/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * lustre/obdclass/obd_mount.c
- * Client/server mount routines
+ * GPL HEADER START
*
- * Copyright (c) 2006 Cluster File Systems, Inc.
- * Author: Nathan Rutman <nathan@clusterfs.com>
+ * 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 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/obd_mount.c
+ *
+ * Client/server mount routines
+ *
+ * Author: Nathan Rutman <nathan@clusterfs.com>
*/
/*********** mount lookup *********/
DECLARE_MUTEX(lustre_mount_info_lock);
-struct list_head server_mount_info_list = LIST_HEAD_INIT(server_mount_info_list);
+static CFS_LIST_HEAD(server_mount_info_list);
static struct lustre_mount_info *server_find_mount(const char *name)
{
"communication errors between this node and "
"the MGS, a bad configuration, or other "
"errors. See the syslog for more "
- "information.\n", mgc->obd_name, logname,
+ "information.\n", mgc->obd_name, logname,
rc);
/* class_obd_list(); */
DECLARE_MUTEX(mgc_start_lock);
-/* Set up a mgcobd to process startup logs */
+/** Set up a mgc obd to process startup logs
+ *
+ * \param sb [in] super block of the mgc obd
+ *
+ * \retval 0 success, otherwise error code
+ */
static int lustre_start_mgc(struct super_block *sb)
{
struct lustre_handle mgc_conn = {0, };
- struct obd_connect_data ocd = { 0 };
+ struct obd_connect_data *data = NULL;
struct lustre_sb_info *lsi = s2lsi(sb);
struct obd_device *obd;
struct obd_export *exp;
mutex_down(&mgc_start_lock);
obd = class_name2obd(mgcname);
- if (obd) {
+ if (obd && !obd->obd_stopping) {
/* Re-using an existing MGC */
atomic_inc(&obd->u.cli.cl_mgc_refcount);
recov_bk++;
CDEBUG(D_MOUNT, "%s: Set MGC reconnect %d\n", mgcname,recov_bk);
rc = obd_set_info_async(obd->obd_self_export,
- strlen(KEY_INIT_RECOV_BACKUP),
+ sizeof(KEY_INIT_RECOV_BACKUP),
KEY_INIT_RECOV_BACKUP,
sizeof(recov_bk), &recov_bk, NULL);
GOTO(out, rc = 0);
/* Try all connections, but only once. */
recov_bk = 1;
rc = obd_set_info_async(obd->obd_self_export,
- strlen(KEY_INIT_RECOV_BACKUP),
+ sizeof(KEY_INIT_RECOV_BACKUP),
KEY_INIT_RECOV_BACKUP,
sizeof(recov_bk), &recov_bk, NULL);
if (rc)
/* nonfatal */
- CERROR("can't set %s %d\n", KEY_INIT_RECOV_BACKUP, rc);
+ CWARN("can't set %s %d\n", KEY_INIT_RECOV_BACKUP, rc);
/* We connect to the MGS at setup, and don't disconnect until cleanup */
-
- ocd.ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_FID;
- ocd.ocd_version = LUSTRE_VERSION_CODE;
-
- rc = obd_connect(NULL, &mgc_conn, obd, &(obd->obd_uuid), &ocd);
+ OBD_ALLOC_PTR(data);
+ if (data == NULL)
+ GOTO(out, rc = -ENOMEM);
+ data->ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_FID |
+ OBD_CONNECT_AT;
+ data->ocd_version = LUSTRE_VERSION_CODE;
+ rc = obd_connect(NULL, &mgc_conn, obd, &(obd->obd_uuid), data, NULL);
+ OBD_FREE_PTR(data);
if (rc) {
CERROR("connect failed %d\n", rc);
GOTO(out, rc);
obd = lsi->lsi_mgc;
if (!obd)
RETURN(-ENOENT);
-
lsi->lsi_mgc = NULL;
+
mutex_down(&mgc_start_lock);
+ LASSERT(atomic_read(&obd->u.cli.cl_mgc_refcount) > 0);
if (!atomic_dec_and_test(&obd->u.cli.cl_mgc_refcount)) {
/* This is not fatal, every client that stops
will call in here. */
GOTO(out, rc = -EBUSY);
}
- /* The MGC has no recoverable data in any case.
+ /* The MGC has no recoverable data in any case.
