]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/gfs2/ops_fstype.c
GFS2: Remove localcaching mount option
[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>
86384605
AD
17#include <linux/namei.h>
18#include <linux/mount.h>
5c676f6d 19#include <linux/gfs2_ondisk.h>
cc632e7f 20#include <linux/quotaops.h>
b3b94faa
DT
21
22#include "gfs2.h"
5c676f6d 23#include "incore.h"
da6dd40d 24#include "bmap.h"
b3b94faa
DT
25#include "glock.h"
26#include "glops.h"
27#include "inode.h"
b3b94faa
DT
28#include "recovery.h"
29#include "rgrp.h"
30#include "super.h"
b3b94faa 31#include "sys.h"
5c676f6d 32#include "util.h"
bb3b0e3d 33#include "log.h"
9ac1b4d9 34#include "quota.h"
b5289681 35#include "dir.h"
63997775 36#include "trace_gfs2.h"
b3b94faa
DT
37
38#define DO 0
39#define UNDO 1
40
9b8df98f
SW
41static const u32 gfs2_old_fs_formats[] = {
42 0
43};
44
45static const u32 gfs2_old_multihost_formats[] = {
46 0
47};
48
49/**
50 * gfs2_tune_init - Fill a gfs2_tune structure with default values
51 * @gt: tune
52 *
53 */
54
55static void gfs2_tune_init(struct gfs2_tune *gt)
56{
57 spin_lock_init(&gt->gt_spin);
58
9b8df98f
SW
59 gt->gt_quota_simul_sync = 64;
60 gt->gt_quota_warn_period = 10;
61 gt->gt_quota_scale_num = 1;
62 gt->gt_quota_scale_den = 1;
9b8df98f
SW
63 gt->gt_new_files_jdata = 0;
64 gt->gt_max_readahead = 1 << 18;
9b8df98f 65 gt->gt_complain_secs = 10;
9b8df98f
SW
66}
67
b3b94faa
DT
68static struct gfs2_sbd *init_sbd(struct super_block *sb)
69{
70 struct gfs2_sbd *sdp;
b3b94faa 71
85d1da67 72 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
73 if (!sdp)
74 return NULL;
75
5c676f6d 76 sb->s_fs_info = sdp;
b3b94faa 77 sdp->sd_vfs = sb;
ba6e9364 78 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
79 gfs2_tune_init(&sdp->sd_tune);
80
e402746a
SW
81 init_waitqueue_head(&sdp->sd_glock_wait);
82 atomic_set(&sdp->sd_glock_disposal, 0);
b3b94faa 83 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
84
85 spin_lock_init(&sdp->sd_rindex_spin);
f55ab26a 86 mutex_init(&sdp->sd_rindex_mutex);
b3b94faa
DT
87 INIT_LIST_HEAD(&sdp->sd_rindex_list);
88 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
b3b94faa
DT
89
90 INIT_LIST_HEAD(&sdp->sd_jindex_list);
91 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 92 mutex_init(&sdp->sd_jindex_mutex);
b3b94faa 93
b3b94faa 94 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 95 mutex_init(&sdp->sd_quota_mutex);
37b2c837 96 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
97 INIT_LIST_HEAD(&sdp->sd_trunc_list);
98 spin_lock_init(&sdp->sd_trunc_lock);
b3b94faa
DT
99
100 spin_lock_init(&sdp->sd_log_lock);
5e687eac 101 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa
DT
102 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
103 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
104 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
105 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
d7b616e2 106 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
b3b94faa 107
5e687eac
BM
108 init_waitqueue_head(&sdp->sd_log_waitq);
109 init_waitqueue_head(&sdp->sd_logd_waitq);
b3b94faa
DT
110 INIT_LIST_HEAD(&sdp->sd_ail1_list);
111 INIT_LIST_HEAD(&sdp->sd_ail2_list);
112
484adff8 113 init_rwsem(&sdp->sd_log_flush_lock);
16615be1
SW
114 atomic_set(&sdp->sd_log_in_flight, 0);
115 init_waitqueue_head(&sdp->sd_log_flush_wait);
b3b94faa
DT
116
117 INIT_LIST_HEAD(&sdp->sd_revoke_list);
118
f55ab26a 119 mutex_init(&sdp->sd_freeze_lock);
b3b94faa
DT
120
121 return sdp;
122}
123
b3b94faa 124
9b8df98f
SW
125/**
126 * gfs2_check_sb - Check superblock
127 * @sdp: the filesystem
128 * @sb: The superblock
129 * @silent: Don't print a message if the check fails
130 *
131 * Checks the version code of the FS is one that we understand how to
132 * read and that the sizes of the various on-disk structures have not
133 * changed.
