1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
38 # define EXPORT_SYMTAB
40 #define DEBUG_SUBSYSTEM S_OSC
42 #include <libcfs/libcfs.h>
45 # include <liblustre.h>
48 #include <lustre_dlm.h>
49 #include <lustre_net.h>
50 #include <lustre/lustre_user.h>
51 #include <obd_cksum.h>
59 #include <lustre_ha.h>
60 #include <lprocfs_status.h>
61 #include <lustre_log.h>
62 #include <lustre_debug.h>
63 #include <lustre_param.h>
64 #include "osc_internal.h"
66 static quota_interface_t *quota_interface = NULL;
67 extern quota_interface_t osc_quota_interface;
69 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
70 static int brw_interpret(const struct lu_env *env,
71 struct ptlrpc_request *req, void *data, int rc);
72 int osc_cleanup(struct obd_device *obd);
74 /* Pack OSC object metadata for disk storage (LE byte order). */
75 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
76 struct lov_stripe_md *lsm)
81 lmm_size = sizeof(**lmmp);
86 OBD_FREE(*lmmp, lmm_size);
92 OBD_ALLOC(*lmmp, lmm_size);
98 LASSERT(lsm->lsm_object_id);
99 LASSERT_SEQ_IS_MDT(lsm->lsm_object_seq);
100 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
101 (*lmmp)->lmm_object_seq = cpu_to_le64(lsm->lsm_object_seq);
107 /* Unpack OSC object metadata from disk storage (LE byte order). */
108 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
109 struct lov_mds_md *lmm, int lmm_bytes)
115 if (lmm_bytes < sizeof (*lmm)) {
116 CERROR("lov_mds_md too small: %d, need %d\n",
117 lmm_bytes, (int)sizeof(*lmm));
120 /* XXX LOV_MAGIC etc check? */
122 if (lmm->lmm_object_id == 0) {
123 CERROR("lov_mds_md: zero lmm_object_id\n");
128 lsm_size = lov_stripe_md_size(1);
132 if (*lsmp != NULL && lmm == NULL) {
133 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
134 OBD_FREE(*lsmp, lsm_size);
140 OBD_ALLOC(*lsmp, lsm_size);
143 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
144 if ((*lsmp)->lsm_oinfo[0] == NULL) {
145 OBD_FREE(*lsmp, lsm_size);
148 loi_init((*lsmp)->lsm_oinfo[0]);
152 /* XXX zero *lsmp? */
153 (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
154 (*lsmp)->lsm_object_seq = le64_to_cpu (lmm->lmm_object_seq);
155 LASSERT((*lsmp)->lsm_object_id);
156 LASSERT_SEQ_IS_MDT((*lsmp)->lsm_object_seq);
159 (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
164 static inline void osc_pack_capa(struct ptlrpc_request *req,
165 struct ost_body *body, void *capa)
167 struct obd_capa *oc = (struct obd_capa *)capa;
168 struct lustre_capa *c;
173 c = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
176 body->oa.o_valid |= OBD_MD_FLOSSCAPA;
177 DEBUG_CAPA(D_SEC, c, "pack");
180 static inline void osc_pack_req_body(struct ptlrpc_request *req,
181 struct obd_info *oinfo)
183 struct ost_body *body;
185 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
188 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
189 osc_pack_capa(req, body, oinfo->oi_capa);
192 static inline void osc_set_capa_size(struct ptlrpc_request *req,
193 const struct req_msg_field *field,
197 req_capsule_set_size(&req->rq_pill, field, RCL_CLIENT, 0);
199 /* it is already calculated as sizeof struct obd_capa */
203 static int osc_getattr_interpret(const struct lu_env *env,
204 struct ptlrpc_request *req,
205 struct osc_async_args *aa, int rc)
207 struct ost_body *body;
213 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
215 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
216 lustre_get_wire_obdo(aa->aa_oi->oi_oa, &body->oa);
218 /* This should really be sent by the OST */
219 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
220 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
222 CDEBUG(D_INFO, "can't unpack ost_body\n");
224 aa->aa_oi->oi_oa->o_valid = 0;
227 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
231 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
232 struct ptlrpc_request_set *set)
234 struct ptlrpc_request *req;
235 struct osc_async_args *aa;
239 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
243 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
244 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
246 ptlrpc_request_free(req);
250 osc_pack_req_body(req, oinfo);
252 ptlrpc_request_set_replen(req);
253 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
255 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
256 aa = ptlrpc_req_async_args(req);
259 ptlrpc_set_add_req(set, req);
263 static int osc_getattr(struct obd_export *exp, struct obd_info *oinfo)
265 struct ptlrpc_request *req;
266 struct ost_body *body;
270 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
274 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
275 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
277 ptlrpc_request_free(req);
281 osc_pack_req_body(req, oinfo);
283 ptlrpc_request_set_replen(req);
285 rc = ptlrpc_queue_wait(req);
289 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
291 GOTO(out, rc = -EPROTO);
293 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
294 lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
296 /* This should really be sent by the OST */
297 oinfo->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
298 oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
302 ptlrpc_req_finished(req);
306 static int osc_setattr(struct obd_export *exp, struct obd_info *oinfo,
307 struct obd_trans_info *oti)
309 struct ptlrpc_request *req;
310 struct ost_body *body;
314 LASSERT(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP);
316 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
320 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
321 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
323 ptlrpc_request_free(req);
327 osc_pack_req_body(req, oinfo);
329 ptlrpc_request_set_replen(req);
331 rc = ptlrpc_queue_wait(req);
335 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
337 GOTO(out, rc = -EPROTO);
339 lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
343 ptlrpc_req_finished(req);
347 static int osc_setattr_interpret(const struct lu_env *env,
348 struct ptlrpc_request *req,
349 struct osc_setattr_args *sa, int rc)
351 struct ost_body *body;
357 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
359 GOTO(out, rc = -EPROTO);
361 lustre_get_wire_obdo(sa->sa_oa, &body->oa);
363 rc = sa->sa_upcall(sa->sa_cookie, rc);
367 int osc_setattr_async_base(struct obd_export *exp, struct obd_info *oinfo,
368 struct obd_trans_info *oti,
369 obd_enqueue_update_f upcall, void *cookie,
370 struct ptlrpc_request_set *rqset)
372 struct ptlrpc_request *req;
373 struct osc_setattr_args *sa;
377 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
381 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
382 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
384 ptlrpc_request_free(req);
388 if (oti && oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
389 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
391 osc_pack_req_body(req, oinfo);
393 ptlrpc_request_set_replen(req);
395 /* do mds to ost setattr asynchronously */
397 /* Do not wait for response. */
398 ptlrpcd_add_req(req, PSCOPE_OTHER);
400 req->rq_interpret_reply =
401 (ptlrpc_interpterer_t)osc_setattr_interpret;
403 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
404 sa = ptlrpc_req_async_args(req);
405 sa->sa_oa = oinfo->oi_oa;
406 sa->sa_upcall = upcall;
407 sa->sa_cookie = cookie;
409 if (rqset == PTLRPCD_SET)
410 ptlrpcd_add_req(req, PSCOPE_OTHER);
412 ptlrpc_set_add_req(rqset, req);
418 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
419 struct obd_trans_info *oti,
420 struct ptlrpc_request_set *rqset)
422 return osc_setattr_async_base(exp, oinfo, oti,
423 oinfo->oi_cb_up, oinfo, rqset);
426 int osc_real_create(struct obd_export *exp, struct obdo *oa,
427 struct lov_stripe_md **ea, struct obd_trans_info *oti)
429 struct ptlrpc_request *req;
430 struct ost_body *body;
431 struct lov_stripe_md *lsm;
440 rc = obd_alloc_memmd(exp, &lsm);
445 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
447 GOTO(out, rc = -ENOMEM);
449 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
451 ptlrpc_request_free(req);
455 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
457 lustre_set_wire_obdo(&body->oa, oa);
459 ptlrpc_request_set_replen(req);
461 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
462 oa->o_flags == OBD_FL_DELORPHAN) {
464 "delorphan from OST integration");
465 /* Don't resend the delorphan req */
466 req->rq_no_resend = req->rq_no_delay = 1;
469 rc = ptlrpc_queue_wait(req);
473 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
475 GOTO(out_req, rc = -EPROTO);
477 lustre_get_wire_obdo(oa, &body->oa);
479 /* This should really be sent by the OST */
480 oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
481 oa->o_valid |= OBD_MD_FLBLKSZ;
483 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
484 * have valid lsm_oinfo data structs, so don't go touching that.
485 * This needs to be fixed in a big way.
487 lsm->lsm_object_id = oa->o_id;
488 lsm->lsm_object_seq = oa->o_seq;
492 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
494 if (oa->o_valid & OBD_MD_FLCOOKIE) {
495 if (!oti->oti_logcookies)
496 oti_alloc_cookies(oti, 1);
497 *oti->oti_logcookies = oa->o_lcookie;
501 CDEBUG(D_HA, "transno: "LPD64"\n",
502 lustre_msg_get_transno(req->rq_repmsg));
504 ptlrpc_req_finished(req);
507 obd_free_memmd(exp, &lsm);
511 int osc_punch_base(struct obd_export *exp, struct obd_info *oinfo,
512 obd_enqueue_update_f upcall, void *cookie,
513 struct ptlrpc_request_set *rqset)
515 struct ptlrpc_request *req;
516 struct osc_setattr_args *sa;
517 struct ost_body *body;
521 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
525 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
526 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
528 ptlrpc_request_free(req);
531 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
532 ptlrpc_at_set_req_timeout(req);
534 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
536 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
537 osc_pack_capa(req, body, oinfo->oi_capa);
539 ptlrpc_request_set_replen(req);
542 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
543 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
544 sa = ptlrpc_req_async_args(req);
545 sa->sa_oa = oinfo->oi_oa;
546 sa->sa_upcall = upcall;
547 sa->sa_cookie = cookie;
548 if (rqset == PTLRPCD_SET)
549 ptlrpcd_add_req(req, PSCOPE_OTHER);
551 ptlrpc_set_add_req(rqset, req);
556 static int osc_punch(struct obd_export *exp, struct obd_info *oinfo,
557 struct obd_trans_info *oti,
558 struct ptlrpc_request_set *rqset)
560 oinfo->oi_oa->o_size = oinfo->oi_policy.l_extent.start;
561 oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
562 oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
563 return osc_punch_base(exp, oinfo,
564 oinfo->oi_cb_up, oinfo, rqset);
567 static int osc_sync(struct obd_export *exp, struct obdo *oa,
568 struct lov_stripe_md *md, obd_size start, obd_size end,
571 struct ptlrpc_request *req;
572 struct ost_body *body;
577 CDEBUG(D_INFO, "oa NULL\n");
581 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
585 osc_set_capa_size(req, &RMF_CAPA1, capa);
586 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
588 ptlrpc_request_free(req);
592 /* overload the size and blocks fields in the oa with start/end */
593 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
595 lustre_set_wire_obdo(&body->oa, oa);
596 body->oa.o_size = start;
597 body->oa.o_blocks = end;
598 body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
599 osc_pack_capa(req, body, capa);
601 ptlrpc_request_set_replen(req);
603 rc = ptlrpc_queue_wait(req);
607 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
609 GOTO(out, rc = -EPROTO);
611 lustre_get_wire_obdo(oa, &body->oa);
615 ptlrpc_req_finished(req);
619 /* Find and cancel locally locks matched by @mode in the resource found by
620 * @objid. Found locks are added into @cancel list. Returns the amount of
621 * locks added to @cancels list. */
622 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
624 ldlm_mode_t mode, int lock_flags)
626 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
627 struct ldlm_res_id res_id;
628 struct ldlm_resource *res;
632 osc_build_res_name(oa->o_id, oa->o_seq, &res_id);
633 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
637 LDLM_RESOURCE_ADDREF(res);
638 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
639 lock_flags, 0, NULL);
640 LDLM_RESOURCE_DELREF(res);
641 ldlm_resource_putref(res);
645 static int osc_destroy_interpret(const struct lu_env *env,
646 struct ptlrpc_request *req, void *data,
649 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
651 cfs_atomic_dec(&cli->cl_destroy_in_flight);
652 cfs_waitq_signal(&cli->cl_destroy_waitq);
656 static int osc_can_send_destroy(struct client_obd *cli)
658 if (cfs_atomic_inc_return(&cli->cl_destroy_in_flight) <=
659 cli->cl_max_rpcs_in_flight) {
660 /* The destroy request can be sent */
663 if (cfs_atomic_dec_return(&cli->cl_destroy_in_flight) <
664 cli->cl_max_rpcs_in_flight) {
666 * The counter has been modified between the two atomic
669 cfs_waitq_signal(&cli->cl_destroy_waitq);
674 /* Destroy requests can be async always on the client, and we don't even really
675 * care about the return code since the client cannot do anything at all about
677 * When the MDS is unlinking a filename, it saves the file objects into a
678 * recovery llog, and these object records are cancelled when the OST reports
679 * they were destroyed and sync'd to disk (i.e. transaction committed).
680 * If the client dies, or the OST is down when the object should be destroyed,
681 * the records are not cancelled, and when the OST reconnects to the MDS next,
682 * it will retrieve the llog unlink logs and then sends the log cancellation
683 * cookies to the MDS after committing destroy transactions. */
684 static int osc_destroy(struct obd_export *exp, struct obdo *oa,
685 struct lov_stripe_md *ea, struct obd_trans_info *oti,
686 struct obd_export *md_export, void *capa)
688 struct client_obd *cli = &exp->exp_obd->u.cli;
689 struct ptlrpc_request *req;
690 struct ost_body *body;
691 CFS_LIST_HEAD(cancels);
696 CDEBUG(D_INFO, "oa NULL\n");
700 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
701 LDLM_FL_DISCARD_DATA);
703 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
705 ldlm_lock_list_put(&cancels, l_bl_ast, count);
709 osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
710 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
713 ptlrpc_request_free(req);
717 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
718 ptlrpc_at_set_req_timeout(req);
720 if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
721 oa->o_lcookie = *oti->oti_logcookies;
722 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
724 lustre_set_wire_obdo(&body->oa, oa);
726 osc_pack_capa(req, body, (struct obd_capa *)capa);
727 ptlrpc_request_set_replen(req);
729 /* don't throttle destroy RPCs for the MDT */
730 if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
731 req->rq_interpret_reply = osc_destroy_interpret;
732 if (!osc_can_send_destroy(cli)) {
733 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP,
737 * Wait until the number of on-going destroy RPCs drops
738 * under max_rpc_in_flight
740 l_wait_event_exclusive(cli->cl_destroy_waitq,
741 osc_can_send_destroy(cli), &lwi);
745 /* Do not wait for response */
746 ptlrpcd_add_req(req, PSCOPE_OTHER);
750 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
753 obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
755 LASSERT(!(oa->o_valid & bits));
758 client_obd_list_lock(&cli->cl_loi_list_lock);
759 oa->o_dirty = cli->cl_dirty;
760 if (cli->cl_dirty - cli->cl_dirty_transit > cli->cl_dirty_max) {
761 CERROR("dirty %lu - %lu > dirty_max %lu\n",
762 cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
764 } else if (cfs_atomic_read(&obd_dirty_pages) -
765 cfs_atomic_read(&obd_dirty_transit_pages) >
766 obd_max_dirty_pages + 1){
767 /* The cfs_atomic_read() allowing the cfs_atomic_inc() are
768 * not covered by a lock thus they may safely race and trip
769 * this CERROR() unless we add in a small fudge factor (+1). */
770 CERROR("dirty %d - %d > system dirty_max %d\n",
771 cfs_atomic_read(&obd_dirty_pages),
772 cfs_atomic_read(&obd_dirty_transit_pages),
773 obd_max_dirty_pages);
775 } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
776 CERROR("dirty %lu - dirty_max %lu too big???\n",
777 cli->cl_dirty, cli->cl_dirty_max);
780 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
781 (cli->cl_max_rpcs_in_flight + 1);
782 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
784 oa->o_grant = cli->cl_avail_grant;
785 oa->o_dropped = cli->cl_lost_grant;
786 cli->cl_lost_grant = 0;
787 client_obd_list_unlock(&cli->cl_loi_list_lock);
788 CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
789 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
793 static void osc_update_next_shrink(struct client_obd *cli)
795 cli->cl_next_shrink_grant =
796 cfs_time_shift(cli->cl_grant_shrink_interval);
797 CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
798 cli->cl_next_shrink_grant);
801 /* caller must hold loi_list_lock */
802 static void osc_consume_write_grant(struct client_obd *cli,
803 struct brw_page *pga)
805 LASSERT_SPIN_LOCKED(&cli->cl_loi_list_lock.lock);
806 LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
807 cfs_atomic_inc(&obd_dirty_pages);
808 cli->cl_dirty += CFS_PAGE_SIZE;
809 cli->cl_avail_grant -= CFS_PAGE_SIZE;
810 pga->flag |= OBD_BRW_FROM_GRANT;
811 CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
812 CFS_PAGE_SIZE, pga, pga->pg);
813 LASSERT(cli->cl_avail_grant >= 0);
814 osc_update_next_shrink(cli);
817 /* the companion to osc_consume_write_grant, called when a brw has completed.
