projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
staging/lustre/libcfs: remove schedule_timeout_and_set_state
[deliverable/linux.git] / drivers / staging / lustre / lnet / lnet / router.c
CommitLineData
d7e09d03
PT		 1/*
2 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
3 *
4 * Copyright (c) 2011, 2012, Intel Corporation.
5 *
6 * This file is part of Portals
7 * http://sourceforge.net/projects/sandiaportals/
8 *
9 * Portals is free software; you can redistribute it and/or
10 * modify it under the terms of version 2 of the GNU General Public
11 * License as published by the Free Software Foundation.
12 *
13 * Portals is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
17 *
18 * You should have received a copy of the GNU General Public License
19 * along with Portals; if not, write to the Free Software
20 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 *
22 */
23
24#define DEBUG_SUBSYSTEM S_LNET
25#include <linux/lnet/lib-lnet.h>
26
27#if defined(LNET_ROUTER)
28
29#define LNET_NRB_TINY_MIN 512 /* min value for each CPT */
30#define LNET_NRB_TINY (LNET_NRB_TINY_MIN * 4)
31#define LNET_NRB_SMALL_MIN 4096 /* min value for each CPT */
32#define LNET_NRB_SMALL (LNET_NRB_SMALL_MIN * 4)
33#define LNET_NRB_LARGE_MIN 256 /* min value for each CPT */
34#define LNET_NRB_LARGE (LNET_NRB_LARGE_MIN * 4)
35
36static char *forwarding = "";
8cc7b4b9
PT		 37module_param(forwarding, charp, 0444);
38MODULE_PARM_DESC(forwarding, "Explicitly enable/disable forwarding between networks");
d7e09d03
PT		 39
40static int tiny_router_buffers;
8cc7b4b9
PT		 41module_param(tiny_router_buffers, int, 0444);
42MODULE_PARM_DESC(tiny_router_buffers, "# of 0 payload messages to buffer in the router");
d7e09d03 43static int small_router_buffers;
8cc7b4b9
PT		 44module_param(small_router_buffers, int, 0444);
45MODULE_PARM_DESC(small_router_buffers, "# of small (1 page) messages to buffer in the router");
d7e09d03 46static int large_router_buffers;
8cc7b4b9
PT		 47module_param(large_router_buffers, int, 0444);
48MODULE_PARM_DESC(large_router_buffers, "# of large messages to buffer in the router");
d7e09d03 49static int peer_buffer_credits = 0;
8cc7b4b9
PT		 50module_param(peer_buffer_credits, int, 0444);
51MODULE_PARM_DESC(peer_buffer_credits, "# router buffer credits per peer");
d7e09d03
PT		 52
53static int auto_down = 1;
8cc7b4b9
PT		 54module_param(auto_down, int, 0444);
55MODULE_PARM_DESC(auto_down, "Automatically mark peers down on comms error");
d7e09d03
PT		 56
57int
58lnet_peer_buffer_credits(lnet_ni_t *ni)
59{
60 /* NI option overrides LNet default */
61 if (ni->ni_peerrtrcredits > 0)
62 return ni->ni_peerrtrcredits;
63 if (peer_buffer_credits > 0)
64 return peer_buffer_credits;
65
66 /* As an approximation, allow this peer the same number of router
67 * buffers as it is allowed outstanding sends */
68 return ni->ni_peertxcredits;
69}
70
71/* forward ref's */
72static int lnet_router_checker(void *);
73#else
74
75int
76lnet_peer_buffer_credits(lnet_ni_t *ni)
77{
78 return 0;
79}
80
81#endif
82
83static int check_routers_before_use = 0;
8cc7b4b9
PT		 84module_param(check_routers_before_use, int, 0444);
85MODULE_PARM_DESC(check_routers_before_use, "Assume routers are down and ping them before use");
d7e09d03
PT		 86
87static int avoid_asym_router_failure = 1;
8cc7b4b9
PT		 88module_param(avoid_asym_router_failure, int, 0644);
89MODULE_PARM_DESC(avoid_asym_router_failure, "Avoid asymmetrical router failures (0 to disable)");
d7e09d03
PT		 90
91static int dead_router_check_interval = 60;
8cc7b4b9
PT		 92module_param(dead_router_check_interval, int, 0644);
93MODULE_PARM_DESC(dead_router_check_interval, "Seconds between dead router health checks (<= 0 to disable)");
d7e09d03
PT		 94
95static int live_router_check_interval = 60;
8cc7b4b9
PT		 96module_param(live_router_check_interval, int, 0644);
97MODULE_PARM_DESC(live_router_check_interval, "Seconds between live router health checks (<= 0 to disable)");
d7e09d03
PT		 98
99static int router_ping_timeout = 50;
8cc7b4b9
PT		 100module_param(router_ping_timeout, int, 0644);
101MODULE_PARM_DESC(router_ping_timeout, "Seconds to wait for the reply to a router health query");
d7e09d03
PT		 102
103int
104lnet_peers_start_down(void)
105{
106 return check_routers_before_use;
107}
108
109void
110lnet_notify_locked(lnet_peer_t *lp, int notifylnd, int alive, cfs_time_t when)
111{
112 if (cfs_time_before(when, lp->lp_timestamp)) { /* out of date information */
113 CDEBUG(D_NET, "Out of date\n");
114 return;
115 }
116
117 lp->lp_timestamp = when; /* update timestamp */
118 lp->lp_ping_deadline = 0; /* disable ping timeout */
119
120 if (lp->lp_alive_count != 0 && /* got old news */
121 (!lp->lp_alive) == (!alive)) { /* new date for old news */
122 CDEBUG(D_NET, "Old news\n");
123 return;
124 }
125
126 /* Flag that notification is outstanding */
127
128 lp->lp_alive_count++;
129 lp->lp_alive = !(!alive); /* 1 bit! */
130 lp->lp_notify = 1;
131 lp->lp_notifylnd |= notifylnd;
132 if (lp->lp_alive)
133 lp->lp_ping_feats = LNET_PING_FEAT_INVAL; /* reset */
134
135 CDEBUG(D_NET, "set %s %d\n", libcfs_nid2str(lp->lp_nid), alive);
136}
137
138void
139lnet_ni_notify_locked(lnet_ni_t *ni, lnet_peer_t *lp)
140{
141 int alive;
142 int notifylnd;
143
144 /* Notify only in 1 thread at any time to ensure ordered notification.
