1 /*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
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.
9 *
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).
15 *
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
19 *
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
22 * have any questions.
23 *
24 * GPL HEADER END
25 */
26 /*
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
29 *
30 * Copyright (c) 2012, Intel Corporation.
31 */
32 /*
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
35 *
36 * lnet/klnds/o2iblnd/o2iblnd_cb.c
37 *
38 * Author: Eric Barton <eric@bartonsoftware.com>
39 */
40
41 #include "o2iblnd.h"
42
43 static void
kiblnd_tx_done(lnet_ni_t * ni,kib_tx_t * tx)44 kiblnd_tx_done(lnet_ni_t *ni, kib_tx_t *tx)
45 {
46 lnet_msg_t *lntmsg[2];
47 kib_net_t *net = ni->ni_data;
48 int rc;
49 int i;
50
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);
57
58 kiblnd_unmap_tx(ni, tx);
59
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;
63 rc = tx->tx_status;
64
65 if (tx->tx_conn != NULL) {
66 LASSERT(ni == tx->tx_conn->ibc_peer->ibp_ni);
67
68 kiblnd_conn_decref(tx->tx_conn);
69 tx->tx_conn = NULL;
70 }
71
72 tx->tx_nwrq = 0;
73 tx->tx_status = 0;
74
75 kiblnd_pool_free_node(&tx->tx_pool->tpo_pool, &tx->tx_list);
76
77 /* delay finalize until my descs have been freed */
78 for (i = 0; i < 2; i++) {
79 if (lntmsg[i] == NULL)
80 continue;
81
82 lnet_finalize(ni, lntmsg[i], rc);
83 }
84 }
85
86 void
kiblnd_txlist_done(lnet_ni_t * ni,struct list_head * txlist,int status)87 kiblnd_txlist_done(lnet_ni_t *ni, struct list_head *txlist, int status)
88 {
89 kib_tx_t *tx;
90
91 while (!list_empty(txlist)) {
92 tx = list_entry(txlist->next, kib_tx_t, tx_list);
93
94 list_del(&tx->tx_list);
95 /* complete now */
96 tx->tx_waiting = 0;
97 tx->tx_status = status;
98 kiblnd_tx_done(ni, tx);
99 }
100 }
101
102 static kib_tx_t *
kiblnd_get_idle_tx(lnet_ni_t * ni,lnet_nid_t target)103 kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target)
104 {
105 kib_net_t *net = (kib_net_t *)ni->ni_data;
106 struct list_head *node;
107 kib_tx_t *tx;
108 kib_tx_poolset_t *tps;
109
110 tps = net->ibn_tx_ps[lnet_cpt_of_nid(target)];
111 node = kiblnd_pool_alloc_node(&tps->tps_poolset);
112 if (node == NULL)
113 return NULL;
114 tx = container_of(node, kib_tx_t, tx_list);
115
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);
126
127 return tx;
128 }
129
130 static void
kiblnd_drop_rx(kib_rx_t * rx)131 kiblnd_drop_rx(kib_rx_t *rx)
132 {
133 kib_conn_t *conn = rx->rx_conn;
134 struct kib_sched_info *sched = conn->ibc_sched;
135 unsigned long flags;
136
137 spin_lock_irqsave(&sched->ibs_lock, flags);
138 LASSERT(conn->ibc_nrx > 0);
139 conn->ibc_nrx--;
140 spin_unlock_irqrestore(&sched->ibs_lock, flags);
141
142 kiblnd_conn_decref(conn);
143 }
144
145 int
kiblnd_post_rx(kib_rx_t * rx,int credit)146 kiblnd_post_rx(kib_rx_t *rx, int credit)
147 {
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;
151 struct ib_mr *mr;
152 int rc;
153
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);
159
160 mr = kiblnd_find_dma_mr(conn->ibc_hdev, rx->rx_msgaddr, IBLND_MSG_SIZE);
161 LASSERT(mr != NULL);
162
163 rx->rx_sge.lkey = mr->lkey;
164 rx->rx_sge.addr = rx->rx_msgaddr;
165 rx->rx_sge.length = IBLND_MSG_SIZE;
166
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);
171
172 LASSERT(conn->ibc_state >= IBLND_CONN_INIT);
173 LASSERT(rx->rx_nob >= 0); /* not posted */
174
175 if (conn->ibc_state > IBLND_CONN_ESTABLISHED) {
176 kiblnd_drop_rx(rx); /* No more posts for this rx */
177 return 0;
178 }
179
180 rx->rx_nob = -1; /* flag posted */
181
182 rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq);
183 if (rc != 0) {
184 CERROR("Can't post rx for %s: %d, bad_wrq: %p\n",
185 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc, bad_wrq);
186 rx->rx_nob = 0;
187 }
188
189 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */
190 return rc;
191
192 if (rc != 0) {
193 kiblnd_close_conn(conn, rc);
194 kiblnd_drop_rx(rx); /* No more posts for this rx */
195 return rc;
196 }
197
198 if (credit == IBLND_POSTRX_NO_CREDIT)
199 return 0;
200
201 spin_lock(&conn->ibc_lock);
202 if (credit == IBLND_POSTRX_PEER_CREDIT)
203 conn->ibc_outstanding_credits++;
204 else
205 conn->ibc_reserved_credits++;
206 spin_unlock(&conn->ibc_lock);
207
208 kiblnd_check_sends(conn);
209 return 0;
210 }
211
212 static kib_tx_t *
kiblnd_find_waiting_tx_locked(kib_conn_t * conn,int txtype,__u64 cookie)213 kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie)
214 {
215 struct list_head *tmp;
216
217 list_for_each(tmp, &conn->ibc_active_txs) {
218 kib_tx_t *tx = list_entry(tmp, kib_tx_t, tx_list);
219
220 LASSERT(!tx->tx_queued);
221 LASSERT(tx->tx_sending != 0 || tx->tx_waiting);
222
223 if (tx->tx_cookie != cookie)
224 continue;
225
226 if (tx->tx_waiting &&
227 tx->tx_msg->ibm_type == txtype)
228 return tx;
229
230 CWARN("Bad completion: %swaiting, type %x (wanted %x)\n",
231 tx->tx_waiting ? "" : "NOT ",
232 tx->tx_msg->ibm_type, txtype);
233 }
234 return NULL;
235 }
236
237 static void
kiblnd_handle_completion(kib_conn_t * conn,int txtype,int status,__u64 cookie)238 kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie)
239 {
240 kib_tx_t *tx;
241 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
242 int idle;
243
244 spin_lock(&conn->ibc_lock);
245
246 tx = kiblnd_find_waiting_tx_locked(conn, txtype, cookie);
247 if (tx == NULL) {
248 spin_unlock(&conn->ibc_lock);
249
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);
253 return;
254 }
255
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);
261 }
262 }
263
264 tx->tx_waiting = 0;
265
266 idle = !tx->tx_queued && (tx->tx_sending == 0);
267 if (idle)
268 list_del(&tx->tx_list);
269
270 spin_unlock(&conn->ibc_lock);
271
272 if (idle)
273 kiblnd_tx_done(ni, tx);
274 }
275
276 static void
kiblnd_send_completion(kib_conn_t * conn,int type,int status,__u64 cookie)277 kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie)
278 {
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);
281
282 if (tx == NULL) {
283 CERROR("Can't get tx for completion %x for %s\n",
284 type, libcfs_nid2str(conn->ibc_peer->ibp_nid));
285 return;
286 }
287
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));
291
292 kiblnd_queue_tx(tx, conn);
293 }
294
295 static void
kiblnd_handle_rx(kib_rx_t * rx)296 kiblnd_handle_rx(kib_rx_t *rx)
297 {
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;
302 kib_tx_t *tx;
303 int rc = 0;
304 int rc2;
305 int post_credit;
306
307 LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
308
309 CDEBUG(D_NET, "Received %x[%d] from %s\n",
310 msg->ibm_type, credits,
311 libcfs_nid2str(conn->ibc_peer->ibp_nid));
312
313 if (credits != 0) {
314 /* Have I received credits that will let me send? */
315 spin_lock(&conn->ibc_lock);
316
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);
321
322 CERROR("Bad credits from %s: %d + %d > %d\n",
323 libcfs_nid2str(conn->ibc_peer->ibp_nid),
324 rc2, credits,
325 IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
326
327 kiblnd_close_conn(conn, -EPROTO);
328 kiblnd_post_rx(rx, IBLND_POSTRX_NO_CREDIT);
329 return;
330 }
331
332 conn->ibc_credits += credits;
333
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++;
338
339 spin_unlock(&conn->ibc_lock);
340 kiblnd_check_sends(conn);
341 }
342
343 switch (msg->ibm_type) {
344 default:
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;
348 rc = -EPROTO;
349 break;
350
351 case IBLND_MSG_NOOP:
352 if (IBLND_OOB_CAPABLE(conn->ibc_version)) {
353 post_credit = IBLND_POSTRX_NO_CREDIT;
354 break;
355 }
356
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;
361 break;
362
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;
369 break;
370
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;
377 break;
378
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);
386 break;
387
388 case IBLND_MSG_PUT_ACK:
389 post_credit = IBLND_POSTRX_RSRVD_CREDIT;
390
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);
394 if (tx != NULL)
395 list_del(&tx->tx_list);
396 spin_unlock(&conn->ibc_lock);
397
398 if (tx == NULL) {
399 CERROR("Unmatched PUT_ACK from %s\n",
400 libcfs_nid2str(conn->ibc_peer->ibp_nid));
401 rc = -EPROTO;
402 break;
403 }
404
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. */
409
410 tx->tx_nwrq = 0; /* overwrite PUT_REQ */
411
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);
416 if (rc2 < 0)
417 CERROR("Can't setup rdma for PUT to %s: %d\n",
418 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc2);
419
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);
424 break;
425
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);
431 break;
432
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;
439 break;
440
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);
446 break;
447 }
448
449 if (rc < 0) /* protocol error */
450 kiblnd_close_conn(conn, rc);
451
452 if (post_credit != IBLND_POSTRX_DONT_POST)
453 kiblnd_post_rx(rx, post_credit);
454 }
455
456 static void
kiblnd_rx_complete(kib_rx_t * rx,int status,int nob)457 kiblnd_rx_complete(kib_rx_t *rx, int status, int nob)
458 {
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;
463 int rc;
464 int err = -EIO;
465
466 LASSERT(net != NULL);
467 LASSERT(rx->rx_nob < 0); /* was posted */
468 rx->rx_nob = 0; /* isn't now */
469
470 if (conn->ibc_state > IBLND_CONN_ESTABLISHED)
471 goto ignore;
472
473 if (status != IB_WC_SUCCESS) {
474 CNETERR("Rx from %s failed: %d\n",
475 libcfs_nid2str(conn->ibc_peer->ibp_nid), status);
476 goto failed;
477 }
478
479 LASSERT(nob >= 0);
480 rx->rx_nob = nob;
481
482 rc = kiblnd_unpack_msg(msg, rx->rx_nob);
483 if (rc != 0) {
484 CERROR("Error %d unpacking rx from %s\n",
485 rc, libcfs_nid2str(conn->ibc_peer->ibp_nid));
486 goto failed;
487 }
488
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));
495 err = -ESTALE;
496 goto failed;
497 }
498
499 /* set time last known alive */
500 kiblnd_peer_alive(conn->ibc_peer);
501
502 /* racing with connection establishment/teardown! */
503
504 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
505 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
506 unsigned long flags;
507
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);
513 return;
514 }
515 write_unlock_irqrestore(g_lock, flags);
516 }
517 kiblnd_handle_rx(rx);
518 return;
519
520 failed:
521 CDEBUG(D_NET, "rx %p conn %p\n", rx, conn);
522 kiblnd_close_conn(conn, err);
523 ignore:
524 kiblnd_drop_rx(rx); /* Don't re-post rx. */
525 }
526
527 static struct page *
kiblnd_kvaddr_to_page(unsigned long vaddr)528 kiblnd_kvaddr_to_page(unsigned long vaddr)
529 {
530 struct page *page;
531
532 if (is_vmalloc_addr((void *)vaddr)) {
533 page = vmalloc_to_page((void *)vaddr);
534 LASSERT(page != NULL);
535 return page;
536 }
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");
542 LBUG();
543 }
544 #endif
545 page = virt_to_page(vaddr);
546 LASSERT(page != NULL);
547 return page;
548 }
549
550 static int
kiblnd_fmr_map_tx(kib_net_t * net,kib_tx_t * tx,kib_rdma_desc_t * rd,int nob)551 kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
552 {
553 kib_hca_dev_t *hdev;
554 __u64 *pages = tx->tx_pages;
