1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
38 # define EXPORT_SYMTAB
40 #define DEBUG_SUBSYSTEM S_OSC
42 #include <libcfs/libcfs.h>
45 # include <liblustre.h>
48 #include <lustre_dlm.h>
49 #include <lustre_net.h>
50 #include <lustre/lustre_user.h>
51 #include <obd_cksum.h>
59 #include <lustre_ha.h>
60 #include <lprocfs_status.h>
61 #include <lustre_log.h>
62 #include <lustre_debug.h>
63 #include <lustre_param.h>
64 #include "osc_internal.h"
66 static quota_interface_t *quota_interface = NULL;
67 extern quota_interface_t osc_quota_interface;
69 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
70 static int brw_interpret(const struct lu_env *env,
71 struct ptlrpc_request *req, void *data, int rc);
72 int osc_cleanup(struct obd_device *obd);
74 /* Pack OSC object metadata for disk storage (LE byte order). */
75 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
76 struct lov_stripe_md *lsm)
81 lmm_size = sizeof(**lmmp);
86 OBD_FREE(*lmmp, lmm_size);
92 OBD_ALLOC(*lmmp, lmm_size);
98 LASSERT(lsm->lsm_object_id);
99 LASSERT_MDS_GROUP(lsm->lsm_object_gr);
100 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
101 (*lmmp)->lmm_object_gr = cpu_to_le64(lsm->lsm_object_gr);
107 /* Unpack OSC object metadata from disk storage (LE byte order). */
108 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
109 struct lov_mds_md *lmm, int lmm_bytes)
115 if (lmm_bytes < sizeof (*lmm)) {
116 CERROR("lov_mds_md too small: %d, need %d\n",
117 lmm_bytes, (int)sizeof(*lmm));
120 /* XXX LOV_MAGIC etc check? */
122 if (lmm->lmm_object_id == 0) {
123 CERROR("lov_mds_md: zero lmm_object_id\n");
128 lsm_size = lov_stripe_md_size(1);
132 if (*lsmp != NULL && lmm == NULL) {
133 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
134 OBD_FREE(*lsmp, lsm_size);
140 OBD_ALLOC(*lsmp, lsm_size);
143 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
144 if ((*lsmp)->lsm_oinfo[0] == NULL) {
145 OBD_FREE(*lsmp, lsm_size);
148 loi_init((*lsmp)->lsm_oinfo[0]);
152 /* XXX zero *lsmp? */
153 (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
154 (*lsmp)->lsm_object_gr = le64_to_cpu (lmm->lmm_object_gr);
155 LASSERT((*lsmp)->lsm_object_id);
156 LASSERT_MDS_GROUP((*lsmp)->lsm_object_gr);
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 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 = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
214 lustre_swab_ost_body);
216 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
217 memcpy(aa->aa_oi->oi_oa, &body->oa, sizeof(*aa->aa_oi->oi_oa));
219 /* This should really be sent by the OST */
220 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
221 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
223 CDEBUG(D_INFO, "can't unpack ost_body\n");
225 aa->aa_oi->oi_oa->o_valid = 0;
228 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
232 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
233 struct ptlrpc_request_set *set)
235 struct ptlrpc_request *req;
236 struct osc_async_args *aa;
240 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
244 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
245 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
247 ptlrpc_request_free(req);
251 osc_pack_req_body(req, oinfo);
253 ptlrpc_request_set_replen(req);
254 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
256 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
257 aa = ptlrpc_req_async_args(req);
260 ptlrpc_set_add_req(set, req);
264 static int osc_getattr(struct obd_export *exp, 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 *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(struct obd_export *exp, struct obd_info *oinfo,
308 struct obd_trans_info *oti)
310 struct ptlrpc_request *req;
311 struct ost_body *body;
315 LASSERTF(!(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP) ||
316 CHECK_MDS_GROUP(oinfo->oi_oa->o_gr),
317 "oinfo->oi_oa->o_valid="LPU64" oinfo->oi_oa->o_gr="LPU64"\n",
318 oinfo->oi_oa->o_valid, oinfo->oi_oa->o_gr);
320 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
324 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
325 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
327 ptlrpc_request_free(req);
331 osc_pack_req_body(req, oinfo);
333 ptlrpc_request_set_replen(req);
335 rc = ptlrpc_queue_wait(req);
339 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
341 GOTO(out, rc = -EPROTO);
343 *oinfo->oi_oa = body->oa;
347 ptlrpc_req_finished(req);
351 static int osc_setattr_interpret(const struct lu_env *env,
352 struct ptlrpc_request *req,
353 struct osc_async_args *aa, int rc)
355 struct ost_body *body;
361 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
363 GOTO(out, rc = -EPROTO);
365 *aa->aa_oi->oi_oa = body->oa;
367 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
371 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
372 struct obd_trans_info *oti,
373 struct ptlrpc_request_set *rqset)
375 struct ptlrpc_request *req;
376 struct osc_async_args *aa;
380 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
384 osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
385 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
387 ptlrpc_request_free(req);
391 osc_pack_req_body(req, oinfo);
393 ptlrpc_request_set_replen(req);
395 if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) {
397 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
400 /* do mds to ost setattr asynchronously */
402 /* Do not wait for response. */
403 ptlrpcd_add_req(req, PSCOPE_OTHER);
405 req->rq_interpret_reply =
406 (ptlrpc_interpterer_t)osc_setattr_interpret;
408 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
409 aa = ptlrpc_req_async_args(req);
412 ptlrpc_set_add_req(rqset, req);
418 int osc_real_create(struct obd_export *exp, struct obdo *oa,
419 struct lov_stripe_md **ea, struct obd_trans_info *oti)
421 struct ptlrpc_request *req;
422 struct ost_body *body;
423 struct lov_stripe_md *lsm;
432 rc = obd_alloc_memmd(exp, &lsm);
437 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
439 GOTO(out, rc = -ENOMEM);
441 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
443 ptlrpc_request_free(req);
447 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
451 ptlrpc_request_set_replen(req);
453 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
454 oa->o_flags == OBD_FL_DELORPHAN) {
456 "delorphan from OST integration");
457 /* Don't resend the delorphan req */
458 req->rq_no_resend = req->rq_no_delay = 1;
461 rc = ptlrpc_queue_wait(req);
465 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
467 GOTO(out_req, rc = -EPROTO);
471 /* This should really be sent by the OST */
472 oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
473 oa->o_valid |= OBD_MD_FLBLKSZ;
475 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
476 * have valid lsm_oinfo data structs, so don't go touching that.
477 * This needs to be fixed in a big way.
479 lsm->lsm_object_id = oa->o_id;
480 lsm->lsm_object_gr = oa->o_gr;
484 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
486 if (oa->o_valid & OBD_MD_FLCOOKIE) {
487 if (!oti->oti_logcookies)
488 oti_alloc_cookies(oti, 1);
489 *oti->oti_logcookies = oa->o_lcookie;
493 CDEBUG(D_HA, "transno: "LPD64"\n",
494 lustre_msg_get_transno(req->rq_repmsg));
496 ptlrpc_req_finished(req);
499 obd_free_memmd(exp, &lsm);
503 static int osc_punch_interpret(const struct lu_env *env,
504 struct ptlrpc_request *req,
505 struct osc_punch_args *aa, int rc)
507 struct ost_body *body;
513 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
515 GOTO(out, rc = -EPROTO);
517 *aa->pa_oa = body->oa;
519 rc = aa->pa_upcall(aa->pa_cookie, rc);
523 int osc_punch_base(struct obd_export *exp, struct obdo *oa,
524 struct obd_capa *capa,
525 obd_enqueue_update_f upcall, void *cookie,
526 struct ptlrpc_request_set *rqset)
528 struct ptlrpc_request *req;
529 struct osc_punch_args *aa;
530 struct ost_body *body;
534 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
538 osc_set_capa_size(req, &RMF_CAPA1, capa);
539 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
541 ptlrpc_request_free(req);
544 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
545 ptlrpc_at_set_req_timeout(req);
547 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
550 osc_pack_capa(req, body, capa);
552 ptlrpc_request_set_replen(req);
555 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_punch_interpret;
556 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
557 aa = ptlrpc_req_async_args(req);
559 aa->pa_upcall = upcall;
560 aa->pa_cookie = cookie;
561 if (rqset == PTLRPCD_SET)
562 ptlrpcd_add_req(req, PSCOPE_OTHER);
564 ptlrpc_set_add_req(rqset, req);
569 static int osc_punch(struct obd_export *exp, struct obd_info *oinfo,
570 struct obd_trans_info *oti,
571 struct ptlrpc_request_set *rqset)
573 oinfo->oi_oa->o_size = oinfo->oi_policy.l_extent.start;
574 oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
575 oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
576 return osc_punch_base(exp, oinfo->oi_oa, oinfo->oi_capa,
577 oinfo->oi_cb_up, oinfo, rqset);
580 static int osc_sync(struct obd_export *exp, struct obdo *oa,
581 struct lov_stripe_md *md, obd_size start, obd_size end,
584 struct ptlrpc_request *req;
585 struct ost_body *body;
590 CDEBUG(D_INFO, "oa NULL\n");
594 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
598 osc_set_capa_size(req, &RMF_CAPA1, capa);
599 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
601 ptlrpc_request_free(req);
605 /* overload the size and blocks fields in the oa with start/end */
606 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
609 body->oa.o_size = start;
610 body->oa.o_blocks = end;
611 body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
612 osc_pack_capa(req, body, capa);
614 ptlrpc_request_set_replen(req);
616 rc = ptlrpc_queue_wait(req);
620 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
622 GOTO(out, rc = -EPROTO);
628 ptlrpc_req_finished(req);
632 /* Find and cancel locally locks matched by @mode in the resource found by
633 * @objid. Found locks are added into @cancel list. Returns the amount of
634 * locks added to @cancels list. */
635 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
636 struct list_head *cancels, ldlm_mode_t mode,
639 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
640 struct ldlm_res_id res_id;
641 struct ldlm_resource *res;
645 osc_build_res_name(oa->o_id, oa->o_gr, &res_id);
646 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
650 LDLM_RESOURCE_ADDREF(res);
651 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
652 lock_flags, 0, NULL);
653 LDLM_RESOURCE_DELREF(res);
654 ldlm_resource_putref(res);
658 static int osc_destroy_interpret(const struct lu_env *env,
659 struct ptlrpc_request *req, void *data,
662 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
664 atomic_dec(&cli->cl_destroy_in_flight);
665 cfs_waitq_signal(&cli->cl_destroy_waitq);
669 static int osc_can_send_destroy(struct client_obd *cli)
671 if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
672 cli->cl_max_rpcs_in_flight) {
673 /* The destroy request can be sent */
676 if (atomic_dec_return(&cli->cl_destroy_in_flight) <
677 cli->cl_max_rpcs_in_flight) {
679 * The counter has been modified between the two atomic
682 cfs_waitq_signal(&cli->cl_destroy_waitq);
687 /* Destroy requests can be async always on the client, and we don't even really
688 * care about the return code since the client cannot do anything at all about
690 * When the MDS is unlinking a filename, it saves the file objects into a
691 * recovery llog, and these object records are cancelled when the OST reports
692 * they were destroyed and sync'd to disk (i.e. transaction committed).
693 * If the client dies, or the OST is down when the object should be destroyed,
694 * the records are not cancelled, and when the OST reconnects to the MDS next,
695 * it will retrieve the llog unlink logs and then sends the log cancellation
696 * cookies to the MDS after committing destroy transactions. */
697 static int osc_destroy(struct obd_export *exp, struct obdo *oa,
698 struct lov_stripe_md *ea, struct obd_trans_info *oti,
699 struct obd_export *md_export, void *capa)
701 struct client_obd *cli = &exp->exp_obd->u.cli;
702 struct ptlrpc_request *req;
703 struct ost_body *body;
704 CFS_LIST_HEAD(cancels);
709 CDEBUG(D_INFO, "oa NULL\n");
713 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
714 LDLM_FL_DISCARD_DATA);
716 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
718 ldlm_lock_list_put(&cancels, l_bl_ast, count);
722 osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
723 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
726 ptlrpc_request_free(req);
730 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
731 ptlrpc_at_set_req_timeout(req);
733 if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
734 oa->o_lcookie = *oti->oti_logcookies;
735 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
739 osc_pack_capa(req, body, (struct obd_capa *)capa);
740 ptlrpc_request_set_replen(req);
742 /* don't throttle destroy RPCs for the MDT */
743 if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
744 req->rq_interpret_reply = osc_destroy_interpret;
745 if (!osc_can_send_destroy(cli)) {
746 struct l_wait_info lwi = { 0 };
749 * Wait until the number of on-going destroy RPCs drops
750 * under max_rpc_in_flight
752 l_wait_event_exclusive(cli->cl_destroy_waitq,
753 osc_can_send_destroy(cli), &lwi);
757 /* Do not wait for response */
758 ptlrpcd_add_req(req, PSCOPE_OTHER);
762 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
765 obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
767 LASSERT(!(oa->o_valid & bits));
770 client_obd_list_lock(&cli->cl_loi_list_lock);
771 oa->o_dirty = cli->cl_dirty;
772 if (cli->cl_dirty - cli->cl_dirty_transit > cli->cl_dirty_max) {
773 CERROR("dirty %lu - %lu > dirty_max %lu\n",
774 cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
776 } else if (atomic_read(&obd_dirty_pages) -
777 atomic_read(&obd_dirty_transit_pages) > obd_max_dirty_pages){
778 CERROR("dirty %d - %d > system dirty_max %d\n",
779 atomic_read(&obd_dirty_pages),
780 atomic_read(&obd_dirty_transit_pages),
781 obd_max_dirty_pages);
783 } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
784 CERROR("dirty %lu - dirty_max %lu too big???\n",
785 cli->cl_dirty, cli->cl_dirty_max);
788 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
789 (cli->cl_max_rpcs_in_flight + 1);
790 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
792 oa->o_grant = cli->cl_avail_grant;
793 oa->o_dropped = cli->cl_lost_grant;
794 cli->cl_lost_grant = 0;
795 client_obd_list_unlock(&cli->cl_loi_list_lock);
796 CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
797 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
801 static void osc_update_next_shrink(struct client_obd *cli)
803 int time = GRANT_SHRINK_INTERVAL;
804 cli->cl_next_shrink_grant = cfs_time_shift(time);
805 CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
806 cli->cl_next_shrink_grant);
809 /* caller must hold loi_list_lock */
810 static void osc_consume_write_grant(struct client_obd *cli,
811 struct brw_page *pga)
813 LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
814 atomic_inc(&obd_dirty_pages);
815 cli->cl_dirty += CFS_PAGE_SIZE;
816 cli->cl_avail_grant -= CFS_PAGE_SIZE;
817 pga->flag |= OBD_BRW_FROM_GRANT;
818 CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
819 CFS_PAGE_SIZE, pga, pga->pg);
820 LASSERT(cli->cl_avail_grant >= 0);
821 osc_update_next_shrink(cli);
824 /* the companion to osc_consume_write_grant, called when a brw has completed.
