4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Whamcloud, Inc.
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 "osc_internal.h"
62 #include "osc_cl_internal.h"
64 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
65 static int brw_interpret(const struct lu_env *env,
66 struct ptlrpc_request *req, void *data, int rc);
67 int osc_cleanup(struct obd_device *obd);
69 /* Pack OSC object metadata for disk storage (LE byte order). */
70 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
71 struct lov_stripe_md *lsm)
76 lmm_size = sizeof(**lmmp);
81 OBD_FREE(*lmmp, lmm_size);
87 OBD_ALLOC(*lmmp, lmm_size);
93 LASSERT(lsm->lsm_object_id);
94 LASSERT_SEQ_IS_MDT(lsm->lsm_object_seq);
95 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
96 (*lmmp)->lmm_object_seq = cpu_to_le64(lsm->lsm_object_seq);
102 /* Unpack OSC object metadata from disk storage (LE byte order). */
103 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
104 struct lov_mds_md *lmm, int lmm_bytes)
107 struct obd_import *imp = class_exp2cliimp(exp);
111 if (lmm_bytes < sizeof (*lmm)) {
112 CERROR("lov_mds_md too small: %d, need %d\n",
113 lmm_bytes, (int)sizeof(*lmm));
116 /* XXX LOV_MAGIC etc check? */
118 if (lmm->lmm_object_id == 0) {
119 CERROR("lov_mds_md: zero lmm_object_id\n");
124 lsm_size = lov_stripe_md_size(1);
128 if (*lsmp != NULL && lmm == NULL) {
129 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
130 OBD_FREE(*lsmp, lsm_size);
136 OBD_ALLOC(*lsmp, lsm_size);
139 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
140 if ((*lsmp)->lsm_oinfo[0] == NULL) {
141 OBD_FREE(*lsmp, lsm_size);
144 loi_init((*lsmp)->lsm_oinfo[0]);
148 /* XXX zero *lsmp? */
149 (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
150 (*lsmp)->lsm_object_seq = le64_to_cpu (lmm->lmm_object_seq);
151 LASSERT((*lsmp)->lsm_object_id);
152 LASSERT_SEQ_IS_MDT((*lsmp)->lsm_object_seq);
156 (imp->imp_connect_data.ocd_connect_flags & OBD_CONNECT_MAXBYTES))
157 (*lsmp)->lsm_maxbytes = imp->imp_connect_data.ocd_maxbytes;
159 (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
164 static inline void osc_pack_capa(struct ptlrpc_request *req,
165 struct ost_body *body, void *capa)
167 struct obd_capa *oc = (struct obd_capa *)capa;
168 struct lustre_capa *c;
173 c = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
176 body->oa.o_valid |= OBD_MD_FLOSSCAPA;
177 DEBUG_CAPA(D_SEC, c, "pack");
180 static inline void osc_pack_req_body(struct ptlrpc_request *req,
181 struct obd_info *oinfo)
183 struct ost_body *body;
185 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
188 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
189 osc_pack_capa(req, body, oinfo->oi_capa);
192 static inline void osc_set_capa_size(struct ptlrpc_request *req,
193 const struct req_msg_field *field,
197 req_capsule_set_size(&req->rq_pill, field, RCL_CLIENT, 0);
199 /* it is already calculated as sizeof struct obd_capa */
203 static int osc_getattr_interpret(const struct lu_env *env,
204 struct ptlrpc_request *req,
205 struct osc_async_args *aa, int rc)
207 struct ost_body *body;
213 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
215 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
216 lustre_get_wire_obdo(aa->aa_oi->oi_oa, &body->oa);
218 /* This should really be sent by the OST */
219 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
220 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
222 CDEBUG(D_INFO, "can't unpack ost_body\n");
224 aa->aa_oi->oi_oa->o_valid = 0;
227 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
231 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
232 struct ptlrpc_request_set *set)
234 struct ptlrpc_request *req;
235 struct osc_async_args *aa;
239 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
243 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
244 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
246 ptlrpc_request_free(req);
250 osc_pack_req_body(req, oinfo);
252 ptlrpc_request_set_replen(req);
253 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
255 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
256 aa = ptlrpc_req_async_args(req);
259 ptlrpc_set_add_req(set, req);
263 static int osc_getattr(const struct lu_env *env, struct obd_export *exp,
264 struct obd_info *oinfo)
266 struct ptlrpc_request *req;
267 struct ost_body *body;
271 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
275 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
276 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
278 ptlrpc_request_free(req);
282 osc_pack_req_body(req, oinfo);
284 ptlrpc_request_set_replen(req);
286 rc = ptlrpc_queue_wait(req);
290 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
292 GOTO(out, rc = -EPROTO);
294 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
295 lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
297 /* This should really be sent by the OST */
298 oinfo->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
299 oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
303 ptlrpc_req_finished(req);
307 static int osc_setattr(const struct lu_env *env, struct obd_export *exp,
308 struct obd_info *oinfo, struct obd_trans_info *oti)
310 struct ptlrpc_request *req;
311 struct ost_body *body;
315 LASSERT(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP);
317 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
321 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
322 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
324 ptlrpc_request_free(req);
328 osc_pack_req_body(req, oinfo);
330 ptlrpc_request_set_replen(req);
332 rc = ptlrpc_queue_wait(req);
336 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
338 GOTO(out, rc = -EPROTO);
340 lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
344 ptlrpc_req_finished(req);
348 static int osc_setattr_interpret(const struct lu_env *env,
349 struct ptlrpc_request *req,
350 struct osc_setattr_args *sa, int rc)
352 struct ost_body *body;
358 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
360 GOTO(out, rc = -EPROTO);
362 lustre_get_wire_obdo(sa->sa_oa, &body->oa);
364 rc = sa->sa_upcall(sa->sa_cookie, rc);
368 int osc_setattr_async_base(struct obd_export *exp, struct obd_info *oinfo,
369 struct obd_trans_info *oti,
370 obd_enqueue_update_f upcall, void *cookie,
371 struct ptlrpc_request_set *rqset)
373 struct ptlrpc_request *req;
374 struct osc_setattr_args *sa;
378 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
382 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
383 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
385 ptlrpc_request_free(req);
389 if (oti && oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
390 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
392 osc_pack_req_body(req, oinfo);
394 ptlrpc_request_set_replen(req);
396 /* do mds to ost setattr asynchronously */
398 /* Do not wait for response. */
399 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
401 req->rq_interpret_reply =
402 (ptlrpc_interpterer_t)osc_setattr_interpret;
404 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
405 sa = ptlrpc_req_async_args(req);
406 sa->sa_oa = oinfo->oi_oa;
407 sa->sa_upcall = upcall;
408 sa->sa_cookie = cookie;
410 if (rqset == PTLRPCD_SET)
411 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
413 ptlrpc_set_add_req(rqset, req);
419 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
420 struct obd_trans_info *oti,
421 struct ptlrpc_request_set *rqset)
423 return osc_setattr_async_base(exp, oinfo, oti,
424 oinfo->oi_cb_up, oinfo, rqset);
427 int osc_real_create(struct obd_export *exp, struct obdo *oa,
428 struct lov_stripe_md **ea, struct obd_trans_info *oti)
430 struct ptlrpc_request *req;
431 struct ost_body *body;
432 struct lov_stripe_md *lsm;
441 rc = obd_alloc_memmd(exp, &lsm);
446 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
448 GOTO(out, rc = -ENOMEM);
450 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
452 ptlrpc_request_free(req);
456 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
458 lustre_set_wire_obdo(&body->oa, oa);
460 ptlrpc_request_set_replen(req);
462 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
463 oa->o_flags == OBD_FL_DELORPHAN) {
465 "delorphan from OST integration");
466 /* Don't resend the delorphan req */
467 req->rq_no_resend = req->rq_no_delay = 1;
470 rc = ptlrpc_queue_wait(req);
474 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
476 GOTO(out_req, rc = -EPROTO);
478 lustre_get_wire_obdo(oa, &body->oa);
480 /* This should really be sent by the OST */
481 oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
482 oa->o_valid |= OBD_MD_FLBLKSZ;
484 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
485 * have valid lsm_oinfo data structs, so don't go touching that.
486 * This needs to be fixed in a big way.
488 lsm->lsm_object_id = oa->o_id;
489 lsm->lsm_object_seq = oa->o_seq;
493 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
495 if (oa->o_valid & OBD_MD_FLCOOKIE) {
496 if (!oti->oti_logcookies)
497 oti_alloc_cookies(oti, 1);
498 *oti->oti_logcookies = oa->o_lcookie;
502 CDEBUG(D_HA, "transno: "LPD64"\n",
503 lustre_msg_get_transno(req->rq_repmsg));
505 ptlrpc_req_finished(req);
508 obd_free_memmd(exp, &lsm);
512 int osc_punch_base(struct obd_export *exp, struct obd_info *oinfo,
513 obd_enqueue_update_f upcall, void *cookie,
514 struct ptlrpc_request_set *rqset)
516 struct ptlrpc_request *req;
517 struct osc_setattr_args *sa;
518 struct ost_body *body;
522 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
526 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
527 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
529 ptlrpc_request_free(req);
532 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
533 ptlrpc_at_set_req_timeout(req);
535 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
537 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
538 osc_pack_capa(req, body, oinfo->oi_capa);
540 ptlrpc_request_set_replen(req);
542 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
543 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
544 sa = ptlrpc_req_async_args(req);
545 sa->sa_oa = oinfo->oi_oa;
546 sa->sa_upcall = upcall;
547 sa->sa_cookie = cookie;
548 if (rqset == PTLRPCD_SET)
549 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
551 ptlrpc_set_add_req(rqset, req);
556 static int osc_punch(const struct lu_env *env, struct obd_export *exp,
557 struct obd_info *oinfo, struct obd_trans_info *oti,
558 struct ptlrpc_request_set *rqset)
560 oinfo->oi_oa->o_size = oinfo->oi_policy.l_extent.start;
561 oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
562 oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
563 return osc_punch_base(exp, oinfo,
564 oinfo->oi_cb_up, oinfo, rqset);
567 static int osc_sync_interpret(const struct lu_env *env,
568 struct ptlrpc_request *req,
571 struct osc_fsync_args *fa = arg;
572 struct ost_body *body;
578 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
580 CERROR ("can't unpack ost_body\n");
581 GOTO(out, rc = -EPROTO);
584 *fa->fa_oi->oi_oa = body->oa;
586 rc = fa->fa_upcall(fa->fa_cookie, rc);
590 int osc_sync_base(struct obd_export *exp, struct obd_info *oinfo,
591 obd_enqueue_update_f upcall, void *cookie,
592 struct ptlrpc_request_set *rqset)
594 struct ptlrpc_request *req;
595 struct ost_body *body;
596 struct osc_fsync_args *fa;
600 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
604 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
605 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
607 ptlrpc_request_free(req);
611 /* overload the size and blocks fields in the oa with start/end */
612 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
614 lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
615 osc_pack_capa(req, body, oinfo->oi_capa);
617 ptlrpc_request_set_replen(req);
618 req->rq_interpret_reply = osc_sync_interpret;
620 CLASSERT(sizeof(*fa) <= sizeof(req->rq_async_args));
621 fa = ptlrpc_req_async_args(req);
623 fa->fa_upcall = upcall;
624 fa->fa_cookie = cookie;
626 if (rqset == PTLRPCD_SET)
627 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
629 ptlrpc_set_add_req(rqset, req);
634 static int osc_sync(const struct lu_env *env, struct obd_export *exp,
635 struct obd_info *oinfo, obd_size start, obd_size end,
636 struct ptlrpc_request_set *set)
641 CDEBUG(D_INFO, "oa NULL\n");
645 oinfo->oi_oa->o_size = start;
646 oinfo->oi_oa->o_blocks = end;
647 oinfo->oi_oa->o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
649 RETURN(osc_sync_base(exp, oinfo, oinfo->oi_cb_up, oinfo, set));
652 /* Find and cancel locally locks matched by @mode in the resource found by
653 * @objid. Found locks are added into @cancel list. Returns the amount of
654 * locks added to @cancels list. */
655 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
657 ldlm_mode_t mode, int lock_flags)
659 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
660 struct ldlm_res_id res_id;
661 struct ldlm_resource *res;
665 osc_build_res_name(oa->o_id, oa->o_seq, &res_id);
666 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
670 LDLM_RESOURCE_ADDREF(res);
671 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
672 lock_flags, 0, NULL);
673 LDLM_RESOURCE_DELREF(res);
674 ldlm_resource_putref(res);
678 static int osc_destroy_interpret(const struct lu_env *env,
679 struct ptlrpc_request *req, void *data,
682 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
684 cfs_atomic_dec(&cli->cl_destroy_in_flight);
685 cfs_waitq_signal(&cli->cl_destroy_waitq);
689 static int osc_can_send_destroy(struct client_obd *cli)
691 if (cfs_atomic_inc_return(&cli->cl_destroy_in_flight) <=
692 cli->cl_max_rpcs_in_flight) {
693 /* The destroy request can be sent */
696 if (cfs_atomic_dec_return(&cli->cl_destroy_in_flight) <
697 cli->cl_max_rpcs_in_flight) {
699 * The counter has been modified between the two atomic
702 cfs_waitq_signal(&cli->cl_destroy_waitq);
707 /* Destroy requests can be async always on the client, and we don't even really
708 * care about the return code since the client cannot do anything at all about
710 * When the MDS is unlinking a filename, it saves the file objects into a
711 * recovery llog, and these object records are cancelled when the OST reports
712 * they were destroyed and sync'd to disk (i.e. transaction committed).
