]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/gfs2/ops_fstype.c
Merge branch 'acpica-gpe' into release
[net-next-2.6.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>
86384605
AD
17#include <linux/namei.h>
18#include <linux/mount.h>
5c676f6d 19#include <linux/gfs2_ondisk.h>
fe64d517 20#include <linux/slow-work.h>
cc632e7f 21#include <linux/quotaops.h>
b3b94faa
DT
22
23#include "gfs2.h"
5c676f6d 24#include "incore.h"
da6dd40d 25#include "bmap.h"
b3b94faa
DT
26#include "glock.h"
27#include "glops.h"
28#include "inode.h"
b3b94faa
DT
29#include "recovery.h"
30#include "rgrp.h"
31#include "super.h"
b3b94faa 32#include "sys.h"
5c676f6d 33#include "util.h"
bb3b0e3d 34#include "log.h"
9ac1b4d9 35#include "quota.h"
b5289681 36#include "dir.h"
63997775 37#include "trace_gfs2.h"
b3b94faa
DT
38
39#define DO 0
40#define UNDO 1
41
9b8df98f
SW
42static const u32 gfs2_old_fs_formats[] = {
43 0
44};
45
46static const u32 gfs2_old_multihost_formats[] = {
47 0
48};
49
50/**
51 * gfs2_tune_init - Fill a gfs2_tune structure with default values
52 * @gt: tune
53 *
54 */
55
56static void gfs2_tune_init(struct gfs2_tune *gt)
57{
58 spin_lock_init(&gt->gt_spin);
59
9b8df98f
SW
60 gt->gt_quota_simul_sync = 64;
61 gt->gt_quota_warn_period = 10;
62 gt->gt_quota_scale_num = 1;
63 gt->gt_quota_scale_den = 1;
9b8df98f
SW
64 gt->gt_new_files_jdata = 0;
65 gt->gt_max_readahead = 1 << 18;
9b8df98f 66 gt->gt_complain_secs = 10;
9b8df98f
SW
67}
68
b3b94faa
DT
69static struct gfs2_sbd *init_sbd(struct super_block *sb)
70{
71 struct gfs2_sbd *sdp;
b3b94faa 72
85d1da67 73 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
74 if (!sdp)
75 return NULL;
76
5c676f6d 77 sb->s_fs_info = sdp;
b3b94faa
DT
78 sdp->sd_vfs = sb;
79
80 gfs2_tune_init(&sdp->sd_tune);
81
e402746a
SW
82 init_waitqueue_head(&sdp->sd_glock_wait);
83 atomic_set(&sdp->sd_glock_disposal, 0);
b3b94faa 84 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
85
86 spin_lock_init(&sdp->sd_rindex_spin);
f55ab26a 87 mutex_init(&sdp->sd_rindex_mutex);
b3b94faa
DT
88 INIT_LIST_HEAD(&sdp->sd_rindex_list);
89 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
b3b94faa
DT
90
91 INIT_LIST_HEAD(&sdp->sd_jindex_list);
92 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 93 mutex_init(&sdp->sd_jindex_mutex);
b3b94faa 94
b3b94faa 95 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 96 mutex_init(&sdp->sd_quota_mutex);
37b2c837 97 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
98 INIT_LIST_HEAD(&sdp->sd_trunc_list);
99 spin_lock_init(&sdp->sd_trunc_lock);
b3b94faa
DT
100
101 spin_lock_init(&sdp->sd_log_lock);
5e687eac 102 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa
DT
103 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
104 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
105 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
106 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
d7b616e2 107 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
b3b94faa 108
5e687eac
BM
109 init_waitqueue_head(&sdp->sd_log_waitq);
110 init_waitqueue_head(&sdp->sd_logd_waitq);
b3b94faa
DT
111 INIT_LIST_HEAD(&sdp->sd_ail1_list);
112 INIT_LIST_HEAD(&sdp->sd_ail2_list);
113
484adff8 114 init_rwsem(&sdp->sd_log_flush_lock);
16615be1
SW
115 atomic_set(&sdp->sd_log_in_flight, 0);
116 init_waitqueue_head(&sdp->sd_log_flush_wait);
b3b94faa
DT
117
118 INIT_LIST_HEAD(&sdp->sd_revoke_list);
119
f55ab26a 120 mutex_init(&sdp->sd_freeze_lock);
b3b94faa
DT
121
122 return sdp;
123}
124
b3b94faa 125
9b8df98f
SW
126/**
127 * gfs2_check_sb - Check superblock
128 * @sdp: the filesystem
129 * @sb: The superblock
130 * @silent: Don't print a message if the check fails
131 *
132 * Checks the version code of the FS is one that we understand how to
133 * read and that the sizes of the various on-disk structures have not
134 * changed.
