#include "o2iblnd.h"
-static struct lnet_lnd the_o2iblnd;
+static const struct lnet_lnd the_o2iblnd;
struct kib_data kiblnd_data;
if (flip) {
/* leave magic unflipped as a clue to peer_ni endianness */
msg->ibm_version = version;
- CLASSERT (sizeof(msg->ibm_type) == 1);
- CLASSERT (sizeof(msg->ibm_credits) == 1);
+ BUILD_BUG_ON(sizeof(msg->ibm_type) != 1);
+ BUILD_BUG_ON(sizeof(msg->ibm_credits) != 1);
msg->ibm_nob = msg_nob;
__swab64s(&msg->ibm_srcnid);
__swab64s(&msg->ibm_srcstamp);
* they are destroyed, so we can be assured that _all_ state to do
* with this peer_ni has been cleaned up when its refcount drops to
* zero. */
- atomic_dec(&net->ibn_npeers);
+ if (atomic_dec_and_test(&net->ibn_npeers))
+ wake_up_var(&net->ibn_npeers);
}
struct kib_peer_ni *
static int
kiblnd_del_peer(struct lnet_ni *ni, lnet_nid_t nid)
{
- struct list_head zombies = LIST_HEAD_INIT(zombies);
+ LIST_HEAD(zombies);
struct list_head *ptmp;
struct list_head *pnxt;
struct kib_peer_ni *peer_ni;
static void
kiblnd_debug_rx(struct kib_rx *rx)
{
- CDEBUG(D_CONSOLE, " %p status %d msg_type %x cred %d\n",
- rx, rx->rx_status, rx->rx_msg->ibm_type,
- rx->rx_msg->ibm_credits);
+ CDEBUG(D_CONSOLE, " %p msg_type %x cred %d\n",
+ rx, rx->rx_msg->ibm_type,
+ rx->rx_msg->ibm_credits);
}
static void
spin_unlock(&conn->ibc_lock);
}
-int
-kiblnd_translate_mtu(int value)
-{
- switch (value) {
- default:
- return -1;
- case 0:
- return 0;
- case 256:
- return IB_MTU_256;
- case 512:
- return IB_MTU_512;
- case 1024:
- return IB_MTU_1024;
- case 2048:
- return IB_MTU_2048;
- case 4096:
- return IB_MTU_4096;
- }
-}
-
static void
kiblnd_setup_mtu_locked(struct rdma_cm_id *cmid)
{
- int mtu;
-
/* XXX There is no path record for iWARP, set by netdev->change_mtu? */
if (cmid->route.path_rec == NULL)
return;
- mtu = kiblnd_translate_mtu(*kiblnd_tunables.kib_ib_mtu);
- LASSERT (mtu >= 0);
- if (mtu != 0)
- cmid->route.path_rec->mtu = mtu;
+ if (*kiblnd_tunables.kib_ib_mtu)
+ cmid->route.path_rec->mtu =
+ ib_mtu_int_to_enum(*kiblnd_tunables.kib_ib_mtu);
}
static int
kiblnd_get_completion_vector(struct kib_conn *conn, int cpt)
{
- cpumask_t *mask;
+ cpumask_var_t *mask;
int vectors;
int off;
int i;
/* hash NID to CPU id in this partition... */
ibp_nid = conn->ibc_peer->ibp_nid;
- off = do_div(ibp_nid, cpumask_weight(mask));
- for_each_cpu(i, mask) {
+ off = do_div(ibp_nid, cpumask_weight(*mask));
+ for_each_cpu(i, *mask) {
if (off-- == 0)
return i % vectors;
}
* One WR for the LNet message
* And ibc_max_frags for the transfer WRs
*/
- unsigned int ret = 1 + conn->ibc_max_frags;
+ int ret;
+ int multiplier = 1 + conn->ibc_max_frags;
enum kib_dev_caps dev_caps = conn->ibc_hdev->ibh_dev->ibd_dev_caps;
/* FastReg needs two extra WRs for map and invalidate */
if (dev_caps & IBLND_DEV_CAPS_FASTREG_ENABLED)
- ret += 2;
+ multiplier += 2;
/* account for a maximum of ibc_queue_depth in-flight transfers */
- ret *= conn->ibc_queue_depth;
- return ret;
+ ret = multiplier * conn->ibc_queue_depth;
+
+ if (ret > conn->ibc_hdev->ibh_max_qp_wr) {
+ CDEBUG(D_NET, "peer_credits %u will result in send work "
+ "request size %d larger than maximum %d device "
+ "can handle\n", conn->ibc_queue_depth, ret,
+ conn->ibc_hdev->ibh_max_qp_wr);
+ conn->ibc_queue_depth =
+ conn->ibc_hdev->ibh_max_qp_wr / multiplier;
+ }
+
+ /* don't go beyond the maximum the device can handle */
+ return min(ret, conn->ibc_hdev->ibh_max_qp_wr);
}
struct kib_conn *
rwlock_t *glock = &kiblnd_data.kib_global_lock;
struct kib_net *net = peer_ni->ibp_ni->ni_data;
struct kib_dev *dev;
- struct ib_qp_init_attr *init_qp_attr;
+ struct ib_qp_init_attr init_qp_attr = {};
struct kib_sched_info *sched;
#ifdef HAVE_IB_CQ_INIT_ATTR
struct ib_cq_init_attr cq_attr = {};
*/
cpt = sched->ibs_cpt;
- LIBCFS_CPT_ALLOC(init_qp_attr, lnet_cpt_table(), cpt,
- sizeof(*init_qp_attr));
- if (init_qp_attr == NULL) {
- CERROR("Can't allocate qp_attr for %s\n",
