/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * Copyright (C) 2002-2006 Cluster File Systems, Inc.
- * Author: Phil Schwan <phil@clusterfs.com>
- * Peter Braam <braam@clusterfs.com>
- * Mike Shaver <shaver@clusterfs.com>
- * Nathan Rutman <nathan@clusterfs.com>
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
*
- * This file is part of the Lustre file system, http://www.lustre.org
- * Lustre is a trademark of Cluster File Systems, Inc.
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
*
- * You may have signed or agreed to another license before downloading
- * this software. If so, you are bound by the terms and conditions
- * of that agreement, and the following does not apply to you. See the
- * LICENSE file included with this distribution for more information.
+ * GPL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved
+ * Use is subject to license terms.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
*
- * If you did not agree to a different license, then this copy of Lustre
- * is open source software; you can redistribute it and/or modify it
- * under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
+ * lustre/lov/lov_obd.c
*
- * In either case, Lustre is distributed in the hope that it will be
- * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
- * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * license text for more details.
+ * Author: Phil Schwan <phil@clusterfs.com>
+ * Author: Peter Braam <braam@clusterfs.com>
+ * Author: Mike Shaver <shaver@clusterfs.com>
+ * Author: Nathan Rutman <nathan@clusterfs.com>
*/
#ifndef EXPORT_SYMTAB
#include <lprocfs_status.h>
#include <lustre_param.h>
#include <lustre_cache.h>
+#include <lustre/ll_fiemap.h>
#include "lov_internal.h"
return rc;
}
+static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid,
+ int activate);
+
+static int lov_notify(struct obd_device *obd, struct obd_device *watched,
+ enum obd_notify_event ev, void *data)
+{
+ int rc = 0;
+ ENTRY;
+
+ if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE) {
+ struct obd_uuid *uuid;
+
+ LASSERT(watched);
+
+ if (strcmp(watched->obd_type->typ_name, LUSTRE_OSC_NAME)) {
+ CERROR("unexpected notification of %s %s!\n",
+ watched->obd_type->typ_name,
+ watched->obd_name);
+ RETURN(-EINVAL);
+ }
+ uuid = &watched->u.cli.cl_target_uuid;
+
+ /* Set OSC as active before notifying the observer, so the
+ * observer can use the OSC normally.
+ */
+ rc = lov_set_osc_active(obd, uuid, ev == OBD_NOTIFY_ACTIVE);
+ if (rc < 0) {
+ CERROR("%sactivation of %s failed: %d\n",
+ (ev == OBD_NOTIFY_ACTIVE) ? "" : "de",
+ obd_uuid2str(uuid), rc);
+ RETURN(rc);
+ }
+ /* active event should be pass lov target index as data */
+ data = &rc;
+ }
+
+ /* Pass the notification up the chain. */
+ if (watched) {
+ rc = obd_notify_observer(obd, watched, ev, data);
+ } else {
+ /* NULL watched means all osc's in the lov (only for syncs) */
+ /* sync event should be send lov idx as data */
+ struct lov_obd *lov = &obd->u.lov;
+ struct obd_device *tgt_obd;
+ int i;
+ lov_getref(obd);
+ for (i = 0; i < lov->desc.ld_tgt_count; i++) {
+ if (!lov->lov_tgts[i])
+ continue;
+ tgt_obd = class_exp2obd(lov->lov_tgts[i]->ltd_exp);
+
+ if ((ev == OBD_NOTIFY_SYNC) ||
+ (ev == OBD_NOTIFY_SYNC_NONBLOCK))
+ data = &i;
+
+ rc = obd_notify_observer(obd, tgt_obd, ev, data);
+ if (rc) {
+ CERROR("%s: notify %s of %s failed %d\n",
+ obd->obd_name,
+ obd->obd_observer->obd_name,
+ tgt_obd->obd_name, rc);
+ break;
+ }
+ }
+ lov_putref(obd);
+ }
+
+ RETURN(rc);
+}
+
#define MAX_STRING_SIZE 128
static int lov_connect_obd(struct obd_device *obd, __u32 index, int activate,
struct obd_connect_data *data)
/* Why should there ever be more than 1 connect? */
lov->lov_connects++;
LASSERT(lov->lov_connects == 1);
-
+
memset(&lov->lov_ocd, 0, sizeof(lov->lov_ocd));
if (data)
lov->lov_ocd = *data;
obd->obd_name, i, rc);
continue;
}
+ /* connect to administrative disabled ost */
+ if (!lov->lov_tgts[i]->ltd_exp)
+ continue;
+
+ rc = lov_notify(obd, lov->lov_tgts[i]->ltd_exp->exp_obd,
+ OBD_NOTIFY_ACTIVE, (void *)&i);
+ if (rc) {
+ CERROR("%s error sending notify %d\n",
+ obd->obd_name, rc);
+ }
}
lov_putref(obd);
-
+
RETURN(0);
}
* -EINVAL : UUID can't be found in the LOV's target list
* -ENOTCONN: The UUID is found, but the target connection is bad (!)
