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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 #define DEBUG_SUBSYSTEM S_OSC
35 #include <lprocfs_status.h>
36 #include <lustre_debug.h>
37 #include <lustre_dlm.h>
38 #include <lustre_fid.h>
39 #include <lustre_ha.h>
40 #include <uapi/linux/lustre/lustre_ioctl.h>
41 #include <lustre_net.h>
42 #include <lustre_obdo.h>
43 #include <uapi/linux/lustre/lustre_param.h>
45 #include <obd_cksum.h>
46 #include <obd_class.h>
47 #include <lustre_osc.h>
49 #include "osc_internal.h"
51 atomic_t osc_pool_req_count;
52 unsigned int osc_reqpool_maxreqcount;
53 struct ptlrpc_request_pool *osc_rq_pool;
55 /* max memory used for request pool, unit is MB */
56 static unsigned int osc_reqpool_mem_max = 5;
57 module_param(osc_reqpool_mem_max, uint, 0444);
59 #define osc_grant_args osc_brw_async_args
61 struct osc_setattr_args {
63 obd_enqueue_update_f sa_upcall;
67 struct osc_fsync_args {
68 struct osc_object *fa_obj;
70 obd_enqueue_update_f fa_upcall;
74 struct osc_ladvise_args {
76 obd_enqueue_update_f la_upcall;
80 static void osc_release_ppga(struct brw_page **ppga, size_t count);
81 static int brw_interpret(const struct lu_env *env, struct ptlrpc_request *req,
84 void osc_pack_req_body(struct ptlrpc_request *req, struct obdo *oa)
86 struct ost_body *body;
88 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
91 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
94 static int osc_getattr(const struct lu_env *env, struct obd_export *exp,
97 struct ptlrpc_request *req;
98 struct ost_body *body;
102 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
106 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
108 ptlrpc_request_free(req);
112 osc_pack_req_body(req, oa);
114 ptlrpc_request_set_replen(req);
116 rc = ptlrpc_queue_wait(req);
120 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
122 GOTO(out, rc = -EPROTO);
124 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
125 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
127 oa->o_blksize = cli_brw_size(exp->exp_obd);
128 oa->o_valid |= OBD_MD_FLBLKSZ;
132 ptlrpc_req_finished(req);
137 static int osc_setattr(const struct lu_env *env, struct obd_export *exp,
140 struct ptlrpc_request *req;
141 struct ost_body *body;
145 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
147 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
151 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
153 ptlrpc_request_free(req);
157 osc_pack_req_body(req, oa);
159 ptlrpc_request_set_replen(req);
161 rc = ptlrpc_queue_wait(req);
165 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
167 GOTO(out, rc = -EPROTO);
169 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
173 ptlrpc_req_finished(req);
178 static int osc_setattr_interpret(const struct lu_env *env,
179 struct ptlrpc_request *req,
180 struct osc_setattr_args *sa, int rc)
182 struct ost_body *body;
188 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
190 GOTO(out, rc = -EPROTO);
192 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, sa->sa_oa,
195 rc = sa->sa_upcall(sa->sa_cookie, rc);
199 int osc_setattr_async(struct obd_export *exp, struct obdo *oa,
200 obd_enqueue_update_f upcall, void *cookie,
201 struct ptlrpc_request_set *rqset)
203 struct ptlrpc_request *req;
204 struct osc_setattr_args *sa;
209 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
213 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
215 ptlrpc_request_free(req);
219 osc_pack_req_body(req, oa);
221 ptlrpc_request_set_replen(req);
223 /* do mds to ost setattr asynchronously */
225 /* Do not wait for response. */
226 ptlrpcd_add_req(req);
228 req->rq_interpret_reply =
229 (ptlrpc_interpterer_t)osc_setattr_interpret;
231 CLASSERT(sizeof(*sa) <= sizeof(req->rq_async_args));
232 sa = ptlrpc_req_async_args(req);
234 sa->sa_upcall = upcall;
235 sa->sa_cookie = cookie;
237 if (rqset == PTLRPCD_SET)
238 ptlrpcd_add_req(req);
240 ptlrpc_set_add_req(rqset, req);
246 static int osc_ladvise_interpret(const struct lu_env *env,
247 struct ptlrpc_request *req,
250 struct osc_ladvise_args *la = arg;
251 struct ost_body *body;
257 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
259 GOTO(out, rc = -EPROTO);
261 *la->la_oa = body->oa;
263 rc = la->la_upcall(la->la_cookie, rc);
268 * If rqset is NULL, do not wait for response. Upcall and cookie could also
269 * be NULL in this case
271 int osc_ladvise_base(struct obd_export *exp, struct obdo *oa,
272 struct ladvise_hdr *ladvise_hdr,
273 obd_enqueue_update_f upcall, void *cookie,
274 struct ptlrpc_request_set *rqset)
276 struct ptlrpc_request *req;
277 struct ost_body *body;
278 struct osc_ladvise_args *la;
280 struct lu_ladvise *req_ladvise;
281 struct lu_ladvise *ladvise = ladvise_hdr->lah_advise;
282 int num_advise = ladvise_hdr->lah_count;
283 struct ladvise_hdr *req_ladvise_hdr;
286 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_LADVISE);
290 req_capsule_set_size(&req->rq_pill, &RMF_OST_LADVISE, RCL_CLIENT,
291 num_advise * sizeof(*ladvise));
292 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_LADVISE);
294 ptlrpc_request_free(req);
297 req->rq_request_portal = OST_IO_PORTAL;
298 ptlrpc_at_set_req_timeout(req);
300 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
302 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
305 req_ladvise_hdr = req_capsule_client_get(&req->rq_pill,
306 &RMF_OST_LADVISE_HDR);
307 memcpy(req_ladvise_hdr, ladvise_hdr, sizeof(*ladvise_hdr));
309 req_ladvise = req_capsule_client_get(&req->rq_pill, &RMF_OST_LADVISE);
310 memcpy(req_ladvise, ladvise, sizeof(*ladvise) * num_advise);
311 ptlrpc_request_set_replen(req);
314 /* Do not wait for response. */
315 ptlrpcd_add_req(req);
319 req->rq_interpret_reply = osc_ladvise_interpret;
320 CLASSERT(sizeof(*la) <= sizeof(req->rq_async_args));
321 la = ptlrpc_req_async_args(req);
323 la->la_upcall = upcall;
324 la->la_cookie = cookie;
326 if (rqset == PTLRPCD_SET)
327 ptlrpcd_add_req(req);
329 ptlrpc_set_add_req(rqset, req);
334 static int osc_create(const struct lu_env *env, struct obd_export *exp,
337 struct ptlrpc_request *req;
338 struct ost_body *body;
343 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
344 LASSERT(fid_seq_is_echo(ostid_seq(&oa->o_oi)));
346 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
348 GOTO(out, rc = -ENOMEM);
350 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
352 ptlrpc_request_free(req);
356 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
359 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
361 ptlrpc_request_set_replen(req);
363 rc = ptlrpc_queue_wait(req);
367 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
369 GOTO(out_req, rc = -EPROTO);
371 CDEBUG(D_INFO, "oa flags %x\n", oa->o_flags);
372 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
374 oa->o_blksize = cli_brw_size(exp->exp_obd);
375 oa->o_valid |= OBD_MD_FLBLKSZ;
377 CDEBUG(D_HA, "transno: %lld\n",
378 lustre_msg_get_transno(req->rq_repmsg));
380 ptlrpc_req_finished(req);
385 int osc_punch_send(struct obd_export *exp, struct obdo *oa,
386 obd_enqueue_update_f upcall, void *cookie)
388 struct ptlrpc_request *req;
389 struct osc_setattr_args *sa;
390 struct obd_import *imp = class_exp2cliimp(exp);
391 struct ost_body *body;
396 req = ptlrpc_request_alloc(imp, &RQF_OST_PUNCH);
400 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
402 ptlrpc_request_free(req);
406 osc_set_io_portal(req);
408 ptlrpc_at_set_req_timeout(req);
410 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
412 lustre_set_wire_obdo(&imp->imp_connect_data, &body->oa, oa);
414 ptlrpc_request_set_replen(req);
416 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
417 CLASSERT(sizeof(*sa) <= sizeof(req->rq_async_args));
418 sa = ptlrpc_req_async_args(req);
420 sa->sa_upcall = upcall;
421 sa->sa_cookie = cookie;
423 ptlrpcd_add_req(req);
427 EXPORT_SYMBOL(osc_punch_send);
429 static int osc_sync_interpret(const struct lu_env *env,
430 struct ptlrpc_request *req,
433 struct osc_fsync_args *fa = arg;
434 struct ost_body *body;
435 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
436 unsigned long valid = 0;
437 struct cl_object *obj;
443 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
445 CERROR("can't unpack ost_body\n");
446 GOTO(out, rc = -EPROTO);
449 *fa->fa_oa = body->oa;
450 obj = osc2cl(fa->fa_obj);
452 /* Update osc object's blocks attribute */
453 cl_object_attr_lock(obj);
454 if (body->oa.o_valid & OBD_MD_FLBLOCKS) {
455 attr->cat_blocks = body->oa.o_blocks;
460 cl_object_attr_update(env, obj, attr, valid);
461 cl_object_attr_unlock(obj);
464 rc = fa->fa_upcall(fa->fa_cookie, rc);
468 int osc_sync_base(struct osc_object *obj, struct obdo *oa,
469 obd_enqueue_update_f upcall, void *cookie,
470 struct ptlrpc_request_set *rqset)
472 struct obd_export *exp = osc_export(obj);
473 struct ptlrpc_request *req;
474 struct ost_body *body;
475 struct osc_fsync_args *fa;
479 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
483 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
485 ptlrpc_request_free(req);
489 /* overload the size and blocks fields in the oa with start/end */
490 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
492 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
494 ptlrpc_request_set_replen(req);
495 req->rq_interpret_reply = osc_sync_interpret;
497 CLASSERT(sizeof(*fa) <= sizeof(req->rq_async_args));
498 fa = ptlrpc_req_async_args(req);
501 fa->fa_upcall = upcall;
502 fa->fa_cookie = cookie;
504 if (rqset == PTLRPCD_SET)
505 ptlrpcd_add_req(req);
507 ptlrpc_set_add_req(rqset, req);
512 /* Find and cancel locally locks matched by @mode in the resource found by
513 * @objid. Found locks are added into @cancel list. Returns the amount of
514 * locks added to @cancels list. */
515 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
516 struct list_head *cancels,
517 enum ldlm_mode mode, __u64 lock_flags)
519 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
520 struct ldlm_res_id res_id;
521 struct ldlm_resource *res;
525 /* Return, i.e. cancel nothing, only if ELC is supported (flag in
526 * export) but disabled through procfs (flag in NS).
