Add the rt linux 4.1.3-rt3 as base

Import the rt linux 4.1.3-rt3 as OPNFV kvm base.

It's from git://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git linux-4.1.y-rt and
the base is:

commit 0917f823c59692d751951bf5ea699a2d1e2f26a2
Author: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Date:   Sat Jul 25 12:13:34 2015 +0200

    Prepare v4.1.3-rt3

    Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

We lose all the git history this way and it's not good. We
should apply another opnfv project repo in future.

Change-Id: I87543d81c9df70d99c5001fbdf646b202c19f423
Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>

17 files changed, 9801 insertions, 0 deletions

diff --git a/kernel/arch/cris/arch-v32/drivers/Kconfig b/kernel/arch/cris/arch-v32/drivers/Kconfig
new file mode 100644
index 000000000..4fc16b44f
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/Kconfig
@@ -0,0 +1,430 @@
+if ETRAX_ARCH_V32
+
+config ETRAX_ETHERNET
+	bool "Ethernet support"
+	depends on ETRAX_ARCH_V32 && NETDEVICES
+	select MII
+	help
+	  This option enables the ETRAX FS built-in 10/100Mbit Ethernet
+	  controller.
+
+config ETRAX_NO_PHY
+	bool "PHY not present"
+	depends on ETRAX_ETHERNET
+	help
+	  This option disables all MDIO communication with an ethernet
+	  transceiver connected to the MII interface. This option shall
+	  typically be enabled if the MII interface is connected to a
+	  switch. This option should normally be disabled. If enabled,
+	  speed and duplex will be locked to 100 Mbit and full duplex.
+
+config ETRAXFS_SERIAL
+	bool "Serial-port support"
+	depends on ETRAX_ARCH_V32
+	select SERIAL_CORE
+	select SERIAL_CORE_CONSOLE
+	help
+	  Enables the ETRAX FS serial driver for ser0 (ttyS0)
+	  You probably want this enabled.
+
+config ETRAX_RS485
+	bool "RS-485 support"
+	depends on ETRAXFS_SERIAL
+	help
+	  Enables support for RS-485 serial communication.
+
+config ETRAX_RS485_DISABLE_RECEIVER
+	bool "Disable serial receiver"
+	depends on ETRAX_RS485
+	help
+	  It is necessary to disable the serial receiver to avoid serial
+	  loopback.  Not all products are able to do this in software only.
+
+config ETRAX_SERIAL_PORT0
+	bool "Serial port 0 enabled"
+	depends on ETRAXFS_SERIAL
+	help
+	  Enables the ETRAX FS serial driver for ser0 (ttyS0)
+	  Normally you want this on. You can control what DMA channels to use
+	  if you do not need DMA to something else.
+	  ser0 can use dma4 or dma6 for output and dma5 or dma7 for input.
+
+config ETRAX_SERIAL_PORT1
+	bool "Serial port 1 enabled"
+	depends on ETRAXFS_SERIAL
+	help
+	  Enables the ETRAX FS serial driver for ser1 (ttyS1).
+
+config ETRAX_SERIAL_PORT2
+	bool "Serial port 2 enabled"
+	depends on ETRAXFS_SERIAL
+	help
+	  Enables the ETRAX FS serial driver for ser2 (ttyS2).
+
+config ETRAX_SERIAL_PORT3
+	bool "Serial port 3 enabled"
+	depends on ETRAXFS_SERIAL
+	help
+	  Enables the ETRAX FS serial driver for ser3 (ttyS3).
+
+config ETRAX_SYNCHRONOUS_SERIAL
+	bool "Synchronous serial-port support"
+	depends on ETRAX_ARCH_V32
+	help
+	  Enables the ETRAX FS synchronous serial driver.
+
+config ETRAX_SYNCHRONOUS_SERIAL_PORT0
+         bool "Synchronous serial port 0 enabled"
+         depends on ETRAX_SYNCHRONOUS_SERIAL
+         help
+           Enabled synchronous serial port 0.
+
+config ETRAX_SYNCHRONOUS_SERIAL0_DMA
+         bool "Enable DMA on synchronous serial port 0."
+         depends on ETRAX_SYNCHRONOUS_SERIAL_PORT0
+         help
+           A synchronous serial port can run in manual or DMA mode.
+           Selecting this option will make it run in DMA mode.
+
+config ETRAX_SYNCHRONOUS_SERIAL_PORT1
+         bool "Synchronous serial port 1 enabled"
+         depends on ETRAX_SYNCHRONOUS_SERIAL && ETRAXFS
+         help
+           Enabled synchronous serial port 1.
+
+config ETRAX_SYNCHRONOUS_SERIAL1_DMA
+         bool "Enable DMA on synchronous serial port 1."
+         depends on ETRAX_SYNCHRONOUS_SERIAL_PORT1
+         help
+           A synchronous serial port can run in manual or DMA mode.
+           Selecting this option will make it run in DMA mode.
+
+config ETRAX_AXISFLASHMAP
+	bool "Axis flash-map support"
+	depends on ETRAX_ARCH_V32
+	select MTD
+	select MTD_CFI
+	select MTD_CFI_AMDSTD
+	select MTD_JEDECPROBE
+	select MTD_BLOCK
+	select MTD_COMPLEX_MAPPINGS
+	select MTD_MTDRAM
+	help
+	  This option enables MTD mapping of flash devices.  Needed to use
+	  flash memories.  If unsure, say Y.
+
+config ETRAX_AXISFLASHMAP_MTD0WHOLE
+	bool "MTD0 is whole boot flash device"
+	depends on ETRAX_AXISFLASHMAP
+	help
+	  When this option is not set, mtd0 refers to the first partition
+	  on the boot flash device. When set, mtd0 refers to the whole
+	  device, with mtd1 referring to the first partition etc.
+
+config ETRAX_PTABLE_SECTOR
+	int "Byte-offset of partition table sector"
+	depends on ETRAX_AXISFLASHMAP
+	default "65536"
+	help
+	  Byte-offset of the partition table in the first flash chip.
+	  The default value is 64kB and should not be changed unless
+	  you know exactly what you are doing. The only valid reason
+	  for changing this is when the flash block size is bigger
+	  than 64kB (e.g. when using two parallel 16 bit flashes).
+
+config ETRAX_NANDFLASH
+	bool "NAND flash support"
+	depends on ETRAX_ARCH_V32
+	select MTD_NAND
+	select MTD_NAND_IDS
+	help
+	  This option enables MTD mapping of NAND flash devices.  Needed to use
+	  NAND flash memories.  If unsure, say Y.
+
+config ETRAX_NANDBOOT
+	bool "Boot from NAND flash"
+	depends on ETRAX_NANDFLASH
+	help
+	  This options enables booting from NAND flash devices.
+	  Say Y if your boot code, kernel and root file system is in
+	  NAND flash. Say N if they are in NOR flash.
+
+config ETRAX_I2C
+	bool "I2C driver"
+	depends on ETRAX_ARCH_V32
+	help
+	  This option enables the I2C driver used by e.g. the RTC driver.
+
+config ETRAX_V32_I2C_DATA_PORT
+	string "I2C data pin"
+	depends on ETRAX_I2C
+	help
+	  The pin to use for I2C data.
+
+config ETRAX_V32_I2C_CLK_PORT
+	string "I2C clock pin"
+	depends on ETRAX_I2C
+	help
+	  The pin to use for I2C clock.
+
+config ETRAX_GPIO
+	bool "GPIO support"
+	depends on ETRAX_ARCH_V32
+	---help---
+	  Enables the ETRAX general port device (major 120, minors 0-4).
+	  You can use this driver to access the general port bits. It supports
+	  these ioctl's:
+	  #include <linux/etraxgpio.h>
+	  fd = open("/dev/gpioa", O_RDWR); // or /dev/gpiob
+	  ioctl(fd, _IO(ETRAXGPIO_IOCTYPE, IO_SETBITS), bits_to_set);
+	  ioctl(fd, _IO(ETRAXGPIO_IOCTYPE, IO_CLRBITS), bits_to_clear);
+	  err = ioctl(fd, _IO(ETRAXGPIO_IOCTYPE, IO_READ_INBITS), &val);
+	  Remember that you need to setup the port directions appropriately in
+	  the General configuration.
+
+config ETRAX_VIRTUAL_GPIO
+	bool "Virtual GPIO support"
+	depends on ETRAX_GPIO
+	help
+	  Enables the virtual Etrax general port device (major 120, minor 6).
+	  It uses an I/O expander for the I2C-bus.
+
+config ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN
+	int "Virtual GPIO interrupt pin on PA pin"
+	range 0 7
+	depends on ETRAX_VIRTUAL_GPIO
+	help
+	  The pin to use on PA for virtual gpio interrupt.
+
+config ETRAX_PA_CHANGEABLE_DIR
+	hex "PA user changeable dir mask"
+	depends on ETRAX_GPIO
+	default "0x00" if ETRAXFS
+	default "0x00000000" if !ETRAXFS
+	help
+	  This is a bitmask (8 bits) with information of what bits in PA that a
+	  user can change direction on using ioctl's.
+	  Bit set = changeable.
+	  You probably want 0 here, but it depends on your hardware.
+
+config ETRAX_PA_CHANGEABLE_BITS
+	hex "PA user changeable bits mask"
+	depends on ETRAX_GPIO
+	default "0x00" if ETRAXFS
+	default "0x00000000" if !ETRAXFS
+	help
+	  This is a bitmask (8 bits) with information of what bits in PA
+	  that a user can change the value on using ioctl's.
+	  Bit set = changeable.
+
+config ETRAX_PB_CHANGEABLE_DIR
+	hex "PB user changeable dir mask"
+	depends on ETRAX_GPIO
+	default "0x00000" if ETRAXFS
+	default "0x00000000" if !ETRAXFS
+	help
+	  This is a bitmask (18 bits) with information of what bits in PB
+	  that a user can change direction on using ioctl's.
+	  Bit set = changeable.
+	  You probably want 0 here, but it depends on your hardware.
+
+config ETRAX_PB_CHANGEABLE_BITS
+	hex "PB user changeable bits mask"
+	depends on ETRAX_GPIO
+	default "0x00000" if ETRAXFS
+	default "0x00000000" if !ETRAXFS
+	help
+	  This is a bitmask (18 bits) with information of what bits in PB
+	  that a user can change the value on using ioctl's.
+	  Bit set = changeable.
+
+config ETRAX_PC_CHANGEABLE_DIR
+	hex "PC user changeable dir mask"
+	depends on ETRAX_GPIO
+	default "0x00000" if ETRAXFS
+	default "0x00000000" if !ETRAXFS
+	help
+	  This is a bitmask (18 bits) with information of what bits in PC
+	  that a user can change direction on using ioctl's.
+	  Bit set = changeable.
+	  You probably want 0 here, but it depends on your hardware.
+
+config ETRAX_PC_CHANGEABLE_BITS
+	hex "PC user changeable bits mask"
+	depends on ETRAX_GPIO
+	default "0x00000" if ETRAXFS
+	default "0x00000000" if ETRAXFS
+	help
+	  This is a bitmask (18 bits) with information of what bits in PC
+	  that a user can change the value on using ioctl's.
+	  Bit set = changeable.
+
+config ETRAX_PD_CHANGEABLE_DIR
+	hex "PD user changeable dir mask"
+	depends on ETRAX_GPIO && ETRAXFS
+	default "0x00000"
+	help
+	  This is a bitmask (18 bits) with information of what bits in PD
+	  that a user can change direction on using ioctl's.
+	  Bit set = changeable.
+	  You probably want 0x00000 here, but it depends on your hardware.
+
+config ETRAX_PD_CHANGEABLE_BITS
+	hex "PD user changeable bits mask"
+	depends on ETRAX_GPIO && ETRAXFS
+	default "0x00000"
+	help
+	  This is a bitmask (18 bits) with information of what bits in PD
+	  that a user can change the value on using ioctl's.
+	  Bit set = changeable.
+
+config ETRAX_PE_CHANGEABLE_DIR
+	hex "PE user changeable dir mask"
+	depends on ETRAX_GPIO && ETRAXFS
+	default "0x00000"
+	help
+	  This is a bitmask (18 bits) with information of what bits in PE
+	  that a user can change direction on using ioctl's.
+	  Bit set = changeable.
+	  You probably want 0x00000 here, but it depends on your hardware.
+
+config ETRAX_PE_CHANGEABLE_BITS
+	hex "PE user changeable bits mask"
+ 	depends on ETRAX_GPIO && ETRAXFS
+	default "0x00000"
+	help
+	  This is a bitmask (18 bits) with information of what bits in PE
+	  that a user can change the value on using ioctl's.
+	  Bit set = changeable.
+
+config ETRAX_PV_CHANGEABLE_DIR
+	hex "PV user changeable dir mask"
+	depends on ETRAX_VIRTUAL_GPIO
+	default "0x0000"
+	help
+	  This is a bitmask (16 bits) with information of what bits in PV
+	  that a user can change direction on using ioctl's.
+	  Bit set = changeable.
+	  You probably want 0x0000 here, but it depends on your hardware.
+
+config ETRAX_PV_CHANGEABLE_BITS
+	hex "PV user changeable bits mask"
+	depends on ETRAX_VIRTUAL_GPIO
+	default "0x0000"
+	help
+	  This is a bitmask (16 bits) with information of what bits in PV
+	  that a user can change the value on using ioctl's.
+	  Bit set = changeable.
+
+config ETRAX_CARDBUS
+        bool "Cardbus support"
+        depends on ETRAX_ARCH_V32
+        help
+	 Enabled the ETRAX Cardbus driver.
+
+config PCI
+       bool
+       depends on ETRAX_CARDBUS
+       default y
+       select HAVE_GENERIC_DMA_COHERENT
+
+config ETRAX_IOP_FW_LOAD
+	tristate "IO-processor hotplug firmware loading support"
+	depends on ETRAX_ARCH_V32
+	select FW_LOADER
+	help
+	  Enables IO-processor hotplug firmware loading support.
+
+config ETRAX_STREAMCOPROC
+	tristate "Stream co-processor driver enabled"
+	depends on ETRAX_ARCH_V32
+	help
+	  This option enables a driver for the stream co-processor
+	  for cryptographic operations.
+
+config ETRAX_MMC_IOP
+	tristate "MMC/SD host driver using IO-processor"
+	depends on ETRAX_ARCH_V32 && MMC
+	help
+	  This option enables the SD/MMC host controller interface.
+	  The host controller is implemented using the built in
+	  IO-Processor. Only the SPU is used in this implementation.
+
+config ETRAX_SPI_MMC
+# Make this one of several "choices" (possible simultaneously but
+# suggested uniquely) when an IOP driver emerges for "real" MMC/SD
+# protocol support.
+	tristate
+	depends on !ETRAX_MMC_IOP
+	default MMC
+	select SPI
+	select MMC_SPI
+
+# While the board info is MMC_SPI only, the drivers are written to be
+# independent of MMC_SPI, so we'll keep SPI non-dependent on the
+# MMC_SPI config choices (well, except for a single depends-on-line
+# for the board-info file until a separate non-MMC SPI board file
+# emerges).
+# FIXME: When that happens, we'll need to be able to ask for and
+# configure non-MMC SPI ports together with MMC_SPI ports (if multiple
+# SPI ports are enabled).
+
+config SPI_ETRAX_SSER
+	tristate
+	depends on SPI_MASTER && ETRAX_ARCH_V32
+	select SPI_BITBANG
+	help
+	  This enables using an synchronous serial (sser) port as a
+	  SPI master controller on Axis ETRAX FS and later.  The
+	  driver can be configured to use any sser port.
+
+config SPI_ETRAX_GPIO
+	tristate
+	depends on SPI_MASTER && ETRAX_ARCH_V32
+	select SPI_BITBANG
+	help
+	  This enables using GPIO pins port as a SPI master controller
+	  on Axis ETRAX FS and later.  The driver can be configured to
+	  use any GPIO pins.
+
+config ETRAX_SPI_SSER0
+	tristate "SPI using synchronous serial port 0 (sser0)"
+	depends on ETRAX_SPI_MMC
+	default m if MMC_SPI=m
+	default y if MMC_SPI=y
+	default y if MMC_SPI=n
+	select SPI_ETRAX_SSER
+	help
+	  Say Y for an MMC/SD socket connected to synchronous serial port 0,
+	  or for devices using the SPI protocol on that port.  Say m if you
+	  want to build it as a module, which will be named spi_crisv32_sser.
+	  (You need to select MMC separately.)
+
+config ETRAX_SPI_SSER1
+	tristate "SPI using synchronous serial port 1 (sser1)"
+	depends on ETRAX_SPI_MMC
+	default m if MMC_SPI=m && ETRAX_SPI_SSER0=n
+	default y if MMC_SPI=y && ETRAX_SPI_SSER0=n
+	default y if MMC_SPI=n && ETRAX_SPI_SSER0=n
+	select SPI_ETRAX_SSER
+	help
+	  Say Y for an MMC/SD socket connected to synchronous serial port 1,
+	  or for devices using the SPI protocol on that port.  Say m if you
+	  want to build it as a module, which will be named spi_crisv32_sser.
+	  (You need to select MMC separately.)
+
+config ETRAX_SPI_GPIO
+	tristate "Bitbanged SPI using gpio pins"
+	depends on ETRAX_SPI_MMC
+	select SPI_ETRAX_GPIO
+	default m if MMC_SPI=m && ETRAX_SPI_SSER0=n && ETRAX_SPI_SSER1=n
+	default y if MMC_SPI=y && ETRAX_SPI_SSER0=n && ETRAX_SPI_SSER1=n
+	default y if MMC_SPI=n && ETRAX_SPI_SSER0=n && ETRAX_SPI_SSER1=n
+	help
+	  Say Y for an MMC/SD socket connected to general I/O pins (but not
+	  a complete synchronous serial ports), or for devices using the SPI
+	  protocol on general I/O pins.  Slow and slows down the system.
+	  Say m to build it as a module, which will be called spi_crisv32_gpio.
+	  (You need to select MMC separately.)
+
+endif
diff --git a/kernel/arch/cris/arch-v32/drivers/Makefile b/kernel/arch/cris/arch-v32/drivers/Makefile
new file mode 100644
index 000000000..15fbfefce
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/Makefile
@@ -0,0 +1,12 @@
+#
+# Makefile for Etrax-specific drivers
+#
+
+obj-$(CONFIG_ETRAX_STREAMCOPROC)        += cryptocop.o
+obj-$(CONFIG_ETRAX_AXISFLASHMAP)        += axisflashmap.o
+obj-$(CONFIG_ETRAXFS)                   += mach-fs/
+obj-$(CONFIG_CRIS_MACH_ARTPEC3)         += mach-a3/
+obj-$(CONFIG_ETRAX_IOP_FW_LOAD)         += iop_fw_load.o
+obj-$(CONFIG_ETRAX_I2C)			+= i2c.o
+obj-$(CONFIG_ETRAX_SYNCHRONOUS_SERIAL)	+= sync_serial.o
+obj-$(CONFIG_PCI)			+= pci/
diff --git a/kernel/arch/cris/arch-v32/drivers/axisflashmap.c b/kernel/arch/cris/arch-v32/drivers/axisflashmap.c
new file mode 100644
index 000000000..28dd77144
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/axisflashmap.c
@@ -0,0 +1,619 @@
+/*
+ * Physical mapping layer for MTD using the Axis partitiontable format
+ *
+ * Copyright (c) 2001-2007 Axis Communications AB
+ *
+ * This file is under the GPL.
+ *
+ * First partition is always sector 0 regardless of if we find a partitiontable
+ * or not. In the start of the next sector, there can be a partitiontable that
+ * tells us what other partitions to define. If there isn't, we use a default
+ * partition split defined below.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+
+#include <linux/mtd/concat.h>
+#include <linux/mtd/map.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/mtdram.h>
+#include <linux/mtd/partitions.h>
+
+#include <asm/axisflashmap.h>
+#include <asm/mmu.h>
+
+#define MEM_CSE0_SIZE (0x04000000)
+#define MEM_CSE1_SIZE (0x04000000)
+
+#define FLASH_UNCACHED_ADDR  KSEG_E
+#define FLASH_CACHED_ADDR    KSEG_F
+
+#define PAGESIZE (512)
+
+#if CONFIG_ETRAX_FLASH_BUSWIDTH==1
+#define flash_data __u8
+#elif CONFIG_ETRAX_FLASH_BUSWIDTH==2
+#define flash_data __u16
+#elif CONFIG_ETRAX_FLASH_BUSWIDTH==4
+#define flash_data __u32
+#endif
+
+/* From head.S */
+extern unsigned long romfs_in_flash; /* 1 when romfs_start, _length in flash */
+extern unsigned long romfs_start, romfs_length;
+extern unsigned long nand_boot; /* 1 when booted from nand flash */
+
+struct partition_name {
+	char name[6];
+};
+
+/* The master mtd for the entire flash. */
+struct mtd_info* axisflash_mtd = NULL;
+
+/* Map driver functions. */
+
+static map_word flash_read(struct map_info *map, unsigned long ofs)
+{
+	map_word tmp;
+	tmp.x[0] = *(flash_data *)(map->map_priv_1 + ofs);
+	return tmp;
+}
+
+static void flash_copy_from(struct map_info *map, void *to,
+			    unsigned long from, ssize_t len)
+{
+	memcpy(to, (void *)(map->map_priv_1 + from), len);
+}
+
+static void flash_write(struct map_info *map, map_word d, unsigned long adr)
+{
+	*(flash_data *)(map->map_priv_1 + adr) = (flash_data)d.x[0];
+}
+
+/*
+ * The map for chip select e0.
+ *
+ * We run into tricky coherence situations if we mix cached with uncached
+ * accesses to we only use the uncached version here.
+ *
+ * The size field is the total size where the flash chips may be mapped on the
+ * chip select. MTD probes should find all devices there and it does not matter
+ * if there are unmapped gaps or aliases (mirrors of flash devices). The MTD
+ * probes will ignore them.
+ *
+ * The start address in map_priv_1 is in virtual memory so we cannot use
+ * MEM_CSE0_START but must rely on that FLASH_UNCACHED_ADDR is the start
+ * address of cse0.
+ */
+static struct map_info map_cse0 = {
+	.name = "cse0",
+	.size = MEM_CSE0_SIZE,
+	.bankwidth = CONFIG_ETRAX_FLASH_BUSWIDTH,
+	.read = flash_read,
+	.copy_from = flash_copy_from,
+	.write = flash_write,
+	.map_priv_1 = FLASH_UNCACHED_ADDR
+};
+
+/*
+ * The map for chip select e1.
+ *
+ * If there was a gap between cse0 and cse1, map_priv_1 would get the wrong
+ * address, but there isn't.
+ */
+static struct map_info map_cse1 = {
+	.name = "cse1",
+	.size = MEM_CSE1_SIZE,
+	.bankwidth = CONFIG_ETRAX_FLASH_BUSWIDTH,
+	.read = flash_read,
+	.copy_from = flash_copy_from,
+	.write = flash_write,
+	.map_priv_1 = FLASH_UNCACHED_ADDR + MEM_CSE0_SIZE
+};
+
+#define MAX_PARTITIONS			7
+#ifdef CONFIG_ETRAX_NANDBOOT
+#define NUM_DEFAULT_PARTITIONS		4
+#define DEFAULT_ROOTFS_PARTITION_NO	2
+#define DEFAULT_MEDIA_SIZE              0x2000000 /* 32 megs */
+#else
+#define NUM_DEFAULT_PARTITIONS		3
+#define DEFAULT_ROOTFS_PARTITION_NO	(-1)
+#define DEFAULT_MEDIA_SIZE              0x800000 /* 8 megs */
+#endif
+
+#if (MAX_PARTITIONS < NUM_DEFAULT_PARTITIONS)
+#error MAX_PARTITIONS must be >= than NUM_DEFAULT_PARTITIONS
+#endif
+
+/* Initialize the ones normally used. */
+static struct mtd_partition axis_partitions[MAX_PARTITIONS] = {
+	{
+		.name = "part0",
+		.size = CONFIG_ETRAX_PTABLE_SECTOR,
+		.offset = 0
+	},
+	{
+		.name = "part1",
+		.size = 0,
+		.offset = 0
+	},
+	{
+		.name = "part2",
+		.size = 0,
+		.offset = 0
+	},
+	{
+		.name = "part3",
+		.size = 0,
+		.offset = 0
+	},
+	{
+		.name = "part4",
+		.size = 0,
+		.offset = 0
+	},
+	{
+		.name = "part5",
+		.size = 0,
+		.offset = 0
+	},
+	{
+		.name = "part6",
+		.size = 0,
+		.offset = 0
+	},
+};
+
+
+/* If no partition-table was found, we use this default-set.
+ * Default flash size is 8MB (NOR). CONFIG_ETRAX_PTABLE_SECTOR is most
+ * likely the size of one flash block and "filesystem"-partition needs
+ * to be >=5 blocks to be able to use JFFS.
+ */
+static struct mtd_partition axis_default_partitions[NUM_DEFAULT_PARTITIONS] = {
+	{
+		.name = "boot firmware",
+		.size = CONFIG_ETRAX_PTABLE_SECTOR,
+		.offset = 0
+	},
+	{
+		.name = "kernel",
+		.size = 10 * CONFIG_ETRAX_PTABLE_SECTOR,
+		.offset = CONFIG_ETRAX_PTABLE_SECTOR
+	},
+#define FILESYSTEM_SECTOR (11 * CONFIG_ETRAX_PTABLE_SECTOR)
+#ifdef CONFIG_ETRAX_NANDBOOT
+	{
+		.name = "rootfs",
+		.size = 10 * CONFIG_ETRAX_PTABLE_SECTOR,
+		.offset = FILESYSTEM_SECTOR
+	},
+#undef FILESYSTEM_SECTOR
+#define FILESYSTEM_SECTOR (21 * CONFIG_ETRAX_PTABLE_SECTOR)
+#endif
+	{
+		.name = "rwfs",
+		.size = DEFAULT_MEDIA_SIZE - FILESYSTEM_SECTOR,
+		.offset = FILESYSTEM_SECTOR
+	}
+};
+
+#ifdef CONFIG_ETRAX_AXISFLASHMAP_MTD0WHOLE
+/* Main flash device */
+static struct mtd_partition main_partition = {
+	.name = "main",
+	.size = 0,
+	.offset = 0
+};
+#endif
+
+/* Auxiliary partition if we find another flash */
+static struct mtd_partition aux_partition = {
+	.name = "aux",
+	.size = 0,
+	.offset = 0
+};
+
+/*
+ * Probe a chip select for AMD-compatible (JEDEC) or CFI-compatible flash
+ * chips in that order (because the amd_flash-driver is faster).
+ */
+static struct mtd_info *probe_cs(struct map_info *map_cs)
+{
+	struct mtd_info *mtd_cs = NULL;
+
+	printk(KERN_INFO
+	       "%s: Probing a 0x%08lx bytes large window at 0x%08lx.\n",
+	       map_cs->name, map_cs->size, map_cs->map_priv_1);
+
+#ifdef CONFIG_MTD_CFI
+	mtd_cs = do_map_probe("cfi_probe", map_cs);
+#endif
+#ifdef CONFIG_MTD_JEDECPROBE
+	if (!mtd_cs)
+		mtd_cs = do_map_probe("jedec_probe", map_cs);
+#endif
+
+	return mtd_cs;
+}
+
+/*
+ * Probe each chip select individually for flash chips. If there are chips on
+ * both cse0 and cse1, the mtd_info structs will be concatenated to one struct
+ * so that MTD partitions can cross chip boundries.
+ *
+ * The only known restriction to how you can mount your chips is that each
+ * chip select must hold similar flash chips. But you need external hardware
+ * to do that anyway and you can put totally different chips on cse0 and cse1
+ * so it isn't really much of a restriction.
+ */
+extern struct mtd_info* __init crisv32_nand_flash_probe (void);
+static struct mtd_info *flash_probe(void)
+{
+	struct mtd_info *mtd_cse0;
+	struct mtd_info *mtd_cse1;
+	struct mtd_info *mtd_total;
+	struct mtd_info *mtds[2];
+	int count = 0;
+
+	if ((mtd_cse0 = probe_cs(&map_cse0)) != NULL)
+		mtds[count++] = mtd_cse0;
+	if ((mtd_cse1 = probe_cs(&map_cse1)) != NULL)
+		mtds[count++] = mtd_cse1;
+
+	if (!mtd_cse0 && !mtd_cse1) {
+		/* No chip found. */
+		return NULL;
+	}
+
+	if (count > 1) {
+		/* Since the concatenation layer adds a small overhead we
+		 * could try to figure out if the chips in cse0 and cse1 are
+		 * identical and reprobe the whole cse0+cse1 window. But since
+		 * flash chips are slow, the overhead is relatively small.
+		 * So we use the MTD concatenation layer instead of further
+		 * complicating the probing procedure.
+		 */
+		mtd_total = mtd_concat_create(mtds, count, "cse0+cse1");
+		if (!mtd_total) {
+			printk(KERN_ERR "%s and %s: Concatenation failed!\n",
+				map_cse0.name, map_cse1.name);
+
+			/* The best we can do now is to only use what we found
+			 * at cse0. */
+			mtd_total = mtd_cse0;
+			map_destroy(mtd_cse1);
+		}
+	} else
+		mtd_total = mtd_cse0 ? mtd_cse0 : mtd_cse1;
+
+	return mtd_total;
+}
+
+/*
+ * Probe the flash chip(s) and, if it succeeds, read the partition-table
+ * and register the partitions with MTD.
+ */
+static int __init init_axis_flash(void)
+{
+	struct mtd_info *main_mtd;
+	struct mtd_info *aux_mtd = NULL;
+	int err = 0;
+	int pidx = 0;
+	struct partitiontable_head *ptable_head = NULL;
+	struct partitiontable_entry *ptable;
+	int ptable_ok = 0;
+	static char page[PAGESIZE];
+	size_t len;
+	int ram_rootfs_partition = -1; /* -1 => no RAM rootfs partition */
+	int part;
+
+	/* We need a root fs. If it resides in RAM, we need to use an
+	 * MTDRAM device, so it must be enabled in the kernel config,
+	 * but its size must be configured as 0 so as not to conflict
+	 * with our usage.
+	 */
+#if !defined(CONFIG_MTD_MTDRAM) || (CONFIG_MTDRAM_TOTAL_SIZE != 0) || (CONFIG_MTDRAM_ABS_POS != 0)
+	if (!romfs_in_flash && !nand_boot) {
+		printk(KERN_EMERG "axisflashmap: Cannot create an MTD RAM "
+		       "device; configure CONFIG_MTD_MTDRAM with size = 0!\n");
+		panic("This kernel cannot boot from RAM!\n");
+	}
+#endif
+
+	main_mtd = flash_probe();
+	if (main_mtd)
+		printk(KERN_INFO "%s: 0x%08x bytes of NOR flash memory.\n",
+		       main_mtd->name, main_mtd->size);
+
+#ifdef CONFIG_ETRAX_NANDFLASH
+	aux_mtd = crisv32_nand_flash_probe();
+	if (aux_mtd)
+		printk(KERN_INFO "%s: 0x%08x bytes of NAND flash memory.\n",
+			aux_mtd->name, aux_mtd->size);
+
+#ifdef CONFIG_ETRAX_NANDBOOT
+	{
+		struct mtd_info *tmp_mtd;
+
+		printk(KERN_INFO "axisflashmap: Set to boot from NAND flash, "
+		       "making NAND flash primary device.\n");
+		tmp_mtd = main_mtd;
+		main_mtd = aux_mtd;
+		aux_mtd = tmp_mtd;
+	}
+#endif /* CONFIG_ETRAX_NANDBOOT */
+#endif /* CONFIG_ETRAX_NANDFLASH */
+
+	if (!main_mtd && !aux_mtd) {
+		/* There's no reason to use this module if no flash chip can
+		 * be identified. Make sure that's understood.
+		 */
+		printk(KERN_INFO "axisflashmap: Found no flash chip.\n");
+	}
+
+#if 0 /* Dump flash memory so we can see what is going on */
+	if (main_mtd) {
+		int sectoraddr, i;
+		for (sectoraddr = 0; sectoraddr < 2*65536+4096;
+				sectoraddr += PAGESIZE) {
+			main_mtd->read(main_mtd, sectoraddr, PAGESIZE, &len,
+				page);
+			printk(KERN_INFO
+			       "Sector at %d (length %d):\n",
+			       sectoraddr, len);
+			for (i = 0; i < PAGESIZE; i += 16) {
+				printk(KERN_INFO
+				       "%02x %02x %02x %02x "
+				       "%02x %02x %02x %02x "
+				       "%02x %02x %02x %02x "
+				       "%02x %02x %02x %02x\n",
+				       page[i] & 255, page[i+1] & 255,
+				       page[i+2] & 255, page[i+3] & 255,
+				       page[i+4] & 255, page[i+5] & 255,
+				       page[i+6] & 255, page[i+7] & 255,
+				       page[i+8] & 255, page[i+9] & 255,
+				       page[i+10] & 255, page[i+11] & 255,
+				       page[i+12] & 255, page[i+13] & 255,
+				       page[i+14] & 255, page[i+15] & 255);
+			}
+		}
+	}
+#endif
+
+	if (main_mtd) {
+		main_mtd->owner = THIS_MODULE;
+		axisflash_mtd = main_mtd;
+
+		loff_t ptable_sector = CONFIG_ETRAX_PTABLE_SECTOR;
+
+		/* First partition (rescue) is always set to the default. */
+		pidx++;
+#ifdef CONFIG_ETRAX_NANDBOOT
+		/* We know where the partition table should be located,
+		 * it will be in first good block after that.
+		 */
+		int blockstat;
+		do {
+			blockstat = mtd_block_isbad(main_mtd, ptable_sector);
+			if (blockstat < 0)
+				ptable_sector = 0; /* read error */
+			else if (blockstat)
+				ptable_sector += main_mtd->erasesize;
+		} while (blockstat && ptable_sector);
+#endif
+		if (ptable_sector) {
+			mtd_read(main_mtd, ptable_sector, PAGESIZE, &len,
+				 page);
+			ptable_head = &((struct partitiontable *) page)->head;
+		}
+
+#if 0 /* Dump partition table so we can see what is going on */
+		printk(KERN_INFO
+		       "axisflashmap: flash read %d bytes at 0x%08x, data: "
+		       "%02x %02x %02x %02x %02x %02x %02x %02x\n",
+		       len, CONFIG_ETRAX_PTABLE_SECTOR,
+		       page[0] & 255, page[1] & 255,
+		       page[2] & 255, page[3] & 255,
+		       page[4] & 255, page[5] & 255,
+		       page[6] & 255, page[7] & 255);
+		printk(KERN_INFO
+		       "axisflashmap: partition table offset %d, data: "
+		       "%02x %02x %02x %02x %02x %02x %02x %02x\n",
+		       PARTITION_TABLE_OFFSET,
+		       page[PARTITION_TABLE_OFFSET+0] & 255,
+		       page[PARTITION_TABLE_OFFSET+1] & 255,
+		       page[PARTITION_TABLE_OFFSET+2] & 255,
+		       page[PARTITION_TABLE_OFFSET+3] & 255,
+		       page[PARTITION_TABLE_OFFSET+4] & 255,
+		       page[PARTITION_TABLE_OFFSET+5] & 255,
+		       page[PARTITION_TABLE_OFFSET+6] & 255,
+		       page[PARTITION_TABLE_OFFSET+7] & 255);
+#endif
+	}
+
+	if (ptable_head && (ptable_head->magic == PARTITION_TABLE_MAGIC)
+	    && (ptable_head->size <
+		(MAX_PARTITIONS * sizeof(struct partitiontable_entry) +
+		PARTITIONTABLE_END_MARKER_SIZE))
+	    && (*(unsigned long*)((void*)ptable_head + sizeof(*ptable_head) +
+				  ptable_head->size -
+				  PARTITIONTABLE_END_MARKER_SIZE)
+		== PARTITIONTABLE_END_MARKER)) {
+		/* Looks like a start, sane length and end of a
+		 * partition table, lets check csum etc.
+		 */
+		struct partitiontable_entry *max_addr =
+			(struct partitiontable_entry *)
+			((unsigned long)ptable_head + sizeof(*ptable_head) +
+			 ptable_head->size);
+		unsigned long offset = CONFIG_ETRAX_PTABLE_SECTOR;
+		unsigned char *p;
+		unsigned long csum = 0;
+
+		ptable = (struct partitiontable_entry *)
+			((unsigned long)ptable_head + sizeof(*ptable_head));
+
+		/* Lets be PARANOID, and check the checksum. */
+		p = (unsigned char*) ptable;
+
+		while (p <= (unsigned char*)max_addr) {
+			csum += *p++;
+			csum += *p++;
+			csum += *p++;
+			csum += *p++;
+		}
+		ptable_ok = (csum == ptable_head->checksum);
+
+		/* Read the entries and use/show the info.  */
+		printk(KERN_INFO "axisflashmap: "
+		       "Found a%s partition table at 0x%p-0x%p.\n",
+		       (ptable_ok ? " valid" : "n invalid"), ptable_head,
+		       max_addr);
+
+		/* We have found a working bootblock.  Now read the
+		 * partition table.  Scan the table.  It ends with 0xffffffff.
+		 */
+		while (ptable_ok
+		       && ptable->offset != PARTITIONTABLE_END_MARKER
+		       && ptable < max_addr
+		       && pidx < MAX_PARTITIONS - 1) {
+
+			axis_partitions[pidx].offset = offset + ptable->offset;
+#ifdef CONFIG_ETRAX_NANDFLASH
+			if (main_mtd->type == MTD_NANDFLASH) {
+				axis_partitions[pidx].size =
+					(((ptable+1)->offset ==
+					  PARTITIONTABLE_END_MARKER) ?
+					  main_mtd->size :
+					  ((ptable+1)->offset + offset)) -
+					(ptable->offset + offset);
+
+			} else
+#endif /* CONFIG_ETRAX_NANDFLASH */
+				axis_partitions[pidx].size = ptable->size;
+#ifdef CONFIG_ETRAX_NANDBOOT
+			/* Save partition number of jffs2 ro partition.
+			 * Needed if RAM booting or root file system in RAM.
+			 */
+			if (!nand_boot &&
+			    ram_rootfs_partition < 0 && /* not already set */
+			    ptable->type == PARTITION_TYPE_JFFS2 &&
+			    (ptable->flags & PARTITION_FLAGS_READONLY_MASK) ==
+				PARTITION_FLAGS_READONLY)
+				ram_rootfs_partition = pidx;
+#endif /* CONFIG_ETRAX_NANDBOOT */
+			pidx++;
+			ptable++;
+		}
+	}
+
+	/* Decide whether to use default partition table. */
+	/* Only use default table if we actually have a device (main_mtd) */
+
+	struct mtd_partition *partition = &axis_partitions[0];
+	if (main_mtd && !ptable_ok) {
+		memcpy(axis_partitions, axis_default_partitions,
+		       sizeof(axis_default_partitions));
+		pidx = NUM_DEFAULT_PARTITIONS;
+		ram_rootfs_partition = DEFAULT_ROOTFS_PARTITION_NO;
+	}
+
+	/* Add artificial partitions for rootfs if necessary */
+	if (romfs_in_flash) {
+		/* rootfs is in directly accessible flash memory = NOR flash.
+		   Add an overlapping device for the rootfs partition. */
+		printk(KERN_INFO "axisflashmap: Adding partition for "
+		       "overlapping root file system image\n");
+		axis_partitions[pidx].size = romfs_length;
+		axis_partitions[pidx].offset = romfs_start - FLASH_CACHED_ADDR;
+		axis_partitions[pidx].name = "romfs";
+		axis_partitions[pidx].mask_flags |= MTD_WRITEABLE;
+		ram_rootfs_partition = -1;
+		pidx++;
+	} else if (romfs_length && !nand_boot) {
+		/* romfs exists in memory, but not in flash, so must be in RAM.
+		 * Configure an MTDRAM partition. */
+		if (ram_rootfs_partition < 0) {
+			/* None set yet, put it at the end */
+			ram_rootfs_partition = pidx;
+			pidx++;
+		}
+		printk(KERN_INFO "axisflashmap: Adding partition for "
+		       "root file system image in RAM\n");
+		axis_partitions[ram_rootfs_partition].size = romfs_length;
+		axis_partitions[ram_rootfs_partition].offset = romfs_start;
+		axis_partitions[ram_rootfs_partition].name = "romfs";
+		axis_partitions[ram_rootfs_partition].mask_flags |=
+			MTD_WRITEABLE;
+	}
+
+#ifdef CONFIG_ETRAX_AXISFLASHMAP_MTD0WHOLE
+	if (main_mtd) {
+		main_partition.size = main_mtd->size;
+		err = mtd_device_register(main_mtd, &main_partition, 1);
+		if (err)
+			panic("axisflashmap: Could not initialize "
+			      "partition for whole main mtd device!\n");
+	}
+#endif
+
+	/* Now, register all partitions with mtd.
