]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/gfs2/ops_fstype.c
GFS2: Move recovery variables to journal structure in memory
[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
10#include <linux/sched.h>
11#include <linux/slab.h>
12#include <linux/spinlock.h>
13#include <linux/completion.h>
14#include <linux/buffer_head.h>
b3b94faa
DT
15#include <linux/blkdev.h>
16#include <linux/kthread.h>
afeacc8c 17#include <linux/export.h>
86384605
AD
18#include <linux/namei.h>
19#include <linux/mount.h>
5c676f6d 20#include <linux/gfs2_ondisk.h>
cc632e7f 21#include <linux/quotaops.h>
56aa72d0 22#include <linux/lockdep.h>
7f78e035 23#include <linux/module.h>
b3b94faa
DT
24
25#include "gfs2.h"
5c676f6d 26#include "incore.h"
da6dd40d 27#include "bmap.h"
b3b94faa
DT
28#include "glock.h"
29#include "glops.h"
30#include "inode.h"
b3b94faa
DT
31#include "recovery.h"
32#include "rgrp.h"
33#include "super.h"
b3b94faa 34#include "sys.h"
5c676f6d 35#include "util.h"
bb3b0e3d 36#include "log.h"
9ac1b4d9 37#include "quota.h"
b5289681 38#include "dir.h"
70d4ee94 39#include "meta_io.h"
63997775 40#include "trace_gfs2.h"
b3b94faa
DT
41
42#define DO 0
43#define UNDO 1
44
9b8df98f
SW
45/**
46 * gfs2_tune_init - Fill a gfs2_tune structure with default values
47 * @gt: tune
48 *
49 */
50
51static void gfs2_tune_init(struct gfs2_tune *gt)
52{
53 spin_lock_init(&gt->gt_spin);
54
9b8df98f
SW
55 gt->gt_quota_warn_period = 10;
56 gt->gt_quota_scale_num = 1;
57 gt->gt_quota_scale_den = 1;
9b8df98f
SW
58 gt->gt_new_files_jdata = 0;
59 gt->gt_max_readahead = 1 << 18;
9b8df98f 60 gt->gt_complain_secs = 10;
9b8df98f
SW
61}
62
b3b94faa
DT
63static struct gfs2_sbd *init_sbd(struct super_block *sb)
64{
65 struct gfs2_sbd *sdp;
70d4ee94 66 struct address_space *mapping;
b3b94faa 67
85d1da67 68 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
69 if (!sdp)
70 return NULL;
71
5c676f6d 72 sb->s_fs_info = sdp;
b3b94faa 73 sdp->sd_vfs = sb;
a245769f
SW
74 sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
75 if (!sdp->sd_lkstats) {
76 kfree(sdp);
77 return NULL;
78 }
79
ba6e9364 80 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
81 gfs2_tune_init(&sdp->sd_tune);
82
e402746a
SW
83 init_waitqueue_head(&sdp->sd_glock_wait);
84 atomic_set(&sdp->sd_glock_disposal, 0);
3942ae53 85 init_completion(&sdp->sd_locking_init);
fd95e81c 86 init_completion(&sdp->sd_wdack);
b3b94faa 87 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
88
89 spin_lock_init(&sdp->sd_rindex_spin);
7c9ca621 90 sdp->sd_rindex_tree.rb_node = NULL;
b3b94faa
DT
91
92 INIT_LIST_HEAD(&sdp->sd_jindex_list);
93 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 94 mutex_init(&sdp->sd_jindex_mutex);
b3b94faa 95
b3b94faa 96 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 97 mutex_init(&sdp->sd_quota_mutex);
e46c772d 98 mutex_init(&sdp->sd_quota_sync_mutex);
37b2c837 99 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
100 INIT_LIST_HEAD(&sdp->sd_trunc_list);
101 spin_lock_init(&sdp->sd_trunc_lock);
2d9e7230 102 spin_lock_init(&sdp->sd_bitmap_lock);
b3b94faa 103
70d4ee94
SW
104 mapping = &sdp->sd_aspace;
105
39849d69 106 address_space_init_once(mapping);
70d4ee94
SW
107 mapping->a_ops = &gfs2_meta_aops;
108 mapping->host = sb->s_bdev->bd_inode;
109 mapping->flags = 0;
110 mapping_set_gfp_mask(mapping, GFP_NOFS);
111 mapping->private_data = NULL;
112 mapping->backing_dev_info = sb->s_bdi;
113 mapping->writeback_index = 0;
114
b3b94faa 115 spin_lock_init(&sdp->sd_log_lock);
5e687eac 116 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa 117 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
d7b616e2 118 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
45138990 119 spin_lock_init(&sdp->sd_ordered_lock);
b3b94faa 120
5e687eac
BM
121 init_waitqueue_head(&sdp->sd_log_waitq);
122 init_waitqueue_head(&sdp->sd_logd_waitq);
d6a079e8 123 spin_lock_init(&sdp->sd_ail_lock);
b3b94faa
DT
124 INIT_LIST_HEAD(&sdp->sd_ail1_list);
125 INIT_LIST_HEAD(&sdp->sd_ail2_list);
126
484adff8 127 init_rwsem(&sdp->sd_log_flush_lock);
16615be1
SW
128 atomic_set(&sdp->sd_log_in_flight, 0);
129 init_waitqueue_head(&sdp->sd_log_flush_wait);
b3b94faa 130
b3b94faa
DT
131 return sdp;
132}
133
b3b94faa 134
9b8df98f
SW
135/**
136 * gfs2_check_sb - Check superblock
137 * @sdp: the filesystem
138 * @sb: The superblock
139 * @silent: Don't print a message if the check fails
140 *
141 * Checks the version code of the FS is one that we understand how to
142 * read and that the sizes of the various on-disk structures have not
143 * changed.
