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/lm_interface.h>
30 #include "ops_fstype.h"
31 #include "ops_dentry.h"
32 #include "ops_super.h"
43 static const u32 gfs2_old_fs_formats
[] = {
47 static const u32 gfs2_old_multihost_formats
[] = {
52 * gfs2_tune_init - Fill a gfs2_tune structure with default values
57 static void gfs2_tune_init(struct gfs2_tune
*gt
)
59 spin_lock_init(>
->gt_spin
);
61 gt
->gt_demote_secs
= 300;
62 gt
->gt_incore_log_blocks
= 1024;
63 gt
->gt_log_flush_secs
= 60;
64 gt
->gt_recoverd_secs
= 60;
66 gt
->gt_quotad_secs
= 5;
67 gt
->gt_quota_simul_sync
= 64;
68 gt
->gt_quota_warn_period
= 10;
69 gt
->gt_quota_scale_num
= 1;
70 gt
->gt_quota_scale_den
= 1;
71 gt
->gt_quota_cache_secs
= 300;
72 gt
->gt_quota_quantum
= 60;
73 gt
->gt_atime_quantum
= 3600;
74 gt
->gt_new_files_jdata
= 0;
75 gt
->gt_max_readahead
= 1 << 18;
76 gt
->gt_stall_secs
= 600;
77 gt
->gt_complain_secs
= 10;
78 gt
->gt_statfs_quantum
= 30;
79 gt
->gt_statfs_slow
= 0;
82 static struct gfs2_sbd
*init_sbd(struct super_block
*sb
)
86 sdp
= kzalloc(sizeof(struct gfs2_sbd
), GFP_KERNEL
);
93 gfs2_tune_init(&sdp
->sd_tune
);
95 INIT_LIST_HEAD(&sdp
->sd_reclaim_list
);
96 spin_lock_init(&sdp
->sd_reclaim_lock
);
97 init_waitqueue_head(&sdp
->sd_reclaim_wq
);
99 mutex_init(&sdp
->sd_inum_mutex
);
100 spin_lock_init(&sdp
->sd_statfs_spin
);
102 spin_lock_init(&sdp
->sd_rindex_spin
);
103 mutex_init(&sdp
->sd_rindex_mutex
);
104 INIT_LIST_HEAD(&sdp
->sd_rindex_list
);
105 INIT_LIST_HEAD(&sdp
->sd_rindex_mru_list
);
107 INIT_LIST_HEAD(&sdp
->sd_jindex_list
);
108 spin_lock_init(&sdp
->sd_jindex_spin
);
109 mutex_init(&sdp
->sd_jindex_mutex
);
111 INIT_LIST_HEAD(&sdp
->sd_quota_list
);
112 spin_lock_init(&sdp
->sd_quota_spin
);
113 mutex_init(&sdp
->sd_quota_mutex
);
115 spin_lock_init(&sdp
->sd_log_lock
);
117 INIT_LIST_HEAD(&sdp
->sd_log_le_buf
);
118 INIT_LIST_HEAD(&sdp
->sd_log_le_revoke
);
119 INIT_LIST_HEAD(&sdp
->sd_log_le_rg
);
120 INIT_LIST_HEAD(&sdp
->sd_log_le_databuf
);
121 INIT_LIST_HEAD(&sdp
->sd_log_le_ordered
);
123 mutex_init(&sdp
->sd_log_reserve_mutex
);
124 INIT_LIST_HEAD(&sdp
->sd_ail1_list
);
125 INIT_LIST_HEAD(&sdp
->sd_ail2_list
);
127 init_rwsem(&sdp
->sd_log_flush_lock
);
128 atomic_set(&sdp
->sd_log_in_flight
, 0);
129 init_waitqueue_head(&sdp
->sd_log_flush_wait
);
131 INIT_LIST_HEAD(&sdp
->sd_revoke_list
);
133 mutex_init(&sdp
->sd_freeze_lock
);
138 static void init_vfs(struct super_block
*sb
, unsigned noatime
)
140 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
142 sb
->s_magic
= GFS2_MAGIC
;
143 sb
->s_op
= &gfs2_super_ops
;
144 sb
->s_export_op
= &gfs2_export_ops
;
146 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
148 if (sb
->s_flags
& (MS_NOATIME
| MS_NODIRATIME
))
149 set_bit(noatime
, &sdp
->sd_flags
);
151 /* Don't let the VFS update atimes. GFS2 handles this itself. */
152 sb
->s_flags
|= MS_NOATIME
| MS_NODIRATIME
;
156 * gfs2_check_sb - Check superblock
157 * @sdp: the filesystem
158 * @sb: The superblock
159 * @silent: Don't print a message if the check fails
161 * Checks the version code of the FS is one that we understand how to
162 * read and that the sizes of the various on-disk structures have not
