]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/gfs2/ops_fstype.c
GFS2: Review bug traps in glops.c
[mirror_ubuntu-bionic-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
f55ab26a 118 mutex_init(&sdp->sd_freeze_lock);
b3b94faa
DT
119
120 return sdp;
121}
122
b3b94faa 123
9b8df98f
SW
124/**
125 * gfs2_check_sb - Check superblock
126 * @sdp: the filesystem
127 * @sb: The superblock
128 * @silent: Don't print a message if the check fails
129 *
130 * Checks the version code of the FS is one that we understand how to
131 * read and that the sizes of the various on-disk structures have not
132 * changed.
133 */
134
32e471ef 135static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 136{
32e471ef
SW
137 struct gfs2_sb_host *sb = &sdp->sd_sb;
138
9b8df98f
SW
139 if (sb->sb_magic != GFS2_MAGIC ||
140 sb->sb_type != GFS2_METATYPE_SB) {
141 if (!silent)
142 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
143 return -EINVAL;
144 }
145
146 /* If format numbers match exactly, we're done. */
147
148 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
149 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
150 return 0;
151
c80dbb58 152 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 153
c80dbb58 154 return -EINVAL;
9b8df98f
SW
155}
156
157static void end_bio_io_page(struct bio *bio, int error)
158{
159 struct page *page = bio->bi_private;
160
161 if (!error)
162 SetPageUptodate(page);
163 else
164 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
165 unlock_page(page);
166}
167
32e471ef 168static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 169{
32e471ef
SW
170 struct gfs2_sb_host *sb = &sdp->sd_sb;
171 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
172 const struct gfs2_sb *str = buf;
173
174 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
175 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
176 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
177 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
178 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
179 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
180 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
181 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
182 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
183 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
184 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
185
186 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
187 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 188 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
189}
190
191/**
192 * gfs2_read_super - Read the gfs2 super block from disk
193 * @sdp: The GFS2 super block
194 * @sector: The location of the super block
195 * @error: The error code to return
196 *
197 * This uses the bio functions to read the super block from disk
198 * because we want to be 100% sure that we never read cached data.
199 * A super block is read twice only during each GFS2 mount and is
200 * never written to by the filesystem. The first time its read no
201 * locks are held, and the only details which are looked at are those
202 * relating to the locking protocol. Once locking is up and working,
203 * the sb is read again under the lock to establish the location of
204 * the master directory (contains pointers to journals etc) and the
205 * root directory.
206 *
207 * Returns: 0 on success or error
208 */
209
32e471ef 210static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
211{
212 struct super_block *sb = sdp->sd_vfs;
213 struct gfs2_sb *p;
214 struct page *page;
215 struct bio *bio;
216
217 page = alloc_page(GFP_NOFS);
218 if (unlikely(!page))
219 return -ENOBUFS;
220
221 ClearPageUptodate(page);
222 ClearPageDirty(page);
223 lock_page(page);
224
225 bio = bio_alloc(GFP_NOFS, 1);
9b8df98f
SW
226 bio->bi_sector = sector * (sb->s_blocksize >> 9);
227 bio->bi_bdev = sb->s_bdev;
228 bio_add_page(bio, page, PAGE_SIZE, 0);
229
230 bio->bi_end_io = end_bio_io_page;
231 bio->bi_private = page;
20ed0535 232 submit_bio(READ_SYNC | REQ_META, bio);
9b8df98f
SW
233 wait_on_page_locked(page);
234 bio_put(bio);
235 if (!PageUptodate(page)) {
236 __free_page(page);
237 return -EIO;
238 }
239 p = kmap(page);
32e471ef 240 gfs2_sb_in(sdp, p);
9b8df98f
SW
241 kunmap(page);
242 __free_page(page);
32e471ef 243 return gfs2_check_sb(sdp, silent);
9b8df98f 244}
6bac243f 245
9b8df98f
SW
246/**
247 * gfs2_read_sb - Read super block
248 * @sdp: The GFS2 superblock
9b8df98f
SW
249 * @silent: Don't print message if mount fails
250 *
251 */
252
6bac243f 253static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
254{
255 u32 hash_blocks, ind_blocks, leaf_blocks;
256 u32 tmp_blocks;
257 unsigned int x;
258 int error;
259
32e471ef 260 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
261 if (error) {
262 if (!silent)
263 fs_err(sdp, "can't read superblock\n");
264 return error;
265 }
266
9b8df98f
SW
267 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
268 GFS2_BASIC_BLOCK_SHIFT;
269 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
270 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
271 sizeof(struct gfs2_dinode)) / sizeof(u64);
272 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
273 sizeof(struct gfs2_meta_header)) / sizeof(u64);
274 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
275 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
276 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
277 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
278 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
279 sizeof(struct gfs2_meta_header)) /
280 sizeof(struct gfs2_quota_change);
281
282 /* Compute maximum reservation required to add a entry to a directory */
283
284 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
285 sdp->sd_jbsize);
286
287 ind_blocks = 0;
288 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
289 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
290 ind_blocks += tmp_blocks;
291 }
292
293 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
294
295 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
296
297 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
298 sizeof(struct gfs2_dinode);
299 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
300 for (x = 2;; x++) {
