#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];
- kib_net_t *net = ni->ni_data;
- int rc;
- int i;
+ lnet_msg_t *lntmsg[2];
+ kib_net_t *net = ni->ni_data;
+ int rc;
+ int i;
- LASSERT (net != NULL);
- LASSERT (!cfs_in_interrupt());
- LASSERT (!tx->tx_queued); /* mustn't be queued for sending */
- LASSERT (tx->tx_sending == 0); /* mustn't be awaiting sent callback */
- LASSERT (!tx->tx_waiting); /* mustn't be awaiting peer response */
- LASSERT (tx->tx_pool != NULL);
+ LASSERT (net != NULL);
+ LASSERT (!in_interrupt());
+ LASSERT (!tx->tx_queued); /* mustn't be queued for sending */
+ LASSERT (tx->tx_sending == 0); /* mustn't be awaiting sent callback */
+ LASSERT (!tx->tx_waiting); /* mustn't be awaiting peer response */
+ LASSERT (tx->tx_pool != NULL);
- kiblnd_unmap_tx(ni, tx);
+ kiblnd_unmap_tx(ni, tx);
- /* tx may have up to 2 lnet msgs to finalise */
- lntmsg[0] = tx->tx_lntmsg[0]; tx->tx_lntmsg[0] = NULL;
- lntmsg[1] = tx->tx_lntmsg[1]; tx->tx_lntmsg[1] = NULL;
- rc = tx->tx_status;
+ /* tx may have up to 2 lnet msgs to finalise */
+ lntmsg[0] = tx->tx_lntmsg[0]; tx->tx_lntmsg[0] = NULL;
+ lntmsg[1] = tx->tx_lntmsg[1]; tx->tx_lntmsg[1] = NULL;
+ rc = tx->tx_status;
- if (tx->tx_conn != NULL) {
- LASSERT (ni == tx->tx_conn->ibc_peer->ibp_ni);
+ if (tx->tx_conn != NULL) {
+ LASSERT (ni == tx->tx_conn->ibc_peer->ibp_ni);
- kiblnd_conn_decref(tx->tx_conn);
- tx->tx_conn = NULL;
- }
+ kiblnd_conn_decref(tx->tx_conn);
+ tx->tx_conn = NULL;
+ }
- tx->tx_nwrq = 0;
- tx->tx_status = 0;
+ tx->tx_nwrq = 0;
+ tx->tx_status = 0;
- kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
+ kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
- /* delay finalize until my descs have been freed */
- for (i = 0; i < 2; i++) {
- if (lntmsg[i] == NULL)
- continue;
+ /* delay finalize until my descs have been freed */
+ for (i = 0; i < 2; i++) {
+ if (lntmsg[i] == NULL)
+ continue;
- lnet_finalize(ni, lntmsg[i], rc);
- }
+ lnet_finalize(ni, lntmsg[i], rc);
+ }
}
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;
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 (!cfs_in_interrupt());
- LASSERT (credit == IBLND_POSTRX_NO_CREDIT ||
- credit == IBLND_POSTRX_PEER_CREDIT ||
- credit == IBLND_POSTRX_RSRVD_CREDIT);
+ LASSERT (net != NULL);
+ LASSERT (!in_interrupt());
+ LASSERT (credit == IBLND_POSTRX_NO_CREDIT ||
+ credit == IBLND_POSTRX_PEER_CREDIT ||
+ credit == IBLND_POSTRX_RSRVD_CREDIT);
- mr = kiblnd_find_dma_mr(conn->ibc_hdev, rx->rx_msgaddr, IBLND_MSG_SIZE);
- LASSERT (mr != NULL);
+ mr = kiblnd_find_dma_mr(conn->ibc_hdev, rx->rx_msgaddr, IBLND_MSG_SIZE);
+ LASSERT (mr != NULL);
rx->rx_sge.lkey = mr->lkey;
rx->rx_sge.addr = rx->rx_msgaddr;
return 0;
}
-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;
- if (vaddr >= VMALLOC_START &&
- vaddr < VMALLOC_END) {
+ if (is_vmalloc_addr((void *)vaddr)) {
page = vmalloc_to_page ((void *)vaddr);
LASSERT (page != NULL);
return 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)
{
kib_msg_t *msg = tx->tx_msg;
}
/* 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;
- 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;
-
- LASSERT (!cfs_in_interrupt());
- LASSERT (tx->tx_nwrq == 0);
- LASSERT (type == IBLND_MSG_GET_DONE ||
- type == IBLND_MSG_PUT_DONE);
-
- srcidx = dstidx = 0;
+ 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;
+
+ LASSERT (!in_interrupt());
+ LASSERT (tx->tx_nwrq == 0);
+ LASSERT (type == IBLND_MSG_GET_DONE ||
+ type == IBLND_MSG_PUT_DONE);
+
+ srcidx = dstidx = 0;
while (resid > 0) {
if (srcidx >= srcrd->rd_nfrags) {
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 * CFS_HZ);
+ tx->tx_queued = 1;
+ tx->tx_deadline = jiffies + (*kiblnd_tunables.kib_timeout * HZ);
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;
