]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/gfs2/ops_fstype.c
block: switch bios to blk_status_t
[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>
95fe66de 26#include <linux/backing-dev.h>
b3b94faa
DT
27
28#include "gfs2.h"
5c676f6d 29#include "incore.h"
da6dd40d 30#include "bmap.h"
b3b94faa
DT
31#include "glock.h"
32#include "glops.h"
33#include "inode.h"
b3b94faa
DT
34#include "recovery.h"
35#include "rgrp.h"
36#include "super.h"
b3b94faa 37#include "sys.h"
5c676f6d 38#include "util.h"
bb3b0e3d 39#include "log.h"
9ac1b4d9 40#include "quota.h"
b5289681 41#include "dir.h"
70d4ee94 42#include "meta_io.h"
63997775 43#include "trace_gfs2.h"
b3b94faa
DT
44
45#define DO 0
46#define UNDO 1
47
9b8df98f
SW
48/**
49 * gfs2_tune_init - Fill a gfs2_tune structure with default values
50 * @gt: tune
51 *
52 */
53
54static void gfs2_tune_init(struct gfs2_tune *gt)
55{
56 spin_lock_init(&gt->gt_spin);
57
9b8df98f
SW
58 gt->gt_quota_warn_period = 10;
59 gt->gt_quota_scale_num = 1;
60 gt->gt_quota_scale_den = 1;
9b8df98f 61 gt->gt_new_files_jdata = 0;
47a9a527 62 gt->gt_max_readahead = BIT(18);
9b8df98f 63 gt->gt_complain_secs = 10;
9b8df98f
SW
64}
65
b3b94faa
DT
66static struct gfs2_sbd *init_sbd(struct super_block *sb)
67{
68 struct gfs2_sbd *sdp;
70d4ee94 69 struct address_space *mapping;
b3b94faa 70
85d1da67 71 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
72 if (!sdp)
73 return NULL;
74
5c676f6d 75 sb->s_fs_info = sdp;
b3b94faa 76 sdp->sd_vfs = sb;
a245769f
SW
77 sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
78 if (!sdp->sd_lkstats) {
79 kfree(sdp);
80 return NULL;
81 }
82
ba6e9364 83 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
84 gfs2_tune_init(&sdp->sd_tune);
85
e402746a
SW
86 init_waitqueue_head(&sdp->sd_glock_wait);
87 atomic_set(&sdp->sd_glock_disposal, 0);
3942ae53 88 init_completion(&sdp->sd_locking_init);
fd95e81c 89 init_completion(&sdp->sd_wdack);
b3b94faa 90 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
91
92 spin_lock_init(&sdp->sd_rindex_spin);
7c9ca621 93 sdp->sd_rindex_tree.rb_node = NULL;
b3b94faa
DT
94
95 INIT_LIST_HEAD(&sdp->sd_jindex_list);
96 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 97 mutex_init(&sdp->sd_jindex_mutex);
0e48e055 98 init_completion(&sdp->sd_journal_ready);
b3b94faa 99
b3b94faa 100 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 101 mutex_init(&sdp->sd_quota_mutex);
e46c772d 102 mutex_init(&sdp->sd_quota_sync_mutex);
37b2c837 103 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
104 INIT_LIST_HEAD(&sdp->sd_trunc_list);
105 spin_lock_init(&sdp->sd_trunc_lock);
2d9e7230 106 spin_lock_init(&sdp->sd_bitmap_lock);
b3b94faa 107
70d4ee94
SW
108 mapping = &sdp->sd_aspace;
109
39849d69 110 address_space_init_once(mapping);
1b2ad412 111 mapping->a_ops = &gfs2_rgrp_aops;
70d4ee94
SW
112 mapping->host = sb->s_bdev->bd_inode;
113 mapping->flags = 0;
114 mapping_set_gfp_mask(mapping, GFP_NOFS);
115 mapping->private_data = NULL;
70d4ee94
SW
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 131 atomic_set(&sdp->sd_log_in_flight, 0);
2e60d768
BM
132 atomic_set(&sdp->sd_reserving_log, 0);
133 init_waitqueue_head(&sdp->sd_reserving_log_wait);
16615be1 134 init_waitqueue_head(&sdp->sd_log_flush_wait);
2e60d768
BM
135 atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN);
136 mutex_init(&sdp->sd_freeze_mutex);
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
4246a0b6 175static void end_bio_io_page(struct bio *bio)
9b8df98f
SW
176{
177 struct page *page = bio->bi_private;
178
4e4cbee9 179 if (!bio->bi_status)
9b8df98f
SW
180 SetPageUptodate(page);
181 else
4e4cbee9 182 pr_warn("error %d reading superblock\n", bio->bi_status);
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;
70fd7614 250 bio_set_op_attrs(bio, REQ_OP_READ, REQ_META);
4e49ea4a 251 submit_bio(bio);
9b8df98f
SW
252 wait_on_page_locked(page);
253 bio_put(bio);
254 if (!PageUptodate(page)) {
255 __free_page(page);
256 return -EIO;
257 }
258 p = kmap(page);
32e471ef 259 gfs2_sb_in(sdp, p);
9b8df98f
SW
260 kunmap(page);
261 __free_page(page);
32e471ef 262 return gfs2_check_sb(sdp, silent);
9b8df98f 263}
6bac243f 264
9b8df98f
SW
265/**
266 * gfs2_read_sb - Read super block
267 * @sdp: The GFS2 superblock
9b8df98f
SW
268 * @silent: Don't print message if mount fails
269 *
270 */
271
6bac243f 272static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
273{
274 u32 hash_blocks, ind_blocks, leaf_blocks;
275 u32 tmp_blocks;
276 unsigned int x;
277 int error;
278
32e471ef 279 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
280 if (error) {
281 if (!silent)
282 fs_err(sdp, "can't read superblock\n");
283 return error;
284 }
285
9b8df98f
SW
286 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
287 GFS2_BASIC_BLOCK_SHIFT;
47a9a527 288 sdp->sd_fsb2bb = BIT(sdp->sd_fsb2bb_shift);
9b8df98f
SW
289 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
290 sizeof(struct gfs2_dinode)) / sizeof(u64);
291 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
292 sizeof(struct gfs2_meta_header)) / sizeof(u64);
293 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
294 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
295 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
296 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
297 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
298 sizeof(struct gfs2_meta_header)) /
299 sizeof(struct gfs2_quota_change);
