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>
39 static const u32 gfs2_old_fs_formats
[] = {
43 static const u32 gfs2_old_multihost_formats
[] = {
48 * gfs2_tune_init - Fill a gfs2_tune structure with default values
53 static void gfs2_tune_init(struct gfs2_tune
*gt
)
55 spin_lock_init(>
->gt_spin
);
57 gt
->gt_incore_log_blocks
= 1024;
58 gt
->gt_log_flush_secs
= 60;
59 gt
->gt_recoverd_secs
= 60;
61 gt
->gt_quota_simul_sync
= 64;
62 gt
->gt_quota_warn_period
= 10;
63 gt
->gt_quota_scale_num
= 1;
64 gt
->gt_quota_scale_den
= 1;
65 gt
->gt_quota_quantum
= 60;
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;
70 gt
->gt_statfs_quantum
= 30;
71 gt
->gt_statfs_slow
= 0;
74 static struct gfs2_sbd
*init_sbd(struct super_block
*sb
)
78 sdp
= kzalloc(sizeof(struct gfs2_sbd
), GFP_KERNEL
);
85 gfs2_tune_init(&sdp
->sd_tune
);
87 mutex_init(&sdp
->sd_inum_mutex
);
88 spin_lock_init(&sdp
->sd_statfs_spin
);
90 spin_lock_init(&sdp
->sd_rindex_spin
);
91 mutex_init(&sdp
->sd_rindex_mutex
);
92 INIT_LIST_HEAD(&sdp
->sd_rindex_list
);
93 INIT_LIST_HEAD(&sdp
->sd_rindex_mru_list
);
95 INIT_LIST_HEAD(&sdp
->sd_jindex_list
);
96 spin_lock_init(&sdp
->sd_jindex_spin
);
97 mutex_init(&sdp
->sd_jindex_mutex
);
99 INIT_LIST_HEAD(&sdp
->sd_quota_list
);
100 mutex_init(&sdp
->sd_quota_mutex
);
101 init_waitqueue_head(&sdp
->sd_quota_wait
);
102 INIT_LIST_HEAD(&sdp
->sd_trunc_list
);
103 spin_lock_init(&sdp
->sd_trunc_lock
);
105 spin_lock_init(&sdp
->sd_log_lock
);
107 INIT_LIST_HEAD(&sdp
->sd_log_le_buf
);
108 INIT_LIST_HEAD(&sdp
->sd_log_le_revoke
);
109 INIT_LIST_HEAD(&sdp
->sd_log_le_rg
);
110 INIT_LIST_HEAD(&sdp
->sd_log_le_databuf
);
111 INIT_LIST_HEAD(&sdp
->sd_log_le_ordered
);
113 mutex_init(&sdp
->sd_log_reserve_mutex
);
114 INIT_LIST_HEAD(&sdp
->sd_ail1_list
);
115 INIT_LIST_HEAD(&sdp
->sd_ail2_list
);
117 init_rwsem(&sdp
->sd_log_flush_lock
);
118 atomic_set(&sdp
->sd_log_in_flight
, 0);
119 init_waitqueue_head(&sdp
->sd_log_flush_wait
);
121 INIT_LIST_HEAD(&sdp
->sd_revoke_list
);
123 mutex_init(&sdp
->sd_freeze_lock
);
130 * gfs2_check_sb - Check superblock
131 * @sdp: the filesystem
132 * @sb: The superblock
133 * @silent: Don't print a message if the check fails
135 * Checks the version code of the FS is one that we understand how to
136 * read and that the sizes of the various on-disk structures have not
140 static int gfs2_check_sb(struct gfs2_sbd
*sdp
, struct gfs2_sb_host
*sb
, int silent
)
144 if (sb
->sb_magic
!= GFS2_MAGIC
||
145 sb
->sb_type
!= GFS2_METATYPE_SB
) {
147 printk(KERN_WARNING
"GFS2: not a GFS2 filesystem\n");
151 /* If format numbers match exactly, we're done. */
153 if (sb
->sb_fs_format
== GFS2_FORMAT_FS
&&
154 sb
->sb_multihost_format
== GFS2_FORMAT_MULTI
)
157 if (sb
->sb_fs_format
!= GFS2_FORMAT_FS
) {
158 for (x
= 0; gfs2_old_fs_formats
[x
]; x
++)
159 if (gfs2_old_fs_formats
[x
] == sb
->sb_fs_format
)
162 if (!gfs2_old_fs_formats
[x
]) {
164 "GFS2: code version (%u, %u) is incompatible "
165 "with ondisk format (%u, %u)\n",
166 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
167 sb
->sb_fs_format
, sb
->sb_multihost_format
);
169 "GFS2: I don't know how to upgrade this FS\n");
174 if (sb
->sb_multihost_format
!= GFS2_FORMAT_MULTI
) {
175 for (x
= 0; gfs2_old_multihost_formats
[x
]; x
++)
176 if (gfs2_old_multihost_formats
[x
] ==
177 sb
->sb_multihost_format
)
180 if (!gfs2_old_multihost_formats
[x
]) {
182 "GFS2: code version (%u, %u) is incompatible "
183 "with ondisk format (%u, %u)\n",
184 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
185 sb
->sb_fs_format
, sb
->sb_multihost_format
