4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lnet/klnds/o2iblnd/o2iblnd_cb.c
38 * Author: Eric Barton <eric@bartonsoftware.com>
44 kiblnd_tx_done (lnet_ni_t *ni, kib_tx_t *tx)
46 lnet_msg_t *lntmsg[2];
47 kib_net_t *net = ni->ni_data;
51 LASSERT (net != NULL);
52 LASSERT (!in_interrupt());
53 LASSERT (!tx->tx_queued); /* mustn't be queued for sending */
54 LASSERT (tx->tx_sending == 0); /* mustn't be awaiting sent callback */
55 LASSERT (!tx->tx_waiting); /* mustn't be awaiting peer response */
56 LASSERT (tx->tx_pool != NULL);
58 kiblnd_unmap_tx(ni, tx);
60 /* tx may have up to 2 lnet msgs to finalise */
61 lntmsg[0] = tx->tx_lntmsg[0]; tx->tx_lntmsg[0] = NULL;
62 lntmsg[1] = tx->tx_lntmsg[1]; tx->tx_lntmsg[1] = NULL;
65 if (tx->tx_conn != NULL) {
66 LASSERT (ni == tx->tx_conn->ibc_peer->ibp_ni);
68 kiblnd_conn_decref(tx->tx_conn);
75 kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
77 /* delay finalize until my descs have been freed */
78 for (i = 0; i < 2; i++) {
79 if (lntmsg[i] == NULL)
82 lnet_finalize(ni, lntmsg[i], rc);
87 kiblnd_txlist_done (lnet_ni_t *ni, struct list_head *txlist, int status)
91 while (!list_empty (txlist)) {
92 tx = list_entry (txlist->next, kib_tx_t, tx_list);
94 list_del(&tx->tx_list);
97 tx->tx_status = status;
98 kiblnd_tx_done(ni, tx);
103 kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target)
105 kib_net_t *net = (kib_net_t *)ni->ni_data;
106 struct list_head *node;
108 kib_tx_poolset_t *tps;
110 tps = net->ibn_tx_ps[lnet_cpt_of_nid(target)];
111 node = kiblnd_pool_alloc_node(&tps->tps_poolset);
114 tx = container_of(node, kib_tx_t, tx_list);
116 LASSERT (tx->tx_nwrq == 0);
117 LASSERT (!tx->tx_queued);
118 LASSERT (tx->tx_sending == 0);
119 LASSERT (!tx->tx_waiting);
120 LASSERT (tx->tx_status == 0);
121 LASSERT (tx->tx_conn == NULL);
122 LASSERT (tx->tx_lntmsg[0] == NULL);
123 LASSERT (tx->tx_lntmsg[1] == NULL);
124 LASSERT (tx->tx_u.pmr == NULL);
125 LASSERT (tx->tx_nfrags == 0);
131 kiblnd_drop_rx(kib_rx_t *rx)
133 kib_conn_t *conn = rx->rx_conn;
134 struct kib_sched_info *sched = conn->ibc_sched;
137 spin_lock_irqsave(&sched->ibs_lock, flags);
138 LASSERT(conn->ibc_nrx > 0);
140 spin_unlock_irqrestore(&sched->ibs_lock, flags);
142 kiblnd_conn_decref(conn);
146 kiblnd_post_rx (kib_rx_t *rx, int credit)
148 kib_conn_t *conn = rx->rx_conn;
149 kib_net_t *net = conn->ibc_peer->ibp_ni->ni_data;
150 struct ib_recv_wr *bad_wrq = NULL;
154 LASSERT (net != NULL);
155 LASSERT (!in_interrupt());
156 LASSERT (credit == IBLND_POSTRX_NO_CREDIT ||
157 credit == IBLND_POSTRX_PEER_CREDIT ||
158 credit == IBLND_POSTRX_RSRVD_CREDIT);
160 mr = kiblnd_find_dma_mr(conn->ibc_hdev, rx->rx_msgaddr, IBLND_MSG_SIZE);
161 LASSERT (mr != NULL);
163 rx->rx_sge.lkey = mr->lkey;
164 rx->rx_sge.addr = rx->rx_msgaddr;
165 rx->rx_sge.length = IBLND_MSG_SIZE;
167 rx->rx_wrq.next = NULL;
168 rx->rx_wrq.sg_list = &rx->rx_sge;
169 rx->rx_wrq.num_sge = 1;
170 rx->rx_wrq.wr_id = kiblnd_ptr2wreqid(rx, IBLND_WID_RX);
172 LASSERT (conn->ibc_state >= IBLND_CONN_INIT);
173 LASSERT (rx->rx_nob >= 0); /* not posted */
175 if (conn->ibc_state > IBLND_CONN_ESTABLISHED) {
176 kiblnd_drop_rx(rx); /* No more posts for this rx */
180 rx->rx_nob = -1; /* flag posted */
182 rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq);
184 CERROR("Can't post rx for %s: %d, bad_wrq: %p\n",
185 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc, bad_wrq);
189 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */
193 kiblnd_close_conn(conn, rc);
194 kiblnd_drop_rx(rx); /* No more posts for this rx */
198 if (credit == IBLND_POSTRX_NO_CREDIT)
201 spin_lock(&conn->ibc_lock);
202 if (credit == IBLND_POSTRX_PEER_CREDIT)
203 conn->ibc_outstanding_credits++;
205 conn->ibc_reserved_credits++;
206 spin_unlock(&conn->ibc_lock);
208 kiblnd_check_sends(conn);
213 kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie)
215 struct list_head *tmp;
217 list_for_each(tmp, &conn->ibc_active_txs) {
218 kib_tx_t *tx = list_entry(tmp, kib_tx_t, tx_list);
220 LASSERT (!tx->tx_queued);
221 LASSERT (tx->tx_sending != 0 || tx->tx_waiting);
223 if (tx->tx_cookie != cookie)
226 if (tx->tx_waiting &&
227 tx->tx_msg->ibm_type == txtype)
230 CWARN("Bad completion: %swaiting, type %x (wanted %x)\n",
231 tx->tx_waiting ? "" : "NOT ",
232 tx->tx_msg->ibm_type, txtype);
238 kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie)
241 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
244 spin_lock(&conn->ibc_lock);
246 tx = kiblnd_find_waiting_tx_locked(conn, txtype, cookie);
248 spin_unlock(&conn->ibc_lock);
250 CWARN("Unmatched completion type %x cookie %#llx from %s\n",
251 txtype, cookie, libcfs_nid2str(conn->ibc_peer->ibp_nid));
252 kiblnd_close_conn(conn, -EPROTO);
256 if (tx->tx_status == 0) { /* success so far */
257 if (status < 0) { /* failed? */
258 tx->tx_status = status;
259 } else if (txtype == IBLND_MSG_GET_REQ) {
260 lnet_set_reply_msg_len(ni, tx->tx_lntmsg[1], status);
266 idle = !tx->tx_queued && (tx->tx_sending == 0);
268 list_del(&tx->tx_list);
270 spin_unlock(&conn->ibc_lock);
273 kiblnd_tx_done(ni, tx);
277 kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie)
279 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
280 kib_tx_t *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
283 CERROR("Can't get tx for completion %x for %s\n",
284 type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
288 tx->tx_msg->ibm_u.completion.ibcm_status = status;
289 tx->tx_msg->ibm_u.completion.ibcm_cookie = cookie;
290 kiblnd_init_tx_msg(ni, tx, type, sizeof(kib_completion_msg_t));
292 kiblnd_queue_tx(tx, conn);
296 kiblnd_handle_rx (kib_rx_t *rx)
298 kib_msg_t *msg = rx->rx_msg;
299 kib_conn_t *conn = rx->rx_conn;
300 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
301 int credits = msg->ibm_credits;
307 LASSERT (conn->ibc_state >= IBLND_CONN_ESTABLISHED);
309 CDEBUG (D_NET, "Received %x[%d] from %s\n",
310 msg->ibm_type, credits,
311 libcfs_nid2str(conn->ibc_peer->ibp_nid));
314 /* Have I received credits that will let me send? */
315 spin_lock(&conn->ibc_lock);
317 if (conn->ibc_credits + credits >
318 IBLND_MSG_QUEUE_SIZE(conn->ibc_version)) {
319 rc2 = conn->ibc_credits;
320 spin_unlock(&conn->ibc_lock);
322 CERROR("Bad credits from %s: %d + %d > %d\n",
323 libcfs_nid2str(conn->ibc_peer->ibp_nid),
325 IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
327 kiblnd_close_conn(conn, -EPROTO);
328 kiblnd_post_rx(rx, IBLND_POSTRX_NO_CREDIT);
332 conn->ibc_credits += credits;
334 /* This ensures the credit taken by NOOP can be returned */
335 if (msg->ibm_type == IBLND_MSG_NOOP &&
336 !IBLND_OOB_CAPABLE(conn->ibc_version)) /* v1 only */
337 conn->ibc_outstanding_credits++;
339 spin_unlock(&conn->ibc_lock);
340 kiblnd_check_sends(conn);
343 switch (msg->ibm_type) {
345 CERROR("Bad IBLND message type %x from %s\n",
346 msg->ibm_type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
347 post_credit = IBLND_POSTRX_NO_CREDIT;
352 if (IBLND_OOB_CAPABLE(conn->ibc_version)) {
353 post_credit = IBLND_POSTRX_NO_CREDIT;
357 if (credits != 0) /* credit already posted */
358 post_credit = IBLND_POSTRX_NO_CREDIT;
359 else /* a keepalive NOOP */
360 post_credit = IBLND_POSTRX_PEER_CREDIT;
363 case IBLND_MSG_IMMEDIATE:
364 post_credit = IBLND_POSTRX_DONT_POST;
365 rc = lnet_parse(ni, &msg->ibm_u.immediate.ibim_hdr,
366 msg->ibm_srcnid, rx, 0);
367 if (rc < 0) /* repost on error */
368 post_credit = IBLND_POSTRX_PEER_CREDIT;
371 case IBLND_MSG_PUT_REQ:
372 post_credit = IBLND_POSTRX_DONT_POST;
373 rc = lnet_parse(ni, &msg->ibm_u.putreq.ibprm_hdr,
374 msg->ibm_srcnid, rx, 1);
375 if (rc < 0) /* repost on error */
376 post_credit = IBLND_POSTRX_PEER_CREDIT;
379 case IBLND_MSG_PUT_NAK:
380 CWARN ("PUT_NACK from %s\n",
381 libcfs_nid2str(conn->ibc_peer->ibp_nid));
382 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
383 kiblnd_handle_completion(conn, IBLND_MSG_PUT_REQ,
384 msg->ibm_u.completion.ibcm_status,
385 msg->ibm_u.completion.ibcm_cookie);
388 case IBLND_MSG_PUT_ACK:
389 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
391 spin_lock(&conn->ibc_lock);
392 tx = kiblnd_find_waiting_tx_locked(conn, IBLND_MSG_PUT_REQ,
393 msg->ibm_u.putack.ibpam_src_cookie);
395 list_del(&tx->tx_list);
396 spin_unlock(&conn->ibc_lock);
399 CERROR("Unmatched PUT_ACK from %s\n",
400 libcfs_nid2str(conn->ibc_peer->ibp_nid));
405 LASSERT (tx->tx_waiting);
406 /* CAVEAT EMPTOR: I could be racing with tx_complete, but...
407 * (a) I can overwrite tx_msg since my peer has received it!