134 */
135
136static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
137{
138 unsigned int x;
139
140 if (sb->sb_magic != GFS2_MAGIC ||
141 sb->sb_type != GFS2_METATYPE_SB) {
142 if (!silent)
143 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
144 return -EINVAL;
145 }
146
147 /* If format numbers match exactly, we're done. */
148
149 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
150 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
151 return 0;
152
153 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
154 for (x = 0; gfs2_old_fs_formats[x]; x++)
155 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
156 break;
157
158 if (!gfs2_old_fs_formats[x]) {
159 printk(KERN_WARNING
160 "GFS2: code version (%u, %u) is incompatible "
161 "with ondisk format (%u, %u)\n",
162 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
163 sb->sb_fs_format, sb->sb_multihost_format);
164 printk(KERN_WARNING
165 "GFS2: I don't know how to upgrade this FS\n");
166 return -EINVAL;
167 }
168 }
169
170 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
171 for (x = 0; gfs2_old_multihost_formats[x]; x++)
172 if (gfs2_old_multihost_formats[x] ==
173 sb->sb_multihost_format)
174 break;
175
176 if (!gfs2_old_multihost_formats[x]) {
177 printk(KERN_WARNING
178 "GFS2: code version (%u, %u) is incompatible "
179 "with ondisk format (%u, %u)\n",
180 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
181 sb->sb_fs_format, sb->sb_multihost_format);
182 printk(KERN_WARNING
183 "GFS2: I don't know how to upgrade this FS\n");
184 return -EINVAL;
185 }
186 }
187
188 if (!sdp->sd_args.ar_upgrade) {
189 printk(KERN_WARNING
190 "GFS2: code version (%u, %u) is incompatible "
191 "with ondisk format (%u, %u)\n",
192 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
193 sb->sb_fs_format, sb->sb_multihost_format);
194 printk(KERN_INFO
195 "GFS2: Use the \"upgrade\" mount option to upgrade "
196 "the FS\n");
197 printk(KERN_INFO "GFS2: See the manual for more details\n");
198 return -EINVAL;
199 }
200
201 return 0;
202}
203
204static void end_bio_io_page(struct bio *bio, int error)
205{
206 struct page *page = bio->bi_private;
207
208 if (!error)
209 SetPageUptodate(page);
210 else
211 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
212 unlock_page(page);
213}
214
215static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
216{
217 const struct gfs2_sb *str = buf;
218
219 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
220 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
221 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
222 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
223 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
224 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
225 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
226 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
227 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
228 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
229 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
230
231 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
232 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
02e3cc70 233 memcpy(sb->sb_uuid, str->sb_uuid, 16);
9b8df98f
SW
234}
235
236/**
237 * gfs2_read_super - Read the gfs2 super block from disk
238 * @sdp: The GFS2 super block
239 * @sector: The location of the super block
240 * @error: The error code to return
241 *
242 * This uses the bio functions to read the super block from disk
243 * because we want to be 100% sure that we never read cached data.
244 * A super block is read twice only during each GFS2 mount and is
245 * never written to by the filesystem. The first time its read no
246 * locks are held, and the only details which are looked at are those
247 * relating to the locking protocol. Once locking is up and working,
248 * the sb is read again under the lock to establish the location of
249 * the master directory (contains pointers to journals etc) and the
250 * root directory.
251 *
252 * Returns: 0 on success or error
253 */
254
255static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
256{
257 struct super_block *sb = sdp->sd_vfs;
258 struct gfs2_sb *p;
259 struct page *page;
260 struct bio *bio;
261
262 page = alloc_page(GFP_NOFS);
263 if (unlikely(!page))
264 return -ENOBUFS;
265
266 ClearPageUptodate(page);
267 ClearPageDirty(page);
268 lock_page(page);
269
270 bio = bio_alloc(GFP_NOFS, 1);
9b8df98f
SW
271 bio->bi_sector = sector * (sb->s_blocksize >> 9);
272 bio->bi_bdev = sb->s_bdev;
273 bio_add_page(bio, page, PAGE_SIZE, 0);
274
275 bio->bi_end_io = end_bio_io_page;
276 bio->bi_private = page;
7b6d91da 277 submit_bio(READ_SYNC | REQ_META, bio);
9b8df98f
SW
278 wait_on_page_locked(page);
279 bio_put(bio);
280 if (!PageUptodate(page)) {
281 __free_page(page);
282 return -EIO;
283 }
284 p = kmap(page);
285 gfs2_sb_in(&sdp->sd_sb, p);
286 kunmap(page);
287 __free_page(page);
288 return 0;
289}
6bac243f 290
9b8df98f
SW
291/**
292 * gfs2_read_sb - Read super block
293 * @sdp: The GFS2 superblock
9b8df98f
SW
294 * @silent: Don't print message if mount fails
295 *
296 */
297
6bac243f 298static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
299{
300 u32 hash_blocks, ind_blocks, leaf_blocks;
301 u32 tmp_blocks;
302 unsigned int x;
303 int error;
304
305 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
306 if (error) {
307 if (!silent)
308 fs_err(sdp, "can't read superblock\n");
309 return error;
310 }
311
312 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
313 if (error)
314 return error;
315
316 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
317 GFS2_BASIC_BLOCK_SHIFT;
318 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
319 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
320 sizeof(struct gfs2_dinode)) / sizeof(u64);
321 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
322 sizeof(struct gfs2_meta_header)) / sizeof(u64);
323 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
324 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
325 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
326 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
327 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
328 sizeof(struct gfs2_meta_header)) /
329 sizeof(struct gfs2_quota_change);
330
331 /* Compute maximum reservation required to add a entry to a directory */
332
333 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
334 sdp->sd_jbsize);
335
336 ind_blocks = 0;
337 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
338 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
339 ind_blocks += tmp_blocks;
340 }
341
342 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
343
344 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
345
346 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
347 sizeof(struct gfs2_dinode);
348 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
349 for (x = 2;; x++) {
350 u64 space, d;