555 kib_fmr_poolset_t *fps;
556 int npages;
557 int size;
558 int cpt;
559 int rc;
560 int i;
561
562 LASSERT(tx->tx_pool != NULL);
563 LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
564
565 hdev = tx->tx_pool->tpo_hdev;
566
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;
572 }
573 }
574
575 cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
576
577 fps = net->ibn_fmr_ps[cpt];
578 rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->tx_u.fmr);
579 if (rc != 0) {
580 CERROR("Can't map %d pages: %d\n", npages, rc);
581 return rc;
582 }
583
584 /* If rd is not tx_rd, it's going to get sent to a peer, who will need
585 * the rkey */
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;
590 rd->rd_nfrags = 1;
591
592 return 0;
593 }
594
595 static int
kiblnd_pmr_map_tx(kib_net_t * net,kib_tx_t * tx,kib_rdma_desc_t * rd,int nob)596 kiblnd_pmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
597 {
598 kib_hca_dev_t *hdev;
599 kib_pmr_poolset_t *pps;
600 __u64 iova;
601 int cpt;
602 int rc;
603
604 LASSERT(tx->tx_pool != NULL);
605 LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
606
607 hdev = tx->tx_pool->tpo_hdev;
608
609 iova = rd->rd_frags[0].rf_addr & ~hdev->ibh_page_mask;
610
611 cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
612
613 pps = net->ibn_pmr_ps[cpt];
614 rc = kiblnd_pmr_pool_map(pps, hdev, rd, &iova, &tx->tx_u.pmr);
615 if (rc != 0) {
616 CERROR("Failed to create MR by phybuf: %d\n", rc);
617 return rc;
618 }
619
620 /* If rd is not tx_rd, it's going to get sent to a peer, who will need
621 * the rkey */
622 rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.pmr->pmr_mr->rkey :
623 tx->tx_u.pmr->pmr_mr->lkey;
624 rd->rd_nfrags = 1;
625 rd->rd_frags[0].rf_addr = iova;
626 rd->rd_frags[0].rf_nob = nob;
627
628 return 0;
629 }
630
631 void
kiblnd_unmap_tx(lnet_ni_t * ni,kib_tx_t * tx)632 kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx)
633 {
634 kib_net_t *net = ni->ni_data;
635
636 LASSERT(net != NULL);
637
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;
641
642 } else if (net->ibn_pmr_ps != NULL && tx->tx_u.pmr != NULL) {
643 kiblnd_pmr_pool_unmap(tx->tx_u.pmr);
644 tx->tx_u.pmr = NULL;
645 }
646
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);
650 tx->tx_nfrags = 0;
651 }
652 }
653
654 int
kiblnd_map_tx(lnet_ni_t * ni,kib_tx_t * tx,kib_rdma_desc_t * rd,int nfrags)655 kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx,
656 kib_rdma_desc_t *rd, int nfrags)
657 {
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;
661 __u32 nob;
662 int i;
663
664 /* If rd is not tx_rd, it's going to get sent to a peer and I'm the
665 * RDMA sink */
666 tx->tx_dmadir = (rd != tx->tx_rd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
667 tx->tx_nfrags = nfrags;
668
669 rd->rd_nfrags =
670 kiblnd_dma_map_sg(hdev->ibh_ibdev,
671 tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir);
672
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;
679 }
680
681 /* looking for pre-mapping MR */
682 mr = kiblnd_find_rd_dma_mr(hdev, rd);
683 if (mr != NULL) {
684 /* found pre-mapping MR */
685 rd->rd_key = (rd != tx->tx_rd) ? mr->rkey : mr->lkey;
686 return 0;
687 }
688
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);
693
694 return -EINVAL;
695 }
696
697
698 static int
kiblnd_setup_rd_iov(lnet_ni_t * ni,kib_tx_t * tx,kib_rdma_desc_t * rd,unsigned int niov,struct kvec * iov,int offset,int nob)699 kiblnd_setup_rd_iov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
700 unsigned int niov, struct kvec *iov, int offset, int nob)
701 {
702 kib_net_t *net = ni->ni_data;
703 struct page *page;
704 struct scatterlist *sg;
705 unsigned long vaddr;
706 int fragnob;
707 int page_offset;
708
709 LASSERT(nob > 0);
710 LASSERT(niov > 0);
711 LASSERT(net != NULL);
712
713 while (offset >= iov->iov_len) {
714 offset -= iov->iov_len;
715 niov--;
716 iov++;
717 LASSERT(niov > 0);
718 }
719
720 sg = tx->tx_frags;
721 do {
722 LASSERT(niov > 0);
723
724 vaddr = ((unsigned long)iov->iov_base) + offset;
725 page_offset = vaddr & (PAGE_SIZE - 1);
726 page = kiblnd_kvaddr_to_page(vaddr);
727 if (page == NULL) {
728 CERROR("Can't find page\n");
729 return -EFAULT;
730 }
731
732 fragnob = min((int)(iov->iov_len - offset), nob);
733 fragnob = min(fragnob, (int)PAGE_SIZE - page_offset);
734
735 sg_set_page(sg, page, fragnob, page_offset);
736 sg++;
737
738 if (offset + fragnob < iov->iov_len) {
739 offset += fragnob;
740 } else {
741 offset = 0;
742 iov++;
743 niov--;
744 }
745 nob -= fragnob;
746 } while (nob > 0);
747
748 return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
749 }
750
751 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)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)
754 {
755 kib_net_t *net = ni->ni_data;
756 struct scatterlist *sg;
757 int fragnob;
758
759 CDEBUG(D_NET, "niov %d offset %d nob %d\n", nkiov, offset, nob);
760
761 LASSERT(nob > 0);
762 LASSERT(nkiov > 0);
763 LASSERT(net != NULL);
764
765 while (offset >= kiov->kiov_len) {
766 offset -= kiov->kiov_len;
767 nkiov--;
768 kiov++;
769 LASSERT(nkiov > 0);
770 }
771
772 sg = tx->tx_frags;
773 do {
774 LASSERT(nkiov > 0);
775
776 fragnob = min((int)(kiov->kiov_len - offset), nob);
777
778 sg_set_page(sg, kiov->kiov_page, fragnob,
779 kiov->kiov_offset + offset);
780 sg++;
781
782 offset = 0;
783 kiov++;
784 nkiov--;
785 nob -= fragnob;
786 } while (nob > 0);
787
788 return kiblnd_map_tx(ni, tx, rd, sg - tx->tx_frags);
789 }
790
791 static int
kiblnd_post_tx_locked(kib_conn_t * conn,kib_tx_t * tx,int credit)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)
795 {
796 kib_msg_t *msg = tx->tx_msg;
797 kib_peer_t *peer = conn->ibc_peer;
798 int ver = conn->ibc_version;
799 int rc;
800 int done;
801 struct ib_send_wr *bad_wrq;
802
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));
807
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));
813
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));
818 return -EAGAIN;
819 }
820
821 if (credit != 0 && conn->ibc_credits == 0) { /* no credits */
822 CDEBUG(D_NET, "%s: no credits\n",
823 libcfs_nid2str(peer->ibp_nid));
824 return -EAGAIN;
825 }
826
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));
832 return -EAGAIN;
833 }
834
835 /* NB don't drop ibc_lock before bumping tx_sending */
836 list_del(&tx->tx_list);
837 tx->tx_queued = 0;
838
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);
852 return 0;
853 }
854
855 kiblnd_pack_msg(peer->ibp_ni, msg, ver, conn->ibc_outstanding_credits,
856 peer->ibp_nid, conn->ibc_incarnation);
857
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++;
863
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. */
870 tx->tx_sending++;
871 list_add(&tx->tx_list, &conn->ibc_active_txs);
872
873 /* I'm still holding ibc_lock! */
874 if (conn->ibc_state != IBLND_CONN_ESTABLISHED) {
875 rc = -ECONNABORTED;
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 */
879 rc = -ENETDOWN;
880 } else {
881 rc = ib_post_send(conn->ibc_cmid->qp,
882 tx->tx_wrq, &bad_wrq);
883 }
884
885 conn->ibc_last_send = jiffies;
886
887 if (rc == 0)
888 return 0;
889
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--;
897
898 tx->tx_status = rc;
899 tx->tx_waiting = 0;
900 tx->tx_sending--;
901
902 done = (tx->tx_sending == 0);
903 if (done)
904 list_del(&tx->tx_list);
905
906 spin_unlock(&conn->ibc_lock);
907
908 if (conn->ibc_state == IBLND_CONN_ESTABLISHED)
909 CERROR("Error %d posting transmit to %s\n",
910 rc, libcfs_nid2str(peer->ibp_nid));
911 else
912 CDEBUG(D_NET, "Error %d posting transmit to %s\n",
913 rc, libcfs_nid2str(peer->ibp_nid));
914
915 kiblnd_close_conn(conn, rc);
916
917 if (done)
918 kiblnd_tx_done(peer->ibp_ni, tx);
919
920 spin_lock(&conn->ibc_lock);
921
922 return -EIO;
923 }
924
925 void
kiblnd_check_sends(kib_conn_t * conn)926 kiblnd_check_sends(kib_conn_t *conn)
927 {
928 int ver = conn->ibc_version;
929 lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
930 kib_tx_t *tx;
931
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));
936 return;
937 }
938
939 spin_lock(&conn->ibc_lock);
940
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);
945
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,
949 kib_tx_t, tx_list);
950 list_del(&tx->tx_list);
951 list_add_tail(&tx->tx_list, &conn->ibc_tx_queue);
952 conn->ibc_reserved_credits--;
953 }
954
955 if (kiblnd_need_noop(conn)) {
956 spin_unlock(&conn->ibc_lock);
957
958 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
959 if (tx != NULL)
960 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_NOOP, 0);
961
962 spin_lock(&conn->ibc_lock);
963 if (tx != NULL)
964 kiblnd_queue_tx_locked(tx, conn);
965 }
966
967 kiblnd_conn_addref(conn); /* 1 ref for me.... (see b21911) */
968
969 for (;;) {
970 int credit;
971
972 if (!list_empty(&conn->ibc_tx_queue_nocred)) {
973 credit = 0;
974 tx = list_entry(conn->ibc_tx_queue_nocred.next,
975 kib_tx_t, tx_list);
976 } else if (!list_empty(&conn->ibc_tx_noops)) {
977 LASSERT(!IBLND_OOB_CAPABLE(ver));
978 credit = 1;
979 tx = list_entry(conn->ibc_tx_noops.next,
980 kib_tx_t, tx_list);
981 } else if (!list_empty(&conn->ibc_tx_queue)) {
982 credit = 1;
983 tx = list_entry(conn->ibc_tx_queue.next,
984 kib_tx_t, tx_list);
985 } else
986 break;
987
988 if (kiblnd_post_tx_locked(conn, tx, credit) != 0)
989 break;
990 }
991
992 spin_unlock(&conn->ibc_lock);
993
994 kiblnd_conn_decref(conn); /* ...until here */
995 }
996
997 static void
kiblnd_tx_complete(kib_tx_t * tx,int status)998 kiblnd_tx_complete(kib_tx_t *tx, int status)
999 {
1000 int failed = (status != IB_WC_SUCCESS);
1001 kib_conn_t *conn = tx->tx_conn;
1002 int idle;
1003
1004 LASSERT(tx->tx_sending > 0);
1005
1006 if (failed) {
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,
1011 status);
1012
1013 kiblnd_close_conn(conn, -EIO);
1014 } else {
1015 kiblnd_peer_alive(conn->ibc_peer);
1016 }
1017
1018 spin_lock(&conn->ibc_lock);
1019
1020 /* I could be racing with rdma completion. Whoever makes 'tx' idle
1021 * gets to free it, which also drops its ref on 'conn'. */
1022
1023 tx->tx_sending--;
1024 conn->ibc_nsends_posted--;
1025 if (tx->tx_msg->ibm_type == IBLND_MSG_NOOP)
1026 conn->ibc_noops_posted--;
1027
1028 if (failed) {
1029 tx->tx_waiting = 0; /* don't wait for peer */
1030 tx->tx_status = -EIO;
1031 }
1032
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) */
1036 if (idle)
1037 list_del(&tx->tx_list);
1038
1039 kiblnd_conn_addref(conn); /* 1 ref for me.... */
1040
1041 spin_unlock(&conn->ibc_lock);
1042
1043 if (idle)
1044 kiblnd_tx_done(conn->ibc_peer->ibp_ni, tx);
1045
1046 kiblnd_check_sends(conn);
1047
1048 kiblnd_conn_decref(conn); /* ...until here */
1049 }
1050
1051 void
kiblnd_init_tx_msg(lnet_ni_t * ni,kib_tx_t * tx,int type,int body_nob)1052 kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob)
1053 {
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;
1058 struct ib_mr *mr;
1059
1060 LASSERT(tx->tx_nwrq >= 0);
1061 LASSERT(tx->tx_nwrq < IBLND_MAX_RDMA_FRAGS + 1);
1062 LASSERT(nob <= IBLND_MSG_SIZE);
1063
1064 kiblnd_init_msg(tx->tx_msg, type, body_nob);
1065
1066 mr = kiblnd_find_dma_mr(hdev, tx->tx_msgaddr, nob);
1067 LASSERT(mr != NULL);
1068