135 */
136
137static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
138{
139 unsigned int x;
140
141 if (sb->sb_magic != GFS2_MAGIC ||
142 sb->sb_type != GFS2_METATYPE_SB) {
143 if (!silent)
144 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
145 return -EINVAL;
146 }
147
148 /* If format numbers match exactly, we're done. */
149
150 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
151 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
152 return 0;
153
154 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
155 for (x = 0; gfs2_old_fs_formats[x]; x++)
156 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
157 break;
158
159 if (!gfs2_old_fs_formats[x]) {
160 printk(KERN_WARNING
161 "GFS2: code version (%u, %u) is incompatible "
162 "with ondisk format (%u, %u)\n",
163 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
164 sb->sb_fs_format, sb->sb_multihost_format);
165 printk(KERN_WARNING
166 "GFS2: I don't know how to upgrade this FS\n");
167 return -EINVAL;
168 }
169 }
170
171 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
172 for (x = 0; gfs2_old_multihost_formats[x]; x++)
173 if (gfs2_old_multihost_formats[x] ==
174 sb->sb_multihost_format)
175 break;
176
177 if (!gfs2_old_multihost_formats[x]) {
178 printk(KERN_WARNING
179 "GFS2: code version (%u, %u) is incompatible "
180 "with ondisk format (%u, %u)\n",
181 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
182 sb->sb_fs_format, sb->sb_multihost_format);
183 printk(KERN_WARNING
184 "GFS2: I don't know how to upgrade this FS\n");
185 return -EINVAL;
186 }
187 }
188
189 if (!sdp->sd_args.ar_upgrade) {
190 printk(KERN_WARNING
191 "GFS2: code version (%u, %u) is incompatible "
192 "with ondisk format (%u, %u)\n",
193 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
194 sb->sb_fs_format, sb->sb_multihost_format);
195 printk(KERN_INFO
196 "GFS2: Use the \"upgrade\" mount option to upgrade "
197 "the FS\n");
198 printk(KERN_INFO "GFS2: See the manual for more details\n");
199 return -EINVAL;
200 }
201
202 return 0;
203}
204
205static void end_bio_io_page(struct bio *bio, int error)
206{
207 struct page *page = bio->bi_private;
208
209 if (!error)
210 SetPageUptodate(page);
211 else
212 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
213 unlock_page(page);
214}
215
216static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
217{
218 const struct gfs2_sb *str = buf;
219
220 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
221 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
222 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
223 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
224 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
225 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
226 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
227 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
228 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
229 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
230 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
231
232 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
233 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
02e3cc70 234 memcpy(sb->sb_uuid, str->sb_uuid, 16);
9b8df98f
SW
235}
236
237/**
238 * gfs2_read_super - Read the gfs2 super block from disk
239 * @sdp: The GFS2 super block
240 * @sector: The location of the super block
241 * @error: The error code to return
242 *
243 * This uses the bio functions to read the super block from disk
244 * because we want to be 100% sure that we never read cached data.
245 * A super block is read twice only during each GFS2 mount and is
246 * never written to by the filesystem. The first time its read no
247 * locks are held, and the only details which are looked at are those
248 * relating to the locking protocol. Once locking is up and working,
249 * the sb is read again under the lock to establish the location of
250 * the master directory (contains pointers to journals etc) and the
251 * root directory.
252 *
253 * Returns: 0 on success or error
254 */
255
256static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
257{
258 struct super_block *sb = sdp->sd_vfs;
259 struct gfs2_sb *p;
260 struct page *page;
261 struct bio *bio;
262
263 page = alloc_page(GFP_NOFS);
264 if (unlikely(!page))
265 return -ENOBUFS;
266
267 ClearPageUptodate(page);
268 ClearPageDirty(page);
269 lock_page(page);
270
271 bio = bio_alloc(GFP_NOFS, 1);
9b8df98f
SW
272 bio->bi_sector = sector * (sb->s_blocksize >> 9);
273 bio->bi_bdev = sb->s_bdev;
274 bio_add_page(bio, page, PAGE_SIZE, 0);
275
276 bio->bi_end_io = end_bio_io_page;
277 bio->bi_private = page;
278 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
279 wait_on_page_locked(page);
280 bio_put(bio);
281 if (!PageUptodate(page)) {
282 __free_page(page);
283 return -EIO;
284 }
285 p = kmap(page);
286 gfs2_sb_in(&sdp->sd_sb, p);
287 kunmap(page);
288 __free_page(page);
289 return 0;
290}
6bac243f 291
9b8df98f
SW
292/**
293 * gfs2_read_sb - Read super block
294 * @sdp: The GFS2 superblock
9b8df98f
SW
295 * @silent: Don't print message if mount fails
296 *
297 */
298
6bac243f 299static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
300{
301 u32 hash_blocks, ind_blocks, leaf_blocks;
302 u32 tmp_blocks;
303 unsigned int x;
304 int error;
305
306 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
307 if (error) {
308 if (!silent)
309 fs_err(sdp, "can't read superblock\n");
310 return error;
311 }
312
313 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
314 if (error)
315 return error;
316
317 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
318 GFS2_BASIC_BLOCK_SHIFT;
319 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
320 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
321 sizeof(struct gfs2_dinode)) / sizeof(u64);
322 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
323 sizeof(struct gfs2_meta_header)) / sizeof(u64);
324 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
325 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
326 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
327 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
328 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
329 sizeof(struct gfs2_meta_header)) /
330 sizeof(struct gfs2_quota_change);
331
332 /* Compute maximum reservation required to add a entry to a directory */
333
334 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
335 sdp->sd_jbsize);
336
337 ind_blocks = 0;
338 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
339 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
340 ind_blocks += tmp_blocks;
341 }
342
343 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
344
345 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
346
347 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
348 sizeof(struct gfs2_dinode);
349 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
350 for (x = 2;; x++) {
351 u64 space, d;
