block/fs/drivers: remove rw argument from submit_bio
[linux-2.6-block.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
d77d1b58
JP
10#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11
b3b94faa
DT
12#include <linux/sched.h>
13#include <linux/slab.h>
14#include <linux/spinlock.h>
15#include <linux/completion.h>
16#include <linux/buffer_head.h>
b3b94faa
DT
17#include <linux/blkdev.h>
18#include <linux/kthread.h>
afeacc8c 19#include <linux/export.h>
86384605
AD
20#include <linux/namei.h>
21#include <linux/mount.h>
5c676f6d 22#include <linux/gfs2_ondisk.h>
cc632e7f 23#include <linux/quotaops.h>
56aa72d0 24#include <linux/lockdep.h>
7f78e035 25#include <linux/module.h>
b3b94faa
DT
26
27#include "gfs2.h"
5c676f6d 28#include "incore.h"
da6dd40d 29#include "bmap.h"
b3b94faa
DT
30#include "glock.h"
31#include "glops.h"
32#include "inode.h"
b3b94faa
DT
33#include "recovery.h"
34#include "rgrp.h"
35#include "super.h"
b3b94faa 36#include "sys.h"
5c676f6d 37#include "util.h"
bb3b0e3d 38#include "log.h"
9ac1b4d9 39#include "quota.h"
b5289681 40#include "dir.h"
70d4ee94 41#include "meta_io.h"
63997775 42#include "trace_gfs2.h"
b3b94faa
DT
43
44#define DO 0
45#define UNDO 1
46
9b8df98f
SW
47/**
48 * gfs2_tune_init - Fill a gfs2_tune structure with default values
49 * @gt: tune
50 *
51 */
52
53static void gfs2_tune_init(struct gfs2_tune *gt)
54{
55 spin_lock_init(&gt->gt_spin);
56
9b8df98f
SW
57 gt->gt_quota_warn_period = 10;
58 gt->gt_quota_scale_num = 1;
59 gt->gt_quota_scale_den = 1;
9b8df98f
SW
60 gt->gt_new_files_jdata = 0;
61 gt->gt_max_readahead = 1 << 18;
9b8df98f 62 gt->gt_complain_secs = 10;
9b8df98f
SW
63}
64
b3b94faa
DT
65static struct gfs2_sbd *init_sbd(struct super_block *sb)
66{
67 struct gfs2_sbd *sdp;
70d4ee94 68 struct address_space *mapping;
b3b94faa 69
85d1da67 70 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
71 if (!sdp)
72 return NULL;
73
5c676f6d 74 sb->s_fs_info = sdp;
b3b94faa 75 sdp->sd_vfs = sb;
a245769f
SW
76 sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
77 if (!sdp->sd_lkstats) {
78 kfree(sdp);
79 return NULL;
80 }
81
ba6e9364 82 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
83 gfs2_tune_init(&sdp->sd_tune);
84
e402746a
SW
85 init_waitqueue_head(&sdp->sd_glock_wait);
86 atomic_set(&sdp->sd_glock_disposal, 0);
3942ae53 87 init_completion(&sdp->sd_locking_init);
fd95e81c 88 init_completion(&sdp->sd_wdack);
b3b94faa 89 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
90
91 spin_lock_init(&sdp->sd_rindex_spin);
7c9ca621 92 sdp->sd_rindex_tree.rb_node = NULL;
b3b94faa
DT
93
94 INIT_LIST_HEAD(&sdp->sd_jindex_list);
95 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 96 mutex_init(&sdp->sd_jindex_mutex);
0e48e055 97 init_completion(&sdp->sd_journal_ready);
b3b94faa 98
b3b94faa 99 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 100 mutex_init(&sdp->sd_quota_mutex);
e46c772d 101 mutex_init(&sdp->sd_quota_sync_mutex);
37b2c837 102 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
103 INIT_LIST_HEAD(&sdp->sd_trunc_list);
104 spin_lock_init(&sdp->sd_trunc_lock);
2d9e7230 105 spin_lock_init(&sdp->sd_bitmap_lock);
b3b94faa 106
70d4ee94
SW
107 mapping = &sdp->sd_aspace;
108
39849d69 109 address_space_init_once(mapping);
1b2ad412 110 mapping->a_ops = &gfs2_rgrp_aops;
70d4ee94
SW
111 mapping->host = sb->s_bdev->bd_inode;
112 mapping->flags = 0;
113 mapping_set_gfp_mask(mapping, GFP_NOFS);
114 mapping->private_data = NULL;
70d4ee94
SW
115 mapping->writeback_index = 0;
116
b3b94faa 117 spin_lock_init(&sdp->sd_log_lock);
5e687eac 118 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa 119 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
d7b616e2 120 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
45138990 121 spin_lock_init(&sdp->sd_ordered_lock);
b3b94faa 122
5e687eac
BM
123 init_waitqueue_head(&sdp->sd_log_waitq);
124 init_waitqueue_head(&sdp->sd_logd_waitq);
d6a079e8 125 spin_lock_init(&sdp->sd_ail_lock);
b3b94faa
DT
126 INIT_LIST_HEAD(&sdp->sd_ail1_list);
127 INIT_LIST_HEAD(&sdp->sd_ail2_list);
128
484adff8 129 init_rwsem(&sdp->sd_log_flush_lock);
16615be1 130 atomic_set(&sdp->sd_log_in_flight, 0);
2e60d768
BM
131 atomic_set(&sdp->sd_reserving_log, 0);
132 init_waitqueue_head(&sdp->sd_reserving_log_wait);
16615be1 133 init_waitqueue_head(&sdp->sd_log_flush_wait);
2e60d768
BM
134 atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN);
135 mutex_init(&sdp->sd_freeze_mutex);
b3b94faa 136
b3b94faa
DT
137 return sdp;
138}
139
b3b94faa 140
9b8df98f
SW
141/**
142 * gfs2_check_sb - Check superblock
143 * @sdp: the filesystem
144 * @sb: The superblock
145 * @silent: Don't print a message if the check fails
146 *
147 * Checks the version code of the FS is one that we understand how to
148 * read and that the sizes of the various on-disk structures have not
149 * changed.
