]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/gfs2/ops_fstype.c
GFS2: Stop block extents at the end of bitmaps
[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>
b3b94faa
DT
22
23#include "gfs2.h"
5c676f6d 24#include "incore.h"
da6dd40d 25#include "bmap.h"
b3b94faa
DT
26#include "glock.h"
27#include "glops.h"
28#include "inode.h"
b3b94faa
DT
29#include "recovery.h"
30#include "rgrp.h"
31#include "super.h"
b3b94faa 32#include "sys.h"
5c676f6d 33#include "util.h"
bb3b0e3d 34#include "log.h"
9ac1b4d9 35#include "quota.h"
b5289681 36#include "dir.h"
63997775 37#include "trace_gfs2.h"
b3b94faa
DT
38
39#define DO 0
40#define UNDO 1
41
9b8df98f
SW
42/**
43 * gfs2_tune_init - Fill a gfs2_tune structure with default values
44 * @gt: tune
45 *
46 */
47
48static void gfs2_tune_init(struct gfs2_tune *gt)
49{
50 spin_lock_init(&gt->gt_spin);
51
9b8df98f
SW
52 gt->gt_quota_simul_sync = 64;
53 gt->gt_quota_warn_period = 10;
54 gt->gt_quota_scale_num = 1;
55 gt->gt_quota_scale_den = 1;
9b8df98f
SW
56 gt->gt_new_files_jdata = 0;
57 gt->gt_max_readahead = 1 << 18;
9b8df98f 58 gt->gt_complain_secs = 10;
9b8df98f
SW
59}
60
b3b94faa
DT
61static struct gfs2_sbd *init_sbd(struct super_block *sb)
62{
63 struct gfs2_sbd *sdp;
b3b94faa 64
85d1da67 65 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
66 if (!sdp)
67 return NULL;
68
5c676f6d 69 sb->s_fs_info = sdp;
b3b94faa 70 sdp->sd_vfs = sb;
a245769f
SW
71 sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
72 if (!sdp->sd_lkstats) {
73 kfree(sdp);
74 return NULL;
75 }
76
ba6e9364 77 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
78 gfs2_tune_init(&sdp->sd_tune);
79
e402746a
SW
80 init_waitqueue_head(&sdp->sd_glock_wait);
81 atomic_set(&sdp->sd_glock_disposal, 0);
3942ae53 82 init_completion(&sdp->sd_locking_init);
b3b94faa 83 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
84
85 spin_lock_init(&sdp->sd_rindex_spin);
7c9ca621 86 sdp->sd_rindex_tree.rb_node = NULL;
b3b94faa
DT
87
88 INIT_LIST_HEAD(&sdp->sd_jindex_list);
89 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 90 mutex_init(&sdp->sd_jindex_mutex);
b3b94faa 91
b3b94faa 92 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 93 mutex_init(&sdp->sd_quota_mutex);
37b2c837 94 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
95 INIT_LIST_HEAD(&sdp->sd_trunc_list);
96 spin_lock_init(&sdp->sd_trunc_lock);
b3b94faa
DT
97
98 spin_lock_init(&sdp->sd_log_lock);
5e687eac 99 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa
DT
100 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
101 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
b3b94faa 102 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
d7b616e2 103 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
b3b94faa 104
5e687eac
BM
105 init_waitqueue_head(&sdp->sd_log_waitq);
106 init_waitqueue_head(&sdp->sd_logd_waitq);
d6a079e8 107 spin_lock_init(&sdp->sd_ail_lock);
b3b94faa
DT
108 INIT_LIST_HEAD(&sdp->sd_ail1_list);
109 INIT_LIST_HEAD(&sdp->sd_ail2_list);
110
484adff8 111 init_rwsem(&sdp->sd_log_flush_lock);
16615be1
SW
112 atomic_set(&sdp->sd_log_in_flight, 0);
113 init_waitqueue_head(&sdp->sd_log_flush_wait);
b3b94faa
DT
114
115 INIT_LIST_HEAD(&sdp->sd_revoke_list);
116
f55ab26a 117 mutex_init(&sdp->sd_freeze_lock);
b3b94faa
DT
118
119 return sdp;
120}
121
b3b94faa 122
9b8df98f
SW
123/**
124 * gfs2_check_sb - Check superblock
125 * @sdp: the filesystem
126 * @sb: The superblock
127 * @silent: Don't print a message if the check fails
128 *
129 * Checks the version code of the FS is one that we understand how to
130 * read and that the sizes of the various on-disk structures have not
131 * changed.
132 */
133
32e471ef 134static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 135{
32e471ef
SW
136 struct gfs2_sb_host *sb = &sdp->sd_sb;
137
9b8df98f
SW
138 if (sb->sb_magic != GFS2_MAGIC ||
139 sb->sb_type != GFS2_METATYPE_SB) {
140 if (!silent)
141 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
142 return -EINVAL;
143 }
144
145 /* If format numbers match exactly, we're done. */
146
147 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
148 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
149 return 0;
150
c80dbb58 151 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 152
c80dbb58 153 return -EINVAL;
9b8df98f
SW
154}
155
156static void end_bio_io_page(struct bio *bio, int error)
157{
158 struct page *page = bio->bi_private;
159
160 if (!error)
161 SetPageUptodate(page);
162 else
163 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
164 unlock_page(page);
165}
166
32e471ef 167static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 168{
32e471ef
SW
169 struct gfs2_sb_host *sb = &sdp->sd_sb;
170 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
171 const struct gfs2_sb *str = buf;
172
173 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
174 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
175 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
176 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
177 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
178 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
179 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
180 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
181 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
182 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
183 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
184
185 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
186 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 187 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
188}
189
190/**
191 * gfs2_read_super - Read the gfs2 super block from disk
192 * @sdp: The GFS2 super block
193 * @sector: The location of the super block
194 * @error: The error code to return
195 *
196 * This uses the bio functions to read the super block from disk
197 * because we want to be 100% sure that we never read cached data.