351 u32 m;
352
353 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
354 d = space;
355 m = do_div(d, sdp->sd_inptrs);
356
357 if (d != sdp->sd_heightsize[x - 1] || m)
358 break;
359 sdp->sd_heightsize[x] = space;
360 }
361 sdp->sd_max_height = x;
362 sdp->sd_heightsize[x] = ~0;
363 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
364
365 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
366 sizeof(struct gfs2_dinode);
367 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
368 for (x = 2;; x++) {
369 u64 space, d;
370 u32 m;
371
372 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
373 d = space;
374 m = do_div(d, sdp->sd_inptrs);
375
376 if (d != sdp->sd_jheightsize[x - 1] || m)
377 break;
378 sdp->sd_jheightsize[x] = space;
379 }
380 sdp->sd_max_jheight = x;
381 sdp->sd_jheightsize[x] = ~0;
382 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
383
384 return 0;
385}
386
b3b94faa
DT
387static int init_names(struct gfs2_sbd *sdp, int silent)
388{
b3b94faa
DT
389 char *proto, *table;
390 int error = 0;
391
392 proto = sdp->sd_args.ar_lockproto;
393 table = sdp->sd_args.ar_locktable;
394
395 /* Try to autodetect */
396
397 if (!proto[0] || !table[0]) {
bb8d8a6f
SW
398 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
399 if (error)
400 return error;
3cf1e7be
SW
401
402 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
b3b94faa
DT
403 if (error)
404 goto out;
405
406 if (!proto[0])
3cf1e7be 407 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 408 if (!table[0])
3cf1e7be 409 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
410 }
411
412 if (!table[0])
413 table = sdp->sd_vfs->s_id;
414
00377d8e
JD
415 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
416 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 417
5d35e31f
DC
418 table = sdp->sd_table_name;
419 while ((table = strchr(table, '/')))
b35997d4
RP
420 *table = '_';
421
a91ea69f 422out:
b3b94faa
DT
423 return error;
424}
425
426static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
427 int undo)
428{
b3b94faa
DT
429 int error = 0;
430
431 if (undo)
432 goto fail_trans;
433
b3b94faa
DT
434 error = gfs2_glock_nq_num(sdp,
435 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
436 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
437 mount_gh);
438 if (error) {
439 fs_err(sdp, "can't acquire mount glock: %d\n", error);
440 goto fail;
441 }
442
443 error = gfs2_glock_nq_num(sdp,
444 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
445 LM_ST_SHARED,
579b78a4 446 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
447 &sdp->sd_live_gh);
448 if (error) {
449 fs_err(sdp, "can't acquire live glock: %d\n", error);
450 goto fail_mount;
451 }
452
453 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
454 CREATE, &sdp->sd_rename_gl);
455 if (error) {
456 fs_err(sdp, "can't create rename glock: %d\n", error);
457 goto fail_live;
458 }
459
460 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
461 CREATE, &sdp->sd_trans_gl);
462 if (error) {
463 fs_err(sdp, "can't create transaction glock: %d\n", error);
464 goto fail_rename;
465 }
b3b94faa
DT
466
467 return 0;
468
feaa7bba 469fail_trans:
b3b94faa 470 gfs2_glock_put(sdp->sd_trans_gl);
feaa7bba 471fail_rename:
b3b94faa 472 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 473fail_live:
b3b94faa 474 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 475fail_mount:
b3b94faa 476 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 477fail:
b3b94faa
DT
478 return error;
479}
480
9b8df98f
SW
481static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
482 u64 no_addr, const char *name)
f42faf4f 483{
9b8df98f
SW
484 struct gfs2_sbd *sdp = sb->s_fs_info;
485 struct dentry *dentry;
486 struct inode *inode;
487
1a0eae88 488 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
9b8df98f
SW
489 if (IS_ERR(inode)) {
490 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
491 return PTR_ERR(inode);
492 }
493 dentry = d_alloc_root(inode);
494 if (!dentry) {
495 fs_err(sdp, "can't alloc %s dentry\n", name);
496 iput(inode);
497 return -ENOMEM;
498 }
499 dentry->d_op = &gfs2_dops;
500 *dptr = dentry;
501 return 0;
f42faf4f
SW
502}
503
9b8df98f 504static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
505{
506 struct super_block *sb = sdp->sd_vfs;
507 struct gfs2_holder sb_gh;
dbb7cae2 508 u64 no_addr;
9b8df98f 509 int ret;
b3b94faa 510
9b8df98f
SW
511 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
512 LM_ST_SHARED, 0, &sb_gh);
513 if (ret) {
514 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
515 return ret;
b3b94faa 516 }
907b9bce 517
6bac243f 518 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
519 if (ret) {
520 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
521 goto out;
522 }
523
524 /* Set up the buffer cache and SB for real */
e1defc4f 525 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 526 ret = -EINVAL;
b3b94faa
DT
527 fs_err(sdp, "FS block size (%u) is too small for device "
528 "block size (%u)\n",
e1defc4f 529 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
530 goto out;
531 }
532 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 533 ret = -EINVAL;
b3b94faa
DT
534 fs_err(sdp, "FS block size (%u) is too big for machine "
535 "page size (%u)\n",
536 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
537 goto out;
538 }
b3b94faa
DT
539 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
540
f42faf4f 541 /* Get the root inode */
dbb7cae2 542 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
543 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
544 if (ret)
f42faf4f 545 goto out;
b3b94faa 546
9b8df98f
SW
547 /* Get the master inode */
548 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
549 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
550 if (ret) {
551 dput(sdp->sd_root_dir);
552 goto out;
553 }
554 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 555out:
b3b94faa 556 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 557 return ret;
b3b94faa
DT
558}
559
da6dd40d
BP
560/**
561 * map_journal_extents - create a reusable "extent" mapping from all logical
562 * blocks to all physical blocks for the given journal. This will save
563 * us time when writing journal blocks. Most journals will have only one
564 * extent that maps all their logical blocks. That's because gfs2.mkfs
565 * arranges the journal blocks sequentially to maximize performance.
566 * So the extent would map the first block for the entire file length.
567 * However, gfs2_jadd can happen while file activity is happening, so
568 * those journals may not be sequential. Less likely is the case where
569 * the users created their own journals by mounting the metafs and
570 * laying it out. But it's still possible. These journals might have
571 * several extents.
572 *
573 * TODO: This should be done in bigger chunks rather than one block at a time,
574 * but since it's only done at mount time, I'm not worried about the
575 * time it takes.
