blob: 922d289aeeb315ce4a5c1131371ea90f266c0df4 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
23#include <linux/jbd.h>
24#include <linux/ext3_fs.h>
25#include <linux/ext3_jbd.h>
26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070031#include <linux/exportfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
Mark Bellon8fc27512005-09-06 15:16:54 -070037#include <linux/seq_file.h>
vignesh babu3fc74262007-07-15 23:41:17 -070038#include <linux/log2.h>
Dan Magenheimerd71bc6d2011-05-26 10:01:49 -060039#include <linux/cleancache.h>
Ben Dooks381be252005-10-30 15:02:56 -080040
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <asm/uaccess.h>
Ben Dooks381be252005-10-30 15:02:56 -080042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include "xattr.h"
44#include "acl.h"
Ben Dooks381be252005-10-30 15:02:56 -080045#include "namei.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
Lukas Czerner785c4bc2011-05-23 18:33:01 +020047#define CREATE_TRACE_POINTS
48#include <trace/events/ext3.h>
49
Linus Torvaldsbbae8bc2009-04-06 17:16:47 -070050#ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
51 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
52#else
53 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
54#endif
55
Johann Lombardi71b96252006-01-08 01:03:20 -080056static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
57 unsigned long journal_devnum);
Linus Torvalds1da177e2005-04-16 15:20:36 -070058static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
Eric Sandeeneee194e2006-09-27 01:49:30 -070059 unsigned int);
Takashi Satoc4be0c12009-01-09 16:40:58 -080060static int ext3_commit_super(struct super_block *sb,
61 struct ext3_super_block *es,
Linus Torvalds1da177e2005-04-16 15:20:36 -070062 int sync);
63static void ext3_mark_recovery_complete(struct super_block * sb,
64 struct ext3_super_block * es);
65static void ext3_clear_journal_err(struct super_block * sb,
66 struct ext3_super_block * es);
67static int ext3_sync_fs(struct super_block *sb, int wait);
68static const char *ext3_decode_error(struct super_block * sb, int errno,
69 char nbuf[16]);
70static int ext3_remount (struct super_block * sb, int * flags, char * data);
David Howells726c3342006-06-23 02:02:58 -070071static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080072static int ext3_unfreeze(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080073static int ext3_freeze(struct super_block *sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Mingming Caoae6ddcc2006-09-27 01:49:27 -070075/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 * Wrappers for journal_start/end.
77 *
78 * The only special thing we need to do here is to make sure that all
79 * journal_end calls result in the superblock being marked dirty, so
80 * that sync() will call the filesystem's write_super callback if
Mingming Caoae6ddcc2006-09-27 01:49:27 -070081 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 */
83handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
84{
85 journal_t *journal;
86
87 if (sb->s_flags & MS_RDONLY)
88 return ERR_PTR(-EROFS);
89
90 /* Special case here: if the journal has aborted behind our
91 * backs (eg. EIO in the commit thread), then we still need to
92 * take the FS itself readonly cleanly. */
93 journal = EXT3_SB(sb)->s_journal;
94 if (is_journal_aborted(journal)) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -070095 ext3_abort(sb, __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 "Detected aborted journal");
97 return ERR_PTR(-EROFS);
98 }
99
100 return journal_start(journal, nblocks);
101}
102
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700103/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 * The only special thing we need to do here is to make sure that all
105 * journal_stop calls result in the superblock being marked dirty, so
106 * that sync() will call the filesystem's write_super callback if
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700107 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 */
109int __ext3_journal_stop(const char *where, handle_t *handle)
110{
111 struct super_block *sb;
112 int err;
113 int rc;
114
115 sb = handle->h_transaction->t_journal->j_private;
116 err = handle->h_err;
117 rc = journal_stop(handle);
118
119 if (!err)
120 err = rc;
121 if (err)
122 __ext3_std_error(sb, where, err);
123 return err;
124}
125
126void ext3_journal_abort_handle(const char *caller, const char *err_fn,
127 struct buffer_head *bh, handle_t *handle, int err)
128{
129 char nbuf[16];
130 const char *errstr = ext3_decode_error(NULL, err, nbuf);
131
132 if (bh)
133 BUFFER_TRACE(bh, "abort");
134
135 if (!handle->h_err)
136 handle->h_err = err;
137
138 if (is_handle_aborted(handle))
139 return;
140
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100141 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
142 caller, errstr, err_fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144 journal_abort_handle(handle);
145}
146
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100147void ext3_msg(struct super_block *sb, const char *prefix,
148 const char *fmt, ...)
149{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800150 struct va_format vaf;
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100151 va_list args;
152
153 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800154
155 vaf.fmt = fmt;
156 vaf.va = &args;
157
158 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
159
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100160 va_end(args);
161}
162
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163/* Deal with the reporting of failure conditions on a filesystem such as
164 * inconsistencies detected or read IO failures.
165 *
166 * On ext2, we can store the error state of the filesystem in the
167 * superblock. That is not possible on ext3, because we may have other
168 * write ordering constraints on the superblock which prevent us from
169 * writing it out straight away; and given that the journal is about to
170 * be aborted, we can't rely on the current, or future, transactions to
171 * write out the superblock safely.
172 *
173 * We'll just use the journal_abort() error code to record an error in
Thadeu Lima de Souza Cascardod6b198b2010-01-17 19:10:07 -0200174 * the journal instead. On recovery, the journal will complain about
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 * that error until we've noted it down and cleared it.
