- /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
+/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * Copyright (C) 2002, 2003 Cluster File Systems, Inc.
- * Author: Phil Schwan <phil@clusterfs.com>
- * Peter Braam <braam@clusterfs.com>
- * Mike Shaver <shaver@clusterfs.com>
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * This file is part of Lustre, http://www.lustre.org.
+ * 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.
*
- * Lustre is free 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.
+ * 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).
*
- * 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
- * GNU General Public License for more details.
+ * 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
*
- * You should have received a copy of the GNU General Public License
- * along with Lustre; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ * 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.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. 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.
+ *
+ * lustre/lov/lov_log.c
+ *
+ * Author: Phil Schwan <phil@clusterfs.com>
+ * Author: Peter Braam <braam@clusterfs.com>
+ * Author: Mike Shaver <shaver@clusterfs.com>
*/
#ifndef EXPORT_SYMTAB
#endif
#define DEBUG_SUBSYSTEM S_LOV
#ifdef __KERNEL__
-#include <linux/slab.h>
-#include <linux/module.h>
-#include <linux/init.h>
-#include <linux/slab.h>
-#include <linux/pagemap.h>
-#include <asm/div64.h>
-#include <linux/seq_file.h>
+#include <libcfs/libcfs.h>
#else
#include <liblustre.h>
#endif
-#include <linux/obd_support.h>
-#include <linux/lustre_lib.h>
-#include <linux/lustre_net.h>
-#include <linux/lustre_idl.h>
-#include <linux/lustre_dlm.h>
-#include <linux/lustre_mds.h>
-#include <linux/obd_class.h>
-#include <linux/obd_lov.h>
-#include <linux/obd_ost.h>
-#include <linux/lprocfs_status.h>
+#include <obd_support.h>
+#include <lustre_lib.h>
+#include <lustre_net.h>
+#include <lustre/lustre_idl.h>
+#include <lustre_dlm.h>
+#include <lustre_mds.h>
+#include <obd_class.h>
+#include <obd_lov.h>
+#include <obd_ost.h>
+#include <lprocfs_status.h>
#include "lov_internal.h"
* we need to keep cookies in stripe order, even if some are NULL, so that
* the right cookies are passed back to the right OSTs at the client side.
* Unset cookies should be all-zero (which will never occur naturally). */
-static int lov_llog_origin_add(struct llog_ctxt *ctxt,
- struct llog_rec_hdr *rec, struct lov_stripe_md *lsm,
- struct llog_cookie *logcookies, int numcookies)
+static int lov_llog_origin_add(struct llog_ctxt *ctxt, struct llog_rec_hdr *rec,
+ struct lov_stripe_md *lsm,
+ struct llog_cookie *logcookies, int numcookies)
{
struct obd_device *obd = ctxt->loc_obd;
struct lov_obd *lov = &obd->u.lov;
- struct lov_oinfo *loi;
- struct llog_unlink_rec *lur;
- int i, rc = 0;
+ int i, rc = 0, cookies = 0;
ENTRY;
- OBD_ALLOC(lur, sizeof(*lur));
- if (!lur)
- RETURN(-ENOMEM);
- lur->lur_hdr.lrh_len = lur->lur_tail.lrt_len = sizeof(*lur);
- lur->lur_hdr.lrh_type = MDS_UNLINK_REC;
-
- LASSERT(logcookies && numcookies >= lsm->lsm_stripe_count);
+ LASSERTF(logcookies && numcookies >= lsm->lsm_stripe_count,
+ "logcookies %p, numcookies %d lsm->lsm_stripe_count %d \n",
+ logcookies, numcookies, lsm->lsm_stripe_count);
- for (i = 0,loi = lsm->lsm_oinfo; i < lsm->lsm_stripe_count; i++,loi++) {
- struct obd_device *child = lov->tgts[loi->loi_ost_idx].ltd_exp->exp_obd;