+	 * We do this one at a time so we can slip in an MTDRAM device
+	 * in the proper place if required. */
+
+	for (part = 0; part < pidx; part++) {
+		if (part == ram_rootfs_partition) {
+			/* add MTDRAM partition here */
+			struct mtd_info *mtd_ram;
+
+			mtd_ram = kmalloc(sizeof(struct mtd_info), GFP_KERNEL);
+			if (!mtd_ram)
+				panic("axisflashmap: Couldn't allocate memory "
+				      "for mtd_info!\n");
+			printk(KERN_INFO "axisflashmap: Adding RAM partition "
+			       "for rootfs image.\n");
+			err = mtdram_init_device(mtd_ram,
+						 (void *)partition[part].offset,
+						 partition[part].size,
+						 partition[part].name);
+			if (err)
+				panic("axisflashmap: Could not initialize "
+				      "MTD RAM device!\n");
+			/* JFFS2 likes to have an erasesize. Keep potential
+			 * JFFS2 rootfs happy by providing one. Since image
+			 * was most likely created for main mtd, use that
+			 * erasesize, if available. Otherwise, make a guess. */
+			mtd_ram->erasesize = (main_mtd ? main_mtd->erasesize :
+				CONFIG_ETRAX_PTABLE_SECTOR);
+		} else {
+			err = mtd_device_register(main_mtd, &partition[part],
+						  1);
+			if (err)
+				panic("axisflashmap: Could not add mtd "
+					"partition %d\n", part);
+		}
+	}
+
+	if (aux_mtd) {
+		aux_partition.size = aux_mtd->size;
+		err = mtd_device_register(aux_mtd, &aux_partition, 1);
+		if (err)
+			panic("axisflashmap: Could not initialize "
+			      "aux mtd device!\n");
+
+	}
+
+	return err;
+}
+
+/* This adds the above to the kernels init-call chain. */
+module_init(init_axis_flash);
+
+EXPORT_SYMBOL(axisflash_mtd);
diff --git a/kernel/arch/cris/arch-v32/drivers/cryptocop.c b/kernel/arch/cris/arch-v32/drivers/cryptocop.c
new file mode 100644
index 000000000..877da1908
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/cryptocop.c
@@ -0,0 +1,3536 @@
+/*
+ * Stream co-processor driver for the ETRAX FS
+ *
+ *    Copyright (C) 2003-2007  Axis Communications AB
+ */
+
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/spinlock.h>
+#include <linux/stddef.h>
+
+#include <asm/uaccess.h>
+#include <asm/io.h>
+#include <linux/atomic.h>
+
+#include <linux/list.h>
+#include <linux/interrupt.h>
+
+#include <asm/signal.h>
+#include <asm/irq.h>
+
+#include <dma.h>
+#include <hwregs/dma.h>
+#include <hwregs/reg_map.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/intr_vect_defs.h>
+
+#include <hwregs/strcop.h>
+#include <hwregs/strcop_defs.h>
+#include <cryptocop.h>
+
+#ifdef CONFIG_ETRAXFS
+#define IN_DMA 9
+#define OUT_DMA 8
+#define IN_DMA_INST regi_dma9
+#define OUT_DMA_INST regi_dma8
+#define DMA_IRQ DMA9_INTR_VECT
+#else
+#define IN_DMA 3
+#define OUT_DMA 2
+#define IN_DMA_INST regi_dma3
+#define OUT_DMA_INST regi_dma2
+#define DMA_IRQ DMA3_INTR_VECT
+#endif
+
+#define DESCR_ALLOC_PAD  (31)
+
+struct cryptocop_dma_desc {
+	char *free_buf; /* If non-null will be kfreed in free_cdesc() */
+	dma_descr_data *dma_descr;
+
+	unsigned char dma_descr_buf[sizeof(dma_descr_data) + DESCR_ALLOC_PAD];
+
+	unsigned int from_pool:1; /* If 1 'allocated' from the descriptor pool. */
+	struct cryptocop_dma_desc *next;
+};
+
+
+struct cryptocop_int_operation{
+	void                        *alloc_ptr;
+	cryptocop_session_id        sid;
+
+	dma_descr_context           ctx_out;
+	dma_descr_context           ctx_in;
+
+	/* DMA descriptors allocated by driver. */
+	struct cryptocop_dma_desc   *cdesc_out;
+	struct cryptocop_dma_desc   *cdesc_in;
+
+	/* Strcop config to use. */
+	cryptocop_3des_mode         tdes_mode;
+	cryptocop_csum_type         csum_mode;
+
+	/* DMA descrs provided by consumer. */
+	dma_descr_data              *ddesc_out;
+	dma_descr_data              *ddesc_in;
+};
+
+
+struct cryptocop_tfrm_ctx {
+	cryptocop_tfrm_id tid;
+	unsigned int blocklength;
+
+	unsigned int start_ix;
+
+	struct cryptocop_tfrm_cfg *tcfg;
+	struct cryptocop_transform_ctx *tctx;
+
+	unsigned char previous_src;
+	unsigned char current_src;
+
+	/* Values to use in metadata out. */
+	unsigned char hash_conf;
+	unsigned char hash_mode;
+	unsigned char ciph_conf;
+	unsigned char cbcmode;
+	unsigned char decrypt;
+
+	unsigned int requires_padding:1;
+	unsigned int strict_block_length:1;
+	unsigned int active:1;
+	unsigned int done:1;
+	size_t consumed;
+	size_t produced;
+
+	/* Pad (input) descriptors to put in the DMA out list when the transform
+	 * output is put on the DMA in list. */
+	struct cryptocop_dma_desc *pad_descs;
+
+	struct cryptocop_tfrm_ctx *prev_src;
+	struct cryptocop_tfrm_ctx *curr_src;
+
+	/* Mapping to HW. */
+	unsigned char unit_no;
+};
+
+
+struct cryptocop_private{
+	cryptocop_session_id sid;
+	struct cryptocop_private *next;
+};
+
+/* Session list. */
+
+struct cryptocop_transform_ctx{
+	struct cryptocop_transform_init init;
+	unsigned char dec_key[CRYPTOCOP_MAX_KEY_LENGTH];
+	unsigned int dec_key_set:1;
+
+	struct cryptocop_transform_ctx *next;
+};
+
+
+struct cryptocop_session{
+	cryptocop_session_id sid;
+
+	struct cryptocop_transform_ctx *tfrm_ctx;
+
+	struct cryptocop_session *next;
+};
+
+/* Priority levels for jobs sent to the cryptocop.  Checksum operations from
+   kernel have highest priority since TCPIP stack processing must not
+   be a bottleneck. */
+typedef enum {
+	cryptocop_prio_kernel_csum = 0,
+	cryptocop_prio_kernel = 1,
+	cryptocop_prio_user = 2,
+	cryptocop_prio_no_prios = 3
+} cryptocop_queue_priority;
+
+struct cryptocop_prio_queue{
+	struct list_head jobs;
+	cryptocop_queue_priority prio;
+};
+
+struct cryptocop_prio_job{
+	struct list_head node;
+	cryptocop_queue_priority prio;
+
+	struct cryptocop_operation *oper;
+	struct cryptocop_int_operation *iop;
+};
+
+struct ioctl_job_cb_ctx {
+	unsigned int processed:1;
+};
+
+
+static struct cryptocop_session *cryptocop_sessions = NULL;
+spinlock_t cryptocop_sessions_lock;
+
+/* Next Session ID to assign. */
+static cryptocop_session_id next_sid = 1;
+
+/* Pad for checksum. */
+static const char csum_zero_pad[1] = {0x00};
+
+/* Trash buffer for mem2mem operations. */
+#define MEM2MEM_DISCARD_BUF_LENGTH  (512)
+static unsigned char mem2mem_discard_buf[MEM2MEM_DISCARD_BUF_LENGTH];
+
+/* Descriptor pool. */
+/* FIXME Tweak this value. */
+#define CRYPTOCOP_DESCRIPTOR_POOL_SIZE   (100)
+static struct cryptocop_dma_desc descr_pool[CRYPTOCOP_DESCRIPTOR_POOL_SIZE];
+static struct cryptocop_dma_desc *descr_pool_free_list;
+static int descr_pool_no_free;
+static spinlock_t descr_pool_lock;
+
+/* Lock to stop cryptocop to start processing of a new operation. The holder
+   of this lock MUST call cryptocop_start_job() after it is unlocked. */
+spinlock_t cryptocop_process_lock;
+
+static struct cryptocop_prio_queue cryptocop_job_queues[cryptocop_prio_no_prios];
+static spinlock_t cryptocop_job_queue_lock;
+static struct cryptocop_prio_job *cryptocop_running_job = NULL;
+static spinlock_t running_job_lock;
+
+/* The interrupt handler appends completed jobs to this list. The scehduled
+ * tasklet removes them upon sending the response to the crypto consumer. */
+static struct list_head cryptocop_completed_jobs;
+static spinlock_t cryptocop_completed_jobs_lock;
+
+DECLARE_WAIT_QUEUE_HEAD(cryptocop_ioc_process_wq);
+
+
+/** Local functions. **/
+
+static int cryptocop_open(struct inode *, struct file *);
+
+static int cryptocop_release(struct inode *, struct file *);
+
+static long cryptocop_ioctl(struct file *file,
+			   unsigned int cmd, unsigned long arg);
+
+static void cryptocop_start_job(void);
+
+static int cryptocop_job_queue_insert(cryptocop_queue_priority prio, struct cryptocop_operation *operation);
+static int cryptocop_job_setup(struct cryptocop_prio_job **pj, struct cryptocop_operation *operation);
+
+static int cryptocop_job_queue_init(void);
+static void cryptocop_job_queue_close(void);
+
+static int create_md5_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length);
+
+static int create_sha1_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length);
+
+static int transform_ok(struct cryptocop_transform_init *tinit);
+
+static struct cryptocop_session *get_session(cryptocop_session_id sid);
+
+static struct cryptocop_transform_ctx *get_transform_ctx(struct cryptocop_session *sess, cryptocop_tfrm_id tid);
+
+static void delete_internal_operation(struct cryptocop_int_operation *iop);
+
+static void get_aes_decrypt_key(unsigned char *dec_key, const unsigned  char *key, unsigned int keylength);
+
+static int init_stream_coprocessor(void);
+
+static void __exit exit_stream_coprocessor(void);
+
+/*#define LDEBUG*/
+#ifdef LDEBUG
+#define DEBUG(s) s
+#define DEBUG_API(s) s
+static void print_cryptocop_operation(struct cryptocop_operation *cop);
+static void print_dma_descriptors(struct cryptocop_int_operation *iop);
+static void print_strcop_crypto_op(struct strcop_crypto_op *cop);
+static void print_lock_status(void);
+static void print_user_dma_lists(struct cryptocop_dma_list_operation *dma_op);
+#define assert(s) do{if (!(s)) panic(#s);} while(0);
+#else
+#define DEBUG(s)
+#define DEBUG_API(s)
+#define assert(s)
+#endif
+
+
+/* Transform constants. */
+#define DES_BLOCK_LENGTH   (8)
+#define AES_BLOCK_LENGTH   (16)
+#define MD5_BLOCK_LENGTH   (64)
+#define SHA1_BLOCK_LENGTH  (64)
+#define CSUM_BLOCK_LENGTH  (2)
+#define MD5_STATE_LENGTH   (16)
+#define SHA1_STATE_LENGTH  (20)
+
+/* The device number. */
+#define CRYPTOCOP_MAJOR    (254)
+#define CRYPTOCOP_MINOR    (0)
+
+
+
+const struct file_operations cryptocop_fops = {
+	.owner		= THIS_MODULE,
+	.open		= cryptocop_open,
+	.release	= cryptocop_release,
+	.unlocked_ioctl = cryptocop_ioctl,
+	.llseek		= noop_llseek,
+};
+
+
+static void free_cdesc(struct cryptocop_dma_desc *cdesc)
+{
+	DEBUG(printk("free_cdesc: cdesc 0x%p, from_pool=%d\n", cdesc, cdesc->from_pool));
+	kfree(cdesc->free_buf);
+
+	if (cdesc->from_pool) {
+		unsigned long int flags;
+		spin_lock_irqsave(&descr_pool_lock, flags);
+		cdesc->next = descr_pool_free_list;
+		descr_pool_free_list = cdesc;
+		++descr_pool_no_free;
+		spin_unlock_irqrestore(&descr_pool_lock, flags);
+	} else {
+		kfree(cdesc);
+	}
+}
+
+
+static struct cryptocop_dma_desc *alloc_cdesc(int alloc_flag)
+{
+	int use_pool = (alloc_flag & GFP_ATOMIC) ? 1 : 0;
+	struct cryptocop_dma_desc *cdesc;
+
+	if (use_pool) {
+		unsigned long int flags;
+		spin_lock_irqsave(&descr_pool_lock, flags);
+		if (!descr_pool_free_list) {
+			spin_unlock_irqrestore(&descr_pool_lock, flags);
+			DEBUG_API(printk("alloc_cdesc: pool is empty\n"));
+			return NULL;
+		}
+		cdesc = descr_pool_free_list;
+		descr_pool_free_list = descr_pool_free_list->next;
+		--descr_pool_no_free;
+		spin_unlock_irqrestore(&descr_pool_lock, flags);
+		cdesc->from_pool = 1;
+	} else {
+		cdesc = kmalloc(sizeof(struct cryptocop_dma_desc), alloc_flag);
+		if (!cdesc) {
+			DEBUG_API(printk("alloc_cdesc: kmalloc\n"));
+			return NULL;
+		}
+		cdesc->from_pool = 0;
+	}
+	cdesc->dma_descr = (dma_descr_data*)(((unsigned long int)cdesc + offsetof(struct cryptocop_dma_desc, dma_descr_buf) + DESCR_ALLOC_PAD) & ~0x0000001F);
+
+	cdesc->next = NULL;
+
+	cdesc->free_buf = NULL;
+	cdesc->dma_descr->out_eop = 0;
+	cdesc->dma_descr->in_eop = 0;
+	cdesc->dma_descr->intr = 0;
+	cdesc->dma_descr->eol = 0;
+	cdesc->dma_descr->wait = 0;
+	cdesc->dma_descr->buf = NULL;
+	cdesc->dma_descr->after = NULL;
+
+	DEBUG_API(printk("alloc_cdesc: return 0x%p, cdesc->dma_descr=0x%p, from_pool=%d\n", cdesc, cdesc->dma_descr, cdesc->from_pool));
+	return cdesc;
+}
+
+
+static void setup_descr_chain(struct cryptocop_dma_desc *cd)
+{
+	DEBUG(printk("setup_descr_chain: entering\n"));
+	while (cd) {
+		if (cd->next) {
+			cd->dma_descr->next = (dma_descr_data*)virt_to_phys(cd->next->dma_descr);
+		} else {
+			cd->dma_descr->next = NULL;
+		}
+		cd = cd->next;
+	}
+	DEBUG(printk("setup_descr_chain: exit\n"));
+}
+
+
+/* Create a pad descriptor for the transform.
+ * Return -1 for error, 0 if pad created. */
+static int create_pad_descriptor(struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **pad_desc, int alloc_flag)
+{
+	struct cryptocop_dma_desc        *cdesc = NULL;
+	int                              error = 0;
+	struct strcop_meta_out           mo = {
+		.ciphsel = src_none,
+		.hashsel = src_none,
+		.csumsel = src_none
+	};
+	char                             *pad;
+	size_t                           plen;
+
+	DEBUG(printk("create_pad_descriptor: start.\n"));
+	/* Setup pad descriptor. */
+
+	DEBUG(printk("create_pad_descriptor: setting up padding.\n"));
+	cdesc = alloc_cdesc(alloc_flag);
+	if (!cdesc){
+		DEBUG_API(printk("create_pad_descriptor: alloc pad desc\n"));
+		goto error_cleanup;
+	}
+	switch (tc->unit_no) {
+	case src_md5:
+		error = create_md5_pad(alloc_flag, tc->consumed, &pad, &plen);
+		if (error){
+			DEBUG_API(printk("create_pad_descriptor: create_md5_pad_failed\n"));
+			goto error_cleanup;
+		}
+		cdesc->free_buf = pad;
+		mo.hashsel = src_dma;
+		mo.hashconf = tc->hash_conf;
+		mo.hashmode = tc->hash_mode;
+		break;
+	case src_sha1:
+		error = create_sha1_pad(alloc_flag, tc->consumed, &pad, &plen);
+		if (error){
+			DEBUG_API(printk("create_pad_descriptor: create_sha1_pad_failed\n"));
+			goto error_cleanup;
+		}
+		cdesc->free_buf = pad;
+		mo.hashsel = src_dma;
+		mo.hashconf = tc->hash_conf;
+		mo.hashmode = tc->hash_mode;
+		break;
+	case src_csum:
+		if (tc->consumed % tc->blocklength){
+			pad = (char*)csum_zero_pad;
+			plen = 1;
+		} else {
+			pad = (char*)cdesc; /* Use any pointer. */
+			plen = 0;
+		}
+		mo.csumsel = src_dma;
+		break;
+	}
+	cdesc->dma_descr->wait = 1;
+	cdesc->dma_descr->out_eop = 1; /* Since this is a pad output is pushed.  EOP is ok here since the padded unit is the only one active. */
+	cdesc->dma_descr->buf = (char*)virt_to_phys((char*)pad);
+	cdesc->dma_descr->after = cdesc->dma_descr->buf + plen;
+
+	cdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, mo);
+	*pad_desc = cdesc;
+
+	return 0;
+
+ error_cleanup:
+	if (cdesc) free_cdesc(cdesc);
+	return -1;
+}
+
+
+static int setup_key_dl_desc(struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **kd, int alloc_flag)
+{
+	struct cryptocop_dma_desc  *key_desc = alloc_cdesc(alloc_flag);
+	struct strcop_meta_out     mo = {0};
+
+	DEBUG(printk("setup_key_dl_desc\n"));
+
+	if (!key_desc) {
+		DEBUG_API(printk("setup_key_dl_desc: failed descriptor allocation.\n"));
+		return -ENOMEM;
+	}
+
+	/* Download key. */
+	if ((tc->tctx->init.alg == cryptocop_alg_aes) && (tc->tcfg->flags & CRYPTOCOP_DECRYPT)) {
+		/* Precook the AES decrypt key. */
+		if (!tc->tctx->dec_key_set){
+			get_aes_decrypt_key(tc->tctx->dec_key, tc->tctx->init.key, tc->tctx->init.keylen);
+			tc->tctx->dec_key_set = 1;
+		}
+		key_desc->dma_descr->buf = (char*)virt_to_phys(tc->tctx->dec_key);
+		key_desc->dma_descr->after = key_desc->dma_descr->buf + tc->tctx->init.keylen/8;
+	} else {
+		key_desc->dma_descr->buf = (char*)virt_to_phys(tc->tctx->init.key);
+		key_desc->dma_descr->after = key_desc->dma_descr->buf + tc->tctx->init.keylen/8;
+	}
+	/* Setup metadata. */
+	mo.dlkey = 1;
+	switch (tc->tctx->init.keylen) {
+	case 64:
+		mo.decrypt = 0;
+		mo.hashmode = 0;
+		break;
+	case 128:
+		mo.decrypt = 0;
+		mo.hashmode = 1;
+		break;
+	case 192:
+		mo.decrypt = 1;
+		mo.hashmode = 0;
+		break;
+	case 256:
+		mo.decrypt = 1;
+		mo.hashmode = 1;
+		break;
+	default:
+		break;
+	}
+	mo.ciphsel = mo.hashsel = mo.csumsel = src_none;
+	key_desc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, mo);
+
+	key_desc->dma_descr->out_eop = 1;
+	key_desc->dma_descr->wait = 1;
+	key_desc->dma_descr->intr = 0;
+
+	*kd = key_desc;
+	return 0;
+}
+
+static int setup_cipher_iv_desc(struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **id, int alloc_flag)
+{
+	struct cryptocop_dma_desc  *iv_desc = alloc_cdesc(alloc_flag);
+	struct strcop_meta_out     mo = {0};
+
+	DEBUG(printk("setup_cipher_iv_desc\n"));
+
+	if (!iv_desc) {
+		DEBUG_API(printk("setup_cipher_iv_desc: failed CBC IV descriptor allocation.\n"));
+		return -ENOMEM;
+	}
+	/* Download IV. */
+	iv_desc->dma_descr->buf = (char*)virt_to_phys(tc->tcfg->iv);
+	iv_desc->dma_descr->after = iv_desc->dma_descr->buf + tc->blocklength;
+
+	/* Setup metadata. */
+	mo.hashsel = mo.csumsel = src_none;
+	mo.ciphsel = src_dma;
+	mo.ciphconf = tc->ciph_conf;
+	mo.cbcmode = tc->cbcmode;
+
+	iv_desc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, mo);
+
+	iv_desc->dma_descr->out_eop = 0;
+	iv_desc->dma_descr->wait = 1;
+	iv_desc->dma_descr->intr = 0;
+
+	*id = iv_desc;
+	return 0;
+}
+
+/* Map the ouput length of the transform to operation output starting on the inject index. */
+static int create_input_descriptors(struct cryptocop_operation *operation, struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **id, int alloc_flag)
+{
+	int                        err = 0;
+	struct cryptocop_dma_desc  head = {0};
+	struct cryptocop_dma_desc  *outdesc = &head;
+	size_t                     iov_offset = 0;
+	size_t                     out_ix = 0;
+	int                        outiov_ix = 0;
+	struct strcop_meta_in      mi = {0};
+
+	size_t                     out_length = tc->produced;
+	int                        rem_length;
+	int                        dlength;
+
+	assert(out_length != 0);
+	if (((tc->produced + tc->tcfg->inject_ix) > operation->tfrm_op.outlen) || (tc->produced && (operation->tfrm_op.outlen == 0))) {
+		DEBUG_API(printk("create_input_descriptors: operation outdata too small\n"));
+		return -EINVAL;
+	}
+	/* Traverse the out iovec until the result inject index is reached. */
+	while ((outiov_ix < operation->tfrm_op.outcount) && ((out_ix + operation->tfrm_op.outdata[outiov_ix].iov_len) <= tc->tcfg->inject_ix)){
+		out_ix += operation->tfrm_op.outdata[outiov_ix].iov_len;
+		outiov_ix++;
+	}
+	if (outiov_ix >= operation->tfrm_op.outcount){
+		DEBUG_API(printk("create_input_descriptors: operation outdata too small\n"));
+		return -EINVAL;
+	}
+	iov_offset = tc->tcfg->inject_ix - out_ix;
+	mi.dmasel = tc->unit_no;
+
+	/* Setup the output descriptors. */
+	while ((out_length > 0) && (outiov_ix < operation->tfrm_op.outcount)) {
+		outdesc->next = alloc_cdesc(alloc_flag);
+		if (!outdesc->next) {
+			DEBUG_API(printk("create_input_descriptors: alloc_cdesc\n"));
+			err = -ENOMEM;
+			goto error_cleanup;
+		}
+		outdesc = outdesc->next;
+		rem_length = operation->tfrm_op.outdata[outiov_ix].iov_len - iov_offset;
+		dlength = (out_length < rem_length) ? out_length : rem_length;
+
+		DEBUG(printk("create_input_descriptors:\n"
+			     "outiov_ix=%d, rem_length=%d, dlength=%d\n"
+			     "iov_offset=%d, outdata[outiov_ix].iov_len=%d\n"
+			     "outcount=%d, outiov_ix=%d\n",
+			     outiov_ix, rem_length, dlength, iov_offset, operation->tfrm_op.outdata[outiov_ix].iov_len, operation->tfrm_op.outcount, outiov_ix));
+
+		outdesc->dma_descr->buf = (char*)virt_to_phys(operation->tfrm_op.outdata[outiov_ix].iov_base + iov_offset);
+		outdesc->dma_descr->after = outdesc->dma_descr->buf + dlength;
+		outdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi);
+
+		out_length -= dlength;
+		iov_offset += dlength;
+		if (iov_offset >= operation->tfrm_op.outdata[outiov_ix].iov_len) {
+			iov_offset = 0;
+			++outiov_ix;
+		}
+	}
+	if (out_length > 0){
+		DEBUG_API(printk("create_input_descriptors: not enough room for output, %d remained\n", out_length));
+		err = -EINVAL;
+		goto error_cleanup;
+	}
+	/* Set sync in last descriptor. */
+	mi.sync = 1;
+	outdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi);
+
+	*id = head.next;
+	return 0;
+
+ error_cleanup:
+	while (head.next) {
+		outdesc = head.next->next;
+		free_cdesc(head.next);
+		head.next = outdesc;
+	}
+	return err;
+}
+
+
+static int create_output_descriptors(struct cryptocop_operation *operation, int *iniov_ix, int *iniov_offset, size_t desc_len, struct cryptocop_dma_desc **current_out_cdesc, struct strcop_meta_out *meta_out, int alloc_flag)
+{
+	while (desc_len != 0) {
+		struct cryptocop_dma_desc  *cdesc;
+		int                        rem_length = operation->tfrm_op.indata[*iniov_ix].iov_len - *iniov_offset;
+		int                        dlength = (desc_len < rem_length) ? desc_len : rem_length;
+
+		cdesc = alloc_cdesc(alloc_flag);
+		if (!cdesc) {
+			DEBUG_API(printk("create_output_descriptors: alloc_cdesc\n"));
+			return -ENOMEM;
+		}
+		(*current_out_cdesc)->next = cdesc;
+		(*current_out_cdesc) = cdesc;
+
+		cdesc->free_buf = NULL;
+
+		cdesc->dma_descr->buf = (char*)virt_to_phys(operation->tfrm_op.indata[*iniov_ix].iov_base + *iniov_offset);
+		cdesc->dma_descr->after = cdesc->dma_descr->buf + dlength;
+
+		assert(desc_len >= dlength);
+		desc_len -= dlength;
+		*iniov_offset += dlength;
+		if (*iniov_offset >= operation->tfrm_op.indata[*iniov_ix].iov_len) {
+			*iniov_offset = 0;
+			++(*iniov_ix);
+			if (*iniov_ix > operation->tfrm_op.incount) {
+				DEBUG_API(printk("create_output_descriptors: not enough indata in operation."));
+				return  -EINVAL;
+			}
+		}
+		cdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, (*meta_out));
+	} /* while (desc_len != 0) */
+	/* Last DMA descriptor gets a 'wait' bit to signal expected change in metadata. */
+	(*current_out_cdesc)->dma_descr->wait = 1; /* This will set extraneous WAIT in some situations, e.g. when padding hashes and checksums. */
+
+	return 0;
+}
+
+
+static int append_input_descriptors(struct cryptocop_operation *operation, struct cryptocop_dma_desc **current_in_cdesc, struct cryptocop_dma_desc **current_out_cdesc, struct cryptocop_tfrm_ctx *tc, int alloc_flag)
+{
+	DEBUG(printk("append_input_descriptors, tc=0x%p, unit_no=%d\n", tc, tc->unit_no));
+	if (tc->tcfg) {
+		int                        failed = 0;
+		struct cryptocop_dma_desc  *idescs = NULL;
+		DEBUG(printk("append_input_descriptors: pushing output, consumed %d produced %d bytes.\n", tc->consumed, tc->produced));
+		if (tc->pad_descs) {
+			DEBUG(printk("append_input_descriptors: append pad descriptors to DMA out list.\n"));
+			while (tc->pad_descs) {
+				DEBUG(printk("append descriptor 0x%p\n", tc->pad_descs));
+				(*current_out_cdesc)->next = tc->pad_descs;
+				tc->pad_descs = tc->pad_descs->next;
+				(*current_out_cdesc) = (*current_out_cdesc)->next;
+			}
+		}
+
+		/* Setup and append output descriptors to DMA in list. */
+		if (tc->unit_no == src_dma){
+			/* mem2mem.  Setup DMA in descriptors to discard all input prior to the requested mem2mem data. */
+			struct strcop_meta_in mi = {.sync = 0, .dmasel = src_dma};
+			unsigned int start_ix = tc->start_ix;
+			while (start_ix){
+				unsigned int desclen = start_ix < MEM2MEM_DISCARD_BUF_LENGTH ? start_ix : MEM2MEM_DISCARD_BUF_LENGTH;
+				(*current_in_cdesc)->next = alloc_cdesc(alloc_flag);
+				if (!(*current_in_cdesc)->next){
+					DEBUG_API(printk("append_input_descriptors: alloc_cdesc mem2mem discard failed\n"));
+					return -ENOMEM;
+				}
+				(*current_in_cdesc) = (*current_in_cdesc)->next;
+				(*current_in_cdesc)->dma_descr->buf = (char*)virt_to_phys(mem2mem_discard_buf);
+				(*current_in_cdesc)->dma_descr->after = (*current_in_cdesc)->dma_descr->buf + desclen;
+				(*current_in_cdesc)->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi);
+				start_ix -= desclen;
+			}
+			mi.sync = 1;
+			(*current_in_cdesc)->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi);
+		}
+
+		failed = create_input_descriptors(operation, tc, &idescs, alloc_flag);
+		if (failed){
+			DEBUG_API(printk("append_input_descriptors: output descriptor setup failed\n"));
+			return failed;
+		}
+		DEBUG(printk("append_input_descriptors: append output descriptors to DMA in list.\n"));
+		while (idescs) {
+			DEBUG(printk("append descriptor 0x%p\n", idescs));
+			(*current_in_cdesc)->next = idescs;
+			idescs = idescs->next;
+			(*current_in_cdesc) = (*current_in_cdesc)->next;
+		}
+	}
+	return 0;
+}
+
+
+
+static int cryptocop_setup_dma_list(struct cryptocop_operation *operation, struct cryptocop_int_operation **int_op, int alloc_flag)
+{
+	struct cryptocop_session *sess;
+	struct cryptocop_transform_ctx *tctx;
+
+	struct cryptocop_tfrm_ctx digest_ctx = {
+		.previous_src = src_none,
+		.current_src = src_none,
+		.start_ix = 0,
+		.requires_padding = 1,
+		.strict_block_length = 0,
+		.hash_conf = 0,
+		.hash_mode = 0,
+		.ciph_conf = 0,
+		.cbcmode = 0,
+		.decrypt = 0,
+		.consumed = 0,
+		.produced = 0,
+		.pad_descs = NULL,
+		.active = 0,
+		.done = 0,
+		.prev_src = NULL,
+		.curr_src = NULL,
+		.tcfg = NULL};
+	struct cryptocop_tfrm_ctx cipher_ctx = {
+		.previous_src = src_none,
+		.current_src = src_none,
+		.start_ix = 0,
+		.requires_padding = 0,
+		.strict_block_length = 1,
+		.hash_conf = 0,
+		.hash_mode = 0,
+		.ciph_conf = 0,
+		.cbcmode = 0,
+		.decrypt = 0,
+		.consumed = 0,
+		.produced = 0,
+		.pad_descs = NULL,
+		.active = 0,
+		.done = 0,
+		.prev_src = NULL,
+		.curr_src = NULL,
+		.tcfg = NULL};
+	struct cryptocop_tfrm_ctx csum_ctx = {
+		.previous_src = src_none,
+		.current_src = src_none,
+		.start_ix = 0,
+		.blocklength = 2,
+		.requires_padding = 1,
+		.strict_block_length = 0,
+		.hash_conf = 0,
+		.hash_mode = 0,
+		.ciph_conf = 0,
+		.cbcmode = 0,
+		.decrypt = 0,
+		.consumed = 0,
+		.produced = 0,
+		.pad_descs = NULL,
+		.active = 0,
+		.done = 0,
+		.tcfg = NULL,
+		.prev_src = NULL,
+		.curr_src = NULL,
+		.unit_no = src_csum};
+	struct cryptocop_tfrm_cfg *tcfg = operation->tfrm_op.tfrm_cfg;
+
+	unsigned int indata_ix = 0;
+
+	/* iovec accounting. */
+	int iniov_ix = 0;
+	int iniov_offset = 0;
+
+	/* Operation descriptor cfg traversal pointer. */
+	struct cryptocop_desc *odsc;
+
+	int failed = 0;
+	/* List heads for allocated descriptors. */
+	struct cryptocop_dma_desc out_cdesc_head = {0};
+	struct cryptocop_dma_desc in_cdesc_head = {0};
+
+	struct cryptocop_dma_desc *current_out_cdesc = &out_cdesc_head;
+	struct cryptocop_dma_desc *current_in_cdesc = &in_cdesc_head;
+
+	struct cryptocop_tfrm_ctx *output_tc = NULL;
+	void                      *iop_alloc_ptr;
+
+	assert(operation != NULL);
+	assert(int_op != NULL);
+
+	DEBUG(printk("cryptocop_setup_dma_list: start\n"));
+	DEBUG(print_cryptocop_operation(operation));
+
+	sess = get_session(operation->sid);
+	if (!sess) {
+		DEBUG_API(printk("cryptocop_setup_dma_list: no session found for operation.\n"));
+		failed = -EINVAL;
+		goto error_cleanup;
+	}
+	iop_alloc_ptr = kmalloc(DESCR_ALLOC_PAD + sizeof(struct cryptocop_int_operation), alloc_flag);
+	if (!iop_alloc_ptr) {
+		DEBUG_API(printk("cryptocop_setup_dma_list:  kmalloc cryptocop_int_operation\n"));
+		failed = -ENOMEM;
+		goto error_cleanup;
+	}
+	(*int_op) = (struct cryptocop_int_operation*)(((unsigned long int)(iop_alloc_ptr + DESCR_ALLOC_PAD + offsetof(struct cryptocop_int_operation, ctx_out)) & ~0x0000001F) - offsetof(struct cryptocop_int_operation, ctx_out));
+	DEBUG(memset((*int_op), 0xff, sizeof(struct cryptocop_int_operation)));
+	(*int_op)->alloc_ptr = iop_alloc_ptr;
+	DEBUG(printk("cryptocop_setup_dma_list: *int_op=0x%p, alloc_ptr=0x%p\n", *int_op, (*int_op)->alloc_ptr));
+
+	(*int_op)->sid = operation->sid;
+	(*int_op)->cdesc_out = NULL;
+	(*int_op)->cdesc_in = NULL;
+	(*int_op)->tdes_mode = cryptocop_3des_ede;
+	(*int_op)->csum_mode = cryptocop_csum_le;
+	(*int_op)->ddesc_out = NULL;
+	(*int_op)->ddesc_in = NULL;
+
+	/* Scan operation->tfrm_op.tfrm_cfg for bad configuration and set up the local contexts. */
+	if (!tcfg) {
+		DEBUG_API(printk("cryptocop_setup_dma_list: no configured transforms in operation.\n"));
+		failed = -EINVAL;
+		goto error_cleanup;
+	}
+	while (tcfg) {
+		tctx = get_transform_ctx(sess, tcfg->tid);
+		if (!tctx) {
+			DEBUG_API(printk("cryptocop_setup_dma_list: no transform id %d in session.\n", tcfg->tid));
+			failed = -EINVAL;
+			goto error_cleanup;
+		}
+		if (tcfg->inject_ix > operation->tfrm_op.outlen){
+			DEBUG_API(printk("cryptocop_setup_dma_list: transform id %d inject_ix (%d) > operation->tfrm_op.outlen(%d)", tcfg->tid, tcfg->inject_ix, operation->tfrm_op.outlen));
+			failed = -EINVAL;
+			goto error_cleanup;
+		}
+		switch (tctx->init.alg){
+		case cryptocop_alg_mem2mem:
+			if (cipher_ctx.tcfg != NULL){
+				DEBUG_API(printk("cryptocop_setup_dma_list: multiple ciphers in operation.\n"));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			/* mem2mem is handled as a NULL cipher. */
+			cipher_ctx.cbcmode = 0;
+			cipher_ctx.decrypt = 0;
+			cipher_ctx.blocklength = 1;
+			cipher_ctx.ciph_conf = 0;
+			cipher_ctx.unit_no = src_dma;
+			cipher_ctx.tcfg = tcfg;
+			cipher_ctx.tctx = tctx;
+			break;
+		case cryptocop_alg_des:
+		case cryptocop_alg_3des:
+		case cryptocop_alg_aes:
+			/* cipher */
+			if (cipher_ctx.tcfg != NULL){
+				DEBUG_API(printk("cryptocop_setup_dma_list: multiple ciphers in operation.\n"));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			cipher_ctx.tcfg = tcfg;
+			cipher_ctx.tctx = tctx;
+			if (cipher_ctx.tcfg->flags & CRYPTOCOP_DECRYPT){
+				cipher_ctx.decrypt = 1;
+			}
+			switch (tctx->init.cipher_mode) {
+			case cryptocop_cipher_mode_ecb:
+				cipher_ctx.cbcmode = 0;
+				break;
+			case cryptocop_cipher_mode_cbc:
+				cipher_ctx.cbcmode = 1;
+				break;
+			default:
+				DEBUG_API(printk("cryptocop_setup_dma_list: cipher_ctx, bad cipher mode==%d\n", tctx->init.cipher_mode));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			DEBUG(printk("cryptocop_setup_dma_list: cipher_ctx, set CBC mode==%d\n", cipher_ctx.cbcmode));
+			switch (tctx->init.alg){
+			case cryptocop_alg_des:
+				cipher_ctx.ciph_conf = 0;
+				cipher_ctx.unit_no = src_des;
+				cipher_ctx.blocklength = DES_BLOCK_LENGTH;
+				break;
+			case cryptocop_alg_3des:
+				cipher_ctx.ciph_conf = 1;
+				cipher_ctx.unit_no = src_des;
+				cipher_ctx.blocklength = DES_BLOCK_LENGTH;
+				break;
+			case cryptocop_alg_aes:
+				cipher_ctx.ciph_conf = 2;
+				cipher_ctx.unit_no = src_aes;
+				cipher_ctx.blocklength = AES_BLOCK_LENGTH;
+				break;
+			default:
+				panic("cryptocop_setup_dma_list: impossible algorithm %d\n", tctx->init.alg);
+			}
+			(*int_op)->tdes_mode = tctx->init.tdes_mode;
+			break;
+		case cryptocop_alg_md5:
+		case cryptocop_alg_sha1:
+			/* digest */
+			if (digest_ctx.tcfg != NULL){
+				DEBUG_API(printk("cryptocop_setup_dma_list: multiple digests in operation.\n"));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			digest_ctx.tcfg = tcfg;
+			digest_ctx.tctx = tctx;
+			digest_ctx.hash_mode = 0; /* Don't use explicit IV in this API. */
+			switch (tctx->init.alg){
+			case cryptocop_alg_md5:
+				digest_ctx.blocklength = MD5_BLOCK_LENGTH;
+				digest_ctx.unit_no = src_md5;
+				digest_ctx.hash_conf = 1; /* 1 => MD-5 */
+				break;
+			case cryptocop_alg_sha1:
+				digest_ctx.blocklength = SHA1_BLOCK_LENGTH;
+				digest_ctx.unit_no = src_sha1;
+				digest_ctx.hash_conf = 0; /* 0 => SHA-1 */
+				break;
+			default:
+				panic("cryptocop_setup_dma_list: impossible digest algorithm\n");
+			}
+			break;
+		case cryptocop_alg_csum:
+			/* digest */
+			if (csum_ctx.tcfg != NULL){
+				DEBUG_API(printk("cryptocop_setup_dma_list: multiple checksums in operation.\n"));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			(*int_op)->csum_mode = tctx->init.csum_mode;
+			csum_ctx.tcfg = tcfg;
+			csum_ctx.tctx = tctx;
+			break;
+		default:
+			/* no algorithm. */
+			DEBUG_API(printk("cryptocop_setup_dma_list: invalid algorithm %d specified in tfrm %d.\n", tctx->init.alg, tcfg->tid));
+			failed = -EINVAL;
+			goto error_cleanup;
+		}
+		tcfg = tcfg->next;
+	}
+	/* Download key if a cipher is used. */
+	if (cipher_ctx.tcfg && (cipher_ctx.tctx->init.alg != cryptocop_alg_mem2mem)){
+		struct cryptocop_dma_desc  *key_desc = NULL;
+
+		failed = setup_key_dl_desc(&cipher_ctx, &key_desc, alloc_flag);
+		if (failed) {
+			DEBUG_API(printk("cryptocop_setup_dma_list: setup key dl\n"));
+			goto error_cleanup;
+		}
+		current_out_cdesc->next = key_desc;
+		current_out_cdesc = key_desc;
+		indata_ix += (unsigned int)(key_desc->dma_descr->after - key_desc->dma_descr->buf);
+
+		/* Download explicit IV if a cipher is used and CBC mode and explicit IV selected. */
+		if ((cipher_ctx.tctx->init.cipher_mode == cryptocop_cipher_mode_cbc) && (cipher_ctx.tcfg->flags & CRYPTOCOP_EXPLICIT_IV)) {
+			struct cryptocop_dma_desc  *iv_desc = NULL;
+
+			DEBUG(printk("cryptocop_setup_dma_list: setup cipher CBC IV descriptor.\n"));
+
+			failed = setup_cipher_iv_desc(&cipher_ctx, &iv_desc, alloc_flag);
+			if (failed) {
+				DEBUG_API(printk("cryptocop_setup_dma_list: CBC IV descriptor.\n"));
+				goto error_cleanup;
+			}
+			current_out_cdesc->next = iv_desc;
+			current_out_cdesc = iv_desc;
+			indata_ix += (unsigned int)(iv_desc->dma_descr->after - iv_desc->dma_descr->buf);
+		}
+	}
+
+	/* Process descriptors. */
+	odsc = operation->tfrm_op.desc;
+	while (odsc) {
+		struct cryptocop_desc_cfg   *dcfg = odsc->cfg;
+		struct strcop_meta_out      meta_out = {0};
+		size_t                      desc_len = odsc->length;
+		int                         active_count, eop_needed_count;
+
+		output_tc = NULL;
+
+		DEBUG(printk("cryptocop_setup_dma_list: parsing an operation descriptor\n"));
+
+		while (dcfg) {
+			struct cryptocop_tfrm_ctx  *tc = NULL;
+
+			DEBUG(printk("cryptocop_setup_dma_list: parsing an operation descriptor configuration.\n"));
+			/* Get the local context for the transform and mark it as the output unit if it produces output. */
+			if (digest_ctx.tcfg && (digest_ctx.tcfg->tid == dcfg->tid)){
+				tc = &digest_ctx;
+			} else if (cipher_ctx.tcfg && (cipher_ctx.tcfg->tid == dcfg->tid)){
+				tc = &cipher_ctx;
+			} else if (csum_ctx.tcfg && (csum_ctx.tcfg->tid == dcfg->tid)){
+				tc = &csum_ctx;
+			}
+			if (!tc) {
+				DEBUG_API(printk("cryptocop_setup_dma_list: invalid transform %d specified in descriptor.\n", dcfg->tid));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			if (tc->done) {
+				DEBUG_API(printk("cryptocop_setup_dma_list: completed transform %d reused.\n", dcfg->tid));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			if (!tc->active) {
+				tc->start_ix = indata_ix;
+				tc->active = 1;
+			}
+
+			tc->previous_src = tc->current_src;
+			tc->prev_src = tc->curr_src;
+			/* Map source unit id to DMA source config. */
+			switch (dcfg->src){
+			case cryptocop_source_dma:
+				tc->current_src = src_dma;
+				break;
+			case cryptocop_source_des:
+				tc->current_src = src_des;
+				break;
+			case cryptocop_source_3des:
+				tc->current_src = src_des;
+				break;
+			case cryptocop_source_aes:
+				tc->current_src = src_aes;
+				break;
+			case cryptocop_source_md5:
+			case cryptocop_source_sha1:
+			case cryptocop_source_csum:
+			case cryptocop_source_none:
+			default:
+				/* We do not allow using accumulating style units (SHA-1, MD5, checksum) as sources to other units.
+				 */
+				DEBUG_API(printk("cryptocop_setup_dma_list: bad unit source configured %d.\n", dcfg->src));
+				failed = -EINVAL;
+				goto error_cleanup;
+			}
+			if (tc->current_src != src_dma) {
+				/* Find the unit we are sourcing from. */
+				if (digest_ctx.unit_no == tc->current_src){
+					tc->curr_src = &digest_ctx;
+				} else if (cipher_ctx.unit_no == tc->current_src){
+					tc->curr_src = &cipher_ctx;
+				} else if (csum_ctx.unit_no == tc->current_src){
+					tc->curr_src = &csum_ctx;
+				}
+				if ((tc->curr_src == tc) && (tc->unit_no != src_dma)){
+					DEBUG_API(printk("cryptocop_setup_dma_list: unit %d configured to source from itself.\n", tc->unit_no));
+					failed = -EINVAL;
+					goto error_cleanup;
+				}
+			} else {
+				tc->curr_src = NULL;
+			}
+
+			/* Detect source switch. */
+			DEBUG(printk("cryptocop_setup_dma_list: tc->active=%d tc->unit_no=%d tc->current_src=%d tc->previous_src=%d, tc->curr_src=0x%p, tc->prev_srv=0x%p\n", tc->active, tc->unit_no, tc->current_src, tc->previous_src, tc->curr_src, tc->prev_src));
+			if (tc->active && (tc->current_src != tc->previous_src)) {
+				/* Only allow source switch when both the old source unit and the new one have
+				 * no pending data to process (i.e. the consumed length must be a multiple of the
+				 * transform blocklength). */
+				/* Note: if the src == NULL we are actually sourcing from DMA out. */
+				if (((tc->prev_src != NULL) && (tc->prev_src->consumed % tc->prev_src->blocklength)) ||
+				    ((tc->curr_src != NULL) && (tc->curr_src->consumed % tc->curr_src->blocklength)))
+				{
+					DEBUG_API(printk("cryptocop_setup_dma_list: can only disconnect from or connect to a unit on a multiple of the blocklength, old: cons=%d, prod=%d, block=%d, new: cons=%d prod=%d, block=%d.\n", tc->prev_src ? tc->prev_src->consumed : INT_MIN, tc->prev_src ? tc->prev_src->produced : INT_MIN, tc->prev_src ? tc->prev_src->blocklength : INT_MIN, tc->curr_src ? tc->curr_src->consumed : INT_MIN, tc->curr_src ? tc->curr_src->produced : INT_MIN, tc->curr_src ? tc->curr_src->blocklength : INT_MIN));
+					failed = -EINVAL;
+					goto error_cleanup;
+				}
+			}
+			/* Detect unit deactivation. */
+			if (dcfg->last) {
+				/* Length check of this is handled below. */
+				tc->done = 1;
+			}
+			dcfg = dcfg->next;
+		} /* while (dcfg) */
+		DEBUG(printk("cryptocop_setup_dma_list: parsing operation descriptor configuration complete.\n"));
+
+		if (cipher_ctx.active && (cipher_ctx.curr_src != NULL) && !cipher_ctx.curr_src->active){
+			DEBUG_API(printk("cryptocop_setup_dma_list: cipher source from inactive unit %d\n", cipher_ctx.curr_src->unit_no));
+			failed = -EINVAL;
+			goto error_cleanup;
+		}
+		if (digest_ctx.active && (digest_ctx.curr_src != NULL) && !digest_ctx.curr_src->active){
+			DEBUG_API(printk("cryptocop_setup_dma_list: digest source from inactive unit %d\n", digest_ctx.curr_src->unit_no));
+			failed = -EINVAL;
+			goto error_cleanup;
+		}
+		if (csum_ctx.active && (csum_ctx.curr_src != NULL) && !csum_ctx.curr_src->active){
+			DEBUG_API(printk("cryptocop_setup_dma_list: cipher source from inactive unit %d\n", csum_ctx.curr_src->unit_no));
+			failed = -EINVAL;
+			goto error_cleanup;
+		}
+
+		/* Update consumed and produced lengths.
+
+		   The consumed length accounting here is actually cheating.  If a unit source from DMA (or any
+		   other unit that process data in blocks of one octet) it is correct, but if it source from a
+		   block processing unit, i.e. a cipher, it will be temporarily incorrect at some times.  However
+		   since it is only allowed--by the HW--to change source to or from a block processing unit at times where that
+		   unit has processed an exact multiple of its block length the end result will be correct.
+		   Beware that if the source change restriction change this code will need to be (much) reworked.
+		*/
+		DEBUG(printk("cryptocop_setup_dma_list: desc->length=%d, desc_len=%d.\n", odsc->length, desc_len));
+
+		if (csum_ctx.active) {
+			csum_ctx.consumed += desc_len;
+			if (csum_ctx.done) {
+				csum_ctx.produced = 2;
+			}
+			DEBUG(printk("cryptocop_setup_dma_list: csum_ctx producing: consumed=%d, produced=%d, blocklength=%d.\n", csum_ctx.consumed, csum_ctx.produced, csum_ctx.blocklength));
+		}
+		if (digest_ctx.active) {
+			digest_ctx.consumed += desc_len;
+			if (digest_ctx.done) {
+				if (digest_ctx.unit_no == src_md5) {
+					digest_ctx.produced = MD5_STATE_LENGTH;
+				} else {
+					digest_ctx.produced = SHA1_STATE_LENGTH;
+				}
+			}
+			DEBUG(printk("cryptocop_setup_dma_list: digest_ctx producing: consumed=%d, produced=%d, blocklength=%d.\n", digest_ctx.consumed, digest_ctx.produced, digest_ctx.blocklength));
+		}
+		if (cipher_ctx.active) {
+			/* Ciphers are allowed only to source from DMA out.  That is filtered above. */
+			assert(cipher_ctx.current_src == src_dma);
+			cipher_ctx.consumed += desc_len;
+			cipher_ctx.produced = cipher_ctx.blocklength * (cipher_ctx.consumed / cipher_ctx.blocklength);
+			if (cipher_ctx.cbcmode && !(cipher_ctx.tcfg->flags & CRYPTOCOP_EXPLICIT_IV) && cipher_ctx.produced){
+				cipher_ctx.produced -= cipher_ctx.blocklength; /* Compensate for CBC iv. */
+			}
+			DEBUG(printk("cryptocop_setup_dma_list: cipher_ctx producing: consumed=%d, produced=%d, blocklength=%d.\n", cipher_ctx.consumed, cipher_ctx.produced, cipher_ctx.blocklength));
+		}
+
+		/* Setup the DMA out descriptors. */
+		/* Configure the metadata. */
+		active_count = 0;
+		eop_needed_count = 0;
+		if (cipher_ctx.active) {
+			++active_count;
+			if (cipher_ctx.unit_no == src_dma){
+				/* mem2mem */
+				meta_out.ciphsel = src_none;
+			} else {
+				meta_out.ciphsel = cipher_ctx.current_src;
+			}
+			meta_out.ciphconf = cipher_ctx.ciph_conf;
+			meta_out.cbcmode = cipher_ctx.cbcmode;
+			meta_out.decrypt = cipher_ctx.decrypt;
+			DEBUG(printk("set ciphsel=%d ciphconf=%d cbcmode=%d decrypt=%d\n", meta_out.ciphsel, meta_out.ciphconf, meta_out.cbcmode, meta_out.decrypt));
+			if (cipher_ctx.done) ++eop_needed_count;
+		} else {
+			meta_out.ciphsel = src_none;
+		}
+
+		if (digest_ctx.active) {
+			++active_count;
+			meta_out.hashsel = digest_ctx.current_src;
+			meta_out.hashconf = digest_ctx.hash_conf;
+			meta_out.hashmode = 0; /* Explicit mode is not used here. */
+			DEBUG(printk("set hashsel=%d hashconf=%d hashmode=%d\n", meta_out.hashsel, meta_out.hashconf, meta_out.hashmode));
+			if (digest_ctx.done) {
+				assert(digest_ctx.pad_descs == NULL);
+				failed = create_pad_descriptor(&digest_ctx, &digest_ctx.pad_descs, alloc_flag);
+				if (failed) {
+					DEBUG_API(printk("cryptocop_setup_dma_list: failed digest pad creation.\n"));
+					goto error_cleanup;
+				}
+			}
+		} else {
+			meta_out.hashsel = src_none;
+		}
+
+		if (csum_ctx.active) {
+			++active_count;
+			meta_out.csumsel = csum_ctx.current_src;
+			if (csum_ctx.done) {
+				assert(csum_ctx.pad_descs == NULL);
+				failed = create_pad_descriptor(&csum_ctx, &csum_ctx.pad_descs, alloc_flag);
+				if (failed) {
+					DEBUG_API(printk("cryptocop_setup_dma_list: failed csum pad creation.\n"));
+					goto error_cleanup;
+				}
+			}
+		} else {
+			meta_out.csumsel = src_none;
+		}
+		DEBUG(printk("cryptocop_setup_dma_list: %d eop needed, %d active units\n", eop_needed_count, active_count));
+		/* Setup DMA out descriptors for the indata. */
+		failed = create_output_descriptors(operation, &iniov_ix, &iniov_offset, desc_len, &current_out_cdesc, &meta_out, alloc_flag);
+		if (failed) {
+			DEBUG_API(printk("cryptocop_setup_dma_list: create_output_descriptors %d\n", failed));
+			goto error_cleanup;
+		}
+		/* Setup out EOP.  If there are active units that are not done here they cannot get an EOP
+		 * so we ust setup a zero length descriptor to DMA to signal EOP only to done units.
+		 * If there is a pad descriptor EOP for the padded unit will be EOPed by it.
