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 /* caller must hold loi_list_lock */
868 void osc_wake_cache_waiters(struct client_obd *cli)
871 struct osc_cache_waiter *ocw;
874 cfs_list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
875 /* if we can't dirty more, we must wait until some is written */
876 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
877 (cfs_atomic_read(&obd_dirty_pages) + 1 >
878 obd_max_dirty_pages)) {
879 CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
880 "osc max %ld, sys max %d\n", cli->cl_dirty,
881 cli->cl_dirty_max, obd_max_dirty_pages);
885 /* if still dirty cache but no grant wait for pending RPCs that
886 * may yet return us some grant before doing sync writes */
887 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
888 CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
889 cli->cl_w_in_flight);
893 ocw = cfs_list_entry(l, struct osc_cache_waiter, ocw_entry);
894 cfs_list_del_init(&ocw->ocw_entry);
895 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
896 /* no more RPCs in flight to return grant, do sync IO */
897 ocw->ocw_rc = -EDQUOT;
898 CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
900 osc_consume_write_grant(cli,
901 &ocw->ocw_oap->oap_brw_page);
904 cfs_waitq_signal(&ocw->ocw_waitq);
910 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
912 client_obd_list_lock(&cli->cl_loi_list_lock);
913 cli->cl_avail_grant += grant;
914 client_obd_list_unlock(&cli->cl_loi_list_lock);
917 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
919 if (body->oa.o_valid & OBD_MD_FLGRANT) {
920 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
921 __osc_update_grant(cli, body->oa.o_grant);
925 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
926 void *key, obd_count vallen, void *val,
927 struct ptlrpc_request_set *set);
929 static int osc_shrink_grant_interpret(const struct lu_env *env,
930 struct ptlrpc_request *req,
933 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
934 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
935 struct ost_body *body;
938 __osc_update_grant(cli, oa->o_grant);
942 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
944 osc_update_grant(cli, body);
950 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
952 client_obd_list_lock(&cli->cl_loi_list_lock);
953 oa->o_grant = cli->cl_avail_grant / 4;
954 cli->cl_avail_grant -= oa->o_grant;
955 client_obd_list_unlock(&cli->cl_loi_list_lock);
956 oa->o_flags |= OBD_FL_SHRINK_GRANT;
957 osc_update_next_shrink(cli);
960 /* Shrink the current grant, either from some large amount to enough for a
961 * full set of in-flight RPCs, or if we have already shrunk to that limit
962 * then to enough for a single RPC. This avoids keeping more grant than
963 * needed, and avoids shrinking the grant piecemeal. */
964 static int osc_shrink_grant(struct client_obd *cli)
966 long target = (cli->cl_max_rpcs_in_flight + 1) *
967 cli->cl_max_pages_per_rpc;
969 client_obd_list_lock(&cli->cl_loi_list_lock);
970 if (cli->cl_avail_grant <= target)
971 target = cli->cl_max_pages_per_rpc;
972 client_obd_list_unlock(&cli->cl_loi_list_lock);
974 return osc_shrink_grant_to_target(cli, target);
977 int osc_shrink_grant_to_target(struct client_obd *cli, long target)
980 struct ost_body *body;
983 client_obd_list_lock(&cli->cl_loi_list_lock);
984 /* Don't shrink if we are already above or below the desired limit
985 * We don't want to shrink below a single RPC, as that will negatively
986 * impact block allocation and long-term performance. */
987 if (target < cli->cl_max_pages_per_rpc)
988 target = cli->cl_max_pages_per_rpc;
990 if (target >= cli->cl_avail_grant) {
991 client_obd_list_unlock(&cli->cl_loi_list_lock);
994 client_obd_list_unlock(&cli->cl_loi_list_lock);
1000 osc_announce_cached(cli, &body->oa, 0);
1002 client_obd_list_lock(&cli->cl_loi_list_lock);
1003 body->oa.o_grant = cli->cl_avail_grant - target;
1004 cli->cl_avail_grant = target;
1005 client_obd_list_unlock(&cli->cl_loi_list_lock);
1006 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
1007 osc_update_next_shrink(cli);
1009 rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
1010 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
1011 sizeof(*body), body, NULL);
1013 __osc_update_grant(cli, body->oa.o_grant);
1018 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
1019 static int osc_should_shrink_grant(struct client_obd *client)
1021 cfs_time_t time = cfs_time_current();
1022 cfs_time_t next_shrink = client->cl_next_shrink_grant;
1024 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
1025 OBD_CONNECT_GRANT_SHRINK) == 0)
1028 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
1029 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
1030 client->cl_avail_grant > GRANT_SHRINK_LIMIT)
1033 osc_update_next_shrink(client);
1038 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1040 struct client_obd *client;
1042 cfs_list_for_each_entry(client, &item->ti_obd_list,
1043 cl_grant_shrink_list) {
1044 if (osc_should_shrink_grant(client))
1045 osc_shrink_grant(client);
1050 static int osc_add_shrink_grant(struct client_obd *client)
1054 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
1056 osc_grant_shrink_grant_cb, NULL,
1057 &client->cl_grant_shrink_list);
1059 CERROR("add grant client %s error %d\n",
1060 client->cl_import->imp_obd->obd_name, rc);
1063 CDEBUG(D_CACHE, "add grant client %s \n",
1064 client->cl_import->imp_obd->obd_name);
1065 osc_update_next_shrink(client);
1069 static int osc_del_shrink_grant(struct client_obd *client)
1071 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1075 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1078 * ocd_grant is the total grant amount we're expect to hold: if we've
1079 * been evicted, it's the new avail_grant amount, cl_dirty will drop
1080 * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1082 * race is tolerable here: if we're evicted, but imp_state already
1083 * left EVICTED state, then cl_dirty must be 0 already.
1085 client_obd_list_lock(&cli->cl_loi_list_lock);
1086 if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1087 cli->cl_avail_grant = ocd->ocd_grant;
1089 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1091 if (cli->cl_avail_grant < 0) {
1092 CWARN("%s: available grant < 0, the OSS is probably not running"
1093 " with patch from bug20278 (%ld) \n",
1094 cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant);
1095 /* workaround for 1.6 servers which do not have
1096 * the patch from bug20278 */
1097 cli->cl_avail_grant = ocd->ocd_grant;
1100 client_obd_list_unlock(&cli->cl_loi_list_lock);
1102 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
1103 cli->cl_import->imp_obd->obd_name,
1104 cli->cl_avail_grant, cli->cl_lost_grant);
1106 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1107 cfs_list_empty(&cli->cl_grant_shrink_list))
1108 osc_add_shrink_grant(cli);
1111 /* We assume that the reason this OSC got a short read is because it read
1112 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1113 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1114 * this stripe never got written at or beyond this stripe offset yet. */
1115 static void handle_short_read(int nob_read, obd_count page_count,
1116 struct brw_page **pga)
1121 /* skip bytes read OK */
1122 while (nob_read > 0) {
1123 LASSERT (page_count > 0);
1125 if (pga[i]->count > nob_read) {
1126 /* EOF inside this page */
1127 ptr = cfs_kmap(pga[i]->pg) +
1128 (pga[i]->off & ~CFS_PAGE_MASK);
1129 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1130 cfs_kunmap(pga[i]->pg);
1136 nob_read -= pga[i]->count;
1141 /* zero remaining pages */
1142 while (page_count-- > 0) {
1143 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1144 memset(ptr, 0, pga[i]->count);
1145 cfs_kunmap(pga[i]->pg);
1150 static int check_write_rcs(struct ptlrpc_request *req,
1151 int requested_nob, int niocount,
1152 obd_count page_count, struct brw_page **pga)
1157 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1158 sizeof(*remote_rcs) *
1160 if (remote_rcs == NULL) {
1161 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1165 /* return error if any niobuf was in error */
1166 for (i = 0; i < niocount; i++) {
1167 if (remote_rcs[i] < 0)
1168 return(remote_rcs[i]);
1170 if (remote_rcs[i] != 0) {
1171 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1172 i, remote_rcs[i], req);
1177 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1178 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1179 req->rq_bulk->bd_nob_transferred, requested_nob);
1186 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1188 if (p1->flag != p2->flag) {
1189 unsigned mask = ~(OBD_BRW_FROM_GRANT|
1190 OBD_BRW_NOCACHE|OBD_BRW_SYNC|OBD_BRW_ASYNC);
1192 /* warn if we try to combine flags that we don't know to be
1193 * safe to combine */
1194 if ((p1->flag & mask) != (p2->flag & mask))
1195 CERROR("is it ok to have flags 0x%x and 0x%x in the "
1196 "same brw?\n", p1->flag, p2->flag);
1200 return (p1->off + p1->count == p2->off);
1203 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1204 struct brw_page **pga, int opc,
1205 cksum_type_t cksum_type)
1210 LASSERT (pg_count > 0);
1211 cksum = init_checksum(cksum_type);
1212 while (nob > 0 && pg_count > 0) {
1213 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1214 int off = pga[i]->off & ~CFS_PAGE_MASK;
1215 int count = pga[i]->count > nob ? nob : pga[i]->count;
1217 /* corrupt the data before we compute the checksum, to
1218 * simulate an OST->client data error */
1219 if (i == 0 && opc == OST_READ &&
1220 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1221 memcpy(ptr + off, "bad1", min(4, nob));
1222 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1223 cfs_kunmap(pga[i]->pg);
1224 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1227 nob -= pga[i]->count;
1231 /* For sending we only compute the wrong checksum instead
1232 * of corrupting the data so it is still correct on a redo */
1233 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1239 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1240 struct lov_stripe_md *lsm, obd_count page_count,
1241 struct brw_page **pga,
1242 struct ptlrpc_request **reqp,
1243 struct obd_capa *ocapa, int reserve)
1245 struct ptlrpc_request *req;
1246 struct ptlrpc_bulk_desc *desc;
1247 struct ost_body *body;
1248 struct obd_ioobj *ioobj;
1249 struct niobuf_remote *niobuf;
1250 int niocount, i, requested_nob, opc, rc;
1251 struct osc_brw_async_args *aa;
1252 struct req_capsule *pill;
1253 struct brw_page *pg_prev;
1256 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1257 RETURN(-ENOMEM); /* Recoverable */
1258 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1259 RETURN(-EINVAL); /* Fatal */
1261 if ((cmd & OBD_BRW_WRITE) != 0) {
1263 req = ptlrpc_request_alloc_pool(cli->cl_import,
1264 cli->cl_import->imp_rq_pool,
1265 &RQF_OST_BRW_WRITE);
1268 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1273 for (niocount = i = 1; i < page_count; i++) {
1274 if (!can_merge_pages(pga[i - 1], pga[i]))
1278 pill = &req->rq_pill;
1279 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1281 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1282 niocount * sizeof(*niobuf));
1283 osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1285 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1287 ptlrpc_request_free(req);
1290 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1291 ptlrpc_at_set_req_timeout(req);
1293 if (opc == OST_WRITE)
1294 desc = ptlrpc_prep_bulk_imp(req, page_count,
1295 BULK_GET_SOURCE, OST_BULK_PORTAL);
1297 desc = ptlrpc_prep_bulk_imp(req, page_count,
1298 BULK_PUT_SINK, OST_BULK_PORTAL);
1301 GOTO(out, rc = -ENOMEM);
1302 /* NB request now owns desc and will free it when it gets freed */
1304 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1305 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1306 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1307 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1309 lustre_set_wire_obdo(&body->oa, oa);
1311 obdo_to_ioobj(oa, ioobj);
1312 ioobj->ioo_bufcnt = niocount;
1313 osc_pack_capa(req, body, ocapa);
1314 LASSERT (page_count > 0);
1316 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1317 struct brw_page *pg = pga[i];
1319 LASSERT(pg->count > 0);
1320 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1321 "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1322 pg->off, pg->count);
1324 LASSERTF(i == 0 || pg->off > pg_prev->off,