198 * A super block is read twice only during each GFS2 mount and is
199 * never written to by the filesystem. The first time its read no
200 * locks are held, and the only details which are looked at are those
201 * relating to the locking protocol. Once locking is up and working,
202 * the sb is read again under the lock to establish the location of
203 * the master directory (contains pointers to journals etc) and the
204 * root directory.
205 *
206 * Returns: 0 on success or error
207 */
208
32e471ef 209static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
210{
211 struct super_block *sb = sdp->sd_vfs;
212 struct gfs2_sb *p;
213 struct page *page;
214 struct bio *bio;
215
216 page = alloc_page(GFP_NOFS);
217 if (unlikely(!page))
218 return -ENOBUFS;
219
220 ClearPageUptodate(page);
221 ClearPageDirty(page);
222 lock_page(page);
223
224 bio = bio_alloc(GFP_NOFS, 1);
9b8df98f
SW
225 bio->bi_sector = sector * (sb->s_blocksize >> 9);
226 bio->bi_bdev = sb->s_bdev;
227 bio_add_page(bio, page, PAGE_SIZE, 0);
228
229 bio->bi_end_io = end_bio_io_page;
230 bio->bi_private = page;
20ed0535 231 submit_bio(READ_SYNC | REQ_META, bio);
9b8df98f
SW
232 wait_on_page_locked(page);
233 bio_put(bio);
234 if (!PageUptodate(page)) {
235 __free_page(page);
236 return -EIO;
237 }
238 p = kmap(page);
32e471ef 239 gfs2_sb_in(sdp, p);
9b8df98f
SW
240 kunmap(page);
241 __free_page(page);
32e471ef 242 return gfs2_check_sb(sdp, silent);
9b8df98f 243}
6bac243f 244
9b8df98f
SW
245/**
246 * gfs2_read_sb - Read super block
247 * @sdp: The GFS2 superblock
9b8df98f
SW
248 * @silent: Don't print message if mount fails
249 *
250 */
251
6bac243f 252static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
253{
254 u32 hash_blocks, ind_blocks, leaf_blocks;
255 u32 tmp_blocks;
256 unsigned int x;
257 int error;
258
32e471ef 259 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
260 if (error) {
261 if (!silent)
262 fs_err(sdp, "can't read superblock\n");
263 return error;
264 }
265
9b8df98f
SW
266 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
267 GFS2_BASIC_BLOCK_SHIFT;
268 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
269 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
270 sizeof(struct gfs2_dinode)) / sizeof(u64);
271 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
272 sizeof(struct gfs2_meta_header)) / sizeof(u64);
273 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
274 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
275 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
276 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
277 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
278 sizeof(struct gfs2_meta_header)) /
279 sizeof(struct gfs2_quota_change);
280
281 /* Compute maximum reservation required to add a entry to a directory */
282
283 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
284 sdp->sd_jbsize);
285
286 ind_blocks = 0;
287 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
288 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
289 ind_blocks += tmp_blocks;
290 }
291
292 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
293
294 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
295
296 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
297 sizeof(struct gfs2_dinode);
298 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
299 for (x = 2;; x++) {
300 u64 space, d;
301 u32 m;
302
303 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
304 d = space;
305 m = do_div(d, sdp->sd_inptrs);
306
307 if (d != sdp->sd_heightsize[x - 1] || m)
308 break;
309 sdp->sd_heightsize[x] = space;
310 }
311 sdp->sd_max_height = x;
312 sdp->sd_heightsize[x] = ~0;
313 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
314
315 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
316 sizeof(struct gfs2_dinode);
317 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
318 for (x = 2;; x++) {
319 u64 space, d;
320 u32 m;
321
322 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
323 d = space;
324 m = do_div(d, sdp->sd_inptrs);
325
326 if (d != sdp->sd_jheightsize[x - 1] || m)
327 break;
328 sdp->sd_jheightsize[x] = space;
329 }
330 sdp->sd_max_jheight = x;
331 sdp->sd_jheightsize[x] = ~0;
332 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
333
334 return 0;
335}
336
b3b94faa
DT
337static int init_names(struct gfs2_sbd *sdp, int silent)
338{
b3b94faa
DT
339 char *proto, *table;
340 int error = 0;
341
342 proto = sdp->sd_args.ar_lockproto;
343 table = sdp->sd_args.ar_locktable;
344
345 /* Try to autodetect */
346
347 if (!proto[0] || !table[0]) {
32e471ef 348 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
349 if (error)
350 return error;
3cf1e7be 351
b3b94faa 352 if (!proto[0])
3cf1e7be 353 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 354 if (!table[0])
3cf1e7be 355 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
356 }
357
358 if (!table[0])
359 table = sdp->sd_vfs->s_id;
360
00377d8e
JD
361 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
362 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 363
5d35e31f
DC
364 table = sdp->sd_table_name;