825 * must be called with the loi lock held. */
826 static void osc_release_write_grant(struct client_obd *cli,
827 struct brw_page *pga, int sent)
829 int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
832 if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
837 pga->flag &= ~OBD_BRW_FROM_GRANT;
838 atomic_dec(&obd_dirty_pages);
839 cli->cl_dirty -= CFS_PAGE_SIZE;
840 if (pga->flag & OBD_BRW_NOCACHE) {
841 pga->flag &= ~OBD_BRW_NOCACHE;
842 atomic_dec(&obd_dirty_transit_pages);
843 cli->cl_dirty_transit -= CFS_PAGE_SIZE;
846 cli->cl_lost_grant += CFS_PAGE_SIZE;
847 CDEBUG(D_CACHE, "lost grant: %lu avail grant: %lu dirty: %lu\n",
848 cli->cl_lost_grant, cli->cl_avail_grant, cli->cl_dirty);
849 } else if (CFS_PAGE_SIZE != blocksize && pga->count != CFS_PAGE_SIZE) {
850 /* For short writes we shouldn't count parts of pages that
851 * span a whole block on the OST side, or our accounting goes
852 * wrong. Should match the code in filter_grant_check. */
853 int offset = pga->off & ~CFS_PAGE_MASK;
854 int count = pga->count + (offset & (blocksize - 1));
855 int end = (offset + pga->count) & (blocksize - 1);
857 count += blocksize - end;
859 cli->cl_lost_grant += CFS_PAGE_SIZE - count;
860 CDEBUG(D_CACHE, "lost %lu grant: %lu avail: %lu dirty: %lu\n",
861 CFS_PAGE_SIZE - count, cli->cl_lost_grant,
862 cli->cl_avail_grant, cli->cl_dirty);
868 static unsigned long rpcs_in_flight(struct client_obd *cli)
870 return cli->cl_r_in_flight + cli->cl_w_in_flight;
873 /* caller must hold loi_list_lock */
874 void osc_wake_cache_waiters(struct client_obd *cli)
876 struct list_head *l, *tmp;
877 struct osc_cache_waiter *ocw;
880 list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
881 /* if we can't dirty more, we must wait until some is written */
882 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
883 (atomic_read(&obd_dirty_pages) + 1 > obd_max_dirty_pages)) {
884 CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
885 "osc max %ld, sys max %d\n", cli->cl_dirty,
886 cli->cl_dirty_max, obd_max_dirty_pages);
890 /* if still dirty cache but no grant wait for pending RPCs that
891 * may yet return us some grant before doing sync writes */
892 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
893 CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
894 cli->cl_w_in_flight);
898 ocw = list_entry(l, struct osc_cache_waiter, ocw_entry);
899 list_del_init(&ocw->ocw_entry);
900 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
901 /* no more RPCs in flight to return grant, do sync IO */
902 ocw->ocw_rc = -EDQUOT;
903 CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
905 osc_consume_write_grant(cli,
906 &ocw->ocw_oap->oap_brw_page);
909 cfs_waitq_signal(&ocw->ocw_waitq);
915 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
917 client_obd_list_lock(&cli->cl_loi_list_lock);
918 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
919 if (body->oa.o_valid & OBD_MD_FLGRANT)
920 cli->cl_avail_grant += body->oa.o_grant;
921 /* waiters are woken in brw_interpret */
922 client_obd_list_unlock(&cli->cl_loi_list_lock);
925 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
926 void *key, obd_count vallen, void *val,
927 struct ptlrpc_request_set *set);
929 static int osc_shrink_grant_interpret(const struct lu_env *env,
930 struct ptlrpc_request *req,
933 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
934 struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
935 struct ost_body *body;
938 client_obd_list_lock(&cli->cl_loi_list_lock);
939 cli->cl_avail_grant += oa->o_grant;
940 client_obd_list_unlock(&cli->cl_loi_list_lock);
944 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
946 osc_update_grant(cli, body);
952 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
954 client_obd_list_lock(&cli->cl_loi_list_lock);
955 oa->o_grant = cli->cl_avail_grant / 4;
956 cli->cl_avail_grant -= oa->o_grant;
957 client_obd_list_unlock(&cli->cl_loi_list_lock);
958 oa->o_flags |= OBD_FL_SHRINK_GRANT;
959 osc_update_next_shrink(cli);
962 static int osc_shrink_grant(struct client_obd *cli)
965 struct ost_body *body;
972 osc_announce_cached(cli, &body->oa, 0);
973 osc_shrink_grant_local(cli, &body->oa);
974 rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
975 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
976 sizeof(*body), body, NULL);
978 client_obd_list_lock(&cli->cl_loi_list_lock);
979 cli->cl_avail_grant += body->oa.o_grant;
980 client_obd_list_unlock(&cli->cl_loi_list_lock);
987 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
988 static int osc_should_shrink_grant(struct client_obd *client)
990 cfs_time_t time = cfs_time_current();
991 cfs_time_t next_shrink = client->cl_next_shrink_grant;
992 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
993 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
994 client->cl_avail_grant > GRANT_SHRINK_LIMIT)
997 osc_update_next_shrink(client);
1002 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1004 struct client_obd *client;
1006 list_for_each_entry(client, &item->ti_obd_list, cl_grant_shrink_list) {
1007 if (osc_should_shrink_grant(client))
1008 osc_shrink_grant(client);
1013 static int osc_add_shrink_grant(struct client_obd *client)
1017 rc = ptlrpc_add_timeout_client(GRANT_SHRINK_INTERVAL,
1019 osc_grant_shrink_grant_cb, NULL,
1020 &client->cl_grant_shrink_list);
1022 CERROR("add grant client %s error %d\n",
1023 client->cl_import->imp_obd->obd_name, rc);
1026 CDEBUG(D_CACHE, "add grant client %s \n",
1027 client->cl_import->imp_obd->obd_name);
1028 osc_update_next_shrink(client);
1032 static int osc_del_shrink_grant(struct client_obd *client)
1034 CDEBUG(D_CACHE, "del grant client %s \n",
1035 client->cl_import->imp_obd->obd_name);
1036 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list);
1039 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1041 client_obd_list_lock(&cli->cl_loi_list_lock);
1042 cli->cl_avail_grant = ocd->ocd_grant;
1043 client_obd_list_unlock(&cli->cl_loi_list_lock);
1045 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1046 list_empty(&cli->cl_grant_shrink_list))
1047 osc_add_shrink_grant(cli);
1049 CDEBUG(D_CACHE, "setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
1050 cli->cl_avail_grant, cli->cl_lost_grant);
1051 LASSERT(cli->cl_avail_grant >= 0);
1054 /* We assume that the reason this OSC got a short read is because it read
1055 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1056 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1057 * this stripe never got written at or beyond this stripe offset yet. */
1058 static void handle_short_read(int nob_read, obd_count page_count,
1059 struct brw_page **pga)
1064 /* skip bytes read OK */
1065 while (nob_read > 0) {
1066 LASSERT (page_count > 0);
1068 if (pga[i]->count > nob_read) {
1069 /* EOF inside this page */
1070 ptr = cfs_kmap(pga[i]->pg) +
1071 (pga[i]->off & ~CFS_PAGE_MASK);
1072 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1073 cfs_kunmap(pga[i]->pg);
1079 nob_read -= pga[i]->count;
1084 /* zero remaining pages */
1085 while (page_count-- > 0) {
1086 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1087 memset(ptr, 0, pga[i]->count);
1088 cfs_kunmap(pga[i]->pg);
1093 static int check_write_rcs(struct ptlrpc_request *req,
1094 int requested_nob, int niocount,
1095 obd_count page_count, struct brw_page **pga)
1099 /* return error if any niobuf was in error */
1100 remote_rcs = lustre_swab_repbuf(req, REQ_REC_OFF + 1,
1101 sizeof(*remote_rcs) * niocount, NULL);
1102 if (remote_rcs == NULL) {
1103 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1106 if (lustre_msg_swabbed(req->rq_repmsg))
1107 for (i = 0; i < niocount; i++)
1108 __swab32s(&remote_rcs[i]);
1110 for (i = 0; i < niocount; i++) {
1111 if (remote_rcs[i] < 0)
1112 return(remote_rcs[i]);
1114 if (remote_rcs[i] != 0) {
1115 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1116 i, remote_rcs[i], req);
1121 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1122 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1123 req->rq_bulk->bd_nob_transferred, requested_nob);
1130 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1132 if (p1->flag != p2->flag) {
1133 unsigned mask = ~(OBD_BRW_FROM_GRANT|
1134 OBD_BRW_NOCACHE|OBD_BRW_SYNC);
1136 /* warn if we try to combine flags that we don't know to be
1137 * safe to combine */
1138 if ((p1->flag & mask) != (p2->flag & mask))
1139 CERROR("is it ok to have flags 0x%x and 0x%x in the "
1140 "same brw?\n", p1->flag, p2->flag);
1144 return (p1->off + p1->count == p2->off);
1147 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1148 struct brw_page **pga, int opc,
1149 cksum_type_t cksum_type)
1154 LASSERT (pg_count > 0);
1155 cksum = init_checksum(cksum_type);
1156 while (nob > 0 && pg_count > 0) {
1157 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1158 int off = pga[i]->off & ~CFS_PAGE_MASK;
1159 int count = pga[i]->count > nob ? nob : pga[i]->count;
1161 /* corrupt the data before we compute the checksum, to
1162 * simulate an OST->client data error */
1163 if (i == 0 && opc == OST_READ &&
1164 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1165 memcpy(ptr + off, "bad1", min(4, nob));
1166 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1167 cfs_kunmap(pga[i]->pg);
1168 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1171 nob -= pga[i]->count;
1175 /* For sending we only compute the wrong checksum instead
1176 * of corrupting the data so it is still correct on a redo */
1177 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1183 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1184 struct lov_stripe_md *lsm, obd_count page_count,
1185 struct brw_page **pga,
1186 struct ptlrpc_request **reqp,
1187 struct obd_capa *ocapa, int reserve)
1189 struct ptlrpc_request *req;
1190 struct ptlrpc_bulk_desc *desc;
1191 struct ost_body *body;
1192 struct obd_ioobj *ioobj;
1193 struct niobuf_remote *niobuf;
1194 int niocount, i, requested_nob, opc, rc;
1195 struct osc_brw_async_args *aa;
1196 struct req_capsule *pill;
1197 struct brw_page *pg_prev;
1200 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1201 RETURN(-ENOMEM); /* Recoverable */
1202 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1203 RETURN(-EINVAL); /* Fatal */
1205 if ((cmd & OBD_BRW_WRITE) != 0) {
1207 req = ptlrpc_request_alloc_pool(cli->cl_import,
1208 cli->cl_import->imp_rq_pool,
1212 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW);
1217 for (niocount = i = 1; i < page_count; i++) {
1218 if (!can_merge_pages(pga[i - 1], pga[i]))
1222 pill = &req->rq_pill;
1223 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1224 niocount * sizeof(*niobuf));
1225 osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1227 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1229 ptlrpc_request_free(req);
1232 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1233 ptlrpc_at_set_req_timeout(req);
1235 if (opc == OST_WRITE)
1236 desc = ptlrpc_prep_bulk_imp(req, page_count,
1237 BULK_GET_SOURCE, OST_BULK_PORTAL);
1239 desc = ptlrpc_prep_bulk_imp(req, page_count,
1240 BULK_PUT_SINK, OST_BULK_PORTAL);
1243 GOTO(out, rc = -ENOMEM);
1244 /* NB request now owns desc and will free it when it gets freed */
1246 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1247 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1248 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1249 LASSERT(body && ioobj && niobuf);
1253 obdo_to_ioobj(oa, ioobj);
1254 ioobj->ioo_bufcnt = niocount;
1255 osc_pack_capa(req, body, ocapa);
1256 LASSERT (page_count > 0);
1258 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1259 struct brw_page *pg = pga[i];
1261 LASSERT(pg->count > 0);
1262 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1263 "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1264 pg->off, pg->count);
1266 LASSERTF(i == 0 || pg->off > pg_prev->off,