- libcfs_nid2str(peer_ni->ibp_nid));
- goto failed_0;
- }
-
LIBCFS_CPT_ALLOC(conn, lnet_cpt_table(), cpt, sizeof(*conn));
if (conn == NULL) {
CERROR("Can't allocate connection for %s\n",
libcfs_nid2str(peer_ni->ibp_nid));
- goto failed_1;
+ goto failed_0;
}
conn->ibc_state = IBLND_CONN_INIT;
goto failed_2;
}
- init_qp_attr->event_handler = kiblnd_qp_event;
- init_qp_attr->qp_context = conn;
- init_qp_attr->cap.max_send_sge = *kiblnd_tunables.kib_wrq_sge;
- init_qp_attr->cap.max_recv_sge = 1;
- init_qp_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
- init_qp_attr->qp_type = IB_QPT_RC;
- init_qp_attr->send_cq = cq;
- init_qp_attr->recv_cq = cq;
-
- conn->ibc_sched = sched;
-
- do {
- init_qp_attr->cap.max_send_wr = kiblnd_send_wrs(conn);
- init_qp_attr->cap.max_recv_wr = IBLND_RECV_WRS(conn);
-
- rc = rdma_create_qp(cmid, conn->ibc_hdev->ibh_pd, init_qp_attr);
- if (!rc || conn->ibc_queue_depth < 2)
- break;
-
- conn->ibc_queue_depth--;
- } while (rc);
+ init_qp_attr.event_handler = kiblnd_qp_event;
+ init_qp_attr.qp_context = conn;
+ init_qp_attr.cap.max_send_sge = *kiblnd_tunables.kib_wrq_sge;
+ init_qp_attr.cap.max_recv_sge = 1;
+ init_qp_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
+ init_qp_attr.qp_type = IB_QPT_RC;
+ init_qp_attr.send_cq = cq;
+ init_qp_attr.recv_cq = cq;
+ /*
+ * kiblnd_send_wrs() can change the connection's queue depth if
+ * the maximum work requests for the device is maxed out
+ */
+ init_qp_attr.cap.max_send_wr = kiblnd_send_wrs(conn);
+ init_qp_attr.cap.max_recv_wr = IBLND_RECV_WRS(conn);
+ rc = rdma_create_qp(cmid, conn->ibc_hdev->ibh_pd, &init_qp_attr);
if (rc) {
CERROR("Can't create QP: %d, send_wr: %d, recv_wr: %d, "
"send_sge: %d, recv_sge: %d\n",
- rc, init_qp_attr->cap.max_send_wr,
- init_qp_attr->cap.max_recv_wr,
- init_qp_attr->cap.max_send_sge,
- init_qp_attr->cap.max_recv_sge);
+ rc, init_qp_attr.cap.max_send_wr,
+ init_qp_attr.cap.max_recv_wr,
+ init_qp_attr.cap.max_send_sge,
+ init_qp_attr.cap.max_recv_sge);
goto failed_2;
}
+ conn->ibc_sched = sched;
+
if (conn->ibc_queue_depth != peer_ni->ibp_queue_depth)
CWARN("peer %s - queue depth reduced from %u to %u"
" to allow for qp creation\n",
kiblnd_map_rx_descs(conn);
- LIBCFS_FREE(init_qp_attr, sizeof(*init_qp_attr));
-
/* 1 ref for caller and each rxmsg */
atomic_set(&conn->ibc_refcount, 1 + IBLND_RX_MSGS(conn));
conn->ibc_nrx = IBLND_RX_MSGS(conn);
failed_2:
kiblnd_destroy_conn(conn);
LIBCFS_FREE(conn, sizeof(*conn));
- failed_1:
- LIBCFS_FREE(init_qp_attr, sizeof(*init_qp_attr));
failed_0:
return NULL;
}
kiblnd_destroy_conn(struct kib_conn *conn)
{
struct rdma_cm_id *cmid = conn->ibc_cmid;
- struct kib_peer_ni *peer_ni = conn->ibc_peer;
- int rc;
+ struct kib_peer_ni *peer_ni = conn->ibc_peer;
LASSERT (!in_interrupt());
LASSERT (atomic_read(&conn->ibc_refcount) == 0);
if (cmid != NULL && cmid->qp != NULL)
rdma_destroy_qp(cmid);
- if (conn->ibc_cq != NULL) {
- rc = ib_destroy_cq(conn->ibc_cq);
- if (rc != 0)
- CWARN("Error destroying CQ: %d\n", rc);
- }
+ if (conn->ibc_cq)
+ ib_destroy_cq(conn->ibc_cq);
kiblnd_txlist_done(&conn->ibc_zombie_txs, -ECONNABORTED,
LNET_MSG_STATUS_OK);
if (conn->ibc_rx_pages != NULL)
kiblnd_unmap_rx_descs(conn);
- if (conn->ibc_rxs != NULL) {
- LIBCFS_FREE(conn->ibc_rxs,
- IBLND_RX_MSGS(conn) * sizeof(struct kib_rx));
- }
+ if (conn->ibc_rxs != NULL)
+ CFS_FREE_PTR_ARRAY(conn->ibc_rxs, IBLND_RX_MSGS(conn));
if (conn->ibc_connvars != NULL)
LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
}
static void
-kiblnd_query(struct lnet_ni *ni, lnet_nid_t nid, time64_t *when)
-{
- time64_t last_alive = 0;
- time64_t now = ktime_get_seconds();
- rwlock_t *glock = &kiblnd_data.kib_global_lock;
- struct kib_peer_ni *peer_ni;
- unsigned long flags;
-
- read_lock_irqsave(glock, flags);
-
- peer_ni = kiblnd_find_peer_locked(ni, nid);
- if (peer_ni != NULL)
- last_alive = peer_ni->ibp_last_alive;