713 * If the client dies, or the OST is down when the object should be destroyed,
714 * the records are not cancelled, and when the OST reconnects to the MDS next,
715 * it will retrieve the llog unlink logs and then sends the log cancellation
716 * cookies to the MDS after committing destroy transactions. */
717 static int osc_destroy(const struct lu_env *env, struct obd_export *exp,
718 struct obdo *oa, struct lov_stripe_md *ea,
719 struct obd_trans_info *oti, struct obd_export *md_export,
722 struct client_obd *cli = &exp->exp_obd->u.cli;
723 struct ptlrpc_request *req;
724 struct ost_body *body;
725 CFS_LIST_HEAD(cancels);
730 CDEBUG(D_INFO, "oa NULL\n");
734 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
735 LDLM_FL_DISCARD_DATA);
737 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
739 ldlm_lock_list_put(&cancels, l_bl_ast, count);
743 osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
744 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
747 ptlrpc_request_free(req);
751 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
752 ptlrpc_at_set_req_timeout(req);
754 if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
755 oa->o_lcookie = *oti->oti_logcookies;
756 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
758 lustre_set_wire_obdo(&body->oa, oa);
760 osc_pack_capa(req, body, (struct obd_capa *)capa);
761 ptlrpc_request_set_replen(req);
763 /* don't throttle destroy RPCs for the MDT */
764 if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
765 req->rq_interpret_reply = osc_destroy_interpret;
766 if (!osc_can_send_destroy(cli)) {
767 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP,
771 * Wait until the number of on-going destroy RPCs drops
772 * under max_rpc_in_flight
774 l_wait_event_exclusive(cli->cl_destroy_waitq,
775 osc_can_send_destroy(cli), &lwi);
779 /* Do not wait for response */
780 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
784 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
787 obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
789 LASSERT(!(oa->o_valid & bits));
792 client_obd_list_lock(&cli->cl_loi_list_lock);
793 oa->o_dirty = cli->cl_dirty;
794 if (cli->cl_dirty - cli->cl_dirty_transit > cli->cl_dirty_max) {
795 CERROR("dirty %lu - %lu > dirty_max %lu\n",
796 cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
798 } else if (cfs_atomic_read(&obd_dirty_pages) -
799 cfs_atomic_read(&obd_dirty_transit_pages) >
800 obd_max_dirty_pages + 1){
801 /* The cfs_atomic_read() allowing the cfs_atomic_inc() are
802 * not covered by a lock thus they may safely race and trip
803 * this CERROR() unless we add in a small fudge factor (+1). */
804 CERROR("dirty %d - %d > system dirty_max %d\n",
805 cfs_atomic_read(&obd_dirty_pages),
806 cfs_atomic_read(&obd_dirty_transit_pages),
807 obd_max_dirty_pages);
809 } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
810 CERROR("dirty %lu - dirty_max %lu too big???\n",
811 cli->cl_dirty, cli->cl_dirty_max);
814 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
815 (cli->cl_max_rpcs_in_flight + 1);
816 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
818 oa->o_grant = cli->cl_avail_grant + cli->cl_reserved_grant;
819 oa->o_dropped = cli->cl_lost_grant;
820 cli->cl_lost_grant = 0;
821 client_obd_list_unlock(&cli->cl_loi_list_lock);
822 CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
823 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
827 void osc_update_next_shrink(struct client_obd *cli)
829 cli->cl_next_shrink_grant =
830 cfs_time_shift(cli->cl_grant_shrink_interval);
831 CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
832 cli->cl_next_shrink_grant);
835 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
837 client_obd_list_lock(&cli->cl_loi_list_lock);
838 cli->cl_avail_grant += grant;
839 client_obd_list_unlock(&cli->cl_loi_list_lock);
842 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
844 if (body->oa.o_valid & OBD_MD_FLGRANT) {
845 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
846 __osc_update_grant(cli, body->oa.o_grant);
850 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
851 obd_count keylen, void *key, obd_count vallen,
852 void *val, struct ptlrpc_request_set *set);
854 static int osc_shrink_grant_interpret(const struct lu_env *env,
855 struct ptlrpc_request *req,
858 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
859 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
860 struct ost_body *body;
863 __osc_update_grant(cli, oa->o_grant);
867 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
869 osc_update_grant(cli, body);
875 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
877 client_obd_list_lock(&cli->cl_loi_list_lock);
878 oa->o_grant = cli->cl_avail_grant / 4;
879 cli->cl_avail_grant -= oa->o_grant;
880 client_obd_list_unlock(&cli->cl_loi_list_lock);
881 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
882 oa->o_valid |= OBD_MD_FLFLAGS;
885 oa->o_flags |= OBD_FL_SHRINK_GRANT;
886 osc_update_next_shrink(cli);
889 /* Shrink the current grant, either from some large amount to enough for a
890 * full set of in-flight RPCs, or if we have already shrunk to that limit
891 * then to enough for a single RPC. This avoids keeping more grant than
892 * needed, and avoids shrinking the grant piecemeal. */
893 static int osc_shrink_grant(struct client_obd *cli)
895 long target = (cli->cl_max_rpcs_in_flight + 1) *
896 cli->cl_max_pages_per_rpc;
898 client_obd_list_lock(&cli->cl_loi_list_lock);
899 if (cli->cl_avail_grant <= target)
900 target = cli->cl_max_pages_per_rpc;
901 client_obd_list_unlock(&cli->cl_loi_list_lock);
903 return osc_shrink_grant_to_target(cli, target);
906 int osc_shrink_grant_to_target(struct client_obd *cli, long target)
909 struct ost_body *body;
912 client_obd_list_lock(&cli->cl_loi_list_lock);
913 /* Don't shrink if we are already above or below the desired limit
914 * We don't want to shrink below a single RPC, as that will negatively
915 * impact block allocation and long-term performance. */
916 if (target < cli->cl_max_pages_per_rpc)
917 target = cli->cl_max_pages_per_rpc;
919 if (target >= cli->cl_avail_grant) {
920 client_obd_list_unlock(&cli->cl_loi_list_lock);
923 client_obd_list_unlock(&cli->cl_loi_list_lock);
929 osc_announce_cached(cli, &body->oa, 0);
931 client_obd_list_lock(&cli->cl_loi_list_lock);
932 body->oa.o_grant = cli->cl_avail_grant - target;
933 cli->cl_avail_grant = target;
934 client_obd_list_unlock(&cli->cl_loi_list_lock);
935 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
936 body->oa.o_valid |= OBD_MD_FLFLAGS;
937 body->oa.o_flags = 0;
939 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
940 osc_update_next_shrink(cli);
942 rc = osc_set_info_async(NULL, cli->cl_import->imp_obd->obd_self_export,
943 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
944 sizeof(*body), body, NULL);
946 __osc_update_grant(cli, body->oa.o_grant);
951 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
952 static int osc_should_shrink_grant(struct client_obd *client)
954 cfs_time_t time = cfs_time_current();
955 cfs_time_t next_shrink = client->cl_next_shrink_grant;
957 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
958 OBD_CONNECT_GRANT_SHRINK) == 0)
961 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
962 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
963 client->cl_avail_grant > GRANT_SHRINK_LIMIT)
966 osc_update_next_shrink(client);
971 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
973 struct client_obd *client;
975 cfs_list_for_each_entry(client, &item->ti_obd_list,
976 cl_grant_shrink_list) {
977 if (osc_should_shrink_grant(client))
978 osc_shrink_grant(client);
983 static int osc_add_shrink_grant(struct client_obd *client)
987 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
989 osc_grant_shrink_grant_cb, NULL,
990 &client->cl_grant_shrink_list);
992 CERROR("add grant client %s error %d\n",
993 client->cl_import->imp_obd->obd_name, rc);
996 CDEBUG(D_CACHE, "add grant client %s \n",
997 client->cl_import->imp_obd->obd_name);
998 osc_update_next_shrink(client);
1002 static int osc_del_shrink_grant(struct client_obd *client)
1004 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1008 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1011 * ocd_grant is the total grant amount we're expect to hold: if we've
1012 * been evicted, it's the new avail_grant amount, cl_dirty will drop
1013 * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1015 * race is tolerable here: if we're evicted, but imp_state already
1016 * left EVICTED state, then cl_dirty must be 0 already.
1018 client_obd_list_lock(&cli->cl_loi_list_lock);
1019 if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1020 cli->cl_avail_grant = ocd->ocd_grant;
1022 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1024 if (cli->cl_avail_grant < 0) {
1025 CWARN("%s: available grant < 0, the OSS is probably not running"
1026 " with patch from bug20278 (%ld) \n",
1027 cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant);
1028 /* workaround for 1.6 servers which do not have
1029 * the patch from bug20278 */
1030 cli->cl_avail_grant = ocd->ocd_grant;
1033 /* determine the appropriate chunk size used by osc_extent. */
1034 cli->cl_chunkbits = max_t(int, CFS_PAGE_SHIFT, ocd->ocd_blocksize);
1035 client_obd_list_unlock(&cli->cl_loi_list_lock);
1037 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld."