1267 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1268 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1270 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1271 pg_prev->pg, page_private(pg_prev->pg),
1272 pg_prev->pg->index, pg_prev->off);
1274 LASSERTF(i == 0 || pg->off > pg_prev->off,
1275 "i %d p_c %u\n", i, page_count);
1277 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1278 (pg->flag & OBD_BRW_SRVLOCK));
1280 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1282 requested_nob += pg->count;
1284 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1286 niobuf->len += pg->count;
1288 niobuf->offset = pg->off;
1289 niobuf->len = pg->count;
1290 niobuf->flags = pg->flag;
1295 LASSERTF((void *)(niobuf - niocount) ==
1296 lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF + 2,
1297 niocount * sizeof(*niobuf)),
1298 "want %p - real %p\n", lustre_msg_buf(req->rq_reqmsg,
1299 REQ_REC_OFF + 2, niocount * sizeof(*niobuf)),
1300 (void *)(niobuf - niocount));
1302 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1303 if (osc_should_shrink_grant(cli))
1304 osc_shrink_grant_local(cli, &body->oa);
1306 /* size[REQ_REC_OFF] still sizeof (*body) */
1307 if (opc == OST_WRITE) {
1308 if (unlikely(cli->cl_checksum) &&
1309 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1310 /* store cl_cksum_type in a local variable since
1311 * it can be changed via lprocfs */
1312 cksum_type_t cksum_type = cli->cl_cksum_type;
1314 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1315 oa->o_flags = body->oa.o_flags = 0;
1316 body->oa.o_flags |= cksum_type_pack(cksum_type);
1317 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1318 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1322 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1324 /* save this in 'oa', too, for later checking */
1325 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1326 oa->o_flags |= cksum_type_pack(cksum_type);
1328 /* clear out the checksum flag, in case this is a
1329 * resend but cl_checksum is no longer set. b=11238 */
1330 oa->o_valid &= ~OBD_MD_FLCKSUM;
1332 oa->o_cksum = body->oa.o_cksum;
1333 /* 1 RC per niobuf */
1334 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_SERVER,
1335 sizeof(__u32) * niocount);
1337 if (unlikely(cli->cl_checksum) &&
1338 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1339 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1340 body->oa.o_flags = 0;
1341 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1342 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1344 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_SERVER, 0);
1345 /* 1 RC for the whole I/O */
1347 ptlrpc_request_set_replen(req);
1349 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1350 aa = ptlrpc_req_async_args(req);
1352 aa->aa_requested_nob = requested_nob;
1353 aa->aa_nio_count = niocount;
1354 aa->aa_page_count = page_count;
1358 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1359 if (ocapa && reserve)
1360 aa->aa_ocapa = capa_get(ocapa);
1366 ptlrpc_req_finished(req);
1370 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1371 __u32 client_cksum, __u32 server_cksum, int nob,
1372 obd_count page_count, struct brw_page **pga,
1373 cksum_type_t client_cksum_type)
1377 cksum_type_t cksum_type;
1379 if (server_cksum == client_cksum) {
1380 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1384 if (oa->o_valid & OBD_MD_FLFLAGS)
1385 cksum_type = cksum_type_unpack(oa->o_flags);
1387 cksum_type = OBD_CKSUM_CRC32;
1389 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1392 if (cksum_type != client_cksum_type)
1393 msg = "the server did not use the checksum type specified in "
1394 "the original request - likely a protocol problem";
1395 else if (new_cksum == server_cksum)
1396 msg = "changed on the client after we checksummed it - "
1397 "likely false positive due to mmap IO (bug 11742)";
1398 else if (new_cksum == client_cksum)
1399 msg = "changed in transit before arrival at OST";
1401 msg = "changed in transit AND doesn't match the original - "
1402 "likely false positive due to mmap IO (bug 11742)";
1404 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inum "
1405 LPU64"/"LPU64" object "LPU64"/"LPU64" extent "
1406 "["LPU64"-"LPU64"]\n",
1407 msg, libcfs_nid2str(peer->nid),
1408 oa->o_valid & OBD_MD_FLFID ? oa->o_fid : (__u64)0,
1409 oa->o_valid & OBD_MD_FLFID ? oa->o_generation :
1412 oa->o_valid & OBD_MD_FLGROUP ? oa->o_gr : (__u64)0,
1414 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1415 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1416 "client csum now %x\n", client_cksum, client_cksum_type,
1417 server_cksum, cksum_type, new_cksum);
1421 /* Note rc enters this function as number of bytes transferred */
1422 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1424 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1425 const lnet_process_id_t *peer =
1426 &req->rq_import->imp_connection->c_peer;
1427 struct client_obd *cli = aa->aa_cli;
1428 struct ost_body *body;
1429 __u32 client_cksum = 0;
1432 if (rc < 0 && rc != -EDQUOT)
1435 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1436 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
1437 lustre_swab_ost_body);
1439 CDEBUG(D_INFO, "Can't unpack body\n");
1443 /* set/clear over quota flag for a uid/gid */
1444 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1445 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA))
1446 lquota_setdq(quota_interface, cli, body->oa.o_uid,
1447 body->oa.o_gid, body->oa.o_valid,
1453 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1454 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1456 osc_update_grant(cli, body);
1458 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1460 CERROR("Unexpected +ve rc %d\n", rc);
1463 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1465 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1468 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1469 check_write_checksum(&body->oa, peer, client_cksum,
1470 body->oa.o_cksum, aa->aa_requested_nob,
1471 aa->aa_page_count, aa->aa_ppga,
1472 cksum_type_unpack(aa->aa_oa->o_flags)))
1475 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1476 aa->aa_page_count, aa->aa_ppga);
1480 /* The rest of this function executes only for OST_READs */
1482 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1486 if (rc > aa->aa_requested_nob) {
1487 CERROR("Unexpected rc %d (%d requested)\n", rc,
1488 aa->aa_requested_nob);
1492 if (rc != req->rq_bulk->bd_nob_transferred) {
1493 CERROR ("Unexpected rc %d (%d transferred)\n",
1494 rc, req->rq_bulk->bd_nob_transferred);
1498 if (rc < aa->aa_requested_nob)
1499 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1501 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1502 static int cksum_counter;
1503 __u32 server_cksum = body->oa.o_cksum;
1506 cksum_type_t cksum_type;
1508 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1509 cksum_type = cksum_type_unpack(body->oa.o_flags);
1511 cksum_type = OBD_CKSUM_CRC32;
1512 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1513 aa->aa_ppga, OST_READ,
1516 if (peer->nid == req->rq_bulk->bd_sender) {
1520 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1523 if (server_cksum == ~0 && rc > 0) {
1524 CERROR("Protocol error: server %s set the 'checksum' "
1525 "bit, but didn't send a checksum. Not fatal, "
1526 "but please notify on http://bugzilla.lustre.org/\n",
1527 libcfs_nid2str(peer->nid));
1528 } else if (server_cksum != client_cksum) {
1529 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1530 "%s%s%s inum "LPU64"/"LPU64" object "
1531 LPU64"/"LPU64" extent "
1532 "["LPU64"-"LPU64"]\n",
1533 req->rq_import->imp_obd->obd_name,
1534 libcfs_nid2str(peer->nid),
1536 body->oa.o_valid & OBD_MD_FLFID ?
1537 body->oa.o_fid : (__u64)0,
1538 body->oa.o_valid & OBD_MD_FLFID ?
1539 body->oa.o_generation :(__u64)0,
1541 body->oa.o_valid & OBD_MD_FLGROUP ?
1542 body->oa.o_gr : (__u64)0,
1543 aa->aa_ppga[0]->off,
1544 aa->aa_ppga[aa->aa_page_count-1]->off +
1545 aa->aa_ppga[aa->aa_page_count-1]->count -
1547 CERROR("client %x, server %x, cksum_type %x\n",
1548 client_cksum, server_cksum, cksum_type);
1550 aa->aa_oa->o_cksum = client_cksum;
1554 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1557 } else if (unlikely(client_cksum)) {
1558 static int cksum_missed;
1561 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1562 CERROR("Checksum %u requested from %s but not sent\n",
1563 cksum_missed, libcfs_nid2str(peer->nid));
1569 *aa->aa_oa = body->oa;
1574 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1575 struct lov_stripe_md *lsm,
1576 obd_count page_count, struct brw_page **pga,
1577 struct obd_capa *ocapa)
1579 struct ptlrpc_request *req;
1583 struct l_wait_info lwi;
1587 cfs_waitq_init(&waitq);
1590 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1591 page_count, pga, &req, ocapa, 0);
1595 rc = ptlrpc_queue_wait(req);
1597 if (rc == -ETIMEDOUT && req->rq_resend) {
1598 DEBUG_REQ(D_HA, req, "BULK TIMEOUT");
1599 ptlrpc_req_finished(req);
1603 rc = osc_brw_fini_request(req, rc);
1605 ptlrpc_req_finished(req);
1606 if (osc_recoverable_error(rc)) {
1608 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1609 CERROR("too many resend retries, returning error\n");
1613 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1614 l_wait_event(waitq, 0, &lwi);
1622 int osc_brw_redo_request(struct ptlrpc_request *request,
1623 struct osc_brw_async_args *aa)
1625 struct ptlrpc_request *new_req;
1626 struct ptlrpc_request_set *set = request->rq_set;
1627 struct osc_brw_async_args *new_aa;
1628 struct osc_async_page *oap;
1632 if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1633 CERROR("too many resend retries, returning error\n");
1637 DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1639 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1640 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1641 aa->aa_cli, aa->aa_oa,
1642 NULL /* lsm unused by osc currently */,
1643 aa->aa_page_count, aa->aa_ppga,
1644 &new_req, aa->aa_ocapa, 0);
1648 client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1650 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1651 if (oap->oap_request != NULL) {
1652 LASSERTF(request == oap->oap_request,
1653 "request %p != oap_request %p\n",
1654 request, oap->oap_request);
1655 if (oap->oap_interrupted) {
1656 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1657 ptlrpc_req_finished(new_req);
1662 /* New request takes over pga and oaps from old request.
1663 * Note that copying a list_head doesn't work, need to move it... */
1665 new_req->rq_interpret_reply = request->rq_interpret_reply;
1666 new_req->rq_async_args = request->rq_async_args;
1667 new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1669 new_aa = ptlrpc_req_async_args(new_req);
1671 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1672 list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1673 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1675 list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1676 if (oap->oap_request) {
1677 ptlrpc_req_finished(oap->oap_request);
1678 oap->oap_request = ptlrpc_request_addref(new_req);
1682 new_aa->aa_ocapa = aa->aa_ocapa;
1683 aa->aa_ocapa = NULL;
1685 /* use ptlrpc_set_add_req is safe because interpret functions work
1686 * in check_set context. only one way exist with access to request
1687 * from different thread got -EINTR - this way protected with
1688 * cl_loi_list_lock */
1689 ptlrpc_set_add_req(set, new_req);
1691 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1693 DEBUG_REQ(D_INFO, new_req, "new request");
1698 * ugh, we want disk allocation on the target to happen in offset order. we'll
1699 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1700 * fine for our small page arrays and doesn't require allocation. its an
1701 * insertion sort that swaps elements that are strides apart, shrinking the
1702 * stride down until its '1' and the array is sorted.