-
- read_unlock_irqrestore(glock, flags);
-
- if (last_alive != 0)
- *when = last_alive;
-
- /* peer_ni is not persistent in hash, trigger peer_ni creation
- * and connection establishment with a NULL tx */
- if (peer_ni == NULL)
- kiblnd_launch_tx(ni, NULL, nid);
-
- CDEBUG(D_NET, "peer_ni %s %p, alive %lld secs ago\n",
- libcfs_nid2str(nid), peer_ni,
- last_alive ? now - last_alive : -1);
- return;
-}
-
-static void
kiblnd_free_pages(struct kib_pages *p)
{
int npages = p->ibp_npages;
if (i++ % 50 == 0)
CDEBUG(D_NET, "%s: Wait for failover\n",
dev->ibd_ifname);
- set_current_state(TASK_INTERRUPTIBLE);
- schedule_timeout(cfs_time_seconds(1) / 100);
+ schedule_timeout_interruptible(cfs_time_seconds(1) / 100);
read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
}
dev = net->ibn_dev;
- /* pre-mapped messages are not bigger than 1 page */
- CLASSERT (IBLND_MSG_SIZE <= PAGE_SIZE);
+ /* pre-mapped messages are not bigger than 1 page */
+ BUILD_BUG_ON(IBLND_MSG_SIZE > PAGE_SIZE);
- /* No fancy arithmetic when we do the buffer calculations */
- CLASSERT (PAGE_SIZE % IBLND_MSG_SIZE == 0);
+ /* No fancy arithmetic when we do the buffer calculations */
+ BUILD_BUG_ON(PAGE_SIZE % IBLND_MSG_SIZE != 0);
tpo->tpo_hdev = kiblnd_current_hdev(dev);
fpo_list);
fpo->fpo_failed = 1;
- list_del(&fpo->fpo_list);
if (fpo->fpo_map_count == 0)
- list_add(&fpo->fpo_list, zombies);
+ list_move(&fpo->fpo_list, zombies);
else
- list_add(&fpo->fpo_list, &fps->fps_failed_pool_list);
+ list_move(&fpo->fpo_list, &fps->fps_failed_pool_list);
}
spin_unlock(&fps->fps_lock);
void
kiblnd_fmr_pool_unmap(struct kib_fmr *fmr, int status)
{
- struct list_head zombies = LIST_HEAD_INIT(zombies);
+ LIST_HEAD(zombies);
struct kib_fmr_pool *fpo = fmr->fmr_pool;
struct kib_fmr_poolset *fps;
time64_t now = ktime_get_seconds();
fps = fpo->fpo_owner;
if (fpo->fpo_is_fmr) {
if (fmr->fmr_pfmr) {
- rc = ib_fmr_pool_unmap(fmr->fmr_pfmr);
- LASSERT(!rc);
+ ib_fmr_pool_unmap(fmr->fmr_pfmr);
fmr->fmr_pfmr = NULL;
}
tx_pages_mapped = 1;
}
- pfmr = ib_fmr_pool_map_phys(fpo->fmr.fpo_fmr_pool,
- pages, npages, iov);
+ pfmr = kib_fmr_pool_map(fpo->fmr.fpo_fmr_pool,
+ pages, npages, iov);
if (likely(!IS_ERR(pfmr))) {
fmr->fmr_key = is_rx ? pfmr->fmr->rkey
: pfmr->fmr->lkey;
#ifdef HAVE_IB_MAP_MR_SG
#ifdef HAVE_IB_MAP_MR_SG_5ARGS
n = ib_map_mr_sg(mr, tx->tx_frags,
- tx->tx_nfrags, NULL, PAGE_SIZE);
+ rd->rd_nfrags, NULL, PAGE_SIZE);
#else
n = ib_map_mr_sg(mr, tx->tx_frags,
- tx->tx_nfrags, PAGE_SIZE);
+ rd->rd_nfrags, PAGE_SIZE);
#endif
- if (unlikely(n != tx->tx_nfrags)) {
+ if (unlikely(n != rd->rd_nfrags)) {
CERROR("Failed to map mr %d/%d "
- "elements\n", n, tx->tx_nfrags);
+ "elements\n", n, rd->rd_nfrags);
return n < 0 ? n : -EINVAL;
}
spin_unlock(&fps->fps_lock);
CDEBUG(D_NET, "Another thread is allocating new "
"FMR pool, waiting for her to complete\n");
- schedule();
+ wait_var_event(fps, !fps->fps_increasing);
goto again;
}
rc = kiblnd_create_fmr_pool(fps, &fpo);
spin_lock(&fps->fps_lock);
fps->fps_increasing = 0;
+ wake_up_var(fps);
if (rc == 0) {
fps->fps_version++;
list_add_tail(&fpo->fpo_list, &fps->fps_pool_list);
struct kib_pool, po_list);
po->po_failed = 1;
- list_del(&po->po_list);
if (po->po_allocated == 0)
- list_add(&po->po_list, zombies);
+ list_move(&po->po_list, zombies);
else
- list_add(&po->po_list, &ps->ps_failed_pool_list);
+ list_move(&po->po_list, &ps->ps_failed_pool_list);
}
spin_unlock(&ps->ps_lock);
}
void
kiblnd_pool_free_node(struct kib_pool *pool, struct list_head *node)
{
- struct list_head zombies = LIST_HEAD_INIT(zombies);
+ LIST_HEAD(zombies);
struct kib_poolset *ps = pool->po_owner;
struct kib_pool *tmp;
time64_t now = ktime_get_seconds();
/* another thread is allocating a new pool */
spin_unlock(&ps->ps_lock);
trips++;
- CDEBUG(D_NET, "Another thread is allocating new "
- "%s pool, waiting %d HZs for her to complete."