1038 "chunk bits: %d.\n", cli->cl_import->imp_obd->obd_name,
1039 cli->cl_avail_grant, cli->cl_lost_grant, cli->cl_chunkbits);
1041 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1042 cfs_list_empty(&cli->cl_grant_shrink_list))
1043 osc_add_shrink_grant(cli);
1046 /* We assume that the reason this OSC got a short read is because it read
1047 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1048 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1049 * this stripe never got written at or beyond this stripe offset yet. */
1050 static void handle_short_read(int nob_read, obd_count page_count,
1051 struct brw_page **pga)
1056 /* skip bytes read OK */
1057 while (nob_read > 0) {
1058 LASSERT (page_count > 0);
1060 if (pga[i]->count > nob_read) {
1061 /* EOF inside this page */
1062 ptr = cfs_kmap(pga[i]->pg) +
1063 (pga[i]->off & ~CFS_PAGE_MASK);
1064 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1065 cfs_kunmap(pga[i]->pg);
1071 nob_read -= pga[i]->count;
1076 /* zero remaining pages */
1077 while (page_count-- > 0) {
1078 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1079 memset(ptr, 0, pga[i]->count);
1080 cfs_kunmap(pga[i]->pg);
1085 static int check_write_rcs(struct ptlrpc_request *req,
1086 int requested_nob, int niocount,
1087 obd_count page_count, struct brw_page **pga)
1092 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1093 sizeof(*remote_rcs) *
1095 if (remote_rcs == NULL) {
1096 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1100 /* return error if any niobuf was in error */
1101 for (i = 0; i < niocount; i++) {
1102 if ((int)remote_rcs[i] < 0)
1103 return(remote_rcs[i]);
1105 if (remote_rcs[i] != 0) {
1106 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1107 i, remote_rcs[i], req);
1112 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1113 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1114 req->rq_bulk->bd_nob_transferred, requested_nob);
1121 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1123 if (p1->flag != p2->flag) {
1124 unsigned mask = ~(OBD_BRW_FROM_GRANT| OBD_BRW_NOCACHE|
1125 OBD_BRW_SYNC|OBD_BRW_ASYNC|OBD_BRW_NOQUOTA);
1127 /* warn if we try to combine flags that we don't know to be
1128 * safe to combine */
1129 if (unlikely((p1->flag & mask) != (p2->flag & mask))) {
1130 CWARN("Saw flags 0x%x and 0x%x in the same brw, please "
1131 "report this at http://bugs.whamcloud.com/\n",
1132 p1->flag, p2->flag);
1137 return (p1->off + p1->count == p2->off);
1140 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1141 struct brw_page **pga, int opc,
1142 cksum_type_t cksum_type)
1147 LASSERT (pg_count > 0);
1148 cksum = init_checksum(cksum_type);
1149 while (nob > 0 && pg_count > 0) {
1150 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1151 int off = pga[i]->off & ~CFS_PAGE_MASK;
1152 int count = pga[i]->count > nob ? nob : pga[i]->count;
1154 /* corrupt the data before we compute the checksum, to
1155 * simulate an OST->client data error */
1156 if (i == 0 && opc == OST_READ &&
1157 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1158 memcpy(ptr + off, "bad1", min(4, nob));
1159 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1160 cfs_kunmap(pga[i]->pg);
1161 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1164 nob -= pga[i]->count;
1168 /* For sending we only compute the wrong checksum instead
1169 * of corrupting the data so it is still correct on a redo */
1170 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1173 return fini_checksum(cksum, cksum_type);
1176 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1177 struct lov_stripe_md *lsm, obd_count page_count,
1178 struct brw_page **pga,
1179 struct ptlrpc_request **reqp,
1180 struct obd_capa *ocapa, int reserve,
1183 struct ptlrpc_request *req;
1184 struct ptlrpc_bulk_desc *desc;
1185 struct ost_body *body;
1186 struct obd_ioobj *ioobj;
1187 struct niobuf_remote *niobuf;
1188 int niocount, i, requested_nob, opc, rc;
1189 struct osc_brw_async_args *aa;
1190 struct req_capsule *pill;
1191 struct brw_page *pg_prev;
1194 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1195 RETURN(-ENOMEM); /* Recoverable */
1196 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1197 RETURN(-EINVAL); /* Fatal */
1199 if ((cmd & OBD_BRW_WRITE) != 0) {
1201 req = ptlrpc_request_alloc_pool(cli->cl_import,
1202 cli->cl_import->imp_rq_pool,
1203 &RQF_OST_BRW_WRITE);
1206 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1211 for (niocount = i = 1; i < page_count; i++) {
1212 if (!can_merge_pages(pga[i - 1], pga[i]))
1216 pill = &req->rq_pill;
1217 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1219 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1220 niocount * sizeof(*niobuf));
1221 osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1223 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1225 ptlrpc_request_free(req);
1228 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1229 ptlrpc_at_set_req_timeout(req);
1231 if (opc == OST_WRITE)
1232 desc = ptlrpc_prep_bulk_imp(req, page_count,
1233 BULK_GET_SOURCE, OST_BULK_PORTAL);
1235 desc = ptlrpc_prep_bulk_imp(req, page_count,
1236 BULK_PUT_SINK, OST_BULK_PORTAL);
1239 GOTO(out, rc = -ENOMEM);
1240 /* NB request now owns desc and will free it when it gets freed */
1242 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1243 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1244 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1245 LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1247 lustre_set_wire_obdo(&body->oa, oa);
1249 obdo_to_ioobj(oa, ioobj);
1250 ioobj->ioo_bufcnt = niocount;
1251 osc_pack_capa(req, body, ocapa);
1252 LASSERT (page_count > 0);
1254 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1255 struct brw_page *pg = pga[i];
1256 int poff = pg->off & ~CFS_PAGE_MASK;
1258 LASSERT(pg->count > 0);
1259 /* make sure there is no gap in the middle of page array */
1260 LASSERTF(page_count == 1 ||
1261 (ergo(i == 0, poff + pg->count == CFS_PAGE_SIZE) &&
1262 ergo(i > 0 && i < page_count - 1,
1263 poff == 0 && pg->count == CFS_PAGE_SIZE) &&
1264 ergo(i == page_count - 1, poff == 0)),
1265 "i: %d/%d pg: %p off: "LPU64", count: %u\n",
1266 i, page_count, pg, pg->off, pg->count);
1268 LASSERTF(i == 0 || pg->off > pg_prev->off,
1269 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1270 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1272 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1273 pg_prev->pg, page_private(pg_prev->pg),
1274 pg_prev->pg->index, pg_prev->off);
1276 LASSERTF(i == 0 || pg->off > pg_prev->off,
1277 "i %d p_c %u\n", i, page_count);
1279 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1280 (pg->flag & OBD_BRW_SRVLOCK));
1282 ptlrpc_prep_bulk_page(desc, pg->pg, poff, pg->count);
1283 requested_nob += pg->count;
1285 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1287 niobuf->len += pg->count;
1289 niobuf->offset = pg->off;
1290 niobuf->len = pg->count;
1291 niobuf->flags = pg->flag;
1296 LASSERTF((void *)(niobuf - niocount) ==
1297 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1298 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1299 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1301 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1303 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1304 body->oa.o_valid |= OBD_MD_FLFLAGS;
1305 body->oa.o_flags = 0;
1307 body->oa.o_flags |= OBD_FL_RECOV_RESEND;
1310 if (osc_should_shrink_grant(cli))
1311 osc_shrink_grant_local(cli, &body->oa);
1313 /* size[REQ_REC_OFF] still sizeof (*body) */
1314 if (opc == OST_WRITE) {
1315 if (cli->cl_checksum &&
1316 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1317 /* store cl_cksum_type in a local variable since
1318 * it can be changed via lprocfs */
1319 cksum_type_t cksum_type = cli->cl_cksum_type;
1321 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1322 oa->o_flags &= OBD_FL_LOCAL_MASK;
1323 body->oa.o_flags = 0;
1325 body->oa.o_flags |= cksum_type_pack(cksum_type);
1326 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1327 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1331 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1333 /* save this in 'oa', too, for later checking */
1334 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1335 oa->o_flags |= cksum_type_pack(cksum_type);
1337 /* clear out the checksum flag, in case this is a
1338 * resend but cl_checksum is no longer set. b=11238 */
1339 oa->o_valid &= ~OBD_MD_FLCKSUM;
1341 oa->o_cksum = body->oa.o_cksum;
1342 /* 1 RC per niobuf */
1343 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1344 sizeof(__u32) * niocount);
1346 if (cli->cl_checksum &&
1347 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1348 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1349 body->oa.o_flags = 0;
1350 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1351 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1354 ptlrpc_request_set_replen(req);
1356 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1357 aa = ptlrpc_req_async_args(req);
1359 aa->aa_requested_nob = requested_nob;
1360 aa->aa_nio_count = niocount;
1361 aa->aa_page_count = page_count;
1365 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1366 if (ocapa && reserve)
1367 aa->aa_ocapa = capa_get(ocapa);
1373 ptlrpc_req_finished(req);
1377 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1378 __u32 client_cksum, __u32 server_cksum, int nob,
1379 obd_count page_count, struct brw_page **pga,
1380 cksum_type_t client_cksum_type)
1384 cksum_type_t cksum_type;
1386 if (server_cksum == client_cksum) {
1387 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1391 cksum_type = cksum_type_unpack(oa->o_valid & OBD_MD_FLFLAGS ?
1393 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1396 if (cksum_type != client_cksum_type)
1397 msg = "the server did not use the checksum type specified in "
1398 "the original request - likely a protocol problem";
1399 else if (new_cksum == server_cksum)
1400 msg = "changed on the client after we checksummed it - "
1401 "likely false positive due to mmap IO (bug 11742)";
1402 else if (new_cksum == client_cksum)
1403 msg = "changed in transit before arrival at OST";
1405 msg = "changed in transit AND doesn't match the original - "
1406 "likely false positive due to mmap IO (bug 11742)";
1408 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1409 " object "LPU64"/"LPU64" extent ["LPU64"-"LPU64"]\n",
1410 msg, libcfs_nid2str(peer->nid),
1411 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1412 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1413 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1415 oa->o_valid & OBD_MD_FLGROUP ? oa->o_seq : (__u64)0,
1417 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1418 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1419 "client csum now %x\n", client_cksum, client_cksum_type,
1420 server_cksum, cksum_type, new_cksum);
1424 /* Note rc enters this function as number of bytes transferred */
1425 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1427 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1428 const lnet_process_id_t *peer =
1429 &req->rq_import->imp_connection->c_peer;
1430 struct client_obd *cli = aa->aa_cli;
1431 struct ost_body *body;
1432 __u32 client_cksum = 0;
1435 if (rc < 0 && rc != -EDQUOT) {
1436 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1440 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1441 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1443 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1447 /* set/clear over quota flag for a uid/gid */
1448 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1449 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1450 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1452 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid "LPX64", flags %x\n",
1453 body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1455 osc_quota_setdq(cli, qid, body->oa.o_valid, body->oa.o_flags);
1458 osc_update_grant(cli, body);
1463 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1464 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1466 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1468 CERROR("Unexpected +ve rc %d\n", rc);
1471 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1473 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1476 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1477 check_write_checksum(&body->oa, peer, client_cksum,
1478 body->oa.o_cksum, aa->aa_requested_nob,
1479 aa->aa_page_count, aa->aa_ppga,
1480 cksum_type_unpack(aa->aa_oa->o_flags)))
1483 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1484 aa->aa_page_count, aa->aa_ppga);
1488 /* The rest of this function executes only for OST_READs */
1490 /* if unwrap_bulk failed, return -EAGAIN to retry */
1491 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1493 GOTO(out, rc = -EAGAIN);
1495 if (rc > aa->aa_requested_nob) {
1496 CERROR("Unexpected rc %d (%d requested)\n", rc,
1497 aa->aa_requested_nob);
1501 if (rc != req->rq_bulk->bd_nob_transferred) {
1502 CERROR ("Unexpected rc %d (%d transferred)\n",
1503 rc, req->rq_bulk->bd_nob_transferred);
1507 if (rc < aa->aa_requested_nob)
1508 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1510 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1511 static int cksum_counter;
1512 __u32 server_cksum = body->oa.o_cksum;
1515 cksum_type_t cksum_type;
1517 cksum_type = cksum_type_unpack(body->oa.o_valid &OBD_MD_FLFLAGS?
1518 body->oa.o_flags : 0);
1519 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1520 aa->aa_ppga, OST_READ,
1523 if (peer->nid == req->rq_bulk->bd_sender) {
1527 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1530 if (server_cksum == ~0 && rc > 0) {
1531 CERROR("Protocol error: server %s set the 'checksum' "
1532 "bit, but didn't send a checksum. Not fatal, "
1533 "but please notify on http://bugs.whamcloud.com/\n",
1534 libcfs_nid2str(peer->nid));
1535 } else if (server_cksum != client_cksum) {
1536 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1537 "%s%s%s inode "DFID" object "
1538 LPU64"/"LPU64" extent "
1539 "["LPU64"-"LPU64"]\n",
1540 req->rq_import->imp_obd->obd_name,
1541 libcfs_nid2str(peer->nid),
1543 body->oa.o_valid & OBD_MD_FLFID ?
1544 body->oa.o_parent_seq : (__u64)0,
1545 body->oa.o_valid & OBD_MD_FLFID ?
1546 body->oa.o_parent_oid : 0,
1547 body->oa.o_valid & OBD_MD_FLFID ?
1548 body->oa.o_parent_ver : 0,
1550 body->oa.o_valid & OBD_MD_FLGROUP ?