408 * (b) tx_waiting set tells tx_complete() it's not done. */
410 tx->tx_nwrq = 0; /* overwrite PUT_REQ */
412 rc2 = kiblnd_init_rdma(conn, tx, IBLND_MSG_PUT_DONE,
413 kiblnd_rd_size(&msg->ibm_u.putack.ibpam_rd),
414 &msg->ibm_u.putack.ibpam_rd,
415 msg->ibm_u.putack.ibpam_dst_cookie);
417 CERROR("Can't setup rdma for PUT to %s: %d\n",
418 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc2);
420 spin_lock(&conn->ibc_lock);
421 tx->tx_waiting = 0; /* clear waiting and queue atomically */
422 kiblnd_queue_tx_locked(tx, conn);
423 spin_unlock(&conn->ibc_lock);
426 case IBLND_MSG_PUT_DONE:
427 post_credit = IBLND_POSTRX_PEER_CREDIT;
428 kiblnd_handle_completion(conn, IBLND_MSG_PUT_ACK,
429 msg->ibm_u.completion.ibcm_status,
430 msg->ibm_u.completion.ibcm_cookie);
433 case IBLND_MSG_GET_REQ:
434 post_credit = IBLND_POSTRX_DONT_POST;
435 rc = lnet_parse(ni, &msg->ibm_u.get.ibgm_hdr,
436 msg->ibm_srcnid, rx, 1);
437 if (rc < 0) /* repost on error */
438 post_credit = IBLND_POSTRX_PEER_CREDIT;
441 case IBLND_MSG_GET_DONE:
442 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
443 kiblnd_handle_completion(conn, IBLND_MSG_GET_REQ,
444 msg->ibm_u.completion.ibcm_status,
445 msg->ibm_u.completion.ibcm_cookie);
449 if (rc < 0) /* protocol error */
450 kiblnd_close_conn(conn, rc);
452 if (post_credit != IBLND_POSTRX_DONT_POST)
453 kiblnd_post_rx(rx, post_credit);
457 kiblnd_rx_complete (kib_rx_t *rx, int status, int nob)
459 kib_msg_t *msg = rx->rx_msg;
460 kib_conn_t *conn = rx->rx_conn;
461 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
462 kib_net_t *net = ni->ni_data;
466 LASSERT (net != NULL);
467 LASSERT (rx->rx_nob < 0); /* was posted */
468 rx->rx_nob = 0; /* isn't now */
470 if (conn->ibc_state > IBLND_CONN_ESTABLISHED)
473 if (status != IB_WC_SUCCESS) {
474 CNETERR("Rx from %s failed: %d\n",
475 libcfs_nid2str(conn->ibc_peer->ibp_nid), status);
482 rc = kiblnd_unpack_msg(msg, rx->rx_nob);
484 CERROR ("Error %d unpacking rx from %s\n",
485 rc, libcfs_nid2str(conn->ibc_peer->ibp_nid));
489 if (msg->ibm_srcnid != conn->ibc_peer->ibp_nid ||
490 msg->ibm_dstnid != ni->ni_nid ||
491 msg->ibm_srcstamp != conn->ibc_incarnation ||
492 msg->ibm_dststamp != net->ibn_incarnation) {
493 CERROR ("Stale rx from %s\n",
494 libcfs_nid2str(conn->ibc_peer->ibp_nid));
499 /* set time last known alive */
500 kiblnd_peer_alive(conn->ibc_peer);
502 /* racing with connection establishment/teardown! */
504 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
505 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
508 write_lock_irqsave(g_lock, flags);
509 /* must check holding global lock to eliminate race */
510 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
511 list_add_tail(&rx->rx_list, &conn->ibc_early_rxs);
512 write_unlock_irqrestore(g_lock, flags);
515 write_unlock_irqrestore(g_lock, flags);
517 kiblnd_handle_rx(rx);
521 CDEBUG(D_NET, "rx %p conn %p\n", rx, conn);
522 kiblnd_close_conn(conn, err);
524 kiblnd_drop_rx(rx); /* Don't re-post rx. */
528 kiblnd_kvaddr_to_page (unsigned long vaddr)
532 if (is_vmalloc_addr((void *)vaddr)) {
533 page = vmalloc_to_page ((void *)vaddr);
534 LASSERT (page != NULL);
537 #ifdef CONFIG_HIGHMEM
538 if (vaddr >= PKMAP_BASE &&
539 vaddr < (PKMAP_BASE + LAST_PKMAP * PAGE_SIZE)) {
540 /* No highmem pages only used for bulk (kiov) I/O */
541 CERROR("find page for address in highmem\n");
545 page = virt_to_page (vaddr);
546 LASSERT (page != NULL);
551 kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
554 __u64 *pages = tx->tx_pages;
555 kib_fmr_poolset_t *fps;
562 LASSERT(tx->tx_pool != NULL);
563 LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
565 hdev = tx->tx_pool->tpo_hdev;
567 for (i = 0, npages = 0; i < rd->rd_nfrags; i++) {
568 for (size = 0; size < rd->rd_frags[i].rf_nob;
569 size += hdev->ibh_page_size) {
570 pages[npages ++] = (rd->rd_frags[i].rf_addr &
571 hdev->ibh_page_mask) + size;
575 cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
577 fps = net->ibn_fmr_ps[cpt];
578 rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->tx_u.fmr);
580 CERROR ("Can't map %d pages: %d\n", npages, rc);
584 /* If rd is not tx_rd, it's going to get sent to a peer, who will need
586 rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.fmr.fmr_pfmr->fmr->rkey :
587 tx->tx_u.fmr.fmr_pfmr->fmr->lkey;
588 rd->rd_frags[0].rf_addr &= ~hdev->ibh_page_mask;
589 rd->rd_frags[0].rf_nob = nob;
596 kiblnd_pmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
599 kib_pmr_poolset_t *pps;
604 LASSERT(tx->tx_pool != NULL);
605 LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
607 hdev = tx->tx_pool->tpo_hdev;
609 iova = rd->rd_frags[0].rf_addr & ~hdev->ibh_page_mask;
611 cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
613 pps = net->ibn_pmr_ps[cpt];
614 rc = kiblnd_pmr_pool_map(pps, hdev, rd, &iova, &tx->tx_u.pmr);
616 CERROR("Failed to create MR by phybuf: %d\n", rc);
620 /* If rd is not tx_rd, it's going to get sent to a peer, who will need
622 rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.pmr->pmr_mr->rkey :
623 tx->tx_u.pmr->pmr_mr->lkey;
625 rd->rd_frags[0].rf_addr = iova;
626 rd->rd_frags[0].rf_nob = nob;
632 kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx)
634 kib_net_t *net = ni->ni_data;
636 LASSERT(net != NULL);
638 if (net->ibn_fmr_ps != NULL && tx->tx_u.fmr.fmr_pfmr != NULL) {
639 kiblnd_fmr_pool_unmap(&tx->tx_u.fmr, tx->tx_status);
640 tx->tx_u.fmr.fmr_pfmr = NULL;
642 } else if (net->ibn_pmr_ps != NULL && tx->tx_u.pmr != NULL) {
643 kiblnd_pmr_pool_unmap(tx->tx_u.pmr);
647 if (tx->tx_nfrags != 0) {
648 kiblnd_dma_unmap_sg(tx->tx_pool->tpo_hdev->ibh_ibdev,
649 tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir);
655 kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx,
656 kib_rdma_desc_t *rd, int nfrags)
658 kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
659 kib_net_t *net = ni->ni_data;
660 struct ib_mr *mr = NULL;
664 /* If rd is not tx_rd, it's going to get sent to a peer and I'm the
666 tx->tx_dmadir = (rd != tx->tx_rd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
667 tx->tx_nfrags = nfrags;
670 kiblnd_dma_map_sg(hdev->ibh_ibdev,
671 tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir);
673 for (i = 0, nob = 0; i < rd->rd_nfrags; i++) {
674 rd->rd_frags[i].rf_nob = kiblnd_sg_dma_len(
675 hdev->ibh_ibdev, &tx->tx_frags[i]);
676 rd->rd_frags[i].rf_addr = kiblnd_sg_dma_address(
677 hdev->ibh_ibdev, &tx->tx_frags[i]);
678 nob += rd->rd_frags[i].rf_nob;
681 /* looking for pre-mapping MR */
682 mr = kiblnd_find_rd_dma_mr(hdev, rd);
684 /* found pre-mapping MR */
685 rd->rd_key = (rd != tx->tx_rd) ? mr->rkey : mr->lkey;
689 if (net->ibn_fmr_ps != NULL)
690 return kiblnd_fmr_map_tx(net, tx, rd, nob);
691 else if (net->ibn_pmr_ps != NULL)
692 return kiblnd_pmr_map_tx(net, tx, rd, nob);
699 kiblnd_setup_rd_iov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
700 unsigned int niov, struct iovec *iov, int offset, int nob)
702 kib_net_t *net = ni->ni_data;
704 struct scatterlist *sg;
711 LASSERT (net != NULL);
713 while (offset >= iov->iov_len) {
714 offset -= iov->iov_len;
724 vaddr = ((unsigned long)iov->iov_base) + offset;
725 page_offset = vaddr & (PAGE_SIZE - 1);
726 page = kiblnd_kvaddr_to_page(vaddr);
728 CERROR ("Can't find page\n");
732 fragnob = min((int)(iov->iov_len - offset), nob);
733 fragnob = min(fragnob, (int)PAGE_SIZE - page_offset);
735 sg_set_page(sg, page, fragnob, page_offset);
738 if (offset + fragnob < iov->iov_len) {
748 return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
752 kiblnd_setup_rd_kiov (lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
753 int nkiov, lnet_kiov_t *kiov, int offset, int nob)
755 kib_net_t *net = ni->ni_data;
756 struct scatterlist *sg;
759 CDEBUG(D_NET, "niov %d offset %d nob %d\n", nkiov, offset, nob);
763 LASSERT (net != NULL);
765 while (offset >= kiov->kiov_len) {
766 offset -= kiov->kiov_len;
776 fragnob = min((int)(kiov->kiov_len - offset), nob);
778 sg_set_page(sg, kiov->kiov_page, fragnob,
779 kiov->kiov_offset + offset);
788 return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
792 kiblnd_post_tx_locked (kib_conn_t *conn, kib_tx_t *tx, int credit)
793 __releases(conn->ibc_lock)
794 __acquires(conn->ibc_lock)
796 kib_msg_t *msg = tx->tx_msg;
797 kib_peer_t *peer = conn->ibc_peer;
798 int ver = conn->ibc_version;
801 struct ib_send_wr *bad_wrq;
803 LASSERT (tx->tx_queued);
804 /* We rely on this for QP sizing */
805 LASSERT (tx->tx_nwrq > 0);
806 LASSERT (tx->tx_nwrq <= 1 + IBLND_RDMA_FRAGS(ver));
808 LASSERT (credit == 0 || credit == 1);
809 LASSERT (conn->ibc_outstanding_credits >= 0);
810 LASSERT (conn->ibc_outstanding_credits <= IBLND_MSG_QUEUE_SIZE(ver));
811 LASSERT (conn->ibc_credits >= 0);
812 LASSERT (conn->ibc_credits <= IBLND_MSG_QUEUE_SIZE(ver));
814 if (conn->ibc_nsends_posted == IBLND_CONCURRENT_SENDS(ver)) {
815 /* tx completions outstanding... */
816 CDEBUG(D_NET, "%s: posted enough\n",
817 libcfs_nid2str(peer->ibp_nid));
821 if (credit != 0 && conn->ibc_credits == 0) { /* no credits */
822 CDEBUG(D_NET, "%s: no credits\n",
823 libcfs_nid2str(peer->ibp_nid));
827 if (credit != 0 && !IBLND_OOB_CAPABLE(ver) &&
828 conn->ibc_credits == 1 && /* last credit reserved */
829 msg->ibm_type != IBLND_MSG_NOOP) { /* for NOOP */
830 CDEBUG(D_NET, "%s: not using last credit\n",
831 libcfs_nid2str(peer->ibp_nid));
835 /* NB don't drop ibc_lock before bumping tx_sending */
836 list_del(&tx->tx_list);
839 if (msg->ibm_type == IBLND_MSG_NOOP &&
840 (!kiblnd_need_noop(conn) || /* redundant NOOP */
841 (IBLND_OOB_CAPABLE(ver) && /* posted enough NOOP */
842 conn->ibc_noops_posted == IBLND_OOB_MSGS(ver)))) {
843 /* OK to drop when posted enough NOOPs, since
844 * kiblnd_check_sends will queue NOOP again when
845 * posted NOOPs complete */
846 spin_unlock(&conn->ibc_lock);
847 kiblnd_tx_done(peer->ibp_ni, tx);
848 spin_lock(&conn->ibc_lock);
849 CDEBUG(D_NET, "%s(%d): redundant or enough NOOP\n",
850 libcfs_nid2str(peer->ibp_nid),
851 conn->ibc_noops_posted);
855 kiblnd_pack_msg(peer->ibp_ni, msg, ver, conn->ibc_outstanding_credits,
856 peer->ibp_nid, conn->ibc_incarnation);
858 conn->ibc_credits -= credit;
859 conn->ibc_outstanding_credits = 0;
860 conn->ibc_nsends_posted++;
861 if (msg->ibm_type == IBLND_MSG_NOOP)
862 conn->ibc_noops_posted++;
864 /* CAVEAT EMPTOR! This tx could be the PUT_DONE of an RDMA
865 * PUT. If so, it was first queued here as a PUT_REQ, sent and
866 * stashed on ibc_active_txs, matched by an incoming PUT_ACK,
867 * and then re-queued here. It's (just) possible that
868 * tx_sending is non-zero if we've not done the tx_complete()
869 * from the first send; hence the ++ rather than = below. */
871 list_add(&tx->tx_list, &conn->ibc_active_txs);
873 /* I'm still holding ibc_lock! */
874 if (conn->ibc_state != IBLND_CONN_ESTABLISHED) {
876 } else if (tx->tx_pool->tpo_pool.po_failed ||
877 conn->ibc_hdev != tx->tx_pool->tpo_hdev) {
878 /* close_conn will launch failover */
881 rc = ib_post_send(conn->ibc_cmid->qp,
882 tx->tx_wrq, &bad_wrq);
885 conn->ibc_last_send = jiffies;
890 /* NB credits are transferred in the actual
891 * message, which can only be the last work item */
892 conn->ibc_credits += credit;
893 conn->ibc_outstanding_credits += msg->ibm_credits;
894 conn->ibc_nsends_posted--;
895 if (msg->ibm_type == IBLND_MSG_NOOP)
896 conn->ibc_noops_posted--;
