-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
+/*
+ * GPL HEADER START
*
- * lustre/mds/mds_log.c
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * Copyright (c) 2001-2003 Cluster File Systems, Inc.
- * Author: Peter Braam <braam@clusterfs.com>
- * Author: Andreas Dilger <adilger@clusterfs.com>
- * Author: Phil Schwan <phil@clusterfs.com>
+ * 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 file is part of the Lustre file system, http://www.lustre.org
- * Lustre is a trademark of Cluster File Systems, Inc.
+ * 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 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.
+ * 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
*
- * 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.
+ * 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.
*
- * 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.
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2012, Whamcloud, Inc.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lustre/mds/mds_log.c
+ *
+ * Author: Peter Braam <braam@clusterfs.com>
+ * Author: Andreas Dilger <adilger@clusterfs.com>
+ * Author: Phil Schwan <phil@clusterfs.com>
*/
#define DEBUG_SUBSYSTEM S_MDS
-#ifdef HAVE_KERNEL_CONFIG_H
+#ifndef AUTOCONF_INCLUDED
#include <linux/config.h>
#endif
#include <linux/module.h>
#include <obd_class.h>
#include <lustre_fsfilt.h>
#include <lustre_mds.h>
-#include <lustre_commit_confd.h>
#include <lustre_log.h>
-
#include "mds_internal.h"
-static int mds_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 mds_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 obd_device *lov_obd = obd->u.mds.mds_osc_obd;
+ struct obd_device *lov_obd = obd->u.mds.mds_lov_obd;
struct llog_ctxt *lctxt;
int rc;
ENTRY;
lctxt = llog_get_context(lov_obd, ctxt->loc_idx);
rc = llog_add(lctxt, rec, lsm, logcookies, numcookies);
+ llog_ctxt_put(lctxt);
+
RETURN(rc);
}
-static int mds_llog_origin_connect(struct llog_ctxt *ctxt, int count,
+static int mds_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 obd_device *lov_obd = obd->u.mds.mds_osc_obd;
+ struct obd_device *lov_obd = obd->u.mds.mds_lov_obd;
struct llog_ctxt *lctxt;
int rc;
ENTRY;
lctxt = llog_get_context(lov_obd, ctxt->loc_idx);
- rc = llog_connect(lctxt, count, logid, gen, uuid);
+ rc = llog_connect(lctxt, logid, gen, uuid);
+ llog_ctxt_put(lctxt);
RETURN(rc);
}
+static struct llog_operations mds_ost_orig_logops = {
+ lop_add: mds_llog_origin_add,
+ lop_connect: mds_llog_origin_connect,
+};
+
static int mds_llog_repl_cancel(struct llog_ctxt *ctxt, struct lov_stripe_md *lsm,
int count, struct llog_cookie *cookies, int flags)
{
struct obd_device *obd = ctxt->loc_obd;
- struct obd_device *lov_obd = obd->u.mds.mds_osc_obd;
+ struct obd_device *lov_obd = obd->u.mds.mds_lov_obd;
struct llog_ctxt *lctxt;
int rc;
ENTRY;
lctxt = llog_get_context(lov_obd, ctxt->loc_idx);
rc = llog_cancel(lctxt, lsm, count, cookies, flags);
+ llog_ctxt_put(lctxt);
RETURN(rc);
}
