2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License version 2.
10 #include <linux/sched.h>
11 #include <linux/slab.h>
12 #include <linux/spinlock.h>
13 #include <linux/completion.h>
14 #include <linux/buffer_head.h>
15 #include <linux/blkdev.h>
16 #include <linux/kthread.h>
17 #include <linux/namei.h>
18 #include <linux/mount.h>
19 #include <linux/gfs2_ondisk.h>
20 #include <linux/slow-work.h>
21 #include <linux/quotaops.h>
37 #include "trace_gfs2.h"
42 static const u32 gfs2_old_fs_formats
[] = {
46 static const u32 gfs2_old_multihost_formats
[] = {
51 * gfs2_tune_init - Fill a gfs2_tune structure with default values
56 static void gfs2_tune_init(struct gfs2_tune
*gt
)
58 spin_lock_init(>
->gt_spin
);
60 gt
->gt_incore_log_blocks
= 1024;
62 gt
->gt_quota_simul_sync
= 64;
63 gt
->gt_quota_warn_period
= 10;
64 gt
->gt_quota_scale_num
= 1;
65 gt
->gt_quota_scale_den
= 1;
66 gt
->gt_new_files_jdata
= 0;
67 gt
->gt_max_readahead
= 1 << 18;
68 gt
->gt_stall_secs
= 600;
69 gt
->gt_complain_secs
= 10;
72 static struct gfs2_sbd
*init_sbd(struct super_block
*sb
)
76 sdp
= kzalloc(sizeof(struct gfs2_sbd
), GFP_KERNEL
);
83 gfs2_tune_init(&sdp
->sd_tune
);
85 spin_lock_init(&sdp
->sd_statfs_spin
);
87 spin_lock_init(&sdp
->sd_rindex_spin
);
88 mutex_init(&sdp
->sd_rindex_mutex
);
89 INIT_LIST_HEAD(&sdp
->sd_rindex_list
);
90 INIT_LIST_HEAD(&sdp
->sd_rindex_mru_list
);
92 INIT_LIST_HEAD(&sdp
->sd_jindex_list
);
93 spin_lock_init(&sdp
->sd_jindex_spin
);
94 mutex_init(&sdp
->sd_jindex_mutex
);
96 INIT_LIST_HEAD(&sdp
->sd_quota_list
);
97 mutex_init(&sdp
->sd_quota_mutex
);
98 init_waitqueue_head(&sdp
->sd_quota_wait
);
99 INIT_LIST_HEAD(&sdp
->sd_trunc_list
);
100 spin_lock_init(&sdp
->sd_trunc_lock
);
102 spin_lock_init(&sdp
->sd_log_lock
);
104 INIT_LIST_HEAD(&sdp
->sd_log_le_buf
);
105 INIT_LIST_HEAD(&sdp
->sd_log_le_revoke
);
106 INIT_LIST_HEAD(&sdp
->sd_log_le_rg
);
107 INIT_LIST_HEAD(&sdp
->sd_log_le_databuf
);
108 INIT_LIST_HEAD(&sdp
->sd_log_le_ordered
);
110 mutex_init(&sdp
->sd_log_reserve_mutex
);
111 INIT_LIST_HEAD(&sdp
->sd_ail1_list
);
112 INIT_LIST_HEAD(&sdp
->sd_ail2_list
);
114 init_rwsem(&sdp
->sd_log_flush_lock
);
115 atomic_set(&sdp
->sd_log_in_flight
, 0);
116 init_waitqueue_head(&sdp
->sd_log_flush_wait
);
118 INIT_LIST_HEAD(&sdp
->sd_revoke_list
);
120 mutex_init(&sdp
->sd_freeze_lock
);
127 * gfs2_check_sb - Check superblock
128 * @sdp: the filesystem
129 * @sb: The superblock
130 * @silent: Don't print a message if the check fails
132 * Checks the version code of the FS is one that we understand how to
133 * read and that the sizes of the various on-disk structures have not
137 static int gfs2_check_sb(struct gfs2_sbd
*sdp
, struct gfs2_sb_host
*sb
, int silent
)
141 if (sb
->sb_magic
!= GFS2_MAGIC
||
142 sb
->sb_type
!= GFS2_METATYPE_SB
) {
144 printk(KERN_WARNING
"GFS2: not a GFS2 filesystem\n");
148 /* If format numbers match exactly, we're done. */
150 if (sb
->sb_fs_format
== GFS2_FORMAT_FS
&&
151 sb
->sb_multihost_format
== GFS2_FORMAT_MULTI
)
154 if (sb
->sb_fs_format
!= GFS2_FORMAT_FS
) {
155 for (x
= 0; gfs2_old_fs_formats
[x
]; x
++)
156 if (gfs2_old_fs_formats
[x
] == sb
->sb_fs_format
)
159 if (!gfs2_old_fs_formats
[x
]) {
161 "GFS2: code version (%u, %u) is incompatible "
162 "with ondisk format (%u, %u)\n",
163 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
164 sb
->sb_fs_format
, sb
->sb_multihost_format
);
166 "GFS2: I don't know how to upgrade this FS\n");
171 if (sb
->sb_multihost_format
!= GFS2_FORMAT_MULTI
) {
172 for (x
= 0; gfs2_old_multihost_formats
[x
]; x
++)
173 if (gfs2_old_multihost_formats
[x
] ==
174 sb
->sb_multihost_format
)
177 if (!gfs2_old_multihost_formats
[x
]) {
179 "GFS2: code version (%u, %u) is incompatible "
180 "with ondisk format (%u, %u)\n",
181 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
182 sb
->sb_fs_format
, sb
->sb_multihost_format
);
184 "GFS2: I don't know how to upgrade this FS\n");
189 if (!sdp
->sd_args
.ar_upgrade
) {
191 "GFS2: code version (%u, %u) is incompatible "