144 */
145
32e471ef 146static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 147{
32e471ef
SW
148 struct gfs2_sb_host *sb = &sdp->sd_sb;
149
9b8df98f
SW
150 if (sb->sb_magic != GFS2_MAGIC ||
151 sb->sb_type != GFS2_METATYPE_SB) {
152 if (!silent)
fc554ed3 153 pr_warn("GFS2: not a GFS2 filesystem\n");
9b8df98f
SW
154 return -EINVAL;
155 }
156
157 /* If format numbers match exactly, we're done. */
158
159 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
160 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
161 return 0;
162
c80dbb58 163 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 164
c80dbb58 165 return -EINVAL;
9b8df98f
SW
166}
167
168static void end_bio_io_page(struct bio *bio, int error)
169{
170 struct page *page = bio->bi_private;
171
172 if (!error)
173 SetPageUptodate(page);
174 else
fc554ed3 175 pr_warn("gfs2: error %d reading superblock\n", error);
9b8df98f
SW
176 unlock_page(page);
177}
178
32e471ef 179static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 180{
32e471ef
SW
181 struct gfs2_sb_host *sb = &sdp->sd_sb;
182 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
183 const struct gfs2_sb *str = buf;
184
185 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
186 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
187 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
188 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
189 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
190 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
191 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
192 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
193 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
194 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
195 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
196
197 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
198 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 199 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
200}
201
202/**
203 * gfs2_read_super - Read the gfs2 super block from disk
204 * @sdp: The GFS2 super block
205 * @sector: The location of the super block
206 * @error: The error code to return
207 *
208 * This uses the bio functions to read the super block from disk
209 * because we want to be 100% sure that we never read cached data.
210 * A super block is read twice only during each GFS2 mount and is
211 * never written to by the filesystem. The first time its read no
212 * locks are held, and the only details which are looked at are those
213 * relating to the locking protocol. Once locking is up and working,
214 * the sb is read again under the lock to establish the location of
215 * the master directory (contains pointers to journals etc) and the
216 * root directory.
217 *
218 * Returns: 0 on success or error
219 */
220
32e471ef 221static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
222{
223 struct super_block *sb = sdp->sd_vfs;
224 struct gfs2_sb *p;
225 struct page *page;
226 struct bio *bio;
227
228 page = alloc_page(GFP_NOFS);
229 if (unlikely(!page))
ac3beb6a 230 return -ENOMEM;
9b8df98f
SW
231
232 ClearPageUptodate(page);
233 ClearPageDirty(page);
234 lock_page(page);
235
236 bio = bio_alloc(GFP_NOFS, 1);
4f024f37 237 bio->bi_iter.bi_sector = sector * (sb->s_blocksize >> 9);
9b8df98f
SW
238 bio->bi_bdev = sb->s_bdev;
239 bio_add_page(bio, page, PAGE_SIZE, 0);
240
241 bio->bi_end_io = end_bio_io_page;
242 bio->bi_private = page;
20ed0535 243 submit_bio(READ_SYNC | REQ_META, bio);
9b8df98f
SW
244 wait_on_page_locked(page);
245 bio_put(bio);
246 if (!PageUptodate(page)) {
247 __free_page(page);
248 return -EIO;
249 }
250 p = kmap(page);
32e471ef 251 gfs2_sb_in(sdp, p);
9b8df98f
SW
252 kunmap(page);
253 __free_page(page);
32e471ef 254 return gfs2_check_sb(sdp, silent);
9b8df98f 255}
6bac243f 256
9b8df98f
SW
257/**
258 * gfs2_read_sb - Read super block
259 * @sdp: The GFS2 superblock
9b8df98f
SW
260 * @silent: Don't print message if mount fails
261 *
262 */
263
6bac243f 264static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
265{
266 u32 hash_blocks, ind_blocks, leaf_blocks;
267 u32 tmp_blocks;
268 unsigned int x;
269 int error;
270
32e471ef 271 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
272 if (error) {
273 if (!silent)
274 fs_err(sdp, "can't read superblock\n");
275 return error;
276 }
277
9b8df98f
SW
278 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
279 GFS2_BASIC_BLOCK_SHIFT;
280 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
281 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
282 sizeof(struct gfs2_dinode)) / sizeof(u64);
283 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
284 sizeof(struct gfs2_meta_header)) / sizeof(u64);
285 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
286 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
287 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
288 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
289 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
290 sizeof(struct gfs2_meta_header)) /
291 sizeof(struct gfs2_quota_change);
a68a0a35
BP
292 sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
293 sizeof(struct gfs2_meta_header))
294 * GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
9b8df98f
SW
295
296 /* Compute maximum reservation required to add a entry to a directory */
297
298 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
299 sdp->sd_jbsize);
300
301 ind_blocks = 0;
302 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
303 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
304 ind_blocks += tmp_blocks;
305 }
306
307 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
308
309 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
310
311 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
312 sizeof(struct gfs2_dinode);
313 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
314 for (x = 2;; x++) {
315 u64 space, d;
316 u32 m;
317
318 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
319 d = space;
320 m = do_div(d, sdp->sd_inptrs);
