 begcyl, endcyl; + u16 heads, beghead, endhead; + u8 pfx_cmd; + + int rc = 0; + int sector = 0; + int dn, d; + + + /* setup prefix data */ + basepriv = (struct dasd_eckd_private *) basedev->private; + startpriv = (struct dasd_eckd_private *) startdev->private; + dedata = &pfxdata.define_extent; + lredata = &pfxdata.locate_record; + + memset(&pfxdata, 0, sizeof(pfxdata)); + pfxdata.format = 1; /* PFX with LRE */ + pfxdata.base_address = basepriv->ned->unit_addr; + pfxdata.base_lss = basepriv->ned->ID; + pfxdata.validity.define_extent = 1; + + /* private uid is kept up to date, conf_data may be outdated */ + if (startpriv->uid.type != UA_BASE_DEVICE) { + pfxdata.validity.verify_base = 1; + if (startpriv->uid.type == UA_HYPER_PAV_ALIAS) + pfxdata.validity.hyper_pav = 1; + } + + switch (cmd) { + case DASD_ECKD_CCW_READ_TRACK_DATA: + dedata->mask.perm = 0x1; + dedata->attributes.operation = basepriv->attrib.operation; + dedata->blk_size = blksize; + dedata->ga_extended |= 0x42; + lredata->operation.orientation = 0x0; + lredata->operation.operation = 0x0C; + lredata->auxiliary.check_bytes = 0x01; + pfx_cmd = DASD_ECKD_CCW_PFX_READ; + break; + case DASD_ECKD_CCW_WRITE_TRACK_DATA: + dedata->mask.perm = 0x02; + dedata->attributes.operation = basepriv->attrib.operation; + dedata->blk_size = blksize; + rc = check_XRC_on_prefix(&pfxdata, basedev); + dedata->ga_extended |= 0x42; + lredata->operation.orientation = 0x0; + lredata->operation.operation = 0x3F; + lredata->extended_operation = 0x23; + lredata->auxiliary.check_bytes = 0x2; + pfx_cmd = DASD_ECKD_CCW_PFX; + break; + default: + DBF_DEV_EVENT(DBF_ERR, basedev, + "prepare itcw, unknown opcode 0x%x", cmd); + BUG(); + break; + } + if (rc) + return rc; + + dedata->attributes.mode = 0x3; /* ECKD */ + + heads = basepriv->rdc_data.trk_per_cyl; + begcyl = trk / heads; + beghead = trk % heads; + endcyl = totrk / heads; + endhead = totrk % heads; + + /* check for sequential prestage - enhance cylinder range */ + if (dedata->attributes.operation == DASD_SEQ_PRESTAGE || + dedata->attributes.operation == DASD_SEQ_ACCESS) { + + if (endcyl + basepriv->attrib.nr_cyl < basepriv->real_cyl) + endcyl += basepriv->attrib.nr_cyl; + else + endcyl = (basepriv->real_cyl - 1); + } + + set_ch_t(&dedata->beg_ext, begcyl, beghead); + set_ch_t(&dedata->end_ext, endcyl, endhead); + + dedata->ep_format = 0x20; /* records per track is valid */ + dedata->ep_rec_per_track = blk_per_trk; + + if (rec_on_trk) { + switch (basepriv->rdc_data.dev_type) { + case 0x3390: + dn = ceil_quot(blksize + 6, 232); + d = 9 + ceil_quot(blksize + 6 * (dn + 1), 34); + sector = (49 + (rec_on_trk - 1) * (10 + d)) / 8; + break; + case 0x3380: + d = 7 + ceil_quot(blksize + 12, 32); + sector = (39 + (rec_on_trk - 1) * (8 + d)) / 7; + break; + } + } + + lredata->auxiliary.length_valid = 1; + lredata->auxiliary.length_scope = 1; + lredata->auxiliary.imbedded_ccw_valid = 1; + lredata->length = tlf; + lredata->imbedded_ccw = cmd; + lredata->count = count; + lredata->sector = sector; + set_ch_t(&lredata->seek_addr, begcyl, beghead); + lredata->search_arg.cyl = lredata->seek_addr.cyl; + lredata->search_arg.head = lredata->seek_addr.head; + lredata->search_arg.record = rec_on_trk; + + dcw = itcw_add_dcw(itcw, pfx_cmd, 0, + &pfxdata, sizeof(pfxdata), total_data_size); + return PTR_RET(dcw); +} + +static struct dasd_ccw_req *dasd_eckd_build_cp_tpm_track( + struct dasd_device *startdev, + struct dasd_block *block, + struct request *req, + sector_t first_rec, + sector_t last_rec, + sector_t first_trk, + sector_t last_trk, + unsigned int first_offs, + unsigned int last_offs, + unsigned int blk_per_trk, + unsigned int blksize) +{ + struct dasd_ccw_req *cqr; + struct req_iterator iter; + struct bio_vec bv; + char *dst; + unsigned int trkcount, ctidaw; + unsigned char cmd; + struct dasd_device *basedev; + unsigned int tlf; + struct itcw *itcw; + struct tidaw *last_tidaw = NULL; + int itcw_op; + size_t itcw_size; + u8 tidaw_flags; + unsigned int seg_len, part_len, len_to_track_end; + unsigned char new_track; + sector_t recid, trkid; + unsigned int offs; + unsigned int count, count_to_trk_end; + int ret; + + basedev = block->base; + if (rq_data_dir(req) == READ) { + cmd = DASD_ECKD_CCW_READ_TRACK_DATA; + itcw_op = ITCW_OP_READ; + } else if (rq_data_dir(req) == WRITE) { + cmd = DASD_ECKD_CCW_WRITE_TRACK_DATA; + itcw_op = ITCW_OP_WRITE; + } else + return ERR_PTR(-EINVAL); + + /* trackbased I/O needs address all memory via TIDAWs, + * not just for 64 bit addresses. This allows us to map + * each segment directly to one tidaw. + * In the case of write requests, additional tidaws may + * be needed when a segment crosses a track boundary. + */ + trkcount = last_trk - first_trk + 1; + ctidaw = 0; + rq_for_each_segment(bv, req, iter) { + ++ctidaw; + } + if (rq_data_dir(req) == WRITE) + ctidaw += (last_trk - first_trk); + + /* Allocate the ccw request. */ + itcw_size = itcw_calc_size(0, ctidaw, 0); + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 0, itcw_size, startdev); + if (IS_ERR(cqr)) + return cqr; + + /* transfer length factor: how many bytes to read from the last track */ + if (first_trk == last_trk) + tlf = last_offs - first_offs + 1; + else + tlf = last_offs + 1; + tlf *= blksize; + + itcw = itcw_init(cqr->data, itcw_size, itcw_op, 0, ctidaw, 0); + if (IS_ERR(itcw)) { + ret = -EINVAL; + goto out_error; + } + cqr->cpaddr = itcw_get_tcw(itcw); + if (prepare_itcw(itcw, first_trk, last_trk, + cmd, basedev, startdev, + first_offs + 1, + trkcount, blksize, + (last_rec - first_rec + 1) * blksize, + tlf, blk_per_trk) == -EAGAIN) { + /* Clock not in sync and XRC is enabled. + * Try again later. + */ + ret = -EAGAIN; + goto out_error; + } + len_to_track_end = 0; + /* + * A tidaw can address 4k of memory, but must not cross page boundaries + * We can let the block layer handle this by setting + * blk_queue_segment_boundary to page boundaries and + * blk_max_segment_size to page size when setting up the request queue. + * For write requests, a TIDAW must not cross track boundaries, because + * we have to set the CBC flag on the last tidaw for each track. + */ + if (rq_data_dir(req) == WRITE) { + new_track = 1; + recid = first_rec; + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + seg_len = bv.bv_len; + while (seg_len) { + if (new_track) { + trkid = recid; + offs = sector_div(trkid, blk_per_trk); + count_to_trk_end = blk_per_trk - offs; + count = min((last_rec - recid + 1), + (sector_t)count_to_trk_end); + len_to_track_end = count * blksize; + recid += count; + new_track = 0; + } + part_len = min(seg_len, len_to_track_end); + seg_len -= part_len; + len_to_track_end -= part_len; + /* We need to end the tidaw at track end */ + if (!len_to_track_end) { + new_track = 1; + tidaw_flags = TIDAW_FLAGS_INSERT_CBC; + } else + tidaw_flags = 0; + last_tidaw = itcw_add_tidaw(itcw, tidaw_flags, + dst, part_len); + if (IS_ERR(last_tidaw)) { + ret = -EINVAL; + goto out_error; + } + dst += part_len; + } + } + } else { + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + last_tidaw = itcw_add_tidaw(itcw, 0x00, + dst, bv.bv_len); + if (IS_ERR(last_tidaw)) { + ret = -EINVAL; + goto out_error; + } + } + } + last_tidaw->flags |= TIDAW_FLAGS_LAST; + last_tidaw->flags &= ~TIDAW_FLAGS_INSERT_CBC; + itcw_finalize(itcw); + + if (blk_noretry_request(req) || + block->base->features & DASD_FEATURE_FAILFAST) + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->cpmode = 1; + cqr->startdev = startdev; + cqr->memdev = startdev; + cqr->block = block; + cqr->expires = startdev->default_expires * HZ; /* default 5 minutes */ + cqr->lpm = startdev->path_data.ppm; + cqr->retries = startdev->default_retries; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +out_error: + dasd_sfree_request(cqr, startdev); + return ERR_PTR(ret); +} + +static struct dasd_ccw_req *dasd_eckd_build_cp(struct dasd_device *startdev, + struct dasd_block *block, + struct request *req) +{ + int cmdrtd, cmdwtd; + int use_prefix; + int fcx_multitrack; + struct dasd_eckd_private *private; + struct dasd_device *basedev; + sector_t first_rec, last_rec; + sector_t first_trk, last_trk; + unsigned int first_offs, last_offs; + unsigned int blk_per_trk, blksize; + int cdlspecial; + unsigned int data_size; + struct dasd_ccw_req *cqr; + + basedev = block->base; + private = (struct dasd_eckd_private *) basedev->private; + + /* Calculate number of blocks/records per track. */ + blksize = block->bp_block; + blk_per_trk = recs_per_track(&private->rdc_data, 0, blksize); + if (blk_per_trk == 0) + return ERR_PTR(-EINVAL); + /* Calculate record id of first and last block. */ + first_rec = first_trk = blk_rq_pos(req) >> block->s2b_shift; + first_offs = sector_div(first_trk, blk_per_trk); + last_rec = last_trk = + (blk_rq_pos(req) + blk_rq_sectors(req) - 1) >> block->s2b_shift; + last_offs = sector_div(last_trk, blk_per_trk); + cdlspecial = (private->uses_cdl && first_rec < 2*blk_per_trk); + + fcx_multitrack = private->features.feature[40] & 0x20; + data_size = blk_rq_bytes(req); + if (data_size % blksize) + return ERR_PTR(-EINVAL); + /* tpm write request add CBC data on each track boundary */ + if (rq_data_dir(req) == WRITE) + data_size += (last_trk - first_trk) * 4; + + /* is read track data and write track data in command mode supported? */ + cmdrtd = private->features.feature[9] & 0x20; + cmdwtd = private->features.feature[12] & 0x40; + use_prefix = private->features.feature[8] & 0x01; + + cqr = NULL; + if (cdlspecial || dasd_page_cache) { + /* do nothing, just fall through to the cmd mode single case */ + } else if ((data_size <= private->fcx_max_data) + && (fcx_multitrack || (first_trk == last_trk))) { + cqr = dasd_eckd_build_cp_tpm_track(startdev, block, req, + first_rec, last_rec, + first_trk, last_trk, + first_offs, last_offs, + blk_per_trk, blksize); + if (IS_ERR(cqr) && (PTR_ERR(cqr) != -EAGAIN) && + (PTR_ERR(cqr) != -ENOMEM)) + cqr = NULL; + } else if (use_prefix && + (((rq_data_dir(req) == READ) && cmdrtd) || + ((rq_data_dir(req) == WRITE) && cmdwtd))) { + cqr = dasd_eckd_build_cp_cmd_track(startdev, block, req, + first_rec, last_rec, + first_trk, last_trk, + first_offs, last_offs, + blk_per_trk, blksize); + if (IS_ERR(cqr) && (PTR_ERR(cqr) != -EAGAIN) && + (PTR_ERR(cqr) != -ENOMEM)) + cqr = NULL; + } + if (!cqr) + cqr = dasd_eckd_build_cp_cmd_single(startdev, block, req, + first_rec, last_rec, + first_trk, last_trk, + first_offs, last_offs, + blk_per_trk, blksize); + return cqr; +} + +static struct dasd_ccw_req *dasd_raw_build_cp(struct dasd_device *startdev, + struct dasd_block *block, + struct request *req) +{ + unsigned long *idaws; + struct dasd_device *basedev; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + struct req_iterator iter; + struct bio_vec bv; + char *dst; + unsigned char cmd; + unsigned int trkcount; + unsigned int seg_len, len_to_track_end; + unsigned int first_offs; + unsigned int cidaw, cplength, datasize; + sector_t first_trk, last_trk, sectors; + sector_t start_padding_sectors, end_sector_offset, end_padding_sectors; + unsigned int pfx_datasize; + + /* + * raw track access needs to be mutiple of 64k and on 64k boundary + * For read requests we can fix an incorrect alignment by padding + * the request with dummy pages. + */ + start_padding_sectors = blk_rq_pos(req) % DASD_RAW_SECTORS_PER_TRACK; + end_sector_offset = (blk_rq_pos(req) + blk_rq_sectors(req)) % + DASD_RAW_SECTORS_PER_TRACK; + end_padding_sectors = (DASD_RAW_SECTORS_PER_TRACK - end_sector_offset) % + DASD_RAW_SECTORS_PER_TRACK; + basedev = block->base; + if ((start_padding_sectors || end_padding_sectors) && + (rq_data_dir(req) == WRITE)) { + DBF_DEV_EVENT(DBF_ERR, basedev, + "raw write not track aligned (%lu,%lu) req %p", + start_padding_sectors, end_padding_sectors, req); + cqr = ERR_PTR(-EINVAL); + goto out; + } + + first_trk = blk_rq_pos(req) / DASD_RAW_SECTORS_PER_TRACK; + last_trk = (blk_rq_pos(req) + blk_rq_sectors(req) - 1) / + DASD_RAW_SECTORS_PER_TRACK; + trkcount = last_trk - first_trk + 1; + first_offs = 0; + + if (rq_data_dir(req) == READ) + cmd = DASD_ECKD_CCW_READ_TRACK; + else if (rq_data_dir(req) == WRITE) + cmd = DASD_ECKD_CCW_WRITE_FULL_TRACK; + else { + cqr = ERR_PTR(-EINVAL); + goto out; + } + + /* + * Raw track based I/O needs IDAWs for each page, + * and not just for 64 bit addresses. + */ + cidaw = trkcount * DASD_RAW_BLOCK_PER_TRACK; + + /* 1x prefix + one read/write ccw per track */ + cplength = 1 + trkcount; + + /* + * struct PFX_eckd_data has up to 2 byte as extended parameter + * this is needed for write full track and has to be mentioned + * separately + * add 8 instead of 2 to keep 8 byte boundary + */ + pfx_datasize = sizeof(struct PFX_eckd_data) + 8; + + datasize = pfx_datasize + cidaw * sizeof(unsigned long long); + + /* Allocate the ccw request. */ + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, cplength, + datasize, startdev); + if (IS_ERR(cqr)) + goto out; + ccw = cqr->cpaddr; + + if (prefix_LRE(ccw++, cqr->data, first_trk, last_trk, cmd, + basedev, startdev, 1 /* format */, first_offs + 1, + trkcount, 0, 0) == -EAGAIN) { + /* Clock not in sync and XRC is enabled. + * Try again later. + */ + dasd_sfree_request(cqr, startdev); + cqr = ERR_PTR(-EAGAIN); + goto out; + } + + idaws = (unsigned long *)(cqr->data + pfx_datasize); + len_to_track_end = 0; + if (start_padding_sectors) { + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = cmd; + /* maximum 3390 track size */ + ccw->count = 57326; + /* 64k map to one track */ + len_to_track_end = 65536 - start_padding_sectors * 512; + ccw->cda = (__u32)(addr_t)idaws; + ccw->flags |= CCW_FLAG_IDA; + ccw->flags |= CCW_FLAG_SLI; + ccw++; + for (sectors = 0; sectors < start_padding_sectors; sectors += 8) + idaws = idal_create_words(idaws, rawpadpage, PAGE_SIZE); + } + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + seg_len = bv.bv_len; + if (cmd == DASD_ECKD_CCW_READ_TRACK) + memset(dst, 0, seg_len); + if (!len_to_track_end) { + ccw[-1].flags |= CCW_FLAG_CC; + ccw->cmd_code = cmd; + /* maximum 3390 track size */ + ccw->count = 57326; + /* 64k map to one track */ + len_to_track_end = 65536; + ccw->cda = (__u32)(addr_t)idaws; + ccw->flags |= CCW_FLAG_IDA; + ccw->flags |= CCW_FLAG_SLI; + ccw++; + } + len_to_track_end -= seg_len; + idaws = idal_create_words(idaws, dst, seg_len); + } + for (sectors = 0; sectors < end_padding_sectors; sectors += 8) + idaws = idal_create_words(idaws, rawpadpage, PAGE_SIZE); + if (blk_noretry_request(req) || + block->base->features & DASD_FEATURE_FAILFAST) + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->startdev = startdev; + cqr->memdev = startdev; + cqr->block = block; + cqr->expires = startdev->default_expires * HZ; + cqr->lpm = startdev->path_data.ppm; + cqr->retries = startdev->default_retries; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + if (IS_ERR(cqr) && PTR_ERR(cqr) != -EAGAIN) + cqr = NULL; +out: + return cqr; +} + + +static int +dasd_eckd_free_cp(struct dasd_ccw_req *cqr, struct request *req) +{ + struct dasd_eckd_private *private; + struct ccw1 *ccw; + struct req_iterator iter; + struct bio_vec bv; + char *dst, *cda; + unsigned int blksize, blk_per_trk, off; + sector_t recid; + int status; + + if (!dasd_page_cache) + goto out; + private = (struct dasd_eckd_private *) cqr->block->base->private; + blksize = cqr->block->bp_block; + blk_per_trk = recs_per_track(&private->rdc_data, 0, blksize); + recid = blk_rq_pos(req) >> cqr->block->s2b_shift; + ccw = cqr->cpaddr; + /* Skip over define extent & locate record. */ + ccw++; + if (private->uses_cdl == 0 || recid > 2*blk_per_trk) + ccw++; + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + for (off = 0; off < bv.bv_len; off += blksize) { + /* Skip locate record. */ + if (private->uses_cdl && recid <= 2*blk_per_trk) + ccw++; + if (dst) { + if (ccw->flags & CCW_FLAG_IDA) + cda = *((char **)((addr_t) ccw->cda)); + else + cda = (char *)((addr_t) ccw->cda); + if (dst != cda) { + if (rq_data_dir(req) == READ) + memcpy(dst, cda, bv.bv_len); + kmem_cache_free(dasd_page_cache, + (void *)((addr_t)cda & PAGE_MASK)); + } + dst = NULL; + } + ccw++; + recid++; + } + } +out: + status = cqr->status == DASD_CQR_DONE; + dasd_sfree_request(cqr, cqr->memdev); + return status; +} + +/* + * Modify ccw/tcw in cqr so it can be started on a base device. + * + * Note that this is not enough to restart the cqr! + * Either reset cqr->startdev as well (summary unit check handling) + * or restart via separate cqr (as in ERP handling). + */ +void dasd_eckd_reset_ccw_to_base_io(struct dasd_ccw_req *cqr) +{ + struct ccw1 *ccw; + struct PFX_eckd_data *pfxdata; + struct tcw *tcw; + struct tccb *tccb; + struct dcw *dcw; + + if (cqr->cpmode == 1) { + tcw = cqr->cpaddr; + tccb = tcw_get_tccb(tcw); + dcw = (struct dcw *)&tccb->tca[0]; + pfxdata = (struct PFX_eckd_data *)&dcw->cd[0]; + pfxdata->validity.verify_base = 0; + pfxdata->validity.hyper_pav = 0; + } else { + ccw = cqr->cpaddr; + pfxdata = cqr->data; + if (ccw->cmd_code == DASD_ECKD_CCW_PFX) { + pfxdata->validity.verify_base = 0; + pfxdata->validity.hyper_pav = 0; + } + } +} + +#define DASD_ECKD_CHANQ_MAX_SIZE 4 + +static struct dasd_ccw_req *dasd_eckd_build_alias_cp(struct dasd_device *base, + struct dasd_block *block, + struct request *req) +{ + struct dasd_eckd_private *private; + struct dasd_device *startdev; + unsigned long flags; + struct dasd_ccw_req *cqr; + + startdev = dasd_alias_get_start_dev(base); + if (!startdev) + startdev = base; + private = (struct dasd_eckd_private *) startdev->private; + if (private->count >= DASD_ECKD_CHANQ_MAX_SIZE) + return ERR_PTR(-EBUSY); + + spin_lock_irqsave(get_ccwdev_lock(startdev->cdev), flags); + private->count++; + if ((base->features & DASD_FEATURE_USERAW)) + cqr = dasd_raw_build_cp(startdev, block, req); + else + cqr = dasd_eckd_build_cp(startdev, block, req); + if (IS_ERR(cqr)) + private->count--; + spin_unlock_irqrestore(get_ccwdev_lock(startdev->cdev), flags); + return cqr; +} + +static int dasd_eckd_free_alias_cp(struct dasd_ccw_req *cqr, + struct request *req) +{ + struct dasd_eckd_private *private; + unsigned long flags; + + spin_lock_irqsave(get_ccwdev_lock(cqr->memdev->cdev), flags); + private = (struct dasd_eckd_private *) cqr->memdev->private; + private->count--; + spin_unlock_irqrestore(get_ccwdev_lock(cqr->memdev->cdev), flags); + return dasd_eckd_free_cp(cqr, req); +} + +static int +dasd_eckd_fill_info(struct dasd_device * device, + struct dasd_information2_t * info) +{ + struct dasd_eckd_private *private; + + private = (struct dasd_eckd_private *) device->private; + info->label_block = 2; + info->FBA_layout = private->uses_cdl ? 0 : 1; + info->format = private->uses_cdl ? DASD_FORMAT_CDL : DASD_FORMAT_LDL; + info->characteristics_size = sizeof(struct dasd_eckd_characteristics); + memcpy(info->characteristics, &private->rdc_data, + sizeof(struct dasd_eckd_characteristics)); + info->confdata_size = min((unsigned long)private->conf_len, + sizeof(info->configuration_data)); + memcpy(info->configuration_data, private->conf_data, + info->confdata_size); + return 0; +} + +/* + * SECTION: ioctl functions for eckd devices. + */ + +/* + * Release device ioctl. + * Buils a channel programm to releases a prior reserved + * (see dasd_eckd_reserve) device. + */ +static int +dasd_eckd_release(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + int rc; + struct ccw1 *ccw; + int useglobal; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + useglobal = 0; + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1, 32, device); + if (IS_ERR(cqr)) { + mutex_lock(&dasd_reserve_mutex); + useglobal = 1; + cqr = &dasd_reserve_req->cqr; + memset(cqr, 0, sizeof(*cqr)); + memset(&dasd_reserve_req->ccw, 0, + sizeof(dasd_reserve_req->ccw)); + cqr->cpaddr = &dasd_reserve_req->ccw; + cqr->data = &dasd_reserve_req->data; + cqr->magic = DASD_ECKD_MAGIC; + } + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_RELEASE; + ccw->flags |= CCW_FLAG_SLI; + ccw->count = 32; + ccw->cda = (__u32)(addr_t) cqr->data; + cqr->startdev = device; + cqr->memdev = device; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->retries = 2; /* set retry counter to enable basic ERP */ + cqr->expires = 2 * HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + rc = dasd_sleep_on_immediatly(cqr); + if (!rc) + clear_bit(DASD_FLAG_IS_RESERVED, &device->flags); + + if (useglobal) + mutex_unlock(&dasd_reserve_mutex); + else + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * Reserve device ioctl. + * Options are set to 'synchronous wait for interrupt' and + * 'timeout the request'. This leads to a terminate IO if + * the interrupt is outstanding for a certain time. + */ +static int +dasd_eckd_reserve(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + int rc; + struct ccw1 *ccw; + int useglobal; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + useglobal = 0; + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1, 32, device); + if (IS_ERR(cqr)) { + mutex_lock(&dasd_reserve_mutex); + useglobal = 1; + cqr = &dasd_reserve_req->cqr; + memset(cqr, 0, sizeof(*cqr)); + memset(&dasd_reserve_req->ccw, 0, + sizeof(dasd_reserve_req->ccw)); + cqr->cpaddr = &dasd_reserve_req->ccw; + cqr->data = &dasd_reserve_req->data; + cqr->magic = DASD_ECKD_MAGIC; + } + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_RESERVE; + ccw->flags |= CCW_FLAG_SLI; + ccw->count = 32; + ccw->cda = (__u32)(addr_t) cqr->data; + cqr->startdev = device; + cqr->memdev = device; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->retries = 2; /* set retry counter to enable basic ERP */ + cqr->expires = 2 * HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + rc = dasd_sleep_on_immediatly(cqr); + if (!rc) + set_bit(DASD_FLAG_IS_RESERVED, &device->flags); + + if (useglobal) + mutex_unlock(&dasd_reserve_mutex); + else + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * Steal lock ioctl - unconditional reserve device. + * Buils a channel programm to break a device's reservation. + * (unconditional reserve) + */ +static int +dasd_eckd_steal_lock(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + int rc; + struct ccw1 *ccw; + int useglobal; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + useglobal = 0; + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1, 32, device); + if (IS_ERR(cqr)) { + mutex_lock(&dasd_reserve_mutex); + useglobal = 1; + cqr = &dasd_reserve_req->cqr; + memset(cqr, 0, sizeof(*cqr)); + memset(&dasd_reserve_req->ccw, 0, + sizeof(dasd_reserve_req->ccw)); + cqr->cpaddr = &dasd_reserve_req->ccw; + cqr->data = &dasd_reserve_req->data; + cqr->magic = DASD_ECKD_MAGIC; + } + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_SLCK; + ccw->flags |= CCW_FLAG_SLI; + ccw->count = 32; + ccw->cda = (__u32)(addr_t) cqr->data; + cqr->startdev = device; + cqr->memdev = device; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->retries = 2; /* set retry counter to enable basic ERP */ + cqr->expires = 2 * HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + rc = dasd_sleep_on_immediatly(cqr); + if (!rc) + set_bit(DASD_FLAG_IS_RESERVED, &device->flags); + + if (useglobal) + mutex_unlock(&dasd_reserve_mutex); + else + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * SNID - Sense Path Group ID + * This ioctl may be used in situations where I/O is stalled due to + * a reserve, so if the normal dasd_smalloc_request fails, we use the + * preallocated dasd_reserve_req. + */ +static int dasd_eckd_snid(struct dasd_device *device, + void __user *argp) +{ + struct dasd_ccw_req *cqr; + int rc; + struct ccw1 *ccw; + int useglobal; + struct dasd_snid_ioctl_data usrparm; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + if (copy_from_user(&usrparm, argp, sizeof(usrparm))) + return -EFAULT; + + useglobal = 0; + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1, + sizeof(struct dasd_snid_data), device); + if (IS_ERR(cqr)) { + mutex_lock(&dasd_reserve_mutex); + useglobal = 1; + cqr = &dasd_reserve_req->cqr; + memset(cqr, 0, sizeof(*cqr)); + memset(&dasd_reserve_req->ccw, 0, + sizeof(dasd_reserve_req->ccw)); + cqr->cpaddr = &dasd_reserve_req->ccw; + cqr->data = &dasd_reserve_req->data; + cqr->magic = DASD_ECKD_MAGIC; + } + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_SNID; + ccw->flags |= CCW_FLAG_SLI; + ccw->count = 12; + ccw->cda = (__u32)(addr_t) cqr->data; + cqr->startdev = device; + cqr->memdev = device; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + set_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags); + cqr->retries = 5; + cqr->expires = 10 * HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + cqr->lpm = usrparm.path_mask; + + rc = dasd_sleep_on_immediatly(cqr); + /* verify that I/O processing didn't modify the path mask */ + if (!rc && usrparm.path_mask && (cqr->lpm != usrparm.path_mask)) + rc = -EIO; + if (!rc) { + usrparm.data = *((struct dasd_snid_data *)cqr->data); + if (copy_to_user(argp, &usrparm, sizeof(usrparm))) + rc = -EFAULT; + } + + if (useglobal) + mutex_unlock(&dasd_reserve_mutex); + else + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * Read performance statistics + */ +static int +dasd_eckd_performance(struct dasd_device *device, void __user *argp) +{ + struct dasd_psf_prssd_data *prssdp; + struct dasd_rssd_perf_stats_t *stats; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int rc; + + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1 /* PSF */ + 1 /* RSSD */, + (sizeof(struct dasd_psf_prssd_data) + + sizeof(struct dasd_rssd_perf_stats_t)), + device); + if (IS_ERR(cqr)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Could not allocate initialization request"); + return PTR_ERR(cqr); + } + cqr->startdev = device; + cqr->memdev = device; + cqr->retries = 0; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + cqr->expires = 10 * HZ; + + /* Prepare for Read Subsystem Data */ + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + memset(prssdp, 0, sizeof(struct dasd_psf_prssd_data)); + prssdp->order = PSF_ORDER_PRSSD; + prssdp->suborder = 0x01; /* Performance Statistics */ + prssdp->varies[1] = 0x01; /* Perf Statistics for the Subsystem */ + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->count = sizeof(struct dasd_psf_prssd_data); + ccw->flags |= CCW_FLAG_CC; + ccw->cda = (__u32)(addr_t) prssdp; + + /* Read Subsystem Data - Performance Statistics */ + stats = (struct dasd_rssd_perf_stats_t *) (prssdp + 1); + memset(stats, 0, sizeof(struct dasd_rssd_perf_stats_t)); + + ccw++; + ccw->cmd_code = DASD_ECKD_CCW_RSSD; + ccw->count = sizeof(struct dasd_rssd_perf_stats_t); + ccw->cda = (__u32)(addr_t) stats; + + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + rc = dasd_sleep_on(cqr); + if (rc == 0) { + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + stats = (struct dasd_rssd_perf_stats_t *) (prssdp + 1); + if (copy_to_user(argp, stats, + sizeof(struct dasd_rssd_perf_stats_t))) + rc = -EFAULT; + } + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * Get attributes (cache operations) + * Returnes the cache attributes used in Define Extend (DE). + */ +static int +dasd_eckd_get_attrib(struct dasd_device *device, void __user *argp) +{ + struct dasd_eckd_private *private = + (struct dasd_eckd_private *)device->private; + struct attrib_data_t attrib = private->attrib; + int rc; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + if (!argp) + return -EINVAL; + + rc = 0; + if (copy_to_user(argp, (long *) &attrib, + sizeof(struct attrib_data_t))) + rc = -EFAULT; + + return rc; +} + +/* + * Set attributes (cache operations) + * Stores the attributes for cache operation to be used in Define Extend (DE). + */ +static int +dasd_eckd_set_attrib(struct dasd_device *device, void __user *argp) +{ + struct dasd_eckd_private *private = + (struct dasd_eckd_private *)device->private; + struct attrib_data_t attrib; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + if (!argp) + return -EINVAL; + + if (copy_from_user(&attrib, argp, sizeof(struct attrib_data_t))) + return -EFAULT; + private->attrib = attrib; + + dev_info(&device->cdev->dev, + "The DASD cache mode was set to %x (%i cylinder prestage)\n", + private->attrib.operation, private->attrib.nr_cyl); + return 0; +} + +/* + * Issue syscall I/O to EMC Symmetrix array. + * CCWs are PSF and RSSD + */ +static int dasd_symm_io(struct dasd_device *device, void __user *argp) +{ + struct dasd_symmio_parms usrparm; + char *psf_data, *rssd_result; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + char psf0, psf1; + int rc; + + if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RAWIO)) + return -EACCES; + psf0 = psf1 = 0; + + /* Copy parms from caller */ + rc = -EFAULT; + if (copy_from_user(&usrparm, argp, sizeof(usrparm))) + goto out; + if (is_compat_task() || sizeof(long) == 4) { + /* Make sure pointers are sane even on 31 bit. */ + rc = -EINVAL; + if ((usrparm.psf_data >> 32) != 0) + goto out; + if ((usrparm.rssd_result >> 32) != 0) + goto out; + usrparm.psf_data &= 0x7fffffffULL; + usrparm.rssd_result &= 0x7fffffffULL; + } + /* alloc I/O data area */ + psf_data = kzalloc(usrparm.psf_data_len, GFP_KERNEL | GFP_DMA); + rssd_result = kzalloc(usrparm.rssd_result_len, GFP_KERNEL | GFP_DMA); + if (!psf_data || !rssd_result) { + rc = -ENOMEM; + goto out_free; + } + + /* get syscall header from user space */ + rc = -EFAULT; + if (copy_from_user(psf_data, + (void __user *)(unsigned long) usrparm.psf_data, + usrparm.psf_data_len)) + goto out_free; + psf0 = psf_data[0]; + psf1 = psf_data[1]; + + /* setup CCWs for PSF + RSSD */ + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 2 , 0, device); + if (IS_ERR(cqr)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Could not allocate initialization request"); + rc = PTR_ERR(cqr); + goto out_free; + } + + cqr->startdev = device; + cqr->memdev = device; + cqr->retries = 3; + cqr->expires = 10 * HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + /* Build the ccws */ + ccw = cqr->cpaddr; + + /* PSF ccw */ + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->count = usrparm.psf_data_len; + ccw->flags |= CCW_FLAG_CC; + ccw->cda = (__u32)(addr_t) psf_data; + + ccw++; + + /* RSSD ccw */ + ccw->cmd_code = DASD_ECKD_CCW_RSSD; + ccw->count = usrparm.rssd_result_len; + ccw->flags = CCW_FLAG_SLI ; + ccw->cda = (__u32)(addr_t) rssd_result; + + rc = dasd_sleep_on(cqr); + if (rc) + goto out_sfree; + + rc = -EFAULT; + if (copy_to_user((void __user *)(unsigned long) usrparm.rssd_result, + rssd_result, usrparm.rssd_result_len)) + goto out_sfree; + rc = 0; + +out_sfree: + dasd_sfree_request(cqr, cqr->memdev); +out_free: + kfree(rssd_result); + kfree(psf_data); +out: + DBF_DEV_EVENT(DBF_WARNING, device, + "Symmetrix ioctl (0x%02x 0x%02x): rc=%d", + (int) psf0, (int) psf1, rc); + return rc; +} + +static int +dasd_eckd_ioctl(struct dasd_block *block, unsigned int cmd, void __user *argp) +{ + struct dasd_device *device = block->base; + + switch (cmd) { + case BIODASDGATTR: + return dasd_eckd_get_attrib(device, argp); + case BIODASDSATTR: + return dasd_eckd_set_attrib(device, argp); + case BIODASDPSRD: + return dasd_eckd_performance(device, argp); + case BIODASDRLSE: + return dasd_eckd_release(device); + case BIODASDRSRV: + return dasd_eckd_reserve(device); + case BIODASDSLCK: + return dasd_eckd_steal_lock(device); + case BIODASDSNID: + return dasd_eckd_snid(device, argp); + case BIODASDSYMMIO: + return dasd_symm_io(device, argp); + default: + return -ENOTTY; + } +} + +/* + * Dump the range of CCWs into 'page' buffer + * and return number of printed chars. + */ +static int +dasd_eckd_dump_ccw_range(struct ccw1 *from, struct ccw1 *to, char *page) +{ + int len, count; + char *datap; + + len = 0; + while (from <= to) { + len += sprintf(page + len, PRINTK_HEADER + " CCW %p: %08X %08X DAT:", + from, ((int *) from)[0], ((int *) from)[1]); + + /* get pointer to data (consider IDALs) */ + if (from->flags & CCW_FLAG_IDA) + datap = (char *) *((addr_t *) (addr_t) from->cda); + else + datap = (char *) ((addr_t) from->cda); + + /* dump data (max 32 bytes) */ + for (count = 0; count < from->count && count < 32; count++) { + if (count % 8 == 0) len += sprintf(page + len, " "); + if (count % 4 == 0) len += sprintf(page + len, " "); + len += sprintf(page + len, "%02x", datap[count]); + } + len += sprintf(page + len, "\n"); + from++; + } + return len; +} + +static void +dasd_eckd_dump_sense_dbf(struct dasd_device *device, struct irb *irb, + char *reason) +{ + u64 *sense; + u64 *stat; + + sense = (u64 *) dasd_get_sense(irb); + stat = (u64 *) &irb->scsw; + if (sense) { + DBF_DEV_EVENT(DBF_EMERG, device, "%s: %016llx %08x : " + "%016llx %016llx %016llx %016llx", + reason, *stat, *((u32 *) (stat + 1)), + sense[0], sense[1], sense[2], sense[3]); + } else { + DBF_DEV_EVENT(DBF_EMERG, device, "%s: %016llx %08x : %s", + reason, *stat, *((u32 *) (stat + 1)), + "NO VALID SENSE"); + } +} + +/* + * Print sense data and related channel program. + * Parts are printed because printk buffer is only 1024 bytes. + */ +static void dasd_eckd_dump_sense_ccw(struct dasd_device *device, + struct dasd_ccw_req *req, struct irb *irb) +{ + char *page; + struct ccw1 *first, *last, *fail, *from, *to; + int len, sl, sct; + + page = (char *) get_zeroed_page(GFP_ATOMIC); + if (page == NULL) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "No memory to dump sense data\n"); + return; + } + /* dump the sense data */ + len = sprintf(page, PRINTK_HEADER + " I/O status report for device %s:\n", + dev_name(&device->cdev->dev)); + len += sprintf(page + len, PRINTK_HEADER + " in req: %p CC:%02X FC:%02X AC:%02X SC:%02X DS:%02X " + "CS:%02X RC:%d\n", + req, scsw_cc(&irb->scsw), scsw_fctl(&irb->scsw), + scsw_actl(&irb->scsw), scsw_stctl(&irb->scsw), + scsw_dstat(&irb->scsw), scsw_cstat(&irb->scsw), + req ? req->intrc : 0); + len += sprintf(page + len, PRINTK_HEADER + " device %s: Failing CCW: %p\n", + dev_name(&device->cdev->dev), + (void *) (addr_t) irb->scsw.cmd.cpa); + if (irb->esw.esw0.erw.cons) { + for (sl = 0; sl < 4; sl++) { + len += sprintf(page + len, PRINTK_HEADER + " Sense(hex) %2d-%2d:", + (8 * sl), ((8 * sl) + 7)); + + for (sct = 0; sct < 8; sct++) { + len += sprintf(page + len, " %02x", + irb->ecw[8 * sl + sct]); + } + len += sprintf(page + len, "\n"); + } + + if (irb->ecw[27] & DASD_SENSE_BIT_0) { + /* 24 Byte Sense Data */ + sprintf(page + len, PRINTK_HEADER + " 24 Byte: %x MSG %x, " + "%s MSGb to SYSOP\n", + irb->ecw[7] >> 4, irb->ecw[7] & 0x0f, + irb->ecw[1] & 0x10 ? "" : "no"); + } else { + /* 32 Byte Sense Data */ + sprintf(page + len, PRINTK_HEADER + " 32 Byte: Format: %x " + "Exception class %x\n", + irb->ecw[6] & 0x0f, irb->ecw[22] >> 4); + } + } else { + sprintf(page + len, PRINTK_HEADER + " SORRY - NO VALID SENSE AVAILABLE\n"); + } + printk(KERN_ERR "%s", page); + + if (req) { + /* req == NULL for unsolicited interrupts */ + /* dump the Channel Program (max 140 Bytes per line) */ + /* Count CCW and print first CCWs (maximum 1024 % 140 = 7) */ + first = req->cpaddr; + for (last = first; last->flags & (CCW_FLAG_CC | CCW_FLAG_DC); last++); + to = min(first + 6, last); + len = sprintf(page, PRINTK_HEADER + " Related CP in req: %p\n", req); + dasd_eckd_dump_ccw_range(first, to, page + len); + printk(KERN_ERR "%s", page); + + /* print failing CCW area (maximum 4) */ + /* scsw->cda is either valid or zero */ + len = 0; + from = ++to; + fail = (struct ccw1 *)(addr_t) + irb->scsw.cmd.cpa; /* failing CCW */ + if (from < fail - 2) { + from = fail - 2; /* there is a gap - print header */ + len += sprintf(page, PRINTK_HEADER "......\n"); + } + to = min(fail + 1, last); + len += dasd_eckd_dump_ccw_range(from, to, page + len); + + /* print last CCWs (maximum 2) */ + from = max(from, ++to); + if (from < last - 1) { + from = last - 1; /* there is a gap - print header */ + len += sprintf(page + len, PRINTK_HEADER "......\n"); + } + len += dasd_eckd_dump_ccw_range(from, last, page + len); + if (len > 0) + printk(KERN_ERR "%s", page); + } + free_page((unsigned long) page); +} + + +/* + * Print sense data from a tcw. + */ +static void dasd_eckd_dump_sense_tcw(struct dasd_device *device, + struct dasd_ccw_req *req, struct irb *irb) +{ + char *page; + int len, sl, sct, residual; + struct tsb *tsb; + u8 *sense, *rcq; + + page = (char *) get_zeroed_page(GFP_ATOMIC); + if (page == NULL) { + DBF_DEV_EVENT(DBF_WARNING, device, " %s", + "No memory to dump sense data"); + return; + } + /* dump the sense data */ + len = sprintf(page, PRINTK_HEADER + " I/O status report for device %s:\n", + dev_name(&device->cdev->dev)); + len += sprintf(page + len, PRINTK_HEADER + " in req: %p CC:%02X FC:%02X AC:%02X SC:%02X DS:%02X " + "CS:%02X fcxs:%02X schxs:%02X RC:%d\n", + req, scsw_cc(&irb->scsw), scsw_fctl(&irb->scsw), + scsw_actl(&irb->scsw), scsw_stctl(&irb->scsw), + scsw_dstat(&irb->scsw), scsw_cstat(&irb->scsw), + irb->scsw.tm.fcxs, irb->scsw.tm.schxs, + req ? req->intrc : 0); + len += sprintf(page + len, PRINTK_HEADER + " device %s: Failing TCW: %p\n", + dev_name(&device->cdev->dev), + (void *) (addr_t) irb->scsw.tm.tcw); + + tsb = NULL; + sense = NULL; + if (irb->scsw.tm.tcw && (irb->scsw.tm.fcxs & 0x01)) + tsb = tcw_get_tsb( + (struct tcw *)(unsigned long)irb->scsw.tm.tcw); + + if (tsb) { + len += sprintf(page + len, PRINTK_HEADER + " tsb->length %d\n", tsb->length); + len += sprintf(page + len, PRINTK_HEADER + " tsb->flags %x\n", tsb->flags); + len += sprintf(page + len, PRINTK_HEADER + " tsb->dcw_offset %d\n", tsb->dcw_offset); + len += sprintf(page + len, PRINTK_HEADER + " tsb->count %d\n", tsb->count); + residual = tsb->count - 28; + len += sprintf(page + len, PRINTK_HEADER + " residual %d\n", residual); + + switch (tsb->flags & 0x07) { + case 1: /* tsa_iostat */ + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.iostat.dev_time %d\n", + tsb->tsa.iostat.dev_time); + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.iostat.def_time %d\n", + tsb->tsa.iostat.def_time); + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.iostat.queue_time %d\n", + tsb->tsa.iostat.queue_time); + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.iostat.dev_busy_time %d\n", + tsb->tsa.iostat.dev_busy_time); + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.iostat.dev_act_time %d\n", + tsb->tsa.iostat.dev_act_time); + sense = tsb->tsa.iostat.sense; + break; + case 2: /* ts_ddpc */ + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.ddpc.rc %d\n", tsb->tsa.ddpc.rc); + for (sl = 0; sl < 2; sl++) { + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.ddpc.rcq %2d-%2d: ", + (8 * sl), ((8 * sl) + 7)); + rcq = tsb->tsa.ddpc.rcq; + for (sct = 0; sct < 8; sct++) { + len += sprintf(page + len, " %02x", + rcq[8 * sl + sct]); + } + len += sprintf(page + len, "\n"); + } + sense = tsb->tsa.ddpc.sense; + break; + case 3: /* tsa_intrg */ + len += sprintf(page + len, PRINTK_HEADER + " tsb->tsa.intrg.: not supportet yet\n"); + break; + } + + if (sense) { + for (sl = 0; sl < 4; sl++) { + len += sprintf(page + len, PRINTK_HEADER + " Sense(hex) %2d-%2d:", + (8 * sl), ((8 * sl) + 7)); + for (sct = 0; sct < 8; sct++) { + len += sprintf(page + len, " %02x", + sense[8 * sl + sct]); + } + len += sprintf(page + len, "\n"); + } + + if (sense[27] & DASD_SENSE_BIT_0) { + /* 24 Byte Sense Data */ + sprintf(page + len, PRINTK_HEADER + " 24 Byte: %x MSG %x, " + "%s MSGb to SYSOP\n", + sense[7] >> 4, sense[7] & 0x0f, + sense[1] & 0x10 ? "" : "no"); + } else { + /* 32 Byte Sense Data */ + sprintf(page + len, PRINTK_HEADER + " 32 Byte: Format: %x " + "Exception class %x\n", + sense[6] & 0x0f, sense[22] >> 4); + } + } else { + sprintf(page + len, PRINTK_HEADER + " SORRY - NO VALID SENSE AVAILABLE\n"); + } + } else { + sprintf(page + len, PRINTK_HEADER + " SORRY - NO TSB DATA AVAILABLE\n"); + } + printk(KERN_ERR "%s", page); + free_page((unsigned long) page); +} + +static void dasd_eckd_dump_sense(struct dasd_device *device, + struct dasd_ccw_req *req, struct irb *irb) +{ + if (scsw_is_tm(&irb->scsw)) + dasd_eckd_dump_sense_tcw(device, req, irb); + else + dasd_eckd_dump_sense_ccw(device, req, irb); +} + +static int dasd_eckd_pm_freeze(struct dasd_device *device) +{ + /* + * the device should be disconnected from our LCU structure + * on restore we will reconnect it and reread LCU specific + * information like PAV support that might have changed + */ + dasd_alias_remove_device(device); + dasd_alias_disconnect_device_from_lcu(device); + + return 0; +} + +static int dasd_eckd_restore_device(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + struct dasd_eckd_characteristics temp_rdc_data; + int rc; + struct dasd_uid temp_uid; + unsigned long flags; + unsigned long cqr_flags = 0; + + private = (struct dasd_eckd_private *) device->private; + + /* Read Configuration Data */ + dasd_eckd_read_conf(device); + + dasd_eckd_get_uid(device, &temp_uid); + /* Generate device unique id */ + rc = dasd_eckd_generate_uid(device); + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + if (memcmp(&private->uid, &temp_uid, sizeof(struct dasd_uid)) != 0) + dev_err(&device->cdev->dev, "The UID of the DASD has " + "changed\n"); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + if (rc) + goto out_err; + + /* register lcu with alias handling, enable PAV if this is a new lcu */ + rc = dasd_alias_make_device_known_to_lcu(device); + if (rc) + return rc; + + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr_flags); + dasd_eckd_validate_server(device, cqr_flags); + + /* RE-Read Configuration Data */ + dasd_eckd_read_conf(device); + + /* Read Feature Codes */ + dasd_eckd_read_features(device); + + /* Read Device Characteristics */ + rc = dasd_generic_read_dev_chars(device, DASD_ECKD_MAGIC, + &temp_rdc_data, 64); + if (rc) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, + "Read device characteristic failed, rc=%d", rc); + goto out_err; + } + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + memcpy(&private->rdc_data, &temp_rdc_data, sizeof(temp_rdc_data)); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + + /* add device to alias management */ + dasd_alias_add_device(device); + + return 0; + +out_err: + return -1; +} + +static int dasd_eckd_reload_device(struct dasd_device *device) +{ + struct dasd_eckd_private *private; + int rc, old_base; + char print_uid[60]; + struct dasd_uid uid; + unsigned long flags; + + private = (struct dasd_eckd_private *) device->private; + + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + old_base = private->uid.base_unit_addr; + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + + /* Read Configuration Data */ + rc = dasd_eckd_read_conf(device); + if (rc) + goto out_err; + + rc = dasd_eckd_generate_uid(device); + if (rc) + goto out_err; + /* + * update unit address configuration and + * add device to alias management + */ + dasd_alias_update_add_device(device); + + dasd_eckd_get_uid(device, &uid); + + if (old_base != uid.base_unit_addr) { + if (strlen(uid.vduit) > 0) + snprintf(print_uid, sizeof(print_uid), + "%s.%s.%04x.%02x.%s", uid.vendor, uid.serial, + uid.ssid, uid.base_unit_addr, uid.vduit); + else + snprintf(print_uid, sizeof(print_uid), + "%s.%s.%04x.%02x", uid.vendor, uid.serial, + uid.ssid, uid.base_unit_addr); + + dev_info(&device->cdev->dev, + "An Alias device was reassigned to a new base device " + "with UID: %s\n", print_uid); + } + return 0; + +out_err: + return -1; +} + +static int dasd_eckd_read_message_buffer(struct dasd_device *device, + struct dasd_rssd_messages *messages, + __u8 lpum) +{ + struct dasd_rssd_messages *message_buf; + struct dasd_psf_prssd_data *prssdp; + struct dasd_eckd_private *private; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int rc; + + private = (struct dasd_eckd_private *) device->private; + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1 /* PSF */ + 1 /* RSSD */, + (sizeof(struct dasd_psf_prssd_data) + + sizeof(struct dasd_rssd_messages)), + device); + if (IS_ERR(cqr)) { + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", + "Could not allocate read message buffer request"); + return PTR_ERR(cqr); + } + + cqr->startdev = device; + cqr->memdev = device; + cqr->block = NULL; + cqr->retries = 256; + cqr->expires = 10 * HZ; + + /* we need to check for messages on exactly this path */ + set_bit(DASD_CQR_VERIFY_PATH, &cqr->flags); + cqr->lpm = lpum; + + /* Prepare for Read Subsystem Data */ + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + memset(prssdp, 0, sizeof(struct dasd_psf_prssd_data)); + prssdp->order = PSF_ORDER_PRSSD; + prssdp->suborder = 0x03; /* Message Buffer */ + /* all other bytes of prssdp must be zero */ + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->count = sizeof(struct dasd_psf_prssd_data); + ccw->flags |= CCW_FLAG_CC; + ccw->flags |= CCW_FLAG_SLI; + ccw->cda = (__u32)(addr_t) prssdp; + + /* Read Subsystem Data - message buffer */ + message_buf = (struct dasd_rssd_messages *) (prssdp + 1); + memset(message_buf, 0, sizeof(struct dasd_rssd_messages)); + + ccw++; + ccw->cmd_code = DASD_ECKD_CCW_RSSD; + ccw->count = sizeof(struct dasd_rssd_messages); + ccw->flags |= CCW_FLAG_SLI; + ccw->cda = (__u32)(addr_t) message_buf; + + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + rc = dasd_sleep_on_immediatly(cqr); + if (rc == 0) { + prssdp = (struct dasd_psf_prssd_data *) cqr->data; + message_buf = (struct dasd_rssd_messages *) + (prssdp + 1); + memcpy(messages, message_buf, + sizeof(struct dasd_rssd_messages)); + } else + DBF_EVENT_DEVID(DBF_WARNING, device->cdev, + "Reading messages failed with rc=%d\n" + , rc); + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +/* + * Perform Subsystem Function - CUIR response + */ +static int +dasd_eckd_psf_cuir_response(struct dasd_device *device, int response, + __u32 message_id, + struct channel_path_desc *desc, + struct subchannel_id sch_id) +{ + struct dasd_psf_cuir_response *psf_cuir; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int rc; + + cqr = dasd_smalloc_request(DASD_ECKD_MAGIC, 1 /* PSF */ , + sizeof(struct dasd_psf_cuir_response), + device); + + if (IS_ERR(cqr)) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Could not allocate PSF-CUIR request"); + return PTR_ERR(cqr); + } + + psf_cuir = (struct dasd_psf_cuir_response *)cqr->data; + psf_cuir->order = PSF_ORDER_CUIR_RESPONSE; + psf_cuir->cc = response; + if (desc) + psf_cuir->chpid = desc->chpid; + psf_cuir->message_id = message_id; + psf_cuir->cssid = sch_id.cssid; + psf_cuir->ssid = sch_id.ssid; + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->cda = (__u32)(addr_t)psf_cuir; + ccw->count = sizeof(struct dasd_psf_cuir_response); + + cqr->startdev = device; + cqr->memdev = device; + cqr->block = NULL; + cqr->retries = 256; + cqr->expires = 10*HZ; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + + rc = dasd_sleep_on(cqr); + + dasd_sfree_request(cqr, cqr->memdev); + return rc; +} + +static int dasd_eckd_cuir_change_state(struct dasd_device *device, __u8 lpum) +{ + unsigned long flags; + __u8 tbcpm; + + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + tbcpm = device->path_data.opm & ~lpum; + if (tbcpm) { + device->path_data.opm = tbcpm; + device->path_data.cuirpm |= lpum; + } + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + return tbcpm ? 0 : PSF_CUIR_LAST_PATH; +} + +/* + * walk through all devices and quiesce them + * if it is the last path return error + * + * if only part of the devices are quiesced and an error + * occurs no onlining necessary, the storage server will + * notify the already set offline devices again + */ +static int dasd_eckd_cuir_quiesce(struct dasd_device *device, __u8 lpum, + struct channel_path_desc *desc, + struct subchannel_id sch_id) +{ + struct alias_pav_group *pavgroup, *tempgroup; + struct dasd_eckd_private *private; + struct dasd_device *dev, *n; + int rc; + + private = (struct dasd_eckd_private *) device->private; + rc = 0; + + /* active devices */ + list_for_each_entry_safe(dev, n, + &private->lcu->active_devices, + alias_list) { + rc = dasd_eckd_cuir_change_state(dev, lpum); + if (rc) + goto out; + } + + /* inactive devices */ + list_for_each_entry_safe(dev, n, + &private->lcu->inactive_devices, + alias_list) { + rc = dasd_eckd_cuir_change_state(dev, lpum); + if (rc) + goto out; + } + + /* devices in PAV groups */ + list_for_each_entry_safe(pavgroup, tempgroup, + &private->lcu->grouplist, group) { + list_for_each_entry_safe(dev, n, &pavgroup->baselist, + alias_list) { + rc = dasd_eckd_cuir_change_state(dev, lpum); + if (rc) + goto out; + } + list_for_each_entry_safe(dev, n, &pavgroup->aliaslist, + alias_list) { + rc = dasd_eckd_cuir_change_state(dev, lpum); + if (rc) + goto out; + } + } + + pr_warn("Service on the storage server caused path %x.%02x to go offline", + sch_id.cssid, desc ? desc->chpid : 0); + rc = PSF_CUIR_COMPLETED; +out: + return rc; +} + +static int dasd_eckd_cuir_resume(struct dasd_device *device, __u8 lpum, + struct channel_path_desc *desc, + struct subchannel_id sch_id) +{ + struct alias_pav_group *pavgroup, *tempgroup; + struct dasd_eckd_private *private; + struct dasd_device *dev, *n; + + pr_info("Path %x.%02x is back online after service on the storage server", + sch_id.cssid, desc ? desc->chpid : 0); + private = (struct dasd_eckd_private *) device->private; + + /* + * the path may have been added through a generic path event before + * only trigger path verification if the path is not already in use + */ + + list_for_each_entry_safe(dev, n, + &private->lcu->active_devices, + alias_list) { + if (!(dev->path_data.opm & lpum)) { + dev->path_data.tbvpm |= lpum; + dasd_schedule_device_bh(dev); + } + } + + list_for_each_entry_safe(dev, n, + &private->lcu->inactive_devices, + alias_list) { + if (!(dev->path_data.opm & lpum)) { + dev->path_data.tbvpm |= lpum; + dasd_schedule_device_bh(dev); + } + } + + /* devices in PAV groups */ + list_for_each_entry_safe(pavgroup, tempgroup, + &private->lcu->grouplist, + group) { + list_for_each_entry_safe(dev, n, + &pavgroup->baselist, + alias_list) { + if (!(dev->path_data.opm & lpum)) { + dev->path_data.tbvpm |= lpum; + dasd_schedule_device_bh(dev); + } + } + list_for_each_entry_safe(dev, n, + &pavgroup->aliaslist, + alias_list) { + if (!(dev->path_data.opm & lpum)) { + dev->path_data.tbvpm |= lpum; + dasd_schedule_device_bh(dev); + } + } + } + return PSF_CUIR_COMPLETED; +} + +static void dasd_eckd_handle_cuir(struct dasd_device *device, void *messages, + __u8 lpum) +{ + struct dasd_cuir_message *cuir = messages; + struct channel_path_desc *desc; + struct subchannel_id sch_id; + int pos, response; + ccw_device_get_schid(device->cdev, &sch_id); + + /* get position of path in mask */ + pos = 8 - ffs(lpum); + /* get channel path descriptor from this position */ + desc = ccw_device_get_chp_desc(device->cdev, pos); + + if (cuir->code == CUIR_QUIESCE) { + /* quiesce */ + response = dasd_eckd_cuir_quiesce(device, lpum, desc, sch_id); + } else if (cuir->code == CUIR_RESUME) { + /* resume */ + response = dasd_eckd_cuir_resume(device, lpum, desc, sch_id); + } else + response = PSF_CUIR_NOT_SUPPORTED; + + dasd_eckd_psf_cuir_response(device, response, cuir->message_id, + desc, sch_id); + + /* free descriptor copy */ + kfree(desc); +} + +static void dasd_eckd_check_attention_work(struct work_struct *work) +{ + struct check_attention_work_data *data; + struct dasd_rssd_messages *messages; + struct dasd_device *device; + int rc; + + data = container_of(work, struct check_attention_work_data, worker); + device = data->device; + + messages = kzalloc(sizeof(*messages), GFP_KERNEL); + if (!messages) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "Could not allocate attention message buffer"); + goto out; + } + + rc = dasd_eckd_read_message_buffer(device, messages, data->lpum); + if (rc) + goto out; + + if (messages->length == ATTENTION_LENGTH_CUIR && + messages->format == ATTENTION_FORMAT_CUIR) + dasd_eckd_handle_cuir(device, messages, data->lpum); + +out: + dasd_put_device(device); + kfree(messages); + kfree(data); +} + +static int dasd_eckd_check_attention(struct dasd_device *device, __u8 lpum) +{ + struct check_attention_work_data *data; + + data = kzalloc(sizeof(*data), GFP_ATOMIC); + if (!data) + return -ENOMEM; + INIT_WORK(&data->worker, dasd_eckd_check_attention_work); + dasd_get_device(device); + data->device = device; + data->lpum = lpum; + schedule_work(&data->worker); + return 0; +} + +static struct ccw_driver dasd_eckd_driver = { + .driver = { + .name = "dasd-eckd", + .owner = THIS_MODULE, + }, + .ids = dasd_eckd_ids, + .probe = dasd_eckd_probe, + .remove = dasd_generic_remove, + .set_offline = dasd_generic_set_offline, + .set_online = dasd_eckd_set_online, + .notify = dasd_generic_notify, + .path_event = dasd_generic_path_event, + .shutdown = dasd_generic_shutdown, + .freeze = dasd_generic_pm_freeze, + .thaw = dasd_generic_restore_device, + .restore = dasd_generic_restore_device, + .uc_handler = dasd_generic_uc_handler, + .int_class = IRQIO_DAS, +}; + +/* + * max_blocks is dependent on the amount of storage that is available + * in the static io buffer for each device. Currently each device has + * 8192 bytes (=2 pages). For 64 bit one dasd_mchunkt_t structure has + * 24 bytes, the struct dasd_ccw_req has 136 bytes and each block can use + * up to 16 bytes (8 for the ccw and 8 for the idal pointer). In + * addition we have one define extent ccw + 16 bytes of data and one + * locate record ccw + 16 bytes of data. That makes: + * (8192 - 24 - 136 - 8 - 16 - 8 - 16) / 16 = 499 blocks at maximum. + * We want to fit two into the available memory so that we can immediately + * start the next request if one finishes off. That makes 249.5 blocks + * for one request. Give a little safety and the result is 240. + */ +static struct dasd_discipline dasd_eckd_discipline = { + .owner = THIS_MODULE, + .name = "ECKD", + .ebcname = "ECKD", + .max_blocks = 190, + .check_device = dasd_eckd_check_characteristics, + .uncheck_device = dasd_eckd_uncheck_device, + .do_analysis = dasd_eckd_do_analysis, + .verify_path = dasd_eckd_verify_path, + .basic_to_ready = dasd_eckd_basic_to_ready, + .online_to_ready = dasd_eckd_online_to_ready, + .basic_to_known = dasd_eckd_basic_to_known, + .fill_geometry = dasd_eckd_fill_geometry, + .start_IO = dasd_start_IO, + .term_IO = dasd_term_IO, + .handle_terminated_request = dasd_eckd_handle_terminated_request, + .format_device = dasd_eckd_format_device, + .erp_action = dasd_eckd_erp_action, + .erp_postaction = dasd_eckd_erp_postaction, + .check_for_device_change = dasd_eckd_check_for_device_change, + .build_cp = dasd_eckd_build_alias_cp, + .free_cp = dasd_eckd_free_alias_cp, + .dump_sense = dasd_eckd_dump_sense, + .dump_sense_dbf = dasd_eckd_dump_sense_dbf, + .fill_info = dasd_eckd_fill_info, + .ioctl = dasd_eckd_ioctl, + .freeze = dasd_eckd_pm_freeze, + .restore = dasd_eckd_restore_device, + .reload = dasd_eckd_reload_device, + .get_uid = dasd_eckd_get_uid, + .kick_validate = dasd_eckd_kick_validate_server, + .check_attention = dasd_eckd_check_attention, +}; + +static int __init +dasd_eckd_init(void) +{ + int ret; + + ASCEBC(dasd_eckd_discipline.ebcname, 4); + dasd_reserve_req = kmalloc(sizeof(*dasd_reserve_req), + GFP_KERNEL | GFP_DMA); + if (!dasd_reserve_req) + return -ENOMEM; + path_verification_worker = kmalloc(sizeof(*path_verification_worker), + GFP_KERNEL | GFP_DMA); + if (!path_verification_worker) { + kfree(dasd_reserve_req); + return -ENOMEM; + } + rawpadpage = (void *)__get_free_page(GFP_KERNEL); + if (!rawpadpage) { + kfree(path_verification_worker); + kfree(dasd_reserve_req); + return -ENOMEM; + } + ret = ccw_driver_register(&dasd_eckd_driver); + if (!ret) + wait_for_device_probe(); + else { + kfree(path_verification_worker); + kfree(dasd_reserve_req); + free_page((unsigned long)rawpadpage); + } + return ret; +} + +static void __exit +dasd_eckd_cleanup(void) +{ + ccw_driver_unregister(&dasd_eckd_driver); + kfree(path_verification_worker); + kfree(dasd_reserve_req); + free_page((unsigned long)rawpadpage); +} + +module_init(dasd_eckd_init); +module_exit(dasd_eckd_cleanup); diff --git a/kernel/drivers/s390/block/dasd_eckd.h b/kernel/drivers/s390/block/dasd_eckd.h new file mode 100644 index 000000000..ddab7df36 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_eckd.h @@ -0,0 +1,533 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2000 + * + */ + +#ifndef DASD_ECKD_H +#define DASD_ECKD_H + +/***************************************************************************** + * SECTION: CCW Definitions + ****************************************************************************/ +#define DASD_ECKD_CCW_WRITE 0x05 +#define DASD_ECKD_CCW_READ 0x06 +#define DASD_ECKD_CCW_WRITE_HOME_ADDRESS 0x09 +#define DASD_ECKD_CCW_READ_HOME_ADDRESS 0x0a +#define DASD_ECKD_CCW_WRITE_KD 0x0d +#define DASD_ECKD_CCW_READ_KD 0x0e +#define DASD_ECKD_CCW_ERASE 0x11 +#define DASD_ECKD_CCW_READ_COUNT 0x12 +#define DASD_ECKD_CCW_SLCK 0x14 +#define DASD_ECKD_CCW_WRITE_RECORD_ZERO 0x15 +#define DASD_ECKD_CCW_READ_RECORD_ZERO 0x16 +#define DASD_ECKD_CCW_WRITE_CKD 0x1d +#define DASD_ECKD_CCW_READ_CKD 0x1e +#define DASD_ECKD_CCW_PSF 0x27 +#define DASD_ECKD_CCW_SNID 0x34 +#define DASD_ECKD_CCW_RSSD 0x3e +#define DASD_ECKD_CCW_LOCATE_RECORD 0x47 +#define DASD_ECKD_CCW_SNSS 0x54 +#define DASD_ECKD_CCW_DEFINE_EXTENT 0x63 +#define DASD_ECKD_CCW_WRITE_MT 0x85 +#define DASD_ECKD_CCW_READ_MT 0x86 +#define DASD_ECKD_CCW_WRITE_KD_MT 0x8d +#define DASD_ECKD_CCW_READ_KD_MT 0x8e +#define DASD_ECKD_CCW_RELEASE 0x94 +#define DASD_ECKD_CCW_WRITE_FULL_TRACK 0x95 +#define DASD_ECKD_CCW_READ_CKD_MT 0x9e +#define DASD_ECKD_CCW_WRITE_CKD_MT 0x9d +#define DASD_ECKD_CCW_WRITE_TRACK_DATA 0xA5 +#define DASD_ECKD_CCW_READ_TRACK_DATA 0xA6 +#define DASD_ECKD_CCW_RESERVE 0xB4 +#define DASD_ECKD_CCW_READ_TRACK 0xDE +#define DASD_ECKD_CCW_PFX 0xE7 +#define DASD_ECKD_CCW_PFX_READ 0xEA +#define DASD_ECKD_CCW_RSCK 0xF9 +#define DASD_ECKD_CCW_RCD 0xFA + +/* + * Perform Subsystem Function / Sub-Orders + */ +#define PSF_ORDER_PRSSD 0x18 +#define PSF_ORDER_CUIR_RESPONSE 0x1A +#define PSF_ORDER_SSC 0x1D + +/* + * CUIR response condition codes + */ +#define PSF_CUIR_INVALID 0x00 +#define PSF_CUIR_COMPLETED 0x01 +#define PSF_CUIR_NOT_SUPPORTED 0x02 +#define PSF_CUIR_ERROR_IN_REQ 0x03 +#define PSF_CUIR_DENIED 0x04 +#define PSF_CUIR_LAST_PATH 0x05 +#define PSF_CUIR_DEVICE_ONLINE 0x06 +#define PSF_CUIR_VARY_FAILURE 0x07 +#define PSF_CUIR_SOFTWARE_FAILURE 0x08 +#define PSF_CUIR_NOT_RECOGNIZED 0x09 + +/* + * CUIR codes + */ +#define CUIR_QUIESCE 0x01 +#define CUIR_RESUME 0x02 + +/* + * attention message definitions + */ +#define ATTENTION_LENGTH_CUIR 0x0e +#define ATTENTION_FORMAT_CUIR 0x01 + +/* + * Size that is reportet for large volumes in the old 16-bit no_cyl field + */ +#define LV_COMPAT_CYL 0xFFFE + + +#define FCX_MAX_DATA_FACTOR 65536 +#define DASD_ECKD_RCD_DATA_SIZE 256 + + +/***************************************************************************** + * SECTION: Type Definitions + ****************************************************************************/ + +struct eckd_count { + __u16 cyl; + __u16 head; + __u8 record; + __u8 kl; + __u16 dl; +} __attribute__ ((packed)); + +struct ch_t { + __u16 cyl; + __u16 head; +} __attribute__ ((packed)); + +struct chs_t { + __u16 cyl; + __u16 head; + __u32 sector; +} __attribute__ ((packed)); + +struct chr_t { + __u16 cyl; + __u16 head; + __u8 record; +} __attribute__ ((packed)); + +struct geom_t { + __u16 cyl; + __u16 head; + __u32 sector; +} __attribute__ ((packed)); + +struct eckd_home { + __u8 skip_control[14]; + __u16 cell_number; + __u8 physical_addr[3]; + __u8 flag; + struct ch_t track_addr; + __u8 reserved; + __u8 key_length; + __u8 reserved2[2]; +} __attribute__ ((packed)); + +struct DE_eckd_data { + struct { + unsigned char perm:2; /* Permissions on this extent */ + unsigned char reserved:1; + unsigned char seek:2; /* Seek control */ + unsigned char auth:2; /* Access authorization */ + unsigned char pci:1; /* PCI Fetch mode */ + } __attribute__ ((packed)) mask; + struct { + unsigned