- "trips = %d\n",
+ CDEBUG(D_NET,
+ "Another thread is allocating new %s pool, waiting %d jiffies for her to complete. trips = %d\n",
ps->ps_name, interval, trips);
- set_current_state(TASK_INTERRUPTIBLE);
- schedule_timeout(interval);
+ schedule_timeout_interruptible(interval);
if (interval < cfs_time_seconds(1))
interval *= 2;
if (tpo->tpo_tx_descs == NULL)
goto out;
- for (i = 0; i < pool->po_size; i++) {
+ for (i = 0; i < pool->po_size; i++) {
struct kib_tx *tx = &tpo->tpo_tx_descs[i];
int wrq_sge = *kiblnd_tunables.kib_wrq_sge;
list_del(&tx->tx_list);
- if (tx->tx_pages != NULL)
- LIBCFS_FREE(tx->tx_pages,
- LNET_MAX_IOV *
- sizeof(*tx->tx_pages));
- if (tx->tx_frags != NULL)
- LIBCFS_FREE(tx->tx_frags,
- (1 + IBLND_MAX_RDMA_FRAGS) *
- sizeof(*tx->tx_frags));
- if (tx->tx_wrq != NULL)
- LIBCFS_FREE(tx->tx_wrq,
- (1 + IBLND_MAX_RDMA_FRAGS) *
- sizeof(*tx->tx_wrq));
+ if (tx->tx_pages != NULL)
+ CFS_FREE_PTR_ARRAY(tx->tx_pages, LNET_MAX_IOV);
+ if (tx->tx_frags != NULL)
+ CFS_FREE_PTR_ARRAY(tx->tx_frags,
+ (1 + IBLND_MAX_RDMA_FRAGS));
+ if (tx->tx_wrq != NULL)
+ CFS_FREE_PTR_ARRAY(tx->tx_wrq,
+ (1 + IBLND_MAX_RDMA_FRAGS));
if (tx->tx_sge != NULL)
- LIBCFS_FREE(tx->tx_sge,
- (1 + IBLND_MAX_RDMA_FRAGS) * wrq_sge *
- sizeof(*tx->tx_sge));
- if (tx->tx_rd != NULL)
- LIBCFS_FREE(tx->tx_rd,
+ CFS_FREE_PTR_ARRAY(tx->tx_sge,
+ (1 + IBLND_MAX_RDMA_FRAGS) *
+ wrq_sge);
+ if (tx->tx_rd != NULL)
+ LIBCFS_FREE(tx->tx_rd,
offsetof(struct kib_rdma_desc,
- rd_frags[IBLND_MAX_RDMA_FRAGS]));
- }
+ rd_frags[IBLND_MAX_RDMA_FRAGS]));
+ }
- LIBCFS_FREE(tpo->tpo_tx_descs,
- pool->po_size * sizeof(struct kib_tx));
+ CFS_FREE_PTR_ARRAY(tpo->tpo_tx_descs, pool->po_size);
out:
- kiblnd_fini_pool(pool);
- LIBCFS_FREE(tpo, sizeof(struct kib_tx_pool));
+ kiblnd_fini_pool(pool);
+ CFS_FREE_PTR(tpo);
}
static int kiblnd_tx_pool_size(struct lnet_ni *ni, int ncpts)
npg = (size * IBLND_MSG_SIZE + PAGE_SIZE - 1) / PAGE_SIZE;
if (kiblnd_alloc_pages(&tpo->tpo_tx_pages, ps->ps_cpt, npg) != 0) {
CERROR("Can't allocate tx pages: %d\n", npg);
- LIBCFS_FREE(tpo, sizeof(struct kib_tx_pool));
+ CFS_FREE_PTR(tpo);
return -ENOMEM;
}
}
static int
+kiblnd_port_get_attr(struct kib_hca_dev *hdev)
+{
+ struct ib_port_attr *port_attr;
+ int rc;
+ unsigned long flags;
+ rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
+
+ LIBCFS_ALLOC(port_attr, sizeof(*port_attr));
+ if (port_attr == NULL) {
+ CDEBUG(D_NETERROR, "Out of memory\n");
+ return -ENOMEM;
+ }
+
+ rc = ib_query_port(hdev->ibh_ibdev, hdev->ibh_port, port_attr);
+
+ write_lock_irqsave(g_lock, flags);
+
+ if (rc == 0)
+ hdev->ibh_state = port_attr->state == IB_PORT_ACTIVE
+ ? IBLND_DEV_PORT_ACTIVE
+ : IBLND_DEV_PORT_DOWN;
+
+ write_unlock_irqrestore(g_lock, flags);
+ LIBCFS_FREE(port_attr, sizeof(*port_attr));
+
+ if (rc != 0) {
+ CDEBUG(D_NETERROR, "Failed to query IB port: %d\n", rc);
+ return rc;
+ }
+ return 0;
+}
+
+static inline void
+kiblnd_set_ni_fatal_on(struct kib_hca_dev *hdev, int val)
+{
+ struct kib_net *net;
+
+ /* for health check */
+ list_for_each_entry(net, &hdev->ibh_dev->ibd_nets, ibn_list) {
+ if (val)
+ CDEBUG(D_NETERROR, "Fatal device error for NI %s\n",
+ libcfs_nid2str(net->ibn_ni->ni_nid));
+ atomic_set(&net->ibn_ni->ni_fatal_error_on, val);
+ }
+}
+
+void
+kiblnd_event_handler(struct ib_event_handler *handler, struct ib_event *event)
+{
+ rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
+ struct kib_hca_dev *hdev;
+ unsigned long flags;
+
+ hdev = container_of(handler, struct kib_hca_dev, ibh_event_handler);
+
