]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/gfs2/ops_fstype.c
Merge tag 'v3.16-rc5' into sched/core, to refresh the branch before applying bigger...
[mirror_ubuntu-artful-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
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;
115 mapping->backing_dev_info = sb->s_bdi;
116 mapping->writeback_index = 0;
117
b3b94faa 118 spin_lock_init(&sdp->sd_log_lock);
5e687eac 119 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa 120 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
d7b616e2 121 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
45138990 122 spin_lock_init(&sdp->sd_ordered_lock);
b3b94faa 123
5e687eac
BM
124 init_waitqueue_head(&sdp->sd_log_waitq);
125 init_waitqueue_head(&sdp->sd_logd_waitq);
d6a079e8 126 spin_lock_init(&sdp->sd_ail_lock);
b3b94faa
DT
127 INIT_LIST_HEAD(&sdp->sd_ail1_list);
128 INIT_LIST_HEAD(&sdp->sd_ail2_list);
129
484adff8 130 init_rwsem(&sdp->sd_log_flush_lock);
16615be1
SW
131 atomic_set(&sdp->sd_log_in_flight, 0);
132 init_waitqueue_head(&sdp->sd_log_flush_wait);
24972557
BM
133 init_waitqueue_head(&sdp->sd_log_frozen_wait);
134 atomic_set(&sdp->sd_log_freeze, 0);
135 atomic_set(&sdp->sd_frozen_root, 0);
136 init_waitqueue_head(&sdp->sd_frozen_root_wait);
b3b94faa 137
b3b94faa
DT
138 return sdp;
139}
140
b3b94faa 141
9b8df98f
SW
142/**
143 * gfs2_check_sb - Check superblock
144 * @sdp: the filesystem
145 * @sb: The superblock
146 * @silent: Don't print a message if the check fails
147 *
148 * Checks the version code of the FS is one that we understand how to
149 * read and that the sizes of the various on-disk structures have not
150 * changed.
151 */
152
32e471ef 153static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 154{
32e471ef
SW
155 struct gfs2_sb_host *sb = &sdp->sd_sb;
156
9b8df98f
SW
157 if (sb->sb_magic != GFS2_MAGIC ||
158 sb->sb_type != GFS2_METATYPE_SB) {
159 if (!silent)
d77d1b58 160 pr_warn("not a GFS2 filesystem\n");
9b8df98f
SW
161 return -EINVAL;
162 }
163
164 /* If format numbers match exactly, we're done. */
165
166 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
167 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
168 return 0;
169
c80dbb58 170 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 171
c80dbb58 172 return -EINVAL;
9b8df98f
SW
173}
174
175static void end_bio_io_page(struct bio *bio, int error)
176{
177 struct page *page = bio->bi_private;
178
179 if (!error)
180 SetPageUptodate(page);
181 else
d77d1b58 182 pr_warn("error %d reading superblock\n", error);
9b8df98f
SW
183 unlock_page(page);
184}
185
32e471ef 186static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 187{
32e471ef
SW
188 struct gfs2_sb_host *sb = &sdp->sd_sb;
189 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
190 const struct gfs2_sb *str = buf;
191
192 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
193 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
194 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
195 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
196 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
197 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
198 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
199 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
200 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
201 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
202 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
203
204 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
205 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 206 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
207}
208
209/**
210 * gfs2_read_super - Read the gfs2 super block from disk
211 * @sdp: The GFS2 super block
212 * @sector: The location of the super block
213 * @error: The error code to return
214 *
215 * This uses the bio functions to read the super block from disk
216 * because we want to be 100% sure that we never read cached data.
217 * A super block is read twice only during each GFS2 mount and is
218 * never written to by the filesystem. The first time its read no
219 * locks are held, and the only details which are looked at are those
220 * relating to the locking protocol. Once locking is up and working,
221 * the sb is read again under the lock to establish the location of
222 * the master directory (contains pointers to journals etc) and the
223 * root directory.