150 */
151
32e471ef 152static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 153{
32e471ef
SW
154 struct gfs2_sb_host *sb = &sdp->sd_sb;
155
9b8df98f
SW
156 if (sb->sb_magic != GFS2_MAGIC ||
157 sb->sb_type != GFS2_METATYPE_SB) {
158 if (!silent)
d77d1b58 159 pr_warn("not a GFS2 filesystem\n");
9b8df98f
SW
160 return -EINVAL;
161 }
162
163 /* If format numbers match exactly, we're done. */
164
165 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
166 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
167 return 0;
168
c80dbb58 169 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 170
c80dbb58 171 return -EINVAL;
9b8df98f
SW
172}
173
4246a0b6 174static void end_bio_io_page(struct bio *bio)
9b8df98f
SW
175{
176 struct page *page = bio->bi_private;
177
4246a0b6 178 if (!bio->bi_error)
9b8df98f
SW
179 SetPageUptodate(page);
180 else
4246a0b6 181 pr_warn("error %d reading superblock\n", bio->bi_error);
9b8df98f
SW
182 unlock_page(page);
183}
184
32e471ef 185static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 186{
32e471ef
SW
187 struct gfs2_sb_host *sb = &sdp->sd_sb;
188 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
189 const struct gfs2_sb *str = buf;
190
191 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
192 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
193 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
194 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
195 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
196 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
197 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
198 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
199 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
200 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
201 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
202
203 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
204 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 205 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
206}
207
208/**
209 * gfs2_read_super - Read the gfs2 super block from disk
210 * @sdp: The GFS2 super block
211 * @sector: The location of the super block
212 * @error: The error code to return
213 *
214 * This uses the bio functions to read the super block from disk
215 * because we want to be 100% sure that we never read cached data.
216 * A super block is read twice only during each GFS2 mount and is
217 * never written to by the filesystem. The first time its read no
218 * locks are held, and the only details which are looked at are those
219 * relating to the locking protocol. Once locking is up and working,
220 * the sb is read again under the lock to establish the location of
221 * the master directory (contains pointers to journals etc) and the
222 * root directory.
223 *
224 * Returns: 0 on success or error
225 */
226
32e471ef 227static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
228{
229 struct super_block *sb = sdp->sd_vfs;
230 struct gfs2_sb *p;
231 struct page *page;
232 struct bio *bio;
233
234 page = alloc_page(GFP_NOFS);
235 if (unlikely(!page))
ac3beb6a 236 return -ENOMEM;
9b8df98f
SW
237
238 ClearPageUptodate(page);
239 ClearPageDirty(page);
240 lock_page(page);
241
242 bio = bio_alloc(GFP_NOFS, 1);
4f024f37 243 bio->bi_iter.bi_sector = sector * (sb->s_blocksize >> 9);
9b8df98f
SW
244 bio->bi_bdev = sb->s_bdev;
245 bio_add_page(bio, page, PAGE_SIZE, 0);
246
247 bio->bi_end_io = end_bio_io_page;
248 bio->bi_private = page;
4e49ea4a
MC
249 bio->bi_rw = READ_SYNC | REQ_META;
250 submit_bio(bio);
9b8df98f
SW
251 wait_on_page_locked(page);
252 bio_put(bio);
253 if (!PageUptodate(page)) {
254 __free_page(page);
255 return -EIO;
256 }
257 p = kmap(page);
32e471ef 258 gfs2_sb_in(sdp, p);
9b8df98f
SW
259 kunmap(page);
260 __free_page(page);
32e471ef 261 return gfs2_check_sb(sdp, silent);
9b8df98f 262}
6bac243f 263
9b8df98f
SW
264/**
265 * gfs2_read_sb - Read super block
266 * @sdp: The GFS2 superblock
9b8df98f
SW
267 * @silent: Don't print message if mount fails
268 *
269 */
270
6bac243f 271static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
272{
273 u32 hash_blocks, ind_blocks, leaf_blocks;
274 u32 tmp_blocks;
275 unsigned int x;
276 int error;
277
32e471ef 278 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
279 if (error) {
280 if (!silent)
281 fs_err(sdp, "can't read superblock\n");
282 return error;
283 }
284
9b8df98f
SW
285 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
286 GFS2_BASIC_BLOCK_SHIFT;
287 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
288 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
289 sizeof(struct gfs2_dinode)) / sizeof(u64);
290 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
291 sizeof(struct gfs2_meta_header)) / sizeof(u64);
292 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
293 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
294 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
295 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
296 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
297 sizeof(struct gfs2_meta_header)) /
298 sizeof(struct gfs2_quota_change);
a68a0a35
BP
299 sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
300 sizeof(struct gfs2_meta_header))
301 * GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
9b8df98f
SW
302
303 /* Compute maximum reservation required to add a entry to a directory */
304
305 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
306 sdp->sd_jbsize);
307
308 ind_blocks = 0;
309 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
310 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
311 ind_blocks += tmp_blocks;
312 }
313
314 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
315
316 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
317
318 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
319 sizeof(struct gfs2_dinode);
320 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
321 for (x = 2;; x++) {
322 u64 space, d;
323 u32 m;
324
325 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
326 d = space;
327 m = do_div(d, sdp->sd_inptrs);
328