1325 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1326 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1328 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1329 pg_prev->pg, page_private(pg_prev->pg),
1330 pg_prev->pg->index, pg_prev->off);
1332 LASSERTF(i == 0 || pg->off > pg_prev->off,
1333 "i %d p_c %u\n", i, page_count);
1335 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1336 (pg->flag & OBD_BRW_SRVLOCK));
1338 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1340 requested_nob += pg->count;
1342 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1344 niobuf->len += pg->count;
1346 niobuf->offset = pg->off;
1347 niobuf->len = pg->count;
1348 niobuf->flags = pg->flag;
1353 LASSERTF((void *)(niobuf - niocount) ==
1354 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1355 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1356 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1358 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1359 if (osc_should_shrink_grant(cli))
1360 osc_shrink_grant_local(cli, &body->oa);
1362 /* size[REQ_REC_OFF] still sizeof (*body) */
1363 if (opc == OST_WRITE) {
1364 if (unlikely(cli->cl_checksum) &&
1365 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1366 /* store cl_cksum_type in a local variable since
1367 * it can be changed via lprocfs */
1368 cksum_type_t cksum_type = cli->cl_cksum_type;
1370 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1371 oa->o_flags &= OBD_FL_LOCAL_MASK;
1372 body->oa.o_flags = 0;
1374 body->oa.o_flags |= cksum_type_pack(cksum_type);
1375 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1376 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1380 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1382 /* save this in 'oa', too, for later checking */
1383 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1384 oa->o_flags |= cksum_type_pack(cksum_type);
1386 /* clear out the checksum flag, in case this is a
1387 * resend but cl_checksum is no longer set. b=11238 */
1388 oa->o_valid &= ~OBD_MD_FLCKSUM;
1390 oa->o_cksum = body->oa.o_cksum;
1391 /* 1 RC per niobuf */
1392 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1393 sizeof(__u32) * niocount);
1395 if (unlikely(cli->cl_checksum) &&
1396 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1397 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1398 body->oa.o_flags = 0;
1399 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1400 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1403 ptlrpc_request_set_replen(req);
1405 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1406 aa = ptlrpc_req_async_args(req);
1408 aa->aa_requested_nob = requested_nob;
1409 aa->aa_nio_count = niocount;
1410 aa->aa_page_count = page_count;
1414 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1415 if (ocapa && reserve)
1416 aa->aa_ocapa = capa_get(ocapa);
1422 ptlrpc_req_finished(req);
1426 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1427 __u32 client_cksum, __u32 server_cksum, int nob,
1428 obd_count page_count, struct brw_page **pga,
1429 cksum_type_t client_cksum_type)
1433 cksum_type_t cksum_type;
1435 if (server_cksum == client_cksum) {
1436 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1440 /* If this is mmaped file - it can be changed at any time */
1441 if (oa->o_valid & OBD_MD_FLFLAGS && oa->o_flags & OBD_FL_MMAP)
1444 if (oa->o_valid & OBD_MD_FLFLAGS)
1445 cksum_type = cksum_type_unpack(oa->o_flags);
1447 cksum_type = OBD_CKSUM_CRC32;
1449 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1452 if (cksum_type != client_cksum_type)
1453 msg = "the server did not use the checksum type specified in "
1454 "the original request - likely a protocol problem";
1455 else if (new_cksum == server_cksum)
1456 msg = "changed on the client after we checksummed it - "
1457 "likely false positive due to mmap IO (bug 11742)";
1458 else if (new_cksum == client_cksum)
1459 msg = "changed in transit before arrival at OST";
1461 msg = "changed in transit AND doesn't match the original - "
1462 "likely false positive due to mmap IO (bug 11742)";
1464 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1465 " object "LPU64"/"LPU64" extent ["LPU64"-"LPU64"]\n",
1466 msg, libcfs_nid2str(peer->nid),
1467 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1468 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1469 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1471 oa->o_valid & OBD_MD_FLGROUP ? oa->o_seq : (__u64)0,
1473 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1474 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1475 "client csum now %x\n", client_cksum, client_cksum_type,
1476 server_cksum, cksum_type, new_cksum);
1480 /* Note rc enters this function as number of bytes transferred */
1481 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1483 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1484 const lnet_process_id_t *peer =
1485 &req->rq_import->imp_connection->c_peer;
1486 struct client_obd *cli = aa->aa_cli;
1487 struct ost_body *body;
1488 __u32 client_cksum = 0;
1491 if (rc < 0 && rc != -EDQUOT) {
1492 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1496 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1497 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1499 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1503 #ifdef HAVE_QUOTA_SUPPORT
1504 /* set/clear over quota flag for a uid/gid */
1505 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1506 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1507 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1509 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid "LPX64", flags %x\n",
1510 body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1512 lquota_setdq(quota_interface, cli, qid, body->oa.o_valid,
1517 osc_update_grant(cli, body);
1522 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1523 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1525 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1527 CERROR("Unexpected +ve rc %d\n", rc);
1530 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1532 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1535 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1536 check_write_checksum(&body->oa, peer, client_cksum,
1537 body->oa.o_cksum, aa->aa_requested_nob,
1538 aa->aa_page_count, aa->aa_ppga,
1539 cksum_type_unpack(aa->aa_oa->o_flags)))
1542 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1543 aa->aa_page_count, aa->aa_ppga);
1547 /* The rest of this function executes only for OST_READs */
1549 /* if unwrap_bulk failed, return -EAGAIN to retry */
1550 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1552 GOTO(out, rc = -EAGAIN);
1554 if (rc > aa->aa_requested_nob) {
1555 CERROR("Unexpected rc %d (%d requested)\n", rc,
1556 aa->aa_requested_nob);
1560 if (rc != req->rq_bulk->bd_nob_transferred) {
1561 CERROR ("Unexpected rc %d (%d transferred)\n",
1562 rc, req->rq_bulk->bd_nob_transferred);
1566 if (rc < aa->aa_requested_nob)
1567 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1569 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1570 static int cksum_counter;
1571 __u32 server_cksum = body->oa.o_cksum;
1574 cksum_type_t cksum_type;
1576 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1577 cksum_type = cksum_type_unpack(body->oa.o_flags);
1579 cksum_type = OBD_CKSUM_CRC32;
1580 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1581 aa->aa_ppga, OST_READ,
1584 if (peer->nid == req->rq_bulk->bd_sender) {
1588 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1591 if (server_cksum == ~0 && rc > 0) {
1592 CERROR("Protocol error: server %s set the 'checksum' "
1593 "bit, but didn't send a checksum. Not fatal, "
1594 "but please notify on http://bugzilla.lustre.org/\n",
1595 libcfs_nid2str(peer->nid));
1596 } else if (server_cksum != client_cksum) {
1597 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1598 "%s%s%s inode "DFID" object "
1599 LPU64"/"LPU64" extent "
1600 "["LPU64"-"LPU64"]\n",
1601 req->rq_import->imp_obd->obd_name,
1602 libcfs_nid2str(peer->nid),
1604 body->oa.o_valid & OBD_MD_FLFID ?
1605 body->oa.o_parent_seq : (__u64)0,
1606 body->oa.o_valid & OBD_MD_FLFID ?
1607 body->oa.o_parent_oid : 0,
1608 body->oa.o_valid & OBD_MD_FLFID ?
1609 body->oa.o_parent_ver : 0,
1611 body->oa.o_valid & OBD_MD_FLGROUP ?
1612 body->oa.o_seq : (__u64)0,
1613 aa->aa_ppga[0]->off,
1614 aa->aa_ppga[aa->aa_page_count-1]->off +
1615 aa->aa_ppga[aa->aa_page_count-1]->count -
1617 CERROR("client %x, server %x, cksum_type %x\n",
1618 client_cksum, server_cksum, cksum_type);
1620 aa->aa_oa->o_cksum = client_cksum;
1624 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1627 } else if (unlikely(client_cksum)) {
1628 static int cksum_missed;
1631 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1632 CERROR("Checksum %u requested from %s but not sent\n",
1633 cksum_missed, libcfs_nid2str(peer->nid));
1639 lustre_get_wire_obdo(aa->aa_oa, &body->oa);
1644 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1645 struct lov_stripe_md *lsm,
1646 obd_count page_count, struct brw_page **pga,
1647 struct obd_capa *ocapa)
1649 struct ptlrpc_request *req;
1653 struct l_wait_info lwi;
1657 cfs_waitq_init(&waitq);
1660 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1661 page_count, pga, &req, ocapa, 0);
1665 rc = ptlrpc_queue_wait(req);
1667 if (rc == -ETIMEDOUT && req->rq_resend) {
1668 DEBUG_REQ(D_HA, req, "BULK TIMEOUT");
1669 ptlrpc_req_finished(req);
1673 rc = osc_brw_fini_request(req, rc);
1675 ptlrpc_req_finished(req);
1676 if (osc_recoverable_error(rc)) {
1678 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1679 CERROR("too many resend retries, returning error\n");
1683 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1684 l_wait_event(waitq, 0, &lwi);
1692 int osc_brw_redo_request(struct ptlrpc_request *request,
1693 struct osc_brw_async_args *aa)
1695 struct ptlrpc_request *new_req;
1696 struct ptlrpc_request_set *set = request->rq_set;
1697 struct osc_brw_async_args *new_aa;
1698 struct osc_async_page *oap;
1702 if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1703 CERROR("too many resent retries, returning error\n");
1707 DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1709 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1710 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1711 aa->aa_cli, aa->aa_oa,
1712 NULL /* lsm unused by osc currently */,
1713 aa->aa_page_count, aa->aa_ppga,
1714 &new_req, aa->aa_ocapa, 0);
1718 client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1720 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1721 if (oap->oap_request != NULL) {
1722 LASSERTF(request == oap->oap_request,
1723 "request %p != oap_request %p\n",
1724 request, oap->oap_request);
1725 if (oap->oap_interrupted) {
1726 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1727 ptlrpc_req_finished(new_req);
1732 /* New request takes over pga and oaps from old request.
1733 * Note that copying a list_head doesn't work, need to move it... */
1735 new_req->rq_interpret_reply = request->rq_interpret_reply;
1736 new_req->rq_async_args = request->rq_async_args;
1737 new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1739 new_aa = ptlrpc_req_async_args(new_req);
1741 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1742 cfs_list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1743 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1745 cfs_list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1746 if (oap->oap_request) {
1747 ptlrpc_req_finished(oap->oap_request);
1748 oap->oap_request = ptlrpc_request_addref(new_req);
1752 new_aa->aa_ocapa = aa->aa_ocapa;
1753 aa->aa_ocapa = NULL;
1755 /* use ptlrpc_set_add_req is safe because interpret functions work
1756 * in check_set context. only one way exist with access to request
1757 * from different thread got -EINTR - this way protected with
1758 * cl_loi_list_lock */
1759 ptlrpc_set_add_req(set, new_req);
1761 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1763 DEBUG_REQ(D_INFO, new_req, "new request");
1768 * ugh, we want disk allocation on the target to happen in offset order. we'll
1769 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1770 * fine for our small page arrays and doesn't require allocation. its an
1771 * insertion sort that swaps elements that are strides apart, shrinking the
1772 * stride down until its '1' and the array is sorted.