* -EBADF : The UUID is found, but the OBD is the wrong type (!)
+ * - any above 0 is lov index
*/
static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid,
int activate)
{
struct lov_obd *lov = &obd->u.lov;
struct lov_tgt_desc *tgt;
- int i, rc = 0;
+ int i = 0;
ENTRY;
CDEBUG(D_INFO, "Searching in lov %p for uuid %s (activate=%d)\n",
}
if (i == lov->desc.ld_tgt_count)
- GOTO(out, rc = -EINVAL);
+ GOTO(out, i = -EINVAL);
if (lov->lov_tgts[i]->ltd_active == activate) {
CDEBUG(D_INFO, "OSC %s already %sactive!\n", uuid->uuid,
activate ? "" : "in");
- GOTO(out, rc);
+ GOTO(out, i);
}
CDEBUG(D_CONFIG, "Marking OSC %s %sactive\n", obd_uuid2str(uuid),
out:
lov_putref(obd);
- RETURN(rc);
+ RETURN(i);
}
-static int lov_notify(struct obd_device *obd, struct obd_device *watched,
- enum obd_notify_event ev, void *data)
-{
- int rc = 0;
- ENTRY;
-
- if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE) {
- struct obd_uuid *uuid;
-
- LASSERT(watched);
-
- if (strcmp(watched->obd_type->typ_name, LUSTRE_OSC_NAME)) {
- CERROR("unexpected notification of %s %s!\n",
- watched->obd_type->typ_name,
- watched->obd_name);
- RETURN(-EINVAL);
- }
- uuid = &watched->u.cli.cl_target_uuid;
-
- /* Set OSC as active before notifying the observer, so the
- * observer can use the OSC normally.
- */
- rc = lov_set_osc_active(obd, uuid, ev == OBD_NOTIFY_ACTIVE);
- if (rc) {
- CERROR("%sactivation of %s failed: %d\n",
- (ev == OBD_NOTIFY_ACTIVE) ? "" : "de",
- obd_uuid2str(uuid), rc);
- RETURN(rc);
- }
- }
-
- /* Pass the notification up the chain. */
- if (watched) {
- rc = obd_notify_observer(obd, watched, ev, data);
- } else {
- /* NULL watched means all osc's in the lov (only for syncs) */
- struct lov_obd *lov = &obd->u.lov;
- struct obd_device *tgt_obd;
- int i;
- lov_getref(obd);
- for (i = 0; i < lov->desc.ld_tgt_count; i++) {
- if (!lov->lov_tgts[i])
- continue;
- tgt_obd = class_exp2obd(lov->lov_tgts[i]->ltd_exp);
- rc = obd_notify_observer(obd, tgt_obd, ev, data);
- if (rc) {
- CERROR("%s: notify %s of %s failed %d\n",
- obd->obd_name,
- obd->obd_observer->obd_name,
- tgt_obd->obd_name, rc);
- break;
- }
- }
- lov_putref(obd);
- }
-
- RETURN(rc);
-}
static int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp,
__u32 index, int gen, int active)
lov->lov_tgts, lov->lov_tgt_size);
}
-
OBD_ALLOC_PTR(tgt);
if (!tgt) {
mutex_up(&lov->lov_lock);
lov->lov_tgts[index] = tgt;
if (index >= lov->desc.ld_tgt_count)
lov->desc.ld_tgt_count = index + 1;
+
+ rc = lov_ost_pool_add(&lov->lov_packed, index, lov->lov_tgt_size);
+ if (rc)
+ RETURN(rc);
+
mutex_up(&lov->lov_lock);
CDEBUG(D_CONFIG, "idx=%d ltd_gen=%d ld_tgt_count=%d\n",
if (rc)
GOTO(out, rc);
+ /* connect to administrative disabled ost */
+ if (!tgt->ltd_exp)
+ GOTO(out, rc = 0);
+
rc = lov_notify(obd, tgt->ltd_exp->exp_obd,
active ? OBD_NOTIFY_ACTIVE : OBD_NOTIFY_INACTIVE,
(void *)&index);
* maximum tgt index for computing the mds_max_easize. So we can't
* shrink it. */
+ lov_ost_pool_remove(&lov->lov_packed, index);
lov->lov_tgts[index] = NULL;
- OBD_FREE_PTR(tgt);
+ OBD_FREE_PTR(tgt);
/* Manual cleanup - no cleanup logs to clean up the osc's. We must
do it ourselves. And we can't do it from lov_cleanup,
void lov_fix_desc_stripe_size(__u64 *val)
{
if (*val < PTLRPC_MAX_BRW_SIZE) {
- LCONSOLE_WARN("Increasing default stripe size to min %u\n",
- PTLRPC_MAX_BRW_SIZE);
+ if (*val)
+ LCONSOLE_WARN("Increasing default stripe size from "
+ LPU64" to %u\n",*val,PTLRPC_MAX_BRW_SIZE);
*val = PTLRPC_MAX_BRW_SIZE;
} else if (*val & (LOV_MIN_STRIPE_SIZE - 1)) {
*val &= ~(LOV_MIN_STRIPE_SIZE - 1);
struct lustre_cfg *lcfg = buf;
struct lov_desc *desc;
struct lov_obd *lov = &obd->u.lov;
- int count;
+ int rc;
ENTRY;
if (LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
lov_fix_desc(desc);
- /* Because of 64-bit divide/mod operations only work with a 32-bit
- * divisor in a 32-bit kernel, we cannot support a stripe width
- * of 4GB or larger on 32-bit CPUs. */
- count = desc->ld_default_stripe_count;
- if ((count > 0 ? count : desc->ld_tgt_count) *
- desc->ld_default_stripe_size > 0xffffffff) {
- CERROR("LOV: stripe width "LPU64"x%u > 4294967295 bytes\n",
- desc->ld_default_stripe_size, count);
- RETURN(-EINVAL);
- }
-
desc->ld_active_tgt_count = 0;
lov->desc = *desc;
lov->lov_tgt_size = 0;
+
sema_init(&lov->lov_lock, 1);
atomic_set(&lov->lov_refcount, 0);
CFS_INIT_LIST_HEAD(&lov->lov_qos.lq_oss_list);
init_rwsem(&lov->lov_qos.lq_rw_sem);
lov->lov_qos.lq_dirty = 1;
- lov->lov_qos.lq_dirty_rr = 1;
+ lov->lov_qos.lq_rr.lqr_dirty = 1;
lov->lov_qos.lq_reset = 1;
/* Default priority is toward free space balance */
lov->lov_qos.lq_prio_free = 232;
+ lov->lov_pools_hash_body = lustre_hash_init("POOLS", 7, 7,
+ &pool_hash_operations, 0);
+ CFS_INIT_LIST_HEAD(&lov->lov_pool_list);
+ lov->lov_pool_count = 0;
+ rc = lov_ost_pool_init(&lov->lov_packed, 0);
+ if (rc)
+ RETURN(rc);
+ rc = lov_ost_pool_init(&lov->lov_qos.lq_rr.lqr_pool, 0);
+ if (rc) {
+ lov_ost_pool_free(&lov->lov_packed);
+ RETURN(rc);
+ }
+
lprocfs_lov_init_vars(&lvars);
lprocfs_obd_setup(obd, lvars.obd_vars);
#ifdef LPROCFS
}
}
#endif
+ lov->lov_pool_proc_entry = lprocfs_register("pools",
+ obd->obd_proc_entry,
+ NULL, NULL);
RETURN(0);
}
static int lov_cleanup(struct obd_device *obd)
{
struct lov_obd *lov = &obd->u.lov;
+ struct list_head *pos, *tmp;
+ struct pool_desc *pool;
lprocfs_obd_cleanup(obd);
+
+ list_for_each_safe(pos, tmp, &lov->lov_pool_list) {
+ pool = list_entry(pos, struct pool_desc, pool_list);
+ /* free the pool structs */
+ lov_pool_del(obd, pool->pool_name);
+ }
+ lov_ost_pool_free(&(lov->lov_qos.lq_rr.lqr_pool));
+ lov_ost_pool_free(&lov->lov_packed);
+ lustre_hash_exit(lov->lov_pools_hash_body);
+
if (lov->lov_tgts) {
int i;
for (i = 0; i < lov->desc.ld_tgt_count; i++) {
lov->lov_tgt_size);
lov->lov_tgt_size = 0;
}
-
- if (lov->lov_qos.lq_rr_size)
- OBD_FREE(lov->lov_qos.lq_rr_array, lov->lov_qos.lq_rr_size);
RETURN(0);
}
lcfg, obd);
GOTO(out, rc);
}
+ case LCFG_POOL_NEW:
+ case LCFG_POOL_ADD:
+ case LCFG_POOL_DEL:
+ case LCFG_POOL_REM:
+ GOTO(out, rc);
+
default: {
CERROR("Unknown command: %d\n", lcfg->lcfg_command);
GOTO(out, rc = -EINVAL);
/* XXX: LOV STACKING: use real "obj_mdp" sub-data */
err = obd_create(lov->lov_tgts[i]->ltd_exp,
tmp_oa, &obj_mdp, oti);
- if (err)
+ if (err) {
/* This export will be disabled until it is recovered,
and then orphan recovery will be completed. */
CERROR("error in orphan recovery on OST idx %d/%d: "
"rc = %d\n", i, lov->desc.ld_tgt_count, err);
+ rc = err;
+ }
if (ost_uuid)
break;
if (!lov->desc.ld_active_tgt_count)
RETURN(-EIO);
+ lov_getref(exp->exp_obd);
/* Recreate a specific object id at the given OST index */
if ((src_oa->o_valid & OBD_MD_FLFLAGS) &&
(src_oa->o_flags & OBD_FL_RECREATE_OBJS)) {
rc = lov_recreate(exp, src_oa, ea, oti);
- RETURN(rc);
+ GOTO(out, rc);
}
maxage = cfs_time_shift_64(-lov->desc.ld_qos_maxage);
rc = lov_prep_create_set(exp, &oinfo, ea, src_oa, oti, &set);
if (rc)
- RETURN(rc);
+ GOTO(out, rc);
list_for_each_entry(req, &set->set_list, rq_link) {
/* XXX: LOV STACKING: use real "obj_mdp" sub-data */
lov_update_create_set(set, req, rc);
}
rc = lov_fini_create_set(set, ea);
+out:
+ lov_putref(exp->exp_obd);
RETURN(rc);
}
#define ASSERT_LSM_MAGIC(lsmp) \
do { \
LASSERT((lsmp) != NULL); \
- LASSERTF(((lsmp)->lsm_magic == LOV_MAGIC || \
+ LASSERTF(((lsmp)->lsm_magic == LOV_MAGIC_V1 || \
+ (lsmp)->lsm_magic == LOV_MAGIC_V3 || \
(lsmp)->lsm_magic == LOV_MAGIC_JOIN), "%p->lsm_magic=%x\n", \
(lsmp), (lsmp)->lsm_magic); \
} while (0)
struct lov_request *req;
struct list_head *pos;
struct lov_obd *lov;
- int rc = 0, err;
+ int rc = 0, err = 0;
ENTRY;
ASSERT_LSM_MAGIC(lsm);
}
lov = &exp->exp_obd->u.lov;
+ lov_getref(exp->exp_obd);
rc = lov_prep_destroy_set(exp, &oinfo, oa, lsm, oti, &set);
if (rc)
- RETURN(rc);
+ GOTO(out, rc);
list_for_each (pos, &set->set_list) {
- int err;
req = list_entry(pos, struct lov_request, rq_link);
if (oa->o_valid & OBD_MD_FLCOOKIE)
rc = lsm_op_find(lsm->lsm_magic)->lsm_destroy(lsm, oa, md_exp);
}
err = lov_fini_destroy_set(set);
+out:
+ lov_putref(exp->exp_obd);
RETURN(rc ? rc : err);
}
if (rc)
RETURN(rc);
- CDEBUG(D_INFO, "objid "LPX64": %ux%u byte stripes\n",
- oinfo->oi_md->lsm_object_id, oinfo->oi_md->lsm_stripe_count,
+ CDEBUG(D_INFO, "objid "LPX64"@"LPX64": %ux%u byte stripes\n",
+ oinfo->oi_md->lsm_object_id,
+ oinfo->oi_md->lsm_object_gr,
+ oinfo->oi_md->lsm_stripe_count,
oinfo->oi_md->lsm_stripe_size);
list_for_each (pos, &set->set_list) {
if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
oti->oti_logcookies = set->set_cookies + req->rq_stripe;
- CDEBUG(D_INFO, "objid "LPX64"[%d] has subobj "LPX64" at idx "
- "%u\n", oinfo->oi_oa->o_id, req->rq_stripe,
- req->rq_oi.oi_oa->o_id, req->rq_idx);
+ CDEBUG(D_INFO, "objid "LPX64"@"LPX64"[%d] has subobj "LPX64
+ " at idx %u\n", oinfo->oi_oa->o_id, oinfo->oi_oa->o_gr,
+ req->rq_stripe, req->rq_oi.oi_oa->o_id, req->rq_idx);
rc = obd_setattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp,
&req->rq_oi, oti, rqset);
RETURN(0);
}
-static int lov_sync(struct obd_export *exp, struct obdo *oa,
- struct lov_stripe_md *lsm, obd_off start, obd_off end)
+static int lov_sync_interpret(struct ptlrpc_request_set *rqset,
+ void *data, int rc)
{
- struct lov_request_set *set;
- struct obd_info oinfo;
+ struct lov_request_set *lovset = (struct lov_request_set *)data;
+ int err;
+ ENTRY;
+
+ if (rc)
+ lovset->set_completes = 0;
+ err = lov_fini_sync_set(lovset);
+ RETURN(rc ? rc : err);
+}
+
+static int lov_sync(struct obd_export *exp, struct obd_info *oinfo,
+ obd_off start, obd_off end,
+ struct ptlrpc_request_set *rqset)
+{
+ struct lov_request_set *set = NULL;
struct lov_obd *lov;
struct list_head *pos;
struct lov_request *req;
- int err = 0, rc = 0;
+ int rc = 0;
ENTRY;
- ASSERT_LSM_MAGIC(lsm);
+ ASSERT_LSM_MAGIC(oinfo->oi_md);
+ LASSERT(rqset != NULL);
if (!exp->exp_obd)
RETURN(-ENODEV);
lov = &exp->exp_obd->u.lov;
- rc = lov_prep_sync_set(exp, &oinfo, oa, lsm, start, end, &set);
+ rc = lov_prep_sync_set(exp, oinfo, start, end, &set);
if (rc)
RETURN(rc);
list_for_each (pos, &set->set_list) {
req = list_entry(pos, struct lov_request, rq_link);
- rc = obd_sync(lov->lov_tgts[req->rq_idx]->ltd_exp,
- req->rq_oi.oi_oa, NULL,
+ rc = obd_sync(lov->lov_tgts[req->rq_idx]->ltd_exp, &req->rq_oi,
req->rq_oi.oi_policy.l_extent.start,
- req->rq_oi.oi_policy.l_extent.end);
- err = lov_update_common_set(set, req, rc);
- if (err) {
+ req->rq_oi.oi_policy.l_extent.end, rqset);
+ if (rc) {
CERROR("error: fsync objid "LPX64" subobj "LPX64
" on OST idx %d: rc = %d\n",
set->set_oi->oi_oa->o_id,
req->rq_oi.oi_oa->o_id, req->rq_idx, rc);
- if (!rc)
- rc = err;
+ break;
}
}
- err = lov_fini_sync_set(set);
- if (!rc)
- rc = err;
- RETURN(rc);
+
+ /* If we are not waiting for responses on async requests, return. */
+ if (rc || list_empty(&rqset->set_requests)) {
+ int err = lov_fini_sync_set(set);
+ RETURN(rc ? rc : err);
+ }
+
+ LASSERT(rqset->set_interpret == NULL);
+ rqset->set_interpret = lov_sync_interpret;
+ rqset->set_arg = (void *)set;
+
+ RETURN(0);
}
static int lov_brw_check(struct lov_obd *lov, struct obd_info *lov_oinfo,
obd_off start, end;
if (!lov_stripe_intersects(lov_oinfo->oi_md, i, pga[i].off,
- pga[i].off + pga[i].count,
+ pga[i].off + pga[i].count - 1,
&start, &end))
continue;
continue;
}
submd.lsm_object_id = loi->loi_id;
+ submd.lsm_object_gr = loi->loi_gr;
submd.lsm_stripe_count = 0;
rc = obd_change_cbdata(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp,
&submd, it, data);
RETURN(rc);
}
+#define FIEMAP_BUFFER_SIZE 4096
+
+/* Non-zero fe_logical indicates that this is a continuation FIEMAP
+ * call. The local end offset and the device are sent in the first
+ * fm_extent. This function calculates the stripe number from the index.
+ * This function returns a stripe_no on which mapping is to be restarted.
+ *
+ * This function returns fm_end_offset which is the in-OST offset at which
+ * mapping should be restarted. If fm_end_offset=0 is returned then caller
+ * will re-calculate proper offset in next stripe.
+ * Note that the first extent is passed to lov_get_info via the value field */
+obd_size fiemap_calc_fm_end_offset(struct ll_user_fiemap *fiemap,
+ struct lov_stripe_md *lsm, obd_size fm_start,
+ obd_size fm_end, int *start_stripe)
+{
+ obd_size local_end = fiemap->fm_extents[0].fe_logical;
+ obd_off lun_start, lun_end;
+ obd_size fm_end_offset;
+ int stripe_no = -1, i;
+
+ if (fiemap->fm_extent_count == 0 ||
+ fiemap->fm_extents[0].fe_logical == 0)
+ return 0;
+
+ /* Find out stripe_no from ost_index saved in the fe_device */
+ for (i = 0; i < lsm->lsm_stripe_count; i++) {
+ if (lsm->lsm_oinfo[i]->loi_ost_idx ==
+ fiemap->fm_extents[0].fe_device) {
+ stripe_no = i;
+ break;
+ }
+ }
+
+ /* If we have finished mapping on previous device, shift logical
+ * offset to start of next device */
+ if ((lov_stripe_intersects(lsm, stripe_no, fm_start, fm_end,
+ &lun_start, &lun_end)) != 0 &&
+ local_end < lun_end) {
+ fm_end_offset = local_end;
+ *start_stripe = stripe_no;
+ } else {
+ /* This is a special value to indicate that caller should
+ * calculate offset in next stripe. */
+ fm_end_offset = 0;
+ *start_stripe = (stripe_no + 1) % lsm->lsm_stripe_count;
+ }
+
+ return fm_end_offset;
+}
+
+/* We calculate on which OST the mapping will end. If the length of mapping
+ * is greater than (stripe_size * stripe_count) then the last_stripe will
+ * will be one just before start_stripe. Else we check if the mapping
+ * intersects each OST and find last_stripe.
+ * This function returns the last_stripe and also sets the stripe_count
+ * over which the mapping is spread */
+int fiemap_calc_last_stripe(struct lov_stripe_md *lsm, obd_size fm_start,
+ obd_size fm_end, int start_stripe,
+ int *stripe_count)
+{
+ int last_stripe;
+ obd_off obd_start, obd_end;
+ int i, j;
+
+ if (fm_end - fm_start > lsm->lsm_stripe_size * lsm->lsm_stripe_count) {
+ last_stripe = (start_stripe < 1 ? lsm->lsm_stripe_count - 1 :
+ start_stripe - 1);
+ *stripe_count = lsm->lsm_stripe_count;
+ } else {
+ for (j = 0, i = start_stripe; j < lsm->lsm_stripe_count;
+ i = (i + 1) % lsm->lsm_stripe_count, j++) {
+ if ((lov_stripe_intersects(lsm, i, fm_start, fm_end,
+ &obd_start, &obd_end)) == 0)
+ break;
+ }
+ *stripe_count = j;
+ last_stripe = (start_stripe + j - 1) %lsm->lsm_stripe_count;
+ }
+
+ return last_stripe;
+}
+
+/* Set fe_device and copy extents from local buffer into main return buffer */
+void fiemap_prepare_and_copy_exts(struct ll_user_fiemap *fiemap,
+ struct ll_fiemap_extent *lcl_fm_ext,
+ int ost_index, unsigned int ext_count,
+ int current_extent)
+{
+ char *to;
+ int ext;
+
+ for (ext = 0; ext < ext_count; ext++) {
+ lcl_fm_ext[ext].fe_device = ost_index;
+ lcl_fm_ext[ext].fe_flags |= FIEMAP_EXTENT_NET;
+ }
+
+ /* Copy fm_extent's from fm_local to return buffer */
+ to = (char *)fiemap + fiemap_count_to_size(current_extent);
+ memcpy(to, lcl_fm_ext, ext_count * sizeof(struct ll_fiemap_extent));
+}
+
+static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key,
+ __u32 *vallen, void *val, struct lov_stripe_md *lsm)
+{
+ struct ll_fiemap_info_key *fm_key = key;
+ struct ll_user_fiemap *fiemap = val;
+ struct ll_user_fiemap *fm_local = NULL;
+ struct ll_fiemap_extent *lcl_fm_ext;
+ int count_local;
+ unsigned int get_num_extents = 0;
+ int ost_index = 0, actual_start_stripe, start_stripe;
+ obd_size fm_start, fm_end, fm_length, fm_end_offset = 0;
+ obd_size curr_loc;
+ int current_extent = 0, rc = 0, i;
+ int ost_eof = 0; /* EOF for object */
+ int ost_done = 0; /* done with required mapping for this OST? */
+ int last_stripe;
+ int cur_stripe = 0, cur_stripe_wrap = 0, stripe_count;
+ unsigned int buffer_size = FIEMAP_BUFFER_SIZE;
+
+ if (lsm == NULL)
+ GOTO(out, rc = 0);
+
+ if (fiemap_count_to_size(fm_key->fiemap.fm_extent_count) < buffer_size)
+ buffer_size = fiemap_count_to_size(fm_key->fiemap.fm_extent_count);
+
+ OBD_ALLOC(fm_local, buffer_size);
+ if (fm_local == NULL)
+ GOTO(out, rc = -ENOMEM);
+ lcl_fm_ext = &fm_local->fm_extents[0];
+
+ count_local = fiemap_size_to_count(buffer_size);
+
+ memcpy(fiemap, &fm_key->fiemap, sizeof(*fiemap));
+ fm_start = fiemap->fm_start;
+ fm_length = fiemap->fm_length;
+ /* Calculate start stripe, last stripe and length of mapping */
+ actual_start_stripe = start_stripe = lov_stripe_number(lsm, fm_start);
+ fm_end = (fm_length == ~0ULL ? fm_key->oa.o_size :
+ fm_start + fm_length - 1);
+ /* If fm_length != ~0ULL but fm_start+fm_length-1 exceeds file size */
+ if (fm_end > fm_key->oa.o_size)
+ fm_end = fm_key->oa.o_size;
+
+ last_stripe = fiemap_calc_last_stripe(lsm, fm_start, fm_end,
+ actual_start_stripe, &stripe_count);
+
+ fm_end_offset = fiemap_calc_fm_end_offset(fiemap, lsm, fm_start, fm_end,
+ &start_stripe);
+
+ if (fiemap->fm_extent_count == 0) {
+ get_num_extents = 1;
+ count_local = 0;
+ }
+
+ /* Check each stripe */
+ for (cur_stripe = start_stripe, i = 0; i < stripe_count;
+ i++, cur_stripe = (cur_stripe + 1) % lsm->lsm_stripe_count) {
+ obd_size req_fm_len; /* Stores length of required mapping */
+ obd_size len_mapped_single_call;
+ obd_off lun_start, lun_end, obd_object_end;
+ unsigned int ext_count;
+
+ cur_stripe_wrap = cur_stripe;
+
+ /* Find out range of mapping on this stripe */
+ if ((lov_stripe_intersects(lsm, cur_stripe, fm_start, fm_end,
+ &lun_start, &obd_object_end)) == 0)
+ continue;
+
+ /* If this is a continuation FIEMAP call and we are on
+ * starting stripe then lun_start needs to be set to
+ * fm_end_offset */
+ if (fm_end_offset != 0 && cur_stripe == start_stripe)
+ lun_start = fm_end_offset;
+
+ if (fm_length != ~0ULL) {
+ /* Handle fm_start + fm_length overflow */
+ if (fm_start + fm_length < fm_start)
+ fm_length = ~0ULL - fm_start;
+ lun_end = lov_size_to_stripe(lsm, fm_start + fm_length,
+ cur_stripe);
+ } else {
+ lun_end = ~0ULL;
+ }
+
+ if (lun_start == lun_end)
+ continue;
+
+ req_fm_len = obd_object_end - lun_start;
+ fm_local->fm_length = 0;
+ len_mapped_single_call = 0;
+
+ /* If the output buffer is very large and the objects have many
+ * extents we may need to loop on a single OST repeatedly */
+ ost_eof = 0;
+ ost_done = 0;
+ do {
+ if (get_num_extents == 0) {
+ /* Don't get too many extents. */
+ if (current_extent + count_local >
+ fiemap->fm_extent_count)
+ count_local = fiemap->fm_extent_count -
+ current_extent;
+ }
+
+ lun_start += len_mapped_single_call;
+ fm_local->fm_length = req_fm_len - len_mapped_single_call;
+ req_fm_len = fm_local->fm_length;
+ fm_local->fm_extent_count = count_local;
+ fm_local->fm_mapped_extents = 0;
+ fm_local->fm_flags = fiemap->fm_flags;
+
+ fm_key->oa.o_id = lsm->lsm_oinfo[cur_stripe]->loi_id;
+ ost_index = lsm->lsm_oinfo[cur_stripe]->loi_ost_idx;
+
+ if (ost_index < 0 || ost_index >=lov->desc.ld_tgt_count)
+ GOTO(out, rc = -EINVAL);
+
+ /* If OST is inactive, return extent with UNKNOWN flag */
+ if (lov && !lov->lov_tgts[ost_index]->ltd_active) {
+ fm_local->fm_flags |= FIEMAP_EXTENT_LAST;
+ fm_local->fm_mapped_extents = 1;
+
+ lcl_fm_ext[0].fe_logical = lun_start;
+ lcl_fm_ext[0].fe_length = obd_object_end -
+ lun_start;
+ lcl_fm_ext[0].fe_flags |= FIEMAP_EXTENT_UNKNOWN;
+
+ goto inactive_tgt;
+ }
+
+ fm_local->fm_start = lun_start;
+ fm_local->fm_flags &= ~FIEMAP_FLAG_DEVICE_ORDER;
+ memcpy(&fm_key->fiemap, fm_local, sizeof(*fm_local));
+ *vallen=fiemap_count_to_size(fm_local->fm_extent_count);
+ rc = obd_get_info(lov->lov_tgts[ost_index]->ltd_exp,
+ keylen, key, vallen, fm_local, lsm);
+ if (rc != 0)
+ GOTO(out, rc);
+
+inactive_tgt:
+ ext_count = fm_local->fm_mapped_extents;
+ if (ext_count == 0) {
+ ost_done = 1;
+ /* If last stripe has hole at the end,
+ * then we need to return */
+ if (cur_stripe_wrap == last_stripe) {
+ fiemap->fm_mapped_extents = 0;
+ goto finish;
+ }
+ break;
+ }
+
+ /* If we just need num of extents then go to next device */
+ if (get_num_extents) {
+ current_extent += ext_count;
+ break;
+ }
+
+ len_mapped_single_call = lcl_fm_ext[ext_count-1].fe_logical -
+ lun_start + lcl_fm_ext[ext_count - 1].fe_length;
+
+ /* Have we finished mapping on this device? */
+ if (req_fm_len <= len_mapped_single_call)
+ ost_done = 1;
+
+ /* Clear the EXTENT_LAST flag which can be present on
+ * last extent */
+ if (lcl_fm_ext[ext_count-1].fe_flags & FIEMAP_EXTENT_LAST)
+ lcl_fm_ext[ext_count - 1].fe_flags &=
+ ~FIEMAP_EXTENT_LAST;
+
+ curr_loc = lov_stripe_size(lsm,
+ lcl_fm_ext[ext_count - 1].fe_logical+
+ lcl_fm_ext[ext_count - 1].fe_length,
+ cur_stripe);
+ if (curr_loc >= fm_key->oa.o_size)
+ ost_eof = 1;
+
+ fiemap_prepare_and_copy_exts(fiemap, lcl_fm_ext,
+ ost_index, ext_count,
+ current_extent);
+
+ current_extent += ext_count;
+
+ /* Ran out of available extents? */
+ if (current_extent >= fiemap->fm_extent_count)
+ goto finish;
+ } while (ost_done == 0 && ost_eof == 0);
+
+ if (cur_stripe_wrap == last_stripe)
+ goto finish;
+ }
+
+finish:
+ /* Indicate that we are returning device offsets unless file just has
+ * single stripe */
+ if (lsm->lsm_stripe_count > 1)
+ fiemap->fm_flags |= FIEMAP_FLAG_DEVICE_ORDER;
+
+ if (get_num_extents)
+ goto skip_last_device_calc;
+
+ /* Check if we have reached the last stripe and whether mapping for that
+ * stripe is done. */
+ if (cur_stripe_wrap == last_stripe) {
+ if (ost_done || ost_eof)
+ fiemap->fm_extents[current_extent - 1].fe_flags |=
+ FIEMAP_EXTENT_LAST;
+ }
+
+skip_last_device_calc:
+ fiemap->fm_mapped_extents = current_extent;
+
+out:
+ OBD_FREE(fm_local, buffer_size);
+ return rc;
+}
+
static int lov_get_info(struct obd_export *exp, __u32 keylen,
- void *key, __u32 *vallen, void *val)
+ void *key, __u32 *vallen, void *val,
+ struct lov_stripe_md *lsm)
{
struct obd_device *obddev = class_exp2obd(exp);
struct lov_obd *lov = &obddev->u.lov;
struct {
char name[16];
struct ldlm_lock *lock;
- struct lov_stripe_md *lsm;
} *data = key;
struct ldlm_res_id *res_id = &data->lock->l_resource->lr_name;
struct lov_oinfo *loi;
/* XXX - it's assumed all the locks for deleted OSTs have
* been cancelled. Also, the export for deleted OSTs will
* be NULL and won't match the lock's export. */
- for (i = 0; i < data->lsm->lsm_stripe_count; i++) {
- loi = data->lsm->lsm_oinfo[i];
+ for (i = 0; i < lsm->lsm_stripe_count; i++) {
+ loi = lsm->lsm_oinfo[i];
if (!lov->lov_tgts[loi->loi_ost_idx])
continue;
if (lov->lov_tgts[loi->loi_ost_idx]->ltd_exp ==
data->lock->l_conn_export &&
- loi->loi_id == res_id->name[0] &&
- loi->loi_gr == res_id->name[1]) {
+ osc_res_name_eq(loi->loi_id, loi->loi_gr, res_id)) {
*stripe = i;
GOTO(out, rc = 0);
}
}
LDLM_ERROR(data->lock, "lock on inode without such object");
- dump_lsm(D_ERROR, data->lsm);
+ dump_lsm(D_ERROR, lsm);
GOTO(out, rc = -ENXIO);
} else if (KEY_IS(KEY_LAST_ID)) {
struct obd_id_info *info = val;
if (!tgt || !tgt->ltd_active)
GOTO(out, rc = -ESRCH);
- rc = obd_get_info(tgt->ltd_exp, keylen, key, &size, info->data);
+ rc = obd_get_info(tgt->ltd_exp, keylen, key, &size, info->data, NULL);
GOTO(out, rc = 0);
} else if (KEY_IS(KEY_LOVDESC)) {
struct lov_desc *desc_ret = val;
*desc_ret = lov->desc;
GOTO(out, rc = 0);
- } else if (KEY_IS(KEY_LOV_IDX)) {
- struct lov_tgt_desc *tgt;
-
- for(i = 0; i < lov->desc.ld_tgt_count; i++) {
- tgt = lov->lov_tgts[i];
- if (tgt && obd_uuid_equals(val, &tgt->ltd_uuid))
- GOTO(out, rc = i);
- }
+ } else if (KEY_IS(KEY_FIEMAP)) {
+ rc = lov_fiemap(lov, keylen, key, vallen, val, lsm);
+ GOTO(out, rc);
}
rc = -EINVAL;
.o_trigger_group_io = lov_trigger_group_io,
.o_teardown_async_page = lov_teardown_async_page,
.o_merge_lvb = lov_merge_lvb,
+ .o_update_lvb = lov_update_lvb,
.o_adjust_kms = lov_adjust_kms,
.o_punch = lov_punch,
.o_sync = lov_sync,
.o_unregister_page_removal_cb = lov_unregister_page_removal_cb,
.o_register_lock_cancel_cb = lov_register_lock_cancel_cb,
.o_unregister_lock_cancel_cb = lov_unregister_lock_cancel_cb,
+ .o_pool_new = lov_pool_new,
+ .o_pool_rem = lov_pool_remove,
+ .o_pool_add = lov_pool_add,
+ .o_pool_del = lov_pool_del,
};
static quota_interface_t *quota_interface;
LASSERT(rc == 0);
}
-MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
+MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
MODULE_DESCRIPTION("Lustre Logical Object Volume OBD driver");
MODULE_LICENSE("GPL");