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, 2013, 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(&req->rq_import->imp_connect_data, &body->oa,
189 osc_pack_capa(req, body, oinfo->oi_capa);
192 static inline void osc_set_capa_size(struct ptlrpc_request *req,
193 const struct req_msg_field *field,
197 req_capsule_set_size(&req->rq_pill, field, RCL_CLIENT, 0);
199 /* it is already calculated as sizeof struct obd_capa */
203 static int osc_getattr_interpret(const struct lu_env *env,
204 struct ptlrpc_request *req,
205 struct osc_async_args *aa, int rc)
207 struct ost_body *body;
213 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
215 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
216 lustre_get_wire_obdo(&req->rq_import->imp_connect_data,
217 aa->aa_oi->oi_oa, &body->oa);
219 /* This should really be sent by the OST */
220 aa->aa_oi->oi_oa->o_blksize = DT_MAX_BRW_SIZE;
221 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
223 CDEBUG(D_INFO, "can't unpack ost_body\n");
225 aa->aa_oi->oi_oa->o_valid = 0;
228 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
232 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
233 struct ptlrpc_request_set *set)
235 struct ptlrpc_request *req;
236 struct osc_async_args *aa;
240 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
244 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
245 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
247 ptlrpc_request_free(req);
251 osc_pack_req_body(req, oinfo);
253 ptlrpc_request_set_replen(req);
254 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
256 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
257 aa = ptlrpc_req_async_args(req);
260 ptlrpc_set_add_req(set, req);
264 static int osc_getattr(const struct lu_env *env, struct obd_export *exp,
265 struct obd_info *oinfo)
267 struct ptlrpc_request *req;
268 struct ost_body *body;
272 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
276 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
277 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
279 ptlrpc_request_free(req);
283 osc_pack_req_body(req, oinfo);
285 ptlrpc_request_set_replen(req);
287 rc = ptlrpc_queue_wait(req);
291 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
293 GOTO(out, rc = -EPROTO);
295 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
296 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oinfo->oi_oa,
299 oinfo->oi_oa->o_blksize = cli_brw_size(exp->exp_obd);
300 oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
304 ptlrpc_req_finished(req);
308 static int osc_setattr(const struct lu_env *env, struct obd_export *exp,
309 struct obd_info *oinfo, struct obd_trans_info *oti)
311 struct ptlrpc_request *req;
312 struct ost_body *body;
316 LASSERT(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP);
318 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
322 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
323 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
325 ptlrpc_request_free(req);
329 osc_pack_req_body(req, oinfo);
331 ptlrpc_request_set_replen(req);
333 rc = ptlrpc_queue_wait(req);
337 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
339 GOTO(out, rc = -EPROTO);
341 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oinfo->oi_oa,
346 ptlrpc_req_finished(req);
350 static int osc_setattr_interpret(const struct lu_env *env,
351 struct ptlrpc_request *req,
352 struct osc_setattr_args *sa, int rc)
354 struct ost_body *body;
360 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
362 GOTO(out, rc = -EPROTO);
364 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, sa->sa_oa,
367 rc = sa->sa_upcall(sa->sa_cookie, rc);
371 int osc_setattr_async_base(struct obd_export *exp, struct obd_info *oinfo,
372 struct obd_trans_info *oti,
373 obd_enqueue_update_f upcall, void *cookie,
374 struct ptlrpc_request_set *rqset)
376 struct ptlrpc_request *req;
377 struct osc_setattr_args *sa;
381 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
385 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
386 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
388 ptlrpc_request_free(req);
392 if (oti && oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
393 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
395 osc_pack_req_body(req, oinfo);
397 ptlrpc_request_set_replen(req);
399 /* do mds to ost setattr asynchronously */
401 /* Do not wait for response. */
402 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
404 req->rq_interpret_reply =
405 (ptlrpc_interpterer_t)osc_setattr_interpret;
407 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
408 sa = ptlrpc_req_async_args(req);
409 sa->sa_oa = oinfo->oi_oa;
410 sa->sa_upcall = upcall;
411 sa->sa_cookie = cookie;
413 if (rqset == PTLRPCD_SET)
414 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
416 ptlrpc_set_add_req(rqset, req);
422 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
423 struct obd_trans_info *oti,
424 struct ptlrpc_request_set *rqset)
426 return osc_setattr_async_base(exp, oinfo, oti,
427 oinfo->oi_cb_up, oinfo, rqset);
430 int osc_real_create(struct obd_export *exp, struct obdo *oa,
431 struct lov_stripe_md **ea, struct obd_trans_info *oti)
433 struct ptlrpc_request *req;
434 struct ost_body *body;
435 struct lov_stripe_md *lsm;
444 rc = obd_alloc_memmd(exp, &lsm);
449 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
451 GOTO(out, rc = -ENOMEM);
453 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
455 ptlrpc_request_free(req);
459 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
462 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
464 ptlrpc_request_set_replen(req);
466 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
467 oa->o_flags == OBD_FL_DELORPHAN) {
469 "delorphan from OST integration");
470 /* Don't resend the delorphan req */
471 req->rq_no_resend = req->rq_no_delay = 1;
474 rc = ptlrpc_queue_wait(req);
478 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
480 GOTO(out_req, rc = -EPROTO);
482 CDEBUG(D_INFO, "oa flags %x\n", oa->o_flags);
483 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
485 oa->o_blksize = cli_brw_size(exp->exp_obd);
486 oa->o_valid |= OBD_MD_FLBLKSZ;
488 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
489 * have valid lsm_oinfo data structs, so don't go touching that.
490 * This needs to be fixed in a big way.
492 lsm->lsm_oi = oa->o_oi;
496 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
498 if (oa->o_valid & OBD_MD_FLCOOKIE) {
499 if (!oti->oti_logcookies)
500 oti_alloc_cookies(oti, 1);
501 *oti->oti_logcookies = oa->o_lcookie;
505 CDEBUG(D_HA, "transno: "LPD64"\n",
506 lustre_msg_get_transno(req->rq_repmsg));
508 ptlrpc_req_finished(req);
511 obd_free_memmd(exp, &lsm);
515 int osc_punch_base(struct obd_export *exp, struct obd_info *oinfo,
516 obd_enqueue_update_f upcall, void *cookie,
517 struct ptlrpc_request_set *rqset)
519 struct ptlrpc_request *req;
520 struct osc_setattr_args *sa;
521 struct ost_body *body;
525 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
529 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
530 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
532 ptlrpc_request_free(req);
535 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
536 ptlrpc_at_set_req_timeout(req);
538 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
540 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
542 osc_pack_capa(req, body, oinfo->oi_capa);
544 ptlrpc_request_set_replen(req);
546 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
547 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
548 sa = ptlrpc_req_async_args(req);
549 sa->sa_oa = oinfo->oi_oa;
550 sa->sa_upcall = upcall;
551 sa->sa_cookie = cookie;
552 if (rqset == PTLRPCD_SET)
553 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
555 ptlrpc_set_add_req(rqset, req);
560 static int osc_punch(const struct lu_env *env, struct obd_export *exp,
561 struct obd_info *oinfo, struct obd_trans_info *oti,
562 struct ptlrpc_request_set *rqset)
564 oinfo->oi_oa->o_size = oinfo->oi_policy.l_extent.start;
565 oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
566 oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
567 return osc_punch_base(exp, oinfo,
568 oinfo->oi_cb_up, oinfo, rqset);
571 static int osc_sync_interpret(const struct lu_env *env,
572 struct ptlrpc_request *req,
575 struct osc_fsync_args *fa = arg;
576 struct ost_body *body;
582 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
584 CERROR ("can't unpack ost_body\n");
585 GOTO(out, rc = -EPROTO);
588 *fa->fa_oi->oi_oa = body->oa;
590 rc = fa->fa_upcall(fa->fa_cookie, rc);
594 int osc_sync_base(struct obd_export *exp, struct obd_info *oinfo,
595 obd_enqueue_update_f upcall, void *cookie,
596 struct ptlrpc_request_set *rqset)
598 struct ptlrpc_request *req;
599 struct ost_body *body;
600 struct osc_fsync_args *fa;
604 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
608 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
609 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
611 ptlrpc_request_free(req);
615 /* overload the size and blocks fields in the oa with start/end */
616 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
618 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
620 osc_pack_capa(req, body, oinfo->oi_capa);
622 ptlrpc_request_set_replen(req);
623 req->rq_interpret_reply = osc_sync_interpret;
625 CLASSERT(sizeof(*fa) <= sizeof(req->rq_async_args));
626 fa = ptlrpc_req_async_args(req);
628 fa->fa_upcall = upcall;
629 fa->fa_cookie = cookie;
631 if (rqset == PTLRPCD_SET)
632 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
634 ptlrpc_set_add_req(rqset, req);
639 static int osc_sync(const struct lu_env *env, struct obd_export *exp,
640 struct obd_info *oinfo, obd_size start, obd_size end,
641 struct ptlrpc_request_set *set)
646 CDEBUG(D_INFO, "oa NULL\n");
650 oinfo->oi_oa->o_size = start;
651 oinfo->oi_oa->o_blocks = end;
652 oinfo->oi_oa->o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
654 RETURN(osc_sync_base(exp, oinfo, oinfo->oi_cb_up, oinfo, set));
657 /* Find and cancel locally locks matched by @mode in the resource found by
658 * @objid. Found locks are added into @cancel list. Returns the amount of
659 * locks added to @cancels list. */
660 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
662 ldlm_mode_t mode, __u64 lock_flags)
664 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
665 struct ldlm_res_id res_id;
666 struct ldlm_resource *res;
670 /* Return, i.e. cancel nothing, only if ELC is supported (flag in
671 * export) but disabled through procfs (flag in NS).
673 * This distinguishes from a case when ELC is not supported originally,
674 * when we still want to cancel locks in advance and just cancel them
675 * locally, without sending any RPC. */
676 if (exp_connect_cancelset(exp) && !ns_connect_cancelset(ns))
679 ostid_build_res_name(&oa->o_oi, &res_id);
680 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
684 LDLM_RESOURCE_ADDREF(res);
685 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
686 lock_flags, 0, NULL);
687 LDLM_RESOURCE_DELREF(res);
688 ldlm_resource_putref(res);
692 static int osc_destroy_interpret(const struct lu_env *env,
693 struct ptlrpc_request *req, void *data,
696 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
698 atomic_dec(&cli->cl_destroy_in_flight);
699 wake_up(&cli->cl_destroy_waitq);
703 static int osc_can_send_destroy(struct client_obd *cli)
705 if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
706 cli->cl_max_rpcs_in_flight) {
707 /* The destroy request can be sent */
710 if (atomic_dec_return(&cli->cl_destroy_in_flight) <
711 cli->cl_max_rpcs_in_flight) {
713 * The counter has been modified between the two atomic
716 wake_up(&cli->cl_destroy_waitq);
721 int osc_create(const struct lu_env *env, struct obd_export *exp,
722 struct obdo *oa, struct lov_stripe_md **ea,
723 struct obd_trans_info *oti)
730 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
732 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
733 oa->o_flags == OBD_FL_RECREATE_OBJS) {
734 RETURN(osc_real_create(exp, oa, ea, oti));
737 if (!fid_seq_is_mdt(ostid_seq(&oa->o_oi)))
738 RETURN(osc_real_create(exp, oa, ea, oti));
740 /* we should not get here anymore */
746 /* Destroy requests can be async always on the client, and we don't even really
747 * care about the return code since the client cannot do anything at all about
749 * When the MDS is unlinking a filename, it saves the file objects into a
750 * recovery llog, and these object records are cancelled when the OST reports
751 * they were destroyed and sync'd to disk (i.e. transaction committed).
