]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - fs/gfs2/ops_fstype.c
GFS2: Clean up freeze code
[mirror_ubuntu-zesty-kernel.git] / fs / gfs2 / ops_fstype.c
CommitLineData
b3b94faa
DT
1/*
2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
cf45b752 3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
b3b94faa
DT
4 *
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
e9fc2aa0 7 * of the GNU General Public License version 2.
b3b94faa
DT
8 */
9
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>
b3b94faa
DT
15#include <linux/blkdev.h>
16#include <linux/kthread.h>
afeacc8c 17#include <linux/export.h>
86384605
AD
18#include <linux/namei.h>
19#include <linux/mount.h>
5c676f6d 20#include <linux/gfs2_ondisk.h>
cc632e7f 21#include <linux/quotaops.h>
56aa72d0 22#include <linux/lockdep.h>
b3b94faa
DT
23
24#include "gfs2.h"
5c676f6d 25#include "incore.h"
da6dd40d 26#include "bmap.h"
b3b94faa
DT
27#include "glock.h"
28#include "glops.h"
29#include "inode.h"
b3b94faa
DT
30#include "recovery.h"
31#include "rgrp.h"
32#include "super.h"
b3b94faa 33#include "sys.h"
5c676f6d 34#include "util.h"
bb3b0e3d 35#include "log.h"
9ac1b4d9 36#include "quota.h"
b5289681 37#include "dir.h"
63997775 38#include "trace_gfs2.h"
b3b94faa
DT
39
40#define DO 0
41#define UNDO 1
42
9b8df98f
SW
43/**
44 * gfs2_tune_init - Fill a gfs2_tune structure with default values
45 * @gt: tune
46 *
47 */
48
49static void gfs2_tune_init(struct gfs2_tune *gt)
50{
51 spin_lock_init(&gt->gt_spin);
52
9b8df98f
SW
53 gt->gt_quota_simul_sync = 64;
54 gt->gt_quota_warn_period = 10;
55 gt->gt_quota_scale_num = 1;
56 gt->gt_quota_scale_den = 1;
9b8df98f
SW
57 gt->gt_new_files_jdata = 0;
58 gt->gt_max_readahead = 1 << 18;
9b8df98f 59 gt->gt_complain_secs = 10;
9b8df98f
SW
60}
61
b3b94faa
DT
62static struct gfs2_sbd *init_sbd(struct super_block *sb)
63{
64 struct gfs2_sbd *sdp;
b3b94faa 65
85d1da67 66 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
67 if (!sdp)
68 return NULL;
69
5c676f6d 70 sb->s_fs_info = sdp;
b3b94faa 71 sdp->sd_vfs = sb;
a245769f
SW
72 sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
73 if (!sdp->sd_lkstats) {
74 kfree(sdp);
75 return NULL;
76 }
77
ba6e9364 78 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
79 gfs2_tune_init(&sdp->sd_tune);
80
e402746a
SW
81 init_waitqueue_head(&sdp->sd_glock_wait);
82 atomic_set(&sdp->sd_glock_disposal, 0);
3942ae53 83 init_completion(&sdp->sd_locking_init);
b3b94faa 84 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
85
86 spin_lock_init(&sdp->sd_rindex_spin);
7c9ca621 87 sdp->sd_rindex_tree.rb_node = NULL;
b3b94faa
DT
88
89 INIT_LIST_HEAD(&sdp->sd_jindex_list);
90 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 91 mutex_init(&sdp->sd_jindex_mutex);
b3b94faa 92
b3b94faa 93 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 94 mutex_init(&sdp->sd_quota_mutex);
37b2c837 95 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
96 INIT_LIST_HEAD(&sdp->sd_trunc_list);
97 spin_lock_init(&sdp->sd_trunc_lock);
b3b94faa
DT
98
99 spin_lock_init(&sdp->sd_log_lock);
5e687eac 100 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa
DT
101 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
102 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
b3b94faa 103 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
d7b616e2 104 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
b3b94faa 105
5e687eac
BM
106 init_waitqueue_head(&sdp->sd_log_waitq);
107 init_waitqueue_head(&sdp->sd_logd_waitq);
d6a079e8 108 spin_lock_init(&sdp->sd_ail_lock);
b3b94faa
DT
109 INIT_LIST_HEAD(&sdp->sd_ail1_list);
110 INIT_LIST_HEAD(&sdp->sd_ail2_list);
111
484adff8 112 init_rwsem(&sdp->sd_log_flush_lock);
16615be1
SW
113 atomic_set(&sdp->sd_log_in_flight, 0);
114 init_waitqueue_head(&sdp->sd_log_flush_wait);
b3b94faa
DT
115
116 INIT_LIST_HEAD(&sdp->sd_revoke_list);
117
b3b94faa
DT
118 return sdp;
119}
120
b3b94faa 121
9b8df98f
SW
122/**
123 * gfs2_check_sb - Check superblock
124 * @sdp: the filesystem
125 * @sb: The superblock
126 * @silent: Don't print a message if the check fails
127 *
128 * Checks the version code of the FS is one that we understand how to
129 * read and that the sizes of the various on-disk structures have not
130 * changed.