576 */
577static int map_journal_extents(struct gfs2_sbd *sdp)
578{
579 struct gfs2_jdesc *jd = sdp->sd_jdesc;
580 unsigned int lb;
581 u64 db, prev_db; /* logical block, disk block, prev disk block */
582 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
583 struct gfs2_journal_extent *jext = NULL;
584 struct buffer_head bh;
585 int rc = 0;
586
da6dd40d
BP
587 prev_db = 0;
588
a2e0f799 589 for (lb = 0; lb < i_size_read(jd->jd_inode) >> sdp->sd_sb.sb_bsize_shift; lb++) {
da6dd40d
BP
590 bh.b_state = 0;
591 bh.b_blocknr = 0;
592 bh.b_size = 1 << ip->i_inode.i_blkbits;
593 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
594 db = bh.b_blocknr;
595 if (rc || !db) {
596 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
597 "%u db=%llu\n", rc, lb, (unsigned long long)db);
598 break;
599 }
600 if (!prev_db || db != prev_db + 1) {
601 jext = kzalloc(sizeof(struct gfs2_journal_extent),
602 GFP_KERNEL);
603 if (!jext) {
604 printk(KERN_INFO "GFS2 error: out of memory "
605 "mapping journal extents.\n");
606 rc = -ENOMEM;
607 break;
608 }
609 jext->dblock = db;
610 jext->lblock = lb;
611 jext->blocks = 1;
612 list_add_tail(&jext->extent_list, &jd->extent_list);
613 } else {
614 jext->blocks++;
615 }
616 prev_db = db;
617 }
618 return rc;
619}
620
f057f6cd 621static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 622{
f057f6cd
SW
623 char *message = "FIRSTMOUNT=Done";
624 char *envp[] = { message, NULL };
625 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
626 ls->ls_first_done = 1;
627 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
628}
629
b5289681
SW
630/**
631 * gfs2_jindex_hold - Grab a lock on the jindex
632 * @sdp: The GFS2 superblock
633 * @ji_gh: the holder for the jindex glock
634 *
635 * Returns: errno
636 */
637
638static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
639{
640 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
641 struct qstr name;
642 char buf[20];
643 struct gfs2_jdesc *jd;
644 int error;
645
646 name.name = buf;
647
648 mutex_lock(&sdp->sd_jindex_mutex);
649
650 for (;;) {
651 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
652 if (error)
653 break;
654
655 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
656 name.hash = gfs2_disk_hash(name.name, name.len);
657
658 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
659 if (error == -ENOENT) {
660 error = 0;
661 break;
662 }
663
664 gfs2_glock_dq_uninit(ji_gh);
665
666 if (error)
667 break;
668
669 error = -ENOMEM;
670 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
671 if (!jd)
672 break;
673
674 INIT_LIST_HEAD(&jd->extent_list);
6ecd7c2d 675 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
676 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
677 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
678 if (!jd->jd_inode)
679 error = -ENOENT;
680 else
681 error = PTR_ERR(jd->jd_inode);
682 kfree(jd);
683 break;
684 }
685
686 spin_lock(&sdp->sd_jindex_spin);
687 jd->jd_jid = sdp->sd_journals++;
688 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
689 spin_unlock(&sdp->sd_jindex_spin);
690 }
691
692 mutex_unlock(&sdp->sd_jindex_mutex);
693
694 return error;
695}
696
b3b94faa
DT
697static int init_journal(struct gfs2_sbd *sdp, int undo)
698{
9b8df98f 699 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa 700 struct gfs2_holder ji_gh;
5c676f6d 701 struct gfs2_inode *ip;
b3b94faa
DT
702 int jindex = 1;
703 int error = 0;
704
705 if (undo) {
706 jindex = 0;
fe64d517 707 goto fail_jinode_gh;
b3b94faa
DT
708 }
709
9b8df98f 710 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 711 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 712 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 713 return PTR_ERR(sdp->sd_jindex);
b3b94faa 714 }
feaa7bba 715 ip = GFS2_I(sdp->sd_jindex);
b3b94faa
DT
716
717 /* Load in the journal index special file */
718
719 error = gfs2_jindex_hold(sdp, &ji_gh);
720 if (error) {
721 fs_err(sdp, "can't read journal index: %d\n", error);
722 goto fail;
723 }
724
0e5a9fb0 725 error = -EUSERS;
b3b94faa
DT
726 if (!gfs2_jindex_size(sdp)) {
727 fs_err(sdp, "no journals!\n");
907b9bce 728 goto fail_jindex;
b3b94faa
DT
729 }
730
731 if (sdp->sd_args.ar_spectator) {
732 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 733 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
734 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
735 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
736 } else {
737 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
738 fs_err(sdp, "can't mount journal #%u\n",
739 sdp->sd_lockstruct.ls_jid);
740 fs_err(sdp, "there are only %u journals (0 - %u)\n",
741 gfs2_jindex_size(sdp),
742 gfs2_jindex_size(sdp) - 1);
743 goto fail_jindex;
744 }
745 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
746
feaa7bba 747 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
748 &gfs2_journal_glops,
749 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
750 &sdp->sd_journal_gh);
751 if (error) {
752 fs_err(sdp, "can't acquire journal glock: %d\n", error);
753 goto fail_jindex;
754 }
755
feaa7bba
SW
756 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
757 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 758 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
759 &sdp->sd_jinode_gh);
760 if (error) {
761 fs_err(sdp, "can't acquire journal inode glock: %d\n",
762 error);