char mode:2; /* Architecture mode */ + unsigned char ckd:1; /* CKD Conversion */ + unsigned char operation:3; /* Operation mode */ + unsigned char cfw:1; /* Cache fast write */ + unsigned char dfw:1; /* DASD fast write */ + } __attribute__ ((packed)) attributes; + __u16 blk_size; /* Blocksize */ + __u16 fast_write_id; + __u8 ga_additional; /* Global Attributes Additional */ + __u8 ga_extended; /* Global Attributes Extended */ + struct ch_t beg_ext; + struct ch_t end_ext; + unsigned long long ep_sys_time; /* Ext Parameter - System Time Stamp */ + __u8 ep_format; /* Extended Parameter format byte */ + __u8 ep_prio; /* Extended Parameter priority I/O byte */ + __u8 ep_reserved1; /* Extended Parameter Reserved */ + __u8 ep_rec_per_track; /* Number of records on a track */ + __u8 ep_reserved[4]; /* Extended Parameter Reserved */ +} __attribute__ ((packed)); + +struct LO_eckd_data { + struct { + unsigned char orientation:2; + unsigned char operation:6; + } __attribute__ ((packed)) operation; + struct { + unsigned char last_bytes_used:1; + unsigned char reserved:6; + unsigned char read_count_suffix:1; + } __attribute__ ((packed)) auxiliary; + __u8 unused; + __u8 count; + struct ch_t seek_addr; + struct chr_t search_arg; + __u8 sector; + __u16 length; +} __attribute__ ((packed)); + +struct LRE_eckd_data { + struct { + unsigned char orientation:2; + unsigned char operation:6; + } __attribute__ ((packed)) operation; + struct { + unsigned char length_valid:1; + unsigned char length_scope:1; + unsigned char imbedded_ccw_valid:1; + unsigned char check_bytes:2; + unsigned char imbedded_count_valid:1; + unsigned char reserved:1; + unsigned char read_count_suffix:1; + } __attribute__ ((packed)) auxiliary; + __u8 imbedded_ccw; + __u8 count; + struct ch_t seek_addr; + struct chr_t search_arg; + __u8 sector; + __u16 length; + __u8 imbedded_count; + __u8 extended_operation; + __u16 extended_parameter_length; + __u8 extended_parameter[0]; +} __attribute__ ((packed)); + +/* Prefix data for format 0x00 and 0x01 */ +struct PFX_eckd_data { + unsigned char format; + struct { + unsigned char define_extent:1; + unsigned char time_stamp:1; + unsigned char verify_base:1; + unsigned char hyper_pav:1; + unsigned char reserved:4; + } __attribute__ ((packed)) validity; + __u8 base_address; + __u8 aux; + __u8 base_lss; + __u8 reserved[7]; + struct DE_eckd_data define_extent; + struct LRE_eckd_data locate_record; +} __attribute__ ((packed)); + +struct dasd_eckd_characteristics { + __u16 cu_type; + struct { + unsigned char support:2; + unsigned char async:1; + unsigned char reserved:1; + unsigned char cache_info:1; + unsigned char model:3; + } __attribute__ ((packed)) cu_model; + __u16 dev_type; + __u8 dev_model; + struct { + unsigned char mult_burst:1; + unsigned char RT_in_LR:1; + unsigned char reserved1:1; + unsigned char RD_IN_LR:1; + unsigned char reserved2:4; + unsigned char reserved3:8; + unsigned char defect_wr:1; + unsigned char XRC_supported:1; + unsigned char reserved4:1; + unsigned char striping:1; + unsigned char reserved5:4; + unsigned char cfw:1; + unsigned char reserved6:2; + unsigned char cache:1; + unsigned char dual_copy:1; + unsigned char dfw:1; + unsigned char reset_alleg:1; + unsigned char sense_down:1; + } __attribute__ ((packed)) facilities; + __u8 dev_class; + __u8 unit_type; + __u16 no_cyl; + __u16 trk_per_cyl; + __u8 sec_per_trk; + __u8 byte_per_track[3]; + __u16 home_bytes; + __u8 formula; + union { + struct { + __u8 f1; + __u16 f2; + __u16 f3; + } __attribute__ ((packed)) f_0x01; + struct { + __u8 f1; + __u8 f2; + __u8 f3; + __u8 f4; + __u8 f5; + } __attribute__ ((packed)) f_0x02; + } __attribute__ ((packed)) factors; + __u16 first_alt_trk; + __u16 no_alt_trk; + __u16 first_dia_trk; + __u16 no_dia_trk; + __u16 first_sup_trk; + __u16 no_sup_trk; + __u8 MDR_ID; + __u8 OBR_ID; + __u8 director; + __u8 rd_trk_set; + __u16 max_rec_zero; + __u8 reserved1; + __u8 RWANY_in_LR; + __u8 factor6; + __u8 factor7; + __u8 factor8; + __u8 reserved2[3]; + __u8 reserved3[6]; + __u32 long_no_cyl; +} __attribute__ ((packed)); + +/* elements of the configuration data */ +struct dasd_ned { + struct { + __u8 identifier:2; + __u8 token_id:1; + __u8 sno_valid:1; + __u8 subst_sno:1; + __u8 recNED:1; + __u8 emuNED:1; + __u8 reserved:1; + } __attribute__ ((packed)) flags; + __u8 descriptor; + __u8 dev_class; + __u8 reserved; + __u8 dev_type[6]; + __u8 dev_model[3]; + __u8 HDA_manufacturer[3]; + __u8 HDA_location[2]; + __u8 HDA_seqno[12]; + __u8 ID; + __u8 unit_addr; +} __attribute__ ((packed)); + +struct dasd_sneq { + struct { + __u8 identifier:2; + __u8 reserved:6; + } __attribute__ ((packed)) flags; + __u8 res1; + __u16 format; + __u8 res2[4]; /* byte 4- 7 */ + __u8 sua_flags; /* byte 8 */ + __u8 base_unit_addr; /* byte 9 */ + __u8 res3[22]; /* byte 10-31 */ +} __attribute__ ((packed)); + +struct vd_sneq { + struct { + __u8 identifier:2; + __u8 reserved:6; + } __attribute__ ((packed)) flags; + __u8 res1; + __u16 format; + __u8 res2[4]; /* byte 4- 7 */ + __u8 uit[16]; /* byte 8-23 */ + __u8 res3[8]; /* byte 24-31 */ +} __attribute__ ((packed)); + +struct dasd_gneq { + struct { + __u8 identifier:2; + __u8 reserved:6; + } __attribute__ ((packed)) flags; + __u8 reserved[5]; + struct { + __u8 value:2; + __u8 number:6; + } __attribute__ ((packed)) timeout; + __u8 reserved3; + __u16 subsystemID; + __u8 reserved2[22]; +} __attribute__ ((packed)); + +struct dasd_rssd_features { + char feature[256]; +} __attribute__((packed)); + +struct dasd_rssd_messages { + __u16 length; + __u8 format; + __u8 code; + __u32 message_id; + __u8 flags; + char messages[4087]; +} __packed; + +struct dasd_cuir_message { + __u16 length; + __u8 format; + __u8 code; + __u32 message_id; + __u8 flags; + __u8 neq_map[3]; + __u8 ned_map; + __u8 record_selector; +} __packed; + +struct dasd_psf_cuir_response { + __u8 order; + __u8 flags; + __u8 cc; + __u8 chpid; + __u16 device_nr; + __u16 reserved; + __u32 message_id; + __u64 system_id; + __u8 cssid; + __u8 ssid; +} __packed; + +/* + * Perform Subsystem Function - Prepare for Read Subsystem Data + */ +struct dasd_psf_prssd_data { + unsigned char order; + unsigned char flags; + unsigned char reserved[4]; + unsigned char suborder; + unsigned char varies[5]; +} __attribute__ ((packed)); + +/* + * Perform Subsystem Function - Set Subsystem Characteristics + */ +struct dasd_psf_ssc_data { + unsigned char order; + unsigned char flags; + unsigned char cu_type[4]; + unsigned char suborder; + unsigned char reserved[59]; +} __attribute__((packed)); + + +/* + * some structures and definitions for alias handling + */ +struct dasd_unit_address_configuration { + struct { + char ua_type; + char base_ua; + } unit[256]; +} __attribute__((packed)); + + +#define MAX_DEVICES_PER_LCU 256 + +/* flags on the LCU */ +#define NEED_UAC_UPDATE 0x01 +#define UPDATE_PENDING 0x02 + +enum pavtype {NO_PAV, BASE_PAV, HYPER_PAV}; + + +struct alias_root { + struct list_head serverlist; + spinlock_t lock; +}; + +struct alias_server { + struct list_head server; + struct dasd_uid uid; + struct list_head lculist; +}; + +struct summary_unit_check_work_data { + char reason; + struct dasd_device *device; + struct work_struct worker; +}; + +struct read_uac_work_data { + struct dasd_device *device; + struct delayed_work dwork; +}; + +struct alias_lcu { + struct list_head lcu; + struct dasd_uid uid; + enum pavtype pav; + char flags; + spinlock_t lock; + struct list_head grouplist; + struct list_head active_devices; + struct list_head inactive_devices; + struct dasd_unit_address_configuration *uac; + struct summary_unit_check_work_data suc_data; + struct read_uac_work_data ruac_data; + struct dasd_ccw_req *rsu_cqr; + struct completion lcu_setup; +}; + +struct alias_pav_group { + struct list_head group; + struct dasd_uid uid; + struct alias_lcu *lcu; + struct list_head baselist; + struct list_head aliaslist; + struct dasd_device *next; +}; + +struct dasd_eckd_private { + struct dasd_eckd_characteristics rdc_data; + u8 *conf_data; + int conf_len; + /* pointers to specific parts in the conf_data */ + struct dasd_ned *ned; + struct dasd_sneq *sneq; + struct vd_sneq *vdsneq; + struct dasd_gneq *gneq; + + struct eckd_count count_area[5]; + int init_cqr_status; + int uses_cdl; + struct attrib_data_t attrib; /* e.g. cache operations */ + struct dasd_rssd_features features; + u32 real_cyl; + + /* alias managemnet */ + struct dasd_uid uid; + struct alias_pav_group *pavgroup; + struct alias_lcu *lcu; + int count; + + u32 fcx_max_data; +}; + + + +int dasd_alias_make_device_known_to_lcu(struct dasd_device *); +void dasd_alias_disconnect_device_from_lcu(struct dasd_device *); +int dasd_alias_add_device(struct dasd_device *); +int dasd_alias_remove_device(struct dasd_device *); +struct dasd_device *dasd_alias_get_start_dev(struct dasd_device *); +void dasd_alias_handle_summary_unit_check(struct dasd_device *, struct irb *); +void dasd_eckd_reset_ccw_to_base_io(struct dasd_ccw_req *); +void dasd_alias_lcu_setup_complete(struct dasd_device *); +void dasd_alias_wait_for_lcu_setup(struct dasd_device *); +int dasd_alias_update_add_device(struct dasd_device *); +#endif /* DASD_ECKD_H */ diff --git a/kernel/drivers/s390/block/dasd_eer.c b/kernel/drivers/s390/block/dasd_eer.c new file mode 100644 index 000000000..21ef63cf0 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_eer.c @@ -0,0 +1,709 @@ +/* + * Character device driver for extended error reporting. + * + * Copyright IBM Corp. 2005 + * extended error reporting for DASD ECKD devices + * Author(s): Stefan Weinhuber <wein@de.ibm.com> + */ + +#define KMSG_COMPONENT "dasd-eckd" + +#include <linux/init.h> +#include <linux/fs.h> +#include <linux/kernel.h> +#include <linux/miscdevice.h> +#include <linux/module.h> +#include <linux/moduleparam.h> +#include <linux/device.h> +#include <linux/poll.h> +#include <linux/mutex.h> +#include <linux/err.h> +#include <linux/slab.h> + +#include <asm/uaccess.h> +#include <linux/atomic.h> +#include <asm/ebcdic.h> + +#include "dasd_int.h" +#include "dasd_eckd.h" + +#ifdef PRINTK_HEADER +#undef PRINTK_HEADER +#endif /* PRINTK_HEADER */ +#define PRINTK_HEADER "dasd(eer):" + +/* + * SECTION: the internal buffer + */ + +/* + * The internal buffer is meant to store obaque blobs of data, so it does + * not know of higher level concepts like triggers. + * It consists of a number of pages that are used as a ringbuffer. Each data + * blob is stored in a simple record that consists of an integer, which + * contains the size of the following data, and the data bytes themselfes. + * + * To allow for multiple independent readers we create one internal buffer + * each time the device is opened and destroy the buffer when the file is + * closed again. The number of pages used for this buffer is determined by + * the module parmeter eer_pages. + * + * One record can be written to a buffer by using the functions + * - dasd_eer_start_record (one time per record to write the size to the + * buffer and reserve the space for the data) + * - dasd_eer_write_buffer (one or more times per record to write the data) + * The data can be written in several steps but you will have to compute + * the total size up front for the invocation of dasd_eer_start_record. + * If the ringbuffer is full, dasd_eer_start_record will remove the required + * number of old records. + * + * A record is typically read in two steps, first read the integer that + * specifies the size of the following data, then read the data. + * Both can be done by + * - dasd_eer_read_buffer + * + * For all mentioned functions you need to get the bufferlock first and keep + * it until a complete record is written or read. + * + * All information necessary to keep track of an internal buffer is kept in + * a struct eerbuffer. The buffer specific to a file pointer is strored in + * the private_data field of that file. To be able to write data to all + * existing buffers, each buffer is also added to the bufferlist. + * If the user does not want to read a complete record in one go, we have to + * keep track of the rest of the record. residual stores the number of bytes + * that are still to deliver. If the rest of the record is invalidated between + * two reads then residual will be set to -1 so that the next read will fail. + * All entries in the eerbuffer structure are protected with the bufferlock. + * To avoid races between writing to a buffer on the one side and creating + * and destroying buffers on the other side, the bufferlock must also be used + * to protect the bufferlist. + */ + +static int eer_pages = 5; +module_param(eer_pages, int, S_IRUGO|S_IWUSR); + +struct eerbuffer { + struct list_head list; + char **buffer; + int buffersize; + int buffer_page_count; + int head; + int tail; + int residual; +}; + +static LIST_HEAD(bufferlist); +static DEFINE_SPINLOCK(bufferlock); +static DECLARE_WAIT_QUEUE_HEAD(dasd_eer_read_wait_queue); + +/* + * How many free bytes are available on the buffer. + * Needs to be called with bufferlock held. + */ +static int dasd_eer_get_free_bytes(struct eerbuffer *eerb) +{ + if (eerb->head < eerb->tail) + return eerb->tail - eerb->head - 1; + return eerb->buffersize - eerb->head + eerb->tail -1; +} + +/* + * How many bytes of buffer space are used. + * Needs to be called with bufferlock held. + */ +static int dasd_eer_get_filled_bytes(struct eerbuffer *eerb) +{ + + if (eerb->head >= eerb->tail) + return eerb->head - eerb->tail; + return eerb->buffersize - eerb->tail + eerb->head; +} + +/* + * The dasd_eer_write_buffer function just copies count bytes of data + * to the buffer. Make sure to call dasd_eer_start_record first, to + * make sure that enough free space is available. + * Needs to be called with bufferlock held. + */ +static void dasd_eer_write_buffer(struct eerbuffer *eerb, + char *data, int count) +{ + + unsigned long headindex,localhead; + unsigned long rest, len; + char *nextdata; + + nextdata = data; + rest = count; + while (rest > 0) { + headindex = eerb->head / PAGE_SIZE; + localhead = eerb->head % PAGE_SIZE; + len = min(rest, PAGE_SIZE - localhead); + memcpy(eerb->buffer[headindex]+localhead, nextdata, len); + nextdata += len; + rest -= len; + eerb->head += len; + if (eerb->head == eerb->buffersize) + eerb->head = 0; /* wrap around */ + BUG_ON(eerb->head > eerb->buffersize); + } +} + +/* + * Needs to be called with bufferlock held. + */ +static int dasd_eer_read_buffer(struct eerbuffer *eerb, char *data, int count) +{ + + unsigned long tailindex,localtail; + unsigned long rest, len, finalcount; + char *nextdata; + + finalcount = min(count, dasd_eer_get_filled_bytes(eerb)); + nextdata = data; + rest = finalcount; + while (rest > 0) { + tailindex = eerb->tail / PAGE_SIZE; + localtail = eerb->tail % PAGE_SIZE; + len = min(rest, PAGE_SIZE - localtail); + memcpy(nextdata, eerb->buffer[tailindex] + localtail, len); + nextdata += len; + rest -= len; + eerb->tail += len; + if (eerb->tail == eerb->buffersize) + eerb->tail = 0; /* wrap around */ + BUG_ON(eerb->tail > eerb->buffersize); + } + return finalcount; +} + +/* + * Whenever you want to write a blob of data to the internal buffer you + * have to start by using this function first. It will write the number + * of bytes that will be written to the buffer. If necessary it will remove + * old records to make room for the new one. + * Needs to be called with bufferlock held. + */ +static int dasd_eer_start_record(struct eerbuffer *eerb, int count) +{ + int tailcount; + + if (count + sizeof(count) > eerb->buffersize) + return -ENOMEM; + while (dasd_eer_get_free_bytes(eerb) < count + sizeof(count)) { + if (eerb->residual > 0) { + eerb->tail += eerb->residual; + if (eerb->tail >= eerb->buffersize) + eerb->tail -= eerb->buffersize; + eerb->residual = -1; + } + dasd_eer_read_buffer(eerb, (char *) &tailcount, + sizeof(tailcount)); + eerb->tail += tailcount; + if (eerb->tail >= eerb->buffersize) + eerb->tail -= eerb->buffersize; + } + dasd_eer_write_buffer(eerb, (char*) &count, sizeof(count)); + + return 0; +}; + +/* + * Release pages that are not used anymore. + */ +static void dasd_eer_free_buffer_pages(char **buf, int no_pages) +{ + int i; + + for (i = 0; i < no_pages; i++) + free_page((unsigned long) buf[i]); +} + +/* + * Allocate a new set of memory pages. + */ +static int dasd_eer_allocate_buffer_pages(char **buf, int no_pages) +{ + int i; + + for (i = 0; i < no_pages; i++) { + buf[i] = (char *) get_zeroed_page(GFP_KERNEL); + if (!buf[i]) { + dasd_eer_free_buffer_pages(buf, i); + return -ENOMEM; + } + } + return 0; +} + +/* + * SECTION: The extended error reporting functionality + */ + +/* + * When a DASD device driver wants to report an error, it calls the + * function dasd_eer_write and gives the respective trigger ID as + * parameter. Currently there are four kinds of triggers: + * + * DASD_EER_FATALERROR: all kinds of unrecoverable I/O problems + * DASD_EER_PPRCSUSPEND: PPRC was suspended + * DASD_EER_NOPATH: There is no path to the device left. + * DASD_EER_STATECHANGE: The state of the device has changed. + * + * For the first three triggers all required information can be supplied by + * the caller. For these triggers a record is written by the function + * dasd_eer_write_standard_trigger. + * + * The DASD_EER_STATECHANGE trigger is special since a sense subsystem + * status ccw need to be executed to gather the necessary sense data first. + * The dasd_eer_snss function will queue the SNSS request and the request + * callback will then call dasd_eer_write with the DASD_EER_STATCHANGE + * trigger. + * + * To avoid memory allocations at runtime, the necessary memory is allocated + * when the extended error reporting is enabled for a device (by + * dasd_eer_probe). There is one sense subsystem status request for each + * eer enabled DASD device. The presence of the cqr in device->eer_cqr + * indicates that eer is enable for the device. The use of the snss request + * is protected by the DASD_FLAG_EER_IN_USE bit. When this flag indicates + * that the cqr is currently in use, dasd_eer_snss cannot start a second + * request but sets the DASD_FLAG_EER_SNSS flag instead. The callback of + * the SNSS request will check the bit and call dasd_eer_snss again. + */ + +#define SNSS_DATA_SIZE 44 + +#define DASD_EER_BUSID_SIZE 10 +struct dasd_eer_header { + __u32 total_size; + __u32 trigger; + __u64 tv_sec; + __u64 tv_usec; + char busid[DASD_EER_BUSID_SIZE]; +} __attribute__ ((packed)); + +/* + * The following function can be used for those triggers that have + * all necessary data available when the function is called. + * If the parameter cqr is not NULL, the chain of requests will be searched + * for valid sense data, and all valid sense data sets will be added to + * the triggers data. + */ +static void dasd_eer_write_standard_trigger(struct dasd_device *device, + struct dasd_ccw_req *cqr, + int trigger) +{ + struct dasd_ccw_req *temp_cqr; + int data_size; + struct timeval tv; + struct dasd_eer_header header; + unsigned long flags; + struct eerbuffer *eerb; + char *sense; + + /* go through cqr chain and count the valid sense data sets */ + data_size = 0; + for (temp_cqr = cqr; temp_cqr; temp_cqr = temp_cqr->refers) + if (dasd_get_sense(&temp_cqr->irb)) + data_size += 32; + + header.total_size = sizeof(header) + data_size + 4; /* "EOR" */ + header.trigger = trigger; + do_gettimeofday(&tv); + header.tv_sec = tv.tv_sec; + header.tv_usec = tv.tv_usec; + strncpy(header.busid, dev_name(&device->cdev->dev), + DASD_EER_BUSID_SIZE); + + spin_lock_irqsave(&bufferlock, flags); + list_for_each_entry(eerb, &bufferlist, list) { + dasd_eer_start_record(eerb, header.total_size); + dasd_eer_write_buffer(eerb, (char *) &header, sizeof(header)); + for (temp_cqr = cqr; temp_cqr; temp_cqr = temp_cqr->refers) { + sense = dasd_get_sense(&temp_cqr->irb); + if (sense) + dasd_eer_write_buffer(eerb, sense, 32); + } + dasd_eer_write_buffer(eerb, "EOR", 4); + } + spin_unlock_irqrestore(&bufferlock, flags); + wake_up_interruptible(&dasd_eer_read_wait_queue); +} + +/* + * This function writes a DASD_EER_STATECHANGE trigger. + */ +static void dasd_eer_write_snss_trigger(struct dasd_device *device, + struct dasd_ccw_req *cqr, + int trigger) +{ + int data_size; + int snss_rc; + struct timeval tv; + struct dasd_eer_header header; + unsigned long flags; + struct eerbuffer *eerb; + + snss_rc = (cqr->status == DASD_CQR_DONE) ? 0 : -EIO; + if (snss_rc) + data_size = 0; + else + data_size = SNSS_DATA_SIZE; + + header.total_size = sizeof(header) + data_size + 4; /* "EOR" */ + header.trigger = DASD_EER_STATECHANGE; + do_gettimeofday(&tv); + header.tv_sec = tv.tv_sec; + header.tv_usec = tv.tv_usec; + strncpy(header.busid, dev_name(&device->cdev->dev), + DASD_EER_BUSID_SIZE); + + spin_lock_irqsave(&bufferlock, flags); + list_for_each_entry(eerb, &bufferlist, list) { + dasd_eer_start_record(eerb, header.total_size); + dasd_eer_write_buffer(eerb, (char *) &header , sizeof(header)); + if (!snss_rc) + dasd_eer_write_buffer(eerb, cqr->data, SNSS_DATA_SIZE); + dasd_eer_write_buffer(eerb, "EOR", 4); + } + spin_unlock_irqrestore(&bufferlock, flags); + wake_up_interruptible(&dasd_eer_read_wait_queue); +} + +/* + * This function is called for all triggers. It calls the appropriate + * function that writes the actual trigger records. + */ +void dasd_eer_write(struct dasd_device *device, struct dasd_ccw_req *cqr, + unsigned int id) +{ + if (!device->eer_cqr) + return; + switch (id) { + case DASD_EER_FATALERROR: + case DASD_EER_PPRCSUSPEND: + dasd_eer_write_standard_trigger(device, cqr, id); + break; + case DASD_EER_NOPATH: + dasd_eer_write_standard_trigger(device, NULL, id); + break; + case DASD_EER_STATECHANGE: + dasd_eer_write_snss_trigger(device, cqr, id); + break; + default: /* unknown trigger, so we write it without any sense data */ + dasd_eer_write_standard_trigger(device, NULL, id); + break; + } +} +EXPORT_SYMBOL(dasd_eer_write); + +/* + * Start a sense subsystem status request. + * Needs to be called with the device held. + */ +void dasd_eer_snss(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + + cqr = device->eer_cqr; + if (!cqr) /* Device not eer enabled. */ + return; + if (test_and_set_bit(DASD_FLAG_EER_IN_USE, &device->flags)) { + /* Sense subsystem status request in use. */ + set_bit(DASD_FLAG_EER_SNSS, &device->flags); + return; + } + /* cdev is already locked, can't use dasd_add_request_head */ + clear_bit(DASD_FLAG_EER_SNSS, &device->flags); + cqr->status = DASD_CQR_QUEUED; + list_add(&cqr->devlist, &device->ccw_queue); + dasd_schedule_device_bh(device); +} + +/* + * Callback function for use with sense subsystem status request. + */ +static void dasd_eer_snss_cb(struct dasd_ccw_req *cqr, void *data) +{ + struct dasd_device *device = cqr->startdev; + unsigned long flags; + + dasd_eer_write(device, cqr, DASD_EER_STATECHANGE); + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + if (device->eer_cqr == cqr) { + clear_bit(DASD_FLAG_EER_IN_USE, &device->flags); + if (test_bit(DASD_FLAG_EER_SNSS, &device->flags)) + /* Another SNSS has been requested in the meantime. */ + dasd_eer_snss(device); + cqr = NULL; + } + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + if (cqr) + /* + * Extended error recovery has been switched off while + * the SNSS request was running. It could even have + * been switched off and on again in which case there + * is a new ccw in device->eer_cqr. Free the "old" + * snss request now. + */ + dasd_kfree_request(cqr, device); +} + +/* + * Enable error reporting on a given device. + */ +int dasd_eer_enable(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + unsigned long flags; + struct ccw1 *ccw; + + if (device->eer_cqr) + return 0; + + if (!device->discipline || strcmp(device->discipline->name, "ECKD")) + return -EPERM; /* FIXME: -EMEDIUMTYPE ? */ + + cqr = dasd_kmalloc_request(DASD_ECKD_MAGIC, 1 /* SNSS */, + SNSS_DATA_SIZE, device); + if (IS_ERR(cqr)) + return -ENOMEM; + + cqr->startdev = device; + cqr->retries = 255; + cqr->expires = 10 * HZ; + clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + set_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags); + + ccw = cqr->cpaddr; + ccw->cmd_code = DASD_ECKD_CCW_SNSS; + ccw->count = SNSS_DATA_SIZE; + ccw->flags = 0; + ccw->cda = (__u32)(addr_t) cqr->data; + + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + cqr->callback = dasd_eer_snss_cb; + + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + if (!device->eer_cqr) { + device->eer_cqr = cqr; + cqr = NULL; + } + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + if (cqr) + dasd_kfree_request(cqr, device); + return 0; +} + +/* + * Disable error reporting on a given device. + */ +void dasd_eer_disable(struct dasd_device *device) +{ + struct dasd_ccw_req *cqr; + unsigned long flags; + int in_use; + + if (!device->eer_cqr) + return; + spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags); + cqr = device->eer_cqr; + device->eer_cqr = NULL; + clear_bit(DASD_FLAG_EER_SNSS, &device->flags); + in_use = test_and_clear_bit(DASD_FLAG_EER_IN_USE, &device->flags); + spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags); + if (cqr && !in_use) + dasd_kfree_request(cqr, device); +} + +/* + * SECTION: the device operations + */ + +/* + * On the one side we need a lock to access our internal buffer, on the + * other side a copy_to_user can sleep. So we need to copy the data we have + * to transfer in a readbuffer, which is protected by the readbuffer_mutex. + */ +static char readbuffer[PAGE_SIZE]; +static DEFINE_MUTEX(readbuffer_mutex); + +static int dasd_eer_open(struct inode *inp, struct file *filp) +{ + struct eerbuffer *eerb; + unsigned long flags; + + eerb = kzalloc(sizeof(struct eerbuffer), GFP_KERNEL); + if (!eerb) + return -ENOMEM; + eerb->buffer_page_count = eer_pages; + if (eerb->buffer_page_count < 1 || + eerb->buffer_page_count > INT_MAX / PAGE_SIZE) { + kfree(eerb); + DBF_EVENT(DBF_WARNING, "can't open device since module " + "parameter eer_pages is smaller than 1 or" + " bigger than %d", (int)(INT_MAX / PAGE_SIZE)); + return -EINVAL; + } + eerb->buffersize = eerb->buffer_page_count * PAGE_SIZE; + eerb->buffer = kmalloc(eerb->buffer_page_count * sizeof(char *), + GFP_KERNEL); + if (!eerb->buffer) { + kfree(eerb); + return -ENOMEM; + } + if (dasd_eer_allocate_buffer_pages(eerb->buffer, + eerb->buffer_page_count)) { + kfree(eerb->buffer); + kfree(eerb); + return -ENOMEM; + } + filp->private_data = eerb; + spin_lock_irqsave(&bufferlock, flags); + list_add(&eerb->list, &bufferlist); + spin_unlock_irqrestore(&bufferlock, flags); + + return nonseekable_open(inp,filp); +} + +static int dasd_eer_close(struct inode *inp, struct file *filp) +{ + struct eerbuffer *eerb; + unsigned long flags; + + eerb = (struct eerbuffer *) filp->private_data; + spin_lock_irqsave(&bufferlock, flags); + list_del(&eerb->list); + spin_unlock_irqrestore(&bufferlock, flags); + dasd_eer_free_buffer_pages(eerb->buffer, eerb->buffer_page_count); + kfree(eerb->buffer); + kfree(eerb); + + return 0; +} + +static ssize_t dasd_eer_read(struct file *filp, char __user *buf, + size_t count, loff_t *ppos) +{ + int tc,rc; + int tailcount,effective_count; + unsigned long flags; + struct eerbuffer *eerb; + + eerb = (struct eerbuffer *) filp->private_data; + if (mutex_lock_interruptible(&readbuffer_mutex)) + return -ERESTARTSYS; + + spin_lock_irqsave(&bufferlock, flags); + + if (eerb->residual < 0) { /* the remainder of this record */ + /* has been deleted */ + eerb->residual = 0; + spin_unlock_irqrestore(&bufferlock, flags); + mutex_unlock(&readbuffer_mutex); + return -EIO; + } else if (eerb->residual > 0) { + /* OK we still have a second half of a record to deliver */ + effective_count = min(eerb->residual, (int) count); + eerb->residual -= effective_count; + } else { + tc = 0; + while (!tc) { + tc = dasd_eer_read_buffer(eerb, (char *) &tailcount, + sizeof(tailcount)); + if (!tc) { + /* no data available */ + spin_unlock_irqrestore(&bufferlock, flags); + mutex_unlock(&readbuffer_mutex); + if (filp->f_flags & O_NONBLOCK) + return -EAGAIN; + rc = wait_event_interruptible( + dasd_eer_read_wait_queue, + eerb->head != eerb->tail); + if (rc) + return rc; + if (mutex_lock_interruptible(&readbuffer_mutex)) + return -ERESTARTSYS; + spin_lock_irqsave(&bufferlock, flags); + } + } + WARN_ON(tc != sizeof(tailcount)); + effective_count = min(tailcount,(int)count); + eerb->residual = tailcount - effective_count; + } + + tc = dasd_eer_read_buffer(eerb, readbuffer, effective_count); + WARN_ON(tc != effective_count); + + spin_unlock_irqrestore(&bufferlock, flags); + + if (copy_to_user(buf, readbuffer, effective_count)) { + mutex_unlock(&readbuffer_mutex); + return -EFAULT; + } + + mutex_unlock(&readbuffer_mutex); + return effective_count; +} + +static unsigned int dasd_eer_poll(struct file *filp, poll_table *ptable) +{ + unsigned int mask; + unsigned long flags; + struct eerbuffer *eerb; + + eerb = (struct eerbuffer *) filp->private_data; + poll_wait(filp, &dasd_eer_read_wait_queue, ptable); + spin_lock_irqsave(&bufferlock, flags); + if (eerb->head != eerb->tail) + mask = POLLIN | POLLRDNORM ; + else + mask = 0; + spin_unlock_irqrestore(&bufferlock, flags); + return mask; +} + +static const struct file_operations dasd_eer_fops = { + .open = &dasd_eer_open, + .release = &dasd_eer_close, + .read = &dasd_eer_read, + .poll = &dasd_eer_poll, + .owner = THIS_MODULE, + .llseek = noop_llseek, +}; + +static struct miscdevice *dasd_eer_dev = NULL; + +int __init dasd_eer_init(void) +{ + int rc; + + dasd_eer_dev = kzalloc(sizeof(*dasd_eer_dev), GFP_KERNEL); + if (!dasd_eer_dev) + return -ENOMEM; + + dasd_eer_dev->minor = MISC_DYNAMIC_MINOR; + dasd_eer_dev->name = "dasd_eer"; + dasd_eer_dev->fops = &dasd_eer_fops; + + rc = misc_register(dasd_eer_dev); + if (rc) { + kfree(dasd_eer_dev); + dasd_eer_dev = NULL; + DBF_EVENT(DBF_ERR, "%s", "dasd_eer_init could not " + "register misc device"); + return rc; + } + + return 