329 if (d != sdp->sd_heightsize[x - 1] || m)
330 break;
331 sdp->sd_heightsize[x] = space;
332 }
333 sdp->sd_max_height = x;
334 sdp->sd_heightsize[x] = ~0;
335 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
336
337 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
338 sizeof(struct gfs2_dinode);
339 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
340 for (x = 2;; x++) {
341 u64 space, d;
342 u32 m;
343
344 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
345 d = space;
346 m = do_div(d, sdp->sd_inptrs);
347
348 if (d != sdp->sd_jheightsize[x - 1] || m)
349 break;
350 sdp->sd_jheightsize[x] = space;
351 }
352 sdp->sd_max_jheight = x;
353 sdp->sd_jheightsize[x] = ~0;
354 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
355
471f3db2
BM
356 sdp->sd_max_dents_per_leaf = (sdp->sd_sb.sb_bsize -
357 sizeof(struct gfs2_leaf)) /
358 GFS2_MIN_DIRENT_SIZE;
9b8df98f
SW
359 return 0;
360}
361
b3b94faa
DT
362static int init_names(struct gfs2_sbd *sdp, int silent)
363{
b3b94faa
DT
364 char *proto, *table;
365 int error = 0;
366
367 proto = sdp->sd_args.ar_lockproto;
368 table = sdp->sd_args.ar_locktable;
369
370 /* Try to autodetect */
371
372 if (!proto[0] || !table[0]) {
32e471ef 373 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
374 if (error)
375 return error;
3cf1e7be 376
b3b94faa 377 if (!proto[0])
3cf1e7be 378 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 379 if (!table[0])
3cf1e7be 380 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
381 }
382
383 if (!table[0])
384 table = sdp->sd_vfs->s_id;
385
00377d8e
JD
386 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
387 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 388
5d35e31f
DC
389 table = sdp->sd_table_name;
390 while ((table = strchr(table, '/')))
b35997d4
RP
391 *table = '_';
392
b3b94faa
DT
393 return error;
394}
395
396static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
397 int undo)
398{
b3b94faa
DT
399 int error = 0;
400
401 if (undo)
402 goto fail_trans;
403
b3b94faa
DT
404 error = gfs2_glock_nq_num(sdp,
405 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
406 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
407 mount_gh);
408 if (error) {
409 fs_err(sdp, "can't acquire mount glock: %d\n", error);
410 goto fail;
411 }
412
413 error = gfs2_glock_nq_num(sdp,
414 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
415 LM_ST_SHARED,
579b78a4 416 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
417 &sdp->sd_live_gh);
418 if (error) {
419 fs_err(sdp, "can't acquire live glock: %d\n", error);
420 goto fail_mount;
421 }
422
423 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
424 CREATE, &sdp->sd_rename_gl);
425 if (error) {
426 fs_err(sdp, "can't create rename glock: %d\n", error);
427 goto fail_live;
428 }
429
24972557
BM
430 error = gfs2_glock_get(sdp, GFS2_FREEZE_LOCK, &gfs2_freeze_glops,
431 CREATE, &sdp->sd_freeze_gl);
b3b94faa
DT
432 if (error) {
433 fs_err(sdp, "can't create transaction glock: %d\n", error);
434 goto fail_rename;
435 }
b3b94faa
DT
436
437 return 0;
438
feaa7bba 439fail_trans:
24972557 440 gfs2_glock_put(sdp->sd_freeze_gl);
feaa7bba 441fail_rename:
b3b94faa 442 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 443fail_live:
b3b94faa 444 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 445fail_mount:
b3b94faa 446 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 447fail:
b3b94faa
DT
448 return error;
449}
450
9b8df98f
SW
451static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
452 u64 no_addr, const char *name)
f42faf4f 453{
9b8df98f
SW
454 struct gfs2_sbd *sdp = sb->s_fs_info;
455 struct dentry *dentry;
456 struct inode *inode;
457
73b462d2 458 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
9b8df98f
SW
459 if (IS_ERR(inode)) {
460 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
461 return PTR_ERR(inode);
462 }
48fde701 463 dentry = d_make_root(inode);
9b8df98f
SW
464 if (!dentry) {
465 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
466 return -ENOMEM;
467 }
9b8df98f
SW
468 *dptr = dentry;
469 return 0;
f42faf4f
SW
470}
471
9b8df98f 472static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
473{
474 struct super_block *sb = sdp->sd_vfs;
475 struct gfs2_holder sb_gh;
dbb7cae2 476 u64 no_addr;
9b8df98f 477 int ret;
b3b94faa 478
9b8df98f
SW
479 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
480 LM_ST_SHARED, 0, &sb_gh);
481 if (ret) {
482 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
483 return ret;
b3b94faa 484 }
907b9bce 485
6bac243f 486 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
487 if (ret) {
488 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
489 goto out;
490 }
491
492 /* Set up the buffer cache and SB for real */
e1defc4f 493 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 494 ret = -EINVAL;
b3b94faa
DT
495 fs_err(sdp, "FS block size (%u) is too small for device "
496 "block size (%u)\n",
e1defc4f 497 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
498 goto out;
499 }
500 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 501 ret = -EINVAL;
b3b94faa
DT
502 fs_err(sdp, "FS block size (%u) is too big for machine "
503 "page size (%u)\n",
504 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
505 goto out;
506 }
b3b94faa
DT
507 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
508
f42faf4f 509 /* Get the root inode */
dbb7cae2 510 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
511 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
512 if (ret)
f42faf4f 513 goto out;
b3b94faa 514
9b8df98f
SW
515 /* Get the master inode */
516 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
517 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
518 if (ret) {
519 dput(sdp->sd_root_dir);
520 goto out;
521 }
522 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 523out:
b3b94faa 524 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 525 return ret;
b3b94faa
DT
526}
527
f057f6cd 528static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 529{
f057f6cd
SW
530 char *message = "FIRSTMOUNT=Done";
531 char *envp[] = { message, NULL };
e0c2a9aa
DT
532
