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 "../../include/linux/lnet/lib-lnet.h"
26
27 #define LNET_NRB_TINY_MIN 512 /* min value for each CPT */
28 #define LNET_NRB_TINY (LNET_NRB_TINY_MIN * 4)
29 #define LNET_NRB_SMALL_MIN 4096 /* min value for each CPT */
30 #define LNET_NRB_SMALL (LNET_NRB_SMALL_MIN * 4)
31 #define LNET_NRB_LARGE_MIN 256 /* min value for each CPT */
32 #define LNET_NRB_LARGE (LNET_NRB_LARGE_MIN * 4)
33
34 static char *forwarding = "";
35 module_param(forwarding, charp, 0444);
36 MODULE_PARM_DESC(forwarding, "Explicitly enable/disable forwarding between networks");
37
38 static int tiny_router_buffers;
39 module_param(tiny_router_buffers, int, 0444);
40 MODULE_PARM_DESC(tiny_router_buffers, "# of 0 payload messages to buffer in the router");
41 static int small_router_buffers;
42 module_param(small_router_buffers, int, 0444);
43 MODULE_PARM_DESC(small_router_buffers, "# of small (1 page) messages to buffer in the router");
44 static int large_router_buffers;
45 module_param(large_router_buffers, int, 0444);
46 MODULE_PARM_DESC(large_router_buffers, "# of large messages to buffer in the router");
47 static int peer_buffer_credits;
48 module_param(peer_buffer_credits, int, 0444);
49 MODULE_PARM_DESC(peer_buffer_credits, "# router buffer credits per peer");
50
51 static int auto_down = 1;
52 module_param(auto_down, int, 0444);
53 MODULE_PARM_DESC(auto_down, "Automatically mark peers down on comms error");
54
55 int
lnet_peer_buffer_credits(lnet_ni_t * ni)56 lnet_peer_buffer_credits(lnet_ni_t *ni)
57 {
58 /* NI option overrides LNet default */
59 if (ni->ni_peerrtrcredits > 0)
60 return ni->ni_peerrtrcredits;
61 if (peer_buffer_credits > 0)
62 return peer_buffer_credits;
63
64 /* As an approximation, allow this peer the same number of router
65 * buffers as it is allowed outstanding sends */
66 return ni->ni_peertxcredits;
67 }
68
69 /* forward ref's */
70 static int lnet_router_checker(void *);
71
72 static int check_routers_before_use;
73 module_param(check_routers_before_use, int, 0444);
74 MODULE_PARM_DESC(check_routers_before_use, "Assume routers are down and ping them before use");
75
76 int avoid_asym_router_failure = 1;
77 module_param(avoid_asym_router_failure, int, 0644);
78 MODULE_PARM_DESC(avoid_asym_router_failure, "Avoid asymmetrical router failures (0 to disable)");
79
80 static int dead_router_check_interval = 60;
81 module_param(dead_router_check_interval, int, 0644);
82 MODULE_PARM_DESC(dead_router_check_interval, "Seconds between dead router health checks (<= 0 to disable)");
83
84 static int live_router_check_interval = 60;
85 module_param(live_router_check_interval, int, 0644);
86 MODULE_PARM_DESC(live_router_check_interval, "Seconds between live router health checks (<= 0 to disable)");
87
88 static int router_ping_timeout = 50;
89 module_param(router_ping_timeout, int, 0644);
90 MODULE_PARM_DESC(router_ping_timeout, "Seconds to wait for the reply to a router health query");
91
92 int
lnet_peers_start_down(void)93 lnet_peers_start_down(void)
94 {
95 return check_routers_before_use;
96 }
97
98 void
lnet_notify_locked(lnet_peer_t * lp,int notifylnd,int alive,unsigned long when)99 lnet_notify_locked(lnet_peer_t *lp, int notifylnd, int alive,
100 unsigned long when)
101 {
102 if (time_before(when, lp->lp_timestamp)) { /* out of date information */
103 CDEBUG(D_NET, "Out of date\n");
104 return;
105 }
106
107 lp->lp_timestamp = when; /* update timestamp */
108 lp->lp_ping_deadline = 0; /* disable ping timeout */
109
110 if (lp->lp_alive_count != 0 && /* got old news */
111 (!lp->lp_alive) == (!alive)) { /* new date for old news */
112 CDEBUG(D_NET, "Old news\n");
113 return;
114 }
115
116 /* Flag that notification is outstanding */
117
118 lp->lp_alive_count++;
119 lp->lp_alive = !(!alive); /* 1 bit! */
120 lp->lp_notify = 1;
121 lp->lp_notifylnd |= notifylnd;
122 if (lp->lp_alive)
123 lp->lp_ping_feats = LNET_PING_FEAT_INVAL; /* reset */
124
125 CDEBUG(D_NET, "set %s %d\n", libcfs_nid2str(lp->lp_nid), alive);
126 }
127
128 static void
lnet_ni_notify_locked(lnet_ni_t * ni,lnet_peer_t * lp)129 lnet_ni_notify_locked(lnet_ni_t *ni, lnet_peer_t *lp)
130 {
131 int alive;
132 int notifylnd;
133
134 /* Notify only in 1 thread at any time to ensure ordered notification.