176 */
177
178static void ext3_handle_error(struct super_block *sb)
179{
180 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
181
182 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
183 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
184
185 if (sb->s_flags & MS_RDONLY)
186 return;
187
Vasily Averin7543fc72006-09-27 01:49:33 -0700188 if (!test_opt (sb, ERRORS_CONT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 journal_t *journal = EXT3_SB(sb)->s_journal;
190
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300191 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 if (journal)
193 journal_abort(journal, -EIO);
194 }
Vasily Averin7543fc72006-09-27 01:49:33 -0700195 if (test_opt (sb, ERRORS_RO)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100196 ext3_msg(sb, KERN_CRIT,
197 "error: remounting filesystem read-only");
Vasily Averin7543fc72006-09-27 01:49:33 -0700198 sb->s_flags |= MS_RDONLY;
199 }
200 ext3_commit_super(sb, es, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 if (test_opt(sb, ERRORS_PANIC))
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100202 panic("EXT3-fs (%s): panic forced after error\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204}
205
Joe Perches99fbb1e2010-11-09 10:18:05 -0800206void ext3_error(struct super_block *sb, const char *function,
207 const char *fmt, ...)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800209 struct va_format vaf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 va_list args;
211
212 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800213
214 vaf.fmt = fmt;
215 vaf.va = &args;
216
217 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
218 sb->s_id, function, &vaf);
219
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 va_end(args);
221
222 ext3_handle_error(sb);
223}
224
225static const char *ext3_decode_error(struct super_block * sb, int errno,
226 char nbuf[16])
227{
228 char *errstr = NULL;
229
230 switch (errno) {
231 case -EIO:
232 errstr = "IO failure";
233 break;
234 case -ENOMEM:
235 errstr = "Out of memory";
236 break;
237 case -EROFS:
238 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
239 errstr = "Journal has aborted";
240 else
241 errstr = "Readonly filesystem";
242 break;
243 default:
244 /* If the caller passed in an extra buffer for unknown
245 * errors, textualise them now. Else we just return
246 * NULL. */
247 if (nbuf) {
248 /* Check for truncated error codes... */
249 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
250 errstr = nbuf;
251 }
252 break;
253 }
254
255 return errstr;
256}
257
258/* __ext3_std_error decodes expected errors from journaling functions
259 * automatically and invokes the appropriate error response. */
260
261void __ext3_std_error (struct super_block * sb, const char * function,
262 int errno)
263{
264 char nbuf[16];
Stephen Tweedie30121622005-05-18 11:47:17 -0400265 const char *errstr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
Stephen Tweedie30121622005-05-18 11:47:17 -0400267 /* Special case: if the error is EROFS, and we're not already
268 * inside a transaction, then there's really no point in logging
269 * an error. */
270 if (errno == -EROFS && journal_current_handle() == NULL &&
271 (sb->s_flags & MS_RDONLY))
272 return;
273
274 errstr = ext3_decode_error(sb, errno, nbuf);
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100275 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
277 ext3_handle_error(sb);
278}
279
280/*
281 * ext3_abort is a much stronger failure handler than ext3_error. The
282 * abort function may be used to deal with unrecoverable failures such
283 * as journal IO errors or ENOMEM at a critical moment in log management.
284 *
285 * We unconditionally force the filesystem into an ABORT|READONLY state,
286 * unless the error response on the fs has been set to panic in which
287 * case we take the easy way out and panic immediately.
288 */
289
Joe Perches99fbb1e2010-11-09 10:18:05 -0800290void ext3_abort(struct super_block *sb, const char *function,
291 const char *fmt, ...)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800293 struct va_format vaf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 va_list args;
295
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800297
298 vaf.fmt = fmt;
299 vaf.va = &args;
300
301 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
302 sb->s_id, function, &vaf);
303
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 va_end(args);
305
306 if (test_opt(sb, ERRORS_PANIC))
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100307 panic("EXT3-fs: panic from previous error\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
309 if (sb->s_flags & MS_RDONLY)
310 return;
311
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100312 ext3_msg(sb, KERN_CRIT,
313 "error: remounting filesystem read-only");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
315 sb->s_flags |= MS_RDONLY;
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300316 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400317 if (EXT3_SB(sb)->s_journal)
318 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319}
320
Joe Perches99fbb1e2010-11-09 10:18:05 -0800321void ext3_warning(struct super_block *sb, const char *function,
322 const char *fmt, ...)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800324 struct va_format vaf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 va_list args;
326
327 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800328
329 vaf.fmt = fmt;
330 vaf.va = &args;
331
332 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
333 sb->s_id, function, &vaf);
334
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 va_end(args);
336}
337
338void ext3_update_dynamic_rev(struct super_block *sb)
339{
340 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
341
342 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
343 return;
344
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100345 ext3_msg(sb, KERN_WARNING,
346 "warning: updating to rev %d because of "
347 "new feature flag, running e2fsck is recommended",
348 EXT3_DYNAMIC_REV);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
350 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
351 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
352 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
353 /* leave es->s_feature_*compat flags alone */
354 /* es->s_uuid will be set by e2fsck if empty */
355
356 /*
357 * The rest of the superblock fields should be zero, and if not it
358 * means they are likely already in use, so leave them alone. We
359 * can leave it up to e2fsck to clean up any inconsistencies there.
360 */
361}
362
363/*
364 * Open the external journal device
365 */
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100366static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
368 struct block_device *bdev;
369 char b[BDEVNAME_SIZE];
370
Tejun Heod4d77622010-11-13 11:55:18 +0100371 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 if (IS_ERR(bdev))
373 goto fail;
374 return bdev;
375
376fail:
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100377 ext3_msg(sb, "error: failed to open journal device %s: %ld",
378 __bdevname(dev, b), PTR_ERR(bdev));
379
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 return NULL;
381}
382
383/*
384 * Release the journal device
385 */
386static int ext3_blkdev_put(struct block_device *bdev)
387{
Tejun Heoe525fd82010-11-13 11:55:17 +0100388 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389}
390
391static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
392{
393 struct block_device *bdev;
394 int ret = -ENODEV;
395
396 bdev = sbi->journal_bdev;
397 if (bdev) {
398 ret = ext3_blkdev_put(bdev);
399 sbi->journal_bdev = NULL;
400 }
401 return ret;
402}
403
404static inline struct inode *orphan_list_entry(struct list_head *l)
405{
406 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
407}
408
409static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
410{
411 struct list_head *l;
412
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100413 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 le32_to_cpu(sbi->s_es->s_last_orphan));
415
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100416 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 list_for_each(l, &sbi->s_orphan) {
418 struct inode *inode = orphan_list_entry(l);
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100419 ext3_msg(sb, KERN_ERR, " "
Eric Sandeeneee194e2006-09-27 01:49:30 -0700420 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 inode->i_sb->s_id, inode->i_ino, inode,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700422 inode->i_mode, inode->i_nlink,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 NEXT_ORPHAN(inode));
424 }
425}
426
427static void ext3_put_super (struct super_block * sb)
428{
429 struct ext3_sb_info *sbi = EXT3_SB(sb);
430 struct ext3_super_block *es = sbi->s_es;
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400431 int i, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400433 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 ext3_xattr_put_super(sb);
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400435 err = journal_destroy(sbi->s_journal);
436 sbi->s_journal = NULL;
437 if (err < 0)
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -0700438 ext3_abort(sb, __func__, "Couldn't clean up the journal");
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 if (!(sb->s_flags & MS_RDONLY)) {
441 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
442 es->s_state = cpu_to_le16(sbi->s_mount_state);
443 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
444 mark_buffer_dirty(sbi->s_sbh);
445 ext3_commit_super(sb, es, 1);
446 }
447
448 for (i = 0; i < sbi->s_gdb_count; i++)
449 brelse(sbi->s_group_desc[i]);
450 kfree(sbi->s_group_desc);
451 percpu_counter_destroy(&sbi->s_freeblocks_counter);
452 percpu_counter_destroy(&sbi->s_freeinodes_counter);
453 percpu_counter_destroy(&sbi->s_dirs_counter);
454 brelse(sbi->s_sbh);
455#ifdef CONFIG_QUOTA
456 for (i = 0; i < MAXQUOTAS; i++)
457 kfree(sbi->s_qf_names[i]);
458#endif
459
460 /* Debugging code just in case the in-memory inode orphan list
461 * isn't empty. The on-disk one can be non-empty if we've
462 * detected an error and taken the fs readonly, but the
463 * in-memory list had better be clean by this point. */
464 if (!list_empty(&sbi->s_orphan))
465 dump_orphan_list(sb, sbi);
466 J_ASSERT(list_empty(&sbi->s_orphan));
467
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700468 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
470 /*
471 * Invalidate the journal device's buffers. We don't want them
472 * floating about in memory - the physical journal device may
473 * hotswapped, and it breaks the `ro-after' testing code.