a68a0a35
BP
300 sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
301 sizeof(struct gfs2_meta_header))
302 * GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
9b8df98f
SW
303
304 /* Compute maximum reservation required to add a entry to a directory */
305
47a9a527 306 hash_blocks = DIV_ROUND_UP(sizeof(u64) * BIT(GFS2_DIR_MAX_DEPTH),
9b8df98f
SW
307 sdp->sd_jbsize);
308
309 ind_blocks = 0;
310 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
311 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
312 ind_blocks += tmp_blocks;
313 }
314
315 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
316
317 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
318
319 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
320 sizeof(struct gfs2_dinode);
321 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
322 for (x = 2;; x++) {
323 u64 space, d;
324 u32 m;
325
326 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
327 d = space;
328 m = do_div(d, sdp->sd_inptrs);
329
330 if (d != sdp->sd_heightsize[x - 1] || m)
331 break;
332 sdp->sd_heightsize[x] = space;
333 }
334 sdp->sd_max_height = x;
335 sdp->sd_heightsize[x] = ~0;
336 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
337
338 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
339 sizeof(struct gfs2_dinode);
340 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
341 for (x = 2;; x++) {
342 u64 space, d;
343 u32 m;
344
345 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
346 d = space;
347 m = do_div(d, sdp->sd_inptrs);
348
349 if (d != sdp->sd_jheightsize[x - 1] || m)
350 break;
351 sdp->sd_jheightsize[x] = space;
352 }
353 sdp->sd_max_jheight = x;
354 sdp->sd_jheightsize[x] = ~0;
355 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
356
471f3db2
BM
357 sdp->sd_max_dents_per_leaf = (sdp->sd_sb.sb_bsize -
358 sizeof(struct gfs2_leaf)) /
359 GFS2_MIN_DIRENT_SIZE;
9b8df98f
SW
360 return 0;
361}
362
b3b94faa
DT
363static int init_names(struct gfs2_sbd *sdp, int silent)
364{
b3b94faa
DT
365 char *proto, *table;
366 int error = 0;
367
368 proto = sdp->sd_args.ar_lockproto;
369 table = sdp->sd_args.ar_locktable;
370
371 /* Try to autodetect */
372
373 if (!proto[0] || !table[0]) {
32e471ef 374 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
375 if (error)
376 return error;
3cf1e7be 377
b3b94faa 378 if (!proto[0])
3cf1e7be 379 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 380 if (!table[0])
3cf1e7be 381 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
382 }
383
384 if (!table[0])
385 table = sdp->sd_vfs->s_id;
386
00377d8e
JD
387 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
388 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 389
5d35e31f
DC
390 table = sdp->sd_table_name;
391 while ((table = strchr(table, '/')))
b35997d4
RP
392 *table = '_';
393
b3b94faa
DT
394 return error;
395}
396
397static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
398 int undo)
399{
b3b94faa
DT
400 int error = 0;
401
402 if (undo)
403 goto fail_trans;
404
b3b94faa
DT
405 error = gfs2_glock_nq_num(sdp,
406 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
407 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
408 mount_gh);
409 if (error) {
410 fs_err(sdp, "can't acquire mount glock: %d\n", error);
411 goto fail;
412 }
413
414 error = gfs2_glock_nq_num(sdp,
415 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
416 LM_ST_SHARED,
579b78a4 417 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
418 &sdp->sd_live_gh);
419 if (error) {
420 fs_err(sdp, "can't acquire live glock: %d\n", error);
421 goto fail_mount;
422 }
423
424 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
425 CREATE, &sdp->sd_rename_gl);
426 if (error) {
427 fs_err(sdp, "can't create rename glock: %d\n", error);
428 goto fail_live;
429 }
430
24972557
BM
431 error = gfs2_glock_get(sdp, GFS2_FREEZE_LOCK, &gfs2_freeze_glops,
432 CREATE, &sdp->sd_freeze_gl);
b3b94faa
DT
433 if (error) {
434 fs_err(sdp, "can't create transaction glock: %d\n", error);
435 goto fail_rename;
436 }
b3b94faa
DT
437
438 return 0;
439
feaa7bba 440fail_trans:
24972557 441 gfs2_glock_put(sdp->sd_freeze_gl);
feaa7bba 442fail_rename:
b3b94faa 443 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 444fail_live:
b3b94faa 445 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 446fail_mount:
b3b94faa 447 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 448fail:
b3b94faa
DT
449 return error;
450}
451
9b8df98f
SW
452static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
453 u64 no_addr, const char *name)
f42faf4f 454{
9b8df98f
SW
455 struct gfs2_sbd *sdp = sb->s_fs_info;
456 struct dentry *dentry;
457 struct inode *inode;
458
3ce37b2c
AG
459 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0,
460 GFS2_BLKST_FREE /* ignore */);
9b8df98f
SW
461 if (IS_ERR(inode)) {
462 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
463 return PTR_ERR(inode);
464 }
48fde701 465 dentry = d_make_root(inode);
9b8df98f
SW
466 if (!dentry) {
467 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
468 return -ENOMEM;
469 }
9b8df98f
SW
470 *dptr = dentry;
471 return 0;
f42faf4f
SW
472}
473
9b8df98f 474static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
475{
476 struct super_block *sb = sdp->sd_vfs;
477 struct gfs2_holder sb_gh;
dbb7cae2 478 u64 no_addr;
9b8df98f 479 int ret;
b3b94faa 480
9b8df98f
SW