-int mds_log_op_unlink(struct obd_device *obd,
- struct lov_mds_md *lmm, int lmm_size,
- struct llog_cookie *logcookies, int cookies_size)
-{
- struct mds_obd *mds = &obd->u.mds;
- struct lov_stripe_md *lsm = NULL;
- struct llog_unlink_rec *lur;
- struct llog_ctxt *ctxt;
- int rc;
- ENTRY;
+static struct llog_operations mds_size_repl_logops = {
+ lop_cancel: mds_llog_repl_cancel,
+};
- if (IS_ERR(mds->mds_osc_obd))
- RETURN(PTR_ERR(mds->mds_osc_obd));
+static struct llog_operations changelog_orig_logops;
- rc = obd_unpackmd(mds->mds_osc_exp, &lsm, lmm, lmm_size);
- if (rc < 0)
- RETURN(rc);
- rc = obd_checkmd(mds->mds_osc_exp, obd->obd_self_export, lsm);
- if (rc)
- GOTO(out, rc);
- /* first prepare unlink log record */
- OBD_ALLOC(lur, sizeof(*lur));
- if (!lur)
- GOTO(out, rc = -ENOMEM);
- lur->lur_hdr.lrh_len = lur->lur_tail.lrt_len = sizeof(*lur);
- lur->lur_hdr.lrh_type = MDS_UNLINK_REC;
-
- ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
- rc = llog_add(ctxt, &lur->lur_hdr, lsm, logcookies,
- cookies_size / sizeof(struct llog_cookie));
+static int llog_changelog_cancel_cb(struct llog_handle *llh,
+ struct llog_rec_hdr *hdr, void *data)
+{
+ struct llog_changelog_rec *rec = (struct llog_changelog_rec *)hdr;
+ struct llog_cookie cookie;
+ long long endrec = *(long long *)data;
+ int rc, err;
+ struct obd_device *obd;
+ void *trans_h;
+ struct inode *inode;
+ ENTRY;
- OBD_FREE(lur, sizeof(*lur));
-out:
- obd_free_memmd(mds->mds_osc_exp, &lsm);
- RETURN(rc);
+ /* This is always a (sub)log, not the catalog */
+ LASSERT(llh->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN);
+
+ if (rec->cr.cr_index > endrec)
+ /* records are in order, so we're done */
+ RETURN(LLOG_PROC_BREAK);
+
+ cookie.lgc_lgl = llh->lgh_id;
+ cookie.lgc_index = hdr->lrh_index;
+ obd = llh->lgh_ctxt->loc_exp->exp_obd;
+ inode = llh->lgh_file->f_dentry->d_inode;
+
+ /* XXX This is a workaround for the deadlock of changelog adding vs.
+ * changelog cancelling. Changelog adding always start transaction
+ * before acquiring the catlog lock (lgh_lock), whereas, changelog
+ * cancelling do start transaction after holding catlog lock.
+ *
+ * We start the transaction earlier here to keep the locking ordering:
+ * 'start transaction -> catlog lock'. LU-81. */
+ trans_h = fsfilt_start_log(obd, inode, FSFILT_OP_CANCEL_UNLINK,
+ NULL, 1);
+ if (IS_ERR(trans_h)) {
+ CERROR("fsfilt_start_log failed: %ld\n", PTR_ERR(trans_h));
+ RETURN(PTR_ERR(trans_h));
+ }
+
+ /* cancel them one at a time. I suppose we could store up the cookies
+ and cancel them all at once; probably more efficient, but this is
+ done as a user call, so who cares... */
+ rc = llog_cat_cancel_records(llh->u.phd.phd_cat_handle, 1, &cookie);
+
+ err = fsfilt_commit(obd, inode, trans_h, 0);
+ if (err) {
+ CERROR("fsfilt_commit failed: %d\n", err);
+ rc = (rc >= 0) ? err : rc;
+ }
+
+ RETURN(rc < 0 ? rc : 0);
}
-int mds_log_op_setattr(struct obd_device *obd, struct inode *inode,
- struct lov_mds_md *lmm, int lmm_size,
