4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #define DEBUG_SUBSYSTEM S_OSC
39 #include <libcfs/libcfs.h>
42 # include <liblustre.h>
45 #include <lustre_dlm.h>
46 #include <lustre_net.h>
47 #include <lustre/lustre_user.h>
48 #include <obd_cksum.h>
56 #include <lustre_ha.h>
57 #include <lprocfs_status.h>
58 #include <lustre_log.h>
59 #include <lustre_debug.h>
60 #include <lustre_param.h>
61 #include <lustre_fid.h>
62 #include "osc_internal.h"
63 #include "osc_cl_internal.h"
65 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
66 static int brw_interpret(const struct lu_env *env,
67 struct ptlrpc_request *req, void *data, int rc);
68 int osc_cleanup(struct obd_device *obd);
70 /* Pack OSC object metadata for disk storage (LE byte order). */
71 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
72 struct lov_stripe_md *lsm)
77 lmm_size = sizeof(**lmmp);
81 if (*lmmp != NULL && lsm == NULL) {
82 OBD_FREE(*lmmp, lmm_size);
85 } else if (unlikely(lsm != NULL && ostid_id(&lsm->lsm_oi) == 0)) {
90 OBD_ALLOC(*lmmp, lmm_size);
96 ostid_cpu_to_le(&lsm->lsm_oi, &(*lmmp)->lmm_oi);
101 /* Unpack OSC object metadata from disk storage (LE byte order). */
102 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
103 struct lov_mds_md *lmm, int lmm_bytes)
106 struct obd_import *imp = class_exp2cliimp(exp);
110 if (lmm_bytes < sizeof(*lmm)) {
111 CERROR("%s: lov_mds_md too small: %d, need %d\n",
112 exp->exp_obd->obd_name, lmm_bytes,
116 /* XXX LOV_MAGIC etc check? */
118 if (unlikely(ostid_id(&lmm->lmm_oi) == 0)) {
119 CERROR("%s: zero lmm_object_id: rc = %d\n",
120 exp->exp_obd->obd_name, -EINVAL);
125 lsm_size = lov_stripe_md_size(1);
129 if (*lsmp != NULL && lmm == NULL) {
130 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
131 OBD_FREE(*lsmp, lsm_size);
137 OBD_ALLOC(*lsmp, lsm_size);
138 if (unlikely(*lsmp == NULL))
140 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
141 if (unlikely((*lsmp)->lsm_oinfo[0] == NULL)) {
142 OBD_FREE(*lsmp, lsm_size);
145 loi_init((*lsmp)->lsm_oinfo[0]);
146 } else if (unlikely(ostid_id(&(*lsmp)->lsm_oi) == 0)) {
151 /* XXX zero *lsmp? */
152 ostid_le_to_cpu(&lmm->lmm_oi, &(*lsmp)->lsm_oi);
155 (imp->imp_connect_data.ocd_connect_flags & OBD_CONNECT_MAXBYTES))
156 (*lsmp)->lsm_maxbytes = imp->imp_connect_data.ocd_maxbytes;
158 (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
163 static inline void osc_pack_capa(struct ptlrpc_request *req,
164 struct ost_body *body, void *capa)
166 struct obd_capa *oc = (struct obd_capa *)capa;
167 struct lustre_capa *c;
172 c = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
175 body->oa.o_valid |= OBD_MD_FLOSSCAPA;
176 DEBUG_CAPA(D_SEC, c, "pack");
179 static inline void osc_pack_req_body(struct ptlrpc_request *req,
180 struct obd_info *oinfo)
182 struct ost_body *body;
184 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
187 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
188 osc_pack_capa(req, body, oinfo->oi_capa);
191 static inline void osc_set_capa_size(struct ptlrpc_request *req,
192 const struct req_msg_field *field,
196 req_capsule_set_size(&req->rq_pill, field, RCL_CLIENT, 0);
198 /* it is already calculated as sizeof struct obd_capa */
202 static int osc_getattr_interpret(const struct lu_env *env,
203 struct ptlrpc_request *req,
204 struct osc_async_args *aa, int rc)
206 struct ost_body *body;
212 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
214 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
215 lustre_get_wire_obdo(aa->aa_oi->oi_oa, &body->oa);
217 /* This should really be sent by the OST */
218 aa->aa_oi->oi_oa->o_blksize = DT_MAX_BRW_SIZE;
219 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
221 CDEBUG(D_INFO, "can't unpack ost_body\n");
223 aa->aa_oi->oi_oa->o_valid = 0;
226 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
230 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
231 struct ptlrpc_request_set *set)
233 struct ptlrpc_request *req;
234 struct osc_async_args *aa;
238 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
242 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
243 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
245 ptlrpc_request_free(req);
249 osc_pack_req_body(req, oinfo);
251 ptlrpc_request_set_replen(req);
252 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
254 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
255 aa = ptlrpc_req_async_args(req);
258 ptlrpc_set_add_req(set, req);
262 static int osc_getattr(const struct lu_env *env, struct obd_export *exp,
263 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 oinfo->oi_oa->o_blksize = cli_brw_size(exp->exp_obd);
297 oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
301 ptlrpc_req_finished(req);
305 static int osc_setattr(const struct lu_env *env, struct obd_export *exp,
306 struct obd_info *oinfo, struct obd_trans_info *oti)
308 struct ptlrpc_request *req;
309 struct ost_body *body;
313 LASSERT(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP);
315 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
319 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
320 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
322 ptlrpc_request_free(req);
326 osc_pack_req_body(req, oinfo);
328 ptlrpc_request_set_replen(req);
330 rc = ptlrpc_queue_wait(req);
334 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
336 GOTO(out, rc = -EPROTO);
338 lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
342 ptlrpc_req_finished(req);
346 static int osc_setattr_interpret(const struct lu_env *env,
347 struct ptlrpc_request *req,
348 struct osc_setattr_args *sa, int rc)
350 struct ost_body *body;
356 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
358 GOTO(out, rc = -EPROTO);
360 lustre_get_wire_obdo(sa->sa_oa, &body->oa);
362 rc = sa->sa_upcall(sa->sa_cookie, rc);
366 int osc_setattr_async_base(struct obd_export *exp, struct obd_info *oinfo,
367 struct obd_trans_info *oti,
368 obd_enqueue_update_f upcall, void *cookie,
369 struct ptlrpc_request_set *rqset)
371 struct ptlrpc_request *req;
372 struct osc_setattr_args *sa;
376 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
380 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
381 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
383 ptlrpc_request_free(req);
387 if (oti && oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
388 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
390 osc_pack_req_body(req, oinfo);
392 ptlrpc_request_set_replen(req);
394 /* do mds to ost setattr asynchronously */
396 /* Do not wait for response. */
397 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
399 req->rq_interpret_reply =
400 (ptlrpc_interpterer_t)osc_setattr_interpret;
402 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
403 sa = ptlrpc_req_async_args(req);
404 sa->sa_oa = oinfo->oi_oa;
405 sa->sa_upcall = upcall;
406 sa->sa_cookie = cookie;
408 if (rqset == PTLRPCD_SET)
409 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
411 ptlrpc_set_add_req(rqset, req);
417 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
418 struct obd_trans_info *oti,
419 struct ptlrpc_request_set *rqset)
421 return osc_setattr_async_base(exp, oinfo, oti,
422 oinfo->oi_cb_up, oinfo, rqset);
425 int osc_real_create(struct obd_export *exp, struct obdo *oa,
426 struct lov_stripe_md **ea, struct obd_trans_info *oti)
428 struct ptlrpc_request *req;
429 struct ost_body *body;
430 struct lov_stripe_md *lsm;
439 rc = obd_alloc_memmd(exp, &lsm);
444 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
446 GOTO(out, rc = -ENOMEM);
448 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
450 ptlrpc_request_free(req);
454 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
456 lustre_set_wire_obdo(&body->oa, oa);
458 ptlrpc_request_set_replen(req);
460 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
461 oa->o_flags == OBD_FL_DELORPHAN) {
463 "delorphan from OST integration");
464 /* Don't resend the delorphan req */
465 req->rq_no_resend = req->rq_no_delay = 1;
468 rc = ptlrpc_queue_wait(req);
472 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
474 GOTO(out_req, rc = -EPROTO);
476 lustre_get_wire_obdo(oa, &body->oa);
478 oa->o_blksize = cli_brw_size(exp->exp_obd);
479 oa->o_valid |= OBD_MD_FLBLKSZ;
481 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
482 * have valid lsm_oinfo data structs, so don't go touching that.
483 * This needs to be fixed in a big way.
485 lsm->lsm_oi = oa->o_oi;
489 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
491 if (oa->o_valid & OBD_MD_FLCOOKIE) {
492 if (!oti->oti_logcookies)
493 oti_alloc_cookies(oti, 1);
494 *oti->oti_logcookies = oa->o_lcookie;
498 CDEBUG(D_HA, "transno: "LPD64"\n",
499 lustre_msg_get_transno(req->rq_repmsg));
501 ptlrpc_req_finished(req);
504 obd_free_memmd(exp, &lsm);
508 int osc_punch_base(struct obd_export *exp, struct obd_info *oinfo,
509 obd_enqueue_update_f upcall, void *cookie,
510 struct ptlrpc_request_set *rqset)
512 struct ptlrpc_request *req;
513 struct osc_setattr_args *sa;
514 struct ost_body *body;
518 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
522 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
523 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
525 ptlrpc_request_free(req);
528 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
529 ptlrpc_at_set_req_timeout(req);
531 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
533 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
534 osc_pack_capa(req, body, oinfo->oi_capa);
536 ptlrpc_request_set_replen(req);
538 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
539 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
540 sa = ptlrpc_req_async_args(req);
541 sa->sa_oa = oinfo->oi_oa;
542 sa->sa_upcall = upcall;
543 sa->sa_cookie = cookie;
544 if (rqset == PTLRPCD_SET)
545 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
547 ptlrpc_set_add_req(rqset, req);
552 static int osc_punch(const struct lu_env *env, struct obd_export *exp,
553 struct obd_info *oinfo, struct obd_trans_info *oti,
554 struct ptlrpc_request_set *rqset)
556 oinfo->oi_oa->o_size = oinfo->oi_policy.l_extent.start;
557 oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
558 oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
559 return osc_punch_base(exp, oinfo,
560 oinfo->oi_cb_up, oinfo, rqset);
563 static int osc_sync_interpret(const struct lu_env *env,
564 struct ptlrpc_request *req,
567 struct osc_fsync_args *fa = arg;
568 struct ost_body *body;
574 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
576 CERROR ("can't unpack ost_body\n");
577 GOTO(out, rc = -EPROTO);
580 *fa->fa_oi->oi_oa = body->oa;
582 rc = fa->fa_upcall(fa->fa_cookie, rc);
586 int osc_sync_base(struct obd_export *exp, struct obd_info *oinfo,
587 obd_enqueue_update_f upcall, void *cookie,
588 struct ptlrpc_request_set *rqset)
590 struct ptlrpc_request *req;
591 struct ost_body *body;
592 struct osc_fsync_args *fa;
596 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
600 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
601 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
603 ptlrpc_request_free(req);
607 /* overload the size and blocks fields in the oa with start/end */
608 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
610 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
611 osc_pack_capa(req, body, oinfo->oi_capa);
613 ptlrpc_request_set_replen(req);
614 req->rq_interpret_reply = osc_sync_interpret;
616 CLASSERT(sizeof(*fa) <= sizeof(req->rq_async_args));
617 fa = ptlrpc_req_async_args(req);
619 fa->fa_upcall = upcall;
620 fa->fa_cookie = cookie;
622 if (rqset == PTLRPCD_SET)
623 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
625 ptlrpc_set_add_req(rqset, req);
630 static int osc_sync(const struct lu_env *env, struct obd_export *exp,
631 struct obd_info *oinfo, obd_size start, obd_size end,
632 struct ptlrpc_request_set *set)
637 CDEBUG(D_INFO, "oa NULL\n");
641 oinfo->oi_oa->o_size = start;
642 oinfo->oi_oa->o_blocks = end;
643 oinfo->oi_oa->o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
645 RETURN(osc_sync_base(exp, oinfo, oinfo->oi_cb_up, oinfo, set));
648 /* Find and cancel locally locks matched by @mode in the resource found by
649 * @objid. Found locks are added into @cancel list. Returns the amount of
650 * locks added to @cancels list. */
651 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
653 ldlm_mode_t mode, int lock_flags)
655 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
656 struct ldlm_res_id res_id;
657 struct ldlm_resource *res;
661 /* Return, i.e. cancel nothing, only if ELC is supported (flag in
662 * export) but disabled through procfs (flag in NS).