192 "with ondisk format (%u, %u)\n",
193 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
194 sb
->sb_fs_format
, sb
->sb_multihost_format
);
196 "GFS2: Use the \"upgrade\" mount option to upgrade "
198 printk(KERN_INFO
"GFS2: See the manual for more details\n");
205 static void end_bio_io_page(struct bio
*bio
, int error
)
207 struct page
*page
= bio
->bi_private
;
210 SetPageUptodate(page
);
212 printk(KERN_WARNING
"gfs2: error %d reading superblock\n", error
);
216 static void gfs2_sb_in(struct gfs2_sb_host
*sb
, const void *buf
)
218 const struct gfs2_sb
*str
= buf
;
220 sb
->sb_magic
= be32_to_cpu(str
->sb_header
.mh_magic
);
221 sb
->sb_type
= be32_to_cpu(str
->sb_header
.mh_type
);
222 sb
->sb_format
= be32_to_cpu(str
->sb_header
.mh_format
);
223 sb
->sb_fs_format
= be32_to_cpu(str
->sb_fs_format
);
224 sb
->sb_multihost_format
= be32_to_cpu(str
->sb_multihost_format
);
225 sb
->sb_bsize
= be32_to_cpu(str
->sb_bsize
);
226 sb
->sb_bsize_shift
= be32_to_cpu(str
->sb_bsize_shift
);
227 sb
->sb_master_dir
.no_addr
= be64_to_cpu(str
->sb_master_dir
.no_addr
);
228 sb
->sb_master_dir
.no_formal_ino
= be64_to_cpu(str
->sb_master_dir
.no_formal_ino
);
229 sb
->sb_root_dir
.no_addr
= be64_to_cpu(str
->sb_root_dir
.no_addr
);
230 sb
->sb_root_dir
.no_formal_ino
= be64_to_cpu(str
->sb_root_dir
.no_formal_ino
);
232 memcpy(sb
->sb_lockproto
, str
->sb_lockproto
, GFS2_LOCKNAME_LEN
);
233 memcpy(sb
->sb_locktable
, str
->sb_locktable
, GFS2_LOCKNAME_LEN
);
234 memcpy(sb
->sb_uuid
, str
->sb_uuid
, 16);
238 * gfs2_read_super - Read the gfs2 super block from disk
239 * @sdp: The GFS2 super block
240 * @sector: The location of the super block
241 * @error: The error code to return
243 * This uses the bio functions to read the super block from disk
244 * because we want to be 100% sure that we never read cached data.
245 * A super block is read twice only during each GFS2 mount and is
246 * never written to by the filesystem. The first time its read no
247 * locks are held, and the only details which are looked at are those
248 * relating to the locking protocol. Once locking is up and working,
249 * the sb is read again under the lock to establish the location of
250 * the master directory (contains pointers to journals etc) and the
253 * Returns: 0 on success or error
256 static int gfs2_read_super(struct gfs2_sbd
*sdp
, sector_t sector
)
258 struct super_block
*sb
= sdp
->sd_vfs
;
263 page
= alloc_page(GFP_NOFS
);
267 ClearPageUptodate(page
);
268 ClearPageDirty(page
);
271 bio
= bio_alloc(GFP_NOFS
, 1);
272 bio
->bi_sector
= sector
* (sb
->s_blocksize
>> 9);
273 bio
->bi_bdev
= sb
->s_bdev
;
274 bio_add_page(bio
, page
, PAGE_SIZE
, 0);
276 bio
->bi_end_io
= end_bio_io_page
;
277 bio
->bi_private
= page
;
278 submit_bio(READ_SYNC
| (1 << BIO_RW_META
), bio
);
279 wait_on_page_locked(page
);
281 if (!PageUptodate(page
)) {
286 gfs2_sb_in(&sdp
->sd_sb
, p
);
293 * gfs2_read_sb - Read super block
294 * @sdp: The GFS2 superblock
295 * @silent: Don't print message if mount fails
299 static int gfs2_read_sb(struct gfs2_sbd
*sdp
, int silent
)
301 u32 hash_blocks
, ind_blocks
, leaf_blocks
;
306 error
= gfs2_read_super(sdp
, GFS2_SB_ADDR
>> sdp
->sd_fsb2bb_shift
);
309 fs_err(sdp
, "can't read superblock\n");
313 error
= gfs2_check_sb(sdp
, &sdp
->sd_sb
, silent
);
317 sdp
->sd_fsb2bb_shift
= sdp
->sd_sb
.sb_bsize_shift
-
318 GFS2_BASIC_BLOCK_SHIFT
;
319 sdp
->sd_fsb2bb
= 1 << sdp
->sd_fsb2bb_shift
;
320 sdp
->sd_diptrs
= (sdp
->sd_sb
.sb_bsize
-
321 sizeof(struct gfs2_dinode
)) / sizeof(u64
);
322 sdp
->sd_inptrs
= (sdp
->sd_sb
.sb_bsize
-
323 sizeof(struct gfs2_meta_header
)) / sizeof(u64
);
324 sdp
->sd_jbsize
= sdp
->sd_sb
.sb_bsize
- sizeof(struct gfs2_meta_header
);
325 sdp
->sd_hash_bsize
= sdp
->sd_sb
.sb_bsize
/ 2;
326 sdp
->sd_hash_bsize_shift
= sdp
->sd_sb
.sb_bsize_shift
- 1;
327 sdp
->sd_hash_ptrs
= sdp
->sd_hash_bsize
/ sizeof(u64
);
328 sdp
->sd_qc_per_block
= (sdp
->sd_sb
.sb_bsize
-
329 sizeof(struct gfs2_meta_header