528 * This distinguishes from a case when ELC is not supported originally,
529 * when we still want to cancel locks in advance and just cancel them
530 * locally, without sending any RPC. */
531 if (exp_connect_cancelset(exp) && !ns_connect_cancelset(ns))
534 ostid_build_res_name(&oa->o_oi, &res_id);
535 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
539 LDLM_RESOURCE_ADDREF(res);
540 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
541 lock_flags, 0, NULL);
542 LDLM_RESOURCE_DELREF(res);
543 ldlm_resource_putref(res);
547 static int osc_destroy_interpret(const struct lu_env *env,
548 struct ptlrpc_request *req, void *data,
551 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
553 atomic_dec(&cli->cl_destroy_in_flight);
554 wake_up(&cli->cl_destroy_waitq);
558 static int osc_can_send_destroy(struct client_obd *cli)
560 if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
561 cli->cl_max_rpcs_in_flight) {
562 /* The destroy request can be sent */
565 if (atomic_dec_return(&cli->cl_destroy_in_flight) <
566 cli->cl_max_rpcs_in_flight) {
568 * The counter has been modified between the two atomic
571 wake_up(&cli->cl_destroy_waitq);
576 static int osc_destroy(const struct lu_env *env, struct obd_export *exp,
579 struct client_obd *cli = &exp->exp_obd->u.cli;
580 struct ptlrpc_request *req;
581 struct ost_body *body;
582 struct list_head cancels = LIST_HEAD_INIT(cancels);
587 CDEBUG(D_INFO, "oa NULL\n");
591 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
592 LDLM_FL_DISCARD_DATA);
594 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
596 ldlm_lock_list_put(&cancels, l_bl_ast, count);
600 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
603 ptlrpc_request_free(req);
607 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
608 ptlrpc_at_set_req_timeout(req);
610 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
612 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
614 ptlrpc_request_set_replen(req);
616 req->rq_interpret_reply = osc_destroy_interpret;
617 if (!osc_can_send_destroy(cli)) {
618 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL);
621 * Wait until the number of on-going destroy RPCs drops
622 * under max_rpc_in_flight
624 rc = l_wait_event_exclusive(cli->cl_destroy_waitq,
625 osc_can_send_destroy(cli), &lwi);
627 ptlrpc_req_finished(req);
632 /* Do not wait for response */
633 ptlrpcd_add_req(req);
637 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
640 u64 bits = OBD_MD_FLBLOCKS | OBD_MD_FLGRANT;
642 LASSERT(!(oa->o_valid & bits));
645 spin_lock(&cli->cl_loi_list_lock);
646 if (OCD_HAS_FLAG(&cli->cl_import->imp_connect_data, GRANT_PARAM))
647 oa->o_dirty = cli->cl_dirty_grant;
649 oa->o_dirty = cli->cl_dirty_pages << PAGE_SHIFT;
650 if (unlikely(cli->cl_dirty_pages - cli->cl_dirty_transit >
651 cli->cl_dirty_max_pages)) {
652 CERROR("dirty %lu - %lu > dirty_max %lu\n",
653 cli->cl_dirty_pages, cli->cl_dirty_transit,
654 cli->cl_dirty_max_pages);
656 } else if (unlikely(atomic_long_read(&obd_dirty_pages) -
657 atomic_long_read(&obd_dirty_transit_pages) >
658 (long)(obd_max_dirty_pages + 1))) {
659 /* The atomic_read() allowing the atomic_inc() are
660 * not covered by a lock thus they may safely race and trip
661 * this CERROR() unless we add in a small fudge factor (+1). */
662 CERROR("%s: dirty %ld - %ld > system dirty_max %ld\n",
663 cli_name(cli), atomic_long_read(&obd_dirty_pages),
664 atomic_long_read(&obd_dirty_transit_pages),
665 obd_max_dirty_pages);
667 } else if (unlikely(cli->cl_dirty_max_pages - cli->cl_dirty_pages >
669 CERROR("dirty %lu - dirty_max %lu too big???\n",
670 cli->cl_dirty_pages, cli->cl_dirty_max_pages);
673 unsigned long nrpages;
674 unsigned long undirty;
676 nrpages = cli->cl_max_pages_per_rpc;
677 nrpages *= cli->cl_max_rpcs_in_flight + 1;
678 nrpages = max(nrpages, cli->cl_dirty_max_pages);
679 undirty = nrpages << PAGE_SHIFT;
680 if (OCD_HAS_FLAG(&cli->cl_import->imp_connect_data,
684 /* take extent tax into account when asking for more
686 nrextents = (nrpages + cli->cl_max_extent_pages - 1) /
687 cli->cl_max_extent_pages;
688 undirty += nrextents * cli->cl_grant_extent_tax;
690 /* Do not ask for more than OBD_MAX_GRANT - a margin for server
691 * to add extent tax, etc.
693 oa->o_undirty = min(undirty, OBD_MAX_GRANT -
694 (PTLRPC_MAX_BRW_PAGES << PAGE_SHIFT)*4UL);
696 oa->o_grant = cli->cl_avail_grant + cli->cl_reserved_grant;
697 oa->o_dropped = cli->cl_lost_grant;
698 cli->cl_lost_grant = 0;
699 spin_unlock(&cli->cl_loi_list_lock);
700 CDEBUG(D_CACHE, "dirty: %llu undirty: %u dropped %u grant: %llu\n",
701 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
704 void osc_update_next_shrink(struct client_obd *cli)
706 cli->cl_next_shrink_grant = ktime_get_seconds() +
707 cli->cl_grant_shrink_interval;
709 CDEBUG(D_CACHE, "next time %lld to shrink grant\n",
710 cli->cl_next_shrink_grant);
713 static void __osc_update_grant(struct client_obd *cli, u64 grant)
715 spin_lock(&cli->cl_loi_list_lock);
716 cli->cl_avail_grant += grant;
717 spin_unlock(&cli->cl_loi_list_lock);
720 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
722 if (body->oa.o_valid & OBD_MD_FLGRANT) {
723 CDEBUG(D_CACHE, "got %llu extra grant\n", body->oa.o_grant);
724 __osc_update_grant(cli, body->oa.o_grant);
728 static int osc_shrink_grant_interpret(const struct lu_env *env,
729 struct ptlrpc_request *req,
732 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
733 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
734 struct ost_body *body;
737 __osc_update_grant(cli, oa->o_grant);
741 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
743 osc_update_grant(cli, body);
749 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
751 spin_lock(&cli->cl_loi_list_lock);
752 oa->o_grant = cli->cl_avail_grant / 4;
753 cli->cl_avail_grant -= oa->o_grant;
754 spin_unlock(&cli->cl_loi_list_lock);
755 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
756 oa->o_valid |= OBD_MD_FLFLAGS;
759 oa->o_flags |= OBD_FL_SHRINK_GRANT;
760 osc_update_next_shrink(cli);
763 /* Shrink the current grant, either from some large amount to enough for a
764 * full set of in-flight RPCs, or if we have already shrunk to that limit
765 * then to enough for a single RPC. This avoids keeping more grant than
766 * needed, and avoids shrinking the grant piecemeal. */
767 static int osc_shrink_grant(struct client_obd *cli)
769 __u64 target_bytes = (cli->cl_max_rpcs_in_flight + 1) *
770 (cli->cl_max_pages_per_rpc << PAGE_SHIFT);
772 spin_lock(&cli->cl_loi_list_lock);
773 if (cli->cl_avail_grant <= target_bytes)
774 target_bytes = cli->cl_max_pages_per_rpc << PAGE_SHIFT;
775 spin_unlock(&cli->cl_loi_list_lock);
777 return osc_shrink_grant_to_target(cli, target_bytes);
780 int osc_shrink_grant_to_target(struct client_obd *cli, __u64 target_bytes)
783 struct ost_body *body;
786 spin_lock(&cli->cl_loi_list_lock);
787 /* Don't shrink if we are already above or below the desired limit
788 * We don't want to shrink below a single RPC, as that will negatively
789 * impact block allocation and long-term performance. */
790 if (target_bytes < cli->cl_max_pages_per_rpc << PAGE_SHIFT)
791 target_bytes = cli->cl_max_pages_per_rpc << PAGE_SHIFT;
793 if (target_bytes >= cli->cl_avail_grant) {
794 spin_unlock(&cli->cl_loi_list_lock);
797 spin_unlock(&cli->cl_loi_list_lock);
803 osc_announce_cached(cli, &body->oa, 0);
805 spin_lock(&cli->cl_loi_list_lock);
806 body->oa.o_grant = cli->cl_avail_grant - target_bytes;
807 cli->cl_avail_grant = target_bytes;
808 spin_unlock(&cli->cl_loi_list_lock);
809 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
810 body->oa.o_valid |= OBD_MD_FLFLAGS;
811 body->oa.o_flags = 0;
813 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
814 osc_update_next_shrink(cli);
816 rc = osc_set_info_async(NULL, cli->cl_import->imp_obd->obd_self_export,
817 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
818 sizeof(*body), body, NULL);
820 __osc_update_grant(cli, body->oa.o_grant);
825 static int osc_should_shrink_grant(struct client_obd *client)
827 time64_t next_shrink = client->cl_next_shrink_grant;
829 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
830 OBD_CONNECT_GRANT_SHRINK) == 0)
833 if (ktime_get_seconds() >= next_shrink - 5) {
834 /* Get the current RPC size directly, instead of going via:
835 * cli_brw_size(obd->u.cli.cl_import->imp_obd->obd_self_export)
836 * Keep comment here so that it can be found by searching. */
837 int brw_size = client->cl_max_pages_per_rpc << PAGE_SHIFT;
839 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
840 client->cl_avail_grant > brw_size)
843 osc_update_next_shrink(client);
848 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
850 struct client_obd *client;
852 list_for_each_entry(client, &item->ti_obd_list, cl_grant_shrink_list) {
853 if (osc_should_shrink_grant(client))
854 osc_shrink_grant(client);
859 static int osc_add_shrink_grant(struct client_obd *client)
863 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
865 osc_grant_shrink_grant_cb, NULL,
866 &client->cl_grant_shrink_list);
868 CERROR("add grant client %s error %d\n", cli_name(client), rc);
871 CDEBUG(D_CACHE, "add grant client %s\n", cli_name(client));
872 osc_update_next_shrink(client);
876 static int osc_del_shrink_grant(struct client_obd *client)
878 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
882 void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
885 * ocd_grant is the total grant amount we're expect to hold: if we've
886 * been evicted, it's the new avail_grant amount, cl_dirty_pages will
887 * drop to 0 as inflight RPCs fail out; otherwise, it's avail_grant +
890 * race is tolerable here: if we're evicted, but imp_state already
891 * left EVICTED state, then cl_dirty_pages must be 0 already.
893 spin_lock(&cli->cl_loi_list_lock);
894 cli->cl_avail_grant = ocd->ocd_grant;
895 if (cli->cl_import->imp_state != LUSTRE_IMP_EVICTED) {
896 cli->cl_avail_grant -= cli->cl_reserved_grant;
897 if (OCD_HAS_FLAG(ocd, GRANT_PARAM))
898 cli->cl_avail_grant -= cli->cl_dirty_grant;
900 cli->cl_avail_grant -=
901 cli->cl_dirty_pages << PAGE_SHIFT;
904 if (OCD_HAS_FLAG(ocd, GRANT_PARAM)) {
908 /* overhead for each extent insertion */
909 cli->cl_grant_extent_tax = ocd->ocd_grant_tax_kb << 10;
910 /* determine the appropriate chunk size used by osc_extent. */
911 cli->cl_chunkbits = max_t(int, PAGE_SHIFT,
912 ocd->ocd_grant_blkbits);
913 /* max_pages_per_rpc must be chunk aligned */
914 chunk_mask = ~((1 << (cli->cl_chunkbits - PAGE_SHIFT)) - 1);
915 cli->cl_max_pages_per_rpc = (cli->cl_max_pages_per_rpc +
916 ~chunk_mask) & chunk_mask;
917 /* determine maximum extent size, in #pages */
918 size = (u64)ocd->ocd_grant_max_blks << ocd->ocd_grant_blkbits;
919 cli->cl_max_extent_pages = size >> PAGE_SHIFT;
920 if (cli->cl_max_extent_pages == 0)
921 cli->cl_max_extent_pages = 1;
923 cli->cl_grant_extent_tax = 0;
924 cli->cl_chunkbits = PAGE_SHIFT;
925 cli->cl_max_extent_pages = DT_MAX_BRW_PAGES;
927 spin_unlock(&cli->cl_loi_list_lock);
929 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld."