224 *
225 * Returns: 0 on success or error
226 */
227
32e471ef 228static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
229{
230 struct super_block *sb = sdp->sd_vfs;
231 struct gfs2_sb *p;
232 struct page *page;
233 struct bio *bio;
234
235 page = alloc_page(GFP_NOFS);
236 if (unlikely(!page))
ac3beb6a 237 return -ENOMEM;
9b8df98f
SW
238
239 ClearPageUptodate(page);
240 ClearPageDirty(page);
241 lock_page(page);
242
243 bio = bio_alloc(GFP_NOFS, 1);
4f024f37 244 bio->bi_iter.bi_sector = sector * (sb->s_blocksize >> 9);
9b8df98f
SW
245 bio->bi_bdev = sb->s_bdev;
246 bio_add_page(bio, page, PAGE_SIZE, 0);
247
248 bio->bi_end_io = end_bio_io_page;
249 bio->bi_private = page;
20ed0535 250 submit_bio(READ_SYNC | REQ_META, 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
356 return 0;
357}
358
b3b94faa
DT
359static int init_names(struct gfs2_sbd *sdp, int silent)
360{
b3b94faa
DT
361 char *proto, *table;
362 int error = 0;
363
364 proto = sdp->sd_args.ar_lockproto;
365 table = sdp->sd_args.ar_locktable;
366
367 /* Try to autodetect */
368
369 if (!proto[0] || !table[0]) {
32e471ef 370 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
371 if (error)
372 return error;
3cf1e7be 373
b3b94faa 374 if (!proto[0])
3cf1e7be 375 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 376 if (!table[0])
3cf1e7be 377 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
378 }
379
380 if (!table[0])
381 table = sdp->sd_vfs->s_id;
382
00377d8e
JD
383 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
384 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 385
5d35e31f
DC
386 table = sdp->sd_table_name;
387 while ((table = strchr(table, '/')))
b35997d4
RP
388 *table = '_';
389
b3b94faa
DT
390 return error;
391}
392
393static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
394 int undo)
395{
b3b94faa
DT
396 int error = 0;
397
398 if (undo)
399 goto fail_trans;
400
b3b94faa
DT
401 error = gfs2_glock_nq_num(sdp,
402 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
403 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
404 mount_gh);
405 if (error) {
406 fs_err(sdp, "can't acquire mount glock: %d\n", error);
407 goto fail;
408 }
409
410 error = gfs2_glock_nq_num(sdp,
411 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
412 LM_ST_SHARED,
579b78a4 413 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
414 &sdp->sd_live_gh);
415 if (error) {
416 fs_err(sdp, "can't acquire live glock: %d\n", error);
417 goto fail_mount;
418 }
419
420 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
421 CREATE, &sdp->sd_rename_gl);
422 if (error) {
423 fs_err(sdp, "can't create rename glock: %d\n", error);
424 goto fail_live;
425 }
426
24972557
BM
427 error = gfs2_glock_get(sdp, GFS2_FREEZE_LOCK, &gfs2_freeze_glops,
428 CREATE, &sdp->sd_freeze_gl);
b3b94faa
DT
429 if (error) {
430 fs_err(sdp, "can't create transaction glock: %d\n", error);
431 goto fail_rename;
432 }
b3b94faa
DT
433
434 return 0;
435
feaa7bba 436fail_trans:
24972557 437 gfs2_glock_put(sdp->sd_freeze_gl);
feaa7bba 438fail_rename:
b3b94faa 439 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 440fail_live:
b3b94faa 441 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 442fail_mount:
b3b94faa 443 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 444fail:
b3b94faa
DT
445 return error;
446}
447
9b8df98f
SW
448static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
449 u64 no_addr, const char *name)
f42faf4f 450{
9b8df98f
SW
451 struct gfs2_sbd *sdp = sb->s_fs_info;
452 struct dentry *dentry;
453 struct inode *inode;
454
44ad37d6 455 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
9b8df98f
SW
456 if (IS_ERR(inode)) {
457 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
458 return PTR_ERR(inode);
459 }
48fde701 460 dentry = d_make_root(inode);
9b8df98f
SW
461 if (!dentry) {
462 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
463 return -ENOMEM;
464 }
9b8df98f
SW
465 *dptr = dentry;
466 return 0;
f42faf4f
SW
467}
468
9b8df98f 469static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
470{
471 struct super_block *sb = sdp->sd_vfs;
472 struct gfs2_holder sb_gh;
dbb7cae2 473 u64 no_addr;
9b8df98f 474 int ret;
b3b94faa 475
9b8df98f
SW
476 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
477 LM_ST_SHARED, 0, &sb_gh);
478 if (ret) {
479 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
480 return ret;
b3b94faa 481 }
907b9bce 482
6bac243f 483 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
484 if (ret) {
485 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
486 goto out;
487 }
488
489 /* Set up the buffer cache and SB for real */
e1defc4f 490 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 491 ret = -EINVAL;
b3b94faa
DT
492 fs_err(sdp, "FS block size (%u) is too small for device "
493 "block size (%u)\n",
e1defc4f 494 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
495 goto out;
496 }
497 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 498 ret = -EINVAL;
b3b94faa
DT
499 fs_err(sdp, "FS block size (%u) is too big for machine "