166 static int gfs2_check_sb(struct gfs2_sbd
*sdp
, struct gfs2_sb_host
*sb
, int silent
)
170 if (sb
->sb_magic
!= GFS2_MAGIC
||
171 sb
->sb_type
!= GFS2_METATYPE_SB
) {
173 printk(KERN_WARNING
"GFS2: not a GFS2 filesystem\n");
177 /* If format numbers match exactly, we're done. */
179 if (sb
->sb_fs_format
== GFS2_FORMAT_FS
&&
180 sb
->sb_multihost_format
== GFS2_FORMAT_MULTI
)
183 if (sb
->sb_fs_format
!= GFS2_FORMAT_FS
) {
184 for (x
= 0; gfs2_old_fs_formats
[x
]; x
++)
185 if (gfs2_old_fs_formats
[x
] == sb
->sb_fs_format
)
188 if (!gfs2_old_fs_formats
[x
]) {
190 "GFS2: code version (%u, %u) is incompatible "
191 "with ondisk format (%u, %u)\n",
192 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
193 sb
->sb_fs_format
, sb
->sb_multihost_format
);
195 "GFS2: I don't know how to upgrade this FS\n");
200 if (sb
->sb_multihost_format
!= GFS2_FORMAT_MULTI
) {
201 for (x
= 0; gfs2_old_multihost_formats
[x
]; x
++)
202 if (gfs2_old_multihost_formats
[x
] ==
203 sb
->sb_multihost_format
)
206 if (!gfs2_old_multihost_formats
[x
]) {
208 "GFS2: code version (%u, %u) is incompatible "
209 "with ondisk format (%u, %u)\n",
210 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
211 sb
->sb_fs_format
, sb
->sb_multihost_format
);
213 "GFS2: I don't know how to upgrade this FS\n");
218 if (!sdp
->sd_args
.ar_upgrade
) {
220 "GFS2: code version (%u, %u) is incompatible "
221 "with ondisk format (%u, %u)\n",
222 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
223 sb
->sb_fs_format
, sb
->sb_multihost_format
);
225 "GFS2: Use the \"upgrade\" mount option to upgrade "
227 printk(KERN_INFO
"GFS2: See the manual for more details\n");
234 static void end_bio_io_page(struct bio
*bio
, int error
)
236 struct page
*page
= bio
->bi_private
;
239 SetPageUptodate(page
);
241 printk(KERN_WARNING
"gfs2: error %d reading superblock\n", error
);
245 static void gfs2_sb_in(struct gfs2_sb_host
*sb
, const void *buf
)
247 const struct gfs2_sb
*str
= buf
;
249 sb
->sb_magic
= be32_to_cpu(str
->sb_header
.mh_magic
);
250 sb
->sb_type
= be32_to_cpu(str
->sb_header
.mh_type
);
251 sb
->sb_format
= be32_to_cpu(str
->sb_header
.mh_format
);
252 sb
->sb_fs_format
= be32_to_cpu(str
->sb_fs_format
);
253 sb
->sb_multihost_format
= be32_to_cpu(str
->sb_multihost_format
);
254 sb
->sb_bsize
= be32_to_cpu(str
->sb_bsize
);
255 sb
->sb_bsize_shift
= be32_to_cpu(str
->sb_bsize_shift
);
256 sb
->sb_master_dir
.no_addr
= be64_to_cpu(str
->sb_master_dir
.no_addr
);
257 sb
->sb_master_dir
.no_formal_ino
= be64_to_cpu(str
->sb_master_dir
.no_formal_ino
);
258 sb
->sb_root_dir
.no_addr
= be64_to_cpu(str
->sb_root_dir
.no_addr
);
259 sb
->sb_root_dir
.no_formal_ino
= be64_to_cpu(str
->sb_root_dir
.no_formal_ino
);
261 memcpy(sb
->sb_lockproto
, str
->sb_lockproto
, GFS2_LOCKNAME_LEN
);
262 memcpy(sb
->sb_locktable
, str
->sb_locktable
, GFS2_LOCKNAME_LEN
);
266 * gfs2_read_super - Read the gfs2 super block from disk
267 * @sdp: The GFS2 super block
268 * @sector: The location of the super block
269 * @error: The error code to return
271 * This uses the bio functions to read the super block from disk
272 * because we want to be 100% sure that we never read cached data.