145 * NB individual events can be missed; the only guarantee is that you
146 * always get the most recent news */
147
148 if (lp->lp_notifying)
149 return;
150
151 lp->lp_notifying = 1;
152
153 while (lp->lp_notify) {
154 alive = lp->lp_alive;
155 notifylnd = lp->lp_notifylnd;
156
157 lp->lp_notifylnd = 0;
158 lp->lp_notify = 0;
159
160 if (notifylnd && ni->ni_lnd->lnd_notify != NULL) {
161 lnet_net_unlock(lp->lp_cpt);
162
163 /* A new notification could happen now; I'll handle it
164 * when control returns to me */
165
166 (ni->ni_lnd->lnd_notify)(ni, lp->lp_nid, alive);
167
168 lnet_net_lock(lp->lp_cpt);
169 }
170 }
171
172 lp->lp_notifying = 0;
173}
174
175
176static void
177lnet_rtr_addref_locked(lnet_peer_t *lp)
178{
179 LASSERT(lp->lp_refcount > 0);
180 LASSERT(lp->lp_rtr_refcount >= 0);
181
182 /* lnet_net_lock must be exclusively locked */
183 lp->lp_rtr_refcount++;
184 if (lp->lp_rtr_refcount == 1) {
185 struct list_head *pos;
186
187 /* a simple insertion sort */
188 list_for_each_prev(pos, &the_lnet.ln_routers) {
189 lnet_peer_t *rtr = list_entry(pos, lnet_peer_t,
190 lp_rtr_list);
191
192 if (rtr->lp_nid < lp->lp_nid)
193 break;
194 }
195
196 list_add(&lp->lp_rtr_list, pos);
197 /* addref for the_lnet.ln_routers */
198 lnet_peer_addref_locked(lp);
199 the_lnet.ln_routers_version++;
200 }
201}
202
203static void
204lnet_rtr_decref_locked(lnet_peer_t *lp)
205{
206 LASSERT(lp->lp_refcount > 0);
207 LASSERT(lp->lp_rtr_refcount > 0);
208
209 /* lnet_net_lock must be exclusively locked */
210 lp->lp_rtr_refcount--;
211 if (lp->lp_rtr_refcount == 0) {
212 LASSERT(list_empty(&lp->lp_routes));
213
214 if (lp->lp_rcd != NULL) {
215 list_add(&lp->lp_rcd->rcd_list,
216 &the_lnet.ln_rcd_deathrow);
217 lp->lp_rcd = NULL;
218 }
219
220 list_del(&lp->lp_rtr_list);
221 /* decref for the_lnet.ln_routers */
222 lnet_peer_decref_locked(lp);
223 the_lnet.ln_routers_version++;
224 }
225}
226
227lnet_remotenet_t *
228lnet_find_net_locked (__u32 net)
229{
230 lnet_remotenet_t *rnet;
231 struct list_head *tmp;
232 struct list_head *rn_list;
233
234 LASSERT(!the_lnet.ln_shutdown);
235
236 rn_list = lnet_net2rnethash(net);
237 list_for_each(tmp, rn_list) {
238 rnet = list_entry(tmp, lnet_remotenet_t, lrn_list);
239
240 if (rnet->lrn_net == net)
241 return rnet;
242 }
243 return NULL;
244}
245
246static void lnet_shuffle_seed(void)
247{
248 static int seeded = 0;
249 int lnd_type, seed[2];
250 struct timeval tv;
251 lnet_ni_t *ni;
252 struct list_head *tmp;
253
254 if (seeded)
255 return;
256
257 cfs_get_random_bytes(seed, sizeof(seed));
258
259 /* Nodes with small feet have little entropy
260 * the NID for this node gives the most entropy in the low bits */
261 list_for_each(tmp, &the_lnet.ln_nis) {
262 ni = list_entry(tmp, lnet_ni_t, ni_list);
263 lnd_type = LNET_NETTYP(LNET_NIDNET(ni->ni_nid));
264
265 if (lnd_type != LOLND)
266 seed[0] ^= (LNET_NIDADDR(ni->ni_nid) | lnd_type);
267 }
268
269 do_gettimeofday(&tv);
270 cfs_srand(tv.tv_sec ^ seed[0], tv.tv_usec ^ seed[1]);
271 seeded = 1;
272 return;
273}
274
275/* NB expects LNET_LOCK held */
276void
277lnet_add_route_to_rnet (lnet_remotenet_t *rnet, lnet_route_t *route)
278{
279 unsigned int len = 0;
280 unsigned int offset = 0;
281 struct list_head *e;
282
283 lnet_shuffle_seed();
284
285 list_for_each (e, &rnet->lrn_routes) {
286 len++;
287 }
288
289 /* len+1 positions to add a new entry, also prevents division by 0 */
290 offset = cfs_rand() % (len + 1);
291 list_for_each (e, &rnet->lrn_routes) {
292 if (offset == 0)
293 break;
294 offset--;
295 }
296 list_add(&route->lr_list, e);
297 list_add(&route->lr_gwlist, &route->lr_gateway->lp_routes);
298
299 the_lnet.ln_remote_nets_version++;
300 lnet_rtr_addref_locked(route->lr_gateway);
301}
302
303int
e75fb87f
DO		 304lnet_add_route(__u32 net, unsigned int hops, lnet_nid_t gateway,
305 unsigned int priority)
d7e09d03
PT		 306{
307 struct list_head *e;
308 lnet_remotenet_t *rnet;
309 lnet_remotenet_t *rnet2;
310 lnet_route_t *route;
311 lnet_ni_t *ni;
312 int add_route;
313 int rc;
314
e75fb87f
DO		 315 CDEBUG(D_NET, "Add route: net %s hops %u priority %u gw %s\n",
316 libcfs_net2str(net), hops, priority, libcfs_nid2str(gateway));
d7e09d03
PT		 317
318 if (gateway == LNET_NID_ANY ||
319 LNET_NETTYP(LNET_NIDNET(gateway)) == LOLND ||
320 net == LNET_NIDNET(LNET_NID_ANY) ||
321 LNET_NETTYP(net) == LOLND ||
322 LNET_NIDNET(gateway) == net ||
323 hops < 1 || hops > 255)
324 return (-EINVAL);
325
326 if (lnet_islocalnet(net)) /* it's a local network */
327 return 0; /* ignore the route entry */
328
329 /* Assume net, route, all new */
330 LIBCFS_ALLOC(route, sizeof(*route));
331 LIBCFS_ALLOC(rnet, sizeof(*rnet));
332 if (route == NULL || rnet == NULL) {
333 CERROR("Out of memory creating route %s %d %s\n",
334 libcfs_net2str(net), hops, libcfs_nid2str(gateway));
335 if (route != NULL)
336 LIBCFS_FREE(route, sizeof(*route));
337 if (rnet != NULL)
338 LIBCFS_FREE(rnet, sizeof(*rnet));
339 return -ENOMEM;
340 }
341
342 INIT_LIST_HEAD(&rnet->lrn_routes);
343 rnet->lrn_net = net;
344 route->lr_hops = hops;
345 route->lr_net = net;
e75fb87f 346 route->lr_priority = priority;
d7e09d03
PT		 347
348 lnet_net_lock(LNET_LOCK_EX);
349
350 rc = lnet_nid2peer_locked(&route->lr_gateway, gateway, LNET_LOCK_EX);
351 if (rc != 0) {
352 lnet_net_unlock(LNET_LOCK_EX);
353
354 LIBCFS_FREE(route, sizeof(*route));
355 LIBCFS_FREE(rnet, sizeof(*rnet));
356
357 if (rc == -EHOSTUNREACH) { /* gateway is not on a local net */
358 return 0; /* ignore the route entry */
359 } else {
360 CERROR("Error %d creating route %s %d %s\n", rc,
361 libcfs_net2str(net), hops,
362 libcfs_nid2str(gateway));
363 }
364 return rc;
365 }
366
367 LASSERT (!the_lnet.ln_shutdown);
368
369 rnet2 = lnet_find_net_locked(net);
370 if (rnet2 == NULL) {