131 */
132
32e471ef 133static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 134{
32e471ef
SW
135 struct gfs2_sb_host *sb = &sdp->sd_sb;
136
9b8df98f
SW
137 if (sb->sb_magic != GFS2_MAGIC ||
138 sb->sb_type != GFS2_METATYPE_SB) {
139 if (!silent)
140 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
141 return -EINVAL;
142 }
143
144 /* If format numbers match exactly, we're done. */
145
146 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
147 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
148 return 0;
149
c80dbb58 150 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 151
c80dbb58 152 return -EINVAL;
9b8df98f
SW
153}
154
155static void end_bio_io_page(struct bio *bio, int error)
156{
157 struct page *page = bio->bi_private;
158
159 if (!error)
160 SetPageUptodate(page);
161 else
162 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
163 unlock_page(page);
164}
165
32e471ef 166static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 167{
32e471ef
SW
168 struct gfs2_sb_host *sb = &sdp->sd_sb;
169 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
170 const struct gfs2_sb *str = buf;
171
172 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
173 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
174 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
175 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
176 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
177 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
178 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
179 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
180 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
181 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
182 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
183
184 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
185 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 186 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
187}
188
189/**
190 * gfs2_read_super - Read the gfs2 super block from disk
191 * @sdp: The GFS2 super block
192 * @sector: The location of the super block
193 * @error: The error code to return
194 *
195 * This uses the bio functions to read the super block from disk
196 * because we want to be 100% sure that we never read cached data.
197 * A super block is read twice only during each GFS2 mount and is
198 * never written to by the filesystem. The first time its read no
199 * locks are held, and the only details which are looked at are those
200 * relating to the locking protocol. Once locking is up and working,
201 * the sb is read again under the lock to establish the location of
202 * the master directory (contains pointers to journals etc) and the
203 * root directory.
204 *
205 * Returns: 0 on success or error
206 */
207
32e471ef 208static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
209{
210 struct super_block *sb = sdp->sd_vfs;
211 struct gfs2_sb *p;
212 struct page *page;
213 struct bio *bio;
214
215 page = alloc_page(GFP_NOFS);
216 if (unlikely(!page))
217 return -ENOBUFS;
218
219 ClearPageUptodate(page);
220 ClearPageDirty(page);
221 lock_page(page);
222
223 bio = bio_alloc(GFP_NOFS, 1);
9b8df98f
SW
224 bio->bi_sector = sector * (sb->s_blocksize >> 9);
225 bio->bi_bdev = sb->s_bdev;
226 bio_add_page(bio, page, PAGE_SIZE, 0);
227
228 bio->bi_end_io = end_bio_io_page;
229 bio->bi_private = page;
20ed0535 230 submit_bio(READ_SYNC | REQ_META, bio);
9b8df98f
SW
231 wait_on_page_locked(page);
232 bio_put(bio);
233 if (!PageUptodate(page)) {
234 __free_page(page);
235 return -EIO;
236 }
237 p = kmap(page);
32e471ef 238 gfs2_sb_in(sdp, p);
9b8df98f
SW
239 kunmap(page);
240 __free_page(page);
32e471ef 241 return gfs2_check_sb(sdp, silent);
9b8df98f 242}
6bac243f 243
9b8df98f
SW
244/**
245 * gfs2_read_sb - Read super block
246 * @sdp: The GFS2 superblock
9b8df98f
SW
247 * @silent: Don't print message if mount fails
248 *
249 */
250
6bac243f 251static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
252{
253 u32 hash_blocks, ind_blocks, leaf_blocks;
254 u32 tmp_blocks;
255 unsigned int x;
256 int error;
257
32e471ef 258 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
259 if (error) {
260 if (!silent)
261 fs_err(sdp, "can't read superblock\n");
262 return error;
263 }
264
9b8df98f
SW
265 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
266 GFS2_BASIC_BLOCK_SHIFT;
267 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
268 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
269 sizeof(struct gfs2_dinode)) / sizeof(u64);
270 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
271 sizeof(struct gfs2_meta_header)) / sizeof(u64);
272 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
273 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
274 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
275 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
276 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
277 sizeof(struct gfs2_meta_header)) /
278 sizeof(struct gfs2_quota_change);
a68a0a35
BP
279 sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
280 sizeof(struct gfs2_meta_header))
281 * GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
9b8df98f
SW
282
283 /* Compute maximum reservation required to add a entry to a directory */
284
285 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
286 sdp->sd_jbsize);
287
288 ind_blocks = 0;
289 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
290 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
291 ind_blocks += tmp_blocks;
292 }
293
294 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
295
296 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
297
298 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
299 sizeof(struct gfs2_dinode);
300 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
301 for (x = 2;; x++) {
302 u64 space, d;
303 u32 m;
304
305 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
306 d = space;
307 m = do_div(d, sdp->sd_inptrs);
308
309 if (d != sdp->sd_heightsize[x - 1] || m)
310 break;
311 sdp->sd_heightsize[x] = space;
312 }
313 sdp->sd_max_height = x;
314 sdp->sd_heightsize[x] = ~0;
315 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
316
317 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
318 sizeof(struct gfs2_dinode);
319 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
