* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2012, Intel Corporation.
+ * Copyright (c) 2012, 2014, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#include "o2iblnd.h"
-void
+static void kiblnd_peer_alive(kib_peer_t *peer);
+static void kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error);
+static void kiblnd_check_sends(kib_conn_t *conn);
+static void kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx,
+ int type, int body_nob);
+static int kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type,
+ int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie);
+static void kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn);
+static void kiblnd_queue_tx(kib_tx_t *tx, kib_conn_t *conn);
+
+static void
kiblnd_tx_done (lnet_ni_t *ni, kib_tx_t *tx)
{
lnet_msg_t *lntmsg[2];
}
void
-kiblnd_txlist_done (lnet_ni_t *ni, cfs_list_t *txlist, int status)
+kiblnd_txlist_done(lnet_ni_t *ni, struct list_head *txlist, int status)
{
- kib_tx_t *tx;
+ kib_tx_t *tx;
- while (!cfs_list_empty (txlist)) {
- tx = cfs_list_entry (txlist->next, kib_tx_t, tx_list);
+ while (!list_empty(txlist)) {
+ tx = list_entry(txlist->next, kib_tx_t, tx_list);
- cfs_list_del(&tx->tx_list);
- /* complete now */
- tx->tx_waiting = 0;
- tx->tx_status = status;
- kiblnd_tx_done(ni, tx);
- }
+ list_del(&tx->tx_list);
+ /* complete now */
+ tx->tx_waiting = 0;
+ tx->tx_status = status;
+ kiblnd_tx_done(ni, tx);
+ }
}
-kib_tx_t *
+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;
- cfs_list_t *node;
+ struct list_head *node;
kib_tx_t *tx;
kib_tx_poolset_t *tps;
return tx;
}
-void
+static void
kiblnd_drop_rx(kib_rx_t *rx)
{
kib_conn_t *conn = rx->rx_conn;
rx->rx_nob = -1; /* flag posted */
- rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq);
- if (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;
- }
+ /* 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 (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;
+ if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */
+ goto out;
- if (rc != 0) {
- kiblnd_close_conn(conn, rc);
- kiblnd_drop_rx(rx); /* No more posts for this rx */
- return rc;
- }
+ if (unlikely(rc != 0)) {
+ kiblnd_close_conn(conn, rc);
+ kiblnd_drop_rx(rx); /* No more posts for this rx */
+ goto out;
+ }
- if (credit == IBLND_POSTRX_NO_CREDIT)
- return 0;
+ if (credit == IBLND_POSTRX_NO_CREDIT)
+ goto out;
spin_lock(&conn->ibc_lock);
if (credit == IBLND_POSTRX_PEER_CREDIT)
spin_unlock(&conn->ibc_lock);
kiblnd_check_sends(conn);
- return 0;
+out:
+ kiblnd_conn_decref(conn);
+ return rc;
}
-kib_tx_t *
+static kib_tx_t *
kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie)
{
- cfs_list_t *tmp;
+ struct list_head *tmp;
- cfs_list_for_each(tmp, &conn->ibc_active_txs) {
- kib_tx_t *tx = cfs_list_entry(tmp, kib_tx_t, tx_list);
+ list_for_each(tmp, &conn->ibc_active_txs) {
+ kib_tx_t *tx = list_entry(tmp, kib_tx_t, tx_list);
- LASSERT (!tx->tx_queued);
- LASSERT (tx->tx_sending != 0 || tx->tx_waiting);
+ LASSERT(!tx->tx_queued);
+ LASSERT(tx->tx_sending != 0 || tx->tx_waiting);
- if (tx->tx_cookie != cookie)
- continue;
+ if (tx->tx_cookie != cookie)
+ continue;
- if (tx->tx_waiting &&
- tx->tx_msg->ibm_type == txtype)
- return tx;
+ if (tx->tx_waiting &&
+ tx->tx_msg->ibm_type == txtype)
+ return tx;
- CWARN("Bad completion: %swaiting, type %x (wanted %x)\n",
- tx->tx_waiting ? "" : "NOT ",
- tx->tx_msg->ibm_type, txtype);
- }
- return NULL;
+ CWARN("Bad completion: %swaiting, type %x (wanted %x)\n",
+ tx->tx_waiting ? "" : "NOT ",
+ tx->tx_msg->ibm_type, txtype);
+ }
+ return NULL;
}
-void
+static void
kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie)
{
kib_tx_t *tx;
idle = !tx->tx_queued && (tx->tx_sending == 0);
if (idle)
- cfs_list_del(&tx->tx_list);