- struct llog_cookie *logcookies, int cookies_size)
+static int llog_changelog_cancel(struct llog_ctxt *ctxt,
+ struct lov_stripe_md *lsm, int count,
+ struct llog_cookie *cookies, int flags)
{
- struct mds_obd *mds = &obd->u.mds;
- struct lov_stripe_md *lsm = NULL;
- struct llog_setattr_rec *lsr;
- struct llog_ctxt *ctxt;
+ struct llog_handle *cathandle = ctxt->loc_handle;
int rc;
ENTRY;
- if (IS_ERR(mds->mds_osc_obd))
- RETURN(PTR_ERR(mds->mds_osc_obd));
+ /* This should only be called with the catalog handle */
+ LASSERT(cathandle->lgh_hdr->llh_flags & LLOG_F_IS_CAT);
- rc = obd_unpackmd(mds->mds_osc_exp, &lsm, lmm, lmm_size);
- if (rc < 0)
- RETURN(rc);
+ rc = llog_cat_process(cathandle, llog_changelog_cancel_cb,
+ (void *)cookies, 0, 0);
+ if (rc >= 0)
+ /* 0 or 1 means we're done */
+ rc = 0;
+ else
+ CERROR("cancel idx %u of catalog "LPX64" rc=%d\n",
+ cathandle->lgh_last_idx, cathandle->lgh_id.lgl_oid, rc);
- rc = obd_checkmd(mds->mds_osc_exp, obd->obd_self_export, lsm);
- if (rc)
- GOTO(out, rc);
+ RETURN(rc);
+}
- OBD_ALLOC(lsr, sizeof(*lsr));
- if (!lsr)
- GOTO(out, rc = -ENOMEM);
+int mds_changelog_llog_init(struct obd_device *obd, struct obd_device *tgt)
+{
+ int rc;
- /* prepare setattr log record */
- lsr->lsr_hdr.lrh_len = lsr->lsr_tail.lrt_len = sizeof(*lsr);
- lsr->lsr_hdr.lrh_type = MDS_SETATTR_REC;
- lsr->lsr_uid = inode->i_uid;
- lsr->lsr_gid = inode->i_gid;
+ /* see osc_llog_init */
+ changelog_orig_logops = llog_lvfs_ops;
+ changelog_orig_logops.lop_setup = llog_obd_origin_setup;
+ changelog_orig_logops.lop_cleanup = llog_obd_origin_cleanup;
+ changelog_orig_logops.lop_add = llog_obd_origin_add;
+ changelog_orig_logops.lop_cancel = llog_changelog_cancel;
+
+ rc = llog_setup_named(obd, &obd->obd_olg, LLOG_CHANGELOG_ORIG_CTXT,
+ tgt, 1, NULL, CHANGELOG_CATALOG,
+ &changelog_orig_logops);
+ if (rc) {
+ CERROR("changelog llog setup failed %d\n", rc);
+ RETURN(rc);
+ }
- /* write setattr log */
- ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
- rc = llog_add(ctxt, &lsr->lsr_hdr, lsm, logcookies,
- cookies_size / sizeof(struct llog_cookie));
+ rc = llog_setup_named(obd, &obd->obd_olg, LLOG_CHANGELOG_USER_ORIG_CTXT,
+ tgt, 1, NULL, CHANGELOG_USERS,
+ &changelog_orig_logops);
+ if (rc) {
+ CERROR("changelog users llog setup failed %d\n", rc);
+ RETURN(rc);
+ }
- OBD_FREE(lsr, sizeof(*lsr));
- out:
- obd_free_memmd(mds->mds_osc_exp, &lsm);
RETURN(rc);
}
+EXPORT_SYMBOL(mds_changelog_llog_init);
-static struct llog_operations mds_ost_orig_logops = {
- lop_add: mds_llog_origin_add,
- lop_connect: mds_llog_origin_connect,
-};
-
-static struct llog_operations mds_size_repl_logops = {
- lop_cancel: mds_llog_repl_cancel,
-};
-
-int mds_llog_init(struct obd_device *obd, struct obd_device *tgt,
- int count, struct llog_catid *logid, struct obd_uuid *uuid)
+int mds_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
+ struct obd_device *disk_obd, int *index)
{
- struct obd_device *lov_obd = obd->u.mds.mds_osc_obd;
+ struct obd_device *lov_obd = obd->u.mds.mds_lov_obd;
+ struct llog_ctxt *ctxt;
int rc;
ENTRY;
- rc = llog_setup(obd, LLOG_MDS_OST_ORIG_CTXT, tgt, 0, NULL,
- &mds_ost_orig_logops);
+ LASSERT(olg == &obd->obd_olg);
+ rc = llog_setup(obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT, disk_obd,
+ 0, NULL, &mds_ost_orig_logops);
if (rc)
RETURN(rc);
- rc = llog_setup(obd, LLOG_SIZE_REPL_CTXT, tgt, 0, NULL,
- &mds_size_repl_logops);
+ rc = llog_setup(obd, &obd->obd_olg, LLOG_SIZE_REPL_CTXT, disk_obd,
+ 0, NULL, &mds_size_repl_logops);
if (rc)
- RETURN(rc);
+ GOTO(err_llog, rc);
- rc = obd_llog_init(lov_obd, tgt, count, logid, uuid);
- if (rc)
+ rc = obd_llog_init(lov_obd, &lov_obd->obd_olg, disk_obd, index);
+ if (rc) {
CERROR("lov_llog_init err %d\n", rc);
+ GOTO(err_cleanup, rc);
+ }
RETURN(rc);
+err_cleanup:
+ ctxt = llog_get_context(obd, LLOG_SIZE_REPL_CTXT);
+ if (ctxt)
+ llog_cleanup(ctxt);
+err_llog:
+ ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
+ if (ctxt)
+ llog_cleanup(ctxt);
+ return rc;
}
int mds_llog_finish(struct obd_device *obd, int count)
if (!rc)
rc = rc2;
+ ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
+ if (ctxt)
+ rc2 = llog_cleanup(ctxt);
+ if (!rc)
+ rc = rc2;
+
+ ctxt = llog_get_context(obd, LLOG_CHANGELOG_USER_ORIG_CTXT);
+ if (ctxt)
+ rc2 = llog_cleanup(ctxt);
+ if (!rc)
+ rc = rc2;
+
+ RETURN(rc);
+}
+
+static int mds_llog_add_unlink(struct obd_device *obd,
+ struct lov_stripe_md *lsm, obd_count count,
+ struct llog_cookie *logcookie, int cookies)
+{
+ struct llog_unlink_rec *lur;
+ struct llog_ctxt *ctxt;
+ int rc;
+
+ if (cookies < lsm->lsm_stripe_count)
+ RETURN(rc = -EFBIG);
+
+ /* first prepare unlink log record */
+ OBD_ALLOC_PTR(lur);
+ if (!lur)
+ RETURN(rc = -ENOMEM);
+ lur->lur_hdr.lrh_len = lur->lur_tail.lrt_len = sizeof(*lur);
+ lur->lur_hdr.lrh_type = MDS_UNLINK_REC;
+ lur->lur_count = count;
+
+ ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
+ rc = llog_add(ctxt, &lur->lur_hdr, lsm, logcookie, cookies);
+ llog_ctxt_put(ctxt);
+
+ OBD_FREE_PTR(lur);
RETURN(rc);
}
+
+int mds_log_op_unlink(struct obd_device *obd,
+ struct lov_mds_md *lmm, int lmm_size,
+ struct llog_cookie *logcookies, int cookies_size)
+{
+ struct mds_obd *mds = &obd->u.mds;
+ struct lov_stripe_md *lsm = NULL;
+ int rc;
+ ENTRY;
+
+ if (IS_ERR(mds->mds_lov_obd))
+ RETURN(PTR_ERR(mds->mds_lov_obd));
+
+ rc = obd_unpackmd(mds->mds_lov_exp, &lsm, lmm, lmm_size);
+ if (rc < 0)
+ RETURN(rc);
+ rc = mds_llog_add_unlink(obd, lsm, 0, logcookies,
+ cookies_size / sizeof(struct llog_cookie));
+ obd_free_memmd(mds->mds_lov_exp, &lsm);
+ RETURN(rc);
+}
+EXPORT_SYMBOL(mds_log_op_unlink);
+
+int mds_log_op_orphan(struct obd_device *obd, struct lov_stripe_md *lsm,
+ obd_count count)
+{
+ struct mds_obd *mds = &obd->u.mds;
+ struct llog_cookie logcookie;
+ int rc;
+ ENTRY;
+
+ if (IS_ERR(mds->mds_lov_obd))
+ RETURN(PTR_ERR(mds->mds_lov_obd));
+
+ rc = mds_llog_add_unlink(obd, lsm, count - 1, &logcookie, 1);
+ RETURN(rc);
+}
+