320 for (x = 2;; x++) {
321 u64 space, d;
322 u32 m;
323
324 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
325 d = space;
326 m = do_div(d, sdp->sd_inptrs);
327
328 if (d != sdp->sd_jheightsize[x - 1] || m)
329 break;
330 sdp->sd_jheightsize[x] = space;
331 }
332 sdp->sd_max_jheight = x;
333 sdp->sd_jheightsize[x] = ~0;
334 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
335
336 return 0;
337}
338
b3b94faa
DT
339static int init_names(struct gfs2_sbd *sdp, int silent)
340{
b3b94faa
DT
341 char *proto, *table;
342 int error = 0;
343
344 proto = sdp->sd_args.ar_lockproto;
345 table = sdp->sd_args.ar_locktable;
346
347 /* Try to autodetect */
348
349 if (!proto[0] || !table[0]) {
32e471ef 350 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
351 if (error)
352 return error;
3cf1e7be 353
b3b94faa 354 if (!proto[0])
3cf1e7be 355 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 356 if (!table[0])
3cf1e7be 357 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
358 }
359
360 if (!table[0])
361 table = sdp->sd_vfs->s_id;
362
00377d8e
JD
363 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
364 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 365
5d35e31f
DC
366 table = sdp->sd_table_name;
367 while ((table = strchr(table, '/')))
b35997d4
RP
368 *table = '_';
369
b3b94faa
DT
370 return error;
371}
372
373static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
374 int undo)
375{
b3b94faa
DT
376 int error = 0;
377
378 if (undo)
379 goto fail_trans;
380
b3b94faa
DT
381 error = gfs2_glock_nq_num(sdp,
382 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
383 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
384 mount_gh);
385 if (error) {
386 fs_err(sdp, "can't acquire mount glock: %d\n", error);
387 goto fail;
388 }
389
390 error = gfs2_glock_nq_num(sdp,
391 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
392 LM_ST_SHARED,
579b78a4 393 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
394 &sdp->sd_live_gh);
395 if (error) {
396 fs_err(sdp, "can't acquire live glock: %d\n", error);
397 goto fail_mount;
398 }
399
400 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
401 CREATE, &sdp->sd_rename_gl);
402 if (error) {
403 fs_err(sdp, "can't create rename glock: %d\n", error);
404 goto fail_live;
405 }
406
407 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
408 CREATE, &sdp->sd_trans_gl);
409 if (error) {
410 fs_err(sdp, "can't create transaction glock: %d\n", error);
411 goto fail_rename;
412 }
b3b94faa
DT
413
414 return 0;
415
feaa7bba 416fail_trans:
b3b94faa 417 gfs2_glock_put(sdp->sd_trans_gl);
feaa7bba 418fail_rename:
b3b94faa 419 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 420fail_live:
b3b94faa 421 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 422fail_mount:
b3b94faa 423 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 424fail:
b3b94faa
DT
425 return error;
426}
427
9b8df98f
SW
428static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
429 u64 no_addr, const char *name)
f42faf4f 430{
9b8df98f
SW
431 struct gfs2_sbd *sdp = sb->s_fs_info;
432 struct dentry *dentry;
433 struct inode *inode;
434
44ad37d6 435 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
9b8df98f
SW
436 if (IS_ERR(inode)) {
437 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
438 return PTR_ERR(inode);
439 }
48fde701 440 dentry = d_make_root(inode);
9b8df98f
SW
441 if (!dentry) {
442 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
443 return -ENOMEM;
444 }
9b8df98f
SW
445 *dptr = dentry;
446 return 0;
f42faf4f
SW
447}
448
9b8df98f 449static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
450{
451 struct super_block *sb = sdp->sd_vfs;
452 struct gfs2_holder sb_gh;
dbb7cae2 453 u64 no_addr;
9b8df98f 454 int ret;
b3b94faa 455
9b8df98f
SW
456 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
457 LM_ST_SHARED, 0, &sb_gh);
458 if (ret) {
459 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
460 return ret;
b3b94faa 461 }
907b9bce 462
6bac243f 463 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
464 if (ret) {
465 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
466 goto out;
467 }
468
469 /* Set up the buffer cache and SB for real */
e1defc4f 470 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 471 ret = -EINVAL;
b3b94faa
DT
472 fs_err(sdp, "FS block size (%u) is too small for device "
473 "block size (%u)\n",
e1defc4f 474 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
475 goto out;
476 }
477 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 478 ret = -EINVAL;
b3b94faa
DT
479 fs_err(sdp, "FS block size (%u) is too big for machine "
480 "page size (%u)\n",
481 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
482 goto out;
483 }
b3b94faa
DT
484 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
485
f42faf4f 486 /* Get the root inode */
dbb7cae2 487 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
488 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
489 if (ret)
f42faf4f 490 goto out;
b3b94faa 491
9b8df98f
SW
492 /* Get the master inode */
493 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
494 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
495 if (ret) {
496 dput(sdp->sd_root_dir);
497 goto out;
498 }
499 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 500out:
b3b94faa 501 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 502 return ret;
b3b94faa
DT
503}
504
da6dd40d
BP
505/**
506 * map_journal_extents - create a reusable "extent" mapping from all logical
507 * blocks to all physical blocks for the given journal. This will save
508 * us time when writing journal blocks. Most journals will have only one
509 * extent that maps all their logical blocks. That's because gfs2.mkfs
510 * arranges the journal blocks sequentially to maximize performance.
511 * So the extent would map the first block for the entire file length.
512 * However, gfs2_jadd can happen while file activity is happening, so
513 * those journals may not be sequential. Less likely is the case where
514 * the users created their own journals by mounting the metafs and
515 * laying it out. But it's still possible. These journals might have
516 * several extents.
517 *
518 * TODO: This should be done in bigger chunks rather than one block at a time,
519 * but since it's only done at mount time, I'm not worried about the
520 * time it takes.