+		 */
+		assert(active_count >= eop_needed_count);
+		assert((eop_needed_count == 0) || (eop_needed_count == 1));
+		if (eop_needed_count) {
+			/* This means that the bulk operation (cipeher/m2m) is terminated. */
+			if (active_count > 1) {
+				/* Use zero length EOP descriptor. */
+				struct cryptocop_dma_desc *ed = alloc_cdesc(alloc_flag);
+				struct strcop_meta_out    ed_mo = {0};
+				if (!ed) {
+					DEBUG_API(printk("cryptocop_setup_dma_list: alloc EOP descriptor for cipher\n"));
+					failed = -ENOMEM;
+					goto error_cleanup;
+				}
+
+				assert(cipher_ctx.active && cipher_ctx.done);
+
+				if (cipher_ctx.unit_no == src_dma){
+					/* mem2mem */
+					ed_mo.ciphsel = src_none;
+				} else {
+					ed_mo.ciphsel = cipher_ctx.current_src;
+				}
+				ed_mo.ciphconf = cipher_ctx.ciph_conf;
+				ed_mo.cbcmode = cipher_ctx.cbcmode;
+				ed_mo.decrypt = cipher_ctx.decrypt;
+
+				ed->free_buf = NULL;
+				ed->dma_descr->wait = 1;
+				ed->dma_descr->out_eop = 1;
+
+				ed->dma_descr->buf = (char*)virt_to_phys(&ed); /* Use any valid physical address for zero length descriptor. */
+				ed->dma_descr->after = ed->dma_descr->buf;
+				ed->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, ed_mo);
+				current_out_cdesc->next = ed;
+				current_out_cdesc = ed;
+			} else {
+				/* Set EOP in the current out descriptor since the only active module is
+				 * the one needing the EOP. */
+
+				current_out_cdesc->dma_descr->out_eop = 1;
+			}
+		}
+
+		if (cipher_ctx.done && cipher_ctx.active) cipher_ctx.active = 0;
+		if (digest_ctx.done && digest_ctx.active) digest_ctx.active = 0;
+		if (csum_ctx.done && csum_ctx.active) csum_ctx.active = 0;
+		indata_ix += odsc->length;
+		odsc = odsc->next;
+	} /* while (odsc) */ /* Process descriptors. */
+	DEBUG(printk("cryptocop_setup_dma_list: done parsing operation descriptors\n"));
+	if (cipher_ctx.tcfg && (cipher_ctx.active || !cipher_ctx.done)){
+		DEBUG_API(printk("cryptocop_setup_dma_list: cipher operation not terminated.\n"));
+		failed = -EINVAL;
+		goto error_cleanup;
+	}
+	if (digest_ctx.tcfg && (digest_ctx.active || !digest_ctx.done)){
+		DEBUG_API(printk("cryptocop_setup_dma_list: digest operation not terminated.\n"));
+		failed = -EINVAL;
+		goto error_cleanup;
+	}
+	if (csum_ctx.tcfg && (csum_ctx.active || !csum_ctx.done)){
+		DEBUG_API(printk("cryptocop_setup_dma_list: csum operation not terminated.\n"));
+		failed = -EINVAL;
+		goto error_cleanup;
+	}
+
+	failed = append_input_descriptors(operation, &current_in_cdesc, &current_out_cdesc, &cipher_ctx, alloc_flag);
+	if (failed){
+		DEBUG_API(printk("cryptocop_setup_dma_list: append_input_descriptors cipher_ctx %d\n", failed));
+		goto error_cleanup;
+	}
+	failed = append_input_descriptors(operation, &current_in_cdesc, &current_out_cdesc, &digest_ctx, alloc_flag);
+	if (failed){
+		DEBUG_API(printk("cryptocop_setup_dma_list: append_input_descriptors cipher_ctx %d\n", failed));
+		goto error_cleanup;
+	}
+	failed = append_input_descriptors(operation, &current_in_cdesc, &current_out_cdesc, &csum_ctx, alloc_flag);
+	if (failed){
+		DEBUG_API(printk("cryptocop_setup_dma_list: append_input_descriptors cipher_ctx %d\n", failed));
+		goto error_cleanup;
+	}
+
+	DEBUG(printk("cryptocop_setup_dma_list: int_op=0x%p, *int_op=0x%p\n", int_op, *int_op));
+	(*int_op)->cdesc_out = out_cdesc_head.next;
+	(*int_op)->cdesc_in = in_cdesc_head.next;
+	DEBUG(printk("cryptocop_setup_dma_list: out_cdesc_head=0x%p in_cdesc_head=0x%p\n", (*int_op)->cdesc_out, (*int_op)->cdesc_in));
+
+	setup_descr_chain(out_cdesc_head.next);
+	setup_descr_chain(in_cdesc_head.next);
+
+	/* Last but not least: mark the last DMA in descriptor for a INTR and EOL and the the
+	 * last DMA out descriptor for EOL.
+	 */
+	current_in_cdesc->dma_descr->intr = 1;
+	current_in_cdesc->dma_descr->eol = 1;
+	current_out_cdesc->dma_descr->eol = 1;
+
+	/* Setup DMA contexts. */
+	(*int_op)->ctx_out.next = NULL;
+	(*int_op)->ctx_out.eol = 1;
+	(*int_op)->ctx_out.intr = 0;
+	(*int_op)->ctx_out.store_mode = 0;
+	(*int_op)->ctx_out.en = 0;
+	(*int_op)->ctx_out.dis = 0;
+	(*int_op)->ctx_out.md0 = 0;
+	(*int_op)->ctx_out.md1 = 0;
+	(*int_op)->ctx_out.md2 = 0;
+	(*int_op)->ctx_out.md3 = 0;
+	(*int_op)->ctx_out.md4 = 0;
+	(*int_op)->ctx_out.saved_data = (dma_descr_data*)virt_to_phys((*int_op)->cdesc_out->dma_descr);
+	(*int_op)->ctx_out.saved_data_buf = (*int_op)->cdesc_out->dma_descr->buf; /* Already physical address. */
+
+	(*int_op)->ctx_in.next = NULL;
+	(*int_op)->ctx_in.eol = 1;
+	(*int_op)->ctx_in.intr = 0;
+	(*int_op)->ctx_in.store_mode = 0;
+	(*int_op)->ctx_in.en = 0;
+	(*int_op)->ctx_in.dis = 0;
+	(*int_op)->ctx_in.md0 = 0;
+	(*int_op)->ctx_in.md1 = 0;
+	(*int_op)->ctx_in.md2 = 0;
+	(*int_op)->ctx_in.md3 = 0;
+	(*int_op)->ctx_in.md4 = 0;
+
+	(*int_op)->ctx_in.saved_data = (dma_descr_data*)virt_to_phys((*int_op)->cdesc_in->dma_descr);
+	(*int_op)->ctx_in.saved_data_buf = (*int_op)->cdesc_in->dma_descr->buf; /* Already physical address. */
+
+	DEBUG(printk("cryptocop_setup_dma_list: done\n"));
+	return 0;
+
+error_cleanup:
+	{
+		/* Free all allocated resources. */
+		struct cryptocop_dma_desc *tmp_cdesc;
+		while (digest_ctx.pad_descs){
+			tmp_cdesc = digest_ctx.pad_descs->next;
+			free_cdesc(digest_ctx.pad_descs);
+			digest_ctx.pad_descs = tmp_cdesc;
+		}
+		while (csum_ctx.pad_descs){
+			tmp_cdesc = csum_ctx.pad_descs->next;
+			free_cdesc(csum_ctx.pad_descs);
+			csum_ctx.pad_descs = tmp_cdesc;
+		}
+		assert(cipher_ctx.pad_descs == NULL); /* The ciphers are never padded. */
+
+		if (*int_op != NULL) delete_internal_operation(*int_op);
+	}
+	DEBUG_API(printk("cryptocop_setup_dma_list: done with error %d\n", failed));
+	return failed;
+}
+
+
+static void delete_internal_operation(struct cryptocop_int_operation *iop)
+{
+	void                      *ptr = iop->alloc_ptr;
+	struct cryptocop_dma_desc *cd = iop->cdesc_out;
+	struct cryptocop_dma_desc *next;
+
+	DEBUG(printk("delete_internal_operation: iop=0x%p, alloc_ptr=0x%p\n", iop, ptr));
+
+	while (cd) {
+		next = cd->next;
+		free_cdesc(cd);
+		cd = next;
+	}
+	cd = iop->cdesc_in;
+	while (cd) {
+		next = cd->next;
+		free_cdesc(cd);
+		cd = next;
+	}
+	kfree(ptr);
+}
+
+#define MD5_MIN_PAD_LENGTH (9)
+#define MD5_PAD_LENGTH_FIELD_LENGTH (8)
+
+static int create_md5_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length)
+{
+	size_t                  padlen = MD5_BLOCK_LENGTH - (hashed_length % MD5_BLOCK_LENGTH);
+	unsigned char           *p;
+	int                     i;
+	unsigned long long int  bit_length = hashed_length << 3;
+
+	if (padlen < MD5_MIN_PAD_LENGTH) padlen += MD5_BLOCK_LENGTH;
+
+	p = kzalloc(padlen, alloc_flag);
+	if (!p) return -ENOMEM;
+
+	*p = 0x80;
+
+	DEBUG(printk("create_md5_pad: hashed_length=%lld bits == %lld bytes\n", bit_length, hashed_length));
+
+	i = padlen - MD5_PAD_LENGTH_FIELD_LENGTH;
+	while (bit_length != 0){
+		p[i++] = bit_length % 0x100;
+		bit_length >>= 8;
+	}
+
+	*pad = (char*)p;
+	*pad_length = padlen;
+
+	return 0;
+}
+
+#define SHA1_MIN_PAD_LENGTH (9)
+#define SHA1_PAD_LENGTH_FIELD_LENGTH (8)
+
+static int create_sha1_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length)
+{
+	size_t                  padlen = SHA1_BLOCK_LENGTH - (hashed_length % SHA1_BLOCK_LENGTH);
+	unsigned char           *p;
+	int                     i;
+	unsigned long long int  bit_length = hashed_length << 3;
+
+	if (padlen < SHA1_MIN_PAD_LENGTH) padlen += SHA1_BLOCK_LENGTH;
+
+	p = kzalloc(padlen, alloc_flag);
+	if (!p) return -ENOMEM;
+
+	*p = 0x80;
+
+	DEBUG(printk("create_sha1_pad: hashed_length=%lld bits == %lld bytes\n", bit_length, hashed_length));
+
+	i = padlen - 1;
+	while (bit_length != 0){
+		p[i--] = bit_length % 0x100;
+		bit_length >>= 8;
+	}
+
+	*pad = (char*)p;
+	*pad_length = padlen;
+
+	return 0;
+}
+
+
+static int transform_ok(struct cryptocop_transform_init *tinit)
+{
+	switch (tinit->alg){
+	case cryptocop_alg_csum:
+		switch (tinit->csum_mode){
+		case cryptocop_csum_le:
+		case cryptocop_csum_be:
+			break;
+		default:
+			DEBUG_API(printk("transform_ok: Bad mode set for csum transform\n"));
+			return -EINVAL;
+		}
+	case cryptocop_alg_mem2mem:
+	case cryptocop_alg_md5:
+	case cryptocop_alg_sha1:
+		if (tinit->keylen != 0) {
+			DEBUG_API(printk("transform_ok: non-zero keylength, %d, for a digest/csum algorithm\n", tinit->keylen));
+			return -EINVAL; /* This check is a bit strict. */
+		}
+		break;
+	case cryptocop_alg_des:
+		if (tinit->keylen != 64) {
+			DEBUG_API(printk("transform_ok: keylen %d invalid for DES\n", tinit->keylen));
+			return -EINVAL;
+		}
+		break;
+	case cryptocop_alg_3des:
+		if (tinit->keylen != 192) {
+			DEBUG_API(printk("transform_ok: keylen %d invalid for 3DES\n", tinit->keylen));
+			return -EINVAL;
+		}
+		break;
+	case cryptocop_alg_aes:
+		if (tinit->keylen != 128 && tinit->keylen != 192 && tinit->keylen != 256) {
+			DEBUG_API(printk("transform_ok: keylen %d invalid for AES\n", tinit->keylen));
+			return -EINVAL;
+		}
+		break;
+	case cryptocop_no_alg:
+	default:
+		DEBUG_API(printk("transform_ok: no such algorithm %d\n", tinit->alg));
+		return -EINVAL;
+	}
+
+	switch (tinit->alg){
+	case cryptocop_alg_des:
+	case cryptocop_alg_3des:
+	case cryptocop_alg_aes:
+		if (tinit->cipher_mode != cryptocop_cipher_mode_ecb && tinit->cipher_mode != cryptocop_cipher_mode_cbc) return -EINVAL;
+	default:
+		 break;
+	}
+	return 0;
+}
+
+
+int cryptocop_new_session(cryptocop_session_id *sid, struct cryptocop_transform_init *tinit, int alloc_flag)
+{
+	struct cryptocop_session         *sess;
+	struct cryptocop_transform_init  *tfrm_in = tinit;
+	struct cryptocop_transform_init  *tmp_in;
+	int                              no_tfrms = 0;
+	int                              i;
+	unsigned long int                flags;
+
+	init_stream_coprocessor(); /* For safety if we are called early */
+
+	while (tfrm_in){
+		int err;
+		++no_tfrms;
+		if ((err = transform_ok(tfrm_in))) {
+			DEBUG_API(printk("cryptocop_new_session, bad transform\n"));
+			return err;
+		}
+		tfrm_in = tfrm_in->next;
+	}
+	if (0 == no_tfrms) {
+		DEBUG_API(printk("cryptocop_new_session, no transforms specified\n"));
+		return -EINVAL;
+	}
+
+	sess = kmalloc(sizeof(struct cryptocop_session), alloc_flag);
+	if (!sess){
+		DEBUG_API(printk("cryptocop_new_session, kmalloc cryptocop_session\n"));
+		return -ENOMEM;
+	}
+
+	sess->tfrm_ctx = kmalloc(no_tfrms * sizeof(struct cryptocop_transform_ctx), alloc_flag);
+	if (!sess->tfrm_ctx) {
+		DEBUG_API(printk("cryptocop_new_session, kmalloc cryptocop_transform_ctx\n"));
+		kfree(sess);
+		return -ENOMEM;
+	}
+
+	tfrm_in = tinit;
+	for (i = 0; i < no_tfrms; i++){
+		tmp_in = tfrm_in->next;
+		while (tmp_in){
+			if (tmp_in->tid == tfrm_in->tid) {
+				DEBUG_API(printk("cryptocop_new_session, duplicate transform ids\n"));
+				kfree(sess->tfrm_ctx);
+				kfree(sess);
+				return -EINVAL;
+			}
+			tmp_in = tmp_in->next;
+		}
+		memcpy(&sess->tfrm_ctx[i].init, tfrm_in, sizeof(struct cryptocop_transform_init));
+		sess->tfrm_ctx[i].dec_key_set = 0;
+		sess->tfrm_ctx[i].next = &sess->tfrm_ctx[i] + 1;
+
+		tfrm_in = tfrm_in->next;
+	}
+	sess->tfrm_ctx[i-1].next = NULL;
+
+	spin_lock_irqsave(&cryptocop_sessions_lock, flags);
+	sess->sid = next_sid;
+	next_sid++;
+	/* TODO If we are really paranoid we should do duplicate check to handle sid wraparound.
+	 *      OTOH 2^64 is a really large number of session. */
+	if (next_sid == 0) next_sid = 1;
+
+	/* Prepend to session list. */
+	sess->next = cryptocop_sessions;
+	cryptocop_sessions = sess;
+	spin_unlock_irqrestore(&cryptocop_sessions_lock, flags);
+	*sid = sess->sid;
+	return 0;
+}
+
+
+int cryptocop_free_session(cryptocop_session_id sid)
+{
+	struct cryptocop_transform_ctx    *tc;
+	struct cryptocop_session          *sess = NULL;
+	struct cryptocop_session          *psess = NULL;
+	unsigned long int                 flags;
+	int                               i;
+	LIST_HEAD(remove_list);
+	struct list_head                  *node, *tmp;
+	struct cryptocop_prio_job         *pj;
+
+	DEBUG(printk("cryptocop_free_session: sid=%lld\n", sid));
+
+	spin_lock_irqsave(&cryptocop_sessions_lock, flags);
+	sess = cryptocop_sessions;
+	while (sess && sess->sid != sid){
+		psess = sess;
+		sess = sess->next;
+	}
+	if (sess){
+		if (psess){
+			psess->next = sess->next;
+		} else {
+			cryptocop_sessions = sess->next;
+		}
+	}
+	spin_unlock_irqrestore(&cryptocop_sessions_lock, flags);
+
+	if (!sess) return -EINVAL;
+
+	/* Remove queued jobs. */
+	spin_lock_irqsave(&cryptocop_job_queue_lock, flags);
+
+	for (i = 0; i < cryptocop_prio_no_prios; i++){
+		if (!list_empty(&(cryptocop_job_queues[i].jobs))){
+			list_for_each_safe(node, tmp, &(cryptocop_job_queues[i].jobs)) {
+				pj = list_entry(node, struct cryptocop_prio_job, node);
+				if (pj->oper->sid == sid) {
+					list_move_tail(node, &remove_list);
+				}
+			}
+		}
+	}
+	spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags);
+
+	list_for_each_safe(node, tmp, &remove_list) {
+		list_del(node);
+		pj = list_entry(node, struct cryptocop_prio_job, node);
+		pj->oper->operation_status = -EAGAIN;  /* EAGAIN is not ideal for job/session terminated but it's the best choice I know of. */
+		DEBUG(printk("cryptocop_free_session: pj=0x%p, pj->oper=0x%p, pj->iop=0x%p\n", pj, pj->oper, pj->iop));
+		pj->oper->cb(pj->oper, pj->oper->cb_data);
+		delete_internal_operation(pj->iop);
+		kfree(pj);
+	}
+
+	tc = sess->tfrm_ctx;
+	/* Erase keying data. */
+	while (tc){
+		DEBUG(printk("cryptocop_free_session: memset keys, tfrm id=%d\n", tc->init.tid));
+		memset(tc->init.key, 0xff, CRYPTOCOP_MAX_KEY_LENGTH);
+		memset(tc->dec_key, 0xff, CRYPTOCOP_MAX_KEY_LENGTH);
+		tc = tc->next;
+	}
+	kfree(sess->tfrm_ctx);
+	kfree(sess);
+
+	return 0;
+}
+
+static struct cryptocop_session *get_session(cryptocop_session_id sid)
+{
+	struct cryptocop_session    *sess;
+	unsigned long int           flags;
+
+	spin_lock_irqsave(&cryptocop_sessions_lock, flags);
+	sess = cryptocop_sessions;
+	while (sess && (sess->sid != sid)){
+		sess = sess->next;
+	}
+	spin_unlock_irqrestore(&cryptocop_sessions_lock, flags);
+
+	return sess;
+}
+
+static struct cryptocop_transform_ctx *get_transform_ctx(struct cryptocop_session *sess, cryptocop_tfrm_id tid)
+{
+	struct cryptocop_transform_ctx *tc = sess->tfrm_ctx;
+
+	DEBUG(printk("get_transform_ctx, sess=0x%p, tid=%d\n", sess, tid));
+	assert(sess != NULL);
+	while (tc && tc->init.tid != tid){
+		DEBUG(printk("tc=0x%p, tc->next=0x%p\n", tc, tc->next));
+		tc = tc->next;
+	}
+	DEBUG(printk("get_transform_ctx, returning tc=0x%p\n", tc));
+	return tc;
+}
+
+
+
+/* The AES s-transform matrix (s-box). */
+static const u8 aes_sbox[256] = {
+	99,  124, 119, 123, 242, 107, 111, 197, 48,  1,   103, 43,  254, 215, 171, 118,
+	202, 130, 201, 125, 250, 89,  71,  240, 173, 212, 162, 175, 156, 164, 114, 192,
+	183, 253, 147, 38,  54,  63,  247, 204, 52,  165, 229, 241, 113, 216, 49,  21,
+	4,   199, 35,  195, 24,  150, 5,   154, 7,   18,  128, 226, 235, 39,  178, 117,
+	9,   131, 44,  26,  27,  110, 90,  160, 82,  59,  214, 179, 41,  227, 47,  132,
+	83,  209, 0,   237, 32,  252, 177, 91,  106, 203, 190, 57,  74,  76,  88,  207,
+	208, 239, 170, 251, 67,  77,  51,  133, 69,  249, 2,   127, 80,  60,  159, 168,
+	81,  163, 64,  143, 146, 157, 56,  245, 188, 182, 218, 33,  16,  255, 243, 210,
+	205, 12,  19,  236, 95,  151, 68,  23,  196, 167, 126, 61,  100, 93,  25,  115,
+	96,  129, 79,  220, 34,  42,  144, 136, 70,  238, 184, 20,  222, 94,  11,  219,
+	224, 50,  58,  10,  73,  6,   36,  92,  194, 211, 172, 98,  145, 149, 228, 121,
+	231, 200, 55,  109, 141, 213, 78,  169, 108, 86,  244, 234, 101, 122, 174, 8,
+	186, 120, 37,  46,  28,  166, 180, 198, 232, 221, 116, 31,  75,  189, 139, 138,
+	112, 62,  181, 102, 72,  3,   246, 14,  97,  53,  87,  185, 134, 193, 29,  158,
+	225, 248, 152, 17,  105, 217, 142, 148, 155, 30,  135, 233, 206, 85,  40,  223,
+	140, 161, 137, 13,  191, 230, 66,  104, 65,  153, 45,  15,  176, 84,  187, 22
+};
+
+/* AES has a 32 bit word round constants for each round in the
+ * key schedule.  round_constant[i] is really Rcon[i+1] in FIPS187.
+ */
+static u32 round_constant[11] = {
+	0x01000000, 0x02000000, 0x04000000, 0x08000000,
+	0x10000000, 0x20000000, 0x40000000, 0x80000000,
+	0x1B000000, 0x36000000, 0x6C000000
+};
+
+/* Apply the s-box to each of the four occtets in w. */
+static u32 aes_ks_subword(const u32 w)
+{
+	u8 bytes[4];
+
+	*(u32*)(&bytes[0]) = w;
+	bytes[0] = aes_sbox[bytes[0]];
+	bytes[1] = aes_sbox[bytes[1]];
+	bytes[2] = aes_sbox[bytes[2]];
+	bytes[3] = aes_sbox[bytes[3]];
+	return *(u32*)(&bytes[0]);
+}
+
+/* The encrypt (forward) Rijndael key schedule algorithm pseudo code:
+ * (Note that AES words are 32 bit long)
+ *
+ * KeyExpansion(byte key[4*Nk], word w[Nb*(Nr+1)], Nk){
+ * word temp
+ * i = 0
+ * while (i < Nk) {
+ *   w[i] = word(key[4*i, 4*i + 1, 4*i + 2, 4*i + 3])
+ *   i = i + 1
+ * }
+ * i = Nk
+ *
+ * while (i < (Nb * (Nr + 1))) {
+ *   temp = w[i - 1]
+ *   if ((i mod Nk) == 0) {
+ *     temp = SubWord(RotWord(temp)) xor Rcon[i/Nk]
+ *   }
+ *   else if ((Nk > 6) && ((i mod Nk) == 4)) {
+ *     temp = SubWord(temp)
+ *   }
+ *   w[i] = w[i - Nk] xor temp
+ * }
+ * RotWord(t) does a 8 bit cyclic shift left on a 32 bit word.
+ * SubWord(t) applies the AES s-box individually to each octet
+ * in a 32 bit word.
+ *
+ * For AES Nk can have the values 4, 6, and 8 (corresponding to
+ * values for Nr of 10, 12, and 14).  Nb is always 4.
+ *
+ * To construct w[i], w[i - 1] and w[i - Nk] must be
+ * available.  Consequently we must keep a state of the last Nk words
+ * to be able to create the last round keys.
+ */
+static void get_aes_decrypt_key(unsigned char *dec_key, const unsigned  char *key, unsigned int keylength)
+{
+	u32 temp;
+	u32 w_ring[8]; /* nk is max 8, use elements 0..(nk - 1) as a ringbuffer */
+	u8  w_last_ix;
+	int i;
+	u8  nr, nk;
+
+	switch (keylength){
+	case 128:
+		nk = 4;
+		nr = 10;
+		break;
+	case 192:
+		nk = 6;
+		nr = 12;
+		break;
+	case 256:
+		nk = 8;
+		nr = 14;
+		break;
+	default:
+		panic("stream co-processor: bad aes key length in get_aes_decrypt_key\n");
+	};
+
+	/* Need to do host byte order correction here since key is byte oriented and the
+	 * kx algorithm is word (u32) oriented. */
+	for (i = 0; i < nk; i+=1) {
+		w_ring[i] = be32_to_cpu(*(u32*)&key[4*i]);
+	}
+
+	i = (int)nk;
+	w_last_ix = i - 1;
+	while (i < (4 * (nr + 2))) {
+		temp = w_ring[w_last_ix];
+		if (!(i % nk)) {
+			/* RotWord(temp) */
+			temp = (temp << 8) | (temp >> 24);
+			temp = aes_ks_subword(temp);
+			temp ^= round_constant[i/nk - 1];
+		} else if ((nk > 6) && ((i % nk) == 4)) {
+			temp = aes_ks_subword(temp);
+		}
+		w_last_ix = (w_last_ix + 1) % nk; /* This is the same as (i-Nk) mod Nk */
+		temp ^= w_ring[w_last_ix];
+		w_ring[w_last_ix] = temp;
+
+		/* We need the round keys for round Nr+1 and Nr+2 (round key
+		 * Nr+2 is the round key beyond the last one used when
+		 * encrypting).  Rounds are numbered starting from 0, Nr=10
+		 * implies 11 rounds are used in encryption/decryption.
+		 */
+		if (i >= (4 * nr)) {
+			/* Need to do host byte order correction here, the key
+			 * is byte oriented. */
+			*(u32*)dec_key = cpu_to_be32(temp);
+			dec_key += 4;
+		}
+		++i;
+	}
+}
+
+
+/**** Job/operation management. ****/
+
+int cryptocop_job_queue_insert_csum(struct cryptocop_operation *operation)
+{
+	return cryptocop_job_queue_insert(cryptocop_prio_kernel_csum, operation);
+}
+
+int cryptocop_job_queue_insert_crypto(struct cryptocop_operation *operation)
+{
+	return cryptocop_job_queue_insert(cryptocop_prio_kernel, operation);
+}
+
+int cryptocop_job_queue_insert_user_job(struct cryptocop_operation *operation)
+{
+	return cryptocop_job_queue_insert(cryptocop_prio_user, operation);
+}
+
+static int cryptocop_job_queue_insert(cryptocop_queue_priority prio, struct cryptocop_operation *operation)
+{
+	int                           ret;
+	struct cryptocop_prio_job     *pj = NULL;
+	unsigned long int             flags;
+
+	DEBUG(printk("cryptocop_job_queue_insert(%d, 0x%p)\n", prio, operation));
+
+	if (!operation || !operation->cb){
+		DEBUG_API(printk("cryptocop_job_queue_insert oper=0x%p, NULL operation or callback\n", operation));
+		return -EINVAL;
+	}
+
+	if ((ret = cryptocop_job_setup(&pj, operation)) != 0){
+		DEBUG_API(printk("cryptocop_job_queue_insert: job setup failed\n"));
+		return ret;
+	}
+	assert(pj != NULL);
+
+	spin_lock_irqsave(&cryptocop_job_queue_lock, flags);
+	list_add_tail(&pj->node, &cryptocop_job_queues[prio].jobs);
+	spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags);
+
+	/* Make sure a job is running */
+	cryptocop_start_job();
+	return 0;
+}
+
+static void cryptocop_do_tasklet(unsigned long unused);
+DECLARE_TASKLET (cryptocop_tasklet, cryptocop_do_tasklet, 0);
+
+static void cryptocop_do_tasklet(unsigned long unused)
+{
+	struct list_head             *node;
+	struct cryptocop_prio_job    *pj = NULL;
+	unsigned long                flags;
+
+	DEBUG(printk("cryptocop_do_tasklet: entering\n"));
+
+	do {
+		spin_lock_irqsave(&cryptocop_completed_jobs_lock, flags);
+		if (!list_empty(&cryptocop_completed_jobs)){
+			node = cryptocop_completed_jobs.next;
+			list_del(node);
+			pj = list_entry(node, struct cryptocop_prio_job, node);
+		} else {
+			pj = NULL;
+		}
+		spin_unlock_irqrestore(&cryptocop_completed_jobs_lock, flags);
+		if (pj) {
+			assert(pj->oper != NULL);
+
+			/* Notify consumer of operation completeness. */
+			DEBUG(printk("cryptocop_do_tasklet: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data));
+
+			pj->oper->operation_status = 0; /* Job is completed. */
+			pj->oper->cb(pj->oper, pj->oper->cb_data);
+			delete_internal_operation(pj->iop);
+			kfree(pj);
+		}
+	} while (pj != NULL);
+
+	DEBUG(printk("cryptocop_do_tasklet: exiting\n"));
+}
+
+static irqreturn_t
+dma_done_interrupt(int irq, void *dev_id)
+{
+	struct cryptocop_prio_job *done_job;
+	reg_dma_rw_ack_intr ack_intr = {
+		.data = 1,
+	};
+
+	REG_WR(dma, IN_DMA_INST, rw_ack_intr, ack_intr);
+
+	DEBUG(printk("cryptocop DMA done\n"));
+
+	spin_lock(&running_job_lock);
+	if (cryptocop_running_job == NULL){
+		printk("stream co-processor got interrupt when not busy\n");
+		spin_unlock(&running_job_lock);
+		return IRQ_HANDLED;
+	}
+	done_job = cryptocop_running_job;
+	cryptocop_running_job = NULL;
+	spin_unlock(&running_job_lock);
+
+	/* Start processing a job. */
+	if (!spin_trylock(&cryptocop_process_lock)){
+		DEBUG(printk("cryptocop irq handler, not starting a job\n"));
+	} else {
+		cryptocop_start_job();
+		spin_unlock(&cryptocop_process_lock);
+	}
+
+	done_job->oper->operation_status = 0; /* Job is completed. */
+	if (done_job->oper->fast_callback){
+		/* This operation wants callback from interrupt. */
+		done_job->oper->cb(done_job->oper, done_job->oper->cb_data);
+		delete_internal_operation(done_job->iop);
+		kfree(done_job);
+	} else {
+		spin_lock(&cryptocop_completed_jobs_lock);
+		list_add_tail(&(done_job->node), &cryptocop_completed_jobs);
+		spin_unlock(&cryptocop_completed_jobs_lock);
+		tasklet_schedule(&cryptocop_tasklet);
+	}
+
+	DEBUG(printk("cryptocop leave irq handler\n"));
+	return IRQ_HANDLED;
+}
+
+
+/* Setup interrupts and DMA channels. */
+static int init_cryptocop(void)
+{
+	unsigned long          flags;
+	reg_dma_rw_cfg         dma_cfg = {.en = 1};
+	reg_dma_rw_intr_mask   intr_mask_in = {.data = regk_dma_yes}; /* Only want descriptor interrupts from the DMA in channel. */
+	reg_dma_rw_ack_intr    ack_intr = {.data = 1,.in_eop = 1 };
+	reg_strcop_rw_cfg      strcop_cfg = {
+		.ipend = regk_strcop_little,
+		.td1 = regk_strcop_e,
+		.td2 = regk_strcop_d,
+		.td3 = regk_strcop_e,
+		.ignore_sync = 0,
+		.en = 1
+	};
+
+	if (request_irq(DMA_IRQ, dma_done_interrupt, 0,
+			"stream co-processor DMA", NULL))
+		panic("request_irq stream co-processor irq dma9");
+
+	(void)crisv32_request_dma(OUT_DMA, "strcop", DMA_PANIC_ON_ERROR,
+		0, dma_strp);
+	(void)crisv32_request_dma(IN_DMA, "strcop", DMA_PANIC_ON_ERROR,
+		0, dma_strp);
+
+	local_irq_save(flags);
+
+	/* Reset and enable the cryptocop. */
+	strcop_cfg.en = 0;
+	REG_WR(strcop, regi_strcop, rw_cfg, strcop_cfg);
+	strcop_cfg.en = 1;
+	REG_WR(strcop, regi_strcop, rw_cfg, strcop_cfg);
+
+	/* Enable DMAs. */
+	REG_WR(dma, IN_DMA_INST, rw_cfg, dma_cfg); /* input DMA */
+	REG_WR(dma, OUT_DMA_INST, rw_cfg, dma_cfg); /* output DMA */
+
+	/* Set up wordsize = 4 for DMAs. */
+	DMA_WR_CMD(OUT_DMA_INST, regk_dma_set_w_size4);
+	DMA_WR_CMD(IN_DMA_INST, regk_dma_set_w_size4);
+
+	/* Enable interrupts. */
+	REG_WR(dma, IN_DMA_INST, rw_intr_mask, intr_mask_in);
+
+	/* Clear intr ack. */
+	REG_WR(dma, IN_DMA_INST, rw_ack_intr, ack_intr);
+
+	local_irq_restore(flags);
+
+	return 0;
+}
+
+/* Free used cryptocop hw resources (interrupt and DMA channels). */
+static void release_cryptocop(void)
+{
+	unsigned long          flags;
+	reg_dma_rw_cfg         dma_cfg = {.en = 0};
+	reg_dma_rw_intr_mask   intr_mask_in = {0};
+	reg_dma_rw_ack_intr    ack_intr = {.data = 1,.in_eop = 1 };
+
+	local_irq_save(flags);
+
+	/* Clear intr ack. */
+	REG_WR(dma, IN_DMA_INST, rw_ack_intr, ack_intr);
+
+	/* Disable DMAs. */
+	REG_WR(dma, IN_DMA_INST, rw_cfg, dma_cfg); /* input DMA */
+	REG_WR(dma, OUT_DMA_INST, rw_cfg, dma_cfg); /* output DMA */
+
+	/* Disable interrupts. */
+	REG_WR(dma, IN_DMA_INST, rw_intr_mask, intr_mask_in);
+
+	local_irq_restore(flags);
+
+	free_irq(DMA_IRQ, NULL);
+
+	(void)crisv32_free_dma(OUT_DMA);
+	(void)crisv32_free_dma(IN_DMA);
+}
+
+
+/* Init job queue. */
+static int cryptocop_job_queue_init(void)
+{
+	int i;
+
+	INIT_LIST_HEAD(&cryptocop_completed_jobs);
+
+	for (i = 0; i < cryptocop_prio_no_prios; i++){
+		cryptocop_job_queues[i].prio = (cryptocop_queue_priority)i;
+		INIT_LIST_HEAD(&cryptocop_job_queues[i].jobs);
+	}
+	return 0;
+}
+
+
+static void cryptocop_job_queue_close(void)
+{
+	struct list_head               *node, *tmp;
+	struct cryptocop_prio_job      *pj = NULL;
+	unsigned long int              process_flags, flags;
+	int                            i;
+
+	/* FIXME: This is as yet untested code. */
+
+	/* Stop strcop from getting an operation to process while we are closing the
+	   module. */
+	spin_lock_irqsave(&cryptocop_process_lock, process_flags);
+
+	/* Empty the job queue. */
+	for (i = 0; i < cryptocop_prio_no_prios; i++){
+		if (!list_empty(&(cryptocop_job_queues[i].jobs))){
+			list_for_each_safe(node, tmp, &(cryptocop_job_queues[i].jobs)) {
+				pj = list_entry(node, struct cryptocop_prio_job, node);
+				list_del(node);
+
+				/* Call callback to notify consumer of job removal. */
+				DEBUG(printk("cryptocop_job_queue_close: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data));
+				pj->oper->operation_status = -EINTR; /* Job is terminated without completion. */
+				pj->oper->cb(pj->oper, pj->oper->cb_data);
+
+				delete_internal_operation(pj->iop);
+				kfree(pj);
+			}
+		}
+	}
+	spin_unlock_irqrestore(&cryptocop_process_lock, process_flags);
+
+	/* Remove the running job, if any. */
+	spin_lock_irqsave(&running_job_lock, flags);
+	if (cryptocop_running_job){
+		reg_strcop_rw_cfg rw_cfg;
+		reg_dma_rw_cfg    dma_out_cfg, dma_in_cfg;
+
+		/* Stop DMA. */
+		dma_out_cfg = REG_RD(dma, OUT_DMA_INST, rw_cfg);
+		dma_out_cfg.en = regk_dma_no;
+		REG_WR(dma, OUT_DMA_INST, rw_cfg, dma_out_cfg);
+
+		dma_in_cfg = REG_RD(dma, IN_DMA_INST, rw_cfg);
+		dma_in_cfg.en = regk_dma_no;
+		REG_WR(dma, IN_DMA_INST, rw_cfg, dma_in_cfg);
+
+		/* Disble the cryptocop. */
+		rw_cfg = REG_RD(strcop, regi_strcop, rw_cfg);
+		rw_cfg.en = 0;
+		REG_WR(strcop, regi_strcop, rw_cfg, rw_cfg);
+
+		pj = cryptocop_running_job;
+		cryptocop_running_job = NULL;
+
+		/* Call callback to notify consumer of job removal. */
+		DEBUG(printk("cryptocop_job_queue_close: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data));
+		pj->oper->operation_status = -EINTR; /* Job is terminated without completion. */
+		pj->oper->cb(pj->oper, pj->oper->cb_data);
+
+		delete_internal_operation(pj->iop);
+		kfree(pj);
+	}
+	spin_unlock_irqrestore(&running_job_lock, flags);
+
+	/* Remove completed jobs, if any. */
+	spin_lock_irqsave(&cryptocop_completed_jobs_lock, flags);
+
+	list_for_each_safe(node, tmp, &cryptocop_completed_jobs) {
+		pj = list_entry(node, struct cryptocop_prio_job, node);
+		list_del(node);
+		/* Call callback to notify consumer of job removal. */
+		DEBUG(printk("cryptocop_job_queue_close: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data));
+		pj->oper->operation_status = -EINTR; /* Job is terminated without completion. */
+		pj->oper->cb(pj->oper, pj->oper->cb_data);
+
+		delete_internal_operation(pj->iop);
+		kfree(pj);
+	}
+	spin_unlock_irqrestore(&cryptocop_completed_jobs_lock, flags);
+}
+
+
+static void cryptocop_start_job(void)
+{
+	int                          i;
+	struct cryptocop_prio_job    *pj;
+	unsigned long int            flags;
+	unsigned long int            running_job_flags;
+	reg_strcop_rw_cfg            rw_cfg = {.en = 1, .ignore_sync = 0};
+
+	DEBUG(printk("cryptocop_start_job: entering\n"));
+
+	spin_lock_irqsave(&running_job_lock, running_job_flags);
+	if (cryptocop_running_job != NULL){
+		/* Already running. */
+		DEBUG(printk("cryptocop_start_job: already running, exit\n"));
+		spin_unlock_irqrestore(&running_job_lock, running_job_flags);
+		return;
+	}
+	spin_lock_irqsave(&cryptocop_job_queue_lock, flags);
+
+	/* Check the queues in priority order. */
+	for (i = cryptocop_prio_kernel_csum; (i < cryptocop_prio_no_prios) && list_empty(&cryptocop_job_queues[i].jobs); i++);
+	if (i == cryptocop_prio_no_prios) {
+		spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags);
+		spin_unlock_irqrestore(&running_job_lock, running_job_flags);
+		DEBUG(printk("cryptocop_start_job: no jobs to run\n"));
+		return; /* No jobs to run */
+	}
+	DEBUG(printk("starting job for prio %d\n", i));
+
+	/* TODO: Do not starve lower priority jobs.  Let in a lower
+	 * prio job for every N-th processed higher prio job or some
+	 * other scheduling policy.  This could reasonably be
+	 * tweakable since the optimal balance would depend on the
+	 * type of load on the system. */
+
+	/* Pull the DMA lists from the job and start the DMA client. */
+	pj = list_entry(cryptocop_job_queues[i].jobs.next, struct cryptocop_prio_job, node);
+	list_del(&pj->node);
+	spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags);
+	cryptocop_running_job = pj;
+
+	/* Set config register (3DES and CSUM modes). */
+	switch (pj->iop->tdes_mode){
+	case cryptocop_3des_eee:
+		rw_cfg.td1 = regk_strcop_e;
+		rw_cfg.td2 = regk_strcop_e;
+		rw_cfg.td3 = regk_strcop_e;
+		break;
+	case cryptocop_3des_eed:
+		rw_cfg.td1 = regk_strcop_e;
+		rw_cfg.td2 = regk_strcop_e;
+		rw_cfg.td3 = regk_strcop_d;
+		break;
+	case cryptocop_3des_ede:
+		rw_cfg.td1 = regk_strcop_e;
+		rw_cfg.td2 = regk_strcop_d;
+		rw_cfg.td3 = regk_strcop_e;
+		break;
+	case cryptocop_3des_edd:
+		rw_cfg.td1 = regk_strcop_e;
+		rw_cfg.td2 = regk_strcop_d;
+		rw_cfg.td3 = regk_strcop_d;
+		break;
+	case cryptocop_3des_dee:
+		rw_cfg.td1 = regk_strcop_d;
+		rw_cfg.td2 = regk_strcop_e;
+		rw_cfg.td3 = regk_strcop_e;
+		break;
+	case cryptocop_3des_ded:
+		rw_cfg.td1 = regk_strcop_d;
+		rw_cfg.td2 = regk_strcop_e;
+		rw_cfg.td3 = regk_strcop_d;
+		break;
+	case cryptocop_3des_dde:
+		rw_cfg.td1 = regk_strcop_d;
+		rw_cfg.td2 = regk_strcop_d;
+		rw_cfg.td3 = regk_strcop_e;
+		break;
+	case cryptocop_3des_ddd:
+		rw_cfg.td1 = regk_strcop_d;
+		rw_cfg.td2 = regk_strcop_d;
+		rw_cfg.td3 = regk_strcop_d;
+		break;
+	default:
+		DEBUG(printk("cryptocop_setup_dma_list: bad 3DES mode\n"));
+	}
+	switch (pj->iop->csum_mode){
+	case cryptocop_csum_le:
+		rw_cfg.ipend = regk_strcop_little;
+		break;
+	case cryptocop_csum_be:
+		rw_cfg.ipend = regk_strcop_big;
+		break;
+	default:
+		DEBUG(printk("cryptocop_setup_dma_list: bad checksum mode\n"));
+	}
+	REG_WR(strcop, regi_strcop, rw_cfg, rw_cfg);
+
+	DEBUG(printk("cryptocop_start_job: starting DMA, new cryptocop_running_job=0x%p\n"
+		     "ctx_in: 0x%p, phys: 0x%p\n"
+		     "ctx_out: 0x%p, phys: 0x%p\n",
+		     pj,
+		     &pj->iop->ctx_in, (char*)virt_to_phys(&pj->iop->ctx_in),
+		     &pj->iop->ctx_out, (char*)virt_to_phys(&pj->iop->ctx_out)));
+
+	/* Start input DMA. */
+	flush_dma_context(&pj->iop->ctx_in);
+	DMA_START_CONTEXT(IN_DMA_INST, virt_to_phys(&pj->iop->ctx_in));
+
+	/* Start output DMA. */
+	DMA_START_CONTEXT(OUT_DMA_INST, virt_to_phys(&pj->iop->ctx_out));
+
+	spin_unlock_irqrestore(&running_job_lock, running_job_flags);
+	DEBUG(printk("cryptocop_start_job: exiting\n"));
+}
+
+
+static int cryptocop_job_setup(struct cryptocop_prio_job **pj, struct cryptocop_operation *operation)
+{
+	int  err;
+	int  alloc_flag = operation->in_interrupt ? GFP_ATOMIC : GFP_KERNEL;
+	void *iop_alloc_ptr = NULL;
+
+	*pj = kmalloc(sizeof (struct cryptocop_prio_job), alloc_flag);
+	if (!*pj) return -ENOMEM;
+
+	DEBUG(printk("cryptocop_job_setup: operation=0x%p\n", operation));
+
+	(*pj)->oper = operation;
+	DEBUG(printk("cryptocop_job_setup, cb=0x%p cb_data=0x%p\n",  (*pj)->oper->cb, (*pj)->oper->cb_data));
+
+	if (operation->use_dmalists) {
+		DEBUG(print_user_dma_lists(&operation->list_op));
+		if (!operation->list_op.inlist || !operation->list_op.outlist || !operation->list_op.out_data_buf || !operation->list_op.in_data_buf){
+			DEBUG_API(printk("cryptocop_job_setup: bad indata (use_dmalists)\n"));
+			kfree(*pj);
+			return -EINVAL;
+		}
+		iop_alloc_ptr = kmalloc(DESCR_ALLOC_PAD + sizeof(struct cryptocop_int_operation), alloc_flag);
+		if (!iop_alloc_ptr) {
+			DEBUG_API(printk("cryptocop_job_setup: kmalloc cryptocop_int_operation\n"));
+			kfree(*pj);
+			return -ENOMEM;
+		}
+		(*pj)->iop = (struct cryptocop_int_operation*)(((unsigned long int)(iop_alloc_ptr + DESCR_ALLOC_PAD + offsetof(struct cryptocop_int_operation, ctx_out)) & ~0x0000001F) - offsetof(struct cryptocop_int_operation, ctx_out));
+		DEBUG(memset((*pj)->iop, 0xff, sizeof(struct cryptocop_int_operation)));
+		(*pj)->iop->alloc_ptr = iop_alloc_ptr;
+		(*pj)->iop->sid = operation->sid;
+		(*pj)->iop->cdesc_out = NULL;
+		(*pj)->iop->cdesc_in = NULL;
+		(*pj)->iop->tdes_mode = operation->list_op.tdes_mode;
+		(*pj)->iop->csum_mode = operation->list_op.csum_mode;
+		(*pj)->iop->ddesc_out = operation->list_op.outlist;
+		(*pj)->iop->ddesc_in = operation->list_op.inlist;
+
+		/* Setup DMA contexts. */
+		(*pj)->iop->ctx_out.next = NULL;
+		(*pj)->iop->ctx_out.eol = 1;
+		(*pj)->iop->ctx_out.saved_data = operation->list_op.outlist;
+		(*pj)->iop->ctx_out.saved_data_buf = operation->list_op.out_data_buf;
+
+		(*pj)->iop->ctx_in.next = NULL;
+		(*pj)->iop->ctx_in.eol = 1;
+		(*pj)->iop->ctx_in.saved_data = operation->list_op.inlist;
+		(*pj)->iop->ctx_in.saved_data_buf = operation->list_op.in_data_buf;
+	} else {
+		if ((err = cryptocop_setup_dma_list(operation, &(*pj)->iop, alloc_flag))) {
+			DEBUG_API(printk("cryptocop_job_setup: cryptocop_setup_dma_list failed %d\n", err));
+			kfree(*pj);
+			return err;
+		}
+	}
+	DEBUG(print_dma_descriptors((*pj)->iop));
+
+	DEBUG(printk("cryptocop_job_setup, DMA list setup successful\n"));
+
+	return 0;
+}
+
+static int cryptocop_open(struct inode *inode, struct file *filp)
+{
+	int p = iminor(inode);
+
+	if (p != CRYPTOCOP_MINOR) return -EINVAL;
+
+	filp->private_data = NULL;
+	return 0;
+}
+
+
+static int cryptocop_release(struct inode *inode, struct file *filp)
+{
+	struct cryptocop_private *dev = filp->private_data;
+	struct cryptocop_private *dev_next;
+
+	while (dev){
+		dev_next = dev->next;
+		if (dev->sid != CRYPTOCOP_SESSION_ID_NONE) {
+			(void)cryptocop_free_session(dev->sid);
+		}
+		kfree(dev);
+		dev = dev_next;
+	}
+
+	return 0;
+}
+
+
+static int cryptocop_ioctl_close_session(struct inode *inode, struct file *filp,
+					 unsigned int cmd, unsigned long arg)
+{
+	struct cryptocop_private  *dev = filp->private_data;
+	struct cryptocop_private  *prev_dev = NULL;
+	struct strcop_session_op  *sess_op = (struct strcop_session_op *)arg;
+	struct strcop_session_op  sop;
+	int                       err;
+
+	DEBUG(printk("cryptocop_ioctl_close_session\n"));
+
+	if (!access_ok(VERIFY_READ, sess_op, sizeof(struct strcop_session_op)))
+		return -EFAULT;
+	err = copy_from_user(&sop, sess_op, sizeof(struct strcop_session_op));
+	if (err) return -EFAULT;
+
+	while (dev && (dev->sid != sop.ses_id)) {
+		prev_dev = dev;
+		dev = dev->next;
+	}
+	if (dev){
+		if (prev_dev){
+			prev_dev->next = dev->next;
+		} else {
+			filp->private_data = dev->next;
+		}
+		err = cryptocop_free_session(dev->sid);
+		if (err) return -EFAULT;
+	} else {
+		DEBUG_API(printk("cryptocop_ioctl_close_session: session %lld not found\n", sop.ses_id));
+		return -EINVAL;
+	}
+	return 0;
+}
+
+
+static void ioctl_process_job_callback(struct cryptocop_operation *op, void*cb_data)
+{
+	struct ioctl_job_cb_ctx *jc = (struct ioctl_job_cb_ctx *)cb_data;
+
+	DEBUG(printk("ioctl_process_job_callback: op=0x%p, cb_data=0x%p\n", op, cb_data));
+
+	jc->processed = 1;
+	wake_up(&cryptocop_ioc_process_wq);
+}
+
+
+#define CRYPTOCOP_IOCTL_CIPHER_TID  (1)
+#define CRYPTOCOP_IOCTL_DIGEST_TID  (2)
+#define CRYPTOCOP_IOCTL_CSUM_TID    (3)
+
+static size_t first_cfg_change_ix(struct strcop_crypto_op *crp_op)
+{
+	size_t ch_ix = 0;
+
+	if (crp_op->do_cipher) ch_ix = crp_op->cipher_start;
+	if (crp_op->do_digest && (crp_op->digest_start < ch_ix)) ch_ix = crp_op->digest_start;
+	if (crp_op->do_csum && (crp_op->csum_start < ch_ix)) ch_ix = crp_op->csum_start;
+
+	DEBUG(printk("first_cfg_change_ix: ix=%d\n", ch_ix));
+	return ch_ix;
+}
+
+
+static size_t next_cfg_change_ix(struct strcop_crypto_op *crp_op, size_t ix)
+{
+	size_t ch_ix = INT_MAX;
+	size_t tmp_ix = 0;
+
+	if (crp_op->do_cipher && ((crp_op->cipher_start + crp_op->cipher_len) > ix)){
+		if (crp_op->cipher_start > ix) {
+			ch_ix = crp_op->cipher_start;
+		} else {
+			ch_ix = crp_op->cipher_start + crp_op->cipher_len;
+		}
+	}
+	if (crp_op->do_digest && ((crp_op->digest_start + crp_op->digest_len) > ix)){
+		if (crp_op->digest_start > ix) {
+			tmp_ix = crp_op->digest_start;
+		} else {
+			tmp_ix = crp_op->digest_start + crp_op->digest_len;
+		}
+		if (tmp_ix < ch_ix) ch_ix = tmp_ix;
+	}
+	if (crp_op->do_csum && ((crp_op->csum_start + crp_op->csum_len) > ix)){
+		if (crp_op->csum_start > ix) {
+			tmp_ix = crp_op->csum_start;
+		} else {
+			tmp_ix = crp_op->csum_start + crp_op->csum_len;
+		}
+		if (tmp_ix < ch_ix) ch_ix = tmp_ix;
+	}
+	if (ch_ix == INT_MAX) ch_ix = ix;
+	DEBUG(printk("next_cfg_change_ix prev ix=%d, next ix=%d\n", ix, ch_ix));
+	return ch_ix;
+}
+
+
+/* Map map_length bytes from the pages starting on *pageix and *pageoffset to iovecs starting on *iovix.