902 done = (tx->tx_sending == 0);
904 list_del(&tx->tx_list);
906 spin_unlock(&conn->ibc_lock);
908 if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
909 CERROR("Error %d posting transmit to %s\n",
910 rc, libcfs_nid2str(peer->ibp_nid));
912 CDEBUG(D_NET, "Error %d posting transmit to %s\n",
913 rc, libcfs_nid2str(peer->ibp_nid));
915 kiblnd_close_conn(conn, rc);
918 kiblnd_tx_done(peer->ibp_ni, tx);
920 spin_lock(&conn->ibc_lock);
926 kiblnd_check_sends (kib_conn_t *conn)
928 int ver = conn->ibc_version;
929 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
932 /* Don't send anything until after the connection is established */
933 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
934 CDEBUG(D_NET, "%s too soon\n",
935 libcfs_nid2str(conn->ibc_peer->ibp_nid));
939 spin_lock(&conn->ibc_lock);
941 LASSERT (conn->ibc_nsends_posted <= IBLND_CONCURRENT_SENDS(ver));
942 LASSERT (!IBLND_OOB_CAPABLE(ver) ||
943 conn->ibc_noops_posted <= IBLND_OOB_MSGS(ver));
944 LASSERT (conn->ibc_reserved_credits >= 0);
946 while (conn->ibc_reserved_credits > 0 &&
947 !list_empty(&conn->ibc_tx_queue_rsrvd)) {
948 tx = list_entry(conn->ibc_tx_queue_rsrvd.next,
950 list_del(&tx->tx_list);
951 list_add_tail(&tx->tx_list, &conn->ibc_tx_queue);
952 conn->ibc_reserved_credits--;
955 if (kiblnd_need_noop(conn)) {
956 spin_unlock(&conn->ibc_lock);
958 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
960 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_NOOP, 0);
962 spin_lock(&conn->ibc_lock);
964 kiblnd_queue_tx_locked(tx, conn);
967 kiblnd_conn_addref(conn); /* 1 ref for me.... (see b21911) */
972 if (!list_empty(&conn->ibc_tx_queue_nocred)) {
974 tx = list_entry(conn->ibc_tx_queue_nocred.next,
976 } else if (!list_empty(&conn->ibc_tx_noops)) {
977 LASSERT (!IBLND_OOB_CAPABLE(ver));
979 tx = list_entry(conn->ibc_tx_noops.next,
981 } else if (!list_empty(&conn->ibc_tx_queue)) {
983 tx = list_entry(conn->ibc_tx_queue.next,
988 if (kiblnd_post_tx_locked(conn, tx, credit) != 0)
992 spin_unlock(&conn->ibc_lock);
994 kiblnd_conn_decref(conn); /* ...until here */
998 kiblnd_tx_complete (kib_tx_t *tx, int status)
1000 int failed = (status != IB_WC_SUCCESS);
1001 kib_conn_t *conn = tx->tx_conn;
1004 LASSERT (tx->tx_sending > 0);
1007 if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
1008 CNETERR("Tx -> %s cookie %#llx sending %d waiting %d: failed %d\n",
1009 libcfs_nid2str(conn->ibc_peer->ibp_nid),
1010 tx->tx_cookie, tx->tx_sending, tx->tx_waiting,
1013 kiblnd_close_conn(conn, -EIO);
1015 kiblnd_peer_alive(conn->ibc_peer);
1018 spin_lock(&conn->ibc_lock);
1020 /* I could be racing with rdma completion. Whoever makes 'tx' idle
1021 * gets to free it, which also drops its ref on 'conn'. */
1024 conn->ibc_nsends_posted--;
1025 if (tx->tx_msg->ibm_type == IBLND_MSG_NOOP)
1026 conn->ibc_noops_posted--;
1029 tx->tx_waiting = 0; /* don't wait for peer */
1030 tx->tx_status = -EIO;
1033 idle = (tx->tx_sending == 0) && /* This is the final callback */
1034 !tx->tx_waiting && /* Not waiting for peer */
1035 !tx->tx_queued; /* Not re-queued (PUT_DONE) */
1037 list_del(&tx->tx_list);
1039 kiblnd_conn_addref(conn); /* 1 ref for me.... */
1041 spin_unlock(&conn->ibc_lock);
1044 kiblnd_tx_done(conn->ibc_peer->ibp_ni, tx);
1046 kiblnd_check_sends(conn);
1048 kiblnd_conn_decref(conn); /* ...until here */
1052 kiblnd_init_tx_msg (lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob)
1054 kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
1055 struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq];
1056 struct ib_send_wr *wrq = &tx->tx_wrq[tx->tx_nwrq];
1057 int nob = offsetof (kib_msg_t, ibm_u) + body_nob;
1060 LASSERT (tx->tx_nwrq >= 0);
1061 LASSERT (tx->tx_nwrq < IBLND_MAX_RDMA_FRAGS + 1);
1062 LASSERT (nob <= IBLND_MSG_SIZE);
1064 kiblnd_init_msg(tx->tx_msg, type, body_nob);
1066 mr = kiblnd_find_dma_mr(hdev, tx->tx_msgaddr, nob);
1067 LASSERT (mr != NULL);
1069 sge->lkey = mr->lkey;
1070 sge->addr = tx->tx_msgaddr;
1073 memset(wrq, 0, sizeof(*wrq));
1076 wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_TX);
1079 wrq->opcode = IB_WR_SEND;
1080 wrq->send_flags = IB_SEND_SIGNALED;
1086 kiblnd_init_rdma (kib_conn_t *conn, kib_tx_t *tx, int type,
1087 int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie)
1089 kib_msg_t *ibmsg = tx->tx_msg;
1090 kib_rdma_desc_t *srcrd = tx->tx_rd;
1091 struct ib_sge *sge = &tx->tx_sge[0];
1092 struct ib_send_wr *wrq = &tx->tx_wrq[0];
1098 LASSERT (!in_interrupt());
1099 LASSERT (tx->tx_nwrq == 0);
1100 LASSERT (type == IBLND_MSG_GET_DONE ||
1101 type == IBLND_MSG_PUT_DONE);
1103 srcidx = dstidx = 0;
1106 if (srcidx >= srcrd->rd_nfrags) {
1107 CERROR("Src buffer exhausted: %d frags\n", srcidx);
1112 if (dstidx == dstrd->rd_nfrags) {
1113 CERROR("Dst buffer exhausted: %d frags\n", dstidx);
1118 if (tx->tx_nwrq == IBLND_RDMA_FRAGS(conn->ibc_version)) {
1119 CERROR("RDMA too fragmented for %s (%d): "
1120 "%d/%d src %d/%d dst frags\n",
1121 libcfs_nid2str(conn->ibc_peer->ibp_nid),
1122 IBLND_RDMA_FRAGS(conn->ibc_version),
1123 srcidx, srcrd->rd_nfrags,
1124 dstidx, dstrd->rd_nfrags);
1129 wrknob = MIN(MIN(kiblnd_rd_frag_size(srcrd, srcidx),
1130 kiblnd_rd_frag_size(dstrd, dstidx)), resid);
1132 sge = &tx->tx_sge[tx->tx_nwrq];
1133 sge->addr = kiblnd_rd_frag_addr(srcrd, srcidx);
1134 sge->lkey = kiblnd_rd_frag_key(srcrd, srcidx);
1135 sge->length = wrknob;
1137 wrq = &tx->tx_wrq[tx->tx_nwrq];
1139 wrq->next = wrq + 1;
1140 wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA);
1143 wrq->opcode = IB_WR_RDMA_WRITE;
1144 wrq->send_flags = 0;
1146 wrq->wr.rdma.remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx);
1147 wrq->wr.rdma.rkey = kiblnd_rd_frag_key(dstrd, dstidx);
1149 srcidx = kiblnd_rd_consume_frag(srcrd, srcidx, wrknob);
1150 dstidx = kiblnd_rd_consume_frag(dstrd, dstidx, wrknob);
1159 if (rc < 0) /* no RDMA if completing with failure */
1162 ibmsg->ibm_u.completion.ibcm_status = rc;
1163 ibmsg->ibm_u.completion.ibcm_cookie = dstcookie;
1164 kiblnd_init_tx_msg(conn->ibc_peer->ibp_ni, tx,
1165 type, sizeof (kib_completion_msg_t));
1171 kiblnd_queue_tx_locked (kib_tx_t *tx, kib_conn_t *conn)
1173 struct list_head *q;
1175 LASSERT (tx->tx_nwrq > 0); /* work items set up */
1176 LASSERT (!tx->tx_queued); /* not queued for sending already */
1177 LASSERT (conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1180 tx->tx_deadline = jiffies + (*kiblnd_tunables.kib_timeout * HZ);
1182 if (tx->tx_conn == NULL) {
1183 kiblnd_conn_addref(conn);
1185 LASSERT (tx->tx_msg->ibm_type != IBLND_MSG_PUT_DONE);
1187 /* PUT_DONE first attached to conn as a PUT_REQ */
1188 LASSERT (tx->tx_conn == conn);
1189 LASSERT (tx->tx_msg->ibm_type == IBLND_MSG_PUT_DONE);
1192 switch (tx->tx_msg->ibm_type) {
1196 case IBLND_MSG_PUT_REQ:
1197 case IBLND_MSG_GET_REQ:
1198 q = &conn->ibc_tx_queue_rsrvd;
1201 case IBLND_MSG_PUT_NAK:
1202 case IBLND_MSG_PUT_ACK:
1203 case IBLND_MSG_PUT_DONE:
1204 case IBLND_MSG_GET_DONE:
1205 q = &conn->ibc_tx_queue_nocred;
1208 case IBLND_MSG_NOOP:
1209 if (IBLND_OOB_CAPABLE(conn->ibc_version))
1210 q = &conn->ibc_tx_queue_nocred;
1212 q = &conn->ibc_tx_noops;
1215 case IBLND_MSG_IMMEDIATE:
1216 q = &conn->ibc_tx_queue;
1220 list_add_tail(&tx->tx_list, q);
1224 kiblnd_queue_tx (kib_tx_t *tx, kib_conn_t *conn)
1226 spin_lock(&conn->ibc_lock);
1227 kiblnd_queue_tx_locked(tx, conn);
1228 spin_unlock(&conn->ibc_lock);
1230 kiblnd_check_sends(conn);
1233 static int kiblnd_resolve_addr(struct rdma_cm_id *cmid,
1234 struct sockaddr_in *srcaddr,
1235 struct sockaddr_in *dstaddr,
1238 unsigned short port;
1241 /* allow the port to be reused */
1242 rc = rdma_set_reuseaddr(cmid, 1);
1244 CERROR("Unable to set reuse on cmid: %d\n", rc);
1248 /* look for a free privileged port */
1249 for (port = PROT_SOCK-1; port > 0; port--) {
1250 srcaddr->sin_port = htons(port);
1251 rc = rdma_resolve_addr(cmid,
1252 (struct sockaddr *)srcaddr,
1253 (struct sockaddr *)dstaddr,
1256 CDEBUG(D_NET, "bound to port %hu\n", port);
1258 } else if (rc == -EADDRINUSE || rc == -EADDRNOTAVAIL) {
1259 CDEBUG(D_NET, "bind to port %hu failed: %d\n",
1266 CERROR("Failed to bind to a free privileged port\n");
1271 kiblnd_connect_peer (kib_peer_t *peer)
1273 struct rdma_cm_id *cmid;
1275 kib_net_t *net = peer->ibp_ni->ni_data;
1276 struct sockaddr_in srcaddr;
1277 struct sockaddr_in dstaddr;
1280 LASSERT (net != NULL);
1281 LASSERT (peer->ibp_connecting > 0);
1283 cmid = kiblnd_rdma_create_id(kiblnd_cm_callback, peer, RDMA_PS_TCP,
1287 CERROR("Can't create CMID for %s: %ld\n",
1288 libcfs_nid2str(peer->ibp_nid), PTR_ERR(cmid));
1294 memset(&srcaddr, 0, sizeof(srcaddr));
1295 srcaddr.sin_family = AF_INET;
1296 srcaddr.sin_addr.s_addr = htonl(dev->ibd_ifip);
1298 memset(&dstaddr, 0, sizeof(dstaddr));
1299 dstaddr.sin_family = AF_INET;
1300 dstaddr.sin_port = htons(*kiblnd_tunables.kib_service);
1301 dstaddr.sin_addr.s_addr = htonl(LNET_NIDADDR(peer->ibp_nid));
1303 kiblnd_peer_addref(peer); /* cmid's ref */
1305 if (*kiblnd_tunables.kib_use_priv_port) {
1306 rc = kiblnd_resolve_addr(cmid, &srcaddr, &dstaddr,
1307 *kiblnd_tunables.kib_timeout * 1000);
1309 rc = rdma_resolve_addr(cmid,
1310 (struct sockaddr *)&srcaddr,
1311 (struct sockaddr *)&dstaddr,
1312 *kiblnd_tunables.kib_timeout * 1000);
1315 /* Can't initiate address resolution: */
1316 CERROR("Can't resolve addr for %s: %d\n",
1317 libcfs_nid2str(peer->ibp_nid), rc);
1321 LASSERT (cmid->device != NULL);
1322 CDEBUG(D_NET, "%s: connection bound to %s:%pI4h:%s\n",
1323 libcfs_nid2str(peer->ibp_nid), dev->ibd_ifname,
1324 &dev->ibd_ifip, cmid->device->name);
1329 kiblnd_peer_decref(peer); /* cmid's ref */
1330 rdma_destroy_id(cmid);
1332 kiblnd_peer_connect_failed(peer, 1, rc);
1336 kiblnd_launch_tx (lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid)
1341 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
1342 unsigned long flags;
1345 /* If I get here, I've committed to send, so I complete the tx with
1346 * failure on any problems */
1348 LASSERT (tx == NULL || tx->tx_conn == NULL); /* only set when assigned a conn */
1349 LASSERT (tx == NULL || tx->tx_nwrq > 0); /* work items have been set up */
1351 /* First time, just use a read lock since I expect to find my peer
1353 read_lock_irqsave(g_lock, flags);
1355 peer = kiblnd_find_peer_locked(nid);
1356 if (peer != NULL && !list_empty(&peer->ibp_conns)) {
1357 /* Found a peer with an established connection */
1358 conn = kiblnd_get_conn_locked(peer);
1359 kiblnd_conn_addref(conn); /* 1 ref for me... */
1361 read_unlock_irqrestore(g_lock, flags);
1364 kiblnd_queue_tx(tx, conn);
1365 kiblnd_conn_decref(conn); /* ...to here */
1369 read_unlock(g_lock);
1370 /* Re-try with a write lock */
1373 peer = kiblnd_find_peer_locked(nid);
1375 if (list_empty(&peer->ibp_conns)) {
1376 /* found a peer, but it's still connecting... */
1377 LASSERT (peer->ibp_connecting != 0 ||
1378 peer->ibp_accepting != 0);
1380 list_add_tail(&tx->tx_list,
1381 &peer->ibp_tx_queue);
1382 write_unlock_irqrestore(g_lock, flags);
1384 conn = kiblnd_get_conn_locked(peer);
1385 kiblnd_conn_addref(conn); /* 1 ref for me... */
1387 write_unlock_irqrestore(g_lock, flags);
1390 kiblnd_queue_tx(tx, conn);
1391 kiblnd_conn_decref(conn); /* ...to here */
1396 write_unlock_irqrestore(g_lock, flags);
1398 /* Allocate a peer ready to add to the peer table and retry */
1399 rc = kiblnd_create_peer(ni, &peer, nid);