763 goto fail_journal_gh;
764 }
765
766 error = gfs2_jdesc_check(sdp->sd_jdesc);
767 if (error) {
768 fs_err(sdp, "my journal (%u) is bad: %d\n",
769 sdp->sd_jdesc->jd_jid, error);
770 goto fail_jinode_gh;
771 }
fd041f0b 772 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
773 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
774 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
775
776 /* Map the extents for this journal's blocks */
777 map_journal_extents(sdp);
b3b94faa 778 }
63997775 779 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
780
781 if (sdp->sd_lockstruct.ls_first) {
782 unsigned int x;
783 for (x = 0; x < sdp->sd_journals; x++) {
6ecd7c2d
TH
784 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
785 true);
b3b94faa
DT
786 if (error) {
787 fs_err(sdp, "error recovering journal %u: %d\n",
788 x, error);
789 goto fail_jinode_gh;
790 }
791 }
792
f057f6cd 793 gfs2_others_may_mount(sdp);
b3b94faa 794 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 795 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
796 if (error) {
797 fs_err(sdp, "error recovering my journal: %d\n", error);
798 goto fail_jinode_gh;
799 }
800 }
801
802 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
803 gfs2_glock_dq_uninit(&ji_gh);
804 jindex = 0;
805
b3b94faa
DT
806 return 0;
807
a91ea69f 808fail_jinode_gh:
b3b94faa
DT
809 if (!sdp->sd_args.ar_spectator)
810 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 811fail_journal_gh:
b3b94faa
DT
812 if (!sdp->sd_args.ar_spectator)
813 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 814fail_jindex:
b3b94faa
DT
815 gfs2_jindex_free(sdp);
816 if (jindex)
817 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 818fail:
f42faf4f 819 iput(sdp->sd_jindex);
b3b94faa
DT
820 return error;
821}
822
b3b94faa
DT
823
824static int init_inodes(struct gfs2_sbd *sdp, int undo)
825{
b3b94faa 826 int error = 0;
5c676f6d 827 struct gfs2_inode *ip;
9b8df98f 828 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
829
830 if (undo)
f42faf4f
SW
831 goto fail_qinode;
832
f42faf4f
SW
833 error = init_journal(sdp, undo);
834 if (error)
9b8df98f 835 goto fail;
b3b94faa 836
b3b94faa 837 /* Read in the master statfs inode */
9b8df98f 838 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
839 if (IS_ERR(sdp->sd_statfs_inode)) {
840 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 841 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 842 goto fail_journal;
b3b94faa
DT
843 }
844
845 /* Read in the resource index inode */
9b8df98f 846 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
847 if (IS_ERR(sdp->sd_rindex)) {
848 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
849 fs_err(sdp, "can't get resource index inode: %d\n", error);
850 goto fail_statfs;
851 }
feaa7bba 852 ip = GFS2_I(sdp->sd_rindex);
cf45b752 853 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
854
855 /* Read in the quota inode */
9b8df98f 856 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
857 if (IS_ERR(sdp->sd_quota_inode)) {
858 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
859 fs_err(sdp, "can't get quota file inode: %d\n", error);
860 goto fail_rindex;
861 }
b3b94faa
DT
862 return 0;
863
f42faf4f
SW
864fail_qinode:
865 iput(sdp->sd_quota_inode);
f42faf4f 866fail_rindex:
b3b94faa 867 gfs2_clear_rgrpd(sdp);
f42faf4f 868 iput(sdp->sd_rindex);
f42faf4f
SW
869fail_statfs:
870 iput(sdp->sd_statfs_inode);
f42faf4f
SW
871fail_journal:
872 init_journal(sdp, UNDO);
f42faf4f 873fail:
b3b94faa
DT
874 return error;
875}
876
877static int init_per_node(struct gfs2_sbd *sdp, int undo)
878{
f42faf4f 879 struct inode *pn = NULL;
b3b94faa
DT
880 char buf[30];
881 int error = 0;
5c676f6d 882 struct gfs2_inode *ip;
9b8df98f 883 struct inode *master = sdp->sd_master_dir->d_inode;
b3b94faa
DT
884
885 if (sdp->sd_args.ar_spectator)
886 return 0;
887
888 if (undo)
889 goto fail_qc_gh;
890
9b8df98f 891 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
892 if (IS_ERR(pn)) {
893 error = PTR_ERR(pn);
b3b94faa
DT
894 fs_err(sdp, "can't find per_node directory: %d\n", error);
895 return error;
896 }
897
b3b94faa 898 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
899 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
900 if (IS_ERR(sdp->sd_sc_inode)) {
901 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 902 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 903 goto fail;
b3b94faa
DT
904 }
905
b3b94faa 906 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
907 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
908 if (IS_ERR(sdp->sd_qc_inode)) {
909 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
910 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
911 goto fail_ut_i;
912 }
913
f42faf4f 914 iput(pn);
b3b94faa
DT
915 pn = NULL;
916
feaa7bba 917 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 918 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
919 &sdp->sd_sc_gh);
920 if (error) {
921 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 922 goto fail_qc_i;
b3b94faa
DT
923 }
924
feaa7bba 925 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 926 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
927 &sdp->sd_qc_gh);
928 if (error) {
929 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
930 goto fail_ut_gh;
931 }
932
933 return 0;
934
a91ea69f 935fail_qc_gh:
b3b94faa 936 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 937fail_ut_gh:
b3b94faa 938 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 939fail_qc_i:
f42faf4f 940 iput(sdp->sd_qc_inode);
a91ea69f 941fail_ut_i:
f42faf4f 942 iput(sdp->sd_sc_inode);
a91ea69f 943fail:
b3b94faa 944 if (pn)
f42faf4f 945 iput(pn);
b3b94faa
DT
946 return error;
947}
948
949static int init_threads(struct gfs2_sbd *sdp, int undo)
950{
951 struct task_struct *p;
952 int error = 0;
953
954 if (undo)
feaa7bba 955 goto fail_quotad;
b3b94faa 956
b3b94faa
DT
957 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
958 error = IS_ERR(p);
959 if (error) {
960 fs_err(sdp, "can't start logd thread: %d\n", error);
961 return error;
962 }
963 sdp->sd_logd_process = p;
964
b3b94faa
DT
965 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
966 error = IS_ERR(p);
967 if (error) {
968 fs_err(sdp, "can't start quotad thread: %d\n", error);
969 goto fail;
970 }
971 sdp->sd_quotad_process = p;
972
b3b94faa
DT
973 return 0;
974
b3b94faa 975
feaa7bba 976fail_quotad:
b3b94faa 977 kthread_stop(sdp->sd_quotad_process);
feaa7bba 978fail:
b3b94faa 979 kthread_stop(sdp->sd_logd_process);
b3b94faa
DT
980 return error;
981}
da755fdb 982
f057f6cd
SW
983static const match_table_t nolock_tokens = {
984 { Opt_jid, "jid=%d\n", },
985 { Opt_err, NULL },
986};
987
8f05228e
SW
988static void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl)
989{
990 struct gfs2_sbd *sdp = gl->gl_sbd;
991 kmem_cache_free(cachep, gl);
992 if (atomic_dec_and_test(&sdp->sd_glock_disposal))
993 wake_up(&sdp->sd_glock_wait);
994}
995
f057f6cd
SW
996static const struct lm_lockops nolock_ops = {
997 .lm_proto_name = "lock_nolock",
8f05228e 998 .lm_put_lock = nolock_put_lock,
f057f6cd
SW
999 .lm_tokens = &nolock_tokens,
1000};
1001
da755fdb
SW
1002/**
1003 * gfs2_lm_mount - mount a locking protocol
1004 * @sdp: the filesystem
3ad2f3fb 1005 * @args: mount arguments
da755fdb
SW
1006 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1007 *
1008 * Returns: errno
1009 */
1010
1011static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1012{
f057f6cd
SW
1013 const struct lm_lockops *lm;
1014 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1015 struct gfs2_args *args = &sdp->sd_args;
1016 const char *proto = sdp->sd_proto_name;
1017 const char *table = sdp->sd_table_name;
1018 const char *fsname;
1019 char *o, *options;
1020 int ret;
da755fdb 1021
f057f6cd
SW
1022 if (!strcmp("lock_nolock", proto)) {
1023 lm = &nolock_ops;
1024 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
1025#ifdef CONFIG_GFS2_FS_LOCKING_DLM
1026 } else if (!strcmp("lock_dlm", proto)) {
1027 lm = &gfs2_dlm_ops;
1028#endif
1029 } else {
1030 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1031 return -ENOENT;
1032 }
da755fdb
SW
1033
1034 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1035
f057f6cd
SW
1036 ls->ls_ops = lm;
1037 ls->ls_first = 1;
da755fdb 1038
f057f6cd
SW
1039 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1040 substring_t tmp[MAX_OPT_ARGS];
1041 int token, option;
1042
1043 if (!o || !*o)
1044 continue;
1045
1046 token = match_token(o, *lm->lm_tokens, tmp);
1047 switch (token) {
1048 case Opt_jid:
1049 ret = match_int(&tmp[0], &option);
1050 if (ret || option < 0)
1051 goto hostdata_error;
ba6e9364
SW
1052 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
1053 ls->ls_jid = option;
f057f6cd
SW
1054 break;
1055 case Opt_id:
2b88f7c5 1056 /* Obsolete, but left for backward compat purposes */
f057f6cd
SW
1057 break;
1058 case Opt_first:
1059 ret = match_int(&tmp[0], &option);
1060 if (ret || (option != 0 && option != 1))
1061 goto hostdata_error;
1062 ls->ls_first = option;
1063 break;
1064 case Opt_nodir:
1065 ret = match_int(&tmp[0], &option);
1066 if (ret || (option != 0 && option != 1))
1067 goto hostdata_error;
1068 ls->ls_nodir = option;
1069 break;
1070 case Opt_err:
1071 default:
1072hostdata_error:
1073 fs_info(sdp, "unknown hostdata (%s)\n", o);
1074 return -EINVAL;
1075 }
da755fdb
SW
1076 }
1077
1078 if (sdp->sd_args.ar_spectator)
1079 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1080 else
1081 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1082 sdp->sd_lockstruct.ls_jid);
1083
f057f6cd
SW
1084 fsname = strchr(table, ':');
1085 if (fsname)
1086 fsname++;
1087 if (lm->lm_mount == NULL) {
1088 fs_info(sdp, "Now mounting FS...\n");
1089 return 0;
da755fdb 1090 }
f057f6cd
SW
1091 ret = lm->lm_mount(sdp, fsname);
1092 if (ret == 0)
1093 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1094 return ret;
da755fdb
SW
1095}
1096
1097void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1098{
f057f6cd
SW
1099 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1100 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1101 lm->lm_unmount)
1102 lm->lm_unmount(sdp);
da755fdb 1103}
b3b94faa 1104
ba6e9364
SW
1105static int gfs2_journalid_wait(void *word)
1106{
1107 if (signal_pending(current))
1108 return -EINTR;
1109 schedule();
1110 return 0;
1111}
1112
1113static int wait_on_journal(struct gfs2_sbd *sdp)
1114{
1115 if (sdp->sd_args.ar_spectator)
1116 return 0;
1117 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1118 return 0;
1119
1120 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1121}
1122
8633ecfa
SW
1123void gfs2_online_uevent(struct gfs2_sbd *sdp)
1124{
1125 struct super_block *sb = sdp->sd_vfs;
1126 char ro[20];
1127 char spectator[20];