481 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
482 LM_ST_SHARED, 0, &sb_gh);
483 if (ret) {
484 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
485 return ret;
b3b94faa 486 }
907b9bce 487
6bac243f 488 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
489 if (ret) {
490 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
491 goto out;
492 }
493
494 /* Set up the buffer cache and SB for real */
e1defc4f 495 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 496 ret = -EINVAL;
b3b94faa
DT
497 fs_err(sdp, "FS block size (%u) is too small for device "
498 "block size (%u)\n",
e1defc4f 499 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
500 goto out;
501 }
502 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 503 ret = -EINVAL;
b3b94faa
DT
504 fs_err(sdp, "FS block size (%u) is too big for machine "
505 "page size (%u)\n",
506 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
507 goto out;
508 }
b3b94faa
DT
509 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
510
f42faf4f 511 /* Get the root inode */
dbb7cae2 512 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
513 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
514 if (ret)
f42faf4f 515 goto out;
b3b94faa 516
9b8df98f
SW
517 /* Get the master inode */
518 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
519 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
520 if (ret) {
521 dput(sdp->sd_root_dir);
522 goto out;
523 }
524 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 525out:
b3b94faa 526 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 527 return ret;
b3b94faa
DT
528}
529
f057f6cd 530static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 531{
f057f6cd
SW
532 char *message = "FIRSTMOUNT=Done";
533 char *envp[] = { message, NULL };
e0c2a9aa
DT
534
535 fs_info(sdp, "first mount done, others may mount\n");
536
537 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
538 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
539
f057f6cd 540 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
541}
542
b5289681
SW
543/**
544 * gfs2_jindex_hold - Grab a lock on the jindex
545 * @sdp: The GFS2 superblock
546 * @ji_gh: the holder for the jindex glock
547 *
548 * Returns: errno
549 */
550
551static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
552{
553 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
554 struct qstr name;
555 char buf[20];
556 struct gfs2_jdesc *jd;
557 int error;
558
559 name.name = buf;
560
561 mutex_lock(&sdp->sd_jindex_mutex);
562
563 for (;;) {
564 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
565 if (error)
566 break;
567
568 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
569 name.hash = gfs2_disk_hash(name.name, name.len);
570
571 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
572 if (error == -ENOENT) {
573 error = 0;
574 break;
575 }
576
577 gfs2_glock_dq_uninit(ji_gh);
578
579 if (error)
580 break;
581
582 error = -ENOMEM;
583 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
584 if (!jd)
585 break;
586
587 INIT_LIST_HEAD(&jd->extent_list);
a17d758b
BP
588 INIT_LIST_HEAD(&jd->jd_revoke_list);
589
6ecd7c2d 590 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
591 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
592 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
593 if (!jd->jd_inode)
594 error = -ENOENT;
595 else
596 error = PTR_ERR(jd->jd_inode);
597 kfree(jd);
598 break;
599 }
600
601 spin_lock(&sdp->sd_jindex_spin);
602 jd->jd_jid = sdp->sd_journals++;
603 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
604 spin_unlock(&sdp->sd_jindex_spin);
605 }
606
607 mutex_unlock(&sdp->sd_jindex_mutex);
608
609 return error;
610}
611
1d12d175
BP
612/**
613 * check_journal_clean - Make sure a journal is clean for a spectator mount
614 * @sdp: The GFS2 superblock
615 * @jd: The journal descriptor
616 *
617 * Returns: 0 if the journal is clean or locked, else an error
618 */
619static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
620{
621 int error;
622 struct gfs2_holder j_gh;
623 struct gfs2_log_header_host head;
624 struct gfs2_inode *ip;
625
626 ip = GFS2_I(jd->jd_inode);
627 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
628 GL_EXACT | GL_NOCACHE, &j_gh);
629 if (error) {
630 fs_err(sdp, "Error locking journal for spectator mount.\n");
631 return -EPERM;
632 }
633 error = gfs2_jdesc_check(jd);
634 if (error) {
635 fs_err(sdp, "Error checking journal for spectator mount.\n");
636 goto out_unlock;
637 }
638 error = gfs2_find_jhead(jd, &head);
639 if (error) {
640 fs_err(sdp, "Error parsing journal for spectator mount.\n");
641 goto out_unlock;
642 }
643 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
644 error = -EPERM;
645 fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
646 "must not be a spectator.\n", jd->jd_jid);
647 }
648
649out_unlock:
650 gfs2_glock_dq_uninit(&j_gh);
651 return error;
652}
653
b3b94faa
DT
654static int init_journal(struct gfs2_sbd *sdp, int undo)
655{
2b0143b5 656 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa 657 struct gfs2_holder ji_gh;
5c676f6d 658 struct gfs2_inode *ip;
b3b94faa
DT
659 int jindex = 1;
660 int error = 0;
661
662 if (undo) {
663 jindex = 0;
fe64d517 664 goto fail_jinode_gh;
b3b94faa
DT
665 }
666
9b8df98f 667 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 668 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 669 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 670 return PTR_ERR(sdp->sd_jindex);