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);
CDEBUG(D_NET, "sending %d bytes in %d frags to %s\n",
payload_nob, payload_niov, libcfs_id2str(target));
- LASSERT (payload_nob == 0 || payload_niov > 0);
- LASSERT (payload_niov <= LNET_MAX_IOV);
+ LASSERT (payload_nob == 0 || payload_niov > 0);
+ LASSERT (payload_niov <= LNET_MAX_IOV);
- /* Thread context */
- LASSERT (!cfs_in_interrupt());
- /* payload is either all vaddrs or all pages */
- LASSERT (!(payload_kiov != NULL && payload_iov != NULL));
+ /* Thread context */
+ LASSERT (!in_interrupt());
+ /* payload is either all vaddrs or all pages */
+ LASSERT (!(payload_kiov != NULL && payload_iov != NULL));
- switch (type) {
- default:
- LBUG();
- return (-EIO);
+ switch (type) {
+ default:
+ LBUG();
+ return (-EIO);
case LNET_MSG_ACK:
LASSERT (payload_nob == 0);
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;
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;
+ int nob;
+ int post_credit = IBLND_POSTRX_PEER_CREDIT;
+ int rc = 0;
- LASSERT (mlen <= rlen);
- LASSERT (!cfs_in_interrupt());
- /* Either all pages or all vaddrs */
- LASSERT (!(kiov != NULL && iov != NULL));
+ LASSERT (mlen <= rlen);
+ LASSERT (!in_interrupt());
+ /* Either all pages or all vaddrs */
+ LASSERT (!(kiov != NULL && iov != NULL));
- switch (rxmsg->ibm_type) {
- default:
- LBUG();
+ switch (rxmsg->ibm_type) {
+ default:
+ LBUG();
case IBLND_MSG_IMMEDIATE:
nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[rlen]);
}
int
-kiblnd_thread_start (int (*fn)(void *arg), void *arg)
+kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name)
{
- long pid = cfs_create_thread (fn, arg, 0);
+ struct task_struct *task = kthread_run(fn, arg, name);
- if (pid < 0)
- return ((int)pid);
+ if (IS_ERR(task))
+ return PTR_ERR(task);
- cfs_atomic_inc (&kiblnd_data.kib_nthreads);
- return (0);
+ atomic_inc(&kiblnd_data.kib_nthreads);
+ return 0;
}
-void
+static void
kiblnd_thread_fini (void)
{
- cfs_atomic_dec (&kiblnd_data.kib_nthreads);
+ 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() */
- peer->ibp_last_alive = cfs_time_current();
- cfs_mb();
+ /* This is racy, but everyone's only writing cfs_time_current() */
+ peer->ibp_last_alive = 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);
kiblnd_set_conn_state(conn, IBLND_CONN_CLOSING);
- if (error != 0 &&
- kiblnd_dev_can_failover(dev)) {
- cfs_list_add_tail(&dev->ibd_fail_list,
- &kiblnd_data.kib_failed_devs);
- cfs_waitq_signal(&kiblnd_data.kib_failover_waitq);
- }
+ if (error != 0 &&
+ kiblnd_dev_can_failover(dev)) {
+ 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);
- cfs_waitq_signal(&kiblnd_data.kib_connd_waitq);
+ 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;
kib_rx_t *rx;
- LASSERT(!cfs_in_interrupt());
+ LASSERT(!in_interrupt());
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_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
- }
+ }
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 (!cfs_in_interrupt());
- LASSERT (conn->ibc_state > IBLND_CONN_INIT);
+ LASSERT (!in_interrupt());
+ LASSERT (conn->ibc_state > IBLND_CONN_INIT);
- kiblnd_set_conn_state(conn, IBLND_CONN_DISCONNECTED);
+ kiblnd_set_conn_state(conn, IBLND_CONN_DISCONNECTED);
- /* abort_receives moves QP state to IB_QPS_ERR. This is only required
- * for connections that didn't get as far as being connected, because
- * rdma_disconnect() does this for free. */
- kiblnd_abort_receives(conn);
+ /* abort_receives moves QP state to IB_QPS_ERR. This is only required
+ * for connections that didn't get as far as being connected, because
+ * rdma_disconnect() does this for free. */
+ kiblnd_abort_receives(conn);
- /* Complete all tx descs not waiting for sends to complete.