+ * Return -1 for ok, 0 for fail. */
+static int map_pages_to_iovec(struct iovec *iov, int iovlen, int *iovix, struct page **pages, int nopages, int *pageix, int *pageoffset, int map_length )
+{
+	int tmplen;
+
+	assert(iov != NULL);
+	assert(iovix != NULL);
+	assert(pages != NULL);
+	assert(pageix != NULL);
+	assert(pageoffset != NULL);
+
+	DEBUG(printk("map_pages_to_iovec, map_length=%d, iovlen=%d, *iovix=%d, nopages=%d, *pageix=%d, *pageoffset=%d\n", map_length, iovlen, *iovix, nopages, *pageix, *pageoffset));
+
+	while (map_length > 0){
+		DEBUG(printk("map_pages_to_iovec, map_length=%d, iovlen=%d, *iovix=%d, nopages=%d, *pageix=%d, *pageoffset=%d\n", map_length, iovlen, *iovix, nopages, *pageix, *pageoffset));
+		if (*iovix >= iovlen){
+			DEBUG_API(printk("map_page_to_iovec: *iovix=%d >= iovlen=%d\n", *iovix, iovlen));
+			return 0;
+		}
+		if (*pageix >= nopages){
+			DEBUG_API(printk("map_page_to_iovec: *pageix=%d >= nopages=%d\n", *pageix, nopages));
+			return 0;
+		}
+		iov[*iovix].iov_base = (unsigned char*)page_address(pages[*pageix]) + *pageoffset;
+		tmplen = PAGE_SIZE - *pageoffset;
+		if (tmplen < map_length){
+			(*pageoffset) = 0;
+			(*pageix)++;
+		} else {
+			tmplen = map_length;
+			(*pageoffset) += map_length;
+		}
+		DEBUG(printk("mapping %d bytes from page %d (or %d) to iovec %d\n", tmplen, *pageix, *pageix-1, *iovix));
+		iov[*iovix].iov_len = tmplen;
+		map_length -= tmplen;
+		(*iovix)++;
+	}
+	DEBUG(printk("map_page_to_iovec, exit, *iovix=%d\n", *iovix));
+	return -1;
+}
+
+
+
+static int cryptocop_ioctl_process(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	int                             i;
+	struct cryptocop_private        *dev = filp->private_data;
+	struct strcop_crypto_op         *crp_oper = (struct strcop_crypto_op *)arg;
+	struct strcop_crypto_op         oper = {0};
+	int                             err = 0;
+	struct cryptocop_operation      *cop = NULL;
+
+	struct ioctl_job_cb_ctx         *jc = NULL;
+
+	struct page                     **inpages = NULL;
+	struct page                     **outpages = NULL;
+	int                             noinpages = 0;
+	int                             nooutpages = 0;
+
+	struct cryptocop_desc           descs[5]; /* Max 5 descriptors are needed, there are three transforms that
+						   * can get connected/disconnected on different places in the indata. */
+	struct cryptocop_desc_cfg       dcfgs[5*3];
+	int                             desc_ix = 0;
+	int                             dcfg_ix = 0;
+	struct cryptocop_tfrm_cfg       ciph_tcfg = {0};
+	struct cryptocop_tfrm_cfg       digest_tcfg = {0};
+	struct cryptocop_tfrm_cfg       csum_tcfg = {0};
+
+	unsigned char                   *digest_result = NULL;
+	int                             digest_length = 0;
+	int                             cblocklen = 0;
+	unsigned char                   csum_result[CSUM_BLOCK_LENGTH];
+	struct cryptocop_session        *sess;
+
+	int    iovlen = 0;
+	int    iovix = 0;
+	int    pageix = 0;
+	int    pageoffset = 0;
+
+	size_t prev_ix = 0;
+	size_t next_ix;
+
+	int    cipher_active, digest_active, csum_active;
+	int    end_digest, end_csum;
+	int    digest_done = 0;
+	int    cipher_done = 0;
+	int    csum_done = 0;
+
+	DEBUG(printk("cryptocop_ioctl_process\n"));
+
+	if (!access_ok(VERIFY_WRITE, crp_oper, sizeof(struct strcop_crypto_op))){
+		DEBUG_API(printk("cryptocop_ioctl_process: !access_ok crp_oper!\n"));
+		return -EFAULT;
+	}
+	if (copy_from_user(&oper, crp_oper, sizeof(struct strcop_crypto_op))) {
+		DEBUG_API(printk("cryptocop_ioctl_process: copy_from_user\n"));
+		return -EFAULT;
+	}
+	DEBUG(print_strcop_crypto_op(&oper));
+
+	while (dev && dev->sid != oper.ses_id) dev = dev->next;
+	if (!dev){
+		DEBUG_API(printk("cryptocop_ioctl_process: session %lld not found\n", oper.ses_id));
+		return -EINVAL;
+	}
+
+	/* Check buffers. */
+	if (((oper.indata + oper.inlen) < oper.indata) || ((oper.cipher_outdata + oper.cipher_outlen) < oper.cipher_outdata)){
+		DEBUG_API(printk("cryptocop_ioctl_process: user buffers wrapped around, bad user!\n"));
+		return -EINVAL;
+	}
+
+	if (!access_ok(VERIFY_WRITE, oper.cipher_outdata, oper.cipher_outlen)){
+		DEBUG_API(printk("cryptocop_ioctl_process: !access_ok out data!\n"));
+		return -EFAULT;
+	}
+	if (!access_ok(VERIFY_READ, oper.indata, oper.inlen)){
+		DEBUG_API(printk("cryptocop_ioctl_process: !access_ok in data!\n"));
+		return -EFAULT;
+	}
+
+	cop = kmalloc(sizeof(struct cryptocop_operation), GFP_KERNEL);
+	if (!cop) {
+		DEBUG_API(printk("cryptocop_ioctl_process: kmalloc\n"));
+		return -ENOMEM;
+	}
+	jc = kmalloc(sizeof(struct ioctl_job_cb_ctx), GFP_KERNEL);
+	if (!jc) {
+		DEBUG_API(printk("cryptocop_ioctl_process: kmalloc\n"));
+		err = -ENOMEM;
+		goto error_cleanup;
+	}
+	jc->processed = 0;
+
+	cop->cb_data = jc;
+	cop->cb = ioctl_process_job_callback;
+	cop->operation_status = 0;
+	cop->use_dmalists = 0;
+	cop->in_interrupt = 0;
+	cop->fast_callback = 0;
+	cop->tfrm_op.tfrm_cfg = NULL;
+	cop->tfrm_op.desc = NULL;
+	cop->tfrm_op.indata = NULL;
+	cop->tfrm_op.incount = 0;
+	cop->tfrm_op.inlen = 0;
+	cop->tfrm_op.outdata = NULL;
+	cop->tfrm_op.outcount = 0;
+	cop->tfrm_op.outlen = 0;
+
+	sess = get_session(oper.ses_id);
+	if (!sess){
+		DEBUG_API(printk("cryptocop_ioctl_process: bad session id.\n"));
+		kfree(cop);
+		kfree(jc);
+		return -EINVAL;
+	}
+
+	if (oper.do_cipher) {
+		unsigned int                    cipher_outlen = 0;
+		struct cryptocop_transform_ctx  *tc = get_transform_ctx(sess, CRYPTOCOP_IOCTL_CIPHER_TID);
+		if (!tc) {
+			DEBUG_API(printk("cryptocop_ioctl_process: no cipher transform in session.\n"));
+			err = -EINVAL;
+			goto error_cleanup;
+		}
+		ciph_tcfg.tid = CRYPTOCOP_IOCTL_CIPHER_TID;
+		ciph_tcfg.inject_ix = 0;
+		ciph_tcfg.flags = 0;
+		if ((oper.cipher_start < 0) || (oper.cipher_len <= 0) || (oper.cipher_start > oper.inlen) || ((oper.cipher_start + oper.cipher_len) > oper.inlen)){
+			DEBUG_API(printk("cryptocop_ioctl_process: bad cipher length\n"));
+			kfree(cop);
+			kfree(jc);
+			return -EINVAL;
+		}
+		cblocklen = tc->init.alg == cryptocop_alg_aes ? AES_BLOCK_LENGTH : DES_BLOCK_LENGTH;
+		if (oper.cipher_len % cblocklen) {
+			kfree(cop);
+			kfree(jc);
+			DEBUG_API(printk("cryptocop_ioctl_process: cipher inlength not multiple of block length.\n"));
+			return -EINVAL;
+		}
+		cipher_outlen = oper.cipher_len;
+		if (tc->init.cipher_mode == cryptocop_cipher_mode_cbc){
+			if (oper.cipher_explicit) {
+				ciph_tcfg.flags |= CRYPTOCOP_EXPLICIT_IV;
+				memcpy(ciph_tcfg.iv, oper.cipher_iv, cblocklen);
+			} else {
+				cipher_outlen = oper.cipher_len - cblocklen;
+			}
+		} else {
+			if (oper.cipher_explicit){
+				kfree(cop);
+				kfree(jc);
+				DEBUG_API(printk("cryptocop_ioctl_process: explicit_iv when not CBC mode\n"));
+				return -EINVAL;
+			}
+		}
+		if (oper.cipher_outlen != cipher_outlen) {
+			kfree(cop);
+			kfree(jc);
+			DEBUG_API(printk("cryptocop_ioctl_process: cipher_outlen incorrect, should be %d not %d.\n", cipher_outlen, oper.cipher_outlen));
+			return -EINVAL;
+		}
+
+		if (oper.decrypt){
+			ciph_tcfg.flags |= CRYPTOCOP_DECRYPT;
+		} else {
+			ciph_tcfg.flags |= CRYPTOCOP_ENCRYPT;
+		}
+		ciph_tcfg.next = cop->tfrm_op.tfrm_cfg;
+		cop->tfrm_op.tfrm_cfg = &ciph_tcfg;
+	}
+	if (oper.do_digest){
+		struct cryptocop_transform_ctx *tc = get_transform_ctx(sess, CRYPTOCOP_IOCTL_DIGEST_TID);
+		if (!tc) {
+			DEBUG_API(printk("cryptocop_ioctl_process: no digest transform in session.\n"));
+			err = -EINVAL;
+			goto error_cleanup;
+		}
+		digest_length = tc->init.alg == cryptocop_alg_md5 ? 16 : 20;
+		digest_result = kmalloc(digest_length, GFP_KERNEL);
+		if (!digest_result) {
+			DEBUG_API(printk("cryptocop_ioctl_process: kmalloc digest_result\n"));
+			err = -EINVAL;
+			goto error_cleanup;
+		}
+		DEBUG(memset(digest_result, 0xff, digest_length));
+
+		digest_tcfg.tid = CRYPTOCOP_IOCTL_DIGEST_TID;
+		digest_tcfg.inject_ix = 0;
+		ciph_tcfg.inject_ix += digest_length;
+		if ((oper.digest_start < 0) || (oper.digest_len <= 0) || (oper.digest_start > oper.inlen) || ((oper.digest_start + oper.digest_len) > oper.inlen)){
+			DEBUG_API(printk("cryptocop_ioctl_process: bad digest length\n"));
+			err = -EINVAL;
+			goto error_cleanup;
+		}
+
+		digest_tcfg.next = cop->tfrm_op.tfrm_cfg;
+		cop->tfrm_op.tfrm_cfg = &digest_tcfg;
+	}
+	if (oper.do_csum){
+		csum_tcfg.tid = CRYPTOCOP_IOCTL_CSUM_TID;
+		csum_tcfg.inject_ix = digest_length;
+		ciph_tcfg.inject_ix += 2;
+
+		if ((oper.csum_start < 0) || (oper.csum_len <= 0) || (oper.csum_start > oper.inlen) || ((oper.csum_start + oper.csum_len) > oper.inlen)){
+			DEBUG_API(printk("cryptocop_ioctl_process: bad csum length\n"));
+			kfree(cop);
+			kfree(jc);
+			return -EINVAL;
+		}
+
+		csum_tcfg.next = cop->tfrm_op.tfrm_cfg;
+		cop->tfrm_op.tfrm_cfg = &csum_tcfg;
+	}
+
+	prev_ix = first_cfg_change_ix(&oper);
+	if (prev_ix > oper.inlen) {
+		DEBUG_API(printk("cryptocop_ioctl_process: length mismatch\n"));
+		nooutpages = noinpages = 0;
+		err = -EINVAL;
+		goto error_cleanup;
+	}
+	DEBUG(printk("cryptocop_ioctl_process: inlen=%d, cipher_outlen=%d\n", oper.inlen, oper.cipher_outlen));
+
+	/* Map user pages for in and out data of the operation. */
+	noinpages = (((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK) + oper.inlen - 1 - prev_ix + ~PAGE_MASK) >> PAGE_SHIFT;
+	DEBUG(printk("cryptocop_ioctl_process: noinpages=%d\n", noinpages));
+	inpages = kmalloc(noinpages * sizeof(struct page*), GFP_KERNEL);
+	if (!inpages){
+		DEBUG_API(printk("cryptocop_ioctl_process: kmalloc inpages\n"));
+		nooutpages = noinpages = 0;
+		err = -ENOMEM;
+		goto error_cleanup;
+	}
+	if (oper.do_cipher){
+		nooutpages = (((unsigned long int)oper.cipher_outdata & ~PAGE_MASK) + oper.cipher_outlen - 1 + ~PAGE_MASK) >> PAGE_SHIFT;
+		DEBUG(printk("cryptocop_ioctl_process: nooutpages=%d\n", nooutpages));
+		outpages = kmalloc(nooutpages * sizeof(struct page*), GFP_KERNEL);
+		if (!outpages){
+			DEBUG_API(printk("cryptocop_ioctl_process: kmalloc outpages\n"));
+			nooutpages = noinpages = 0;
+			err = -ENOMEM;
+			goto error_cleanup;
+		}
+	}
+
+	/* Acquire the mm page semaphore. */
+	down_read(&current->mm->mmap_sem);
+
+	err = get_user_pages(current,
+			     current->mm,
+			     (unsigned long int)(oper.indata + prev_ix),
+			     noinpages,
+			     0,  /* read access only for in data */
+			     0, /* no force */
+			     inpages,
+			     NULL);
+
+	if (err < 0) {
+		up_read(&current->mm->mmap_sem);
+		nooutpages = noinpages = 0;
+		DEBUG_API(printk("cryptocop_ioctl_process: get_user_pages indata\n"));
+		goto error_cleanup;
+	}
+	noinpages = err;
+	if (oper.do_cipher){
+		err = get_user_pages(current,
+				     current->mm,
+				     (unsigned long int)oper.cipher_outdata,
+				     nooutpages,
+				     1, /* write access for out data */
+				     0, /* no force */
+				     outpages,
+				     NULL);
+		up_read(&current->mm->mmap_sem);
+		if (err < 0) {
+			nooutpages = 0;
+			DEBUG_API(printk("cryptocop_ioctl_process: get_user_pages outdata\n"));
+			goto error_cleanup;
+		}
+		nooutpages = err;
+	} else {
+		up_read(&current->mm->mmap_sem);
+	}
+
+	/* Add 6 to nooutpages to make room for possibly inserted buffers for storing digest and
+	 * csum output and splits when units are (dis-)connected. */
+	cop->tfrm_op.indata = kmalloc((noinpages) * sizeof(struct iovec), GFP_KERNEL);
+	cop->tfrm_op.outdata = kmalloc((6 + nooutpages) * sizeof(struct iovec), GFP_KERNEL);
+	if (!cop->tfrm_op.indata || !cop->tfrm_op.outdata) {
+		DEBUG_API(printk("cryptocop_ioctl_process: kmalloc iovecs\n"));
+		err = -ENOMEM;
+		goto error_cleanup;
+	}
+
+	cop->tfrm_op.inlen = oper.inlen - prev_ix;
+	cop->tfrm_op.outlen = 0;
+	if (oper.do_cipher) cop->tfrm_op.outlen += oper.cipher_outlen;
+	if (oper.do_digest) cop->tfrm_op.outlen += digest_length;
+	if (oper.do_csum) cop->tfrm_op.outlen += 2;
+
+	/* Setup the in iovecs. */
+	cop->tfrm_op.incount = noinpages;
+	if (noinpages > 1){
+		size_t tmplen = cop->tfrm_op.inlen;
+
+		cop->tfrm_op.indata[0].iov_len = PAGE_SIZE - ((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK);
+		cop->tfrm_op.indata[0].iov_base = (unsigned char*)page_address(inpages[0]) + ((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK);
+		tmplen -= cop->tfrm_op.indata[0].iov_len;
+		for (i = 1; i<noinpages; i++){
+			cop->tfrm_op.indata[i].iov_len = tmplen < PAGE_SIZE ? tmplen : PAGE_SIZE;
+			cop->tfrm_op.indata[i].iov_base = (unsigned char*)page_address(inpages[i]);
+			tmplen -= PAGE_SIZE;
+		}
+	} else {
+		cop->tfrm_op.indata[0].iov_len = oper.inlen - prev_ix;
+		cop->tfrm_op.indata[0].iov_base = (unsigned char*)page_address(inpages[0]) + ((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK);
+	}
+
+	iovlen = nooutpages + 6;
+	pageoffset = oper.do_cipher ? ((unsigned long int)oper.cipher_outdata & ~PAGE_MASK) : 0;
+
+	next_ix = next_cfg_change_ix(&oper, prev_ix);
+	if (prev_ix == next_ix){
+		DEBUG_API(printk("cryptocop_ioctl_process: length configuration broken.\n"));
+		err = -EINVAL;  /* This should be impossible barring bugs. */
+		goto error_cleanup;
+	}
+	while (prev_ix != next_ix){
+		end_digest = end_csum = cipher_active = digest_active = csum_active = 0;
+		descs[desc_ix].cfg = NULL;
+		descs[desc_ix].length = next_ix - prev_ix;
+
+		if (oper.do_cipher && (oper.cipher_start < next_ix) && (prev_ix < (oper.cipher_start + oper.cipher_len))) {
+			dcfgs[dcfg_ix].tid = CRYPTOCOP_IOCTL_CIPHER_TID;
+			dcfgs[dcfg_ix].src = cryptocop_source_dma;
+			cipher_active = 1;
+
+			if (next_ix == (oper.cipher_start + oper.cipher_len)){
+				cipher_done = 1;
+				dcfgs[dcfg_ix].last = 1;
+			} else {
+				dcfgs[dcfg_ix].last = 0;
+			}
+			dcfgs[dcfg_ix].next = descs[desc_ix].cfg;
+			descs[desc_ix].cfg = &dcfgs[dcfg_ix];
+			++dcfg_ix;
+		}
+		if (oper.do_digest && (oper.digest_start < next_ix) && (prev_ix < (oper.digest_start + oper.digest_len))) {
+			digest_active = 1;
+			dcfgs[dcfg_ix].tid = CRYPTOCOP_IOCTL_DIGEST_TID;
+			dcfgs[dcfg_ix].src = cryptocop_source_dma;
+			if (next_ix == (oper.digest_start + oper.digest_len)){
+				assert(!digest_done);
+				digest_done = 1;
+				dcfgs[dcfg_ix].last = 1;
+			} else {
+				dcfgs[dcfg_ix].last = 0;
+			}
+			dcfgs[dcfg_ix].next = descs[desc_ix].cfg;
+			descs[desc_ix].cfg = &dcfgs[dcfg_ix];
+			++dcfg_ix;
+		}
+		if (oper.do_csum && (oper.csum_start < next_ix) && (prev_ix < (oper.csum_start + oper.csum_len))){
+			csum_active = 1;
+			dcfgs[dcfg_ix].tid = CRYPTOCOP_IOCTL_CSUM_TID;
+			dcfgs[dcfg_ix].src = cryptocop_source_dma;
+			if (next_ix == (oper.csum_start + oper.csum_len)){
+				csum_done = 1;
+				dcfgs[dcfg_ix].last = 1;
+			} else {
+				dcfgs[dcfg_ix].last = 0;
+			}
+			dcfgs[dcfg_ix].next = descs[desc_ix].cfg;
+			descs[desc_ix].cfg = &dcfgs[dcfg_ix];
+			++dcfg_ix;
+		}
+		if (!descs[desc_ix].cfg){
+			DEBUG_API(printk("cryptocop_ioctl_process: data segment %d (%d to %d) had no active transforms\n", desc_ix, prev_ix, next_ix));
+			err = -EINVAL;
+			goto error_cleanup;
+		}
+		descs[desc_ix].next = &(descs[desc_ix]) + 1;
+		++desc_ix;
+		prev_ix = next_ix;
+		next_ix = next_cfg_change_ix(&oper, prev_ix);
+	}
+	if (desc_ix > 0){
+		descs[desc_ix-1].next = NULL;
+	} else {
+		descs[0].next = NULL;
+	}
+	if (oper.do_digest) {
+		DEBUG(printk("cryptocop_ioctl_process: mapping %d byte digest output to iovec %d\n", digest_length, iovix));
+		/* Add outdata iovec, length == <length of type of digest> */
+		cop->tfrm_op.outdata[iovix].iov_base = digest_result;
+		cop->tfrm_op.outdata[iovix].iov_len = digest_length;
+		++iovix;
+	}
+	if (oper.do_csum) {
+		/* Add outdata iovec, length == 2, the length of csum. */
+		DEBUG(printk("cryptocop_ioctl_process: mapping 2 byte csum output to iovec %d\n", iovix));
+		/* Add outdata iovec, length == <length of type of digest> */
+		cop->tfrm_op.outdata[iovix].iov_base = csum_result;
+		cop->tfrm_op.outdata[iovix].iov_len = 2;
+		++iovix;
+	}
+	if (oper.do_cipher) {
+		if (!map_pages_to_iovec(cop->tfrm_op.outdata, iovlen, &iovix, outpages, nooutpages, &pageix, &pageoffset, oper.cipher_outlen)){
+			DEBUG_API(printk("cryptocop_ioctl_process: failed to map pages to iovec.\n"));
+			err = -ENOSYS; /* This should be impossible barring bugs. */
+			goto error_cleanup;
+		}
+	}
+	DEBUG(printk("cryptocop_ioctl_process: setting cop->tfrm_op.outcount %d\n", iovix));
+	cop->tfrm_op.outcount = iovix;
+	assert(iovix <= (nooutpages + 6));
+
+	cop->sid = oper.ses_id;
+	cop->tfrm_op.desc = &descs[0];
+
+	DEBUG(printk("cryptocop_ioctl_process: inserting job, cb_data=0x%p\n", cop->cb_data));
+
+	if ((err = cryptocop_job_queue_insert_user_job(cop)) != 0) {
+		DEBUG_API(printk("cryptocop_ioctl_process: insert job %d\n", err));
+		err = -EINVAL;
+		goto error_cleanup;
+	}
+
+	DEBUG(printk("cryptocop_ioctl_process: begin wait for result\n"));
+
+	wait_event(cryptocop_ioc_process_wq, (jc->processed != 0));
+	DEBUG(printk("cryptocop_ioctl_process: end wait for result\n"));
+        if (!jc->processed){
+		printk(KERN_WARNING "cryptocop_ioctl_process: job not processed at completion\n");
+		err = -EIO;
+		goto error_cleanup;
+	}
+
+	/* Job process done.  Cipher output should already be correct in job so no post processing of outdata. */
+	DEBUG(printk("cryptocop_ioctl_process: operation_status = %d\n", cop->operation_status));
+	if (cop->operation_status == 0){
+		if (oper.do_digest){
+			DEBUG(printk("cryptocop_ioctl_process: copy %d bytes digest to user\n", digest_length));
+			err = copy_to_user((unsigned char*)crp_oper + offsetof(struct strcop_crypto_op, digest), digest_result, digest_length);
+			if (0 != err){
+				DEBUG_API(printk("cryptocop_ioctl_process: copy_to_user, digest length %d, err %d\n", digest_length, err));
+				err = -EFAULT;
+				goto error_cleanup;
+			}
+		}
+		if (oper.do_csum){
+			DEBUG(printk("cryptocop_ioctl_process: copy 2 bytes checksum to user\n"));
+			err = copy_to_user((unsigned char*)crp_oper + offsetof(struct strcop_crypto_op, csum), csum_result, 2);
+			if (0 != err){
+				DEBUG_API(printk("cryptocop_ioctl_process: copy_to_user, csum, err %d\n", err));
+				err = -EFAULT;
+				goto error_cleanup;
+			}
+		}
+		err = 0;
+	} else {
+		DEBUG(printk("cryptocop_ioctl_process: returning err = operation_status = %d\n", cop->operation_status));
+		err = cop->operation_status;
+	}
+
+ error_cleanup:
+	/* Release page caches. */
+	for (i = 0; i < noinpages; i++){
+		put_page(inpages[i]);
+	}
+	for (i = 0; i < nooutpages; i++){
+		int spdl_err;
+		/* Mark output pages dirty. */
+		spdl_err = set_page_dirty_lock(outpages[i]);
+		DEBUG(if (spdl_err < 0)printk("cryptocop_ioctl_process: set_page_dirty_lock returned %d\n", spdl_err));
+	}
+	for (i = 0; i < nooutpages; i++){
+		put_page(outpages[i]);
+	}
+
+	kfree(digest_result);
+	kfree(inpages);
+	kfree(outpages);
+	if (cop){
+		kfree(cop->tfrm_op.indata);
+		kfree(cop->tfrm_op.outdata);
+		kfree(cop);
+	}
+	kfree(jc);
+
+	DEBUG(print_lock_status());
+
+	return err;
+}
+
+
+static int cryptocop_ioctl_create_session(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	cryptocop_session_id             sid;
+	int                              err;
+	struct cryptocop_private         *dev;
+	struct strcop_session_op         *sess_op = (struct strcop_session_op *)arg;
+	struct strcop_session_op         sop;
+	struct cryptocop_transform_init  *tis = NULL;
+	struct cryptocop_transform_init  ti_cipher = {0};
+	struct cryptocop_transform_init  ti_digest = {0};
+	struct cryptocop_transform_init  ti_csum = {0};
+
+	if (!access_ok(VERIFY_WRITE, sess_op, sizeof(struct strcop_session_op)))
+		return -EFAULT;
+	err = copy_from_user(&sop, sess_op, sizeof(struct strcop_session_op));
+	if (err) return -EFAULT;
+	if (sop.cipher != cryptocop_cipher_none) {
+		if (!access_ok(VERIFY_READ, sop.key, sop.keylen)) return -EFAULT;
+	}
+	DEBUG(printk("cryptocop_ioctl_create_session, sess_op:\n"));
+
+	DEBUG(printk("\tcipher:%d\n"
+		     "\tcipher_mode:%d\n"
+		     "\tdigest:%d\n"
+		     "\tcsum:%d\n",
+		     (int)sop.cipher,
+		     (int)sop.cmode,
+		     (int)sop.digest,
+		     (int)sop.csum));
+
+	if (sop.cipher != cryptocop_cipher_none){
+		/* Init the cipher. */
+		switch (sop.cipher){
+		case cryptocop_cipher_des:
+			ti_cipher.alg = cryptocop_alg_des;
+			break;
+		case cryptocop_cipher_3des:
+			ti_cipher.alg = cryptocop_alg_3des;
+			break;
+		case cryptocop_cipher_aes:
+			ti_cipher.alg = cryptocop_alg_aes;
+			break;
+		default:
+			DEBUG_API(printk("create session, bad cipher algorithm %d\n", sop.cipher));
+			return -EINVAL;
+		};
+		DEBUG(printk("setting cipher transform %d\n", ti_cipher.alg));
+		copy_from_user(ti_cipher.key, sop.key, sop.keylen/8);
+		ti_cipher.keylen = sop.keylen;
+		switch (sop.cmode){
+		case cryptocop_cipher_mode_cbc:
+		case cryptocop_cipher_mode_ecb:
+			ti_cipher.cipher_mode = sop.cmode;
+			break;
+		default:
+			DEBUG_API(printk("create session, bad cipher mode %d\n", sop.cmode));
+			return -EINVAL;
+		}
+		DEBUG(printk("cryptocop_ioctl_create_session: setting CBC mode %d\n", ti_cipher.cipher_mode));
+		switch (sop.des3_mode){
+		case cryptocop_3des_eee:
+		case cryptocop_3des_eed:
+		case cryptocop_3des_ede:
+		case cryptocop_3des_edd:
+		case cryptocop_3des_dee:
+		case cryptocop_3des_ded:
+		case cryptocop_3des_dde:
+		case cryptocop_3des_ddd:
+			ti_cipher.tdes_mode = sop.des3_mode;
+			break;
+		default:
+			DEBUG_API(printk("create session, bad 3DES mode %d\n", sop.des3_mode));
+			return -EINVAL;
+		}
+		ti_cipher.tid = CRYPTOCOP_IOCTL_CIPHER_TID;
+		ti_cipher.next = tis;
+		tis = &ti_cipher;
+	} /* if (sop.cipher != cryptocop_cipher_none) */
+	if (sop.digest != cryptocop_digest_none){
+		DEBUG(printk("setting digest transform\n"));
+		switch (sop.digest){
+		case cryptocop_digest_md5:
+			ti_digest.alg = cryptocop_alg_md5;
+			break;
+		case cryptocop_digest_sha1:
+			ti_digest.alg = cryptocop_alg_sha1;
+			break;
+		default:
+			DEBUG_API(printk("create session, bad digest algorithm %d\n", sop.digest));
+			return -EINVAL;
+		}
+		ti_digest.tid = CRYPTOCOP_IOCTL_DIGEST_TID;
+		ti_digest.next = tis;
+		tis = &ti_digest;
+	} /* if (sop.digest != cryptocop_digest_none) */
+	if (sop.csum != cryptocop_csum_none){
+		DEBUG(printk("setting csum transform\n"));
+		switch (sop.csum){
+		case cryptocop_csum_le:
+		case cryptocop_csum_be:
+			ti_csum.csum_mode = sop.csum;
+			break;
+		default:
+			DEBUG_API(printk("create session, bad checksum algorithm %d\n", sop.csum));
+			return -EINVAL;
+		}
+		ti_csum.alg = cryptocop_alg_csum;
+		ti_csum.tid = CRYPTOCOP_IOCTL_CSUM_TID;
+		ti_csum.next = tis;
+		tis = &ti_csum;
+	} /* (sop.csum != cryptocop_csum_none) */
+	dev = kmalloc(sizeof(struct cryptocop_private), GFP_KERNEL);
+	if (!dev){
+		DEBUG_API(printk("create session, alloc dev\n"));
+		return -ENOMEM;
+	}
+
+	err = cryptocop_new_session(&sid, tis, GFP_KERNEL);
+	DEBUG({ if (err) printk("create session, cryptocop_new_session %d\n", err);});
+
+	if (err) {
+		kfree(dev);
+		return err;
+	}
+	sess_op->ses_id = sid;
+	dev->sid = sid;
+	dev->next = filp->private_data;
+	filp->private_data = dev;
+
+	return 0;
+}
+
+static long cryptocop_ioctl_unlocked(struct inode *inode,
+	struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	int err = 0;
+	if (_IOC_TYPE(cmd) != ETRAXCRYPTOCOP_IOCTYPE) {
+		DEBUG_API(printk("cryptocop_ioctl: wrong type\n"));
+		return -ENOTTY;
+	}
+	if (_IOC_NR(cmd) > CRYPTOCOP_IO_MAXNR){
+		return -ENOTTY;
+	}
+	/* Access check of the argument.  Some commands, e.g. create session and process op,
+	   needs additional checks.  Those are handled in the command handling functions. */
+	if (_IOC_DIR(cmd) & _IOC_READ)
+		err = !access_ok(VERIFY_WRITE, (void *)arg, _IOC_SIZE(cmd));
+	else if (_IOC_DIR(cmd) & _IOC_WRITE)
+		err = !access_ok(VERIFY_READ, (void *)arg, _IOC_SIZE(cmd));
+	if (err) return -EFAULT;
+
+	switch (cmd) {
+	case CRYPTOCOP_IO_CREATE_SESSION:
+		return cryptocop_ioctl_create_session(inode, filp, cmd, arg);
+	case CRYPTOCOP_IO_CLOSE_SESSION:
+		return cryptocop_ioctl_close_session(inode, filp, cmd, arg);
+	case CRYPTOCOP_IO_PROCESS_OP:
+		return cryptocop_ioctl_process(inode, filp, cmd, arg);
+	default:
+		DEBUG_API(printk("cryptocop_ioctl: unknown command\n"));
+		return -ENOTTY;
+	}
+	return 0;
+}
+
+static long
+cryptocop_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
+{
+       long ret;
+
+       mutex_lock(&cryptocop_mutex);
+       ret = cryptocop_ioctl_unlocked(file_inode(filp), filp, cmd, arg);
+       mutex_unlock(&cryptocop_mutex);
+
+       return ret;
+}
+
+
+#ifdef LDEBUG
+static void print_dma_descriptors(struct cryptocop_int_operation *iop)
+{
+	struct cryptocop_dma_desc *cdesc_out = iop->cdesc_out;
+	struct cryptocop_dma_desc *cdesc_in = iop->cdesc_in;
+	int                       i;
+
+	printk("print_dma_descriptors start\n");
+
+	printk("iop:\n");
+	printk("\tsid: 0x%lld\n", iop->sid);
+
+	printk("\tcdesc_out: 0x%p\n", iop->cdesc_out);
+	printk("\tcdesc_in: 0x%p\n", iop->cdesc_in);
+	printk("\tddesc_out: 0x%p\n", iop->ddesc_out);
+	printk("\tddesc_in: 0x%p\n", iop->ddesc_in);
+
+	printk("\niop->ctx_out: 0x%p phys: 0x%p\n", &iop->ctx_out, (char*)virt_to_phys(&iop->ctx_out));
+	printk("\tnext: 0x%p\n"
+	       "\tsaved_data: 0x%p\n"
+	       "\tsaved_data_buf: 0x%p\n",
+	       iop->ctx_out.next,
+	       iop->ctx_out.saved_data,
+	       iop->ctx_out.saved_data_buf);
+
+	printk("\niop->ctx_in: 0x%p phys: 0x%p\n", &iop->ctx_in, (char*)virt_to_phys(&iop->ctx_in));
+	printk("\tnext: 0x%p\n"
+	       "\tsaved_data: 0x%p\n"
+	       "\tsaved_data_buf: 0x%p\n",
+	       iop->ctx_in.next,
+	       iop->ctx_in.saved_data,
+	       iop->ctx_in.saved_data_buf);
+
+	i = 0;
+	while (cdesc_out) {
+		dma_descr_data *td;
+		printk("cdesc_out %d, desc=0x%p\n", i, cdesc_out->dma_descr);
+		printk("\n\tvirt_to_phys(desc): 0x%p\n", (char*)virt_to_phys(cdesc_out->dma_descr));
+		td = cdesc_out->dma_descr;
+		printk("\n\tbuf: 0x%p\n"
+		       "\tafter: 0x%p\n"
+		       "\tmd: 0x%04x\n"
+		       "\tnext: 0x%p\n",
+		       td->buf,
+		       td->after,
+		       td->md,
+		       td->next);
+		printk("flags:\n"
+		       "\twait:\t%d\n"
+		       "\teol:\t%d\n"
+		       "\touteop:\t%d\n"
+		       "\tineop:\t%d\n"
+		       "\tintr:\t%d\n",
+		       td->wait,
+		       td->eol,
+		       td->out_eop,
+		       td->in_eop,
+		       td->intr);
+		cdesc_out = cdesc_out->next;
+		i++;
+	}
+	i = 0;
+	while (cdesc_in) {
+		dma_descr_data *td;
+		printk("cdesc_in %d, desc=0x%p\n", i, cdesc_in->dma_descr);
+		printk("\n\tvirt_to_phys(desc): 0x%p\n", (char*)virt_to_phys(cdesc_in->dma_descr));
+		td = cdesc_in->dma_descr;
+		printk("\n\tbuf: 0x%p\n"
+		       "\tafter: 0x%p\n"
+		       "\tmd: 0x%04x\n"
+		       "\tnext: 0x%p\n",
+		       td->buf,
+		       td->after,
+		       td->md,
+		       td->next);
+		printk("flags:\n"
+		       "\twait:\t%d\n"
+		       "\teol:\t%d\n"
+		       "\touteop:\t%d\n"
+		       "\tineop:\t%d\n"
+		       "\tintr:\t%d\n",
+		       td->wait,
+		       td->eol,
+		       td->out_eop,
+		       td->in_eop,
+		       td->intr);
+		cdesc_in = cdesc_in->next;
+		i++;
+	}
+
+	printk("print_dma_descriptors end\n");
+}
+
+
+static void print_strcop_crypto_op(struct strcop_crypto_op *cop)
+{
+	printk("print_strcop_crypto_op, 0x%p\n", cop);
+
+	/* Indata. */
+	printk("indata=0x%p\n"
+	       "inlen=%d\n"
+	       "do_cipher=%d\n"
+	       "decrypt=%d\n"
+	       "cipher_explicit=%d\n"
+	       "cipher_start=%d\n"
+	       "cipher_len=%d\n"
+	       "outdata=0x%p\n"
+	       "outlen=%d\n",
+	       cop->indata,
+	       cop->inlen,
+	       cop->do_cipher,
+	       cop->decrypt,
+	       cop->cipher_explicit,
+	       cop->cipher_start,
+	       cop->cipher_len,
+	       cop->cipher_outdata,
+	       cop->cipher_outlen);
+
+	printk("do_digest=%d\n"
+	       "digest_start=%d\n"
+	       "digest_len=%d\n",
+	       cop->do_digest,
+	       cop->digest_start,
+	       cop->digest_len);
+
+	printk("do_csum=%d\n"
+	       "csum_start=%d\n"
+	       "csum_len=%d\n",
+	       cop->do_csum,
+	       cop->csum_start,
+	       cop->csum_len);
+}
+
+static void print_cryptocop_operation(struct cryptocop_operation *cop)
+{
+	struct cryptocop_desc      *d;
+	struct cryptocop_tfrm_cfg  *tc;
+	struct cryptocop_desc_cfg  *dc;
+	int                        i;
+
+	printk("print_cryptocop_operation, cop=0x%p\n\n", cop);
+	printk("sid: %lld\n", cop->sid);
+	printk("operation_status=%d\n"
+	       "use_dmalists=%d\n"
+	       "in_interrupt=%d\n"
+	       "fast_callback=%d\n",
+	       cop->operation_status,
+	       cop->use_dmalists,
+	       cop->in_interrupt,
+	       cop->fast_callback);
+
+	if (cop->use_dmalists){
+		print_user_dma_lists(&cop->list_op);
+	} else {
+		printk("cop->tfrm_op\n"
+		       "tfrm_cfg=0x%p\n"
+		       "desc=0x%p\n"
+		       "indata=0x%p\n"
+		       "incount=%d\n"
+		       "inlen=%d\n"
+		       "outdata=0x%p\n"
+		       "outcount=%d\n"
+		       "outlen=%d\n\n",
+		       cop->tfrm_op.tfrm_cfg,
+		       cop->tfrm_op.desc,
+		       cop->tfrm_op.indata,
+		       cop->tfrm_op.incount,
+		       cop->tfrm_op.inlen,
+		       cop->tfrm_op.outdata,
+		       cop->tfrm_op.outcount,
+		       cop->tfrm_op.outlen);
+
+		tc = cop->tfrm_op.tfrm_cfg;
+		while (tc){
+			printk("tfrm_cfg, 0x%p\n"
+			       "tid=%d\n"
+			       "flags=%d\n"
+			       "inject_ix=%d\n"
+			       "next=0x%p\n",
+			       tc,
+			       tc->tid,
+			       tc->flags,
+			       tc->inject_ix,
+			       tc->next);
+			tc = tc->next;
+		}
+		d = cop->tfrm_op.desc;
+		while (d){
+			printk("\n======================desc, 0x%p\n"
+			       "length=%d\n"
+			       "cfg=0x%p\n"
+			       "next=0x%p\n",
+			       d,
+			       d->length,
+			       d->cfg,
+			       d->next);
+			dc = d->cfg;
+			while (dc){
+				printk("=========desc_cfg, 0x%p\n"
+				       "tid=%d\n"
+				       "src=%d\n"
+				       "last=%d\n"
+				       "next=0x%p\n",
+				       dc,
+				       dc->tid,
+				       dc->src,
+				       dc->last,
+				       dc->next);
+				dc = dc->next;
+			}
+			d = d->next;
+		}
+		printk("\n====iniov\n");
+		for (i = 0; i < cop->tfrm_op.incount; i++){
+			printk("indata[%d]\n"
+			       "base=0x%p\n"
+			       "len=%d\n",
+			       i,
+			       cop->tfrm_op.indata[i].iov_base,
+			       cop->tfrm_op.indata[i].iov_len);
+		}
+		printk("\n====outiov\n");
+		for (i = 0; i < cop->tfrm_op.outcount; i++){
+			printk("outdata[%d]\n"
+			       "base=0x%p\n"
+			       "len=%d\n",
+			       i,
+			       cop->tfrm_op.outdata[i].iov_base,
+			       cop->tfrm_op.outdata[i].iov_len);
+		}
+	}
+	printk("------------end print_cryptocop_operation\n");
+}
+
+
+static void print_user_dma_lists(struct cryptocop_dma_list_operation *dma_op)
+{
+	dma_descr_data *dd;
+	int i;
+
+	printk("print_user_dma_lists, dma_op=0x%p\n", dma_op);
+
+	printk("out_data_buf = 0x%p, phys_to_virt(out_data_buf) = 0x%p\n", dma_op->out_data_buf, phys_to_virt((unsigned long int)dma_op->out_data_buf));
+	printk("in_data_buf = 0x%p, phys_to_virt(in_data_buf) = 0x%p\n", dma_op->in_data_buf, phys_to_virt((unsigned long int)dma_op->in_data_buf));
+
+	printk("##############outlist\n");
+	dd = phys_to_virt((unsigned long int)dma_op->outlist);
+	i = 0;
+	while (dd != NULL) {
+		printk("#%d phys_to_virt(desc) 0x%p\n", i, dd);
+		printk("\n\tbuf: 0x%p\n"
+		       "\tafter: 0x%p\n"
+		       "\tmd: 0x%04x\n"
+		       "\tnext: 0x%p\n",
+		       dd->buf,
+		       dd->after,
+		       dd->md,
+		       dd->next);
+		printk("flags:\n"
+		       "\twait:\t%d\n"
+		       "\teol:\t%d\n"
+		       "\touteop:\t%d\n"
+		       "\tineop:\t%d\n"
+		       "\tintr:\t%d\n",
+		       dd->wait,
+		       dd->eol,
+		       dd->out_eop,
+		       dd->in_eop,
+		       dd->intr);
+		if (dd->eol)
+			dd = NULL;
+		else
+			dd = phys_to_virt((unsigned long int)dd->next);
+		++i;
+	}
+
+	printk("##############inlist\n");
+	dd = phys_to_virt((unsigned long int)dma_op->inlist);
+	i = 0;
+	while (dd != NULL) {
+		printk("#%d phys_to_virt(desc) 0x%p\n", i, dd);
+		printk("\n\tbuf: 0x%p\n"
+		       "\tafter: 0x%p\n"
+		       "\tmd: 0x%04x\n"
+		       "\tnext: 0x%p\n",
+		       dd->buf,
+		       dd->after,
+		       dd->md,
+		       dd->next);
+		printk("flags:\n"
+		       "\twait:\t%d\n"
+		       "\teol:\t%d\n"
+		       "\touteop:\t%d\n"
+		       "\tineop:\t%d\n"
+		       "\tintr:\t%d\n",
+		       dd->wait,
+		       dd->eol,
+		       dd->out_eop,
+		       dd->in_eop,
+		       dd->intr);
+		if (dd->eol)
+			dd = NULL;
+		else
+			dd = phys_to_virt((unsigned long int)dd->next);
+		++i;
+	}
+}
+
+
+static void print_lock_status(void)
+{
+	printk("**********************print_lock_status\n");
+	printk("cryptocop_completed_jobs_lock %d\n", spin_is_locked(&cryptocop_completed_jobs_lock));
+	printk("cryptocop_job_queue_lock %d\n", spin_is_locked(&cryptocop_job_queue_lock));
+	printk("descr_pool_lock %d\n", spin_is_locked(&descr_pool_lock));
+	printk("cryptocop_sessions_lock %d\n", spin_is_locked(cryptocop_sessions_lock));
+	printk("running_job_lock %d\n", spin_is_locked(running_job_lock));
+	printk("cryptocop_process_lock %d\n", spin_is_locked(cryptocop_process_lock));
+}
+#endif /* LDEBUG */
+
+
+static const char cryptocop_name[] = "ETRAX FS stream co-processor";
+
+static int init_stream_coprocessor(void)
+{
+	int err;
+	int i;
+	static int initialized = 0;
+
+	if (initialized)
+		return 0;
+
+	initialized = 1;
+
+	printk("ETRAX FS stream co-processor driver v0.01, (c) 2003 Axis Communications AB\n");
+
+	err = register_chrdev(CRYPTOCOP_MAJOR, cryptocop_name, &cryptocop_fops);
+	if (err < 0) {
+		printk(KERN_ERR "stream co-processor: could not get major number.\n");
+		return err;
+	}
+
+	err = init_cryptocop();
+	if (err) {
+		(void)unregister_chrdev(CRYPTOCOP_MAJOR, cryptocop_name);
+		return err;
+	}
+	err = cryptocop_job_queue_init();
+	if (err) {
+		release_cryptocop();
+		(void)unregister_chrdev(CRYPTOCOP_MAJOR, cryptocop_name);
+		return err;
+	}
+	/* Init the descriptor pool. */
+	for (i = 0; i < CRYPTOCOP_DESCRIPTOR_POOL_SIZE - 1; i++) {
+		descr_pool[i].from_pool = 1;
+		descr_pool[i].next = &descr_pool[i + 1];
+	}
+	descr_pool[i].from_pool = 1;
+	descr_pool[i].next = NULL;
+	descr_pool_free_list = &descr_pool[0];
+	descr_pool_no_free = CRYPTOCOP_DESCRIPTOR_POOL_SIZE;
+
+	spin_lock_init(&cryptocop_completed_jobs_lock);
+	spin_lock_init(&cryptocop_job_queue_lock);
+	spin_lock_init(&descr_pool_lock);
+	spin_lock_init(&cryptocop_sessions_lock);
+	spin_lock_init(&running_job_lock);
+	spin_lock_init(&cryptocop_process_lock);
+
+	cryptocop_sessions = NULL;
+	next_sid = 1;
+
+	cryptocop_running_job = NULL;
+
+	printk("stream co-processor: init done.\n");
+	return 0;
+}
+
+static void __exit exit_stream_coprocessor(void)
+{
+	release_cryptocop();
+	cryptocop_job_queue_close();
+}
+
+module_init(init_stream_coprocessor);
+module_exit(exit_stream_coprocessor);
+
diff --git a/kernel/arch/cris/arch-v32/drivers/i2c.c b/kernel/arch/cris/arch-v32/drivers/i2c.c
new file mode 100644
index 000000000..3b2c82ce8
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/i2c.c
@@ -0,0 +1,751 @@
+/*!***************************************************************************
+*!