371 /* new network */
372 list_add_tail(&rnet->lrn_list, lnet_net2rnethash(net));
373 rnet2 = rnet;
374 }
375
376 /* Search for a duplicate route (it's a NOOP if it is) */
377 add_route = 1;
378 list_for_each (e, &rnet2->lrn_routes) {
379 lnet_route_t *route2 = list_entry(e, lnet_route_t, lr_list);
380
381 if (route2->lr_gateway == route->lr_gateway) {
382 add_route = 0;
383 break;
384 }
385
386 /* our lookups must be true */
387 LASSERT (route2->lr_gateway->lp_nid != gateway);
388 }
389
390 if (add_route) {
391 lnet_peer_addref_locked(route->lr_gateway); /* +1 for notify */
392 lnet_add_route_to_rnet(rnet2, route);
393
394 ni = route->lr_gateway->lp_ni;
395 lnet_net_unlock(LNET_LOCK_EX);
396
397 /* XXX Assume alive */
398 if (ni->ni_lnd->lnd_notify != NULL)
399 (ni->ni_lnd->lnd_notify)(ni, gateway, 1);
400
401 lnet_net_lock(LNET_LOCK_EX);
402 }
403
404 /* -1 for notify or !add_route */
405 lnet_peer_decref_locked(route->lr_gateway);
406 lnet_net_unlock(LNET_LOCK_EX);
407
408 if (!add_route)
409 LIBCFS_FREE(route, sizeof(*route));
410
411 if (rnet != rnet2)
412 LIBCFS_FREE(rnet, sizeof(*rnet));
413
414 return 0;
415}
416
417int
418lnet_check_routes(void)
419{
420 lnet_remotenet_t *rnet;
421 lnet_route_t *route;
422 lnet_route_t *route2;
423 struct list_head *e1;
424 struct list_head *e2;
425 int cpt;
426 struct list_head *rn_list;
427 int i;
428
429 cpt = lnet_net_lock_current();
430
431 for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) {
432 rn_list = &the_lnet.ln_remote_nets_hash[i];
433 list_for_each(e1, rn_list) {
434 rnet = list_entry(e1, lnet_remotenet_t, lrn_list);
435
436 route2 = NULL;
437 list_for_each(e2, &rnet->lrn_routes) {
438 lnet_nid_t nid1;
439 lnet_nid_t nid2;
440 int net;
441
442 route = list_entry(e2, lnet_route_t,
443 lr_list);
444
445 if (route2 == NULL) {
446 route2 = route;
447 continue;
448 }
449
450 if (route->lr_gateway->lp_ni ==
451 route2->lr_gateway->lp_ni)
452 continue;
453
454 nid1 = route->lr_gateway->lp_nid;
455 nid2 = route2->lr_gateway->lp_nid;
456 net = rnet->lrn_net;
457
458 lnet_net_unlock(cpt);
459
460 CERROR("Routes to %s via %s and %s not "
461 "supported\n",
462 libcfs_net2str(net),
463 libcfs_nid2str(nid1),
464 libcfs_nid2str(nid2));
465 return -EINVAL;
466 }
467 }
468 }
469
470 lnet_net_unlock(cpt);
471 return 0;
472}
473
474int
475lnet_del_route(__u32 net, lnet_nid_t gw_nid)
476{
477 struct lnet_peer *gateway;
478 lnet_remotenet_t *rnet;
479 lnet_route_t *route;
480 struct list_head *e1;
481 struct list_head *e2;
482 int rc = -ENOENT;
483 struct list_head *rn_list;
484 int idx = 0;
485
486 CDEBUG(D_NET, "Del route: net %s : gw %s\n",
487 libcfs_net2str(net), libcfs_nid2str(gw_nid));
488
489 /* NB Caller may specify either all routes via the given gateway
490 * or a specific route entry actual NIDs) */
491
492 lnet_net_lock(LNET_LOCK_EX);
493 if (net == LNET_NIDNET(LNET_NID_ANY))
494 rn_list = &the_lnet.ln_remote_nets_hash[0];
495 else
496 rn_list = lnet_net2rnethash(net);
497
498 again:
499 list_for_each(e1, rn_list) {
500 rnet = list_entry(e1, lnet_remotenet_t, lrn_list);
501
502 if (!(net == LNET_NIDNET(LNET_NID_ANY) ||
503 net == rnet->lrn_net))
504 continue;
505
506 list_for_each(e2, &rnet->lrn_routes) {
507 route = list_entry(e2, lnet_route_t, lr_list);
508
509 gateway = route->lr_gateway;
510 if (!(gw_nid == LNET_NID_ANY ||
511 gw_nid == gateway->lp_nid))
512 continue;
513
514 list_del(&route->lr_list);
515 list_del(&route->lr_gwlist);
516 the_lnet.ln_remote_nets_version++;
517
518 if (list_empty(&rnet->lrn_routes))
519 list_del(&rnet->lrn_list);
520 else
521 rnet = NULL;
522
523 lnet_rtr_decref_locked(gateway);
524 lnet_peer_decref_locked(gateway);
525
526 lnet_net_unlock(LNET_LOCK_EX);
527
528 LIBCFS_FREE(route, sizeof(*route));
529
530 if (rnet != NULL)
531 LIBCFS_FREE(rnet, sizeof(*rnet));
532
533 rc = 0;
534 lnet_net_lock(LNET_LOCK_EX);
535 goto again;
536 }
537 }
538
539 if (net == LNET_NIDNET(LNET_NID_ANY) &&
540 ++idx < LNET_REMOTE_NETS_HASH_SIZE) {
541 rn_list = &the_lnet.ln_remote_nets_hash[idx];
542 goto again;
543 }
544 lnet_net_unlock(LNET_LOCK_EX);
545
546 return rc;
547}
548
549void
550lnet_destroy_routes (void)
551{
552 lnet_del_route(LNET_NIDNET(LNET_NID_ANY), LNET_NID_ANY);
553}
554
555int
556lnet_get_route(int idx, __u32 *net, __u32 *hops,
e75fb87f 557 lnet_nid_t *gateway, __u32 *alive, __u32 *priority)
d7e09d03
PT		 558{
559 struct list_head *e1;
560 struct list_head *e2;
561 lnet_remotenet_t *rnet;
562 lnet_route_t *route;
563 int cpt;
564 int i;
565 struct list_head *rn_list;
566
567 cpt = lnet_net_lock_current();
568
569 for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) {
570 rn_list = &the_lnet.ln_remote_nets_hash[i];
571 list_for_each(e1, rn_list) {
572 rnet = list_entry(e1, lnet_remotenet_t, lrn_list);
573
574 list_for_each(e2, &rnet->lrn_routes) {
575 route = list_entry(e2, lnet_route_t,
576 lr_list);
577
578 if (idx-- == 0) {
e75fb87f
DO		 579 *net = rnet->lrn_net;
580 *hops = route->lr_hops;
581 *priority = route->lr_priority;
582 *gateway = route->lr_gateway->lp_nid;
583 *alive = route->lr_gateway->lp_alive;
d7e09d03
PT		 584 lnet_net_unlock(cpt);
585 return 0;
586 }
587 }
588 }
589 }
590
591 lnet_net_unlock(cpt);
592 return -ENOENT;
593}
594
595void
596lnet_swap_pinginfo(lnet_ping_info_t *info)
597{
598 int i;
599 lnet_ni_status_t *stat;
600
601 __swab32s(&info->pi_magic);
602 __swab32s(&info->pi_features);
603 __swab32s(&info->pi_pid);
604 __swab32s(&info->pi_nnis);
605 for (i = 0; i < info->pi_nnis && i < LNET_MAX_RTR_NIS; i++) {
606 stat = &info->pi_ni[i];
607 __swab64s(&stat->ns_nid);
608 __swab32s(&stat->ns_status);
609 }
610 return;
611}
612
613/**
614 * parse router-checker pinginfo, record number of down NIs for remote
615 * networks on that router.
