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) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
29 *
30 * Copyright (c) 2011, 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/lnet/lib-move.c
37 *
38 * Data movement routines
39 */
40
41 #define DEBUG_SUBSYSTEM S_LNET
42
43 #include "../../include/linux/lnet/lib-lnet.h"
44
45 static int local_nid_dist_zero = 1;
46 module_param(local_nid_dist_zero, int, 0444);
47 MODULE_PARM_DESC(local_nid_dist_zero, "Reserved");
48
49 int
lnet_fail_nid(lnet_nid_t nid,unsigned int threshold)50 lnet_fail_nid(lnet_nid_t nid, unsigned int threshold)
51 {
52 lnet_test_peer_t *tp;
53 struct list_head *el;
54 struct list_head *next;
55 struct list_head cull;
56
57 LASSERT(the_lnet.ln_init);
58
59 /* NB: use lnet_net_lock(0) to serialize operations on test peers */
60 if (threshold != 0) {
61 /* Adding a new entry */
62 LIBCFS_ALLOC(tp, sizeof(*tp));
63 if (tp == NULL)
64 return -ENOMEM;
65
66 tp->tp_nid = nid;
67 tp->tp_threshold = threshold;
68
69 lnet_net_lock(0);
70 list_add_tail(&tp->tp_list, &the_lnet.ln_test_peers);
71 lnet_net_unlock(0);
72 return 0;
73 }
74
75 /* removing entries */
76 INIT_LIST_HEAD(&cull);
77
78 lnet_net_lock(0);
79
80 list_for_each_safe(el, next, &the_lnet.ln_test_peers) {
81 tp = list_entry(el, lnet_test_peer_t, tp_list);
82
83 if (tp->tp_threshold == 0 || /* needs culling anyway */
84 nid == LNET_NID_ANY || /* removing all entries */
85 tp->tp_nid == nid) { /* matched this one */
86 list_del(&tp->tp_list);
87 list_add(&tp->tp_list, &cull);
88 }
89 }
90
91 lnet_net_unlock(0);
92
93 while (!list_empty(&cull)) {
94 tp = list_entry(cull.next, lnet_test_peer_t, tp_list);
95
96 list_del(&tp->tp_list);
97 LIBCFS_FREE(tp, sizeof(*tp));
98 }
99 return 0;
100 }
101
102 static int
fail_peer(lnet_nid_t nid,int outgoing)103 fail_peer(lnet_nid_t nid, int outgoing)
104 {
105 lnet_test_peer_t *tp;
106 struct list_head *el;
107 struct list_head *next;
108 struct list_head cull;
109 int fail = 0;
110
111 INIT_LIST_HEAD(&cull);
112
113 /* NB: use lnet_net_lock(0) to serialize operations on test peers */
114 lnet_net_lock(0);
115
116 list_for_each_safe(el, next, &the_lnet.ln_test_peers) {
117 tp = list_entry(el, lnet_test_peer_t, tp_list);
118
119 if (tp->tp_threshold == 0) {
120 /* zombie entry */
121 if (outgoing) {
122 /* only cull zombies on outgoing tests,
123 * since we may be at interrupt priority on
124 * incoming messages. */
125 list_del(&tp->tp_list);
126 list_add(&tp->tp_list, &cull);
127 }
128 continue;
129 }
130
131 if (tp->tp_nid == LNET_NID_ANY || /* fail every peer */
132 nid == tp->tp_nid) { /* fail this peer */
133 fail = 1;
134
135 if (tp->tp_threshold != LNET_MD_THRESH_INF) {
136 tp->tp_threshold--;
137 if (outgoing &&
138 tp->tp_threshold == 0) {
139 /* see above */
140 list_del(&tp->tp_list);
141 list_add(&tp->tp_list, &cull);
142 }
143 }
144 break;
145 }
146 }
147
148 lnet_net_unlock(0);
149
150 while (!list_empty(&cull)) {
151 tp = list_entry(cull.next, lnet_test_peer_t, tp_list);
152 list_del(&tp->tp_list);
153
154 LIBCFS_FREE(tp, sizeof(*tp));
155 }
156
157 return fail;
158 }
159
160 unsigned int
lnet_iov_nob(unsigned int niov,struct kvec * iov)161 lnet_iov_nob(unsigned int niov, struct kvec *iov)
162 {
163 unsigned int nob = 0;
164
165 while (niov-- > 0)
166 nob += (iov++)->iov_len;
167
168 return nob;
169 }
170 EXPORT_SYMBOL(lnet_iov_nob);
171
172 void
lnet_copy_iov2iov(unsigned int ndiov,struct kvec * diov,unsigned int doffset,unsigned int nsiov,struct kvec * siov,unsigned int soffset,unsigned int nob)173 lnet_copy_iov2iov(unsigned int ndiov, struct kvec *diov, unsigned int doffset,
174 unsigned int nsiov, struct kvec *siov, unsigned int soffset,
175 unsigned int nob)
176 {
177 /* NB diov, siov are READ-ONLY */
178 unsigned int this_nob;
179
180 if (nob == 0)
181 return;
182
183 /* skip complete frags before 'doffset' */
184 LASSERT(ndiov > 0);
185 while (doffset >= diov->iov_len) {
186 doffset -= diov->iov_len;
187 diov++;
188 ndiov--;
189 LASSERT(ndiov > 0);
190 }
191
192 /* skip complete frags before 'soffset' */
193 LASSERT(nsiov > 0);
194 while (soffset >= siov->iov_len) {
195 soffset -= siov->iov_len;
196 siov++;
197 nsiov--;
198 LASSERT(nsiov > 0);
199 }
200
201 do {
202 LASSERT(ndiov > 0);
203 LASSERT(nsiov > 0);
204 this_nob = min(diov->iov_len - doffset,
205 siov->iov_len - soffset);
206 this_nob = min(this_nob, nob);
207
208 memcpy((char *)diov->iov_base + doffset,
209 (char *)siov->iov_base + soffset, this_nob);
210 nob -= this_nob;
211
212 if (diov->iov_len > doffset + this_nob) {
213 doffset += this_nob;
214 } else {
215 diov++;
216 ndiov--;
217 doffset = 0;
218 }
219
220 if (siov->iov_len > soffset + this_nob) {
221 soffset += this_nob;
222 } else {
223 siov++;
224 nsiov--;
225 soffset = 0;
226 }
227 } while (nob > 0);
228 }
229 EXPORT_SYMBOL(lnet_copy_iov2iov);
230
231 int
lnet_extract_iov(int dst_niov,struct kvec * dst,int src_niov,struct kvec * src,unsigned int offset,unsigned int len)232 lnet_extract_iov(int dst_niov, struct kvec *dst,
233 int src_niov, struct kvec *src,
234 unsigned int offset, unsigned int len)
235 {
236 /* Initialise 'dst' to the subset of 'src' starting at 'offset',
237 * for exactly 'len' bytes, and return the number of entries.
238 * NB not destructive to 'src' */
239 unsigned int frag_len;
240 unsigned int niov;
241
242 if (len == 0) /* no data => */
243 return 0; /* no frags */
244
245 LASSERT(src_niov > 0);
246 while (offset >= src->iov_len) { /* skip initial frags */
247 offset -= src->iov_len;
248 src_niov--;
249 src++;
250 LASSERT(src_niov > 0);
251 }
252
253 niov = 1;
254 for (;;) {
255 LASSERT(src_niov > 0);
256 LASSERT((int)niov <= dst_niov);
257
258 frag_len = src->iov_len - offset;
259 dst->iov_base = ((char *)src->iov_base) + offset;
260
261 if (len <= frag_len) {
262 dst->iov_len = len;
263 return niov;
264 }
265
266 dst->iov_len = frag_len;
267
268 len -= frag_len;
269 dst++;
270 src++;
271 niov++;
272 src_niov--;
273 offset = 0;
274 }
275 }
276 EXPORT_SYMBOL(lnet_extract_iov);
277
278
279 unsigned int
lnet_kiov_nob(unsigned int niov,lnet_kiov_t * kiov)280 lnet_kiov_nob(unsigned int niov, lnet_kiov_t *kiov)
281 {
282 unsigned int nob = 0;
283
284 while (niov-- > 0)
285 nob += (kiov++)->kiov_len;
286
287 return nob;
288 }
289 EXPORT_SYMBOL(lnet_kiov_nob);
290
291 void
lnet_copy_kiov2kiov(unsigned int ndiov,lnet_kiov_t * diov,unsigned int doffset,unsigned int nsiov,lnet_kiov_t * siov,unsigned int soffset,unsigned int nob)292 lnet_copy_kiov2kiov(unsigned int ndiov, lnet_kiov_t *diov, unsigned int doffset,
293 unsigned int nsiov, lnet_kiov_t *siov, unsigned int soffset,
294 unsigned int nob)
295 {
296 /* NB diov, siov are READ-ONLY */
297 unsigned int this_nob;
298 char *daddr = NULL;
299 char *saddr = NULL;
300
301 if (nob == 0)
302 return;
303
304 LASSERT(!in_interrupt());
305
306 LASSERT(ndiov > 0);
307 while (doffset >= diov->kiov_len) {
308 doffset -= diov->kiov_len;
309 diov++;
310 ndiov--;
311 LASSERT(ndiov > 0);
312 }
313
314 LASSERT(nsiov > 0);
315 while (soffset >= siov->kiov_len) {
316 soffset -= siov->kiov_len;
317 siov++;
318 nsiov--;
319 LASSERT(nsiov > 0);
320 }
321
322 do {
323 LASSERT(ndiov > 0);
324 LASSERT(nsiov > 0);
325 this_nob = min(diov->kiov_len - doffset,
326 siov->kiov_len - soffset);
327 this_nob = min(this_nob, nob);
328
329 if (daddr == NULL)
330 daddr = ((char *)kmap(diov->kiov_page)) +
331 diov->kiov_offset + doffset;
332 if (saddr == NULL)
333 saddr = ((char *)kmap(siov->kiov_page)) +
334 siov->kiov_offset + soffset;
335
336 /* Vanishing risk of kmap deadlock when mapping 2 pages.