752 * If the client dies, or the OST is down when the object should be destroyed,
753 * the records are not cancelled, and when the OST reconnects to the MDS next,
754 * it will retrieve the llog unlink logs and then sends the log cancellation
755 * cookies to the MDS after committing destroy transactions. */
756 static int osc_destroy(const struct lu_env *env, struct obd_export *exp,
757 struct obdo *oa, struct lov_stripe_md *ea,
758 struct obd_trans_info *oti, struct obd_export *md_export,
761 struct client_obd *cli = &exp->exp_obd->u.cli;
762 struct ptlrpc_request *req;
763 struct ost_body *body;
764 CFS_LIST_HEAD(cancels);
769 CDEBUG(D_INFO, "oa NULL\n");
773 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
774 LDLM_FL_DISCARD_DATA);
776 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
778 ldlm_lock_list_put(&cancels, l_bl_ast, count);
782 osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
783 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
786 ptlrpc_request_free(req);
790 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
791 ptlrpc_at_set_req_timeout(req);
793 if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
794 oa->o_lcookie = *oti->oti_logcookies;
795 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
797 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
799 osc_pack_capa(req, body, (struct obd_capa *)capa);
800 ptlrpc_request_set_replen(req);
802 /* If osc_destory is for destroying the unlink orphan,
803 * sent from MDT to OST, which should not be blocked here,
804 * because the process might be triggered by ptlrpcd, and
805 * it is not good to block ptlrpcd thread (b=16006)*/
806 if (!(oa->o_flags & OBD_FL_DELORPHAN)) {
807 req->rq_interpret_reply = osc_destroy_interpret;
808 if (!osc_can_send_destroy(cli)) {
809 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP,
813 * Wait until the number of on-going destroy RPCs drops
814 * under max_rpc_in_flight
816 l_wait_event_exclusive(cli->cl_destroy_waitq,
817 osc_can_send_destroy(cli), &lwi);
821 /* Do not wait for response */
822 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
826 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
829 obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
831 LASSERT(!(oa->o_valid & bits));
834 client_obd_list_lock(&cli->cl_loi_list_lock);
835 oa->o_dirty = cli->cl_dirty;
836 if (unlikely(cli->cl_dirty - cli->cl_dirty_transit >
837 cli->cl_dirty_max)) {
838 CERROR("dirty %lu - %lu > dirty_max %lu\n",
839 cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
841 } else if (unlikely(atomic_read(&obd_unstable_pages) +
842 atomic_read(&obd_dirty_pages) -
843 atomic_read(&obd_dirty_transit_pages) >
844 (long)(obd_max_dirty_pages + 1))) {
845 /* The atomic_read() allowing the atomic_inc() are
846 * not covered by a lock thus they may safely race and trip
847 * this CERROR() unless we add in a small fudge factor (+1). */
848 CERROR("%s: dirty %d + %d - %d > system dirty_max %d\n",
849 cli->cl_import->imp_obd->obd_name,
850 atomic_read(&obd_unstable_pages),
851 atomic_read(&obd_dirty_pages),
852 atomic_read(&obd_dirty_transit_pages),
853 obd_max_dirty_pages);
855 } else if (unlikely(cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff)) {
856 CERROR("dirty %lu - dirty_max %lu too big???\n",
857 cli->cl_dirty, cli->cl_dirty_max);
860 long max_in_flight = (cli->cl_max_pages_per_rpc <<
862 (cli->cl_max_rpcs_in_flight + 1);
863 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
865 oa->o_grant = cli->cl_avail_grant + cli->cl_reserved_grant;
866 oa->o_dropped = cli->cl_lost_grant;
867 cli->cl_lost_grant = 0;
868 client_obd_list_unlock(&cli->cl_loi_list_lock);
869 CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
870 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
874 void osc_update_next_shrink(struct client_obd *cli)
876 cli->cl_next_shrink_grant =
877 cfs_time_shift(cli->cl_grant_shrink_interval);
878 CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
879 cli->cl_next_shrink_grant);
882 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
884 client_obd_list_lock(&cli->cl_loi_list_lock);
885 cli->cl_avail_grant += grant;
886 client_obd_list_unlock(&cli->cl_loi_list_lock);
889 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
891 if (body->oa.o_valid & OBD_MD_FLGRANT) {
892 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
893 __osc_update_grant(cli, body->oa.o_grant);
897 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
898 obd_count keylen, void *key, obd_count vallen,
899 void *val, struct ptlrpc_request_set *set);
901 static int osc_shrink_grant_interpret(const struct lu_env *env,
902 struct ptlrpc_request *req,
905 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
906 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
907 struct ost_body *body;
910 __osc_update_grant(cli, oa->o_grant);
914 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
916 osc_update_grant(cli, body);
922 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
924 client_obd_list_lock(&cli->cl_loi_list_lock);
925 oa->o_grant = cli->cl_avail_grant / 4;
926 cli->cl_avail_grant -= oa->o_grant;
927 client_obd_list_unlock(&cli->cl_loi_list_lock);
928 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
929 oa->o_valid |= OBD_MD_FLFLAGS;
932 oa->o_flags |= OBD_FL_SHRINK_GRANT;
933 osc_update_next_shrink(cli);
936 /* Shrink the current grant, either from some large amount to enough for a
937 * full set of in-flight RPCs, or if we have already shrunk to that limit
938 * then to enough for a single RPC. This avoids keeping more grant than
939 * needed, and avoids shrinking the grant piecemeal. */
940 static int osc_shrink_grant(struct client_obd *cli)
942 __u64 target_bytes = (cli->cl_max_rpcs_in_flight + 1) *
943 (cli->cl_max_pages_per_rpc << PAGE_CACHE_SHIFT);
945 client_obd_list_lock(&cli->cl_loi_list_lock);
946 if (cli->cl_avail_grant <= target_bytes)
947 target_bytes = cli->cl_max_pages_per_rpc << PAGE_CACHE_SHIFT;
948 client_obd_list_unlock(&cli->cl_loi_list_lock);
950 return osc_shrink_grant_to_target(cli, target_bytes);
953 int osc_shrink_grant_to_target(struct client_obd *cli, __u64 target_bytes)
956 struct ost_body *body;
959 client_obd_list_lock(&cli->cl_loi_list_lock);
960 /* Don't shrink if we are already above or below the desired limit
961 * We don't want to shrink below a single RPC, as that will negatively
962 * impact block allocation and long-term performance. */
963 if (target_bytes < cli->cl_max_pages_per_rpc << PAGE_CACHE_SHIFT)
964 target_bytes = cli->cl_max_pages_per_rpc << PAGE_CACHE_SHIFT;
966 if (target_bytes >= cli->cl_avail_grant) {
967 client_obd_list_unlock(&cli->cl_loi_list_lock);
970 client_obd_list_unlock(&cli->cl_loi_list_lock);
976 osc_announce_cached(cli, &body->oa, 0);
978 client_obd_list_lock(&cli->cl_loi_list_lock);
979 body->oa.o_grant = cli->cl_avail_grant - target_bytes;
980 cli->cl_avail_grant = target_bytes;
981 client_obd_list_unlock(&cli->cl_loi_list_lock);
982 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
983 body->oa.o_valid |= OBD_MD_FLFLAGS;
984 body->oa.o_flags = 0;
986 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
987 osc_update_next_shrink(cli);
989 rc = osc_set_info_async(NULL, cli->cl_import->imp_obd->obd_self_export,
990 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
991 sizeof(*body), body, NULL);
993 __osc_update_grant(cli, body->oa.o_grant);
998 static int osc_should_shrink_grant(struct client_obd *client)
1000 cfs_time_t time = cfs_time_current();
1001 cfs_time_t next_shrink = client->cl_next_shrink_grant;
1003 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
1004 OBD_CONNECT_GRANT_SHRINK) == 0)
1007 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
1008 /* Get the current RPC size directly, instead of going via:
1009 * cli_brw_size(obd->u.cli.cl_import->imp_obd->obd_self_export)
1010 * Keep comment here so that it can be found by searching. */
1011 int brw_size = client->cl_max_pages_per_rpc << PAGE_CACHE_SHIFT;
1013 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
1014 client->cl_avail_grant > brw_size)
1017 osc_update_next_shrink(client);
1022 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1024 struct client_obd *client;
1026 cfs_list_for_each_entry(client, &item->ti_obd_list,
1027 cl_grant_shrink_list) {
1028 if (osc_should_shrink_grant(client))
1029 osc_shrink_grant(client);
1034 static int osc_add_shrink_grant(struct client_obd *client)
1038 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
1040 osc_grant_shrink_grant_cb, NULL,
1041 &client->cl_grant_shrink_list);
1043 CERROR("add grant client %s error %d\n",
1044 client->cl_import->imp_obd->obd_name, rc);
1047 CDEBUG(D_CACHE, "add grant client %s \n",
1048 client->cl_import->imp_obd->obd_name);
1049 osc_update_next_shrink(client);
1053 static int osc_del_shrink_grant(struct client_obd *client)
1055 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1059 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1062 * ocd_grant is the total grant amount we're expect to hold: if we've
1063 * been evicted, it's the new avail_grant amount, cl_dirty will drop
1064 * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1066 * race is tolerable here: if we're evicted, but imp_state already
1067 * left EVICTED state, then cl_dirty must be 0 already.
1069 client_obd_list_lock(&cli->cl_loi_list_lock);
1070 if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1071 cli->cl_avail_grant = ocd->ocd_grant;
1073 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1075 if (cli->cl_avail_grant < 0) {
1076 CWARN("%s: available grant < 0: avail/ocd/dirty %ld/%u/%ld\n",
1077 cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant,
1078 ocd->ocd_grant, cli->cl_dirty);
1079 /* workaround for servers which do not have the patch from
1081 cli->cl_avail_grant = ocd->ocd_grant;
1084 /* determine the appropriate chunk size used by osc_extent. */
1085 cli->cl_chunkbits = max_t(int, PAGE_CACHE_SHIFT, ocd->ocd_blocksize);
1086 client_obd_list_unlock(&cli->cl_loi_list_lock);
1088 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld."