1704 static void sort_brw_pages(struct brw_page **array, int num)
1707 struct brw_page *tmp;
1711 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1716 for (i = stride ; i < num ; i++) {
1719 while (j >= stride && array[j - stride]->off > tmp->off) {
1720 array[j] = array[j - stride];
1725 } while (stride > 1);
1728 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1734 LASSERT (pages > 0);
1735 offset = pg[i]->off & ~CFS_PAGE_MASK;
1739 if (pages == 0) /* that's all */
1742 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1743 return count; /* doesn't end on page boundary */
1746 offset = pg[i]->off & ~CFS_PAGE_MASK;
1747 if (offset != 0) /* doesn't start on page boundary */
1754 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1756 struct brw_page **ppga;
1759 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1763 for (i = 0; i < count; i++)
1768 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1770 LASSERT(ppga != NULL);
1771 OBD_FREE(ppga, sizeof(*ppga) * count);
1774 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1775 obd_count page_count, struct brw_page *pga,
1776 struct obd_trans_info *oti)
1778 struct obdo *saved_oa = NULL;
1779 struct brw_page **ppga, **orig;
1780 struct obd_import *imp = class_exp2cliimp(exp);
1781 struct client_obd *cli = &imp->imp_obd->u.cli;
1782 int rc, page_count_orig;
1785 if (cmd & OBD_BRW_CHECK) {
1786 /* The caller just wants to know if there's a chance that this
1787 * I/O can succeed */
1789 if (imp == NULL || imp->imp_invalid)
1794 /* test_brw with a failed create can trip this, maybe others. */
1795 LASSERT(cli->cl_max_pages_per_rpc);
1799 orig = ppga = osc_build_ppga(pga, page_count);
1802 page_count_orig = page_count;
1804 sort_brw_pages(ppga, page_count);
1805 while (page_count) {
1806 obd_count pages_per_brw;
1808 if (page_count > cli->cl_max_pages_per_rpc)
1809 pages_per_brw = cli->cl_max_pages_per_rpc;
1811 pages_per_brw = page_count;
1813 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1815 if (saved_oa != NULL) {
1816 /* restore previously saved oa */
1817 *oinfo->oi_oa = *saved_oa;
1818 } else if (page_count > pages_per_brw) {
1819 /* save a copy of oa (brw will clobber it) */
1820 OBDO_ALLOC(saved_oa);
1821 if (saved_oa == NULL)
1822 GOTO(out, rc = -ENOMEM);
1823 *saved_oa = *oinfo->oi_oa;
1826 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1827 pages_per_brw, ppga, oinfo->oi_capa);
1832 page_count -= pages_per_brw;
1833 ppga += pages_per_brw;
1837 osc_release_ppga(orig, page_count_orig);
1839 if (saved_oa != NULL)
1840 OBDO_FREE(saved_oa);
1845 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1846 * the dirty accounting. Writeback completes or truncate happens before
1847 * writing starts. Must be called with the loi lock held. */
1848 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1851 osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1855 /* This maintains the lists of pending pages to read/write for a given object
1856 * (lop). This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1857 * to quickly find objects that are ready to send an RPC. */
1858 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1864 if (lop->lop_num_pending == 0)
1867 /* if we have an invalid import we want to drain the queued pages
1868 * by forcing them through rpcs that immediately fail and complete
1869 * the pages. recovery relies on this to empty the queued pages
1870 * before canceling the locks and evicting down the llite pages */
1871 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1874 /* stream rpcs in queue order as long as as there is an urgent page
1875 * queued. this is our cheap solution for good batching in the case
1876 * where writepage marks some random page in the middle of the file
1877 * as urgent because of, say, memory pressure */
1878 if (!list_empty(&lop->lop_urgent)) {
1879 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1882 /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1883 optimal = cli->cl_max_pages_per_rpc;
1884 if (cmd & OBD_BRW_WRITE) {
1885 /* trigger a write rpc stream as long as there are dirtiers
1886 * waiting for space. as they're waiting, they're not going to
1887 * create more pages to coallesce with what's waiting.. */
1888 if (!list_empty(&cli->cl_cache_waiters)) {
1889 CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1892 /* +16 to avoid triggering rpcs that would want to include pages
1893 * that are being queued but which can't be made ready until
1894 * the queuer finishes with the page. this is a wart for
1895 * llite::commit_write() */
1898 if (lop->lop_num_pending >= optimal)
1904 static void on_list(struct list_head *item, struct list_head *list,
1907 if (list_empty(item) && should_be_on)
1908 list_add_tail(item, list);
1909 else if (!list_empty(item) && !should_be_on)
1910 list_del_init(item);
1913 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
1914 * can find pages to build into rpcs quickly */
1915 void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
1917 on_list(&loi->loi_cli_item, &cli->cl_loi_ready_list,
1918 lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE) ||
1919 lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ));
1921 on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
1922 loi->loi_write_lop.lop_num_pending);
1924 on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
1925 loi->loi_read_lop.lop_num_pending);
1928 static void lop_update_pending(struct client_obd *cli,
1929 struct loi_oap_pages *lop, int cmd, int delta)
1931 lop->lop_num_pending += delta;
1932 if (cmd & OBD_BRW_WRITE)
1933 cli->cl_pending_w_pages += delta;
1935 cli->cl_pending_r_pages += delta;
1939 * this is called when a sync waiter receives an interruption. Its job is to
1940 * get the caller woken as soon as possible. If its page hasn't been put in an
1941 * rpc yet it can dequeue immediately. Otherwise it has to mark the rpc as
1942 * desiring interruption which will forcefully complete the rpc once the rpc
1945 int osc_oap_interrupted(const struct lu_env *env, struct osc_async_page *oap)
1947 struct loi_oap_pages *lop;
1948 struct lov_oinfo *loi;
1952 LASSERT(!oap->oap_interrupted);
1953 oap->oap_interrupted = 1;
1955 /* ok, it's been put in an rpc. only one oap gets a request reference */
1956 if (oap->oap_request != NULL) {
1957 ptlrpc_mark_interrupted(oap->oap_request);
1958 ptlrpcd_wake(oap->oap_request);
1959 ptlrpc_req_finished(oap->oap_request);
1960 oap->oap_request = NULL;
1964 * page completion may be called only if ->cpo_prep() method was
1965 * executed by osc_io_submit(), that also adds page the to pending list
1967 if (!list_empty(&oap->oap_pending_item)) {
1968 list_del_init(&oap->oap_pending_item);
1969 list_del_init(&oap->oap_urgent_item);
1972 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
1973 &loi->loi_write_lop : &loi->loi_read_lop;
1974 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
1975 loi_list_maint(oap->oap_cli, oap->oap_loi);
1976 rc = oap->oap_caller_ops->ap_completion(env,
1977 oap->oap_caller_data,
1978 oap->oap_cmd, NULL, -EINTR);
1984 /* this is trying to propogate async writeback errors back up to the
1985 * application. As an async write fails we record the error code for later if
1986 * the app does an fsync. As long as errors persist we force future rpcs to be
1987 * sync so that the app can get a sync error and break the cycle of queueing
1988 * pages for which writeback will fail. */
1989 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
1996 ar->ar_force_sync = 1;
1997 ar->ar_min_xid = ptlrpc_sample_next_xid();
2002 if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2003 ar->ar_force_sync = 0;
2006 void osc_oap_to_pending(struct osc_async_page *oap)
2008 struct loi_oap_pages *lop;
2010 if (oap->oap_cmd & OBD_BRW_WRITE)
2011 lop = &oap->oap_loi->loi_write_lop;
2013 lop = &oap->oap_loi->loi_read_lop;
2015 if (oap->oap_async_flags & ASYNC_URGENT)
2016 list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2017 list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2018 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2021 /* this must be called holding the loi list lock to give coverage to exit_cache,
2022 * async_flag maintenance, and oap_request */
2023 static void osc_ap_completion(const struct lu_env *env,
2024 struct client_obd *cli, struct obdo *oa,
2025 struct osc_async_page *oap, int sent, int rc)
2030 if (oap->oap_request != NULL) {
2031 xid = ptlrpc_req_xid(oap->oap_request);
2032 ptlrpc_req_finished(oap->oap_request);
2033 oap->oap_request = NULL;
2036 oap->oap_async_flags = 0;
2037 oap->oap_interrupted = 0;
2039 if (oap->oap_cmd & OBD_BRW_WRITE) {
2040 osc_process_ar(&cli->cl_ar, xid, rc);
2041 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2044 if (rc == 0 && oa != NULL) {
2045 if (oa->o_valid & OBD_MD_FLBLOCKS)
2046 oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2047 if (oa->o_valid & OBD_MD_FLMTIME)
2048 oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2049 if (oa->o_valid & OBD_MD_FLATIME)
2050 oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2051 if (oa->o_valid & OBD_MD_FLCTIME)
2052 oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2055 rc = oap->oap_caller_ops->ap_completion(env, oap->oap_caller_data,
2056 oap->oap_cmd, oa, rc);
2058 /* ll_ap_completion (from llite) drops PG_locked. so, a new
2059 * I/O on the page could start, but OSC calls it under lock
2060 * and thus we can add oap back to pending safely */
2062 /* upper layer wants to leave the page on pending queue */
2063 osc_oap_to_pending(oap);
2065 osc_exit_cache(cli, oap, sent);
2069 static int brw_interpret(const struct lu_env *env,
2070 struct ptlrpc_request *req, void *data, int rc)
2072 struct osc_brw_async_args *aa = data;
2073 struct client_obd *cli;
2077 rc = osc_brw_fini_request(req, rc);
2078 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2079 if (osc_recoverable_error(rc)) {
2080 rc = osc_brw_redo_request(req, aa);
2086 capa_put(aa->aa_ocapa);
2087 aa->aa_ocapa = NULL;
2092 client_obd_list_lock(&cli->cl_loi_list_lock);
2094 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2095 * is called so we know whether to go to sync BRWs or wait for more
2096 * RPCs to complete */
2097 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2098 cli->cl_w_in_flight--;
2100 cli->cl_r_in_flight--;
2102 async = list_empty(&aa->aa_oaps);
2103 if (!async) { /* from osc_send_oap_rpc() */
2104 struct osc_async_page *oap, *tmp;
2105 /* the caller may re-use the oap after the completion call so
2106 * we need to clean it up a little */
2107 list_for_each_entry_safe(oap, tmp, &aa->aa_oaps, oap_rpc_item) {
2108 list_del_init(&oap->oap_rpc_item);
2109 osc_ap_completion(env, cli, aa->aa_oa, oap, 1, rc);
2111 OBDO_FREE(aa->aa_oa);
2112 } else { /* from async_internal() */
2114 for (i = 0; i < aa->aa_page_count; i++)
2115 osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2117 osc_wake_cache_waiters(cli);
2118 osc_check_rpcs(env, cli);
2119 client_obd_list_unlock(&cli->cl_loi_list_lock);
2121 cl_req_completion(env, aa->aa_clerq, rc);
2122 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2126 static struct ptlrpc_request *osc_build_req(const struct lu_env *env,
2127 struct client_obd *cli,
2128 struct list_head *rpc_list,
2129 int page_count, int cmd)
2131 struct ptlrpc_request *req;
2132 struct brw_page **pga = NULL;
2133 struct osc_brw_async_args *aa;
2134 struct obdo *oa = NULL;
2135 const struct obd_async_page_ops *ops = NULL;
2136 void *caller_data = NULL;
2137 struct osc_async_page *oap;
2138 struct osc_async_page *tmp;
2139 struct ost_body *body;
2140 struct cl_req *clerq = NULL;
2141 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2142 struct ldlm_lock *lock = NULL;
2143 struct cl_req_attr crattr;
2147 LASSERT(!list_empty(rpc_list));
2149 memset(&crattr, 0, sizeof crattr);
2150 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2152 GOTO(out, req = ERR_PTR(-ENOMEM));
2156 GOTO(out, req = ERR_PTR(-ENOMEM));
2159 list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2160 struct cl_page *page = osc_oap2cl_page(oap);
2162 ops = oap->oap_caller_ops;
2163 caller_data = oap->oap_caller_data;
2165 clerq = cl_req_alloc(env, page, crt,
2166 1 /* only 1-object rpcs for
2169 GOTO(out, req = (void *)clerq);
2170 lock = oap->oap_ldlm_lock;
2172 pga[i] = &oap->oap_brw_page;
2173 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2174 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2175 pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2177 cl_req_page_add(env, clerq, page);
2180 /* always get the data for the obdo for the rpc */
2181 LASSERT(ops != NULL);
2183 crattr.cra_capa = NULL;
2184 cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2186 oa->o_handle = lock->l_remote_handle;
2187 oa->o_valid |= OBD_MD_FLHANDLE;
2190 rc = cl_req_prep(env, clerq);
2192 CERROR("cl_req_prep failed: %d\n", rc);
2193 GOTO(out, req = ERR_PTR(rc));
2196 sort_brw_pages(pga, page_count);
2197 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2198 pga, &req, crattr.cra_capa, 1);
2200 CERROR("prep_req failed: %d\n", rc);
2201 GOTO(out, req = ERR_PTR(rc));
2204 /* Need to update the timestamps after the request is built in case
2205 * we race with setattr (locally or in queue at OST). If OST gets
2206 * later setattr before earlier BRW (as determined by the request xid),
2207 * the OST will not use BRW timestamps. Sadly, there is no obvious
2208 * way to do this in a single call. bug 10150 */
2209 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2210 cl_req_attr_set(env, clerq, &crattr,
2211 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2213 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2214 aa = ptlrpc_req_async_args(req);
2215 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2216 list_splice(rpc_list, &aa->aa_oaps);
2217 CFS_INIT_LIST_HEAD(rpc_list);
2218 aa->aa_clerq = clerq;
2220 capa_put(crattr.cra_capa);
2225 OBD_FREE(pga, sizeof(*pga) * page_count);
2226 /* this should happen rarely and is pretty bad, it makes the
2227 * pending list not follow the dirty order */
2228 client_obd_list_lock(&cli->cl_loi_list_lock);
2229 list_for_each_entry_safe(oap, tmp, rpc_list, oap_rpc_item) {
2230 list_del_init(&oap->oap_rpc_item);
2232 /* queued sync pages can be torn down while the pages
2233 * were between the pending list and the rpc */
2234 if (oap->oap_interrupted) {
2235 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2236 osc_ap_completion(env, cli, NULL, oap, 0,
2240 osc_ap_completion(env, cli, NULL, oap, 0, PTR_ERR(req));
2242 if (clerq && !IS_ERR(clerq))
2243 cl_req_completion(env, clerq, PTR_ERR(req));
2249 * prepare pages for ASYNC io and put pages in send queue.
2253 * \param cmd - OBD_BRW_* macroses
2254 * \param lop - pending pages
2256 * \return zero if pages successfully add to send queue.
2257 * \return not zere if error occurring.