135 * NB individual events can be missed; the only guarantee is that you
136 * always get the most recent news */
137
138 if (lp->lp_notifying || ni == NULL)
139 return;
140
141 lp->lp_notifying = 1;
142
143 while (lp->lp_notify) {
144 alive = lp->lp_alive;
145 notifylnd = lp->lp_notifylnd;
146
147 lp->lp_notifylnd = 0;
148 lp->lp_notify = 0;
149
150 if (notifylnd && ni->ni_lnd->lnd_notify != NULL) {
151 lnet_net_unlock(lp->lp_cpt);
152
153 /* A new notification could happen now; I'll handle it
154 * when control returns to me */
155
156 (ni->ni_lnd->lnd_notify)(ni, lp->lp_nid, alive);
157
158 lnet_net_lock(lp->lp_cpt);
159 }
160 }
161
162 lp->lp_notifying = 0;
163 }
164
165 static void
lnet_rtr_addref_locked(lnet_peer_t * lp)166 lnet_rtr_addref_locked(lnet_peer_t *lp)
167 {
168 LASSERT(lp->lp_refcount > 0);
169 LASSERT(lp->lp_rtr_refcount >= 0);
170
171 /* lnet_net_lock must be exclusively locked */
172 lp->lp_rtr_refcount++;
173 if (lp->lp_rtr_refcount == 1) {
174 struct list_head *pos;
175
176 /* a simple insertion sort */
177 list_for_each_prev(pos, &the_lnet.ln_routers) {
178 lnet_peer_t *rtr = list_entry(pos, lnet_peer_t,
179 lp_rtr_list);
180
181 if (rtr->lp_nid < lp->lp_nid)
182 break;
183 }
184
185 list_add(&lp->lp_rtr_list, pos);
186 /* addref for the_lnet.ln_routers */
187 lnet_peer_addref_locked(lp);
188 the_lnet.ln_routers_version++;
189 }
190 }
191
192 static void
lnet_rtr_decref_locked(lnet_peer_t * lp)193 lnet_rtr_decref_locked(lnet_peer_t *lp)
194 {
195 LASSERT(lp->lp_refcount > 0);
196 LASSERT(lp->lp_rtr_refcount > 0);
197
198 /* lnet_net_lock must be exclusively locked */
199 lp->lp_rtr_refcount--;
200 if (lp->lp_rtr_refcount == 0) {
201 LASSERT(list_empty(&lp->lp_routes));
202
203 if (lp->lp_rcd != NULL) {
204 list_add(&lp->lp_rcd->rcd_list,
205 &the_lnet.ln_rcd_deathrow);
206 lp->lp_rcd = NULL;
207 }
208
209 list_del(&lp->lp_rtr_list);
210 /* decref for the_lnet.ln_routers */
211 lnet_peer_decref_locked(lp);
212 the_lnet.ln_routers_version++;
213 }
214 }
215
216 lnet_remotenet_t *
lnet_find_net_locked(__u32 net)217 lnet_find_net_locked(__u32 net)
218 {
219 lnet_remotenet_t *rnet;
220 struct list_head *tmp;
221 struct list_head *rn_list;
222
223 LASSERT(!the_lnet.ln_shutdown);
224
225 rn_list = lnet_net2rnethash(net);
226 list_for_each(tmp, rn_list) {
227 rnet = list_entry(tmp, lnet_remotenet_t, lrn_list);
228
229 if (rnet->lrn_net == net)
230 return rnet;
231 }
232 return NULL;
233 }
234
lnet_shuffle_seed(void)235 static void lnet_shuffle_seed(void)
236 {
237 static int seeded;
238 __u32 lnd_type, seed[2];
239 struct timespec64 ts;
240 lnet_ni_t *ni;
241 struct list_head *tmp;
242
243 if (seeded)
244 return;
245
246 cfs_get_random_bytes(seed, sizeof(seed));
247
248 /* Nodes with small feet have little entropy
249 * the NID for this node gives the most entropy in the low bits */
250 list_for_each(tmp, &the_lnet.ln_nis) {
251 ni = list_entry(tmp, lnet_ni_t, ni_list);
252 lnd_type = LNET_NETTYP(LNET_NIDNET(ni->ni_nid));
253
254 if (lnd_type != LOLND)
255 seed[0] ^= (LNET_NIDADDR(ni->ni_nid) | lnd_type);
256 }
257
258 ktime_get_ts64(&ts);
259 cfs_srand(ts.tv_sec ^ seed[0], ts.tv_nsec ^ seed[1]);
260 seeded = 1;
261 }
262
263 /* NB expects LNET_LOCK held */
264 static void
lnet_add_route_to_rnet(lnet_remotenet_t * rnet,lnet_route_t * route)265 lnet_add_route_to_rnet(lnet_remotenet_t *rnet, lnet_route_t *route)
266 {
267 unsigned int len = 0;
268 unsigned int offset = 0;
269 struct list_head *e;
270
271 lnet_shuffle_seed();
272
273 list_for_each(e, &rnet->lrn_routes) {
274 len++;
275 }
276
277 /* len+1 positions to add a new entry, also prevents division by 0 */
278 offset = cfs_rand() % (len + 1);
279 list_for_each(e, &rnet->lrn_routes) {
280 if (offset == 0)
281 break;
282 offset--;
283 }
284 list_add(&route->lr_list, e);
285 list_add(&route->lr_gwlist, &route->lr_gateway->lp_routes);
286
287 the_lnet.ln_remote_nets_version++;
288 lnet_rtr_addref_locked(route->lr_gateway);
289 }
290
291 int
lnet_add_route(__u32 net,unsigned int hops,lnet_nid_t gateway,unsigned int priority)292 lnet_add_route(__u32 net, unsigned int hops, lnet_nid_t gateway,
293 unsigned int priority)
294 {
295 struct list_head *e;
296 lnet_remotenet_t *rnet;
297 lnet_remotenet_t *rnet2;
298 lnet_route_t *route;
299 lnet_ni_t *ni;
300 int add_route;
301 int rc;
302
303 CDEBUG(D_NET, "Add route: net %s hops %u priority %u gw %s\n",
304 libcfs_net2str(net), hops, priority, libcfs_nid2str(gateway));
305
306 if (gateway == LNET_NID_ANY ||
307 LNET_NETTYP(LNET_NIDNET(gateway)) == LOLND ||
308 net == LNET_NIDNET(LNET_NID_ANY) ||
309 LNET_NETTYP(net) == LOLND ||
310 LNET_NIDNET(gateway) == net ||
311 hops < 1 || hops > 255)
312 return -EINVAL;
313
314 if (lnet_islocalnet(net)) /* it's a local network */
315 return 0; /* ignore the route entry */
316
317 /* Assume net, route, all new */
318 LIBCFS_ALLOC(route, sizeof(*route));
319 LIBCFS_ALLOC(rnet, sizeof(*rnet));
320 if (route == NULL || rnet == NULL) {
321 CERROR("Out of memory creating route %s %d %s\n",
322 libcfs_net2str(net), hops, libcfs_nid2str(gateway));
323 if (route != NULL)
324 LIBCFS_FREE(route, sizeof(*route));
325 if (rnet != NULL)
326 LIBCFS_FREE(rnet, sizeof(*rnet));
327 return -ENOMEM;
328 }
329
330 INIT_LIST_HEAD(&rnet->lrn_routes);
331 rnet->lrn_net = net;
332 route->lr_hops = hops;
333 route->lr_net = net;
334 route->lr_priority = priority;
335
336 lnet_net_lock(LNET_LOCK_EX);
337
338 rc = lnet_nid2peer_locked(&route->lr_gateway, gateway, LNET_LOCK_EX);
339 if (rc != 0) {
340 lnet_net_unlock(LNET_LOCK_EX);
341
342 LIBCFS_FREE(route, sizeof(*route));
343 LIBCFS_FREE(rnet, sizeof(*rnet));
344
345 if (rc == -EHOSTUNREACH) /* gateway is not on a local net */
346 return 0; /* ignore the route entry */
347 CERROR("Error %d creating route %s %d %s\n", rc,
348 libcfs_net2str(net), hops,
349 libcfs_nid2str(gateway));
350
351 return rc;
352 }
353
354 LASSERT(!the_lnet.ln_shutdown);
355
356 rnet2 = lnet_find_net_locked(net);