1069 sge->lkey = mr->lkey;
1070 sge->addr = tx->tx_msgaddr;
1071 sge->length = nob;
1072
1073 memset(wrq, 0, sizeof(*wrq));
1074
1075 wrq->next = NULL;
1076 wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_TX);
1077 wrq->sg_list = sge;
1078 wrq->num_sge = 1;
1079 wrq->opcode = IB_WR_SEND;
1080 wrq->send_flags = IB_SEND_SIGNALED;
1081
1082 tx->tx_nwrq++;
1083 }
1084
1085 int
kiblnd_init_rdma(kib_conn_t * conn,kib_tx_t * tx,int type,int resid,kib_rdma_desc_t * dstrd,__u64 dstcookie)1086 kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type,
1087 int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie)
1088 {
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];
1093 int rc = resid;
1094 int srcidx;
1095 int dstidx;
1096 int wrknob;
1097
1098 LASSERT(!in_interrupt());
1099 LASSERT(tx->tx_nwrq == 0);
1100 LASSERT(type == IBLND_MSG_GET_DONE ||
1101 type == IBLND_MSG_PUT_DONE);
1102
1103 srcidx = dstidx = 0;
1104
1105 while (resid > 0) {
1106 if (srcidx >= srcrd->rd_nfrags) {
1107 CERROR("Src buffer exhausted: %d frags\n", srcidx);
1108 rc = -EPROTO;
1109 break;
1110 }
1111
1112 if (dstidx == dstrd->rd_nfrags) {
1113 CERROR("Dst buffer exhausted: %d frags\n", dstidx);
1114 rc = -EPROTO;
1115 break;
1116 }
1117
1118 if (tx->tx_nwrq == IBLND_RDMA_FRAGS(conn->ibc_version)) {
1119 CERROR("RDMA too fragmented for %s (%d): %d/%d src %d/%d dst frags\n",
1120 libcfs_nid2str(conn->ibc_peer->ibp_nid),
1121 IBLND_RDMA_FRAGS(conn->ibc_version),
1122 srcidx, srcrd->rd_nfrags,
1123 dstidx, dstrd->rd_nfrags);
1124 rc = -EMSGSIZE;
1125 break;
1126 }
1127
1128 wrknob = min(min(kiblnd_rd_frag_size(srcrd, srcidx),
1129 kiblnd_rd_frag_size(dstrd, dstidx)),
1130 (__u32) resid);
1131
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;
1136
1137 wrq = &tx->tx_wrq[tx->tx_nwrq];
1138
1139 wrq->next = wrq + 1;
1140 wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA);
1141 wrq->sg_list = sge;
1142 wrq->num_sge = 1;
1143 wrq->opcode = IB_WR_RDMA_WRITE;
1144 wrq->send_flags = 0;
1145
1146 wrq->wr.rdma.remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx);
1147 wrq->wr.rdma.rkey = kiblnd_rd_frag_key(dstrd, dstidx);
1148
1149 srcidx = kiblnd_rd_consume_frag(srcrd, srcidx, wrknob);
1150 dstidx = kiblnd_rd_consume_frag(dstrd, dstidx, wrknob);
1151
1152 resid -= wrknob;
1153
1154 tx->tx_nwrq++;
1155 wrq++;
1156 sge++;
1157 }
1158
1159 if (rc < 0) /* no RDMA if completing with failure */
1160 tx->tx_nwrq = 0;
1161
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));
1166
1167 return rc;
1168 }
1169
1170 void
kiblnd_queue_tx_locked(kib_tx_t * tx,kib_conn_t * conn)1171 kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn)
1172 {
1173 struct list_head *q;
1174
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);
1178
1179 tx->tx_queued = 1;
1180 tx->tx_deadline = jiffies + (*kiblnd_tunables.kib_timeout * HZ);
1181
1182 if (tx->tx_conn == NULL) {
1183 kiblnd_conn_addref(conn);
1184 tx->tx_conn = conn;
1185 LASSERT(tx->tx_msg->ibm_type != IBLND_MSG_PUT_DONE);
1186 } else {
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);
1190 }
1191
1192 switch (tx->tx_msg->ibm_type) {
1193 default:
1194 LBUG();
1195
1196 case IBLND_MSG_PUT_REQ:
1197 case IBLND_MSG_GET_REQ:
1198 q = &conn->ibc_tx_queue_rsrvd;
1199 break;
1200
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;
1206 break;
1207
1208 case IBLND_MSG_NOOP:
1209 if (IBLND_OOB_CAPABLE(conn->ibc_version))
1210 q = &conn->ibc_tx_queue_nocred;
1211 else
1212 q = &conn->ibc_tx_noops;
1213 break;
1214
1215 case IBLND_MSG_IMMEDIATE:
1216 q = &conn->ibc_tx_queue;
1217 break;
1218 }
1219
1220 list_add_tail(&tx->tx_list, q);
1221 }
1222
1223 void
kiblnd_queue_tx(kib_tx_t * tx,kib_conn_t * conn)1224 kiblnd_queue_tx(kib_tx_t *tx, kib_conn_t *conn)
1225 {
1226 spin_lock(&conn->ibc_lock);
1227 kiblnd_queue_tx_locked(tx, conn);
1228 spin_unlock(&conn->ibc_lock);
1229
1230 kiblnd_check_sends(conn);
1231 }
1232
kiblnd_resolve_addr(struct rdma_cm_id * cmid,struct sockaddr_in * srcaddr,struct sockaddr_in * dstaddr,int timeout_ms)1233 static int kiblnd_resolve_addr(struct rdma_cm_id *cmid,
1234 struct sockaddr_in *srcaddr,
1235 struct sockaddr_in *dstaddr,
1236 int timeout_ms)
1237 {
1238 unsigned short port;
1239 int rc;
1240
1241 /* allow the port to be reused */
1242 rc = rdma_set_reuseaddr(cmid, 1);
1243 if (rc != 0) {
1244 CERROR("Unable to set reuse on cmid: %d\n", rc);
1245 return rc;
1246 }
1247
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,
1254 timeout_ms);
1255 if (rc == 0) {
1256 CDEBUG(D_NET, "bound to port %hu\n", port);
1257 return 0;
1258 } else if (rc == -EADDRINUSE || rc == -EADDRNOTAVAIL) {
1259 CDEBUG(D_NET, "bind to port %hu failed: %d\n",
1260 port, rc);
1261 } else {
1262 return rc;
1263 }
1264 }
1265
1266 CERROR("Failed to bind to a free privileged port\n");
1267 return rc;
1268 }
1269
1270 static void
kiblnd_connect_peer(kib_peer_t * peer)1271 kiblnd_connect_peer(kib_peer_t *peer)
1272 {
1273 struct rdma_cm_id *cmid;
1274 kib_dev_t *dev;
1275 kib_net_t *net = peer->ibp_ni->ni_data;
1276 struct sockaddr_in srcaddr;
1277 struct sockaddr_in dstaddr;
1278 int rc;
1279
1280 LASSERT(net != NULL);
1281 LASSERT(peer->ibp_connecting > 0);
1282
1283 cmid = kiblnd_rdma_create_id(kiblnd_cm_callback, peer, RDMA_PS_TCP,
1284 IB_QPT_RC);
1285
1286 if (IS_ERR(cmid)) {
1287 CERROR("Can't create CMID for %s: %ld\n",
1288 libcfs_nid2str(peer->ibp_nid), PTR_ERR(cmid));
1289 rc = PTR_ERR(cmid);
1290 goto failed;
1291 }
1292
1293 dev = net->ibn_dev;
1294 memset(&srcaddr, 0, sizeof(srcaddr));
1295 srcaddr.sin_family = AF_INET;
1296 srcaddr.sin_addr.s_addr = htonl(dev->ibd_ifip);
1297
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));
1302
1303 kiblnd_peer_addref(peer); /* cmid's ref */
1304
1305 if (*kiblnd_tunables.kib_use_priv_port) {
1306 rc = kiblnd_resolve_addr(cmid, &srcaddr, &dstaddr,
1307 *kiblnd_tunables.kib_timeout * 1000);
1308 } else {
1309 rc = rdma_resolve_addr(cmid,
1310 (struct sockaddr *)&srcaddr,
1311 (struct sockaddr *)&dstaddr,
1312 *kiblnd_tunables.kib_timeout * 1000);
1313 }
1314 if (rc != 0) {
1315 /* Can't initiate address resolution: */
1316 CERROR("Can't resolve addr for %s: %d\n",
1317 libcfs_nid2str(peer->ibp_nid), rc);
1318 goto failed2;
1319 }
1320
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);
1325
1326 return;
1327
1328 failed2:
1329 kiblnd_peer_decref(peer); /* cmid's ref */
1330 rdma_destroy_id(cmid);
1331 failed:
1332 kiblnd_peer_connect_failed(peer, 1, rc);
1333 }
1334
1335 void
kiblnd_launch_tx(lnet_ni_t * ni,kib_tx_t * tx,lnet_nid_t nid)1336 kiblnd_launch_tx(lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid)
1337 {
1338 kib_peer_t *peer;
1339 kib_peer_t *peer2;
1340 kib_conn_t *conn;
1341 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
1342 unsigned long flags;
1343 int rc;
1344
1345 /* If I get here, I've committed to send, so I complete the tx with
1346 * failure on any problems */
1347
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 */
1350
1351 /* First time, just use a read lock since I expect to find my peer
1352 * connected */
1353 read_lock_irqsave(g_lock, flags);
1354
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... */
1360
1361 read_unlock_irqrestore(g_lock, flags);
1362
1363 if (tx != NULL)
1364 kiblnd_queue_tx(tx, conn);
1365 kiblnd_conn_decref(conn); /* ...to here */
1366 return;
1367 }
1368
1369 read_unlock(g_lock);
1370 /* Re-try with a write lock */
1371 write_lock(g_lock);
1372
1373 peer = kiblnd_find_peer_locked(nid);
1374 if (peer != NULL) {
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);
1379 if (tx != NULL)
1380 list_add_tail(&tx->tx_list,
1381 &peer->ibp_tx_queue);
1382 write_unlock_irqrestore(g_lock, flags);
1383 } else {
1384 conn = kiblnd_get_conn_locked(peer);
1385 kiblnd_conn_addref(conn); /* 1 ref for me... */
1386
1387 write_unlock_irqrestore(g_lock, flags);
1388
1389 if (tx != NULL)
1390 kiblnd_queue_tx(tx, conn);
1391 kiblnd_conn_decref(conn); /* ...to here */
1392 }
1393 return;
1394 }
1395
1396 write_unlock_irqrestore(g_lock, flags);
1397
1398 /* Allocate a peer ready to add to the peer table and retry */
1399 rc = kiblnd_create_peer(ni, &peer, nid);
1400 if (rc != 0) {
1401 CERROR("Can't create peer %s\n", libcfs_nid2str(nid));
1402 if (tx != NULL) {
1403 tx->tx_status = -EHOSTUNREACH;
1404 tx->tx_waiting = 0;
1405 kiblnd_tx_done(ni, tx);
1406 }
1407 return;
1408 }
1409
1410 write_lock_irqsave(g_lock, flags);
1411
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);
1418 if (tx != NULL)
1419 list_add_tail(&tx->tx_list,
1420 &peer2->ibp_tx_queue);
1421 write_unlock_irqrestore(g_lock, flags);
1422 } else {
1423 conn = kiblnd_get_conn_locked(peer2);
1424 kiblnd_conn_addref(conn); /* 1 ref for me... */
1425
1426 write_unlock_irqrestore(g_lock, flags);
1427
1428 if (tx != NULL)
1429 kiblnd_queue_tx(tx, conn);
1430 kiblnd_conn_decref(conn); /* ...to here */
1431 }
1432
1433 kiblnd_peer_decref(peer);
1434 return;
1435 }
1436
1437 /* Brand new peer */
1438 LASSERT(peer->ibp_connecting == 0);
1439 peer->ibp_connecting = 1;
1440
1441 /* always called with a ref on ni, which prevents ni being shutdown */
1442 LASSERT(((kib_net_t *)ni->ni_data)->ibn_shutdown == 0);
1443
1444 if (tx != NULL)
1445 list_add_tail(&tx->tx_list, &peer->ibp_tx_queue);
1446
1447 kiblnd_peer_addref(peer);
1448 list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
1449
1450 write_unlock_irqrestore(g_lock, flags);
1451
1452 kiblnd_connect_peer(peer);
1453 kiblnd_peer_decref(peer);
1454 }
1455
1456 int
kiblnd_send(lnet_ni_t * ni,void * private,lnet_msg_t * lntmsg)1457 kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg)
1458 {
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 kvec *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;
1469 kib_msg_t *ibmsg;
1470 kib_tx_t *tx;
1471 int nob;
1472 int rc;
1473
1474 /* NB 'private' is different depending on what we're sending.... */
1475
1476 CDEBUG(D_NET, "sending %d bytes in %d frags to %s\n",
1477 payload_nob, payload_niov, libcfs_id2str(target));
1478
1479 LASSERT(payload_nob == 0 || payload_niov > 0);
1480 LASSERT(payload_niov <= LNET_MAX_IOV);
1481
1482 /* Thread context */
1483 LASSERT(!in_interrupt());
1484 /* payload is either all vaddrs or all pages */
1485 LASSERT(!(payload_kiov != NULL && payload_iov != NULL));
1486
1487 switch (type) {
1488 default:
1489 LBUG();
1490 return -EIO;
1491
1492 case LNET_MSG_ACK:
1493 LASSERT(payload_nob == 0);
1494 break;
1495
1496 case LNET_MSG_GET:
1497 if (routing || target_is_router)
1498 break; /* send IMMEDIATE */
1499
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 */
1504
1505 tx = kiblnd_get_idle_tx(ni, target.nid);
1506 if (tx == NULL) {
1507 CERROR("Can't allocate txd for GET to %s\n",
1508 libcfs_nid2str(target.nid));
1509 return -ENOMEM;
1510 }
1511
1512 ibmsg = tx->tx_msg;
1513
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);
1520 else
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);
1526 if (rc != 0) {
1527 CERROR("Can't setup GET sink for %s: %d\n",
1528 libcfs_nid2str(target.nid), rc);
1529 kiblnd_tx_done(ni, tx);
1530 return -EIO;
1531 }
1532
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;
1536