)) /
330 sizeof(struct gfs2_quota_change
);
332 /* Compute maximum reservation required to add a entry to a directory */
334 hash_blocks
= DIV_ROUND_UP(sizeof(u64
) * (1 << GFS2_DIR_MAX_DEPTH
),
338 for (tmp_blocks
= hash_blocks
; tmp_blocks
> sdp
->sd_diptrs
;) {
339 tmp_blocks
= DIV_ROUND_UP(tmp_blocks
, sdp
->sd_inptrs
);
340 ind_blocks
+= tmp_blocks
;
343 leaf_blocks
= 2 + GFS2_DIR_MAX_DEPTH
;
345 sdp
->sd_max_dirres
= hash_blocks
+ ind_blocks
+ leaf_blocks
;
347 sdp
->sd_heightsize
[0] = sdp
->sd_sb
.sb_bsize
-
348 sizeof(struct gfs2_dinode
);
349 sdp
->sd_heightsize
[1] = sdp
->sd_sb
.sb_bsize
* sdp
->sd_diptrs
;
354 space
= sdp
->sd_heightsize
[x
- 1] * sdp
->sd_inptrs
;
356 m
= do_div(d
, sdp
->sd_inptrs
);
358 if (d
!= sdp
->sd_heightsize
[x
- 1] || m
)
360 sdp
->sd_heightsize
[x
] = space
;
362 sdp
->sd_max_height
= x
;
363 sdp
->sd_heightsize
[x
] = ~0;
364 gfs2_assert(sdp
, sdp
->sd_max_height
<= GFS2_MAX_META_HEIGHT
);
366 sdp
->sd_jheightsize
[0] = sdp
->sd_sb
.sb_bsize
-
367 sizeof(struct gfs2_dinode
);
368 sdp
->sd_jheightsize
[1] = sdp
->sd_jbsize
* sdp
->sd_diptrs
;
373 space
= sdp
->sd_jheightsize
[x
- 1] * sdp
->sd_inptrs
;
375 m
= do_div(d
, sdp
->sd_inptrs
);
377 if (d
!= sdp
->sd_jheightsize
[x
- 1] || m
)
379 sdp
->sd_jheightsize
[x
] = space
;
381 sdp
->sd_max_jheight
= x
;
382 sdp
->sd_jheightsize
[x
] = ~0;
383 gfs2_assert(sdp
, sdp
->sd_max_jheight
<= GFS2_MAX_META_HEIGHT
);
388 static int init_names(struct gfs2_sbd
*sdp
, int silent
)
393 proto
= sdp
->sd_args
.ar_lockproto
;
394 table
= sdp
->sd_args
.ar_locktable
;
396 /* Try to autodetect */
398 if (!proto
[0] || !table
[0]) {
399 error
= gfs2_read_super(sdp
, GFS2_SB_ADDR
>> sdp
->sd_fsb2bb_shift
);
403 error
= gfs2_check_sb(sdp
, &sdp
->sd_sb
, silent
);
408 proto
= sdp
->sd_sb
.sb_lockproto
;
410 table
= sdp
->sd_sb
.sb_locktable
;
414 table
= sdp
->sd_vfs
->s_id
;
416 strlcpy(sdp
->sd_proto_name
, proto
, GFS2_FSNAME_LEN
);
417 strlcpy(sdp
->sd_table_name
, table
, GFS2_FSNAME_LEN
);
419 table
= sdp
->sd_table_name
;
420 while ((table
= strchr(table
, '/')))
427 static int init_locking(struct gfs2_sbd
*sdp
, struct gfs2_holder
*mount_gh
,
435 error
= gfs2_glock_nq_num(sdp
,
436 GFS2_MOUNT_LOCK
, &gfs2_nondisk_glops
,
437 LM_ST_EXCLUSIVE
, LM_FLAG_NOEXP
| GL_NOCACHE
,
440 fs_err(sdp
, "can't acquire mount glock: %d\n", error
);
444 error
= gfs2_glock_nq_num(sdp
,
445 GFS2_LIVE_LOCK
, &gfs2_nondisk_glops
,
447 LM_FLAG_NOEXP
| GL_EXACT
,
450 fs_err(sdp
, "can't acquire live glock: %d\n", error
);
454 error
= gfs2_glock_get(sdp
, GFS2_RENAME_LOCK
, &gfs2_nondisk_glops
,
455 CREATE
, &sdp
->sd_rename_gl
);
457 fs_err(sdp
, "can't create rename glock: %d\n", error
);
461 error
= gfs2_glock_get(sdp
, GFS2_TRANS_LOCK
, &gfs2_trans_glops
,
462 CREATE
, &sdp
->sd_trans_gl
);
464 fs_err(sdp
, "can't create transaction glock: %d\n", error
);
471 gfs2_glock_put(sdp
->sd_trans_gl
);
473 gfs2_glock_put(sdp
->sd_rename_gl
);
475 gfs2_glock_dq_uninit(&sdp
->sd_live_gh
);
477 gfs2_glock_dq_uninit(mount_gh
);
482 static int gfs2_lookup_root(struct super_block
*sb
, struct dentry
**dptr
,
483 u64 no_addr
, const char *name
)
485 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
486 struct dentry
*dentry
;
489 inode
= gfs2_inode_lookup(sb
, DT_DIR
, no_addr
, 0, 0);
491 fs_err(sdp
, "can't read in %s inode: %ld\n", name
, PTR_ERR(inode
));
492 return PTR_ERR(inode
);
494 dentry
= d_alloc_root(inode
);
496 fs_err(sdp
, "can't alloc %s dentry\n", name
);
500 dentry
->d_op
= &gfs2_dops
;
505 static int init_sb(struct gfs2_sbd
*sdp
, int silent
)
507 struct super_block
*sb
= sdp
->sd_vfs
;
508 struct gfs2_holder sb_gh
;
512 ret
= gfs2_glock_nq_num(sdp
, GFS2_SB_LOCK
, &gfs2_meta_glops
,
513 LM_ST_SHARED
, 0, &sb_gh
);
515 fs_err(sdp
, "can't acquire superblock glock: %d\n", ret
);
519 ret
= gfs2_read_sb(sdp
, silent
);
521 fs_err(sdp
, "can't read superblock: %d\n", ret
);
525 /* Set up the buffer cache and SB for real */
526 if (sdp
->sd_sb
.sb_bsize
< bdev_logical_block_size(sb
