diff options
Diffstat (limited to 'kernel/drivers/staging/lustre/lnet/klnds/o2iblnd')
4 files changed, 930 insertions, 1266 deletions
diff --git a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c index 3bad441de..7c730e3f7 100644 --- a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c +++ b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c @@ -38,25 +38,26 @@ * Author: Eric Barton <eric@bartonsoftware.com> */ -#include "o2iblnd.h" #include <asm/div64.h> +#include <asm/page.h> +#include "o2iblnd.h" static lnd_t the_o2iblnd = { - .lnd_type = O2IBLND, - .lnd_startup = kiblnd_startup, - .lnd_shutdown = kiblnd_shutdown, - .lnd_ctl = kiblnd_ctl, - .lnd_query = kiblnd_query, - .lnd_send = kiblnd_send, - .lnd_recv = kiblnd_recv, + .lnd_type = O2IBLND, + .lnd_startup = kiblnd_startup, + .lnd_shutdown = kiblnd_shutdown, + .lnd_ctl = kiblnd_ctl, + .lnd_query = kiblnd_query, + .lnd_send = kiblnd_send, + .lnd_recv = kiblnd_recv, }; -kib_data_t kiblnd_data; +kib_data_t kiblnd_data; static __u32 kiblnd_cksum(void *ptr, int nob) { - char *c = ptr; - __u32 sum = 0; + char *c = ptr; + __u32 sum = 0; while (nob-- > 0) sum = ((sum << 1) | (sum >> 31)) + *c++; @@ -138,10 +139,10 @@ static int kiblnd_msgtype2size(int type) static int kiblnd_unpack_rd(kib_msg_t *msg, int flip) { - kib_rdma_desc_t *rd; - int nob; - int n; - int i; + kib_rdma_desc_t *rd; + int nob; + int n; + int i; LASSERT(msg->ibm_type == IBLND_MSG_GET_REQ || msg->ibm_type == IBLND_MSG_PUT_ACK); @@ -210,10 +211,10 @@ void kiblnd_pack_msg(lnet_ni_t *ni, kib_msg_t *msg, int version, int kiblnd_unpack_msg(kib_msg_t *msg, int nob) { const int hdr_size = offsetof(kib_msg_t, ibm_u); - __u32 msg_cksum; - __u16 version; - int msg_nob; - int flip; + __u32 msg_cksum; + __u16 version; + int msg_nob; + int flip; /* 6 bytes are enough to have received magic + version */ if (nob < 6) { @@ -320,10 +321,10 @@ int kiblnd_unpack_msg(kib_msg_t *msg, int nob) int kiblnd_create_peer(lnet_ni_t *ni, kib_peer_t **peerp, lnet_nid_t nid) { - kib_peer_t *peer; - kib_net_t *net = ni->ni_data; - int cpt = lnet_cpt_of_nid(nid); - unsigned long flags; + kib_peer_t *peer; + kib_net_t *net = ni->ni_data; + int cpt = lnet_cpt_of_nid(nid); + unsigned long flags; LASSERT(net != NULL); LASSERT(nid != LNET_NID_ANY); @@ -385,9 +386,9 @@ kib_peer_t *kiblnd_find_peer_locked(lnet_nid_t nid) { /* the caller is responsible for accounting the additional reference * that this creates */ - struct list_head *peer_list = kiblnd_nid2peerlist(nid); - struct list_head *tmp; - kib_peer_t *peer; + struct list_head *peer_list = kiblnd_nid2peerlist(nid); + struct list_head *tmp; + kib_peer_t *peer; list_for_each(tmp, peer_list) { @@ -422,10 +423,10 @@ void kiblnd_unlink_peer_locked(kib_peer_t *peer) static int kiblnd_get_peer_info(lnet_ni_t *ni, int index, lnet_nid_t *nidp, int *count) { - kib_peer_t *peer; - struct list_head *ptmp; - int i; - unsigned long flags; + kib_peer_t *peer; + struct list_head *ptmp; + int i; + unsigned long flags; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); @@ -459,9 +460,9 @@ static int kiblnd_get_peer_info(lnet_ni_t *ni, int index, static void kiblnd_del_peer_locked(kib_peer_t *peer) { - struct list_head *ctmp; - struct list_head *cnxt; - kib_conn_t *conn; + struct list_head *ctmp; + struct list_head *cnxt; + kib_conn_t *conn; if (list_empty(&peer->ibp_conns)) { kiblnd_unlink_peer_locked(peer); @@ -480,14 +481,14 @@ static void kiblnd_del_peer_locked(kib_peer_t *peer) static int kiblnd_del_peer(lnet_ni_t *ni, lnet_nid_t nid) { LIST_HEAD(zombies); - struct list_head *ptmp; - struct list_head *pnxt; - kib_peer_t *peer; - int lo; - int hi; - int i; - unsigned long flags; - int rc = -ENOENT; + struct list_head *ptmp; + struct list_head *pnxt; + kib_peer_t *peer; + int lo; + int hi; + int i; + unsigned long flags; + int rc = -ENOENT; write_lock_irqsave(&kiblnd_data.kib_global_lock, flags); @@ -532,12 +533,12 @@ static int kiblnd_del_peer(lnet_ni_t *ni, lnet_nid_t nid) static kib_conn_t *kiblnd_get_conn_by_idx(lnet_ni_t *ni, int index) { - kib_peer_t *peer; - struct list_head *ptmp; - kib_conn_t *conn; - struct list_head *ctmp; - int i; - unsigned long flags; + kib_peer_t *peer; + struct list_head *ptmp; + kib_conn_t *conn; + struct list_head *ctmp; + int i; + unsigned long flags; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); @@ -593,7 +594,7 @@ int kiblnd_translate_mtu(int value) static void kiblnd_setup_mtu_locked(struct rdma_cm_id *cmid) { - int mtu; + int mtu; /* XXX There is no path record for iWARP, set by netdev->change_mtu? */ if (cmid->route.path_rec == NULL) @@ -607,11 +608,11 @@ static void kiblnd_setup_mtu_locked(struct rdma_cm_id *cmid) static int kiblnd_get_completion_vector(kib_conn_t *conn, int cpt) { - cpumask_t *mask; - int vectors; - int off; - int i; - lnet_nid_t nid = conn->ibc_peer->ibp_nid; + cpumask_t *mask; + int vectors; + int off; + int i; + lnet_nid_t nid = conn->ibc_peer->ibp_nid; vectors = conn->ibc_cmid->device->num_comp_vectors; if (vectors <= 1) @@ -642,17 +643,18 @@ kib_conn_t *kiblnd_create_conn(kib_peer_t *peer, struct rdma_cm_id *cmid, * she must dispose of 'cmid'. (Actually I'd block forever if I tried * to destroy 'cmid' here since I'm called from the CM which still has * its ref on 'cmid'). */ - rwlock_t *glock = &kiblnd_data.kib_global_lock; - kib_net_t *net = peer->ibp_ni->ni_data; - kib_dev_t *dev; + rwlock_t *glock = &kiblnd_data.kib_global_lock; + kib_net_t *net = peer->ibp_ni->ni_data; + kib_dev_t *dev; struct ib_qp_init_attr *init_qp_attr; - struct kib_sched_info *sched; - kib_conn_t *conn; - struct ib_cq *cq; - unsigned long flags; - int cpt; - int rc; - int i; + struct kib_sched_info *sched; + struct ib_cq_init_attr cq_attr = {}; + kib_conn_t *conn; + struct ib_cq *cq; + unsigned long flags; + int cpt; + int rc; + int i; LASSERT(net != NULL); LASSERT(!in_interrupt()); @@ -742,10 +744,11 @@ kib_conn_t *kiblnd_create_conn(kib_peer_t *peer, struct rdma_cm_id *cmid, kiblnd_map_rx_descs(conn); + cq_attr.cqe = IBLND_CQ_ENTRIES(version); + cq_attr.comp_vector = kiblnd_get_completion_vector(conn, cpt); cq = ib_create_cq(cmid->device, kiblnd_cq_completion, kiblnd_cq_event, conn, - IBLND_CQ_ENTRIES(version), - kiblnd_get_completion_vector(conn, cpt)); + &cq_attr); if (IS_ERR(cq)) { CERROR("Can't create CQ: %ld, cqe: %d\n", PTR_ERR(cq), IBLND_CQ_ENTRIES(version)); @@ -837,8 +840,8 @@ kib_conn_t *kiblnd_create_conn(kib_peer_t *peer, struct rdma_cm_id *cmid, void kiblnd_destroy_conn(kib_conn_t *conn) { struct rdma_cm_id *cmid = conn->ibc_cmid; - kib_peer_t *peer = conn->ibc_peer; - int rc; + kib_peer_t *peer = conn->ibc_peer; + int rc; LASSERT(!in_interrupt()); LASSERT(atomic_read(&conn->ibc_refcount) == 0); @@ -904,10 +907,10 @@ void kiblnd_destroy_conn(kib_conn_t *conn) int kiblnd_close_peer_conns_locked(kib_peer_t *peer, int why) { - kib_conn_t *conn; - struct list_head *ctmp; - struct list_head *cnxt; - int count = 0; + kib_conn_t *conn; + struct list_head *ctmp; + struct list_head *cnxt; + int count = 0; list_for_each_safe(ctmp, cnxt, &peer->ibp_conns) { conn = list_entry(ctmp, kib_conn_t, ibc_list); @@ -926,10 +929,10 @@ int kiblnd_close_peer_conns_locked(kib_peer_t *peer, int why) int kiblnd_close_stale_conns_locked(kib_peer_t *peer, int version, __u64 incarnation) { - kib_conn_t *conn; - struct list_head *ctmp; - struct list_head *cnxt; - int count = 0; + kib_conn_t *conn; + struct list_head *ctmp; + struct list_head *cnxt; + int count = 0; list_for_each_safe(ctmp, cnxt, &peer->ibp_conns) { conn = list_entry(ctmp, kib_conn_t, ibc_list); @@ -953,14 +956,14 @@ int kiblnd_close_stale_conns_locked(kib_peer_t *peer, static int kiblnd_close_matching_conns(lnet_ni_t *ni, lnet_nid_t nid) { - kib_peer_t *peer; - struct list_head *ptmp; - struct list_head *pnxt; - int lo; - int hi; - int i; - unsigned long flags; - int count = 0; + kib_peer_t *peer; + struct list_head *ptmp; + struct list_head *pnxt; + int lo; + int hi; + int i; + unsigned long flags; + int count = 0; write_lock_irqsave(&kiblnd_data.kib_global_lock, flags); @@ -1001,17 +1004,17 @@ static int kiblnd_close_matching_conns(lnet_ni_t *ni, lnet_nid_t nid) int kiblnd_ctl(lnet_ni_t *ni, unsigned int cmd, void *arg) { struct libcfs_ioctl_data *data = arg; - int rc = -EINVAL; + int rc = -EINVAL; switch (cmd) { case IOC_LIBCFS_GET_PEER: { - lnet_nid_t nid = 0; - int count = 0; + lnet_nid_t nid = 0; + int count = 0; rc = kiblnd_get_peer_info(ni, data->ioc_count, &nid, &count); - data->ioc_nid = nid; - data->ioc_count = count; + data->ioc_nid = nid; + data->ioc_count = count; break; } @@ -1053,11 +1056,11 @@ int kiblnd_ctl(lnet_ni_t *ni, unsigned int cmd, void *arg) void kiblnd_query(lnet_ni_t *ni, lnet_nid_t nid, unsigned long *when) { - unsigned long last_alive = 0; - unsigned long now = cfs_time_current(); - rwlock_t *glock = &kiblnd_data.kib_global_lock; - kib_peer_t *peer; - unsigned long flags; + unsigned long last_alive = 0; + unsigned long now = cfs_time_current(); + rwlock_t *glock = &kiblnd_data.kib_global_lock; + kib_peer_t *peer; + unsigned long flags; read_lock_irqsave(glock, flags); @@ -1086,8 +1089,8 @@ void kiblnd_query(lnet_ni_t *ni, lnet_nid_t nid, unsigned long *when) void kiblnd_free_pages(kib_pages_t *p) { - int npages = p->ibp_npages; - int i; + int npages = p->ibp_npages; + int i; for (i = 0; i < npages; i++) { if (p->ibp_pages[i] != NULL) @@ -1099,8 +1102,8 @@ void kiblnd_free_pages(kib_pages_t *p) int kiblnd_alloc_pages(kib_pages_t **pp, int cpt, int npages) { - kib_pages_t *p; - int i; + kib_pages_t *p; + int i; LIBCFS_CPT_ALLOC(p, lnet_cpt_table(), cpt, offsetof(kib_pages_t, ibp_pages[npages])); @@ -1130,7 +1133,7 @@ int kiblnd_alloc_pages(kib_pages_t **pp, int cpt, int npages) void kiblnd_unmap_rx_descs(kib_conn_t *conn) { kib_rx_t *rx; - int i; + int i; LASSERT(conn->ibc_rxs != NULL); LASSERT(conn->ibc_hdev != NULL); @@ -1153,14 +1156,13 @@ void kiblnd_unmap_rx_descs(kib_conn_t *conn) void kiblnd_map_rx_descs(kib_conn_t *conn) { - kib_rx_t *rx; - struct page *pg; - int pg_off; - int ipg; - int i; + kib_rx_t *rx; + struct page *pg; + int pg_off; + int ipg; + int i; - for (pg_off = ipg = i = 0; - i < IBLND_RX_MSGS(conn->ibc_version); i++) { + for (pg_off = ipg = i = 0; i < IBLND_RX_MSGS(conn->ibc_version); i++) { pg = conn->ibc_rx_pages->ibp_pages[ipg]; rx = &conn->ibc_rxs[i]; @@ -1177,7 +1179,7 @@ void kiblnd_map_rx_descs(kib_conn_t *conn) CDEBUG(D_NET, "rx %d: %p %#llx(%#llx)\n", i, rx->rx_msg, rx->rx_msgaddr, - lnet_page2phys(pg) + pg_off); + (__u64)(page_to_phys(pg) + pg_off)); pg_off += IBLND_MSG_SIZE; LASSERT(pg_off <= PAGE_SIZE); @@ -1192,9 +1194,9 @@ void kiblnd_map_rx_descs(kib_conn_t *conn) static void kiblnd_unmap_tx_pool(kib_tx_pool_t *tpo) { - kib_hca_dev_t *hdev = tpo->tpo_hdev; - kib_tx_t *tx; - int i; + kib_hca_dev_t *hdev = tpo->tpo_hdev; + kib_tx_t *tx; + int i; LASSERT(tpo->tpo_pool.po_allocated == 0); @@ -1216,8 +1218,8 @@ static void kiblnd_unmap_tx_pool(kib_tx_pool_t *tpo) static kib_hca_dev_t *kiblnd_current_hdev(kib_dev_t *dev) { kib_hca_dev_t *hdev; - unsigned long flags; - int i = 0; + unsigned long flags; + int i = 0; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); while (dev->ibd_failover) { @@ -1240,15 +1242,15 @@ static kib_hca_dev_t *kiblnd_current_hdev(kib_dev_t *dev) static void kiblnd_map_tx_pool(kib_tx_pool_t *tpo) { - kib_pages_t *txpgs = tpo->tpo_tx_pages; - kib_pool_t *pool = &tpo->tpo_pool; - kib_net_t *net = pool->po_owner->ps_net; - kib_dev_t *dev; - struct page *page; - kib_tx_t *tx; - int page_offset; - int ipage; - int i; + kib_pages_t *txpgs = tpo->tpo_tx_pages; + kib_pool_t *pool = &tpo->tpo_pool; + kib_net_t *net = pool->po_owner->ps_net; + kib_dev_t *dev; + struct page *page; + kib_tx_t *tx; + int page_offset; + int ipage; + int i; LASSERT(net != NULL); @@ -1291,7 +1293,7 @@ static void kiblnd_map_tx_pool(kib_tx_pool_t *tpo) struct ib_mr *kiblnd_find_dma_mr(kib_hca_dev_t *hdev, __u64 addr, __u64 size) { - __u64 index; + __u64 index; LASSERT(hdev->ibh_mrs[0] != NULL); @@ -1311,7 +1313,7 @@ struct ib_mr *kiblnd_find_rd_dma_mr(kib_hca_dev_t *hdev, kib_rdma_desc_t *rd) { struct ib_mr *prev_mr; struct ib_mr *mr; - int i; + int i; LASSERT(hdev->ibh_mrs[0] != NULL); @@ -1382,18 +1384,18 @@ static int kiblnd_create_fmr_pool(kib_fmr_poolset_t *fps, kib_fmr_pool_t **pp_fpo) { /* FMR pool for RDMA */ - kib_dev_t *dev = fps->fps_net->ibn_dev; - kib_fmr_pool_t *fpo; + kib_dev_t *dev = fps->fps_net->ibn_dev; + kib_fmr_pool_t *fpo; struct ib_fmr_pool_param param = { .max_pages_per_fmr = LNET_MAX_PAYLOAD/PAGE_SIZE, - .page_shift = PAGE_SHIFT, - .access = (IB_ACCESS_LOCAL_WRITE | + .page_shift = PAGE_SHIFT, + .access = (IB_ACCESS_LOCAL_WRITE | IB_ACCESS_REMOTE_WRITE), - .pool_size = fps->fps_pool_size, + .pool_size = fps->fps_pool_size, .dirty_watermark = fps->fps_flush_trigger, .flush_function = NULL, - .flush_arg = NULL, - .cache = !!*kiblnd_tunables.kib_fmr_cache}; + .flush_arg = NULL, + .cache = !!