500 "page size (%u)\n",
501 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
502 goto out;
503 }
b3b94faa
DT
504 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
505
f42faf4f 506 /* Get the root inode */
dbb7cae2 507 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
508 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
509 if (ret)
f42faf4f 510 goto out;
b3b94faa 511
9b8df98f
SW
512 /* Get the master inode */
513 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
514 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
515 if (ret) {
516 dput(sdp->sd_root_dir);
517 goto out;
518 }
519 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 520out:
b3b94faa 521 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 522 return ret;
b3b94faa
DT
523}
524
f057f6cd 525static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 526{
f057f6cd
SW
527 char *message = "FIRSTMOUNT=Done";
528 char *envp[] = { message, NULL };
e0c2a9aa
DT
529
530 fs_info(sdp, "first mount done, others may mount\n");
531
532 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
533 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
534
f057f6cd 535 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
536}
537
b5289681
SW
538/**
539 * gfs2_jindex_hold - Grab a lock on the jindex
540 * @sdp: The GFS2 superblock
541 * @ji_gh: the holder for the jindex glock
542 *
543 * Returns: errno
544 */
545
546static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
547{
548 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
549 struct qstr name;
550 char buf[20];
551 struct gfs2_jdesc *jd;
552 int error;
553
554 name.name = buf;
555
556 mutex_lock(&sdp->sd_jindex_mutex);
557
558 for (;;) {
559 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
560 if (error)
561 break;
562
563 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
564 name.hash = gfs2_disk_hash(name.name, name.len);
565
566 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
567 if (error == -ENOENT) {
568 error = 0;
569 break;
570 }
571
572 gfs2_glock_dq_uninit(ji_gh);
573
574 if (error)
575 break;
576
577 error = -ENOMEM;
578 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
579 if (!jd)
580 break;
581
582 INIT_LIST_HEAD(&jd->extent_list);
a17d758b
BP
583 INIT_LIST_HEAD(&jd->jd_revoke_list);
584
6ecd7c2d 585 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
586 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
587 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
588 if (!jd->jd_inode)
589 error = -ENOENT;
590 else
591 error = PTR_ERR(jd->jd_inode);
592 kfree(jd);
593 break;
594 }
595
596 spin_lock(&sdp->sd_jindex_spin);
597 jd->jd_jid = sdp->sd_journals++;
598 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
599 spin_unlock(&sdp->sd_jindex_spin);
600 }
601
602 mutex_unlock(&sdp->sd_jindex_mutex);
603
604 return error;
605}
606
1d12d175
BP
607/**
608 * check_journal_clean - Make sure a journal is clean for a spectator mount
609 * @sdp: The GFS2 superblock
610 * @jd: The journal descriptor
611 *
612 * Returns: 0 if the journal is clean or locked, else an error
613 */
614static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
615{
616 int error;
617 struct gfs2_holder j_gh;
618 struct gfs2_log_header_host head;
619 struct gfs2_inode *ip;
620
621 ip = GFS2_I(jd->jd_inode);
622 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
623 GL_EXACT | GL_NOCACHE, &j_gh);
624 if (error) {
625 fs_err(sdp, "Error locking journal for spectator mount.\n");
626 return -EPERM;
627 }
628 error = gfs2_jdesc_check(jd);
629 if (error) {
630 fs_err(sdp, "Error checking journal for spectator mount.\n");
631 goto out_unlock;
632 }
633 error = gfs2_find_jhead(jd, &head);
634 if (error) {
635 fs_err(sdp, "Error parsing journal for spectator mount.\n");
636 goto out_unlock;
637 }
638 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
639 error = -EPERM;
640 fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
641 "must not be a spectator.\n", jd->jd_jid);
642 }
643
644out_unlock:
645 gfs2_glock_dq_uninit(&j_gh);
646 return error;
647}
648
b3b94faa
DT
649static int init_journal(struct gfs2_sbd *sdp, int undo)
650{
9b8df98f 651 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 652 struct gfs2_holder ji_gh;
5c676f6d 653 struct gfs2_inode *ip;
b3b94faa
DT
654 int jindex = 1;
655 int error = 0;
656
657 if (undo) {
658 jindex = 0;
fe64d517 659 goto fail_jinode_gh;
b3b94faa
DT
660 }
661
9b8df98f 662 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 663 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 664 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 665 return PTR_ERR(sdp->sd_jindex);
b3b94faa 666 }
b3b94faa
DT
667
668 /* Load in the journal index special file */
669
670 error = gfs2_jindex_hold(sdp, &ji_gh);
671 if (error) {
672 fs_err(sdp, "can't read journal index: %d\n", error);
673 goto fail;
674 }
675
0e5a9fb0 676 error = -EUSERS;
b3b94faa
DT
677 if (!gfs2_jindex_size(sdp)) {
678 fs_err(sdp, "no journals!\n");
907b9bce 679 goto fail_jindex;
b3b94faa
DT
680 }
681