301 u64 space, d;
302 u32 m;
303
304 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
305 d = space;
306 m = do_div(d, sdp->sd_inptrs);
307
308 if (d != sdp->sd_heightsize[x - 1] || m)
309 break;
310 sdp->sd_heightsize[x] = space;
311 }
312 sdp->sd_max_height = x;
313 sdp->sd_heightsize[x] = ~0;
314 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
315
316 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
317 sizeof(struct gfs2_dinode);
318 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
319 for (x = 2;; x++) {
320 u64 space, d;
321 u32 m;
322
323 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
324 d = space;
325 m = do_div(d, sdp->sd_inptrs);
326
327 if (d != sdp->sd_jheightsize[x - 1] || m)
328 break;
329 sdp->sd_jheightsize[x] = space;
330 }
331 sdp->sd_max_jheight = x;
332 sdp->sd_jheightsize[x] = ~0;
333 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
334
335 return 0;
336}
337
b3b94faa
DT
338static int init_names(struct gfs2_sbd *sdp, int silent)
339{
b3b94faa
DT
340 char *proto, *table;
341 int error = 0;
342
343 proto = sdp->sd_args.ar_lockproto;
344 table = sdp->sd_args.ar_locktable;
345
346 /* Try to autodetect */
347
348 if (!proto[0] || !table[0]) {
32e471ef 349 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
350 if (error)
351 return error;
3cf1e7be 352
b3b94faa 353 if (!proto[0])
3cf1e7be 354 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 355 if (!table[0])
3cf1e7be 356 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
357 }
358
359 if (!table[0])
360 table = sdp->sd_vfs->s_id;
361
00377d8e
JD
362 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
363 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 364
5d35e31f
DC
365 table = sdp->sd_table_name;
366 while ((table = strchr(table, '/')))
b35997d4
RP
367 *table = '_';
368
b3b94faa
DT
369 return error;
370}
371
372static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
373 int undo)
374{
b3b94faa
DT
375 int error = 0;
376
377 if (undo)
378 goto fail_trans;
379
b3b94faa
DT
380 error = gfs2_glock_nq_num(sdp,
381 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
382 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
383 mount_gh);
384 if (error) {
385 fs_err(sdp, "can't acquire mount glock: %d\n", error);
386 goto fail;
387 }
388
389 error = gfs2_glock_nq_num(sdp,
390 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
391 LM_ST_SHARED,
579b78a4 392 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
393 &sdp->sd_live_gh);
394 if (error) {
395 fs_err(sdp, "can't acquire live glock: %d\n", error);
396 goto fail_mount;
397 }
398
399 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
400 CREATE, &sdp->sd_rename_gl);
401 if (error) {
402 fs_err(sdp, "can't create rename glock: %d\n", error);
403 goto fail_live;
404 }
405
406 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
407 CREATE, &sdp->sd_trans_gl);
408 if (error) {
409 fs_err(sdp, "can't create transaction glock: %d\n", error);
410 goto fail_rename;
411 }
b3b94faa
DT
412
413 return 0;
414
feaa7bba 415fail_trans:
b3b94faa 416 gfs2_glock_put(sdp->sd_trans_gl);
feaa7bba 417fail_rename:
b3b94faa 418 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 419fail_live:
b3b94faa 420 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 421fail_mount:
b3b94faa 422 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 423fail:
b3b94faa
DT
424 return error;
425}
426
9b8df98f
SW
427static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
428 u64 no_addr, const char *name)
f42faf4f 429{
9b8df98f
SW
430 struct gfs2_sbd *sdp = sb->s_fs_info;
431 struct dentry *dentry;
432 struct inode *inode;
433
44ad37d6 434 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
9b8df98f
SW
435 if (IS_ERR(inode)) {
436 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
437 return PTR_ERR(inode);
438 }
48fde701 439 dentry = d_make_root(inode);
9b8df98f
SW
440 if (!dentry) {
441 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
442 return -ENOMEM;
443 }
9b8df98f
SW
444 *dptr = dentry;
445 return 0;
f42faf4f
SW
446}
447
9b8df98f 448static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
449{
450 struct super_block *sb = sdp->sd_vfs;
451 struct gfs2_holder sb_gh;
dbb7cae2 452 u64 no_addr;
9b8df98f 453 int ret;
b3b94faa 454
9b8df98f
SW
455 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
456 LM_ST_SHARED, 0, &sb_gh);
457 if (ret) {
458 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
459 return ret;
b3b94faa 460 }
907b9bce 461
6bac243f 462 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
463 if (ret) {
464 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
465 goto out;
466 }
467
468 /* Set up the buffer cache and SB for real */
e1defc4f 469 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 470 ret = -EINVAL;
b3b94faa
DT
471 fs_err(sdp, "FS block size (%u) is too small for device "
472 "block size (%u)\n",
e1defc4f 473 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
474 goto out;
475 }
476 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 477 ret = -EINVAL;
b3b94faa
DT
478 fs_err(sdp, "FS block size (%u) is too big for machine "
479 "page size (%u)\n",
480 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
481 goto out;
482 }
b3b94faa
DT
483 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
484
f42faf4f 485 /* Get the root inode */
dbb7cae2 486 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
487 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
488 if (ret)
f42faf4f 489 goto out;
b3b94faa 490
9b8df98f
SW
491 /* Get the master inode */
492 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
493 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
494 if (ret) {
495 dput(sdp->sd_root_dir);
496 goto out;
497 }
498 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 499out:
b3b94faa 500 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 501 return ret;
b3b94faa
DT
502}
503
da6dd40d
BP
504/**
505 * map_journal_extents - create a reusable "extent" mapping from all logical
506 * blocks to all physical blocks for the given journal. This will save
507 * us time when writing journal blocks. Most journals will have only one
508 * extent that maps all their logical blocks. That's because gfs2.mkfs
509 * arranges the journal blocks sequentially to maximize performance.