321
322 if (d != sdp->sd_heightsize[x - 1] || m)
323 break;
324 sdp->sd_heightsize[x] = space;
325 }
326 sdp->sd_max_height = x;
327 sdp->sd_heightsize[x] = ~0;
328 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
329
330 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
331 sizeof(struct gfs2_dinode);
332 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
333 for (x = 2;; x++) {
334 u64 space, d;
335 u32 m;
336
337 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
338 d = space;
339 m = do_div(d, sdp->sd_inptrs);
340
341 if (d != sdp->sd_jheightsize[x - 1] || m)
342 break;
343 sdp->sd_jheightsize[x] = space;
344 }
345 sdp->sd_max_jheight = x;
346 sdp->sd_jheightsize[x] = ~0;
347 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
348
349 return 0;
350}
351
b3b94faa
DT
352static int init_names(struct gfs2_sbd *sdp, int silent)
353{
b3b94faa
DT
354 char *proto, *table;
355 int error = 0;
356
357 proto = sdp->sd_args.ar_lockproto;
358 table = sdp->sd_args.ar_locktable;
359
360 /* Try to autodetect */
361
362 if (!proto[0] || !table[0]) {
32e471ef 363 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
364 if (error)
365 return error;
3cf1e7be 366
b3b94faa 367 if (!proto[0])
3cf1e7be 368 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 369 if (!table[0])
3cf1e7be 370 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
371 }
372
373 if (!table[0])
374 table = sdp->sd_vfs->s_id;
375
00377d8e
JD
376 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
377 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 378
5d35e31f
DC
379 table = sdp->sd_table_name;
380 while ((table = strchr(table, '/')))
b35997d4
RP
381 *table = '_';
382
b3b94faa
DT
383 return error;
384}
385
386static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
387 int undo)
388{
b3b94faa
DT
389 int error = 0;
390
391 if (undo)
392 goto fail_trans;
393
b3b94faa
DT
394 error = gfs2_glock_nq_num(sdp,
395 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
396 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
397 mount_gh);
398 if (error) {
399 fs_err(sdp, "can't acquire mount glock: %d\n", error);
400 goto fail;
401 }
402
403 error = gfs2_glock_nq_num(sdp,
404 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
405 LM_ST_SHARED,
579b78a4 406 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
407 &sdp->sd_live_gh);
408 if (error) {
409 fs_err(sdp, "can't acquire live glock: %d\n", error);
410 goto fail_mount;
411 }
412
413 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
414 CREATE, &sdp->sd_rename_gl);
415 if (error) {
416 fs_err(sdp, "can't create rename glock: %d\n", error);
417 goto fail_live;
418 }
419
420 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
421 CREATE, &sdp->sd_trans_gl);
422 if (error) {
423 fs_err(sdp, "can't create transaction glock: %d\n", error);
424 goto fail_rename;
425 }
b3b94faa
DT
426
427 return 0;
428
feaa7bba 429fail_trans:
b3b94faa 430 gfs2_glock_put(sdp->sd_trans_gl);
feaa7bba 431fail_rename:
b3b94faa 432 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 433fail_live:
b3b94faa 434 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 435fail_mount:
b3b94faa 436 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 437fail:
b3b94faa
DT
438 return error;
439}
440
9b8df98f
SW
441static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
442 u64 no_addr, const char *name)
f42faf4f 443{
9b8df98f
SW
444 struct gfs2_sbd *sdp = sb->s_fs_info;
445 struct dentry *dentry;
446 struct inode *inode;
447
44ad37d6 448 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
9b8df98f
SW
449 if (IS_ERR(inode)) {
450 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
451 return PTR_ERR(inode);
452 }
48fde701 453 dentry = d_make_root(inode);
9b8df98f
SW
454 if (!dentry) {
455 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
456 return -ENOMEM;
457 }
9b8df98f
SW
458 *dptr = dentry;
459 return 0;
f42faf4f
SW
460}
461
9b8df98f 462static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
463{
464 struct super_block *sb = sdp->sd_vfs;
465 struct gfs2_holder sb_gh;
dbb7cae2 466 u64 no_addr;
9b8df98f 467 int ret;
b3b94faa 468
9b8df98f
SW
469 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
470 LM_ST_SHARED, 0, &sb_gh);
471 if (ret) {
472 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
473 return ret;
b3b94faa 474 }
907b9bce 475
6bac243f 476 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
477 if (ret) {
478 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
479 goto out;
480 }
481
482 /* Set up the buffer cache and SB for real */
e1defc4f 483 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 484 ret = -EINVAL;
b3b94faa
DT
485 fs_err(sdp, "FS block size (%u) is too small for device "
486 "block size (%u)\n",
e1defc4f 487 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
488 goto out;
489 }
490 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 491 ret = -EINVAL;
b3b94faa
DT
492 fs_err(sdp, "FS block size (%u) is too big for machine "
493 "page size (%u)\n",
494 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
495 goto out;
496 }
b3b94faa
DT
497 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
498
f42faf4f 499 /* Get the root inode */
dbb7cae2 500 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
501 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
502 if (ret)
f42faf4f 503 goto out;
b3b94faa 504
9b8df98f
SW
505 /* Get the master inode */
506 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
507 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
508 if (ret) {
509 dput(sdp->sd_root_dir);
510 goto out;
511 }
512 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 513out:
b3b94faa 514 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 515 return ret;
b3b94faa
DT
516}
517
f057f6cd 518static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 519{
f057f6cd
SW
520 char *message = "FIRSTMOUNT=Done";
521 char *envp[] = { message, NULL };
e0c2a9aa
DT
522
523 fs_info(sdp, "first mount done, others may mount\n");
524
525 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