);
187 "GFS2: I don't know how to upgrade this FS\n");
192 if (!sdp
->sd_args
.ar_upgrade
) {
194 "GFS2: code version (%u, %u) is incompatible "
195 "with ondisk format (%u, %u)\n",
196 GFS2_FORMAT_FS
, GFS2_FORMAT_MULTI
,
197 sb
->sb_fs_format
, sb
->sb_multihost_format
);
199 "GFS2: Use the \"upgrade\" mount option to upgrade "
201 printk(KERN_INFO
"GFS2: See the manual for more details\n");
208 static void end_bio_io_page(struct bio
*bio
, int error
)
210 struct page
*page
= bio
->bi_private
;
213 SetPageUptodate(page
);
215 printk(KERN_WARNING
"gfs2: error %d reading superblock\n", error
);
219 static void gfs2_sb_in(struct gfs2_sb_host
*sb
, const void *buf
)
221 const struct gfs2_sb
*str
= buf
;
223 sb
->sb_magic
= be32_to_cpu(str
->sb_header
.mh_magic
);
224 sb
->sb_type
= be32_to_cpu(str
->sb_header
.mh_type
);
225 sb
->sb_format
= be32_to_cpu(str
->sb_header
.mh_format
);
226 sb
->sb_fs_format
= be32_to_cpu(str
->sb_fs_format
);
227 sb
->sb_multihost_format
= be32_to_cpu(str
->sb_multihost_format
);
228 sb
->sb_bsize
= be32_to_cpu(str
->sb_bsize
);
229 sb
->sb_bsize_shift
= be32_to_cpu(str
->sb_bsize_shift
);
230 sb
->sb_master_dir
.no_addr
= be64_to_cpu(str
->sb_master_dir
.no_addr
);
231 sb
->sb_master_dir
.no_formal_ino
= be64_to_cpu(str
->sb_master_dir
.no_formal_ino
);
232 sb
->sb_root_dir
.no_addr
= be64_to_cpu(str
->sb_root_dir
.no_addr
);
233 sb
->sb_root_dir
.no_formal_ino
= be64_to_cpu(str
->sb_root_dir
.no_formal_ino
);
235 memcpy(sb
->sb_lockproto
, str
->sb_lockproto
, GFS2_LOCKNAME_LEN
);
236 memcpy(sb
->sb_locktable
, str
->sb_locktable
, GFS2_LOCKNAME_LEN
);
237 memcpy(sb
->sb_uuid
, str
->sb_uuid
, 16);
241 * gfs2_read_super - Read the gfs2 super block from disk
242 * @sdp: The GFS2 super block
243 * @sector: The location of the super block
244 * @error: The error code to return
246 * This uses the bio functions to read the super block from disk
247 * because we want to be 100% sure that we never read cached data.
248 * A super block is read twice only during each GFS2 mount and is
249 * never written to by the filesystem. The first time its read no
250 * locks are held, and the only details which are looked at are those
251 * relating to the locking protocol. Once locking is up and working,
252 * the sb is read again under the lock to establish the location of
253 * the master directory (contains pointers to journals etc) and the
256 * Returns: 0 on success or error
259 static int gfs2_read_super(struct gfs2_sbd
*sdp
, sector_t sector
)
261 struct super_block
*sb
= sdp
->sd_vfs
;
266 page
= alloc_page(GFP_NOFS
);
270 ClearPageUptodate(page
);
271 ClearPageDirty(page
);
274 bio
= bio_alloc(GFP_NOFS
, 1);
275 if (unlikely(!bio
)) {
280 bio
->bi_sector
= sector
* (sb
->s_blocksize
>> 9);
281 bio
->bi_bdev
= sb
->s_bdev
;
282 bio_add_page(bio
, page
, PAGE_SIZE
, 0);
284 bio
->bi_end_io
= end_bio_io_page
;
285 bio
->bi_private
= page
;
286 submit_bio(READ_SYNC
| (1 << BIO_RW_META
), bio
);
287 wait_on_page_locked(page
);
289 if (!PageUptodate(page
)) {
294 gfs2_sb_in(&sdp
->sd_sb
, p
);
301 * gfs2_read_sb - Read super block
302 * @sdp: The GFS2 superblock
303 * @silent: Don't print message if mount fails
307 static int gfs2_read_sb(struct gfs2_sbd
*sdp
, int silent
)
309 u32 hash_blocks
, ind_blocks
, leaf_blocks
;
314 error
= gfs2_read_super(sdp
, GFS2_SB_ADDR
>> sdp
->sd_fsb2bb_shift
);
317 fs_err(sdp
, "can't read superblock\n");
321 error
= gfs2_check_sb(sdp
, &sdp
->sd_sb
, silent
);
325 sdp
->sd_fsb2bb_shift
= sdp
->sd_sb
.sb_bsize_shift
-
326 GFS2_BASIC_BLOCK_SHIFT
;
327 sdp
->sd_fsb2bb