474 */
475 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700476 invalidate_bdev(sbi->journal_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 ext3_blkdev_remove(sbi);
478 }
479 sb->s_fs_info = NULL;
Pekka Enberg5df096d2009-01-07 18:07:25 -0800480 kfree(sbi->s_blockgroup_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 kfree(sbi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482}
483
Christoph Lametere18b8902006-12-06 20:33:20 -0800484static struct kmem_cache *ext3_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
486/*
487 * Called inside transaction, so use GFP_NOFS
488 */
489static struct inode *ext3_alloc_inode(struct super_block *sb)
490{
491 struct ext3_inode_info *ei;
492
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800493 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 if (!ei)
495 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 ei->i_block_alloc_info = NULL;
497 ei->vfs_inode.i_version = 1;
Jan Karafe8bc912009-10-16 19:26:15 +0200498 atomic_set(&ei->i_datasync_tid, 0);
499 atomic_set(&ei->i_sync_tid, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 return &ei->vfs_inode;
501}
502
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200503static int ext3_drop_inode(struct inode *inode)
504{
505 int drop = generic_drop_inode(inode);
506
507 trace_ext3_drop_inode(inode, drop);
508 return drop;
509}
510
Nick Pigginfa0d7e32011-01-07 17:49:49 +1100511static void ext3_i_callback(struct rcu_head *head)
512{
513 struct inode *inode = container_of(head, struct inode, i_rcu);
514 INIT_LIST_HEAD(&inode->i_dentry);
515 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
516}
517
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518static void ext3_destroy_inode(struct inode *inode)
519{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700520 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
521 printk("EXT3 Inode %p: orphan list check failed!\n",
522 EXT3_I(inode));
523 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
524 EXT3_I(inode), sizeof(struct ext3_inode_info),
525 false);
526 dump_stack();
527 }
Nick Pigginfa0d7e32011-01-07 17:49:49 +1100528 call_rcu(&inode->i_rcu, ext3_i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529}
530
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700531static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532{
533 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
534
Christoph Lametera35afb82007-05-16 22:10:57 -0700535 INIT_LIST_HEAD(&ei->i_orphan);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536#ifdef CONFIG_EXT3_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700537 init_rwsem(&ei->xattr_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700539 mutex_init(&ei->truncate_mutex);
540 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541}
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700542
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543static int init_inodecache(void)
544{
545 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
546 sizeof(struct ext3_inode_info),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800547 0, (SLAB_RECLAIM_ACCOUNT|
548 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900549 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 if (ext3_inode_cachep == NULL)
551 return -ENOMEM;
552 return 0;
553}
554
555static void destroy_inodecache(void)
556{
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -0700557 kmem_cache_destroy(ext3_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558}
559
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200560static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
Mark Bellon8fc27512005-09-06 15:16:54 -0700561{
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200562#if defined(CONFIG_QUOTA)
OGAWA Hirofumi275abf52005-09-22 21:44:03 -0700563 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564
Jan Kara1aeec432009-11-30 22:22:41 +0100565 if (sbi->s_jquota_fmt) {
566 char *fmtname = "";
567
568 switch (sbi->s_jquota_fmt) {
569 case QFMT_VFS_OLD:
570 fmtname = "vfsold";
571 break;
572 case QFMT_VFS_V0:
573 fmtname = "vfsv0";
574 break;
575 case QFMT_VFS_V1:
576 fmtname = "vfsv1";
577 break;
578 }
579 seq_printf(seq, ",jqfmt=%s", fmtname);
580 }
Mark Bellon8fc27512005-09-06 15:16:54 -0700581
582 if (sbi->s_qf_names[USRQUOTA])
583 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
584
585 if (sbi->s_qf_names[GRPQUOTA])
586 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
587
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300588 if (test_opt(sb, USRQUOTA))
Mark Bellon8fc27512005-09-06 15:16:54 -0700589 seq_puts(seq, ",usrquota");
590
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300591 if (test_opt(sb, GRPQUOTA))
Mark Bellon8fc27512005-09-06 15:16:54 -0700592 seq_puts(seq, ",grpquota");
593#endif
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200594}
595
Jan Kara3c4cec62009-08-24 16:38:43 +0200596static char *data_mode_string(unsigned long mode)
597{
598 switch (mode) {
599 case EXT3_MOUNT_JOURNAL_DATA:
600 return "journal";
601 case EXT3_MOUNT_ORDERED_DATA:
602 return "ordered";
603 case EXT3_MOUNT_WRITEBACK_DATA:
604 return "writeback";
605 }
606 return "unknown";
607}
608
Miklos Szeredi571beed2007-10-16 23:26:26 -0700609/*