273 * A super block is read twice only during each GFS2 mount and is
274 * never written to by the filesystem. The first time its read no
275 * locks are held, and the only details which are looked at are those
276 * relating to the locking protocol. Once locking is up and working,
277 * the sb is read again under the lock to establish the location of
278 * the master directory (contains pointers to journals etc) and the
281 * Returns: 0 on success or error
284 static int gfs2_read_super(struct gfs2_sbd
*sdp
, sector_t sector
)
286 struct super_block
*sb
= sdp
->sd_vfs
;
291 page
= alloc_page(GFP_NOFS
);
295 ClearPageUptodate(page
);
296 ClearPageDirty(page
);
299 bio
= bio_alloc(GFP_NOFS
, 1);
300 if (unlikely(!bio
)) {
305 bio
->bi_sector
= sector
* (sb
->s_blocksize
>> 9);
306 bio
->bi_bdev
= sb
->s_bdev
;
307 bio_add_page(bio
, page
, PAGE_SIZE
, 0);
309 bio
->bi_end_io
= end_bio_io_page
;
310 bio
->bi_private
= page
;
311 submit_bio(READ_SYNC
| (1 << BIO_RW_META
), bio
);
312 wait_on_page_locked(page
);
314 if (!PageUptodate(page
)) {
319 gfs2_sb_in(&sdp
->sd_sb
, p
);
325 * gfs2_read_sb - Read super block
326 * @sdp: The GFS2 superblock
327 * @gl: the glock for the superblock (assumed to be held)
328 * @silent: Don't print message if mount fails
332 static int gfs2_read_sb(struct gfs2_sbd
*sdp
, struct gfs2_glock
*gl
, int silent
)
334 u32 hash_blocks
, ind_blocks
, leaf_blocks
;
339 error
= gfs2_read_super(sdp
, GFS2_SB_ADDR
>> sdp
->sd_fsb2bb_shift
);
342 fs_err(sdp
, "can't read superblock\n");
346 error
= gfs2_check_sb(sdp
, &sdp
->sd_sb
, silent
);
350 sdp
->sd_fsb2bb_shift
= sdp
->sd_sb
.sb_bsize_shift
-
351 GFS2_BASIC_BLOCK_SHIFT
;
352 sdp
->sd_fsb2bb
= 1 << sdp
->sd_fsb2bb_shift
;
353 sdp
->sd_diptrs
= (sdp
->sd_sb
.sb_bsize
-
354 sizeof(struct gfs2_dinode
)) / sizeof(u64
);
355 sdp
->sd_inptrs
= (sdp
->sd_sb
.sb_bsize
-
356 sizeof(struct gfs2_meta_header
)) / sizeof(u64
);
357 sdp
->sd_jbsize
= sdp
->sd_sb
.sb_bsize
- sizeof(struct gfs2_meta_header
);
358 sdp
->sd_hash_bsize
= sdp
->sd_sb
.sb_bsize
/ 2;
359 sdp
->sd_hash_bsize_shift
= sdp
->sd_sb
.sb_bsize_shift
- 1;
360 sdp
->sd_hash_ptrs
= sdp
->sd_hash_bsize
/ sizeof(u64
);
361 sdp
->sd_qc_per_block
= (sdp
->sd_sb
.sb_bsize
-
362 sizeof(struct gfs2_meta_header
)) /
363 sizeof(struct gfs2_quota_change
);
365 /* Compute maximum reservation required to add a entry to a directory */
367 hash_blocks
= DIV_ROUND_UP(sizeof(u64
) * (1 << GFS2_DIR_MAX_DEPTH
),
371 for (tmp_blocks
= hash_blocks
; tmp_blocks
> sdp
->sd_diptrs
;) {
372 tmp_blocks
= DIV_ROUND_UP(tmp_blocks
, sdp
->sd_inptrs
);
373 ind_blocks
+= tmp_blocks
;
376 leaf_blocks
= 2 + GFS2_DIR_MAX_DEPTH
;
378 sdp
->sd_max_dirres
= hash_blocks
+ ind_blocks
+ leaf_blocks
;
380 sdp
->sd_heightsize
[0] = sdp
->sd_sb
.sb_bsize
-
381 sizeof(struct gfs2_dinode
);
382 sdp
->sd_heightsize
[1] = sdp
->sd_sb
.sb_bsize
* sdp
->sd_diptrs
;
387 space
= sdp
->sd_heightsize
[x
- 1] * sdp
->sd_inptrs
;
389 m
= do_div(d
, sdp
->sd_inptrs
);
391 if (d
!= sdp
->sd_heightsize
[x
- 1] || m
)
393 sdp
->sd_heightsize
[x
] = space
;
395 sdp
->sd_max_height
= x
;
396 sdp
->sd_heightsize
[x
] = ~0;
397 gfs2_assert(sdp
, sdp
->sd_max_height
<= GFS2_MAX_META_HEIGHT
);
399 sdp
->sd_jheightsize
[0] = sdp
->sd_sb
.sb_bsize
-
400 sizeof(struct gfs2_dinode
);
401 sdp
->sd_jheightsize
[1] = sdp
->sd_jbsize
* sdp
->sd_diptrs
;
406 space
= sdp
->sd_jheightsize
[x
- 1] * sdp
->sd_inptrs
;
408 m
= do_div(d
, sdp
->sd_inptrs
);
410 if (d
!= sdp
->sd_jheightsize
[x
- 1] || m
)
412 sdp
->sd_jheightsize
[x
] = space
;
414 sdp
->sd_max_jheight
= x
;
415 sdp
->sd_jheightsize
[x
] = ~0;
416 gfs2_assert(sdp
, sdp
->sd_max_jheight
<= GFS2_MAX_META_HEIGHT
);
421 static int init_names(struct gfs2_sbd
*sdp
, int silent
)
426 proto
= sdp
->sd_args
.ar_lockproto
;
427 table
= sdp
->sd_args
.ar_locktable
;
429 /* Try to autodetect */
431 if (!proto
[0] || !table
[0]) {
432 error
= gfs2_read_super(sdp
, GFS2_SB_ADDR
>> sdp
->sd_fsb2bb_shift
);
436 error