526 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
527
f057f6cd 528 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
529}
530
b5289681
SW
531/**
532 * gfs2_jindex_hold - Grab a lock on the jindex
533 * @sdp: The GFS2 superblock
534 * @ji_gh: the holder for the jindex glock
535 *
536 * Returns: errno
537 */
538
539static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
540{
541 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
542 struct qstr name;
543 char buf[20];
544 struct gfs2_jdesc *jd;
545 int error;
546
547 name.name = buf;
548
549 mutex_lock(&sdp->sd_jindex_mutex);
550
551 for (;;) {
552 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
553 if (error)
554 break;
555
556 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
557 name.hash = gfs2_disk_hash(name.name, name.len);
558
559 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
560 if (error == -ENOENT) {
561 error = 0;
562 break;
563 }
564
565 gfs2_glock_dq_uninit(ji_gh);
566
567 if (error)
568 break;
569
570 error = -ENOMEM;
571 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
572 if (!jd)
573 break;
574
575 INIT_LIST_HEAD(&jd->extent_list);
a17d758b
BP
576 INIT_LIST_HEAD(&jd->jd_revoke_list);
577
6ecd7c2d 578 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
579 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
580 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
581 if (!jd->jd_inode)
582 error = -ENOENT;
583 else
584 error = PTR_ERR(jd->jd_inode);
585 kfree(jd);
586 break;
587 }
588
589 spin_lock(&sdp->sd_jindex_spin);
590 jd->jd_jid = sdp->sd_journals++;
591 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
592 spin_unlock(&sdp->sd_jindex_spin);
593 }
594
595 mutex_unlock(&sdp->sd_jindex_mutex);
596
597 return error;
598}
599
1d12d175
BP
600/**
601 * check_journal_clean - Make sure a journal is clean for a spectator mount
602 * @sdp: The GFS2 superblock
603 * @jd: The journal descriptor
604 *
605 * Returns: 0 if the journal is clean or locked, else an error
606 */
607static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
608{
609 int error;
610 struct gfs2_holder j_gh;
611 struct gfs2_log_header_host head;
612 struct gfs2_inode *ip;
613
614 ip = GFS2_I(jd->jd_inode);
615 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
616 GL_EXACT | GL_NOCACHE, &j_gh);
617 if (error) {
618 fs_err(sdp, "Error locking journal for spectator mount.\n");
619 return -EPERM;
620 }
621 error = gfs2_jdesc_check(jd);
622 if (error) {
623 fs_err(sdp, "Error checking journal for spectator mount.\n");
624 goto out_unlock;
625 }
626 error = gfs2_find_jhead(jd, &head);
627 if (error) {
628 fs_err(sdp, "Error parsing journal for spectator mount.\n");
629 goto out_unlock;
630 }
631 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
632 error = -EPERM;
633 fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
634 "must not be a spectator.\n", jd->jd_jid);
635 }
636
637out_unlock:
638 gfs2_glock_dq_uninit(&j_gh);
639 return error;
640}
641
b3b94faa
DT
642static int init_journal(struct gfs2_sbd *sdp, int undo)
643{
9b8df98f 644 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 645 struct gfs2_holder ji_gh;
5c676f6d 646 struct gfs2_inode *ip;
b3b94faa
DT
647 int jindex = 1;
648 int error = 0;
649
650 if (undo) {
651 jindex = 0;
fe64d517 652 goto fail_jinode_gh;
b3b94faa
DT
653 }
654
9b8df98f 655 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 656 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 657 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 658 return PTR_ERR(sdp->sd_jindex);
b3b94faa 659 }
b3b94faa
DT
660
661 /* Load in the journal index special file */
662
663 error = gfs2_jindex_hold(sdp, &ji_gh);
664 if (error) {
665 fs_err(sdp, "can't read journal index: %d\n", error);
666 goto fail;
667 }
668
0e5a9fb0 669 error = -EUSERS;
b3b94faa
DT
670 if (!gfs2_jindex_size(sdp)) {
671 fs_err(sdp, "no journals!\n");
907b9bce 672 goto fail_jindex;
b3b94faa
DT
673 }
674
675 if (sdp->sd_args.ar_spectator) {
676 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 677 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
678 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
679 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
680 } else {
681 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
682 fs_err(sdp, "can't mount journal #%u\n",
683 sdp->sd_lockstruct.ls_jid);
684 fs_err(sdp, "there are only %u journals (0 - %u)\n",
685 gfs2_jindex_size(sdp),
686 gfs2_jindex_size(sdp) - 1);
687 goto fail_jindex;
688 }
689 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
690
feaa7bba 691 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
692 &gfs2_journal_glops,
693 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
694 &sdp->sd_journal_gh);
695 if (error) {
696 fs_err(sdp, "can't acquire journal glock: %d\n", error);
697 goto fail_jindex;
698 }
699
feaa7bba
SW
700 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
701 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 702 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
703 &sdp->sd_jinode_gh);
704 if (error) {
705 fs_err(sdp, "can't acquire journal inode glock: %d\n",
706 error);
707 goto fail_journal_gh;
708 }
709
710 error = gfs2_jdesc_check(sdp->sd_jdesc);
711 if (error) {
712 fs_err(sdp, "my journal (%u) is bad: %d\n",
713 sdp->sd_jdesc->jd_jid, error);
714 goto fail_jinode_gh;
715 }
fd041f0b 716 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
717 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
718 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
719
720 /* Map the extents for this journal's blocks */
b50f227b 721 gfs2_map_journal_extents(sdp, sdp->sd_jdesc);
b3b94faa 722 }
63997775 723 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
724
725 if (sdp->sd_lockstruct.ls_first) {
726 unsigned int x;
727 for (x = 0; x < sdp->sd_journals; x++) {
1d12d175
BP
728 struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
729
730 if (sdp->sd_args.ar_spectator) {
731 error = check_journal_clean(sdp, jd);
732 if (error)