337 * However in practice at least one of the kiovs will be mapped
338 * kernel pages and the map/unmap will be NOOPs */
339
340 memcpy(daddr, saddr, this_nob);
341 nob -= this_nob;
342
343 if (diov->kiov_len > doffset + this_nob) {
344 daddr += this_nob;
345 doffset += this_nob;
346 } else {
347 kunmap(diov->kiov_page);
348 daddr = NULL;
349 diov++;
350 ndiov--;
351 doffset = 0;
352 }
353
354 if (siov->kiov_len > soffset + this_nob) {
355 saddr += this_nob;
356 soffset += this_nob;
357 } else {
358 kunmap(siov->kiov_page);
359 saddr = NULL;
360 siov++;
361 nsiov--;
362 soffset = 0;
363 }
364 } while (nob > 0);
365
366 if (daddr != NULL)
367 kunmap(diov->kiov_page);
368 if (saddr != NULL)
369 kunmap(siov->kiov_page);
370 }
371 EXPORT_SYMBOL(lnet_copy_kiov2kiov);
372
373 void
lnet_copy_kiov2iov(unsigned int niov,struct kvec * iov,unsigned int iovoffset,unsigned int nkiov,lnet_kiov_t * kiov,unsigned int kiovoffset,unsigned int nob)374 lnet_copy_kiov2iov(unsigned int niov, struct kvec *iov, unsigned int iovoffset,
375 unsigned int nkiov, lnet_kiov_t *kiov,
376 unsigned int kiovoffset, unsigned int nob)
377 {
378 /* NB iov, kiov are READ-ONLY */
379 unsigned int this_nob;
380 char *addr = NULL;
381
382 if (nob == 0)
383 return;
384
385 LASSERT(!in_interrupt());
386
387 LASSERT(niov > 0);
388 while (iovoffset >= iov->iov_len) {
389 iovoffset -= iov->iov_len;
390 iov++;
391 niov--;
392 LASSERT(niov > 0);
393 }
394
395 LASSERT(nkiov > 0);
396 while (kiovoffset >= kiov->kiov_len) {
397 kiovoffset -= kiov->kiov_len;
398 kiov++;
399 nkiov--;
400 LASSERT(nkiov > 0);
401 }
402
403 do {
404 LASSERT(niov > 0);
405 LASSERT(nkiov > 0);
406 this_nob = min(iov->iov_len - iovoffset,
407 (__kernel_size_t) kiov->kiov_len - kiovoffset);
408 this_nob = min(this_nob, nob);
409
410 if (addr == NULL)
411 addr = ((char *)kmap(kiov->kiov_page)) +
412 kiov->kiov_offset + kiovoffset;
413
414 memcpy((char *)iov->iov_base + iovoffset, addr, this_nob);
415 nob -= this_nob;
416
417 if (iov->iov_len > iovoffset + this_nob) {
418 iovoffset += this_nob;
419 } else {
420 iov++;
421 niov--;
422 iovoffset = 0;
423 }
424
425 if (kiov->kiov_len > kiovoffset + this_nob) {
426 addr += this_nob;
427 kiovoffset += this_nob;
428 } else {
429 kunmap(kiov->kiov_page);
430 addr = NULL;
431 kiov++;
432 nkiov--;
433 kiovoffset = 0;
434 }
435
436 } while (nob > 0);
437
438 if (addr != NULL)
439 kunmap(kiov->kiov_page);
440 }
441 EXPORT_SYMBOL(lnet_copy_kiov2iov);
442
443 void
lnet_copy_iov2kiov(unsigned int nkiov,lnet_kiov_t * kiov,unsigned int kiovoffset,unsigned int niov,struct kvec * iov,unsigned int iovoffset,unsigned int nob)444 lnet_copy_iov2kiov(unsigned int nkiov, lnet_kiov_t *kiov,
445 unsigned int kiovoffset, unsigned int niov,
446 struct kvec *iov, unsigned int iovoffset,
447 unsigned int nob)
448 {
449 /* NB kiov, iov are READ-ONLY */
450 unsigned int this_nob;
451 char *addr = NULL;
452
453 if (nob == 0)
454 return;
455
456 LASSERT(!in_interrupt());
457
458 LASSERT(nkiov > 0);
459 while (kiovoffset >= kiov->kiov_len) {
460 kiovoffset -= kiov->kiov_len;
461 kiov++;
462 nkiov--;
463 LASSERT(nkiov > 0);
464 }
465
466 LASSERT(niov > 0);
467 while (iovoffset >= iov->iov_len) {
468 iovoffset -= iov->iov_len;
469 iov++;
470 niov--;
471 LASSERT(niov > 0);
472 }
473
474 do {
475 LASSERT(nkiov > 0);
476 LASSERT(niov > 0);
477 this_nob = min((__kernel_size_t) kiov->kiov_len - kiovoffset,
478 iov->iov_len - iovoffset);
479 this_nob = min(this_nob, nob);
480
481 if (addr == NULL)
482 addr = ((char *)kmap(kiov->kiov_page)) +
483 kiov->kiov_offset + kiovoffset;
484
485 memcpy(addr, (char *)iov->iov_base + iovoffset, this_nob);
486 nob -= this_nob;
487
488 if (kiov->kiov_len > kiovoffset + this_nob) {
489 addr += this_nob;
490 kiovoffset += this_nob;
491 } else {
492 kunmap(kiov->kiov_page);
493 addr = NULL;
494 kiov++;
495 nkiov--;
496 kiovoffset = 0;
497 }
498
499 if (iov->iov_len > iovoffset + this_nob) {
500 iovoffset += this_nob;
501 } else {
502 iov++;
503 niov--;
504 iovoffset = 0;
505 }
506 } while (nob > 0);
507
508 if (addr != NULL)
509 kunmap(kiov->kiov_page);
510 }
511 EXPORT_SYMBOL(lnet_copy_iov2kiov);
512
513 int
lnet_extract_kiov(int dst_niov,lnet_kiov_t * dst,int src_niov,lnet_kiov_t * src,unsigned int offset,unsigned int len)514 lnet_extract_kiov(int dst_niov, lnet_kiov_t *dst,
515 int src_niov, lnet_kiov_t *src,
516 unsigned int offset, unsigned int len)
517 {
518 /* Initialise 'dst' to the subset of 'src' starting at 'offset',
519 * for exactly 'len' bytes, and return the number of entries.
520 * NB not destructive to 'src' */
521 unsigned int frag_len;
522 unsigned int niov;
523
524 if (len == 0) /* no data => */
525 return 0; /* no frags */
526
527 LASSERT(src_niov > 0);
528 while (offset >= src->kiov_len) { /* skip initial frags */
529 offset -= src->kiov_len;
530 src_niov--;
531 src++;
532 LASSERT(src_niov > 0);
533 }
534
535 niov = 1;
536 for (;;) {
537 LASSERT(src_niov > 0);
538 LASSERT((int)niov <= dst_niov);
539
540 frag_len = src->kiov_len - offset;
541 dst->kiov_page = src->kiov_page;
542 dst->kiov_offset = src->kiov_offset + offset;
543
544 if (len <= frag_len) {
545 dst->kiov_len = len;
546 LASSERT(dst->kiov_offset + dst->kiov_len
547 <= PAGE_CACHE_SIZE);
548 return niov;
549 }
550
551 dst->kiov_len = frag_len;
552 LASSERT(dst->kiov_offset + dst->kiov_len <= PAGE_CACHE_SIZE);
553
554 len -= frag_len;
555 dst++;
556 src++;
557 niov++;
558 src_niov--;
559 offset = 0;
560 }
561 }
562 EXPORT_SYMBOL(lnet_extract_kiov);
563
564 static void
lnet_ni_recv(lnet_ni_t * ni,void * private,lnet_msg_t * msg,int delayed,unsigned int offset,unsigned int mlen,unsigned int rlen)565 lnet_ni_recv(lnet_ni_t *ni, void *private, lnet_msg_t *msg, int delayed,
566 unsigned int offset, unsigned int mlen, unsigned int rlen)
567 {
568 unsigned int niov = 0;
569 struct kvec *iov = NULL;
570 lnet_kiov_t *kiov = NULL;
571 int rc;
572
573 LASSERT(!in_interrupt());
574 LASSERT(mlen == 0 || msg != NULL);
575
576 if (msg != NULL) {
577 LASSERT(msg->msg_receiving);
578 LASSERT(!msg->msg_sending);
579 LASSERT(rlen == msg->msg_len);
580 LASSERT(mlen <= msg->msg_len);
581 LASSERT(msg->msg_offset == offset);
582 LASSERT(msg->msg_wanted == mlen);
583
584 msg->msg_receiving = 0;
585
586 if (mlen != 0) {
587 niov = msg->msg_niov;
588 iov = msg->msg_iov;
589 kiov = msg->msg_kiov;
590
591 LASSERT(niov > 0);
592 LASSERT((iov == NULL) != (kiov == NULL));
593 }
594 }
595
596 rc = (ni->ni_lnd->lnd_recv)(ni, private, msg, delayed,
597 niov, iov, kiov, offset, mlen, rlen);
598 if (rc < 0)
599 lnet_finalize(ni, msg, rc);
600 }
601
602 static void
lnet_setpayloadbuffer(lnet_msg_t * msg)603 lnet_setpayloadbuffer(lnet_msg_t *msg)
604 {
605 lnet_libmd_t *md = msg->msg_md;
606
607 LASSERT(msg->msg_len > 0);
608 LASSERT(!msg->msg_routing);
609 LASSERT(md != NULL);
610 LASSERT(msg->msg_niov == 0);
611 LASSERT(msg->msg_iov == NULL);
612 LASSERT(msg->msg_kiov == NULL);
613
614 msg->msg_niov = md->md_niov;
615 if ((md->md_options & LNET_MD_KIOV) != 0)
616 msg->msg_kiov = md->md_iov.kiov;
617 else
618 msg->msg_iov = md->md_iov.iov;
619 }
620
621 void
lnet_prep_send(lnet_msg_t * msg,int type,lnet_process_id_t target,unsigned int offset,unsigned int len)622 lnet_prep_send(lnet_msg_t *msg, int type, lnet_process_id_t target,
623 unsigned int offset, unsigned int len)
624 {
625 msg->msg_type = type;
626 msg->msg_target = target;
627 msg->msg_len = len;
628 msg->msg_offset = offset;
629
630 if (len != 0)
631 lnet_setpayloadbuffer(msg);
632
633 memset(&msg->msg_hdr, 0, sizeof(msg->msg_hdr));
634 msg->msg_hdr.type = cpu_to_le32(type);
635 msg->msg_hdr.dest_nid = cpu_to_le64(target.nid);
636 msg->msg_hdr.dest_pid = cpu_to_le32(target.pid);
637 /* src_nid will be set later */
638 msg->msg_hdr.src_pid = cpu_to_le32(the_lnet.ln_pid);
639 msg->msg_hdr.payload_length = cpu_to_le32(len);
640 }
641
642 static void
lnet_ni_send(lnet_ni_t * ni,lnet_msg_t * msg)643 lnet_ni_send(lnet_ni_t *ni, lnet_msg_t *msg)
644 {
645 void *priv = msg->msg_private;
646 int rc;
647
648 LASSERT(!in_interrupt());
649 LASSERT(LNET_NETTYP(LNET_NIDNET(ni->ni_nid)) == LOLND ||
650 (msg->msg_txcredit && msg->msg_peertxcredit));
651
652 rc = (ni->ni_lnd->lnd_send)(ni, priv, msg);
653 if (rc < 0)
654 lnet_finalize(ni, msg, rc);
655 }
656
657 static int
lnet_ni_eager_recv(lnet_ni_t * ni,lnet_msg_t * msg)658 lnet_ni_eager_recv(lnet_ni_t *ni, lnet_msg_t *msg)
659 {
660 int rc;
661
662 LASSERT(!msg->msg_sending);
663 LASSERT(msg->msg_receiving);
664 LASSERT(!msg->msg_rx_ready_delay);
665 LASSERT(ni->ni_lnd->lnd_eager_recv != NULL);
666
667 msg->msg_rx_ready_delay = 1;
668 rc = (ni->ni_lnd->lnd_eager_recv)(ni, msg->msg_private, msg,
669 &msg->msg_private);
670 if (rc != 0) {
671 CERROR("recv from %s / send to %s aborted: eager_recv failed %d\n",
672 libcfs_nid2str(msg->msg_rxpeer->lp_nid),
673 libcfs_id2str(msg->msg_target), rc);
674 LASSERT(rc < 0); /* required by my callers */
675 }
676
677 return rc;
678 }
679
680 /* NB: caller shall hold a ref on 'lp' as I'd drop lnet_net_lock */
681 static void
lnet_ni_query_locked(lnet_ni_t * ni,lnet_peer_t * lp)682 lnet_ni_query_locked(lnet_ni_t *ni, lnet_peer_t *lp)
683 {
684 unsigned long last_alive = 0;
685
686 LASSERT(lnet_peer_aliveness_enabled(lp));
687 LASSERT(ni->ni_lnd->lnd_query != NULL);
688
689 lnet_net_unlock(lp->lp_cpt);
690 (ni->ni_lnd->lnd_query)(ni, lp->lp_nid, &last_alive);
691 lnet_net_lock(lp->lp_cpt);
692
693 lp->lp_last_query = cfs_time_current();
694
695 if (last_alive != 0) /* NI has updated timestamp */
696 lp->lp_last_alive = last_alive;
697 }
698
699 /* NB: always called with lnet_net_lock held */
700 static inline int
lnet_peer_is_alive(lnet_peer_t * lp,unsigned long now)701 lnet_peer_is_alive(lnet_peer_t *lp, unsigned long now)
702 {
703 int alive;
704 unsigned long deadline;
705
706 LASSERT(lnet_peer_aliveness_enabled(lp));
707
708 /* Trust lnet_notify() if it has more recent aliveness news, but
709 * ignore the initial assumed death (see lnet_peers_start_down()).
710 */
711 if (!lp->lp_alive && lp->lp_alive_count > 0 &&
712 cfs_time_aftereq(lp->lp_timestamp, lp->lp_last_alive))
713 return 0;
714
715 deadline = cfs_time_add(lp->lp_last_alive,
716 cfs_time_seconds(lp->lp_ni->ni_peertimeout));
717 alive = cfs_time_after(deadline, now);
718
719 /* Update obsolete lp_alive except for routers assumed to be dead
720 * initially, because router checker would update aliveness in this
721 * case, and moreover lp_last_alive at peer creation is assumed.
