1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #define DEBUG_SUBSYSTEM S_RPC
40 # include <liblustre.h>
42 # include <libcfs/libcfs.h>
44 # include <linux/kernel.h>
48 #include <obd_class.h>
49 #include <lustre_net.h>
50 #include <lustre_sec.h>
51 #include "ptlrpc_internal.h"
53 lnet_handle_eq_t ptlrpc_eq_h;
56 * Client's outgoing request callback
58 void request_out_callback(lnet_event_t *ev)
60 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
61 struct ptlrpc_request *req = cbid->cbid_arg;
64 LASSERT (ev->type == LNET_EVENT_SEND ||
65 ev->type == LNET_EVENT_UNLINK);
66 LASSERT (ev->unlinked);
68 DEBUG_REQ((ev->status == 0) ? D_NET : D_ERROR, req,
69 "type %d, status %d", ev->type, ev->status);
71 sptlrpc_request_out_callback(req);
73 if (ev->type == LNET_EVENT_UNLINK || ev->status != 0) {
75 /* Failed send: make it seem like the reply timed out, just
76 * like failing sends in client.c does currently... */
78 spin_lock(&req->rq_lock);
80 spin_unlock(&req->rq_lock);
82 ptlrpc_client_wake_req(req);
85 ptlrpc_req_finished(req);
91 * Client's incoming reply callback
93 void reply_in_callback(lnet_event_t *ev)
95 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
96 struct ptlrpc_request *req = cbid->cbid_arg;
99 DEBUG_REQ((ev->status == 0) ? D_NET : D_ERROR, req,
100 "type %d, status %d", ev->type, ev->status);
102 LASSERT (ev->type == LNET_EVENT_PUT || ev->type == LNET_EVENT_UNLINK);
103 LASSERT (ev->md.start == req->rq_repbuf);
104 LASSERT (ev->mlength <= req->rq_repbuf_len);
105 /* We've set LNET_MD_MANAGE_REMOTE for all outgoing requests
106 for adaptive timeouts' early reply. */
107 LASSERT((ev->md.options & LNET_MD_MANAGE_REMOTE) != 0);
109 spin_lock(&req->rq_lock);
111 req->rq_receiving_reply = 0;
114 req->rq_must_unlink = 0;
119 if (ev->type == LNET_EVENT_UNLINK) {
120 LASSERT(ev->unlinked);
121 DEBUG_REQ(D_RPCTRACE, req, "unlink");
125 if (ev->mlength < ev->rlength ) {
126 CDEBUG(D_RPCTRACE, "truncate req %p rpc %d - %d+%d\n", req,
127 req->rq_replen, ev->rlength, ev->offset);
128 req->rq_reply_truncate = 1;
130 req->rq_status = -EOVERFLOW;
131 req->rq_nob_received = ev->rlength + ev->offset;
135 if ((ev->offset == 0) &&
136 ((lustre_msghdr_get_flags(req->rq_reqmsg) & MSGHDR_AT_SUPPORT))) {
138 DEBUG_REQ(D_ADAPTTO, req,
139 "Early reply received: mlen=%u offset=%d replen=%d "
140 "replied=%d unlinked=%d", ev->mlength, ev->offset,
141 req->rq_replen, req->rq_replied, ev->unlinked);
143 req->rq_early_count++; /* number received, client side */
145 if (req->rq_replied) /* already got the real reply */
149 req->rq_reply_off = ev->offset;
150 req->rq_nob_received = ev->mlength;
151 /* And we're still receiving */
152 req->rq_receiving_reply = 1;
155 req->rq_rep_swab_mask = 0;
157 req->rq_reply_off = ev->offset;
158 req->rq_nob_received = ev->mlength;
159 /* LNetMDUnlink can't be called under the LNET_LOCK,
160 so we must unlink in ptlrpc_unregister_reply */
161 DEBUG_REQ(D_INFO, req,
162 "reply in flags=%x mlen=%u offset=%d replen=%d",
163 lustre_msg_get_flags(req->rq_reqmsg),
164 ev->mlength, ev->offset, req->rq_replen);
167 req->rq_import->imp_last_reply_time = cfs_time_current_sec();