+ write_lock_irqsave(g_lock, flags);
+
+ switch (event->event) {
+ case IB_EVENT_DEVICE_FATAL:
+ CDEBUG(D_NET, "IB device fatal\n");
+ hdev->ibh_state = IBLND_DEV_FATAL;
+ kiblnd_set_ni_fatal_on(hdev, 1);
+ break;
+ case IB_EVENT_PORT_ACTIVE:
+ CDEBUG(D_NET, "IB port active\n");
+ if (event->element.port_num == hdev->ibh_port) {
+ hdev->ibh_state = IBLND_DEV_PORT_ACTIVE;
+ kiblnd_set_ni_fatal_on(hdev, 0);
+ }
+ break;
+ case IB_EVENT_PORT_ERR:
+ CDEBUG(D_NET, "IB port err\n");
+ if (event->element.port_num == hdev->ibh_port) {
+ hdev->ibh_state = IBLND_DEV_PORT_DOWN;
+ kiblnd_set_ni_fatal_on(hdev, 1);
+ }
+ break;
+ default:
+ break;
+ }
+ write_unlock_irqrestore(g_lock, flags);
+}
+
+static int
kiblnd_hdev_get_attr(struct kib_hca_dev *hdev)
{
struct ib_device_attr *dev_attr;
int rc = 0;
+ int rc2 = 0;
/* It's safe to assume a HCA can handle a page size
* matching that of the native system */
#endif
hdev->ibh_mr_size = dev_attr->max_mr_size;
+ hdev->ibh_max_qp_wr = dev_attr->max_qp_wr;
/* Setup device Memory Registration capabilities */
+#ifdef HAVE_IB_DEVICE_OPS
+ if (hdev->ibh_ibdev->ops.alloc_fmr &&
+ hdev->ibh_ibdev->ops.dealloc_fmr &&
+ hdev->ibh_ibdev->ops.map_phys_fmr &&
+ hdev->ibh_ibdev->ops.unmap_fmr) {
+#else
if (hdev->ibh_ibdev->alloc_fmr &&
hdev->ibh_ibdev->dealloc_fmr &&
hdev->ibh_ibdev->map_phys_fmr &&
hdev->ibh_ibdev->unmap_fmr) {
+#endif
LCONSOLE_INFO("Using FMR for registration\n");
hdev->ibh_dev->ibd_dev_caps |= IBLND_DEV_CAPS_FMR_ENABLED;
} else if (dev_attr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) {
rc = -ENOSYS;
}
- if (rc == 0 && hdev->ibh_mr_size == ~0ULL)
- hdev->ibh_mr_shift = 64;
- else if (rc != 0)
+ rc2 = kiblnd_port_get_attr(hdev);
+ if (rc2 != 0)
+ return rc2;
+
+ if (rc != 0)
rc = -EINVAL;
#ifndef HAVE_IB_DEVICE_ATTRS
void
kiblnd_hdev_destroy(struct kib_hca_dev *hdev)
{
+ if (hdev->ibh_event_handler.device != NULL)
+ ib_unregister_event_handler(&hdev->ibh_event_handler);
+
#ifdef HAVE_IB_GET_DMA_MR
kiblnd_hdev_cleanup_mrs(hdev);
#endif
}
static int
-kiblnd_dev_need_failover(struct kib_dev *dev)
+kiblnd_dev_need_failover(struct kib_dev *dev, struct net *ns)
{
struct rdma_cm_id *cmid;
struct sockaddr_in srcaddr;
*
* a. rdma_bind_addr(), it will conflict with listener cmid
* b. rdma_resolve_addr() to zero addr */
- cmid = kiblnd_rdma_create_id(kiblnd_dummy_callback, dev, RDMA_PS_TCP,
- IB_QPT_RC);
+ cmid = kiblnd_rdma_create_id(ns, kiblnd_dummy_callback, dev,
+ RDMA_PS_TCP, IB_QPT_RC);
if (IS_ERR(cmid)) {
rc = PTR_ERR(cmid);
CERROR("Failed to create cmid for failover: %d\n", rc);
}
int
-kiblnd_dev_failover(struct kib_dev *dev)
+kiblnd_dev_failover(struct kib_dev *dev, struct net *ns)
{
- struct list_head zombie_tpo = LIST_HEAD_INIT(zombie_tpo);
- struct list_head zombie_ppo = LIST_HEAD_INIT(zombie_ppo);
- struct list_head zombie_fpo = LIST_HEAD_INIT(zombie_fpo);
+ LIST_HEAD(zombie_tpo);
+ LIST_HEAD(zombie_ppo);
+ LIST_HEAD(zombie_fpo);
struct rdma_cm_id *cmid = NULL;
struct kib_hca_dev *hdev = NULL;
struct kib_hca_dev *old;
dev->ibd_can_failover ||
dev->ibd_hdev == NULL);
- rc = kiblnd_dev_need_failover(dev);
+ rc = kiblnd_dev_need_failover(dev, ns);
if (rc <= 0)
goto out;
rdma_destroy_id(cmid);
}
- cmid = kiblnd_rdma_create_id(kiblnd_cm_callback, dev, RDMA_PS_TCP,
- IB_QPT_RC);
+ cmid = kiblnd_rdma_create_id(ns, kiblnd_cm_callback, dev, RDMA_PS_TCP,
+ IB_QPT_RC);
if (IS_ERR(cmid)) {
rc = PTR_ERR(cmid);
CERROR("Failed to create cmid for failover: %d\n", rc);
hdev->ibh_dev = dev;
hdev->ibh_cmid = cmid;
hdev->ibh_ibdev = cmid->device;