= 1 << sdp
->sd_fsb2bb_shift
;
328 sdp
->sd_diptrs
= (sdp
->sd_sb
.sb_bsize
-
329 sizeof(struct gfs2_dinode
)) / sizeof(u64
);
330 sdp
->sd_inptrs
= (sdp
->sd_sb
.sb_bsize
-
331 sizeof(struct gfs2_meta_header
)) / sizeof(u64
);
332 sdp
->sd_jbsize
= sdp
->sd_sb
.sb_bsize
- sizeof(struct gfs2_meta_header
);
333 sdp
->sd_hash_bsize
= sdp
->sd_sb
.sb_bsize
/ 2;
334 sdp
->sd_hash_bsize_shift
= sdp
->sd_sb
.sb_bsize_shift
- 1;
335 sdp
->sd_hash_ptrs
= sdp
->sd_hash_bsize
/ sizeof(u64
);
336 sdp
->sd_qc_per_block
= (sdp
->sd_sb
.sb_bsize
-
337 sizeof(struct gfs2_meta_header
)) /
338 sizeof(struct gfs2_quota_change
);
340 /* Compute maximum reservation required to add a entry to a directory */
342 hash_blocks
= DIV_ROUND_UP(sizeof(u64
) * (1 << GFS2_DIR_MAX_DEPTH
),
346 for (tmp_blocks
= hash_blocks
; tmp_blocks
> sdp
->sd_diptrs
;) {
347 tmp_blocks
= DIV_ROUND_UP(tmp_blocks
, sdp
->sd_inptrs
);
348 ind_blocks
+= tmp_blocks
;
351 leaf_blocks
= 2 + GFS2_DIR_MAX_DEPTH
;
353 sdp
->sd_max_dirres
= hash_blocks
+ ind_blocks
+ leaf_blocks
;
355 sdp
->sd_heightsize
[0] = sdp
->sd_sb
.sb_bsize
-
356 sizeof(struct gfs2_dinode
);
357 sdp
->sd_heightsize
[1] = sdp
->sd_sb
.sb_bsize
* sdp
->sd_diptrs
;
362 space
= sdp
->sd_heightsize
[x
- 1] * sdp
->sd_inptrs
;
364 m
= do_div(d
, sdp
->sd_inptrs
);
366 if (d
!= sdp
->sd_heightsize
[x
- 1] || m
)
368 sdp
->sd_heightsize
[x
] = space
;
370 sdp
->sd_max_height
= x
;
371 sdp
->sd_heightsize
[x
] = ~0;
372 gfs2_assert(sdp
, sdp
->sd_max_height
<= GFS2_MAX_META_HEIGHT
);
374 sdp
->sd_jheightsize
[0] = sdp
->sd_sb
.sb_bsize
-
375 sizeof(struct gfs2_dinode
);
376 sdp
->sd_jheightsize
[1] = sdp
->sd_jbsize
* sdp
->sd_diptrs
;
381 space
= sdp
->sd_jheightsize
[x
- 1] * sdp
->sd_inptrs
;
383 m
= do_div(d
, sdp
->sd_inptrs
);
385 if (d
!= sdp
->sd_jheightsize
[x
- 1] || m
)
387 sdp
->sd_jheightsize
[x
] = space
;
389 sdp
->sd_max_jheight
= x
;
390 sdp
->sd_jheightsize
[x
] = ~0;
391 gfs2_assert(sdp
, sdp
->sd_max_jheight
<= GFS2_MAX_META_HEIGHT
);
396 static int init_names(struct gfs2_sbd
*sdp
, int silent
)
401 proto
= sdp
->sd_args
.ar_lockproto
;
402 table
= sdp
->sd_args
.ar_locktable
;
404 /* Try to autodetect */
406 if (!proto
[0] || !table
[0]) {
407 error
= gfs2_read_super(sdp
, GFS2_SB_ADDR
>> sdp
->sd_fsb2bb_shift
);
411 error
= gfs2_check_sb(sdp
, &sdp
->sd_sb
, silent
);
416 proto
= sdp
->sd_sb
.sb_lockproto
;
418 table
= sdp
->sd_sb
.sb_locktable
;
422 table
= sdp
->sd_vfs
->s_id
;
424 strlcpy(sdp
->sd_proto_name
, proto
, GFS2_FSNAME_LEN
);
425 strlcpy(sdp
->sd_table_name
, table
, GFS2_FSNAME_LEN
);
427 table
= sdp
->sd_table_name
;
428 while ((table
= strchr(table
, '/')))
435 static int init_locking(struct gfs2_sbd
*sdp
, struct gfs2_holder
*mount_gh
,
443 error
= gfs2_glock_nq_num(sdp
,
444 GFS2_MOUNT_LOCK
, &gfs2_nondisk_glops
,
445 LM_ST_EXCLUSIVE
, LM_FLAG_NOEXP
| GL_NOCACHE
,
448 fs_err(sdp
, "can't acquire mount glock: %d\n", error
);
452 error
= gfs2_glock_nq_num(sdp
,
453 GFS2_LIVE_LOCK
, &gfs2_nondisk_glops
,
455 LM_FLAG_NOEXP
| GL_EXACT
,
458 fs_err(sdp
, "can't acquire live glock: %d\n", error
);
462 error
= gfs2_glock_get(sdp
, GFS2_RENAME_LOCK
, &gfs2_nondisk_glops
,
463 CREATE
, &sdp
->sd_rename_gl
);
465 fs_err(sdp
, "can't create rename glock: %d\n", error
);
469 error
= gfs2_glock_get(sdp
, GFS2_TRANS_LOCK
, &gfs2_trans_glops
,
470 CREATE
, &sdp
->sd_trans_gl
);
472 fs_err(sdp
, "can't create transaction glock: %d\n", error
);
479 gfs2_glock_put(sdp
->sd_trans_gl
);
481 gfs2_glock_put(sdp
->sd_rename_gl
);
483 gfs2_glock_dq_uninit(&sdp
->sd_live_gh
);
485 gfs2_glock_dq_uninit(mount_gh