664 * This distinguishes from a case when ELC is not supported originally,
665 * when we still want to cancel locks in advance and just cancel them
666 * locally, without sending any RPC. */
667 if (exp_connect_cancelset(exp) && !ns_connect_cancelset(ns))
670 ostid_build_res_name(&oa->o_oi, &res_id);
671 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
675 LDLM_RESOURCE_ADDREF(res);
676 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
677 lock_flags, 0, NULL);
678 LDLM_RESOURCE_DELREF(res);
679 ldlm_resource_putref(res);
683 static int osc_destroy_interpret(const struct lu_env *env,
684 struct ptlrpc_request *req, void *data,
687 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
689 cfs_atomic_dec(&cli->cl_destroy_in_flight);
690 cfs_waitq_signal(&cli->cl_destroy_waitq);
694 static int osc_can_send_destroy(struct client_obd *cli)
696 if (cfs_atomic_inc_return(&cli->cl_destroy_in_flight) <=
697 cli->cl_max_rpcs_in_flight) {
698 /* The destroy request can be sent */
701 if (cfs_atomic_dec_return(&cli->cl_destroy_in_flight) <
702 cli->cl_max_rpcs_in_flight) {
704 * The counter has been modified between the two atomic
707 cfs_waitq_signal(&cli->cl_destroy_waitq);
712 int osc_create(const struct lu_env *env, struct obd_export *exp,
713 struct obdo *oa, struct lov_stripe_md **ea,
714 struct obd_trans_info *oti)
721 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
723 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
724 oa->o_flags == OBD_FL_RECREATE_OBJS) {
725 RETURN(osc_real_create(exp, oa, ea, oti));
728 if (!fid_seq_is_mdt(ostid_seq(&oa->o_oi)))
729 RETURN(osc_real_create(exp, oa, ea, oti));
731 /* we should not get here anymore */
737 /* Destroy requests can be async always on the client, and we don't even really
738 * care about the return code since the client cannot do anything at all about
740 * When the MDS is unlinking a filename, it saves the file objects into a
741 * recovery llog, and these object records are cancelled when the OST reports
742 * they were destroyed and sync'd to disk (i.e. transaction committed).
743 * If the client dies, or the OST is down when the object should be destroyed,
744 * the records are not cancelled, and when the OST reconnects to the MDS next,
745 * it will retrieve the llog unlink logs and then sends the log cancellation
746 * cookies to the MDS after committing destroy transactions. */
747 static int osc_destroy(const struct lu_env *env, struct obd_export *exp,
748 struct obdo *oa, struct lov_stripe_md *ea,
749 struct obd_trans_info *oti, struct obd_export *md_export,
752 struct client_obd *cli = &exp->exp_obd->u.cli;
753 struct ptlrpc_request *req;
754 struct ost_body *body;
755 CFS_LIST_HEAD(cancels);
760 CDEBUG(D_INFO, "oa NULL\n");
764 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
765 LDLM_FL_DISCARD_DATA);
767 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
769 ldlm_lock_list_put(&cancels, l_bl_ast, count);
773 osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
774 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
777 ptlrpc_request_free(req);
781 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
782 ptlrpc_at_set_req_timeout(req);
784 if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
785 oa->o_lcookie = *oti->oti_logcookies;
786 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
788 lustre_set_wire_obdo(&body->oa, oa);
790 osc_pack_capa(req, body, (struct obd_capa *)capa);
791 ptlrpc_request_set_replen(req);
793 /* If osc_destory is for destroying the unlink orphan,
794 * sent from MDT to OST, which should not be blocked here,
795 * because the process might be triggered by ptlrpcd, and
796 * it is not good to block ptlrpcd thread (b=16006)*/
797 if (!(oa->o_flags & OBD_FL_DELORPHAN)) {
798 req->rq_interpret_reply = osc_destroy_interpret;
799 if (!osc_can_send_destroy(cli)) {
800 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP,
804 * Wait until the number of on-going destroy RPCs drops
805 * under max_rpc_in_flight
807 l_wait_event_exclusive(cli->cl_destroy_waitq,
808 osc_can_send_destroy(cli), &lwi);
812 /* Do not wait for response */
813 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
817 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
820 obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
822 LASSERT(!(oa->o_valid & bits));
825 client_obd_list_lock(&cli->cl_loi_list_lock);
826 oa->o_dirty = cli->cl_dirty;
827 if (unlikely(cli->cl_dirty - cli->cl_dirty_transit >
828 cli->cl_dirty_max)) {
829 CERROR("dirty %lu - %lu > dirty_max %lu\n",
830 cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
832 } else if (unlikely(cfs_atomic_read(&obd_dirty_pages) -
833 cfs_atomic_read(&obd_dirty_transit_pages) >
834 (long)(obd_max_dirty_pages + 1))) {
835 /* The cfs_atomic_read() allowing the cfs_atomic_inc() are
836 * not covered by a lock thus they may safely race and trip
837 * this CERROR() unless we add in a small fudge factor (+1). */
838 CERROR("dirty %d - %d > system dirty_max %d\n",
839 cfs_atomic_read(&obd_dirty_pages),
840 cfs_atomic_read(&obd_dirty_transit_pages),
841 obd_max_dirty_pages);
843 } else if (unlikely(cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff)) {
844 CERROR("dirty %lu - dirty_max %lu too big???\n",
845 cli->cl_dirty, cli->cl_dirty_max);
848 long max_in_flight = (cli->cl_max_pages_per_rpc <<
850 (cli->cl_max_rpcs_in_flight + 1);
851 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
853 oa->o_grant = cli->cl_avail_grant + cli->cl_reserved_grant;
854 oa->o_dropped = cli->cl_lost_grant;
855 cli->cl_lost_grant = 0;
856 client_obd_list_unlock(&cli->cl_loi_list_lock);
857 CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
858 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
862 void osc_update_next_shrink(struct client_obd *cli)
864 cli->cl_next_shrink_grant =
865 cfs_time_shift(cli->cl_grant_shrink_interval);
866 CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
867 cli->cl_next_shrink_grant);
870 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
872 client_obd_list_lock(&cli->cl_loi_list_lock);
873 cli->cl_avail_grant += grant;
874 client_obd_list_unlock(&cli->cl_loi_list_lock);
877 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
879 if (body->oa.o_valid & OBD_MD_FLGRANT) {
880 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
881 __osc_update_grant(cli, body->oa.o_grant);
885 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
886 obd_count keylen, void *key, obd_count vallen,
887 void *val, struct ptlrpc_request_set *set);
889 static int osc_shrink_grant_interpret(const struct lu_env *env,
890 struct ptlrpc_request *req,
893 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
894 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
895 struct ost_body *body;
898 __osc_update_grant(cli, oa->o_grant);
902 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
904 osc_update_grant(cli, body);
910 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
912 client_obd_list_lock(&cli->cl_loi_list_lock);
913 oa->o_grant = cli->cl_avail_grant / 4;
914 cli->cl_avail_grant -= oa->o_grant;
915 client_obd_list_unlock(&cli->cl_loi_list_lock);
916 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
917 oa->o_valid |= OBD_MD_FLFLAGS;
920 oa->o_flags |= OBD_FL_SHRINK_GRANT;
921 osc_update_next_shrink(cli);
924 /* Shrink the current grant, either from some large amount to enough for a
925 * full set of in-flight RPCs, or if we have already shrunk to that limit
926 * then to enough for a single RPC. This avoids keeping more grant than
927 * needed, and avoids shrinking the grant piecemeal. */
928 static int osc_shrink_grant(struct client_obd *cli)
930 __u64 target_bytes = (cli->cl_max_rpcs_in_flight + 1) *
931 (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT);
933 client_obd_list_lock(&cli->cl_loi_list_lock);
934 if (cli->cl_avail_grant <= target_bytes)
935 target_bytes = cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
936 client_obd_list_unlock(&cli->cl_loi_list_lock);
938 return osc_shrink_grant_to_target(cli, target_bytes);
941 int osc_shrink_grant_to_target(struct client_obd *cli, __u64 target_bytes)
944 struct ost_body *body;
947 client_obd_list_lock(&cli->cl_loi_list_lock);
948 /* Don't shrink if we are already above or below the desired limit
949 * We don't want to shrink below a single RPC, as that will negatively
950 * impact block allocation and long-term performance. */
951 if (target_bytes < cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)
952 target_bytes = cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
954 if (target_bytes >= cli->cl_avail_grant) {
955 client_obd_list_unlock(&cli->cl_loi_list_lock);
958 client_obd_list_unlock(&cli->cl_loi_list_lock);
964 osc_announce_cached(cli, &body->oa, 0);
966 client_obd_list_lock(&cli->cl_loi_list_lock);
967 body->oa.o_grant = cli->cl_avail_grant - target_bytes;
968 cli->cl_avail_grant = target_bytes;
969 client_obd_list_unlock(&cli->cl_loi_list_lock);
970 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
971 body->oa.o_valid |= OBD_MD_FLFLAGS;
972 body->oa.o_flags = 0;
974 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
975 osc_update_next_shrink(cli);
977 rc = osc_set_info_async(NULL, cli->cl_import->imp_obd->obd_self_export,
978 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
979 sizeof(*body), body, NULL);
981 __osc_update_grant(cli, body->oa.o_grant);
986 static int osc_should_shrink_grant(struct client_obd *client)
988 cfs_time_t time = cfs_time_current();
989 cfs_time_t next_shrink = client->cl_next_shrink_grant;
991 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
992 OBD_CONNECT_GRANT_SHRINK) == 0)
995 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
996 /* Get the current RPC size directly, instead of going via:
997 * cli_brw_size(obd->u.cli.cl_import->imp_obd->obd_self_export)
998 * Keep comment here so that it can be found by searching. */
999 int brw_size = client->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
1001 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
1002 client->cl_avail_grant > brw_size)
1005 osc_update_next_shrink(client);
1010 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1012 struct client_obd *client;
1014 cfs_list_for_each_entry(client, &item->ti_obd_list,
1015 cl_grant_shrink_list) {
1016 if (osc_should_shrink_grant(client))
1017 osc_shrink_grant(client);
1022 static int osc_add_shrink_grant(struct client_obd *client)
1026 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
1028 osc_grant_shrink_grant_cb, NULL,
1029 &client->cl_grant_shrink_list);
1031 CERROR("add grant client %s error %d\n",
1032 client->cl_import->imp_obd->obd_name, rc);
1035 CDEBUG(D_CACHE, "add grant client %s \n",
1036 client->cl_import->imp_obd->obd_name);
1037 osc_update_next_shrink(client);
1041 static int osc_del_shrink_grant(struct client_obd *client)
1043 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1047 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1050 * ocd_grant is the total grant amount we're expect to hold: if we've
1051 * been evicted, it's the new avail_grant amount, cl_dirty will drop
1052 * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1054 * race is tolerable here: if we're evicted, but imp_state already
1055 * left EVICTED state, then cl_dirty must be 0 already.