+ for (i = 0; i < lsm->lsm_stripe_count; i++) {
+ struct lov_oinfo *loi = lsm->lsm_oinfo[i];
+ struct obd_device *child =
+ lov->lov_tgts[loi->loi_ost_idx]->ltd_exp->exp_obd;
struct llog_ctxt *cctxt = llog_get_context(child, ctxt->loc_idx);
- lur->lur_oid = loi->loi_id;
- lur->lur_ogen = loi->loi_gr;
- rc += llog_add(cctxt, &lur->lur_hdr, NULL, logcookies + rc,
- numcookies - rc);
-
+ /* fill mds unlink/setattr log record */
+ switch (rec->lrh_type) {
+ case MDS_UNLINK_REC: {
+ struct llog_unlink_rec *lur = (struct llog_unlink_rec *)rec;
+ lur->lur_oid = loi->loi_id;
+ lur->lur_oseq = loi->loi_seq;
+ break;
+ }
+ case MDS_SETATTR_REC: {
+ struct llog_setattr_rec *lsr = (struct llog_setattr_rec *)rec;
+ lsr->lsr_oid = loi->loi_id;
+ lsr->lsr_oseq = loi->loi_seq;
+ break;
+ }
+ case MDS_SETATTR64_REC: {
+ struct llog_setattr64_rec *lsr = (struct llog_setattr64_rec *)rec;
+ lsr->lsr_oid = loi->loi_id;
+ lsr->lsr_oseq = loi->loi_seq;
+ break;
+ }
+ default:
+ break;
+ }
+ /* inject error in llog_add() below */
+ if (OBD_FAIL_CHECK(OBD_FAIL_MDS_FAIL_LOV_LOG_ADD)) {
+ llog_ctxt_put(cctxt);
+ cctxt = NULL;
+ }
+ rc = llog_add(cctxt, rec, NULL, logcookies + cookies,
+ numcookies - cookies);
+ llog_ctxt_put(cctxt);
+ if (rc < 0) {
+ CERROR("Can't add llog (rc = %d) for stripe %d\n",
+ rc, cookies);
+ memset(logcookies + cookies, 0,
+ sizeof(struct llog_cookie));
+ rc = 1; /* skip this cookie */
+ }
+ /* Note that rc is always 1 if llog_add was successful */
+ cookies += rc;
}
- OBD_FREE(lur, sizeof(*lur));
-
- RETURN(rc);
+ RETURN(cookies);
}
-static int lov_llog_origin_connect(struct llog_ctxt *ctxt, int count,
- struct llog_logid *logid,
- struct llog_gen *gen)
+static int lov_llog_origin_connect(struct llog_ctxt *ctxt,
+ struct llog_logid *logid,
+ struct llog_gen *gen,
+ struct obd_uuid *uuid)
{
struct obd_device *obd = ctxt->loc_obd;
struct lov_obd *lov = &obd->u.lov;
- int i, rc = 0;
+ int i, rc = 0, err = 0;
ENTRY;
- LASSERT(lov->desc.ld_tgt_count == count);
+ obd_getref(obd);
for (i = 0; i < lov->desc.ld_tgt_count; i++) {
- struct obd_device *child = lov->tgts[i].ltd_exp->exp_obd;
- struct llog_ctxt *cctxt = llog_get_context(child, ctxt->loc_idx);
- rc = llog_connect(cctxt, 1, logid, gen);
+ struct obd_device *child;
+ struct llog_ctxt *cctxt;
+
+ if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_active)
+ continue;
+ if (uuid && !obd_uuid_equals(uuid, &lov->lov_tgts[i]->ltd_uuid))
+ continue;
+ CDEBUG(D_CONFIG, "connect %d/%d\n", i, lov->desc.ld_tgt_count);
+ child = lov->lov_tgts[i]->ltd_exp->exp_obd;
+ cctxt = llog_get_context(child, ctxt->loc_idx);
+ rc = llog_connect(cctxt, logid, gen, uuid);
+ llog_ctxt_put(cctxt);
+
if (rc) {
- CERROR("error osc_llog_connect %d\n", i);
- break;
+ CERROR("error osc_llog_connect tgt %d (%d)\n", i, rc);
+ if (!err)
+ err = rc;
}
}
+ obd_putref(obd);
- RETURN(rc);
+ RETURN(err);
}
/* the replicators commit callback */
{
struct lov_obd *lov;
struct obd_device *obd = ctxt->loc_obd;
- struct lov_oinfo *loi;
int rc = 0, i;
ENTRY;
LASSERT(lsm != NULL);
LASSERT(count == lsm->lsm_stripe_count);
- loi = lsm->lsm_oinfo;
lov = &obd->u.lov;
- for (i = 0; i < count; i++, cookies++, loi++) {
- struct obd_device *child = lov->tgts[loi->loi_ost_idx].ltd_exp->exp_obd;
- struct llog_ctxt *cctxt = llog_get_context(child, ctxt->loc_idx);
+ obd_getref(obd);
+ for (i = 0; i < count; i++, cookies++) {
+ struct lov_oinfo *loi = lsm->lsm_oinfo[i];
+ struct obd_device *child =