357 if (rnet2 == NULL) {
358 /* new network */
359 list_add_tail(&rnet->lrn_list, lnet_net2rnethash(net));
360 rnet2 = rnet;
361 }
362
363 /* Search for a duplicate route (it's a NOOP if it is) */
364 add_route = 1;
365 list_for_each(e, &rnet2->lrn_routes) {
366 lnet_route_t *route2 = list_entry(e, lnet_route_t, lr_list);
367
368 if (route2->lr_gateway == route->lr_gateway) {
369 add_route = 0;
370 break;
371 }
372
373 /* our lookups must be true */
374 LASSERT(route2->lr_gateway->lp_nid != gateway);
375 }
376
377 if (add_route) {
378 lnet_peer_addref_locked(route->lr_gateway); /* +1 for notify */
379 lnet_add_route_to_rnet(rnet2, route);
380
381 ni = route->lr_gateway->lp_ni;
382 lnet_net_unlock(LNET_LOCK_EX);
383
384 /* XXX Assume alive */
385 if (ni->ni_lnd->lnd_notify != NULL)
386 (ni->ni_lnd->lnd_notify)(ni, gateway, 1);
387
388 lnet_net_lock(LNET_LOCK_EX);
389 }
390
391 /* -1 for notify or !add_route */
392 lnet_peer_decref_locked(route->lr_gateway);
393 lnet_net_unlock(LNET_LOCK_EX);
394
395 if (!add_route)
396 LIBCFS_FREE(route, sizeof(*route));
397
398 if (rnet != rnet2)
399 LIBCFS_FREE(rnet, sizeof(*rnet));
400
401 return 0;
402 }
403
404 int
lnet_check_routes(void)405 lnet_check_routes(void)
406 {
407 lnet_remotenet_t *rnet;
408 lnet_route_t *route;
409 lnet_route_t *route2;
410 struct list_head *e1;
411 struct list_head *e2;
412 int cpt;
413 struct list_head *rn_list;
414 int i;
415
416 cpt = lnet_net_lock_current();
417
418 for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) {
419 rn_list = &the_lnet.ln_remote_nets_hash[i];
420 list_for_each(e1, rn_list) {
421 rnet = list_entry(e1, lnet_remotenet_t, lrn_list);
422
423 route2 = NULL;
424 list_for_each(e2, &rnet->lrn_routes) {
425 lnet_nid_t nid1;
426 lnet_nid_t nid2;
427 int net;
428
429 route = list_entry(e2, lnet_route_t,
430 lr_list);
431
432 if (route2 == NULL) {
433 route2 = route;
434 continue;
435 }
436
437 if (route->lr_gateway->lp_ni ==
438 route2->lr_gateway->lp_ni)
439 continue;
440
441 nid1 = route->lr_gateway->lp_nid;
442 nid2 = route2->lr_gateway->lp_nid;
443 net = rnet->lrn_net;
444
445 lnet_net_unlock(cpt);
446
447 CERROR("Routes to %s via %s and %s not supported\n",
448 libcfs_net2str(net),
449 libcfs_nid2str(nid1),
450 libcfs_nid2str(nid2));
451 return -EINVAL;
452 }
453 }
454 }
455
456 lnet_net_unlock(cpt);
457 return 0;
458 }
459
460 int
lnet_del_route(__u32 net,lnet_nid_t gw_nid)461 lnet_del_route(__u32 net, lnet_nid_t gw_nid)
462 {
463 struct lnet_peer *gateway;
464 lnet_remotenet_t *rnet;
465 lnet_route_t *route;
466 struct list_head *e1;
467 struct list_head *e2;
468 int rc = -ENOENT;
469 struct list_head *rn_list;
470 int idx = 0;
471
472 CDEBUG(D_NET, "Del route: net %s : gw %s\n",
473 libcfs_net2str(net), libcfs_nid2str(gw_nid));
474
475 /* NB Caller may specify either all routes via the given gateway
476 * or a specific route entry actual NIDs) */
477
478 lnet_net_lock(LNET_LOCK_EX);
479 if (net == LNET_NIDNET(LNET_NID_ANY))
480 rn_list = &the_lnet.ln_remote_nets_hash[0];
481 else
482 rn_list = lnet_net2rnethash(net);
483
484 again:
485 list_for_each(e1, rn_list) {
486 rnet = list_entry(e1, lnet_remotenet_t, lrn_list);
487
488 if (!(net == LNET_NIDNET(LNET_NID_ANY) ||
489 net == rnet->lrn_net))
490 continue;
491
492 list_for_each(e2, &rnet->lrn_routes) {
493 route = list_entry(e2, lnet_route_t, lr_list);
494
495 gateway = route->lr_gateway;
496 if (!(gw_nid == LNET_NID_ANY ||
497 gw_nid == gateway->lp_nid))
498 continue;
499
500 list_del(&route->lr_list);
501 list_del(&route->lr_gwlist);
502 the_lnet.ln_remote_nets_version++;
503
504 if (list_empty(&rnet->lrn_routes))
505 list_del(&rnet->lrn_list);
506 else
507 rnet = NULL;
508
509 lnet_rtr_decref_locked(gateway);
510 lnet_peer_decref_locked(gateway);
511
512 lnet_net_unlock(LNET_LOCK_EX);
513
514 LIBCFS_FREE(route, sizeof(*route));
515
516 if (rnet != NULL)
517 LIBCFS_FREE(rnet, sizeof(*rnet));
518
519 rc = 0;
520 lnet_net_lock(LNET_LOCK_EX);
521 goto again;
522 }
523 }
524
525 if (net == LNET_NIDNET(LNET_NID_ANY) &&
526 ++idx < LNET_REMOTE_NETS_HASH_SIZE) {
527 rn_list = &the_lnet.ln_remote_nets_hash[idx];
528 goto again;
529 }
530 lnet_net_unlock(LNET_LOCK_EX);
531
532 return rc;
533 }
534
535 void
lnet_destroy_routes(void)536 lnet_destroy_routes(void)
537 {
538 lnet_del_route(LNET_NIDNET(LNET_NID_ANY), LNET_NID_ANY);
539 }
540
541 int
lnet_get_route(int idx,__u32 * net,__u32 * hops,lnet_nid_t * gateway,__u32 * alive,__u32 * priority)542 lnet_get_route(int idx, __u32 *net, __u32 *hops,
543 lnet_nid_t *gateway, __u32 *alive, __u32 *priority)
544 {
545 struct list_head *e1;
546 struct list_head *e2;
547 lnet_remotenet_t *rnet;
548 lnet_route_t *route;
549 int cpt;
550 int i;
551 struct list_head *rn_list;
552
553 cpt = lnet_net_lock_current();
554
555 for (i = 0; i < LNET_REMOTE_NETS_HASH_SIZE; i++) {
556 rn_list = &the_lnet.ln_remote_nets_hash[i];
557 list_for_each(e1, rn_list) {
558 rnet = list_entry(e1, lnet_remotenet_t, lrn_list);
559
560 list_for_each(e2, &rnet->lrn_routes) {
561 route = list_entry(e2, lnet_route_t,
562 lr_list);
563
564 if (idx-- == 0) {
565 *net = rnet->lrn_net;
566 *hops = route->lr_hops;
567 *priority = route->lr_priority;
568 *gateway = route->lr_gateway->lp_nid;
569 *alive = route->lr_gateway->lp_alive;
570 lnet_net_unlock(cpt);
571 return 0;
572 }
573 }
574 }
575 }
576
577 lnet_net_unlock(cpt);
578 return -ENOENT;
579 }
580
581 void
lnet_swap_pinginfo(lnet_ping_info_t * info)582 lnet_swap_pinginfo(lnet_ping_info_t *info)
583 {
584 int i;
585 lnet_ni_status_t *stat;
586
587 __swab32s(&info->pi_magic);
588 __swab32s(&info->pi_features);
589 __swab32s(&info->pi_pid);
590 __swab32s(&info->pi_nnis);
591 for (i = 0; i < info->pi_nnis && i < LNET_MAX_RTR_NIS; i++) {
592 stat = &info->pi_ni[i];
593 __swab64s(&stat->ns_nid);
594 __swab32s(&stat->ns_status);
595 }
596 }
597
598 /**
599 * parse router-checker pinginfo, record number of down NIs for remote
600 * networks on that router.