510 * So the extent would map the first block for the entire file length.
511 * However, gfs2_jadd can happen while file activity is happening, so
512 * those journals may not be sequential. Less likely is the case where
513 * the users created their own journals by mounting the metafs and
514 * laying it out. But it's still possible. These journals might have
515 * several extents.
516 *
517 * TODO: This should be done in bigger chunks rather than one block at a time,
518 * but since it's only done at mount time, I'm not worried about the
519 * time it takes.
520 */
521static int map_journal_extents(struct gfs2_sbd *sdp)
522{
523 struct gfs2_jdesc *jd = sdp->sd_jdesc;
524 unsigned int lb;
525 u64 db, prev_db; /* logical block, disk block, prev disk block */
526 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
527 struct gfs2_journal_extent *jext = NULL;
528 struct buffer_head bh;
529 int rc = 0;
530
da6dd40d
BP
531 prev_db = 0;
532
a2e0f799 533 for (lb = 0; lb < i_size_read(jd->jd_inode) >> sdp->sd_sb.sb_bsize_shift; lb++) {
da6dd40d
BP
534 bh.b_state = 0;
535 bh.b_blocknr = 0;
536 bh.b_size = 1 << ip->i_inode.i_blkbits;
537 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
538 db = bh.b_blocknr;
539 if (rc || !db) {
540 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
541 "%u db=%llu\n", rc, lb, (unsigned long long)db);
542 break;
543 }
544 if (!prev_db || db != prev_db + 1) {
545 jext = kzalloc(sizeof(struct gfs2_journal_extent),
546 GFP_KERNEL);
547 if (!jext) {
548 printk(KERN_INFO "GFS2 error: out of memory "
549 "mapping journal extents.\n");
550 rc = -ENOMEM;
551 break;
552 }
553 jext->dblock = db;
554 jext->lblock = lb;
555 jext->blocks = 1;
556 list_add_tail(&jext->extent_list, &jd->extent_list);
557 } else {
558 jext->blocks++;
559 }
560 prev_db = db;
561 }
562 return rc;
563}
564
f057f6cd 565static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 566{
f057f6cd
SW
567 char *message = "FIRSTMOUNT=Done";
568 char *envp[] = { message, NULL };
e0c2a9aa
DT
569
570 fs_info(sdp, "first mount done, others may mount\n");
571
572 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
573 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
574
f057f6cd 575 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
576}
577
b5289681
SW
578/**
579 * gfs2_jindex_hold - Grab a lock on the jindex
580 * @sdp: The GFS2 superblock
581 * @ji_gh: the holder for the jindex glock
582 *
583 * Returns: errno
584 */
585
586static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
587{
588 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
589 struct qstr name;
590 char buf[20];
591 struct gfs2_jdesc *jd;
592 int error;
593
594 name.name = buf;
595
596 mutex_lock(&sdp->sd_jindex_mutex);
597
598 for (;;) {
599 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
600 if (error)
601 break;
602
603 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
604 name.hash = gfs2_disk_hash(name.name, name.len);
605
606 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
607 if (error == -ENOENT) {
608 error = 0;
609 break;
610 }
611
612 gfs2_glock_dq_uninit(ji_gh);
613
614 if (error)
615 break;
616
617 error = -ENOMEM;
618 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
619 if (!jd)
620 break;
621
622 INIT_LIST_HEAD(&jd->extent_list);
6ecd7c2d 623 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
624 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
625 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
626 if (!jd->jd_inode)
627 error = -ENOENT;
628 else
629 error = PTR_ERR(jd->jd_inode);
630 kfree(jd);
631 break;
632 }
633
634 spin_lock(&sdp->sd_jindex_spin);
635 jd->jd_jid = sdp->sd_journals++;
636 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
637 spin_unlock(&sdp->sd_jindex_spin);
638 }
639
640 mutex_unlock(&sdp->sd_jindex_mutex);
641
642 return error;
643}
644
b3b94faa
DT
645static int init_journal(struct gfs2_sbd *sdp, int undo)
646{
9b8df98f 647 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 648 struct gfs2_holder ji_gh;
5c676f6d 649 struct gfs2_inode *ip;
b3b94faa
DT
650 int jindex = 1;
651 int error = 0;
652
653 if (undo) {
654 jindex = 0;
fe64d517 655 goto fail_jinode_gh;
b3b94faa
DT
656 }
657
9b8df98f 658 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 659 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 660 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 661 return PTR_ERR(sdp->sd_jindex);
b3b94faa 662 }
b3b94faa
DT
663
664 /* Load in the journal index special file */
665
666 error = gfs2_jindex_hold(sdp, &ji_gh);
667 if (error) {
668 fs_err(sdp, "can't read journal index: %d\n", error);
669 goto fail;
670 }
671