->s_bdev
)) {
528 fs_err(sdp
, "FS block size (%u) is too small for device "
530 sdp
->sd_sb
.sb_bsize
, bdev_logical_block_size(sb
->s_bdev
));
533 if (sdp
->sd_sb
.sb_bsize
> PAGE_SIZE
) {
535 fs_err(sdp
, "FS block size (%u) is too big for machine "
537 sdp
->sd_sb
.sb_bsize
, (unsigned int)PAGE_SIZE
);
540 sb_set_blocksize(sb
, sdp
->sd_sb
.sb_bsize
);
542 /* Get the root inode */
543 no_addr
= sdp
->sd_sb
.sb_root_dir
.no_addr
;
544 ret
= gfs2_lookup_root(sb
, &sdp
->sd_root_dir
, no_addr
, "root");
548 /* Get the master inode */
549 no_addr
= sdp
->sd_sb
.sb_master_dir
.no_addr
;
550 ret
= gfs2_lookup_root(sb
, &sdp
->sd_master_dir
, no_addr
, "master");
552 dput(sdp
->sd_root_dir
);
555 sb
->s_root
= dget(sdp
->sd_args
.ar_meta
? sdp
->sd_master_dir
: sdp
->sd_root_dir
);
557 gfs2_glock_dq_uninit(&sb_gh
);
562 * map_journal_extents - create a reusable "extent" mapping from all logical
563 * blocks to all physical blocks for the given journal. This will save
564 * us time when writing journal blocks. Most journals will have only one
565 * extent that maps all their logical blocks. That's because gfs2.mkfs
566 * arranges the journal blocks sequentially to maximize performance.
567 * So the extent would map the first block for the entire file length.
568 * However, gfs2_jadd can happen while file activity is happening, so
569 * those journals may not be sequential. Less likely is the case where
570 * the users created their own journals by mounting the metafs and
571 * laying it out. But it's still possible. These journals might have
574 * TODO: This should be done in bigger chunks rather than one block at a time,
575 * but since it's only done at mount time, I'm not worried about the
578 static int map_journal_extents(struct gfs2_sbd
*sdp
)
580 struct gfs2_jdesc
*jd
= sdp
->sd_jdesc
;
582 u64 db
, prev_db
; /* logical block, disk block, prev disk block */
583 struct gfs2_inode
*ip
= GFS2_I(jd
->jd_inode
);
584 struct gfs2_journal_extent
*jext
= NULL
;
585 struct buffer_head bh
;
590 for (lb
= 0; lb
< ip
->i_disksize
>> sdp
->sd_sb
.sb_bsize_shift
; lb
++) {
593 bh
.b_size
= 1 << ip
->i_inode
.i_blkbits
;
594 rc
= gfs2_block_map(jd
->jd_inode
, lb
, &bh
, 0);
597 printk(KERN_INFO
"GFS2 journal mapping error %d: lb="
598 "%u db=%llu\n", rc
, lb
, (unsigned long long)db
);
601 if (!prev_db
|| db
!= prev_db
+ 1) {
602 jext
= kzalloc(sizeof(struct gfs2_journal_extent
),
605 printk(KERN_INFO
"GFS2 error: out of memory "
606 "mapping journal extents.\n");
613 list_add_tail(&jext
->extent_list
, &jd
->extent_list
);
622 static void gfs2_others_may_mount(struct gfs2_sbd
*sdp
)
624 char *message
= "FIRSTMOUNT=Done";
625 char *envp
[] = { message
, NULL
};
626 struct lm_lockstruct
*ls
= &sdp
->sd_lockstruct
;
627 ls
->ls_first_done
= 1;
628 kobject_uevent_env(&sdp
->sd_kobj
, KOBJ_CHANGE
, envp
);
632 * gfs2_jindex_hold - Grab a lock on the jindex
633 * @sdp: The GFS2 superblock
634 * @ji_gh: the holder for the jindex glock
639 static int gfs2_jindex_hold(struct gfs2_sbd
*sdp
, struct gfs2_holder
*ji_gh
)
641 struct gfs2_inode
*dip
= GFS2_I(sdp
->sd_jindex
);
644 struct gfs2_jdesc
*jd
;
649 mutex_lock(&sdp
->sd_jindex_mutex
);
652 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, ji_gh
);
656 name
.len
= sprintf(buf
, "journal%u", sdp
->sd_journals
);
657 name
.hash
= gfs2_disk_hash(name
.name
, name
.len
);
659 error
= gfs2_dir_check(sdp
->sd_jindex
, &name
, NULL
);
660 if (error
== -ENOENT
) {
665 gfs2_glock_dq_uninit(ji_gh
);
671 jd
= kzalloc(sizeof(struct gfs2_jdesc
), GFP_KERNEL
);
675 INIT_LIST_HEAD(&jd
->extent_list
);
676 slow_work_init(&jd
->jd_work
, &gfs2_recover_ops
);
677 jd
->jd_inode
= gfs2_lookupi(sdp
->sd_jindex
, &name
, 1);
678 if (!jd
->jd_inode
|| IS_ERR(jd
->jd_inode
)) {
682 error
= PTR_ERR(jd
->jd_inode
);
687 spin_lock(&sdp
->sd_jindex_spin
);
688 jd
->jd_jid
= sdp
->sd_journals
++;
689 list_add_tail(&jd
->jd_list
, &sdp
->sd_jindex_list
);
690 spin_unlock(&sdp
->sd_jindex_spin
);
693 mutex_unlock(&sdp