722 */
723 if (alive && !lp->lp_alive &&
724 !(lnet_isrouter(lp) && lp->lp_alive_count == 0))
725 lnet_notify_locked(lp, 0, 1, lp->lp_last_alive);
726
727 return alive;
728 }
729
730
731 /* NB: returns 1 when alive, 0 when dead, negative when error;
732 * may drop the lnet_net_lock */
733 static int
lnet_peer_alive_locked(lnet_peer_t * lp)734 lnet_peer_alive_locked(lnet_peer_t *lp)
735 {
736 unsigned long now = cfs_time_current();
737
738 if (!lnet_peer_aliveness_enabled(lp))
739 return -ENODEV;
740
741 if (lnet_peer_is_alive(lp, now))
742 return 1;
743
744 /* Peer appears dead, but we should avoid frequent NI queries (at
745 * most once per lnet_queryinterval seconds). */
746 if (lp->lp_last_query != 0) {
747 static const int lnet_queryinterval = 1;
748
749 unsigned long next_query =
750 cfs_time_add(lp->lp_last_query,
751 cfs_time_seconds(lnet_queryinterval));
752
753 if (time_before(now, next_query)) {
754 if (lp->lp_alive)
755 CWARN("Unexpected aliveness of peer %s: %d < %d (%d/%d)\n",
756 libcfs_nid2str(lp->lp_nid),
757 (int)now, (int)next_query,
758 lnet_queryinterval,
759 lp->lp_ni->ni_peertimeout);
760 return 0;
761 }
762 }
763
764 /* query NI for latest aliveness news */
765 lnet_ni_query_locked(lp->lp_ni, lp);
766
767 if (lnet_peer_is_alive(lp, now))
768 return 1;
769
770 lnet_notify_locked(lp, 0, 0, lp->lp_last_alive);
771 return 0;
772 }
773
774 /**
775 * \param msg The message to be sent.
776 * \param do_send True if lnet_ni_send() should be called in this function.
777 * lnet_send() is going to lnet_net_unlock immediately after this, so
778 * it sets do_send FALSE and I don't do the unlock/send/lock bit.
779 *
780 * \retval 0 If \a msg sent or OK to send.
781 * \retval EAGAIN If \a msg blocked for credit.
782 * \retval EHOSTUNREACH If the next hop of the message appears dead.
783 * \retval ECANCELED If the MD of the message has been unlinked.
784 */
785 static int
lnet_post_send_locked(lnet_msg_t * msg,int do_send)786 lnet_post_send_locked(lnet_msg_t *msg, int do_send)
787 {
788 lnet_peer_t *lp = msg->msg_txpeer;
789 lnet_ni_t *ni = lp->lp_ni;
790 int cpt = msg->msg_tx_cpt;
791 struct lnet_tx_queue *tq = ni->ni_tx_queues[cpt];
792
793 /* non-lnet_send() callers have checked before */
794 LASSERT(!do_send || msg->msg_tx_delayed);
795 LASSERT(!msg->msg_receiving);
796 LASSERT(msg->msg_tx_committed);
797
798 /* NB 'lp' is always the next hop */
799 if ((msg->msg_target.pid & LNET_PID_USERFLAG) == 0 &&
800 lnet_peer_alive_locked(lp) == 0) {
801 the_lnet.ln_counters[cpt]->drop_count++;
802 the_lnet.ln_counters[cpt]->drop_length += msg->msg_len;
803 lnet_net_unlock(cpt);
804
805 CNETERR("Dropping message for %s: peer not alive\n",
806 libcfs_id2str(msg->msg_target));
807 if (do_send)
808 lnet_finalize(ni, msg, -EHOSTUNREACH);
809
810 lnet_net_lock(cpt);
811 return EHOSTUNREACH;
812 }
813
814 if (msg->msg_md != NULL &&
815 (msg->msg_md->md_flags & LNET_MD_FLAG_ABORTED) != 0) {
816 lnet_net_unlock(cpt);
817
818 CNETERR("Aborting message for %s: LNetM[DE]Unlink() already called on the MD/ME.\n",
819 libcfs_id2str(msg->msg_target));
820 if (do_send)
821 lnet_finalize(ni, msg, -ECANCELED);
822
823 lnet_net_lock(cpt);
824 return ECANCELED;
825 }
826
827 if (!msg->msg_peertxcredit) {
828 LASSERT((lp->lp_txcredits < 0) ==
829 !list_empty(&lp->lp_txq));
830
831 msg->msg_peertxcredit = 1;
832 lp->lp_txqnob += msg->msg_len + sizeof(lnet_hdr_t);
833 lp->lp_txcredits--;
834
835 if (lp->lp_txcredits < lp->lp_mintxcredits)
836 lp->lp_mintxcredits = lp->lp_txcredits;
837
838 if (lp->lp_txcredits < 0) {
839 msg->msg_tx_delayed = 1;
840 list_add_tail(&msg->msg_list, &lp->lp_txq);
841 return EAGAIN;
842 }
843 }
844
845 if (!msg->msg_txcredit) {
846 LASSERT((tq->tq_credits < 0) ==
847 !list_empty(&tq->tq_delayed));
848
849 msg->msg_txcredit = 1;
850 tq->tq_credits--;
851
852 if (tq->tq_credits < tq->tq_credits_min)
853 tq->tq_credits_min = tq->tq_credits;
854
855 if (tq->tq_credits < 0) {
856 msg->msg_tx_delayed = 1;
857 list_add_tail(&msg->msg_list, &tq->tq_delayed);
858 return EAGAIN;
859 }
860 }
861
862 if (do_send) {
863 lnet_net_unlock(cpt);
864 lnet_ni_send(ni, msg);
865 lnet_net_lock(cpt);
866 }
867 return 0;
868 }
869
870
871 static lnet_rtrbufpool_t *
lnet_msg2bufpool(lnet_msg_t * msg)872 lnet_msg2bufpool(lnet_msg_t *msg)
873 {
874 lnet_rtrbufpool_t *rbp;
875 int cpt;
876
877 LASSERT(msg->msg_rx_committed);
878
879 cpt = msg->msg_rx_cpt;
880 rbp = &the_lnet.ln_rtrpools[cpt][0];
881
882 LASSERT(msg->msg_len <= LNET_MTU);
883 while (msg->msg_len > (unsigned int)rbp->rbp_npages * PAGE_CACHE_SIZE) {
884 rbp++;
885 LASSERT(rbp < &the_lnet.ln_rtrpools[cpt][LNET_NRBPOOLS]);
886 }
887
888 return rbp;
889 }
890
891 static int
lnet_post_routed_recv_locked(lnet_msg_t * msg,int do_recv)892 lnet_post_routed_recv_locked(lnet_msg_t *msg, int do_recv)
893 {
894 /* lnet_parse is going to lnet_net_unlock immediately after this, so it
895 * sets do_recv FALSE and I don't do the unlock/send/lock bit. I
896 * return EAGAIN if msg blocked and 0 if received or OK to receive */
897 lnet_peer_t *lp = msg->msg_rxpeer;
898 lnet_rtrbufpool_t *rbp;
899 lnet_rtrbuf_t *rb;
900
901 LASSERT(msg->msg_iov == NULL);
902 LASSERT(msg->msg_kiov == NULL);
903 LASSERT(msg->msg_niov == 0);
904 LASSERT(msg->msg_routing);
905 LASSERT(msg->msg_receiving);
906 LASSERT(!msg->msg_sending);
907
908 /* non-lnet_parse callers only receive delayed messages */
909 LASSERT(!do_recv || msg->msg_rx_delayed);
910
911 if (!msg->msg_peerrtrcredit) {
912 LASSERT((lp->lp_rtrcredits < 0) ==
913 !list_empty(&lp->lp_rtrq));
914
915 msg->msg_peerrtrcredit = 1;
916 lp->lp_rtrcredits--;
917 if (lp->lp_rtrcredits < lp->lp_minrtrcredits)
918 lp->lp_minrtrcredits = lp->lp_rtrcredits;
919
920 if (lp->lp_rtrcredits < 0) {
921 /* must have checked eager_recv before here */
922 LASSERT(msg->msg_rx_ready_delay);
923 msg->msg_rx_delayed = 1;
924 list_add_tail(&msg->msg_list, &lp->lp_rtrq);
925 return EAGAIN;
926 }
927 }
928
929 rbp = lnet_msg2bufpool(msg);
930
931 if (!msg->msg_rtrcredit) {
932 LASSERT((rbp->rbp_credits < 0) ==
933 !list_empty(&rbp->rbp_msgs));
934
935 msg->msg_rtrcredit = 1;
936 rbp->rbp_credits--;
937 if (rbp->rbp_credits < rbp->rbp_mincredits)
938 rbp->rbp_mincredits = rbp->rbp_credits;
939
940 if (rbp->rbp_credits < 0) {
941 /* must have checked eager_recv before here */
942 LASSERT(msg->msg_rx_ready_delay);
943 msg->msg_rx_delayed = 1;
944 list_add_tail(&msg->msg_list, &rbp->rbp_msgs);
945 return EAGAIN;
946 }
947 }
948
949 LASSERT(!list_empty(&rbp->rbp_bufs));
950 rb = list_entry(rbp->rbp_bufs.next, lnet_rtrbuf_t, rb_list);
951 list_del(&rb->rb_list);
952
953 msg->msg_niov = rbp->rbp_npages;
954 msg->msg_kiov = &rb->rb_kiov[0];
955
956 if (do_recv) {
957 int cpt = msg->msg_rx_cpt;
958
959 lnet_net_unlock(cpt);
960 lnet_ni_recv(lp->lp_ni, msg->msg_private, msg, 1,
961 0, msg->msg_len, msg->msg_len);
962 lnet_net_lock(cpt);
963 }
964 return 0;
965 }
966
967 void
lnet_return_tx_credits_locked(lnet_msg_t * msg)968 lnet_return_tx_credits_locked(lnet_msg_t *msg)
969 {
970 lnet_peer_t *txpeer = msg->msg_txpeer;
971 lnet_msg_t *msg2;
972
973 if (msg->msg_txcredit) {
974 struct lnet_ni *ni = txpeer->lp_ni;
975 struct lnet_tx_queue *tq = ni->ni_tx_queues[msg->msg_tx_cpt];
976
977 /* give back NI txcredits */
978 msg->msg_txcredit = 0;
979
980 LASSERT((tq->tq_credits < 0) ==
981 !list_empty(&tq->tq_delayed));
982
983 tq->tq_credits++;
984 if (tq->tq_credits <= 0) {
985 msg2 = list_entry(tq->tq_delayed.next,
986 lnet_msg_t, msg_list);
987 list_del(&msg2->msg_list);
988
989 LASSERT(msg2->msg_txpeer->lp_ni == ni);
990 LASSERT(msg2->msg_tx_delayed);
991
992 (void) lnet_post_send_locked(msg2, 1);
993 }
994 }
995
996 if (msg->msg_peertxcredit) {
997 /* give back peer txcredits */
998 msg->msg_peertxcredit = 0;
999
1000 LASSERT((txpeer->lp_txcredits < 0) ==
1001 !list_empty(&txpeer->lp_txq));
1002
1003 txpeer->lp_txqnob -= msg->msg_len + sizeof(lnet_hdr_t);
1004 LASSERT(txpeer->lp_txqnob >= 0);
1005
1006 txpeer->lp_txcredits++;
1007 if (txpeer->lp_txcredits <= 0) {
1008 msg2 = list_entry(txpeer->lp_txq.next,
1009 lnet_msg_t, msg_list);
1010 list_del(&msg2->msg_list);
1011
1012 LASSERT(msg2->msg_txpeer == txpeer);
1013 LASSERT(msg2->msg_tx_delayed);
1014
1015 (void) lnet_post_send_locked(msg2, 1);
1016 }
1017 }
1018
1019 if (txpeer != NULL) {
1020 msg->msg_txpeer = NULL;
1021 lnet_peer_decref_locked(txpeer);
1022 }
1023 }
1024
1025 void
lnet_return_rx_credits_locked(lnet_msg_t * msg)1026 lnet_return_rx_credits_locked(lnet_msg_t *msg)
1027 {
1028 lnet_peer_t *rxpeer = msg->msg_rxpeer;
1029 lnet_msg_t *msg2;