+ list_del(&tx->tx_list);
spin_unlock(&conn->ibc_lock);
kiblnd_tx_done(ni, tx);
}
-void
+static void
kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie)
{
lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
kiblnd_queue_tx(tx, conn);
}
-void
+static void
kiblnd_handle_rx (kib_rx_t *rx)
{
kib_msg_t *msg = rx->rx_msg;
tx = kiblnd_find_waiting_tx_locked(conn, IBLND_MSG_PUT_REQ,
msg->ibm_u.putack.ibpam_src_cookie);
if (tx != NULL)
- cfs_list_del(&tx->tx_list);
+ list_del(&tx->tx_list);
spin_unlock(&conn->ibc_lock);
if (tx == NULL) {
kiblnd_post_rx(rx, post_credit);
}
-void
+static void
kiblnd_rx_complete (kib_rx_t *rx, int status, int nob)
{
kib_msg_t *msg = rx->rx_msg;
write_lock_irqsave(g_lock, flags);
/* must check holding global lock to eliminate race */
if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
- cfs_list_add_tail(&rx->rx_list, &conn->ibc_early_rxs);
+ list_add_tail(&rx->rx_list, &conn->ibc_early_rxs);
write_unlock_irqrestore(g_lock, flags);
return;
}
kiblnd_drop_rx(rx); /* Don't re-post rx. */
}
-struct page *
+static struct page *
kiblnd_kvaddr_to_page (unsigned long vaddr)
{
struct page *page;
}
-int
+static int
kiblnd_setup_rd_iov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
unsigned int niov, struct iovec *iov, int offset, int nob)
{
return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
}
-int
+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)
{
return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
}
-int
+static int
kiblnd_post_tx_locked (kib_conn_t *conn, kib_tx_t *tx, int credit)
+__must_hold(&conn->ibc_lock)
{
kib_msg_t *msg = tx->tx_msg;
kib_peer_t *peer = conn->ibc_peer;
}
/* NB don't drop ibc_lock before bumping tx_sending */
- cfs_list_del(&tx->tx_list);
+ list_del(&tx->tx_list);
tx->tx_queued = 0;
if (msg->ibm_type == IBLND_MSG_NOOP &&
* tx_sending is non-zero if we've not done the tx_complete()
* from the first send; hence the ++ rather than = below. */
tx->tx_sending++;
- cfs_list_add(&tx->tx_list, &conn->ibc_active_txs);
+ list_add(&tx->tx_list, &conn->ibc_active_txs);
/* I'm still holding ibc_lock! */
if (conn->ibc_state != IBLND_CONN_ESTABLISHED) {
done = (tx->tx_sending == 0);
if (done)
- cfs_list_del(&tx->tx_list);
+ list_del(&tx->tx_list);
spin_unlock(&conn->ibc_lock);
return -EIO;
}
-void
+static void
kiblnd_check_sends (kib_conn_t *conn)
{
int ver = conn->ibc_version;
LASSERT (conn->ibc_reserved_credits >= 0);
while (conn->ibc_reserved_credits > 0 &&
- !cfs_list_empty(&conn->ibc_tx_queue_rsrvd)) {
- tx = cfs_list_entry(conn->ibc_tx_queue_rsrvd.next,
+ !list_empty(&conn->ibc_tx_queue_rsrvd)) {
+ tx = list_entry(conn->ibc_tx_queue_rsrvd.next,
kib_tx_t, tx_list);
- cfs_list_del(&tx->tx_list);
- cfs_list_add_tail(&tx->tx_list, &conn->ibc_tx_queue);
+ list_del(&tx->tx_list);
+ list_add_tail(&tx->tx_list, &conn->ibc_tx_queue);
conn->ibc_reserved_credits--;
}
for (;;) {
int credit;
- if (!cfs_list_empty(&conn->ibc_tx_queue_nocred)) {
+ if (!list_empty(&conn->ibc_tx_queue_nocred)) {
credit = 0;
- tx = cfs_list_entry(conn->ibc_tx_queue_nocred.next,
+ tx = list_entry(conn->ibc_tx_queue_nocred.next,
kib_tx_t, tx_list);
- } else if (!cfs_list_empty(&conn->ibc_tx_noops)) {
+ } else if (!list_empty(&conn->ibc_tx_noops)) {
LASSERT (!IBLND_OOB_CAPABLE(ver));
credit = 1;
- tx = cfs_list_entry(conn->ibc_tx_noops.next,
+ tx = list_entry(conn->ibc_tx_noops.next,
kib_tx_t, tx_list);
- } else if (!cfs_list_empty(&conn->ibc_tx_queue)) {
+ } else if (!list_empty(&conn->ibc_tx_queue)) {
credit = 1;
- tx = cfs_list_entry(conn->ibc_tx_queue.next,
+ tx = list_entry(conn->ibc_tx_queue.next,
kib_tx_t, tx_list);
} else
break;
kiblnd_conn_decref(conn); /* ...until here */
}
-void
+static void
kiblnd_tx_complete (kib_tx_t *tx, int status)