601 */
602 static void
lnet_parse_rc_info(lnet_rc_data_t * rcd)603 lnet_parse_rc_info(lnet_rc_data_t *rcd)
604 {
605 lnet_ping_info_t *info = rcd->rcd_pinginfo;
606 struct lnet_peer *gw = rcd->rcd_gateway;
607 lnet_route_t *rtr;
608
609 if (!gw->lp_alive)
610 return;
611
612 if (info->pi_magic == __swab32(LNET_PROTO_PING_MAGIC))
613 lnet_swap_pinginfo(info);
614
615 /* NB always racing with network! */
616 if (info->pi_magic != LNET_PROTO_PING_MAGIC) {
617 CDEBUG(D_NET, "%s: Unexpected magic %08x\n",
618 libcfs_nid2str(gw->lp_nid), info->pi_magic);
619 gw->lp_ping_feats = LNET_PING_FEAT_INVAL;
620 return;
621 }
622
623 gw->lp_ping_feats = info->pi_features;
624 if ((gw->lp_ping_feats & LNET_PING_FEAT_MASK) == 0) {
625 CDEBUG(D_NET, "%s: Unexpected features 0x%x\n",
626 libcfs_nid2str(gw->lp_nid), gw->lp_ping_feats);
627 return; /* nothing I can understand */
628 }
629
630 if ((gw->lp_ping_feats & LNET_PING_FEAT_NI_STATUS) == 0)
631 return; /* can't carry NI status info */
632
633 list_for_each_entry(rtr, &gw->lp_routes, lr_gwlist) {
634 int ptl_status = LNET_NI_STATUS_INVALID;
635 int down = 0;
636 int up = 0;
637 int i;
638
639 for (i = 0; i < info->pi_nnis && i < LNET_MAX_RTR_NIS; i++) {
640 lnet_ni_status_t *stat = &info->pi_ni[i];
641 lnet_nid_t nid = stat->ns_nid;
642
643 if (nid == LNET_NID_ANY) {
644 CDEBUG(D_NET, "%s: unexpected LNET_NID_ANY\n",
645 libcfs_nid2str(gw->lp_nid));
646 gw->lp_ping_feats = LNET_PING_FEAT_INVAL;
647 return;
648 }
649
650 if (LNET_NETTYP(LNET_NIDNET(nid)) == LOLND)
651 continue;
652
653 if (stat->ns_status == LNET_NI_STATUS_DOWN) {
654 if (LNET_NETTYP(LNET_NIDNET(nid)) != PTLLND)
655 down++;
656 else if (ptl_status != LNET_NI_STATUS_UP)
657 ptl_status = LNET_NI_STATUS_DOWN;
658 continue;
659 }
660
661 if (stat->ns_status == LNET_NI_STATUS_UP) {
662 if (LNET_NIDNET(nid) == rtr->lr_net) {
663 up = 1;
664 break;
665 }
666 /* ptl NIs are considered down only when
667 * they're all down */
668 if (LNET_NETTYP(LNET_NIDNET(nid)) == PTLLND)
669 ptl_status = LNET_NI_STATUS_UP;
670 continue;
671 }
672
673 CDEBUG(D_NET, "%s: Unexpected status 0x%x\n",
674 libcfs_nid2str(gw->lp_nid), stat->ns_status);
675 gw->lp_ping_feats = LNET_PING_FEAT_INVAL;
676 return;
677 }
678
679 if (up) { /* ignore downed NIs if NI for dest network is up */
680 rtr->lr_downis = 0;
681 continue;
682 }
683 rtr->lr_downis = down + (ptl_status == LNET_NI_STATUS_DOWN);
684 }
685 }
686
687 static void
lnet_router_checker_event(lnet_event_t * event)688 lnet_router_checker_event(lnet_event_t *event)
689 {
690 lnet_rc_data_t *rcd = event->md.user_ptr;
691 struct lnet_peer *lp;
692
693 LASSERT(rcd != NULL);
694
695 if (event->unlinked) {
696 LNetInvalidateHandle(&rcd->rcd_mdh);
697 return;
698 }
699
700 LASSERT(event->type == LNET_EVENT_SEND ||
701 event->type == LNET_EVENT_REPLY);
702
703 lp = rcd->rcd_gateway;
704 LASSERT(lp != NULL);
705
706 /* NB: it's called with holding lnet_res_lock, we have a few
707 * places need to hold both locks at the same time, please take
708 * care of lock ordering */
709 lnet_net_lock(lp->lp_cpt);
710 if (!lnet_isrouter(lp) || lp->lp_rcd != rcd) {
711 /* ignore if no longer a router or rcd is replaced */
712 goto out;
713 }
714
715 if (event->type == LNET_EVENT_SEND) {
716 lp->lp_ping_notsent = 0;
717 if (event->status == 0)
718 goto out;
719 }
720
721 /* LNET_EVENT_REPLY */
722 /* A successful REPLY means the router is up. If _any_ comms
723 * to the router fail I assume it's down (this will happen if
724 * we ping alive routers to try to detect router death before
725 * apps get burned). */
726
727 lnet_notify_locked(lp, 1, (event->status == 0), cfs_time_current());
728 /* The router checker will wake up very shortly and do the
729 * actual notification.