352 u32 m;
353
354 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
355 d = space;
356 m = do_div(d, sdp->sd_inptrs);
357
358 if (d != sdp->sd_heightsize[x - 1] || m)
359 break;
360 sdp->sd_heightsize[x] = space;
361 }
362 sdp->sd_max_height = x;
363 sdp->sd_heightsize[x] = ~0;
364 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
365
366 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
367 sizeof(struct gfs2_dinode);
368 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
369 for (x = 2;; x++) {
370 u64 space, d;
371 u32 m;
372
373 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
374 d = space;
375 m = do_div(d, sdp->sd_inptrs);
376
377 if (d != sdp->sd_jheightsize[x - 1] || m)
378 break;
379 sdp->sd_jheightsize[x] = space;
380 }
381 sdp->sd_max_jheight = x;
382 sdp->sd_jheightsize[x] = ~0;
383 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
384
385 return 0;
386}
387
b3b94faa
DT
388static int init_names(struct gfs2_sbd *sdp, int silent)
389{
b3b94faa
DT
390 char *proto, *table;
391 int error = 0;
392
393 proto = sdp->sd_args.ar_lockproto;
394 table = sdp->sd_args.ar_locktable;
395
396 /* Try to autodetect */
397
398 if (!proto[0] || !table[0]) {
bb8d8a6f
SW
399 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
400 if (error)
401 return error;
3cf1e7be
SW
402
403 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
b3b94faa
DT
404 if (error)
405 goto out;
406
407 if (!proto[0])
3cf1e7be 408 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 409 if (!table[0])
3cf1e7be 410 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
411 }
412
413 if (!table[0])
414 table = sdp->sd_vfs->s_id;
415
00377d8e
JD
416 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
417 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 418
5d35e31f
DC
419 table = sdp->sd_table_name;
420 while ((table = strchr(table, '/')))
b35997d4
RP
421 *table = '_';
422
a91ea69f 423out:
b3b94faa
DT
424 return error;
425}
426
427static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
428 int undo)
429{
b3b94faa
DT
430 int error = 0;
431
432 if (undo)
433 goto fail_trans;
434
b3b94faa
DT
435 error = gfs2_glock_nq_num(sdp,
436 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
437 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
438 mount_gh);
439 if (error) {
440 fs_err(sdp, "can't acquire mount glock: %d\n", error);
441 goto fail;
442 }
443
444 error = gfs2_glock_nq_num(sdp,
445 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
446 LM_ST_SHARED,
579b78a4 447 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
448 &sdp->sd_live_gh);
449 if (error) {
450 fs_err(sdp, "can't acquire live glock: %d\n", error);
451 goto fail_mount;
452 }
453
454 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
455 CREATE, &sdp->sd_rename_gl);
456 if (error) {
457 fs_err(sdp, "can't create rename glock: %d\n", error);
458 goto fail_live;
459 }
460
461 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
462 CREATE, &sdp->sd_trans_gl);
463 if (error) {
464 fs_err(sdp, "can't create transaction glock: %d\n", error);
465 goto fail_rename;
466 }
b3b94faa
DT
467
468 return 0;
469
feaa7bba 470fail_trans:
b3b94faa 471 gfs2_glock_put(sdp->sd_trans_gl);
feaa7bba 472fail_rename:
b3b94faa 473 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 474fail_live:
b3b94faa 475 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 476fail_mount:
b3b94faa 477 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 478fail:
b3b94faa
DT
479 return error;
480}
481
9b8df98f
SW
482static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
483 u64 no_addr, const char *name)
f42faf4f 484{
9b8df98f
SW
485 struct gfs2_sbd *sdp = sb->s_fs_info;
486 struct dentry *dentry;
487 struct inode *inode;
488
1a0eae88 489 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
9b8df98f
SW
490 if (IS_ERR(inode)) {
491 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
492 return PTR_ERR(inode);
493 }
494 dentry = d_alloc_root(inode);
495 if (!dentry) {
496 fs_err(sdp, "can't alloc %s dentry\n", name);
497 iput(inode);
498 return -ENOMEM;
499 }
500 dentry->d_op = &gfs2_dops;
501 *dptr = dentry;
502 return 0;
f42faf4f
SW
503}
504
9b8df98f 505static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
506{
507 struct super_block *sb = sdp->sd_vfs;
508 struct gfs2_holder sb_gh;
dbb7cae2 509 u64 no_addr;
9b8df98f 510 int ret;
b3b94faa 511
9b8df98f
SW
512 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
513 LM_ST_SHARED, 0, &sb_gh);
514 if (ret) {
515 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
516 return ret;
b3b94faa 517 }
907b9bce 518
6bac243f 519 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
520 if (ret) {
521 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
522 goto out;
523 }
524
525 /* Set up the buffer cache and SB for real */
e1defc4f 526 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 527 ret = -EINVAL;
b3b94faa
DT
528 fs_err(sdp, "FS block size (%u) is too small for device "
529 "block size (%u)\n",
e1defc4f 530 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
531 goto out;
532 }
533 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 534 ret = -EINVAL;
b3b94faa
DT
535 fs_err(sdp, "FS block size (%u) is too big for machine "
536 "page size (%u)\n",
537 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
538 goto out;
539 }
b3b94faa
DT
540 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
541
f42faf4f 542 /* Get the root inode */
dbb7cae2 543 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
544 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
545 if (ret)
f42faf4f 546 goto out;
b3b94faa 547
9b8df98f
SW
548 /* Get the master inode */
549 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
550 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
551 if (ret) {
552 dput(sdp->sd_root_dir);
553 goto out;
554 }
555 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 556out:
b3b94faa 557 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 558 return ret;
b3b94faa
DT
559}
560
da6dd40d
BP
561/**
562 * map_journal_extents - create a reusable "extent" mapping from all logical
563 * blocks to all physical blocks for the given journal. This will save
564 * us time when writing journal blocks. Most journals will have only one
565 * extent that maps all their logical blocks. That's because gfs2.mkfs
566 * arranges the journal blocks sequentially to maximize performance.
567 * So the extent would map the first block for the entire file length.
568 * However, gfs2_jadd can happen while file activity is happening, so
569 * those journals may not be sequential. Less likely is the case where
570 * the users created their own journals by mounting the metafs and
571 * laying it out. But it's still possible. These journals might have
572 * several extents.
573 *
574 * TODO: This should be done in bigger chunks rather than one block at a time,
575 * but since it's only done at mount time, I'm not worried about the
576 * time it takes.