1774 static void sort_brw_pages(struct brw_page **array, int num)
1777 struct brw_page *tmp;
1781 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1786 for (i = stride ; i < num ; i++) {
1789 while (j >= stride && array[j - stride]->off > tmp->off) {
1790 array[j] = array[j - stride];
1795 } while (stride > 1);
1798 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1804 LASSERT (pages > 0);
1805 offset = pg[i]->off & ~CFS_PAGE_MASK;
1809 if (pages == 0) /* that's all */
1812 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1813 return count; /* doesn't end on page boundary */
1816 offset = pg[i]->off & ~CFS_PAGE_MASK;
1817 if (offset != 0) /* doesn't start on page boundary */
1824 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1826 struct brw_page **ppga;
1829 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1833 for (i = 0; i < count; i++)
1838 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1840 LASSERT(ppga != NULL);
1841 OBD_FREE(ppga, sizeof(*ppga) * count);
1844 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1845 obd_count page_count, struct brw_page *pga,
1846 struct obd_trans_info *oti)
1848 struct obdo *saved_oa = NULL;
1849 struct brw_page **ppga, **orig;
1850 struct obd_import *imp = class_exp2cliimp(exp);
1851 struct client_obd *cli;
1852 int rc, page_count_orig;
1855 LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1856 cli = &imp->imp_obd->u.cli;
1858 if (cmd & OBD_BRW_CHECK) {
1859 /* The caller just wants to know if there's a chance that this
1860 * I/O can succeed */
1862 if (imp->imp_invalid)
1867 /* test_brw with a failed create can trip this, maybe others. */
1868 LASSERT(cli->cl_max_pages_per_rpc);
1872 orig = ppga = osc_build_ppga(pga, page_count);
1875 page_count_orig = page_count;
1877 sort_brw_pages(ppga, page_count);
1878 while (page_count) {
1879 obd_count pages_per_brw;
1881 if (page_count > cli->cl_max_pages_per_rpc)
1882 pages_per_brw = cli->cl_max_pages_per_rpc;
1884 pages_per_brw = page_count;
1886 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1888 if (saved_oa != NULL) {
1889 /* restore previously saved oa */
1890 *oinfo->oi_oa = *saved_oa;
1891 } else if (page_count > pages_per_brw) {
1892 /* save a copy of oa (brw will clobber it) */
1893 OBDO_ALLOC(saved_oa);
1894 if (saved_oa == NULL)
1895 GOTO(out, rc = -ENOMEM);
1896 *saved_oa = *oinfo->oi_oa;
1899 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1900 pages_per_brw, ppga, oinfo->oi_capa);
1905 page_count -= pages_per_brw;
1906 ppga += pages_per_brw;
1910 osc_release_ppga(orig, page_count_orig);
1912 if (saved_oa != NULL)
1913 OBDO_FREE(saved_oa);
1918 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1919 * the dirty accounting. Writeback completes or truncate happens before
1920 * writing starts. Must be called with the loi lock held. */
1921 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1924 osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1928 /* This maintains the lists of pending pages to read/write for a given object
1929 * (lop). This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1930 * to quickly find objects that are ready to send an RPC. */
1931 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1937 if (lop->lop_num_pending == 0)
1940 /* if we have an invalid import we want to drain the queued pages
1941 * by forcing them through rpcs that immediately fail and complete
1942 * the pages. recovery relies on this to empty the queued pages
1943 * before canceling the locks and evicting down the llite pages */
1944 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1947 /* stream rpcs in queue order as long as as there is an urgent page
1948 * queued. this is our cheap solution for good batching in the case
1949 * where writepage marks some random page in the middle of the file
1950 * as urgent because of, say, memory pressure */
1951 if (!cfs_list_empty(&lop->lop_urgent)) {
1952 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1955 /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1956 optimal = cli->cl_max_pages_per_rpc;
1957 if (cmd & OBD_BRW_WRITE) {
1958 /* trigger a write rpc stream as long as there are dirtiers
1959 * waiting for space. as they're waiting, they're not going to
1960 * create more pages to coalesce with what's waiting.. */
1961 if (!cfs_list_empty(&cli->cl_cache_waiters)) {
1962 CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1965 /* +16 to avoid triggering rpcs that would want to include pages
1966 * that are being queued but which can't be made ready until
1967 * the queuer finishes with the page. this is a wart for
1968 * llite::commit_write() */
1971 if (lop->lop_num_pending >= optimal)
1977 static int lop_makes_hprpc(struct loi_oap_pages *lop)
1979 struct osc_async_page *oap;
1982 if (cfs_list_empty(&lop->lop_urgent))
1985 oap = cfs_list_entry(lop->lop_urgent.next,
1986 struct osc_async_page, oap_urgent_item);
1988 if (oap->oap_async_flags & ASYNC_HP) {
1989 CDEBUG(D_CACHE, "hp request forcing RPC\n");
1996 static void on_list(cfs_list_t *item, cfs_list_t *list,
1999 if (cfs_list_empty(item) && should_be_on)
2000 cfs_list_add_tail(item, list);
2001 else if (!cfs_list_empty(item) && !should_be_on)
2002 cfs_list_del_init(item);
2005 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
2006 * can find pages to build into rpcs quickly */
2007 void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
2009 if (lop_makes_hprpc(&loi->loi_write_lop) ||
2010 lop_makes_hprpc(&loi->loi_read_lop)) {
2012 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list, 0);
2013 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
2015 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 0);
2016 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list,
2017 lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)||
2018 lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ));
2021 on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
2022 loi->loi_write_lop.lop_num_pending);
2024 on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
2025 loi->loi_read_lop.lop_num_pending);
2028 static void lop_update_pending(struct client_obd *cli,
2029 struct loi_oap_pages *lop, int cmd, int delta)
2031 lop->lop_num_pending += delta;
2032 if (cmd & OBD_BRW_WRITE)
2033 cli->cl_pending_w_pages += delta;
2035 cli->cl_pending_r_pages += delta;
2039 * this is called when a sync waiter receives an interruption. Its job is to
2040 * get the caller woken as soon as possible. If its page hasn't been put in an
2041 * rpc yet it can dequeue immediately. Otherwise it has to mark the rpc as
2042 * desiring interruption which will forcefully complete the rpc once the rpc
2045 int osc_oap_interrupted(const struct lu_env *env, struct osc_async_page *oap)
2047 struct loi_oap_pages *lop;
2048 struct lov_oinfo *loi;
2052 LASSERT(!oap->oap_interrupted);
2053 oap->oap_interrupted = 1;
2055 /* ok, it's been put in an rpc. only one oap gets a request reference */
2056 if (oap->oap_request != NULL) {
2057 ptlrpc_mark_interrupted(oap->oap_request);
2058 ptlrpcd_wake(oap->oap_request);
2059 ptlrpc_req_finished(oap->oap_request);
2060 oap->oap_request = NULL;
2064 * page completion may be called only if ->cpo_prep() method was
2065 * executed by osc_io_submit(), that also adds page the to pending list
2067 if (!cfs_list_empty(&oap->oap_pending_item)) {
2068 cfs_list_del_init(&oap->oap_pending_item);
2069 cfs_list_del_init(&oap->oap_urgent_item);
2072 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
2073 &loi->loi_write_lop : &loi->loi_read_lop;
2074 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
2075 loi_list_maint(oap->oap_cli, oap->oap_loi);
2076 rc = oap->oap_caller_ops->ap_completion(env,
2077 oap->oap_caller_data,
2078 oap->oap_cmd, NULL, -EINTR);
2084 /* this is trying to propogate async writeback errors back up to the
2085 * application. As an async write fails we record the error code for later if
2086 * the app does an fsync. As long as errors persist we force future rpcs to be
2087 * sync so that the app can get a sync error and break the cycle of queueing
2088 * pages for which writeback will fail. */
2089 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
2096 ar->ar_force_sync = 1;
2097 ar->ar_min_xid = ptlrpc_sample_next_xid();
2102 if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2103 ar->ar_force_sync = 0;
2106 void osc_oap_to_pending(struct osc_async_page *oap)
2108 struct loi_oap_pages *lop;
2110 if (oap->oap_cmd & OBD_BRW_WRITE)
2111 lop = &oap->oap_loi->loi_write_lop;
2113 lop = &oap->oap_loi->loi_read_lop;
2115 if (oap->oap_async_flags & ASYNC_HP)
2116 cfs_list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2117 else if (oap->oap_async_flags & ASYNC_URGENT)
2118 cfs_list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2119 cfs_list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2120 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2123 /* this must be called holding the loi list lock to give coverage to exit_cache,
2124 * async_flag maintenance, and oap_request */
2125 static void osc_ap_completion(const struct lu_env *env,
2126 struct client_obd *cli, struct obdo *oa,
2127 struct osc_async_page *oap, int sent, int rc)
2132 if (oap->oap_request != NULL) {
2133 xid = ptlrpc_req_xid(oap->oap_request);
2134 ptlrpc_req_finished(oap->oap_request);
2135 oap->oap_request = NULL;
2138 cfs_spin_lock(&oap->oap_lock);
2139 oap->oap_async_flags = 0;
2140 cfs_spin_unlock(&oap->oap_lock);
2141 oap->oap_interrupted = 0;
2143 if (oap->oap_cmd & OBD_BRW_WRITE) {
2144 osc_process_ar(&cli->cl_ar, xid, rc);
2145 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2148 if (rc == 0 && oa != NULL) {
2149 if (oa->o_valid & OBD_MD_FLBLOCKS)
2150 oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2151 if (oa->o_valid & OBD_MD_FLMTIME)
2152 oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2153 if (oa->o_valid & OBD_MD_FLATIME)
2154 oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2155 if (oa->o_valid & OBD_MD_FLCTIME)
2156 oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2159 rc = oap->oap_caller_ops->ap_completion(env, oap->oap_caller_data,
2160 oap->oap_cmd, oa, rc);
2162 /* ll_ap_completion (from llite) drops PG_locked. so, a new
2163 * I/O on the page could start, but OSC calls it under lock
2164 * and thus we can add oap back to pending safely */
2166 /* upper layer wants to leave the page on pending queue */
2167 osc_oap_to_pending(oap);
2169 osc_exit_cache(cli, oap, sent);
2173 static int brw_interpret(const struct lu_env *env,
2174 struct ptlrpc_request *req, void *data, int rc)
2176 struct osc_brw_async_args *aa = data;
2177 struct client_obd *cli;
2181 rc = osc_brw_fini_request(req, rc);
2182 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2183 if (osc_recoverable_error(rc)) {
2184 /* Only retry once for mmaped files since the mmaped page
2185 * might be modified at anytime. We have to retry at least
2186 * once in case there WAS really a corruption of the page
2187 * on the network, that was not caused by mmap() modifying
2188 * the page. Bug11742 */
2189 if ((rc == -EAGAIN) && (aa->aa_resends > 0) &&
2190 aa->aa_oa->o_valid & OBD_MD_FLFLAGS &&
2191 aa->aa_oa->o_flags & OBD_FL_MMAP) {
2194 rc = osc_brw_redo_request(req, aa);
2201 capa_put(aa->aa_ocapa);
2202 aa->aa_ocapa = NULL;
2207 client_obd_list_lock(&cli->cl_loi_list_lock);
2209 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2210 * is called so we know whether to go to sync BRWs or wait for more
2211 * RPCs to complete */
2212 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2213 cli->cl_w_in_flight--;
2215 cli->cl_r_in_flight--;
2217 async = cfs_list_empty(&aa->aa_oaps);
2218 if (!async) { /* from osc_send_oap_rpc() */
2219 struct osc_async_page *oap, *tmp;
2220 /* the caller may re-use the oap after the completion call so
2221 * we need to clean it up a little */
2222 cfs_list_for_each_entry_safe(oap, tmp, &aa->aa_oaps,
2224 cfs_list_del_init(&oap->oap_rpc_item);
2225 osc_ap_completion(env, cli, aa->aa_oa, oap, 1, rc);
2227 OBDO_FREE(aa->aa_oa);
2228 } else { /* from async_internal() */
2230 for (i = 0; i < aa->aa_page_count; i++)
2231 osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2233 osc_wake_cache_waiters(cli);
2234 osc_check_rpcs(env, cli);
2235 client_obd_list_unlock(&cli->cl_loi_list_lock);
2237 cl_req_completion(env, aa->aa_clerq, rc);
2238 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2243 static struct ptlrpc_request *osc_build_req(const struct lu_env *env,
2244 struct client_obd *cli,
2245 cfs_list_t *rpc_list,
2246 int page_count, int cmd)
2248 struct ptlrpc_request *req;
2249 struct brw_page **pga = NULL;
2250 struct osc_brw_async_args *aa;
2251 struct obdo *oa = NULL;
2252 const struct obd_async_page_ops *ops = NULL;
2253 void *caller_data = NULL;
2254 struct osc_async_page *oap;
2255 struct osc_async_page *tmp;
2256 struct ost_body *body;
2257 struct cl_req *clerq = NULL;
2258 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2259 struct ldlm_lock *lock = NULL;
2260 struct cl_req_attr crattr;
2261 int i, rc, mpflag = 0;
2264 LASSERT(!cfs_list_empty(rpc_list));
2266 if (cmd & OBD_BRW_MEMALLOC)
2267 mpflag = cfs_memory_pressure_get_and_set();
2269 memset(&crattr, 0, sizeof crattr);
2270 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2272 GOTO(out, req = ERR_PTR(-ENOMEM));
2276 GOTO(out, req = ERR_PTR(-ENOMEM));
2279 cfs_list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2280 struct cl_page *page = osc_oap2cl_page(oap);
2282 ops = oap->oap_caller_ops;
2283 caller_data = oap->oap_caller_data;
2285 clerq = cl_req_alloc(env, page, crt,
2286 1 /* only 1-object rpcs for
2289 GOTO(out, req = (void *)clerq);
2290 lock = oap->oap_ldlm_lock;
2292 pga[i] = &oap->oap_brw_page;
2293 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2294 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2295 pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2297 cl_req_page_add(env, clerq, page);
2300 /* always get the data for the obdo for the rpc */
2301 LASSERT(ops != NULL);
2303 crattr.cra_capa = NULL;
2304 cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2306 oa->o_handle = lock->l_remote_handle;
2307 oa->o_valid |= OBD_MD_FLHANDLE;
2310 rc = cl_req_prep(env, clerq);
2312 CERROR("cl_req_prep failed: %d\n", rc);
2313 GOTO(out, req = ERR_PTR(rc));
2316 sort_brw_pages(pga, page_count);
2317 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2318 pga, &req, crattr.cra_capa, 1);
2320 CERROR("prep_req failed: %d\n", rc);
2321 GOTO(out, req = ERR_PTR(rc));
2324 if (cmd & OBD_BRW_MEMALLOC)
2325 req->rq_memalloc = 1;
2327 /* Need to update the timestamps after the request is built in case
2328 * we race with setattr (locally or in queue at OST). If OST gets
2329 * later setattr before earlier BRW (as determined by the request xid),
2330 * the OST will not use BRW timestamps. Sadly, there is no obvious
2331 * way to do this in a single call. bug 10150 */
2332 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2333 cl_req_attr_set(env, clerq, &crattr,
2334 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2336 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2337 aa = ptlrpc_req_async_args(req);
2338 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2339 cfs_list_splice(rpc_list, &aa->aa_oaps);
2340 CFS_INIT_LIST_HEAD(rpc_list);
2341 aa->aa_clerq = clerq;
2343 if (cmd & OBD_BRW_MEMALLOC)
2344 cfs_memory_pressure_restore(mpflag);
2346 capa_put(crattr.cra_capa);
2351 OBD_FREE(pga, sizeof(*pga) * page_count);
2352 /* this should happen rarely and is pretty bad, it makes the
2353 * pending list not follow the dirty order */
2354 client_obd_list_lock(&cli->cl_loi_list_lock);
2355 cfs_list_for_each_entry_safe(oap, tmp, rpc_list, oap_rpc_item) {
2356 cfs_list_del_init(&oap->oap_rpc_item);
2358 /* queued sync pages can be torn down while the pages
2359 * were between the pending list and the rpc */
2360 if (oap->oap_interrupted) {
2361 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2362 osc_ap_completion(env, cli, NULL, oap, 0,
2366 osc_ap_completion(env, cli, NULL, oap, 0, PTR_ERR(req));
2368 if (clerq && !IS_ERR(clerq))
2369 cl_req_completion(env, clerq, PTR_ERR(req));
2375 * prepare pages for ASYNC io and put pages in send queue.