616 */
617static void
618lnet_parse_rc_info(lnet_rc_data_t *rcd)
619{
620 lnet_ping_info_t *info = rcd->rcd_pinginfo;
621 struct lnet_peer *gw = rcd->rcd_gateway;
622 lnet_route_t *rtr;
623
624 if (!gw->lp_alive)
625 return;
626
627 if (info->pi_magic == __swab32(LNET_PROTO_PING_MAGIC))
628 lnet_swap_pinginfo(info);
629
630 /* NB always racing with network! */
631 if (info->pi_magic != LNET_PROTO_PING_MAGIC) {
632 CDEBUG(D_NET, "%s: Unexpected magic %08x\n",
633 libcfs_nid2str(gw->lp_nid), info->pi_magic);
634 gw->lp_ping_feats = LNET_PING_FEAT_INVAL;
635 return;
636 }
637
638 gw->lp_ping_feats = info->pi_features;
639 if ((gw->lp_ping_feats & LNET_PING_FEAT_MASK) == 0) {
640 CDEBUG(D_NET, "%s: Unexpected features 0x%x\n",
641 libcfs_nid2str(gw->lp_nid), gw->lp_ping_feats);
642 return; /* nothing I can understand */
643 }
644
645 if ((gw->lp_ping_feats & LNET_PING_FEAT_NI_STATUS) == 0)
646 return; /* can't carry NI status info */
647
648 list_for_each_entry(rtr, &gw->lp_routes, lr_gwlist) {
649 int ptl_status = LNET_NI_STATUS_INVALID;
650 int down = 0;
651 int up = 0;
652 int i;
653
654 for (i = 0; i < info->pi_nnis && i < LNET_MAX_RTR_NIS; i++) {
655 lnet_ni_status_t *stat = &info->pi_ni[i];
656 lnet_nid_t nid = stat->ns_nid;
657
658 if (nid == LNET_NID_ANY) {
659 CDEBUG(D_NET, "%s: unexpected LNET_NID_ANY\n",
660 libcfs_nid2str(gw->lp_nid));
661 gw->lp_ping_feats = LNET_PING_FEAT_INVAL;
662 return;
663 }
664
665 if (LNET_NETTYP(LNET_NIDNET(nid)) == LOLND)
666 continue;
667
668 if (stat->ns_status == LNET_NI_STATUS_DOWN) {
669 if (LNET_NETTYP(LNET_NIDNET(nid)) != PTLLND)
670 down++;
671 else if (ptl_status != LNET_NI_STATUS_UP)
672 ptl_status = LNET_NI_STATUS_DOWN;
673 continue;
674 }
675
676 if (stat->ns_status == LNET_NI_STATUS_UP) {
677 if (LNET_NIDNET(nid) == rtr->lr_net) {
678 up = 1;
679 break;
680 }
681 /* ptl NIs are considered down only when
682 * they're all down */
683 if (LNET_NETTYP(LNET_NIDNET(nid)) == PTLLND)
684 ptl_status = LNET_NI_STATUS_UP;
685 continue;
686 }
687
688 CDEBUG(D_NET, "%s: Unexpected status 0x%x\n",
689 libcfs_nid2str(gw->lp_nid), stat->ns_status);
690 gw->lp_ping_feats = LNET_PING_FEAT_INVAL;
691 return;
692 }
693
694 if (up) { /* ignore downed NIs if NI for dest network is up */
695 rtr->lr_downis = 0;
696 continue;
697 }
698 rtr->lr_downis = down + (ptl_status == LNET_NI_STATUS_DOWN);
699 }
700}
701
702static void
703lnet_router_checker_event(lnet_event_t *event)
704{
705 lnet_rc_data_t *rcd = event->md.user_ptr;
706 struct lnet_peer *lp;
707
708 LASSERT(rcd != NULL);
709
710 if (event->unlinked) {
711 LNetInvalidateHandle(&rcd->rcd_mdh);
712 return;
713 }
714
715 LASSERT(event->type == LNET_EVENT_SEND ||
716 event->type == LNET_EVENT_REPLY);
717
718 lp = rcd->rcd_gateway;
719 LASSERT(lp != NULL);
720
721 /* NB: it's called with holding lnet_res_lock, we have a few
722 * places need to hold both locks at the same time, please take
723 * care of lock ordering */
724 lnet_net_lock(lp->lp_cpt);
725 if (!lnet_isrouter(lp) || lp->lp_rcd != rcd) {
726 /* ignore if no longer a router or rcd is replaced */
727 goto out;
728 }
729
730 if (event->type == LNET_EVENT_SEND) {
731 lp->lp_ping_notsent = 0;
732 if (event->status == 0)
733 goto out;
734 }
735
736 /* LNET_EVENT_REPLY */
737 /* A successful REPLY means the router is up. If _any_ comms
738 * to the router fail I assume it's down (this will happen if
739 * we ping alive routers to try to detect router death before
740 * apps get burned). */
741
742 lnet_notify_locked(lp, 1, (event->status == 0), cfs_time_current());
743 /* The router checker will wake up very shortly and do the
744 * actual notification.
745 * XXX If 'lp' stops being a router before then, it will still
746 * have the notification pending!!! */
747
748 if (avoid_asym_router_failure && event->status == 0)
749 lnet_parse_rc_info(rcd);
750
751 out:
752 lnet_net_unlock(lp->lp_cpt);
753}
754
755void
756lnet_wait_known_routerstate(void)
757{
758 lnet_peer_t *rtr;
759 struct list_head *entry;
760 int all_known;
761
762 LASSERT (the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
763
764 for (;;) {
765 int cpt = lnet_net_lock_current();
766
767 all_known = 1;
768 list_for_each (entry, &the_lnet.ln_routers) {
769 rtr = list_entry(entry, lnet_peer_t, lp_rtr_list);
770
771 if (rtr->lp_alive_count == 0) {
772 all_known = 0;
773 break;
774 }
775 }
776
777 lnet_net_unlock(cpt);
778
779 if (all_known)
780 return;
781
782 cfs_pause(cfs_time_seconds(1));
783 }
784}
785
786void
787lnet_update_ni_status_locked(void)
788{
789 lnet_ni_t *ni;
790 long now;
791 int timeout;
792
793 LASSERT(the_lnet.ln_routing);
794
795 timeout = router_ping_timeout +
796 MAX(live_router_check_interval, dead_router_check_interval);
797
798 now = cfs_time_current_sec();
799 list_for_each_entry(ni, &the_lnet.ln_nis, ni_list) {
800 if (ni->ni_lnd->lnd_type == LOLND)
801 continue;
802
803 if (now < ni->ni_last_alive + timeout)
804 continue;
805
806 lnet_ni_lock(ni);
807 /* re-check with lock */
808 if (now < ni->ni_last_alive + timeout) {
809 lnet_ni_unlock(ni);
810 continue;
811 }
812
813 LASSERT(ni->ni_status != NULL);
814
815 if (ni->ni_status->ns_status != LNET_NI_STATUS_DOWN) {
816 CDEBUG(D_NET, "NI(%s:%d) status changed to down\n",
817 libcfs_nid2str(ni->ni_nid), timeout);
818 /* NB: so far, this is the only place to set
819 * NI status to "down" */
820 ni->ni_status->ns_status = LNET_NI_STATUS_DOWN;
821 }
822 lnet_ni_unlock(ni);
823 }
824}
825
826void
827lnet_destroy_rc_data(lnet_rc_data_t *rcd)
828{
829 LASSERT(list_empty(&rcd->rcd_list));
830 /* detached from network */