1551 body->oa.o_seq : (__u64)0,
1552 aa->aa_ppga[0]->off,
1553 aa->aa_ppga[aa->aa_page_count-1]->off +
1554 aa->aa_ppga[aa->aa_page_count-1]->count -
1556 CERROR("client %x, server %x, cksum_type %x\n",
1557 client_cksum, server_cksum, cksum_type);
1559 aa->aa_oa->o_cksum = client_cksum;
1563 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1566 } else if (unlikely(client_cksum)) {
1567 static int cksum_missed;
1570 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1571 CERROR("Checksum %u requested from %s but not sent\n",
1572 cksum_missed, libcfs_nid2str(peer->nid));
1578 lustre_get_wire_obdo(aa->aa_oa, &body->oa);
1583 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1584 struct lov_stripe_md *lsm,
1585 obd_count page_count, struct brw_page **pga,
1586 struct obd_capa *ocapa)
1588 struct ptlrpc_request *req;
1591 int generation, resends = 0;
1592 struct l_wait_info lwi;
1596 cfs_waitq_init(&waitq);
1597 generation = exp->exp_obd->u.cli.cl_import->imp_generation;
1600 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1601 page_count, pga, &req, ocapa, 0, resends);
1606 req->rq_generation_set = 1;
1607 req->rq_import_generation = generation;
1608 req->rq_sent = cfs_time_current_sec() + resends;
1611 rc = ptlrpc_queue_wait(req);
1613 if (rc == -ETIMEDOUT && req->rq_resend) {
1614 DEBUG_REQ(D_HA, req, "BULK TIMEOUT");
1615 ptlrpc_req_finished(req);
1619 rc = osc_brw_fini_request(req, rc);
1621 ptlrpc_req_finished(req);
1622 /* When server return -EINPROGRESS, client should always retry
1623 * regardless of the number of times the bulk was resent already.*/
1624 if (osc_recoverable_error(rc)) {
1626 if (rc != -EINPROGRESS &&
1627 !client_should_resend(resends, &exp->exp_obd->u.cli)) {
1628 CERROR("%s: too many resend retries for object: "
1629 ""LPU64":"LPU64", rc = %d.\n",
1630 exp->exp_obd->obd_name, oa->o_id, oa->o_seq, rc);
1634 exp->exp_obd->u.cli.cl_import->imp_generation) {
1635 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1636 ""LPU64":"LPU64", rc = %d.\n",
1637 exp->exp_obd->obd_name, oa->o_id, oa->o_seq, rc);
1641 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL,
1643 l_wait_event(waitq, 0, &lwi);
1648 if (rc == -EAGAIN || rc == -EINPROGRESS)
1653 int osc_brw_redo_request(struct ptlrpc_request *request,
1654 struct osc_brw_async_args *aa)
1656 struct ptlrpc_request *new_req;
1657 struct osc_brw_async_args *new_aa;
1658 struct osc_async_page *oap;
1662 DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1664 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1665 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1666 aa->aa_cli, aa->aa_oa,
1667 NULL /* lsm unused by osc currently */,
1668 aa->aa_page_count, aa->aa_ppga,
1669 &new_req, aa->aa_ocapa, 0, 1);
1673 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1674 if (oap->oap_request != NULL) {
1675 LASSERTF(request == oap->oap_request,
1676 "request %p != oap_request %p\n",
1677 request, oap->oap_request);
1678 if (oap->oap_interrupted) {
1679 ptlrpc_req_finished(new_req);
1684 /* New request takes over pga and oaps from old request.
1685 * Note that copying a list_head doesn't work, need to move it... */
1687 new_req->rq_interpret_reply = request->rq_interpret_reply;
1688 new_req->rq_async_args = request->rq_async_args;
1689 new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1690 new_req->rq_generation_set = 1;
1691 new_req->rq_import_generation = request->rq_import_generation;
1693 new_aa = ptlrpc_req_async_args(new_req);
1695 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1696 cfs_list_splice_init(&aa->aa_oaps, &new_aa->aa_oaps);
1697 CFS_INIT_LIST_HEAD(&new_aa->aa_exts);
1698 cfs_list_splice_init(&aa->aa_exts, &new_aa->aa_exts);
1700 cfs_list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1701 if (oap->oap_request) {
1702 ptlrpc_req_finished(oap->oap_request);
1703 oap->oap_request = ptlrpc_request_addref(new_req);
1707 new_aa->aa_ocapa = aa->aa_ocapa;
1708 aa->aa_ocapa = NULL;
1710 /* XXX: This code will run into problem if we're going to support
1711 * to add a series of BRW RPCs into a self-defined ptlrpc_request_set
1712 * and wait for all of them to be finished. We should inherit request
1713 * set from old request. */
1714 ptlrpcd_add_req(new_req, PDL_POLICY_SAME, -1);
1716 DEBUG_REQ(D_INFO, new_req, "new request");
1721 * ugh, we want disk allocation on the target to happen in offset order. we'll
1722 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1723 * fine for our small page arrays and doesn't require allocation. its an
1724 * insertion sort that swaps elements that are strides apart, shrinking the
1725 * stride down until its '1' and the array is sorted.
1727 static void sort_brw_pages(struct brw_page **array, int num)
1730 struct brw_page *tmp;
1734 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1739 for (i = stride ; i < num ; i++) {
1742 while (j >= stride && array[j - stride]->off > tmp->off) {
1743 array[j] = array[j - stride];
1748 } while (stride > 1);
1751 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1757 LASSERT (pages > 0);
1758 offset = pg[i]->off & ~CFS_PAGE_MASK;
1762 if (pages == 0) /* that's all */
1765 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1766 return count; /* doesn't end on page boundary */
1769 offset = pg[i]->off & ~CFS_PAGE_MASK;
1770 if (offset != 0) /* doesn't start on page boundary */
1777 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1779 struct brw_page **ppga;
1782 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1786 for (i = 0; i < count; i++)
1791 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1793 LASSERT(ppga != NULL);
1794 OBD_FREE(ppga, sizeof(*ppga) * count);
1797 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1798 obd_count page_count, struct brw_page *pga,
1799 struct obd_trans_info *oti)
1801 struct obdo *saved_oa = NULL;
1802 struct brw_page **ppga, **orig;
1803 struct obd_import *imp = class_exp2cliimp(exp);
1804 struct client_obd *cli;
1805 int rc, page_count_orig;
1808 LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1809 cli = &imp->imp_obd->u.cli;
1811 if (cmd & OBD_BRW_CHECK) {
1812 /* The caller just wants to know if there's a chance that this
1813 * I/O can succeed */
1815 if (imp->imp_invalid)
1820 /* test_brw with a failed create can trip this, maybe others. */
1821 LASSERT(cli->cl_max_pages_per_rpc);
1825 orig = ppga = osc_build_ppga(pga, page_count);
1828 page_count_orig = page_count;
1830 sort_brw_pages(ppga, page_count);
1831 while (page_count) {
1832 obd_count pages_per_brw;
1834 if (page_count > cli->cl_max_pages_per_rpc)
1835 pages_per_brw = cli->cl_max_pages_per_rpc;
1837 pages_per_brw = page_count;
1839 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1841 if (saved_oa != NULL) {
1842 /* restore previously saved oa */
1843 *oinfo->oi_oa = *saved_oa;
1844 } else if (page_count > pages_per_brw) {
1845 /* save a copy of oa (brw will clobber it) */
1846 OBDO_ALLOC(saved_oa);
1847 if (saved_oa == NULL)
1848 GOTO(out, rc = -ENOMEM);
1849 *saved_oa = *oinfo->oi_oa;
1852 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1853 pages_per_brw, ppga, oinfo->oi_capa);
1858 page_count -= pages_per_brw;
1859 ppga += pages_per_brw;
1863 osc_release_ppga(orig, page_count_orig);
1865 if (saved_oa != NULL)
1866 OBDO_FREE(saved_oa);
1871 static int brw_interpret(const struct lu_env *env,
1872 struct ptlrpc_request *req, void *data, int rc)
1874 struct osc_brw_async_args *aa = data;
1875 struct osc_extent *ext;
1876 struct osc_extent *tmp;
1877 struct cl_object *obj = NULL;
1878 struct client_obd *cli = aa->aa_cli;
1881 rc = osc_brw_fini_request(req, rc);
1882 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
1883 /* When server return -EINPROGRESS, client should always retry
1884 * regardless of the number of times the bulk was resent already. */
1885 if (osc_recoverable_error(rc)) {
1886 if (req->rq_import_generation !=
1887 req->rq_import->imp_generation) {
1888 CDEBUG(D_HA, "%s: resend cross eviction for object: "
1889 ""LPU64":"LPU64", rc = %d.\n",
1890 req->rq_import->imp_obd->obd_name,
1891 aa->aa_oa->o_id, aa->aa_oa->o_seq, rc);
1892 } else if (rc == -EINPROGRESS ||
1893 client_should_resend(aa->aa_resends, aa->aa_cli)) {
1894 rc = osc_brw_redo_request(req, aa);
1896 CERROR("%s: too many resent retries for object: "
1897 ""LPU64":"LPU64", rc = %d.\n",
1898 req->rq_import->imp_obd->obd_name,
1899 aa->aa_oa->o_id, aa->aa_oa->o_seq, rc);
1904 else if (rc == -EAGAIN || rc == -EINPROGRESS)
1909 capa_put(aa->aa_ocapa);
1910 aa->aa_ocapa = NULL;
1913 cfs_list_for_each_entry_safe(ext, tmp, &aa->aa_exts, oe_link) {
1914 if (obj == NULL && rc == 0) {
1915 obj = osc2cl(ext->oe_obj);
1919 cfs_list_del_init(&ext->oe_link);
1920 osc_extent_finish(env, ext, 1, rc);
1922 LASSERT(cfs_list_empty(&aa->aa_exts));
1923 LASSERT(cfs_list_empty(&aa->aa_oaps));
1926 struct obdo *oa = aa->aa_oa;
1927 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
1928 unsigned long valid = 0;
1931 if (oa->o_valid & OBD_MD_FLBLOCKS) {
1932 attr->cat_blocks = oa->o_blocks;
1933 valid |= CAT_BLOCKS;
1935 if (oa->o_valid & OBD_MD_FLMTIME) {
1936 attr->cat_mtime = oa->o_mtime;
1939 if (oa->o_valid & OBD_MD_FLATIME) {
1940 attr->cat_atime = oa->o_atime;
1943 if (oa->o_valid & OBD_MD_FLCTIME) {
1944 attr->cat_ctime = oa->o_ctime;
1948 cl_object_attr_lock(obj);
1949 cl_object_attr_set(env, obj, attr, valid);
1950 cl_object_attr_unlock(obj);
1952 cl_object_put(env, obj);
1954 OBDO_FREE(aa->aa_oa);
1956 cl_req_completion(env, aa->aa_clerq, rc < 0 ? rc :
1957 req->rq_bulk->bd_nob_transferred);
1958 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
1959 ptlrpc_lprocfs_brw(req, req->rq_bulk->bd_nob_transferred);
1961 client_obd_list_lock(&cli->cl_loi_list_lock);
1962 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
1963 * is called so we know whether to go to sync BRWs or wait for more
1964 * RPCs to complete */
1965 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
1966 cli->cl_w_in_flight--;
1968 cli->cl_r_in_flight--;
1969 osc_wake_cache_waiters(cli);
1970 client_obd_list_unlock(&cli->cl_loi_list_lock);
1972 osc_io_unplug(env, cli, NULL, PDL_POLICY_SAME);
1977 * Build an RPC by the list of extent @ext_list. The caller must ensure
1978 * that the total pages in this list are NOT over max pages per RPC.
1979 * Extents in the list must be in OES_RPC state.