521 */
522static int map_journal_extents(struct gfs2_sbd *sdp)
523{
524 struct gfs2_jdesc *jd = sdp->sd_jdesc;
525 unsigned int lb;
526 u64 db, prev_db; /* logical block, disk block, prev disk block */
527 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
528 struct gfs2_journal_extent *jext = NULL;
529 struct buffer_head bh;
530 int rc = 0;
531
da6dd40d
BP
532 prev_db = 0;
533
a2e0f799 534 for (lb = 0; lb < i_size_read(jd->jd_inode) >> sdp->sd_sb.sb_bsize_shift; lb++) {
da6dd40d
BP
535 bh.b_state = 0;
536 bh.b_blocknr = 0;
537 bh.b_size = 1 << ip->i_inode.i_blkbits;
538 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
539 db = bh.b_blocknr;
540 if (rc || !db) {
541 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
542 "%u db=%llu\n", rc, lb, (unsigned long long)db);
543 break;
544 }
545 if (!prev_db || db != prev_db + 1) {
546 jext = kzalloc(sizeof(struct gfs2_journal_extent),
547 GFP_KERNEL);
548 if (!jext) {
549 printk(KERN_INFO "GFS2 error: out of memory "
550 "mapping journal extents.\n");
551 rc = -ENOMEM;
552 break;
553 }
554 jext->dblock = db;
555 jext->lblock = lb;
556 jext->blocks = 1;
557 list_add_tail(&jext->extent_list, &jd->extent_list);
558 } else {
559 jext->blocks++;
560 }
561 prev_db = db;
562 }
563 return rc;
564}
565
f057f6cd 566static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 567{
f057f6cd
SW
568 char *message = "FIRSTMOUNT=Done";
569 char *envp[] = { message, NULL };
e0c2a9aa
DT
570
571 fs_info(sdp, "first mount done, others may mount\n");
572
573 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
574 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
575
f057f6cd 576 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
577}
578
b5289681
SW
579/**
580 * gfs2_jindex_hold - Grab a lock on the jindex
581 * @sdp: The GFS2 superblock
582 * @ji_gh: the holder for the jindex glock
583 *
584 * Returns: errno
585 */
586
587static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
588{
589 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
590 struct qstr name;
591 char buf[20];
592 struct gfs2_jdesc *jd;
593 int error;
594
595 name.name = buf;
596
597 mutex_lock(&sdp->sd_jindex_mutex);
598
599 for (;;) {
600 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
601 if (error)
602 break;
603
604 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
605 name.hash = gfs2_disk_hash(name.name, name.len);
606
607 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
608 if (error == -ENOENT) {
609 error = 0;
610 break;
611 }
612
613 gfs2_glock_dq_uninit(ji_gh);
614
615 if (error)
616 break;
617
618 error = -ENOMEM;
619 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
620 if (!jd)
621 break;
622
623 INIT_LIST_HEAD(&jd->extent_list);
6ecd7c2d 624 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
625 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
626 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
627 if (!jd->jd_inode)
628 error = -ENOENT;
629 else
630 error = PTR_ERR(jd->jd_inode);
631 kfree(jd);
632 break;
633 }
634
635 spin_lock(&sdp->sd_jindex_spin);
636 jd->jd_jid = sdp->sd_journals++;
637 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
638 spin_unlock(&sdp->sd_jindex_spin);
639 }
640
641 mutex_unlock(&sdp->sd_jindex_mutex);
642
643 return error;
644}
645
b3b94faa
DT
646static int init_journal(struct gfs2_sbd *sdp, int undo)
647{
9b8df98f 648 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 649 struct gfs2_holder ji_gh;
5c676f6d 650 struct gfs2_inode *ip;
b3b94faa
DT
651 int jindex = 1;
652 int error = 0;
653
654 if (undo) {
655 jindex = 0;
fe64d517 656 goto fail_jinode_gh;
b3b94faa
DT
657 }
658
9b8df98f 659 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 660 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 661 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 662 return PTR_ERR(sdp->sd_jindex);
b3b94faa 663 }
b3b94faa
DT
664
665 /* Load in the journal index special file */
666
667 error = gfs2_jindex_hold(sdp, &ji_gh);
668 if (error) {
669 fs_err(sdp, "can't read journal index: %d\n", error);
670 goto fail;
671 }
672
0e5a9fb0 673 error = -EUSERS;
b3b94faa
DT
674 if (!gfs2_jindex_size(sdp)) {
675 fs_err(sdp, "no journals!\n");
907b9bce 676 goto fail_jindex;
b3b94faa
DT
677 }
678
679 if (sdp->sd_args.ar_spectator) {
680 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 681 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
682 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
683 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
684 } else {
685 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
686 fs_err(sdp, "can't mount journal #%u\n",
687 sdp->sd_lockstruct.ls_jid);
688 fs_err(sdp, "there are only %u journals (0 - %u)\n",
689 gfs2_jindex_size(sdp),
690 gfs2_jindex_size(sdp) - 1);
691 goto fail_jindex;
692 }
693 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
694
feaa7bba 695 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
696 &gfs2_journal_glops,
697 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
698 &sdp->sd_journal_gh);
699 if (error) {
700 fs_err(sdp, "can't acquire journal glock: %d\n", error);
701 goto fail_jindex;
702 }
703
feaa7bba
SW
704 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
705 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 706 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
707 &sdp->sd_jinode_gh);
708 if (error) {
709 fs_err(sdp, "can't acquire journal inode glock: %d\n",
710 error);
711 goto fail_journal_gh;
712 }
713
714 error = gfs2_jdesc_check(sdp->sd_jdesc);
715 if (error) {
716 fs_err(sdp, "my journal (%u) is bad: %d\n",
717 sdp->sd_jdesc->jd_jid, error);
718 goto fail_jinode_gh;
719 }
fd041f0b 720 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
721 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
722 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
723
724 /* Map the extents for this journal's blocks */
725 map_journal_extents(sdp);
b3b94faa 726 }
63997775 727 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
728
729 if (sdp->sd_lockstruct.ls_first) {
730 unsigned int x;
731 for (x = 0; x < sdp->sd_journals; x++) {
6ecd7c2d
TH
732 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
733 true);
b3b94faa
DT
734 if (error) {
735 fs_err(sdp, "error recovering journal %u: %d\n",
736 x, error);
737 goto fail_jinode_gh;
738 }
739 }
740