0; +} + +void dasd_eer_exit(void) +{ + if (dasd_eer_dev) { + misc_deregister(dasd_eer_dev); + kfree(dasd_eer_dev); + dasd_eer_dev = NULL; + } +} diff --git a/kernel/drivers/s390/block/dasd_erp.c b/kernel/drivers/s390/block/dasd_erp.c new file mode 100644 index 000000000..e1e88486b --- /dev/null +++ b/kernel/drivers/s390/block/dasd_erp.c @@ -0,0 +1,202 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2001 + * + */ + +#define KMSG_COMPONENT "dasd" + +#include <linux/ctype.h> +#include <linux/init.h> + +#include <asm/debug.h> +#include <asm/ebcdic.h> +#include <asm/uaccess.h> + +/* This is ugly... */ +#define PRINTK_HEADER "dasd_erp:" + +#include "dasd_int.h" + +struct dasd_ccw_req * +dasd_alloc_erp_request(char *magic, int cplength, int datasize, + struct dasd_device * device) +{ + unsigned long flags; + struct dasd_ccw_req *cqr; + char *data; + int size; + + /* Sanity checks */ + BUG_ON( magic == NULL || datasize > PAGE_SIZE || + (cplength*sizeof(struct ccw1)) > PAGE_SIZE); + + size = (sizeof(struct dasd_ccw_req) + 7L) & -8L; + if (cplength > 0) + size += cplength * sizeof(struct ccw1); + if (datasize > 0) + size += datasize; + spin_lock_irqsave(&device->mem_lock, flags); + cqr = (struct dasd_ccw_req *) + dasd_alloc_chunk(&device->erp_chunks, size); + spin_unlock_irqrestore(&device->mem_lock, flags); + if (cqr == NULL) + return ERR_PTR(-ENOMEM); + memset(cqr, 0, sizeof(struct dasd_ccw_req)); + INIT_LIST_HEAD(&cqr->devlist); + INIT_LIST_HEAD(&cqr->blocklist); + data = (char *) cqr + ((sizeof(struct dasd_ccw_req) + 7L) & -8L); + cqr->cpaddr = NULL; + if (cplength > 0) { + cqr->cpaddr = (struct ccw1 *) data; + data += cplength*sizeof(struct ccw1); + memset(cqr->cpaddr, 0, cplength*sizeof(struct ccw1)); + } + cqr->data = NULL; + if (datasize > 0) { + cqr->data = data; + memset(cqr->data, 0, datasize); + } + strncpy((char *) &cqr->magic, magic, 4); + ASCEBC((char *) &cqr->magic, 4); + set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); + dasd_get_device(device); + return cqr; +} + +void +dasd_free_erp_request(struct dasd_ccw_req *cqr, struct dasd_device * device) +{ + unsigned long flags; + + spin_lock_irqsave(&device->mem_lock, flags); + dasd_free_chunk(&device->erp_chunks, cqr); + spin_unlock_irqrestore(&device->mem_lock, flags); + atomic_dec(&device->ref_count); +} + + +/* + * dasd_default_erp_action just retries the current cqr + */ +struct dasd_ccw_req * +dasd_default_erp_action(struct dasd_ccw_req *cqr) +{ + struct dasd_device *device; + + device = cqr->startdev; + + /* just retry - there is nothing to save ... I got no sense data.... */ + if (cqr->retries > 0) { + DBF_DEV_EVENT(DBF_DEBUG, device, + "default ERP called (%i retries left)", + cqr->retries); + if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) + cqr->lpm = device->path_data.opm; + cqr->status = DASD_CQR_FILLED; + } else { + pr_err("%s: default ERP has run out of retries and failed\n", + dev_name(&device->cdev->dev)); + cqr->status = DASD_CQR_FAILED; + cqr->stopclk = get_tod_clock(); + } + return cqr; +} /* end dasd_default_erp_action */ + +/* + * DESCRIPTION + * Frees all ERPs of the current ERP Chain and set the status + * of the original CQR either to DASD_CQR_DONE if ERP was successful + * or to DASD_CQR_FAILED if ERP was NOT successful. + * NOTE: This function is only called if no discipline postaction + * is available + * + * PARAMETER + * erp current erp_head + * + * RETURN VALUES + * cqr pointer to the original CQR + */ +struct dasd_ccw_req *dasd_default_erp_postaction(struct dasd_ccw_req *cqr) +{ + int success; + unsigned long long startclk, stopclk; + struct dasd_device *startdev; + + BUG_ON(cqr->refers == NULL || cqr->function == NULL); + + success = cqr->status == DASD_CQR_DONE; + startclk = cqr->startclk; + stopclk = cqr->stopclk; + startdev = cqr->startdev; + + /* free all ERPs - but NOT the original cqr */ + while (cqr->refers != NULL) { + struct dasd_ccw_req *refers; + + refers = cqr->refers; + /* remove the request from the block queue */ + list_del(&cqr->blocklist); + /* free the finished erp request */ + dasd_free_erp_request(cqr, cqr->memdev); + cqr = refers; + } + + /* set corresponding status to original cqr */ + cqr->startclk = startclk; + cqr->stopclk = stopclk; + cqr->startdev = startdev; + if (success) + cqr->status = DASD_CQR_DONE; + else { + cqr->status = DASD_CQR_FAILED; + cqr->stopclk = get_tod_clock(); + } + + return cqr; + +} /* end default_erp_postaction */ + +void +dasd_log_sense(struct dasd_ccw_req *cqr, struct irb *irb) +{ + struct dasd_device *device; + + device = cqr->startdev; + if (cqr->intrc == -ETIMEDOUT) { + dev_err(&device->cdev->dev, + "A timeout error occurred for cqr %p", cqr); + return; + } + if (cqr->intrc == -ENOLINK) { + dev_err(&device->cdev->dev, + "A transport error occurred for cqr %p", cqr); + return; + } + /* dump sense data */ + if (device->discipline && device->discipline->dump_sense) + device->discipline->dump_sense(device, cqr, irb); +} + +void +dasd_log_sense_dbf(struct dasd_ccw_req *cqr, struct irb *irb) +{ + struct dasd_device *device; + + device = cqr->startdev; + /* dump sense data to s390 debugfeature*/ + if (device->discipline && device->discipline->dump_sense_dbf) + device->discipline->dump_sense_dbf(device, irb, "log"); +} +EXPORT_SYMBOL(dasd_log_sense_dbf); + +EXPORT_SYMBOL(dasd_default_erp_action); +EXPORT_SYMBOL(dasd_default_erp_postaction); +EXPORT_SYMBOL(dasd_alloc_erp_request); +EXPORT_SYMBOL(dasd_free_erp_request); +EXPORT_SYMBOL(dasd_log_sense); + diff --git a/kernel/drivers/s390/block/dasd_fba.c b/kernel/drivers/s390/block/dasd_fba.c new file mode 100644 index 000000000..c9262e789 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_fba.c @@ -0,0 +1,633 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2009 + */ + +#define KMSG_COMPONENT "dasd-fba" + +#include <linux/stddef.h> +#include <linux/kernel.h> +#include <asm/debug.h> + +#include <linux/slab.h> +#include <linux/hdreg.h> /* HDIO_GETGEO */ +#include <linux/bio.h> +#include <linux/module.h> +#include <linux/init.h> + +#include <asm/idals.h> +#include <asm/ebcdic.h> +#include <asm/io.h> +#include <asm/ccwdev.h> + +#include "dasd_int.h" +#include "dasd_fba.h" + +#ifdef PRINTK_HEADER +#undef PRINTK_HEADER +#endif /* PRINTK_HEADER */ +#define PRINTK_HEADER "dasd(fba):" + +#define FBA_DEFAULT_RETRIES 32 + +#define DASD_FBA_CCW_WRITE 0x41 +#define DASD_FBA_CCW_READ 0x42 +#define DASD_FBA_CCW_LOCATE 0x43 +#define DASD_FBA_CCW_DEFINE_EXTENT 0x63 + +MODULE_LICENSE("GPL"); + +static struct dasd_discipline dasd_fba_discipline; + +struct dasd_fba_private { + struct dasd_fba_characteristics rdc_data; +}; + +static struct ccw_device_id dasd_fba_ids[] = { + { CCW_DEVICE_DEVTYPE (0x6310, 0, 0x9336, 0), .driver_info = 0x1}, + { CCW_DEVICE_DEVTYPE (0x3880, 0, 0x3370, 0), .driver_info = 0x2}, + { /* end of list */ }, +}; + +MODULE_DEVICE_TABLE(ccw, dasd_fba_ids); + +static struct ccw_driver dasd_fba_driver; /* see below */ +static int +dasd_fba_probe(struct ccw_device *cdev) +{ + return dasd_generic_probe(cdev, &dasd_fba_discipline); +} + +static int +dasd_fba_set_online(struct ccw_device *cdev) +{ + return dasd_generic_set_online(cdev, &dasd_fba_discipline); +} + +static struct ccw_driver dasd_fba_driver = { + .driver = { + .name = "dasd-fba", + .owner = THIS_MODULE, + }, + .ids = dasd_fba_ids, + .probe = dasd_fba_probe, + .remove = dasd_generic_remove, + .set_offline = dasd_generic_set_offline, + .set_online = dasd_fba_set_online, + .notify = dasd_generic_notify, + .path_event = dasd_generic_path_event, + .freeze = dasd_generic_pm_freeze, + .thaw = dasd_generic_restore_device, + .restore = dasd_generic_restore_device, + .int_class = IRQIO_DAS, +}; + +static void +define_extent(struct ccw1 * ccw, struct DE_fba_data *data, int rw, + int blksize, int beg, int nr) +{ + ccw->cmd_code = DASD_FBA_CCW_DEFINE_EXTENT; + ccw->flags = 0; + ccw->count = 16; + ccw->cda = (__u32) __pa(data); + memset(data, 0, sizeof (struct DE_fba_data)); + if (rw == WRITE) + (data->mask).perm = 0x0; + else if (rw == READ) + (data->mask).perm = 0x1; + else + data->mask.perm = 0x2; + data->blk_size = blksize; + data->ext_loc = beg; + data->ext_end = nr - 1; +} + +static void +locate_record(struct ccw1 * ccw, struct LO_fba_data *data, int rw, + int block_nr, int block_ct) +{ + ccw->cmd_code = DASD_FBA_CCW_LOCATE; + ccw->flags = 0; + ccw->count = 8; + ccw->cda = (__u32) __pa(data); + memset(data, 0, sizeof (struct LO_fba_data)); + if (rw == WRITE) + data->operation.cmd = 0x5; + else if (rw == READ) + data->operation.cmd = 0x6; + else + data->operation.cmd = 0x8; + data->blk_nr = block_nr; + data->blk_ct = block_ct; +} + +static int +dasd_fba_check_characteristics(struct dasd_device *device) +{ + struct dasd_block *block; + struct dasd_fba_private *private; + struct ccw_device *cdev = device->cdev; + int rc; + int readonly; + + private = (struct dasd_fba_private *) device->private; + if (!private) { + private = kzalloc(sizeof(*private), GFP_KERNEL | GFP_DMA); + if (!private) { + dev_warn(&device->cdev->dev, + "Allocating memory for private DASD " + "data failed\n"); + return -ENOMEM; + } + device->private = (void *) private; + } else { + memset(private, 0, sizeof(*private)); + } + block = dasd_alloc_block(); + if (IS_ERR(block)) { + DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s", "could not allocate " + "dasd block structure"); + device->private = NULL; + kfree(private); + return PTR_ERR(block); + } + device->block = block; + block->base = device; + + /* Read Device Characteristics */ + rc = dasd_generic_read_dev_chars(device, DASD_FBA_MAGIC, + &private->rdc_data, 32); + if (rc) { + DBF_EVENT_DEVID(DBF_WARNING, cdev, "Read device " + "characteristics returned error %d", rc); + device->block = NULL; + dasd_free_block(block); + device->private = NULL; + kfree(private); + return rc; + } + + device->default_expires = DASD_EXPIRES; + device->default_retries = FBA_DEFAULT_RETRIES; + device->path_data.opm = LPM_ANYPATH; + + readonly = dasd_device_is_ro(device); + if (readonly) + set_bit(DASD_FLAG_DEVICE_RO, &device->flags); + + dev_info(&device->cdev->dev, + "New FBA DASD %04X/%02X (CU %04X/%02X) with %d MB " + "and %d B/blk%s\n", + cdev->id.dev_type, + cdev->id.dev_model, + cdev->id.cu_type, + cdev->id.cu_model, + ((private->rdc_data.blk_bdsa * + (private->rdc_data.blk_size >> 9)) >> 11), + private->rdc_data.blk_size, + readonly ? ", read-only device" : ""); + return 0; +} + +static int dasd_fba_do_analysis(struct dasd_block *block) +{ + struct dasd_fba_private *private; + int sb, rc; + + private = (struct dasd_fba_private *) block->base->private; + rc = dasd_check_blocksize(private->rdc_data.blk_size); + if (rc) { + DBF_DEV_EVENT(DBF_WARNING, block->base, "unknown blocksize %d", + private->rdc_data.blk_size); + return rc; + } + block->blocks = private->rdc_data.blk_bdsa; + block->bp_block = private->rdc_data.blk_size; + block->s2b_shift = 0; /* bits to shift 512 to get a block */ + for (sb = 512; sb < private->rdc_data.blk_size; sb = sb << 1) + block->s2b_shift++; + return 0; +} + +static int dasd_fba_fill_geometry(struct dasd_block *block, + struct hd_geometry *geo) +{ + if (dasd_check_blocksize(block->bp_block) != 0) + return -EINVAL; + geo->cylinders = (block->blocks << block->s2b_shift) >> 10; + geo->heads = 16; + geo->sectors = 128 >> block->s2b_shift; + return 0; +} + +static dasd_erp_fn_t +dasd_fba_erp_action(struct dasd_ccw_req * cqr) +{ + return dasd_default_erp_action; +} + +static dasd_erp_fn_t +dasd_fba_erp_postaction(struct dasd_ccw_req * cqr) +{ + if (cqr->function == dasd_default_erp_action) + return dasd_default_erp_postaction; + + DBF_DEV_EVENT(DBF_WARNING, cqr->startdev, "unknown ERP action %p", + cqr->function); + return NULL; +} + +static void dasd_fba_check_for_device_change(struct dasd_device *device, + struct dasd_ccw_req *cqr, + struct irb *irb) +{ + char mask; + + /* first of all check for state change pending interrupt */ + mask = DEV_STAT_ATTENTION | DEV_STAT_DEV_END | DEV_STAT_UNIT_EXCEP; + if ((irb->scsw.cmd.dstat & mask) == mask) + dasd_generic_handle_state_change(device); +}; + +static struct dasd_ccw_req *dasd_fba_build_cp(struct dasd_device * memdev, + struct dasd_block *block, + struct request *req) +{ + struct dasd_fba_private *private; + unsigned long *idaws; + struct LO_fba_data *LO_data; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + struct req_iterator iter; + struct bio_vec bv; + char *dst; + int count, cidaw, cplength, datasize; + sector_t recid, first_rec, last_rec; + unsigned int blksize, off; + unsigned char cmd; + + private = (struct dasd_fba_private *) block->base->private; + if (rq_data_dir(req) == READ) { + cmd = DASD_FBA_CCW_READ; + } else if (rq_data_dir(req) == WRITE) { + cmd = DASD_FBA_CCW_WRITE; + } else + return ERR_PTR(-EINVAL); + blksize = block->bp_block; + /* Calculate record id of first and last block. */ + first_rec = blk_rq_pos(req) >> block->s2b_shift; + last_rec = + (blk_rq_pos(req) + blk_rq_sectors(req) - 1) >> block->s2b_shift; + /* Check struct bio and count the number of blocks for the request. */ + count = 0; + cidaw = 0; + rq_for_each_segment(bv, req, iter) { + if (bv.bv_len & (blksize - 1)) + /* Fba can only do full blocks. */ + return ERR_PTR(-EINVAL); + count += bv.bv_len >> (block->s2b_shift + 9); + if (idal_is_needed (page_address(bv.bv_page), bv.bv_len)) + cidaw += bv.bv_len / blksize; + } + /* Paranoia. */ + if (count != last_rec - first_rec + 1) + return ERR_PTR(-EINVAL); + /* 1x define extent + 1x locate record + number of blocks */ + cplength = 2 + count; + /* 1x define extent + 1x locate record */ + datasize = sizeof(struct DE_fba_data) + sizeof(struct LO_fba_data) + + cidaw * sizeof(unsigned long); + /* + * Find out number of additional locate record ccws if the device + * can't do data chaining. + */ + if (private->rdc_data.mode.bits.data_chain == 0) { + cplength += count - 1; + datasize += (count - 1)*sizeof(struct LO_fba_data); + } + /* Allocate the ccw request. */ + cqr = dasd_smalloc_request(DASD_FBA_MAGIC, cplength, datasize, memdev); + if (IS_ERR(cqr)) + return cqr; + ccw = cqr->cpaddr; + /* First ccw is define extent. */ + define_extent(ccw++, cqr->data, rq_data_dir(req), + block->bp_block, blk_rq_pos(req), blk_rq_sectors(req)); + /* Build locate_record + read/write ccws. */ + idaws = (unsigned long *) (cqr->data + sizeof(struct DE_fba_data)); + LO_data = (struct LO_fba_data *) (idaws + cidaw); + /* Locate record for all blocks for smart devices. */ + if (private->rdc_data.mode.bits.data_chain != 0) { + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw++, LO_data++, rq_data_dir(req), 0, count); + } + recid = first_rec; + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + if (dasd_page_cache) { + char *copy = kmem_cache_alloc(dasd_page_cache, + GFP_DMA | __GFP_NOWARN); + if (copy && rq_data_dir(req) == WRITE) + memcpy(copy + bv.bv_offset, dst, bv.bv_len); + if (copy) + dst = copy + bv.bv_offset; + } + for (off = 0; off < bv.bv_len; off += blksize) { + /* Locate record for stupid devices. */ + if (private->rdc_data.mode.bits.data_chain == 0) { + ccw[-1].flags |= CCW_FLAG_CC; + locate_record(ccw, LO_data++, + rq_data_dir(req), + recid - first_rec, 1); + ccw->flags = CCW_FLAG_CC; + ccw++; + } else { + if (recid > first_rec) + ccw[-1].flags |= CCW_FLAG_DC; + else + ccw[-1].flags |= CCW_FLAG_CC; + } + ccw->cmd_code = cmd; + ccw->count = block->bp_block; + if (idal_is_needed(dst, blksize)) { + ccw->cda = (__u32)(addr_t) idaws; + ccw->flags = CCW_FLAG_IDA; + idaws = idal_create_words(idaws, dst, blksize); + } else { + ccw->cda = (__u32)(addr_t) dst; + ccw->flags = 0; + } + ccw++; + dst += blksize; + recid++; + } + } + if (blk_noretry_request(req) || + block->base->features & DASD_FEATURE_FAILFAST) + set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags); + cqr->startdev = memdev; + cqr->memdev = memdev; + cqr->block = block; + cqr->expires = memdev->default_expires * HZ; /* default 5 minutes */ + cqr->retries = memdev->default_retries; + cqr->buildclk = get_tod_clock(); + cqr->status = DASD_CQR_FILLED; + return cqr; +} + +static int +dasd_fba_free_cp(struct dasd_ccw_req *cqr, struct request *req) +{ + struct dasd_fba_private *private; + struct ccw1 *ccw; + struct req_iterator iter; + struct bio_vec bv; + char *dst, *cda; + unsigned int blksize, off; + int status; + + if (!dasd_page_cache) + goto out; + private = (struct dasd_fba_private *) cqr->block->base->private; + blksize = cqr->block->bp_block; + ccw = cqr->cpaddr; + /* Skip over define extent & locate record. */ + ccw++; + if (private->rdc_data.mode.bits.data_chain != 0) + ccw++; + rq_for_each_segment(bv, req, iter) { + dst = page_address(bv.bv_page) + bv.bv_offset; + for (off = 0; off < bv.bv_len; off += blksize) { + /* Skip locate record. */ + if (private->rdc_data.mode.bits.data_chain == 0) + ccw++; + if (dst) { + if (ccw->flags & CCW_FLAG_IDA) + cda = *((char **)((addr_t) ccw->cda)); + else + cda = (char *)((addr_t) ccw->cda); + if (dst != cda) { + if (rq_data_dir(req) == READ) + memcpy(dst, cda, bv.bv_len); + kmem_cache_free(dasd_page_cache, + (void *)((addr_t)cda & PAGE_MASK)); + } + dst = NULL; + } + ccw++; + } + } +out: + status = cqr->status == DASD_CQR_DONE; + dasd_sfree_request(cqr, cqr->memdev); + return status; +} + +static void dasd_fba_handle_terminated_request(struct dasd_ccw_req *cqr) +{ + if (cqr->retries < 0) + cqr->status = DASD_CQR_FAILED; + else + cqr->status = DASD_CQR_FILLED; +}; + +static int +dasd_fba_fill_info(struct dasd_device * device, + struct dasd_information2_t * info) +{ + info->label_block = 1; + info->FBA_layout = 1; + info->format = DASD_FORMAT_LDL; + info->characteristics_size = sizeof(struct dasd_fba_characteristics); + memcpy(info->characteristics, + &((struct dasd_fba_private *) device->private)->rdc_data, + sizeof (struct dasd_fba_characteristics)); + info->confdata_size = 0; + return 0; +} + +static void +dasd_fba_dump_sense_dbf(struct dasd_device *device, struct irb *irb, + char *reason) +{ + u64 *sense; + + sense = (u64 *) dasd_get_sense(irb); + if (sense) { + DBF_DEV_EVENT(DBF_EMERG, device, + "%s: %s %02x%02x%02x %016llx %016llx %016llx " + "%016llx", reason, + scsw_is_tm(&irb->scsw) ? "t" : "c", + scsw_cc(&irb->scsw), scsw_cstat(&irb->scsw), + scsw_dstat(&irb->scsw), sense[0], sense[1], + sense[2], sense[3]); + } else { + DBF_DEV_EVENT(DBF_EMERG, device, "%s", + "SORRY - NO VALID SENSE AVAILABLE\n"); + } +} + + +static void +dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req, + struct irb *irb) +{ + char *page; + struct ccw1 *act, *end, *last; + int len, sl, sct, count; + + page = (char *) get_zeroed_page(GFP_ATOMIC); + if (page == NULL) { + DBF_DEV_EVENT(DBF_WARNING, device, "%s", + "No memory to dump sense data"); + return; + } + len = sprintf(page, PRINTK_HEADER + " I/O status report for device %s:\n", + dev_name(&device->cdev->dev)); + len += sprintf(page + len, PRINTK_HEADER + " in req: %p CS: 0x%02X DS: 0x%02X\n", req, + irb->scsw.cmd.cstat, irb->scsw.cmd.dstat); + len += sprintf(page + len, PRINTK_HEADER + " device %s: Failing CCW: %p\n", + dev_name(&device->cdev->dev), + (void *) (addr_t) irb->scsw.cmd.cpa); + if (irb->esw.esw0.erw.cons) { + for (sl = 0; sl < 4; sl++) { + len += sprintf(page + len, PRINTK_HEADER + " Sense(hex) %2d-%2d:", + (8 * sl), ((8 * sl) + 7)); + + for (sct = 0; sct < 8; sct++) { + len += sprintf(page + len, " %02x", + irb->ecw[8 * sl + sct]); + } + len += sprintf(page + len, "\n"); + } + } else { + len += sprintf(page + len, PRINTK_HEADER + " SORRY - NO VALID SENSE AVAILABLE\n"); + } + printk(KERN_ERR "%s", page); + + /* dump the Channel Program */ + /* print first CCWs (maximum 8) */ + act = req->cpaddr; + for (last = act; last->flags & (CCW_FLAG_CC | CCW_FLAG_DC); last++); + end = min(act + 8, last); + len = sprintf(page, PRINTK_HEADER " Related CP in req: %p\n", req); + while (act <= end) { + len += sprintf(page + len, PRINTK_HEADER + " CCW %p: %08X %08X DAT:", + act, ((int *) act)[0], ((int *) act)[1]); + for (count = 0; count < 32 && count < act->count; + count += sizeof(int)) + len += sprintf(page + len, " %08X", + ((int *) (addr_t) act->cda) + [(count>>2)]); + len += sprintf(page + len, "\n"); + act++; + } + printk(KERN_ERR "%s", page); + + + /* print failing CCW area */ + len = 0; + if (act < ((struct ccw1 *)(addr_t) irb->scsw.cmd.cpa) - 2) { + act = ((struct ccw1 *)(addr_t) irb->scsw.cmd.cpa) - 2; + len += sprintf(page + len, PRINTK_HEADER "......\n"); + } + end = min((struct ccw1 *)(addr_t) irb->scsw.cmd.cpa + 2, last); + while (act <= end) { + len += sprintf(page + len, PRINTK_HEADER + " CCW %p: %08X %08X DAT:", + act, ((int *) act)[0], ((int *) act)[1]); + for (count = 0; count < 32 && count < act->count; + count += sizeof(int)) + len += sprintf(page + len, " %08X", + ((int *) (addr_t) act->cda) + [(count>>2)]); + len += sprintf(page + len, "\n"); + act++; + } + + /* print last CCWs */ + if (act < last - 2) { + act = last - 2; + len += sprintf(page + len, PRINTK_HEADER "......\n"); + } + while (act <= last) { + len += sprintf(page + len, PRINTK_HEADER + " CCW %p: %08X %08X DAT:", + act, ((int *) act)[0], ((int *) act)[1]); + for (count = 0; count < 32 && count < act->count; + count += sizeof(int)) + len += sprintf(page + len, " %08X", + ((int *) (addr_t) act->cda) + [(count>>2)]); + len += sprintf(page + len, "\n"); + act++; + } + if (len > 0) + printk(KERN_ERR "%s", page); + free_page((unsigned long) page); +} + +/* + * max_blocks is dependent on the amount of storage that is available + * in the static io buffer for each device. Currently each device has + * 8192 bytes (=2 pages). For 64 bit one dasd_mchunkt_t structure has + * 24 bytes, the struct dasd_ccw_req has 136 bytes and each block can use + * up to 16 bytes (8 for the ccw and 8 for the idal pointer). In + * addition we have one define extent ccw + 16 bytes of data and a + * locate record ccw for each block (stupid devices!) + 16 bytes of data. + * That makes: + * (8192 - 24 - 136 - 8 - 16) / 40 = 200.2 blocks at maximum. + * We want to fit two into the available memory so that we can immediately + * start the next request if one finishes off. That makes 100.1 blocks + * for one request. Give a little safety and the result is 96. + */ +static struct dasd_discipline dasd_fba_discipline = { + .owner = THIS_MODULE, + .name = "FBA ", + .ebcname = "FBA ", + .max_blocks = 96, + .check_device = dasd_fba_check_characteristics, + .do_analysis = dasd_fba_do_analysis, + .verify_path = dasd_generic_verify_path, + .fill_geometry = dasd_fba_fill_geometry, + .start_IO = dasd_start_IO, + .term_IO = dasd_term_IO, + .handle_terminated_request = dasd_fba_handle_terminated_request, + .erp_action = dasd_fba_erp_action, + .erp_postaction = dasd_fba_erp_postaction, + .check_for_device_change = dasd_fba_check_for_device_change, + .build_cp = dasd_fba_build_cp, + .free_cp = dasd_fba_free_cp, + .dump_sense = dasd_fba_dump_sense, + .dump_sense_dbf = dasd_fba_dump_sense_dbf, + .fill_info = dasd_fba_fill_info, +}; + +static int __init +dasd_fba_init(void) +{ + int ret; + + ASCEBC(dasd_fba_discipline.ebcname, 4); + ret = ccw_driver_register(&dasd_fba_driver); + if (!ret) + wait_for_device_probe(); + + return ret; +} + +static void __exit +dasd_fba_cleanup(void) +{ + ccw_driver_unregister(&dasd_fba_driver); +} + +module_init(dasd_fba_init); +module_exit(dasd_fba_cleanup); diff --git a/kernel/drivers/s390/block/dasd_fba.h b/kernel/drivers/s390/block/dasd_fba.h new file mode 100644 index 000000000..b5d3db0e5 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_fba.h @@ -0,0 +1,71 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Coypright IBM Corp. 1999, 2000 + * + */ + +#ifndef DASD_FBA_H +#define DASD_FBA_H + +struct DE_fba_data { + struct { + unsigned char perm:2; /* Permissions on this extent */ + unsigned char zero:2; /* Must be zero */ + unsigned char da:1; /* usually zero */ + unsigned char diag:1; /* allow diagnose */ + unsigned char zero2:2; /* zero */ + } __attribute__ ((packed)) mask; + __u8 zero; /* Must be zero */ + __u16 blk_size; /* Blocksize */ + __u32 ext_loc; /* Extent locator */ + __u32 ext_beg; /* logical number of block 0 in extent */ + __u32 ext_end; /* logocal number of last block in extent */ +} __attribute__ ((packed)); + +struct LO_fba_data { + struct { + unsigned char zero:4; + unsigned char cmd:4; + } __attribute__ ((packed)) operation; + __u8 auxiliary; + __u16 blk_ct; + __u32 blk_nr; +} __attribute__ ((packed)); + +struct dasd_fba_characteristics { + union { + __u8 c; + struct { + unsigned char reserved:1; + unsigned char overrunnable:1; + unsigned char burst_byte:1; + unsigned char data_chain:1; + unsigned char zeros:4; + } __attribute__ ((packed)) bits; + } __attribute__ ((packed)) mode; + union { + __u8 c; + struct { + unsigned char zero0:1; + unsigned char removable:1; + unsigned char shared:1; + unsigned char zero1:1; + unsigned char mam:1; + unsigned char zeros:3; + } __attribute__ ((packed)) bits; + } __attribute__ ((packed)) features; + __u8 dev_class; + __u8 unit_type; + __u16 blk_size; + __u32 blk_per_cycl; + __u32 blk_per_bound; + __u32 blk_bdsa; + __u32 reserved0; + __u16 reserved1; + __u16 blk_ce; + __u32 reserved2; + __u16 reserved3; +} __attribute__ ((packed)); + +#endif /* DASD_FBA_H */ diff --git a/kernel/drivers/s390/block/dasd_genhd.c b/kernel/drivers/s390/block/dasd_genhd.c new file mode 100644 index 000000000..90f39f79f --- /dev/null +++ b/kernel/drivers/s390/block/dasd_genhd.c @@ -0,0 +1,200 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2001 + * + * gendisk related functions for the dasd driver. + * + */ + +#define KMSG_COMPONENT "dasd" + +#include <linux/interrupt.h> +#include <linux/fs.h> +#include <linux/blkpg.h> + +#include <asm/uaccess.h> + +/* This is ugly... */ +#define PRINTK_HEADER "dasd_gendisk:" + +#include "dasd_int.h" + +/* + * Allocate and register gendisk structure for device. + */ +int dasd_gendisk_alloc(struct dasd_block *block) +{ + struct gendisk *gdp; + struct dasd_device *base; + int len; + + /* Make sure the minor for this device exists. */ + base = block->base; + if (base->devindex >= DASD_PER_MAJOR) + return -EBUSY; + + gdp = alloc_disk(1 << DASD_PARTN_BITS); + if (!gdp) + return -ENOMEM; + + /* Initialize gendisk structure. */ + gdp->major = DASD_MAJOR; + gdp->first_minor = base->devindex << DASD_PARTN_BITS; + gdp->fops = &dasd_device_operations; + gdp->driverfs_dev = &base->cdev->dev; + + /* + * Set device name. + * dasda - dasdz : 26 devices + * dasdaa - dasdzz : 676 devices, added up = 702 + * dasdaaa - dasdzzz : 17576 devices, added up = 18278 + * dasdaaaa - dasdzzzz : 456976 devices, added up = 475252 + */ + len = sprintf(gdp->disk_name, "dasd"); + if (base->devindex > 25) { + if (base->devindex > 701) { + if (base->devindex > 18277) + len += sprintf(gdp->disk_name + len, "%c", + 'a'+(((base->devindex-18278) + /17576)%26)); + len += sprintf(gdp->disk_name + len, "%c", + 'a'+(((base->devindex-702)/676)%26)); + } + len += sprintf(gdp->disk_name + len, "%c", + 'a'+(((base->devindex-26)/26)%26)); + } + len += sprintf(gdp->disk_name + len, "%c", 'a'+(base->devindex%26)); + + if (base->features & DASD_FEATURE_READONLY || + test_bit(DASD_FLAG_DEVICE_RO, &base->flags)) + set_disk_ro(gdp, 1); + dasd_add_link_to_gendisk(gdp, base); + gdp->queue = block->request_queue; + block->gdp = gdp; + set_capacity(block->gdp, 0); + add_disk(block->gdp); + return 0; +} + +/* + * Unregister and free gendisk structure for device. + */ +void dasd_gendisk_free(struct dasd_block *block) +{ + if (block->gdp) { + del_gendisk(block->gdp); + block->gdp->private_data = NULL; + put_disk(block->gdp); + block->gdp = NULL; + } +} + +/* + * Trigger a partition detection. + */ +int dasd_scan_partitions(struct dasd_block *block) +{ + struct block_device *bdev; + int retry, rc; + + retry = 5; + bdev = bdget_disk(block->gdp, 0); + if (!bdev) { + DBF_DEV_EVENT(DBF_ERR, block->base, "%s", + "scan partitions error, bdget returned NULL"); + return -ENODEV; + } + + rc = blkdev_get(bdev, FMODE_READ, NULL); + if (rc < 0) { + DBF_DEV_EVENT(DBF_ERR, block->base, + "scan partitions error, blkdev_get returned %d", + rc); + return -ENODEV; + } + /* + * See fs/partition/check.c:register_disk,rescan_partitions + * Can't call rescan_partitions directly. Use ioctl. + */ + rc = ioctl_by_bdev(bdev, BLKRRPART, 0); + while (rc == -EBUSY && retry > 0) { + schedule(); + rc = ioctl_by_bdev(bdev, BLKRRPART, 0); + retry--; + DBF_DEV_EVENT(DBF_ERR, block->base, + "scan partitions error, retry %d rc %d", + retry, rc); + } + + /* + * Since the matching blkdev_put call to the blkdev_get in + * this function is not called before dasd_destroy_partitions + * the offline open_count limit needs to be increased from + * 0 to 1. This is done by setting device->bdev (see + * dasd_generic_set_offline). As long as the partition + * detection is running no offline should be allowed. That + * is why the assignment to device->bdev is done AFTER + * the BLKRRPART ioctl. + */ + block->bdev = bdev; + return 0; +} + +/* + * Remove all inodes in the system for a device, delete the + * partitions and make device unusable by setting its size to zero. + */ +void dasd_destroy_partitions(struct dasd_block *block) +{ + /* The two structs have 168/176 byte on 31/64 bit. */ + struct blkpg_partition bpart; + struct blkpg_ioctl_arg barg; + struct block_device *bdev; + + /* + * Get the bdev pointer from the device structure and clear + * device->bdev to lower the offline open_count limit again. + */ + bdev = block->bdev; + block->bdev = NULL; + + /* + * See fs/partition/check.c:delete_partition + * Can't call delete_partitions directly. Use ioctl. + * The ioctl also does locking and invalidation. + */ + memset(&bpart, 0, sizeof(struct blkpg_partition)); + memset(&barg, 0, sizeof(struct blkpg_ioctl_arg)); + barg.data = (void __force __user *) &bpart; + barg.op = BLKPG_DEL_PARTITION; + for (bpart.pno = block->gdp->minors - 1; bpart.pno > 0; bpart.pno--) + ioctl_by_bdev(bdev, BLKPG, (unsigned long) &barg); + + invalidate_partition(block->gdp, 0); + /* Matching blkdev_put to the blkdev_get in dasd_scan_partitions. */ + blkdev_put(bdev, FMODE_READ); + set_capacity(block->gdp, 0); +} + +int dasd_gendisk_init(void) +{ + int rc; + + /* Register to static dasd major 94 */ + rc = register_blkdev(DASD_MAJOR, "dasd"); + if (rc != 0) { + pr_warning("Registering the device driver with major number " + "%d failed\n", DASD_MAJOR); + return rc; + } + return 0; +} + +void dasd_gendisk_exit(void) +{ + unregister_blkdev(DASD_MAJOR, "dasd"); +} diff --git a/kernel/drivers/s390/block/dasd_int.h b/kernel/drivers/s390/block/dasd_int.h new file mode 100644 index 000000000..227e3dea3 --- /dev/null +++ b/kernel/drivers/s390/block/dasd_int.h @@ -0,0 +1,814 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2009 + */ + +#ifndef DASD_INT_H +#define DASD_INT_H + +/* we keep old device allocation scheme; IOW, minors are still in 0..255 */ +#define DASD_PER_MAJOR (1U << (MINORBITS - DASD_PARTN_BITS)) +#define DASD_PARTN_MASK ((1 << DASD_PARTN_BITS) - 1) + +/* + * States a dasd device can have: + * new: the dasd_device structure is allocated. + * known: the discipline for the device is identified. + * basic: the device can do basic i/o. + * unfmt: the device could not be analyzed (format is unknown). + * ready: partition detection is done and the device is can do block io. + * online: the device accepts requests from the block device queue. + * + * Things to do for startup state transitions: + * new -> known: find discipline for the device and create devfs entries. + * known -> basic: request irq line for the device. + * basic -> ready: do the initial analysis, e.g. format detection, + * do block device setup and detect partitions. + * ready -> online: schedule the device tasklet. + * Things to do for shutdown state transitions: + * online -> ready: just set the new device state. + * ready -> basic: flush requests from the block device layer, clear + * partition information and reset format information. + * basic -> known: terminate all requests and free irq. + * known -> new: remove devfs entries and forget discipline. + */ + +#define DASD_STATE_NEW 0 +#define DASD_STATE_KNOWN 1 +#define DASD_STATE_BASIC 2 +#define DASD_STATE_UNFMT 3 +#define DASD_STATE_READY 4 +#define DASD_STATE_ONLINE 5 + +#include <linux/module.h> +#include <linux/wait.h> +#include <linux/blkdev.h> +#include <linux/genhd.h> +#include <linux/hdreg.h> +#include <linux/interrupt.h> +#include <linux/log2.h> +#include <asm/ccwdev.h> +#include <linux/workqueue.h> +#include <asm/debug.h> +#include <asm/dasd.h> +#include <asm/idals.h> + +/* DASD discipline magic */ +#define DASD_ECKD_MAGIC 0xC5C3D2C4 +#define DASD_DIAG_MAGIC 0xC4C9C1C7 +#define DASD_FBA_MAGIC 0xC6C2C140 + +/* + * SECTION: Type definitions + */ +struct dasd_device; +struct dasd_block; + +/* BIT DEFINITIONS FOR SENSE DATA */ +#define DASD_SENSE_BIT_0 0x80 +#define DASD_SENSE_BIT_1 0x40 +#define DASD_SENSE_BIT_2 0x20 +#define DASD_SENSE_BIT_3 0x10 + +/* BIT DEFINITIONS FOR SIM SENSE */ +#define DASD_SIM_SENSE 0x0F +#define DASD_SIM_MSG_TO_OP 0x03 +#define DASD_SIM_LOG 0x0C + +/* lock class for nested cdev lock */ +#define CDEV_NESTED_FIRST 1 +#define CDEV_NESTED_SECOND 2 + +/* + * SECTION: MACROs for klogd and s390 debug feature (dbf) + */ +#define DBF_DEV_EVENT(d_level, d_device, d_str, d_data...) \ +do { \ + debug_sprintf_event(d_device->debug_area, \ + d_level, \ + d_str "\n", \ + d_data); \ +} while(0) + +#define DBF_DEV_EXC(d_level, d_device, d_str, d_data...) \ +do { \ + debug_sprintf_exception(d_device->debug_area, \ + d_level, \ + d_str "\n", \ + d_data); \ +} while(0) + +#define DBF_EVENT(d_level, d_str, d_data...)