2260 osc_send_oap_rpc(const struct lu_env *env, struct client_obd *cli,
2261 struct lov_oinfo *loi,
2262 int cmd, struct loi_oap_pages *lop)
2264 struct ptlrpc_request *req;
2265 obd_count page_count = 0;
2266 struct osc_async_page *oap = NULL, *tmp;
2267 struct osc_brw_async_args *aa;
2268 const struct obd_async_page_ops *ops;
2269 CFS_LIST_HEAD(rpc_list);
2270 unsigned int ending_offset;
2271 unsigned starting_offset = 0;
2273 struct cl_object *clob = NULL;
2276 /* first we find the pages we're allowed to work with */
2277 list_for_each_entry_safe(oap, tmp, &lop->lop_pending,
2279 ops = oap->oap_caller_ops;
2281 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2282 "magic 0x%x\n", oap, oap->oap_magic);
2285 /* pin object in memory, so that completion call-backs
2286 * can be safely called under client_obd_list lock. */
2287 clob = osc_oap2cl_page(oap)->cp_obj;
2288 cl_object_get(clob);
2291 if (page_count != 0 &&
2292 srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2293 CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2294 " oap %p, page %p, srvlock %u\n",
2295 oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2298 /* in llite being 'ready' equates to the page being locked
2299 * until completion unlocks it. commit_write submits a page
2300 * as not ready because its unlock will happen unconditionally
2301 * as the call returns. if we race with commit_write giving
2302 * us that page we dont' want to create a hole in the page
2303 * stream, so we stop and leave the rpc to be fired by
2304 * another dirtier or kupdated interval (the not ready page
2305 * will still be on the dirty list). we could call in
2306 * at the end of ll_file_write to process the queue again. */
2307 if (!(oap->oap_async_flags & ASYNC_READY)) {
2308 int rc = ops->ap_make_ready(env, oap->oap_caller_data,
2311 CDEBUG(D_INODE, "oap %p page %p returned %d "
2312 "instead of ready\n", oap,
2316 /* llite is telling us that the page is still
2317 * in commit_write and that we should try
2318 * and put it in an rpc again later. we
2319 * break out of the loop so we don't create
2320 * a hole in the sequence of pages in the rpc
2325 /* the io isn't needed.. tell the checks
2326 * below to complete the rpc with EINTR */
2327 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2328 oap->oap_count = -EINTR;
2331 oap->oap_async_flags |= ASYNC_READY;
2334 LASSERTF(0, "oap %p page %p returned %d "
2335 "from make_ready\n", oap,
2343 * Page submitted for IO has to be locked. Either by
2344 * ->ap_make_ready() or by higher layers.
2346 #if defined(__KERNEL__) && defined(__linux__)
2348 struct cl_page *page;
2350 page = osc_oap2cl_page(oap);
2352 if (page->cp_type == CPT_CACHEABLE &&
2353 !(PageLocked(oap->oap_page) &&
2354 (CheckWriteback(oap->oap_page, cmd)))) {
2355 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2357 (long)oap->oap_page->flags,
2358 oap->oap_async_flags);
2363 /* If there is a gap at the start of this page, it can't merge
2364 * with any previous page, so we'll hand the network a
2365 * "fragmented" page array that it can't transfer in 1 RDMA */
2366 if (page_count != 0 && oap->oap_page_off != 0)
2369 /* take the page out of our book-keeping */
2370 list_del_init(&oap->oap_pending_item);
2371 lop_update_pending(cli, lop, cmd, -1);
2372 list_del_init(&oap->oap_urgent_item);
2374 if (page_count == 0)
2375 starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2376 (PTLRPC_MAX_BRW_SIZE - 1);
2378 /* ask the caller for the size of the io as the rpc leaves. */
2379 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
2381 ops->ap_refresh_count(env, oap->oap_caller_data,
2383 LASSERT(oap->oap_page_off + oap->oap_count <= CFS_PAGE_SIZE);
2385 if (oap->oap_count <= 0) {
2386 CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2388 osc_ap_completion(env, cli, NULL,
2389 oap, 0, oap->oap_count);
2393 /* now put the page back in our accounting */
2394 list_add_tail(&oap->oap_rpc_item, &rpc_list);
2395 if (page_count == 0)
2396 srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2397 if (++page_count >= cli->cl_max_pages_per_rpc)
2400 /* End on a PTLRPC_MAX_BRW_SIZE boundary. We want full-sized
2401 * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2402 * have the same alignment as the initial writes that allocated
2403 * extents on the server. */
2404 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2405 oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2406 if (ending_offset == 0)
2409 /* If there is a gap at the end of this page, it can't merge
2410 * with any subsequent pages, so we'll hand the network a
2411 * "fragmented" page array that it can't transfer in 1 RDMA */
2412 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2416 osc_wake_cache_waiters(cli);
2418 loi_list_maint(cli, loi);
2420 client_obd_list_unlock(&cli->cl_loi_list_lock);
2423 cl_object_put(env, clob);
2425 if (page_count == 0) {
2426 client_obd_list_lock(&cli->cl_loi_list_lock);
2430 req = osc_build_req(env, cli, &rpc_list, page_count, cmd);
2432 LASSERT(list_empty(&rpc_list));
2433 loi_list_maint(cli, loi);
2434 RETURN(PTR_ERR(req));
2437 aa = ptlrpc_req_async_args(req);
2439 if (cmd == OBD_BRW_READ) {
2440 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2441 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2442 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2443 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2445 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2446 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2447 cli->cl_w_in_flight);
2448 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2449 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2451 ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2453 client_obd_list_lock(&cli->cl_loi_list_lock);
2455 if (cmd == OBD_BRW_READ)
2456 cli->cl_r_in_flight++;
2458 cli->cl_w_in_flight++;
2460 /* queued sync pages can be torn down while the pages
2461 * were between the pending list and the rpc */
2463 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2464 /* only one oap gets a request reference */
2467 if (oap->oap_interrupted && !req->rq_intr) {
2468 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2470 ptlrpc_mark_interrupted(req);
2474 tmp->oap_request = ptlrpc_request_addref(req);
2476 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2477 page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2479 req->rq_interpret_reply = brw_interpret;
2480 ptlrpcd_add_req(req, PSCOPE_BRW);
2484 #define LOI_DEBUG(LOI, STR, args...) \
2485 CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR, \
2486 !list_empty(&(LOI)->loi_cli_item), \
2487 (LOI)->loi_write_lop.lop_num_pending, \
2488 !list_empty(&(LOI)->loi_write_lop.lop_urgent), \
2489 (LOI)->loi_read_lop.lop_num_pending, \
2490 !list_empty(&(LOI)->loi_read_lop.lop_urgent), \
2493 /* This is called by osc_check_rpcs() to find which objects have pages that
2494 * we could be sending. These lists are maintained by lop_makes_rpc(). */
2495 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2498 /* first return all objects which we already know to have
2499 * pages ready to be stuffed into rpcs */
2500 if (!list_empty(&cli->cl_loi_ready_list))
2501 RETURN(list_entry(cli->cl_loi_ready_list.next,
2502 struct lov_oinfo, loi_cli_item));
2504 /* then if we have cache waiters, return all objects with queued
2505 * writes. This is especially important when many small files
2506 * have filled up the cache and not been fired into rpcs because
2507 * they don't pass the nr_pending/object threshhold */
2508 if (!list_empty(&cli->cl_cache_waiters) &&
2509 !list_empty(&cli->cl_loi_write_list))
2510 RETURN(list_entry(cli->cl_loi_write_list.next,
2511 struct lov_oinfo, loi_write_item));
2513 /* then return all queued objects when we have an invalid import
2514 * so that they get flushed */
2515 if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2516 if (!list_empty(&cli->cl_loi_write_list))
2517 RETURN(list_entry(cli->cl_loi_write_list.next,
2518 struct lov_oinfo, loi_write_item));
2519 if (!list_empty(&cli->cl_loi_read_list))
2520 RETURN(list_entry(cli->cl_loi_read_list.next,
2521 struct lov_oinfo, loi_read_item));
2526 /* called with the loi list lock held */
2527 void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2529 struct lov_oinfo *loi;
2530 int rc = 0, race_counter = 0;
2533 while ((loi = osc_next_loi(cli)) != NULL) {
2534 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2536 if (rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight)
2539 /* attempt some read/write balancing by alternating between
2540 * reads and writes in an object. The makes_rpc checks here
2541 * would be redundant if we were getting read/write work items
2542 * instead of objects. we don't want send_oap_rpc to drain a
2543 * partial read pending queue when we're given this object to
2544 * do io on writes while there are cache waiters */
2545 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2546 rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_WRITE,
2547 &loi->loi_write_lop);
2555 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2556 rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_READ,
2557 &loi->loi_read_lop);
2566 /* attempt some inter-object balancing by issueing rpcs
2567 * for each object in turn */
2568 if (!list_empty(&loi->loi_cli_item))
2569 list_del_init(&loi->loi_cli_item);
2570 if (!list_empty(&loi->loi_write_item))
2571 list_del_init(&loi->loi_write_item);
2572 if (!list_empty(&loi->loi_read_item))
2573 list_del_init(&loi->loi_read_item);
2575 loi_list_maint(cli, loi);
2577 /* send_oap_rpc fails with 0 when make_ready tells it to
2578 * back off. llite's make_ready does this when it tries
2579 * to lock a page queued for write that is already locked.
2580 * we want to try sending rpcs from many objects, but we
2581 * don't want to spin failing with 0. */
2582 if (race_counter == 10)
2588 /* we're trying to queue a page in the osc so we're subject to the
2589 * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2590 * If the osc's queued pages are already at that limit, then we want to sleep
2591 * until there is space in the osc's queue for us. We also may be waiting for
2592 * write credits from the OST if there are RPCs in flight that may return some
2593 * before we fall back to sync writes.
2595 * We need this know our allocation was granted in the presence of signals */
2596 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2600 client_obd_list_lock(&cli->cl_loi_list_lock);
2601 rc = list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2602 client_obd_list_unlock(&cli->cl_loi_list_lock);
2607 * Non-blocking version of osc_enter_cache() that consumes grant only when it
2610 int osc_enter_cache_try(const struct lu_env *env,
2611 struct client_obd *cli, struct lov_oinfo *loi,
2612 struct osc_async_page *oap, int transient)
2616 has_grant = cli->cl_avail_grant >= CFS_PAGE_SIZE;
2618 osc_consume_write_grant(cli, &oap->oap_brw_page);
2620 cli->cl_dirty_transit += CFS_PAGE_SIZE;
2621 atomic_inc(&obd_dirty_transit_pages);
2622 oap->oap_brw_flags |= OBD_BRW_NOCACHE;
2628 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2629 * grant or cache space. */
2630 static int osc_enter_cache(const struct lu_env *env,
2631 struct client_obd *cli, struct lov_oinfo *loi,
2632 struct osc_async_page *oap)
2634 struct osc_cache_waiter ocw;
2635 struct l_wait_info lwi = { 0 };
2639 CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2640 "grant: %lu\n", cli->cl_dirty, atomic_read(&obd_dirty_pages),
2641 cli->cl_dirty_max, obd_max_dirty_pages,
2642 cli->cl_lost_grant, cli->cl_avail_grant);
2644 /* force the caller to try sync io. this can jump the list
2645 * of queued writes and create a discontiguous rpc stream */
2646 if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2647 loi->loi_ar.ar_force_sync)
2650 /* Hopefully normal case - cache space and write credits available */
2651 if (cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max &&
2652 atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages &&
2653 osc_enter_cache_try(env, cli, loi, oap, 0))
2656 /* Make sure that there are write rpcs in flight to wait for. This
2657 * is a little silly as this object may not have any pending but
2658 * other objects sure might. */
2659 if (cli->cl_w_in_flight) {
2660 list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2661 cfs_waitq_init(&ocw.ocw_waitq);
2665 loi_list_maint(cli, loi);
2666 osc_check_rpcs(env, cli);
2667 client_obd_list_unlock(&cli->cl_loi_list_lock);
2669 CDEBUG(D_CACHE, "sleeping for cache space\n");
2670 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2672 client_obd_list_lock(&cli->cl_loi_list_lock);
2673 if (!list_empty(&ocw.ocw_entry)) {
2674 list_del(&ocw.ocw_entry);
2684 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2685 struct lov_oinfo *loi, cfs_page_t *page,
2686 obd_off offset, const struct obd_async_page_ops *ops,
2687 void *data, void **res, int nocache,
2688 struct lustre_handle *lockh)
2690 struct osc_async_page *oap;
2695 return size_round(sizeof(*oap));
2698 oap->oap_magic = OAP_MAGIC;
2699 oap->oap_cli = &exp->exp_obd->u.cli;
2702 oap->oap_caller_ops = ops;
2703 oap->oap_caller_data = data;
2705 oap->oap_page = page;
2706 oap->oap_obj_off = offset;
2707 if (!client_is_remote(exp) &&
2708 cfs_capable(CFS_CAP_SYS_RESOURCE))
2709 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2711 LASSERT(!(offset & ~CFS_PAGE_MASK));
2713 CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2714 CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2715 CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2716 CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2718 spin_lock_init(&oap->oap_lock);
2719 CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
2723 struct osc_async_page *oap_from_cookie(void *cookie)
2725 struct osc_async_page *oap = cookie;
2726 if (oap->oap_magic != OAP_MAGIC)
2727 return ERR_PTR(-EINVAL);
2731 int osc_queue_async_io(const struct lu_env *env,
2732 struct obd_export *exp, struct lov_stripe_md *lsm,
2733 struct lov_oinfo *loi, void *cookie,
2734 int cmd, obd_off off, int count,
2735 obd_flag brw_flags, enum async_flags async_flags)
2737 struct client_obd *cli = &exp->exp_obd->u.cli;
2738 struct osc_async_page *oap;
2742 oap = oap_from_cookie(cookie);
2744 RETURN(PTR_ERR(oap));
2746 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2749 if (!list_empty(&oap->oap_pending_item) ||
2750 !list_empty(&oap->oap_urgent_item) ||
2751 !list_empty(&oap->oap_rpc_item))
2754 /* check if the file's owner/group is over quota */
2755 if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)) {
2756 struct cl_object *obj;
2757 struct cl_attr attr; /* XXX put attr into thread info */
2759 obj = cl_object_top(osc_oap2cl_page(oap)->cp_obj);
2761 cl_object_attr_lock(obj);