818 * must be called with the loi lock held. */
819 static void osc_release_write_grant(struct client_obd *cli,
820 struct brw_page *pga, int sent)
822 int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
825 LASSERT_SPIN_LOCKED(&cli->cl_loi_list_lock.lock);
826 if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
831 pga->flag &= ~OBD_BRW_FROM_GRANT;
832 cfs_atomic_dec(&obd_dirty_pages);
833 cli->cl_dirty -= CFS_PAGE_SIZE;
834 if (pga->flag & OBD_BRW_NOCACHE) {
835 pga->flag &= ~OBD_BRW_NOCACHE;
836 cfs_atomic_dec(&obd_dirty_transit_pages);
837 cli->cl_dirty_transit -= CFS_PAGE_SIZE;
840 cli->cl_lost_grant += CFS_PAGE_SIZE;
841 CDEBUG(D_CACHE, "lost grant: %lu avail grant: %lu dirty: %lu\n",
842 cli->cl_lost_grant, cli->cl_avail_grant, cli->cl_dirty);
843 } else if (CFS_PAGE_SIZE != blocksize && pga->count != CFS_PAGE_SIZE) {
844 /* For short writes we shouldn't count parts of pages that
845 * span a whole block on the OST side, or our accounting goes
846 * wrong. Should match the code in filter_grant_check. */
847 int offset = pga->off & ~CFS_PAGE_MASK;
848 int count = pga->count + (offset & (blocksize - 1));
849 int end = (offset + pga->count) & (blocksize - 1);
851 count += blocksize - end;
853 cli->cl_lost_grant += CFS_PAGE_SIZE - count;
854 CDEBUG(D_CACHE, "lost %lu grant: %lu avail: %lu dirty: %lu\n",
855 CFS_PAGE_SIZE - count, cli->cl_lost_grant,
856 cli->cl_avail_grant, cli->cl_dirty);
862 static unsigned long rpcs_in_flight(struct client_obd *cli)
864 return cli->cl_r_in_flight + cli->cl_w_in_flight;
867 int osc_wake_sync_fs(struct client_obd *cli)
870 if (cfs_list_empty(&cli->cl_loi_sync_fs_list) &&
871 cli->cl_sf_wait.started) {
872 cli->cl_sf_wait.sfw_upcall(cli->cl_sf_wait.sfw_oi, 0);
873 cli->cl_sf_wait.started = 0;
878 /* caller must hold loi_list_lock */
879 void osc_wake_cache_waiters(struct client_obd *cli)
882 struct osc_cache_waiter *ocw;
885 cfs_list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
886 /* if we can't dirty more, we must wait until some is written */
887 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
888 (cfs_atomic_read(&obd_dirty_pages) + 1 >
889 obd_max_dirty_pages)) {
890 CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
891 "osc max %ld, sys max %d\n", cli->cl_dirty,
892 cli->cl_dirty_max, obd_max_dirty_pages);
896 /* if still dirty cache but no grant wait for pending RPCs that
897 * may yet return us some grant before doing sync writes */
898 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
899 CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
900 cli->cl_w_in_flight);
904 ocw = cfs_list_entry(l, struct osc_cache_waiter, ocw_entry);
905 cfs_list_del_init(&ocw->ocw_entry);
906 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
907 /* no more RPCs in flight to return grant, do sync IO */
908 ocw->ocw_rc = -EDQUOT;
909 CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
911 osc_consume_write_grant(cli,
912 &ocw->ocw_oap->oap_brw_page);
915 cfs_waitq_signal(&ocw->ocw_waitq);
921 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
923 client_obd_list_lock(&cli->cl_loi_list_lock);
924 cli->cl_avail_grant += grant;
925 client_obd_list_unlock(&cli->cl_loi_list_lock);
928 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
930 if (body->oa.o_valid & OBD_MD_FLGRANT) {
931 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
932 __osc_update_grant(cli, body->oa.o_grant);
936 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
937 void *key, obd_count vallen, void *val,
938 struct ptlrpc_request_set *set);
940 static int osc_shrink_grant_interpret(const struct lu_env *env,
941 struct ptlrpc_request *req,
944 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
945 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
946 struct ost_body *body;
949 __osc_update_grant(cli, oa->o_grant);
953 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
955 osc_update_grant(cli, body);
961 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
963 client_obd_list_lock(&cli->cl_loi_list_lock);
964 oa->o_grant = cli->cl_avail_grant / 4;
965 cli->cl_avail_grant -= oa->o_grant;
966 client_obd_list_unlock(&cli->cl_loi_list_lock);
967 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
968 oa->o_valid |= OBD_MD_FLFLAGS;
971 oa->o_flags |= OBD_FL_SHRINK_GRANT;
972 osc_update_next_shrink(cli);
975 /* Shrink the current grant, either from some large amount to enough for a
976 * full set of in-flight RPCs, or if we have already shrunk to that limit
977 * then to enough for a single RPC. This avoids keeping more grant than
978 * needed, and avoids shrinking the grant piecemeal. */
979 static int osc_shrink_grant(struct client_obd *cli)
981 long target = (cli->cl_max_rpcs_in_flight + 1) *
982 cli->cl_max_pages_per_rpc;
984 client_obd_list_lock(&cli->cl_loi_list_lock);
985 if (cli->cl_avail_grant <= target)
986 target = cli->cl_max_pages_per_rpc;
987 client_obd_list_unlock(&cli->cl_loi_list_lock);
989 return osc_shrink_grant_to_target(cli, target);
992 int osc_shrink_grant_to_target(struct client_obd *cli, long target)
995 struct ost_body *body;
998 client_obd_list_lock(&cli->cl_loi_list_lock);
999 /* Don't shrink if we are already above or below the desired limit
1000 * We don't want to shrink below a single RPC, as that will negatively
1001 * impact block allocation and long-term performance. */
1002 if (target < cli->cl_max_pages_per_rpc)
1003 target = cli->cl_max_pages_per_rpc;
1005 if (target >= cli->cl_avail_grant) {
1006 client_obd_list_unlock(&cli->cl_loi_list_lock);
1009 client_obd_list_unlock(&cli->cl_loi_list_lock);
1011 OBD_ALLOC_PTR(body);
1015 osc_announce_cached(cli, &body->oa, 0);
1017 client_obd_list_lock(&cli->cl_loi_list_lock);
1018 body->oa.o_grant = cli->cl_avail_grant - target;
1019 cli->cl_avail_grant = target;
1020 client_obd_list_unlock(&cli->cl_loi_list_lock);
1021 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
1022 body->oa.o_valid |= OBD_MD_FLFLAGS;
1023 body->oa.o_flags = 0;
1025 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
1026 osc_update_next_shrink(cli);
1028 rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
1029 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
1030 sizeof(*body), body, NULL);
1032 __osc_update_grant(cli, body->oa.o_grant);
1037 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
1038 static int osc_should_shrink_grant(struct client_obd *client)
1040 cfs_time_t time = cfs_time_current();
1041 cfs_time_t next_shrink = client->cl_next_shrink_grant;
1043 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
1044 OBD_CONNECT_GRANT_SHRINK) == 0)
1047 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
1048 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
1049 client->cl_avail_grant > GRANT_SHRINK_LIMIT)
1052 osc_update_next_shrink(client);
1057 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1059 struct client_obd *client;
1061 cfs_list_for_each_entry(client, &item->ti_obd_list,
1062 cl_grant_shrink_list) {
1063 if (osc_should_shrink_grant(client))
1064 osc_shrink_grant(client);
1069 static int osc_add_shrink_grant(struct client_obd *client)
1073 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
1075 osc_grant_shrink_grant_cb, NULL,
1076 &client->cl_grant_shrink_list);
1078 CERROR("add grant client %s error %d\n",
1079 client->cl_import->imp_obd->obd_name, rc);
1082 CDEBUG(D_CACHE, "add grant client %s \n",
1083 client->cl_import->imp_obd->obd_name);
1084 osc_update_next_shrink(client);
1088 static int osc_del_shrink_grant(struct client_obd *client)
1090 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1094 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1097 * ocd_grant is the total grant amount we're expect to hold: if we've
1098 * been evicted, it's the new avail_grant amount, cl_dirty will drop
1099 * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1101 * race is tolerable here: if we're evicted, but imp_state already
1102 * left EVICTED state, then cl_dirty must be 0 already.
1104 client_obd_list_lock(&cli->cl_loi_list_lock);
1105 if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1106 cli->cl_avail_grant = ocd->ocd_grant;
1108 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1110 if (cli->cl_avail_grant < 0) {
1111 CWARN("%s: available grant < 0, the OSS is probably not running"
1112 " with patch from bug20278 (%ld) \n",
1113 cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant);
1114 /* workaround for 1.6 servers which do not have
1115 * the patch from bug20278 */
1116 cli->cl_avail_grant = ocd->ocd_grant;
1119 client_obd_list_unlock(&cli->cl_loi_list_lock);
1121 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
1122 cli->cl_import->imp_obd->obd_name,
1123 cli->cl_avail_grant, cli->cl_lost_grant);
1125 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1126 cfs_list_empty(&cli->cl_grant_shrink_list))
1127 osc_add_shrink_grant(cli);
1130 /* We assume that the reason this OSC got a short read is because it read
1131 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1132 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1133 * this stripe never got written at or beyond this stripe offset yet. */
1134 static void handle_short_read(int nob_read, obd_count page_count,
1135 struct brw_page **pga)
1140 /* skip bytes read OK */
1141 while (nob_read > 0) {
1142 LASSERT (page_count > 0);
1144 if (pga[i]->count > nob_read) {
1145 /* EOF inside this page */
1146 ptr = cfs_kmap(pga[i]->pg) +
1147 (pga[i]->off & ~CFS_PAGE_MASK);
1148 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1149 cfs_kunmap(pga[i]->pg);
1155 nob_read -= pga[i]->count;
1160 /* zero remaining pages */
1161 while (page_count-- > 0) {
1162 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1163 memset(ptr, 0, pga[i]->count);
1164 cfs_kunmap(pga[i]->pg);
1169 static int check_write_rcs(struct ptlrpc_request *req,
1170 int requested_nob, int niocount,
1171 obd_count page_count, struct brw_page **pga)
1176 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1177 sizeof(*remote_rcs) *
1179 if (remote_rcs == NULL) {
1180 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1184 /* return error if any niobuf was in error */
1185 for (i = 0; i < niocount; i++) {
1186 if (remote_rcs[i] < 0)
1187 return(remote_rcs[i]);
1189 if (remote_rcs[i] != 0) {
1190 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1191 i, remote_rcs[i], req);
1196 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1197 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1198 req->rq_bulk->bd_nob_transferred, requested_nob);
1205 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1207 if (p1->flag != p2->flag) {
1208 unsigned mask = ~(OBD_BRW_FROM_GRANT|
1209 OBD_BRW_NOCACHE|OBD_BRW_SYNC|OBD_BRW_ASYNC);
1211 /* warn if we try to combine flags that we don't know to be
1212 * safe to combine */
1213 if ((p1->flag & mask) != (p2->flag & mask))
1214 CERROR("is it ok to have flags 0x%x and 0x%x in the "
1215 "same brw?\n", p1->flag, p2->flag);
1219 return (p1->off + p1->count == p2->off);
1222 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1223 struct brw_page **pga, int opc,
1224 cksum_type_t cksum_type)
1229 LASSERT (pg_count > 0);
1230 cksum = init_checksum(cksum_type);
1231 while (nob > 0 && pg_count > 0) {
1232 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1233 int off = pga[i]->off & ~CFS_PAGE_MASK;
1234 int count = pga[i]->count > nob ? nob : pga[i]->count;
1236 /* corrupt the data before we compute the checksum, to
1237 * simulate an OST->client data error */
1238 if (i == 0 && opc == OST_READ &&
1239 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1240 memcpy(ptr + off, "bad1", min(4, nob));
1241 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1242 cfs_kunmap(pga[i]->pg);
1243 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1246 nob -= pga[i]->count;
1250 /* For sending we only compute the wrong checksum instead
1251 * of corrupting the data so it is still correct on a redo */
1252 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1258 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1259 struct lov_stripe_md *lsm, obd_count page_count,
1260 struct brw_page **pga,
1261 struct ptlrpc_request **reqp,
1262 struct obd_capa *ocapa, int reserve)
1264 struct ptlrpc_request *req;
1265 struct ptlrpc_bulk_desc *desc;
1266 struct ost_body *body;
1267 struct obd_ioobj *ioobj;
1268 struct niobuf_remote *niobuf;
1269 int niocount, i, requested_nob, opc, rc;
1270 struct osc_brw_async_args *aa;
1271 struct req_capsule *pill;
1272 struct brw_page *pg_prev;
1275 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1276 RETURN(-ENOMEM); /* Recoverable */
1277 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1278 RETURN(-EINVAL); /* Fatal */
1280 if ((cmd & OBD_BRW_WRITE) != 0) {
1282 req = ptlrpc_request_alloc_pool(cli->cl_import,
1283 cli->cl_import->imp_rq_pool,
1284 &RQF_OST_BRW_WRITE);
1287 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1292 for (niocount = i = 1; i < page_count; i++) {
1293 if (!can_merge_pages(pga[i - 1], pga[i]))
1297 pill = &req->rq_pill;
1298 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1300 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1301 niocount * sizeof(*niobuf));
1302 osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1304 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1306 ptlrpc_request_free(req);
1309 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1310 ptlrpc_at_set_req_timeout(req);
1312 if (opc == OST_WRITE)
1313 desc = ptlrpc_prep_bulk_imp(req, page_count,
1314 BULK_GET_SOURCE, OST_BULK_PORTAL);
1316 desc = ptlrpc_prep_bulk_imp(req, page_count,
1317 BULK_PUT_SINK, OST_BULK_PORTAL);
1320 GOTO(out, rc = -ENOMEM);
1321 /* NB request now owns desc and will free it when it gets freed */
1323 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1324 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1325 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1326 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1328 lustre_set_wire_obdo(&body->oa, oa);
1330 obdo_to_ioobj(oa, ioobj);
1331 ioobj->ioo_bufcnt = niocount;
1332 osc_pack_capa(req, body, ocapa);
1333 LASSERT (page_count > 0);
1335 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1336 struct brw_page *pg = pga[i];
1338 LASSERT(pg->count > 0);
1339 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1340 "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1341 pg->off, pg->count);
1343 LASSERTF(i == 0 || pg->off > pg_prev->off,
1344 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1345 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1347 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1348 pg_prev->pg, page_private(pg_prev->pg),
1349 pg_prev->pg->index, pg_prev->off);
1351 LASSERTF(i == 0 || pg->off > pg_prev->off,
1352 "i %d p_c %u\n", i, page_count);
1354 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1355 (pg->flag & OBD_BRW_SRVLOCK));
1357 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1359 requested_nob += pg->count;
1361 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1363 niobuf->len += pg->count;
1365 niobuf->offset = pg->off;
1366 niobuf->len = pg->count;
1367 niobuf->flags = pg->flag;
1372 LASSERTF((void *)(niobuf - niocount) ==
1373 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1374 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1375 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1377 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1378 if (osc_should_shrink_grant(cli))
1379 osc_shrink_grant_local(cli, &body->oa);
1381 /* size[REQ_REC_OFF] still sizeof (*body) */
1382 if (opc == OST_WRITE) {
1383 if (unlikely(cli->cl_checksum) &&
1384 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1385 /* store cl_cksum_type in a local variable since
1386 * it can be changed via lprocfs */
1387 cksum_type_t cksum_type = cli->cl_cksum_type;
1389 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1390 oa->o_flags &= OBD_FL_LOCAL_MASK;
1391 body->oa.o_flags = 0;
