aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorJoel Becker <joel.becker@oracle.com>2008-02-01 18:08:23 -0500
committerMark Fasheh <mfasheh@suse.com>2008-04-18 11:56:05 -0400
commitb61817e1166c5e19c08baf05196477cc345e1b1a (patch)
tree8da1c387086313aecdbb8f96fd0ab33417860620 /fs
parent74ae4e104dfc57017783fc07d5f2f9129062207f (diff)
ocfs2: Add the USERSPACE_STACK incompat bit.
The filesystem gains the USERSPACE_STACK incomat bit and the s_cluster_info field on the superblock. When a userspace stack is in use, the name of the stack is stored on-disk for mount-time verification. The "cluster_stack" option is added to mount(2) processing. The mount process needs to pass the matching stack name. If the passed name and the on-disk name do not match, the mount is failed. When using the classic o2cb stack, the incompat bit is *not* set and no mount option is used other than the usual heartbeat=local. Thus, the filesystem is compatible with older tools. Signed-off-by: Joel Becker <joel.becker@oracle.com> Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Diffstat (limited to 'fs')
-rw-r--r--fs/ocfs2/ocfs2.h7
-rw-r--r--fs/ocfs2/ocfs2_fs.h40
-rw-r--r--fs/ocfs2/super.c90
3 files changed, 134 insertions, 3 deletions
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index af929eca5412..9ff5811345a9 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -248,6 +248,7 @@ struct ocfs2_super
248 struct ocfs2_alloc_stats alloc_stats; 248 struct ocfs2_alloc_stats alloc_stats;
249 char dev_str[20]; /* "major,minor" of the device */ 249 char dev_str[20]; /* "major,minor" of the device */
250 250
251 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
251 struct ocfs2_cluster_connection *cconn; 252 struct ocfs2_cluster_connection *cconn;
252 struct ocfs2_lock_res osb_super_lockres; 253 struct ocfs2_lock_res osb_super_lockres;
253 struct ocfs2_lock_res osb_rename_lockres; 254 struct ocfs2_lock_res osb_rename_lockres;
@@ -368,6 +369,12 @@ static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
368 return ret; 369 return ret;
369} 370}
370 371
372static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
373{
374 return (osb->s_feature_incompat &
375 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
376}
377
371static inline int ocfs2_mount_local(struct ocfs2_super *osb) 378static inline int ocfs2_mount_local(struct ocfs2_super *osb)
372{ 379{
373 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); 380 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index c49502329ab0..52c426665154 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -89,7 +89,8 @@
89#define OCFS2_FEATURE_INCOMPAT_SUPP (OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \ 89#define OCFS2_FEATURE_INCOMPAT_SUPP (OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \
90 | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \ 90 | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \
91 | OCFS2_FEATURE_INCOMPAT_INLINE_DATA \ 91 | OCFS2_FEATURE_INCOMPAT_INLINE_DATA \
92 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP) 92 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP \
93 | OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK)
93#define OCFS2_FEATURE_RO_COMPAT_SUPP OCFS2_FEATURE_RO_COMPAT_UNWRITTEN 94#define OCFS2_FEATURE_RO_COMPAT_SUPP OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
94 95
95/* 96/*
@@ -131,6 +132,17 @@
131 132
132 133
133/* 134/*
135 * Support for alternate, userspace cluster stacks. If set, the superblock
136 * field s_cluster_info contains a tag for the alternate stack in use as
137 * well as the name of the cluster being joined.
138 * mount.ocfs2 must pass in a matching stack name.
139 *
140 * If not set, the classic stack will be used. This is compatbile with
141 * all older versions.
142 */
143#define OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK 0x0080
144
145/*
134 * backup superblock flag is used to indicate that this volume 146 * backup superblock flag is used to indicate that this volume
135 * has backup superblocks. 147 * has backup superblocks.