->sd_jindex_mutex
);
698 static int init_journal(struct gfs2_sbd
*sdp
, int undo
)
700 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
701 struct gfs2_holder ji_gh
;
702 struct gfs2_inode
*ip
;
711 sdp
->sd_jindex
= gfs2_lookup_simple(master
, "jindex");
712 if (IS_ERR(sdp
->sd_jindex
)) {
713 fs_err(sdp
, "can't lookup journal index: %d\n", error
);
714 return PTR_ERR(sdp
->sd_jindex
);
716 ip
= GFS2_I(sdp
->sd_jindex
);
718 /* Load in the journal index special file */
720 error
= gfs2_jindex_hold(sdp
, &ji_gh
);
722 fs_err(sdp
, "can't read journal index: %d\n", error
);
727 if (!gfs2_jindex_size(sdp
)) {
728 fs_err(sdp
, "no journals!\n");
732 if (sdp
->sd_args
.ar_spectator
) {
733 sdp
->sd_jdesc
= gfs2_jdesc_find(sdp
, 0);
734 atomic_set(&sdp
->sd_log_blks_free
, sdp
->sd_jdesc
->jd_blocks
);
736 if (sdp
->sd_lockstruct
.ls_jid
>= gfs2_jindex_size(sdp
)) {
737 fs_err(sdp
, "can't mount journal #%u\n",
738 sdp
->sd_lockstruct
.ls_jid
);
739 fs_err(sdp
, "there are only %u journals (0 - %u)\n",
740 gfs2_jindex_size(sdp
),
741 gfs2_jindex_size(sdp
) - 1);
744 sdp
->sd_jdesc
= gfs2_jdesc_find(sdp
, sdp
->sd_lockstruct
.ls_jid
);
746 error
= gfs2_glock_nq_num(sdp
, sdp
->sd_lockstruct
.ls_jid
,
748 LM_ST_EXCLUSIVE
, LM_FLAG_NOEXP
,
749 &sdp
->sd_journal_gh
);
751 fs_err(sdp
, "can't acquire journal glock: %d\n", error
);
755 ip
= GFS2_I(sdp
->sd_jdesc
->jd_inode
);
756 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
,
757 LM_FLAG_NOEXP
| GL_EXACT
| GL_NOCACHE
,
760 fs_err(sdp
, "can't acquire journal inode glock: %d\n",
762 goto fail_journal_gh
;
765 error
= gfs2_jdesc_check(sdp
->sd_jdesc
);
767 fs_err(sdp
, "my journal (%u) is bad: %d\n",
768 sdp
->sd_jdesc
->jd_jid
, error
);
771 atomic_set(&sdp
->sd_log_blks_free
, sdp
->sd_jdesc
->jd_blocks
);
773 /* Map the extents for this journal's blocks */
774 map_journal_extents(sdp
);
776 trace_gfs2_log_blocks(sdp
, atomic_read(&sdp
->sd_log_blks_free
));
778 if (sdp
->sd_lockstruct
.ls_first
) {
780 for (x
= 0; x
< sdp
->sd_journals
; x
++) {
781 error
= gfs2_recover_journal(gfs2_jdesc_find(sdp
, x
));
783 fs_err(sdp
, "error recovering journal %u: %d\n",
789 gfs2_others_may_mount(sdp
);
790 } else if (!sdp
->sd_args
.ar_spectator
) {
791 error
= gfs2_recover_journal(sdp
->sd_jdesc
);
793 fs_err(sdp
, "error recovering my journal: %d\n", error
);
798 set_bit(SDF_JOURNAL_CHECKED
, &sdp
->sd_flags
);
799 gfs2_glock_dq_uninit(&ji_gh
);
805 if (!sdp
->sd_args
.ar_spectator
)
806 gfs2_glock_dq_uninit(&sdp
->sd_jinode_gh
);
808 if (!sdp
->sd_args
.ar_spectator
)
809 gfs2_glock_dq_uninit(&sdp
->sd_journal_gh
);
811 gfs2_jindex_free(sdp
);
813 gfs2_glock_dq_uninit(&ji_gh
);
815 iput(sdp
->sd_jindex
);
820 static int init_inodes(struct gfs2_sbd
*sdp
, int undo
)
823 struct gfs2_inode
*ip
;
824 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
829 error
= init_journal(sdp
, undo
);
833 /* Read in the master statfs inode */
834 sdp
->sd_statfs_inode
= gfs2_lookup_simple(master
, "statfs");
835 if (IS_ERR(sdp
->sd_statfs_inode
)) {
836 error
= PTR_ERR(sdp
->sd_statfs_inode
);
837 fs_err(sdp
, "can't read in statfs inode: %d\n", error
);
841 /* Read in the resource index inode */
842 sdp
->sd_rindex
= gfs2_lookup_simple(master
, "rindex");
843 if (IS_ERR(sdp
->sd_rindex
)) {
844 error
= PTR_ERR(sdp
->sd_rindex
);
845 fs_err(sdp
, "can't get resource index inode: %d\n", error
);
848 ip
= GFS2_I(sdp
->sd_rindex
);
849 sdp
->sd_rindex_uptodate
= 0;
851 /* Read in the quota inode */
852 sdp
->sd_quota_inode
= gfs2_lookup_simple(master
, "quota");
853 if (IS_ERR(sdp
->sd_quota_inode
)) {
854 error
= PTR_ERR(sdp
->sd_quota_inode
);
855 fs_err(sdp
, "can't get quota file inode: %d\n", error
);
861 iput(sdp
->sd_quota_inode
);
863 gfs2_clear_rgrpd(sdp
);
864 iput(sdp
->sd_rindex
);
866 iput(sdp
->sd_statfs_inode
);
868 init_journal(sdp
, UNDO
);
873 static int init_per_node(struct gfs2_sbd
*sdp
, int undo
)
875 struct inode
*pn
= NULL
;
878 struct gfs2_inode
*ip
;
879 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