1537 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_GET_REQ, nob);
1538
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);
1544 return -EIO;
1545 }
1546
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);
1550 return 0;
1551
1552 case LNET_MSG_REPLY:
1553 case LNET_MSG_PUT:
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 */
1558
1559 tx = kiblnd_get_idle_tx(ni, target.nid);
1560 if (tx == NULL) {
1561 CERROR("Can't allocate %s txd for %s\n",
1562 type == LNET_MSG_PUT ? "PUT" : "REPLY",
1563 libcfs_nid2str(target.nid));
1564 return -ENOMEM;
1565 }
1566
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);
1571 else
1572 rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1573 payload_niov, payload_kiov,
1574 payload_offset, payload_nob);
1575 if (rc != 0) {
1576 CERROR("Can't setup PUT src for %s: %d\n",
1577 libcfs_nid2str(target.nid), rc);
1578 kiblnd_tx_done(ni, tx);
1579 return -EIO;
1580 }
1581
1582 ibmsg = tx->tx_msg;
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));
1586
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);
1590 return 0;
1591 }
1592
1593 /* send IMMEDIATE */
1594
1595 LASSERT(offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob])
1596 <= IBLND_MSG_SIZE);
1597
1598 tx = kiblnd_get_idle_tx(ni, target.nid);
1599 if (tx == NULL) {
1600 CERROR("Can't send %d to %s: tx descs exhausted\n",
1601 type, libcfs_nid2str(target.nid));
1602 return -ENOMEM;
1603 }
1604
1605 ibmsg = tx->tx_msg;
1606 ibmsg->ibm_u.immediate.ibim_hdr = *hdr;
1607
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);
1613 else
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);
1618
1619 nob = offsetof(kib_immediate_msg_t, ibim_payload[payload_nob]);
1620 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_IMMEDIATE, nob);
1621
1622 tx->tx_lntmsg[0] = lntmsg; /* finalise lntmsg on completion */
1623 kiblnd_launch_tx(ni, tx, target.nid);
1624 return 0;
1625 }
1626
1627 static void
kiblnd_reply(lnet_ni_t * ni,kib_rx_t * rx,lnet_msg_t * lntmsg)1628 kiblnd_reply(lnet_ni_t *ni, kib_rx_t *rx, lnet_msg_t *lntmsg)
1629 {
1630 lnet_process_id_t target = lntmsg->msg_target;
1631 unsigned int niov = lntmsg->msg_niov;
1632 struct kvec *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;
1636 kib_tx_t *tx;
1637 int rc;
1638
1639 tx = kiblnd_get_idle_tx(ni, rx->rx_conn->ibc_peer->ibp_nid);
1640 if (tx == NULL) {
1641 CERROR("Can't get tx for REPLY to %s\n",
1642 libcfs_nid2str(target.nid));
1643 goto failed_0;
1644 }
1645
1646 if (nob == 0)
1647 rc = 0;
1648 else if (kiov == NULL)
1649 rc = kiblnd_setup_rd_iov(ni, tx, tx->tx_rd,
1650 niov, iov, offset, nob);
1651 else
1652 rc = kiblnd_setup_rd_kiov(ni, tx, tx->tx_rd,
1653 niov, kiov, offset, nob);
1654
1655 if (rc != 0) {
1656 CERROR("Can't setup GET src for %s: %d\n",
1657 libcfs_nid2str(target.nid), rc);
1658 goto failed_1;
1659 }
1660
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);
1665 if (rc < 0) {
1666 CERROR("Can't setup rdma for GET from %s: %d\n",
1667 libcfs_nid2str(target.nid), rc);
1668 goto failed_1;
1669 }
1670
1671 if (nob == 0) {
1672 /* No RDMA: local completion may happen now! */
1673 lnet_finalize(ni, lntmsg, 0);
1674 } else {
1675 /* RDMA: lnet_finalize(lntmsg) when it
1676 * completes */
1677 tx->tx_lntmsg[0] = lntmsg;
1678 }
1679
1680 kiblnd_queue_tx(tx, rx->rx_conn);
1681 return;
1682
1683 failed_1:
1684 kiblnd_tx_done(ni, tx);
1685 failed_0:
1686 lnet_finalize(ni, lntmsg, -EIO);
1687 }
1688
1689 int
kiblnd_recv(lnet_ni_t * ni,void * private,lnet_msg_t * lntmsg,int delayed,unsigned int niov,struct kvec * iov,lnet_kiov_t * kiov,unsigned int offset,unsigned int mlen,unsigned int rlen)1690 kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed,
1691 unsigned int niov, struct kvec *iov, lnet_kiov_t *kiov,
1692 unsigned int offset, unsigned int mlen, unsigned int rlen)
1693 {
1694 kib_rx_t *rx = private;
1695 kib_msg_t *rxmsg = rx->rx_msg;
1696 kib_conn_t *conn = rx->rx_conn;
1697 kib_tx_t *tx;
1698 kib_msg_t *txmsg;
1699 int nob;
1700 int post_credit = IBLND_POSTRX_PEER_CREDIT;
1701 int rc = 0;
1702
1703 LASSERT(mlen <= rlen);
1704 LASSERT(!in_interrupt());
1705 /* Either all pages or all vaddrs */
1706 LASSERT(!(kiov != NULL && iov != NULL));
1707
1708 switch (rxmsg->ibm_type) {
1709 default:
1710 LBUG();
1711
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),
1717 nob, rx->rx_nob);
1718 rc = -EPROTO;
1719 break;
1720 }
1721
1722 if (kiov != NULL)
1723 lnet_copy_flat2kiov(niov, kiov, offset,
1724 IBLND_MSG_SIZE, rxmsg,
1725 offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1726 mlen);
1727 else
1728 lnet_copy_flat2iov(niov, iov, offset,
1729 IBLND_MSG_SIZE, rxmsg,
1730 offsetof(kib_msg_t, ibm_u.immediate.ibim_payload),
1731 mlen);
1732 lnet_finalize(ni, lntmsg, 0);
1733 break;
1734
1735 case IBLND_MSG_PUT_REQ:
1736 if (mlen == 0) {
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);
1740 break;
1741 }
1742
1743 tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
1744 if (tx == NULL) {
1745 CERROR("Can't allocate tx for %s\n",
1746 libcfs_nid2str(conn->ibc_peer->ibp_nid));
1747 /* Not replying will break the connection */
1748 rc = -ENOMEM;
1749 break;
1750 }
1751
1752 txmsg = tx->tx_msg;
1753 if (kiov == NULL)
1754 rc = kiblnd_setup_rd_iov(ni, tx,
1755 &txmsg->ibm_u.putack.ibpam_rd,
1756 niov, iov, offset, mlen);
1757 else
1758 rc = kiblnd_setup_rd_kiov(ni, tx,
1759 &txmsg->ibm_u.putack.ibpam_rd,
1760 niov, kiov, offset, mlen);
1761 if (rc != 0) {
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);
1768 break;
1769 }
1770
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;
1774
1775 kiblnd_init_tx_msg(ni, tx, IBLND_MSG_PUT_ACK, nob);
1776
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);
1780
1781 /* reposted buffer reserved for PUT_DONE */
1782 post_credit = IBLND_POSTRX_NO_CREDIT;
1783 break;
1784
1785 case IBLND_MSG_GET_REQ:
1786 if (lntmsg != NULL) {
1787 /* Optimized GET; RDMA lntmsg's payload */
1788 kiblnd_reply(ni, rx, lntmsg);
1789 } else {
1790 /* GET didn't match anything */
1791 kiblnd_send_completion(rx->rx_conn, IBLND_MSG_GET_DONE,
1792 -ENODATA,
1793 rxmsg->ibm_u.get.ibgm_cookie);
1794 }
1795 break;
1796 }
1797
1798 kiblnd_post_rx(rx, post_credit);
1799 return rc;
1800 }
1801
1802 int
kiblnd_thread_start(int (* fn)(void * arg),void * arg,char * name)1803 kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name)
1804 {
1805 struct task_struct *task = kthread_run(fn, arg, "%s", name);
1806
1807 if (IS_ERR(task))
1808 return PTR_ERR(task);
1809
1810 atomic_inc(&kiblnd_data.kib_nthreads);
1811 return 0;
1812 }
1813
1814 static void
kiblnd_thread_fini(void)1815 kiblnd_thread_fini(void)
1816 {
1817 atomic_dec(&kiblnd_data.kib_nthreads);
1818 }
1819
1820 void
kiblnd_peer_alive(kib_peer_t * peer)1821 kiblnd_peer_alive(kib_peer_t *peer)
1822 {
1823 /* This is racy, but everyone's only writing cfs_time_current() */
1824 peer->ibp_last_alive = cfs_time_current();
1825 mb();
1826 }
1827
1828 static void
kiblnd_peer_notify(kib_peer_t * peer)1829 kiblnd_peer_notify(kib_peer_t *peer)
1830 {
1831 int error = 0;
1832 unsigned long last_alive = 0;
1833 unsigned long flags;
1834
1835 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1836
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;
1843
1844 last_alive = peer->ibp_last_alive;
1845 }
1846
1847 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1848
1849 if (error != 0)
1850 lnet_notify(peer->ibp_ni,
1851 peer->ibp_nid, 0, last_alive);
1852 }
1853
1854 void
kiblnd_close_conn_locked(kib_conn_t * conn,int error)1855 kiblnd_close_conn_locked(kib_conn_t *conn, int error)
1856 {
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;
1864 kib_dev_t *dev;
1865 unsigned long flags;
1866
1867 LASSERT(error != 0 || conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1868
1869 if (error != 0 && conn->ibc_comms_error == 0)
1870 conn->ibc_comms_error = error;
1871
1872 if (conn->ibc_state != IBLND_CONN_ESTABLISHED)
1873 return; /* already being handled */
1874
1875 if (error == 0 &&
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));
1883 } else {
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)");
1891 }
1892
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 */
1896
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);
1900
1901 /* set/clear error on last conn */
1902 peer->ibp_error = conn->ibc_comms_error;
1903 }
1904
1905 kiblnd_set_conn_state(conn, IBLND_CONN_CLOSING);
1906
1907 if (error != 0 &&
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);
1912 }
1913
1914 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
1915
1916 list_add_tail(&conn->ibc_list, &kiblnd_data.kib_connd_conns);
1917 wake_up(&kiblnd_data.kib_connd_waitq);
1918
1919 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
1920 }
1921
1922 void
kiblnd_close_conn(kib_conn_t * conn,int error)1923 kiblnd_close_conn(kib_conn_t *conn, int error)
1924 {
1925 unsigned long flags;
1926
1927 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1928
1929 kiblnd_close_conn_locked(conn, error);
1930
1931 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1932 }
1933
1934 static void
kiblnd_handle_early_rxs(kib_conn_t * conn)1935 kiblnd_handle_early_rxs(kib_conn_t *conn)
1936 {
1937 unsigned long flags;
1938 kib_rx_t *rx;
1939 kib_rx_t *tmp;
1940
1941 LASSERT(!in_interrupt());
1942 LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
1943
1944 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1945 list_for_each_entry_safe(rx, tmp, &conn->ibc_early_rxs, rx_list) {
1946 list_del(&rx->rx_list);
1947 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1948
1949 kiblnd_handle_rx(rx);
1950
1951 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
1952 }
1953 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
1954 }
1955
1956 static void
kiblnd_abort_txs(kib_conn_t * conn,struct list_head * txs)1957 kiblnd_abort_txs(kib_conn_t *conn, struct list_head *txs)
1958 {
1959 LIST_HEAD(zombies);
1960 struct list_head *tmp;
1961 struct list_head *nxt;
1962 kib_tx_t *tx;
1963
1964 spin_lock(&conn->ibc_lock);
1965
1966 list_for_each_safe(tmp, nxt, txs) {
1967 tx = list_entry(tmp, kib_tx_t, tx_list);
1968
1969 if (txs == &conn->ibc_active_txs) {
1970 LASSERT(!tx->tx_queued);
1971 LASSERT(tx->tx_waiting ||
1972 tx->tx_sending != 0);
1973 } else {
1974 LASSERT(tx->tx_queued);
1975 }
1976
1977 tx->tx_status = -ECONNABORTED;
1978 tx->tx_waiting = 0;
1979
1980 if (tx->tx_sending == 0) {
1981 tx->tx_queued = 0;
1982 list_del(&tx->tx_list);
1983 list_add(&tx->tx_list, &zombies);
1984 }
1985 }
1986
1987 spin_unlock(&conn->ibc_lock);
1988
1989 kiblnd_txlist_done(conn->ibc_peer->ibp_ni, &zombies, -ECONNABORTED);
1990 }
1991
1992 static void
kiblnd_finalise_conn(kib_conn_t * conn)1993 kiblnd_finalise_conn(kib_conn_t *conn)
1994 {
1995 LASSERT(!in_interrupt());
1996 LASSERT(conn->ibc_state > IBLND_CONN_INIT);
1997
1998 kiblnd_set_conn_state(conn, IBLND_CONN_DISCONNECTED);
1999
2000 /* abort_receives moves QP state to IB_QPS_ERR. This is only required
2001 * for connections that didn't get as far as being connected, because
2002 * rdma_disconnect() does this for free. */
2003 kiblnd_abort_receives(conn);
2004
2005 /* Complete all tx descs not waiting for sends to complete.
