1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2002, 2003 Cluster File Systems, Inc.
5 * Author: Phil Schwan <phil@clusterfs.com>
6 * Peter Braam <braam@clusterfs.com>
7 * Mike Shaver <shaver@clusterfs.com>
9 * This file is part of Lustre, http://www.lustre.org.
11 * Lustre is free software; you can redistribute it and/or
12 * modify it under the terms of version 2 of the GNU General Public
13 * License as published by the Free Software Foundation.
15 * Lustre is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with Lustre; if not, write to the Free Software
22 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
26 # define EXPORT_SYMTAB
28 #define DEBUG_SUBSYSTEM S_LOV
30 #include <linux/slab.h>
31 #include <linux/module.h>
32 #include <linux/init.h>
33 #include <linux/slab.h>
34 #include <linux/pagemap.h>
35 #include <linux/seq_file.h>
36 #include <asm/div64.h>
38 #include <liblustre.h>
41 #include <linux/obd_support.h>
42 #include <linux/lustre_lib.h>
43 #include <linux/lustre_net.h>
44 #include <linux/lustre_idl.h>
45 #include <linux/lustre_dlm.h>
46 #include <linux/lustre_mds.h>
47 #include <linux/obd_class.h>
48 #include <linux/obd_lov.h>
49 #include <linux/obd_ost.h>
50 #include <linux/lprocfs_status.h>
52 #include "lov_internal.h"
55 #define MAX_STRING_SIZE 128
56 static int lov_connect_obd(struct obd_device *obd, struct lov_tgt_desc *tgt,
57 int activate, struct obd_connect_data *conn_data,
58 unsigned long connect_flags)
60 struct obd_uuid lov_osc_uuid = { "LOV_OSC_UUID" };
61 struct obd_uuid *tgt_uuid = &tgt->uuid;
64 struct proc_dir_entry *lov_proc_dir;
66 struct lov_obd *lov = &obd->u.lov;
67 struct lustre_handle conn = {0, };
68 struct obd_device *tgt_obd;
72 tgt_obd = class_find_client_obd(tgt_uuid, LUSTRE_OSC_NAME,
76 CERROR("Target %s not attached\n", tgt_uuid->uuid);
80 if (!tgt_obd->obd_set_up) {
81 CERROR("Target %s not set up\n", tgt_uuid->uuid);
86 tgt_obd->obd_no_recov = 0;
87 ptlrpc_activate_import(tgt_obd->u.cli.cl_import);
90 if (tgt_obd->u.cli.cl_import->imp_invalid) {
91 CERROR("not connecting OSC %s; administratively "
92 "disabled\n", tgt_uuid->uuid);
93 rc = obd_register_observer(tgt_obd, obd);
95 CERROR("Target %s register_observer error %d; "
96 "will not be able to reactivate\n",
102 rc = obd_connect(&conn, tgt_obd, &lov_osc_uuid, conn_data,
105 CERROR("Target %s connect error %d\n", tgt_uuid->uuid, rc);
108 tgt->ltd_exp = class_conn2export(&conn);
110 rc = obd_register_observer(tgt_obd, obd);
112 CERROR("Target %s register_observer error %d\n",
114 obd_disconnect(tgt->ltd_exp, 0);
120 lov->desc.ld_active_tgt_count++;
123 lov_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
125 struct obd_device *osc_obd = class_conn2obd(&conn);
126 struct proc_dir_entry *osc_symlink;
127 char name[MAX_STRING_SIZE + 1];
129 LASSERT(osc_obd != NULL);
130 LASSERT(osc_obd->obd_type != NULL);
131 LASSERT(osc_obd->obd_type->typ_name != NULL);
132 name[MAX_STRING_SIZE] = '\0';
133 snprintf(name, MAX_STRING_SIZE, "../../../%s/%s",
134 osc_obd->obd_type->typ_name,
136 osc_symlink = proc_symlink(osc_obd->obd_name, lov_proc_dir,
138 if (osc_symlink == NULL) {
139 CERROR("could not register LOV target "
140 "/proc/fs/lustre/%s/%s/target_obds/%s\n",
141 obd->obd_type->typ_name, obd->obd_name,
143 lprocfs_remove(lov_proc_dir);
152 static int lov_connect(struct lustre_handle *conn, struct obd_device *obd,
153 struct obd_uuid *cluuid, struct obd_connect_data *data,
157 struct proc_dir_entry *lov_proc_dir;
159 struct lov_obd *lov = &obd->u.lov;
160 struct lov_tgt_desc *tgt;
161 struct obd_export *exp;
165 rc = class_connect(conn, obd, cluuid);
169 exp = class_conn2export(conn);
171 /* We don't want to actually do the underlying connections more than
172 * once, so keep track. */
174 if (lov->refcount > 1) {
175 class_export_put(exp);
180 lov_proc_dir = lprocfs_register("target_obds", obd->obd_proc_entry,
182 if (IS_ERR(lov_proc_dir)) {
183 CERROR("could not register /proc/fs/lustre/%s/%s/target_obds.",
184 obd->obd_type->typ_name, obd->obd_name);
189 /* connect_flags is the MDS number, save for use in lov_add_obd */
190 lov->lov_connect_flags = flags;
191 for (i = 0, tgt = lov->tgts; i < lov->desc.ld_tgt_count; i++, tgt++) {
192 if (obd_uuid_empty(&tgt->uuid))
194 rc = lov_connect_obd(obd, tgt, 0, data, flags);
199 class_export_put(exp);
205 lprocfs_remove(lov_proc_dir);
209 struct obd_uuid uuid;
211 --lov->desc.ld_active_tgt_count;
213 /* save for CERROR below; (we know it's terminated) */
215 rc2 = obd_disconnect(tgt->ltd_exp, 0);
217 CERROR("error: LOV target %s disconnect on OST idx %d: "
218 "rc = %d\n", uuid.uuid, i, rc2);
220 class_disconnect(exp, 0);
224 static int lov_disconnect_obd(struct obd_device *obd,
225 struct lov_tgt_desc *tgt,
229 struct proc_dir_entry *lov_proc_dir;
231 struct obd_device *osc_obd = class_exp2obd(tgt->ltd_exp);
232 struct lov_obd *lov = &obd->u.lov;
237 lov_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
239 struct proc_dir_entry *osc_symlink;
241 osc_symlink = lprocfs_srch(lov_proc_dir, osc_obd->obd_name);
243 lprocfs_remove(osc_symlink);
245 CERROR("/proc/fs/lustre/%s/%s/target_obds/%s missing\n",
246 obd->obd_type->typ_name, obd->obd_name,
251 if (obd->obd_no_recov) {
252 /* Pass it on to our clients.
253 * XXX This should be an argument to disconnect,
254 * XXX not a back-door flag on the OBD. Ah well.
257 osc_obd->obd_no_recov = 1;
260 obd_register_observer(tgt->ltd_exp->exp_obd, NULL);
261 rc = obd_disconnect(tgt->ltd_exp, flags);
264 CERROR("Target %s disconnect error %d\n",
272 lov->desc.ld_active_tgt_count--;
278 static int lov_disconnect(struct obd_export *exp, unsigned long flags)
280 struct obd_device *obd = class_exp2obd(exp);
282 struct proc_dir_entry *lov_proc_dir;
284 struct lov_obd *lov = &obd->u.lov;
285 struct lov_tgt_desc *tgt;
292 /* Only disconnect the underlying layers on the final disconnect. */
294 if (lov->refcount != 0)
297 for (i = 0, tgt = lov->tgts; i < lov->desc.ld_tgt_count; i++, tgt++) {
299 lov_disconnect_obd(obd, tgt, flags);
303 lov_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
305 lprocfs_remove(lov_proc_dir);
307 CERROR("/proc/fs/lustre/%s/%s/target_obds missing.",
308 obd->obd_type->typ_name, obd->obd_name);
313 rc = class_disconnect(exp, 0);
319 * -EINVAL : UUID can't be found in the LOV's target list
320 * -ENOTCONN: The UUID is found, but the target connection is bad (!)
321 * -EBADF : The UUID is found, but the OBD is the wrong type (!)
323 static int lov_set_osc_active(struct lov_obd *lov, struct obd_uuid *uuid,
326 struct lov_tgt_desc *tgt;
330 CDEBUG(D_INFO, "Searching in lov %p for uuid %s (activate=%d)\n",
331 lov, uuid->uuid, activate);
333 spin_lock(&lov->lov_lock);
334 for (i = 0, tgt = lov->tgts; i < lov->desc.ld_tgt_count; i++, tgt++) {
335 if (tgt->ltd_exp == NULL)
338 CDEBUG(D_INFO, "lov idx %d is %s conn "LPX64"\n",
339 i, tgt->uuid.uuid, tgt->ltd_exp->exp_handle.h_cookie);
340 if (strncmp(uuid->uuid, tgt->uuid.uuid, sizeof uuid->uuid) == 0)
344 if (i == lov->desc.ld_tgt_count)
345 GOTO(out, rc = -EINVAL);
348 if (tgt->active == activate) {
349 CDEBUG(D_INFO, "OSC %s already %sactive!\n", uuid->uuid,
350 activate ? "" : "in");
354 CDEBUG(D_INFO, "Marking OSC %s %sactive\n", uuid->uuid,
355 activate ? "" : "in");
357 tgt->active = activate;
359 lov->desc.ld_active_tgt_count++;
361 lov->desc.ld_active_tgt_count--;
365 spin_unlock(&lov->lov_lock);
369 static int lov_notify(struct obd_device *obd, struct obd_device *watched,
370 int active, void *data)
372 struct obd_uuid *uuid;
376 if (strcmp(watched->obd_type->typ_name, LUSTRE_OSC_NAME)) {
377 CERROR("unexpected notification of %s %s!\n",
378 watched->obd_type->typ_name,
382 uuid = &watched->u.cli.cl_import->imp_target_uuid;
384 /* Set OSC as active before notifying the observer, so the
385 * observer can use the OSC normally.
387 rc = lov_set_osc_active(&obd->u.lov, uuid, active);
389 CERROR("%sactivation of %s failed: %d\n",
390 active ? "" : "de", uuid->uuid, rc);
394 if (obd->obd_observer)
395 /* Pass the notification up the chain. */
396 rc = obd_notify(obd->obd_observer, watched, active, data);
401 int lov_attach(struct obd_device *dev, obd_count len, void *data)
403 struct lprocfs_static_vars lvars;
406 lprocfs_init_vars(lov, &lvars);
407 rc = lprocfs_obd_attach(dev, lvars.obd_vars);
410 struct proc_dir_entry *entry;
412 entry = create_proc_entry("target_obd_status", 0444,
413 dev->obd_proc_entry);
417 entry->proc_fops = &lov_proc_target_fops;
425 int lov_detach(struct obd_device *dev)
427 return lprocfs_obd_detach(dev);
430 static int lov_setup(struct obd_device *obd, obd_count len, void *buf)
432 struct lov_obd *lov = &obd->u.lov;
433 struct lustre_cfg *lcfg = buf;
434 struct lov_desc *desc;
438 if (LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
439 CERROR("LOV setup requires a descriptor\n");
443 desc = (struct lov_desc *)lustre_cfg_string(lcfg, 1);
444 if (sizeof(*desc) > LUSTRE_CFG_BUFLEN(lcfg, 1)) {
445 CERROR("descriptor size wrong: %d > %d\n",
446 (int)sizeof(*desc), LUSTRE_CFG_BUFLEN(lcfg, 1));
450 /* Because of 64-bit divide/mod operations only work with a 32-bit
451 * divisor in a 32-bit kernel, we cannot support a stripe width
452 * of 4GB or larger on 32-bit CPUs.
455 count = desc->ld_default_stripe_count;
456 if (count && (count * desc->ld_default_stripe_size) > ~0UL) {
457 CERROR("LOV: stripe width "LPU64"x%u > %lu on 32-bit system\n",
458 desc->ld_default_stripe_size, count, ~0UL);
461 if (desc->ld_tgt_count > 0) {
462 lov->bufsize= sizeof(struct lov_tgt_desc) * desc->ld_tgt_count;
464 lov->bufsize = sizeof(struct lov_tgt_desc) * LOV_MAX_TGT_COUNT;
466 OBD_ALLOC(lov->tgts, lov->bufsize);
467 if (lov->tgts == NULL) {
469 CERROR("couldn't allocate %d bytes for target table.\n",
474 desc->ld_tgt_count = 0;
475 desc->ld_active_tgt_count = 0;
477 spin_lock_init(&lov->lov_lock);
478 sema_init(&lov->lov_llog_sem, 1);
483 static int lov_cleanup(struct obd_device *obd, int flags)
485 struct lov_obd *lov = &obd->u.lov;
487 OBD_FREE(lov->tgts, lov->bufsize);
492 lov_add_obd(struct obd_device *obd, struct obd_uuid *uuidp, int index, int gen)
494 struct lov_obd *lov = &obd->u.lov;
495 struct lov_tgt_desc *tgt;
499 CDEBUG(D_CONFIG, "uuid: %s idx: %d gen: %d\n",
500 uuidp->uuid, index, gen);
502 if ((index < 0) || (index >= LOV_MAX_TGT_COUNT)) {
503 CERROR("request to add OBD %s at invalid index: %d\n",
509 CERROR("request to add OBD %s with invalid generation: %d\n",
514 tgt = lov->tgts + index;
515 if (!obd_uuid_empty(&tgt->uuid)) {
516 CERROR("OBD already assigned at LOV target index %d\n",
522 /* XXX - add a sanity check on the generation number. */
525 if (index >= lov->desc.ld_tgt_count)
526 lov->desc.ld_tgt_count = index + 1;
528 CDEBUG(D_CONFIG, "idx: %d ltd_gen: %d ld_tgt_count: %d\n",
529 index, tgt->ltd_gen, lov->desc.ld_tgt_count);
531 if (lov->refcount == 0)
535 struct obd_device *osc_obd;
537 osc_obd = class_exp2obd(tgt->ltd_exp);
539 osc_obd->obd_no_recov = 0;
542 rc = lov_connect_obd(obd, tgt, 1, NULL, lov->lov_connect_flags);
546 if (obd->obd_observer) {
547 /* tell the mds_lov about the new target */
548 rc = obd_notify(obd->obd_observer, tgt->ltd_exp->exp_obd, 1,
554 if (rc && tgt->ltd_exp != NULL)
555 lov_disconnect_obd(obd, tgt, 0);
560 lov_del_obd(struct obd_device *obd, struct obd_uuid *uuidp, int index, int gen)
562 struct lov_obd *lov = &obd->u.lov;
563 struct lov_tgt_desc *tgt;
564 int count = lov->desc.ld_tgt_count;
568 CDEBUG(D_CONFIG, "uuid: %s idx: %d gen: %d\n",
569 uuidp->uuid, index, gen);
571 if (index >= count) {
572 CERROR("LOV target index %d >= number of LOV OBDs %d.\n",
577 tgt = lov->tgts + index;
579 if (obd_uuid_empty(&tgt->uuid)) {
580 CERROR("LOV target at index %d is not setup.\n", index);
584 if (strncmp(uuidp->uuid, tgt->uuid.uuid, sizeof uuidp->uuid) != 0) {
585 CERROR("LOV target UUID %s at index %d doesn't match %s.\n",
586 tgt->uuid.uuid, index, uuidp->uuid);
591 struct obd_device *osc_obd;
593 osc_obd = class_exp2obd(tgt->ltd_exp);
595 osc_obd->obd_no_recov = 1;
596 rc = obd_llog_finish(osc_obd, &osc_obd->obd_llogs, 1);
598 CERROR("osc_llog_finish error: %d\n", rc);
600 lov_disconnect_obd(obd, tgt, 0);
603 /* XXX - right now there is a dependency on ld_tgt_count being the
604 * maximum tgt index for computing the mds_max_easize. So we can't
607 /* lt_gen = 0 will mean it will not match the gen of any valid loi */
608 memset(tgt, 0, sizeof(*tgt));
610 CDEBUG(D_CONFIG, "uuid: %s idx: %d gen: %d exp: %p active: %d\n",
611 tgt->uuid.uuid, index, tgt->ltd_gen, tgt->ltd_exp, tgt->active);
616 static int lov_process_config(struct obd_device *obd, obd_count len, void *buf)
618 struct lustre_cfg *lcfg = buf;
619 struct obd_uuid obd_uuid;
626 switch(cmd = lcfg->lcfg_command) {
627 case LCFG_LOV_ADD_OBD:
628 case LCFG_LOV_DEL_OBD: {
629 if (LUSTRE_CFG_BUFLEN(lcfg, 1) > sizeof(obd_uuid.uuid))
630 GOTO(out, rc = -EINVAL);
632 obd_str2uuid(&obd_uuid, lustre_cfg_string(lcfg, 1));
634 if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", &index) != 1)
635 GOTO(out, rc = -EINVAL);
636 if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", &gen) != 1)
637 GOTO(out, rc = -EINVAL);
638 if (cmd == LCFG_LOV_ADD_OBD)
639 rc = lov_add_obd(obd, &obd_uuid, index, gen);
641 rc = lov_del_obd(obd, &obd_uuid, index, gen);
645 CERROR("Unknown command: %d\n", lcfg->lcfg_command);
646 GOTO(out, rc = -EINVAL);
655 #define log2(n) ffz(~(n))
658 static int lov_clear_orphans(struct obd_export *export, struct obdo *src_oa,
659 struct lov_stripe_md **ea,
660 struct obd_trans_info *oti)
664 struct obd_uuid *ost_uuid = NULL;
668 LASSERT(src_oa->o_valid & OBD_MD_FLFLAGS &&
669 src_oa->o_flags == OBD_FL_DELORPHAN);
671 lov = &export->exp_obd->u.lov;
673 tmp_oa = obdo_alloc();
677 if (src_oa->o_valid & OBD_MD_FLINLINE) {
678 ost_uuid = (struct obd_uuid *)src_oa->o_inline;
679 CDEBUG(D_HA, "clearing orphans only for %s\n",
683 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
684 struct lov_stripe_md obj_md;
685 struct lov_stripe_md *obj_mdp = &obj_md;
688 /* if called for a specific target, we don't
689 care if it is not active. */
690 if (lov->tgts[i].active == 0 && ost_uuid == NULL) {
691 CDEBUG(D_HA, "lov idx %d inactive\n", i);
695 if (ost_uuid && !obd_uuid_equals(ost_uuid, &lov->tgts[i].uuid))
698 memcpy(tmp_oa, src_oa, sizeof(*tmp_oa));
700 /* XXX: LOV STACKING: use real "obj_mdp" sub-data */
701 err = obd_create(lov->tgts[i].ltd_exp, tmp_oa, NULL, 0,
704 /* This export will be disabled until it is recovered,
705 and then orphan recovery will be completed. */
706 CERROR("error in orphan recovery on OST idx %d/%d: "
707 "rc = %d\n", i, lov->desc.ld_tgt_count, err);
716 static int lov_recreate(struct obd_export *exp, struct obdo *src_oa,
717 void *acl, int acl_size,
718 struct lov_stripe_md **ea, struct obd_trans_info *oti)
720 struct lov_stripe_md *obj_mdp, *lsm;
721 struct lov_obd *lov = &exp->exp_obd->u.lov;
726 LASSERT(src_oa->o_valid & OBD_MD_FLFLAGS &&
727 src_oa->o_flags & OBD_FL_RECREATE_OBJS);
729 OBD_ALLOC(obj_mdp, sizeof(*obj_mdp));
733 ost_idx = src_oa->o_nlink;
736 GOTO(out, rc = -EINVAL);
737 if (ost_idx >= lov->desc.ld_tgt_count)
738 GOTO(out, rc = -EINVAL);
740 for (i = 0; i < lsm->lsm_stripe_count; i++) {
741 if (lsm->lsm_oinfo[i].loi_ost_idx == ost_idx) {
742 if (lsm->lsm_oinfo[i].loi_id != src_oa->o_id)
743 GOTO(out, rc = -EINVAL);
747 if (i == lsm->lsm_stripe_count)
748 GOTO(out, rc = -EINVAL);
750 rc = obd_create(lov->tgts[ost_idx].ltd_exp, src_oa, acl, acl_size,
753 OBD_FREE(obj_mdp, sizeof(*obj_mdp));
757 /* the LOV expects oa->o_id to be set to the LOV object id */
758 static int lov_create(struct obd_export *exp, struct obdo *src_oa,
759 void *acl, int acl_size,
760 struct lov_stripe_md **ea, struct obd_trans_info *oti)
762 struct lov_request_set *set = NULL;
763 struct list_head *pos;
772 if ((src_oa->o_valid & OBD_MD_FLFLAGS) &&
773 src_oa->o_flags == OBD_FL_DELORPHAN) {
774 rc = lov_clear_orphans(exp, src_oa, ea, oti);
778 lov = &exp->exp_obd->u.lov;
779 if (!lov->desc.ld_active_tgt_count)
782 /* Recreate a specific object id at the given OST index */
783 if ((src_oa->o_valid & OBD_MD_FLFLAGS) &&
784 (src_oa->o_flags & OBD_FL_RECREATE_OBJS)) {
785 rc = lov_recreate(exp, src_oa, acl, acl_size, ea, oti);
789 rc = lov_prep_create_set(exp, ea, src_oa, oti, &set);
793 list_for_each (pos, &set->set_list) {
794 struct lov_request *req =
795 list_entry(pos, struct lov_request, rq_link);
797 /* XXX: LOV STACKING: use real "obj_mdp" sub-data */
798 rc = obd_create(lov->tgts[req->rq_idx].ltd_exp, req->rq_oa,
799 acl, acl_size, &req->rq_md, oti);
800 lov_update_create_set(set, req, rc);
802 rc = lov_fini_create_set(set, ea);
806 #define lsm_bad_magic(LSMP) \
808 struct lov_stripe_md *_lsm__ = (LSMP); \
811 CERROR("LOV requires striping ea\n"); \
813 } else if (_lsm__->lsm_magic != LOV_MAGIC) { \
814 CERROR("LOV striping magic bad %#x != %#x\n", \
815 _lsm__->lsm_magic, LOV_MAGIC); \
821 static int lov_destroy(struct obd_export *exp, struct obdo *oa,
822 struct lov_stripe_md *lsm, struct obd_trans_info *oti)
824 struct lov_request_set *set;
825 struct lov_request *req;
826 struct list_head *pos;
831 if (lsm_bad_magic(lsm))
834 if (!exp || !exp->exp_obd)
837 lov = &exp->exp_obd->u.lov;
838 rc = lov_prep_destroy_set(exp, oa, lsm, oti, &set);
842 list_for_each (pos, &set->set_list) {
844 req = list_entry(pos, struct lov_request, rq_link);
846 /* XXX update the cookie position */
847 oti->oti_logcookies = set->set_cookies + req->rq_stripe;
848 rc = obd_destroy(lov->tgts[req->rq_idx].ltd_exp, req->rq_oa,
850 err = lov_update_common_set(set, req, rc);
852 CERROR("error: destroying objid "LPX64" subobj "
853 LPX64" on OST idx %d: rc = %d\n",
854 set->set_oa->o_id, req->rq_oa->o_id,
860 lov_fini_destroy_set(set);
864 static int lov_getattr(struct obd_export *exp, struct obdo *oa,
865 struct lov_stripe_md *lsm)
867 struct lov_request_set *set;
868 struct lov_request *req;
869 struct list_head *pos;
874 if (lsm_bad_magic(lsm))
877 if (!exp || !exp->exp_obd)
880 lov = &exp->exp_obd->u.lov;
882 rc = lov_prep_getattr_set(exp, oa, lsm, &set);
886 list_for_each (pos, &set->set_list) {
887 req = list_entry(pos, struct lov_request, rq_link);
889 CDEBUG(D_INFO, "objid "LPX64"[%d] has subobj "LPX64" at idx "
890 "%u\n", oa->o_id, req->rq_stripe, req->rq_oa->o_id,
893 rc = obd_getattr(lov->tgts[req->rq_idx].ltd_exp,
895 err = lov_update_common_set(set, req, rc);
897 CERROR("error: getattr objid "LPX64" subobj "
898 LPX64" on OST idx %d: rc = %d\n",
899 set->set_oa->o_id, req->rq_oa->o_id,
905 rc = lov_fini_getattr_set(set);
911 static int lov_getattr_interpret(struct ptlrpc_request_set *rqset, void *data,
914 struct lov_request_set *lovset = (struct lov_request_set *)data;
917 /* don't do attribute merge if this aysnc op failed */
919 lovset->set_completes = 0;
920 lov_fini_getattr_set(lovset);
922 rc = lov_fini_getattr_set(lovset);
927 static int lov_getattr_async(struct obd_export *exp, struct obdo *oa,
928 struct lov_stripe_md *lsm,
929 struct ptlrpc_request_set *rqset)
931 struct lov_request_set *lovset;
933 struct list_head *pos;
934 struct lov_request *req;
938 if (lsm_bad_magic(lsm))
941 if (!exp || !exp->exp_obd)
944 lov = &exp->exp_obd->u.lov;
946 rc = lov_prep_getattr_set(exp, oa, lsm, &lovset);
950 CDEBUG(D_INFO, "objid "LPX64": %ux%u byte stripes\n",
951 lsm->lsm_object_id, lsm->lsm_stripe_count, lsm->lsm_stripe_size);
953 list_for_each (pos, &lovset->set_list) {
954 req = list_entry(pos, struct lov_request, rq_link);
956 CDEBUG(D_INFO, "objid "LPX64"[%d] has subobj "LPX64" at idx "
957 "%u\n", oa->o_id, req->rq_stripe, req->rq_oa->o_id,
959 rc = obd_getattr_async(lov->tgts[req->rq_idx].ltd_exp,
960 req->rq_oa, NULL, rqset);
962 CERROR("error: getattr objid "LPX64" subobj "
963 LPX64" on OST idx %d: rc = %d\n",
964 lovset->set_oa->o_id, req->rq_oa->o_id,
968 lov_update_common_set(lovset, req, rc);
972 LASSERT (rqset->set_interpret == NULL);
973 rqset->set_interpret = lov_getattr_interpret;
974 rqset->set_arg = (void *)lovset;
978 lov_fini_getattr_set(lovset);
982 static int lov_setattr(struct obd_export *exp, struct obdo *src_oa,
983 struct lov_stripe_md *lsm, struct obd_trans_info *oti)
985 struct lov_request_set *set;
987 struct list_head *pos;
988 struct lov_request *req;
992 if (lsm_bad_magic(lsm))
995 if (!exp || !exp->exp_obd)
998 /* for now, we only expect time updates here */
999 LASSERT(!(src_oa->o_valid & ~(OBD_MD_FLID|OBD_MD_FLTYPE | OBD_MD_FLMODE|
1000 OBD_MD_FLATIME | OBD_MD_FLMTIME |
1001 OBD_MD_FLCTIME | OBD_MD_FLFLAGS |
1002 OBD_MD_FLSIZE | OBD_MD_FLGROUP)));
1004 LASSERT(!(src_oa->o_valid & OBD_MD_FLGROUP) || src_oa->o_gr > 0);
1006 lov = &exp->exp_obd->u.lov;
1007 rc = lov_prep_setattr_set(exp, src_oa, lsm, NULL, &set);
1011 list_for_each (pos, &set->set_list) {
1012 req = list_entry(pos, struct lov_request, rq_link);
1014 rc = obd_setattr(lov->tgts[req->rq_idx].ltd_exp, req->rq_oa,
1016 err = lov_update_common_set(set, req, rc);
1018 CERROR("error: setattr objid "LPX64" subobj "
1019 LPX64" on OST idx %d: rc = %d\n",
1020 set->set_oa->o_id, req->rq_oa->o_id,
1026 err = lov_fini_setattr_set(set);
1032 static int lov_revalidate_policy(struct lov_obd *lov, struct lov_stripe_md *lsm)
1034 static int next_idx = 0;
1035 struct lov_tgt_desc *tgt;
1038 /* XXX - we should do something clever and take lsm
1039 * into account but just do round robin for now. */
1041 /* last_idx must always be less that count because
1042 * ld_tgt_count currently cannot shrink. */
1043 count = lov->desc.ld_tgt_count;
1045 for (i = next_idx, tgt = lov->tgts + i; i < count; i++, tgt++) {
1047 next_idx = (i + 1) % count;
1052 for (i = 0, tgt = lov->tgts; i < next_idx; i++, tgt++) {
1054 next_idx = (i + 1) % count;
1062 static int lov_revalidate_md(struct obd_export *exp, struct obdo *src_oa,
1063 struct lov_stripe_md *ea,
1064 struct obd_trans_info *oti)
1066 struct obd_export *osc_exp;
1067 struct lov_obd *lov = &exp->exp_obd->u.lov;
1068 struct lov_stripe_md *lsm = ea;
1069 struct lov_stripe_md obj_md;
1070 struct lov_stripe_md *obj_mdp = &obj_md;
1071 struct lov_oinfo *loi;
1072 struct obdo *tmp_oa;
1073 int ost_idx, updates = 0, i;
1076 tmp_oa = obdo_alloc();
1080 loi = lsm->lsm_oinfo;
1081 for (i = 0; i < lsm->lsm_stripe_count; i++, loi++) {
1083 if (!obd_uuid_empty(&lov->tgts[loi->loi_ost_idx].uuid))
1086 ost_idx = lov_revalidate_policy(lov, lsm);
1088 /* FIXME: punt for now. */
1089 CERROR("lov_revalidate_policy failed; no active "
1094 /* create a new object */
1095 memcpy(tmp_oa, src_oa, sizeof(*tmp_oa));
1096 /* XXX: LOV STACKING: use real "obj_mdp" sub-data */
1097 osc_exp = lov->tgts[ost_idx].ltd_exp;
1098 rc = obd_create(osc_exp, tmp_oa, NULL, 0, &obj_mdp, oti);
1100 CERROR("error creating new subobj at idx %d; "
1101 "rc = %d\n", ost_idx, rc);
1105 oti->oti_objid[ost_idx] = tmp_oa->o_id;
1106 loi->loi_id = tmp_oa->o_id;
1107 loi->loi_gr = tmp_oa->o_gr;
1108 loi->loi_ost_idx = ost_idx;
1109 loi->loi_ost_gen = lov->tgts[ost_idx].ltd_gen;
1110 CDEBUG(D_INODE, "replacing objid "LPX64" subobj "LPX64
1111 " with idx %d gen %d.\n", lsm->lsm_object_id,
1112 loi->loi_id, ost_idx, loi->loi_ost_gen);
1116 /* If we got an error revalidating an entry there's no need to
1117 * cleanup up objects we allocated here because the bad entry
1118 * still points to a deleted OST. */
1124 /* FIXME: maybe we'll just make one node the authoritative attribute node, then
1125 * we can send this 'punch' to just the authoritative node and the nodes
1126 * that the punch will affect. */
1127 static int lov_punch(struct obd_export *exp, struct obdo *oa,
1128 struct lov_stripe_md *lsm,
1129 obd_off start, obd_off end, struct obd_trans_info *oti)
1131 struct lov_request_set *set;
1132 struct lov_obd *lov;
1133 struct list_head *pos;
1134 struct lov_request *req;
1135 int err = 0, rc = 0;
1138 if (lsm_bad_magic(lsm))
1141 if (!exp || !exp->exp_obd)
1144 lov = &exp->exp_obd->u.lov;
1145 rc = lov_prep_punch_set(exp, oa, lsm, start, end, oti, &set);
1149 list_for_each (pos, &set->set_list) {
1150 req = list_entry(pos, struct lov_request, rq_link);
1152 rc = obd_punch(lov->tgts[req->rq_idx].ltd_exp, req->rq_oa,
1153 NULL, req->rq_extent.start,
1154 req->rq_extent.end, NULL);
1155 err = lov_update_punch_set(set, req, rc);
1157 CERROR("error: punch objid "LPX64" subobj "LPX64
1158 " on OST idx %d: rc = %d\n", set->set_oa->o_id,
1159 req->rq_oa->o_id, req->rq_idx, rc);
1164 err = lov_fini_punch_set(set);
1170 static int lov_sync(struct obd_export *exp, struct obdo *oa,
1171 struct lov_stripe_md *lsm, obd_off start, obd_off end)
1173 struct lov_request_set *set;
1174 struct lov_obd *lov;
1175 struct list_head *pos;
1176 struct lov_request *req;
1177 int err = 0, rc = 0;
1180 if (lsm_bad_magic(lsm))
1186 lov = &exp->exp_obd->u.lov;
1187 rc = lov_prep_sync_set(exp, oa, lsm, start, end, &set);
1191 list_for_each (pos, &set->set_list) {
1192 req = list_entry(pos, struct lov_request, rq_link);
1194 rc = obd_sync(lov->tgts[req->rq_idx].ltd_exp, req->rq_oa,
1195 NULL, req->rq_extent.start, req->rq_extent.end);
1196 err = lov_update_common_set(set, req, rc);
1198 CERROR("error: fsync objid "LPX64" subobj "LPX64
1199 " on OST idx %d: rc = %d\n", set->set_oa->o_id,
1200 req->rq_oa->o_id, req->rq_idx, rc);
1205 err = lov_fini_sync_set(set);
1211 static int lov_brw_check(struct lov_obd *lov, struct obdo *oa,
1212 struct lov_stripe_md *lsm,
1213 obd_count oa_bufs, struct brw_page *pga)
1218 /* The caller just wants to know if there's a chance that this
1219 * I/O can succeed */
1220 for (i = 0; i < oa_bufs; i++) {
1221 int stripe = lov_stripe_number(lsm, pga[i].disk_offset);
1222 int ost = lsm->lsm_oinfo[stripe].loi_ost_idx;
1225 if (!lov_stripe_intersects(lsm, i, pga[i].disk_offset,
1226 pga[i].disk_offset + pga[i].count,
1230 if (lov->tgts[ost].active == 0) {
1231 CDEBUG(D_HA, "lov idx %d inactive\n", ost);
1234 rc = obd_brw(OBD_BRW_CHECK, lov->tgts[ost].ltd_exp, oa,
1235 NULL, 1, &pga[i], NULL);
1242 static int lov_brw(int cmd, struct obd_export *exp, struct obdo *src_oa,
1243 struct lov_stripe_md *lsm, obd_count oa_bufs,
1244 struct brw_page *pga, struct obd_trans_info *oti)
1246 struct lov_request_set *set;
1247 struct lov_request *req;
1248 struct list_head *pos;
1249 struct lov_obd *lov = &exp->exp_obd->u.lov;
1253 if (lsm_bad_magic(lsm))
1256 if (cmd == OBD_BRW_CHECK) {
1257 rc = lov_brw_check(lov, src_oa, lsm, oa_bufs, pga);
1261 rc = lov_prep_brw_set(exp, src_oa, lsm, oa_bufs, pga, oti, &set);
1265 list_for_each (pos, &set->set_list) {
1266 struct obd_export *sub_exp;
1267 struct brw_page *sub_pga;
1268 req = list_entry(pos, struct lov_request, rq_link);
1270 sub_exp = lov->tgts[req->rq_idx].ltd_exp;
1271 sub_pga = set->set_pga + req->rq_pgaidx;
1272 rc = obd_brw(cmd, sub_exp, req->rq_oa, req->rq_md,
1273 req->rq_oabufs, sub_pga, oti);
1276 lov_update_common_set(set, req, rc);
1279 err = lov_fini_brw_set(set);
1285 static int lov_brw_interpret(struct ptlrpc_request_set *reqset, void *data,
1288 struct lov_request_set *lovset = (struct lov_request_set *)data;
1292 lovset->set_completes = 0;
1293 lov_fini_brw_set(lovset);
1295 rc = lov_fini_brw_set(lovset);
1301 static int lov_brw_async(int cmd, struct obd_export *exp, struct obdo *oa,
1302 struct lov_stripe_md *lsm, obd_count oa_bufs,
1303 struct brw_page *pga, struct ptlrpc_request_set *set,
1304 struct obd_trans_info *oti)
1306 struct lov_request_set *lovset;
1307 struct lov_request *req;
1308 struct list_head *pos;
1309 struct lov_obd *lov = &exp->exp_obd->u.lov;
1313 if (lsm_bad_magic(lsm))
1316 if (cmd == OBD_BRW_CHECK) {
1317 rc = lov_brw_check(lov, oa, lsm, oa_bufs, pga);
1321 rc = lov_prep_brw_set(exp, oa, lsm, oa_bufs, pga, oti, &lovset);
1325 list_for_each (pos, &lovset->set_list) {
1326 struct obd_export *sub_exp;
1327 struct brw_page *sub_pga;
1328 req = list_entry(pos, struct lov_request, rq_link);
1330 sub_exp = lov->tgts[req->rq_idx].ltd_exp;
1331 sub_pga = lovset->set_pga + req->rq_pgaidx;
1332 rc = obd_brw_async(cmd, sub_exp, req->rq_oa, req->rq_md,
1333 req->rq_oabufs, sub_pga, set, oti);
1336 lov_update_common_set(lovset, req, rc);
1339 LASSERT(set->set_interpret == NULL);
1340 set->set_interpret = (set_interpreter_func)lov_brw_interpret;
1341 set->set_arg = (void *)lovset;
1345 lov_fini_brw_set(lovset);
1349 static int lov_ap_make_ready(void *data, int cmd)
1351 struct lov_async_page *lap = LAP_FROM_COOKIE(data);
1353 return lap->lap_caller_ops->ap_make_ready(lap->lap_caller_data, cmd);
1355 static int lov_ap_refresh_count(void *data, int cmd)
1357 struct lov_async_page *lap = LAP_FROM_COOKIE(data);
1359 return lap->lap_caller_ops->ap_refresh_count(lap->lap_caller_data,
1362 static void lov_ap_fill_obdo(void *data, int cmd, struct obdo *oa)
1364 struct lov_async_page *lap = LAP_FROM_COOKIE(data);
1366 lap->lap_caller_ops->ap_fill_obdo(lap->lap_caller_data, cmd, oa);
1367 /* XXX woah, shouldn't we be altering more here? size? */
1368 oa->o_id = lap->lap_loi_id;
1371 static void lov_ap_completion(void *data, int cmd, struct obdo *oa, int rc)
1373 struct lov_async_page *lap = LAP_FROM_COOKIE(data);
1375 /* in a raid1 regime this would down a count of many ios
1376 * in flight, onl calling the caller_ops completion when all
1377 * the raid1 ios are complete */
1378 lap->lap_caller_ops->ap_completion(lap->lap_caller_data, cmd, oa, rc);
1381 static struct obd_async_page_ops lov_async_page_ops = {
1382 .ap_make_ready = lov_ap_make_ready,
1383 .ap_refresh_count = lov_ap_refresh_count,
1384 .ap_fill_obdo = lov_ap_fill_obdo,
1385 .ap_completion = lov_ap_completion,
1388 static int lov_prep_async_page(struct obd_export *exp,
1389 struct lov_stripe_md *lsm,
1390 struct lov_oinfo *loi, struct page *page,
1391 obd_off offset, struct obd_async_page_ops *ops,
1392 void *data, void **res)
1394 struct lov_obd *lov = &exp->exp_obd->u.lov;
1395 struct lov_async_page *lap;
1399 if (lsm_bad_magic(lsm))
1401 LASSERT(loi == NULL);
1403 stripe = lov_stripe_number(lsm, offset);
1404 loi = &lsm->lsm_oinfo[stripe];
1406 if (obd_uuid_empty(&lov->tgts[loi->loi_ost_idx].uuid))
1408 if (lov->tgts[loi->loi_ost_idx].active == 0)
1410 if (lov->tgts[loi->loi_ost_idx].ltd_exp == NULL) {
1411 CERROR("ltd_exp == NULL, but OST idx %d doesn't appear to be "
1412 "deleted or inactive.\n", loi->loi_ost_idx);
1416 OBD_ALLOC(lap, sizeof(*lap));
1420 lap->lap_magic = LAP_MAGIC;
1421 lap->lap_caller_ops = ops;
1422 lap->lap_caller_data = data;
1424 /* FIXME handle multiple oscs after landing b_raid1 */
1425 lap->lap_stripe = stripe;
1426 switch (lsm->lsm_pattern) {
1427 case LOV_PATTERN_RAID0:
1428 lov_stripe_offset(lsm, offset, lap->lap_stripe,
1429 &lap->lap_sub_offset);
1431 case LOV_PATTERN_CMOBD:
1432 lap->lap_sub_offset = offset;
1438 /* so the callback doesn't need the lsm */
1439 lap->lap_loi_id = loi->loi_id;
1441 rc = obd_prep_async_page(lov->tgts[loi->loi_ost_idx].ltd_exp,
1442 lsm, loi, page, lap->lap_sub_offset,
1443 &lov_async_page_ops, lap,
1444 &lap->lap_sub_cookie);
1446 OBD_FREE(lap, sizeof(*lap));
1449 CDEBUG(D_CACHE, "lap %p page %p cookie %p off "LPU64"\n", lap, page,
1450 lap->lap_sub_cookie, offset);
1455 static int lov_queue_async_io(struct obd_export *exp,
1456 struct lov_stripe_md *lsm,
1457 struct lov_oinfo *loi, void *cookie,
1458 int cmd, obd_off off, int count,
1459 obd_flags brw_flags, obd_flags async_flags)
1461 struct lov_obd *lov = &exp->exp_obd->u.lov;
1462 struct lov_async_page *lap;
1465 LASSERT(loi == NULL);
1467 if (lsm_bad_magic(lsm))
1470 lap = LAP_FROM_COOKIE(cookie);
1472 loi = &lsm->lsm_oinfo[lap->lap_stripe];
1474 rc = obd_queue_async_io(lov->tgts[loi->loi_ost_idx].ltd_exp, lsm,
1475 loi, lap->lap_sub_cookie, cmd, off, count,
1476 brw_flags, async_flags);
1480 static int lov_set_async_flags(struct obd_export *exp,
1481 struct lov_stripe_md *lsm,
1482 struct lov_oinfo *loi, void *cookie,
1483 obd_flags async_flags)
1485 struct lov_obd *lov = &exp->exp_obd->u.lov;
1486 struct lov_async_page *lap;
1489 LASSERT(loi == NULL);
1491 if (lsm_bad_magic(lsm))
1494 lap = LAP_FROM_COOKIE(cookie);
1496 loi = &lsm->lsm_oinfo[lap->lap_stripe];
1498 rc = obd_set_async_flags(lov->tgts[loi->loi_ost_idx].ltd_exp,
1499 lsm, loi, lap->lap_sub_cookie, async_flags);
1503 static int lov_queue_group_io(struct obd_export *exp,
1504 struct lov_stripe_md *lsm,
1505 struct lov_oinfo *loi,
1506 struct obd_io_group *oig, void *cookie,
1507 int cmd, obd_off off, int count,
1508 obd_flags brw_flags, obd_flags async_flags)
1510 struct lov_obd *lov = &exp->exp_obd->u.lov;
1511 struct lov_async_page *lap;
1514 LASSERT(loi == NULL);
1516 if (lsm_bad_magic(lsm))
1519 lap = LAP_FROM_COOKIE(cookie);
1521 loi = &lsm->lsm_oinfo[lap->lap_stripe];
1523 rc = obd_queue_group_io(lov->tgts[loi->loi_ost_idx].ltd_exp, lsm, loi,
1524 oig, lap->lap_sub_cookie, cmd, off, count,
1525 brw_flags, async_flags);
1529 /* this isn't exactly optimal. we may have queued sync io in oscs on
1530 * all stripes, but we don't record that fact at queue time. so we
1531 * trigger sync io on all stripes. */
1532 static int lov_trigger_group_io(struct obd_export *exp,
1533 struct lov_stripe_md *lsm,
1534 struct lov_oinfo *loi,
1535 struct obd_io_group *oig)
1537 struct lov_obd *lov = &exp->exp_obd->u.lov;
1540 LASSERT(loi == NULL);
1542 if (lsm_bad_magic(lsm))
1545 loi = lsm->lsm_oinfo;
1546 for (i = 0; i < lsm->lsm_stripe_count; i++, loi++) {
1547 if (lov->tgts[loi->loi_ost_idx].active == 0) {
1548 CDEBUG(D_HA, "lov idx %d inactive\n", loi->loi_ost_idx);
1552 err = obd_trigger_group_io(lov->tgts[loi->loi_ost_idx].ltd_exp,
1554 if (rc == 0 && err != 0)
1560 static int lov_teardown_async_page(struct obd_export *exp,
1561 struct lov_stripe_md *lsm,
1562 struct lov_oinfo *loi, void *cookie)
1564 struct lov_obd *lov = &exp->exp_obd->u.lov;
1565 struct lov_async_page *lap;
1568 LASSERT(loi == NULL);
1570 if (lsm_bad_magic(lsm))
1573 lap = LAP_FROM_COOKIE(cookie);
1575 loi = &lsm->lsm_oinfo[lap->lap_stripe];
1577 rc = obd_teardown_async_page(lov->tgts[loi->loi_ost_idx].ltd_exp,
1578 lsm, loi, lap->lap_sub_cookie);
1580 CERROR("unable to teardown sub cookie %p: %d\n",
1581 lap->lap_sub_cookie, rc);
1584 OBD_FREE(lap, sizeof(*lap));
1588 static int lov_enqueue(struct obd_export *exp, struct lov_stripe_md *lsm,
1589 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
1590 int *flags, void *bl_cb, void *cp_cb, void *gl_cb,
1591 void *data,__u32 lvb_len, void *lvb_swabber,
1592 struct lustre_handle *lockh)
1594 struct lov_request_set *set;
1595 struct lov_request *req;
1596 struct list_head *pos;
1597 struct lustre_handle *lov_lockhp;
1598 struct lov_obd *lov;
1600 int save_flags = *flags;
1603 if (lsm_bad_magic(lsm))
1606 /* we should never be asked to replay a lock this way. */
1607 LASSERT((*flags & LDLM_FL_REPLAY) == 0);
1609 if (!exp || !exp->exp_obd)
1612 lov = &exp->exp_obd->u.lov;
1613 rc = lov_prep_enqueue_set(exp, lsm, policy, mode, lockh, &set);
1617 list_for_each (pos, &set->set_list) {
1618 ldlm_policy_data_t sub_policy;
1619 req = list_entry(pos, struct lov_request, rq_link);
1620 lov_lockhp = set->set_lockh->llh_handles + req->rq_stripe;
1621 LASSERT(lov_lockhp);
1623 *flags = save_flags;
1624 sub_policy.l_extent.start = req->rq_extent.start;
1625 sub_policy.l_extent.end = req->rq_extent.end;
1627 rc = obd_enqueue(lov->tgts[req->rq_idx].ltd_exp, req->rq_md,
1628 type, &sub_policy, mode, flags, bl_cb,
1629 cp_cb, gl_cb, data, lvb_len, lvb_swabber,
1631 rc = lov_update_enqueue_set(set, req, rc, save_flags);
1636 lov_fini_enqueue_set(set, mode);
1640 static int lov_match(struct obd_export *exp, struct lov_stripe_md *lsm,
1641 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
1642 int *flags, void *data, struct lustre_handle *lockh)
1644 struct lov_request_set *set;
1645 struct lov_request *req;
1646 struct list_head *pos;
1647 struct lov_obd *lov = &exp->exp_obd->u.lov;
1648 struct lustre_handle *lov_lockhp;
1649 int lov_flags, rc = 0;
1652 if (lsm_bad_magic(lsm))
1655 if (!exp || !exp->exp_obd)
1658 lov = &exp->exp_obd->u.lov;
1659 rc = lov_prep_match_set(exp, lsm, policy, mode, lockh, &set);
1663 list_for_each (pos, &set->set_list) {
1664 ldlm_policy_data_t sub_policy;
1665 req = list_entry(pos, struct lov_request, rq_link);
1666 lov_lockhp = set->set_lockh->llh_handles + req->rq_stripe;
1667 LASSERT(lov_lockhp);
1669 sub_policy.l_extent.start = req->rq_extent.start;
1670 sub_policy.l_extent.end = req->rq_extent.end;
1673 rc = obd_match(lov->tgts[req->rq_idx].ltd_exp, req->rq_md,
1674 type, &sub_policy, mode, &lov_flags, data,
1676 rc = lov_update_match_set(set, req, rc);
1680 lov_fini_match_set(set, mode, *flags);
1684 static int lov_change_cbdata(struct obd_export *exp,
1685 struct lov_stripe_md *lsm, ldlm_iterator_t it,
1688 struct lov_obd *lov;
1689 struct lov_oinfo *loi;
1693 if (lsm_bad_magic(lsm))
1696 if (!exp || !exp->exp_obd)
1699 LASSERT(lsm->lsm_object_gr > 0);
1701 lov = &exp->exp_obd->u.lov;
1702 for (i = 0,loi = lsm->lsm_oinfo; i < lsm->lsm_stripe_count; i++,loi++) {
1703 struct lov_stripe_md submd;
1704 if (lov->tgts[loi->loi_ost_idx].active == 0) {
1705 CDEBUG(D_HA, "lov idx %d inactive\n", loi->loi_ost_idx);
1709 submd.lsm_object_id = loi->loi_id;
1710 submd.lsm_object_gr = lsm->lsm_object_gr;
1711 submd.lsm_stripe_count = 0;
1712 rc = obd_change_cbdata(lov->tgts[loi->loi_ost_idx].ltd_exp,
1718 static int lov_cancel(struct obd_export *exp, struct lov_stripe_md *lsm,
1719 __u32 mode, struct lustre_handle *lockh)
1721 struct lov_request_set *set;
1722 struct lov_request *req;
1723 struct list_head *pos;
1724 struct lov_obd *lov = &exp->exp_obd->u.lov;
1725 struct lustre_handle *lov_lockhp;
1726 int err = 0, rc = 0;
1729 if (lsm_bad_magic(lsm))
1732 if (!exp || !exp->exp_obd)
1735 LASSERT(lsm->lsm_object_gr > 0);
1738 lov = &exp->exp_obd->u.lov;
1739 rc = lov_prep_cancel_set(exp, lsm, mode, lockh, &set);
1743 list_for_each (pos, &set->set_list) {
1744 req = list_entry(pos, struct lov_request, rq_link);
1745 lov_lockhp = set->set_lockh->llh_handles + req->rq_stripe;
1747 rc = obd_cancel(lov->tgts[req->rq_idx].ltd_exp, req->rq_md,
1749 rc = lov_update_common_set(set, req, rc);
1751 CERROR("error: cancel objid "LPX64" subobj "
1752 LPX64" on OST idx %d: rc = %d\n",
1754 req->rq_md->lsm_object_id, req->rq_idx, rc);
1759 lov_fini_cancel_set(set);
1763 static int lov_cancel_unused(struct obd_export *exp,
1764 struct lov_stripe_md *lsm,
1765 int flags, void *opaque)
1767 struct lov_obd *lov;
1768 struct lov_oinfo *loi;
1772 lov = &exp->exp_obd->u.lov;
1774 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
1775 int err = obd_cancel_unused(lov->tgts[i].ltd_exp, NULL,
1783 if (lsm_bad_magic(lsm))
1786 if (!exp || !exp->exp_obd)
1789 LASSERT(lsm->lsm_object_gr > 0);
1791 for (i = 0,loi = lsm->lsm_oinfo; i < lsm->lsm_stripe_count; i++,loi++) {
1792 struct lov_stripe_md submd;
1795 if (lov->tgts[loi->loi_ost_idx].active == 0)
1796 CDEBUG(D_HA, "lov idx %d inactive\n", loi->loi_ost_idx);
1798 submd.lsm_object_id = loi->loi_id;
1799 submd.lsm_object_gr = lsm->lsm_object_gr;
1800 submd.lsm_stripe_count = 0;
1801 err = obd_cancel_unused(lov->tgts[loi->loi_ost_idx].ltd_exp,
1802 &submd, flags, opaque);
1803 if (err && lov->tgts[loi->loi_ost_idx].active) {
1804 CERROR("error: cancel unused objid "LPX64" subobj "LPX64
1805 " on OST idx %d: rc = %d\n", lsm->lsm_object_id,
1806 loi->loi_id, loi->loi_ost_idx, err);
1814 #define LOV_U64_MAX ((__u64)~0ULL)
1815 #define LOV_SUM_MAX(tot, add) \
1817 if ((tot) + (add) < (tot)) \
1818 (tot) = LOV_U64_MAX; \
1823 static int lov_statfs(struct obd_device *obd, struct obd_statfs *osfs,
1824 unsigned long max_age)
1826 struct lov_obd *lov = &obd->u.lov;
1827 struct obd_statfs lov_sfs;
1834 /* We only get block data from the OBD */
1835 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
1837 if (!lov->tgts[i].active) {
1838 CDEBUG(D_HA, "lov idx %d inactive\n", i);
1842 err = obd_statfs(class_exp2obd(lov->tgts[i].ltd_exp), &lov_sfs,
1845 if (lov->tgts[i].active && !rc)
1851 memcpy(osfs, &lov_sfs, sizeof(lov_sfs));
1854 osfs->os_bfree += lov_sfs.os_bfree;
1855 osfs->os_bavail += lov_sfs.os_bavail;
1856 osfs->os_blocks += lov_sfs.os_blocks;
1857 /* XXX not sure about this one - depends on policy.
1858 * - could be minimum if we always stripe on all OBDs
1859 * (but that would be wrong for any other policy,
1860 * if one of the OBDs has no more objects left)
1861 * - could be sum if we stripe whole objects
1862 * - could be average, just to give a nice number
1864 * To give a "reasonable" (if not wholly accurate)
1865 * number, we divide the total number of free objects
1866 * by expected stripe count (watch out for overflow).
1868 LOV_SUM_MAX(osfs->os_files, lov_sfs.os_files);
1869 LOV_SUM_MAX(osfs->os_ffree, lov_sfs.os_ffree);
1874 __u32 expected_stripes = lov->desc.ld_default_stripe_count ?
1875 lov->desc.ld_default_stripe_count :
1876 lov->desc.ld_active_tgt_count;
1878 if (osfs->os_files != LOV_U64_MAX)
1879 do_div(osfs->os_files, expected_stripes);
1880 if (osfs->os_ffree != LOV_U64_MAX)
1881 do_div(osfs->os_ffree, expected_stripes);
1888 static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
1889 void *karg, void *uarg)
1891 struct obd_device *obddev = class_exp2obd(exp);
1892 struct lov_obd *lov = &obddev->u.lov;
1893 int i, rc, count = lov->desc.ld_tgt_count;
1894 struct obd_uuid *uuidp;
1898 case OBD_IOC_LOV_GET_CONFIG: {
1899 struct obd_ioctl_data *data = karg;
1900 struct lov_tgt_desc *tgtdesc;
1901 struct lov_desc *desc;
1907 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
1910 data = (struct obd_ioctl_data *)buf;
1912 if (sizeof(*desc) > data->ioc_inllen1) {
1913 obd_ioctl_freedata(buf, len);
1917 if (sizeof(uuidp->uuid) * count > data->ioc_inllen2) {
1918 obd_ioctl_freedata(buf, len);
1922 if (sizeof(__u32) * count > data->ioc_inllen3) {
1923 obd_ioctl_freedata(buf, len);
1927 desc = (struct lov_desc *)data->ioc_inlbuf1;
1928 memcpy(desc, &(lov->desc), sizeof(*desc));
1930 uuidp = (struct obd_uuid *)data->ioc_inlbuf2;
1931 genp = (__u32 *)data->ioc_inlbuf3;
1932 tgtdesc = lov->tgts;
1933 /* the uuid will be empty for deleted OSTs */
1934 for (i = 0; i < count; i++, uuidp++, genp++, tgtdesc++) {
1935 obd_str2uuid(uuidp, tgtdesc->uuid.uuid);
1936 *genp = tgtdesc->ltd_gen;
1939 rc = copy_to_user((void *)uarg, buf, len);
1942 obd_ioctl_freedata(buf, len);
1945 case LL_IOC_LOV_SETSTRIPE:
1946 rc = lov_setstripe(exp, karg, uarg);
1948 case LL_IOC_LOV_GETSTRIPE:
1949 rc = lov_getstripe(exp, karg, uarg);
1951 case LL_IOC_LOV_SETEA:
1952 rc = lov_setea(exp, karg, uarg);
1959 for (i = 0; i < count; i++) {
1962 /* OST was deleted */
1963 if (obd_uuid_empty(&lov->tgts[i].uuid))
1966 err = obd_iocontrol(cmd, lov->tgts[i].ltd_exp,
1969 if (lov->tgts[i].active) {
1970 CERROR("error: iocontrol OSC %s on OST "
1971 "idx %d cmd %x: err = %d\n",
1972 lov->tgts[i].uuid.uuid, i,
1988 static int lov_get_info(struct obd_export *exp, __u32 keylen,
1989 void *key, __u32 *vallen, void *val)
1991 struct obd_device *obddev = class_exp2obd(exp);
1992 struct lov_obd *lov = &obddev->u.lov;
1996 if (!vallen || !val)
1999 if (keylen > strlen("lock_to_stripe") &&
2000 strcmp(key, "lock_to_stripe") == 0) {
2003 struct ldlm_lock *lock;
2004 struct lov_stripe_md *lsm;
2006 struct lov_oinfo *loi;
2007 struct ldlm_res_id *res_id = &data->lock->l_resource->lr_name;
2008 __u32 *stripe = val;
2010 if (*vallen < sizeof(*stripe))
2012 *vallen = sizeof(*stripe);
2014 /* XXX This is another one of those bits that will need to
2015 * change if we ever actually support nested LOVs. It uses
2016 * the lock's export to find out which stripe it is. */
2017 /* XXX - it's assumed all the locks for deleted OSTs have
2018 * been cancelled. Also, the export for deleted OSTs will
2019 * be NULL and won't match the lock's export. */
2020 for (i = 0, loi = data->lsm->lsm_oinfo;
2021 i < data->lsm->lsm_stripe_count;
2023 if (lov->tgts[loi->loi_ost_idx].ltd_exp ==
2024 data->lock->l_conn_export &&
2025 loi->loi_id == res_id->name[0] &&
2026 loi->loi_gr == res_id->name[2]) {
2031 LDLM_ERROR(data->lock, "lock on inode without such object\n");
2032 dump_lsm(D_ERROR, data->lsm);
2034 } else if (keylen >= strlen("size_to_stripe") &&
2035 strcmp(key, "size_to_stripe") == 0) {
2039 struct lov_stripe_md *lsm;
2042 if (*vallen < sizeof(*data))
2045 data->size = lov_size_to_stripe(data->lsm, data->size,
2046 data->stripe_number);