);
490 static int gfs2_lookup_root(struct super_block
*sb
, struct dentry
**dptr
,
491 u64 no_addr
, const char *name
)
493 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
494 struct dentry
*dentry
;
497 inode
= gfs2_inode_lookup(sb
, DT_DIR
, no_addr
, 0, 0);
499 fs_err(sdp
, "can't read in %s inode: %ld\n", name
, PTR_ERR(inode
));
500 return PTR_ERR(inode
);
502 dentry
= d_alloc_root(inode
);
504 fs_err(sdp
, "can't alloc %s dentry\n", name
);
508 dentry
->d_op
= &gfs2_dops
;
513 static int init_sb(struct gfs2_sbd
*sdp
, int silent
)
515 struct super_block
*sb
= sdp
->sd_vfs
;
516 struct gfs2_holder sb_gh
;
520 ret
= gfs2_glock_nq_num(sdp
, GFS2_SB_LOCK
, &gfs2_meta_glops
,
521 LM_ST_SHARED
, 0, &sb_gh
);
523 fs_err(sdp
, "can't acquire superblock glock: %d\n", ret
);
527 ret
= gfs2_read_sb(sdp
, silent
);
529 fs_err(sdp
, "can't read superblock: %d\n", ret
);
533 /* Set up the buffer cache and SB for real */
534 if (sdp
->sd_sb
.sb_bsize
< bdev_hardsect_size(sb
->s_bdev
)) {
536 fs_err(sdp
, "FS block size (%u) is too small for device "
538 sdp
->sd_sb
.sb_bsize
, bdev_hardsect_size(sb
->s_bdev
));
541 if (sdp
->sd_sb
.sb_bsize
> PAGE_SIZE
) {
543 fs_err(sdp
, "FS block size (%u) is too big for machine "
545 sdp
->sd_sb
.sb_bsize
, (unsigned int)PAGE_SIZE
);
548 sb_set_blocksize(sb
, sdp
->sd_sb
.sb_bsize
);
550 /* Get the root inode */
551 no_addr
= sdp
->sd_sb
.sb_root_dir
.no_addr
;
552 ret
= gfs2_lookup_root(sb
, &sdp
->sd_root_dir
, no_addr
, "root");
556 /* Get the master inode */
557 no_addr
= sdp
->sd_sb
.sb_master_dir
.no_addr
;
558 ret
= gfs2_lookup_root(sb
, &sdp
->sd_master_dir
, no_addr
, "master");
560 dput(sdp
->sd_root_dir
);
563 sb
->s_root
= dget(sdp
->sd_args
.ar_meta
? sdp
->sd_master_dir
: sdp
->sd_root_dir
);
565 gfs2_glock_dq_uninit(&sb_gh
);
570 * map_journal_extents - create a reusable "extent" mapping from all logical
571 * blocks to all physical blocks for the given journal. This will save
572 * us time when writing journal blocks. Most journals will have only one
573 * extent that maps all their logical blocks. That's because gfs2.mkfs
574 * arranges the journal blocks sequentially to maximize performance.
575 * So the extent would map the first block for the entire file length.
576 * However, gfs2_jadd can happen while file activity is happening, so
577 * those journals may not be sequential. Less likely is the case where
578 * the users created their own journals by mounting the metafs and
579 * laying it out. But it's still possible. These journals might have
582 * TODO: This should be done in bigger chunks rather than one block at a time,
583 * but since it's only done at mount time, I'm not worried about the
586 static int map_journal_extents(struct gfs2_sbd
*sdp
)
588 struct gfs2_jdesc
*jd
= sdp
->sd_jdesc
;
590 u64 db
, prev_db
; /* logical block, disk block, prev disk block */
591 struct gfs2_inode
*ip
= GFS2_I(jd
->jd_inode
);
592 struct gfs2_journal_extent
*jext
= NULL
;
593 struct buffer_head bh
;
598 for (lb
= 0; lb
< ip
->i_disksize
>> sdp
->sd_sb
.sb_bsize_shift
; lb
++) {
601 bh
.b_size
= 1 << ip
->i_inode
.i_blkbits
;
602 rc
= gfs2_block_map(jd
->jd_inode
, lb
, &bh
, 0);
605 printk(KERN_INFO
"GFS2 journal mapping error %d: lb="
606 "%u db=%llu\n", rc
, lb
, (unsigned long long)db
);
609 if (!prev_db
|| db
!= prev_db
+ 1) {
610 jext
= kzalloc(sizeof(struct gfs2_journal_extent
),
613 printk(KERN_INFO
"GFS2 error: out of memory "
614 "mapping journal extents.\n");
621 list_add_tail(&jext
->extent_list
, &jd
->extent_list
);
630 static void gfs2_others_may_mount(struct gfs2_sbd