1981 int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
1982 cfs_list_t *ext_list, int cmd, pdl_policy_t pol)
1984 struct ptlrpc_request *req = NULL;
1985 struct osc_extent *ext;
1986 CFS_LIST_HEAD(rpc_list);
1987 struct brw_page **pga = NULL;
1988 struct osc_brw_async_args *aa = NULL;
1989 struct obdo *oa = NULL;
1990 struct osc_async_page *oap;
1991 struct osc_async_page *tmp;
1992 struct cl_req *clerq = NULL;
1993 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
1994 struct ldlm_lock *lock = NULL;
1995 struct cl_req_attr crattr;
1996 obd_off starting_offset = OBD_OBJECT_EOF;
1997 obd_off ending_offset = 0;
1998 int i, rc, mpflag = 0, mem_tight = 0, page_count = 0;
2001 LASSERT(!cfs_list_empty(ext_list));
2003 /* add pages into rpc_list to build BRW rpc */
2004 cfs_list_for_each_entry(ext, ext_list, oe_link) {
2005 LASSERT(ext->oe_state == OES_RPC);
2006 mem_tight |= ext->oe_memalloc;
2007 cfs_list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
2009 cfs_list_add_tail(&oap->oap_rpc_item, &rpc_list);
2010 if (starting_offset > oap->oap_obj_off)
2011 starting_offset = oap->oap_obj_off;
2013 LASSERT(oap->oap_page_off == 0);
2014 if (ending_offset < oap->oap_obj_off + oap->oap_count)
2015 ending_offset = oap->oap_obj_off +
2018 LASSERT(oap->oap_page_off + oap->oap_count ==
2024 mpflag = cfs_memory_pressure_get_and_set();
2026 memset(&crattr, 0, sizeof crattr);
2027 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2029 GOTO(out, rc = -ENOMEM);
2033 GOTO(out, rc = -ENOMEM);
2036 cfs_list_for_each_entry(oap, &rpc_list, oap_rpc_item) {
2037 struct cl_page *page = oap2cl_page(oap);
2038 if (clerq == NULL) {
2039 clerq = cl_req_alloc(env, page, crt,
2040 1 /* only 1-object rpcs for
2043 GOTO(out, rc = PTR_ERR(clerq));
2044 lock = oap->oap_ldlm_lock;
2047 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
2048 pga[i] = &oap->oap_brw_page;
2049 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2050 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2051 pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2053 cl_req_page_add(env, clerq, page);
2056 /* always get the data for the obdo for the rpc */
2057 LASSERT(clerq != NULL);
2059 crattr.cra_capa = NULL;
2060 memset(crattr.cra_jobid, 0, JOBSTATS_JOBID_SIZE);
2061 cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2063 oa->o_handle = lock->l_remote_handle;
2064 oa->o_valid |= OBD_MD_FLHANDLE;
2067 rc = cl_req_prep(env, clerq);
2069 CERROR("cl_req_prep failed: %d\n", rc);
2073 sort_brw_pages(pga, page_count);
2074 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2075 pga, &req, crattr.cra_capa, 1, 0);
2077 CERROR("prep_req failed: %d\n", rc);
2081 req->rq_interpret_reply = brw_interpret;
2083 req->rq_memalloc = 1;
2085 /* Need to update the timestamps after the request is built in case
2086 * we race with setattr (locally or in queue at OST). If OST gets
2087 * later setattr before earlier BRW (as determined by the request xid),
2088 * the OST will not use BRW timestamps. Sadly, there is no obvious
2089 * way to do this in a single call. bug 10150 */
2090 cl_req_attr_set(env, clerq, &crattr,
2091 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2093 lustre_msg_set_jobid(req->rq_reqmsg, crattr.cra_jobid);
2095 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2096 aa = ptlrpc_req_async_args(req);
2097 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2098 cfs_list_splice_init(&rpc_list, &aa->aa_oaps);
2099 CFS_INIT_LIST_HEAD(&aa->aa_exts);
2100 cfs_list_splice_init(ext_list, &aa->aa_exts);
2101 aa->aa_clerq = clerq;
2103 /* queued sync pages can be torn down while the pages
2104 * were between the pending list and the rpc */
2106 cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2107 /* only one oap gets a request reference */
2110 if (oap->oap_interrupted && !req->rq_intr) {
2111 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2113 ptlrpc_mark_interrupted(req);
2117 tmp->oap_request = ptlrpc_request_addref(req);
2119 client_obd_list_lock(&cli->cl_loi_list_lock);
2120 starting_offset >>= CFS_PAGE_SHIFT;
2121 if (cmd == OBD_BRW_READ) {
2122 cli->cl_r_in_flight++;
2123 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2124 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2125 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2126 starting_offset + 1);
2128 cli->cl_w_in_flight++;
2129 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2130 lprocfs_oh_tally(&cli->cl_write_rpc_hist, cli->cl_w_in_flight);
2131 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2132 starting_offset + 1);
2134 client_obd_list_unlock(&cli->cl_loi_list_lock);
2136 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2137 page_count, aa, cli->cl_r_in_flight,
2138 cli->cl_w_in_flight);
2140 /* XXX: Maybe the caller can check the RPC bulk descriptor to
2141 * see which CPU/NUMA node the majority of pages were allocated
2142 * on, and try to assign the async RPC to the CPU core
2143 * (PDL_POLICY_PREFERRED) to reduce cross-CPU memory traffic.
2145 * But on the other hand, we expect that multiple ptlrpcd
2146 * threads and the initial write sponsor can run in parallel,
2147 * especially when data checksum is enabled, which is CPU-bound
2148 * operation and single ptlrpcd thread cannot process in time.
2149 * So more ptlrpcd threads sharing BRW load
2150 * (with PDL_POLICY_ROUND) seems better.
2152 ptlrpcd_add_req(req, pol, -1);
2158 cfs_memory_pressure_restore(mpflag);
2160 capa_put(crattr.cra_capa);
2162 LASSERT(req == NULL);
2167 OBD_FREE(pga, sizeof(*pga) * page_count);
2168 /* this should happen rarely and is pretty bad, it makes the
2169 * pending list not follow the dirty order */
2170 while (!cfs_list_empty(ext_list)) {
2171 ext = cfs_list_entry(ext_list->next, struct osc_extent,
2173 cfs_list_del_init(&ext->oe_link);
2174 osc_extent_finish(env, ext, 0, rc);
2176 if (clerq && !IS_ERR(clerq))
2177 cl_req_completion(env, clerq, rc);
2182 static int osc_set_lock_data_with_check(struct ldlm_lock *lock,
2183 struct ldlm_enqueue_info *einfo)
2185 void *data = einfo->ei_cbdata;
2188 LASSERT(lock != NULL);
2189 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2190 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2191 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2192 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2194 lock_res_and_lock(lock);
2195 cfs_spin_lock(&osc_ast_guard);
2197 if (lock->l_ast_data == NULL)
2198 lock->l_ast_data = data;
2199 if (lock->l_ast_data == data)
2202 cfs_spin_unlock(&osc_ast_guard);
2203 unlock_res_and_lock(lock);
2208 static int osc_set_data_with_check(struct lustre_handle *lockh,
2209 struct ldlm_enqueue_info *einfo)
2211 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2215 set = osc_set_lock_data_with_check(lock, einfo);
2216 LDLM_LOCK_PUT(lock);
2218 CERROR("lockh %p, data %p - client evicted?\n",
2219 lockh, einfo->ei_cbdata);
2223 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2224 ldlm_iterator_t replace, void *data)
2226 struct ldlm_res_id res_id;
2227 struct obd_device *obd = class_exp2obd(exp);
2229 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
2230 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2234 /* find any ldlm lock of the inode in osc
2238 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2239 ldlm_iterator_t replace, void *data)
2241 struct ldlm_res_id res_id;
2242 struct obd_device *obd = class_exp2obd(exp);
2245 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
2246 rc = ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2247 if (rc == LDLM_ITER_STOP)
2249 if (rc == LDLM_ITER_CONTINUE)
2254 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
2255 obd_enqueue_update_f upcall, void *cookie,
2256 int *flags, int agl, int rc)
2258 int intent = *flags & LDLM_FL_HAS_INTENT;
2262 /* The request was created before ldlm_cli_enqueue call. */
2263 if (rc == ELDLM_LOCK_ABORTED) {
2264 struct ldlm_reply *rep;
2265 rep = req_capsule_server_get(&req->rq_pill,
2268 LASSERT(rep != NULL);
2269 if (rep->lock_policy_res1)
2270 rc = rep->lock_policy_res1;
2274 if ((intent != 0 && rc == ELDLM_LOCK_ABORTED && agl == 0) ||
2276 *flags |= LDLM_FL_LVB_READY;
2277 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
2278 lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
2281 /* Call the update callback. */
2282 rc = (*upcall)(cookie, rc);
2286 static int osc_enqueue_interpret(const struct lu_env *env,
2287 struct ptlrpc_request *req,
2288 struct osc_enqueue_args *aa, int rc)
2290 struct ldlm_lock *lock;
2291 struct lustre_handle handle;
2293 struct ost_lvb *lvb;
2295 int *flags = aa->oa_flags;
2297 /* Make a local copy of a lock handle and a mode, because aa->oa_*
2298 * might be freed anytime after lock upcall has been called. */
2299 lustre_handle_copy(&handle, aa->oa_lockh);
2300 mode = aa->oa_ei->ei_mode;
2302 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2304 lock = ldlm_handle2lock(&handle);
2306 /* Take an additional reference so that a blocking AST that
2307 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2308 * to arrive after an upcall has been executed by
2309 * osc_enqueue_fini(). */
2310 ldlm_lock_addref(&handle, mode);
2312 /* Let CP AST to grant the lock first. */
2313 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
2315 if (aa->oa_agl && rc == ELDLM_LOCK_ABORTED) {
2320 lvb_len = sizeof(*aa->oa_lvb);
2323 /* Complete obtaining the lock procedure. */
2324 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
2325 mode, flags, lvb, lvb_len, &handle, rc);
2326 /* Complete osc stuff. */
2327 rc = osc_enqueue_fini(req, aa->oa_lvb, aa->oa_upcall, aa->oa_cookie,
2328 flags, aa->oa_agl, rc);
2330 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
2332 /* Release the lock for async request. */
2333 if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
2335 * Releases a reference taken by ldlm_cli_enqueue(), if it is
2336 * not already released by
2337 * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
2339 ldlm_lock_decref(&handle, mode);
2341 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
2342 aa->oa_lockh, req, aa);
2343 ldlm_lock_decref(&handle, mode);
2344 LDLM_LOCK_PUT(lock);
2348 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
2349 struct lov_oinfo *loi, int flags,
2350 struct ost_lvb *lvb, __u32 mode, int rc)
2352 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
2354 if (rc == ELDLM_OK) {
2357 LASSERT(lock != NULL);
2358 loi->loi_lvb = *lvb;
2359 tmp = loi->loi_lvb.lvb_size;
2360 /* Extend KMS up to the end of this lock and no further
2361 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
2362 if (tmp > lock->l_policy_data.l_extent.end)
2363 tmp = lock->l_policy_data.l_extent.end + 1;
2364 if (tmp >= loi->loi_kms) {
2365 LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
2366 ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
2367 loi_kms_set(loi, tmp);
2369 LDLM_DEBUG(lock, "lock acquired, setting rss="
2370 LPU64"; leaving kms="LPU64", end="LPU64,
2371 loi->loi_lvb.lvb_size, loi->loi_kms,
2372 lock->l_policy_data.l_extent.end);
2374 ldlm_lock_allow_match(lock);
2375 } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
2376 LASSERT(lock != NULL);
2377 loi->loi_lvb = *lvb;
2378 ldlm_lock_allow_match(lock);
2379 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
2380 " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
2386 ldlm_lock_fail_match(lock);
2388 LDLM_LOCK_PUT(lock);
2391 EXPORT_SYMBOL(osc_update_enqueue);
2393 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
2395 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
2396 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
2397 * other synchronous requests, however keeping some locks and trying to obtain
2398 * others may take a considerable amount of time in a case of ost failure; and
2399 * when other sync requests do not get released lock from a client, the client
2400 * is excluded from the cluster -- such scenarious make the life difficult, so
2401 * release locks just after they are obtained. */
2402 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2403 int *flags, ldlm_policy_data_t *policy,
2404 struct ost_lvb *lvb, int kms_valid,
2405 obd_enqueue_update_f upcall, void *cookie,
2406 struct ldlm_enqueue_info *einfo,
2407 struct lustre_handle *lockh,
2408 struct ptlrpc_request_set *rqset, int async, int agl)
2410 struct obd_device *obd = exp->exp_obd;
2411 struct ptlrpc_request *req = NULL;
2412 int intent = *flags & LDLM_FL_HAS_INTENT;
2413 int match_lvb = (agl != 0 ? 0 : LDLM_FL_LVB_READY);
2418 /* Filesystem lock extents are extended to page boundaries so that
2419 * dealing with the page cache is a little smoother. */
2420 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
2421 policy->l_extent.end |= ~CFS_PAGE_MASK;
2424 * kms is not valid when either object is completely fresh (so that no
2425 * locks are cached), or object was evicted. In the latter case cached
2426 * lock cannot be used, because it would prime inode state with
2427 * potentially stale LVB.