2006 * NB we should be safe from RDMA now that the QP has changed state */
2007
2008 kiblnd_abort_txs(conn, &conn->ibc_tx_noops);
2009 kiblnd_abort_txs(conn, &conn->ibc_tx_queue);
2010 kiblnd_abort_txs(conn, &conn->ibc_tx_queue_rsrvd);
2011 kiblnd_abort_txs(conn, &conn->ibc_tx_queue_nocred);
2012 kiblnd_abort_txs(conn, &conn->ibc_active_txs);
2013
2014 kiblnd_handle_early_rxs(conn);
2015 }
2016
2017 void
kiblnd_peer_connect_failed(kib_peer_t * peer,int active,int error)2018 kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error)
2019 {
2020 LIST_HEAD(zombies);
2021 unsigned long flags;
2022
2023 LASSERT(error != 0);
2024 LASSERT(!in_interrupt());
2025
2026 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2027
2028 if (active) {
2029 LASSERT(peer->ibp_connecting > 0);
2030 peer->ibp_connecting--;
2031 } else {
2032 LASSERT(peer->ibp_accepting > 0);
2033 peer->ibp_accepting--;
2034 }
2035
2036 if (peer->ibp_connecting != 0 ||
2037 peer->ibp_accepting != 0) {
2038 /* another connection attempt under way... */
2039 write_unlock_irqrestore(&kiblnd_data.kib_global_lock,
2040 flags);
2041 return;
2042 }
2043
2044 if (list_empty(&peer->ibp_conns)) {
2045 /* Take peer's blocked transmits to complete with error */
2046 list_add(&zombies, &peer->ibp_tx_queue);
2047 list_del_init(&peer->ibp_tx_queue);
2048
2049 if (kiblnd_peer_active(peer))
2050 kiblnd_unlink_peer_locked(peer);
2051
2052 peer->ibp_error = error;
2053 } else {
2054 /* Can't have blocked transmits if there are connections */
2055 LASSERT(list_empty(&peer->ibp_tx_queue));
2056 }
2057
2058 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2059
2060 kiblnd_peer_notify(peer);
2061
2062 if (list_empty(&zombies))
2063 return;
2064
2065 CNETERR("Deleting messages for %s: connection failed\n",
2066 libcfs_nid2str(peer->ibp_nid));
2067
2068 kiblnd_txlist_done(peer->ibp_ni, &zombies, -EHOSTUNREACH);
2069 }
2070
2071 void
kiblnd_connreq_done(kib_conn_t * conn,int status)2072 kiblnd_connreq_done(kib_conn_t *conn, int status)
2073 {
2074 kib_peer_t *peer = conn->ibc_peer;
2075 kib_tx_t *tx;
2076 kib_tx_t *tmp;
2077 struct list_head txs;
2078 unsigned long flags;
2079 int active;
2080
2081 active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2082
2083 CDEBUG(D_NET, "%s: active(%d), version(%x), status(%d)\n",
2084 libcfs_nid2str(peer->ibp_nid), active,
2085 conn->ibc_version, status);
2086
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));
2092
2093 LIBCFS_FREE(conn->ibc_connvars, sizeof(*conn->ibc_connvars));
2094 conn->ibc_connvars = NULL;
2095
2096 if (status != 0) {
2097 /* failed to establish connection */
2098 kiblnd_peer_connect_failed(peer, active, status);
2099 kiblnd_finalise_conn(conn);
2100 return;
2101 }
2102
2103 /* connection established */
2104 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2105
2106 conn->ibc_last_send = jiffies;
2107 kiblnd_set_conn_state(conn, IBLND_CONN_ESTABLISHED);
2108 kiblnd_peer_alive(peer);
2109
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);
2114 if (active)
2115 peer->ibp_connecting--;
2116 else
2117 peer->ibp_accepting--;
2118
2119 if (peer->ibp_version == 0) {
2120 peer->ibp_version = conn->ibc_version;
2121 peer->ibp_incarnation = conn->ibc_incarnation;
2122 }
2123
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;
2130 }
2131
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);
2135
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;
2139
2140 /* start to shut down connection */
2141 kiblnd_close_conn_locked(conn, -ECONNABORTED);
2142 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2143
2144 kiblnd_txlist_done(ni, &txs, -ECONNABORTED);
2145
2146 return;
2147 }
2148
2149 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2150
2151 /* Schedule blocked txs */
2152 spin_lock(&conn->ibc_lock);
2153 list_for_each_entry_safe(tx, tmp, &txs, tx_list) {
2154 list_del(&tx->tx_list);
2155
2156 kiblnd_queue_tx_locked(tx, conn);
2157 }
2158 spin_unlock(&conn->ibc_lock);
2159
2160 kiblnd_check_sends(conn);
2161
2162 /* schedule blocked rxs */
2163 kiblnd_handle_early_rxs(conn);
2164 }
2165
2166 static void
kiblnd_reject(struct rdma_cm_id * cmid,kib_rej_t * rej)2167 kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej)
2168 {
2169 int rc;
2170
2171 rc = rdma_reject(cmid, rej, sizeof(*rej));
2172
2173 if (rc != 0)
2174 CWARN("Error %d sending reject\n", rc);
2175 }
2176
2177 static int
kiblnd_passive_connect(struct rdma_cm_id * cmid,void * priv,int priv_nob)2178 kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob)
2179 {
2180 rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
2181 kib_msg_t *reqmsg = priv;
2182 kib_msg_t *ackmsg;
2183 kib_dev_t *ibdev;
2184 kib_peer_t *peer;
2185 kib_peer_t *peer2;
2186 kib_conn_t *conn;
2187 lnet_ni_t *ni = NULL;
2188 kib_net_t *net = NULL;
2189 lnet_nid_t nid;
2190 struct rdma_conn_param cp;
2191 kib_rej_t rej;
2192 int version = IBLND_MSG_VERSION;
2193 unsigned long flags;
2194 int rc;
2195 struct sockaddr_in *peer_addr;
2196 LASSERT(!in_interrupt());
2197
2198 /* cmid inherits 'context' from the corresponding listener id */
2199 ibdev = (kib_dev_t *)cmid->context;
2200 LASSERT(ibdev != NULL);
2201
2202 memset(&rej, 0, sizeof(rej));
2203 rej.ibr_magic = IBLND_MSG_MAGIC;
2204 rej.ibr_why = IBLND_REJECT_FATAL;
2205 rej.ibr_cp.ibcp_max_msg_size = IBLND_MSG_SIZE;
2206
2207 peer_addr = (struct sockaddr_in *)&(cmid->route.addr.dst_addr);
2208 if (*kiblnd_tunables.kib_require_priv_port &&
2209 ntohs(peer_addr->sin_port) >= PROT_SOCK) {
2210 __u32 ip = ntohl(peer_addr->sin_addr.s_addr);
2211 CERROR("Peer's port (%pI4h:%hu) is not privileged\n",
2212 &ip, ntohs(peer_addr->sin_port));
2213 goto failed;
2214 }
2215
2216 if (priv_nob < offsetof(kib_msg_t, ibm_type)) {
2217 CERROR("Short connection request\n");
2218 goto failed;
2219 }
2220
2221 /* Future protocol version compatibility support! If the
2222 * o2iblnd-specific protocol changes, or when LNET unifies
2223 * protocols over all LNDs, the initial connection will
2224 * negotiate a protocol version. I trap this here to avoid
2225 * console errors; the reject tells the peer which protocol I
2226 * speak. */
2227 if (reqmsg->ibm_magic == LNET_PROTO_MAGIC ||
2228 reqmsg->ibm_magic == __swab32(LNET_PROTO_MAGIC))
2229 goto failed;
2230 if (reqmsg->ibm_magic == IBLND_MSG_MAGIC &&
2231 reqmsg->ibm_version != IBLND_MSG_VERSION &&
2232 reqmsg->ibm_version != IBLND_MSG_VERSION_1)
2233 goto failed;
2234 if (reqmsg->ibm_magic == __swab32(IBLND_MSG_MAGIC) &&
2235 reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION) &&
2236 reqmsg->ibm_version != __swab16(IBLND_MSG_VERSION_1))
2237 goto failed;
2238
2239 rc = kiblnd_unpack_msg(reqmsg, priv_nob);
2240 if (rc != 0) {
2241 CERROR("Can't parse connection request: %d\n", rc);
2242 goto failed;
2243 }
2244
2245 nid = reqmsg->ibm_srcnid;
2246 ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid));
2247
2248 if (ni != NULL) {
2249 net = (kib_net_t *)ni->ni_data;
2250 rej.ibr_incarnation = net->ibn_incarnation;
2251 }
2252
2253 if (ni == NULL || /* no matching net */
2254 ni->ni_nid != reqmsg->ibm_dstnid || /* right NET, wrong NID! */
2255 net->ibn_dev != ibdev) { /* wrong device */
2256 CERROR("Can't accept %s on %s (%s:%d:%pI4h): bad dst nid %s\n",
2257 libcfs_nid2str(nid),
2258 ni == NULL ? "NA" : libcfs_nid2str(ni->ni_nid),
2259 ibdev->ibd_ifname, ibdev->ibd_nnets,
2260 &ibdev->ibd_ifip,
2261 libcfs_nid2str(reqmsg->ibm_dstnid));
2262
2263 goto failed;
2264 }
2265
2266 /* check time stamp as soon as possible */
2267 if (reqmsg->ibm_dststamp != 0 &&
2268 reqmsg->ibm_dststamp != net->ibn_incarnation) {
2269 CWARN("Stale connection request\n");
2270 rej.ibr_why = IBLND_REJECT_CONN_STALE;
2271 goto failed;
2272 }
2273
2274 /* I can accept peer's version */
2275 version = reqmsg->ibm_version;
2276
2277 if (reqmsg->ibm_type != IBLND_MSG_CONNREQ) {
2278 CERROR("Unexpected connreq msg type: %x from %s\n",
2279 reqmsg->ibm_type, libcfs_nid2str(nid));
2280 goto failed;
2281 }
2282
2283 if (reqmsg->ibm_u.connparams.ibcp_queue_depth !=
2284 IBLND_MSG_QUEUE_SIZE(version)) {
2285 CERROR("Can't accept %s: incompatible queue depth %d (%d wanted)\n",
2286 libcfs_nid2str(nid), reqmsg->ibm_u.connparams.ibcp_queue_depth,
2287 IBLND_MSG_QUEUE_SIZE(version));
2288
2289 if (version == IBLND_MSG_VERSION)
2290 rej.ibr_why = IBLND_REJECT_MSG_QUEUE_SIZE;
2291
2292 goto failed;
2293 }
2294
2295 if (reqmsg->ibm_u.connparams.ibcp_max_frags !=
2296 IBLND_RDMA_FRAGS(version)) {
2297 CERROR("Can't accept %s(version %x): incompatible max_frags %d (%d wanted)\n",
2298 libcfs_nid2str(nid), version,
2299 reqmsg->ibm_u.connparams.ibcp_max_frags,
2300 IBLND_RDMA_FRAGS(version));
2301
2302 if (version == IBLND_MSG_VERSION)
2303 rej.ibr_why = IBLND_REJECT_RDMA_FRAGS;
2304
2305 goto failed;
2306
2307 }
2308
2309 if (reqmsg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2310 CERROR("Can't accept %s: message size %d too big (%d max)\n",
2311 libcfs_nid2str(nid),
2312 reqmsg->ibm_u.connparams.ibcp_max_msg_size,
2313 IBLND_MSG_SIZE);
2314 goto failed;
2315 }
2316
2317 /* assume 'nid' is a new peer; create */
2318 rc = kiblnd_create_peer(ni, &peer, nid);
2319 if (rc != 0) {
2320 CERROR("Can't create peer for %s\n", libcfs_nid2str(nid));
2321 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2322 goto failed;
2323 }
2324
2325 write_lock_irqsave(g_lock, flags);
2326
2327 peer2 = kiblnd_find_peer_locked(nid);
2328 if (peer2 != NULL) {
2329 if (peer2->ibp_version == 0) {
2330 peer2->ibp_version = version;
2331 peer2->ibp_incarnation = reqmsg->ibm_srcstamp;
2332 }
2333
2334 /* not the guy I've talked with */
2335 if (peer2->ibp_incarnation != reqmsg->ibm_srcstamp ||
2336 peer2->ibp_version != version) {
2337 kiblnd_close_peer_conns_locked(peer2, -ESTALE);
2338 write_unlock_irqrestore(g_lock, flags);
2339
2340 CWARN("Conn stale %s [old ver: %x, new ver: %x]\n",
2341 libcfs_nid2str(nid), peer2->ibp_version, version);
2342
2343 kiblnd_peer_decref(peer);
2344 rej.ibr_why = IBLND_REJECT_CONN_STALE;
2345 goto failed;
2346 }
2347
2348 /* tie-break connection race in favour of the higher NID */
2349 if (peer2->ibp_connecting != 0 &&
2350 nid < ni->ni_nid) {
2351 write_unlock_irqrestore(g_lock, flags);
2352
2353 CWARN("Conn race %s\n", libcfs_nid2str(peer2->ibp_nid));
2354
2355 kiblnd_peer_decref(peer);
2356 rej.ibr_why = IBLND_REJECT_CONN_RACE;
2357 goto failed;
2358 }
2359
2360 peer2->ibp_accepting++;
2361 kiblnd_peer_addref(peer2);
2362
2363 write_unlock_irqrestore(g_lock, flags);
2364 kiblnd_peer_decref(peer);
2365 peer = peer2;
2366 } else {
2367 /* Brand new peer */
2368 LASSERT(peer->ibp_accepting == 0);
2369 LASSERT(peer->ibp_version == 0 &&
2370 peer->ibp_incarnation == 0);
2371
2372 peer->ibp_accepting = 1;
2373 peer->ibp_version = version;
2374 peer->ibp_incarnation = reqmsg->ibm_srcstamp;
2375
2376 /* I have a ref on ni that prevents it being shutdown */
2377 LASSERT(net->ibn_shutdown == 0);
2378
2379 kiblnd_peer_addref(peer);
2380 list_add_tail(&peer->ibp_list, kiblnd_nid2peerlist(nid));
2381
2382 write_unlock_irqrestore(g_lock, flags);
2383 }
2384
2385 conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_PASSIVE_WAIT, version);
2386 if (conn == NULL) {
2387 kiblnd_peer_connect_failed(peer, 0, -ENOMEM);
2388 kiblnd_peer_decref(peer);
2389 rej.ibr_why = IBLND_REJECT_NO_RESOURCES;
2390 goto failed;
2391 }
2392
2393 /* conn now "owns" cmid, so I return success from here on to ensure the
2394 * CM callback doesn't destroy cmid. */
2395
2396 conn->ibc_incarnation = reqmsg->ibm_srcstamp;
2397 conn->ibc_credits = IBLND_MSG_QUEUE_SIZE(version);
2398 conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(version);
2399 LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(version)
2400 <= IBLND_RX_MSGS(version));
2401
2402 ackmsg = &conn->ibc_connvars->cv_msg;
2403 memset(ackmsg, 0, sizeof(*ackmsg));
2404
2405 kiblnd_init_msg(ackmsg, IBLND_MSG_CONNACK,
2406 sizeof(ackmsg->ibm_u.connparams));
2407 ackmsg->ibm_u.connparams.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2408 ackmsg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2409 ackmsg->ibm_u.connparams.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
2410
2411 kiblnd_pack_msg(ni, ackmsg, version, 0, nid, reqmsg->ibm_srcstamp);
2412
2413 memset(&cp, 0, sizeof(cp));
2414 cp.private_data = ackmsg;
2415 cp.private_data_len = ackmsg->ibm_nob;
2416 cp.responder_resources = 0; /* No atomic ops or RDMA reads */
2417 cp.initiator_depth = 0;
2418 cp.flow_control = 1;
2419 cp.retry_count = *kiblnd_tunables.kib_retry_count;
2420 cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
2421
2422 CDEBUG(D_NET, "Accept %s\n", libcfs_nid2str(nid));
2423
2424 rc = rdma_accept(cmid, &cp);
2425 if (rc != 0) {
2426 CERROR("Can't accept %s: %d\n", libcfs_nid2str(nid), rc);
2427 rej.ibr_version = version;
2428 rej.ibr_why = IBLND_REJECT_FATAL;
2429
2430 kiblnd_reject(cmid, &rej);
2431 kiblnd_connreq_done(conn, rc);
2432 kiblnd_conn_decref(conn);
2433 }
2434
2435 lnet_ni_decref(ni);
2436 return 0;
2437
2438 failed:
2439 if (ni != NULL)
2440 lnet_ni_decref(ni);
2441
2442 rej.ibr_version = version;
2443 rej.ibr_cp.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2444 rej.ibr_cp.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
2445 kiblnd_reject(cmid, &rej);
2446
2447 return -ECONNREFUSED;
2448 }
2449
2450 static void
kiblnd_reconnect(kib_conn_t * conn,int version,__u64 incarnation,int why,kib_connparams_t * cp)2451 kiblnd_reconnect(kib_conn_t *conn, int version,
2452 __u64 incarnation, int why, kib_connparams_t *cp)
2453 {
2454 kib_peer_t *peer = conn->ibc_peer;
2455 char *reason;
2456 int retry = 0;
2457 unsigned long flags;
2458
2459 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2460 LASSERT(peer->ibp_connecting > 0); /* 'conn' at least */
2461
2462 write_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2463
2464 /* retry connection if it's still needed and no other connection
2465 * attempts (active or passive) are in progress
2466 * NB: reconnect is still needed even when ibp_tx_queue is
2467 * empty if ibp_version != version because reconnect may be
2468 * initiated by kiblnd_query() */
2469 if ((!list_empty(&peer->ibp_tx_queue) ||
2470 peer->ibp_version != version) &&
2471 peer->ibp_connecting == 1 &&
2472 peer->ibp_accepting == 0) {
2473 retry = 1;
2474 peer->ibp_connecting++;
2475
2476 peer->ibp_version = version;
2477 peer->ibp_incarnation = incarnation;
2478 }
2479
2480 write_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2481
2482 if (!retry)
2483 return;
2484
2485 switch (why) {
2486 default:
2487 reason = "Unknown";
2488 break;
2489
2490 case IBLND_REJECT_CONN_STALE:
2491 reason = "stale";
2492 break;
2493
2494 case IBLND_REJECT_CONN_RACE:
2495 reason = "conn race";
2496 break;
2497
2498 case IBLND_REJECT_CONN_UNCOMPAT:
2499 reason = "version negotiation";
2500 break;
2501 }
2502
2503 CNETERR("%s: retrying (%s), %x, %x, queue_dep: %d, max_frag: %d, msg_size: %d\n",
2504 libcfs_nid2str(peer->ibp_nid),
2505 reason, IBLND_MSG_VERSION, version,
2506 cp != NULL ? cp->ibcp_queue_depth : IBLND_MSG_QUEUE_SIZE(version),
2507 cp != NULL ? cp->ibcp_max_frags : IBLND_RDMA_FRAGS(version),
2508 cp != NULL ? cp->ibcp_max_msg_size : IBLND_MSG_SIZE);
2509
2510 kiblnd_connect_peer(peer);
2511 }
2512
2513 static void
kiblnd_rejected(kib_conn_t * conn,int reason,void * priv,int priv_nob)2514 kiblnd_rejected(kib_conn_t *conn, int reason, void *priv, int priv_nob)
2515 {
2516 kib_peer_t *peer = conn->ibc_peer;
2517
2518 LASSERT(!in_interrupt());
2519 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
2520
2521 switch (reason) {
2522 case IB_CM_REJ_STALE_CONN:
2523 kiblnd_reconnect(conn, IBLND_MSG_VERSION, 0,
2524 IBLND_REJECT_CONN_STALE, NULL);
2525 break;
2526
2527 case IB_CM_REJ_INVALID_SERVICE_ID:
2528 CNETERR("%s rejected: no listener at %d\n",
2529 libcfs_nid2str(peer->ibp_nid),
2530 *kiblnd_tunables.kib_service);
2531 break;
2532
2533 case IB_CM_REJ_CONSUMER_DEFINED:
2534 if (priv_nob >= offsetof(kib_rej_t, ibr_padding)) {
2535 kib_rej_t *rej = priv;
2536 kib_connparams_t *cp = NULL;
2537 int flip = 0;
2538 __u64 incarnation = -1;
2539
2540 /* NB. default incarnation is -1 because:
2541 * a) V1 will ignore dst incarnation in connreq.