* force shotdown set in umount_begin */
obd->obd_no_recov = 1;
/* Clean the nid uuids */
if (!niduuid)
- RETURN(-ENOMEM);
+ GOTO(out, rc = -ENOMEM);
+
for (i = 0; i < lsi->lsi_lmd->lmd_mgs_failnodes; i++) {
sprintf(ptr, "_%x", i);
rc = do_lcfg(LUSTRE_MGC_OBDNAME, 0, LCFG_DEL_UUID,
CERROR("del MDC UUID %s failed: rc = %d\n",
niduuid, rc);
}
- OBD_FREE(niduuid, len);
- /* class_import_put will get rid of the additional connections */
-
out:
+ if (niduuid)
+ OBD_FREE(niduuid, len);
+
+ /* class_import_put will get rid of the additional connections */
mutex_up(&mgc_start_lock);
RETURN(rc);
}
/* cl_mgc_sem in mgc insures we sleep if the mgc_fs is busy */
rc = obd_set_info_async(mgc->obd_self_export,
- strlen("set_fs"), "set_fs",
+ sizeof(KEY_SET_FS), KEY_SET_FS,
sizeof(*sb), sb, NULL);
if (rc) {
CERROR("can't set_fs %d\n", rc);
CDEBUG(D_MOUNT, "Unassign mgc disk\n");
rc = obd_set_info_async(mgc->obd_self_export,
- strlen("clear_fs"), "clear_fs",
+ sizeof(KEY_CLEAR_FS), KEY_CLEAR_FS,
0, NULL, NULL);
RETURN(rc);
}
/* Register the target */
/* FIXME use mgc_process_config instead */
rc = obd_set_info_async(mgc->u.cli.cl_mgc_mgsexp,
- strlen("register_target"), "register_target",
+ sizeof(KEY_REGISTER_TARGET), KEY_REGISTER_TARGET,
sizeof(*mti), mti, NULL);
- if (rc) {
- CERROR("registration with the MGS failed (%d)\n", rc);
+ if (rc)
GOTO(out, rc);
- }
/* Always update our flags */
ldd->ldd_flags = mti->mti_flags & ~LDD_F_REWRITE_LDD;
lsi->lsi_ldd->ldd_svname);
GOTO(out_mgc, rc);
}
+ /* non-fatal error of registeration with MGS */
+ if (rc)
+ CDEBUG(D_MOUNT, "Cannot register with MGS: %d\n", rc);
/* Let the target look up the mount using the target's name
(we can't pass the sb or mnt through class_process_config.) */
if (rc) {
CERROR("failed to start server %s: %d\n",
lsi->lsi_ldd->ldd_svname, rc);
+ server_deregister_mount(lsi->lsi_ldd->ldd_svname);
GOTO(out_mgc, rc);
}
struct lustre_sb_info *lustre_init_lsi(struct super_block *sb)
{
- struct lustre_sb_info *lsi = NULL;
+ struct lustre_sb_info *lsi;
ENTRY;
- OBD_ALLOC(lsi, sizeof(*lsi));
+ OBD_ALLOC_PTR(lsi);
if (!lsi)
RETURN(NULL);
- OBD_ALLOC(lsi->lsi_lmd, sizeof(*lsi->lsi_lmd));
+ OBD_ALLOC_PTR(lsi->lsi_lmd);
if (!lsi->lsi_lmd) {
- OBD_FREE(lsi, sizeof(*lsi));
+ OBD_FREE_PTR(lsi);
RETURN(NULL);
}
struct lustre_sb_info *lsi = s2lsi(sb);
ENTRY;
- if (!lsi)
- RETURN(0);
-
- CDEBUG(D_MOUNT, "Freeing lsi\n");
+ LASSERT(lsi != NULL);
+ CDEBUG(D_MOUNT, "Freeing lsi %p\n", lsi);
/* someone didn't call server_put_mount. */
LASSERT(atomic_read(&lsi->lsi_mounts) == 0);
struct lustre_sb_info *lsi = s2lsi(sb);
ENTRY;
- LASSERT(lsi);
+ LASSERT(lsi != NULL);
CDEBUG(D_MOUNT, "put %p %d\n", sb, atomic_read(&lsi->lsi_mounts));
-
if (atomic_dec_and_test(&lsi->lsi_mounts)) {
lustre_free_lsi(sb);
RETURN(1);
Note ext3/ldiskfs can't be mounted ro. */
s_flags = sb->s_flags;
+ /* allocate memory for options */
+ OBD_PAGE_ALLOC(__page, CFS_ALLOC_STD);
+ if (!__page)
+ GOTO(out_free, rc = -ENOMEM);
+ page = (unsigned long)cfs_page_address(__page);
+ options = (char *)page;
+ memset(options, 0, CFS_PAGE_SIZE);
+
+ /* mount-line options must be added for pre-mount because it may
+ * contain mount options such as journal_dev which are required