533 fs_info(sdp, "first mount done, others may mount\n");
534
535 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
536 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
537
f057f6cd 538 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
539}
540
b5289681
SW
541/**
542 * gfs2_jindex_hold - Grab a lock on the jindex
543 * @sdp: The GFS2 superblock
544 * @ji_gh: the holder for the jindex glock
545 *
546 * Returns: errno
547 */
548
549static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
550{
551 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
552 struct qstr name;
553 char buf[20];
554 struct gfs2_jdesc *jd;
555 int error;
556
557 name.name = buf;
558
559 mutex_lock(&sdp->sd_jindex_mutex);
560
561 for (;;) {
562 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
563 if (error)
564 break;
565
566 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
567 name.hash = gfs2_disk_hash(name.name, name.len);
568
569 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
570 if (error == -ENOENT) {
571 error = 0;
572 break;
573 }
574
575 gfs2_glock_dq_uninit(ji_gh);
576
577 if (error)
578 break;
579
580 error = -ENOMEM;
581 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
582 if (!jd)
583 break;
584
585 INIT_LIST_HEAD(&jd->extent_list);
a17d758b
BP
586 INIT_LIST_HEAD(&jd->jd_revoke_list);
587
6ecd7c2d 588 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
589 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
590 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
591 if (!jd->jd_inode)
592 error = -ENOENT;
593 else
594 error = PTR_ERR(jd->jd_inode);
595 kfree(jd);
596 break;
597 }
598
599 spin_lock(&sdp->sd_jindex_spin);
600 jd->jd_jid = sdp->sd_journals++;
601 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
602 spin_unlock(&sdp->sd_jindex_spin);
603 }
604
605 mutex_unlock(&sdp->sd_jindex_mutex);
606
607 return error;
608}
609
1d12d175
BP
610/**
611 * check_journal_clean - Make sure a journal is clean for a spectator mount
612 * @sdp: The GFS2 superblock
613 * @jd: The journal descriptor
614 *
615 * Returns: 0 if the journal is clean or locked, else an error
616 */
617static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
618{
619 int error;
620 struct gfs2_holder j_gh;
621 struct gfs2_log_header_host head;
622 struct gfs2_inode *ip;
623
624 ip = GFS2_I(jd->jd_inode);
625 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
626 GL_EXACT | GL_NOCACHE, &j_gh);
627 if (error) {
628 fs_err(sdp, "Error locking journal for spectator mount.\n");
629 return -EPERM;
630 }
631 error = gfs2_jdesc_check(jd);
632 if (error) {
633 fs_err(sdp, "Error checking journal for spectator mount.\n");
634 goto out_unlock;
635 }
636 error = gfs2_find_jhead(jd, &head);
637 if (error) {
638 fs_err(sdp, "Error parsing journal for spectator mount.\n");
639 goto out_unlock;
640 }
641 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
642 error = -EPERM;
643 fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
644 "must not be a spectator.\n", jd->jd_jid);
645 }
646
647out_unlock:
648 gfs2_glock_dq_uninit(&j_gh);
649 return error;
650}
651
b3b94faa
DT
652static int init_journal(struct gfs2_sbd *sdp, int undo)
653{
2b0143b5 654 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa 655 struct gfs2_holder ji_gh;
5c676f6d 656 struct gfs2_inode *ip;
b3b94faa
DT
657 int jindex = 1;
658 int error = 0;
659
660 if (undo) {
661 jindex = 0;
fe64d517 662 goto fail_jinode_gh;
b3b94faa
DT
663 }
664
9b8df98f 665 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 666 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 667 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 668 return PTR_ERR(sdp->sd_jindex);
b3b94faa 669 }
b3b94faa
DT
670
671 /* Load in the journal index special file */
672
673 error = gfs2_jindex_hold(sdp, &ji_gh);
674 if (error) {
675 fs_err(sdp, "can't read journal index: %d\n", error);
676 goto fail;
677 }
678
0e5a9fb0 679 error = -EUSERS;
b3b94faa
DT
680 if (!gfs2_jindex_size(sdp)) {
681 fs_err(sdp, "no journals!\n");
907b9bce 682 goto fail_jindex;
b3b94faa
DT
683 }
684
685 if (sdp->sd_args.ar_spectator) {
686 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 687 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
688 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
689 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
690 } else {
691 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
692 fs_err(sdp, "can't mount journal #%u\n",
693 sdp->sd_lockstruct.ls_jid);
694 fs_err(sdp, "there are only %u journals (0 - %u)\n",
695 gfs2_jindex_size(sdp),
696 gfs2_jindex_size(sdp) - 1);
697 goto fail_jindex;
698 }
699 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
700
feaa7bba 701 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
702 &gfs2_journal_glops,
703 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
704 &sdp->sd_journal_gh);
705 if (error) {
706 fs_err(sdp, "can't acquire journal glock: %d\n", error);
707 goto fail_jindex;
708 }
709
feaa7bba
SW
710 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
711 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 712 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
713 &sdp->sd_jinode_gh);
714 if (error) {
715 fs_err(sdp, "can't acquire journal inode glock: %d\n",
716 error);
717 goto fail_journal_gh;
718 }
719
720 error = gfs2_jdesc_check(sdp->sd_jdesc);
721 if (error) {
722 fs_err(sdp, "my journal (%u) is bad: %d\n",
723 sdp->sd_jdesc->jd_jid, error);
724 goto fail_jinode_gh;
725 }
fd041f0b 726 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
727 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
728 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
729
730 /* Map the extents for this journal's blocks */
b50f227b 731 gfs2_map_journal_extents(sdp, sdp->sd_jdesc);
b3b94faa 732 }
63997775 733 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
734
735 if (sdp->sd_lockstruct.ls_first) {
736 unsigned int x;
737 for (x = 0; x < sdp->sd_journals; x++) {
1d12d175
BP
738 struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
739
740 if (sdp->sd_args.ar_spectator) {
741 error = check_journal_clean(sdp, jd);
742 if (error)