730 * XXX If 'lp' stops being a router before then, it will still
731 * have the notification pending!!! */
732
733 if (avoid_asym_router_failure && event->status == 0)
734 lnet_parse_rc_info(rcd);
735
736 out:
737 lnet_net_unlock(lp->lp_cpt);
738 }
739
740 static void
lnet_wait_known_routerstate(void)741 lnet_wait_known_routerstate(void)
742 {
743 lnet_peer_t *rtr;
744 struct list_head *entry;
745 int all_known;
746
747 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
748
749 for (;;) {
750 int cpt = lnet_net_lock_current();
751
752 all_known = 1;
753 list_for_each(entry, &the_lnet.ln_routers) {
754 rtr = list_entry(entry, lnet_peer_t, lp_rtr_list);
755
756 if (rtr->lp_alive_count == 0) {
757 all_known = 0;
758 break;
759 }
760 }
761
762 lnet_net_unlock(cpt);
763
764 if (all_known)
765 return;
766
767 set_current_state(TASK_UNINTERRUPTIBLE);
768 schedule_timeout(cfs_time_seconds(1));
769 }
770 }
771
772 void
lnet_router_ni_update_locked(lnet_peer_t * gw,__u32 net)773 lnet_router_ni_update_locked(lnet_peer_t *gw, __u32 net)
774 {
775 lnet_route_t *rte;
776
777 if ((gw->lp_ping_feats & LNET_PING_FEAT_NI_STATUS) != 0) {
778 list_for_each_entry(rte, &gw->lp_routes, lr_gwlist) {
779 if (rte->lr_net == net) {
780 rte->lr_downis = 0;
781 break;
782 }
783 }
784 }
785 }
786
787 static void
lnet_update_ni_status_locked(void)788 lnet_update_ni_status_locked(void)
789 {
790 lnet_ni_t *ni;
791 time64_t now;
792 int timeout;
793
794 LASSERT(the_lnet.ln_routing);
795
796 timeout = router_ping_timeout +
797 max(live_router_check_interval, dead_router_check_interval);
798
799 now = ktime_get_real_seconds();
800 list_for_each_entry(ni, &the_lnet.ln_nis, ni_list) {
801 if (ni->ni_lnd->lnd_type == LOLND)
802 continue;
803
804 if (now < ni->ni_last_alive + timeout)
805 continue;
806
807 lnet_ni_lock(ni);
808 /* re-check with lock */
809 if (now < ni->ni_last_alive + timeout) {
810 lnet_ni_unlock(ni);
811 continue;
812 }
813
814 LASSERT(ni->ni_status != NULL);
815
816 if (ni->ni_status->ns_status != LNET_NI_STATUS_DOWN) {
817 CDEBUG(D_NET, "NI(%s:%d) status changed to down\n",
818 libcfs_nid2str(ni->ni_nid), timeout);
819 /* NB: so far, this is the only place to set
820 * NI status to "down" */
821 ni->ni_status->ns_status = LNET_NI_STATUS_DOWN;
822 }
823 lnet_ni_unlock(ni);
824 }
825 }
826
827 static void
lnet_destroy_rc_data(lnet_rc_data_t * rcd)828 lnet_destroy_rc_data(lnet_rc_data_t *rcd)
829 {
830 LASSERT(list_empty(&rcd->rcd_list));
831 /* detached from network */
832 LASSERT(LNetHandleIsInvalid(rcd->rcd_mdh));
833
834 if (rcd->rcd_gateway != NULL) {
835 int cpt = rcd->rcd_gateway->lp_cpt;
836
837 lnet_net_lock(cpt);
838 lnet_peer_decref_locked(rcd->rcd_gateway);
839 lnet_net_unlock(cpt);
840 }
841
842 if (rcd->rcd_pinginfo != NULL)
843 LIBCFS_FREE(rcd->rcd_pinginfo, LNET_PINGINFO_SIZE);
844
845 LIBCFS_FREE(rcd, sizeof(*rcd));
846 }
847
848 static lnet_rc_data_t *
lnet_create_rc_data_locked(lnet_peer_t * gateway)849 lnet_create_rc_data_locked(lnet_peer_t *gateway)
850 {
851 lnet_rc_data_t *rcd = NULL;
852 lnet_ping_info_t *pi;
853 int rc;
854 int i;
855
856 lnet_net_unlock(gateway->lp_cpt);
857
858 LIBCFS_ALLOC(rcd, sizeof(*rcd));
859 if (rcd == NULL)
860 goto out;
861
862 LNetInvalidateHandle(&rcd->rcd_mdh);
863 INIT_LIST_HEAD(&rcd->rcd_list);
864
865 LIBCFS_ALLOC(pi, LNET_PINGINFO_SIZE);
866 if (pi == NULL)
867 goto out;
868
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
917 static int
lnet_router_check_interval(lnet_peer_t * rtr)918 lnet_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
930 static void
lnet_ping_router_locked(lnet_peer_t * rtr)931 lnet_ping_router_locked(lnet_peer_t *rtr)
932 {
933 lnet_rc_data_t *rcd = NULL;
934 unsigned long 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 alive_count %d lp_ping_timestamp %lu\n",
963 libcfs_nid2str(rtr->lp_nid), secs,
964 rtr->lp_ping_deadline, rtr->lp_ping_notsent,
965 rtr->lp_alive, rtr->lp_alive_count, rtr->lp_ping_timestamp);
966
967 if (secs != 0 && !rtr->lp_ping_notsent &&
968 cfs_time_after(now, cfs_time_add(rtr->lp_ping_timestamp,
969 cfs_time_seconds(secs)))) {
970 int rc;
971 lnet_process_id_t id;
972 lnet_handle_md_t mdh;
973
974 id.nid = rtr->lp_nid;
975 id.pid = LUSTRE_SRV_LNET_PID;
976 CDEBUG(D_NET, "Check: %s\n", libcfs_id2str(id));
977
978 rtr->lp_ping_notsent = 1;
979 rtr->lp_ping_timestamp = now;
980
981 mdh = rcd->rcd_mdh;
982
983 if (rtr->lp_ping_deadline == 0) {
984 rtr->lp_ping_deadline =
985 cfs_time_shift(router_ping_timeout);
986 }
987
988 lnet_net_unlock(rtr->lp_cpt);
989
990 rc = LNetGet(LNET_NID_ANY, mdh, id, LNET_RESERVED_PORTAL,
991 LNET_PROTO_PING_MATCHBITS, 0);
992
993 lnet_net_lock(rtr->lp_cpt);
994 if (rc != 0)
995 rtr->lp_ping_notsent = 0; /* no event pending */
996 }
997
998 lnet_peer_decref_locked(rtr);
999 }
1000
1001 int
lnet_router_checker_start(void)1002 lnet_router_checker_start(void)
1003 {
1004 int rc;
1005 int eqsz;
1006
1007 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_SHUTDOWN);
1008
1009 if (check_routers_before_use &&
1010 dead_router_check_interval <= 0) {