1401 CERROR("Can't create peer %s\n", libcfs_nid2str(nid));
1403 tx->tx_status = -EHOSTUNREACH;
1405 kiblnd_tx_done(ni, tx);
1410 write_lock_irqsave(g_lock, flags);
1412 peer2 = kiblnd_find_peer_locked(nid);
1413 if (peer2 != NULL) {
1414 if (list_empty(&peer2->ibp_conns)) {
1415 /* found a peer, but it's still connecting... */
1416 LASSERT (peer2->ibp_connecting != 0 ||
1417 peer2->ibp_accepting != 0);
1419 list_add_tail(&tx->tx_list,
1420 &peer2->ibp_tx_queue);
1421 write_unlock_irqrestore(g_lock, flags);
1423 conn = kiblnd_get_conn_locked(peer2);
1424 kiblnd_conn_addref(conn); /* 1 ref for me... */
1426 write_unlock_irqrestore(g_lock, flags);
1429 kiblnd_queue_tx(tx, conn);
1430 kiblnd_conn_decref(conn); /* ...to here */
1433 kiblnd_peer_decref(peer);
1437 /* Brand new peer */
1438 LASSERT (peer->ibp_connecting == 0);
1439 peer->ibp_connecting = 1;
1441 /* always called with a ref on ni, which prevents ni being shutdown */
1442 LASSERT (((kib_net_t *)ni->ni_data)->ibn_shutdown == 0);
1445 list_add_tail(&tx->tx_list, &peer->ibp_tx_queue);
1447 kiblnd_peer_addref(peer);
1448 list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
1450 write_unlock_irqrestore(g_lock, flags);
1452 kiblnd_connect_peer(peer);
1453 kiblnd_peer_decref(peer);
1457 kiblnd_send (lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg)
1459 lnet_hdr_t *hdr = &lntmsg->msg_hdr;
1460 int type = lntmsg->msg_type;
1461 lnet_process_id_t target = lntmsg->msg_target;
1462 int target_is_router = lntmsg->msg_target_is_router;
1463 int routing = lntmsg->msg_routing;
1464 unsigned int payload_niov = lntmsg->msg_niov;
1465 struct iovec *payload_iov = lntmsg->msg_iov;
1466 lnet_kiov_t *payload_kiov = lntmsg->msg_kiov;
1467 unsigned int payload_offset = lntmsg->msg_offset;
1468 unsigned int payload_nob = lntmsg->msg_len;
1474 /* NB 'private' is different depending on what we're sending.... */
1476 CDEBUG(D_NET, "sending %d bytes in %d frags to %s\n",
1477 payload_nob, payload_niov, libcfs_id2str(target));
1479 LASSERT (payload_nob == 0 || payload_niov > 0);
1480 LASSERT (payload_niov <= LNET_MAX_IOV);
1482 /* Thread context */
1483 LASSERT (!in_interrupt());
1484 /* payload is either all vaddrs or all pages */
1485 LASSERT (!(payload_kiov != NULL && payload_iov != NULL));
1493 LASSERT (payload_nob == 0);
1497 if (routing || target_is_router)
1498 break; /* send IMMEDIATE */
1500 /* is the REPLY message too small for RDMA? */
1501 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[lntmsg->msg_md->md_length]);
1502 if (nob <= IBLND_MSG_SIZE)
1503 break; /* send IMMEDIATE */
1505 tx = kiblnd_get_idle_tx(ni, target.nid);
1507 CERROR("Can't allocate txd for GET to %s\n",
1508 libcfs_nid2str(target.nid));
1514 if ((lntmsg->msg_md->md_options & LNET_MD_KIOV) == 0)
1515 rc = kiblnd_setup_rd_iov(ni, tx,
1516 &ibmsg->ibm_u.get.ibgm_rd,
1517 lntmsg->msg_md->md_niov,
1518 lntmsg->msg_md->md_iov.iov,
1519 0, lntmsg->msg_md->md_length);
1521 rc = kiblnd_setup_rd_kiov(ni, tx,
1522 &ibmsg->ibm_u.get.ibgm_rd,
1523 lntmsg->msg_md->md_niov,
1524 lntmsg->msg_md->md_iov.kiov,
1525 0, lntmsg->msg_md->md_length);
1527 CERROR("Can't setup GET sink for %s: %d\n",
1528 libcfs_nid2str(target.nid), rc);
1529 kiblnd_tx_done(ni, tx);
1533 nob = offsetof(kib_get_msg_t, ibgm_rd.rd_frags[tx->tx_nfrags]);
1534 ibmsg->ibm_u.get.ibgm_cookie = tx->tx_cookie;
1535 ibmsg->ibm_u.get.ibgm_hdr = *hdr;
1537 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_GET_REQ, nob);
1539 tx->tx_lntmsg[1] = lnet_create_reply_msg(ni, lntmsg);
1540 if (tx->tx_lntmsg[1] == NULL) {
1541 CERROR("Can't create reply for GET -> %s\n",
1542 libcfs_nid2str(target.nid));
1543 kiblnd_tx_done(ni, tx);
1547 tx->tx_lntmsg[0] = lntmsg; /* finalise lntmsg[0,1] on completion */
1548 tx->tx_waiting = 1; /* waiting for GET_DONE */
1549 kiblnd_launch_tx(ni, tx, target.nid);
1552 case LNET_MSG_REPLY:
1554 /* Is the payload small enough not to need RDMA? */
1555 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob]);
1556 if (nob <= IBLND_MSG_SIZE)
1557 break; /* send IMMEDIATE */
1559 tx = kiblnd_get_idle_tx(ni, target.nid);
1561 CERROR("Can't allocate %s txd for %s\n",
1562 type == LNET_MSG_PUT ? "PUT" : "REPLY",
1563 libcfs_nid2str(target.nid));
1567 if (payload_kiov == NULL)
1568 rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1569 payload_niov, payload_iov,
1570 payload_offset, payload_nob);
1572 rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1573 payload_niov, payload_kiov,
1574 payload_offset, payload_nob);
1576 CERROR("Can't setup PUT src for %s: %d\n",
1577 libcfs_nid2str(target.nid), rc);
1578 kiblnd_tx_done(ni, tx);
1583 ibmsg->ibm_u.putreq.ibprm_hdr = *hdr;
1584 ibmsg->ibm_u.putreq.ibprm_cookie = tx->tx_cookie;
1585 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_REQ, sizeof(kib_putreq_msg_t));
1587 tx->tx_lntmsg[0] = lntmsg; /* finalise lntmsg on completion */
1588 tx->tx_waiting = 1; /* waiting for PUT_{ACK,NAK} */
1589 kiblnd_launch_tx(ni, tx, target.nid);
1593 /* send IMMEDIATE */
1595 LASSERT (offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob])
1598 tx = kiblnd_get_idle_tx(ni, target.nid);
1600 CERROR ("Can't send %d to %s: tx descs exhausted\n",
1601 type, libcfs_nid2str(target.nid));
1606 ibmsg->ibm_u.immediate.ibim_hdr = *hdr;
1608 if (payload_kiov != NULL)
1609 lnet_copy_kiov2flat(IBLND_MSG_SIZE, ibmsg,
1610 offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1611 payload_niov, payload_kiov,
1612 payload_offset, payload_nob);
1614 lnet_copy_iov2flat(IBLND_MSG_SIZE, ibmsg,
1615 offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1616 payload_niov, payload_iov,
1617 payload_offset, payload_nob);
1619 nob = offsetof(kib_immediate_msg_t, ibim_payload[payload_nob]);
1620 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_IMMEDIATE, nob);
1622 tx->tx_lntmsg[0] = lntmsg; /* finalise lntmsg on completion */
1623 kiblnd_launch_tx(ni, tx, target.nid);
1628 kiblnd_reply (lnet_ni_t *ni, kib_rx_t *rx, lnet_msg_t *lntmsg)
1630 lnet_process_id_t target = lntmsg->msg_target;
1631 unsigned int niov = lntmsg->msg_niov;
1632 struct iovec *iov = lntmsg->msg_iov;
1633 lnet_kiov_t *kiov = lntmsg->msg_kiov;
1634 unsigned int offset = lntmsg->msg_offset;
1635 unsigned int nob = lntmsg->msg_len;
1639 tx = kiblnd_get_idle_tx(ni, rx->rx_conn->ibc_peer->ibp_nid);
1641 CERROR("Can't get tx for REPLY to %s\n",
1642 libcfs_nid2str(target.nid));
1648 else if (kiov == NULL)
1649 rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1650 niov, iov, offset, nob);
1652 rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1653 niov, kiov, offset, nob);
1656 CERROR("Can't setup GET src for %s: %d\n",
1657 libcfs_nid2str(target.nid), rc);
1661 rc = kiblnd_init_rdma(rx->rx_conn, tx,
1662 IBLND_MSG_GET_DONE, nob,
1663 &rx->rx_msg->ibm_u.get.ibgm_rd,
1664 rx->rx_msg->ibm_u.get.ibgm_cookie);
1666 CERROR("Can't setup rdma for GET from %s: %d\n",
1667 libcfs_nid2str(target.nid), rc);
1672 /* No RDMA: local completion may happen now! */
1673 lnet_finalize(ni, lntmsg, 0);
1675 /* RDMA: lnet_finalize(lntmsg) when it
1677 tx->tx_lntmsg[0] = lntmsg;
1680 kiblnd_queue_tx(tx, rx->rx_conn);
1684 kiblnd_tx_done(ni, tx);
1686 lnet_finalize(ni, lntmsg, -EIO);
1690 kiblnd_recv (lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed,
1691 unsigned int niov, struct iovec *iov, lnet_kiov_t *kiov,
1692 unsigned int offset, unsigned int mlen, unsigned int rlen)
1694 kib_rx_t *rx = private;
1695 kib_msg_t *rxmsg = rx->rx_msg;
1696 kib_conn_t *conn = rx->rx_conn;
1700 int post_credit = IBLND_POSTRX_PEER_CREDIT;
1703 LASSERT (mlen <= rlen);
1704 LASSERT (!in_interrupt());
1705 /* Either all pages or all vaddrs */
1706 LASSERT (!(kiov != NULL && iov != NULL));
1708 switch (rxmsg->ibm_type) {
1712 case IBLND_MSG_IMMEDIATE:
1713 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[rlen]);
1714 if (nob > rx->rx_nob) {
1715 CERROR ("Immediate message from %s too big: %d(%d)\n",
1716 libcfs_nid2str(rxmsg->ibm_u.immediate.ibim_hdr.src_nid),
1723 lnet_copy_flat2kiov(niov, kiov, offset,
1724 IBLND_MSG_SIZE, rxmsg,
1725 offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1728 lnet_copy_flat2iov(niov, iov, offset,
1729 IBLND_MSG_SIZE, rxmsg,
1730 offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1732 lnet_finalize (ni, lntmsg, 0);
1735 case IBLND_MSG_PUT_REQ:
1737 lnet_finalize(ni, lntmsg, 0);
1738 kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, 0,
1739 rxmsg->ibm_u.putreq.ibprm_cookie);
1743 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
1745 CERROR("Can't allocate tx for %s\n",
1746 libcfs_nid2str(conn->ibc_peer->ibp_nid));
1747 /* Not replying will break the connection */
1754 rc = kiblnd_setup_rd_iov(ni, tx,
1755 &txmsg->ibm_u.putack.ibpam_rd,
1756 niov, iov, offset, mlen);
1758 rc = kiblnd_setup_rd_kiov(ni, tx,
1759 &txmsg->ibm_u.putack.ibpam_rd,
1760 niov, kiov, offset, mlen);
1762 CERROR("Can't setup PUT sink for %s: %d\n",
1763 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
1764 kiblnd_tx_done(ni, tx);
1765 /* tell peer it's over */
1766 kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, rc,
1767 rxmsg->ibm_u.putreq.ibprm_cookie);
1771 nob = offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[tx->tx_nfrags]);
1772 txmsg->ibm_u.putack.ibpam_src_cookie = rxmsg->ibm_u.putreq.ibprm_cookie;
1773 txmsg->ibm_u.putack.ibpam_dst_cookie = tx->tx_cookie;
1775 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_ACK, nob);
1777 tx->tx_lntmsg[0] = lntmsg; /* finalise lntmsg on completion */
1778 tx->tx_waiting = 1; /* waiting for PUT_DONE */
1779 kiblnd_queue_tx(tx, conn);
1781 /* reposted buffer reserved for PUT_DONE */
1782 post_credit = IBLND_POSTRX_NO_CREDIT;
1785 case IBLND_MSG_GET_REQ:
1786 if (lntmsg != NULL) {
1787 /* Optimized GET; RDMA lntmsg's payload */
1788 kiblnd_reply(ni, rx, lntmsg);
1790 /* GET didn't match anything */
1791 kiblnd_send_completion(rx->rx_conn, IBLND_MSG_GET_DONE,
1793 rxmsg->ibm_u.get.ibgm_cookie);
1798 kiblnd_post_rx(rx, post_credit);
1803 kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name)
1805 struct task_struct *task = kthread_run(fn, arg, "%s", name);
1808 return PTR_ERR(task);
1810 atomic_inc(&kiblnd_data.kib_nthreads);
1815 kiblnd_thread_fini (void)
1817 atomic_dec (&kiblnd_data.kib_nthreads);
1821 kiblnd_peer_alive (kib_peer_t *peer)
1823 /* This is racy, but everyone's only writing cfs_time_current() */
1824 peer->ibp_last_alive = cfs_time_current();
1829 kiblnd_peer_notify (kib_peer_t *peer)
1832 unsigned long last_alive = 0;
1833 unsigned long flags;
1835 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1837 if (list_empty(&peer->ibp_conns) &&
1838 peer->ibp_accepting == 0 &&
1839 peer->ibp_connecting == 0 &&
1840 peer->ibp_error != 0) {
1841 error = peer->ibp_error;
1842 peer->ibp_error = 0;
1844 last_alive = peer->ibp_last_alive;
1847 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1850 lnet_notify(peer->ibp_ni,
1851 peer->ibp_nid, 0, last_alive);
1855 kiblnd_close_conn_locked (kib_conn_t *conn, int error)
1857 /* This just does the immediate housekeeping. 'error' is zero for a
1858 * normal shutdown which can happen only after the connection has been
1859 * established. If the connection is established, schedule the
1860 * connection to be finished off by the connd. Otherwise the connd is
1861 * already dealing with it (either to set it up or tear it down).