1393 body->oa.o_flags |= cksum_type_pack(cksum_type);
1394 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1395 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1399 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1401 /* save this in 'oa', too, for later checking */
1402 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1403 oa->o_flags |= cksum_type_pack(cksum_type);
1405 /* clear out the checksum flag, in case this is a
1406 * resend but cl_checksum is no longer set. b=11238 */
1407 oa->o_valid &= ~OBD_MD_FLCKSUM;
1409 oa->o_cksum = body->oa.o_cksum;
1410 /* 1 RC per niobuf */
1411 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1412 sizeof(__u32) * niocount);
1414 if (unlikely(cli->cl_checksum) &&
1415 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1416 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1417 body->oa.o_flags = 0;
1418 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1419 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1422 ptlrpc_request_set_replen(req);
1424 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1425 aa = ptlrpc_req_async_args(req);
1427 aa->aa_requested_nob = requested_nob;
1428 aa->aa_nio_count = niocount;
1429 aa->aa_page_count = page_count;
1433 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1434 if (ocapa && reserve)
1435 aa->aa_ocapa = capa_get(ocapa);
1441 ptlrpc_req_finished(req);
1445 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1446 __u32 client_cksum, __u32 server_cksum, int nob,
1447 obd_count page_count, struct brw_page **pga,
1448 cksum_type_t client_cksum_type)
1452 cksum_type_t cksum_type;
1454 if (server_cksum == client_cksum) {
1455 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1459 /* If this is mmaped file - it can be changed at any time */
1460 if (oa->o_valid & OBD_MD_FLFLAGS && oa->o_flags & OBD_FL_MMAP)
1463 if (oa->o_valid & OBD_MD_FLFLAGS)
1464 cksum_type = cksum_type_unpack(oa->o_flags);
1466 cksum_type = OBD_CKSUM_CRC32;
1468 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1471 if (cksum_type != client_cksum_type)
1472 msg = "the server did not use the checksum type specified in "
1473 "the original request - likely a protocol problem";
1474 else if (new_cksum == server_cksum)
1475 msg = "changed on the client after we checksummed it - "
1476 "likely false positive due to mmap IO (bug 11742)";
1477 else if (new_cksum == client_cksum)
1478 msg = "changed in transit before arrival at OST";
1480 msg = "changed in transit AND doesn't match the original - "
1481 "likely false positive due to mmap IO (bug 11742)";
1483 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1484 " object "LPU64"/"LPU64" extent ["LPU64"-"LPU64"]\n",
1485 msg, libcfs_nid2str(peer->nid),
1486 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1487 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1488 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1490 oa->o_valid & OBD_MD_FLGROUP ? oa->o_seq : (__u64)0,
1492 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1493 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1494 "client csum now %x\n", client_cksum, client_cksum_type,
1495 server_cksum, cksum_type, new_cksum);
1499 /* Note rc enters this function as number of bytes transferred */
1500 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1502 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1503 const lnet_process_id_t *peer =
1504 &req->rq_import->imp_connection->c_peer;
1505 struct client_obd *cli = aa->aa_cli;
1506 struct ost_body *body;
1507 __u32 client_cksum = 0;
1510 if (rc < 0 && rc != -EDQUOT) {
1511 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1515 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1516 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1518 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1522 #ifdef HAVE_QUOTA_SUPPORT
1523 /* set/clear over quota flag for a uid/gid */
1524 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1525 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1526 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1528 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid "LPX64", flags %x\n",
1529 body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1531 lquota_setdq(quota_interface, cli, qid, body->oa.o_valid,
1536 osc_update_grant(cli, body);
1541 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1542 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1544 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1546 CERROR("Unexpected +ve rc %d\n", rc);
1549 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1551 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1554 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1555 check_write_checksum(&body->oa, peer, client_cksum,
1556 body->oa.o_cksum, aa->aa_requested_nob,
1557 aa->aa_page_count, aa->aa_ppga,
1558 cksum_type_unpack(aa->aa_oa->o_flags)))
1561 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1562 aa->aa_page_count, aa->aa_ppga);
1566 /* The rest of this function executes only for OST_READs */
1568 /* if unwrap_bulk failed, return -EAGAIN to retry */
1569 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1571 GOTO(out, rc = -EAGAIN);
1573 if (rc > aa->aa_requested_nob) {
1574 CERROR("Unexpected rc %d (%d requested)\n", rc,
1575 aa->aa_requested_nob);
1579 if (rc != req->rq_bulk->bd_nob_transferred) {
1580 CERROR ("Unexpected rc %d (%d transferred)\n",
1581 rc, req->rq_bulk->bd_nob_transferred);
1585 if (rc < aa->aa_requested_nob)
1586 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1588 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1589 static int cksum_counter;
1590 __u32 server_cksum = body->oa.o_cksum;
1593 cksum_type_t cksum_type;
1595 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1596 cksum_type = cksum_type_unpack(body->oa.o_flags);
1598 cksum_type = OBD_CKSUM_CRC32;
1599 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1600 aa->aa_ppga, OST_READ,
1603 if (peer->nid == req->rq_bulk->bd_sender) {
1607 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1610 if (server_cksum == ~0 && rc > 0) {
1611 CERROR("Protocol error: server %s set the 'checksum' "
1612 "bit, but didn't send a checksum. Not fatal, "
1613 "but please notify on http://bugzilla.lustre.org/\n",
1614 libcfs_nid2str(peer->nid));
1615 } else if (server_cksum != client_cksum) {
1616 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1617 "%s%s%s inode "DFID" object "
1618 LPU64"/"LPU64" extent "
1619 "["LPU64"-"LPU64"]\n",
1620 req->rq_import->imp_obd->obd_name,
1621 libcfs_nid2str(peer->nid),
1623 body->oa.o_valid & OBD_MD_FLFID ?
1624 body->oa.o_parent_seq : (__u64)0,
1625 body->oa.o_valid & OBD_MD_FLFID ?
1626 body->oa.o_parent_oid : 0,
1627 body->oa.o_valid & OBD_MD_FLFID ?
1628 body->oa.o_parent_ver : 0,
1630 body->oa.o_valid & OBD_MD_FLGROUP ?
1631 body->oa.o_seq : (__u64)0,
1632 aa->aa_ppga[0]->off,
1633 aa->aa_ppga[aa->aa_page_count-1]->off +
1634 aa->aa_ppga[aa->aa_page_count-1]->count -
1636 CERROR("client %x, server %x, cksum_type %x\n",
1637 client_cksum, server_cksum, cksum_type);
1639 aa->aa_oa->o_cksum = client_cksum;
1643 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1646 } else if (unlikely(client_cksum)) {
1647 static int cksum_missed;
1650 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1651 CERROR("Checksum %u requested from %s but not sent\n",
1652 cksum_missed, libcfs_nid2str(peer->nid));
1658 lustre_get_wire_obdo(aa->aa_oa, &body->oa);
1663 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1664 struct lov_stripe_md *lsm,
1665 obd_count page_count, struct brw_page **pga,
1666 struct obd_capa *ocapa)
1668 struct ptlrpc_request *req;
1672 struct l_wait_info lwi;
1676 cfs_waitq_init(&waitq);
1679 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1680 page_count, pga, &req, ocapa, 0);
1684 rc = ptlrpc_queue_wait(req);
1686 if (rc == -ETIMEDOUT && req->rq_resend) {
1687 DEBUG_REQ(D_HA, req, "BULK TIMEOUT");
1688 ptlrpc_req_finished(req);
1692 rc = osc_brw_fini_request(req, rc);
1694 ptlrpc_req_finished(req);
1695 if (osc_recoverable_error(rc)) {
1697 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1698 CERROR("too many resend retries, returning error\n");
1702 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1703 l_wait_event(waitq, 0, &lwi);
1711 int osc_brw_redo_request(struct ptlrpc_request *request,
1712 struct osc_brw_async_args *aa)
1714 struct ptlrpc_request *new_req;
1715 struct ptlrpc_request_set *set = request->rq_set;
1716 struct osc_brw_async_args *new_aa;
1717 struct osc_async_page *oap;
1721 if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1722 CERROR("too many resent retries, returning error\n");
1726 DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1728 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1729 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1730 aa->aa_cli, aa->aa_oa,
1731 NULL /* lsm unused by osc currently */,
1732 aa->aa_page_count, aa->aa_ppga,
1733 &new_req, aa->aa_ocapa, 0);
1737 client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1739 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1740 if (oap->oap_request != NULL) {
1741 LASSERTF(request == oap->oap_request,
1742 "request %p != oap_request %p\n",
1743 request, oap->oap_request);
1744 if (oap->oap_interrupted) {
1745 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1746 ptlrpc_req_finished(new_req);
1751 /* New request takes over pga and oaps from old request.
1752 * Note that copying a list_head doesn't work, need to move it... */
1754 new_req->rq_interpret_reply = request->rq_interpret_reply;
1755 new_req->rq_async_args = request->rq_async_args;
1756 new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1758 new_aa = ptlrpc_req_async_args(new_req);
1760 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1761 cfs_list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1762 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1764 cfs_list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1765 if (oap->oap_request) {
1766 ptlrpc_req_finished(oap->oap_request);
1767 oap->oap_request = ptlrpc_request_addref(new_req);
1771 new_aa->aa_ocapa = aa->aa_ocapa;
1772 aa->aa_ocapa = NULL;
1774 /* use ptlrpc_set_add_req is safe because interpret functions work
1775 * in check_set context. only one way exist with access to request
1776 * from different thread got -EINTR - this way protected with
1777 * cl_loi_list_lock */
1778 ptlrpc_set_add_req(set, new_req);
1780 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1782 DEBUG_REQ(D_INFO, new_req, "new request");
1787 * ugh, we want disk allocation on the target to happen in offset order. we'll
1788 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1789 * fine for our small page arrays and doesn't require allocation. its an
1790 * insertion sort that swaps elements that are strides apart, shrinking the
1791 * stride down until its '1' and the array is sorted.
1793 static void sort_brw_pages(struct brw_page **array, int num)
1796 struct brw_page *tmp;
1800 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1805 for (i = stride ; i < num ; i++) {
1808 while (j >= stride && array[j - stride]->off > tmp->off) {
1809 array[j] = array[j - stride];
1814 } while (stride > 1);
1817 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1823 LASSERT (pages > 0);
1824 offset = pg[i]->off & ~CFS_PAGE_MASK;
1828 if (pages == 0) /* that's all */
1831 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1832 return count; /* doesn't end on page boundary */
1835 offset = pg[i]->off & ~CFS_PAGE_MASK;
1836 if (offset != 0) /* doesn't start on page boundary */
1843 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1845 struct brw_page **ppga;
1848 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1852 for (i = 0; i < count; i++)
1857 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1859 LASSERT(ppga != NULL);
1860 OBD_FREE(ppga, sizeof(*ppga) * count);
1863 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1864 obd_count page_count, struct brw_page *pga,
1865 struct obd_trans_info *oti)
1867 struct obdo *saved_oa = NULL;
1868 struct brw_page **ppga, **orig;
1869 struct obd_import *imp = class_exp2cliimp(exp);
1870 struct client_obd *cli;
1871 int rc, page_count_orig;
1874 LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1875 cli = &imp->imp_obd->u.cli;
1877 if (cmd & OBD_BRW_CHECK) {
1878 /* The caller just wants to know if there's a chance that this
1879 * I/O can succeed */
1881 if (imp->imp_invalid)
1886 /* test_brw with a failed create can trip this, maybe others. */
1887 LASSERT(cli->cl_max_pages_per_rpc);
1891 orig = ppga = osc_build_ppga(pga, page_count);
1894 page_count_orig = page_count;
1896 sort_brw_pages(ppga, page_count);
1897 while (page_count) {
1898 obd_count pages_per_brw;
1900 if (page_count > cli->cl_max_pages_per_rpc)
1901 pages_per_brw = cli->cl_max_pages_per_rpc;
1903 pages_per_brw = page_count;
1905 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1907 if (saved_oa != NULL) {
1908 /* restore previously saved oa */
1909 *oinfo->oi_oa = *saved_oa;
1910 } else if (page_count > pages_per_brw) {
1911 /* save a copy of oa (brw will clobber it) */
1912 OBDO_ALLOC(saved_oa);
1913 if (saved_oa == NULL)
1914 GOTO(out, rc = -ENOMEM);
1915 *saved_oa = *oinfo->oi_oa;
1918 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1919 pages_per_brw, ppga, oinfo->oi_capa);
1924 page_count -= pages_per_brw;
1925 ppga += pages_per_brw;
1929 osc_release_ppga(orig, page_count_orig);
1931 if (saved_oa != NULL)
1932 OBDO_FREE(saved_oa);
1937 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1938 * the dirty accounting. Writeback completes or truncate happens before
1939 * writing starts. Must be called with the loi lock held. */
1940 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1943 osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1946 static int lop_makes_syncfs_rpc(struct loi_oap_pages *lop)
1948 struct osc_async_page *oap;
1951 if (cfs_list_empty(&lop->lop_urgent))
1954 oap = cfs_list_entry(lop->lop_urgent.next,
1955 struct osc_async_page, oap_urgent_item);
1957 if (oap->oap_async_flags & ASYNC_SYNCFS) {
1958 CDEBUG(D_CACHE, "syncfs request forcing RPC\n");
1965 /* This maintains the lists of pending pages to read/write for a given object
1966 * (lop). This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1967 * to quickly find objects that are ready to send an RPC. */
1968 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1974 if (lop->lop_num_pending == 0)
1977 /* if we have an invalid import we want to drain the queued pages
1978 * by forcing them through rpcs that immediately fail and complete
1979 * the pages. recovery relies on this to empty the queued pages
1980 * before canceling the locks and evicting down the llite pages */
1981 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1984 /* stream rpcs in queue order as long as as there is an urgent page
1985 * queued. this is our cheap solution for good batching in the case
1986 * where writepage marks some random page in the middle of the file
1987 * as urgent because of, say, memory pressure */
1988 if (!cfs_list_empty(&lop->lop_urgent)) {
1989 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1992 /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1993 optimal = cli->cl_max_pages_per_rpc;
1994 if (cmd & OBD_BRW_WRITE) {
1995 /* trigger a write rpc stream as long as there are dirtiers
1996 * waiting for space. as they're waiting, they're not going to
1997 * create more pages to coalesce with what's waiting.. */
1998 if (!cfs_list_empty(&cli->cl_cache_waiters)) {
1999 CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
2002 /* +16 to avoid triggering rpcs that would want to include pages
2003 * that are being queued but which can't be made ready until
2004 * the queuer finishes with the page. this is a wart for
2005 * llite::commit_write() */
2008 if (lop->lop_num_pending >= optimal)
2014 static int lop_makes_hprpc(struct loi_oap_pages *lop)
2016 struct osc_async_page *oap;