1057 client_obd_list_lock(&cli->cl_loi_list_lock);
1058 if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1059 cli->cl_avail_grant = ocd->ocd_grant;
1061 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1063 if (cli->cl_avail_grant < 0) {
1064 CWARN("%s: available grant < 0: avail/ocd/dirty %ld/%u/%ld\n",
1065 cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant,
1066 ocd->ocd_grant, cli->cl_dirty);
1067 /* workaround for servers which do not have the patch from
1069 cli->cl_avail_grant = ocd->ocd_grant;
1072 /* determine the appropriate chunk size used by osc_extent. */
1073 cli->cl_chunkbits = max_t(int, CFS_PAGE_SHIFT, ocd->ocd_blocksize);
1074 client_obd_list_unlock(&cli->cl_loi_list_lock);
1076 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld."
1077 "chunk bits: %d.\n", cli->cl_import->imp_obd->obd_name,
1078 cli->cl_avail_grant, cli->cl_lost_grant, cli->cl_chunkbits);
1080 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1081 cfs_list_empty(&cli->cl_grant_shrink_list))
1082 osc_add_shrink_grant(cli);
1085 /* We assume that the reason this OSC got a short read is because it read
1086 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1087 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1088 * this stripe never got written at or beyond this stripe offset yet. */
1089 static void handle_short_read(int nob_read, obd_count page_count,
1090 struct brw_page **pga)
1095 /* skip bytes read OK */
1096 while (nob_read > 0) {
1097 LASSERT (page_count > 0);
1099 if (pga[i]->count > nob_read) {
1100 /* EOF inside this page */
1101 ptr = cfs_kmap(pga[i]->pg) +
1102 (pga[i]->off & ~CFS_PAGE_MASK);
1103 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1104 cfs_kunmap(pga[i]->pg);
1110 nob_read -= pga[i]->count;
1115 /* zero remaining pages */
1116 while (page_count-- > 0) {
1117 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1118 memset(ptr, 0, pga[i]->count);
1119 cfs_kunmap(pga[i]->pg);
1124 static int check_write_rcs(struct ptlrpc_request *req,
1125 int requested_nob, int niocount,
1126 obd_count page_count, struct brw_page **pga)
1131 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1132 sizeof(*remote_rcs) *
1134 if (remote_rcs == NULL) {
1135 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1139 /* return error if any niobuf was in error */
1140 for (i = 0; i < niocount; i++) {
1141 if ((int)remote_rcs[i] < 0)
1142 return(remote_rcs[i]);
1144 if (remote_rcs[i] != 0) {
1145 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1146 i, remote_rcs[i], req);
1151 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1152 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1153 req->rq_bulk->bd_nob_transferred, requested_nob);
1160 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1162 if (p1->flag != p2->flag) {
1163 unsigned mask = ~(OBD_BRW_FROM_GRANT| OBD_BRW_NOCACHE|
1164 OBD_BRW_SYNC|OBD_BRW_ASYNC|OBD_BRW_NOQUOTA);
1166 /* warn if we try to combine flags that we don't know to be
1167 * safe to combine */
1168 if (unlikely((p1->flag & mask) != (p2->flag & mask))) {
1169 CWARN("Saw flags 0x%x and 0x%x in the same brw, please "
1170 "report this at http://bugs.whamcloud.com/\n",
1171 p1->flag, p2->flag);
1176 return (p1->off + p1->count == p2->off);
1179 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1180 struct brw_page **pga, int opc,
1181 cksum_type_t cksum_type)
1185 struct cfs_crypto_hash_desc *hdesc;
1186 unsigned int bufsize;
1188 unsigned char cfs_alg = cksum_obd2cfs(cksum_type);
1190 LASSERT(pg_count > 0);
1192 hdesc = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1193 if (IS_ERR(hdesc)) {
1194 CERROR("Unable to initialize checksum hash %s\n",
1195 cfs_crypto_hash_name(cfs_alg));
1196 return PTR_ERR(hdesc);
1199 while (nob > 0 && pg_count > 0) {
1200 int count = pga[i]->count > nob ? nob : pga[i]->count;
1202 /* corrupt the data before we compute the checksum, to
1203 * simulate an OST->client data error */
1204 if (i == 0 && opc == OST_READ &&
1205 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE)) {
1206 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1207 int off = pga[i]->off & ~CFS_PAGE_MASK;
1208 memcpy(ptr + off, "bad1", min(4, nob));
1209 cfs_kunmap(pga[i]->pg);
1211 cfs_crypto_hash_update_page(hdesc, pga[i]->pg,
1212 pga[i]->off & ~CFS_PAGE_MASK,
1214 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1215 (int)(pga[i]->off & ~CFS_PAGE_MASK), cksum);
1217 nob -= pga[i]->count;
1223 err = cfs_crypto_hash_final(hdesc, (unsigned char *)&cksum, &bufsize);
1226 cfs_crypto_hash_final(hdesc, NULL, NULL);
1228 /* For sending we only compute the wrong checksum instead
1229 * of corrupting the data so it is still correct on a redo */
1230 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1236 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1237 struct lov_stripe_md *lsm, obd_count page_count,
1238 struct brw_page **pga,
1239 struct ptlrpc_request **reqp,
1240 struct obd_capa *ocapa, int reserve,
1243 struct ptlrpc_request *req;
1244 struct ptlrpc_bulk_desc *desc;
1245 struct ost_body *body;
1246 struct obd_ioobj *ioobj;
1247 struct niobuf_remote *niobuf;
1248 int niocount, i, requested_nob, opc, rc;
1249 struct osc_brw_async_args *aa;
1250 struct req_capsule *pill;
1251 struct brw_page *pg_prev;
1254 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1255 RETURN(-ENOMEM); /* Recoverable */
1256 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1257 RETURN(-EINVAL); /* Fatal */
1259 if ((cmd & OBD_BRW_WRITE) != 0) {
1261 req = ptlrpc_request_alloc_pool(cli->cl_import,
1262 cli->cl_import->imp_rq_pool,
1263 &RQF_OST_BRW_WRITE);
1266 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1271 for (niocount = i = 1; i < page_count; i++) {
1272 if (!can_merge_pages(pga[i - 1], pga[i]))
1276 pill = &req->rq_pill;
1277 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1279 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1280 niocount * sizeof(*niobuf));
1281 osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1283 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1285 ptlrpc_request_free(req);
1288 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1289 ptlrpc_at_set_req_timeout(req);
1290 /* ask ptlrpc not to resend on EINPROGRESS since BRWs have their own
1292 req->rq_no_retry_einprogress = 1;
1294 desc = ptlrpc_prep_bulk_imp(req, page_count,
1295 cli->cl_import->imp_connect_data.ocd_brw_size >> LNET_MTU_BITS,
1296 opc == OST_WRITE ? BULK_GET_SOURCE : BULK_PUT_SINK,
1300 GOTO(out, rc = -ENOMEM);
1301 /* NB request now owns desc and will free it when it gets freed */
1303 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1304 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1305 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1306 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1308 lustre_set_wire_obdo(&body->oa, oa);
1310 obdo_to_ioobj(oa, ioobj);
1311 ioobj->ioo_bufcnt = niocount;
1312 /* The high bits of ioo_max_brw tells server _maximum_ number of bulks
1313 * that might be send for this request. The actual number is decided
1314 * when the RPC is finally sent in ptlrpc_register_bulk(). It sends
1315 * "max - 1" for old client compatibility sending "0", and also so the
1316 * the actual maximum is a power-of-two number, not one less. LU-1431 */
1317 ioobj_max_brw_set(ioobj, desc->bd_md_max_brw);
1318 osc_pack_capa(req, body, ocapa);
1319 LASSERT(page_count > 0);
1321 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1322 struct brw_page *pg = pga[i];
1323 int poff = pg->off & ~CFS_PAGE_MASK;
1325 LASSERT(pg->count > 0);
1326 /* make sure there is no gap in the middle of page array */
1327 LASSERTF(page_count == 1 ||
1328 (ergo(i == 0, poff + pg->count == CFS_PAGE_SIZE) &&
1329 ergo(i > 0 && i < page_count - 1,
1330 poff == 0 && pg->count == CFS_PAGE_SIZE) &&
1331 ergo(i == page_count - 1, poff == 0)),
1332 "i: %d/%d pg: %p off: "LPU64", count: %u\n",
1333 i, page_count, pg, pg->off, pg->count);
1335 LASSERTF(i == 0 || pg->off > pg_prev->off,
1336 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1337 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1339 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1340 pg_prev->pg, page_private(pg_prev->pg),
1341 pg_prev->pg->index, pg_prev->off);
1343 LASSERTF(i == 0 || pg->off > pg_prev->off,
1344 "i %d p_c %u\n", i, page_count);
1346 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1347 (pg->flag & OBD_BRW_SRVLOCK));
1349 ptlrpc_prep_bulk_page_pin(desc, pg->pg, poff, pg->count);
1350 requested_nob += pg->count;
1352 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1354 niobuf->len += pg->count;
1356 niobuf->offset = pg->off;
1357 niobuf->len = pg->count;
1358 niobuf->flags = pg->flag;
1363 LASSERTF((void *)(niobuf - niocount) ==
1364 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1365 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1366 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1368 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1370 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1371 body->oa.o_valid |= OBD_MD_FLFLAGS;
1372 body->oa.o_flags = 0;
1374 body->oa.o_flags |= OBD_FL_RECOV_RESEND;
1377 if (osc_should_shrink_grant(cli))
1378 osc_shrink_grant_local(cli, &body->oa);
1380 /* size[REQ_REC_OFF] still sizeof (*body) */
1381 if (opc == OST_WRITE) {
1382 if (cli->cl_checksum &&
1383 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1384 /* store cl_cksum_type in a local variable since
1385 * it can be changed via lprocfs */
1386 cksum_type_t cksum_type = cli->cl_cksum_type;
1388 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1389 oa->o_flags &= OBD_FL_LOCAL_MASK;
1390 body->oa.o_flags = 0;
1392 body->oa.o_flags |= cksum_type_pack(cksum_type);
1393 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1394 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1398 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1400 /* save this in 'oa', too, for later checking */
1401 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1402 oa->o_flags |= cksum_type_pack(cksum_type);
1404 /* clear out the checksum flag, in case this is a
1405 * resend but cl_checksum is no longer set. b=11238 */
1406 oa->o_valid &= ~OBD_MD_FLCKSUM;
1408 oa->o_cksum = body->oa.o_cksum;
1409 /* 1 RC per niobuf */
1410 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1411 sizeof(__u32) * niocount);
1413 if (cli->cl_checksum &&
1414 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1415 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1416 body->oa.o_flags = 0;
1417 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1418 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1421 ptlrpc_request_set_replen(req);
1423 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1424 aa = ptlrpc_req_async_args(req);
1426 aa->aa_requested_nob = requested_nob;
1427 aa->aa_nio_count = niocount;
1428 aa->aa_page_count = page_count;
1432 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1433 if (ocapa && reserve)
1434 aa->aa_ocapa = capa_get(ocapa);
1440 ptlrpc_req_finished(req);
1444 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1445 __u32 client_cksum, __u32 server_cksum, int nob,
1446 obd_count page_count, struct brw_page **pga,
1447 cksum_type_t client_cksum_type)
1451 cksum_type_t cksum_type;
1453 if (server_cksum == client_cksum) {
1454 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1458 cksum_type = cksum_type_unpack(oa->o_valid & OBD_MD_FLFLAGS ?