1030
1031 if (msg->msg_rtrcredit) {
1032 /* give back global router credits */
1033 lnet_rtrbuf_t *rb;
1034 lnet_rtrbufpool_t *rbp;
1035
1036 /* NB If a msg ever blocks for a buffer in rbp_msgs, it stays
1037 * there until it gets one allocated, or aborts the wait
1038 * itself */
1039 LASSERT(msg->msg_kiov != NULL);
1040
1041 rb = list_entry(msg->msg_kiov, lnet_rtrbuf_t, rb_kiov[0]);
1042 rbp = rb->rb_pool;
1043 LASSERT(rbp == lnet_msg2bufpool(msg));
1044
1045 msg->msg_kiov = NULL;
1046 msg->msg_rtrcredit = 0;
1047
1048 LASSERT((rbp->rbp_credits < 0) ==
1049 !list_empty(&rbp->rbp_msgs));
1050 LASSERT((rbp->rbp_credits > 0) ==
1051 !list_empty(&rbp->rbp_bufs));
1052
1053 list_add(&rb->rb_list, &rbp->rbp_bufs);
1054 rbp->rbp_credits++;
1055 if (rbp->rbp_credits <= 0) {
1056 msg2 = list_entry(rbp->rbp_msgs.next,
1057 lnet_msg_t, msg_list);
1058 list_del(&msg2->msg_list);
1059
1060 (void) lnet_post_routed_recv_locked(msg2, 1);
1061 }
1062 }
1063
1064 if (msg->msg_peerrtrcredit) {
1065 /* give back peer router credits */
1066 msg->msg_peerrtrcredit = 0;
1067
1068 LASSERT((rxpeer->lp_rtrcredits < 0) ==
1069 !list_empty(&rxpeer->lp_rtrq));
1070
1071 rxpeer->lp_rtrcredits++;
1072 if (rxpeer->lp_rtrcredits <= 0) {
1073 msg2 = list_entry(rxpeer->lp_rtrq.next,
1074 lnet_msg_t, msg_list);
1075 list_del(&msg2->msg_list);
1076
1077 (void) lnet_post_routed_recv_locked(msg2, 1);
1078 }
1079 }
1080 if (rxpeer != NULL) {
1081 msg->msg_rxpeer = NULL;
1082 lnet_peer_decref_locked(rxpeer);
1083 }
1084 }
1085
1086 static int
lnet_compare_routes(lnet_route_t * r1,lnet_route_t * r2)1087 lnet_compare_routes(lnet_route_t *r1, lnet_route_t *r2)
1088 {
1089 lnet_peer_t *p1 = r1->lr_gateway;
1090 lnet_peer_t *p2 = r2->lr_gateway;
1091
1092 if (r1->lr_priority < r2->lr_priority)
1093 return 1;
1094
1095 if (r1->lr_priority > r2->lr_priority)
1096 return -1;
1097
1098 if (r1->lr_hops < r2->lr_hops)
1099 return 1;
1100
1101 if (r1->lr_hops > r2->lr_hops)
1102 return -1;
1103
1104 if (p1->lp_txqnob < p2->lp_txqnob)
1105 return 1;
1106
1107 if (p1->lp_txqnob > p2->lp_txqnob)
1108 return -1;
1109
1110 if (p1->lp_txcredits > p2->lp_txcredits)
1111 return 1;
1112
1113 if (p1->lp_txcredits < p2->lp_txcredits)
1114 return -1;
1115
1116 if (r1->lr_seq - r2->lr_seq <= 0)
1117 return 1;
1118
1119 return -1;
1120 }
1121
1122 static lnet_peer_t *
lnet_find_route_locked(lnet_ni_t * ni,lnet_nid_t target,lnet_nid_t rtr_nid)1123 lnet_find_route_locked(lnet_ni_t *ni, lnet_nid_t target, lnet_nid_t rtr_nid)
1124 {
1125 lnet_remotenet_t *rnet;
1126 lnet_route_t *rtr;
1127 lnet_route_t *rtr_best;
1128 lnet_route_t *rtr_last;
1129 struct lnet_peer *lp_best;
1130 struct lnet_peer *lp;
1131 int rc;
1132
1133 /* If @rtr_nid is not LNET_NID_ANY, return the gateway with
1134 * rtr_nid nid, otherwise find the best gateway I can use */
1135
1136 rnet = lnet_find_net_locked(LNET_NIDNET(target));
1137 if (rnet == NULL)
1138 return NULL;
1139
1140 lp_best = NULL;
1141 rtr_best = rtr_last = NULL;
1142 list_for_each_entry(rtr, &rnet->lrn_routes, lr_list) {
1143 lp = rtr->lr_gateway;
1144
1145 if (!lp->lp_alive || /* gateway is down */
1146 ((lp->lp_ping_feats & LNET_PING_FEAT_NI_STATUS) != 0 &&
1147 rtr->lr_downis != 0)) /* NI to target is down */
1148 continue;
1149
1150 if (ni != NULL && lp->lp_ni != ni)
1151 continue;
1152
1153 if (lp->lp_nid == rtr_nid) /* it's pre-determined router */
1154 return lp;
1155
1156 if (lp_best == NULL) {
1157 rtr_best = rtr_last = rtr;
1158 lp_best = lp;
1159 continue;
1160 }
1161
1162 /* no protection on below fields, but it's harmless */
1163 if (rtr_last->lr_seq - rtr->lr_seq < 0)
1164 rtr_last = rtr;
1165
1166 rc = lnet_compare_routes(rtr, rtr_best);
1167 if (rc < 0)
1168 continue;
1169
1170 rtr_best = rtr;
1171 lp_best = lp;
1172 }
1173
1174 /* set sequence number on the best router to the latest sequence + 1
1175 * so we can round-robin all routers, it's race and inaccurate but
1176 * harmless and functional */
1177 if (rtr_best != NULL)
1178 rtr_best->lr_seq = rtr_last->lr_seq + 1;
1179 return lp_best;
1180 }
1181
1182 int
lnet_send(lnet_nid_t src_nid,lnet_msg_t * msg,lnet_nid_t rtr_nid)1183 lnet_send(lnet_nid_t src_nid, lnet_msg_t *msg, lnet_nid_t rtr_nid)
1184 {
1185 lnet_nid_t dst_nid = msg->msg_target.nid;
1186 struct lnet_ni *src_ni;
1187 struct lnet_ni *local_ni;
1188 struct lnet_peer *lp;
1189 int cpt;
1190 int cpt2;
1191 int rc;
1192
1193 /* NB: rtr_nid is set to LNET_NID_ANY for all current use-cases,
1194 * but we might want to use pre-determined router for ACK/REPLY
1195 * in the future */
1196 /* NB: ni != NULL == interface pre-determined (ACK/REPLY) */
1197 LASSERT(msg->msg_txpeer == NULL);
1198 LASSERT(!msg->msg_sending);
1199 LASSERT(!msg->msg_target_is_router);
1200 LASSERT(!msg->msg_receiving);
1201
1202 msg->msg_sending = 1;
1203
1204 LASSERT(!msg->msg_tx_committed);
1205 cpt = lnet_cpt_of_nid(rtr_nid == LNET_NID_ANY ? dst_nid : rtr_nid);
1206 again:
1207 lnet_net_lock(cpt);
1208
1209 if (the_lnet.ln_shutdown) {
1210 lnet_net_unlock(cpt);
1211 return -ESHUTDOWN;
1212 }
1213
1214 if (src_nid == LNET_NID_ANY) {
1215 src_ni = NULL;
1216 } else {
1217 src_ni = lnet_nid2ni_locked(src_nid, cpt);
1218 if (src_ni == NULL) {
1219 lnet_net_unlock(cpt);
1220 LCONSOLE_WARN("Can't send to %s: src %s is not a local nid\n",
1221 libcfs_nid2str(dst_nid),
1222 libcfs_nid2str(src_nid));
1223 return -EINVAL;
1224 }
1225 LASSERT(!msg->msg_routing);
1226 }
1227
1228 /* Is this for someone on a local network? */
1229 local_ni = lnet_net2ni_locked(LNET_NIDNET(dst_nid), cpt);
1230
1231 if (local_ni != NULL) {
1232 if (src_ni == NULL) {
1233 src_ni = local_ni;
1234 src_nid = src_ni->ni_nid;
1235 } else if (src_ni == local_ni) {
1236 lnet_ni_decref_locked(local_ni, cpt);
1237 } else {
1238 lnet_ni_decref_locked(local_ni, cpt);
1239 lnet_ni_decref_locked(src_ni, cpt);
1240 lnet_net_unlock(cpt);
1241 LCONSOLE_WARN("No route to %s via from %s\n",
1242 libcfs_nid2str(dst_nid),
1243 libcfs_nid2str(src_nid));
1244 return -EINVAL;
1245 }
1246
1247 LASSERT(src_nid != LNET_NID_ANY);
1248 lnet_msg_commit(msg, cpt);
1249
1250 if (!msg->msg_routing)
1251 msg->msg_hdr.src_nid = cpu_to_le64(src_nid);
1252
1253 if (src_ni == the_lnet.ln_loni) {
1254 /* No send credit hassles with LOLND */
1255 lnet_net_unlock(cpt);
1256 lnet_ni_send(src_ni, msg);
1257
1258 lnet_net_lock(cpt);
1259 lnet_ni_decref_locked(src_ni, cpt);
1260 lnet_net_unlock(cpt);
1261 return 0;
1262 }
1263
1264 rc = lnet_nid2peer_locked(&lp, dst_nid, cpt);
1265 /* lp has ref on src_ni; lose mine */
1266 lnet_ni_decref_locked(src_ni, cpt);
1267 if (rc != 0) {
1268 lnet_net_unlock(cpt);
1269 LCONSOLE_WARN("Error %d finding peer %s\n", rc,
1270 libcfs_nid2str(dst_nid));
1271 /* ENOMEM or shutting down */
1272 return rc;
1273 }
1274 LASSERT(lp->lp_ni == src_ni);
1275 } else {
1276 /* sending to a remote network */
1277 lp = lnet_find_route_locked(src_ni, dst_nid, rtr_nid);
1278 if (lp == NULL) {
1279 if (src_ni != NULL)
1280 lnet_ni_decref_locked(src_ni, cpt);
1281 lnet_net_unlock(cpt);
1282
1283 LCONSOLE_WARN("No route to %s via %s (all routers down)\n",
1284 libcfs_id2str(msg->msg_target),
1285 libcfs_nid2str(src_nid));
1286 return -EHOSTUNREACH;
1287 }
1288
1289 /* rtr_nid is LNET_NID_ANY or NID of pre-determined router,
1290 * it's possible that rtr_nid isn't LNET_NID_ANY and lp isn't
1291 * pre-determined router, this can happen if router table
1292 * was changed when we release the lock */
1293 if (rtr_nid != lp->lp_nid) {
1294 cpt2 = lnet_cpt_of_nid_locked(lp->lp_nid);
1295 if (cpt2 != cpt) {
1296 if (src_ni != NULL)
1297 lnet_ni_decref_locked(src_ni, cpt);
1298 lnet_net_unlock(cpt);
1299
1300 rtr_nid = lp->lp_nid;
1301 cpt = cpt2;
1302 goto again;
1303 }
1304 }
1305
1306 CDEBUG(D_NET, "Best route to %s via %s for %s %d\n",
1307 libcfs_nid2str(dst_nid), libcfs_nid2str(lp->lp_nid),
1308 lnet_msgtyp2str(msg->msg_type), msg->msg_len);
1309
1310 if (src_ni == NULL) {
1311 src_ni = lp->lp_ni;
1312 src_nid = src_ni->ni_nid;
1313 } else {
1314 LASSERT(src_ni == lp->lp_ni);
1315 lnet_ni_decref_locked(src_ni, cpt);
1316 }
1317
1318 lnet_peer_addref_locked(lp);
1319
1320 LASSERT(src_nid != LNET_NID_ANY);
1321 lnet_msg_commit(msg, cpt);
1322
1323 if (!msg->msg_routing) {
1324 /* I'm the source and now I know which NI to send on */
1325 msg->msg_hdr.src_nid = cpu_to_le64(src_nid);
1326 }
1327
1328 msg->msg_target_is_router = 1;
1329 msg->msg_target.nid = lp->lp_nid;
1330 msg->msg_target.pid = LUSTRE_SRV_LNET_PID;
1331 }
1332
1333 /* 'lp' is our best choice of peer */
1334
1335 LASSERT(!msg->msg_peertxcredit);