*kiblnd_tunables.kib_fmr_cache}; int rc; LIBCFS_CPT_ALLOC(fpo, lnet_cpt_table(), fps->fps_cpt, sizeof(*fpo)); @@ -1454,7 +1456,7 @@ static int kiblnd_init_fmr_poolset(kib_fmr_poolset_t *fps, int cpt, int flush_trigger) { kib_fmr_pool_t *fpo; - int rc; + int rc; memset(fps, 0, sizeof(kib_fmr_poolset_t)); @@ -1485,11 +1487,11 @@ static int kiblnd_fmr_pool_is_idle(kib_fmr_pool_t *fpo, unsigned long now) void kiblnd_fmr_pool_unmap(kib_fmr_t *fmr, int status) { LIST_HEAD(zombies); - kib_fmr_pool_t *fpo = fmr->fmr_pool; + kib_fmr_pool_t *fpo = fmr->fmr_pool; kib_fmr_poolset_t *fps = fpo->fpo_owner; - unsigned long now = cfs_time_current(); - kib_fmr_pool_t *tmp; - int rc; + unsigned long now = cfs_time_current(); + kib_fmr_pool_t *tmp; + int rc; rc = ib_fmr_pool_unmap(fmr->fmr_pfmr); LASSERT(rc == 0); @@ -1525,9 +1527,9 @@ int kiblnd_fmr_pool_map(kib_fmr_poolset_t *fps, __u64 *pages, int npages, __u64 iov, kib_fmr_t *fmr) { struct ib_pool_fmr *pfmr; - kib_fmr_pool_t *fpo; - __u64 version; - int rc; + kib_fmr_pool_t *fpo; + __u64 version; + int rc; again: spin_lock(&fps->fps_lock); @@ -1658,13 +1660,13 @@ static int kiblnd_init_poolset(kib_poolset_t *ps, int cpt, kib_ps_node_init_t nd_init, kib_ps_node_fini_t nd_fini) { - kib_pool_t *pool; - int rc; + kib_pool_t *pool; + int rc; memset(ps, 0, sizeof(kib_poolset_t)); - ps->ps_cpt = cpt; - ps->ps_net = net; + ps->ps_cpt = cpt; + ps->ps_net = net; ps->ps_pool_create = po_create; ps->ps_pool_destroy = po_destroy; ps->ps_node_init = nd_init; @@ -1698,9 +1700,9 @@ static int kiblnd_pool_is_idle(kib_pool_t *pool, unsigned long now) void kiblnd_pool_free_node(kib_pool_t *pool, struct list_head *node) { LIST_HEAD(zombies); - kib_poolset_t *ps = pool->po_owner; - kib_pool_t *tmp; - unsigned long now = cfs_time_current(); + kib_poolset_t *ps = pool->po_owner; + kib_pool_t *tmp; + unsigned long now = cfs_time_current(); spin_lock(&ps->ps_lock); @@ -1727,9 +1729,9 @@ void kiblnd_pool_free_node(kib_pool_t *pool, struct list_head *node) struct list_head *kiblnd_pool_alloc_node(kib_poolset_t *ps) { - struct list_head *node; - kib_pool_t *pool; - int rc; + struct list_head *node; + kib_pool_t *pool; + int rc; again: spin_lock(&ps->ps_lock); @@ -1787,144 +1789,10 @@ struct list_head *kiblnd_pool_alloc_node(kib_poolset_t *ps) goto again; } -void kiblnd_pmr_pool_unmap(kib_phys_mr_t *pmr) -{ - kib_pmr_pool_t *ppo = pmr->pmr_pool; - struct ib_mr *mr = pmr->pmr_mr; - - pmr->pmr_mr = NULL; - kiblnd_pool_free_node(&ppo->ppo_pool, &pmr->pmr_list); - if (mr != NULL) - ib_dereg_mr(mr); -} - -int kiblnd_pmr_pool_map(kib_pmr_poolset_t *pps, kib_hca_dev_t *hdev, - kib_rdma_desc_t *rd, __u64 *iova, kib_phys_mr_t **pp_pmr) -{ - kib_phys_mr_t *pmr; - struct list_head *node; - int rc; - int i; - - node = kiblnd_pool_alloc_node(&pps->pps_poolset); - if (node == NULL) { - CERROR("Failed to allocate PMR descriptor\n"); - return -ENOMEM; - } - - pmr = container_of(node, kib_phys_mr_t, pmr_list); - if (pmr->pmr_pool->ppo_hdev != hdev) { - kiblnd_pool_free_node(&pmr->pmr_pool->ppo_pool, node); - return -EAGAIN; - } - - for (i = 0; i < rd->rd_nfrags; i++) { - pmr->pmr_ipb[i].addr = rd->rd_frags[i].rf_addr; - pmr->pmr_ipb[i].size = rd->rd_frags[i].rf_nob; - } - - pmr->pmr_mr = ib_reg_phys_mr(hdev->ibh_pd, - pmr->pmr_ipb, rd->rd_nfrags, - IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE, - iova); - if (!IS_ERR(pmr->pmr_mr)) { - pmr->pmr_iova = *iova; - *pp_pmr = pmr; - return 0; - } - - rc = PTR_ERR(pmr->pmr_mr); - CERROR("Failed ib_reg_phys_mr: %d\n", rc); - - pmr->pmr_mr = NULL; - kiblnd_pool_free_node(&pmr->pmr_pool->ppo_pool, node); - - return rc; -} - -static void kiblnd_destroy_pmr_pool(kib_pool_t *pool) -{ - kib_pmr_pool_t *ppo = container_of(pool, kib_pmr_pool_t, ppo_pool); - kib_phys_mr_t *pmr; - kib_phys_mr_t *tmp; - - LASSERT(pool->po_allocated == 0); - - list_for_each_entry_safe(pmr, tmp, &pool->po_free_list, pmr_list) { - LASSERT(pmr->pmr_mr == NULL); - list_del(&pmr->pmr_list); - - if (pmr->pmr_ipb != NULL) { - LIBCFS_FREE(pmr->pmr_ipb, - IBLND_MAX_RDMA_FRAGS * - sizeof(struct ib_phys_buf)); - } - - LIBCFS_FREE(pmr, sizeof(kib_phys_mr_t)); - } - - kiblnd_fini_pool(pool); - if (ppo->ppo_hdev != NULL) - kiblnd_hdev_decref(ppo->ppo_hdev); - - LIBCFS_FREE(ppo, sizeof(kib_pmr_pool_t)); -} - -static inline int kiblnd_pmr_pool_size(int ncpts) -{ - int size = *kiblnd_tunables.kib_pmr_pool_size / ncpts; - - return max(IBLND_PMR_POOL, size); -} - -static int kiblnd_create_pmr_pool(kib_poolset_t *ps, int size, - kib_pool_t **pp_po) -{ - struct kib_pmr_pool *ppo; - struct kib_pool *pool; - kib_phys_mr_t *pmr; - int i; - - LIBCFS_CPT_ALLOC(ppo, lnet_cpt_table(), - ps->ps_cpt, sizeof(kib_pmr_pool_t)); - if (ppo == NULL) { - CERROR("Failed to allocate PMR pool\n"); - return -ENOMEM; - } - - pool = &ppo->ppo_pool; - kiblnd_init_pool(ps, pool, size); - - for (i = 0; i < size; i++) { - LIBCFS_CPT_ALLOC(pmr, lnet_cpt_table(), - ps->ps_cpt, sizeof(kib_phys_mr_t)); - if (pmr == NULL) - break; - - pmr->pmr_pool = ppo; - LIBCFS_CPT_ALLOC(pmr->pmr_ipb, lnet_cpt_table(), ps->ps_cpt, - IBLND_MAX_RDMA_FRAGS * sizeof(*pmr->pmr_ipb)); - if (pmr->pmr_ipb == NULL) - break; - - list_add(&pmr->pmr_list, &pool->po_free_list); - } - - if (i < size) { - ps->ps_pool_destroy(pool); - return -ENOMEM; - } - - ppo->ppo_hdev = kiblnd_current_hdev(ps->ps_net->ibn_dev); - *pp_po = pool; - return 0; -} - static void kiblnd_destroy_tx_pool(kib_pool_t *pool) { - kib_tx_pool_t *tpo = container_of(pool, kib_tx_pool_t, tpo_pool); - int i; + kib_tx_pool_t *tpo = container_of(pool, kib_tx_pool_t, tpo_pool); + int i; LASSERT(pool->po_allocated == 0); @@ -1979,9 +1847,9 @@ static int kiblnd_tx_pool_size(int ncpts) static int kiblnd_create_tx_pool(kib_poolset_t *ps, int size, kib_pool_t **pp_po) { - int i; - int npg; - kib_pool_t *pool; + int i; + int npg; + kib_pool_t *pool; kib_tx_pool_t *tpo; LIBCFS_CPT_ALLOC(tpo, lnet_cpt_table(), ps->ps_cpt, sizeof(*tpo)); @@ -2064,19 +1932,18 @@ static void kiblnd_tx_init(kib_pool_t *pool, struct list_head *node) { kib_tx_poolset_t *tps = container_of(pool->po_owner, kib_tx_poolset_t, tps_poolset); - kib_tx_t *tx = list_entry(node, kib_tx_t, tx_list); + kib_tx_t *tx = list_entry(node, kib_tx_t, tx_list); tx->tx_cookie = tps->tps_next_tx_cookie++; } static void kiblnd_net_fini_pools(kib_net_t *net) { - int i; + int i; cfs_cpt_for_each(i, lnet_cpt_table()) { - kib_tx_poolset_t *tps; - kib_fmr_poolset_t *fps; - kib_pmr_poolset_t *pps; + kib_tx_poolset_t *tps; + kib_fmr_poolset_t *fps; if (net->ibn_tx_ps != NULL) { tps = net->ibn_tx_ps[i]; @@ -2087,11 +1954,6 @@ static void kiblnd_net_fini_pools(kib_net_t *net) fps = net->ibn_fmr_ps[i]; kiblnd_fini_fmr_poolset(fps); } - - if (net->ibn_pmr_ps != NULL) { - pps = net->ibn_pmr_ps[i]; - kiblnd_fini_poolset(&pps->pps_poolset); - } } if (net->ibn_tx_ps != NULL) { @@ -2103,25 +1965,19 @@ static void kiblnd_net_fini_pools(kib_net_t *net) cfs_percpt_free(net->ibn_fmr_ps); net->ibn_fmr_ps = NULL; } - - if (net->ibn_pmr_ps != NULL) { - cfs_percpt_free(net->ibn_pmr_ps); - net->ibn_pmr_ps = NULL; - } } static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) { - unsigned long flags; - int cpt; - int rc; - int i; + unsigned long flags; + int cpt; + int rc = 0; + int i; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); if (*kiblnd_tunables.kib_map_on_demand == 0 && net->ibn_dev->ibd_hdev->ibh_nmrs == 1) { - read_unlock_irqrestore(&kiblnd_data.kib_global_lock, - flags); + read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags); goto create_tx_pool; } @@ -2136,12 +1992,16 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) goto failed; } - /* TX pool must be created later than FMR/PMR, see LU-2268 - * for details */ + /* + * TX pool must be created later than FMR, see LU-2268 + * for details + */ LASSERT(net->ibn_tx_ps == NULL); - /* premapping can fail if ibd_nmr > 1, so we always create - * FMR/PMR pool and map-on-demand if premapping failed */ + /* + * premapping can fail if ibd_nmr > 1, so we always create + * FMR pool and map-on-demand if premapping failed + */ net->ibn_fmr_ps = cfs_percpt_alloc(lnet_cpt_table(), sizeof(kib_fmr_poolset_t)); @@ -2157,7 +2017,7 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) kiblnd_fmr_pool_size(ncpts), kiblnd_fmr_flush_trigger(ncpts)); if (rc == -ENOSYS && i == 0) /* no FMR */ - break; /* create PMR pool */ + break; if (rc != 0) { /* a real error */ CERROR("Can't initialize FMR pool for CPT %d: %d\n", @@ -2174,38 +2034,8 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) cfs_percpt_free(net->ibn_fmr_ps); net->ibn_fmr_ps = NULL; - CWARN("Device does not support FMR, failing back to PMR\n"); - - if (*kiblnd_tunables.kib_pmr_pool_size < - *kiblnd_tunables.kib_ntx / 4) { - CERROR("Can't set pmr pool size (%d) < ntx / 4(%d)\n", - *kiblnd_tunables.kib_pmr_pool_size, - *kiblnd_tunables.kib_ntx / 4); - rc = -EINVAL; - goto failed; - } - - net->ibn_pmr_ps = cfs_percpt_alloc(lnet_cpt_table(), - sizeof(kib_pmr_poolset_t)); - if (net->ibn_pmr_ps == NULL) { - CERROR("Failed to allocate PMR pool array\n"); - rc = -ENOMEM; - goto failed; - } - - for (i = 0; i < ncpts; i++) { - cpt = (cpts == NULL) ? i : cpts[i]; - rc = kiblnd_init_poolset(&net->ibn_pmr_ps[cpt]->pps_poolset, - cpt, net, "PMR", - kiblnd_pmr_pool_size(ncpts), - kiblnd_create_pmr_pool, - kiblnd_destroy_pmr_pool, NULL, NULL); - if (rc != 0) { - CERROR("Can't initialize PMR pool for CPT %d: %d\n", - cpt, rc); + CWARN("Device does not support FMR\n"); goto failed; - } - } create_tx_pool: net->ibn_tx_ps = cfs_percpt_alloc(lnet_cpt_table(), @@ -2241,7 +2071,7 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) static int kiblnd_hdev_get_attr(kib_hca_dev_t *hdev) { struct ib_device_attr *attr; - int rc; + int rc; /* It's safe to assume a HCA can handle a page size * matching that of the native system */ @@ -2284,7 +2114,7 @@ static int kiblnd_hdev_get_attr(kib_hca_dev_t *hdev) static void kiblnd_hdev_cleanup_mrs(kib_hca_dev_t *hdev) { - int i; + int i; if (hdev->ibh_nmrs == 0 || hdev->ibh_mrs == NULL) return; @@ -2317,83 +2147,31 @@ void kiblnd_hdev_destroy(kib_hca_dev_t *hdev) static int kiblnd_hdev_setup_mrs(kib_hca_dev_t *hdev) { struct ib_mr *mr; - int i; - int rc; - __u64 mm_size; - __u64 mr_size; - int acflags = IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE; + int rc; + int acflags = IB_ACCESS_LOCAL_WRITE | IB_ACCESS_REMOTE_WRITE; rc = kiblnd_hdev_get_attr(hdev); if (rc != 0) return rc; - if (hdev->ibh_mr_shift == 64) { - LIBCFS_ALLOC(hdev->ibh_mrs, 1 * sizeof(*hdev->ibh_mrs)); - if (hdev->ibh_mrs == NULL) { - CERROR("Failed to allocate MRs table\n"); - return -ENOMEM; - } - - hdev->ibh_mrs[0] = NULL; - hdev->ibh_nmrs = 1; - - mr = ib_get_dma_mr(hdev->ibh_pd, acflags); - if (IS_ERR(mr)) { - CERROR("Failed ib_get_dma_mr : %ld\n", PTR_ERR(mr)); - kiblnd_hdev_cleanup_mrs(hdev); - return PTR_ERR(mr); - } - - hdev->ibh_mrs[0] = mr; - - goto out; - } - - mr_size = 1ULL << hdev->ibh_mr_shift; - mm_size = (unsigned long)high_memory - PAGE_OFFSET; - - hdev->ibh_nmrs = (int)((mm_size + mr_size - 1) >> hdev->ibh_mr_shift); - - if (hdev->ibh_mr_shift < 32 || hdev->ibh_nmrs > 1024) { - /* it's 4T..., assume we will re-code at that time */ - CERROR("Can't support memory size: x%#llx with MR size: x%#llx\n", - mm_size, mr_size); - return -EINVAL; - } - - /* create an array of MRs to cover all memory */ - LIBCFS_ALLOC(hdev->ibh_mrs, sizeof(*hdev->ibh_mrs) * hdev->ibh_nmrs); + LIBCFS_ALLOC(hdev->ibh_mrs, 1 * sizeof(*hdev->ibh_mrs)); if (hdev->ibh_mrs == NULL) { - CERROR("Failed to allocate MRs' table\n"); + CERROR("Failed to allocate MRs table\n"); return -ENOMEM; } - for (i = 0; i < hdev->ibh_nmrs; i++) { - struct ib_phys_buf ipb; - __u64 iova; - - ipb.size = hdev->ibh_mr_size; - ipb.addr = i * mr_size; - iova = ipb.addr; - - mr = ib_reg_phys_mr(hdev->ibh_pd, &ipb, 1, acflags, &iova); - if (IS_ERR(mr)) { - CERROR("Failed ib_reg_phys_mr addr %#llx size %#llx : %ld\n", - ipb.addr, ipb.size, PTR_ERR(mr)); - kiblnd_hdev_cleanup_mrs(hdev); - return PTR_ERR(mr); - } - - LASSERT(iova == ipb.addr); + hdev->ibh_mrs[0] = NULL; + hdev->ibh_nmrs = 1; - hdev->ibh_mrs[i] = mr; + mr = ib_get_dma_mr(hdev->ibh_pd, acflags); + if (IS_ERR(mr)) { + CERROR("Failed ib_get_dma_mr : %ld\n", PTR_ERR(mr)); + kiblnd_hdev_cleanup_mrs(hdev); + return PTR_ERR(mr); } -out: - if (hdev->ibh_mr_size != ~0ULL || hdev->ibh_nmrs != 1) - LCONSOLE_INFO("Register global MR array, MR size: %#llx, array size: %d\n", - hdev->ibh_mr_size, hdev->ibh_nmrs); + hdev->ibh_mrs[0] = mr; + return 0; } @@ -2406,10 +2184,10 @@ static int kiblnd_dummy_callback(struct rdma_cm_id *cmid, static int kiblnd_dev_need_failover(kib_dev_t *dev) { - struct rdma_cm_id *cmid; - struct sockaddr_in srcaddr; - struct sockaddr_in dstaddr; - int rc; + struct rdma_cm_id *cmid; + struct sockaddr_in srcaddr; + struct sockaddr_in dstaddr; + int rc; if (dev->ibd_hdev == NULL || /* initializing */ dev->ibd_hdev->ibh_cmid == NULL || /* listener is dead */ @@ -2435,7 +2213,7 @@ static int kiblnd_dev_need_failover(kib_dev_t *dev) } memset(&srcaddr, 0, sizeof(srcaddr)); - srcaddr.sin_family = AF_INET; + srcaddr.sin_family = AF_INET; srcaddr.sin_addr.s_addr = (__force u32)htonl(dev->ibd_ifip); memset(&dstaddr, 0, sizeof(dstaddr)); @@ -2450,13 +2228,10 @@ static int kiblnd_dev_need_failover(kib_dev_t *dev) return rc; } - if (dev->ibd_hdev->ibh_ibdev == cmid->device) { - /* don't need device failover */ - rdma_destroy_id(cmid); - return 0; - } + rc = dev->ibd_hdev->ibh_ibdev != cmid->device; /* true for failover */ + rdma_destroy_id(cmid); - return 1; + return rc; } int kiblnd_dev_failover(kib_dev_t *dev) @@ -2464,15 +2239,14 @@ int kiblnd_dev_failover(kib_dev_t *dev) LIST_HEAD(zombie_tpo); LIST_HEAD(zombie_ppo); LIST_HEAD(zombie_fpo); - struct rdma_cm_id *cmid = NULL; - kib_hca_dev_t *hdev = NULL; - kib_hca_dev_t *old; - struct ib_pd *pd; - kib_net_t *net; - struct sockaddr_in addr; - unsigned long flags; - int rc = 0; - int i; + struct rdma_cm_id *cmid = NULL; + kib_hca_dev_t *hdev = NULL; + struct ib_pd *pd; + kib_net_t *net; + struct sockaddr_in addr; + unsigned long flags; + int rc = 0; + int i; LASSERT(*kiblnd_tunables.kib_dev_failover > 1 || dev->ibd_can_failover || @@ -2558,23 +2332,16 @@ int kiblnd_dev_failover(kib_dev_t *dev) write_lock_irqsave(&kiblnd_data.kib_global_lock, flags); - old = dev->ibd_hdev; - dev->ibd_hdev = hdev; /* take over the refcount */ - hdev = old; + swap(dev->ibd_hdev, hdev); /* take over the refcount */ list_for_each_entry(net, &dev->ibd_nets, ibn_list) { cfs_cpt_for_each(i, lnet_cpt_table()) { kiblnd_fail_poolset(&net->ibn_tx_ps[i]->tps_poolset, &zombie_tpo); - if (net->ibn_fmr_ps != NULL) { + if (net->ibn_fmr_ps) kiblnd_fail_fmr_poolset(net->ibn_fmr_ps[i], &zombie_fpo); - - } else if (net->ibn_pmr_ps != NULL) { - kiblnd_fail_poolset(&net->ibn_pmr_ps[i]-> - pps_poolset, &zombie_ppo); - } } } @@ -2614,13 +2381,13 @@ void kiblnd_destroy_dev(kib_dev_t *dev) static kib_dev_t *kiblnd_create_dev(char *ifname) { struct net_device *netdev; - kib_dev_t *dev; - __u32 netmask; - __u32 ip; - int up; - int rc; + kib_dev_t *dev; + __u32 netmask; + __u32 ip; + int up; + int rc; - rc = libcfs_ipif_query(ifname, &up, &ip, &netmask); + rc = lnet_ipif_query(ifname, &up, &ip, &netmask); if (rc != 0) { CERROR("Can't query IPoIB interface %s: %d\n", ifname, rc); @@ -2665,14 +2432,11 @@ static kib_dev_t *kiblnd_create_dev(char *ifname) static void kiblnd_base_shutdown(void) { - struct kib_sched_info *sched; - int i; + struct kib_sched_info *sched; + int i; LASSERT(list_empty(&kiblnd_data.kib_devs)); - CDEBUG(D_MALLOC, "before LND base cleanup: kmem %d\n", - atomic_read(&libcfs_kmemory)); - switch (kiblnd_data.kib_init) { default: LBUG(); @@ -2723,28 +2487,22 @@ static void kiblnd_base_shutdown(void) if (kiblnd_data.kib_scheds != NULL) cfs_percpt_free(kiblnd_data.kib_scheds); - CDEBUG(D_MALLOC, "after