/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * Copyright (C) 2001-2003 Cluster File Systems, Inc.
- * Author: Andreas Dilger <adilger@clusterfs.com>
+ * GPL HEADER START
*
- * This file is part of Lustre, http://www.lustre.org.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * 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 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 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.
+ * 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
- * along with Lustre; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ * 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
+ *
+ * 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 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.
+ *
+ * lustre/obdclass/llog_cat.c
*
* OST<->MDS recovery logging infrastructure.
*
* Invariants in implementation:
* - we do not share logs among different OST<->MDS connections, so that
* if an OST or MDS fails it need only look at log(s) relevant to itself
+ *
+ * Author: Andreas Dilger <adilger@clusterfs.com>
*/
#define DEBUG_SUBSYSTEM S_LOG
#define EXPORT_SYMTAB
#endif
-#ifdef __KERNEL__
-#include <linux/fs.h>
-#else
+#ifndef __KERNEL__
#include <liblustre.h>
#endif
-#include <linux/obd_class.h>
-#include <linux/lustre_log.h>
-#include <portals/list.h>
+#include <obd_class.h>
+#include <lustre_log.h>
+#include <libcfs/list.h>
/* Create a new log handle and add it to the open list.
* This log handle will be closed when all of the records in it are removed.
ENTRY;
llh = cathandle->lgh_hdr;
- bitmap_size = sizeof(llh->llh_bitmap) * 8;
+ bitmap_size = LLOG_BITMAP_SIZE(llh);
index = (cathandle->lgh_last_idx + 1) % bitmap_size;
if (llh->llh_cat_idx == index) {
CERROR("no free catalog slots for log...\n");
RETURN(ERR_PTR(-ENOSPC));
- } else {
- if (index == 0)
- index = 1;
- if (ext2_set_bit(index, llh->llh_bitmap)) {
- CERROR("argh, index %u already set in log bitmap?\n",
- index);
- LBUG(); /* should never happen */
- }
- cathandle->lgh_last_idx = index;
- llh->llh_count++;
- llh->llh_tail.lrt_index = index;
}
+ if (OBD_FAIL_CHECK(OBD_FAIL_MDS_LLOG_CREATE_FAILED))
+ RETURN(ERR_PTR(-ENOSPC));
+
rc = llog_create(cathandle->lgh_ctxt, &loghandle, NULL, NULL);
if (rc)
RETURN(ERR_PTR(rc));
if (rc)
GOTO(out_destroy, rc);
- CDEBUG(D_HA, "new recovery log "LPX64":%x for index %u of catalog "
+ if (index == 0)
+ index = 1;
+ if (ext2_set_bit(index, llh->llh_bitmap)) {
+ CERROR("argh, index %u already set in log bitmap?\n",
+ index);
+ LBUG(); /* should never happen */
+ }
+ cathandle->lgh_last_idx = index;
+ llh->llh_count++;
+ llh->llh_tail.lrt_index = index;
+
+ CDEBUG(D_RPCTRACE,"new recovery log "LPX64":%x for index %u of catalog "
LPX64"\n", loghandle->lgh_id.lgl_oid, loghandle->lgh_id.lgl_ogen,
index, cathandle->lgh_id.lgl_oid);
/* build the record for this log in the catalog */
LASSERT(list_empty(&loghandle->u.phd.phd_entry));
list_add_tail(&loghandle->u.phd.phd_entry, &cathandle->u.chd.chd_head);
- out_destroy:
+out_destroy:
if (rc < 0)
llog_destroy(loghandle);
RETURN(loghandle);
}
-EXPORT_SYMBOL(llog_cat_new_log);
/* Open an existent log handle and add it to the open list.
* This log handle will be closed when all of the records in it are removed.
if (!rc) {
loghandle->u.phd.phd_cat_handle = cathandle;
loghandle->u.phd.phd_cookie.lgc_lgl = cathandle->lgh_id;
- loghandle->u.phd.phd_cookie.lgc_index =
+ loghandle->u.phd.phd_cookie.lgc_index =
loghandle->lgh_hdr->llh_cat_idx;
}
}
EXPORT_SYMBOL(llog_cat_put);
-/* Return the currently active log handle. If the current log handle doesn't
+/**
+ * lockdep markers for nested struct llog_handle::lgh_lock locking.