\ +do { \ + debug_sprintf_event(dasd_debug_area, \ + d_level,\ + d_str "\n", \ + d_data); \ +} while(0) + +#define DBF_EVENT_DEVID(d_level, d_cdev, d_str, d_data...) \ +do { \ + struct ccw_dev_id __dev_id; \ + ccw_device_get_id(d_cdev, &__dev_id); \ + debug_sprintf_event(dasd_debug_area, \ + d_level, \ + "0.%x.%04x " d_str "\n", \ + __dev_id.ssid, __dev_id.devno, d_data); \ +} while (0) + +#define DBF_EXC(d_level, d_str, d_data...)\ +do { \ + debug_sprintf_exception(dasd_debug_area, \ + d_level,\ + d_str "\n", \ + d_data); \ +} while(0) + +/* limit size for an errorstring */ +#define ERRORLENGTH 30 + +/* definition of dbf debug levels */ +#define DBF_EMERG 0 /* system is unusable */ +#define DBF_ALERT 1 /* action must be taken immediately */ +#define DBF_CRIT 2 /* critical conditions */ +#define DBF_ERR 3 /* error conditions */ +#define DBF_WARNING 4 /* warning conditions */ +#define DBF_NOTICE 5 /* normal but significant condition */ +#define DBF_INFO 6 /* informational */ +#define DBF_DEBUG 6 /* debug-level messages */ + +/* messages to be written via klogd and dbf */ +#define DEV_MESSAGE(d_loglevel,d_device,d_string,d_args...)\ +do { \ + printk(d_loglevel PRINTK_HEADER " %s: " d_string "\n", \ + dev_name(&d_device->cdev->dev), d_args); \ + DBF_DEV_EVENT(DBF_ALERT, d_device, d_string, d_args); \ +} while(0) + +#define MESSAGE(d_loglevel,d_string,d_args...)\ +do { \ + printk(d_loglevel PRINTK_HEADER " " d_string "\n", d_args); \ + DBF_EVENT(DBF_ALERT, d_string, d_args); \ +} while(0) + +/* messages to be written via klogd only */ +#define DEV_MESSAGE_LOG(d_loglevel,d_device,d_string,d_args...)\ +do { \ + printk(d_loglevel PRINTK_HEADER " %s: " d_string "\n", \ + dev_name(&d_device->cdev->dev), d_args); \ +} while(0) + +#define MESSAGE_LOG(d_loglevel,d_string,d_args...)\ +do { \ + printk(d_loglevel PRINTK_HEADER " " d_string "\n", d_args); \ +} while(0) + +struct dasd_ccw_req { + unsigned int magic; /* Eye catcher */ + struct list_head devlist; /* for dasd_device request queue */ + struct list_head blocklist; /* for dasd_block request queue */ + + /* Where to execute what... */ + struct dasd_block *block; /* the originating block device */ + struct dasd_device *memdev; /* the device used to allocate this */ + struct dasd_device *startdev; /* device the request is started on */ + struct dasd_device *basedev; /* base device if no block->base */ + void *cpaddr; /* address of ccw or tcw */ + unsigned char cpmode; /* 0 = cmd mode, 1 = itcw */ + char status; /* status of this request */ + short retries; /* A retry counter */ + unsigned long flags; /* flags of this request */ + + /* ... and how */ + unsigned long starttime; /* jiffies time of request start */ + unsigned long expires; /* expiration period in jiffies */ + char lpm; /* logical path mask */ + void *data; /* pointer to data area */ + + /* these are important for recovering erroneous requests */ + int intrc; /* internal error, e.g. from start_IO */ + struct irb irb; /* device status in case of an error */ + struct dasd_ccw_req *refers; /* ERP-chain queueing. */ + void *function; /* originating ERP action */ + + /* these are for statistics only */ + unsigned long long buildclk; /* TOD-clock of request generation */ + unsigned long long startclk; /* TOD-clock of request start */ + unsigned long long stopclk; /* TOD-clock of request interrupt */ + unsigned long long endclk; /* TOD-clock of request termination */ + + /* Callback that is called after reaching final status. */ + void (*callback)(struct dasd_ccw_req *, void *data); + void *callback_data; +}; + +/* + * dasd_ccw_req -> status can be: + */ +#define DASD_CQR_FILLED 0x00 /* request is ready to be processed */ +#define DASD_CQR_DONE 0x01 /* request is completed successfully */ +#define DASD_CQR_NEED_ERP 0x02 /* request needs recovery action */ +#define DASD_CQR_IN_ERP 0x03 /* request is in recovery */ +#define DASD_CQR_FAILED 0x04 /* request is finally failed */ +#define DASD_CQR_TERMINATED 0x05 /* request was stopped by driver */ + +#define DASD_CQR_QUEUED 0x80 /* request is queued to be processed */ +#define DASD_CQR_IN_IO 0x81 /* request is currently in IO */ +#define DASD_CQR_ERROR 0x82 /* request is completed with error */ +#define DASD_CQR_CLEAR_PENDING 0x83 /* request is clear pending */ +#define DASD_CQR_CLEARED 0x84 /* request was cleared */ +#define DASD_CQR_SUCCESS 0x85 /* request was successful */ + +/* default expiration time*/ +#define DASD_EXPIRES 300 +#define DASD_EXPIRES_MAX 40000000 +#define DASD_RETRIES 256 +#define DASD_RETRIES_MAX 32768 + +/* per dasd_ccw_req flags */ +#define DASD_CQR_FLAGS_USE_ERP 0 /* use ERP for this request */ +#define DASD_CQR_FLAGS_FAILFAST 1 /* FAILFAST */ +#define DASD_CQR_VERIFY_PATH 2 /* path verification request */ +#define DASD_CQR_ALLOW_SLOCK 3 /* Try this request even when lock was + * stolen. Should not be combined with + * DASD_CQR_FLAGS_USE_ERP + */ + +/* Signature for error recovery functions. */ +typedef struct dasd_ccw_req *(*dasd_erp_fn_t) (struct dasd_ccw_req *); + +/* + * Unique identifier for dasd device. + */ +#define UA_NOT_CONFIGURED 0x00 +#define UA_BASE_DEVICE 0x01 +#define UA_BASE_PAV_ALIAS 0x02 +#define UA_HYPER_PAV_ALIAS 0x03 + +struct dasd_uid { + __u8 type; + char vendor[4]; + char serial[15]; + __u16 ssid; + __u8 real_unit_addr; + __u8 base_unit_addr; + char vduit[33]; +}; + +/* + * the struct dasd_discipline is + * sth like a table of virtual functions, if you think of dasd_eckd + * inheriting dasd... + * no, currently we are not planning to reimplement the driver in C++ + */ +struct dasd_discipline { + struct module *owner; + char ebcname[8]; /* a name used for tagging and printks */ + char name[8]; /* a name used for tagging and printks */ + int max_blocks; /* maximum number of blocks to be chained */ + + struct list_head list; /* used for list of disciplines */ + + /* + * Device recognition functions. check_device is used to verify + * the sense data and the information returned by read device + * characteristics. It returns 0 if the discipline can be used + * for the device in question. uncheck_device is called during + * device shutdown to deregister a device from its discipline. + */ + int (*check_device) (struct dasd_device *); + void (*uncheck_device) (struct dasd_device *); + + /* + * do_analysis is used in the step from device state "basic" to + * state "accept". It returns 0 if the device can be made ready, + * it returns -EMEDIUMTYPE if the device can't be made ready or + * -EAGAIN if do_analysis started a ccw that needs to complete + * before the analysis may be repeated. + */ + int (*do_analysis) (struct dasd_block *); + + /* + * This function is called, when new paths become available. + * Disciplins may use this callback to do necessary setup work, + * e.g. verify that new path is compatible with the current + * configuration. + */ + int (*verify_path)(struct dasd_device *, __u8); + + /* + * Last things to do when a device is set online, and first things + * when it is set offline. + */ + int (*basic_to_ready) (struct dasd_device *); + int (*online_to_ready) (struct dasd_device *); + int (*basic_to_known)(struct dasd_device *); + + /* (struct dasd_device *); + * Device operation functions. build_cp creates a ccw chain for + * a block device request, start_io starts the request and + * term_IO cancels it (e.g. in case of a timeout). format_device + * returns a ccw chain to be used to format the device. + * handle_terminated_request allows to examine a cqr and prepare + * it for retry. + */ + struct dasd_ccw_req *(*build_cp) (struct dasd_device *, + struct dasd_block *, + struct request *); + int (*start_IO) (struct dasd_ccw_req *); + int (*term_IO) (struct dasd_ccw_req *); + void (*handle_terminated_request) (struct dasd_ccw_req *); + int (*format_device) (struct dasd_device *, + struct format_data_t *, int enable_pav); + int (*free_cp) (struct dasd_ccw_req *, struct request *); + + /* + * Error recovery functions. examine_error() returns a value that + * indicates what to do for an error condition. If examine_error() + * returns 'dasd_era_recover' erp_action() is called to create a + * special error recovery ccw. erp_postaction() is called after + * an error recovery ccw has finished its execution. dump_sense + * is called for every error condition to print the sense data + * to the console. + */ + dasd_erp_fn_t(*erp_action) (struct dasd_ccw_req *); + dasd_erp_fn_t(*erp_postaction) (struct dasd_ccw_req *); + void (*dump_sense) (struct dasd_device *, struct dasd_ccw_req *, + struct irb *); + void (*dump_sense_dbf) (struct dasd_device *, struct irb *, char *); + void (*check_for_device_change) (struct dasd_device *, + struct dasd_ccw_req *, + struct irb *); + + /* i/o control functions. */ + int (*fill_geometry) (struct dasd_block *, struct hd_geometry *); + int (*fill_info) (struct dasd_device *, struct dasd_information2_t *); + int (*ioctl) (struct dasd_block *, unsigned int, void __user *); + + /* suspend/resume functions */ + int (*freeze) (struct dasd_device *); + int (*restore) (struct dasd_device *); + + /* reload device after state change */ + int (*reload) (struct dasd_device *); + + int (*get_uid) (struct dasd_device *, struct dasd_uid *); + void (*kick_validate) (struct dasd_device *); + int (*check_attention)(struct dasd_device *, __u8); +}; + +extern struct dasd_discipline *dasd_diag_discipline_pointer; + +/* + * Notification numbers for extended error reporting notifications: + * The DASD_EER_DISABLE notification is sent before a dasd_device (and it's + * eer pointer) is freed. The error reporting module needs to do all necessary + * cleanup steps. + * The DASD_EER_TRIGGER notification sends the actual error reports (triggers). + */ +#define DASD_EER_DISABLE 0 +#define DASD_EER_TRIGGER 1 + +/* Trigger IDs for extended error reporting DASD_EER_TRIGGER notification */ +#define DASD_EER_FATALERROR 1 +#define DASD_EER_NOPATH 2 +#define DASD_EER_STATECHANGE 3 +#define DASD_EER_PPRCSUSPEND 4 + +struct dasd_path { + __u8 opm; + __u8 tbvpm; + __u8 ppm; + __u8 npm; + /* paths that are not used because of a special condition */ + __u8 cablepm; /* miss-cabled */ + __u8 hpfpm; /* the HPF requirements of the other paths are not met */ + __u8 cuirpm; /* CUIR varied offline */ +}; + +struct dasd_profile_info { + /* legacy part of profile data, as in dasd_profile_info_t */ + unsigned int dasd_io_reqs; /* number of requests processed */ + unsigned int dasd_io_sects; /* number of sectors processed */ + unsigned int dasd_io_secs[32]; /* histogram of request's sizes */ + unsigned int dasd_io_times[32]; /* histogram of requests's times */ + unsigned int dasd_io_timps[32]; /* h. of requests's times per sector */ + unsigned int dasd_io_time1[32]; /* hist. of time from build to start */ + unsigned int dasd_io_time2[32]; /* hist. of time from start to irq */ + unsigned int dasd_io_time2ps[32]; /* hist. of time from start to irq */ + unsigned int dasd_io_time3[32]; /* hist. of time from irq to end */ + unsigned int dasd_io_nr_req[32]; /* hist. of # of requests in chanq */ + + /* new data */ + struct timespec starttod; /* time of start or last reset */ + unsigned int dasd_io_alias; /* requests using an alias */ + unsigned int dasd_io_tpm; /* requests using transport mode */ + unsigned int dasd_read_reqs; /* total number of read requests */ + unsigned int dasd_read_sects; /* total number read sectors */ + unsigned int dasd_read_alias; /* read request using an alias */ + unsigned int dasd_read_tpm; /* read requests in transport mode */ + unsigned int dasd_read_secs[32]; /* histogram of request's sizes */ + unsigned int dasd_read_times[32]; /* histogram of requests's times */ + unsigned int dasd_read_time1[32]; /* hist. time from build to start */ + unsigned int dasd_read_time2[32]; /* hist. of time from start to irq */ + unsigned int dasd_read_time3[32]; /* hist. of time from irq to end */ + unsigned int dasd_read_nr_req[32]; /* hist. of # of requests in chanq */ +}; + +struct dasd_profile { + struct dentry *dentry; + struct dasd_profile_info *data; + spinlock_t lock; +}; + +struct dasd_device { + /* Block device stuff. */ + struct dasd_block *block; + + unsigned int devindex; + unsigned long flags; /* per device flags */ + unsigned short features; /* copy of devmap-features (read-only!) */ + + /* extended error reporting stuff (eer) */ + struct dasd_ccw_req *eer_cqr; + + /* Device discipline stuff. */ + struct dasd_discipline *discipline; + struct dasd_discipline *base_discipline; + char *private; + struct dasd_path path_data; + + /* Device state and target state. */ + int state, target; + struct mutex state_mutex; + int stopped; /* device (ccw_device_start) was stopped */ + + /* reference count. */ + atomic_t ref_count; + + /* ccw queue and memory for static ccw/erp buffers. */ + struct list_head ccw_queue; + spinlock_t mem_lock; + void *ccw_mem; + void *erp_mem; + struct list_head ccw_chunks; + struct list_head erp_chunks; + + atomic_t tasklet_scheduled; + struct tasklet_struct tasklet; + struct work_struct kick_work; + struct work_struct restore_device; + struct work_struct reload_device; + struct work_struct kick_validate; + struct timer_list timer; + + debug_info_t *debug_area; + + struct ccw_device *cdev; + + /* hook for alias management */ + struct list_head alias_list; + + /* default expiration time in s */ + unsigned long default_expires; + unsigned long default_retries; + + unsigned long blk_timeout; + + struct dentry *debugfs_dentry; + struct dasd_profile profile; +}; + +struct dasd_block { + /* Block device stuff. */ + struct gendisk *gdp; + struct request_queue *request_queue; + spinlock_t request_queue_lock; + struct block_device *bdev; + atomic_t open_count; + + unsigned long long blocks; /* size of volume in blocks */ + unsigned int bp_block; /* bytes per block */ + unsigned int s2b_shift; /* log2 (bp_block/512) */ + + struct dasd_device *base; + struct list_head ccw_queue; + spinlock_t queue_lock; + + atomic_t tasklet_scheduled; + struct tasklet_struct tasklet; + struct timer_list timer; + + struct dentry *debugfs_dentry; + struct dasd_profile profile; +}; + +struct dasd_attention_data { + struct dasd_device *device; + __u8 lpum; +}; + +/* reasons why device (ccw_device_start) was stopped */ +#define DASD_STOPPED_NOT_ACC 1 /* not accessible */ +#define DASD_STOPPED_QUIESCE 2 /* Quiesced */ +#define DASD_STOPPED_PENDING 4 /* long busy */ +#define DASD_STOPPED_DC_WAIT 8 /* disconnected, wait */ +#define DASD_STOPPED_SU 16 /* summary unit check handling */ +#define DASD_STOPPED_PM 32 /* pm state transition */ +#define DASD_UNRESUMED_PM 64 /* pm resume failed state */ + +/* per device flags */ +#define DASD_FLAG_OFFLINE 3 /* device is in offline processing */ +#define DASD_FLAG_EER_SNSS 4 /* A SNSS is required */ +#define DASD_FLAG_EER_IN_USE 5 /* A SNSS request is running */ +#define DASD_FLAG_DEVICE_RO 6 /* The device itself is read-only. Don't + * confuse this with the user specified + * read-only feature. + */ +#define DASD_FLAG_IS_RESERVED 7 /* The device is reserved */ +#define DASD_FLAG_LOCK_STOLEN 8 /* The device lock was stolen */ +#define DASD_FLAG_SUSPENDED 9 /* The device was suspended */ +#define DASD_FLAG_SAFE_OFFLINE 10 /* safe offline processing requested*/ +#define DASD_FLAG_SAFE_OFFLINE_RUNNING 11 /* safe offline running */ +#define DASD_FLAG_ABORTALL 12 /* Abort all noretry requests */ + +#define DASD_SLEEPON_START_TAG ((void *) 1) +#define DASD_SLEEPON_END_TAG ((void *) 2) + +void dasd_put_device_wake(struct dasd_device *); + +/* + * Reference count inliners + */ +static inline void +dasd_get_device(struct dasd_device *device) +{ + atomic_inc(&device->ref_count); +} + +static inline void +dasd_put_device(struct dasd_device *device) +{ + if (atomic_dec_return(&device->ref_count) == 0) + dasd_put_device_wake(device); +} + +/* + * The static memory in ccw_mem and erp_mem is managed by a sorted + * list of free memory chunks. + */ +struct dasd_mchunk +{ + struct list_head list; + unsigned long size; +} __attribute__ ((aligned(8))); + +static inline void +dasd_init_chunklist(struct list_head *chunk_list, void *mem, + unsigned long size) +{ + struct dasd_mchunk *chunk; + + INIT_LIST_HEAD(chunk_list); + chunk = (struct dasd_mchunk *) mem; + chunk->size = size - sizeof(struct dasd_mchunk); + list_add(&chunk->list, chunk_list); +} + +static inline void * +dasd_alloc_chunk(struct list_head *chunk_list, unsigned long size) +{ + struct dasd_mchunk *chunk, *tmp; + + size = (size + 7L) & -8L; + list_for_each_entry(chunk, chunk_list, list) { + if (chunk->size < size) + continue; + if (chunk->size > size + sizeof(struct dasd_mchunk)) { + char *endaddr = (char *) (chunk + 1) + chunk->size; + tmp = (struct dasd_mchunk *) (endaddr - size) - 1; + tmp->size = size; + chunk->size -= size + sizeof(struct dasd_mchunk); + chunk = tmp; + } else + list_del(&chunk->list); + return (void *) (chunk + 1); + } + return NULL; +} + +static inline void +dasd_free_chunk(struct list_head *chunk_list, void *mem) +{ + struct dasd_mchunk *chunk, *tmp; + struct list_head *p, *left; + + chunk = (struct dasd_mchunk *) + ((char *) mem - sizeof(struct dasd_mchunk)); + /* Find out the left neighbour in chunk_list. */ + left = chunk_list; + list_for_each(p, chunk_list) { + if (list_entry(p, struct dasd_mchunk, list) > chunk) + break; + left = p; + } + /* Try to merge with right neighbour = next element from left. */ + if (left->next != chunk_list) { + tmp = list_entry(left->next, struct dasd_mchunk, list); + if ((char *) (chunk + 1) + chunk->size == (char *) tmp) { + list_del(&tmp->list); + chunk->size += tmp->size + sizeof(struct dasd_mchunk); + } + } + /* Try to merge with left neighbour. */ + if (left != chunk_list) { + tmp = list_entry(left, struct dasd_mchunk, list); + if ((char *) (tmp + 1) + tmp->size == (char *) chunk) { + tmp->size += chunk->size + sizeof(struct dasd_mchunk); + return; + } + } + __list_add(&chunk->list, left, left->next); +} + +/* + * Check if bsize is in { 512, 1024, 2048, 4096 } + */ +static inline int +dasd_check_blocksize(int bsize) +{ + if (bsize < 512 || bsize > 4096 || !is_power_of_2(bsize)) + return -EMEDIUMTYPE; + return 0; +} + +/* externals in dasd.c */ +#define DASD_PROFILE_OFF 0 +#define DASD_PROFILE_ON 1 +#define DASD_PROFILE_GLOBAL_ONLY 2 + +extern debug_info_t *dasd_debug_area; +extern struct dasd_profile dasd_global_profile; +extern unsigned int dasd_global_profile_level; +extern const struct block_device_operations dasd_device_operations; + +extern struct kmem_cache *dasd_page_cache; + +struct dasd_ccw_req * +dasd_kmalloc_request(int , int, int, struct dasd_device *); +struct dasd_ccw_req * +dasd_smalloc_request(int , int, int, struct dasd_device *); +void dasd_kfree_request(struct dasd_ccw_req *, struct dasd_device *); +void dasd_sfree_request(struct dasd_ccw_req *, struct dasd_device *); +void dasd_wakeup_cb(struct dasd_ccw_req *, void *); + +static inline int +dasd_kmalloc_set_cda(struct ccw1 *ccw, void *cda, struct dasd_device *device) +{ + return set_normalized_cda(ccw, cda); +} + +struct dasd_device *dasd_alloc_device(void); +void dasd_free_device(struct dasd_device *); + +struct dasd_block *dasd_alloc_block(void); +void dasd_free_block(struct dasd_block *); + +enum blk_eh_timer_return dasd_times_out(struct request *req); + +void dasd_enable_device(struct dasd_device *); +void dasd_set_target_state(struct dasd_device *, int); +void dasd_kick_device(struct dasd_device *); +void dasd_restore_device(struct dasd_device *); +void dasd_reload_device(struct dasd_device *); + +void dasd_add_request_head(struct dasd_ccw_req *); +void dasd_add_request_tail(struct dasd_ccw_req *); +int dasd_start_IO(struct dasd_ccw_req *); +int dasd_term_IO(struct dasd_ccw_req *); +void dasd_schedule_device_bh(struct dasd_device *); +void dasd_schedule_block_bh(struct dasd_block *); +int dasd_sleep_on(struct dasd_ccw_req *); +int dasd_sleep_on_queue(struct list_head *); +int dasd_sleep_on_immediatly(struct dasd_ccw_req *); +int dasd_sleep_on_interruptible(struct dasd_ccw_req *); +void dasd_device_set_timer(struct dasd_device *, int); +void dasd_device_clear_timer(struct dasd_device *); +void dasd_block_set_timer(struct dasd_block *, int); +void dasd_block_clear_timer(struct dasd_block *); +int dasd_cancel_req(struct dasd_ccw_req *); +int dasd_flush_device_queue(struct dasd_device *); +int dasd_generic_probe (struct ccw_device *, struct dasd_discipline *); +void dasd_generic_remove (struct ccw_device *cdev); +int dasd_generic_set_online(struct ccw_device *, struct dasd_discipline *); +int dasd_generic_set_offline (struct ccw_device *cdev); +int dasd_generic_notify(struct ccw_device *, int); +int dasd_generic_last_path_gone(struct dasd_device *); +int dasd_generic_path_operational(struct dasd_device *); +void dasd_generic_shutdown(struct ccw_device *); + +void dasd_generic_handle_state_change(struct dasd_device *); +int dasd_generic_pm_freeze(struct ccw_device *); +int dasd_generic_restore_device(struct ccw_device *); +enum uc_todo dasd_generic_uc_handler(struct ccw_device *, struct irb *); +void dasd_generic_path_event(struct ccw_device *, int *); +int dasd_generic_verify_path(struct dasd_device *, __u8); + +int dasd_generic_read_dev_chars(struct dasd_device *, int, void *, int); +char *dasd_get_sense(struct irb *); + +void dasd_device_set_stop_bits(struct dasd_device *, int); +void dasd_device_remove_stop_bits(struct dasd_device *, int); + +int dasd_device_is_ro(struct dasd_device *); + +void dasd_profile_reset(struct dasd_profile *); +int dasd_profile_on(struct dasd_profile *); +void dasd_profile_off(struct dasd_profile *); +char *dasd_get_user_string(const char __user *, size_t); + +/* externals in dasd_devmap.c */ +extern int dasd_max_devindex; +extern int dasd_probeonly; +extern int dasd_autodetect; +extern int dasd_nopav; +extern int dasd_nofcx; + +int dasd_devmap_init(void); +void dasd_devmap_exit(void); + +struct dasd_device *dasd_create_device(struct ccw_device *); +void dasd_delete_device(struct dasd_device *); + +int dasd_get_feature(struct ccw_device *, int); +int dasd_set_feature(struct ccw_device *, int, int); + +int dasd_add_sysfs_files(struct ccw_device *); +void dasd_remove_sysfs_files(struct ccw_device *); + +struct dasd_device *dasd_device_from_cdev(struct ccw_device *); +struct dasd_device *dasd_device_from_cdev_locked(struct ccw_device *); +struct dasd_device *dasd_device_from_devindex(int); + +void dasd_add_link_to_gendisk(struct gendisk *, struct dasd_device *); +struct dasd_device *dasd_device_from_gendisk(struct gendisk *); + +int dasd_parse(void); +int dasd_busid_known(const char *); + +/* externals in dasd_gendisk.c */ +int dasd_gendisk_init(void); +void dasd_gendisk_exit(void); +int dasd_gendisk_alloc(struct dasd_block *); +void dasd_gendisk_free(struct dasd_block *); +int dasd_scan_partitions(struct dasd_block *); +void dasd_destroy_partitions(struct dasd_block *); + +/* externals in dasd_ioctl.c */ +int dasd_ioctl(struct block_device *, fmode_t, unsigned int, unsigned long); + +/* externals in dasd_proc.c */ +int dasd_proc_init(void); +void dasd_proc_exit(void); + +/* externals in dasd_erp.c */ +struct dasd_ccw_req *dasd_default_erp_action(struct dasd_ccw_req *); +struct dasd_ccw_req *dasd_default_erp_postaction(struct dasd_ccw_req *); +struct dasd_ccw_req *dasd_alloc_erp_request(char *, int, int, + struct dasd_device *); +void dasd_free_erp_request(struct dasd_ccw_req *, struct dasd_device *); +void dasd_log_sense(struct dasd_ccw_req *, struct irb *); +void dasd_log_sense_dbf(struct dasd_ccw_req *cqr, struct irb *irb); + +/* externals in dasd_3990_erp.c */ +struct dasd_ccw_req *dasd_3990_erp_action(struct dasd_ccw_req *); +void dasd_3990_erp_handle_sim(struct dasd_device *, char *); + +/* externals in dasd_eer.c */ +#ifdef CONFIG_DASD_EER +int dasd_eer_init(void); +void dasd_eer_exit(void); +int dasd_eer_enable(struct dasd_device *); +void dasd_eer_disable(struct dasd_device *); +void dasd_eer_write(struct dasd_device *, struct dasd_ccw_req *cqr, + unsigned int id); +void dasd_eer_snss(struct dasd_device *); + +static inline int dasd_eer_enabled(struct dasd_device *device) +{ + return device->eer_cqr != NULL; +} +#else +#define dasd_eer_init() (0) +#define dasd_eer_exit() do { } while (0) +#define dasd_eer_enable(d) (0) +#define dasd_eer_disable(d) do { } while (0) +#define dasd_eer_write(d,c,i) do { } while (0) +#define dasd_eer_snss(d) do { } while (0) +#define dasd_eer_enabled(d) (0) +#endif /* CONFIG_DASD_ERR */ + +#endif /* DASD_H */ diff --git a/kernel/drivers/s390/block/dasd_ioctl.c b/kernel/drivers/s390/block/dasd_ioctl.c new file mode 100644 index 000000000..02837d0ad --- /dev/null +++ b/kernel/drivers/s390/block/dasd_ioctl.c @@ -0,0 +1,579 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Copyright IBM Corp. 1999, 2001 + * + * i/o controls for the dasd driver. + */ + +#define KMSG_COMPONENT "dasd" + +#include <linux/interrupt.h> +#include <linux/compat.h> +#include <linux/major.h> +#include <linux/fs.h> +#include <linux/blkpg.h> +#include <linux/slab.h> +#include <asm/compat.h> +#include <asm/ccwdev.h> +#include <asm/schid.h> +#include <asm/cmb.h> +#include <asm/uaccess.h> + +/* This is ugly... */ +#define PRINTK_HEADER "dasd_ioctl:" + +#include "dasd_int.h" + + +static int +dasd_ioctl_api_version(void __user *argp) +{ + int ver = DASD_API_VERSION; + return put_user(ver, (int __user *)argp); +} + +/* + * Enable device. + * used by dasdfmt after BIODASDDISABLE to retrigger blocksize detection + */ +static int +dasd_ioctl_enable(struct block_device *bdev) +{ + struct dasd_device *base; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + + dasd_enable_device(base); + /* Formatting the dasd device can change the capacity. */ + mutex_lock(&bdev->bd_mutex); + i_size_write(bdev->bd_inode, + (loff_t)get_capacity(base->block->gdp) << 9); + mutex_unlock(&bdev->bd_mutex); + dasd_put_device(base); + return 0; +} + +/* + * Disable device. + * Used by dasdfmt. Disable I/O operations but allow ioctls. + */ +static int +dasd_ioctl_disable(struct block_device *bdev) +{ + struct dasd_device *base; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + /* + * Man this is sick. We don't do a