610 * Show an option if
611 * - it's set to a non-default value OR
612 * - if the per-sb default is different from the global default
613 */
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200614static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
615{
616 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredi571beed2007-10-16 23:26:26 -0700617 struct ext3_sb_info *sbi = EXT3_SB(sb);
618 struct ext3_super_block *es = sbi->s_es;
619 unsigned long def_mount_opts;
620
621 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
622
623 if (sbi->s_sb_block != 1)
624 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
625 if (test_opt(sb, MINIX_DF))
626 seq_puts(seq, ",minixdf");
627 if (test_opt(sb, GRPID))
628 seq_puts(seq, ",grpid");
629 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
630 seq_puts(seq, ",nogrpid");
631 if (sbi->s_resuid != EXT3_DEF_RESUID ||
632 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
633 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
634 }
635 if (sbi->s_resgid != EXT3_DEF_RESGID ||
636 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
637 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
638 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800639 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredi571beed2007-10-16 23:26:26 -0700640 int def_errors = le16_to_cpu(es->s_errors);
641
642 if (def_errors == EXT3_ERRORS_PANIC ||
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800643 def_errors == EXT3_ERRORS_CONTINUE) {
644 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700645 }
646 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800647 if (test_opt(sb, ERRORS_CONT))
648 seq_puts(seq, ",errors=continue");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700649 if (test_opt(sb, ERRORS_PANIC))
650 seq_puts(seq, ",errors=panic");
651 if (test_opt(sb, NO_UID32))
652 seq_puts(seq, ",nouid32");
653 if (test_opt(sb, DEBUG))
654 seq_puts(seq, ",debug");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700655#ifdef CONFIG_EXT3_FS_XATTR
656 if (test_opt(sb, XATTR_USER))
657 seq_puts(seq, ",user_xattr");
658 if (!test_opt(sb, XATTR_USER) &&
659 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
660 seq_puts(seq, ",nouser_xattr");
661 }
662#endif
663#ifdef CONFIG_EXT3_FS_POSIX_ACL
664 if (test_opt(sb, POSIX_ACL))
665 seq_puts(seq, ",acl");
666 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
667 seq_puts(seq, ",noacl");
668#endif
669 if (!test_opt(sb, RESERVATION))
670 seq_puts(seq, ",noreservation");
671 if (sbi->s_commit_interval) {
672 seq_printf(seq, ",commit=%u",
673 (unsigned) (sbi->s_commit_interval / HZ));
674 }
Eric Sandeen0636c732010-04-30 11:09:34 -0500675
676 /*
677 * Always display barrier state so it's clear what the status is.
678 */
679 seq_puts(seq, ",barrier=");
680 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300681 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700682 if (test_opt(sb, DATA_ERR_ABORT))
683 seq_puts(seq, ",data_err=abort");
684
Eric Sandeendee1d3b2009-11-16 16:50:49 -0600685 if (test_opt(sb, NOLOAD))
686 seq_puts(seq, ",norecovery");
687
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200688 ext3_show_quota_options(seq, sb);
Mark Bellon8fc27512005-09-06 15:16:54 -0700689
690 return 0;
691}
692
NeilBrownfdb36672006-09-16 12:15:38 -0700693
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700694static struct inode *ext3_nfs_get_inode(struct super_block *sb,
695 u64 ino, u32 generation)
NeilBrownfdb36672006-09-16 12:15:38 -0700696{
NeilBrownfdb36672006-09-16 12:15:38 -0700697 struct inode *inode;
NeilBrownfdb36672006-09-16 12:15:38 -0700698
699 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
700 return ERR_PTR(-ESTALE);
701 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
702 return ERR_PTR(-ESTALE);
703
704 /* iget isn't really right if the inode is currently unallocated!!
705 *
706 * ext3_read_inode will return a bad_inode if the inode had been
707 * deleted, so we should be safe.
708 *
709 * Currently we don't know the generation for parent directory, so
710 * a generation of 0 means "accept any"
711 */
David Howells473043d2008-02-07 00:15:36 -0800712 inode = ext3_iget(sb, ino);
713 if (IS_ERR(inode))
714 return ERR_CAST(inode);
715 if (generation && inode->i_generation != generation) {
NeilBrownfdb36672006-09-16 12:15:38 -0700716 iput(inode);
717 return ERR_PTR(-ESTALE);
718 }
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700719
720 return inode;
721}
722
723static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
724 int fh_len, int fh_type)
725{
726 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
727 ext3_nfs_get_inode);
728}
729
730static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
731 int fh_len, int fh_type)
732{
733 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
734 ext3_nfs_get_inode);
NeilBrownfdb36672006-09-16 12:15:38 -0700735}
736
Toshiyuki Okajima6b082b52009-01-05 22:38:14 -0500737/*
738 * Try to release metadata pages (indirect blocks, directories) which are
739 * mapped via the block device. Since these pages could have journal heads
740 * which would prevent try_to_free_buffers() from freeing them, we must use
741 * jbd layer's try_to_free_buffers() function to release them.