682 if (sdp->sd_args.ar_spectator) {
683 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 684 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
685 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
686 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
687 } else {
688 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
689 fs_err(sdp, "can't mount journal #%u\n",
690 sdp->sd_lockstruct.ls_jid);
691 fs_err(sdp, "there are only %u journals (0 - %u)\n",
692 gfs2_jindex_size(sdp),
693 gfs2_jindex_size(sdp) - 1);
694 goto fail_jindex;
695 }
696 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
697
feaa7bba 698 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
699 &gfs2_journal_glops,
700 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
701 &sdp->sd_journal_gh);
702 if (error) {
703 fs_err(sdp, "can't acquire journal glock: %d\n", error);
704 goto fail_jindex;
705 }
706
feaa7bba
SW
707 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
708 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 709 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
710 &sdp->sd_jinode_gh);
711 if (error) {
712 fs_err(sdp, "can't acquire journal inode glock: %d\n",
713 error);
714 goto fail_journal_gh;
715 }
716
717 error = gfs2_jdesc_check(sdp->sd_jdesc);
718 if (error) {
719 fs_err(sdp, "my journal (%u) is bad: %d\n",
720 sdp->sd_jdesc->jd_jid, error);
721 goto fail_jinode_gh;
722 }
fd041f0b 723 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
724 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
725 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
726
727 /* Map the extents for this journal's blocks */
b50f227b 728 gfs2_map_journal_extents(sdp, sdp->sd_jdesc);
b3b94faa 729 }
63997775 730 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
731
732 if (sdp->sd_lockstruct.ls_first) {
733 unsigned int x;
734 for (x = 0; x < sdp->sd_journals; x++) {
1d12d175
BP
735 struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
736
737 if (sdp->sd_args.ar_spectator) {
738 error = check_journal_clean(sdp, jd);
739 if (error)
740 goto fail_jinode_gh;
741 continue;
742 }
743 error = gfs2_recover_journal(jd, true);
b3b94faa
DT
744 if (error) {
745 fs_err(sdp, "error recovering journal %u: %d\n",
746 x, error);
747 goto fail_jinode_gh;
748 }
749 }
750
f057f6cd 751 gfs2_others_may_mount(sdp);
b3b94faa 752 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 753 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
754 if (error) {
755 fs_err(sdp, "error recovering my journal: %d\n", error);
756 goto fail_jinode_gh;
757 }
758 }
759
760 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
761 gfs2_glock_dq_uninit(&ji_gh);
762 jindex = 0;
24972557
BM
763 if (!sdp->sd_args.ar_spectator) {
764 error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, 0,
765 &sdp->sd_thaw_gh);
766 if (error) {
767 fs_err(sdp, "can't acquire freeze glock: %d\n", error);
768 goto fail_jinode_gh;
769 }
770 }
771 gfs2_glock_dq_uninit(&sdp->sd_thaw_gh);
b3b94faa
DT
772 return 0;
773
a91ea69f 774fail_jinode_gh:
b3b94faa
DT
775 if (!sdp->sd_args.ar_spectator)
776 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 777fail_journal_gh:
b3b94faa
DT
778 if (!sdp->sd_args.ar_spectator)
779 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 780fail_jindex:
b3b94faa
DT
781 gfs2_jindex_free(sdp);
782 if (jindex)
783 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 784fail:
f42faf4f 785 iput(sdp->sd_jindex);
b3b94faa
DT
786 return error;
787}
788
56aa72d0 789static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
790
791static int init_inodes(struct gfs2_sbd *sdp, int undo)
792{
b3b94faa 793 int error = 0;
9b8df98f 794 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
795
796 if (undo)
f42faf4f
SW
797 goto fail_qinode;
798
f42faf4f 799 error = init_journal(sdp, undo);
0e48e055 800 complete_all(&sdp->sd_journal_ready);
f42faf4f 801 if (error)
9b8df98f 802 goto fail;
b3b94faa 803
b3b94faa 804 /* Read in the master statfs inode */
9b8df98f 805 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
806 if (IS_ERR(sdp->sd_statfs_inode)) {
807 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 808 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 809 goto fail_journal;
b3b94faa
DT
810 }
811
812 /* Read in the resource index inode */
9b8df98f 813 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
814 if (IS_ERR(sdp->sd_rindex)) {
815 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
816 fs_err(sdp, "can't get resource index inode: %d\n", error);
817 goto fail_statfs;
818 }
cf45b752 819 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
820
821 /* Read in the quota inode */
9b8df98f 822 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
823 if (IS_ERR(sdp->sd_quota_inode)) {
824 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
825 fs_err(sdp, "can't get quota file inode: %d\n", error);
826 goto fail_rindex;
827 }
56aa72d0
JK
828 /*
829 * i_mutex on quota files is special. Since this inode is hidden system
830 * file, we are safe to define locking ourselves.