930 "chunk bits: %d cl_max_extent_pages: %d\n",
932 cli->cl_avail_grant, cli->cl_lost_grant, cli->cl_chunkbits,
933 cli->cl_max_extent_pages);
935 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
936 list_empty(&cli->cl_grant_shrink_list))
937 osc_add_shrink_grant(cli);
939 EXPORT_SYMBOL(osc_init_grant);
941 /* We assume that the reason this OSC got a short read is because it read
942 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
943 * via the LOV, and it _knows_ it's reading inside the file, it's just that
944 * this stripe never got written at or beyond this stripe offset yet. */
945 static void handle_short_read(int nob_read, size_t page_count,
946 struct brw_page **pga)
951 /* skip bytes read OK */
952 while (nob_read > 0) {
953 LASSERT (page_count > 0);
955 if (pga[i]->count > nob_read) {
956 /* EOF inside this page */
957 ptr = kmap(pga[i]->pg) +
958 (pga[i]->off & ~PAGE_MASK);
959 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
966 nob_read -= pga[i]->count;
971 /* zero remaining pages */
972 while (page_count-- > 0) {
973 ptr = kmap(pga[i]->pg) + (pga[i]->off & ~PAGE_MASK);
974 memset(ptr, 0, pga[i]->count);
980 static int check_write_rcs(struct ptlrpc_request *req,
981 int requested_nob, int niocount,
982 size_t page_count, struct brw_page **pga)
987 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
988 sizeof(*remote_rcs) *
990 if (remote_rcs == NULL) {
991 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
995 /* return error if any niobuf was in error */
996 for (i = 0; i < niocount; i++) {
997 if ((int)remote_rcs[i] < 0)
998 return(remote_rcs[i]);
1000 if (remote_rcs[i] != 0) {
1001 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1002 i, remote_rcs[i], req);
1006 if (req->rq_bulk != NULL &&
1007 req->rq_bulk->bd_nob_transferred != requested_nob) {
1008 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1009 req->rq_bulk->bd_nob_transferred, requested_nob);
1016 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1018 if (p1->flag != p2->flag) {
1019 unsigned mask = ~(OBD_BRW_FROM_GRANT | OBD_BRW_NOCACHE |
1020 OBD_BRW_SYNC | OBD_BRW_ASYNC |
1021 OBD_BRW_NOQUOTA | OBD_BRW_SOFT_SYNC);
1023 /* warn if we try to combine flags that we don't know to be
1024 * safe to combine */
1025 if (unlikely((p1->flag & mask) != (p2->flag & mask))) {
1026 CWARN("Saw flags 0x%x and 0x%x in the same brw, please "
1027 "report this at https://jira.hpdd.intel.com/\n",
1028 p1->flag, p2->flag);
1033 return (p1->off + p1->count == p2->off);
1036 static int osc_checksum_bulk_t10pi(const char *obd_name, int nob,
1037 size_t pg_count, struct brw_page **pga,
1038 int opc, obd_dif_csum_fn *fn,
1042 struct cfs_crypto_hash_desc *hdesc;
1043 /* Used Adler as the default checksum type on top of DIF tags */
1044 unsigned char cfs_alg = cksum_obd2cfs(OBD_CKSUM_T10_TOP);
1045 struct page *__page;
1046 unsigned char *buffer;
1048 unsigned int bufsize;
1050 int used_number = 0;
1056 LASSERT(pg_count > 0);
1058 __page = alloc_page(GFP_KERNEL);
1062 hdesc = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1063 if (IS_ERR(hdesc)) {
1064 rc = PTR_ERR(hdesc);
1065 CERROR("%s: unable to initialize checksum hash %s: rc = %d\n",
1066 obd_name, cfs_crypto_hash_name(cfs_alg), rc);
1070 buffer = kmap(__page);
1071 guard_start = (__u16 *)buffer;
1072 guard_number = PAGE_SIZE / sizeof(*guard_start);
1073 while (nob > 0 && pg_count > 0) {
1074 unsigned int count = pga[i]->count > nob ? nob : pga[i]->count;
1076 /* corrupt the data before we compute the checksum, to
1077 * simulate an OST->client data error */
1078 if (unlikely(i == 0 && opc == OST_READ &&
1079 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))) {
1080 unsigned char *ptr = kmap(pga[i]->pg);
1081 int off = pga[i]->off & ~PAGE_MASK;
1083 memcpy(ptr + off, "bad1", min_t(typeof(nob), 4, nob));
1088 * The left guard number should be able to hold checksums of a
1091 rc = obd_page_dif_generate_buffer(obd_name, pga[i]->pg, 0,
1093 guard_start + used_number,
1094 guard_number - used_number,
1100 used_number += used;
1101 if (used_number == guard_number) {
1102 cfs_crypto_hash_update_page(hdesc, __page, 0,
1103 used_number * sizeof(*guard_start));
1107 nob -= pga[i]->count;
1115 if (used_number != 0)
1116 cfs_crypto_hash_update_page(hdesc, __page, 0,
1117 used_number * sizeof(*guard_start));
1119 bufsize = sizeof(cksum);
1120 cfs_crypto_hash_final(hdesc, (unsigned char *)&cksum, &bufsize);
1122 /* For sending we only compute the wrong checksum instead
1123 * of corrupting the data so it is still correct on a redo */
1124 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1129 __free_page(__page);
1133 static int osc_checksum_bulk(int nob, size_t pg_count,
1134 struct brw_page **pga, int opc,
1135 enum cksum_types cksum_type,
1139 struct cfs_crypto_hash_desc *hdesc;
1140 unsigned int bufsize;
1141 unsigned char cfs_alg = cksum_obd2cfs(cksum_type);
1143 LASSERT(pg_count > 0);
1145 hdesc = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1146 if (IS_ERR(hdesc)) {
1147 CERROR("Unable to initialize checksum hash %s\n",
1148 cfs_crypto_hash_name(cfs_alg));
1149 return PTR_ERR(hdesc);
1152 while (nob > 0 && pg_count > 0) {
1153 unsigned int count = pga[i]->count > nob ? nob : pga[i]->count;
1155 /* corrupt the data before we compute the checksum, to
1156 * simulate an OST->client data error */
1157 if (i == 0 && opc == OST_READ &&
1158 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE)) {
1159 unsigned char *ptr = kmap(pga[i]->pg);
1160 int off = pga[i]->off & ~PAGE_MASK;
1162 memcpy(ptr + off, "bad1", min_t(typeof(nob), 4, nob));
1165 cfs_crypto_hash_update_page(hdesc, pga[i]->pg,
1166 pga[i]->off & ~PAGE_MASK,
1168 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d\n",
1169 (int)(pga[i]->off & ~PAGE_MASK));
1171 nob -= pga[i]->count;
1176 bufsize = sizeof(*cksum);
1177 cfs_crypto_hash_final(hdesc, (unsigned char *)cksum, &bufsize);
1179 /* For sending we only compute the wrong checksum instead
1180 * of corrupting the data so it is still correct on a redo */
1181 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1187 static int osc_checksum_bulk_rw(const char *obd_name,
1188 enum cksum_types cksum_type,
1189 int nob, size_t pg_count,
1190 struct brw_page **pga, int opc,
1193 obd_dif_csum_fn *fn = NULL;
1194 int sector_size = 0;
1198 obd_t10_cksum2dif(cksum_type, &fn, §or_size);
1201 rc = osc_checksum_bulk_t10pi(obd_name, nob, pg_count, pga,
1202 opc, fn, sector_size, check_sum);
1204 rc = osc_checksum_bulk(nob, pg_count, pga, opc, cksum_type,
1211 osc_brw_prep_request(int cmd, struct client_obd *cli, struct obdo *oa,
1212 u32 page_count, struct brw_page **pga,
1213 struct ptlrpc_request **reqp, int resend)
1215 struct ptlrpc_request *req;
1216 struct ptlrpc_bulk_desc *desc;
1217 struct ost_body *body;
1218 struct obd_ioobj *ioobj;
1219 struct niobuf_remote *niobuf;
1220 int niocount, i, requested_nob, opc, rc, short_io_size = 0;
1221 struct osc_brw_async_args *aa;
1222 struct req_capsule *pill;
1223 struct brw_page *pg_prev;
1225 const char *obd_name = cli->cl_import->imp_obd->obd_name;
1228 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1229 RETURN(-ENOMEM); /* Recoverable */
1230 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1231 RETURN(-EINVAL); /* Fatal */
1233 if ((cmd & OBD_BRW_WRITE) != 0) {
1235 req = ptlrpc_request_alloc_pool(cli->cl_import,
1237 &RQF_OST_BRW_WRITE);
1240 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1245 for (niocount = i = 1; i < page_count; i++) {
1246 if (!can_merge_pages(pga[i - 1], pga[i]))
1250 pill = &req->rq_pill;
1251 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1253 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1254 niocount * sizeof(*niobuf));
1256 for (i = 0; i < page_count; i++)
1257 short_io_size += pga[i]->count;
1259 /* Check if we can do a short io. */
1260 if (!(short_io_size <= cli->cl_short_io_bytes && niocount == 1 &&
1261 imp_connect_shortio(cli->cl_import)))
1264 req_capsule_set_size(pill, &RMF_SHORT_IO, RCL_CLIENT,
1265 opc == OST_READ ? 0 : short_io_size);
1266 if (opc == OST_READ)
1267 req_capsule_set_size(pill, &RMF_SHORT_IO, RCL_SERVER,
1270 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1272 ptlrpc_request_free(req);
1275 osc_set_io_portal(req);
1277 ptlrpc_at_set_req_timeout(req);
1278 /* ask ptlrpc not to resend on EINPROGRESS since BRWs have their own
1280 req->rq_no_retry_einprogress = 1;
1282 if (short_io_size != 0) {
1284 short_io_buf = NULL;
1288 desc = ptlrpc_prep_bulk_imp(req, page_count,
1289 cli->cl_import->imp_connect_data.ocd_brw_size >> LNET_MTU_BITS,
1290 (opc == OST_WRITE ? PTLRPC_BULK_GET_SOURCE :
1291 PTLRPC_BULK_PUT_SINK) |
1292 PTLRPC_BULK_BUF_KIOV,
1294 &ptlrpc_bulk_kiov_pin_ops);
1297 GOTO(out, rc = -ENOMEM);
1298 /* NB request now owns desc and will free it when it gets freed */
1300 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1301 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1302 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1303 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1305 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
1307 /* For READ and WRITE, we can't fill o_uid and o_gid using from_kuid()
1308 * and from_kgid(), because they are asynchronous. Fortunately, variable
1309 * oa contains valid o_uid and o_gid in these two operations.
1310 * Besides, filling o_uid and o_gid is enough for nrs-tbf, see LU-9658.