1460 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1463 if (cksum_type != client_cksum_type)
1464 msg = "the server did not use the checksum type specified in "
1465 "the original request - likely a protocol problem";
1466 else if (new_cksum == server_cksum)
1467 msg = "changed on the client after we checksummed it - "
1468 "likely false positive due to mmap IO (bug 11742)";
1469 else if (new_cksum == client_cksum)
1470 msg = "changed in transit before arrival at OST";
1472 msg = "changed in transit AND doesn't match the original - "
1473 "likely false positive due to mmap IO (bug 11742)";
1475 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1476 " object "DOSTID" extent ["LPU64"-"LPU64"]\n",
1477 msg, libcfs_nid2str(peer->nid),
1478 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1479 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1480 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1481 POSTID(&oa->o_oi), pga[0]->off,
1482 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1483 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1484 "client csum now %x\n", client_cksum, client_cksum_type,
1485 server_cksum, cksum_type, new_cksum);
1489 /* Note rc enters this function as number of bytes transferred */
1490 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1492 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1493 const lnet_process_id_t *peer =
1494 &req->rq_import->imp_connection->c_peer;
1495 struct client_obd *cli = aa->aa_cli;
1496 struct ost_body *body;
1497 __u32 client_cksum = 0;
1500 if (rc < 0 && rc != -EDQUOT) {
1501 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1505 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1506 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1508 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1512 /* set/clear over quota flag for a uid/gid */
1513 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1514 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1515 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1517 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid "LPX64", flags %x\n",
1518 body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1520 osc_quota_setdq(cli, qid, body->oa.o_valid, body->oa.o_flags);
1523 osc_update_grant(cli, body);
1528 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1529 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1531 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1533 CERROR("Unexpected +ve rc %d\n", rc);
1536 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1538 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1541 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1542 check_write_checksum(&body->oa, peer, client_cksum,
1543 body->oa.o_cksum, aa->aa_requested_nob,
1544 aa->aa_page_count, aa->aa_ppga,
1545 cksum_type_unpack(aa->aa_oa->o_flags)))
1548 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1549 aa->aa_page_count, aa->aa_ppga);
1553 /* The rest of this function executes only for OST_READs */
1555 /* if unwrap_bulk failed, return -EAGAIN to retry */
1556 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1558 GOTO(out, rc = -EAGAIN);
1560 if (rc > aa->aa_requested_nob) {
1561 CERROR("Unexpected rc %d (%d requested)\n", rc,
1562 aa->aa_requested_nob);
1566 if (rc != req->rq_bulk->bd_nob_transferred) {
1567 CERROR ("Unexpected rc %d (%d transferred)\n",
1568 rc, req->rq_bulk->bd_nob_transferred);
1572 if (rc < aa->aa_requested_nob)
1573 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1575 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1576 static int cksum_counter;
1577 __u32 server_cksum = body->oa.o_cksum;
1580 cksum_type_t cksum_type;
1582 cksum_type = cksum_type_unpack(body->oa.o_valid &OBD_MD_FLFLAGS?
1583 body->oa.o_flags : 0);
1584 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1585 aa->aa_ppga, OST_READ,
1588 if (peer->nid == req->rq_bulk->bd_sender) {
1592 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1595 if (server_cksum == ~0 && rc > 0) {
1596 CERROR("Protocol error: server %s set the 'checksum' "
1597 "bit, but didn't send a checksum. Not fatal, "
1598 "but please notify on http://bugs.whamcloud.com/\n",
1599 libcfs_nid2str(peer->nid));
1600 } else if (server_cksum != client_cksum) {
1601 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1602 "%s%s%s inode "DFID" object "DOSTID
1603 " extent ["LPU64"-"LPU64"]\n",
1604 req->rq_import->imp_obd->obd_name,
1605 libcfs_nid2str(peer->nid),
1607 body->oa.o_valid & OBD_MD_FLFID ?
1608 body->oa.o_parent_seq : (__u64)0,
1609 body->oa.o_valid & OBD_MD_FLFID ?
1610 body->oa.o_parent_oid : 0,
1611 body->oa.o_valid & OBD_MD_FLFID ?
1612 body->oa.o_parent_ver : 0,
1613 POSTID(&body->oa.o_oi),
1614 aa->aa_ppga[0]->off,
1615 aa->aa_ppga[aa->aa_page_count-1]->off +
1616 aa->aa_ppga[aa->aa_page_count-1]->count -
1618 CERROR("client %x, server %x, cksum_type %x\n",
1619 client_cksum, server_cksum, cksum_type);
1621 aa->aa_oa->o_cksum = client_cksum;
1625 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1628 } else if (unlikely(client_cksum)) {
1629 static int cksum_missed;
1632 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1633 CERROR("Checksum %u requested from %s but not sent\n",
1634 cksum_missed, libcfs_nid2str(peer->nid));
1640 lustre_get_wire_obdo(aa->aa_oa, &body->oa);
1645 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1646 struct lov_stripe_md *lsm,
1647 obd_count page_count, struct brw_page **pga,
1648 struct obd_capa *ocapa)
1650 struct ptlrpc_request *req;
1653 int generation, resends = 0;
1654 struct l_wait_info lwi;
1658 cfs_waitq_init(&waitq);
1659 generation = exp->exp_obd->u.cli.cl_import->imp_generation;
1662 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1663 page_count, pga, &req, ocapa, 0, resends);
1668 req->rq_generation_set = 1;
1669 req->rq_import_generation = generation;
1670 req->rq_sent = cfs_time_current_sec() + resends;
1673 rc = ptlrpc_queue_wait(req);
1675 if (rc == -ETIMEDOUT && req->rq_resend) {
1676 DEBUG_REQ(D_HA, req, "BULK TIMEOUT");
1677 ptlrpc_req_finished(req);
1681 rc = osc_brw_fini_request(req, rc);
1683 ptlrpc_req_finished(req);
1684 /* When server return -EINPROGRESS, client should always retry
1685 * regardless of the number of times the bulk was resent already.*/
1686 if (osc_recoverable_error(rc)) {
1688 if (rc != -EINPROGRESS &&
1689 !client_should_resend(resends, &exp->exp_obd->u.cli)) {
1690 CERROR("%s: too many resend retries for object: "
1691 ""DOSTID", rc = %d.\n", exp->exp_obd->obd_name,
1692 POSTID(&oa->o_oi), rc);
1696 exp->exp_obd->u.cli.cl_import->imp_generation) {
1697 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1698 ""DOSTID", rc = %d.\n", exp->exp_obd->obd_name,
1699 POSTID(&oa->o_oi), rc);
1703 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL,
1705 l_wait_event(waitq, 0, &lwi);
1710 if (rc == -EAGAIN || rc == -EINPROGRESS)
1715 static int osc_brw_redo_request(struct ptlrpc_request *request,
1716 struct osc_brw_async_args *aa, int rc)
1718 struct ptlrpc_request *new_req;
1719 struct osc_brw_async_args *new_aa;
1720 struct osc_async_page *oap;
1723 DEBUG_REQ(rc == -EINPROGRESS ? D_RPCTRACE : D_ERROR, request,
1724 "redo for recoverable error %d", rc);
1726 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1727 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1728 aa->aa_cli, aa->aa_oa,
1729 NULL /* lsm unused by osc currently */,
1730 aa->aa_page_count, aa->aa_ppga,
1731 &new_req, aa->aa_ocapa, 0, 1);
1735 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1736 if (oap->oap_request != NULL) {
1737 LASSERTF(request == oap->oap_request,
1738 "request %p != oap_request %p\n",
1739 request, oap->oap_request);
1740 if (oap->oap_interrupted) {
1741 ptlrpc_req_finished(new_req);
1746 /* New request takes over pga and oaps from old request.
1747 * Note that copying a list_head doesn't work, need to move it... */
1749 new_req->rq_interpret_reply = request->rq_interpret_reply;
1750 new_req->rq_async_args = request->rq_async_args;
1751 /* cap resend delay to the current request timeout, this is similar to
1752 * what ptlrpc does (see after_reply()) */
1753 if (aa->aa_resends > new_req->rq_timeout)
1754 new_req->rq_sent = cfs_time_current_sec() + new_req->rq_timeout;
1756 new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1757 new_req->rq_generation_set = 1;
1758 new_req->rq_import_generation = request->rq_import_generation;
1760 new_aa = ptlrpc_req_async_args(new_req);
1762 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1763 cfs_list_splice_init(&aa->aa_oaps, &new_aa->aa_oaps);
1764 CFS_INIT_LIST_HEAD(&new_aa->aa_exts);
1765 cfs_list_splice_init(&aa->aa_exts, &new_aa->aa_exts);
1766 new_aa->aa_resends = aa->aa_resends;
1768 cfs_list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1769 if (oap->oap_request) {
1770 ptlrpc_req_finished(oap->oap_request);
1771 oap->oap_request = ptlrpc_request_addref(new_req);
1775 new_aa->aa_ocapa = aa->aa_ocapa;
1776 aa->aa_ocapa = NULL;
1778 /* XXX: This code will run into problem if we're going to support
1779 * to add a series of BRW RPCs into a self-defined ptlrpc_request_set
1780 * and wait for all of them to be finished. We should inherit request
1781 * set from old request. */
1782 ptlrpcd_add_req(new_req, PDL_POLICY_SAME, -1);
1784 DEBUG_REQ(D_INFO, new_req, "new request");
1789 * ugh, we want disk allocation on the target to happen in offset order. we'll
1790 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1791 * fine for our small page arrays and doesn't require allocation. its an
1792 * insertion sort that swaps elements that are strides apart, shrinking the
1793 * stride down until its '1' and the array is sorted.