2432 /* Next, search for already existing extent locks that will cover us */
2433 /* If we're trying to read, we also search for an existing PW lock. The
2434 * VFS and page cache already protect us locally, so lots of readers/
2435 * writers can share a single PW lock.
2437 * There are problems with conversion deadlocks, so instead of
2438 * converting a read lock to a write lock, we'll just enqueue a new
2441 * At some point we should cancel the read lock instead of making them
2442 * send us a blocking callback, but there are problems with canceling
2443 * locks out from other users right now, too. */
2444 mode = einfo->ei_mode;
2445 if (einfo->ei_mode == LCK_PR)
2447 mode = ldlm_lock_match(obd->obd_namespace, *flags | match_lvb, res_id,
2448 einfo->ei_type, policy, mode, lockh, 0);
2450 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
2452 if ((agl != 0) && !(matched->l_flags & LDLM_FL_LVB_READY)) {
2453 /* For AGL, if enqueue RPC is sent but the lock is not
2454 * granted, then skip to process this strpe.
2455 * Return -ECANCELED to tell the caller. */
2456 ldlm_lock_decref(lockh, mode);
2457 LDLM_LOCK_PUT(matched);
2459 } else if (osc_set_lock_data_with_check(matched, einfo)) {
2460 *flags |= LDLM_FL_LVB_READY;
2461 /* addref the lock only if not async requests and PW
2462 * lock is matched whereas we asked for PR. */
2463 if (!rqset && einfo->ei_mode != mode)
2464 ldlm_lock_addref(lockh, LCK_PR);
2466 /* I would like to be able to ASSERT here that
2467 * rss <= kms, but I can't, for reasons which
2468 * are explained in lov_enqueue() */
2471 /* We already have a lock, and it's referenced */
2472 (*upcall)(cookie, ELDLM_OK);
2474 if (einfo->ei_mode != mode)
2475 ldlm_lock_decref(lockh, LCK_PW);
2477 /* For async requests, decref the lock. */
2478 ldlm_lock_decref(lockh, einfo->ei_mode);
2479 LDLM_LOCK_PUT(matched);
2482 ldlm_lock_decref(lockh, mode);
2483 LDLM_LOCK_PUT(matched);
2489 CFS_LIST_HEAD(cancels);
2490 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2491 &RQF_LDLM_ENQUEUE_LVB);
2495 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
2497 ptlrpc_request_free(req);
2501 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
2503 ptlrpc_request_set_replen(req);
2506 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
2507 *flags &= ~LDLM_FL_BLOCK_GRANTED;
2509 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
2510 sizeof(*lvb), lockh, async);
2513 struct osc_enqueue_args *aa;
2514 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
2515 aa = ptlrpc_req_async_args(req);
2518 aa->oa_flags = flags;
2519 aa->oa_upcall = upcall;
2520 aa->oa_cookie = cookie;
2522 aa->oa_lockh = lockh;
2525 req->rq_interpret_reply =
2526 (ptlrpc_interpterer_t)osc_enqueue_interpret;
2527 if (rqset == PTLRPCD_SET)
2528 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
2530 ptlrpc_set_add_req(rqset, req);
2531 } else if (intent) {
2532 ptlrpc_req_finished(req);
2537 rc = osc_enqueue_fini(req, lvb, upcall, cookie, flags, agl, rc);
2539 ptlrpc_req_finished(req);
2544 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
2545 struct ldlm_enqueue_info *einfo,
2546 struct ptlrpc_request_set *rqset)
2548 struct ldlm_res_id res_id;
2552 osc_build_res_name(oinfo->oi_md->lsm_object_id,
2553 oinfo->oi_md->lsm_object_seq, &res_id);
2555 rc = osc_enqueue_base(exp, &res_id, &oinfo->oi_flags, &oinfo->oi_policy,
2556 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
2557 oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid,
2558 oinfo->oi_cb_up, oinfo, einfo, oinfo->oi_lockh,
2559 rqset, rqset != NULL, 0);
2563 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2564 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
2565 int *flags, void *data, struct lustre_handle *lockh,
2568 struct obd_device *obd = exp->exp_obd;
2569 int lflags = *flags;
2573 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
2576 /* Filesystem lock extents are extended to page boundaries so that
2577 * dealing with the page cache is a little smoother */
2578 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
2579 policy->l_extent.end |= ~CFS_PAGE_MASK;
2581 /* Next, search for already existing extent locks that will cover us */
2582 /* If we're trying to read, we also search for an existing PW lock. The
2583 * VFS and page cache already protect us locally, so lots of readers/
2584 * writers can share a single PW lock. */
2588 rc = ldlm_lock_match(obd->obd_namespace, lflags,
2589 res_id, type, policy, rc, lockh, unref);
2592 if (!osc_set_data_with_check(lockh, data)) {
2593 if (!(lflags & LDLM_FL_TEST_LOCK))
2594 ldlm_lock_decref(lockh, rc);
2598 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
2599 ldlm_lock_addref(lockh, LCK_PR);
2600 ldlm_lock_decref(lockh, LCK_PW);
2607 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
2611 if (unlikely(mode == LCK_GROUP))
2612 ldlm_lock_decref_and_cancel(lockh, mode);
2614 ldlm_lock_decref(lockh, mode);
2619 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
2620 __u32 mode, struct lustre_handle *lockh)
2623 RETURN(osc_cancel_base(lockh, mode));
2626 static int osc_cancel_unused(struct obd_export *exp,
2627 struct lov_stripe_md *lsm,
2628 ldlm_cancel_flags_t flags,
2631 struct obd_device *obd = class_exp2obd(exp);
2632 struct ldlm_res_id res_id, *resp = NULL;
2635 resp = osc_build_res_name(lsm->lsm_object_id,
2636 lsm->lsm_object_seq, &res_id);
2639 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
2642 static int osc_statfs_interpret(const struct lu_env *env,
2643 struct ptlrpc_request *req,
2644 struct osc_async_args *aa, int rc)
2646 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
2647 struct obd_statfs *msfs;
2652 /* The request has in fact never been sent
2653 * due to issues at a higher level (LOV).
2654 * Exit immediately since the caller is
2655 * aware of the problem and takes care
2656 * of the clean up */
2659 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
2660 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY))
2666 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2668 GOTO(out, rc = -EPROTO);
2671 /* Reinitialize the RDONLY and DEGRADED flags at the client
2672 * on each statfs, so they don't stay set permanently. */
2673 cfs_spin_lock(&cli->cl_oscc.oscc_lock);
2675 if (unlikely(msfs->os_state & OS_STATE_DEGRADED))
2676 cli->cl_oscc.oscc_flags |= OSCC_FLAG_DEGRADED;
2677 else if (unlikely(cli->cl_oscc.oscc_flags & OSCC_FLAG_DEGRADED))
2678 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_DEGRADED;
2680 if (unlikely(msfs->os_state & OS_STATE_READONLY))
2681 cli->cl_oscc.oscc_flags |= OSCC_FLAG_RDONLY;
2682 else if (unlikely(cli->cl_oscc.oscc_flags & OSCC_FLAG_RDONLY))
2683 cli->cl_oscc.oscc_flags &= ~OSCC_FLAG_RDONLY;
2685 /* Add a bit of hysteresis so this flag isn't continually flapping,
2686 * and ensure that new files don't get extremely fragmented due to
2687 * only a small amount of available space in the filesystem.
2688 * We want to set the NOSPC flag when there is less than ~0.1% free
2689 * and clear it when there is at least ~0.2% free space, so:
2690 * avail < ~0.1% max max = avail + used
2691 * 1025 * avail < avail + used used = blocks - free
2692 * 1024 * avail < used
2693 * 1024 * avail < blocks - free
2694 * avail < ((blocks - free) >> 10)
2696 * On very large disk, say 16TB 0.1% will be 16 GB. We don't want to
2697 * lose that amount of space so in those cases we report no space left
2698 * if their is less than 1 GB left. */
2699 used = min_t(__u64,(msfs->os_blocks - msfs->os_bfree) >> 10, 1 << 30);
2700 if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) == 0) &&
2701 ((msfs->os_ffree < 32) || (msfs->os_bavail < used))))
2702 cli->cl_oscc.oscc_flags |= OSCC_FLAG_NOSPC;
2703 else if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) != 0) &&
2704 (msfs->os_ffree > 64) &&
2705 (msfs->os_bavail > (used << 1)))) {
2706 cli->cl_oscc.oscc_flags &= ~(OSCC_FLAG_NOSPC |
2707 OSCC_FLAG_NOSPC_BLK);
2710 if (unlikely(((cli->cl_oscc.oscc_flags & OSCC_FLAG_NOSPC) != 0) &&
2711 (msfs->os_bavail < used)))
2712 cli->cl_oscc.oscc_flags |= OSCC_FLAG_NOSPC_BLK;
2714 cfs_spin_unlock(&cli->cl_oscc.oscc_lock);
2716 *aa->aa_oi->oi_osfs = *msfs;
2718 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
2722 static int osc_statfs_async(struct obd_export *exp,
2723 struct obd_info *oinfo, __u64 max_age,
2724 struct ptlrpc_request_set *rqset)
2726 struct obd_device *obd = class_exp2obd(exp);
2727 struct ptlrpc_request *req;
2728 struct osc_async_args *aa;
2732 /* We could possibly pass max_age in the request (as an absolute
2733 * timestamp or a "seconds.usec ago") so the target can avoid doing
2734 * extra calls into the filesystem if that isn't necessary (e.g.
2735 * during mount that would help a bit). Having relative timestamps
2736 * is not so great if request processing is slow, while absolute
2737 * timestamps are not ideal because they need time synchronization. */
2738 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
2742 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2744 ptlrpc_request_free(req);
2747 ptlrpc_request_set_replen(req);
2748 req->rq_request_portal = OST_CREATE_PORTAL;
2749 ptlrpc_at_set_req_timeout(req);
2751 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
2752 /* procfs requests not want stat in wait for avoid deadlock */
2753 req->rq_no_resend = 1;
2754 req->rq_no_delay = 1;
2757 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
2758 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
2759 aa = ptlrpc_req_async_args(req);
2762 ptlrpc_set_add_req(rqset, req);
2766 static int osc_statfs(const struct lu_env *env, struct obd_export *exp,
2767 struct obd_statfs *osfs, __u64 max_age, __u32 flags)
2769 struct obd_device *obd = class_exp2obd(exp);
2770 struct obd_statfs *msfs;
2771 struct ptlrpc_request *req;
2772 struct obd_import *imp = NULL;
2776 /*Since the request might also come from lprocfs, so we need
2777 *sync this with client_disconnect_export Bug15684*/
2778 cfs_down_read(&obd->u.cli.cl_sem);
2779 if (obd->u.cli.cl_import)
2780 imp = class_import_get(obd->u.cli.cl_import);
2781 cfs_up_read(&obd->u.cli.cl_sem);
2785 /* We could possibly pass max_age in the request (as an absolute
2786 * timestamp or a "seconds.usec ago") so the target can avoid doing
2787 * extra calls into the filesystem if that isn't necessary (e.g.
2788 * during mount that would help a bit). Having relative timestamps
2789 * is not so great if request processing is slow, while absolute
2790 * timestamps are not ideal because they need time synchronization. */
2791 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
2793 class_import_put(imp);
2798 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2800 ptlrpc_request_free(req);
2803 ptlrpc_request_set_replen(req);
2804 req->rq_request_portal = OST_CREATE_PORTAL;
2805 ptlrpc_at_set_req_timeout(req);
2807 if (flags & OBD_STATFS_NODELAY) {
2808 /* procfs requests not want stat in wait for avoid deadlock */
2809 req->rq_no_resend = 1;
2810 req->rq_no_delay = 1;
2813 rc = ptlrpc_queue_wait(req);
2817 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2819 GOTO(out, rc = -EPROTO);
2826 ptlrpc_req_finished(req);
2830 /* Retrieve object striping information.