2019 if (cfs_list_empty(&lop->lop_urgent))
2022 oap = cfs_list_entry(lop->lop_urgent.next,
2023 struct osc_async_page, oap_urgent_item);
2025 if (oap->oap_async_flags & ASYNC_HP) {
2026 CDEBUG(D_CACHE, "hp request forcing RPC\n");
2033 static void on_list(cfs_list_t *item, cfs_list_t *list,
2036 if (cfs_list_empty(item) && should_be_on)
2037 cfs_list_add_tail(item, list);
2038 else if (!cfs_list_empty(item) && !should_be_on)
2039 cfs_list_del_init(item);
2042 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
2043 * can find pages to build into rpcs quickly */
2044 void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
2046 if (lop_makes_hprpc(&loi->loi_write_lop) ||
2047 lop_makes_hprpc(&loi->loi_read_lop)) {
2049 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list, 0);
2050 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
2052 if (lop_makes_syncfs_rpc(&loi->loi_write_lop)) {
2053 on_list(&loi->loi_sync_fs_item,
2054 &cli->cl_loi_sync_fs_list,
2055 loi->loi_write_lop.lop_num_pending);
2057 on_list(&loi->loi_hp_ready_item,
2058 &cli->cl_loi_hp_ready_list, 0);
2059 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list,
2060 lop_makes_rpc(cli, &loi->loi_write_lop,
2062 lop_makes_rpc(cli, &loi->loi_read_lop,
2067 on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
2068 loi->loi_write_lop.lop_num_pending);
2070 on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
2071 loi->loi_read_lop.lop_num_pending);
2074 static void lop_update_pending(struct client_obd *cli,
2075 struct loi_oap_pages *lop, int cmd, int delta)
2077 lop->lop_num_pending += delta;
2078 if (cmd & OBD_BRW_WRITE)
2079 cli->cl_pending_w_pages += delta;
2081 cli->cl_pending_r_pages += delta;
2085 * this is called when a sync waiter receives an interruption. Its job is to
2086 * get the caller woken as soon as possible. If its page hasn't been put in an
2087 * rpc yet it can dequeue immediately. Otherwise it has to mark the rpc as
2088 * desiring interruption which will forcefully complete the rpc once the rpc
2091 int osc_oap_interrupted(const struct lu_env *env, struct osc_async_page *oap)
2093 struct loi_oap_pages *lop;
2094 struct lov_oinfo *loi;
2098 LASSERT(!oap->oap_interrupted);
2099 oap->oap_interrupted = 1;
2101 /* ok, it's been put in an rpc. only one oap gets a request reference */
2102 if (oap->oap_request != NULL) {
2103 ptlrpc_mark_interrupted(oap->oap_request);
2104 ptlrpcd_wake(oap->oap_request);
2105 ptlrpc_req_finished(oap->oap_request);
2106 oap->oap_request = NULL;
2110 * page completion may be called only if ->cpo_prep() method was
2111 * executed by osc_io_submit(), that also adds page the to pending list
2113 if (!cfs_list_empty(&oap->oap_pending_item)) {
2114 cfs_list_del_init(&oap->oap_pending_item);
2115 cfs_list_del_init(&oap->oap_urgent_item);
2118 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
2119 &loi->loi_write_lop : &loi->loi_read_lop;
2120 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
2121 loi_list_maint(oap->oap_cli, oap->oap_loi);
2122 rc = oap->oap_caller_ops->ap_completion(env,
2123 oap->oap_caller_data,
2124 oap->oap_cmd, NULL, -EINTR);
2130 /* this is trying to propogate async writeback errors back up to the
2131 * application. As an async write fails we record the error code for later if
2132 * the app does an fsync. As long as errors persist we force future rpcs to be
2133 * sync so that the app can get a sync error and break the cycle of queueing
2134 * pages for which writeback will fail. */
2135 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
2142 ar->ar_force_sync = 1;
2143 ar->ar_min_xid = ptlrpc_sample_next_xid();
2148 if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2149 ar->ar_force_sync = 0;
2152 static int osc_add_to_lop_urgent(struct loi_oap_pages *lop,
2153 struct osc_async_page *oap,
2154 obd_flag async_flags)
2157 /* If true, then already present in lop urgent */
2158 if (!cfs_list_empty(&oap->oap_urgent_item)) {
2159 CWARN("Request to add duplicate oap_urgent for flag = %d\n",
2160 oap->oap_async_flags);
2164 /* item from sync_fs, to avoid duplicates check the existing flags */
2165 if (async_flags & ASYNC_SYNCFS) {
2166 cfs_list_add_tail(&oap->oap_urgent_item,
2171 if (oap->oap_async_flags & ASYNC_HP)
2172 cfs_list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2173 else if (oap->oap_async_flags & ASYNC_URGENT ||
2174 async_flags & ASYNC_URGENT)
2175 cfs_list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2180 void osc_oap_to_pending(struct osc_async_page *oap)
2182 struct loi_oap_pages *lop;
2184 if (oap->oap_cmd & OBD_BRW_WRITE)
2185 lop = &oap->oap_loi->loi_write_lop;
2187 lop = &oap->oap_loi->loi_read_lop;
2189 osc_add_to_lop_urgent(lop, oap, 0);
2190 cfs_list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2191 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2194 /* this must be called holding the loi list lock to give coverage to exit_cache,
2195 * async_flag maintenance, and oap_request */
2196 static void osc_ap_completion(const struct lu_env *env,
2197 struct client_obd *cli, struct obdo *oa,
2198 struct osc_async_page *oap, int sent, int rc)
2203 if (oap->oap_request != NULL) {
2204 xid = ptlrpc_req_xid(oap->oap_request);
2205 ptlrpc_req_finished(oap->oap_request);
2206 oap->oap_request = NULL;
2209 cfs_spin_lock(&oap->oap_lock);
2210 oap->oap_async_flags = 0;
2211 cfs_spin_unlock(&oap->oap_lock);
2212 oap->oap_interrupted = 0;
2214 if (oap->oap_cmd & OBD_BRW_WRITE) {
2215 osc_process_ar(&cli->cl_ar, xid, rc);
2216 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2219 if (rc == 0 && oa != NULL) {
2220 if (oa->o_valid & OBD_MD_FLBLOCKS)
2221 oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2222 if (oa->o_valid & OBD_MD_FLMTIME)
2223 oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2224 if (oa->o_valid & OBD_MD_FLATIME)
2225 oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2226 if (oa->o_valid & OBD_MD_FLCTIME)
2227 oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2230 rc = oap->oap_caller_ops->ap_completion(env, oap->oap_caller_data,
2231 oap->oap_cmd, oa, rc);
2233 /* ll_ap_completion (from llite) drops PG_locked. so, a new
2234 * I/O on the page could start, but OSC calls it under lock
2235 * and thus we can add oap back to pending safely */
2237 /* upper layer wants to leave the page on pending queue */
2238 osc_oap_to_pending(oap);
2240 osc_exit_cache(cli, oap, sent);
2244 static int brw_interpret(const struct lu_env *env,
2245 struct ptlrpc_request *req, void *data, int rc)
2247 struct osc_brw_async_args *aa = data;
2248 struct client_obd *cli;
2252 rc = osc_brw_fini_request(req, rc);
2253 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2254 if (osc_recoverable_error(rc)) {
2255 /* Only retry once for mmaped files since the mmaped page
2256 * might be modified at anytime. We have to retry at least
2257 * once in case there WAS really a corruption of the page
2258 * on the network, that was not caused by mmap() modifying
2259 * the page. Bug11742 */
2260 if ((rc == -EAGAIN) && (aa->aa_resends > 0) &&
2261 aa->aa_oa->o_valid & OBD_MD_FLFLAGS &&
2262 aa->aa_oa->o_flags & OBD_FL_MMAP) {
2265 rc = osc_brw_redo_request(req, aa);
2272 capa_put(aa->aa_ocapa);
2273 aa->aa_ocapa = NULL;
2278 client_obd_list_lock(&cli->cl_loi_list_lock);
2280 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2281 * is called so we know whether to go to sync BRWs or wait for more
2282 * RPCs to complete */
2283 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2284 cli->cl_w_in_flight--;
2286 cli->cl_r_in_flight--;
2288 async = cfs_list_empty(&aa->aa_oaps);
2289 if (!async) { /* from osc_send_oap_rpc() */
2290 struct osc_async_page *oap, *tmp;
2291 /* the caller may re-use the oap after the completion call so
2292 * we need to clean it up a little */
2293 cfs_list_for_each_entry_safe(oap, tmp, &aa->aa_oaps,
2295 cfs_list_del_init(&oap->oap_rpc_item);
2296 osc_ap_completion(env, cli, aa->aa_oa, oap, 1, rc);
2298 OBDO_FREE(aa->aa_oa);
2299 } else { /* from async_internal() */
2301 for (i = 0; i < aa->aa_page_count; i++)
2302 osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2304 osc_wake_cache_waiters(cli);
2305 osc_wake_sync_fs(cli);
2306 osc_check_rpcs(env, cli);
2307 client_obd_list_unlock(&cli->cl_loi_list_lock);
2309 cl_req_completion(env, aa->aa_clerq, rc);
2310 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2315 static struct ptlrpc_request *osc_build_req(const struct lu_env *env,
2316 struct client_obd *cli,
2317 cfs_list_t *rpc_list,
2318 int page_count, int cmd)
2320 struct ptlrpc_request *req;
2321 struct brw_page **pga = NULL;
2322 struct osc_brw_async_args *aa;
2323 struct obdo *oa = NULL;
2324 const struct obd_async_page_ops *ops = NULL;
2325 void *caller_data = NULL;
2326 struct osc_async_page *oap;
2327 struct osc_async_page *tmp;
2328 struct ost_body *body;
2329 struct cl_req *clerq = NULL;
2330 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2331 struct ldlm_lock *lock = NULL;
2332 struct cl_req_attr crattr;
2333 int i, rc, mpflag = 0;
2336 LASSERT(!cfs_list_empty(rpc_list));
2338 if (cmd & OBD_BRW_MEMALLOC)
2339 mpflag = cfs_memory_pressure_get_and_set();
2341 memset(&crattr, 0, sizeof crattr);
2342 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2344 GOTO(out, req = ERR_PTR(-ENOMEM));
2348 GOTO(out, req = ERR_PTR(-ENOMEM));
2351 cfs_list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2352 struct cl_page *page = osc_oap2cl_page(oap);
2354 ops = oap->oap_caller_ops;
2355 caller_data = oap->oap_caller_data;
2357 clerq = cl_req_alloc(env, page, crt,
2358 1 /* only 1-object rpcs for
2361 GOTO(out, req = (void *)clerq);
2362 lock = oap->oap_ldlm_lock;
2364 pga[i] = &oap->oap_brw_page;
2365 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2366 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2367 pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2369 cl_req_page_add(env, clerq, page);
2372 /* always get the data for the obdo for the rpc */
2373 LASSERT(ops != NULL);
2375 crattr.cra_capa = NULL;
2376 cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2378 oa->o_handle = lock->l_remote_handle;
2379 oa->o_valid |= OBD_MD_FLHANDLE;
2382 rc = cl_req_prep(env, clerq);
2384 CERROR("cl_req_prep failed: %d\n", rc);
2385 GOTO(out, req = ERR_PTR(rc));
2388 sort_brw_pages(pga, page_count);
2389 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2390 pga, &req, crattr.cra_capa, 1);
2392 CERROR("prep_req failed: %d\n", rc);
2393 GOTO(out, req = ERR_PTR(rc));
2396 if (cmd & OBD_BRW_MEMALLOC)
2397 req->rq_memalloc = 1;
2399 /* Need to update the timestamps after the request is built in case
2400 * we race with setattr (locally or in queue at OST). If OST gets
2401 * later setattr before earlier BRW (as determined by the request xid),
2402 * the OST will not use BRW timestamps. Sadly, there is no obvious
2403 * way to do this in a single call. bug 10150 */
2404 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2405 cl_req_attr_set(env, clerq, &crattr,
2406 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2408 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2409 aa = ptlrpc_req_async_args(req);
2410 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2411 cfs_list_splice(rpc_list, &aa->aa_oaps);
2412 CFS_INIT_LIST_HEAD(rpc_list);
2413 aa->aa_clerq = clerq;
2415 if (cmd & OBD_BRW_MEMALLOC)
2416 cfs_memory_pressure_restore(mpflag);
2418 capa_put(crattr.cra_capa);
2423 OBD_FREE(pga, sizeof(*pga) * page_count);
2424 /* this should happen rarely and is pretty bad, it makes the
2425 * pending list not follow the dirty order */
2426 client_obd_list_lock(&cli->cl_loi_list_lock);
2427 cfs_list_for_each_entry_safe(oap, tmp, rpc_list, oap_rpc_item) {
2428 cfs_list_del_init(&oap->oap_rpc_item);
2430 /* queued sync pages can be torn down while the pages
2431 * were between the pending list and the rpc */
2432 if (oap->oap_interrupted) {
2433 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2434 osc_ap_completion(env, cli, NULL, oap, 0,
2438 osc_ap_completion(env, cli, NULL, oap, 0, PTR_ERR(req));
2440 if (clerq && !IS_ERR(clerq))
2441 cl_req_completion(env, clerq, PTR_ERR(req));
2447 * prepare pages for ASYNC io and put pages in send queue.
2449 * \param cmd OBD_BRW_* macroses
2450 * \param lop pending pages
2452 * \return zero if no page added to send queue.
2453 * \return 1 if pages successfully added to send queue.
2454 * \return negative on errors.
2457 osc_send_oap_rpc(const struct lu_env *env, struct client_obd *cli,
2458 struct lov_oinfo *loi,
2459 int cmd, struct loi_oap_pages *lop)
2461 struct ptlrpc_request *req;
2462 obd_count page_count = 0;
2463 struct osc_async_page *oap = NULL, *tmp;
2464 struct osc_brw_async_args *aa;
2465 const struct obd_async_page_ops *ops;
2466 CFS_LIST_HEAD(rpc_list);
2467 CFS_LIST_HEAD(tmp_list);
2468 unsigned int ending_offset;
2469 unsigned starting_offset = 0;
2470 int srvlock = 0, mem_tight = 0;
2471 struct cl_object *clob = NULL;
2474 /* ASYNC_HP pages first. At present, when the lock the pages is
2475 * to be canceled, the pages covered by the lock will be sent out
2476 * with ASYNC_HP. We have to send out them as soon as possible. */
2477 cfs_list_for_each_entry_safe(oap, tmp, &lop->lop_urgent, oap_urgent_item) {
2478 if (oap->oap_async_flags & ASYNC_HP)
2479 cfs_list_move(&oap->oap_pending_item, &tmp_list);
2481 cfs_list_move_tail(&oap->oap_pending_item, &tmp_list);
2482 if (++page_count >= cli->cl_max_pages_per_rpc)
2486 cfs_list_splice(&tmp_list, &lop->lop_pending);
2489 /* first we find the pages we're allowed to work with */
2490 cfs_list_for_each_entry_safe(oap, tmp, &lop->lop_pending,
2492 ops = oap->oap_caller_ops;
2494 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2495 "magic 0x%x\n", oap, oap->oap_magic);
2498 /* pin object in memory, so that completion call-backs
2499 * can be safely called under client_obd_list lock. */
2500 clob = osc_oap2cl_page(oap)->cp_obj;
2501 cl_object_get(clob);
2504 if (page_count != 0 &&
2505 srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2506 CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2507 " oap %p, page %p, srvlock %u\n",
2508 oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2512 /* If there is a gap at the start of this page, it can't merge
2513 * with any previous page, so we'll hand the network a
2514 * "fragmented" page array that it can't transfer in 1 RDMA */
2515 if (page_count != 0 && oap->oap_page_off != 0)
2518 /* in llite being 'ready' equates to the page being locked
2519 * until completion unlocks it. commit_write submits a page
2520 * as not ready because its unlock will happen unconditionally
2521 * as the call returns. if we race with commit_write giving
2522 * us that page we don't want to create a hole in the page
2523 * stream, so we stop and leave the rpc to be fired by
2524 * another dirtier or kupdated interval (the not ready page
2525 * will still be on the dirty list). we could call in
2526 * at the end of ll_file_write to process the queue again. */
2527 if (!(oap->oap_async_flags & ASYNC_READY)) {
2528 int rc = ops->ap_make_ready(env, oap->oap_caller_data,
2531 CDEBUG(D_INODE, "oap %p page %p returned %d "
2532 "instead of ready\n", oap,
2536 /* llite is telling us that the page is still
2537 * in commit_write and that we should try
2538 * and put it in an rpc again later. we
2539 * break out of the loop so we don't create
2540 * a hole in the sequence of pages in the rpc
2545 /* the io isn't needed.. tell the checks
2546 * below to complete the rpc with EINTR */
2547 cfs_spin_lock(&oap->oap_lock);
2548 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2549 cfs_spin_unlock(&oap->oap_lock);
2550 oap->oap_count = -EINTR;
2553 cfs_spin_lock(&oap->oap_lock);
2554 oap->oap_async_flags |= ASYNC_READY;
2555 cfs_spin_unlock(&oap->oap_lock);
2558 LASSERTF(0, "oap %p page %p returned %d "
2559 "from make_ready\n", oap,
2567 * Page submitted for IO has to be locked. Either by
2568 * ->ap_make_ready() or by higher layers.