1089 "chunk bits: %d.\n", cli->cl_import->imp_obd->obd_name,
1090 cli->cl_avail_grant, cli->cl_lost_grant, cli->cl_chunkbits);
1092 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1093 cfs_list_empty(&cli->cl_grant_shrink_list))
1094 osc_add_shrink_grant(cli);
1097 /* We assume that the reason this OSC got a short read is because it read
1098 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1099 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1100 * this stripe never got written at or beyond this stripe offset yet. */
1101 static void handle_short_read(int nob_read, obd_count page_count,
1102 struct brw_page **pga)
1107 /* skip bytes read OK */
1108 while (nob_read > 0) {
1109 LASSERT (page_count > 0);
1111 if (pga[i]->count > nob_read) {
1112 /* EOF inside this page */
1113 ptr = kmap(pga[i]->pg) +
1114 (pga[i]->off & ~CFS_PAGE_MASK);
1115 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1122 nob_read -= pga[i]->count;
1127 /* zero remaining pages */
1128 while (page_count-- > 0) {
1129 ptr = kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1130 memset(ptr, 0, pga[i]->count);
1136 static int check_write_rcs(struct ptlrpc_request *req,
1137 int requested_nob, int niocount,
1138 obd_count page_count, struct brw_page **pga)
1143 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1144 sizeof(*remote_rcs) *
1146 if (remote_rcs == NULL) {
1147 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1151 /* return error if any niobuf was in error */
1152 for (i = 0; i < niocount; i++) {
1153 if ((int)remote_rcs[i] < 0)
1154 return(remote_rcs[i]);
1156 if (remote_rcs[i] != 0) {
1157 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1158 i, remote_rcs[i], req);
1163 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1164 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1165 req->rq_bulk->bd_nob_transferred, requested_nob);
1172 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1174 if (p1->flag != p2->flag) {
1175 unsigned mask = ~(OBD_BRW_FROM_GRANT | OBD_BRW_NOCACHE |
1176 OBD_BRW_SYNC | OBD_BRW_ASYNC |
1177 OBD_BRW_NOQUOTA | OBD_BRW_SOFT_SYNC);
1179 /* warn if we try to combine flags that we don't know to be
1180 * safe to combine */
1181 if (unlikely((p1->flag & mask) != (p2->flag & mask))) {
1182 CWARN("Saw flags 0x%x and 0x%x in the same brw, please "
1183 "report this at http://bugs.whamcloud.com/\n",
1184 p1->flag, p2->flag);
1189 return (p1->off + p1->count == p2->off);
1192 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1193 struct brw_page **pga, int opc,
1194 cksum_type_t cksum_type)
1198 struct cfs_crypto_hash_desc *hdesc;
1199 unsigned int bufsize;
1201 unsigned char cfs_alg = cksum_obd2cfs(cksum_type);
1203 LASSERT(pg_count > 0);
1205 hdesc = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1206 if (IS_ERR(hdesc)) {
1207 CERROR("Unable to initialize checksum hash %s\n",
1208 cfs_crypto_hash_name(cfs_alg));
1209 return PTR_ERR(hdesc);
1212 while (nob > 0 && pg_count > 0) {
1213 int count = pga[i]->count > nob ? nob : pga[i]->count;
1215 /* corrupt the data before we compute the checksum, to
1216 * simulate an OST->client data error */
1217 if (i == 0 && opc == OST_READ &&
1218 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE)) {
1219 unsigned char *ptr = kmap(pga[i]->pg);
1220 int off = pga[i]->off & ~CFS_PAGE_MASK;
1221 memcpy(ptr + off, "bad1", min(4, nob));
1224 cfs_crypto_hash_update_page(hdesc, pga[i]->pg,
1225 pga[i]->off & ~CFS_PAGE_MASK,
1227 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d\n",
1228 (int)(pga[i]->off & ~CFS_PAGE_MASK));
1230 nob -= pga[i]->count;
1236 err = cfs_crypto_hash_final(hdesc, (unsigned char *)&cksum, &bufsize);
1239 cfs_crypto_hash_final(hdesc, NULL, NULL);
1241 /* For sending we only compute the wrong checksum instead
1242 * of corrupting the data so it is still correct on a redo */
1243 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1249 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1250 struct lov_stripe_md *lsm, obd_count page_count,
1251 struct brw_page **pga,
1252 struct ptlrpc_request **reqp,
1253 struct obd_capa *ocapa, int reserve,
1256 struct ptlrpc_request *req;
1257 struct ptlrpc_bulk_desc *desc;
1258 struct ost_body *body;
1259 struct obd_ioobj *ioobj;
1260 struct niobuf_remote *niobuf;
1261 int niocount, i, requested_nob, opc, rc;
1262 struct osc_brw_async_args *aa;
1263 struct req_capsule *pill;
1264 struct brw_page *pg_prev;
1267 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1268 RETURN(-ENOMEM); /* Recoverable */
1269 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1270 RETURN(-EINVAL); /* Fatal */
1272 if ((cmd & OBD_BRW_WRITE) != 0) {
1274 req = ptlrpc_request_alloc_pool(cli->cl_import,
1275 cli->cl_import->imp_rq_pool,
1276 &RQF_OST_BRW_WRITE);
1279 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1284 for (niocount = i = 1; i < page_count; i++) {
1285 if (!can_merge_pages(pga[i - 1], pga[i]))
1289 pill = &req->rq_pill;
1290 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1292 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1293 niocount * sizeof(*niobuf));
1294 osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1296 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1298 ptlrpc_request_free(req);
1301 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1302 ptlrpc_at_set_req_timeout(req);
1303 /* ask ptlrpc not to resend on EINPROGRESS since BRWs have their own
1305 req->rq_no_retry_einprogress = 1;
1307 desc = ptlrpc_prep_bulk_imp(req, page_count,
1308 cli->cl_import->imp_connect_data.ocd_brw_size >> LNET_MTU_BITS,
1309 opc == OST_WRITE ? BULK_GET_SOURCE : BULK_PUT_SINK,
1313 GOTO(out, rc = -ENOMEM);
1314 /* NB request now owns desc and will free it when it gets freed */
1316 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1317 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1318 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1319 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1321 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
1323 obdo_to_ioobj(oa, ioobj);
1324 ioobj->ioo_bufcnt = niocount;
1325 /* The high bits of ioo_max_brw tells server _maximum_ number of bulks
1326 * that might be send for this request. The actual number is decided
1327 * when the RPC is finally sent in ptlrpc_register_bulk(). It sends
1328 * "max - 1" for old client compatibility sending "0", and also so the
1329 * the actual maximum is a power-of-two number, not one less. LU-1431 */
1330 ioobj_max_brw_set(ioobj, desc->bd_md_max_brw);
1331 osc_pack_capa(req, body, ocapa);
1332 LASSERT(page_count > 0);
1334 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1335 struct brw_page *pg = pga[i];
1336 int poff = pg->off & ~CFS_PAGE_MASK;
1338 LASSERT(pg->count > 0);
1339 /* make sure there is no gap in the middle of page array */
1340 LASSERTF(page_count == 1 ||
1341 (ergo(i == 0, poff + pg->count == PAGE_CACHE_SIZE) &&
1342 ergo(i > 0 && i < page_count - 1,
1343 poff == 0 && pg->count == PAGE_CACHE_SIZE) &&
1344 ergo(i == page_count - 1, poff == 0)),
1345 "i: %d/%d pg: %p off: "LPU64", count: %u\n",
1346 i, page_count, pg, pg->off, pg->count);
1348 LASSERTF(i == 0 || pg->off > pg_prev->off,
1349 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1350 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1352 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1353 pg_prev->pg, page_private(pg_prev->pg),
1354 pg_prev->pg->index, pg_prev->off);
1356 LASSERTF(i == 0 || pg->off > pg_prev->off,
1357 "i %d p_c %u\n", i, page_count);
1359 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1360 (pg->flag & OBD_BRW_SRVLOCK));
1362 ptlrpc_prep_bulk_page_pin(desc, pg->pg, poff, pg->count);
1363 requested_nob += pg->count;
1365 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1367 niobuf->len += pg->count;
1369 niobuf->offset = pg->off;
1370 niobuf->len = pg->count;
1371 niobuf->flags = pg->flag;
1376 LASSERTF((void *)(niobuf - niocount) ==
1377 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1378 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1379 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1381 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1383 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1384 body->oa.o_valid |= OBD_MD_FLFLAGS;
1385 body->oa.o_flags = 0;
1387 body->oa.o_flags |= OBD_FL_RECOV_RESEND;
1390 if (osc_should_shrink_grant(cli))
1391 osc_shrink_grant_local(cli, &body->oa);
1393 /* size[REQ_REC_OFF] still sizeof (*body) */
1394 if (opc == OST_WRITE) {
1395 if (cli->cl_checksum &&
1396 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1397 /* store cl_cksum_type in a local variable since
1398 * it can be changed via lprocfs */
1399 cksum_type_t cksum_type = cli->cl_cksum_type;
1401 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1402 oa->o_flags &= OBD_FL_LOCAL_MASK;
1403 body->oa.o_flags = 0;
1405 body->oa.o_flags |= cksum_type_pack(cksum_type);
1406 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1407 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1411 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1413 /* save this in 'oa', too, for later checking */
1414 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1415 oa->o_flags |= cksum_type_pack(cksum_type);
1417 /* clear out the checksum flag, in case this is a
1418 * resend but cl_checksum is no longer set. b=11238 */
1419 oa->o_valid &= ~OBD_MD_FLCKSUM;
1421 oa->o_cksum = body->oa.o_cksum;
1422 /* 1 RC per niobuf */
1423 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1424 sizeof(__u32) * niocount);
1426 if (cli->cl_checksum &&
1427 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1428 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1429 body->oa.o_flags = 0;
1430 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1431 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1434 ptlrpc_request_set_replen(req);
1436 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1437 aa = ptlrpc_req_async_args(req);
1439 aa->aa_requested_nob = requested_nob;
1440 aa->aa_nio_count = niocount;
1441 aa->aa_page_count = page_count;
1445 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1446 if (ocapa && reserve)
1447 aa->aa_ocapa = capa_get(ocapa);
1453 ptlrpc_req_finished(req);
1457 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1458 __u32 client_cksum, __u32 server_cksum, int nob,
1459 obd_count page_count, struct brw_page **pga,
1460 cksum_type_t client_cksum_type)
1464 cksum_type_t cksum_type;
1466 if (server_cksum == client_cksum) {
1467 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1471 cksum_type = cksum_type_unpack(oa->o_valid & OBD_MD_FLFLAGS ?
1473 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1476 if (cksum_type != client_cksum_type)
1477 msg = "the server did not use the checksum type specified in "
1478 "the original request - likely a protocol problem";
1479 else if (new_cksum == server_cksum)
1480 msg = "changed on the client after we checksummed it - "
1481 "likely false positive due to mmap IO (bug 11742)";
1482 else if (new_cksum == client_cksum)
1483 msg = "changed in transit before arrival at OST";
1485 msg = "changed in transit AND doesn't match the original - "
1486 "likely false positive due to mmap IO (bug 11742)";
1488 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1489 " object "DOSTID" extent ["LPU64"-"LPU64"]\n",
1490 msg, libcfs_nid2str(peer->nid),
1491 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1492 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1493 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1494 POSTID(&oa->o_oi), pga[0]->off,
1495 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1496 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1497 "client csum now %x\n", client_cksum, client_cksum_type,
1498 server_cksum, cksum_type, new_cksum);
1502 /* Note rc enters this function as number of bytes transferred */
1503 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1505 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1506 const lnet_process_id_t *peer =
1507 &req->rq_import->imp_connection->c_peer;
1508 struct client_obd *cli = aa->aa_cli;
1509 struct ost_body *body;
1510 __u32 client_cksum = 0;
1513 if (rc < 0 && rc != -EDQUOT) {
1514 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1518 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1519 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1521 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1525 /* set/clear over quota flag for a uid/gid */
1526 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1527 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1528 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1530 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid "LPX64", flags %x\n",
1531 body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1533 osc_quota_setdq(cli, qid, body->oa.o_valid, body->oa.o_flags);
1536 osc_update_grant(cli, body);
1541 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1542 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1544 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1546 CERROR("Unexpected +ve rc %d\n", rc);
1549 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1551 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1554 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1555 check_write_checksum(&body->oa, peer, client_cksum,
1556 body->oa.o_cksum, aa->aa_requested_nob,
1557 aa->aa_page_count, aa->aa_ppga,
1558 cksum_type_unpack(aa->aa_oa->o_flags)))
1561 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1562 aa->aa_page_count, aa->aa_ppga);
1566 /* The rest of this function executes only for OST_READs */
1568 /* if unwrap_bulk failed, return -EAGAIN to retry */
1569 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1571 GOTO(out, rc = -EAGAIN);
1573 if (rc > aa->aa_requested_nob) {
1574 CERROR("Unexpected rc %d (%d requested)\n", rc,
1575 aa->aa_requested_nob);
1579 if (rc != req->rq_bulk->bd_nob_transferred) {
1580 CERROR ("Unexpected rc %d (%d transferred)\n",
1581 rc, req->rq_bulk->bd_nob_transferred);
1585 if (rc < aa->aa_requested_nob)
1586 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1588 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1589 static int cksum_counter;
1590 __u32 server_cksum = body->oa.o_cksum;
1593 cksum_type_t cksum_type;
1595 cksum_type = cksum_type_unpack(body->oa.o_valid &OBD_MD_FLFLAGS?
1596 body->oa.o_flags : 0);
1597 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1598 aa->aa_ppga, OST_READ,
1601 if (peer->nid == req->rq_bulk->bd_sender) {
1605 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1608 if (server_cksum == ~0 && rc > 0) {
1609 CERROR("Protocol error: server %s set the 'checksum' "
1610 "bit, but didn't send a checksum. Not fatal, "
1611 "but please notify on http://bugs.whamcloud.com/\n",
1612 libcfs_nid2str(peer->nid));
1613 } else if (server_cksum != client_cksum) {
1614 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1615 "%s%s%s inode "DFID" object "DOSTID
1616 " extent ["LPU64"-"LPU64"]\n",
1617 req->rq_import->imp_obd->obd_name,
1618 libcfs_nid2str(peer->nid),
1620 body->oa.o_valid & OBD_MD_FLFID ?