1795 static void sort_brw_pages(struct brw_page **array, int num)
1798 struct brw_page *tmp;
1802 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1807 for (i = stride ; i < num ; i++) {
1810 while (j >= stride && array[j - stride]->off > tmp->off) {
1811 array[j] = array[j - stride];
1816 } while (stride > 1);
1819 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1825 LASSERT (pages > 0);
1826 offset = pg[i]->off & ~CFS_PAGE_MASK;
1830 if (pages == 0) /* that's all */
1833 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1834 return count; /* doesn't end on page boundary */
1837 offset = pg[i]->off & ~CFS_PAGE_MASK;
1838 if (offset != 0) /* doesn't start on page boundary */
1845 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1847 struct brw_page **ppga;
1850 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1854 for (i = 0; i < count; i++)
1859 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1861 LASSERT(ppga != NULL);
1862 OBD_FREE(ppga, sizeof(*ppga) * count);
1865 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1866 obd_count page_count, struct brw_page *pga,
1867 struct obd_trans_info *oti)
1869 struct obdo *saved_oa = NULL;
1870 struct brw_page **ppga, **orig;
1871 struct obd_import *imp = class_exp2cliimp(exp);
1872 struct client_obd *cli;
1873 int rc, page_count_orig;
1876 LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1877 cli = &imp->imp_obd->u.cli;
1879 if (cmd & OBD_BRW_CHECK) {
1880 /* The caller just wants to know if there's a chance that this
1881 * I/O can succeed */
1883 if (imp->imp_invalid)
1888 /* test_brw with a failed create can trip this, maybe others. */
1889 LASSERT(cli->cl_max_pages_per_rpc);
1893 orig = ppga = osc_build_ppga(pga, page_count);
1896 page_count_orig = page_count;
1898 sort_brw_pages(ppga, page_count);
1899 while (page_count) {
1900 obd_count pages_per_brw;
1902 if (page_count > cli->cl_max_pages_per_rpc)
1903 pages_per_brw = cli->cl_max_pages_per_rpc;
1905 pages_per_brw = page_count;
1907 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1909 if (saved_oa != NULL) {
1910 /* restore previously saved oa */
1911 *oinfo->oi_oa = *saved_oa;
1912 } else if (page_count > pages_per_brw) {
1913 /* save a copy of oa (brw will clobber it) */
1914 OBDO_ALLOC(saved_oa);
1915 if (saved_oa == NULL)
1916 GOTO(out, rc = -ENOMEM);
1917 *saved_oa = *oinfo->oi_oa;
1920 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1921 pages_per_brw, ppga, oinfo->oi_capa);
1926 page_count -= pages_per_brw;
1927 ppga += pages_per_brw;
1931 osc_release_ppga(orig, page_count_orig);
1933 if (saved_oa != NULL)
1934 OBDO_FREE(saved_oa);
1939 static int brw_interpret(const struct lu_env *env,
1940 struct ptlrpc_request *req, void *data, int rc)
1942 struct osc_brw_async_args *aa = data;
1943 struct osc_extent *ext;
1944 struct osc_extent *tmp;
1945 struct cl_object *obj = NULL;
1946 struct client_obd *cli = aa->aa_cli;
1949 rc = osc_brw_fini_request(req, rc);
1950 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
1951 /* When server return -EINPROGRESS, client should always retry
1952 * regardless of the number of times the bulk was resent already. */
1953 if (osc_recoverable_error(rc)) {
1954 if (req->rq_import_generation !=
1955 req->rq_import->imp_generation) {
1956 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1957 ""DOSTID", rc = %d.\n",
1958 req->rq_import->imp_obd->obd_name,
1959 POSTID(&aa->aa_oa->o_oi), rc);
1960 } else if (rc == -EINPROGRESS ||
1961 client_should_resend(aa->aa_resends, aa->aa_cli)) {
1962 rc = osc_brw_redo_request(req, aa, rc);
1964 CERROR("%s: too many resent retries for object: "
1965 ""LPU64":"LPU64", rc = %d.\n",
1966 req->rq_import->imp_obd->obd_name,
1967 POSTID(&aa->aa_oa->o_oi), rc);
1972 else if (rc == -EAGAIN || rc == -EINPROGRESS)
1977 capa_put(aa->aa_ocapa);
1978 aa->aa_ocapa = NULL;
1981 cfs_list_for_each_entry_safe(ext, tmp, &aa->aa_exts, oe_link) {
1982 if (obj == NULL && rc == 0) {
1983 obj = osc2cl(ext->oe_obj);
1987 cfs_list_del_init(&ext->oe_link);
1988 osc_extent_finish(env, ext, 1, rc);
1990 LASSERT(cfs_list_empty(&aa->aa_exts));
1991 LASSERT(cfs_list_empty(&aa->aa_oaps));
1994 struct obdo *oa = aa->aa_oa;
1995 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
1996 unsigned long valid = 0;
1999 if (oa->o_valid & OBD_MD_FLBLOCKS) {
2000 attr->cat_blocks = oa->o_blocks;
2001 valid |= CAT_BLOCKS;
2003 if (oa->o_valid & OBD_MD_FLMTIME) {
2004 attr->cat_mtime = oa->o_mtime;
2007 if (oa->o_valid & OBD_MD_FLATIME) {
2008 attr->cat_atime = oa->o_atime;
2011 if (oa->o_valid & OBD_MD_FLCTIME) {
2012 attr->cat_ctime = oa->o_ctime;
2016 cl_object_attr_lock(obj);
2017 cl_object_attr_set(env, obj, attr, valid);
2018 cl_object_attr_unlock(obj);
2020 cl_object_put(env, obj);
2022 OBDO_FREE(aa->aa_oa);
2024 cl_req_completion(env, aa->aa_clerq, rc < 0 ? rc :
2025 req->rq_bulk->bd_nob_transferred);
2026 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2027 ptlrpc_lprocfs_brw(req, req->rq_bulk->bd_nob_transferred);
2029 client_obd_list_lock(&cli->cl_loi_list_lock);
2030 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2031 * is called so we know whether to go to sync BRWs or wait for more
2032 * RPCs to complete */
2033 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2034 cli->cl_w_in_flight--;
2036 cli->cl_r_in_flight--;
2037 osc_wake_cache_waiters(cli);
2038 client_obd_list_unlock(&cli->cl_loi_list_lock);
2040 osc_io_unplug(env, cli, NULL, PDL_POLICY_SAME);
2045 * Build an RPC by the list of extent @ext_list. The caller must ensure
2046 * that the total pages in this list are NOT over max pages per RPC.
2047 * Extents in the list must be in OES_RPC state.
2049 int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
2050 cfs_list_t *ext_list, int cmd, pdl_policy_t pol)
2052 struct ptlrpc_request *req = NULL;
2053 struct osc_extent *ext;
2054 CFS_LIST_HEAD(rpc_list);
2055 struct brw_page **pga = NULL;
2056 struct osc_brw_async_args *aa = NULL;
2057 struct obdo *oa = NULL;
2058 struct osc_async_page *oap;
2059 struct osc_async_page *tmp;
2060 struct cl_req *clerq = NULL;
2061 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2062 struct ldlm_lock *lock = NULL;
2063 struct cl_req_attr crattr;
2064 obd_off starting_offset = OBD_OBJECT_EOF;
2065 obd_off ending_offset = 0;
2066 int i, rc, mpflag = 0, mem_tight = 0, page_count = 0;
2069 LASSERT(!cfs_list_empty(ext_list));
2071 /* add pages into rpc_list to build BRW rpc */
2072 cfs_list_for_each_entry(ext, ext_list, oe_link) {
2073 LASSERT(ext->oe_state == OES_RPC);
2074 mem_tight |= ext->oe_memalloc;
2075 cfs_list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
2077 cfs_list_add_tail(&oap->oap_rpc_item, &rpc_list);
2078 if (starting_offset > oap->oap_obj_off)
2079 starting_offset = oap->oap_obj_off;
2081 LASSERT(oap->oap_page_off == 0);
2082 if (ending_offset < oap->oap_obj_off + oap->oap_count)
2083 ending_offset = oap->oap_obj_off +
2086 LASSERT(oap->oap_page_off + oap->oap_count ==
2092 mpflag = cfs_memory_pressure_get_and_set();
2094 memset(&crattr, 0, sizeof crattr);
2095 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2097 GOTO(out, rc = -ENOMEM);
2101 GOTO(out, rc = -ENOMEM);
2104 cfs_list_for_each_entry(oap, &rpc_list, oap_rpc_item) {
2105 struct cl_page *page = oap2cl_page(oap);
2106 if (clerq == NULL) {
2107 clerq = cl_req_alloc(env, page, crt,
2108 1 /* only 1-object rpcs for
2111 GOTO(out, rc = PTR_ERR(clerq));
2112 lock = oap->oap_ldlm_lock;
2115 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
2116 pga[i] = &oap->oap_brw_page;
2117 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2118 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2119 pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2121 cl_req_page_add(env, clerq, page);
2124 /* always get the data for the obdo for the rpc */
2125 LASSERT(clerq != NULL);
2127 crattr.cra_capa = NULL;
2128 memset(crattr.cra_jobid, 0, JOBSTATS_JOBID_SIZE);
2129 cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2131 oa->o_handle = lock->l_remote_handle;
2132 oa->o_valid |= OBD_MD_FLHANDLE;
2135 rc = cl_req_prep(env, clerq);
2137 CERROR("cl_req_prep failed: %d\n", rc);
2141 sort_brw_pages(pga, page_count);
2142 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2143 pga, &req, crattr.cra_capa, 1, 0);
2145 CERROR("prep_req failed: %d\n", rc);
2149 req->rq_interpret_reply = brw_interpret;
2151 req->rq_memalloc = 1;
2153 /* Need to update the timestamps after the request is built in case
2154 * we race with setattr (locally or in queue at OST). If OST gets
2155 * later setattr before earlier BRW (as determined by the request xid),
2156 * the OST will not use BRW timestamps. Sadly, there is no obvious
2157 * way to do this in a single call. bug 10150 */
2158 cl_req_attr_set(env, clerq, &crattr,
2159 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2161 lustre_msg_set_jobid(req->rq_reqmsg, crattr.cra_jobid);
2163 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2164 aa = ptlrpc_req_async_args(req);
2165 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2166 cfs_list_splice_init(&rpc_list, &aa->aa_oaps);
2167 CFS_INIT_LIST_HEAD(&aa->aa_exts);
2168 cfs_list_splice_init(ext_list, &aa->aa_exts);
2169 aa->aa_clerq = clerq;
2171 /* queued sync pages can be torn down while the pages
2172 * were between the pending list and the rpc */
2174 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2175 /* only one oap gets a request reference */
2178 if (oap->oap_interrupted && !req->rq_intr) {
2179 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2181 ptlrpc_mark_interrupted(req);
2185 tmp->oap_request = ptlrpc_request_addref(req);
2187 client_obd_list_lock(&cli->cl_loi_list_lock);
2188 starting_offset >>= CFS_PAGE_SHIFT;
2189 if (cmd == OBD_BRW_READ) {
2190 cli->cl_r_in_flight++;
2191 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2192 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2193 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2194 starting_offset + 1);
2196 cli->cl_w_in_flight++;
2197 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2198 lprocfs_oh_tally(&cli->cl_write_rpc_hist, cli->cl_w_in_flight);
2199 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2200 starting_offset + 1);
2202 client_obd_list_unlock(&cli->cl_loi_list_lock);
2204 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2205 page_count, aa, cli->cl_r_in_flight,
2206 cli->cl_w_in_flight);
2208 /* XXX: Maybe the caller can check the RPC bulk descriptor to
2209 * see which CPU/NUMA node the majority of pages were allocated
2210 * on, and try to assign the async RPC to the CPU core
2211 * (PDL_POLICY_PREFERRED) to reduce cross-CPU memory traffic.
2213 * But on the other hand, we expect that multiple ptlrpcd
2214 * threads and the initial write sponsor can run in parallel,
2215 * especially when data checksum is enabled, which is CPU-bound
2216 * operation and single ptlrpcd thread cannot process in time.
2217 * So more ptlrpcd threads sharing BRW load
2218 * (with PDL_POLICY_ROUND) seems better.