831 LASSERT(LNetHandleIsInvalid(rcd->rcd_mdh));
832
833 if (rcd->rcd_gateway != NULL) {
834 int cpt = rcd->rcd_gateway->lp_cpt;
835
836 lnet_net_lock(cpt);
837 lnet_peer_decref_locked(rcd->rcd_gateway);
838 lnet_net_unlock(cpt);
839 }
840
841 if (rcd->rcd_pinginfo != NULL)
842 LIBCFS_FREE(rcd->rcd_pinginfo, LNET_PINGINFO_SIZE);
843
844 LIBCFS_FREE(rcd, sizeof(*rcd));
845}
846
847lnet_rc_data_t *
848lnet_create_rc_data_locked(lnet_peer_t *gateway)
849{
850 lnet_rc_data_t *rcd = NULL;
851 lnet_ping_info_t *pi;
852 int rc;
853 int i;
854
855 lnet_net_unlock(gateway->lp_cpt);
856
857 LIBCFS_ALLOC(rcd, sizeof(*rcd));
858 if (rcd == NULL)
859 goto out;
860
861 LNetInvalidateHandle(&rcd->rcd_mdh);
862 INIT_LIST_HEAD(&rcd->rcd_list);
863
864 LIBCFS_ALLOC(pi, LNET_PINGINFO_SIZE);
865 if (pi == NULL)
866 goto out;
867
868 memset(pi, 0, LNET_PINGINFO_SIZE);
869 for (i = 0; i < LNET_MAX_RTR_NIS; i++) {
870 pi->pi_ni[i].ns_nid = LNET_NID_ANY;
871 pi->pi_ni[i].ns_status = LNET_NI_STATUS_INVALID;
872 }
873 rcd->rcd_pinginfo = pi;
874
875 LASSERT (!LNetHandleIsInvalid(the_lnet.ln_rc_eqh));
876 rc = LNetMDBind((lnet_md_t){.start = pi,
877 .user_ptr = rcd,
878 .length = LNET_PINGINFO_SIZE,
879 .threshold = LNET_MD_THRESH_INF,
880 .options = LNET_MD_TRUNCATE,
881 .eq_handle = the_lnet.ln_rc_eqh},
882 LNET_UNLINK,
883 &rcd->rcd_mdh);
884 if (rc < 0) {
885 CERROR("Can't bind MD: %d\n", rc);
886 goto out;
887 }
888 LASSERT(rc == 0);
889
890 lnet_net_lock(gateway->lp_cpt);
891 /* router table changed or someone has created rcd for this gateway */
892 if (!lnet_isrouter(gateway) || gateway->lp_rcd != NULL) {
893 lnet_net_unlock(gateway->lp_cpt);
894 goto out;
895 }
896
897 lnet_peer_addref_locked(gateway);
898 rcd->rcd_gateway = gateway;
899 gateway->lp_rcd = rcd;
900 gateway->lp_ping_notsent = 0;
901
902 return rcd;
903
904 out:
905 if (rcd != NULL) {
906 if (!LNetHandleIsInvalid(rcd->rcd_mdh)) {
907 rc = LNetMDUnlink(rcd->rcd_mdh);
908 LASSERT(rc == 0);
909 }
910 lnet_destroy_rc_data(rcd);
911 }
912
913 lnet_net_lock(gateway->lp_cpt);
914 return gateway->lp_rcd;
915}
916
917static int
918lnet_router_check_interval (lnet_peer_t *rtr)
919{
920 int secs;
921
922 secs = rtr->lp_alive ? live_router_check_interval :
923 dead_router_check_interval;
924 if (secs < 0)
925 secs = 0;
926
927 return secs;
928}
929
930static void
931lnet_ping_router_locked (lnet_peer_t *rtr)
932{
933 lnet_rc_data_t *rcd = NULL;
934 cfs_time_t now = cfs_time_current();
935 int secs;
936
937 lnet_peer_addref_locked(rtr);
938
939 if (rtr->lp_ping_deadline != 0 && /* ping timed out? */
940 cfs_time_after(now, rtr->lp_ping_deadline))
941 lnet_notify_locked(rtr, 1, 0, now);
942
943 /* Run any outstanding notifications */
944 lnet_ni_notify_locked(rtr->lp_ni, rtr);
945
946 if (!lnet_isrouter(rtr) ||
947 the_lnet.ln_rc_state != LNET_RC_STATE_RUNNING) {
948 /* router table changed or router checker is shutting down */
949 lnet_peer_decref_locked(rtr);
950 return;
951 }
952
953 rcd = rtr->lp_rcd != NULL ?
954 rtr->lp_rcd : lnet_create_rc_data_locked(rtr);
955
956 if (rcd == NULL)
957 return;
958
959 secs = lnet_router_check_interval(rtr);
960
961 CDEBUG(D_NET,
962 "rtr %s %d: deadline %lu ping_notsent %d alive %d "
963 "alive_count %d lp_ping_timestamp %lu\n",
964 libcfs_nid2str(rtr->lp_nid), secs,
965 rtr->lp_ping_deadline, rtr->lp_ping_notsent,
966 rtr->lp_alive, rtr->lp_alive_count, rtr->lp_ping_timestamp);
967
968 if (secs != 0 && !rtr->lp_ping_notsent &&
969 cfs_time_after(now, cfs_time_add(rtr->lp_ping_timestamp,
970 cfs_time_seconds(secs)))) {
971 int rc;
972 lnet_process_id_t id;
973 lnet_handle_md_t mdh;
974
975 id.nid = rtr->lp_nid;
976 id.pid = LUSTRE_SRV_LNET_PID;
977 CDEBUG(D_NET, "Check: %s\n", libcfs_id2str(id));
978
979 rtr->lp_ping_notsent = 1;
980 rtr->lp_ping_timestamp = now;
981
982 mdh = rcd->rcd_mdh;
983
984 if (rtr->lp_ping_deadline == 0) {
985 rtr->lp_ping_deadline =
986 cfs_time_shift(router_ping_timeout);
987 }
988
989 lnet_net_unlock(rtr->lp_cpt);
990
991 rc = LNetGet(LNET_NID_ANY, mdh, id, LNET_RESERVED_PORTAL,
992 LNET_PROTO_PING_MATCHBITS, 0);
993
994 lnet_net_lock(rtr->lp_cpt);
995 if (rc != 0)
996 rtr->lp_ping_notsent = 0; /* no event pending */
997 }
998
999 lnet_peer_decref_locked(rtr);
1000 return;
1001}
1002
1003int
1004lnet_router_checker_start(void)
1005{
1006 int rc;
1007 int eqsz;
1008
1009 LASSERT (the_lnet.ln_rc_state == LNET_RC_STATE_SHUTDOWN);
1010
1011 if (check_routers_before_use &&
1012 dead_router_check_interval <= 0) {
1013 LCONSOLE_ERROR_MSG(0x10a, "'dead_router_check_interval' must be"
1014 " set if 'check_routers_before_use' is set"
1015 "\n");
1016 return -EINVAL;
1017 }
1018
1019 if (!the_lnet.ln_routing &&
1020 live_router_check_interval <= 0 &&
1021 dead_router_check_interval <= 0)
1022 return 0;
1023
1024 sema_init(&the_lnet.ln_rc_signal, 0);
1025 /* EQ size doesn't matter; the callback is guaranteed to get every
1026 * event */
1027 eqsz = 0;
1028 rc = LNetEQAlloc(eqsz, lnet_router_checker_event,
1029 &the_lnet.ln_rc_eqh);
1030 if (rc != 0) {
1031 CERROR("Can't allocate EQ(%d): %d\n", eqsz, rc);
1032 return -ENOMEM;
1033 }
1034
1035 the_lnet.ln_rc_state = LNET_RC_STATE_RUNNING;
1036 rc = PTR_ERR(kthread_run(lnet_router_checker,
1037 NULL, "router_checker"));
1038 if (IS_ERR_VALUE(rc)) {
1039 CERROR("Can't start router checker thread: %d\n", rc);
1040 /* block until event callback signals exit */
1041 down(&the_lnet.ln_rc_signal);
1042 rc = LNetEQFree(the_lnet.ln_rc_eqh);
1043 LASSERT(rc == 0);
1044 the_lnet.ln_rc_state = LNET_RC_STATE_SHUTDOWN;
1045 return -ENOMEM;
1046 }
1047