1011 LCONSOLE_ERROR_MSG(0x10a, "'dead_router_check_interval' must be set if 'check_routers_before_use' is set\n");
1012 return -EINVAL;
1013 }
1014
1015 if (!the_lnet.ln_routing &&
1016 live_router_check_interval <= 0 &&
1017 dead_router_check_interval <= 0)
1018 return 0;
1019
1020 sema_init(&the_lnet.ln_rc_signal, 0);
1021 /* EQ size doesn't matter; the callback is guaranteed to get every
1022 * event */
1023 eqsz = 0;
1024 rc = LNetEQAlloc(eqsz, lnet_router_checker_event,
1025 &the_lnet.ln_rc_eqh);
1026 if (rc != 0) {
1027 CERROR("Can't allocate EQ(%d): %d\n", eqsz, rc);
1028 return -ENOMEM;
1029 }
1030
1031 the_lnet.ln_rc_state = LNET_RC_STATE_RUNNING;
1032 rc = PTR_ERR(kthread_run(lnet_router_checker,
1033 NULL, "router_checker"));
1034 if (IS_ERR_VALUE(rc)) {
1035 CERROR("Can't start router checker thread: %d\n", rc);
1036 /* block until event callback signals exit */
1037 down(&the_lnet.ln_rc_signal);
1038 rc = LNetEQFree(the_lnet.ln_rc_eqh);
1039 LASSERT(rc == 0);
1040 the_lnet.ln_rc_state = LNET_RC_STATE_SHUTDOWN;
1041 return -ENOMEM;
1042 }
1043
1044 if (check_routers_before_use) {
1045 /* Note that a helpful side-effect of pinging all known routers
1046 * at startup is that it makes them drop stale connections they
1047 * may have to a previous instance of me. */
1048 lnet_wait_known_routerstate();
1049 }
1050
1051 return 0;
1052 }
1053
1054 void
lnet_router_checker_stop(void)1055 lnet_router_checker_stop(void)
1056 {
1057 int rc;
1058
1059 if (the_lnet.ln_rc_state == LNET_RC_STATE_SHUTDOWN)
1060 return;
1061
1062 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
1063 the_lnet.ln_rc_state = LNET_RC_STATE_STOPPING;
1064
1065 /* block until event callback signals exit */
1066 down(&the_lnet.ln_rc_signal);
1067 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_SHUTDOWN);
1068
1069 rc = LNetEQFree(the_lnet.ln_rc_eqh);
1070 LASSERT(rc == 0);
1071 }
1072
1073 static void
lnet_prune_rc_data(int wait_unlink)1074 lnet_prune_rc_data(int wait_unlink)
1075 {
1076 lnet_rc_data_t *rcd;
1077 lnet_rc_data_t *tmp;
1078 lnet_peer_t *lp;
1079 struct list_head head;
1080 int i = 2;
1081
1082 if (likely(the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING &&
1083 list_empty(&the_lnet.ln_rcd_deathrow) &&
1084 list_empty(&the_lnet.ln_rcd_zombie)))
1085 return;
1086
1087 INIT_LIST_HEAD(&head);
1088
1089 lnet_net_lock(LNET_LOCK_EX);
1090
1091 if (the_lnet.ln_rc_state != LNET_RC_STATE_RUNNING) {
1092 /* router checker is stopping, prune all */
1093 list_for_each_entry(lp, &the_lnet.ln_routers,
1094 lp_rtr_list) {
1095 if (lp->lp_rcd == NULL)
1096 continue;
1097
1098 LASSERT(list_empty(&lp->lp_rcd->rcd_list));
1099 list_add(&lp->lp_rcd->rcd_list,
1100 &the_lnet.ln_rcd_deathrow);
1101 lp->lp_rcd = NULL;
1102 }
1103 }
1104
1105 /* unlink all RCDs on deathrow list */
1106 list_splice_init(&the_lnet.ln_rcd_deathrow, &head);
1107
1108 if (!list_empty(&head)) {
1109 lnet_net_unlock(LNET_LOCK_EX);
1110
1111 list_for_each_entry(rcd, &head, rcd_list)
1112 LNetMDUnlink(rcd->rcd_mdh);
1113
1114 lnet_net_lock(LNET_LOCK_EX);
1115 }
1116
1117 list_splice_init(&head, &the_lnet.ln_rcd_zombie);
1118
1119 /* release all zombie RCDs */
1120 while (!list_empty(&the_lnet.ln_rcd_zombie)) {
1121 list_for_each_entry_safe(rcd, tmp, &the_lnet.ln_rcd_zombie,
1122 rcd_list) {
1123 if (LNetHandleIsInvalid(rcd->rcd_mdh))
1124 list_move(&rcd->rcd_list, &head);
1125 }
1126
1127 wait_unlink = wait_unlink &&
1128 !list_empty(&the_lnet.ln_rcd_zombie);
1129
1130 lnet_net_unlock(LNET_LOCK_EX);
1131
1132 while (!list_empty(&head)) {
1133 rcd = list_entry(head.next,
1134 lnet_rc_data_t, rcd_list);
1135 list_del_init(&rcd->rcd_list);
1136 lnet_destroy_rc_data(rcd);
1137 }
1138
1139 if (!wait_unlink)
1140 return;
1141
1142 i++;
1143 CDEBUG(((i & (-i)) == i) ? D_WARNING : D_NET,
1144 "Waiting for rc buffers to unlink\n");
1145 set_current_state(TASK_UNINTERRUPTIBLE);
1146 schedule_timeout(cfs_time_seconds(1) / 4);
1147
1148 lnet_net_lock(LNET_LOCK_EX);
1149 }
1150
1151 lnet_net_unlock(LNET_LOCK_EX);
1152 }
1153
1154 static int
lnet_router_checker(void * arg)1155 lnet_router_checker(void *arg)
1156 {
1157 lnet_peer_t *rtr;
1158 struct list_head *entry;
1159
1160 cfs_block_allsigs();
1161
1162 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING);
1163
1164 while (the_lnet.ln_rc_state == LNET_RC_STATE_RUNNING) {
1165 __u64 version;
1166 int cpt;
1167 int cpt2;
1168
1169 cpt = lnet_net_lock_current();
1170 rescan:
1171 version = the_lnet.ln_routers_version;
1172
1173 list_for_each(entry, &the_lnet.ln_routers) {
1174 rtr = list_entry(entry, lnet_peer_t, lp_rtr_list);
1175
1176 cpt2 = lnet_cpt_of_nid_locked(rtr->lp_nid);
1177 if (cpt != cpt2) {
1178 lnet_net_unlock(cpt);
1179 cpt = cpt2;
1180 lnet_net_lock(cpt);
1181 /* the routers list has changed */
1182 if (version != the_lnet.ln_routers_version)
1183 goto rescan;
1184 }
1185
1186 lnet_ping_router_locked(rtr);
1187
1188 /* NB dropped lock */
1189 if (version != the_lnet.ln_routers_version) {
1190 /* the routers list has changed */