1311 * OBD_MD_FLUID and OBD_MD_FLUID is not set in order to avoid breaking
1312 * other process logic */
1313 body->oa.o_uid = oa->o_uid;
1314 body->oa.o_gid = oa->o_gid;
1316 obdo_to_ioobj(oa, ioobj);
1317 ioobj->ioo_bufcnt = niocount;
1318 /* The high bits of ioo_max_brw tells server _maximum_ number of bulks
1319 * that might be send for this request. The actual number is decided
1320 * when the RPC is finally sent in ptlrpc_register_bulk(). It sends
1321 * "max - 1" for old client compatibility sending "0", and also so the
1322 * the actual maximum is a power-of-two number, not one less. LU-1431 */
1324 ioobj_max_brw_set(ioobj, desc->bd_md_max_brw);
1326 ioobj_max_brw_set(ioobj, 0);
1328 if (short_io_size != 0) {
1329 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1330 body->oa.o_valid |= OBD_MD_FLFLAGS;
1331 body->oa.o_flags = 0;
1333 body->oa.o_flags |= OBD_FL_SHORT_IO;
1334 CDEBUG(D_CACHE, "Using short io for data transfer, size = %d\n",
1336 if (opc == OST_WRITE) {
1337 short_io_buf = req_capsule_client_get(pill,
1339 LASSERT(short_io_buf != NULL);
1343 LASSERT(page_count > 0);
1345 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1346 struct brw_page *pg = pga[i];
1347 int poff = pg->off & ~PAGE_MASK;
1349 LASSERT(pg->count > 0);
1350 /* make sure there is no gap in the middle of page array */
1351 LASSERTF(page_count == 1 ||
1352 (ergo(i == 0, poff + pg->count == PAGE_SIZE) &&
1353 ergo(i > 0 && i < page_count - 1,
1354 poff == 0 && pg->count == PAGE_SIZE) &&
1355 ergo(i == page_count - 1, poff == 0)),
1356 "i: %d/%d pg: %p off: %llu, count: %u\n",
1357 i, page_count, pg, pg->off, pg->count);
1358 LASSERTF(i == 0 || pg->off > pg_prev->off,
1359 "i %d p_c %u pg %p [pri %lu ind %lu] off %llu"
1360 " prev_pg %p [pri %lu ind %lu] off %llu\n",
1362 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1363 pg_prev->pg, page_private(pg_prev->pg),
1364 pg_prev->pg->index, pg_prev->off);
1365 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1366 (pg->flag & OBD_BRW_SRVLOCK));
1367 if (short_io_size != 0 && opc == OST_WRITE) {
1368 unsigned char *ptr = ll_kmap_atomic(pg->pg, KM_USER0);
1370 LASSERT(short_io_size >= requested_nob + pg->count);
1371 memcpy(short_io_buf + requested_nob,
1374 ll_kunmap_atomic(ptr, KM_USER0);
1375 } else if (short_io_size == 0) {
1376 desc->bd_frag_ops->add_kiov_frag(desc, pg->pg, poff,
1379 requested_nob += pg->count;
1381 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1383 niobuf->rnb_len += pg->count;
1385 niobuf->rnb_offset = pg->off;
1386 niobuf->rnb_len = pg->count;
1387 niobuf->rnb_flags = pg->flag;
1392 LASSERTF((void *)(niobuf - niocount) ==
1393 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1394 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1395 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1397 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1399 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1400 body->oa.o_valid |= OBD_MD_FLFLAGS;
1401 body->oa.o_flags = 0;
1403 body->oa.o_flags |= OBD_FL_RECOV_RESEND;
1406 if (osc_should_shrink_grant(cli))
1407 osc_shrink_grant_local(cli, &body->oa);
1409 /* size[REQ_REC_OFF] still sizeof (*body) */
1410 if (opc == OST_WRITE) {
1411 if (cli->cl_checksum &&
1412 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1413 /* store cl_cksum_type in a local variable since
1414 * it can be changed via lprocfs */
1415 enum cksum_types cksum_type = cli->cl_cksum_type;
1417 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1418 body->oa.o_flags = 0;
1420 body->oa.o_flags |= obd_cksum_type_pack(obd_name,
1422 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1424 rc = osc_checksum_bulk_rw(obd_name, cksum_type,
1425 requested_nob, page_count,
1429 CDEBUG(D_PAGE, "failed to checksum, rc = %d\n",
1433 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1436 /* save this in 'oa', too, for later checking */
1437 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1438 oa->o_flags |= obd_cksum_type_pack(obd_name,
1441 /* clear out the checksum flag, in case this is a
1442 * resend but cl_checksum is no longer set. b=11238 */
1443 oa->o_valid &= ~OBD_MD_FLCKSUM;
1445 oa->o_cksum = body->oa.o_cksum;
1446 /* 1 RC per niobuf */
1447 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1448 sizeof(__u32) * niocount);
1450 if (cli->cl_checksum &&
1451 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1452 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1453 body->oa.o_flags = 0;
1454 body->oa.o_flags |= obd_cksum_type_pack(obd_name,
1455 cli->cl_cksum_type);
1456 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1459 /* Client cksum has been already copied to wire obdo in previous
1460 * lustre_set_wire_obdo(), and in the case a bulk-read is being
1461 * resent due to cksum error, this will allow Server to
1462 * check+dump pages on its side */
1464 ptlrpc_request_set_replen(req);
1466 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1467 aa = ptlrpc_req_async_args(req);
1469 aa->aa_requested_nob = requested_nob;
1470 aa->aa_nio_count = niocount;
1471 aa->aa_page_count = page_count;
1475 INIT_LIST_HEAD(&aa->aa_oaps);
1478 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1479 CDEBUG(D_RPCTRACE, "brw rpc %p - object "DOSTID" offset %lld<>%lld\n",
1480 req, POSTID(&oa->o_oi), niobuf[0].rnb_offset,
1481 niobuf[niocount - 1].rnb_offset + niobuf[niocount - 1].rnb_len);
1485 ptlrpc_req_finished(req);
1489 char dbgcksum_file_name[PATH_MAX];
1491 static void dump_all_bulk_pages(struct obdo *oa, __u32 page_count,
1492 struct brw_page **pga, __u32 server_cksum,
1500 /* will only keep dump of pages on first error for the same range in
1501 * file/fid, not during the resends/retries. */
1502 snprintf(dbgcksum_file_name, sizeof(dbgcksum_file_name),
1503 "%s-checksum_dump-osc-"DFID":[%llu-%llu]-%x-%x",
1504 (strncmp(libcfs_debug_file_path_arr, "NONE", 4) != 0 ?
1505 libcfs_debug_file_path_arr :
1506 LIBCFS_DEBUG_FILE_PATH_DEFAULT),
1507 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : 0ULL,
1508 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1509 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1511 pga[page_count-1]->off + pga[page_count-1]->count - 1,
1512 client_cksum, server_cksum);
1513 filp = filp_open(dbgcksum_file_name,
1514 O_CREAT | O_EXCL | O_WRONLY | O_LARGEFILE, 0600);
1518 CDEBUG(D_INFO, "%s: can't open to dump pages with "
1519 "checksum error: rc = %d\n", dbgcksum_file_name,
1522 CERROR("%s: can't open to dump pages with checksum "
1523 "error: rc = %d\n", dbgcksum_file_name, rc);
1527 for (i = 0; i < page_count; i++) {
1528 len = pga[i]->count;
1529 buf = kmap(pga[i]->pg);
1531 rc = cfs_kernel_write(filp, buf, len, &filp->f_pos);
1533 CERROR("%s: wanted to write %u but got %d "
1534 "error\n", dbgcksum_file_name, len, rc);
1539 CDEBUG(D_INFO, "%s: wrote %d bytes\n",
1540 dbgcksum_file_name, rc);
1545 rc = ll_vfs_fsync_range(filp, 0, LLONG_MAX, 1);
1547 CERROR("%s: sync returns %d\n", dbgcksum_file_name, rc);
1548 filp_close(filp, NULL);
1553 check_write_checksum(struct obdo *oa, const struct lnet_process_id *peer,
1554 __u32 client_cksum, __u32 server_cksum,
1555 struct osc_brw_async_args *aa)
1557 const char *obd_name = aa->aa_cli->cl_import->imp_obd->obd_name;
1558 enum cksum_types cksum_type;
1559 obd_dif_csum_fn *fn = NULL;
1560 int sector_size = 0;
1566 if (server_cksum == client_cksum) {
1567 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1571 if (aa->aa_cli->cl_checksum_dump)
1572 dump_all_bulk_pages(oa, aa->aa_page_count, aa->aa_ppga,
1573 server_cksum, client_cksum);
1575 cksum_type = obd_cksum_type_unpack(oa->o_valid & OBD_MD_FLFLAGS ?
1578 switch (cksum_type) {
1579 case OBD_CKSUM_T10IP512:
1584 case OBD_CKSUM_T10IP4K:
1589 case OBD_CKSUM_T10CRC512:
1591 fn = obd_dif_crc_fn;
1594 case OBD_CKSUM_T10CRC4K:
1596 fn = obd_dif_crc_fn;
1604 rc = osc_checksum_bulk_t10pi(obd_name, aa->aa_requested_nob,
1612 rc = osc_checksum_bulk(aa->aa_requested_nob, aa->aa_page_count,
1613 aa->aa_ppga, OST_WRITE, cksum_type,
1617 msg = "failed to calculate the client write checksum";
1618 else if (cksum_type != obd_cksum_type_unpack(aa->aa_oa->o_flags))
1619 msg = "the server did not use the checksum type specified in "
1620 "the original request - likely a protocol problem";
1621 else if (new_cksum == server_cksum)
1622 msg = "changed on the client after we checksummed it - "
1623 "likely false positive due to mmap IO (bug 11742)";
1624 else if (new_cksum == client_cksum)
1625 msg = "changed in transit before arrival at OST";
1627 msg = "changed in transit AND doesn't match the original - "
1628 "likely false positive due to mmap IO (bug 11742)";
1630 LCONSOLE_ERROR_MSG(0x132, "%s: BAD WRITE CHECKSUM: %s: from %s inode "
1631 DFID " object "DOSTID" extent [%llu-%llu], original "
1632 "client csum %x (type %x), server csum %x (type %x),"
1633 " client csum now %x\n",
1634 obd_name, msg, libcfs_nid2str(peer->nid),
1635 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1636 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1637 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1638 POSTID(&oa->o_oi), aa->aa_ppga[0]->off,
1639 aa->aa_ppga[aa->aa_page_count - 1]->off +
1640 aa->aa_ppga[aa->aa_page_count-1]->count - 1,
1642 obd_cksum_type_unpack(aa->aa_oa->o_flags),
1643 server_cksum, cksum_type, new_cksum);
1647 /* Note rc enters this function as number of bytes transferred */
1648 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1650 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1651 struct client_obd *cli = aa->aa_cli;
1652 const char *obd_name = cli->cl_import->imp_obd->obd_name;
1653 const struct lnet_process_id *peer =
1654 &req->rq_import->imp_connection->c_peer;
1655 struct ost_body *body;
1656 u32 client_cksum = 0;
1659 if (rc < 0 && rc != -EDQUOT) {
1660 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1664 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1665 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1667 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1671 /* set/clear over quota flag for a uid/gid/projid */
1672 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1673 body->oa.o_valid & (OBD_MD_FLALLQUOTA)) {
1674 unsigned qid[LL_MAXQUOTAS] = {
1675 body->oa.o_uid, body->oa.o_gid,
1676 body->oa.o_projid };
1677 CDEBUG(D_QUOTA, "setdq for [%u %u %u] with valid %#llx, flags %x\n",
1678 body->oa.o_uid, body->oa.o_gid, body->oa.o_projid,
1679 body->oa.o_valid, body->oa.o_flags);
1680 osc_quota_setdq(cli, qid, body->oa.o_valid,
1684 osc_update_grant(cli, body);
1689 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1690 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1692 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1694 CERROR("Unexpected +ve rc %d\n", rc);
1698 if (req->rq_bulk != NULL &&
1699 sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1702 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1703 check_write_checksum(&body->oa, peer, client_cksum,
1704 body->oa.o_cksum, aa))
1707 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1708 aa->aa_page_count, aa->aa_ppga);
1712 /* The rest of this function executes only for OST_READs */
1714 if (req->rq_bulk == NULL) {
1715 rc = req_capsule_get_size(&req->rq_pill, &RMF_SHORT_IO,
1717 LASSERT(rc == req->rq_status);
1719 /* if unwrap_bulk failed, return -EAGAIN to retry */
1720 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1723 GOTO(out, rc = -EAGAIN);
1725 if (rc > aa->aa_requested_nob) {
1726 CERROR("Unexpected rc %d (%d requested)\n", rc,
1727 aa->aa_requested_nob);
1731 if (req->rq_bulk != NULL && rc != req->rq_bulk->bd_nob_transferred) {
1732 CERROR ("Unexpected rc %d (%d transferred)\n",
1733 rc, req->rq_bulk->bd_nob_transferred);
1737 if (req->rq_bulk == NULL) {
1739 int nob, pg_count, i = 0;
1742 CDEBUG(D_CACHE, "Using short io read, size %d\n", rc);
1743 pg_count = aa->aa_page_count;
1744 buf = req_capsule_server_sized_get(&req->rq_pill, &RMF_SHORT_IO,
1747 while (nob > 0 && pg_count > 0) {
1749 int count = aa->aa_ppga[i]->count > nob ?
1750 nob : aa->aa_ppga[i]->count;
1752 CDEBUG(D_CACHE, "page %p count %d\n",
1753 aa->aa_ppga[i]->pg, count);
1754 ptr = ll_kmap_atomic(aa->aa_ppga[i]->pg, KM_USER0);
1755 memcpy(ptr + (aa->aa_ppga[i]->off & ~PAGE_MASK), buf,
1757 ll_kunmap_atomic((void *) ptr, KM_USER0);
1766 if (rc < aa->aa_requested_nob)
1767 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1769 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1770 static int cksum_counter;
1771 u32 server_cksum = body->oa.o_cksum;
1774 enum cksum_types cksum_type;
1775 u32 o_flags = body->oa.o_valid & OBD_MD_FLFLAGS ?
1776 body->oa.o_flags : 0;
1778 cksum_type = obd_cksum_type_unpack(o_flags);
1779 rc = osc_checksum_bulk_rw(obd_name, cksum_type, rc,
1780 aa->aa_page_count, aa->aa_ppga,
1781 OST_READ, &client_cksum);
1785 if (req->rq_bulk != NULL &&
1786 peer->nid != req->rq_bulk->bd_sender) {
1788 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1791 if (server_cksum != client_cksum) {
1792 struct ost_body *clbody;
1793 u32 page_count = aa->aa_page_count;
1795 clbody = req_capsule_client_get(&req->rq_pill,
1797 if (cli->cl_checksum_dump)
1798 dump_all_bulk_pages(&clbody->oa, page_count,
1799 aa->aa_ppga, server_cksum,
1802 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1803 "%s%s%s inode "DFID" object "DOSTID
1804 " extent [%llu-%llu], client %x, "
1805 "server %x, cksum_type %x\n",
1807 libcfs_nid2str(peer->nid),
1809 clbody->oa.o_valid & OBD_MD_FLFID ?