+*! FILE NAME  : i2c.c
+*!
+*! DESCRIPTION: implements an interface for IIC/I2C, both directly from other
+*!              kernel modules (i2c_writereg/readreg) and from userspace using
+*!              ioctl()'s
+*!
+*! Nov 30 1998  Torbjorn Eliasson  Initial version.
+*!              Bjorn Wesen        Elinux kernel version.
+*! Jan 14 2000  Johan Adolfsson    Fixed PB shadow register stuff -
+*!                                 don't use PB_I2C if DS1302 uses same bits,
+*!                                 use PB.
+*| June 23 2003 Pieter Grimmerink  Added 'i2c_sendnack'. i2c_readreg now
+*|                                 generates nack on last received byte,
+*|                                 instead of ack.
+*|                                 i2c_getack changed data level while clock
+*|                                 was high, causing DS75 to see  a stop condition
+*!
+*! ---------------------------------------------------------------------------
+*!
+*! (C) Copyright 1999-2007 Axis Communications AB, LUND, SWEDEN
+*!
+*!***************************************************************************/
+
+/****************** INCLUDE FILES SECTION ***********************************/
+
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/string.h>
+#include <linux/init.h>
+#include <linux/mutex.h>
+
+#include <asm/etraxi2c.h>
+
+#include <asm/io.h>
+#include <asm/delay.h>
+
+#include "i2c.h"
+
+/****************** I2C DEFINITION SECTION *************************/
+
+#define D(x)
+
+#define I2C_MAJOR 123  /* LOCAL/EXPERIMENTAL */
+static DEFINE_MUTEX(i2c_mutex);
+static const char i2c_name[] = "i2c";
+
+#define CLOCK_LOW_TIME            8
+#define CLOCK_HIGH_TIME           8
+#define START_CONDITION_HOLD_TIME 8
+#define STOP_CONDITION_HOLD_TIME  8
+#define ENABLE_OUTPUT 0x01
+#define ENABLE_INPUT 0x00
+#define I2C_CLOCK_HIGH 1
+#define I2C_CLOCK_LOW 0
+#define I2C_DATA_HIGH 1
+#define I2C_DATA_LOW 0
+
+#define i2c_enable()
+#define i2c_disable()
+
+/* enable or disable output-enable, to select output or input on the i2c bus */
+
+#define i2c_dir_out() crisv32_io_set_dir(&cris_i2c_data, crisv32_io_dir_out)
+#define i2c_dir_in() crisv32_io_set_dir(&cris_i2c_data, crisv32_io_dir_in)
+
+/* control the i2c clock and data signals */
+
+#define i2c_clk(x) crisv32_io_set(&cris_i2c_clk, x)
+#define i2c_data(x) crisv32_io_set(&cris_i2c_data, x)
+
+/* read a bit from the i2c interface */
+
+#define i2c_getbit() crisv32_io_rd(&cris_i2c_data)
+
+#define i2c_delay(usecs) udelay(usecs)
+
+static DEFINE_SPINLOCK(i2c_lock); /* Protect directions etc */
+
+/****************** VARIABLE SECTION ************************************/
+
+static struct crisv32_iopin cris_i2c_clk;
+static struct crisv32_iopin cris_i2c_data;
+
+/****************** FUNCTION DEFINITION SECTION *************************/
+
+
+/* generate i2c start condition */
+
+void
+i2c_start(void)
+{
+	/*
+	 * SCL=1 SDA=1
+	 */
+	i2c_dir_out();
+	i2c_delay(CLOCK_HIGH_TIME/6);
+	i2c_data(I2C_DATA_HIGH);
+	i2c_clk(I2C_CLOCK_HIGH);
+	i2c_delay(CLOCK_HIGH_TIME);
+	/*
+	 * SCL=1 SDA=0
+	 */
+	i2c_data(I2C_DATA_LOW);
+	i2c_delay(START_CONDITION_HOLD_TIME);
+	/*
+	 * SCL=0 SDA=0
+	 */
+	i2c_clk(I2C_CLOCK_LOW);
+	i2c_delay(CLOCK_LOW_TIME);
+}
+
+/* generate i2c stop condition */
+
+void
+i2c_stop(void)
+{
+	i2c_dir_out();
+
+	/*
+	 * SCL=0 SDA=0
+	 */
+	i2c_clk(I2C_CLOCK_LOW);
+	i2c_data(I2C_DATA_LOW);
+	i2c_delay(CLOCK_LOW_TIME*2);
+	/*
+	 * SCL=1 SDA=0
+	 */
+	i2c_clk(I2C_CLOCK_HIGH);
+	i2c_delay(CLOCK_HIGH_TIME*2);
+	/*
+	 * SCL=1 SDA=1
+	 */
+	i2c_data(I2C_DATA_HIGH);
+	i2c_delay(STOP_CONDITION_HOLD_TIME);
+
+	i2c_dir_in();
+}
+
+/* write a byte to the i2c interface */
+
+void
+i2c_outbyte(unsigned char x)
+{
+	int i;
+
+	i2c_dir_out();
+
+	for (i = 0; i < 8; i++) {
+		if (x & 0x80) {
+			i2c_data(I2C_DATA_HIGH);
+		} else {
+			i2c_data(I2C_DATA_LOW);
+		}
+
+		i2c_delay(CLOCK_LOW_TIME/2);
+		i2c_clk(I2C_CLOCK_HIGH);
+		i2c_delay(CLOCK_HIGH_TIME);
+		i2c_clk(I2C_CLOCK_LOW);
+		i2c_delay(CLOCK_LOW_TIME/2);
+		x <<= 1;
+	}
+	i2c_data(I2C_DATA_LOW);
+	i2c_delay(CLOCK_LOW_TIME/2);
+
+	/*
+	 * enable input
+	 */
+	i2c_dir_in();
+}
+
+/* read a byte from the i2c interface */
+
+unsigned char
+i2c_inbyte(void)
+{
+	unsigned char aBitByte = 0;
+	int i;
+
+	/* Switch off I2C to get bit */
+	i2c_disable();
+	i2c_dir_in();
+	i2c_delay(CLOCK_HIGH_TIME/2);
+
+	/* Get bit */
+	aBitByte |= i2c_getbit();
+
+	/* Enable I2C */
+	i2c_enable();
+	i2c_delay(CLOCK_LOW_TIME/2);
+
+	for (i = 1; i < 8; i++) {
+		aBitByte <<= 1;
+		/* Clock pulse */
+		i2c_clk(I2C_CLOCK_HIGH);
+		i2c_delay(CLOCK_HIGH_TIME);
+		i2c_clk(I2C_CLOCK_LOW);
+		i2c_delay(CLOCK_LOW_TIME);
+
+		/* Switch off I2C to get bit */
+		i2c_disable();
+		i2c_dir_in();
+		i2c_delay(CLOCK_HIGH_TIME/2);
+
+		/* Get bit */
+		aBitByte |= i2c_getbit();
+
+		/* Enable I2C */
+		i2c_enable();
+		i2c_delay(CLOCK_LOW_TIME/2);
+	}
+	i2c_clk(I2C_CLOCK_HIGH);
+	i2c_delay(CLOCK_HIGH_TIME);
+
+	/*
+	 * we leave the clock low, getbyte is usually followed
+	 * by sendack/nack, they assume the clock to be low
+	 */
+	i2c_clk(I2C_CLOCK_LOW);
+	return aBitByte;
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: i2c_getack
+*#
+*# DESCRIPTION  : checks if ack was received from ic2
+*#
+*#--------------------------------------------------------------------------*/
+
+int
+i2c_getack(void)
+{
+	int ack = 1;
+	/*
+	 * enable output
+	 */
+	i2c_dir_out();
+	/*
+	 * Release data bus by setting
+	 * data high
+	 */
+	i2c_data(I2C_DATA_HIGH);
+	/*
+	 * enable input
+	 */
+	i2c_dir_in();
+	i2c_delay(CLOCK_HIGH_TIME/4);
+	/*
+	 * generate ACK clock pulse
+	 */
+	i2c_clk(I2C_CLOCK_HIGH);
+#if 0
+	/*
+	 * Use PORT PB instead of I2C
+	 * for input. (I2C not working)
+	 */
+	i2c_clk(1);
+	i2c_data(1);
+	/*
+	 * switch off I2C
+	 */
+	i2c_data(1);
+	i2c_disable();
+	i2c_dir_in();
+#endif
+
+	/*
+	 * now wait for ack
+	 */
+	i2c_delay(CLOCK_HIGH_TIME/2);
+	/*
+	 * check for ack
+	 */
+	if (i2c_getbit())
+		ack = 0;
+	i2c_delay(CLOCK_HIGH_TIME/2);
+	if (!ack) {
+		if (!i2c_getbit()) /* receiver pulld SDA low */
+			ack = 1;
+		i2c_delay(CLOCK_HIGH_TIME/2);
+	}
+
+   /*
+    * our clock is high now, make sure data is low
+    * before we enable our output. If we keep data high
+    * and enable output, we would generate a stop condition.
+    */
+#if 0
+   i2c_data(I2C_DATA_LOW);
+
+	/*
+	 * end clock pulse
+	 */
+	i2c_enable();
+	i2c_dir_out();
+#endif
+	i2c_clk(I2C_CLOCK_LOW);
+	i2c_delay(CLOCK_HIGH_TIME/4);
+	/*
+	 * enable output
+	 */
+	i2c_dir_out();
+	/*
+	 * remove ACK clock pulse
+	 */
+	i2c_data(I2C_DATA_HIGH);
+	i2c_delay(CLOCK_LOW_TIME/2);
+	return ack;
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: I2C::sendAck
+*#
+*# DESCRIPTION  : Send ACK on received data
+*#
+*#--------------------------------------------------------------------------*/
+void
+i2c_sendack(void)
+{
+	/*
+	 * enable output
+	 */
+	i2c_delay(CLOCK_LOW_TIME);
+	i2c_dir_out();
+	/*
+	 * set ack pulse high
+	 */
+	i2c_data(I2C_DATA_LOW);
+	/*
+	 * generate clock pulse
+	 */
+	i2c_delay(CLOCK_HIGH_TIME/6);
+	i2c_clk(I2C_CLOCK_HIGH);
+	i2c_delay(CLOCK_HIGH_TIME);
+	i2c_clk(I2C_CLOCK_LOW);
+	i2c_delay(CLOCK_LOW_TIME/6);
+	/*
+	 * reset data out
+	 */
+	i2c_data(I2C_DATA_HIGH);
+	i2c_delay(CLOCK_LOW_TIME);
+
+	i2c_dir_in();
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: i2c_sendnack
+*#
+*# DESCRIPTION  : Sends NACK on received data
+*#
+*#--------------------------------------------------------------------------*/
+void
+i2c_sendnack(void)
+{
+	/*
+	 * enable output
+	 */
+	i2c_delay(CLOCK_LOW_TIME);
+	i2c_dir_out();
+	/*
+	 * set data high
+	 */
+	i2c_data(I2C_DATA_HIGH);
+	/*
+	 * generate clock pulse
+	 */
+	i2c_delay(CLOCK_HIGH_TIME/6);
+	i2c_clk(I2C_CLOCK_HIGH);
+	i2c_delay(CLOCK_HIGH_TIME);
+	i2c_clk(I2C_CLOCK_LOW);
+	i2c_delay(CLOCK_LOW_TIME);
+
+	i2c_dir_in();
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: i2c_write
+*#
+*# DESCRIPTION  : Writes a value to an I2C device
+*#
+*#--------------------------------------------------------------------------*/
+int
+i2c_write(unsigned char theSlave, void *data, size_t nbytes)
+{
+	int error, cntr = 3;
+	unsigned char bytes_wrote = 0;
+	unsigned char value;
+	unsigned long flags;
+
+	spin_lock_irqsave(&i2c_lock, flags);
+
+	do {
+		error = 0;
+
+		i2c_start();
+		/*
+		 * send slave address
+		 */
+		i2c_outbyte((theSlave & 0xfe));
+		/*
+		 * wait for ack
+		 */
+		if (!i2c_getack())
+			error = 1;
+		/*
+		 * send data
+		 */
+		for (bytes_wrote = 0; bytes_wrote < nbytes; bytes_wrote++) {
+			memcpy(&value, data + bytes_wrote, sizeof value);
+			i2c_outbyte(value);
+			/*
+			 * now it's time to wait for ack
+			 */
+			if (!i2c_getack())
+				error |= 4;
+		}
+		/*
+		 * end byte stream
+		 */
+		i2c_stop();
+
+	} while (error && cntr--);
+
+	i2c_delay(CLOCK_LOW_TIME);
+
+	spin_unlock_irqrestore(&i2c_lock, flags);
+
+	return -error;
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: i2c_read
+*#
+*# DESCRIPTION  : Reads a value from an I2C device
+*#
+*#--------------------------------------------------------------------------*/
+int
+i2c_read(unsigned char theSlave, void *data, size_t nbytes)
+{
+	unsigned char b = 0;
+	unsigned char bytes_read = 0;
+	int error, cntr = 3;
+	unsigned long flags;
+
+	spin_lock_irqsave(&i2c_lock, flags);
+
+	do {
+		error = 0;
+		memset(data, 0, nbytes);
+		/*
+		 * generate start condition
+		 */
+		i2c_start();
+		/*
+		 * send slave address
+		 */
+		i2c_outbyte((theSlave | 0x01));
+		/*
+		 * wait for ack
+		 */
+		if (!i2c_getack())
+			error = 1;
+		/*
+		 * fetch data
+		 */
+		for (bytes_read = 0; bytes_read < nbytes; bytes_read++) {
+			b = i2c_inbyte();
+			memcpy(data + bytes_read, &b, sizeof b);
+
+			if (bytes_read < (nbytes - 1))
+				i2c_sendack();
+		}
+		/*
+		 * last received byte needs to be nacked
+		 * instead of acked
+		 */
+		i2c_sendnack();
+		/*
+		 * end sequence
+		 */
+		i2c_stop();
+	} while (error && cntr--);
+
+	spin_unlock_irqrestore(&i2c_lock, flags);
+
+	return -error;
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: i2c_writereg
+*#
+*# DESCRIPTION  : Writes a value to an I2C device
+*#
+*#--------------------------------------------------------------------------*/
+int
+i2c_writereg(unsigned char theSlave, unsigned char theReg,
+	     unsigned char theValue)
+{
+	int error, cntr = 3;
+	unsigned long flags;
+
+	spin_lock_irqsave(&i2c_lock, flags);
+
+	do {
+		error = 0;
+
+		i2c_start();
+		/*
+		 * send slave address
+		 */
+		i2c_outbyte((theSlave & 0xfe));
+		/*
+		 * wait for ack
+		 */
+		if(!i2c_getack())
+			error = 1;
+		/*
+		 * now select register
+		 */
+		i2c_dir_out();
+		i2c_outbyte(theReg);
+		/*
+		 * now it's time to wait for ack
+		 */
+		if(!i2c_getack())
+			error |= 2;
+		/*
+		 * send register register data
+		 */
+		i2c_outbyte(theValue);
+		/*
+		 * now it's time to wait for ack
+		 */
+		if(!i2c_getack())
+			error |= 4;
+		/*
+		 * end byte stream
+		 */
+		i2c_stop();
+	} while(error && cntr--);
+
+	i2c_delay(CLOCK_LOW_TIME);
+
+	spin_unlock_irqrestore(&i2c_lock, flags);
+
+	return -error;
+}
+
+/*#---------------------------------------------------------------------------
+*#
+*# FUNCTION NAME: i2c_readreg
+*#
+*# DESCRIPTION  : Reads a value from the decoder registers.
+*#
+*#--------------------------------------------------------------------------*/
+unsigned char
+i2c_readreg(unsigned char theSlave, unsigned char theReg)
+{
+	unsigned char b = 0;
+	int error, cntr = 3;
+	unsigned long flags;
+
+	spin_lock_irqsave(&i2c_lock, flags);
+
+	do {
+		error = 0;
+		/*
+		 * generate start condition
+		 */
+		i2c_start();
+
+		/*
+		 * send slave address
+		 */
+		i2c_outbyte((theSlave & 0xfe));
+		/*
+		 * wait for ack
+		 */
+		if(!i2c_getack())
+			error = 1;
+		/*
+		 * now select register
+		 */
+		i2c_dir_out();
+		i2c_outbyte(theReg);
+		/*
+		 * now it's time to wait for ack
+		 */
+		if(!i2c_getack())
+			error |= 2;
+		/*
+		 * repeat start condition
+		 */
+		i2c_delay(CLOCK_LOW_TIME);
+		i2c_start();
+		/*
+		 * send slave address
+		 */
+		i2c_outbyte(theSlave | 0x01);
+		/*
+		 * wait for ack
+		 */
+		if(!i2c_getack())
+			error |= 4;
+		/*
+		 * fetch register
+		 */
+		b = i2c_inbyte();
+		/*
+		 * last received byte needs to be nacked
+		 * instead of acked
+		 */
+		i2c_sendnack();
+		/*
+		 * end sequence
+		 */
+		i2c_stop();
+
+	} while(error && cntr--);
+
+	spin_unlock_irqrestore(&i2c_lock, flags);
+
+	return b;
+}
+
+static int
+i2c_open(struct inode *inode, struct file *filp)
+{
+	return 0;
+}
+
+static int
+i2c_release(struct inode *inode, struct file *filp)
+{
+	return 0;
+}
+
+/* Main device API. ioctl's to write or read to/from i2c registers.
+ */
+
+static long
+i2c_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	int ret;
+	if(_IOC_TYPE(cmd) != ETRAXI2C_IOCTYPE) {
+		return -ENOTTY;
+	}
+
+	switch (_IOC_NR(cmd)) {
+		case I2C_WRITEREG:
+			/* write to an i2c slave */
+			D(printk("i2cw %d %d %d\n",
+				 I2C_ARGSLAVE(arg),
+				 I2C_ARGREG(arg),
+				 I2C_ARGVALUE(arg)));
+
+			mutex_lock(&i2c_mutex);
+			ret = i2c_writereg(I2C_ARGSLAVE(arg),
+					    I2C_ARGREG(arg),
+					    I2C_ARGVALUE(arg));
+			mutex_unlock(&i2c_mutex);
+			return ret;
+
+		case I2C_READREG:
+		{
+			unsigned char val;
+			/* read from an i2c slave */
+			D(printk("i2cr %d %d ",
+				I2C_ARGSLAVE(arg),
+				I2C_ARGREG(arg)));
+			mutex_lock(&i2c_mutex);
+			val = i2c_readreg(I2C_ARGSLAVE(arg), I2C_ARGREG(arg));
+			mutex_unlock(&i2c_mutex);
+			D(printk("= %d\n", val));
+			return val;
+		}
+		default:
+			return -EINVAL;
+
+	}
+
+	return 0;
+}
+
+static const struct file_operations i2c_fops = {
+	.owner		= THIS_MODULE,
+	.unlocked_ioctl = i2c_ioctl,
+	.open		= i2c_open,
+	.release	= i2c_release,
+	.llseek		= noop_llseek,
+};
+
+static int __init i2c_init(void)
+{
+	static int res;
+	static int first = 1;
+
+	if (!first)
+		return res;
+
+	first = 0;
+
+	/* Setup and enable the DATA and CLK pins */
+
+	res = crisv32_io_get_name(&cris_i2c_data,
+		CONFIG_ETRAX_V32_I2C_DATA_PORT);
+	if (res < 0)
+		return res;
+
+	res = crisv32_io_get_name(&cris_i2c_clk, CONFIG_ETRAX_V32_I2C_CLK_PORT);
+	crisv32_io_set_dir(&cris_i2c_clk, crisv32_io_dir_out);
+
+	return res;
+}
+
+
+static int __init i2c_register(void)
+{
+	int res;
+
+	res = i2c_init();
+	if (res < 0)
+		return res;
+
+	/* register char device */
+
+	res = register_chrdev(I2C_MAJOR, i2c_name, &i2c_fops);
+	if (res < 0) {
+		printk(KERN_ERR "i2c: couldn't get a major number.\n");
+		return res;
+	}
+
+	printk(KERN_INFO
+		"I2C driver v2.2, (c) 1999-2007 Axis Communications AB\n");
+
+	return 0;
+}
+/* this makes sure that i2c_init is called during boot */
+module_init(i2c_register);
+
+/****************** END OF FILE i2c.c ********************************/
diff --git a/kernel/arch/cris/arch-v32/drivers/i2c.h b/kernel/arch/cris/arch-v32/drivers/i2c.h
new file mode 100644
index 000000000..d9cc856f8
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/i2c.h
@@ -0,0 +1,16 @@
+
+#include <linux/init.h>
+
+/* High level I2C actions */
+int i2c_write(unsigned char theSlave, void *data, size_t nbytes);
+int i2c_read(unsigned char theSlave, void *data, size_t nbytes);
+int i2c_writereg(unsigned char theSlave, unsigned char theReg, unsigned char theValue);
+unsigned char i2c_readreg(unsigned char theSlave, unsigned char theReg);
+
+/* Low level I2C */
+void i2c_start(void);
+void i2c_stop(void);
+void i2c_outbyte(unsigned char x);
+unsigned char i2c_inbyte(void);
+int i2c_getack(void);
+void i2c_sendack(void);
diff --git a/kernel/arch/cris/arch-v32/drivers/iop_fw_load.c b/kernel/arch/cris/arch-v32/drivers/iop_fw_load.c
new file mode 100644
index 000000000..2f8ea0f7a
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/iop_fw_load.c
@@ -0,0 +1,230 @@
+/*
+ * Firmware loader for ETRAX FS IO-Processor
+ *
+ * Copyright (C) 2004  Axis Communications AB
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/device.h>
+#include <linux/firmware.h>
+
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/reg_map.h>
+#include <hwregs/iop/iop_reg_space.h>
+#include <hwregs/iop/iop_mpu_macros.h>
+#include <hwregs/iop/iop_mpu_defs.h>
+#include <hwregs/iop/iop_spu_defs.h>
+#include <hwregs/iop/iop_sw_cpu_defs.h>
+
+#define IOP_TIMEOUT 100
+
+#error "This driver is broken with regard to its driver core usage."
+#error "Please contact <greg@kroah.com> for details on how to fix it properly."
+
+static struct device iop_spu_device[2] = {
+	{ .init_name =     "iop-spu0", },
+	{ .init_name =     "iop-spu1", },
+};
+
+static struct device iop_mpu_device = {
+	.init_name =       "iop-mpu",
+};
+
+static int wait_mpu_idle(void)
+{
+	reg_iop_mpu_r_stat mpu_stat;
+	unsigned int timeout = IOP_TIMEOUT;
+
+	do {
+		mpu_stat = REG_RD(iop_mpu, regi_iop_mpu, r_stat);
+	} while (mpu_stat.instr_reg_busy == regk_iop_mpu_yes && --timeout > 0);
+	if (timeout == 0) {
+		printk(KERN_ERR "Timeout waiting for MPU to be idle\n");
+		return -EBUSY;
+	}
+	return 0;
+}
+
+int iop_fw_load_spu(const unsigned char *fw_name, unsigned int spu_inst)
+{
+	reg_iop_sw_cpu_rw_mc_ctrl mc_ctrl = {
+		.wr_spu0_mem =    regk_iop_sw_cpu_no,
+		.wr_spu1_mem =    regk_iop_sw_cpu_no,
+		.size =           4,
+		.cmd =            regk_iop_sw_cpu_reg_copy,
+		.keep_owner =     regk_iop_sw_cpu_yes
+	};
+	reg_iop_spu_rw_ctrl spu_ctrl = {
+		.en  =            regk_iop_spu_no,
+		.fsm =            regk_iop_spu_no,
+	};
+	reg_iop_sw_cpu_r_mc_stat mc_stat;
+        const struct firmware *fw_entry;
+	u32 *data;
+	unsigned int timeout;
+	int retval, i;
+
+	if (spu_inst > 1)
+		return -ENODEV;
+
+	/* get firmware */
+	retval = request_firmware(&fw_entry,
+				  fw_name,
+				  &iop_spu_device[spu_inst]);
+	if (retval != 0)
+	{
+		printk(KERN_ERR
+		       "iop_load_spu: Failed to load firmware \"%s\"\n",
+		       fw_name);
+		return retval;
+	}
+	data = (u32 *) fw_entry->data;
+
+	/* acquire ownership of memory controller */
+	switch (spu_inst) {
+	case 0:
+		mc_ctrl.wr_spu0_mem = regk_iop_sw_cpu_yes;
+		REG_WR(iop_spu, regi_iop_spu0, rw_ctrl, spu_ctrl);
+		break;
+	case 1:
+		mc_ctrl.wr_spu1_mem = regk_iop_sw_cpu_yes;
+		REG_WR(iop_spu, regi_iop_spu1, rw_ctrl, spu_ctrl);
+		break;
+	}
+	timeout = IOP_TIMEOUT;
+	do {
+		REG_WR(iop_sw_cpu, regi_iop_sw_cpu, rw_mc_ctrl, mc_ctrl);
+		mc_stat = REG_RD(iop_sw_cpu, regi_iop_sw_cpu, r_mc_stat);
+	} while (mc_stat.owned_by_cpu == regk_iop_sw_cpu_no && --timeout > 0);
+	if (timeout == 0) {
+		printk(KERN_ERR "Timeout waiting to acquire MC\n");
+		retval = -EBUSY;
+		goto out;
+	}
+
+	/* write to SPU memory */
+	for (i = 0; i < (fw_entry->size/4); i++) {
+		switch (spu_inst) {
+		case 0:
+			REG_WR_INT(iop_spu, regi_iop_spu0, rw_seq_pc, (i*4));
+			break;
+		case 1:
+			REG_WR_INT(iop_spu, regi_iop_spu1, rw_seq_pc, (i*4));
+			break;
+		}
+		REG_WR_INT(iop_sw_cpu, regi_iop_sw_cpu, rw_mc_data, *data);
+		data++;
+	}
+
+	/* release ownership of memory controller */
+	(void) REG_RD(iop_sw_cpu, regi_iop_sw_cpu, rs_mc_data);
+
+ out:
+	release_firmware(fw_entry);
+	return retval;
+}
+
+int iop_fw_load_mpu(unsigned char *fw_name)
+{
+	const unsigned int start_addr = 0;
+	reg_iop_mpu_rw_ctrl mpu_ctrl;
+        const struct firmware *fw_entry;
+	u32 *data;
+	int retval, i;
+
+	/* get firmware */
+	retval = request_firmware(&fw_entry, fw_name, &iop_mpu_device);
+	if (retval != 0)
+	{
+		printk(KERN_ERR
+		       "iop_load_spu: Failed to load firmware \"%s\"\n",
+		       fw_name);
+		return retval;
+	}
+	data = (u32 *) fw_entry->data;
+
+	/* disable MPU */
+	mpu_ctrl.en = regk_iop_mpu_no;
+	REG_WR(iop_mpu, regi_iop_mpu, rw_ctrl, mpu_ctrl);
+	/* put start address in R0 */
+	REG_WR_VECT(iop_mpu, regi_iop_mpu, rw_r, 0, start_addr);
+	/* write to memory by executing 'SWX i, 4, R0' for each word */
+	if ((retval = wait_mpu_idle()) != 0)
+		goto out;
+	REG_WR(iop_mpu, regi_iop_mpu, rw_instr, MPU_SWX_IIR_INSTR(0, 4, 0));
+	for (i = 0; i < (fw_entry->size / 4); i++) {
+		REG_WR_INT(iop_mpu, regi_iop_mpu, rw_immediate, *data);
+		if ((retval = wait_mpu_idle()) != 0)
+			goto out;
+		data++;
+	}
+
+ out:
+	release_firmware(fw_entry);
+	return retval;
+}
+
+int iop_start_mpu(unsigned int start_addr)
+{
+	reg_iop_mpu_rw_ctrl mpu_ctrl = { .en = regk_iop_mpu_yes };
+	int retval;
+
+	/* disable MPU */
+	if ((retval = wait_mpu_idle()) != 0)
+		goto out;
+	REG_WR(iop_mpu, regi_iop_mpu, rw_instr, MPU_HALT());
+	if ((retval = wait_mpu_idle()) != 0)
+		goto out;
+	/* set PC and wait for it to bite */
+	if ((retval = wait_mpu_idle()) != 0)
+		goto out;
+	REG_WR_INT(iop_mpu, regi_iop_mpu, rw_instr, MPU_BA_I(start_addr));
+	if ((retval = wait_mpu_idle()) != 0)
+		goto out;
+	/* make sure the MPU starts executing with interrupts disabled */
+	REG_WR(iop_mpu, regi_iop_mpu, rw_instr, MPU_DI());
+	if ((retval = wait_mpu_idle()) != 0)
+		goto out;
+	/* enable MPU */
+	REG_WR(iop_mpu, regi_iop_mpu, rw_ctrl, mpu_ctrl);
+ out:
+	return retval;
+}
+
+static int __init iop_fw_load_init(void)
+{
+#if 0
+	/*
+	 * static struct devices can not be added directly to sysfs by ignoring
+	 * the driver model infrastructure.  To fix this properly, please use
+	 * the platform_bus to register these devices to be able to properly
+	 * use the firmware infrastructure.
+	 */
+	device_initialize(&iop_spu_device[0]);
+	kobject_set_name(&iop_spu_device[0].kobj, "iop-spu0");
+	kobject_add(&iop_spu_device[0].kobj);
+	device_initialize(&iop_spu_device[1]);
+	kobject_set_name(&iop_spu_device[1].kobj, "iop-spu1");
+	kobject_add(&iop_spu_device[1].kobj);
+	device_initialize(&iop_mpu_device);
+	kobject_set_name(&iop_mpu_device.kobj, "iop-mpu");
+	kobject_add(&iop_mpu_device.kobj);
+#endif
+	return 0;
+}
+
+static void __exit iop_fw_load_exit(void)
+{
+}
+
+module_init(iop_fw_load_init);
+module_exit(iop_fw_load_exit);
+
+MODULE_DESCRIPTION("ETRAX FS IO-Processor Firmware Loader");
+MODULE_LICENSE("GPL");
+
+EXPORT_SYMBOL(iop_fw_load_spu);
+EXPORT_SYMBOL(iop_fw_load_mpu);
+EXPORT_SYMBOL(iop_start_mpu);
diff --git a/kernel/arch/cris/arch-v32/drivers/mach-a3/Makefile b/kernel/arch/cris/arch-v32/drivers/mach-a3/Makefile
new file mode 100644
index 000000000..5c6d2a2a0
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/mach-a3/Makefile
@@ -0,0 +1,6 @@
+#
+# Makefile for Etrax-specific drivers
+#
+
+obj-$(CONFIG_ETRAX_NANDFLASH)   += nandflash.o
+obj-$(CONFIG_ETRAX_GPIO)        += gpio.o
diff --git a/kernel/arch/cris/arch-v32/drivers/mach-a3/gpio.c b/kernel/arch/cris/arch-v32/drivers/mach-a3/gpio.c
new file mode 100644
index 000000000..74f9fe809
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/mach-a3/gpio.c
@@ -0,0 +1,999 @@
+/*
+ * Artec-3 general port I/O device
+ *
+ * Copyright (c) 2007 Axis Communications AB
+ *
+ * Authors:    Bjorn Wesen      (initial version)
+ *             Ola Knutsson     (LED handling)
+ *             Johan Adolfsson  (read/set directions, write, port G,
+ *                               port to ETRAX FS.
+ *             Ricard Wanderlof (PWM for Artpec-3)
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/ioport.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/string.h>
+#include <linux/poll.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/spinlock.h>
+#include <linux/mutex.h>
+
+#include <asm/etraxgpio.h>
+#include <hwregs/reg_map.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/gio_defs.h>
+#include <hwregs/intr_vect_defs.h>
+#include <asm/io.h>
+#include <asm/irq.h>
+#include <mach/pinmux.h>
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+#include "../i2c.h"
+
+#define VIRT_I2C_ADDR 0x40
+#endif
+
+/* The following gio ports on ARTPEC-3 is available:
+ * pa 32 bits
+ * pb 32 bits
+ * pc 16 bits
+ * each port has a rw_px_dout, r_px_din and rw_px_oe register.
+ */
+
+#define GPIO_MAJOR 120  /* experimental MAJOR number */
+
+#define I2C_INTERRUPT_BITS 0x300 /* i2c0_done and i2c1_done bits */
+
+#define D(x)
+
+#if 0
+static int dp_cnt;
+#define DP(x) \
+	do { \
+		dp_cnt++; \
+		if (dp_cnt % 1000 == 0) \
+			x; \
+	} while (0)
+#else
+#define DP(x)
+#endif
+
+static DEFINE_MUTEX(gpio_mutex);
+static char gpio_name[] = "etrax gpio";
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static int virtual_gpio_ioctl(struct file *file, unsigned int cmd,
+			      unsigned long arg);
+#endif
+static long gpio_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
+static ssize_t gpio_write(struct file *file, const char __user *buf,
+	size_t count, loff_t *off);
+static int gpio_open(struct inode *inode, struct file *filp);
+static int gpio_release(struct inode *inode, struct file *filp);
+static unsigned int gpio_poll(struct file *filp,
+	struct poll_table_struct *wait);
+
+/* private data per open() of this driver */
+
+struct gpio_private {
+	struct gpio_private *next;
+	/* The IO_CFG_WRITE_MODE_VALUE only support 8 bits: */
+	unsigned char clk_mask;
+	unsigned char data_mask;
+	unsigned char write_msb;
+	unsigned char pad1;
+	/* These fields are generic */
+	unsigned long highalarm, lowalarm;
+	wait_queue_head_t alarm_wq;
+	int minor;
+};
+
+static void gpio_set_alarm(struct gpio_private *priv);
+static int gpio_leds_ioctl(unsigned int cmd, unsigned long arg);
+static int gpio_pwm_ioctl(struct gpio_private *priv, unsigned int cmd,
+	unsigned long arg);
+
+
+/* linked list of alarms to check for */
+
+static struct gpio_private *alarmlist;
+
+static int wanted_interrupts;
+
+static DEFINE_SPINLOCK(gpio_lock);
+
+#define NUM_PORTS (GPIO_MINOR_LAST+1)
+#define GIO_REG_RD_ADDR(reg) \
+	(unsigned long *)(regi_gio + REG_RD_ADDR_gio_##reg)
+#define GIO_REG_WR_ADDR(reg) \
+	(unsigned long *)(regi_gio + REG_WR_ADDR_gio_##reg)
+static unsigned long led_dummy;
+static unsigned long port_d_dummy;	/* Only input on Artpec-3 */
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static unsigned long port_e_dummy;	/* Non existent on Artpec-3 */
+static unsigned long virtual_dummy;
+static unsigned long virtual_rw_pv_oe = CONFIG_ETRAX_DEF_GIO_PV_OE;
+static unsigned short cached_virtual_gpio_read;
+#endif
+
+static unsigned long *data_out[NUM_PORTS] = {
+	GIO_REG_WR_ADDR(rw_pa_dout),
+	GIO_REG_WR_ADDR(rw_pb_dout),
+	&led_dummy,
+	GIO_REG_WR_ADDR(rw_pc_dout),
+	&port_d_dummy,
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	&port_e_dummy,
+	&virtual_dummy,
+#endif
+};
+
+static unsigned long *data_in[NUM_PORTS] = {
+	GIO_REG_RD_ADDR(r_pa_din),
+	GIO_REG_RD_ADDR(r_pb_din),
+	&led_dummy,
+	GIO_REG_RD_ADDR(r_pc_din),
+	GIO_REG_RD_ADDR(r_pd_din),
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	&port_e_dummy,
+	&virtual_dummy,
+#endif
+};
+
+static unsigned long changeable_dir[NUM_PORTS] = {
+	CONFIG_ETRAX_PA_CHANGEABLE_DIR,
+	CONFIG_ETRAX_PB_CHANGEABLE_DIR,
+	0,
+	CONFIG_ETRAX_PC_CHANGEABLE_DIR,
+	0,
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	0,
+	CONFIG_ETRAX_PV_CHANGEABLE_DIR,
+#endif
+};
+
+static unsigned long changeable_bits[NUM_PORTS] = {
+	CONFIG_ETRAX_PA_CHANGEABLE_BITS,
+	CONFIG_ETRAX_PB_CHANGEABLE_BITS,
+	0,
+	CONFIG_ETRAX_PC_CHANGEABLE_BITS,
+	0,
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	0,
+	CONFIG_ETRAX_PV_CHANGEABLE_BITS,
+#endif
+};
+
+static unsigned long *dir_oe[NUM_PORTS] = {
+	GIO_REG_WR_ADDR(rw_pa_oe),
+	GIO_REG_WR_ADDR(rw_pb_oe),
+	&led_dummy,
+	GIO_REG_WR_ADDR(rw_pc_oe),
+	&port_d_dummy,
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	&port_e_dummy,
+	&virtual_rw_pv_oe,
+#endif
+};
+
+static void gpio_set_alarm(struct gpio_private *priv)
+{
+	int bit;
+	int intr_cfg;
+	int mask;
+	int pins;
+	unsigned long flags;
+
+	spin_lock_irqsave(&gpio_lock, flags);
+	intr_cfg = REG_RD_INT(gio, regi_gio, rw_intr_cfg);
+	pins = REG_RD_INT(gio, regi_gio, rw_intr_pins);
+	mask = REG_RD_INT(gio, regi_gio, rw_intr_mask) & I2C_INTERRUPT_BITS;
+
+	for (bit = 0; bit < 32; bit++) {
+		int intr = bit % 8;
+		int pin = bit / 8;
+		if (priv->minor < GPIO_MINOR_LEDS)
+			pin += priv->minor * 4;
+		else
+			pin += (priv->minor - 1) * 4;
+
+		if (priv->highalarm & (1<<bit)) {
+			intr_cfg |= (regk_gio_hi << (intr * 3));
+			mask |= 1 << intr;
+			wanted_interrupts = mask & 0xff;
+			pins |= pin << (intr * 4);
+		} else if (priv->lowalarm & (1<<bit)) {
+			intr_cfg |= (regk_gio_lo << (intr * 3));
+			mask |= 1 << intr;
+			wanted_interrupts = mask & 0xff;
+			pins |= pin << (intr * 4);
+		}
+	}
+
+	REG_WR_INT(gio, regi_gio, rw_intr_cfg, intr_cfg);
+	REG_WR_INT(gio, regi_gio, rw_intr_pins, pins);
+	REG_WR_INT(gio, regi_gio, rw_intr_mask, mask);
+
+	spin_unlock_irqrestore(&gpio_lock, flags);
+}
+
+static unsigned int gpio_poll(struct file *file, struct poll_table_struct *wait)
+{
+	unsigned int mask = 0;
+	struct gpio_private *priv = file->private_data;
+	unsigned long data;
+	unsigned long tmp;
+
+	if (priv->minor >= GPIO_MINOR_PWM0 &&
+	    priv->minor <= GPIO_MINOR_LAST_PWM)
+		return 0;
+
+	poll_wait(file, &priv->alarm_wq, wait);
+	if (priv->minor <= GPIO_MINOR_D) {
+		data = readl(data_in[priv->minor]);
+		REG_WR_INT(gio, regi_gio, rw_ack_intr, wanted_interrupts);
+		tmp = REG_RD_INT(gio, regi_gio, rw_intr_mask);
+		tmp &= I2C_INTERRUPT_BITS;
+		tmp |= wanted_interrupts;
+		REG_WR_INT(gio, regi_gio, rw_intr_mask, tmp);
+	} else
+		return 0;
+
+	if ((data & priv->highalarm) || (~data & priv->lowalarm))
+		mask = POLLIN|POLLRDNORM;
+
+	DP(printk(KERN_DEBUG "gpio_poll ready: mask 0x%08X\n", mask));
+	return mask;
+}
+
+static irqreturn_t gpio_interrupt(int irq, void *dev_id)
+{
+	reg_gio_rw_intr_mask intr_mask;
+	reg_gio_r_masked_intr masked_intr;
+	reg_gio_rw_ack_intr ack_intr;
+	unsigned long flags;
+	unsigned long tmp;
+	unsigned long tmp2;
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	unsigned char enable_gpiov_ack = 0;
+#endif
+
+	/* Find what PA interrupts are active */
+	masked_intr = REG_RD(gio, regi_gio, r_masked_intr);
+	tmp = REG_TYPE_CONV(unsigned long, reg_gio_r_masked_intr, masked_intr);
+
+	/* Find those that we have enabled */
+	spin_lock_irqsave(&gpio_lock, flags);
+	tmp &= wanted_interrupts;
+	spin_unlock_irqrestore(&gpio_lock, flags);
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	/* Something changed on virtual GPIO. Interrupt is acked by
+	 * reading the device.
+	 */
+	if (tmp & (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN)) {
+		i2c_read(VIRT_I2C_ADDR, (void *)&cached_virtual_gpio_read,
+			sizeof(cached_virtual_gpio_read));
+		enable_gpiov_ack = 1;
+	}
+#endif
+
+	/* Ack them */
+	ack_intr = REG_TYPE_CONV(reg_gio_rw_ack_intr, unsigned long, tmp);
+	REG_WR(gio, regi_gio, rw_ack_intr, ack_intr);
+
+	/* Disable those interrupts.. */
+	intr_mask = REG_RD(gio, regi_gio, rw_intr_mask);
+	tmp2 = REG_TYPE_CONV(unsigned long, reg_gio_rw_intr_mask, intr_mask);
+	tmp2 &= ~tmp;
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	/* Do not disable interrupt on virtual GPIO. Changes on virtual
+	 * pins are only noticed by an interrupt.