1191 goto rescan;
1192 }
1193 }
1194
1195 if (the_lnet.ln_routing)
1196 lnet_update_ni_status_locked();
1197
1198 lnet_net_unlock(cpt);
1199
1200 lnet_prune_rc_data(0); /* don't wait for UNLINK */
1201
1202 /* Call schedule_timeout() here always adds 1 to load average
1203 * because kernel counts # active tasks as nr_running
1204 * + nr_uninterruptible. */
1205 set_current_state(TASK_INTERRUPTIBLE);
1206 schedule_timeout(cfs_time_seconds(1));
1207 }
1208
1209 LASSERT(the_lnet.ln_rc_state == LNET_RC_STATE_STOPPING);
1210
1211 lnet_prune_rc_data(1); /* wait for UNLINK */
1212
1213 the_lnet.ln_rc_state = LNET_RC_STATE_SHUTDOWN;
1214 up(&the_lnet.ln_rc_signal);
1215 /* The unlink event callback will signal final completion */
1216 return 0;
1217 }
1218
1219 static void
lnet_destroy_rtrbuf(lnet_rtrbuf_t * rb,int npages)1220 lnet_destroy_rtrbuf(lnet_rtrbuf_t *rb, int npages)
1221 {
1222 int sz = offsetof(lnet_rtrbuf_t, rb_kiov[npages]);
1223
1224 while (--npages >= 0)
1225 __free_page(rb->rb_kiov[npages].kiov_page);
1226
1227 LIBCFS_FREE(rb, sz);
1228 }
1229
1230 static lnet_rtrbuf_t *
lnet_new_rtrbuf(lnet_rtrbufpool_t * rbp,int cpt)1231 lnet_new_rtrbuf(lnet_rtrbufpool_t *rbp, int cpt)
1232 {
1233 int npages = rbp->rbp_npages;
1234 int sz = offsetof(lnet_rtrbuf_t, rb_kiov[npages]);
1235 struct page *page;
1236 lnet_rtrbuf_t *rb;
1237 int i;
1238
1239 LIBCFS_CPT_ALLOC(rb, lnet_cpt_table(), cpt, sz);
1240 if (rb == NULL)
1241 return NULL;
1242
1243 rb->rb_pool = rbp;
1244
1245 for (i = 0; i < npages; i++) {
1246 page = alloc_pages_node(
1247 cfs_cpt_spread_node(lnet_cpt_table(), cpt),
1248 GFP_KERNEL | __GFP_ZERO, 0);
1249 if (page == NULL) {
1250 while (--i >= 0)
1251 __free_page(rb->rb_kiov[i].kiov_page);
1252
1253 LIBCFS_FREE(rb, sz);
1254 return NULL;
1255 }
1256
1257 rb->rb_kiov[i].kiov_len = PAGE_CACHE_SIZE;
1258 rb->rb_kiov[i].kiov_offset = 0;
1259 rb->rb_kiov[i].kiov_page = page;
1260 }
1261
1262 return rb;
1263 }
1264
1265 static void
lnet_rtrpool_free_bufs(lnet_rtrbufpool_t * rbp)1266 lnet_rtrpool_free_bufs(lnet_rtrbufpool_t *rbp)
1267 {
1268 int npages = rbp->rbp_npages;
1269 int nbuffers = 0;
1270 lnet_rtrbuf_t *rb;
1271
1272 if (rbp->rbp_nbuffers == 0) /* not initialized or already freed */
1273 return;
1274
1275 LASSERT(list_empty(&rbp->rbp_msgs));
1276 LASSERT(rbp->rbp_credits == rbp->rbp_nbuffers);
1277
1278 while (!list_empty(&rbp->rbp_bufs)) {
1279 LASSERT(rbp->rbp_credits > 0);
1280
1281 rb = list_entry(rbp->rbp_bufs.next,
1282 lnet_rtrbuf_t, rb_list);
1283 list_del(&rb->rb_list);
1284 lnet_destroy_rtrbuf(rb, npages);
1285 nbuffers++;
1286 }
1287
1288 LASSERT(rbp->rbp_nbuffers == nbuffers);
1289 LASSERT(rbp->rbp_credits == nbuffers);
1290
1291 rbp->rbp_nbuffers = rbp->rbp_credits = 0;
1292 }
1293
1294 static int
lnet_rtrpool_alloc_bufs(lnet_rtrbufpool_t * rbp,int nbufs,int cpt)1295 lnet_rtrpool_alloc_bufs(lnet_rtrbufpool_t *rbp, int nbufs, int cpt)
1296 {
1297 lnet_rtrbuf_t *rb;
1298 int i;
1299
1300 if (rbp->rbp_nbuffers != 0) {
1301 LASSERT(rbp->rbp_nbuffers == nbufs);
1302 return 0;
1303 }
1304
1305 for (i = 0; i < nbufs; i++) {
1306 rb = lnet_new_rtrbuf(rbp, cpt);
1307
1308 if (rb == NULL) {
1309 CERROR("Failed to allocate %d router bufs of %d pages\n",
1310 nbufs, rbp->rbp_npages);
1311 return -ENOMEM;
1312 }
1313
1314 rbp->rbp_nbuffers++;
1315 rbp->rbp_credits++;
1316 rbp->rbp_mincredits++;
1317 list_add(&rb->rb_list, &rbp->rbp_bufs);
1318
1319 /* No allocation "under fire" */
1320 /* Otherwise we'd need code to schedule blocked msgs etc */
1321 LASSERT(!the_lnet.ln_routing);
1322 }
1323
1324 LASSERT(rbp->rbp_credits == nbufs);
1325 return 0;
1326 }
1327
1328 static void
lnet_rtrpool_init(lnet_rtrbufpool_t * rbp,int npages)1329 lnet_rtrpool_init(lnet_rtrbufpool_t *rbp, int npages)
1330 {
1331 INIT_LIST_HEAD(&rbp->rbp_msgs);
1332 INIT_LIST_HEAD(&rbp->rbp_bufs);
1333
1334 rbp->rbp_npages = npages;
1335 rbp->rbp_credits = 0;
1336 rbp->rbp_mincredits = 0;
1337 }
1338
1339 void
lnet_rtrpools_free(void)1340 lnet_rtrpools_free(void)
1341 {
1342 lnet_rtrbufpool_t *rtrp;
1343 int i;
1344
1345 if (the_lnet.ln_rtrpools == NULL) /* uninitialized or freed */
1346 return;
1347
1348 cfs_percpt_for_each(rtrp, i, the_lnet.ln_rtrpools) {
1349 lnet_rtrpool_free_bufs(&rtrp[0]);
1350 lnet_rtrpool_free_bufs(&rtrp[1]);
1351 lnet_rtrpool_free_bufs(&rtrp[2]);
1352 }
1353
1354 cfs_percpt_free(the_lnet.ln_rtrpools);
1355 the_lnet.ln_rtrpools = NULL;
1356 }
1357
1358 static int
lnet_nrb_tiny_calculate(int npages)1359 lnet_nrb_tiny_calculate(int npages)
1360 {
1361 int nrbs = LNET_NRB_TINY;
1362
1363 if (tiny_router_buffers < 0) {
1364 LCONSOLE_ERROR_MSG(0x10c,
1365 "tiny_router_buffers=%d invalid when routing enabled\n",
1366 tiny_router_buffers);
1367 return -1;
1368 }
1369
1370 if (tiny_router_buffers > 0)
1371 nrbs = tiny_router_buffers;
1372
1373 nrbs /= LNET_CPT_NUMBER;
1374 return max(nrbs, LNET_NRB_TINY_MIN);
1375 }
1376
1377 static int
lnet_nrb_small_calculate(int npages)1378 lnet_nrb_small_calculate(int npages)
1379 {
1380 int nrbs = LNET_NRB_SMALL;
1381