2762 rc = cl_object_attr_get(env, obj, &attr);
2763 cl_object_attr_unlock(obj);
2765 if (rc == 0 && lquota_chkdq(quota_interface, cli, attr.cat_uid,
2766 attr.cat_gid) == NO_QUOTA)
2773 loi = lsm->lsm_oinfo[0];
2775 client_obd_list_lock(&cli->cl_loi_list_lock);
2777 LASSERT(off + count <= CFS_PAGE_SIZE);
2779 oap->oap_page_off = off;
2780 oap->oap_count = count;
2781 oap->oap_brw_flags = brw_flags;
2782 oap->oap_async_flags = async_flags;
2784 if (cmd & OBD_BRW_WRITE) {
2785 rc = osc_enter_cache(env, cli, loi, oap);
2787 client_obd_list_unlock(&cli->cl_loi_list_lock);
2792 osc_oap_to_pending(oap);
2793 loi_list_maint(cli, loi);
2795 LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
2798 osc_check_rpcs(env, cli);
2799 client_obd_list_unlock(&cli->cl_loi_list_lock);
2804 /* aka (~was & now & flag), but this is more clear :) */
2805 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
2807 int osc_set_async_flags_base(struct client_obd *cli,
2808 struct lov_oinfo *loi, struct osc_async_page *oap,
2809 obd_flag async_flags)
2811 struct loi_oap_pages *lop;
2814 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2817 if (oap->oap_cmd & OBD_BRW_WRITE) {
2818 lop = &loi->loi_write_lop;
2820 lop = &loi->loi_read_lop;
2823 if (list_empty(&oap->oap_pending_item))
2826 if ((oap->oap_async_flags & async_flags) == async_flags)
2829 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
2830 oap->oap_async_flags |= ASYNC_READY;
2832 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT)) {
2833 if (list_empty(&oap->oap_rpc_item)) {
2834 list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2835 loi_list_maint(cli, loi);
2839 LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
2840 oap->oap_async_flags);
2844 int osc_teardown_async_page(struct obd_export *exp,
2845 struct lov_stripe_md *lsm,
2846 struct lov_oinfo *loi, void *cookie)
2848 struct client_obd *cli = &exp->exp_obd->u.cli;
2849 struct loi_oap_pages *lop;
2850 struct osc_async_page *oap;
2854 oap = oap_from_cookie(cookie);
2856 RETURN(PTR_ERR(oap));
2859 loi = lsm->lsm_oinfo[0];
2861 if (oap->oap_cmd & OBD_BRW_WRITE) {
2862 lop = &loi->loi_write_lop;
2864 lop = &loi->loi_read_lop;
2867 client_obd_list_lock(&cli->cl_loi_list_lock);
2869 if (!list_empty(&oap->oap_rpc_item))
2870 GOTO(out, rc = -EBUSY);
2872 osc_exit_cache(cli, oap, 0);
2873 osc_wake_cache_waiters(cli);
2875 if (!list_empty(&oap->oap_urgent_item)) {
2876 list_del_init(&oap->oap_urgent_item);
2877 oap->oap_async_flags &= ~ASYNC_URGENT;
2879 if (!list_empty(&oap->oap_pending_item)) {
2880 list_del_init(&oap->oap_pending_item);
2881 lop_update_pending(cli, lop, oap->oap_cmd, -1);
2883 loi_list_maint(cli, loi);
2884 LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
2886 client_obd_list_unlock(&cli->cl_loi_list_lock);
2890 static void osc_set_lock_data_with_check(struct ldlm_lock *lock,
2891 struct ldlm_enqueue_info *einfo,
2894 void *data = einfo->ei_cbdata;
2896 LASSERT(lock != NULL);
2897 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2898 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2899 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2900 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2902 lock_res_and_lock(lock);
2903 spin_lock(&osc_ast_guard);
2904 LASSERT(lock->l_ast_data == NULL || lock->l_ast_data == data);
2905 lock->l_ast_data = data;
2906 spin_unlock(&osc_ast_guard);
2907 unlock_res_and_lock(lock);
2910 static void osc_set_data_with_check(struct lustre_handle *lockh,
2911 struct ldlm_enqueue_info *einfo,
2914 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2917 osc_set_lock_data_with_check(lock, einfo, flags);
2918 LDLM_LOCK_PUT(lock);
2920 CERROR("lockh %p, data %p - client evicted?\n",
2921 lockh, einfo->ei_cbdata);
2924 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2925 ldlm_iterator_t replace, void *data)
2927 struct ldlm_res_id res_id;
2928 struct obd_device *obd = class_exp2obd(exp);
2930 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_gr, &res_id);
2931 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2935 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
2936 obd_enqueue_update_f upcall, void *cookie,
2939 int intent = *flags & LDLM_FL_HAS_INTENT;
2943 /* The request was created before ldlm_cli_enqueue call. */
2944 if (rc == ELDLM_LOCK_ABORTED) {
2945 struct ldlm_reply *rep;
2946 rep = req_capsule_server_get(&req->rq_pill,
2949 LASSERT(rep != NULL);
2950 if (rep->lock_policy_res1)
2951 rc = rep->lock_policy_res1;
2955 if ((intent && rc == ELDLM_LOCK_ABORTED) || !rc) {
2956 *flags |= LDLM_FL_LVB_READY;
2957 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
2958 lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
2961 /* Call the update callback. */
2962 rc = (*upcall)(cookie, rc);
2966 static int osc_enqueue_interpret(const struct lu_env *env,
2967 struct ptlrpc_request *req,
2968 struct osc_enqueue_args *aa, int rc)
2970 struct ldlm_lock *lock;
2971 struct lustre_handle handle;
2974 /* Make a local copy of a lock handle and a mode, because aa->oa_*
2975 * might be freed anytime after lock upcall has been called. */
2976 lustre_handle_copy(&handle, aa->oa_lockh);
2977 mode = aa->oa_ei->ei_mode;
2979 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2981 lock = ldlm_handle2lock(&handle);
2983 /* Take an additional reference so that a blocking AST that
2984 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2985 * to arrive after an upcall has been executed by
2986 * osc_enqueue_fini(). */
2987 ldlm_lock_addref(&handle, mode);
2989 /* Complete obtaining the lock procedure. */
2990 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
2991 mode, aa->oa_flags, aa->oa_lvb,
2992 sizeof(*aa->oa_lvb), lustre_swab_ost_lvb,
2994 /* Complete osc stuff. */
2995 rc = osc_enqueue_fini(req, aa->oa_lvb,
2996 aa->oa_upcall, aa->oa_cookie, aa->oa_flags, rc);
2997 /* Release the lock for async request. */
2998 if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
3000 * Releases a reference taken by ldlm_cli_enqueue(), if it is
3001 * not already released by
3002 * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
3004 ldlm_lock_decref(&handle, mode);
3006 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
3007 aa->oa_lockh, req, aa);
3008 ldlm_lock_decref(&handle, mode);
3009 LDLM_LOCK_PUT(lock);
3013 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
3014 struct lov_oinfo *loi, int flags,
3015 struct ost_lvb *lvb, __u32 mode, int rc)
3017 if (rc == ELDLM_OK) {
3018 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
3021 LASSERT(lock != NULL);
3022 loi->loi_lvb = *lvb;
3023 tmp = loi->loi_lvb.lvb_size;
3024 /* Extend KMS up to the end of this lock and no further
3025 * A lock on [x,y] means a KMS of up to y + 1 bytes! */
3026 if (tmp > lock->l_policy_data.l_extent.end)
3027 tmp = lock->l_policy_data.l_extent.end + 1;
3028 if (tmp >= loi->loi_kms) {
3029 LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
3030 ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
3031 loi_kms_set(loi, tmp);
3033 LDLM_DEBUG(lock, "lock acquired, setting rss="
3034 LPU64"; leaving kms="LPU64", end="LPU64,
3035 loi->loi_lvb.lvb_size, loi->loi_kms,
3036 lock->l_policy_data.l_extent.end);
3038 ldlm_lock_allow_match(lock);
3039 LDLM_LOCK_PUT(lock);
3040 } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
3041 loi->loi_lvb = *lvb;
3042 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
3043 " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
3047 EXPORT_SYMBOL(osc_update_enqueue);
3049 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
3051 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
3052 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
3053 * other synchronous requests, however keeping some locks and trying to obtain
3054 * others may take a considerable amount of time in a case of ost failure; and
3055 * when other sync requests do not get released lock from a client, the client
3056 * is excluded from the cluster -- such scenarious make the life difficult, so
3057 * release locks just after they are obtained. */
3058 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3059 int *flags, ldlm_policy_data_t *policy,
3060 struct ost_lvb *lvb, int kms_valid,
3061 obd_enqueue_update_f upcall, void *cookie,
3062 struct ldlm_enqueue_info *einfo,
3063 struct lustre_handle *lockh,
3064 struct ptlrpc_request_set *rqset, int async)
3066 struct obd_device *obd = exp->exp_obd;
3067 struct ptlrpc_request *req = NULL;
3068 int intent = *flags & LDLM_FL_HAS_INTENT;
3073 /* Filesystem lock extents are extended to page boundaries so that
3074 * dealing with the page cache is a little smoother. */
3075 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3076 policy->l_extent.end |= ~CFS_PAGE_MASK;
3079 * kms is not valid when either object is completely fresh (so that no
3080 * locks are cached), or object was evicted. In the latter case cached
3081 * lock cannot be used, because it would prime inode state with
3082 * potentially stale LVB.
3087 /* Next, search for already existing extent locks that will cover us */
3088 /* If we're trying to read, we also search for an existing PW lock. The
3089 * VFS and page cache already protect us locally, so lots of readers/
3090 * writers can share a single PW lock.
3092 * There are problems with conversion deadlocks, so instead of
3093 * converting a read lock to a write lock, we'll just enqueue a new
3096 * At some point we should cancel the read lock instead of making them
3097 * send us a blocking callback, but there are problems with canceling
3098 * locks out from other users right now, too. */
3099 mode = einfo->ei_mode;
3100 if (einfo->ei_mode == LCK_PR)
3102 mode = ldlm_lock_match(obd->obd_namespace,
3103 *flags | LDLM_FL_LVB_READY, res_id,
3104 einfo->ei_type, policy, mode, lockh, 0);
3106 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
3108 if (matched->l_ast_data == NULL ||
3109 matched->l_ast_data == einfo->ei_cbdata) {
3110 /* addref the lock only if not async requests and PW
3111 * lock is matched whereas we asked for PR. */
3112 if (!rqset && einfo->ei_mode != mode)
3113 ldlm_lock_addref(lockh, LCK_PR);
3114 osc_set_lock_data_with_check(matched, einfo, *flags);
3116 /* I would like to be able to ASSERT here that
3117 * rss <= kms, but I can't, for reasons which
3118 * are explained in lov_enqueue() */
3121 /* We already have a lock, and it's referenced */
3122 (*upcall)(cookie, ELDLM_OK);
3124 /* For async requests, decref the lock. */
3125 if (einfo->ei_mode != mode)
3126 ldlm_lock_decref(lockh, LCK_PW);
3128 ldlm_lock_decref(lockh, einfo->ei_mode);
3129 LDLM_LOCK_PUT(matched);
3132 ldlm_lock_decref(lockh, mode);
3133 LDLM_LOCK_PUT(matched);
3138 CFS_LIST_HEAD(cancels);
3139 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3140 &RQF_LDLM_ENQUEUE_LVB);
3144 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
3148 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
3150 ptlrpc_request_set_replen(req);
3153 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
3154 *flags &= ~LDLM_FL_BLOCK_GRANTED;
3156 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
3157 sizeof(*lvb), lustre_swab_ost_lvb, lockh, async);
3160 struct osc_enqueue_args *aa;
3161 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3162 aa = ptlrpc_req_async_args(req);
3165 aa->oa_flags = flags;
3166 aa->oa_upcall = upcall;
3167 aa->oa_cookie = cookie;
3169 aa->oa_lockh = lockh;
3171 req->rq_interpret_reply =
3172 (ptlrpc_interpterer_t)osc_enqueue_interpret;
3173 if (rqset == PTLRPCD_SET)
3174 ptlrpcd_add_req(req, PSCOPE_OTHER);
3176 ptlrpc_set_add_req(rqset, req);
3177 } else if (intent) {
3178 ptlrpc_req_finished(req);
3183 rc = osc_enqueue_fini(req, lvb, upcall, cookie, flags, rc);
3185 ptlrpc_req_finished(req);
3190 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
3191 struct ldlm_enqueue_info *einfo,
3192 struct ptlrpc_request_set *rqset)
3194 struct ldlm_res_id res_id;
3198 osc_build_res_name(oinfo->oi_md->lsm_object_id,
3199 oinfo->oi_md->lsm_object_gr, &res_id);
3201 rc = osc_enqueue_base(exp, &res_id, &oinfo->oi_flags, &oinfo->oi_policy,
3202 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
3203 oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid,
3204 oinfo->oi_cb_up, oinfo, einfo, oinfo->oi_lockh,
3205 rqset, rqset != NULL);
3209 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3210 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
3211 int *flags, void *data, struct lustre_handle *lockh,
3214 struct obd_device *obd = exp->exp_obd;
3215 int lflags = *flags;
3219 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
3222 /* Filesystem lock extents are extended to page boundaries so that
3223 * dealing with the page cache is a little smoother */
3224 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3225 policy->l_extent.end |= ~CFS_PAGE_MASK;
3227 /* Next, search for already existing extent locks that will cover us */
3228 /* If we're trying to read, we also search for an existing PW lock. The
3229 * VFS and page cache already protect us locally, so lots of readers/
3230 * writers can share a single PW lock. */
3234 rc = ldlm_lock_match(obd->obd_namespace, lflags | LDLM_FL_LVB_READY,
3235 res_id, type, policy, rc, lockh, unref);
3238 osc_set_data_with_check(lockh, data, lflags);
3239 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
3240 ldlm_lock_addref(lockh, LCK_PR);
3241 ldlm_lock_decref(lockh, LCK_PW);
3248 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
3252 if (unlikely(mode == LCK_GROUP))
3253 ldlm_lock_decref_and_cancel(lockh, mode);
3255 ldlm_lock_decref(lockh, mode);
3260 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
3261 __u32 mode, struct lustre_handle *lockh)
3264 RETURN(osc_cancel_base(lockh, mode));
3267 static int osc_cancel_unused(struct obd_export *exp,
3268 struct lov_stripe_md *lsm, int flags,
3271 struct obd_device *obd = class_exp2obd(exp);
3272 struct ldlm_res_id res_id, *resp = NULL;
3275 resp = osc_build_res_name(lsm->lsm_object_id,
3276 lsm->lsm_object_gr, &res_id);
3279 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
3282 static int osc_statfs_interpret(const struct lu_env *env,
3283 struct ptlrpc_request *req,
3284 struct osc_async_args *aa, int rc)
3286 struct obd_statfs *msfs;
3292 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3294 GOTO(out, rc = -EPROTO);
3297 *aa->aa_oi->oi_osfs = *msfs;
3299 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
3303 static int osc_statfs_async(struct obd_device *obd, struct obd_info *oinfo,
3304 __u64 max_age, struct ptlrpc_request_set *rqset)
3306 struct ptlrpc_request *req;
3307 struct osc_async_args *aa;
3311 /* We could possibly pass max_age in the request (as an absolute
3312 * timestamp or a "seconds.usec ago") so the target can avoid doing
3313 * extra calls into the filesystem if that isn't necessary (e.g.