= gfs2_check_sb(sdp
, &sdp
->sd_sb
, silent
);
441 proto
= sdp
->sd_sb
.sb_lockproto
;
443 table
= sdp
->sd_sb
.sb_locktable
;
447 table
= sdp
->sd_vfs
->s_id
;
449 strlcpy(sdp
->sd_proto_name
, proto
, GFS2_FSNAME_LEN
);
450 strlcpy(sdp
->sd_table_name
, table
, GFS2_FSNAME_LEN
);
452 table
= sdp
->sd_table_name
;
453 while ((table
= strchr(table
, '/')))
460 static int init_locking(struct gfs2_sbd
*sdp
, struct gfs2_holder
*mount_gh
,
463 struct task_struct
*p
;
469 for (sdp
->sd_glockd_num
= 0;
470 sdp
->sd_glockd_num
< sdp
->sd_args
.ar_num_glockd
;
471 sdp
->sd_glockd_num
++) {
472 p
= kthread_run(gfs2_glockd
, sdp
, "gfs2_glockd");
475 fs_err(sdp
, "can't start glockd thread: %d\n", error
);
478 sdp
->sd_glockd_process
[sdp
->sd_glockd_num
] = p
;
481 error
= gfs2_glock_nq_num(sdp
,
482 GFS2_MOUNT_LOCK
, &gfs2_nondisk_glops
,
483 LM_ST_EXCLUSIVE
, LM_FLAG_NOEXP
| GL_NOCACHE
,
486 fs_err(sdp
, "can't acquire mount glock: %d\n", error
);
490 error
= gfs2_glock_nq_num(sdp
,
491 GFS2_LIVE_LOCK
, &gfs2_nondisk_glops
,
493 LM_FLAG_NOEXP
| GL_EXACT
,
496 fs_err(sdp
, "can't acquire live glock: %d\n", error
);
500 error
= gfs2_glock_get(sdp
, GFS2_RENAME_LOCK
, &gfs2_nondisk_glops
,
501 CREATE
, &sdp
->sd_rename_gl
);
503 fs_err(sdp
, "can't create rename glock: %d\n", error
);
507 error
= gfs2_glock_get(sdp
, GFS2_TRANS_LOCK
, &gfs2_trans_glops
,
508 CREATE
, &sdp
->sd_trans_gl
);
510 fs_err(sdp
, "can't create transaction glock: %d\n", error
);
513 set_bit(GLF_STICKY
, &sdp
->sd_trans_gl
->gl_flags
);
518 gfs2_glock_put(sdp
->sd_trans_gl
);
520 gfs2_glock_put(sdp
->sd_rename_gl
);
522 gfs2_glock_dq_uninit(&sdp
->sd_live_gh
);
524 gfs2_glock_dq_uninit(mount_gh
);
526 while (sdp
->sd_glockd_num
--)
527 kthread_stop(sdp
->sd_glockd_process
[sdp
->sd_glockd_num
]);
532 static int gfs2_lookup_root(struct super_block
*sb
, struct dentry
**dptr
,
533 u64 no_addr
, const char *name
)
535 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
536 struct dentry
*dentry
;
539 inode
= gfs2_inode_lookup(sb
, DT_DIR
, no_addr
, 0, 0);
541 fs_err(sdp
, "can't read in %s inode: %ld\n", name
, PTR_ERR(inode
));
542 return PTR_ERR(inode
);
544 dentry
= d_alloc_root(inode
);
546 fs_err(sdp
, "can't alloc %s dentry\n", name
);
550 dentry
->d_op
= &gfs2_dops
;
555 static int init_sb(struct gfs2_sbd
*sdp
, int silent
)
557 struct super_block
*sb
= sdp
->sd_vfs
;
558 struct gfs2_holder sb_gh
;
562 ret
= gfs2_glock_nq_num(sdp
, GFS2_SB_LOCK
, &gfs2_meta_glops
,
563 LM_ST_SHARED
, 0, &sb_gh
);
565 fs_err(sdp
, "can't acquire superblock glock: %d\n", ret
);
569 ret
= gfs2_read_sb(sdp
, sb_gh
.gh_gl
, silent
);
571 fs_err(sdp
, "can't read superblock: %d\n", ret
);
575 /* Set up the buffer cache and SB for real */
576 if (sdp
->sd_sb
.sb_bsize
< bdev_hardsect_size(sb
->s_bdev
)) {
578 fs_err(sdp
, "FS block size (%u) is too small for device "
580 sdp
->sd_sb
.sb_bsize
, bdev_hardsect_size(sb
->s_bdev
));
583 if (sdp
->sd_sb
.sb_bsize
> PAGE_SIZE
) {
585 fs_err(sdp
, "FS block size (%u) is too big for machine "
587 sdp
->sd_sb
.sb_bsize
, (unsigned int)PAGE_SIZE
);
590 sb_set_blocksize(sb
, sdp
->sd_sb
.sb_bsize
);
592 /* Get the root inode */
593 no_addr
= sdp
->sd_sb
.sb_root_dir
.no_addr
;
594 ret
= gfs2_lookup_root(sb
, &sdp
->sd_root_dir
, no_addr
, "root");
598 /* Get the master inode */
599 no_addr
= sdp
->sd_sb
.sb_master_dir
.no_addr
;
600 ret
= gfs2_lookup_root(sb
, &sdp
->sd_master_dir
, no_addr
, "master");
602 dput(sdp
->sd_root_dir
);
605 sb
->s_root
= dget(sdp
->sd_args
.ar_meta
? sdp
->sd_master_dir
: sdp
->sd_root_dir
);
607 gfs2_glock_dq_uninit(&sb_gh
);
612 * map_journal_extents - create a reusable "extent" mapping from all logical
613 * blocks to all physical blocks for the given journal. This will save
614 * us time when writing journal blocks. Most journals will have only one
615 * extent that maps all their logical blocks. That's because gfs2.mkfs
616 * arranges the journal blocks sequentially to maximize performance.