365 while ((table = strchr(table, '/')))
b35997d4
RP
366 *table = '_';
367
b3b94faa
DT
368 return error;
369}
370
371static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
372 int undo)
373{
b3b94faa
DT
374 int error = 0;
375
376 if (undo)
377 goto fail_trans;
378
b3b94faa
DT
379 error = gfs2_glock_nq_num(sdp,
380 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
381 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
382 mount_gh);
383 if (error) {
384 fs_err(sdp, "can't acquire mount glock: %d\n", error);
385 goto fail;
386 }
387
388 error = gfs2_glock_nq_num(sdp,
389 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
390 LM_ST_SHARED,
579b78a4 391 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
392 &sdp->sd_live_gh);
393 if (error) {
394 fs_err(sdp, "can't acquire live glock: %d\n", error);
395 goto fail_mount;
396 }
397
398 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
399 CREATE, &sdp->sd_rename_gl);
400 if (error) {
401 fs_err(sdp, "can't create rename glock: %d\n", error);
402 goto fail_live;
403 }
404
405 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
406 CREATE, &sdp->sd_trans_gl);
407 if (error) {
408 fs_err(sdp, "can't create transaction glock: %d\n", error);
409 goto fail_rename;
410 }
b3b94faa
DT
411
412 return 0;
413
feaa7bba 414fail_trans:
b3b94faa 415 gfs2_glock_put(sdp->sd_trans_gl);
feaa7bba 416fail_rename:
b3b94faa 417 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 418fail_live:
b3b94faa 419 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 420fail_mount:
b3b94faa 421 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 422fail:
b3b94faa
DT
423 return error;
424}
425
9b8df98f
SW
426static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
427 u64 no_addr, const char *name)
f42faf4f 428{
9b8df98f
SW
429 struct gfs2_sbd *sdp = sb->s_fs_info;
430 struct dentry *dentry;
431 struct inode *inode;
432
44ad37d6 433 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
9b8df98f
SW
434 if (IS_ERR(inode)) {
435 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
436 return PTR_ERR(inode);
437 }
48fde701 438 dentry = d_make_root(inode);
9b8df98f
SW
439 if (!dentry) {
440 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
441 return -ENOMEM;
442 }
9b8df98f
SW
443 *dptr = dentry;
444 return 0;
f42faf4f
SW
445}
446
9b8df98f 447static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
448{
449 struct super_block *sb = sdp->sd_vfs;
450 struct gfs2_holder sb_gh;
dbb7cae2 451 u64 no_addr;
9b8df98f 452 int ret;
b3b94faa 453
9b8df98f
SW
454 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
455 LM_ST_SHARED, 0, &sb_gh);
456 if (ret) {
457 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
458 return ret;
b3b94faa 459 }
907b9bce 460
6bac243f 461 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
462 if (ret) {
463 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
464 goto out;
465 }
466
467 /* Set up the buffer cache and SB for real */
e1defc4f 468 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 469 ret = -EINVAL;
b3b94faa
DT
470 fs_err(sdp, "FS block size (%u) is too small for device "
471 "block size (%u)\n",
e1defc4f 472 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
473 goto out;
474 }
475 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 476 ret = -EINVAL;
b3b94faa
DT
477 fs_err(sdp, "FS block size (%u) is too big for machine "
478 "page size (%u)\n",
479 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
480 goto out;
481 }
b3b94faa
DT
482 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
483
f42faf4f 484 /* Get the root inode */
dbb7cae2 485 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
486 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
487 if (ret)
f42faf4f 488 goto out;
b3b94faa 489
9b8df98f
SW
490 /* Get the master inode */
491 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
492 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
493 if (ret) {
494 dput(sdp->sd_root_dir);
495 goto out;
496 }
497 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 498out:
b3b94faa 499 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 500 return ret;
b3b94faa
DT
501}
502
da6dd40d
BP
503/**
504 * map_journal_extents - create a reusable "extent" mapping from all logical
505 * blocks to all physical blocks for the given journal. This will save
506 * us time when writing journal blocks. Most journals will have only one
507 * extent that maps all their logical blocks. That's because gfs2.mkfs
508 * arranges the journal blocks sequentially to maximize performance.
509 * So the extent would map the first block for the entire file length.
510 * However, gfs2_jadd can happen while file activity is happening, so
511 * those journals may not be sequential. Less likely is the case where
512 * the users created their own journals by mounting the metafs and
513 * laying it out. But it's still possible. These journals might have
514 * several extents.
515 *
516 * TODO: This should be done in bigger chunks rather than one block at a time,
517 * but since it's only done at mount time, I'm not worried about the
518 * time it takes.