1621 body->oa.o_parent_seq : (__u64)0,
1622 body->oa.o_valid & OBD_MD_FLFID ?
1623 body->oa.o_parent_oid : 0,
1624 body->oa.o_valid & OBD_MD_FLFID ?
1625 body->oa.o_parent_ver : 0,
1626 POSTID(&body->oa.o_oi),
1627 aa->aa_ppga[0]->off,
1628 aa->aa_ppga[aa->aa_page_count-1]->off +
1629 aa->aa_ppga[aa->aa_page_count-1]->count -
1631 CERROR("client %x, server %x, cksum_type %x\n",
1632 client_cksum, server_cksum, cksum_type);
1634 aa->aa_oa->o_cksum = client_cksum;
1638 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1641 } else if (unlikely(client_cksum)) {
1642 static int cksum_missed;
1645 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1646 CERROR("Checksum %u requested from %s but not sent\n",
1647 cksum_missed, libcfs_nid2str(peer->nid));
1653 lustre_get_wire_obdo(&req->rq_import->imp_connect_data,
1654 aa->aa_oa, &body->oa);
1659 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1660 struct lov_stripe_md *lsm,
1661 obd_count page_count, struct brw_page **pga,
1662 struct obd_capa *ocapa)
1664 struct ptlrpc_request *req;
1666 wait_queue_head_t waitq;
1667 int generation, resends = 0;
1668 struct l_wait_info lwi;
1672 init_waitqueue_head(&waitq);
1673 generation = exp->exp_obd->u.cli.cl_import->imp_generation;
1676 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1677 page_count, pga, &req, ocapa, 0, resends);
1682 req->rq_generation_set = 1;
1683 req->rq_import_generation = generation;
1684 req->rq_sent = cfs_time_current_sec() + resends;
1687 rc = ptlrpc_queue_wait(req);
1689 if (rc == -ETIMEDOUT && req->rq_resend) {
1690 DEBUG_REQ(D_HA, req, "BULK TIMEOUT");
1691 ptlrpc_req_finished(req);
1695 rc = osc_brw_fini_request(req, rc);
1697 ptlrpc_req_finished(req);
1698 /* When server return -EINPROGRESS, client should always retry
1699 * regardless of the number of times the bulk was resent already.*/
1700 if (osc_recoverable_error(rc)) {
1702 if (rc != -EINPROGRESS &&
1703 !client_should_resend(resends, &exp->exp_obd->u.cli)) {
1704 CERROR("%s: too many resend retries for object: "
1705 ""DOSTID", rc = %d.\n", exp->exp_obd->obd_name,
1706 POSTID(&oa->o_oi), rc);
1710 exp->exp_obd->u.cli.cl_import->imp_generation) {
1711 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1712 ""DOSTID", rc = %d.\n", exp->exp_obd->obd_name,
1713 POSTID(&oa->o_oi), rc);
1717 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL,
1719 l_wait_event(waitq, 0, &lwi);
1724 if (rc == -EAGAIN || rc == -EINPROGRESS)
1729 static int osc_brw_redo_request(struct ptlrpc_request *request,
1730 struct osc_brw_async_args *aa, int rc)
1732 struct ptlrpc_request *new_req;
1733 struct osc_brw_async_args *new_aa;
1734 struct osc_async_page *oap;
1737 DEBUG_REQ(rc == -EINPROGRESS ? D_RPCTRACE : D_ERROR, request,
1738 "redo for recoverable error %d", rc);
1740 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1741 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1742 aa->aa_cli, aa->aa_oa,
1743 NULL /* lsm unused by osc currently */,
1744 aa->aa_page_count, aa->aa_ppga,
1745 &new_req, aa->aa_ocapa, 0, 1);
1749 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1750 if (oap->oap_request != NULL) {
1751 LASSERTF(request == oap->oap_request,
1752 "request %p != oap_request %p\n",
1753 request, oap->oap_request);
1754 if (oap->oap_interrupted) {
1755 ptlrpc_req_finished(new_req);
1760 /* New request takes over pga and oaps from old request.
1761 * Note that copying a list_head doesn't work, need to move it... */
1763 new_req->rq_interpret_reply = request->rq_interpret_reply;
1764 new_req->rq_async_args = request->rq_async_args;
1765 new_req->rq_commit_cb = request->rq_commit_cb;
1766 /* cap resend delay to the current request timeout, this is similar to
1767 * what ptlrpc does (see after_reply()) */
1768 if (aa->aa_resends > new_req->rq_timeout)
1769 new_req->rq_sent = cfs_time_current_sec() + new_req->rq_timeout;
1771 new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1772 new_req->rq_generation_set = 1;
1773 new_req->rq_import_generation = request->rq_import_generation;
1775 new_aa = ptlrpc_req_async_args(new_req);
1777 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1778 cfs_list_splice_init(&aa->aa_oaps, &new_aa->aa_oaps);
1779 CFS_INIT_LIST_HEAD(&new_aa->aa_exts);
1780 cfs_list_splice_init(&aa->aa_exts, &new_aa->aa_exts);
1781 new_aa->aa_resends = aa->aa_resends;
1783 cfs_list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1784 if (oap->oap_request) {
1785 ptlrpc_req_finished(oap->oap_request);
1786 oap->oap_request = ptlrpc_request_addref(new_req);
1790 new_aa->aa_ocapa = aa->aa_ocapa;
1791 aa->aa_ocapa = NULL;
1793 /* XXX: This code will run into problem if we're going to support
1794 * to add a series of BRW RPCs into a self-defined ptlrpc_request_set
1795 * and wait for all of them to be finished. We should inherit request
1796 * set from old request. */
1797 ptlrpcd_add_req(new_req, PDL_POLICY_SAME, -1);
1799 DEBUG_REQ(D_INFO, new_req, "new request");
1804 * ugh, we want disk allocation on the target to happen in offset order. we'll
1805 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1806 * fine for our small page arrays and doesn't require allocation. its an
1807 * insertion sort that swaps elements that are strides apart, shrinking the
1808 * stride down until its '1' and the array is sorted.
1810 static void sort_brw_pages(struct brw_page **array, int num)
1813 struct brw_page *tmp;
1817 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1822 for (i = stride ; i < num ; i++) {
1825 while (j >= stride && array[j - stride]->off > tmp->off) {
1826 array[j] = array[j - stride];
1831 } while (stride > 1);
1834 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1840 LASSERT (pages > 0);
1841 offset = pg[i]->off & ~CFS_PAGE_MASK;
1845 if (pages == 0) /* that's all */
1848 if (offset + pg[i]->count < PAGE_CACHE_SIZE)
1849 return count; /* doesn't end on page boundary */
1852 offset = pg[i]->off & ~CFS_PAGE_MASK;
1853 if (offset != 0) /* doesn't start on page boundary */
1860 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1862 struct brw_page **ppga;
1865 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1869 for (i = 0; i < count; i++)
1874 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1876 LASSERT(ppga != NULL);
1877 OBD_FREE(ppga, sizeof(*ppga) * count);
1880 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1881 obd_count page_count, struct brw_page *pga,
1882 struct obd_trans_info *oti)
1884 struct obdo *saved_oa = NULL;
1885 struct brw_page **ppga, **orig;
1886 struct obd_import *imp = class_exp2cliimp(exp);
1887 struct client_obd *cli;
1888 int rc, page_count_orig;
1891 LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1892 cli = &imp->imp_obd->u.cli;
1894 if (cmd & OBD_BRW_CHECK) {
1895 /* The caller just wants to know if there's a chance that this
1896 * I/O can succeed */
1898 if (imp->imp_invalid)
1903 /* test_brw with a failed create can trip this, maybe others. */
1904 LASSERT(cli->cl_max_pages_per_rpc);
1908 orig = ppga = osc_build_ppga(pga, page_count);
1911 page_count_orig = page_count;
1913 sort_brw_pages(ppga, page_count);
1914 while (page_count) {
1915 obd_count pages_per_brw;
1917 if (page_count > cli->cl_max_pages_per_rpc)
1918 pages_per_brw = cli->cl_max_pages_per_rpc;
1920 pages_per_brw = page_count;
1922 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1924 if (saved_oa != NULL) {
1925 /* restore previously saved oa */
1926 *oinfo->oi_oa = *saved_oa;
1927 } else if (page_count > pages_per_brw) {
1928 /* save a copy of oa (brw will clobber it) */
1929 OBDO_ALLOC(saved_oa);
1930 if (saved_oa == NULL)
1931 GOTO(out, rc = -ENOMEM);
1932 *saved_oa = *oinfo->oi_oa;
1935 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1936 pages_per_brw, ppga, oinfo->oi_capa);
1941 page_count -= pages_per_brw;
1942 ppga += pages_per_brw;
1946 osc_release_ppga(orig, page_count_orig);
1948 if (saved_oa != NULL)
1949 OBDO_FREE(saved_oa);
1954 static int brw_interpret(const struct lu_env *env,
1955 struct ptlrpc_request *req, void *data, int rc)
1957 struct osc_brw_async_args *aa = data;
1958 struct osc_extent *ext;
1959 struct osc_extent *tmp;
1960 struct client_obd *cli = aa->aa_cli;
1963 rc = osc_brw_fini_request(req, rc);
1964 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
1965 /* When server return -EINPROGRESS, client should always retry
1966 * regardless of the number of times the bulk was resent already. */
1967 if (osc_recoverable_error(rc)) {
1968 if (req->rq_import_generation !=
1969 req->rq_import->imp_generation) {
1970 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1971 ""DOSTID", rc = %d.\n",
1972 req->rq_import->imp_obd->obd_name,
1973 POSTID(&aa->aa_oa->o_oi), rc);
1974 } else if (rc == -EINPROGRESS ||
1975 client_should_resend(aa->aa_resends, aa->aa_cli)) {
1976 rc = osc_brw_redo_request(req, aa, rc);
1978 CERROR("%s: too many resent retries for object: "
1979 ""LPU64":"LPU64", rc = %d.\n",
1980 req->rq_import->imp_obd->obd_name,
1981 POSTID(&aa->aa_oa->o_oi), rc);
1986 else if (rc == -EAGAIN || rc == -EINPROGRESS)
1991 capa_put(aa->aa_ocapa);
1992 aa->aa_ocapa = NULL;
1996 struct obdo *oa = aa->aa_oa;
1997 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
1998 unsigned long valid = 0;
1999 struct cl_object *obj;
2000 struct osc_async_page *last;
2002 last = brw_page2oap(aa->aa_ppga[aa->aa_page_count - 1]);
2003 obj = osc2cl(last->oap_obj);
2005 cl_object_attr_lock(obj);
2006 if (oa->o_valid & OBD_MD_FLBLOCKS) {
2007 attr->cat_blocks = oa->o_blocks;
2008 valid |= CAT_BLOCKS;
2010 if (oa->o_valid & OBD_MD_FLMTIME) {
2011 attr->cat_mtime = oa->o_mtime;
2014 if (oa->o_valid & OBD_MD_FLATIME) {
2015 attr->cat_atime = oa->o_atime;
2018 if (oa->o_valid & OBD_MD_FLCTIME) {
2019 attr->cat_ctime = oa->o_ctime;
2023 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
2024 struct lov_oinfo *loi = cl2osc(obj)->oo_oinfo;
2025 loff_t last_off = last->oap_count + last->oap_obj_off;
2027 /* Change file size if this is an out of quota or
2028 * direct IO write and it extends the file size */
2029 if (loi->loi_lvb.lvb_size < last_off) {
2030 attr->cat_size = last_off;
2033 /* Extend KMS if it's not a lockless write */
2034 if (loi->loi_kms < last_off &&
2035 oap2osc_page(last)->ops_srvlock == 0) {
2036 attr->cat_kms = last_off;
2042 cl_object_attr_set(env, obj, attr, valid);
2043 cl_object_attr_unlock(obj);
2045 OBDO_FREE(aa->aa_oa);
2047 cfs_list_for_each_entry_safe(ext, tmp, &aa->aa_exts, oe_link) {
2048 cfs_list_del_init(&ext->oe_link);
2049 osc_extent_finish(env, ext, 1, rc);
2051 LASSERT(cfs_list_empty(&aa->aa_exts));
2052 LASSERT(cfs_list_empty(&aa->aa_oaps));
2054 cl_req_completion(env, aa->aa_clerq, rc < 0 ? rc :
2055 req->rq_bulk->bd_nob_transferred);
2056 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2057 ptlrpc_lprocfs_brw(req, req->rq_bulk->bd_nob_transferred);
2059 client_obd_list_lock(&cli->cl_loi_list_lock);
2060 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2061 * is called so we know whether to go to sync BRWs or wait for more
2062 * RPCs to complete */
2063 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2064 cli->cl_w_in_flight--;
2066 cli->cl_r_in_flight--;
2067 osc_wake_cache_waiters(cli);
2068 client_obd_list_unlock(&cli->cl_loi_list_lock);
2070 osc_io_unplug(env, cli, NULL, PDL_POLICY_SAME);
2074 static void brw_commit(struct ptlrpc_request *req)
2076 spin_lock(&req->rq_lock);
2077 /* If osc_inc_unstable_pages (via osc_extent_finish) races with
2078 * this called via the rq_commit_cb, I need to ensure
2079 * osc_dec_unstable_pages is still called. Otherwise unstable
2080 * pages may be leaked. */
2081 if (req->rq_unstable) {
2082 spin_unlock(&req->rq_lock);
2083 osc_dec_unstable_pages(req);
2084 spin_lock(&req->rq_lock);
2086 req->rq_committed = 1;
2088 spin_unlock(&req->rq_lock);
2092 * Build an RPC by the list of extent @ext_list. The caller must ensure
2093 * that the total pages in this list are NOT over max pages per RPC.