2542 * b) V2 will provide incarnation while rejecting me,
2543 * -1 will be overwrote.
2544 *
2545 * if I try to connect to a V1 peer with V2 protocol,
2546 * it rejected me then upgrade to V2, I have no idea
2547 * about the upgrading and try to reconnect with V1,
2548 * in this case upgraded V2 can find out I'm trying to
2549 * talk to the old guy and reject me(incarnation is -1).
2550 */
2551
2552 if (rej->ibr_magic == __swab32(IBLND_MSG_MAGIC) ||
2553 rej->ibr_magic == __swab32(LNET_PROTO_MAGIC)) {
2554 __swab32s(&rej->ibr_magic);
2555 __swab16s(&rej->ibr_version);
2556 flip = 1;
2557 }
2558
2559 if (priv_nob >= sizeof(kib_rej_t) &&
2560 rej->ibr_version > IBLND_MSG_VERSION_1) {
2561 /* priv_nob is always 148 in current version
2562 * of OFED, so we still need to check version.
2563 * (define of IB_CM_REJ_PRIVATE_DATA_SIZE) */
2564 cp = &rej->ibr_cp;
2565
2566 if (flip) {
2567 __swab64s(&rej->ibr_incarnation);
2568 __swab16s(&cp->ibcp_queue_depth);
2569 __swab16s(&cp->ibcp_max_frags);
2570 __swab32s(&cp->ibcp_max_msg_size);
2571 }
2572
2573 incarnation = rej->ibr_incarnation;
2574 }
2575
2576 if (rej->ibr_magic != IBLND_MSG_MAGIC &&
2577 rej->ibr_magic != LNET_PROTO_MAGIC) {
2578 CERROR("%s rejected: consumer defined fatal error\n",
2579 libcfs_nid2str(peer->ibp_nid));
2580 break;
2581 }
2582
2583 if (rej->ibr_version != IBLND_MSG_VERSION &&
2584 rej->ibr_version != IBLND_MSG_VERSION_1) {
2585 CERROR("%s rejected: o2iblnd version %x error\n",
2586 libcfs_nid2str(peer->ibp_nid),
2587 rej->ibr_version);
2588 break;
2589 }
2590
2591 if (rej->ibr_why == IBLND_REJECT_FATAL &&
2592 rej->ibr_version == IBLND_MSG_VERSION_1) {
2593 CDEBUG(D_NET, "rejected by old version peer %s: %x\n",
2594 libcfs_nid2str(peer->ibp_nid), rej->ibr_version);
2595
2596 if (conn->ibc_version != IBLND_MSG_VERSION_1)
2597 rej->ibr_why = IBLND_REJECT_CONN_UNCOMPAT;
2598 }
2599
2600 switch (rej->ibr_why) {
2601 case IBLND_REJECT_CONN_RACE:
2602 case IBLND_REJECT_CONN_STALE:
2603 case IBLND_REJECT_CONN_UNCOMPAT:
2604 kiblnd_reconnect(conn, rej->ibr_version,
2605 incarnation, rej->ibr_why, cp);
2606 break;
2607
2608 case IBLND_REJECT_MSG_QUEUE_SIZE:
2609 CERROR("%s rejected: incompatible message queue depth %d, %d\n",
2610 libcfs_nid2str(peer->ibp_nid),
2611 cp != NULL ? cp->ibcp_queue_depth :
2612 IBLND_MSG_QUEUE_SIZE(rej->ibr_version),
2613 IBLND_MSG_QUEUE_SIZE(conn->ibc_version));
2614 break;
2615
2616 case IBLND_REJECT_RDMA_FRAGS:
2617 CERROR("%s rejected: incompatible # of RDMA fragments %d, %d\n",
2618 libcfs_nid2str(peer->ibp_nid),
2619 cp != NULL ? cp->ibcp_max_frags :
2620 IBLND_RDMA_FRAGS(rej->ibr_version),
2621 IBLND_RDMA_FRAGS(conn->ibc_version));
2622 break;
2623
2624 case IBLND_REJECT_NO_RESOURCES:
2625 CERROR("%s rejected: o2iblnd no resources\n",
2626 libcfs_nid2str(peer->ibp_nid));
2627 break;
2628
2629 case IBLND_REJECT_FATAL:
2630 CERROR("%s rejected: o2iblnd fatal error\n",
2631 libcfs_nid2str(peer->ibp_nid));
2632 break;
2633
2634 default:
2635 CERROR("%s rejected: o2iblnd reason %d\n",
2636 libcfs_nid2str(peer->ibp_nid),
2637 rej->ibr_why);
2638 break;
2639 }
2640 break;
2641 }
2642 /* fall through */
2643 default:
2644 CNETERR("%s rejected: reason %d, size %d\n",
2645 libcfs_nid2str(peer->ibp_nid), reason, priv_nob);
2646 break;
2647 }
2648
2649 kiblnd_connreq_done(conn, -ECONNREFUSED);
2650 }
2651
2652 static void
kiblnd_check_connreply(kib_conn_t * conn,void * priv,int priv_nob)2653 kiblnd_check_connreply(kib_conn_t *conn, void *priv, int priv_nob)
2654 {
2655 kib_peer_t *peer = conn->ibc_peer;
2656 lnet_ni_t *ni = peer->ibp_ni;
2657 kib_net_t *net = ni->ni_data;
2658 kib_msg_t *msg = priv;
2659 int ver = conn->ibc_version;
2660 int rc = kiblnd_unpack_msg(msg, priv_nob);
2661 unsigned long flags;
2662
2663 LASSERT(net != NULL);
2664
2665 if (rc != 0) {
2666 CERROR("Can't unpack connack from %s: %d\n",
2667 libcfs_nid2str(peer->ibp_nid), rc);
2668 goto failed;
2669 }
2670
2671 if (msg->ibm_type != IBLND_MSG_CONNACK) {
2672 CERROR("Unexpected message %d from %s\n",
2673 msg->ibm_type, libcfs_nid2str(peer->ibp_nid));
2674 rc = -EPROTO;
2675 goto failed;
2676 }
2677
2678 if (ver != msg->ibm_version) {
2679 CERROR("%s replied version %x is different with requested version %x\n",
2680 libcfs_nid2str(peer->ibp_nid), msg->ibm_version, ver);
2681 rc = -EPROTO;
2682 goto failed;
2683 }
2684
2685 if (msg->ibm_u.connparams.ibcp_queue_depth !=
2686 IBLND_MSG_QUEUE_SIZE(ver)) {
2687 CERROR("%s has incompatible queue depth %d(%d wanted)\n",
2688 libcfs_nid2str(peer->ibp_nid),
2689 msg->ibm_u.connparams.ibcp_queue_depth,
2690 IBLND_MSG_QUEUE_SIZE(ver));
2691 rc = -EPROTO;
2692 goto failed;
2693 }
2694
2695 if (msg->ibm_u.connparams.ibcp_max_frags !=
2696 IBLND_RDMA_FRAGS(ver)) {
2697 CERROR("%s has incompatible max_frags %d (%d wanted)\n",
2698 libcfs_nid2str(peer->ibp_nid),
2699 msg->ibm_u.connparams.ibcp_max_frags,
2700 IBLND_RDMA_FRAGS(ver));
2701 rc = -EPROTO;
2702 goto failed;
2703 }
2704
2705 if (msg->ibm_u.connparams.ibcp_max_msg_size > IBLND_MSG_SIZE) {
2706 CERROR("%s max message size %d too big (%d max)\n",
2707 libcfs_nid2str(peer->ibp_nid),
2708 msg->ibm_u.connparams.ibcp_max_msg_size,
2709 IBLND_MSG_SIZE);
2710 rc = -EPROTO;
2711 goto failed;
2712 }
2713
2714 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2715 if (msg->ibm_dstnid == ni->ni_nid &&
2716 msg->ibm_dststamp == net->ibn_incarnation)
2717 rc = 0;
2718 else
2719 rc = -ESTALE;
2720 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2721
2722 if (rc != 0) {
2723 CERROR("Bad connection reply from %s, rc = %d, version: %x max_frags: %d\n",
2724 libcfs_nid2str(peer->ibp_nid), rc,
2725 msg->ibm_version, msg->ibm_u.connparams.ibcp_max_frags);
2726 goto failed;
2727 }
2728
2729 conn->ibc_incarnation = msg->ibm_srcstamp;
2730 conn->ibc_credits =
2731 conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(ver);
2732 LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(ver)
2733 <= IBLND_RX_MSGS(ver));
2734
2735 kiblnd_connreq_done(conn, 0);
2736 return;
2737
2738 failed:
2739 /* NB My QP has already established itself, so I handle anything going
2740 * wrong here by setting ibc_comms_error.