1810 clbody->oa.o_parent_seq : 0ULL,
1811 clbody->oa.o_valid & OBD_MD_FLFID ?
1812 clbody->oa.o_parent_oid : 0,
1813 clbody->oa.o_valid & OBD_MD_FLFID ?
1814 clbody->oa.o_parent_ver : 0,
1815 POSTID(&body->oa.o_oi),
1816 aa->aa_ppga[0]->off,
1817 aa->aa_ppga[page_count-1]->off +
1818 aa->aa_ppga[page_count-1]->count - 1,
1819 client_cksum, server_cksum,
1822 aa->aa_oa->o_cksum = client_cksum;
1826 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1829 } else if (unlikely(client_cksum)) {
1830 static int cksum_missed;
1833 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1834 CERROR("Checksum %u requested from %s but not sent\n",
1835 cksum_missed, libcfs_nid2str(peer->nid));
1841 lustre_get_wire_obdo(&req->rq_import->imp_connect_data,
1842 aa->aa_oa, &body->oa);
1847 static int osc_brw_redo_request(struct ptlrpc_request *request,
1848 struct osc_brw_async_args *aa, int rc)
1850 struct ptlrpc_request *new_req;
1851 struct osc_brw_async_args *new_aa;
1852 struct osc_async_page *oap;
1855 DEBUG_REQ(rc == -EINPROGRESS ? D_RPCTRACE : D_ERROR, request,
1856 "redo for recoverable error %d", rc);
1858 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1859 OST_WRITE ? OBD_BRW_WRITE : OBD_BRW_READ,
1860 aa->aa_cli, aa->aa_oa, aa->aa_page_count,
1861 aa->aa_ppga, &new_req, 1);
1865 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1866 if (oap->oap_request != NULL) {
1867 LASSERTF(request == oap->oap_request,
1868 "request %p != oap_request %p\n",
1869 request, oap->oap_request);
1870 if (oap->oap_interrupted) {
1871 ptlrpc_req_finished(new_req);
1876 /* New request takes over pga and oaps from old request.
1877 * Note that copying a list_head doesn't work, need to move it... */
1879 new_req->rq_interpret_reply = request->rq_interpret_reply;
1880 new_req->rq_async_args = request->rq_async_args;
1881 new_req->rq_commit_cb = request->rq_commit_cb;
1882 /* cap resend delay to the current request timeout, this is similar to
1883 * what ptlrpc does (see after_reply()) */
1884 if (aa->aa_resends > new_req->rq_timeout)
1885 new_req->rq_sent = ktime_get_real_seconds() + new_req->rq_timeout;
1887 new_req->rq_sent = ktime_get_real_seconds() + aa->aa_resends;
1888 new_req->rq_generation_set = 1;
1889 new_req->rq_import_generation = request->rq_import_generation;
1891 new_aa = ptlrpc_req_async_args(new_req);
1893 INIT_LIST_HEAD(&new_aa->aa_oaps);
1894 list_splice_init(&aa->aa_oaps, &new_aa->aa_oaps);
1895 INIT_LIST_HEAD(&new_aa->aa_exts);
1896 list_splice_init(&aa->aa_exts, &new_aa->aa_exts);
1897 new_aa->aa_resends = aa->aa_resends;
1899 list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1900 if (oap->oap_request) {
1901 ptlrpc_req_finished(oap->oap_request);
1902 oap->oap_request = ptlrpc_request_addref(new_req);
1906 /* XXX: This code will run into problem if we're going to support
1907 * to add a series of BRW RPCs into a self-defined ptlrpc_request_set
1908 * and wait for all of them to be finished. We should inherit request
1909 * set from old request. */
1910 ptlrpcd_add_req(new_req);
1912 DEBUG_REQ(D_INFO, new_req, "new request");
1917 * ugh, we want disk allocation on the target to happen in offset order. we'll
1918 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1919 * fine for our small page arrays and doesn't require allocation. its an
1920 * insertion sort that swaps elements that are strides apart, shrinking the
1921 * stride down until its '1' and the array is sorted.
1923 static void sort_brw_pages(struct brw_page **array, int num)
1926 struct brw_page *tmp;
1930 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1935 for (i = stride ; i < num ; i++) {
1938 while (j >= stride && array[j - stride]->off > tmp->off) {
1939 array[j] = array[j - stride];
1944 } while (stride > 1);
1947 static void osc_release_ppga(struct brw_page **ppga, size_t count)
1949 LASSERT(ppga != NULL);
1950 OBD_FREE(ppga, sizeof(*ppga) * count);
1953 static int brw_interpret(const struct lu_env *env,
1954 struct ptlrpc_request *req, void *data, int rc)
1956 struct osc_brw_async_args *aa = data;
1957 struct osc_extent *ext;
1958 struct osc_extent *tmp;
1959 struct client_obd *cli = aa->aa_cli;
1960 unsigned long transferred = 0;
1963 rc = osc_brw_fini_request(req, rc);
1964 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
1965 /* When server return -EINPROGRESS, client should always retry
1966 * regardless of the number of times the bulk was resent already. */
1967 if (osc_recoverable_error(rc) && !req->rq_no_delay) {
1968 if (req->rq_import_generation !=
1969 req->rq_import->imp_generation) {
1970 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1971 ""DOSTID", rc = %d.\n",
1972 req->rq_import->imp_obd->obd_name,
1973 POSTID(&aa->aa_oa->o_oi), rc);
1974 } else if (rc == -EINPROGRESS ||
1975 client_should_resend(aa->aa_resends, aa->aa_cli)) {
1976 rc = osc_brw_redo_request(req, aa, rc);
1978 CERROR("%s: too many resent retries for object: "
1979 "%llu:%llu, rc = %d.\n",
1980 req->rq_import->imp_obd->obd_name,
1981 POSTID(&aa->aa_oa->o_oi), rc);
1986 else if (rc == -EAGAIN || rc == -EINPROGRESS)
1991 struct obdo *oa = aa->aa_oa;
1992 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
1993 unsigned long valid = 0;
1994 struct cl_object *obj;
1995 struct osc_async_page *last;
1997 last = brw_page2oap(aa->aa_ppga[aa->aa_page_count - 1]);
1998 obj = osc2cl(last->oap_obj);
2000 cl_object_attr_lock(obj);
2001 if (oa->o_valid & OBD_MD_FLBLOCKS) {
2002 attr->cat_blocks = oa->o_blocks;
2003 valid |= CAT_BLOCKS;
2005 if (oa->o_valid & OBD_MD_FLMTIME) {
2006 attr->cat_mtime = oa->o_mtime;
2009 if (oa->o_valid & OBD_MD_FLATIME) {
2010 attr->cat_atime = oa->o_atime;
2013 if (oa->o_valid & OBD_MD_FLCTIME) {
2014 attr->cat_ctime = oa->o_ctime;
2018 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
2019 struct lov_oinfo *loi = cl2osc(obj)->oo_oinfo;
2020 loff_t last_off = last->oap_count + last->oap_obj_off +
2023 /* Change file size if this is an out of quota or
2024 * direct IO write and it extends the file size */
2025 if (loi->loi_lvb.lvb_size < last_off) {
2026 attr->cat_size = last_off;
2029 /* Extend KMS if it's not a lockless write */
2030 if (loi->loi_kms < last_off &&
2031 oap2osc_page(last)->ops_srvlock == 0) {
2032 attr->cat_kms = last_off;
2038 cl_object_attr_update(env, obj, attr, valid);
2039 cl_object_attr_unlock(obj);
2041 OBDO_FREE(aa->aa_oa);
2043 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE && rc == 0)
2044 osc_inc_unstable_pages(req);
2046 list_for_each_entry_safe(ext, tmp, &aa->aa_exts, oe_link) {
2047 list_del_init(&ext->oe_link);
2048 osc_extent_finish(env, ext, 1,
2049 rc && req->rq_no_delay ? -EWOULDBLOCK : rc);
2051 LASSERT(list_empty(&aa->aa_exts));
2052 LASSERT(list_empty(&aa->aa_oaps));
2054 transferred = (req->rq_bulk == NULL ? /* short io */
2055 aa->aa_requested_nob :
2056 req->rq_bulk->bd_nob_transferred);
2058 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2059 ptlrpc_lprocfs_brw(req, transferred);
2061 spin_lock(&cli->cl_loi_list_lock);
2062 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2063 * is called so we know whether to go to sync BRWs or wait for more
2064 * RPCs to complete */
2065 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2066 cli->cl_w_in_flight--;
2068 cli->cl_r_in_flight--;
2069 osc_wake_cache_waiters(cli);
2070 spin_unlock(&cli->cl_loi_list_lock);
2072 osc_io_unplug(env, cli, NULL);
2076 static void brw_commit(struct ptlrpc_request *req)
2078 /* If osc_inc_unstable_pages (via osc_extent_finish) races with
2079 * this called via the rq_commit_cb, I need to ensure
2080 * osc_dec_unstable_pages is still called. Otherwise unstable
2081 * pages may be leaked. */
2082 spin_lock(&req->rq_lock);
2083 if (likely(req->rq_unstable)) {
2084 req->rq_unstable = 0;
2085 spin_unlock(&req->rq_lock);
2087 osc_dec_unstable_pages(req);
2089 req->rq_committed = 1;
2090 spin_unlock(&req->rq_lock);
2095 * Build an RPC by the list of extent @ext_list. The caller must ensure
2096 * that the total pages in this list are NOT over max pages per RPC.
2097 * Extents in the list must be in OES_RPC state.
2099 int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
2100 struct list_head *ext_list, int cmd)
2102 struct ptlrpc_request *req = NULL;
2103 struct osc_extent *ext;
2104 struct brw_page **pga = NULL;
2105 struct osc_brw_async_args *aa = NULL;
2106 struct obdo *oa = NULL;
2107 struct osc_async_page *oap;
2108 struct osc_object *obj = NULL;
2109 struct cl_req_attr *crattr = NULL;
2110 loff_t starting_offset = OBD_OBJECT_EOF;
2111 loff_t ending_offset = 0;
2115 bool soft_sync = false;
2116 bool interrupted = false;
2117 bool ndelay = false;
2121 __u32 layout_version = 0;
2122 struct list_head rpc_list = LIST_HEAD_INIT(rpc_list);
2123 struct ost_body *body;
2125 LASSERT(!list_empty(ext_list));
2127 /* add pages into rpc_list to build BRW rpc */
2128 list_for_each_entry(ext, ext_list, oe_link) {
2129 LASSERT(ext->oe_state == OES_RPC);
2130 mem_tight |= ext->oe_memalloc;
2131 grant += ext->oe_grants;
2132 page_count += ext->oe_nr_pages;
2133 layout_version = MAX(layout_version, ext->oe_layout_version);
2138 soft_sync = osc_over_unstable_soft_limit(cli);
2140 mpflag = cfs_memory_pressure_get_and_set();
2142 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2144 GOTO(out, rc = -ENOMEM);
2148 GOTO(out, rc = -ENOMEM);
2151 list_for_each_entry(ext, ext_list, oe_link) {
2152 list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
2154 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
2156 oap->oap_brw_flags |= OBD_BRW_SOFT_SYNC;
2157 pga[i] = &oap->oap_brw_page;
2158 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2161 list_add_tail(&oap->oap_rpc_item, &rpc_list);
2162 if (starting_offset == OBD_OBJECT_EOF ||
2163 starting_offset > oap->oap_obj_off)
2164 starting_offset = oap->oap_obj_off;
2166 LASSERT(oap->oap_page_off == 0);
2167 if (ending_offset < oap->oap_obj_off + oap->oap_count)
2168 ending_offset = oap->oap_obj_off +
2171 LASSERT(oap->oap_page_off + oap->oap_count ==
2173 if (oap->oap_interrupted)
2180 /* first page in the list */
2181 oap = list_entry(rpc_list.next, typeof(*oap), oap_rpc_item);
2183 crattr = &osc_env_info(env)->oti_req_attr;
2184 memset(crattr, 0, sizeof(*crattr));
2185 crattr->cra_type = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2186 crattr->cra_flags = ~0ULL;
2187 crattr->cra_page = oap2cl_page(oap);
2188 crattr->cra_oa = oa;
2189 cl_req_attr_set(env, osc2cl(obj), crattr);
2191 if (cmd == OBD_BRW_WRITE) {
2192 oa->o_grant_used = grant;
2193 if (layout_version > 0) {
2194 CDEBUG(D_LAYOUT, DFID": write with layout version %u\n",
2195 PFID(&oa->o_oi.oi_fid), layout_version);
2197 oa->o_layout_version = layout_version;
2198 oa->o_valid |= OBD_MD_LAYOUT_VERSION;
2202 sort_brw_pages(pga, page_count);
2203 rc = osc_brw_prep_request(cmd, cli, oa, page_count, pga, &req, 0);
2205 CERROR("prep_req failed: %d\n", rc);
2209 req->rq_commit_cb = brw_commit;
2210 req->rq_interpret_reply = brw_interpret;
2211 req->rq_memalloc = mem_tight != 0;
2212 oap->oap_request = ptlrpc_request_addref(req);
2213 if (interrupted && !req->rq_intr)
2214 ptlrpc_mark_interrupted(req);
2216 req->rq_no_resend = req->rq_no_delay = 1;
2217 /* probably set a shorter timeout value.