2832 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
2833 * the maximum number of OST indices which will fit in the user buffer.
2834 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
2836 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
2838 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
2839 struct lov_user_md_v3 lum, *lumk;
2840 struct lov_user_ost_data_v1 *lmm_objects;
2841 int rc = 0, lum_size;
2847 /* we only need the header part from user space to get lmm_magic and
2848 * lmm_stripe_count, (the header part is common to v1 and v3) */
2849 lum_size = sizeof(struct lov_user_md_v1);
2850 if (cfs_copy_from_user(&lum, lump, lum_size))
2853 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
2854 (lum.lmm_magic != LOV_USER_MAGIC_V3))
2857 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
2858 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
2859 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
2860 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
2862 /* we can use lov_mds_md_size() to compute lum_size
2863 * because lov_user_md_vX and lov_mds_md_vX have the same size */
2864 if (lum.lmm_stripe_count > 0) {
2865 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
2866 OBD_ALLOC(lumk, lum_size);
2870 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
2871 lmm_objects = &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
2873 lmm_objects = &(lumk->lmm_objects[0]);
2874 lmm_objects->l_object_id = lsm->lsm_object_id;
2876 lum_size = lov_mds_md_size(0, lum.lmm_magic);
2880 lumk->lmm_object_id = lsm->lsm_object_id;
2881 lumk->lmm_object_seq = lsm->lsm_object_seq;
2882 lumk->lmm_stripe_count = 1;
2884 if (cfs_copy_to_user(lump, lumk, lum_size))
2888 OBD_FREE(lumk, lum_size);
2894 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
2895 void *karg, void *uarg)
2897 struct obd_device *obd = exp->exp_obd;
2898 struct obd_ioctl_data *data = karg;
2902 if (!cfs_try_module_get(THIS_MODULE)) {
2903 CERROR("Can't get module. Is it alive?");
2907 case OBD_IOC_LOV_GET_CONFIG: {
2909 struct lov_desc *desc;
2910 struct obd_uuid uuid;
2914 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
2915 GOTO(out, err = -EINVAL);
2917 data = (struct obd_ioctl_data *)buf;
2919 if (sizeof(*desc) > data->ioc_inllen1) {
2920 obd_ioctl_freedata(buf, len);
2921 GOTO(out, err = -EINVAL);
2924 if (data->ioc_inllen2 < sizeof(uuid)) {
2925 obd_ioctl_freedata(buf, len);
2926 GOTO(out, err = -EINVAL);
2929 desc = (struct lov_desc *)data->ioc_inlbuf1;
2930 desc->ld_tgt_count = 1;
2931 desc->ld_active_tgt_count = 1;
2932 desc->ld_default_stripe_count = 1;
2933 desc->ld_default_stripe_size = 0;
2934 desc->ld_default_stripe_offset = 0;
2935 desc->ld_pattern = 0;
2936 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
2938 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
2940 err = cfs_copy_to_user((void *)uarg, buf, len);
2943 obd_ioctl_freedata(buf, len);
2946 case LL_IOC_LOV_SETSTRIPE:
2947 err = obd_alloc_memmd(exp, karg);
2951 case LL_IOC_LOV_GETSTRIPE:
2952 err = osc_getstripe(karg, uarg);
2954 case OBD_IOC_CLIENT_RECOVER:
2955 err = ptlrpc_recover_import(obd->u.cli.cl_import,
2956 data->ioc_inlbuf1, 0);
2960 case IOC_OSC_SET_ACTIVE:
2961 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
2964 case OBD_IOC_POLL_QUOTACHECK:
2965 err = osc_quota_poll_check(exp, (struct if_quotacheck *)karg);
2967 case OBD_IOC_PING_TARGET:
2968 err = ptlrpc_obd_ping(obd);
2971 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
2972 cmd, cfs_curproc_comm());
2973 GOTO(out, err = -ENOTTY);
2976 cfs_module_put(THIS_MODULE);
2980 static int osc_get_info(const struct lu_env *env, struct obd_export *exp,
2981 obd_count keylen, void *key, __u32 *vallen, void *val,
2982 struct lov_stripe_md *lsm)
2985 if (!vallen || !val)
2988 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
2989 __u32 *stripe = val;
2990 *vallen = sizeof(*stripe);
2993 } else if (KEY_IS(KEY_LAST_ID)) {
2994 struct ptlrpc_request *req;
2999 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3000 &RQF_OST_GET_INFO_LAST_ID);
3004 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3005 RCL_CLIENT, keylen);
3006 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3008 ptlrpc_request_free(req);
3012 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3013 memcpy(tmp, key, keylen);
3015 req->rq_no_delay = req->rq_no_resend = 1;
3016 ptlrpc_request_set_replen(req);
3017 rc = ptlrpc_queue_wait(req);
3021 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
3023 GOTO(out, rc = -EPROTO);
3025 *((obd_id *)val) = *reply;
3027 ptlrpc_req_finished(req);
3029 } else if (KEY_IS(KEY_FIEMAP)) {
3030 struct ptlrpc_request *req;
3031 struct ll_user_fiemap *reply;
3035 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3036 &RQF_OST_GET_INFO_FIEMAP);
3040 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
3041 RCL_CLIENT, keylen);
3042 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3043 RCL_CLIENT, *vallen);
3044 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3045 RCL_SERVER, *vallen);
3047 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3049 ptlrpc_request_free(req);
3053 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
3054 memcpy(tmp, key, keylen);
3055 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3056 memcpy(tmp, val, *vallen);
3058 ptlrpc_request_set_replen(req);
3059 rc = ptlrpc_queue_wait(req);
3063 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3065 GOTO(out1, rc = -EPROTO);
3067 memcpy(val, reply, *vallen);
3069 ptlrpc_req_finished(req);
3077 static int osc_setinfo_mds_connect_import(struct obd_import *imp)
3079 struct llog_ctxt *ctxt;
3083 ctxt = llog_get_context(imp->imp_obd, LLOG_MDS_OST_ORIG_CTXT);
3085 rc = llog_initiator_connect(ctxt);
3086 llog_ctxt_put(ctxt);
3088 /* XXX return an error? skip setting below flags? */
3091 cfs_spin_lock(&imp->imp_lock);
3092 imp->imp_server_timeout = 1;
3093 imp->imp_pingable = 1;
3094 cfs_spin_unlock(&imp->imp_lock);
3095 CDEBUG(D_RPCTRACE, "pinging OST %s\n", obd2cli_tgt(imp->imp_obd));
3100 static int osc_setinfo_mds_conn_interpret(const struct lu_env *env,
3101 struct ptlrpc_request *req,
3108 RETURN(osc_setinfo_mds_connect_import(req->rq_import));
3111 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
3112 obd_count keylen, void *key, obd_count vallen,
3113 void *val, struct ptlrpc_request_set *set)
3115 struct ptlrpc_request *req;
3116 struct obd_device *obd = exp->exp_obd;
3117 struct obd_import *imp = class_exp2cliimp(exp);
3122 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
3124 if (KEY_IS(KEY_NEXT_ID)) {
3126 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3128 if (vallen != sizeof(obd_id))
3133 if (vallen != sizeof(obd_id))
3136 /* avoid race between allocate new object and set next id
3137 * from ll_sync thread */
3138 cfs_spin_lock(&oscc->oscc_lock);
3139 new_val = *((obd_id*)val) + 1;
3140 if (new_val > oscc->oscc_next_id)
3141 oscc->oscc_next_id = new_val;
3142 cfs_spin_unlock(&oscc->oscc_lock);
3143 CDEBUG(D_HA, "%s: set oscc_next_id = "LPU64"\n",
3144 exp->exp_obd->obd_name,
3145 obd->u.cli.cl_oscc.oscc_next_id);
3150 if (KEY_IS(KEY_CHECKSUM)) {
3151 if (vallen != sizeof(int))
3153 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
3157 if (KEY_IS(KEY_SPTLRPC_CONF)) {
3158 sptlrpc_conf_client_adapt(obd);
3162 if (KEY_IS(KEY_FLUSH_CTX)) {
3163 sptlrpc_import_flush_my_ctx(imp);
3167 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
3170 /* We pass all other commands directly to OST. Since nobody calls osc
3171 methods directly and everybody is supposed to go through LOV, we
3172 assume lov checked invalid values for us.
3173 The only recognised values so far are evict_by_nid and mds_conn.
3174 Even if something bad goes through, we'd get a -EINVAL from OST
3177 if (KEY_IS(KEY_GRANT_SHRINK))
3178 req = ptlrpc_request_alloc(imp, &RQF_OST_SET_GRANT_INFO);
3180 req = ptlrpc_request_alloc(imp, &RQF_OBD_SET_INFO);
3185 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3186 RCL_CLIENT, keylen);
3187 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
3188 RCL_CLIENT, vallen);
3189 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
3191 ptlrpc_request_free(req);
3195 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3196 memcpy(tmp, key, keylen);
3197 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_VAL);
3198 memcpy(tmp, val, vallen);
3200 if (KEY_IS(KEY_MDS_CONN)) {
3201 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3203 oscc->oscc_oa.o_seq = (*(__u32 *)val);
3204 oscc->oscc_oa.o_valid |= OBD_MD_FLGROUP;
3205 LASSERT_SEQ_IS_MDT(oscc->oscc_oa.o_seq);
3206 req->rq_no_delay = req->rq_no_resend = 1;
3207 req->rq_interpret_reply = osc_setinfo_mds_conn_interpret;
3208 } else if (KEY_IS(KEY_GRANT_SHRINK)) {
3209 struct osc_grant_args *aa;
3212 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3213 aa = ptlrpc_req_async_args(req);
3216 ptlrpc_req_finished(req);
3219 *oa = ((struct ost_body *)val)->oa;
3221 req->rq_interpret_reply = osc_shrink_grant_interpret;
3224 ptlrpc_request_set_replen(req);
3225 if (!KEY_IS(KEY_GRANT_SHRINK)) {
3226 LASSERT(set != NULL);
3227 ptlrpc_set_add_req(set, req);
3228 ptlrpc_check_set(NULL, set);
3230 ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
3236 static struct llog_operations osc_size_repl_logops = {
3237 lop_cancel: llog_obd_repl_cancel
3240 static struct llog_operations osc_mds_ost_orig_logops;
3242 static int __osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
3243 struct obd_device *tgt, struct llog_catid *catid)
3248 rc = llog_setup(obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT, tgt, 1,
3249 &catid->lci_logid, &osc_mds_ost_orig_logops);
3251 CERROR("failed LLOG_MDS_OST_ORIG_CTXT\n");
3255 rc = llog_setup(obd, &obd->obd_olg, LLOG_SIZE_REPL_CTXT, tgt, 1,
3256 NULL, &osc_size_repl_logops);
3258 struct llog_ctxt *ctxt =
3259 llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3262 CERROR("failed LLOG_SIZE_REPL_CTXT\n");
3267 CERROR("osc '%s' tgt '%s' catid %p rc=%d\n",
3268 obd->obd_name, tgt->obd_name, catid, rc);
3269 CERROR("logid "LPX64":0x%x\n",
3270 catid->lci_logid.lgl_oid, catid->lci_logid.lgl_ogen);
3275 static int osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
3276 struct obd_device *disk_obd, int *index)
3278 struct llog_catid catid;
3279 static char name[32] = CATLIST;
3283 LASSERT(olg == &obd->obd_olg);
3285 cfs_mutex_lock(&olg->olg_cat_processing);
3286 rc = llog_get_cat_list(disk_obd, name, *index, 1, &catid);
3288 CERROR("rc: %d\n", rc);
3292 CDEBUG(D_INFO, "%s: Init llog for %d - catid "LPX64"/"LPX64":%x\n",
3293 obd->obd_name, *index, catid.lci_logid.lgl_oid,
3294 catid.lci_logid.lgl_oseq, catid.lci_logid.lgl_ogen);
3296 rc = __osc_llog_init(obd, olg, disk_obd, &catid);
3298 CERROR("rc: %d\n", rc);
3302 rc = llog_put_cat_list(disk_obd, name, *index, 1, &catid);
3304 CERROR("rc: %d\n", rc);
3309 cfs_mutex_unlock(&olg->olg_cat_processing);
3314 static int osc_llog_finish(struct obd_device *obd, int count)
3316 struct llog_ctxt *ctxt;
3317 int rc = 0, rc2 = 0;
3320 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3322 rc = llog_cleanup(ctxt);
3324 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3326 rc2 = llog_cleanup(ctxt);
3333 static int osc_reconnect(const struct lu_env *env,
3334 struct obd_export *exp, struct obd_device *obd,
3335 struct obd_uuid *cluuid,
3336 struct obd_connect_data *data,
3339 struct client_obd *cli = &obd->u.cli;
3341 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
3344 client_obd_list_lock(&cli->cl_loi_list_lock);
3345 data->ocd_grant = (cli->cl_avail_grant + cli->cl_dirty) ?:
3346 2 * cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
3347 lost_grant = cli->cl_lost_grant;
3348 cli->cl_lost_grant = 0;
3349 client_obd_list_unlock(&cli->cl_loi_list_lock);
3351 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
3352 " ocd_grant: %d, lost: %ld.\n", data->ocd_connect_flags,
3353 data->ocd_version, data->ocd_grant, lost_grant);
3359 static int osc_disconnect(struct obd_export *exp)
3361 struct obd_device *obd = class_exp2obd(exp);
3362 struct llog_ctxt *ctxt;
3365 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3367 if (obd->u.cli.cl_conn_count == 1) {
3368 /* Flush any remaining cancel messages out to the
3370 llog_sync(ctxt, exp, 0);
3372 llog_ctxt_put(ctxt);
3374 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
3378 rc = client_disconnect_export(exp);
3380 * Initially we put del_shrink_grant before disconnect_export, but it
3381 * causes the following problem if setup (connect) and cleanup
3382 * (disconnect) are tangled together.