519 */
520static int map_journal_extents(struct gfs2_sbd *sdp)
521{
522 struct gfs2_jdesc *jd = sdp->sd_jdesc;
523 unsigned int lb;
524 u64 db, prev_db; /* logical block, disk block, prev disk block */
525 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
526 struct gfs2_journal_extent *jext = NULL;
527 struct buffer_head bh;
528 int rc = 0;
529
da6dd40d
BP
530 prev_db = 0;
531
a2e0f799 532 for (lb = 0; lb < i_size_read(jd->jd_inode) >> sdp->sd_sb.sb_bsize_shift; lb++) {
da6dd40d
BP
533 bh.b_state = 0;
534 bh.b_blocknr = 0;
535 bh.b_size = 1 << ip->i_inode.i_blkbits;
536 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
537 db = bh.b_blocknr;
538 if (rc || !db) {
539 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
540 "%u db=%llu\n", rc, lb, (unsigned long long)db);
541 break;
542 }
543 if (!prev_db || db != prev_db + 1) {
544 jext = kzalloc(sizeof(struct gfs2_journal_extent),
545 GFP_KERNEL);
546 if (!jext) {
547 printk(KERN_INFO "GFS2 error: out of memory "
548 "mapping journal extents.\n");
549 rc = -ENOMEM;
550 break;
551 }
552 jext->dblock = db;
553 jext->lblock = lb;
554 jext->blocks = 1;
555 list_add_tail(&jext->extent_list, &jd->extent_list);
556 } else {
557 jext->blocks++;
558 }
559 prev_db = db;
560 }
561 return rc;
562}
563
f057f6cd 564static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 565{
f057f6cd
SW
566 char *message = "FIRSTMOUNT=Done";
567 char *envp[] = { message, NULL };
e0c2a9aa
DT
568
569 fs_info(sdp, "first mount done, others may mount\n");
570
571 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
572 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
573
f057f6cd 574 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
575}
576
b5289681
SW
577/**
578 * gfs2_jindex_hold - Grab a lock on the jindex
579 * @sdp: The GFS2 superblock
580 * @ji_gh: the holder for the jindex glock
581 *
582 * Returns: errno
583 */
584
585static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
586{
587 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
588 struct qstr name;
589 char buf[20];
590 struct gfs2_jdesc *jd;
591 int error;
592
593 name.name = buf;
594
595 mutex_lock(&sdp->sd_jindex_mutex);
596
597 for (;;) {
598 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
599 if (error)
600 break;
601
602 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
603 name.hash = gfs2_disk_hash(name.name, name.len);
604
605 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
606 if (error == -ENOENT) {
607 error = 0;
608 break;
609 }
610
611 gfs2_glock_dq_uninit(ji_gh);
612
613 if (error)
614 break;
615
616 error = -ENOMEM;
617 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
618 if (!jd)
619 break;
620
621 INIT_LIST_HEAD(&jd->extent_list);
6ecd7c2d 622 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
623 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
624 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
625 if (!jd->jd_inode)
626 error = -ENOENT;
627 else
628 error = PTR_ERR(jd->jd_inode);
629 kfree(jd);
630 break;
631 }
632
633 spin_lock(&sdp->sd_jindex_spin);
634 jd->jd_jid = sdp->sd_journals++;
635 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
636 spin_unlock(&sdp->sd_jindex_spin);
637 }
638
639 mutex_unlock(&sdp->sd_jindex_mutex);
640
641 return error;
642}
643
b3b94faa
DT
644static int init_journal(struct gfs2_sbd *sdp, int undo)
645{
9b8df98f 646 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 647 struct gfs2_holder ji_gh;
5c676f6d 648 struct gfs2_inode *ip;
b3b94faa
DT
649 int jindex = 1;
650 int error = 0;
651
652 if (undo) {
653 jindex = 0;
fe64d517 654 goto fail_jinode_gh;
b3b94faa
DT
655 }
656
9b8df98f 657 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 658 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 659 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 660 return PTR_ERR(sdp->sd_jindex);
b3b94faa 661 }
b3b94faa
DT
662
663 /* Load in the journal index special file */
664
665 error = gfs2_jindex_hold(sdp, &ji_gh);
666 if (error) {
667 fs_err(sdp, "can't read journal index: %d\n", error);
668 goto fail;
669 }
670
0e5a9fb0 671 error = -EUSERS;
b3b94faa
DT
672 if (!gfs2_jindex_size(sdp)) {
673 fs_err(sdp, "no journals!\n");
907b9bce 674 goto fail_jindex;
b3b94faa
DT
675 }
676
677 if (sdp->sd_args.ar_spectator) {
678 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 679 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
680 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
681 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
682 } else {
683 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
684 fs_err(sdp, "can't mount journal #%u\n",
685 sdp->sd_lockstruct.ls_jid);
686 fs_err(sdp, "there are only %u journals (0 - %u)\n",
687 gfs2_jindex_size(sdp),
688 gfs2_jindex_size(sdp) - 1);
689 goto fail_jindex;
690 }
691 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
692
feaa7bba 693 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
694 &gfs2_journal_glops,
695 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
696 &sdp->sd_journal_gh);
697 if (error) {
698 fs_err(sdp, "can't acquire journal glock: %d\n", error);
699 goto fail_jindex;
700 }
701
feaa7bba
SW
702 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
703 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 704 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
705 &sdp->sd_jinode_gh);
706 if (error) {
707 fs_err(sdp, "can't acquire journal inode glock: %d\n",
708 error);
709 goto fail_journal_gh;
710 }
711
712 error = gfs2_jdesc_check(sdp->sd_jdesc);
713 if (error) {
714 fs_err(sdp, "my journal (%u) is bad: %d\n",
715 sdp->sd_jdesc->jd_jid, error);
716 goto fail_jinode_gh;
717 }
fd041f0b 718 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
719 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
720 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
721
722 /* Map the extents for this journal's blocks */
723 map_journal_extents(sdp);
b3b94faa 724 }
63997775 725 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
726
727 if (sdp->sd_lockstruct.ls_first) {
728 unsigned int x;
729 for (x = 0; x < sdp->sd_journals; x++) {
6ecd7c2d
TH
730 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
731 true);
b3b94faa
DT
732 if (error) {
733 fs_err(sdp, "error recovering journal %u: %d\n",
734 x, error);
735 goto fail_jinode_gh;
736 }
737 }
738
f057f6cd 739 gfs2_others_may_mount(sdp);
b3b94faa 740 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 741 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
742 if (error) {
743 fs_err(sdp, "error recovering my journal: %d\n", error);
744 goto fail_jinode_gh;
745 }
746 }
747
748 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
749 gfs2_glock_dq_uninit(&ji_gh);
750 jindex = 0;
751
b3b94faa
DT
752 return 0;
753
a91ea69f 754fail_jinode_gh:
b3b94faa
DT
755 if (!sdp->sd_args.ar_spectator)
756 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 757fail_journal_gh:
b3b94faa
DT
758 if (!sdp->sd_args.ar_spectator)
759 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 760fail_jindex:
b3b94faa
DT
761 gfs2_jindex_free(sdp);
762 if (jindex)
763 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 764fail:
f42faf4f 765 iput(sdp->sd_jindex);
b3b94faa
DT
766 return error;
767}
768
b3b94faa
DT
769
770static int init_inodes(struct gfs2_sbd *sdp, int undo)
771{
b3b94faa 772 int error = 0;
9b8df98f 773 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
774
775 if (undo)
f42faf4f
SW
776 goto fail_qinode;
777
f42faf4f
SW
778 error = init_journal(sdp, undo);
779 if (error)
9b8df98f 780 goto fail;
b3b94faa 781
b3b94faa 782 /* Read in the master statfs inode */
9b8df98f 783 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
784 if (IS_ERR(sdp->sd_statfs_inode)) {
785 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 786 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 787 goto fail_journal;
b3b94faa
DT
788 }
789
790 /* Read in the resource index inode */
9b8df98f 791 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
792 if (IS_ERR(sdp->sd_rindex)) {
793 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
794 fs_err(sdp, "can't get resource index inode: %d\n", error);
795 goto fail_statfs;
796 }
cf45b752 797 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
798
799 /* Read in the quota inode */
9b8df98f 800 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
801 if (IS_ERR(sdp->sd_quota_inode)) {
802 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
803 fs_err(sdp, "can't get quota file inode: %d\n", error);
804 goto fail_rindex;
805 }
a365fbf3
SW
806
807 error = gfs2_rindex_update(sdp);
808 if (error)
809 goto fail_qinode;
810
b3b94faa
DT
811 return 0;
812
f42faf4f
SW
813fail_qinode:
814 iput(sdp->sd_quota_inode);
f42faf4f 815fail_rindex:
b3b94faa 816 gfs2_clear_rgrpd(sdp);
f42faf4f 817 iput(sdp->sd_rindex);
f42faf4f
SW
818fail_statfs:
819 iput(sdp->sd_statfs_inode);
f42faf4f
SW
820fail_journal:
821 init_journal(sdp, UNDO);
f42faf4f 822fail:
b3b94faa
DT
823 return error;
824}
825
826static int init_per_node(struct gfs2_sbd *sdp, int undo)
827{
f42faf4f 828 struct inode *pn = NULL;
b3b94faa
DT
829 char buf[30];
830 int error = 0;
5c676f6d 831 struct gfs2_inode *ip;
9b8df98f 832 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
833
834 if (sdp->sd_args.ar_spectator)
835 return 0;
836
837 if (undo)
838 goto fail_qc_gh;
839
9b8df98f 840 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
841 if (IS_ERR(pn)) {
842 error = PTR_ERR(pn);
b3b94faa
DT
843 fs_err(sdp, "can't find per_node directory: %d\n", error);
844 return error;
845 }
846
b3b94faa 847 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
848 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
849 if (IS_ERR(sdp->sd_sc_inode)) {
850 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 851 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 852 goto fail;
b3b94faa
DT
853 }
854
b3b94faa 855 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
856 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
857 if (IS_ERR(sdp->sd_qc_inode)) {
858 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
859 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
860 goto fail_ut_i;
861 }
862
f42faf4f 863 iput(pn);
b3b94faa
DT
864 pn = NULL;
865
feaa7bba 866 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 867 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
868 &sdp->sd_sc_gh);
869 if (error) {
870 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 871 goto fail_qc_i;
b3b94faa
DT
872 }
873
feaa7bba 874 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 875 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
876 &sdp->sd_qc_gh);
877 if (error) {
878 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
879 goto fail_ut_gh;
880 }
881
882 return 0;
883
a91ea69f 884fail_qc_gh:
b3b94faa 885 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 886fail_ut_gh:
b3b94faa 887 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 888fail_qc_i:
f42faf4f 889 iput(sdp->sd_qc_inode);
a91ea69f 890fail_ut_i:
f42faf4f 891 iput(sdp->sd_sc_inode);
a91ea69f 892fail:
b3b94faa 893 if (pn)
f42faf4f 894 iput(pn);
b3b94faa
DT
895 return error;
896}
897
898static int init_threads(struct gfs2_sbd *sdp, int undo)
899{
900 struct task_struct *p;
901 int error = 0;
902
903 if (undo)
feaa7bba 904 goto fail_quotad;
b3b94faa 905
b3b94faa
DT
906 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
907 error = IS_ERR(p);
908 if (error) {
909 fs_err(sdp, "can't start logd thread: %d\n", error);
910 return error;
911 }
912 sdp->sd_logd_process = p;
913
b3b94faa
DT
914 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
915 error = IS_ERR(p);
916 if (error) {
917 fs_err(sdp, "can't start quotad thread: %d\n", error);
918 goto fail;
919 }
920 sdp->sd_quotad_process = p;
921
b3b94faa
DT
922 return 0;
923
b3b94faa 924
feaa7bba 925fail_quotad:
b3b94faa 926 kthread_stop(sdp->sd_quotad_process);
feaa7bba 927fail:
b3b94faa 928 kthread_stop(sdp->sd_logd_process);
b3b94faa
DT
929 return error;
930}
da755fdb 931
f057f6cd
SW
932static const match_table_t nolock_tokens = {
933 { Opt_jid, "jid=%d\n", },
934 { Opt_err, NULL },
935};
936
937static const struct lm_lockops nolock_ops = {
938 .lm_proto_name = "lock_nolock",
fc0e38da 939 .lm_put_lock = gfs2_glock_free,
f057f6cd
SW
940 .lm_tokens = &nolock_tokens,
941};
942
da755fdb
SW
943/**
944 * gfs2_lm_mount - mount a locking protocol
945 * @sdp: the filesystem
3ad2f3fb 946 * @args: mount arguments
da755fdb
SW