0e5a9fb0 672 error = -EUSERS;
b3b94faa
DT
673 if (!gfs2_jindex_size(sdp)) {
674 fs_err(sdp, "no journals!\n");
907b9bce 675 goto fail_jindex;
b3b94faa
DT
676 }
677
678 if (sdp->sd_args.ar_spectator) {
679 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 680 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
681 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
682 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
683 } else {
684 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
685 fs_err(sdp, "can't mount journal #%u\n",
686 sdp->sd_lockstruct.ls_jid);
687 fs_err(sdp, "there are only %u journals (0 - %u)\n",
688 gfs2_jindex_size(sdp),
689 gfs2_jindex_size(sdp) - 1);
690 goto fail_jindex;
691 }
692 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
693
feaa7bba 694 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
695 &gfs2_journal_glops,
696 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
697 &sdp->sd_journal_gh);
698 if (error) {
699 fs_err(sdp, "can't acquire journal glock: %d\n", error);
700 goto fail_jindex;
701 }
702
feaa7bba
SW
703 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
704 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 705 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
706 &sdp->sd_jinode_gh);
707 if (error) {
708 fs_err(sdp, "can't acquire journal inode glock: %d\n",
709 error);
710 goto fail_journal_gh;
711 }
712
713 error = gfs2_jdesc_check(sdp->sd_jdesc);
714 if (error) {
715 fs_err(sdp, "my journal (%u) is bad: %d\n",
716 sdp->sd_jdesc->jd_jid, error);
717 goto fail_jinode_gh;
718 }
fd041f0b 719 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
720 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
721 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
722
723 /* Map the extents for this journal's blocks */
724 map_journal_extents(sdp);
b3b94faa 725 }
63997775 726 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
727
728 if (sdp->sd_lockstruct.ls_first) {
729 unsigned int x;
730 for (x = 0; x < sdp->sd_journals; x++) {
6ecd7c2d
TH
731 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
732 true);
b3b94faa
DT
733 if (error) {
734 fs_err(sdp, "error recovering journal %u: %d\n",
735 x, error);
736 goto fail_jinode_gh;
737 }
738 }
739
f057f6cd 740 gfs2_others_may_mount(sdp);
b3b94faa 741 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 742 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
743 if (error) {
744 fs_err(sdp, "error recovering my journal: %d\n", error);
745 goto fail_jinode_gh;
746 }
747 }
748
749 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
750 gfs2_glock_dq_uninit(&ji_gh);
751 jindex = 0;
752
b3b94faa
DT
753 return 0;
754
a91ea69f 755fail_jinode_gh:
b3b94faa
DT
756 if (!sdp->sd_args.ar_spectator)
757 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 758fail_journal_gh:
b3b94faa
DT
759 if (!sdp->sd_args.ar_spectator)
760 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 761fail_jindex:
b3b94faa
DT
762 gfs2_jindex_free(sdp);
763 if (jindex)
764 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 765fail:
f42faf4f 766 iput(sdp->sd_jindex);
b3b94faa
DT
767 return error;
768}
769
56aa72d0 770static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
771
772static int init_inodes(struct gfs2_sbd *sdp, int undo)
773{
b3b94faa 774 int error = 0;
9b8df98f 775 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
776
777 if (undo)
f42faf4f
SW
778 goto fail_qinode;
779
f42faf4f
SW
780 error = init_journal(sdp, undo);
781 if (error)
9b8df98f 782 goto fail;
b3b94faa 783
b3b94faa 784 /* Read in the master statfs inode */
9b8df98f 785 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
786 if (IS_ERR(sdp->sd_statfs_inode)) {
787 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 788 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 789 goto fail_journal;
b3b94faa
DT
790 }
791
792 /* Read in the resource index inode */
9b8df98f 793 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
794 if (IS_ERR(sdp->sd_rindex)) {
795 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
796 fs_err(sdp, "can't get resource index inode: %d\n", error);
797 goto fail_statfs;
798 }
cf45b752 799 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
800
801 /* Read in the quota inode */
9b8df98f 802 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
803 if (IS_ERR(sdp->sd_quota_inode)) {
804 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
805 fs_err(sdp, "can't get quota file inode: %d\n", error);
806 goto fail_rindex;
807 }
56aa72d0
JK
808 /*
809 * i_mutex on quota files is special. Since this inode is hidden system
810 * file, we are safe to define locking ourselves.