577 */
578static int map_journal_extents(struct gfs2_sbd *sdp)
579{
580 struct gfs2_jdesc *jd = sdp->sd_jdesc;
581 unsigned int lb;
582 u64 db, prev_db; /* logical block, disk block, prev disk block */
583 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
584 struct gfs2_journal_extent *jext = NULL;
585 struct buffer_head bh;
586 int rc = 0;
587
da6dd40d
BP
588 prev_db = 0;
589
c9e98886 590 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
da6dd40d
BP
591 bh.b_state = 0;
592 bh.b_blocknr = 0;
593 bh.b_size = 1 << ip->i_inode.i_blkbits;
594 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
595 db = bh.b_blocknr;
596 if (rc || !db) {
597 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
598 "%u db=%llu\n", rc, lb, (unsigned long long)db);
599 break;
600 }
601 if (!prev_db || db != prev_db + 1) {
602 jext = kzalloc(sizeof(struct gfs2_journal_extent),
603 GFP_KERNEL);
604 if (!jext) {
605 printk(KERN_INFO "GFS2 error: out of memory "
606 "mapping journal extents.\n");
607 rc = -ENOMEM;
608 break;
609 }
610 jext->dblock = db;
611 jext->lblock = lb;
612 jext->blocks = 1;
613 list_add_tail(&jext->extent_list, &jd->extent_list);
614 } else {
615 jext->blocks++;
616 }
617 prev_db = db;
618 }
619 return rc;
620}
621
f057f6cd 622static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 623{
f057f6cd
SW
624 char *message = "FIRSTMOUNT=Done";
625 char *envp[] = { message, NULL };
626 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
627 ls->ls_first_done = 1;
628 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
629}
630
b5289681
SW
631/**
632 * gfs2_jindex_hold - Grab a lock on the jindex
633 * @sdp: The GFS2 superblock
634 * @ji_gh: the holder for the jindex glock
635 *
636 * Returns: errno
637 */
638
639static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
640{
641 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
642 struct qstr name;
643 char buf[20];
644 struct gfs2_jdesc *jd;
645 int error;
646
647 name.name = buf;
648
649 mutex_lock(&sdp->sd_jindex_mutex);
650
651 for (;;) {
652 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
653 if (error)
654 break;
655
656 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
657 name.hash = gfs2_disk_hash(name.name, name.len);
658
659 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
660 if (error == -ENOENT) {
661 error = 0;
662 break;
663 }
664
665 gfs2_glock_dq_uninit(ji_gh);
666
667 if (error)
668 break;
669
670 error = -ENOMEM;
671 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
672 if (!jd)
673 break;
674
675 INIT_LIST_HEAD(&jd->extent_list);
fe64d517 676 slow_work_init(&jd->jd_work, &gfs2_recover_ops);
b5289681
SW
677 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
678 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
679 if (!jd->jd_inode)
680 error = -ENOENT;
681 else
682 error = PTR_ERR(jd->jd_inode);
683 kfree(jd);
684 break;
685 }
686
687 spin_lock(&sdp->sd_jindex_spin);
688 jd->jd_jid = sdp->sd_journals++;
689 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
690 spin_unlock(&sdp->sd_jindex_spin);
691 }
692
693 mutex_unlock(&sdp->sd_jindex_mutex);
694
695 return error;
696}
697
b3b94faa
DT
698static int init_journal(struct gfs2_sbd *sdp, int undo)
699{
9b8df98f 700 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 701 struct gfs2_holder ji_gh;
5c676f6d 702 struct gfs2_inode *ip;
b3b94faa
DT
703 int jindex = 1;
704 int error = 0;
705
706 if (undo) {
707 jindex = 0;
fe64d517 708 goto fail_jinode_gh;
b3b94faa
DT
709 }
710
9b8df98f 711 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 712 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 713 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 714 return PTR_ERR(sdp->sd_jindex);
b3b94faa 715 }
feaa7bba 716 ip = GFS2_I(sdp->sd_jindex);
b3b94faa
DT
717
718 /* Load in the journal index special file */
719
720 error = gfs2_jindex_hold(sdp, &ji_gh);
721 if (error) {
722 fs_err(sdp, "can't read journal index: %d\n", error);
723 goto fail;
724 }
725
0e5a9fb0 726 error = -EUSERS;
b3b94faa
DT
727 if (!gfs2_jindex_size(sdp)) {
728 fs_err(sdp, "no journals!\n");
907b9bce 729 goto fail_jindex;
b3b94faa
DT
730 }
731
732 if (sdp->sd_args.ar_spectator) {
733 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 734 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
735 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
736 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
737 } else {
738 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
739 fs_err(sdp, "can't mount journal #%u\n",
740 sdp->sd_lockstruct.ls_jid);
741 fs_err(sdp, "there are only %u journals (0 - %u)\n",
742 gfs2_jindex_size(sdp),
743 gfs2_jindex_size(sdp) - 1);
744 goto fail_jindex;
745 }
746 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
747
feaa7bba 748 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
749 &gfs2_journal_glops,
750 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
751 &sdp->sd_journal_gh);
752 if (error) {
753 fs_err(sdp, "can't acquire journal glock: %d\n", error);
754 goto fail_jindex;
755 }
756
feaa7bba
SW
757 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
758 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 759 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