+ * to mount successfuly the underlying filesystem */
+ if (lmd->lmd_opts && (*(lmd->lmd_opts) != 0))
+ strncat(options, lmd->lmd_opts, CFS_PAGE_SIZE - 1);
+
/* Pre-mount ldiskfs to read the MOUNT_DATA_FILE */
CDEBUG(D_MOUNT, "Pre-mount ldiskfs %s\n", lmd->lmd_dev);
- mnt = ll_kern_mount("ldiskfs", s_flags, lmd->lmd_dev, 0);
+ mnt = ll_kern_mount("ldiskfs", s_flags, lmd->lmd_dev, (void *)options);
if (IS_ERR(mnt)) {
rc = PTR_ERR(mnt);
CERROR("premount %s:%#lx ldiskfs failed: %d "
/* Done with our pre-mount, now do the real mount. */
/* Glom up mount options */
- OBD_PAGE_ALLOC(__page, CFS_ALLOC_STD);
- if (!__page)
- GOTO(out_free, rc = -ENOMEM);
- page = (unsigned long)cfs_page_address(__page);
-
- options = (char *)page;
memset(options, 0, CFS_PAGE_SIZE);
strncpy(options, ldd->ldd_mount_opts, CFS_PAGE_SIZE - 2);
MT_STR(ldd), lmd->lmd_dev, options);
mnt = ll_kern_mount(MT_STR(ldd), s_flags, lmd->lmd_dev,
(void *)options);
- OBD_PAGE_FREE(__page);
if (IS_ERR(mnt)) {
rc = PTR_ERR(mnt);
CERROR("ll_kern_mount failed: rc = %d\n", rc);
GOTO(out_free, rc);
}
+ OBD_PAGE_FREE(__page);
lsi->lsi_ldd = ldd; /* freed at lsi cleanup */
CDEBUG(D_SUPER, "%s: mnt = %p\n", lmd->lmd_dev, mnt);
RETURN(mnt);
out_free:
+ if (__page)
+ OBD_PAGE_FREE(__page);
OBD_FREE(ldd, sizeof(*ldd));
lsi->lsi_ldd = NULL;
RETURN(ERR_PTR(rc));
int tmpname_sz;
int lddflags = lsi->lsi_ldd->ldd_flags;
int lsiflags = lsi->lsi_flags;
- int rc;
ENTRY;
LASSERT(lsiflags & LSI_SERVER);
CDEBUG(D_MOUNT, "server put_super %s\n", tmpname);
/* Stop the target */
- if (IS_MDT(lsi->lsi_ldd) || IS_OST(lsi->lsi_ldd)) {
+ if (!(lsi->lsi_lmd->lmd_flags & LMD_FLG_NOSVC) &&
+ (IS_MDT(lsi->lsi_ldd) || IS_OST(lsi->lsi_ldd))) {
struct lustre_profile *lprof = NULL;
/* tell the mgc to drop the config log */
/* If they wanted the mgs to stop separately from the mdt, they
should have put it on a different device. */
if (IS_MGS(lsi->lsi_ldd)) {
- /* stop the mgc before the mgs so the connection gets cleaned
- up */
- lustre_stop_mgc(sb);
- server_stop_mgs(sb);
+ /* if MDS start with --nomgs, don't stop MGS then */
+ if (!(lsi->lsi_lmd->lmd_flags & LMD_FLG_NOMGS))
+ server_stop_mgs(sb);
}
/* Clean the mgc and sb */
- rc = lustre_common_put_super(sb);
- /* FIXME how can I report a failure to umount? */
+ lustre_common_put_super(sb);
/* Wait for the targets to really clean up - can't exit (and let the
sb get destroyed) while the mount is still in use */
if (IS_ERR(mnt)) {
rc = PTR_ERR(mnt);
CERROR("Unable to mount device %s: %d\n",
- lsi->lsi_lmd->lmd_dev, rc);
+ lsi->lsi_lmd->lmd_dev, rc);
lustre_put_lsi(sb);
- GOTO(out, rc);
+ RETURN(rc);
}
lsi->lsi_srv_mnt = mnt;
"running. Double-mount may have compromised"
" the disk journal.\n",
lsi->lsi_ldd->ldd_svname);
- unlock_mntput(mnt);
lustre_put_lsi(sb);
- GOTO(out, rc = -EALREADY);
+ unlock_mntput(mnt);
+ RETURN(-EALREADY);
}
- /* start MGS before MGC */
- if (IS_MGS(lsi->lsi_ldd)) {
+ /* Start MGS before MGC */
+ if (IS_MGS(lsi->lsi_ldd) && !(lsi->lsi_lmd->lmd_flags & LMD_FLG_NOMGS)) {
rc = server_start_mgs(sb);
if (rc)
GOTO(out_mnt, rc);
lsi->lsi_ldd->ldd_svname, lsi->lsi_lmd->lmd_dev);
RETURN(0);
-
out_mnt:
+ /* We jump here in case of failure while starting targets or MGS.