811 */
812 lockdep_set_class(&sdp->sd_quota_inode->i_mutex,
813 &gfs2_quota_imutex_key);
a365fbf3
SW
814
815 error = gfs2_rindex_update(sdp);
816 if (error)
817 goto fail_qinode;
818
b3b94faa
DT
819 return 0;
820
f42faf4f
SW
821fail_qinode:
822 iput(sdp->sd_quota_inode);
f42faf4f 823fail_rindex:
b3b94faa 824 gfs2_clear_rgrpd(sdp);
f42faf4f 825 iput(sdp->sd_rindex);
f42faf4f
SW
826fail_statfs:
827 iput(sdp->sd_statfs_inode);
f42faf4f
SW
828fail_journal:
829 init_journal(sdp, UNDO);
f42faf4f 830fail:
b3b94faa
DT
831 return error;
832}
833
834static int init_per_node(struct gfs2_sbd *sdp, int undo)
835{
f42faf4f 836 struct inode *pn = NULL;
b3b94faa
DT
837 char buf[30];
838 int error = 0;
5c676f6d 839 struct gfs2_inode *ip;
9b8df98f 840 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
841
842 if (sdp->sd_args.ar_spectator)
843 return 0;
844
845 if (undo)
846 goto fail_qc_gh;
847
9b8df98f 848 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
849 if (IS_ERR(pn)) {
850 error = PTR_ERR(pn);
b3b94faa
DT
851 fs_err(sdp, "can't find per_node directory: %d\n", error);
852 return error;
853 }
854
b3b94faa 855 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
856 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
857 if (IS_ERR(sdp->sd_sc_inode)) {
858 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 859 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 860 goto fail;
b3b94faa
DT
861 }
862
b3b94faa 863 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
864 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
865 if (IS_ERR(sdp->sd_qc_inode)) {
866 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
867 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
868 goto fail_ut_i;
869 }
870
f42faf4f 871 iput(pn);
b3b94faa
DT
872 pn = NULL;
873
feaa7bba 874 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 875 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
876 &sdp->sd_sc_gh);
877 if (error) {
878 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 879 goto fail_qc_i;
b3b94faa
DT
880 }
881
feaa7bba 882 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 883 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
884 &sdp->sd_qc_gh);
885 if (error) {
886 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
887 goto fail_ut_gh;
888 }
889
890 return 0;
891
a91ea69f 892fail_qc_gh:
b3b94faa 893 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 894fail_ut_gh:
b3b94faa 895 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 896fail_qc_i:
f42faf4f 897 iput(sdp->sd_qc_inode);
a91ea69f 898fail_ut_i:
f42faf4f 899 iput(sdp->sd_sc_inode);
a91ea69f 900fail:
b3b94faa 901 if (pn)
f42faf4f 902 iput(pn);
b3b94faa
DT
903 return error;
904}
905
906static int init_threads(struct gfs2_sbd *sdp, int undo)
907{
908 struct task_struct *p;
909 int error = 0;
910
911 if (undo)
feaa7bba 912 goto fail_quotad;
b3b94faa 913
b3b94faa
DT
914 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
915 error = IS_ERR(p);
916 if (error) {
917 fs_err(sdp, "can't start logd thread: %d\n", error);
918 return error;
919 }
920 sdp->sd_logd_process = p;
921
b3b94faa
DT
922 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
923 error = IS_ERR(p);
924 if (error) {
925 fs_err(sdp, "can't start quotad thread: %d\n", error);
926 goto fail;
927 }
928 sdp->sd_quotad_process = p;
929
b3b94faa
DT
930 return 0;
931
b3b94faa 932
feaa7bba 933fail_quotad:
b3b94faa 934 kthread_stop(sdp->sd_quotad_process);
feaa7bba 935fail:
b3b94faa 936 kthread_stop(sdp->sd_logd_process);
b3b94faa
DT
937 return error;
938}
da755fdb 939
f057f6cd
SW
940static const match_table_t nolock_tokens = {
941 { Opt_jid, "jid=%d\n", },
942 { Opt_err, NULL },
943};
944
945static const struct lm_lockops nolock_ops = {
946 .lm_proto_name = "lock_nolock",
fc0e38da 947 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
948 .lm_tokens = &nolock_tokens,
949};
950
da755fdb
SW
951/**
952 * gfs2_lm_mount - mount a locking protocol
953 * @sdp: the filesystem
3ad2f3fb 954 * @args: mount arguments
da755fdb
SW
955 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
956 *
957 * Returns: errno
958 */
959
960static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
961{
f057f6cd
SW
962 const struct lm_lockops *lm;
963 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
964 struct gfs2_args *args = &sdp->sd_args;
965 const char *proto = sdp->sd_proto_name;
966 const char *table = sdp->sd_table_name;
f057f6cd
SW
967 char *o, *options;
968 int ret;
da755fdb 969
f057f6cd
SW
970 if (!strcmp("lock_nolock", proto)) {
971 lm = &nolock_ops;
972 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
973#ifdef CONFIG_GFS2_FS_LOCKING_DLM
974 } else if (!strcmp("lock_dlm", proto)) {
975 lm = &gfs2_dlm_ops;
976#endif
977 } else {
978 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
979 return -ENOENT;
980 }
da755fdb
SW
981
982 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
983
f057f6cd
SW
984 ls->ls_ops = lm;
985 ls->ls_first = 1;
da755fdb 986
f057f6cd
SW
987 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
988 substring_t tmp[MAX_OPT_ARGS];
989 int token, option;
990
991 if (!o || !*o)
992 continue;
993