743 goto fail_jinode_gh;
744 continue;
745 }
746 error = gfs2_recover_journal(jd, true);
b3b94faa
DT
747 if (error) {
748 fs_err(sdp, "error recovering journal %u: %d\n",
749 x, error);
750 goto fail_jinode_gh;
751 }
752 }
753
f057f6cd 754 gfs2_others_may_mount(sdp);
b3b94faa 755 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 756 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
757 if (error) {
758 fs_err(sdp, "error recovering my journal: %d\n", error);
759 goto fail_jinode_gh;
760 }
761 }
762
086cc672 763 sdp->sd_log_idle = 1;
b3b94faa
DT
764 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
765 gfs2_glock_dq_uninit(&ji_gh);
766 jindex = 0;
2e60d768 767 INIT_WORK(&sdp->sd_freeze_work, gfs2_freeze_func);
b3b94faa
DT
768 return 0;
769
a91ea69f 770fail_jinode_gh:
b3b94faa
DT
771 if (!sdp->sd_args.ar_spectator)
772 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 773fail_journal_gh:
b3b94faa
DT
774 if (!sdp->sd_args.ar_spectator)
775 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 776fail_jindex:
b3b94faa
DT
777 gfs2_jindex_free(sdp);
778 if (jindex)
779 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 780fail:
f42faf4f 781 iput(sdp->sd_jindex);
b3b94faa
DT
782 return error;
783}
784
56aa72d0 785static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
786
787static int init_inodes(struct gfs2_sbd *sdp, int undo)
788{
b3b94faa 789 int error = 0;
2b0143b5 790 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa
DT
791
792 if (undo)
f42faf4f
SW
793 goto fail_qinode;
794
f42faf4f 795 error = init_journal(sdp, undo);
0e48e055 796 complete_all(&sdp->sd_journal_ready);
f42faf4f 797 if (error)
9b8df98f 798 goto fail;
b3b94faa 799
b3b94faa 800 /* Read in the master statfs inode */
9b8df98f 801 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
802 if (IS_ERR(sdp->sd_statfs_inode)) {
803 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 804 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 805 goto fail_journal;
b3b94faa
DT
806 }
807
808 /* Read in the resource index inode */
9b8df98f 809 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
810 if (IS_ERR(sdp->sd_rindex)) {
811 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
812 fs_err(sdp, "can't get resource index inode: %d\n", error);
813 goto fail_statfs;
814 }
cf45b752 815 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
816
817 /* Read in the quota inode */
9b8df98f 818 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
819 if (IS_ERR(sdp->sd_quota_inode)) {
820 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
821 fs_err(sdp, "can't get quota file inode: %d\n", error);
822 goto fail_rindex;
823 }
56aa72d0
JK
824 /*
825 * i_mutex on quota files is special. Since this inode is hidden system
826 * file, we are safe to define locking ourselves.
827 */
9902af79 828 lockdep_set_class(&sdp->sd_quota_inode->i_rwsem,
56aa72d0 829 &gfs2_quota_imutex_key);
a365fbf3
SW
830
831 error = gfs2_rindex_update(sdp);
832 if (error)
833 goto fail_qinode;
834
b3b94faa
DT
835 return 0;
836
f42faf4f
SW
837fail_qinode:
838 iput(sdp->sd_quota_inode);
f42faf4f 839fail_rindex:
b3b94faa 840 gfs2_clear_rgrpd(sdp);
f42faf4f 841 iput(sdp->sd_rindex);
f42faf4f
SW
842fail_statfs:
843 iput(sdp->sd_statfs_inode);
f42faf4f
SW
844fail_journal:
845 init_journal(sdp, UNDO);
f42faf4f 846fail:
b3b94faa
DT
847 return error;
848}
849
850static int init_per_node(struct gfs2_sbd *sdp, int undo)
851{
f42faf4f 852 struct inode *pn = NULL;
b3b94faa
DT
853 char buf[30];
854 int error = 0;
5c676f6d 855 struct gfs2_inode *ip;
2b0143b5 856 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa
DT
857
858 if (sdp->sd_args.ar_spectator)
859 return 0;
860
861 if (undo)
862 goto fail_qc_gh;
863
9b8df98f 864 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
865 if (IS_ERR(pn)) {
866 error = PTR_ERR(pn);
b3b94faa
DT
867 fs_err(sdp, "can't find per_node directory: %d\n", error);
868 return error;
869 }
870
b3b94faa 871 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
872 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
873 if (IS_ERR(sdp->sd_sc_inode)) {
874 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 875 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 876 goto fail;
b3b94faa
DT
877 }
878
b3b94faa 879 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
880 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
881 if (IS_ERR(sdp->sd_qc_inode)) {
882 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
883 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
884 goto fail_ut_i;
885 }
886
f42faf4f 887 iput(pn);
b3b94faa
DT
888 pn = NULL;
889
feaa7bba 890 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 891 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
892 &sdp->sd_sc_gh);
893 if (error) {
894 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 895 goto fail_qc_i;
b3b94faa
DT
896 }
897
feaa7bba 898 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 899 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
900 &sdp->sd_qc_gh);
901 if (error) {
902 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
903 goto fail_ut_gh;
904 }
905
906 return 0;
907
a91ea69f 908fail_qc_gh:
b3b94faa 909 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 910fail_ut_gh:
b3b94faa 911 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 912fail_qc_i:
f42faf4f 913 iput(sdp->sd_qc_inode);
a91ea69f 914fail_ut_i:
f42faf4f 915 iput(sdp->sd_sc_inode);
a91ea69f 916fail:
6fde2242 917 iput(pn);
b3b94faa
DT
918 return error;
919}
920
f057f6cd
SW
921static const match_table_t nolock_tokens = {
922 { Opt_jid, "jid=%d\n", },
923 { Opt_err, NULL },
924};
925
926static const struct lm_lockops nolock_ops = {
927 .lm_proto_name = "lock_nolock",
fc0e38da 928 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
929 .lm_tokens = &nolock_tokens,
930};
931
da755fdb
SW
932/**
933 * gfs2_lm_mount - mount a locking protocol
934 * @sdp: the filesystem
3ad2f3fb 935 * @args: mount arguments
da755fdb
SW
936 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
937 *
938 * Returns: errno
939 */
940
941static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