2048 } else if (keylen >= strlen("last_id") && strcmp(key, "last_id") == 0) {
2050 int rc, size = sizeof(obd_id);
2051 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
2052 if (!lov->tgts[i].active)
2054 rc = obd_get_info(lov->tgts[i].ltd_exp,
2055 keylen, key, &size, &(ids[i]));
2060 } else if (keylen >= strlen("lovdesc") && strcmp(key, "lovdesc") == 0) {
2061 struct lov_desc *desc_ret = val;
2062 *desc_ret = lov->desc;
2070 static int lov_set_info(struct obd_export *exp, obd_count keylen,
2071 void *key, obd_count vallen, void *val)
2073 struct obd_device *obddev = class_exp2obd(exp);
2074 struct lov_obd *lov = &obddev->u.lov;
2078 #define KEY_IS(str) \
2079 (keylen == strlen(str) && memcmp(key, str, keylen) == 0)
2081 if (KEY_IS("next_id")) {
2082 if (vallen != lov->desc.ld_tgt_count)
2084 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
2085 /* initialize all OSCs, even inactive ones */
2087 err = obd_set_info(lov->tgts[i].ltd_exp,
2088 keylen, key, sizeof(obd_id),
2089 ((obd_id*)val) + i);
2095 if (KEY_IS("async")) {
2096 struct lov_desc *desc = &lov->desc;
2097 struct lov_tgt_desc *tgts = lov->tgts;
2099 if (vallen != sizeof(int))
2101 lov->async = *((int*) val);
2103 for (i = 0; i < desc->ld_tgt_count; i++, tgts++) {
2104 struct obd_uuid *tgt_uuid = &tgts->uuid;
2105 struct obd_device *tgt_obd;
2107 tgt_obd = class_find_client_obd(tgt_uuid,
2111 CERROR("Target %s not attached\n",
2118 err = obd_set_info(tgt_obd->obd_self_export,
2119 keylen, key, vallen, val);
2121 CERROR("Failed to set async on target %s\n",
2130 if (KEY_IS("growth_count")) {
2131 if (vallen != sizeof(int))
2133 } else if (KEY_IS("mds_conn")) {
2134 if (vallen != sizeof(__u32))
2136 } else if (KEY_IS("unlinked") || KEY_IS("unrecovery")) {
2139 } else if (KEY_IS("sec")) {
2140 struct lov_tgt_desc *tgt;
2141 struct obd_export *exp;
2144 spin_lock(&lov->lov_lock);
2145 for (i = 0, tgt = lov->tgts; i < lov->desc.ld_tgt_count;
2148 /* during setup time the connections to osc might
2149 * haven't been established.
2152 struct obd_device *tgt_obd;
2154 tgt_obd = class_find_client_obd(&tgt->uuid,
2158 CERROR("can't set security flavor, "
2159 "device %s not attached?\n",
2164 exp = tgt_obd->obd_self_export;
2167 err = obd_set_info(exp, keylen, key, vallen, val);
2171 spin_unlock(&lov->lov_lock);
2178 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
2179 if (val && !obd_uuid_equals(val, &lov->tgts[i].uuid))
2182 if (!val && !lov->tgts[i].active)
2185 err = obd_set_info(lov->tgts[i].ltd_exp,
2186 keylen, key, vallen, val);
2195 struct lov_multi_wait {
2196 struct ldlm_lock *lock;
2202 int lov_complete_many(struct obd_export *exp, struct lov_stripe_md *lsm,
2203 struct lustre_handle *lockh)
2205 struct lov_lock_handles *lov_lockh = NULL;
2206 struct lustre_handle *lov_lockhp;
2207 struct lov_obd *lov;
2208 struct lov_oinfo *loi;
2209 struct lov_multi_wait *queues;
2213 if (lsm_bad_magic(lsm))
2216 if (!exp || !exp->exp_obd)
2219 LASSERT(lockh != NULL);
2220 if (lsm->lsm_stripe_count > 1) {
2221 lov_lockh = lov_handle2llh(lockh);
2222 if (lov_lockh == NULL) {
2223 CERROR("LOV: invalid lov lock handle %p\n", lockh);
2227 lov_lockhp = lov_lockh->llh_handles;
2232 OBD_ALLOC(queues, lsm->lsm_stripe_count * sizeof(*queues));
2234 GOTO(out, rc = -ENOMEM);
2236 lov = &exp->exp_obd->u.lov;
2237 for (i = 0, loi = lsm->lsm_oinfo; i < lsm->lsm_stripe_count;
2238 i++, loi++, lov_lockhp++) {
2239 struct ldlm_lock *lock;
2240 struct obd_device *obd;
2241 unsigned long irqflags;
2243 lock = ldlm_handle2lock(lov_lockhp);
2245 CDEBUG(D_HA, "lov idx %d subobj "LPX64" no lock?\n",
2246 loi->loi_ost_idx, loi->loi_id);
2247 queues[i].completed = 1;
2251 queues[i].lock = lock;
2252 init_waitqueue_entry(&(queues[i].wait), current);
2253 add_wait_queue(lock->l_waitq, &(queues[i].wait));
2255 obd = class_exp2obd(lock->l_conn_export);
2257 imp = obd->u.cli.cl_import;
2259 spin_lock_irqsave(&imp->imp_lock, irqflags);
2260 queues[i].generation = imp->imp_generation;
2261 spin_unlock_irqrestore(&imp->imp_lock, irqflags);
2265 lwi = LWI_TIMEOUT_INTR(obd_timeout * HZ, ldlm_expired_completion_wait,
2266 interrupted_completion_wait, &lwd);
2267 rc = l_wait_event_added(check_multi_complete(queues, lsm), &lwi);
2269 for (i = 0; i < lsm->lsm_stripe_count; i++)
2270 remove_wait_queue(lock->l_waitq, &(queues[i].wait));
2272 if (rc == -EINTR || rc == -ETIMEDOUT) {
2278 if (lov_lockh != NULL)
2279 lov_llh_put(lov_lockh);
2284 struct obd_ops lov_obd_ops = {
2285 .o_owner = THIS_MODULE,
2286 .o_attach = lov_attach,
2287 .o_detach = lov_detach,
2288 .o_setup = lov_setup,
2289 .o_cleanup = lov_cleanup,
2290 .o_process_config = lov_process_config,
2291 .o_connect = lov_connect,
2292 .o_disconnect = lov_disconnect,
2293 .o_statfs = lov_statfs,
2294 .o_packmd = lov_packmd,
2295 .o_unpackmd = lov_unpackmd,
2296 .o_revalidate_md = lov_revalidate_md,
2297 .o_create = lov_create,
2298 .o_destroy = lov_destroy,
2299 .o_getattr = lov_getattr,
2300 .o_getattr_async = lov_getattr_async,
2301 .o_setattr = lov_setattr,
2303 .o_brw_async = lov_brw_async,
2304 .o_prep_async_page = lov_prep_async_page,
2305 .o_queue_async_io = lov_queue_async_io,
2306 .o_set_async_flags = lov_set_async_flags,
2307 .o_queue_group_io = lov_queue_group_io,
2308 .o_trigger_group_io = lov_trigger_group_io,
2309 .o_teardown_async_page = lov_teardown_async_page,
2310 .o_adjust_kms = lov_adjust_kms,
2311 .o_punch = lov_punch,
2313 .o_enqueue = lov_enqueue,
2314 .o_match = lov_match,
2315 .o_change_cbdata = lov_change_cbdata,
2316 .o_cancel = lov_cancel,
2317 .o_cancel_unused = lov_cancel_unused,
2318 .o_iocontrol = lov_iocontrol,
2319 .o_get_info = lov_get_info,
2320 .o_set_info = lov_set_info,
2321 .o_llog_init = lov_llog_init,
2322 .o_llog_finish = lov_llog_finish,
2323 .o_notify = lov_notify,
2326 int __init lov_init(void)
2328 struct lprocfs_static_vars lvars;
2332 lprocfs_init_vars(lov, &lvars);
2333 rc = class_register_type(&lov_obd_ops, NULL, lvars.module_vars,
2334 OBD_LOV_DEVICENAME);
2339 static void /*__exit*/ lov_exit(void)
2341 class_unregister_type(OBD_LOV_DEVICENAME);
2344 MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
2345 MODULE_DESCRIPTION("Lustre Logical Object Volume OBD driver");
2346 MODULE_LICENSE("GPL");
2348 module_init(lov_init);
2349 module_exit(lov_exit);