*sdp
)
632 char *message
= "FIRSTMOUNT=Done";
633 char *envp
[] = { message
, NULL
};
634 struct lm_lockstruct
*ls
= &sdp
->sd_lockstruct
;
635 ls
->ls_first_done
= 1;
636 kobject_uevent_env(&sdp
->sd_kobj
, KOBJ_CHANGE
, envp
);
640 * gfs2_jindex_hold - Grab a lock on the jindex
641 * @sdp: The GFS2 superblock
642 * @ji_gh: the holder for the jindex glock
647 static int gfs2_jindex_hold(struct gfs2_sbd
*sdp
, struct gfs2_holder
*ji_gh
)
649 struct gfs2_inode
*dip
= GFS2_I(sdp
->sd_jindex
);
652 struct gfs2_jdesc
*jd
;
657 mutex_lock(&sdp
->sd_jindex_mutex
);
660 error
= gfs2_glock_nq_init(dip
->i_gl
, LM_ST_SHARED
, 0, ji_gh
);
664 name
.len
= sprintf(buf
, "journal%u", sdp
->sd_journals
);
665 name
.hash
= gfs2_disk_hash(name
.name
, name
.len
);
667 error
= gfs2_dir_check(sdp
->sd_jindex
, &name
, NULL
);
668 if (error
== -ENOENT
) {
673 gfs2_glock_dq_uninit(ji_gh
);
679 jd
= kzalloc(sizeof(struct gfs2_jdesc
), GFP_KERNEL
);
683 INIT_LIST_HEAD(&jd
->extent_list
);
684 jd
->jd_inode
= gfs2_lookupi(sdp
->sd_jindex
, &name
, 1);
685 if (!jd
->jd_inode
|| IS_ERR(jd
->jd_inode
)) {
689 error
= PTR_ERR(jd
->jd_inode
);
694 spin_lock(&sdp
->sd_jindex_spin
);
695 jd
->jd_jid
= sdp
->sd_journals
++;
696 list_add_tail(&jd
->jd_list
, &sdp
->sd_jindex_list
);
697 spin_unlock(&sdp
->sd_jindex_spin
);
700 mutex_unlock(&sdp
->sd_jindex_mutex
);
705 static int init_journal(struct gfs2_sbd
*sdp
, int undo
)
707 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
708 struct gfs2_holder ji_gh
;
709 struct task_struct
*p
;
710 struct gfs2_inode
*ip
;
719 sdp
->sd_jindex
= gfs2_lookup_simple(master
, "jindex");
720 if (IS_ERR(sdp
->sd_jindex
)) {
721 fs_err(sdp
, "can't lookup journal index: %d\n", error
);
722 return PTR_ERR(sdp
->sd_jindex
);
724 ip
= GFS2_I(sdp
->sd_jindex
);
726 /* Load in the journal index special file */
728 error
= gfs2_jindex_hold(sdp
, &ji_gh
);
730 fs_err(sdp
, "can't read journal index: %d\n", error
);
735 if (!gfs2_jindex_size(sdp
)) {
736 fs_err(sdp
, "no journals!\n");
740 if (sdp
->sd_args
.ar_spectator
) {
741 sdp
->sd_jdesc
= gfs2_jdesc_find(sdp
, 0);
742 atomic_set(&sdp
->sd_log_blks_free
, sdp
->sd_jdesc
->jd_blocks
);
744 if (sdp
->sd_lockstruct
.ls_jid
>= gfs2_jindex_size(sdp
)) {
745 fs_err(sdp
, "can't mount journal #%u\n",
746 sdp
->sd_lockstruct
.ls_jid
);
747 fs_err(sdp
, "there are only %u journals (0 - %u)\n",
748 gfs2_jindex_size(sdp
),
749 gfs2_jindex_size(sdp
) - 1);
752 sdp
->sd_jdesc
= gfs2_jdesc_find(sdp
, sdp
->sd_lockstruct
.ls_jid
);
754 error
= gfs2_glock_nq_num(sdp
, sdp
->sd_lockstruct
.ls_jid
,
756 LM_ST_EXCLUSIVE
, LM_FLAG_NOEXP
,
757 &sdp
->sd_journal_gh
);
759 fs_err(sdp
, "can't acquire journal glock: %d\n", error
);
763 ip
= GFS2_I(sdp
->sd_jdesc
->jd_inode
);
764 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
,
765 LM_FLAG_NOEXP
| GL_EXACT
| GL_NOCACHE
,
768 fs_err(sdp
, "can't acquire journal inode glock: %d\n",
770 goto fail_journal_gh
;
773 error
= gfs2_jdesc_check(sdp
->sd_jdesc
);
775 fs_err(sdp
, "my journal (%u) is bad: %d\n",
776 sdp
->sd_jdesc
->jd_jid
, error
);
779 atomic_set(&sdp
->sd_log_blks_free
, sdp
->sd_jdesc
->jd_blocks
);
781 /* Map the extents for this journal's blocks */
782 map_journal_extents(sdp
);
785 if (sdp
->sd_lockstruct
.ls_first
) {
787 for (x
= 0; x
< sdp
->sd_journals
; x
++) {
788 error
= gfs2_recover_journal(gfs2_jdesc_find(sdp
, x
));
790 fs_err(sdp
, "error recovering journal %u: %d\n",
796 gfs2_others_may_mount(sdp
);
797 } else if (!sdp
->sd_args
.ar_spectator
) {
798 error
= gfs2_recover_journal(sdp
->sd_jdesc
);