+	 */
+	if (enable_gpiov_ack)
+		tmp2 |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+#endif
+	intr_mask = REG_TYPE_CONV(reg_gio_rw_intr_mask, unsigned long, tmp2);
+	REG_WR(gio, regi_gio, rw_intr_mask, intr_mask);
+
+	return IRQ_RETVAL(tmp);
+}
+
+static void gpio_write_bit(unsigned long *port, unsigned char data, int bit,
+	unsigned char clk_mask, unsigned char data_mask)
+{
+	unsigned long shadow = readl(port) & ~clk_mask;
+	writel(shadow, port);
+	if (data & 1 << bit)
+		shadow |= data_mask;
+	else
+		shadow &= ~data_mask;
+	writel(shadow, port);
+	/* For FPGA: min 5.0ns (DCC) before CCLK high */
+	shadow |= clk_mask;
+	writel(shadow, port);
+}
+
+static void gpio_write_byte(struct gpio_private *priv, unsigned long *port,
+		unsigned char data)
+{
+	int i;
+
+	if (priv->write_msb)
+		for (i = 7; i >= 0; i--)
+			gpio_write_bit(port, data, i, priv->clk_mask,
+				priv->data_mask);
+	else
+		for (i = 0; i <= 7; i++)
+			gpio_write_bit(port, data, i, priv->clk_mask,
+				priv->data_mask);
+}
+
+
+static ssize_t gpio_write(struct file *file, const char __user *buf,
+	size_t count, loff_t *off)
+{
+	struct gpio_private *priv = file->private_data;
+	unsigned long flags;
+	ssize_t retval = count;
+	/* Only bits 0-7 may be used for write operations but allow all
+	   devices except leds... */
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	if (priv->minor == GPIO_MINOR_V)
+		return -EFAULT;
+#endif
+	if (priv->minor == GPIO_MINOR_LEDS)
+		return -EFAULT;
+
+	if (priv->minor >= GPIO_MINOR_PWM0 &&
+	    priv->minor <= GPIO_MINOR_LAST_PWM)
+		return -EFAULT;
+
+	if (!access_ok(VERIFY_READ, buf, count))
+		return -EFAULT;
+
+	/* It must have been configured using the IO_CFG_WRITE_MODE */
+	/* Perhaps a better error code? */
+	if (priv->clk_mask == 0 || priv->data_mask == 0)
+		return -EPERM;
+
+	D(printk(KERN_DEBUG "gpio_write: %lu to data 0x%02X clk 0x%02X "
+		"msb: %i\n",
+		count, priv->data_mask, priv->clk_mask, priv->write_msb));
+
+	spin_lock_irqsave(&gpio_lock, flags);
+
+	while (count--)
+		gpio_write_byte(priv, data_out[priv->minor], *buf++);
+
+	spin_unlock_irqrestore(&gpio_lock, flags);
+	return retval;
+}
+
+static int gpio_open(struct inode *inode, struct file *filp)
+{
+	struct gpio_private *priv;
+	int p = iminor(inode);
+
+	if (p > GPIO_MINOR_LAST_PWM ||
+	    (p > GPIO_MINOR_LAST && p < GPIO_MINOR_PWM0))
+		return -EINVAL;
+
+	priv = kmalloc(sizeof(struct gpio_private), GFP_KERNEL);
+
+	if (!priv)
+		return -ENOMEM;
+
+	mutex_lock(&gpio_mutex);
+	memset(priv, 0, sizeof(*priv));
+
+	priv->minor = p;
+	filp->private_data = priv;
+
+	/* initialize the io/alarm struct, not for PWM ports though  */
+	if (p <= GPIO_MINOR_LAST) {
+
+		priv->clk_mask = 0;
+		priv->data_mask = 0;
+		priv->highalarm = 0;
+		priv->lowalarm = 0;
+
+		init_waitqueue_head(&priv->alarm_wq);
+
+		/* link it into our alarmlist */
+		spin_lock_irq(&gpio_lock);
+		priv->next = alarmlist;
+		alarmlist = priv;
+		spin_unlock_irq(&gpio_lock);
+	}
+
+	mutex_unlock(&gpio_mutex);
+	return 0;
+}
+
+static int gpio_release(struct inode *inode, struct file *filp)
+{
+	struct gpio_private *p;
+	struct gpio_private *todel;
+	/* local copies while updating them: */
+	unsigned long a_high, a_low;
+
+	/* prepare to free private structure */
+	todel = filp->private_data;
+
+	/* unlink from alarmlist - only for non-PWM ports though */
+	if (todel->minor <= GPIO_MINOR_LAST) {
+		spin_lock_irq(&gpio_lock);
+		p = alarmlist;
+
+		if (p == todel)
+			alarmlist = todel->next;
+		 else {
+			while (p->next != todel)
+				p = p->next;
+			p->next = todel->next;
+		}
+
+		/* Check if there are still any alarms set */
+		p = alarmlist;
+		a_high = 0;
+		a_low = 0;
+		while (p) {
+			if (p->minor == GPIO_MINOR_A) {
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+				p->lowalarm |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+#endif
+				a_high |= p->highalarm;
+				a_low |= p->lowalarm;
+			}
+
+			p = p->next;
+		}
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	/* Variable 'a_low' needs to be set here again
+	 * to ensure that interrupt for virtual GPIO is handled.
+	 */
+		a_low |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+#endif
+
+		spin_unlock_irq(&gpio_lock);
+	}
+	kfree(todel);
+
+	return 0;
+}
+
+/* Main device API. ioctl's to read/set/clear bits, as well as to
+ * set alarms to wait for using a subsequent select().
+ */
+
+inline unsigned long setget_input(struct gpio_private *priv, unsigned long arg)
+{
+	/* Set direction 0=unchanged 1=input,
+	 * return mask with 1=input
+	 */
+	unsigned long flags;
+	unsigned long dir_shadow;
+
+	spin_lock_irqsave(&gpio_lock, flags);
+
+	dir_shadow = readl(dir_oe[priv->minor]) &
+		~(arg & changeable_dir[priv->minor]);
+	writel(dir_shadow, dir_oe[priv->minor]);
+
+	spin_unlock_irqrestore(&gpio_lock, flags);
+
+	if (priv->minor == GPIO_MINOR_C)
+		dir_shadow ^= 0xFFFF;		/* Only 16 bits */
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	else if (priv->minor == GPIO_MINOR_V)
+		dir_shadow ^= 0xFFFF;		/* Only 16 bits */
+#endif
+	else
+		dir_shadow ^= 0xFFFFFFFF;	/* PA, PB and PD 32 bits */
+
+	return dir_shadow;
+
+} /* setget_input */
+
+static inline unsigned long setget_output(struct gpio_private *priv,
+	unsigned long arg)
+{
+	unsigned long flags;
+	unsigned long dir_shadow;
+
+	spin_lock_irqsave(&gpio_lock, flags);
+
+	dir_shadow = readl(dir_oe[priv->minor]) |
+		(arg & changeable_dir[priv->minor]);
+	writel(dir_shadow, dir_oe[priv->minor]);
+
+	spin_unlock_irqrestore(&gpio_lock, flags);
+	return dir_shadow;
+} /* setget_output */
+
+static long gpio_ioctl_unlocked(struct file *file,
+	unsigned int cmd, unsigned long arg)
+{
+	unsigned long flags;
+	unsigned long val;
+	unsigned long shadow;
+	struct gpio_private *priv = file->private_data;
+
+	if (_IOC_TYPE(cmd) != ETRAXGPIO_IOCTYPE)
+		return -ENOTTY;
+
+	/* Check for special ioctl handlers first */
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	if (priv->minor == GPIO_MINOR_V)
+		return virtual_gpio_ioctl(file, cmd, arg);
+#endif
+
+	if (priv->minor == GPIO_MINOR_LEDS)
+		return gpio_leds_ioctl(cmd, arg);
+
+	if (priv->minor >= GPIO_MINOR_PWM0 &&
+	    priv->minor <= GPIO_MINOR_LAST_PWM)
+		return gpio_pwm_ioctl(priv, cmd, arg);
+
+	switch (_IOC_NR(cmd)) {
+	case IO_READBITS: /* Use IO_READ_INBITS and IO_READ_OUTBITS instead */
+		/* Read the port. */
+		return readl(data_in[priv->minor]);
+	case IO_SETBITS:
+		spin_lock_irqsave(&gpio_lock, flags);
+		/* Set changeable bits with a 1 in arg. */
+		shadow = readl(data_out[priv->minor]) |
+			(arg & changeable_bits[priv->minor]);
+		writel(shadow, data_out[priv->minor]);
+		spin_unlock_irqrestore(&gpio_lock, flags);
+		break;
+	case IO_CLRBITS:
+		spin_lock_irqsave(&gpio_lock, flags);
+		/* Clear changeable bits with a 1 in arg. */
+		shadow = readl(data_out[priv->minor]) &
+			~(arg & changeable_bits[priv->minor]);
+		writel(shadow, data_out[priv->minor]);
+		spin_unlock_irqrestore(&gpio_lock, flags);
+		break;
+	case IO_HIGHALARM:
+		/* Set alarm when bits with 1 in arg go high. */
+		priv->highalarm |= arg;
+		gpio_set_alarm(priv);
+		break;
+	case IO_LOWALARM:
+		/* Set alarm when bits with 1 in arg go low. */
+		priv->lowalarm |= arg;
+		gpio_set_alarm(priv);
+		break;
+	case IO_CLRALARM:
+		/* Clear alarm for bits with 1 in arg. */
+		priv->highalarm &= ~arg;
+		priv->lowalarm  &= ~arg;
+		gpio_set_alarm(priv);
+		break;
+	case IO_READDIR: /* Use IO_SETGET_INPUT/OUTPUT instead! */
+		/* Read direction 0=input 1=output */
+		return readl(dir_oe[priv->minor]);
+
+	case IO_SETINPUT: /* Use IO_SETGET_INPUT instead! */
+		/* Set direction 0=unchanged 1=input,
+		 * return mask with 1=input
+		 */
+		return setget_input(priv, arg);
+
+	case IO_SETOUTPUT: /* Use IO_SETGET_OUTPUT instead! */
+		/* Set direction 0=unchanged 1=output,
+		 * return mask with 1=output
+		 */
+		return setget_output(priv, arg);
+
+	case IO_CFG_WRITE_MODE:
+	{
+		int res = -EPERM;
+		unsigned long dir_shadow, clk_mask, data_mask, write_msb;
+
+		clk_mask = arg & 0xFF;
+		data_mask = (arg >> 8) & 0xFF;
+		write_msb = (arg >> 16) & 0x01;
+
+		/* Check if we're allowed to change the bits and
+		 * the direction is correct
+		 */
+		spin_lock_irqsave(&gpio_lock, flags);
+		dir_shadow = readl(dir_oe[priv->minor]);
+		if ((clk_mask & changeable_bits[priv->minor]) &&
+		    (data_mask & changeable_bits[priv->minor]) &&
+		    (clk_mask & dir_shadow) &&
+		    (data_mask & dir_shadow)) {
+			priv->clk_mask = clk_mask;
+			priv->data_mask = data_mask;
+			priv->write_msb = write_msb;
+			res = 0;
+		}
+		spin_unlock_irqrestore(&gpio_lock, flags);
+
+		return res;
+	}
+	case IO_READ_INBITS:
+		/* *arg is result of reading the input pins */
+		val = readl(data_in[priv->minor]);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		return 0;
+	case IO_READ_OUTBITS:
+		 /* *arg is result of reading the output shadow */
+		val = *data_out[priv->minor];
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	case IO_SETGET_INPUT:
+		/* bits set in *arg is set to input,
+		 * *arg updated with current input pins.
+		 */
+		if (copy_from_user(&val, (void __user *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_input(priv, val);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	case IO_SETGET_OUTPUT:
+		/* bits set in *arg is set to output,
+		 * *arg updated with current output pins.
+		 */
+		if (copy_from_user(&val, (void __user *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_output(priv, val);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	default:
+		return -EINVAL;
+	} /* switch */
+
+	return 0;
+}
+
+static long gpio_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+       long ret;
+
+       mutex_lock(&gpio_mutex);
+       ret = gpio_ioctl_unlocked(file, cmd, arg);
+       mutex_unlock(&gpio_mutex);
+
+       return ret;
+}
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static int virtual_gpio_ioctl(struct file *file, unsigned int cmd,
+	unsigned long arg)
+{
+	unsigned long flags;
+	unsigned short val;
+	unsigned short shadow;
+	struct gpio_private *priv = file->private_data;
+
+	switch (_IOC_NR(cmd)) {
+	case IO_SETBITS:
+		spin_lock_irqsave(&gpio_lock, flags);
+		/* Set changeable bits with a 1 in arg. */
+		i2c_read(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		shadow |= ~readl(dir_oe[priv->minor]) |
+			(arg & changeable_bits[priv->minor]);
+		i2c_write(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		spin_unlock_irqrestore(&gpio_lock, flags);
+		break;
+	case IO_CLRBITS:
+		spin_lock_irqsave(&gpio_lock, flags);
+		/* Clear changeable bits with a 1 in arg. */
+		i2c_read(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		shadow |= ~readl(dir_oe[priv->minor]) &
+			~(arg & changeable_bits[priv->minor]);
+		i2c_write(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		spin_unlock_irqrestore(&gpio_lock, flags);
+		break;
+	case IO_HIGHALARM:
+		/* Set alarm when bits with 1 in arg go high. */
+		priv->highalarm |= arg;
+		break;
+	case IO_LOWALARM:
+		/* Set alarm when bits with 1 in arg go low. */
+		priv->lowalarm |= arg;
+		break;
+	case IO_CLRALARM:
+		/* Clear alarm for bits with 1 in arg. */
+		priv->highalarm &= ~arg;
+		priv->lowalarm  &= ~arg;
+		break;
+	case IO_CFG_WRITE_MODE:
+	{
+		unsigned long dir_shadow;
+		dir_shadow = readl(dir_oe[priv->minor]);
+
+		priv->clk_mask = arg & 0xFF;
+		priv->data_mask = (arg >> 8) & 0xFF;
+		priv->write_msb = (arg >> 16) & 0x01;
+		/* Check if we're allowed to change the bits and
+		 * the direction is correct
+		 */
+		if (!((priv->clk_mask & changeable_bits[priv->minor]) &&
+		      (priv->data_mask & changeable_bits[priv->minor]) &&
+		      (priv->clk_mask & dir_shadow) &&
+		      (priv->data_mask & dir_shadow))) {
+			priv->clk_mask = 0;
+			priv->data_mask = 0;
+			return -EPERM;
+		}
+		break;
+	}
+	case IO_READ_INBITS:
+		/* *arg is result of reading the input pins */
+		val = cached_virtual_gpio_read & ~readl(dir_oe[priv->minor]);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		return 0;
+
+	case IO_READ_OUTBITS:
+		 /* *arg is result of reading the output shadow */
+		i2c_read(VIRT_I2C_ADDR, (void *)&val, sizeof(val));
+		val &= readl(dir_oe[priv->minor]);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	case IO_SETGET_INPUT:
+	{
+		/* bits set in *arg is set to input,
+		 * *arg updated with current input pins.
+		 */
+		unsigned short input_mask = ~readl(dir_oe[priv->minor]);
+		if (copy_from_user(&val, (void __user *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_input(priv, val);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		if ((input_mask & val) != input_mask) {
+			/* Input pins changed. All ports desired as input
+			 * should be set to logic 1.
+			 */
+			unsigned short change = input_mask ^ val;
+			i2c_read(VIRT_I2C_ADDR, (void *)&shadow,
+				sizeof(shadow));
+			shadow &= ~change;
+			shadow |= val;
+			i2c_write(VIRT_I2C_ADDR, (void *)&shadow,
+				sizeof(shadow));
+		}
+		break;
+	}
+	case IO_SETGET_OUTPUT:
+		/* bits set in *arg is set to output,
+		 * *arg updated with current output pins.
+		 */
+		if (copy_from_user(&val, (void __user *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_output(priv, val);
+		if (copy_to_user((void __user *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	default:
+		return -EINVAL;
+	} /* switch */
+	return 0;
+}
+#endif /* CONFIG_ETRAX_VIRTUAL_GPIO */
+
+static int gpio_leds_ioctl(unsigned int cmd, unsigned long arg)
+{
+	unsigned char green;
+	unsigned char red;
+
+	switch (_IOC_NR(cmd)) {
+	case IO_LEDACTIVE_SET:
+		green = ((unsigned char) arg) & 1;
+		red   = (((unsigned char) arg) >> 1) & 1;
+		CRIS_LED_ACTIVE_SET_G(green);
+		CRIS_LED_ACTIVE_SET_R(red);
+		break;
+
+	default:
+		return -EINVAL;
+	} /* switch */
+
+	return 0;
+}
+
+static int gpio_pwm_set_mode(unsigned long arg, int pwm_port)
+{
+	int pinmux_pwm = pinmux_pwm0 + pwm_port;
+	int mode;
+	reg_gio_rw_pwm0_ctrl rw_pwm_ctrl = {
+		.ccd_val = 0,
+		.ccd_override = regk_gio_no,
+		.mode = regk_gio_no
+	};
+	int allocstatus;
+
+	if (get_user(mode, &((struct io_pwm_set_mode *) arg)->mode))
+		return -EFAULT;
+	rw_pwm_ctrl.mode = mode;
+	if (mode != PWM_OFF)
+		allocstatus = crisv32_pinmux_alloc_fixed(pinmux_pwm);
+	else
+		allocstatus = crisv32_pinmux_dealloc_fixed(pinmux_pwm);
+	if (allocstatus)
+		return allocstatus;
+	REG_WRITE(reg_gio_rw_pwm0_ctrl, REG_ADDR(gio, regi_gio, rw_pwm0_ctrl) +
+		12 * pwm_port, rw_pwm_ctrl);
+	return 0;
+}
+
+static int gpio_pwm_set_period(unsigned long arg, int pwm_port)
+{
+	struct io_pwm_set_period periods;
+	reg_gio_rw_pwm0_var rw_pwm_widths;
+
+	if (copy_from_user(&periods, (void __user *)arg, sizeof(periods)))
+		return -EFAULT;
+	if (periods.lo > 8191 || periods.hi > 8191)
+		return -EINVAL;
+	rw_pwm_widths.lo = periods.lo;
+	rw_pwm_widths.hi = periods.hi;
+	REG_WRITE(reg_gio_rw_pwm0_var, REG_ADDR(gio, regi_gio, rw_pwm0_var) +
+		12 * pwm_port, rw_pwm_widths);
+	return 0;
+}
+
+static int gpio_pwm_set_duty(unsigned long arg, int pwm_port)
+{
+	unsigned int duty;
+	reg_gio_rw_pwm0_data rw_pwm_duty;
+
+	if (get_user(duty, &((struct io_pwm_set_duty *) arg)->duty))
+		return -EFAULT;
+	if (duty > 255)
+		return -EINVAL;
+	rw_pwm_duty.data = duty;
+	REG_WRITE(reg_gio_rw_pwm0_data, REG_ADDR(gio, regi_gio, rw_pwm0_data) +
+		12 * pwm_port, rw_pwm_duty);
+	return 0;
+}
+
+static int gpio_pwm_ioctl(struct gpio_private *priv, unsigned int cmd,
+	unsigned long arg)
+{
+	int pwm_port = priv->minor - GPIO_MINOR_PWM0;
+
+	switch (_IOC_NR(cmd)) {
+	case IO_PWM_SET_MODE:
+		return gpio_pwm_set_mode(arg, pwm_port);
+	case IO_PWM_SET_PERIOD:
+		return gpio_pwm_set_period(arg, pwm_port);
+	case IO_PWM_SET_DUTY:
+		return gpio_pwm_set_duty(arg, pwm_port);
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static const struct file_operations gpio_fops = {
+	.owner		= THIS_MODULE,
+	.poll		= gpio_poll,
+	.unlocked_ioctl	= gpio_ioctl,
+	.write		= gpio_write,
+	.open		= gpio_open,
+	.release	= gpio_release,
+	.llseek		= noop_llseek,
+};
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static void __init virtual_gpio_init(void)
+{
+	reg_gio_rw_intr_cfg intr_cfg;
+	reg_gio_rw_intr_mask intr_mask;
+	unsigned short shadow;
+
+	shadow = ~virtual_rw_pv_oe; /* Input ports should be set to logic 1 */
+	shadow |= CONFIG_ETRAX_DEF_GIO_PV_OUT;
+	i2c_write(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+
+	/* Set interrupt mask and on what state the interrupt shall trigger.
+	 * For virtual gpio the interrupt shall trigger on logic '0'.
+	 */
+	intr_cfg = REG_RD(gio, regi_gio, rw_intr_cfg);
+	intr_mask = REG_RD(gio, regi_gio, rw_intr_mask);
+
+	switch (CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN) {
+	case 0:
+		intr_cfg.pa0 = regk_gio_lo;
+		intr_mask.pa0 = regk_gio_yes;
+		break;
+	case 1:
+		intr_cfg.pa1 = regk_gio_lo;
+		intr_mask.pa1 = regk_gio_yes;
+		break;
+	case 2:
+		intr_cfg.pa2 = regk_gio_lo;
+		intr_mask.pa2 = regk_gio_yes;
+		break;
+	case 3:
+		intr_cfg.pa3 = regk_gio_lo;
+		intr_mask.pa3 = regk_gio_yes;
+		break;
+	case 4:
+		intr_cfg.pa4 = regk_gio_lo;
+		intr_mask.pa4 = regk_gio_yes;
+		break;
+	case 5:
+		intr_cfg.pa5 = regk_gio_lo;
+		intr_mask.pa5 = regk_gio_yes;
+		break;
+	case 6:
+		intr_cfg.pa6 = regk_gio_lo;
+		intr_mask.pa6 = regk_gio_yes;
+		break;
+	case 7:
+		intr_cfg.pa7 = regk_gio_lo;
+		intr_mask.pa7 = regk_gio_yes;
+		break;
+	}
+
+	REG_WR(gio, regi_gio, rw_intr_cfg, intr_cfg);
+	REG_WR(gio, regi_gio, rw_intr_mask, intr_mask);
+}
+#endif
+
+/* main driver initialization routine, called from mem.c */
+
+static int __init gpio_init(void)
+{
+	int res;
+
+	printk(KERN_INFO "ETRAX FS GPIO driver v2.7, (c) 2003-2008 "
+		"Axis Communications AB\n");
+
+	/* do the formalities */
+
+	res = register_chrdev(GPIO_MAJOR, gpio_name, &gpio_fops);
+	if (res < 0) {
+		printk(KERN_ERR "gpio: couldn't get a major number.\n");
+		return res;
+	}
+
+	/* Clear all leds */
+	CRIS_LED_NETWORK_GRP0_SET(0);
+	CRIS_LED_NETWORK_GRP1_SET(0);
+	CRIS_LED_ACTIVE_SET(0);
+	CRIS_LED_DISK_READ(0);
+	CRIS_LED_DISK_WRITE(0);
+
+	int res2 = request_irq(GIO_INTR_VECT, gpio_interrupt,
+		IRQF_SHARED, "gpio", &alarmlist);
+	if (res2) {
+		printk(KERN_ERR "err: irq for gpio\n");
+		return res2;
+	}
+
+	/* No IRQs by default. */
+	REG_WR_INT(gio, regi_gio, rw_intr_pins, 0);
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	virtual_gpio_init();
+#endif
+
+	return res;
+}
+
+/* this makes sure that gpio_init is called during kernel boot */
+
+module_init(gpio_init);
diff --git a/kernel/arch/cris/arch-v32/drivers/mach-a3/nandflash.c b/kernel/arch/cris/arch-v32/drivers/mach-a3/nandflash.c
new file mode 100644
index 000000000..7fb52128d
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/mach-a3/nandflash.c
@@ -0,0 +1,180 @@
+/*
+ *  arch/cris/arch-v32/drivers/nandflash.c
+ *
+ *  Copyright (c) 2007
+ *
+ *  Derived from drivers/mtd/nand/spia.c
+ *	  Copyright (C) 2000 Steven J. Hill (sjhill@realitydiluted.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/slab.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/nand.h>
+#include <linux/mtd/partitions.h>
+#include <arch/memmap.h>
+#include <hwregs/reg_map.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/pio_defs.h>
+#include <pinmux.h>
+#include <asm/io.h>
+
+#define MANUAL_ALE_CLE_CONTROL 1
+
+#define regf_ALE	a0
+#define regf_CLE	a1
+#define regf_NCE	ce0_n
+
+#define CLE_BIT 10
+#define ALE_BIT 11
+#define CE_BIT 12
+
+struct mtd_info_wrapper {
+	struct mtd_info info;
+	struct nand_chip chip;
+};
+
+/* Bitmask for control pins */
+#define PIN_BITMASK ((1 << CE_BIT) | (1 << CLE_BIT) | (1 << ALE_BIT))
+
+static struct mtd_info *crisv32_mtd;
+/*
+ *	hardware specific access to control-lines
+ */
+static void crisv32_hwcontrol(struct mtd_info *mtd, int cmd,
+			      unsigned int ctrl)
+{
+	unsigned long flags;
+	reg_pio_rw_dout dout;
+	struct nand_chip *this = mtd->priv;
+
+	local_irq_save(flags);
+
+	/* control bits change */
+	if (ctrl & NAND_CTRL_CHANGE) {
+		dout = REG_RD(pio, regi_pio, rw_dout);
+		dout.regf_NCE = (ctrl & NAND_NCE) ? 0 : 1;
+
+#if !MANUAL_ALE_CLE_CONTROL
+		if (ctrl & NAND_ALE) {
+			/* A0 = ALE high */
+			this->IO_ADDR_W = (void __iomem *)REG_ADDR(pio,
+				regi_pio, rw_io_access1);
+		} else if (ctrl & NAND_CLE) {
+			/* A1 = CLE high */
+			this->IO_ADDR_W = (void __iomem *)REG_ADDR(pio,
+				regi_pio, rw_io_access2);
+		} else {
+			/* A1 = CLE and A0 = ALE low */
+			this->IO_ADDR_W = (void __iomem *)REG_ADDR(pio,
+				regi_pio, rw_io_access0);
+		}
+#else
+
+		dout.regf_CLE = (ctrl & NAND_CLE) ? 1 : 0;
+		dout.regf_ALE = (ctrl & NAND_ALE) ? 1 : 0;
+#endif
+		REG_WR(pio, regi_pio, rw_dout, dout);
+	}
+
+	/* command to chip */
+	if (cmd != NAND_CMD_NONE)
+		writeb(cmd, this->IO_ADDR_W);
+
+	local_irq_restore(flags);
+}
+
+/*
+*	read device ready pin
+*/
+static int crisv32_device_ready(struct mtd_info *mtd)
+{
+	reg_pio_r_din din = REG_RD(pio, regi_pio, r_din);
+	return din.rdy;
+}
+
+/*
+ * Main initialization routine
+ */
+struct mtd_info *__init crisv32_nand_flash_probe(void)
+{
+	void __iomem *read_cs;
+	void __iomem *write_cs;
+
+	struct mtd_info_wrapper *wrapper;
+	struct nand_chip *this;
+	int err = 0;
+
+	reg_pio_rw_man_ctrl man_ctrl = {
+		.regf_NCE = regk_pio_yes,
+#if MANUAL_ALE_CLE_CONTROL
+		.regf_ALE = regk_pio_yes,
+		.regf_CLE = regk_pio_yes
+#endif
+	};
+	reg_pio_rw_oe oe = {
+		.regf_NCE = regk_pio_yes,
+#if MANUAL_ALE_CLE_CONTROL
+		.regf_ALE = regk_pio_yes,
+		.regf_CLE = regk_pio_yes
+#endif
+	};
+	reg_pio_rw_dout dout = { .regf_NCE = 1 };
+
+	/* Allocate pio pins to pio */
+	crisv32_pinmux_alloc_fixed(pinmux_pio);
+	/* Set up CE, ALE, CLE (ce0_n, a0, a1) for manual control and output */
+	REG_WR(pio, regi_pio, rw_man_ctrl, man_ctrl);
+	REG_WR(pio, regi_pio, rw_dout, dout);
+	REG_WR(pio, regi_pio, rw_oe, oe);
+
+	/* Allocate memory for MTD device structure and private data */
+	wrapper = kzalloc(sizeof(struct mtd_info_wrapper), GFP_KERNEL);
+	if (!wrapper) {
+		printk(KERN_ERR "Unable to allocate CRISv32 NAND MTD "
+			"device structure.\n");
+		err = -ENOMEM;
+		return NULL;
+	}
+
+	read_cs = write_cs = (void __iomem *)REG_ADDR(pio, regi_pio,
+		rw_io_access0);
+
+	/* Get pointer to private data */
+	this = &wrapper->chip;
+	crisv32_mtd = &wrapper->info;
+
+	/* Link the private data with the MTD structure */
+	crisv32_mtd->priv = this;
+
+	/* Set address of NAND IO lines */
+	this->IO_ADDR_R = read_cs;
+	this->IO_ADDR_W = write_cs;
+	this->cmd_ctrl = crisv32_hwcontrol;
+	this->dev_ready = crisv32_device_ready;
+	/* 20 us command delay time */
+	this->chip_delay = 20;
+	this->ecc.mode = NAND_ECC_SOFT;
+
+	/* Enable the following for a flash based bad block table */
+	/* this->bbt_options = NAND_BBT_USE_FLASH; */
+
+	/* Scan to find existence of the device */
+	if (nand_scan(crisv32_mtd, 1)) {
+		err = -ENXIO;
+		goto out_mtd;
+	}
+
+	return crisv32_mtd;
+
+out_mtd:
+	kfree(wrapper);
+	return NULL;
+}
+
diff --git a/kernel/arch/cris/arch-v32/drivers/mach-fs/Makefile b/kernel/arch/cris/arch-v32/drivers/mach-fs/Makefile
new file mode 100644
index 000000000..5c6d2a2a0
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/mach-fs/Makefile
@@ -0,0 +1,6 @@
+#
+# Makefile for Etrax-specific drivers
+#
+
+obj-$(CONFIG_ETRAX_NANDFLASH)   += nandflash.o
+obj-$(CONFIG_ETRAX_GPIO)        += gpio.o
diff --git a/kernel/arch/cris/arch-v32/drivers/mach-fs/gpio.c b/kernel/arch/cris/arch-v32/drivers/mach-fs/gpio.c
new file mode 100644
index 000000000..009f4ee1b
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/mach-fs/gpio.c
@@ -0,0 +1,979 @@
+/*
+ * ETRAX CRISv32 general port I/O device
+ *
+ * Copyright (c) 1999-2006 Axis Communications AB
+ *
+ * Authors:    Bjorn Wesen      (initial version)
+ *             Ola Knutsson     (LED handling)
+ *             Johan Adolfsson  (read/set directions, write, port G,
+ *                               port to ETRAX FS.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/ioport.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/string.h>
+#include <linux/poll.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/spinlock.h>
+#include <linux/mutex.h>
+
+#include <asm/etraxgpio.h>
+#include <hwregs/reg_map.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/gio_defs.h>
+#include <hwregs/intr_vect_defs.h>
+#include <asm/io.h>
+#include <asm/irq.h>
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+#include "../i2c.h"
+
+#define VIRT_I2C_ADDR 0x40
+#endif
+
+/* The following gio ports on ETRAX FS is available:
+ * pa  8 bits, supports interrupts off, hi, low, set, posedge, negedge anyedge
+ * pb 18 bits
+ * pc 18 bits
+ * pd 18 bits
+ * pe 18 bits
+ * each port has a rw_px_dout, r_px_din and rw_px_oe register.
+ */
+
+#define GPIO_MAJOR 120  /* experimental MAJOR number */
+
+#define D(x)
+
+#if 0
+static int dp_cnt;
+#define DP(x) \
+	do { \
+		dp_cnt++; \
+		if (dp_cnt % 1000 == 0) \
+			x; \
+	} while (0)
+#else
+#define DP(x)
+#endif
+
+static DEFINE_MUTEX(gpio_mutex);
+static char gpio_name[] = "etrax gpio";
+
+#if 0
+static wait_queue_head_t *gpio_wq;
+#endif
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static int virtual_gpio_ioctl(struct file *file, unsigned int cmd,
+	unsigned long arg);
+#endif
+static long gpio_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
+static ssize_t gpio_write(struct file *file, const char *buf, size_t count,
+	loff_t *off);
+static int gpio_open(struct inode *inode, struct file *filp);
+static int gpio_release(struct inode *inode, struct file *filp);
+static unsigned int gpio_poll(struct file *filp,
+	struct poll_table_struct *wait);
+
+/* private data per open() of this driver */
+
+struct gpio_private {
+	struct gpio_private *next;
+	/* The IO_CFG_WRITE_MODE_VALUE only support 8 bits: */
+	unsigned char clk_mask;
+	unsigned char data_mask;
+	unsigned char write_msb;
+	unsigned char pad1;
+	/* These fields are generic */
+	unsigned long highalarm, lowalarm;
+	wait_queue_head_t alarm_wq;
+	int minor;
+};
+
+/* linked list of alarms to check for */
+
+static struct gpio_private *alarmlist;
+
+static int gpio_some_alarms; /* Set if someone uses alarm */
+static unsigned long gpio_pa_high_alarms;
+static unsigned long gpio_pa_low_alarms;
+
+static DEFINE_SPINLOCK(alarm_lock);
+
+#define NUM_PORTS (GPIO_MINOR_LAST+1)
+#define GIO_REG_RD_ADDR(reg) \
+	(volatile unsigned long *)(regi_gio + REG_RD_ADDR_gio_##reg)
+#define GIO_REG_WR_ADDR(reg) \
+	(volatile unsigned long *)(regi_gio + REG_RD_ADDR_gio_##reg)
+unsigned long led_dummy;
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static unsigned long virtual_dummy;
+static unsigned long virtual_rw_pv_oe = CONFIG_ETRAX_DEF_GIO_PV_OE;
+static unsigned short cached_virtual_gpio_read;
+#endif
+
+static volatile unsigned long *data_out[NUM_PORTS] = {
+	GIO_REG_WR_ADDR(rw_pa_dout),
+	GIO_REG_WR_ADDR(rw_pb_dout),
+	&led_dummy,
+	GIO_REG_WR_ADDR(rw_pc_dout),
+	GIO_REG_WR_ADDR(rw_pd_dout),
+	GIO_REG_WR_ADDR(rw_pe_dout),
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	&virtual_dummy,
+#endif
+};
+
+static volatile unsigned long *data_in[NUM_PORTS] = {
+	GIO_REG_RD_ADDR(r_pa_din),
+	GIO_REG_RD_ADDR(r_pb_din),
+	&led_dummy,
+	GIO_REG_RD_ADDR(r_pc_din),
+	GIO_REG_RD_ADDR(r_pd_din),
+	GIO_REG_RD_ADDR(r_pe_din),
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	&virtual_dummy,
+#endif
+};
+
+static unsigned long changeable_dir[NUM_PORTS] = {
+	CONFIG_ETRAX_PA_CHANGEABLE_DIR,
+	CONFIG_ETRAX_PB_CHANGEABLE_DIR,
+	0,
+	CONFIG_ETRAX_PC_CHANGEABLE_DIR,
+	CONFIG_ETRAX_PD_CHANGEABLE_DIR,
+	CONFIG_ETRAX_PE_CHANGEABLE_DIR,
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	CONFIG_ETRAX_PV_CHANGEABLE_DIR,
+#endif
+};
+
+static unsigned long changeable_bits[NUM_PORTS] = {
+	CONFIG_ETRAX_PA_CHANGEABLE_BITS,
+	CONFIG_ETRAX_PB_CHANGEABLE_BITS,
+	0,
+	CONFIG_ETRAX_PC_CHANGEABLE_BITS,
+	CONFIG_ETRAX_PD_CHANGEABLE_BITS,
+	CONFIG_ETRAX_PE_CHANGEABLE_BITS,
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	CONFIG_ETRAX_PV_CHANGEABLE_BITS,
+#endif
+};
+
+static volatile unsigned long *dir_oe[NUM_PORTS] = {
+	GIO_REG_WR_ADDR(rw_pa_oe),
+	GIO_REG_WR_ADDR(rw_pb_oe),
+	&led_dummy,
+	GIO_REG_WR_ADDR(rw_pc_oe),
+	GIO_REG_WR_ADDR(rw_pd_oe),
+	GIO_REG_WR_ADDR(rw_pe_oe),
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	&virtual_rw_pv_oe,
+#endif
+};
+
+
+
+static unsigned int gpio_poll(struct file *file, struct poll_table_struct *wait)
+{
+	unsigned int mask = 0;
+	struct gpio_private *priv = file->private_data;
+	unsigned long data;
+	poll_wait(file, &priv->alarm_wq, wait);
+	if (priv->minor == GPIO_MINOR_A) {
+		reg_gio_rw_intr_cfg intr_cfg;
+		unsigned long tmp;
+		unsigned long flags;
+
+		local_irq_save(flags);
+		data = REG_TYPE_CONV(unsigned long, reg_gio_r_pa_din,
+			REG_RD(gio, regi_gio, r_pa_din));
+		/* PA has support for interrupt
+		 * lets activate high for those low and with highalarm set
+		 */
+		intr_cfg = REG_RD(gio, regi_gio, rw_intr_cfg);
+
+		tmp = ~data & priv->highalarm & 0xFF;
+		if (tmp & (1 << 0))
+			intr_cfg.pa0 = regk_gio_hi;
+		if (tmp & (1 << 1))
+			intr_cfg.pa1 = regk_gio_hi;
+		if (tmp & (1 << 2))
+			intr_cfg.pa2 = regk_gio_hi;
+		if (tmp & (1 << 3))
+			intr_cfg.pa3 = regk_gio_hi;
+		if (tmp & (1 << 4))
+			intr_cfg.pa4 = regk_gio_hi;
+		if (tmp & (1 << 5))
+			intr_cfg.pa5 = regk_gio_hi;
+		if (tmp & (1 << 6))
+			intr_cfg.pa6 = regk_gio_hi;
+		if (tmp & (1 << 7))
+			intr_cfg.pa7 = regk_gio_hi;
+		/*
+		 * lets activate low for those high and with lowalarm set
+		 */
+		tmp = data & priv->lowalarm & 0xFF;
+		if (tmp & (1 << 0))
+			intr_cfg.pa0 = regk_gio_lo;
+		if (tmp & (1 << 1))
+			intr_cfg.pa1 = regk_gio_lo;
+		if (tmp & (1 << 2))
+			intr_cfg.pa2 = regk_gio_lo;
+		if (tmp & (1 << 3))
+			intr_cfg.pa3 = regk_gio_lo;
+		if (tmp & (1 << 4))
+			intr_cfg.pa4 = regk_gio_lo;
+		if (tmp & (1 << 5))
+			intr_cfg.pa5 = regk_gio_lo;
+		if (tmp & (1 << 6))
+			intr_cfg.pa6 = regk_gio_lo;
+		if (tmp & (1 << 7))
+			intr_cfg.pa7 = regk_gio_lo;
+
+		REG_WR(gio, regi_gio, rw_intr_cfg, intr_cfg);
+		local_irq_restore(flags);
+	} else if (priv->minor <= GPIO_MINOR_E)
+		data = *data_in[priv->minor];
+	else
+		return 0;
+
+	if ((data & priv->highalarm) || (~data & priv->lowalarm))
+		mask = POLLIN|POLLRDNORM;
+
+	DP(printk(KERN_DEBUG "gpio_poll ready: mask 0x%08X\n", mask));
+	return mask;
+}
+
+int etrax_gpio_wake_up_check(void)
+{
+	struct gpio_private *priv;
+	unsigned long data = 0;
+	unsigned long flags;
+	int ret = 0;
+	spin_lock_irqsave(&alarm_lock, flags);
+	priv = alarmlist;
+	while (priv) {
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+		if (priv->minor == GPIO_MINOR_V)
+			data = (unsigned long)cached_virtual_gpio_read;
+		else {
+			data = *data_in[priv->minor];
+			if (priv->minor == GPIO_MINOR_A)
+				priv->lowalarm |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+		}
+#else
+		data = *data_in[priv->minor];
+#endif
+		if ((data & priv->highalarm) ||
+		    (~data & priv->lowalarm)) {
+			DP(printk(KERN_DEBUG
+				"etrax_gpio_wake_up_check %i\n", priv->minor));
+			wake_up_interruptible(&priv->alarm_wq);
+			ret = 1;
+		}
+		priv = priv->next;
+	}
+	spin_unlock_irqrestore(&alarm_lock, flags);
+	return ret;
+}
+
+static irqreturn_t
+gpio_poll_timer_interrupt(int irq, void *dev_id)
+{
+	if (gpio_some_alarms)
+		return IRQ_RETVAL(etrax_gpio_wake_up_check());
+	return IRQ_NONE;
+}
+
+static irqreturn_t
+gpio_pa_interrupt(int irq, void *dev_id)
+{
+	reg_gio_rw_intr_mask intr_mask;
+	reg_gio_r_masked_intr masked_intr;
+	reg_gio_rw_ack_intr ack_intr;
+	unsigned long tmp;
+	unsigned long tmp2;
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	unsigned char enable_gpiov_ack = 0;
+#endif
+
+	/* Find what PA interrupts are active */
+	masked_intr = REG_RD(gio, regi_gio, r_masked_intr);
+	tmp = REG_TYPE_CONV(unsigned long, reg_gio_r_masked_intr, masked_intr);
+
+	/* Find those that we have enabled */
+	spin_lock(&alarm_lock);
+	tmp &= (gpio_pa_high_alarms | gpio_pa_low_alarms);
+	spin_unlock(&alarm_lock);
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	/* Something changed on virtual GPIO. Interrupt is acked by
+	 * reading the device.
+	 */
+	if (tmp & (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN)) {
+		i2c_read(VIRT_I2C_ADDR, (void *)&cached_virtual_gpio_read,
+			sizeof(cached_virtual_gpio_read));
+		enable_gpiov_ack = 1;
+	}
+#endif
+
+	/* Ack them */
+	ack_intr = REG_TYPE_CONV(reg_gio_rw_ack_intr, unsigned long, tmp);
+	REG_WR(gio, regi_gio, rw_ack_intr, ack_intr);
+
+	/* Disable those interrupts.. */
+	intr_mask = REG_RD(gio, regi_gio, rw_intr_mask);
+	tmp2 = REG_TYPE_CONV(unsigned long, reg_gio_rw_intr_mask, intr_mask);
+	tmp2 &= ~tmp;
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	/* Do not disable interrupt on virtual GPIO. Changes on virtual
+	 * pins are only noticed by an interrupt.
+	 */
+	if (enable_gpiov_ack)
+		tmp2 |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+#endif
+	intr_mask = REG_TYPE_CONV(reg_gio_rw_intr_mask, unsigned long, tmp2);
+	REG_WR(gio, regi_gio, rw_intr_mask, intr_mask);
+
+	if (gpio_some_alarms)
+		return IRQ_RETVAL(etrax_gpio_wake_up_check());
+	return IRQ_NONE;
+}
+
+
+static ssize_t gpio_write(struct file *file, const char *buf, size_t count,
+	loff_t *off)
+{
+	struct gpio_private *priv = file->private_data;
+	unsigned char data, clk_mask, data_mask, write_msb;
+	unsigned long flags;
+	unsigned long shadow;
+	volatile unsigned long *port;
+	ssize_t retval = count;
+	/* Only bits 0-7 may be used for write operations but allow all
+	   devices except leds... */
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	if (priv->minor == GPIO_MINOR_V)
+		return -EFAULT;
+#endif
+	if (priv->minor == GPIO_MINOR_LEDS)
+		return -EFAULT;
+
+	if (!access_ok(VERIFY_READ, buf, count))
+		return -EFAULT;
+	clk_mask = priv->clk_mask;
+	data_mask = priv->data_mask;
+	/* It must have been configured using the IO_CFG_WRITE_MODE */
+	/* Perhaps a better error code? */
+	if (clk_mask == 0 || data_mask == 0)
+		return -EPERM;
+	write_msb = priv->write_msb;
+	D(printk(KERN_DEBUG "gpio_write: %lu to data 0x%02X clk 0x%02X "
+		"msb: %i\n", count, data_mask, clk_mask, write_msb));
+	port = data_out[priv->minor];
+
+	while (count--) {
+		int i;
+		data = *buf++;
+		if (priv->write_msb) {
+			for (i = 7; i >= 0; i--) {
+				local_irq_save(flags);
+				shadow = *port;
+				*port = shadow &= ~clk_mask;
+				if (data & 1<<i)
+					*port = shadow |= data_mask;
+				else
+					*port = shadow &= ~data_mask;
+			/* For FPGA: min 5.0ns (DCC) before CCLK high */
+				*port = shadow |= clk_mask;
+				local_irq_restore(flags);
+			}
+		} else {
+			for (i = 0; i <= 7; i++) {
+				local_irq_save(flags);
+				shadow = *port;
+				*port = shadow &= ~clk_mask;
+				if (data & 1<<i)
+					*port = shadow |= data_mask;
+				else
+					*port = shadow &= ~data_mask;
+			/* For FPGA: min 5.0ns (DCC) before CCLK high */
+				*port = shadow |= clk_mask;
+				local_irq_restore(flags);
+			}
+		}
+	}
+	return retval;
+}
+
+
+
+static int
+gpio_open(struct inode *inode, struct file *filp)
+{
+	struct gpio_private *priv;
+	int p = iminor(inode);
+
+	if (p > GPIO_MINOR_LAST)
+		return -EINVAL;
+
+	priv = kmalloc(sizeof(struct gpio_private), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
+	mutex_lock(&gpio_mutex);
+	memset(priv, 0, sizeof(*priv));
+
+	priv->minor = p;
+
+	/* initialize the io/alarm struct */
+
+	priv->clk_mask = 0;
+	priv->data_mask = 0;
+	priv->highalarm = 0;
+	priv->lowalarm = 0;
+	init_waitqueue_head(&priv->alarm_wq);
+
+	filp->private_data = (void *)priv;
+
+	/* link it into our alarmlist */
+	spin_lock_irq(&alarm_lock);
+	priv->next = alarmlist;
+	alarmlist = priv;
+	spin_unlock_irq(&alarm_lock);
+
+	mutex_unlock(&gpio_mutex);
+	return 0;
+}
+
+static int
+gpio_release(struct inode *inode, struct file *filp)
+{
+	struct gpio_private *p;
+	struct gpio_private *todel;
+	/* local copies while updating them: */
+	unsigned long a_high, a_low;
+	unsigned long some_alarms;
+
+	/* unlink from alarmlist and free the private structure */
+
+	spin_lock_irq(&alarm_lock);
+	p = alarmlist;
+	todel = filp->private_data;
+
+	if (p == todel) {
+		alarmlist = todel->next;
+	} else {
+		while (p->next != todel)
+			p = p->next;
+		p->next = todel->next;
+	}
+
+	kfree(todel);
+	/* Check if there are still any alarms set */
+	p = alarmlist;
+	some_alarms = 0;
+	a_high = 0;
+	a_low = 0;
+	while (p) {
+		if (p->minor == GPIO_MINOR_A) {
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+			p->lowalarm |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+#endif
+			a_high |= p->highalarm;
+			a_low |= p->lowalarm;
+		}
+
+		if (p->highalarm | p->lowalarm)
+			some_alarms = 1;
+		p = p->next;
+	}
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	/* Variables 'some_alarms' and 'a_low' needs to be set here again
+	 * to ensure that interrupt for virtual GPIO is handled.
+	 */
+	some_alarms = 1;
+	a_low |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+#endif
+
+	gpio_some_alarms = some_alarms;
+	gpio_pa_high_alarms = a_high;
+	gpio_pa_low_alarms = a_low;
+	spin_unlock_irq(&alarm_lock);
+
+	return 0;
+}
+
+/* Main device API. ioctl's to read/set/clear bits, as well as to
+ * set alarms to wait for using a subsequent select().