942{
f057f6cd
SW
943 const struct lm_lockops *lm;
944 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
945 struct gfs2_args *args = &sdp->sd_args;
946 const char *proto = sdp->sd_proto_name;
947 const char *table = sdp->sd_table_name;
f057f6cd
SW
948 char *o, *options;
949 int ret;
da755fdb 950
f057f6cd
SW
951 if (!strcmp("lock_nolock", proto)) {
952 lm = &nolock_ops;
953 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
954#ifdef CONFIG_GFS2_FS_LOCKING_DLM
955 } else if (!strcmp("lock_dlm", proto)) {
956 lm = &gfs2_dlm_ops;
957#endif
958 } else {
d77d1b58 959 pr_info("can't find protocol %s\n", proto);
f057f6cd
SW
960 return -ENOENT;
961 }
da755fdb
SW
962
963 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
964
f057f6cd
SW
965 ls->ls_ops = lm;
966 ls->ls_first = 1;
da755fdb 967
f057f6cd
SW
968 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
969 substring_t tmp[MAX_OPT_ARGS];
970 int token, option;
971
972 if (!o || !*o)
973 continue;
974
975 token = match_token(o, *lm->lm_tokens, tmp);
976 switch (token) {
977 case Opt_jid:
978 ret = match_int(&tmp[0], &option);
979 if (ret || option < 0)
980 goto hostdata_error;
ba6e9364
SW
981 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
982 ls->ls_jid = option;
f057f6cd
SW
983 break;
984 case Opt_id:
4875647a 985 case Opt_nodir:
2b88f7c5 986 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
987 break;
988 case Opt_first:
989 ret = match_int(&tmp[0], &option);
990 if (ret || (option != 0 && option != 1))
991 goto hostdata_error;
992 ls->ls_first = option;
993 break;
f057f6cd
SW
994 case Opt_err:
995 default:
996hostdata_error:
997 fs_info(sdp, "unknown hostdata (%s)\n", o);
998 return -EINVAL;
999 }
da755fdb
SW
1000 }
1001
f057f6cd
SW
1002 if (lm->lm_mount == NULL) {
1003 fs_info(sdp, "Now mounting FS...\n");
19237039 1004 complete_all(&sdp->sd_locking_init);
f057f6cd 1005 return 0;
da755fdb 1006 }
e0c2a9aa 1007 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1008 if (ret == 0)
1009 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1010 complete_all(&sdp->sd_locking_init);
f057f6cd 1011 return ret;
da755fdb
SW
1012}
1013
1014void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1015{
f057f6cd
SW
1016 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1017 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1018 lm->lm_unmount)
1019 lm->lm_unmount(sdp);
da755fdb 1020}
b3b94faa 1021
ba6e9364
SW
1022static int wait_on_journal(struct gfs2_sbd *sdp)
1023{
ba6e9364
SW
1024 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1025 return 0;
1026
74316201
N
1027 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, TASK_INTERRUPTIBLE)
1028 ? -EINTR : 0;
ba6e9364
SW
1029}
1030
8633ecfa
SW
1031void gfs2_online_uevent(struct gfs2_sbd *sdp)
1032{
1033 struct super_block *sb = sdp->sd_vfs;
1034 char ro[20];
1035 char spectator[20];
1036 char *envp[] = { ro, spectator, NULL };
1037 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1038 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1039 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1040}
1041
b3b94faa
DT
1042/**
1043 * fill_super - Read in superblock
1044 * @sb: The VFS superblock
1045 * @data: Mount options
1046 * @silent: Don't complain if it's not a GFS2 filesystem
1047 *
1048 * Returns: errno
1049 */
1050
f55073ff 1051static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1052{
1053 struct gfs2_sbd *sdp;
1054 struct gfs2_holder mount_gh;
1055 int error;
1056
1057 sdp = init_sbd(sb);
1058 if (!sdp) {
d77d1b58 1059 pr_warn("can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1060 return -ENOMEM;
1061 }
f55073ff 1062 sdp->sd_args = *args;
b3b94faa 1063
fe64d517 1064 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1065 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1066 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1067 }
6f04c1c7
SW
1068 if (sdp->sd_args.ar_posix_acl)
1069 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1070 if (sdp->sd_args.ar_nobarrier)
1071 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1072
9964afbb 1073 sb->s_flags |= MS_NOSEC;
719ee344
SW
1074 sb->s_magic = GFS2_MAGIC;
1075 sb->s_op = &gfs2_super_ops;
41ced6dc 1076 sb->s_d_op = &gfs2_dops;
719ee344 1077 sb->s_export_op = &gfs2_export_ops;
40b78a32 1078 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f 1079 sb->s_qcop = &gfs2_quotactl_ops;
de3b08d3 1080 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cc632e7f 1081 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1082 sb->s_time_gran = 1;
1083 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1084
1085 /* Set up the buffer cache and fill in some fake block size values
1086 to allow us to read-in the on-disk superblock. */
1087 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1088 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1089 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1090 GFS2_BASIC_BLOCK_SHIFT;
1091 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1092
5e687eac 1093 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1094 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1095 if (sdp->sd_args.ar_statfs_quantum) {
1096 sdp->sd_tune.gt_statfs_slow = 0;
1097 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1098 } else {
3d3c10f2
BM
1099 sdp->sd_tune.gt_statfs_slow = 1;
1100 sdp->sd_tune.gt_statfs_quantum = 30;
1101 }
48c2b613 1102
b3b94faa 1103 error = init_names(sdp, silent);
0d515210
BP
1104 if (error) {
1105 /* In this case, we haven't initialized sysfs, so we have to
1106 manually free the sdp. */
1107 free_percpu(sdp->sd_lkstats);
1108 kfree(sdp);
1109 sb->s_fs_info = NULL;
1110 return error;
1111 }
b3b94faa 1112
e0c2a9aa
DT
1113 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1114
b3b94faa 1115 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1116 /*
1117 * If we hit an error here, gfs2_sys_fs_add will have called function
1118 * kobject_put which causes the sysfs usage count to go to zero, which
1119 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1120 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1121 * kobject_put to free sdp.