881 if (sdp
->sd_args
.ar_spectator
)
887 pn
= gfs2_lookup_simple(master
, "per_node");
890 fs_err(sdp
, "can't find per_node directory: %d\n", error
);
894 sprintf(buf
, "statfs_change%u", sdp
->sd_jdesc
->jd_jid
);
895 sdp
->sd_sc_inode
= gfs2_lookup_simple(pn
, buf
);
896 if (IS_ERR(sdp
->sd_sc_inode
)) {
897 error
= PTR_ERR(sdp
->sd_sc_inode
);
898 fs_err(sdp
, "can't find local \"sc\" file: %d\n", error
);
902 sprintf(buf
, "quota_change%u", sdp
->sd_jdesc
->jd_jid
);
903 sdp
->sd_qc_inode
= gfs2_lookup_simple(pn
, buf
);
904 if (IS_ERR(sdp
->sd_qc_inode
)) {
905 error
= PTR_ERR(sdp
->sd_qc_inode
);
906 fs_err(sdp
, "can't find local \"qc\" file: %d\n", error
);
913 ip
= GFS2_I(sdp
->sd_sc_inode
);
914 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0,
917 fs_err(sdp
, "can't lock local \"sc\" file: %d\n", error
);
921 ip
= GFS2_I(sdp
->sd_qc_inode
);
922 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0,
925 fs_err(sdp
, "can't lock local \"qc\" file: %d\n", error
);
932 gfs2_glock_dq_uninit(&sdp
->sd_qc_gh
);
934 gfs2_glock_dq_uninit(&sdp
->sd_sc_gh
);
936 iput(sdp
->sd_qc_inode
);
938 iput(sdp
->sd_sc_inode
);
945 static int init_threads(struct gfs2_sbd
*sdp
, int undo
)
947 struct task_struct
*p
;
953 sdp
->sd_log_flush_time
= jiffies
;
955 p
= kthread_run(gfs2_logd
, sdp
, "gfs2_logd");
958 fs_err(sdp
, "can't start logd thread: %d\n", error
);
961 sdp
->sd_logd_process
= p
;
963 p
= kthread_run(gfs2_quotad
, sdp
, "gfs2_quotad");
966 fs_err(sdp
, "can't start quotad thread: %d\n", error
);
969 sdp
->sd_quotad_process
= p
;
975 kthread_stop(sdp
->sd_quotad_process
);
977 kthread_stop(sdp
->sd_logd_process
);
981 static const match_table_t nolock_tokens
= {
982 { Opt_jid
, "jid=%d\n", },
986 static const struct lm_lockops nolock_ops
= {
987 .lm_proto_name
= "lock_nolock",
988 .lm_put_lock
= kmem_cache_free
,
989 .lm_tokens
= &nolock_tokens
,
993 * gfs2_lm_mount - mount a locking protocol
994 * @sdp: the filesystem
995 * @args: mount arguements
996 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1001 static int gfs2_lm_mount(struct gfs2_sbd
*sdp
, int silent
)
1003 const struct lm_lockops
*lm
;
1004 struct lm_lockstruct
*ls
= &sdp
->sd_lockstruct
;
1005 struct gfs2_args
*args
= &sdp
->sd_args
;
1006 const char *proto
= sdp
->sd_proto_name
;
1007 const char *table
= sdp
->sd_table_name
;
1012 if (!strcmp("lock_nolock", proto
)) {
1014 sdp
->sd_args
.ar_localflocks
= 1;
1015 sdp
->sd_args
.ar_localcaching
= 1;
1016 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1017 } else if (!strcmp("lock_dlm", proto
)) {
1021 printk(KERN_INFO
"GFS2: can't find protocol %s\n", proto
);
1025 fs_info(sdp
, "Trying to join cluster \"%s\", \"%s\"\n", proto
, table
);
1030 for (options
= args
->ar_hostdata
; (o
= strsep(&options
, ":")); ) {
1031 substring_t tmp
[MAX_OPT_ARGS
];
1037 token
= match_token(o
, *lm
->lm_tokens
, tmp
);
1040 ret
= match_int(&tmp
[0], &option
);
1041 if (ret
|| option
< 0)
1042 goto hostdata_error
;
1043 ls
->ls_jid
= option
;
1046 /* Obsolete, but left for backward compat purposes */
1049 ret
= match_int(&tmp
[0], &option
);
1050 if (ret
|| (option
!= 0 && option
!= 1))
1051 goto hostdata_error
;
1052 ls
->ls_first
= option
;
1055 ret
= match_int(&tmp
[0], &option
);
1056 if (ret
|| (option
!= 0 && option
!= 1))
1057 goto hostdata_error
;
1058 ls
->ls_nodir
= option
;
1063 fs_info(sdp
, "unknown hostdata (%s)\n", o
);
1068 if (sdp
->sd_args
.ar_spectator
)
1069 snprintf(sdp
->sd_fsname
, GFS2_FSNAME_LEN
, "%s.s", table
);
1071 snprintf(sdp
->sd_fsname
, GFS2_FSNAME_LEN
, "%s.%u", table
,
1072 sdp
->sd_lockstruct
.ls_jid
);
1074 fsname
= strchr(table
, ':');
1077 if (lm
->lm_mount
== NULL
) {
1078 fs_info(sdp
, "Now mounting FS...\n");
1081 ret
= lm
->lm_mount(sdp
, fsname
);
1083 fs_info(sdp
, "Joined cluster. Now mounting FS...\n");
1087 void gfs2_lm_unmount(struct gfs2_sbd
*sdp
)
1089 const struct lm_lockops
*lm
= sdp
->sd_lockstruct
.ls_ops
;
1090 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)) &&