- * NB we should be safe from RDMA now that the QP has changed state */
+ /* Complete all tx descs not waiting for sends to complete.
+ * NB we should be safe from RDMA now that the QP has changed state */
- kiblnd_abort_txs(conn, &conn->ibc_tx_noops);
- kiblnd_abort_txs(conn, &conn->ibc_tx_queue);
- kiblnd_abort_txs(conn, &conn->ibc_tx_queue_rsrvd);
- kiblnd_abort_txs(conn, &conn->ibc_tx_queue_nocred);
- kiblnd_abort_txs(conn, &conn->ibc_active_txs);
+ kiblnd_abort_txs(conn, &conn->ibc_tx_noops);
+ kiblnd_abort_txs(conn, &conn->ibc_tx_queue);
+ kiblnd_abort_txs(conn, &conn->ibc_tx_queue_rsrvd);
+ kiblnd_abort_txs(conn, &conn->ibc_tx_queue_nocred);
+ kiblnd_abort_txs(conn, &conn->ibc_active_txs);
- kiblnd_handle_early_rxs(conn);
+ 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 (!cfs_in_interrupt());
+ LASSERT (error != 0);
+ LASSERT (!in_interrupt());
write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
- if (active) {
- LASSERT (peer->ibp_connecting > 0);
- peer->ibp_connecting--;
- } else {
- LASSERT (peer->ibp_accepting > 0);
- peer->ibp_accepting--;
- }
+ if (active) {
+ LASSERT (peer->ibp_connecting > 0);
+ peer->ibp_connecting--;
+ } else {
+ LASSERT (peer->ibp_accepting > 0);
+ peer->ibp_accepting--;
+ }
- if (peer->ibp_connecting != 0 ||
- peer->ibp_accepting != 0) {
- /* another connection attempt under way... */
+ if (peer->ibp_connecting != 0 ||
+ 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);
- CDEBUG(D_NET,"%s: active(%d), version(%x), status(%d)\n",
- libcfs_nid2str(peer->ibp_nid), active,
- conn->ibc_version, status);
+ CDEBUG(D_NET,"%s: active(%d), version(%x), status(%d)\n",
+ libcfs_nid2str(peer->ibp_nid), active,
+ conn->ibc_version, status);
- LASSERT (!cfs_in_interrupt());
- LASSERT ((conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT &&
- peer->ibp_connecting > 0) ||
- (conn->ibc_state == IBLND_CONN_PASSIVE_WAIT &&
- peer->ibp_accepting > 0));
+ LASSERT (!in_interrupt());
+ LASSERT ((conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT &&
+ peer->ibp_connecting > 0) ||
+ (conn->ibc_state == IBLND_CONN_PASSIVE_WAIT &&
+ peer->ibp_accepting > 0));
LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
conn->ibc_connvars = NULL;
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
+static int
kiblnd_passive_connect (struct rdma_cm_id *cmid, void *priv, int priv_nob)
{
rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
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;
- LASSERT (!cfs_in_interrupt());
+ 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 */
- ibdev = (kib_dev_t *)cmid->context;
- LASSERT (ibdev != NULL);
+ /* cmid inherits 'context' from the corresponding listener id */
+ ibdev = (kib_dev_t *)cmid->context;
+ LASSERT (ibdev != NULL);
memset(&rej, 0, sizeof(rej));
rej.ibr_magic = IBLND_MSG_MAGIC;
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;
+ kib_peer_t *peer = conn->ibc_peer;
- LASSERT (!cfs_in_interrupt());
- LASSERT (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
+ LASSERT (!in_interrupt());
+ LASSERT (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
- switch (reason) {
- case IB_CM_REJ_STALE_CONN:
- kiblnd_reconnect(conn, IBLND_MSG_VERSION, 0,