2741 * kiblnd_connreq_done(0) moves the conn state to ESTABLISHED, but then
2742 * immediately tears it down. */
2743
2744 LASSERT(rc != 0);
2745 conn->ibc_comms_error = rc;
2746 kiblnd_connreq_done(conn, 0);
2747 }
2748
2749 static int
kiblnd_active_connect(struct rdma_cm_id * cmid)2750 kiblnd_active_connect(struct rdma_cm_id *cmid)
2751 {
2752 kib_peer_t *peer = (kib_peer_t *)cmid->context;
2753 kib_conn_t *conn;
2754 kib_msg_t *msg;
2755 struct rdma_conn_param cp;
2756 int version;
2757 __u64 incarnation;
2758 unsigned long flags;
2759 int rc;
2760
2761 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
2762
2763 incarnation = peer->ibp_incarnation;
2764 version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION :
2765 peer->ibp_version;
2766
2767 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
2768
2769 conn = kiblnd_create_conn(peer, cmid, IBLND_CONN_ACTIVE_CONNECT, version);
2770 if (conn == NULL) {
2771 kiblnd_peer_connect_failed(peer, 1, -ENOMEM);
2772 kiblnd_peer_decref(peer); /* lose cmid's ref */
2773 return -ENOMEM;
2774 }
2775
2776 /* conn "owns" cmid now, so I return success from here on to ensure the
2777 * CM callback doesn't destroy cmid. conn also takes over cmid's ref
2778 * on peer */
2779
2780 msg = &conn->ibc_connvars->cv_msg;
2781
2782 memset(msg, 0, sizeof(*msg));
2783 kiblnd_init_msg(msg, IBLND_MSG_CONNREQ, sizeof(msg->ibm_u.connparams));
2784 msg->ibm_u.connparams.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
2785 msg->ibm_u.connparams.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
2786 msg->ibm_u.connparams.ibcp_max_msg_size = IBLND_MSG_SIZE;
2787
2788 kiblnd_pack_msg(peer->ibp_ni, msg, version,
2789 0, peer->ibp_nid, incarnation);
2790
2791 memset(&cp, 0, sizeof(cp));
2792 cp.private_data = msg;
2793 cp.private_data_len = msg->ibm_nob;
2794 cp.responder_resources = 0; /* No atomic ops or RDMA reads */
2795 cp.initiator_depth = 0;
2796 cp.flow_control = 1;
2797 cp.retry_count = *kiblnd_tunables.kib_retry_count;
2798 cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
2799
2800 LASSERT(cmid->context == (void *)conn);
2801 LASSERT(conn->ibc_cmid == cmid);
2802
2803 rc = rdma_connect(cmid, &cp);
2804 if (rc != 0) {
2805 CERROR("Can't connect to %s: %d\n",
2806 libcfs_nid2str(peer->ibp_nid), rc);
2807 kiblnd_connreq_done(conn, rc);
2808 kiblnd_conn_decref(conn);
2809 }
2810
2811 return 0;
2812 }
2813
2814 int
kiblnd_cm_callback(struct rdma_cm_id * cmid,struct rdma_cm_event * event)2815 kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event)
2816 {
2817 kib_peer_t *peer;
2818 kib_conn_t *conn;
2819 int rc;
2820
2821 switch (event->event) {
2822 default:
2823 CERROR("Unexpected event: %d, status: %d\n",
2824 event->event, event->status);
2825 LBUG();
2826
2827 case RDMA_CM_EVENT_CONNECT_REQUEST:
2828 /* destroy cmid on failure */
2829 rc = kiblnd_passive_connect(cmid,
2830 (void *)KIBLND_CONN_PARAM(event),
2831 KIBLND_CONN_PARAM_LEN(event));
2832 CDEBUG(D_NET, "connreq: %d\n", rc);
2833 return rc;
2834
2835 case RDMA_CM_EVENT_ADDR_ERROR:
2836 peer = (kib_peer_t *)cmid->context;
2837 CNETERR("%s: ADDR ERROR %d\n",
2838 libcfs_nid2str(peer->ibp_nid), event->status);
2839 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2840 kiblnd_peer_decref(peer);
2841 return -EHOSTUNREACH; /* rc != 0 destroys cmid */
2842
2843 case RDMA_CM_EVENT_ADDR_RESOLVED:
2844 peer = (kib_peer_t *)cmid->context;
2845
2846 CDEBUG(D_NET, "%s Addr resolved: %d\n",
2847 libcfs_nid2str(peer->ibp_nid), event->status);
2848
2849 if (event->status != 0) {
2850 CNETERR("Can't resolve address for %s: %d\n",
2851 libcfs_nid2str(peer->ibp_nid), event->status);
2852 rc = event->status;
2853 } else {
2854 rc = rdma_resolve_route(
2855 cmid, *kiblnd_tunables.kib_timeout * 1000);
2856 if (rc == 0)
2857 return 0;
2858 /* Can't initiate route resolution */
2859 CERROR("Can't resolve route for %s: %d\n",
2860 libcfs_nid2str(peer->ibp_nid), rc);
2861 }
2862 kiblnd_peer_connect_failed(peer, 1, rc);
2863 kiblnd_peer_decref(peer);
2864 return rc; /* rc != 0 destroys cmid */
2865
2866 case RDMA_CM_EVENT_ROUTE_ERROR:
2867 peer = (kib_peer_t *)cmid->context;
2868 CNETERR("%s: ROUTE ERROR %d\n",
2869 libcfs_nid2str(peer->ibp_nid), event->status);
2870 kiblnd_peer_connect_failed(peer, 1, -EHOSTUNREACH);
2871 kiblnd_peer_decref(peer);
2872 return -EHOSTUNREACH; /* rc != 0 destroys cmid */
2873
2874 case RDMA_CM_EVENT_ROUTE_RESOLVED:
2875 peer = (kib_peer_t *)cmid->context;
2876 CDEBUG(D_NET, "%s Route resolved: %d\n",
2877 libcfs_nid2str(peer->ibp_nid), event->status);
2878
2879 if (event->status == 0)
2880 return kiblnd_active_connect(cmid);
2881
2882 CNETERR("Can't resolve route for %s: %d\n",
2883 libcfs_nid2str(peer->ibp_nid), event->status);
2884 kiblnd_peer_connect_failed(peer, 1, event->status);
2885 kiblnd_peer_decref(peer);
2886 return event->status; /* rc != 0 destroys cmid */
2887
2888 case RDMA_CM_EVENT_UNREACHABLE:
2889 conn = (kib_conn_t *)cmid->context;
2890 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
2891 conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
2892 CNETERR("%s: UNREACHABLE %d\n",
2893 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
2894 kiblnd_connreq_done(conn, -ENETDOWN);
2895 kiblnd_conn_decref(conn);
2896 return 0;
2897
2898 case RDMA_CM_EVENT_CONNECT_ERROR:
2899 conn = (kib_conn_t *)cmid->context;
2900 LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT ||
2901 conn->ibc_state == IBLND_CONN_PASSIVE_WAIT);
2902 CNETERR("%s: CONNECT ERROR %d\n",
2903 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->status);
2904 kiblnd_connreq_done(conn, -ENOTCONN);
2905 kiblnd_conn_decref(conn);
2906 return 0;
2907
2908 case RDMA_CM_EVENT_REJECTED:
2909 conn = (kib_conn_t *)cmid->context;
2910 switch (conn->ibc_state) {
2911 default:
2912 LBUG();
2913
2914 case IBLND_CONN_PASSIVE_WAIT:
2915 CERROR("%s: REJECTED %d\n",
2916 libcfs_nid2str(conn->ibc_peer->ibp_nid),
2917 event->status);
2918 kiblnd_connreq_done(conn, -ECONNRESET);
2919 break;
2920
2921 case IBLND_CONN_ACTIVE_CONNECT:
2922 kiblnd_rejected(conn, event->status,
2923 (void *)KIBLND_CONN_PARAM(event),
2924 KIBLND_CONN_PARAM_LEN(event));
2925 break;
2926 }
2927 kiblnd_conn_decref(conn);
2928 return 0;
2929
2930 case RDMA_CM_EVENT_ESTABLISHED:
2931 conn = (kib_conn_t *)cmid->context;
2932 switch (conn->ibc_state) {
2933 default:
2934 LBUG();
2935
2936 case IBLND_CONN_PASSIVE_WAIT:
2937 CDEBUG(D_NET, "ESTABLISHED (passive): %s\n",
2938 libcfs_nid2str(conn->ibc_peer->ibp_nid));
2939 kiblnd_connreq_done(conn, 0);
2940 break;
2941
2942 case IBLND_CONN_ACTIVE_CONNECT:
2943 CDEBUG(D_NET, "ESTABLISHED(active): %s\n",
2944 libcfs_nid2str(conn->ibc_peer->ibp_nid));
2945 kiblnd_check_connreply(conn,
2946 (void *)KIBLND_CONN_PARAM(event),
2947 KIBLND_CONN_PARAM_LEN(event));
2948 break;
2949 }
2950 /* net keeps its ref on conn! */
2951 return 0;
2952
2953 case RDMA_CM_EVENT_TIMEWAIT_EXIT:
2954 CDEBUG(D_NET, "Ignore TIMEWAIT_EXIT event\n");
2955 return 0;
2956 case RDMA_CM_EVENT_DISCONNECTED:
2957 conn = (kib_conn_t *)cmid->context;
2958 if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
2959 CERROR("%s DISCONNECTED\n",
2960 libcfs_nid2str(conn->ibc_peer->ibp_nid));
2961 kiblnd_connreq_done(conn, -ECONNRESET);
2962 } else {
2963 kiblnd_close_conn(conn, 0);
2964 }
2965 kiblnd_conn_decref(conn);
2966 cmid->context = NULL;
2967 return 0;
2968
2969 case RDMA_CM_EVENT_DEVICE_REMOVAL:
2970 LCONSOLE_ERROR_MSG(0x131,
2971 "Received notification of device removal\n"
2972 "Please shutdown LNET to allow this to proceed\n");
2973 /* Can't remove network from underneath LNET for now, so I have
2974 * to ignore this */
2975 return 0;
2976
2977 case RDMA_CM_EVENT_ADDR_CHANGE:
2978 LCONSOLE_INFO("Physical link changed (eg hca/port)\n");
2979 return 0;
2980 }
2981 }
2982
2983 static int
kiblnd_check_txs_locked(kib_conn_t * conn,struct list_head * txs)2984 kiblnd_check_txs_locked(kib_conn_t *conn, struct list_head *txs)
2985 {
2986 kib_tx_t *tx;
2987 struct list_head *ttmp;
2988
2989 list_for_each(ttmp, txs) {
2990 tx = list_entry(ttmp, kib_tx_t, tx_list);
2991
2992 if (txs != &conn->ibc_active_txs) {
2993 LASSERT(tx->tx_queued);
2994 } else {
2995 LASSERT(!tx->tx_queued);
2996 LASSERT(tx->tx_waiting || tx->tx_sending != 0);
2997 }
2998
2999 if (cfs_time_aftereq(jiffies, tx->tx_deadline)) {
3000 CERROR("Timed out tx: %s, %lu seconds\n",
3001 kiblnd_queue2str(conn, txs),
3002 cfs_duration_sec(jiffies - tx->tx_deadline));
3003 return 1;
3004 }
3005 }
3006
3007 return 0;
3008 }
3009
3010 static int
kiblnd_conn_timed_out_locked(kib_conn_t * conn)3011 kiblnd_conn_timed_out_locked(kib_conn_t *conn)
3012 {
3013 return kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue) ||
3014 kiblnd_check_txs_locked(conn, &conn->ibc_tx_noops) ||
3015 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_rsrvd) ||
3016 kiblnd_check_txs_locked(conn, &conn->ibc_tx_queue_nocred) ||
3017 kiblnd_check_txs_locked(conn, &conn->ibc_active_txs);
3018 }
3019
3020 static void
kiblnd_check_conns(int idx)3021 kiblnd_check_conns(int idx)
3022 {
3023 LIST_HEAD(closes);
3024 LIST_HEAD(checksends);
3025 struct list_head *peers = &kiblnd_data.kib_peers[idx];
3026 struct list_head *ptmp;
3027 kib_peer_t *peer;
3028 kib_conn_t *conn;
3029 kib_conn_t *tmp;
3030 struct list_head *ctmp;
3031 unsigned long flags;
3032
3033 /* NB. We expect to have a look at all the peers and not find any
3034 * RDMAs to time out, so we just use a shared lock while we
3035 * take a look... */
3036 read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
3037
3038 list_for_each(ptmp, peers) {
3039 peer = list_entry(ptmp, kib_peer_t, ibp_list);
3040
3041 list_for_each(ctmp, &peer->ibp_conns) {
3042 int timedout;
3043 int sendnoop;
3044
3045 conn = list_entry(ctmp, kib_conn_t, ibc_list);
3046
3047 LASSERT(conn->ibc_state == IBLND_CONN_ESTABLISHED);
3048
3049 spin_lock(&conn->ibc_lock);
3050
3051 sendnoop = kiblnd_need_noop(conn);
3052 timedout = kiblnd_conn_timed_out_locked(conn);
3053 if (!sendnoop && !timedout) {
3054 spin_unlock(&conn->ibc_lock);
3055 continue;
3056 }
3057
3058 if (timedout) {
3059 CERROR("Timed out RDMA with %s (%lu): c: %u, oc: %u, rc: %u\n",
3060 libcfs_nid2str(peer->ibp_nid),
3061 cfs_duration_sec(cfs_time_current() -
3062 peer->ibp_last_alive),
3063 conn->ibc_credits,
3064 conn->ibc_outstanding_credits,
3065 conn->ibc_reserved_credits);
3066 list_add(&conn->ibc_connd_list, &closes);
3067 } else {
3068 list_add(&conn->ibc_connd_list,
3069 &checksends);
3070 }
3071 /* +ref for 'closes' or 'checksends' */
3072 kiblnd_conn_addref(conn);
3073
3074 spin_unlock(&conn->ibc_lock);
3075 }
3076 }
3077
3078 read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
3079
3080 /* Handle timeout by closing the whole
3081 * connection. We can only be sure RDMA activity
3082 * has ceased once the QP has been modified. */
3083 list_for_each_entry_safe(conn, tmp, &closes, ibc_connd_list) {
3084 list_del(&conn->ibc_connd_list);
3085 kiblnd_close_conn(conn, -ETIMEDOUT);
3086 kiblnd_conn_decref(conn);
3087 }
3088
3089 /* In case we have enough credits to return via a
3090 * NOOP, but there were no non-blocking tx descs
3091 * free to do it last time... */
3092 while (!list_empty(&checksends)) {
3093 conn = list_entry(checksends.next,
3094 kib_conn_t, ibc_connd_list);
3095 list_del(&conn->ibc_connd_list);
3096 kiblnd_check_sends(conn);
3097 kiblnd_conn_decref(conn);
3098 }
3099 }
3100
3101 static void
kiblnd_disconnect_conn(kib_conn_t * conn)3102 kiblnd_disconnect_conn(kib_conn_t *conn)
3103 {
3104 LASSERT(!in_interrupt());
3105 LASSERT(current == kiblnd_data.kib_connd);
3106 LASSERT(conn->ibc_state == IBLND_CONN_CLOSING);
3107