2220 ptlrpcd_add_req(req, pol, -1);
2226 cfs_memory_pressure_restore(mpflag);
2228 capa_put(crattr.cra_capa);
2230 LASSERT(req == NULL);
2235 OBD_FREE(pga, sizeof(*pga) * page_count);
2236 /* this should happen rarely and is pretty bad, it makes the
2237 * pending list not follow the dirty order */
2238 while (!cfs_list_empty(ext_list)) {
2239 ext = cfs_list_entry(ext_list->next, struct osc_extent,
2241 cfs_list_del_init(&ext->oe_link);
2242 osc_extent_finish(env, ext, 0, rc);
2244 if (clerq && !IS_ERR(clerq))
2245 cl_req_completion(env, clerq, rc);
2250 static int osc_set_lock_data_with_check(struct ldlm_lock *lock,
2251 struct ldlm_enqueue_info *einfo)
2253 void *data = einfo->ei_cbdata;
2256 LASSERT(lock != NULL);
2257 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2258 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2259 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2260 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2262 lock_res_and_lock(lock);
2263 spin_lock(&osc_ast_guard);
2265 if (lock->l_ast_data == NULL)
2266 lock->l_ast_data = data;
2267 if (lock->l_ast_data == data)
2270 spin_unlock(&osc_ast_guard);
2271 unlock_res_and_lock(lock);
2276 static int osc_set_data_with_check(struct lustre_handle *lockh,
2277 struct ldlm_enqueue_info *einfo)
2279 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2283 set = osc_set_lock_data_with_check(lock, einfo);
2284 LDLM_LOCK_PUT(lock);
2286 CERROR("lockh %p, data %p - client evicted?\n",
2287 lockh, einfo->ei_cbdata);
2291 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2292 ldlm_iterator_t replace, void *data)
2294 struct ldlm_res_id res_id;
2295 struct obd_device *obd = class_exp2obd(exp);
2297 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2298 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2302 /* find any ldlm lock of the inode in osc
2306 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2307 ldlm_iterator_t replace, void *data)
2309 struct ldlm_res_id res_id;
2310 struct obd_device *obd = class_exp2obd(exp);
2313 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2314 rc = ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2315 if (rc == LDLM_ITER_STOP)
2317 if (rc == LDLM_ITER_CONTINUE)
2322 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
2323 obd_enqueue_update_f upcall, void *cookie,
2324 __u64 *flags, int agl, int rc)
2326 int intent = *flags & LDLM_FL_HAS_INTENT;
2330 /* The request was created before ldlm_cli_enqueue call. */
2331 if (rc == ELDLM_LOCK_ABORTED) {
2332 struct ldlm_reply *rep;
2333 rep = req_capsule_server_get(&req->rq_pill,
2336 LASSERT(rep != NULL);
2337 if (rep->lock_policy_res1)
2338 rc = rep->lock_policy_res1;
2342 if ((intent != 0 && rc == ELDLM_LOCK_ABORTED && agl == 0) ||
2344 *flags |= LDLM_FL_LVB_READY;
2345 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
2346 lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
2349 /* Call the update callback. */
2350 rc = (*upcall)(cookie, rc);
2354 static int osc_enqueue_interpret(const struct lu_env *env,
2355 struct ptlrpc_request *req,
2356 struct osc_enqueue_args *aa, int rc)
2358 struct ldlm_lock *lock;
2359 struct lustre_handle handle;
2361 struct ost_lvb *lvb;
2363 __u64 *flags = aa->oa_flags;
2365 /* Make a local copy of a lock handle and a mode, because aa->oa_*
2366 * might be freed anytime after lock upcall has been called. */
2367 lustre_handle_copy(&handle, aa->oa_lockh);
2368 mode = aa->oa_ei->ei_mode;
2370 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2372 lock = ldlm_handle2lock(&handle);
2374 /* Take an additional reference so that a blocking AST that
2375 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2376 * to arrive after an upcall has been executed by
2377 * osc_enqueue_fini(). */
2378 ldlm_lock_addref(&handle, mode);
2380 /* Let CP AST to grant the lock first. */
2381 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
2383 if (aa->oa_agl && rc == ELDLM_LOCK_ABORTED) {
2388 lvb_len = sizeof(*aa->oa_lvb);
2391 /* Complete obtaining the lock procedure. */
2392 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
2393 mode, flags, lvb, lvb_len, &handle, rc);
2394 /* Complete osc stuff. */
2395 rc = osc_enqueue_fini(req, aa->oa_lvb, aa->oa_upcall, aa->oa_cookie,
2396 flags, aa->oa_agl, rc);
2398 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
2400 /* Release the lock for async request. */
2401 if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
2403 * Releases a reference taken by ldlm_cli_enqueue(), if it is
2404 * not already released by
2405 * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
2407 ldlm_lock_decref(&handle, mode);
2409 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
2410 aa->oa_lockh, req, aa);
2411 ldlm_lock_decref(&handle, mode);
2412 LDLM_LOCK_PUT(lock);
2416 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
2417 struct lov_oinfo *loi, int flags,
2418 struct ost_lvb *lvb, __u32 mode, int rc)
2420 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
2422 if (rc == ELDLM_OK) {
2425 LASSERT(lock != NULL);
2426 loi->loi_lvb = *lvb;
2427 tmp = loi->loi_lvb.lvb_size;
2428 /* Extend KMS up to the end of this lock and no further
2429 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
2430 if (tmp > lock->l_policy_data.l_extent.end)
2431 tmp = lock->l_policy_data.l_extent.end + 1;
2432 if (tmp >= loi->loi_kms) {
2433 LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
2434 ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
2435 loi_kms_set(loi, tmp);
2437 LDLM_DEBUG(lock, "lock acquired, setting rss="
2438 LPU64"; leaving kms="LPU64", end="LPU64,
2439 loi->loi_lvb.lvb_size, loi->loi_kms,
2440 lock->l_policy_data.l_extent.end);
2442 ldlm_lock_allow_match(lock);
2443 } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
2444 LASSERT(lock != NULL);
2445 loi->loi_lvb = *lvb;
2446 ldlm_lock_allow_match(lock);
2447 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
2448 " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
2454 ldlm_lock_fail_match(lock);
2456 LDLM_LOCK_PUT(lock);
2459 EXPORT_SYMBOL(osc_update_enqueue);
2461 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
2463 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
2464 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
2465 * other synchronous requests, however keeping some locks and trying to obtain
2466 * others may take a considerable amount of time in a case of ost failure; and
2467 * when other sync requests do not get released lock from a client, the client
2468 * is excluded from the cluster -- such scenarious make the life difficult, so
2469 * release locks just after they are obtained. */
2470 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2471 __u64 *flags, ldlm_policy_data_t *policy,
2472 struct ost_lvb *lvb, int kms_valid,
2473 obd_enqueue_update_f upcall, void *cookie,
2474 struct ldlm_enqueue_info *einfo,
2475 struct lustre_handle *lockh,
2476 struct ptlrpc_request_set *rqset, int async, int agl)
2478 struct obd_device *obd = exp->exp_obd;
2479 struct ptlrpc_request *req = NULL;
2480 int intent = *flags & LDLM_FL_HAS_INTENT;
2481 int match_lvb = (agl != 0 ? 0 : LDLM_FL_LVB_READY);
2486 /* Filesystem lock extents are extended to page boundaries so that
2487 * dealing with the page cache is a little smoother. */
2488 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
2489 policy->l_extent.end |= ~CFS_PAGE_MASK;
2492 * kms is not valid when either object is completely fresh (so that no
2493 * locks are cached), or object was evicted. In the latter case cached
2494 * lock cannot be used, because it would prime inode state with
2495 * potentially stale LVB.
2500 /* Next, search for already existing extent locks that will cover us */
2501 /* If we're trying to read, we also search for an existing PW lock. The
2502 * VFS and page cache already protect us locally, so lots of readers/
2503 * writers can share a single PW lock.
2505 * There are problems with conversion deadlocks, so instead of
2506 * converting a read lock to a write lock, we'll just enqueue a new
2509 * At some point we should cancel the read lock instead of making them
2510 * send us a blocking callback, but there are problems with canceling
2511 * locks out from other users right now, too. */
2512 mode = einfo->ei_mode;
2513 if (einfo->ei_mode == LCK_PR)
2515 mode = ldlm_lock_match(obd->obd_namespace, *flags | match_lvb, res_id,
2516 einfo->ei_type, policy, mode, lockh, 0);
2518 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
2520 if ((agl != 0) && !(matched->l_flags & LDLM_FL_LVB_READY)) {
2521 /* For AGL, if enqueue RPC is sent but the lock is not
2522 * granted, then skip to process this strpe.
2523 * Return -ECANCELED to tell the caller. */
2524 ldlm_lock_decref(lockh, mode);
2525 LDLM_LOCK_PUT(matched);
2527 } else if (osc_set_lock_data_with_check(matched, einfo)) {
2528 *flags |= LDLM_FL_LVB_READY;
2529 /* addref the lock only if not async requests and PW
2530 * lock is matched whereas we asked for PR. */
2531 if (!rqset && einfo->ei_mode != mode)
2532 ldlm_lock_addref(lockh, LCK_PR);
2534 /* I would like to be able to ASSERT here that
2535 * rss <= kms, but I can't, for reasons which
2536 * are explained in lov_enqueue() */
2539 /* We already have a lock, and it's referenced.
2541 * At this point, the cl_lock::cll_state is CLS_QUEUING,
2542 * AGL upcall may change it to CLS_HELD directly. */
2543 (*upcall)(cookie, ELDLM_OK);
2545 if (einfo->ei_mode != mode)
2546 ldlm_lock_decref(lockh, LCK_PW);
2548 /* For async requests, decref the lock. */
2549 ldlm_lock_decref(lockh, einfo->ei_mode);
2550 LDLM_LOCK_PUT(matched);
2553 ldlm_lock_decref(lockh, mode);
2554 LDLM_LOCK_PUT(matched);
2560 CFS_LIST_HEAD(cancels);
2561 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2562 &RQF_LDLM_ENQUEUE_LVB);
2566 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
2568 ptlrpc_request_free(req);
2572 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
2574 ptlrpc_request_set_replen(req);
2577 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
2578 *flags &= ~LDLM_FL_BLOCK_GRANTED;
2580 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
2581 sizeof(*lvb), LVB_T_OST, lockh, async);
2584 struct osc_enqueue_args *aa;
2585 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
2586 aa = ptlrpc_req_async_args(req);
2589 aa->oa_flags = flags;
2590 aa->oa_upcall = upcall;
2591 aa->oa_cookie = cookie;
2593 aa->oa_lockh = lockh;
2596 req->rq_interpret_reply =
2597 (ptlrpc_interpterer_t)osc_enqueue_interpret;
2598 if (rqset == PTLRPCD_SET)
2599 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
2601 ptlrpc_set_add_req(rqset, req);
2602 } else if (intent) {
2603 ptlrpc_req_finished(req);
2608 rc = osc_enqueue_fini(req, lvb, upcall, cookie, flags, agl, rc);
2610 ptlrpc_req_finished(req);
2615 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
2616 struct ldlm_enqueue_info *einfo,
2617 struct ptlrpc_request_set *rqset)
2619 struct ldlm_res_id res_id;
2623 ostid_build_res_name(&oinfo->oi_md->lsm_oi, &res_id);
2624 rc = osc_enqueue_base(exp, &res_id, &oinfo->oi_flags, &oinfo->oi_policy,
2625 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
2626 oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid,
2627 oinfo->oi_cb_up, oinfo, einfo, oinfo->oi_lockh,
2628 rqset, rqset != NULL, 0);
2632 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2633 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
2634 int *flags, void *data, struct lustre_handle *lockh,
2637 struct obd_device *obd = exp->exp_obd;
2638 int lflags = *flags;
2642 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
2645 /* Filesystem lock extents are extended to page boundaries so that
2646 * dealing with the page cache is a little smoother */
2647 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
2648 policy->l_extent.end |= ~CFS_PAGE_MASK;
2650 /* Next, search for already existing extent locks that will cover us */
2651 /* If we're trying to read, we also search for an existing PW lock. The
2652 * VFS and page cache already protect us locally, so lots of readers/
2653 * writers can share a single PW lock. */
2657 rc = ldlm_lock_match(obd->obd_namespace, lflags,
2658 res_id, type, policy, rc, lockh, unref);
2661 if (!osc_set_data_with_check(lockh, data)) {
2662 if (!(lflags & LDLM_FL_TEST_LOCK))
2663 ldlm_lock_decref(lockh, rc);
2667 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
2668 ldlm_lock_addref(lockh, LCK_PR);
2669 ldlm_lock_decref(lockh, LCK_PW);
2676 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
2680 if (unlikely(mode == LCK_GROUP))
2681 ldlm_lock_decref_and_cancel(lockh, mode);
2683 ldlm_lock_decref(lockh, mode);
2688 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
2689 __u32 mode, struct lustre_handle *lockh)
2692 RETURN(osc_cancel_base(lockh, mode));
2695 static int osc_cancel_unused(struct obd_export *exp,
2696 struct lov_stripe_md *lsm,
2697 ldlm_cancel_flags_t flags,
2700 struct obd_device *obd = class_exp2obd(exp);
2701 struct ldlm_res_id res_id, *resp = NULL;
2704 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2708 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
2711 static int osc_statfs_interpret(const struct lu_env *env,
2712 struct ptlrpc_request *req,
2713 struct osc_async_args *aa, int rc)
2715 struct obd_statfs *msfs;
2719 /* The request has in fact never been sent
2720 * due to issues at a higher level (LOV).
2721 * Exit immediately since the caller is
2722 * aware of the problem and takes care
2723 * of the clean up */
2726 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
2727 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
2733 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2735 GOTO(out, rc = -EPROTO);
2738 *aa->aa_oi->oi_osfs = *msfs;
2740 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
2744 static int osc_statfs_async(struct obd_export *exp,
2745 struct obd_info *oinfo, __u64 max_age,
2746 struct ptlrpc_request_set *rqset)
2748 struct obd_device *obd = class_exp2obd(exp);
2749 struct ptlrpc_request *req;
2750 struct osc_async_args *aa;
2754 /* We could possibly pass max_age in the request (as an absolute
2755 * timestamp or a "seconds.usec ago") so the target can avoid doing
2756 * extra calls into the filesystem if that isn't necessary (e.g.