994 token = match_token(o, *lm->lm_tokens, tmp);
995 switch (token) {
996 case Opt_jid:
997 ret = match_int(&tmp[0], &option);
998 if (ret || option < 0)
999 goto hostdata_error;
ba6e9364
SW
1000 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
1001 ls->ls_jid = option;
f057f6cd
SW
1002 break;
1003 case Opt_id:
4875647a 1004 case Opt_nodir:
2b88f7c5 1005 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
1006 break;
1007 case Opt_first:
1008 ret = match_int(&tmp[0], &option);
1009 if (ret || (option != 0 && option != 1))
1010 goto hostdata_error;
1011 ls->ls_first = option;
1012 break;
f057f6cd
SW
1013 case Opt_err:
1014 default:
1015hostdata_error:
1016 fs_info(sdp, "unknown hostdata (%s)\n", o);
1017 return -EINVAL;
1018 }
da755fdb
SW
1019 }
1020
f057f6cd
SW
1021 if (lm->lm_mount == NULL) {
1022 fs_info(sdp, "Now mounting FS...\n");
19237039 1023 complete_all(&sdp->sd_locking_init);
f057f6cd 1024 return 0;
da755fdb 1025 }
e0c2a9aa 1026 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1027 if (ret == 0)
1028 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1029 complete_all(&sdp->sd_locking_init);
f057f6cd 1030 return ret;
da755fdb
SW
1031}
1032
1033void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1034{
f057f6cd
SW
1035 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1036 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1037 lm->lm_unmount)
1038 lm->lm_unmount(sdp);
da755fdb 1039}
b3b94faa 1040
ba6e9364
SW
1041static int gfs2_journalid_wait(void *word)
1042{
1043 if (signal_pending(current))
1044 return -EINTR;
1045 schedule();
1046 return 0;
1047}
1048
1049static int wait_on_journal(struct gfs2_sbd *sdp)
1050{
ba6e9364
SW
1051 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1052 return 0;
1053
1054 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1055}
1056
8633ecfa
SW
1057void gfs2_online_uevent(struct gfs2_sbd *sdp)
1058{
1059 struct super_block *sb = sdp->sd_vfs;
1060 char ro[20];
1061 char spectator[20];
1062 char *envp[] = { ro, spectator, NULL };
1063 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1064 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1065 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1066}
1067
b3b94faa
DT
1068/**
1069 * fill_super - Read in superblock
1070 * @sb: The VFS superblock
1071 * @data: Mount options
1072 * @silent: Don't complain if it's not a GFS2 filesystem
1073 *
1074 * Returns: errno
1075 */
1076
f55073ff 1077static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1078{
1079 struct gfs2_sbd *sdp;
1080 struct gfs2_holder mount_gh;
1081 int error;
1082
1083 sdp = init_sbd(sb);
1084 if (!sdp) {
d92a8d48 1085 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1086 return -ENOMEM;
1087 }
f55073ff 1088 sdp->sd_args = *args;
b3b94faa 1089
fe64d517 1090 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1091 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1092 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1093 }
6f04c1c7
SW
1094 if (sdp->sd_args.ar_posix_acl)
1095 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1096 if (sdp->sd_args.ar_nobarrier)
1097 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1098
9964afbb 1099 sb->s_flags |= MS_NOSEC;
719ee344
SW
1100 sb->s_magic = GFS2_MAGIC;
1101 sb->s_op = &gfs2_super_ops;
41ced6dc 1102 sb->s_d_op = &gfs2_dops;
719ee344 1103 sb->s_export_op = &gfs2_export_ops;
40b78a32 1104 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1105 sb->s_qcop = &gfs2_quotactl_ops;
1106 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1107 sb->s_time_gran = 1;
1108 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1109
1110 /* Set up the buffer cache and fill in some fake block size values
1111 to allow us to read-in the on-disk superblock. */
1112 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1113 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1114 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1115 GFS2_BASIC_BLOCK_SHIFT;
1116 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1117
5e687eac 1118 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1119 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1120 if (sdp->sd_args.ar_statfs_quantum) {
1121 sdp->sd_tune.gt_statfs_slow = 0;
1122 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1123 } else {
3d3c10f2
BM
1124 sdp->sd_tune.gt_statfs_slow = 1;
1125 sdp->sd_tune.gt_statfs_quantum = 30;
1126 }
48c2b613 1127
b3b94faa 1128 error = init_names(sdp, silent);
0d515210
BP
1129 if (error) {
1130 /* In this case, we haven't initialized sysfs, so we have to
1131 manually free the sdp. */
1132 free_percpu(sdp->sd_lkstats);
1133 kfree(sdp);
1134 sb->s_fs_info = NULL;
1135 return error;
1136 }
b3b94faa 1137
e0c2a9aa
DT
1138 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1139
b3b94faa 1140 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1141 /*
1142 * If we hit an error here, gfs2_sys_fs_add will have called function
1143 * kobject_put which causes the sysfs usage count to go to zero, which
1144 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1145 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1146 * kobject_put to free sdp.