617 * So the extent would map the first block for the entire file length.
618 * However, gfs2_jadd can happen while file activity is happening, so
619 * those journals may not be sequential. Less likely is the case where
620 * the users created their own journals by mounting the metafs and
621 * laying it out. But it's still possible. These journals might have
624 * TODO: This should be done in bigger chunks rather than one block at a time,
625 * but since it's only done at mount time, I'm not worried about the
628 static int map_journal_extents(struct gfs2_sbd
*sdp
)
630 struct gfs2_jdesc
*jd
= sdp
->sd_jdesc
;
632 u64 db
, prev_db
; /* logical block, disk block, prev disk block */
633 struct gfs2_inode
*ip
= GFS2_I(jd
->jd_inode
);
634 struct gfs2_journal_extent
*jext
= NULL
;
635 struct buffer_head bh
;
640 for (lb
= 0; lb
< ip
->i_di
.di_size
>> sdp
->sd_sb
.sb_bsize_shift
; lb
++) {
643 bh
.b_size
= 1 << ip
->i_inode
.i_blkbits
;
644 rc
= gfs2_block_map(jd
->jd_inode
, lb
, &bh
, 0);
647 printk(KERN_INFO
"GFS2 journal mapping error %d: lb="
648 "%u db=%llu\n", rc
, lb
, (unsigned long long)db
);
651 if (!prev_db
|| db
!= prev_db
+ 1) {
652 jext
= kzalloc(sizeof(struct gfs2_journal_extent
),
655 printk(KERN_INFO
"GFS2 error: out of memory "
656 "mapping journal extents.\n");
663 list_add_tail(&jext
->extent_list
, &jd
->extent_list
);
672 static void gfs2_lm_others_may_mount(struct gfs2_sbd
*sdp
)
674 if (!sdp
->sd_lockstruct
.ls_ops
->lm_others_may_mount
)
676 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
677 sdp
->sd_lockstruct
.ls_ops
->lm_others_may_mount(
678 sdp
->sd_lockstruct
.ls_lockspace
);
681 static int init_journal(struct gfs2_sbd
*sdp
, int undo
)
683 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
684 struct gfs2_holder ji_gh
;
685 struct task_struct
*p
;
686 struct gfs2_inode
*ip
;
695 sdp
->sd_jindex
= gfs2_lookup_simple(master
, "jindex");
696 if (IS_ERR(sdp
->sd_jindex
)) {
697 fs_err(sdp
, "can't lookup journal index: %d\n", error
);
698 return PTR_ERR(sdp
->sd_jindex
);
700 ip
= GFS2_I(sdp
->sd_jindex
);
701 set_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
703 /* Load in the journal index special file */
705 error
= gfs2_jindex_hold(sdp
, &ji_gh
);
707 fs_err(sdp
, "can't read journal index: %d\n", error
);
712 if (!gfs2_jindex_size(sdp
)) {
713 fs_err(sdp
, "no journals!\n");
717 if (sdp
->sd_args
.ar_spectator
) {
718 sdp
->sd_jdesc
= gfs2_jdesc_find(sdp
, 0);
719 atomic_set(&sdp
->sd_log_blks_free
, sdp
->sd_jdesc
->jd_blocks
);
721 if (sdp
->sd_lockstruct
.ls_jid
>= gfs2_jindex_size(sdp
)) {
722 fs_err(sdp
, "can't mount journal #%u\n",
723 sdp
->sd_lockstruct
.ls_jid
);
724 fs_err(sdp
, "there are only %u journals (0 - %u)\n",
725 gfs2_jindex_size(sdp
),
726 gfs2_jindex_size(sdp
) - 1);
729 sdp
->sd_jdesc
= gfs2_jdesc_find(sdp
, sdp
->sd_lockstruct
.ls_jid
);
731 error
= gfs2_glock_nq_num(sdp
, sdp
->sd_lockstruct
.ls_jid
,
733 LM_ST_EXCLUSIVE
, LM_FLAG_NOEXP
,
734 &sdp
->sd_journal_gh
);
736 fs_err(sdp
, "can't acquire journal glock: %d\n", error
);
740 ip
= GFS2_I(sdp
->sd_jdesc
->jd_inode
);
741 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
,
742 LM_FLAG_NOEXP
| GL_EXACT
| GL_NOCACHE
,
745 fs_err(sdp
, "can't acquire journal inode glock: %d\n",
747 goto fail_journal_gh
;
750 error
= gfs2_jdesc_check(sdp
->sd_jdesc
);
752 fs_err(sdp
, "my journal (%u) is bad: %d\n",
753 sdp
->sd_jdesc
->jd_jid
, error
);