170 /* NB don't unlock till after wakeup; req can disappear under us
171 * since we don't have our own ref */
172 ptlrpc_client_wake_req(req);
173 spin_unlock(&req->rq_lock);
178 * Client's bulk has been written/read
180 void client_bulk_callback (lnet_event_t *ev)
182 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
183 struct ptlrpc_bulk_desc *desc = cbid->cbid_arg;
186 LASSERT ((desc->bd_type == BULK_PUT_SINK &&
187 ev->type == LNET_EVENT_PUT) ||
188 (desc->bd_type == BULK_GET_SOURCE &&
189 ev->type == LNET_EVENT_GET) ||
190 ev->type == LNET_EVENT_UNLINK);
191 LASSERT (ev->unlinked);
193 CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
194 "event type %d, status %d, desc %p\n",
195 ev->type, ev->status, desc);
197 spin_lock(&desc->bd_lock);
199 LASSERT(desc->bd_network_rw);
200 desc->bd_network_rw = 0;
202 if (ev->type != LNET_EVENT_UNLINK && ev->status == 0) {
203 desc->bd_success = 1;
204 desc->bd_nob_transferred = ev->mlength;
205 desc->bd_sender = ev->sender;
208 /* release the encrypted pages for write */
209 if (desc->bd_req->rq_bulk_write)
210 sptlrpc_enc_pool_put_pages(desc);
212 /* NB don't unlock till after wakeup; desc can disappear under us
214 ptlrpc_client_wake_req(desc->bd_req);
216 spin_unlock(&desc->bd_lock);
221 * Server's incoming request callback
223 void request_in_callback(lnet_event_t *ev)
225 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
226 struct ptlrpc_request_buffer_desc *rqbd = cbid->cbid_arg;
227 struct ptlrpc_service *service = rqbd->rqbd_service;
228 struct ptlrpc_request *req;
231 LASSERT (ev->type == LNET_EVENT_PUT ||
232 ev->type == LNET_EVENT_UNLINK);
233 LASSERT ((char *)ev->md.start >= rqbd->rqbd_buffer);
234 LASSERT ((char *)ev->md.start + ev->offset + ev->mlength <=
235 rqbd->rqbd_buffer + service->srv_buf_size);
237 CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
238 "event type %d, status %d, service %s\n",
239 ev->type, ev->status, service->srv_name);
242 /* If this is the last request message to fit in the
243 * request buffer we can use the request object embedded in
244 * rqbd. Note that if we failed to allocate a request,
245 * we'd have to re-post the rqbd, which we can't do in this
247 req = &rqbd->rqbd_req;
248 memset(req, 0, sizeof (*req));
250 LASSERT (ev->type == LNET_EVENT_PUT);
251 if (ev->status != 0) {
252 /* We moaned above already... */
255 OBD_ALLOC_GFP(req, sizeof(*req), CFS_ALLOC_ATOMIC_TRY);
257 CERROR("Can't allocate incoming request descriptor: "
258 "Dropping %s RPC from %s\n",
260 libcfs_id2str(ev->initiator));
265 /* NB we ABSOLUTELY RELY on req being zeroed, so pointers are NULL,
266 * flags are reset and scalars are zero. We only set the message
267 * size to non-zero if this was a successful receive. */
268 req->rq_xid = ev->match_bits;
269 req->rq_reqbuf = ev->md.start + ev->offset;
270 if (ev->type == LNET_EVENT_PUT && ev->status == 0)
271 req->rq_reqdata_len = ev->mlength;
272 do_gettimeofday(&req->rq_arrival_time);
273 req->rq_peer = ev->initiator;
274 req->rq_self = ev->target.nid;
276 req->rq_phase = RQ_PHASE_NEW;
278 req->rq_uid = ev->uid;
280 spin_lock_init(&req->rq_lock);