2757 * during mount that would help a bit). Having relative timestamps
2758 * is not so great if request processing is slow, while absolute
2759 * timestamps are not ideal because they need time synchronization. */
2760 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
2764 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2766 ptlrpc_request_free(req);
2769 ptlrpc_request_set_replen(req);
2770 req->rq_request_portal = OST_CREATE_PORTAL;
2771 ptlrpc_at_set_req_timeout(req);
2773 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
2774 /* procfs requests not want stat in wait for avoid deadlock */
2775 req->rq_no_resend = 1;
2776 req->rq_no_delay = 1;
2779 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
2780 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
2781 aa = ptlrpc_req_async_args(req);
2784 ptlrpc_set_add_req(rqset, req);
2788 static int osc_statfs(const struct lu_env *env, struct obd_export *exp,
2789 struct obd_statfs *osfs, __u64 max_age, __u32 flags)
2791 struct obd_device *obd = class_exp2obd(exp);
2792 struct obd_statfs *msfs;
2793 struct ptlrpc_request *req;
2794 struct obd_import *imp = NULL;
2798 /*Since the request might also come from lprocfs, so we need
2799 *sync this with client_disconnect_export Bug15684*/
2800 down_read(&obd->u.cli.cl_sem);
2801 if (obd->u.cli.cl_import)
2802 imp = class_import_get(obd->u.cli.cl_import);
2803 up_read(&obd->u.cli.cl_sem);
2807 /* We could possibly pass max_age in the request (as an absolute
2808 * timestamp or a "seconds.usec ago") so the target can avoid doing
2809 * extra calls into the filesystem if that isn't necessary (e.g.
2810 * during mount that would help a bit). Having relative timestamps
2811 * is not so great if request processing is slow, while absolute
2812 * timestamps are not ideal because they need time synchronization. */
2813 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
2815 class_import_put(imp);
2820 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2822 ptlrpc_request_free(req);
2825 ptlrpc_request_set_replen(req);
2826 req->rq_request_portal = OST_CREATE_PORTAL;
2827 ptlrpc_at_set_req_timeout(req);
2829 if (flags & OBD_STATFS_NODELAY) {
2830 /* procfs requests not want stat in wait for avoid deadlock */
2831 req->rq_no_resend = 1;
2832 req->rq_no_delay = 1;
2835 rc = ptlrpc_queue_wait(req);
2839 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2841 GOTO(out, rc = -EPROTO);
2848 ptlrpc_req_finished(req);
2852 /* Retrieve object striping information.
2854 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
2855 * the maximum number of OST indices which will fit in the user buffer.
2856 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
2858 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
2860 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
2861 struct lov_user_md_v3 lum, *lumk;
2862 struct lov_user_ost_data_v1 *lmm_objects;
2863 int rc = 0, lum_size;
2869 /* we only need the header part from user space to get lmm_magic and
2870 * lmm_stripe_count, (the header part is common to v1 and v3) */
2871 lum_size = sizeof(struct lov_user_md_v1);
2872 if (cfs_copy_from_user(&lum, lump, lum_size))
2875 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
2876 (lum.lmm_magic != LOV_USER_MAGIC_V3))
2879 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
2880 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
2881 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
2882 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
2884 /* we can use lov_mds_md_size() to compute lum_size
2885 * because lov_user_md_vX and lov_mds_md_vX have the same size */
2886 if (lum.lmm_stripe_count > 0) {
2887 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
2888 OBD_ALLOC(lumk, lum_size);
2892 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
2894 &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
2896 lmm_objects = &(lumk->lmm_objects[0]);
2897 lmm_objects->l_ost_oi = lsm->lsm_oi;
2899 lum_size = lov_mds_md_size(0, lum.lmm_magic);
2903 lumk->lmm_oi = lsm->lsm_oi;
2904 lumk->lmm_stripe_count = 1;
2906 if (cfs_copy_to_user(lump, lumk, lum_size))
2910 OBD_FREE(lumk, lum_size);
2916 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
2917 void *karg, void *uarg)
2919 struct obd_device *obd = exp->exp_obd;
2920 struct obd_ioctl_data *data = karg;
2924 if (!cfs_try_module_get(THIS_MODULE)) {
2925 CERROR("Can't get module. Is it alive?");
2929 case OBD_IOC_LOV_GET_CONFIG: {
2931 struct lov_desc *desc;
2932 struct obd_uuid uuid;
2936 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
2937 GOTO(out, err = -EINVAL);
2939 data = (struct obd_ioctl_data *)buf;
2941 if (sizeof(*desc) > data->ioc_inllen1) {
2942 obd_ioctl_freedata(buf, len);
2943 GOTO(out, err = -EINVAL);
2946 if (data->ioc_inllen2 < sizeof(uuid)) {
2947 obd_ioctl_freedata(buf, len);
2948 GOTO(out, err = -EINVAL);
2951 desc = (struct lov_desc *)data->ioc_inlbuf1;
2952 desc->ld_tgt_count = 1;
2953 desc->ld_active_tgt_count = 1;
2954 desc->ld_default_stripe_count = 1;
2955 desc->ld_default_stripe_size = 0;
2956 desc->ld_default_stripe_offset = 0;
2957 desc->ld_pattern = 0;
2958 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
2960 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
2962 err = cfs_copy_to_user((void *)uarg, buf, len);
2965 obd_ioctl_freedata(buf, len);
2968 case LL_IOC_LOV_SETSTRIPE:
2969 err = obd_alloc_memmd(exp, karg);
2973 case LL_IOC_LOV_GETSTRIPE:
2974 err = osc_getstripe(karg, uarg);
2976 case OBD_IOC_CLIENT_RECOVER:
2977 err = ptlrpc_recover_import(obd->u.cli.cl_import,
2978 data->ioc_inlbuf1, 0);
2982 case IOC_OSC_SET_ACTIVE:
2983 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
2986 case OBD_IOC_POLL_QUOTACHECK:
2987 err = osc_quota_poll_check(exp, (struct if_quotacheck *)karg);
2989 case OBD_IOC_PING_TARGET:
2990 err = ptlrpc_obd_ping(obd);
2993 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
2994 cmd, cfs_curproc_comm());
2995 GOTO(out, err = -ENOTTY);
2998 cfs_module_put(THIS_MODULE);
3002 static int osc_get_info(const struct lu_env *env, struct obd_export *exp,
3003 obd_count keylen, void *key, __u32 *vallen, void *val,
3004 struct lov_stripe_md *lsm)
3007 if (!vallen || !val)
3010 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
3011 __u32 *stripe = val;
3012 *vallen = sizeof(*stripe);
3015 } else if (KEY_IS(KEY_LAST_ID)) {
3016 struct ptlrpc_request *req;
3021 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3022 &RQF_OST_GET_INFO_LAST_ID);
3026 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3027 RCL_CLIENT, keylen);
3028 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3030 ptlrpc_request_free(req);
3034 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3035 memcpy(tmp, key, keylen);
3037 req->rq_no_delay = req->rq_no_resend = 1;
3038 ptlrpc_request_set_replen(req);
3039 rc = ptlrpc_queue_wait(req);
3043 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
3045 GOTO(out, rc = -EPROTO);
3047 *((obd_id *)val) = *reply;
3049 ptlrpc_req_finished(req);
3051 } else if (KEY_IS(KEY_FIEMAP)) {
3052 struct ptlrpc_request *req;
3053 struct ll_user_fiemap *reply;
3057 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3058 &RQF_OST_GET_INFO_FIEMAP);
3062 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
3063 RCL_CLIENT, keylen);
3064 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3065 RCL_CLIENT, *vallen);
3066 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3067 RCL_SERVER, *vallen);
3069 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3071 ptlrpc_request_free(req);
3075 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
3076 memcpy(tmp, key, keylen);
3077 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3078 memcpy(tmp, val, *vallen);
3080 ptlrpc_request_set_replen(req);
3081 rc = ptlrpc_queue_wait(req);
3085 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3087 GOTO(out1, rc = -EPROTO);
3089 memcpy(val, reply, *vallen);
3091 ptlrpc_req_finished(req);
3099 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
3100 obd_count keylen, void *key, obd_count vallen,
3101 void *val, struct ptlrpc_request_set *set)
3103 struct ptlrpc_request *req;
3104 struct obd_device *obd = exp->exp_obd;
3105 struct obd_import *imp = class_exp2cliimp(exp);
3110 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
3112 if (KEY_IS(KEY_CHECKSUM)) {
3113 if (vallen != sizeof(int))
3115 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
3119 if (KEY_IS(KEY_SPTLRPC_CONF)) {
3120 sptlrpc_conf_client_adapt(obd);
3124 if (KEY_IS(KEY_FLUSH_CTX)) {
3125 sptlrpc_import_flush_my_ctx(imp);
3129 if (KEY_IS(KEY_CACHE_SET)) {
3130 struct client_obd *cli = &obd->u.cli;
3132 LASSERT(cli->cl_cache == NULL); /* only once */
3133 cli->cl_cache = (struct cl_client_cache *)val;
3134 cfs_atomic_inc(&cli->cl_cache->ccc_users);
3135 cli->cl_lru_left = &cli->cl_cache->ccc_lru_left;
3137 /* add this osc into entity list */
3138 LASSERT(cfs_list_empty(&cli->cl_lru_osc));
3139 spin_lock(&cli->cl_cache->ccc_lru_lock);
3140 cfs_list_add(&cli->cl_lru_osc, &cli->cl_cache->ccc_lru);
3141 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3146 if (KEY_IS(KEY_CACHE_LRU_SHRINK)) {
3147 struct client_obd *cli = &obd->u.cli;
3148 int nr = cfs_atomic_read(&cli->cl_lru_in_list) >> 1;
3149 int target = *(int *)val;
3151 nr = osc_lru_shrink(cli, min(nr, target));
3156 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
3159 /* We pass all other commands directly to OST. Since nobody calls osc
3160 methods directly and everybody is supposed to go through LOV, we
3161 assume lov checked invalid values for us.
3162 The only recognised values so far are evict_by_nid and mds_conn.
3163 Even if something bad goes through, we'd get a -EINVAL from OST
3166 req = ptlrpc_request_alloc(imp, KEY_IS(KEY_GRANT_SHRINK) ?
3167 &RQF_OST_SET_GRANT_INFO :
3172 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3173 RCL_CLIENT, keylen);
3174 if (!KEY_IS(KEY_GRANT_SHRINK))
3175 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
3176 RCL_CLIENT, vallen);
3177 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
3179 ptlrpc_request_free(req);
3183 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3184 memcpy(tmp, key, keylen);
3185 tmp = req_capsule_client_get(&req->rq_pill, KEY_IS(KEY_GRANT_SHRINK) ?