800 fs_err(sdp
, "error recovering my journal: %d\n", error
);
805 set_bit(SDF_JOURNAL_CHECKED
, &sdp
->sd_flags
);
806 gfs2_glock_dq_uninit(&ji_gh
);
809 p
= kthread_run(gfs2_recoverd
, sdp
, "gfs2_recoverd");
812 fs_err(sdp
, "can't start recoverd thread: %d\n", error
);
815 sdp
->sd_recoverd_process
= p
;
820 kthread_stop(sdp
->sd_recoverd_process
);
822 if (!sdp
->sd_args
.ar_spectator
)
823 gfs2_glock_dq_uninit(&sdp
->sd_jinode_gh
);
825 if (!sdp
->sd_args
.ar_spectator
)
826 gfs2_glock_dq_uninit(&sdp
->sd_journal_gh
);
828 gfs2_jindex_free(sdp
);
830 gfs2_glock_dq_uninit(&ji_gh
);
832 iput(sdp
->sd_jindex
);
837 static int init_inodes(struct gfs2_sbd
*sdp
, int undo
)
840 struct gfs2_inode
*ip
;
841 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
846 error
= init_journal(sdp
, undo
);
850 /* Read in the master inode number inode */
851 sdp
->sd_inum_inode
= gfs2_lookup_simple(master
, "inum");
852 if (IS_ERR(sdp
->sd_inum_inode
)) {
853 error
= PTR_ERR(sdp
->sd_inum_inode
);
854 fs_err(sdp
, "can't read in inum inode: %d\n", error
);
859 /* Read in the master statfs inode */
860 sdp
->sd_statfs_inode
= gfs2_lookup_simple(master
, "statfs");
861 if (IS_ERR(sdp
->sd_statfs_inode
)) {
862 error
= PTR_ERR(sdp
->sd_statfs_inode
);
863 fs_err(sdp
, "can't read in statfs inode: %d\n", error
);
867 /* Read in the resource index inode */
868 sdp
->sd_rindex
= gfs2_lookup_simple(master
, "rindex");
869 if (IS_ERR(sdp
->sd_rindex
)) {
870 error
= PTR_ERR(sdp
->sd_rindex
);
871 fs_err(sdp
, "can't get resource index inode: %d\n", error
);
874 ip
= GFS2_I(sdp
->sd_rindex
);
875 sdp
->sd_rindex_uptodate
= 0;
877 /* Read in the quota inode */
878 sdp
->sd_quota_inode
= gfs2_lookup_simple(master
, "quota");
879 if (IS_ERR(sdp
->sd_quota_inode
)) {
880 error
= PTR_ERR(sdp
->sd_quota_inode
);
881 fs_err(sdp
, "can't get quota file inode: %d\n", error
);
887 iput(sdp
->sd_quota_inode
);
889 gfs2_clear_rgrpd(sdp
);
890 iput(sdp
->sd_rindex
);
892 iput(sdp
->sd_statfs_inode
);
894 iput(sdp
->sd_inum_inode
);
896 init_journal(sdp
, UNDO
);
901 static int init_per_node(struct gfs2_sbd
*sdp
, int undo
)
903 struct inode
*pn
= NULL
;
906 struct gfs2_inode
*ip
;
907 struct inode
*master
= sdp
->sd_master_dir
->d_inode
;
909 if (sdp
->sd_args
.ar_spectator
)
915 pn
= gfs2_lookup_simple(master
, "per_node");
918 fs_err(sdp
, "can't find per_node directory: %d\n", error
);
922 sprintf(buf
, "inum_range%u", sdp
->sd_jdesc
->jd_jid
);
923 sdp
->sd_ir_inode
= gfs2_lookup_simple(pn
, buf
);
924 if (IS_ERR(sdp
->sd_ir_inode
)) {
925 error
= PTR_ERR(sdp
->sd_ir_inode
);
926 fs_err(sdp
, "can't find local \"ir\" file: %d\n", error
);
930 sprintf(buf
, "statfs_change%u", sdp
->sd_jdesc
->jd_jid
);
931 sdp
->sd_sc_inode
= gfs2_lookup_simple(pn
, buf
);
932 if (IS_ERR(sdp
->sd_sc_inode
)) {
933 error
= PTR_ERR(sdp
->sd_sc_inode
);
934 fs_err(sdp
, "can't find local \"sc\" file: %d\n", error
);
938 sprintf(buf
, "quota_change%u", sdp
->sd_jdesc
->jd_jid
);
939 sdp
->sd_qc_inode
= gfs2_lookup_simple(pn
, buf
);
940 if (IS_ERR(sdp
->sd_qc_inode
)) {
941 error
= PTR_ERR(sdp
->sd_qc_inode
);
942 fs_err(sdp
, "can't find local \"qc\" file: %d\n", error
);
949 ip
= GFS2_I(sdp
->sd_ir_inode
);
950 error
= gfs2_glock_nq_init(ip
->i_gl
,
954 fs_err(sdp
, "can't lock local \"ir\" file: %d\n", error
);
958 ip
= GFS2_I(sdp
->sd_sc_inode
);
959 error
= gfs2_glock_nq_init(ip
->i_gl
,
963 fs_err(sdp
, "can't lock local \"sc\" file: %d\n", error
);
967 ip
= GFS2_I(sdp
->sd_qc_inode
);
968 error
= gfs2_glock_nq_init(ip
->i_gl
,
972 fs_err(sdp
, "can't lock local \"qc\" file: %d\n", error