2094 * Extents in the list must be in OES_RPC state.
2096 int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
2097 cfs_list_t *ext_list, int cmd, pdl_policy_t pol)
2099 struct ptlrpc_request *req = NULL;
2100 struct osc_extent *ext;
2101 struct brw_page **pga = NULL;
2102 struct osc_brw_async_args *aa = NULL;
2103 struct obdo *oa = NULL;
2104 struct osc_async_page *oap;
2105 struct osc_async_page *tmp;
2106 struct cl_req *clerq = NULL;
2107 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE :
2109 struct ldlm_lock *lock = NULL;
2110 struct cl_req_attr *crattr = NULL;
2111 obd_off starting_offset = OBD_OBJECT_EOF;
2112 obd_off ending_offset = 0;
2118 CFS_LIST_HEAD(rpc_list);
2121 LASSERT(!cfs_list_empty(ext_list));
2123 /* add pages into rpc_list to build BRW rpc */
2124 cfs_list_for_each_entry(ext, ext_list, oe_link) {
2125 LASSERT(ext->oe_state == OES_RPC);
2126 mem_tight |= ext->oe_memalloc;
2127 cfs_list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
2129 cfs_list_add_tail(&oap->oap_rpc_item, &rpc_list);
2130 if (starting_offset > oap->oap_obj_off)
2131 starting_offset = oap->oap_obj_off;
2133 LASSERT(oap->oap_page_off == 0);
2134 if (ending_offset < oap->oap_obj_off + oap->oap_count)
2135 ending_offset = oap->oap_obj_off +
2138 LASSERT(oap->oap_page_off + oap->oap_count ==
2144 mpflag = cfs_memory_pressure_get_and_set();
2146 OBD_ALLOC(crattr, sizeof(*crattr));
2148 GOTO(out, rc = -ENOMEM);
2150 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2152 GOTO(out, rc = -ENOMEM);
2156 GOTO(out, rc = -ENOMEM);
2159 cfs_list_for_each_entry(oap, &rpc_list, oap_rpc_item) {
2160 struct cl_page *page = oap2cl_page(oap);
2161 if (clerq == NULL) {
2162 clerq = cl_req_alloc(env, page, crt,
2163 1 /* only 1-object rpcs for now */);
2165 GOTO(out, rc = PTR_ERR(clerq));
2166 lock = oap->oap_ldlm_lock;
2169 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
2170 pga[i] = &oap->oap_brw_page;
2171 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2172 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2173 pga[i]->pg, page_index(oap->oap_page), oap,
2176 cl_req_page_add(env, clerq, page);
2179 /* always get the data for the obdo for the rpc */
2180 LASSERT(clerq != NULL);
2181 crattr->cra_oa = oa;
2182 cl_req_attr_set(env, clerq, crattr, ~0ULL);
2184 oa->o_handle = lock->l_remote_handle;
2185 oa->o_valid |= OBD_MD_FLHANDLE;
2188 rc = cl_req_prep(env, clerq);
2190 CERROR("cl_req_prep failed: %d\n", rc);
2194 sort_brw_pages(pga, page_count);
2195 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2196 pga, &req, crattr->cra_capa, 1, 0);
2198 CERROR("prep_req failed: %d\n", rc);
2202 req->rq_commit_cb = brw_commit;
2203 req->rq_interpret_reply = brw_interpret;
2206 req->rq_memalloc = 1;
2208 /* Need to update the timestamps after the request is built in case
2209 * we race with setattr (locally or in queue at OST). If OST gets
2210 * later setattr before earlier BRW (as determined by the request xid),
2211 * the OST will not use BRW timestamps. Sadly, there is no obvious
2212 * way to do this in a single call. bug 10150 */
2213 cl_req_attr_set(env, clerq, crattr,
2214 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2216 lustre_msg_set_jobid(req->rq_reqmsg, crattr->cra_jobid);
2218 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2219 aa = ptlrpc_req_async_args(req);
2220 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2221 cfs_list_splice_init(&rpc_list, &aa->aa_oaps);
2222 CFS_INIT_LIST_HEAD(&aa->aa_exts);
2223 cfs_list_splice_init(ext_list, &aa->aa_exts);
2224 aa->aa_clerq = clerq;
2226 /* queued sync pages can be torn down while the pages
2227 * were between the pending list and the rpc */
2229 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2230 /* only one oap gets a request reference */
2233 if (oap->oap_interrupted && !req->rq_intr) {
2234 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2236 ptlrpc_mark_interrupted(req);
2240 tmp->oap_request = ptlrpc_request_addref(req);
2242 client_obd_list_lock(&cli->cl_loi_list_lock);
2243 starting_offset >>= PAGE_CACHE_SHIFT;
2244 if (cmd == OBD_BRW_READ) {
2245 cli->cl_r_in_flight++;
2246 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2247 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2248 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2249 starting_offset + 1);
2251 cli->cl_w_in_flight++;
2252 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2253 lprocfs_oh_tally(&cli->cl_write_rpc_hist, cli->cl_w_in_flight);
2254 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2255 starting_offset + 1);
2257 client_obd_list_unlock(&cli->cl_loi_list_lock);
2259 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2260 page_count, aa, cli->cl_r_in_flight,
2261 cli->cl_w_in_flight);
2263 /* XXX: Maybe the caller can check the RPC bulk descriptor to
2264 * see which CPU/NUMA node the majority of pages were allocated
2265 * on, and try to assign the async RPC to the CPU core
2266 * (PDL_POLICY_PREFERRED) to reduce cross-CPU memory traffic.
2268 * But on the other hand, we expect that multiple ptlrpcd
2269 * threads and the initial write sponsor can run in parallel,
2270 * especially when data checksum is enabled, which is CPU-bound
2271 * operation and single ptlrpcd thread cannot process in time.
2272 * So more ptlrpcd threads sharing BRW load
2273 * (with PDL_POLICY_ROUND) seems better.
2275 ptlrpcd_add_req(req, pol, -1);
2281 cfs_memory_pressure_restore(mpflag);
2283 if (crattr != NULL) {
2284 capa_put(crattr->cra_capa);
2285 OBD_FREE(crattr, sizeof(*crattr));
2289 LASSERT(req == NULL);
2294 OBD_FREE(pga, sizeof(*pga) * page_count);
2295 /* this should happen rarely and is pretty bad, it makes the
2296 * pending list not follow the dirty order */
2297 while (!cfs_list_empty(ext_list)) {
2298 ext = cfs_list_entry(ext_list->next, struct osc_extent,
2300 cfs_list_del_init(&ext->oe_link);
2301 osc_extent_finish(env, ext, 0, rc);
2303 if (clerq && !IS_ERR(clerq))
2304 cl_req_completion(env, clerq, rc);
2309 static int osc_set_lock_data_with_check(struct ldlm_lock *lock,
2310 struct ldlm_enqueue_info *einfo)
2312 void *data = einfo->ei_cbdata;
2315 LASSERT(lock != NULL);
2316 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2317 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2318 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2319 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2321 lock_res_and_lock(lock);
2322 spin_lock(&osc_ast_guard);
2324 if (lock->l_ast_data == NULL)
2325 lock->l_ast_data = data;
2326 if (lock->l_ast_data == data)
2329 spin_unlock(&osc_ast_guard);
2330 unlock_res_and_lock(lock);
2335 static int osc_set_data_with_check(struct lustre_handle *lockh,
2336 struct ldlm_enqueue_info *einfo)
2338 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2342 set = osc_set_lock_data_with_check(lock, einfo);
2343 LDLM_LOCK_PUT(lock);
2345 CERROR("lockh %p, data %p - client evicted?\n",
2346 lockh, einfo->ei_cbdata);
2350 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2351 ldlm_iterator_t replace, void *data)
2353 struct ldlm_res_id res_id;
2354 struct obd_device *obd = class_exp2obd(exp);
2356 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2357 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2361 /* find any ldlm lock of the inode in osc
2365 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2366 ldlm_iterator_t replace, void *data)
2368 struct ldlm_res_id res_id;
2369 struct obd_device *obd = class_exp2obd(exp);
2372 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2373 rc = ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2374 if (rc == LDLM_ITER_STOP)
2376 if (rc == LDLM_ITER_CONTINUE)
2381 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
2382 obd_enqueue_update_f upcall, void *cookie,
2383 __u64 *flags, int agl, int rc)
2385 int intent = *flags & LDLM_FL_HAS_INTENT;
2389 /* The request was created before ldlm_cli_enqueue call. */
2390 if (rc == ELDLM_LOCK_ABORTED) {
2391 struct ldlm_reply *rep;
2392 rep = req_capsule_server_get(&req->rq_pill,
2395 LASSERT(rep != NULL);
2396 rep->lock_policy_res1 =
2397 ptlrpc_status_ntoh(rep->lock_policy_res1);
2398 if (rep->lock_policy_res1)
2399 rc = rep->lock_policy_res1;
2403 if ((intent != 0 && rc == ELDLM_LOCK_ABORTED && agl == 0) ||
2405 *flags |= LDLM_FL_LVB_READY;
2406 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
2407 lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
2410 /* Call the update callback. */
2411 rc = (*upcall)(cookie, rc);
2415 static int osc_enqueue_interpret(const struct lu_env *env,
2416 struct ptlrpc_request *req,
2417 struct osc_enqueue_args *aa, int rc)
2419 struct ldlm_lock *lock;
2420 struct lustre_handle handle;
2422 struct ost_lvb *lvb;
2424 __u64 *flags = aa->oa_flags;
2426 /* Make a local copy of a lock handle and a mode, because aa->oa_*
2427 * might be freed anytime after lock upcall has been called. */
2428 lustre_handle_copy(&handle, aa->oa_lockh);
2429 mode = aa->oa_ei->ei_mode;
2431 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2433 lock = ldlm_handle2lock(&handle);
2435 /* Take an additional reference so that a blocking AST that
2436 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2437 * to arrive after an upcall has been executed by
2438 * osc_enqueue_fini(). */
2439 ldlm_lock_addref(&handle, mode);
2441 /* Let cl_lock_state_wait fail with -ERESTARTSYS to unuse sublocks. */
2442 OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_ENQUEUE_HANG, 2);
2444 /* Let CP AST to grant the lock first. */
2445 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
2447 if (aa->oa_agl && rc == ELDLM_LOCK_ABORTED) {
2452 lvb_len = sizeof(*aa->oa_lvb);
2455 /* Complete obtaining the lock procedure. */
2456 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
2457 mode, flags, lvb, lvb_len, &handle, rc);
2458 /* Complete osc stuff. */
2459 rc = osc_enqueue_fini(req, aa->oa_lvb, aa->oa_upcall, aa->oa_cookie,
2460 flags, aa->oa_agl, rc);
2462 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
2464 /* Release the lock for async request. */
2465 if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
2467 * Releases a reference taken by ldlm_cli_enqueue(), if it is
2468 * not already released by
2469 * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
2471 ldlm_lock_decref(&handle, mode);
2473 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
2474 aa->oa_lockh, req, aa);
2475 ldlm_lock_decref(&handle, mode);
2476 LDLM_LOCK_PUT(lock);
2480 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
2481 struct lov_oinfo *loi, __u64 flags,
2482 struct ost_lvb *lvb, __u32 mode, int rc)
2484 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
2486 if (rc == ELDLM_OK) {
2489 LASSERT(lock != NULL);
2490 loi->loi_lvb = *lvb;
2491 tmp = loi->loi_lvb.lvb_size;
2492 /* Extend KMS up to the end of this lock and no further
2493 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
2494 if (tmp > lock->l_policy_data.l_extent.end)
2495 tmp = lock->l_policy_data.l_extent.end + 1;
2496 if (tmp >= loi->loi_kms) {
2497 LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
2498 ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
2499 loi_kms_set(loi, tmp);
2501 LDLM_DEBUG(lock, "lock acquired, setting rss="
2502 LPU64"; leaving kms="LPU64", end="LPU64,
2503 loi->loi_lvb.lvb_size, loi->loi_kms,
2504 lock->l_policy_data.l_extent.end);
2506 ldlm_lock_allow_match(lock);
2507 } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
2508 LASSERT(lock != NULL);
2509 loi->loi_lvb = *lvb;
2510 ldlm_lock_allow_match(lock);
2511 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
2512 " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
2518 ldlm_lock_fail_match(lock);
2520 LDLM_LOCK_PUT(lock);
2523 EXPORT_SYMBOL(osc_update_enqueue);
2525 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
2527 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
2528 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
2529 * other synchronous requests, however keeping some locks and trying to obtain
2530 * others may take a considerable amount of time in a case of ost failure; and
2531 * when other sync requests do not get released lock from a client, the client
2532 * is excluded from the cluster -- such scenarious make the life difficult, so
2533 * release locks just after they are obtained. */
2534 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2535 __u64 *flags, ldlm_policy_data_t *policy,
2536 struct ost_lvb *lvb, int kms_valid,
2537 obd_enqueue_update_f upcall, void *cookie,
2538 struct ldlm_enqueue_info *einfo,
2539 struct lustre_handle *lockh,
2540 struct ptlrpc_request_set *rqset, int async, int agl)
2542 struct obd_device *obd = exp->exp_obd;
2543 struct ptlrpc_request *req = NULL;
2544 int intent = *flags & LDLM_FL_HAS_INTENT;
2545 __u64 match_lvb = (agl != 0 ? 0 : LDLM_FL_LVB_READY);
2550 /* Filesystem lock extents are extended to page boundaries so that
2551 * dealing with the page cache is a little smoother. */
2552 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
2553 policy->l_extent.end |= ~CFS_PAGE_MASK;
2556 * kms is not valid when either object is completely fresh (so that no
2557 * locks are cached), or object was evicted. In the latter case cached
2558 * lock cannot be used, because it would prime inode state with
2559 * potentially stale LVB.