1336 LASSERT(!msg->msg_txcredit);
1337 LASSERT(msg->msg_txpeer == NULL);
1338
1339 msg->msg_txpeer = lp; /* msg takes my ref on lp */
1340
1341 rc = lnet_post_send_locked(msg, 0);
1342 lnet_net_unlock(cpt);
1343
1344 if (rc == EHOSTUNREACH || rc == ECANCELED)
1345 return -rc;
1346
1347 if (rc == 0)
1348 lnet_ni_send(src_ni, msg);
1349
1350 return 0; /* rc == 0 or EAGAIN */
1351 }
1352
1353 static void
lnet_drop_message(lnet_ni_t * ni,int cpt,void * private,unsigned int nob)1354 lnet_drop_message(lnet_ni_t *ni, int cpt, void *private, unsigned int nob)
1355 {
1356 lnet_net_lock(cpt);
1357 the_lnet.ln_counters[cpt]->drop_count++;
1358 the_lnet.ln_counters[cpt]->drop_length += nob;
1359 lnet_net_unlock(cpt);
1360
1361 lnet_ni_recv(ni, private, NULL, 0, 0, 0, nob);
1362 }
1363
1364 static void
lnet_recv_put(lnet_ni_t * ni,lnet_msg_t * msg)1365 lnet_recv_put(lnet_ni_t *ni, lnet_msg_t *msg)
1366 {
1367 lnet_hdr_t *hdr = &msg->msg_hdr;
1368
1369 if (msg->msg_wanted != 0)
1370 lnet_setpayloadbuffer(msg);
1371
1372 lnet_build_msg_event(msg, LNET_EVENT_PUT);
1373
1374 /* Must I ACK? If so I'll grab the ack_wmd out of the header and put
1375 * it back into the ACK during lnet_finalize() */
1376 msg->msg_ack = (!lnet_is_wire_handle_none(&hdr->msg.put.ack_wmd) &&
1377 (msg->msg_md->md_options & LNET_MD_ACK_DISABLE) == 0);
1378
1379 lnet_ni_recv(ni, msg->msg_private, msg, msg->msg_rx_delayed,
1380 msg->msg_offset, msg->msg_wanted, hdr->payload_length);
1381 }
1382
1383 static int
lnet_parse_put(lnet_ni_t * ni,lnet_msg_t * msg)1384 lnet_parse_put(lnet_ni_t *ni, lnet_msg_t *msg)
1385 {
1386 lnet_hdr_t *hdr = &msg->msg_hdr;
1387 struct lnet_match_info info;
1388 int rc;
1389
1390 /* Convert put fields to host byte order */
1391 hdr->msg.put.match_bits = le64_to_cpu(hdr->msg.put.match_bits);
1392 hdr->msg.put.ptl_index = le32_to_cpu(hdr->msg.put.ptl_index);
1393 hdr->msg.put.offset = le32_to_cpu(hdr->msg.put.offset);
1394
1395 info.mi_id.nid = hdr->src_nid;
1396 info.mi_id.pid = hdr->src_pid;
1397 info.mi_opc = LNET_MD_OP_PUT;
1398 info.mi_portal = hdr->msg.put.ptl_index;
1399 info.mi_rlength = hdr->payload_length;
1400 info.mi_roffset = hdr->msg.put.offset;
1401 info.mi_mbits = hdr->msg.put.match_bits;
1402
1403 msg->msg_rx_ready_delay = ni->ni_lnd->lnd_eager_recv == NULL;
1404
1405 again:
1406 rc = lnet_ptl_match_md(&info, msg);
1407 switch (rc) {
1408 default:
1409 LBUG();
1410
1411 case LNET_MATCHMD_OK:
1412 lnet_recv_put(ni, msg);
1413 return 0;
1414
1415 case LNET_MATCHMD_NONE:
1416 if (msg->msg_rx_delayed) /* attached on delayed list */
1417 return 0;
1418
1419 rc = lnet_ni_eager_recv(ni, msg);
1420 if (rc == 0)
1421 goto again;
1422 /* fall through */
1423
1424 case LNET_MATCHMD_DROP:
1425 CNETERR("Dropping PUT from %s portal %d match %llu offset %d length %d: %d\n",
1426 libcfs_id2str(info.mi_id), info.mi_portal,
1427 info.mi_mbits, info.mi_roffset, info.mi_rlength, rc);
1428
1429 return ENOENT; /* +ve: OK but no match */
1430 }
1431 }
1432
1433 static int
lnet_parse_get(lnet_ni_t * ni,lnet_msg_t * msg,int rdma_get)1434 lnet_parse_get(lnet_ni_t *ni, lnet_msg_t *msg, int rdma_get)
1435 {
1436 struct lnet_match_info info;
1437 lnet_hdr_t *hdr = &msg->msg_hdr;
1438 lnet_handle_wire_t reply_wmd;
1439 int rc;
1440
1441 /* Convert get fields to host byte order */
1442 hdr->msg.get.match_bits = le64_to_cpu(hdr->msg.get.match_bits);
1443 hdr->msg.get.ptl_index = le32_to_cpu(hdr->msg.get.ptl_index);
1444 hdr->msg.get.sink_length = le32_to_cpu(hdr->msg.get.sink_length);
1445 hdr->msg.get.src_offset = le32_to_cpu(hdr->msg.get.src_offset);
1446
1447 info.mi_id.nid = hdr->src_nid;
1448 info.mi_id.pid = hdr->src_pid;
1449 info.mi_opc = LNET_MD_OP_GET;
1450 info.mi_portal = hdr->msg.get.ptl_index;
1451 info.mi_rlength = hdr->msg.get.sink_length;
1452 info.mi_roffset = hdr->msg.get.src_offset;
1453 info.mi_mbits = hdr->msg.get.match_bits;
1454
1455 rc = lnet_ptl_match_md(&info, msg);
1456 if (rc == LNET_MATCHMD_DROP) {
1457 CNETERR("Dropping GET from %s portal %d match %llu offset %d length %d\n",
1458 libcfs_id2str(info.mi_id), info.mi_portal,
1459 info.mi_mbits, info.mi_roffset, info.mi_rlength);
1460 return ENOENT; /* +ve: OK but no match */
1461 }
1462
1463 LASSERT(rc == LNET_MATCHMD_OK);
1464
1465 lnet_build_msg_event(msg, LNET_EVENT_GET);
1466
1467 reply_wmd = hdr->msg.get.return_wmd;
1468
1469 lnet_prep_send(msg, LNET_MSG_REPLY, info.mi_id,
1470 msg->msg_offset, msg->msg_wanted);
1471
1472 msg->msg_hdr.msg.reply.dst_wmd = reply_wmd;
1473
1474 if (rdma_get) {
1475 /* The LND completes the REPLY from her recv procedure */
1476 lnet_ni_recv(ni, msg->msg_private, msg, 0,
1477 msg->msg_offset, msg->msg_len, msg->msg_len);
1478 return 0;
1479 }
1480
1481 lnet_ni_recv(ni, msg->msg_private, NULL, 0, 0, 0, 0);
1482 msg->msg_receiving = 0;
1483
1484 rc = lnet_send(ni->ni_nid, msg, LNET_NID_ANY);
1485 if (rc < 0) {
1486 /* didn't get as far as lnet_ni_send() */
1487 CERROR("%s: Unable to send REPLY for GET from %s: %d\n",
1488 libcfs_nid2str(ni->ni_nid),
1489 libcfs_id2str(info.mi_id), rc);
1490
1491 lnet_finalize(ni, msg, rc);
1492 }
1493
1494 return 0;
1495 }
1496
1497 static int
lnet_parse_reply(lnet_ni_t * ni,lnet_msg_t * msg)1498 lnet_parse_reply(lnet_ni_t *ni, lnet_msg_t *msg)
1499 {
1500 void *private = msg->msg_private;
1501 lnet_hdr_t *hdr = &msg->msg_hdr;
1502 lnet_process_id_t src = {0};
1503 lnet_libmd_t *md;
1504 int rlength;
1505 int mlength;
1506 int cpt;
1507
1508 cpt = lnet_cpt_of_cookie(hdr->msg.reply.dst_wmd.wh_object_cookie);
1509 lnet_res_lock(cpt);
1510
1511 src.nid = hdr->src_nid;
1512 src.pid = hdr->src_pid;
1513
1514 /* NB handles only looked up by creator (no flips) */
1515 md = lnet_wire_handle2md(&hdr->msg.reply.dst_wmd);
1516 if (md == NULL || md->md_threshold == 0 || md->md_me != NULL) {
1517 CNETERR("%s: Dropping REPLY from %s for %s MD %#llx.%#llx\n",
1518 libcfs_nid2str(ni->ni_nid), libcfs_id2str(src),
1519 (md == NULL) ? "invalid" : "inactive",
1520 hdr->msg.reply.dst_wmd.wh_interface_cookie,
1521 hdr->msg.reply.dst_wmd.wh_object_cookie);
1522 if (md != NULL && md->md_me != NULL)
1523 CERROR("REPLY MD also attached to portal %d\n",
1524 md->md_me->me_portal);
1525
1526 lnet_res_unlock(cpt);
1527 return ENOENT; /* +ve: OK but no match */
1528 }
1529
1530 LASSERT(md->md_offset == 0);
1531
1532 rlength = hdr->payload_length;
1533 mlength = min_t(uint, rlength, md->md_length);
1534
1535 if (mlength < rlength &&
1536 (md->md_options & LNET_MD_TRUNCATE) == 0) {
1537 CNETERR("%s: Dropping REPLY from %s length %d for MD %#llx would overflow (%d)\n",
1538 libcfs_nid2str(ni->ni_nid), libcfs_id2str(src),
1539 rlength, hdr->msg.reply.dst_wmd.wh_object_cookie,
1540 mlength);
1541 lnet_res_unlock(cpt);
1542 return ENOENT; /* +ve: OK but no match */
1543 }
1544
1545 CDEBUG(D_NET, "%s: Reply from %s of length %d/%d into md %#llx\n",
1546 libcfs_nid2str(ni->ni_nid), libcfs_id2str(src),
1547 mlength, rlength, hdr->msg.reply.dst_wmd.wh_object_cookie);
1548
1549 lnet_msg_attach_md(msg, md, 0, mlength);
1550
1551 if (mlength != 0)
1552 lnet_setpayloadbuffer(msg);
1553
1554 lnet_res_unlock(cpt);
1555
1556 lnet_build_msg_event(msg, LNET_EVENT_REPLY);
1557
1558 lnet_ni_recv(ni, private, msg, 0, 0, mlength, rlength);
1559 return 0;
1560 }
1561
1562 static int
lnet_parse_ack(lnet_ni_t * ni,lnet_msg_t * msg)1563 lnet_parse_ack(lnet_ni_t *ni, lnet_msg_t *msg)
1564 {
1565 lnet_hdr_t *hdr = &msg->msg_hdr;
1566 lnet_process_id_t src = {0};
1567 lnet_libmd_t *md;
1568 int cpt;
1569
1570 src.nid = hdr->src_nid;
1571 src.pid = hdr->src_pid;
1572
1573 /* Convert ack fields to host byte order */
1574 hdr->msg.ack.match_bits = le64_to_cpu(hdr->msg.ack.match_bits);
1575 hdr->msg.ack.mlength = le32_to_cpu(hdr->msg.ack.mlength);
1576
1577 cpt = lnet_cpt_of_cookie(hdr->msg.ack.dst_wmd.wh_object_cookie);
1578 lnet_res_lock(cpt);
1579
1580 /* NB handles only looked up by creator (no flips) */
1581 md = lnet_wire_handle2md(&hdr->msg.ack.dst_wmd);
1582 if (md == NULL || md->md_threshold == 0 || md->md_me != NULL) {
1583 /* Don't moan; this is expected */
1584 CDEBUG(D_NET,
1585 "%s: Dropping ACK from %s to %s MD %#llx.%#llx\n",
1586 libcfs_nid2str(ni->ni_nid), libcfs_id2str(src),
1587 (md == NULL) ? "invalid" : "inactive",
1588 hdr->msg.ack.dst_wmd.wh_interface_cookie,
1589 hdr->msg.ack.dst_wmd.wh_object_cookie);
1590 if (md != NULL && md->md_me != NULL)
1591 CERROR("Source MD also attached to portal %d\n",
1592 md->md_me->me_portal);
1593
1594 lnet_res_unlock(cpt);
1595 return ENOENT; /* +ve! */
1596 }
1597
1598 CDEBUG(D_NET, "%s: ACK from %s into md %#llx\n",