1128 char *envp[] = { ro, spectator, NULL };
1129 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1130 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1131 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1132}
1133
b3b94faa
DT
1134/**
1135 * fill_super - Read in superblock
1136 * @sb: The VFS superblock
1137 * @data: Mount options
1138 * @silent: Don't complain if it's not a GFS2 filesystem
1139 *
1140 * Returns: errno
1141 */
1142
f55073ff 1143static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1144{
1145 struct gfs2_sbd *sdp;
1146 struct gfs2_holder mount_gh;
1147 int error;
1148
1149 sdp = init_sbd(sb);
1150 if (!sdp) {
d92a8d48 1151 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1152 return -ENOMEM;
1153 }
f55073ff 1154 sdp->sd_args = *args;
b3b94faa 1155
fe64d517 1156 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1157 sb->s_flags |= MS_RDONLY;
fe64d517
SW
1158 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1159 }
6f04c1c7
SW
1160 if (sdp->sd_args.ar_posix_acl)
1161 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1162 if (sdp->sd_args.ar_nobarrier)
1163 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1164
719ee344
SW
1165 sb->s_magic = GFS2_MAGIC;
1166 sb->s_op = &gfs2_super_ops;
1167 sb->s_export_op = &gfs2_export_ops;
40b78a32 1168 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f
SW
1169 sb->s_qcop = &gfs2_quotactl_ops;
1170 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1171 sb->s_time_gran = 1;
1172 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1173
1174 /* Set up the buffer cache and fill in some fake block size values
1175 to allow us to read-in the on-disk superblock. */
1176 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1177 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1178 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1179 GFS2_BASIC_BLOCK_SHIFT;
1180 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
b3b94faa 1181
5e687eac 1182 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1183 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1184 if (sdp->sd_args.ar_statfs_quantum) {
1185 sdp->sd_tune.gt_statfs_slow = 0;
1186 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
1187 }
1188 else {
1189 sdp->sd_tune.gt_statfs_slow = 1;
1190 sdp->sd_tune.gt_statfs_quantum = 30;
1191 }
48c2b613 1192
b3b94faa
DT
1193 error = init_names(sdp, silent);
1194 if (error)
1195 goto fail;
1196
7c52b166
RP
1197 gfs2_create_debugfs_file(sdp);
1198
b3b94faa
DT
1199 error = gfs2_sys_fs_add(sdp);
1200 if (error)
1201 goto fail;
1202
1203 error = gfs2_lm_mount(sdp, silent);
1204 if (error)
1205 goto fail_sys;
1206
1207 error = init_locking(sdp, &mount_gh, DO);
1208 if (error)
1209 goto fail_lm;
1210
9b8df98f 1211 error = init_sb(sdp, silent);
b3b94faa
DT
1212 if (error)
1213 goto fail_locking;
b3b94faa 1214
ba6e9364
SW
1215 error = wait_on_journal(sdp);
1216 if (error)
1217 goto fail_sb;
1218
b3b94faa
DT
1219 error = init_inodes(sdp, DO);
1220 if (error)
f42faf4f 1221 goto fail_sb;
b3b94faa
DT
1222
1223 error = init_per_node(sdp, DO);
1224 if (error)
1225 goto fail_inodes;
1226
1227 error = gfs2_statfs_init(sdp);
1228 if (error) {
1229 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1230 goto fail_per_node;
1231 }
1232
1233 error = init_threads(sdp, DO);
1234 if (error)
1235 goto fail_per_node;
1236
1237 if (!(sb->s_flags & MS_RDONLY)) {
1238 error = gfs2_make_fs_rw(sdp);
1239 if (error) {
1240 fs_err(sdp, "can't make FS RW: %d\n", error);
1241 goto fail_threads;
1242 }
1243 }
1244
1245 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1246 gfs2_online_uevent(sdp);
b3b94faa
DT
1247 return 0;
1248
a91ea69f 1249fail_threads:
b3b94faa 1250 init_threads(sdp, UNDO);
a91ea69f 1251fail_per_node:
b3b94faa 1252 init_per_node(sdp, UNDO);
a91ea69f 1253fail_inodes:
b3b94faa 1254 init_inodes(sdp, UNDO);
a91ea69f 1255fail_sb:
9b8df98f
SW
1256 if (sdp->sd_root_dir)
1257 dput(sdp->sd_root_dir);
1258 if (sdp->sd_master_dir)
1259 dput(sdp->sd_master_dir);
e7c8707e
SW
1260 if (sb->s_root)
1261 dput(sb->s_root);
9b8df98f 1262 sb->s_root = NULL;
a91ea69f 1263fail_locking:
b3b94faa 1264 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1265fail_lm:
c1184f8a 1266 invalidate_inodes(sb);
fefc03bf 1267 gfs2_gl_hash_clear(sdp);
b3b94faa 1268 gfs2_lm_unmount(sdp);
a91ea69f 1269fail_sys:
b3b94faa 1270 gfs2_sys_fs_del(sdp);
a91ea69f 1271fail:
5f882096 1272 gfs2_delete_debugfs_file(sdp);
a2c45807 1273 kfree(sdp);
5c676f6d 1274 sb->s_fs_info = NULL;
b3b94faa
DT
1275 return error;
1276}
1277
f55073ff 1278static int set_gfs2_super(struct super_block *s, void *data)
b3b94faa 1279{
f55073ff
SW
1280 s->s_bdev = data;
1281 s->s_dev = s->s_bdev->bd_dev;
1282
1283 /*
1284 * We set the bdi here to the queue backing, file systems can
1285 * overwrite this in ->fill_super()
1286 */
1287 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1288 return 0;
86384605
AD
1289}
1290
f55073ff 1291static int test_gfs2_super(struct super_block *s, void *ptr)
f6d03139
SW
1292{
1293 struct block_device *bdev = ptr;
1294 return (bdev == s->s_bdev);
1295}
1296
f55073ff
SW
1297/**
1298 * gfs2_get_sb - Get the GFS2 superblock
1299 * @fs_type: The GFS2 filesystem type
1300 * @flags: Mount flags
1301 * @dev_name: The name of the device
1302 * @data: The mount arguments
1303 * @mnt: The vfsmnt for this mount
1304 *
1305 * Q. Why not use get_sb_bdev() ?