2564 /* Next, search for already existing extent locks that will cover us */
2565 /* If we're trying to read, we also search for an existing PW lock. The
2566 * VFS and page cache already protect us locally, so lots of readers/
2567 * writers can share a single PW lock.
2569 * There are problems with conversion deadlocks, so instead of
2570 * converting a read lock to a write lock, we'll just enqueue a new
2573 * At some point we should cancel the read lock instead of making them
2574 * send us a blocking callback, but there are problems with canceling
2575 * locks out from other users right now, too. */
2576 mode = einfo->ei_mode;
2577 if (einfo->ei_mode == LCK_PR)
2579 mode = ldlm_lock_match(obd->obd_namespace, *flags | match_lvb, res_id,
2580 einfo->ei_type, policy, mode, lockh, 0);
2582 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
2584 if ((agl != 0) && !ldlm_is_lvb_ready(matched)) {
2585 /* For AGL, if enqueue RPC is sent but the lock is not
2586 * granted, then skip to process this strpe.
2587 * Return -ECANCELED to tell the caller. */
2588 ldlm_lock_decref(lockh, mode);
2589 LDLM_LOCK_PUT(matched);
2591 } else if (osc_set_lock_data_with_check(matched, einfo)) {
2592 *flags |= LDLM_FL_LVB_READY;
2593 /* addref the lock only if not async requests and PW
2594 * lock is matched whereas we asked for PR. */
2595 if (!rqset && einfo->ei_mode != mode)
2596 ldlm_lock_addref(lockh, LCK_PR);
2598 /* I would like to be able to ASSERT here that
2599 * rss <= kms, but I can't, for reasons which
2600 * are explained in lov_enqueue() */
2603 /* We already have a lock, and it's referenced.
2605 * At this point, the cl_lock::cll_state is CLS_QUEUING,
2606 * AGL upcall may change it to CLS_HELD directly. */
2607 (*upcall)(cookie, ELDLM_OK);
2609 if (einfo->ei_mode != mode)
2610 ldlm_lock_decref(lockh, LCK_PW);
2612 /* For async requests, decref the lock. */
2613 ldlm_lock_decref(lockh, einfo->ei_mode);
2614 LDLM_LOCK_PUT(matched);
2617 ldlm_lock_decref(lockh, mode);
2618 LDLM_LOCK_PUT(matched);
2624 CFS_LIST_HEAD(cancels);
2625 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2626 &RQF_LDLM_ENQUEUE_LVB);
2630 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
2632 ptlrpc_request_free(req);
2636 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
2638 ptlrpc_request_set_replen(req);
2641 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
2642 *flags &= ~LDLM_FL_BLOCK_GRANTED;
2644 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
2645 sizeof(*lvb), LVB_T_OST, lockh, async);
2648 struct osc_enqueue_args *aa;
2649 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
2650 aa = ptlrpc_req_async_args(req);
2653 aa->oa_flags = flags;
2654 aa->oa_upcall = upcall;
2655 aa->oa_cookie = cookie;
2657 aa->oa_lockh = lockh;
2660 req->rq_interpret_reply =
2661 (ptlrpc_interpterer_t)osc_enqueue_interpret;
2662 if (rqset == PTLRPCD_SET)
2663 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
2665 ptlrpc_set_add_req(rqset, req);
2666 } else if (intent) {
2667 ptlrpc_req_finished(req);
2672 rc = osc_enqueue_fini(req, lvb, upcall, cookie, flags, agl, rc);
2674 ptlrpc_req_finished(req);
2679 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
2680 struct ldlm_enqueue_info *einfo,
2681 struct ptlrpc_request_set *rqset)
2683 struct ldlm_res_id res_id;
2687 ostid_build_res_name(&oinfo->oi_md->lsm_oi, &res_id);
2688 rc = osc_enqueue_base(exp, &res_id, &oinfo->oi_flags, &oinfo->oi_policy,
2689 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
2690 oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid,
2691 oinfo->oi_cb_up, oinfo, einfo, oinfo->oi_lockh,
2692 rqset, rqset != NULL, 0);
2696 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2697 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
2698 __u64 *flags, void *data, struct lustre_handle *lockh,
2701 struct obd_device *obd = exp->exp_obd;
2702 __u64 lflags = *flags;
2706 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
2709 /* Filesystem lock extents are extended to page boundaries so that
2710 * dealing with the page cache is a little smoother */
2711 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
2712 policy->l_extent.end |= ~CFS_PAGE_MASK;
2714 /* Next, search for already existing extent locks that will cover us */
2715 /* If we're trying to read, we also search for an existing PW lock. The
2716 * VFS and page cache already protect us locally, so lots of readers/
2717 * writers can share a single PW lock. */
2721 rc = ldlm_lock_match(obd->obd_namespace, lflags,
2722 res_id, type, policy, rc, lockh, unref);
2725 if (!osc_set_data_with_check(lockh, data)) {
2726 if (!(lflags & LDLM_FL_TEST_LOCK))
2727 ldlm_lock_decref(lockh, rc);
2731 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
2732 ldlm_lock_addref(lockh, LCK_PR);
2733 ldlm_lock_decref(lockh, LCK_PW);
2740 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
2744 if (unlikely(mode == LCK_GROUP))
2745 ldlm_lock_decref_and_cancel(lockh, mode);
2747 ldlm_lock_decref(lockh, mode);
2752 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
2753 __u32 mode, struct lustre_handle *lockh)
2756 RETURN(osc_cancel_base(lockh, mode));
2759 static int osc_cancel_unused(struct obd_export *exp,
2760 struct lov_stripe_md *lsm,
2761 ldlm_cancel_flags_t flags,
2764 struct obd_device *obd = class_exp2obd(exp);
2765 struct ldlm_res_id res_id, *resp = NULL;
2768 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2772 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
2775 static int osc_statfs_interpret(const struct lu_env *env,
2776 struct ptlrpc_request *req,
2777 struct osc_async_args *aa, int rc)
2779 struct obd_statfs *msfs;
2783 /* The request has in fact never been sent
2784 * due to issues at a higher level (LOV).
2785 * Exit immediately since the caller is
2786 * aware of the problem and takes care
2787 * of the clean up */
2790 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
2791 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
2797 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2799 GOTO(out, rc = -EPROTO);
2802 *aa->aa_oi->oi_osfs = *msfs;
2804 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
2808 static int osc_statfs_async(struct obd_export *exp,
2809 struct obd_info *oinfo, __u64 max_age,
2810 struct ptlrpc_request_set *rqset)
2812 struct obd_device *obd = class_exp2obd(exp);
2813 struct ptlrpc_request *req;
2814 struct osc_async_args *aa;
2818 /* We could possibly pass max_age in the request (as an absolute
2819 * timestamp or a "seconds.usec ago") so the target can avoid doing
2820 * extra calls into the filesystem if that isn't necessary (e.g.
2821 * during mount that would help a bit). Having relative timestamps
2822 * is not so great if request processing is slow, while absolute
2823 * timestamps are not ideal because they need time synchronization. */
2824 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
2828 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2830 ptlrpc_request_free(req);
2833 ptlrpc_request_set_replen(req);
2834 req->rq_request_portal = OST_CREATE_PORTAL;
2835 ptlrpc_at_set_req_timeout(req);
2837 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
2838 /* procfs requests not want stat in wait for avoid deadlock */
2839 req->rq_no_resend = 1;
2840 req->rq_no_delay = 1;
2843 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
2844 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
2845 aa = ptlrpc_req_async_args(req);
2848 ptlrpc_set_add_req(rqset, req);
2852 static int osc_statfs(const struct lu_env *env, struct obd_export *exp,
2853 struct obd_statfs *osfs, __u64 max_age, __u32 flags)
2855 struct obd_device *obd = class_exp2obd(exp);
2856 struct obd_statfs *msfs;
2857 struct ptlrpc_request *req;
2858 struct obd_import *imp = NULL;
2862 /*Since the request might also come from lprocfs, so we need
2863 *sync this with client_disconnect_export Bug15684*/
2864 down_read(&obd->u.cli.cl_sem);
2865 if (obd->u.cli.cl_import)
2866 imp = class_import_get(obd->u.cli.cl_import);
2867 up_read(&obd->u.cli.cl_sem);
2871 /* We could possibly pass max_age in the request (as an absolute
2872 * timestamp or a "seconds.usec ago") so the target can avoid doing
2873 * extra calls into the filesystem if that isn't necessary (e.g.