2377 * \param cmd OBD_BRW_* macroses
2378 * \param lop pending pages
2380 * \return zero if no page added to send queue.
2381 * \return 1 if pages successfully added to send queue.
2382 * \return negative on errors.
2385 osc_send_oap_rpc(const struct lu_env *env, struct client_obd *cli,
2386 struct lov_oinfo *loi,
2387 int cmd, struct loi_oap_pages *lop)
2389 struct ptlrpc_request *req;
2390 obd_count page_count = 0;
2391 struct osc_async_page *oap = NULL, *tmp;
2392 struct osc_brw_async_args *aa;
2393 const struct obd_async_page_ops *ops;
2394 CFS_LIST_HEAD(rpc_list);
2395 CFS_LIST_HEAD(tmp_list);
2396 unsigned int ending_offset;
2397 unsigned starting_offset = 0;
2398 int srvlock = 0, mem_tight = 0;
2399 struct cl_object *clob = NULL;
2402 /* ASYNC_HP pages first. At present, when the lock the pages is
2403 * to be canceled, the pages covered by the lock will be sent out
2404 * with ASYNC_HP. We have to send out them as soon as possible. */
2405 cfs_list_for_each_entry_safe(oap, tmp, &lop->lop_urgent, oap_urgent_item) {
2406 if (oap->oap_async_flags & ASYNC_HP)
2407 cfs_list_move(&oap->oap_pending_item, &tmp_list);
2409 cfs_list_move_tail(&oap->oap_pending_item, &tmp_list);
2410 if (++page_count >= cli->cl_max_pages_per_rpc)
2414 cfs_list_splice(&tmp_list, &lop->lop_pending);
2417 /* first we find the pages we're allowed to work with */
2418 cfs_list_for_each_entry_safe(oap, tmp, &lop->lop_pending,
2420 ops = oap->oap_caller_ops;
2422 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2423 "magic 0x%x\n", oap, oap->oap_magic);
2426 /* pin object in memory, so that completion call-backs
2427 * can be safely called under client_obd_list lock. */
2428 clob = osc_oap2cl_page(oap)->cp_obj;
2429 cl_object_get(clob);
2432 if (page_count != 0 &&
2433 srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2434 CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2435 " oap %p, page %p, srvlock %u\n",
2436 oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2440 /* If there is a gap at the start of this page, it can't merge
2441 * with any previous page, so we'll hand the network a
2442 * "fragmented" page array that it can't transfer in 1 RDMA */
2443 if (page_count != 0 && oap->oap_page_off != 0)
2446 /* in llite being 'ready' equates to the page being locked
2447 * until completion unlocks it. commit_write submits a page
2448 * as not ready because its unlock will happen unconditionally
2449 * as the call returns. if we race with commit_write giving
2450 * us that page we don't want to create a hole in the page
2451 * stream, so we stop and leave the rpc to be fired by
2452 * another dirtier or kupdated interval (the not ready page
2453 * will still be on the dirty list). we could call in
2454 * at the end of ll_file_write to process the queue again. */
2455 if (!(oap->oap_async_flags & ASYNC_READY)) {
2456 int rc = ops->ap_make_ready(env, oap->oap_caller_data,
2459 CDEBUG(D_INODE, "oap %p page %p returned %d "
2460 "instead of ready\n", oap,
2464 /* llite is telling us that the page is still
2465 * in commit_write and that we should try
2466 * and put it in an rpc again later. we
2467 * break out of the loop so we don't create
2468 * a hole in the sequence of pages in the rpc
2473 /* the io isn't needed.. tell the checks
2474 * below to complete the rpc with EINTR */
2475 cfs_spin_lock(&oap->oap_lock);
2476 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2477 cfs_spin_unlock(&oap->oap_lock);
2478 oap->oap_count = -EINTR;
2481 cfs_spin_lock(&oap->oap_lock);
2482 oap->oap_async_flags |= ASYNC_READY;
2483 cfs_spin_unlock(&oap->oap_lock);
2486 LASSERTF(0, "oap %p page %p returned %d "
2487 "from make_ready\n", oap,
2495 * Page submitted for IO has to be locked. Either by
2496 * ->ap_make_ready() or by higher layers.
2498 #if defined(__KERNEL__) && defined(__linux__)
2500 struct cl_page *page;
2502 page = osc_oap2cl_page(oap);
2504 if (page->cp_type == CPT_CACHEABLE &&
2505 !(PageLocked(oap->oap_page) &&
2506 (CheckWriteback(oap->oap_page, cmd)))) {
2507 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2509 (long)oap->oap_page->flags,
2510 oap->oap_async_flags);
2516 /* take the page out of our book-keeping */
2517 cfs_list_del_init(&oap->oap_pending_item);
2518 lop_update_pending(cli, lop, cmd, -1);
2519 cfs_list_del_init(&oap->oap_urgent_item);
2521 if (page_count == 0)
2522 starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2523 (PTLRPC_MAX_BRW_SIZE - 1);
2525 /* ask the caller for the size of the io as the rpc leaves. */
2526 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
2528 ops->ap_refresh_count(env, oap->oap_caller_data,
2530 LASSERT(oap->oap_page_off + oap->oap_count <= CFS_PAGE_SIZE);
2532 if (oap->oap_count <= 0) {
2533 CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2535 osc_ap_completion(env, cli, NULL,
2536 oap, 0, oap->oap_count);
2540 /* now put the page back in our accounting */
2541 cfs_list_add_tail(&oap->oap_rpc_item, &rpc_list);
2542 if (oap->oap_brw_flags & OBD_BRW_MEMALLOC)
2544 if (page_count == 0)
2545 srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2546 if (++page_count >= cli->cl_max_pages_per_rpc)
2549 /* End on a PTLRPC_MAX_BRW_SIZE boundary. We want full-sized
2550 * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2551 * have the same alignment as the initial writes that allocated
2552 * extents on the server. */
2553 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2554 oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2555 if (ending_offset == 0)
2558 /* If there is a gap at the end of this page, it can't merge
2559 * with any subsequent pages, so we'll hand the network a
2560 * "fragmented" page array that it can't transfer in 1 RDMA */
2561 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2565 osc_wake_cache_waiters(cli);
2567 loi_list_maint(cli, loi);
2569 client_obd_list_unlock(&cli->cl_loi_list_lock);
2572 cl_object_put(env, clob);
2574 if (page_count == 0) {
2575 client_obd_list_lock(&cli->cl_loi_list_lock);
2579 req = osc_build_req(env, cli, &rpc_list, page_count,
2580 mem_tight ? (cmd | OBD_BRW_MEMALLOC) : cmd);
2582 LASSERT(cfs_list_empty(&rpc_list));
2583 loi_list_maint(cli, loi);
2584 RETURN(PTR_ERR(req));
2587 aa = ptlrpc_req_async_args(req);
2589 if (cmd == OBD_BRW_READ) {
2590 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2591 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2592 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2593 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2595 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2596 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2597 cli->cl_w_in_flight);
2598 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2599 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2601 ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2603 client_obd_list_lock(&cli->cl_loi_list_lock);
2605 if (cmd == OBD_BRW_READ)
2606 cli->cl_r_in_flight++;
2608 cli->cl_w_in_flight++;
2610 /* queued sync pages can be torn down while the pages
2611 * were between the pending list and the rpc */
2613 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2614 /* only one oap gets a request reference */
2617 if (oap->oap_interrupted && !req->rq_intr) {
2618 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2620 ptlrpc_mark_interrupted(req);
2624 tmp->oap_request = ptlrpc_request_addref(req);
2626 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2627 page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2629 req->rq_interpret_reply = brw_interpret;
2630 ptlrpcd_add_req(req, PSCOPE_BRW);
2634 #define LOI_DEBUG(LOI, STR, args...) \
2635 CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR, \
2636 !cfs_list_empty(&(LOI)->loi_ready_item) || \
2637 !cfs_list_empty(&(LOI)->loi_hp_ready_item), \
2638 (LOI)->loi_write_lop.lop_num_pending, \
2639 !cfs_list_empty(&(LOI)->loi_write_lop.lop_urgent), \
2640 (LOI)->loi_read_lop.lop_num_pending, \
2641 !cfs_list_empty(&(LOI)->loi_read_lop.lop_urgent), \
2644 /* This is called by osc_check_rpcs() to find which objects have pages that
2645 * we could be sending. These lists are maintained by lop_makes_rpc(). */
2646 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2650 /* First return objects that have blocked locks so that they
2651 * will be flushed quickly and other clients can get the lock,
2652 * then objects which have pages ready to be stuffed into RPCs */
2653 if (!cfs_list_empty(&cli->cl_loi_hp_ready_list))
2654 RETURN(cfs_list_entry(cli->cl_loi_hp_ready_list.next,
2655 struct lov_oinfo, loi_hp_ready_item));
2656 if (!cfs_list_empty(&cli->cl_loi_ready_list))
2657 RETURN(cfs_list_entry(cli->cl_loi_ready_list.next,
2658 struct lov_oinfo, loi_ready_item));
2660 /* then if we have cache waiters, return all objects with queued
2661 * writes. This is especially important when many small files
2662 * have filled up the cache and not been fired into rpcs because
2663 * they don't pass the nr_pending/object threshhold */
2664 if (!cfs_list_empty(&cli->cl_cache_waiters) &&
2665 !cfs_list_empty(&cli->cl_loi_write_list))
2666 RETURN(cfs_list_entry(cli->cl_loi_write_list.next,
2667 struct lov_oinfo, loi_write_item));
2669 /* then return all queued objects when we have an invalid import
2670 * so that they get flushed */
2671 if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2672 if (!cfs_list_empty(&cli->cl_loi_write_list))
2673 RETURN(cfs_list_entry(cli->cl_loi_write_list.next,
2676 if (!cfs_list_empty(&cli->cl_loi_read_list))
2677 RETURN(cfs_list_entry(cli->cl_loi_read_list.next,
2678 struct lov_oinfo, loi_read_item));
2683 static int osc_max_rpc_in_flight(struct client_obd *cli, struct lov_oinfo *loi)
2685 struct osc_async_page *oap;
2688 if (!cfs_list_empty(&loi->loi_write_lop.lop_urgent)) {
2689 oap = cfs_list_entry(loi->loi_write_lop.lop_urgent.next,
2690 struct osc_async_page, oap_urgent_item);
2691 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2694 if (!hprpc && !cfs_list_empty(&loi->loi_read_lop.lop_urgent)) {
2695 oap = cfs_list_entry(loi->loi_read_lop.lop_urgent.next,
2696 struct osc_async_page, oap_urgent_item);
2697 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2700 return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
2703 /* called with the loi list lock held */
2704 void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2706 struct lov_oinfo *loi;
2707 int rc = 0, race_counter = 0;
2710 while ((loi = osc_next_loi(cli)) != NULL) {
2711 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2713 if (osc_max_rpc_in_flight(cli, loi))
2716 /* attempt some read/write balancing by alternating between
2717 * reads and writes in an object. The makes_rpc checks here
2718 * would be redundant if we were getting read/write work items
2719 * instead of objects. we don't want send_oap_rpc to drain a
2720 * partial read pending queue when we're given this object to
2721 * do io on writes while there are cache waiters */
2722 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2723 rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_WRITE,
2724 &loi->loi_write_lop);
2726 CERROR("Write request failed with %d\n", rc);
2728 /* osc_send_oap_rpc failed, mostly because of
2731 * It can't break here, because if:
2732 * - a page was submitted by osc_io_submit, so
2734 * - no request in flight
2735 * - no subsequent request
2736 * The system will be in live-lock state,
2737 * because there is no chance to call
2738 * osc_io_unplug() and osc_check_rpcs() any
2739 * more. pdflush can't help in this case,
2740 * because it might be blocked at grabbing
2741 * the page lock as we mentioned.
2743 * Anyway, continue to drain pages. */
2752 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2753 rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_READ,
2754 &loi->loi_read_lop);
2756 CERROR("Read request failed with %d\n", rc);
2764 /* attempt some inter-object balancing by issuing rpcs
2765 * for each object in turn */
2766 if (!cfs_list_empty(&loi->loi_hp_ready_item))
2767 cfs_list_del_init(&loi->loi_hp_ready_item);
2768 if (!cfs_list_empty(&loi->loi_ready_item))
2769 cfs_list_del_init(&loi->loi_ready_item);
2770 if (!cfs_list_empty(&loi->loi_write_item))
2771 cfs_list_del_init(&loi->loi_write_item);
2772 if (!cfs_list_empty(&loi->loi_read_item))
2773 cfs_list_del_init(&loi->loi_read_item);
2775 loi_list_maint(cli, loi);
2777 /* send_oap_rpc fails with 0 when make_ready tells it to
2778 * back off. llite's make_ready does this when it tries
2779 * to lock a page queued for write that is already locked.
