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
43 # include <libcfs/libcfs.h>
44 #else /* __KERNEL__ */
45 # include <liblustre.h>
48 # include <lustre_dlm.h>
49 #include <libcfs/kp30.h>
50 #include <lustre_net.h>
51 #include <lustre/lustre_user.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 <lustre_cache.h>
65 #include "osc_internal.h"
67 static quota_interface_t *quota_interface = NULL;
68 extern quota_interface_t osc_quota_interface;
70 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
71 static int brw_interpret(struct ptlrpc_request *request, void *data, int rc);
72 int osc_cleanup(struct obd_device *obd);
74 static quota_interface_t *quota_interface;
75 extern quota_interface_t osc_quota_interface;
78 atomic_t osc_resend_time;
80 /* Pack OSC object metadata for disk storage (LE byte order). */
81 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
82 struct lov_stripe_md *lsm)
87 lmm_size = sizeof(**lmmp);
92 OBD_FREE(*lmmp, lmm_size);
98 OBD_ALLOC(*lmmp, lmm_size);
104 LASSERT(lsm->lsm_object_id);
105 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
111 /* Unpack OSC object metadata from disk storage (LE byte order). */
112 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
113 struct lov_mds_md *lmm, int lmm_bytes)
119 if (lmm_bytes < sizeof (*lmm)) {
120 CERROR("lov_mds_md too small: %d, need %d\n",
121 lmm_bytes, (int)sizeof(*lmm));
124 /* XXX LOV_MAGIC etc check? */
126 if (lmm->lmm_object_id == 0) {
127 CERROR("lov_mds_md: zero lmm_object_id\n");
132 lsm_size = lov_stripe_md_size(1);
136 if (*lsmp != NULL && lmm == NULL) {
137 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
138 OBD_FREE(*lsmp, lsm_size);
144 OBD_ALLOC(*lsmp, lsm_size);
147 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
148 if ((*lsmp)->lsm_oinfo[0] == NULL) {
149 OBD_FREE(*lsmp, lsm_size);
152 loi_init((*lsmp)->lsm_oinfo[0]);
156 /* XXX zero *lsmp? */
157 (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
158 LASSERT((*lsmp)->lsm_object_id);
161 (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
166 static int osc_getattr_interpret(struct ptlrpc_request *req,
167 struct osc_async_args *aa, int rc)
169 struct ost_body *body;
175 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
176 lustre_swab_ost_body);
178 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
179 memcpy(aa->aa_oi->oi_oa, &body->oa, sizeof(*aa->aa_oi->oi_oa));
181 /* This should really be sent by the OST */
182 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
183 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
185 CERROR("can't unpack ost_body\n");
187 aa->aa_oi->oi_oa->o_valid = 0;
190 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
194 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
195 struct ptlrpc_request_set *set)
197 struct ptlrpc_request *req;
198 struct ost_body *body;
199 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*body) };
200 struct osc_async_args *aa;
203 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
204 OST_GETATTR, 2, size,NULL);
208 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
209 memcpy(&body->oa, oinfo->oi_oa, sizeof(*oinfo->oi_oa));
211 ptlrpc_req_set_repsize(req, 2, size);
212 req->rq_interpret_reply = osc_getattr_interpret;
214 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
215 aa = ptlrpc_req_async_args(req);
218 ptlrpc_set_add_req(set, req);
222 static int osc_getattr(struct obd_export *exp, struct obd_info *oinfo)
224 struct ptlrpc_request *req;
225 struct ost_body *body;
226 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*body) };
230 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
231 OST_GETATTR, 2, size, NULL);
235 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
236 memcpy(&body->oa, oinfo->oi_oa, sizeof(*oinfo->oi_oa));
238 ptlrpc_req_set_repsize(req, 2, size);
240 rc = ptlrpc_queue_wait(req);
242 CERROR("%s failed: rc = %d\n", __FUNCTION__, rc);
246 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
247 lustre_swab_ost_body);
249 CERROR ("can't unpack ost_body\n");
250 GOTO (out, rc = -EPROTO);
253 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
254 memcpy(oinfo->oi_oa, &body->oa, sizeof(*oinfo->oi_oa));
256 /* This should really be sent by the OST */
257 oinfo->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
258 oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
262 ptlrpc_req_finished(req);
266 static int osc_setattr(struct obd_export *exp, struct obd_info *oinfo,
267 struct obd_trans_info *oti)
269 struct ptlrpc_request *req;
270 struct ost_body *body;
271 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*body) };
275 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
276 OST_SETATTR, 2, size, NULL);
280 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
281 memcpy(&body->oa, oinfo->oi_oa, sizeof(*oinfo->oi_oa));
283 ptlrpc_req_set_repsize(req, 2, size);
285 rc = ptlrpc_queue_wait(req);
289 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
290 lustre_swab_ost_body);
292 GOTO(out, rc = -EPROTO);
294 memcpy(oinfo->oi_oa, &body->oa, sizeof(*oinfo->oi_oa));
298 ptlrpc_req_finished(req);
302 static int osc_setattr_interpret(struct ptlrpc_request *req,
303 struct osc_async_args *aa, int rc)
305 struct ost_body *body;
311 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
312 lustre_swab_ost_body);
314 CERROR("can't unpack ost_body\n");
315 GOTO(out, rc = -EPROTO);
318 memcpy(aa->aa_oi->oi_oa, &body->oa, sizeof(*aa->aa_oi->oi_oa));
320 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
324 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
325 struct obd_trans_info *oti,
326 struct ptlrpc_request_set *rqset)
328 struct ptlrpc_request *req;
329 struct ost_body *body;
330 __u32 size[3] = { sizeof(struct ptlrpc_body), sizeof(*body), 0 };
332 struct osc_async_args *aa;
335 if (osc_exp_is_2_0_server(exp)) {
339 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
340 OST_SETATTR, bufcount, size, NULL);
344 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
346 if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) {
348 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
351 memcpy(&body->oa, oinfo->oi_oa, sizeof(*oinfo->oi_oa));
352 ptlrpc_req_set_repsize(req, 2, size);
353 /* do mds to ost setattr asynchronouly */
355 /* Do not wait for response. */
356 ptlrpcd_add_req(req);
358 req->rq_interpret_reply = osc_setattr_interpret;
360 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
361 aa = ptlrpc_req_async_args(req);
364 ptlrpc_set_add_req(rqset, req);
370 int osc_real_create(struct obd_export *exp, struct obdo *oa,
371 struct lov_stripe_md **ea, struct obd_trans_info *oti)
373 struct ptlrpc_request *req;
374 struct ost_body *body;
375 struct lov_stripe_md *lsm;
376 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*body) };
385 rc = obd_alloc_memmd(exp, &lsm);
390 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
391 OST_CREATE, 2, size, NULL);
393 GOTO(out, rc = -ENOMEM);
395 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
396 memcpy(&body->oa, oa, sizeof(body->oa));
398 ptlrpc_req_set_repsize(req, 2, size);
399 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
400 oa->o_flags == OBD_FL_DELORPHAN) {
402 "delorphan from OST integration");
403 /* Don't resend the delorphan req */
404 req->rq_no_resend = req->rq_no_delay = 1;
407 rc = ptlrpc_queue_wait(req);
411 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
412 lustre_swab_ost_body);
414 CERROR ("can't unpack ost_body\n");
415 GOTO (out_req, rc = -EPROTO);
418 memcpy(oa, &body->oa, sizeof(*oa));
420 /* This should really be sent by the OST */
421 oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
422 oa->o_valid |= OBD_MD_FLBLKSZ;
424 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
425 * have valid lsm_oinfo data structs, so don't go touching that.
426 * This needs to be fixed in a big way.
428 lsm->lsm_object_id = oa->o_id;
432 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
434 if (oa->o_valid & OBD_MD_FLCOOKIE) {
435 if (!oti->oti_logcookies)
436 oti_alloc_cookies(oti, 1);
437 *oti->oti_logcookies = oa->o_lcookie;
441 CDEBUG(D_HA, "transno: "LPD64"\n",
442 lustre_msg_get_transno(req->rq_repmsg));
444 ptlrpc_req_finished(req);
447 obd_free_memmd(exp, &lsm);
451 static int osc_punch_interpret(struct ptlrpc_request *req,
452 struct osc_async_args *aa, int rc)
454 struct ost_body *body;
460 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof (*body),
461 lustre_swab_ost_body);
463 CERROR ("can't unpack ost_body\n");
464 GOTO(out, rc = -EPROTO);
467 memcpy(aa->aa_oi->oi_oa, &body->oa, sizeof(*aa->aa_oi->oi_oa));
469 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
473 static int osc_punch(struct obd_export *exp, struct obd_info *oinfo,
474 struct obd_trans_info *oti,
475 struct ptlrpc_request_set *rqset)
477 struct ptlrpc_request *req;
478 struct osc_async_args *aa;
479 struct ost_body *body;
480 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*body) };
488 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
489 OST_PUNCH, 2, size, NULL);
493 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
494 ptlrpc_at_set_req_timeout(req);
496 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
497 memcpy(&body->oa, oinfo->oi_oa, sizeof(*oinfo->oi_oa));
499 /* overload the size and blocks fields in the oa with start/end */
500 body->oa.o_size = oinfo->oi_policy.l_extent.start;
501 body->oa.o_blocks = oinfo->oi_policy.l_extent.end;
502 body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
504 ptlrpc_req_set_repsize(req, 2, size);
506 req->rq_interpret_reply = osc_punch_interpret;
507 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
508 aa = ptlrpc_req_async_args(req);
510 ptlrpc_set_add_req(rqset, req);
515 static int osc_sync_interpret(struct ptlrpc_request *req,
516 struct osc_async_args *aa, int rc)
518 struct ost_body *body;
524 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
525 lustre_swab_ost_body);
527 CERROR ("can't unpack ost_body\n");
528 GOTO(out, rc = -EPROTO);
531 *aa->aa_oi->oi_oa = body->oa;
533 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
537 static int osc_sync(struct obd_export *exp, struct obd_info *oinfo,
538 obd_size start, obd_size end,
539 struct ptlrpc_request_set *set)
541 struct ptlrpc_request *req;
542 struct ost_body *body;
543 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*body) };
544 struct osc_async_args *aa;
552 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
553 OST_SYNC, 2, size, NULL);
557 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
558 memcpy(&body->oa, oinfo->oi_oa, sizeof(*oinfo->oi_oa));
560 /* overload the size and blocks fields in the oa with start/end */
561 body->oa.o_size = start;
562 body->oa.o_blocks = end;
563 body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
565 ptlrpc_req_set_repsize(req, 2, size);
566 req->rq_interpret_reply = osc_sync_interpret;
568 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
569 aa = ptlrpc_req_async_args(req);
572 ptlrpc_set_add_req(set, req);
576 /* Find and cancel locally locks matched by @mode in the resource found by
577 * @objid. Found locks are added into @cancel list. Returns the amount of
578 * locks added to @cancels list. */
579 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
580 struct list_head *cancels, ldlm_mode_t mode,
583 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
584 struct ldlm_res_id res_id;
585 struct ldlm_resource *res;
589 osc_build_res_name(oa->o_id, oa->o_gr, &res_id);
590 res = ldlm_resource_get(ns, NULL, res_id, 0, 0);
594 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
595 lock_flags, 0, NULL);
596 ldlm_resource_putref(res);
600 static int osc_destroy_interpret(struct ptlrpc_request *req, void *data,
603 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
605 atomic_dec(&cli->cl_destroy_in_flight);
606 cfs_waitq_signal(&cli->cl_destroy_waitq);
610 static int osc_can_send_destroy(struct client_obd *cli)
612 if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
613 cli->cl_max_rpcs_in_flight) {
614 /* The destroy request can be sent */
617 if (atomic_dec_return(&cli->cl_destroy_in_flight) <
618 cli->cl_max_rpcs_in_flight) {
620 * The counter has been modified between the two atomic
623 cfs_waitq_signal(&cli->cl_destroy_waitq);
628 /* Destroy requests can be async always on the client, and we don't even really
629 * care about the return code since the client cannot do anything at all about
631 * When the MDS is unlinking a filename, it saves the file objects into a
632 * recovery llog, and these object records are cancelled when the OST reports
633 * they were destroyed and sync'd to disk (i.e. transaction committed).
634 * If the client dies, or the OST is down when the object should be destroyed,
635 * the records are not cancelled, and when the OST reconnects to the MDS next,
636 * it will retrieve the llog unlink logs and then sends the log cancellation
637 * cookies to the MDS after committing destroy transactions. */
638 static int osc_destroy(struct obd_export *exp, struct obdo *oa,
639 struct lov_stripe_md *ea, struct obd_trans_info *oti,
640 struct obd_export *md_export)
642 CFS_LIST_HEAD(cancels);
643 struct ptlrpc_request *req;
644 struct ost_body *body;
645 __u32 size[3] = { sizeof(struct ptlrpc_body), sizeof(*body),
646 sizeof(struct ldlm_request) };
647 int count, bufcount = 2;
648 struct client_obd *cli = &exp->exp_obd->u.cli;
656 LASSERT(oa->o_id != 0);
658 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
659 LDLM_FL_DISCARD_DATA);
660 if (exp_connect_cancelset(exp))
662 req = ldlm_prep_elc_req(exp, LUSTRE_OST_VERSION, OST_DESTROY, bufcount,
663 size, REQ_REC_OFF + 1, 0, &cancels, count);
667 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
668 ptlrpc_at_set_req_timeout(req);
670 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
672 if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE) {
673 oa->o_lcookie = *oti->oti_logcookies;
676 memcpy(&body->oa, oa, sizeof(*oa));
677 ptlrpc_req_set_repsize(req, 2, size);
679 /* don't throttle destroy RPCs for the MDT */
680 if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
681 req->rq_interpret_reply = osc_destroy_interpret;
682 if (!osc_can_send_destroy(cli)) {
683 struct l_wait_info lwi = { 0 };
686 * Wait until the number of on-going destroy RPCs drops
687 * under max_rpc_in_flight
689 l_wait_event_exclusive(cli->cl_destroy_waitq,
690 osc_can_send_destroy(cli), &lwi);
694 /* Do not wait for response */
695 ptlrpcd_add_req(req);
699 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
702 obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
704 LASSERT(!(oa->o_valid & bits));
707 client_obd_list_lock(&cli->cl_loi_list_lock);
708 oa->o_dirty = cli->cl_dirty;
709 if (cli->cl_dirty > cli->cl_dirty_max) {
710 CERROR("dirty %lu > dirty_max %lu\n",
711 cli->cl_dirty, cli->cl_dirty_max);
713 } else if (atomic_read(&obd_dirty_pages) > obd_max_dirty_pages) {
714 CERROR("dirty %d > system dirty_max %d\n",
715 atomic_read(&obd_dirty_pages), obd_max_dirty_pages);
717 } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
718 CERROR("dirty %lu - dirty_max %lu too big???\n",
719 cli->cl_dirty, cli->cl_dirty_max);
722 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
723 (cli->cl_max_rpcs_in_flight + 1);
724 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
726 oa->o_grant = cli->cl_avail_grant;
727 oa->o_dropped = cli->cl_lost_grant;
728 cli->cl_lost_grant = 0;
729 client_obd_list_unlock(&cli->cl_loi_list_lock);
730 CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
731 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
735 static void osc_update_next_shrink(struct client_obd *cli)
737 int time = GRANT_SHRINK_INTERVAL;
738 cli->cl_next_shrink_grant = cfs_time_shift(time);
739 CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
740 cli->cl_next_shrink_grant);
743 /* caller must hold loi_list_lock */
744 static void osc_consume_write_grant(struct client_obd *cli,struct brw_page *pga)
746 atomic_inc(&obd_dirty_pages);
747 cli->cl_dirty += CFS_PAGE_SIZE;
748 cli->cl_avail_grant -= CFS_PAGE_SIZE;
749 pga->flag |= OBD_BRW_FROM_GRANT;
750 CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
751 CFS_PAGE_SIZE, pga, pga->pg);
752 LASSERTF(cli->cl_avail_grant >= 0, "invalid avail grant is %ld \n",
753 cli->cl_avail_grant);
754 osc_update_next_shrink(cli);
757 /* the companion to osc_consume_write_grant, called when a brw has completed.