756 atomic_set(&sdp
->sd_log_blks_free
, sdp
->sd_jdesc
->jd_blocks
);
758 /* Map the extents for this journal's blocks */
759 map_journal_extents(sdp
);
762 if (sdp
->sd_lockstruct
.ls_first
) {
764 for (x
= 0; x
< sdp
->sd_journals
; x
++) {
765 error
= gfs2_recover_journal(gfs2_jdesc_find(sdp
, x
));
767 fs_err(sdp
, "error recovering journal %u: %d\n",
773 gfs2_lm_others_may_mount(sdp
);
774 } else if (!sdp
->sd_args
.ar_spectator
) {
775 error
= gfs2_recover_journal(sdp
->sd_jdesc
);
777 fs_err(sdp
, "error recovering my journal: %d\n", error
);
782 set_bit(SDF_JOURNAL_CHECKED
, &sdp
->sd_flags
);
783 gfs2_glock_dq_uninit(&ji_gh
);
786 p
= kthread_run(gfs2_recoverd
, sdp
, "gfs2_recoverd");
789 fs_err(sdp
, "can't start recoverd thread: %d\n", error
);
792 sdp
->sd_recoverd_process
= p
;
797 kthread_stop(sdp
->sd_recoverd_process
);
799 if (!sdp
->sd_args
.ar_spectator
)
800 gfs2_glock_dq_uninit(&sdp
->sd_jinode_gh
);
802 if (!sdp
->sd_args
.ar_spectator
)
803 gfs2_glock_dq_uninit(&sdp
->sd_journal_gh
);
805 gfs2_jindex_free(sdp
);
807 gfs2_glock_dq_uninit(&ji_gh
);
809 iput(sdp
->sd_jindex
);
814 static int init_inodes(struct gfs2_sbd
*sdp
, int undo
)
817 struct gfs2_inode
*ip
;
818 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
823 error
= init_journal(sdp
, undo
);
827 /* Read in the master inode number inode */
828 sdp
->sd_inum_inode
= gfs2_lookup_simple(master
, "inum");
829 if (IS_ERR(sdp
->sd_inum_inode
)) {
830 error
= PTR_ERR(sdp
->sd_inum_inode
);
831 fs_err(sdp
, "can't read in inum inode: %d\n", error
);
836 /* Read in the master statfs inode */
837 sdp
->sd_statfs_inode
= gfs2_lookup_simple(master
, "statfs");
838 if (IS_ERR(sdp
->sd_statfs_inode
)) {
839 error
= PTR_ERR(sdp
->sd_statfs_inode
);
840 fs_err(sdp
, "can't read in statfs inode: %d\n", error
);
844 /* Read in the resource index inode */
845 sdp
->sd_rindex
= gfs2_lookup_simple(master
, "rindex");
846 if (IS_ERR(sdp
->sd_rindex
)) {
847 error
= PTR_ERR(sdp
->sd_rindex
);
848 fs_err(sdp
, "can't get resource index inode: %d\n", error
);
851 ip
= GFS2_I(sdp
->sd_rindex
);
852 set_bit(GLF_STICKY
, &ip
->i_gl
->gl_flags
);
853 sdp
->sd_rindex_uptodate
= 0;
855 /* Read in the quota inode */
856 sdp
->sd_quota_inode
= gfs2_lookup_simple(master
, "quota");
857 if (IS_ERR(sdp
->sd_quota_inode
)) {
858 error
= PTR_ERR(sdp
->sd_quota_inode
);
859 fs_err(sdp
, "can't get quota file inode: %d\n", error
);
865 iput(sdp
->sd_quota_inode
);
867 gfs2_clear_rgrpd(sdp
);
868 iput(sdp
->sd_rindex
);
870 iput(sdp
->sd_statfs_inode
);
872 iput(sdp
->sd_inum_inode
);
874 init_journal(sdp
, UNDO
);
879 static int init_per_node(struct gfs2_sbd
*sdp
, int undo
)
881 struct inode
*pn
= NULL
;
884 struct gfs2_inode
*ip
;
885 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
887 if (sdp
->sd_args
.ar_spectator
)
893 pn
= gfs2_lookup_simple(master
, "per_node");
896 fs_err(sdp
, "can't find per_node directory: %d\n", error
);
900 sprintf(buf
, "inum_range%u", sdp
->sd_jdesc
->jd_jid
);
901 sdp
->sd_ir_inode
= gfs2_lookup_simple(pn
, buf
);
902 if (IS_ERR(sdp
->sd_ir_inode
)) {
903 error
= PTR_ERR(sdp
->sd_ir_inode
);
904 fs_err(sdp
, "can't find local \"ir\" file: %d\n", error
);
908 sprintf(buf
, "statfs_change%u", sdp
->sd_jdesc
->jd_jid
);
909 sdp
->sd_sc_inode
= gfs2_lookup_simple(pn
, buf
);
910 if (IS_ERR(sdp
->sd_sc_inode
)) {
911 error
= PTR_ERR(sdp
->sd_sc_inode
);
912 fs_err(sdp