1092 lm
->lm_unmount(sdp
);
1095 void gfs2_online_uevent(struct gfs2_sbd
*sdp
)
1097 struct super_block
*sb
= sdp
->sd_vfs
;
1100 char *envp
[] = { ro
, spectator
, NULL
};
1101 sprintf(ro
, "RDONLY=%d", (sb
->s_flags
& MS_RDONLY
) ? 1 : 0);
1102 sprintf(spectator
, "SPECTATOR=%d", sdp
->sd_args
.ar_spectator
? 1 : 0);
1103 kobject_uevent_env(&sdp
->sd_kobj
, KOBJ_ONLINE
, envp
);
1107 * fill_super - Read in superblock
1108 * @sb: The VFS superblock
1109 * @data: Mount options
1110 * @silent: Don't complain if it's not a GFS2 filesystem
1115 static int fill_super(struct super_block
*sb
, struct gfs2_args
*args
, int silent
)
1117 struct gfs2_sbd
*sdp
;
1118 struct gfs2_holder mount_gh
;
1123 printk(KERN_WARNING
"GFS2: can't alloc struct gfs2_sbd\n");
1126 sdp
->sd_args
= *args
;
1128 if (sdp
->sd_args
.ar_spectator
) {
1129 sb
->s_flags
|= MS_RDONLY
;
1130 set_bit(SDF_NORECOVERY
, &sdp
->sd_flags
);
1132 if (sdp
->sd_args
.ar_posix_acl
)
1133 sb
->s_flags
|= MS_POSIXACL
;
1134 if (sdp
->sd_args
.ar_nobarrier
)
1135 set_bit(SDF_NOBARRIERS
, &sdp
->sd_flags
);
1137 sb
->s_magic
= GFS2_MAGIC
;
1138 sb
->s_op
= &gfs2_super_ops
;
1139 sb
->s_export_op
= &gfs2_export_ops
;
1140 sb
->s_xattr
= gfs2_xattr_handlers
;
1141 sb
->s_qcop
= &gfs2_quotactl_ops
;
1142 sb_dqopt(sb
)->flags
|= DQUOT_QUOTA_SYS_FILE
;
1143 sb
->s_time_gran
= 1;
1144 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
1146 /* Set up the buffer cache and fill in some fake block size values
1147 to allow us to read-in the on-disk superblock. */
1148 sdp
->sd_sb
.sb_bsize
= sb_min_blocksize(sb
, GFS2_BASIC_BLOCK
);
1149 sdp
->sd_sb
.sb_bsize_shift
= sb
->s_blocksize_bits
;
1150 sdp
->sd_fsb2bb_shift
= sdp
->sd_sb
.sb_bsize_shift
-
1151 GFS2_BASIC_BLOCK_SHIFT
;
1152 sdp
->sd_fsb2bb
= 1 << sdp
->sd_fsb2bb_shift
;
1154 sdp
->sd_tune
.gt_log_flush_secs
= sdp
->sd_args
.ar_commit
;
1155 sdp
->sd_tune
.gt_quota_quantum
= sdp
->sd_args
.ar_quota_quantum
;
1156 if (sdp
->sd_args
.ar_statfs_quantum
) {
1157 sdp
->sd_tune
.gt_statfs_slow
= 0;
1158 sdp
->sd_tune
.gt_statfs_quantum
= sdp
->sd_args
.ar_statfs_quantum
;
1161 sdp
->sd_tune
.gt_statfs_slow
= 1;
1162 sdp
->sd_tune
.gt_statfs_quantum
= 30;
1165 error
= init_names(sdp
, silent
);
1169 gfs2_create_debugfs_file(sdp
);
1171 error
= gfs2_sys_fs_add(sdp
);
1175 error
= gfs2_lm_mount(sdp
, silent
);
1179 error
= init_locking(sdp
, &mount_gh
, DO
);
1183 error
= init_sb(sdp
, silent
);
1187 error
= init_inodes(sdp
, DO
);
1191 error
= init_per_node(sdp
, DO
);
1195 error
= gfs2_statfs_init(sdp
);
1197 fs_err(sdp
, "can't initialize statfs subsystem: %d\n", error
);
1201 error
= init_threads(sdp
, DO
);
1205 if (!(sb
->s_flags
& MS_RDONLY
)) {
1206 error
= gfs2_make_fs_rw(sdp
);
1208 fs_err(sdp
, "can't make FS RW: %d\n", error
);
1213 gfs2_glock_dq_uninit(&mount_gh
);
1214 gfs2_online_uevent(sdp
);
1218 init_threads(sdp
, UNDO
);
1220 init_per_node(sdp
, UNDO
);
1222 init_inodes(sdp
, UNDO
);
1224 if (sdp
->sd_root_dir
)
1225 dput(sdp
->sd_root_dir
);
1226 if (sdp
->sd_master_dir
)
1227 dput(sdp
->sd_master_dir
);
1232 init_locking(sdp
, &mount_gh
, UNDO
);
1234 gfs2_gl_hash_clear(sdp
);
1235 gfs2_lm_unmount(sdp
);
1236 while (invalidate_inodes(sb
))
1239 gfs2_sys_fs_del(sdp
);
1241 gfs2_delete_debugfs_file(sdp
);
1243 sb
->s_fs_info
= NULL
;
1247 static int set_gfs2_super(struct super_block
*s
, void *data
)
1250 s
->s_dev
= s
->s_bdev
->bd_dev
;
1253 * We set the bdi here to the queue backing, file systems can
1254 * overwrite this in ->fill_super()
1256 s
->s_bdi
= &bdev_get_queue(s
->s_bdev
)->backing_dev_info
;
1260 static int test_gfs2_super(struct super_block
*s
, void *ptr
)
1262 struct block_device
*bdev
= ptr
;
1263 return (bdev
== s
->s_bdev
);
1267 * gfs2_get_sb - Get the GFS2 superblock
1268 * @fs_type: The GFS2 filesystem type
1269 * @flags: Mount flags
1270 * @dev_name: The name of the device
1271 * @data: The mount arguments
1272 * @mnt: The vfsmnt for this mount
1274 * Q. Why not use get_sb_bdev() ?