b3b94faa 671 }
b3b94faa
DT
672
673 /* Load in the journal index special file */
674
675 error = gfs2_jindex_hold(sdp, &ji_gh);
676 if (error) {
677 fs_err(sdp, "can't read journal index: %d\n", error);
678 goto fail;
679 }
680
0e5a9fb0 681 error = -EUSERS;
b3b94faa
DT
682 if (!gfs2_jindex_size(sdp)) {
683 fs_err(sdp, "no journals!\n");
907b9bce 684 goto fail_jindex;
b3b94faa
DT
685 }
686
f07b3520 687 atomic_set(&sdp->sd_log_blks_needed, 0);
b3b94faa
DT
688 if (sdp->sd_args.ar_spectator) {
689 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 690 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
691 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
692 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
693 } else {
694 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
695 fs_err(sdp, "can't mount journal #%u\n",
696 sdp->sd_lockstruct.ls_jid);
697 fs_err(sdp, "there are only %u journals (0 - %u)\n",
698 gfs2_jindex_size(sdp),
699 gfs2_jindex_size(sdp) - 1);
700 goto fail_jindex;
701 }
702 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
703
feaa7bba 704 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
705 &gfs2_journal_glops,
706 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
707 &sdp->sd_journal_gh);
708 if (error) {
709 fs_err(sdp, "can't acquire journal glock: %d\n", error);
710 goto fail_jindex;
711 }
712
feaa7bba
SW
713 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
714 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 715 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
716 &sdp->sd_jinode_gh);
717 if (error) {
718 fs_err(sdp, "can't acquire journal inode glock: %d\n",
719 error);
720 goto fail_journal_gh;
721 }
722
723 error = gfs2_jdesc_check(sdp->sd_jdesc);
724 if (error) {
725 fs_err(sdp, "my journal (%u) is bad: %d\n",
726 sdp->sd_jdesc->jd_jid, error);
727 goto fail_jinode_gh;
728 }
fd041f0b 729 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
730 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
731 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
732
733 /* Map the extents for this journal's blocks */
b50f227b 734 gfs2_map_journal_extents(sdp, sdp->sd_jdesc);
b3b94faa 735 }
63997775 736 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
737
738 if (sdp->sd_lockstruct.ls_first) {
739 unsigned int x;
740 for (x = 0; x < sdp->sd_journals; x++) {
1d12d175
BP
741 struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
742
743 if (sdp->sd_args.ar_spectator) {
744 error = check_journal_clean(sdp, jd);
745 if (error)
746 goto fail_jinode_gh;
747 continue;
748 }
749 error = gfs2_recover_journal(jd, true);
b3b94faa
DT
750 if (error) {
751 fs_err(sdp, "error recovering journal %u: %d\n",
752 x, error);
753 goto fail_jinode_gh;
754 }
755 }
756
f057f6cd 757 gfs2_others_may_mount(sdp);
b3b94faa 758 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 759 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
760 if (error) {
761 fs_err(sdp, "error recovering my journal: %d\n", error);
762 goto fail_jinode_gh;
763 }
764 }
765
086cc672 766 sdp->sd_log_idle = 1;
b3b94faa
DT
767 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
768 gfs2_glock_dq_uninit(&ji_gh);
769 jindex = 0;
2e60d768 770 INIT_WORK(&sdp->sd_freeze_work, gfs2_freeze_func);
b3b94faa
DT
771 return 0;
772
a91ea69f 773fail_jinode_gh:
b3b94faa
DT
774 if (!sdp->sd_args.ar_spectator)
775 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 776fail_journal_gh:
b3b94faa
DT
777 if (!sdp->sd_args.ar_spectator)
778 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 779fail_jindex:
b3b94faa
DT
780 gfs2_jindex_free(sdp);
781 if (jindex)
782 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 783fail:
f42faf4f 784 iput(sdp->sd_jindex);
b3b94faa
DT
785 return error;
786}
787
56aa72d0 788static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
789
790static int init_inodes(struct gfs2_sbd *sdp, int undo)
791{
b3b94faa 792 int error = 0;
2b0143b5 793 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa
DT
794
795 if (undo)
f42faf4f
SW
796 goto fail_qinode;
797
f42faf4f 798 error = init_journal(sdp, undo);
0e48e055 799 complete_all(&sdp->sd_journal_ready);
f42faf4f 800 if (error)
9b8df98f 801 goto fail;
b3b94faa 802
b3b94faa 803 /* Read in the master statfs inode */
9b8df98f 804 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
805 if (IS_ERR(sdp->sd_statfs_inode)) {
806 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 807 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 808 goto fail_journal;
b3b94faa
DT
809 }
810
811 /* Read in the resource index inode */
9b8df98f 812 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
813 if (IS_ERR(sdp->sd_rindex)) {
814 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
815 fs_err(sdp, "can't get resource index inode: %d\n", error);
816 goto fail_statfs;
817 }
cf45b752 818 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
819
820 /* Read in the quota inode */
9b8df98f 821 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
822 if (IS_ERR(sdp->sd_quota_inode)) {
823 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
824 fs_err(sdp, "can't get quota file inode: %d\n", error);
825 goto fail_rindex;
826 }
56aa72d0
JK
827 /*
828 * i_mutex on quota files is special. Since this inode is hidden system
829 * file, we are safe to define locking ourselves.