1147 */
b3b94faa 1148 if (error)
0d515210
BP
1149 return error;
1150
1151 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1152
1153 error = gfs2_lm_mount(sdp, silent);
1154 if (error)
0d515210 1155 goto fail_debug;
b3b94faa
DT
1156
1157 error = init_locking(sdp, &mount_gh, DO);
1158 if (error)
1159 goto fail_lm;
1160
9b8df98f 1161 error = init_sb(sdp, silent);
b3b94faa
DT
1162 if (error)
1163 goto fail_locking;
b3b94faa 1164
ba6e9364
SW
1165 error = wait_on_journal(sdp);
1166 if (error)
1167 goto fail_sb;
1168
feb47ca9
SW
1169 /*
1170 * If user space has failed to join the cluster or some similar
1171 * failure has occurred, then the journal id will contain a
1172 * negative (error) number. This will then be returned to the
1173 * caller (of the mount syscall). We do this even for spectator
1174 * mounts (which just write a jid of 0 to indicate "ok" even though
1175 * the jid is unused in the spectator case)
1176 */
1177 if (sdp->sd_lockstruct.ls_jid < 0) {
1178 error = sdp->sd_lockstruct.ls_jid;
1179 sdp->sd_lockstruct.ls_jid = 0;
1180 goto fail_sb;
1181 }
1182
e0c2a9aa
DT
1183 if (sdp->sd_args.ar_spectator)
1184 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1185 sdp->sd_table_name);
1186 else
1187 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1188 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1189
b3b94faa
DT
1190 error = init_inodes(sdp, DO);
1191 if (error)
f42faf4f 1192 goto fail_sb;
b3b94faa
DT
1193
1194 error = init_per_node(sdp, DO);
1195 if (error)
1196 goto fail_inodes;
1197
1198 error = gfs2_statfs_init(sdp);
1199 if (error) {
1200 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1201 goto fail_per_node;
1202 }
1203
1204 error = init_threads(sdp, DO);
1205 if (error)
1206 goto fail_per_node;
1207
1208 if (!(sb->s_flags & MS_RDONLY)) {
1209 error = gfs2_make_fs_rw(sdp);
1210 if (error) {
1211 fs_err(sdp, "can't make FS RW: %d\n", error);
1212 goto fail_threads;
1213 }
1214 }
1215
1216 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1217 gfs2_online_uevent(sdp);
b3b94faa
DT
1218 return 0;
1219
a91ea69f 1220fail_threads:
b3b94faa 1221 init_threads(sdp, UNDO);
a91ea69f 1222fail_per_node:
b3b94faa 1223 init_per_node(sdp, UNDO);
a91ea69f 1224fail_inodes:
b3b94faa 1225 init_inodes(sdp, UNDO);
a91ea69f 1226fail_sb:
9b8df98f
SW
1227 if (sdp->sd_root_dir)
1228 dput(sdp->sd_root_dir);
1229 if (sdp->sd_master_dir)
1230 dput(sdp->sd_master_dir);
e7c8707e
SW
1231 if (sb->s_root)
1232 dput(sb->s_root);
9b8df98f 1233 sb->s_root = NULL;
a91ea69f 1234fail_locking:
b3b94faa 1235 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1236fail_lm:
fefc03bf 1237 gfs2_gl_hash_clear(sdp);
b3b94faa 1238 gfs2_lm_unmount(sdp);
0d515210 1239fail_debug:
5f882096 1240 gfs2_delete_debugfs_file(sdp);
a245769f 1241 free_percpu(sdp->sd_lkstats);
0d515210
BP
1242 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1243 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1244 gfs2_sys_fs_del(sdp);
5c676f6d 1245 sb->s_fs_info = NULL;
b3b94faa
DT
1246 return error;
1247}
1248
f55073ff 1249static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1250{
f55073ff
SW
1251 s->s_bdev = data;
1252 s->s_dev = s->s_bdev->bd_dev;
1253
1254 /*
1255 * We set the bdi here to the queue backing, file systems can
1256 * overwrite this in ->fill_super()
1257 */
1258 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1259 return 0;
86384605
AD
1260}
1261
f55073ff 1262static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1263{
1264 struct block_device *bdev = ptr;
1265 return (bdev == s->s_bdev);
1266}
1267
f55073ff 1268/**
8bcbbf00 1269 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1270 * @fs_type: The GFS2 filesystem type
1271 * @flags: Mount flags
1272 * @dev_name: The name of the device
1273 * @data: The mount arguments
f55073ff
SW
1274 *
1275 * Q. Why not use get_sb_bdev() ?