1275 * A. We need to select one of two root directories to mount, independent
1276 * of whether this is the initial, or subsequent, mount of this sb
1278 * Returns: 0 or -ve on error
1281 static int gfs2_get_sb(struct file_system_type
*fs_type
, int flags
,
1282 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1284 struct block_device
*bdev
;
1285 struct super_block
*s
;
1286 fmode_t mode
= FMODE_READ
;
1288 struct gfs2_args args
;
1289 struct gfs2_sbd
*sdp
;
1291 if (!(flags
& MS_RDONLY
))
1292 mode
|= FMODE_WRITE
;
1294 bdev
= open_bdev_exclusive(dev_name
, mode
, fs_type
);
1296 return PTR_ERR(bdev
);
1299 * once the super is inserted into the list by sget, s_umount
1300 * will protect the lockfs code from trying to start a snapshot
1301 * while we are mounting
1303 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
1304 if (bdev
->bd_fsfreeze_count
> 0) {
1305 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
1309 s
= sget(fs_type
, test_gfs2_super
, set_gfs2_super
, bdev
);
1310 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
1315 memset(&args
, 0, sizeof(args
));
1316 args
.ar_quota
= GFS2_QUOTA_DEFAULT
;
1317 args
.ar_data
= GFS2_DATA_DEFAULT
;
1318 args
.ar_commit
= 60;
1319 args
.ar_statfs_quantum
= 30;
1320 args
.ar_quota_quantum
= 60;
1321 args
.ar_errors
= GFS2_ERRORS_DEFAULT
;
1323 error
= gfs2_mount_args(&args
, data
);
1325 printk(KERN_WARNING
"GFS2: can't parse mount arguments\n");
1328 deactivate_locked_super(s
);
1334 if ((flags
^ s
->s_flags
) & MS_RDONLY
)
1336 close_bdev_exclusive(bdev
, mode
);
1338 char b
[BDEVNAME_SIZE
];
1342 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1343 sb_set_blocksize(s
, block_size(bdev
));
1344 error
= fill_super(s
, &args
, flags
& MS_SILENT
? 1 : 0);
1346 deactivate_locked_super(s
);
1349 s
->s_flags
|= MS_ACTIVE
;
1356 mnt
->mnt_root
= dget(sdp
->sd_master_dir
);
1358 mnt
->mnt_root
= dget(sdp
->sd_root_dir
);
1362 deactivate_locked_super(s
);
1364 close_bdev_exclusive(bdev
, mode
);
1368 static int set_meta_super(struct super_block
*s
, void *ptr
)
1373 static int gfs2_get_sb_meta(struct file_system_type
*fs_type
, int flags
,
1374 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1376 struct super_block
*s
;
1377 struct gfs2_sbd
*sdp
;
1381 error
= kern_path(dev_name
, LOOKUP_FOLLOW
, &path
);
1383 printk(KERN_WARNING
"GFS2: path_lookup on %s returned error %d\n",
1387 s
= sget(&gfs2_fs_type
, test_gfs2_super
, set_meta_super
,
1388 path
.dentry
->d_inode
->i_sb
->s_bdev
);
1391 printk(KERN_WARNING
"GFS2: gfs2 mount does not exist\n");
1394 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
1395 deactivate_locked_super(s
);
1400 mnt
->mnt_root
= dget(sdp
->sd_master_dir
);
1404 static void gfs2_kill_sb(struct super_block
*sb
)
1406 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1409 kill_block_super(sb
);
1413 gfs2_meta_syncfs(sdp
);
1414 dput(sdp
->sd_root_dir
);
1415 dput(sdp
->sd_master_dir
);
1416 sdp
->sd_root_dir
= NULL
;
1417 sdp
->sd_master_dir
= NULL
;
1418 shrink_dcache_sb(sb
);
1419 kill_block_super(sb
);
1420 gfs2_delete_debugfs_file(sdp
);
1424 struct file_system_type gfs2_fs_type
= {
1426 .fs_flags
= FS_REQUIRES_DEV
,
1427 .get_sb
= gfs2_get_sb
,
1428 .kill_sb
= gfs2_kill_sb
,
1429 .owner
= THIS_MODULE
,
1432 struct file_system_type gfs2meta_fs_type
= {
1434 .fs_flags
= FS_REQUIRES_DEV
,
1435 .get_sb
= gfs2_get_sb_meta
,
1436 .owner
= THIS_MODULE
,