1862 * Caller holds kib_global_lock exclusively in irq context */
1863 kib_peer_t *peer = conn->ibc_peer;
1865 unsigned long flags;
1867 LASSERT (error != 0 || conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1869 if (error != 0 && conn->ibc_comms_error == 0)
1870 conn->ibc_comms_error = error;
1872 if (conn->ibc_state != IBLND_CONN_ESTABLISHED)
1873 return; /* already being handled */
1876 list_empty(&conn->ibc_tx_noops) &&
1877 list_empty(&conn->ibc_tx_queue) &&
1878 list_empty(&conn->ibc_tx_queue_rsrvd) &&
1879 list_empty(&conn->ibc_tx_queue_nocred) &&
1880 list_empty(&conn->ibc_active_txs)) {
1881 CDEBUG(D_NET, "closing conn to %s\n",
1882 libcfs_nid2str(peer->ibp_nid));
1884 CNETERR("Closing conn to %s: error %d%s%s%s%s%s\n",
1885 libcfs_nid2str(peer->ibp_nid), error,
1886 list_empty(&conn->ibc_tx_queue) ? "" : "(sending)",
1887 list_empty(&conn->ibc_tx_noops) ? "" : "(sending_noops)",
1888 list_empty(&conn->ibc_tx_queue_rsrvd) ? "" : "(sending_rsrvd)",
1889 list_empty(&conn->ibc_tx_queue_nocred) ? "" : "(sending_nocred)",
1890 list_empty(&conn->ibc_active_txs) ? "" : "(waiting)");
1893 dev = ((kib_net_t *)peer->ibp_ni->ni_data)->ibn_dev;
1894 list_del(&conn->ibc_list);
1895 /* connd (see below) takes over ibc_list's ref */
1897 if (list_empty (&peer->ibp_conns) && /* no more conns */
1898 kiblnd_peer_active(peer)) { /* still in peer table */
1899 kiblnd_unlink_peer_locked(peer);
1901 /* set/clear error on last conn */
1902 peer->ibp_error = conn->ibc_comms_error;
1905 kiblnd_set_conn_state(conn, IBLND_CONN_CLOSING);
1908 kiblnd_dev_can_failover(dev)) {
1909 list_add_tail(&dev->ibd_fail_list,
1910 &kiblnd_data.kib_failed_devs);
1911 wake_up(&kiblnd_data.kib_failover_waitq);
1914 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
1916 list_add_tail(&conn->ibc_list, &kiblnd_data.kib_connd_conns);
1917 wake_up(&kiblnd_data.kib_connd_waitq);
1919 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
1923 kiblnd_close_conn(kib_conn_t *conn, int error)
1925 unsigned long flags;
1927 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1929 kiblnd_close_conn_locked(conn, error);
1931 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1935 kiblnd_handle_early_rxs(kib_conn_t *conn)
1937 unsigned long flags;
1940 LASSERT(!in_interrupt());
1941 LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1943 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1944 while (!list_empty(&conn->ibc_early_rxs)) {
1945 rx = list_entry(conn->ibc_early_rxs.next,
1947 list_del(&rx->rx_list);
1948 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1950 kiblnd_handle_rx(rx);
1952 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1954 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1958 kiblnd_abort_txs(kib_conn_t *conn, struct list_head *txs)
1960 LIST_HEAD (zombies);
1961 struct list_head *tmp;
1962 struct list_head *nxt;
1965 spin_lock(&conn->ibc_lock);
1967 list_for_each_safe (tmp, nxt, txs) {
1968 tx = list_entry (tmp, kib_tx_t, tx_list);
1970 if (txs == &conn->ibc_active_txs) {
1971 LASSERT (!tx->tx_queued);
1972 LASSERT (tx->tx_waiting ||
1973 tx->tx_sending != 0);
1975 LASSERT (tx->tx_queued);
1978 tx->tx_status = -ECONNABORTED;
1981 if (tx->tx_sending == 0) {
1983 list_del (&tx->tx_list);
1984 list_add (&tx->tx_list, &zombies);
1988 spin_unlock(&conn->ibc_lock);
1990 kiblnd_txlist_done(conn->ibc_peer->ibp_ni, &zombies, -ECONNABORTED);
1994 kiblnd_finalise_conn (kib_conn_t *conn)
1996 LASSERT (!in_interrupt());
1997 LASSERT (conn->ibc_state > IBLND_CONN_INIT);
1999 kiblnd_set_conn_state(conn, IBLND_CONN_DISCONNECTED);
2001 /* abort_receives moves QP state to IB_QPS_ERR. This is only required
2002 * for connections that didn't get as far as being connected, because
2003 * rdma_disconnect() does this for free. */
2004 kiblnd_abort_receives(conn);
2006 /* Complete all tx descs not waiting for sends to complete.
2007 * NB we should be safe from RDMA now that the QP has changed state */
2009 kiblnd_abort_txs(conn, &conn->ibc_tx_noops);
2010 kiblnd_abort_txs(conn, &conn->ibc_tx_queue);
2011 kiblnd_abort_txs(conn, &conn->ibc_tx_queue_rsrvd);
2012 kiblnd_abort_txs(conn, &conn->ibc_tx_queue_nocred);
2013 kiblnd_abort_txs(conn, &conn->ibc_active_txs);
2015 kiblnd_handle_early_rxs(conn);
2019 kiblnd_peer_connect_failed (kib_peer_t *peer, int active, int error)
2021 LIST_HEAD (zombies);
2022 unsigned long flags;
2024 LASSERT (error != 0);
2025 LASSERT (!in_interrupt());
2027 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2030 LASSERT (peer->ibp_connecting > 0);
2031 peer->ibp_connecting--;
2033 LASSERT (peer->ibp_accepting > 0);
2034 peer->ibp_accepting--;
2037 if (peer->ibp_connecting != 0 ||
2038 peer->ibp_accepting != 0) {
2039 /* another connection attempt under way... */
2040 write_unlock_irqrestore(&kiblnd_data.kib_global_lock,
2045 if (list_empty(&peer->ibp_conns)) {
2046 /* Take peer's blocked transmits to complete with error */
2047 list_add(&zombies, &peer->ibp_tx_queue);
2048 list_del_init(&peer->ibp_tx_queue);
2050 if (kiblnd_peer_active(peer))
2051 kiblnd_unlink_peer_locked(peer);
2053 peer->ibp_error = error;
2055 /* Can't have blocked transmits if there are connections */
2056 LASSERT (list_empty(&peer->ibp_tx_queue));
2059 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2061 kiblnd_peer_notify(peer);
2063 if (list_empty (&zombies))
2066 CNETERR("Deleting messages for %s: connection failed\n",
2067 libcfs_nid2str(peer->ibp_nid));
2069 kiblnd_txlist_done(peer->ibp_ni, &zombies, -EHOSTUNREACH);
2073 kiblnd_connreq_done(kib_conn_t *conn, int status)
2075 kib_peer_t *peer = conn->ibc_peer;
2077 struct list_head txs;
2078 unsigned long flags;
2081 active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2083 CDEBUG(D_NET, "%s: active(%d), version(%x), status(%d)\n",
2084 libcfs_nid2str(peer->ibp_nid), active,
2085 conn->ibc_version, status);
2087 LASSERT (!in_interrupt());
2088 LASSERT ((conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT &&
2089 peer->ibp_connecting > 0) ||
2090 (conn->ibc_state == IBLND_CONN_PASSIVE_WAIT &&
2091 peer->ibp_accepting > 0));
2093 LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
2094 conn->ibc_connvars = NULL;
2097 /* failed to establish connection */
2098 kiblnd_peer_connect_failed(peer, active, status);
2099 kiblnd_finalise_conn(conn);
2103 /* connection established */
2104 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2106 conn->ibc_last_send = jiffies;
2107 kiblnd_set_conn_state(conn, IBLND_CONN_ESTABLISHED);
2108 kiblnd_peer_alive(peer);
2110 /* Add conn to peer's list and nuke any dangling conns from a different
2111 * peer instance... */
2112 kiblnd_conn_addref(conn); /* +1 ref for ibc_list */
2113 list_add(&conn->ibc_list, &peer->ibp_conns);
2115 peer->ibp_connecting--;
2117 peer->ibp_accepting--;
2119 if (peer->ibp_version == 0) {
2120 peer->ibp_version = conn->ibc_version;
2121 peer->ibp_incarnation = conn->ibc_incarnation;
2124 if (peer->ibp_version != conn->ibc_version ||
2125 peer->ibp_incarnation != conn->ibc_incarnation) {
2126 kiblnd_close_stale_conns_locked(peer, conn->ibc_version,
2127 conn->ibc_incarnation);
2128 peer->ibp_version = conn->ibc_version;
2129 peer->ibp_incarnation = conn->ibc_incarnation;
2132 /* grab pending txs while I have the lock */
2133 list_add(&txs, &peer->ibp_tx_queue);
2134 list_del_init(&peer->ibp_tx_queue);
2136 if (!kiblnd_peer_active(peer) || /* peer has been deleted */
2137 conn->ibc_comms_error != 0) { /* error has happened already */
2138 lnet_ni_t *ni = peer->ibp_ni;
2140 /* start to shut down connection */
2141 kiblnd_close_conn_locked(conn, -ECONNABORTED);
2142 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2144 kiblnd_txlist_done(ni, &txs, -ECONNABORTED);
2149 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2151 /* Schedule blocked txs */
2152 spin_lock(&conn->ibc_lock);
2153 while (!list_empty(&txs)) {
2154 tx = list_entry(txs.next, kib_tx_t, tx_list);
2155 list_del(&tx->tx_list);
2157 kiblnd_queue_tx_locked(tx, conn);
2159 spin_unlock(&conn->ibc_lock);
2161 kiblnd_check_sends(conn);
2163 /* schedule blocked rxs */
2164 kiblnd_handle_early_rxs(conn);
2168 kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej)
2172 rc = rdma_reject(cmid, rej, sizeof(*rej));
2175 CWARN("Error %d sending reject\n", rc);
2179 kiblnd_passive_connect (struct rdma_cm_id *cmid, void *priv, int priv_nob)
2181 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
2182 kib_msg_t *reqmsg = priv;
2188 lnet_ni_t *ni = NULL;
2189 kib_net_t *net = NULL;
2191 struct rdma_conn_param cp;
2193 int version = IBLND_MSG_VERSION;
2194 unsigned long flags;
2196 struct sockaddr_in *peer_addr;
2197 LASSERT (!in_interrupt());
2199 /* cmid inherits 'context' from the corresponding listener id */
2200 ibdev = (kib_dev_t *)cmid->context;
2201 LASSERT (ibdev != NULL);
2203 memset(&rej, 0, sizeof(rej));
2204 rej.ibr_magic = IBLND_MSG_MAGIC;
2205 rej.ibr_why = IBLND_REJECT_FATAL;
2206 rej.ibr_cp.ibcp_max_msg_size = IBLND_MSG_SIZE;
2208 peer_addr = (struct sockaddr_in *)&(cmid->route.addr.dst_addr);
2209 if (*kiblnd_tunables.kib_require_priv_port &&
2210 ntohs(peer_addr->sin_port) >= PROT_SOCK) {
2211 __u32 ip = ntohl(peer_addr->sin_addr.s_addr);
2212 CERROR("Peer's port (%pI4h:%hu) is not privileged\n",
2213 &ip, ntohs(peer_addr->sin_port));
2217 if (priv_nob < offsetof(kib_msg_t, ibm_type)) {
2218 CERROR("Short connection request\n");
2222 /* Future protocol version compatibility support! If the
2223 * o2iblnd-specific protocol changes, or when LNET unifies