);
979 gfs2_glock_dq_uninit(&sdp
->sd_qc_gh
);
981 gfs2_glock_dq_uninit(&sdp
->sd_sc_gh
);
983 gfs2_glock_dq_uninit(&sdp
->sd_ir_gh
);
985 iput(sdp
->sd_qc_inode
);
987 iput(sdp
->sd_sc_inode
);
989 iput(sdp
->sd_ir_inode
);
996 static int init_threads(struct gfs2_sbd
*sdp
, int undo
)
998 struct task_struct
*p
;
1004 sdp
->sd_log_flush_time
= jiffies
;
1006 p
= kthread_run(gfs2_logd
, sdp
, "gfs2_logd");
1009 fs_err(sdp
, "can't start logd thread: %d\n", error
);
1012 sdp
->sd_logd_process
= p
;
1014 p
= kthread_run(gfs2_quotad
, sdp
, "gfs2_quotad");
1017 fs_err(sdp
, "can't start quotad thread: %d\n", error
);
1020 sdp
->sd_quotad_process
= p
;
1026 kthread_stop(sdp
->sd_quotad_process
);
1028 kthread_stop(sdp
->sd_logd_process
);
1032 static const match_table_t nolock_tokens
= {
1033 { Opt_jid
, "jid=%d\n", },
1037 static const struct lm_lockops nolock_ops
= {
1038 .lm_proto_name
= "lock_nolock",
1039 .lm_put_lock
= kmem_cache_free
,
1040 .lm_tokens
= &nolock_tokens
,
1044 * gfs2_lm_mount - mount a locking protocol
1045 * @sdp: the filesystem
1046 * @args: mount arguements
1047 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1052 static int gfs2_lm_mount(struct gfs2_sbd
*sdp
, int silent
)
1054 const struct lm_lockops
*lm
;
1055 struct lm_lockstruct
*ls
= &sdp
->sd_lockstruct
;
1056 struct gfs2_args
*args
= &sdp
->sd_args
;
1057 const char *proto
= sdp
->sd_proto_name
;
1058 const char *table
= sdp
->sd_table_name
;
1063 if (!strcmp("lock_nolock", proto
)) {
1065 sdp
->sd_args
.ar_localflocks
= 1;
1066 sdp
->sd_args
.ar_localcaching
= 1;
1067 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1068 } else if (!strcmp("lock_dlm", proto
)) {
1072 printk(KERN_INFO
"GFS2: can't find protocol %s\n", proto
);
1076 fs_info(sdp
, "Trying to join cluster \"%s\", \"%s\"\n", proto
, table
);
1082 for (options
= args
->ar_hostdata
; (o
= strsep(&options
, ":")); ) {
1083 substring_t tmp
[MAX_OPT_ARGS
];
1089 token
= match_token(o
, *lm
->lm_tokens
, tmp
);
1092 ret
= match_int(&tmp
[0], &option
);
1093 if (ret
|| option
< 0)
1094 goto hostdata_error
;
1095 ls
->ls_jid
= option
;
1098 ret
= match_int(&tmp
[0], &option
);
1100 goto hostdata_error
;
1104 ret
= match_int(&tmp
[0], &option
);
1105 if (ret
|| (option
!= 0 && option
!= 1))
1106 goto hostdata_error
;
1107 ls
->ls_first
= option
;
1110 ret
= match_int(&tmp
[0], &option
);
1111 if (ret
|| (option
!= 0 && option
!= 1))
1112 goto hostdata_error
;
1113 ls
->ls_nodir
= option
;
1118 fs_info(sdp
, "unknown hostdata (%s)\n", o
);
1123 if (sdp
->sd_args
.ar_spectator
)
1124 snprintf(sdp
->sd_fsname
, GFS2_FSNAME_LEN
, "%s.s", table
);
1126 snprintf(sdp
->sd_fsname
, GFS2_FSNAME_LEN
, "%s.%u", table
,
1127 sdp
->sd_lockstruct
.ls_jid
);
1129 fsname
= strchr(table
, ':');
1132 if (lm
->lm_mount
== NULL
) {
1133 fs_info(sdp
, "Now mounting FS...\n");
1136 ret
= lm
->lm_mount(sdp
, fsname
);
1138 fs_info(sdp
, "Joined cluster. Now mounting FS...\n");
1142 void gfs2_lm_unmount(struct gfs2_sbd
*sdp
)
1144 const struct lm_lockops
*lm
= sdp
->sd_lockstruct
.ls_ops
;
1145 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)) &&
1147 lm
->lm_unmount(sdp
);
1151 * fill_super - Read in superblock
1152 * @sb: The VFS superblock
1153 * @data: Mount options
1154 * @silent: Don't complain if it's not a GFS2 filesystem
1159 static int fill_super(struct super_block
*sb
, void *data
, int silent
)
1161 struct gfs2_sbd
*sdp
;
1162 struct gfs2_holder mount_gh
;
1167 printk(KERN_WARNING
"GFS2: can't alloc struct gfs2_sbd\n");
1171 sdp
->sd_args
.ar_quota
= GFS2_QUOTA_DEFAULT
;
1172 sdp
->sd_args
.ar_data
= GFS2_DATA_DEFAULT
;
1174 error
= gfs2_mount_args(sdp
, &sdp
->sd_args
, data
);
1176 printk(KERN_WARNING
"GFS2: can't parse mount arguments\n");
1180 if (sdp
->sd_args
.ar_spectator
)
1181 sb
->s_flags
|= MS_RDONLY
;
1182 if (sdp
->sd_args
.ar_posix_acl
)
1183 sb
->s_flags
|= MS_POSIXACL
;
1185 sb
->s_magic
= GFS2_MAGIC
;
1186 sb
->s_op
= &gfs2_super_ops
;
1187 sb
->s_export_op
= &gfs2_export_ops
;
1188 sb
->s_time_gran
= 1;
1189 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
1191 /* Set up the buffer cache and fill in some fake block size values
1192 to allow us to read-in the on-disk superblock. */
1193 sdp
->sd_sb
.sb_bsize
= sb_min_blocksize(sb
, GFS2_BASIC_BLOCK
);
1194 sdp
->sd_sb
.sb_bsize_shift
= sb
->s_blocksize_bits
;
1195 sdp
->sd_fsb2bb_shift
= sdp
->sd_sb
.sb_bsize_shift
-
1196 GFS2_BASIC_BLOCK_SHIFT
;
1197 sdp
->sd_fsb2bb
= 1 << sdp
->sd_fsb2bb_shift
;
1199 error
= init_names(sdp
, silent
);
1203 gfs2_create_debugfs_file(sdp
);
1205 error
= gfs2_sys_fs_add(sdp
);
1209 error
= gfs2_lm_mount(sdp
, silent
);
1213 error
= init_locking(sdp
, &mount_gh
, DO
);
1217 error
= init_sb(sdp
, silent
);
1221 error
= init_inodes(sdp
, DO
);
1225 error
= init_per_node(sdp
, DO
);
1229 error
= gfs2_statfs_init(sdp
);
1231 fs_err(sdp
, "can't initialize statfs subsystem: %d\n", error
);
1235 error
= init_threads(sdp
, DO
);
1239 if (!(sb
->s_flags
& MS_RDONLY
)) {
1240 error
= gfs2_make_fs_rw(sdp
);
1242 fs_err(sdp
, "can't make FS RW: %d\n", error
);
1247 gfs2_glock_dq_uninit(&mount_gh
);
1252 init_threads(sdp
, UNDO
);
1254 init_per_node(sdp
, UNDO
);
1256 init_inodes(sdp
, UNDO
);
1258 if (sdp
->sd_root_dir
)
1259 dput(sdp
->sd_root_dir
);
1260 if (sdp
->sd_master_dir
)
1261 dput(sdp
->sd_master_dir
);
1266 init_locking(sdp
, &mount_gh
, UNDO
);
1268 gfs2_gl_hash_clear(sdp
);
1269 gfs2_lm_unmount(sdp
);
1270 while (invalidate_inodes(sb
))
1273 gfs2_sys_fs_del(sdp
);
1275 gfs2_delete_debugfs_file(sdp
);
1277 sb
->s_fs_info
= NULL
;
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 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, fill_super
, mnt
);
1287 static struct super_block
*get_gfs2_sb(const char *dev_name
)
1289 struct super_block
*sb
;
1290 struct nameidata nd
;
1293 error
= path_lookup(dev_name
, LOOKUP_FOLLOW
, &nd
);
1295 printk(KERN_WARNING
"GFS2: path_lookup on %s returned error %d\n",
1299 sb
= nd
.path
.dentry
->d_inode
->i_sb
;
1300 if (sb
&& (sb
->s_type
== &gfs2_fs_type
))
1301 atomic_inc(&sb
->s_active
);
1308 static int gfs2_get_sb_meta(struct file_system_type
*fs_type
, int flags
,
1309 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
1311 struct super_block
*sb
= NULL
;
1312 struct gfs2_sbd
*sdp
;
1314 sb
= get_gfs2_sb(dev_name
);
1316 printk(KERN_WARNING
"GFS2: gfs2 mount does not exist\n");
1319 sdp
= sb
->s_fs_info
;
1321 mnt
->mnt_root
= dget(sdp
->sd_master_dir
);
1325 static void gfs2_kill_sb(struct super_block
*sb
)
1327 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1330 kill_block_super(sb
);
1334 gfs2_meta_syncfs(sdp
);
1335 dput(sdp
->sd_root_dir
);
1336 dput(sdp
->sd_master_dir
);
1337 sdp
->sd_root_dir
= NULL
;
1338 sdp
->sd_master_dir
= NULL
;
1339 shrink_dcache_sb(sb
);
1340 kill_block_super(sb
);
1341 gfs2_delete_debugfs_file(sdp
);
1345 struct file_system_type gfs2_fs_type
= {
1347 .fs_flags
= FS_REQUIRES_DEV
,
1348 .get_sb
= gfs2_get_sb
,
1349 .kill_sb
= gfs2_kill_sb
,
1350 .owner
= THIS_MODULE
,
1353 struct file_system_type gfs2meta_fs_type
= {
1355 .fs_flags
= FS_REQUIRES_DEV
,
1356 .get_sb
= gfs2_get_sb_meta
,
1357 .owner
= THIS_MODULE
,