{
int failed = (status != IB_WC_SUCCESS);
!tx->tx_waiting && /* Not waiting for peer */
!tx->tx_queued; /* Not re-queued (PUT_DONE) */
if (idle)
- cfs_list_del(&tx->tx_list);
+ list_del(&tx->tx_list);
kiblnd_conn_addref(conn); /* 1 ref for me.... */
kiblnd_conn_decref(conn); /* ...until here */
}
-void
+static 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;
tx->tx_nwrq++;
}
-int
-kiblnd_init_rdma (kib_conn_t *conn, kib_tx_t *tx, int type,
- int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie)
+static 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;
return rc;
}
-void
-kiblnd_queue_tx_locked (kib_tx_t *tx, kib_conn_t *conn)
+static void
+kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn)
{
- cfs_list_t *q;
+ struct list_head *q;
- LASSERT (tx->tx_nwrq > 0); /* work items set up */
- LASSERT (!tx->tx_queued); /* not queued for sending already */
- LASSERT (conn->ibc_state >= IBLND_CONN_ESTABLISHED);
+ LASSERT(tx->tx_nwrq > 0); /* work items set up */
+ LASSERT(!tx->tx_queued); /* not queued for sending already */
+ LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
tx->tx_queued = 1;
- tx->tx_deadline = jiffies + (*kiblnd_tunables.kib_timeout * HZ);
+ tx->tx_deadline = jiffies +
+ msecs_to_jiffies(*kiblnd_tunables.kib_timeout *
+ MSEC_PER_SEC);
if (tx->tx_conn == NULL) {
kiblnd_conn_addref(conn);
break;
}
- cfs_list_add_tail(&tx->tx_list, q);
+ list_add_tail(&tx->tx_list, q);
}
-void
+static void
kiblnd_queue_tx (kib_tx_t *tx, kib_conn_t *conn)
{
spin_lock(&conn->ibc_lock);
return rc;
}
-void
+static void
kiblnd_connect_peer (kib_peer_t *peer)
{
struct rdma_cm_id *cmid;
}
LASSERT (cmid->device != NULL);
- CDEBUG(D_NET, "%s: connection bound to %s:%u.%u.%u.%u:%s\n",
+ CDEBUG(D_NET, "%s: connection bound to %s:%pI4h:%s\n",
libcfs_nid2str(peer->ibp_nid), dev->ibd_ifname,
- HIPQUAD(dev->ibd_ifip), cmid->device->name);
+ &dev->ibd_ifip, cmid->device->name);
return;
read_lock_irqsave(g_lock, flags);
peer = kiblnd_find_peer_locked(nid);
- if (peer != NULL && !cfs_list_empty(&peer->ibp_conns)) {
+ if (peer != NULL && !list_empty(&peer->ibp_conns)) {
/* Found a peer with an established connection */
conn = kiblnd_get_conn_locked(peer);
kiblnd_conn_addref(conn); /* 1 ref for me... */
peer = kiblnd_find_peer_locked(nid);
if (peer != NULL) {
- if (cfs_list_empty(&peer->ibp_conns)) {
+ if (list_empty(&peer->ibp_conns)) {
/* found a peer, but it's still connecting... */
LASSERT (peer->ibp_connecting != 0 ||
peer->ibp_accepting != 0);
if (tx != NULL)
- cfs_list_add_tail(&tx->tx_list,
+ list_add_tail(&tx->tx_list,
&peer->ibp_tx_queue);
write_unlock_irqrestore(g_lock, flags);
} else {
peer2 = kiblnd_find_peer_locked(nid);
if (peer2 != NULL) {
- if (cfs_list_empty(&peer2->ibp_conns)) {
+ if (list_empty(&peer2->ibp_conns)) {
/* found a peer, but it's still connecting... */
LASSERT (peer2->ibp_connecting != 0 ||
peer2->ibp_accepting != 0);
if (tx != NULL)
- cfs_list_add_tail(&tx->tx_list,
+ list_add_tail(&tx->tx_list,
&peer2->ibp_tx_queue);
write_unlock_irqrestore(g_lock, flags);
} else {
LASSERT (((kib_net_t *)ni->ni_data)->ibn_shutdown == 0);
if (tx != NULL)
- cfs_list_add_tail(&tx->tx_list, &peer->ibp_tx_queue);
+ list_add_tail(&tx->tx_list, &peer->ibp_tx_queue);
kiblnd_peer_addref(peer);
- cfs_list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
+ list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
write_unlock_irqrestore(g_lock, flags);
return 0;
}
-void
+static void
kiblnd_reply (lnet_ni_t *ni, kib_rx_t *rx, lnet_msg_t *lntmsg)
{
lnet_process_id_t target = lntmsg->msg_target;
return 0;
}
-void
+static void
kiblnd_thread_fini (void)
{
atomic_dec (&kiblnd_data.kib_nthreads);
}
-void
+static void
kiblnd_peer_alive (kib_peer_t *peer)