758 * must be called with the loi lock held. */
759 static void osc_release_write_grant(struct client_obd *cli,
760 struct brw_page *pga, int sent)
762 int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
765 if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
770 pga->flag &= ~OBD_BRW_FROM_GRANT;
771 atomic_dec(&obd_dirty_pages);
772 cli->cl_dirty -= CFS_PAGE_SIZE;
774 cli->cl_lost_grant += CFS_PAGE_SIZE;
775 CDEBUG(D_CACHE, "lost grant: %lu avail grant: %lu dirty: %lu\n",
776 cli->cl_lost_grant, cli->cl_avail_grant, cli->cl_dirty);
777 } else if (CFS_PAGE_SIZE != blocksize && pga->count != CFS_PAGE_SIZE) {
778 /* For short writes we shouldn't count parts of pages that
779 * span a whole block on the OST side, or our accounting goes
780 * wrong. Should match the code in filter_grant_check. */
781 int offset = pga->off & ~CFS_PAGE_MASK;
782 int count = pga->count + (offset & (blocksize - 1));
783 int end = (offset + pga->count) & (blocksize - 1);
785 count += blocksize - end;
787 cli->cl_lost_grant += CFS_PAGE_SIZE - count;
788 CDEBUG(D_CACHE, "lost %lu grant: %lu avail: %lu dirty: %lu\n",
789 CFS_PAGE_SIZE - count, cli->cl_lost_grant,
790 cli->cl_avail_grant, cli->cl_dirty);
796 static unsigned long rpcs_in_flight(struct client_obd *cli)
798 return cli->cl_r_in_flight + cli->cl_w_in_flight;
801 /* caller must hold loi_list_lock */
802 void osc_wake_cache_waiters(struct client_obd *cli)
804 struct list_head *l, *tmp;
805 struct osc_cache_waiter *ocw;
808 list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
809 /* if we can't dirty more, we must wait until some is written */
810 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
811 ((atomic_read(&obd_dirty_pages)+1)>(obd_max_dirty_pages))) {
812 CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
813 "osc max %ld, sys max %d\n", cli->cl_dirty,
814 cli->cl_dirty_max, obd_max_dirty_pages);
818 /* if still dirty cache but no grant wait for pending RPCs that
819 * may yet return us some grant before doing sync writes */
820 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
821 CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
822 cli->cl_w_in_flight);
826 ocw = list_entry(l, struct osc_cache_waiter, ocw_entry);
827 list_del_init(&ocw->ocw_entry);
828 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
829 /* no more RPCs in flight to return grant, do sync IO */
830 ocw->ocw_rc = -EDQUOT;
831 CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
833 osc_consume_write_grant(cli,
834 &ocw->ocw_oap->oap_brw_page);
837 cfs_waitq_signal(&ocw->ocw_waitq);
843 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
845 client_obd_list_lock(&cli->cl_loi_list_lock);
846 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
847 if (body->oa.o_valid & OBD_MD_FLGRANT)
848 cli->cl_avail_grant += body->oa.o_grant;
849 /* waiters are woken in brw_interpret */
850 client_obd_list_unlock(&cli->cl_loi_list_lock);
853 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
854 void *key, obd_count vallen, void *val,
855 struct ptlrpc_request_set *set);
857 static int osc_shrink_grant_interpret(struct ptlrpc_request *req,
858 struct osc_grant_args *aa, int rc)
860 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
861 struct obdo *oa = aa->aa_oa;
862 struct ost_body *body;
865 client_obd_list_lock(&cli->cl_loi_list_lock);
866 cli->cl_avail_grant += oa->o_grant;
867 client_obd_list_unlock(&cli->cl_loi_list_lock);
870 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*oa),
871 lustre_swab_ost_body);
872 osc_update_grant(cli, body);
878 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
880 client_obd_list_lock(&cli->cl_loi_list_lock);
881 oa->o_grant = cli->cl_avail_grant / 4;
882 cli->cl_avail_grant -= oa->o_grant;
883 client_obd_list_unlock(&cli->cl_loi_list_lock);
884 oa->o_flags |= OBD_FL_SHRINK_GRANT;
885 osc_update_next_shrink(cli);
888 static int osc_shrink_grant(struct client_obd *cli)
891 struct ost_body *body;
898 osc_announce_cached(cli, &body->oa, 0);
899 osc_shrink_grant_local(cli, &body->oa);
900 rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
901 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
902 sizeof(*body), body, NULL);
904 client_obd_list_lock(&cli->cl_loi_list_lock);
905 cli->cl_avail_grant += body->oa.o_grant;
906 client_obd_list_unlock(&cli->cl_loi_list_lock);
913 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
914 static int osc_should_shrink_grant(struct client_obd *client)
916 cfs_time_t time = cfs_time_current();
917 cfs_time_t next_shrink = client->cl_next_shrink_grant;
918 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
919 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
920 client->cl_avail_grant > GRANT_SHRINK_LIMIT)
923 osc_update_next_shrink(client);
928 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
930 struct client_obd *client;
932 list_for_each_entry(client, &item->ti_obd_list, cl_grant_shrink_list) {
933 if (osc_should_shrink_grant(client))
934 osc_shrink_grant(client);
939 static int osc_add_shrink_grant(struct client_obd *client)
943 rc = ptlrpc_add_timeout_client(GRANT_SHRINK_INTERVAL,
945 osc_grant_shrink_grant_cb, NULL,
946 &client->cl_grant_shrink_list);
948 CERROR("add grant client %s error %d\n",
949 client->cl_import->imp_obd->obd_name, rc);
952 CDEBUG(D_CACHE, "add grant client %s \n",
953 client->cl_import->imp_obd->obd_name);
954 osc_update_next_shrink(client);
958 static int osc_del_shrink_grant(struct client_obd *client)
960 CDEBUG(D_CACHE, "del grant client %s \n",
961 client->cl_import->imp_obd->obd_name);
962 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list);
965 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
967 client_obd_list_lock(&cli->cl_loi_list_lock);
968 cli->cl_avail_grant = ocd->ocd_grant;
969 client_obd_list_unlock(&cli->cl_loi_list_lock);
971 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
972 list_empty(&cli->cl_grant_shrink_list))
973 osc_add_shrink_grant(cli);
975 CDEBUG(D_CACHE, "setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
976 cli->cl_avail_grant, cli->cl_lost_grant);
977 LASSERT(cli->cl_avail_grant >= 0);
980 /* We assume that the reason this OSC got a short read is because it read
981 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
982 * via the LOV, and it _knows_ it's reading inside the file, it's just that
983 * this stripe never got written at or beyond this stripe offset yet. */
984 static void handle_short_read(int nob_read, obd_count page_count,
985 struct brw_page **pga)
990 /* skip bytes read OK */
991 while (nob_read > 0) {
992 LASSERT (page_count > 0);
994 if (pga[i]->count > nob_read) {
995 /* EOF inside this page */
996 ptr = cfs_kmap(pga[i]->pg) +
997 (pga[i]->off & ~CFS_PAGE_MASK);
998 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
999 cfs_kunmap(pga[i]->pg);
1005 nob_read -= pga[i]->count;
1010 /* zero remaining pages */
1011 while (page_count-- > 0) {
1012 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1013 memset(ptr, 0, pga[i]->count);
1014 cfs_kunmap(pga[i]->pg);
1019 static int check_write_rcs(struct ptlrpc_request *req,
1020 int requested_nob, int niocount,
1021 obd_count page_count, struct brw_page **pga)
1025 /* return error if any niobuf was in error */
1026 remote_rcs = lustre_swab_repbuf(req, REQ_REC_OFF + 1,
1027 sizeof(*remote_rcs) * niocount, NULL);
1028 if (remote_rcs == NULL) {
1029 CERROR("Missing/short RC vector on BRW_WRITE reply\n");
1032 if (lustre_rep_need_swab(req))
1033 for (i = 0; i < niocount; i++)
1034 __swab32s(&remote_rcs[i]);
1036 for (i = 0; i < niocount; i++) {
1037 if (remote_rcs[i] < 0)
1038 return(remote_rcs[i]);
1040 if (remote_rcs[i] != 0) {
1041 CERROR("rc[%d] invalid (%d) req %p\n",
1042 i, remote_rcs[i], req);
1047 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1048 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1049 req->rq_bulk->bd_nob_transferred, requested_nob);
1056 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1058 if (p1->flag != p2->flag) {
1059 unsigned mask = ~(OBD_BRW_FROM_GRANT | OBD_BRW_ASYNC);
1061 /* warn if we try to combine flags that we don't know to be
1062 * safe to combine */
1063 if ((p1->flag & mask) != (p2->flag & mask))
1064 CERROR("is it ok to have flags 0x%x and 0x%x in the "
1065 "same brw?\n", p1->flag, p2->flag);
1069 return (p1->off + p1->count == p2->off);
1072 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1073 struct brw_page **pga, int opc,
1074 cksum_type_t cksum_type)
1079 LASSERT (pg_count > 0);
1080 cksum = init_checksum(cksum_type);
1081 while (nob > 0 && pg_count > 0) {
1082 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1083 int off = pga[i]->off & ~CFS_PAGE_MASK;
1084 int count = pga[i]->count > nob ? nob : pga[i]->count;
1086 /* corrupt the data before we compute the checksum, to
1087 * simulate an OST->client data error */
1088 if (i == 0 && opc == OST_READ &&
1089 OBD_FAIL_CHECK_ONCE(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1090 memcpy(ptr + off, "bad1", min(4, nob));
1091 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1092 cfs_kunmap(pga[i]->pg);
1093 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1096 nob -= pga[i]->count;
1100 /* For sending we only compute the wrong checksum instead
1101 * of corrupting the data so it is still correct on a redo */
1102 if (opc == OST_WRITE && OBD_FAIL_CHECK_ONCE(OBD_FAIL_OSC_CHECKSUM_SEND))
1108 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1109 struct lov_stripe_md *lsm, obd_count page_count,
1110 struct brw_page **pga,
1111 struct ptlrpc_request **reqp)
1113 struct ptlrpc_request *req;
1114 struct ptlrpc_bulk_desc *desc;
1115 struct ost_body *body;
1116 struct obd_ioobj *ioobj;
1117 struct niobuf_remote *niobuf;
1118 __u32 size[4] = { sizeof(struct ptlrpc_body), sizeof(*body) };
1119 int niocount, i, requested_nob, opc, rc;
1120 struct ptlrpc_request_pool *pool;
1121 struct osc_brw_async_args *aa;
1122 struct brw_page *pg_prev;
1125 OBD_FAIL_RETURN(OBD_FAIL_OSC_BRW_PREP_REQ, -ENOMEM); /* Recoverable */
1126 OBD_FAIL_RETURN(OBD_FAIL_OSC_BRW_PREP_REQ2, -EINVAL); /* Fatal */
1128 opc = ((cmd & OBD_BRW_WRITE) != 0) ? OST_WRITE : OST_READ;
1129 pool = ((cmd & OBD_BRW_WRITE) != 0) ? cli->cl_import->imp_rq_pool :NULL;
1131 for (niocount = i = 1; i < page_count; i++) {
1132 if (!can_merge_pages(pga[i - 1], pga[i]))
1136 size[REQ_REC_OFF + 1] = sizeof(*ioobj);
1137 size[REQ_REC_OFF + 2] = niocount * sizeof(*niobuf);
1139 req = ptlrpc_prep_req_pool(cli->cl_import, LUSTRE_OST_VERSION, opc, 4, size,
1144 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1145 ptlrpc_at_set_req_timeout(req);
1147 if (opc == OST_WRITE)
1148 desc = ptlrpc_prep_bulk_imp (req, page_count,
1149 BULK_GET_SOURCE, OST_BULK_PORTAL);
1151 desc = ptlrpc_prep_bulk_imp (req, page_count,
1152 BULK_PUT_SINK, OST_BULK_PORTAL);
1154 GOTO(out, rc = -ENOMEM);
1155 /* NB request now owns desc and will free it when it gets freed */
1157 body = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF, sizeof(*body));
1158 ioobj = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF + 1, sizeof(*ioobj));
1159 niobuf = lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF + 2,
1160 niocount * sizeof(*niobuf));
1162 memcpy(&body->oa, oa, sizeof(*oa));
1164 obdo_to_ioobj(oa, ioobj);
1165 ioobj->ioo_bufcnt = niocount;
1167 LASSERT (page_count > 0);
1169 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1170 struct brw_page *pg = pga[i];
1172 LASSERT(pg->count > 0);
1173 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1174 "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1175 pg->off, pg->count);
1177 LASSERTF(i == 0 || pg->off > pg_prev->off,
1178 "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1179 " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1181 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1182 pg_prev->pg, page_private(pg_prev->pg),
1183 pg_prev->pg->index, pg_prev->off);
1185 LASSERTF(i == 0 || pg->off > pg_prev->off,
1186 "i %d p_c %u\n", i, page_count);
1188 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1189 (pg->flag & OBD_BRW_SRVLOCK));
1191 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1193 requested_nob += pg->count;
1195 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1197 niobuf->len += pg->count;
1199 niobuf->offset = pg->off;
1200 niobuf->len = pg->count;
1201 niobuf->flags = pg->flag;
1206 LASSERTF((void *)(niobuf - niocount) ==
1207 lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF + 2,
1208 niocount * sizeof(*niobuf)),
1209 "want %p - real %p\n", lustre_msg_buf(req->rq_reqmsg,
1210 REQ_REC_OFF + 2, niocount * sizeof(*niobuf)),