real disable but only downgrade + * the device to DASD_STATE_BASIC. The reason is that dasdfmt uses + * BIODASDDISABLE to disable accesses to the device via the block + * device layer but it still wants to do i/o on the device by + * using the BIODASDFMT ioctl. Therefore the correct state for the + * device is DASD_STATE_BASIC that allows to do basic i/o. + */ + dasd_set_target_state(base, DASD_STATE_BASIC); + /* + * Set i_size to zero, since read, write, etc. check against this + * value. + */ + mutex_lock(&bdev->bd_mutex); + i_size_write(bdev->bd_inode, 0); + mutex_unlock(&bdev->bd_mutex); + dasd_put_device(base); + return 0; +} + +/* + * Quiesce device. + */ +static int dasd_ioctl_quiesce(struct dasd_block *block) +{ + unsigned long flags; + struct dasd_device *base; + + base = block->base; + if (!capable (CAP_SYS_ADMIN)) + return -EACCES; + + pr_info("%s: The DASD has been put in the quiesce " + "state\n", dev_name(&base->cdev->dev)); + spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags); + dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE); + spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), flags); + return 0; +} + + +/* + * Resume device. + */ +static int dasd_ioctl_resume(struct dasd_block *block) +{ + unsigned long flags; + struct dasd_device *base; + + base = block->base; + if (!capable (CAP_SYS_ADMIN)) + return -EACCES; + + pr_info("%s: I/O operations have been resumed " + "on the DASD\n", dev_name(&base->cdev->dev)); + spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags); + dasd_device_remove_stop_bits(base, DASD_STOPPED_QUIESCE); + spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), flags); + + dasd_schedule_block_bh(block); + return 0; +} + +/* + * Abort all failfast I/O on a device. + */ +static int dasd_ioctl_abortio(struct dasd_block *block) +{ + unsigned long flags; + struct dasd_device *base; + struct dasd_ccw_req *cqr, *n; + + base = block->base; + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + if (test_and_set_bit(DASD_FLAG_ABORTALL, &base->flags)) + return 0; + DBF_DEV_EVENT(DBF_NOTICE, base, "%s", "abortall flag set"); + + spin_lock_irqsave(&block->request_queue_lock, flags); + spin_lock(&block->queue_lock); + list_for_each_entry_safe(cqr, n, &block->ccw_queue, blocklist) { + if (test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) && + cqr->callback_data && + cqr->callback_data != DASD_SLEEPON_START_TAG && + cqr->callback_data != DASD_SLEEPON_END_TAG) { + spin_unlock(&block->queue_lock); + blk_abort_request(cqr->callback_data); + spin_lock(&block->queue_lock); + } + } + spin_unlock(&block->queue_lock); + spin_unlock_irqrestore(&block->request_queue_lock, flags); + + dasd_schedule_block_bh(block); + return 0; +} + +/* + * Allow I/O on a device + */ +static int dasd_ioctl_allowio(struct dasd_block *block) +{ + struct dasd_device *base; + + base = block->base; + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + + if (test_and_clear_bit(DASD_FLAG_ABORTALL, &base->flags)) + DBF_DEV_EVENT(DBF_NOTICE, base, "%s", "abortall flag unset"); + + return 0; +} + +/* + * performs formatting of _device_ according to _fdata_ + * Note: The discipline's format_function is assumed to deliver formatting + * commands to format multiple units of the device. In terms of the ECKD + * devices this means CCWs are generated to format multiple tracks. + */ +static int +dasd_format(struct dasd_block *block, struct format_data_t *fdata) +{ + struct dasd_device *base; + int enable_pav = 1; + int rc, retries; + int start, stop; + + base = block->base; + if (base->discipline->format_device == NULL) + return -EPERM; + + if (base->state != DASD_STATE_BASIC) { + pr_warn("%s: The DASD cannot be formatted while it is enabled\n", + dev_name(&base->cdev->dev)); + return -EBUSY; + } + + DBF_DEV_EVENT(DBF_NOTICE, base, + "formatting units %u to %u (%u B blocks) flags %u", + fdata->start_unit, + fdata->stop_unit, fdata->blksize, fdata->intensity); + + /* Since dasdfmt keeps the device open after it was disabled, + * there still exists an inode for this device. + * We must update i_blkbits, otherwise we might get errors when + * enabling the device later. + */ + if (fdata->start_unit == 0) { + struct block_device *bdev = bdget_disk(block->gdp, 0); + bdev->bd_inode->i_blkbits = blksize_bits(fdata->blksize); + bdput(bdev); + } + + retries = 255; + /* backup start- and endtrack for retries */ + start = fdata->start_unit; + stop = fdata->stop_unit; + do { + rc = base->discipline->format_device(base, fdata, enable_pav); + if (rc) { + if (rc == -EAGAIN) { + retries--; + /* disable PAV in case of errors */ + enable_pav = 0; + fdata->start_unit = start; + fdata->stop_unit = stop; + } else + return rc; + } else + /* success */ + break; + } while (retries); + + if (!retries) + return -EIO; + else + return 0; +} + +/* + * Format device. + */ +static int +dasd_ioctl_format(struct block_device *bdev, void __user *argp) +{ + struct dasd_device *base; + struct format_data_t fdata; + int rc; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + if (!argp) + return -EINVAL; + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + if (base->features & DASD_FEATURE_READONLY || + test_bit(DASD_FLAG_DEVICE_RO, &base->flags)) { + dasd_put_device(base); + return -EROFS; + } + if (copy_from_user(&fdata, argp, sizeof(struct format_data_t))) { + dasd_put_device(base); + return -EFAULT; + } + if (bdev != bdev->bd_contains) { + pr_warning("%s: The specified DASD is a partition and cannot " + "be formatted\n", + dev_name(&base->cdev->dev)); + dasd_put_device(base); + return -EINVAL; + } + rc = dasd_format(base->block, &fdata); + dasd_put_device(base); + return rc; +} + +#ifdef CONFIG_DASD_PROFILE +/* + * Reset device profile information + */ +static int dasd_ioctl_reset_profile(struct dasd_block *block) +{ + dasd_profile_reset(&block->profile); + return 0; +} + +/* + * Return device profile information + */ +static int dasd_ioctl_read_profile(struct dasd_block *block, void __user *argp) +{ + struct dasd_profile_info_t *data; + int rc = 0; + + data = kmalloc(sizeof(*data), GFP_KERNEL); + if (!data) + return -ENOMEM; + + spin_lock_bh(&block->profile.lock); + if (block->profile.data) { + data->dasd_io_reqs = block->profile.data->dasd_io_reqs; + data->dasd_io_sects = block->profile.data->dasd_io_sects; + memcpy(data->dasd_io_secs, block->profile.data->dasd_io_secs, + sizeof(data->dasd_io_secs)); + memcpy(data->dasd_io_times, block->profile.data->dasd_io_times, + sizeof(data->dasd_io_times)); + memcpy(data->dasd_io_timps, block->profile.data->dasd_io_timps, + sizeof(data->dasd_io_timps)); + memcpy(data->dasd_io_time1, block->profile.data->dasd_io_time1, + sizeof(data->dasd_io_time1)); + memcpy(data->dasd_io_time2, block->profile.data->dasd_io_time2, + sizeof(data->dasd_io_time2)); + memcpy(data->dasd_io_time2ps, + block->profile.data->dasd_io_time2ps, + sizeof(data->dasd_io_time2ps)); + memcpy(data->dasd_io_time3, block->profile.data->dasd_io_time3, + sizeof(data->dasd_io_time3)); + memcpy(data->dasd_io_nr_req, + block->profile.data->dasd_io_nr_req, + sizeof(data->dasd_io_nr_req)); + spin_unlock_bh(&block->profile.lock); + } else { + spin_unlock_bh(&block->profile.lock); + rc = -EIO; + goto out; + } + if (copy_to_user(argp, data, sizeof(*data))) + rc = -EFAULT; +out: + kfree(data); + return rc; +} +#else +static int dasd_ioctl_reset_profile(struct dasd_block *block) +{ + return -ENOTTY; +} + +static int dasd_ioctl_read_profile(struct dasd_block *block, void __user *argp) +{ + return -ENOTTY; +} +#endif + +/* + * Return dasd information. Used for BIODASDINFO and BIODASDINFO2. + */ +static int dasd_ioctl_information(struct dasd_block *block, + unsigned int cmd, void __user *argp) +{ + struct dasd_information2_t *dasd_info; + struct subchannel_id sch_id; + struct ccw_dev_id dev_id; + struct dasd_device *base; + struct ccw_device *cdev; + unsigned long flags; + int rc; + + base = block->base; + if (!base->discipline || !base->discipline->fill_info) + return -EINVAL; + + dasd_info = kzalloc(sizeof(struct dasd_information2_t), GFP_KERNEL); + if (dasd_info == NULL) + return -ENOMEM; + + rc = base->discipline->fill_info(base, dasd_info); + if (rc) { + kfree(dasd_info); + return rc; + } + + cdev = base->cdev; + ccw_device_get_id(cdev, &dev_id); + ccw_device_get_schid(cdev, &sch_id); + + dasd_info->devno = dev_id.devno; + dasd_info->schid = sch_id.sch_no; + dasd_info->cu_type = cdev->id.cu_type; + dasd_info->cu_model = cdev->id.cu_model; + dasd_info->dev_type = cdev->id.dev_type; + dasd_info->dev_model = cdev->id.dev_model; + dasd_info->status = base->state; + /* + * The open_count is increased for every opener, that includes + * the blkdev_get in dasd_scan_partitions. + * This must be hidden from user-space. + */ + dasd_info->open_count = atomic_read(&block->open_count); + if (!block->bdev) + dasd_info->open_count++; + + /* + * check if device is really formatted + * LDL / CDL was returned by 'fill_info' + */ + if ((base->state < DASD_STATE_READY) || + (dasd_check_blocksize(block->bp_block))) + dasd_info->format = DASD_FORMAT_NONE; + + dasd_info->features |= + ((base->features & DASD_FEATURE_READONLY) != 0); + + memcpy(dasd_info->type, base->discipline->name, 4); + + if (block->request_queue->request_fn) { + struct list_head *l; +#ifdef DASD_EXTENDED_PROFILING + { + struct list_head *l; + spin_lock_irqsave(&block->lock, flags); + list_for_each(l, &block->request_queue->queue_head) + dasd_info->req_queue_len++; + spin_unlock_irqrestore(&block->lock, flags); + } +#endif /* DASD_EXTENDED_PROFILING */ + spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags); + list_for_each(l, &base->ccw_queue) + dasd_info->chanq_len++; + spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), + flags); + } + + rc = 0; + if (copy_to_user(argp, dasd_info, + ((cmd == (unsigned int) BIODASDINFO2) ? + sizeof(struct dasd_information2_t) : + sizeof(struct dasd_information_t)))) + rc = -EFAULT; + kfree(dasd_info); + return rc; +} + +/* + * Set read only + */ +static int +dasd_ioctl_set_ro(struct block_device *bdev, void __user *argp) +{ + struct dasd_device *base; + int intval, rc; + + if (!capable(CAP_SYS_ADMIN)) + return -EACCES; + if (bdev != bdev->bd_contains) + // ro setting is not allowed for partitions + return -EINVAL; + if (get_user(intval, (int __user *)argp)) + return -EFAULT; + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + if (!intval && test_bit(DASD_FLAG_DEVICE_RO, &base->flags)) { + dasd_put_device(base); + return -EROFS; + } + set_disk_ro(bdev->bd_disk, intval); + rc = dasd_set_feature(base->cdev, DASD_FEATURE_READONLY, intval); + dasd_put_device(base); + return rc; +} + +static int dasd_ioctl_readall_cmb(struct dasd_block *block, unsigned int cmd, + struct cmbdata __user *argp) +{ + size_t size = _IOC_SIZE(cmd); + struct cmbdata data; + int ret; + + ret = cmf_readall(block->base->cdev, &data); + if (!ret && copy_to_user(argp, &data, min(size, sizeof(*argp)))) + return -EFAULT; + return ret; +} + +int dasd_ioctl(struct block_device *bdev, fmode_t mode, + unsigned int cmd, unsigned long arg) +{ + struct dasd_block *block; + struct dasd_device *base; + void __user *argp; + int rc; + + if (is_compat_task()) + argp = compat_ptr(arg); + else + argp = (void __user *)arg; + + if ((_IOC_DIR(cmd) != _IOC_NONE) && !arg) { + PRINT_DEBUG("empty data ptr"); + return -EINVAL; + } + + base = dasd_device_from_gendisk(bdev->bd_disk); + if (!base) + return -ENODEV; + block = base->block; + rc = 0; + switch (cmd) { + case BIODASDDISABLE: + rc = dasd_ioctl_disable(bdev); + break; + case BIODASDENABLE: + rc = dasd_ioctl_enable(bdev); + break; + case BIODASDQUIESCE: + rc = dasd_ioctl_quiesce(block); + break; + case BIODASDRESUME: + rc = dasd_ioctl_resume(block); + break; + case BIODASDABORTIO: + rc = dasd_ioctl_abortio(block); + break; + case BIODASDALLOWIO: + rc = dasd_ioctl_allowio(block); + break; + case BIODASDFMT: + rc = dasd_ioctl_format(bdev, argp); + break; + case BIODASDINFO: + rc = dasd_ioctl_information(block, cmd, argp); + break; + case BIODASDINFO2: + rc = dasd_ioctl_information(block, cmd, argp); + break; + case BIODASDPRRD: + rc = dasd_ioctl_read_profile(block, argp); + break; + case BIODASDPRRST: + rc = dasd_ioctl_reset_profile(block); + break; + case BLKROSET: + rc = dasd_ioctl_set_ro(bdev, argp); + break; + case DASDAPIVER: + rc = dasd_ioctl_api_version(argp); + break; + case BIODASDCMFENABLE: + rc = enable_cmf(base->cdev); + break; + case BIODASDCMFDISABLE: + rc = disable_cmf(base->cdev); + break; + case BIODASDREADALLCMB: + rc = dasd_ioctl_readall_cmb(block, cmd, argp); + break; + default: + /* if the discipline has an ioctl method try it. */ + rc = -ENOTTY; + if (base->discipline->ioctl) + rc = base->discipline->ioctl(block, cmd, argp); + } + dasd_put_device(base); + return rc; +} diff --git a/kernel/drivers/s390/block/dasd_proc.c b/kernel/drivers/s390/block/dasd_proc.c new file mode 100644 index 000000000..aa7bb2d1d --- /dev/null +++ b/kernel/drivers/s390/block/dasd_proc.c @@ -0,0 +1,383 @@ +/* + * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com> + * Horst Hummel <Horst.Hummel@de.ibm.com> + * Carsten Otte <Cotte@de.ibm.com> + * Martin Schwidefsky <schwidefsky@de.ibm.com> + * Bugreports.to..: <Linux390@de.ibm.com> + * Coypright IBM Corp. 1999, 2002 + * + * /proc interface for the dasd driver. + * + */ + +#define KMSG_COMPONENT "dasd" + +#include <linux/ctype.h> +#include <linux/slab.h> +#include <linux/string.h> +#include <linux/seq_file.h> +#include <linux/vmalloc.h> +#include <linux/proc_fs.h> + +#include <asm/debug.h> +#include <asm/uaccess.h> + +/* This is ugly... */ +#define PRINTK_HEADER "dasd_proc:" + +#include "dasd_int.h" + +static struct proc_dir_entry *dasd_proc_root_entry = NULL; +static struct proc_dir_entry *dasd_devices_entry = NULL; +static struct proc_dir_entry *dasd_statistics_entry = NULL; + +static int +dasd_devices_show(struct seq_file *m, void *v) +{ + struct dasd_device *device; + struct dasd_block *block; + char *substr; + + device = dasd_device_from_devindex((unsigned long) v - 1); + if (IS_ERR(device)) + return 0; + if (device->block) + block = device->block; + else { + dasd_put_device(device); + return 0; + } + /* Print device number. */ + seq_printf(m, "%s", dev_name(&device->cdev->dev)); + /* Print discipline string. */ + if (device->discipline != NULL) + seq_printf(m, "(%s)", device->discipline->name); + else + seq_printf(m, "(none)"); + /* Print kdev. */ + if (block->gdp) + seq_printf(m, " at (%3d:%6d)", + MAJOR(disk_devt(block->gdp)), + MINOR(disk_devt(block->gdp))); + else + seq_printf(m, " at (???:??????)"); + /* Print device name. */ + if (block->gdp) + seq_printf(m, " is %-8s", block->gdp->disk_name); + else + seq_printf(m, " is ????????"); + /* Print devices features. */ + substr = (device->features & DASD_FEATURE_READONLY) ? "(ro)" : " "; + seq_printf(m, "%4s: ", substr); + /* Print device status information. */ + switch (device->state) { + case DASD_STATE_NEW: + seq_printf(m, "new"); + break; + case DASD_STATE_KNOWN: + seq_printf(m, "detected"); + break; + case DASD_STATE_BASIC: + seq_printf(m, "basic"); + break; + case DASD_STATE_UNFMT: + seq_printf(m, "unformatted"); + break; + case DASD_STATE_READY: + case DASD_STATE_ONLINE: + seq_printf(m, "active "); + if (dasd_check_blocksize(block->bp_block)) + seq_printf(m, "n/f "); + else + seq_printf(m, + "at blocksize: %d, %lld blocks, %lld MB", + block->bp_block, block->blocks, + ((block->bp_block >> 9) * + block->blocks) >> 11); + break; + default: + seq_printf(m, "no stat"); + break; + } + dasd_put_device(device); + if (dasd_probeonly) + seq_printf(m, "(probeonly)"); + seq_printf(m, "\n"); + return 0; +} + +static void *dasd_devices_start(struct seq_file *m, loff_t *pos) +{ + if (*pos >= dasd_max_devindex) + return NULL; + return (void *)((unsigned long) *pos + 1); +} + +static void *dasd_devices_next(struct seq_file *m, void *v, loff_t *pos) +{ + ++*pos; + return dasd_devices_start(m, pos); +} + +static void dasd_devices_stop(struct seq_file *m, void *v) +{ +} + +static const struct seq_operations dasd_devices_seq_ops = { + .start = dasd_devices_start, + .next = dasd_devices_next, + .stop = dasd_devices_stop, + .show = dasd_devices_show, +}; + +static int dasd_devices_open(struct inode *inode, struct file *file) +{ + return seq_open(file, &dasd_devices_seq_ops); +} + +static const struct file_operations dasd_devices_file_ops = { + .owner = THIS_MODULE, + .open = dasd_devices_open, + .read = seq_read, + .llseek = seq_lseek, + .release = seq_release, +}; + +#ifdef CONFIG_DASD_PROFILE +static int dasd_stats_all_block_on(void) +{ + int i, rc; + struct dasd_device *device; + + rc = 0; + for (i = 0; i < dasd_max_devindex; ++i) { + device = dasd_device_from_devindex(i); + if (IS_ERR(device)) + continue; + if (device->block) + rc = dasd_profile_on(&device->block->profile); + dasd_put_device(device); + if (rc) + return rc; + } + return 0; +} + +static void dasd_stats_all_block_off(void) +{ + int i; + struct dasd_device *device; + + for (i = 0; i < dasd_max_devindex; ++i) { + device = dasd_device_from_devindex(i); + if (IS_ERR(device)) + continue; + if (device->block) + dasd_profile_off(&device->block->profile); + dasd_put_device(device); + } +} + +static void dasd_stats_all_block_reset(void) +{ + int i; + struct dasd_device *device; + + for (i = 0; i < dasd_max_devindex; ++i) { + device = dasd_device_from_devindex(i); + if (IS_ERR(device)) + continue; + if (device->block) + dasd_profile_reset(&device->block->profile); + dasd_put_device(device); + } +} + +static void dasd_statistics_array(struct seq_file *m, unsigned int *array, int factor) +{ + int i; + + for (i = 0; i < 32; i++) { + seq_printf(m, "%7d ", array[i] / factor); + if (i == 15) + seq_putc(m, '\n'); + } + seq_putc(m, '\n'); +} +#endif /* CONFIG_DASD_PROFILE */ + +static int dasd_stats_proc_show(struct seq_file *m, void *v) +{ +#ifdef CONFIG_DASD_PROFILE + struct dasd_profile_info *prof; + int factor; + + spin_lock_bh(&dasd_global_profile.lock); + prof = dasd_global_profile.data; + if (!prof) { + spin_unlock_bh(&dasd_global_profile.lock); + seq_printf(m, "Statistics are off - they might be " + "switched on using 'echo set on > " + "/proc/dasd/statistics'\n"); + return 0; + } + + /* prevent counter 'overflow' on output */ + for (factor = 1; (prof->dasd_io_reqs / factor) > 9999999; + factor *= 10); + + seq_printf(m, "%d dasd I/O requests\n", prof->dasd_io_reqs); + seq_printf(m, "with %u sectors(512B each)\n", + prof->dasd_io_sects); + seq_printf(m, "Scale Factor is %d\n", factor); + seq_printf(m, + " __<4 ___8 __16 __32 __64 _128 " + " _256 _512 __1k __2k __4k __8k " + " _16k _32k _64k 128k\n"); + seq_printf(m, + " _256 _512 __1M __2M __4M __8M " + " _16M _32M _64M 128M 256M 512M " + " __1G __2G __4G " " _>4G\n"); + + seq_printf(m, "Histogram of sizes (512B secs)\n"); + dasd_statistics_array(m, prof->dasd_io_secs, factor); + seq_printf(m, "Histogram of I/O times (microseconds)\n"); + dasd_statistics_array(m, prof->dasd_io_times, factor); + seq_printf(m, "Histogram of I/O times per sector\n"); + dasd_statistics_array(m, prof->dasd_io_timps, factor); + seq_printf(m, "Histogram of I/O time till ssch\n"); + dasd_statistics_array(m, prof->dasd_io_time1, factor); + seq_printf(m, "Histogram of I/O time between ssch and irq\n"); + dasd_statistics_array(m, prof->dasd_io_time2, factor); + seq_printf(m, "Histogram of I/O time between ssch " + "and irq per sector\n"); + dasd_statistics_array(m, prof->dasd_io_time2ps, factor); + seq_printf(m, "Histogram of I/O time between irq and end\n"); + dasd_statistics_array(m, prof->dasd_io_time3, factor); + seq_printf(m, "# of req in chanq at enqueuing (1..32) \n"); + dasd_statistics_array(m, prof->dasd_io_nr_req, factor); + spin_unlock_bh(&dasd_global_profile.lock); +#else + seq_printf(m, "Statistics are not activated in this kernel\n"); +#endif + return 0; +} + +static int dasd_stats_proc_open(struct inode *inode, struct file *file) +{ + return single_open(file, dasd_stats_proc_show, NULL); +} + +static ssize_t dasd_stats_proc_write(struct file *file, + const char __user *user_buf, size_t user_len, loff_t *pos) +{ +#ifdef CONFIG_DASD_PROFILE + char *buffer, *str; + int rc; + + if (user_len > 65536) + user_len = 65536; + buffer = dasd_get_user_string(user_buf, user_len); + if (IS_ERR(buffer)) + return PTR_ERR(buffer); + + /* check for valid verbs */ + str = skip_spaces(buffer); + if (strncmp(str, "set", 3) == 0 && isspace(str[3])) { + /* 'set xxx' was given */ + str = skip_spaces(str + 4); + if (strcmp(str, "on") == 0) { + /* switch on statistics profiling */ + rc = dasd_stats_all_block_on(); + if (rc) { + dasd_stats_all_block_off(); + goto out_error; + } + rc = dasd_profile_on(&dasd_global_profile); + if (rc) { + dasd_stats_all_block_off(); + goto out_error; + } + dasd_profile_reset(&dasd_global_profile); + dasd_global_profile_level = DASD_PROFILE_ON; + pr_info("The statistics feature has been switched " + "on\n"); + } else if (strcmp(str, "off") == 0) { + /* switch off statistics profiling */ + dasd_global_profile_level = DASD_PROFILE_OFF; + dasd_profile_off(&dasd_global_profile); + dasd_stats_all_block_off(); + pr_info("The statistics feature has been switched " + "off\n"); + } else + goto out_parse_error; + } else if (strncmp(str, "reset", 5) == 0) { + /* reset the statistics */ + dasd_profile_reset(&dasd_global_profile); + dasd_stats_all_block_reset(); + pr_info("The statistics have been reset\n"); + } else + goto out_parse_error; + vfree(buffer); + return user_len; +out_parse_error: + rc = -EINVAL; + pr_warning("%s is not a supported value for /proc/dasd/statistics\n", + str); +out_error: + vfree(buffer); + return rc; +#else + pr_warning("/proc/dasd/statistics: is not activated in this kernel\n"); + return user_len; +#endif /* CONFIG_DASD_PROFILE */ +} + +static const struct file_operations dasd_stats_proc_fops = { + .owner = THIS_MODULE, + .open = dasd_stats_proc_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, + .write = dasd_stats_proc_write, +}; + +/* + * Create dasd proc-fs entries. + * In case creation failed, cleanup and return -ENOENT. + */ +int +dasd_proc_init(void) +{ + dasd_proc_root_entry = proc_mkdir("dasd", NULL); + if (!dasd_proc_root_entry) + goto out_nodasd; + dasd_devices_entry = proc_create("devices", + S_IFREG | S_IRUGO | S_IWUSR, + dasd_proc_root_entry, + &dasd_devices_file_ops); + if (!dasd_devices_entry) + goto out_nodevices; + dasd_statistics_entry = proc_create("statistics", + S_IFREG | S_IRUGO | S_IWUSR, + dasd_proc_root_entry, + &dasd_stats_proc_fops); + if (!dasd_statistics_entry) + goto out_nostatistics; + return 0; + + out_nostatistics: + remove_proc_entry("devices", dasd_proc_root_entry); + out_nodevices: + remove_proc_entry("dasd", NULL); + out_nodasd: + return -ENOENT; +} + +void +dasd_proc_exit(void) +{ + remove_proc_entry("devices", dasd_proc_root_entry); + remove_proc_entry("statistics", dasd_proc_root_entry); + remove_proc_entry("dasd", NULL); +} diff --git a/kernel/drivers/s390/block/dcssblk.c b/kernel/drivers/s390/block/dcssblk.c new file mode 100644 index 000000000..da212813f --- /dev/null +++ b/kernel/drivers/s390/block/dcssblk.c @@ -0,0 +1,1091 @@ +/* + * dcssblk.c -- the S/390 block driver for dcss memory + * + * Authors: Carsten Otte, Stefan Weinhuber, Gerald Schaefer + */ + +#define KMSG_COMPONENT "dcssblk" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include <linux/module.h> +#include <linux/moduleparam.h> +#include <linux/ctype.h> +#include <linux/errno.h> +#include <linux/init.h> +#include <linux/slab.h> +#include <linux/blkdev.h> +#include <linux/completion.h> +#include <linux/interrupt.h> +#include <linux/platform_device.h> +#include <asm/extmem.h> +#include <asm/io.h> + +#define DCSSBLK_NAME "dcssblk" +#define DCSSBLK_MINORS_PER_DISK 1 +#define DCSSBLK_PARM_LEN 400 +#define DCSS_BUS_ID_SIZE 20 + +static int dcssblk_open(struct block_device *bdev, fmode_t mode); +static void dcssblk_release(struct gendisk *disk, fmode_t mode); +static void dcssblk_make_request(struct request_queue *q, struct bio *bio); +static long dcssblk_direct_access(struct block_device *bdev, sector_t secnum, + void **kaddr, unsigned long *pfn, long size); + +static char dcssblk_segments[DCSSBLK_PARM_LEN] = "\0"; + +static int dcssblk_major; +static const struct block_device_operations dcssblk_devops = { + .owner = THIS_MODULE, + .open = dcssblk_open, + .release = dcssblk_release, + .direct_access = dcssblk_direct_access, +}; + +struct dcssblk_dev_info { + struct list_head lh; + struct device dev; + char segment_name[DCSS_BUS_ID_SIZE]; + atomic_t use_count; + struct gendisk *gd; + unsigned long start; + unsigned long end; + int segment_type; + unsigned char save_pending; + unsigned char is_shared; + struct request_queue *dcssblk_queue; + int num_of_segments; + struct list_head seg_list; +}; + +struct segment_info { + struct list_head lh; + char segment_name[DCSS_BUS_ID_SIZE]; + unsigned long start; + unsigned long end; + int segment_type; +}; + +static ssize_t dcssblk_add_store(struct device * dev, struct device_attribute *attr, const char * buf, + size_t count); +static ssize_t dcssblk_remove_store(struct device * dev, struct device_attribute *attr, const char * buf, + size_t count); + +static DEVICE_ATTR(add, S_IWUSR, NULL, dcssblk_add_store); +static DEVICE_ATTR(remove, S_IWUSR, NULL, dcssblk_remove_store); + +static struct device *dcssblk_root_dev; + +static LIST_HEAD(dcssblk_devices); +static struct rw_semaphore dcssblk_devices_sem; + +/* + * release function for segment device. + */ +static void +dcssblk_release_segment(struct device *dev) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry, *temp; + + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + list_for_each_entry_safe(entry, temp, &dev_info->seg_list, lh) { + list_del(&entry->lh); + kfree(entry); + } + kfree(dev_info); + module_put(THIS_MODULE); +} + +/* + * get a minor number. needs to be called with + * down_write(&dcssblk_devices_sem) and the + * device needs to be enqueued before the semaphore is + * freed. + */ +static int +dcssblk_assign_free_minor(struct dcssblk_dev_info *dev_info) +{ + int minor, found; + struct dcssblk_dev_info *entry; + + if (dev_info == NULL) + return -EINVAL; + for (minor = 0; minor < (1<<MINORBITS); minor++) { + found = 0; + // test if minor available + list_for_each_entry(entry, &dcssblk_devices, lh) + if (minor == entry->gd->first_minor) + found++; + if (!found) break; // got unused minor + } + if (found) + return -EBUSY; + dev_info->gd->first_minor = minor; + return 0; +} + +/* + * get the struct dcssblk_dev_info from dcssblk_devices + * for the given name. + * down_read(&dcssblk_devices_sem) must be held. + */ +static struct dcssblk_dev_info * +dcssblk_get_device_by_name(char *name) +{ + struct dcssblk_dev_info *entry; + + list_for_each_entry(entry, &dcssblk_devices, lh) { + if (!strcmp(name, entry->segment_name)) { + return entry; + } + } + return NULL; +} + +/* + * get the struct segment_info from seg_list + * for the given name. + * down_read(&dcssblk_devices_sem) must be held. + */ +static struct segment_info * +dcssblk_get_segment_by_name(char *name) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + + list_for_each_entry(dev_info, &dcssblk_devices, lh) { + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (!strcmp(name, entry->segment_name)) + return entry; + } + } + return NULL; +} + +/* + * get the highest address of the multi-segment block. + */ +static unsigned long +dcssblk_find_highest_addr(struct dcssblk_dev_info *dev_info) +{ + unsigned long highest_addr; + struct segment_info *entry; + + highest_addr = 0; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (highest_addr < entry->end) + highest_addr = entry->end; + } + return highest_addr; +} + +/* + * get the lowest address of the multi-segment block. + */ +static unsigned long +dcssblk_find_lowest_addr(struct dcssblk_dev_info *dev_info) +{ + int set_first; + unsigned long lowest_addr; + struct segment_info *entry; + + set_first = 0; + lowest_addr = 0; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (set_first == 0) { + lowest_addr = entry->start; + set_first = 1; + } else { + if (lowest_addr > entry->start) + lowest_addr = entry->start; + } + } + return lowest_addr; +} + +/* + * Check continuity of segments. + */ +static int +dcssblk_is_continuous(struct dcssblk_dev_info *dev_info) +{ + int i, j, rc; + struct segment_info *sort_list, *entry, temp; + + if (dev_info->num_of_segments <= 1) + return 0; + + sort_list = kzalloc( + sizeof(struct segment_info) * dev_info->num_of_segments, + GFP_KERNEL); + if (sort_list == NULL) + return -ENOMEM; + i = 0; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + memcpy(&sort_list[i], entry, sizeof(struct segment_info)); + i++; + } + + /* sort segments */ + for (i = 0; i < dev_info->num_of_segments; i++) + for (j = 0; j < dev_info->num_of_segments; j++) + if (sort_list[j].start > sort_list[i].start) { + memcpy(&temp, &sort_list[i], + sizeof(struct segment_info)); + memcpy(&sort_list[i], &sort_list[j], + sizeof(struct segment_info)); + memcpy(&sort_list[j], &temp, + sizeof(struct segment_info)); + } + + /* check continuity */ + for (i = 0; i < dev_info->num_of_segments - 1; i++) { + if ((sort_list[i].end + 1) != sort_list[i+1].start) { + pr_err("Adjacent DCSSs %s and %s are not " + "contiguous\n", sort_list[i].segment_name, + sort_list[i+1].segment_name); + rc = -EINVAL; + goto out; + } + /* EN and EW are allowed in a block device */ + if (sort_list[i].segment_type != sort_list[i+1].segment_type) { + if (!(sort_list[i].segment_type & SEGMENT_EXCLUSIVE) || + (sort_list[i].segment_type == SEG_TYPE_ER) || + !