1048 if (check_routers_before_use) {
1049 /* Note that a helpful side-effect of pinging all known routers
1050 * at startup is that it makes them drop stale connections they
1051 * may have to a previous instance of me. */
1052 lnet_wait_known_routerstate();
1053 }
1054
1055 return 0;
1056}
1057
1058void
1059lnet_router_checker_stop (void)
1060{
1061 int rc;
1062
1063 if (the_lnet.ln_rc_state == LNET_RC_STATE_SHUTDOWN)
1064 return;
1065
1066 LASSERT (the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
1067 the_lnet.ln_rc_state = LNET_RC_STATE_STOPPING;
1068
1069 /* block until event callback signals exit */
1070 down(&the_lnet.ln_rc_signal);
1071 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_SHUTDOWN);
1072
1073 rc = LNetEQFree(the_lnet.ln_rc_eqh);
1074 LASSERT (rc == 0);
1075 return;
1076}
1077
1078static void
1079lnet_prune_rc_data(int wait_unlink)
1080{
1081 lnet_rc_data_t *rcd;
1082 lnet_rc_data_t *tmp;
1083 lnet_peer_t *lp;
1084 struct list_head head;
1085 int i = 2;
1086
1087 if (likely(the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING &&
1088 list_empty(&the_lnet.ln_rcd_deathrow) &&
1089 list_empty(&the_lnet.ln_rcd_zombie)))
1090 return;
1091
1092 INIT_LIST_HEAD(&head);
1093
1094 lnet_net_lock(LNET_LOCK_EX);
1095
1096 if (the_lnet.ln_rc_state != LNET_RC_STATE_RUNNING) {
1097 /* router checker is stopping, prune all */
1098 list_for_each_entry(lp, &the_lnet.ln_routers,
1099 lp_rtr_list) {
1100 if (lp->lp_rcd == NULL)
1101 continue;
1102
1103 LASSERT(list_empty(&lp->lp_rcd->rcd_list));
1104 list_add(&lp->lp_rcd->rcd_list,
1105 &the_lnet.ln_rcd_deathrow);
1106 lp->lp_rcd = NULL;
1107 }
1108 }
1109
1110 /* unlink all RCDs on deathrow list */
1111 list_splice_init(&the_lnet.ln_rcd_deathrow, &head);
1112
1113 if (!list_empty(&head)) {
1114 lnet_net_unlock(LNET_LOCK_EX);
1115
1116 list_for_each_entry(rcd, &head, rcd_list)
1117 LNetMDUnlink(rcd->rcd_mdh);
1118
1119 lnet_net_lock(LNET_LOCK_EX);
1120 }
1121
1122 list_splice_init(&head, &the_lnet.ln_rcd_zombie);
1123
1124 /* release all zombie RCDs */
1125 while (!list_empty(&the_lnet.ln_rcd_zombie)) {
1126 list_for_each_entry_safe(rcd, tmp, &the_lnet.ln_rcd_zombie,
1127 rcd_list) {
1128 if (LNetHandleIsInvalid(rcd->rcd_mdh))
1129 list_move(&rcd->rcd_list, &head);
1130 }
1131
1132 wait_unlink = wait_unlink &&
1133 !list_empty(&the_lnet.ln_rcd_zombie);
1134
1135 lnet_net_unlock(LNET_LOCK_EX);
1136
1137 while (!list_empty(&head)) {
1138 rcd = list_entry(head.next,
1139 lnet_rc_data_t, rcd_list);
1140 list_del_init(&rcd->rcd_list);
1141 lnet_destroy_rc_data(rcd);
1142 }
1143
1144 if (!wait_unlink)
1145 return;
1146
1147 i++;
1148 CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET,
1149 "Waiting for rc buffers to unlink\n");
1150 cfs_pause(cfs_time_seconds(1) / 4);
1151
1152 lnet_net_lock(LNET_LOCK_EX);
1153 }
1154
1155 lnet_net_unlock(LNET_LOCK_EX);
1156}
1157
1158
1159#if defined(LNET_ROUTER)
1160
1161static int
1162lnet_router_checker(void *arg)
1163{
1164 lnet_peer_t *rtr;
1165 struct list_head *entry;
1166
1167 cfs_block_allsigs();
1168
1169 LASSERT (the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
1170
1171 while (the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING) {
1172 __u64 version;
1173 int cpt;
1174 int cpt2;
1175
1176 cpt = lnet_net_lock_current();
1177rescan:
1178 version = the_lnet.ln_routers_version;
1179
1180 list_for_each(entry, &the_lnet.ln_routers) {
1181 rtr = list_entry(entry, lnet_peer_t, lp_rtr_list);
1182
1183 cpt2 = lnet_cpt_of_nid_locked(rtr->lp_nid);
1184 if (cpt != cpt2) {
1185 lnet_net_unlock(cpt);
1186 cpt = cpt2;
1187 lnet_net_lock(cpt);
1188 /* the routers list has changed */
1189 if (version != the_lnet.ln_routers_version)
1190 goto rescan;
1191 }
1192
1193 lnet_ping_router_locked(rtr);
1194
1195 /* NB dropped lock */
1196 if (version != the_lnet.ln_routers_version) {
1197 /* the routers list has changed */
1198 goto rescan;
1199 }
1200 }
1201
1202 if (the_lnet.ln_routing)
1203 lnet_update_ni_status_locked();
1204
1205 lnet_net_unlock(cpt);
1206
1207 lnet_prune_rc_data(0); /* don't wait for UNLINK */
1208
1209 /* Call cfs_pause() here always adds 1 to load average
1210 * because kernel counts # active tasks as nr_running
1211 * + nr_uninterruptible. */
18fd5baa
PT		 1212 set_current_state(TASK_INTERRUPTIBLE);
1213 schedule_timeout(cfs_time_seconds(1));
d7e09d03
PT		 1214 }
1215
1216 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_STOPPING);
1217
1218 lnet_prune_rc_data(1); /* wait for UNLINK */
1219
1220 the_lnet.ln_rc_state = LNET_RC_STATE_SHUTDOWN;
1221 up(&the_lnet.ln_rc_signal);
1222 /* The unlink event callback will signal final completion */
1223 return 0;
1224}
1225
1226void
1227lnet_destroy_rtrbuf(lnet_rtrbuf_t *rb, int npages)
1228{
1229 int sz = offsetof(lnet_rtrbuf_t, rb_kiov[npages]);
1230
1231 while (--npages >= 0)
1232 __free_page(rb->rb_kiov[npages].kiov_page);
1233
1234 LIBCFS_FREE(rb, sz);
1235}
1236
1237lnet_rtrbuf_t *
1238lnet_new_rtrbuf(lnet_rtrbufpool_t *rbp, int cpt)
1239{
1240 int npages = rbp->rbp_npages;
1241 int sz = offsetof(lnet_rtrbuf_t, rb_kiov[npages]);
1242 struct page *page;
1243 lnet_rtrbuf_t *rb;
1244 int i;
1245
1246 LIBCFS_CPT_ALLOC(rb, lnet_cpt_table(), cpt, sz);
1247 if (rb == NULL)
1248 return NULL;
1249
1250 rb->rb_pool = rbp;
1251
1252 for (i = 0; i < npages; i++) {
49c02a75
PT		 1253 page = alloc_pages_node(
1254 cfs_cpt_spread_node(lnet_cpt_table(), cpt),
1255 __GFP_ZERO | GFP_IOFS, 0);
d7e09d03
PT		 1256 if (page == NULL) {
1257 while (--i >= 0)
1258 __free_page(rb->rb_kiov[i].kiov_page);
1259
1260 LIBCFS_FREE(rb, sz);
