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 <linux/workqueue.h>
36 #include <libcfs/libcfs.h>
37 #include <linux/falloc.h>
38 #include <lprocfs_status.h>
39 #include <lustre_debug.h>
40 #include <lustre_dlm.h>
41 #include <lustre_fid.h>
42 #include <lustre_ha.h>
43 #include <uapi/linux/lustre/lustre_ioctl.h>
44 #include <lustre_net.h>
45 #include <lustre_obdo.h>
47 #include <obd_cksum.h>
48 #include <obd_class.h>
49 #include <lustre_osc.h>
50 #include <linux/falloc.h>
52 #include "osc_internal.h"
54 atomic_t osc_pool_req_count;
55 unsigned int osc_reqpool_maxreqcount;
56 struct ptlrpc_request_pool *osc_rq_pool;
58 /* max memory used for request pool, unit is MB */
59 static unsigned int osc_reqpool_mem_max = 5;
60 module_param(osc_reqpool_mem_max, uint, 0444);
62 static int osc_idle_timeout = 20;
63 module_param(osc_idle_timeout, uint, 0644);
65 #define osc_grant_args osc_brw_async_args
67 struct osc_setattr_args {
69 obd_enqueue_update_f sa_upcall;
73 struct osc_fsync_args {
74 struct osc_object *fa_obj;
76 obd_enqueue_update_f fa_upcall;
80 struct osc_ladvise_args {
82 obd_enqueue_update_f la_upcall;
86 static void osc_release_ppga(struct brw_page **ppga, size_t count);
87 static int brw_interpret(const struct lu_env *env, struct ptlrpc_request *req,
90 void osc_pack_req_body(struct ptlrpc_request *req, struct obdo *oa)
92 struct ost_body *body;
94 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
97 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
100 static int osc_getattr(const struct lu_env *env, struct obd_export *exp,
103 struct ptlrpc_request *req;
104 struct ost_body *body;
108 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
112 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
114 ptlrpc_request_free(req);
118 osc_pack_req_body(req, oa);
120 ptlrpc_request_set_replen(req);
122 rc = ptlrpc_queue_wait(req);
126 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
128 GOTO(out, rc = -EPROTO);
130 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
131 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
133 oa->o_blksize = cli_brw_size(exp->exp_obd);
134 oa->o_valid |= OBD_MD_FLBLKSZ;
138 ptlrpc_req_finished(req);
143 static int osc_setattr(const struct lu_env *env, struct obd_export *exp,
146 struct ptlrpc_request *req;
147 struct ost_body *body;
151 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
153 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
157 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
159 ptlrpc_request_free(req);
163 osc_pack_req_body(req, oa);
165 ptlrpc_request_set_replen(req);
167 rc = ptlrpc_queue_wait(req);
171 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
173 GOTO(out, rc = -EPROTO);
175 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
179 ptlrpc_req_finished(req);
184 static int osc_setattr_interpret(const struct lu_env *env,
185 struct ptlrpc_request *req, void *args, int rc)
187 struct osc_setattr_args *sa = args;
188 struct ost_body *body;
195 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
197 GOTO(out, rc = -EPROTO);
199 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, sa->sa_oa,
202 rc = sa->sa_upcall(sa->sa_cookie, rc);
206 int osc_setattr_async(struct obd_export *exp, struct obdo *oa,
207 obd_enqueue_update_f upcall, void *cookie,
208 struct ptlrpc_request_set *rqset)
210 struct ptlrpc_request *req;
211 struct osc_setattr_args *sa;
216 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
220 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
222 ptlrpc_request_free(req);
226 osc_pack_req_body(req, oa);
228 ptlrpc_request_set_replen(req);
230 /* do mds to ost setattr asynchronously */
232 /* Do not wait for response. */
233 ptlrpcd_add_req(req);
235 req->rq_interpret_reply = osc_setattr_interpret;
237 sa = ptlrpc_req_async_args(sa, req);
239 sa->sa_upcall = upcall;
240 sa->sa_cookie = cookie;
242 ptlrpc_set_add_req(rqset, req);
248 static int osc_ladvise_interpret(const struct lu_env *env,
249 struct ptlrpc_request *req,
252 struct osc_ladvise_args *la = arg;
253 struct ost_body *body;
259 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
261 GOTO(out, rc = -EPROTO);
263 *la->la_oa = body->oa;
265 rc = la->la_upcall(la->la_cookie, rc);
270 * If rqset is NULL, do not wait for response. Upcall and cookie could also
271 * be NULL in this case
273 int osc_ladvise_base(struct obd_export *exp, struct obdo *oa,
274 struct ladvise_hdr *ladvise_hdr,
275 obd_enqueue_update_f upcall, void *cookie,
276 struct ptlrpc_request_set *rqset)
278 struct ptlrpc_request *req;
279 struct ost_body *body;
280 struct osc_ladvise_args *la;
282 struct lu_ladvise *req_ladvise;
283 struct lu_ladvise *ladvise = ladvise_hdr->lah_advise;
284 int num_advise = ladvise_hdr->lah_count;
285 struct ladvise_hdr *req_ladvise_hdr;
288 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_LADVISE);
292 req_capsule_set_size(&req->rq_pill, &RMF_OST_LADVISE, RCL_CLIENT,
293 num_advise * sizeof(*ladvise));
294 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_LADVISE);
296 ptlrpc_request_free(req);
299 req->rq_request_portal = OST_IO_PORTAL;
300 ptlrpc_at_set_req_timeout(req);
302 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
304 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
307 req_ladvise_hdr = req_capsule_client_get(&req->rq_pill,
308 &RMF_OST_LADVISE_HDR);
309 memcpy(req_ladvise_hdr, ladvise_hdr, sizeof(*ladvise_hdr));
311 req_ladvise = req_capsule_client_get(&req->rq_pill, &RMF_OST_LADVISE);
312 memcpy(req_ladvise, ladvise, sizeof(*ladvise) * num_advise);
313 ptlrpc_request_set_replen(req);
316 /* Do not wait for response. */
317 ptlrpcd_add_req(req);
321 req->rq_interpret_reply = osc_ladvise_interpret;
322 la = ptlrpc_req_async_args(la, req);
324 la->la_upcall = upcall;
325 la->la_cookie = cookie;
327 ptlrpc_set_add_req(rqset, req);
332 static int osc_create(const struct lu_env *env, struct obd_export *exp,
335 struct ptlrpc_request *req;
336 struct ost_body *body;
341 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
342 LASSERT(fid_seq_is_echo(ostid_seq(&oa->o_oi)));
344 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
346 GOTO(out, rc = -ENOMEM);
348 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
350 ptlrpc_request_free(req);
354 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
357 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
359 ptlrpc_request_set_replen(req);
361 rc = ptlrpc_queue_wait(req);
365 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
367 GOTO(out_req, rc = -EPROTO);
369 CDEBUG(D_INFO, "oa flags %x\n", oa->o_flags);
370 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
372 oa->o_blksize = cli_brw_size(exp->exp_obd);
373 oa->o_valid |= OBD_MD_FLBLKSZ;
375 CDEBUG(D_HA, "transno: %lld\n",
376 lustre_msg_get_transno(req->rq_repmsg));
378 ptlrpc_req_finished(req);
383 int osc_punch_send(struct obd_export *exp, struct obdo *oa,
384 obd_enqueue_update_f upcall, void *cookie)
386 struct ptlrpc_request *req;
387 struct osc_setattr_args *sa;
388 struct obd_import *imp = class_exp2cliimp(exp);
389 struct ost_body *body;
394 req = ptlrpc_request_alloc(imp, &RQF_OST_PUNCH);
398 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
400 ptlrpc_request_free(req);
404 osc_set_io_portal(req);
406 ptlrpc_at_set_req_timeout(req);
408 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
410 lustre_set_wire_obdo(&imp->imp_connect_data, &body->oa, oa);
412 ptlrpc_request_set_replen(req);
414 req->rq_interpret_reply = osc_setattr_interpret;
415 sa = ptlrpc_req_async_args(sa, req);
417 sa->sa_upcall = upcall;
418 sa->sa_cookie = cookie;
420 ptlrpcd_add_req(req);
424 EXPORT_SYMBOL(osc_punch_send);
427 * osc_fallocate_base() - Handles fallocate request.
429 * @exp: Export structure
430 * @oa: Attributes passed to OSS from client (obdo structure)
431 * @upcall: Primary & supplementary group information
432 * @cookie: Exclusive identifier
433 * @rqset: Request list.
434 * @mode: Operation done on given range.
436 * osc_fallocate_base() - Handles fallocate requests only. Only block
437 * allocation or standard preallocate operation is supported currently.
438 * Other mode flags is not supported yet. ftruncate(2) or truncate(2)
439 * is supported via SETATTR request.
441 * Return: Non-zero on failure and O on success.
443 int osc_fallocate_base(struct obd_export *exp, struct obdo *oa,
444 obd_enqueue_update_f upcall, void *cookie, int mode)
446 struct ptlrpc_request *req;
447 struct osc_setattr_args *sa;
448 struct ost_body *body;
449 struct obd_import *imp = class_exp2cliimp(exp);
454 * Only mode == 0 (which is standard prealloc) is supported now.
455 * Punch is not supported yet.
457 if (mode & ~FALLOC_FL_KEEP_SIZE)
459 oa->o_falloc_mode = mode;
461 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
466 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_FALLOCATE);
468 ptlrpc_request_free(req);
472 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
475 lustre_set_wire_obdo(&imp->imp_connect_data, &body->oa, oa);
477 ptlrpc_request_set_replen(req);
479 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
480 BUILD_BUG_ON(sizeof(*sa) > sizeof(req->rq_async_args));
481 sa = ptlrpc_req_async_args(sa, req);
483 sa->sa_upcall = upcall;
484 sa->sa_cookie = cookie;
486 ptlrpcd_add_req(req);
491 static int osc_sync_interpret(const struct lu_env *env,
492 struct ptlrpc_request *req, void *args, int rc)
494 struct osc_fsync_args *fa = args;
495 struct ost_body *body;
496 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
497 unsigned long valid = 0;
498 struct cl_object *obj;
504 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
506 CERROR("can't unpack ost_body\n");
507 GOTO(out, rc = -EPROTO);
510 *fa->fa_oa = body->oa;
511 obj = osc2cl(fa->fa_obj);
513 /* Update osc object's blocks attribute */
514 cl_object_attr_lock(obj);
515 if (body->oa.o_valid & OBD_MD_FLBLOCKS) {
516 attr->cat_blocks = body->oa.o_blocks;
521 cl_object_attr_update(env, obj, attr, valid);
522 cl_object_attr_unlock(obj);
525 rc = fa->fa_upcall(fa->fa_cookie, rc);
529 int osc_sync_base(struct osc_object *obj, struct obdo *oa,
530 obd_enqueue_update_f upcall, void *cookie,
531 struct ptlrpc_request_set *rqset)
533 struct obd_export *exp = osc_export(obj);
534 struct ptlrpc_request *req;
535 struct ost_body *body;
536 struct osc_fsync_args *fa;
540 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
544 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
546 ptlrpc_request_free(req);
550 /* overload the size and blocks fields in the oa with start/end */
551 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
553 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
555 ptlrpc_request_set_replen(req);
556 req->rq_interpret_reply = osc_sync_interpret;
558 fa = ptlrpc_req_async_args(fa, req);
561 fa->fa_upcall = upcall;
562 fa->fa_cookie = cookie;
564 ptlrpc_set_add_req(rqset, req);
569 /* Find and cancel locally locks matched by @mode in the resource found by
570 * @objid. Found locks are added into @cancel list. Returns the amount of
571 * locks added to @cancels list. */
572 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
573 struct list_head *cancels,
574 enum ldlm_mode mode, __u64 lock_flags)
576 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
577 struct ldlm_res_id res_id;
578 struct ldlm_resource *res;
582 /* Return, i.e. cancel nothing, only if ELC is supported (flag in
583 * export) but disabled through procfs (flag in NS).