(sort_list[i+1].segment_type & + SEGMENT_EXCLUSIVE) || + (sort_list[i+1].segment_type == SEG_TYPE_ER)) { + pr_err("DCSS %s and DCSS %s have " + "incompatible types\n", + sort_list[i].segment_name, + sort_list[i+1].segment_name); + rc = -EINVAL; + goto out; + } + } + } + rc = 0; +out: + kfree(sort_list); + return rc; +} + +/* + * Load a segment + */ +static int +dcssblk_load_segment(char *name, struct segment_info **seg_info) +{ + int rc; + + /* already loaded? */ + down_read(&dcssblk_devices_sem); + *seg_info = dcssblk_get_segment_by_name(name); + up_read(&dcssblk_devices_sem); + if (*seg_info != NULL) + return -EEXIST; + + /* get a struct segment_info */ + *seg_info = kzalloc(sizeof(struct segment_info), GFP_KERNEL); + if (*seg_info == NULL) + return -ENOMEM; + + strcpy((*seg_info)->segment_name, name); + + /* load the segment */ + rc = segment_load(name, SEGMENT_SHARED, + &(*seg_info)->start, &(*seg_info)->end); + if (rc < 0) { + segment_warning(rc, (*seg_info)->segment_name); + kfree(*seg_info); + } else { + INIT_LIST_HEAD(&(*seg_info)->lh); + (*seg_info)->segment_type = rc; + } + return rc; +} + +/* + * device attribute for switching shared/nonshared (exclusive) + * operation (show + store) + */ +static ssize_t +dcssblk_shared_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dcssblk_dev_info *dev_info; + + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + return sprintf(buf, dev_info->is_shared ? "1\n" : "0\n"); +} + +static ssize_t +dcssblk_shared_store(struct device *dev, struct device_attribute *attr, const char *inbuf, size_t count) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry, *temp; + int rc; + + if ((count > 1) && (inbuf[1] != '\n') && (inbuf[1] != '\0')) + return -EINVAL; + down_write(&dcssblk_devices_sem); + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + if (atomic_read(&dev_info->use_count)) { + rc = -EBUSY; + goto out; + } + if (inbuf[0] == '1') { + /* reload segments in shared mode */ + list_for_each_entry(entry, &dev_info->seg_list, lh) { + rc = segment_modify_shared(entry->segment_name, + SEGMENT_SHARED); + if (rc < 0) { + BUG_ON(rc == -EINVAL); + if (rc != -EAGAIN) + goto removeseg; + } + } + dev_info->is_shared = 1; + switch (dev_info->segment_type) { + case SEG_TYPE_SR: + case SEG_TYPE_ER: + case SEG_TYPE_SC: + set_disk_ro(dev_info->gd, 1); + } + } else if (inbuf[0] == '0') { + /* reload segments in exclusive mode */ + if (dev_info->segment_type == SEG_TYPE_SC) { + pr_err("DCSS %s is of type SC and cannot be " + "loaded as exclusive-writable\n", + dev_info->segment_name); + rc = -EINVAL; + goto out; + } + list_for_each_entry(entry, &dev_info->seg_list, lh) { + rc = segment_modify_shared(entry->segment_name, + SEGMENT_EXCLUSIVE); + if (rc < 0) { + BUG_ON(rc == -EINVAL); + if (rc != -EAGAIN) + goto removeseg; + } + } + dev_info->is_shared = 0; + set_disk_ro(dev_info->gd, 0); + } else { + rc = -EINVAL; + goto out; + } + rc = count; + goto out; + +removeseg: + pr_err("DCSS device %s is removed after a failed access mode " + "change\n", dev_info->segment_name); + temp = entry; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (entry != temp) + segment_unload(entry->segment_name); + } + list_del(&dev_info->lh); + + del_gendisk(dev_info->gd); + blk_cleanup_queue(dev_info->dcssblk_queue); + dev_info->gd->queue = NULL; + put_disk(dev_info->gd); + up_write(&dcssblk_devices_sem); + + if (device_remove_file_self(dev, attr)) { + device_unregister(dev); + put_device(dev); + } + return rc; +out: + up_write(&dcssblk_devices_sem); + return rc; +} +static DEVICE_ATTR(shared, S_IWUSR | S_IRUSR, dcssblk_shared_show, + dcssblk_shared_store); + +/* + * device attribute for save operation on current copy + * of the segment. If the segment is busy, saving will + * become pending until it gets released, which can be + * undone by storing a non-true value to this entry. + * (show + store) + */ +static ssize_t +dcssblk_save_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct dcssblk_dev_info *dev_info; + + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + return sprintf(buf, dev_info->save_pending ? "1\n" : "0\n"); +} + +static ssize_t +dcssblk_save_store(struct device *dev, struct device_attribute *attr, const char *inbuf, size_t count) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + + if ((count > 1) && (inbuf[1] != '\n') && (inbuf[1] != '\0')) + return -EINVAL; + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + + down_write(&dcssblk_devices_sem); + if (inbuf[0] == '1') { + if (atomic_read(&dev_info->use_count) == 0) { + // device is idle => we save immediately + pr_info("All DCSSs that map to device %s are " + "saved\n", dev_info->segment_name); + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (entry->segment_type == SEG_TYPE_EN || + entry->segment_type == SEG_TYPE_SN) + pr_warn("DCSS %s is of type SN or EN" + " and cannot be saved\n", + entry->segment_name); + else + segment_save(entry->segment_name); + } + } else { + // device is busy => we save it when it becomes + // idle in dcssblk_release + pr_info("Device %s is in use, its DCSSs will be " + "saved when it becomes idle\n", + dev_info->segment_name); + dev_info->save_pending = 1; + } + } else if (inbuf[0] == '0') { + if (dev_info->save_pending) { + // device is busy & the user wants to undo his save + // request + dev_info->save_pending = 0; + pr_info("A pending save request for device %s " + "has been canceled\n", + dev_info->segment_name); + } + } else { + up_write(&dcssblk_devices_sem); + return -EINVAL; + } + up_write(&dcssblk_devices_sem); + return count; +} +static DEVICE_ATTR(save, S_IWUSR | S_IRUSR, dcssblk_save_show, + dcssblk_save_store); + +/* + * device attribute for showing all segments in a device + */ +static ssize_t +dcssblk_seglist_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + int i; + + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + + down_read(&dcssblk_devices_sem); + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + i = 0; + buf[0] = '\0'; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + strcpy(&buf[i], entry->segment_name); + i += strlen(entry->segment_name); + buf[i] = '\n'; + i++; + } + up_read(&dcssblk_devices_sem); + return i; +} +static DEVICE_ATTR(seglist, S_IRUSR, dcssblk_seglist_show, NULL); + +static struct attribute *dcssblk_dev_attrs[] = { + &dev_attr_shared.attr, + &dev_attr_save.attr, + &dev_attr_seglist.attr, + NULL, +}; +static struct attribute_group dcssblk_dev_attr_group = { + .attrs = dcssblk_dev_attrs, +}; +static const struct attribute_group *dcssblk_dev_attr_groups[] = { + &dcssblk_dev_attr_group, + NULL, +}; + +/* + * device attribute for adding devices + */ +static ssize_t +dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) +{ + int rc, i, j, num_of_segments; + struct dcssblk_dev_info *dev_info; + struct segment_info *seg_info, *temp; + char *local_buf; + unsigned long seg_byte_size; + + dev_info = NULL; + seg_info = NULL; + if (dev != dcssblk_root_dev) { + rc = -EINVAL; + goto out_nobuf; + } + if ((count < 1) || (buf[0] == '\0') || (buf[0] == '\n')) { + rc = -ENAMETOOLONG; + goto out_nobuf; + } + + local_buf = kmalloc(count + 1, GFP_KERNEL); + if (local_buf == NULL) { + rc = -ENOMEM; + goto out_nobuf; + } + + /* + * parse input + */ + num_of_segments = 0; + for (i = 0; (i < count && (buf[i] != '\0') && (buf[i] != '\n')); i++) { + for (j = i; (buf[j] != ':') && + (buf[j] != '\0') && + (buf[j] != '\n') && + j < count; j++) { + local_buf[j-i] = toupper(buf[j]); + } + local_buf[j-i] = '\0'; + if (((j - i) == 0) || ((j - i) > 8)) { + rc = -ENAMETOOLONG; + goto seg_list_del; + } + + rc = dcssblk_load_segment(local_buf, &seg_info); + if (rc < 0) + goto seg_list_del; + /* + * get a struct dcssblk_dev_info + */ + if (num_of_segments == 0) { + dev_info = kzalloc(sizeof(struct dcssblk_dev_info), + GFP_KERNEL); + if (dev_info == NULL) { + rc = -ENOMEM; + goto out; + } + strcpy(dev_info->segment_name, local_buf); + dev_info->segment_type = seg_info->segment_type; + INIT_LIST_HEAD(&dev_info->seg_list); + } + list_add_tail(&seg_info->lh, &dev_info->seg_list); + num_of_segments++; + i = j; + + if ((buf[j] == '\0') || (buf[j] == '\n')) + break; + } + + /* no trailing colon at the end of the input */ + if ((i > 0) && (buf[i-1] == ':')) { + rc = -ENAMETOOLONG; + goto seg_list_del; + } + strlcpy(local_buf, buf, i + 1); + dev_info->num_of_segments = num_of_segments; + rc = dcssblk_is_continuous(dev_info); + if (rc < 0) + goto seg_list_del; + + dev_info->start = dcssblk_find_lowest_addr(dev_info); + dev_info->end = dcssblk_find_highest_addr(dev_info); + + dev_set_name(&dev_info->dev, "%s", dev_info->segment_name); + dev_info->dev.release = dcssblk_release_segment; + dev_info->dev.groups = dcssblk_dev_attr_groups; + INIT_LIST_HEAD(&dev_info->lh); + dev_info->gd = alloc_disk(DCSSBLK_MINORS_PER_DISK); + if (dev_info->gd == NULL) { + rc = -ENOMEM; + goto seg_list_del; + } + dev_info->gd->major = dcssblk_major; + dev_info->gd->fops = &dcssblk_devops; + dev_info->dcssblk_queue = blk_alloc_queue(GFP_KERNEL); + dev_info->gd->queue = dev_info->dcssblk_queue; + dev_info->gd->private_data = dev_info; + dev_info->gd->driverfs_dev = &dev_info->dev; + blk_queue_make_request(dev_info->dcssblk_queue, dcssblk_make_request); + blk_queue_logical_block_size(dev_info->dcssblk_queue, 4096); + + seg_byte_size = (dev_info->end - dev_info->start + 1); + set_capacity(dev_info->gd, seg_byte_size >> 9); // size in sectors + pr_info("Loaded %s with total size %lu bytes and capacity %lu " + "sectors\n", local_buf, seg_byte_size, seg_byte_size >> 9); + + dev_info->save_pending = 0; + dev_info->is_shared = 1; + dev_info->dev.parent = dcssblk_root_dev; + + /* + *get minor, add to list + */ + down_write(&dcssblk_devices_sem); + if (dcssblk_get_segment_by_name(local_buf)) { + rc = -EEXIST; + goto release_gd; + } + rc = dcssblk_assign_free_minor(dev_info); + if (rc) + goto release_gd; + sprintf(dev_info->gd->disk_name, "dcssblk%d", + dev_info->gd->first_minor); + list_add_tail(&dev_info->lh, &dcssblk_devices); + + if (!try_module_get(THIS_MODULE)) { + rc = -ENODEV; + goto dev_list_del; + } + /* + * register the device + */ + rc = device_register(&dev_info->dev); + if (rc) + goto put_dev; + + get_device(&dev_info->dev); + add_disk(dev_info->gd); + + switch (dev_info->segment_type) { + case SEG_TYPE_SR: + case SEG_TYPE_ER: + case SEG_TYPE_SC: + set_disk_ro(dev_info->gd,1); + break; + default: + set_disk_ro(dev_info->gd,0); + break; + } + up_write(&dcssblk_devices_sem); + rc = count; + goto out; + +put_dev: + list_del(&dev_info->lh); + blk_cleanup_queue(dev_info->dcssblk_queue); + dev_info->gd->queue = NULL; + put_disk(dev_info->gd); + list_for_each_entry(seg_info, &dev_info->seg_list, lh) { + segment_unload(seg_info->segment_name); + } + put_device(&dev_info->dev); + up_write(&dcssblk_devices_sem); + goto out; +dev_list_del: + list_del(&dev_info->lh); +release_gd: + blk_cleanup_queue(dev_info->dcssblk_queue); + dev_info->gd->queue = NULL; + put_disk(dev_info->gd); + up_write(&dcssblk_devices_sem); +seg_list_del: + if (dev_info == NULL) + goto out; + list_for_each_entry_safe(seg_info, temp, &dev_info->seg_list, lh) { + list_del(&seg_info->lh); + segment_unload(seg_info->segment_name); + kfree(seg_info); + } + kfree(dev_info); +out: + kfree(local_buf); +out_nobuf: + return rc; +} + +/* + * device attribute for removing devices + */ +static ssize_t +dcssblk_remove_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + int rc, i; + char *local_buf; + + if (dev != dcssblk_root_dev) { + return -EINVAL; + } + local_buf = kmalloc(count + 1, GFP_KERNEL); + if (local_buf == NULL) { + return -ENOMEM; + } + /* + * parse input + */ + for (i = 0; ((*(buf+i)!='\0') && (*(buf+i)!='\n') && i < count); i++) { + local_buf[i] = toupper(buf[i]); + } + local_buf[i] = '\0'; + if ((i == 0) || (i > 8)) { + rc = -ENAMETOOLONG; + goto out_buf; + } + + down_write(&dcssblk_devices_sem); + dev_info = dcssblk_get_device_by_name(local_buf); + if (dev_info == NULL) { + up_write(&dcssblk_devices_sem); + pr_warning("Device %s cannot be removed because it is not a " + "known device\n", local_buf); + rc = -ENODEV; + goto out_buf; + } + if (atomic_read(&dev_info->use_count) != 0) { + up_write(&dcssblk_devices_sem); + pr_warning("Device %s cannot be removed while it is in " + "use\n", local_buf); + rc = -EBUSY; + goto out_buf; + } + + list_del(&dev_info->lh); + del_gendisk(dev_info->gd); + blk_cleanup_queue(dev_info->dcssblk_queue); + dev_info->gd->queue = NULL; + put_disk(dev_info->gd); + device_unregister(&dev_info->dev); + + /* unload all related segments */ + list_for_each_entry(entry, &dev_info->seg_list, lh) + segment_unload(entry->segment_name); + + put_device(&dev_info->dev); + up_write(&dcssblk_devices_sem); + + rc = count; +out_buf: + kfree(local_buf); + return rc; +} + +static int +dcssblk_open(struct block_device *bdev, fmode_t mode) +{ + struct dcssblk_dev_info *dev_info; + int rc; + + dev_info = bdev->bd_disk->private_data; + if (NULL == dev_info) { + rc = -ENODEV; + goto out; + } + atomic_inc(&dev_info->use_count); + bdev->bd_block_size = 4096; + rc = 0; +out: + return rc; +} + +static void +dcssblk_release(struct gendisk *disk, fmode_t mode) +{ + struct dcssblk_dev_info *dev_info = disk->private_data; + struct segment_info *entry; + + if (!dev_info) { + WARN_ON(1); + return; + } + down_write(&dcssblk_devices_sem); + if (atomic_dec_and_test(&dev_info->use_count) + && (dev_info->save_pending)) { + pr_info("Device %s has become idle and is being saved " + "now\n", dev_info->segment_name); + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (entry->segment_type == SEG_TYPE_EN || + entry->segment_type == SEG_TYPE_SN) + pr_warn("DCSS %s is of type SN or EN and cannot" + " be saved\n", entry->segment_name); + else + segment_save(entry->segment_name); + } + dev_info->save_pending = 0; + } + up_write(&dcssblk_devices_sem); +} + +static void +dcssblk_make_request(struct request_queue *q, struct bio *bio) +{ + struct dcssblk_dev_info *dev_info; + struct bio_vec bvec; + struct bvec_iter iter; + unsigned long index; + unsigned long page_addr; + unsigned long source_addr; + unsigned long bytes_done; + + bytes_done = 0; + dev_info = bio->bi_bdev->bd_disk->private_data; + if (dev_info == NULL) + goto fail; + if ((bio->bi_iter.bi_sector & 7) != 0 || + (bio->bi_iter.bi_size & 4095) != 0) + /* Request is not page-aligned. */ + goto fail; + if (bio_end_sector(bio) > get_capacity(bio->bi_bdev->bd_disk)) { + /* Request beyond end of DCSS segment. */ + goto fail; + } + /* verify data transfer direction */ + if (dev_info->is_shared) { + switch (dev_info->segment_type) { + case SEG_TYPE_SR: + case SEG_TYPE_ER: + case SEG_TYPE_SC: + /* cannot write to these segments */ + if (bio_data_dir(bio) == WRITE) { + pr_warning("Writing to %s failed because it " + "is a read-only device\n", + dev_name(&dev_info->dev)); + goto fail; + } + } + } + + index = (bio->bi_iter.bi_sector >> 3); + bio_for_each_segment(bvec, bio, iter) { + page_addr = (unsigned long) + page_address(bvec.bv_page) + bvec.bv_offset; + source_addr = dev_info->start + (index<<12) + bytes_done; + if (unlikely((page_addr & 4095) != 0) || (bvec.bv_len & 4095) != 0) + // More paranoia. + goto fail; + if (bio_data_dir(bio) == READ) { + memcpy((void*)page_addr, (void*)source_addr, + bvec.bv_len); + } else { + memcpy((void*)source_addr, (void*)page_addr, + bvec.bv_len); + } + bytes_done += bvec.bv_len; + } + bio_endio(bio, 0); + return; +fail: + bio_io_error(bio); +} + +static long +dcssblk_direct_access (struct block_device *bdev, sector_t secnum, + void **kaddr, unsigned long *pfn, long size) +{ + struct dcssblk_dev_info *dev_info; + unsigned long offset, dev_sz; + + dev_info = bdev->bd_disk->private_data; + if (!dev_info) + return -ENODEV; + dev_sz = dev_info->end - dev_info->start; + offset = secnum * 512; + *kaddr = (void *) (dev_info->start + offset); + *pfn = virt_to_phys(*kaddr) >> PAGE_SHIFT; + + return dev_sz - offset; +} + +static void +dcssblk_check_params(void) +{ + int rc, i, j, k; + char buf[DCSSBLK_PARM_LEN + 1]; + struct dcssblk_dev_info *dev_info; + + for (i = 0; (i < DCSSBLK_PARM_LEN) && (dcssblk_segments[i] != '\0'); + i++) { + for (j = i; (dcssblk_segments[j] != ',') && + (dcssblk_segments[j] != '\0') && + (dcssblk_segments[j] != '(') && + (j < DCSSBLK_PARM_LEN); j++) + { + buf[j-i] = dcssblk_segments[j]; + } + buf[j-i] = '\0'; + rc = dcssblk_add_store(dcssblk_root_dev, NULL, buf, j-i); + if ((rc >= 0) && (dcssblk_segments[j] == '(')) { + for (k = 0; (buf[k] != ':') && (buf[k] != '\0'); k++) + buf[k] = toupper(buf[k]); + buf[k] = '\0'; + if (!strncmp(&dcssblk_segments[j], "(local)", 7)) { + down_read(&dcssblk_devices_sem); + dev_info = dcssblk_get_device_by_name(buf); + up_read(&dcssblk_devices_sem); + if (dev_info) + dcssblk_shared_store(&dev_info->dev, + NULL, "0\n", 2); + } + } + while ((dcssblk_segments[j] != ',') && + (dcssblk_segments[j] != '\0')) + { + j++; + } + if (dcssblk_segments[j] == '\0') + break; + i = j; + } +} + +/* + * Suspend / Resume + */ +static int dcssblk_freeze(struct device *dev) +{ + struct dcssblk_dev_info *dev_info; + int rc = 0; + + list_for_each_entry(dev_info, &dcssblk_devices, lh) { + switch (dev_info->segment_type) { + case SEG_TYPE_SR: + case SEG_TYPE_ER: + case SEG_TYPE_SC: + if (!dev_info->is_shared) + rc = -EINVAL; + break; + default: + rc = -EINVAL; + break; + } + if (rc) + break; + } + if (rc) + pr_err("Suspending the system failed because DCSS device %s " + "is writable\n", + dev_info->segment_name); + return rc; +} + +static int dcssblk_restore(struct device *dev) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + unsigned long start, end; + int rc = 0; + + list_for_each_entry(dev_info, &dcssblk_devices, lh) { + list_for_each_entry(entry, &dev_info->seg_list, lh) { + segment_unload(entry->segment_name); + rc = segment_load(entry->segment_name, SEGMENT_SHARED, + &start, &end); + if (rc < 0) { +// TODO in_use check ? + segment_warning(rc, entry->segment_name); + goto out_panic; + } + if (start != entry->start || end != entry->end) { + pr_err("The address range of DCSS %s changed " + "while the system was suspended\n", + entry->segment_name); + goto out_panic; + } + } + } + return 0; +out_panic: + panic("fatal dcssblk resume error\n"); +} + +static int dcssblk_thaw(struct device *dev) +{ + return 0; +} + +static const struct dev_pm_ops dcssblk_pm_ops = { + .freeze = dcssblk_freeze, + .thaw = dcssblk_thaw, + .restore = dcssblk_restore, +}; + +static struct platform_driver dcssblk_pdrv = { + .driver = { + .name = "dcssblk", + .pm = &dcssblk_pm_ops, + }, +}; + +static struct platform_device *dcssblk_pdev; + + +/* + * The init/exit functions. + */ +static void __exit +dcssblk_exit(void) +{ + platform_device_unregister(dcssblk_pdev); + platform_driver_unregister(&dcssblk_pdrv); + root_device_unregister(dcssblk_root_dev); + unregister_blkdev(dcssblk_major, DCSSBLK_NAME); +} + +static int __init +dcssblk_init(void) +{ + int rc; + + rc = platform_driver_register(&dcssblk_pdrv); + if (rc) + return rc; + + dcssblk_pdev = platform_device_register_simple("dcssblk", -1, NULL, + 0); + if (IS_ERR(dcssblk_pdev)) { + rc = PTR_ERR(dcssblk_pdev); + goto out_pdrv; + } + + dcssblk_root_dev = root_device_register("dcssblk"); + if (IS_ERR(dcssblk_root_dev)) { + rc = PTR_ERR(dcssblk_root_dev); + goto out_pdev; + } + rc = device_create_file(dcssblk_root_dev, &dev_attr_add); + if (rc) + goto out_root; + rc = device_create_file(dcssblk_root_dev, &dev_attr_remove); + if (rc) + goto out_root; + rc = register_blkdev(0, DCSSBLK_NAME); + if (rc < 0) + goto out_root; + dcssblk_major = rc; + init_rwsem(&dcssblk_devices_sem); + + dcssblk_check_params(); + return 0; + +out_root: + root_device_unregister(dcssblk_root_dev); +out_pdev: + platform_device_unregister(dcssblk_pdev); +out_pdrv: + platform_driver_unregister(&dcssblk_pdrv); + return rc; +} + +module_init(dcssblk_init); +module_exit(dcssblk_exit); + +module_param_string(segments, dcssblk_segments, DCSSBLK_PARM_LEN, 0444); +MODULE_PARM_DESC(segments, "Name of DCSS segment(s) to be loaded, " + "comma-separated list, names in each set separated " + "by commas are separated by colons, each set contains " + "names of contiguous segments and each name max. 8 chars.\n" + "Adding \"(local)\" to the end of each set equals echoing 0 " + "to /sys/devices/dcssblk/<device name>/shared after loading " + "the contiguous segments - \n" + "e.g. segments=\"mydcss1,mydcss2:mydcss3,mydcss4(local)\""); + +MODULE_LICENSE("GPL"); diff --git a/kernel/drivers/s390/block/scm_blk.c b/kernel/drivers/s390/block/scm_blk.c new file mode 100644 index 000000000..75d9896de --- /dev/null +++ b/kernel/drivers/s390/block/scm_blk.c @@ -0,0 +1,620 @@ +/* + * Block driver for s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott <sebott@linux.vnet.ibm.com> + */ + +#define KMSG_COMPONENT "scm_block" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include <linux/interrupt.h> +#include <linux/spinlock.h> +#include <linux/mempool.h> +#include <linux/module.h> +#include <linux/blkdev.h> +#include <linux/genhd.h> +#include <linux/slab.h> +#include <linux/list.h> +#include <asm/eadm.h> +#include "scm_blk.h" + +debug_info_t *scm_debug; +static int scm_major; +static mempool_t *aidaw_pool; +static DEFINE_SPINLOCK(list_lock); +static LIST_HEAD(inactive_requests); +static unsigned int nr_requests = 64; +static unsigned int nr_requests_per_io = 8; +static atomic_t nr_devices = ATOMIC_INIT(0); +module_param(nr_requests, uint, S_IRUGO); +MODULE_PARM_DESC(nr_requests, "Number of parallel requests."); + +module_param(nr_requests_per_io, uint, S_IRUGO); +MODULE_PARM_DESC(nr_requests_per_io, "Number of requests per IO."); + +MODULE_DESCRIPTION("Block driver for s390 storage class memory."); +MODULE_LICENSE("GPL"); +MODULE_ALIAS("scm:scmdev*"); + +static void __scm_free_rq(struct scm_request *scmrq) +{ + struct aob_rq_header *aobrq = to_aobrq(scmrq); + + free_page((unsigned long) scmrq->aob); + __scm_free_rq_cluster(scmrq); + kfree(scmrq->request); + kfree(aobrq); +} + +static void scm_free_rqs(void) +{ + struct list_head *iter, *safe; + struct scm_request *scmrq; + + spin_lock_irq(&list_lock); + list_for_each_safe(iter, safe, &inactive_requests) { + scmrq = list_entry(iter, struct scm_request, list); + list_del(&scmrq->list); + __scm_free_rq(scmrq); + } + spin_unlock_irq(&list_lock); + + mempool_destroy(aidaw_pool); +} + +static int __scm_alloc_rq(void) +{ + struct aob_rq_header *aobrq; + struct scm_request *scmrq; + + aobrq = kzalloc(sizeof(*aobrq) + sizeof(*scmrq), GFP_KERNEL); + if (!aobrq) + return -ENOMEM; + + scmrq = (void *) aobrq->data; + scmrq->aob = (void *) get_zeroed_page(GFP_DMA); + if (!scmrq->aob) + goto free; + + scmrq->request = kcalloc(nr_requests_per_io, sizeof(scmrq->request[0]), + GFP_KERNEL); + if (!scmrq->request) + goto free; + + if (__scm_alloc_rq_cluster(scmrq)) + goto free; + + INIT_LIST_HEAD(&scmrq->list); + spin_lock_irq(&list_lock); + list_add(&scmrq->list, &inactive_requests); + spin_unlock_irq(&list_lock); + + return 0; +free: + __scm_free_rq(scmrq); + return -ENOMEM; +} + +static int scm_alloc_rqs(unsigned int nrqs) +{ + int ret = 0; + + aidaw_pool = mempool_create_page_pool(max(nrqs/8, 1U), 0); + if (!aidaw_pool) + return -ENOMEM; + + while (nrqs-- && !ret) + ret = __scm_alloc_rq(); + + return ret; +} + +static struct scm_request *scm_request_fetch(void) +{ + struct scm_request *scmrq = NULL; + + spin_lock(&list_lock); + if (list_empty(&inactive_requests)) + goto out; + scmrq = list_first_entry(&inactive_requests, struct scm_request, list); + list_del(&scmrq->list); +out: + spin_unlock(&list_lock); + return scmrq; +} + +static void scm_request_done(struct scm_request *scmrq) +{ + unsigned long flags; + struct msb *msb; + u64 aidaw; + int i; + + for (i = 0; i < nr_requests_per_io && scmrq->request[i]; i++) { + msb = &scmrq->aob->msb[i]; + aidaw = msb->data_addr; + + if ((msb->flags & MSB_FLAG_IDA) && aidaw && + IS_ALIGNED(aidaw, PAGE_SIZE)) + mempool_free(virt_to_page(aidaw), aidaw_pool); + } + + spin_lock_irqsave(&list_lock, flags); + list_add(&scmrq->list, &inactive_requests); + spin_unlock_irqrestore(&list_lock, flags); +} + +static bool scm_permit_request(struct scm_blk_dev *bdev, struct request *req) +{ + return rq_data_dir(req) != WRITE || bdev->state != SCM_WR_PROHIBIT; +} + +static inline struct aidaw *scm_aidaw_alloc(void) +{ + struct page *page = mempool_alloc(aidaw_pool, GFP_ATOMIC); + + return page ? page_address(page) : NULL; +} + +static inline unsigned long scm_aidaw_bytes(struct aidaw *aidaw) +{ + unsigned long _aidaw = (unsigned long) aidaw; + unsigned long bytes = ALIGN(_aidaw, PAGE_SIZE) - _aidaw; + + return (bytes / sizeof(*aidaw)) * PAGE_SIZE; +} + +struct aidaw *scm_aidaw_fetch(struct scm_request *scmrq, unsigned int bytes) +{ + struct aidaw *aidaw; + + if (scm_aidaw_bytes(scmrq->next_aidaw) >= bytes) + return scmrq->next_aidaw; + + aidaw = scm_aidaw_alloc(); + if (aidaw) + memset(aidaw, 0, PAGE_SIZE); + return aidaw; +} + +static int scm_request_prepare(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + struct scm_device *scmdev = bdev->gendisk->private_data; + int pos = scmrq->aob->request.msb_count; + struct msb *msb = &scmrq->aob->msb[pos]; + struct request *req = scmrq->request[pos]; + struct req_iterator iter; + struct aidaw *aidaw; + struct bio_vec bv; + + aidaw = scm_aidaw_fetch(scmrq, blk_rq_bytes(req)); + if (!aidaw) + return -ENOMEM; + + msb->bs = MSB_BS_4K; + scmrq->aob->request.msb_count++; + msb->scm_addr = scmdev->address + ((u64) blk_rq_pos(req) << 9); + msb->oc = (rq_data_dir(req) == READ) ? MSB_OC_READ : MSB_OC_WRITE; + msb->flags |= MSB_FLAG_IDA; + msb->data_addr = (u64) aidaw; + + rq_for_each_segment(bv, req, iter) { + WARN_ON(bv.bv_offset); + msb->blk_count += bv.bv_len >> 12; + aidaw->data_addr = (u64) page_address(bv.bv_page); + aidaw++; + } + + scmrq->next_aidaw = aidaw; + return 0; +} + +static inline void scm_request_set(struct scm_request *scmrq, + struct request *req) +{ + scmrq->request[scmrq->aob->request.msb_count] = req; +} + +static inline void scm_request_init(struct scm_blk_dev *bdev, + struct scm_request *scmrq) +{ + struct aob_rq_header *aobrq = to_aobrq(scmrq); + struct aob *aob = scmrq->aob; + + memset(scmrq->request, 0, + nr_requests_per_io * sizeof(scmrq->request[0])); + memset(aob, 0, sizeof(*aob)); + aobrq->scmdev = bdev->scmdev; + aob->request.cmd_code = ARQB_CMD_MOVE; + aob->request.data = (u64) aobrq; + scmrq->bdev = bdev; + scmrq->retries = 4; + scmrq->error = 0; + /* We don't use all msbs - place aidaws at the end of the aob page. */ + scmrq->next_aidaw = (void *) &aob->msb[nr_requests_per_io]; + scm_request_cluster_init(scmrq); +} + +static void scm_ensure_queue_restart(struct scm_blk_dev *bdev) +{ + if (atomic_read(&bdev->queued_reqs)) { + /* Queue restart is triggered by the next interrupt. */ + return; + } + blk_delay_queue(bdev->rq, SCM_QUEUE_DELAY); +} + +void scm_request_requeue(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + int i; + + scm_release_cluster(scmrq); + for (i = 0; i < nr_requests_per_io && scmrq->request[i]; i++) + blk_requeue_request(bdev->rq, scmrq->request[i]); + + atomic_dec(&bdev->queued_reqs); + scm_request_done(scmrq); + scm_ensure_queue_restart(bdev); +} + +void scm_request_finish(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + int i; + + scm_release_cluster(scmrq); + for (i = 0; i < nr_requests_per_io && scmrq->request[i]; i++) + blk_end_request_all(scmrq->request[i], scmrq->error); + + atomic_dec(&bdev->queued_reqs); + scm_request_done(scmrq); +} + +static int scm_request_start(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + int ret; + + atomic_inc(&bdev->queued_reqs); + if (!scmrq->aob->request.msb_count) { + scm_request_requeue(scmrq); + return -EINVAL; + } + + ret = eadm_start_aob(scmrq->aob); + if (ret) { + SCM_LOG(5, "no subchannel"); + scm_request_requeue(scmrq); + } + return ret; +} + +static void scm_blk_request(struct request_queue *rq) +{ + struct scm_device *scmdev = rq->queuedata; + struct scm_blk_dev *bdev = dev_get_drvdata(&scmdev->dev); + struct scm_request *scmrq = NULL; + struct request *req; + + while ((req = blk_peek_request(rq))) { + if (req->cmd_type != REQ_TYPE_FS) { + blk_start_request(req); + blk_dump_rq_flags(req, KMSG_COMPONENT " bad request"); + blk_end_request_all(req, -EIO); + continue; + } + + if (!scm_permit_request(bdev, req)) + goto out; + + if (!scmrq) { + scmrq = scm_request_fetch(); + if (!scmrq) { + SCM_LOG(5, "no request"); + goto out; + } + scm_request_init(bdev, scmrq); + } + scm_request_set(scmrq, req); + + if (!scm_reserve_cluster(scmrq)) { + SCM_LOG(5, "cluster busy"); + scm_request_set(scmrq, NULL); + if (scmrq->aob->request.msb_count) + goto out; + + scm_request_done(scmrq); + return; + } + + if (scm_need_cluster_request(scmrq)) { + if (scmrq->aob->request.msb_count) { + /* Start cluster requests separately. */ + scm_request_set(scmrq, NULL); + if (scm_request_start(scmrq)) + return; + } else { + atomic_inc(&bdev->queued_reqs); + blk_start_request(req); + scm_initiate_cluster_request(scmrq); + } + scmrq = NULL; + continue; + } + + if (scm_request_prepare(scmrq)) { + SCM_LOG(5, "aidaw alloc failed"); + scm_request_set(scmrq, NULL); + goto out; + } + blk_start_request(req); + + if (scmrq->aob->request.msb_count < nr_requests_per_io) + continue; + + if (scm_request_start(scmrq)) + return; + + scmrq = NULL; + } +out: + if (scmrq) + scm_request_start(scmrq); + else + scm_ensure_queue_restart(bdev); +} + +static void __scmrq_log_error(struct scm_request *scmrq) +{ + struct aob *aob = scmrq->aob; + + if (scmrq->error == -ETIMEDOUT) + SCM_LOG(1, "Request timeout"); + else { + SCM_LOG(1, "Request error"); + SCM_LOG_HEX(1, &aob->response, sizeof(aob->response)); + } + if (scmrq->retries) + SCM_LOG(1, "Retry request"); + else + pr_err("An I/O operation to SCM failed with rc=%d\n", + scmrq->error); +} + +void scm_blk_irq(struct scm_device *scmdev, void *data, int error) +{ + struct scm_request *scmrq = data; + struct scm_blk_dev *bdev = scmrq->bdev; + + scmrq->error = error; + if (error) + __scmrq_log_error(scmrq); + + spin_lock(&bdev->lock); + list_add_tail(&scmrq->list, &bdev->finished_requests); + spin_unlock(&bdev->lock); + tasklet_hi_schedule(&bdev->tasklet); +} + +static