1261 return NULL;
1262 }
1263
1264 rb->rb_kiov[i].kiov_len = PAGE_CACHE_SIZE;
1265 rb->rb_kiov[i].kiov_offset = 0;
1266 rb->rb_kiov[i].kiov_page = page;
1267 }
1268
1269 return rb;
1270}
1271
1272void
1273lnet_rtrpool_free_bufs(lnet_rtrbufpool_t *rbp)
1274{
1275 int npages = rbp->rbp_npages;
1276 int nbuffers = 0;
1277 lnet_rtrbuf_t *rb;
1278
1279 if (rbp->rbp_nbuffers == 0) /* not initialized or already freed */
1280 return;
1281
1282 LASSERT (list_empty(&rbp->rbp_msgs));
1283 LASSERT (rbp->rbp_credits == rbp->rbp_nbuffers);
1284
1285 while (!list_empty(&rbp->rbp_bufs)) {
1286 LASSERT (rbp->rbp_credits > 0);
1287
1288 rb = list_entry(rbp->rbp_bufs.next,
1289 lnet_rtrbuf_t, rb_list);
1290 list_del(&rb->rb_list);
1291 lnet_destroy_rtrbuf(rb, npages);
1292 nbuffers++;
1293 }
1294
1295 LASSERT (rbp->rbp_nbuffers == nbuffers);
1296 LASSERT (rbp->rbp_credits == nbuffers);
1297
1298 rbp->rbp_nbuffers = rbp->rbp_credits = 0;
1299}
1300
1301int
1302lnet_rtrpool_alloc_bufs(lnet_rtrbufpool_t *rbp, int nbufs, int cpt)
1303{
1304 lnet_rtrbuf_t *rb;
1305 int i;
1306
1307 if (rbp->rbp_nbuffers != 0) {
1308 LASSERT (rbp->rbp_nbuffers == nbufs);
1309 return 0;
1310 }
1311
1312 for (i = 0; i < nbufs; i++) {
1313 rb = lnet_new_rtrbuf(rbp, cpt);
1314
1315 if (rb == NULL) {
1316 CERROR("Failed to allocate %d router bufs of %d pages\n",
1317 nbufs, rbp->rbp_npages);
1318 return -ENOMEM;
1319 }
1320
1321 rbp->rbp_nbuffers++;
1322 rbp->rbp_credits++;
1323 rbp->rbp_mincredits++;
1324 list_add(&rb->rb_list, &rbp->rbp_bufs);
1325
1326 /* No allocation "under fire" */
1327 /* Otherwise we'd need code to schedule blocked msgs etc */
1328 LASSERT (!the_lnet.ln_routing);
1329 }
1330
1331 LASSERT (rbp->rbp_credits == nbufs);
1332 return 0;
1333}
1334
1335void
1336lnet_rtrpool_init(lnet_rtrbufpool_t *rbp, int npages)
1337{
1338 INIT_LIST_HEAD(&rbp->rbp_msgs);
1339 INIT_LIST_HEAD(&rbp->rbp_bufs);
1340
1341 rbp->rbp_npages = npages;
1342 rbp->rbp_credits = 0;
1343 rbp->rbp_mincredits = 0;
1344}
1345
1346void
1347lnet_rtrpools_free(void)
1348{
1349 lnet_rtrbufpool_t *rtrp;
1350 int i;
1351
1352 if (the_lnet.ln_rtrpools == NULL) /* uninitialized or freed */
1353 return;
1354
1355 cfs_percpt_for_each(rtrp, i, the_lnet.ln_rtrpools) {
1356 lnet_rtrpool_free_bufs(&rtrp[0]);
1357 lnet_rtrpool_free_bufs(&rtrp[1]);
1358 lnet_rtrpool_free_bufs(&rtrp[2]);
1359 }
1360
1361 cfs_percpt_free(the_lnet.ln_rtrpools);
1362 the_lnet.ln_rtrpools = NULL;
1363}
1364
1365static int
1366lnet_nrb_tiny_calculate(int npages)
1367{
1368 int nrbs = LNET_NRB_TINY;
1369
1370 if (tiny_router_buffers < 0) {
1371 LCONSOLE_ERROR_MSG(0x10c,
1372 "tiny_router_buffers=%d invalid when "
1373 "routing enabled\n", tiny_router_buffers);
1374 return -1;
1375 }
1376
1377 if (tiny_router_buffers > 0)
1378 nrbs = tiny_router_buffers;
1379
1380 nrbs /= LNET_CPT_NUMBER;
1381 return max(nrbs, LNET_NRB_TINY_MIN);
1382}
1383
1384static int
1385lnet_nrb_small_calculate(int npages)
1386{
1387 int nrbs = LNET_NRB_SMALL;
1388
1389 if (small_router_buffers < 0) {
1390 LCONSOLE_ERROR_MSG(0x10c,
1391 "small_router_buffers=%d invalid when "
1392 "routing enabled\n", small_router_buffers);
1393 return -1;
1394 }
1395
1396 if (small_router_buffers > 0)
1397 nrbs = small_router_buffers;
1398
1399 nrbs /= LNET_CPT_NUMBER;
1400 return max(nrbs, LNET_NRB_SMALL_MIN);
1401}
1402
1403static int
1404lnet_nrb_large_calculate(int npages)
1405{
1406 int nrbs = LNET_NRB_LARGE;
1407
1408 if (large_router_buffers < 0) {
1409 LCONSOLE_ERROR_MSG(0x10c,
1410 "large_router_buffers=%d invalid when "
1411 "routing enabled\n", large_router_buffers);
1412 return -1;
1413 }
1414
1415 if (large_router_buffers > 0)
1416 nrbs = large_router_buffers;
1417
1418 nrbs /= LNET_CPT_NUMBER;
1419 return max(nrbs, LNET_NRB_LARGE_MIN);
1420}
1421
1422int
1423lnet_rtrpools_alloc(int im_a_router)
1424{
1425 lnet_rtrbufpool_t *rtrp;
1426 int large_pages = (LNET_MTU + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1427 int small_pages = 1;
1428 int nrb_tiny;
1429 int nrb_small;
1430 int nrb_large;
1431 int rc;
1432 int i;
1433
1434 if (!strcmp(forwarding, "")) {
1435 /* not set either way */
1436 if (!im_a_router)
1437 return 0;
1438 } else if (!strcmp(forwarding, "disabled")) {
1439 /* explicitly disabled */
1440 return 0;
1441 } else if (!strcmp(forwarding, "enabled")) {
1442 /* explicitly enabled */
1443 } else {
1444 LCONSOLE_ERROR_MSG(0x10b, "'forwarding' not set to either "
1445 "'enabled' or 'disabled'\n");
1446 return -EINVAL;
1447 }
1448
1449 nrb_tiny = lnet_nrb_tiny_calculate(0);
1450 if (nrb_tiny < 0)
1451 return -EINVAL;
1452
1453 nrb_small = lnet_nrb_small_calculate(small_pages);
1454 if (nrb_small < 0)
1455 return -EINVAL;
1456
1457 nrb_large = lnet_nrb_large_calculate(large_pages);
1458 if (nrb_large < 0)
1459 return -EINVAL;
1460
1461 the_lnet.ln_rtrpools = cfs_percpt_alloc(lnet_cpt_table(),
1462 LNET_NRBPOOLS *
1463 sizeof(lnet_rtrbufpool_t));
1464 if (the_lnet.ln_rtrpools == NULL) {
1465 LCONSOLE_ERROR_MSG(0x10c,
1466 "Failed to initialize router buffe pool\n");
1467 return -ENOMEM;
1468 }
1469
1470 cfs_percpt_for_each(rtrp, i, the_lnet.ln_rtrpools) {
1471 lnet_rtrpool_init(&rtrp[0], 0);
1472 rc = lnet_rtrpool_alloc_bufs(&rtrp[0], nrb_tiny, i);
1473 if (rc != 0)
1474 goto failed;
1475
1476 lnet_rtrpool_init(&rtrp[1], small_pages);
1477 rc = lnet_rtrpool_alloc_bufs(&rtrp[1], nrb_small, i);
1478 if (rc != 0)
1479 goto failed;
1480
1481 lnet_rtrpool_init(&rtrp[2], large_pages);
1482 rc = lnet_rtrpool_alloc_bufs(&rtrp[2], nrb_large, i);
1483 if (rc != 0)
1484 goto failed;
1485 }
1486