2218 * to handle ETIMEDOUT in brw_interpret() correctly. */
2219 /* lustre_msg_set_timeout(req, req->rq_timeout / 2); */
2222 /* Need to update the timestamps after the request is built in case
2223 * we race with setattr (locally or in queue at OST). If OST gets
2224 * later setattr before earlier BRW (as determined by the request xid),
2225 * the OST will not use BRW timestamps. Sadly, there is no obvious
2226 * way to do this in a single call. bug 10150 */
2227 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2228 crattr->cra_oa = &body->oa;
2229 crattr->cra_flags = OBD_MD_FLMTIME | OBD_MD_FLCTIME | OBD_MD_FLATIME;
2230 cl_req_attr_set(env, osc2cl(obj), crattr);
2231 lustre_msg_set_jobid(req->rq_reqmsg, crattr->cra_jobid);
2233 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2234 aa = ptlrpc_req_async_args(req);
2235 INIT_LIST_HEAD(&aa->aa_oaps);
2236 list_splice_init(&rpc_list, &aa->aa_oaps);
2237 INIT_LIST_HEAD(&aa->aa_exts);
2238 list_splice_init(ext_list, &aa->aa_exts);
2240 spin_lock(&cli->cl_loi_list_lock);
2241 starting_offset >>= PAGE_SHIFT;
2242 if (cmd == OBD_BRW_READ) {
2243 cli->cl_r_in_flight++;
2244 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2245 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2246 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2247 starting_offset + 1);
2249 cli->cl_w_in_flight++;
2250 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2251 lprocfs_oh_tally(&cli->cl_write_rpc_hist, cli->cl_w_in_flight);
2252 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2253 starting_offset + 1);
2255 spin_unlock(&cli->cl_loi_list_lock);
2257 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %ur/%uw in flight",
2258 page_count, aa, cli->cl_r_in_flight,
2259 cli->cl_w_in_flight);
2260 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_DELAY_IO, cfs_fail_val);
2262 ptlrpcd_add_req(req);
2268 cfs_memory_pressure_restore(mpflag);
2271 LASSERT(req == NULL);
2276 OBD_FREE(pga, sizeof(*pga) * page_count);
2277 /* this should happen rarely and is pretty bad, it makes the
2278 * pending list not follow the dirty order */
2279 while (!list_empty(ext_list)) {
2280 ext = list_entry(ext_list->next, struct osc_extent,
2282 list_del_init(&ext->oe_link);
2283 osc_extent_finish(env, ext, 0, rc);
2289 static int osc_set_lock_data(struct ldlm_lock *lock, void *data)
2293 LASSERT(lock != NULL);
2295 lock_res_and_lock(lock);
2297 if (lock->l_ast_data == NULL)
2298 lock->l_ast_data = data;
2299 if (lock->l_ast_data == data)
2302 unlock_res_and_lock(lock);
2307 int osc_enqueue_fini(struct ptlrpc_request *req, osc_enqueue_upcall_f upcall,
2308 void *cookie, struct lustre_handle *lockh,
2309 enum ldlm_mode mode, __u64 *flags, bool speculative,
2312 bool intent = *flags & LDLM_FL_HAS_INTENT;
2316 /* The request was created before ldlm_cli_enqueue call. */
2317 if (intent && errcode == ELDLM_LOCK_ABORTED) {
2318 struct ldlm_reply *rep;
2320 rep = req_capsule_server_get(&req->rq_pill, &RMF_DLM_REP);
2321 LASSERT(rep != NULL);
2323 rep->lock_policy_res1 =
2324 ptlrpc_status_ntoh(rep->lock_policy_res1);
2325 if (rep->lock_policy_res1)
2326 errcode = rep->lock_policy_res1;
2328 *flags |= LDLM_FL_LVB_READY;
2329 } else if (errcode == ELDLM_OK) {
2330 *flags |= LDLM_FL_LVB_READY;
2333 /* Call the update callback. */
2334 rc = (*upcall)(cookie, lockh, errcode);
2336 /* release the reference taken in ldlm_cli_enqueue() */
2337 if (errcode == ELDLM_LOCK_MATCHED)
2339 if (errcode == ELDLM_OK && lustre_handle_is_used(lockh))
2340 ldlm_lock_decref(lockh, mode);
2345 int osc_enqueue_interpret(const struct lu_env *env, struct ptlrpc_request *req,
2346 struct osc_enqueue_args *aa, int rc)
2348 struct ldlm_lock *lock;
2349 struct lustre_handle *lockh = &aa->oa_lockh;
2350 enum ldlm_mode mode = aa->oa_mode;
2351 struct ost_lvb *lvb = aa->oa_lvb;
2352 __u32 lvb_len = sizeof(*lvb);
2357 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2359 lock = ldlm_handle2lock(lockh);
2360 LASSERTF(lock != NULL,
2361 "lockh %#llx, req %p, aa %p - client evicted?\n",
2362 lockh->cookie, req, aa);
2364 /* Take an additional reference so that a blocking AST that
2365 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2366 * to arrive after an upcall has been executed by
2367 * osc_enqueue_fini(). */
2368 ldlm_lock_addref(lockh, mode);
2370 /* Let cl_lock_state_wait fail with -ERESTARTSYS to unuse sublocks. */
2371 OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_ENQUEUE_HANG, 2);
2373 /* Let CP AST to grant the lock first. */
2374 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
2376 if (aa->oa_speculative) {
2377 LASSERT(aa->oa_lvb == NULL);
2378 LASSERT(aa->oa_flags == NULL);
2379 aa->oa_flags = &flags;
2382 /* Complete obtaining the lock procedure. */
2383 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_type, 1,
2384 aa->oa_mode, aa->oa_flags, lvb, lvb_len,
2386 /* Complete osc stuff. */
2387 rc = osc_enqueue_fini(req, aa->oa_upcall, aa->oa_cookie, lockh, mode,
2388 aa->oa_flags, aa->oa_speculative, rc);
2390 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
2392 ldlm_lock_decref(lockh, mode);
2393 LDLM_LOCK_PUT(lock);
2397 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
2399 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
2400 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
2401 * other synchronous requests, however keeping some locks and trying to obtain
2402 * others may take a considerable amount of time in a case of ost failure; and
2403 * when other sync requests do not get released lock from a client, the client
2404 * is evicted from the cluster -- such scenarious make the life difficult, so
2405 * release locks just after they are obtained. */
2406 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2407 __u64 *flags, union ldlm_policy_data *policy,
2408 struct ost_lvb *lvb, int kms_valid,
2409 osc_enqueue_upcall_f upcall, void *cookie,
2410 struct ldlm_enqueue_info *einfo,
2411 struct ptlrpc_request_set *rqset, int async,
2414 struct obd_device *obd = exp->exp_obd;
2415 struct lustre_handle lockh = { 0 };
2416 struct ptlrpc_request *req = NULL;
2417 int intent = *flags & LDLM_FL_HAS_INTENT;
2418 __u64 match_flags = *flags;
2419 enum ldlm_mode mode;
2423 /* Filesystem lock extents are extended to page boundaries so that
2424 * dealing with the page cache is a little smoother. */
2425 policy->l_extent.start -= policy->l_extent.start & ~PAGE_MASK;
2426 policy->l_extent.end |= ~PAGE_MASK;
2429 * kms is not valid when either object is completely fresh (so that no
2430 * locks are cached), or object was evicted. In the latter case cached
2431 * lock cannot be used, because it would prime inode state with
2432 * potentially stale LVB.
2437 /* Next, search for already existing extent locks that will cover us */
2438 /* If we're trying to read, we also search for an existing PW lock. The
2439 * VFS and page cache already protect us locally, so lots of readers/
2440 * writers can share a single PW lock.
2442 * There are problems with conversion deadlocks, so instead of
2443 * converting a read lock to a write lock, we'll just enqueue a new
2446 * At some point we should cancel the read lock instead of making them
2447 * send us a blocking callback, but there are problems with canceling
2448 * locks out from other users right now, too. */
2449 mode = einfo->ei_mode;
2450 if (einfo->ei_mode == LCK_PR)
2452 /* Normal lock requests must wait for the LVB to be ready before
2453 * matching a lock; speculative lock requests do not need to,
2454 * because they will not actually use the lock. */
2456 match_flags |= LDLM_FL_LVB_READY;
2458 match_flags |= LDLM_FL_BLOCK_GRANTED;
2459 mode = ldlm_lock_match(obd->obd_namespace, match_flags, res_id,
2460 einfo->ei_type, policy, mode, &lockh, 0);
2462 struct ldlm_lock *matched;
2464 if (*flags & LDLM_FL_TEST_LOCK)
2467 matched = ldlm_handle2lock(&lockh);
2469 /* This DLM lock request is speculative, and does not
2470 * have an associated IO request. Therefore if there
2471 * is already a DLM lock, it wll just inform the
2472 * caller to cancel the request for this stripe.*/
2473 lock_res_and_lock(matched);
2474 if (ldlm_extent_equal(&policy->l_extent,
2475 &matched->l_policy_data.l_extent))
2479 unlock_res_and_lock(matched);
2481 ldlm_lock_decref(&lockh, mode);
2482 LDLM_LOCK_PUT(matched);
2484 } else if (osc_set_lock_data(matched, einfo->ei_cbdata)) {
2485 *flags |= LDLM_FL_LVB_READY;
2487 /* We already have a lock, and it's referenced. */
2488 (*upcall)(cookie, &lockh, ELDLM_LOCK_MATCHED);
2490 ldlm_lock_decref(&lockh, mode);
2491 LDLM_LOCK_PUT(matched);
2494 ldlm_lock_decref(&lockh, mode);
2495 LDLM_LOCK_PUT(matched);
2500 if (*flags & (LDLM_FL_TEST_LOCK | LDLM_FL_MATCH_LOCK))
2504 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2505 &RQF_LDLM_ENQUEUE_LVB);
2509 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
2511 ptlrpc_request_free(req);
2515 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
2517 ptlrpc_request_set_replen(req);
2520 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
2521 *flags &= ~LDLM_FL_BLOCK_GRANTED;
2523 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
2524 sizeof(*lvb), LVB_T_OST, &lockh, async);
2527 struct osc_enqueue_args *aa;
2528 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2529 aa = ptlrpc_req_async_args(req);
2531 aa->oa_mode = einfo->ei_mode;
2532 aa->oa_type = einfo->ei_type;
2533 lustre_handle_copy(&aa->oa_lockh, &lockh);
2534 aa->oa_upcall = upcall;
2535 aa->oa_cookie = cookie;
2536 aa->oa_speculative = speculative;
2538 aa->oa_flags = flags;
2541 /* speculative locks are essentially to enqueue
2542 * a DLM lock in advance, so we don't care
2543 * about the result of the enqueue. */
2545 aa->oa_flags = NULL;
2548 req->rq_interpret_reply =
2549 (ptlrpc_interpterer_t)osc_enqueue_interpret;
2550 if (rqset == PTLRPCD_SET)
2551 ptlrpcd_add_req(req);
2553 ptlrpc_set_add_req(rqset, req);
2554 } else if (intent) {
2555 ptlrpc_req_finished(req);
2560 rc = osc_enqueue_fini(req, upcall, cookie, &lockh, einfo->ei_mode,
2561 flags, speculative, rc);
2563 ptlrpc_req_finished(req);
2568 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2569 enum ldlm_type type, union ldlm_policy_data *policy,
2570 enum ldlm_mode mode, __u64 *flags, void *data,
2571 struct lustre_handle *lockh, int unref)
2573 struct obd_device *obd = exp->exp_obd;
2574 __u64 lflags = *flags;
2578 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
2581 /* Filesystem lock extents are extended to page boundaries so that
2582 * dealing with the page cache is a little smoother */
2583 policy->l_extent.start -= policy->l_extent.start & ~PAGE_MASK;
2584 policy->l_extent.end |= ~PAGE_MASK;
2586 /* Next, search for already existing extent locks that will cover us */
2587 /* If we're trying to read, we also search for an existing PW lock. The
2588 * VFS and page cache already protect us locally, so lots of readers/
2589 * writers can share a single PW lock. */
2593 rc = ldlm_lock_match(obd->obd_namespace, lflags,
2594 res_id, type, policy, rc, lockh, unref);
2595 if (rc == 0 || lflags & LDLM_FL_TEST_LOCK)
2599 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2601 LASSERT(lock != NULL);
2602 if (!osc_set_lock_data(lock, data)) {
2603 ldlm_lock_decref(lockh, rc);
2606 LDLM_LOCK_PUT(lock);
2611 static int osc_statfs_interpret(const struct lu_env *env,
2612 struct ptlrpc_request *req,
2613 struct osc_async_args *aa, int rc)
2615 struct obd_statfs *msfs;
2619 /* The request has in fact never been sent
2620 * due to issues at a higher level (LOV).