void scm_blk_handle_error(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + unsigned long flags; + + if (scmrq->error != -EIO) + goto restart; + + /* For -EIO the response block is valid. */ + switch (scmrq->aob->response.eqc) { + case EQC_WR_PROHIBIT: + spin_lock_irqsave(&bdev->lock, flags); + if (bdev->state != SCM_WR_PROHIBIT) + pr_info("%lx: Write access to the SCM increment is suspended\n", + (unsigned long) bdev->scmdev->address); + bdev->state = SCM_WR_PROHIBIT; + spin_unlock_irqrestore(&bdev->lock, flags); + goto requeue; + default: + break; + } + +restart: + if (!eadm_start_aob(scmrq->aob)) + return; + +requeue: + spin_lock_irqsave(&bdev->rq_lock, flags); + scm_request_requeue(scmrq); + spin_unlock_irqrestore(&bdev->rq_lock, flags); +} + +static void scm_blk_tasklet(struct scm_blk_dev *bdev) +{ + struct scm_request *scmrq; + unsigned long flags; + + spin_lock_irqsave(&bdev->lock, flags); + while (!list_empty(&bdev->finished_requests)) { + scmrq = list_first_entry(&bdev->finished_requests, + struct scm_request, list); + list_del(&scmrq->list); + spin_unlock_irqrestore(&bdev->lock, flags); + + if (scmrq->error && scmrq->retries-- > 0) { + scm_blk_handle_error(scmrq); + + /* Request restarted or requeued, handle next. */ + spin_lock_irqsave(&bdev->lock, flags); + continue; + } + + if (scm_test_cluster_request(scmrq)) { + scm_cluster_request_irq(scmrq); + spin_lock_irqsave(&bdev->lock, flags); + continue; + } + + scm_request_finish(scmrq); + spin_lock_irqsave(&bdev->lock, flags); + } + spin_unlock_irqrestore(&bdev->lock, flags); + /* Look out for more requests. */ + blk_run_queue(bdev->rq); +} + +static const struct block_device_operations scm_blk_devops = { + .owner = THIS_MODULE, +}; + +int scm_blk_dev_setup(struct scm_blk_dev *bdev, struct scm_device *scmdev) +{ + struct request_queue *rq; + int len, ret = -ENOMEM; + unsigned int devindex, nr_max_blk; + + devindex = atomic_inc_return(&nr_devices) - 1; + /* scma..scmz + scmaa..scmzz */ + if (devindex > 701) { + ret = -ENODEV; + goto out; + } + + bdev->scmdev = scmdev; + bdev->state = SCM_OPER; + spin_lock_init(&bdev->rq_lock); + spin_lock_init(&bdev->lock); + INIT_LIST_HEAD(&bdev->finished_requests); + atomic_set(&bdev->queued_reqs, 0); + tasklet_init(&bdev->tasklet, + (void (*)(unsigned long)) scm_blk_tasklet, + (unsigned long) bdev); + + rq = blk_init_queue(scm_blk_request, &bdev->rq_lock); + if (!rq) + goto out; + + bdev->rq = rq; + nr_max_blk = min(scmdev->nr_max_block, + (unsigned int) (PAGE_SIZE / sizeof(struct aidaw))); + + blk_queue_logical_block_size(rq, 1 << 12); + blk_queue_max_hw_sectors(rq, nr_max_blk << 3); /* 8 * 512 = blk_size */ + blk_queue_max_segments(rq, nr_max_blk); + queue_flag_set_unlocked(QUEUE_FLAG_NONROT, rq); + queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, rq); + scm_blk_dev_cluster_setup(bdev); + + bdev->gendisk = alloc_disk(SCM_NR_PARTS); + if (!bdev->gendisk) + goto out_queue; + + rq->queuedata = scmdev; + bdev->gendisk->driverfs_dev = &scmdev->dev; + bdev->gendisk->private_data = scmdev; + bdev->gendisk->fops = &scm_blk_devops; + bdev->gendisk->queue = rq; + bdev->gendisk->major = scm_major; + bdev->gendisk->first_minor = devindex * SCM_NR_PARTS; + + len = snprintf(bdev->gendisk->disk_name, DISK_NAME_LEN, "scm"); + if (devindex > 25) { + len += snprintf(bdev->gendisk->disk_name + len, + DISK_NAME_LEN - len, "%c", + 'a' + (devindex / 26) - 1); + devindex = devindex % 26; + } + snprintf(bdev->gendisk->disk_name + len, DISK_NAME_LEN - len, "%c", + 'a' + devindex); + + /* 512 byte sectors */ + set_capacity(bdev->gendisk, scmdev->size >> 9); + add_disk(bdev->gendisk); + return 0; + +out_queue: + blk_cleanup_queue(rq); +out: + atomic_dec(&nr_devices); + return ret; +} + +void scm_blk_dev_cleanup(struct scm_blk_dev *bdev) +{ + tasklet_kill(&bdev->tasklet); + del_gendisk(bdev->gendisk); + blk_cleanup_queue(bdev->gendisk->queue); + put_disk(bdev->gendisk); +} + +void scm_blk_set_available(struct scm_blk_dev *bdev) +{ + unsigned long flags; + + spin_lock_irqsave(&bdev->lock, flags); + if (bdev->state == SCM_WR_PROHIBIT) + pr_info("%lx: Write access to the SCM increment is restored\n", + (unsigned long) bdev->scmdev->address); + bdev->state = SCM_OPER; + spin_unlock_irqrestore(&bdev->lock, flags); +} + +static bool __init scm_blk_params_valid(void) +{ + if (!nr_requests_per_io || nr_requests_per_io > 64) + return false; + + return scm_cluster_size_valid(); +} + +static int __init scm_blk_init(void) +{ + int ret = -EINVAL; + + if (!scm_blk_params_valid()) + goto out; + + ret = register_blkdev(0, "scm"); + if (ret < 0) + goto out; + + scm_major = ret; + ret = scm_alloc_rqs(nr_requests); + if (ret) + goto out_free; + + scm_debug = debug_register("scm_log", 16, 1, 16); + if (!scm_debug) { + ret = -ENOMEM; + goto out_free; + } + + debug_register_view(scm_debug, &debug_hex_ascii_view); + debug_set_level(scm_debug, 2); + + ret = scm_drv_init(); + if (ret) + goto out_dbf; + + return ret; + +out_dbf: + debug_unregister(scm_debug); +out_free: + scm_free_rqs(); + unregister_blkdev(scm_major, "scm"); +out: + return ret; +} +module_init(scm_blk_init); + +static void __exit scm_blk_cleanup(void) +{ + scm_drv_cleanup(); + debug_unregister(scm_debug); + scm_free_rqs(); + unregister_blkdev(scm_major, "scm"); +} +module_exit(scm_blk_cleanup); diff --git a/kernel/drivers/s390/block/scm_blk.h b/kernel/drivers/s390/block/scm_blk.h new file mode 100644 index 000000000..09218cdc5 --- /dev/null +++ b/kernel/drivers/s390/block/scm_blk.h @@ -0,0 +1,136 @@ +#ifndef SCM_BLK_H +#define SCM_BLK_H + +#include <linux/interrupt.h> +#include <linux/spinlock.h> +#include <linux/blkdev.h> +#include <linux/genhd.h> +#include <linux/list.h> + +#include <asm/debug.h> +#include <asm/eadm.h> + +#define SCM_NR_PARTS 8 +#define SCM_QUEUE_DELAY 5 + +struct scm_blk_dev { + struct tasklet_struct tasklet; + struct request_queue *rq; + struct gendisk *gendisk; + struct scm_device *scmdev; + spinlock_t rq_lock; /* guard the request queue */ + spinlock_t lock; /* guard the rest of the blockdev */ + atomic_t queued_reqs; + enum {SCM_OPER, SCM_WR_PROHIBIT} state; + struct list_head finished_requests; +#ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE + struct list_head cluster_list; +#endif +}; + +struct scm_request { + struct scm_blk_dev *bdev; + struct aidaw *next_aidaw; + struct request **request; + struct aob *aob; + struct list_head list; + u8 retries; + int error; +#ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE + struct { + enum {CLUSTER_NONE, CLUSTER_READ, CLUSTER_WRITE} state; + struct list_head list; + void **buf; + } cluster; +#endif +}; + +#define to_aobrq(rq) container_of((void *) rq, struct aob_rq_header, data) + +int scm_blk_dev_setup(struct scm_blk_dev *, struct scm_device *); +void scm_blk_dev_cleanup(struct scm_blk_dev *); +void scm_blk_set_available(struct scm_blk_dev *); +void scm_blk_irq(struct scm_device *, void *, int); + +void scm_request_finish(struct scm_request *); +void scm_request_requeue(struct scm_request *); + +struct aidaw *scm_aidaw_fetch(struct scm_request *scmrq, unsigned int bytes); + +int scm_drv_init(void); +void scm_drv_cleanup(void); + +#ifdef CONFIG_SCM_BLOCK_CLUSTER_WRITE +void __scm_free_rq_cluster(struct scm_request *); +int __scm_alloc_rq_cluster(struct scm_request *); +void scm_request_cluster_init(struct scm_request *); +bool scm_reserve_cluster(struct scm_request *); +void scm_release_cluster(struct scm_request *); +void scm_blk_dev_cluster_setup(struct scm_blk_dev *); +bool scm_need_cluster_request(struct scm_request *); +void scm_initiate_cluster_request(struct scm_request *); +void scm_cluster_request_irq(struct scm_request *); +bool scm_test_cluster_request(struct scm_request *); +bool scm_cluster_size_valid(void); +#else /* CONFIG_SCM_BLOCK_CLUSTER_WRITE */ +static inline void __scm_free_rq_cluster(struct scm_request *scmrq) {} +static inline int __scm_alloc_rq_cluster(struct scm_request *scmrq) +{ + return 0; +} +static inline void scm_request_cluster_init(struct scm_request *scmrq) {} +static inline bool scm_reserve_cluster(struct scm_request *scmrq) +{ + return true; +} +static inline void scm_release_cluster(struct scm_request *scmrq) {} +static inline void scm_blk_dev_cluster_setup(struct scm_blk_dev *bdev) {} +static inline bool scm_need_cluster_request(struct scm_request *scmrq) +{ + return false; +} +static inline void scm_initiate_cluster_request(struct scm_request *scmrq) {} +static inline void scm_cluster_request_irq(struct scm_request *scmrq) {} +static inline bool scm_test_cluster_request(struct scm_request *scmrq) +{ + return false; +} +static inline bool scm_cluster_size_valid(void) +{ + return true; +} +#endif /* CONFIG_SCM_BLOCK_CLUSTER_WRITE */ + +extern debug_info_t *scm_debug; + +#define SCM_LOG(imp, txt) do { \ + debug_text_event(scm_debug, imp, txt); \ + } while (0) + +static inline void SCM_LOG_HEX(int level, void *data, int length) +{ + if (!debug_level_enabled(scm_debug, level)) + return; + while (length > 0) { + debug_event(scm_debug, level, data, length); + length -= scm_debug->buf_size; + data += scm_debug->buf_size; + } +} + +static inline void SCM_LOG_STATE(int level, struct scm_device *scmdev) +{ + struct { + u64 address; + u8 oper_state; + u8 rank; + } __packed data = { + .address = scmdev->address, + .oper_state = scmdev->attrs.oper_state, + .rank = scmdev->attrs.rank, + }; + + SCM_LOG_HEX(level, &data, sizeof(data)); +} + +#endif /* SCM_BLK_H */ diff --git a/kernel/drivers/s390/block/scm_blk_cluster.c b/kernel/drivers/s390/block/scm_blk_cluster.c new file mode 100644 index 000000000..7497ddde2 --- /dev/null +++ b/kernel/drivers/s390/block/scm_blk_cluster.c @@ -0,0 +1,255 @@ +/* + * Block driver for s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott <sebott@linux.vnet.ibm.com> + */ + +#include <linux/spinlock.h> +#include <linux/module.h> +#include <linux/blkdev.h> +#include <linux/genhd.h> +#include <linux/slab.h> +#include <linux/list.h> +#include <asm/eadm.h> +#include "scm_blk.h" + +static unsigned int write_cluster_size = 64; +module_param(write_cluster_size, uint, S_IRUGO); +MODULE_PARM_DESC(write_cluster_size, + "Number of pages used for contiguous writes."); + +#define CLUSTER_SIZE (write_cluster_size * PAGE_SIZE) + +void __scm_free_rq_cluster(struct scm_request *scmrq) +{ + int i; + + if (!scmrq->cluster.buf) + return; + + for (i = 0; i < 2 * write_cluster_size; i++) + free_page((unsigned long) scmrq->cluster.buf[i]); + + kfree(scmrq->cluster.buf); +} + +int __scm_alloc_rq_cluster(struct scm_request *scmrq) +{ + int i; + + scmrq->cluster.buf = kzalloc(sizeof(void *) * 2 * write_cluster_size, + GFP_KERNEL); + if (!scmrq->cluster.buf) + return -ENOMEM; + + for (i = 0; i < 2 * write_cluster_size; i++) { + scmrq->cluster.buf[i] = (void *) get_zeroed_page(GFP_DMA); + if (!scmrq->cluster.buf[i]) + return -ENOMEM; + } + INIT_LIST_HEAD(&scmrq->cluster.list); + return 0; +} + +void scm_request_cluster_init(struct scm_request *scmrq) +{ + scmrq->cluster.state = CLUSTER_NONE; +} + +static bool clusters_intersect(struct request *A, struct request *B) +{ + unsigned long firstA, lastA, firstB, lastB; + + firstA = ((u64) blk_rq_pos(A) << 9) / CLUSTER_SIZE; + lastA = (((u64) blk_rq_pos(A) << 9) + + blk_rq_bytes(A) - 1) / CLUSTER_SIZE; + + firstB = ((u64) blk_rq_pos(B) << 9) / CLUSTER_SIZE; + lastB = (((u64) blk_rq_pos(B) << 9) + + blk_rq_bytes(B) - 1) / CLUSTER_SIZE; + + return (firstB <= lastA && firstA <= lastB); +} + +bool scm_reserve_cluster(struct scm_request *scmrq) +{ + struct request *req = scmrq->request[scmrq->aob->request.msb_count]; + struct scm_blk_dev *bdev = scmrq->bdev; + struct scm_request *iter; + int pos, add = 1; + + if (write_cluster_size == 0) + return true; + + spin_lock(&bdev->lock); + list_for_each_entry(iter, &bdev->cluster_list, cluster.list) { + if (iter == scmrq) { + /* + * We don't have to use clusters_intersect here, since + * cluster requests are always started separately. + */ + add = 0; + continue; + } + for (pos = 0; pos < iter->aob->request.msb_count; pos++) { + if (clusters_intersect(req, iter->request[pos]) && + (rq_data_dir(req) == WRITE || + rq_data_dir(iter->request[pos]) == WRITE)) { + spin_unlock(&bdev->lock); + return false; + } + } + } + if (add) + list_add(&scmrq->cluster.list, &bdev->cluster_list); + spin_unlock(&bdev->lock); + + return true; +} + +void scm_release_cluster(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + unsigned long flags; + + if (write_cluster_size == 0) + return; + + spin_lock_irqsave(&bdev->lock, flags); + list_del(&scmrq->cluster.list); + spin_unlock_irqrestore(&bdev->lock, flags); +} + +void scm_blk_dev_cluster_setup(struct scm_blk_dev *bdev) +{ + INIT_LIST_HEAD(&bdev->cluster_list); + blk_queue_io_opt(bdev->rq, CLUSTER_SIZE); +} + +static int scm_prepare_cluster_request(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + struct scm_device *scmdev = bdev->gendisk->private_data; + struct request *req = scmrq->request[0]; + struct msb *msb = &scmrq->aob->msb[0]; + struct req_iterator iter; + struct aidaw *aidaw; + struct bio_vec bv; + int i = 0; + u64 addr; + + switch (scmrq->cluster.state) { + case CLUSTER_NONE: + scmrq->cluster.state = CLUSTER_READ; + /* fall through */ + case CLUSTER_READ: + msb->bs = MSB_BS_4K; + msb->oc = MSB_OC_READ; + msb->flags = MSB_FLAG_IDA; + msb->blk_count = write_cluster_size; + + addr = scmdev->address + ((u64) blk_rq_pos(req) << 9); + msb->scm_addr = round_down(addr, CLUSTER_SIZE); + + if (msb->scm_addr != + round_down(addr + (u64) blk_rq_bytes(req) - 1, + CLUSTER_SIZE)) + msb->blk_count = 2 * write_cluster_size; + + aidaw = scm_aidaw_fetch(scmrq, msb->blk_count * PAGE_SIZE); + if (!aidaw) + return -ENOMEM; + + scmrq->aob->request.msb_count = 1; + msb->data_addr = (u64) aidaw; + for (i = 0; i < msb->blk_count; i++) { + aidaw->data_addr = (u64) scmrq->cluster.buf[i]; + aidaw++; + } + + break; + case CLUSTER_WRITE: + aidaw = (void *) msb->data_addr; + msb->oc = MSB_OC_WRITE; + + for (addr = msb->scm_addr; + addr < scmdev->address + ((u64) blk_rq_pos(req) << 9); + addr += PAGE_SIZE) { + aidaw->data_addr = (u64) scmrq->cluster.buf[i]; + aidaw++; + i++; + } + rq_for_each_segment(bv, req, iter) { + aidaw->data_addr = (u64) page_address(bv.bv_page); + aidaw++; + i++; + } + for (; i < msb->blk_count; i++) { + aidaw->data_addr = (u64) scmrq->cluster.buf[i]; + aidaw++; + } + break; + } + return 0; +} + +bool scm_need_cluster_request(struct scm_request *scmrq) +{ + int pos = scmrq->aob->request.msb_count; + + if (rq_data_dir(scmrq->request[pos]) == READ) + return false; + + return blk_rq_bytes(scmrq->request[pos]) < CLUSTER_SIZE; +} + +/* Called with queue lock held. */ +void scm_initiate_cluster_request(struct scm_request *scmrq) +{ + if (scm_prepare_cluster_request(scmrq)) + goto requeue; + if (eadm_start_aob(scmrq->aob)) + goto requeue; + return; +requeue: + scm_request_requeue(scmrq); +} + +bool scm_test_cluster_request(struct scm_request *scmrq) +{ + return scmrq->cluster.state != CLUSTER_NONE; +} + +void scm_cluster_request_irq(struct scm_request *scmrq) +{ + struct scm_blk_dev *bdev = scmrq->bdev; + unsigned long flags; + + switch (scmrq->cluster.state) { + case CLUSTER_NONE: + BUG(); + break; + case CLUSTER_READ: + if (scmrq->error) { + scm_request_finish(scmrq); + break; + } + scmrq->cluster.state = CLUSTER_WRITE; + spin_lock_irqsave(&bdev->rq_lock, flags); + scm_initiate_cluster_request(scmrq); + spin_unlock_irqrestore(&bdev->rq_lock, flags); + break; + case CLUSTER_WRITE: + scm_request_finish(scmrq); + break; + } +} + +bool scm_cluster_size_valid(void) +{ + if (write_cluster_size == 1 || write_cluster_size > 128) + return false; + + return !(write_cluster_size & (write_cluster_size - 1)); +} diff --git a/kernel/drivers/s390/block/scm_drv.c b/kernel/drivers/s390/block/scm_drv.c new file mode 100644 index 000000000..c98cf52d7 --- /dev/null +++ b/kernel/drivers/s390/block/scm_drv.c @@ -0,0 +1,92 @@ +/* + * Device driver for s390 storage class memory. + * + * Copyright IBM Corp. 2012 + * Author(s): Sebastian Ott <sebott@linux.vnet.ibm.com> + */ + +#define KMSG_COMPONENT "scm_block" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include <linux/module.h> +#include <linux/slab.h> +#include <asm/eadm.h> +#include "scm_blk.h" + +static void scm_notify(struct scm_device *scmdev, enum scm_event event) +{ + struct scm_blk_dev *bdev = dev_get_drvdata(&scmdev->dev); + + switch (event) { + case SCM_CHANGE: + pr_info("%lx: The capabilities of the SCM increment changed\n", + (unsigned long) scmdev->address); + SCM_LOG(2, "State changed"); + SCM_LOG_STATE(2, scmdev); + break; + case SCM_AVAIL: + SCM_LOG(2, "Increment available"); + SCM_LOG_STATE(2, scmdev); + scm_blk_set_available(bdev); + break; + } +} + +static int scm_probe(struct scm_device *scmdev) +{ + struct scm_blk_dev *bdev; + int ret; + + SCM_LOG(2, "probe"); + SCM_LOG_STATE(2, scmdev); + + if (scmdev->attrs.oper_state != OP_STATE_GOOD) + return -EINVAL; + + bdev = kzalloc(sizeof(*bdev), GFP_KERNEL); + if (!bdev) + return -ENOMEM; + + dev_set_drvdata(&scmdev->dev, bdev); + ret = scm_blk_dev_setup(bdev, scmdev); + if (ret) { + dev_set_drvdata(&scmdev->dev, NULL); + kfree(bdev); + goto out; + } + +out: + return ret; +} + +static int scm_remove(struct scm_device *scmdev) +{ + struct scm_blk_dev *bdev = dev_get_drvdata(&scmdev->dev); + + scm_blk_dev_cleanup(bdev); + dev_set_drvdata(&scmdev->dev, NULL); + kfree(bdev); + + return 0; +} + +static struct scm_driver scm_drv = { + .drv = { + .name = "scm_block", + .owner = THIS_MODULE, + }, + .notify = scm_notify, + .probe = scm_probe, + .remove = scm_remove, + .handler = scm_blk_irq, +}; + +int __init scm_drv_init(void) +{ + return scm_driver_register(&scm_drv); +} + +void scm_drv_cleanup(void) +{ + scm_driver_unregister(&scm_drv); +} diff --git a/kernel/drivers/s390/block/xpram.c b/kernel/drivers/s390/block/xpram.c new file mode 100644 index 000000000..7d4e9397a --- /dev/null +++ b/kernel/drivers/s390/block/xpram.c @@ -0,0 +1,478 @@ +/* + * Xpram.c -- the S/390 expanded memory RAM-disk + * + * significant parts of this code are based on + * the sbull device driver presented in + * A. Rubini: Linux Device Drivers + * + * Author of XPRAM specific coding: Reinhard Buendgen + * buendgen@de.ibm.com + * Rewrite for 2.5: Martin Schwidefsky <schwidefsky@de.ibm.com> + * + * External interfaces: + * Interfaces to linux kernel + * xpram_setup: read kernel parameters + * Device specific file operations + * xpram_iotcl + * xpram_open + * + * "ad-hoc" partitioning: + * the expanded memory can be partitioned among several devices + * (with different minors). The partitioning set up can be + * set by kernel or module parameters (int devs & int sizes[]) + * + * Potential future improvements: + * generic hard disk support to replace ad-hoc partitioning + */ + +#define KMSG_COMPONENT "xpram" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include <linux/module.h> +#include <linux/moduleparam.h> +#include <linux/ctype.h> /* isdigit, isxdigit */ +#include <linux/errno.h> +#include <linux/init.h> +#include <linux/blkdev.h> +#include <linux/blkpg.h> +#include <linux/hdreg.h> /* HDIO_GETGEO */ +#include <linux/device.h> +#include <linux/bio.h> +#include <linux/suspend.h> +#include <linux/platform_device.h> +#include <linux/gfp.h> +#include <asm/uaccess.h> + +#define XPRAM_NAME "xpram" +#define XPRAM_DEVS 1 /* one partition */ +#define XPRAM_MAX_DEVS 32 /* maximal number of devices (partitions) */ + +typedef struct { + unsigned int size; /* size of xpram segment in pages */ + unsigned int offset; /* start page of xpram segment */ +} xpram_device_t; + +static xpram_device_t xpram_devices[XPRAM_MAX_DEVS]; +static unsigned int xpram_sizes[XPRAM_MAX_DEVS]; +static struct gendisk *xpram_disks[XPRAM_MAX_DEVS]; +static struct request_queue *xpram_queues[XPRAM_MAX_DEVS]; +static unsigned int xpram_pages; +static int xpram_devs; + +/* + * Parameter parsing functions. + */ +static int devs = XPRAM_DEVS; +static char *sizes[XPRAM_MAX_DEVS]; + +module_param(devs, int, 0); +module_param_array(sizes, charp, NULL, 0); + +MODULE_PARM_DESC(devs, "number of devices (\"partitions\"), " \ + "the default is " __MODULE_STRING(XPRAM_DEVS) "\n"); +MODULE_PARM_DESC(sizes, "list of device (partition) sizes " \ + "the defaults are 0s \n" \ + "All devices with size 0 equally partition the " + "remaining space on the expanded strorage not " + "claimed by explicit sizes\n"); +MODULE_LICENSE("GPL"); + +/* + * Copy expanded memory page (4kB) into main memory + * Arguments + * page_addr: address of target page + * xpage_index: index of expandeded memory page + * Return value + * 0: if operation succeeds + * -EIO: if pgin failed + * -ENXIO: if xpram has vanished + */ +static int xpram_page_in (unsigned long page_addr, unsigned int xpage_index) +{ + int cc = 2; /* return unused cc 2 if pgin traps */ + + asm volatile( + " .insn rre,0xb22e0000,%1,%2\n" /* pgin %1,%2 */ + "0: ipm %0\n" + " srl %0,28\n" + "1:\n" + EX_TABLE(0b,1b) + : "+d" (cc) : "a" (__pa(page_addr)), "d" (xpage_index) : "cc"); + if (cc == 3) + return -ENXIO; + if (cc == 2) + return -ENXIO; + if (cc == 1) + return -EIO; + return 0; +} + +/* + * Copy a 4kB page of main memory to an expanded memory page + * Arguments + * page_addr: address of source page + * xpage_index: index of expandeded memory page + * Return value + * 0: if operation succeeds + * -EIO: if pgout failed + * -ENXIO: if xpram has vanished + */ +static long xpram_page_out (unsigned long page_addr, unsigned int xpage_index) +{ + int cc = 2; /* return unused cc 2 if pgin traps */ + + asm volatile( + " .insn rre,0xb22f0000,%1,%2\n" /* pgout %1,%2 */ + "0: ipm %0\n" + " srl %0,28\n" + "1:\n" + EX_TABLE(0b,1b) + : "+d" (cc) : "a" (__pa(page_addr)), "d" (xpage_index) : "cc"); + if (cc == 3) + return -ENXIO; + if (cc == 2) + return -ENXIO; + if (cc == 1) + return -EIO; + return 0; +} + +/* + * Check if xpram is available. + */ +static int xpram_present(void) +{ + unsigned long mem_page; + int rc; + + mem_page = (unsigned long) __get_free_page(GFP_KERNEL); + if (!mem_page) + return -ENOMEM; + rc = xpram_page_in(mem_page, 0); + free_page(mem_page); + return rc ? -ENXIO : 0; +} + +/* + * Return index of the last available xpram page. + */ +static unsigned long xpram_highest_page_index(void) +{ + unsigned int page_index, add_bit; + unsigned long mem_page; + + mem_page = (unsigned long) __get_free_page(GFP_KERNEL); + if (!mem_page) + return 0; + + page_index = 0; + add_bit = 1ULL << (sizeof(unsigned int)*8 - 1); + while (add_bit > 0) { + if (xpram_page_in(mem_page, page_index | add_bit) == 0) + page_index |= add_bit; + add_bit >>= 1; + } + + free_page (mem_page); + + return page_index; +} + +/* + * Block device make request function. + */ +static void xpram_make_request(struct request_queue *q, struct bio *bio) +{ + xpram_device_t *xdev = bio->bi_bdev->bd_disk->private_data; + struct bio_vec bvec; + struct bvec_iter iter; + unsigned int index; + unsigned long page_addr; + unsigned long bytes; + + if ((bio->bi_iter.bi_sector & 7) != 0 || + (bio->bi_iter.bi_size & 4095) != 0) + /* Request is not page-aligned. */ + goto fail; + if ((bio->bi_iter.bi_size >> 12) > xdev->size) + /* Request size is no page-aligned. */ + goto fail; + if ((bio->bi_iter.bi_sector >> 3) > 0xffffffffU - xdev->offset) + goto fail; + index = (bio->bi_iter.bi_sector >> 3) + xdev->offset; + bio_for_each_segment(bvec, bio, iter) { + page_addr = (unsigned long) + kmap(bvec.bv_page) + bvec.bv_offset; + bytes = bvec.bv_len; + if ((page_addr & 4095) != 0 || (bytes & 4095) != 0) + /* More paranoia. */ + goto fail; + while (bytes > 0) { + if (bio_data_dir(bio) == READ) { + if (xpram_page_in(page_addr, index) != 0) + goto fail; + } else { + if (xpram_page_out(page_addr, index) != 0) + goto fail; + } + page_addr += 4096; + bytes -= 4096; + index++; + } + } + set_bit(BIO_UPTODATE, &bio->bi_flags); + bio_endio(bio, 0); + return; +fail: + bio_io_error(bio); +} + +static int xpram_getgeo(struct block_device *bdev, struct hd_geometry *geo) +{ + unsigned long size; + + /* + * get geometry: we have to fake one... trim the size to a + * multiple of 64 (32k): tell we have 16 sectors, 4 heads, + * whatever cylinders. Tell also that data starts at sector. 4. + */ + size = (xpram_pages * 8) & ~0x3f; + geo->cylinders = size >> 6; + geo->heads = 4; + geo->sectors = 16; + geo->start = 4; + return 0; +} + +static const struct block_device_operations xpram_devops = +{ + .owner = THIS_MODULE, + .getgeo = xpram_getgeo, +}; + +/* + * Setup xpram_sizes array. + */ +static int __init xpram_setup_sizes(unsigned long pages) +{ + unsigned long mem_needed; + unsigned long mem_auto; + unsigned long long size; + char *sizes_end; + int mem_auto_no; + int i; + + /* Check number of devices. */ + if (devs <= 0 || devs > XPRAM_MAX_DEVS) { + pr_err("%d is not a valid number of XPRAM devices\n",devs); + return -EINVAL; + } + xpram_devs = devs; + + /* + * Copy sizes array to xpram_sizes and align partition + * sizes to page boundary. + */ + mem_needed = 0; + mem_auto_no = 0; + for (i = 0; i < xpram_devs; i++) { + if (sizes[i]) { + size = simple_strtoull(sizes[i], &sizes_end, 0); + switch (*sizes_end) { + case 'g': + case 'G': + size <<= 20; + break; + case 'm': + case 'M': + size <<= 10; + } + xpram_sizes[i] = (size + 3) & -4UL; + } + if (xpram_sizes[i]) + mem_needed += xpram_sizes[i]; + else + mem_auto_no++; + } + + pr_info(" number of devices (partitions): %d \n", xpram_devs); + for (i = 0; i < xpram_devs; i++) { + if (xpram_sizes[i]) + pr_info(" size of partition %d: %u kB\n", + i, xpram_sizes[i]); + else + pr_info(" size of partition %d to be set " + "automatically\n",i); + } + pr_info(" memory needed (for sized partitions): %lu kB\n", + mem_needed); + pr_info(" partitions to be sized automatically: %d\n", + mem_auto_no); + + if (mem_needed > pages * 4) { + pr_err("Not enough expanded memory available\n"); + return -EINVAL; + } + + /* + * partitioning: + * xpram_sizes[i] != 0; partition i has size xpram_sizes[i] kB + * else: ; all partitions with zero xpram_sizes[i] + * partition equally the remaining space + */ + if (mem_auto_no) { + mem_auto = ((pages - mem_needed / 4) / mem_auto_no) * 4; + pr_info(" automatically determined " + "partition size: %lu kB\n", mem_auto); + for (i = 0; i < xpram_devs; i++) + if (xpram_sizes[i] == 0) + xpram_sizes[i] = mem_auto; + } + return 0; +} + +static int __init xpram_setup_blkdev(void) +{ + unsigned long offset; + int i, rc = -ENOMEM; + + for (i = 0; i < xpram_devs; i++) { + xpram_disks[i] = alloc_disk(1); + if (!xpram_disks[i]) + goto out; + xpram_queues[i] = blk_alloc_queue(GFP_KERNEL); + if (!xpram_queues[i]) { + put_disk(xpram_disks[i]); + goto out; + } + queue_flag_set_unlocked(QUEUE_FLAG_NONROT, xpram_queues[i]); + queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, xpram_queues[i]); + blk_queue_make_request(xpram_queues[i], xpram_make_request); + blk_queue_logical_block_size(xpram_queues[i], 4096); + } + + /* + * Register xpram major. + */ + rc = register_blkdev(XPRAM_MAJOR, XPRAM_NAME); + if (rc < 0) + goto out; + + /* + * Setup device structures. + */ + offset = 0; + for (i = 0; i < xpram_devs; i++) { + struct gendisk *disk = xpram_disks[i]; + + xpram_devices[i].size = xpram_sizes[i] / 4; + xpram_devices[i].offset = offset; + offset += xpram_devices[i].size; + disk->major = XPRAM_MAJOR; + disk->first_minor = i; + disk->fops = &xpram_devops; + disk->private_data = &xpram_devices[i]; + disk->queue = xpram_queues[i]; + sprintf(disk->disk_name, "slram%d", i); + set_capacity(disk, xpram_sizes[i] << 1); + add_disk(disk); + } + + return 0; +out: + while (i--) { + blk_cleanup_queue(xpram_queues[i]); + put_disk(xpram_disks[i]); + } + return rc; +} + +/* + * Resume failed: Print error message and call panic. + */ +static void xpram_resume_error(const char *message) +{ + pr_err("Resuming the system failed: %s\n", message); + panic("xpram resume error\n"); +} + +/* + * Check if xpram setup changed between suspend and resume. + */ +static int xpram_restore(struct device *dev) +{ + if (!xpram_pages) + return 0; + if (xpram_present() != 0) + xpram_resume_error("xpram disappeared"); + if (xpram_pages != xpram_highest_page_index() + 1) + xpram_resume_error("Size of xpram changed"); + return 0; +} + +static const struct dev_pm_ops xpram_pm_ops = { + .restore = xpram_restore, +}; + +static struct platform_driver xpram_pdrv = { + .driver = { + .name = XPRAM_NAME, + .pm = &xpram_pm_ops, + }, +}; + +static struct platform_device *xpram_pdev; + +/* + * Finally, the init/exit functions. + */ +static void __exit xpram_exit(void) +{ + int i; + for (i = 0; i < xpram_devs; i++) { + del_gendisk(xpram_disks[i]); + blk_cleanup_queue(xpram_queues[i]); + put_disk(xpram_disks[i]); + } + unregister_blkdev(XPRAM_MAJOR, XPRAM_NAME); + platform_device_unregister(xpram_pdev); + platform_driver_unregister(&xpram_pdrv); +} + +static int __init xpram_init(void) +{ + int rc; + + /* Find out size of expanded memory. */ + if (xpram_present() != 0) { + pr_err("No expanded memory available\n"); + return -ENODEV; + } + xpram_pages = xpram_highest_page_index() + 1; + pr_info(" %u pages expanded memory found (%lu KB).\n", + xpram_pages, (unsigned long) xpram_pages*4); + rc = xpram_setup_sizes(xpram_pages); + if (rc) + return rc; + rc = platform_driver_register(&xpram_pdrv); + if (rc) + return rc; + xpram_pdev = platform_device_register_simple(XPRAM_NAME, -1, NULL, 0); + if (IS_ERR(xpram_pdev)) { + rc = PTR_ERR(xpram_pdev); + goto fail_platform_driver_unregister; + } + rc = xpram_setup_blkdev(); + if (rc) + goto fail_platform_device_unregister; + return 0; + +fail_platform_device_unregister: + platform_device_unregister(xpram_pdev); +fail_platform_driver_unregister: + platform_driver_unregister(&xpram_pdrv); + return rc; +} + +module_init(xpram_init); +module_exit(xpram_exit); |