1211 (void *)(niobuf - niocount));
1213 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1214 if (osc_should_shrink_grant(cli))
1215 osc_shrink_grant_local(cli, &body->oa);
1217 /* size[REQ_REC_OFF] still sizeof (*body) */
1218 if (opc == OST_WRITE) {
1219 if (cli->cl_checksum) {
1220 /* store cl_cksum_type in a local variable since
1221 * it can be changed via lprocfs */
1222 cksum_type_t cksum_type = cli->cl_cksum_type;
1224 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1225 oa->o_flags = body->oa.o_flags = 0;
1226 body->oa.o_flags |= cksum_type_pack(cksum_type);
1227 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1228 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1232 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1234 /* save this in 'oa', too, for later checking */
1235 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1236 oa->o_flags |= cksum_type_pack(cksum_type);
1238 /* clear out the checksum flag, in case this is a
1239 * resend but cl_checksum is no longer set. b=11238 */
1240 oa->o_valid &= ~OBD_MD_FLCKSUM;
1242 oa->o_cksum = body->oa.o_cksum;
1243 /* 1 RC per niobuf */
1244 size[REPLY_REC_OFF + 1] = sizeof(__u32) * niocount;
1245 ptlrpc_req_set_repsize(req, 3, size);
1247 if (cli->cl_checksum) {
1248 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1249 body->oa.o_flags = 0;
1250 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1251 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1253 /* 1 RC for the whole I/O */
1254 ptlrpc_req_set_repsize(req, 2, size);
1257 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1258 aa = ptlrpc_req_async_args(req);
1260 aa->aa_requested_nob = requested_nob;
1261 aa->aa_nio_count = niocount;
1262 aa->aa_page_count = page_count;
1266 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1272 ptlrpc_req_finished (req);
1276 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1277 __u32 client_cksum, __u32 server_cksum, int nob,
1278 obd_count page_count, struct brw_page **pga,
1279 cksum_type_t client_cksum_type)
1283 cksum_type_t cksum_type;
1285 if (server_cksum == client_cksum) {
1286 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1290 if (oa->o_valid & OBD_MD_FLFLAGS)
1291 cksum_type = cksum_type_unpack(oa->o_flags);
1293 cksum_type = OBD_CKSUM_CRC32;
1295 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1298 if (cksum_type != client_cksum_type)
1299 msg = "the server did not use the checksum type specified in "
1300 "the original request - likely a protocol problem";
1301 else if (new_cksum == server_cksum)
1302 msg = "changed on the client after we checksummed it - "
1303 "likely false positive due to mmap IO (bug 11742)";
1304 else if (new_cksum == client_cksum)
1305 msg = "changed in transit before arrival at OST";
1307 msg = "changed in transit AND doesn't match the original - "
1308 "likely false positive due to mmap IO (bug 11742)";
1310 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inum "
1311 LPU64"/"LPU64" object "LPU64"/"LPU64" extent "
1312 "["LPU64"-"LPU64"]\n",
1313 msg, libcfs_nid2str(peer->nid),
1314 oa->o_valid & OBD_MD_FLFID ? oa->o_fid : (__u64)0,
1315 oa->o_valid & OBD_MD_FLFID ? oa->o_generation :
1318 oa->o_valid & OBD_MD_FLGROUP ? oa->o_gr : (__u64)0,
1320 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1321 CERROR("original client csum %x (type %x), server csum %x (type %x), "
1322 "client csum now %x\n", client_cksum, client_cksum_type,
1323 server_cksum, cksum_type, new_cksum);
1328 /* Note rc enters this function as number of bytes transferred */
1329 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1331 struct osc_brw_async_args *aa = ptlrpc_req_async_args(req);
1332 const lnet_process_id_t *peer =
1333 &req->rq_import->imp_connection->c_peer;
1334 struct client_obd *cli = aa->aa_cli;
1335 struct ost_body *body;
1336 __u32 client_cksum = 0;
1339 if (rc < 0 && rc != -EDQUOT)
1342 LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1343 body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
1344 lustre_swab_ost_body);
1346 CERROR ("Can't unpack body\n");
1350 /* set/clear over quota flag for a uid/gid */
1351 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1352 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA))
1353 lquota_setdq(quota_interface, cli, body->oa.o_uid,
1354 body->oa.o_gid, body->oa.o_valid,
1360 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1361 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1363 osc_update_grant(cli, body);
1365 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1367 CERROR ("Unexpected +ve rc %d\n", rc);
1370 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1372 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1373 check_write_checksum(&body->oa, peer, client_cksum,
1374 body->oa.o_cksum, aa->aa_requested_nob,
1375 aa->aa_page_count, aa->aa_ppga,
1376 cksum_type_unpack(aa->aa_oa->o_flags)))
1379 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1380 aa->aa_page_count, aa->aa_ppga);
1384 /* The rest of this function executes only for OST_READs */
1385 if (rc > aa->aa_requested_nob) {
1386 CERROR("Unexpected rc %d (%d requested)\n", rc,
1387 aa->aa_requested_nob);
1391 if (rc != req->rq_bulk->bd_nob_transferred) {
1392 CERROR ("Unexpected rc %d (%d transferred)\n",
1393 rc, req->rq_bulk->bd_nob_transferred);
1397 if (rc < aa->aa_requested_nob)
1398 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1400 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1401 static int cksum_counter;
1402 __u32 server_cksum = body->oa.o_cksum;
1405 cksum_type_t cksum_type;
1407 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1408 cksum_type = cksum_type_unpack(body->oa.o_flags);
1410 cksum_type = OBD_CKSUM_CRC32;
1411 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1412 aa->aa_ppga, OST_READ,
1415 if (peer->nid == req->rq_bulk->bd_sender) {
1419 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1422 if (server_cksum == ~0 && rc > 0) {
1423 CERROR("Protocol error: server %s set the 'checksum' "
1424 "bit, but didn't send a checksum. Not fatal, "
1425 "but please notify on http://bugzilla.lustre.org/\n",
1426 libcfs_nid2str(peer->nid));
1427 } else if (server_cksum != client_cksum) {
1428 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1429 "%s%s%s inum "LPU64"/"LPU64" object "
1430 LPU64"/"LPU64" extent "
1431 "["LPU64"-"LPU64"]\n",
1432 req->rq_import->imp_obd->obd_name,
1433 libcfs_nid2str(peer->nid),
1435 body->oa.o_valid & OBD_MD_FLFID ?
1436 body->oa.o_fid : (__u64)0,
1437 body->oa.o_valid & OBD_MD_FLFID ?
1438 body->oa.o_generation :(__u64)0,
1440 body->oa.o_valid & OBD_MD_FLGROUP ?
1441 body->oa.o_gr : (__u64)0,
1442 aa->aa_ppga[0]->off,
1443 aa->aa_ppga[aa->aa_page_count-1]->off +
1444 aa->aa_ppga[aa->aa_page_count-1]->count -
1446 CERROR("client %x, server %x, cksum_type %x\n",
1447 client_cksum, server_cksum, cksum_type);
1449 aa->aa_oa->o_cksum = client_cksum;
1453 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1456 } else if (unlikely(client_cksum)) {
1457 static int cksum_missed;
1460 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1461 CERROR("Checksum %u requested from %s but not sent\n",
1462 cksum_missed, libcfs_nid2str(peer->nid));
1468 memcpy(aa->aa_oa, &body->oa, sizeof(*aa->aa_oa));
1473 static int osc_brw_internal(int cmd, struct obd_export *exp,struct obdo *oa,
1474 struct lov_stripe_md *lsm,
1475 obd_count page_count, struct brw_page **pga)
1477 struct ptlrpc_request *request;
1481 struct l_wait_info lwi;
1484 init_waitqueue_head(&waitq);
1487 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1488 page_count, pga, &request);
1492 rc = ptlrpc_queue_wait(request);
1494 if (rc == -ETIMEDOUT && request->rq_resend) {
1495 DEBUG_REQ(D_HA, request, "BULK TIMEOUT");
1496 ptlrpc_req_finished(request);
1500 rc = osc_brw_fini_request(request, rc);
1502 ptlrpc_req_finished(request);
1503 if (osc_recoverable_error(rc)) {
1505 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1506 CERROR("too many resend retries, returning error\n");
1510 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1511 l_wait_event(waitq, 0, &lwi);
1518 int osc_brw_redo_request(struct ptlrpc_request *request,
1519 struct osc_brw_async_args *aa)
1521 struct ptlrpc_request *new_req;
1522 struct ptlrpc_request_set *set = request->rq_set;
1523 struct osc_brw_async_args *new_aa;
1524 struct osc_async_page *oap;
1528 if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1529 CERROR("too many resend retries, returning error\n");
1533 DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1535 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1536 OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1537 aa->aa_cli, aa->aa_oa,
1538 NULL /* lsm unused by osc currently */,
1539 aa->aa_page_count, aa->aa_ppga, &new_req);
1543 client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1545 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1546 if (oap->oap_request != NULL) {
1547 LASSERTF(request == oap->oap_request,
1548 "request %p != oap_request %p\n",
1549 request, oap->oap_request);
1550 if (oap->oap_interrupted) {
1551 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1552 ptlrpc_req_finished(new_req);
1557 /* New request takes over pga and oaps from old request.
1558 * Note that copying a list_head doesn't work, need to move it... */
1560 new_req->rq_interpret_reply = request->rq_interpret_reply;
1561 new_req->rq_async_args = request->rq_async_args;
1562 new_req->rq_sent = CURRENT_SECONDS + aa->aa_resends;
1564 new_aa = ptlrpc_req_async_args(new_req);
1566 CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1567 list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1568 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1570 list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1571 if (oap->oap_request) {
1572 ptlrpc_req_finished(oap->oap_request);
1573 oap->oap_request = ptlrpc_request_addref(new_req);
1577 /* use ptlrpc_set_add_req is safe because interpret functions work
1578 * in check_set context. only one way exist with access to request
1579 * from different thread got -EINTR - this way protected with
1580 * cl_loi_list_lock */
1581 ptlrpc_set_add_req(set, new_req);
1583 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1585 DEBUG_REQ(D_INFO, new_req, "new request");
1589 static int async_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1590 struct lov_stripe_md *lsm, obd_count page_count,
1591 struct brw_page **pga, struct ptlrpc_request_set *set)
1593 struct ptlrpc_request *request;
1594 struct client_obd *cli = &exp->exp_obd->u.cli;
1596 struct osc_brw_async_args *aa;
1599 /* Consume write credits even if doing a sync write -
1600 * otherwise we may run out of space on OST due to grant. */
1601 if (cmd == OBD_BRW_WRITE) {
1602 client_obd_list_lock(&cli->cl_loi_list_lock);
1603 for (i = 0; i < page_count; i++) {
1604 if (cli->cl_avail_grant >= CFS_PAGE_SIZE)
1605 osc_consume_write_grant(cli, pga[i]);
1607 client_obd_list_unlock(&cli->cl_loi_list_lock);
1610 rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1611 page_count, pga, &request);
1613 CLASSERT(sizeof(*aa) <= sizeof(request->rq_async_args));
1614 aa = ptlrpc_req_async_args(request);
1615 if (cmd == OBD_BRW_READ) {
1616 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
1617 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
1619 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
1620 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
1621 cli->cl_w_in_flight);
1623 ptlrpc_lprocfs_brw(request, aa->aa_requested_nob);
1625 LASSERT(list_empty(&aa->aa_oaps));
1628 request->rq_interpret_reply = brw_interpret;
1629 ptlrpc_set_add_req(set, request);
1630 client_obd_list_lock(&cli->cl_loi_list_lock);
1631 if (cmd == OBD_BRW_READ)
1632 cli->cl_r_in_flight++;
1634 cli->cl_w_in_flight++;
1635 client_obd_list_unlock(&cli->cl_loi_list_lock);
1636 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_DIO_PAUSE, 3);
1637 } else if (cmd == OBD_BRW_WRITE) {
1638 client_obd_list_lock(&cli->cl_loi_list_lock);
1639 for (i = 0; i < page_count; i++)
1640 osc_release_write_grant(cli, pga[i], 0);
1641 osc_wake_cache_waiters(cli);
1642 client_obd_list_unlock(&cli->cl_loi_list_lock);
1649 * ugh, we want disk allocation on the target to happen in offset order. we'll
1650 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1651 * fine for our small page arrays and doesn't require allocation. its an
1652 * insertion sort that swaps elements that are strides apart, shrinking the
1653 * stride down until its '1' and the array is sorted.