2874 * during mount that would help a bit). Having relative timestamps
2875 * is not so great if request processing is slow, while absolute
2876 * timestamps are not ideal because they need time synchronization. */
2877 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
2879 class_import_put(imp);
2884 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2886 ptlrpc_request_free(req);
2889 ptlrpc_request_set_replen(req);
2890 req->rq_request_portal = OST_CREATE_PORTAL;
2891 ptlrpc_at_set_req_timeout(req);
2893 if (flags & OBD_STATFS_NODELAY) {
2894 /* procfs requests not want stat in wait for avoid deadlock */
2895 req->rq_no_resend = 1;
2896 req->rq_no_delay = 1;
2899 rc = ptlrpc_queue_wait(req);
2903 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2905 GOTO(out, rc = -EPROTO);
2912 ptlrpc_req_finished(req);
2916 /* Retrieve object striping information.
2918 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
2919 * the maximum number of OST indices which will fit in the user buffer.
2920 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
2922 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
2924 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
2925 struct lov_user_md_v3 lum, *lumk;
2926 struct lov_user_ost_data_v1 *lmm_objects;
2927 int rc = 0, lum_size;
2933 /* we only need the header part from user space to get lmm_magic and
2934 * lmm_stripe_count, (the header part is common to v1 and v3) */
2935 lum_size = sizeof(struct lov_user_md_v1);
2936 if (copy_from_user(&lum, lump, lum_size))
2939 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
2940 (lum.lmm_magic != LOV_USER_MAGIC_V3))
2943 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
2944 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
2945 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
2946 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
2948 /* we can use lov_mds_md_size() to compute lum_size
2949 * because lov_user_md_vX and lov_mds_md_vX have the same size */
2950 if (lum.lmm_stripe_count > 0) {
2951 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
2952 OBD_ALLOC(lumk, lum_size);
2956 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
2958 &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
2960 lmm_objects = &(lumk->lmm_objects[0]);
2961 lmm_objects->l_ost_oi = lsm->lsm_oi;
2963 lum_size = lov_mds_md_size(0, lum.lmm_magic);
2967 lumk->lmm_oi = lsm->lsm_oi;
2968 lumk->lmm_stripe_count = 1;
2970 if (copy_to_user(lump, lumk, lum_size))
2974 OBD_FREE(lumk, lum_size);
2980 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
2981 void *karg, void *uarg)
2983 struct obd_device *obd = exp->exp_obd;
2984 struct obd_ioctl_data *data = karg;
2988 if (!try_module_get(THIS_MODULE)) {
2989 CERROR("Can't get module. Is it alive?");
2993 case OBD_IOC_LOV_GET_CONFIG: {
2995 struct lov_desc *desc;
2996 struct obd_uuid uuid;
3000 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
3001 GOTO(out, err = -EINVAL);
3003 data = (struct obd_ioctl_data *)buf;
3005 if (sizeof(*desc) > data->ioc_inllen1) {
3006 obd_ioctl_freedata(buf, len);
3007 GOTO(out, err = -EINVAL);
3010 if (data->ioc_inllen2 < sizeof(uuid)) {
3011 obd_ioctl_freedata(buf, len);
3012 GOTO(out, err = -EINVAL);
3015 desc = (struct lov_desc *)data->ioc_inlbuf1;
3016 desc->ld_tgt_count = 1;
3017 desc->ld_active_tgt_count = 1;
3018 desc->ld_default_stripe_count = 1;
3019 desc->ld_default_stripe_size = 0;
3020 desc->ld_default_stripe_offset = 0;
3021 desc->ld_pattern = 0;
3022 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
3024 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
3026 err = copy_to_user((void *)uarg, buf, len);
3029 obd_ioctl_freedata(buf, len);
3032 case LL_IOC_LOV_SETSTRIPE:
3033 err = obd_alloc_memmd(exp, karg);
3037 case LL_IOC_LOV_GETSTRIPE:
3038 err = osc_getstripe(karg, uarg);
3040 case OBD_IOC_CLIENT_RECOVER:
3041 err = ptlrpc_recover_import(obd->u.cli.cl_import,
3042 data->ioc_inlbuf1, 0);
3046 case IOC_OSC_SET_ACTIVE:
3047 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
3050 case OBD_IOC_POLL_QUOTACHECK:
3051 err = osc_quota_poll_check(exp, (struct if_quotacheck *)karg);
3053 case OBD_IOC_PING_TARGET:
3054 err = ptlrpc_obd_ping(obd);
3057 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
3058 cmd, current_comm());
3059 GOTO(out, err = -ENOTTY);
3062 module_put(THIS_MODULE);
3066 static int osc_get_info(const struct lu_env *env, struct obd_export *exp,
3067 obd_count keylen, void *key, __u32 *vallen, void *val,
3068 struct lov_stripe_md *lsm)
3071 if (!vallen || !val)
3074 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
3075 __u32 *stripe = val;
3076 *vallen = sizeof(*stripe);
3079 } else if (KEY_IS(KEY_LAST_ID)) {
3080 struct ptlrpc_request *req;
3085 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3086 &RQF_OST_GET_INFO_LAST_ID);
3090 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3091 RCL_CLIENT, keylen);
3092 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3094 ptlrpc_request_free(req);
3098 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3099 memcpy(tmp, key, keylen);
3101 req->rq_no_delay = req->rq_no_resend = 1;
3102 ptlrpc_request_set_replen(req);
3103 rc = ptlrpc_queue_wait(req);
3107 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
3109 GOTO(out, rc = -EPROTO);
3111 *((obd_id *)val) = *reply;
3113 ptlrpc_req_finished(req);
3115 } else if (KEY_IS(KEY_FIEMAP)) {
3116 struct ll_fiemap_info_key *fm_key =
3117 (struct ll_fiemap_info_key *)key;
3118 struct ldlm_res_id res_id;
3119 ldlm_policy_data_t policy;
3120 struct lustre_handle lockh;
3121 ldlm_mode_t mode = 0;
3122 struct ptlrpc_request *req;
3123 struct ll_user_fiemap *reply;
3127 if (!(fm_key->fiemap.fm_flags & FIEMAP_FLAG_SYNC))
3130 policy.l_extent.start = fm_key->fiemap.fm_start &
3133 if (OBD_OBJECT_EOF - fm_key->fiemap.fm_length <=
3134 fm_key->fiemap.fm_start + PAGE_CACHE_SIZE - 1)
3135 policy.l_extent.end = OBD_OBJECT_EOF;
3137 policy.l_extent.end = (fm_key->fiemap.fm_start +
3138 fm_key->fiemap.fm_length +
3139 PAGE_CACHE_SIZE - 1) & CFS_PAGE_MASK;
3141 ostid_build_res_name(&fm_key->oa.o_oi, &res_id);
3142 mode = ldlm_lock_match(exp->exp_obd->obd_namespace,
3143 LDLM_FL_BLOCK_GRANTED |
3145 &res_id, LDLM_EXTENT, &policy,
3146 LCK_PR | LCK_PW, &lockh, 0);
3147 if (mode) { /* lock is cached on client */
3148 if (mode != LCK_PR) {
3149 ldlm_lock_addref(&lockh, LCK_PR);
3150 ldlm_lock_decref(&lockh, LCK_PW);
3152 } else { /* no cached lock, needs acquire lock on server side */
3153 fm_key->oa.o_valid |= OBD_MD_FLFLAGS;
3154 fm_key->oa.o_flags |= OBD_FL_SRVLOCK;
3158 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3159 &RQF_OST_GET_INFO_FIEMAP);
3161 GOTO(drop_lock, rc = -ENOMEM);
3163 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
3164 RCL_CLIENT, keylen);
3165 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3166 RCL_CLIENT, *vallen);
3167 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3168 RCL_SERVER, *vallen);
3170 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3172 ptlrpc_request_free(req);
3173 GOTO(drop_lock, rc);
3176 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
3177 memcpy(tmp, key, keylen);
3178 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3179 memcpy(tmp, val, *vallen);
3181 ptlrpc_request_set_replen(req);
3182 rc = ptlrpc_queue_wait(req);
3186 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3188 GOTO(fini_req, rc = -EPROTO);
3190 memcpy(val, reply, *vallen);
3192 ptlrpc_req_finished(req);
3195 ldlm_lock_decref(&lockh, LCK_PR);
3202 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
3203 obd_count keylen, void *key, obd_count vallen,
3204 void *val, struct ptlrpc_request_set *set)
3206 struct ptlrpc_request *req;
3207 struct obd_device *obd = exp->exp_obd;
3208 struct obd_import *imp = class_exp2cliimp(exp);
3213 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
3215 if (KEY_IS(KEY_CHECKSUM)) {
3216 if (vallen != sizeof(int))
3218 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
3222 if (KEY_IS(KEY_SPTLRPC_CONF)) {
3223 sptlrpc_conf_client_adapt(obd);
3227 if (KEY_IS(KEY_FLUSH_CTX)) {
3228 sptlrpc_import_flush_my_ctx(imp);
3232 if (KEY_IS(KEY_CACHE_SET)) {
3233 struct client_obd *cli = &obd->u.cli;
3235 LASSERT(cli->cl_cache == NULL); /* only once */
3236 cli->cl_cache = (struct cl_client_cache *)val;
3237 atomic_inc(&cli->cl_cache->ccc_users);
3238 cli->cl_lru_left = &cli->cl_cache->ccc_lru_left;
3240 /* add this osc into entity list */
3241 LASSERT(cfs_list_empty(&cli->cl_lru_osc));
3242 spin_lock(&cli->cl_cache->ccc_lru_lock);
3243 cfs_list_add(&cli->cl_lru_osc, &cli->cl_cache->ccc_lru);
3244 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3249 if (KEY_IS(KEY_CACHE_LRU_SHRINK)) {
3250 struct client_obd *cli = &obd->u.cli;
3251 int nr = atomic_read(&cli->cl_lru_in_list) >> 1;
3252 int target = *(int *)val;
3254 nr = osc_lru_shrink(env, cli, min(nr, target), true);
3259 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
3262 /* We pass all other commands directly to OST. Since nobody calls osc
3263 methods directly and everybody is supposed to go through LOV, we
3264 assume lov checked invalid values for us.
3265 The only recognised values so far are evict_by_nid and mds_conn.
3266 Even if something bad goes through, we'd get a -EINVAL from OST
3269 req = ptlrpc_request_alloc(imp, KEY_IS(KEY_GRANT_SHRINK) ?
3270 &RQF_OST_SET_GRANT_INFO :
3275 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3276 RCL_CLIENT, keylen);
3277 if (!KEY_IS(KEY_GRANT_SHRINK))
3278 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
3279 RCL_CLIENT, vallen);
3280 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
3282 ptlrpc_request_free(req);
3286 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3287 memcpy(tmp, key, keylen);
3288 tmp = req_capsule_client_get(&req->rq_pill, KEY_IS(KEY_GRANT_SHRINK) ?