733 goto fail_jinode_gh;
734 continue;
735 }
736 error = gfs2_recover_journal(jd, true);
b3b94faa
DT
737 if (error) {
738 fs_err(sdp, "error recovering journal %u: %d\n",
739 x, error);
740 goto fail_jinode_gh;
741 }
742 }
743
f057f6cd 744 gfs2_others_may_mount(sdp);
b3b94faa 745 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 746 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
747 if (error) {
748 fs_err(sdp, "error recovering my journal: %d\n", error);
749 goto fail_jinode_gh;
750 }
751 }
752
753 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
754 gfs2_glock_dq_uninit(&ji_gh);
755 jindex = 0;
756
b3b94faa
DT
757 return 0;
758
a91ea69f 759fail_jinode_gh:
b3b94faa
DT
760 if (!sdp->sd_args.ar_spectator)
761 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 762fail_journal_gh:
b3b94faa
DT
763 if (!sdp->sd_args.ar_spectator)
764 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 765fail_jindex:
b3b94faa
DT
766 gfs2_jindex_free(sdp);
767 if (jindex)
768 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 769fail:
f42faf4f 770 iput(sdp->sd_jindex);
b3b94faa
DT
771 return error;
772}
773
56aa72d0 774static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
775
776static int init_inodes(struct gfs2_sbd *sdp, int undo)
777{
b3b94faa 778 int error = 0;
9b8df98f 779 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
780
781 if (undo)
f42faf4f
SW
782 goto fail_qinode;
783
f42faf4f
SW
784 error = init_journal(sdp, undo);
785 if (error)
9b8df98f 786 goto fail;
b3b94faa 787
b3b94faa 788 /* Read in the master statfs inode */
9b8df98f 789 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
790 if (IS_ERR(sdp->sd_statfs_inode)) {
791 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 792 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 793 goto fail_journal;
b3b94faa
DT
794 }
795
796 /* Read in the resource index inode */
9b8df98f 797 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
798 if (IS_ERR(sdp->sd_rindex)) {
799 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
800 fs_err(sdp, "can't get resource index inode: %d\n", error);
801 goto fail_statfs;
802 }
cf45b752 803 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
804
805 /* Read in the quota inode */
9b8df98f 806 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
807 if (IS_ERR(sdp->sd_quota_inode)) {
808 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
809 fs_err(sdp, "can't get quota file inode: %d\n", error);
810 goto fail_rindex;
811 }
56aa72d0
JK
812 /*
813 * i_mutex on quota files is special. Since this inode is hidden system
814 * file, we are safe to define locking ourselves.
815 */
816 lockdep_set_class(&sdp->sd_quota_inode->i_mutex,
817 &gfs2_quota_imutex_key);
a365fbf3
SW
818
819 error = gfs2_rindex_update(sdp);
820 if (error)
821 goto fail_qinode;
822
b3b94faa
DT
823 return 0;
824
f42faf4f
SW
825fail_qinode:
826 iput(sdp->sd_quota_inode);
f42faf4f 827fail_rindex:
b3b94faa 828 gfs2_clear_rgrpd(sdp);
f42faf4f 829 iput(sdp->sd_rindex);
f42faf4f
SW
830fail_statfs:
831 iput(sdp->sd_statfs_inode);
f42faf4f
SW
832fail_journal:
833 init_journal(sdp, UNDO);
f42faf4f 834fail:
b3b94faa
DT
835 return error;
836}
837
838static int init_per_node(struct gfs2_sbd *sdp, int undo)
839{
f42faf4f 840 struct inode *pn = NULL;
b3b94faa
DT
841 char buf[30];
842 int error = 0;
5c676f6d 843 struct gfs2_inode *ip;
9b8df98f 844 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
845
846 if (sdp->sd_args.ar_spectator)
847 return 0;
848
849 if (undo)
850 goto fail_qc_gh;
851
9b8df98f 852 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
853 if (IS_ERR(pn)) {
854 error = PTR_ERR(pn);
b3b94faa
DT
855 fs_err(sdp, "can't find per_node directory: %d\n", error);
856 return error;
857 }
858
b3b94faa 859 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
860 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
861 if (IS_ERR(sdp->sd_sc_inode)) {
862 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 863 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 864 goto fail;
b3b94faa
DT
865 }
866
b3b94faa 867 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
868 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
869 if (IS_ERR(sdp->sd_qc_inode)) {
870 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
871 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
872 goto fail_ut_i;
873 }
874
f42faf4f 875 iput(pn);
b3b94faa
DT
876 pn = NULL;
877
feaa7bba 878 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 879 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
880 &sdp->sd_sc_gh);
881 if (error) {
882 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 883 goto fail_qc_i;
b3b94faa
DT
884 }
885
feaa7bba 886 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 887 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
888 &sdp->sd_qc_gh);
889 if (error) {
890 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
891 goto fail_ut_gh;
892 }
893
894 return 0;
895
a91ea69f 896fail_qc_gh:
b3b94faa 897 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 898fail_ut_gh:
b3b94faa 899 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 900fail_qc_i:
f42faf4f 901 iput(sdp->sd_qc_inode);
a91ea69f 902fail_ut_i:
f42faf4f 903 iput(sdp->sd_sc_inode);
a91ea69f 904fail:
b3b94faa 905 if (pn)
f42faf4f 906 iput(pn);
b3b94faa
DT
907 return error;
908}
909
f057f6cd
SW
910static const match_table_t nolock_tokens = {
911 { Opt_jid, "jid=%d\n", },
912 { Opt_err, NULL },
913};
914
915static const struct lm_lockops nolock_ops = {
916 .lm_proto_name = "lock_nolock",
fc0e38da 917 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
918 .lm_tokens = &nolock_tokens,
919};
920
da755fdb
SW
921/**
922 * gfs2_lm_mount - mount a locking protocol
923 * @sdp: the filesystem
3ad2f3fb 924 * @args: mount arguments
da755fdb
SW
925 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
926 *
927 * Returns: errno
928 */
929
930static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
931{
f057f6cd
SW
932 const struct lm_lockops *lm;
933 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