136 */ 148 */
@@ -272,6 +284,10 @@ struct ocfs2_new_group_input {
272#define OCFS2_VOL_UUID_LEN 16 284#define OCFS2_VOL_UUID_LEN 16
273#define OCFS2_MAX_VOL_LABEL_LEN 64 285#define OCFS2_MAX_VOL_LABEL_LEN 64
274 286
287/* The alternate, userspace stack fields */
288#define OCFS2_STACK_LABEL_LEN 4
289#define OCFS2_CLUSTER_NAME_LEN 16
290
275/* Journal limits (in bytes) */ 291/* Journal limits (in bytes) */
276#define OCFS2_MIN_JOURNAL_SIZE (4 * 1024 * 1024) 292#define OCFS2_MIN_JOURNAL_SIZE (4 * 1024 * 1024)
277 293
@@ -513,6 +529,13 @@ struct ocfs2_slot_map_extended {
513 */ 529 */
514}; 530};
515 531
532struct ocfs2_cluster_info {
533/*00*/ __u8 ci_stack[OCFS2_STACK_LABEL_LEN];
534 __le32 ci_reserved;
535/*08*/ __u8 ci_cluster[OCFS2_CLUSTER_NAME_LEN];
536/*18*/
537};
538
516/* 539/*
517 * On disk superblock for OCFS2 540 * On disk superblock for OCFS2
518 * Note that it is contained inside an ocfs2_dinode, so all offsets 541 * Note that it is contained inside an ocfs2_dinode, so all offsets
@@ -545,7 +568,20 @@ struct ocfs2_super_block {
545 * group header */ 568 * group header */
546/*50*/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */ 569/*50*/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */
547/*90*/ __u8 s_uuid[OCFS2_VOL_UUID_LEN]; /* 128-bit uuid */ 570/*90*/ __u8 s_uuid[OCFS2_VOL_UUID_LEN]; /* 128-bit uuid */
548/*A0*/ 571/*A0*/ struct ocfs2_cluster_info s_cluster_info; /* Selected userspace
572 stack. Only valid
573 with INCOMPAT flag. */
574/*B8*/ __le64 s_reserved2[17]; /* Fill out superblock */
575/*140*/
576
577 /*
578 * NOTE: As stated above, all offsets are relative to
579 * ocfs2_dinode.id2, which is at 0xC0 in the inode.
580 * 0xC0 + 0x140 = 0x200 or 512 bytes. A superblock must fit within
581 * our smallest blocksize, which is 512 bytes. To ensure this,
582 * we reserve the space in s_reserved2. Anything past s_reserved2
583 * will not be available on the smallest blocksize.
584 */
549}; 585};
550 586
551/* 587/*
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index e27a0d47ea2b..96ebe36d5d77 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -87,6 +87,7 @@ struct mount_options
87 unsigned int atime_quantum; 87 unsigned int atime_quantum;
88 signed short slot; 88 signed short slot;
89 unsigned int localalloc_opt; 89 unsigned int localalloc_opt;
90 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
90}; 91};
91 92
92static int ocfs2_parse_options(struct super_block *sb, char *options, 93static int ocfs2_parse_options(struct super_block *sb, char *options,
@@ -152,6 +153,7 @@ enum {
152 Opt_commit, 153 Opt_commit,
153 Opt_localalloc, 154 Opt_localalloc,
154 Opt_localflocks, 155 Opt_localflocks,
156 Opt_stack,
155 Opt_err, 157 Opt_err,
156}; 158};
157 159
@@ -170,6 +172,7 @@ static match_table_t tokens = {
170 {Opt_commit, "commit=%u"}, 172 {Opt_commit, "commit=%u"},
171 {Opt_localalloc, "localalloc=%d"}, 173 {Opt_localalloc, "localalloc=%d"},
172 {Opt_localflocks, "localflocks"}, 174 {Opt_localflocks, "localflocks"},
175 {Opt_stack, "cluster_stack=%s"},
173 {Opt_err, NULL} 176 {Opt_err, NULL}
174}; 177};
175 178
@@ -549,8 +552,17 @@ static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
549 } 552 }
550 } 553 }
551 554
555 if (ocfs2_userspace_stack(osb)) {
556 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
557 mlog(ML_ERROR, "Userspace stack expected, but "
558 "o2cb heartbeat arguments passed to mount\n");
559 return -EINVAL;
560 }
561 }
562
552 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 563 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
553 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) { 564 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
565 !ocfs2_userspace_stack(osb)) {
554 mlog(ML_ERROR, "Heartbeat has to be started to mount " 566 mlog(ML_ERROR, "Heartbeat has to be started to mount "
555 "a read-write clustered device.\n"); 567 "a read-write clustered device.\n");
556 return -EINVAL; 568 return -EINVAL;
@@ -560,6 +572,35 @@ static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
560 return 0; 572 return 0;
561} 573}
562 574
575/*
576 * If we're using a userspace stack, mount should have passed
577 * a name that matches the disk. If not, mount should not
578 * have passed a stack.