1276 * A. We need to select one of two root directories to mount, independent
1277 * of whether this is the initial, or subsequent, mount of this sb
1278 *
1279 * Returns: 0 or -ve on error
1280 */
1281
8bcbbf00
AV
1282static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1283 const char *dev_name, void *data)
f55073ff
SW
1284{
1285 struct block_device *bdev;
1286 struct super_block *s;
d4d77629 1287 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1288 int error;
1289 struct gfs2_args args;
1290 struct gfs2_sbd *sdp;
1291
1292 if (!(flags & MS_RDONLY))
1293 mode |= FMODE_WRITE;
1294
d4d77629 1295 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1296 if (IS_ERR(bdev))
8bcbbf00 1297 return ERR_CAST(bdev);
f55073ff
SW
1298
1299 /*
1300 * once the super is inserted into the list by sget, s_umount
1301 * will protect the lockfs code from trying to start a snapshot
1302 * while we are mounting
1303 */
1304 mutex_lock(&bdev->bd_fsfreeze_mutex);
1305 if (bdev->bd_fsfreeze_count > 0) {
1306 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1307 error = -EBUSY;
1308 goto error_bdev;
1309 }
9249e17f 1310 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1311 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1312 error = PTR_ERR(s);
1313 if (IS_ERR(s))
1314 goto error_bdev;
1315
8bcbbf00 1316 if (s->s_root)
d4d77629 1317 blkdev_put(bdev, mode);
8bcbbf00 1318
f55073ff
SW
1319 memset(&args, 0, sizeof(args));
1320 args.ar_quota = GFS2_QUOTA_DEFAULT;
1321 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1322 args.ar_commit = 30;
3d3c10f2
BM
1323 args.ar_statfs_quantum = 30;
1324 args.ar_quota_quantum = 60;
f55073ff
SW
1325 args.ar_errors = GFS2_ERRORS_DEFAULT;
1326
1327 error = gfs2_mount_args(&args, data);
1328 if (error) {
1329 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
8bcbbf00 1330 goto error_super;
f55073ff
SW
1331 }
1332
1333 if (s->s_root) {
1334 error = -EBUSY;
1335 if ((flags ^ s->s_flags) & MS_RDONLY)
1336 goto error_super;
f55073ff
SW
1337 } else {
1338 char b[BDEVNAME_SIZE];
1339
f55073ff
SW
1340 s->s_mode = mode;
1341 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1342 sb_set_blocksize(s, block_size(bdev));
1343 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1344 if (error)
1345 goto error_super;
f55073ff
SW
1346 s->s_flags |= MS_ACTIVE;
1347 bdev->bd_super = s;
1348 }
1349
1350 sdp = s->s_fs_info;
f55073ff 1351 if (args.ar_meta)
8bcbbf00 1352 return dget(sdp->sd_master_dir);
f55073ff 1353 else
8bcbbf00 1354 return dget(sdp->sd_root_dir);
f55073ff
SW
1355
1356error_super:
1357 deactivate_locked_super(s);
8bcbbf00 1358 return ERR_PTR(error);
f55073ff 1359error_bdev:
d4d77629 1360 blkdev_put(bdev, mode);
8bcbbf00 1361 return ERR_PTR(error);
f55073ff
SW
1362}
1363
f6d03139
SW
1364static int set_meta_super(struct super_block *s, void *ptr)
1365{
1366 return -EINVAL;
1367}
1368
8bcbbf00
AV
1369static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1370 int flags, const char *dev_name, void *data)
86384605 1371{
f6d03139 1372 struct super_block *s;
003dec89 1373 struct gfs2_sbd *sdp;
e24977d4 1374 struct path path;
86384605 1375 int error;
907b9bce 1376
e24977d4 1377 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1378 if (error) {
9b8df98f
SW
1379 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1380 dev_name, error);
8bcbbf00 1381 return ERR_PTR(error);
86384605 1382 }
9249e17f 1383 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
f6d03139 1384 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1385 path_put(&path);
f6d03139 1386 if (IS_ERR(s)) {
86384605 1387 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
8bcbbf00 1388 return ERR_CAST(s);
86384605 1389 }
f55073ff
SW
1390 if ((flags ^ s->s_flags) & MS_RDONLY) {
1391 deactivate_locked_super(s);
8bcbbf00 1392 return ERR_PTR(-EBUSY);
f55073ff 1393 }
f6d03139 1394 sdp = s->s_fs_info;
8bcbbf00 1395 return dget(sdp->sd_master_dir);
b3b94faa
DT
1396}
1397
419c93e0 1398static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1399{
1400 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1401
1402 if (sdp == NULL) {
1403 kill_block_super(sb);
1404 return;
3af165ac 1405 }
88a19ad0
SW
1406
1407 gfs2_meta_syncfs(sdp);
1408 dput(sdp->sd_root_dir);
1409 dput(sdp->sd_master_dir);
1410 sdp->sd_root_dir = NULL;
1411 sdp->sd_master_dir = NULL;
fefc03bf 1412 shrink_dcache_sb(sb);
88a19ad0 1413 gfs2_delete_debugfs_file(sdp);
a245769f 1414 free_percpu(sdp->sd_lkstats);
0d515210 1415 kill_block_super(sb);
86384605
AD
1416}
1417
b3b94faa
DT
1418struct file_system_type gfs2_fs_type = {
1419 .name = "gfs2",
1420 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1421 .mount = gfs2_mount,
419c93e0
SW
1422 .kill_sb = gfs2_kill_sb,
1423 .owner = THIS_MODULE,
1424};
1425
1426struct file_system_type gfs2meta_fs_type = {
1427 .name = "gfs2meta",
1428 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1429 .mount = gfs2_mount_meta,
b3b94faa
DT
1430 .owner = THIS_MODULE,
1431};
1432