LND base cleanup: kmem %d\n", - atomic_read(&libcfs_kmemory)); - kiblnd_data.kib_init = IBLND_INIT_NOTHING; module_put(THIS_MODULE); } void kiblnd_shutdown(lnet_ni_t *ni) { - kib_net_t *net = ni->ni_data; - rwlock_t *g_lock = &kiblnd_data.kib_global_lock; - int i; - unsigned long flags; + kib_net_t *net = ni->ni_data; + rwlock_t *g_lock = &kiblnd_data.kib_global_lock; + int i; + unsigned long flags; LASSERT(kiblnd_data.kib_init == IBLND_INIT_ALL); if (net == NULL) goto out; - CDEBUG(D_MALLOC, "before LND net cleanup: kmem %d\n", - atomic_read(&libcfs_kmemory)); - write_lock_irqsave(g_lock, flags); net->ibn_shutdown = 1; write_unlock_irqrestore(g_lock, flags); @@ -2789,9 +2547,6 @@ void kiblnd_shutdown(lnet_ni_t *ni) break; } - CDEBUG(D_MALLOC, "after LND net cleanup: kmem %d\n", - atomic_read(&libcfs_kmemory)); - net->ibn_init = IBLND_INIT_NOTHING; ni->ni_data = NULL; @@ -2804,9 +2559,9 @@ out: static int kiblnd_base_startup(void) { - struct kib_sched_info *sched; - int rc; - int i; + struct kib_sched_info *sched; + int rc; + int i; LASSERT(kiblnd_data.kib_init == IBLND_INIT_NOTHING); @@ -2821,8 +2576,7 @@ static int kiblnd_base_startup(void) kiblnd_data.kib_peer_hash_size = IBLND_PEER_HASH_SIZE; LIBCFS_ALLOC(kiblnd_data.kib_peers, - sizeof(struct list_head) * - kiblnd_data.kib_peer_hash_size); + sizeof(struct list_head) * kiblnd_data.kib_peer_hash_size); if (kiblnd_data.kib_peers == NULL) goto failed; for (i = 0; i < kiblnd_data.kib_peer_hash_size; i++) @@ -2840,7 +2594,7 @@ static int kiblnd_base_startup(void) goto failed; cfs_percpt_for_each(sched, i, kiblnd_data.kib_scheds) { - int nthrs; + int nthrs; spin_lock_init(&sched->ibs_lock); INIT_LIST_HEAD(&sched->ibs_conns); @@ -2893,9 +2647,9 @@ static int kiblnd_base_startup(void) static int kiblnd_start_schedulers(struct kib_sched_info *sched) { - int rc = 0; - int nthrs; - int i; + int rc = 0; + int nthrs; + int i; if (sched->ibs_nthreads == 0) { if (*kiblnd_tunables.kib_nscheds > 0) { @@ -2913,8 +2667,8 @@ static int kiblnd_start_schedulers(struct kib_sched_info *sched) } for (i = 0; i < nthrs; i++) { - long id; - char name[20]; + long id; + char name[20]; id = KIB_THREAD_ID(sched->ibs_cpt, sched->ibs_nthreads + i); snprintf(name, sizeof(name), "kiblnd_sd_%02ld_%02ld", @@ -2935,9 +2689,9 @@ static int kiblnd_start_schedulers(struct kib_sched_info *sched) static int kiblnd_dev_start_threads(kib_dev_t *dev, int newdev, __u32 *cpts, int ncpts) { - int cpt; - int rc; - int i; + int cpt; + int rc; + int i; for (i = 0; i < ncpts; i++) { struct kib_sched_info *sched; @@ -2960,10 +2714,10 @@ static int kiblnd_dev_start_threads(kib_dev_t *dev, int newdev, __u32 *cpts, static kib_dev_t *kiblnd_dev_search(char *ifname) { - kib_dev_t *alias = NULL; - kib_dev_t *dev; - char *colon; - char *colon2; + kib_dev_t *alias = NULL; + kib_dev_t *dev; + char *colon; + char *colon2; colon = strchr(ifname, ':'); list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) { @@ -2992,13 +2746,13 @@ static kib_dev_t *kiblnd_dev_search(char *ifname) int kiblnd_startup(lnet_ni_t *ni) { - char *ifname; - kib_dev_t *ibdev = NULL; - kib_net_t *net; - struct timeval tv; - unsigned long flags; - int rc; - int newdev; + char *ifname; + kib_dev_t *ibdev = NULL; + kib_net_t *net; + struct timespec64 tv; + unsigned long flags; + int rc; + int newdev; LASSERT(ni->ni_lnd == &the_o2iblnd); @@ -3013,8 +2767,9 @@ int kiblnd_startup(lnet_ni_t *ni) if (net == NULL) goto net_failed; - do_gettimeofday(&tv); - net->ibn_incarnation = (((__u64)tv.tv_sec) * 1000000) + tv.tv_usec; + ktime_get_real_ts64(&tv); + net->ibn_incarnation = tv.tv_sec * USEC_PER_SEC + + tv.tv_nsec / NSEC_PER_USEC; ni->ni_peertimeout = *kiblnd_tunables.kib_peertimeout; ni->ni_maxtxcredits = *kiblnd_tunables.kib_credits; @@ -3091,7 +2846,7 @@ static void __exit kiblnd_module_fini(void) static int __init kiblnd_module_init(void) { - int rc; + int rc; CLASSERT(sizeof(kib_msg_t) <= IBLND_MSG_SIZE); CLASSERT(offsetof(kib_msg_t, diff --git a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h index cd664d025..263db37de 100644 --- a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h +++ b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h @@ -48,7 +48,7 @@ #include <linux/uio.h> #include <linux/uaccess.h> -#include <asm/io.h> +#include <linux/io.h> #include <linux/fs.h> #include <linux/file.h> @@ -65,7 +65,6 @@ #include "../../../include/linux/libcfs/libcfs.h" #include "../../../include/linux/lnet/lnet.h" #include "../../../include/linux/lnet/lib-lnet.h" -#include "../../../include/linux/lnet/lnet-sysctl.h" #include <rdma/rdma_cm.h> #include <rdma/ib_cm.h> @@ -80,43 +79,42 @@ #define IBLND_N_SCHED_HIGH 4 typedef struct { - int *kib_dev_failover; /* HCA failover */ - unsigned int *kib_service; /* IB service number */ - int *kib_min_reconnect_interval; /* first failed connection retry... */ - int *kib_max_reconnect_interval; /* ...exponentially increasing to this */ - int *kib_cksum; /* checksum kib_msg_t? */ - int *kib_timeout; /* comms timeout (seconds) */ - int *kib_keepalive; /* keepalive timeout (seconds) */ - int *kib_ntx; /* # tx descs */ - int *kib_credits; /* # concurrent sends */ - int *kib_peertxcredits; /* # concurrent sends to 1 peer */ - int *kib_peerrtrcredits; /* # per-peer router buffer credits */ - int *kib_peercredits_hiw; /* # when eagerly to return credits */ - int *kib_peertimeout; /* seconds to consider peer dead */ - char **kib_default_ipif; /* default IPoIB interface */ - int *kib_retry_count; - int *kib_rnr_retry_count; - int *kib_concurrent_sends; /* send work queue sizing */ - int *kib_ib_mtu; /* IB MTU */ - int *kib_map_on_demand; /* map-on-demand if RD has more fragments - * than this value, 0 disable map-on-demand */ - int *kib_pmr_pool_size; /* # physical MR in pool */ - int *kib_fmr_pool_size; /* # FMRs in pool */ - int *kib_fmr_flush_trigger; /* When to trigger FMR flush */ - int *kib_fmr_cache; /* enable FMR pool cache? */ - int *kib_require_priv_port;/* accept only privileged ports */ - int *kib_use_priv_port; /* use privileged port for active connect */ - /* # threads on each CPT */ - int *kib_nscheds; + int *kib_dev_failover; /* HCA failover */ + unsigned int *kib_service; /* IB service number */ + int *kib_min_reconnect_interval; /* first failed connection retry... */ + int *kib_max_reconnect_interval; /* exponentially increasing to this */ + int *kib_cksum; /* checksum kib_msg_t? */ + int *kib_timeout; /* comms timeout (seconds) */ + int *kib_keepalive; /* keepalive timeout (seconds) */ + int *kib_ntx; /* # tx descs */ + int *kib_credits; /* # concurrent sends */ + int *kib_peertxcredits; /* # concurrent sends to 1 peer */ + int *kib_peerrtrcredits; /* # per-peer router buffer credits */ + int *kib_peercredits_hiw; /* # when eagerly to return credits */ + int *kib_peertimeout; /* seconds to consider peer dead */ + char **kib_default_ipif; /* default IPoIB interface */ + int *kib_retry_count; + int *kib_rnr_retry_count; + int *kib_concurrent_sends; /* send work queue sizing */ + int *kib_ib_mtu; /* IB MTU */ + int *kib_map_on_demand; /* map-on-demand if RD has more */ + /* fragments than this value, 0 */ + /* disable map-on-demand */ + int *kib_fmr_pool_size; /* # FMRs in pool */ + int *kib_fmr_flush_trigger; /* When to trigger FMR flush */ + int *kib_fmr_cache; /* enable FMR pool cache? */ + int *kib_require_priv_port; /* accept only privileged ports */ + int *kib_use_priv_port; /* use privileged port for active connect */ + int *kib_nscheds; /* # threads on each CPT */ } kib_tunables_t; extern kib_tunables_t kiblnd_tunables; -#define IBLND_MSG_QUEUE_SIZE_V1 8 /* V1 only : # messages/RDMAs in-flight */ -#define IBLND_CREDIT_HIGHWATER_V1 7 /* V1 only : when eagerly to return credits */ +#define IBLND_MSG_QUEUE_SIZE_V1 8 /* V1 only : # messages/RDMAs in-flight */ +#define IBLND_CREDIT_HIGHWATER_V1 7 /* V1 only : when eagerly to return credits */ -#define IBLND_CREDITS_DEFAULT 8 /* default # of peer credits */ -#define IBLND_CREDITS_MAX ((typeof(((kib_msg_t*) 0)->ibm_credits)) - 1) /* Max # of peer credits */ +#define IBLND_CREDITS_DEFAULT 8 /* default # of peer credits */ +#define IBLND_CREDITS_MAX ((typeof(((kib_msg_t *) 0)->ibm_credits)) - 1) /* Max # of peer credits */ #define IBLND_MSG_QUEUE_SIZE(v) ((v) == IBLND_MSG_VERSION_1 ? \ IBLND_MSG_QUEUE_SIZE_V1 : \ @@ -125,7 +123,9 @@ extern kib_tunables_t kiblnd_tunables; IBLND_CREDIT_HIGHWATER_V1 : \ *kiblnd_tunables.kib_peercredits_hiw) /* when eagerly to return credits */ -#define kiblnd_rdma_create_id(cb, dev, ps, qpt) rdma_create_id(cb, dev, ps, qpt) +#define kiblnd_rdma_create_id(cb, dev, ps, qpt) rdma_create_id(&init_net, \ + cb, dev, \ + ps, qpt) static inline int kiblnd_concurrent_sends_v1(void) @@ -159,7 +159,6 @@ kiblnd_concurrent_sends_v1(void) /* Pools (shared by connections on each CPT) */ /* These pools can grow at runtime, so don't need give a very large value */ #define IBLND_TX_POOL 256 -#define IBLND_PMR_POOL 256 #define IBLND_FMR_POOL 256 #define IBLND_FMR_POOL_FLUSH 192 @@ -186,34 +185,35 @@ struct kib_hca_dev; #endif typedef struct { - struct list_head ibd_list; /* chain on kib_devs */ - struct list_head ibd_fail_list; /* chain on kib_failed_devs */ - __u32 ibd_ifip; /* IPoIB interface IP */ - /** IPoIB interface name */ - char ibd_ifname[KIB_IFNAME_SIZE]; - int ibd_nnets; /* # nets extant */ - - unsigned long ibd_next_failover; - int ibd_failed_failover; /* # failover failures */ - unsigned int ibd_failover; /* failover in progress */ - unsigned int ibd_can_failover; /* IPoIB interface is a bonding master */ - struct list_head ibd_nets; - struct kib_hca_dev *ibd_hdev; + struct list_head ibd_list; /* chain on kib_devs */ + struct list_head ibd_fail_list; /* chain on kib_failed_devs */ + __u32 ibd_ifip; /* IPoIB interface IP */ + + /* IPoIB interface name */ + char ibd_ifname[KIB_IFNAME_SIZE]; + int ibd_nnets; /* # nets extant */ + + unsigned long ibd_next_failover; + int ibd_failed_failover; /* # failover failures */ + unsigned int ibd_failover; /* failover in progress */ + unsigned int ibd_can_failover; /* IPoIB interface is a bonding master */ + struct list_head ibd_nets; + struct kib_hca_dev *ibd_hdev; } kib_dev_t; typedef struct kib_hca_dev { - struct rdma_cm_id *ibh_cmid; /* listener cmid */ - struct ib_device *ibh_ibdev; /* IB device */ - int ibh_page_shift; /* page shift of current HCA */ - int ibh_page_size; /* page size of current HCA */ - __u64 ibh_page_mask; /* page mask of current HCA */ - int ibh_mr_shift; /* bits shift of max MR size */ - __u64 ibh_mr_size; /* size of MR */ - int ibh_nmrs; /* # of global MRs */ - struct ib_mr **ibh_mrs; /* global MR */ - struct ib_pd *ibh_pd; /* PD */ - kib_dev_t *ibh_dev; /* owner */ - atomic_t ibh_ref; /* refcount */ + struct rdma_cm_id *ibh_cmid; /* listener cmid */ + struct ib_device *ibh_ibdev; /* IB device */ + int ibh_page_shift; /* page shift of current HCA */ + int ibh_page_size; /* page size of current HCA */ + __u64 ibh_page_mask; /* page mask of current HCA */ + int ibh_mr_shift; /* bits shift of max MR size */ + __u64 ibh_mr_size; /* size of MR */ + int ibh_nmrs; /* # of global MRs */ + struct ib_mr **ibh_mrs; /* global MR */ + struct ib_pd *ibh_pd; /* PD */ + kib_dev_t *ibh_dev; /* owner */ + atomic_t ibh_ref; /* refcount */ } kib_hca_dev_t; /** # of seconds to keep pool alive */ @@ -222,21 +222,10 @@ typedef struct kib_hca_dev { #define IBLND_POOL_RETRY 1 typedef struct { - int ibp_npages; /* # pages */ - struct page *ibp_pages[0]; /* page array */ + int ibp_npages; /* # pages */ + struct page *ibp_pages[0]; /* page array */ } kib_pages_t; -struct kib_pmr_pool; - -typedef struct { - struct list_head pmr_list; /* chain node */ - struct ib_phys_buf *pmr_ipb; /* physical buffer */ - struct ib_mr *pmr_mr; /* IB MR */ - struct kib_pmr_pool *pmr_pool; /* owner of this MR */ - __u64 pmr_iova; /* Virtual I/O address */ - int pmr_refcount; /* reference count */ -} kib_phys_mr_t; - struct kib_pool; struct kib_poolset; @@ -251,97 +240,87 @@ struct kib_net; #define IBLND_POOL_NAME_LEN 32 typedef struct kib_poolset { - spinlock_t ps_lock; /* serialize */ - struct kib_net *ps_net; /* network it belongs to */ - char ps_name[IBLND_POOL_NAME_LEN]; /* pool set name */ - struct list_head ps_pool_list; /* list of pools */ - struct list_head ps_failed_pool_list; /* failed pool list */ - unsigned long ps_next_retry; /* time stamp for retry if failed to allocate */ - int ps_increasing; /* is allocating new pool */ - int ps_pool_size; /* new pool size */ - int ps_cpt; /* CPT id */ - - kib_ps_pool_create_t ps_pool_create; /* create a new pool */ - kib_ps_pool_destroy_t ps_pool_destroy; /* destroy a pool */ - kib_ps_node_init_t ps_node_init; /* initialize new allocated node */ - kib_ps_node_fini_t ps_node_fini; /* finalize node */ + spinlock_t ps_lock; /* serialize */ + struct kib_net *ps_net; /* network it belongs to */ + char ps_name[IBLND_POOL_NAME_LEN]; /* pool set name */ + struct list_head ps_pool_list; /* list of pools */ + struct list_head ps_failed_pool_list;/* failed pool list */ + unsigned long ps_next_retry; /* time stamp for retry if */ + /* failed to allocate */ + int ps_increasing; /* is allocating new pool */ + int ps_pool_size; /* new pool size */ + int ps_cpt; /* CPT id */ + + kib_ps_pool_create_t ps_pool_create; /* create a new pool */ + kib_ps_pool_destroy_t ps_pool_destroy; /* destroy a pool */ + kib_ps_node_init_t ps_node_init; /* initialize new allocated node */ + kib_ps_node_fini_t ps_node_fini; /* finalize node */ } kib_poolset_t; typedef struct kib_pool { - struct list_head po_list; /* chain on pool list */ - struct list_head po_free_list; /* pre-allocated node */ - kib_poolset_t *po_owner; /* pool_set of this pool */ - unsigned long po_deadline; /* deadline of this pool */ - int po_allocated; /* # of elements in use */ - int po_failed; /* pool is created on failed HCA */ - int po_size; /* # of pre-allocated elements */ + struct list_head po_list; /* chain on pool list */ + struct list_head po_free_list; /* pre-allocated node */ + kib_poolset_t *po_owner; /* pool_set of this pool */ + unsigned long po_deadline; /* deadline of this pool */ + int po_allocated; /* # of elements in use */ + int po_failed; /* pool is created on failed HCA */ + int po_size; /* # of pre-allocated elements */ } kib_pool_t; typedef struct { - kib_poolset_t