742 */
743static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
744 gfp_t wait)
745{
746 journal_t *journal = EXT3_SB(sb)->s_journal;
747
748 WARN_ON(PageChecked(page));
749 if (!page_has_buffers(page))
750 return 0;
751 if (journal)
752 return journal_try_to_free_buffers(journal, page,
753 wait & ~__GFP_WAIT);
754 return try_to_free_buffers(page);
755}
756
Mark Bellon8fc27512005-09-06 15:16:54 -0700757#ifdef CONFIG_QUOTA
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
759#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
760
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761static int ext3_write_dquot(struct dquot *dquot);
762static int ext3_acquire_dquot(struct dquot *dquot);
763static int ext3_release_dquot(struct dquot *dquot);
764static int ext3_mark_dquot_dirty(struct dquot *dquot);
765static int ext3_write_info(struct super_block *sb, int type);
Jan Kara2fd83a42008-04-28 02:14:34 -0700766static int ext3_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +0200767 struct path *path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768static int ext3_quota_on_mount(struct super_block *sb, int type);
769static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
770 size_t len, loff_t off);
771static ssize_t ext3_quota_write(struct super_block *sb, int type,
772 const char *data, size_t len, loff_t off);
773
Alexey Dobriyan61e225d2009-09-21 17:01:08 -0700774static const struct dquot_operations ext3_quota_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 .write_dquot = ext3_write_dquot,
776 .acquire_dquot = ext3_acquire_dquot,
777 .release_dquot = ext3_release_dquot,
778 .mark_dirty = ext3_mark_dquot_dirty,
Jan Kara157091a2008-11-25 15:31:34 +0100779 .write_info = ext3_write_info,
780 .alloc_dquot = dquot_alloc,
781 .destroy_dquot = dquot_destroy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782};
783
Alexey Dobriyan0d54b212009-09-21 17:01:09 -0700784static const struct quotactl_ops ext3_qctl_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 .quota_on = ext3_quota_on,
Christoph Hellwig287a8092010-05-19 07:16:45 -0400786 .quota_off = dquot_quota_off,
787 .quota_sync = dquot_quota_sync,
788 .get_info = dquot_get_dqinfo,
789 .set_info = dquot_set_dqinfo,
790 .get_dqblk = dquot_get_dqblk,
791 .set_dqblk = dquot_set_dqblk
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792};
793#endif
794
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800795static const struct super_operations ext3_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 .alloc_inode = ext3_alloc_inode,
797 .destroy_inode = ext3_destroy_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 .write_inode = ext3_write_inode,
799 .dirty_inode = ext3_dirty_inode,
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200800 .drop_inode = ext3_drop_inode,
Al Viroac14a952010-06-06 07:08:19 -0400801 .evict_inode = ext3_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 .put_super = ext3_put_super,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 .sync_fs = ext3_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -0800804 .freeze_fs = ext3_freeze,
805 .unfreeze_fs = ext3_unfreeze,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 .statfs = ext3_statfs,
807 .remount_fs = ext3_remount,
Mark Bellon8fc27512005-09-06 15:16:54 -0700808 .show_options = ext3_show_options,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809#ifdef CONFIG_QUOTA
810 .quota_read = ext3_quota_read,
811 .quota_write = ext3_quota_write,
812#endif
Toshiyuki Okajima6b082b52009-01-05 22:38:14 -0500813 .bdev_try_to_free_page = bdev_try_to_free_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814};
815
Christoph Hellwig39655162007-10-21 16:42:17 -0700816static const struct export_operations ext3_export_ops = {
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700817 .fh_to_dentry = ext3_fh_to_dentry,
818 .fh_to_parent = ext3_fh_to_parent,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 .get_parent = ext3_get_parent,
820};
821
822enum {
823 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
824 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Adrian Bunk2860b732005-11-08 21:34:59 -0800825 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Badari Pulavartyade1a292006-06-26 00:25:04 -0700827 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
Johann Lombardi71b96252006-01-08 01:03:20 -0800828 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700830 Opt_data_err_abort, Opt_data_err_ignore,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara1aeec432009-11-30 22:22:41 +0100832 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
Eric Sandeen0636c732010-04-30 11:09:34 -0500833 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
834 Opt_resize, Opt_usrquota, Opt_grpquota
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835};
836
Steven Whitehousea447c092008-10-13 10:46:57 +0100837static const match_table_t tokens = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 {Opt_bsd_df, "bsddf"},
839 {Opt_minix_df, "minixdf"},
840 {Opt_grpid, "grpid"},
841 {Opt_grpid, "bsdgroups"},
842 {Opt_nogrpid, "nogrpid"},
843 {Opt_nogrpid, "sysvgroups"},
844 {Opt_resgid, "resgid=%u"},
845 {Opt_resuid, "resuid=%u"},
846 {Opt_sb, "sb=%u"},
847 {Opt_err_cont, "errors=continue"},
848 {Opt_err_panic, "errors=panic"},
849 {Opt_err_ro, "errors=remount-ro"},
850 {Opt_nouid32, "nouid32"},
851 {Opt_nocheck, "nocheck"},
852 {Opt_nocheck, "check=none"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 {Opt_debug, "debug"},
854 {Opt_oldalloc, "oldalloc"},
855 {Opt_orlov, "orlov"},
856 {Opt_user_xattr, "user_xattr"},
857 {Opt_nouser_xattr, "nouser_xattr"},
858 {Opt_acl, "acl"},
859 {Opt_noacl, "noacl"},
860 {Opt_reservation, "reservation"},
861 {Opt_noreservation, "noreservation"},
862 {Opt_noload, "noload"},
Eric Sandeendee1d3b2009-11-16 16:50:49 -0600863 {Opt_noload, "norecovery"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 {Opt_nobh, "nobh"},
Badari Pulavartyade1a292006-06-26 00:25:04 -0700865 {Opt_bh, "bh"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 {Opt_commit, "commit=%u"},
867 {Opt_journal_update, "journal=update"},
868 {Opt_journal_inum, "journal=%u"},
Johann Lombardi71b96252006-01-08 01:03:20 -0800869 {Opt_journal_dev, "journal_dev=%u"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 {Opt_abort, "abort"},
871 {Opt_data_journal, "data=journal"},
872 {Opt_data_ordered, "data=ordered"},
873 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700874 {Opt_data_err_abort, "data_err=abort"},
875 {Opt_data_err_ignore, "data_err=ignore"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 {Opt_offusrjquota, "usrjquota="},
877 {Opt_usrjquota, "usrjquota=%s"},
878 {Opt_offgrpjquota, "grpjquota="},
879 {Opt_grpjquota, "grpjquota=%s"},
880 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
881 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara1aeec432009-11-30 22:22:41 +0100882 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700883 {Opt_grpquota, "grpquota"},
Jan Kara1f545872005-06-23 22:01:04 -0700884 {Opt_noquota, "noquota"},