f057f6cd 741 gfs2_others_may_mount(sdp);
b3b94faa 742 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 743 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
744 if (error) {
745 fs_err(sdp, "error recovering my journal: %d\n", error);
746 goto fail_jinode_gh;
747 }
748 }
749
750 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
751 gfs2_glock_dq_uninit(&ji_gh);
752 jindex = 0;
753
b3b94faa
DT
754 return 0;
755
a91ea69f 756fail_jinode_gh:
b3b94faa
DT
757 if (!sdp->sd_args.ar_spectator)
758 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 759fail_journal_gh:
b3b94faa
DT
760 if (!sdp->sd_args.ar_spectator)
761 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 762fail_jindex:
b3b94faa
DT
763 gfs2_jindex_free(sdp);
764 if (jindex)
765 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 766fail:
f42faf4f 767 iput(sdp->sd_jindex);
b3b94faa
DT
768 return error;
769}
770
56aa72d0 771static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
772
773static int init_inodes(struct gfs2_sbd *sdp, int undo)
774{
b3b94faa 775 int error = 0;
9b8df98f 776 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
777
778 if (undo)
f42faf4f
SW
779 goto fail_qinode;
780
f42faf4f
SW
781 error = init_journal(sdp, undo);
782 if (error)
9b8df98f 783 goto fail;
b3b94faa 784
b3b94faa 785 /* Read in the master statfs inode */
9b8df98f 786 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
787 if (IS_ERR(sdp->sd_statfs_inode)) {
788 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 789 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 790 goto fail_journal;
b3b94faa
DT
791 }
792
793 /* Read in the resource index inode */
9b8df98f 794 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
795 if (IS_ERR(sdp->sd_rindex)) {
796 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
797 fs_err(sdp, "can't get resource index inode: %d\n", error);
798 goto fail_statfs;
799 }
cf45b752 800 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
801
802 /* Read in the quota inode */
9b8df98f 803 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
804 if (IS_ERR(sdp->sd_quota_inode)) {
805 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
806 fs_err(sdp, "can't get quota file inode: %d\n", error);
807 goto fail_rindex;
808 }
56aa72d0
JK
809 /*
810 * i_mutex on quota files is special. Since this inode is hidden system
811 * file, we are safe to define locking ourselves.
812 */
813 lockdep_set_class(&sdp->sd_quota_inode->i_mutex,
814 &gfs2_quota_imutex_key);
a365fbf3
SW
815
816 error = gfs2_rindex_update(sdp);
817 if (error)
818 goto fail_qinode;
819
b3b94faa
DT
820 return 0;
821
f42faf4f
SW
822fail_qinode:
823 iput(sdp->sd_quota_inode);
f42faf4f 824fail_rindex:
b3b94faa 825 gfs2_clear_rgrpd(sdp);
f42faf4f 826 iput(sdp->sd_rindex);
f42faf4f
SW
827fail_statfs:
828 iput(sdp->sd_statfs_inode);
f42faf4f
SW
829fail_journal:
830 init_journal(sdp, UNDO);
f42faf4f 831fail:
b3b94faa
DT
832 return error;
833}
834
835static int init_per_node(struct gfs2_sbd *sdp, int undo)
836{
f42faf4f 837 struct inode *pn = NULL;
b3b94faa
DT
838 char buf[30];
839 int error = 0;
5c676f6d 840 struct gfs2_inode *ip;
9b8df98f 841 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
842
843 if (sdp->sd_args.ar_spectator)
844 return 0;
845
846 if (undo)
847 goto fail_qc_gh;
848
9b8df98f 849 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
850 if (IS_ERR(pn)) {
851 error = PTR_ERR(pn);
b3b94faa
DT
852 fs_err(sdp, "can't find per_node directory: %d\n", error);
853 return error;
854 }
855
b3b94faa 856 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
857 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
858 if (IS_ERR(sdp->sd_sc_inode)) {
859 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 860 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 861 goto fail;
b3b94faa
DT
862 }
863
b3b94faa 864 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
865 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
866 if (IS_ERR(sdp->sd_qc_inode)) {
867 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
868 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
869 goto fail_ut_i;
870 }
871
f42faf4f 872 iput(pn);
b3b94faa
DT
873 pn = NULL;
874
feaa7bba 875 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 876 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
877 &sdp->sd_sc_gh);
878 if (error) {
879 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 880 goto fail_qc_i;
b3b94faa
DT
881 }
882
feaa7bba 883 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 884 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
885 &sdp->sd_qc_gh);
886 if (error) {
887 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
888 goto fail_ut_gh;
889 }
890
891 return 0;
892
a91ea69f 893fail_qc_gh:
b3b94faa 894 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 895fail_ut_gh:
b3b94faa 896 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 897fail_qc_i:
f42faf4f 898 iput(sdp->sd_qc_inode);
a91ea69f 899fail_ut_i:
f42faf4f 900 iput(sdp->sd_sc_inode);
a91ea69f 901fail:
b3b94faa 902 if (pn)
f42faf4f 903 iput(pn);
b3b94faa
DT
904 return error;
905}
906
907static int init_threads(struct gfs2_sbd *sdp, int undo)
908{
909 struct task_struct *p;
910 int error = 0;
911
912 if (undo)
feaa7bba 913 goto fail_quotad;
b3b94faa 914
b3b94faa
DT
915 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
916 error = IS_ERR(p);
917 if (error) {
918 fs_err(sdp, "can't start logd thread: %d\n", error);
919 return error;
920 }
921 sdp->sd_logd_process = p;
922
b3b94faa
DT
923 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
924 error = IS_ERR(p);
925 if (error) {
926 fs_err(sdp, "can't start quotad thread: %d\n", error);
927 goto fail;
928 }
929 sdp->sd_quotad_process = p;
930
b3b94faa
DT
931 return 0;
932
b3b94faa 933
feaa7bba 934fail_quotad:
b3b94faa 935 kthread_stop(sdp->sd_quotad_process);
feaa7bba 936fail:
b3b94faa 937 kthread_stop(sdp->sd_logd_process);
b3b94faa
DT
938 return error;
939}
da755fdb 940
f057f6cd
SW
941static const match_table_t nolock_tokens = {
942 { Opt_jid, "jid=%d\n", },
943 { Opt_err, NULL },
944};
945
946static const struct lm_lockops nolock_ops = {
947 .lm_proto_name = "lock_nolock",
fc0e38da 948 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
949 .lm_tokens = &nolock_tokens,
950};
951
da755fdb
SW
952/**
953 * gfs2_lm_mount - mount a locking protocol