+ */
+
+inline unsigned long setget_input(struct gpio_private *priv, unsigned long arg)
+{
+	/* Set direction 0=unchanged 1=input,
+	 * return mask with 1=input
+	 */
+	unsigned long flags;
+	unsigned long dir_shadow;
+
+	local_irq_save(flags);
+	dir_shadow = *dir_oe[priv->minor];
+	dir_shadow &= ~(arg & changeable_dir[priv->minor]);
+	*dir_oe[priv->minor] = dir_shadow;
+	local_irq_restore(flags);
+
+	if (priv->minor == GPIO_MINOR_A)
+		dir_shadow ^= 0xFF;    /* Only 8 bits */
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	else if (priv->minor == GPIO_MINOR_V)
+		dir_shadow ^= 0xFFFF;  /* Only 16 bits */
+#endif
+	else
+		dir_shadow ^= 0x3FFFF; /* Only 18 bits */
+	return dir_shadow;
+
+} /* setget_input */
+
+inline unsigned long setget_output(struct gpio_private *priv, unsigned long arg)
+{
+	unsigned long flags;
+	unsigned long dir_shadow;
+
+	local_irq_save(flags);
+	dir_shadow = *dir_oe[priv->minor];
+	dir_shadow |=  (arg & changeable_dir[priv->minor]);
+	*dir_oe[priv->minor] = dir_shadow;
+	local_irq_restore(flags);
+	return dir_shadow;
+} /* setget_output */
+
+static int gpio_leds_ioctl(unsigned int cmd, unsigned long arg);
+
+static int
+gpio_ioctl_unlocked(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	unsigned long flags;
+	unsigned long val;
+	unsigned long shadow;
+	struct gpio_private *priv = file->private_data;
+	if (_IOC_TYPE(cmd) != ETRAXGPIO_IOCTYPE)
+		return -EINVAL;
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	if (priv->minor == GPIO_MINOR_V)
+		return virtual_gpio_ioctl(file, cmd, arg);
+#endif
+
+	switch (_IOC_NR(cmd)) {
+	case IO_READBITS: /* Use IO_READ_INBITS and IO_READ_OUTBITS instead */
+		/* Read the port. */
+		return *data_in[priv->minor];
+		break;
+	case IO_SETBITS:
+		local_irq_save(flags);
+		/* Set changeable bits with a 1 in arg. */
+		shadow = *data_out[priv->minor];
+		shadow |=  (arg & changeable_bits[priv->minor]);
+		*data_out[priv->minor] = shadow;
+		local_irq_restore(flags);
+		break;
+	case IO_CLRBITS:
+		local_irq_save(flags);
+		/* Clear changeable bits with a 1 in arg. */
+		shadow = *data_out[priv->minor];
+		shadow &=  ~(arg & changeable_bits[priv->minor]);
+		*data_out[priv->minor] = shadow;
+		local_irq_restore(flags);
+		break;
+	case IO_HIGHALARM:
+		/* Set alarm when bits with 1 in arg go high. */
+		priv->highalarm |= arg;
+		spin_lock_irqsave(&alarm_lock, flags);
+		gpio_some_alarms = 1;
+		if (priv->minor == GPIO_MINOR_A)
+			gpio_pa_high_alarms |= arg;
+		spin_unlock_irqrestore(&alarm_lock, flags);
+		break;
+	case IO_LOWALARM:
+		/* Set alarm when bits with 1 in arg go low. */
+		priv->lowalarm |= arg;
+		spin_lock_irqsave(&alarm_lock, flags);
+		gpio_some_alarms = 1;
+		if (priv->minor == GPIO_MINOR_A)
+			gpio_pa_low_alarms |= arg;
+		spin_unlock_irqrestore(&alarm_lock, flags);
+		break;
+	case IO_CLRALARM:
+		/* Clear alarm for bits with 1 in arg. */
+		priv->highalarm &= ~arg;
+		priv->lowalarm  &= ~arg;
+		spin_lock_irqsave(&alarm_lock, flags);
+		if (priv->minor == GPIO_MINOR_A) {
+			if (gpio_pa_high_alarms & arg ||
+			    gpio_pa_low_alarms & arg)
+				/* Must update the gpio_pa_*alarms masks */
+				;
+		}
+		spin_unlock_irqrestore(&alarm_lock, flags);
+		break;
+	case IO_READDIR: /* Use IO_SETGET_INPUT/OUTPUT instead! */
+		/* Read direction 0=input 1=output */
+		return *dir_oe[priv->minor];
+	case IO_SETINPUT: /* Use IO_SETGET_INPUT instead! */
+		/* Set direction 0=unchanged 1=input,
+		 * return mask with 1=input
+		 */
+		return setget_input(priv, arg);
+		break;
+	case IO_SETOUTPUT: /* Use IO_SETGET_OUTPUT instead! */
+		/* Set direction 0=unchanged 1=output,
+		 * return mask with 1=output
+		 */
+		return setget_output(priv, arg);
+
+	case IO_CFG_WRITE_MODE:
+	{
+		unsigned long dir_shadow;
+		dir_shadow = *dir_oe[priv->minor];
+
+		priv->clk_mask = arg & 0xFF;
+		priv->data_mask = (arg >> 8) & 0xFF;
+		priv->write_msb = (arg >> 16) & 0x01;
+		/* Check if we're allowed to change the bits and
+		 * the direction is correct
+		 */
+		if (!((priv->clk_mask & changeable_bits[priv->minor]) &&
+		      (priv->data_mask & changeable_bits[priv->minor]) &&
+		      (priv->clk_mask & dir_shadow) &&
+		      (priv->data_mask & dir_shadow))) {
+			priv->clk_mask = 0;
+			priv->data_mask = 0;
+			return -EPERM;
+		}
+		break;
+	}
+	case IO_READ_INBITS:
+		/* *arg is result of reading the input pins */
+		val = *data_in[priv->minor];
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		return 0;
+		break;
+	case IO_READ_OUTBITS:
+		 /* *arg is result of reading the output shadow */
+		val = *data_out[priv->minor];
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	case IO_SETGET_INPUT:
+		/* bits set in *arg is set to input,
+		 * *arg updated with current input pins.
+		 */
+		if (copy_from_user(&val, (unsigned long *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_input(priv, val);
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	case IO_SETGET_OUTPUT:
+		/* bits set in *arg is set to output,
+		 * *arg updated with current output pins.
+		 */
+		if (copy_from_user(&val, (unsigned long *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_output(priv, val);
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	default:
+		if (priv->minor == GPIO_MINOR_LEDS)
+			return gpio_leds_ioctl(cmd, arg);
+		else
+			return -EINVAL;
+	} /* switch */
+
+	return 0;
+}
+
+static long gpio_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+       long ret;
+
+       mutex_lock(&gpio_mutex);
+       ret = gpio_ioctl_unlocked(file, cmd, arg);
+       mutex_unlock(&gpio_mutex);
+
+       return ret;
+}
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static int
+virtual_gpio_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	unsigned long flags;
+	unsigned short val;
+	unsigned short shadow;
+	struct gpio_private *priv = file->private_data;
+
+	switch (_IOC_NR(cmd)) {
+	case IO_SETBITS:
+		local_irq_save(flags);
+		/* Set changeable bits with a 1 in arg. */
+		i2c_read(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		shadow |= ~*dir_oe[priv->minor];
+		shadow |= (arg & changeable_bits[priv->minor]);
+		i2c_write(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		local_irq_restore(flags);
+		break;
+	case IO_CLRBITS:
+		local_irq_save(flags);
+		/* Clear changeable bits with a 1 in arg. */
+		i2c_read(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		shadow |= ~*dir_oe[priv->minor];
+		shadow &= ~(arg & changeable_bits[priv->minor]);
+		i2c_write(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+		local_irq_restore(flags);
+		break;
+	case IO_HIGHALARM:
+		/* Set alarm when bits with 1 in arg go high. */
+		priv->highalarm |= arg;
+		spin_lock(&alarm_lock);
+		gpio_some_alarms = 1;
+		spin_unlock(&alarm_lock);
+		break;
+	case IO_LOWALARM:
+		/* Set alarm when bits with 1 in arg go low. */
+		priv->lowalarm |= arg;
+		spin_lock(&alarm_lock);
+		gpio_some_alarms = 1;
+		spin_unlock(&alarm_lock);
+		break;
+	case IO_CLRALARM:
+		/* Clear alarm for bits with 1 in arg. */
+		priv->highalarm &= ~arg;
+		priv->lowalarm  &= ~arg;
+		spin_lock(&alarm_lock);
+		spin_unlock(&alarm_lock);
+		break;
+	case IO_CFG_WRITE_MODE:
+	{
+		unsigned long dir_shadow;
+		dir_shadow = *dir_oe[priv->minor];
+
+		priv->clk_mask = arg & 0xFF;
+		priv->data_mask = (arg >> 8) & 0xFF;
+		priv->write_msb = (arg >> 16) & 0x01;
+		/* Check if we're allowed to change the bits and
+		 * the direction is correct
+		 */
+		if (!((priv->clk_mask & changeable_bits[priv->minor]) &&
+		      (priv->data_mask & changeable_bits[priv->minor]) &&
+		      (priv->clk_mask & dir_shadow) &&
+		      (priv->data_mask & dir_shadow))) {
+			priv->clk_mask = 0;
+			priv->data_mask = 0;
+			return -EPERM;
+		}
+		break;
+	}
+	case IO_READ_INBITS:
+		/* *arg is result of reading the input pins */
+		val = cached_virtual_gpio_read;
+		val &= ~*dir_oe[priv->minor];
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		return 0;
+		break;
+	case IO_READ_OUTBITS:
+		 /* *arg is result of reading the output shadow */
+		i2c_read(VIRT_I2C_ADDR, (void *)&val, sizeof(val));
+		val &= *dir_oe[priv->minor];
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	case IO_SETGET_INPUT:
+	{
+		/* bits set in *arg is set to input,
+		 * *arg updated with current input pins.
+		 */
+		unsigned short input_mask = ~*dir_oe[priv->minor];
+		if (copy_from_user(&val, (unsigned long *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_input(priv, val);
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		if ((input_mask & val) != input_mask) {
+			/* Input pins changed. All ports desired as input
+			 * should be set to logic 1.
+			 */
+			unsigned short change = input_mask ^ val;
+			i2c_read(VIRT_I2C_ADDR, (void *)&shadow,
+				sizeof(shadow));
+			shadow &= ~change;
+			shadow |= val;
+			i2c_write(VIRT_I2C_ADDR, (void *)&shadow,
+				sizeof(shadow));
+		}
+		break;
+	}
+	case IO_SETGET_OUTPUT:
+		/* bits set in *arg is set to output,
+		 * *arg updated with current output pins.
+		 */
+		if (copy_from_user(&val, (unsigned long *)arg, sizeof(val)))
+			return -EFAULT;
+		val = setget_output(priv, val);
+		if (copy_to_user((unsigned long *)arg, &val, sizeof(val)))
+			return -EFAULT;
+		break;
+	default:
+		return -EINVAL;
+	} /* switch */
+  return 0;
+}
+#endif /* CONFIG_ETRAX_VIRTUAL_GPIO */
+
+static int
+gpio_leds_ioctl(unsigned int cmd, unsigned long arg)
+{
+	unsigned char green;
+	unsigned char red;
+
+	switch (_IOC_NR(cmd)) {
+	case IO_LEDACTIVE_SET:
+		green = ((unsigned char) arg) & 1;
+		red   = (((unsigned char) arg) >> 1) & 1;
+		CRIS_LED_ACTIVE_SET_G(green);
+		CRIS_LED_ACTIVE_SET_R(red);
+		break;
+
+	default:
+		return -EINVAL;
+	} /* switch */
+
+	return 0;
+}
+
+static const struct file_operations gpio_fops = {
+	.owner		= THIS_MODULE,
+	.poll		= gpio_poll,
+	.unlocked_ioctl	= gpio_ioctl,
+	.write		= gpio_write,
+	.open		= gpio_open,
+	.release	= gpio_release,
+	.llseek		= noop_llseek,
+};
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+static void
+virtual_gpio_init(void)
+{
+	reg_gio_rw_intr_cfg intr_cfg;
+	reg_gio_rw_intr_mask intr_mask;
+	unsigned short shadow;
+
+	shadow = ~virtual_rw_pv_oe; /* Input ports should be set to logic 1 */
+	shadow |= CONFIG_ETRAX_DEF_GIO_PV_OUT;
+	i2c_write(VIRT_I2C_ADDR, (void *)&shadow, sizeof(shadow));
+
+	/* Set interrupt mask and on what state the interrupt shall trigger.
+	 * For virtual gpio the interrupt shall trigger on logic '0'.
+	 */
+	intr_cfg = REG_RD(gio, regi_gio, rw_intr_cfg);
+	intr_mask = REG_RD(gio, regi_gio, rw_intr_mask);
+
+	switch (CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN) {
+	case 0:
+		intr_cfg.pa0 = regk_gio_lo;
+		intr_mask.pa0 = regk_gio_yes;
+		break;
+	case 1:
+		intr_cfg.pa1 = regk_gio_lo;
+		intr_mask.pa1 = regk_gio_yes;
+		break;
+	case 2:
+		intr_cfg.pa2 = regk_gio_lo;
+		intr_mask.pa2 = regk_gio_yes;
+		break;
+	case 3:
+		intr_cfg.pa3 = regk_gio_lo;
+		intr_mask.pa3 = regk_gio_yes;
+		break;
+	case 4:
+		intr_cfg.pa4 = regk_gio_lo;
+		intr_mask.pa4 = regk_gio_yes;
+		break;
+	case 5:
+		intr_cfg.pa5 = regk_gio_lo;
+		intr_mask.pa5 = regk_gio_yes;
+		break;
+	case 6:
+		intr_cfg.pa6 = regk_gio_lo;
+		intr_mask.pa6 = regk_gio_yes;
+		break;
+	case 7:
+		intr_cfg.pa7 = regk_gio_lo;
+		intr_mask.pa7 = regk_gio_yes;
+	break;
+	}
+
+	REG_WR(gio, regi_gio, rw_intr_cfg, intr_cfg);
+	REG_WR(gio, regi_gio, rw_intr_mask, intr_mask);
+
+	gpio_pa_low_alarms |= (1 << CONFIG_ETRAX_VIRTUAL_GPIO_INTERRUPT_PA_PIN);
+	gpio_some_alarms = 1;
+}
+#endif
+
+/* main driver initialization routine, called from mem.c */
+
+static __init int
+gpio_init(void)
+{
+	int res;
+
+	/* do the formalities */
+
+	res = register_chrdev(GPIO_MAJOR, gpio_name, &gpio_fops);
+	if (res < 0) {
+		printk(KERN_ERR "gpio: couldn't get a major number.\n");
+		return res;
+	}
+
+	/* Clear all leds */
+	CRIS_LED_NETWORK_GRP0_SET(0);
+	CRIS_LED_NETWORK_GRP1_SET(0);
+	CRIS_LED_ACTIVE_SET(0);
+	CRIS_LED_DISK_READ(0);
+	CRIS_LED_DISK_WRITE(0);
+
+	printk(KERN_INFO "ETRAX FS GPIO driver v2.5, (c) 2003-2007 "
+		"Axis Communications AB\n");
+	/* We call etrax_gpio_wake_up_check() from timer interrupt */
+	if (request_irq(TIMER0_INTR_VECT, gpio_poll_timer_interrupt,
+			IRQF_SHARED, "gpio poll", &alarmlist))
+		printk(KERN_ERR "timer0 irq for gpio\n");
+
+	if (request_irq(GIO_INTR_VECT, gpio_pa_interrupt,
+			IRQF_SHARED, "gpio PA", &alarmlist))
+		printk(KERN_ERR "PA irq for gpio\n");
+
+#ifdef CONFIG_ETRAX_VIRTUAL_GPIO
+	virtual_gpio_init();
+#endif
+
+	return res;
+}
+
+/* this makes sure that gpio_init is called during kernel boot */
+
+module_init(gpio_init);
diff --git a/kernel/arch/cris/arch-v32/drivers/mach-fs/nandflash.c b/kernel/arch/cris/arch-v32/drivers/mach-fs/nandflash.c
new file mode 100644
index 000000000..e03238454
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/mach-fs/nandflash.c
@@ -0,0 +1,174 @@
+/*
+ *  arch/cris/arch-v32/drivers/nandflash.c
+ *
+ *  Copyright (c) 2004
+ *
+ *  Derived from drivers/mtd/nand/spia.c
+ *	  Copyright (C) 2000 Steven J. Hill (sjhill@realitydiluted.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/slab.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/nand.h>
+#include <linux/mtd/partitions.h>
+#include <arch/memmap.h>
+#include <hwregs/reg_map.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/gio_defs.h>
+#include <hwregs/bif_core_defs.h>
+#include <asm/io.h>
+
+#define CE_BIT 4
+#define CLE_BIT 5
+#define ALE_BIT 6
+#define BY_BIT 7
+
+struct mtd_info_wrapper {
+	struct mtd_info info;
+	struct nand_chip chip;
+};
+
+/* Bitmask for control pins */
+#define PIN_BITMASK ((1 << CE_BIT) | (1 << CLE_BIT) | (1 << ALE_BIT))
+
+/* Bitmask for mtd nand control bits */
+#define CTRL_BITMASK (NAND_NCE | NAND_CLE | NAND_ALE)
+
+
+static struct mtd_info *crisv32_mtd;
+/*
+ *	hardware specific access to control-lines
+ */
+static void crisv32_hwcontrol(struct mtd_info *mtd, int cmd,
+			      unsigned int ctrl)
+{
+	unsigned long flags;
+	reg_gio_rw_pa_dout dout;
+	struct nand_chip *this = mtd->priv;
+
+	local_irq_save(flags);
+
+	/* control bits change */
+	if (ctrl & NAND_CTRL_CHANGE) {
+		dout = REG_RD(gio, regi_gio, rw_pa_dout);
+		dout.data &= ~PIN_BITMASK;
+
+#if (CE_BIT == 4 && NAND_NCE == 1 &&  \
+     CLE_BIT == 5 && NAND_CLE == 2 && \
+     ALE_BIT == 6 && NAND_ALE == 4)
+		/* Pins in same order as control bits, but shifted.
+		 * Optimize for this case; works for 2.6.18 */
+		dout.data |= ((ctrl & CTRL_BITMASK) ^ NAND_NCE) << CE_BIT;
+#else
+		/* the slow way */
+		if (!(ctrl & NAND_NCE))
+			dout.data |= (1 << CE_BIT);
+		if (ctrl & NAND_CLE)
+			dout.data |= (1 << CLE_BIT);
+		if (ctrl & NAND_ALE)
+			dout.data |= (1 << ALE_BIT);
+#endif
+		REG_WR(gio, regi_gio, rw_pa_dout, dout);
+	}
+
+	/* command to chip */
+	if (cmd != NAND_CMD_NONE)
+		writeb(cmd, this->IO_ADDR_W);
+
+	local_irq_restore(flags);
+}
+
+/*
+*	read device ready pin
+*/
+static int crisv32_device_ready(struct mtd_info *mtd)
+{
+	reg_gio_r_pa_din din = REG_RD(gio, regi_gio, r_pa_din);
+	return ((din.data & (1 << BY_BIT)) >> BY_BIT);
+}
+
+/*
+ * Main initialization routine
+ */
+struct mtd_info *__init crisv32_nand_flash_probe(void)
+{
+	void __iomem *read_cs;
+	void __iomem *write_cs;
+
+	reg_bif_core_rw_grp3_cfg bif_cfg = REG_RD(bif_core, regi_bif_core,
+		rw_grp3_cfg);
+	reg_gio_rw_pa_oe pa_oe = REG_RD(gio, regi_gio, rw_pa_oe);
+	struct mtd_info_wrapper *wrapper;
+	struct nand_chip *this;
+	int err = 0;
+
+	/* Allocate memory for MTD device structure and private data */
+	wrapper = kzalloc(sizeof(struct mtd_info_wrapper), GFP_KERNEL);
+	if (!wrapper) {
+		printk(KERN_ERR "Unable to allocate CRISv32 NAND MTD "
+			"device structure.\n");
+		err = -ENOMEM;
+		return NULL;
+	}
+
+	read_cs = ioremap(MEM_CSP0_START | MEM_NON_CACHEABLE, 8192);
+	write_cs = ioremap(MEM_CSP1_START | MEM_NON_CACHEABLE, 8192);
+
+	if (!read_cs || !write_cs) {
+		printk(KERN_ERR "CRISv32 NAND ioremap failed\n");
+		err = -EIO;
+		goto out_mtd;
+	}
+
+	/* Get pointer to private data */
+	this = &wrapper->chip;
+	crisv32_mtd = &wrapper->info;
+
+	pa_oe.oe |= 1 << CE_BIT;
+	pa_oe.oe |= 1 << ALE_BIT;
+	pa_oe.oe |= 1 << CLE_BIT;
+	pa_oe.oe &= ~(1 << BY_BIT);
+	REG_WR(gio, regi_gio, rw_pa_oe, pa_oe);
+
+	bif_cfg.gated_csp0 = regk_bif_core_rd;
+	bif_cfg.gated_csp1 = regk_bif_core_wr;
+	REG_WR(bif_core, regi_bif_core, rw_grp3_cfg, bif_cfg);
+
+	/* Link the private data with the MTD structure */
+	crisv32_mtd->priv = this;
+
+	/* Set address of NAND IO lines */
+	this->IO_ADDR_R = read_cs;
+	this->IO_ADDR_W = write_cs;
+	this->cmd_ctrl = crisv32_hwcontrol;
+	this->dev_ready = crisv32_device_ready;
+	/* 20 us command delay time */
+	this->chip_delay = 20;
+	this->ecc.mode = NAND_ECC_SOFT;
+
+	/* Enable the following for a flash based bad block table */
+	/* this->bbt_options = NAND_BBT_USE_FLASH; */
+
+	/* Scan to find existence of the device */
+	if (nand_scan(crisv32_mtd, 1)) {
+		err = -ENXIO;
+		goto out_ior;
+	}
+
+	return crisv32_mtd;
+
+out_ior:
+	iounmap((void *)read_cs);
+	iounmap((void *)write_cs);
+out_mtd:
+	kfree(wrapper);
+	return NULL;
+}
+
diff --git a/kernel/arch/cris/arch-v32/drivers/pci/Makefile b/kernel/arch/cris/arch-v32/drivers/pci/Makefile
new file mode 100644
index 000000000..bff7482f2
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/pci/Makefile
@@ -0,0 +1,5 @@
+#
+# Makefile for Etrax cardbus driver
+#
+
+obj-$(CONFIG_ETRAX_CARDBUS)        += bios.o dma.o
diff --git a/kernel/arch/cris/arch-v32/drivers/pci/bios.c b/kernel/arch/cris/arch-v32/drivers/pci/bios.c
new file mode 100644
index 000000000..64a5fb937
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/pci/bios.c
@@ -0,0 +1,99 @@
+#include <linux/pci.h>
+#include <linux/kernel.h>
+#include <arch/hwregs/intr_vect.h>
+
+void pcibios_fixup_bus(struct pci_bus *b)
+{
+}
+
+void pcibios_set_master(struct pci_dev *dev)
+{
+	u8 lat;
+	pci_read_config_byte(dev, PCI_LATENCY_TIMER, &lat);
+	printk(KERN_DEBUG "PCI: Setting latency timer of device %s to %d\n", pci_name(dev), lat);
+	pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat);
+}
+
+int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
+			enum pci_mmap_state mmap_state, int write_combine)
+{
+	unsigned long prot;
+
+	/* Leave vm_pgoff as-is, the PCI space address is the physical
+	 * address on this platform.
+	 */
+	prot = pgprot_val(vma->vm_page_prot);
+	vma->vm_page_prot = __pgprot(prot);
+
+	/* Write-combine setting is ignored, it is changed via the mtrr
+	 * interfaces on this platform.
+	 */
+	if (remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff,
+			     vma->vm_end - vma->vm_start,
+			     vma->vm_page_prot))
+		return -EAGAIN;
+
+	return 0;
+}
+
+resource_size_t
+pcibios_align_resource(void *data, const struct resource *res,
+		       resource_size_t size, resource_size_t align)
+{
+	resource_size_t start = res->start;
+
+	if ((res->flags & IORESOURCE_IO) && (start & 0x300))
+		start = (start + 0x3ff) & ~0x3ff;
+
+	return start;
+}
+
+int pcibios_enable_resources(struct pci_dev *dev, int mask)
+{
+	u16 cmd, old_cmd;
+	int idx;
+	struct resource *r;
+
+	pci_read_config_word(dev, PCI_COMMAND, &cmd);
+	old_cmd = cmd;
+	for(idx=0; idx<6; idx++) {
+		/* Only set up the requested stuff */
+		if (!(mask & (1<<idx)))
+			continue;
+
+		r = &dev->resource[idx];
+		if (!r->start && r->end) {
+			printk(KERN_ERR "PCI: Device %s not available because of resource collisions\n", pci_name(dev));
+			return -EINVAL;
+		}
+		if (r->flags & IORESOURCE_IO)
+			cmd |= PCI_COMMAND_IO;
+		if (r->flags & IORESOURCE_MEM)
+			cmd |= PCI_COMMAND_MEMORY;
+	}
+	if (dev->resource[PCI_ROM_RESOURCE].start)
+		cmd |= PCI_COMMAND_MEMORY;
+	if (cmd != old_cmd) {
+		printk("PCI: Enabling device %s (%04x -> %04x)\n", pci_name(dev), old_cmd, cmd);
+		pci_write_config_word(dev, PCI_COMMAND, cmd);
+	}
+	return 0;
+}
+
+int pcibios_enable_irq(struct pci_dev *dev)
+{
+	dev->irq = EXT_INTR_VECT;
+	return 0;
+}
+
+int pcibios_enable_device(struct pci_dev *dev, int mask)
+{
+	int err;
+
+	if ((err = pcibios_enable_resources(dev, mask)) < 0)
+		return err;
+
+	if (!dev->msi_enabled)
+		pcibios_enable_irq(dev);
+	return 0;
+}
diff --git a/kernel/arch/cris/arch-v32/drivers/pci/dma.c b/kernel/arch/cris/arch-v32/drivers/pci/dma.c
new file mode 100644
index 000000000..ee55578d9
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/pci/dma.c
@@ -0,0 +1,50 @@
+/*
+ * Dynamic DMA mapping support.
+ *
+ * On cris there is no hardware dynamic DMA address translation,
+ * so consistent alloc/free are merely page allocation/freeing.
+ * The rest of the dynamic DMA mapping interface is implemented
+ * in asm/pci.h.
+ *
+ * Borrowed from i386.
+ */
+
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/string.h>
+#include <linux/pci.h>
+#include <linux/gfp.h>
+#include <asm/io.h>
+
+void *dma_alloc_coherent(struct device *dev, size_t size,
+			   dma_addr_t *dma_handle, gfp_t gfp)
+{
+	void *ret;
+	int order = get_order(size);
+	/* ignore region specifiers */
+	gfp &= ~(__GFP_DMA | __GFP_HIGHMEM);
+
+	if (dma_alloc_from_coherent(dev, size, dma_handle, &ret))
+		return ret;
+
+	if (dev == NULL || (dev->coherent_dma_mask < 0xffffffff))
+		gfp |= GFP_DMA;
+
+	ret = (void *)__get_free_pages(gfp, order);
+
+	if (ret != NULL) {
+		memset(ret, 0, size);
+		*dma_handle = virt_to_phys(ret);
+	}
+	return ret;
+}
+
+void dma_free_coherent(struct device *dev, size_t size,
+			 void *vaddr, dma_addr_t dma_handle)
+{
+	int order = get_order(size);
+
+	if (!dma_release_from_coherent(dev, order, vaddr))
+		free_pages((unsigned long)vaddr, order);
+}
+
diff --git a/kernel/arch/cris/arch-v32/drivers/sync_serial.c b/kernel/arch/cris/arch-v32/drivers/sync_serial.c
new file mode 100644
index 000000000..4dda9bd6b
--- /dev/null
+++ b/kernel/arch/cris/arch-v32/drivers/sync_serial.c
@@ -0,0 +1,1709 @@
+/*
+ * Simple synchronous serial port driver for ETRAX FS and ARTPEC-3.
+ *
+ * Copyright (c) 2005, 2008 Axis Communications AB
+ * Author: Mikael Starvik
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/major.h>
+#include <linux/sched.h>
+#include <linux/mutex.h>
+#include <linux/interrupt.h>
+#include <linux/poll.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/device.h>
+#include <linux/wait.h>
+
+#include <asm/io.h>
+#include <mach/dma.h>
+#include <pinmux.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/sser_defs.h>
+#include <hwregs/timer_defs.h>
+#include <hwregs/dma_defs.h>
+#include <hwregs/dma.h>
+#include <hwregs/intr_vect_defs.h>
+#include <hwregs/intr_vect.h>
+#include <hwregs/reg_map.h>
+#include <asm/sync_serial.h>
+
+
+/* The receiver is a bit tricky because of the continuous stream of data.*/
+/*                                                                       */
+/* Three DMA descriptors are linked together. Each DMA descriptor is     */
+/* responsible for port->bufchunk of a common buffer.                    */
+/*                                                                       */
+/* +---------------------------------------------+                       */
+/* |   +----------+   +----------+   +----------+ |                      */
+/* +-> | Descr[0] |-->| Descr[1] |-->| Descr[2] |-+                      */
+/*     +----------+   +----------+   +----------+                        */
+/*         |            |              |                                 */
+/*         v            v              v                                 */
+/*   +-------------------------------------+                             */
+/*   |        BUFFER                       |                             */
+/*   +-------------------------------------+                             */
+/*      |<- data_avail ->|                                               */
+/*    readp          writep                                              */
+/*                                                                       */
+/* If the application keeps up the pace readp will be right after writep.*/
+/* If the application can't keep the pace we have to throw away data.    */
+/* The idea is that readp should be ready with the data pointed out by	 */
+/* Descr[i] when the DMA has filled in Descr[i+1].                       */
+/* Otherwise we will discard	                                         */
+/* the rest of the data pointed out by Descr1 and set readp to the start */
+/* of Descr2                                                             */
+
+/* IN_BUFFER_SIZE should be a multiple of 6 to make sure that 24 bit */
+/* words can be handled */
+#define IN_DESCR_SIZE SSP_INPUT_CHUNK_SIZE
+#define NBR_IN_DESCR (8*6)
+#define IN_BUFFER_SIZE (IN_DESCR_SIZE * NBR_IN_DESCR)
+
+#define NBR_OUT_DESCR 8
+#define OUT_BUFFER_SIZE (1024 * NBR_OUT_DESCR)
+
+#define DEFAULT_FRAME_RATE 0
+#define DEFAULT_WORD_RATE 7
+
+/* To be removed when we move to pure udev. */
+#define SYNC_SERIAL_MAJOR 125
+
+/* NOTE: Enabling some debug will likely cause overrun or underrun,
+ * especially if manual mode is used.
+ */
+#define DEBUG(x)
+#define DEBUGREAD(x)
+#define DEBUGWRITE(x)
+#define DEBUGPOLL(x)
+#define DEBUGRXINT(x)
+#define DEBUGTXINT(x)
+#define DEBUGTRDMA(x)
+#define DEBUGOUTBUF(x)
+
+enum syncser_irq_setup {
+	no_irq_setup = 0,
+	dma_irq_setup = 1,
+	manual_irq_setup = 2,
+};
+
+struct sync_port {
+	unsigned long regi_sser;
+	unsigned long regi_dmain;
+	unsigned long regi_dmaout;
+
+	/* Interrupt vectors. */
+	unsigned long dma_in_intr_vect; /* Used for DMA in. */
+	unsigned long dma_out_intr_vect; /* Used for DMA out. */
+	unsigned long syncser_intr_vect; /* Used when no DMA. */
+
+	/* DMA number for in and out. */
+	unsigned int dma_in_nbr;
+	unsigned int dma_out_nbr;
+
+	/* DMA owner. */
+	enum dma_owner req_dma;
+
+	char started; /* 1 if port has been started */
+	char port_nbr; /* Port 0 or 1 */
+	char busy; /* 1 if port is busy */
+
+	char enabled;  /* 1 if port is enabled */
+	char use_dma;  /* 1 if port uses dma */
+	char tr_running;
+
+	enum syncser_irq_setup init_irqs;
+	int output;
+	int input;
+
+	/* Next byte to be read by application */
+	unsigned char *readp;
+	/* Next byte to be written by etrax */
+	unsigned char *writep;
+
+	unsigned int in_buffer_size;
+	unsigned int in_buffer_len;
+	unsigned int inbufchunk;
+	/* Data buffers for in and output. */
+	unsigned char out_buffer[OUT_BUFFER_SIZE] __aligned(32);
+	unsigned char in_buffer[IN_BUFFER_SIZE] __aligned(32);
+	unsigned char flip[IN_BUFFER_SIZE] __aligned(32);
+	struct timespec timestamp[NBR_IN_DESCR];
+	struct dma_descr_data *next_rx_desc;
+	struct dma_descr_data *prev_rx_desc;
+
+	struct timeval last_timestamp;
+	int read_ts_idx;
+	int write_ts_idx;
+
+	/* Pointer to the first available descriptor in the ring,
+	 * unless active_tr_descr == catch_tr_descr and a dma
+	 * transfer is active */
+	struct dma_descr_data *active_tr_descr;
+
+	/* Pointer to the first allocated descriptor in the ring */
+	struct dma_descr_data *catch_tr_descr;
+
+	/* Pointer to the descriptor with the current end-of-list */
+	struct dma_descr_data *prev_tr_descr;
+	int full;
+
+	/* Pointer to the first byte being read by DMA
+	 * or current position in out_buffer if not using DMA. */
+	unsigned char *out_rd_ptr;
+
+	/* Number of bytes currently locked for being read by DMA */
+	int out_buf_count;
+
+	dma_descr_context in_context __aligned(32);
+	dma_descr_context out_context __aligned(32);
+	dma_descr_data in_descr[NBR_IN_DESCR] __aligned(16);
+	dma_descr_data out_descr[NBR_OUT_DESCR] __aligned(16);
+
+	wait_queue_head_t out_wait_q;
+	wait_queue_head_t in_wait_q;
+
+	spinlock_t lock;
+};
+
+static DEFINE_MUTEX(sync_serial_mutex);
+static int etrax_sync_serial_init(void);
+static void initialize_port(int portnbr);
+static inline int sync_data_avail(struct sync_port *port);
+
+static int sync_serial_open(struct inode *, struct file *);
+static int sync_serial_release(struct inode *, struct file *);
+static unsigned int sync_serial_poll(struct file *filp, poll_table *wait);
+
+static long sync_serial_ioctl(struct file *file,
+			      unsigned int cmd, unsigned long arg);
+static int sync_serial_ioctl_unlocked(struct file *file,
+				      unsigned int cmd, unsigned long arg);
+static ssize_t sync_serial_write(struct file *file, const char __user *buf,
+				 size_t count, loff_t *ppos);
+static ssize_t sync_serial_read(struct file *file, char __user *buf,
+				size_t count, loff_t *ppos);
+
+#if ((defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT0) && \
+	defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL0_DMA)) || \
+	(defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT1) && \
+	defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL1_DMA)))
+#define SYNC_SER_DMA
+#else
+#define SYNC_SER_MANUAL
+#endif
+
+#ifdef SYNC_SER_DMA
+static void start_dma_out(struct sync_port *port, const char *data, int count);
+static void start_dma_in(struct sync_port *port);
+static irqreturn_t tr_interrupt(int irq, void *dev_id);
+static irqreturn_t rx_interrupt(int irq, void *dev_id);
+#endif
+#ifdef SYNC_SER_MANUAL
+static void send_word(struct sync_port *port);
+static irqreturn_t manual_interrupt(int irq, void *dev_id);
+#endif
+
+#define artpec_pinmux_alloc_fixed crisv32_pinmux_alloc_fixed
+#define artpec_request_dma crisv32_request_dma
+#define artpec_free_dma crisv32_free_dma
+
+#ifdef CONFIG_ETRAXFS
+/* ETRAX FS */
+#define DMA_OUT_NBR0		SYNC_SER0_TX_DMA_NBR
+#define DMA_IN_NBR0		SYNC_SER0_RX_DMA_NBR
+#define DMA_OUT_NBR1		SYNC_SER1_TX_DMA_NBR
+#define DMA_IN_NBR1		SYNC_SER1_RX_DMA_NBR
+#define PINMUX_SSER0		pinmux_sser0
+#define PINMUX_SSER1		pinmux_sser1
+#define SYNCSER_INST0		regi_sser0
+#define SYNCSER_INST1		regi_sser1
+#define SYNCSER_INTR_VECT0	SSER0_INTR_VECT
+#define SYNCSER_INTR_VECT1	SSER1_INTR_VECT
+#define OUT_DMA_INST0		regi_dma4
+#define IN_DMA_INST0		regi_dma5
+#define DMA_OUT_INTR_VECT0	DMA4_INTR_VECT
+#define DMA_OUT_INTR_VECT1	DMA7_INTR_VECT
+#define DMA_IN_INTR_VECT0	DMA5_INTR_VECT
+#define DMA_IN_INTR_VECT1	DMA6_INTR_VECT
+#define REQ_DMA_SYNCSER0	dma_sser0
+#define REQ_DMA_SYNCSER1	dma_sser1
+#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL1_DMA)
+#define PORT1_DMA 1
+#else
+#define PORT1_DMA 0
+#endif
+#elif defined(CONFIG_CRIS_MACH_ARTPEC3)
+/* ARTPEC-3 */
+#define DMA_OUT_NBR0		SYNC_SER_TX_DMA_NBR
+#define DMA_IN_NBR0		SYNC_SER_RX_DMA_NBR
+#define PINMUX_SSER0		pinmux_sser
+#define SYNCSER_INST0		regi_sser
+#define SYNCSER_INTR_VECT0	SSER_INTR_VECT
+#define OUT_DMA_INST0		regi_dma6
+#define IN_DMA_INST0		regi_dma7
+#define DMA_OUT_INTR_VECT0	DMA6_INTR_VECT
+#define DMA_IN_INTR_VECT0	DMA7_INTR_VECT
+#define REQ_DMA_SYNCSER0	dma_sser
+#define REQ_DMA_SYNCSER1	dma_sser
+#endif
+
+#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL0_DMA)
+#define PORT0_DMA 1
+#else
+#define PORT0_DMA 0
+#endif
+
+/* The ports */
+static struct sync_port ports[] = {
+	{
+		.regi_sser		= SYNCSER_INST0,
+		.regi_dmaout		= OUT_DMA_INST0,
+		.regi_dmain		= IN_DMA_INST0,
+		.use_dma		= PORT0_DMA,
+		.dma_in_intr_vect	= DMA_IN_INTR_VECT0,
+		.dma_out_intr_vect	= DMA_OUT_INTR_VECT0,
+		.dma_in_nbr		= DMA_IN_NBR0,
+		.dma_out_nbr		= DMA_OUT_NBR0,
+		.req_dma		= REQ_DMA_SYNCSER0,
+		.syncser_intr_vect	= SYNCSER_INTR_VECT0,
+	},
+#ifdef CONFIG_ETRAXFS
+	{
+		.regi_sser		= SYNCSER_INST1,
+		.regi_dmaout		= regi_dma6,
+		.regi_dmain		= regi_dma7,
+		.use_dma		= PORT1_DMA,
+		.dma_in_intr_vect	= DMA_IN_INTR_VECT1,
+		.dma_out_intr_vect	= DMA_OUT_INTR_VECT1,
+		.dma_in_nbr		= DMA_IN_NBR1,
+		.dma_out_nbr		= DMA_OUT_NBR1,
+		.req_dma		= REQ_DMA_SYNCSER1,
+		.syncser_intr_vect	= SYNCSER_INTR_VECT1,
+	},
+#endif
+};
+
+#define NBR_PORTS ARRAY_SIZE(ports)
+
+static const struct file_operations syncser_fops = {
+	.owner		= THIS_MODULE,
+	.write		= sync_serial_write,
+	.read		= sync_serial_read,
+	.poll		= sync_serial_poll,
+	.unlocked_ioctl	= sync_serial_ioctl,
+	.open		= sync_serial_open,
+	.release	= sync_serial_release,
+	.llseek		= noop_llseek,
+};
+
+static dev_t syncser_first;
+static int minor_count = NBR_PORTS;
+#define SYNCSER_NAME "syncser"
+static struct cdev *syncser_cdev;
+static struct class *syncser_class;
+
+static void sync_serial_start_port(struct sync_port *port)
+{
+	reg_sser_rw_cfg cfg = REG_RD(sser, port->regi_sser, rw_cfg);
+	reg_sser_rw_tr_cfg tr_cfg =
+		REG_RD(sser, port->regi_sser, rw_tr_cfg);
+	reg_sser_rw_rec_cfg rec_cfg =
+		REG_RD(sser, port->regi_sser, rw_rec_cfg);
+	cfg.en = regk_sser_yes;
+	tr_cfg.tr_en = regk_sser_yes;
+	rec_cfg.rec_en = regk_sser_yes;
+	REG_WR(sser, port->regi_sser, rw_cfg, cfg);
+	REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg);
+	REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg);
+	port->started = 1;
+}
+
+static void __init initialize_port(int portnbr)
+{
+	struct sync_port *port = &ports[portnbr];
+	reg_sser_rw_cfg cfg = { 0 };
+	reg_sser_rw_frm_cfg frm_cfg = { 0 };
+	reg_sser_rw_tr_cfg tr_cfg = { 0 };
+	reg_sser_rw_rec_cfg rec_cfg = { 0 };
+
+	DEBUG(pr_info("Init sync serial port %d\n", portnbr));
+
+	port->port_nbr = portnbr;
+	port->init_irqs = no_irq_setup;
+
+	port->out_rd_ptr = port->out_buffer;
+	port->out_buf_count = 0;
+
+	port->output = 1;
+	port->input = 0;
+
+	port->readp = port->flip;
+	port->writep = port->flip;
+	port->in_buffer_size = IN_BUFFER_SIZE;
+	port->in_buffer_len = 0;
+	port->inbufchunk = IN_DESCR_SIZE;
+
+	port->read_ts_idx = 0;
+	port->write_ts_idx = 0;
+
+	init_waitqueue_head(&port->out_wait_q);
+	init_waitqueue_head(&port->in_wait_q);
+
+	spin_lock_init(&port->lock);
+
+	cfg.out_clk_src = regk_sser_intern_clk;
+	cfg.out_clk_pol = regk_sser_pos;
+	cfg.clk_od_mode = regk_sser_no;
+	cfg.clk_dir = regk_sser_out;
+	cfg.gate_clk = regk_sser_no;
+	cfg.base_freq = regk_sser_f29_493;
+	cfg.clk_div = 256;
+	REG_WR(sser, port->regi_sser, rw_cfg, cfg);
+
+	frm_cfg.wordrate = DEFAULT_WORD_RATE;
+	frm_cfg.type = regk_sser_edge;
+	frm_cfg.frame_pin_dir = regk_sser_out;
+	frm_cfg.frame_pin_use = regk_sser_frm;
+	frm_cfg.status_pin_dir = regk_sser_in;
+	frm_cfg.status_pin_use = regk_sser_hold;
+	frm_cfg.out_on = regk_sser_tr;
+	frm_cfg.tr_delay = 1;
+	REG_WR(sser, port->regi_sser, rw_frm_cfg, frm_cfg);
+
+	tr_cfg.urun_stop = regk_sser_no;
+	tr_cfg.sample_size = 7;
+	tr_cfg.sh_dir = regk_sser_msbfirst;
+	tr_cfg.use_dma = port->use_dma ? regk_sser_yes : regk_sser_no;
+#if 0
+	tr_cfg.rate_ctrl = regk_sser_bulk;
+	tr_cfg.data_pin_use = regk_sser_dout;
+#else
+	tr_cfg.rate_ctrl = regk_sser_iso;
+	tr_cfg.data_pin_use = regk_sser_dout;
+#endif
+	tr_cfg.bulk_wspace = 1;
+	REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg);
+
+	rec_cfg.sample_size = 7;
+	rec_cfg.sh_dir = regk_sser_msbfirst;
+	rec_cfg.use_dma = port->use_dma ? regk_sser_yes : regk_sser_no;
+	rec_cfg.fifo_thr = regk_sser_inf;
+	REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg);
+
+#ifdef SYNC_SER_DMA
+	{
+		int i;
+		/* Setup the descriptor ring for dma out/transmit. */
+		for (i = 0; i < NBR_OUT_DESCR; i++) {
+			dma_descr_data *descr = &port->out_descr[i];
+			descr->wait = 0;
+			descr->intr = 1;
+			descr->eol = 0;
+			descr->out_eop = 0;
+			descr->next =
+				(dma_descr_data *)virt_to_phys(&descr[i+1]);
+		}
+	}
+
+	/* Create a ring from the list. */
+	port->out_descr[NBR_OUT_DESCR-1].next =
+		(dma_descr_data *)virt_to_phys(&port->out_descr[0]);
+
+	/* Setup context for traversing the ring. */
+	port->active_tr_descr = &port->out_descr[0];
+	port->prev_tr_descr = &port->out_descr[NBR_OUT_DESCR-1];
+	port->catch_tr_descr = &port->out_descr[0];
+#endif
+}
+
+static inline int sync_data_avail(struct sync_port *port)
+{
+	return port->in_buffer_len;
+}
+
+static int sync_serial_open(struct inode *inode, struct file *file)
+{
+	int ret = 0;
+	int dev = iminor(inode);
+	struct sync_port *port;
+#ifdef SYNC_SER_DMA
+	reg_dma_rw_cfg cfg = { .en = regk_dma_yes };
+	reg_dma_rw_intr_mask intr_mask = { .data = regk_dma_yes };
+#endif
+
+	DEBUG(pr_debug("Open sync serial port %d\n", dev));
+
+	if (dev < 0 || dev >= NBR_PORTS || !ports[dev].enabled) {
+		DEBUG(pr_info("Invalid minor %d\n", dev));
+		return -ENODEV;
+	}
+	port = &ports[dev];
+	/* Allow open this device twice (assuming one reader and one writer) */
+	if (port->busy == 2) {
+		DEBUG(pr_info("syncser%d is busy\n", dev));
+		return -EBUSY;
+	}
+
+	mutex_lock(&sync_serial_mutex);
+
+	/* Clear any stale date left in the flip buffer */
+	port->readp = port->writep = port->flip;
+	port->in_buffer_len = 0;
+	port->read_ts_idx = 0;
+	port->write_ts_idx = 0;
+
+	if (port->init_irqs != no_irq_setup) {
+		/* Init only on first call. */
+		port->busy++;
+		mutex_unlock(&sync_serial_mutex);
+		return 0;
+	}
+	if (port->use_dma) {
+#ifdef SYNC_SER_DMA
+		const char *tmp;
+		DEBUG(pr_info("Using DMA for syncser%d\n", dev));
+
+		tmp = dev == 0 ? "syncser0 tx" : "syncser1 tx";
+		if (request_irq(port->dma_out_intr_vect, tr_interrupt, 0,
+				tmp, port)) {
+			pr_err("Can't alloc syncser%d TX IRQ", dev);
+			ret = -EBUSY;
+			goto unlock_and_exit;
+		}
+		if (artpec_request_dma(port->dma_out_nbr, tmp,
+				DMA_VERBOSE_ON_ERROR, 0, port->req_dma)) {
+			free_irq(port->dma_out_intr_vect, port);
+			pr_err("Can't alloc syncser%d TX DMA", dev);
+			ret = -EBUSY;
+			goto unlock_and_exit;
+		}
+		tmp = dev == 0 ? "syncser0 rx" : "syncser1 rx";
+		if (request_irq(port->dma_in_intr_vect, rx_interrupt, 0,
+				tmp, port)) {
+			artpec_free_dma(port->dma_out_nbr);
+			free_irq(port->dma_out_intr_vect, port);
+			pr_err("Can't alloc syncser%d RX IRQ", dev);
+			ret = -EBUSY;
+			goto unlock_and_exit;
+		}
+		if (artpec_request_dma(port->dma_in_nbr, tmp,
+				DMA_VERBOSE_ON_ERROR, 0, port->req_dma)) {
+			artpec_free_dma(port->dma_out_nbr);
+			free_irq(port->dma_out_intr_vect, port);
+			free_irq(port->dma_in_intr_vect, port);
+			pr_err("Can't alloc syncser%d RX DMA", dev);
+			ret = -EBUSY;
+			goto unlock_and_exit;
+		}
+		/* Enable DMAs */
+		REG_WR(dma, port->regi_dmain, rw_cfg, cfg);
+		REG_WR(dma, port->regi_dmaout, rw_cfg, cfg);
+		/* Enable DMA IRQs */
+		REG_WR(dma, port->regi_dmain, rw_intr_mask, intr_mask);
+		REG_WR(dma, port->regi_dmaout, rw_intr_mask, intr_mask);
+		/* Set up wordsize = 1 for DMAs. */
+		DMA_WR_CMD(port->regi_dmain, regk_dma_set_w_size1);
+		DMA_WR_CMD(port->regi_dmaout, regk_dma_set_w_size1);
+
+		start_dma_in(port);
+		port->init_irqs = dma_irq_setup;
+#endif
+	} else { /* !port->use_dma */
+#ifdef SYNC_SER_MANUAL
+		const char *tmp = dev == 0 ? "syncser0 manual irq" :
+					     "syncser1 manual irq";
+		if (request_irq(port->syncser_intr_vect, manual_interrupt,
+				0, tmp, port)) {
+			pr_err("Can't alloc syncser%d manual irq",
+				dev);
+			ret = -EBUSY;
+			goto unlock_and_exit;
+		}
+		port->init_irqs = manual_irq_setup;
+#else
+		panic("sync_serial: Manual mode not supported\n");
+#endif /* SYNC_SER_MANUAL */
+	}
+	port->busy++;
+	ret = 0;
+
+unlock_and_exit:
+	mutex_unlock(&sync_serial_mutex);
+	return ret;
+}
+
+static int sync_serial_release(struct inode *inode, struct file *file)
+{
+	int dev = iminor(inode);
+	struct sync_port *port;
+
+	if (dev < 0 || dev >= NBR_PORTS || !ports[dev].enabled) {
+		DEBUG(pr_info("Invalid minor %d\n", dev));
+		return -ENODEV;
+	}
+	port = &ports[dev];
+	if (port->busy)
+		port->busy--;
+	if (!port->busy)
+		/* XXX */;
+	return 0;
+}
+
+static unsigned int sync_serial_poll(struct file *file, poll_table *wait)
+{
+	int dev = iminor(file_inode(file));
+	unsigned int mask = 0;
+	struct sync_port *port;
+	DEBUGPOLL(
+	static unsigned int prev_mask;
+	);
+
+	port = &ports[dev];
+
+	if (!port->started)
+		sync_serial_start_port(port);
+
+	poll_wait(file, &port->out_wait_q, wait);
+	poll_wait(file, &port->in_wait_q, wait);
+
+	/* No active transfer, descriptors are available */
+	if (port->output && !port->tr_running)
+		mask |= POLLOUT | POLLWRNORM;
+
+	/* Descriptor and buffer space available. */
+	if (port->output &&
+	    port->active_tr_descr != port->catch_tr_descr &&
+	    port->out_buf_count < OUT_BUFFER_SIZE)
+		mask |=  POLLOUT | POLLWRNORM;
+
+	/* At least an inbufchunk of data */
+	if (port->input && sync_data_avail(port) >= port->inbufchunk)
+		mask |= POLLIN | POLLRDNORM;
+
+	DEBUGPOLL(
+	if (mask != prev_mask)
+		pr_info("sync_serial_poll: mask 0x%08X %s %s\n",
+			mask,
+			mask & POLLOUT ? "POLLOUT" : "",
+			mask & POLLIN ? "POLLIN" : "");
+		prev_mask = mask;
+	);
+	return mask;
+}
+
+static ssize_t __sync_serial_read(struct file *file,
+				  char __user *buf,
+				  size_t count,
+				  loff_t *ppos,
+				  struct timespec *ts)
+{
+	unsigned long flags;
+	int dev = MINOR(file_inode(file)->i_rdev);
+	int avail;
+	struct sync_port *port;
+	unsigned char *start;
+	unsigned char *end;
+
+	if (dev < 0 || dev >= NBR_PORTS || !ports[dev].enabled) {
+		DEBUG(pr_info("Invalid minor %d\n", dev));
+		return -ENODEV;
+	}
+	port = &ports[dev];
+
+	if (!port->started)
+		sync_serial_start_port(port);
+
+	/* Calculate number of available bytes */
+	/* Save pointers to avoid that they are modified by interrupt */
+	spin_lock_irqsave(&port->lock, flags);
+	start = port->readp;
+	end = port->writep;
+	spin_unlock_irqrestore(&port->lock, flags);
+
+	while ((start == end) && !port->in_buffer_len) {
+		if (file->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		wait_event_interruptible(port->in_wait_q,
+					 !(start == end && !port->full));
+
+		if (signal_pending(current))
+			return -EINTR;
+
+		spin_lock_irqsave(&port->lock, flags);
+		start = port->readp;
+		end = port->writep;
+		spin_unlock_irqrestore(&port->lock, flags);
+	}
+
+	DEBUGREAD(pr_info("R%d c %d ri %u wi %u /%u\n",
+			  dev, count,
+			  start - port->flip, end - port->flip,
+			  port->in_buffer_size));
+
+	/* Lazy read, never return wrapped data. */
+	if (end > start)
+		avail = end - start;
+	else
+		avail = port->flip + port->in_buffer_size - start;
+
+	count = count > avail ? avail : count;
+	if (copy_to_user(buf, start, count))
+		return -EFAULT;
+
+	/* If timestamp requested, find timestamp of first returned byte
+	 * and copy it.