885 {Opt_quota, "quota"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700886 {Opt_usrquota, "usrquota"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 {Opt_barrier, "barrier=%u"},
Eric Sandeen0636c732010-04-30 11:09:34 -0500888 {Opt_barrier, "barrier"},
889 {Opt_nobarrier, "nobarrier"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 {Opt_resize, "resize"},
Josef Bacik92587212008-03-04 14:29:43 -0800891 {Opt_err, NULL},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892};
893
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100894static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895{
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700896 ext3_fsblk_t sb_block;
897 char *options = (char *) *data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
899 if (!options || strncmp(options, "sb=", 3) != 0)
900 return 1; /* Default location */
901 options += 3;
Mingming Cao43d23f92006-06-25 05:48:07 -0700902 /*todo: use simple_strtoll with >32bit ext3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 sb_block = simple_strtoul(options, &options, 0);
904 if (*options && *options != ',') {
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100905 ext3_msg(sb, "error: invalid sb specification: %s",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 (char *) *data);
907 return 1;
908 }
909 if (*options == ',')
910 options++;
911 *data = (void *) options;
912 return sb_block;
913}
914
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300915#ifdef CONFIG_QUOTA
916static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
917{
918 struct ext3_sb_info *sbi = EXT3_SB(sb);
919 char *qname;
920
921 if (sb_any_quota_loaded(sb) &&
922 !sbi->s_qf_names[qtype]) {
923 ext3_msg(sb, KERN_ERR,
924 "Cannot change journaled "
925 "quota options when quota turned on");
926 return 0;
927 }
928 qname = match_strdup(args);
929 if (!qname) {
930 ext3_msg(sb, KERN_ERR,
931 "Not enough memory for storing quotafile name");
932 return 0;
933 }
934 if (sbi->s_qf_names[qtype] &&
935 strcmp(sbi->s_qf_names[qtype], qname)) {
936 ext3_msg(sb, KERN_ERR,
937 "%s quota file already specified", QTYPE2NAME(qtype));
938 kfree(qname);
939 return 0;
940 }
941 sbi->s_qf_names[qtype] = qname;
942 if (strchr(sbi->s_qf_names[qtype], '/')) {
943 ext3_msg(sb, KERN_ERR,
944 "quotafile must be on filesystem root");
945 kfree(sbi->s_qf_names[qtype]);
946 sbi->s_qf_names[qtype] = NULL;
947 return 0;
948 }
949 set_opt(sbi->s_mount_opt, QUOTA);
950 return 1;
951}
952
953static int clear_qf_name(struct super_block *sb, int qtype) {
954
955 struct ext3_sb_info *sbi = EXT3_SB(sb);
956
957 if (sb_any_quota_loaded(sb) &&
958 sbi->s_qf_names[qtype]) {
959 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
960 " when quota turned on");
961 return 0;
962 }
963 /*
964 * The space will be released later when all options are confirmed
965 * to be correct
966 */
967 sbi->s_qf_names[qtype] = NULL;
968 return 1;
969}
970#endif
971
Johann Lombardi71b96252006-01-08 01:03:20 -0800972static int parse_options (char *options, struct super_block *sb,
Eric Sandeeneee194e2006-09-27 01:49:30 -0700973 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao43d23f92006-06-25 05:48:07 -0700974 ext3_fsblk_t *n_blocks_count, int is_remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975{
976 struct ext3_sb_info *sbi = EXT3_SB(sb);
977 char * p;
978 substring_t args[MAX_OPT_ARGS];
979 int data_opt = 0;
980 int option;
981#ifdef CONFIG_QUOTA
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300982 int qfmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983#endif
984
985 if (!options)
986 return 1;
987
988 while ((p = strsep (&options, ",")) != NULL) {
989 int token;
990 if (!*p)
991 continue;
Eric Sandeen0636c732010-04-30 11:09:34 -0500992 /*
993 * Initialize args struct so we know whether arg was
994 * found; some options take optional arguments.
995 */
996 args[0].to = args[0].from = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 token = match_token(p, tokens, args);
998 switch (token) {
999 case Opt_bsd_df:
1000 clear_opt (sbi->s_mount_opt, MINIX_DF);
1001 break;
1002 case Opt_minix_df:
1003 set_opt (sbi->s_mount_opt, MINIX_DF);
1004 break;
1005 case Opt_grpid:
1006 set_opt (sbi->s_mount_opt, GRPID);
1007 break;
1008 case Opt_nogrpid:
1009 clear_opt (sbi->s_mount_opt, GRPID);
1010 break;
1011 case Opt_resuid:
1012 if (match_int(&args[0], &option))
1013 return 0;
1014 sbi->s_resuid = option;
1015 break;
1016 case Opt_resgid:
1017 if (match_int(&args[0], &option))
1018 return 0;
1019 sbi->s_resgid = option;
1020 break;
1021 case Opt_sb:
1022 /* handled by get_sb_block() instead of here */
1023 /* *sb_block = match_int(&args[0]); */
1024 break;
1025 case Opt_err_panic:
1026 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1027 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1028 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1029 break;
1030 case Opt_err_ro:
1031 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1032 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1033 set_opt (sbi->s_mount_opt, ERRORS_RO);
1034 break;
1035 case Opt_err_cont:
1036 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1037 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1038 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1039 break;
1040 case Opt_nouid32:
1041 set_opt (sbi->s_mount_opt, NO_UID32);
1042 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 case Opt_nocheck:
1044 clear_opt (sbi->s_mount_opt, CHECK);
1045 break;
1046 case Opt_debug:
1047 set_opt (sbi->s_mount_opt, DEBUG);
1048 break;
1049 case Opt_oldalloc:
Lukas Czernerfbc85402011-08-16 18:08:06 +02001050 ext3_msg(sb, KERN_WARNING,
1051 "Ignoring deprecated oldalloc option");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 break;
1053 case Opt_orlov:
Lukas Czernerfbc85402011-08-16 18:08:06 +02001054 ext3_msg(sb, KERN_WARNING,
1055 "Ignoring deprecated orlov option");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 break;
1057#ifdef CONFIG_EXT3_FS_XATTR
1058 case Opt_user_xattr:
1059 set_opt (sbi->s_mount_opt, XATTR_USER);
1060 break;
1061 case Opt_nouser_xattr:
1062 clear_opt (sbi->s_mount_opt, XATTR_USER);
1063 break;
1064#else
1065 case Opt_user_xattr:
1066 case Opt_nouser_xattr:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001067 ext3_msg(sb, KERN_INFO,
1068 "(no)user_xattr options not supported");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 break;
1070#endif
1071#ifdef CONFIG_EXT3_FS_POSIX_ACL
1072 case Opt_acl:
1073 set_opt(sbi->s_mount_opt, POSIX_ACL);
1074 break;
1075 case Opt_noacl:
1076 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1077 break;
1078#else
1079 case Opt_acl:
1080 case Opt_noacl:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001081 ext3_msg(sb, KERN_INFO,
1082 "(no)acl options not supported");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 break;
1084#endif
1085 case Opt_reservation:
1086 set_opt(sbi->s_mount_opt, RESERVATION);
1087 break;
1088 case Opt_noreservation:
1089 clear_opt(sbi->s_mount_opt, RESERVATION);
1090 break;
1091 case Opt_journal_update:
1092 /* @@@ FIXME */
1093 /* Eventually we will want to be able to create
1094 a journal file here. For now, only allow the