+ */
+enum {
+ LLOGH_CAT,
+ LLOGH_LOG
+};
+
+/** Return the currently active log handle. If the current log handle doesn't
* have enough space left for the current record, start a new one.
*
* If reclen is 0, we only want to know what the currently active log is,
struct llog_handle *loghandle = NULL;
ENTRY;
- down_read(&cathandle->lgh_lock);
+ down_read_nested(&cathandle->lgh_lock, LLOGH_CAT);
loghandle = cathandle->u.chd.chd_current_log;
if (loghandle) {
struct llog_log_hdr *llh = loghandle->lgh_hdr;
- if (loghandle->lgh_last_idx < (sizeof(llh->llh_bitmap)*8) - 1) {
- down_write(&loghandle->lgh_lock);
+ down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+ if (loghandle->lgh_last_idx < LLOG_BITMAP_SIZE(llh) - 1) {
up_read(&cathandle->lgh_lock);
RETURN(loghandle);
+ } else {
+ up_write(&loghandle->lgh_lock);
}
}
if (!create) {
/* time to create new log */
/* first, we have to make sure the state hasn't changed */
- down_write(&cathandle->lgh_lock);
+ down_write_nested(&cathandle->lgh_lock, LLOGH_CAT);
loghandle = cathandle->u.chd.chd_current_log;
if (loghandle) {
struct llog_log_hdr *llh = loghandle->lgh_hdr;
- if (loghandle->lgh_last_idx < (sizeof(llh->llh_bitmap)*8) - 1) {
- down_write(&loghandle->lgh_lock);
+ down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
+ if (loghandle->lgh_last_idx < LLOG_BITMAP_SIZE(llh) - 1) {
up_write(&cathandle->lgh_lock);
RETURN(loghandle);
+ } else {
+ up_write(&loghandle->lgh_lock);
}
}
CDEBUG(D_INODE, "creating new log\n");
loghandle = llog_cat_new_log(cathandle);
if (!IS_ERR(loghandle))
- down_write(&loghandle->lgh_lock);
+ down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
up_write(&cathandle->lgh_lock);
RETURN(loghandle);
}
* Assumes caller has already pushed us into the kernel context.
*/
int llog_cat_cancel_records(struct llog_handle *cathandle, int count,
- struct llog_cookie *cookies)
+ struct llog_cookie *cookies)
{
int i, index, rc = 0;
ENTRY;
- down_write(&cathandle->lgh_lock);
+ down_write_nested(&cathandle->lgh_lock, LLOGH_CAT);
for (i = 0; i < count; i++, cookies++) {
struct llog_handle *loghandle;
struct llog_logid *lgl = &cookies->lgc_lgl;
break;
}
- down_write(&loghandle->lgh_lock);
+ down_write_nested(&loghandle->lgh_lock, LLOGH_LOG);
rc = llog_cancel_rec(loghandle, cookies->lgc_index);
up_write(&loghandle->lgh_lock);
llog_cat_set_first_idx(cathandle, index);
rc = llog_cancel_rec(cathandle, index);
if (rc == 0)
- CDEBUG(D_HA, "cancel plain log at index %u "
- "of catalog "LPX64"\n",
+ CDEBUG(D_RPCTRACE,"cancel plain log at index %u"
+ " of catalog "LPX64"\n",
index, cathandle->lgh_id.lgl_oid);
}
}
struct llog_handle *llh;
int rc;
+ ENTRY;
if (rec->lrh_type != LLOG_LOGID_MAGIC) {