tps_poolset; /* pool-set */ - __u64 tps_next_tx_cookie; /* cookie of TX */ + kib_poolset_t tps_poolset; /* pool-set */ + __u64 tps_next_tx_cookie; /* cookie of TX */ } kib_tx_poolset_t; typedef struct { - kib_pool_t tpo_pool; /* pool */ - struct kib_hca_dev *tpo_hdev; /* device for this pool */ - struct kib_tx *tpo_tx_descs; /* all the tx descriptors */ - kib_pages_t *tpo_tx_pages; /* premapped tx msg pages */ + kib_pool_t tpo_pool; /* pool */ + struct kib_hca_dev *tpo_hdev; /* device for this pool */ + struct kib_tx *tpo_tx_descs; /* all the tx descriptors */ + kib_pages_t *tpo_tx_pages; /* premapped tx msg pages */ } kib_tx_pool_t; typedef struct { - kib_poolset_t pps_poolset; /* pool-set */ -} kib_pmr_poolset_t; - -typedef struct kib_pmr_pool { - struct kib_hca_dev *ppo_hdev; /* device for this pool */ - kib_pool_t ppo_pool; /* pool */ -} kib_pmr_pool_t; - -typedef struct { - spinlock_t fps_lock; /* serialize */ - struct kib_net *fps_net; /* IB network */ - struct list_head fps_pool_list; /* FMR pool list */ - struct list_head fps_failed_pool_list; /* FMR pool list */ - __u64 fps_version; /* validity stamp */ - int fps_cpt; /* CPT id */ - int fps_pool_size; - int fps_flush_trigger; - /* is allocating new pool */ - int fps_increasing; - /* time stamp for retry if failed to allocate */ - unsigned long fps_next_retry; + spinlock_t fps_lock; /* serialize */ + struct kib_net *fps_net; /* IB network */ + struct list_head fps_pool_list; /* FMR pool list */ + struct list_head fps_failed_pool_list;/* FMR pool list */ + __u64 fps_version; /* validity stamp */ + int fps_cpt; /* CPT id */ + int fps_pool_size; + int fps_flush_trigger; + int fps_increasing; /* is allocating new pool */ + unsigned long fps_next_retry; /* time stamp for retry if*/ + /* failed to allocate */ } kib_fmr_poolset_t; typedef struct { - struct list_head fpo_list; /* chain on pool list */ - struct kib_hca_dev *fpo_hdev; /* device for this pool */ - kib_fmr_poolset_t *fpo_owner; /* owner of this pool */ - struct ib_fmr_pool *fpo_fmr_pool; /* IB FMR pool */ - unsigned long fpo_deadline; /* deadline of this pool */ - int fpo_failed; /* fmr pool is failed */ - int fpo_map_count; /* # of mapped FMR */ + struct list_head fpo_list; /* chain on pool list */ + struct kib_hca_dev *fpo_hdev; /* device for this pool */ + kib_fmr_poolset_t *fpo_owner; /* owner of this pool */ + struct ib_fmr_pool *fpo_fmr_pool; /* IB FMR pool */ + unsigned long fpo_deadline; /* deadline of this pool */ + int fpo_failed; /* fmr pool is failed */ + int fpo_map_count; /* # of mapped FMR */ } kib_fmr_pool_t; typedef struct { - struct ib_pool_fmr *fmr_pfmr; /* IB pool fmr */ - kib_fmr_pool_t *fmr_pool; /* pool of FMR */ + struct ib_pool_fmr *fmr_pfmr; /* IB pool fmr */ + kib_fmr_pool_t *fmr_pool; /* pool of FMR */ } kib_fmr_t; typedef struct kib_net { - struct list_head ibn_list; /* chain on kib_dev_t::ibd_nets */ - __u64 ibn_incarnation; /* my epoch */ - int ibn_init; /* initialisation state */ - int ibn_shutdown; /* shutting down? */ + struct list_head ibn_list; /* chain on kib_dev_t::ibd_nets */ + __u64 ibn_incarnation;/* my epoch */ + int ibn_init; /* initialisation state */ + int ibn_shutdown; /* shutting down? */ - atomic_t ibn_npeers; /* # peers extant */ - atomic_t ibn_nconns; /* # connections extant */ + atomic_t ibn_npeers; /* # peers extant */ + atomic_t ibn_nconns; /* # connections extant */ - kib_tx_poolset_t **ibn_tx_ps; /* tx pool-set */ - kib_fmr_poolset_t **ibn_fmr_ps; /* fmr pool-set */ - kib_pmr_poolset_t **ibn_pmr_ps; /* pmr pool-set */ + kib_tx_poolset_t **ibn_tx_ps; /* tx pool-set */ + kib_fmr_poolset_t **ibn_fmr_ps; /* fmr pool-set */ - kib_dev_t *ibn_dev; /* underlying IB device */ + kib_dev_t *ibn_dev; /* underlying IB device */ } kib_net_t; #define KIB_THREAD_SHIFT 16 @@ -350,51 +329,36 @@ typedef struct kib_net { #define KIB_THREAD_TID(id) ((id) & ((1UL << KIB_THREAD_SHIFT) - 1)) struct kib_sched_info { - /* serialise */ - spinlock_t ibs_lock; - /* schedulers sleep here */ - wait_queue_head_t ibs_waitq; - /* conns to check for rx completions */ - struct list_head ibs_conns; - /* number of scheduler threads */ - int ibs_nthreads; - /* max allowed scheduler threads */ - int ibs_nthreads_max; - int ibs_cpt; /* CPT id */ + spinlock_t ibs_lock; /* serialise */ + wait_queue_head_t ibs_waitq; /* schedulers sleep here */ + struct list_head ibs_conns; /* conns to check for rx completions */ + int ibs_nthreads; /* number of scheduler threads */ + int ibs_nthreads_max; /* max allowed scheduler threads */ + int ibs_cpt; /* CPT id */ }; typedef struct { - int kib_init; /* initialisation state */ - int kib_shutdown; /* shut down? */ - struct list_head kib_devs; /* IB devices extant */ - /* list head of failed devices */ - struct list_head kib_failed_devs; - /* schedulers sleep here */ - wait_queue_head_t kib_failover_waitq; - atomic_t kib_nthreads; /* # live threads */ - /* stabilize net/dev/peer/conn ops */ - rwlock_t kib_global_lock; - /* hash table of all my known peers */ - struct list_head *kib_peers; - /* size of kib_peers */ - int kib_peer_hash_size; - /* the connd task (serialisation assertions) */ - void *kib_connd; - /* connections to setup/teardown */ - struct list_head kib_connd_conns; - /* connections with zero refcount */ - struct list_head kib_connd_zombies; - /* connection daemon sleeps here */ - wait_queue_head_t kib_connd_waitq; - spinlock_t kib_connd_lock; /* serialise */ - struct ib_qp_attr kib_error_qpa; /* QP->ERROR */ - /* percpt data for schedulers */ - struct kib_sched_info **kib_scheds; + int kib_init; /* initialisation state */ + int kib_shutdown; /* shut down? */ + struct list_head kib_devs; /* IB devices extant */ + struct list_head kib_failed_devs; /* list head of failed devices */ + wait_queue_head_t kib_failover_waitq; /* schedulers sleep here */ + atomic_t kib_nthreads; /* # live threads */ + rwlock_t kib_global_lock; /* stabilize net/dev/peer/conn ops */ + struct list_head *kib_peers; /* hash table of all my known peers */ + int kib_peer_hash_size; /* size of kib_peers */ + void *kib_connd; /* the connd task (serialisation assertions) */ + struct list_head kib_connd_conns; /* connections to setup/teardown */ + struct list_head kib_connd_zombies; /* connections with zero refcount */ + wait_queue_head_t kib_connd_waitq; /* connection daemon sleeps here */ + spinlock_t kib_connd_lock; /* serialise */ + struct ib_qp_attr kib_error_qpa; /* QP->ERROR */ + struct kib_sched_info **kib_scheds; /* percpt data for schedulers */ } kib_data_t; -#define IBLND_INIT_NOTHING 0 -#define IBLND_INIT_DATA 1 -#define IBLND_INIT_ALL 2 +#define IBLND_INIT_NOTHING 0 +#define IBLND_INIT_DATA 1 +#define IBLND_INIT_ALL 2 /************************************************************************ * IB Wire message format. @@ -402,242 +366,241 @@ typedef struct { */ typedef struct kib_connparams { - __u16 ibcp_queue_depth; - __u16 ibcp_max_frags; - __u32 ibcp_max_msg_size; + __u16 ibcp_queue_depth; + __u16 ibcp_max_frags; + __u32 ibcp_max_msg_size; } WIRE_ATTR kib_connparams_t; typedef struct { - lnet_hdr_t ibim_hdr; /* portals header */ - char ibim_payload[0]; /* piggy-backed payload */ + lnet_hdr_t ibim_hdr; /* portals header */ + char ibim_payload[0]; /* piggy-backed payload */ } WIRE_ATTR kib_immediate_msg_t; typedef struct { - __u32 rf_nob; /* # bytes this frag */ - __u64 rf_addr; /* CAVEAT EMPTOR: misaligned!! */ + __u32 rf_nob; /* # bytes this frag */ + __u64 rf_addr; /* CAVEAT EMPTOR: misaligned!! */ } WIRE_ATTR kib_rdma_frag_t; typedef struct { - __u32 rd_key; /* local/remote key */ - __u32 rd_nfrags; /* # fragments */ - kib_rdma_frag_t rd_frags[0]; /* buffer frags */ + __u32 rd_key; /* local/remote key */ + __u32 rd_nfrags; /* # fragments */ + kib_rdma_frag_t rd_frags[0]; /* buffer frags */ } WIRE_ATTR kib_rdma_desc_t; typedef struct { - lnet_hdr_t ibprm_hdr; /* portals header */ - __u64 ibprm_cookie; /* opaque completion cookie */ + lnet_hdr_t ibprm_hdr; /* portals header */ + __u64 ibprm_cookie; /* opaque completion cookie */ } WIRE_ATTR kib_putreq_msg_t; typedef struct { - __u64 ibpam_src_cookie; /* reflected completion cookie */ - __u64 ibpam_dst_cookie; /* opaque completion cookie */ - kib_rdma_desc_t ibpam_rd; /* sender's sink buffer */ + __u64 ibpam_src_cookie; /* reflected completion cookie */ + __u64 ibpam_dst_cookie; /* opaque completion cookie */ + kib_rdma_desc_t ibpam_rd; /* sender's sink buffer */ } WIRE_ATTR kib_putack_msg_t; typedef struct { - lnet_hdr_t ibgm_hdr; /* portals header */ - __u64 ibgm_cookie; /* opaque completion cookie */ - kib_rdma_desc_t ibgm_rd; /* rdma descriptor */ + lnet_hdr_t ibgm_hdr; /* portals header */ + __u64 ibgm_cookie; /* opaque completion cookie */ + kib_rdma_desc_t ibgm_rd; /* rdma descriptor */ } WIRE_ATTR kib_get_msg_t; typedef struct { - __u64 ibcm_cookie; /* opaque completion cookie */ - __s32 ibcm_status; /* < 0 failure: >= 0 length */ + __u64 ibcm_cookie; /* opaque completion cookie */ + __s32 ibcm_status; /* < 0 failure: >= 0 length */ } WIRE_ATTR kib_completion_msg_t; typedef struct { /* First 2 fields fixed FOR ALL TIME */ - __u32 ibm_magic; /* I'm an ibnal message */ - __u16 ibm_version; /* this is my version number */ - - __u8 ibm_type; /* msg type */ - __u8 ibm_credits; /* returned credits */ - __u32 ibm_nob; /* # bytes in whole message */ - __u32 ibm_cksum; /* checksum (0 == no checksum) */ - __u64 ibm_srcnid; /* sender's NID */ - __u64 ibm_srcstamp; /* sender's incarnation */ - __u64 ibm_dstnid; /* destination's NID */ - __u64 ibm_dststamp; /* destination's incarnation */ + __u32 ibm_magic; /* I'm an ibnal message */ + __u16 ibm_version; /* this is my version number */ + + __u8 ibm_type; /* msg type */ + __u8 ibm_credits; /* returned credits */ + __u32 ibm_nob; /* # bytes in whole message */ + __u32 ibm_cksum; /* checksum (0 == no checksum) */ + __u64 ibm_srcnid; /* sender's NID */ + __u64 ibm_srcstamp; /* sender's incarnation */ + __u64 ibm_dstnid; /* destination's NID */ + __u64 ibm_dststamp; /* destination's incarnation */ union { - kib_connparams_t connparams; - kib_immediate_msg_t immediate; - kib_putreq_msg_t putreq; - kib_putack_msg_t putack; - kib_get_msg_t get; - kib_completion_msg_t completion; + kib_connparams_t connparams; + kib_immediate_msg_t immediate; + kib_putreq_msg_t putreq; + kib_putack_msg_t putack; + kib_get_msg_t get; + kib_completion_msg_t completion; } WIRE_ATTR ibm_u; } WIRE_ATTR kib_msg_t; -#define IBLND_MSG_MAGIC LNET_PROTO_IB_MAGIC /* unique magic */ +#define IBLND_MSG_MAGIC LNET_PROTO_IB_MAGIC /* unique magic */ -#define IBLND_MSG_VERSION_1 0x11 -#define IBLND_MSG_VERSION_2 0x12 -#define IBLND_MSG_VERSION IBLND_MSG_VERSION_2 +#define IBLND_MSG_VERSION_1 0x11 +#define IBLND_MSG_VERSION_2 0x12 +#define IBLND_MSG_VERSION IBLND_MSG_VERSION_2 -#define IBLND_MSG_CONNREQ 0xc0 /* connection request */ -#define IBLND_MSG_CONNACK 0xc1 /* connection acknowledge */ -#define IBLND_MSG_NOOP 0xd0 /* nothing (just credits) */ -#define IBLND_MSG_IMMEDIATE 0xd1 /* immediate */ -#define IBLND_MSG_PUT_REQ 0xd2 /* putreq (src->sink) */ -#define IBLND_MSG_PUT_NAK 0xd3 /* completion (sink->src) */ -#define IBLND_MSG_PUT_ACK 0xd4 /* putack (sink->src) */ -#define IBLND_MSG_PUT_DONE 0xd5 /* completion (src->sink) */ -#define IBLND_MSG_GET_REQ 0xd6 /* getreq (sink->src) */ -#define IBLND_MSG_GET_DONE 0xd7 /* completion (src->sink: all OK) */ +#define IBLND_MSG_CONNREQ 0xc0 /* connection request */ +#define IBLND_MSG_CONNACK 0xc1 /* connection acknowledge */ +#define IBLND_MSG_NOOP 0xd0 /* nothing (just credits) */ +#define IBLND_MSG_IMMEDIATE 0xd1 /* immediate */ +#define IBLND_MSG_PUT_REQ 0xd2 /* putreq (src->sink) */ +#define IBLND_MSG_PUT_NAK 0xd3 /* completion (sink->src) */ +#define IBLND_MSG_PUT_ACK 0xd4 /* putack (sink->src) */ +#define IBLND_MSG_PUT_DONE 0xd5 /* completion (src->sink) */ +#define IBLND_MSG_GET_REQ 0xd6 /* getreq (sink->src) */ +#define IBLND_MSG_GET_DONE 0xd7 /* completion (src->sink: all OK) */ typedef struct { - __u32 ibr_magic; /* sender's magic */ - __u16 ibr_version; /* sender's version */ - __u8 ibr_why; /* reject reason */ - __u8 ibr_padding; /* padding */ - __u64 ibr_incarnation; /* incarnation of peer */ - kib_connparams_t ibr_cp; /* connection parameters */ + __u32 ibr_magic; /* sender's magic */ + __u16 ibr_version; /* sender's version */ + __u8 ibr_why; /* reject reason */ + __u8 ibr_padding; /* padding */ + __u64 ibr_incarnation; /* incarnation of peer */ + kib_connparams_t ibr_cp; /* connection parameters */ } WIRE_ATTR kib_rej_t; /* connection rejection reasons */ -#define IBLND_REJECT_CONN_RACE 1 /* You lost connection race */ -#define IBLND_REJECT_NO_RESOURCES 2 /* Out of memory/conns etc */ -#define IBLND_REJECT_FATAL 3 /* Anything else */ - -#define IBLND_REJECT_CONN_UNCOMPAT 4 /* incompatible version peer */ -#define IBLND_REJECT_CONN_STALE 5 /* stale peer */ - -#define IBLND_REJECT_RDMA_FRAGS 6 /* Fatal: peer's rdma frags can't match mine */ -#define IBLND_REJECT_MSG_QUEUE_SIZE 7 /* Fatal: peer's msg queue size can't match mine */ +#define IBLND_REJECT_CONN_RACE 1 /* You lost connection race */ +#define IBLND_REJECT_NO_RESOURCES 2 /* Out of memory/conns etc */ +#define IBLND_REJECT_FATAL 3 /* Anything else */ +#define IBLND_REJECT_CONN_UNCOMPAT 4 /* incompatible version peer */ +#define IBLND_REJECT_CONN_STALE 5 /* stale peer */ +#define IBLND_REJECT_RDMA_FRAGS 6 /* Fatal: peer's rdma frags can't match */ + /* mine */ +#define IBLND_REJECT_MSG_QUEUE_SIZE 7 /* Fatal: peer's msg queue size can't */ + /* match mine */ /***********************************************************************/ -typedef struct kib_rx /* receive message */ +typedef struct kib_rx /* receive message */ { - struct list_head rx_list; /* queue for attention */ - struct kib_conn *rx_conn; /* owning conn */ - int rx_nob; /* # bytes received (-1 while posted) */ - enum ib_wc_status rx_status; /* completion status */ - kib_msg_t *rx_msg; /* message buffer (host vaddr) */ - __u64 rx_msgaddr; /* message buffer (I/O addr) */ - DECLARE_PCI_UNMAP_ADDR (rx_msgunmap); /* for dma_unmap_single() */ - struct ib_recv_wr rx_wrq; /* receive work item... */ - struct ib_sge rx_sge; /* ...and its memory */ + struct list_head rx_list; /* queue for attention */ + struct kib_conn *rx_conn; /* owning conn */ + int rx_nob; /* # bytes received (-1 while posted) */ + enum ib_wc_status rx_status; /* completion status */ + kib_msg_t *rx_msg; /* message buffer (host vaddr) */ + __u64 rx_msgaddr; /* message buffer (I/O addr) */ + DECLARE_PCI_UNMAP_ADDR(rx_msgunmap); /* for dma_unmap_single() */ + struct ib_recv_wr rx_wrq; /* receive work item... */ + struct ib_sge rx_sge; /* ...and its memory */ } kib_rx_t; -#define IBLND_POSTRX_DONT_POST 0 /* don't post */ -#define IBLND_POSTRX_NO_CREDIT 1 /* post: no credits */ -#define IBLND_POSTRX_PEER_CREDIT 2 /* post: give peer back 1 credit */ -#define IBLND_POSTRX_RSRVD_CREDIT 3 /* post: give myself back 1 reserved credit */ +#define IBLND_POSTRX_DONT_POST 0 /* don't post */ +#define IBLND_POSTRX_NO_CREDIT 1 /* post: no credits */ +#define IBLND_POSTRX_PEER_CREDIT 2 /* post: give peer back 1 credit */ +#define IBLND_POSTRX_RSRVD_CREDIT 3 /* post: give self back 1 reserved credit */ -typedef struct kib_tx /* transmit message */ +typedef struct kib_tx /* transmit message */ { - struct list_head tx_list; /* queue on idle_txs ibc_tx_queue etc. */ - kib_tx_pool_t *tx_pool; /* pool I'm from */ - struct kib_conn *tx_conn; /* owning conn */ - short tx_sending; /* # tx callbacks outstanding */ - short tx_queued; /* queued for sending */ - short tx_waiting; /* waiting for peer */ - int tx_status; /* LNET completion status */ - unsigned long tx_deadline; /* completion deadline */ - __u64 tx_cookie; /* completion cookie */ - lnet_msg_t *tx_lntmsg[2]; /* lnet msgs to finalize on completion */ - kib_msg_t *tx_msg; /* message buffer (host vaddr) */ - __u64 tx_msgaddr; /* message buffer (I/O addr) */ - DECLARE_PCI_UNMAP_ADDR (tx_msgunmap); /* for dma_unmap_single() */ - int tx_nwrq; /* # send work items */ - struct ib_send_wr *tx_wrq; /* send work items... */ - struct ib_sge *tx_sge; /* ...and their memory */ - kib_rdma_desc_t *tx_rd; /* rdma descriptor */ - int tx_nfrags; /* # entries in... */ - struct scatterlist *tx_frags; /* dma_map_sg descriptor */ - __u64 *tx_pages; /* rdma phys page addrs */ - union { - kib_phys_mr_t *pmr; /* MR for physical buffer */ - kib_fmr_t fmr; /* FMR */ - } tx_u; - int tx_dmadir; /* dma direction */ + struct list_head tx_list; /* queue on idle_txs ibc_tx_queue etc. */ + kib_tx_pool_t *tx_pool; /* pool I'm from */ + struct kib_conn *tx_conn; /* owning conn */ + short tx_sending; /* # tx callbacks outstanding */ + short tx_queued; /* queued for sending */ + short tx_waiting; /* waiting for peer */ + int tx_status; /* LNET completion status */ + unsigned long tx_deadline; /* completion deadline */ + __u64 tx_cookie; /* completion cookie */ + lnet_msg_t *tx_lntmsg[2]; /* lnet msgs to finalize on completion */ + kib_msg_t *tx_msg; /* message buffer (host vaddr) */ + __u64 tx_msgaddr; /* message buffer (I/O addr) */ + DECLARE_PCI_UNMAP_ADDR(tx_msgunmap); /* for dma_unmap_single() */ + int tx_nwrq; /* # send work items */ + struct ib_rdma_wr *tx_wrq; /* send work items... */ + struct ib_sge *tx_sge; /* ...and their memory */ + kib_rdma_desc_t *tx_rd; /* rdma descriptor */ + int tx_nfrags; /* # entries in... */ + struct scatterlist *tx_frags; /* dma_map_sg descriptor */ + __u64 *tx_pages; /* rdma phys page addrs */ + kib_fmr_t fmr; /* FMR */ + int tx_dmadir; /* dma direction */ } kib_tx_t; typedef struct kib_connvars { - /* connection-in-progress variables */ - kib_msg_t cv_msg; + kib_msg_t cv_msg; /* connection-in-progress variables */ } kib_connvars_t; typedef struct kib_conn { - struct kib_sched_info *ibc_sched; /* scheduler information */ - struct kib_peer *ibc_peer; /* owning peer */ - kib_hca_dev_t *ibc_hdev; /* HCA bound on */ - struct list_head ibc_list; /* stash on peer's conn list */ - struct list_head ibc_sched_list; /* schedule for attention */ - __u16 ibc_version; /* version of connection */ - __u64 ibc_incarnation; /* which instance of the peer */ - atomic_t ibc_refcount; /* # users */ - int ibc_state; /* what's happening */ - int ibc_nsends_posted; /* # uncompleted sends */ - int ibc_noops_posted; /* # uncompleted NOOPs */ - int ibc_credits; /* # credits I have */ - int ibc_outstanding_credits; /* # credits to return */ - int ibc_reserved_credits;/* # ACK/DONE msg credits */ - int ibc_comms_error; /* set on comms error */ - unsigned int ibc_nrx:16; /* receive buffers owned */ - unsigned int ibc_scheduled:1; /* scheduled for attention */ - unsigned int ibc_ready:1; /* CQ callback fired */ - /* time of last send */ - unsigned long ibc_last_send; - /** link chain for kiblnd_check_conns only */ - struct list_head ibc_connd_list; - /** rxs completed before ESTABLISHED */ - struct list_head ibc_early_rxs; - /** IBLND_MSG_NOOPs for IBLND_MSG_VERSION_1 */ - struct list_head ibc_tx_noops; - struct list_head ibc_tx_queue; /* sends that need a credit */ - struct list_head ibc_tx_queue_nocred;/* sends that don't need a credit */ - struct list_head ibc_tx_queue_rsrvd; /* sends that need to reserve an ACK/DONE msg */ - struct list_head ibc_active_txs; /* active tx awaiting completion */ - spinlock_t ibc_lock; /* serialise */ - kib_rx_t *ibc_rxs; /* the rx descs */ - kib_pages_t *ibc_rx_pages; /* premapped rx msg pages */ - - struct rdma_cm_id *ibc_cmid; /* CM id */ - struct ib_cq *ibc_cq; /* completion queue */ - - kib_connvars_t *ibc_connvars; /* in-progress connection state */ + struct kib_sched_info *ibc_sched; /* scheduler information */ + struct kib_peer *ibc_peer; /* owning peer */ + kib_hca_dev_t *ibc_hdev; /* HCA bound on */ + struct list_head ibc_list; /* stash on peer's conn list */ + struct list_head ibc_sched_list; /* schedule for attention */ + __u16 ibc_version; /* version of connection */ + __u64 ibc_incarnation; /* which instance of the peer */ + atomic_t ibc_refcount; /* # users */ + int ibc_state; /* what's happening */ + int ibc_nsends_posted; /* # uncompleted sends */ + int ibc_noops_posted; /* # uncompleted NOOPs */ + int ibc_credits; /* # credits I have */ + int ibc_outstanding_credits; /* # credits to return */ + int ibc_reserved_credits; /* # ACK/DONE msg credits */ + int ibc_comms_error; /* set on comms error */ + unsigned int ibc_nrx:16; /* receive buffers owned */ + unsigned int ibc_scheduled:1; /* scheduled for attention */ + unsigned int ibc_ready:1; /* CQ callback fired */ + unsigned long ibc_last_send; /* time of last send */ + struct list_head ibc_connd_list; /* link chain for */ + /* kiblnd_check_conns only */ + struct list_head ibc_early_rxs; /* rxs completed before ESTABLISHED */ + struct list_head ibc_tx_noops; /* IBLND_MSG_NOOPs for */ + /* IBLND_MSG_VERSION_1 */ + struct list_head ibc_tx_queue; /* sends that need a credit */ + struct list_head ibc_tx_queue_nocred; /* sends that don't need a */ + /* credit */ + struct list_head ibc_tx_queue_rsrvd; /* sends that need to */ + /* reserve an ACK/DONE msg */ + struct list_head ibc_active_txs; /* active tx awaiting completion */ + spinlock_t ibc_lock; /* serialise */ + kib_rx_t *ibc_rxs; /* the rx descs */ + kib_pages_t *ibc_rx_pages; /* premapped rx msg pages */ + + struct rdma_cm_id *ibc_cmid; /* CM id */ + struct ib_cq *ibc_cq; /* completion queue */ + + kib_connvars_t *ibc_connvars; /* in-progress connection state */ } kib_conn_t; -#define IBLND_CONN_INIT 0 /* being initialised */ -#define IBLND_CONN_ACTIVE_CONNECT 1 /* active sending req */ -#define IBLND_CONN_PASSIVE_WAIT 2 /* passive waiting for rtu */ -#define IBLND_CONN_ESTABLISHED 3 /* connection established */ -#define IBLND_CONN_CLOSING 4 /* being closed */ -#define IBLND_CONN_DISCONNECTED 5 /* disconnected */ +#define IBLND_CONN_INIT 0 /* being initialised */ +#define IBLND_CONN_ACTIVE_CONNECT 1 /* active sending req */ +#define IBLND_CONN_PASSIVE_WAIT 2 /* passive waiting for rtu */ +#define IBLND_CONN_ESTABLISHED 3 /* connection established */ +#define IBLND_CONN_CLOSING 4 /* being closed */ +#define IBLND_CONN_DISCONNECTED 5 /* disconnected */ typedef struct kib_peer { - struct list_head ibp_list; /* stash on global peer list */ - lnet_nid_t ibp_nid; /* who's on the other end(s) */ - lnet_ni_t *ibp_ni; /* LNet interface */ - atomic_t ibp_refcount; /* # users */ - struct list_head ibp_conns; /* all active connections */ - struct list_head ibp_tx_queue; /* msgs waiting for a conn */ - __u16 ibp_version; /* version of peer */ - __u64 ibp_incarnation; /* incarnation of peer */ - int ibp_connecting; /* current active connection attempts */ - int ibp_accepting; /* current passive connection attempts */ - int ibp_error; /* errno on closing this peer */ - unsigned long ibp_last_alive; /* when (in jiffies) I was last alive */ + struct list_head ibp_list; /* stash on global peer list */ + lnet_nid_t ibp_nid; /* who's on the other end(s) */ + lnet_ni_t *ibp_ni; /* LNet interface */ + atomic_t ibp_refcount; /* # users */ + struct list_head ibp_conns; /* all active connections */ + struct list_head ibp_tx_queue; /* msgs waiting for a conn */ + __u16 ibp_version; /* version of peer */ + __u64 ibp_incarnation; /* incarnation of peer */ + int ibp_connecting; /* current active connection attempts + */ + int ibp_accepting; /* current passive connection attempts + */ + int ibp_error; /* errno on closing this peer */ + unsigned long ibp_last_alive; /* when (in jiffies) I was last alive + */ } kib_peer_t; -extern kib_data_t kiblnd_data; +extern kib_data_t kiblnd_data; -extern void kiblnd_hdev_destroy(kib_hca_dev_t *hdev); +void kiblnd_hdev_destroy(kib_hca_dev_t *hdev); static inline void kiblnd_hdev_addref_locked(kib_hca_dev_t *hdev) { - LASSERT (atomic_read(&hdev->ibh_ref) > 0); + LASSERT(atomic_read(&hdev->ibh_ref) > 0); atomic_inc(&hdev->ibh_ref); } static inline void kiblnd_hdev_decref(kib_hca_dev_t *hdev) { - LASSERT (atomic_read(&hdev->ibh_ref) > 0); + LASSERT(atomic_read(&hdev->ibh_ref) > 0); if (atomic_dec_and_test(&hdev->ibh_ref)) kiblnd_hdev_destroy(hdev); } @@ -684,7 +647,7 @@ do { \ do { \ CDEBUG(D_NET, "peer[%p] -> %s (%d)++\n", \ (peer), libcfs_nid2str((peer)->ibp_nid), \ - atomic_read (&(peer)->ibp_refcount)); \ + atomic_read(&(peer)->ibp_refcount)); \ atomic_inc(&(peer)->ibp_refcount); \ } while (0) @@ -692,32 +655,32 @@ do { \ do { \ CDEBUG(D_NET, "peer[%p] -> %s (%d)--\n", \ (peer), libcfs_nid2str((peer)->ibp_nid), \ - atomic_read (&(peer)->ibp_refcount)); \ + atomic_read(&(peer)->ibp_refcount)); \ LASSERT_ATOMIC_POS(&(peer)->ibp_refcount); \ if (atomic_dec_and_test(&(peer)->ibp_refcount)) \ kiblnd_destroy_peer(peer); \ } while (0) static inline struct list_head * -kiblnd_nid2peerlist (lnet_nid_t nid) +kiblnd_nid2peerlist(lnet_nid_t nid) { unsigned int hash = ((unsigned int)nid) % kiblnd_data.kib_peer_hash_size; - return (&kiblnd_data.kib_peers [hash]); + return &kiblnd_data.kib_peers[hash]; } static inline int -kiblnd_peer_active (kib_peer_t *peer) +kiblnd_peer_active(kib_peer_t *peer) { /* Am I in the peer hash table? */ - return (!list_empty(&peer->ibp_list)); + return !list_empty(&peer->ibp_list); } static inline kib_conn_t * -kiblnd_get_conn_locked (kib_peer_t *peer) +kiblnd_get_conn_locked(kib_peer_t *peer) { - LASSERT (!list_empty(&peer->ibp_conns)); + LASSERT(!list_empty(&peer->ibp_conns)); /* just return the first connection */ return list_entry(peer->ibp_conns.next, kib_conn_t, ibc_list); @@ -734,7 +697,7 @@ kiblnd_send_keepalive(kib_conn_t *conn) static inline int kiblnd_need_noop(kib_conn_t *conn) { - LASSERT (conn->ibc_state >= IBLND_CONN_ESTABLISHED); + LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED); if (conn->ibc_outstanding_credits < IBLND_CREDITS_HIGHWATER(conn->ibc_version) && @@ -771,7 +734,7 @@ kiblnd_abort_receives(kib_conn_t *conn) } static inline const char * -kiblnd_queue2str (kib_conn_t *conn, struct list_head *q) +kiblnd_queue2str(kib_conn_t *conn, struct list_head *q) { if (q == &conn->ibc_tx_queue) return "tx_queue"; @@ -789,8 +752,8 @@ kiblnd_queue2str (kib_conn_t *conn, struct list_head *q) return NULL; } -/* CAVEAT EMPTOR: We rely on descriptor alignment to allow us to use the - * lowest bits of the work request id to stash the work item type. */ +/* CAVEAT EMPTOR: We rely on descriptor alignment to allow us to use the */ +/* lowest bits of the work request id to stash the work item type. */ #define IBLND_WID_TX 0 #define IBLND_WID_RDMA 1 @@ -798,43 +761,43 @@ kiblnd_queue2str (kib_conn_t *conn, struct list_head *q) #define IBLND_WID_MASK 3UL static inline __u64 -kiblnd_ptr2wreqid (void *ptr, int type) +kiblnd_ptr2wreqid(void *ptr, int type) { unsigned long lptr = (unsigned long)ptr; - LASSERT ((lptr & IBLND_WID_MASK) == 0); - LASSERT ((type & ~IBLND_WID_MASK) == 0); + LASSERT((lptr & IBLND_WID_MASK) == 0); + LASSERT((type & ~IBLND_WID_MASK) == 0); return (__u64)(lptr | type); } static inline void * -kiblnd_wreqid2ptr (__u64 wreqid) +kiblnd_wreqid2ptr(__u64 wreqid) { return (void *)(((unsigned long)wreqid) & ~IBLND_WID_MASK); } static inline int -kiblnd_wreqid2type (__u64 wreqid) +kiblnd_wreqid2type(__u64 wreqid) { - return (wreqid & IBLND_WID_MASK); + return wreqid & IBLND_WID_MASK; } static inline void -kiblnd_set_conn_state (kib_conn_t *conn, int state) +kiblnd_set_conn_state(kib_conn_t *conn, int state) { conn->ibc_state = state; mb(); } static inline void -kiblnd_init_msg (kib_msg_t *msg, int type, int body_nob) +kiblnd_init_msg(kib_msg_t *msg, int type, int body_nob) { msg->ibm_type = type; msg->ibm_nob = offsetof(kib_msg_t, ibm_u) + body_nob; } static inline int -kiblnd_rd_size (kib_rdma_desc_t *rd) +kiblnd_rd_size(kib_rdma_desc_t *rd) { int i; int size; @@ -870,7 +833,7 @@ kiblnd_rd_consume_frag(kib_rdma_desc_t *rd, int index, __u32 nob) rd->rd_frags[index].rf_addr += nob; rd->rd_frags[index].rf_nob -= nob; } else { - index ++; + index++; } return index; @@ -879,15 +842,14 @@ kiblnd_rd_consume_frag(kib_rdma_desc_t *rd, int index, __u32 nob) static inline int kiblnd_rd_msg_size(kib_rdma_desc_t *rd, int msgtype, int n) { - LASSERT (msgtype == IBLND_MSG_GET_REQ || - msgtype == IBLND_MSG_PUT_ACK); + LASSERT(msgtype == IBLND_MSG_GET_REQ || + msgtype == IBLND_MSG_PUT_ACK); return msgtype == IBLND_MSG_GET_REQ ? offsetof(kib_get_msg_t, ibgm_rd.rd_frags[n]) : offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[n]); } - static inline __u64 kiblnd_dma_mapping_error(struct ib_device *dev, u64 dma_addr) { @@ -937,13 +899,12 @@ static inline unsigned int kiblnd_sg_dma_len(struct ib_device *dev, return ib_sg_dma_len(dev, sg); } -/* XXX We use KIBLND_CONN_PARAM(e) as writable buffer, it's not strictly - * right because OFED1.2 defines it as const, to use it we have to add - * (void *) cast to overcome "const" */ - -#define KIBLND_CONN_PARAM(e) ((e)->param.conn.private_data) -#define KIBLND_CONN_PARAM_LEN(e) ((e)->param.conn.private_data_len) +/* XXX We use KIBLND_CONN_PARAM(e) as writable buffer, it's not strictly */ +/* right because OFED1.2 defines it as const, to use it we have to add */ +/* (void *) cast to overcome "const" */ +#define KIBLND_CONN_PARAM(e) ((e)->param.conn.private_data) +#define KIBLND_CONN_PARAM_LEN(e) ((e)->param.conn.private_data_len) struct ib_mr *kiblnd_find_rd_dma_mr(kib_hca_dev_t *hdev, kib_rdma_desc_t *rd); @@ -951,9 +912,6 @@ struct ib_mr *kiblnd_find_dma_mr(kib_hca_dev_t *hdev, __u64 addr, __u64 size); void kiblnd_map_rx_descs(kib_conn_t *conn); void kiblnd_unmap_rx_descs(kib_conn_t *conn); -int kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, - kib_rdma_desc_t *rd, int nfrags); -void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx); void kiblnd_pool_free_node(kib_pool_t *pool, struct list_head *node); struct list_head *kiblnd_pool_alloc_node(kib_poolset_t *ps); @@ -961,57 +919,53 @@ int kiblnd_fmr_pool_map(kib_fmr_poolset_t *fps, __u64 *pages, int npages, __u64 iov, kib_fmr_t *fmr); void kiblnd_fmr_pool_unmap(kib_fmr_t *fmr, int status); -int kiblnd_pmr_pool_map(kib_pmr_poolset_t *pps, kib_hca_dev_t *hdev, - kib_rdma_desc_t *rd, __u64 *iova, kib_phys_mr_t **pp_pmr); -void kiblnd_pmr_pool_unmap(kib_phys_mr_t *pmr); - -int kiblnd_startup (lnet_ni_t *ni); -void kiblnd_shutdown (lnet_ni_t *ni); -int kiblnd_ctl (lnet_ni_t *ni, unsigned int cmd, void *arg); -void kiblnd_query (struct lnet_ni *ni, lnet_nid_t nid, unsigned long *when); +int kiblnd_startup(lnet_ni_t *ni); +void kiblnd_shutdown(lnet_ni_t *ni); +int kiblnd_ctl(lnet_ni_t *ni, unsigned int cmd, void *arg); +void kiblnd_query(struct lnet_ni *ni, lnet_nid_t nid, unsigned long *when); int kiblnd_tunables_init(void); void kiblnd_tunables_fini(void); -int kiblnd_connd (void *arg); +int kiblnd_connd(void *arg); int kiblnd_scheduler(void *arg); int kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name); -int kiblnd_failover_thread (void *arg); +int kiblnd_failover_thread(void *arg); int kiblnd_alloc_pages(kib_pages_t **pp, int cpt, int npages); -void kiblnd_free_pages (kib_pages_t *p); +void kiblnd_free_pages(kib_pages_t *p); int kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event); int kiblnd_translate_mtu(int value); int kiblnd_dev_failover(kib_dev_t *dev); -int kiblnd_create_peer (lnet_ni_t *ni, kib_peer_t **peerp, lnet_nid_t nid); -void kiblnd_destroy_peer (kib_peer_t *peer); -void kiblnd_destroy_dev (kib_dev_t *dev); -void kiblnd_unlink_peer_locked (kib_peer_t *peer); -void kiblnd_peer_alive (kib_peer_t *peer); -kib_peer_t *kiblnd_find_peer_locked (lnet_nid_t nid); -void kiblnd_peer_connect_failed (kib_peer_t *peer, int active, int error); -int kiblnd_close_stale_conns_locked (kib_peer_t *peer, +int kiblnd_create_peer(lnet_ni_t *ni, kib_peer_t **peerp, lnet_nid_t nid); +void kiblnd_destroy_peer(kib_peer_t *peer); +void kiblnd_destroy_dev(kib_dev_t *dev); +void kiblnd_unlink_peer_locked(kib_peer_t *peer); +void kiblnd_peer_alive(kib_peer_t *peer); +kib_peer_t *kiblnd_find_peer_locked(lnet_nid_t nid); +void kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error); +int kiblnd_close_stale_conns_locked(kib_peer_t *peer, int version, __u64 incarnation); -int kiblnd_close_peer_conns_locked (kib_peer_t *peer, int why); +int kiblnd_close_peer_conns_locked(kib_peer_t *peer, int why); void kiblnd_connreq_done(kib_conn_t *conn, int status); -kib_conn_t *kiblnd_create_conn (kib_peer_t *peer, struct rdma_cm_id *cmid, +kib_conn_t *kiblnd_create_conn(kib_peer_t *peer, struct rdma_cm_id *cmid, int state, int version); -void kiblnd_destroy_conn (kib_conn_t *conn); -void kiblnd_close_conn (kib_conn_t *conn, int error); -void kiblnd_close_conn_locked (kib_conn_t *conn, int error); +void kiblnd_destroy_conn(kib_conn_t *conn); +void kiblnd_close_conn(kib_conn_t *conn, int error); +void kiblnd_close_conn_locked(kib_conn_t *conn, int error); -int kiblnd_init_rdma (kib_conn_t *conn, kib_tx_t *tx, int type, +int kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type, int nob, kib_rdma_desc_t *dstrd, __u64 dstcookie); -void kiblnd_launch_tx (lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid); -void kiblnd_queue_tx_locked (kib_tx_t *tx, kib_conn_t *conn); -void kiblnd_queue_tx (kib_tx_t *tx, kib_conn_t *conn); -void kiblnd_init_tx_msg (lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob); -void kiblnd_txlist_done (lnet_ni_t *ni, struct list_head *txlist, +void kiblnd_launch_tx(lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid); +void kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn); +void kiblnd_queue_tx(kib_tx_t *tx, kib_conn_t *conn); +void kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob); +void kiblnd_txlist_done(lnet_ni_t *ni, struct list_head *txlist, int status); void kiblnd_check_sends (kib_conn_t *conn); @@ -1019,10 +973,10 @@ void kiblnd_qp_event(struct ib_event *event, void *arg); void kiblnd_cq_event(struct ib_event *event, void *arg); void kiblnd_cq_completion(struct ib_cq *cq, void *arg); -void kiblnd_pack_msg (lnet_ni_t *ni, kib_msg_t *msg, int version, +void kiblnd_pack_msg(lnet_ni_t *ni, kib_msg_t *msg, int version, int credits, lnet_nid_t dstnid, __u64 dststamp); int kiblnd_unpack_msg(kib_msg_t *msg, int nob); -int kiblnd_post_rx (kib_rx_t *rx, int credit); +int kiblnd_post_rx(kib_rx_t *rx, int credit); int kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg); int kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, diff --git a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c index dbf374983..260750354 100644 --- a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c +++ b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c @@ -40,13 +40,15 @@ #include "o2iblnd.h" +static void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx); + static void kiblnd_tx_done(lnet_ni_t *ni, kib_tx_t *tx) { lnet_msg_t *lntmsg[2]; - kib_net_t *net = ni->ni_data; - int rc; - int i; + kib_net_t *net = ni->ni_data; + int rc; + int i; LASSERT(net != NULL); LASSERT(!in_interrupt()); @@ -102,10 +104,10 @@ kiblnd_txlist_done(lnet_ni_t *ni, struct list_head *txlist, int status) static kib_tx_t * kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target) { - kib_net_t *net = (kib_net_t *)ni->ni_data; - struct list_head *node; - kib_tx_t *tx; - kib_tx_poolset_t *tps; + kib_net_t *net = (kib_net_t *)ni->ni_data; + struct list_head *node; + kib_tx_t *tx; + kib_tx_poolset_t *tps; tps = net->ibn_tx_ps[lnet_cpt_of_nid(target)]; node = kiblnd_pool_alloc_node(&tps->tps_poolset); @@ -121,7 +123,6 @@ kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target) LASSERT(tx->tx_conn == NULL); LASSERT(tx->tx_lntmsg[0] == NULL); LASSERT(tx->tx_lntmsg[1] == NULL); - LASSERT(tx->tx_u.pmr == NULL); LASSERT(tx->tx_nfrags == 0); return tx; @@ -130,9 +131,9 @@ kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target) static void kiblnd_drop_rx(kib_rx_t *rx) { - kib_conn_t *conn = rx->rx_conn; - struct kib_sched_info *sched = conn->ibc_sched; - unsigned long flags; + kib_conn_t *conn = rx->rx_conn; + struct kib_sched_info *sched = conn->ibc_sched; + unsigned long flags; spin_lock_irqsave(&sched->ibs_lock, flags); LASSERT(conn->ibc_nrx > 0); @@ -145,11 +146,11 @@ kiblnd_drop_rx(kib_rx_t *rx) int kiblnd_post_rx(kib_rx_t *rx, int credit) { - kib_conn_t *conn = rx->rx_conn; - kib_net_t *net = conn->ibc_peer->ibp_ni->ni_data; - struct ib_recv_wr *bad_wrq = NULL; - struct ib_mr *mr; - int rc; + kib_conn_t *conn = rx->rx_conn; + kib_net_t *net = conn->ibc_peer->ibp_ni->ni_data; + struct ib_recv_wr *bad_wrq = NULL; + struct ib_mr *mr; + int rc; LASSERT(net != NULL); LASSERT(!in_interrupt()); @@ -164,10 +165,10 @@ kiblnd_post_rx(kib_rx_t *rx, int credit) rx->rx_sge.addr = rx->rx_msgaddr; rx->rx_sge.length = IBLND_MSG_SIZE; - rx->rx_wrq.next = NULL; + rx->rx_wrq.next = NULL; rx->rx_wrq.sg_list = &rx->rx_sge; rx->rx_wrq.num_sge = 1; - rx->rx_wrq.wr_id = kiblnd_ptr2wreqid(rx, IBLND_WID_RX); + rx->rx_wrq.wr_id = kiblnd_ptr2wreqid(rx, IBLND_WID_RX); LASSERT(conn->ibc_state >= IBLND_CONN_INIT); LASSERT(rx->rx_nob >= 0); /* not posted */ @@ -179,24 +180,28 @@ kiblnd_post_rx(kib_rx_t *rx, int credit) rx->rx_nob = -1; /* flag posted */ + /* NB: need an extra reference after ib_post_recv because we don't + * own this rx (and rx::rx_conn) anymore, LU-5678. + */ + kiblnd_conn_addref(conn); rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq); - if (rc != 0) { + if (unlikely(rc != 0)) { CERROR("Can't post rx for %s: %d, bad_wrq: %p\n", libcfs_nid2str(conn->ibc_peer->ibp_nid), rc, bad_wrq); rx->rx_nob = 0; } if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */ - return rc; + goto out; - if (rc != 0) { + if (unlikely(rc != 0)) { kiblnd_close_conn(conn, rc); kiblnd_drop_rx(rx); /* No more posts for this rx */ - return rc; + goto out; } if (credit == IBLND_POSTRX_NO_CREDIT) - return 0; + goto out; spin_lock(&conn->ibc_lock); if (credit == IBLND_POSTRX_PEER_CREDIT) @@ -206,13 +211,15 @@ kiblnd_post_rx(kib_rx_t *rx, int credit) spin_unlock(&conn->ibc_lock); kiblnd_check_sends(conn); - return 0; +out: + kiblnd_conn_decref(conn); + return rc; } static kib_tx_t * kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie) { - struct list_head *tmp; + struct list_head *tmp; list_for_each(tmp, &conn->ibc_active_txs) { kib_tx_t *tx = list_entry(tmp, kib_tx_t, tx_list); @@ -237,9 +244,9 @@ kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie) static void kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie) { - kib_tx_t *tx; - lnet_ni_t *ni = conn->ibc_peer->ibp_ni; - int idle; + kib_tx_t *tx; + lnet_ni_t *ni = conn->ibc_peer->ibp_ni; + int idle; spin_lock(&conn->ibc_lock); @@ -254,11 +261,10 @@ kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie) } if (tx->tx_status == 0) { /* success so far */ - if (status < 0) { /* failed? */ + if (status < 0) /* failed? */ tx->tx_status = status; - } else if (txtype == IBLND_MSG_GET_REQ) { + else if (txtype == IBLND_MSG_GET_REQ) lnet_set_reply_msg_len(ni, tx->tx_lntmsg[1], status); - } } tx->tx_waiting = 0; @@ -276,8 +282,8 @@ kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie) static void kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie) { - lnet_ni_t *ni = conn->ibc_peer->ibp_ni; - kib_tx_t *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid); + lnet_ni_t *ni = conn->ibc_peer->ibp_ni; + kib_tx_t *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid); if (tx == NULL) { CERROR("Can't get tx for completion %x for %s\n", @@ -295,14 +301,14 @@ kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie) static void kiblnd_handle_rx(kib_rx_t *rx) { - kib_msg_t *msg = rx->rx_msg; - kib_conn_t *conn = rx->rx_conn; - lnet_ni_t *ni = conn->ibc_peer->ibp_ni; - int credits = msg->ibm_credits; - kib_tx_t *tx; - int rc = 0; - int rc2; - int post_credit; + kib_msg_t *msg = rx->rx_msg; + kib_conn_t *conn = rx->rx_conn; + lnet_ni_t *ni = conn->ibc_peer->ibp_ni; + int credits = msg->ibm_credits; + kib_tx_t *tx; + int rc = 0; + int rc2; + int post_credit; LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED); @@ -456,12 +462,12 @@ kiblnd_handle_rx(kib_rx_t *rx) static void kiblnd_rx_complete(kib_rx_t *rx, int status, int nob) { - kib_msg_t *msg = rx->rx_msg; - kib_conn_t *conn = rx->rx_conn; - lnet_ni_t *ni = conn->ibc_peer->ibp_ni; - kib_net_t *net = ni->ni_data; - int rc; - int err = -EIO; + kib_msg_t *msg = rx->rx_msg; + kib_conn_t *conn = rx->rx_conn; + lnet_ni_t *ni = conn->ibc_peer->ibp_ni; + kib_net_t *net = ni->ni_data; + int rc; + int err = -EIO; LASSERT(net != NULL); LASSERT(rx->rx_nob < 0); /* was posted */ @@ -502,8 +508,8 @@ kiblnd_rx_complete(kib_rx_t *rx, int status, int nob) /* racing with connection establishment/teardown! */ if (conn->ibc_state < IBLND_CONN_ESTABLISHED) { - rwlock_t *g_lock = &kiblnd_data.kib_global_lock; - unsigned long flags; + rwlock_t *g_lock = &kiblnd_data.kib_global_lock; + unsigned long flags; write_lock_irqsave(g_lock, flags); /* must check holding global lock to eliminate race */ @@ -550,19 +556,19 @@ kiblnd_kvaddr_to_page(unsigned long vaddr) static int kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) { - kib_hca_dev_t *hdev; - __u64 *pages = tx->tx_pages; - kib_fmr_poolset_t *fps; - int npages; - int size; - int cpt; - int rc; - int i; + kib_hca_dev_t *hdev; + __u64 *pages = tx->tx_pages; + kib_fmr_poolset_t *fps; + int npages; + int size; + int cpt; + int rc; + int i; LASSERT(tx->tx_pool != NULL); LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL); - hdev = tx->tx_pool->tpo_hdev; + hdev = tx->tx_pool->tpo_hdev; for (i = 0, npages = 0; i < rd->rd_nfrags; i++) { for (size = 0; size < rd->rd_frags[i].rf_nob; @@ -575,7 +581,7 @@ kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt; fps = net->ibn_fmr_ps[cpt]; - rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->tx_u.fmr); + rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->fmr); if (rc != 0) { CERROR("Can't map %d pages: %d\n", npages, rc); return rc; @@ -583,65 +589,24 @@ kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) /* If rd is not tx_rd, it's going to get sent to a peer, who will need * the rkey */ - rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.fmr.fmr_pfmr->fmr->rkey : - tx->tx_u.fmr.fmr_pfmr->fmr->lkey; + rd->rd_key = (rd != tx->tx_rd) ? tx->fmr.fmr_pfmr->fmr->rkey : + tx->fmr.fmr_pfmr->fmr->lkey; rd->rd_frags[0].rf_addr &= ~hdev->ibh_page_mask; - rd->rd_frags[0].rf_nob = nob; + rd->rd_frags[0].rf_nob = nob; rd->rd_nfrags = 1; return 0; } -static int -kiblnd_pmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) +static void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx) { - kib_hca_dev_t *hdev; - kib_pmr_poolset_t *pps; - __u64 iova; - int cpt; - int rc; - - LASSERT(tx->tx_pool != NULL); - LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL); - - hdev = tx->tx_pool->tpo_hdev; - - iova = rd->rd_frags[0].rf_addr & ~hdev->ibh_page_mask; - - cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt; - - pps = net->ibn_pmr_ps[cpt]; - rc = kiblnd_pmr_pool_map(pps, hdev, rd, &iova, &tx->tx_u.pmr); - if (rc != 0) { - CERROR("Failed to create MR by phybuf: %d\n", rc); - return rc; - } - - /* If rd is not tx_rd, it's going to get sent to a peer, who will need - * the rkey */ - rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.pmr->pmr_mr->rkey : - tx->tx_u.pmr->pmr_mr->lkey; - rd->rd_nfrags = 1; - rd->rd_frags[0].rf_addr = iova; - rd->rd_frags[0].rf_nob = nob; - - return 0; -} - -void -kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx) -{ - kib_net_t *net = ni->ni_data; + kib_net_t *net = ni->ni_data; LASSERT(net != NULL); - if (net->ibn_fmr_ps != NULL && tx->tx_u.fmr.fmr_pfmr != NULL) { - kiblnd_fmr_pool_unmap(&tx->tx_u.fmr, tx->tx_status); - tx->tx_u.fmr.fmr_pfmr = NULL; - - } else if (net->ibn_pmr_ps != NULL && tx->tx_u.pmr != NULL) { - kiblnd_pmr_pool_unmap(tx->tx_u.pmr); - tx->tx_u.pmr = NULL; + if (net->ibn_fmr_ps && tx->fmr.fmr_pfmr) { + kiblnd_fmr_pool_unmap(&tx->fmr, tx->tx_status); + tx->fmr.fmr_pfmr = NULL; } if (tx->tx_nfrags != 0) { @@ -651,24 +616,22 @@ kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx) } } -int -kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, - kib_rdma_desc_t *rd, int nfrags) +static int kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd, + int nfrags) { - kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev; - kib_net_t *net = ni->ni_data; - struct ib_mr *mr = NULL; - __u32 nob; - int i; + kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev; + kib_net_t *net = ni->ni_data; + struct ib_mr *mr = NULL; + __u32 nob; + int i; /* If rd is not tx_rd, it's going to get sent to a peer and I'm the * RDMA sink */ tx->tx_dmadir = (rd != tx->tx_rd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE; tx->tx_nfrags = nfrags; - rd->rd_nfrags = - kiblnd_dma_map_sg(hdev->ibh_ibdev, - tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir); + rd->rd_nfrags = kiblnd_dma_map_sg(hdev->ibh_ibdev, tx->tx_frags, + tx->tx_nfrags, tx->tx_dmadir); for (i = 0, nob = 0; i < rd->rd_nfrags; i++) { rd->rd_frags[i].rf_nob = kiblnd_sg_dma_len( @@ -688,23 +651,20 @@ kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, if (net->ibn_fmr_ps != NULL) return kiblnd_fmr_map_tx(net, tx, rd, nob); - else if (net->ibn_pmr_ps != NULL) - return kiblnd_pmr_map_tx(net, tx, rd, nob); return -EINVAL; } - static int kiblnd_setup_rd_iov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd, unsigned int niov, struct kvec *iov, int offset, int nob) { - kib_net_t *net = ni->ni_data; - struct page *page; + kib_net_t *net = ni->ni_data; + struct page *page; struct scatterlist *sg; - unsigned long vaddr; - int fragnob; - int page_offset; + unsigned long vaddr; + int fragnob; + int page_offset; LASSERT(nob > 0); LASSERT(niov > 0); @@ -752,9 +712,9 @@ static int kiblnd_setup_rd_kiov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd, int nkiov, lnet_kiov_t *kiov, int offset, int nob) { - kib_net_t *net = ni->ni_data; + kib_net_t *net = ni->ni_data; struct scatterlist *sg; - int fragnob; + int fragnob; CDEBUG(D_NET, "niov %d offset %d nob %d\n", nkiov, offset, nob); @@ -793,11 +753,11 @@ kiblnd_post_tx_locked(kib_conn_t *conn, kib_tx_t *tx, int credit) __releases(conn->ibc_lock) __acquires(conn->ibc_lock) { - kib_msg_t *msg = tx->tx_msg; - kib_peer_t *peer = conn->ibc_peer; - int ver = conn->ibc_version; - int rc; - int done; + kib_msg_t *msg = tx->tx_msg; + kib_peer_t *peer = conn->ibc_peer; + int ver = conn->ibc_version; + int rc; + int done; struct ib_send_wr *bad_wrq; LASSERT(tx->tx_queued); @@ -878,8 +838,7 @@ kiblnd_post_tx_locked(kib_conn_t *conn, kib_tx_t *tx, int credit) /* close_conn will launch failover */ rc = -ENETDOWN; } else { - rc = ib_post_send(conn->ibc_cmid->qp, - tx->tx_wrq, &bad_wrq); + rc = ib_post_send(conn->ibc_cmid->qp, &tx->tx_wrq->wr, &bad_wrq); } conn->ibc_last_send = jiffies; @@ -925,9 +884,9 @@ kiblnd_post_tx_locked(kib_conn_t *conn, kib_tx_t *tx, int credit) void kiblnd_check_sends(kib_conn_t *conn) { - int ver = conn->ibc_version; + int ver = conn->ibc_version; lnet_ni_t *ni = conn->ibc_peer->ibp_ni; - kib_tx_t *tx; + kib_tx_t *tx; /* Don't send anything until after the connection is established */ if (conn->ibc_state < IBLND_CONN_ESTABLISHED) { @@ -997,9 +956,9 @@ kiblnd_check_sends(kib_conn_t *conn) static void kiblnd_tx_complete(kib_tx_t *tx, int status) { - int failed = (status != IB_WC_SUCCESS); - kib_conn_t *conn = tx->tx_conn; - int idle; + int failed = (status != IB_WC_SUCCESS); + kib_conn_t *conn = tx->tx_conn; + int idle; LASSERT(tx->tx_sending > 0); @@ -1051,11 +1010,11 @@ kiblnd_tx_complete(kib_tx_t *tx, int status) void kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob) { - kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev; - struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq]; - struct ib_send_wr *wrq = &tx->tx_wrq[tx->tx_nwrq]; - int nob = offsetof(kib_msg_t, ibm_u) + body_nob; - struct ib_mr *mr; + kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev; + struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq]; + struct ib_rdma_wr *wrq = &tx->tx_wrq[tx->tx_nwrq]; + int nob = offsetof(kib_msg_t, ibm_u) + body_nob; + struct ib_mr *mr; LASSERT(tx->tx_nwrq >= 0); LASSERT(tx->tx_nwrq < IBLND_MAX_RDMA_FRAGS + 1); @@ -1072,12 +1031,12 @@ kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob) memset(wrq, 0, sizeof(*wrq)); - wrq->next = NULL; - wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_TX); - wrq->sg_list = sge; - wrq->num_sge = 1; - wrq->opcode = IB_WR_SEND; - wrq->send_flags = IB_SEND_SIGNALED; + wrq->wr.next = NULL; + wrq->wr.wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_TX); + wrq->wr.sg_list = sge; + wrq->wr.num_sge = 1; + wrq->wr.opcode = IB_WR_SEND; + wrq->wr.send_flags = IB_SEND_SIGNALED; tx->tx_nwrq++; } @@ -1086,14 +1045,14 @@ int kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type, int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie) { - kib_msg_t *ibmsg = tx->tx_msg; - kib_rdma_desc_t *srcrd = tx->tx_rd; - struct ib_sge *sge = &tx->tx_sge[0]; - struct ib_send_wr *wrq = &tx->tx_wrq[0]; - int rc = resid; - int srcidx; - int dstidx; - int wrknob; + kib_msg_t *ibmsg = tx->tx_msg; + kib_rdma_desc_t *srcrd = tx->tx_rd; + struct ib_sge *sge = &tx->tx_sge[0]; + struct ib_rdma_wr *wrq = &tx->tx_wrq[0], *next; + int rc = resid; + int srcidx; + int dstidx; + int wrknob; LASSERT(!in_interrupt()); LASSERT(tx->tx_nwrq == 0); @@ -1135,16 +1094,17 @@ kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type, sge->length = wrknob; wrq = &tx->tx_wrq[tx->tx_nwrq]; + next = wrq + 1; - wrq->next = wrq + 1; - wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA); - wrq->sg_list = sge; - wrq->num_sge = 1; - wrq->opcode = IB_WR_RDMA_WRITE; - wrq->send_flags = 0; + wrq->wr.next = &next->wr; + wrq->wr.wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA); + wrq->wr.sg_list = sge; + wrq->wr.num_sge = 1; + wrq->wr.opcode = IB_WR_RDMA_WRITE; + wrq->wr.send_flags = 0; - wrq->wr.rdma.remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx); - wrq->wr.rdma.rkey = kiblnd_rd_frag_key(dstrd, dstidx); + wrq->remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx); + wrq->rkey = kiblnd_rd_frag_key(dstrd, dstidx); srcidx = kiblnd_rd_consume_frag(srcrd, srcidx, wrknob); dstidx = kiblnd_rd_consume_frag(dstrd, dstidx, wrknob); @@ -1170,7 +1130,7 @@ kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type, void kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn) { - struct list_head *q; + struct list_head *q; LASSERT(tx->tx_nwrq > 0); /* work items set up */ LASSERT(!tx->tx_queued); /* not queued for sending already */ @@ -1271,11 +1231,11 @@ static void kiblnd_connect_peer(kib_peer_t *peer) { struct rdma_cm_id *cmid; - kib_dev_t *dev; - kib_net_t *net = peer->ibp_ni->ni_data; + kib_dev_t *dev; + kib_net_t *net = peer->ibp_ni->ni_data; struct sockaddr_in srcaddr; struct sockaddr_in dstaddr; - int rc; + int rc; LASSERT(net != NULL); LASSERT(peer->ibp_connecting > 0); @@ -1335,12 +1295,12 @@ kiblnd_connect_peer(kib_peer_t *peer) void kiblnd_launch_tx(lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid) { - kib_peer_t *peer; - kib_peer_t *peer2; - kib_conn_t *conn; - rwlock_t *g_lock = &kiblnd_data.kib_global_lock; - unsigned long flags; - int rc; + kib_peer_t *peer; + kib_peer_t *peer2; + kib_conn_t *conn; + rwlock_t *g_lock = &kiblnd_data.kib_global_lock; + unsigned long flags; + int rc; /* If I get here, I've committed to send, so I complete the tx with * failure on any problems */ @@ -1456,20 +1416,21 @@ kiblnd_launch_tx(lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid) int kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg) { - lnet_hdr_t *hdr = &lntmsg->msg_hdr; - int type = lntmsg->msg_type; + lnet_hdr_t *hdr = &lntmsg->msg_hdr; + int type = lntmsg->msg_type; lnet_process_id_t target = lntmsg->msg_target; - int target_is_router = lntmsg->msg_target_is_router; - int routing = lntmsg->msg_routing; - unsigned int payload_niov = lntmsg->msg_niov; - struct kvec *payload_iov = lntmsg->msg_iov; - lnet_kiov_t *payload_kiov = lntmsg->msg_kiov; - unsigned int payload_offset = lntmsg->msg_offset; - unsigned int payload_nob = lntmsg->msg_len; - kib_msg_t *ibmsg; - kib_tx_t *tx; - int nob; - int rc; + int target_is_router = lntmsg->msg_target_is_router; + int routing = lntmsg->msg_routing; + unsigned int payload_niov = lntmsg->msg_niov; + struct kvec *payload_iov = lntmsg->msg_iov; + lnet_kiov_t *payload_kiov = lntmsg->msg_kiov; + unsigned int payload_offset = lntmsg->msg_offset; + unsigned int payload_nob = lntmsg->msg_len; + kib_msg_t *ibmsg; + kib_rdma_desc_t *rd; + kib_tx_t *tx; + int nob; + int rc; /* NB 'private' is different depending on what we're sending.... */ @@ -1510,16 +1471,14 @@ kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg) } ibmsg = tx->tx_msg; - + rd = &ibmsg->ibm_u.get.ibgm_rd; if ((lntmsg->msg_md->md_options & LNET_MD_KIOV) == 0) - rc = kiblnd_setup_rd_iov(ni, tx, - &ibmsg->ibm_u.get.ibgm_rd, + rc = kiblnd_setup_rd_iov(ni, tx, rd, lntmsg->msg_md->md_niov, lntmsg->msg_md->md_iov.iov, 0, lntmsg->msg_md->md_length); else - rc = kiblnd_setup_rd_kiov(ni, tx, - &ibmsg->ibm_u.get.ibgm_rd, + rc = kiblnd_setup_rd_kiov(ni, tx, rd, lntmsg->msg_md->md_niov, lntmsg->msg_md->md_iov.kiov, 0, lntmsg->msg_md->md_length); @@ -1530,7 +1489,7 @@ kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg) return -EIO; } - nob = offsetof(kib_get_msg_t, ibgm_rd.rd_frags[tx->tx_nfrags]); + nob = offsetof(kib_get_msg_t, ibgm_rd.rd_frags[rd->rd_nfrags]); ibmsg->ibm_u.get.ibgm_cookie = tx->tx_cookie; ibmsg->ibm_u.get.ibgm_hdr = *hdr; @@ -1628,13 +1587,13 @@ static void kiblnd_reply(lnet_ni_t *ni, kib_rx_t *rx, lnet_msg_t *lntmsg) { lnet_process_id_t target = lntmsg->msg_target; - unsigned int niov = lntmsg->msg_niov; - struct kvec *iov = lntmsg->msg_iov; - lnet_kiov_t *kiov = lntmsg->msg_kiov; - unsigned int offset = lntmsg->msg_offset; - unsigned int nob = lntmsg->msg_len; - kib_tx_t *tx; - int rc; + unsigned int niov = lntmsg->msg_niov; + struct kvec *iov = lntmsg->msg_iov; + lnet_kiov_t *kiov = lntmsg->msg_kiov; + unsigned int offset = lntmsg->msg_offset; + unsigned int nob = lntmsg->msg_len; + kib_tx_t *tx; + int rc; tx = kiblnd_get_idle_tx(ni, rx->rx_conn->ibc_peer->ibp_nid); if (tx == NULL) { @@ -1691,14 +1650,13 @@ kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, unsigned int niov, struct kvec *iov, lnet_kiov_t *kiov, unsigned int offset, unsigned int mlen, unsigned int rlen) { - kib_rx_t *rx = private; - kib_msg_t *rxmsg = rx->rx_msg; - kib_conn_t *conn = rx->rx_conn; - kib_tx_t *tx; - kib_msg_t *txmsg; - int nob; - int post_credit = IBLND_POSTRX_PEER_CREDIT; - int rc = 0; + kib_rx_t *rx = private; + kib_msg_t *rxmsg = rx->rx_msg; + kib_conn_t *conn = rx->rx_conn; + kib_tx_t *tx; + int nob; + int post_credit = IBLND_POSTRX_PEER_CREDIT; + int rc = 0; LASSERT(mlen <= rlen); LASSERT(!in_interrupt()); @@ -1732,7 +1690,10 @@ kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, lnet_finalize(ni, lntmsg, 0); break; - case IBLND_MSG_PUT_REQ: + case IBLND_MSG_PUT_REQ: { + kib_msg_t *txmsg; + kib_rdma_desc_t *rd; + if (mlen == 0) { lnet_finalize(ni, lntmsg, 0); kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, 0, @@ -1750,13 +1711,12 @@ kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, } txmsg = tx->tx_msg; + rd = &txmsg->ibm_u.putack.ibpam_rd; if (kiov == NULL) - rc = kiblnd_setup_rd_iov(ni, tx, - &txmsg->ibm_u.putack.ibpam_rd, + rc = kiblnd_setup_rd_iov(ni, tx, rd, niov, iov, offset, mlen); else - rc = kiblnd_setup_rd_kiov(ni, tx, - &txmsg->ibm_u.putack.ibpam_rd, + rc = kiblnd_setup_rd_kiov(ni, tx, rd, niov, kiov, offset, mlen); if (rc != 0) { CERROR("Can't setup PUT sink for %s: %d\n", @@ -1768,7 +1728,7 @@ kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, break; } - nob = offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[tx->tx_nfrags]); + nob = offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[rd->rd_nfrags]); txmsg->ibm_u.putack.ibpam_src_cookie = rxmsg->ibm_u.putreq.ibprm_cookie; txmsg->ibm_u.putack.ibpam_dst_cookie = tx->tx_cookie; @@ -1781,6 +1741,7 @@ kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, /* reposted buffer reserved for PUT_DONE */ post_credit = IBLND_POSTRX_NO_CREDIT; break; + } case IBLND_MSG_GET_REQ: if (lntmsg != NULL) { @@ -1828,8 +1789,8 @@ kiblnd_peer_alive(kib_peer_t *peer) static void kiblnd_peer_notify(kib_peer_t *peer) { - int error = 0; - unsigned long last_alive = 0; + int error = 0; + unsigned long last_alive = 0; unsigned long flags; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); @@ -1860,9 +1821,9 @@ kiblnd_close_conn_locked(kib_conn_t *conn, int error) * connection to be finished off by the connd. Otherwise the connd is * already dealing with it (either to set it up or tear it down). * Caller holds kib_global_lock exclusively in irq context */ - kib_peer_t *peer = conn->ibc_peer; - kib_dev_t *dev; - unsigned long flags; + kib_peer_t *peer = conn->ibc_peer; + kib_dev_t *dev; + unsigned long flags; LASSERT(error != 0 || conn->ibc_state >= IBLND_CONN_ESTABLISHED); @@ -1934,8 +1895,8 @@ kiblnd_close_conn(kib_conn_t *conn, int error) static void kiblnd_handle_early_rxs(kib_conn_t *conn) { - unsigned long flags; - kib_rx_t *rx; + unsigned long flags; + kib_rx_t *rx; kib_rx_t *tmp; LASSERT(!in_interrupt()); @@ -1957,9 +1918,9 @@ static void kiblnd_abort_txs(kib_conn_t *conn, struct list_head *txs) { LIST_HEAD(zombies); - struct list_head *tmp; - struct list_head *nxt; - kib_tx_t *tx; + struct list_head *tmp; + struct list_head *nxt; + kib_tx_t *tx; spin_lock(&conn->ibc_lock); @@ -2018,7 +1979,7 @@ void kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error) { LIST_HEAD(zombies); - unsigned long flags; + unsigned long flags; LASSERT(error != 0); LASSERT(!in_interrupt()); @@ -2071,12 +2032,12 @@ kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error) void kiblnd_connreq_done(kib_conn_t *conn, int status) { - kib_peer_t *peer = conn->ibc_peer; - kib_tx_t *tx; + kib_peer_t *peer = conn->ibc_peer; + kib_tx_t *tx; kib_tx_t *tmp; - struct list_head txs; - unsigned long flags; - int active; + struct list_head txs; + unsigned long flags; + int active; active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT); @@ -2166,7 +2127,7 @@ kiblnd_connreq_done(kib_conn_t *conn, int status) static void kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej) { - int rc; + int rc; rc = rdma_reject(cmid, rej, sizeof(*rej)); @@ -2177,22 +2138,23 @@ kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej) static int kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob) { - rwlock_t *g_lock = &kiblnd_data.kib_global_lock; - kib_msg_t *reqmsg = priv; - kib_msg_t *ackmsg; - kib_dev_t *ibdev; - kib_peer_t *peer; - kib_peer_t *peer2; - kib_conn_t *conn; - lnet_ni_t *ni = NULL; - kib_net_t *net = NULL; - lnet_nid_t nid; + rwlock_t *g_lock = &kiblnd_data.kib_global_lock; + kib_msg_t *reqmsg = priv; + kib_msg_t *ackmsg; + kib_dev_t *ibdev; + kib_peer_t *peer; + kib_peer_t *peer2; + kib_conn_t *conn; + lnet_ni_t *ni = NULL; + kib_net_t *net = NULL; + lnet_nid_t nid; struct rdma_conn_param cp; - kib_rej_t rej; - int version = IBLND_MSG_VERSION; - unsigned long flags; - int rc; - struct sockaddr_in *peer_addr; + kib_rej_t rej; + int version = IBLND_MSG_VERSION; + unsigned long flags; + int rc; + struct sockaddr_in *peer_addr; + LASSERT(!in_interrupt()); /* cmid inherits 'context' from the corresponding listener id */ @@ -2200,14 +2162,15 @@ kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob) LASSERT(ibdev != NULL); memset(&rej, 0, sizeof(rej)); - rej.ibr_magic = IBLND_MSG_MAGIC; - rej.ibr_why = IBLND_REJECT_FATAL; + rej.ibr_magic = IBLND_MSG_MAGIC; + rej.ibr_why = IBLND_REJECT_FATAL; rej.ibr_cp.ibcp_max_msg_size = IBLND_MSG_SIZE; peer_addr = (struct sockaddr_in *)&(cmid->route.addr.dst_addr); if (*kiblnd_tunables.kib_require_priv_port && ntohs(peer_addr->sin_port) >= PROT_SOCK) { __u32 ip = ntohl(peer_addr->sin_addr.s_addr); + CERROR("Peer's port (%pI4h:%hu) is not privileged\n", &ip, ntohs(peer_addr->sin_port)); goto failed; @@ -2243,7 +2206,7 @@ kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob) } nid = reqmsg->ibm_srcnid; - ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid)); + ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid)); if (ni != NULL) { net = (kib_net_t *)ni->ni_data; @@ -2394,7 +2357,7 @@ kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob) * CM callback doesn't destroy cmid. */ conn->ibc_incarnation = reqmsg->ibm_srcstamp; - conn->ibc_credits = IBLND_MSG_QUEUE_SIZE(version); + conn->ibc_credits = IBLND_MSG_QUEUE_SIZE(version); conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(version); LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(version) <= IBLND_RX_MSGS(version)); @@ -2412,12 +2375,12 @@ kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob) memset(&cp, 0, sizeof(cp)); cp.private_data = ackmsg; - cp.private_data_len = ackmsg->ibm_nob; + cp.private_data_len = ackmsg->ibm_nob; cp.responder_resources = 0; /* No atomic ops or RDMA reads */ - cp.initiator_depth = 0; + cp.initiator_depth = 0; cp.flow_control = 1; - cp.retry_count = *kiblnd_tunables.kib_retry_count; - cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count; + cp.retry_count = *kiblnd_tunables.kib_retry_count; + cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count; CDEBUG(D_NET, "Accept %s\n", libcfs_nid2str(nid)); @@ -2439,7 +2402,7 @@ kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob) if (ni != NULL) lnet_ni_decref(ni); - rej.ibr_version = version; + rej.ibr_version = version; rej.ibr_cp.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version); rej.ibr_cp.ibcp_max_frags = IBLND_RDMA_FRAGS(version); kiblnd_reject(cmid, &rej); @@ -2451,10 +2414,10 @@ static void kiblnd_reconnect(kib_conn_t *conn, int version, __u64 incarnation, int why, kib_connparams_t *cp) { - kib_peer_t *peer = conn->ibc_peer; - char *reason; - int retry = 0; - unsigned long flags; + kib_peer_t *peer = conn->ibc_peer; + char *reason; + int retry = 0; + unsigned long flags; LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT); LASSERT(peer->ibp_connecting > 0); /* 'conn' at least */ @@ -2513,7 +2476,7 @@ kiblnd_reconnect(kib_conn_t *conn, int version, static void kiblnd_rejected(kib_conn_t *conn, int reason, void *priv, int priv_nob) { - kib_peer_t *peer = conn->ibc_peer; + kib_peer_t *peer = conn->ibc_peer; LASSERT(!in_interrupt()); LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT); @@ -2532,10 +2495,10 @@ kiblnd_rejected(kib_conn_t *conn, int reason, void *priv, int priv_nob) case IB_CM_REJ_CONSUMER_DEFINED: if (priv_nob >= offsetof(kib_rej_t, ibr_padding)) { - kib_rej_t *rej = priv; - kib_connparams_t *cp = NULL; - int flip = 0; - __u64 incarnation = -1; + kib_rej_t *rej = priv; + kib_connparams_t *cp = NULL; + int flip = 0; + __u64 incarnation = -1; /* NB. default incarnation is -1 because: * a) V1 will ignore dst incarnation in connreq. @@ -2652,13 +2615,13 @@ kiblnd_rejected(kib_conn_t *conn, int reason, void *priv, int priv_nob) static void kiblnd_check_connreply(kib_conn_t *conn, void *priv, int priv_nob) { - kib_peer_t *peer = conn->ibc_peer; - lnet_ni_t *ni = peer->ibp_ni; - kib_net_t *net = ni->ni_data; - kib_msg_t *msg = priv; - int ver = conn->ibc_version; - int rc = kiblnd_unpack_msg(msg, priv_nob); - unsigned long flags; + kib_peer_t *peer = conn->ibc_peer; + lnet_ni_t *ni = peer->ibp_ni; + kib_net_t *net = ni->ni_data; + kib_msg_t *msg = priv; + int ver = conn->ibc_version; + int rc = kiblnd_unpack_msg(msg, priv_nob); + unsigned long flags; LASSERT(net != NULL); @@ -2726,8 +2689,8 @@ kiblnd_check_connreply(kib_conn_t *conn, void *priv, int priv_nob) goto failed; } - conn->ibc_incarnation = msg->ibm_srcstamp; - conn->ibc_credits = + conn->ibc_incarnation = msg->ibm_srcstamp; + conn->ibc_credits = conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(ver); LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(ver) <= IBLND_RX_MSGS(ver)); @@ -2749,20 +2712,20 @@ kiblnd_check_connreply(kib_conn_t *conn, void *priv, int priv_nob) static int kiblnd_active_connect(struct rdma_cm_id *cmid) { - kib_peer_t *peer = (kib_peer_t *)cmid->context; - kib_conn_t *conn; - kib_msg_t *msg; - struct rdma_conn_param cp; - int version; - __u64 incarnation; - unsigned long flags; - int rc; + kib_peer_t *peer = (kib_peer_t *)cmid->context; + kib_conn_t *conn; + kib_msg_t *msg; + struct rdma_conn_param cp; + int version; + __u64 incarnation; + unsigned long flags; + int rc; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); incarnation = peer->ibp_incarnation; - version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION : - peer->ibp_version; + version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION : + peer->ibp_version; read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags); @@ -2793,8 +2756,8 @@ kiblnd_active_connect(struct rdma_cm_id *cmid) cp.private_data_len = msg->ibm_nob; cp.responder_resources = 0; /* No atomic ops or RDMA reads */ cp.initiator_depth = 0; - cp.flow_control = 1; - cp.retry_count = *kiblnd_tunables.kib_retry_count; + cp.flow_control = 1; + cp.retry_count = *kiblnd_tunables.kib_retry_count; cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count; LASSERT(cmid->context == (void *)conn); @@ -2814,9 +2777,9 @@ kiblnd_active_connect(struct rdma_cm_id *cmid) int kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event) { - kib_peer_t *peer; - kib_conn_t *conn; - int rc; + kib_peer_t *peer; + kib_conn_t *conn; + int rc; switch (event->event) { default: @@ -2983,8 +2946,8 @@ kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event) static int kiblnd_check_txs_locked(kib_conn_t *conn, struct list_head *txs) { - kib_tx_t *tx; - struct list_head *ttmp; + kib_tx_t *tx; + struct list_head *ttmp; list_for_each(ttmp, txs) { tx = list_entry(ttmp, kib_tx_t, tx_list); @@ -3022,13 +2985,13 @@ kiblnd_check_conns(int idx) { LIST_HEAD(closes); LIST_HEAD(checksends); - struct list_head *peers = &kiblnd_data.kib_peers[idx]; - struct list_head *ptmp; - kib_peer_t *peer; - kib_conn_t *conn; + struct list_head *peers = &kiblnd_data.kib_peers[idx]; + struct list_head *ptmp; + kib_peer_t *peer; + kib_conn_t *conn; kib_conn_t *tmp; - struct list_head *ctmp; - unsigned long flags; + struct list_head *ctmp; + unsigned long flags; /* NB. We expect to have a look at all the peers and not find any * RDMAs to time out, so we just use a shared lock while we @@ -3114,14 +3077,14 @@ kiblnd_disconnect_conn(kib_conn_t *conn) int kiblnd_connd(void *arg) { - wait_queue_t wait; - unsigned long flags; - kib_conn_t *conn; - int timeout; - int i; - int dropped_lock; - int peer_index = 0; - unsigned long deadline = jiffies; + wait_queue_t wait; + unsigned long flags; + kib_conn_t *conn; + int timeout; + int i; + int dropped_lock; + int peer_index = 0; + unsigned long deadline = jiffies; cfs_block_allsigs(); @@ -3135,8 +3098,7 @@ kiblnd_connd(void *arg) dropped_lock = 0; if (!list_empty(&kiblnd_data.kib_connd_zombies)) { - conn = list_entry(kiblnd_data. \ - kib_connd_zombies.next, + conn = list_entry(kiblnd_data.kib_connd_zombies.next, kib_conn_t, ibc_list); list_del(&conn->ibc_list); @@ -3169,7 +3131,7 @@ kiblnd_connd(void *arg) if (timeout <= 0) { const int n = 4; const int p = 1; - int chunk = kiblnd_data.kib_peer_hash_size; + int chunk = kiblnd_data.kib_peer_hash_size; spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags); dropped_lock = 1; @@ -3273,9 +3235,9 @@ kiblnd_cq_completion(struct ib_cq *cq, void *arg) * consuming my CQ I could be called after all completions have * occurred. But in this case, ibc_nrx == 0 && ibc_nsends_posted == 0 * and this CQ is about to be destroyed so I NOOP. */ - kib_conn_t *conn = (kib_conn_t *)arg; - struct kib_sched_info *sched = conn->ibc_sched; - unsigned long flags; + kib_conn_t *conn = arg; + struct kib_sched_info *sched = conn->ibc_sched; + unsigned long flags; LASSERT(cq == conn->ibc_cq); @@ -3309,15 +3271,15 @@ kiblnd_cq_event(struct ib_event *event, void *arg) int kiblnd_scheduler(void *arg) { - long id = (long)arg; - struct kib_sched_info *sched; - kib_conn_t *conn; - wait_queue_t wait; - unsigned long flags; - struct ib_wc wc; - int did_something; - int busy_loops = 0; - int rc; + long id = (long)arg; + struct kib_sched_info *sched; + kib_conn_t *conn; + wait_queue_t wait; + unsigned long flags; + struct ib_wc wc; + int did_something; + int busy_loops = 0; + int rc; cfs_block_allsigs(); @@ -3432,11 +3394,11 @@ kiblnd_scheduler(void *arg) int kiblnd_failover_thread(void *arg) { - rwlock_t *glock = &kiblnd_data.kib_global_lock; - kib_dev_t *dev; - wait_queue_t wait; - unsigned long flags; - int rc; + rwlock_t *glock = &kiblnd_data.kib_global_lock; + kib_dev_t *dev; + wait_queue_t wait; + unsigned long flags; + int rc; LASSERT(*kiblnd_tunables.kib_dev_failover != 0); @@ -3446,8 +3408,8 @@ kiblnd_failover_thread(void *arg) write_lock_irqsave(glock, flags); while (!kiblnd_data.kib_shutdown) { - int do_failover = 0; - int long_sleep; + int do_failover = 0; + int long_sleep; list_for_each_entry(dev, &kiblnd_data.kib_failed_devs, ibd_fail_list) { diff --git a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c index eedf01afd..1d4e7efb5 100644 --- a/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c +++ b/kernel/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c @@ -126,11 +126,6 @@ static int fmr_cache = 1; module_param(fmr_cache, int, 0444); MODULE_PARM_DESC(fmr_cache, "non-zero to enable FMR caching"); -/* NB: this value is shared by all CPTs, it can grow at runtime */ -static int pmr_pool_size = 512; -module_param(pmr_pool_size, int, 0444); -MODULE_PARM_DESC(pmr_pool_size, "size of MR cache pmr pool on each CPT"); - /* * 0: disable failover * 1: enable failover if necessary @@ -140,7 +135,6 @@ static int dev_failover; module_param(dev_failover, int, 0444); MODULE_PARM_DESC(dev_failover, "HCA failover for bonding (0 off, 1 on, other values reserved)"); - static int require_privileged_port; module_param(require_privileged_port, int, 0644); MODULE_PARM_DESC(require_privileged_port, "require privileged port when accepting connection"); @@ -150,30 +144,29 @@ module_param(use_privileged_port, int, 0644); MODULE_PARM_DESC(use_privileged_port, "use privileged port when initiating connection"); kib_tunables_t kiblnd_tunables = { - .kib_dev_failover = &dev_failover, - .kib_service = &service, - .kib_cksum = &cksum, - .kib_timeout = &timeout, - .kib_keepalive = &keepalive, - .kib_ntx = &ntx, - .kib_credits = &credits, - .kib_peertxcredits = &peer_credits, - .kib_peercredits_hiw = &peer_credits_hiw, - .kib_peerrtrcredits = &peer_buffer_credits, - .kib_peertimeout = &peer_timeout, - .kib_default_ipif = &ipif_name, - .kib_retry_count = &retry_count, - .kib_rnr_retry_count = &rnr_retry_count, - .kib_concurrent_sends = &concurrent_sends, - .kib_ib_mtu = &ib_mtu, - .kib_map_on_demand = &map_on_demand, - .kib_fmr_pool_size = &fmr_pool_size, - .kib_fmr_flush_trigger = &fmr_flush_trigger, - .kib_fmr_cache = &fmr_cache, - .kib_pmr_pool_size = &pmr_pool_size, - .kib_require_priv_port = &require_privileged_port, - .kib_use_priv_port = &use_privileged_port, - .kib_nscheds = &nscheds + .kib_dev_failover = &dev_failover, + .kib_service = &service, + .kib_cksum = &cksum, + .kib_timeout = &timeout, + .kib_keepalive = &keepalive, + .kib_ntx = &ntx, + .kib_credits = &credits, + .kib_peertxcredits = &peer_credits, + .kib_peercredits_hiw = &peer_credits_hiw, + .kib_peerrtrcredits = &peer_buffer_credits, + .kib_peertimeout = &peer_timeout, + .kib_default_ipif = &ipif_name, + .kib_retry_count = &retry_count, + .kib_rnr_retry_count = &rnr_retry_count, + .kib_concurrent_sends = &concurrent_sends, + .kib_ib_mtu = &ib_mtu, + .kib_map_on_demand = &map_on_demand, + .kib_fmr_pool_size = &fmr_pool_size, + .kib_fmr_flush_trigger = &fmr_flush_trigger, + .kib_fmr_cache = &fmr_cache, + .kib_require_priv_port = &require_privileged_port, + .kib_use_priv_port = &use_privileged_port, + .kib_nscheds = &nscheds }; int |