- IBLND_REJECT_CONN_STALE, NULL);
- break;
+ switch (reason) {
+ case IB_CM_REJ_STALE_CONN:
+ kiblnd_reconnect(conn, IBLND_MSG_VERSION, 0,
+ IBLND_REJECT_CONN_STALE, NULL);
+ break;
case IB_CM_REJ_INVALID_SERVICE_ID:
CNETERR("%s rejected: no listener at %d\n",
case IBLND_REJECT_MSG_QUEUE_SIZE:
CERROR("%s rejected: incompatible message queue depth %d, %d\n",
- libcfs_nid2str(peer->ibp_nid), cp->ibcp_queue_depth,
- IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
+ libcfs_nid2str(peer->ibp_nid),
+ cp != NULL ? cp->ibcp_queue_depth :
+ IBLND_MSG_QUEUE_SIZE(rej->ibr_version),
+ IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
break;
case IBLND_REJECT_RDMA_FRAGS:
CERROR("%s rejected: incompatible # of RDMA fragments %d, %d\n",
- libcfs_nid2str(peer->ibp_nid), cp->ibcp_max_frags,
- IBLND_RDMA_FRAGS(conn->ibc_version));
+ libcfs_nid2str(peer->ibp_nid),
+ cp != NULL ? cp->ibcp_max_frags :
+ IBLND_RDMA_FRAGS(rej->ibr_version),
+ IBLND_RDMA_FRAGS(conn->ibc_version));
break;
case IBLND_REJECT_NO_RESOURCES:
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 (!cfs_in_interrupt());
- LASSERT (current == kiblnd_data.kib_connd);
- LASSERT (conn->ibc_state == IBLND_CONN_CLOSING);
+ LASSERT (!in_interrupt());
+ LASSERT (current == kiblnd_data.kib_connd);
+ LASSERT (conn->ibc_state == IBLND_CONN_CLOSING);
- rdma_disconnect(conn->ibc_cmid);
- kiblnd_finalise_conn(conn);
+ rdma_disconnect(conn->ibc_cmid);
+ kiblnd_finalise_conn(conn);
- kiblnd_peer_notify(conn->ibc_peer);
+ kiblnd_peer_notify(conn->ibc_peer);
}
int
kiblnd_connd (void *arg)
{
- cfs_waitlink_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_daemonize ("kiblnd_connd");
- cfs_block_allsigs ();
+ cfs_block_allsigs ();
- cfs_waitlink_init (&wait);
- kiblnd_data.kib_connd = current;
+ init_waitqueue_entry_current (&wait);
+ kiblnd_data.kib_connd = current;
spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
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);
if (chunk == 0)
chunk = 1;
- for (i = 0; i < chunk; i++) {
- kiblnd_check_conns(peer_index);
- peer_index = (peer_index + 1) %
- kiblnd_data.kib_peer_hash_size;
- }
+ for (i = 0; i < chunk; i++) {
+ kiblnd_check_conns(peer_index);
+ peer_index = (peer_index + 1) %
+ kiblnd_data.kib_peer_hash_size;
+ }
- deadline += p * CFS_HZ;
+ deadline += p * HZ;
spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
}
continue;
/* Nothing to do for 'timeout' */
- cfs_set_current_state(CFS_TASK_INTERRUPTIBLE);
- cfs_waitq_add(&kiblnd_data.kib_connd_waitq, &wait);
+ set_current_state(TASK_INTERRUPTIBLE);
+ add_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
- cfs_waitq_timedwait(&wait, CFS_TASK_INTERRUPTIBLE, timeout);
+ waitq_timedwait(&wait, TASK_INTERRUPTIBLE, timeout);
- cfs_set_current_state(CFS_TASK_RUNNING);
- cfs_waitq_del(&kiblnd_data.kib_connd_waitq, &wait);
+ set_current_state(TASK_RUNNING);
+ remove_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
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 (cfs_waitq_active(&sched->ibs_waitq))
- cfs_waitq_signal(&sched->ibs_waitq);
+ if (waitqueue_active(&sched->ibs_waitq))
+ wake_up(&sched->ibs_waitq);
}
spin_unlock_irqrestore(&sched->ibs_lock, flags);