3108 rdma_disconnect(conn->ibc_cmid);
3109 kiblnd_finalise_conn(conn);
3110
3111 kiblnd_peer_notify(conn->ibc_peer);
3112 }
3113
3114 int
kiblnd_connd(void * arg)3115 kiblnd_connd(void *arg)
3116 {
3117 wait_queue_t wait;
3118 unsigned long flags;
3119 kib_conn_t *conn;
3120 int timeout;
3121 int i;
3122 int dropped_lock;
3123 int peer_index = 0;
3124 unsigned long deadline = jiffies;
3125
3126 cfs_block_allsigs();
3127
3128 init_waitqueue_entry(&wait, current);
3129 kiblnd_data.kib_connd = current;
3130
3131 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3132
3133 while (!kiblnd_data.kib_shutdown) {
3134
3135 dropped_lock = 0;
3136
3137 if (!list_empty(&kiblnd_data.kib_connd_zombies)) {
3138 conn = list_entry(kiblnd_data. \
3139 kib_connd_zombies.next,
3140 kib_conn_t, ibc_list);
3141 list_del(&conn->ibc_list);
3142
3143 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
3144 flags);
3145 dropped_lock = 1;
3146
3147 kiblnd_destroy_conn(conn);
3148
3149 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3150 }
3151
3152 if (!list_empty(&kiblnd_data.kib_connd_conns)) {
3153 conn = list_entry(kiblnd_data.kib_connd_conns.next,
3154 kib_conn_t, ibc_list);
3155 list_del(&conn->ibc_list);
3156
3157 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock,
3158 flags);
3159 dropped_lock = 1;
3160
3161 kiblnd_disconnect_conn(conn);
3162 kiblnd_conn_decref(conn);
3163
3164 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3165 }
3166
3167 /* careful with the jiffy wrap... */
3168 timeout = (int)(deadline - jiffies);
3169 if (timeout <= 0) {
3170 const int n = 4;
3171 const int p = 1;
3172 int chunk = kiblnd_data.kib_peer_hash_size;
3173
3174 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3175 dropped_lock = 1;
3176
3177 /* Time to check for RDMA timeouts on a few more
3178 * peers: I do checks every 'p' seconds on a
3179 * proportion of the peer table and I need to check
3180 * every connection 'n' times within a timeout
3181 * interval, to ensure I detect a timeout on any
3182 * connection within (n+1)/n times the timeout
3183 * interval. */
3184
3185 if (*kiblnd_tunables.kib_timeout > n * p)
3186 chunk = (chunk * n * p) /
3187 *kiblnd_tunables.kib_timeout;
3188 if (chunk == 0)
3189 chunk = 1;
3190
3191 for (i = 0; i < chunk; i++) {
3192 kiblnd_check_conns(peer_index);
3193 peer_index = (peer_index + 1) %
3194 kiblnd_data.kib_peer_hash_size;
3195 }
3196
3197 deadline += p * HZ;
3198 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3199 }
3200
3201 if (dropped_lock)
3202 continue;
3203
3204 /* Nothing to do for 'timeout' */
3205 set_current_state(TASK_INTERRUPTIBLE);
3206 add_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3207 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3208
3209 schedule_timeout(timeout);
3210
3211 remove_wait_queue(&kiblnd_data.kib_connd_waitq, &wait);
3212 spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags);
3213 }
3214
3215 spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
3216
3217 kiblnd_thread_fini();
3218 return 0;
3219 }
3220
3221 void
kiblnd_qp_event(struct ib_event * event,void * arg)3222 kiblnd_qp_event(struct ib_event *event, void *arg)
3223 {
3224 kib_conn_t *conn = arg;
3225
3226 switch (event->event) {
3227 case IB_EVENT_COMM_EST:
3228 CDEBUG(D_NET, "%s established\n",
3229 libcfs_nid2str(conn->ibc_peer->ibp_nid));
3230 return;
3231
3232 default:
3233 CERROR("%s: Async QP event type %d\n",
3234 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3235 return;
3236 }
3237 }
3238
3239 static void
kiblnd_complete(struct ib_wc * wc)3240 kiblnd_complete(struct ib_wc *wc)
3241 {
3242 switch (kiblnd_wreqid2type(wc->wr_id)) {
3243 default:
3244 LBUG();
3245
3246 case IBLND_WID_RDMA:
3247 /* We only get RDMA completion notification if it fails. All
3248 * subsequent work items, including the final SEND will fail
3249 * too. However we can't print out any more info about the
3250 * failing RDMA because 'tx' might be back on the idle list or
3251 * even reused already if we didn't manage to post all our work
3252 * items */
3253 CNETERR("RDMA (tx: %p) failed: %d\n",
3254 kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3255 return;
3256
3257 case IBLND_WID_TX:
3258 kiblnd_tx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status);
3259 return;
3260
3261 case IBLND_WID_RX:
3262 kiblnd_rx_complete(kiblnd_wreqid2ptr(wc->wr_id), wc->status,
3263 wc->byte_len);
3264 return;
3265 }
3266 }
3267
3268 void
kiblnd_cq_completion(struct ib_cq * cq,void * arg)3269 kiblnd_cq_completion(struct ib_cq *cq, void *arg)
3270 {
3271 /* NB I'm not allowed to schedule this conn once its refcount has
3272 * reached 0. Since fundamentally I'm racing with scheduler threads
3273 * consuming my CQ I could be called after all completions have
3274 * occurred. But in this case, ibc_nrx == 0 && ibc_nsends_posted == 0
3275 * and this CQ is about to be destroyed so I NOOP. */
3276 kib_conn_t *conn = (kib_conn_t *)arg;
3277 struct kib_sched_info *sched = conn->ibc_sched;
3278 unsigned long flags;
3279
3280 LASSERT(cq == conn->ibc_cq);
3281
3282 spin_lock_irqsave(&sched->ibs_lock, flags);
3283
3284 conn->ibc_ready = 1;
3285
3286 if (!conn->ibc_scheduled &&
3287 (conn->ibc_nrx > 0 ||
3288 conn->ibc_nsends_posted > 0)) {
3289 kiblnd_conn_addref(conn); /* +1 ref for sched_conns */
3290 conn->ibc_scheduled = 1;
3291 list_add_tail(&conn->ibc_sched_list, &sched->ibs_conns);
3292
3293 if (waitqueue_active(&sched->ibs_waitq))
3294 wake_up(&sched->ibs_waitq);
3295 }
3296
3297 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3298 }
3299
3300 void
kiblnd_cq_event(struct ib_event * event,void * arg)3301 kiblnd_cq_event(struct ib_event *event, void *arg)
3302 {
3303 kib_conn_t *conn = arg;
3304
3305 CERROR("%s: async CQ event type %d\n",
3306 libcfs_nid2str(conn->ibc_peer->ibp_nid), event->event);
3307 }
3308
3309 int
kiblnd_scheduler(void * arg)3310 kiblnd_scheduler(void *arg)
3311 {
3312 long id = (long)arg;
3313 struct kib_sched_info *sched;
3314 kib_conn_t *conn;
3315 wait_queue_t wait;
3316 unsigned long flags;
3317 struct ib_wc wc;
3318 int did_something;
3319 int busy_loops = 0;
3320 int rc;
3321
3322 cfs_block_allsigs();
3323
3324 init_waitqueue_entry(&wait, current);
3325
3326 sched = kiblnd_data.kib_scheds[KIB_THREAD_CPT(id)];
3327
3328 rc = cfs_cpt_bind(lnet_cpt_table(), sched->ibs_cpt);
3329 if (rc != 0) {
3330 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",
3331 sched->ibs_cpt);
3332 }
3333
3334 spin_lock_irqsave(&sched->ibs_lock, flags);
3335
3336 while (!kiblnd_data.kib_shutdown) {
3337 if (busy_loops++ >= IBLND_RESCHED) {
3338 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3339
3340 cond_resched();
3341 busy_loops = 0;
3342
3343 spin_lock_irqsave(&sched->ibs_lock, flags);
3344 }
3345
3346 did_something = 0;
3347
3348 if (!list_empty(&sched->ibs_conns)) {
3349 conn = list_entry(sched->ibs_conns.next,
3350 kib_conn_t, ibc_sched_list);
3351 /* take over kib_sched_conns' ref on conn... */
3352 LASSERT(conn->ibc_scheduled);
3353 list_del(&conn->ibc_sched_list);
3354 conn->ibc_ready = 0;
3355
3356 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3357
3358 rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3359 if (rc == 0) {
3360 rc = ib_req_notify_cq(conn->ibc_cq,
3361 IB_CQ_NEXT_COMP);
3362 if (rc < 0) {
3363 CWARN("%s: ib_req_notify_cq failed: %d, closing connection\n",
3364 libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
3365 kiblnd_close_conn(conn, -EIO);
3366 kiblnd_conn_decref(conn);
3367 spin_lock_irqsave(&sched->ibs_lock,
3368 flags);
3369 continue;
3370 }
3371
3372 rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
3373 }
3374
3375 if (rc < 0) {
3376 CWARN("%s: ib_poll_cq failed: %d, closing connection\n",
3377 libcfs_nid2str(conn->ibc_peer->ibp_nid),
3378 rc);
3379 kiblnd_close_conn(conn, -EIO);
3380 kiblnd_conn_decref(conn);
3381 spin_lock_irqsave(&sched->ibs_lock, flags);
3382 continue;
3383 }
3384
3385 spin_lock_irqsave(&sched->ibs_lock, flags);
3386
3387 if (rc != 0 || conn->ibc_ready) {
3388 /* There may be another completion waiting; get
3389 * another scheduler to check while I handle
3390 * this one... */
3391 /* +1 ref for sched_conns */
3392 kiblnd_conn_addref(conn);
3393 list_add_tail(&conn->ibc_sched_list,
3394 &sched->ibs_conns);
3395 if (waitqueue_active(&sched->ibs_waitq))
3396 wake_up(&sched->ibs_waitq);
3397 } else {
3398 conn->ibc_scheduled = 0;
3399 }
3400
3401 if (rc != 0) {
3402 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3403 kiblnd_complete(&wc);
3404
3405 spin_lock_irqsave(&sched->ibs_lock, flags);
3406 }
3407
3408 kiblnd_conn_decref(conn); /* ...drop my ref from above */
3409 did_something = 1;
3410 }
3411
3412 if (did_something)
3413 continue;
3414
3415 set_current_state(TASK_INTERRUPTIBLE);
3416 add_wait_queue_exclusive(&sched->ibs_waitq, &wait);
3417 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3418
3419 schedule();
3420 busy_loops = 0;
3421
3422 remove_wait_queue(&sched->ibs_waitq, &wait);
3423 spin_lock_irqsave(&sched->ibs_lock, flags);
3424 }
3425
3426 spin_unlock_irqrestore(&sched->ibs_lock, flags);
3427
3428 kiblnd_thread_fini();
3429 return 0;
3430 }
3431
3432 int
kiblnd_failover_thread(void * arg)3433 kiblnd_failover_thread(void *arg)
3434 {
3435 rwlock_t *glock = &kiblnd_data.kib_global_lock;
3436 kib_dev_t *dev;
3437 wait_queue_t wait;
3438 unsigned long flags;
3439 int rc;
3440
3441 LASSERT(*kiblnd_tunables.kib_dev_failover != 0);
3442
3443 cfs_block_allsigs();
3444
3445 init_waitqueue_entry(&wait, current);
3446 write_lock_irqsave(glock, flags);
3447
3448 while (!kiblnd_data.kib_shutdown) {
3449 int do_failover = 0;
3450 int long_sleep;
3451
3452 list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
3453 ibd_fail_list) {
3454 if (time_before(cfs_time_current(),
3455 dev->ibd_next_failover))
3456 continue;
3457 do_failover = 1;
3458 break;
3459 }
3460
3461 if (do_failover) {
3462 list_del_init(&dev->ibd_fail_list);
3463 dev->ibd_failover = 1;
3464 write_unlock_irqrestore(glock, flags);
3465
3466 rc = kiblnd_dev_failover(dev);
3467
3468 write_lock_irqsave(glock, flags);
3469
3470 LASSERT(dev->ibd_failover);
3471 dev->ibd_failover = 0;
3472 if (rc >= 0) { /* Device is OK or failover succeed */
3473 dev->ibd_next_failover = cfs_time_shift(3);
3474 continue;
3475 }
3476
3477 /* failed to failover, retry later */
3478 dev->ibd_next_failover =
3479 cfs_time_shift(min(dev->ibd_failed_failover, 10));
3480 if (kiblnd_dev_can_failover(dev)) {
3481 list_add_tail(&dev->ibd_fail_list,
3482 &kiblnd_data.kib_failed_devs);
3483 }
3484
3485 continue;
3486 }
3487
3488 /* long sleep if no more pending failover */
3489 long_sleep = list_empty(&kiblnd_data.kib_failed_devs);
3490
3491 set_current_state(TASK_INTERRUPTIBLE);
3492 add_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3493 write_unlock_irqrestore(glock, flags);
3494
3495 rc = schedule_timeout(long_sleep ? cfs_time_seconds(10) :
3496 cfs_time_seconds(1));
3497 remove_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
3498 write_lock_irqsave(glock, flags);
3499
3500 if (!long_sleep || rc != 0)
3501 continue;
3502
3503 /* have a long sleep, routine check all active devices,
3504 * we need checking like this because if there is not active
3505 * connection on the dev and no SEND from local, we may listen
3506 * on wrong HCA for ever while there is a bonding failover */
3507 list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
3508 if (kiblnd_dev_can_failover(dev)) {
3509 list_add_tail(&dev->ibd_fail_list,
3510 &kiblnd_data.kib_failed_devs);
3511 }
3512 }
3513 }
3514
3515 write_unlock_irqrestore(glock, flags);
3516
3517 kiblnd_thread_fini();
3518 return 0;
3519 }
3520