3188 memcpy(tmp, val, vallen);
3190 if (KEY_IS(KEY_GRANT_SHRINK)) {
3191 struct osc_grant_args *aa;
3194 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3195 aa = ptlrpc_req_async_args(req);
3198 ptlrpc_req_finished(req);
3201 *oa = ((struct ost_body *)val)->oa;
3203 req->rq_interpret_reply = osc_shrink_grant_interpret;
3206 ptlrpc_request_set_replen(req);
3207 if (!KEY_IS(KEY_GRANT_SHRINK)) {
3208 LASSERT(set != NULL);
3209 ptlrpc_set_add_req(set, req);
3210 ptlrpc_check_set(NULL, set);
3212 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
3218 static int osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
3219 struct obd_device *disk_obd, int *index)
3221 /* this code is not supposed to be used with LOD/OSP
3222 * to be removed soon */
3227 static int osc_llog_finish(struct obd_device *obd, int count)
3229 struct llog_ctxt *ctxt;
3233 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3235 llog_cat_close(NULL, ctxt->loc_handle);
3236 llog_cleanup(NULL, ctxt);
3239 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3241 llog_cleanup(NULL, ctxt);
3245 static int osc_reconnect(const struct lu_env *env,
3246 struct obd_export *exp, struct obd_device *obd,
3247 struct obd_uuid *cluuid,
3248 struct obd_connect_data *data,
3251 struct client_obd *cli = &obd->u.cli;
3253 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
3256 client_obd_list_lock(&cli->cl_loi_list_lock);
3257 data->ocd_grant = (cli->cl_avail_grant + cli->cl_dirty) ?:
3258 2 * cli_brw_size(obd);
3259 lost_grant = cli->cl_lost_grant;
3260 cli->cl_lost_grant = 0;
3261 client_obd_list_unlock(&cli->cl_loi_list_lock);
3263 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
3264 " ocd_grant: %d, lost: %ld.\n", data->ocd_connect_flags,
3265 data->ocd_version, data->ocd_grant, lost_grant);
3271 static int osc_disconnect(struct obd_export *exp)
3273 struct obd_device *obd = class_exp2obd(exp);
3274 struct llog_ctxt *ctxt;
3277 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3279 if (obd->u.cli.cl_conn_count == 1) {
3280 /* Flush any remaining cancel messages out to the
3282 llog_sync(ctxt, exp, 0);
3284 llog_ctxt_put(ctxt);
3286 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
3290 rc = client_disconnect_export(exp);
3292 * Initially we put del_shrink_grant before disconnect_export, but it
3293 * causes the following problem if setup (connect) and cleanup
3294 * (disconnect) are tangled together.
3295 * connect p1 disconnect p2
3296 * ptlrpc_connect_import
3297 * ............... class_manual_cleanup
3300 * ptlrpc_connect_interrupt
3302 * add this client to shrink list
3304 * Bang! pinger trigger the shrink.
3305 * So the osc should be disconnected from the shrink list, after we
3306 * are sure the import has been destroyed. BUG18662
3308 if (obd->u.cli.cl_import == NULL)
3309 osc_del_shrink_grant(&obd->u.cli);
3313 static int osc_import_event(struct obd_device *obd,
3314 struct obd_import *imp,
3315 enum obd_import_event event)
3317 struct client_obd *cli;
3321 LASSERT(imp->imp_obd == obd);
3324 case IMP_EVENT_DISCON: {
3326 client_obd_list_lock(&cli->cl_loi_list_lock);
3327 cli->cl_avail_grant = 0;
3328 cli->cl_lost_grant = 0;
3329 client_obd_list_unlock(&cli->cl_loi_list_lock);
3332 case IMP_EVENT_INACTIVE: {
3333 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
3336 case IMP_EVENT_INVALIDATE: {
3337 struct ldlm_namespace *ns = obd->obd_namespace;
3341 env = cl_env_get(&refcheck);
3345 /* all pages go to failing rpcs due to the invalid
3347 osc_io_unplug(env, cli, NULL, PDL_POLICY_ROUND);
3349 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3350 cl_env_put(env, &refcheck);
3355 case IMP_EVENT_ACTIVE: {
3356 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
3359 case IMP_EVENT_OCD: {
3360 struct obd_connect_data *ocd = &imp->imp_connect_data;
3362 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
3363 osc_init_grant(&obd->u.cli, ocd);
3366 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
3367 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
3369 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
3372 case IMP_EVENT_DEACTIVATE: {
3373 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DEACTIVATE, NULL);
3376 case IMP_EVENT_ACTIVATE: {
3377 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVATE, NULL);
3381 CERROR("Unknown import event %d\n", event);
3388 * Determine whether the lock can be canceled before replaying the lock
3389 * during recovery, see bug16774 for detailed information.
3391 * \retval zero the lock can't be canceled
3392 * \retval other ok to cancel
3394 static int osc_cancel_for_recovery(struct ldlm_lock *lock)
3396 check_res_locked(lock->l_resource);
3399 * Cancel all unused extent lock in granted mode LCK_PR or LCK_CR.
3401 * XXX as a future improvement, we can also cancel unused write lock
3402 * if it doesn't have dirty data and active mmaps.
3404 if (lock->l_resource->lr_type == LDLM_EXTENT &&
3405 (lock->l_granted_mode == LCK_PR ||
3406 lock->l_granted_mode == LCK_CR) &&
3407 (osc_dlm_lock_pageref(lock) == 0))
3413 static int brw_queue_work(const struct lu_env *env, void *data)
3415 struct client_obd *cli = data;
3417 CDEBUG(D_CACHE, "Run writeback work for client obd %p.\n", cli);
3419 osc_io_unplug(env, cli, NULL, PDL_POLICY_SAME);
3423 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
3425 struct lprocfs_static_vars lvars = { 0 };
3426 struct client_obd *cli = &obd->u.cli;
3431 rc = ptlrpcd_addref();
3435 rc = client_obd_setup(obd, lcfg);
3437 GOTO(out_ptlrpcd, rc);
3439 handler = ptlrpcd_alloc_work(cli->cl_import, brw_queue_work, cli);
3440 if (IS_ERR(handler))
3441 GOTO(out_client_setup, rc = PTR_ERR(handler));
3442 cli->cl_writeback_work = handler;
3444 rc = osc_quota_setup(obd);
3446 GOTO(out_ptlrpcd_work, rc);
3448 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
3449 lprocfs_osc_init_vars(&lvars);
3450 if (lprocfs_obd_setup(obd, lvars.obd_vars) == 0) {
3451 lproc_osc_attach_seqstat(obd);
3452 sptlrpc_lprocfs_cliobd_attach(obd);
3453 ptlrpc_lprocfs_register_obd(obd);
3456 /* We need to allocate a few requests more, because
3457 * brw_interpret tries to create new requests before freeing
3458 * previous ones, Ideally we want to have 2x max_rpcs_in_flight
3459 * reserved, but I'm afraid that might be too much wasted RAM
3460 * in fact, so 2 is just my guess and still should work. */
3461 cli->cl_import->imp_rq_pool =
3462 ptlrpc_init_rq_pool(cli->cl_max_rpcs_in_flight + 2,
3464 ptlrpc_add_rqs_to_pool);
3466 CFS_INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
3467 ns_register_cancel(obd->obd_namespace, osc_cancel_for_recovery);
3471 ptlrpcd_destroy_work(handler);
3473 client_obd_cleanup(obd);
3479 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
3485 case OBD_CLEANUP_EARLY: {
3486 struct obd_import *imp;
3487 imp = obd->u.cli.cl_import;
3488 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
3489 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
3490 ptlrpc_deactivate_import(imp);
3491 spin_lock(&imp->imp_lock);
3492 imp->imp_pingable = 0;
3493 spin_unlock(&imp->imp_lock);
3496 case OBD_CLEANUP_EXPORTS: {
3497 struct client_obd *cli = &obd->u.cli;
3499 * for echo client, export may be on zombie list, wait for
3500 * zombie thread to cull it, because cli.cl_import will be
3501 * cleared in client_disconnect_export():
3502 * class_export_destroy() -> obd_cleanup() ->
3503 * echo_device_free() -> echo_client_cleanup() ->
3504 * obd_disconnect() -> osc_disconnect() ->
3505 * client_disconnect_export()
3507 obd_zombie_barrier();
3508 if (cli->cl_writeback_work) {
3509 ptlrpcd_destroy_work(cli->cl_writeback_work);
3510 cli->cl_writeback_work = NULL;
3512 obd_cleanup_client_import(obd);
3513 ptlrpc_lprocfs_unregister_obd(obd);
3514 lprocfs_obd_cleanup(obd);
3515 rc = obd_llog_finish(obd, 0);
3517 CERROR("failed to cleanup llogging subsystems\n");
3524 int osc_cleanup(struct obd_device *obd)
3526 struct client_obd *cli = &obd->u.cli;
3532 if (cli->cl_cache != NULL) {
3533 LASSERT(cfs_atomic_read(&cli->cl_cache->ccc_users) > 0);
3534 spin_lock(&cli->cl_cache->ccc_lru_lock);
3535 cfs_list_del_init(&cli->cl_lru_osc);
3536 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3537 cli->cl_lru_left = NULL;
3538 cfs_atomic_dec(&cli->cl_cache->ccc_users);
3539 cli->cl_cache = NULL;
3542 /* free memory of osc quota cache */
3543 osc_quota_cleanup(obd);
3545 rc = client_obd_cleanup(obd);
3551 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
3553 struct lprocfs_static_vars lvars = { 0 };
3556 lprocfs_osc_init_vars(&lvars);
3558 switch (lcfg->lcfg_command) {
3560 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars,
3570 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
3572 return osc_process_config_base(obd, buf);
3575 struct obd_ops osc_obd_ops = {
3576 .o_owner = THIS_MODULE,
3577 .o_setup = osc_setup,
3578 .o_precleanup = osc_precleanup,
3579 .o_cleanup = osc_cleanup,
3580 .o_add_conn = client_import_add_conn,
3581 .o_del_conn = client_import_del_conn,
3582 .o_connect = client_connect_import,
3583 .o_reconnect = osc_reconnect,
3584 .o_disconnect = osc_disconnect,
3585 .o_statfs = osc_statfs,
3586 .o_statfs_async = osc_statfs_async,
3587 .o_packmd = osc_packmd,
3588 .o_unpackmd = osc_unpackmd,
3589 .o_create = osc_create,
3590 .o_destroy = osc_destroy,
3591 .o_getattr = osc_getattr,
3592 .o_getattr_async = osc_getattr_async,
3593 .o_setattr = osc_setattr,
3594 .o_setattr_async = osc_setattr_async,
3596 .o_punch = osc_punch,
3598 .o_enqueue = osc_enqueue,
3599 .o_change_cbdata = osc_change_cbdata,
3600 .o_find_cbdata = osc_find_cbdata,
3601 .o_cancel = osc_cancel,
3602 .o_cancel_unused = osc_cancel_unused,
3603 .o_iocontrol = osc_iocontrol,
3604 .o_get_info = osc_get_info,
3605 .o_set_info_async = osc_set_info_async,
3606 .o_import_event = osc_import_event,
3607 .o_llog_init = osc_llog_init,
3608 .o_llog_finish = osc_llog_finish,
3609 .o_process_config = osc_process_config,
3610 .o_quotactl = osc_quotactl,
3611 .o_quotacheck = osc_quotacheck,
3614 extern struct lu_kmem_descr osc_caches[];
3615 extern spinlock_t osc_ast_guard;
3616 extern struct lock_class_key osc_ast_guard_class;
3618 int __init osc_init(void)
3620 struct lprocfs_static_vars lvars = { 0 };
3624 /* print an address of _any_ initialized kernel symbol from this
3625 * module, to allow debugging with gdb that doesn't support data
3626 * symbols from modules.*/
3627 CDEBUG(D_INFO, "Lustre OSC module (%p).\n", &osc_caches);
3629 rc = lu_kmem_init(osc_caches);
3631 lprocfs_osc_init_vars(&lvars);
3633 rc = class_register_type(&osc_obd_ops, NULL, lvars.module_vars,
3634 LUSTRE_OSC_NAME, &osc_device_type);
3636 lu_kmem_fini(osc_caches);
3640 spin_lock_init(&osc_ast_guard);
3641 lockdep_set_class(&osc_ast_guard, &osc_ast_guard_class);
3647 static void /*__exit*/ osc_exit(void)
3649 class_unregister_type(LUSTRE_OSC_NAME);
3650 lu_kmem_fini(osc_caches);
3653 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
3654 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
3655 MODULE_LICENSE("GPL");
3657 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);
git://git.whamcloud.com / fs / lustre-release.git / blob