2780 * we want to try sending rpcs from many objects, but we
2781 * don't want to spin failing with 0. */
2782 if (race_counter == 10)
2788 /* we're trying to queue a page in the osc so we're subject to the
2789 * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2790 * If the osc's queued pages are already at that limit, then we want to sleep
2791 * until there is space in the osc's queue for us. We also may be waiting for
2792 * write credits from the OST if there are RPCs in flight that may return some
2793 * before we fall back to sync writes.
2795 * We need this know our allocation was granted in the presence of signals */
2796 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2800 client_obd_list_lock(&cli->cl_loi_list_lock);
2801 rc = cfs_list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2802 client_obd_list_unlock(&cli->cl_loi_list_lock);
2807 * Non-blocking version of osc_enter_cache() that consumes grant only when it
2810 int osc_enter_cache_try(const struct lu_env *env,
2811 struct client_obd *cli, struct lov_oinfo *loi,
2812 struct osc_async_page *oap, int transient)
2816 has_grant = cli->cl_avail_grant >= CFS_PAGE_SIZE;
2818 osc_consume_write_grant(cli, &oap->oap_brw_page);
2820 cli->cl_dirty_transit += CFS_PAGE_SIZE;
2821 cfs_atomic_inc(&obd_dirty_transit_pages);
2822 oap->oap_brw_flags |= OBD_BRW_NOCACHE;
2828 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2829 * grant or cache space. */
2830 static int osc_enter_cache(const struct lu_env *env,
2831 struct client_obd *cli, struct lov_oinfo *loi,
2832 struct osc_async_page *oap)
2834 struct osc_cache_waiter ocw;
2835 struct l_wait_info lwi = { 0 };
2839 CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2840 "grant: %lu\n", cli->cl_dirty, cfs_atomic_read(&obd_dirty_pages),
2841 cli->cl_dirty_max, obd_max_dirty_pages,
2842 cli->cl_lost_grant, cli->cl_avail_grant);
2844 /* force the caller to try sync io. this can jump the list
2845 * of queued writes and create a discontiguous rpc stream */
2846 if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2847 loi->loi_ar.ar_force_sync)
2850 /* Hopefully normal case - cache space and write credits available */
2851 if (cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max &&
2852 cfs_atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages &&
2853 osc_enter_cache_try(env, cli, loi, oap, 0))
2856 /* It is safe to block as a cache waiter as long as there is grant
2857 * space available or the hope of additional grant being returned
2858 * when an in flight write completes. Using the write back cache
2859 * if possible is preferable to sending the data synchronously
2860 * because write pages can then be merged in to large requests.
2861 * The addition of this cache waiter will causing pending write
2862 * pages to be sent immediately. */
2863 if (cli->cl_w_in_flight || cli->cl_avail_grant >= CFS_PAGE_SIZE) {
2864 cfs_list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2865 cfs_waitq_init(&ocw.ocw_waitq);
2869 loi_list_maint(cli, loi);
2870 osc_check_rpcs(env, cli);
2871 client_obd_list_unlock(&cli->cl_loi_list_lock);
2873 CDEBUG(D_CACHE, "sleeping for cache space\n");
2874 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2876 client_obd_list_lock(&cli->cl_loi_list_lock);
2877 if (!cfs_list_empty(&ocw.ocw_entry)) {
2878 cfs_list_del(&ocw.ocw_entry);
2888 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2889 struct lov_oinfo *loi, cfs_page_t *page,
2890 obd_off offset, const struct obd_async_page_ops *ops,
2891 void *data, void **res, int nocache,
2892 struct lustre_handle *lockh)
2894 struct osc_async_page *oap;
2899 return cfs_size_round(sizeof(*oap));
2902 oap->oap_magic = OAP_MAGIC;
2903 oap->oap_cli = &exp->exp_obd->u.cli;
2906 oap->oap_caller_ops = ops;
2907 oap->oap_caller_data = data;
2909 oap->oap_page = page;
2910 oap->oap_obj_off = offset;
2911 if (!client_is_remote(exp) &&
2912 cfs_capable(CFS_CAP_SYS_RESOURCE))
2913 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2915 LASSERT(!(offset & ~CFS_PAGE_MASK));
2917 CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2918 CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2919 CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2920 CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2922 cfs_spin_lock_init(&oap->oap_lock);
2923 CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
2927 struct osc_async_page *oap_from_cookie(void *cookie)
2929 struct osc_async_page *oap = cookie;
2930 if (oap->oap_magic != OAP_MAGIC)
2931 return ERR_PTR(-EINVAL);
2935 int osc_queue_async_io(const struct lu_env *env,
2936 struct obd_export *exp, struct lov_stripe_md *lsm,
2937 struct lov_oinfo *loi, void *cookie,
2938 int cmd, obd_off off, int count,
2939 obd_flag brw_flags, enum async_flags async_flags)
2941 struct client_obd *cli = &exp->exp_obd->u.cli;
2942 struct osc_async_page *oap;
2946 oap = oap_from_cookie(cookie);
2948 RETURN(PTR_ERR(oap));
2950 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2953 if (!cfs_list_empty(&oap->oap_pending_item) ||
2954 !cfs_list_empty(&oap->oap_urgent_item) ||
2955 !cfs_list_empty(&oap->oap_rpc_item))
2958 /* check if the file's owner/group is over quota */
2959 if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)) {
2960 struct cl_object *obj;
2961 struct cl_attr attr; /* XXX put attr into thread info */
2962 unsigned int qid[MAXQUOTAS];
2964 obj = cl_object_top(osc_oap2cl_page(oap)->cp_obj);
2966 cl_object_attr_lock(obj);
2967 rc = cl_object_attr_get(env, obj, &attr);
2968 cl_object_attr_unlock(obj);
2970 qid[USRQUOTA] = attr.cat_uid;
2971 qid[GRPQUOTA] = attr.cat_gid;
2973 lquota_chkdq(quota_interface, cli, qid) == NO_QUOTA)
2980 loi = lsm->lsm_oinfo[0];
2982 client_obd_list_lock(&cli->cl_loi_list_lock);
2984 LASSERT(off + count <= CFS_PAGE_SIZE);
2986 oap->oap_page_off = off;
2987 oap->oap_count = count;
2988 oap->oap_brw_flags = brw_flags;
2989 /* Give a hint to OST that requests are coming from kswapd - bug19529 */
2990 if (cfs_memory_pressure_get())
2991 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
2992 cfs_spin_lock(&oap->oap_lock);
2993 oap->oap_async_flags = async_flags;
2994 cfs_spin_unlock(&oap->oap_lock);
2996 if (cmd & OBD_BRW_WRITE) {
2997 rc = osc_enter_cache(env, cli, loi, oap);
2999 client_obd_list_unlock(&cli->cl_loi_list_lock);
3004 osc_oap_to_pending(oap);
3005 loi_list_maint(cli, loi);
3007 LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
3010 osc_check_rpcs(env, cli);
3011 client_obd_list_unlock(&cli->cl_loi_list_lock);
3016 /* aka (~was & now & flag), but this is more clear :) */
3017 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
3019 int osc_set_async_flags_base(struct client_obd *cli,
3020 struct lov_oinfo *loi, struct osc_async_page *oap,
3021 obd_flag async_flags)
3023 struct loi_oap_pages *lop;
3027 LASSERT(!cfs_list_empty(&oap->oap_pending_item));
3029 if (oap->oap_cmd & OBD_BRW_WRITE) {
3030 lop = &loi->loi_write_lop;
3032 lop = &loi->loi_read_lop;
3035 if ((oap->oap_async_flags & async_flags) == async_flags)
3038 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
3039 flags |= ASYNC_READY;
3041 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT) &&
3042 cfs_list_empty(&oap->oap_rpc_item)) {
3043 if (oap->oap_async_flags & ASYNC_HP)
3044 cfs_list_add(&oap->oap_urgent_item, &lop->lop_urgent);
3046 cfs_list_add_tail(&oap->oap_urgent_item,
3048 flags |= ASYNC_URGENT;
3049 loi_list_maint(cli, loi);
3051 cfs_spin_lock(&oap->oap_lock);
3052 oap->oap_async_flags |= flags;
3053 cfs_spin_unlock(&oap->oap_lock);
3055 LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
3056 oap->oap_async_flags);
3060 int osc_teardown_async_page(struct obd_export *exp,
3061 struct lov_stripe_md *lsm,
3062 struct lov_oinfo *loi, void *cookie)
3064 struct client_obd *cli = &exp->exp_obd->u.cli;
3065 struct loi_oap_pages *lop;
3066 struct osc_async_page *oap;
3070 oap = oap_from_cookie(cookie);
3072 RETURN(PTR_ERR(oap));
3075 loi = lsm->lsm_oinfo[0];
3077 if (oap->oap_cmd & OBD_BRW_WRITE) {
3078 lop = &loi->loi_write_lop;
3080 lop = &loi->loi_read_lop;
3083 client_obd_list_lock(&cli->cl_loi_list_lock);
3085 if (!cfs_list_empty(&oap->oap_rpc_item))
3086 GOTO(out, rc = -EBUSY);
3088 osc_exit_cache(cli, oap, 0);
3089 osc_wake_cache_waiters(cli);
3091 if (!cfs_list_empty(&oap->oap_urgent_item)) {
3092 cfs_list_del_init(&oap->oap_urgent_item);
3093 cfs_spin_lock(&oap->oap_lock);
3094 oap->oap_async_flags &= ~(ASYNC_URGENT | ASYNC_HP);
3095 cfs_spin_unlock(&oap->oap_lock);
3097 if (!cfs_list_empty(&oap->oap_pending_item)) {
3098 cfs_list_del_init(&oap->oap_pending_item);
3099 lop_update_pending(cli, lop, oap->oap_cmd, -1);
3101 loi_list_maint(cli, loi);
3102 LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
3104 client_obd_list_unlock(&cli->cl_loi_list_lock);
3108 static void osc_set_lock_data_with_check(struct ldlm_lock *lock,
3109 struct ldlm_enqueue_info *einfo,
3112 void *data = einfo->ei_cbdata;
3114 LASSERT(lock != NULL);
3115 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
3116 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
3117 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
3118 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
3120 lock_res_and_lock(lock);
3121 cfs_spin_lock(&osc_ast_guard);
3122 LASSERT(lock->l_ast_data == NULL || lock->l_ast_data == data);
3123 lock->l_ast_data = data;
3124 cfs_spin_unlock(&osc_ast_guard);
3125 unlock_res_and_lock(lock);
3128 static void osc_set_data_with_check(struct lustre_handle *lockh,
3129 struct ldlm_enqueue_info *einfo,
3132 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
3135 osc_set_lock_data_with_check(lock, einfo, flags);
3136 LDLM_LOCK_PUT(lock);
3138 CERROR("lockh %p, data %p - client evicted?\n",
3139 lockh, einfo->ei_cbdata);
3142 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3143 ldlm_iterator_t replace, void *data)
3145 struct ldlm_res_id res_id;
3146 struct obd_device *obd = class_exp2obd(exp);
3148 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
3149 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3153 /* find any ldlm lock of the inode in osc
3157 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3158 ldlm_iterator_t replace, void *data)
3160 struct ldlm_res_id res_id;
3161 struct obd_device *obd = class_exp2obd(exp);
3164 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
3165 rc = ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3166 if (rc == LDLM_ITER_STOP)
3168 if (rc == LDLM_ITER_CONTINUE)
3173 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
3174 obd_enqueue_update_f upcall, void *cookie,
3177 int intent = *flags & LDLM_FL_HAS_INTENT;
3181 /* The request was created before ldlm_cli_enqueue call. */
3182 if (rc == ELDLM_LOCK_ABORTED) {
3183 struct ldlm_reply *rep;
3184 rep = req_capsule_server_get(&req->rq_pill,
3187 LASSERT(rep != NULL);
3188 if (rep->lock_policy_res1)
3189 rc = rep->lock_policy_res1;
3193 if ((intent && rc == ELDLM_LOCK_ABORTED) || !rc) {
3194 *flags |= LDLM_FL_LVB_READY;
3195 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
3196 lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
3199 /* Call the update callback. */
3200 rc = (*upcall)(cookie, rc);
3204 static int osc_enqueue_interpret(const struct lu_env *env,
3205 struct ptlrpc_request *req,
3206 struct osc_enqueue_args *aa, int rc)
3208 struct ldlm_lock *lock;
3209 struct lustre_handle handle;
3212 /* Make a local copy of a lock handle and a mode, because aa->oa_*
3213 * might be freed anytime after lock upcall has been called. */
3214 lustre_handle_copy(&handle, aa->oa_lockh);
3215 mode = aa->oa_ei->ei_mode;
3217 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
3219 lock = ldlm_handle2lock(&handle);
3221 /* Take an additional reference so that a blocking AST that
3222 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
3223 * to arrive after an upcall has been executed by
3224 * osc_enqueue_fini(). */
3225 ldlm_lock_addref(&handle, mode);
3227 /* Let CP AST to grant the lock first. */
3228 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
3230 /* Complete obtaining the lock procedure. */
3231 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
3232 mode, aa->oa_flags, aa->oa_lvb,
3233 sizeof(*aa->oa_lvb), &handle, rc);
3234 /* Complete osc stuff. */
3235 rc = osc_enqueue_fini(req, aa->oa_lvb,
3236 aa->oa_upcall, aa->oa_cookie, aa->oa_flags, rc);
3238 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
3240 /* Release the lock for async request. */
3241 if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
3243 * Releases a reference taken by ldlm_cli_enqueue(), if it is
3244 * not already released by
3245 * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
3247 ldlm_lock_decref(&handle, mode);
3249 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
3250 aa->oa_lockh, req, aa);
3251 ldlm_lock_decref(&handle, mode);
3252 LDLM_LOCK_PUT(lock);
3256 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
3257 struct lov_oinfo *loi, int flags,
3258 struct ost_lvb *lvb, __u32 mode, int rc)
3260 if (rc == ELDLM_OK) {
3261 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
3264 LASSERT(lock != NULL);
3265 loi->loi_lvb = *lvb;
3266 tmp = loi->loi_lvb.lvb_size;
3267 /* Extend KMS up to the end of this lock and no further
3268 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
3269 if (tmp > lock->l_policy_data.l_extent.end)
3270 tmp = lock->l_policy_data.l_extent.end + 1;
3271 if (tmp >= loi->loi_kms) {
3272 LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
3273 ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
3274 loi_kms_set(loi, tmp);
3276 LDLM_DEBUG(lock, "lock acquired, setting rss="
3277 LPU64"; leaving kms="LPU64", end="LPU64,
3278 loi->loi_lvb.lvb_size, loi->loi_kms,
3279 lock->l_policy_data.l_extent.end);
3281 ldlm_lock_allow_match(lock);
3282 LDLM_LOCK_PUT(lock);
3283 } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
3284 loi->loi_lvb = *lvb;
3285 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
3286 " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
3290 EXPORT_SYMBOL(osc_update_enqueue);
3292 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
3294 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
3295 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
3296 * other synchronous requests, however keeping some locks and trying to obtain
3297 * others may take a considerable amount of time in a case of ost failure; and
3298 * when other sync requests do not get released lock from a client, the client
3299 * is excluded from the cluster -- such scenarious make the life difficult, so
3300 * release locks just after they are obtained. */
3301 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3302 int *flags, ldlm_policy_data_t *policy,
3303 struct ost_lvb *lvb, int kms_valid,
3304 obd_enqueue_update_f upcall, void *cookie,
3305 struct ldlm_enqueue_info *einfo,
3306 struct lustre_handle *lockh,
3307 struct ptlrpc_request_set *rqset, int async)
3309 struct obd_device *obd = exp->exp_obd;
3310 struct ptlrpc_request *req = NULL;
3311 int intent = *flags & LDLM_FL_HAS_INTENT;
3316 /* Filesystem lock extents are extended to page boundaries so that
3317 * dealing with the page cache is a little smoother. */
3318 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3319 policy->l_extent.end |= ~CFS_PAGE_MASK;
3322 * kms is not valid when either object is completely fresh (so that no
3323 * locks are cached), or object was evicted. In the latter case cached
3324 * lock cannot be used, because it would prime inode state with
3325 * potentially stale LVB.