934 struct gfs2_args *args = &sdp->sd_args;
935 const char *proto = sdp->sd_proto_name;
936 const char *table = sdp->sd_table_name;
f057f6cd
SW
937 char *o, *options;
938 int ret;
da755fdb 939
f057f6cd
SW
940 if (!strcmp("lock_nolock", proto)) {
941 lm = &nolock_ops;
942 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
943#ifdef CONFIG_GFS2_FS_LOCKING_DLM
944 } else if (!strcmp("lock_dlm", proto)) {
945 lm = &gfs2_dlm_ops;
946#endif
947 } else {
fc554ed3 948 pr_info("GFS2: can't find protocol %s\n", proto);
f057f6cd
SW
949 return -ENOENT;
950 }
da755fdb
SW
951
952 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
953
f057f6cd
SW
954 ls->ls_ops = lm;
955 ls->ls_first = 1;
da755fdb 956
f057f6cd
SW
957 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
958 substring_t tmp[MAX_OPT_ARGS];
959 int token, option;
960
961 if (!o || !*o)
962 continue;
963
964 token = match_token(o, *lm->lm_tokens, tmp);
965 switch (token) {
966 case Opt_jid:
967 ret = match_int(&tmp[0], &option);
968 if (ret || option < 0)
969 goto hostdata_error;
ba6e9364
SW
970 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
971 ls->ls_jid = option;
f057f6cd
SW
972 break;
973 case Opt_id:
4875647a 974 case Opt_nodir:
2b88f7c5 975 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
976 break;
977 case Opt_first:
978 ret = match_int(&tmp[0], &option);
979 if (ret || (option != 0 && option != 1))
980 goto hostdata_error;
981 ls->ls_first = option;
982 break;
f057f6cd
SW
983 case Opt_err:
984 default:
985hostdata_error:
986 fs_info(sdp, "unknown hostdata (%s)\n", o);
987 return -EINVAL;
988 }
da755fdb
SW
989 }
990
f057f6cd
SW
991 if (lm->lm_mount == NULL) {
992 fs_info(sdp, "Now mounting FS...\n");
19237039 993 complete_all(&sdp->sd_locking_init);
f057f6cd 994 return 0;
da755fdb 995 }
e0c2a9aa 996 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
997 if (ret == 0)
998 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 999 complete_all(&sdp->sd_locking_init);
f057f6cd 1000 return ret;
da755fdb
SW
1001}
1002
1003void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1004{
f057f6cd
SW
1005 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1006 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1007 lm->lm_unmount)
1008 lm->lm_unmount(sdp);
da755fdb 1009}
b3b94faa 1010
ba6e9364
SW
1011static int gfs2_journalid_wait(void *word)
1012{
1013 if (signal_pending(current))
1014 return -EINTR;
1015 schedule();
1016 return 0;
1017}
1018
1019static int wait_on_journal(struct gfs2_sbd *sdp)
1020{
ba6e9364
SW
1021 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1022 return 0;
1023
1024 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1025}
1026
8633ecfa
SW
1027void gfs2_online_uevent(struct gfs2_sbd *sdp)
1028{
1029 struct super_block *sb = sdp->sd_vfs;
1030 char ro[20];
1031 char spectator[20];
1032 char *envp[] = { ro, spectator, NULL };
1033 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1034 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1035 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1036}
1037
b3b94faa
DT
1038/**
1039 * fill_super - Read in superblock
1040 * @sb: The VFS superblock
1041 * @data: Mount options
1042 * @silent: Don't complain if it's not a GFS2 filesystem
1043 *
1044 * Returns: errno
1045 */
1046
f55073ff 1047static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1048{
1049 struct gfs2_sbd *sdp;
1050 struct gfs2_holder mount_gh;
1051 int error;
1052
1053 sdp = init_sbd(sb);
1054 if (!sdp) {
fc554ed3 1055 pr_warn("GFS2: can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1056 return -ENOMEM;
1057 }
f55073ff 1058 sdp->sd_args = *args;
b3b94faa 1059
fe64d517 1060 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1061 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1062 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1063 }
6f04c1c7
SW
1064 if (sdp->sd_args.ar_posix_acl)
1065 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1066 if (sdp->sd_args.ar_nobarrier)
1067 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1068
9964afbb 1069 sb->s_flags |= MS_NOSEC;
719ee344
SW
1070 sb->s_magic = GFS2_MAGIC;
1071 sb->s_op = &gfs2_super_ops;
41ced6dc 1072 sb->s_d_op = &gfs2_dops;
719ee344 1073 sb->s_export_op = &gfs2_export_ops;
40b78a32 1074 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1075 sb->s_qcop = &gfs2_quotactl_ops;
1076 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1077 sb->s_time_gran = 1;
1078 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1079
1080 /* Set up the buffer cache and fill in some fake block size values
1081 to allow us to read-in the on-disk superblock. */
1082 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1083 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1084 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1085 GFS2_BASIC_BLOCK_SHIFT;
1086 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1087
5e687eac 1088 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1089 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1090 if (sdp->sd_args.ar_statfs_quantum) {
1091 sdp->sd_tune.gt_statfs_slow = 0;
1092 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1093 } else {
3d3c10f2
BM
1094 sdp->sd_tune.gt_statfs_slow = 1;
1095 sdp->sd_tune.gt_statfs_quantum = 30;
1096 }
48c2b613 1097
b3b94faa 1098 error = init_names(sdp, silent);
0d515210
BP
1099 if (error) {
1100 /* In this case, we haven't initialized sysfs, so we have to
1101 manually free the sdp. */
1102 free_percpu(sdp->sd_lkstats);
1103 kfree(sdp);
1104 sb->s_fs_info = NULL;
1105 return error;
1106 }
b3b94faa 1107
e0c2a9aa
DT
1108 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1109
b3b94faa 1110 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1111 /*
1112 * If we hit an error here, gfs2_sys_fs_add will have called function
1113 * kobject_put which causes the sysfs usage count to go to zero, which
1114 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1115 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1116 * kobject_put to free sdp.
1117 */
b3b94faa 1118 if (error)
0d515210
BP
1119 return error;
1120
1121 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1122
1123 error = gfs2_lm_mount(sdp, silent);
1124 if (error)
0d515210 1125 goto fail_debug;
b3b94faa
DT
1126
1127 error = init_locking(sdp, &mount_gh, DO);
1128 if (error)
1129 goto fail_lm;
1130
9b8df98f 1131 error = init_sb(sdp, silent);
b3b94faa
DT
1132 if (error)
1133 goto fail_locking;
b3b94faa 1134
ba6e9364
SW
1135 error = wait_on_journal(sdp);
1136 if (error)
1137 goto fail_sb;
1138
feb47ca9
SW
1139 /*
1140 * If user space has failed to join the cluster or some similar
1141 * failure has occurred, then the journal id will contain a
1142 * negative (error) number. This will then be returned to the
1143 * caller (of the mount syscall). We do this even for spectator
1144 * mounts (which just write a jid of 0 to indicate "ok" even though
1145 * the jid is unused in the spectator case)
1146 */
1147 if (sdp->sd_lockstruct.ls_jid < 0) {
1148 error = sdp->sd_lockstruct.ls_jid;
1149 sdp->sd_lockstruct.ls_jid = 0;
1150 goto fail_sb;
1151 }
1152
e0c2a9aa
DT
1153 if (sdp->sd_args.ar_spectator)
1154 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1155 sdp->sd_table_name);
1156 else
1157 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1158 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1159
b3b94faa
DT
1160 error = init_inodes(sdp, DO);
1161 if (error)
f42faf4f 1162 goto fail_sb;
b3b94faa
DT
1163
1164 error = init_per_node(sdp, DO);
1165 if (error)
1166 goto fail_inodes;
1167
1168 error = gfs2_statfs_init(sdp);
1169 if (error) {
1170 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1171 goto fail_per_node;
1172 }
1173
b3b94faa
DT
1174 if (!(sb->s_flags & MS_RDONLY)) {
1175 error = gfs2_make_fs_rw(sdp);
1176 if (error) {
1177 fs_err(sdp, "can't make FS RW: %d\n", error);
8ad151c2 1178 goto fail_per_node;
b3b94faa
DT
1179 }
1180 }
1181
1182 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1183 gfs2_online_uevent(sdp);
b3b94faa
DT
1184 return 0;
1185
a91ea69f 1186fail_per_node:
b3b94faa 1187 init_per_node(sdp, UNDO);
a91ea69f 1188fail_inodes:
b3b94faa 1189 init_inodes(sdp, UNDO);
a91ea69f 1190fail_sb:
9b8df98f
SW
1191 if (sdp->sd_root_dir)
1192 dput(sdp->sd_root_dir);
1193 if (sdp->sd_master_dir)
1194 dput(sdp->sd_master_dir);
e7c8707e
SW
1195 if (sb->s_root)
1196 dput(sb->s_root);
9b8df98f 1197 sb->s_root = NULL;
a91ea69f 1198fail_locking:
b3b94faa 1199 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1200fail_lm:
fefc03bf 1201 gfs2_gl_hash_clear(sdp);
b3b94faa 1202 gfs2_lm_unmount(sdp);
0d515210 1203fail_debug:
5f882096 1204 gfs2_delete_debugfs_file(sdp);
a245769f 1205 free_percpu(sdp->sd_lkstats);
0d515210
BP
1206 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1207 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1208 gfs2_sys_fs_del(sdp);
5c676f6d 1209 sb->s_fs_info = NULL;
b3b94faa
DT
1210 return error;
1211}
1212
f55073ff 1213static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1214{
f55073ff
SW
1215 s->s_bdev = data;
1216 s->s_dev = s->s_bdev->bd_dev;
1217
1218 /*
1219 * We set the bdi here to the queue backing, file systems can
1220 * overwrite this in ->fill_super()
1221 */
1222 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1223 return 0;
86384605
AD
1224}
1225
f55073ff 1226static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1227{
1228 struct block_device *bdev = ptr;
1229 return (bdev == s->s_bdev);
1230}
1231
f55073ff 1232/**
8bcbbf00 1233 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1234 * @fs_type: The GFS2 filesystem type
1235 * @flags: Mount flags
1236 * @dev_name: The name of the device
1237 * @data: The mount arguments
f55073ff
SW
1238 *
1239 * Q. Why not use get_sb_bdev() ?
1240 * A. We need to select one of two root directories to mount, independent
1241 * of whether this is the initial, or subsequent, mount of this sb
1242 *
1243 * Returns: 0 or -ve on error
1244 */
1245
8bcbbf00
AV
1246static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1247 const char *dev_name, void *data)
f55073ff
SW
1248{
1249 struct block_device *bdev;
1250 struct super_block *s;
d4d77629 1251 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1252 int error;
1253 struct gfs2_args args;
1254 struct gfs2_sbd *sdp;
1255
1256 if (!(flags & MS_RDONLY))
1257 mode |= FMODE_WRITE;
1258
d4d77629 1259 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1260 if (IS_ERR(bdev))
8bcbbf00 1261 return ERR_CAST(bdev);
f55073ff
SW
1262
1263 /*
1264 * once the super is inserted into the list by sget, s_umount
1265 * will protect the lockfs code from trying to start a snapshot
1266 * while we are mounting
1267 */
1268 mutex_lock(&bdev->bd_fsfreeze_mutex);
1269 if (bdev->bd_fsfreeze_count > 0) {
1270 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1271 error = -EBUSY;
1272 goto error_bdev;
1273 }
9249e17f 1274 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1275 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1276 error = PTR_ERR(s);
1277 if (IS_ERR(s))
1278 goto error_bdev;
1279
dfe5b9ad
SW
1280 if (s->s_root) {
1281 /*
1282 * s_umount nests inside bd_mutex during
1283 * __invalidate_device(). blkdev_put() acquires
1284 * bd_mutex and can't be called under s_umount. Drop
1285 * s_umount temporarily. This is safe as we're
1286 * holding an active reference.