1122 */
b3b94faa 1123 if (error)
0d515210
BP
1124 return error;
1125
1126 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1127
1128 error = gfs2_lm_mount(sdp, silent);
1129 if (error)
0d515210 1130 goto fail_debug;
b3b94faa
DT
1131
1132 error = init_locking(sdp, &mount_gh, DO);
1133 if (error)
1134 goto fail_lm;
1135
9b8df98f 1136 error = init_sb(sdp, silent);
b3b94faa
DT
1137 if (error)
1138 goto fail_locking;
b3b94faa 1139
ba6e9364
SW
1140 error = wait_on_journal(sdp);
1141 if (error)
1142 goto fail_sb;
1143
feb47ca9
SW
1144 /*
1145 * If user space has failed to join the cluster or some similar
1146 * failure has occurred, then the journal id will contain a
1147 * negative (error) number. This will then be returned to the
1148 * caller (of the mount syscall). We do this even for spectator
1149 * mounts (which just write a jid of 0 to indicate "ok" even though
1150 * the jid is unused in the spectator case)
1151 */
1152 if (sdp->sd_lockstruct.ls_jid < 0) {
1153 error = sdp->sd_lockstruct.ls_jid;
1154 sdp->sd_lockstruct.ls_jid = 0;
1155 goto fail_sb;
1156 }
1157
e0c2a9aa
DT
1158 if (sdp->sd_args.ar_spectator)
1159 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1160 sdp->sd_table_name);
1161 else
1162 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1163 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1164
b3b94faa
DT
1165 error = init_inodes(sdp, DO);
1166 if (error)
f42faf4f 1167 goto fail_sb;
b3b94faa
DT
1168
1169 error = init_per_node(sdp, DO);
1170 if (error)
1171 goto fail_inodes;
1172
1173 error = gfs2_statfs_init(sdp);
1174 if (error) {
1175 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1176 goto fail_per_node;
1177 }
1178
b3b94faa
DT
1179 if (!(sb->s_flags & MS_RDONLY)) {
1180 error = gfs2_make_fs_rw(sdp);
1181 if (error) {
1182 fs_err(sdp, "can't make FS RW: %d\n", error);
8ad151c2 1183 goto fail_per_node;
b3b94faa
DT
1184 }
1185 }
1186
1187 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1188 gfs2_online_uevent(sdp);
b3b94faa
DT
1189 return 0;
1190
a91ea69f 1191fail_per_node:
b3b94faa 1192 init_per_node(sdp, UNDO);
a91ea69f 1193fail_inodes:
b3b94faa 1194 init_inodes(sdp, UNDO);
a91ea69f 1195fail_sb:
9b8df98f
SW
1196 if (sdp->sd_root_dir)
1197 dput(sdp->sd_root_dir);
1198 if (sdp->sd_master_dir)
1199 dput(sdp->sd_master_dir);
e7c8707e
SW
1200 if (sb->s_root)
1201 dput(sb->s_root);
9b8df98f 1202 sb->s_root = NULL;
a91ea69f 1203fail_locking:
b3b94faa 1204 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1205fail_lm:
0e48e055 1206 complete_all(&sdp->sd_journal_ready);
fefc03bf 1207 gfs2_gl_hash_clear(sdp);
b3b94faa 1208 gfs2_lm_unmount(sdp);
0d515210 1209fail_debug:
5f882096 1210 gfs2_delete_debugfs_file(sdp);
a245769f 1211 free_percpu(sdp->sd_lkstats);
0d515210
BP
1212 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1213 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1214 gfs2_sys_fs_del(sdp);
5c676f6d 1215 sb->s_fs_info = NULL;
b3b94faa
DT
1216 return error;
1217}
1218
f55073ff 1219static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1220{
f55073ff
SW
1221 s->s_bdev = data;
1222 s->s_dev = s->s_bdev->bd_dev;
1223
1224 /*
1225 * We set the bdi here to the queue backing, file systems can
1226 * overwrite this in ->fill_super()
1227 */
1228 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1229 return 0;
86384605
AD
1230}
1231
f55073ff 1232static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1233{
1234 struct block_device *bdev = ptr;
1235 return (bdev == s->s_bdev);
1236}
1237
f55073ff 1238/**
8bcbbf00 1239 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1240 * @fs_type: The GFS2 filesystem type
1241 * @flags: Mount flags
1242 * @dev_name: The name of the device
1243 * @data: The mount arguments
f55073ff
SW
1244 *
1245 * Q. Why not use get_sb_bdev() ?
1246 * A. We need to select one of two root directories to mount, independent
1247 * of whether this is the initial, or subsequent, mount of this sb
1248 *
1249 * Returns: 0 or -ve on error
1250 */
1251
8bcbbf00
AV
1252static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1253 const char *dev_name, void *data)
f55073ff
SW
1254{
1255 struct block_device *bdev;
1256 struct super_block *s;
d4d77629 1257 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1258 int error;
1259 struct gfs2_args args;
1260 struct gfs2_sbd *sdp;
1261
1262 if (!(flags & MS_RDONLY))
1263 mode |= FMODE_WRITE;
1264
d4d77629 1265 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1266 if (IS_ERR(bdev))
8bcbbf00 1267 return ERR_CAST(bdev);
f55073ff
SW
1268
1269 /*
1270 * once the super is inserted into the list by sget, s_umount
1271 * will protect the lockfs code from trying to start a snapshot
1272 * while we are mounting
1273 */
1274 mutex_lock(&bdev->bd_fsfreeze_mutex);
1275 if (bdev->bd_fsfreeze_count > 0) {
1276 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1277 error = -EBUSY;
1278 goto error_bdev;
1279 }
9249e17f 1280 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1281 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1282 error = PTR_ERR(s);
1283 if (IS_ERR(s))
1284 goto error_bdev;
1285
dfe5b9ad
SW
1286 if (s->s_root) {
1287 /*
1288 * s_umount nests inside bd_mutex during
1289 * __invalidate_device(). blkdev_put() acquires
1290 * bd_mutex and can't be called under s_umount. Drop
1291 * s_umount temporarily. This is safe as we're
1292 * holding an active reference.