2621 * Exit immediately since the caller is
2622 * aware of the problem and takes care
2623 * of the clean up */
2626 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
2627 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
2633 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2635 GOTO(out, rc = -EPROTO);
2638 *aa->aa_oi->oi_osfs = *msfs;
2640 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
2644 static int osc_statfs_async(struct obd_export *exp,
2645 struct obd_info *oinfo, time64_t max_age,
2646 struct ptlrpc_request_set *rqset)
2648 struct obd_device *obd = class_exp2obd(exp);
2649 struct ptlrpc_request *req;
2650 struct osc_async_args *aa;
2654 /* We could possibly pass max_age in the request (as an absolute
2655 * timestamp or a "seconds.usec ago") so the target can avoid doing
2656 * extra calls into the filesystem if that isn't necessary (e.g.
2657 * during mount that would help a bit). Having relative timestamps
2658 * is not so great if request processing is slow, while absolute
2659 * timestamps are not ideal because they need time synchronization. */
2660 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
2664 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2666 ptlrpc_request_free(req);
2669 ptlrpc_request_set_replen(req);
2670 req->rq_request_portal = OST_CREATE_PORTAL;
2671 ptlrpc_at_set_req_timeout(req);
2673 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
2674 /* procfs requests not want stat in wait for avoid deadlock */
2675 req->rq_no_resend = 1;
2676 req->rq_no_delay = 1;
2679 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
2680 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2681 aa = ptlrpc_req_async_args(req);
2684 ptlrpc_set_add_req(rqset, req);
2688 static int osc_statfs(const struct lu_env *env, struct obd_export *exp,
2689 struct obd_statfs *osfs, time64_t max_age, __u32 flags)
2691 struct obd_device *obd = class_exp2obd(exp);
2692 struct obd_statfs *msfs;
2693 struct ptlrpc_request *req;
2694 struct obd_import *imp = NULL;
2698 /*Since the request might also come from lprocfs, so we need
2699 *sync this with client_disconnect_export Bug15684*/
2700 down_read(&obd->u.cli.cl_sem);
2701 if (obd->u.cli.cl_import)
2702 imp = class_import_get(obd->u.cli.cl_import);
2703 up_read(&obd->u.cli.cl_sem);
2707 /* We could possibly pass max_age in the request (as an absolute
2708 * timestamp or a "seconds.usec ago") so the target can avoid doing
2709 * extra calls into the filesystem if that isn't necessary (e.g.
2710 * during mount that would help a bit). Having relative timestamps
2711 * is not so great if request processing is slow, while absolute
2712 * timestamps are not ideal because they need time synchronization. */
2713 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
2715 class_import_put(imp);
2720 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2722 ptlrpc_request_free(req);
2725 ptlrpc_request_set_replen(req);
2726 req->rq_request_portal = OST_CREATE_PORTAL;
2727 ptlrpc_at_set_req_timeout(req);
2729 if (flags & OBD_STATFS_NODELAY) {
2730 /* procfs requests not want stat in wait for avoid deadlock */
2731 req->rq_no_resend = 1;
2732 req->rq_no_delay = 1;
2735 rc = ptlrpc_queue_wait(req);
2739 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2741 GOTO(out, rc = -EPROTO);
2748 ptlrpc_req_finished(req);
2752 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
2753 void *karg, void __user *uarg)
2755 struct obd_device *obd = exp->exp_obd;
2756 struct obd_ioctl_data *data = karg;
2760 if (!try_module_get(THIS_MODULE)) {
2761 CERROR("%s: cannot get module '%s'\n", obd->obd_name,
2762 module_name(THIS_MODULE));
2766 case OBD_IOC_CLIENT_RECOVER:
2767 err = ptlrpc_recover_import(obd->u.cli.cl_import,
2768 data->ioc_inlbuf1, 0);
2772 case IOC_OSC_SET_ACTIVE:
2773 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
2776 case OBD_IOC_PING_TARGET:
2777 err = ptlrpc_obd_ping(obd);
2780 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
2781 cmd, current_comm());
2782 GOTO(out, err = -ENOTTY);
2785 module_put(THIS_MODULE);
2789 int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
2790 u32 keylen, void *key, u32 vallen, void *val,
2791 struct ptlrpc_request_set *set)
2793 struct ptlrpc_request *req;
2794 struct obd_device *obd = exp->exp_obd;
2795 struct obd_import *imp = class_exp2cliimp(exp);
2800 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
2802 if (KEY_IS(KEY_CHECKSUM)) {
2803 if (vallen != sizeof(int))
2805 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
2809 if (KEY_IS(KEY_SPTLRPC_CONF)) {
2810 sptlrpc_conf_client_adapt(obd);
2814 if (KEY_IS(KEY_FLUSH_CTX)) {
2815 sptlrpc_import_flush_my_ctx(imp);
2819 if (KEY_IS(KEY_CACHE_SET)) {
2820 struct client_obd *cli = &obd->u.cli;
2822 LASSERT(cli->cl_cache == NULL); /* only once */
2823 cli->cl_cache = (struct cl_client_cache *)val;
2824 cl_cache_incref(cli->cl_cache);
2825 cli->cl_lru_left = &cli->cl_cache->ccc_lru_left;
2827 /* add this osc into entity list */
2828 LASSERT(list_empty(&cli->cl_lru_osc));
2829 spin_lock(&cli->cl_cache->ccc_lru_lock);
2830 list_add(&cli->cl_lru_osc, &cli->cl_cache->ccc_lru);
2831 spin_unlock(&cli->cl_cache->ccc_lru_lock);
2836 if (KEY_IS(KEY_CACHE_LRU_SHRINK)) {
2837 struct client_obd *cli = &obd->u.cli;
2838 long nr = atomic_long_read(&cli->cl_lru_in_list) >> 1;
2839 long target = *(long *)val;
2841 nr = osc_lru_shrink(env, cli, min(nr, target), true);
2846 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
2849 /* We pass all other commands directly to OST. Since nobody calls osc
2850 methods directly and everybody is supposed to go through LOV, we
2851 assume lov checked invalid values for us.
2852 The only recognised values so far are evict_by_nid and mds_conn.
2853 Even if something bad goes through, we'd get a -EINVAL from OST
2856 req = ptlrpc_request_alloc(imp, KEY_IS(KEY_GRANT_SHRINK) ?
2857 &RQF_OST_SET_GRANT_INFO :
2862 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
2863 RCL_CLIENT, keylen);
2864 if (!KEY_IS(KEY_GRANT_SHRINK))
2865 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
2866 RCL_CLIENT, vallen);
2867 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
2869 ptlrpc_request_free(req);
2873 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
2874 memcpy(tmp, key, keylen);
2875 tmp = req_capsule_client_get(&req->rq_pill, KEY_IS(KEY_GRANT_SHRINK) ?
2878 memcpy(tmp, val, vallen);
2880 if (KEY_IS(KEY_GRANT_SHRINK)) {
2881 struct osc_grant_args *aa;
2884 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2885 aa = ptlrpc_req_async_args(req);
2888 ptlrpc_req_finished(req);
2891 *oa = ((struct ost_body *)val)->oa;
2893 req->rq_interpret_reply = osc_shrink_grant_interpret;
2896 ptlrpc_request_set_replen(req);
2897 if (!KEY_IS(KEY_GRANT_SHRINK)) {
2898 LASSERT(set != NULL);
2899 ptlrpc_set_add_req(set, req);
2900 ptlrpc_check_set(NULL, set);
2902 ptlrpcd_add_req(req);
2907 EXPORT_SYMBOL(osc_set_info_async);
2909 int osc_reconnect(const struct lu_env *env, struct obd_export *exp,
2910 struct obd_device *obd, struct obd_uuid *cluuid,
2911 struct obd_connect_data *data, void *localdata)
2913 struct client_obd *cli = &obd->u.cli;
2915 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
2919 spin_lock(&cli->cl_loi_list_lock);
2920 grant = cli->cl_avail_grant + cli->cl_reserved_grant;
2921 if (data->ocd_connect_flags & OBD_CONNECT_GRANT_PARAM)
2922 grant += cli->cl_dirty_grant;
2924 grant += cli->cl_dirty_pages << PAGE_SHIFT;
2925 data->ocd_grant = grant ? : 2 * cli_brw_size(obd);
2926 lost_grant = cli->cl_lost_grant;
2927 cli->cl_lost_grant = 0;
2928 spin_unlock(&cli->cl_loi_list_lock);
2930 CDEBUG(D_RPCTRACE, "ocd_connect_flags: %#llx ocd_version: %d"
2931 " ocd_grant: %d, lost: %ld.\n", data->ocd_connect_flags,
2932 data->ocd_version, data->ocd_grant, lost_grant);
2937 EXPORT_SYMBOL(osc_reconnect);
2939 int osc_disconnect(struct obd_export *exp)
2941 struct obd_device *obd = class_exp2obd(exp);
2944 rc = client_disconnect_export(exp);
2946 * Initially we put del_shrink_grant before disconnect_export, but it
2947 * causes the following problem if setup (connect) and cleanup
2948 * (disconnect) are tangled together.
2949 * connect p1 disconnect p2
2950 * ptlrpc_connect_import
2951 * ............... class_manual_cleanup
2954 * ptlrpc_connect_interrupt
2956 * add this client to shrink list
2958 * Bang! pinger trigger the shrink.