1382 if (small_router_buffers < 0) {
1383 LCONSOLE_ERROR_MSG(0x10c,
1384 "small_router_buffers=%d invalid when routing enabled\n",
1385 small_router_buffers);
1386 return -1;
1387 }
1388
1389 if (small_router_buffers > 0)
1390 nrbs = small_router_buffers;
1391
1392 nrbs /= LNET_CPT_NUMBER;
1393 return max(nrbs, LNET_NRB_SMALL_MIN);
1394 }
1395
1396 static int
lnet_nrb_large_calculate(int npages)1397 lnet_nrb_large_calculate(int npages)
1398 {
1399 int nrbs = LNET_NRB_LARGE;
1400
1401 if (large_router_buffers < 0) {
1402 LCONSOLE_ERROR_MSG(0x10c,
1403 "large_router_buffers=%d invalid when routing enabled\n",
1404 large_router_buffers);
1405 return -1;
1406 }
1407
1408 if (large_router_buffers > 0)
1409 nrbs = large_router_buffers;
1410
1411 nrbs /= LNET_CPT_NUMBER;
1412 return max(nrbs, LNET_NRB_LARGE_MIN);
1413 }
1414
1415 int
lnet_rtrpools_alloc(int im_a_router)1416 lnet_rtrpools_alloc(int im_a_router)
1417 {
1418 lnet_rtrbufpool_t *rtrp;
1419 int large_pages;
1420 int small_pages = 1;
1421 int nrb_tiny;
1422 int nrb_small;
1423 int nrb_large;
1424 int rc;
1425 int i;
1426
1427 large_pages = (LNET_MTU + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1428
1429 if (!strcmp(forwarding, "")) {
1430 /* not set either way */
1431 if (!im_a_router)
1432 return 0;
1433 } else if (!strcmp(forwarding, "disabled")) {
1434 /* explicitly disabled */
1435 return 0;
1436 } else if (!strcmp(forwarding, "enabled")) {
1437 /* explicitly enabled */
1438 } else {
1439 LCONSOLE_ERROR_MSG(0x10b, "'forwarding' not set to either 'enabled' or 'disabled'\n");
1440 return -EINVAL;
1441 }
1442
1443 nrb_tiny = lnet_nrb_tiny_calculate(0);
1444 if (nrb_tiny < 0)
1445 return -EINVAL;
1446
1447 nrb_small = lnet_nrb_small_calculate(small_pages);
1448 if (nrb_small < 0)
1449 return -EINVAL;
1450
1451 nrb_large = lnet_nrb_large_calculate(large_pages);
1452 if (nrb_large < 0)
1453 return -EINVAL;
1454
1455 the_lnet.ln_rtrpools = cfs_percpt_alloc(lnet_cpt_table(),
1456 LNET_NRBPOOLS *
1457 sizeof(lnet_rtrbufpool_t));
1458 if (the_lnet.ln_rtrpools == NULL) {
1459 LCONSOLE_ERROR_MSG(0x10c,
1460 "Failed to initialize router buffe pool\n");
1461 return -ENOMEM;
1462 }
1463
1464 cfs_percpt_for_each(rtrp, i, the_lnet.ln_rtrpools) {
1465 lnet_rtrpool_init(&rtrp[0], 0);
1466 rc = lnet_rtrpool_alloc_bufs(&rtrp[0], nrb_tiny, i);
1467 if (rc != 0)
1468 goto failed;
1469
1470 lnet_rtrpool_init(&rtrp[1], small_pages);
1471 rc = lnet_rtrpool_alloc_bufs(&rtrp[1], nrb_small, i);
1472 if (rc != 0)
1473 goto failed;
1474
1475 lnet_rtrpool_init(&rtrp[2], large_pages);
1476 rc = lnet_rtrpool_alloc_bufs(&rtrp[2], nrb_large, i);
1477 if (rc != 0)
1478 goto failed;
1479 }
1480
1481 lnet_net_lock(LNET_LOCK_EX);
1482 the_lnet.ln_routing = 1;
1483 lnet_net_unlock(LNET_LOCK_EX);
1484
1485 return 0;
1486
1487 failed:
1488 lnet_rtrpools_free();
1489 return rc;
1490 }
1491
1492 int
lnet_notify(lnet_ni_t * ni,lnet_nid_t nid,int alive,unsigned long when)1493 lnet_notify(lnet_ni_t *ni, lnet_nid_t nid, int alive, unsigned long when)
1494 {
1495 struct lnet_peer *lp = NULL;
1496 unsigned long now = cfs_time_current();
1497 int cpt = lnet_cpt_of_nid(nid);
1498
1499 LASSERT(!in_interrupt());
1500
1501 CDEBUG(D_NET, "%s notifying %s: %s\n",
1502 (ni == NULL) ? "userspace" : libcfs_nid2str(ni->ni_nid),
1503 libcfs_nid2str(nid),
1504 alive ? "up" : "down");
1505
1506 if (ni != NULL &&
1507 LNET_NIDNET(ni->ni_nid) != LNET_NIDNET(nid)) {
1508 CWARN("Ignoring notification of %s %s by %s (different net)\n",
1509 libcfs_nid2str(nid), alive ? "birth" : "death",
1510 libcfs_nid2str(ni->ni_nid));
1511 return -EINVAL;
1512 }
1513
1514 /* can't do predictions... */
1515 if (cfs_time_after(when, now)) {
1516 CWARN("Ignoring prediction from %s of %s %s %ld seconds in the future\n",
1517 (ni == NULL) ? "userspace" : libcfs_nid2str(ni->ni_nid),
1518 libcfs_nid2str(nid), alive ? "up" : "down",
1519 cfs_duration_sec(cfs_time_sub(when, now)));
1520 return -EINVAL;
1521 }
1522
1523 if (ni != NULL && !alive && /* LND telling me she's down */
1524 !auto_down) { /* auto-down disabled */
1525 CDEBUG(D_NET, "Auto-down disabled\n");
1526 return 0;
1527 }
1528
1529 lnet_net_lock(cpt);
1530
1531 if (the_lnet.ln_shutdown) {
1532 lnet_net_unlock(cpt);
1533 return -ESHUTDOWN;
1534 }
1535
1536 lp = lnet_find_peer_locked(the_lnet.ln_peer_tables[cpt], nid);
1537 if (lp == NULL) {
1538 /* nid not found */
1539 lnet_net_unlock(cpt);
1540 CDEBUG(D_NET, "%s not found\n", libcfs_nid2str(nid));
1541 return 0;
1542 }
1543
1544 /* We can't fully trust LND on reporting exact peer last_alive
1545 * if he notifies us about dead peer. For example ksocklnd can
1546 * call us with when == _time_when_the_node_was_booted_ if
1547 * no connections were successfully established */
1548 if (ni != NULL && !alive && when < lp->lp_last_alive)
1549 when = lp->lp_last_alive;
1550
1551 lnet_notify_locked(lp, ni == NULL, alive, when);
1552
1553 lnet_ni_notify_locked(ni, lp);
1554
1555 lnet_peer_decref_locked(lp);
1556
1557 lnet_net_unlock(cpt);
1558 return 0;
1559 }
1560 EXPORT_SYMBOL(lnet_notify);
1561