831 */
832 lockdep_set_class(&sdp->sd_quota_inode->i_mutex,
833 &gfs2_quota_imutex_key);
a365fbf3
SW
834
835 error = gfs2_rindex_update(sdp);
836 if (error)
837 goto fail_qinode;
838
b3b94faa
DT
839 return 0;
840
f42faf4f
SW
841fail_qinode:
842 iput(sdp->sd_quota_inode);
f42faf4f 843fail_rindex:
b3b94faa 844 gfs2_clear_rgrpd(sdp);
f42faf4f 845 iput(sdp->sd_rindex);
f42faf4f
SW
846fail_statfs:
847 iput(sdp->sd_statfs_inode);
f42faf4f
SW
848fail_journal:
849 init_journal(sdp, UNDO);
f42faf4f 850fail:
b3b94faa
DT
851 return error;
852}
853
854static int init_per_node(struct gfs2_sbd *sdp, int undo)
855{
f42faf4f 856 struct inode *pn = NULL;
b3b94faa
DT
857 char buf[30];
858 int error = 0;
5c676f6d 859 struct gfs2_inode *ip;
9b8df98f 860 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
861
862 if (sdp->sd_args.ar_spectator)
863 return 0;
864
865 if (undo)
866 goto fail_qc_gh;
867
9b8df98f 868 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
869 if (IS_ERR(pn)) {
870 error = PTR_ERR(pn);
b3b94faa
DT
871 fs_err(sdp, "can't find per_node directory: %d\n", error);
872 return error;
873 }
874
b3b94faa 875 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
876 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
877 if (IS_ERR(sdp->sd_sc_inode)) {
878 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 879 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 880 goto fail;
b3b94faa
DT
881 }
882
b3b94faa 883 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
884 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
885 if (IS_ERR(sdp->sd_qc_inode)) {
886 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
887 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
888 goto fail_ut_i;
889 }
890
f42faf4f 891 iput(pn);
b3b94faa
DT
892 pn = NULL;
893
feaa7bba 894 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 895 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
896 &sdp->sd_sc_gh);
897 if (error) {
898 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 899 goto fail_qc_i;
b3b94faa
DT
900 }
901
feaa7bba 902 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 903 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
904 &sdp->sd_qc_gh);
905 if (error) {
906 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
907 goto fail_ut_gh;
908 }
909
910 return 0;
911
a91ea69f 912fail_qc_gh:
b3b94faa 913 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 914fail_ut_gh:
b3b94faa 915 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 916fail_qc_i:
f42faf4f 917 iput(sdp->sd_qc_inode);
a91ea69f 918fail_ut_i:
f42faf4f 919 iput(sdp->sd_sc_inode);
a91ea69f 920fail:
b3b94faa 921 if (pn)
f42faf4f 922 iput(pn);
b3b94faa
DT
923 return error;
924}
925
f057f6cd
SW
926static const match_table_t nolock_tokens = {
927 { Opt_jid, "jid=%d\n", },
928 { Opt_err, NULL },
929};
930
931static const struct lm_lockops nolock_ops = {
932 .lm_proto_name = "lock_nolock",
fc0e38da 933 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
934 .lm_tokens = &nolock_tokens,
935};
936
da755fdb
SW
937/**
938 * gfs2_lm_mount - mount a locking protocol
939 * @sdp: the filesystem
3ad2f3fb 940 * @args: mount arguments
da755fdb
SW
941 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
942 *
943 * Returns: errno
944 */
945
946static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
947{
f057f6cd
SW
948 const struct lm_lockops *lm;
949 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
950 struct gfs2_args *args = &sdp->sd_args;
951 const char *proto = sdp->sd_proto_name;
952 const char *table = sdp->sd_table_name;
f057f6cd
SW
953 char *o, *options;
954 int ret;
da755fdb 955
f057f6cd
SW
956 if (!strcmp("lock_nolock", proto)) {
957 lm = &nolock_ops;
958 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
959#ifdef CONFIG_GFS2_FS_LOCKING_DLM
960 } else if (!strcmp("lock_dlm", proto)) {
961 lm = &gfs2_dlm_ops;
962#endif
963 } else {
d77d1b58 964 pr_info("can't find protocol %s\n", proto);
f057f6cd
SW
965 return -ENOENT;
966 }
da755fdb
SW
967
968 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
969
f057f6cd
SW
970 ls->ls_ops = lm;
971 ls->ls_first = 1;
da755fdb 972
f057f6cd
SW
973 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
974 substring_t tmp[MAX_OPT_ARGS];
975 int token, option;
976
977 if (!o || !*o)
978 continue;
979
980 token = match_token(o, *lm->lm_tokens, tmp);
981 switch (token) {
982 case Opt_jid:
983 ret = match_int(&tmp[0], &option);
984 if (ret || option < 0)
985 goto hostdata_error;
ba6e9364
SW
986 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
987 ls->ls_jid = option;
f057f6cd
SW
988 break;
989 case Opt_id:
4875647a 990 case Opt_nodir:
2b88f7c5 991 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
992 break;