2959 * So the osc should be disconnected from the shrink list, after we
2960 * are sure the import has been destroyed. BUG18662
2962 if (obd->u.cli.cl_import == NULL)
2963 osc_del_shrink_grant(&obd->u.cli);
2966 EXPORT_SYMBOL(osc_disconnect);
2968 int osc_ldlm_resource_invalidate(struct cfs_hash *hs, struct cfs_hash_bd *bd,
2969 struct hlist_node *hnode, void *arg)
2971 struct lu_env *env = arg;
2972 struct ldlm_resource *res = cfs_hash_object(hs, hnode);
2973 struct ldlm_lock *lock;
2974 struct osc_object *osc = NULL;
2978 list_for_each_entry(lock, &res->lr_granted, l_res_link) {
2979 if (lock->l_ast_data != NULL && osc == NULL) {
2980 osc = lock->l_ast_data;
2981 cl_object_get(osc2cl(osc));
2984 /* clear LDLM_FL_CLEANED flag to make sure it will be canceled
2985 * by the 2nd round of ldlm_namespace_clean() call in
2986 * osc_import_event(). */
2987 ldlm_clear_cleaned(lock);
2992 osc_object_invalidate(env, osc);
2993 cl_object_put(env, osc2cl(osc));
2998 EXPORT_SYMBOL(osc_ldlm_resource_invalidate);
3000 static int osc_import_event(struct obd_device *obd,
3001 struct obd_import *imp,
3002 enum obd_import_event event)
3004 struct client_obd *cli;
3008 LASSERT(imp->imp_obd == obd);
3011 case IMP_EVENT_DISCON: {
3013 spin_lock(&cli->cl_loi_list_lock);
3014 cli->cl_avail_grant = 0;
3015 cli->cl_lost_grant = 0;
3016 spin_unlock(&cli->cl_loi_list_lock);
3019 case IMP_EVENT_INACTIVE: {
3020 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE);
3023 case IMP_EVENT_INVALIDATE: {
3024 struct ldlm_namespace *ns = obd->obd_namespace;
3028 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3030 env = cl_env_get(&refcheck);
3032 osc_io_unplug(env, &obd->u.cli, NULL);
3034 cfs_hash_for_each_nolock(ns->ns_rs_hash,
3035 osc_ldlm_resource_invalidate,
3037 cl_env_put(env, &refcheck);
3039 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3044 case IMP_EVENT_ACTIVE: {
3045 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE);
3048 case IMP_EVENT_OCD: {
3049 struct obd_connect_data *ocd = &imp->imp_connect_data;
3051 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
3052 osc_init_grant(&obd->u.cli, ocd);
3055 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
3056 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
3058 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD);
3061 case IMP_EVENT_DEACTIVATE: {
3062 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DEACTIVATE);
3065 case IMP_EVENT_ACTIVATE: {
3066 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVATE);
3070 CERROR("Unknown import event %d\n", event);
3077 * Determine whether the lock can be canceled before replaying the lock
3078 * during recovery, see bug16774 for detailed information.
3080 * \retval zero the lock can't be canceled
3081 * \retval other ok to cancel
3083 static int osc_cancel_weight(struct ldlm_lock *lock)
3086 * Cancel all unused and granted extent lock.
3088 if (lock->l_resource->lr_type == LDLM_EXTENT &&
3089 lock->l_granted_mode == lock->l_req_mode &&
3090 osc_ldlm_weigh_ast(lock) == 0)
3096 static int brw_queue_work(const struct lu_env *env, void *data)
3098 struct client_obd *cli = data;
3100 CDEBUG(D_CACHE, "Run writeback work for client obd %p.\n", cli);
3102 osc_io_unplug(env, cli, NULL);
3106 int osc_setup_common(struct obd_device *obd, struct lustre_cfg *lcfg)
3108 struct client_obd *cli = &obd->u.cli;
3114 rc = ptlrpcd_addref();
3118 rc = client_obd_setup(obd, lcfg);
3120 GOTO(out_ptlrpcd, rc);
3123 handler = ptlrpcd_alloc_work(cli->cl_import, brw_queue_work, cli);
3124 if (IS_ERR(handler))
3125 GOTO(out_ptlrpcd_work, rc = PTR_ERR(handler));
3126 cli->cl_writeback_work = handler;
3128 handler = ptlrpcd_alloc_work(cli->cl_import, lru_queue_work, cli);
3129 if (IS_ERR(handler))
3130 GOTO(out_ptlrpcd_work, rc = PTR_ERR(handler));
3131 cli->cl_lru_work = handler;
3133 rc = osc_quota_setup(obd);
3135 GOTO(out_ptlrpcd_work, rc);
3137 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
3139 INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
3143 if (cli->cl_writeback_work != NULL) {
3144 ptlrpcd_destroy_work(cli->cl_writeback_work);
3145 cli->cl_writeback_work = NULL;
3147 if (cli->cl_lru_work != NULL) {
3148 ptlrpcd_destroy_work(cli->cl_lru_work);
3149 cli->cl_lru_work = NULL;
3151 client_obd_cleanup(obd);
3156 EXPORT_SYMBOL(osc_setup_common);
3158 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
3160 struct client_obd *cli = &obd->u.cli;
3168 rc = osc_setup_common(obd, lcfg);
3172 rc = osc_tunables_init(obd);
3177 * We try to control the total number of requests with a upper limit
3178 * osc_reqpool_maxreqcount. There might be some race which will cause
3179 * over-limit allocation, but it is fine.
3181 req_count = atomic_read(&osc_pool_req_count);
3182 if (req_count < osc_reqpool_maxreqcount) {
3183 adding = cli->cl_max_rpcs_in_flight + 2;
3184 if (req_count + adding > osc_reqpool_maxreqcount)
3185 adding = osc_reqpool_maxreqcount - req_count;
3187 added = ptlrpc_add_rqs_to_pool(osc_rq_pool, adding);
3188 atomic_add(added, &osc_pool_req_count);
3191 INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
3192 ns_register_cancel(obd->obd_namespace, osc_cancel_weight);
3194 spin_lock(&osc_shrink_lock);
3195 list_add_tail(&cli->cl_shrink_list, &osc_shrink_list);
3196 spin_unlock(&osc_shrink_lock);
3201 int osc_precleanup_common(struct obd_device *obd)
3203 struct client_obd *cli = &obd->u.cli;
3207 * for echo client, export may be on zombie list, wait for
3208 * zombie thread to cull it, because cli.cl_import will be
3209 * cleared in client_disconnect_export():
3210 * class_export_destroy() -> obd_cleanup() ->
3211 * echo_device_free() -> echo_client_cleanup() ->
3212 * obd_disconnect() -> osc_disconnect() ->
3213 * client_disconnect_export()
3215 obd_zombie_barrier();
3216 if (cli->cl_writeback_work) {
3217 ptlrpcd_destroy_work(cli->cl_writeback_work);
3218 cli->cl_writeback_work = NULL;
3221 if (cli->cl_lru_work) {
3222 ptlrpcd_destroy_work(cli->cl_lru_work);
3223 cli->cl_lru_work = NULL;
3226 obd_cleanup_client_import(obd);
3229 EXPORT_SYMBOL(osc_precleanup_common);
3231 static int osc_precleanup(struct obd_device *obd)
3235 osc_precleanup_common(obd);
3237 ptlrpc_lprocfs_unregister_obd(obd);
3241 int osc_cleanup_common(struct obd_device *obd)
3243 struct client_obd *cli = &obd->u.cli;
3248 spin_lock(&osc_shrink_lock);
3249 list_del(&cli->cl_shrink_list);
3250 spin_unlock(&osc_shrink_lock);
3253 if (cli->cl_cache != NULL) {
3254 LASSERT(atomic_read(&cli->cl_cache->ccc_users) > 0);
3255 spin_lock(&cli->cl_cache->ccc_lru_lock);
3256 list_del_init(&cli->cl_lru_osc);
3257 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3258 cli->cl_lru_left = NULL;
3259 cl_cache_decref(cli->cl_cache);
3260 cli->cl_cache = NULL;
3263 /* free memory of osc quota cache */
3264 osc_quota_cleanup(obd);
3266 rc = client_obd_cleanup(obd);
3271 EXPORT_SYMBOL(osc_cleanup_common);
3273 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
3275 ssize_t count = class_modify_config(lcfg, PARAM_OSC,
3276 &obd->obd_kset.kobj);
3277 return count > 0 ? 0 : count;
3280 static int osc_process_config(struct obd_device *obd, size_t len, void *buf)
3282 return osc_process_config_base(obd, buf);
3285 static struct obd_ops osc_obd_ops = {
3286 .o_owner = THIS_MODULE,
3287 .o_setup = osc_setup,
3288 .o_precleanup = osc_precleanup,
3289 .o_cleanup = osc_cleanup_common,
3290 .o_add_conn = client_import_add_conn,
3291 .o_del_conn = client_import_del_conn,
3292 .o_connect = client_connect_import,
3293 .o_reconnect = osc_reconnect,
3294 .o_disconnect = osc_disconnect,
3295 .o_statfs = osc_statfs,
3296 .o_statfs_async = osc_statfs_async,
3297 .o_create = osc_create,
3298 .o_destroy = osc_destroy,
3299 .o_getattr = osc_getattr,
3300 .o_setattr = osc_setattr,
3301 .o_iocontrol = osc_iocontrol,
3302 .o_set_info_async = osc_set_info_async,
3303 .o_import_event = osc_import_event,
3304 .o_process_config = osc_process_config,
3305 .o_quotactl = osc_quotactl,
3308 static struct shrinker *osc_cache_shrinker;
3309 struct list_head osc_shrink_list = LIST_HEAD_INIT(osc_shrink_list);
3310 DEFINE_SPINLOCK(osc_shrink_lock);
3312 #ifndef HAVE_SHRINKER_COUNT
3313 static int osc_cache_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask))
3315 struct shrink_control scv = {
3316 .nr_to_scan = shrink_param(sc, nr_to_scan),
3317 .gfp_mask = shrink_param(sc, gfp_mask)
3319 #if !defined(HAVE_SHRINKER_WANT_SHRINK_PTR) && !defined(HAVE_SHRINK_CONTROL)
3320 struct shrinker *shrinker = NULL;
3323 (void)osc_cache_shrink_scan(shrinker, &scv);
3325 return osc_cache_shrink_count(shrinker, &scv);
3329 static int __init osc_init(void)
3331 bool enable_proc = true;
3332 struct obd_type *type;
3333 unsigned int reqpool_size;
3334 unsigned int reqsize;
3336 DEF_SHRINKER_VAR(osc_shvar, osc_cache_shrink,
3337 osc_cache_shrink_count, osc_cache_shrink_scan);
3340 /* print an address of _any_ initialized kernel symbol from this
3341 * module, to allow debugging with gdb that doesn't support data
3342 * symbols from modules.*/
3343 CDEBUG(D_INFO, "Lustre OSC module (%p).\n", &osc_caches);
3345 rc = lu_kmem_init(osc_caches);
3349 type = class_search_type(LUSTRE_OSP_NAME);
3350 if (type != NULL && type->typ_procsym != NULL)
3351 enable_proc = false;
3353 rc = class_register_type(&osc_obd_ops, NULL, enable_proc, NULL,
3354 LUSTRE_OSC_NAME, &osc_device_type);
3358 osc_cache_shrinker = set_shrinker(DEFAULT_SEEKS, &osc_shvar);
3360 /* This is obviously too much memory, only prevent overflow here */
3361 if (osc_reqpool_mem_max >= 1 << 12 || osc_reqpool_mem_max == 0)
3362 GOTO(out_type, rc = -EINVAL);
3364 reqpool_size = osc_reqpool_mem_max << 20;
3367 while (reqsize < OST_IO_MAXREQSIZE)
3368 reqsize = reqsize << 1;
3371 * We don't enlarge the request count in OSC pool according to
3372 * cl_max_rpcs_in_flight. The allocation from the pool will only be
3373 * tried after normal allocation failed. So a small OSC pool won't
3374 * cause much performance degression in most of cases.
3376 osc_reqpool_maxreqcount = reqpool_size / reqsize;
3378 atomic_set(&osc_pool_req_count, 0);
3379 osc_rq_pool = ptlrpc_init_rq_pool(0, OST_IO_MAXREQSIZE,
3380 ptlrpc_add_rqs_to_pool);
3382 if (osc_rq_pool != NULL)
3386 class_unregister_type(LUSTRE_OSC_NAME);
3388 lu_kmem_fini(osc_caches);
3393 static void __exit osc_exit(void)
3395 remove_shrinker(osc_cache_shrinker);
3396 class_unregister_type(LUSTRE_OSC_NAME);
3397 lu_kmem_fini(osc_caches);
3398 ptlrpc_free_rq_pool(osc_rq_pool);
3401 MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
3402 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
3403 MODULE_VERSION(LUSTRE_VERSION_STRING);
3404 MODULE_LICENSE("GPL");
3406 module_init(osc_init);
3407 module_exit(osc_exit);