1599 libcfs_nid2str(ni->ni_nid), libcfs_id2str(src),
1600 hdr->msg.ack.dst_wmd.wh_object_cookie);
1601
1602 lnet_msg_attach_md(msg, md, 0, 0);
1603
1604 lnet_res_unlock(cpt);
1605
1606 lnet_build_msg_event(msg, LNET_EVENT_ACK);
1607
1608 lnet_ni_recv(ni, msg->msg_private, msg, 0, 0, 0, msg->msg_len);
1609 return 0;
1610 }
1611
1612 static int
lnet_parse_forward_locked(lnet_ni_t * ni,lnet_msg_t * msg)1613 lnet_parse_forward_locked(lnet_ni_t *ni, lnet_msg_t *msg)
1614 {
1615 int rc = 0;
1616
1617 if (msg->msg_rxpeer->lp_rtrcredits <= 0 ||
1618 lnet_msg2bufpool(msg)->rbp_credits <= 0) {
1619 if (ni->ni_lnd->lnd_eager_recv == NULL) {
1620 msg->msg_rx_ready_delay = 1;
1621 } else {
1622 lnet_net_unlock(msg->msg_rx_cpt);
1623 rc = lnet_ni_eager_recv(ni, msg);
1624 lnet_net_lock(msg->msg_rx_cpt);
1625 }
1626 }
1627
1628 if (rc == 0)
1629 rc = lnet_post_routed_recv_locked(msg, 0);
1630 return rc;
1631 }
1632
1633 char *
lnet_msgtyp2str(int type)1634 lnet_msgtyp2str(int type)
1635 {
1636 switch (type) {
1637 case LNET_MSG_ACK:
1638 return "ACK";
1639 case LNET_MSG_PUT:
1640 return "PUT";
1641 case LNET_MSG_GET:
1642 return "GET";
1643 case LNET_MSG_REPLY:
1644 return "REPLY";
1645 case LNET_MSG_HELLO:
1646 return "HELLO";
1647 default:
1648 return "<UNKNOWN>";
1649 }
1650 }
1651 EXPORT_SYMBOL(lnet_msgtyp2str);
1652
1653 void
lnet_print_hdr(lnet_hdr_t * hdr)1654 lnet_print_hdr(lnet_hdr_t *hdr)
1655 {
1656 lnet_process_id_t src = {0};
1657 lnet_process_id_t dst = {0};
1658 char *type_str = lnet_msgtyp2str(hdr->type);
1659
1660 src.nid = hdr->src_nid;
1661 src.pid = hdr->src_pid;
1662
1663 dst.nid = hdr->dest_nid;
1664 dst.pid = hdr->dest_pid;
1665
1666 CWARN("P3 Header at %p of type %s\n", hdr, type_str);
1667 CWARN(" From %s\n", libcfs_id2str(src));
1668 CWARN(" To %s\n", libcfs_id2str(dst));
1669
1670 switch (hdr->type) {
1671 default:
1672 break;
1673
1674 case LNET_MSG_PUT:
1675 CWARN(" Ptl index %d, ack md %#llx.%#llx, match bits %llu\n",
1676 hdr->msg.put.ptl_index,
1677 hdr->msg.put.ack_wmd.wh_interface_cookie,
1678 hdr->msg.put.ack_wmd.wh_object_cookie,
1679 hdr->msg.put.match_bits);
1680 CWARN(" Length %d, offset %d, hdr data %#llx\n",
1681 hdr->payload_length, hdr->msg.put.offset,
1682 hdr->msg.put.hdr_data);
1683 break;
1684
1685 case LNET_MSG_GET:
1686 CWARN(" Ptl index %d, return md %#llx.%#llx, match bits %llu\n",
1687 hdr->msg.get.ptl_index,
1688 hdr->msg.get.return_wmd.wh_interface_cookie,
1689 hdr->msg.get.return_wmd.wh_object_cookie,
1690 hdr->msg.get.match_bits);
1691 CWARN(" Length %d, src offset %d\n",
1692 hdr->msg.get.sink_length,
1693 hdr->msg.get.src_offset);
1694 break;
1695
1696 case LNET_MSG_ACK:
1697 CWARN(" dst md %#llx.%#llx, manipulated length %d\n",
1698 hdr->msg.ack.dst_wmd.wh_interface_cookie,
1699 hdr->msg.ack.dst_wmd.wh_object_cookie,
1700 hdr->msg.ack.mlength);
1701 break;
1702
1703 case LNET_MSG_REPLY:
1704 CWARN(" dst md %#llx.%#llx, length %d\n",
1705 hdr->msg.reply.dst_wmd.wh_interface_cookie,
1706 hdr->msg.reply.dst_wmd.wh_object_cookie,
1707 hdr->payload_length);
1708 }
1709
1710 }
1711
1712 int
lnet_parse(lnet_ni_t * ni,lnet_hdr_t * hdr,lnet_nid_t from_nid,void * private,int rdma_req)1713 lnet_parse(lnet_ni_t *ni, lnet_hdr_t *hdr, lnet_nid_t from_nid,
1714 void *private, int rdma_req)
1715 {
1716 int rc = 0;
1717 int cpt;
1718 int for_me;
1719 struct lnet_msg *msg;
1720 lnet_pid_t dest_pid;
1721 lnet_nid_t dest_nid;
1722 lnet_nid_t src_nid;
1723 __u32 payload_length;
1724 __u32 type;
1725
1726 LASSERT(!in_interrupt());
1727
1728 type = le32_to_cpu(hdr->type);
1729 src_nid = le64_to_cpu(hdr->src_nid);
1730 dest_nid = le64_to_cpu(hdr->dest_nid);
1731 dest_pid = le32_to_cpu(hdr->dest_pid);
1732 payload_length = le32_to_cpu(hdr->payload_length);
1733
1734 for_me = (ni->ni_nid == dest_nid);
1735 cpt = lnet_cpt_of_nid(from_nid);
1736
1737 switch (type) {
1738 case LNET_MSG_ACK:
1739 case LNET_MSG_GET:
1740 if (payload_length > 0) {
1741 CERROR("%s, src %s: bad %s payload %d (0 expected)\n",
1742 libcfs_nid2str(from_nid),
1743 libcfs_nid2str(src_nid),
1744 lnet_msgtyp2str(type), payload_length);
1745 return -EPROTO;
1746 }
1747 break;
1748
1749 case LNET_MSG_PUT:
1750 case LNET_MSG_REPLY:
1751 if (payload_length >
1752 (__u32)(for_me ? LNET_MAX_PAYLOAD : LNET_MTU)) {
1753 CERROR("%s, src %s: bad %s payload %d (%d max expected)\n",
1754 libcfs_nid2str(from_nid),
1755 libcfs_nid2str(src_nid),
1756 lnet_msgtyp2str(type),
1757 payload_length,
1758 for_me ? LNET_MAX_PAYLOAD : LNET_MTU);
1759 return -EPROTO;
1760 }
1761 break;
1762
1763 default:
1764 CERROR("%s, src %s: Bad message type 0x%x\n",
1765 libcfs_nid2str(from_nid),
1766 libcfs_nid2str(src_nid), type);
1767 return -EPROTO;
1768 }
1769
1770 if (the_lnet.ln_routing &&
1771 ni->ni_last_alive != get_seconds()) {
1772 lnet_ni_lock(ni);
1773
1774 /* NB: so far here is the only place to set NI status to "up */
1775 ni->ni_last_alive = get_seconds();
1776 if (ni->ni_status != NULL &&
1777 ni->ni_status->ns_status == LNET_NI_STATUS_DOWN)
1778 ni->ni_status->ns_status = LNET_NI_STATUS_UP;
1779 lnet_ni_unlock(ni);
1780 }
1781
1782 /* Regard a bad destination NID as a protocol error. Senders should
1783 * know what they're doing; if they don't they're misconfigured, buggy
1784 * or malicious so we chop them off at the knees :) */
1785
1786 if (!for_me) {
1787 if (LNET_NIDNET(dest_nid) == LNET_NIDNET(ni->ni_nid)) {
1788 /* should have gone direct */
1789 CERROR("%s, src %s: Bad dest nid %s (should have been sent direct)\n",
1790 libcfs_nid2str(from_nid),
1791 libcfs_nid2str(src_nid),
1792 libcfs_nid2str(dest_nid));
1793 return -EPROTO;
1794 }
1795
1796 if (lnet_islocalnid(dest_nid)) {
1797 /* dest is another local NI; sender should have used
1798 * this node's NID on its own network */
1799 CERROR("%s, src %s: Bad dest nid %s (it's my nid but on a different network)\n",
1800 libcfs_nid2str(from_nid),
1801 libcfs_nid2str(src_nid),
1802 libcfs_nid2str(dest_nid));
1803 return -EPROTO;
1804 }
1805
1806 if (rdma_req && type == LNET_MSG_GET) {
1807 CERROR("%s, src %s: Bad optimized GET for %s (final destination must be me)\n",
1808 libcfs_nid2str(from_nid),
1809 libcfs_nid2str(src_nid),
1810 libcfs_nid2str(dest_nid));
1811 return -EPROTO;
1812 }
1813
1814 if (!the_lnet.ln_routing) {
1815 CERROR("%s, src %s: Dropping message for %s (routing not enabled)\n",
1816 libcfs_nid2str(from_nid),
1817 libcfs_nid2str(src_nid),
1818 libcfs_nid2str(dest_nid));
1819 goto drop;
1820 }
1821 }
1822
1823 /* Message looks OK; we're not going to return an error, so we MUST
1824 * call back lnd_recv() come what may... */
1825
1826 if (!list_empty(&the_lnet.ln_test_peers) && /* normally we don't */
1827 fail_peer(src_nid, 0)) { /* shall we now? */
1828 CERROR("%s, src %s: Dropping %s to simulate failure\n",
1829 libcfs_nid2str(from_nid), libcfs_nid2str(src_nid),
1830 lnet_msgtyp2str(type));
1831 goto drop;
1832 }
1833
1834 msg = lnet_msg_alloc();
1835 if (msg == NULL) {
1836 CERROR("%s, src %s: Dropping %s (out of memory)\n",
1837 libcfs_nid2str(from_nid), libcfs_nid2str(src_nid),
1838 lnet_msgtyp2str(type));
1839 goto drop;
1840 }
1841
1842 /* msg zeroed in lnet_msg_alloc;
1843 * i.e. flags all clear, pointers NULL etc
1844 */
1845
1846 msg->msg_type = type;
1847 msg->msg_private = private;
1848 msg->msg_receiving = 1;
1849 msg->msg_len = msg->msg_wanted = payload_length;
1850 msg->msg_offset = 0;
1851 msg->msg_hdr = *hdr;
1852 /* for building message event */
1853 msg->msg_from = from_nid;
1854 if (!for_me) {
1855 msg->msg_target.pid = dest_pid;
1856 msg->msg_target.nid = dest_nid;
1857 msg->msg_routing = 1;
1858
1859 } else {
1860 /* convert common msg->hdr fields to host byteorder */
1861 msg->msg_hdr.type = type;
1862 msg->msg_hdr.src_nid = src_nid;
1863 msg->msg_hdr.src_pid = le32_to_cpu(msg->msg_hdr.src_pid);
1864 msg->msg_hdr.dest_nid = dest_nid;
1865 msg->msg_hdr.dest_pid = dest_pid;
1866 msg->msg_hdr.payload_length = payload_length;
1867 }
1868
1869 lnet_net_lock(cpt);
1870 rc = lnet_nid2peer_locked(&msg->msg_rxpeer, from_nid, cpt);
1871 if (rc != 0) {
1872 lnet_net_unlock(cpt);
1873 CERROR("%s, src %s: Dropping %s (error %d looking up sender)\n",
1874 libcfs_nid2str(from_nid), libcfs_nid2str(src_nid),
1875 lnet_msgtyp2str(type), rc);
1876 lnet_msg_free(msg);
1877 goto drop;
1878 }
1879
1880 if (lnet_isrouter(msg->msg_rxpeer)) {
1881 lnet_peer_set_alive(msg->msg_rxpeer);
1882 if (avoid_asym_router_failure &&
1883 LNET_NIDNET(src_nid) != LNET_NIDNET(from_nid)) {
1884 /* received a remote message from router, update
1885 * remote NI status on this router.
1886 * NB: multi-hop routed message will be ignored.