1287 */
1288 up_write(&s->s_umount);
d4d77629 1289 blkdev_put(bdev, mode);
dfe5b9ad
SW
1290 down_write(&s->s_umount);
1291 }
8bcbbf00 1292
f55073ff
SW
1293 memset(&args, 0, sizeof(args));
1294 args.ar_quota = GFS2_QUOTA_DEFAULT;
1295 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1296 args.ar_commit = 30;
3d3c10f2
BM
1297 args.ar_statfs_quantum = 30;
1298 args.ar_quota_quantum = 60;
f55073ff
SW
1299 args.ar_errors = GFS2_ERRORS_DEFAULT;
1300
1301 error = gfs2_mount_args(&args, data);
1302 if (error) {
fc554ed3 1303 pr_warn("GFS2: can't parse mount arguments\n");
8bcbbf00 1304 goto error_super;
f55073ff
SW
1305 }
1306
1307 if (s->s_root) {
1308 error = -EBUSY;
1309 if ((flags ^ s->s_flags) & MS_RDONLY)
1310 goto error_super;
f55073ff
SW
1311 } else {
1312 char b[BDEVNAME_SIZE];
1313
f55073ff
SW
1314 s->s_mode = mode;
1315 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1316 sb_set_blocksize(s, block_size(bdev));
1317 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1318 if (error)
1319 goto error_super;
f55073ff
SW
1320 s->s_flags |= MS_ACTIVE;
1321 bdev->bd_super = s;
1322 }
1323
1324 sdp = s->s_fs_info;
f55073ff 1325 if (args.ar_meta)
8bcbbf00 1326 return dget(sdp->sd_master_dir);
f55073ff 1327 else
8bcbbf00 1328 return dget(sdp->sd_root_dir);
f55073ff
SW
1329
1330error_super:
1331 deactivate_locked_super(s);
8bcbbf00 1332 return ERR_PTR(error);
f55073ff 1333error_bdev:
d4d77629 1334 blkdev_put(bdev, mode);
8bcbbf00 1335 return ERR_PTR(error);
f55073ff
SW
1336}
1337
f6d03139
SW
1338static int set_meta_super(struct super_block *s, void *ptr)
1339{
1340 return -EINVAL;
1341}
1342
8bcbbf00
AV
1343static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1344 int flags, const char *dev_name, void *data)
86384605 1345{
f6d03139 1346 struct super_block *s;
003dec89 1347 struct gfs2_sbd *sdp;
e24977d4 1348 struct path path;
86384605 1349 int error;
907b9bce 1350
e24977d4 1351 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1352 if (error) {
fc554ed3 1353 pr_warn("GFS2: path_lookup on %s returned error %d\n",
9b8df98f 1354 dev_name, error);
8bcbbf00 1355 return ERR_PTR(error);
86384605 1356 }
9249e17f 1357 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
f6d03139 1358 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1359 path_put(&path);
f6d03139 1360 if (IS_ERR(s)) {
fc554ed3 1361 pr_warn("GFS2: gfs2 mount does not exist\n");
8bcbbf00 1362 return ERR_CAST(s);
86384605 1363 }
f55073ff
SW
1364 if ((flags ^ s->s_flags) & MS_RDONLY) {
1365 deactivate_locked_super(s);
8bcbbf00 1366 return ERR_PTR(-EBUSY);
f55073ff 1367 }
f6d03139 1368 sdp = s->s_fs_info;
8bcbbf00 1369 return dget(sdp->sd_master_dir);
b3b94faa
DT
1370}
1371
419c93e0 1372static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1373{
1374 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1375
1376 if (sdp == NULL) {
1377 kill_block_super(sb);
1378 return;
3af165ac 1379 }
88a19ad0
SW
1380
1381 gfs2_meta_syncfs(sdp);
1382 dput(sdp->sd_root_dir);
1383 dput(sdp->sd_master_dir);
1384 sdp->sd_root_dir = NULL;
1385 sdp->sd_master_dir = NULL;
fefc03bf 1386 shrink_dcache_sb(sb);
88a19ad0 1387 gfs2_delete_debugfs_file(sdp);
a245769f 1388 free_percpu(sdp->sd_lkstats);
0d515210 1389 kill_block_super(sb);
86384605
AD
1390}
1391
b3b94faa
DT
1392struct file_system_type gfs2_fs_type = {
1393 .name = "gfs2",
1394 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1395 .mount = gfs2_mount,
419c93e0
SW
1396 .kill_sb = gfs2_kill_sb,
1397 .owner = THIS_MODULE,
1398};
7f78e035 1399MODULE_ALIAS_FS("gfs2");
419c93e0
SW
1400
1401struct file_system_type gfs2meta_fs_type = {
1402 .name = "gfs2meta",
1403 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1404 .mount = gfs2_mount_meta,
b3b94faa
DT
1405 .owner = THIS_MODULE,
1406};
7f78e035 1407MODULE_ALIAS_FS("gfs2meta");