830 */
9902af79 831 lockdep_set_class(&sdp->sd_quota_inode->i_rwsem,
56aa72d0 832 &gfs2_quota_imutex_key);
a365fbf3
SW
833
834 error = gfs2_rindex_update(sdp);
835 if (error)
836 goto fail_qinode;
837
b3b94faa
DT
838 return 0;
839
f42faf4f
SW
840fail_qinode:
841 iput(sdp->sd_quota_inode);
f42faf4f 842fail_rindex:
b3b94faa 843 gfs2_clear_rgrpd(sdp);
f42faf4f 844 iput(sdp->sd_rindex);
f42faf4f
SW
845fail_statfs:
846 iput(sdp->sd_statfs_inode);
f42faf4f
SW
847fail_journal:
848 init_journal(sdp, UNDO);
f42faf4f 849fail:
b3b94faa
DT
850 return error;
851}
852
853static int init_per_node(struct gfs2_sbd *sdp, int undo)
854{
f42faf4f 855 struct inode *pn = NULL;
b3b94faa
DT
856 char buf[30];
857 int error = 0;
5c676f6d 858 struct gfs2_inode *ip;
2b0143b5 859 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa
DT
860
861 if (sdp->sd_args.ar_spectator)
862 return 0;
863
864 if (undo)
865 goto fail_qc_gh;
866
9b8df98f 867 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
868 if (IS_ERR(pn)) {
869 error = PTR_ERR(pn);
b3b94faa
DT
870 fs_err(sdp, "can't find per_node directory: %d\n", error);
871 return error;
872 }
873
b3b94faa 874 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
875 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
876 if (IS_ERR(sdp->sd_sc_inode)) {
877 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 878 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 879 goto fail;
b3b94faa
DT
880 }
881
b3b94faa 882 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
883 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
884 if (IS_ERR(sdp->sd_qc_inode)) {
885 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
886 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
887 goto fail_ut_i;
888 }
889
f42faf4f 890 iput(pn);
b3b94faa
DT
891 pn = NULL;
892
feaa7bba 893 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 894 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
895 &sdp->sd_sc_gh);
896 if (error) {
897 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 898 goto fail_qc_i;
b3b94faa
DT
899 }
900
feaa7bba 901 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 902 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
903 &sdp->sd_qc_gh);
904 if (error) {
905 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
906 goto fail_ut_gh;
907 }
908
909 return 0;
910
a91ea69f 911fail_qc_gh:
b3b94faa 912 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 913fail_ut_gh:
b3b94faa 914 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 915fail_qc_i:
f42faf4f 916 iput(sdp->sd_qc_inode);
a91ea69f 917fail_ut_i:
f42faf4f 918 iput(sdp->sd_sc_inode);
a91ea69f 919fail:
6fde2242 920 iput(pn);
b3b94faa
DT
921 return error;
922}
923
f057f6cd
SW
924static const match_table_t nolock_tokens = {
925 { Opt_jid, "jid=%d\n", },
926 { Opt_err, NULL },
927};
928
929static const struct lm_lockops nolock_ops = {
930 .lm_proto_name = "lock_nolock",
fc0e38da 931 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
932 .lm_tokens = &nolock_tokens,
933};
934
da755fdb
SW
935/**
936 * gfs2_lm_mount - mount a locking protocol
937 * @sdp: the filesystem
3ad2f3fb 938 * @args: mount arguments
da755fdb
SW
939 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
940 *
941 * Returns: errno
942 */
943
944static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
945{
f057f6cd
SW
946 const struct lm_lockops *lm;
947 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
948 struct gfs2_args *args = &sdp->sd_args;
949 const char *proto = sdp->sd_proto_name;
950 const char *table = sdp->sd_table_name;
f057f6cd
SW
951 char *o, *options;
952 int ret;
da755fdb 953
f057f6cd
SW
954 if (!strcmp("lock_nolock", proto)) {
955 lm = &nolock_ops;
956 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
957#ifdef CONFIG_GFS2_FS_LOCKING_DLM
958 } else if (!strcmp("lock_dlm", proto)) {
959 lm = &gfs2_dlm_ops;
960#endif
961 } else {
d77d1b58 962 pr_info("can't find protocol %s\n", proto);
f057f6cd
SW
963 return -ENOENT;
964 }
da755fdb
SW
965
966 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
967
f057f6cd
SW
968 ls->ls_ops = lm;
969 ls->ls_first = 1;
da755fdb 970
f057f6cd
SW
971 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
972 substring_t tmp[MAX_OPT_ARGS];
973 int token, option;
974
975 if (!o || !*o)
976 continue;
977
978 token = match_token(o, *lm->lm_tokens, tmp);
979 switch (token) {
980 case Opt_jid:
981 ret = match_int(&tmp[0], &option);
982 if (ret || option < 0)
983 goto hostdata_error;
ba6e9364
SW
984 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
985 ls->ls_jid = option;
f057f6cd
SW
986 break;
987 case Opt_id:
4875647a 988 case Opt_nodir:
2b88f7c5 989 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
990 break;