1655 static void sort_brw_pages(struct brw_page **array, int num)
1658 struct brw_page *tmp;
1662 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1667 for (i = stride ; i < num ; i++) {
1670 while (j >= stride && array[j-stride]->off > tmp->off) {
1671 array[j] = array[j - stride];
1676 } while (stride > 1);
1679 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1685 LASSERT (pages > 0);
1686 offset = pg[i]->off & (~CFS_PAGE_MASK);
1690 if (pages == 0) /* that's all */
1693 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1694 return count; /* doesn't end on page boundary */
1697 offset = pg[i]->off & (~CFS_PAGE_MASK);
1698 if (offset != 0) /* doesn't start on page boundary */
1705 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1707 struct brw_page **ppga;
1710 OBD_ALLOC(ppga, sizeof(*ppga) * count);
1714 for (i = 0; i < count; i++)
1719 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1721 LASSERT(ppga != NULL);
1722 OBD_FREE(ppga, sizeof(*ppga) * count);
1725 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1726 obd_count page_count, struct brw_page *pga,
1727 struct obd_trans_info *oti)
1729 struct obdo *saved_oa = NULL;
1730 struct brw_page **ppga, **orig;
1731 struct obd_import *imp = class_exp2cliimp(exp);
1732 struct client_obd *cli = &imp->imp_obd->u.cli;
1733 int rc, page_count_orig;
1736 if (cmd & OBD_BRW_CHECK) {
1737 /* The caller just wants to know if there's a chance that this
1738 * I/O can succeed */
1740 if (imp == NULL || imp->imp_invalid)
1745 /* test_brw with a failed create can trip this, maybe others. */
1746 LASSERT(cli->cl_max_pages_per_rpc);
1750 orig = ppga = osc_build_ppga(pga, page_count);
1753 page_count_orig = page_count;
1755 sort_brw_pages(ppga, page_count);
1756 while (page_count) {
1757 obd_count pages_per_brw;
1759 if (page_count > cli->cl_max_pages_per_rpc)
1760 pages_per_brw = cli->cl_max_pages_per_rpc;
1762 pages_per_brw = page_count;
1764 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1766 if (saved_oa != NULL) {
1767 /* restore previously saved oa */
1768 *oinfo->oi_oa = *saved_oa;
1769 } else if (page_count > pages_per_brw) {
1770 /* save a copy of oa (brw will clobber it) */
1771 OBDO_ALLOC(saved_oa);
1772 if (saved_oa == NULL)
1773 GOTO(out, rc = -ENOMEM);
1774 *saved_oa = *oinfo->oi_oa;
1777 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1778 pages_per_brw, ppga);
1783 page_count -= pages_per_brw;
1784 ppga += pages_per_brw;
1788 osc_release_ppga(orig, page_count_orig);
1790 if (saved_oa != NULL)
1791 OBDO_FREE(saved_oa);
1796 static int osc_brw_async(int cmd, struct obd_export *exp,
1797 struct obd_info *oinfo, obd_count page_count,
1798 struct brw_page *pga, struct obd_trans_info *oti,
1799 struct ptlrpc_request_set *set)
1801 struct brw_page **ppga, **orig;
1802 int page_count_orig;
1806 if (cmd & OBD_BRW_CHECK) {
1807 /* The caller just wants to know if there's a chance that this
1808 * I/O can succeed */
1809 struct obd_import *imp = class_exp2cliimp(exp);
1811 if (imp == NULL || imp->imp_invalid)
1816 orig = ppga = osc_build_ppga(pga, page_count);
1819 page_count_orig = page_count;
1821 sort_brw_pages(ppga, page_count);
1822 while (page_count) {
1823 struct brw_page **copy;
1824 obd_count pages_per_brw;
1826 pages_per_brw = min_t(obd_count, page_count,
1827 class_exp2cliimp(exp)->imp_obd->u.cli.cl_max_pages_per_rpc);
1829 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1831 /* use ppga only if single RPC is going to fly */
1832 if (pages_per_brw != page_count_orig || ppga != orig) {
1833 OBD_ALLOC(copy, pages_per_brw * sizeof(*copy));
1835 GOTO(out, rc = -ENOMEM);
1836 memcpy(copy, ppga, pages_per_brw * sizeof(*copy));
1840 rc = async_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1841 pages_per_brw, copy, set);
1845 OBD_FREE(copy, pages_per_brw * sizeof(*copy));
1850 /* we passed it to async_internal() which is
1851 * now responsible for releasing memory */
1855 page_count -= pages_per_brw;
1856 ppga += pages_per_brw;
1860 osc_release_ppga(orig, page_count_orig);
1864 static void osc_check_rpcs(struct client_obd *cli);
1866 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1867 * the dirty accounting. Writeback completes or truncate happens before
1868 * writing starts. Must be called with the loi lock held. */
1869 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1872 osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1875 /* This maintains the lists of pending pages to read/write for a given object
1876 * (lop). This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1877 * to quickly find objects that are ready to send an RPC. */
1878 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1884 if (lop->lop_num_pending == 0)
1887 /* if we have an invalid import we want to drain the queued pages
1888 * by forcing them through rpcs that immediately fail and complete
1889 * the pages. recovery relies on this to empty the queued pages
1890 * before canceling the locks and evicting down the llite pages */
1891 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1894 /* stream rpcs in queue order as long as as there is an urgent page
1895 * queued. this is our cheap solution for good batching in the case
1896 * where writepage marks some random page in the middle of the file
1897 * as urgent because of, say, memory pressure */
1898 if (!list_empty(&lop->lop_urgent)) {
1899 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1903 /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1904 optimal = cli->cl_max_pages_per_rpc;
1905 if (cmd & OBD_BRW_WRITE) {
1906 /* trigger a write rpc stream as long as there are dirtiers
1907 * waiting for space. as they're waiting, they're not going to
1908 * create more pages to coallesce with what's waiting.. */
1909 if (!list_empty(&cli->cl_cache_waiters)) {
1910 CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1914 /* +16 to avoid triggering rpcs that would want to include pages
1915 * that are being queued but which can't be made ready until
1916 * the queuer finishes with the page. this is a wart for
1917 * llite::commit_write() */
1920 if (lop->lop_num_pending >= optimal)
1926 static int lop_makes_hprpc(struct loi_oap_pages *lop)
1928 struct osc_async_page *oap;
1931 if (list_empty(&lop->lop_urgent))
1934 oap = list_entry(lop->lop_urgent.next,
1935 struct osc_async_page, oap_urgent_item);
1937 if (oap->oap_async_flags & ASYNC_HP) {
1938 CDEBUG(D_CACHE, "hp request forcing RPC\n");
1945 static void on_list(struct list_head *item, struct list_head *list,
1948 if (list_empty(item) && should_be_on)
1949 list_add_tail(item, list);
1950 else if (!list_empty(item) && !should_be_on)
1951 list_del_init(item);
1954 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
1955 * can find pages to build into rpcs quickly */
1956 static void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
1958 if (lop_makes_hprpc(&loi->loi_write_lop) ||
1959 lop_makes_hprpc(&loi->loi_read_lop)) {
1961 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list, 0);
1962 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
1964 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 0);
1965 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list,
1966 lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)||
1967 lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ));
1970 on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
1971 loi->loi_write_lop.lop_num_pending);
1973 on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
1974 loi->loi_read_lop.lop_num_pending);
1977 static void lop_update_pending(struct client_obd *cli,
1978 struct loi_oap_pages *lop, int cmd, int delta)
1980 lop->lop_num_pending += delta;
1981 if (cmd & OBD_BRW_WRITE)
1982 cli->cl_pending_w_pages += delta;
1984 cli->cl_pending_r_pages += delta;
1987 /* this is called when a sync waiter receives an interruption. Its job is to
1988 * get the caller woken as soon as possible. If its page hasn't been put in an
1989 * rpc yet it can dequeue immediately. Otherwise it has to mark the rpc as
1990 * desiring interruption which will forcefully complete the rpc once the rpc
1992 static void osc_occ_interrupted(struct oig_callback_context *occ)
1994 struct osc_async_page *oap;
1995 struct loi_oap_pages *lop;
1996 struct lov_oinfo *loi;
1999 /* XXX member_of() */
2000 oap = list_entry(occ, struct osc_async_page, oap_occ);
2002 client_obd_list_lock(&oap->oap_cli->cl_loi_list_lock);
2004 oap->oap_interrupted = 1;
2006 /* ok, it's been put in an rpc. only one oap gets a request reference */
2007 if (oap->oap_request != NULL) {
2008 ptlrpc_mark_interrupted(oap->oap_request);
2009 ptlrpcd_wake(oap->oap_request);
2013 /* we don't get interruption callbacks until osc_trigger_group_io()
2014 * has been called and put the sync oaps in the pending/urgent lists.*/
2015 if (!list_empty(&oap->oap_pending_item)) {
2016 list_del_init(&oap->oap_pending_item);
2017 list_del_init(&oap->oap_urgent_item);
2020 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
2021 &loi->loi_write_lop : &loi->loi_read_lop;
2022 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
2023 loi_list_maint(oap->oap_cli, oap->oap_loi);
2025 oig_complete_one(oap->oap_oig, &oap->oap_occ, -EINTR);
2026 oap->oap_oig = NULL;
2030 client_obd_list_unlock(&oap->oap_cli->cl_loi_list_lock);
2033 /* this is trying to propogate async writeback errors back up to the
2034 * application. As an async write fails we record the error code for later if
2035 * the app does an fsync. As long as errors persist we force future rpcs to be
2036 * sync so that the app can get a sync error and break the cycle of queueing
2037 * pages for which writeback will fail. */
2038 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
2045 ar->ar_force_sync = 1;
2046 ar->ar_min_xid = ptlrpc_sample_next_xid();
2051 if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2052 ar->ar_force_sync = 0;
2055 static void osc_oap_to_pending(struct osc_async_page *oap)
2057 struct loi_oap_pages *lop;
2059 if (oap->oap_cmd & OBD_BRW_WRITE)
2060 lop = &oap->oap_loi->loi_write_lop;
2062 lop = &oap->oap_loi->loi_read_lop;
2064 if (oap->oap_async_flags & ASYNC_HP)
2065 list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2066 else if (oap->oap_async_flags & ASYNC_URGENT)
2067 list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2068 list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2069 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2072 /* this must be called holding the loi list lock to give coverage to exit_cache,
2073 * async_flag maintenance, and oap_request */
2074 static void osc_ap_completion(struct client_obd *cli, struct obdo *oa,
2075 struct osc_async_page *oap, int sent, int rc)
2080 if (oap->oap_request != NULL) {
2081 xid = ptlrpc_req_xid(oap->oap_request);
2082 ptlrpc_req_finished(oap->oap_request);
2083 oap->oap_request = NULL;
2086 oap->oap_async_flags = 0;
2087 oap->oap_interrupted = 0;
2089 if (oap->oap_cmd & OBD_BRW_WRITE) {
2090 osc_process_ar(&cli->cl_ar, xid, rc);
2091 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2094 if (rc == 0 && oa != NULL) {
2095 if (oa->o_valid & OBD_MD_FLBLOCKS)
2096 oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2097 if (oa->o_valid & OBD_MD_FLMTIME)
2098 oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2099 if (oa->o_valid & OBD_MD_FLATIME)
2100 oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2101 if (oa->o_valid & OBD_MD_FLCTIME)
2102 oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2106 osc_exit_cache(cli, oap, sent);
2107 oig_complete_one(oap->oap_oig, &oap->oap_occ, rc);
2108 oap->oap_oig = NULL;
2113 rc = oap->oap_caller_ops->ap_completion(oap->oap_caller_data,
2114 oap->oap_cmd, oa, rc);
2116 /* ll_ap_completion (from llite) drops PG_locked. so, a new
2117 * I/O on the page could start, but OSC calls it under lock
2118 * and thus we can add oap back to pending safely */
2120 /* upper layer wants to leave the page on pending queue */
2121 osc_oap_to_pending(oap);
2123 osc_exit_cache(cli, oap, sent);
2127 static int brw_interpret(struct ptlrpc_request *request, void *data, int rc)
2129 struct osc_brw_async_args *aa = data;
2130 struct client_obd *cli;
2133 rc = osc_brw_fini_request(request, rc);
2134 CDEBUG(D_INODE, "request %p aa %p rc %d\n", request, aa, rc);
2136 if (osc_recoverable_error(rc)) {
2137 rc = osc_brw_redo_request(request, aa);
2143 client_obd_list_lock(&cli->cl_loi_list_lock);
2144 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2145 * is called so we know whether to go to sync BRWs or wait for more
2146 * RPCs to complete */
2147 if (lustre_msg_get_opc(request->rq_reqmsg) == OST_WRITE)
2148 cli->cl_w_in_flight--;
2150 cli->cl_r_in_flight--;
2152 if (!list_empty(&aa->aa_oaps)) { /* from osc_send_oap_rpc() */
2153 struct osc_async_page *oap, *tmp;
2154 /* the caller may re-use the oap after the completion call so
2155 * we need to clean it up a little */
2156 list_for_each_entry_safe(oap, tmp, &aa->aa_oaps, oap_rpc_item) {
2157 list_del_init(&oap->oap_rpc_item);
2158 osc_ap_completion(cli, aa->aa_oa, oap, 1, rc);
2160 OBDO_FREE(aa->aa_oa);
2161 } else { /* from async_internal() */
2163 for (i = 0; i < aa->aa_page_count; i++)
2164 osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2166 osc_wake_cache_waiters(cli);
2167 osc_check_rpcs(cli);
2168 client_obd_list_unlock(&cli->cl_loi_list_lock);
2170 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2175 static struct ptlrpc_request *osc_build_req(struct client_obd *cli,
2176 struct list_head *rpc_list,
2177 int page_count, int cmd)
2179 struct ptlrpc_request *req;
2180 struct brw_page **pga = NULL;
2181 struct osc_brw_async_args *aa;
2182 struct obdo *oa = NULL;
2183 struct obd_async_page_ops *ops = NULL;
2184 void *caller_data = NULL;
2185 struct osc_async_page *oap;
2186 struct ldlm_lock *lock = NULL;
2191 LASSERT(!list_empty(rpc_list));
2193 OBD_ALLOC(pga, sizeof(*pga) * page_count);
2195 RETURN(ERR_PTR(-ENOMEM));
2199 GOTO(out, req = ERR_PTR(-ENOMEM));
2202 list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2204 ops = oap->oap_caller_ops;
2205 caller_data = oap->oap_caller_data;
2206 lock = oap->oap_ldlm_lock;
2208 pga[i] = &oap->oap_brw_page;
2209 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2210 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2211 pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2215 /* always get the data for the obdo for the rpc */
2216 LASSERT(ops != NULL);
2217 ops->ap_fill_obdo(caller_data, cmd, oa);
2219 oa->o_handle = lock->l_remote_handle;
2220 oa->o_valid |= OBD_MD_FLHANDLE;
2223 sort_brw_pages(pga, page_count);
2224 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count, pga, &req);
2226 CERROR("prep_req failed: %d\n", rc);
2227 GOTO(out, req = ERR_PTR(rc));
2229 oa = &((struct ost_body *)lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF,
2230 sizeof(struct ost_body)))->oa;
2232 /* Need to update the timestamps after the request is built in case
2233 * we race with setattr (locally or in queue at OST). If OST gets
2234 * later setattr before earlier BRW (as determined by the request xid),
2235 * the OST will not use BRW timestamps. Sadly, there is no obvious
2236 * way to do this in a single call. bug 10150 */
2237 if (pga[0]->flag & OBD_BRW_SRVLOCK) {
2238 /* in case of lockless read/write do not use inode's
2239 * timestamps because concurrent stat might fill the
2240 * inode with out-of-date times, send current
2242 if (cmd & OBD_BRW_WRITE) {
2243 oa->o_mtime = oa->o_ctime = LTIME_S(CURRENT_TIME);
2244 oa->o_valid |= OBD_MD_FLMTIME | OBD_MD_FLCTIME;
2245 valid = OBD_MD_FLATIME;
2247 oa->o_atime = LTIME_S(CURRENT_TIME);
2248 oa->o_valid |= OBD_MD_FLATIME;
2249 valid = OBD_MD_FLMTIME | OBD_MD_FLCTIME;
2252 valid = OBD_MD_FLMTIME | OBD_MD_FLCTIME | OBD_MD_FLATIME;
2254 ops->ap_update_obdo(caller_data, cmd, oa, valid);
2256 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2257 aa = ptlrpc_req_async_args(req);
2258 CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2259 list_splice(rpc_list, &aa->aa_oaps);
2260 CFS_INIT_LIST_HEAD(rpc_list);
2267 OBD_FREE(pga, sizeof(*pga) * page_count);
2272 /* the loi lock is held across this function but it's allowed to release
2273 * and reacquire it during its work */
2275 * prepare pages for ASYNC io and put pages in send queue.
2279 * \param cmd - OBD_BRW_* macroses
2280 * \param lop - pending pages
2282 * \return zero if pages successfully add to send queue.
2283 * \return not zere if error occurring.
2285 static int osc_send_oap_rpc(struct client_obd *cli, struct lov_oinfo *loi,
2286 int cmd, struct loi_oap_pages *lop)
2288 struct ptlrpc_request *req;
2289 obd_count page_count = 0;
2290 struct osc_async_page *oap = NULL, *tmp;
2291 struct osc_brw_async_args *aa;
2292 struct obd_async_page_ops *ops;
2293 CFS_LIST_HEAD(rpc_list);
2294 unsigned int ending_offset;
2295 unsigned starting_offset = 0;
2299 /* If there are HP OAPs we need to handle at least 1 of them,
2300 * move it the beginning of the pending list for that. */
2301 if (!list_empty(&lop->lop_urgent)) {
2302 oap = list_entry(lop->lop_urgent.next,
2303 struct osc_async_page, oap_urgent_item);
2304 if (oap->oap_async_flags & ASYNC_HP)
2305 list_move(&oap->oap_pending_item, &lop->lop_pending);
2308 /* first we find the pages we're allowed to work with */
2309 list_for_each_entry_safe(oap, tmp, &lop->lop_pending, oap_pending_item){
2310 ops = oap->oap_caller_ops;
2312 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2313 "magic 0x%x\n", oap, oap->oap_magic);
2315 if (page_count != 0 &&
2316 srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2317 CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2318 " oap %p, page %p, srvlock %u\n",
2319 oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2322 /* in llite being 'ready' equates to the page being locked
2323 * until completion unlocks it. commit_write submits a page
2324 * as not ready because its unlock will happen unconditionally
2325 * as the call returns. if we race with commit_write giving
2326 * us that page we dont' want to create a hole in the page
2327 * stream, so we stop and leave the rpc to be fired by
2328 * another dirtier or kupdated interval (the not ready page
2329 * will still be on the dirty list). we could call in
2330 * at the end of ll_file_write to process the queue again. */
2331 if (!(oap->oap_async_flags & ASYNC_READY)) {
2332 int rc = ops->ap_make_ready(oap->oap_caller_data, cmd);
2334 CDEBUG(D_INODE, "oap %p page %p returned %d "
2335 "instead of ready\n", oap,
2339 /* llite is telling us that the page is still
2340 * in commit_write and that we should try
2341 * and put it in an rpc again later. we
2342 * break out of the loop so we don't create
2343 * a hole in the sequence of pages in the rpc
2348 /* the io isn't needed.. tell the checks
2349 * below to complete the rpc with EINTR */
2350 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2351 oap->oap_count = -EINTR;
2354 oap->oap_async_flags |= ASYNC_READY;
2357 LASSERTF(0, "oap %p page %p returned %d "
2358 "from make_ready\n", oap,
2366 * Page submitted for IO has to be locked. Either by
2367 * ->ap_make_ready() or by higher layers.