281 CFS_INIT_LIST_HEAD(&req->rq_timed_list);
282 atomic_set(&req->rq_refcount, 1);
283 if (ev->type == LNET_EVENT_PUT)
284 CDEBUG(D_RPCTRACE, "incoming req@%p x"LPU64" msgsize %u\n",
285 req, req->rq_xid, ev->mlength);
287 CDEBUG(D_RPCTRACE, "peer: %s\n", libcfs_id2str(req->rq_peer));
289 spin_lock(&service->srv_lock);
291 req->rq_history_seq = service->srv_request_seq++;
292 list_add_tail(&req->rq_history_list, &service->srv_request_history);
295 service->srv_nrqbd_receiving--;
296 CDEBUG(D_INFO, "Buffer complete: %d buffers still posted\n",
297 service->srv_nrqbd_receiving);
299 /* Normally, don't complain about 0 buffers posted; LNET won't
300 * drop incoming reqs since we set the portal lazy */
301 if (test_req_buffer_pressure &&
302 ev->type != LNET_EVENT_UNLINK &&
303 service->srv_nrqbd_receiving == 0)
304 CWARN("All %s request buffers busy\n",
307 /* req takes over the network's ref on rqbd */
309 /* req takes a ref on rqbd */
310 rqbd->rqbd_refcount++;
313 list_add_tail(&req->rq_list, &service->srv_req_in_queue);
314 service->srv_n_queued_reqs++;
316 /* NB everything can disappear under us once the request
317 * has been queued and we unlock, so do the wake now... */
318 cfs_waitq_signal(&service->srv_waitq);
320 spin_unlock(&service->srv_lock);
325 * Server's outgoing reply callback
327 void reply_out_callback(lnet_event_t *ev)
329 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
330 struct ptlrpc_reply_state *rs = cbid->cbid_arg;
331 struct ptlrpc_service *svc = rs->rs_service;
334 LASSERT (ev->type == LNET_EVENT_SEND ||
335 ev->type == LNET_EVENT_ACK ||
336 ev->type == LNET_EVENT_UNLINK);
338 if (!rs->rs_difficult) {
339 /* 'Easy' replies have no further processing so I drop the
340 * net's ref on 'rs' */
341 LASSERT (ev->unlinked);
342 ptlrpc_rs_decref(rs);
343 atomic_dec (&svc->srv_outstanding_replies);
348 LASSERT (rs->rs_on_net);
351 /* Last network callback. The net's ref on 'rs' stays put
352 * until ptlrpc_handle_rs() is done with it */
353 spin_lock(&svc->srv_lock);
354 spin_lock(&rs->rs_lock);
356 if (!rs->rs_no_ack ||
357 rs->rs_transno <= rs->rs_export->exp_obd->obd_last_committed)
358 ptlrpc_schedule_difficult_reply (rs);
359 spin_unlock(&rs->rs_lock);
360 spin_unlock(&svc->srv_lock);
367 * Server's bulk completion callback
369 void server_bulk_callback (lnet_event_t *ev)
371 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
372 struct ptlrpc_bulk_desc *desc = cbid->cbid_arg;
375 LASSERT (ev->type == LNET_EVENT_SEND ||
376 ev->type == LNET_EVENT_UNLINK ||
377 (desc->bd_type == BULK_PUT_SOURCE &&
378 ev->type == LNET_EVENT_ACK) ||
379 (desc->bd_type == BULK_GET_SINK &&
380 ev->type == LNET_EVENT_REPLY));
382 CDEBUG((ev->status == 0) ? D_NET : D_ERROR,
383 "event type %d, status %d, desc %p\n",
384 ev->type, ev->status, desc);
386 spin_lock(&desc->bd_lock);
388 if ((ev->type == LNET_EVENT_ACK ||
389 ev->type == LNET_EVENT_REPLY) &&
391 /* We heard back from the peer, so even if we get this
392 * before the SENT event (oh yes we can), we know we
393 * read/wrote the peer buffer and how much... */
394 desc->bd_success = 1;
395 desc->bd_nob_transferred = ev->mlength;
396 desc->bd_sender = ev->sender;
400 /* This is the last callback no matter what... */
401 desc->bd_network_rw = 0;
402 cfs_waitq_signal(&desc->bd_waitq);
405 spin_unlock(&desc->bd_lock);
409 static void ptlrpc_master_callback(lnet_event_t *ev)
411 struct ptlrpc_cb_id *cbid = ev->md.user_ptr;
412 void (*callback)(lnet_event_t *ev) = cbid->cbid_fn;
414 /* Honestly, it's best to find out early. */
415 LASSERT (cbid->cbid_arg != LP_POISON);
416 LASSERT (callback == request_out_callback ||
417 callback == reply_in_callback ||
418 callback == client_bulk_callback ||
419 callback == request_in_callback ||
420 callback == reply_out_callback ||
421 callback == server_bulk_callback);
426 int ptlrpc_uuid_to_peer (struct obd_uuid *uuid,
427 lnet_process_id_t *peer, lnet_nid_t *self)
430 __u32 best_order = 0;
433 int portals_compatibility;
439 portals_compatibility = LNetCtl(IOC_LIBCFS_PORTALS_COMPATIBILITY, NULL);
441 peer->pid = LUSTRE_SRV_LNET_PID;
443 /* Choose the matching UUID that's closest */
444 while (lustre_uuid_to_peer(uuid->uuid, &dst_nid, count++) == 0) {
445 dist = LNetDist(dst_nid, &src_nid, &order);
449 if (dist == 0) { /* local! use loopback LND */
450 peer->nid = *self = LNET_MKNID(LNET_MKNET(LOLND, 0), 0);
457 (dist == best_dist && order < best_order)) {
461 if (portals_compatibility > 1) {
462 /* Strong portals compatibility: Zero the nid's
463 * NET, so if I'm reading new config logs, or
464 * getting configured by (new) lconf I can
465 * still talk to old servers. */
466 dst_nid = LNET_MKNID(0, LNET_NIDADDR(dst_nid));
467 src_nid = LNET_MKNID(0, LNET_NIDADDR(src_nid));
475 CDEBUG(D_NET,"%s->%s\n", uuid->uuid, libcfs_id2str(*peer));
477 CERROR("No NID found for %s\n", uuid->uuid);
481 void ptlrpc_ni_fini(void)
484 struct l_wait_info lwi;
488 /* Wait for the event queue to become idle since there may still be
489 * messages in flight with pending events (i.e. the fire-and-forget
490 * messages == client requests and "non-difficult" server
493 for (retries = 0;; retries++) {
494 rc = LNetEQFree(ptlrpc_eq_h);
505 CWARN("Event queue still busy\n");
508 cfs_waitq_init(&waitq);
509 lwi = LWI_TIMEOUT(cfs_time_seconds(2), NULL, NULL);
510 l_wait_event(waitq, 0, &lwi);
517 lnet_pid_t ptl_get_pid(void)
524 pid = LUSTRE_SRV_LNET_PID;
529 int ptlrpc_ni_init(void)
535 CDEBUG(D_NET, "My pid is: %x\n", pid);
537 /* We're not passing any limits yet... */
538 rc = LNetNIInit(pid);
540 CDEBUG (D_NET, "Can't init network interface: %d\n", rc);
544 /* CAVEAT EMPTOR: how we process portals events is _radically_
545 * different depending on... */
547 /* kernel portals calls our master callback when events are added to
548 * the event queue. In fact lustre never pulls events off this queue,
549 * so it's only sized for some debug history. */
550 rc = LNetEQAlloc(1024, ptlrpc_master_callback, &ptlrpc_eq_h);
552 /* liblustre calls the master callback when it removes events from the
553 * event queue. The event queue has to be big enough not to drop
555 rc = LNetEQAlloc(10240, LNET_EQ_HANDLER_NONE, &ptlrpc_eq_h);
560 CERROR ("Failed to allocate event queue: %d\n", rc);
567 CFS_LIST_HEAD(liblustre_wait_callbacks);
568 CFS_LIST_HEAD(liblustre_idle_callbacks);
569 void *liblustre_services_callback;
572 liblustre_register_waitidle_callback (struct list_head *callback_list,
574 int (*fn)(void *arg), void *arg)
576 struct liblustre_wait_callback *llwc;
578 OBD_ALLOC(llwc, sizeof(*llwc));
579 LASSERT (llwc != NULL);
581 llwc->llwc_name = name;
583 llwc->llwc_arg = arg;
584 list_add_tail(&llwc->llwc_list, callback_list);
590 liblustre_deregister_waitidle_callback (void *opaque)
592 struct liblustre_wait_callback *llwc = opaque;
594 list_del(&llwc->llwc_list);
595 OBD_FREE(llwc, sizeof(*llwc));
599 liblustre_register_wait_callback (const char *name,
600 int (*fn)(void *arg), void *arg)
602 return liblustre_register_waitidle_callback(&liblustre_wait_callbacks,
607 liblustre_deregister_wait_callback (void *opaque)
609 liblustre_deregister_waitidle_callback(opaque);
613 liblustre_register_idle_callback (const char *name,
614 int (*fn)(void *arg), void *arg)
616 return liblustre_register_waitidle_callback(&liblustre_idle_callbacks,
621 liblustre_deregister_idle_callback (void *opaque)
623 liblustre_deregister_waitidle_callback(opaque);
627 liblustre_check_events (int timeout)
634 rc = LNetEQPoll(&ptlrpc_eq_h, 1, timeout * 1000, &ev, &i);
638 LASSERT (rc == -EOVERFLOW || rc == 1);
640 /* liblustre: no asynch callback so we can't afford to miss any
642 if (rc == -EOVERFLOW) {
643 CERROR ("Dropped an event!!!\n");
647 ptlrpc_master_callback (&ev);
651 int liblustre_waiting = 0;
654 liblustre_wait_event (int timeout)
656 struct list_head *tmp;
657 struct liblustre_wait_callback *llwc;
658 int found_something = 0;
660 /* single threaded recursion check... */
661 liblustre_waiting = 1;
664 /* Deal with all pending events */
665 while (liblustre_check_events(0))
668 /* Give all registered callbacks a bite at the cherry */
669 list_for_each(tmp, &liblustre_wait_callbacks) {
670 llwc = list_entry(tmp, struct liblustre_wait_callback,
673 if (llwc->llwc_fn(llwc->llwc_arg))
677 if (found_something || timeout == 0)
680 /* Nothing so far, but I'm allowed to block... */
681 found_something = liblustre_check_events(timeout);
682 if (!found_something) /* still nothing */
683 break; /* I timed out */
686 liblustre_waiting = 0;
688 return found_something;
692 liblustre_wait_idle(void)
694 static int recursed = 0;
696 struct list_head *tmp;
697 struct liblustre_wait_callback *llwc;
704 liblustre_wait_event(0);
708 list_for_each(tmp, &liblustre_idle_callbacks) {
709 llwc = list_entry(tmp, struct liblustre_wait_callback,
712 if (!llwc->llwc_fn(llwc->llwc_arg)) {
723 #endif /* __KERNEL__ */
725 int ptlrpc_init_portals(void)
727 int rc = ptlrpc_ni_init();
730 CERROR("network initialisation failed\n");
734 liblustre_services_callback =
735 liblustre_register_wait_callback("liblustre_check_services",
736 &liblustre_check_services,
738 init_completion_module(liblustre_wait_event);
740 rc = ptlrpcd_addref();
744 CERROR("rpcd initialisation failed\n");
746 liblustre_deregister_wait_callback(liblustre_services_callback);
752 void ptlrpc_exit_portals(void)
755 liblustre_deregister_wait_callback(liblustre_services_callback);