{
/* This is racy, but everyone's only writing cfs_time_current() */
smp_mb();
}
-void
+static void
kiblnd_peer_notify (kib_peer_t *peer)
{
int error = 0;
read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
- if (cfs_list_empty(&peer->ibp_conns) &&
+ if (list_empty(&peer->ibp_conns) &&
peer->ibp_accepting == 0 &&
peer->ibp_connecting == 0 &&
peer->ibp_error != 0) {
return; /* already being handled */
if (error == 0 &&
- cfs_list_empty(&conn->ibc_tx_noops) &&
- cfs_list_empty(&conn->ibc_tx_queue) &&
- cfs_list_empty(&conn->ibc_tx_queue_rsrvd) &&
- cfs_list_empty(&conn->ibc_tx_queue_nocred) &&
- cfs_list_empty(&conn->ibc_active_txs)) {
+ list_empty(&conn->ibc_tx_noops) &&
+ list_empty(&conn->ibc_tx_queue) &&
+ list_empty(&conn->ibc_tx_queue_rsrvd) &&
+ list_empty(&conn->ibc_tx_queue_nocred) &&
+ list_empty(&conn->ibc_active_txs)) {
CDEBUG(D_NET, "closing conn to %s\n",
libcfs_nid2str(peer->ibp_nid));
} else {
CNETERR("Closing conn to %s: error %d%s%s%s%s%s\n",
libcfs_nid2str(peer->ibp_nid), error,
- cfs_list_empty(&conn->ibc_tx_queue) ? "" : "(sending)",
- cfs_list_empty(&conn->ibc_tx_noops) ? "" : "(sending_noops)",
- cfs_list_empty(&conn->ibc_tx_queue_rsrvd) ? "" : "(sending_rsrvd)",
- cfs_list_empty(&conn->ibc_tx_queue_nocred) ? "" : "(sending_nocred)",
- cfs_list_empty(&conn->ibc_active_txs) ? "" : "(waiting)");
+ list_empty(&conn->ibc_tx_queue) ? "" : "(sending)",
+ list_empty(&conn->ibc_tx_noops) ? "" : "(sending_noops)",
+ list_empty(&conn->ibc_tx_queue_rsrvd) ?
+ "" : "(sending_rsrvd)",
+ list_empty(&conn->ibc_tx_queue_nocred) ?
+ "" : "(sending_nocred)",
+ list_empty(&conn->ibc_active_txs) ? "" : "(waiting)");
}
dev = ((kib_net_t *)peer->ibp_ni->ni_data)->ibn_dev;
- cfs_list_del(&conn->ibc_list);
+ list_del(&conn->ibc_list);
/* connd (see below) takes over ibc_list's ref */
- if (cfs_list_empty (&peer->ibp_conns) && /* no more conns */
+ if (list_empty(&peer->ibp_conns) && /* no more conns */
kiblnd_peer_active(peer)) { /* still in peer table */
kiblnd_unlink_peer_locked(peer);
if (error != 0 &&
kiblnd_dev_can_failover(dev)) {
- cfs_list_add_tail(&dev->ibd_fail_list,
+ list_add_tail(&dev->ibd_fail_list,
&kiblnd_data.kib_failed_devs);
wake_up(&kiblnd_data.kib_failover_waitq);
}
spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
- cfs_list_add_tail(&conn->ibc_list, &kiblnd_data.kib_connd_conns);
+ list_add_tail(&conn->ibc_list, &kiblnd_data.kib_connd_conns);
wake_up(&kiblnd_data.kib_connd_waitq);
spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
}
-void
+static void
kiblnd_handle_early_rxs(kib_conn_t *conn)
{
unsigned long flags;
LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
- while (!cfs_list_empty(&conn->ibc_early_rxs)) {
- rx = cfs_list_entry(conn->ibc_early_rxs.next,
+ while (!list_empty(&conn->ibc_early_rxs)) {
+ rx = list_entry(conn->ibc_early_rxs.next,
kib_rx_t, rx_list);
- cfs_list_del(&rx->rx_list);
+ list_del(&rx->rx_list);
write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
kiblnd_handle_rx(rx);
write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
}
-void
-kiblnd_abort_txs(kib_conn_t *conn, cfs_list_t *txs)
+static void
+kiblnd_abort_txs(kib_conn_t *conn, struct list_head *txs)
{
- CFS_LIST_HEAD (zombies);
- cfs_list_t *tmp;
- cfs_list_t *nxt;
- kib_tx_t *tx;
+ struct list_head zombies = LIST_HEAD_INIT(zombies);
+ struct list_head *tmp;
+ struct list_head *nxt;
+ kib_tx_t *tx;
spin_lock(&conn->ibc_lock);
- cfs_list_for_each_safe (tmp, nxt, txs) {
- tx = cfs_list_entry (tmp, kib_tx_t, tx_list);
+ list_for_each_safe(tmp, nxt, txs) {
+ tx = list_entry(tmp, kib_tx_t, tx_list);
- if (txs == &conn->ibc_active_txs) {
- LASSERT (!tx->tx_queued);
- LASSERT (tx->tx_waiting ||
- tx->tx_sending != 0);
- } else {
- LASSERT (tx->tx_queued);