1095 user to specify an existing inode to be the
1096 journal file. */
1097 if (is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001098 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1099 "journal on remount");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 return 0;
1101 }
1102 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1103 break;
1104 case Opt_journal_inum:
1105 if (is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001106 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1107 "journal on remount");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 return 0;
1109 }
1110 if (match_int(&args[0], &option))
1111 return 0;
1112 *inum = option;
1113 break;
Johann Lombardi71b96252006-01-08 01:03:20 -08001114 case Opt_journal_dev:
1115 if (is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001116 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1117 "journal on remount");
Johann Lombardi71b96252006-01-08 01:03:20 -08001118 return 0;
1119 }
1120 if (match_int(&args[0], &option))
1121 return 0;
1122 *journal_devnum = option;
1123 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 case Opt_noload:
1125 set_opt (sbi->s_mount_opt, NOLOAD);
1126 break;
1127 case Opt_commit:
1128 if (match_int(&args[0], &option))
1129 return 0;
1130 if (option < 0)
1131 return 0;
1132 if (option == 0)
1133 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1134 sbi->s_commit_interval = HZ * option;
1135 break;
1136 case Opt_data_journal:
1137 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1138 goto datacheck;
1139 case Opt_data_ordered:
1140 data_opt = EXT3_MOUNT_ORDERED_DATA;
1141 goto datacheck;
1142 case Opt_data_writeback:
1143 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1144 datacheck:
1145 if (is_remount) {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001146 if (test_opt(sb, DATA_FLAGS) == data_opt)
Jan Kara3c4cec62009-08-24 16:38:43 +02001147 break;
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001148 ext3_msg(sb, KERN_ERR,
1149 "error: cannot change "
Jan Kara3c4cec62009-08-24 16:38:43 +02001150 "data mode on remount. The filesystem "
1151 "is mounted in data=%s mode and you "
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001152 "try to remount it in data=%s mode.",
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001153 data_mode_string(test_opt(sb,
1154 DATA_FLAGS)),
Jan Kara3c4cec62009-08-24 16:38:43 +02001155 data_mode_string(data_opt));
1156 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 } else {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001158 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 sbi->s_mount_opt |= data_opt;
1160 }
1161 break;
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -07001162 case Opt_data_err_abort:
1163 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1164 break;
1165 case Opt_data_err_ignore:
1166 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1167 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168#ifdef CONFIG_QUOTA
1169 case Opt_usrjquota:
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001170 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1171 return 0;
1172 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 case Opt_grpjquota:
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001174 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 break;
1177 case Opt_offusrjquota:
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001178 if (!clear_qf_name(sb, USRQUOTA))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 return 0;
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001180 break;
1181 case Opt_offgrpjquota:
1182 if (!clear_qf_name(sb, GRPQUOTA))
1183 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 break;
1185 case Opt_jqfmt_vfsold:
Jan Karad06bf1d2008-07-25 01:46:18 -07001186 qfmt = QFMT_VFS_OLD;
1187 goto set_qf_format;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 case Opt_jqfmt_vfsv0:
Jan Karad06bf1d2008-07-25 01:46:18 -07001189 qfmt = QFMT_VFS_V0;
Jan Kara1aeec432009-11-30 22:22:41 +01001190 goto set_qf_format;
1191 case Opt_jqfmt_vfsv1:
1192 qfmt = QFMT_VFS_V1;
Jan Karad06bf1d2008-07-25 01:46:18 -07001193set_qf_format:
Jan Karaee0d5ff2008-08-20 18:11:50 +02001194 if (sb_any_quota_loaded(sb) &&
Jan Karad06bf1d2008-07-25 01:46:18 -07001195 sbi->s_jquota_fmt != qfmt) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001196 ext3_msg(sb, KERN_ERR, "error: cannot change "
Jan Karad06bf1d2008-07-25 01:46:18 -07001197 "journaled quota options when "
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001198 "quota turned on.");
Jan Karad06bf1d2008-07-25 01:46:18 -07001199 return 0;
1200 }
1201 sbi->s_jquota_fmt = qfmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 break;
Jan Kara1f545872005-06-23 22:01:04 -07001203 case Opt_quota:
Mark Bellon8fc27512005-09-06 15:16:54 -07001204 case Opt_usrquota:
Jan Kara1f545872005-06-23 22:01:04 -07001205 set_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001206 set_opt(sbi->s_mount_opt, USRQUOTA);
1207 break;
1208 case Opt_grpquota:
1209 set_opt(sbi->s_mount_opt, QUOTA);
1210 set_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001211 break;
1212 case Opt_noquota:
Jan Karaee0d5ff2008-08-20 18:11:50 +02001213 if (sb_any_quota_loaded(sb)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001214 ext3_msg(sb, KERN_ERR, "error: cannot change "
1215 "quota options when quota turned on.");
Jan Kara1f545872005-06-23 22:01:04 -07001216 return 0;
1217 }
1218 clear_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001219 clear_opt(sbi->s_mount_opt, USRQUOTA);
1220 clear_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001221 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222#else
Mark Bellon8fc27512005-09-06 15:16:54 -07001223 case Opt_quota:
1224 case Opt_usrquota:
1225 case Opt_grpquota:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001226 ext3_msg(sb, KERN_ERR,
1227 "error: quota options not supported.");
Jan Karafa1ff1e2008-04-28 02:16:14 -07001228 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 case Opt_usrjquota:
1230 case Opt_grpjquota:
1231 case Opt_offusrjquota:
1232 case Opt_offgrpjquota:
1233 case Opt_jqfmt_vfsold:
1234 case Opt_jqfmt_vfsv0:
Jan Kara1aeec432009-11-30 22:22:41 +01001235 case Opt_jqfmt_vfsv1:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001236 ext3_msg(sb, KERN_ERR,
1237 "error: journaled quota options not "
1238 "supported.");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 break;
Jan Kara1f545872005-06-23 22:01:04 -07001240 case Opt_noquota:
1241 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242#endif
1243 case Opt_abort:
1244 set_opt(sbi->s_mount_opt, ABORT);
1245 break;
Eric Sandeen0636c732010-04-30 11:09:34 -05001246 case Opt_nobarrier:
1247 clear_opt(sbi->s_mount_opt, BARRIER);
1248 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 case Opt_barrier:
Eric Sandeen0636c732010-04-30 11:09:34 -05001250 if (args[0].from) {
1251 if (match_int(&args[0], &option))
1252 return 0;
1253 } else
1254 option = 1; /* No argument, default to 1 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 if (option)