3330 /* Next, search for already existing extent locks that will cover us */
3331 /* If we're trying to read, we also search for an existing PW lock. The
3332 * VFS and page cache already protect us locally, so lots of readers/
3333 * writers can share a single PW lock.
3335 * There are problems with conversion deadlocks, so instead of
3336 * converting a read lock to a write lock, we'll just enqueue a new
3339 * At some point we should cancel the read lock instead of making them
3340 * send us a blocking callback, but there are problems with canceling
3341 * locks out from other users right now, too. */
3342 mode = einfo->ei_mode;
3343 if (einfo->ei_mode == LCK_PR)
3345 mode = ldlm_lock_match(obd->obd_namespace,
3346 *flags | LDLM_FL_LVB_READY, res_id,
3347 einfo->ei_type, policy, mode, lockh, 0);
3349 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
3351 if (matched->l_ast_data == NULL ||
3352 matched->l_ast_data == einfo->ei_cbdata) {
3353 /* addref the lock only if not async requests and PW
3354 * lock is matched whereas we asked for PR. */
3355 if (!rqset && einfo->ei_mode != mode)
3356 ldlm_lock_addref(lockh, LCK_PR);
3357 osc_set_lock_data_with_check(matched, einfo, *flags);
3359 /* I would like to be able to ASSERT here that
3360 * rss <= kms, but I can't, for reasons which
3361 * are explained in lov_enqueue() */
3364 /* We already have a lock, and it's referenced */
3365 (*upcall)(cookie, ELDLM_OK);
3367 /* For async requests, decref the lock. */
3368 if (einfo->ei_mode != mode)
3369 ldlm_lock_decref(lockh, LCK_PW);
3371 ldlm_lock_decref(lockh, einfo->ei_mode);
3372 LDLM_LOCK_PUT(matched);
3375 ldlm_lock_decref(lockh, mode);
3376 LDLM_LOCK_PUT(matched);
3381 CFS_LIST_HEAD(cancels);
3382 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3383 &RQF_LDLM_ENQUEUE_LVB);
3387 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
3389 ptlrpc_request_free(req);
3393 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
3395 ptlrpc_request_set_replen(req);
3398 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
3399 *flags &= ~LDLM_FL_BLOCK_GRANTED;
3401 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
3402 sizeof(*lvb), lockh, async);
3405 struct osc_enqueue_args *aa;
3406 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3407 aa = ptlrpc_req_async_args(req);
3410 aa->oa_flags = flags;
3411 aa->oa_upcall = upcall;
3412 aa->oa_cookie = cookie;
3414 aa->oa_lockh = lockh;
3416 req->rq_interpret_reply =
3417 (ptlrpc_interpterer_t)osc_enqueue_interpret;
3418 if (rqset == PTLRPCD_SET)
3419 ptlrpcd_add_req(req, PSCOPE_OTHER);
3421 ptlrpc_set_add_req(rqset, req);
3422 } else if (intent) {
3423 ptlrpc_req_finished(req);
3428 rc = osc_enqueue_fini(req, lvb, upcall, cookie, flags, rc);
3430 ptlrpc_req_finished(req);
3435 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
3436 struct ldlm_enqueue_info *einfo,
3437 struct ptlrpc_request_set *rqset)
3439 struct ldlm_res_id res_id;
3443 osc_build_res_name(oinfo->oi_md->lsm_object_id,
3444 oinfo->oi_md->lsm_object_seq, &res_id);
3446 rc = osc_enqueue_base(exp, &res_id, &oinfo->oi_flags, &oinfo->oi_policy,
3447 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
3448 oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid,
3449 oinfo->oi_cb_up, oinfo, einfo, oinfo->oi_lockh,
3450 rqset, rqset != NULL);
3454 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3455 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
3456 int *flags, void *data, struct lustre_handle *lockh,
3459 struct obd_device *obd = exp->exp_obd;
3460 int lflags = *flags;
3464 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
3467 /* Filesystem lock extents are extended to page boundaries so that
3468 * dealing with the page cache is a little smoother */
3469 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3470 policy->l_extent.end |= ~CFS_PAGE_MASK;
3472 /* Next, search for already existing extent locks that will cover us */
3473 /* If we're trying to read, we also search for an existing PW lock. The
3474 * VFS and page cache already protect us locally, so lots of readers/
3475 * writers can share a single PW lock. */
3479 rc = ldlm_lock_match(obd->obd_namespace, lflags,
3480 res_id, type, policy, rc, lockh, unref);
3483 osc_set_data_with_check(lockh, data, lflags);
3484 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
3485 ldlm_lock_addref(lockh, LCK_PR);
3486 ldlm_lock_decref(lockh, LCK_PW);
3493 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
3497 if (unlikely(mode == LCK_GROUP))
3498 ldlm_lock_decref_and_cancel(lockh, mode);
3500 ldlm_lock_decref(lockh, mode);
3505 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
3506 __u32 mode, struct lustre_handle *lockh)
3509 RETURN(osc_cancel_base(lockh, mode));
3512 static int osc_cancel_unused(struct obd_export *exp,
3513 struct lov_stripe_md *lsm,
3514 ldlm_cancel_flags_t flags,
3517 struct obd_device *obd = class_exp2obd(exp);
3518 struct ldlm_res_id res_id, *resp = NULL;
3521 resp = osc_build_res_name(lsm->lsm_object_id,
3522 lsm->lsm_object_seq, &res_id);
3525 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
3528 static int osc_statfs_interpret(const struct lu_env *env,
3529 struct ptlrpc_request *req,
3530 struct osc_async_args *aa, int rc)
3532 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
3533 struct obd_statfs *msfs;
3538 /* The request has in fact never been sent
3539 * due to issues at a higher level (LOV).
3540 * Exit immediately since the caller is
3541 * aware of the problem and takes care
3542 * of the clean up */
3545 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
3546 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
3552 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3554 GOTO(out, rc = -EPROTO);
3557 /* Reinitialize the RDONLY and DEGRADED flags at the client
3558 * on each statfs, so they don't stay set permanently. */
3559 cfs_spin_lock(&cli->cl_oscc.oscc_lock);
3561 if (unlikely(msfs->os_state & OS_STATE_DEGRADED))
3562 cli->cl_oscc.oscc_flags |= OSCC_FLAG_DEGRADED;
3563 else if (unlikely(cli->cl_oscc.oscc_flags & OSCC_FLAG_DEGRADED))
3564 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_DEGRADED;
3566 if (unlikely(msfs->os_state & OS_STATE_READONLY))
3567 cli->cl_oscc.oscc_flags |= OSCC_FLAG_RDONLY;
3568 else if (unlikely(cli->cl_oscc.oscc_flags & OSCC_FLAG_RDONLY))
3569 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_RDONLY;
3571 /* Add a bit of hysteresis so this flag isn't continually flapping,
3572 * and ensure that new files don't get extremely fragmented due to
3573 * only a small amount of available space in the filesystem.
3574 * We want to set the NOSPC flag when there is less than ~0.1% free
3575 * and clear it when there is at least ~0.2% free space, so:
3576 * avail < ~0.1% max max = avail + used
3577 * 1025 * avail < avail + used used = blocks - free
3578 * 1024 * avail < used
3579 * 1024 * avail < blocks - free
3580 * avail < ((blocks - free) >> 10)
3582 * On very large disk, say 16TB 0.1% will be 16 GB. We don't want to
3583 * lose that amount of space so in those cases we report no space left
3584 * if their is less than 1 GB left. */
3585 used = min_t(__u64,(msfs->os_blocks - msfs->os_bfree) >> 10, 1 << 30);
3586 if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) == 0) &&
3587 ((msfs->os_ffree < 32) || (msfs->os_bavail < used))))
3588 cli->cl_oscc.oscc_flags |= OSCC_FLAG_NOSPC;
3589 else if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) != 0) &&
3590 (msfs->os_ffree > 64) && (msfs->os_bavail > (used << 1))))
3591 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_NOSPC;
3593 cfs_spin_unlock(&cli->cl_oscc.oscc_lock);
3595 *aa->aa_oi->oi_osfs = *msfs;
3597 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
3601 static int osc_statfs_async(struct obd_device *obd, struct obd_info *oinfo,
3602 __u64 max_age, struct ptlrpc_request_set *rqset)
3604 struct ptlrpc_request *req;
3605 struct osc_async_args *aa;
3609 /* We could possibly pass max_age in the request (as an absolute
3610 * timestamp or a "seconds.usec ago") so the target can avoid doing
3611 * extra calls into the filesystem if that isn't necessary (e.g.
3612 * during mount that would help a bit). Having relative timestamps
3613 * is not so great if request processing is slow, while absolute
3614 * timestamps are not ideal because they need time synchronization. */
3615 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
3619 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3621 ptlrpc_request_free(req);
3624 ptlrpc_request_set_replen(req);
3625 req->rq_request_portal = OST_CREATE_PORTAL;
3626 ptlrpc_at_set_req_timeout(req);
3628 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
3629 /* procfs requests not want stat in wait for avoid deadlock */
3630 req->rq_no_resend = 1;
3631 req->rq_no_delay = 1;
3634 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
3635 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3636 aa = ptlrpc_req_async_args(req);
3639 ptlrpc_set_add_req(rqset, req);
3643 static int osc_statfs(struct obd_device *obd, struct obd_statfs *osfs,
3644 __u64 max_age, __u32 flags)
3646 struct obd_statfs *msfs;
3647 struct ptlrpc_request *req;
3648 struct obd_import *imp = NULL;
3652 /*Since the request might also come from lprocfs, so we need
3653 *sync this with client_disconnect_export Bug15684*/
3654 cfs_down_read(&obd->u.cli.cl_sem);
3655 if (obd->u.cli.cl_import)
3656 imp = class_import_get(obd->u.cli.cl_import);
3657 cfs_up_read(&obd->u.cli.cl_sem);
3661 /* We could possibly pass max_age in the request (as an absolute
3662 * timestamp or a "seconds.usec ago") so the target can avoid doing
3663 * extra calls into the filesystem if that isn't necessary (e.g.