+ lov->lov_tgts[loi->loi_ost_idx]->ltd_exp->exp_obd;
+ struct llog_ctxt *cctxt =
+ llog_get_context(child, ctxt->loc_idx);
int err;
err = llog_cancel(cctxt, NULL, 1, cookies, flags);
- if (err && lov->tgts[loi->loi_ost_idx].active) {
+ llog_ctxt_put(cctxt);
+ if (err && lov->lov_tgts[loi->loi_ost_idx]->ltd_active) {
CERROR("error: objid "LPX64" subobj "LPX64
" on OST idx %d: rc = %d\n", lsm->lsm_object_id,
loi->loi_id, loi->loi_ost_idx, err);
rc = err;
}
}
+ obd_putref(obd);
RETURN(rc);
}
-static struct llog_operations lov_unlink_orig_logops = {
+static struct llog_operations lov_mds_ost_orig_logops = {
lop_add: lov_llog_origin_add,
lop_connect: lov_llog_origin_connect
};
lop_cancel: lov_llog_repl_cancel
};
-
-int lov_llog_init(struct obd_device *obd, struct obd_device *tgt,
- int count, struct llog_logid *logid)
+int lov_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
+ struct obd_device *disk_obd, int *index)
{
struct lov_obd *lov = &obd->u.lov;
+ struct obd_device *child;
int i, rc = 0;
ENTRY;
-
- rc = llog_setup(obd, LLOG_UNLINK_ORIG_CTXT, tgt, 0, NULL,
- &lov_unlink_orig_logops);
+
+ LASSERT(olg == &obd->obd_olg);
+ rc = llog_setup(obd, olg, LLOG_MDS_OST_ORIG_CTXT, disk_obd, 0, NULL,
+ &lov_mds_ost_orig_logops);
if (rc)
RETURN(rc);
- rc = llog_setup(obd, LLOG_SIZE_REPL_CTXT, tgt, 0, NULL,
+ rc = llog_setup(obd, olg, LLOG_SIZE_REPL_CTXT, disk_obd, 0, NULL,
&lov_size_repl_logops);
if (rc)
- RETURN(rc);
+ GOTO(err_cleanup, rc);
- LASSERT(lov->desc.ld_tgt_count == count);
+ obd_getref(obd);
+ /* count may not match lov->desc.ld_tgt_count during dynamic ost add */
for (i = 0; i < lov->desc.ld_tgt_count; i++) {
- struct obd_device *child = lov->tgts[i].ltd_exp->exp_obd;
- rc = obd_llog_init(child, tgt, 1, logid + i);
- if (rc) {
- CERROR("error osc_llog_init %d\n", i);
- break;
- }
+ if (!lov->lov_tgts[i])
+ continue;
+
+ if (index && i != *index)
+ continue;
+
+ child = lov->lov_tgts[i]->ltd_obd;
+ rc = obd_llog_init(child, &child->obd_olg, disk_obd, &i);
+ if (rc)
+ CERROR("error osc_llog_init idx %d osc '%s' tgt '%s' "
+ "(rc=%d)\n", i, child->obd_name,
+ disk_obd->obd_name, rc);
+ rc = 0;
}
- RETURN(rc);
+ obd_putref(obd);
+ GOTO(err_cleanup, rc);
+err_cleanup:
+ if (rc) {
+ struct llog_ctxt *ctxt =
+ llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
+ if (ctxt)
+ llog_cleanup(ctxt);
+ ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
+ if (ctxt)
+ llog_cleanup(ctxt);
+ }
+ return rc;
}
int lov_llog_finish(struct obd_device *obd, int count)
{
- struct lov_obd *lov = &obd->u.lov;
- int i, rc = 0;
+ struct llog_ctxt *ctxt;
+ int rc = 0, rc2 = 0;
ENTRY;
-
- rc = llog_cleanup(llog_get_context(obd, LLOG_UNLINK_ORIG_CTXT));
- if (rc)
- RETURN(rc);
- rc = llog_cleanup(llog_get_context(obd, LLOG_SIZE_REPL_CTXT));
- if (rc)
- RETURN(rc);
+ /* cleanup our llogs only if the ctxts have been setup
+ * (client lov doesn't setup, mds lov does). */
+ ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
+ if (ctxt)
+ rc = llog_cleanup(ctxt);
- LASSERT(lov->desc.ld_tgt_count == count);
- for (i = 0; i < lov->desc.ld_tgt_count; i++) {
- struct obd_device *child = lov->tgts[i].ltd_exp->exp_obd;
- rc = obd_llog_finish(child, 1);
- if (rc) {
- CERROR("error osc_llog_finish %d\n", i);
- break;
- }
- }
+ ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
+ if (ctxt)
+ rc2 = llog_cleanup(ctxt);
+ if (!rc)
+ rc = rc2;
+
+ /* lov->tgt llogs are cleaned during osc_cleanup. */
RETURN(rc);
}