760 &sdp->sd_jinode_gh);
761 if (error) {
762 fs_err(sdp, "can't acquire journal inode glock: %d\n",
763 error);
764 goto fail_journal_gh;
765 }
766
767 error = gfs2_jdesc_check(sdp->sd_jdesc);
768 if (error) {
769 fs_err(sdp, "my journal (%u) is bad: %d\n",
770 sdp->sd_jdesc->jd_jid, error);
771 goto fail_jinode_gh;
772 }
fd041f0b 773 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
774 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
775 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
776
777 /* Map the extents for this journal's blocks */
778 map_journal_extents(sdp);
b3b94faa 779 }
63997775 780 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
781
782 if (sdp->sd_lockstruct.ls_first) {
783 unsigned int x;
784 for (x = 0; x < sdp->sd_journals; x++) {
c63e31c2 785 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
b3b94faa
DT
786 if (error) {
787 fs_err(sdp, "error recovering journal %u: %d\n",
788 x, error);
789 goto fail_jinode_gh;
790 }
791 }
792
f057f6cd 793 gfs2_others_may_mount(sdp);
b3b94faa 794 } else if (!sdp->sd_args.ar_spectator) {
c63e31c2 795 error = gfs2_recover_journal(sdp->sd_jdesc);
b3b94faa
DT
796 if (error) {
797 fs_err(sdp, "error recovering my journal: %d\n", error);
798 goto fail_jinode_gh;
799 }
800 }
801
802 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
803 gfs2_glock_dq_uninit(&ji_gh);
804 jindex = 0;
805
b3b94faa
DT
806 return 0;
807
a91ea69f 808fail_jinode_gh:
b3b94faa
DT
809 if (!sdp->sd_args.ar_spectator)
810 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 811fail_journal_gh:
b3b94faa
DT
812 if (!sdp->sd_args.ar_spectator)
813 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 814fail_jindex:
b3b94faa
DT
815 gfs2_jindex_free(sdp);
816 if (jindex)
817 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 818fail:
f42faf4f 819 iput(sdp->sd_jindex);
b3b94faa
DT
820 return error;
821}
822
b3b94faa
DT
823
824static int init_inodes(struct gfs2_sbd *sdp, int undo)
825{
b3b94faa 826 int error = 0;
5c676f6d 827 struct gfs2_inode *ip;
9b8df98f 828 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
829
830 if (undo)
f42faf4f
SW
831 goto fail_qinode;
832
f42faf4f
SW
833 error = init_journal(sdp, undo);
834 if (error)
9b8df98f 835 goto fail;
b3b94faa 836
b3b94faa 837 /* Read in the master statfs inode */
9b8df98f 838 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
839 if (IS_ERR(sdp->sd_statfs_inode)) {
840 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 841 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 842 goto fail_journal;
b3b94faa
DT
843 }
844
845 /* Read in the resource index inode */
9b8df98f 846 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
847 if (IS_ERR(sdp->sd_rindex)) {
848 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
849 fs_err(sdp, "can't get resource index inode: %d\n", error);
850 goto fail_statfs;
851 }
feaa7bba 852 ip = GFS2_I(sdp->sd_rindex);
cf45b752 853 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
854
855 /* Read in the quota inode */
9b8df98f 856 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
857 if (IS_ERR(sdp->sd_quota_inode)) {
858 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
859 fs_err(sdp, "can't get quota file inode: %d\n", error);
860 goto fail_rindex;
861 }
b3b94faa
DT
862 return 0;
863
f42faf4f
SW
864fail_qinode:
865 iput(sdp->sd_quota_inode);
f42faf4f 866fail_rindex:
b3b94faa 867 gfs2_clear_rgrpd(sdp);
f42faf4f 868 iput(sdp->sd_rindex);
f42faf4f
SW
869fail_statfs:
870 iput(sdp->sd_statfs_inode);
f42faf4f
SW
871fail_journal:
872 init_journal(sdp, UNDO);
f42faf4f 873fail:
b3b94faa
DT
874 return error;
875}
876
877static int init_per_node(struct gfs2_sbd *sdp, int undo)
878{
f42faf4f 879 struct inode *pn = NULL;
b3b94faa
DT
880 char buf[30];
881 int error = 0;
5c676f6d 882 struct gfs2_inode *ip;
9b8df98f 883 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
884
885 if (sdp->sd_args.ar_spectator)
886 return 0;
887
888 if (undo)
889 goto fail_qc_gh;
890
9b8df98f 891 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
892 if (IS_ERR(pn)) {
893 error = PTR_ERR(pn);
b3b94faa
DT
894 fs_err(sdp, "can't find per_node directory: %d\n", error);
895 return error;
896 }
897
b3b94faa 898 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
899 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
900 if (IS_ERR(sdp->sd_sc_inode)) {
901 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 902 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 903 goto fail;
b3b94faa
DT
904 }
905
b3b94faa 906 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
907 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
908 if (IS_ERR(sdp->sd_qc_inode)) {
909 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
910 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
911 goto fail_ut_i;
912 }
913
f42faf4f 914 iput(pn);
b3b94faa
DT
915 pn = NULL;
916
feaa7bba 917 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 918 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
919 &sdp->sd_sc_gh);
920 if (error) {
921 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 922 goto fail_qc_i;
b3b94faa
DT
923 }
924