- }
+ if (txs == &conn->ibc_active_txs) {
+ LASSERT(!tx->tx_queued);
+ LASSERT(tx->tx_waiting ||
+ tx->tx_sending != 0);
+ } else {
+ LASSERT(tx->tx_queued);
+ }
- tx->tx_status = -ECONNABORTED;
- tx->tx_waiting = 0;
+ tx->tx_status = -ECONNABORTED;
+ tx->tx_waiting = 0;
- if (tx->tx_sending == 0) {
- tx->tx_queued = 0;
- cfs_list_del (&tx->tx_list);
- cfs_list_add (&tx->tx_list, &zombies);
- }
- }
+ if (tx->tx_sending == 0) {
+ tx->tx_queued = 0;
+ list_del(&tx->tx_list);
+ list_add(&tx->tx_list, &zombies);
+ }
+ }
spin_unlock(&conn->ibc_lock);
kiblnd_txlist_done(conn->ibc_peer->ibp_ni, &zombies, -ECONNABORTED);
}
-void
+static void
kiblnd_finalise_conn (kib_conn_t *conn)
{
LASSERT (!in_interrupt());
kiblnd_handle_early_rxs(conn);
}
-void
-kiblnd_peer_connect_failed (kib_peer_t *peer, int active, int error)
+static void
+kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error)
{
- CFS_LIST_HEAD (zombies);
- unsigned long flags;
+ struct list_head zombies = LIST_HEAD_INIT(zombies);
+ unsigned long flags;
LASSERT (error != 0);
LASSERT (!in_interrupt());
}
if (peer->ibp_connecting != 0 ||
- peer->ibp_accepting != 0) {
- /* another connection attempt under way... */
+ peer->ibp_accepting != 0) {
+ /* another connection attempt under way... */
write_unlock_irqrestore(&kiblnd_data.kib_global_lock,
- flags);
- return;
- }
+ flags);
+ return;
+ }
- if (cfs_list_empty(&peer->ibp_conns)) {
- /* Take peer's blocked transmits to complete with error */
- cfs_list_add(&zombies, &peer->ibp_tx_queue);
- cfs_list_del_init(&peer->ibp_tx_queue);
+ if (list_empty(&peer->ibp_conns)) {
+ /* Take peer's blocked transmits to complete with error */
+ list_add(&zombies, &peer->ibp_tx_queue);
+ list_del_init(&peer->ibp_tx_queue);
- if (kiblnd_peer_active(peer))
- kiblnd_unlink_peer_locked(peer);
+ if (kiblnd_peer_active(peer))
+ kiblnd_unlink_peer_locked(peer);
- peer->ibp_error = error;
- } else {
- /* Can't have blocked transmits if there are connections */
- LASSERT (cfs_list_empty(&peer->ibp_tx_queue));
- }
+ peer->ibp_error = error;
+ } else {
+ /* Can't have blocked transmits if there are connections */
+ LASSERT(list_empty(&peer->ibp_tx_queue));
+ }
write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
- kiblnd_peer_notify(peer);
+ kiblnd_peer_notify(peer);
- if (cfs_list_empty (&zombies))
- return;
+ if (list_empty(&zombies))
+ return;
- CNETERR("Deleting messages for %s: connection failed\n",
- libcfs_nid2str(peer->ibp_nid));
+ CNETERR("Deleting messages for %s: connection failed\n",
+ libcfs_nid2str(peer->ibp_nid));
- kiblnd_txlist_done(peer->ibp_ni, &zombies, -EHOSTUNREACH);
+ kiblnd_txlist_done(peer->ibp_ni, &zombies, -EHOSTUNREACH);
}
-void
+static void
kiblnd_connreq_done(kib_conn_t *conn, int status)
{
- kib_peer_t *peer = conn->ibc_peer;
- kib_tx_t *tx;
- cfs_list_t txs;
- unsigned long flags;
- int active;
+ kib_peer_t *peer = conn->ibc_peer;
+ kib_tx_t *tx;
+ struct list_head txs;
+ unsigned long flags;
+ int active;
active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
kiblnd_set_conn_state(conn, IBLND_CONN_ESTABLISHED);
kiblnd_peer_alive(peer);
- /* Add conn to peer's list and nuke any dangling conns from a different
- * peer instance... */
- kiblnd_conn_addref(conn); /* +1 ref for ibc_list */
- cfs_list_add(&conn->ibc_list, &peer->ibp_conns);
- if (active)
- peer->ibp_connecting--;
- else
- peer->ibp_accepting--;
+ /* Add conn to peer's list and nuke any dangling conns from a different
+ * peer instance... */
+ kiblnd_conn_addref(conn); /* +1 ref for ibc_list */
+ list_add(&conn->ibc_list, &peer->ibp_conns);
+ if (active)
+ peer->ibp_connecting--;
+ else
+ peer->ibp_accepting--;
if (peer->ibp_version == 0) {
peer->ibp_version = conn->ibc_version;