991 case Opt_first:
992 ret = match_int(&tmp[0], &option);
993 if (ret || (option != 0 && option != 1))
994 goto hostdata_error;
995 ls->ls_first = option;
996 break;
f057f6cd
SW
997 case Opt_err:
998 default:
999hostdata_error:
1000 fs_info(sdp, "unknown hostdata (%s)\n", o);
1001 return -EINVAL;
1002 }
da755fdb
SW
1003 }
1004
f057f6cd
SW
1005 if (lm->lm_mount == NULL) {
1006 fs_info(sdp, "Now mounting FS...\n");
19237039 1007 complete_all(&sdp->sd_locking_init);
f057f6cd 1008 return 0;
da755fdb 1009 }
e0c2a9aa 1010 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1011 if (ret == 0)
1012 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1013 complete_all(&sdp->sd_locking_init);
f057f6cd 1014 return ret;
da755fdb
SW
1015}
1016
1017void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1018{
f057f6cd
SW
1019 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1020 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1021 lm->lm_unmount)
1022 lm->lm_unmount(sdp);
da755fdb 1023}
b3b94faa 1024
ba6e9364
SW
1025static int wait_on_journal(struct gfs2_sbd *sdp)
1026{
ba6e9364
SW
1027 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1028 return 0;
1029
74316201
N
1030 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, TASK_INTERRUPTIBLE)
1031 ? -EINTR : 0;
ba6e9364
SW
1032}
1033
8633ecfa
SW
1034void gfs2_online_uevent(struct gfs2_sbd *sdp)
1035{
1036 struct super_block *sb = sdp->sd_vfs;
1037 char ro[20];
1038 char spectator[20];
1039 char *envp[] = { ro, spectator, NULL };
1040 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1041 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1042 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1043}
1044
b3b94faa
DT
1045/**
1046 * fill_super - Read in superblock
1047 * @sb: The VFS superblock
1048 * @data: Mount options
1049 * @silent: Don't complain if it's not a GFS2 filesystem
1050 *
1051 * Returns: errno
1052 */
1053
f55073ff 1054static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1055{
1056 struct gfs2_sbd *sdp;
1057 struct gfs2_holder mount_gh;
1058 int error;
1059
1060 sdp = init_sbd(sb);
1061 if (!sdp) {
d77d1b58 1062 pr_warn("can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1063 return -ENOMEM;
1064 }
f55073ff 1065 sdp->sd_args = *args;
b3b94faa 1066
fe64d517 1067 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1068 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1069 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1070 }
6f04c1c7
SW
1071 if (sdp->sd_args.ar_posix_acl)
1072 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1073 if (sdp->sd_args.ar_nobarrier)
1074 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1075
9964afbb 1076 sb->s_flags |= MS_NOSEC;
719ee344
SW
1077 sb->s_magic = GFS2_MAGIC;
1078 sb->s_op = &gfs2_super_ops;
41ced6dc 1079 sb->s_d_op = &gfs2_dops;
719ee344 1080 sb->s_export_op = &gfs2_export_ops;
40b78a32 1081 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f 1082 sb->s_qcop = &gfs2_quotactl_ops;
de3b08d3 1083 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cc632e7f 1084 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1085 sb->s_time_gran = 1;
1086 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1087
1088 /* Set up the buffer cache and fill in some fake block size values
1089 to allow us to read-in the on-disk superblock. */
1090 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1091 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1092 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1093 GFS2_BASIC_BLOCK_SHIFT;
47a9a527 1094 sdp->sd_fsb2bb = BIT(sdp->sd_fsb2bb_shift);
b3b94faa 1095
5e687eac 1096 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1097 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1098 if (sdp->sd_args.ar_statfs_quantum) {
1099 sdp->sd_tune.gt_statfs_slow = 0;
1100 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1101 } else {
3d3c10f2
BM
1102 sdp->sd_tune.gt_statfs_slow = 1;
1103 sdp->sd_tune.gt_statfs_quantum = 30;
1104 }
48c2b613 1105
b3b94faa 1106 error = init_names(sdp, silent);
0d515210
BP
1107 if (error) {
1108 /* In this case, we haven't initialized sysfs, so we have to
1109 manually free the sdp. */
1110 free_percpu(sdp->sd_lkstats);
1111 kfree(sdp);
1112 sb->s_fs_info = NULL;
1113 return error;
1114 }
b3b94faa 1115
e0c2a9aa
DT
1116 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1117
b3b94faa 1118 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1119 /*
1120 * If we hit an error here, gfs2_sys_fs_add will have called function
1121 * kobject_put which causes the sysfs usage count to go to zero, which
1122 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1123 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1124 * kobject_put to free sdp.