3314 * during mount that would help a bit). Having relative timestamps
3315 * is not so great if request processing is slow, while absolute
3316 * timestamps are not ideal because they need time synchronization. */
3317 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
3321 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3323 ptlrpc_request_free(req);
3326 ptlrpc_request_set_replen(req);
3327 req->rq_request_portal = OST_CREATE_PORTAL;
3328 ptlrpc_at_set_req_timeout(req);
3330 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
3331 /* procfs requests not want stat in wait for avoid deadlock */
3332 req->rq_no_resend = 1;
3333 req->rq_no_delay = 1;
3336 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
3337 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3338 aa = ptlrpc_req_async_args(req);
3341 ptlrpc_set_add_req(rqset, req);
3345 static int osc_statfs(struct obd_device *obd, struct obd_statfs *osfs,
3346 __u64 max_age, __u32 flags)
3348 struct obd_statfs *msfs;
3349 struct ptlrpc_request *req;
3350 struct obd_import *imp = NULL;
3354 /*Since the request might also come from lprocfs, so we need
3355 *sync this with client_disconnect_export Bug15684*/
3356 down_read(&obd->u.cli.cl_sem);
3357 if (obd->u.cli.cl_import)
3358 imp = class_import_get(obd->u.cli.cl_import);
3359 up_read(&obd->u.cli.cl_sem);
3363 /* We could possibly pass max_age in the request (as an absolute
3364 * timestamp or a "seconds.usec ago") so the target can avoid doing
3365 * extra calls into the filesystem if that isn't necessary (e.g.
3366 * during mount that would help a bit). Having relative timestamps
3367 * is not so great if request processing is slow, while absolute
3368 * timestamps are not ideal because they need time synchronization. */
3369 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
3371 class_import_put(imp);
3376 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
3378 ptlrpc_request_free(req);
3381 ptlrpc_request_set_replen(req);
3382 req->rq_request_portal = OST_CREATE_PORTAL;
3383 ptlrpc_at_set_req_timeout(req);
3385 if (flags & OBD_STATFS_NODELAY) {
3386 /* procfs requests not want stat in wait for avoid deadlock */
3387 req->rq_no_resend = 1;
3388 req->rq_no_delay = 1;
3391 rc = ptlrpc_queue_wait(req);
3395 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
3397 GOTO(out, rc = -EPROTO);
3404 ptlrpc_req_finished(req);
3408 /* Retrieve object striping information.
3410 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
3411 * the maximum number of OST indices which will fit in the user buffer.
3412 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
3414 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
3416 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
3417 struct lov_user_md_v3 lum, *lumk;
3418 struct lov_user_ost_data_v1 *lmm_objects;
3419 int rc = 0, lum_size;
3425 /* we only need the header part from user space to get lmm_magic and
3426 * lmm_stripe_count, (the header part is common to v1 and v3) */
3427 lum_size = sizeof(struct lov_user_md_v1);
3428 if (copy_from_user(&lum, lump, lum_size))
3431 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
3432 (lum.lmm_magic != LOV_USER_MAGIC_V3))
3435 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
3436 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
3437 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
3438 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
3440 /* we can use lov_mds_md_size() to compute lum_size
3441 * because lov_user_md_vX and lov_mds_md_vX have the same size */
3442 if (lum.lmm_stripe_count > 0) {
3443 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
3444 OBD_ALLOC(lumk, lum_size);
3448 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
3449 lmm_objects = &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
3451 lmm_objects = &(lumk->lmm_objects[0]);
3452 lmm_objects->l_object_id = lsm->lsm_object_id;
3454 lum_size = lov_mds_md_size(0, lum.lmm_magic);
3458 lumk->lmm_object_id = lsm->lsm_object_id;
3459 lumk->lmm_object_gr = lsm->lsm_object_gr;
3460 lumk->lmm_stripe_count = 1;
3462 if (copy_to_user(lump, lumk, lum_size))
3466 OBD_FREE(lumk, lum_size);
3472 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
3473 void *karg, void *uarg)
3475 struct obd_device *obd = exp->exp_obd;
3476 struct obd_ioctl_data *data = karg;
3480 if (!try_module_get(THIS_MODULE)) {
3481 CERROR("Can't get module. Is it alive?");
3485 case OBD_IOC_LOV_GET_CONFIG: {
3487 struct lov_desc *desc;
3488 struct obd_uuid uuid;
3492 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
3493 GOTO(out, err = -EINVAL);
3495 data = (struct obd_ioctl_data *)buf;
3497 if (sizeof(*desc) > data->ioc_inllen1) {
3498 obd_ioctl_freedata(buf, len);
3499 GOTO(out, err = -EINVAL);
3502 if (data->ioc_inllen2 < sizeof(uuid)) {
3503 obd_ioctl_freedata(buf, len);
3504 GOTO(out, err = -EINVAL);
3507 desc = (struct lov_desc *)data->ioc_inlbuf1;
3508 desc->ld_tgt_count = 1;
3509 desc->ld_active_tgt_count = 1;
3510 desc->ld_default_stripe_count = 1;
3511 desc->ld_default_stripe_size = 0;
3512 desc->ld_default_stripe_offset = 0;
3513 desc->ld_pattern = 0;
3514 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
3516 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
3518 err = copy_to_user((void *)uarg, buf, len);
3521 obd_ioctl_freedata(buf, len);
3524 case LL_IOC_LOV_SETSTRIPE:
3525 err = obd_alloc_memmd(exp, karg);
3529 case LL_IOC_LOV_GETSTRIPE:
3530 err = osc_getstripe(karg, uarg);
3532 case OBD_IOC_CLIENT_RECOVER:
3533 err = ptlrpc_recover_import(obd->u.cli.cl_import,
3538 case IOC_OSC_SET_ACTIVE:
3539 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
3542 case OBD_IOC_POLL_QUOTACHECK:
3543 err = lquota_poll_check(quota_interface, exp,
3544 (struct if_quotacheck *)karg);
3546 case OBD_IOC_PING_TARGET:
3547 err = ptlrpc_obd_ping(obd);
3550 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
3551 cmd, cfs_curproc_comm());
3552 GOTO(out, err = -ENOTTY);
3555 module_put(THIS_MODULE);
3559 static int osc_get_info(struct obd_export *exp, obd_count keylen,
3560 void *key, __u32 *vallen, void *val,
3561 struct lov_stripe_md *lsm)
3564 if (!vallen || !val)
3567 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
3568 __u32 *stripe = val;
3569 *vallen = sizeof(*stripe);
3572 } else if (KEY_IS(KEY_LAST_ID)) {
3573 struct ptlrpc_request *req;
3578 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3579 &RQF_OST_GET_INFO_LAST_ID);
3583 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3584 RCL_CLIENT, keylen);
3585 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3587 ptlrpc_request_free(req);
3591 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3592 memcpy(tmp, key, keylen);
3594 ptlrpc_request_set_replen(req);
3595 rc = ptlrpc_queue_wait(req);
3599 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
3601 GOTO(out, rc = -EPROTO);
3603 *((obd_id *)val) = *reply;
3605 ptlrpc_req_finished(req);
3607 } else if (KEY_IS(KEY_FIEMAP)) {
3608 struct ptlrpc_request *req;
3609 struct ll_user_fiemap *reply;
3613 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3614 &RQF_OST_GET_INFO_FIEMAP);
3618 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
3619 RCL_CLIENT, keylen);
3620 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3621 RCL_CLIENT, *vallen);
3622 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
3623 RCL_SERVER, *vallen);
3625 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
3627 ptlrpc_request_free(req);
3631 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
3632 memcpy(tmp, key, keylen);
3633 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3634 memcpy(tmp, val, *vallen);
3636 ptlrpc_request_set_replen(req);
3637 rc = ptlrpc_queue_wait(req);
3641 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
3643 GOTO(out1, rc = -EPROTO);
3645 memcpy(val, reply, *vallen);
3647 ptlrpc_req_finished(req);
3655 static int osc_setinfo_mds_conn_interpret(const struct lu_env *env,
3656 struct ptlrpc_request *req,
3659 struct llog_ctxt *ctxt;
3660 struct obd_import *imp = req->rq_import;
3666 ctxt = llog_get_context(imp->imp_obd, LLOG_MDS_OST_ORIG_CTXT);
3669 rc = llog_initiator_connect(ctxt);
3671 CERROR("cannot establish connection for "
3672 "ctxt %p: %d\n", ctxt, rc);
3675 llog_ctxt_put(ctxt);
3676 spin_lock(&imp->imp_lock);
3677 imp->imp_server_timeout = 1;
3678 imp->imp_pingable = 1;
3679 spin_unlock(&imp->imp_lock);
3680 CDEBUG(D_RPCTRACE, "pinging OST %s\n", obd2cli_tgt(imp->imp_obd));
3685 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
3686 void *key, obd_count vallen, void *val,
3687 struct ptlrpc_request_set *set)
3689 struct ptlrpc_request *req;
3690 struct obd_device *obd = exp->exp_obd;
3691 struct obd_import *imp = class_exp2cliimp(exp);
3696 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
3698 if (KEY_IS(KEY_NEXT_ID)) {
3699 if (vallen != sizeof(obd_id))
3703 obd->u.cli.cl_oscc.oscc_next_id = *((obd_id*)val) + 1;
3704 CDEBUG(D_HA, "%s: set oscc_next_id = "LPU64"\n",
3705 exp->exp_obd->obd_name,
3706 obd->u.cli.cl_oscc.oscc_next_id);
3711 if (KEY_IS(KEY_UNLINKED)) {
3712 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3713 spin_lock(&oscc->oscc_lock);
3714 oscc->oscc_flags &= ~OSCC_FLAG_NOSPC;
3715 spin_unlock(&oscc->oscc_lock);
3719 if (KEY_IS(KEY_INIT_RECOV)) {
3720 if (vallen != sizeof(int))
3722 spin_lock(&imp->imp_lock);
3723 imp->imp_initial_recov = *(int *)val;
3724 spin_unlock(&imp->imp_lock);
3725 CDEBUG(D_HA, "%s: set imp_initial_recov = %d\n",
3726 exp->exp_obd->obd_name,
3727 imp->imp_initial_recov);
3731 if (KEY_IS(KEY_CHECKSUM)) {
3732 if (vallen != sizeof(int))
3734 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
3738 if (KEY_IS(KEY_SPTLRPC_CONF)) {
3739 sptlrpc_conf_client_adapt(obd);
3743 if (KEY_IS(KEY_FLUSH_CTX)) {
3744 sptlrpc_import_flush_my_ctx(imp);
3748 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
3751 /* We pass all other commands directly to OST. Since nobody calls osc
3752 methods directly and everybody is supposed to go through LOV, we
3753 assume lov checked invalid values for us.