2224 * protocols over all LNDs, the initial connection will
2225 * negotiate a protocol version. I trap this here to avoid
2226 * console errors; the reject tells the peer which protocol I
2228 if (reqmsg->ibm_magic == LNET_PROTO_MAGIC ||
2229 reqmsg->ibm_magic == __swab32(LNET_PROTO_MAGIC))
2231 if (reqmsg->ibm_magic == IBLND_MSG_MAGIC &&
2232 reqmsg->ibm_version != IBLND_MSG_VERSION &&
2233 reqmsg->ibm_version != IBLND_MSG_VERSION_1)
2235 if (reqmsg->ibm_magic == __swab32(IBLND_MSG_MAGIC) &&
2236 reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION) &&
2237 reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION_1))
2240 rc = kiblnd_unpack_msg(reqmsg, priv_nob);
2242 CERROR("Can't parse connection request: %d\n", rc);
2246 nid = reqmsg->ibm_srcnid;
2247 ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid));
2250 net = (kib_net_t *)ni->ni_data;
2251 rej.ibr_incarnation = net->ibn_incarnation;
2254 if (ni == NULL || /* no matching net */
2255 ni->ni_nid != reqmsg->ibm_dstnid || /* right NET, wrong NID! */
2256 net->ibn_dev != ibdev) { /* wrong device */
2257 CERROR("Can't accept %s on %s (%s:%d:%pI4h): "
2258 "bad dst nid %s\n", libcfs_nid2str(nid),
2259 ni == NULL ? "NA" : libcfs_nid2str(ni->ni_nid),
2260 ibdev->ibd_ifname, ibdev->ibd_nnets,
2262 libcfs_nid2str(reqmsg->ibm_dstnid));
2267 /* check time stamp as soon as possible */
2268 if (reqmsg->ibm_dststamp != 0 &&
2269 reqmsg->ibm_dststamp != net->ibn_incarnation) {
2270 CWARN("Stale connection request\n");
2271 rej.ibr_why = IBLND_REJECT_CONN_STALE;
2275 /* I can accept peer's version */
2276 version = reqmsg->ibm_version;
2278 if (reqmsg->ibm_type != IBLND_MSG_CONNREQ) {
2279 CERROR("Unexpected connreq msg type: %x from %s\n",
2280 reqmsg->ibm_type, libcfs_nid2str(nid));
2284 if (reqmsg->ibm_u.connparams.ibcp_queue_depth !=
2285 IBLND_MSG_QUEUE_SIZE(version)) {
2286 CERROR("Can't accept %s: incompatible queue depth %d (%d wanted)\n",
2287 libcfs_nid2str(nid), reqmsg->ibm_u.connparams.ibcp_queue_depth,
2288 IBLND_MSG_QUEUE_SIZE(version));
2290 if (version == IBLND_MSG_VERSION)
2291 rej.ibr_why = IBLND_REJECT_MSG_QUEUE_SIZE;
2296 if (reqmsg->ibm_u.connparams.ibcp_max_frags !=
2297 IBLND_RDMA_FRAGS(version)) {
2298 CERROR("Can't accept %s(version %x): "
2299 "incompatible max_frags %d (%d wanted)\n",
2300 libcfs_nid2str(nid), version,
2301 reqmsg->ibm_u.connparams.ibcp_max_frags,
2302 IBLND_RDMA_FRAGS(version));
2304 if (version == IBLND_MSG_VERSION)
2305 rej.ibr_why = IBLND_REJECT_RDMA_FRAGS;
2311 if (reqmsg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2312 CERROR("Can't accept %s: message size %d too big (%d max)\n",
2313 libcfs_nid2str(nid),
2314 reqmsg->ibm_u.connparams.ibcp_max_msg_size,
2319 /* assume 'nid' is a new peer; create */
2320 rc = kiblnd_create_peer(ni, &peer, nid);
2322 CERROR("Can't create peer for %s\n", libcfs_nid2str(nid));
2323 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2327 write_lock_irqsave(g_lock, flags);
2329 peer2 = kiblnd_find_peer_locked(nid);
2330 if (peer2 != NULL) {
2331 if (peer2->ibp_version == 0) {
2332 peer2->ibp_version = version;
2333 peer2->ibp_incarnation = reqmsg->ibm_srcstamp;
2336 /* not the guy I've talked with */
2337 if (peer2->ibp_incarnation != reqmsg->ibm_srcstamp ||
2338 peer2->ibp_version != version) {
2339 kiblnd_close_peer_conns_locked(peer2, -ESTALE);
2340 write_unlock_irqrestore(g_lock, flags);
2342 CWARN("Conn stale %s [old ver: %x, new ver: %x]\n",
2343 libcfs_nid2str(nid), peer2->ibp_version, version);
2345 kiblnd_peer_decref(peer);
2346 rej.ibr_why = IBLND_REJECT_CONN_STALE;
2350 /* tie-break connection race in favour of the higher NID */
2351 if (peer2->ibp_connecting != 0 &&
2353 write_unlock_irqrestore(g_lock, flags);
2355 CWARN("Conn race %s\n", libcfs_nid2str(peer2->ibp_nid));
2357 kiblnd_peer_decref(peer);
2358 rej.ibr_why = IBLND_REJECT_CONN_RACE;
2362 peer2->ibp_accepting++;
2363 kiblnd_peer_addref(peer2);
2365 write_unlock_irqrestore(g_lock, flags);
2366 kiblnd_peer_decref(peer);
2369 /* Brand new peer */
2370 LASSERT (peer->ibp_accepting == 0);
2371 LASSERT (peer->ibp_version == 0 &&
2372 peer->ibp_incarnation == 0);
2374 peer->ibp_accepting = 1;
2375 peer->ibp_version = version;
2376 peer->ibp_incarnation = reqmsg->ibm_srcstamp;
2378 /* I have a ref on ni that prevents it being shutdown */
2379 LASSERT (net->ibn_shutdown == 0);
2381 kiblnd_peer_addref(peer);
2382 list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
2384 write_unlock_irqrestore(g_lock, flags);
2387 conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_PASSIVE_WAIT, version);
2389 kiblnd_peer_connect_failed(peer, 0, -ENOMEM);
2390 kiblnd_peer_decref(peer);
2391 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2395 /* conn now "owns" cmid, so I return success from here on to ensure the
2396 * CM callback doesn't destroy cmid. */
2398 conn->ibc_incarnation = reqmsg->ibm_srcstamp;
2399 conn->ibc_credits = IBLND_MSG_QUEUE_SIZE(version);
2400 conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(version);
2401 LASSERT (conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(version)
2402 <= IBLND_RX_MSGS(version));
2404 ackmsg = &conn->ibc_connvars->cv_msg;
2405 memset(ackmsg, 0, sizeof(*ackmsg));
2407 kiblnd_init_msg(ackmsg, IBLND_MSG_CONNACK,
2408 sizeof(ackmsg->ibm_u.connparams));
2409 ackmsg->ibm_u.connparams.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2410 ackmsg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2411 ackmsg->ibm_u.connparams.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
2413 kiblnd_pack_msg(ni, ackmsg, version, 0, nid, reqmsg->ibm_srcstamp);
2415 memset(&cp, 0, sizeof(cp));
2416 cp.private_data = ackmsg;
2417 cp.private_data_len = ackmsg->ibm_nob;
2418 cp.responder_resources = 0; /* No atomic ops or RDMA reads */
2419 cp.initiator_depth = 0;
2420 cp.flow_control = 1;
2421 cp.retry_count = *kiblnd_tunables.kib_retry_count;
2422 cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
2424 CDEBUG(D_NET, "Accept %s\n", libcfs_nid2str(nid));
2426 rc = rdma_accept(cmid, &cp);
2428 CERROR("Can't accept %s: %d\n", libcfs_nid2str(nid), rc);
2429 rej.ibr_version = version;
2430 rej.ibr_why = IBLND_REJECT_FATAL;
2432 kiblnd_reject(cmid, &rej);
2433 kiblnd_connreq_done(conn, rc);
2434 kiblnd_conn_decref(conn);
2444 rej.ibr_version = version;
2445 rej.ibr_cp.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2446 rej.ibr_cp.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
2447 kiblnd_reject(cmid, &rej);
2449 return -ECONNREFUSED;
2453 kiblnd_reconnect (kib_conn_t *conn, int version,
2454 __u64 incarnation, int why, kib_connparams_t *cp)
2456 kib_peer_t *peer = conn->ibc_peer;
2459 unsigned long flags;
2461 LASSERT (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2462 LASSERT (peer->ibp_connecting > 0); /* 'conn' at least */
2464 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2466 /* retry connection if it's still needed and no other connection
2467 * attempts (active or passive) are in progress
2468 * NB: reconnect is still needed even when ibp_tx_queue is
2469 * empty if ibp_version != version because reconnect may be
2470 * initiated by kiblnd_query() */
2471 if ((!list_empty(&peer->ibp_tx_queue) ||
2472 peer->ibp_version != version) &&
2473 peer->ibp_connecting == 1 &&
2474 peer->ibp_accepting == 0) {
2476 peer->ibp_connecting++;
2478 peer->ibp_version = version;
2479 peer->ibp_incarnation = incarnation;
2482 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2492 case IBLND_REJECT_CONN_STALE:
2496 case IBLND_REJECT_CONN_RACE:
2497 reason = "conn race";
2500 case IBLND_REJECT_CONN_UNCOMPAT:
2501 reason = "version negotiation";
2505 CNETERR("%s: retrying (%s), %x, %x, "
2506 "queue_dep: %d, max_frag: %d, msg_size: %d\n",
2507 libcfs_nid2str(peer->ibp_nid),
2508 reason, IBLND_MSG_VERSION, version,
2509 cp != NULL? cp->ibcp_queue_depth :IBLND_MSG_QUEUE_SIZE(version),
2510 cp != NULL? cp->ibcp_max_frags : IBLND_RDMA_FRAGS(version),
2511 cp != NULL? cp->ibcp_max_msg_size: IBLND_MSG_SIZE);
2513 kiblnd_connect_peer(peer);
2517 kiblnd_rejected (kib_conn_t *conn, int reason, void *priv, int priv_nob)
2519 kib_peer_t *peer = conn->ibc_peer;
2521 LASSERT (!in_interrupt());
2522 LASSERT (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2525 case IB_CM_REJ_STALE_CONN:
2526 kiblnd_reconnect(conn, IBLND_MSG_VERSION, 0,
2527 IBLND_REJECT_CONN_STALE, NULL);
2530 case IB_CM_REJ_INVALID_SERVICE_ID:
2531 CNETERR("%s rejected: no listener at %d\n",
2532 libcfs_nid2str(peer->ibp_nid),
2533 *kiblnd_tunables.kib_service);
2536 case IB_CM_REJ_CONSUMER_DEFINED:
2537 if (priv_nob >= offsetof(kib_rej_t, ibr_padding)) {
2538 kib_rej_t *rej = priv;
2539 kib_connparams_t *cp = NULL;
2541 __u64 incarnation = -1;
2543 /* NB. default incarnation is -1 because:
2544 * a) V1 will ignore dst incarnation in connreq.
2545 * b) V2 will provide incarnation while rejecting me,
2546 * -1 will be overwrote.
2548 * if I try to connect to a V1 peer with V2 protocol,
2549 * it rejected me then upgrade to V2, I have no idea
2550 * about the upgrading and try to reconnect with V1,
2551 * in this case upgraded V2 can find out I'm trying to
2552 * talk to the old guy and reject me(incarnation is -1).
2555 if (rej->ibr_magic == __swab32(IBLND_MSG_MAGIC) ||
2556 rej->ibr_magic == __swab32(LNET_PROTO_MAGIC)) {
2557 __swab32s(&rej->ibr_magic);
2558 __swab16s(&rej->ibr_version);
2562 if (priv_nob >= sizeof(kib_rej_t) &&
2563 rej->ibr_version > IBLND_MSG_VERSION_1) {
2564 /* priv_nob is always 148 in current version
2565 * of OFED, so we still need to check version.