+ hdev->ibh_port = cmid->port_num;
#ifdef HAVE_IB_ALLOC_PD_2ARGS
pd = ib_alloc_pd(cmid->device, 0);
}
#endif
+ INIT_IB_EVENT_HANDLER(&hdev->ibh_event_handler,
+ hdev->ibh_ibdev, kiblnd_event_handler);
+ ib_register_event_handler(&hdev->ibh_event_handler);
+
write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
old = dev->ibd_hdev;
LIBCFS_FREE(dev, sizeof(*dev));
}
-static struct kib_dev *
-kiblnd_create_dev(char *ifname)
-{
- struct net_device *netdev;
- struct kib_dev *dev = NULL;
- int flags;
- int rc;
-
- rtnl_lock();
- for_each_netdev(&init_net, netdev) {
- struct in_device *in_dev;
-
- if (strcmp(netdev->name, "lo") == 0) /* skip the loopback IF */
- continue;
-
- flags = dev_get_flags(netdev);
- if (!(flags & IFF_UP)) {
- CERROR("Can't query IPoIB interface %s: it's down\n",
- netdev->name);
- goto unlock;
- }
-
- in_dev = __in_dev_get_rtnl(netdev);
- if (!in_dev) {
- CWARN("Interface %s has no IPv4 status.\n",
- netdev->name);
- goto unlock;
- }
-
- for_ifa(in_dev)
- if (strcmp(ifname, ifa->ifa_label) == 0) {
- LIBCFS_ALLOC(dev, sizeof(*dev));
- if (!dev)
- goto unlock;
-
- dev->ibd_can_failover = !!(flags & IFF_MASTER);
- dev->ibd_ifip = ntohl(ifa->ifa_local);
-
- INIT_LIST_HEAD(&dev->ibd_nets);
- INIT_LIST_HEAD(&dev->ibd_list); /* not yet in kib_devs */
- INIT_LIST_HEAD(&dev->ibd_fail_list);
- break;
- }
- endfor_ifa(in_dev);
- }
- rtnl_unlock();
-
- if (!dev) {
- CERROR("Can't find any usable interfaces\n");
- return NULL;
- }
-
- if (dev->ibd_ifip == 0) {
- CERROR("Can't initialize device: no IP address\n");
- goto free_dev;
- }
- strcpy(&dev->ibd_ifname[0], ifname);
-
- /* initialize the device */
- rc = kiblnd_dev_failover(dev);
- if (rc != 0) {
- CERROR("Can't initialize device: %d\n", rc);
- goto free_dev;
- }
-
- list_add_tail(&dev->ibd_list, &kiblnd_data.kib_devs);
- return dev;
-unlock:
- rtnl_unlock();
-free_dev:
- LIBCFS_FREE(dev, sizeof(*dev));
- return NULL;
-}
-
static void
kiblnd_base_shutdown(void)
{
wake_up_all(&kiblnd_data.kib_connd_waitq);
wake_up_all(&kiblnd_data.kib_failover_waitq);
- i = 2;
- while (atomic_read(&kiblnd_data.kib_nthreads) != 0) {
- i++;
- /* power of 2? */
- CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET,
- "Waiting for %d threads to terminate\n",
- atomic_read(&kiblnd_data.kib_nthreads));
- set_current_state(TASK_UNINTERRUPTIBLE);
- schedule_timeout(cfs_time_seconds(1));
- }
-
- /* fall through */
+ wait_var_event_warning(&kiblnd_data.kib_nthreads,
+ !atomic_read(&kiblnd_data.kib_nthreads),
+ "Waiting for %d threads to terminate\n",
+ atomic_read(&kiblnd_data.kib_nthreads));
+ /* fall through */
case IBLND_INIT_NOTHING:
break;
}
- if (kiblnd_data.kib_peers != NULL) {
- LIBCFS_FREE(kiblnd_data.kib_peers,
- sizeof(struct list_head) *
- kiblnd_data.kib_peer_hash_size);
- }
+ if (kiblnd_data.kib_peers)
+ CFS_FREE_PTR_ARRAY(kiblnd_data.kib_peers,
+ kiblnd_data.kib_peer_hash_size);
if (kiblnd_data.kib_scheds != NULL)
cfs_percpt_free(kiblnd_data.kib_scheds);
{
struct kib_net *net = ni->ni_data;
rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
- int i;
- unsigned long flags;
+ unsigned long flags;
LASSERT(kiblnd_data.kib_init == IBLND_INIT_ALL);
default:
LBUG();
- case IBLND_INIT_ALL:
- /* nuke all existing peers within this net */
- kiblnd_del_peer(ni, LNET_NID_ANY);
+ case IBLND_INIT_ALL:
+ /* nuke all existing peers within this net */
+ kiblnd_del_peer(ni, LNET_NID_ANY);
/* Wait for all peer_ni state to clean up */
- i = 2;
- while (atomic_read(&net->ibn_npeers) != 0) {
- i++;
- /* power of 2? */
- CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET,
- "%s: waiting for %d peers to disconnect\n",
- libcfs_nid2str(ni->ni_nid),
- atomic_read(&net->ibn_npeers));
- set_current_state(TASK_UNINTERRUPTIBLE);
- schedule_timeout(cfs_time_seconds(1));
- }
+ wait_var_event_warning(&net->ibn_npeers,
+ atomic_read(&net->ibn_npeers) == 0,
+ "%s: waiting for %d peers to disconnect\n",
+ libcfs_nid2str(ni->ni_nid),
+ atomic_read(&net->ibn_npeers));
kiblnd_net_fini_pools(net);
list_del(&net->ibn_list);
write_unlock_irqrestore(g_lock, flags);
- /* fall through */
+ /* fall through */
case IBLND_INIT_NOTHING:
LASSERT (atomic_read(&net->ibn_nconns) == 0);
out:
if (list_empty(&kiblnd_data.kib_devs))
kiblnd_base_shutdown();
- return;
}
static int
-kiblnd_base_startup(void)
+kiblnd_base_startup(struct net *ns)
{
struct kib_sched_info *sched;
int rc;
LASSERT(kiblnd_data.kib_init == IBLND_INIT_NOTHING);
- try_module_get(THIS_MODULE);
+ if (!try_module_get(THIS_MODULE))
+ goto failed;
+
memset(&kiblnd_data, 0, sizeof(kiblnd_data)); /* zero pointers, flags etc */
rwlock_init(&kiblnd_data.kib_global_lock);
INIT_LIST_HEAD(&kiblnd_data.kib_failed_devs);
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);
+ CFS_ALLOC_PTR_ARRAY(kiblnd_data.kib_peers,
+ kiblnd_data.kib_peer_hash_size);
if (kiblnd_data.kib_peers == NULL)
goto failed;
}
if (*kiblnd_tunables.kib_dev_failover != 0)
- rc = kiblnd_thread_start(kiblnd_failover_thread, NULL,
+ rc = kiblnd_thread_start(kiblnd_failover_thread, ns,
"kiblnd_failover");
if (rc != 0) {
return rc;
}
-static int
-kiblnd_dev_start_threads(struct kib_dev *dev, int newdev, u32 *cpts, int ncpts)
+static int kiblnd_dev_start_threads(struct kib_dev *dev, bool newdev, u32 *cpts,
+ int ncpts)
{
int cpt;
int rc;
{
struct kib_dev *alias = NULL;
struct kib_dev *dev;
- char *colon;
- char *colon2;
+ char *colon;
+ char *colon2;
colon = strchr(ifname, ':');
list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
static int
kiblnd_startup(struct lnet_ni *ni)
{
- char *ifname;
+ char *ifname = NULL;
+ struct lnet_inetdev *ifaces = NULL;
struct kib_dev *ibdev = NULL;
- struct kib_net *net;
- unsigned long flags;
- int rc;
- int newdev;
- int node_id;
+ struct kib_net *net = NULL;
+ unsigned long flags;
+ int rc;
+ int i;
+ bool newdev;
- LASSERT (ni->ni_net->net_lnd == &the_o2iblnd);
+ LASSERT(ni->ni_net->net_lnd == &the_o2iblnd);
- if (kiblnd_data.kib_init == IBLND_INIT_NOTHING) {
- rc = kiblnd_base_startup();
- if (rc != 0)
- return rc;
- }
+ if (kiblnd_data.kib_init == IBLND_INIT_NOTHING) {
+ rc = kiblnd_base_startup(ni->ni_net_ns);
+ if (rc != 0)
+ return rc;
+ }
- LIBCFS_ALLOC(net, sizeof(*net));
- ni->ni_data = net;
- if (net == NULL)
- goto failed;
+ LIBCFS_ALLOC(net, sizeof(*net));
+ ni->ni_data = net;
+ if (net == NULL) {
+ rc = -ENOMEM;
+ goto failed;
+ }
+ net->ibn_ni = ni;
net->ibn_incarnation = ktime_get_real_ns() / NSEC_PER_USEC;
kiblnd_tunables_setup(ni);
*/
if (ni->ni_interfaces[0] != NULL) {
/* Use the IPoIB interface specified in 'networks=' */
-
- CLASSERT(LNET_INTERFACES_NUM > 1);
if (ni->ni_interfaces[1] != NULL) {
- CERROR("Multiple interfaces not supported\n");
+ CERROR("ko2iblnd: Multiple interfaces not supported\n");
+ rc = -EINVAL;
goto failed;
}
ifname = *kiblnd_tunables.kib_default_ipif;
}
- if (strlen(ifname) >= sizeof(ibdev->ibd_ifname)) {
- CERROR("IPoIB interface name too long: %s\n", ifname);
- goto failed;
- }
+ if (strlen(ifname) >= sizeof(ibdev->ibd_ifname)) {
+ CERROR("IPoIB interface name too long: %s\n", ifname);
+ rc = -E2BIG;
+ goto failed;