1125 */
b3b94faa 1126 if (error)
0d515210
BP
1127 return error;
1128
1129 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1130
1131 error = gfs2_lm_mount(sdp, silent);
1132 if (error)
0d515210 1133 goto fail_debug;
b3b94faa
DT
1134
1135 error = init_locking(sdp, &mount_gh, DO);
1136 if (error)
1137 goto fail_lm;
1138
9b8df98f 1139 error = init_sb(sdp, silent);
b3b94faa
DT
1140 if (error)
1141 goto fail_locking;
b3b94faa 1142
ba6e9364
SW
1143 error = wait_on_journal(sdp);
1144 if (error)
1145 goto fail_sb;
1146
feb47ca9
SW
1147 /*
1148 * If user space has failed to join the cluster or some similar
1149 * failure has occurred, then the journal id will contain a
1150 * negative (error) number. This will then be returned to the
1151 * caller (of the mount syscall). We do this even for spectator
1152 * mounts (which just write a jid of 0 to indicate "ok" even though
1153 * the jid is unused in the spectator case)
1154 */
1155 if (sdp->sd_lockstruct.ls_jid < 0) {
1156 error = sdp->sd_lockstruct.ls_jid;
1157 sdp->sd_lockstruct.ls_jid = 0;
1158 goto fail_sb;
1159 }
1160
e0c2a9aa
DT
1161 if (sdp->sd_args.ar_spectator)
1162 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1163 sdp->sd_table_name);
1164 else
1165 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1166 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1167
b3b94faa
DT
1168 error = init_inodes(sdp, DO);
1169 if (error)
f42faf4f 1170 goto fail_sb;
b3b94faa
DT
1171
1172 error = init_per_node(sdp, DO);
1173 if (error)
1174 goto fail_inodes;
1175
1176 error = gfs2_statfs_init(sdp);
1177 if (error) {
1178 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1179 goto fail_per_node;
1180 }
1181
b3b94faa
DT
1182 if (!(sb->s_flags & MS_RDONLY)) {
1183 error = gfs2_make_fs_rw(sdp);
1184 if (error) {
1185 fs_err(sdp, "can't make FS RW: %d\n", error);
8ad151c2 1186 goto fail_per_node;
b3b94faa
DT
1187 }
1188 }
1189
1190 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1191 gfs2_online_uevent(sdp);
b3b94faa
DT
1192 return 0;
1193
a91ea69f 1194fail_per_node:
b3b94faa 1195 init_per_node(sdp, UNDO);
a91ea69f 1196fail_inodes:
b3b94faa 1197 init_inodes(sdp, UNDO);
a91ea69f 1198fail_sb:
9b8df98f
SW
1199 if (sdp->sd_root_dir)
1200 dput(sdp->sd_root_dir);
1201 if (sdp->sd_master_dir)
1202 dput(sdp->sd_master_dir);
e7c8707e
SW
1203 if (sb->s_root)
1204 dput(sb->s_root);
9b8df98f 1205 sb->s_root = NULL;
a91ea69f 1206fail_locking:
b3b94faa 1207 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1208fail_lm:
0e48e055 1209 complete_all(&sdp->sd_journal_ready);
fefc03bf 1210 gfs2_gl_hash_clear(sdp);
b3b94faa 1211 gfs2_lm_unmount(sdp);
0d515210 1212fail_debug:
5f882096 1213 gfs2_delete_debugfs_file(sdp);
a245769f 1214 free_percpu(sdp->sd_lkstats);
0d515210
BP
1215 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1216 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1217 gfs2_sys_fs_del(sdp);
5c676f6d 1218 sb->s_fs_info = NULL;
b3b94faa
DT
1219 return error;
1220}
1221
f55073ff 1222static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1223{
f55073ff
SW
1224 s->s_bdev = data;
1225 s->s_dev = s->s_bdev->bd_dev;
95fe66de 1226 s->s_bdi = bdi_get(s->s_bdev->bd_bdi);
f55073ff 1227 return 0;
86384605
AD
1228}
1229
f55073ff 1230static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1231{
1232 struct block_device *bdev = ptr;
1233 return (bdev == s->s_bdev);
1234}
1235
f55073ff 1236/**
8bcbbf00 1237 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1238 * @fs_type: The GFS2 filesystem type
1239 * @flags: Mount flags
1240 * @dev_name: The name of the device
1241 * @data: The mount arguments
f55073ff
SW
1242 *
1243 * Q. Why not use get_sb_bdev() ?
1244 * A. We need to select one of two root directories to mount, independent
1245 * of whether this is the initial, or subsequent, mount of this sb
1246 *
1247 * Returns: 0 or -ve on error
1248 */
1249
8bcbbf00
AV
1250static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1251 const char *dev_name, void *data)
f55073ff
SW
1252{
1253 struct block_device *bdev;
1254 struct super_block *s;
d4d77629 1255 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1256 int error;
1257 struct gfs2_args args;
1258 struct gfs2_sbd *sdp;
1259
1260 if (!(flags & MS_RDONLY))
1261 mode |= FMODE_WRITE;
1262
d4d77629 1263 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1264 if (IS_ERR(bdev))
8bcbbf00 1265 return ERR_CAST(bdev);
f55073ff
SW
1266
1267 /*
1268 * once the super is inserted into the list by sget, s_umount
1269 * will protect the lockfs code from trying to start a snapshot
1270 * while we are mounting
1271 */
1272 mutex_lock(&bdev->bd_fsfreeze_mutex);
1273 if (bdev->bd_fsfreeze_count > 0) {
1274 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1275 error = -EBUSY;
1276 goto error_bdev;
1277 }
9249e17f 1278 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1279 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1280 error = PTR_ERR(s);
1281 if (IS_ERR(s))
1282 goto error_bdev;
1283
dfe5b9ad
SW
1284 if (s->s_root) {
1285 /*
1286 * s_umount nests inside bd_mutex during
1287 * __invalidate_device(). blkdev_put() acquires
1288 * bd_mutex and can't be called under s_umount. Drop
1289 * s_umount temporarily. This is safe as we're
1290 * holding an active reference.