2566 * (define of IB_CM_REJ_PRIVATE_DATA_SIZE) */
2570 __swab64s(&rej->ibr_incarnation);
2571 __swab16s(&cp->ibcp_queue_depth);
2572 __swab16s(&cp->ibcp_max_frags);
2573 __swab32s(&cp->ibcp_max_msg_size);
2576 incarnation = rej->ibr_incarnation;
2579 if (rej->ibr_magic != IBLND_MSG_MAGIC &&
2580 rej->ibr_magic != LNET_PROTO_MAGIC) {
2581 CERROR("%s rejected: consumer defined fatal error\n",
2582 libcfs_nid2str(peer->ibp_nid));
2586 if (rej->ibr_version != IBLND_MSG_VERSION &&
2587 rej->ibr_version != IBLND_MSG_VERSION_1) {
2588 CERROR("%s rejected: o2iblnd version %x error\n",
2589 libcfs_nid2str(peer->ibp_nid),
2594 if (rej->ibr_why == IBLND_REJECT_FATAL &&
2595 rej->ibr_version == IBLND_MSG_VERSION_1) {
2596 CDEBUG(D_NET, "rejected by old version peer %s: %x\n",
2597 libcfs_nid2str(peer->ibp_nid), rej->ibr_version);
2599 if (conn->ibc_version != IBLND_MSG_VERSION_1)
2600 rej->ibr_why = IBLND_REJECT_CONN_UNCOMPAT;
2603 switch (rej->ibr_why) {
2604 case IBLND_REJECT_CONN_RACE:
2605 case IBLND_REJECT_CONN_STALE:
2606 case IBLND_REJECT_CONN_UNCOMPAT:
2607 kiblnd_reconnect(conn, rej->ibr_version,
2608 incarnation, rej->ibr_why, cp);
2611 case IBLND_REJECT_MSG_QUEUE_SIZE:
2612 CERROR("%s rejected: incompatible message queue depth %d, %d\n",
2613 libcfs_nid2str(peer->ibp_nid),
2614 cp != NULL ? cp->ibcp_queue_depth :
2615 IBLND_MSG_QUEUE_SIZE(rej->ibr_version),
2616 IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
2619 case IBLND_REJECT_RDMA_FRAGS:
2620 CERROR("%s rejected: incompatible # of RDMA fragments %d, %d\n",
2621 libcfs_nid2str(peer->ibp_nid),
2622 cp != NULL ? cp->ibcp_max_frags :
2623 IBLND_RDMA_FRAGS(rej->ibr_version),
2624 IBLND_RDMA_FRAGS(conn->ibc_version));
2627 case IBLND_REJECT_NO_RESOURCES:
2628 CERROR("%s rejected: o2iblnd no resources\n",
2629 libcfs_nid2str(peer->ibp_nid));
2632 case IBLND_REJECT_FATAL:
2633 CERROR("%s rejected: o2iblnd fatal error\n",
2634 libcfs_nid2str(peer->ibp_nid));
2638 CERROR("%s rejected: o2iblnd reason %d\n",
2639 libcfs_nid2str(peer->ibp_nid),
2647 CNETERR("%s rejected: reason %d, size %d\n",
2648 libcfs_nid2str(peer->ibp_nid), reason, priv_nob);
2652 kiblnd_connreq_done(conn, -ECONNREFUSED);
2656 kiblnd_check_connreply (kib_conn_t *conn, void *priv, int priv_nob)
2658 kib_peer_t *peer = conn->ibc_peer;
2659 lnet_ni_t *ni = peer->ibp_ni;
2660 kib_net_t *net = ni->ni_data;
2661 kib_msg_t *msg = priv;
2662 int ver = conn->ibc_version;
2663 int rc = kiblnd_unpack_msg(msg, priv_nob);
2664 unsigned long flags;
2666 LASSERT (net != NULL);
2669 CERROR("Can't unpack connack from %s: %d\n",
2670 libcfs_nid2str(peer->ibp_nid), rc);
2674 if (msg->ibm_type != IBLND_MSG_CONNACK) {
2675 CERROR("Unexpected message %d from %s\n",
2676 msg->ibm_type, libcfs_nid2str(peer->ibp_nid));
2681 if (ver != msg->ibm_version) {
2682 CERROR("%s replied version %x is different with "
2683 "requested version %x\n",
2684 libcfs_nid2str(peer->ibp_nid), msg->ibm_version, ver);
2689 if (msg->ibm_u.connparams.ibcp_queue_depth !=
2690 IBLND_MSG_QUEUE_SIZE(ver)) {
2691 CERROR("%s has incompatible queue depth %d(%d wanted)\n",
2692 libcfs_nid2str(peer->ibp_nid),
2693 msg->ibm_u.connparams.ibcp_queue_depth,
2694 IBLND_MSG_QUEUE_SIZE(ver));
2699 if (msg->ibm_u.connparams.ibcp_max_frags !=
2700 IBLND_RDMA_FRAGS(ver)) {
2701 CERROR("%s has incompatible max_frags %d (%d wanted)\n",
2702 libcfs_nid2str(peer->ibp_nid),
2703 msg->ibm_u.connparams.ibcp_max_frags,
2704 IBLND_RDMA_FRAGS(ver));
2709 if (msg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2710 CERROR("%s max message size %d too big (%d max)\n",
2711 libcfs_nid2str(peer->ibp_nid),
2712 msg->ibm_u.connparams.ibcp_max_msg_size,
2718 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2719 if (msg->ibm_dstnid == ni->ni_nid &&
2720 msg->ibm_dststamp == net->ibn_incarnation)
2724 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2727 CERROR("Bad connection reply from %s, rc = %d, "
2728 "version: %x max_frags: %d\n",
2729 libcfs_nid2str(peer->ibp_nid), rc,
2730 msg->ibm_version, msg->ibm_u.connparams.ibcp_max_frags);
2734 conn->ibc_incarnation = msg->ibm_srcstamp;
2736 conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(ver);
2737 LASSERT (conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(ver)
2738 <= IBLND_RX_MSGS(ver));
2740 kiblnd_connreq_done(conn, 0);
2744 /* NB My QP has already established itself, so I handle anything going
2745 * wrong here by setting ibc_comms_error.
2746 * kiblnd_connreq_done(0) moves the conn state to ESTABLISHED, but then
2747 * immediately tears it down. */
2750 conn->ibc_comms_error = rc;
2751 kiblnd_connreq_done(conn, 0);
2755 kiblnd_active_connect (struct rdma_cm_id *cmid)
2757 kib_peer_t *peer = (kib_peer_t *)cmid->context;
2760 struct rdma_conn_param cp;
2763 unsigned long flags;
2766 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2768 incarnation = peer->ibp_incarnation;
2769 version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION :
2772 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2774 conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_ACTIVE_CONNECT, version);
2776 kiblnd_peer_connect_failed(peer, 1, -ENOMEM);
2777 kiblnd_peer_decref(peer); /* lose cmid's ref */
2781 /* conn "owns" cmid now, so I return success from here on to ensure the
2782 * CM callback doesn't destroy cmid. conn also takes over cmid's ref
2785 msg = &conn->ibc_connvars->cv_msg;
2787 memset(msg, 0, sizeof(*msg));
2788 kiblnd_init_msg(msg, IBLND_MSG_CONNREQ, sizeof(msg->ibm_u.connparams));
2789 msg->ibm_u.connparams.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2790 msg->ibm_u.connparams.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
2791 msg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2793 kiblnd_pack_msg(peer->ibp_ni, msg, version,
2794 0, peer->ibp_nid, incarnation);
2796 memset(&cp, 0, sizeof(cp));
2797 cp.private_data = msg;
2798 cp.private_data_len = msg->ibm_nob;
2799 cp.responder_resources = 0; /* No atomic ops or RDMA reads */
2800 cp.initiator_depth = 0;
2801 cp.flow_control = 1;
2802 cp.retry_count = *kiblnd_tunables.kib_retry_count;
2803 cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
2805 LASSERT(cmid->context == (void *)conn);
2806 LASSERT(conn->ibc_cmid == cmid);
2808 rc = rdma_connect(cmid, &cp);
2810 CERROR("Can't connect to %s: %d\n",
2811 libcfs_nid2str(peer->ibp_nid), rc);
2812 kiblnd_connreq_done(conn, rc);
2813 kiblnd_conn_decref(conn);
2820 kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event)
2826 switch (event->event) {
2828 CERROR("Unexpected event: %d, status: %d\n",
2829 event->event, event->status);
2832 case RDMA_CM_EVENT_CONNECT_REQUEST:
2833 /* destroy cmid on failure */
2834 rc = kiblnd_passive_connect(cmid,
2835 (void *)KIBLND_CONN_PARAM(event),
2836 KIBLND_CONN_PARAM_LEN(event));
2837 CDEBUG(D_NET, "connreq: %d\n", rc);
2840 case RDMA_CM_EVENT_ADDR_ERROR:
2841 peer = (kib_peer_t *)cmid->context;
2842 CNETERR("%s: ADDR ERROR %d\n",
2843 libcfs_nid2str(peer->ibp_nid), event->status);
2844 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2845 kiblnd_peer_decref(peer);
2846 return -EHOSTUNREACH; /* rc != 0 destroys cmid */
2848 case RDMA_CM_EVENT_ADDR_RESOLVED:
2849 peer = (kib_peer_t *)cmid->context;
2851 CDEBUG(D_NET, "%s Addr resolved: %d\n",
2852 libcfs_nid2str(peer->ibp_nid), event->status);
2854 if (event->status != 0) {
2855 CNETERR("Can't resolve address for %s: %d\n",
2856 libcfs_nid2str(peer->ibp_nid), event->status);
2859 rc = rdma_resolve_route(
2860 cmid, *kiblnd_tunables.kib_timeout * 1000);
2863 /* Can't initiate route resolution */
2864 CERROR("Can't resolve route for %s: %d\n",
2865 libcfs_nid2str(peer->ibp_nid), rc);
2867 kiblnd_peer_connect_failed(peer, 1, rc);
2868 kiblnd_peer_decref(peer);
2869 return rc; /* rc != 0 destroys cmid */
2871 case RDMA_CM_EVENT_ROUTE_ERROR:
2872 peer = (kib_peer_t *)cmid->context;
2873 CNETERR("%s: ROUTE ERROR %d\n",
2874 libcfs_nid2str(peer->ibp_nid), event->status);
2875 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2876 kiblnd_peer_decref(peer);
2877 return -EHOSTUNREACH; /* rc != 0 destroys cmid */
2879 case RDMA_CM_EVENT_ROUTE_RESOLVED:
2880 peer = (kib_peer_t *)cmid->context;
2881 CDEBUG(D_NET, "%s Route resolved: %d\n",
2882 libcfs_nid2str(peer->ibp_nid), event->status);
2884 if (event->status == 0)
2885 return kiblnd_active_connect(cmid);
2887 CNETERR("Can't resolve route for %s: %d\n",
2888 libcfs_nid2str(peer->ibp_nid), event->status);
2889 kiblnd_peer_connect_failed(peer, 1, event->status);
2890 kiblnd_peer_decref(peer);
2891 return event->status; /* rc != 0 destroys cmid */
2893 case RDMA_CM_EVENT_UNREACHABLE:
2894 conn = (kib_conn_t *)cmid->context;
2895 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
2896 conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
2897 CNETERR("%s: UNREACHABLE %d\n",
2898 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
2899 kiblnd_connreq_done(conn, -ENETDOWN);
2900 kiblnd_conn_decref(conn);
2903 case RDMA_CM_EVENT_CONNECT_ERROR:
2904 conn = (kib_conn_t *)cmid->context;
2905 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
2906 conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
2907 CNETERR("%s: CONNECT ERROR %d\n",
2908 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
2909 kiblnd_connreq_done(conn, -ENOTCONN);
2910 kiblnd_conn_decref(conn);
2913 case RDMA_CM_EVENT_REJECTED:
2914 conn = (kib_conn_t *)cmid->context;
2915 switch (conn->ibc_state) {
2919 case IBLND_CONN_PASSIVE_WAIT:
2920 CERROR ("%s: REJECTED %d\n",
2921 libcfs_nid2str(conn->ibc_peer->ibp_nid),
2923 kiblnd_connreq_done(conn, -ECONNRESET);
2926 case IBLND_CONN_ACTIVE_CONNECT:
2927 kiblnd_rejected(conn, event->status,
2928 (void *)KIBLND_CONN_PARAM(event),
2929 KIBLND_CONN_PARAM_LEN(event));
2932 kiblnd_conn_decref(conn);
2935 case RDMA_CM_EVENT_ESTABLISHED:
2936 conn = (kib_conn_t *)cmid->context;
2937 switch (conn->ibc_state) {
2941 case IBLND_CONN_PASSIVE_WAIT:
2942 CDEBUG(D_NET, "ESTABLISHED (passive): %s\n",
2943 libcfs_nid2str(conn->ibc_peer->ibp_nid));
2944 kiblnd_connreq_done(conn, 0);
2947 case IBLND_CONN_ACTIVE_CONNECT:
2948 CDEBUG(D_NET, "ESTABLISHED(active): %s\n",
2949 libcfs_nid2str(conn->ibc_peer->ibp_nid));
2950 kiblnd_check_connreply(conn,
2951 (void *)KIBLND_CONN_PARAM(event),
2952 KIBLND_CONN_PARAM_LEN(event));
2955 /* net keeps its ref on conn! */
2958 case RDMA_CM_EVENT_TIMEWAIT_EXIT:
2959 CDEBUG(D_NET, "Ignore TIMEWAIT_EXIT event\n");
2961 case RDMA_CM_EVENT_DISCONNECTED:
2962 conn = (kib_conn_t *)cmid->context;
2963 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
2964 CERROR("%s DISCONNECTED\n",
2965 libcfs_nid2str(conn->ibc_peer->ibp_nid));
2966 kiblnd_connreq_done(conn, -ECONNRESET);
2968 kiblnd_close_conn(conn, 0);
2970 kiblnd_conn_decref(conn);
2971 cmid->context = NULL;
2974 case RDMA_CM_EVENT_DEVICE_REMOVAL:
2975 LCONSOLE_ERROR_MSG(0x131,
2976 "Received notification of device removal\n"
2977 "Please shutdown LNET to allow this to proceed\n");
2978 /* Can't remove network from underneath LNET for now, so I have
2982 case RDMA_CM_EVENT_ADDR_CHANGE:
2983 LCONSOLE_INFO("Physical link changed (eg hca/port)\n");
2989 kiblnd_check_txs_locked(kib_conn_t *conn, struct list_head *txs)
2992 struct list_head *ttmp;
2994 list_for_each (ttmp, txs) {
2995 tx = list_entry (ttmp, kib_tx_t, tx_list);
2997 if (txs != &conn->ibc_active_txs) {
2998 LASSERT (tx->tx_queued);
3000 LASSERT (!tx->tx_queued);
3001 LASSERT (tx->tx_waiting || tx->tx_sending != 0);
3004 if (cfs_time_aftereq (jiffies, tx->tx_deadline)) {
3005 CERROR("Timed out tx: %s, %lu seconds\n",
3006 kiblnd_queue2str(conn, txs),
3007 cfs_duration_sec(jiffies - tx->tx_deadline));
3016 kiblnd_conn_timed_out_locked(kib_conn_t *conn)
3018 return kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue) ||
3019 kiblnd_check_txs_locked(conn, &conn->ibc_tx_noops) ||