feaa7bba 925 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 926 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
927 &sdp->sd_qc_gh);
928 if (error) {
929 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
930 goto fail_ut_gh;
931 }
932
933 return 0;
934
a91ea69f 935fail_qc_gh:
b3b94faa 936 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 937fail_ut_gh:
b3b94faa 938 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 939fail_qc_i:
f42faf4f 940 iput(sdp->sd_qc_inode);
a91ea69f 941fail_ut_i:
f42faf4f 942 iput(sdp->sd_sc_inode);
a91ea69f 943fail:
b3b94faa 944 if (pn)
f42faf4f 945 iput(pn);
b3b94faa
DT
946 return error;
947}
948
949static int init_threads(struct gfs2_sbd *sdp, int undo)
950{
951 struct task_struct *p;
952 int error = 0;
953
954 if (undo)
feaa7bba 955 goto fail_quotad;
b3b94faa 956
b3b94faa
DT
957 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
958 error = IS_ERR(p);
959 if (error) {
960 fs_err(sdp, "can't start logd thread: %d\n", error);
961 return error;
962 }
963 sdp->sd_logd_process = p;
964
b3b94faa
DT
965 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
966 error = IS_ERR(p);
967 if (error) {
968 fs_err(sdp, "can't start quotad thread: %d\n", error);
969 goto fail;
970 }
971 sdp->sd_quotad_process = p;
972
b3b94faa
DT
973 return 0;
974
b3b94faa 975
feaa7bba 976fail_quotad:
b3b94faa 977 kthread_stop(sdp->sd_quotad_process);
feaa7bba 978fail:
b3b94faa 979 kthread_stop(sdp->sd_logd_process);
b3b94faa
DT
980 return error;
981}
da755fdb 982
f057f6cd
SW
983static const match_table_t nolock_tokens = {
984 { Opt_jid, "jid=%d\n", },
985 { Opt_err, NULL },
986};
987
8f05228e
SW
988static void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl)
989{
990 struct gfs2_sbd *sdp = gl->gl_sbd;
991 kmem_cache_free(cachep, gl);
992 if (atomic_dec_and_test(&sdp->sd_glock_disposal))
993 wake_up(&sdp->sd_glock_wait);
994}
995
f057f6cd
SW
996static const struct lm_lockops nolock_ops = {
997 .lm_proto_name = "lock_nolock",
8f05228e 998 .lm_put_lock = nolock_put_lock,
f057f6cd
SW
999 .lm_tokens = &nolock_tokens,
1000};
1001
da755fdb
SW
1002/**
1003 * gfs2_lm_mount - mount a locking protocol
1004 * @sdp: the filesystem
3ad2f3fb 1005 * @args: mount arguments
da755fdb
SW
1006 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1007 *
1008 * Returns: errno
1009 */
1010
1011static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1012{
f057f6cd
SW
1013 const struct lm_lockops *lm;
1014 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1015 struct gfs2_args *args = &sdp->sd_args;
1016 const char *proto = sdp->sd_proto_name;
1017 const char *table = sdp->sd_table_name;
1018 const char *fsname;
1019 char *o, *options;
1020 int ret;
da755fdb 1021
f057f6cd
SW
1022 if (!strcmp("lock_nolock", proto)) {
1023 lm = &nolock_ops;
1024 sdp->sd_args.ar_localflocks = 1;
1025 sdp->sd_args.ar_localcaching = 1;
1026#ifdef CONFIG_GFS2_FS_LOCKING_DLM
1027 } else if (!strcmp("lock_dlm", proto)) {
1028 lm = &gfs2_dlm_ops;
1029#endif
1030 } else {
1031 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1032 return -ENOENT;
1033 }
da755fdb
SW
1034
1035 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1036
f057f6cd
SW
1037 ls->ls_ops = lm;
1038 ls->ls_first = 1;
da755fdb 1039
f057f6cd
SW
1040 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1041 substring_t tmp[MAX_OPT_ARGS];
1042 int token, option;
1043
1044 if (!o || !*o)
1045 continue;
1046
1047 token = match_token(o, *lm->lm_tokens, tmp);
1048 switch (token) {
1049 case Opt_jid:
1050 ret = match_int(&tmp[0], &option);
1051 if (ret || option < 0)
1052 goto hostdata_error;
1053 ls->ls_jid = option;
1054 break;
1055 case Opt_id:
2b88f7c5 1056 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
1057 break;
1058 case Opt_first:
1059 ret = match_int(&tmp[0], &option);
1060 if (ret || (option != 0 && option != 1))
1061 goto hostdata_error;
1062 ls->ls_first = option;
1063 break;
1064 case Opt_nodir:
1065 ret = match_int(&tmp[0], &option);
1066 if (ret || (option != 0 && option != 1))
1067 goto hostdata_error;
1068 ls->ls_nodir = option;
1069 break;
1070 case Opt_err:
1071 default:
1072hostdata_error:
1073 fs_info(sdp, "unknown hostdata (%s)\n", o);
1074 return -EINVAL;
1075 }
da755fdb
SW
1076 }
1077
1078 if (sdp->sd_args.ar_spectator)
1079 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1080 else
1081 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1082 sdp->sd_lockstruct.ls_jid);
1083
f057f6cd
SW
1084 fsname = strchr(table, ':');
1085 if (fsname)
1086 fsname++;
1087 if (lm->lm_mount == NULL) {
1088 fs_info(sdp, "Now mounting FS...\n");
1089 return 0;
da755fdb 1090 }
f057f6cd
SW
1091 ret = lm->lm_mount(sdp, fsname);
1092 if (ret == 0)
1093 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1094 return ret;
da755fdb
SW
1095}
1096
1097void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1098{
f057f6cd
SW
1099 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1100 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1101 lm->lm_unmount)
1102 lm->lm_unmount(sdp);
da755fdb 1103}
b3b94faa 1104
8633ecfa
SW
1105void gfs2_online_uevent(struct gfs2_sbd *sdp)
1106{
1107 struct super_block *sb = sdp->sd_vfs;
1108 char ro[20];
1109 char spectator[20];
1110 char *envp[] = { ro, spectator, NULL };