947 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
948 *
949 * Returns: errno
950 */
951
952static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
953{
f057f6cd
SW
954 const struct lm_lockops *lm;
955 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
956 struct gfs2_args *args = &sdp->sd_args;
957 const char *proto = sdp->sd_proto_name;
958 const char *table = sdp->sd_table_name;
f057f6cd
SW
959 char *o, *options;
960 int ret;
da755fdb 961
f057f6cd
SW
962 if (!strcmp("lock_nolock", proto)) {
963 lm = &nolock_ops;
964 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
965#ifdef CONFIG_GFS2_FS_LOCKING_DLM
966 } else if (!strcmp("lock_dlm", proto)) {
967 lm = &gfs2_dlm_ops;
968#endif
969 } else {
970 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
971 return -ENOENT;
972 }
da755fdb
SW
973
974 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
975
f057f6cd
SW
976 ls->ls_ops = lm;
977 ls->ls_first = 1;
da755fdb 978
f057f6cd
SW
979 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
980 substring_t tmp[MAX_OPT_ARGS];
981 int token, option;
982
983 if (!o || !*o)
984 continue;
985
986 token = match_token(o, *lm->lm_tokens, tmp);
987 switch (token) {
988 case Opt_jid:
989 ret = match_int(&tmp[0], &option);
990 if (ret || option < 0)
991 goto hostdata_error;
ba6e9364
SW
992 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
993 ls->ls_jid = option;
f057f6cd
SW
994 break;
995 case Opt_id:
4875647a 996 case Opt_nodir:
2b88f7c5 997 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
998 break;
999 case Opt_first:
1000 ret = match_int(&tmp[0], &option);
1001 if (ret || (option != 0 && option != 1))
1002 goto hostdata_error;
1003 ls->ls_first = option;
1004 break;
f057f6cd
SW
1005 case Opt_err:
1006 default:
1007hostdata_error:
1008 fs_info(sdp, "unknown hostdata (%s)\n", o);
1009 return -EINVAL;
1010 }
da755fdb
SW
1011 }
1012
f057f6cd
SW
1013 if (lm->lm_mount == NULL) {
1014 fs_info(sdp, "Now mounting FS...\n");
19237039 1015 complete_all(&sdp->sd_locking_init);
f057f6cd 1016 return 0;
da755fdb 1017 }
e0c2a9aa 1018 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1019 if (ret == 0)
1020 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1021 complete_all(&sdp->sd_locking_init);
f057f6cd 1022 return ret;
da755fdb
SW
1023}
1024
1025void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1026{
f057f6cd
SW
1027 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1028 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1029 lm->lm_unmount)
1030 lm->lm_unmount(sdp);
da755fdb 1031}
b3b94faa 1032
ba6e9364
SW
1033static int gfs2_journalid_wait(void *word)
1034{
1035 if (signal_pending(current))
1036 return -EINTR;
1037 schedule();
1038 return 0;
1039}
1040
1041static int wait_on_journal(struct gfs2_sbd *sdp)
1042{
ba6e9364
SW
1043 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1044 return 0;
1045
1046 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1047}
1048
8633ecfa
SW
1049void gfs2_online_uevent(struct gfs2_sbd *sdp)
1050{
1051 struct super_block *sb = sdp->sd_vfs;
1052 char ro[20];
1053 char spectator[20];
1054 char *envp[] = { ro, spectator, NULL };
1055 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1056 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1057 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1058}
1059
b3b94faa
DT
1060/**
1061 * fill_super - Read in superblock
1062 * @sb: The VFS superblock
1063 * @data: Mount options
1064 * @silent: Don't complain if it's not a GFS2 filesystem
1065 *
1066 * Returns: errno
1067 */
1068
f55073ff 1069static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1070{
1071 struct gfs2_sbd *sdp;
1072 struct gfs2_holder mount_gh;
1073 int error;
1074
1075 sdp = init_sbd(sb);
1076 if (!sdp) {
d92a8d48 1077 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1078 return -ENOMEM;
1079 }
f55073ff 1080 sdp->sd_args = *args;
b3b94faa 1081
fe64d517 1082 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1083 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1084 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1085 }
6f04c1c7
SW
1086 if (sdp->sd_args.ar_posix_acl)
1087 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1088 if (sdp->sd_args.ar_nobarrier)
1089 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1090
9964afbb 1091 sb->s_flags |= MS_NOSEC;
719ee344
SW
1092 sb->s_magic = GFS2_MAGIC;
1093 sb->s_op = &gfs2_super_ops;
41ced6dc 1094 sb->s_d_op = &gfs2_dops;
719ee344 1095 sb->s_export_op = &gfs2_export_ops;
40b78a32 1096 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1097 sb->s_qcop = &gfs2_quotactl_ops;
1098 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1099 sb->s_time_gran = 1;
1100 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1101
1102 /* Set up the buffer cache and fill in some fake block size values
1103 to allow us to read-in the on-disk superblock. */
1104 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1105 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1106 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1107 GFS2_BASIC_BLOCK_SHIFT;
1108 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1109
5e687eac 1110 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1111 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1112 if (sdp->sd_args.ar_statfs_quantum) {
1113 sdp->sd_tune.gt_statfs_slow = 0;
1114 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1115 } else {
3d3c10f2
BM
1116 sdp->sd_tune.gt_statfs_slow = 1;
1117 sdp->sd_tune.gt_statfs_quantum = 30;
1118 }
48c2b613 1119
b3b94faa 1120 error = init_names(sdp, silent);
0d515210
BP
1121 if (error) {
1122 /* In this case, we haven't initialized sysfs, so we have to
1123 manually free the sdp. */
1124 free_percpu(sdp->sd_lkstats);
1125 kfree(sdp);
1126 sb->s_fs_info = NULL;
1127 return error;
1128 }
b3b94faa 1129
e0c2a9aa
DT
1130 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1131
b3b94faa 1132 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1133 /*
1134 * If we hit an error here, gfs2_sys_fs_add will have called function
1135 * kobject_put which causes the sysfs usage count to go to zero, which
1136 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1137 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1138 * kobject_put to free sdp.
1139 */
b3b94faa 1140 if (error)
0d515210
BP
1141 return error;
1142
1143 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1144
1145 error = gfs2_lm_mount(sdp, silent);
1146 if (error)
0d515210 1147 goto fail_debug;
b3b94faa
DT
1148
1149 error = init_locking(sdp, &mount_gh, DO);
1150 if (error)
1151 goto fail_lm;
1152
9b8df98f 1153 error = init_sb(sdp, silent);
b3b94faa
DT
1154 if (error)
1155 goto fail_locking;
b3b94faa 1156
ba6e9364
SW
1157 error = wait_on_journal(sdp);
1158 if (error)
1159 goto fail_sb;
1160
feb47ca9
SW
1161 /*
1162 * If user space has failed to join the cluster or some similar
1163 * failure has occurred, then the journal id will contain a
1164 * negative (error) number. This will then be returned to the
1165 * caller (of the mount syscall). We do this even for spectator
1166 * mounts (which just write a jid of 0 to indicate "ok" even though
1167 * the jid is unused in the spectator case)
1168 */
1169 if (sdp->sd_lockstruct.ls_jid < 0) {
1170 error = sdp->sd_lockstruct.ls_jid;
1171 sdp->sd_lockstruct.ls_jid = 0;
1172 goto fail_sb;
1173 }
1174
e0c2a9aa
DT
1175 if (sdp->sd_args.ar_spectator)
1176 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1177 sdp->sd_table_name);
1178 else
1179 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1180 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1181
b3b94faa
DT
1182 error = init_inodes(sdp, DO);
1183 if (error)
f42faf4f 1184 goto fail_sb;
b3b94faa
DT
1185
1186 error = init_per_node(sdp, DO);
1187 if (error)
1188 goto fail_inodes;
1189
1190 error = gfs2_statfs_init(sdp);
1191 if (error) {
1192 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1193 goto fail_per_node;
1194 }
1195
1196 error = init_threads(sdp, DO);
1197 if (error)
1198 goto fail_per_node;
1199
1200 if (!(sb->s_flags & MS_RDONLY)) {
1201 error = gfs2_make_fs_rw(sdp);
1202 if (error) {
1203 fs_err(sdp, "can't make FS RW: %d\n", error);
1204 goto fail_threads;
1205 }
1206 }
1207
1208 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1209 gfs2_online_uevent(sdp);
b3b94faa
DT
1210 return 0;
1211
a91ea69f 1212fail_threads:
b3b94faa 1213 init_threads(sdp, UNDO);
a91ea69f 1214fail_per_node:
b3b94faa 1215 init_per_node(sdp, UNDO);
a91ea69f 1216fail_inodes:
b3b94faa 1217 init_inodes(sdp, UNDO);
a91ea69f 1218fail_sb:
9b8df98f
SW
1219 if (sdp->sd_root_dir)
1220 dput(sdp->sd_root_dir);
1221 if (sdp->sd_master_dir)
1222 dput(sdp->sd_master_dir);
e7c8707e
SW
1223 if (sb->s_root)
1224 dput(sb->s_root);
9b8df98f 1225 sb->s_root = NULL;
a91ea69f 1226fail_locking:
b3b94faa 1227 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1228fail_lm:
fefc03bf 1229 gfs2_gl_hash_clear(sdp);
b3b94faa 1230 gfs2_lm_unmount(sdp);
0d515210 1231fail_debug:
5f882096 1232 gfs2_delete_debugfs_file(sdp);
a245769f 1233 free_percpu(sdp->sd_lkstats);
0d515210
BP
1234 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1235 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1236 gfs2_sys_fs_del(sdp);
5c676f6d 1237 sb->s_fs_info = NULL;
b3b94faa
DT
1238 return error;
1239}
1240
f55073ff 1241static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1242{
f55073ff
SW
1243 s->s_bdev = data;
1244 s->s_dev = s->s_bdev->bd_dev;
1245
1246 /*
1247 * We set the bdi here to the queue backing, file systems can
1248 * overwrite this in ->fill_super()
1249 */
1250 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1251 return 0;
86384605
AD
1252}
1253
f55073ff 1254static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1255{
1256 struct block_device *bdev = ptr;
1257 return (bdev == s->s_bdev);
1258}
1259
f55073ff 1260/**
8bcbbf00 1261 * gfs2_mount - Get the GFS2 superblock
f55073ff
SW
1262 * @fs_type: The GFS2 filesystem type
1263 * @flags: Mount flags
1264 * @dev_name: The name of the device
1265 * @data: The mount arguments
f55073ff
SW
1266 *
1267 * Q. Why not use get_sb_bdev() ?