1887 */
1888 lnet_router_ni_update_locked(msg->msg_rxpeer,
1889 LNET_NIDNET(src_nid));
1890 }
1891 }
1892
1893 lnet_msg_commit(msg, cpt);
1894
1895 if (!for_me) {
1896 rc = lnet_parse_forward_locked(ni, msg);
1897 lnet_net_unlock(cpt);
1898
1899 if (rc < 0)
1900 goto free_drop;
1901 if (rc == 0) {
1902 lnet_ni_recv(ni, msg->msg_private, msg, 0,
1903 0, payload_length, payload_length);
1904 }
1905 return 0;
1906 }
1907
1908 lnet_net_unlock(cpt);
1909
1910 switch (type) {
1911 case LNET_MSG_ACK:
1912 rc = lnet_parse_ack(ni, msg);
1913 break;
1914 case LNET_MSG_PUT:
1915 rc = lnet_parse_put(ni, msg);
1916 break;
1917 case LNET_MSG_GET:
1918 rc = lnet_parse_get(ni, msg, rdma_req);
1919 break;
1920 case LNET_MSG_REPLY:
1921 rc = lnet_parse_reply(ni, msg);
1922 break;
1923 default:
1924 LASSERT(0);
1925 rc = -EPROTO;
1926 goto free_drop; /* prevent an unused label if !kernel */
1927 }
1928
1929 if (rc == 0)
1930 return 0;
1931
1932 LASSERT(rc == ENOENT);
1933
1934 free_drop:
1935 LASSERT(msg->msg_md == NULL);
1936 lnet_finalize(ni, msg, rc);
1937
1938 drop:
1939 lnet_drop_message(ni, cpt, private, payload_length);
1940 return 0;
1941 }
1942 EXPORT_SYMBOL(lnet_parse);
1943
1944 void
lnet_drop_delayed_msg_list(struct list_head * head,char * reason)1945 lnet_drop_delayed_msg_list(struct list_head *head, char *reason)
1946 {
1947 while (!list_empty(head)) {
1948 lnet_process_id_t id = {0};
1949 lnet_msg_t *msg;
1950
1951 msg = list_entry(head->next, lnet_msg_t, msg_list);
1952 list_del(&msg->msg_list);
1953
1954 id.nid = msg->msg_hdr.src_nid;
1955 id.pid = msg->msg_hdr.src_pid;
1956
1957 LASSERT(msg->msg_md == NULL);
1958 LASSERT(msg->msg_rx_delayed);
1959 LASSERT(msg->msg_rxpeer != NULL);
1960 LASSERT(msg->msg_hdr.type == LNET_MSG_PUT);
1961
1962 CWARN("Dropping delayed PUT from %s portal %d match %llu offset %d length %d: %s\n",
1963 libcfs_id2str(id),
1964 msg->msg_hdr.msg.put.ptl_index,
1965 msg->msg_hdr.msg.put.match_bits,
1966 msg->msg_hdr.msg.put.offset,
1967 msg->msg_hdr.payload_length, reason);
1968
1969 /* NB I can't drop msg's ref on msg_rxpeer until after I've
1970 * called lnet_drop_message(), so I just hang onto msg as well
1971 * until that's done */
1972
1973 lnet_drop_message(msg->msg_rxpeer->lp_ni,
1974 msg->msg_rxpeer->lp_cpt,
1975 msg->msg_private, msg->msg_len);
1976 /*
1977 * NB: message will not generate event because w/o attached MD,
1978 * but we still should give error code so lnet_msg_decommit()
1979 * can skip counters operations and other checks.
1980 */
1981 lnet_finalize(msg->msg_rxpeer->lp_ni, msg, -ENOENT);
1982 }
1983 }
1984
1985 void
lnet_recv_delayed_msg_list(struct list_head * head)1986 lnet_recv_delayed_msg_list(struct list_head *head)
1987 {
1988 while (!list_empty(head)) {
1989 lnet_msg_t *msg;
1990 lnet_process_id_t id;
1991
1992 msg = list_entry(head->next, lnet_msg_t, msg_list);
1993 list_del(&msg->msg_list);
1994
1995 /* md won't disappear under me, since each msg
1996 * holds a ref on it */
1997
1998 id.nid = msg->msg_hdr.src_nid;
1999 id.pid = msg->msg_hdr.src_pid;
2000
2001 LASSERT(msg->msg_rx_delayed);
2002 LASSERT(msg->msg_md != NULL);
2003 LASSERT(msg->msg_rxpeer != NULL);
2004 LASSERT(msg->msg_hdr.type == LNET_MSG_PUT);
2005
2006 CDEBUG(D_NET, "Resuming delayed PUT from %s portal %d match %llu offset %d length %d.\n",
2007 libcfs_id2str(id), msg->msg_hdr.msg.put.ptl_index,
2008 msg->msg_hdr.msg.put.match_bits,
2009 msg->msg_hdr.msg.put.offset,
2010 msg->msg_hdr.payload_length);
2011
2012 lnet_recv_put(msg->msg_rxpeer->lp_ni, msg);
2013 }
2014 }
2015
2016 /**
2017 * Initiate an asynchronous PUT operation.
2018 *
2019 * There are several events associated with a PUT: completion of the send on
2020 * the initiator node (LNET_EVENT_SEND), and when the send completes
2021 * successfully, the receipt of an acknowledgment (LNET_EVENT_ACK) indicating
2022 * that the operation was accepted by the target. The event LNET_EVENT_PUT is
2023 * used at the target node to indicate the completion of incoming data
2024 * delivery.
2025 *
2026 * The local events will be logged in the EQ associated with the MD pointed to
2027 * by \a mdh handle. Using a MD without an associated EQ results in these
2028 * events being discarded. In this case, the caller must have another
2029 * mechanism (e.g., a higher level protocol) for determining when it is safe
2030 * to modify the memory region associated with the MD.
2031 *
2032 * Note that LNet does not guarantee the order of LNET_EVENT_SEND and
2033 * LNET_EVENT_ACK, though intuitively ACK should happen after SEND.
2034 *
2035 * \param self Indicates the NID of a local interface through which to send
2036 * the PUT request. Use LNET_NID_ANY to let LNet choose one by itself.
2037 * \param mdh A handle for the MD that describes the memory to be sent. The MD
2038 * must be "free floating" (See LNetMDBind()).
2039 * \param ack Controls whether an acknowledgment is requested.
2040 * Acknowledgments are only sent when they are requested by the initiating
2041 * process and the target MD enables them.
2042 * \param target A process identifier for the target process.
2043 * \param portal The index in the \a target's portal table.
2044 * \param match_bits The match bits to use for MD selection at the target
2045 * process.
2046 * \param offset The offset into the target MD (only used when the target
2047 * MD has the LNET_MD_MANAGE_REMOTE option set).
2048 * \param hdr_data 64 bits of user data that can be included in the message
2049 * header. This data is written to an event queue entry at the target if an
2050 * EQ is present on the matching MD.
2051 *
2052 * \retval 0 Success, and only in this case events will be generated
2053 * and logged to EQ (if it exists).
2054 * \retval -EIO Simulated failure.
2055 * \retval -ENOMEM Memory allocation failure.
2056 * \retval -ENOENT Invalid MD object.
2057 *
2058 * \see lnet_event_t::hdr_data and lnet_event_kind_t.
2059 */
2060 int
LNetPut(lnet_nid_t self,lnet_handle_md_t mdh,lnet_ack_req_t ack,lnet_process_id_t target,unsigned int portal,__u64 match_bits,unsigned int offset,__u64 hdr_data)2061 LNetPut(lnet_nid_t self, lnet_handle_md_t mdh, lnet_ack_req_t ack,
2062 lnet_process_id_t target, unsigned int portal,
2063 __u64 match_bits, unsigned int offset,
2064 __u64 hdr_data)
2065 {
2066 struct lnet_msg *msg;
2067 struct lnet_libmd *md;
2068 int cpt;
2069 int rc;
2070
2071 LASSERT(the_lnet.ln_init);
2072 LASSERT(the_lnet.ln_refcount > 0);
2073
2074 if (!list_empty(&the_lnet.ln_test_peers) && /* normally we don't */
2075 fail_peer(target.nid, 1)) { /* shall we now? */
2076 CERROR("Dropping PUT to %s: simulated failure\n",
2077 libcfs_id2str(target));
2078 return -EIO;
2079 }
2080
2081 msg = lnet_msg_alloc();
2082 if (msg == NULL) {
2083 CERROR("Dropping PUT to %s: ENOMEM on lnet_msg_t\n",
2084 libcfs_id2str(target));
2085 return -ENOMEM;
2086 }
2087 msg->msg_vmflush = !!memory_pressure_get();
2088
2089 cpt = lnet_cpt_of_cookie(mdh.cookie);
2090 lnet_res_lock(cpt);
2091
2092 md = lnet_handle2md(&mdh);
2093 if (md == NULL || md->md_threshold == 0 || md->md_me != NULL) {
2094 CERROR("Dropping PUT (%llu:%d:%s): MD (%d) invalid\n",
2095 match_bits, portal, libcfs_id2str(target),
2096 md == NULL ? -1 : md->md_threshold);
2097 if (md != NULL && md->md_me != NULL)
2098 CERROR("Source MD also attached to portal %d\n",
2099 md->md_me->me_portal);
2100 lnet_res_unlock(cpt);
2101
2102 lnet_msg_free(msg);
2103 return -ENOENT;
2104 }
2105
2106 CDEBUG(D_NET, "LNetPut -> %s\n", libcfs_id2str(target));
2107
2108 lnet_msg_attach_md(msg, md, 0, 0);
2109
2110 lnet_prep_send(msg, LNET_MSG_PUT, target, 0, md->md_length);
2111
2112 msg->msg_hdr.msg.put.match_bits = cpu_to_le64(match_bits);
2113 msg->msg_hdr.msg.put.ptl_index = cpu_to_le32(portal);
2114 msg->msg_hdr.msg.put.offset = cpu_to_le32(offset);
2115 msg->msg_hdr.msg.put.hdr_data = hdr_data;
2116
2117 /* NB handles only looked up by creator (no flips) */
2118 if (ack == LNET_ACK_REQ) {
2119 msg->msg_hdr.msg.put.ack_wmd.wh_interface_cookie =
2120 the_lnet.ln_interface_cookie;
2121 msg->msg_hdr.msg.put.ack_wmd.wh_object_cookie =
2122 md->md_lh.lh_cookie;
2123 } else {
2124 msg->msg_hdr.msg.put.ack_wmd.wh_interface_cookie =
2125 LNET_WIRE_HANDLE_COOKIE_NONE;
2126 msg->msg_hdr.msg.put.ack_wmd.wh_object_cookie =
2127 LNET_WIRE_HANDLE_COOKIE_NONE;
2128 }
2129
2130 lnet_res_unlock(cpt);
2131
2132 lnet_build_msg_event(msg, LNET_EVENT_SEND);
2133
2134 rc = lnet_send(self, msg, LNET_NID_ANY);
2135 if (rc != 0) {
2136 CNETERR("Error sending PUT to %s: %d\n",
2137 libcfs_id2str(target), rc);
2138 lnet_finalize(NULL, msg, rc);
2139 }
2140
2141 /* completion will be signalled by an event */
2142 return 0;
2143 }
2144 EXPORT_SYMBOL(LNetPut);
2145
2146 lnet_msg_t *
lnet_create_reply_msg(lnet_ni_t * ni,lnet_msg_t * getmsg)2147 lnet_create_reply_msg(lnet_ni_t *ni, lnet_msg_t *getmsg)
2148 {
2149 /* The LND can DMA direct to the GET md (i.e. no REPLY msg). This
2150 * returns a msg for the LND to pass to lnet_finalize() when the sink
2151 * data has been received.
2152 *
2153 * CAVEAT EMPTOR: 'getmsg' is the original GET, which is freed when
2154 * lnet_finalize() is called on it, so the LND must call this first */
2155
2156 struct lnet_msg *msg = lnet_msg_alloc();
2157 struct lnet_libmd *getmd = getmsg->msg_md;
2158 lnet_process_id_t peer_id = getmsg->msg_target;
2159 int cpt;
2160
2161 LASSERT(!getmsg->msg_target_is_router);
2162 LASSERT(!getmsg->msg_routing);
2163
2164 cpt = lnet_cpt_of_cookie(getmd->md_lh.lh_cookie);
2165 lnet_res_lock(cpt);
2166
2167 LASSERT(getmd->md_refcount > 0);
2168
2169 if (msg == NULL) {
2170 CERROR("%s: Dropping REPLY from %s: can't allocate msg\n",
2171 libcfs_nid2str(ni->ni_nid), libcfs_id2str(peer_id));
2172 goto drop;
2173 }
2174
2175 if (getmd->md_threshold == 0) {
2176 CERROR("%s: Dropping REPLY from %s for inactive MD %p\n",
2177 libcfs_nid2str(ni->ni_nid), libcfs_id2str(peer_id),
2178 getmd);
2179 lnet_res_unlock(cpt);
2180 goto drop;
2181 }
2182
2183 LASSERT(getmd->md_offset == 0);
2184
2185 CDEBUG(D_NET, "%s: Reply from %s md %p\n",
2186 libcfs_nid2str(ni->ni_nid), libcfs_id2str(peer_id), getmd);
2187
2188 /* setup information for lnet_build_msg_event */
2189 msg->msg_from = peer_id.nid;
2190 msg->msg_type = LNET_MSG_GET; /* flag this msg as an "optimized" GET */
2191 msg->msg_hdr.src_nid = peer_id.nid;
2192 msg->msg_hdr.payload_length = getmd->md_length;
2193 msg->msg_receiving = 1; /* required by lnet_msg_attach_md */
2194
2195 lnet_msg_attach_md(msg, getmd, getmd->md_offset, getmd->md_length);
2196 lnet_res_unlock(cpt);
2197
2198 cpt = lnet_cpt_of_nid(peer_id.nid);
2199
2200 lnet_net_lock(cpt);
2201 lnet_msg_commit(msg, cpt);
2202 lnet_net_unlock(cpt);
2203
2204 lnet_build_msg_event(msg, LNET_EVENT_REPLY);
2205
2206 return msg;
2207
2208 drop:
2209 cpt = lnet_cpt_of_nid(peer_id.nid);
2210
2211 lnet_net_lock(cpt);
2212 the_lnet.ln_counters[cpt]->drop_count++;
2213 the_lnet.ln_counters[cpt]->drop_length += getmd->md_length;
2214 lnet_net_unlock(cpt);
2215
2216 if (msg != NULL)
2217 lnet_msg_free(msg);
2218
2219 return NULL;
2220 }
2221 EXPORT_SYMBOL(lnet_create_reply_msg);
2222
2223 void
lnet_set_reply_msg_len(lnet_ni_t * ni,lnet_msg_t * reply,unsigned int len)2224 lnet_set_reply_msg_len(lnet_ni_t *ni, lnet_msg_t *reply, unsigned int len)
2225 {
2226 /* Set the REPLY length, now the RDMA that elides the REPLY message has
2227 * completed and I know it. */
2228 LASSERT(reply != NULL);
2229 LASSERT(reply->msg_type == LNET_MSG_GET);
2230 LASSERT(reply->msg_ev.type == LNET_EVENT_REPLY);
2231
2232 /* NB I trusted my peer to RDMA. If she tells me she's written beyond
2233 * the end of my buffer, I might as well be dead. */
2234 LASSERT(len <= reply->msg_ev.mlength);
2235
2236 reply->msg_ev.mlength = len;
2237 }
2238 EXPORT_SYMBOL(lnet_set_reply_msg_len);
2239
2240 /**
2241 * Initiate an asynchronous GET operation.
2242 *
2243 * On the initiator node, an LNET_EVENT_SEND is logged when the GET request
2244 * is sent, and an LNET_EVENT_REPLY is logged when the data returned from
2245 * the target node in the REPLY has been written to local MD.
2246 *
2247 * On the target node, an LNET_EVENT_GET is logged when the GET request
2248 * arrives and is accepted into a MD.
2249 *
2250 * \param self,target,portal,match_bits,offset See the discussion in LNetPut().
2251 * \param mdh A handle for the MD that describes the memory into which the
2252 * requested data will be received. The MD must be "free floating"
2253 * (See LNetMDBind()).
2254 *
2255 * \retval 0 Success, and only in this case events will be generated
2256 * and logged to EQ (if it exists) of the MD.
2257 * \retval -EIO Simulated failure.
2258 * \retval -ENOMEM Memory allocation failure.
2259 * \retval -ENOENT Invalid MD object.
2260 */
2261 int
LNetGet(lnet_nid_t self,lnet_handle_md_t mdh,lnet_process_id_t target,unsigned int portal,__u64 match_bits,unsigned int offset)2262 LNetGet(lnet_nid_t self, lnet_handle_md_t mdh,
2263 lnet_process_id_t target, unsigned int portal,
2264 __u64 match_bits, unsigned int offset)
2265 {
2266 struct lnet_msg *msg;
2267 struct lnet_libmd *md;
2268 int cpt;
2269 int rc;
2270
2271 LASSERT(the_lnet.ln_init);
2272 LASSERT(the_lnet.ln_refcount > 0);
2273
2274 if (!list_empty(&the_lnet.ln_test_peers) && /* normally we don't */
2275 fail_peer(target.nid, 1)) { /* shall we now? */
2276 CERROR("Dropping GET to %s: simulated failure\n",
2277 libcfs_id2str(target));
2278 return -EIO;
2279 }
2280
2281 msg = lnet_msg_alloc();
2282 if (msg == NULL) {
2283 CERROR("Dropping GET to %s: ENOMEM on lnet_msg_t\n",
2284 libcfs_id2str(target));
2285 return -ENOMEM;
2286 }
2287
2288 cpt = lnet_cpt_of_cookie(mdh.cookie);
2289 lnet_res_lock(cpt);
2290
2291 md = lnet_handle2md(&mdh);
2292 if (md == NULL || md->md_threshold == 0 || md->md_me != NULL) {
2293 CERROR("Dropping GET (%llu:%d:%s): MD (%d) invalid\n",
2294 match_bits, portal, libcfs_id2str(target),
2295 md == NULL ? -1 : md->md_threshold);
2296 if (md != NULL && md->md_me != NULL)
2297 CERROR("REPLY MD also attached to portal %d\n",
2298 md->md_me->me_portal);
2299
2300 lnet_res_unlock(cpt);
2301
2302 lnet_msg_free(msg);
2303 return -ENOENT;
2304 }
2305
2306 CDEBUG(D_NET, "LNetGet -> %s\n", libcfs_id2str(target));
2307
2308 lnet_msg_attach_md(msg, md, 0, 0);
2309
2310 lnet_prep_send(msg, LNET_MSG_GET, target, 0, 0);
2311
2312 msg->msg_hdr.msg.get.match_bits = cpu_to_le64(match_bits);
2313 msg->msg_hdr.msg.get.ptl_index = cpu_to_le32(portal);
2314 msg->msg_hdr.msg.get.src_offset = cpu_to_le32(offset);
2315 msg->msg_hdr.msg.get.sink_length = cpu_to_le32(md->md_length);
2316
2317 /* NB handles only looked up by creator (no flips) */
2318 msg->msg_hdr.msg.get.return_wmd.wh_interface_cookie =
2319 the_lnet.ln_interface_cookie;
2320 msg->msg_hdr.msg.get.return_wmd.wh_object_cookie =
2321 md->md_lh.lh_cookie;
2322
2323 lnet_res_unlock(cpt);
2324
2325 lnet_build_msg_event(msg, LNET_EVENT_SEND);
2326
2327 rc = lnet_send(self, msg, LNET_NID_ANY);
2328 if (rc < 0) {
2329 CNETERR("Error sending GET to %s: %d\n",
2330 libcfs_id2str(target), rc);
2331 lnet_finalize(NULL, msg, rc);
2332 }
2333
2334 /* completion will be signalled by an event */
2335 return 0;
2336 }
2337 EXPORT_SYMBOL(LNetGet);
2338
2339 /**
2340 * Calculate distance to node at \a dstnid.
2341 *
2342 * \param dstnid Target NID.
2343 * \param srcnidp If not NULL, NID of the local interface to reach \a dstnid
2344 * is saved here.
2345 * \param orderp If not NULL, order of the route to reach \a dstnid is saved
2346 * here.
2347 *
2348 * \retval 0 If \a dstnid belongs to a local interface, and reserved option
2349 * local_nid_dist_zero is set, which is the default.
2350 * \retval positives Distance to target NID, i.e. number of hops plus one.
2351 * \retval -EHOSTUNREACH If \a dstnid is not reachable.
2352 */
2353 int
LNetDist(lnet_nid_t dstnid,lnet_nid_t * srcnidp,__u32 * orderp)2354 LNetDist(lnet_nid_t dstnid, lnet_nid_t *srcnidp, __u32 *orderp)
2355 {
2356 struct list_head *e;
2357 struct lnet_ni *ni;
2358 lnet_remotenet_t *rnet;
2359 __u32 dstnet = LNET_NIDNET(dstnid);
2360 int hops;
2361 int cpt;
2362 __u32 order = 2;
2363 struct list_head *rn_list;
2364
2365 /* if !local_nid_dist_zero, I don't return a distance of 0 ever
2366 * (when lustre sees a distance of 0, it substitutes 0@lo), so I
2367 * keep order 0 free for 0@lo and order 1 free for a local NID
2368 * match */
2369
2370 LASSERT(the_lnet.ln_init);
2371 LASSERT(the_lnet.ln_refcount > 0);
2372
2373 cpt = lnet_net_lock_current();
2374
2375 list_for_each(e, &the_lnet.ln_nis) {
2376 ni = list_entry(e, lnet_ni_t, ni_list);
2377
2378 if (ni->ni_nid == dstnid) {
2379 if (srcnidp != NULL)
2380 *srcnidp = dstnid;
2381 if (orderp != NULL) {
2382 if (LNET_NETTYP(LNET_NIDNET(dstnid)) == LOLND)
2383 *orderp = 0;
2384 else
2385 *orderp = 1;
2386 }
2387 lnet_net_unlock(cpt);
2388
2389 return local_nid_dist_zero ? 0 : 1;
2390 }
2391
2392 if (LNET_NIDNET(ni->ni_nid) == dstnet) {
2393 if (srcnidp != NULL)
2394 *srcnidp = ni->ni_nid;
2395 if (orderp != NULL)
2396 *orderp = order;
2397 lnet_net_unlock(cpt);
2398 return 1;
2399 }
2400
2401 order++;
2402 }
2403
2404 rn_list = lnet_net2rnethash(dstnet);
2405 list_for_each(e, rn_list) {
2406 rnet = list_entry(e, lnet_remotenet_t, lrn_list);
2407
2408 if (rnet->lrn_net == dstnet) {
2409 lnet_route_t *route;
2410 lnet_route_t *shortest = NULL;
2411
2412 LASSERT(!list_empty(&rnet->lrn_routes));
2413
2414 list_for_each_entry(route, &rnet->lrn_routes,
2415 lr_list) {
2416 if (shortest == NULL ||
2417 route->lr_hops < shortest->lr_hops)
2418 shortest = route;
2419 }
2420
2421 LASSERT(shortest != NULL);
2422 hops = shortest->lr_hops;
2423 if (srcnidp != NULL)
2424 *srcnidp = shortest->lr_gateway->lp_ni->ni_nid;
2425 if (orderp != NULL)
2426 *orderp = order;
2427 lnet_net_unlock(cpt);
2428 return hops + 1;
2429 }
2430 order++;
2431 }
2432
2433 lnet_net_unlock(cpt);
2434 return -EHOSTUNREACH;
2435 }
2436 EXPORT_SYMBOL(LNetDist);
2437
2438 /**
2439 * Set the number of asynchronous messages expected from a target process.
2440 *
2441 * This function is only meaningful for userspace callers. It's a no-op when
2442 * called from kernel.
2443 *
2444 * Asynchronous messages are those that can come from a target when the
2445 * userspace process is not waiting for IO to complete; e.g., AST callbacks
2446 * from Lustre servers. Specifying the expected number of such messages
2447 * allows them to be eagerly received when user process is not running in
2448 * LNet; otherwise network errors may occur.
2449 *
2450 * \param id Process ID of the target process.
2451 * \param nasync Number of asynchronous messages expected from the target.
2452 *
2453 * \return 0 on success, and an error code otherwise.
2454 */
2455 int
LNetSetAsync(lnet_process_id_t id,int nasync)2456 LNetSetAsync(lnet_process_id_t id, int nasync)
2457 {
2458 return 0;
2459 }
2460 EXPORT_SYMBOL(LNetSetAsync);
2461