2570 #if defined(__KERNEL__) && defined(__linux__)
2572 struct cl_page *page;
2574 page = osc_oap2cl_page(oap);
2576 if (page->cp_type == CPT_CACHEABLE &&
2577 !(PageLocked(oap->oap_page) &&
2578 (CheckWriteback(oap->oap_page, cmd)))) {
2579 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2581 (long)oap->oap_page->flags,
2582 oap->oap_async_flags);
2588 /* take the page out of our book-keeping */
2589 cfs_list_del_init(&oap->oap_pending_item);
2590 lop_update_pending(cli, lop, cmd, -1);
2591 cfs_list_del_init(&oap->oap_urgent_item);
2593 if (page_count == 0)
2594 starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2595 (PTLRPC_MAX_BRW_SIZE - 1);
2597 /* ask the caller for the size of the io as the rpc leaves. */
2598 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
2600 ops->ap_refresh_count(env, oap->oap_caller_data,
2602 LASSERT(oap->oap_page_off + oap->oap_count <= CFS_PAGE_SIZE);
2604 if (oap->oap_count <= 0) {
2605 CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2607 osc_ap_completion(env, cli, NULL,
2608 oap, 0, oap->oap_count);
2612 /* now put the page back in our accounting */
2613 cfs_list_add_tail(&oap->oap_rpc_item, &rpc_list);
2614 if (oap->oap_brw_flags & OBD_BRW_MEMALLOC)
2616 if (page_count == 0)
2617 srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2618 if (++page_count >= cli->cl_max_pages_per_rpc)
2621 /* End on a PTLRPC_MAX_BRW_SIZE boundary. We want full-sized
2622 * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2623 * have the same alignment as the initial writes that allocated
2624 * extents on the server. */
2625 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2626 oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2627 if (ending_offset == 0)
2630 /* If there is a gap at the end of this page, it can't merge
2631 * with any subsequent pages, so we'll hand the network a
2632 * "fragmented" page array that it can't transfer in 1 RDMA */
2633 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2637 osc_wake_cache_waiters(cli);
2638 osc_wake_sync_fs(cli);
2639 loi_list_maint(cli, loi);
2641 client_obd_list_unlock(&cli->cl_loi_list_lock);
2644 cl_object_put(env, clob);
2646 if (page_count == 0) {
2647 client_obd_list_lock(&cli->cl_loi_list_lock);
2651 req = osc_build_req(env, cli, &rpc_list, page_count,
2652 mem_tight ? (cmd | OBD_BRW_MEMALLOC) : cmd);
2654 LASSERT(cfs_list_empty(&rpc_list));
2655 loi_list_maint(cli, loi);
2656 RETURN(PTR_ERR(req));
2659 aa = ptlrpc_req_async_args(req);
2661 if (cmd == OBD_BRW_READ) {
2662 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2663 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2664 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2665 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2667 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2668 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2669 cli->cl_w_in_flight);
2670 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2671 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2673 ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2675 client_obd_list_lock(&cli->cl_loi_list_lock);
2677 if (cmd == OBD_BRW_READ)
2678 cli->cl_r_in_flight++;
2680 cli->cl_w_in_flight++;
2682 /* queued sync pages can be torn down while the pages
2683 * were between the pending list and the rpc */
2685 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2686 /* only one oap gets a request reference */
2689 if (oap->oap_interrupted && !req->rq_intr) {
2690 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2692 ptlrpc_mark_interrupted(req);
2696 tmp->oap_request = ptlrpc_request_addref(req);
2698 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2699 page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2701 req->rq_interpret_reply = brw_interpret;
2702 ptlrpcd_add_req(req, PSCOPE_BRW);
2706 #define LOI_DEBUG(LOI, STR, args...) \
2707 CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR, \
2708 !cfs_list_empty(&(LOI)->loi_ready_item) || \
2709 !cfs_list_empty(&(LOI)->loi_hp_ready_item), \
2710 (LOI)->loi_write_lop.lop_num_pending, \
2711 !cfs_list_empty(&(LOI)->loi_write_lop.lop_urgent), \
2712 (LOI)->loi_read_lop.lop_num_pending, \
2713 !cfs_list_empty(&(LOI)->loi_read_lop.lop_urgent), \
2716 /* This is called by osc_check_rpcs() to find which objects have pages that
2717 * we could be sending. These lists are maintained by lop_makes_rpc(). */
2718 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2722 /* First return objects that have blocked locks so that they
2723 * will be flushed quickly and other clients can get the lock,
2724 * then objects which have pages ready to be stuffed into RPCs */
2725 if (!cfs_list_empty(&cli->cl_loi_hp_ready_list))
2726 RETURN(cfs_list_entry(cli->cl_loi_hp_ready_list.next,
2727 struct lov_oinfo, loi_hp_ready_item));
2728 if (!cfs_list_empty(&cli->cl_loi_ready_list))
2729 RETURN(cfs_list_entry(cli->cl_loi_ready_list.next,
2730 struct lov_oinfo, loi_ready_item));
2731 if (!cfs_list_empty(&cli->cl_loi_sync_fs_list))
2732 RETURN(cfs_list_entry(cli->cl_loi_sync_fs_list.next,
2733 struct lov_oinfo, loi_sync_fs_item));
2735 /* then if we have cache waiters, return all objects with queued
2736 * writes. This is especially important when many small files
2737 * have filled up the cache and not been fired into rpcs because
2738 * they don't pass the nr_pending/object threshhold */
2739 if (!cfs_list_empty(&cli->cl_cache_waiters) &&
2740 !cfs_list_empty(&cli->cl_loi_write_list))
2741 RETURN(cfs_list_entry(cli->cl_loi_write_list.next,
2742 struct lov_oinfo, loi_write_item));
2744 /* then return all queued objects when we have an invalid import
2745 * so that they get flushed */
2746 if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2747 if (!cfs_list_empty(&cli->cl_loi_write_list))
2748 RETURN(cfs_list_entry(cli->cl_loi_write_list.next,
2751 if (!cfs_list_empty(&cli->cl_loi_read_list))
2752 RETURN(cfs_list_entry(cli->cl_loi_read_list.next,
2753 struct lov_oinfo, loi_read_item));
2758 static int osc_max_rpc_in_flight(struct client_obd *cli, struct lov_oinfo *loi)
2760 struct osc_async_page *oap;
2763 if (!cfs_list_empty(&loi->loi_write_lop.lop_urgent)) {
2764 oap = cfs_list_entry(loi->loi_write_lop.lop_urgent.next,
2765 struct osc_async_page, oap_urgent_item);
2766 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2769 if (!hprpc && !cfs_list_empty(&loi->loi_read_lop.lop_urgent)) {
2770 oap = cfs_list_entry(loi->loi_read_lop.lop_urgent.next,
2771 struct osc_async_page, oap_urgent_item);
2772 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2775 return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
2778 /* called with the loi list lock held */
2779 void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2781 struct lov_oinfo *loi;
2782 int rc = 0, race_counter = 0;
2785 while ((loi = osc_next_loi(cli)) != NULL) {
2786 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2788 if (osc_max_rpc_in_flight(cli, loi))
2791 /* attempt some read/write balancing by alternating between
2792 * reads and writes in an object. The makes_rpc checks here
2793 * would be redundant if we were getting read/write work items
2794 * instead of objects. we don't want send_oap_rpc to drain a
2795 * partial read pending queue when we're given this object to
2796 * do io on writes while there are cache waiters */
2797 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2798 rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_WRITE,
2799 &loi->loi_write_lop);
2801 CERROR("Write request failed with %d\n", rc);
2803 /* osc_send_oap_rpc failed, mostly because of
2806 * It can't break here, because if:
2807 * - a page was submitted by osc_io_submit, so
2809 * - no request in flight
2810 * - no subsequent request
2811 * The system will be in live-lock state,
2812 * because there is no chance to call
2813 * osc_io_unplug() and osc_check_rpcs() any
2814 * more. pdflush can't help in this case,
2815 * because it might be blocked at grabbing
2816 * the page lock as we mentioned.
2818 * Anyway, continue to drain pages. */
2827 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2828 rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_READ,
2829 &loi->loi_read_lop);
2831 CERROR("Read request failed with %d\n", rc);
2839 /* attempt some inter-object balancing by issuing rpcs
2840 * for each object in turn */
2841 if (!cfs_list_empty(&loi->loi_hp_ready_item))
2842 cfs_list_del_init(&loi->loi_hp_ready_item);
2843 if (!cfs_list_empty(&loi->loi_ready_item))
2844 cfs_list_del_init(&loi->loi_ready_item);
2845 if (!cfs_list_empty(&loi->loi_write_item))
2846 cfs_list_del_init(&loi->loi_write_item);
2847 if (!cfs_list_empty(&loi->loi_read_item))
2848 cfs_list_del_init(&loi->loi_read_item);
2849 if (!cfs_list_empty(&loi->loi_sync_fs_item))
2850 cfs_list_del_init(&loi->loi_sync_fs_item);
2852 loi_list_maint(cli, loi);
2854 /* send_oap_rpc fails with 0 when make_ready tells it to
2855 * back off. llite's make_ready does this when it tries
2856 * to lock a page queued for write that is already locked.
2857 * we want to try sending rpcs from many objects, but we
2858 * don't want to spin failing with 0. */
2859 if (race_counter == 10)
2865 /* we're trying to queue a page in the osc so we're subject to the
2866 * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2867 * If the osc's queued pages are already at that limit, then we want to sleep
2868 * until there is space in the osc's queue for us. We also may be waiting for
2869 * write credits from the OST if there are RPCs in flight that may return some
2870 * before we fall back to sync writes.
2872 * We need this know our allocation was granted in the presence of signals */
2873 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2877 client_obd_list_lock(&cli->cl_loi_list_lock);
2878 rc = cfs_list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2879 client_obd_list_unlock(&cli->cl_loi_list_lock);
2884 * Non-blocking version of osc_enter_cache() that consumes grant only when it
2887 int osc_enter_cache_try(const struct lu_env *env,
2888 struct client_obd *cli, struct lov_oinfo *loi,
2889 struct osc_async_page *oap, int transient)
2893 has_grant = cli->cl_avail_grant >= CFS_PAGE_SIZE;
2895 osc_consume_write_grant(cli, &oap->oap_brw_page);
2897 cli->cl_dirty_transit += CFS_PAGE_SIZE;
2898 cfs_atomic_inc(&obd_dirty_transit_pages);
2899 oap->oap_brw_flags |= OBD_BRW_NOCACHE;
2905 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2906 * grant or cache space. */
2907 static int osc_enter_cache(const struct lu_env *env,
2908 struct client_obd *cli, struct lov_oinfo *loi,
2909 struct osc_async_page *oap)
2911 struct osc_cache_waiter ocw;
2912 struct l_wait_info lwi = { 0 };
2916 CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2917 "grant: %lu\n", cli->cl_dirty, cfs_atomic_read(&obd_dirty_pages),
2918 cli->cl_dirty_max, obd_max_dirty_pages,
2919 cli->cl_lost_grant, cli->cl_avail_grant);
2921 /* force the caller to try sync io. this can jump the list
2922 * of queued writes and create a discontiguous rpc stream */
2923 if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2924 loi->loi_ar.ar_force_sync)
2927 /* Hopefully normal case - cache space and write credits available */
2928 if (cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max &&
2929 cfs_atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages &&
2930 osc_enter_cache_try(env, cli, loi, oap, 0))
2933 /* It is safe to block as a cache waiter as long as there is grant
2934 * space available or the hope of additional grant being returned
2935 * when an in flight write completes. Using the write back cache
2936 * if possible is preferable to sending the data synchronously
2937 * because write pages can then be merged in to large requests.
2938 * The addition of this cache waiter will causing pending write
2939 * pages to be sent immediately. */
2940 if (cli->cl_w_in_flight || cli->cl_avail_grant >= CFS_PAGE_SIZE) {
2941 cfs_list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2942 cfs_waitq_init(&ocw.ocw_waitq);
2946 loi_list_maint(cli, loi);
2947 osc_check_rpcs(env, cli);
2948 client_obd_list_unlock(&cli->cl_loi_list_lock);
2950 CDEBUG(D_CACHE, "sleeping for cache space\n");
2951 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2953 client_obd_list_lock(&cli->cl_loi_list_lock);
2954 if (!cfs_list_empty(&ocw.ocw_entry)) {
2955 cfs_list_del(&ocw.ocw_entry);
2965 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2966 struct lov_oinfo *loi, cfs_page_t *page,
2967 obd_off offset, const struct obd_async_page_ops *ops,
2968 void *data, void **res, int nocache,
2969 struct lustre_handle *lockh)
2971 struct osc_async_page *oap;
2976 return cfs_size_round(sizeof(*oap));
2979 oap->oap_magic = OAP_MAGIC;
2980 oap->oap_cli = &exp->exp_obd->u.cli;
2983 oap->oap_caller_ops = ops;
2984 oap->oap_caller_data = data;
2986 oap->oap_page = page;
2987 oap->oap_obj_off = offset;
2988 if (!client_is_remote(exp) &&
2989 cfs_capable(CFS_CAP_SYS_RESOURCE))
2990 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2992 LASSERT(!(offset & ~CFS_PAGE_MASK));
2994 CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2995 CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2996 CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2997 CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2999 cfs_spin_lock_init(&oap->oap_lock);
3000 CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
3004 struct osc_async_page *oap_from_cookie(void *cookie)
3006 struct osc_async_page *oap = cookie;
3007 if (oap->oap_magic != OAP_MAGIC)
3008 return ERR_PTR(-EINVAL);
3012 int osc_queue_async_io(const struct lu_env *env,
3013 struct obd_export *exp, struct lov_stripe_md *lsm,
3014 struct lov_oinfo *loi, void *cookie,
3015 int cmd, obd_off off, int count,
3016 obd_flag brw_flags, enum async_flags async_flags)
3018 struct client_obd *cli = &exp->exp_obd->u.cli;
3019 struct osc_async_page *oap;
3023 oap = oap_from_cookie(cookie);
3025 RETURN(PTR_ERR(oap));
3027 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
3030 if (!cfs_list_empty(&oap->oap_pending_item) ||
3031 !cfs_list_empty(&oap->oap_urgent_item) ||
3032 !cfs_list_empty(&oap->oap_rpc_item))
3035 /* check if the file's owner/group is over quota */
3036 if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)) {
3037 struct cl_object *obj;
3038 struct cl_attr attr; /* XXX put attr into thread info */
3039 unsigned int qid[MAXQUOTAS];
3041 obj = cl_object_top(osc_oap2cl_page(oap)->cp_obj);
3043 cl_object_attr_lock(obj);
3044 rc = cl_object_attr_get(env, obj, &attr);
3045 cl_object_attr_unlock(obj);
3047 qid[USRQUOTA] = attr.cat_uid;
3048 qid[GRPQUOTA] = attr.cat_gid;
3050 lquota_chkdq(quota_interface, cli, qid) == NO_QUOTA)
3057 loi = lsm->lsm_oinfo[0];
3059 client_obd_list_lock(&cli->cl_loi_list_lock);
3061 LASSERT(off + count <= CFS_PAGE_SIZE);
3063 oap->oap_page_off = off;
3064 oap->oap_count = count;
3065 oap->oap_brw_flags = brw_flags;
3066 /* Give a hint to OST that requests are coming from kswapd - bug19529 */
3067 if (cfs_memory_pressure_get())
3068 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
3069 cfs_spin_lock(&oap->oap_lock);
3070 oap->oap_async_flags = async_flags;
3071 cfs_spin_unlock(&oap->oap_lock);
3073 if (cmd & OBD_BRW_WRITE) {
3074 rc = osc_enter_cache(env, cli, loi, oap);
3076 client_obd_list_unlock(&cli->cl_loi_list_lock);
3081 osc_oap_to_pending(oap);
3082 loi_list_maint(cli, loi);
3084 LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
3087 osc_check_rpcs(env, cli);
3088 client_obd_list_unlock(&cli->cl_loi_list_lock);
3093 /* aka (~was & now & flag), but this is more clear :) */
3094 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
3096 int osc_set_async_flags_base(struct client_obd *cli,
3097 struct lov_oinfo *loi, struct osc_async_page *oap,
3098 obd_flag async_flags)
3100 struct loi_oap_pages *lop;
3104 LASSERT(!cfs_list_empty(&oap->oap_pending_item));
3106 if (oap->oap_cmd & OBD_BRW_WRITE) {
3107 lop = &loi->loi_write_lop;
3109 lop = &loi->loi_read_lop;
3112 if ((oap->oap_async_flags & async_flags) == async_flags)
3115 /* XXX: This introduces a tiny insignificant race for the case if this
3116 * loi already had other urgent items.