993 case Opt_first:
994 ret = match_int(&tmp[0], &option);
995 if (ret || (option != 0 && option != 1))
996 goto hostdata_error;
997 ls->ls_first = option;
998 break;
f057f6cd
SW
999 case Opt_err:
1000 default:
1001hostdata_error:
1002 fs_info(sdp, "unknown hostdata (%s)\n", o);
1003 return -EINVAL;
1004 }
da755fdb
SW
1005 }
1006
f057f6cd
SW
1007 if (lm->lm_mount == NULL) {
1008 fs_info(sdp, "Now mounting FS...\n");
19237039 1009 complete_all(&sdp->sd_locking_init);
f057f6cd 1010 return 0;
da755fdb 1011 }
e0c2a9aa 1012 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1013 if (ret == 0)
1014 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1015 complete_all(&sdp->sd_locking_init);
f057f6cd 1016 return ret;
da755fdb
SW
1017}
1018
1019void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1020{
f057f6cd
SW
1021 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1022 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1023 lm->lm_unmount)
1024 lm->lm_unmount(sdp);
da755fdb 1025}
b3b94faa 1026
ba6e9364
SW
1027static int gfs2_journalid_wait(void *word)
1028{
1029 if (signal_pending(current))
1030 return -EINTR;
1031 schedule();
1032 return 0;
1033}
1034
1035static int wait_on_journal(struct gfs2_sbd *sdp)
1036{
ba6e9364
SW
1037 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1038 return 0;
1039
1040 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1041}
1042
8633ecfa
SW
1043void gfs2_online_uevent(struct gfs2_sbd *sdp)
1044{
1045 struct super_block *sb = sdp->sd_vfs;
1046 char ro[20];
1047 char spectator[20];
1048 char *envp[] = { ro, spectator, NULL };
1049 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1050 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1051 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1052}
1053
b3b94faa
DT
1054/**
1055 * fill_super - Read in superblock
1056 * @sb: The VFS superblock
1057 * @data: Mount options
1058 * @silent: Don't complain if it's not a GFS2 filesystem
1059 *
1060 * Returns: errno
1061 */
1062
f55073ff 1063static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1064{
1065 struct gfs2_sbd *sdp;
1066 struct gfs2_holder mount_gh;
1067 int error;
1068
1069 sdp = init_sbd(sb);
1070 if (!sdp) {
d77d1b58 1071 pr_warn("can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1072 return -ENOMEM;
1073 }
f55073ff 1074 sdp->sd_args = *args;
b3b94faa 1075
fe64d517 1076 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1077 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1078 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1079 }
6f04c1c7
SW
1080 if (sdp->sd_args.ar_posix_acl)
1081 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1082 if (sdp->sd_args.ar_nobarrier)
1083 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1084
9964afbb 1085 sb->s_flags |= MS_NOSEC;
719ee344
SW
1086 sb->s_magic = GFS2_MAGIC;
1087 sb->s_op = &gfs2_super_ops;
41ced6dc 1088 sb->s_d_op = &gfs2_dops;
719ee344 1089 sb->s_export_op = &gfs2_export_ops;
40b78a32 1090 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1091 sb->s_qcop = &gfs2_quotactl_ops;
1092 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1093 sb->s_time_gran = 1;
1094 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1095
1096 /* Set up the buffer cache and fill in some fake block size values
1097 to allow us to read-in the on-disk superblock. */
1098 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1099 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1100 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1101 GFS2_BASIC_BLOCK_SHIFT;
1102 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1103
5e687eac 1104 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1105 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1106 if (sdp->sd_args.ar_statfs_quantum) {
1107 sdp->sd_tune.gt_statfs_slow = 0;
1108 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1109 } else {
3d3c10f2
BM
1110 sdp->sd_tune.gt_statfs_slow = 1;
1111 sdp->sd_tune.gt_statfs_quantum = 30;
1112 }
48c2b613 1113
b3b94faa 1114 error = init_names(sdp, silent);
0d515210
BP
1115 if (error) {
1116 /* In this case, we haven't initialized sysfs, so we have to
1117 manually free the sdp. */
1118 free_percpu(sdp->sd_lkstats);
1119 kfree(sdp);
1120 sb->s_fs_info = NULL;
1121 return error;
1122 }
b3b94faa 1123
e0c2a9aa
DT
1124 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1125
b3b94faa 1126 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1127 /*
1128 * If we hit an error here, gfs2_sys_fs_add will have called function
1129 * kobject_put which causes the sysfs usage count to go to zero, which
1130 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1131 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1132 * kobject_put to free sdp.