3664 * during mount that would help a bit). Having relative timestamps
3665 * is not so great if request processing is slow, while absolute
3666 * timestamps are not ideal because they need time synchronization. */
3667 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
3669 class_import_put(imp);
3674 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3676 ptlrpc_request_free(req);
3679 ptlrpc_request_set_replen(req);
3680 req->rq_request_portal = OST_CREATE_PORTAL;
3681 ptlrpc_at_set_req_timeout(req);
3683 if (flags & OBD_STATFS_NODELAY) {
3684 /* procfs requests not want stat in wait for avoid deadlock */
3685 req->rq_no_resend = 1;
3686 req->rq_no_delay = 1;
3689 rc = ptlrpc_queue_wait(req);
3693 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3695 GOTO(out, rc = -EPROTO);
3702 ptlrpc_req_finished(req);
3706 /* Retrieve object striping information.
3708 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
3709 * the maximum number of OST indices which will fit in the user buffer.
3710 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
3712 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
3714 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
3715 struct lov_user_md_v3 lum, *lumk;
3716 struct lov_user_ost_data_v1 *lmm_objects;
3717 int rc = 0, lum_size;
3723 /* we only need the header part from user space to get lmm_magic and
3724 * lmm_stripe_count, (the header part is common to v1 and v3) */
3725 lum_size = sizeof(struct lov_user_md_v1);
3726 if (cfs_copy_from_user(&lum, lump, lum_size))
3729 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
3730 (lum.lmm_magic != LOV_USER_MAGIC_V3))
3733 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
3734 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
3735 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
3736 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
3738 /* we can use lov_mds_md_size() to compute lum_size
3739 * because lov_user_md_vX and lov_mds_md_vX have the same size */
3740 if (lum.lmm_stripe_count > 0) {
3741 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
3742 OBD_ALLOC(lumk, lum_size);
3746 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
3747 lmm_objects = &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
3749 lmm_objects = &(lumk->lmm_objects[0]);
3750 lmm_objects->l_object_id = lsm->lsm_object_id;
3752 lum_size = lov_mds_md_size(0, lum.lmm_magic);
3756 lumk->lmm_object_id = lsm->lsm_object_id;
3757 lumk->lmm_object_seq = lsm->lsm_object_seq;
3758 lumk->lmm_stripe_count = 1;
3760 if (cfs_copy_to_user(lump, lumk, lum_size))
3764 OBD_FREE(lumk, lum_size);
3770 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
3771 void *karg, void *uarg)
3773 struct obd_device *obd = exp->exp_obd;
3774 struct obd_ioctl_data *data = karg;
3778 if (!cfs_try_module_get(THIS_MODULE)) {
3779 CERROR("Can't get module. Is it alive?");
3783 case OBD_IOC_LOV_GET_CONFIG: {
3785 struct lov_desc *desc;
3786 struct obd_uuid uuid;
3790 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
3791 GOTO(out, err = -EINVAL);
3793 data = (struct obd_ioctl_data *)buf;
3795 if (sizeof(*desc) > data->ioc_inllen1) {
3796 obd_ioctl_freedata(buf, len);
3797 GOTO(out, err = -EINVAL);
3800 if (data->ioc_inllen2 < sizeof(uuid)) {
3801 obd_ioctl_freedata(buf, len);
3802 GOTO(out, err = -EINVAL);
3805 desc = (struct lov_desc *)data->ioc_inlbuf1;
3806 desc->ld_tgt_count = 1;
3807 desc->ld_active_tgt_count = 1;
3808 desc->ld_default_stripe_count = 1;
3809 desc->ld_default_stripe_size = 0;
3810 desc->ld_default_stripe_offset = 0;
3811 desc->ld_pattern = 0;
3812 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
3814 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
3816 err = cfs_copy_to_user((void *)uarg, buf, len);
3819 obd_ioctl_freedata(buf, len);
3822 case LL_IOC_LOV_SETSTRIPE:
3823 err = obd_alloc_memmd(exp, karg);
3827 case LL_IOC_LOV_GETSTRIPE:
3828 err = osc_getstripe(karg, uarg);
3830 case OBD_IOC_CLIENT_RECOVER:
3831 err = ptlrpc_recover_import(obd->u.cli.cl_import,
3836 case IOC_OSC_SET_ACTIVE:
3837 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
3840 case OBD_IOC_POLL_QUOTACHECK:
3841 err = lquota_poll_check(quota_interface, exp,
3842 (struct if_quotacheck *)karg);
3844 case OBD_IOC_PING_TARGET:
3845 err = ptlrpc_obd_ping(obd);
3848 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
3849 cmd, cfs_curproc_comm());
3850 GOTO(out, err = -ENOTTY);
3853 cfs_module_put(THIS_MODULE);
3857 static int osc_get_info(struct obd_export *exp, obd_count keylen,
3858 void *key, __u32 *vallen, void *val,
3859 struct lov_stripe_md *lsm)
3862 if (!vallen || !val)
3865 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
3866 __u32 *stripe = val;
3867 *vallen = sizeof(*stripe);
3870 } else if (KEY_IS(KEY_LAST_ID)) {
3871 struct ptlrpc_request *req;
3876 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3877 &RQF_OST_GET_INFO_LAST_ID);
3881 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3882 RCL_CLIENT, keylen);
3883 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3885 ptlrpc_request_free(req);
3889 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3890 memcpy(tmp, key, keylen);
3892 req->rq_no_delay = req->rq_no_resend = 1;
3893 ptlrpc_request_set_replen(req);
3894 rc = ptlrpc_queue_wait(req);
3898 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
3900 GOTO(out, rc = -EPROTO);
3902 *((obd_id *)val) = *reply;
3904 ptlrpc_req_finished(req);
3906 } else if (KEY_IS(KEY_FIEMAP)) {
3907 struct ptlrpc_request *req;
3908 struct ll_user_fiemap *reply;
3912 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3913 &RQF_OST_GET_INFO_FIEMAP);
3917 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
3918 RCL_CLIENT, keylen);
3919 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3920 RCL_CLIENT, *vallen);
3921 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3922 RCL_SERVER, *vallen);
3924 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3926 ptlrpc_request_free(req);
3930 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
3931 memcpy(tmp, key, keylen);
3932 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3933 memcpy(tmp, val, *vallen);
3935 ptlrpc_request_set_replen(req);
3936 rc = ptlrpc_queue_wait(req);
3940 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3942 GOTO(out1, rc = -EPROTO);
3944 memcpy(val, reply, *vallen);
3946 ptlrpc_req_finished(req);
3954 static int osc_setinfo_mds_connect_import(struct obd_import *imp)
3956 struct llog_ctxt *ctxt;
3960 ctxt = llog_get_context(imp->imp_obd, LLOG_MDS_OST_ORIG_CTXT);
3962 rc = llog_initiator_connect(ctxt);
3963 llog_ctxt_put(ctxt);
3965 /* XXX return an error? skip setting below flags? */
3968 cfs_spin_lock(&imp->imp_lock);
3969 imp->imp_server_timeout = 1;
3970 imp->imp_pingable = 1;
3971 cfs_spin_unlock(&imp->imp_lock);
3972 CDEBUG(D_RPCTRACE, "pinging OST %s\n", obd2cli_tgt(imp->imp_obd));
3977 static int osc_setinfo_mds_conn_interpret(const struct lu_env *env,
3978 struct ptlrpc_request *req,
3985 RETURN(osc_setinfo_mds_connect_import(req->rq_import));
3988 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
3989 void *key, obd_count vallen, void *val,
3990 struct ptlrpc_request_set *set)
3992 struct ptlrpc_request *req;
3993 struct obd_device *obd = exp->exp_obd;
3994 struct obd_import *imp = class_exp2cliimp(exp);
3999 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
4001 if (KEY_IS(KEY_NEXT_ID)) {
4003 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4005 if (vallen != sizeof(obd_id))
4010 if (vallen != sizeof(obd_id))
4013 /* avoid race between allocate new object and set next id
4014 * from ll_sync thread */
4015 cfs_spin_lock(&oscc->oscc_lock);
4016 new_val = *((obd_id*)val) + 1;
4017 if (new_val > oscc->oscc_next_id)
4018 oscc->oscc_next_id = new_val;
4019 cfs_spin_unlock(&oscc->oscc_lock);
4020 CDEBUG(D_HA, "%s: set oscc_next_id = "LPU64"\n",
4021 exp->exp_obd->obd_name,
4022 obd->u.cli.cl_oscc.oscc_next_id);
4027 if (KEY_IS(KEY_CHECKSUM)) {
4028 if (vallen != sizeof(int))
4030 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
4034 if (KEY_IS(KEY_SPTLRPC_CONF)) {
4035 sptlrpc_conf_client_adapt(obd);
4039 if (KEY_IS(KEY_FLUSH_CTX)) {
4040 sptlrpc_import_flush_my_ctx(imp);
4044 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
4047 /* We pass all other commands directly to OST. Since nobody calls osc
4048 methods directly and everybody is supposed to go through LOV, we
4049 assume lov checked invalid values for us.
4050 The only recognised values so far are evict_by_nid and mds_conn.
4051 Even if something bad goes through, we'd get a -EINVAL from OST
4054 if (KEY_IS(KEY_GRANT_SHRINK))
4055 req = ptlrpc_request_alloc(imp, &RQF_OST_SET_GRANT_INFO);
4057 req = ptlrpc_request_alloc(imp, &RQF_OBD_SET_INFO);
4062 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
4063 RCL_CLIENT, keylen);
4064 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
4065 RCL_CLIENT, vallen);
4066 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
4068 ptlrpc_request_free(req);
4072 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
4073 memcpy(tmp, key, keylen);
4074 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_VAL);
4075 memcpy(tmp, val, vallen);
4077 if (KEY_IS(KEY_MDS_CONN)) {
4078 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4080 oscc->oscc_oa.o_seq = (*(__u32 *)val);
4081 oscc->oscc_oa.o_valid |= OBD_MD_FLGROUP;
4082 LASSERT_SEQ_IS_MDT(oscc->oscc_oa.o_seq);
4083 req->rq_no_delay = req->rq_no_resend = 1;
4084 req->rq_interpret_reply = osc_setinfo_mds_conn_interpret;
4085 } else if (KEY_IS(KEY_GRANT_SHRINK)) {
4086 struct osc_grant_args *aa;
4089 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
4090 aa = ptlrpc_req_async_args(req);
4093 ptlrpc_req_finished(req);
4096 *oa = ((struct ost_body *)val)->oa;
4098 req->rq_interpret_reply = osc_shrink_grant_interpret;
4101 ptlrpc_request_set_replen(req);
4102 if (!KEY_IS(KEY_GRANT_SHRINK)) {
4103 LASSERT(set != NULL);
4104 ptlrpc_set_add_req(set, req);
4105 ptlrpc_check_set(NULL, set);
4107 ptlrpcd_add_req(req, PSCOPE_OTHER);
4113 static struct llog_operations osc_size_repl_logops = {
4114 lop_cancel: llog_obd_repl_cancel
4117 static struct llog_operations osc_mds_ost_orig_logops;
4119 static int __osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
4120 struct obd_device *tgt, struct llog_catid *catid)
4125 rc = llog_setup(obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT, tgt, 1,
4126 &catid->lci_logid, &osc_mds_ost_orig_logops);
4128 CERROR("failed LLOG_MDS_OST_ORIG_CTXT\n");
4132 rc = llog_setup(obd, &obd->obd_olg, LLOG_SIZE_REPL_CTXT, tgt, 1,
4133 NULL, &osc_size_repl_logops);
4135 struct llog_ctxt *ctxt =
4136 llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
4139 CERROR("failed LLOG_SIZE_REPL_CTXT\n");
4144 CERROR("osc '%s' tgt '%s' catid %p rc=%d\n",
4145 obd->obd_name, tgt->obd_name, catid, rc);
4146 CERROR("logid "LPX64":0x%x\n",
4147 catid->lci_logid.lgl_oid, catid->lci_logid.lgl_ogen);
4152 static int osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
4153 struct obd_device *disk_obd, int *index)
4155 struct llog_catid catid;
4156 static char name[32] = CATLIST;
4160 LASSERT(olg == &obd->obd_olg);
4162 cfs_mutex_down(&olg->olg_cat_processing);
4163 rc = llog_get_cat_list(disk_obd, name, *index, 1, &catid);
4165 CERROR("rc: %d\n", rc);
4169 CDEBUG(D_INFO, "%s: Init llog for %d - catid "LPX64"/"LPX64":%x\n",
4170 obd->obd_name, *index, catid.lci_logid.lgl_oid,
4171 catid.lci_logid.lgl_oseq, catid.lci_logid.lgl_ogen);
4173 rc = __osc_llog_init(obd, olg, disk_obd, &catid);
4175 CERROR("rc: %d\n", rc);
4179 rc = llog_put_cat_list(disk_obd, name, *index, 1, &catid);
4181 CERROR("rc: %d\n", rc);
4186 cfs_mutex_up(&olg->olg_cat_processing);
4191 static int osc_llog_finish(struct obd_device *obd, int count)
4193 struct llog_ctxt *ctxt;
4194 int rc = 0, rc2 = 0;
4197 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
4199 rc = llog_cleanup(ctxt);
4201 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
4203 rc2 = llog_cleanup(ctxt);
4210 static int osc_reconnect(const struct lu_env *env,
4211 struct obd_export *exp, struct obd_device *obd,
4212 struct obd_uuid *cluuid,
4213 struct obd_connect_data *data,
4216 struct client_obd *cli = &obd->u.cli;
4218 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
4221 client_obd_list_lock(&cli->cl_loi_list_lock);
4222 data->ocd_grant = (cli->cl_avail_grant + cli->cl_dirty) ?:
4223 2 * cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
4224 lost_grant = cli->cl_lost_grant;
4225 cli->cl_lost_grant = 0;
4226 client_obd_list_unlock(&cli->cl_loi_list_lock);
4228 CDEBUG(D_CACHE, "request ocd_grant: %d cl_avail_grant: %ld "
4229 "cl_dirty: %ld cl_lost_grant: %ld\n", data->ocd_grant,
4230 cli->cl_avail_grant, cli->cl_dirty, lost_grant);
4231 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
4232 " ocd_grant: %d\n", data->ocd_connect_flags,
4233 data->ocd_version, data->ocd_grant);
4239 static int osc_disconnect(struct obd_export *exp)
4241 struct obd_device *obd = class_exp2obd(exp);
4242 struct llog_ctxt *ctxt;
4245 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
4247 if (obd->u.cli.cl_conn_count == 1) {
4248 /* Flush any remaining cancel messages out to the
4250 llog_sync(ctxt, exp);
4252 llog_ctxt_put(ctxt);
4254 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
4258 rc = client_disconnect_export(exp);
4260 * Initially we put del_shrink_grant before disconnect_export, but it
4261 * causes the following problem if setup (connect) and cleanup
4262 * (disconnect) are tangled together.