3020 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_rsrvd) ||
3021 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_nocred) ||
3022 kiblnd_check_txs_locked(conn, &conn->ibc_active_txs);
3026 kiblnd_check_conns (int idx)
3029 LIST_HEAD (checksends);
3030 struct list_head *peers = &kiblnd_data.kib_peers[idx];
3031 struct list_head *ptmp;
3034 struct list_head *ctmp;
3035 unsigned long flags;
3037 /* NB. We expect to have a look at all the peers and not find any
3038 * RDMAs to time out, so we just use a shared lock while we
3040 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
3042 list_for_each (ptmp, peers) {
3043 peer = list_entry (ptmp, kib_peer_t, ibp_list);
3045 list_for_each (ctmp, &peer->ibp_conns) {
3049 conn = list_entry(ctmp, kib_conn_t, ibc_list);
3051 LASSERT (conn->ibc_state == IBLND_CONN_ESTABLISHED);
3053 spin_lock(&conn->ibc_lock);
3055 sendnoop = kiblnd_need_noop(conn);
3056 timedout = kiblnd_conn_timed_out_locked(conn);
3057 if (!sendnoop && !timedout) {
3058 spin_unlock(&conn->ibc_lock);
3063 CERROR("Timed out RDMA with %s (%lu): "
3064 "c: %u, oc: %u, rc: %u\n",
3065 libcfs_nid2str(peer->ibp_nid),
3066 cfs_duration_sec(cfs_time_current() -
3067 peer->ibp_last_alive),
3069 conn->ibc_outstanding_credits,
3070 conn->ibc_reserved_credits);
3071 list_add(&conn->ibc_connd_list, &closes);
3073 list_add(&conn->ibc_connd_list,
3076 /* +ref for 'closes' or 'checksends' */
3077 kiblnd_conn_addref(conn);
3079 spin_unlock(&conn->ibc_lock);
3083 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
3085 /* Handle timeout by closing the whole
3086 * connection. We can only be sure RDMA activity
3087 * has ceased once the QP has been modified. */
3088 while (!list_empty(&closes)) {
3089 conn = list_entry(closes.next,
3090 kib_conn_t, ibc_connd_list);
3091 list_del(&conn->ibc_connd_list);
3092 kiblnd_close_conn(conn, -ETIMEDOUT);
3093 kiblnd_conn_decref(conn);
3096 /* In case we have enough credits to return via a
3097 * NOOP, but there were no non-blocking tx descs
3098 * free to do it last time... */
3099 while (!list_empty(&checksends)) {
3100 conn = list_entry(checksends.next,
3101 kib_conn_t, ibc_connd_list);
3102 list_del(&conn->ibc_connd_list);
3103 kiblnd_check_sends(conn);
3104 kiblnd_conn_decref(conn);
3109 kiblnd_disconnect_conn (kib_conn_t *conn)
3111 LASSERT (!in_interrupt());
3112 LASSERT (current == kiblnd_data.kib_connd);
3113 LASSERT (conn->ibc_state == IBLND_CONN_CLOSING);
3115 rdma_disconnect(conn->ibc_cmid);
3116 kiblnd_finalise_conn(conn);
3118 kiblnd_peer_notify(conn->ibc_peer);
3122 kiblnd_connd (void *arg)
3125 unsigned long flags;
3131 unsigned long deadline = jiffies;
3133 cfs_block_allsigs ();
3135 init_waitqueue_entry(&wait, current);
3136 kiblnd_data.kib_connd = current;
3138 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3140 while (!kiblnd_data.kib_shutdown) {
3144 if (!list_empty (&kiblnd_data.kib_connd_zombies)) {
3145 conn = list_entry(kiblnd_data. \
3146 kib_connd_zombies.next,
3147 kib_conn_t, ibc_list);
3148 list_del(&conn->ibc_list);
3150 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
3154 kiblnd_destroy_conn(conn);
3156 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3159 if (!list_empty(&kiblnd_data.kib_connd_conns)) {
3160 conn = list_entry(kiblnd_data.kib_connd_conns.next,
3161 kib_conn_t, ibc_list);
3162 list_del(&conn->ibc_list);
3164 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
3168 kiblnd_disconnect_conn(conn);
3169 kiblnd_conn_decref(conn);
3171 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3174 /* careful with the jiffy wrap... */
3175 timeout = (int)(deadline - jiffies);
3179 int chunk = kiblnd_data.kib_peer_hash_size;
3181 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3184 /* Time to check for RDMA timeouts on a few more
3185 * peers: I do checks every 'p' seconds on a
3186 * proportion of the peer table and I need to check
3187 * every connection 'n' times within a timeout
3188 * interval, to ensure I detect a timeout on any
3189 * connection within (n+1)/n times the timeout
3192 if (*kiblnd_tunables.kib_timeout > n * p)
3193 chunk = (chunk * n * p) /
3194 *kiblnd_tunables.kib_timeout;
3198 for (i = 0; i < chunk; i++) {
3199 kiblnd_check_conns(peer_index);
3200 peer_index = (peer_index + 1) %
3201 kiblnd_data.kib_peer_hash_size;
3205 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3211 /* Nothing to do for 'timeout' */
3212 set_current_state(TASK_INTERRUPTIBLE);
3213 add_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3214 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3216 schedule_timeout(timeout);
3218 remove_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3219 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3222 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3224 kiblnd_thread_fini();
3229 kiblnd_qp_event(struct ib_event *event, void *arg)
3231 kib_conn_t *conn = arg;
3233 switch (event->event) {
3234 case IB_EVENT_COMM_EST:
3235 CDEBUG(D_NET, "%s established\n",
3236 libcfs_nid2str(conn->ibc_peer->ibp_nid));
3240 CERROR("%s: Async QP event type %d\n",
3241 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3247 kiblnd_complete (struct ib_wc *wc)
3249 switch (kiblnd_wreqid2type(wc->wr_id)) {
3253 case IBLND_WID_RDMA:
3254 /* We only get RDMA completion notification if it fails. All
3255 * subsequent work items, including the final SEND will fail
3256 * too. However we can't print out any more info about the
3257 * failing RDMA because 'tx' might be back on the idle list or
3258 * even reused already if we didn't manage to post all our work
3260 CNETERR("RDMA (tx: %p) failed: %d\n",
3261 kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3265 kiblnd_tx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3269 kiblnd_rx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status,
3276 kiblnd_cq_completion(struct ib_cq *cq, void *arg)
3278 /* NB I'm not allowed to schedule this conn once its refcount has
3279 * reached 0. Since fundamentally I'm racing with scheduler threads
3280 * consuming my CQ I could be called after all completions have
3281 * occurred. But in this case, ibc_nrx == 0 && ibc_nsends_posted == 0
3282 * and this CQ is about to be destroyed so I NOOP. */
3283 kib_conn_t *conn = (kib_conn_t *)arg;
3284 struct kib_sched_info *sched = conn->ibc_sched;
3285 unsigned long flags;
3287 LASSERT(cq == conn->ibc_cq);
3289 spin_lock_irqsave(&sched->ibs_lock, flags);
3291 conn->ibc_ready = 1;
3293 if (!conn->ibc_scheduled &&
3294 (conn->ibc_nrx > 0 ||
3295 conn->ibc_nsends_posted > 0)) {
3296 kiblnd_conn_addref(conn); /* +1 ref for sched_conns */
3297 conn->ibc_scheduled = 1;
3298 list_add_tail(&conn->ibc_sched_list, &sched->ibs_conns);
3300 if (waitqueue_active(&sched->ibs_waitq))
3301 wake_up(&sched->ibs_waitq);
3304 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3308 kiblnd_cq_event(struct ib_event *event, void *arg)
3310 kib_conn_t *conn = arg;
3312 CERROR("%s: async CQ event type %d\n",
3313 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3317 kiblnd_scheduler(void *arg)
3319 long id = (long)arg;
3320 struct kib_sched_info *sched;
3323 unsigned long flags;
3329 cfs_block_allsigs();
3331 init_waitqueue_entry(&wait, current);
3333 sched = kiblnd_data.kib_scheds[KIB_THREAD_CPT(id)];
3335 rc = cfs_cpt_bind(lnet_cpt_table(), sched->ibs_cpt);
3337 CWARN("Failed to bind on CPT %d, please verify whether "
3338 "all CPUs are healthy and reload modules if necessary, "
3339 "otherwise your system might under risk of low "
3340 "performance\n", sched->ibs_cpt);
3343 spin_lock_irqsave(&sched->ibs_lock, flags);
3345 while (!kiblnd_data.kib_shutdown) {
3346 if (busy_loops++ >= IBLND_RESCHED) {
3347 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3352 spin_lock_irqsave(&sched->ibs_lock, flags);
3357 if (!list_empty(&sched->ibs_conns)) {
3358 conn = list_entry(sched->ibs_conns.next,
3359 kib_conn_t, ibc_sched_list);
3360 /* take over kib_sched_conns' ref on conn... */
3361 LASSERT(conn->ibc_scheduled);
3362 list_del(&conn->ibc_sched_list);
3363 conn->ibc_ready = 0;
3365 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3367 rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3369 rc = ib_req_notify_cq(conn->ibc_cq,
3372 CWARN("%s: ib_req_notify_cq failed: %d, "
3373 "closing connection\n",
3374 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
3375 kiblnd_close_conn(conn, -EIO);
3376 kiblnd_conn_decref(conn);
3377 spin_lock_irqsave(&sched->ibs_lock,
3382 rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3386 CWARN("%s: ib_poll_cq failed: %d, "
3387 "closing connection\n",
3388 libcfs_nid2str(conn->ibc_peer->ibp_nid),
3390 kiblnd_close_conn(conn, -EIO);
3391 kiblnd_conn_decref(conn);
3392 spin_lock_irqsave(&sched->ibs_lock, flags);
3396 spin_lock_irqsave(&sched->ibs_lock, flags);
3398 if (rc != 0 || conn->ibc_ready) {
3399 /* There may be another completion waiting; get
3400 * another scheduler to check while I handle
3402 /* +1 ref for sched_conns */
3403 kiblnd_conn_addref(conn);
3404 list_add_tail(&conn->ibc_sched_list,
3406 if (waitqueue_active(&sched->ibs_waitq))
3407 wake_up(&sched->ibs_waitq);
3409 conn->ibc_scheduled = 0;
3413 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3414 kiblnd_complete(&wc);
3416 spin_lock_irqsave(&sched->ibs_lock, flags);
3419 kiblnd_conn_decref(conn); /* ...drop my ref from above */
3426 set_current_state(TASK_INTERRUPTIBLE);
3427 add_wait_queue_exclusive(&sched->ibs_waitq, &wait);
3428 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3433 remove_wait_queue(&sched->ibs_waitq, &wait);
3434 spin_lock_irqsave(&sched->ibs_lock, flags);
3437 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3439 kiblnd_thread_fini();
3444 kiblnd_failover_thread(void *arg)
3446 rwlock_t *glock = &kiblnd_data.kib_global_lock;
3449 unsigned long flags;
3452 LASSERT (*kiblnd_tunables.kib_dev_failover != 0);
3454 cfs_block_allsigs ();
3456 init_waitqueue_entry(&wait, current);
3457 write_lock_irqsave(glock, flags);
3459 while (!kiblnd_data.kib_shutdown) {
3460 int do_failover = 0;
3463 list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
3465 if (time_before(cfs_time_current(),
3466 dev->ibd_next_failover))
3473 list_del_init(&dev->ibd_fail_list);
3474 dev->ibd_failover = 1;
3475 write_unlock_irqrestore(glock, flags);
3477 rc = kiblnd_dev_failover(dev);
3479 write_lock_irqsave(glock, flags);
3481 LASSERT (dev->ibd_failover);
3482 dev->ibd_failover = 0;
3483 if (rc >= 0) { /* Device is OK or failover succeed */
3484 dev->ibd_next_failover = cfs_time_shift(3);
3488 /* failed to failover, retry later */
3489 dev->ibd_next_failover =
3490 cfs_time_shift(min(dev->ibd_failed_failover, 10));
3491 if (kiblnd_dev_can_failover(dev)) {
3492 list_add_tail(&dev->ibd_fail_list,
3493 &kiblnd_data.kib_failed_devs);
3499 /* long sleep if no more pending failover */
3500 long_sleep = list_empty(&kiblnd_data.kib_failed_devs);
3502 set_current_state(TASK_INTERRUPTIBLE);
3503 add_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3504 write_unlock_irqrestore(glock, flags);
3506 rc = schedule_timeout(long_sleep ? cfs_time_seconds(10) :
3507 cfs_time_seconds(1));
3508 remove_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3509 write_lock_irqsave(glock, flags);
3511 if (!long_sleep || rc != 0)
3514 /* have a long sleep, routine check all active devices,
3515 * we need checking like this because if there is not active
3516 * connection on the dev and no SEND from local, we may listen
3517 * on wrong HCA for ever while there is a bonding failover */
3518 list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
3519 if (kiblnd_dev_can_failover(dev)) {
3520 list_add_tail(&dev->ibd_fail_list,
3521 &kiblnd_data.kib_failed_devs);
3526 write_unlock_irqrestore(glock, flags);
3528 kiblnd_thread_fini();