1133 */
b3b94faa 1134 if (error)
0d515210
BP
1135 return error;
1136
1137 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1138
1139 error = gfs2_lm_mount(sdp, silent);
1140 if (error)
0d515210 1141 goto fail_debug;
b3b94faa
DT
1142
1143 error = init_locking(sdp, &mount_gh, DO);
1144 if (error)
1145 goto fail_lm;
1146
9b8df98f 1147 error = init_sb(sdp, silent);
b3b94faa
DT
1148 if (error)
1149 goto fail_locking;
b3b94faa 1150
ba6e9364
SW
1151 error = wait_on_journal(sdp);
1152 if (error)
1153 goto fail_sb;
1154
feb47ca9
SW
1155 /*
1156 * If user space has failed to join the cluster or some similar
1157 * failure has occurred, then the journal id will contain a
1158 * negative (error) number. This will then be returned to the
1159 * caller (of the mount syscall). We do this even for spectator
1160 * mounts (which just write a jid of 0 to indicate "ok" even though
1161 * the jid is unused in the spectator case)
1162 */
1163 if (sdp->sd_lockstruct.ls_jid < 0) {
1164 error = sdp->sd_lockstruct.ls_jid;
1165 sdp->sd_lockstruct.ls_jid = 0;
1166 goto fail_sb;
1167 }
1168
e0c2a9aa
DT
1169 if (sdp->sd_args.ar_spectator)
1170 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1171 sdp->sd_table_name);
1172 else
1173 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1174 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1175
b3b94faa
DT
1176 error = init_inodes(sdp, DO);
1177 if (error)
f42faf4f 1178 goto fail_sb;
b3b94faa
DT
1179
1180 error = init_per_node(sdp, DO);
1181 if (error)
1182 goto fail_inodes;
1183
1184 error = gfs2_statfs_init(sdp);
1185 if (error) {
1186 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1187 goto fail_per_node;
1188 }
1189
b3b94faa
DT
1190 if (!(sb->s_flags & MS_RDONLY)) {
1191 error = gfs2_make_fs_rw(sdp);
1192 if (error) {
1193 fs_err(sdp, "can't make FS RW: %d\n", error);
8ad151c2 1194 goto fail_per_node;
b3b94faa
DT
1195 }
1196 }
1197
1198 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1199 gfs2_online_uevent(sdp);
b3b94faa
DT
1200 return 0;
1201
a91ea69f 1202fail_per_node:
b3b94faa 1203 init_per_node(sdp, UNDO);
a91ea69f 1204fail_inodes:
b3b94faa 1205 init_inodes(sdp, UNDO);
a91ea69f 1206fail_sb:
9b8df98f
SW
1207 if (sdp->sd_root_dir)
1208 dput(sdp->sd_root_dir);
1209 if (sdp->sd_master_dir)
1210 dput(sdp->sd_master_dir);
e7c8707e
SW
1211 if (sb->s_root)
1212 dput(sb->s_root);
9b8df98f 1213 sb->s_root = NULL;
a91ea69f 1214fail_locking:
b3b94faa 1215 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1216fail_lm:
0e48e055 1217 complete_all(&sdp->sd_journal_ready);
fefc03bf 1218 gfs2_gl_hash_clear(sdp);
b3b94faa 1219 gfs2_lm_unmount(sdp);
0d515210 1220fail_debug:
5f882096 1221 gfs2_delete_debugfs_file(sdp);
a245769f 1222 free_percpu(sdp->sd_lkstats);
0d515210
BP
1223 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1224 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1225 gfs2_sys_fs_del(sdp);
5c676f6d 1226 sb->s_fs_info = NULL;
b3b94faa
DT
1227 return error;
1228}
1229
f55073ff 1230static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1231{
f55073ff
SW
1232 s->s_bdev = data;
1233 s->s_dev = s->s_bdev->bd_dev;
1234
1235 /*
1236 * We set the bdi here to the queue backing, file systems can
1237 * overwrite this in ->fill_super()
1238 */
1239 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1240 return 0;
86384605
AD
1241}
1242
f55073ff 1243static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1244{
1245 struct block_device *bdev = ptr;
1246 return (bdev == s->s_bdev);
1247}
1248
f55073ff 1249/**
8bcbbf00 1250 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1251 * @fs_type: The GFS2 filesystem type
1252 * @flags: Mount flags
1253 * @dev_name: The name of the device
1254 * @data: The mount arguments
f55073ff
SW
1255 *
1256 * Q. Why not use get_sb_bdev() ?
1257 * A. We need to select one of two root directories to mount, independent
1258 * of whether this is the initial, or subsequent, mount of this sb
1259 *
1260 * Returns: 0 or -ve on error
1261 */
1262
8bcbbf00
AV
1263static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1264 const char *dev_name, void *data)
f55073ff
SW
1265{
1266 struct block_device *bdev;
1267 struct super_block *s;
d4d77629 1268 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1269 int error;
1270 struct gfs2_args args;
1271 struct gfs2_sbd *sdp;
1272
1273 if (!(flags & MS_RDONLY))
1274 mode |= FMODE_WRITE;
1275
d4d77629 1276 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1277 if (IS_ERR(bdev))
8bcbbf00 1278 return ERR_CAST(bdev);
f55073ff
SW
1279
1280 /*
1281 * once the super is inserted into the list by sget, s_umount
1282 * will protect the lockfs code from trying to start a snapshot
1283 * while we are mounting
1284 */
1285 mutex_lock(&bdev->bd_fsfreeze_mutex);
1286 if (bdev->bd_fsfreeze_count > 0) {
1287 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1288 error = -EBUSY;
1289 goto error_bdev;
1290 }
9249e17f 1291 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1292 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1293 error = PTR_ERR(s);
1294 if (IS_ERR(s))
1295 goto error_bdev;
1296
dfe5b9ad
SW
1297 if (s->s_root) {
1298 /*
1299 * s_umount nests inside bd_mutex during
1300 * __invalidate_device(). blkdev_put() acquires
1301 * bd_mutex and can't be called under s_umount. Drop
1302 * s_umount temporarily. This is safe as we're
1303 * holding an active reference.