1111 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1112 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1113 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1114}
1115
b3b94faa
DT
1116/**
1117 * fill_super - Read in superblock
1118 * @sb: The VFS superblock
1119 * @data: Mount options
1120 * @silent: Don't complain if it's not a GFS2 filesystem
1121 *
1122 * Returns: errno
1123 */
1124
f55073ff 1125static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1126{
1127 struct gfs2_sbd *sdp;
1128 struct gfs2_holder mount_gh;
1129 int error;
1130
1131 sdp = init_sbd(sb);
1132 if (!sdp) {
d92a8d48 1133 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1134 return -ENOMEM;
1135 }
f55073ff 1136 sdp->sd_args = *args;
b3b94faa 1137
fe64d517 1138 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1139 sb->s_flags |= MS_RDONLY;
fe64d517
SW
1140 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1141 }
6f04c1c7
SW
1142 if (sdp->sd_args.ar_posix_acl)
1143 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1144 if (sdp->sd_args.ar_nobarrier)
1145 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1146
719ee344
SW
1147 sb->s_magic = GFS2_MAGIC;
1148 sb->s_op = &gfs2_super_ops;
1149 sb->s_export_op = &gfs2_export_ops;
40b78a32 1150 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1151 sb->s_qcop = &gfs2_quotactl_ops;
1152 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1153 sb->s_time_gran = 1;
1154 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1155
1156 /* Set up the buffer cache and fill in some fake block size values
1157 to allow us to read-in the on-disk superblock. */
1158 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1159 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1160 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1161 GFS2_BASIC_BLOCK_SHIFT;
1162 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1163
5e687eac 1164 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1165 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1166 if (sdp->sd_args.ar_statfs_quantum) {
1167 sdp->sd_tune.gt_statfs_slow = 0;
1168 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
1169 }
1170 else {
1171 sdp->sd_tune.gt_statfs_slow = 1;
1172 sdp->sd_tune.gt_statfs_quantum = 30;
1173 }
48c2b613 1174
b3b94faa
DT
1175 error = init_names(sdp, silent);
1176 if (error)
1177 goto fail;
1178
7c52b166
RP
1179 gfs2_create_debugfs_file(sdp);
1180
b3b94faa
DT
1181 error = gfs2_sys_fs_add(sdp);
1182 if (error)
1183 goto fail;
1184
1185 error = gfs2_lm_mount(sdp, silent);
1186 if (error)
1187 goto fail_sys;
1188
1189 error = init_locking(sdp, &mount_gh, DO);
1190 if (error)
1191 goto fail_lm;
1192
9b8df98f 1193 error = init_sb(sdp, silent);
b3b94faa
DT
1194 if (error)
1195 goto fail_locking;
b3b94faa
DT
1196
1197 error = init_inodes(sdp, DO);
1198 if (error)
f42faf4f 1199 goto fail_sb;
b3b94faa
DT
1200
1201 error = init_per_node(sdp, DO);
1202 if (error)
1203 goto fail_inodes;
1204
1205 error = gfs2_statfs_init(sdp);
1206 if (error) {
1207 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1208 goto fail_per_node;
1209 }
1210
1211 error = init_threads(sdp, DO);
1212 if (error)
1213 goto fail_per_node;
1214
1215 if (!(sb->s_flags & MS_RDONLY)) {
1216 error = gfs2_make_fs_rw(sdp);
1217 if (error) {
1218 fs_err(sdp, "can't make FS RW: %d\n", error);
1219 goto fail_threads;
1220 }
1221 }
1222
1223 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1224 gfs2_online_uevent(sdp);
b3b94faa
DT
1225 return 0;
1226
a91ea69f 1227fail_threads:
b3b94faa 1228 init_threads(sdp, UNDO);
a91ea69f 1229fail_per_node:
b3b94faa 1230 init_per_node(sdp, UNDO);
a91ea69f 1231fail_inodes:
b3b94faa 1232 init_inodes(sdp, UNDO);
a91ea69f 1233fail_sb:
9b8df98f
SW
1234 if (sdp->sd_root_dir)
1235 dput(sdp->sd_root_dir);
1236 if (sdp->sd_master_dir)
1237 dput(sdp->sd_master_dir);
e7c8707e
SW
1238 if (sb->s_root)
1239 dput(sb->s_root);
9b8df98f 1240 sb->s_root = NULL;
a91ea69f 1241fail_locking:
b3b94faa 1242 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1243fail_lm:
c1184f8a 1244 invalidate_inodes(sb);
fefc03bf 1245 gfs2_gl_hash_clear(sdp);
b3b94faa 1246 gfs2_lm_unmount(sdp);
a91ea69f 1247fail_sys:
b3b94faa 1248 gfs2_sys_fs_del(sdp);
a91ea69f 1249fail:
5f882096 1250 gfs2_delete_debugfs_file(sdp);
a2c45807 1251 kfree(sdp);
5c676f6d 1252 sb->s_fs_info = NULL;
b3b94faa
DT
1253 return error;
1254}
1255
f55073ff 1256static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1257{
f55073ff
SW
1258 s->s_bdev = data;
1259 s->s_dev = s->s_bdev->bd_dev;
1260
1261 /*
1262 * We set the bdi here to the queue backing, file systems can
1263 * overwrite this in ->fill_super()
1264 */
1265 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1266 return 0;
86384605
AD
1267}
1268
f55073ff 1269static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1270{
1271 struct block_device *bdev = ptr;
1272 return (bdev == s->s_bdev);
1273}
1274
f55073ff
SW
1275/**
1276 * gfs2_get_sb - Get the GFS2 superblock
1277 * @fs_type: The GFS2 filesystem type
1278 * @flags: Mount flags
1279 * @dev_name: The name of the device
1280 * @data: The mount arguments
1281 * @mnt: The vfsmnt for this mount
1282 *
1283 * Q. Why not use get_sb_bdev() ?