peer->ibp_incarnation = conn->ibc_incarnation;
}
- /* grab pending txs while I have the lock */
- cfs_list_add(&txs, &peer->ibp_tx_queue);
- cfs_list_del_init(&peer->ibp_tx_queue);
+ /* grab pending txs while I have the lock */
+ list_add(&txs, &peer->ibp_tx_queue);
+ list_del_init(&peer->ibp_tx_queue);
if (!kiblnd_peer_active(peer) || /* peer has been deleted */
conn->ibc_comms_error != 0) { /* error has happened already */
/* Schedule blocked txs */
spin_lock(&conn->ibc_lock);
- while (!cfs_list_empty(&txs)) {
- tx = cfs_list_entry(txs.next, kib_tx_t, tx_list);
- cfs_list_del(&tx->tx_list);
+ while (!list_empty(&txs)) {
+ tx = list_entry(txs.next, kib_tx_t, tx_list);
+ list_del(&tx->tx_list);
kiblnd_queue_tx_locked(tx, conn);
}
kiblnd_handle_early_rxs(conn);
}
-void
+static void
kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej)
{
int rc;
CWARN("Error %d sending reject\n", rc);
}
-int
-kiblnd_passive_connect (struct rdma_cm_id *cmid, void *priv, int priv_nob)
+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;
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 (%u.%u.%u.%u:%hu) is not privileged\n",
- HIPQUAD(ip), ntohs(peer_addr->sin_port));
+ __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;
}
if (ni == NULL || /* no matching net */
ni->ni_nid != reqmsg->ibm_dstnid || /* right NET, wrong NID! */
net->ibn_dev != ibdev) { /* wrong device */
- CERROR("Can't accept %s on %s (%s:%d:%u.%u.%u.%u): "
+ CERROR("Can't accept %s on %s (%s:%d:%pI4h): "
"bad dst nid %s\n", libcfs_nid2str(nid),
ni == NULL ? "NA" : libcfs_nid2str(ni->ni_nid),
ibdev->ibd_ifname, ibdev->ibd_nnets,
- HIPQUAD(ibdev->ibd_ifip),
+ &ibdev->ibd_ifip,
libcfs_nid2str(reqmsg->ibm_dstnid));
goto failed;
LASSERT (net->ibn_shutdown == 0);
kiblnd_peer_addref(peer);
- cfs_list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
+ list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
write_unlock_irqrestore(g_lock, flags);
}
return -ECONNREFUSED;
}
-void
+static void
kiblnd_reconnect (kib_conn_t *conn, int version,
__u64 incarnation, int why, kib_connparams_t *cp)
{
* NB: reconnect is still needed even when ibp_tx_queue is
* empty if ibp_version != version because reconnect may be
* initiated by kiblnd_query() */
- if ((!cfs_list_empty(&peer->ibp_tx_queue) ||
+ if ((!list_empty(&peer->ibp_tx_queue) ||
peer->ibp_version != version) &&
peer->ibp_connecting == 1 &&
peer->ibp_accepting == 0) {
kiblnd_connect_peer(peer);
}
-void
+static void
kiblnd_rejected (kib_conn_t *conn, int reason, void *priv, int priv_nob)
{
kib_peer_t *peer = conn->ibc_peer;
kiblnd_connreq_done(conn, -ECONNREFUSED);
}
-void
+static void
kiblnd_check_connreply (kib_conn_t *conn, void *priv, int priv_nob)
{
kib_peer_t *peer = conn->ibc_peer;
kiblnd_connreq_done(conn, 0);
}
-int
+static int
kiblnd_active_connect (struct rdma_cm_id *cmid)
{
kib_peer_t *peer = (kib_peer_t *)cmid->context;
}
static int
-kiblnd_check_txs_locked(kib_conn_t *conn, cfs_list_t *txs)
+kiblnd_check_txs_locked(kib_conn_t *conn, struct list_head *txs)
{
- kib_tx_t *tx;
- cfs_list_t *ttmp;
+ kib_tx_t *tx;
+ struct list_head *ttmp;
- cfs_list_for_each (ttmp, txs) {
- tx = cfs_list_entry (ttmp, kib_tx_t, tx_list);
+ list_for_each(ttmp, txs) {
+ tx = list_entry(ttmp, kib_tx_t, tx_list);
- if (txs != &conn->ibc_active_txs) {
- LASSERT (tx->tx_queued);
- } else {
- LASSERT (!tx->tx_queued);
- LASSERT (tx->tx_waiting || tx->tx_sending != 0);
- }
+ if (txs != &conn->ibc_active_txs) {
+ LASSERT(tx->tx_queued);
+ } else {
+ LASSERT(!tx->tx_queued);
+ LASSERT(tx->tx_waiting || tx->tx_sending != 0);
+ }
- if (cfs_time_aftereq (jiffies, tx->tx_deadline)) {
- CERROR("Timed out tx: %s, %lu seconds\n",
- kiblnd_queue2str(conn, txs),
- cfs_duration_sec(jiffies - tx->tx_deadline));