2369 #if defined(__KERNEL__) && defined(__linux__)
2370 if(!(PageLocked(oap->oap_page) &&
2371 (CheckWriteback(oap->oap_page, cmd) || oap->oap_oig !=NULL))) {
2372 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2373 oap->oap_page, (long)oap->oap_page->flags, oap->oap_async_flags);
2377 /* If there is a gap at the start of this page, it can't merge
2378 * with any previous page, so we'll hand the network a
2379 * "fragmented" page array that it can't transfer in 1 RDMA */
2380 if (page_count != 0 && oap->oap_page_off != 0)
2383 /* take the page out of our book-keeping */
2384 list_del_init(&oap->oap_pending_item);
2385 lop_update_pending(cli, lop, cmd, -1);
2386 list_del_init(&oap->oap_urgent_item);
2388 if (page_count == 0)
2389 starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2390 (PTLRPC_MAX_BRW_SIZE - 1);
2392 /* ask the caller for the size of the io as the rpc leaves. */
2393 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE))
2395 ops->ap_refresh_count(oap->oap_caller_data,cmd);
2396 if (oap->oap_count <= 0) {
2397 CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2399 osc_ap_completion(cli, NULL, oap, 0, oap->oap_count);
2403 /* now put the page back in our accounting */
2404 list_add_tail(&oap->oap_rpc_item, &rpc_list);
2405 if (page_count == 0)
2406 srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2407 if (++page_count >= cli->cl_max_pages_per_rpc)
2410 /* End on a PTLRPC_MAX_BRW_SIZE boundary. We want full-sized
2411 * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2412 * have the same alignment as the initial writes that allocated
2413 * extents on the server. */
2414 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2415 oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2416 if (ending_offset == 0)
2419 /* If there is a gap at the end of this page, it can't merge
2420 * with any subsequent pages, so we'll hand the network a
2421 * "fragmented" page array that it can't transfer in 1 RDMA */
2422 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2426 osc_wake_cache_waiters(cli);
2428 if (page_count == 0)
2431 loi_list_maint(cli, loi);
2433 client_obd_list_unlock(&cli->cl_loi_list_lock);
2435 req = osc_build_req(cli, &rpc_list, page_count, cmd);
2437 /* this should happen rarely and is pretty bad, it makes the
2438 * pending list not follow the dirty order */
2439 client_obd_list_lock(&cli->cl_loi_list_lock);
2440 list_for_each_entry_safe(oap, tmp, &rpc_list, oap_rpc_item) {
2441 list_del_init(&oap->oap_rpc_item);
2443 /* queued sync pages can be torn down while the pages
2444 * were between the pending list and the rpc */
2445 if (oap->oap_interrupted) {
2446 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2447 osc_ap_completion(cli, NULL, oap, 0,
2451 osc_ap_completion(cli, NULL, oap, 0, PTR_ERR(req));
2453 loi_list_maint(cli, loi);
2454 RETURN(PTR_ERR(req));
2457 aa = ptlrpc_req_async_args(req);
2458 if (cmd == OBD_BRW_READ) {
2459 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2460 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2461 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2462 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2464 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2465 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2466 cli->cl_w_in_flight);
2467 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2468 (starting_offset >> CFS_PAGE_SHIFT) + 1);
2470 ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2472 client_obd_list_lock(&cli->cl_loi_list_lock);
2474 if (cmd == OBD_BRW_READ)
2475 cli->cl_r_in_flight++;
2477 cli->cl_w_in_flight++;
2479 /* queued sync pages can be torn down while the pages
2480 * were between the pending list and the rpc */
2482 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2483 /* only one oap gets a request reference */
2486 if (oap->oap_interrupted && !req->rq_intr) {
2487 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2489 ptlrpc_mark_interrupted(req);
2493 tmp->oap_request = ptlrpc_request_addref(req);
2495 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2496 page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2498 req->rq_interpret_reply = brw_interpret;
2499 ptlrpcd_add_req(req);
2503 #define LOI_DEBUG(LOI, STR, args...) \
2504 CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR, \
2505 !list_empty(&(LOI)->loi_ready_item) || \
2506 !list_empty(&(LOI)->loi_hp_ready_item), \
2507 (LOI)->loi_write_lop.lop_num_pending, \
2508 !list_empty(&(LOI)->loi_write_lop.lop_urgent), \
2509 (LOI)->loi_read_lop.lop_num_pending, \
2510 !list_empty(&(LOI)->loi_read_lop.lop_urgent), \
2513 /* This is called by osc_check_rpcs() to find which objects have pages that
2514 * we could be sending. These lists are maintained by lop_makes_rpc(). */
2515 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2518 /* First return objects that have blocked locks so that they
2519 * will be flushed quickly and other clients can get the lock,
2520 * then objects which have pages ready to be stuffed into RPCs */
2521 if (!list_empty(&cli->cl_loi_hp_ready_list))
2522 RETURN(list_entry(cli->cl_loi_hp_ready_list.next,
2523 struct lov_oinfo, loi_hp_ready_item));
2524 if (!list_empty(&cli->cl_loi_ready_list))
2525 RETURN(list_entry(cli->cl_loi_ready_list.next,
2526 struct lov_oinfo, loi_ready_item));
2528 /* then if we have cache waiters, return all objects with queued
2529 * writes. This is especially important when many small files
2530 * have filled up the cache and not been fired into rpcs because
2531 * they don't pass the nr_pending/object threshhold */
2532 if (!list_empty(&cli->cl_cache_waiters) &&
2533 !list_empty(&cli->cl_loi_write_list))
2534 RETURN(list_entry(cli->cl_loi_write_list.next,
2535 struct lov_oinfo, loi_write_item));
2537 /* then return all queued objects when we have an invalid import
2538 * so that they get flushed */
2539 if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2540 if (!list_empty(&cli->cl_loi_write_list))
2541 RETURN(list_entry(cli->cl_loi_write_list.next,
2542 struct lov_oinfo, loi_write_item));
2543 if (!list_empty(&cli->cl_loi_read_list))
2544 RETURN(list_entry(cli->cl_loi_read_list.next,
2545 struct lov_oinfo, loi_read_item));
2550 static int osc_max_rpc_in_flight(struct client_obd *cli, struct lov_oinfo *loi)
2552 struct osc_async_page *oap;
2555 if (!list_empty(&loi->loi_write_lop.lop_urgent)) {
2556 oap = list_entry(loi->loi_write_lop.lop_urgent.next,
2557 struct osc_async_page, oap_urgent_item);
2558 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2561 if (!hprpc && !list_empty(&loi->loi_read_lop.lop_urgent)) {
2562 oap = list_entry(loi->loi_write_lop.lop_urgent.next,
2563 struct osc_async_page, oap_urgent_item);
2564 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2567 return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
2570 /* called with the loi list lock held */
2571 static void osc_check_rpcs(struct client_obd *cli)
2573 struct lov_oinfo *loi;
2574 int rc = 0, race_counter = 0;
2577 while ((loi = osc_next_loi(cli)) != NULL) {
2578 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2580 if (osc_max_rpc_in_flight(cli, loi))
2583 /* attempt some read/write balancing by alternating between
2584 * reads and writes in an object. The makes_rpc checks here
2585 * would be redundant if we were getting read/write work items
2586 * instead of objects. we don't want send_oap_rpc to drain a
2587 * partial read pending queue when we're given this object to
2588 * do io on writes while there are cache waiters */
2589 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2590 rc = osc_send_oap_rpc(cli, loi, OBD_BRW_WRITE,
2591 &loi->loi_write_lop);
2599 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2600 rc = osc_send_oap_rpc(cli, loi, OBD_BRW_READ,
2601 &loi->loi_read_lop);
2610 /* attempt some inter-object balancing by issueing rpcs
2611 * for each object in turn */
2612 if (!list_empty(&loi->loi_hp_ready_item))
2613 list_del_init(&loi->loi_hp_ready_item);
2614 if (!list_empty(&loi->loi_ready_item))
2615 list_del_init(&loi->loi_ready_item);
2616 if (!list_empty(&loi->loi_write_item))
2617 list_del_init(&loi->loi_write_item);
2618 if (!list_empty(&loi->loi_read_item))
2619 list_del_init(&loi->loi_read_item);
2621 loi_list_maint(cli, loi);
2623 /* send_oap_rpc fails with 0 when make_ready tells it to
2624 * back off. llite's make_ready does this when it tries
2625 * to lock a page queued for write that is already locked.
2626 * we want to try sending rpcs from many objects, but we
2627 * don't want to spin failing with 0. */
2628 if (race_counter == 10)
2634 /* we're trying to queue a page in the osc so we're subject to the
2635 * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2636 * If the osc's queued pages are already at that limit, then we want to sleep
2637 * until there is space in the osc's queue for us. We also may be waiting for
2638 * write credits from the OST if there are RPCs in flight that may return some
2639 * before we fall back to sync writes.
2641 * We need this know our allocation was granted in the presence of signals */
2642 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2646 client_obd_list_lock(&cli->cl_loi_list_lock);
2647 rc = list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2648 client_obd_list_unlock(&cli->cl_loi_list_lock);
2652 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2653 * grant or cache space. */
2654 static int osc_enter_cache(struct client_obd *cli, struct lov_oinfo *loi,
2655 struct osc_async_page *oap)
2657 struct osc_cache_waiter ocw;
2658 struct l_wait_info lwi = { 0 };
2661 CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2662 "grant: %lu\n", cli->cl_dirty, atomic_read(&obd_dirty_pages),
2663 cli->cl_dirty_max, obd_max_dirty_pages,
2664 cli->cl_lost_grant, cli->cl_avail_grant);
2666 /* force the caller to try sync io. this can jump the list
2667 * of queued writes and create a discontiguous rpc stream */
2668 if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2669 loi->loi_ar.ar_force_sync)
2672 /* Hopefully normal case - cache space and write credits available */
2673 if ((cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max) &&
2674 (atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages) &&
2675 (cli->cl_avail_grant >= CFS_PAGE_SIZE)) {
2676 /* account for ourselves */
2677 osc_consume_write_grant(cli, &oap->oap_brw_page);
2681 /* Make sure that there are write rpcs in flight to wait for. This
2682 * is a little silly as this object may not have any pending but
2683 * other objects sure might. */
2684 if (cli->cl_w_in_flight) {
2685 list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2686 cfs_waitq_init(&ocw.ocw_waitq);
2690 loi_list_maint(cli, loi);
2691 osc_check_rpcs(cli);
2692 client_obd_list_unlock(&cli->cl_loi_list_lock);
2694 CDEBUG(D_CACHE, "sleeping for cache space\n");
2695 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2697 client_obd_list_lock(&cli->cl_loi_list_lock);
2698 if (!list_empty(&ocw.ocw_entry)) {
2699 list_del(&ocw.ocw_entry);
2708 static int osc_reget_short_lock(struct obd_export *exp,
2709 struct lov_stripe_md *lsm,
2711 obd_off start, obd_off end,
2714 struct osc_async_page *oap = *res;
2719 spin_lock(&oap->oap_lock);
2720 rc = ldlm_lock_fast_match(oap->oap_ldlm_lock, rw,
2721 start, end, cookie);
2722 spin_unlock(&oap->oap_lock);
2727 static int osc_release_short_lock(struct obd_export *exp,
2728 struct lov_stripe_md *lsm, obd_off end,
2729 void *cookie, int rw)
2732 ldlm_lock_fast_release(cookie, rw);
2733 /* no error could have happened at this layer */
2737 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2738 struct lov_oinfo *loi, cfs_page_t *page,
2739 obd_off offset, struct obd_async_page_ops *ops,
2740 void *data, void **res, int nocache,
2741 struct lustre_handle *lockh)
2743 struct osc_async_page *oap;
2744 struct ldlm_res_id oid = {{0}};
2750 return size_round(sizeof(*oap));
2753 oap->oap_magic = OAP_MAGIC;
2754 oap->oap_cli = &exp->exp_obd->u.cli;
2757 oap->oap_caller_ops = ops;
2758 oap->oap_caller_data = data;
2760 oap->oap_page = page;
2761 oap->oap_obj_off = offset;
2763 CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2764 CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2765 CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2766 CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2768 oap->oap_occ.occ_interrupted = osc_occ_interrupted;
2770 spin_lock_init(&oap->oap_lock);
2772 /* If the page was marked as notcacheable - don't add to any locks */
2774 osc_build_res_name(loi->loi_id, loi->loi_gr, &oid);
2775 /* This is the only place where we can call cache_add_extent
2776 without oap_lock, because this page is locked now, and
2777 the lock we are adding it to is referenced, so cannot lose
2778 any pages either. */
2779 rc = cache_add_extent(oap->oap_cli->cl_cache, &oid, oap, lockh);
2784 CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
2788 struct osc_async_page *oap_from_cookie(void *cookie)
2790 struct osc_async_page *oap = cookie;
2791 if (oap->oap_magic != OAP_MAGIC)
2792 return ERR_PTR(-EINVAL);
2796 static int osc_queue_async_io(struct obd_export *exp, struct lov_stripe_md *lsm,
2797 struct lov_oinfo *loi, void *cookie,
2798 int cmd, obd_off off, int count,
2799 obd_flag brw_flags, enum async_flags async_flags)
2801 struct client_obd *cli = &exp->exp_obd->u.cli;
2802 struct osc_async_page *oap;
2806 oap = oap_from_cookie(cookie);
2808 RETURN(PTR_ERR(oap));
2810 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2813 if (!list_empty(&oap->oap_pending_item) ||
2814 !list_empty(&oap->oap_urgent_item) ||
2815 !list_empty(&oap->oap_rpc_item))
2818 /* check if the file's owner/group is over quota */
2819 if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)){
2820 struct obd_async_page_ops *ops;
2827 ops = oap->oap_caller_ops;
2828 ops->ap_fill_obdo(oap->oap_caller_data, cmd, oa);
2829 if (lquota_chkdq(quota_interface, cli, oa->o_uid, oa->o_gid) ==
2839 loi = lsm->lsm_oinfo[0];
2841 client_obd_list_lock(&cli->cl_loi_list_lock);
2844 oap->oap_page_off = off;
2845 oap->oap_count = count;
2846 oap->oap_brw_flags = brw_flags;
2847 oap->oap_async_flags = async_flags;
2849 if (cmd & OBD_BRW_WRITE) {
2850 rc = osc_enter_cache(cli, loi, oap);
2852 client_obd_list_unlock(&cli->cl_loi_list_lock);
2857 osc_oap_to_pending(oap);
2858 loi_list_maint(cli, loi);
2860 LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
2863 osc_check_rpcs(cli);
2864 client_obd_list_unlock(&cli->cl_loi_list_lock);
2869 /* aka (~was & now & flag), but this is more clear :) */
2870 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
2872 static int osc_set_async_flags(struct obd_export *exp,
2873 struct lov_stripe_md *lsm,
2874 struct lov_oinfo *loi, void *cookie,
2875 obd_flag async_flags)
2877 struct client_obd *cli = &exp->exp_obd->u.cli;
2878 struct loi_oap_pages *lop;
2879 struct osc_async_page *oap;
2883 oap = oap_from_cookie(cookie);
2885 RETURN(PTR_ERR(oap));
2888 * bug 7311: OST-side locking is only supported for liblustre for now
2889 * (and liblustre never calls obd_set_async_flags(). I hope.), generic
2890 * implementation has to handle case where OST-locked page was picked
2891 * up by, e.g., ->writepage().