, "can't find local \"sc\" file: %d\n", error
);
916 sprintf(buf
, "quota_change%u", sdp
->sd_jdesc
->jd_jid
);
917 sdp
->sd_qc_inode
= gfs2_lookup_simple(pn
, buf
);
918 if (IS_ERR(sdp
->sd_qc_inode
)) {
919 error
= PTR_ERR(sdp
->sd_qc_inode
);
920 fs_err(sdp
, "can't find local \"qc\" file: %d\n", error
);
927 ip
= GFS2_I(sdp
->sd_ir_inode
);
928 error
= gfs2_glock_nq_init(ip
->i_gl
,
932 fs_err(sdp
, "can't lock local \"ir\" file: %d\n", error
);
936 ip
= GFS2_I(sdp
->sd_sc_inode
);
937 error
= gfs2_glock_nq_init(ip
->i_gl
,
941 fs_err(sdp
, "can't lock local \"sc\" file: %d\n", error
);
945 ip
= GFS2_I(sdp
->sd_qc_inode
);
946 error
= gfs2_glock_nq_init(ip
->i_gl
,
950 fs_err(sdp
, "can't lock local \"qc\" file: %d\n", error
);
957 gfs2_glock_dq_uninit(&sdp
->sd_qc_gh
);
959 gfs2_glock_dq_uninit(&sdp
->sd_sc_gh
);
961 gfs2_glock_dq_uninit(&sdp
->sd_ir_gh
);
963 iput(sdp
->sd_qc_inode
);
965 iput(sdp
->sd_sc_inode
);
967 iput(sdp
->sd_ir_inode
);
974 static int init_threads(struct gfs2_sbd
*sdp
, int undo
)
976 struct task_struct
*p
;
982 sdp
->sd_log_flush_time
= jiffies
;
983 sdp
->sd_jindex_refresh_time
= jiffies
;
985 p
= kthread_run(gfs2_logd
, sdp
, "gfs2_logd");
988 fs_err(sdp
, "can't start logd thread: %d\n", error
);
991 sdp
->sd_logd_process
= p
;
993 sdp
->sd_statfs_sync_time
= jiffies
;
994 sdp
->sd_quota_sync_time
= jiffies
;
996 p
= kthread_run(gfs2_quotad
, sdp
, "gfs2_quotad");
999 fs_err(sdp
, "can't start quotad thread: %d\n", error
);
1002 sdp
->sd_quotad_process
= p
;
1008 kthread_stop(sdp
->sd_quotad_process
);
1010 kthread_stop(sdp
->sd_logd_process
);
1015 * gfs2_lm_mount - mount a locking protocol
1016 * @sdp: the filesystem
1017 * @args: mount arguements
1018 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1023 static int gfs2_lm_mount(struct gfs2_sbd
*sdp
, int silent
)
1025 char *proto
= sdp
->sd_proto_name
;
1026 char *table
= sdp
->sd_table_name
;
1027 int flags
= LM_MFLAG_CONV_NODROP
;
1030 if (sdp
->sd_args
.ar_spectator
)
1031 flags
|= LM_MFLAG_SPECTATOR
;
1033 fs_info(sdp
, "Trying to join cluster \"%s\", \"%s\"\n", proto
, table
);
1035 error
= gfs2_mount_lockproto(proto
, table
, sdp
->sd_args
.ar_hostdata
,
1037 GFS2_MIN_LVB_SIZE
, flags
,
1038 &sdp
->sd_lockstruct
, &sdp
->sd_kobj
);
1040 fs_info(sdp
, "can't mount proto=%s, table=%s, hostdata=%s\n",
1041 proto
, table
, sdp
->sd_args
.ar_hostdata
);
1045 if (gfs2_assert_warn(sdp
, sdp
->sd_lockstruct
.ls_ops
) ||
1046 gfs2_assert_warn(sdp
, sdp
->sd_lockstruct
.ls_lvb_size
>=
1047 GFS2_MIN_LVB_SIZE
)) {
1048 gfs2_unmount_lockproto(&sdp
->sd_lockstruct
);
1052 if (sdp
->sd_args
.ar_spectator
)
1053 snprintf(sdp
->sd_fsname
, GFS2_FSNAME_LEN
, "%s.s", table
);
1055 snprintf(sdp
->sd_fsname
, GFS2_FSNAME_LEN
, "%s.%u", table
,
1056 sdp
->sd_lockstruct
.ls_jid
);
1058 fs_info(sdp
, "Joined cluster. Now mounting FS...\n");
1060 if ((sdp
->sd_lockstruct
.ls_flags
& LM_LSFLAG_LOCAL
) &&
1061 !sdp
->sd_args
.ar_ignore_local_fs
) {
1062 sdp
->sd_args
.ar_localflocks
= 1;
1063 sdp
->sd_args
.ar_localcaching
= 1;
1070 void gfs2_lm_unmount(struct gfs2_sbd
*sdp
)
1072 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
1073 gfs2_unmount_lockproto(&sdp
->sd_lockstruct
);
1077 * fill_super - Read in superblock
1078 * @sb: The VFS superblock
1079 * @data: Mount options
1080 * @silent: Don't complain if it's not a GFS2 filesystem
1085 static int fill_super(struct super_block
*sb
, void *data