- return 1;
- }
- }
+ if (cfs_time_aftereq(jiffies, tx->tx_deadline)) {
+ CERROR("Timed out tx: %s, %lu seconds\n",
+ kiblnd_queue2str(conn, txs),
+ cfs_duration_sec(jiffies - tx->tx_deadline));
+ return 1;
+ }
+ }
- return 0;
+ return 0;
}
static int
kiblnd_check_txs_locked(conn, &conn->ibc_active_txs);
}
-void
+static void
kiblnd_check_conns (int idx)
{
- CFS_LIST_HEAD (closes);
- CFS_LIST_HEAD (checksends);
- cfs_list_t *peers = &kiblnd_data.kib_peers[idx];
- cfs_list_t *ptmp;
- kib_peer_t *peer;
- kib_conn_t *conn;
- cfs_list_t *ctmp;
- unsigned long flags;
+ struct list_head closes = LIST_HEAD_INIT(closes);
+ struct list_head checksends = LIST_HEAD_INIT(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 *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
* take a look... */
read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
- cfs_list_for_each (ptmp, peers) {
- peer = cfs_list_entry (ptmp, kib_peer_t, ibp_list);
+ list_for_each(ptmp, peers) {
+ peer = list_entry(ptmp, kib_peer_t, ibp_list);
- cfs_list_for_each (ctmp, &peer->ibp_conns) {
- int timedout;
- int sendnoop;
+ list_for_each(ctmp, &peer->ibp_conns) {
+ int timedout;
+ int sendnoop;
- conn = cfs_list_entry(ctmp, kib_conn_t, ibc_list);
+ conn = list_entry(ctmp, kib_conn_t, ibc_list);
- LASSERT (conn->ibc_state == IBLND_CONN_ESTABLISHED);
+ LASSERT(conn->ibc_state == IBLND_CONN_ESTABLISHED);
spin_lock(&conn->ibc_lock);
timedout = kiblnd_conn_timed_out_locked(conn);
if (!sendnoop && !timedout) {
spin_unlock(&conn->ibc_lock);
- continue;
- }
+ continue;
+ }
- if (timedout) {
- CERROR("Timed out RDMA with %s (%lu): "
- "c: %u, oc: %u, rc: %u\n",
- libcfs_nid2str(peer->ibp_nid),
- cfs_duration_sec(cfs_time_current() -
- peer->ibp_last_alive),
- conn->ibc_credits,
- conn->ibc_outstanding_credits,
- conn->ibc_reserved_credits);
- cfs_list_add(&conn->ibc_connd_list, &closes);
- } else {
- cfs_list_add(&conn->ibc_connd_list,
- &checksends);
- }
- /* +ref for 'closes' or 'checksends' */
- kiblnd_conn_addref(conn);
+ if (timedout) {
+ CERROR("Timed out RDMA with %s (%lu): "
+ "c: %u, oc: %u, rc: %u\n",
+ libcfs_nid2str(peer->ibp_nid),
+ cfs_duration_sec(cfs_time_current() -
+ peer->ibp_last_alive),
+ conn->ibc_credits,
+ conn->ibc_outstanding_credits,
+ conn->ibc_reserved_credits);
+ list_add(&conn->ibc_connd_list, &closes);
+ } else {
+ list_add(&conn->ibc_connd_list, &checksends);
+ }
+ /* +ref for 'closes' or 'checksends' */
+ kiblnd_conn_addref(conn);
spin_unlock(&conn->ibc_lock);
}
read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
- /* Handle timeout by closing the whole
- * connection. We can only be sure RDMA activity
- * has ceased once the QP has been modified. */
- while (!cfs_list_empty(&closes)) {
- conn = cfs_list_entry(closes.next,
- kib_conn_t, ibc_connd_list);
- cfs_list_del(&conn->ibc_connd_list);
- kiblnd_close_conn(conn, -ETIMEDOUT);
- kiblnd_conn_decref(conn);
- }
+ /* Handle timeout by closing the whole
+ * connection. We can only be sure RDMA activity
+ * has ceased once the QP has been modified. */
+ while (!list_empty(&closes)) {
+ conn = list_entry(closes.next,
+ kib_conn_t, ibc_connd_list);
+ list_del(&conn->ibc_connd_list);
+ kiblnd_close_conn(conn, -ETIMEDOUT);
+ kiblnd_conn_decref(conn);
+ }
- /* In case we have enough credits to return via a
- * NOOP, but there were no non-blocking tx descs
- * free to do it last time... */
- while (!cfs_list_empty(&checksends)) {
- conn = cfs_list_entry(checksends.next,
- kib_conn_t, ibc_connd_list);
- cfs_list_del(&conn->ibc_connd_list);
- kiblnd_check_sends(conn);
- kiblnd_conn_decref(conn);
- }
+ /* In case we have enough credits to return via a
+ * NOOP, but there were no non-blocking tx descs
+ * free to do it last time... */