954 * @sdp: the filesystem
3ad2f3fb 955 * @args: mount arguments
da755fdb
SW
956 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
957 *
958 * Returns: errno
959 */
960
961static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
962{
f057f6cd
SW
963 const struct lm_lockops *lm;
964 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
965 struct gfs2_args *args = &sdp->sd_args;
966 const char *proto = sdp->sd_proto_name;
967 const char *table = sdp->sd_table_name;
f057f6cd
SW
968 char *o, *options;
969 int ret;
da755fdb 970
f057f6cd
SW
971 if (!strcmp("lock_nolock", proto)) {
972 lm = &nolock_ops;
973 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
974#ifdef CONFIG_GFS2_FS_LOCKING_DLM
975 } else if (!strcmp("lock_dlm", proto)) {
976 lm = &gfs2_dlm_ops;
977#endif
978 } else {
979 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
980 return -ENOENT;
981 }
da755fdb
SW
982
983 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
984
f057f6cd
SW
985 ls->ls_ops = lm;
986 ls->ls_first = 1;
da755fdb 987
f057f6cd
SW
988 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
989 substring_t tmp[MAX_OPT_ARGS];
990 int token, option;
991
992 if (!o || !*o)
993 continue;
994
995 token = match_token(o, *lm->lm_tokens, tmp);
996 switch (token) {
997 case Opt_jid:
998 ret = match_int(&tmp[0], &option);
999 if (ret || option < 0)
1000 goto hostdata_error;
ba6e9364
SW
1001 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
1002 ls->ls_jid = option;
f057f6cd
SW
1003 break;
1004 case Opt_id:
4875647a 1005 case Opt_nodir:
2b88f7c5 1006 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
1007 break;
1008 case Opt_first:
1009 ret = match_int(&tmp[0], &option);
1010 if (ret || (option != 0 && option != 1))
1011 goto hostdata_error;
1012 ls->ls_first = option;
1013 break;
f057f6cd
SW
1014 case Opt_err:
1015 default:
1016hostdata_error:
1017 fs_info(sdp, "unknown hostdata (%s)\n", o);
1018 return -EINVAL;
1019 }
da755fdb
SW
1020 }
1021
f057f6cd
SW
1022 if (lm->lm_mount == NULL) {
1023 fs_info(sdp, "Now mounting FS...\n");
19237039 1024 complete_all(&sdp->sd_locking_init);
f057f6cd 1025 return 0;
da755fdb 1026 }
e0c2a9aa 1027 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1028 if (ret == 0)
1029 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1030 complete_all(&sdp->sd_locking_init);
f057f6cd 1031 return ret;
da755fdb
SW
1032}
1033
1034void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1035{
f057f6cd
SW
1036 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1037 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1038 lm->lm_unmount)
1039 lm->lm_unmount(sdp);
da755fdb 1040}
b3b94faa 1041
ba6e9364
SW
1042static int gfs2_journalid_wait(void *word)
1043{
1044 if (signal_pending(current))
1045 return -EINTR;
1046 schedule();
1047 return 0;
1048}
1049
1050static int wait_on_journal(struct gfs2_sbd *sdp)
1051{
ba6e9364
SW
1052 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1053 return 0;
1054
1055 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1056}
1057
8633ecfa
SW
1058void gfs2_online_uevent(struct gfs2_sbd *sdp)
1059{
1060 struct super_block *sb = sdp->sd_vfs;
1061 char ro[20];
1062 char spectator[20];
1063 char *envp[] = { ro, spectator, NULL };
1064 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1065 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1066 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1067}
1068
b3b94faa
DT
1069/**
1070 * fill_super - Read in superblock
1071 * @sb: The VFS superblock
1072 * @data: Mount options
1073 * @silent: Don't complain if it's not a GFS2 filesystem
1074 *
1075 * Returns: errno
1076 */
1077
f55073ff 1078static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1079{
1080 struct gfs2_sbd *sdp;
1081 struct gfs2_holder mount_gh;
1082 int error;
1083
1084 sdp = init_sbd(sb);
1085 if (!sdp) {
d92a8d48 1086 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1087 return -ENOMEM;
1088 }
f55073ff 1089 sdp->sd_args = *args;
b3b94faa 1090
fe64d517 1091 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1092 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1093 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1094 }
6f04c1c7
SW
1095 if (sdp->sd_args.ar_posix_acl)
1096 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1097 if (sdp->sd_args.ar_nobarrier)
1098 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1099
9964afbb 1100 sb->s_flags |= MS_NOSEC;
719ee344
SW
1101 sb->s_magic = GFS2_MAGIC;
1102 sb->s_op = &gfs2_super_ops;
41ced6dc 1103 sb->s_d_op = &gfs2_dops;
719ee344 1104 sb->s_export_op = &gfs2_export_ops;
40b78a32 1105 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1106 sb->s_qcop = &gfs2_quotactl_ops;
1107 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1108 sb->s_time_gran = 1;
1109 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1110
1111 /* Set up the buffer cache and fill in some fake block size values
1112 to allow us to read-in the on-disk superblock. */
1113 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1114 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1115 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1116 GFS2_BASIC_BLOCK_SHIFT;
1117 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1118
5e687eac 1119 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1120 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1121 if (sdp->sd_args.ar_statfs_quantum) {
1122 sdp->sd_tune.gt_statfs_slow = 0;
1123 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1124 } else {
3d3c10f2
BM
1125 sdp->sd_tune.gt_statfs_slow = 1;
1126 sdp->sd_tune.gt_statfs_quantum = 30;
1127 }
48c2b613 1128
b3b94faa 1129 error = init_names(sdp, silent);
0d515210
BP
1130 if (error) {
1131 /* In this case, we haven't initialized sysfs, so we have to
1132 manually free the sdp. */
1133 free_percpu(sdp->sd_lkstats);
1134 kfree(sdp);
1135 sb->s_fs_info = NULL;
1136 return error;
1137 }
b3b94faa 1138
e0c2a9aa
DT
1139 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1140
b3b94faa 1141 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1142 /*
1143 * If we hit an error here, gfs2_sys_fs_add will have called function
1144 * kobject_put which causes the sysfs usage count to go to zero, which
1145 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1146 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1147 * kobject_put to free sdp.
1148 */
b3b94faa 1149 if (error)
0d515210
BP
1150 return error;
1151
1152 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1153
1154 error = gfs2_lm_mount(sdp, silent);
1155 if (error)
0d515210 1156 goto fail_debug;
b3b94faa
DT
1157
1158 error = init_locking(sdp, &mount_gh, DO);
1159 if (error)
1160 goto fail_lm;
1161
9b8df98f 1162 error = init_sb(sdp, silent);
b3b94faa
DT
1163 if (error)
1164 goto fail_locking;
b3b94faa 1165
ba6e9364
SW
1166 error = wait_on_journal(sdp);
1167 if (error)
1168 goto fail_sb;
1169
feb47ca9
SW
1170 /*
1171 * If user space has failed to join the cluster or some similar
1172 * failure has occurred, then the journal id will contain a
1173 * negative (error) number. This will then be returned to the
1174 * caller (of the mount syscall). We do this even for spectator
1175 * mounts (which just write a jid of 0 to indicate "ok" even though
1176 * the jid is unused in the spectator case)
1177 */
1178 if (sdp->sd_lockstruct.ls_jid < 0) {
1179 error = sdp->sd_lockstruct.ls_jid;
1180 sdp->sd_lockstruct.ls_jid = 0;
1181 goto fail_sb;
1182 }
1183
e0c2a9aa
DT
1184 if (sdp->sd_args.ar_spectator)
1185 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1186 sdp->sd_table_name);
1187 else
1188 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1189 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1190
b3b94faa
DT
1191 error = init_inodes(sdp, DO);
1192 if (error)
f42faf4f 1193 goto fail_sb;
b3b94faa
DT
1194
1195 error = init_per_node(sdp, DO);
1196 if (error)
1197 goto fail_inodes;
1198
1199 error = gfs2_statfs_init(sdp);
1200 if (error) {
1201 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1202 goto fail_per_node;
1203 }
1204
1205 error = init_threads(sdp, DO);
1206 if (error)
1207 goto fail_per_node;
1208
1209 if (!(sb->s_flags & MS_RDONLY)) {
1210 error = gfs2_make_fs_rw(sdp);
1211 if (error) {
1212 fs_err(sdp, "can't make FS RW: %d\n", error);
1213 goto fail_threads;
1214 }
1215 }
1216
1217 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1218 gfs2_online_uevent(sdp);
b3b94faa
DT
1219 return 0;
1220
a91ea69f 1221fail_threads:
b3b94faa 1222 init_threads(sdp, UNDO);
a91ea69f 1223fail_per_node:
b3b94faa 1224 init_per_node(sdp, UNDO);
a91ea69f 1225fail_inodes:
b3b94faa 1226 init_inodes(sdp, UNDO);
a91ea69f 1227fail_sb:
9b8df98f
SW
1228 if (sdp->sd_root_dir)
1229 dput(sdp->sd_root_dir);
1230 if (sdp->sd_master_dir)
1231 dput(sdp->sd_master_dir);
e7c8707e
SW
1232 if (sb->s_root)
1233 dput(sb->s_root);
9b8df98f 1234 sb->s_root = NULL;
a91ea69f 1235fail_locking:
b3b94faa 1236 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1237fail_lm:
fefc03bf 1238 gfs2_gl_hash_clear(sdp);
b3b94faa 1239 gfs2_lm_unmount(sdp);
0d515210 1240fail_debug:
5f882096 1241 gfs2_delete_debugfs_file(sdp);
a245769f 1242 free_percpu(sdp->sd_lkstats);
0d515210
BP
1243 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1244 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1245 gfs2_sys_fs_del(sdp);
5c676f6d 1246 sb->s_fs_info = NULL;
b3b94faa
DT
1247 return error;
1248}
1249
f55073ff 1250static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1251{
f55073ff
SW
1252 s->s_bdev = data;
1253 s->s_dev = s->s_bdev->bd_dev;
1254
1255 /*
1256 * We set the bdi here to the queue backing, file systems can
1257 * overwrite this in ->fill_super()
1258 */
1259 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1260 return 0;
86384605
AD
1261}
1262
f55073ff 1263static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1264{
1265 struct block_device *bdev = ptr;
1266 return (bdev == s->s_bdev);
1267}
1268
f55073ff 1269/**
8bcbbf00 1270 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1271 * @fs_type: The GFS2 filesystem type
1272 * @flags: Mount flags
1273 * @dev_name: The name of the device
1274 * @data: The mount arguments
f55073ff
SW
1275 *
1276 * Q. Why not use get_sb_bdev() ?
1277 * A. We need to select one of two root directories to mount, independent
1278 * of whether this is the initial, or subsequent, mount of this sb
1279 *
1280 * Returns: 0 or -ve on error
1281 */
1282
8bcbbf00
AV
1283static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1284 const char *dev_name, void *data)
f55073ff
SW
1285{
1286 struct block_device *bdev;
1287 struct super_block *s;
d4d77629 1288 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1289 int error;
1290 struct gfs2_args args;
1291 struct gfs2_sbd *sdp;
1292
1293 if (!(flags & MS_RDONLY))
1294 mode |= FMODE_WRITE;
1295
d4d77629 1296 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1297 if (IS_ERR(bdev))
8bcbbf00 1298 return ERR_CAST(bdev);
f55073ff
SW
1299
1300 /*
1301 * once the super is inserted into the list by sget, s_umount
1302 * will protect the lockfs code from trying to start a snapshot
1303 * while we are mounting
1304 */
1305 mutex_lock(&bdev->bd_fsfreeze_mutex);
1306 if (bdev->bd_fsfreeze_count > 0) {
1307 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1308 error = -EBUSY;
1309 goto error_bdev;
1310 }
9249e17f 1311 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1312 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1313 error = PTR_ERR(s);
1314 if (IS_ERR(s))
1315 goto error_bdev;
1316
8bcbbf00 1317 if (s->s_root)
d4d77629 1318 blkdev_put(bdev, mode);
8bcbbf00 1319
f55073ff
SW
1320 memset(&args, 0, sizeof(args));
1321 args.ar_quota = GFS2_QUOTA_DEFAULT;
1322 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1323 args.ar_commit = 30;
3d3c10f2
BM
1324 args.ar_statfs_quantum = 30;
1325 args.ar_quota_quantum = 60;
f55073ff
SW
1326 args.ar_errors = GFS2_ERRORS_DEFAULT;
1327
1328 error = gfs2_mount_args(&args, data);
1329 if (error) {
1330 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
8bcbbf00 1331 goto error_super;
f55073ff
SW
1332 }
1333
1334 if (s->s_root) {
1335 error = -EBUSY;
1336 if ((flags ^ s->s_flags) & MS_RDONLY)
1337 goto error_super;
f55073ff
SW
1338 } else {
1339 char b[BDEVNAME_SIZE];
1340
f55073ff
SW
1341 s->s_mode = mode;
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);
8bcbbf00
AV
1345 if (error)
1346 goto error_super;
f55073ff
SW
1347 s->s_flags |= MS_ACTIVE;
1348 bdev->bd_super = s;
1349 }
1350
1351 sdp = s->s_fs_info;
f55073ff 1352 if (args.ar_meta)
8bcbbf00 1353 return dget(sdp->sd_master_dir);
f55073ff 1354 else
8bcbbf00 1355 return dget(sdp->sd_root_dir);
f55073ff
SW
1356
1357error_super:
1358 deactivate_locked_super(s);
8bcbbf00 1359 return ERR_PTR(error);
f55073ff 1360error_bdev:
d4d77629 1361 blkdev_put(bdev, mode);
8bcbbf00 1362 return ERR_PTR(error);
f55073ff
SW
1363}
1364
f6d03139
SW
1365static int set_meta_super(struct super_block *s, void *ptr)
1366{
1367 return -EINVAL;
1368}
1369
8bcbbf00
AV
1370static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1371 int flags, const char *dev_name, void *data)
86384605 1372{
f6d03139 1373 struct super_block *s;
003dec89 1374 struct gfs2_sbd *sdp;
e24977d4 1375 struct path path;
86384605 1376 int error;
907b9bce 1377
e24977d4 1378 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1379 if (error) {
9b8df98f
SW
1380 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1381 dev_name, error);
8bcbbf00 1382 return ERR_PTR(error);
86384605 1383 }
9249e17f 1384 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
f6d03139 1385 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1386 path_put(&path);
f6d03139 1387 if (IS_ERR(s)) {
86384605 1388 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
8bcbbf00 1389 return ERR_CAST(s);
86384605 1390 }
f55073ff
SW
1391 if ((flags ^ s->s_flags) & MS_RDONLY) {
1392 deactivate_locked_super(s);
8bcbbf00 1393 return ERR_PTR(-EBUSY);
f55073ff 1394 }
f6d03139 1395 sdp = s->s_fs_info;
8bcbbf00 1396 return dget(sdp->sd_master_dir);
b3b94faa
DT
1397}
1398
419c93e0 1399static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1400{
1401 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1402
1403 if (sdp == NULL) {
1404 kill_block_super(sb);
1405 return;
3af165ac 1406 }
88a19ad0
SW
1407
1408 gfs2_meta_syncfs(sdp);
1409 dput(sdp->sd_root_dir);
1410 dput(sdp->sd_master_dir);
1411 sdp->sd_root_dir = NULL;
1412 sdp->sd_master_dir = NULL;
fefc03bf 1413 shrink_dcache_sb(sb);
88a19ad0 1414 gfs2_delete_debugfs_file(sdp);
a245769f 1415 free_percpu(sdp->sd_lkstats);
0d515210 1416 kill_block_super(sb);
86384605
AD
1417}
1418
b3b94faa
DT
1419struct file_system_type gfs2_fs_type = {
1420 .name = "gfs2",
1421 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1422 .mount = gfs2_mount,
419c93e0
SW
1423 .kill_sb = gfs2_kill_sb,
1424 .owner = THIS_MODULE,
1425};
1426
1427struct file_system_type gfs2meta_fs_type = {
1428 .name = "gfs2meta",
1429 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1430 .mount = gfs2_mount_meta,
b3b94faa
DT
1431 .owner = THIS_MODULE,
1432};
1433