585 * This distinguishes from a case when ELC is not supported originally,
586 * when we still want to cancel locks in advance and just cancel them
587 * locally, without sending any RPC. */
588 if (exp_connect_cancelset(exp) && !ns_connect_cancelset(ns))
591 ostid_build_res_name(&oa->o_oi, &res_id);
592 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
596 LDLM_RESOURCE_ADDREF(res);
597 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
598 lock_flags, 0, NULL);
599 LDLM_RESOURCE_DELREF(res);
600 ldlm_resource_putref(res);
604 static int osc_destroy_interpret(const struct lu_env *env,
605 struct ptlrpc_request *req, void *args, int rc)
607 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
609 atomic_dec(&cli->cl_destroy_in_flight);
610 wake_up(&cli->cl_destroy_waitq);
615 static int osc_can_send_destroy(struct client_obd *cli)
617 if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
618 cli->cl_max_rpcs_in_flight) {
619 /* The destroy request can be sent */
622 if (atomic_dec_return(&cli->cl_destroy_in_flight) <
623 cli->cl_max_rpcs_in_flight) {
625 * The counter has been modified between the two atomic
628 wake_up(&cli->cl_destroy_waitq);
633 static int osc_destroy(const struct lu_env *env, struct obd_export *exp,
636 struct client_obd *cli = &exp->exp_obd->u.cli;
637 struct ptlrpc_request *req;
638 struct ost_body *body;
644 CDEBUG(D_INFO, "oa NULL\n");
648 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
649 LDLM_FL_DISCARD_DATA);
651 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
653 ldlm_lock_list_put(&cancels, l_bl_ast, count);
657 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
660 ptlrpc_request_free(req);
664 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
665 ptlrpc_at_set_req_timeout(req);
667 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
669 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
671 ptlrpc_request_set_replen(req);
673 req->rq_interpret_reply = osc_destroy_interpret;
674 if (!osc_can_send_destroy(cli)) {
676 * Wait until the number of on-going destroy RPCs drops
677 * under max_rpc_in_flight
679 rc = l_wait_event_abortable_exclusive(
680 cli->cl_destroy_waitq,
681 osc_can_send_destroy(cli));
683 ptlrpc_req_finished(req);
688 /* Do not wait for response */
689 ptlrpcd_add_req(req);
693 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
696 u64 bits = OBD_MD_FLBLOCKS | OBD_MD_FLGRANT;
698 LASSERT(!(oa->o_valid & bits));
701 spin_lock(&cli->cl_loi_list_lock);
702 if (OCD_HAS_FLAG(&cli->cl_import->imp_connect_data, GRANT_PARAM))
703 oa->o_dirty = cli->cl_dirty_grant;
705 oa->o_dirty = cli->cl_dirty_pages << PAGE_SHIFT;
706 if (unlikely(cli->cl_dirty_pages > cli->cl_dirty_max_pages)) {
707 CERROR("dirty %lu > dirty_max %lu\n",
709 cli->cl_dirty_max_pages);
711 } else if (unlikely(atomic_long_read(&obd_dirty_pages) >
712 (long)(obd_max_dirty_pages + 1))) {
713 /* The atomic_read() allowing the atomic_inc() are
714 * not covered by a lock thus they may safely race and trip
715 * this CERROR() unless we add in a small fudge factor (+1). */
716 CERROR("%s: dirty %ld > system dirty_max %ld\n",
717 cli_name(cli), atomic_long_read(&obd_dirty_pages),
718 obd_max_dirty_pages);
720 } else if (unlikely(cli->cl_dirty_max_pages - cli->cl_dirty_pages >
722 CERROR("dirty %lu - dirty_max %lu too big???\n",
723 cli->cl_dirty_pages, cli->cl_dirty_max_pages);
726 unsigned long nrpages;
727 unsigned long undirty;
729 nrpages = cli->cl_max_pages_per_rpc;
730 nrpages *= cli->cl_max_rpcs_in_flight + 1;
731 nrpages = max(nrpages, cli->cl_dirty_max_pages);
732 undirty = nrpages << PAGE_SHIFT;
733 if (OCD_HAS_FLAG(&cli->cl_import->imp_connect_data,
737 /* take extent tax into account when asking for more
739 nrextents = (nrpages + cli->cl_max_extent_pages - 1) /
740 cli->cl_max_extent_pages;
741 undirty += nrextents * cli->cl_grant_extent_tax;
743 /* Do not ask for more than OBD_MAX_GRANT - a margin for server
744 * to add extent tax, etc.
746 oa->o_undirty = min(undirty, OBD_MAX_GRANT &
747 ~(PTLRPC_MAX_BRW_SIZE * 4UL));
749 oa->o_grant = cli->cl_avail_grant + cli->cl_reserved_grant;
750 oa->o_dropped = cli->cl_lost_grant;
751 cli->cl_lost_grant = 0;
752 spin_unlock(&cli->cl_loi_list_lock);
753 CDEBUG(D_CACHE, "dirty: %llu undirty: %u dropped %u grant: %llu\n",
754 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
757 void osc_update_next_shrink(struct client_obd *cli)
759 cli->cl_next_shrink_grant = ktime_get_seconds() +
760 cli->cl_grant_shrink_interval;
762 CDEBUG(D_CACHE, "next time %lld to shrink grant\n",
763 cli->cl_next_shrink_grant);
766 static void __osc_update_grant(struct client_obd *cli, u64 grant)
768 spin_lock(&cli->cl_loi_list_lock);
769 cli->cl_avail_grant += grant;
770 spin_unlock(&cli->cl_loi_list_lock);
773 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
775 if (body->oa.o_valid & OBD_MD_FLGRANT) {
776 CDEBUG(D_CACHE, "got %llu extra grant\n", body->oa.o_grant);
777 __osc_update_grant(cli, body->oa.o_grant);
782 * grant thread data for shrinking space.
784 struct grant_thread_data {
785 struct list_head gtd_clients;
786 struct mutex gtd_mutex;
787 unsigned long gtd_stopped:1;
789 static struct grant_thread_data client_gtd;
791 static int osc_shrink_grant_interpret(const struct lu_env *env,
792 struct ptlrpc_request *req,
795 struct osc_grant_args *aa = args;
796 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
797 struct ost_body *body;
800 __osc_update_grant(cli, aa->aa_oa->o_grant);
804 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
806 osc_update_grant(cli, body);
808 OBD_SLAB_FREE_PTR(aa->aa_oa, osc_obdo_kmem);
814 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
816 spin_lock(&cli->cl_loi_list_lock);
817 oa->o_grant = cli->cl_avail_grant / 4;
818 cli->cl_avail_grant -= oa->o_grant;
819 spin_unlock(&cli->cl_loi_list_lock);
820 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
821 oa->o_valid |= OBD_MD_FLFLAGS;
824 oa->o_flags |= OBD_FL_SHRINK_GRANT;
825 osc_update_next_shrink(cli);
828 /* Shrink the current grant, either from some large amount to enough for a
829 * full set of in-flight RPCs, or if we have already shrunk to that limit
830 * then to enough for a single RPC. This avoids keeping more grant than
831 * needed, and avoids shrinking the grant piecemeal. */
832 static int osc_shrink_grant(struct client_obd *cli)
834 __u64 target_bytes = (cli->cl_max_rpcs_in_flight + 1) *
835 (cli->cl_max_pages_per_rpc << PAGE_SHIFT);
837 spin_lock(&cli->cl_loi_list_lock);
838 if (cli->cl_avail_grant <= target_bytes)
839 target_bytes = cli->cl_max_pages_per_rpc << PAGE_SHIFT;
840 spin_unlock(&cli->cl_loi_list_lock);
842 return osc_shrink_grant_to_target(cli, target_bytes);
845 int osc_shrink_grant_to_target(struct client_obd *cli, __u64 target_bytes)
848 struct ost_body *body;
851 spin_lock(&cli->cl_loi_list_lock);
852 /* Don't shrink if we are already above or below the desired limit
853 * We don't want to shrink below a single RPC, as that will negatively
854 * impact block allocation and long-term performance. */
855 if (target_bytes < cli->cl_max_pages_per_rpc << PAGE_SHIFT)
856 target_bytes = cli->cl_max_pages_per_rpc << PAGE_SHIFT;
858 if (target_bytes >= cli->cl_avail_grant) {
859 spin_unlock(&cli->cl_loi_list_lock);
862 spin_unlock(&cli->cl_loi_list_lock);
868 osc_announce_cached(cli, &body->oa, 0);
870 spin_lock(&cli->cl_loi_list_lock);
871 if (target_bytes >= cli->cl_avail_grant) {
872 /* available grant has changed since target calculation */
873 spin_unlock(&cli->cl_loi_list_lock);
874 GOTO(out_free, rc = 0);
876 body->oa.o_grant = cli->cl_avail_grant - target_bytes;
877 cli->cl_avail_grant = target_bytes;
878 spin_unlock(&cli->cl_loi_list_lock);
879 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
880 body->oa.o_valid |= OBD_MD_FLFLAGS;
881 body->oa.o_flags = 0;
883 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
884 osc_update_next_shrink(cli);
886 rc = osc_set_info_async(NULL, cli->cl_import->imp_obd->obd_self_export,
887 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
888 sizeof(*body), body, NULL);
890 __osc_update_grant(cli, body->oa.o_grant);
896 static int osc_should_shrink_grant(struct client_obd *client)
898 time64_t next_shrink = client->cl_next_shrink_grant;
900 if (client->cl_import == NULL)
903 if (!OCD_HAS_FLAG(&client->cl_import->imp_connect_data, GRANT_SHRINK) ||
904 client->cl_import->imp_grant_shrink_disabled) {
905 osc_update_next_shrink(client);
909 if (ktime_get_seconds() >= next_shrink - 5) {
910 /* Get the current RPC size directly, instead of going via:
911 * cli_brw_size(obd->u.cli.cl_import->imp_obd->obd_self_export)
912 * Keep comment here so that it can be found by searching. */
913 int brw_size = client->cl_max_pages_per_rpc << PAGE_SHIFT;
915 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
916 client->cl_avail_grant > brw_size)
919 osc_update_next_shrink(client);
924 #define GRANT_SHRINK_RPC_BATCH 100
926 static struct delayed_work work;
928 static void osc_grant_work_handler(struct work_struct *data)
930 struct client_obd *cli;
932 bool init_next_shrink = true;
933 time64_t next_shrink = ktime_get_seconds() + GRANT_SHRINK_INTERVAL;
936 mutex_lock(&client_gtd.gtd_mutex);
937 list_for_each_entry(cli, &client_gtd.gtd_clients,
939 if (rpc_sent < GRANT_SHRINK_RPC_BATCH &&
940 osc_should_shrink_grant(cli)) {
941 osc_shrink_grant(cli);
945 if (!init_next_shrink) {
946 if (cli->cl_next_shrink_grant < next_shrink &&
947 cli->cl_next_shrink_grant > ktime_get_seconds())
948 next_shrink = cli->cl_next_shrink_grant;
950 init_next_shrink = false;
951 next_shrink = cli->cl_next_shrink_grant;
954 mutex_unlock(&client_gtd.gtd_mutex);
956 if (client_gtd.gtd_stopped == 1)
959 if (next_shrink > ktime_get_seconds()) {
960 time64_t delay = next_shrink - ktime_get_seconds();
962 schedule_delayed_work(&work, cfs_time_seconds(delay));
964 schedule_work(&work.work);
968 void osc_schedule_grant_work(void)
970 cancel_delayed_work_sync(&work);
971 schedule_work(&work.work);
975 * Start grant thread for returing grant to server for idle clients.
977 static int osc_start_grant_work(void)
979 client_gtd.gtd_stopped = 0;
980 mutex_init(&client_gtd.gtd_mutex);
981 INIT_LIST_HEAD(&client_gtd.gtd_clients);
983 INIT_DELAYED_WORK(&work, osc_grant_work_handler);
984 schedule_work(&work.work);
989 static void osc_stop_grant_work(void)
991 client_gtd.gtd_stopped = 1;
992 cancel_delayed_work_sync(&work);
995 static void osc_add_grant_list(struct client_obd *client)
997 mutex_lock(&client_gtd.gtd_mutex);
998 list_add(&client->cl_grant_chain, &client_gtd.gtd_clients);
999 mutex_unlock(&client_gtd.gtd_mutex);
1002 static void osc_del_grant_list(struct client_obd *client)
1004 if (list_empty(&client->cl_grant_chain))
1007 mutex_lock(&client_gtd.gtd_mutex);
1008 list_del_init(&client->cl_grant_chain);
1009 mutex_unlock(&client_gtd.gtd_mutex);
1012 void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1015 * ocd_grant is the total grant amount we're expect to hold: if we've
1016 * been evicted, it's the new avail_grant amount, cl_dirty_pages will
1017 * drop to 0 as inflight RPCs fail out; otherwise, it's avail_grant +
1020 * race is tolerable here: if we're evicted, but imp_state already
1021 * left EVICTED state, then cl_dirty_pages must be 0 already.
1023 spin_lock(&cli->cl_loi_list_lock);
1024 cli->cl_avail_grant = ocd->ocd_grant;
1025 if (cli->cl_import->imp_state != LUSTRE_IMP_EVICTED) {
1026 unsigned long consumed = cli->cl_reserved_grant;
1028 if (OCD_HAS_FLAG(ocd, GRANT_PARAM))
1029 consumed += cli->cl_dirty_grant;
1031 consumed += cli->cl_dirty_pages << PAGE_SHIFT;
1032 if (cli->cl_avail_grant < consumed) {
1033 CERROR("%s: granted %ld but already consumed %ld\n",
1034 cli_name(cli), cli->cl_avail_grant, consumed);
1035 cli->cl_avail_grant = 0;
1037 cli->cl_avail_grant -= consumed;
1041 if (OCD_HAS_FLAG(ocd, GRANT_PARAM)) {
1045 /* overhead for each extent insertion */
1046 cli->cl_grant_extent_tax = ocd->ocd_grant_tax_kb << 10;
1047 /* determine the appropriate chunk size used by osc_extent. */
1048 cli->cl_chunkbits = max_t(int, PAGE_SHIFT,
1049 ocd->ocd_grant_blkbits);
1050 /* max_pages_per_rpc must be chunk aligned */
1051 chunk_mask = ~((1 << (cli->cl_chunkbits - PAGE_SHIFT)) - 1);
1052 cli->cl_max_pages_per_rpc = (cli->cl_max_pages_per_rpc +
1053 ~chunk_mask) & chunk_mask;
1054 /* determine maximum extent size, in #pages */
1055 size = (u64)ocd->ocd_grant_max_blks << ocd->ocd_grant_blkbits;
1056 cli->cl_max_extent_pages = size >> PAGE_SHIFT;
1057 if (cli->cl_max_extent_pages == 0)
1058 cli->cl_max_extent_pages = 1;
1060 cli->cl_grant_extent_tax = 0;
1061 cli->cl_chunkbits = PAGE_SHIFT;
1062 cli->cl_max_extent_pages = DT_MAX_BRW_PAGES;
1064 spin_unlock(&cli->cl_loi_list_lock);
1067 "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld. chunk bits: %d cl_max_extent_pages: %d\n",
1069 cli->cl_avail_grant, cli->cl_lost_grant, cli->cl_chunkbits,
1070 cli->cl_max_extent_pages);
1072 if (OCD_HAS_FLAG(ocd, GRANT_SHRINK) && list_empty(&cli->cl_grant_chain))
1073 osc_add_grant_list(cli);
1075 EXPORT_SYMBOL(osc_init_grant);
1077 /* We assume that the reason this OSC got a short read is because it read
1078 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1079 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1080 * this stripe never got written at or beyond this stripe offset yet. */
1081 static void handle_short_read(int nob_read, size_t page_count,
1082 struct brw_page **pga)
1087 /* skip bytes read OK */
1088 while (nob_read > 0) {
1089 LASSERT (page_count > 0);
1091 if (pga[i]->count > nob_read) {
1092 /* EOF inside this page */
1093 ptr = kmap(pga[i]->pg) +
1094 (pga[i]->off & ~PAGE_MASK);
1095 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1102 nob_read -= pga[i]->count;
1107 /* zero remaining pages */
1108 while (page_count-- > 0) {
1109 ptr = kmap(pga[i]->pg) + (pga[i]->off & ~PAGE_MASK);
1110 memset(ptr, 0, pga[i]->count);
1116 static int check_write_rcs(struct ptlrpc_request *req,
1117 int requested_nob, int niocount,
1118 size_t page_count, struct brw_page **pga)
1123 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1124 sizeof(*remote_rcs) *
1126 if (remote_rcs == NULL) {
1127 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1131 /* return error if any niobuf was in error */
1132 for (i = 0; i < niocount; i++) {
1133 if ((int)remote_rcs[i] < 0) {
1134 CDEBUG(D_INFO, "rc[%d]: %d req %p\n",
1135 i, remote_rcs[i], req);
1136 return remote_rcs[i];
1139 if (remote_rcs[i] != 0) {
1140 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1141 i, remote_rcs[i], req);
1145 if (req->rq_bulk != NULL &&
1146 req->rq_bulk->bd_nob_transferred != requested_nob) {
1147 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1148 req->rq_bulk->bd_nob_transferred, requested_nob);
1155 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1157 if (p1->flag != p2->flag) {
1158 unsigned mask = ~(OBD_BRW_FROM_GRANT | OBD_BRW_NOCACHE |
1159 OBD_BRW_SYNC | OBD_BRW_ASYNC |
1160 OBD_BRW_NOQUOTA | OBD_BRW_SOFT_SYNC);
1162 /* warn if we try to combine flags that we don't know to be
1163 * safe to combine */
1164 if (unlikely((p1->flag & mask) != (p2->flag & mask))) {
1165 CWARN("Saw flags 0x%x and 0x%x in the same brw, please "
1166 "report this at https://jira.whamcloud.com/\n",
1167 p1->flag, p2->flag);
1172 return (p1->off + p1->count == p2->off);
1175 #if IS_ENABLED(CONFIG_CRC_T10DIF)
1176 static int osc_checksum_bulk_t10pi(const char *obd_name, int nob,
1177 size_t pg_count, struct brw_page **pga,
1178 int opc, obd_dif_csum_fn *fn,
1182 struct ahash_request *req;
1183 /* Used Adler as the default checksum type on top of DIF tags */
1184 unsigned char cfs_alg = cksum_obd2cfs(OBD_CKSUM_T10_TOP);
1185 struct page *__page;
1186 unsigned char *buffer;
1188 unsigned int bufsize;
1190 int used_number = 0;
1196 LASSERT(pg_count > 0);
1198 __page = alloc_page(GFP_KERNEL);
1202 req = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1205 CERROR("%s: unable to initialize checksum hash %s: rc = %d\n",
1206 obd_name, cfs_crypto_hash_name(cfs_alg), rc);
1210 buffer = kmap(__page);
1211 guard_start = (__u16 *)buffer;
1212 guard_number = PAGE_SIZE / sizeof(*guard_start);
1213 while (nob > 0 && pg_count > 0) {
1214 unsigned int count = pga[i]->count > nob ? nob : pga[i]->count;
1216 /* corrupt the data before we compute the checksum, to
1217 * simulate an OST->client data error */
1218 if (unlikely(i == 0 && opc == OST_READ &&
1219 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))) {
1220 unsigned char *ptr = kmap(pga[i]->pg);
1221 int off = pga[i]->off & ~PAGE_MASK;
1223 memcpy(ptr + off, "bad1", min_t(typeof(nob), 4, nob));
1228 * The left guard number should be able to hold checksums of a
1231 rc = obd_page_dif_generate_buffer(obd_name, pga[i]->pg,
1232 pga[i]->off & ~PAGE_MASK,
1234 guard_start + used_number,
1235 guard_number - used_number,
1241 used_number += used;
1242 if (used_number == guard_number) {
1243 cfs_crypto_hash_update_page(req, __page, 0,
1244 used_number * sizeof(*guard_start));
1248 nob -= pga[i]->count;
1256 if (used_number != 0)
1257 cfs_crypto_hash_update_page(req, __page, 0,
1258 used_number * sizeof(*guard_start));
1260 bufsize = sizeof(cksum);
1261 cfs_crypto_hash_final(req, (unsigned char *)&cksum, &bufsize);
1263 /* For sending we only compute the wrong checksum instead
1264 * of corrupting the data so it is still correct on a redo */
1265 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1270 __free_page(__page);
1273 #else /* !CONFIG_CRC_T10DIF */
1274 #define obd_dif_ip_fn NULL
1275 #define obd_dif_crc_fn NULL
1276 #define osc_checksum_bulk_t10pi(name, nob, pgc, pga, opc, fn, ssize, csum) \
1278 #endif /* CONFIG_CRC_T10DIF */
1280 static int osc_checksum_bulk(int nob, size_t pg_count,
1281 struct brw_page **pga, int opc,
1282 enum cksum_types cksum_type,
1286 struct ahash_request *req;
1287 unsigned int bufsize;
1288 unsigned char cfs_alg = cksum_obd2cfs(cksum_type);
1290 LASSERT(pg_count > 0);
1292 req = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1294 CERROR("Unable to initialize checksum hash %s\n",
1295 cfs_crypto_hash_name(cfs_alg));
1296 return PTR_ERR(req);
1299 while (nob > 0 && pg_count > 0) {
1300 unsigned int count = pga[i]->count > nob ? nob : pga[i]->count;
1302 /* corrupt the data before we compute the checksum, to
1303 * simulate an OST->client data error */
1304 if (i == 0 && opc == OST_READ &&
1305 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE)) {
1306 unsigned char *ptr = kmap(pga[i]->pg);
1307 int off = pga[i]->off & ~PAGE_MASK;
1309 memcpy(ptr + off, "bad1", min_t(typeof(nob), 4, nob));
1312 cfs_crypto_hash_update_page(req, pga[i]->pg,
1313 pga[i]->off & ~PAGE_MASK,
1315 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d\n",
1316 (int)(pga[i]->off & ~PAGE_MASK));
1318 nob -= pga[i]->count;
1323 bufsize = sizeof(*cksum);
1324 cfs_crypto_hash_final(req, (unsigned char *)cksum, &bufsize);
1326 /* For sending we only compute the wrong checksum instead
1327 * of corrupting the data so it is still correct on a redo */
1328 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1334 static int osc_checksum_bulk_rw(const char *obd_name,
1335 enum cksum_types cksum_type,
1336 int nob, size_t pg_count,
1337 struct brw_page **pga, int opc,
1340 obd_dif_csum_fn *fn = NULL;
1341 int sector_size = 0;
1345 obd_t10_cksum2dif(cksum_type, &fn, §or_size);
1348 rc = osc_checksum_bulk_t10pi(obd_name, nob, pg_count, pga,
1349 opc, fn, sector_size, check_sum);
1351 rc = osc_checksum_bulk(nob, pg_count, pga, opc, cksum_type,
1357 static inline void osc_release_bounce_pages(struct brw_page **pga,
1360 #ifdef HAVE_LUSTRE_CRYPTO
1363 for (i = 0; i < page_count; i++) {
1364 if (!pga[i]->pg->mapping)
1365 /* bounce pages are unmapped */
1366 llcrypt_finalize_bounce_page(&pga[i]->pg);
1367 pga[i]->count -= pga[i]->bp_count_diff;
1368 pga[i]->off += pga[i]->bp_off_diff;
1374 osc_brw_prep_request(int cmd, struct client_obd *cli, struct obdo *oa,
1375 u32 page_count, struct brw_page **pga,
1376 struct ptlrpc_request **reqp, int resend)
1378 struct ptlrpc_request *req;
1379 struct ptlrpc_bulk_desc *desc;
1380 struct ost_body *body;
1381 struct obd_ioobj *ioobj;
1382 struct niobuf_remote *niobuf;
1383 int niocount, i, requested_nob, opc, rc, short_io_size = 0;
1384 struct osc_brw_async_args *aa;
1385 struct req_capsule *pill;
1386 struct brw_page *pg_prev;
1388 const char *obd_name = cli->cl_import->imp_obd->obd_name;
1389 struct inode *inode;
1392 inode = page2inode(pga[0]->pg);
1393 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1394 RETURN(-ENOMEM); /* Recoverable */
1395 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1396 RETURN(-EINVAL); /* Fatal */
1398 if ((cmd & OBD_BRW_WRITE) != 0) {
1400 req = ptlrpc_request_alloc_pool(cli->cl_import,
1402 &RQF_OST_BRW_WRITE);
1405 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1410 if (opc == OST_WRITE && inode && IS_ENCRYPTED(inode)) {
1411 for (i = 0; i < page_count; i++) {
1412 struct brw_page *pg = pga[i];
1413 struct page *data_page = NULL;
1414 bool retried = false;
1415 bool lockedbymyself;
1416 u32 nunits = (pg->off & ~PAGE_MASK) + pg->count;
1419 if (nunits & ~LUSTRE_ENCRYPTION_MASK)
1420 nunits = (nunits & LUSTRE_ENCRYPTION_MASK) +
1421 LUSTRE_ENCRYPTION_UNIT_SIZE;
1422 /* The page can already be locked when we arrive here.
1423 * This is possible when cl_page_assume/vvp_page_assume
1424 * is stuck on wait_on_page_writeback with page lock
1425 * held. In this case there is no risk for the lock to
1426 * be released while we are doing our encryption
1427 * processing, because writeback against that page will
1428 * end in vvp_page_completion_write/cl_page_completion,
1429 * which means only once the page is fully processed.
1431 lockedbymyself = trylock_page(pg->pg);
1433 llcrypt_encrypt_pagecache_blocks(pg->pg,
1437 unlock_page(pg->pg);
1438 if (IS_ERR(data_page)) {
1439 rc = PTR_ERR(data_page);
1440 if (rc == -ENOMEM && !retried) {
1445 ptlrpc_request_free(req);
1449 /* there should be no gap in the middle of page array */
1450 if (i == page_count - 1) {
1451 struct osc_async_page *oap = brw_page2oap(pg);
1453 oa->o_size = oap->oap_count +
1454 oap->oap_obj_off + oap->oap_page_off;
1456 /* len is forced to nunits, and relative offset to 0
1457 * so store the old, clear text info
1459 pg->bp_count_diff = nunits - pg->count;
1461 pg->bp_off_diff = pg->off & ~PAGE_MASK;
1462 pg->off = pg->off & PAGE_MASK;
1464 } else if (opc == OST_READ && inode && IS_ENCRYPTED(inode)) {
1465 for (i = 0; i < page_count; i++) {
1466 struct brw_page *pg = pga[i];
1467 u32 nunits = (pg->off & ~PAGE_MASK) + pg->count;
1469 if (nunits & ~LUSTRE_ENCRYPTION_MASK)
1470 nunits = (nunits & LUSTRE_ENCRYPTION_MASK) +
1471 LUSTRE_ENCRYPTION_UNIT_SIZE;
1472 /* count/off are forced to cover the whole encryption
1473 * unit size so that all encrypted data is stored on the
1474 * OST, so adjust bp_{count,off}_diff for the size of
1477 pg->bp_count_diff = nunits - pg->count;
1479 pg->bp_off_diff = pg->off & ~PAGE_MASK;
1480 pg->off = pg->off & PAGE_MASK;
1484 for (niocount = i = 1; i < page_count; i++) {
1485 if (!can_merge_pages(pga[i - 1], pga[i]))
1489 pill = &req->rq_pill;
1490 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1492 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1493 niocount * sizeof(*niobuf));
1495 for (i = 0; i < page_count; i++) {
1496 short_io_size += pga[i]->count;
1497 if (!inode || !IS_ENCRYPTED(inode)) {
1498 pga[i]->bp_count_diff = 0;
1499 pga[i]->bp_off_diff = 0;
1503 /* Check if read/write is small enough to be a short io. */
1504 if (short_io_size > cli->cl_max_short_io_bytes || niocount > 1 ||
1505 !imp_connect_shortio(cli->cl_import))
1508 req_capsule_set_size(pill, &RMF_SHORT_IO, RCL_CLIENT,
1509 opc == OST_READ ? 0 : short_io_size);
1510 if (opc == OST_READ)
1511 req_capsule_set_size(pill, &RMF_SHORT_IO, RCL_SERVER,
1514 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1516 ptlrpc_request_free(req);
1519 osc_set_io_portal(req);
1521 ptlrpc_at_set_req_timeout(req);
1522 /* ask ptlrpc not to resend on EINPROGRESS since BRWs have their own
1524 req->rq_no_retry_einprogress = 1;
1526 if (short_io_size != 0) {
1528 short_io_buf = NULL;
1532 desc = ptlrpc_prep_bulk_imp(req, page_count,
1533 cli->cl_import->imp_connect_data.ocd_brw_size >> LNET_MTU_BITS,
1534 (opc == OST_WRITE ? PTLRPC_BULK_GET_SOURCE :
1535 PTLRPC_BULK_PUT_SINK),
1537 &ptlrpc_bulk_kiov_pin_ops);
1540 GOTO(out, rc = -ENOMEM);
1541 /* NB request now owns desc and will free it when it gets freed */
1543 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1544 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1545 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1546 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1548 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
1550 /* For READ and WRITE, we can't fill o_uid and o_gid using from_kuid()
1551 * and from_kgid(), because they are asynchronous. Fortunately, variable
1552 * oa contains valid o_uid and o_gid in these two operations.
1553 * Besides, filling o_uid and o_gid is enough for nrs-tbf, see LU-9658.
1554 * OBD_MD_FLUID and OBD_MD_FLUID is not set in order to avoid breaking
1555 * other process logic */
1556 body->oa.o_uid = oa->o_uid;
1557 body->oa.o_gid = oa->o_gid;
1559 obdo_to_ioobj(oa, ioobj);
1560 ioobj->ioo_bufcnt = niocount;
1561 /* The high bits of ioo_max_brw tells server _maximum_ number of bulks
1562 * that might be send for this request. The actual number is decided
1563 * when the RPC is finally sent in ptlrpc_register_bulk(). It sends
1564 * "max - 1" for old client compatibility sending "0", and also so the
1565 * the actual maximum is a power-of-two number, not one less. LU-1431 */
1567 ioobj_max_brw_set(ioobj, desc->bd_md_max_brw);
1569 ioobj_max_brw_set(ioobj, 0);
1571 if (short_io_size != 0) {
1572 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1573 body->oa.o_valid |= OBD_MD_FLFLAGS;
1574 body->oa.o_flags = 0;
1576 body->oa.o_flags |= OBD_FL_SHORT_IO;
1577 CDEBUG(D_CACHE, "Using short io for data transfer, size = %d\n",
1579 if (opc == OST_WRITE) {
1580 short_io_buf = req_capsule_client_get(pill,
1582 LASSERT(short_io_buf != NULL);
1586 LASSERT(page_count > 0);
1588 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1589 struct brw_page *pg = pga[i];
1590 int poff = pg->off & ~PAGE_MASK;
1592 LASSERT(pg->count > 0);
1593 /* make sure there is no gap in the middle of page array */
1594 LASSERTF(page_count == 1 ||
1595 (ergo(i == 0, poff + pg->count == PAGE_SIZE) &&
1596 ergo(i > 0 && i < page_count - 1,
1597 poff == 0 && pg->count == PAGE_SIZE) &&
1598 ergo(i == page_count - 1, poff == 0)),
1599 "i: %d/%d pg: %p off: %llu, count: %u\n",
1600 i, page_count, pg, pg->off, pg->count);
1601 LASSERTF(i == 0 || pg->off > pg_prev->off,
1602 "i %d p_c %u pg %p [pri %lu ind %lu] off %llu"
1603 " prev_pg %p [pri %lu ind %lu] off %llu\n",
1605 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1606 pg_prev->pg, page_private(pg_prev->pg),
1607 pg_prev->pg->index, pg_prev->off);
1608 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1609 (pg->flag & OBD_BRW_SRVLOCK));
1610 if (short_io_size != 0 && opc == OST_WRITE) {
1611 unsigned char *ptr = kmap_atomic(pg->pg);
1613 LASSERT(short_io_size >= requested_nob + pg->count);
1614 memcpy(short_io_buf + requested_nob,
1618 } else if (short_io_size == 0) {
1619 desc->bd_frag_ops->add_kiov_frag(desc, pg->pg, poff,
1622 requested_nob += pg->count;
1624 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1626 niobuf->rnb_len += pg->count;
1628 niobuf->rnb_offset = pg->off;
1629 niobuf->rnb_len = pg->count;
1630 niobuf->rnb_flags = pg->flag;
1635 LASSERTF((void *)(niobuf - niocount) ==
1636 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1637 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1638 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1640 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1642 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1643 body->oa.o_valid |= OBD_MD_FLFLAGS;
1644 body->oa.o_flags = 0;
1646 body->oa.o_flags |= OBD_FL_RECOV_RESEND;
1649 if (osc_should_shrink_grant(cli))
1650 osc_shrink_grant_local(cli, &body->oa);
1652 /* size[REQ_REC_OFF] still sizeof (*body) */
1653 if (opc == OST_WRITE) {
1654 if (cli->cl_checksum &&
1655 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1656 /* store cl_cksum_type in a local variable since
1657 * it can be changed via lprocfs */
1658 enum cksum_types cksum_type = cli->cl_cksum_type;
1660 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1661 body->oa.o_flags = 0;
1663 body->oa.o_flags |= obd_cksum_type_pack(obd_name,
1665 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1667 rc = osc_checksum_bulk_rw(obd_name, cksum_type,
1668 requested_nob, page_count,
1672 CDEBUG(D_PAGE, "failed to checksum, rc = %d\n",
1676 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1679 /* save this in 'oa', too, for later checking */
1680 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1681 oa->o_flags |= obd_cksum_type_pack(obd_name,
1684 /* clear out the checksum flag, in case this is a
1685 * resend but cl_checksum is no longer set. b=11238 */
1686 oa->o_valid &= ~OBD_MD_FLCKSUM;
1688 oa->o_cksum = body->oa.o_cksum;
1689 /* 1 RC per niobuf */
1690 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1691 sizeof(__u32) * niocount);
1693 if (cli->cl_checksum &&
1694 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1695 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1696 body->oa.o_flags = 0;
1697 body->oa.o_flags |= obd_cksum_type_pack(obd_name,
1698 cli->cl_cksum_type);
1699 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1702 /* Client cksum has been already copied to wire obdo in previous
1703 * lustre_set_wire_obdo(), and in the case a bulk-read is being
1704 * resent due to cksum error, this will allow Server to
1705 * check+dump pages on its side */
1707 ptlrpc_request_set_replen(req);
1709 aa = ptlrpc_req_async_args(aa, req);
1711 aa->aa_requested_nob = requested_nob;
1712 aa->aa_nio_count = niocount;
1713 aa->aa_page_count = page_count;
1717 INIT_LIST_HEAD(&aa->aa_oaps);
1720 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1721 CDEBUG(D_RPCTRACE, "brw rpc %p - object "DOSTID" offset %lld<>%lld\n",
1722 req, POSTID(&oa->o_oi), niobuf[0].rnb_offset,
1723 niobuf[niocount - 1].rnb_offset + niobuf[niocount - 1].rnb_len);
1727 ptlrpc_req_finished(req);
1731 char dbgcksum_file_name[PATH_MAX];
1733 static void dump_all_bulk_pages(struct obdo *oa, __u32 page_count,
1734 struct brw_page **pga, __u32 server_cksum,
1742 /* will only keep dump of pages on first error for the same range in
1743 * file/fid, not during the resends/retries. */
1744 snprintf(dbgcksum_file_name, sizeof(dbgcksum_file_name),
1745 "%s-checksum_dump-osc-"DFID":[%llu-%llu]-%x-%x",
1746 (strncmp(libcfs_debug_file_path_arr, "NONE", 4) != 0 ?
1747 libcfs_debug_file_path_arr :
1748 LIBCFS_DEBUG_FILE_PATH_DEFAULT),
1749 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : 0ULL,
1750 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1751 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1753 pga[page_count-1]->off + pga[page_count-1]->count - 1,
1754 client_cksum, server_cksum);
1755 filp = filp_open(dbgcksum_file_name,
1756 O_CREAT | O_EXCL | O_WRONLY | O_LARGEFILE, 0600);
1760 CDEBUG(D_INFO, "%s: can't open to dump pages with "
1761 "checksum error: rc = %d\n", dbgcksum_file_name,
1764 CERROR("%s: can't open to dump pages with checksum "
1765 "error: rc = %d\n", dbgcksum_file_name, rc);
1769 for (i = 0; i < page_count; i++) {
1770 len = pga[i]->count;
1771 buf = kmap(pga[i]->pg);
1773 rc = cfs_kernel_write(filp, buf, len, &filp->f_pos);
1775 CERROR("%s: wanted to write %u but got %d "
1776 "error\n", dbgcksum_file_name, len, rc);
1781 CDEBUG(D_INFO, "%s: wrote %d bytes\n",
1782 dbgcksum_file_name, rc);
1787 rc = vfs_fsync_range(filp, 0, LLONG_MAX, 1);
1789 CERROR("%s: sync returns %d\n", dbgcksum_file_name, rc);
1790 filp_close(filp, NULL);
1794 check_write_checksum(struct obdo *oa, const struct lnet_process_id *peer,
1795 __u32 client_cksum, __u32 server_cksum,
1796 struct osc_brw_async_args *aa)
1798 const char *obd_name = aa->aa_cli->cl_import->imp_obd->obd_name;
1799 enum cksum_types cksum_type;
1800 obd_dif_csum_fn *fn = NULL;
1801 int sector_size = 0;
1806 if (server_cksum == client_cksum) {
1807 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1811 if (aa->aa_cli->cl_checksum_dump)
1812 dump_all_bulk_pages(oa, aa->aa_page_count, aa->aa_ppga,
1813 server_cksum, client_cksum);
1815 cksum_type = obd_cksum_type_unpack(oa->o_valid & OBD_MD_FLFLAGS ?
1818 switch (cksum_type) {
1819 case OBD_CKSUM_T10IP512:
1823 case OBD_CKSUM_T10IP4K:
1827 case OBD_CKSUM_T10CRC512:
1828 fn = obd_dif_crc_fn;
1831 case OBD_CKSUM_T10CRC4K:
1832 fn = obd_dif_crc_fn;
1840 rc = osc_checksum_bulk_t10pi(obd_name, aa->aa_requested_nob,
1841 aa->aa_page_count, aa->aa_ppga,
1842 OST_WRITE, fn, sector_size,
1845 rc = osc_checksum_bulk(aa->aa_requested_nob, aa->aa_page_count,
1846 aa->aa_ppga, OST_WRITE, cksum_type,
1850 msg = "failed to calculate the client write checksum";
1851 else if (cksum_type != obd_cksum_type_unpack(aa->aa_oa->o_flags))
1852 msg = "the server did not use the checksum type specified in "
1853 "the original request - likely a protocol problem";
1854 else if (new_cksum == server_cksum)
1855 msg = "changed on the client after we checksummed it - "
1856 "likely false positive due to mmap IO (bug 11742)";
1857 else if (new_cksum == client_cksum)
1858 msg = "changed in transit before arrival at OST";
1860 msg = "changed in transit AND doesn't match the original - "
1861 "likely false positive due to mmap IO (bug 11742)";
1863 LCONSOLE_ERROR_MSG(0x132, "%s: BAD WRITE CHECKSUM: %s: from %s inode "
1864 DFID " object "DOSTID" extent [%llu-%llu], original "
1865 "client csum %x (type %x), server csum %x (type %x),"
1866 " client csum now %x\n",
1867 obd_name, msg, libcfs_nid2str(peer->nid),
1868 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1869 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1870 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1871 POSTID(&oa->o_oi), aa->aa_ppga[0]->off,
1872 aa->aa_ppga[aa->aa_page_count - 1]->off +
1873 aa->aa_ppga[aa->aa_page_count-1]->count - 1,
1875 obd_cksum_type_unpack(aa->aa_oa->o_flags),
1876 server_cksum, cksum_type, new_cksum);
1880 /* Note rc enters this function as number of bytes transferred */
1881 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1883 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1884 struct client_obd *cli = aa->aa_cli;
1885 const char *obd_name = cli->cl_import->imp_obd->obd_name;
1886 const struct lnet_process_id *peer =
1887 &req->rq_import->imp_connection->c_peer;
1888 struct ost_body *body;
1889 u32 client_cksum = 0;
1890 struct inode *inode;
1894 if (rc < 0 && rc != -EDQUOT) {
1895 DEBUG_REQ(D_INFO, req, "Failed request: rc = %d", rc);
1899 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1900 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1902 DEBUG_REQ(D_INFO, req, "cannot unpack body");
1906 /* set/clear over quota flag for a uid/gid/projid */
1907 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1908 body->oa.o_valid & (OBD_MD_FLALLQUOTA)) {
1909 unsigned qid[LL_MAXQUOTAS] = {
1910 body->oa.o_uid, body->oa.o_gid,
1911 body->oa.o_projid };
1913 "setdq for [%u %u %u] with valid %#llx, flags %x\n",
1914 body->oa.o_uid, body->oa.o_gid, body->oa.o_projid,
1915 body->oa.o_valid, body->oa.o_flags);
1916 osc_quota_setdq(cli, req->rq_xid, qid, body->oa.o_valid,
1920 osc_update_grant(cli, body);
1925 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1926 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1928 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1930 CERROR("%s: unexpected positive size %d\n",
1935 if (req->rq_bulk != NULL &&
1936 sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1939 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1940 check_write_checksum(&body->oa, peer, client_cksum,
1941 body->oa.o_cksum, aa))
1944 rc = check_write_rcs(req, aa->aa_requested_nob,
1945 aa->aa_nio_count, aa->aa_page_count,
1950 /* The rest of this function executes only for OST_READs */
1952 if (req->rq_bulk == NULL) {
1953 rc = req_capsule_get_size(&req->rq_pill, &RMF_SHORT_IO,
1955 LASSERT(rc == req->rq_status);
1957 /* if unwrap_bulk failed, return -EAGAIN to retry */
1958 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1961 GOTO(out, rc = -EAGAIN);
1963 if (rc > aa->aa_requested_nob) {
1964 CERROR("%s: unexpected size %d, requested %d\n", obd_name,
1965 rc, aa->aa_requested_nob);
1969 if (req->rq_bulk != NULL && rc != req->rq_bulk->bd_nob_transferred) {
1970 CERROR("%s: unexpected size %d, transferred %d\n", obd_name,
1971 rc, req->rq_bulk->bd_nob_transferred);
1975 if (req->rq_bulk == NULL) {
1977 int nob, pg_count, i = 0;
1980 CDEBUG(D_CACHE, "Using short io read, size %d\n", rc);
1981 pg_count = aa->aa_page_count;
1982 buf = req_capsule_server_sized_get(&req->rq_pill, &RMF_SHORT_IO,
1985 while (nob > 0 && pg_count > 0) {
1987 int count = aa->aa_ppga[i]->count > nob ?
1988 nob : aa->aa_ppga[i]->count;
1990 CDEBUG(D_CACHE, "page %p count %d\n",
1991 aa->aa_ppga[i]->pg, count);
1992 ptr = kmap_atomic(aa->aa_ppga[i]->pg);
1993 memcpy(ptr + (aa->aa_ppga[i]->off & ~PAGE_MASK), buf,
1995 kunmap_atomic((void *) ptr);
2004 if (rc < aa->aa_requested_nob)
2005 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
2007 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
2008 static int cksum_counter;
2009 u32 server_cksum = body->oa.o_cksum;
2012 enum cksum_types cksum_type;
2013 u32 o_flags = body->oa.o_valid & OBD_MD_FLFLAGS ?
2014 body->oa.o_flags : 0;
2016 cksum_type = obd_cksum_type_unpack(o_flags);
2017 rc = osc_checksum_bulk_rw(obd_name, cksum_type, rc,
2018 aa->aa_page_count, aa->aa_ppga,
2019 OST_READ, &client_cksum);
2023 if (req->rq_bulk != NULL &&
2024 peer->nid != req->rq_bulk->bd_sender) {
2026 router = libcfs_nid2str(req->rq_bulk->bd_sender);
2029 if (server_cksum != client_cksum) {
2030 struct ost_body *clbody;
2031 u32 page_count = aa->aa_page_count;
2033 clbody = req_capsule_client_get(&req->rq_pill,
2035 if (cli->cl_checksum_dump)
2036 dump_all_bulk_pages(&clbody->oa, page_count,
2037 aa->aa_ppga, server_cksum,
2040 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
2041 "%s%s%s inode "DFID" object "DOSTID
2042 " extent [%llu-%llu], client %x, "
2043 "server %x, cksum_type %x\n",
2045 libcfs_nid2str(peer->nid),
2047 clbody->oa.o_valid & OBD_MD_FLFID ?
2048 clbody->oa.o_parent_seq : 0ULL,
2049 clbody->oa.o_valid & OBD_MD_FLFID ?
2050 clbody->oa.o_parent_oid : 0,
2051 clbody->oa.o_valid & OBD_MD_FLFID ?
2052 clbody->oa.o_parent_ver : 0,
2053 POSTID(&body->oa.o_oi),
2054 aa->aa_ppga[0]->off,
2055 aa->aa_ppga[page_count-1]->off +
2056 aa->aa_ppga[page_count-1]->count - 1,
2057 client_cksum, server_cksum,
2060 aa->aa_oa->o_cksum = client_cksum;
2064 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
2067 } else if (unlikely(client_cksum)) {
2068 static int cksum_missed;
2071 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
2072 CERROR("%s: checksum %u requested from %s but not sent\n",
2073 obd_name, cksum_missed,
2074 libcfs_nid2str(peer->nid));
2079 inode = page2inode(aa->aa_ppga[0]->pg);
2080 if (inode && IS_ENCRYPTED(inode)) {
2083 if (!llcrypt_has_encryption_key(inode)) {
2084 CDEBUG(D_SEC, "no enc key for ino %lu\n", inode->i_ino);
2087 for (idx = 0; idx < aa->aa_page_count; idx++) {
2088 struct brw_page *pg = aa->aa_ppga[idx];
2089 unsigned int offs = 0;
2091 while (offs < PAGE_SIZE) {
2092 /* do not decrypt if page is all 0s */
2093 if (memchr_inv(page_address(pg->pg) + offs, 0,
2094 LUSTRE_ENCRYPTION_UNIT_SIZE) == NULL) {
2095 /* if page is empty forward info to
2096 * upper layers (ll_io_zero_page) by
2097 * clearing PagePrivate2
2100 ClearPagePrivate2(pg->pg);
2104 /* The page is already locked when we arrive here,
2105 * except when we deal with a twisted page for
2106 * specific Direct IO support, in which case
2107 * PageChecked flag is set on page.
2109 if (PageChecked(pg->pg))
2111 rc = llcrypt_decrypt_pagecache_blocks(pg->pg,
2112 LUSTRE_ENCRYPTION_UNIT_SIZE,
2114 if (PageChecked(pg->pg))
2115 unlock_page(pg->pg);
2119 offs += LUSTRE_ENCRYPTION_UNIT_SIZE;
2126 lustre_get_wire_obdo(&req->rq_import->imp_connect_data,
2127 aa->aa_oa, &body->oa);
2132 static int osc_brw_redo_request(struct ptlrpc_request *request,
2133 struct osc_brw_async_args *aa, int rc)
2135 struct ptlrpc_request *new_req;
2136 struct osc_brw_async_args *new_aa;
2137 struct osc_async_page *oap;
2140 /* The below message is checked in replay-ost-single.sh test_8ae*/
2141 DEBUG_REQ(rc == -EINPROGRESS ? D_RPCTRACE : D_ERROR, request,
2142 "redo for recoverable error %d", rc);
2144 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
2145 OST_WRITE ? OBD_BRW_WRITE : OBD_BRW_READ,
2146 aa->aa_cli, aa->aa_oa, aa->aa_page_count,
2147 aa->aa_ppga, &new_req, 1);
2151 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2152 if (oap->oap_request != NULL) {
2153 LASSERTF(request == oap->oap_request,
2154 "request %p != oap_request %p\n",
2155 request, oap->oap_request);
2159 * New request takes over pga and oaps from old request.
2160 * Note that copying a list_head doesn't work, need to move it...
2163 new_req->rq_interpret_reply = request->rq_interpret_reply;
2164 new_req->rq_async_args = request->rq_async_args;
2165 new_req->rq_commit_cb = request->rq_commit_cb;
2166 /* cap resend delay to the current request timeout, this is similar to
2167 * what ptlrpc does (see after_reply()) */
2168 if (aa->aa_resends > new_req->rq_timeout)
2169 new_req->rq_sent = ktime_get_real_seconds() + new_req->rq_timeout;
2171 new_req->rq_sent = ktime_get_real_seconds() + aa->aa_resends;
2172 new_req->rq_generation_set = 1;
2173 new_req->rq_import_generation = request->rq_import_generation;
2175 new_aa = ptlrpc_req_async_args(new_aa, new_req);
2177 INIT_LIST_HEAD(&new_aa->aa_oaps);
2178 list_splice_init(&aa->aa_oaps, &new_aa->aa_oaps);
2179 INIT_LIST_HEAD(&new_aa->aa_exts);
2180 list_splice_init(&aa->aa_exts, &new_aa->aa_exts);
2181 new_aa->aa_resends = aa->aa_resends;
2183 list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
2184 if (oap->oap_request) {
2185 ptlrpc_req_finished(oap->oap_request);
2186 oap->oap_request = ptlrpc_request_addref(new_req);
2190 /* XXX: This code will run into problem if we're going to support
2191 * to add a series of BRW RPCs into a self-defined ptlrpc_request_set
2192 * and wait for all of them to be finished. We should inherit request
2193 * set from old request. */
2194 ptlrpcd_add_req(new_req);
2196 DEBUG_REQ(D_INFO, new_req, "new request");
2201 * ugh, we want disk allocation on the target to happen in offset order. we'll
2202 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
2203 * fine for our small page arrays and doesn't require allocation. its an
2204 * insertion sort that swaps elements that are strides apart, shrinking the
2205 * stride down until its '1' and the array is sorted.
2207 static void sort_brw_pages(struct brw_page **array, int num)
2210 struct brw_page *tmp;
2214 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
2219 for (i = stride ; i < num ; i++) {
2222 while (j >= stride && array[j - stride]->off > tmp->off) {
2223 array[j] = array[j - stride];
2228 } while (stride > 1);
2231 static void osc_release_ppga(struct brw_page **ppga, size_t count)
2233 LASSERT(ppga != NULL);
2234 OBD_FREE_PTR_ARRAY(ppga, count);
2237 static int brw_interpret(const struct lu_env *env,
2238 struct ptlrpc_request *req, void *args, int rc)
2240 struct osc_brw_async_args *aa = args;
2241 struct osc_extent *ext;
2242 struct osc_extent *tmp;
2243 struct client_obd *cli = aa->aa_cli;
2244 unsigned long transferred = 0;
2248 rc = osc_brw_fini_request(req, rc);
2249 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2251 /* restore clear text pages */
2252 osc_release_bounce_pages(aa->aa_ppga, aa->aa_page_count);
2255 * When server returns -EINPROGRESS, client should always retry
2256 * regardless of the number of times the bulk was resent already.
2258 if (osc_recoverable_error(rc) && !req->rq_no_delay) {
2259 if (req->rq_import_generation !=
2260 req->rq_import->imp_generation) {
2261 CDEBUG(D_HA, "%s: resend cross eviction for object: "
2262 ""DOSTID", rc = %d.\n",
2263 req->rq_import->imp_obd->obd_name,
2264 POSTID(&aa->aa_oa->o_oi), rc);
2265 } else if (rc == -EINPROGRESS ||
2266 client_should_resend(aa->aa_resends, aa->aa_cli)) {
2267 rc = osc_brw_redo_request(req, aa, rc);
2269 CERROR("%s: too many resent retries for object: "
2270 "%llu:%llu, rc = %d.\n",
2271 req->rq_import->imp_obd->obd_name,
2272 POSTID(&aa->aa_oa->o_oi), rc);
2277 else if (rc == -EAGAIN || rc == -EINPROGRESS)
2282 struct obdo *oa = aa->aa_oa;
2283 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
2284 unsigned long valid = 0;
2285 struct cl_object *obj;
2286 struct osc_async_page *last;
2288 last = brw_page2oap(aa->aa_ppga[aa->aa_page_count - 1]);
2289 obj = osc2cl(last->oap_obj);
2291 cl_object_attr_lock(obj);
2292 if (oa->o_valid & OBD_MD_FLBLOCKS) {
2293 attr->cat_blocks = oa->o_blocks;
2294 valid |= CAT_BLOCKS;
2296 if (oa->o_valid & OBD_MD_FLMTIME) {
2297 attr->cat_mtime = oa->o_mtime;
2300 if (oa->o_valid & OBD_MD_FLATIME) {
2301 attr->cat_atime = oa->o_atime;
2304 if (oa->o_valid & OBD_MD_FLCTIME) {
2305 attr->cat_ctime = oa->o_ctime;
2309 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
2310 struct lov_oinfo *loi = cl2osc(obj)->oo_oinfo;
2311 loff_t last_off = last->oap_count + last->oap_obj_off +
2314 /* Change file size if this is an out of quota or
2315 * direct IO write and it extends the file size */
2316 if (loi->loi_lvb.lvb_size < last_off) {
2317 attr->cat_size = last_off;
2320 /* Extend KMS if it's not a lockless write */
2321 if (loi->loi_kms < last_off &&
2322 oap2osc_page(last)->ops_srvlock == 0) {
2323 attr->cat_kms = last_off;
2329 cl_object_attr_update(env, obj, attr, valid);
2330 cl_object_attr_unlock(obj);
2332 OBD_SLAB_FREE_PTR(aa->aa_oa, osc_obdo_kmem);
2335 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE && rc == 0)
2336 osc_inc_unstable_pages(req);
2338 list_for_each_entry_safe(ext, tmp, &aa->aa_exts, oe_link) {
2339 list_del_init(&ext->oe_link);
2340 osc_extent_finish(env, ext, 1,
2341 rc && req->rq_no_delay ? -EWOULDBLOCK : rc);
2343 LASSERT(list_empty(&aa->aa_exts));
2344 LASSERT(list_empty(&aa->aa_oaps));
2346 transferred = (req->rq_bulk == NULL ? /* short io */
2347 aa->aa_requested_nob :
2348 req->rq_bulk->bd_nob_transferred);
2350 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2351 ptlrpc_lprocfs_brw(req, transferred);
2353 spin_lock(&cli->cl_loi_list_lock);
2354 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2355 * is called so we know whether to go to sync BRWs or wait for more
2356 * RPCs to complete */
2357 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2358 cli->cl_w_in_flight--;
2360 cli->cl_r_in_flight--;
2361 osc_wake_cache_waiters(cli);
2362 spin_unlock(&cli->cl_loi_list_lock);
2364 osc_io_unplug(env, cli, NULL);
2368 static void brw_commit(struct ptlrpc_request *req)
2370 /* If osc_inc_unstable_pages (via osc_extent_finish) races with
2371 * this called via the rq_commit_cb, I need to ensure
2372 * osc_dec_unstable_pages is still called. Otherwise unstable
2373 * pages may be leaked. */
2374 spin_lock(&req->rq_lock);
2375 if (likely(req->rq_unstable)) {
2376 req->rq_unstable = 0;
2377 spin_unlock(&req->rq_lock);
2379 osc_dec_unstable_pages(req);
2381 req->rq_committed = 1;
2382 spin_unlock(&req->rq_lock);
2387 * Build an RPC by the list of extent @ext_list. The caller must ensure
2388 * that the total pages in this list are NOT over max pages per RPC.
2389 * Extents in the list must be in OES_RPC state.
2391 int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
2392 struct list_head *ext_list, int cmd)
2394 struct ptlrpc_request *req = NULL;
2395 struct osc_extent *ext;
2396 struct brw_page **pga = NULL;
2397 struct osc_brw_async_args *aa = NULL;
2398 struct obdo *oa = NULL;
2399 struct osc_async_page *oap;
2400 struct osc_object *obj = NULL;
2401 struct cl_req_attr *crattr = NULL;
2402 loff_t starting_offset = OBD_OBJECT_EOF;
2403 loff_t ending_offset = 0;
2404 /* '1' for consistency with code that checks !mpflag to restore */
2408 bool soft_sync = false;
2409 bool ndelay = false;
2413 __u32 layout_version = 0;
2414 LIST_HEAD(rpc_list);
2415 struct ost_body *body;
2417 LASSERT(!list_empty(ext_list));
2419 /* add pages into rpc_list to build BRW rpc */
2420 list_for_each_entry(ext, ext_list, oe_link) {
2421 LASSERT(ext->oe_state == OES_RPC);
2422 mem_tight |= ext->oe_memalloc;
2423 grant += ext->oe_grants;
2424 page_count += ext->oe_nr_pages;
2425 layout_version = max(layout_version, ext->oe_layout_version);
2430 soft_sync = osc_over_unstable_soft_limit(cli);
2432 mpflag = memalloc_noreclaim_save();
2434 OBD_ALLOC_PTR_ARRAY(pga, page_count);
2436 GOTO(out, rc = -ENOMEM);
2438 OBD_SLAB_ALLOC_PTR_GFP(oa, osc_obdo_kmem, GFP_NOFS);
2440 GOTO(out, rc = -ENOMEM);
2443 list_for_each_entry(ext, ext_list, oe_link) {
2444 list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
2446 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
2448 oap->oap_brw_flags |= OBD_BRW_SOFT_SYNC;
2449 pga[i] = &oap->oap_brw_page;
2450 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2453 list_add_tail(&oap->oap_rpc_item, &rpc_list);
2454 if (starting_offset == OBD_OBJECT_EOF ||
2455 starting_offset > oap->oap_obj_off)
2456 starting_offset = oap->oap_obj_off;
2458 LASSERT(oap->oap_page_off == 0);
2459 if (ending_offset < oap->oap_obj_off + oap->oap_count)
2460 ending_offset = oap->oap_obj_off +
2463 LASSERT(oap->oap_page_off + oap->oap_count ==
2470 /* first page in the list */
2471 oap = list_entry(rpc_list.next, typeof(*oap), oap_rpc_item);
2473 crattr = &osc_env_info(env)->oti_req_attr;
2474 memset(crattr, 0, sizeof(*crattr));
2475 crattr->cra_type = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2476 crattr->cra_flags = ~0ULL;
2477 crattr->cra_page = oap2cl_page(oap);
2478 crattr->cra_oa = oa;
2479 cl_req_attr_set(env, osc2cl(obj), crattr);
2481 if (cmd == OBD_BRW_WRITE) {
2482 oa->o_grant_used = grant;
2483 if (layout_version > 0) {
2484 CDEBUG(D_LAYOUT, DFID": write with layout version %u\n",
2485 PFID(&oa->o_oi.oi_fid), layout_version);
2487 oa->o_layout_version = layout_version;
2488 oa->o_valid |= OBD_MD_LAYOUT_VERSION;
2492 sort_brw_pages(pga, page_count);
2493 rc = osc_brw_prep_request(cmd, cli, oa, page_count, pga, &req, 0);
2495 CERROR("prep_req failed: %d\n", rc);
2499 req->rq_commit_cb = brw_commit;
2500 req->rq_interpret_reply = brw_interpret;
2501 req->rq_memalloc = mem_tight != 0;
2502 oap->oap_request = ptlrpc_request_addref(req);
2504 req->rq_no_resend = req->rq_no_delay = 1;
2505 /* probably set a shorter timeout value.
2506 * to handle ETIMEDOUT in brw_interpret() correctly. */
2507 /* lustre_msg_set_timeout(req, req->rq_timeout / 2); */
2510 /* Need to update the timestamps after the request is built in case
2511 * we race with setattr (locally or in queue at OST). If OST gets
2512 * later setattr before earlier BRW (as determined by the request xid),
2513 * the OST will not use BRW timestamps. Sadly, there is no obvious
2514 * way to do this in a single call. bug 10150 */
2515 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2516 crattr->cra_oa = &body->oa;
2517 crattr->cra_flags = OBD_MD_FLMTIME | OBD_MD_FLCTIME | OBD_MD_FLATIME;
2518 cl_req_attr_set(env, osc2cl(obj), crattr);
2519 lustre_msg_set_jobid(req->rq_reqmsg, crattr->cra_jobid);
2521 aa = ptlrpc_req_async_args(aa, req);
2522 INIT_LIST_HEAD(&aa->aa_oaps);
2523 list_splice_init(&rpc_list, &aa->aa_oaps);
2524 INIT_LIST_HEAD(&aa->aa_exts);
2525 list_splice_init(ext_list, &aa->aa_exts);
2527 spin_lock(&cli->cl_loi_list_lock);
2528 starting_offset >>= PAGE_SHIFT;
2529 if (cmd == OBD_BRW_READ) {
2530 cli->cl_r_in_flight++;
2531 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2532 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2533 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2534 starting_offset + 1);
2536 cli->cl_w_in_flight++;
2537 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2538 lprocfs_oh_tally(&cli->cl_write_rpc_hist, cli->cl_w_in_flight);
2539 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2540 starting_offset + 1);
2542 spin_unlock(&cli->cl_loi_list_lock);
2544 DEBUG_REQ(D_INODE, req, "%d pages, aa %p, now %ur/%uw in flight",
2545 page_count, aa, cli->cl_r_in_flight,
2546 cli->cl_w_in_flight);
2547 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_DELAY_IO, cfs_fail_val);
2549 ptlrpcd_add_req(req);
2555 memalloc_noreclaim_restore(mpflag);
2558 LASSERT(req == NULL);
2561 OBD_SLAB_FREE_PTR(oa, osc_obdo_kmem);
2563 osc_release_bounce_pages(pga, page_count);
2564 osc_release_ppga(pga, page_count);
2566 /* this should happen rarely and is pretty bad, it makes the
2567 * pending list not follow the dirty order */
2568 while (!list_empty(ext_list)) {
2569 ext = list_entry(ext_list->next, struct osc_extent,
2571 list_del_init(&ext->oe_link);
2572 osc_extent_finish(env, ext, 0, rc);
2578 static int osc_set_lock_data(struct ldlm_lock *lock, void *data)
2582 LASSERT(lock != NULL);
2584 lock_res_and_lock(lock);
2586 if (lock->l_ast_data == NULL)
2587 lock->l_ast_data = data;
2588 if (lock->l_ast_data == data)
2591 unlock_res_and_lock(lock);
2596 int osc_enqueue_fini(struct ptlrpc_request *req, osc_enqueue_upcall_f upcall,
2597 void *cookie, struct lustre_handle *lockh,
2598 enum ldlm_mode mode, __u64 *flags, bool speculative,
2601 bool intent = *flags & LDLM_FL_HAS_INTENT;
2605 /* The request was created before ldlm_cli_enqueue call. */
2606 if (intent && errcode == ELDLM_LOCK_ABORTED) {
2607 struct ldlm_reply *rep;
2609 rep = req_capsule_server_get(&req->rq_pill, &RMF_DLM_REP);
2610 LASSERT(rep != NULL);
2612 rep->lock_policy_res1 =
2613 ptlrpc_status_ntoh(rep->lock_policy_res1);
2614 if (rep->lock_policy_res1)
2615 errcode = rep->lock_policy_res1;
2617 *flags |= LDLM_FL_LVB_READY;
2618 } else if (errcode == ELDLM_OK) {
2619 *flags |= LDLM_FL_LVB_READY;
2622 /* Call the update callback. */
2623 rc = (*upcall)(cookie, lockh, errcode);
2625 /* release the reference taken in ldlm_cli_enqueue() */
2626 if (errcode == ELDLM_LOCK_MATCHED)
2628 if (errcode == ELDLM_OK && lustre_handle_is_used(lockh))
2629 ldlm_lock_decref(lockh, mode);
2634 int osc_enqueue_interpret(const struct lu_env *env, struct ptlrpc_request *req,
2637 struct osc_enqueue_args *aa = args;
2638 struct ldlm_lock *lock;
2639 struct lustre_handle *lockh = &aa->oa_lockh;
2640 enum ldlm_mode mode = aa->oa_mode;
2641 struct ost_lvb *lvb = aa->oa_lvb;
2642 __u32 lvb_len = sizeof(*lvb);
2647 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2649 lock = ldlm_handle2lock(lockh);
2650 LASSERTF(lock != NULL,
2651 "lockh %#llx, req %p, aa %p - client evicted?\n",
2652 lockh->cookie, req, aa);
2654 /* Take an additional reference so that a blocking AST that
2655 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2656 * to arrive after an upcall has been executed by
2657 * osc_enqueue_fini(). */
2658 ldlm_lock_addref(lockh, mode);
2660 /* Let cl_lock_state_wait fail with -ERESTARTSYS to unuse sublocks. */
2661 OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_ENQUEUE_HANG, 2);
2663 /* Let CP AST to grant the lock first. */
2664 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
2666 if (aa->oa_speculative) {
2667 LASSERT(aa->oa_lvb == NULL);
2668 LASSERT(aa->oa_flags == NULL);
2669 aa->oa_flags = &flags;
2672 /* Complete obtaining the lock procedure. */
2673 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_type, 1,
2674 aa->oa_mode, aa->oa_flags, lvb, lvb_len,
2676 /* Complete osc stuff. */
2677 rc = osc_enqueue_fini(req, aa->oa_upcall, aa->oa_cookie, lockh, mode,
2678 aa->oa_flags, aa->oa_speculative, rc);
2680 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
2682 ldlm_lock_decref(lockh, mode);
2683 LDLM_LOCK_PUT(lock);
2687 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
2688 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
2689 * other synchronous requests, however keeping some locks and trying to obtain
2690 * others may take a considerable amount of time in a case of ost failure; and
2691 * when other sync requests do not get released lock from a client, the client
2692 * is evicted from the cluster -- such scenarious make the life difficult, so
2693 * release locks just after they are obtained. */
2694 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2695 __u64 *flags, union ldlm_policy_data *policy,
2696 struct ost_lvb *lvb, osc_enqueue_upcall_f upcall,
2697 void *cookie, struct ldlm_enqueue_info *einfo,
2698 struct ptlrpc_request_set *rqset, int async,
2701 struct obd_device *obd = exp->exp_obd;
2702 struct lustre_handle lockh = { 0 };
2703 struct ptlrpc_request *req = NULL;
2704 int intent = *flags & LDLM_FL_HAS_INTENT;
2705 __u64 match_flags = *flags;
2706 enum ldlm_mode mode;
2710 /* Filesystem lock extents are extended to page boundaries so that
2711 * dealing with the page cache is a little smoother. */
2712 policy->l_extent.start -= policy->l_extent.start & ~PAGE_MASK;
2713 policy->l_extent.end |= ~PAGE_MASK;
2715 /* Next, search for already existing extent locks that will cover us */
2716 /* If we're trying to read, we also search for an existing PW lock. The
2717 * VFS and page cache already protect us locally, so lots of readers/
2718 * writers can share a single PW lock.
2720 * There are problems with conversion deadlocks, so instead of
2721 * converting a read lock to a write lock, we'll just enqueue a new
2724 * At some point we should cancel the read lock instead of making them
2725 * send us a blocking callback, but there are problems with canceling
2726 * locks out from other users right now, too. */
2727 mode = einfo->ei_mode;
2728 if (einfo->ei_mode == LCK_PR)
2730 /* Normal lock requests must wait for the LVB to be ready before
2731 * matching a lock; speculative lock requests do not need to,
2732 * because they will not actually use the lock. */
2734 match_flags |= LDLM_FL_LVB_READY;
2736 match_flags |= LDLM_FL_BLOCK_GRANTED;
2737 mode = ldlm_lock_match(obd->obd_namespace, match_flags, res_id,
2738 einfo->ei_type, policy, mode, &lockh);
2740 struct ldlm_lock *matched;
2742 if (*flags & LDLM_FL_TEST_LOCK)
2745 matched = ldlm_handle2lock(&lockh);
2747 /* This DLM lock request is speculative, and does not
2748 * have an associated IO request. Therefore if there
2749 * is already a DLM lock, it wll just inform the
2750 * caller to cancel the request for this stripe.*/
2751 lock_res_and_lock(matched);
2752 if (ldlm_extent_equal(&policy->l_extent,
2753 &matched->l_policy_data.l_extent))
2757 unlock_res_and_lock(matched);
2759 ldlm_lock_decref(&lockh, mode);
2760 LDLM_LOCK_PUT(matched);
2762 } else if (osc_set_lock_data(matched, einfo->ei_cbdata)) {
2763 *flags |= LDLM_FL_LVB_READY;
2765 /* We already have a lock, and it's referenced. */
2766 (*upcall)(cookie, &lockh, ELDLM_LOCK_MATCHED);
2768 ldlm_lock_decref(&lockh, mode);
2769 LDLM_LOCK_PUT(matched);
2772 ldlm_lock_decref(&lockh, mode);
2773 LDLM_LOCK_PUT(matched);
2777 if (*flags & (LDLM_FL_TEST_LOCK | LDLM_FL_MATCH_LOCK))
2781 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2782 &RQF_LDLM_ENQUEUE_LVB);
2786 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
2788 ptlrpc_request_free(req);
2792 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
2794 ptlrpc_request_set_replen(req);
2797 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
2798 *flags &= ~LDLM_FL_BLOCK_GRANTED;
2800 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
2801 sizeof(*lvb), LVB_T_OST, &lockh, async);
2804 struct osc_enqueue_args *aa;
2805 aa = ptlrpc_req_async_args(aa, req);
2807 aa->oa_mode = einfo->ei_mode;
2808 aa->oa_type = einfo->ei_type;
2809 lustre_handle_copy(&aa->oa_lockh, &lockh);
2810 aa->oa_upcall = upcall;
2811 aa->oa_cookie = cookie;
2812 aa->oa_speculative = speculative;
2814 aa->oa_flags = flags;
2817 /* speculative locks are essentially to enqueue
2818 * a DLM lock in advance, so we don't care
2819 * about the result of the enqueue. */
2821 aa->oa_flags = NULL;
2824 req->rq_interpret_reply = osc_enqueue_interpret;
2825 ptlrpc_set_add_req(rqset, req);
2826 } else if (intent) {
2827 ptlrpc_req_finished(req);
2832 rc = osc_enqueue_fini(req, upcall, cookie, &lockh, einfo->ei_mode,
2833 flags, speculative, rc);
2835 ptlrpc_req_finished(req);
2840 int osc_match_base(const struct lu_env *env, struct obd_export *exp,
2841 struct ldlm_res_id *res_id, enum ldlm_type type,
2842 union ldlm_policy_data *policy, enum ldlm_mode mode,
2843 __u64 *flags, struct osc_object *obj,
2844 struct lustre_handle *lockh, enum ldlm_match_flags match_flags)
2846 struct obd_device *obd = exp->exp_obd;
2847 __u64 lflags = *flags;
2851 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
2854 /* Filesystem lock extents are extended to page boundaries so that
2855 * dealing with the page cache is a little smoother */
2856 policy->l_extent.start -= policy->l_extent.start & ~PAGE_MASK;
2857 policy->l_extent.end |= ~PAGE_MASK;
2859 /* Next, search for already existing extent locks that will cover us */
2860 /* If we're trying to read, we also search for an existing PW lock. The
2861 * VFS and page cache already protect us locally, so lots of readers/
2862 * writers can share a single PW lock. */
2867 rc = ldlm_lock_match_with_skip(obd->obd_namespace, lflags, 0,
2868 res_id, type, policy, rc, lockh,
2870 if (rc == 0 || lflags & LDLM_FL_TEST_LOCK)
2874 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2876 LASSERT(lock != NULL);
2877 if (osc_set_lock_data(lock, obj)) {
2878 lock_res_and_lock(lock);
2879 if (!ldlm_is_lvb_cached(lock)) {
2880 LASSERT(lock->l_ast_data == obj);
2881 osc_lock_lvb_update(env, obj, lock, NULL);
2882 ldlm_set_lvb_cached(lock);
2884 unlock_res_and_lock(lock);
2886 ldlm_lock_decref(lockh, rc);
2889 LDLM_LOCK_PUT(lock);
2894 static int osc_statfs_interpret(const struct lu_env *env,
2895 struct ptlrpc_request *req, void *args, int rc)
2897 struct osc_async_args *aa = args;
2898 struct obd_statfs *msfs;
2903 * The request has in fact never been sent due to issues at
2904 * a higher level (LOV). Exit immediately since the caller
2905 * is aware of the problem and takes care of the clean up.
2909 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
2910 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
2916 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2918 GOTO(out, rc = -EPROTO);
2920 *aa->aa_oi->oi_osfs = *msfs;
2922 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
2927 static int osc_statfs_async(struct obd_export *exp,
2928 struct obd_info *oinfo, time64_t max_age,
2929 struct ptlrpc_request_set *rqset)
2931 struct obd_device *obd = class_exp2obd(exp);
2932 struct ptlrpc_request *req;
2933 struct osc_async_args *aa;
2937 if (obd->obd_osfs_age >= max_age) {
2939 "%s: use %p cache blocks %llu/%llu objects %llu/%llu\n",
2940 obd->obd_name, &obd->obd_osfs,
2941 obd->obd_osfs.os_bavail, obd->obd_osfs.os_blocks,
2942 obd->obd_osfs.os_ffree, obd->obd_osfs.os_files);
2943 spin_lock(&obd->obd_osfs_lock);
2944 memcpy(oinfo->oi_osfs, &obd->obd_osfs, sizeof(*oinfo->oi_osfs));
2945 spin_unlock(&obd->obd_osfs_lock);
2946 oinfo->oi_flags |= OBD_STATFS_FROM_CACHE;
2947 if (oinfo->oi_cb_up)
2948 oinfo->oi_cb_up(oinfo, 0);
2953 /* We could possibly pass max_age in the request (as an absolute
2954 * timestamp or a "seconds.usec ago") so the target can avoid doing
2955 * extra calls into the filesystem if that isn't necessary (e.g.
2956 * during mount that would help a bit). Having relative timestamps
2957 * is not so great if request processing is slow, while absolute
2958 * timestamps are not ideal because they need time synchronization. */
2959 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
2963 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2965 ptlrpc_request_free(req);
2968 ptlrpc_request_set_replen(req);
2969 req->rq_request_portal = OST_CREATE_PORTAL;
2970 ptlrpc_at_set_req_timeout(req);
2972 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
2973 /* procfs requests not want stat in wait for avoid deadlock */
2974 req->rq_no_resend = 1;
2975 req->rq_no_delay = 1;
2978 req->rq_interpret_reply = osc_statfs_interpret;
2979 aa = ptlrpc_req_async_args(aa, req);
2982 ptlrpc_set_add_req(rqset, req);
2986 static int osc_statfs(const struct lu_env *env, struct obd_export *exp,
2987 struct obd_statfs *osfs, time64_t max_age, __u32 flags)
2989 struct obd_device *obd = class_exp2obd(exp);
2990 struct obd_statfs *msfs;
2991 struct ptlrpc_request *req;
2992 struct obd_import *imp = NULL;
2997 /*Since the request might also come from lprocfs, so we need
2998 *sync this with client_disconnect_export Bug15684*/
2999 down_read(&obd->u.cli.cl_sem);
3000 if (obd->u.cli.cl_import)
3001 imp = class_import_get(obd->u.cli.cl_import);
3002 up_read(&obd->u.cli.cl_sem);
3006 /* We could possibly pass max_age in the request (as an absolute
3007 * timestamp or a "seconds.usec ago") so the target can avoid doing
3008 * extra calls into the filesystem if that isn't necessary (e.g.
3009 * during mount that would help a bit). Having relative timestamps
3010 * is not so great if request processing is slow, while absolute
3011 * timestamps are not ideal because they need time synchronization. */
3012 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
3014 class_import_put(imp);
3019 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3021 ptlrpc_request_free(req);
3024 ptlrpc_request_set_replen(req);
3025 req->rq_request_portal = OST_CREATE_PORTAL;
3026 ptlrpc_at_set_req_timeout(req);
3028 if (flags & OBD_STATFS_NODELAY) {
3029 /* procfs requests not want stat in wait for avoid deadlock */
3030 req->rq_no_resend = 1;
3031 req->rq_no_delay = 1;
3034 rc = ptlrpc_queue_wait(req);
3038 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3040 GOTO(out, rc = -EPROTO);
3046 ptlrpc_req_finished(req);
3050 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
3051 void *karg, void __user *uarg)
3053 struct obd_device *obd = exp->exp_obd;
3054 struct obd_ioctl_data *data = karg;
3058 if (!try_module_get(THIS_MODULE)) {
3059 CERROR("%s: cannot get module '%s'\n", obd->obd_name,
3060 module_name(THIS_MODULE));
3064 case OBD_IOC_CLIENT_RECOVER:
3065 rc = ptlrpc_recover_import(obd->u.cli.cl_import,
3066 data->ioc_inlbuf1, 0);
3070 case IOC_OSC_SET_ACTIVE:
3071 rc = ptlrpc_set_import_active(obd->u.cli.cl_import,
3076 CDEBUG(D_INODE, "%s: unrecognised ioctl %#x by %s: rc = %d\n",
3077 obd->obd_name, cmd, current->comm, rc);
3081 module_put(THIS_MODULE);
3085 int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
3086 u32 keylen, void *key, u32 vallen, void *val,
3087 struct ptlrpc_request_set *set)
3089 struct ptlrpc_request *req;
3090 struct obd_device *obd = exp->exp_obd;
3091 struct obd_import *imp = class_exp2cliimp(exp);
3096 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
3098 if (KEY_IS(KEY_CHECKSUM)) {
3099 if (vallen != sizeof(int))
3101 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
3105 if (KEY_IS(KEY_SPTLRPC_CONF)) {
3106 sptlrpc_conf_client_adapt(obd);
3110 if (KEY_IS(KEY_FLUSH_CTX)) {
3111 sptlrpc_import_flush_my_ctx(imp);
3115 if (KEY_IS(KEY_CACHE_LRU_SHRINK)) {
3116 struct client_obd *cli = &obd->u.cli;
3117 long nr = atomic_long_read(&cli->cl_lru_in_list) >> 1;
3118 long target = *(long *)val;
3120 nr = osc_lru_shrink(env, cli, min(nr, target), true);
3125 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
3128 /* We pass all other commands directly to OST. Since nobody calls osc
3129 methods directly and everybody is supposed to go through LOV, we
3130 assume lov checked invalid values for us.
3131 The only recognised values so far are evict_by_nid and mds_conn.
3132 Even if something bad goes through, we'd get a -EINVAL from OST
3135 req = ptlrpc_request_alloc(imp, KEY_IS(KEY_GRANT_SHRINK) ?
3136 &RQF_OST_SET_GRANT_INFO :
3141 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3142 RCL_CLIENT, keylen);
3143 if (!KEY_IS(KEY_GRANT_SHRINK))
3144 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
3145 RCL_CLIENT, vallen);
3146 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
3148 ptlrpc_request_free(req);
3152 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3153 memcpy(tmp, key, keylen);
3154 tmp = req_capsule_client_get(&req->rq_pill, KEY_IS(KEY_GRANT_SHRINK) ?
3157 memcpy(tmp, val, vallen);
3159 if (KEY_IS(KEY_GRANT_SHRINK)) {
3160 struct osc_grant_args *aa;
3163 aa = ptlrpc_req_async_args(aa, req);
3164 OBD_SLAB_ALLOC_PTR_GFP(oa, osc_obdo_kmem, GFP_NOFS);
3166 ptlrpc_req_finished(req);
3169 *oa = ((struct ost_body *)val)->oa;
3171 req->rq_interpret_reply = osc_shrink_grant_interpret;
3174 ptlrpc_request_set_replen(req);
3175 if (!KEY_IS(KEY_GRANT_SHRINK)) {
3176 LASSERT(set != NULL);
3177 ptlrpc_set_add_req(set, req);
3178 ptlrpc_check_set(NULL, set);
3180 ptlrpcd_add_req(req);
3185 EXPORT_SYMBOL(osc_set_info_async);
3187 int osc_reconnect(const struct lu_env *env, struct obd_export *exp,
3188 struct obd_device *obd, struct obd_uuid *cluuid,
3189 struct obd_connect_data *data, void *localdata)
3191 struct client_obd *cli = &obd->u.cli;
3193 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
3197 spin_lock(&cli->cl_loi_list_lock);
3198 grant = cli->cl_avail_grant + cli->cl_reserved_grant;
3199 if (data->ocd_connect_flags & OBD_CONNECT_GRANT_PARAM) {
3200 /* restore ocd_grant_blkbits as client page bits */
3201 data->ocd_grant_blkbits = PAGE_SHIFT;
3202 grant += cli->cl_dirty_grant;
3204 grant += cli->cl_dirty_pages << PAGE_SHIFT;
3206 data->ocd_grant = grant ? : 2 * cli_brw_size(obd);
3207 lost_grant = cli->cl_lost_grant;
3208 cli->cl_lost_grant = 0;
3209 spin_unlock(&cli->cl_loi_list_lock);
3211 CDEBUG(D_RPCTRACE, "ocd_connect_flags: %#llx ocd_version: %d"
3212 " ocd_grant: %d, lost: %ld.\n", data->ocd_connect_flags,
3213 data->ocd_version, data->ocd_grant, lost_grant);
3218 EXPORT_SYMBOL(osc_reconnect);
3220 int osc_disconnect(struct obd_export *exp)
3222 struct obd_device *obd = class_exp2obd(exp);
3225 rc = client_disconnect_export(exp);
3227 * Initially we put del_shrink_grant before disconnect_export, but it
3228 * causes the following problem if setup (connect) and cleanup
3229 * (disconnect) are tangled together.
3230 * connect p1 disconnect p2
3231 * ptlrpc_connect_import
3232 * ............... class_manual_cleanup
3235 * ptlrpc_connect_interrupt
3237 * add this client to shrink list
3239 * Bang! grant shrink thread trigger the shrink. BUG18662
3241 osc_del_grant_list(&obd->u.cli);
3244 EXPORT_SYMBOL(osc_disconnect);
3246 int osc_ldlm_resource_invalidate(struct cfs_hash *hs, struct cfs_hash_bd *bd,
3247 struct hlist_node *hnode, void *arg)
3249 struct lu_env *env = arg;
3250 struct ldlm_resource *res = cfs_hash_object(hs, hnode);
3251 struct ldlm_lock *lock;
3252 struct osc_object *osc = NULL;
3256 list_for_each_entry(lock, &res->lr_granted, l_res_link) {
3257 if (lock->l_ast_data != NULL && osc == NULL) {
3258 osc = lock->l_ast_data;
3259 cl_object_get(osc2cl(osc));
3262 /* clear LDLM_FL_CLEANED flag to make sure it will be canceled
3263 * by the 2nd round of ldlm_namespace_clean() call in
3264 * osc_import_event(). */
3265 ldlm_clear_cleaned(lock);
3270 osc_object_invalidate(env, osc);
3271 cl_object_put(env, osc2cl(osc));
3276 EXPORT_SYMBOL(osc_ldlm_resource_invalidate);
3278 static int osc_import_event(struct obd_device *obd,
3279 struct obd_import *imp,
3280 enum obd_import_event event)
3282 struct client_obd *cli;
3286 LASSERT(imp->imp_obd == obd);
3289 case IMP_EVENT_DISCON: {
3291 spin_lock(&cli->cl_loi_list_lock);
3292 cli->cl_avail_grant = 0;
3293 cli->cl_lost_grant = 0;
3294 spin_unlock(&cli->cl_loi_list_lock);
3297 case IMP_EVENT_INACTIVE: {
3298 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE);
3301 case IMP_EVENT_INVALIDATE: {
3302 struct ldlm_namespace *ns = obd->obd_namespace;
3306 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3308 env = cl_env_get(&refcheck);
3310 osc_io_unplug(env, &obd->u.cli, NULL);
3312 cfs_hash_for_each_nolock(ns->ns_rs_hash,
3313 osc_ldlm_resource_invalidate,
3315 cl_env_put(env, &refcheck);
3317 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3322 case IMP_EVENT_ACTIVE: {
3323 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE);
3326 case IMP_EVENT_OCD: {
3327 struct obd_connect_data *ocd = &imp->imp_connect_data;
3329 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
3330 osc_init_grant(&obd->u.cli, ocd);
3333 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
3334 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
3336 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD);
3339 case IMP_EVENT_DEACTIVATE: {
3340 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DEACTIVATE);
3343 case IMP_EVENT_ACTIVATE: {
3344 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVATE);
3348 CERROR("Unknown import event %d\n", event);
3355 * Determine whether the lock can be canceled before replaying the lock
3356 * during recovery, see bug16774 for detailed information.
3358 * \retval zero the lock can't be canceled
3359 * \retval other ok to cancel
3361 static int osc_cancel_weight(struct ldlm_lock *lock)
3364 * Cancel all unused and granted extent lock.
3366 if (lock->l_resource->lr_type == LDLM_EXTENT &&
3367 ldlm_is_granted(lock) &&
3368 osc_ldlm_weigh_ast(lock) == 0)
3374 static int brw_queue_work(const struct lu_env *env, void *data)
3376 struct client_obd *cli = data;
3378 CDEBUG(D_CACHE, "Run writeback work for client obd %p.\n", cli);
3380 osc_io_unplug(env, cli, NULL);
3384 int osc_setup_common(struct obd_device *obd, struct lustre_cfg *lcfg)
3386 struct client_obd *cli = &obd->u.cli;
3392 rc = ptlrpcd_addref();
3396 rc = client_obd_setup(obd, lcfg);
3398 GOTO(out_ptlrpcd, rc);
3401 handler = ptlrpcd_alloc_work(cli->cl_import, brw_queue_work, cli);
3402 if (IS_ERR(handler))
3403 GOTO(out_ptlrpcd_work, rc = PTR_ERR(handler));
3404 cli->cl_writeback_work = handler;
3406 handler = ptlrpcd_alloc_work(cli->cl_import, lru_queue_work, cli);
3407 if (IS_ERR(handler))
3408 GOTO(out_ptlrpcd_work, rc = PTR_ERR(handler));
3409 cli->cl_lru_work = handler;
3411 rc = osc_quota_setup(obd);
3413 GOTO(out_ptlrpcd_work, rc);
3415 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
3416 osc_update_next_shrink(cli);
3421 if (cli->cl_writeback_work != NULL) {
3422 ptlrpcd_destroy_work(cli->cl_writeback_work);
3423 cli->cl_writeback_work = NULL;
3425 if (cli->cl_lru_work != NULL) {
3426 ptlrpcd_destroy_work(cli->cl_lru_work);
3427 cli->cl_lru_work = NULL;
3429 client_obd_cleanup(obd);
3434 EXPORT_SYMBOL(osc_setup_common);
3436 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
3438 struct client_obd *cli = &obd->u.cli;
3446 rc = osc_setup_common(obd, lcfg);
3450 rc = osc_tunables_init(obd);
3455 * We try to control the total number of requests with a upper limit
3456 * osc_reqpool_maxreqcount. There might be some race which will cause
3457 * over-limit allocation, but it is fine.
3459 req_count = atomic_read(&osc_pool_req_count);
3460 if (req_count < osc_reqpool_maxreqcount) {
3461 adding = cli->cl_max_rpcs_in_flight + 2;
3462 if (req_count + adding > osc_reqpool_maxreqcount)
3463 adding = osc_reqpool_maxreqcount - req_count;
3465 added = ptlrpc_add_rqs_to_pool(osc_rq_pool, adding);
3466 atomic_add(added, &osc_pool_req_count);
3469 ns_register_cancel(obd->obd_namespace, osc_cancel_weight);
3471 spin_lock(&osc_shrink_lock);
3472 list_add_tail(&cli->cl_shrink_list, &osc_shrink_list);
3473 spin_unlock(&osc_shrink_lock);
3474 cli->cl_import->imp_idle_timeout = osc_idle_timeout;
3475 cli->cl_import->imp_idle_debug = D_HA;
3480 int osc_precleanup_common(struct obd_device *obd)
3482 struct client_obd *cli = &obd->u.cli;
3486 * for echo client, export may be on zombie list, wait for
3487 * zombie thread to cull it, because cli.cl_import will be
3488 * cleared in client_disconnect_export():
3489 * class_export_destroy() -> obd_cleanup() ->
3490 * echo_device_free() -> echo_client_cleanup() ->
3491 * obd_disconnect() -> osc_disconnect() ->
3492 * client_disconnect_export()
3494 obd_zombie_barrier();
3495 if (cli->cl_writeback_work) {
3496 ptlrpcd_destroy_work(cli->cl_writeback_work);
3497 cli->cl_writeback_work = NULL;
3500 if (cli->cl_lru_work) {
3501 ptlrpcd_destroy_work(cli->cl_lru_work);
3502 cli->cl_lru_work = NULL;
3505 obd_cleanup_client_import(obd);
3508 EXPORT_SYMBOL(osc_precleanup_common);
3510 static int osc_precleanup(struct obd_device *obd)
3514 osc_precleanup_common(obd);
3516 ptlrpc_lprocfs_unregister_obd(obd);
3520 int osc_cleanup_common(struct obd_device *obd)
3522 struct client_obd *cli = &obd->u.cli;
3527 spin_lock(&osc_shrink_lock);
3528 list_del(&cli->cl_shrink_list);
3529 spin_unlock(&osc_shrink_lock);
3532 if (cli->cl_cache != NULL) {
3533 LASSERT(atomic_read(&cli->cl_cache->ccc_users) > 0);
3534 spin_lock(&cli->cl_cache->ccc_lru_lock);
3535 list_del_init(&cli->cl_lru_osc);
3536 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3537 cli->cl_lru_left = NULL;
3538 cl_cache_decref(cli->cl_cache);
3539 cli->cl_cache = NULL;
3542 /* free memory of osc quota cache */
3543 osc_quota_cleanup(obd);
3545 rc = client_obd_cleanup(obd);
3550 EXPORT_SYMBOL(osc_cleanup_common);
3552 static const struct obd_ops osc_obd_ops = {
3553 .o_owner = THIS_MODULE,
3554 .o_setup = osc_setup,
3555 .o_precleanup = osc_precleanup,
3556 .o_cleanup = osc_cleanup_common,
3557 .o_add_conn = client_import_add_conn,
3558 .o_del_conn = client_import_del_conn,
3559 .o_connect = client_connect_import,
3560 .o_reconnect = osc_reconnect,
3561 .o_disconnect = osc_disconnect,
3562 .o_statfs = osc_statfs,
3563 .o_statfs_async = osc_statfs_async,
3564 .o_create = osc_create,
3565 .o_destroy = osc_destroy,
3566 .o_getattr = osc_getattr,
3567 .o_setattr = osc_setattr,
3568 .o_iocontrol = osc_iocontrol,
3569 .o_set_info_async = osc_set_info_async,
3570 .o_import_event = osc_import_event,
3571 .o_quotactl = osc_quotactl,
3574 static struct shrinker *osc_cache_shrinker;
3575 LIST_HEAD(osc_shrink_list);
3576 DEFINE_SPINLOCK(osc_shrink_lock);
3578 #ifndef HAVE_SHRINKER_COUNT
3579 static int osc_cache_shrink(SHRINKER_ARGS(sc, nr_to_scan, gfp_mask))
3581 struct shrink_control scv = {
3582 .nr_to_scan = shrink_param(sc, nr_to_scan),
3583 .gfp_mask = shrink_param(sc, gfp_mask)
3585 (void)osc_cache_shrink_scan(shrinker, &scv);
3587 return osc_cache_shrink_count(shrinker, &scv);
3591 static int __init osc_init(void)
3593 unsigned int reqpool_size;
3594 unsigned int reqsize;
3596 DEF_SHRINKER_VAR(osc_shvar, osc_cache_shrink,
3597 osc_cache_shrink_count, osc_cache_shrink_scan);
3600 /* print an address of _any_ initialized kernel symbol from this
3601 * module, to allow debugging with gdb that doesn't support data
3602 * symbols from modules.*/
3603 CDEBUG(D_INFO, "Lustre OSC module (%p).\n", &osc_caches);
3605 rc = lu_kmem_init(osc_caches);
3609 rc = class_register_type(&osc_obd_ops, NULL, true, NULL,
3610 LUSTRE_OSC_NAME, &osc_device_type);
3614 osc_cache_shrinker = set_shrinker(DEFAULT_SEEKS, &osc_shvar);
3616 /* This is obviously too much memory, only prevent overflow here */
3617 if (osc_reqpool_mem_max >= 1 << 12 || osc_reqpool_mem_max == 0)
3618 GOTO(out_type, rc = -EINVAL);
3620 reqpool_size = osc_reqpool_mem_max << 20;
3623 while (reqsize < OST_IO_MAXREQSIZE)
3624 reqsize = reqsize << 1;
3627 * We don't enlarge the request count in OSC pool according to
3628 * cl_max_rpcs_in_flight. The allocation from the pool will only be
3629 * tried after normal allocation failed. So a small OSC pool won't
3630 * cause much performance degression in most of cases.
3632 osc_reqpool_maxreqcount = reqpool_size / reqsize;
3634 atomic_set(&osc_pool_req_count, 0);
3635 osc_rq_pool = ptlrpc_init_rq_pool(0, OST_IO_MAXREQSIZE,
3636 ptlrpc_add_rqs_to_pool);
3638 if (osc_rq_pool == NULL)
3639 GOTO(out_type, rc = -ENOMEM);
3641 rc = osc_start_grant_work();
3643 GOTO(out_req_pool, rc);
3648 ptlrpc_free_rq_pool(osc_rq_pool);
3650 class_unregister_type(LUSTRE_OSC_NAME);
3652 lu_kmem_fini(osc_caches);
3657 static void __exit osc_exit(void)
3659 osc_stop_grant_work();
3660 remove_shrinker(osc_cache_shrinker);
3661 class_unregister_type(LUSTRE_OSC_NAME);
3662 lu_kmem_fini(osc_caches);
3663 ptlrpc_free_rq_pool(osc_rq_pool);
3666 MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
3667 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
3668 MODULE_VERSION(LUSTRE_VERSION_STRING);
3669 MODULE_LICENSE("GPL");
3671 module_init(osc_init);
3672 module_exit(osc_exit);