+ while (!list_empty(&checksends)) {
+ conn = list_entry(checksends.next,
+ kib_conn_t, ibc_connd_list);
+ list_del(&conn->ibc_connd_list);
+ kiblnd_check_sends(conn);
+ kiblnd_conn_decref(conn);
+ }
}
-void
+static void
kiblnd_disconnect_conn (kib_conn_t *conn)
{
LASSERT (!in_interrupt());
dropped_lock = 0;
- if (!cfs_list_empty (&kiblnd_data.kib_connd_zombies)) {
- conn = cfs_list_entry(kiblnd_data. \
+ if (!list_empty(&kiblnd_data.kib_connd_zombies)) {
+ conn = list_entry(kiblnd_data. \
kib_connd_zombies.next,
kib_conn_t, ibc_list);
- cfs_list_del(&conn->ibc_list);
+ list_del(&conn->ibc_list);
spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
flags);
spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
}
- if (!cfs_list_empty(&kiblnd_data.kib_connd_conns)) {
- conn = cfs_list_entry(kiblnd_data.kib_connd_conns.next,
+ if (!list_empty(&kiblnd_data.kib_connd_conns)) {
+ conn = list_entry(kiblnd_data.kib_connd_conns.next,
kib_conn_t, ibc_list);
- cfs_list_del(&conn->ibc_list);
+ list_del(&conn->ibc_list);
spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
flags);
kiblnd_data.kib_peer_hash_size;
}
- deadline += p * HZ;
+ deadline += msecs_to_jiffies(p * MSEC_PER_SEC);
spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
}
}
}
-void
+static void
kiblnd_complete (struct ib_wc *wc)
{
switch (kiblnd_wreqid2type(wc->wr_id)) {
conn->ibc_nsends_posted > 0)) {
kiblnd_conn_addref(conn); /* +1 ref for sched_conns */
conn->ibc_scheduled = 1;
- cfs_list_add_tail(&conn->ibc_sched_list, &sched->ibs_conns);
+ list_add_tail(&conn->ibc_sched_list, &sched->ibs_conns);
if (waitqueue_active(&sched->ibs_waitq))
wake_up(&sched->ibs_waitq);
did_something = 0;
- if (!cfs_list_empty(&sched->ibs_conns)) {
- conn = cfs_list_entry(sched->ibs_conns.next,
+ if (!list_empty(&sched->ibs_conns)) {
+ conn = list_entry(sched->ibs_conns.next,
kib_conn_t, ibc_sched_list);
/* take over kib_sched_conns' ref on conn... */
LASSERT(conn->ibc_scheduled);
- cfs_list_del(&conn->ibc_sched_list);
+ list_del(&conn->ibc_sched_list);
conn->ibc_ready = 0;
spin_unlock_irqrestore(&sched->ibs_lock, flags);
* this one... */
/* +1 ref for sched_conns */
kiblnd_conn_addref(conn);
- cfs_list_add_tail(&conn->ibc_sched_list,
+ list_add_tail(&conn->ibc_sched_list,
&sched->ibs_conns);
if (waitqueue_active(&sched->ibs_waitq))
wake_up(&sched->ibs_waitq);
int do_failover = 0;
int long_sleep;
- cfs_list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
+ list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
ibd_fail_list) {
if (cfs_time_before(cfs_time_current(),
dev->ibd_next_failover))
}
if (do_failover) {
- cfs_list_del_init(&dev->ibd_fail_list);
+ list_del_init(&dev->ibd_fail_list);
dev->ibd_failover = 1;
write_unlock_irqrestore(glock, flags);
dev->ibd_next_failover =
cfs_time_shift(min(dev->ibd_failed_failover, 10));
if (kiblnd_dev_can_failover(dev)) {
- cfs_list_add_tail(&dev->ibd_fail_list,
+ list_add_tail(&dev->ibd_fail_list,
&kiblnd_data.kib_failed_devs);
}
}
/* long sleep if no more pending failover */
- long_sleep = cfs_list_empty(&kiblnd_data.kib_failed_devs);
+ long_sleep = list_empty(&kiblnd_data.kib_failed_devs);
set_current_state(TASK_INTERRUPTIBLE);
add_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
* we need checking like this because if there is not active
* connection on the dev and no SEND from local, we may listen
* on wrong HCA for ever while there is a bonding failover */
- cfs_list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
+ list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
if (kiblnd_dev_can_failover(dev)) {
- cfs_list_add_tail(&dev->ibd_fail_list,
+ list_add_tail(&dev->ibd_fail_list,
&kiblnd_data.kib_failed_devs);
}
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff