4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Intel Corporation.
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
38 #include <obd_support.h>
39 #include <lustre_net.h>
40 #include <lustre_lib.h>
42 #include <obd_class.h>
43 #include "ptlrpc_internal.h"
46 * Helper function. Sends \a len bytes from \a base at offset \a offset
47 * over \a conn connection to portal \a portal.
48 * Returns 0 on success or error code.
50 static int ptl_send_buf(lnet_handle_md_t *mdh, void *base, int len,
51 lnet_ack_req_t ack, struct ptlrpc_cb_id *cbid,
52 struct ptlrpc_connection *conn, int portal, __u64 xid,
59 LASSERT(conn != NULL);
60 CDEBUG(D_INFO, "conn=%p id %s\n", conn, libcfs_id2str(conn->c_peer));
63 md.threshold = (ack == LNET_ACK_REQ) ? 2 : 1;
64 md.options = PTLRPC_MD_OPTIONS;
66 md.eq_handle = ptlrpc_eq_h;
68 if (unlikely(ack == LNET_ACK_REQ &&
69 OBD_FAIL_CHECK_ORSET(OBD_FAIL_PTLRPC_ACK,
71 /* don't ask for the ack to simulate failing client */
75 rc = LNetMDBind(md, LNET_UNLINK, mdh);
76 if (unlikely(rc != 0)) {
77 CERROR("LNetMDBind failed: %d\n", rc);
78 LASSERT(rc == -ENOMEM);
82 CDEBUG(D_NET, "Sending %d bytes to portal %d, xid "LPD64", offset %u\n",
83 len, portal, xid, offset);
85 rc = LNetPut(conn->c_self, *mdh, ack,
86 conn->c_peer, portal, xid, offset, 0);
87 if (unlikely(rc != 0)) {
89 /* We're going to get an UNLINK event when I unlink below,
90 * which will complete just like any other failed send, so
91 * I fall through and return success here! */
92 CERROR("LNetPut(%s, %d, "LPD64") failed: %d\n",
93 libcfs_id2str(conn->c_peer), portal, xid, rc);
94 rc2 = LNetMDUnlink(*mdh);
95 LASSERTF(rc2 == 0, "rc2 = %d\n", rc2);
101 static void mdunlink_iterate_helper(lnet_handle_md_t *bd_mds, int count)
105 for (i = 0; i < count; i++)
106 LNetMDUnlink(bd_mds[i]);
111 * Register bulk at the sender for later transfer.
112 * Returns 0 on success or error code.
114 int ptlrpc_register_bulk(struct ptlrpc_request *req)
116 struct ptlrpc_bulk_desc *desc = req->rq_bulk;
117 lnet_process_id_t peer;
123 lnet_handle_me_t me_h;
126 if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_BULK_GET_NET))
129 /* NB no locking required until desc is on the network */
130 LASSERT(desc->bd_nob > 0);
131 LASSERT(desc->bd_md_count == 0);
132 LASSERT(desc->bd_md_max_brw <= PTLRPC_BULK_OPS_COUNT);
133 LASSERT(desc->bd_iov_count <= PTLRPC_MAX_BRW_PAGES);
134 LASSERT(desc->bd_req != NULL);
135 LASSERT(desc->bd_type == BULK_PUT_SINK ||
136 desc->bd_type == BULK_GET_SOURCE);
138 /* cleanup the state of the bulk for it will be reused */
139 if (req->rq_resend || req->rq_send_state == LUSTRE_IMP_REPLAY)
140 desc->bd_nob_transferred = 0;
142 LASSERT(desc->bd_nob_transferred == 0);
144 desc->bd_failure = 0;
146 peer = desc->bd_import->imp_connection->c_peer;
148 LASSERT(desc->bd_cbid.cbid_fn == client_bulk_callback);
149 LASSERT(desc->bd_cbid.cbid_arg == desc);
151 /* An XID is only used for a single request from the client.
152 * For retried bulk transfers, a new XID will be allocated in
153 * in ptlrpc_check_set() if it needs to be resent, so it is not
154 * using the same RDMA match bits after an error.
156 * For multi-bulk RPCs, rq_xid is the last XID needed for bulks. The
157 * first bulk XID is power-of-two aligned before rq_xid. LU-1431 */
158 xid = req->rq_xid & ~((__u64)desc->bd_md_max_brw - 1);
159 LASSERTF(!(desc->bd_registered &&
160 req->rq_send_state != LUSTRE_IMP_REPLAY) ||
161 xid != desc->bd_last_xid,
162 "registered: %d rq_xid: "LPU64" bd_last_xid: "LPU64"\n",
163 desc->bd_registered, xid, desc->bd_last_xid);
165 total_md = (desc->bd_iov_count + LNET_MAX_IOV - 1) / LNET_MAX_IOV;
166 desc->bd_registered = 1;
167 desc->bd_last_xid = xid;
168 desc->bd_md_count = total_md;
169 md.user_ptr = &desc->bd_cbid;
170 md.eq_handle = ptlrpc_eq_h;
171 md.threshold = 1; /* PUT or GET */
173 for (posted_md = 0; posted_md < total_md; posted_md++, xid++) {
174 md.options = PTLRPC_MD_OPTIONS |
175 ((desc->bd_type == BULK_GET_SOURCE) ?
176 LNET_MD_OP_GET : LNET_MD_OP_PUT);
177 ptlrpc_fill_bulk_md(&md, desc, posted_md);
179 rc = LNetMEAttach(desc->bd_portal, peer, xid, 0,
180 LNET_UNLINK, LNET_INS_AFTER, &me_h);
182 CERROR("%s: LNetMEAttach failed x"LPU64"/%d: rc = %d\n",
183 desc->bd_import->imp_obd->obd_name, xid,
188 /* About to let the network at it... */
189 rc = LNetMDAttach(me_h, md, LNET_UNLINK,
190 &desc->bd_mds[posted_md]);
192 CERROR("%s: LNetMDAttach failed x"LPU64"/%d: rc = %d\n",
193 desc->bd_import->imp_obd->obd_name, xid,
195 rc2 = LNetMEUnlink(me_h);
202 LASSERT(rc == -ENOMEM);
203 spin_lock(&desc->bd_lock);
204 desc->bd_md_count -= total_md - posted_md;
205 spin_unlock(&desc->bd_lock);
206 LASSERT(desc->bd_md_count >= 0);
207 mdunlink_iterate_helper(desc->bd_mds, desc->bd_md_max_brw);
208 req->rq_status = -ENOMEM;
212 /* Set rq_xid to matchbits of the final bulk so that server can
213 * infer the number of bulks that were prepared */
215 LASSERTF(desc->bd_last_xid == (req->rq_xid & PTLRPC_BULK_OPS_MASK),
216 "bd_last_xid = x"LPU64", rq_xid = x"LPU64"\n",
217 desc->bd_last_xid, req->rq_xid);
219 spin_lock(&desc->bd_lock);
220 /* Holler if peer manages to touch buffers before he knows the xid */
221 if (desc->bd_md_count != total_md)
222 CWARN("%s: Peer %s touched %d buffers while I registered\n",
223 desc->bd_import->imp_obd->obd_name, libcfs_id2str(peer),
224 total_md - desc->bd_md_count);
225 spin_unlock(&desc->bd_lock);
227 CDEBUG(D_NET, "Setup %u bulk %s buffers: %u pages %u bytes, "
228 "xid x"LPX64"-"LPX64", portal %u\n", desc->bd_md_count,
229 desc->bd_type == BULK_GET_SOURCE ? "get-source" : "put-sink",
230 desc->bd_iov_count, desc->bd_nob,
231 desc->bd_last_xid, req->rq_xid, desc->bd_portal);
235 EXPORT_SYMBOL(ptlrpc_register_bulk);
238 * Disconnect a bulk desc from the network. Idempotent. Not
239 * thread-safe (i.e. only interlocks with completion callback).
240 * Returns 1 on success or 0 if network unregistration failed for whatever
243 int ptlrpc_unregister_bulk(struct ptlrpc_request *req, int async)
245 struct ptlrpc_bulk_desc *desc = req->rq_bulk;
246 wait_queue_head_t *wq;
247 struct l_wait_info lwi;
250 LASSERT(!in_interrupt()); /* might sleep */
252 /* Let's setup deadline for reply unlink. */
253 if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_BULK_UNLINK) &&
254 async && req->rq_bulk_deadline == 0)
255 req->rq_bulk_deadline = cfs_time_current_sec() + LONG_UNLINK;
257 if (ptlrpc_client_bulk_active(req) == 0) /* completed or */
258 return 1; /* never registered */
260 LASSERT(desc->bd_req == req); /* bd_req NULL until registered */
262 /* the unlink ensures the callback happens ASAP and is the last
263 * one. If it fails, it must be because completion just happened,
264 * but we must still l_wait_event() in this case to give liblustre
265 * a chance to run client_bulk_callback() */
266 mdunlink_iterate_helper(desc->bd_mds, desc->bd_md_max_brw);
268 if (ptlrpc_client_bulk_active(req) == 0) /* completed or */
269 return 1; /* never registered */
271 /* Move to "Unregistering" phase as bulk was not unlinked yet. */
272 ptlrpc_rqphase_move(req, RQ_PHASE_UNREGISTERING);
274 /* Do not wait for unlink to finish. */
278 if (req->rq_set != NULL)
279 wq = &req->rq_set->set_waitq;
281 wq = &req->rq_reply_waitq;
284 /* Network access will complete in finite time but the HUGE
285 * timeout lets us CWARN for visibility of sluggish NALs */
286 lwi = LWI_TIMEOUT_INTERVAL(cfs_time_seconds(LONG_UNLINK),
287 cfs_time_seconds(1), NULL, NULL);
288 rc = l_wait_event(*wq, !ptlrpc_client_bulk_active(req), &lwi);
290 ptlrpc_rqphase_move(req, req->rq_next_phase);
294 LASSERT(rc == -ETIMEDOUT);
295 DEBUG_REQ(D_WARNING, req, "Unexpectedly long timeout: desc %p",
300 EXPORT_SYMBOL(ptlrpc_unregister_bulk);
302 static void ptlrpc_at_set_reply(struct ptlrpc_request *req, int flags)
304 struct ptlrpc_service_part *svcpt = req->rq_rqbd->rqbd_svcpt;
305 struct ptlrpc_service *svc = svcpt->scp_service;
306 int service_time = max_t(int, cfs_time_current_sec() -
307 req->rq_arrival_time.tv_sec, 1);
309 if (!(flags & PTLRPC_REPLY_EARLY) &&
310 (req->rq_type != PTL_RPC_MSG_ERR) &&
311 (req->rq_reqmsg != NULL) &&
312 !(lustre_msg_get_flags(req->rq_reqmsg) &
313 (MSG_RESENT | MSG_REPLAY |
314 MSG_REQ_REPLAY_DONE | MSG_LOCK_REPLAY_DONE))) {
315 /* early replies, errors and recovery requests don't count
316 * toward our service time estimate */
317 int oldse = at_measured(&svcpt->scp_at_estimate, service_time);
320 DEBUG_REQ(D_ADAPTTO, req,
321 "svc %s changed estimate from %d to %d",
322 svc->srv_name, oldse,
323 at_get(&svcpt->scp_at_estimate));
326 /* Report actual service time for client latency calc */
327 lustre_msg_set_service_time(req->rq_repmsg, service_time);
328 /* Report service time estimate for future client reqs, but report 0
329 * (to be ignored by client) if it's a error reply during recovery.
331 if (req->rq_type == PTL_RPC_MSG_ERR &&
332 (req->rq_export == NULL || req->rq_export->exp_obd->obd_recovering))
333 lustre_msg_set_timeout(req->rq_repmsg, 0);
335 lustre_msg_set_timeout(req->rq_repmsg,
336 at_get(&svcpt->scp_at_estimate));
338 if (req->rq_reqmsg &&
339 !(lustre_msghdr_get_flags(req->rq_reqmsg) & MSGHDR_AT_SUPPORT)) {
340 CDEBUG(D_ADAPTTO, "No early reply support: flags=%#x "
341 "req_flags=%#x magic=%d:%x/%x len=%d\n",
342 flags, lustre_msg_get_flags(req->rq_reqmsg),
343 lustre_msg_is_v1(req->rq_reqmsg),
344 lustre_msg_get_magic(req->rq_reqmsg),
345 lustre_msg_get_magic(req->rq_repmsg), req->rq_replen);
350 * Send request reply from request \a req reply buffer.
351 * \a flags defines reply types
352 * Returns 0 on sucess or error code
354 int ptlrpc_send_reply(struct ptlrpc_request *req, int flags)
356 struct ptlrpc_reply_state *rs = req->rq_reply_state;
357 struct ptlrpc_connection *conn;
360 /* We must already have a reply buffer (only ptlrpc_error() may be
361 * called without one). The reply generated by sptlrpc layer (e.g.
362 * error notify, etc.) might have NULL rq->reqmsg; Otherwise we must
363 * have a request buffer which is either the actual (swabbed) incoming
364 * request, or a saved copy if this is a req saved in
365 * target_queue_final_reply().
367 LASSERT(req->rq_no_reply == 0);
368 LASSERT(req->rq_reqbuf != NULL);
370 LASSERT((flags & PTLRPC_REPLY_MAYBE_DIFFICULT) || !rs->rs_difficult);
371 LASSERT(req->rq_repmsg != NULL);
372 LASSERT(req->rq_repmsg == rs->rs_msg);
373 LASSERT(rs->rs_cb_id.cbid_fn == reply_out_callback);
374 LASSERT(rs->rs_cb_id.cbid_arg == rs);
376 /* There may be no rq_export during failover */
378 if (unlikely(req->rq_export && req->rq_export->exp_obd &&
379 req->rq_export->exp_obd->obd_fail)) {
380 /* Failed obd's only send ENODEV */
381 req->rq_type = PTL_RPC_MSG_ERR;
382 req->rq_status = -ENODEV;
383 CDEBUG(D_HA, "sending ENODEV from failed obd %d\n",
384 req->rq_export->exp_obd->obd_minor);
387 /* In order to keep interoprability with the client (< 2.3) which
388 * doesn't have pb_jobid in ptlrpc_body, We have to shrink the
389 * ptlrpc_body in reply buffer to ptlrpc_body_v2, otherwise, the
390 * reply buffer on client will be overflow.
392 * XXX Remove this whenver we drop the interoprability with such client.
394 req->rq_replen = lustre_shrink_msg(req->rq_repmsg, 0,
395 sizeof(struct ptlrpc_body_v2), 1);
397 if (req->rq_type != PTL_RPC_MSG_ERR)
398 req->rq_type = PTL_RPC_MSG_REPLY;
400 lustre_msg_set_type(req->rq_repmsg, req->rq_type);
401 lustre_msg_set_status(req->rq_repmsg,
402 ptlrpc_status_hton(req->rq_status));
403 lustre_msg_set_opc(req->rq_repmsg,
404 req->rq_reqmsg ? lustre_msg_get_opc(req->rq_reqmsg) : 0);
406 target_pack_pool_reply(req);
408 ptlrpc_at_set_reply(req, flags);
410 if (req->rq_export == NULL || req->rq_export->exp_connection == NULL)
411 conn = ptlrpc_connection_get(req->rq_peer, req->rq_self, NULL);
413 conn = ptlrpc_connection_addref(req->rq_export->exp_connection);
415 if (unlikely(conn == NULL)) {
416 CERROR("not replying on NULL connection\n"); /* bug 9635 */
419 ptlrpc_rs_addref(rs); /* +1 ref for the network */
421 rc = sptlrpc_svc_wrap_reply(req);
425 req->rq_sent = cfs_time_current_sec();
427 rc = ptl_send_buf(&rs->rs_md_h, rs->rs_repbuf, rs->rs_repdata_len,
428 (rs->rs_difficult && !rs->rs_no_ack) ?
429 LNET_ACK_REQ : LNET_NOACK_REQ,
431 ptlrpc_req2svc(req)->srv_rep_portal,
432 req->rq_xid, req->rq_reply_off);
434 if (unlikely(rc != 0))
435 ptlrpc_req_drop_rs(req);
436 ptlrpc_connection_put(conn);
439 EXPORT_SYMBOL(ptlrpc_send_reply);
441 int ptlrpc_reply(struct ptlrpc_request *req)
443 if (req->rq_no_reply)
446 return (ptlrpc_send_reply(req, 0));
448 EXPORT_SYMBOL(ptlrpc_reply);
451 * For request \a req send an error reply back. Create empty
452 * reply buffers if necessary.
454 int ptlrpc_send_error(struct ptlrpc_request *req, int may_be_difficult)
458 if (req->rq_no_reply)
461 if (!req->rq_repmsg) {
462 rc = lustre_pack_reply(req, 1, NULL, NULL);
467 if (req->rq_status != -ENOSPC && req->rq_status != -EACCES &&
468 req->rq_status != -EPERM && req->rq_status != -ENOENT &&
469 req->rq_status != -EINPROGRESS && req->rq_status != -EDQUOT)
470 req->rq_type = PTL_RPC_MSG_ERR;
472 rc = ptlrpc_send_reply(req, may_be_difficult);
475 EXPORT_SYMBOL(ptlrpc_send_error);
477 int ptlrpc_error(struct ptlrpc_request *req)
479 return ptlrpc_send_error(req, 0);
481 EXPORT_SYMBOL(ptlrpc_error);
484 * Send request \a request.
485 * if \a noreply is set, don't expect any reply back and don't set up
487 * Returns 0 on success or error code.
489 int ptl_send_rpc(struct ptlrpc_request *request, int noreply)
494 struct ptlrpc_connection *connection;
495 lnet_handle_me_t reply_me_h;
497 struct obd_device *obd = request->rq_import->imp_obd;
499 if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_DROP_RPC))
502 LASSERT(request->rq_type == PTL_RPC_MSG_REQUEST);
503 LASSERT(request->rq_wait_ctx == 0);
505 /* If this is a re-transmit, we're required to have disengaged
506 * cleanly from the previous attempt */
507 LASSERT(!request->rq_receiving_reply);
509 if (request->rq_import->imp_obd &&
510 request->rq_import->imp_obd->obd_fail) {
511 CDEBUG(D_HA, "muting rpc for failed imp obd %s\n",
512 request->rq_import->imp_obd->obd_name);
513 /* this prevents us from waiting in ptlrpc_queue_wait */
515 request->rq_status = -ENODEV;
519 connection = request->rq_import->imp_connection;
521 lustre_msg_set_handle(request->rq_reqmsg,
522 &request->rq_import->imp_remote_handle);
523 lustre_msg_set_type(request->rq_reqmsg, PTL_RPC_MSG_REQUEST);
524 lustre_msg_set_conn_cnt(request->rq_reqmsg,
525 request->rq_import->imp_conn_cnt);
526 lustre_msghdr_set_flags(request->rq_reqmsg,
527 request->rq_import->imp_msghdr_flags);
529 if (request->rq_resend)
530 lustre_msg_add_flags(request->rq_reqmsg, MSG_RESENT);
532 if (request->rq_memalloc)
533 mpflag = cfs_memory_pressure_get_and_set();
535 rc = sptlrpc_cli_wrap_request(request);
539 /* bulk register should be done after wrap_request() */
540 if (request->rq_bulk != NULL) {
541 rc = ptlrpc_register_bulk(request);
547 LASSERT(request->rq_replen != 0);
548 if (request->rq_repbuf == NULL) {
549 LASSERT(request->rq_repdata == NULL);
550 LASSERT(request->rq_repmsg == NULL);
551 rc = sptlrpc_cli_alloc_repbuf(request,
554 /* this prevents us from looping in
555 * ptlrpc_queue_wait */
557 request->rq_status = rc;
558 GOTO(cleanup_bulk, rc);
561 request->rq_repdata = NULL;
562 request->rq_repmsg = NULL;
565 rc = LNetMEAttach(request->rq_reply_portal,/*XXX FIXME bug 249*/
566 connection->c_peer, request->rq_xid, 0,
567 LNET_UNLINK, LNET_INS_AFTER, &reply_me_h);
569 CERROR("LNetMEAttach failed: %d\n", rc);
570 LASSERT(rc == -ENOMEM);
571 GOTO(cleanup_bulk, rc = -ENOMEM);
575 spin_lock(&request->rq_lock);
576 /* If the MD attach succeeds, there _will_ be a reply_in callback */
577 request->rq_receiving_reply = !noreply;
578 /* We are responsible for unlinking the reply buffer */
579 request->rq_must_unlink = !noreply;
580 /* Clear any flags that may be present from previous sends. */
581 request->rq_replied = 0;
583 request->rq_timedout = 0;
584 request->rq_net_err = 0;
585 request->rq_resend = 0;
586 request->rq_restart = 0;
587 request->rq_reply_truncate = 0;
588 spin_unlock(&request->rq_lock);
591 reply_md.start = request->rq_repbuf;
592 reply_md.length = request->rq_repbuf_len;
593 /* Allow multiple early replies */
594 reply_md.threshold = LNET_MD_THRESH_INF;
595 /* Manage remote for early replies */
596 reply_md.options = PTLRPC_MD_OPTIONS | LNET_MD_OP_PUT |
597 LNET_MD_MANAGE_REMOTE |
598 LNET_MD_TRUNCATE; /* allow to make EOVERFLOW error */;
599 reply_md.user_ptr = &request->rq_reply_cbid;
600 reply_md.eq_handle = ptlrpc_eq_h;
602 /* We must see the unlink callback to unset rq_must_unlink,
603 so we can't auto-unlink */
604 rc = LNetMDAttach(reply_me_h, reply_md, LNET_RETAIN,
605 &request->rq_reply_md_h);
607 CERROR("LNetMDAttach failed: %d\n", rc);
608 LASSERT(rc == -ENOMEM);
609 spin_lock(&request->rq_lock);
610 /* ...but the MD attach didn't succeed... */
611 request->rq_receiving_reply = 0;
612 spin_unlock(&request->rq_lock);
613 GOTO(cleanup_me, rc = -ENOMEM);
616 CDEBUG(D_NET, "Setup reply buffer: %u bytes, xid "LPU64
618 request->rq_repbuf_len, request->rq_xid,
619 request->rq_reply_portal);
622 /* add references on request for request_out_callback */
623 ptlrpc_request_addref(request);
624 if (obd->obd_svc_stats != NULL)
625 lprocfs_counter_add(obd->obd_svc_stats, PTLRPC_REQACTIVE_CNTR,
626 atomic_read(&request->rq_import->imp_inflight));
628 OBD_FAIL_TIMEOUT(OBD_FAIL_PTLRPC_DELAY_SEND, request->rq_timeout + 5);
630 do_gettimeofday(&request->rq_arrival_time);
631 request->rq_sent = cfs_time_current_sec();
632 /* We give the server rq_timeout secs to process the req, and
633 add the network latency for our local timeout. */
634 request->rq_deadline = request->rq_sent + request->rq_timeout +
635 ptlrpc_at_get_net_latency(request);
637 ptlrpc_pinger_sending_on_import(request->rq_import);
639 DEBUG_REQ(D_INFO, request, "send flg=%x",
640 lustre_msg_get_flags(request->rq_reqmsg));
641 rc = ptl_send_buf(&request->rq_req_md_h,
642 request->rq_reqbuf, request->rq_reqdata_len,
643 LNET_NOACK_REQ, &request->rq_req_cbid,
645 request->rq_request_portal,
650 ptlrpc_req_finished(request);
655 /* MEUnlink is safe; the PUT didn't even get off the ground, and
656 * nobody apart from the PUT's target has the right nid+XID to
657 * access the reply buffer. */
658 rc2 = LNetMEUnlink(reply_me_h);
660 /* UNLINKED callback called synchronously */
661 LASSERT(!request->rq_receiving_reply);
664 /* We do sync unlink here as there was no real transfer here so
665 * the chance to have long unlink to sluggish net is smaller here. */
666 ptlrpc_unregister_bulk(request, 0);
668 if (request->rq_memalloc)
669 cfs_memory_pressure_restore(mpflag);
672 EXPORT_SYMBOL(ptl_send_rpc);
675 * Register request buffer descriptor for request receiving.
677 int ptlrpc_register_rqbd(struct ptlrpc_request_buffer_desc *rqbd)
679 struct ptlrpc_service *service = rqbd->rqbd_svcpt->scp_service;
680 static lnet_process_id_t match_id = {LNET_NID_ANY, LNET_PID_ANY};
683 lnet_handle_me_t me_h;
685 CDEBUG(D_NET, "LNetMEAttach: portal %d\n",
686 service->srv_req_portal);
688 if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_RQBD))
691 /* NB: CPT affinity service should use new LNet flag LNET_INS_LOCAL,
692 * which means buffer can only be attached on local CPT, and LND
693 * threads can find it by grabbing a local lock */
694 rc = LNetMEAttach(service->srv_req_portal,
695 match_id, 0, ~0, LNET_UNLINK,
696 rqbd->rqbd_svcpt->scp_cpt >= 0 ?
697 LNET_INS_LOCAL : LNET_INS_AFTER, &me_h);
699 CERROR("LNetMEAttach failed: %d\n", rc);
703 LASSERT(rqbd->rqbd_refcount == 0);
704 rqbd->rqbd_refcount = 1;
706 md.start = rqbd->rqbd_buffer;
707 md.length = service->srv_buf_size;
708 md.max_size = service->srv_max_req_size;
709 md.threshold = LNET_MD_THRESH_INF;
710 md.options = PTLRPC_MD_OPTIONS | LNET_MD_OP_PUT | LNET_MD_MAX_SIZE;
711 md.user_ptr = &rqbd->rqbd_cbid;
712 md.eq_handle = ptlrpc_eq_h;
714 rc = LNetMDAttach(me_h, md, LNET_UNLINK, &rqbd->rqbd_md_h);
718 CERROR("LNetMDAttach failed: %d;\n", rc);
719 LASSERT(rc == -ENOMEM);
720 rc = LNetMEUnlink(me_h);
722 rqbd->rqbd_refcount = 0;