+	 * N.B: Applications that request timstamps MUST read data in
+	 * chunks that are multiples of IN_DESCR_SIZE.
+	 * Otherwise the timestamps will not be aligned to the data read.
+	 */
+	if (ts != NULL) {
+		int idx = port->read_ts_idx;
+		memcpy(ts, &port->timestamp[idx], sizeof(struct timespec));
+		port->read_ts_idx += count / IN_DESCR_SIZE;
+		if (port->read_ts_idx >= NBR_IN_DESCR)
+			port->read_ts_idx = 0;
+	}
+
+	spin_lock_irqsave(&port->lock, flags);
+	port->readp += count;
+	/* Check for wrap */
+	if (port->readp >= port->flip + port->in_buffer_size)
+		port->readp = port->flip;
+	port->in_buffer_len -= count;
+	port->full = 0;
+	spin_unlock_irqrestore(&port->lock, flags);
+
+	DEBUGREAD(pr_info("r %d\n", count));
+
+	return count;
+}
+
+static ssize_t sync_serial_input(struct file *file, unsigned long arg)
+{
+	struct ssp_request req;
+	int count;
+	int ret;
+
+	/* Copy the request structure from user-mode. */
+	ret = copy_from_user(&req, (struct ssp_request __user *)arg,
+		sizeof(struct ssp_request));
+
+	if (ret) {
+		DEBUG(pr_info("sync_serial_input copy from user failed\n"));
+		return -EFAULT;
+	}
+
+	/* To get the timestamps aligned, make sure that 'len'
+	 * is a multiple of IN_DESCR_SIZE.
+	 */
+	if ((req.len % IN_DESCR_SIZE) != 0) {
+		DEBUG(pr_info("sync_serial: req.len %x, IN_DESCR_SIZE %x\n",
+			      req.len, IN_DESCR_SIZE));
+		return -EFAULT;
+	}
+
+	/* Do the actual read. */
+	/* Note that req.buf is actually a pointer to user space. */
+	count = __sync_serial_read(file, req.buf, req.len,
+				   NULL, &req.ts);
+
+	if (count < 0) {
+		DEBUG(pr_info("sync_serial_input read failed\n"));
+		return count;
+	}
+
+	/* Copy the request back to user-mode. */
+	ret = copy_to_user((struct ssp_request __user *)arg, &req,
+		sizeof(struct ssp_request));
+
+	if (ret) {
+		DEBUG(pr_info("syncser input copy2user failed\n"));
+		return -EFAULT;
+	}
+
+	/* Return the number of bytes read. */
+	return count;
+}
+
+
+static int sync_serial_ioctl_unlocked(struct file *file,
+				      unsigned int cmd, unsigned long arg)
+{
+	int return_val = 0;
+	int dma_w_size = regk_dma_set_w_size1;
+	int dev = iminor(file_inode(file));
+	struct sync_port *port;
+	reg_sser_rw_tr_cfg tr_cfg;
+	reg_sser_rw_rec_cfg rec_cfg;
+	reg_sser_rw_frm_cfg frm_cfg;
+	reg_sser_rw_cfg gen_cfg;
+	reg_sser_rw_intr_mask intr_mask;
+
+	if (dev < 0 || dev >= NBR_PORTS || !ports[dev].enabled) {
+		DEBUG(pr_info("Invalid minor %d\n", dev));
+		return -1;
+	}
+
+	if (cmd == SSP_INPUT)
+		return sync_serial_input(file, arg);
+
+	port = &ports[dev];
+	spin_lock_irq(&port->lock);
+
+	tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg);
+	rec_cfg = REG_RD(sser, port->regi_sser, rw_rec_cfg);
+	frm_cfg = REG_RD(sser, port->regi_sser, rw_frm_cfg);
+	gen_cfg = REG_RD(sser, port->regi_sser, rw_cfg);
+	intr_mask = REG_RD(sser, port->regi_sser, rw_intr_mask);
+
+	switch (cmd) {
+	case SSP_SPEED:
+		if (GET_SPEED(arg) == CODEC) {
+			unsigned int freq;
+
+			gen_cfg.base_freq = regk_sser_f32;
+
+			/* Clock divider will internally be
+			 * gen_cfg.clk_div + 1.
+			 */
+
+			freq = GET_FREQ(arg);
+			switch (freq) {
+			case FREQ_32kHz:
+			case FREQ_64kHz:
+			case FREQ_128kHz:
+			case FREQ_256kHz:
+				gen_cfg.clk_div = 125 *
+					(1 << (freq - FREQ_256kHz)) - 1;
+				break;
+			case FREQ_512kHz:
+				gen_cfg.clk_div = 62;
+				break;
+			case FREQ_1MHz:
+			case FREQ_2MHz:
+			case FREQ_4MHz:
+				gen_cfg.clk_div = 8 * (1 << freq) - 1;
+				break;
+			}
+		} else if (GET_SPEED(arg) == CODEC_f32768) {
+			gen_cfg.base_freq = regk_sser_f32_768;
+			switch (GET_FREQ(arg)) {
+			case FREQ_4096kHz:
+				gen_cfg.clk_div = 7;
+				break;
+			default:
+				spin_unlock_irq(&port->lock);
+				return -EINVAL;
+			}
+		} else {
+			gen_cfg.base_freq = regk_sser_f29_493;
+			switch (GET_SPEED(arg)) {
+			case SSP150:
+				gen_cfg.clk_div = 29493000 / (150 * 8) - 1;
+				break;
+			case SSP300:
+				gen_cfg.clk_div = 29493000 / (300 * 8) - 1;
+				break;
+			case SSP600:
+				gen_cfg.clk_div = 29493000 / (600 * 8) - 1;
+				break;
+			case SSP1200:
+				gen_cfg.clk_div = 29493000 / (1200 * 8) - 1;
+				break;
+			case SSP2400:
+				gen_cfg.clk_div = 29493000 / (2400 * 8) - 1;
+				break;
+			case SSP4800:
+				gen_cfg.clk_div = 29493000 / (4800 * 8) - 1;
+				break;
+			case SSP9600:
+				gen_cfg.clk_div = 29493000 / (9600 * 8) - 1;
+				break;
+			case SSP19200:
+				gen_cfg.clk_div = 29493000 / (19200 * 8) - 1;
+				break;
+			case SSP28800:
+				gen_cfg.clk_div = 29493000 / (28800 * 8) - 1;
+				break;
+			case SSP57600:
+				gen_cfg.clk_div = 29493000 / (57600 * 8) - 1;
+				break;
+			case SSP115200:
+				gen_cfg.clk_div = 29493000 / (115200 * 8) - 1;
+				break;
+			case SSP230400:
+				gen_cfg.clk_div = 29493000 / (230400 * 8) - 1;
+				break;
+			case SSP460800:
+				gen_cfg.clk_div = 29493000 / (460800 * 8) - 1;
+				break;
+			case SSP921600:
+				gen_cfg.clk_div = 29493000 / (921600 * 8) - 1;
+				break;
+			case SSP3125000:
+				gen_cfg.base_freq = regk_sser_f100;
+				gen_cfg.clk_div = 100000000 / (3125000 * 8) - 1;
+				break;
+
+			}
+		}
+		frm_cfg.wordrate = GET_WORD_RATE(arg);
+
+		break;
+	case SSP_MODE:
+		switch (arg) {
+		case MASTER_OUTPUT:
+			port->output = 1;
+			port->input = 0;
+			frm_cfg.out_on = regk_sser_tr;
+			frm_cfg.frame_pin_dir = regk_sser_out;
+			gen_cfg.clk_dir = regk_sser_out;
+			break;
+		case SLAVE_OUTPUT:
+			port->output = 1;
+			port->input = 0;
+			frm_cfg.frame_pin_dir = regk_sser_in;
+			gen_cfg.clk_dir = regk_sser_in;
+			break;
+		case MASTER_INPUT:
+			port->output = 0;
+			port->input = 1;
+			frm_cfg.frame_pin_dir = regk_sser_out;
+			frm_cfg.out_on = regk_sser_intern_tb;
+			gen_cfg.clk_dir = regk_sser_out;
+			break;
+		case SLAVE_INPUT:
+			port->output = 0;
+			port->input = 1;
+			frm_cfg.frame_pin_dir = regk_sser_in;
+			gen_cfg.clk_dir = regk_sser_in;
+			break;
+		case MASTER_BIDIR:
+			port->output = 1;
+			port->input = 1;
+			frm_cfg.frame_pin_dir = regk_sser_out;
+			frm_cfg.out_on = regk_sser_intern_tb;
+			gen_cfg.clk_dir = regk_sser_out;
+			break;
+		case SLAVE_BIDIR:
+			port->output = 1;
+			port->input = 1;
+			frm_cfg.frame_pin_dir = regk_sser_in;
+			gen_cfg.clk_dir = regk_sser_in;
+			break;
+		default:
+			spin_unlock_irq(&port->lock);
+			return -EINVAL;
+		}
+		if (!port->use_dma || arg == MASTER_OUTPUT ||
+				arg == SLAVE_OUTPUT)
+			intr_mask.rdav = regk_sser_yes;
+		break;
+	case SSP_FRAME_SYNC:
+		if (arg & NORMAL_SYNC) {
+			frm_cfg.rec_delay = 1;
+			frm_cfg.tr_delay = 1;
+		} else if (arg & EARLY_SYNC)
+			frm_cfg.rec_delay = frm_cfg.tr_delay = 0;
+		else if (arg & LATE_SYNC) {
+			frm_cfg.tr_delay = 2;
+			frm_cfg.rec_delay = 2;
+		} else if (arg & SECOND_WORD_SYNC) {
+			frm_cfg.rec_delay = 7;
+			frm_cfg.tr_delay = 1;
+		}
+
+		tr_cfg.bulk_wspace = frm_cfg.tr_delay;
+		frm_cfg.early_wend = regk_sser_yes;
+		if (arg & BIT_SYNC)
+			frm_cfg.type = regk_sser_edge;
+		else if (arg & WORD_SYNC)
+			frm_cfg.type = regk_sser_level;
+		else if (arg & EXTENDED_SYNC)
+			frm_cfg.early_wend = regk_sser_no;
+
+		if (arg & SYNC_ON)
+			frm_cfg.frame_pin_use = regk_sser_frm;
+		else if (arg & SYNC_OFF)
+			frm_cfg.frame_pin_use = regk_sser_gio0;
+
+		dma_w_size = regk_dma_set_w_size2;
+		if (arg & WORD_SIZE_8) {
+			rec_cfg.sample_size = tr_cfg.sample_size = 7;
+			dma_w_size = regk_dma_set_w_size1;
+		} else if (arg & WORD_SIZE_12)
+			rec_cfg.sample_size = tr_cfg.sample_size = 11;
+		else if (arg & WORD_SIZE_16)
+			rec_cfg.sample_size = tr_cfg.sample_size = 15;
+		else if (arg & WORD_SIZE_24)
+			rec_cfg.sample_size = tr_cfg.sample_size = 23;
+		else if (arg & WORD_SIZE_32)
+			rec_cfg.sample_size = tr_cfg.sample_size = 31;
+
+		if (arg & BIT_ORDER_MSB)
+			rec_cfg.sh_dir = tr_cfg.sh_dir = regk_sser_msbfirst;
+		else if (arg & BIT_ORDER_LSB)
+			rec_cfg.sh_dir = tr_cfg.sh_dir = regk_sser_lsbfirst;
+
+		if (arg & FLOW_CONTROL_ENABLE) {
+			frm_cfg.status_pin_use = regk_sser_frm;
+			rec_cfg.fifo_thr = regk_sser_thr16;
+		} else if (arg & FLOW_CONTROL_DISABLE) {
+			frm_cfg.status_pin_use = regk_sser_gio0;
+			rec_cfg.fifo_thr = regk_sser_inf;
+		}
+
+		if (arg & CLOCK_NOT_GATED)
+			gen_cfg.gate_clk = regk_sser_no;
+		else if (arg & CLOCK_GATED)
+			gen_cfg.gate_clk = regk_sser_yes;
+
+		break;
+	case SSP_IPOLARITY:
+		/* NOTE!! negedge is considered NORMAL */
+		if (arg & CLOCK_NORMAL)
+			rec_cfg.clk_pol = regk_sser_neg;
+		else if (arg & CLOCK_INVERT)
+			rec_cfg.clk_pol = regk_sser_pos;
+
+		if (arg & FRAME_NORMAL)
+			frm_cfg.level = regk_sser_pos_hi;
+		else if (arg & FRAME_INVERT)
+			frm_cfg.level = regk_sser_neg_lo;
+
+		if (arg & STATUS_NORMAL)
+			gen_cfg.hold_pol = regk_sser_pos;
+		else if (arg & STATUS_INVERT)
+			gen_cfg.hold_pol = regk_sser_neg;
+		break;
+	case SSP_OPOLARITY:
+		if (arg & CLOCK_NORMAL)
+			gen_cfg.out_clk_pol = regk_sser_pos;
+		else if (arg & CLOCK_INVERT)
+			gen_cfg.out_clk_pol = regk_sser_neg;
+
+		if (arg & FRAME_NORMAL)
+			frm_cfg.level = regk_sser_pos_hi;
+		else if (arg & FRAME_INVERT)
+			frm_cfg.level = regk_sser_neg_lo;
+
+		if (arg & STATUS_NORMAL)
+			gen_cfg.hold_pol = regk_sser_pos;
+		else if (arg & STATUS_INVERT)
+			gen_cfg.hold_pol = regk_sser_neg;
+		break;
+	case SSP_SPI:
+		rec_cfg.fifo_thr = regk_sser_inf;
+		rec_cfg.sh_dir = tr_cfg.sh_dir = regk_sser_msbfirst;
+		rec_cfg.sample_size = tr_cfg.sample_size = 7;
+		frm_cfg.frame_pin_use = regk_sser_frm;
+		frm_cfg.type = regk_sser_level;
+		frm_cfg.tr_delay = 1;
+		frm_cfg.level = regk_sser_neg_lo;
+		if (arg & SPI_SLAVE) {
+			rec_cfg.clk_pol = regk_sser_neg;
+			gen_cfg.clk_dir = regk_sser_in;
+			port->input = 1;
+			port->output = 0;
+		} else {
+			gen_cfg.out_clk_pol = regk_sser_pos;
+			port->input = 0;
+			port->output = 1;
+			gen_cfg.clk_dir = regk_sser_out;
+		}
+		break;
+	case SSP_INBUFCHUNK:
+		break;
+	default:
+		return_val = -1;
+	}
+
+
+	if (port->started) {
+		rec_cfg.rec_en = port->input;
+		gen_cfg.en = (port->output | port->input);
+	}
+
+	REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg);
+	REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg);
+	REG_WR(sser, port->regi_sser, rw_frm_cfg, frm_cfg);
+	REG_WR(sser, port->regi_sser, rw_intr_mask, intr_mask);
+	REG_WR(sser, port->regi_sser, rw_cfg, gen_cfg);
+
+
+	if (cmd == SSP_FRAME_SYNC && (arg & (WORD_SIZE_8 | WORD_SIZE_12 |
+			WORD_SIZE_16 | WORD_SIZE_24 | WORD_SIZE_32))) {
+		int en = gen_cfg.en;
+		gen_cfg.en = 0;
+		REG_WR(sser, port->regi_sser, rw_cfg, gen_cfg);
+		/* ##### Should DMA be stoped before we change dma size? */
+		DMA_WR_CMD(port->regi_dmain, dma_w_size);
+		DMA_WR_CMD(port->regi_dmaout, dma_w_size);
+		gen_cfg.en = en;
+		REG_WR(sser, port->regi_sser, rw_cfg, gen_cfg);
+	}
+
+	spin_unlock_irq(&port->lock);
+	return return_val;
+}
+
+static long sync_serial_ioctl(struct file *file,
+		unsigned int cmd, unsigned long arg)
+{
+	long ret;
+
+	mutex_lock(&sync_serial_mutex);
+	ret = sync_serial_ioctl_unlocked(file, cmd, arg);
+	mutex_unlock(&sync_serial_mutex);
+
+	return ret;
+}
+
+/* NOTE: sync_serial_write does not support concurrency */
+static ssize_t sync_serial_write(struct file *file, const char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	int dev = iminor(file_inode(file));
+	DECLARE_WAITQUEUE(wait, current);
+	struct sync_port *port;
+	int trunc_count;
+	unsigned long flags;
+	int bytes_free;
+	int out_buf_count;
+
+	unsigned char *rd_ptr;       /* First allocated byte in the buffer */
+	unsigned char *wr_ptr;       /* First free byte in the buffer */
+	unsigned char *buf_stop_ptr; /* Last byte + 1 */
+
+	if (dev < 0 || dev >= NBR_PORTS || !ports[dev].enabled) {
+		DEBUG(pr_info("Invalid minor %d\n", dev));
+		return -ENODEV;
+	}
+	port = &ports[dev];
+
+	/* |<-         OUT_BUFFER_SIZE                          ->|
+	 *           |<- out_buf_count ->|
+	 *                               |<- trunc_count ->| ...->|
+	 *  ______________________________________________________
+	 * |  free   |   data            | free                   |
+	 * |_________|___________________|________________________|
+	 *           ^ rd_ptr            ^ wr_ptr
+	 */
+	DEBUGWRITE(pr_info("W d%d c %u a: %p c: %p\n",
+			   port->port_nbr, count, port->active_tr_descr,
+			   port->catch_tr_descr));
+
+	/* Read variables that may be updated by interrupts */
+	spin_lock_irqsave(&port->lock, flags);
+	rd_ptr = port->out_rd_ptr;
+	out_buf_count = port->out_buf_count;
+	spin_unlock_irqrestore(&port->lock, flags);
+
+	/* Check if resources are available */
+	if (port->tr_running &&
+	    ((port->use_dma && port->active_tr_descr == port->catch_tr_descr) ||
+	     out_buf_count >= OUT_BUFFER_SIZE)) {
+		DEBUGWRITE(pr_info("sser%d full\n", dev));
+		return -EAGAIN;
+	}
+
+	buf_stop_ptr = port->out_buffer + OUT_BUFFER_SIZE;
+
+	/* Determine pointer to the first free byte, before copying. */
+	wr_ptr = rd_ptr + out_buf_count;
+	if (wr_ptr >= buf_stop_ptr)
+		wr_ptr -= OUT_BUFFER_SIZE;
+
+	/* If we wrap the ring buffer, let the user space program handle it by
+	 * truncating the data. This could be more elegant, small buffer
+	 * fragments may occur.
+	 */
+	bytes_free = OUT_BUFFER_SIZE - out_buf_count;
+	if (wr_ptr + bytes_free > buf_stop_ptr)
+		bytes_free = buf_stop_ptr - wr_ptr;
+	trunc_count = (count < bytes_free) ? count : bytes_free;
+
+	if (copy_from_user(wr_ptr, buf, trunc_count))
+		return -EFAULT;
+
+	DEBUGOUTBUF(pr_info("%-4d + %-4d = %-4d     %p %p %p\n",
+			    out_buf_count, trunc_count,
+			    port->out_buf_count, port->out_buffer,
+			    wr_ptr, buf_stop_ptr));
+
+	/* Make sure transmitter/receiver is running */
+	if (!port->started) {
+		reg_sser_rw_cfg cfg = REG_RD(sser, port->regi_sser, rw_cfg);
+		reg_sser_rw_rec_cfg rec_cfg =
+			REG_RD(sser, port->regi_sser, rw_rec_cfg);
+		cfg.en = regk_sser_yes;
+		rec_cfg.rec_en = port->input;
+		REG_WR(sser, port->regi_sser, rw_cfg, cfg);
+		REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg);
+		port->started = 1;
+	}
+
+	/* Setup wait if blocking */
+	if (!(file->f_flags & O_NONBLOCK)) {
+		add_wait_queue(&port->out_wait_q, &wait);
+		set_current_state(TASK_INTERRUPTIBLE);
+	}
+
+	spin_lock_irqsave(&port->lock, flags);
+	port->out_buf_count += trunc_count;
+	if (port->use_dma) {
+#ifdef SYNC_SER_DMA
+		start_dma_out(port, wr_ptr, trunc_count);
+#endif
+	} else if (!port->tr_running) {
+#ifdef SYNC_SER_MANUAL
+		reg_sser_rw_intr_mask intr_mask;
+		intr_mask = REG_RD(sser, port->regi_sser, rw_intr_mask);
+		/* Start sender by writing data */
+		send_word(port);
+		/* and enable transmitter ready IRQ */
+		intr_mask.trdy = 1;
+		REG_WR(sser, port->regi_sser, rw_intr_mask, intr_mask);
+#endif
+	}
+	spin_unlock_irqrestore(&port->lock, flags);
+
+	/* Exit if non blocking */
+	if (file->f_flags & O_NONBLOCK) {
+		DEBUGWRITE(pr_info("w d%d c %u  %08x\n",
+				   port->port_nbr, trunc_count,
+				   REG_RD_INT(dma, port->regi_dmaout, r_intr)));
+		return trunc_count;
+	}
+
+	schedule();
+	remove_wait_queue(&port->out_wait_q, &wait);
+
+	if (signal_pending(current))
+		return -EINTR;
+
+	DEBUGWRITE(pr_info("w d%d c %u\n", port->port_nbr, trunc_count));
+	return trunc_count;
+}
+
+static ssize_t sync_serial_read(struct file *file, char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	return __sync_serial_read(file, buf, count, ppos, NULL);
+}
+
+#ifdef SYNC_SER_MANUAL
+static void send_word(struct sync_port *port)
+{
+	reg_sser_rw_tr_cfg tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg);
+	reg_sser_rw_tr_data tr_data =  {0};
+
+	switch (tr_cfg.sample_size) {
+	case 8:
+		port->out_buf_count--;
+		tr_data.data = *port->out_rd_ptr++;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		if (port->out_rd_ptr >= port->out_buffer + OUT_BUFFER_SIZE)
+			port->out_rd_ptr = port->out_buffer;
+		break;
+	case 12:
+	{
+		int data = (*port->out_rd_ptr++) << 8;
+		data |= *port->out_rd_ptr++;
+		port->out_buf_count -= 2;
+		tr_data.data = data;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		if (port->out_rd_ptr >= port->out_buffer + OUT_BUFFER_SIZE)
+			port->out_rd_ptr = port->out_buffer;
+		break;
+	}
+	case 16:
+		port->out_buf_count -= 2;
+		tr_data.data = *(unsigned short *)port->out_rd_ptr;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		port->out_rd_ptr += 2;
+		if (port->out_rd_ptr >= port->out_buffer + OUT_BUFFER_SIZE)
+			port->out_rd_ptr = port->out_buffer;
+		break;
+	case 24:
+		port->out_buf_count -= 3;
+		tr_data.data = *(unsigned short *)port->out_rd_ptr;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		port->out_rd_ptr += 2;
+		tr_data.data = *port->out_rd_ptr++;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		if (port->out_rd_ptr >= port->out_buffer + OUT_BUFFER_SIZE)
+			port->out_rd_ptr = port->out_buffer;
+		break;
+	case 32:
+		port->out_buf_count -= 4;
+		tr_data.data = *(unsigned short *)port->out_rd_ptr;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		port->out_rd_ptr += 2;
+		tr_data.data = *(unsigned short *)port->out_rd_ptr;
+		REG_WR(sser, port->regi_sser, rw_tr_data, tr_data);
+		port->out_rd_ptr += 2;
+		if (port->out_rd_ptr >= port->out_buffer + OUT_BUFFER_SIZE)
+			port->out_rd_ptr = port->out_buffer;
+		break;
+	}
+}
+#endif
+
+#ifdef SYNC_SER_DMA
+static void start_dma_out(struct sync_port *port, const char *data, int count)
+{
+	port->active_tr_descr->buf = (char *)virt_to_phys((char *)data);
+	port->active_tr_descr->after = port->active_tr_descr->buf + count;
+	port->active_tr_descr->intr = 1;
+
+	port->active_tr_descr->eol = 1;
+	port->prev_tr_descr->eol = 0;
+
+	DEBUGTRDMA(pr_info("Inserting eolr:%p eol@:%p\n",
+		port->prev_tr_descr, port->active_tr_descr));
+	port->prev_tr_descr = port->active_tr_descr;
+	port->active_tr_descr = phys_to_virt((int)port->active_tr_descr->next);
+
+	if (!port->tr_running) {
+		reg_sser_rw_tr_cfg tr_cfg = REG_RD(sser, port->regi_sser,
+			rw_tr_cfg);
+
+		port->out_context.next = NULL;
+		port->out_context.saved_data =
+			(dma_descr_data *)virt_to_phys(port->prev_tr_descr);
+		port->out_context.saved_data_buf = port->prev_tr_descr->buf;
+
+		DMA_START_CONTEXT(port->regi_dmaout,
+			virt_to_phys((char *)&port->out_context));
+
+		tr_cfg.tr_en = regk_sser_yes;
+		REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg);
+		DEBUGTRDMA(pr_info("dma s\n"););
+	} else {
+		DMA_CONTINUE_DATA(port->regi_dmaout);
+		DEBUGTRDMA(pr_info("dma c\n"););
+	}
+
+	port->tr_running = 1;
+}
+
+static void start_dma_in(struct sync_port *port)
+{
+	int i;
+	char *buf;
+	unsigned long flags;
+	spin_lock_irqsave(&port->lock, flags);
+	port->writep = port->flip;
+	spin_unlock_irqrestore(&port->lock, flags);
+
+	buf = (char *)virt_to_phys(port->in_buffer);
+	for (i = 0; i < NBR_IN_DESCR; i++) {
+		port->in_descr[i].buf = buf;
+		port->in_descr[i].after = buf + port->inbufchunk;
+		port->in_descr[i].intr = 1;
+		port->in_descr[i].next =
+			(dma_descr_data *)virt_to_phys(&port->in_descr[i+1]);
+		port->in_descr[i].buf = buf;
+		buf += port->inbufchunk;
+	}
+	/* Link the last descriptor to the first */
+	port->in_descr[i-1].next =
+		(dma_descr_data *)virt_to_phys(&port->in_descr[0]);
+	port->in_descr[i-1].eol = regk_sser_yes;
+	port->next_rx_desc = &port->in_descr[0];
+	port->prev_rx_desc = &port->in_descr[NBR_IN_DESCR - 1];
+	port->in_context.saved_data =
+		(dma_descr_data *)virt_to_phys(&port->in_descr[0]);
+	port->in_context.saved_data_buf = port->in_descr[0].buf;
+	DMA_START_CONTEXT(port->regi_dmain, virt_to_phys(&port->in_context));
+}
+
+static irqreturn_t tr_interrupt(int irq, void *dev_id)
+{
+	reg_dma_r_masked_intr masked;
+	reg_dma_rw_ack_intr ack_intr = { .data = regk_dma_yes };
+	reg_dma_rw_stat stat;
+	int i;
+	int found = 0;
+	int stop_sser = 0;
+
+	for (i = 0; i < NBR_PORTS; i++) {
+		struct sync_port *port = &ports[i];
+		if (!port->enabled || !port->use_dma)
+			continue;
+
+		/* IRQ active for the port? */
+		masked = REG_RD(dma, port->regi_dmaout, r_masked_intr);
+		if (!masked.data)
+			continue;
+
+		found = 1;
+
+		/* Check if we should stop the DMA transfer */
+		stat = REG_RD(dma, port->regi_dmaout, rw_stat);
+		if (stat.list_state == regk_dma_data_at_eol)
+			stop_sser = 1;
+
+		/* Clear IRQ */
+		REG_WR(dma, port->regi_dmaout, rw_ack_intr, ack_intr);
+
+		if (!stop_sser) {
+			/* The DMA has completed a descriptor, EOL was not
+			 * encountered, so step relevant descriptor and
+			 * datapointers forward. */
+			int sent;
+			sent = port->catch_tr_descr->after -
+				port->catch_tr_descr->buf;
+			DEBUGTXINT(pr_info("%-4d - %-4d = %-4d\t"
+					   "in descr %p (ac: %p)\n",
+					   port->out_buf_count, sent,
+					   port->out_buf_count - sent,
+					   port->catch_tr_descr,
+					   port->active_tr_descr););
+			port->out_buf_count -= sent;
+			port->catch_tr_descr =
+				phys_to_virt((int) port->catch_tr_descr->next);
+			port->out_rd_ptr =
+				phys_to_virt((int) port->catch_tr_descr->buf);
+		} else {
+			reg_sser_rw_tr_cfg tr_cfg;
+			int j, sent;
+			/* EOL handler.
+			 * Note that if an EOL was encountered during the irq
+			 * locked section of sync_ser_write the DMA will be
+			 * restarted and the eol flag will be cleared.
+			 * The remaining descriptors will be traversed by
+			 * the descriptor interrupts as usual.
+			 */
+			j = 0;
+			while (!port->catch_tr_descr->eol) {
+				sent = port->catch_tr_descr->after -
+					port->catch_tr_descr->buf;
+				DEBUGOUTBUF(pr_info(
+					"traversing descr %p -%d (%d)\n",
+					port->catch_tr_descr,
+					sent,
+					port->out_buf_count));
+				port->out_buf_count -= sent;
+				port->catch_tr_descr = phys_to_virt(
+					(int)port->catch_tr_descr->next);
+				j++;
+				if (j >= NBR_OUT_DESCR) {
+					/* TODO: Reset and recover */
+					panic("sync_serial: missing eol");
+				}
+			}
+			sent = port->catch_tr_descr->after -
+				port->catch_tr_descr->buf;
+			DEBUGOUTBUF(pr_info("eol at descr %p -%d (%d)\n",
+				port->catch_tr_descr,
+				sent,
+				port->out_buf_count));
+
+			port->out_buf_count -= sent;
+
+			/* Update read pointer to first free byte, we
+			 * may already be writing data there. */
+			port->out_rd_ptr =
+				phys_to_virt((int) port->catch_tr_descr->after);
+			if (port->out_rd_ptr > port->out_buffer +
+					OUT_BUFFER_SIZE)
+				port->out_rd_ptr = port->out_buffer;
+
+			tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg);
+			DEBUGTXINT(pr_info(
+				"tr_int DMA stop %d, set catch @ %p\n",
+				port->out_buf_count,
+				port->active_tr_descr));
+			if (port->out_buf_count != 0)
+				pr_err("sync_ser: buf not empty after eol\n");
+			port->catch_tr_descr = port->active_tr_descr;
+			port->tr_running = 0;
+			tr_cfg.tr_en = regk_sser_no;
+			REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg);
+		}
+		/* wake up the waiting process */
+		wake_up_interruptible(&port->out_wait_q);
+	}
+	return IRQ_RETVAL(found);
+} /* tr_interrupt */
+
+
+static inline void handle_rx_packet(struct sync_port *port)
+{
+	int idx;
+	reg_dma_rw_ack_intr ack_intr = { .data = regk_dma_yes };
+	unsigned long flags;
+
+	DEBUGRXINT(pr_info("!"));
+	spin_lock_irqsave(&port->lock, flags);
+
+	/* If we overrun the user experience is crap regardless if we
+	 * drop new or old data. Its much easier to get it right when
+	 * dropping new data so lets do that.
+	 */
+	if ((port->writep + port->inbufchunk <=
+	     port->flip + port->in_buffer_size) &&
+	    (port->in_buffer_len + port->inbufchunk < IN_BUFFER_SIZE)) {
+		memcpy(port->writep,
+		       phys_to_virt((unsigned)port->next_rx_desc->buf),
+		       port->inbufchunk);
+		port->writep += port->inbufchunk;
+		if (port->writep >= port->flip + port->in_buffer_size)
+			port->writep = port->flip;
+
+		/* Timestamp the new data chunk. */
+		if (port->write_ts_idx == NBR_IN_DESCR)
+			port->write_ts_idx = 0;
+		idx = port->write_ts_idx++;
+		do_posix_clock_monotonic_gettime(&port->timestamp[idx]);
+		port->in_buffer_len += port->inbufchunk;
+	}
+	spin_unlock_irqrestore(&port->lock, flags);
+
+	port->next_rx_desc->eol = 1;
+	port->prev_rx_desc->eol = 0;
+	/* Cache bug workaround */
+	flush_dma_descr(port->prev_rx_desc, 0);
+	port->prev_rx_desc = port->next_rx_desc;
+	port->next_rx_desc = phys_to_virt((unsigned)port->next_rx_desc->next);
+	/* Cache bug workaround */
+	flush_dma_descr(port->prev_rx_desc, 1);
+	/* wake up the waiting process */
+	wake_up_interruptible(&port->in_wait_q);
+	DMA_CONTINUE(port->regi_dmain);
+	REG_WR(dma, port->regi_dmain, rw_ack_intr, ack_intr);
+
+}
+
+static irqreturn_t rx_interrupt(int irq, void *dev_id)
+{
+	reg_dma_r_masked_intr masked;
+
+	int i;
+	int found = 0;
+
+	DEBUG(pr_info("rx_interrupt\n"));
+
+	for (i = 0; i < NBR_PORTS; i++) {
+		struct sync_port *port = &ports[i];
+
+		if (!port->enabled || !port->use_dma)
+			continue;
+
+		masked = REG_RD(dma, port->regi_dmain, r_masked_intr);
+
+		if (!masked.data)
+			continue;
+
+		/* Descriptor interrupt */
+		found = 1;
+		while (REG_RD(dma, port->regi_dmain, rw_data) !=
+				virt_to_phys(port->next_rx_desc))
+			handle_rx_packet(port);
+	}
+	return IRQ_RETVAL(found);
+} /* rx_interrupt */
+#endif /* SYNC_SER_DMA */
+
+#ifdef SYNC_SER_MANUAL
+static irqreturn_t manual_interrupt(int irq, void *dev_id)
+{
+	unsigned long flags;
+	int i;
+	int found = 0;
+	reg_sser_r_masked_intr masked;
+
+	for (i = 0; i < NBR_PORTS; i++) {
+		struct sync_port *port = &ports[i];
+
+		if (!port->enabled || port->use_dma)
+			continue;
+
+		masked = REG_RD(sser, port->regi_sser, r_masked_intr);
+		/* Data received? */
+		if (masked.rdav) {
+			reg_sser_rw_rec_cfg rec_cfg =
+				REG_RD(sser, port->regi_sser, rw_rec_cfg);
+			reg_sser_r_rec_data data = REG_RD(sser,
+				port->regi_sser, r_rec_data);
+			found = 1;
+			/* Read data */
+			spin_lock_irqsave(&port->lock, flags);
+			switch (rec_cfg.sample_size) {
+			case 8:
+				*port->writep++ = data.data & 0xff;
+				break;
+			case 12:
+				*port->writep = (data.data & 0x0ff0) >> 4;
+				*(port->writep + 1) = data.data & 0x0f;
+				port->writep += 2;
+				break;
+			case 16:
+				*(unsigned short *)port->writep = data.data;
+				port->writep += 2;
+				break;
+			case 24:
+				*(unsigned int *)port->writep = data.data;
+				port->writep += 3;
+				break;
+			case 32:
+				*(unsigned int *)port->writep = data.data;
+				port->writep += 4;
+				break;
+			}
+
+			/* Wrap? */
+			if (port->writep >= port->flip + port->in_buffer_size)
+				port->writep = port->flip;
+			if (port->writep == port->readp) {
+				/* Receive buf overrun, discard oldest data */
+				port->readp++;
+				/* Wrap? */
+				if (port->readp >= port->flip +
+						port->in_buffer_size)
+					port->readp = port->flip;
+			}
+			spin_unlock_irqrestore(&port->lock, flags);
+			if (sync_data_avail(port) >= port->inbufchunk)
+				/* Wake up application */
+				wake_up_interruptible(&port->in_wait_q);
+		}
+
+		/* Transmitter ready? */
+		if (masked.trdy) {
+			found = 1;
+			/* More data to send */
+			if (port->out_buf_count > 0)
+				send_word(port);
+			else {
+				/* Transmission finished */
+				reg_sser_rw_intr_mask intr_mask;
+				intr_mask = REG_RD(sser, port->regi_sser,
+					rw_intr_mask);
+				intr_mask.trdy = 0;
+				REG_WR(sser, port->regi_sser,
+					rw_intr_mask, intr_mask);
+				/* Wake up application */
+				wake_up_interruptible(&port->out_wait_q);
+			}
+		}
+	}
+	return IRQ_RETVAL(found);
+}
+#endif
+
+static int __init etrax_sync_serial_init(void)
+{
+#if 1
+	/* This code will be removed when we move to udev for all devices. */
+	syncser_first = MKDEV(SYNC_SERIAL_MAJOR, 0);
+	if (register_chrdev_region(syncser_first, minor_count, SYNCSER_NAME)) {
+		pr_err("Failed to register major %d\n", SYNC_SERIAL_MAJOR);
+		return -1;
+	}
+#else
+	/* Allocate dynamic major number. */
+	if (alloc_chrdev_region(&syncser_first, 0, minor_count, SYNCSER_NAME)) {
+		pr_err("Failed to allocate character device region\n");
+		return -1;
+	}
+#endif
+	syncser_cdev = cdev_alloc();
+	if (!syncser_cdev) {
+		pr_err("Failed to allocate cdev for syncser\n");
+		unregister_chrdev_region(syncser_first, minor_count);
+		return -1;
+	}
+	cdev_init(syncser_cdev, &syncser_fops);
+
+	/* Create a sysfs class for syncser */
+	syncser_class = class_create(THIS_MODULE, "syncser_class");
+
+	/* Initialize Ports */
+#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT0)
+	if (artpec_pinmux_alloc_fixed(PINMUX_SSER0)) {
+		pr_warn("Unable to alloc pins for synchronous serial port 0\n");
+		unregister_chrdev_region(syncser_first, minor_count);
+		return -EIO;
+	}
+	initialize_port(0);
+	ports[0].enabled = 1;
+	/* Register with sysfs so udev can pick it up. */
+	device_create(syncser_class, NULL, syncser_first, NULL,
+		      "%s%d", SYNCSER_NAME, 0);
+#endif
+
+#if defined(CONFIG_ETRAXFS) && defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT1)
+	if (artpec_pinmux_alloc_fixed(PINMUX_SSER1)) {
+		pr_warn("Unable to alloc pins for synchronous serial port 1\n");
+		unregister_chrdev_region(syncser_first, minor_count);
+		class_destroy(syncser_class);
+		return -EIO;
+	}
+	initialize_port(1);
+	ports[1].enabled = 1;
+	/* Register with sysfs so udev can pick it up. */
+	device_create(syncser_class, NULL, syncser_first, NULL,
+		      "%s%d", SYNCSER_NAME, 0);
+#endif
+
+	/* Add it to system */
+	if (cdev_add(syncser_cdev, syncser_first, minor_count) < 0) {
+		pr_err("Failed to add syncser as char device\n");
+		device_destroy(syncser_class, syncser_first);
+		class_destroy(syncser_class);
+		cdev_del(syncser_cdev);
+		unregister_chrdev_region(syncser_first, minor_count);
+		return -1;
+	}
+
+
+	pr_info("ARTPEC synchronous serial port (%s: %d, %d)\n",
+		SYNCSER_NAME, MAJOR(syncser_first), MINOR(syncser_first));
+
+	return 0;
+}
+
+static void __exit etrax_sync_serial_exit(void)
+{
+	int i;
+	device_destroy(syncser_class, syncser_first);
+	class_destroy(syncser_class);
+
+	if (syncser_cdev) {
+		cdev_del(syncser_cdev);
+		unregister_chrdev_region(syncser_first, minor_count);
+	}
+	for (i = 0; i < NBR_PORTS; i++) {
+		struct sync_port *port = &ports[i];
+		if (port->init_irqs == dma_irq_setup) {
+			/* Free dma irqs and dma channels. */
+#ifdef SYNC_SER_DMA
+			artpec_free_dma(port->dma_in_nbr);
+			artpec_free_dma(port->dma_out_nbr);
+			free_irq(port->dma_out_intr_vect, port);
+			free_irq(port->dma_in_intr_vect, port);
+#endif
+		} else if (port->init_irqs == manual_irq_setup) {
+			/* Free manual irq. */
+			free_irq(port->syncser_intr_vect, port);
+		}
+	}
+
+	pr_info("ARTPEC synchronous serial port unregistered\n");
+}
+
+module_init(etrax_sync_serial_init);
+module_exit(etrax_sync_serial_exit);
+
+MODULE_LICENSE("GPL");
+