, int silent
)
1087 struct gfs2_sbd
*sdp
;
1088 struct gfs2_holder mount_gh
;
1093 printk(KERN_WARNING
"GFS2: can't alloc struct gfs2_sbd\n");
1097 error
= gfs2_mount_args(sdp
, (char *)data
, 0);
1099 printk(KERN_WARNING
"GFS2: can't parse mount arguments\n");
1103 init_vfs(sb
, SDF_NOATIME
);
1105 /* Set up the buffer cache and fill in some fake block size values
1106 to allow us to read-in the on-disk superblock. */
1107 sdp
->sd_sb
.sb_bsize
= sb_min_blocksize(sb
, GFS2_BASIC_BLOCK
);
1108 sdp
->sd_sb
.sb_bsize_shift
= sb
->s_blocksize_bits
;
1109 sdp
->sd_fsb2bb_shift
= sdp
->sd_sb
.sb_bsize_shift
-
1110 GFS2_BASIC_BLOCK_SHIFT
;
1111 sdp
->sd_fsb2bb
= 1 << sdp
->sd_fsb2bb_shift
;
1113 error
= init_names(sdp
, silent
);
1117 gfs2_create_debugfs_file(sdp
);
1119 error
= gfs2_sys_fs_add(sdp
);
1123 error
= gfs2_lm_mount(sdp
, silent
);
1127 error
= init_locking(sdp
, &mount_gh
, DO
);
1131 error
= init_sb(sdp
, silent
);
1135 error
= init_inodes(sdp
, DO
);
1139 error
= init_per_node(sdp
, DO
);
1143 error
= gfs2_statfs_init(sdp
);
1145 fs_err(sdp
, "can't initialize statfs subsystem: %d\n", error
);
1149 error
= init_threads(sdp
, DO
);
1153 if (!(sb
->s_flags
& MS_RDONLY
)) {
1154 error
= gfs2_make_fs_rw(sdp
);
1156 fs_err(sdp
, "can't make FS RW: %d\n", error
);
1161 gfs2_glock_dq_uninit(&mount_gh
);
1166 init_threads(sdp
, UNDO
);
1168 init_per_node(sdp
, UNDO
);
1170 init_inodes(sdp
, UNDO
);
1172 if (sdp
->sd_root_dir
)
1173 dput(sdp
->sd_root_dir
);
1174 if (sdp
->sd_master_dir
)
1175 dput(sdp
->sd_master_dir
);
1178 init_locking(sdp
, &mount_gh
, UNDO
);
1180 gfs2_gl_hash_clear(sdp
);
1181 gfs2_lm_unmount(sdp
);
1182 while (invalidate_inodes(sb
))
1185 gfs2_sys_fs_del(sdp
);
1187 gfs2_delete_debugfs_file(sdp
);
1189 sb
->s_fs_info
= NULL
;
1193 static int gfs2_get_sb(struct file_system_type
*fs_type
, int flags
,
1194 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1196 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, fill_super
, mnt
);
1199 static struct super_block
*get_gfs2_sb(const char *dev_name
)
1201 struct super_block
*sb
;
1202 struct nameidata nd
;
1205 error
= path_lookup(dev_name
, LOOKUP_FOLLOW
, &nd
);
1207 printk(KERN_WARNING
"GFS2: path_lookup on %s returned error %d\n",
1211 sb
= nd
.path
.dentry
->d_inode
->i_sb
;
1212 if (sb
&& (sb
->s_type
== &gfs2_fs_type
))
1213 atomic_inc(&sb
->s_active
);
1220 static int gfs2_get_sb_meta(struct file_system_type
*fs_type
, int flags
,
1221 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1223 struct super_block
*sb
= NULL
;
1224 struct gfs2_sbd
*sdp
;
1226 sb
= get_gfs2_sb(dev_name
);
1228 printk(KERN_WARNING
"GFS2: gfs2 mount does not exist\n");
1231 sdp
= sb
->s_fs_info
;
1233 mnt
->mnt_root
= dget(sdp
->sd_master_dir
);
1237 static void gfs2_kill_sb(struct super_block
*sb
)
1239 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1241 gfs2_meta_syncfs(sdp
);
1242 dput(sdp
->sd_root_dir
);
1243 dput(sdp
->sd_master_dir
);
1244 sdp
->sd_root_dir
= NULL
;
1245 sdp
->sd_master_dir
= NULL
;
1247 shrink_dcache_sb(sb
);
1248 kill_block_super(sb
);
1250 gfs2_delete_debugfs_file(sdp
);
1253 struct file_system_type gfs2_fs_type
= {
1255 .fs_flags
= FS_REQUIRES_DEV
,
1256 .get_sb
= gfs2_get_sb
,
1257 .kill_sb
= gfs2_kill_sb
,
1258 .owner
= THIS_MODULE
,
1261 struct file_system_type gfs2meta_fs_type
= {
1263 .fs_flags
= FS_REQUIRES_DEV
,
1264 .get_sb
= gfs2_get_sb_meta
,
1265 .owner
= THIS_MODULE
,