long id = (long)arg;
struct kib_sched_info *sched;
kib_conn_t *conn;
- cfs_waitlink_t wait;
+ wait_queue_t wait;
unsigned long flags;
struct ib_wc wc;
- char name[20];
int did_something;
int busy_loops = 0;
int rc;
- snprintf(name, sizeof(name), "kiblnd_sd_%02ld_%02ld",
- KIB_THREAD_CPT(id), KIB_THREAD_TID(id));
-
- cfs_daemonize(name);
cfs_block_allsigs();
- cfs_waitlink_init(&wait);
+ init_waitqueue_entry_current(&wait);
sched = kiblnd_data.kib_scheds[KIB_THREAD_CPT(id)];
rc = cfs_cpt_bind(lnet_cpt_table(), sched->ibs_cpt);
if (rc != 0) {
- CWARN("Failed to bind %s on CPT %d, please verify whether "
+ CWARN("Failed to bind on CPT %d, please verify whether "
"all CPUs are healthy and reload modules if necessary, "
"otherwise your system might under risk of low "
- "performance\n", name, sched->ibs_cpt);
+ "performance\n", sched->ibs_cpt);
}
spin_lock_irqsave(&sched->ibs_lock, flags);
if (busy_loops++ >= IBLND_RESCHED) {
spin_unlock_irqrestore(&sched->ibs_lock, flags);
- cfs_cond_resched();
+ cond_resched();
busy_loops = 0;
spin_lock_irqsave(&sched->ibs_lock, flags);
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 (cfs_waitq_active(&sched->ibs_waitq))
- cfs_waitq_signal(&sched->ibs_waitq);
+ if (waitqueue_active(&sched->ibs_waitq))
+ wake_up(&sched->ibs_waitq);
} else {
conn->ibc_scheduled = 0;
}
if (did_something)
continue;
- cfs_set_current_state(CFS_TASK_INTERRUPTIBLE);
- cfs_waitq_add_exclusive(&sched->ibs_waitq, &wait);
+ set_current_state(TASK_INTERRUPTIBLE);
+ add_wait_queue_exclusive(&sched->ibs_waitq, &wait);
spin_unlock_irqrestore(&sched->ibs_lock, flags);
- cfs_waitq_wait(&wait, CFS_TASK_INTERRUPTIBLE);
+ waitq_wait(&wait, TASK_INTERRUPTIBLE);
busy_loops = 0;
- cfs_waitq_del(&sched->ibs_waitq, &wait);
- cfs_set_current_state(CFS_TASK_RUNNING);
+ remove_wait_queue(&sched->ibs_waitq, &wait);
+ set_current_state(TASK_RUNNING);
spin_lock_irqsave(&sched->ibs_lock, flags);
}
kiblnd_failover_thread(void *arg)
{
rwlock_t *glock = &kiblnd_data.kib_global_lock;
- kib_dev_t *dev;
- cfs_waitlink_t wait;
- unsigned long flags;
- int rc;
+ kib_dev_t *dev;
+ wait_queue_t wait;
+ unsigned long flags;
+ int rc;
- LASSERT (*kiblnd_tunables.kib_dev_failover != 0);
+ LASSERT (*kiblnd_tunables.kib_dev_failover != 0);
- cfs_daemonize ("kiblnd_failover");
- cfs_block_allsigs ();
+ cfs_block_allsigs ();
- cfs_waitlink_init(&wait);
+ init_waitqueue_entry_current(&wait);
write_lock_irqsave(glock, flags);
while (!kiblnd_data.kib_shutdown) {
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);
- cfs_set_current_state(CFS_TASK_INTERRUPTIBLE);
- cfs_waitq_add(&kiblnd_data.kib_failover_waitq, &wait);
+ set_current_state(TASK_INTERRUPTIBLE);
+ add_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
write_unlock_irqrestore(glock, flags);
rc = schedule_timeout(long_sleep ? cfs_time_seconds(10) :
cfs_time_seconds(1));
- cfs_set_current_state(CFS_TASK_RUNNING);
- cfs_waitq_del(&kiblnd_data.kib_failover_waitq, &wait);
+ set_current_state(TASK_RUNNING);
+ remove_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
write_lock_irqsave(glock, flags);
if (!long_sleep || rc != 0)
* 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);
}
}