3118 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_SYNCFS) &&
3119 cfs_list_empty(&oap->oap_rpc_item) &&
3120 cfs_list_empty(&oap->oap_urgent_item)) {
3121 osc_add_to_lop_urgent(lop, oap, ASYNC_SYNCFS);
3122 flags |= ASYNC_SYNCFS;
3123 cfs_spin_lock(&oap->oap_lock);
3124 oap->oap_async_flags |= flags;
3125 cfs_spin_unlock(&oap->oap_lock);
3126 loi_list_maint(cli, loi);
3130 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
3131 flags |= ASYNC_READY;
3133 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT) &&
3134 cfs_list_empty(&oap->oap_rpc_item)) {
3135 osc_add_to_lop_urgent(lop, oap, ASYNC_URGENT);
3136 flags |= ASYNC_URGENT;
3137 loi_list_maint(cli, loi);
3139 cfs_spin_lock(&oap->oap_lock);
3140 oap->oap_async_flags |= flags;
3141 cfs_spin_unlock(&oap->oap_lock);
3143 LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
3144 oap->oap_async_flags);
3148 int osc_teardown_async_page(struct obd_export *exp,
3149 struct lov_stripe_md *lsm,
3150 struct lov_oinfo *loi, void *cookie)
3152 struct client_obd *cli = &exp->exp_obd->u.cli;
3153 struct loi_oap_pages *lop;
3154 struct osc_async_page *oap;
3158 oap = oap_from_cookie(cookie);
3160 RETURN(PTR_ERR(oap));
3163 loi = lsm->lsm_oinfo[0];
3165 if (oap->oap_cmd & OBD_BRW_WRITE) {
3166 lop = &loi->loi_write_lop;
3168 lop = &loi->loi_read_lop;
3171 client_obd_list_lock(&cli->cl_loi_list_lock);
3173 if (!cfs_list_empty(&oap->oap_rpc_item))
3174 GOTO(out, rc = -EBUSY);
3176 osc_exit_cache(cli, oap, 0);
3177 osc_wake_cache_waiters(cli);
3179 if (!cfs_list_empty(&oap->oap_urgent_item)) {
3180 cfs_list_del_init(&oap->oap_urgent_item);
3181 cfs_spin_lock(&oap->oap_lock);
3182 oap->oap_async_flags &= ~(ASYNC_URGENT | ASYNC_HP |
3184 cfs_spin_unlock(&oap->oap_lock);
3186 if (!cfs_list_empty(&oap->oap_pending_item)) {
3187 cfs_list_del_init(&oap->oap_pending_item);
3188 lop_update_pending(cli, lop, oap->oap_cmd, -1);
3190 loi_list_maint(cli, loi);
3191 LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
3193 client_obd_list_unlock(&cli->cl_loi_list_lock);
3197 static void osc_set_lock_data_with_check(struct ldlm_lock *lock,
3198 struct ldlm_enqueue_info *einfo,
3201 void *data = einfo->ei_cbdata;
3203 LASSERT(lock != NULL);
3204 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
3205 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
3206 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
3207 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
3209 lock_res_and_lock(lock);
3210 cfs_spin_lock(&osc_ast_guard);
3211 LASSERT(lock->l_ast_data == NULL || lock->l_ast_data == data);
3212 lock->l_ast_data = data;
3213 cfs_spin_unlock(&osc_ast_guard);
3214 unlock_res_and_lock(lock);
3217 static void osc_set_data_with_check(struct lustre_handle *lockh,
3218 struct ldlm_enqueue_info *einfo,
3221 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
3224 osc_set_lock_data_with_check(lock, einfo, flags);
3225 LDLM_LOCK_PUT(lock);
3227 CERROR("lockh %p, data %p - client evicted?\n",
3228 lockh, einfo->ei_cbdata);
3231 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3232 ldlm_iterator_t replace, void *data)
3234 struct ldlm_res_id res_id;
3235 struct obd_device *obd = class_exp2obd(exp);
3237 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
3238 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3242 /* find any ldlm lock of the inode in osc
3246 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3247 ldlm_iterator_t replace, void *data)
3249 struct ldlm_res_id res_id;
3250 struct obd_device *obd = class_exp2obd(exp);
3253 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
3254 rc = ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3255 if (rc == LDLM_ITER_STOP)
3257 if (rc == LDLM_ITER_CONTINUE)
3262 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
3263 obd_enqueue_update_f upcall, void *cookie,
3266 int intent = *flags & LDLM_FL_HAS_INTENT;
3270 /* The request was created before ldlm_cli_enqueue call. */
3271 if (rc == ELDLM_LOCK_ABORTED) {
3272 struct ldlm_reply *rep;
3273 rep = req_capsule_server_get(&req->rq_pill,
3276 LASSERT(rep != NULL);
3277 if (rep->lock_policy_res1)
3278 rc = rep->lock_policy_res1;
3282 if ((intent && rc == ELDLM_LOCK_ABORTED) || !rc) {
3283 *flags |= LDLM_FL_LVB_READY;
3284 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
3285 lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
3288 /* Call the update callback. */
3289 rc = (*upcall)(cookie, rc);
3293 static int osc_enqueue_interpret(const struct lu_env *env,
3294 struct ptlrpc_request *req,
3295 struct osc_enqueue_args *aa, int rc)
3297 struct ldlm_lock *lock;
3298 struct lustre_handle handle;
3301 /* Make a local copy of a lock handle and a mode, because aa->oa_*
3302 * might be freed anytime after lock upcall has been called. */
3303 lustre_handle_copy(&handle, aa->oa_lockh);
3304 mode = aa->oa_ei->ei_mode;
3306 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
3308 lock = ldlm_handle2lock(&handle);
3310 /* Take an additional reference so that a blocking AST that
3311 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
3312 * to arrive after an upcall has been executed by
3313 * osc_enqueue_fini(). */
3314 ldlm_lock_addref(&handle, mode);
3316 /* Let CP AST to grant the lock first. */
3317 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
3319 /* Complete obtaining the lock procedure. */
3320 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
3321 mode, aa->oa_flags, aa->oa_lvb,
3322 sizeof(*aa->oa_lvb), &handle, rc);
3323 /* Complete osc stuff. */
3324 rc = osc_enqueue_fini(req, aa->oa_lvb,
3325 aa->oa_upcall, aa->oa_cookie, aa->oa_flags, rc);
3327 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
3329 /* Release the lock for async request. */
3330 if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
3332 * Releases a reference taken by ldlm_cli_enqueue(), if it is
3333 * not already released by
3334 * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
3336 ldlm_lock_decref(&handle, mode);
3338 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
3339 aa->oa_lockh, req, aa);
3340 ldlm_lock_decref(&handle, mode);
3341 LDLM_LOCK_PUT(lock);
3345 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
3346 struct lov_oinfo *loi, int flags,
3347 struct ost_lvb *lvb, __u32 mode, int rc)
3349 if (rc == ELDLM_OK) {
3350 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
3353 LASSERT(lock != NULL);
3354 loi->loi_lvb = *lvb;
3355 tmp = loi->loi_lvb.lvb_size;
3356 /* Extend KMS up to the end of this lock and no further
3357 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
3358 if (tmp > lock->l_policy_data.l_extent.end)
3359 tmp = lock->l_policy_data.l_extent.end + 1;
3360 if (tmp >= loi->loi_kms) {
3361 LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
3362 ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
3363 loi_kms_set(loi, tmp);
3365 LDLM_DEBUG(lock, "lock acquired, setting rss="
3366 LPU64"; leaving kms="LPU64", end="LPU64,
3367 loi->loi_lvb.lvb_size, loi->loi_kms,
3368 lock->l_policy_data.l_extent.end);
3370 ldlm_lock_allow_match(lock);
3371 LDLM_LOCK_PUT(lock);
3372 } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
3373 loi->loi_lvb = *lvb;
3374 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
3375 " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
3379 EXPORT_SYMBOL(osc_update_enqueue);
3381 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
3383 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
3384 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
3385 * other synchronous requests, however keeping some locks and trying to obtain
3386 * others may take a considerable amount of time in a case of ost failure; and
3387 * when other sync requests do not get released lock from a client, the client
3388 * is excluded from the cluster -- such scenarious make the life difficult, so
3389 * release locks just after they are obtained. */
3390 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3391 int *flags, ldlm_policy_data_t *policy,
3392 struct ost_lvb *lvb, int kms_valid,
3393 obd_enqueue_update_f upcall, void *cookie,
3394 struct ldlm_enqueue_info *einfo,
3395 struct lustre_handle *lockh,
3396 struct ptlrpc_request_set *rqset, int async)
3398 struct obd_device *obd = exp->exp_obd;
3399 struct ptlrpc_request *req = NULL;
3400 int intent = *flags & LDLM_FL_HAS_INTENT;
3405 /* Filesystem lock extents are extended to page boundaries so that
3406 * dealing with the page cache is a little smoother. */
3407 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3408 policy->l_extent.end |= ~CFS_PAGE_MASK;
3411 * kms is not valid when either object is completely fresh (so that no
3412 * locks are cached), or object was evicted. In the latter case cached
3413 * lock cannot be used, because it would prime inode state with
3414 * potentially stale LVB.
3419 /* Next, search for already existing extent locks that will cover us */
3420 /* If we're trying to read, we also search for an existing PW lock. The
3421 * VFS and page cache already protect us locally, so lots of readers/
3422 * writers can share a single PW lock.
3424 * There are problems with conversion deadlocks, so instead of
3425 * converting a read lock to a write lock, we'll just enqueue a new
3428 * At some point we should cancel the read lock instead of making them
3429 * send us a blocking callback, but there are problems with canceling
3430 * locks out from other users right now, too. */
3431 mode = einfo->ei_mode;
3432 if (einfo->ei_mode == LCK_PR)
3434 mode = ldlm_lock_match(obd->obd_namespace,
3435 *flags | LDLM_FL_LVB_READY, res_id,
3436 einfo->ei_type, policy, mode, lockh, 0);
3438 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
3440 if (matched->l_ast_data == NULL ||
3441 matched->l_ast_data == einfo->ei_cbdata) {
3442 /* addref the lock only if not async requests and PW
3443 * lock is matched whereas we asked for PR. */
3444 if (!rqset && einfo->ei_mode != mode)
3445 ldlm_lock_addref(lockh, LCK_PR);
3446 osc_set_lock_data_with_check(matched, einfo, *flags);
3448 /* I would like to be able to ASSERT here that
3449 * rss <= kms, but I can't, for reasons which
3450 * are explained in lov_enqueue() */
3453 /* We already have a lock, and it's referenced */
3454 (*upcall)(cookie, ELDLM_OK);
3456 /* For async requests, decref the lock. */
3457 if (einfo->ei_mode != mode)
3458 ldlm_lock_decref(lockh, LCK_PW);
3460 ldlm_lock_decref(lockh, einfo->ei_mode);
3461 LDLM_LOCK_PUT(matched);
3464 ldlm_lock_decref(lockh, mode);
3465 LDLM_LOCK_PUT(matched);
3470 CFS_LIST_HEAD(cancels);
3471 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3472 &RQF_LDLM_ENQUEUE_LVB);
3476 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
3478 ptlrpc_request_free(req);
3482 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
3484 ptlrpc_request_set_replen(req);
3487 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
3488 *flags &= ~LDLM_FL_BLOCK_GRANTED;
3490 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
3491 sizeof(*lvb), lockh, async);
3494 struct osc_enqueue_args *aa;
3495 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3496 aa = ptlrpc_req_async_args(req);
3499 aa->oa_flags = flags;
3500 aa->oa_upcall = upcall;
3501 aa->oa_cookie = cookie;
3503 aa->oa_lockh = lockh;
3505 req->rq_interpret_reply =
3506 (ptlrpc_interpterer_t)osc_enqueue_interpret;
3507 if (rqset == PTLRPCD_SET)
3508 ptlrpcd_add_req(req, PSCOPE_OTHER);
3510 ptlrpc_set_add_req(rqset, req);
3511 } else if (intent) {
3512 ptlrpc_req_finished(req);
3517 rc = osc_enqueue_fini(req, lvb, upcall, cookie, flags, rc);
3519 ptlrpc_req_finished(req);
3524 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
3525 struct ldlm_enqueue_info *einfo,
3526 struct ptlrpc_request_set *rqset)
3528 struct ldlm_res_id res_id;
3532 osc_build_res_name(oinfo->oi_md->lsm_object_id,
3533 oinfo->oi_md->lsm_object_seq, &res_id);
3535 rc = osc_enqueue_base(exp, &res_id, &oinfo->oi_flags, &oinfo->oi_policy,
3536 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
3537 oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid,
3538 oinfo->oi_cb_up, oinfo, einfo, oinfo->oi_lockh,
3539 rqset, rqset != NULL);
3543 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3544 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
3545 int *flags, void *data, struct lustre_handle *lockh,
3548 struct obd_device *obd = exp->exp_obd;
3549 int lflags = *flags;
3553 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
3556 /* Filesystem lock extents are extended to page boundaries so that
3557 * dealing with the page cache is a little smoother */
3558 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3559 policy->l_extent.end |= ~CFS_PAGE_MASK;
3561 /* Next, search for already existing extent locks that will cover us */
3562 /* If we're trying to read, we also search for an existing PW lock. The
3563 * VFS and page cache already protect us locally, so lots of readers/
3564 * writers can share a single PW lock. */
3568 rc = ldlm_lock_match(obd->obd_namespace, lflags,
3569 res_id, type, policy, rc, lockh, unref);
3572 osc_set_data_with_check(lockh, data, lflags);
3573 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
3574 ldlm_lock_addref(lockh, LCK_PR);
3575 ldlm_lock_decref(lockh, LCK_PW);
3582 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
3586 if (unlikely(mode == LCK_GROUP))
3587 ldlm_lock_decref_and_cancel(lockh, mode);
3589 ldlm_lock_decref(lockh, mode);
3594 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
3595 __u32 mode, struct lustre_handle *lockh)
3598 RETURN(osc_cancel_base(lockh, mode));
3601 static int osc_cancel_unused(struct obd_export *exp,
3602 struct lov_stripe_md *lsm,
3603 ldlm_cancel_flags_t flags,
3606 struct obd_device *obd = class_exp2obd(exp);
3607 struct ldlm_res_id res_id, *resp = NULL;
3610 resp = osc_build_res_name(lsm->lsm_object_id,
3611 lsm->lsm_object_seq, &res_id);
3614 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
3617 static int osc_statfs_interpret(const struct lu_env *env,
3618 struct ptlrpc_request *req,
3619 struct osc_async_args *aa, int rc)
3621 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
3622 struct obd_statfs *msfs;
3627 /* The request has in fact never been sent
3628 * due to issues at a higher level (LOV).
3629 * Exit immediately since the caller is
3630 * aware of the problem and takes care
3631 * of the clean up */
3634 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
3635 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
3641 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3643 GOTO(out, rc = -EPROTO);
3646 /* Reinitialize the RDONLY and DEGRADED flags at the client
3647 * on each statfs, so they don't stay set permanently. */
3648 cfs_spin_lock(&cli->cl_oscc.oscc_lock);
3650 if (unlikely(msfs->os_state & OS_STATE_DEGRADED))
3651 cli->cl_oscc.oscc_flags |= OSCC_FLAG_DEGRADED;
3652 else if (unlikely(cli->cl_oscc.oscc_flags & OSCC_FLAG_DEGRADED))
3653 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_DEGRADED;
3655 if (unlikely(msfs->os_state & OS_STATE_READONLY))
3656 cli->cl_oscc.oscc_flags |= OSCC_FLAG_RDONLY;
3657 else if (unlikely(cli->cl_oscc.oscc_flags & OSCC_FLAG_RDONLY))
3658 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_RDONLY;
3660 /* Add a bit of hysteresis so this flag isn't continually flapping,
3661 * and ensure that new files don't get extremely fragmented due to
3662 * only a small amount of available space in the filesystem.
3663 * We want to set the NOSPC flag when there is less than ~0.1% free
3664 * and clear it when there is at least ~0.2% free space, so:
3665 * avail < ~0.1% max max = avail + used
3666 * 1025 * avail < avail + used used = blocks - free
3667 * 1024 * avail < used
3668 * 1024 * avail < blocks - free
3669 * avail < ((blocks - free) >> 10)
3671 * On very large disk, say 16TB 0.1% will be 16 GB. We don't want to
3672 * lose that amount of space so in those cases we report no space left
3673 * if their is less than 1 GB left. */
3674 used = min_t(__u64,(msfs->os_blocks - msfs->os_bfree) >> 10, 1 << 30);
3675 if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) == 0) &&
3676 ((msfs->os_ffree < 32) || (msfs->os_bavail < used))))
3677 cli->cl_oscc.oscc_flags |= OSCC_FLAG_NOSPC;
3678 else if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) != 0) &&
3679 (msfs->os_ffree > 64) && (msfs->os_bavail > (used << 1))))
3680 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_NOSPC;
3682 cfs_spin_unlock(&cli->cl_oscc.oscc_lock);
3684 *aa->aa_oi->oi_osfs = *msfs;
3686 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
3690 static int osc_statfs_async(struct obd_device *obd, struct obd_info *oinfo,
3691 __u64 max_age, struct ptlrpc_request_set *rqset)
3693 struct ptlrpc_request *req;
3694 struct osc_async_args *aa;
3698 /* We could possibly pass max_age in the request (as an absolute
3699 * timestamp or a "seconds.usec ago") so the target can avoid doing
3700 * extra calls into the filesystem if that isn't necessary (e.g.
3701 * during mount that would help a bit). Having relative timestamps
3702 * is not so great if request processing is slow, while absolute
3703 * timestamps are not ideal because they need time synchronization. */
3704 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
3708 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3710 ptlrpc_request_free(req);
3713 ptlrpc_request_set_replen(req);
3714 req->rq_request_portal = OST_CREATE_PORTAL;
3715 ptlrpc_at_set_req_timeout(req);
3717 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
3718 /* procfs requests not want stat in wait for avoid deadlock */
3719 req->rq_no_resend = 1;
3720 req->rq_no_delay = 1;
3723 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
3724 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3725 aa = ptlrpc_req_async_args(req);
3728 ptlrpc_set_add_req(rqset, req);
3732 static int osc_statfs(struct obd_device *obd, struct obd_statfs *osfs,
3733 __u64 max_age, __u32 flags)
3735 struct obd_statfs *msfs;
3736 struct ptlrpc_request *req;
3737 struct obd_import *imp = NULL;
3741 /*Since the request might also come from lprocfs, so we need
3742 *sync this with client_disconnect_export Bug15684*/
3743 cfs_down_read(&obd->u.cli.cl_sem);
3744 if (obd->u.cli.cl_import)
3745 imp = class_import_get(obd->u.cli.cl_import);
3746 cfs_up_read(&obd->u.cli.cl_sem);
3750 /* We could possibly pass max_age in the request (as an absolute
3751 * timestamp or a "seconds.usec ago") so the target can avoid doing
3752 * extra calls into the filesystem if that isn't necessary (e.g.
3753 * during mount that would help a bit). Having relative timestamps
3754 * is not so great if request processing is slow, while absolute
3755 * timestamps are not ideal because they need time synchronization. */
3756 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
3758 class_import_put(imp);
3763 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3765 ptlrpc_request_free(req);
3768 ptlrpc_request_set_replen(req);
3769 req->rq_request_portal = OST_CREATE_PORTAL;
3770 ptlrpc_at_set_req_timeout(req);
3772 if (flags & OBD_STATFS_NODELAY) {
3773 /* procfs requests not want stat in wait for avoid deadlock */
3774 req->rq_no_resend = 1;
3775 req->rq_no_delay = 1;
3778 rc = ptlrpc_queue_wait(req);
3782 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3784 GOTO(out, rc = -EPROTO);
3791 ptlrpc_req_finished(req);
3795 /* Retrieve object striping information.
3797 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
3798 * the maximum number of OST indices which will fit in the user buffer.
3799 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
3801 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
3803 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
3804 struct lov_user_md_v3 lum, *lumk;
3805 struct lov_user_ost_data_v1 *lmm_objects;
3806 int rc = 0, lum_size;
3812 /* we only need the header part from user space to get lmm_magic and
3813 * lmm_stripe_count, (the header part is common to v1 and v3) */
3814 lum_size = sizeof(struct lov_user_md_v1);
3815 if (cfs_copy_from_user(&lum, lump, lum_size))
3818 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
3819 (lum.lmm_magic != LOV_USER_MAGIC_V3))
3822 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
3823 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
3824 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
3825 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
3827 /* we can use lov_mds_md_size() to compute lum_size
3828 * because lov_user_md_vX and lov_mds_md_vX have the same size */
3829 if (lum.lmm_stripe_count > 0) {
3830 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
3831 OBD_ALLOC(lumk, lum_size);
3835 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
3836 lmm_objects = &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
3838 lmm_objects = &(lumk->lmm_objects[0]);
3839 lmm_objects->l_object_id = lsm->lsm_object_id;
3841 lum_size = lov_mds_md_size(0, lum.lmm_magic);
3845 lumk->lmm_object_id = lsm->lsm_object_id;
3846 lumk->lmm_object_seq = lsm->lsm_object_seq;
3847 lumk->lmm_stripe_count = 1;
3849 if (cfs_copy_to_user(lump, lumk, lum_size))
3853 OBD_FREE(lumk, lum_size);
3859 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
3860 void *karg, void *uarg)
3862 struct obd_device *obd = exp->exp_obd;
3863 struct obd_ioctl_data *data = karg;
3867 if (!cfs_try_module_get(THIS_MODULE)) {
3868 CERROR("Can't get module. Is it alive?");
3872 case OBD_IOC_LOV_GET_CONFIG: {
3874 struct lov_desc *desc;
3875 struct obd_uuid uuid;
3879 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
3880 GOTO(out, err = -EINVAL);
3882 data = (struct obd_ioctl_data *)buf;
3884 if (sizeof(*desc) > data->ioc_inllen1) {
3885 obd_ioctl_freedata(buf, len);
3886 GOTO(out, err = -EINVAL);
3889 if (data->ioc_inllen2 < sizeof(uuid)) {
3890 obd_ioctl_freedata(buf, len);
3891 GOTO(out, err = -EINVAL);
3894 desc = (struct lov_desc *)data->ioc_inlbuf1;
3895 desc->ld_tgt_count = 1;
3896 desc->ld_active_tgt_count = 1;
3897 desc->ld_default_stripe_count = 1;
3898 desc->ld_default_stripe_size = 0;
3899 desc->ld_default_stripe_offset = 0;
3900 desc->ld_pattern = 0;
3901 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
3903 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
3905 err = cfs_copy_to_user((void *)uarg, buf, len);
3908 obd_ioctl_freedata(buf, len);
3911 case LL_IOC_LOV_SETSTRIPE:
3912 err = obd_alloc_memmd(exp, karg);
3916 case LL_IOC_LOV_GETSTRIPE:
3917 err = osc_getstripe(karg, uarg);
3919 case OBD_IOC_CLIENT_RECOVER:
3920 err = ptlrpc_recover_import(obd->u.cli.cl_import,
3925 case IOC_OSC_SET_ACTIVE:
3926 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
3929 case OBD_IOC_POLL_QUOTACHECK:
3930 err = lquota_poll_check(quota_interface, exp,
3931 (struct if_quotacheck *)karg);
3933 case OBD_IOC_PING_TARGET:
3934 err = ptlrpc_obd_ping(obd);
3937 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
3938 cmd, cfs_curproc_comm());
3939 GOTO(out, err = -ENOTTY);
3942 cfs_module_put(THIS_MODULE);
3946 static int osc_get_info(struct obd_export *exp, obd_count keylen,
3947 void *key, __u32 *vallen, void *val,
3948 struct lov_stripe_md *lsm)
3951 if (!vallen || !val)
3954 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
3955 __u32 *stripe = val;
3956 *vallen = sizeof(*stripe);
3959 } else if (KEY_IS(KEY_LAST_ID)) {
3960 struct ptlrpc_request *req;
3965 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3966 &RQF_OST_GET_INFO_LAST_ID);
3970 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3971 RCL_CLIENT, keylen);
3972 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3974 ptlrpc_request_free(req);
3978 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3979 memcpy(tmp, key, keylen);
3981 req->rq_no_delay = req->rq_no_resend = 1;
3982 ptlrpc_request_set_replen(req);
3983 rc = ptlrpc_queue_wait(req);
3987 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
3989 GOTO(out, rc = -EPROTO);
3991 *((obd_id *)val) = *reply;
3993 ptlrpc_req_finished(req);
3995 } else if (KEY_IS(KEY_FIEMAP)) {
3996 struct ptlrpc_request *req;
3997 struct ll_user_fiemap *reply;
4001 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
4002 &RQF_OST_GET_INFO_FIEMAP);
4006 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
4007 RCL_CLIENT, keylen);
4008 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
4009 RCL_CLIENT, *vallen);
4010 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
4011 RCL_SERVER, *vallen);
4013 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
4015 ptlrpc_request_free(req);
4019 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
4020 memcpy(tmp, key, keylen);
4021 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
4022 memcpy(tmp, val, *vallen);
4024 ptlrpc_request_set_replen(req);
4025 rc = ptlrpc_queue_wait(req);
4029 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
4031 GOTO(out1, rc = -EPROTO);
4033 memcpy(val, reply, *vallen);
4035 ptlrpc_req_finished(req);
4043 static int osc_setinfo_mds_connect_import(struct obd_import *imp)
4045 struct llog_ctxt *ctxt;
4049 ctxt = llog_get_context(imp->imp_obd, LLOG_MDS_OST_ORIG_CTXT);
4051 rc = llog_initiator_connect(ctxt);
4052 llog_ctxt_put(ctxt);
4054 /* XXX return an error? skip setting below flags? */
4057 cfs_spin_lock(&imp->imp_lock);
4058 imp->imp_server_timeout = 1;
4059 imp->imp_pingable = 1;
4060 cfs_spin_unlock(&imp->imp_lock);
4061 CDEBUG(D_RPCTRACE, "pinging OST %s\n", obd2cli_tgt(imp->imp_obd));
4066 static int osc_setinfo_mds_conn_interpret(const struct lu_env *env,
4067 struct ptlrpc_request *req,
4074 RETURN(osc_setinfo_mds_connect_import(req->rq_import));
4077 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
4078 void *key, obd_count vallen, void *val,
4079 struct ptlrpc_request_set *set)
4081 struct ptlrpc_request *req;
4082 struct obd_device *obd = exp->exp_obd;
4083 struct obd_import *imp = class_exp2cliimp(exp);
4088 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
4090 if (KEY_IS(KEY_NEXT_ID)) {
4092 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4094 if (vallen != sizeof(obd_id))
4099 if (vallen != sizeof(obd_id))
4102 /* avoid race between allocate new object and set next id
4103 * from ll_sync thread */
4104 cfs_spin_lock(&oscc->oscc_lock);
4105 new_val = *((obd_id*)val) + 1;
4106 if (new_val > oscc->oscc_next_id)
4107 oscc->oscc_next_id = new_val;
4108 cfs_spin_unlock(&oscc->oscc_lock);
4109 CDEBUG(D_HA, "%s: set oscc_next_id = "LPU64"\n",
4110 exp->exp_obd->obd_name,
4111 obd->u.cli.cl_oscc.oscc_next_id);
4116 if (KEY_IS(KEY_CHECKSUM)) {
4117 if (vallen != sizeof(int))
4119 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
4123 if (KEY_IS(KEY_SPTLRPC_CONF)) {
4124 sptlrpc_conf_client_adapt(obd);
4128 if (KEY_IS(KEY_FLUSH_CTX)) {
4129 sptlrpc_import_flush_my_ctx(imp);
4133 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
4136 /* We pass all other commands directly to OST. Since nobody calls osc
4137 methods directly and everybody is supposed to go through LOV, we
4138 assume lov checked invalid values for us.
4139 The only recognised values so far are evict_by_nid and mds_conn.
4140 Even if something bad goes through, we'd get a -EINVAL from OST
4143 if (KEY_IS(KEY_GRANT_SHRINK))
4144 req = ptlrpc_request_alloc(imp, &RQF_OST_SET_GRANT_INFO);
4146 req = ptlrpc_request_alloc(imp, &RQF_OBD_SET_INFO);
4151 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
4152 RCL_CLIENT, keylen);
4153 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
4154 RCL_CLIENT, vallen);
4155 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
4157 ptlrpc_request_free(req);
4161 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
4162 memcpy(tmp, key, keylen);
4163 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_VAL);
4164 memcpy(tmp, val, vallen);
4166 if (KEY_IS(KEY_MDS_CONN)) {
4167 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4169 oscc->oscc_oa.o_seq = (*(__u32 *)val);
4170 oscc->oscc_oa.o_valid |= OBD_MD_FLGROUP;
4171 LASSERT_SEQ_IS_MDT(oscc->oscc_oa.o_seq);
4172 req->rq_no_delay = req->rq_no_resend = 1;
4173 req->rq_interpret_reply = osc_setinfo_mds_conn_interpret;
4174 } else if (KEY_IS(KEY_GRANT_SHRINK)) {
4175 struct osc_grant_args *aa;
4178 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
4179 aa = ptlrpc_req_async_args(req);
4182 ptlrpc_req_finished(req);
4185 *oa = ((struct ost_body *)val)->oa;
4187 req->rq_interpret_reply = osc_shrink_grant_interpret;
4190 ptlrpc_request_set_replen(req);
4191 if (!KEY_IS(KEY_GRANT_SHRINK)) {
4192 LASSERT(set != NULL);
4193 ptlrpc_set_add_req(set, req);
4194 ptlrpc_check_set(NULL, set);
4196 ptlrpcd_add_req(req, PSCOPE_OTHER);
4202 static struct llog_operations osc_size_repl_logops = {
4203 lop_cancel: llog_obd_repl_cancel
4206 static struct llog_operations osc_mds_ost_orig_logops;
4208 static int __osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
4209 struct obd_device *tgt, struct llog_catid *catid)
4214 rc = llog_setup(obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT, tgt, 1,
4215 &catid->lci_logid, &osc_mds_ost_orig_logops);
4217 CERROR("failed LLOG_MDS_OST_ORIG_CTXT\n");
4221 rc = llog_setup(obd, &obd->obd_olg, LLOG_SIZE_REPL_CTXT, tgt, 1,
4222 NULL, &osc_size_repl_logops);
4224 struct llog_ctxt *ctxt =
4225 llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
4228 CERROR("failed LLOG_SIZE_REPL_CTXT\n");
4233 CERROR("osc '%s' tgt '%s' catid %p rc=%d\n",
4234 obd->obd_name, tgt->obd_name, catid, rc);
4235 CERROR("logid "LPX64":0x%x\n",
4236 catid->lci_logid.lgl_oid, catid->lci_logid.lgl_ogen);
4241 static int osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
4242 struct obd_device *disk_obd, int *index)
4244 struct llog_catid catid;
4245 static char name[32] = CATLIST;
4249 LASSERT(olg == &obd->obd_olg);
4251 cfs_mutex_down(&olg->olg_cat_processing);
4252 rc = llog_get_cat_list(disk_obd, name, *index, 1, &catid);
4254 CERROR("rc: %d\n", rc);
4258 CDEBUG(D_INFO, "%s: Init llog for %d - catid "LPX64"/"LPX64":%x\n",
4259 obd->obd_name, *index, catid.lci_logid.lgl_oid,
4260 catid.lci_logid.lgl_oseq, catid.lci_logid.lgl_ogen);
4262 rc = __osc_llog_init(obd, olg, disk_obd, &catid);
4264 CERROR("rc: %d\n", rc);
4268 rc = llog_put_cat_list(disk_obd, name, *index, 1, &catid);
4270 CERROR("rc: %d\n", rc);
4275 cfs_mutex_up(&olg->olg_cat_processing);
4280 static int osc_llog_finish(struct obd_device *obd, int count)
4282 struct llog_ctxt *ctxt;
4283 int rc = 0, rc2 = 0;
4286 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
4288 rc = llog_cleanup(ctxt);
4290 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
4292 rc2 = llog_cleanup(ctxt);
4299 static int osc_reconnect(const struct lu_env *env,
4300 struct obd_export *exp, struct obd_device *obd,
4301 struct obd_uuid *cluuid,
4302 struct obd_connect_data *data,
4305 struct client_obd *cli = &obd->u.cli;
4307 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
4310 client_obd_list_lock(&cli->cl_loi_list_lock);
4311 data->ocd_grant = (cli->cl_avail_grant + cli->cl_dirty) ?:
4312 2 * cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
4313 lost_grant = cli->cl_lost_grant;
4314 cli->cl_lost_grant = 0;
4315 client_obd_list_unlock(&cli->cl_loi_list_lock);
4317 CDEBUG(D_CACHE, "request ocd_grant: %d cl_avail_grant: %ld "
4318 "cl_dirty: %ld cl_lost_grant: %ld\n", data->ocd_grant,
4319 cli->cl_avail_grant, cli->cl_dirty, lost_grant);
4320 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
4321 " ocd_grant: %d\n", data->ocd_connect_flags,
4322 data->ocd_version, data->ocd_grant);
4328 static int osc_disconnect(struct obd_export *exp)
4330 struct obd_device *obd = class_exp2obd(exp);
4331 struct llog_ctxt *ctxt;
4334 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
4336 if (obd->u.cli.cl_conn_count == 1) {
4337 /* Flush any remaining cancel messages out to the
4339 llog_sync(ctxt, exp);
4341 llog_ctxt_put(ctxt);
4343 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
4347 rc = client_disconnect_export(exp);
4349 * Initially we put del_shrink_grant before disconnect_export, but it
4350 * causes the following problem if setup (connect) and cleanup
4351 * (disconnect) are tangled together.
4352 * connect p1 disconnect p2
4353 * ptlrpc_connect_import
4354 * ............... class_manual_cleanup
4357 * ptlrpc_connect_interrupt
4359 * add this client to shrink list
4361 * Bang! pinger trigger the shrink.
4362 * So the osc should be disconnected from the shrink list, after we
4363 * are sure the import has been destroyed. BUG18662
4365 if (obd->u.cli.cl_import == NULL)
4366 osc_del_shrink_grant(&obd->u.cli);
4370 static int osc_import_event(struct obd_device *obd,
4371 struct obd_import *imp,
4372 enum obd_import_event event)
4374 struct client_obd *cli;
4378 LASSERT(imp->imp_obd == obd);
4381 case IMP_EVENT_DISCON: {
4382 /* Only do this on the MDS OSC's */
4383 if (imp->imp_server_timeout) {
4384 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4386 cfs_spin_lock(&oscc->oscc_lock);
4387 oscc->oscc_flags |= OSCC_FLAG_RECOVERING;
4388 cfs_spin_unlock(&oscc->oscc_lock);
4391 client_obd_list_lock(&cli->cl_loi_list_lock);
4392 cli->cl_avail_grant = 0;
4393 cli->cl_lost_grant = 0;
4394 client_obd_list_unlock(&cli->cl_loi_list_lock);
4397 case IMP_EVENT_INACTIVE: {
4398 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
4401 case IMP_EVENT_INVALIDATE: {
4402 struct ldlm_namespace *ns = obd->obd_namespace;
4406 env = cl_env_get(&refcheck);
4410 client_obd_list_lock(&cli->cl_loi_list_lock);
4411 /* all pages go to failing rpcs due to the invalid
4413 osc_check_rpcs(env, cli);
4414 client_obd_list_unlock(&cli->cl_loi_list_lock);
4416 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
4417 cl_env_put(env, &refcheck);
4422 case IMP_EVENT_ACTIVE: {
4423 /* Only do this on the MDS OSC's */
4424 if (imp->imp_server_timeout) {
4425 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4427 cfs_spin_lock(&oscc->oscc_lock);
4428 oscc->oscc_flags &= ~OSCC_FLAG_NOSPC;
4429 cfs_spin_unlock(&oscc->oscc_lock);
4431 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
4434 case IMP_EVENT_OCD: {
4435 struct obd_connect_data *ocd = &imp->imp_connect_data;
4437 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
4438 osc_init_grant(&obd->u.cli, ocd);
4441 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
4442 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
4444 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
4448 CERROR("Unknown import event %d\n", event);
4455 * Determine whether the lock can be canceled before replaying the lock
4456 * during recovery, see bug16774 for detailed information.
4458 * \retval zero the lock can't be canceled
4459 * \retval other ok to cancel
4461 static int osc_cancel_for_recovery(struct ldlm_lock *lock)
4463 check_res_locked(lock->l_resource);
4466 * Cancel all unused extent lock in granted mode LCK_PR or LCK_CR.
4468 * XXX as a future improvement, we can also cancel unused write lock
4469 * if it doesn't have dirty data and active mmaps.
4471 if (lock->l_resource->lr_type == LDLM_EXTENT &&
4472 (lock->l_granted_mode == LCK_PR ||
4473 lock->l_granted_mode == LCK_CR) &&
4474 (osc_dlm_lock_pageref(lock) == 0))
4480 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
4486 rc = ptlrpcd_addref();
4490 rc = client_obd_setup(obd, lcfg);
4494 struct lprocfs_static_vars lvars = { 0 };
4495 struct client_obd *cli = &obd->u.cli;
4497 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
4498 lprocfs_osc_init_vars(&lvars);
4499 if (lprocfs_obd_setup(obd, lvars.obd_vars) == 0) {
4500 lproc_osc_attach_seqstat(obd);
4501 sptlrpc_lprocfs_cliobd_attach(obd);
4502 ptlrpc_lprocfs_register_obd(obd);
4506 /* We need to allocate a few requests more, because
4507 brw_interpret tries to create new requests before freeing
4508 previous ones. Ideally we want to have 2x max_rpcs_in_flight
4509 reserved, but I afraid that might be too much wasted RAM
4510 in fact, so 2 is just my guess and still should work. */
4511 cli->cl_import->imp_rq_pool =
4512 ptlrpc_init_rq_pool(cli->cl_max_rpcs_in_flight + 2,
4514 ptlrpc_add_rqs_to_pool);
4516 CFS_INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
4517 cfs_sema_init(&cli->cl_grant_sem, 1);
4519 ns_register_cancel(obd->obd_namespace, osc_cancel_for_recovery);
4525 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
4531 case OBD_CLEANUP_EARLY: {
4532 struct obd_import *imp;
4533 imp = obd->u.cli.cl_import;
4534 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
4535 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
4536 ptlrpc_deactivate_import(imp);
4537 cfs_spin_lock(&imp->imp_lock);
4538 imp->imp_pingable = 0;
4539 cfs_spin_unlock(&imp->imp_lock);
4542 case OBD_CLEANUP_EXPORTS: {
4543 /* If we set up but never connected, the
4544 client import will not have been cleaned. */
4545 if (obd->u.cli.cl_import) {
4546 struct obd_import *imp;
4547 cfs_down_write(&obd->u.cli.cl_sem);
4548 imp = obd->u.cli.cl_import;
4549 CDEBUG(D_CONFIG, "%s: client import never connected\n",
4551 ptlrpc_invalidate_import(imp);
4552 if (imp->imp_rq_pool) {
4553 ptlrpc_free_rq_pool(imp->imp_rq_pool);
4554 imp->imp_rq_pool = NULL;
4556 class_destroy_import(imp);
4557 cfs_up_write(&obd->u.cli.cl_sem);
4558 obd->u.cli.cl_import = NULL;
4560 rc = obd_llog_finish(obd, 0);
4562 CERROR("failed to cleanup llogging subsystems\n");
4569 int osc_cleanup(struct obd_device *obd)
4574 ptlrpc_lprocfs_unregister_obd(obd);
4575 lprocfs_obd_cleanup(obd);
4577 /* free memory of osc quota cache */
4578 lquota_cleanup(quota_interface, obd);
4580 rc = client_obd_cleanup(obd);
4586 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
4588 struct lprocfs_static_vars lvars = { 0 };
4591 lprocfs_osc_init_vars(&lvars);
4593 switch (lcfg->lcfg_command) {
4595 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars,
4605 static int osc_sync_fs(struct obd_export *exp, struct obd_info *oinfo,
4608 struct obd_device *obd = class_exp2obd(exp);
4609 struct client_obd *cli;
4610 struct lov_oinfo *loi;
4611 struct lov_oinfo *tloi;
4612 struct osc_async_page *oap;
4613 struct osc_async_page *toap;
4614 struct loi_oap_pages *lop;
4620 env = cl_env_get(&refcheck);
4622 RETURN(PTR_ERR(env));
4625 client_obd_list_lock(&cli->cl_loi_list_lock);
4626 cli->cl_sf_wait.sfw_oi = oinfo;
4627 cli->cl_sf_wait.sfw_upcall = oinfo->oi_cb_up;
4628 cli->cl_sf_wait.started = 1;
4629 /* creating cl_loi_sync_fs list */
4630 cfs_list_for_each_entry_safe(loi, tloi, &cli->cl_loi_write_list,
4632 lop = &loi->loi_write_lop;
4633 cfs_list_for_each_entry_safe(oap, toap, &lop->lop_pending,
4635 osc_set_async_flags_base(cli, loi, oap, ASYNC_SYNCFS);
4637 osc_check_rpcs(env, cli);
4638 osc_wake_sync_fs(cli);
4639 client_obd_list_unlock(&cli->cl_loi_list_lock);
4640 cl_env_put(env, &refcheck);
4645 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
4647 return osc_process_config_base(obd, buf);
4650 struct obd_ops osc_obd_ops = {
4651 .o_owner = THIS_MODULE,
4652 .o_setup = osc_setup,
4653 .o_precleanup = osc_precleanup,
4654 .o_cleanup = osc_cleanup,
4655 .o_add_conn = client_import_add_conn,
4656 .o_del_conn = client_import_del_conn,
4657 .o_connect = client_connect_import,
4658 .o_reconnect = osc_reconnect,
4659 .o_disconnect = osc_disconnect,
4660 .o_statfs = osc_statfs,
4661 .o_statfs_async = osc_statfs_async,
4662 .o_packmd = osc_packmd,
4663 .o_unpackmd = osc_unpackmd,
4664 .o_precreate = osc_precreate,
4665 .o_create = osc_create,
4666 .o_create_async = osc_create_async,
4667 .o_destroy = osc_destroy,
4668 .o_getattr = osc_getattr,
4669 .o_getattr_async = osc_getattr_async,
4670 .o_setattr = osc_setattr,
4671 .o_setattr_async = osc_setattr_async,
4673 .o_punch = osc_punch,
4675 .o_enqueue = osc_enqueue,
4676 .o_change_cbdata = osc_change_cbdata,
4677 .o_find_cbdata = osc_find_cbdata,
4678 .o_cancel = osc_cancel,
4679 .o_cancel_unused = osc_cancel_unused,
4680 .o_iocontrol = osc_iocontrol,
4681 .o_get_info = osc_get_info,
4682 .o_set_info_async = osc_set_info_async,
4683 .o_import_event = osc_import_event,
4684 .o_llog_init = osc_llog_init,
4685 .o_llog_finish = osc_llog_finish,
4686 .o_process_config = osc_process_config,
4687 .o_sync_fs = osc_sync_fs,
4690 extern struct lu_kmem_descr osc_caches[];
4691 extern cfs_spinlock_t osc_ast_guard;
4692 extern cfs_lock_class_key_t osc_ast_guard_class;
4694 int __init osc_init(void)
4696 struct lprocfs_static_vars lvars = { 0 };
4700 /* print an address of _any_ initialized kernel symbol from this
4701 * module, to allow debugging with gdb that doesn't support data
4702 * symbols from modules.*/
4703 CDEBUG(D_CONSOLE, "Lustre OSC module (%p).\n", &osc_caches);
4705 rc = lu_kmem_init(osc_caches);
4707 lprocfs_osc_init_vars(&lvars);
4709 cfs_request_module("lquota");
4710 quota_interface = PORTAL_SYMBOL_GET(osc_quota_interface);
4711 lquota_init(quota_interface);
4712 init_obd_quota_ops(quota_interface, &osc_obd_ops);
4714 rc = class_register_type(&osc_obd_ops, NULL, lvars.module_vars,
4715 LUSTRE_OSC_NAME, &osc_device_type);
4717 if (quota_interface)
4718 PORTAL_SYMBOL_PUT(osc_quota_interface);
4719 lu_kmem_fini(osc_caches);
4723 cfs_spin_lock_init(&osc_ast_guard);
4724 cfs_lockdep_set_class(&osc_ast_guard, &osc_ast_guard_class);
4726 osc_mds_ost_orig_logops = llog_lvfs_ops;
4727 osc_mds_ost_orig_logops.lop_setup = llog_obd_origin_setup;
4728 osc_mds_ost_orig_logops.lop_cleanup = llog_obd_origin_cleanup;
4729 osc_mds_ost_orig_logops.lop_add = llog_obd_origin_add;
4730 osc_mds_ost_orig_logops.lop_connect = llog_origin_connect;
4736 static void /*__exit*/ osc_exit(void)
4738 lu_device_type_fini(&osc_device_type);
4740 lquota_exit(quota_interface);
4741 if (quota_interface)
4742 PORTAL_SYMBOL_PUT(osc_quota_interface);
4744 class_unregister_type(LUSTRE_OSC_NAME);
4745 lu_kmem_fini(osc_caches);
4748 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
4749 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
4750 MODULE_LICENSE("GPL");
4752 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);