+ }
- ibdev = kiblnd_dev_search(ifname);
+ rc = lnet_inet_enumerate(&ifaces, ni->ni_net_ns);
+ if (rc < 0)
+ goto failed;
+
+ for (i = 0; i < rc; i++) {
+ if (strcmp(ifname, ifaces[i].li_name) == 0)
+ break;
+ }
+
+ if (i == rc) {
+ CERROR("ko2iblnd: No matching interfaces\n");
+ rc = -ENOENT;
+ goto failed;
+ }
+ ibdev = kiblnd_dev_search(ifname);
newdev = ibdev == NULL;
/* hmm...create kib_dev even for alias */
- if (ibdev == NULL || strcmp(&ibdev->ibd_ifname[0], ifname) != 0)
- ibdev = kiblnd_create_dev(ifname);
+ if (ibdev == NULL || strcmp(&ibdev->ibd_ifname[0], ifname) != 0) {
+ LIBCFS_ALLOC(ibdev, sizeof(*ibdev));
+ if (!ibdev) {
+ rc = -ENOMEM;
+ goto failed;
+ }
- if (ibdev == NULL)
- goto failed;
+ ibdev->ibd_ifip = ifaces[i].li_ipaddr;
+ strlcpy(ibdev->ibd_ifname, ifaces[i].li_name,
+ sizeof(ibdev->ibd_ifname));
+ ibdev->ibd_can_failover = !!(ifaces[i].li_flags & IFF_MASTER);
+
+ INIT_LIST_HEAD(&ibdev->ibd_nets);
+ INIT_LIST_HEAD(&ibdev->ibd_list); /* not yet in kib_devs */
+ INIT_LIST_HEAD(&ibdev->ibd_fail_list);
+
+ /* initialize the device */
+ rc = kiblnd_dev_failover(ibdev, ni->ni_net_ns);
+ if (rc) {
+ CERROR("ko2iblnd: Can't initialize device: rc = %d\n",
+ rc);
+ goto failed;
+ }
- node_id = dev_to_node(ibdev->ibd_hdev->ibh_ibdev->dma_device);
- ni->ni_dev_cpt = cfs_cpt_of_node(lnet_cpt_table(), node_id);
+ list_add_tail(&ibdev->ibd_list, &kiblnd_data.kib_devs);
+ }
net->ibn_dev = ibdev;
ni->ni_nid = LNET_MKNID(LNET_NIDNET(ni->ni_nid), ibdev->ibd_ifip);
- rc = kiblnd_dev_start_threads(ibdev, newdev,
- ni->ni_cpts, ni->ni_ncpts);
+ ni->ni_dev_cpt = ifaces[i].li_cpt;
+
+ rc = kiblnd_dev_start_threads(ibdev, newdev, ni->ni_cpts, ni->ni_ncpts);
if (rc != 0)
goto failed;
rc = kiblnd_net_init_pools(net, ni, ni->ni_cpts, ni->ni_ncpts);
- if (rc != 0) {
- CERROR("Failed to initialize NI pools: %d\n", rc);
- goto failed;
- }
+ if (rc != 0) {
+ CERROR("Failed to initialize NI pools: %d\n", rc);
+ goto failed;
+ }
write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
ibdev->ibd_nnets++;
list_add_tail(&net->ibn_list, &ibdev->ibd_nets);
+ /* for health check */
+ if (ibdev->ibd_hdev->ibh_state == IBLND_DEV_PORT_DOWN)
+ kiblnd_set_ni_fatal_on(ibdev->ibd_hdev, 1);
write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
- net->ibn_init = IBLND_INIT_ALL;
+ net->ibn_init = IBLND_INIT_ALL;
- return 0;
+ return 0;
failed:
if (net != NULL && net->ibn_dev == NULL && ibdev != NULL)
- kiblnd_destroy_dev(ibdev);
+ kiblnd_destroy_dev(ibdev);
- kiblnd_shutdown(ni);
+ kfree(ifaces);
+ kiblnd_shutdown(ni);
- CDEBUG(D_NET, "kiblnd_startup failed\n");
- return -ENETDOWN;
+ CDEBUG(D_NET, "Configuration of device %s failed: rc = %d\n",
+ ifname ? ifname : "", rc);
+
+ return -ENETDOWN;
}
-static struct lnet_lnd the_o2iblnd = {
+static const struct lnet_lnd 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,
};
{
int rc;
- CLASSERT(sizeof(struct kib_msg) <= IBLND_MSG_SIZE);
- CLASSERT(offsetof(struct kib_msg,
- ibm_u.get.ibgm_rd.rd_frags[IBLND_MAX_RDMA_FRAGS]) <=
- IBLND_MSG_SIZE);
- CLASSERT(offsetof(struct kib_msg,
- ibm_u.putack.ibpam_rd.rd_frags[IBLND_MAX_RDMA_FRAGS])
- <= IBLND_MSG_SIZE);
+ BUILD_BUG_ON(sizeof(struct kib_msg) > IBLND_MSG_SIZE);
+ BUILD_BUG_ON(offsetof(struct kib_msg,
+ ibm_u.get.ibgm_rd.rd_frags[IBLND_MAX_RDMA_FRAGS]) >
+ IBLND_MSG_SIZE);
+ BUILD_BUG_ON(offsetof(struct kib_msg,
+ ibm_u.putack.ibpam_rd.rd_frags[IBLND_MAX_RDMA_FRAGS]) >
+ IBLND_MSG_SIZE);
rc = kiblnd_tunables_init();
if (rc != 0)
git://git.whamcloud.com / fs / lustre-release.git / blobdiff