1304 */
1305 up_write(&s->s_umount);
d4d77629 1306 blkdev_put(bdev, mode);
dfe5b9ad
SW
1307 down_write(&s->s_umount);
1308 }
8bcbbf00 1309
f55073ff
SW
1310 memset(&args, 0, sizeof(args));
1311 args.ar_quota = GFS2_QUOTA_DEFAULT;
1312 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1313 args.ar_commit = 30;
3d3c10f2
BM
1314 args.ar_statfs_quantum = 30;
1315 args.ar_quota_quantum = 60;
f55073ff
SW
1316 args.ar_errors = GFS2_ERRORS_DEFAULT;
1317
1318 error = gfs2_mount_args(&args, data);
1319 if (error) {
d77d1b58 1320 pr_warn("can't parse mount arguments\n");
8bcbbf00 1321 goto error_super;
f55073ff
SW
1322 }
1323
1324 if (s->s_root) {
1325 error = -EBUSY;
1326 if ((flags ^ s->s_flags) & MS_RDONLY)
1327 goto error_super;
f55073ff
SW
1328 } else {
1329 char b[BDEVNAME_SIZE];
1330
f55073ff
SW
1331 s->s_mode = mode;
1332 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1333 sb_set_blocksize(s, block_size(bdev));
1334 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1335 if (error)
1336 goto error_super;
f55073ff
SW
1337 s->s_flags |= MS_ACTIVE;
1338 bdev->bd_super = s;
1339 }
1340
1341 sdp = s->s_fs_info;
f55073ff 1342 if (args.ar_meta)
8bcbbf00 1343 return dget(sdp->sd_master_dir);
f55073ff 1344 else
8bcbbf00 1345 return dget(sdp->sd_root_dir);
f55073ff
SW
1346
1347error_super:
1348 deactivate_locked_super(s);
8bcbbf00 1349 return ERR_PTR(error);
f55073ff 1350error_bdev:
d4d77629 1351 blkdev_put(bdev, mode);
8bcbbf00 1352 return ERR_PTR(error);
f55073ff
SW
1353}
1354
f6d03139
SW
1355static int set_meta_super(struct super_block *s, void *ptr)
1356{
1357 return -EINVAL;
1358}
1359
8bcbbf00
AV
1360static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1361 int flags, const char *dev_name, void *data)
86384605 1362{
f6d03139 1363 struct super_block *s;
003dec89 1364 struct gfs2_sbd *sdp;
e24977d4 1365 struct path path;
86384605 1366 int error;
907b9bce 1367
e24977d4 1368 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1369 if (error) {
d77d1b58
JP
1370 pr_warn("path_lookup on %s returned error %d\n",
1371 dev_name, error);
8bcbbf00 1372 return ERR_PTR(error);
86384605 1373 }
9249e17f 1374 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
f6d03139 1375 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1376 path_put(&path);
f6d03139 1377 if (IS_ERR(s)) {
d77d1b58 1378 pr_warn("gfs2 mount does not exist\n");
8bcbbf00 1379 return ERR_CAST(s);
86384605 1380 }
f55073ff
SW
1381 if ((flags ^ s->s_flags) & MS_RDONLY) {
1382 deactivate_locked_super(s);
8bcbbf00 1383 return ERR_PTR(-EBUSY);
f55073ff 1384 }
f6d03139 1385 sdp = s->s_fs_info;
8bcbbf00 1386 return dget(sdp->sd_master_dir);
b3b94faa
DT
1387}
1388
419c93e0 1389static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1390{
1391 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1392
1393 if (sdp == NULL) {
1394 kill_block_super(sb);
1395 return;
3af165ac 1396 }
88a19ad0 1397
24972557 1398 gfs2_log_flush(sdp, NULL, SYNC_FLUSH);
88a19ad0
SW
1399 dput(sdp->sd_root_dir);
1400 dput(sdp->sd_master_dir);
1401 sdp->sd_root_dir = NULL;
1402 sdp->sd_master_dir = NULL;
fefc03bf 1403 shrink_dcache_sb(sb);
88a19ad0 1404 gfs2_delete_debugfs_file(sdp);
a245769f 1405 free_percpu(sdp->sd_lkstats);
0d515210 1406 kill_block_super(sb);
86384605
AD
1407}
1408
b3b94faa
DT
1409struct file_system_type gfs2_fs_type = {
1410 .name = "gfs2",
1411 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1412 .mount = gfs2_mount,
419c93e0
SW
1413 .kill_sb = gfs2_kill_sb,
1414 .owner = THIS_MODULE,
1415};
7f78e035 1416MODULE_ALIAS_FS("gfs2");
419c93e0
SW
1417
1418struct file_system_type gfs2meta_fs_type = {
1419 .name = "gfs2meta",
1420 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1421 .mount = gfs2_mount_meta,
b3b94faa
DT
1422 .owner = THIS_MODULE,
1423};
7f78e035 1424MODULE_ALIAS_FS("gfs2meta");