4263 * connect p1 disconnect p2
4264 * ptlrpc_connect_import
4265 * ............... class_manual_cleanup
4268 * ptlrpc_connect_interrupt
4270 * add this client to shrink list
4272 * Bang! pinger trigger the shrink.
4273 * So the osc should be disconnected from the shrink list, after we
4274 * are sure the import has been destroyed. BUG18662
4276 if (obd->u.cli.cl_import == NULL)
4277 osc_del_shrink_grant(&obd->u.cli);
4281 static int osc_import_event(struct obd_device *obd,
4282 struct obd_import *imp,
4283 enum obd_import_event event)
4285 struct client_obd *cli;
4289 LASSERT(imp->imp_obd == obd);
4292 case IMP_EVENT_DISCON: {
4293 /* Only do this on the MDS OSC's */
4294 if (imp->imp_server_timeout) {
4295 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4297 cfs_spin_lock(&oscc->oscc_lock);
4298 oscc->oscc_flags |= OSCC_FLAG_RECOVERING;
4299 cfs_spin_unlock(&oscc->oscc_lock);
4302 client_obd_list_lock(&cli->cl_loi_list_lock);
4303 cli->cl_avail_grant = 0;
4304 cli->cl_lost_grant = 0;
4305 client_obd_list_unlock(&cli->cl_loi_list_lock);
4308 case IMP_EVENT_INACTIVE: {
4309 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
4312 case IMP_EVENT_INVALIDATE: {
4313 struct ldlm_namespace *ns = obd->obd_namespace;
4317 env = cl_env_get(&refcheck);
4321 client_obd_list_lock(&cli->cl_loi_list_lock);
4322 /* all pages go to failing rpcs due to the invalid
4324 osc_check_rpcs(env, cli);
4325 client_obd_list_unlock(&cli->cl_loi_list_lock);
4327 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
4328 cl_env_put(env, &refcheck);
4333 case IMP_EVENT_ACTIVE: {
4334 /* Only do this on the MDS OSC's */
4335 if (imp->imp_server_timeout) {
4336 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4338 cfs_spin_lock(&oscc->oscc_lock);
4339 oscc->oscc_flags &= ~OSCC_FLAG_NOSPC;
4340 cfs_spin_unlock(&oscc->oscc_lock);
4342 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
4345 case IMP_EVENT_OCD: {
4346 struct obd_connect_data *ocd = &imp->imp_connect_data;
4348 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
4349 osc_init_grant(&obd->u.cli, ocd);
4352 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
4353 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
4355 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
4359 CERROR("Unknown import event %d\n", event);
4366 * Determine whether the lock can be canceled before replaying the lock
4367 * during recovery, see bug16774 for detailed information.
4369 * \retval zero the lock can't be canceled
4370 * \retval other ok to cancel
4372 static int osc_cancel_for_recovery(struct ldlm_lock *lock)
4374 check_res_locked(lock->l_resource);
4377 * Cancel all unused extent lock in granted mode LCK_PR or LCK_CR.
4379 * XXX as a future improvement, we can also cancel unused write lock
4380 * if it doesn't have dirty data and active mmaps.
4382 if (lock->l_resource->lr_type == LDLM_EXTENT &&
4383 (lock->l_granted_mode == LCK_PR ||
4384 lock->l_granted_mode == LCK_CR) &&
4385 (osc_dlm_lock_pageref(lock) == 0))
4391 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
4397 rc = ptlrpcd_addref();
4401 rc = client_obd_setup(obd, lcfg);
4405 struct lprocfs_static_vars lvars = { 0 };
4406 struct client_obd *cli = &obd->u.cli;
4408 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
4409 lprocfs_osc_init_vars(&lvars);
4410 if (lprocfs_obd_setup(obd, lvars.obd_vars) == 0) {
4411 lproc_osc_attach_seqstat(obd);
4412 sptlrpc_lprocfs_cliobd_attach(obd);
4413 ptlrpc_lprocfs_register_obd(obd);
4417 /* We need to allocate a few requests more, because
4418 brw_interpret tries to create new requests before freeing
4419 previous ones. Ideally we want to have 2x max_rpcs_in_flight
4420 reserved, but I afraid that might be too much wasted RAM
4421 in fact, so 2 is just my guess and still should work. */
4422 cli->cl_import->imp_rq_pool =
4423 ptlrpc_init_rq_pool(cli->cl_max_rpcs_in_flight + 2,
4425 ptlrpc_add_rqs_to_pool);
4427 CFS_INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
4428 cfs_sema_init(&cli->cl_grant_sem, 1);
4430 ns_register_cancel(obd->obd_namespace, osc_cancel_for_recovery);
4436 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
4442 case OBD_CLEANUP_EARLY: {
4443 struct obd_import *imp;
4444 imp = obd->u.cli.cl_import;
4445 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
4446 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
4447 ptlrpc_deactivate_import(imp);
4448 cfs_spin_lock(&imp->imp_lock);
4449 imp->imp_pingable = 0;
4450 cfs_spin_unlock(&imp->imp_lock);
4453 case OBD_CLEANUP_EXPORTS: {
4454 /* If we set up but never connected, the
4455 client import will not have been cleaned. */
4456 if (obd->u.cli.cl_import) {
4457 struct obd_import *imp;
4458 cfs_down_write(&obd->u.cli.cl_sem);
4459 imp = obd->u.cli.cl_import;
4460 CDEBUG(D_CONFIG, "%s: client import never connected\n",
4462 ptlrpc_invalidate_import(imp);
4463 if (imp->imp_rq_pool) {
4464 ptlrpc_free_rq_pool(imp->imp_rq_pool);
4465 imp->imp_rq_pool = NULL;
4467 class_destroy_import(imp);
4468 cfs_up_write(&obd->u.cli.cl_sem);
4469 obd->u.cli.cl_import = NULL;
4471 rc = obd_llog_finish(obd, 0);
4473 CERROR("failed to cleanup llogging subsystems\n");
4480 int osc_cleanup(struct obd_device *obd)
4485 ptlrpc_lprocfs_unregister_obd(obd);
4486 lprocfs_obd_cleanup(obd);
4488 /* free memory of osc quota cache */
4489 lquota_cleanup(quota_interface, obd);
4491 rc = client_obd_cleanup(obd);
4497 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
4499 struct lprocfs_static_vars lvars = { 0 };
4502 lprocfs_osc_init_vars(&lvars);
4504 switch (lcfg->lcfg_command) {
4506 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars,
4516 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
4518 return osc_process_config_base(obd, buf);
4521 struct obd_ops osc_obd_ops = {
4522 .o_owner = THIS_MODULE,
4523 .o_setup = osc_setup,
4524 .o_precleanup = osc_precleanup,
4525 .o_cleanup = osc_cleanup,
4526 .o_add_conn = client_import_add_conn,
4527 .o_del_conn = client_import_del_conn,
4528 .o_connect = client_connect_import,
4529 .o_reconnect = osc_reconnect,
4530 .o_disconnect = osc_disconnect,
4531 .o_statfs = osc_statfs,
4532 .o_statfs_async = osc_statfs_async,
4533 .o_packmd = osc_packmd,
4534 .o_unpackmd = osc_unpackmd,
4535 .o_precreate = osc_precreate,
4536 .o_create = osc_create,
4537 .o_create_async = osc_create_async,
4538 .o_destroy = osc_destroy,
4539 .o_getattr = osc_getattr,
4540 .o_getattr_async = osc_getattr_async,
4541 .o_setattr = osc_setattr,
4542 .o_setattr_async = osc_setattr_async,
4544 .o_punch = osc_punch,
4546 .o_enqueue = osc_enqueue,
4547 .o_change_cbdata = osc_change_cbdata,
4548 .o_find_cbdata = osc_find_cbdata,
4549 .o_cancel = osc_cancel,
4550 .o_cancel_unused = osc_cancel_unused,
4551 .o_iocontrol = osc_iocontrol,
4552 .o_get_info = osc_get_info,
4553 .o_set_info_async = osc_set_info_async,
4554 .o_import_event = osc_import_event,
4555 .o_llog_init = osc_llog_init,
4556 .o_llog_finish = osc_llog_finish,
4557 .o_process_config = osc_process_config,
4560 extern struct lu_kmem_descr osc_caches[];
4561 extern cfs_spinlock_t osc_ast_guard;
4562 extern cfs_lock_class_key_t osc_ast_guard_class;
4564 int __init osc_init(void)
4566 struct lprocfs_static_vars lvars = { 0 };
4570 /* print an address of _any_ initialized kernel symbol from this
4571 * module, to allow debugging with gdb that doesn't support data
4572 * symbols from modules.*/
4573 CDEBUG(D_CONSOLE, "Lustre OSC module (%p).\n", &osc_caches);
4575 rc = lu_kmem_init(osc_caches);
4577 lprocfs_osc_init_vars(&lvars);
4579 cfs_request_module("lquota");
4580 quota_interface = PORTAL_SYMBOL_GET(osc_quota_interface);
4581 lquota_init(quota_interface);
4582 init_obd_quota_ops(quota_interface, &osc_obd_ops);
4584 rc = class_register_type(&osc_obd_ops, NULL, lvars.module_vars,
4585 LUSTRE_OSC_NAME, &osc_device_type);
4587 if (quota_interface)
4588 PORTAL_SYMBOL_PUT(osc_quota_interface);
4589 lu_kmem_fini(osc_caches);
4593 cfs_spin_lock_init(&osc_ast_guard);
4594 cfs_lockdep_set_class(&osc_ast_guard, &osc_ast_guard_class);
4596 osc_mds_ost_orig_logops = llog_lvfs_ops;
4597 osc_mds_ost_orig_logops.lop_setup = llog_obd_origin_setup;
4598 osc_mds_ost_orig_logops.lop_cleanup = llog_obd_origin_cleanup;
4599 osc_mds_ost_orig_logops.lop_add = llog_obd_origin_add;
4600 osc_mds_ost_orig_logops.lop_connect = llog_origin_connect;
4606 static void /*__exit*/ osc_exit(void)
4608 lu_device_type_fini(&osc_device_type);
4610 lquota_exit(quota_interface);
4611 if (quota_interface)
4612 PORTAL_SYMBOL_PUT(osc_quota_interface);
4614 class_unregister_type(LUSTRE_OSC_NAME);
4615 lu_kmem_fini(osc_caches);
4618 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
4619 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
4620 MODULE_LICENSE("GPL");
4622 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);