+ * In this case we can't just put @mnt and have to do real cleanup
+ * with stoping targets, etc. */
server_put_super(sb);
-out:
- RETURN(rc);
+ return rc;
}
/* Get the index from the obd name.
}
/* Drop a ref to the mounted disk */
lustre_put_lsi(sb);
+ lu_types_stop();
RETURN(rc);
}
} else if (strncmp(s1, "nosvc", 5) == 0) {
lmd->lmd_flags |= LMD_FLG_NOSVC;
clear++;
+ } else if (strncmp(s1, "nomgs", 5) == 0) {
+ lmd->lmd_flags |= LMD_FLG_NOMGS;
+ clear++;
/* ost exclusion list */
} else if (strncmp(s1, "exclude=", 8) == 0) {
rc = lmd_make_exclusion(lmd, s1 + 7);
goto invalid;
}
- s1 = strrchr(devname, ':');
+ s1 = strstr(devname, ":/");
if (s1) {
+ ++s1;
lmd->lmd_flags = LMD_FLG_CLIENT;
/* Remove leading /s from fsname */
while (*++s1 == '/') ;
RETURN(-ENOMEM);
lmd = lsi->lsi_lmd;
+ /*
+ * Disable lockdep during mount, because mount locking patterns are
+ * `special'.
+ */
+ lockdep_off();
+
/* Figure out the lmd from the mount options */
if (lmd_parse((char *)data, lmd)) {
lustre_put_lsi(sb);
- RETURN(-EINVAL);
+ GOTO(out, rc = -EINVAL);
}
if (lmd_is_client(lmd)) {
LCONSOLE_ERROR_MSG(0x165, "Nothing registered for "
"client mount! Is the 'lustre' "
"module loaded?\n");
+ lustre_put_lsi(sb);
rc = -ENODEV;
} else {
rc = lustre_start_mgc(sb);
if (rc) {
- lustre_stop_mgc(sb);
- goto out;
+ lustre_put_lsi(sb);
+ GOTO(out, rc);
}
/* Connect and start */
/* (should always be ll_fill_super) */
/* s_f_s will call server_put_super on failure */
}
+ /* If error happens in fill_super() call, @lsi will be killed there.
+ * This is why we do not put it here. */
+ GOTO(out, rc);
out:
- if (rc){
+ if (rc) {
CERROR("Unable to mount %s (%d)\n",
s2lsi(sb) ? lmd->lmd_dev : "", rc);
} else {
- CDEBUG(D_SUPER, "mount %s complete\n", lmd->lmd_dev);
+ CDEBUG(D_SUPER, "Mount %s complete\n",
+ lmd->lmd_dev);
}
- RETURN(rc);
+ lockdep_on();
+ return rc;
}
struct super_block * lustre_get_sb(struct file_system_type *fs_type,
int flags, const char *devname, void * data)
{
- /* calls back in fill super */
- /* we could append devname= onto options (*data) here,
- but 2.4 doesn't get devname. So we do it in mount_lustre.c */
return get_sb_nodev(fs_type, flags, data, lustre_fill_super);
}
#else
int flags, const char *devname, void * data,
struct vfsmount *mnt)
{
- /* calls back in fill super */
- /* we could append devname= onto options (*data) here,
- but 2.4 doesn't get devname. So we do it in mount_lustre.c */
return get_sb_nodev(fs_type, flags, data, lustre_fill_super, mnt);
}
#endif
{
struct lustre_sb_info *lsi = s2lsi(sb);
- if (kill_super_cb && lsi &&(lsi->lsi_flags & LSI_SERVER))
+ if (kill_super_cb && lsi && !(lsi->lsi_flags & LSI_SERVER))
(*kill_super_cb)(sb);
kill_anon_super(sb);
.name = "lustre",
.get_sb = lustre_get_sb,
.kill_sb = lustre_kill_super,
- .fs_flags = FS_BINARY_MOUNTDATA | FS_REQUIRES_DEV,
+ .fs_flags = FS_BINARY_MOUNTDATA | FS_REQUIRES_DEV |
+ LL_RENAME_DOES_D_MOVE,
};
int lustre_register_fs(void)
EXPORT_SYMBOL(server_name2index);
EXPORT_SYMBOL(server_mti_print);
EXPORT_SYMBOL(do_lcfg);
-
-
git://git.whamcloud.com / fs / lustre-release.git / blobdiff