1291 */
1292 up_write(&s->s_umount);
d4d77629 1293 blkdev_put(bdev, mode);
dfe5b9ad 1294 down_write(&s->s_umount);
6de20eb0
AP
1295 } else {
1296 /* s_mode must be set before deactivate_locked_super calls */
1297 s->s_mode = mode;
dfe5b9ad 1298 }
8bcbbf00 1299
f55073ff
SW
1300 memset(&args, 0, sizeof(args));
1301 args.ar_quota = GFS2_QUOTA_DEFAULT;
1302 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1303 args.ar_commit = 30;
3d3c10f2
BM
1304 args.ar_statfs_quantum = 30;
1305 args.ar_quota_quantum = 60;
f55073ff
SW
1306 args.ar_errors = GFS2_ERRORS_DEFAULT;
1307
1308 error = gfs2_mount_args(&args, data);
1309 if (error) {
d77d1b58 1310 pr_warn("can't parse mount arguments\n");
8bcbbf00 1311 goto error_super;
f55073ff
SW
1312 }
1313
1314 if (s->s_root) {
1315 error = -EBUSY;
1316 if ((flags ^ s->s_flags) & MS_RDONLY)
1317 goto error_super;
f55073ff 1318 } else {
a1c6f057 1319 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
f55073ff
SW
1320 sb_set_blocksize(s, block_size(bdev));
1321 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1322 if (error)
1323 goto error_super;
f55073ff
SW
1324 s->s_flags |= MS_ACTIVE;
1325 bdev->bd_super = s;
1326 }
1327
1328 sdp = s->s_fs_info;
f55073ff 1329 if (args.ar_meta)
8bcbbf00 1330 return dget(sdp->sd_master_dir);
f55073ff 1331 else
8bcbbf00 1332 return dget(sdp->sd_root_dir);
f55073ff
SW
1333
1334error_super:
1335 deactivate_locked_super(s);
8bcbbf00 1336 return ERR_PTR(error);
f55073ff 1337error_bdev:
d4d77629 1338 blkdev_put(bdev, mode);
8bcbbf00 1339 return ERR_PTR(error);
f55073ff
SW
1340}
1341
f6d03139
SW
1342static int set_meta_super(struct super_block *s, void *ptr)
1343{
1344 return -EINVAL;
1345}
1346
8bcbbf00
AV
1347static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1348 int flags, const char *dev_name, void *data)
86384605 1349{
f6d03139 1350 struct super_block *s;
003dec89 1351 struct gfs2_sbd *sdp;
e24977d4 1352 struct path path;
86384605 1353 int error;
907b9bce 1354
e24977d4 1355 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1356 if (error) {
d77d1b58
JP
1357 pr_warn("path_lookup on %s returned error %d\n",
1358 dev_name, error);
8bcbbf00 1359 return ERR_PTR(error);
86384605 1360 }
9249e17f 1361 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
fc64005c 1362 path.dentry->d_sb->s_bdev);
e24977d4 1363 path_put(&path);
f6d03139 1364 if (IS_ERR(s)) {
d77d1b58 1365 pr_warn("gfs2 mount does not exist\n");
8bcbbf00 1366 return ERR_CAST(s);
86384605 1367 }
f55073ff
SW
1368 if ((flags ^ s->s_flags) & MS_RDONLY) {
1369 deactivate_locked_super(s);
8bcbbf00 1370 return ERR_PTR(-EBUSY);
f55073ff 1371 }
f6d03139 1372 sdp = s->s_fs_info;
8bcbbf00 1373 return dget(sdp->sd_master_dir);
b3b94faa
DT
1374}
1375
419c93e0 1376static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1377{
1378 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1379
1380 if (sdp == NULL) {
1381 kill_block_super(sb);
1382 return;
3af165ac 1383 }
88a19ad0 1384
24972557 1385 gfs2_log_flush(sdp, NULL, SYNC_FLUSH);
88a19ad0
SW
1386 dput(sdp->sd_root_dir);
1387 dput(sdp->sd_master_dir);
1388 sdp->sd_root_dir = NULL;
1389 sdp->sd_master_dir = NULL;
fefc03bf 1390 shrink_dcache_sb(sb);
88a19ad0 1391 gfs2_delete_debugfs_file(sdp);
a245769f 1392 free_percpu(sdp->sd_lkstats);
0d515210 1393 kill_block_super(sb);
86384605
AD
1394}
1395
b3b94faa
DT
1396struct file_system_type gfs2_fs_type = {
1397 .name = "gfs2",
1398 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1399 .mount = gfs2_mount,
419c93e0
SW
1400 .kill_sb = gfs2_kill_sb,
1401 .owner = THIS_MODULE,
1402};
7f78e035 1403MODULE_ALIAS_FS("gfs2");
419c93e0
SW
1404
1405struct file_system_type gfs2meta_fs_type = {
1406 .name = "gfs2meta",
1407 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1408 .mount = gfs2_mount_meta,
b3b94faa
DT
1409 .owner = THIS_MODULE,
1410};
7f78e035 1411MODULE_ALIAS_FS("gfs2meta");