579 */
580static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
581 struct mount_options *mopt)
582{
583 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
584 mlog(ML_ERROR,
585 "cluster stack passed to mount, but this filesystem "
586 "does not support it\n");
587 return -EINVAL;
588 }
589
590 if (ocfs2_userspace_stack(osb) &&
591 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
592 OCFS2_STACK_LABEL_LEN)) {
593 mlog(ML_ERROR,
594 "cluster stack passed to mount (\"%s\") does not "
595 "match the filesystem (\"%s\")\n",
596 mopt->cluster_stack,
597 osb->osb_cluster_stack);
598 return -EINVAL;
599 }
600
601 return 0;
602}
603
563static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 604static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
564{ 605{
565 struct dentry *root; 606 struct dentry *root;
@@ -598,6 +639,10 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
598 osb->osb_commit_interval = parsed_options.commit_interval; 639 osb->osb_commit_interval = parsed_options.commit_interval;
599 osb->local_alloc_size = parsed_options.localalloc_opt; 640 osb->local_alloc_size = parsed_options.localalloc_opt;
600 641
642 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
643 if (status)
644 goto read_super_error;
645
601 sb->s_magic = OCFS2_SUPER_MAGIC; 646 sb->s_magic = OCFS2_SUPER_MAGIC;
602 647
603 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, 648 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
@@ -752,6 +797,7 @@ static int ocfs2_parse_options(struct super_block *sb,
752 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 797 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
753 mopt->slot = OCFS2_INVALID_SLOT; 798 mopt->slot = OCFS2_INVALID_SLOT;
754 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE; 799 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
800 mopt->cluster_stack[0] = '\0';
755 801
756 if (!options) { 802 if (!options) {
757 status = 1; 803 status = 1;
@@ -853,6 +899,25 @@ static int ocfs2_parse_options(struct super_block *sb,
853 if (!is_remount) 899 if (!is_remount)
854 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 900 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
855 break; 901 break;
902 case Opt_stack:
903 /* Check both that the option we were passed
904 * is of the right length and that it is a proper
905 * string of the right length.
906 */
907 if (((args[0].to - args[0].from) !=
908 OCFS2_STACK_LABEL_LEN) ||
909 (strnlen(args[0].from,
910 OCFS2_STACK_LABEL_LEN) !=
911 OCFS2_STACK_LABEL_LEN)) {
912 mlog(ML_ERROR,
913 "Invalid cluster_stack option\n");
914 status = 0;
915 goto bail;
916 }
917 memcpy(mopt->cluster_stack, args[0].from,
918 OCFS2_STACK_LABEL_LEN);
919 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
920 break;
856 default: 921 default:
857 mlog(ML_ERROR, 922 mlog(ML_ERROR,
858 "Unrecognized mount option \"%s\" " 923 "Unrecognized mount option \"%s\" "
@@ -911,6 +976,10 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
911 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 976 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
912 seq_printf(s, ",localflocks,"); 977 seq_printf(s, ",localflocks,");
913 978
979 if (osb->osb_cluster_stack[0])
980 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
981 osb->osb_cluster_stack);
982
914 return 0; 983 return 0;
915} 984}
916 985
@@ -1403,6 +1472,25 @@ static int ocfs2_initialize_super(struct super_block *sb,
1403 goto bail; 1472 goto bail;
1404 } 1473 }
1405 1474
1475 if (ocfs2_userspace_stack(osb)) {
1476 memcpy(osb->osb_cluster_stack,
1477 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
1478 OCFS2_STACK_LABEL_LEN);
1479 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1480 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
1481 mlog(ML_ERROR,
1482 "couldn't mount because of an invalid "
1483 "cluster stack label (%s) \n",
1484 osb->osb_cluster_stack);
1485 status = -EINVAL;
1486 goto bail;
1487 }
1488 } else {
1489 /* The empty string is identical with classic tools that
1490 * don't know about s_cluster_info. */
1491 osb->osb_cluster_stack[0] = '\0';
1492 }
1493
1406 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 1494 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1407 1495
1408 /* FIXME 1496 /* FIXME