1487 lnet_net_lock(LNET_LOCK_EX);
1488 the_lnet.ln_routing = 1;
1489 lnet_net_unlock(LNET_LOCK_EX);
1490
1491 return 0;
1492
1493 failed:
1494 lnet_rtrpools_free();
1495 return rc;
1496}
1497
1498int
1499lnet_notify(lnet_ni_t *ni, lnet_nid_t nid, int alive, cfs_time_t when)
1500{
1501 struct lnet_peer *lp = NULL;
1502 cfs_time_t now = cfs_time_current();
1503 int cpt = lnet_cpt_of_nid(nid);
1504
1505 LASSERT (!in_interrupt ());
1506
1507 CDEBUG (D_NET, "%s notifying %s: %s\n",
1508 (ni == NULL) ? "userspace" : libcfs_nid2str(ni->ni_nid),
1509 libcfs_nid2str(nid),
1510 alive ? "up" : "down");
1511
1512 if (ni != NULL &&
1513 LNET_NIDNET(ni->ni_nid) != LNET_NIDNET(nid)) {
1514 CWARN ("Ignoring notification of %s %s by %s (different net)\n",
1515 libcfs_nid2str(nid), alive ? "birth" : "death",
1516 libcfs_nid2str(ni->ni_nid));
1517 return -EINVAL;
1518 }
1519
1520 /* can't do predictions... */
1521 if (cfs_time_after(when, now)) {
1522 CWARN ("Ignoring prediction from %s of %s %s "
1523 "%ld seconds in the future\n",
1524 (ni == NULL) ? "userspace" : libcfs_nid2str(ni->ni_nid),
1525 libcfs_nid2str(nid), alive ? "up" : "down",
1526 cfs_duration_sec(cfs_time_sub(when, now)));
1527 return -EINVAL;
1528 }
1529
1530 if (ni != NULL && !alive && /* LND telling me she's down */
1531 !auto_down) { /* auto-down disabled */
1532 CDEBUG(D_NET, "Auto-down disabled\n");
1533 return 0;
1534 }
1535
1536 lnet_net_lock(cpt);
1537
1538 if (the_lnet.ln_shutdown) {
1539 lnet_net_unlock(cpt);
1540 return -ESHUTDOWN;
1541 }
1542
1543 lp = lnet_find_peer_locked(the_lnet.ln_peer_tables[cpt], nid);
1544 if (lp == NULL) {
1545 /* nid not found */
1546 lnet_net_unlock(cpt);
1547 CDEBUG(D_NET, "%s not found\n", libcfs_nid2str(nid));
1548 return 0;
1549 }
1550
1551 /* We can't fully trust LND on reporting exact peer last_alive
1552 * if he notifies us about dead peer. For example ksocklnd can
1553 * call us with when == _time_when_the_node_was_booted_ if
1554 * no connections were successfully established */
1555 if (ni != NULL && !alive && when < lp->lp_last_alive)
1556 when = lp->lp_last_alive;
1557
1558 lnet_notify_locked(lp, ni == NULL, alive, when);
1559
1560 lnet_ni_notify_locked(ni, lp);
1561
1562 lnet_peer_decref_locked(lp);
1563
1564 lnet_net_unlock(cpt);
1565 return 0;
1566}
1567EXPORT_SYMBOL(lnet_notify);
1568
1569void
1570lnet_get_tunables (void)
1571{
1572 return;
1573}
1574
1575#else
1576
1577int
1578lnet_notify (lnet_ni_t *ni, lnet_nid_t nid, int alive, cfs_time_t when)
1579{
1580 return -EOPNOTSUPP;
1581}
1582
1583void
1584lnet_router_checker (void)
1585{
1586 static time_t last = 0;
1587 static int running = 0;
1588
1589 time_t now = cfs_time_current_sec();
1590 int interval = now - last;
1591 int rc;
1592 __u64 version;
1593 lnet_peer_t *rtr;
1594
1595 /* It's no use to call me again within a sec - all intervals and
1596 * timeouts are measured in seconds */
1597 if (last != 0 && interval < 2)
1598 return;
1599
1600 if (last != 0 &&
1601 interval > MAX(live_router_check_interval,
1602 dead_router_check_interval))
1603 CNETERR("Checker(%d/%d) not called for %d seconds\n",
1604 live_router_check_interval, dead_router_check_interval,
1605 interval);
1606
1607 LASSERT(LNET_CPT_NUMBER == 1);
1608
1609 lnet_net_lock(0);
1610 LASSERT(!running); /* recursion check */
1611 running = 1;
1612 lnet_net_unlock(0);
1613
1614 last = now;
1615
1616 if (the_lnet.ln_rc_state == LNET_RC_STATE_STOPPING)
1617 lnet_prune_rc_data(0); /* unlink all rcd and nowait */
1618
1619 /* consume all pending events */
1620 while (1) {
1621 int i;
1622 lnet_event_t ev;
1623
1624 /* NB ln_rc_eqh must be the 1st in 'eventqs' otherwise the
1625 * recursion breaker in LNetEQPoll would fail */
1626 rc = LNetEQPoll(&the_lnet.ln_rc_eqh, 1, 0, &ev, &i);
1627 if (rc == 0) /* no event pending */
1628 break;
1629
1630 /* NB a lost SENT prevents me from pinging a router again */
1631 if (rc == -EOVERFLOW) {
1632 CERROR("Dropped an event!!!\n");
1633 abort();
1634 }
1635
1636 LASSERT (rc == 1);
1637
1638 lnet_router_checker_event(&ev);
1639 }
1640
1641 if (the_lnet.ln_rc_state == LNET_RC_STATE_STOPPING) {
1642 lnet_prune_rc_data(1); /* release rcd */
1643 the_lnet.ln_rc_state = LNET_RC_STATE_SHUTDOWN;
1644 running = 0;
1645 return;
1646 }
1647
1648 LASSERT (the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
1649
1650 lnet_net_lock(0);
1651
1652 version = the_lnet.ln_routers_version;
1653 list_for_each_entry (rtr, &the_lnet.ln_routers, lp_rtr_list) {
1654 lnet_ping_router_locked(rtr);
1655 LASSERT (version == the_lnet.ln_routers_version);
1656 }
1657
1658 lnet_net_unlock(0);
1659
1660 running = 0; /* lock only needed for the recursion check */
1661 return;
1662}
1663
1664/* NB lnet_peers_start_down depends on me,
1665 * so must be called before any peer creation */
1666void
1667lnet_get_tunables (void)
1668{
1669 char *s;
1670
1671 s = getenv("LNET_ROUTER_PING_TIMEOUT");
1672 if (s != NULL) router_ping_timeout = atoi(s);
1673
1674 s = getenv("LNET_LIVE_ROUTER_CHECK_INTERVAL");
1675 if (s != NULL) live_router_check_interval = atoi(s);
1676
1677 s = getenv("LNET_DEAD_ROUTER_CHECK_INTERVAL");
1678 if (s != NULL) dead_router_check_interval = atoi(s);
1679
1680 /* This replaces old lnd_notify mechanism */
1681 check_routers_before_use = 1;
1682 if (dead_router_check_interval <= 0)
1683 dead_router_check_interval = 30;
1684}
1685
1686void
1687lnet_rtrpools_free(void)
1688{
1689}
1690
1691int
1692lnet_rtrpools_alloc(int im_a_arouter)
1693{
1694 return 0;
1695}
1696
1697#endif
Linux kernel - Deliverables
This page took 0.17677 seconds and 5 git commands to generate.