3383 * connect p1 disconnect p2
3384 * ptlrpc_connect_import
3385 * ............... class_manual_cleanup
3388 * ptlrpc_connect_interrupt
3390 * add this client to shrink list
3392 * Bang! pinger trigger the shrink.
3393 * So the osc should be disconnected from the shrink list, after we
3394 * are sure the import has been destroyed. BUG18662
3396 if (obd->u.cli.cl_import == NULL)
3397 osc_del_shrink_grant(&obd->u.cli);
3401 static int osc_import_event(struct obd_device *obd,
3402 struct obd_import *imp,
3403 enum obd_import_event event)
3405 struct client_obd *cli;
3409 LASSERT(imp->imp_obd == obd);
3412 case IMP_EVENT_DISCON: {
3413 /* Only do this on the MDS OSC's */
3414 if (imp->imp_server_timeout) {
3415 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3417 cfs_spin_lock(&oscc->oscc_lock);
3418 oscc->oscc_flags |= OSCC_FLAG_RECOVERING;
3419 cfs_spin_unlock(&oscc->oscc_lock);
3422 client_obd_list_lock(&cli->cl_loi_list_lock);
3423 cli->cl_avail_grant = 0;
3424 cli->cl_lost_grant = 0;
3425 client_obd_list_unlock(&cli->cl_loi_list_lock);
3428 case IMP_EVENT_INACTIVE: {
3429 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
3432 case IMP_EVENT_INVALIDATE: {
3433 struct ldlm_namespace *ns = obd->obd_namespace;
3437 env = cl_env_get(&refcheck);
3441 /* all pages go to failing rpcs due to the invalid
3443 osc_io_unplug(env, cli, NULL, PDL_POLICY_ROUND);
3445 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3446 cl_env_put(env, &refcheck);
3451 case IMP_EVENT_ACTIVE: {
3452 /* Only do this on the MDS OSC's */
3453 if (imp->imp_server_timeout) {
3454 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3456 cfs_spin_lock(&oscc->oscc_lock);
3457 oscc->oscc_flags &= ~(OSCC_FLAG_NOSPC |
3458 OSCC_FLAG_NOSPC_BLK);
3459 cfs_spin_unlock(&oscc->oscc_lock);
3461 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
3464 case IMP_EVENT_OCD: {
3465 struct obd_connect_data *ocd = &imp->imp_connect_data;
3467 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
3468 osc_init_grant(&obd->u.cli, ocd);
3471 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
3472 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
3474 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
3477 case IMP_EVENT_DEACTIVATE: {
3478 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DEACTIVATE, NULL);
3481 case IMP_EVENT_ACTIVATE: {
3482 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVATE, NULL);
3486 CERROR("Unknown import event %d\n", event);
3493 * Determine whether the lock can be canceled before replaying the lock
3494 * during recovery, see bug16774 for detailed information.
3496 * \retval zero the lock can't be canceled
3497 * \retval other ok to cancel
3499 static int osc_cancel_for_recovery(struct ldlm_lock *lock)
3501 check_res_locked(lock->l_resource);
3504 * Cancel all unused extent lock in granted mode LCK_PR or LCK_CR.
3506 * XXX as a future improvement, we can also cancel unused write lock
3507 * if it doesn't have dirty data and active mmaps.
3509 if (lock->l_resource->lr_type == LDLM_EXTENT &&
3510 (lock->l_granted_mode == LCK_PR ||
3511 lock->l_granted_mode == LCK_CR) &&
3512 (osc_dlm_lock_pageref(lock) == 0))
3518 static int brw_queue_work(const struct lu_env *env, void *data)
3520 struct client_obd *cli = data;
3522 CDEBUG(D_CACHE, "Run writeback work for client obd %p.\n", cli);
3524 osc_io_unplug(env, cli, NULL, PDL_POLICY_SAME);
3528 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
3530 struct client_obd *cli = &obd->u.cli;
3535 rc = ptlrpcd_addref();
3539 rc = client_obd_setup(obd, lcfg);
3542 handler = ptlrpcd_alloc_work(cli->cl_import,
3543 brw_queue_work, cli);
3544 if (!IS_ERR(handler))
3545 cli->cl_writeback_work = handler;
3547 rc = PTR_ERR(handler);
3551 struct lprocfs_static_vars lvars = { 0 };
3553 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
3554 lprocfs_osc_init_vars(&lvars);
3555 if (lprocfs_obd_setup(obd, lvars.obd_vars) == 0) {
3556 lproc_osc_attach_seqstat(obd);
3557 sptlrpc_lprocfs_cliobd_attach(obd);
3558 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 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);
3574 ns_register_cancel(obd->obd_namespace, osc_cancel_for_recovery);
3582 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
3588 case OBD_CLEANUP_EARLY: {
3589 struct obd_import *imp;
3590 imp = obd->u.cli.cl_import;
3591 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
3592 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
3593 ptlrpc_deactivate_import(imp);
3594 cfs_spin_lock(&imp->imp_lock);
3595 imp->imp_pingable = 0;
3596 cfs_spin_unlock(&imp->imp_lock);
3599 case OBD_CLEANUP_EXPORTS: {
3600 struct client_obd *cli = &obd->u.cli;
3602 * for echo client, export may be on zombie list, wait for
3603 * zombie thread to cull it, because cli.cl_import will be
3604 * cleared in client_disconnect_export():
3605 * class_export_destroy() -> obd_cleanup() ->
3606 * echo_device_free() -> echo_client_cleanup() ->
3607 * obd_disconnect() -> osc_disconnect() ->
3608 * client_disconnect_export()
3610 obd_zombie_barrier();
3611 if (cli->cl_writeback_work) {
3612 ptlrpcd_destroy_work(cli->cl_writeback_work);
3613 cli->cl_writeback_work = NULL;
3615 obd_cleanup_client_import(obd);
3616 ptlrpc_lprocfs_unregister_obd(obd);
3617 lprocfs_obd_cleanup(obd);
3618 rc = obd_llog_finish(obd, 0);
3620 CERROR("failed to cleanup llogging subsystems\n");
3627 int osc_cleanup(struct obd_device *obd)
3633 /* free memory of osc quota cache */
3634 osc_quota_cleanup(obd);
3636 rc = client_obd_cleanup(obd);
3642 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
3644 struct lprocfs_static_vars lvars = { 0 };
3647 lprocfs_osc_init_vars(&lvars);
3649 switch (lcfg->lcfg_command) {
3651 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars,
3661 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
3663 return osc_process_config_base(obd, buf);
3666 struct obd_ops osc_obd_ops = {
3667 .o_owner = THIS_MODULE,
3668 .o_setup = osc_setup,
3669 .o_precleanup = osc_precleanup,
3670 .o_cleanup = osc_cleanup,
3671 .o_add_conn = client_import_add_conn,
3672 .o_del_conn = client_import_del_conn,
3673 .o_connect = client_connect_import,
3674 .o_reconnect = osc_reconnect,
3675 .o_disconnect = osc_disconnect,
3676 .o_statfs = osc_statfs,
3677 .o_statfs_async = osc_statfs_async,
3678 .o_packmd = osc_packmd,
3679 .o_unpackmd = osc_unpackmd,
3680 .o_precreate = osc_precreate,
3681 .o_create = osc_create,
3682 .o_create_async = osc_create_async,
3683 .o_destroy = osc_destroy,
3684 .o_getattr = osc_getattr,
3685 .o_getattr_async = osc_getattr_async,
3686 .o_setattr = osc_setattr,
3687 .o_setattr_async = osc_setattr_async,
3689 .o_punch = osc_punch,
3691 .o_enqueue = osc_enqueue,
3692 .o_change_cbdata = osc_change_cbdata,
3693 .o_find_cbdata = osc_find_cbdata,
3694 .o_cancel = osc_cancel,
3695 .o_cancel_unused = osc_cancel_unused,
3696 .o_iocontrol = osc_iocontrol,
3697 .o_get_info = osc_get_info,
3698 .o_set_info_async = osc_set_info_async,
3699 .o_import_event = osc_import_event,
3700 .o_llog_init = osc_llog_init,
3701 .o_llog_finish = osc_llog_finish,
3702 .o_process_config = osc_process_config,
3703 .o_quotactl = osc_quotactl,
3704 .o_quotacheck = osc_quotacheck,
3705 .o_quota_adjust_qunit = osc_quota_adjust_qunit,
3708 extern struct lu_kmem_descr osc_caches[];
3709 extern cfs_spinlock_t osc_ast_guard;
3710 extern cfs_lock_class_key_t osc_ast_guard_class;
3712 int __init osc_init(void)
3714 struct lprocfs_static_vars lvars = { 0 };
3718 /* print an address of _any_ initialized kernel symbol from this
3719 * module, to allow debugging with gdb that doesn't support data
3720 * symbols from modules.*/
3721 CDEBUG(D_INFO, "Lustre OSC module (%p).\n", &osc_caches);
3723 rc = lu_kmem_init(osc_caches);
3725 lprocfs_osc_init_vars(&lvars);
3728 rc = class_register_type(&osc_obd_ops, NULL, lvars.module_vars,
3729 LUSTRE_OSC_NAME, &osc_device_type);
3731 lu_kmem_fini(osc_caches);
3735 cfs_spin_lock_init(&osc_ast_guard);
3736 cfs_lockdep_set_class(&osc_ast_guard, &osc_ast_guard_class);
3738 osc_mds_ost_orig_logops = llog_lvfs_ops;
3739 osc_mds_ost_orig_logops.lop_setup = llog_obd_origin_setup;
3740 osc_mds_ost_orig_logops.lop_cleanup = llog_obd_origin_cleanup;
3741 osc_mds_ost_orig_logops.lop_add = llog_obd_origin_add;
3742 osc_mds_ost_orig_logops.lop_connect = llog_origin_connect;
3748 static void /*__exit*/ osc_exit(void)
3751 class_unregister_type(LUSTRE_OSC_NAME);
3752 lu_kmem_fini(osc_caches);
3755 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
3756 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
3757 MODULE_LICENSE("GPL");
3759 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);
git://git.whamcloud.com / fs / lustre-release.git / blob