2893 LASSERT(!(oap->oap_brw_flags & OBD_BRW_SRVLOCK));
2894 LASSERT(!LIBLUSTRE_CLIENT); /* check that liblustre angels do fear to
2897 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2901 loi = lsm->lsm_oinfo[0];
2903 if (oap->oap_cmd & OBD_BRW_WRITE) {
2904 lop = &loi->loi_write_lop;
2906 lop = &loi->loi_read_lop;
2909 client_obd_list_lock(&cli->cl_loi_list_lock);
2911 if (list_empty(&oap->oap_pending_item))
2912 GOTO(out, rc = -EINVAL);
2914 if ((oap->oap_async_flags & async_flags) == async_flags)
2917 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
2918 oap->oap_async_flags |= ASYNC_READY;
2920 if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT) &&
2921 list_empty(&oap->oap_rpc_item)) {
2922 if (oap->oap_async_flags & ASYNC_HP)
2923 list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2925 list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2926 oap->oap_async_flags |= ASYNC_URGENT;
2927 loi_list_maint(cli, loi);
2930 LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
2931 oap->oap_async_flags);
2933 osc_check_rpcs(cli);
2934 client_obd_list_unlock(&cli->cl_loi_list_lock);
2938 static int osc_queue_group_io(struct obd_export *exp, struct lov_stripe_md *lsm,
2939 struct lov_oinfo *loi,
2940 struct obd_io_group *oig, void *cookie,
2941 int cmd, obd_off off, int count,
2943 obd_flag async_flags)
2945 struct client_obd *cli = &exp->exp_obd->u.cli;
2946 struct osc_async_page *oap;
2947 struct loi_oap_pages *lop;
2951 oap = oap_from_cookie(cookie);
2953 RETURN(PTR_ERR(oap));
2955 if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2958 if (!list_empty(&oap->oap_pending_item) ||
2959 !list_empty(&oap->oap_urgent_item) ||
2960 !list_empty(&oap->oap_rpc_item))
2964 loi = lsm->lsm_oinfo[0];
2966 client_obd_list_lock(&cli->cl_loi_list_lock);
2969 oap->oap_page_off = off;
2970 oap->oap_count = count;
2971 oap->oap_brw_flags = brw_flags;
2972 oap->oap_async_flags = async_flags;
2974 if (cmd & OBD_BRW_WRITE)
2975 lop = &loi->loi_write_lop;
2977 lop = &loi->loi_read_lop;
2979 list_add_tail(&oap->oap_pending_item, &lop->lop_pending_group);
2980 if (oap->oap_async_flags & ASYNC_GROUP_SYNC) {
2982 rc = oig_add_one(oig, &oap->oap_occ);
2985 LOI_DEBUG(loi, "oap %p page %p on group pending: rc %d\n",
2986 oap, oap->oap_page, rc);
2988 client_obd_list_unlock(&cli->cl_loi_list_lock);
2993 static void osc_group_to_pending(struct client_obd *cli, struct lov_oinfo *loi,
2994 struct loi_oap_pages *lop, int cmd)
2996 struct list_head *pos, *tmp;
2997 struct osc_async_page *oap;
2999 list_for_each_safe(pos, tmp, &lop->lop_pending_group) {
3000 oap = list_entry(pos, struct osc_async_page, oap_pending_item);
3001 list_del(&oap->oap_pending_item);
3002 osc_oap_to_pending(oap);
3004 loi_list_maint(cli, loi);
3007 static int osc_trigger_group_io(struct obd_export *exp,
3008 struct lov_stripe_md *lsm,
3009 struct lov_oinfo *loi,
3010 struct obd_io_group *oig)
3012 struct client_obd *cli = &exp->exp_obd->u.cli;
3016 loi = lsm->lsm_oinfo[0];
3018 client_obd_list_lock(&cli->cl_loi_list_lock);
3020 osc_group_to_pending(cli, loi, &loi->loi_write_lop, OBD_BRW_WRITE);
3021 osc_group_to_pending(cli, loi, &loi->loi_read_lop, OBD_BRW_READ);
3023 osc_check_rpcs(cli);
3024 client_obd_list_unlock(&cli->cl_loi_list_lock);
3029 static int osc_teardown_async_page(struct obd_export *exp,
3030 struct lov_stripe_md *lsm,
3031 struct lov_oinfo *loi, void *cookie)
3033 struct client_obd *cli = &exp->exp_obd->u.cli;
3034 struct loi_oap_pages *lop;
3035 struct osc_async_page *oap;
3039 oap = oap_from_cookie(cookie);
3041 RETURN(PTR_ERR(oap));
3044 loi = lsm->lsm_oinfo[0];
3046 if (oap->oap_cmd & OBD_BRW_WRITE) {
3047 lop = &loi->loi_write_lop;
3049 lop = &loi->loi_read_lop;
3052 client_obd_list_lock(&cli->cl_loi_list_lock);
3054 if (!list_empty(&oap->oap_rpc_item))
3055 GOTO(out, rc = -EBUSY);
3057 osc_exit_cache(cli, oap, 0);
3058 osc_wake_cache_waiters(cli);
3060 if (!list_empty(&oap->oap_urgent_item)) {
3061 list_del_init(&oap->oap_urgent_item);
3062 oap->oap_async_flags &= ~(ASYNC_URGENT | ASYNC_HP);
3065 if (!list_empty(&oap->oap_pending_item)) {
3066 list_del_init(&oap->oap_pending_item);
3067 lop_update_pending(cli, lop, oap->oap_cmd, -1);
3069 loi_list_maint(cli, loi);
3070 cache_remove_extent(cli->cl_cache, oap);
3072 LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
3074 client_obd_list_unlock(&cli->cl_loi_list_lock);
3078 int osc_extent_blocking_cb(struct ldlm_lock *lock,
3079 struct ldlm_lock_desc *new, void *data,
3082 struct lustre_handle lockh = { 0 };
3086 if ((unsigned long)data > 0 && (unsigned long)data < 0x1000) {
3087 LDLM_ERROR(lock, "cancelling lock with bad data %p", data);
3092 case LDLM_CB_BLOCKING:
3093 ldlm_lock2handle(lock, &lockh);
3094 rc = ldlm_cli_cancel(&lockh);
3096 CERROR("ldlm_cli_cancel failed: %d\n", rc);
3098 case LDLM_CB_CANCELING: {
3100 ldlm_lock2handle(lock, &lockh);
3101 /* This lock wasn't granted, don't try to do anything */
3102 if (lock->l_req_mode != lock->l_granted_mode)
3105 cache_remove_lock(lock->l_conn_export->exp_obd->u.cli.cl_cache,
3108 if (lock->l_conn_export->exp_obd->u.cli.cl_ext_lock_cancel_cb)
3109 lock->l_conn_export->exp_obd->u.cli.cl_ext_lock_cancel_cb(
3110 lock, new, data,flag);
3119 EXPORT_SYMBOL(osc_extent_blocking_cb);
3121 static void osc_set_data_with_check(struct lustre_handle *lockh, void *data,
3124 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
3127 CERROR("lockh %p, data %p - client evicted?\n", lockh, data);
3130 lock_res_and_lock(lock);
3131 #if defined (__KERNEL__) && defined (__linux__)
3132 /* Liang XXX: Darwin and Winnt checking should be added */
3133 if (lock->l_ast_data && lock->l_ast_data != data) {
3134 struct inode *new_inode = data;
3135 struct inode *old_inode = lock->l_ast_data;
3136 if (!(old_inode->i_state & I_FREEING))
3137 LDLM_ERROR(lock, "inconsistent l_ast_data found");
3138 LASSERTF(old_inode->i_state & I_FREEING,
3139 "Found existing inode %p/%lu/%u state %lu in lock: "
3140 "setting data to %p/%lu/%u\n", old_inode,
3141 old_inode->i_ino, old_inode->i_generation,
3143 new_inode, new_inode->i_ino, new_inode->i_generation);
3146 lock->l_ast_data = data;
3147 lock->l_flags |= (flags & LDLM_FL_NO_LRU);
3148 unlock_res_and_lock(lock);
3149 LDLM_LOCK_PUT(lock);
3152 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3153 ldlm_iterator_t replace, void *data)
3155 struct ldlm_res_id res_id;
3156 struct obd_device *obd = class_exp2obd(exp);
3158 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_gr, &res_id);
3159 ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3163 static int osc_enqueue_fini(struct obd_device *obd, struct ptlrpc_request *req,
3164 struct obd_info *oinfo, int intent, int rc)
3169 /* The request was created before ldlm_cli_enqueue call. */
3170 if (rc == ELDLM_LOCK_ABORTED) {
3171 struct ldlm_reply *rep;
3173 /* swabbed by ldlm_cli_enqueue() */
3174 LASSERT(lustre_rep_swabbed(req, DLM_LOCKREPLY_OFF));
3175 rep = lustre_msg_buf(req->rq_repmsg, DLM_LOCKREPLY_OFF,
3177 LASSERT(rep != NULL);
3178 if (rep->lock_policy_res1)
3179 rc = rep->lock_policy_res1;
3183 if ((intent && rc == ELDLM_LOCK_ABORTED) || !rc) {
3184 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
3185 oinfo->oi_md->lsm_oinfo[0]->loi_lvb.lvb_size,
3186 oinfo->oi_md->lsm_oinfo[0]->loi_lvb.lvb_blocks,
3187 oinfo->oi_md->lsm_oinfo[0]->loi_lvb.lvb_mtime);
3191 cache_add_lock(obd->u.cli.cl_cache, oinfo->oi_lockh);
3193 /* Call the update callback. */
3194 rc = oinfo->oi_cb_up(oinfo, rc);
3198 static int osc_enqueue_interpret(struct ptlrpc_request *req,
3199 struct osc_enqueue_args *aa, int rc)
3201 int intent = aa->oa_oi->oi_flags & LDLM_FL_HAS_INTENT;
3202 struct lov_stripe_md *lsm = aa->oa_oi->oi_md;
3203 struct ldlm_lock *lock;
3205 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
3207 lock = ldlm_handle2lock(aa->oa_oi->oi_lockh);
3209 /* Complete obtaining the lock procedure. */
3210 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
3212 &aa->oa_oi->oi_flags,
3213 &lsm->lsm_oinfo[0]->loi_lvb,
3214 sizeof(lsm->lsm_oinfo[0]->loi_lvb),
3215 lustre_swab_ost_lvb,
3216 aa->oa_oi->oi_lockh, rc);
3218 /* Complete osc stuff. */
3219 rc = osc_enqueue_fini(aa->oa_exp->exp_obd, req, aa->oa_oi, intent, rc);
3221 /* Release the lock for async request. */
3222 if (lustre_handle_is_used(aa->oa_oi->oi_lockh) && rc == ELDLM_OK)
3223 ldlm_lock_decref(aa->oa_oi->oi_lockh, aa->oa_ei->ei_mode);
3225 LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
3226 aa->oa_oi->oi_lockh, req, aa);
3227 LDLM_LOCK_PUT(lock);
3231 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
3232 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
3233 * other synchronous requests, however keeping some locks and trying to obtain
3234 * others may take a considerable amount of time in a case of ost failure; and
3235 * when other sync requests do not get released lock from a client, the client
3236 * is excluded from the cluster -- such scenarious make the life difficult, so
3237 * release locks just after they are obtained. */
3238 static int osc_enqueue(struct obd_export *exp, struct obd_info *oinfo,
3239 struct ldlm_enqueue_info *einfo,
3240 struct ptlrpc_request_set *rqset)
3242 struct ldlm_res_id res_id;
3243 struct obd_device *obd = exp->exp_obd;
3244 struct ldlm_reply *rep;
3245 struct ptlrpc_request *req = NULL;
3246 int intent = oinfo->oi_flags & LDLM_FL_HAS_INTENT;
3251 osc_build_res_name(oinfo->oi_md->lsm_object_id,
3252 oinfo->oi_md->lsm_object_gr, &res_id);
3253 /* Filesystem lock extents are extended to page boundaries so that
3254 * dealing with the page cache is a little smoother. */
3255 oinfo->oi_policy.l_extent.start -=
3256 oinfo->oi_policy.l_extent.start & ~CFS_PAGE_MASK;
3257 oinfo->oi_policy.l_extent.end |= ~CFS_PAGE_MASK;
3259 if (oinfo->oi_md->lsm_oinfo[0]->loi_kms_valid == 0)
3262 /* Next, search for already existing extent locks that will cover us */
3263 /* If we're trying to read, we also search for an existing PW lock. The
3264 * VFS and page cache already protect us locally, so lots of readers/
3265 * writers can share a single PW lock.
3267 * There are problems with conversion deadlocks, so instead of
3268 * converting a read lock to a write lock, we'll just enqueue a new
3271 * At some point we should cancel the read lock instead of making them
3272 * send us a blocking callback, but there are problems with canceling
3273 * locks out from other users right now, too. */
3274 mode = einfo->ei_mode;
3275 if (einfo->ei_mode == LCK_PR)
3277 mode = ldlm_lock_match(obd->obd_namespace,
3278 oinfo->oi_flags | LDLM_FL_LVB_READY, &res_id,
3279 einfo->ei_type, &oinfo->oi_policy, mode,
3282 /* addref the lock only if not async requests and PW lock is
3283 * matched whereas we asked for PR. */
3284 if (!rqset && einfo->ei_mode != mode)
3285 ldlm_lock_addref(oinfo->oi_lockh, LCK_PR);
3286 osc_set_data_with_check(oinfo->oi_lockh, einfo->ei_cbdata,
3289 /* I would like to be able to ASSERT here that rss <=
3290 * kms, but I can't, for reasons which are explained in
3294 /* We already have a lock, and it's referenced */
3295 oinfo->oi_cb_up(oinfo, ELDLM_OK);
3297 /* For async requests, decref the lock. */
3298 if (einfo->ei_mode != mode)
3299 ldlm_lock_decref(oinfo->oi_lockh, LCK_PW);
3301 ldlm_lock_decref(oinfo->oi_lockh, einfo->ei_mode);
3309 [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
3310 [DLM_LOCKREQ_OFF] = sizeof(struct ldlm_request),
3311 [DLM_LOCKREQ_OFF + 1] = 0 };
3313 req = ldlm_prep_enqueue_req(exp, 2, size, NULL, 0);
3317 size[DLM_LOCKREPLY_OFF] = sizeof(*rep);
3318 size[DLM_REPLY_REC_OFF] =
3319 sizeof(oinfo->oi_md->lsm_oinfo[0]->loi_lvb);
3320 ptlrpc_req_set_repsize(req, 3, size);
3323 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
3324 oinfo->oi_flags &= ~LDLM_FL_BLOCK_GRANTED;
3326 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id,
3327 &oinfo->oi_policy, &oinfo->oi_flags,
3328 &oinfo->oi_md->lsm_oinfo[0]->loi_lvb,
3329 sizeof(oinfo->oi_md->lsm_oinfo[0]->loi_lvb),
3330 lustre_swab_ost_lvb, oinfo->oi_lockh,
3334 struct osc_enqueue_args *aa;
3335 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3336 aa = ptlrpc_req_async_args(req);
3341 req->rq_interpret_reply = osc_enqueue_interpret;
3342 ptlrpc_set_add_req(rqset, req);
3343 } else if (intent) {
3344 ptlrpc_req_finished(req);
3349 rc = osc_enqueue_fini(obd, req, oinfo, intent, rc);
3351 ptlrpc_req_finished(req);
3356 static int osc_match(struct obd_export *exp, struct lov_stripe_md *lsm,
3357 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
3358 int *flags, void *data, struct lustre_handle *lockh,
3361 struct ldlm_res_id res_id;
3362 struct obd_device *obd = exp->exp_obd;
3363 int lflags = *flags;
3367 osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_gr, &res_id);
3369 OBD_FAIL_RETURN(OBD_FAIL_OSC_MATCH, -EIO);
3371 /* Filesystem lock extents are extended to page boundaries so that
3372 * dealing with the page cache is a little smoother */
3373 policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3374 policy->l_extent.end |= ~CFS_PAGE_MASK;
3376 /* Next, search for already existing extent locks that will cover us */
3377 /* If we're trying to read, we also search for an existing PW lock. The
3378 * VFS and page cache already protect us locally, so lots of readers/
3379 * writers can share a single PW lock. */
3383 rc = ldlm_lock_match(obd->obd_namespace, lflags | LDLM_FL_LVB_READY,
3384 &res_id, type, policy, rc, lockh);
3386 osc_set_data_with_check(lockh, data, lflags);
3387 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
3388 ldlm_lock_addref(lockh, LCK_PR);
3389 ldlm_lock_decref(lockh, LCK_PW);
3391 if (n_matches != NULL)
3398 static int osc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
3399 __u32 mode, struct lustre_handle *lockh)
3403 if (unlikely(mode == LCK_GROUP))
3404 ldlm_lock_decref_and_cancel(lockh, mode);
3406 ldlm_lock_decref(lockh, mode);
3411 static int osc_cancel_unused(struct obd_export *exp,
3412 struct lov_stripe_md *lsm, int flags, void *opaque)
3414 struct obd_device *obd = class_exp2obd(exp);
3415 struct ldlm_res_id res_id, *resp = NULL;
3418 resp = osc_build_res_name(lsm->lsm_object_id,
3419 lsm->lsm_object_gr, &res_id);
3422 return ldlm_cli_cancel_unused(obd->obd_namespace, resp, flags, opaque);
3426 static int osc_join_lru(struct obd_export *exp,
3427 struct lov_stripe_md *lsm, int join)
3429 struct obd_device *obd = class_exp2obd(exp);
3430 struct ldlm_res_id res_id, *resp = NULL;
3433 resp = osc_build_res_name(lsm->lsm_object_id,
3434 lsm->lsm_object_gr, &res_id);
3437 return ldlm_cli_join_lru(obd->obd_namespace, resp, join);
3441 static int osc_statfs_interpret(struct ptlrpc_request *req,
3442 struct osc_async_args *aa, int rc)
3444 struct obd_statfs *msfs;
3450 msfs = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*msfs),
3451 lustre_swab_obd_statfs);
3453 CERROR("Can't unpack obd_statfs\n");
3454 GOTO(out, rc = -EPROTO);
3457 memcpy(aa->aa_oi->oi_osfs, msfs, sizeof(*msfs));
3459 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
3463 static int osc_statfs_async(struct obd_device *obd, struct obd_info *oinfo,
3464 __u64 max_age, struct ptlrpc_request_set *rqset)
3466 struct ptlrpc_request *req;
3467 struct osc_async_args *aa;
3468 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*oinfo->oi_osfs) };
3471 /* We could possibly pass max_age in the request (as an absolute
3472 * timestamp or a "seconds.usec ago") so the target can avoid doing
3473 * extra calls into the filesystem if that isn't necessary (e.g.
3474 * during mount that would help a bit). Having relative timestamps
3475 * is not so great if request processing is slow, while absolute
3476 * timestamps are not ideal because they need time synchronization. */
3477 req = ptlrpc_prep_req(obd->u.cli.cl_import, LUSTRE_OST_VERSION,
3478 OST_STATFS, 1, NULL, NULL);
3482 ptlrpc_req_set_repsize(req, 2, size);
3483 req->rq_request_portal = OST_CREATE_PORTAL;
3484 ptlrpc_at_set_req_timeout(req);
3485 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
3486 /* procfs requests not want stat in wait for avoid deadlock */
3487 req->rq_no_resend = 1;
3488 req->rq_no_delay = 1;
3491 req->rq_interpret_reply = osc_statfs_interpret;
3492 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3493 aa = ptlrpc_req_async_args(req);
3496 ptlrpc_set_add_req(rqset, req);
3500 static int osc_statfs(struct obd_device *obd, struct obd_statfs *osfs,
3501 __u64 max_age, __u32 flags)
3503 struct obd_statfs *msfs;
3504 struct ptlrpc_request *req;
3505 struct obd_import *imp = NULL;
3506 __u32 size[2] = { sizeof(struct ptlrpc_body), sizeof(*osfs) };
3510 /*Since the request might also come from lprocfs, so we need
3511 *sync this with client_disconnect_export Bug15684*/
3512 down_read(&obd->u.cli.cl_sem);
3513 if (obd->u.cli.cl_import)
3514 imp = class_import_get(obd->u.cli.cl_import);
3515 up_read(&obd->u.cli.cl_sem);
3519 /* We could possibly pass max_age in the request (as an absolute
3520 * timestamp or a "seconds.usec ago") so the target can avoid doing
3521 * extra calls into the filesystem if that isn't necessary (e.g.
3522 * during mount that would help a bit). Having relative timestamps
3523 * is not so great if request processing is slow, while absolute
3524 * timestamps are not ideal because they need time synchronization. */
3525 req = ptlrpc_prep_req(imp, LUSTRE_OST_VERSION,
3526 OST_STATFS, 1, NULL, NULL);
3528 class_import_put(imp);
3532 ptlrpc_req_set_repsize(req, 2, size);
3533 req->rq_request_portal = OST_CREATE_PORTAL;
3534 ptlrpc_at_set_req_timeout(req);
3536 if (flags & OBD_STATFS_NODELAY) {
3537 /* procfs requests not want stat in wait for avoid deadlock */
3538 req->rq_no_resend = 1;
3539 req->rq_no_delay = 1;
3542 rc = ptlrpc_queue_wait(req);
3546 msfs = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*msfs),
3547 lustre_swab_obd_statfs);
3549 CERROR("Can't unpack obd_statfs\n");
3550 GOTO(out, rc = -EPROTO);
3553 memcpy(osfs, msfs, sizeof(*osfs));
3557 ptlrpc_req_finished(req);
3561 /* Retrieve object striping information.
3563 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
3564 * the maximum number of OST indices which will fit in the user buffer.
3565 * lmm_magic must be LOV_MAGIC_V1 or LOV_MAGIC_V3 (we only use 1 slot here).
3567 static int osc_getstripe(struct lov_stripe_md *lsm, struct lov_user_md *lump)
3569 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
3570 struct lov_user_md_v3 lum, *lumk;
3571 int rc = 0, lum_size;
3572 struct lov_user_ost_data_v1 *lmm_objects;
3578 /* we only need the header part from user space to get lmm_magic and
3579 * lmm_stripe_count, (the header part is common to v1 and v3) */
3580 lum_size = sizeof(struct lov_user_md_v1);
3581 if (copy_from_user(&lum, lump, lum_size))
3584 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
3585 (lum.lmm_magic != LOV_USER_MAGIC_V3))
3588 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
3589 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
3590 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
3591 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
3593 /* we can use lov_mds_md_size() to compute lum_size
3594 * because lov_user_md_vX and lov_mds_md_vX have the same size */
3595 if (lum.lmm_stripe_count > 0) {
3596 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
3597 OBD_ALLOC(lumk, lum_size);
3600 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
3601 lmm_objects = &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
3603 lmm_objects = &(lumk->lmm_objects[0]);
3604 lmm_objects->l_object_id = lsm->lsm_object_id;
3606 lum_size = lov_mds_md_size(0, lum.lmm_magic);
3610 lumk->lmm_object_id = lsm->lsm_object_id;
3611 lumk->lmm_stripe_count = 1;
3613 if (copy_to_user(lump, lumk, lum_size))
3617 OBD_FREE(lumk, lum_size);
3623 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
3624 void *karg, void *uarg)
3626 struct obd_device *obd = exp->exp_obd;
3627 struct obd_ioctl_data *data = karg;
3631 if (!try_module_get(THIS_MODULE)) {
3632 CERROR("Can't get module. Is it alive?");
3636 case OBD_IOC_LOV_GET_CONFIG: {
3638 struct lov_desc *desc;
3639 struct obd_uuid uuid;
3643 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
3644 GOTO(out, err = -EINVAL);
3646 data = (struct obd_ioctl_data *)buf;
3648 if (sizeof(*desc) > data->ioc_inllen1) {
3649 obd_ioctl_freedata(buf, len);
3650 GOTO(out, err = -EINVAL);
3653 if (data->ioc_inllen2 < sizeof(uuid)) {
3654 obd_ioctl_freedata(buf, len);
3655 GOTO(out, err = -EINVAL);
3658 desc = (struct lov_desc *)data->ioc_inlbuf1;
3659 desc->ld_tgt_count = 1;
3660 desc->ld_active_tgt_count = 1;
3661 desc->ld_default_stripe_count = 1;
3662 desc->ld_default_stripe_size = 0;
3663 desc->ld_default_stripe_offset = 0;
3664 desc->ld_pattern = 0;
3665 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
3667 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
3669 err = copy_to_user((void *)uarg, buf, len);
3672 obd_ioctl_freedata(buf, len);
3675 case LL_IOC_LOV_SETSTRIPE:
3676 err = obd_alloc_memmd(exp, karg);
3680 case LL_IOC_LOV_GETSTRIPE:
3681 err = osc_getstripe(karg, uarg);
3683 case OBD_IOC_CLIENT_RECOVER:
3684 err = ptlrpc_recover_import(obd->u.cli.cl_import,
3689 case IOC_OSC_SET_ACTIVE:
3690 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
3693 case OBD_IOC_POLL_QUOTACHECK:
3694 err = lquota_poll_check(quota_interface, exp,
3695 (struct if_quotacheck *)karg);
3697 case OBD_IOC_DESTROY: {
3700 if (!cfs_capable(CFS_CAP_SYS_ADMIN))
3701 GOTO (out, err = -EPERM);
3702 oa = &data->ioc_obdo1;
3705 GOTO(out, err = -EINVAL);
3707 oa->o_valid |= OBD_MD_FLGROUP;
3709 err = osc_destroy(exp, oa, NULL, NULL, NULL);
3712 case OBD_IOC_PING_TARGET:
3713 err = ptlrpc_obd_ping(obd);
3716 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
3717 cmd, cfs_curproc_comm());
3718 GOTO(out, err = -ENOTTY);
3721 module_put(THIS_MODULE);
3725 static int osc_get_info(struct obd_export *exp, obd_count keylen,
3726 void *key, __u32 *vallen, void *val, struct lov_stripe_md *lsm)
3729 if (!vallen || !val)
3732 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
3733 __u32 *stripe = val;
3734 *vallen = sizeof(*stripe);
3737 } else if (KEY_IS(KEY_LAST_ID)) {
3738 struct ptlrpc_request *req;
3740 char *bufs[2] = { NULL, key };
3741 __u32 size[2] = { sizeof(struct ptlrpc_body), keylen };
3744 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
3745 OST_GET_INFO, 2, size, bufs);
3749 size[REPLY_REC_OFF] = *vallen;
3750 ptlrpc_req_set_repsize(req, 2, size);
3751 rc = ptlrpc_queue_wait(req);
3755 reply = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*reply),
3756 lustre_swab_ost_last_id);
3757 if (reply == NULL) {
3758 CERROR("Can't unpack OST last ID\n");
3759 GOTO(out, rc = -EPROTO);
3761 *((obd_id *)val) = *reply;
3763 ptlrpc_req_finished(req);
3765 } else if (KEY_IS(KEY_FIEMAP)) {
3766 struct ptlrpc_request *req;
3767 struct ll_user_fiemap *reply;
3768 char *bufs[2] = { NULL, key };
3769 __u32 size[2] = { sizeof(struct ptlrpc_body), keylen };
3772 req = ptlrpc_prep_req(class_exp2cliimp(exp), LUSTRE_OST_VERSION,
3773 OST_GET_INFO, 2, size, bufs);
3777 size[REPLY_REC_OFF] = *vallen;
3778 ptlrpc_req_set_repsize(req, 2, size);
3780 rc = ptlrpc_queue_wait(req);
3783 reply = lustre_swab_repbuf(req, REPLY_REC_OFF, *vallen,
3784 lustre_swab_fiemap);
3785 if (reply == NULL) {
3786 CERROR("Can't unpack FIEMAP reply.\n");
3787 GOTO(out1, rc = -EPROTO);
3790 memcpy(val, reply, *vallen);
3793 ptlrpc_req_finished(req);
3801 static int osc_setinfo_mds_conn_interpret(struct ptlrpc_request *req,
3804 struct llog_ctxt *ctxt;
3805 struct obd_import *imp = req->rq_import;
3811 ctxt = llog_get_context(imp->imp_obd, LLOG_MDS_OST_ORIG_CTXT);
3814 rc = llog_initiator_connect(ctxt);
3816 CERROR("cannot establish connection for "
3817 "ctxt %p: %d\n", ctxt, rc);
3820 llog_ctxt_put(ctxt);
3821 spin_lock(&imp->imp_lock);
3822 imp->imp_server_timeout = 1;
3823 imp->imp_pingable = 1;
3824 spin_unlock(&imp->imp_lock);
3825 CDEBUG(D_RPCTRACE, "pinging OST %s\n", obd2cli_tgt(imp->imp_obd));
3830 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
3831 void *key, obd_count vallen, void *val,
3832 struct ptlrpc_request_set *set)
3834 struct ptlrpc_request *req;
3835 struct obd_device *obd = exp->exp_obd;
3836 struct obd_import *imp = class_exp2cliimp(exp);
3837 __u32 size[3] = { sizeof(struct ptlrpc_body), keylen, vallen };
3838 char *bufs[3] = { NULL, key, val };
3841 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
3843 if (KEY_IS(KEY_NEXT_ID)) {
3844 if (vallen != sizeof(obd_id))
3846 obd->u.cli.cl_oscc.oscc_next_id = *((obd_id*)val) + 1;
3847 CDEBUG(D_HA, "%s: set oscc_next_id = "LPU64"\n",
3848 exp->exp_obd->obd_name,
3849 obd->u.cli.cl_oscc.oscc_next_id);
3854 if (KEY_IS(KEY_UNLINKED)) {
3855 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
3856 spin_lock(&oscc->oscc_lock);
3857 oscc->oscc_flags &= ~OSCC_FLAG_NOSPC;
3858 spin_unlock(&oscc->oscc_lock);
3862 if (KEY_IS(KEY_INIT_RECOV)) {
3863 if (vallen != sizeof(int))
3865 spin_lock(&imp->imp_lock);
3866 imp->imp_initial_recov = *(int *)val;
3867 spin_unlock(&imp->imp_lock);
3868 CDEBUG(D_HA, "%s: set imp_initial_recov = %d\n",
3869 exp->exp_obd->obd_name,
3870 imp->imp_initial_recov);
3874 if (KEY_IS(KEY_CHECKSUM)) {
3875 if (vallen != sizeof(int))
3877 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
3881 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
3884 /* We pass all other commands directly to OST. Since nobody calls osc
3885 methods directly and everybody is supposed to go through LOV, we
3886 assume lov checked invalid values for us.