1268 * A. We need to select one of two root directories to mount, independent
1269 * of whether this is the initial, or subsequent, mount of this sb
1270 *
1271 * Returns: 0 or -ve on error
1272 */
1273
8bcbbf00
AV
1274static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags,
1275 const char *dev_name, void *data)
f55073ff
SW
1276{
1277 struct block_device *bdev;
1278 struct super_block *s;
d4d77629 1279 fmode_t mode = FMODE_READ | FMODE_EXCL;
f55073ff
SW
1280 int error;
1281 struct gfs2_args args;
1282 struct gfs2_sbd *sdp;
1283
1284 if (!(flags & MS_RDONLY))
1285 mode |= FMODE_WRITE;
1286
d4d77629 1287 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
f55073ff 1288 if (IS_ERR(bdev))
8bcbbf00 1289 return ERR_CAST(bdev);
f55073ff
SW
1290
1291 /*
1292 * once the super is inserted into the list by sget, s_umount
1293 * will protect the lockfs code from trying to start a snapshot
1294 * while we are mounting
1295 */
1296 mutex_lock(&bdev->bd_fsfreeze_mutex);
1297 if (bdev->bd_fsfreeze_count > 0) {
1298 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1299 error = -EBUSY;
1300 goto error_bdev;
1301 }
9249e17f 1302 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1303 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1304 error = PTR_ERR(s);
1305 if (IS_ERR(s))
1306 goto error_bdev;
1307
8bcbbf00 1308 if (s->s_root)
d4d77629 1309 blkdev_put(bdev, mode);
8bcbbf00 1310
f55073ff
SW
1311 memset(&args, 0, sizeof(args));
1312 args.ar_quota = GFS2_QUOTA_DEFAULT;
1313 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1314 args.ar_commit = 30;
3d3c10f2
BM
1315 args.ar_statfs_quantum = 30;
1316 args.ar_quota_quantum = 60;
f55073ff
SW
1317 args.ar_errors = GFS2_ERRORS_DEFAULT;
1318
1319 error = gfs2_mount_args(&args, data);
1320 if (error) {
1321 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
8bcbbf00 1322 goto error_super;
f55073ff
SW
1323 }
1324
1325 if (s->s_root) {
1326 error = -EBUSY;
1327 if ((flags ^ s->s_flags) & MS_RDONLY)
1328 goto error_super;
f55073ff
SW
1329 } else {
1330 char b[BDEVNAME_SIZE];
1331
f55073ff
SW
1332 s->s_mode = mode;
1333 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1334 sb_set_blocksize(s, block_size(bdev));
1335 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1336 if (error)
1337 goto error_super;
f55073ff
SW
1338 s->s_flags |= MS_ACTIVE;
1339 bdev->bd_super = s;
1340 }
1341
1342 sdp = s->s_fs_info;
f55073ff 1343 if (args.ar_meta)
8bcbbf00 1344 return dget(sdp->sd_master_dir);
f55073ff 1345 else
8bcbbf00 1346 return dget(sdp->sd_root_dir);
f55073ff
SW
1347
1348error_super:
1349 deactivate_locked_super(s);
8bcbbf00 1350 return ERR_PTR(error);
f55073ff 1351error_bdev:
d4d77629 1352 blkdev_put(bdev, mode);
8bcbbf00 1353 return ERR_PTR(error);
f55073ff
SW
1354}
1355
f6d03139
SW
1356static int set_meta_super(struct super_block *s, void *ptr)
1357{
1358 return -EINVAL;
1359}
1360
8bcbbf00
AV
1361static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1362 int flags, const char *dev_name, void *data)
86384605 1363{
f6d03139 1364 struct super_block *s;
003dec89 1365 struct gfs2_sbd *sdp;
e24977d4 1366 struct path path;
86384605 1367 int error;
907b9bce 1368
e24977d4 1369 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1370 if (error) {
9b8df98f
SW
1371 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1372 dev_name, error);
8bcbbf00 1373 return ERR_PTR(error);
86384605 1374 }
9249e17f 1375 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
f6d03139 1376 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1377 path_put(&path);
f6d03139 1378 if (IS_ERR(s)) {
86384605 1379 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
8bcbbf00 1380 return ERR_CAST(s);
86384605 1381 }
f55073ff
SW
1382 if ((flags ^ s->s_flags) & MS_RDONLY) {
1383 deactivate_locked_super(s);
8bcbbf00 1384 return ERR_PTR(-EBUSY);
f55073ff 1385 }
f6d03139 1386 sdp = s->s_fs_info;
8bcbbf00 1387 return dget(sdp->sd_master_dir);
b3b94faa
DT
1388}
1389
419c93e0 1390static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1391{
1392 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1393
1394 if (sdp == NULL) {
1395 kill_block_super(sb);
1396 return;
3af165ac 1397 }
88a19ad0
SW
1398
1399 gfs2_meta_syncfs(sdp);
1400 dput(sdp->sd_root_dir);
1401 dput(sdp->sd_master_dir);
1402 sdp->sd_root_dir = NULL;
1403 sdp->sd_master_dir = NULL;
fefc03bf 1404 shrink_dcache_sb(sb);
88a19ad0 1405 gfs2_delete_debugfs_file(sdp);
a245769f 1406 free_percpu(sdp->sd_lkstats);
0d515210 1407 kill_block_super(sb);
86384605
AD
1408}
1409
b3b94faa
DT
1410struct file_system_type gfs2_fs_type = {
1411 .name = "gfs2",
1412 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1413 .mount = gfs2_mount,
419c93e0
SW
1414 .kill_sb = gfs2_kill_sb,
1415 .owner = THIS_MODULE,
1416};
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};
1424