1293 */
1294 up_write(&s->s_umount);
d4d77629 1295 blkdev_put(bdev, mode);
dfe5b9ad 1296 down_write(&s->s_umount);
6de20eb0
AP
1297 } else {
1298 /* s_mode must be set before deactivate_locked_super calls */
1299 s->s_mode = mode;
dfe5b9ad 1300 }
8bcbbf00 1301
f55073ff
SW
1302 memset(&args, 0, sizeof(args));
1303 args.ar_quota = GFS2_QUOTA_DEFAULT;
1304 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1305 args.ar_commit = 30;
3d3c10f2
BM
1306 args.ar_statfs_quantum = 30;
1307 args.ar_quota_quantum = 60;
f55073ff
SW
1308 args.ar_errors = GFS2_ERRORS_DEFAULT;
1309
1310 error = gfs2_mount_args(&args, data);
1311 if (error) {
d77d1b58 1312 pr_warn("can't parse mount arguments\n");
8bcbbf00 1313 goto error_super;
f55073ff
SW
1314 }
1315
1316 if (s->s_root) {
1317 error = -EBUSY;
1318 if ((flags ^ s->s_flags) & MS_RDONLY)
1319 goto error_super;
f55073ff 1320 } else {
a1c6f057 1321 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
f55073ff
SW
1322 sb_set_blocksize(s, block_size(bdev));
1323 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1324 if (error)
1325 goto error_super;
f55073ff
SW
1326 s->s_flags |= MS_ACTIVE;
1327 bdev->bd_super = s;
1328 }
1329
1330 sdp = s->s_fs_info;
f55073ff 1331 if (args.ar_meta)
8bcbbf00 1332 return dget(sdp->sd_master_dir);
f55073ff 1333 else
8bcbbf00 1334 return dget(sdp->sd_root_dir);
f55073ff
SW
1335
1336error_super:
1337 deactivate_locked_super(s);
8bcbbf00 1338 return ERR_PTR(error);
f55073ff 1339error_bdev:
d4d77629 1340 blkdev_put(bdev, mode);
8bcbbf00 1341 return ERR_PTR(error);
f55073ff
SW
1342}
1343
f6d03139
SW
1344static int set_meta_super(struct super_block *s, void *ptr)
1345{
1346 return -EINVAL;
1347}
1348
8bcbbf00
AV
1349static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1350 int flags, const char *dev_name, void *data)
86384605 1351{
f6d03139 1352 struct super_block *s;
003dec89 1353 struct gfs2_sbd *sdp;
e24977d4 1354 struct path path;
86384605 1355 int error;
907b9bce 1356
e24977d4 1357 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1358 if (error) {
d77d1b58
JP
1359 pr_warn("path_lookup on %s returned error %d\n",
1360 dev_name, error);
8bcbbf00 1361 return ERR_PTR(error);
86384605 1362 }
9249e17f 1363 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
fc64005c 1364 path.dentry->d_sb->s_bdev);
e24977d4 1365 path_put(&path);
f6d03139 1366 if (IS_ERR(s)) {
d77d1b58 1367 pr_warn("gfs2 mount does not exist\n");
8bcbbf00 1368 return ERR_CAST(s);
86384605 1369 }
f55073ff
SW
1370 if ((flags ^ s->s_flags) & MS_RDONLY) {
1371 deactivate_locked_super(s);
8bcbbf00 1372 return ERR_PTR(-EBUSY);
f55073ff 1373 }
f6d03139 1374 sdp = s->s_fs_info;
8bcbbf00 1375 return dget(sdp->sd_master_dir);
b3b94faa
DT
1376}
1377
419c93e0 1378static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1379{
1380 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1381
1382 if (sdp == NULL) {
1383 kill_block_super(sb);
1384 return;
3af165ac 1385 }
88a19ad0 1386
24972557 1387 gfs2_log_flush(sdp, NULL, SYNC_FLUSH);
88a19ad0
SW
1388 dput(sdp->sd_root_dir);
1389 dput(sdp->sd_master_dir);
1390 sdp->sd_root_dir = NULL;
1391 sdp->sd_master_dir = NULL;
fefc03bf 1392 shrink_dcache_sb(sb);
88a19ad0 1393 gfs2_delete_debugfs_file(sdp);
a245769f 1394 free_percpu(sdp->sd_lkstats);
0d515210 1395 kill_block_super(sb);
86384605
AD
1396}
1397
b3b94faa
DT
1398struct file_system_type gfs2_fs_type = {
1399 .name = "gfs2",
1400 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1401 .mount = gfs2_mount,
419c93e0
SW
1402 .kill_sb = gfs2_kill_sb,
1403 .owner = THIS_MODULE,
1404};
7f78e035 1405MODULE_ALIAS_FS("gfs2");
419c93e0
SW
1406
1407struct file_system_type gfs2meta_fs_type = {
1408 .name = "gfs2meta",
1409 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1410 .mount = gfs2_mount_meta,
b3b94faa
DT
1411 .owner = THIS_MODULE,
1412};
7f78e035 1413MODULE_ALIAS_FS("gfs2meta");