3887 The only recognised values so far are evict_by_nid and mds_conn.
3888 Even if something bad goes through, we'd get a -EINVAL from OST
3891 req = ptlrpc_prep_req(imp, LUSTRE_OST_VERSION, OST_SET_INFO, 3, size,
3896 if (KEY_IS(KEY_MDS_CONN))
3897 req->rq_interpret_reply = osc_setinfo_mds_conn_interpret;
3898 else if (KEY_IS(KEY_GRANT_SHRINK))
3899 req->rq_interpret_reply = osc_shrink_grant_interpret;
3901 if (KEY_IS(KEY_GRANT_SHRINK)) {
3902 struct osc_grant_args *aa;
3905 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
3906 aa = ptlrpc_req_async_args(req);
3909 ptlrpc_req_finished(req);
3912 *oa = ((struct ost_body *)val)->oa;
3916 ptlrpc_req_set_repsize(req, 2, size);
3917 ptlrpcd_add_req(req);
3919 ptlrpc_req_set_repsize(req, 1, NULL);
3920 ptlrpc_set_add_req(set, req);
3921 ptlrpc_check_set(set);
3928 static struct llog_operations osc_size_repl_logops = {
3929 lop_cancel: llog_obd_repl_cancel
3932 static struct llog_operations osc_mds_ost_orig_logops;
3933 static int osc_llog_init(struct obd_device *obd, struct obd_device *tgt,
3934 int count, struct llog_catid *catid,
3935 struct obd_uuid *uuid)
3940 spin_lock(&obd->obd_dev_lock);
3941 if (osc_mds_ost_orig_logops.lop_setup != llog_obd_origin_setup) {
3942 osc_mds_ost_orig_logops = llog_lvfs_ops;
3943 osc_mds_ost_orig_logops.lop_setup = llog_obd_origin_setup;
3944 osc_mds_ost_orig_logops.lop_cleanup = llog_obd_origin_cleanup;
3945 osc_mds_ost_orig_logops.lop_add = llog_obd_origin_add;
3946 osc_mds_ost_orig_logops.lop_connect = llog_origin_connect;
3948 spin_unlock(&obd->obd_dev_lock);
3950 rc = llog_setup(obd, LLOG_MDS_OST_ORIG_CTXT, tgt, count,
3951 &catid->lci_logid, &osc_mds_ost_orig_logops);
3953 CERROR("failed LLOG_MDS_OST_ORIG_CTXT\n");
3957 rc = llog_setup(obd, LLOG_SIZE_REPL_CTXT, tgt, count, NULL,
3958 &osc_size_repl_logops);
3960 struct llog_ctxt *ctxt =
3961 llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3964 CERROR("failed LLOG_SIZE_REPL_CTXT\n");
3968 CERROR("osc '%s' tgt '%s' cnt %d catid %p rc=%d\n",
3969 obd->obd_name, tgt->obd_name, count, catid, rc);
3970 CERROR("logid "LPX64":0x%x\n",
3971 catid->lci_logid.lgl_oid, catid->lci_logid.lgl_ogen);
3976 static int osc_llog_finish(struct obd_device *obd, int count)
3978 struct llog_ctxt *ctxt;
3979 int rc = 0, rc2 = 0;
3982 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
3984 rc = llog_cleanup(ctxt);
3986 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
3988 rc2 = llog_cleanup(ctxt);
3995 static int osc_reconnect(struct obd_export *exp, struct obd_device *obd,
3996 struct obd_uuid *cluuid,
3997 struct obd_connect_data *data,
4000 struct client_obd *cli = &obd->u.cli;
4002 if (data != NULL && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
4005 client_obd_list_lock(&cli->cl_loi_list_lock);
4006 data->ocd_grant = cli->cl_avail_grant ?:
4007 2 * cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT;
4008 lost_grant = cli->cl_lost_grant;
4009 cli->cl_lost_grant = 0;
4010 client_obd_list_unlock(&cli->cl_loi_list_lock);
4012 CDEBUG(D_CACHE, "request ocd_grant: %d cl_avail_grant: %ld "
4013 "cl_lost_grant: %ld\n", data->ocd_grant,
4014 cli->cl_avail_grant, lost_grant);
4015 CDEBUG(D_RPCTRACE, "ocd_connect_flags: "LPX64" ocd_version: %d"
4016 " ocd_grant: %d\n", data->ocd_connect_flags,
4017 data->ocd_version, data->ocd_grant);
4023 static int osc_disconnect(struct obd_export *exp)
4025 struct obd_device *obd = class_exp2obd(exp);
4026 struct llog_ctxt *ctxt;
4029 ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
4031 if (obd->u.cli.cl_conn_count == 1) {
4032 /* Flush any remaining cancel messages out to the
4034 llog_sync(ctxt, exp);
4036 llog_ctxt_put(ctxt);
4038 CDEBUG(D_HA, "No LLOG_SIZE_REPL_CTXT found in obd %p\n",
4042 osc_del_shrink_grant(&obd->u.cli);
4043 rc = client_disconnect_export(exp);
4047 static int osc_import_event(struct obd_device *obd,
4048 struct obd_import *imp,
4049 enum obd_import_event event)
4051 struct client_obd *cli;
4055 LASSERT(imp->imp_obd == obd);
4058 case IMP_EVENT_DISCON: {
4059 /* Only do this on the MDS OSC's */
4060 if (imp->imp_server_timeout) {
4061 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4063 spin_lock(&oscc->oscc_lock);
4064 oscc->oscc_flags |= OSCC_FLAG_RECOVERING;
4065 spin_unlock(&oscc->oscc_lock);
4068 client_obd_list_lock(&cli->cl_loi_list_lock);
4069 cli->cl_avail_grant = 0;
4070 cli->cl_lost_grant = 0;
4071 client_obd_list_unlock(&cli->cl_loi_list_lock);
4072 ptlrpc_import_setasync(imp, -1);
4076 case IMP_EVENT_INACTIVE: {
4077 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
4080 case IMP_EVENT_INVALIDATE: {
4081 struct ldlm_namespace *ns = obd->obd_namespace;
4085 client_obd_list_lock(&cli->cl_loi_list_lock);
4086 /* all pages go to failing rpcs due to the invalid import */
4087 osc_check_rpcs(cli);
4088 client_obd_list_unlock(&cli->cl_loi_list_lock);
4090 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
4094 case IMP_EVENT_ACTIVE: {
4095 /* Only do this on the MDS OSC's */
4096 if (imp->imp_server_timeout) {
4097 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4099 spin_lock(&oscc->oscc_lock);
4100 oscc->oscc_flags &= ~OSCC_FLAG_NOSPC;
4101 spin_unlock(&oscc->oscc_lock);
4103 CDEBUG(D_INFO, "notify server \n");
4104 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
4107 case IMP_EVENT_OCD: {
4108 struct obd_connect_data *ocd = &imp->imp_connect_data;
4110 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
4111 osc_init_grant(&obd->u.cli, ocd);
4114 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
4115 imp->imp_client->cli_request_portal =OST_REQUEST_PORTAL;
4117 ptlrpc_import_setasync(imp, 1);
4118 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
4122 CERROR("Unknown import event %d\n", event);
4128 int osc_setup(struct obd_device *obd, obd_count len, void *buf)
4134 rc = ptlrpcd_addref();
4138 rc = client_obd_setup(obd, len, buf);
4142 struct lprocfs_static_vars lvars = { 0 };
4143 struct client_obd *cli = &obd->u.cli;
4145 lprocfs_osc_init_vars(&lvars);
4146 if (lprocfs_obd_setup(obd, lvars.obd_vars) == 0) {
4147 lproc_osc_attach_seqstat(obd);
4148 ptlrpc_lprocfs_register_obd(obd);
4152 /* We need to allocate a few requests more, because
4153 brw_interpret tries to create new requests before freeing
4154 previous ones. Ideally we want to have 2x max_rpcs_in_flight
4155 reserved, but I afraid that might be too much wasted RAM
4156 in fact, so 2 is just my guess and still should work. */
4157 cli->cl_import->imp_rq_pool =
4158 ptlrpc_init_rq_pool(cli->cl_max_rpcs_in_flight + 2,
4160 ptlrpc_add_rqs_to_pool);
4161 cli->cl_cache = cache_create(obd);
4162 if (!cli->cl_cache) {
4166 CFS_INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
4167 sema_init(&cli->cl_grant_sem, 1);
4173 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
4179 case OBD_CLEANUP_EARLY: {
4180 struct obd_import *imp;
4181 imp = obd->u.cli.cl_import;
4182 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
4183 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
4184 ptlrpc_deactivate_import(imp);
4187 case OBD_CLEANUP_EXPORTS: {
4188 /* If we set up but never connected, the
4189 client import will not have been cleaned. */
4190 if (obd->u.cli.cl_import) {
4191 struct obd_import *imp;
4192 down_write(&obd->u.cli.cl_sem);
4193 imp = obd->u.cli.cl_import;
4194 CDEBUG(D_CONFIG, "%s: client import never connected\n",
4196 ptlrpc_invalidate_import(imp);
4197 if (imp->imp_rq_pool) {
4198 ptlrpc_free_rq_pool(imp->imp_rq_pool);
4199 imp->imp_rq_pool = NULL;
4201 class_destroy_import(imp);
4202 up_write(&obd->u.cli.cl_sem);
4203 obd->u.cli.cl_import = NULL;
4205 rc = obd_llog_finish(obd, 0);
4207 CERROR("failed to cleanup llogging subsystems\n");
4210 case OBD_CLEANUP_SELF_EXP:
4212 case OBD_CLEANUP_OBD:
4218 int osc_cleanup(struct obd_device *obd)
4220 struct osc_creator *oscc = &obd->u.cli.cl_oscc;
4224 ptlrpc_lprocfs_unregister_obd(obd);
4225 lprocfs_obd_cleanup(obd);
4227 spin_lock(&oscc->oscc_lock);
4228 oscc->oscc_flags &= ~OSCC_FLAG_RECOVERING;
4229 oscc->oscc_flags |= OSCC_FLAG_EXITING;
4230 spin_unlock(&oscc->oscc_lock);
4232 /* free memory of osc quota cache */
4233 lquota_cleanup(quota_interface, obd);
4235 cache_destroy(obd->u.cli.cl_cache);
4236 rc = client_obd_cleanup(obd);
4242 static int osc_register_page_removal_cb(struct obd_device *obd,
4243 obd_page_removal_cb_t func,
4244 obd_pin_extent_cb pin_cb)
4248 /* this server - not need init */
4252 return cache_add_extent_removal_cb(obd->u.cli.cl_cache, func,
4256 static int osc_unregister_page_removal_cb(struct obd_device *obd,
4257 obd_page_removal_cb_t func)
4260 return cache_del_extent_removal_cb(obd->u.cli.cl_cache, func);
4263 static int osc_register_lock_cancel_cb(struct obd_device *obd,
4264 obd_lock_cancel_cb cb)
4267 LASSERT(obd->u.cli.cl_ext_lock_cancel_cb == NULL);
4269 /* this server - not need init */
4273 obd->u.cli.cl_ext_lock_cancel_cb = cb;
4277 static int osc_unregister_lock_cancel_cb(struct obd_device *obd,
4278 obd_lock_cancel_cb cb)
4282 if (obd->u.cli.cl_ext_lock_cancel_cb != cb) {
4283 CERROR("Unregistering cancel cb %p, while only %p was "
4285 obd->u.cli.cl_ext_lock_cancel_cb);
4289 obd->u.cli.cl_ext_lock_cancel_cb = NULL;
4293 static int osc_process_config(struct obd_device *obd, obd_count len, void *buf)
4295 struct lustre_cfg *lcfg = buf;
4296 struct lprocfs_static_vars lvars = { 0 };
4299 lprocfs_osc_init_vars(&lvars);
4301 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars, lcfg, obd);
4305 struct obd_ops osc_obd_ops = {
4306 .o_owner = THIS_MODULE,
4307 .o_setup = osc_setup,
4308 .o_precleanup = osc_precleanup,
4309 .o_cleanup = osc_cleanup,
4310 .o_add_conn = client_import_add_conn,
4311 .o_del_conn = client_import_del_conn,
4312 .o_connect = client_connect_import,
4313 .o_reconnect = osc_reconnect,
4314 .o_disconnect = osc_disconnect,
4315 .o_statfs = osc_statfs,
4316 .o_statfs_async = osc_statfs_async,
4317 .o_packmd = osc_packmd,
4318 .o_unpackmd = osc_unpackmd,
4319 .o_precreate = osc_precreate,
4320 .o_create = osc_create,
4321 .o_destroy = osc_destroy,
4322 .o_getattr = osc_getattr,
4323 .o_getattr_async = osc_getattr_async,
4324 .o_setattr = osc_setattr,
4325 .o_setattr_async = osc_setattr_async,
4327 .o_brw_async = osc_brw_async,
4328 .o_prep_async_page = osc_prep_async_page,
4329 .o_reget_short_lock = osc_reget_short_lock,
4330 .o_release_short_lock = osc_release_short_lock,
4331 .o_queue_async_io = osc_queue_async_io,
4332 .o_set_async_flags = osc_set_async_flags,
4333 .o_queue_group_io = osc_queue_group_io,
4334 .o_trigger_group_io = osc_trigger_group_io,
4335 .o_teardown_async_page = osc_teardown_async_page,
4336 .o_punch = osc_punch,
4338 .o_enqueue = osc_enqueue,
4339 .o_match = osc_match,
4340 .o_change_cbdata = osc_change_cbdata,
4341 .o_cancel = osc_cancel,
4342 .o_cancel_unused = osc_cancel_unused,
4343 .o_join_lru = osc_join_lru,
4344 .o_iocontrol = osc_iocontrol,
4345 .o_get_info = osc_get_info,
4346 .o_set_info_async = osc_set_info_async,
4347 .o_import_event = osc_import_event,
4348 .o_llog_init = osc_llog_init,
4349 .o_llog_finish = osc_llog_finish,
4350 .o_process_config = osc_process_config,
4351 .o_register_page_removal_cb = osc_register_page_removal_cb,
4352 .o_unregister_page_removal_cb = osc_unregister_page_removal_cb,
4353 .o_register_lock_cancel_cb = osc_register_lock_cancel_cb,
4354 .o_unregister_lock_cancel_cb = osc_unregister_lock_cancel_cb,
4356 int __init osc_init(void)
4358 struct lprocfs_static_vars lvars = { 0 };
4362 lprocfs_osc_init_vars(&lvars);
4364 request_module("lquota");
4365 quota_interface = PORTAL_SYMBOL_GET(osc_quota_interface);
4366 lquota_init(quota_interface);
4367 init_obd_quota_ops(quota_interface, &osc_obd_ops);
4369 rc = class_register_type(&osc_obd_ops, lvars.module_vars,
4372 if (quota_interface)
4373 PORTAL_SYMBOL_PUT(osc_quota_interface);
4381 static void /*__exit*/ osc_exit(void)
4383 lquota_exit(quota_interface);
4384 if (quota_interface)
4385 PORTAL_SYMBOL_PUT(osc_quota_interface);
4387 class_unregister_type(LUSTRE_OSC_NAME);
4390 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
4391 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
4392 MODULE_LICENSE("GPL");
4394 cfs_module(osc, LUSTRE_VERSION_STRING, osc_init, osc_exit);
git://git.whamcloud.com / fs / lustre-release.git / blob