1306 * A. We need to select one of two root directories to mount, independent
1307 * of whether this is the initial, or subsequent, mount of this sb
1308 *
1309 * Returns: 0 or -ve on error
1310 */
1311
1312static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1313 const char *dev_name, void *data, struct vfsmount *mnt)
1314{
1315 struct block_device *bdev;
1316 struct super_block *s;
1317 fmode_t mode = FMODE_READ;
1318 int error;
1319 struct gfs2_args args;
1320 struct gfs2_sbd *sdp;
1321
1322 if (!(flags & MS_RDONLY))
1323 mode |= FMODE_WRITE;
1324
1325 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1326 if (IS_ERR(bdev))
1327 return PTR_ERR(bdev);
1328
1329 /*
1330 * once the super is inserted into the list by sget, s_umount
1331 * will protect the lockfs code from trying to start a snapshot
1332 * while we are mounting
1333 */
1334 mutex_lock(&bdev->bd_fsfreeze_mutex);
1335 if (bdev->bd_fsfreeze_count > 0) {
1336 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1337 error = -EBUSY;
1338 goto error_bdev;
1339 }
1340 s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev);
1341 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1342 error = PTR_ERR(s);
1343 if (IS_ERR(s))
1344 goto error_bdev;
1345
1346 memset(&args, 0, sizeof(args));
1347 args.ar_quota = GFS2_QUOTA_DEFAULT;
1348 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1349 args.ar_commit = 30;
3d3c10f2
BM
1350 args.ar_statfs_quantum = 30;
1351 args.ar_quota_quantum = 60;
f55073ff
SW
1352 args.ar_errors = GFS2_ERRORS_DEFAULT;
1353
1354 error = gfs2_mount_args(&args, data);
1355 if (error) {
1356 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1357 if (s->s_root)
1358 goto error_super;
1359 deactivate_locked_super(s);
1360 return error;
1361 }
1362
1363 if (s->s_root) {
1364 error = -EBUSY;
1365 if ((flags ^ s->s_flags) & MS_RDONLY)
1366 goto error_super;
1367 close_bdev_exclusive(bdev, mode);
1368 } else {
1369 char b[BDEVNAME_SIZE];
1370
1371 s->s_flags = flags;
1372 s->s_mode = mode;
1373 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1374 sb_set_blocksize(s, block_size(bdev));
1375 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
1376 if (error) {
1377 deactivate_locked_super(s);
1378 return error;
1379 }
1380 s->s_flags |= MS_ACTIVE;
1381 bdev->bd_super = s;
1382 }
1383
1384 sdp = s->s_fs_info;
1385 mnt->mnt_sb = s;
1386 if (args.ar_meta)
1387 mnt->mnt_root = dget(sdp->sd_master_dir);
1388 else
1389 mnt->mnt_root = dget(sdp->sd_root_dir);
1390 return 0;
1391
1392error_super:
1393 deactivate_locked_super(s);
1394error_bdev:
1395 close_bdev_exclusive(bdev, mode);
1396 return error;
1397}
1398
f6d03139
SW
1399static int set_meta_super(struct super_block *s, void *ptr)
1400{
1401 return -EINVAL;
1402}
1403
003dec89
SW
1404static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1405 const char *dev_name, void *data, struct vfsmount *mnt)
86384605 1406{
f6d03139 1407 struct super_block *s;
003dec89 1408 struct gfs2_sbd *sdp;
e24977d4 1409 struct path path;
86384605 1410 int error;
907b9bce 1411
e24977d4 1412 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1413 if (error) {
9b8df98f
SW
1414 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1415 dev_name, error);
003dec89 1416 return error;
86384605 1417 }
f55073ff 1418 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super,
f6d03139 1419 path.dentry->d_inode->i_sb->s_bdev);
e24977d4 1420 path_put(&path);
f6d03139 1421 if (IS_ERR(s)) {
86384605 1422 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
f6d03139 1423 return PTR_ERR(s);
86384605 1424 }
f55073ff
SW
1425 if ((flags ^ s->s_flags) & MS_RDONLY) {
1426 deactivate_locked_super(s);
1427 return -EBUSY;
1428 }
f6d03139
SW
1429 sdp = s->s_fs_info;
1430 mnt->mnt_sb = s;
9b8df98f
SW
1431 mnt->mnt_root = dget(sdp->sd_master_dir);
1432 return 0;
b3b94faa
DT
1433}
1434
419c93e0 1435static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1436{
1437 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1438
1439 if (sdp == NULL) {
1440 kill_block_super(sb);
1441 return;
3af165ac 1442 }
88a19ad0
SW
1443
1444 gfs2_meta_syncfs(sdp);
1445 dput(sdp->sd_root_dir);
1446 dput(sdp->sd_master_dir);
1447 sdp->sd_root_dir = NULL;
1448 sdp->sd_master_dir = NULL;
fefc03bf 1449 shrink_dcache_sb(sb);
9b8df98f 1450 kill_block_super(sb);
88a19ad0
SW
1451 gfs2_delete_debugfs_file(sdp);
1452 kfree(sdp);
86384605
AD
1453}
1454
b3b94faa
DT
1455struct file_system_type gfs2_fs_type = {
1456 .name = "gfs2",
1457 .fs_flags = FS_REQUIRES_DEV,
1458 .get_sb = gfs2_get_sb,
419c93e0
SW
1459 .kill_sb = gfs2_kill_sb,
1460 .owner = THIS_MODULE,
1461};
1462
1463struct file_system_type gfs2meta_fs_type = {
1464 .name = "gfs2meta",
1465 .fs_flags = FS_REQUIRES_DEV,
86384605 1466 .get_sb = gfs2_get_sb_meta,
b3b94faa
DT
1467 .owner = THIS_MODULE,
1468};
1469