3291 memcpy(tmp, val, vallen);
3293 if (KEY_IS(KEY_GRANT_SHRINK)) {
3294 struct osc_grant_args *aa;
3297 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3298 aa = ptlrpc_req_async_args(req);
3301 ptlrpc_req_finished(req);
3304 *oa = ((struct ost_body *)val)->oa;
3306 req->rq_interpret_reply = osc_shrink_grant_interpret;
3309 ptlrpc_request_set_replen(req);
3310 if (!KEY_IS(KEY_GRANT_SHRINK)) {
3311 LASSERT(set != NULL);
3312 ptlrpc_set_add_req(set, req);
3313 ptlrpc_check_set(NULL, set);
3315 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
3321 static int osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
3322 struct obd_device *disk_obd, int *index)
3324 /* this code is not supposed to be used with LOD/OSP
3325 * to be removed soon */
3330 static int osc_llog_finish(struct obd_device *obd, int count)
3332 struct llog_ctxt *ctxt;
3336 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3338 llog_cat_close(NULL, ctxt->loc_handle);
3339 llog_cleanup(NULL, ctxt);
3342 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3344 llog_cleanup(NULL, ctxt);
3348 static int osc_reconnect(const struct lu_env *env,
3349 struct obd_export *exp, struct obd_device *obd,
3350 struct obd_uuid *cluuid,
3351 struct obd_connect_data *data,
3354 struct client_obd *cli = &obd->u.cli;
3356 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
3359 client_obd_list_lock(&cli->cl_loi_list_lock);
3360 data->ocd_grant = (cli->cl_avail_grant + cli->cl_dirty) ?:
3361 2 * cli_brw_size(obd);
3362 lost_grant = cli->cl_lost_grant;
3363 cli->cl_lost_grant = 0;
3364 client_obd_list_unlock(&cli->cl_loi_list_lock);
3366 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
3367 " ocd_grant: %d, lost: %ld.\n", data->ocd_connect_flags,
3368 data->ocd_version, data->ocd_grant, lost_grant);
3374 static int osc_disconnect(struct obd_export *exp)
3376 struct obd_device *obd = class_exp2obd(exp);
3377 struct llog_ctxt *ctxt;
3380 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3382 if (obd->u.cli.cl_conn_count == 1) {
3383 /* Flush any remaining cancel messages out to the
3385 llog_sync(ctxt, exp, 0);
3387 llog_ctxt_put(ctxt);
3389 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
3393 rc = client_disconnect_export(exp);
3395 * Initially we put del_shrink_grant before disconnect_export, but it
3396 * causes the following problem if setup (connect) and cleanup
3397 * (disconnect) are tangled together.
3398 * connect p1 disconnect p2
3399 * ptlrpc_connect_import
3400 * ............... class_manual_cleanup
3403 * ptlrpc_connect_interrupt
3405 * add this client to shrink list
3407 * Bang! pinger trigger the shrink.
3408 * So the osc should be disconnected from the shrink list, after we
3409 * are sure the import has been destroyed. BUG18662
3411 if (obd->u.cli.cl_import == NULL)
3412 osc_del_shrink_grant(&obd->u.cli);
3416 static int osc_import_event(struct obd_device *obd,
3417 struct obd_import *imp,
3418 enum obd_import_event event)
3420 struct client_obd *cli;
3424 LASSERT(imp->imp_obd == obd);
3427 case IMP_EVENT_DISCON: {
3429 client_obd_list_lock(&cli->cl_loi_list_lock);
3430 cli->cl_avail_grant = 0;
3431 cli->cl_lost_grant = 0;
3432 client_obd_list_unlock(&cli->cl_loi_list_lock);
3435 case IMP_EVENT_INACTIVE: {
3436 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
3439 case IMP_EVENT_INVALIDATE: {
3440 struct ldlm_namespace *ns = obd->obd_namespace;
3444 env = cl_env_get(&refcheck);
3448 /* all pages go to failing rpcs due to the invalid
3450 osc_io_unplug(env, cli, NULL, PDL_POLICY_ROUND);
3452 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3453 cl_env_put(env, &refcheck);
3458 case IMP_EVENT_ACTIVE: {
3459 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
3462 case IMP_EVENT_OCD: {
3463 struct obd_connect_data *ocd = &imp->imp_connect_data;
3465 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
3466 osc_init_grant(&obd->u.cli, ocd);
3469 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
3470 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
3472 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
3475 case IMP_EVENT_DEACTIVATE: {
3476 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DEACTIVATE, NULL);
3479 case IMP_EVENT_ACTIVATE: {
3480 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVATE, NULL);
3484 CERROR("Unknown import event %d\n", event);
3491 * Determine whether the lock can be canceled before replaying the lock
3492 * during recovery, see bug16774 for detailed information.
3494 * \retval zero the lock can't be canceled
3495 * \retval other ok to cancel
3497 static int osc_cancel_for_recovery(struct ldlm_lock *lock)
3500 * Cancel all unused extent lock in granted mode LCK_PR or LCK_CR.
3502 * XXX as a future improvement, we can also cancel unused write lock
3503 * if it doesn't have dirty data and active mmaps.
3505 if (lock->l_resource->lr_type == LDLM_EXTENT &&
3506 (lock->l_granted_mode == LCK_PR || lock->l_granted_mode == LCK_CR)&&
3507 osc_ldlm_weigh_ast(lock) == 0)
3513 static int brw_queue_work(const struct lu_env *env, void *data)
3515 struct client_obd *cli = data;
3517 CDEBUG(D_CACHE, "Run writeback work for client obd %p.\n", cli);
3519 osc_io_unplug(env, cli, NULL, PDL_POLICY_SAME);
3523 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
3525 struct client_obd *cli = &obd->u.cli;
3530 rc = ptlrpcd_addref();
3534 rc = client_obd_setup(obd, lcfg);
3536 GOTO(out_ptlrpcd, rc);
3538 handler = ptlrpcd_alloc_work(cli->cl_import, brw_queue_work, cli);
3539 if (IS_ERR(handler))
3540 GOTO(out_client_setup, rc = PTR_ERR(handler));
3541 cli->cl_writeback_work = handler;
3543 handler = ptlrpcd_alloc_work(cli->cl_import, lru_queue_work, cli);
3544 if (IS_ERR(handler))
3545 GOTO(out_ptlrpcd_work, rc = PTR_ERR(handler));
3546 cli->cl_lru_work = handler;
3548 rc = osc_quota_setup(obd);
3550 GOTO(out_ptlrpcd_work, rc);
3552 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
3554 obd->obd_vars = lprocfs_osc_obd_vars;
3556 if (lprocfs_seq_obd_setup(obd) == 0) {
3557 lproc_osc_attach_seqstat(obd);
3558 sptlrpc_lprocfs_cliobd_attach(obd);
3559 ptlrpc_lprocfs_register_obd(obd);
3562 /* We need to allocate a few requests more, because
3563 * brw_interpret tries to create new requests before freeing
3564 * previous ones, Ideally we want to have 2x max_rpcs_in_flight
3565 * reserved, but I'm afraid that might be too much wasted RAM
3566 * in fact, so 2 is just my guess and still should work. */
3567 cli->cl_import->imp_rq_pool =
3568 ptlrpc_init_rq_pool(cli->cl_max_rpcs_in_flight + 2,
3570 ptlrpc_add_rqs_to_pool);
3572 CFS_INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
3573 ns_register_cancel(obd->obd_namespace, osc_cancel_for_recovery);
3577 if (cli->cl_writeback_work != NULL) {
3578 ptlrpcd_destroy_work(cli->cl_writeback_work);
3579 cli->cl_writeback_work = NULL;
3581 if (cli->cl_lru_work != NULL) {
3582 ptlrpcd_destroy_work(cli->cl_lru_work);
3583 cli->cl_lru_work = NULL;
3586 client_obd_cleanup(obd);
3592 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
3598 case OBD_CLEANUP_EARLY: {
3599 struct obd_import *imp;
3600 imp = obd->u.cli.cl_import;
3601 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
3602 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
3603 ptlrpc_deactivate_import(imp);
3604 spin_lock(&imp->imp_lock);
3605 imp->imp_pingable = 0;
3606 spin_unlock(&imp->imp_lock);
3609 case OBD_CLEANUP_EXPORTS: {
3610 struct client_obd *cli = &obd->u.cli;
3612 * for echo client, export may be on zombie list, wait for
3613 * zombie thread to cull it, because cli.cl_import will be
3614 * cleared in client_disconnect_export():
3615 * class_export_destroy() -> obd_cleanup() ->
3616 * echo_device_free() -> echo_client_cleanup() ->
3617 * obd_disconnect() -> osc_disconnect() ->
3618 * client_disconnect_export()
3620 obd_zombie_barrier();
3621 if (cli->cl_writeback_work) {
3622 ptlrpcd_destroy_work(cli->cl_writeback_work);
3623 cli->cl_writeback_work = NULL;
3625 if (cli->cl_lru_work) {
3626 ptlrpcd_destroy_work(cli->cl_lru_work);
3627 cli->cl_lru_work = NULL;
3629 obd_cleanup_client_import(obd);
3630 ptlrpc_lprocfs_unregister_obd(obd);
3631 lprocfs_obd_cleanup(obd);
3632 rc = obd_llog_finish(obd, 0);
3634 CERROR("failed to cleanup llogging subsystems\n");
3641 int osc_cleanup(struct obd_device *obd)
3643 struct client_obd *cli = &obd->u.cli;
3649 if (cli->cl_cache != NULL) {
3650 LASSERT(atomic_read(&cli->cl_cache->ccc_users) > 0);
3651 spin_lock(&cli->cl_cache->ccc_lru_lock);
3652 cfs_list_del_init(&cli->cl_lru_osc);
3653 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3654 cli->cl_lru_left = NULL;
3655 atomic_dec(&cli->cl_cache->ccc_users);
3656 cli->cl_cache = NULL;
3659 /* free memory of osc quota cache */
3660 osc_quota_cleanup(obd);
3662 rc = client_obd_cleanup(obd);
3668 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
3670 int rc = class_process_proc_seq_param(PARAM_OSC, obd->obd_vars,
3672 return rc > 0 ? 0: rc;
3675 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
3677 return osc_process_config_base(obd, buf);
3680 struct obd_ops osc_obd_ops = {
3681 .o_owner = THIS_MODULE,
3682 .o_setup = osc_setup,
3683 .o_precleanup = osc_precleanup,
3684 .o_cleanup = osc_cleanup,
3685 .o_add_conn = client_import_add_conn,
3686 .o_del_conn = client_import_del_conn,
3687 .o_connect = client_connect_import,
3688 .o_reconnect = osc_reconnect,
3689 .o_disconnect = osc_disconnect,
3690 .o_statfs = osc_statfs,
3691 .o_statfs_async = osc_statfs_async,
3692 .o_packmd = osc_packmd,
3693 .o_unpackmd = osc_unpackmd,
3694 .o_create = osc_create,
3695 .o_destroy = osc_destroy,
3696 .o_getattr = osc_getattr,
3697 .o_getattr_async = osc_getattr_async,
3698 .o_setattr = osc_setattr,
3699 .o_setattr_async = osc_setattr_async,
3701 .o_punch = osc_punch,
3703 .o_enqueue = osc_enqueue,
3704 .o_change_cbdata = osc_change_cbdata,
3705 .o_find_cbdata = osc_find_cbdata,
3706 .o_cancel = osc_cancel,
3707 .o_cancel_unused = osc_cancel_unused,
3708 .o_iocontrol = osc_iocontrol,
3709 .o_get_info = osc_get_info,
3710 .o_set_info_async = osc_set_info_async,
3711 .o_import_event = osc_import_event,
3712 .o_llog_init = osc_llog_init,
3713 .o_llog_finish = osc_llog_finish,
3714 .o_process_config = osc_process_config,
3715 .o_quotactl = osc_quotactl,
3716 .o_quotacheck = osc_quotacheck,
3719 extern struct lu_kmem_descr osc_caches[];
3720 extern spinlock_t osc_ast_guard;
3721 extern struct lock_class_key osc_ast_guard_class;
3723 int __init osc_init(void)
3728 /* print an address of _any_ initialized kernel symbol from this
3729 * module, to allow debugging with gdb that doesn't support data
3730 * symbols from modules.*/
3731 CDEBUG(D_INFO, "Lustre OSC module (%p).\n", &osc_caches);
3733 rc = lu_kmem_init(osc_caches);
3737 rc = class_register_type(&osc_obd_ops, NULL, NULL,
3738 #ifndef HAVE_ONLY_PROCFS_SEQ
3741 LUSTRE_OSC_NAME, &osc_device_type);
3743 lu_kmem_fini(osc_caches);
3747 spin_lock_init(&osc_ast_guard);
3748 lockdep_set_class(&osc_ast_guard, &osc_ast_guard_class);
3754 static void /*__exit*/ osc_exit(void)
3756 class_unregister_type(LUSTRE_OSC_NAME);
3757 lu_kmem_fini(osc_caches);
3760 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
3761 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
3762 MODULE_LICENSE("GPL");
3764 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);
git://git.whamcloud.com / fs / lustre-release.git / blob