3754 The only recognised values so far are evict_by_nid and mds_conn.
3755 Even if something bad goes through, we'd get a -EINVAL from OST
3758 if (KEY_IS(KEY_GRANT_SHRINK))
3759 req = ptlrpc_request_alloc(imp, &RQF_OST_SET_GRANT_INFO);
3761 req = ptlrpc_request_alloc(imp, &RQF_OST_SET_INFO);
3766 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
3767 RCL_CLIENT, keylen);
3768 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
3769 RCL_CLIENT, vallen);
3770 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
3772 ptlrpc_request_free(req);
3776 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
3777 memcpy(tmp, key, keylen);
3778 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_VAL);
3779 memcpy(tmp, val, vallen);
3781 if (KEY_IS(KEY_MDS_CONN)) {
3782 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3784 oscc->oscc_oa.o_gr = (*(__u32 *)val);
3785 oscc->oscc_oa.o_valid |= OBD_MD_FLGROUP;
3786 LASSERT_MDS_GROUP(oscc->oscc_oa.o_gr);
3787 req->rq_interpret_reply = osc_setinfo_mds_conn_interpret;
3788 } else if (KEY_IS(KEY_GRANT_SHRINK)) {
3789 struct osc_grant_args *aa;
3792 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3793 aa = ptlrpc_req_async_args(req);
3796 ptlrpc_req_finished(req);
3799 *oa = ((struct ost_body *)val)->oa;
3801 req->rq_interpret_reply = osc_shrink_grant_interpret;
3804 ptlrpc_request_set_replen(req);
3805 if (!KEY_IS(KEY_GRANT_SHRINK)) {
3806 LASSERT(set != NULL);
3807 ptlrpc_set_add_req(set, req);
3808 ptlrpc_check_set(NULL, set);
3810 ptlrpcd_add_req(req, PSCOPE_OTHER);
3816 static struct llog_operations osc_size_repl_logops = {
3817 lop_cancel: llog_obd_repl_cancel
3820 static struct llog_operations osc_mds_ost_orig_logops;
3821 static int osc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
3822 struct obd_device *tgt, int count,
3823 struct llog_catid *catid, struct obd_uuid *uuid)
3828 LASSERT(olg == &obd->obd_olg);
3829 spin_lock(&obd->obd_dev_lock);
3830 if (osc_mds_ost_orig_logops.lop_setup != llog_obd_origin_setup) {
3831 osc_mds_ost_orig_logops = llog_lvfs_ops;
3832 osc_mds_ost_orig_logops.lop_setup = llog_obd_origin_setup;
3833 osc_mds_ost_orig_logops.lop_cleanup = llog_obd_origin_cleanup;
3834 osc_mds_ost_orig_logops.lop_add = llog_obd_origin_add;
3835 osc_mds_ost_orig_logops.lop_connect = llog_origin_connect;
3837 spin_unlock(&obd->obd_dev_lock);
3839 rc = llog_setup(obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT, tgt, count,
3840 &catid->lci_logid, &osc_mds_ost_orig_logops);
3842 CERROR("failed LLOG_MDS_OST_ORIG_CTXT\n");
3846 rc = llog_setup(obd, &obd->obd_olg, LLOG_SIZE_REPL_CTXT, tgt, count,
3847 NULL, &osc_size_repl_logops);
3849 struct llog_ctxt *ctxt =
3850 llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3853 CERROR("failed LLOG_SIZE_REPL_CTXT\n");
3858 CERROR("osc '%s' tgt '%s' cnt %d catid %p rc=%d\n",
3859 obd->obd_name, tgt->obd_name, count, catid, rc);
3860 CERROR("logid "LPX64":0x%x\n",
3861 catid->lci_logid.lgl_oid, catid->lci_logid.lgl_ogen);
3866 static int osc_llog_finish(struct obd_device *obd, int count)
3868 struct llog_ctxt *ctxt;
3869 int rc = 0, rc2 = 0;
3872 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3874 rc = llog_cleanup(ctxt);
3876 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3878 rc2 = llog_cleanup(ctxt);
3885 static int osc_reconnect(const struct lu_env *env,
3886 struct obd_export *exp, struct obd_device *obd,
3887 struct obd_uuid *cluuid,
3888 struct obd_connect_data *data,
3891 struct client_obd *cli = &obd->u.cli;
3893 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
3896 client_obd_list_lock(&cli->cl_loi_list_lock);
3897 data->ocd_grant = cli->cl_avail_grant ?:
3898 2 * cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
3899 lost_grant = cli->cl_lost_grant;
3900 cli->cl_lost_grant = 0;
3901 client_obd_list_unlock(&cli->cl_loi_list_lock);
3903 CDEBUG(D_CACHE, "request ocd_grant: %d cl_avail_grant: %ld "
3904 "cl_lost_grant: %ld\n", data->ocd_grant,
3905 cli->cl_avail_grant, lost_grant);
3906 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
3907 " ocd_grant: %d\n", data->ocd_connect_flags,
3908 data->ocd_version, data->ocd_grant);
3914 static int osc_disconnect(struct obd_export *exp)
3916 struct obd_device *obd = class_exp2obd(exp);
3917 struct llog_ctxt *ctxt;
3920 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3922 if (obd->u.cli.cl_conn_count == 1) {
3923 /* Flush any remaining cancel messages out to the
3925 llog_sync(ctxt, exp);
3927 llog_ctxt_put(ctxt);
3929 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
3933 osc_del_shrink_grant(&obd->u.cli);
3934 rc = client_disconnect_export(exp);
3938 static int osc_import_event(struct obd_device *obd,
3939 struct obd_import *imp,
3940 enum obd_import_event event)
3942 struct client_obd *cli;
3946 LASSERT(imp->imp_obd == obd);
3949 case IMP_EVENT_DISCON: {
3950 /* Only do this on the MDS OSC's */
3951 if (imp->imp_server_timeout) {
3952 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3954 spin_lock(&oscc->oscc_lock);
3955 oscc->oscc_flags |= OSCC_FLAG_RECOVERING;
3956 spin_unlock(&oscc->oscc_lock);
3959 client_obd_list_lock(&cli->cl_loi_list_lock);
3960 cli->cl_avail_grant = 0;
3961 cli->cl_lost_grant = 0;
3962 client_obd_list_unlock(&cli->cl_loi_list_lock);
3965 case IMP_EVENT_INACTIVE: {
3966 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
3969 case IMP_EVENT_INVALIDATE: {
3970 struct ldlm_namespace *ns = obd->obd_namespace;
3974 env = cl_env_get(&refcheck);
3978 client_obd_list_lock(&cli->cl_loi_list_lock);
3979 /* all pages go to failing rpcs due to the invalid
3981 osc_check_rpcs(env, cli);
3982 client_obd_list_unlock(&cli->cl_loi_list_lock);
3984 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3985 cl_env_put(env, &refcheck);
3990 case IMP_EVENT_ACTIVE: {
3991 /* Only do this on the MDS OSC's */
3992 if (imp->imp_server_timeout) {
3993 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3995 spin_lock(&oscc->oscc_lock);
3996 oscc->oscc_flags &= ~OSCC_FLAG_NOSPC;
3997 spin_unlock(&oscc->oscc_lock);
3999 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
4002 case IMP_EVENT_OCD: {
4003 struct obd_connect_data *ocd = &imp->imp_connect_data;
4005 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
4006 osc_init_grant(&obd->u.cli, ocd);
4009 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
4010 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
4012 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
4016 CERROR("Unknown import event %d\n", event);
4022 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
4028 rc = ptlrpcd_addref();
4032 rc = client_obd_setup(obd, lcfg);
4036 struct lprocfs_static_vars lvars = { 0 };
4037 struct client_obd *cli = &obd->u.cli;
4039 lprocfs_osc_init_vars(&lvars);
4040 if (lprocfs_obd_setup(obd, lvars.obd_vars) == 0) {
4041 lproc_osc_attach_seqstat(obd);
4042 sptlrpc_lprocfs_cliobd_attach(obd);
4043 ptlrpc_lprocfs_register_obd(obd);
4047 /* We need to allocate a few requests more, because
4048 brw_interpret tries to create new requests before freeing
4049 previous ones. Ideally we want to have 2x max_rpcs_in_flight
4050 reserved, but I afraid that might be too much wasted RAM
4051 in fact, so 2 is just my guess and still should work. */
4052 cli->cl_import->imp_rq_pool =
4053 ptlrpc_init_rq_pool(cli->cl_max_rpcs_in_flight + 2,
4055 ptlrpc_add_rqs_to_pool);
4057 CFS_INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
4058 sema_init(&cli->cl_grant_sem, 1);
4064 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
4070 case OBD_CLEANUP_EARLY: {
4071 struct obd_import *imp;
4072 imp = obd->u.cli.cl_import;
4073 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
4074 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
4075 ptlrpc_deactivate_import(imp);
4076 spin_lock(&imp->imp_lock);
4077 imp->imp_pingable = 0;
4078 spin_unlock(&imp->imp_lock);
4081 case OBD_CLEANUP_EXPORTS: {
4082 /* If we set up but never connected, the
4083 client import will not have been cleaned. */
4084 if (obd->u.cli.cl_import) {
4085 struct obd_import *imp;
4086 down_write(&obd->u.cli.cl_sem);
4087 imp = obd->u.cli.cl_import;
4088 CDEBUG(D_CONFIG, "%s: client import never connected\n",
4090 ptlrpc_invalidate_import(imp);
4091 if (imp->imp_rq_pool) {
4092 ptlrpc_free_rq_pool(imp->imp_rq_pool);
4093 imp->imp_rq_pool = NULL;
4095 class_destroy_import(imp);
4096 up_write(&obd->u.cli.cl_sem);
4097 obd->u.cli.cl_import = NULL;
4099 rc = obd_llog_finish(obd, 0);
4101 CERROR("failed to cleanup llogging subsystems\n");
4108 int osc_cleanup(struct obd_device *obd)
4110 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4114 ptlrpc_lprocfs_unregister_obd(obd);
4115 lprocfs_obd_cleanup(obd);
4117 spin_lock(&oscc->oscc_lock);
4118 oscc->oscc_flags &= ~OSCC_FLAG_RECOVERING;
4119 oscc->oscc_flags |= OSCC_FLAG_EXITING;
4120 spin_unlock(&oscc->oscc_lock);
4122 /* free memory of osc quota cache */
4123 lquota_cleanup(quota_interface, obd);
4125 rc = client_obd_cleanup(obd);
4131 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
4133 struct lprocfs_static_vars lvars = { 0 };
4136 lprocfs_osc_init_vars(&lvars);
4138 switch (lcfg->lcfg_command) {
4140 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars,
4150 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
4152 return osc_process_config_base(obd, buf);
4155 struct obd_ops osc_obd_ops = {
4156 .o_owner = THIS_MODULE,
4157 .o_setup = osc_setup,
4158 .o_precleanup = osc_precleanup,
4159 .o_cleanup = osc_cleanup,
4160 .o_add_conn = client_import_add_conn,
4161 .o_del_conn = client_import_del_conn,
4162 .o_connect = client_connect_import,
4163 .o_reconnect = osc_reconnect,
4164 .o_disconnect = osc_disconnect,
4165 .o_statfs = osc_statfs,
4166 .o_statfs_async = osc_statfs_async,
4167 .o_packmd = osc_packmd,
4168 .o_unpackmd = osc_unpackmd,
4169 .o_precreate = osc_precreate,
4170 .o_create = osc_create,
4171 .o_destroy = osc_destroy,
4172 .o_getattr = osc_getattr,
4173 .o_getattr_async = osc_getattr_async,
4174 .o_setattr = osc_setattr,
4175 .o_setattr_async = osc_setattr_async,
4177 .o_punch = osc_punch,
4179 .o_enqueue = osc_enqueue,
4180 .o_change_cbdata = osc_change_cbdata,
4181 .o_cancel = osc_cancel,
4182 .o_cancel_unused = osc_cancel_unused,
4183 .o_iocontrol = osc_iocontrol,
4184 .o_get_info = osc_get_info,
4185 .o_set_info_async = osc_set_info_async,
4186 .o_import_event = osc_import_event,
4187 .o_llog_init = osc_llog_init,
4188 .o_llog_finish = osc_llog_finish,
4189 .o_process_config = osc_process_config,
4192 extern struct lu_kmem_descr osc_caches[];
4193 extern spinlock_t osc_ast_guard;
4194 extern struct lock_class_key osc_ast_guard_class;
4196 int __init osc_init(void)
4198 struct lprocfs_static_vars lvars = { 0 };
4202 /* print an address of _any_ initialized kernel symbol from this
4203 * module, to allow debugging with gdb that doesn't support data
4204 * symbols from modules.*/
4205 CDEBUG(D_CONSOLE, "Lustre OSC module (%p).\n", &osc_caches);
4207 rc = lu_kmem_init(osc_caches);
4209 lprocfs_osc_init_vars(&lvars);
4211 request_module("lquota");
4212 quota_interface = PORTAL_SYMBOL_GET(osc_quota_interface);
4213 lquota_init(quota_interface);
4214 init_obd_quota_ops(quota_interface, &osc_obd_ops);
4216 rc = class_register_type(&osc_obd_ops, NULL, lvars.module_vars,
4217 LUSTRE_OSC_NAME, &osc_device_type);
4219 if (quota_interface)
4220 PORTAL_SYMBOL_PUT(osc_quota_interface);
4221 lu_kmem_fini(osc_caches);
4225 spin_lock_init(&osc_ast_guard);
4226 lockdep_set_class(&osc_ast_guard, &osc_ast_guard_class);
4232 static void /*__exit*/ osc_exit(void)
4234 lu_device_type_fini(&osc_device_type);
4236 lquota_exit(quota_interface);
4237 if (quota_interface)
4238 PORTAL_SYMBOL_PUT(osc_quota_interface);
4240 class_unregister_type(LUSTRE_OSC_NAME);
4241 lu_kmem_fini(osc_caches);
4244 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
4245 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
4246 MODULE_LICENSE("GPL");
4248 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);
git://git.whamcloud.com / fs / lustre-release.git / blob