1284 * A. We need to select one of two root directories to mount, independent
1285 * of whether this is the initial, or subsequent, mount of this sb
1286 *
1287 * Returns: 0 or -ve on error
1288 */
1289
1290static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1291 const char *dev_name, void *data, struct vfsmount *mnt)
1292{
1293 struct block_device *bdev;
1294 struct super_block *s;
1295 fmode_t mode = FMODE_READ;
1296 int error;
1297 struct gfs2_args args;
1298 struct gfs2_sbd *sdp;
1299
1300 if (!(flags & MS_RDONLY))
1301 mode |= FMODE_WRITE;
1302
1303 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1304 if (IS_ERR(bdev))
1305 return PTR_ERR(bdev);
1306
1307 /*
1308 * once the super is inserted into the list by sget, s_umount
1309 * will protect the lockfs code from trying to start a snapshot
1310 * while we are mounting
1311 */
1312 mutex_lock(&bdev->bd_fsfreeze_mutex);
1313 if (bdev->bd_fsfreeze_count > 0) {
1314 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1315 error = -EBUSY;
1316 goto error_bdev;
1317 }
1318 s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev);
1319 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1320 error = PTR_ERR(s);
1321 if (IS_ERR(s))
1322 goto error_bdev;
1323
1324 memset(&args, 0, sizeof(args));
1325 args.ar_quota = GFS2_QUOTA_DEFAULT;
1326 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1327 args.ar_commit = 30;
3d3c10f2
BM
1328 args.ar_statfs_quantum = 30;
1329 args.ar_quota_quantum = 60;
f55073ff
SW
1330 args.ar_errors = GFS2_ERRORS_DEFAULT;
1331
1332 error = gfs2_mount_args(&args, data);
1333 if (error) {
1334 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1335 if (s->s_root)
1336 goto error_super;
1337 deactivate_locked_super(s);
1338 return error;
1339 }
1340
1341 if (s->s_root) {
1342 error = -EBUSY;
1343 if ((flags ^ s->s_flags) & MS_RDONLY)
1344 goto error_super;
1345 close_bdev_exclusive(bdev, mode);
1346 } else {
1347 char b[BDEVNAME_SIZE];
1348
1349 s->s_flags = flags;
1350 s->s_mode = mode;
1351 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1352 sb_set_blocksize(s, block_size(bdev));
1353 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
1354 if (error) {
1355 deactivate_locked_super(s);
1356 return error;
1357 }
1358 s->s_flags |= MS_ACTIVE;
1359 bdev->bd_super = s;
1360 }
1361
1362 sdp = s->s_fs_info;
1363 mnt->mnt_sb = s;
1364 if (args.ar_meta)
1365 mnt->mnt_root = dget(sdp->sd_master_dir);
1366 else
1367 mnt->mnt_root = dget(sdp->sd_root_dir);
1368 return 0;
1369
1370error_super:
1371 deactivate_locked_super(s);
1372error_bdev:
1373 close_bdev_exclusive(bdev, mode);
1374 return error;
1375}
1376
f6d03139
SW
1377static int set_meta_super(struct super_block *s, void *ptr)
1378{
1379 return -EINVAL;
1380}
1381
003dec89
SW
1382static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1383 const char *dev_name, void *data, struct vfsmount *mnt)
86384605 1384{
f6d03139 1385 struct super_block *s;
003dec89 1386 struct gfs2_sbd *sdp;
e24977d4 1387 struct path path;
86384605 1388 int error;
907b9bce 1389
e24977d4 1390 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1391 if (error) {
9b8df98f
SW
1392 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1393 dev_name, error);
003dec89 1394 return error;
86384605 1395 }
f55073ff 1396 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super,
f6d03139 1397 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1398 path_put(&path);
f6d03139 1399 if (IS_ERR(s)) {
86384605 1400 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
f6d03139 1401 return PTR_ERR(s);
86384605 1402 }
f55073ff
SW
1403 if ((flags ^ s->s_flags) & MS_RDONLY) {
1404 deactivate_locked_super(s);
1405 return -EBUSY;
1406 }
f6d03139
SW
1407 sdp = s->s_fs_info;
1408 mnt->mnt_sb = s;
9b8df98f
SW
1409 mnt->mnt_root = dget(sdp->sd_master_dir);
1410 return 0;
b3b94faa
DT
1411}
1412
419c93e0 1413static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1414{
1415 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1416
1417 if (sdp == NULL) {
1418 kill_block_super(sb);
1419 return;
3af165ac 1420 }
88a19ad0
SW
1421
1422 gfs2_meta_syncfs(sdp);
1423 dput(sdp->sd_root_dir);
1424 dput(sdp->sd_master_dir);
1425 sdp->sd_root_dir = NULL;
1426 sdp->sd_master_dir = NULL;
fefc03bf 1427 shrink_dcache_sb(sb);
9b8df98f 1428 kill_block_super(sb);
88a19ad0
SW
1429 gfs2_delete_debugfs_file(sdp);
1430 kfree(sdp);
86384605
AD
1431}
1432
b3b94faa
DT
1433struct file_system_type gfs2_fs_type = {
1434 .name = "gfs2",
1435 .fs_flags = FS_REQUIRES_DEV,
1436 .get_sb = gfs2_get_sb,
419c93e0
SW
1437 .kill_sb = gfs2_kill_sb,
1438 .owner = THIS_MODULE,
1439};
1440
1441struct file_system_type gfs2meta_fs_type = {
1442 .name = "gfs2meta",
1443 .fs_flags = FS_REQUIRES_DEV,
86384605 1444 .get_sb = gfs2_get_sb_meta,
b3b94faa
DT
1445 .owner = THIS_MODULE,
1446};
1447