CERROR("invalid record in catalog\n");
RETURN(-EINVAL);
}
- CWARN("processing log "LPX64":%x at index %u of catalog "LPX64"\n",
- lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
+ CDEBUG(D_HA, "processing log "LPX64":%x at index %u of catalog "
+ LPX64"\n", lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
rec->lrh_index, cat_llh->lgh_id.lgl_oid);
rc = llog_cat_id2handle(cat_llh, &llh, &lir->lid_id);
CWARN("catlog "LPX64" crosses index zero\n",
cat_llh->lgh_id.lgl_oid);
- cd.first_idx = llh->llh_cat_idx;
- cd.last_idx = 0;
+ cd.lpcd_first_idx = llh->llh_cat_idx;
+ cd.lpcd_last_idx = 0;
rc = llog_process(cat_llh, llog_cat_process_cb, &d, &cd);
if (rc != 0)
RETURN(rc);
- cd.first_idx = 0;
- cd.last_idx = cat_llh->lgh_last_idx;
+ cd.lpcd_first_idx = 0;
+ cd.lpcd_last_idx = cat_llh->lgh_last_idx;
rc = llog_process(cat_llh, llog_cat_process_cb, &d, &cd);
} else {
rc = llog_process(cat_llh, llog_cat_process_cb, &d, NULL);
}
EXPORT_SYMBOL(llog_cat_process);
+#ifdef __KERNEL__
+int llog_cat_process_thread(void *data)
+{
+ struct llog_process_cat_args *args = data;
+ struct llog_ctxt *ctxt = args->lpca_ctxt;
+ struct llog_handle *llh = NULL;
+ void *cb = args->lpca_cb;
+ struct llog_logid logid;
+ int rc;
+ ENTRY;
+
+ cfs_daemonize_ctxt("ll_log_process");
+
+ logid = *(struct llog_logid *)(args->lpca_arg);
+ rc = llog_create(ctxt, &llh, &logid, NULL);
+ if (rc) {
+ CERROR("llog_create() failed %d\n", rc);
+ GOTO(out, rc);
+ }
+ rc = llog_init_handle(llh, LLOG_F_IS_CAT, NULL);
+ if (rc) {
+ CERROR("llog_init_handle failed %d\n", rc);
+ GOTO(release_llh, rc);
+ }
+
+ if (cb) {
+ rc = llog_cat_process(llh, (llog_cb_t)cb, NULL);
+ if (rc != LLOG_PROC_BREAK && rc != 0)
+ CERROR("llog_cat_process() failed %d\n", rc);
+ } else {
+ CWARN("No callback function for recovery\n");
+ }
+
+ /*
+ * Make sure that all cached data is sent.
+ */
+ llog_sync(ctxt, NULL);
+ GOTO(release_llh, rc);
+release_llh:
+ rc = llog_cat_put(llh);
+ if (rc)
+ CERROR("llog_cat_put() failed %d\n", rc);
+out:
+ llog_ctxt_put(ctxt);
+ OBD_FREE_PTR(args);
+ return rc;
+}
+EXPORT_SYMBOL(llog_cat_process_thread);
+#endif
+
+static int llog_cat_reverse_process_cb(struct llog_handle *cat_llh,
+ struct llog_rec_hdr *rec, void *data)
+{
+ struct llog_process_data *d = data;
+ struct llog_logid_rec *lir = (struct llog_logid_rec *)rec;
+ struct llog_handle *llh;
+ int rc;
+
+ if (le32_to_cpu(rec->lrh_type) != LLOG_LOGID_MAGIC) {
+ CERROR("invalid record in catalog\n");
+ RETURN(-EINVAL);
+ }
+ CDEBUG(D_HA, "processing log "LPX64":%x at index %u of catalog "
+ LPX64"\n", lir->lid_id.lgl_oid, lir->lid_id.lgl_ogen,
+ le32_to_cpu(rec->lrh_index), cat_llh->lgh_id.lgl_oid);
+
+ rc = llog_cat_id2handle(cat_llh, &llh, &lir->lid_id);
+ if (rc) {
+ CERROR("Cannot find handle for log "LPX64"\n",
+ lir->lid_id.lgl_oid);
+ RETURN(rc);
+ }
+
+ rc = llog_reverse_process(llh, d->lpd_cb, d->lpd_data, NULL);
+ RETURN(rc);
+}
+
+int llog_cat_reverse_process(struct llog_handle *cat_llh,
+ llog_cb_t cb, void *data)
+{
+ struct llog_process_data d;
+ struct llog_process_cat_data cd;
+ struct llog_log_hdr *llh = cat_llh->lgh_hdr;
+ int rc;
+ ENTRY;
+
+ LASSERT(llh->llh_flags & LLOG_F_IS_CAT);
+ d.lpd_data = data;
+ d.lpd_cb = cb;
+
+ if (llh->llh_cat_idx > cat_llh->lgh_last_idx) {
+ CWARN("catalog "LPX64" crosses index zero\n",
+ cat_llh->lgh_id.lgl_oid);
+
+ cd.lpcd_first_idx = 0;
+ cd.lpcd_last_idx = cat_llh->lgh_last_idx;
+ rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
+ &d, &cd);
+ if (rc != 0)
+ RETURN(rc);
+
+ cd.lpcd_first_idx = le32_to_cpu(llh->llh_cat_idx);
+ cd.lpcd_last_idx = 0;
+ rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
+ &d, &cd);
+ } else {
+ rc = llog_reverse_process(cat_llh, llog_cat_reverse_process_cb,
+ &d, NULL);
+ }
+
+ RETURN(rc);
+}
+EXPORT_SYMBOL(llog_cat_reverse_process);
+
int llog_cat_set_first_idx(struct llog_handle *cathandle, int index)
{
struct llog_log_hdr *llh = cathandle->lgh_hdr;
int i, bitmap_size, idx;
ENTRY;
- bitmap_size = sizeof(llh->llh_bitmap) * 8;
+ bitmap_size = LLOG_BITMAP_SIZE(llh);
if (llh->llh_cat_idx == (index - 1)) {
idx = llh->llh_cat_idx + 1;
llh->llh_cat_idx = idx;
}
}
out:
- CDEBUG(D_HA, "set catlog "LPX64" first idx %u\n",
+ CDEBUG(D_RPCTRACE, "set catlog "LPX64" first idx %u\n",
cathandle->lgh_id.lgl_oid, llh->llh_cat_idx);
}
RETURN(0);
}
-
-#if 0
-/* Assumes caller has already pushed us into the kernel context. */
-int llog_cat_init(struct llog_handle *cathandle, struct obd_uuid *tgtuuid)
-{
- struct llog_log_hdr *llh;
- loff_t offset = 0;
- int rc = 0;
- ENTRY;
-
- LASSERT(sizeof(*llh) == LLOG_CHUNK_SIZE);
-
- down(&cathandle->lgh_lock);
- llh = cathandle->lgh_hdr;
-
- if (cathandle->lgh_file->f_dentry->d_inode->i_size == 0) {
- llog_write_rec(cathandle, &llh->llh_hdr, NULL, 0, NULL, 0);
-
-write_hdr:
- rc = lustre_fwrite(cathandle->lgh_file, llh, LLOG_CHUNK_SIZE,
- &offset);
- if (rc != LLOG_CHUNK_SIZE) {
- CERROR("error writing catalog header: rc %d\n", rc);
- OBD_FREE(llh, sizeof(*llh));
- if (rc >= 0)
- rc = -ENOSPC;
- } else
- rc = 0;
- } else {
- rc = lustre_fread(cathandle->lgh_file, llh, LLOG_CHUNK_SIZE,
- &offset);
- if (rc != LLOG_CHUNK_SIZE) {
- CERROR("error reading catalog header: rc %d\n", rc);
- /* Can we do much else if the header is bad? */
- goto write_hdr;
- } else
- rc = 0;
- }
-
- cathandle->lgh_tgtuuid = &llh->llh_tgtuuid;
- up(&cathandle->lgh_lock);
- RETURN(rc);
-}
-EXPORT_SYMBOL(llog_cat_init);
-
-#endif