1256 set_opt(sbi->s_mount_opt, BARRIER);
1257 else
1258 clear_opt(sbi->s_mount_opt, BARRIER);
1259 break;
1260 case Opt_ignore:
1261 break;
1262 case Opt_resize:
Jan Kara08c6a962005-07-12 13:58:28 -07001263 if (!is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001264 ext3_msg(sb, KERN_ERR,
1265 "error: resize option only available "
1266 "for remount");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 return 0;
1268 }
KAMBAROV, ZAURc7f17212005-06-28 20:45:11 -07001269 if (match_int(&args[0], &option) != 0)
1270 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 *n_blocks_count = option;
1272 break;
1273 case Opt_nobh:
Christoph Hellwig4c4d3902010-06-07 10:20:39 +02001274 ext3_msg(sb, KERN_WARNING,
1275 "warning: ignoring deprecated nobh option");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 break;
Badari Pulavartyade1a292006-06-26 00:25:04 -07001277 case Opt_bh:
Christoph Hellwig4c4d3902010-06-07 10:20:39 +02001278 ext3_msg(sb, KERN_WARNING,
1279 "warning: ignoring deprecated bh option");
Badari Pulavartyade1a292006-06-26 00:25:04 -07001280 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 default:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001282 ext3_msg(sb, KERN_ERR,
1283 "error: unrecognized mount option \"%s\" "
1284 "or missing value", p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 return 0;
1286 }
1287 }
1288#ifdef CONFIG_QUOTA
Mark Bellon8fc27512005-09-06 15:16:54 -07001289 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001290 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Mark Bellon8fc27512005-09-06 15:16:54 -07001291 clear_opt(sbi->s_mount_opt, USRQUOTA);
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001292 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Mark Bellon8fc27512005-09-06 15:16:54 -07001293 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1294
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001295 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001296 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1297 "format mixing.");
Mark Bellon8fc27512005-09-06 15:16:54 -07001298 return 0;
1299 }
1300
1301 if (!sbi->s_jquota_fmt) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001302 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1303 "not specified.");
Mark Bellon8fc27512005-09-06 15:16:54 -07001304 return 0;
1305 }
1306 } else {
1307 if (sbi->s_jquota_fmt) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001308 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
Jan Kara99aeaf62008-07-25 01:46:17 -07001309 "specified with no journaling "
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001310 "enabled.");
Mark Bellon8fc27512005-09-06 15:16:54 -07001311 return 0;
1312 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 }
1314#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 return 1;
1316}
1317
1318static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1319 int read_only)
1320{
1321 struct ext3_sb_info *sbi = EXT3_SB(sb);
1322 int res = 0;
1323
1324 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001325 ext3_msg(sb, KERN_ERR,
1326 "error: revision level too high, "
1327 "forcing read-only mode");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 res = MS_RDONLY;
1329 }
1330 if (read_only)
1331 return res;
1332 if (!(sbi->s_mount_state & EXT3_VALID_FS))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001333 ext3_msg(sb, KERN_WARNING,
1334 "warning: mounting unchecked fs, "
1335 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001337 ext3_msg(sb, KERN_WARNING,
1338 "warning: mounting fs with errors, "
1339 "running e2fsck is recommended");
Namhyung Kimdf0d6b82010-10-10 21:36:59 +09001340 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 le16_to_cpu(es->s_mnt_count) >=
Namhyung Kimdf0d6b82010-10-10 21:36:59 +09001342 le16_to_cpu(es->s_max_mnt_count))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001343 ext3_msg(sb, KERN_WARNING,
1344 "warning: maximal mount count reached, "
1345 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 else if (le32_to_cpu(es->s_checkinterval) &&
1347 (le32_to_cpu(es->s_lastcheck) +
1348 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001349 ext3_msg(sb, KERN_WARNING,
1350 "warning: checktime reached, "
1351 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352#if 0
1353 /* @@@ We _will_ want to clear the valid bit if we find
1354 inconsistencies, to force a fsck at reboot. But for
1355 a plain journaled filesystem we can keep it set as
1356 valid forever! :) */
Marcin Slusarze7f23eb2008-04-28 02:16:06 -07001357 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358#endif
Namhyung Kimdf0d6b82010-10-10 21:36:59 +09001359 if (!le16_to_cpu(es->s_max_mnt_count))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
Marcin Slusarz50e8a282008-02-08 04:20:13 -08001361 le16_add_cpu(&es->s_mnt_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 es->s_mtime = cpu_to_le32(get_seconds());
1363 ext3_update_dynamic_rev(sb);
1364 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1365
1366 ext3_commit_super(sb, es, 1);
1367 if (test_opt(sb, DEBUG))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001368 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1369 "bpg=%lu, ipg=%lu, mo=%04lx]",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 sb->s_blocksize,
1371 sbi->s_groups_count,
1372 EXT3_BLOCKS_PER_GROUP(sb),
1373 EXT3_INODES_PER_GROUP(sb),
1374 sbi->s_mount_opt);
1375
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1377 char b[BDEVNAME_SIZE];
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001378 ext3_msg(sb, KERN_INFO, "using external journal on %s",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1380 } else {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001381 ext3_msg(sb, KERN_INFO, "using internal journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 }
Dan Magenheimerd71bc6d2011-05-26 10:01:49 -06001383 cleancache_init_fs(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 return res;
1385}
1386
1387/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001388static int ext3_check_descriptors(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389{
1390 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 int i;
1392
1393 ext3_debug ("Checking group descriptors");
1394
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001395 for (i = 0; i < sbi->s_groups_count; i++) {
1396 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
Akinobu Mita1eaafea2008-04-28 02:16:07 -07001397 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1398 ext3_fsblk_t last_block;
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001399
Eric Sandeen855565e2006-09-27 01:49:29 -07001400 if (i == sbi->s_groups_count - 1)
1401 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1402 else
1403 last_block = first_block +
1404 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1405
Eric Sandeen855565e2006-09-27 01:49:29 -07001406 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1407 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 {
1409 ext3_error (sb, "ext3_check_descriptors",
1410 "Block bitmap for group %d"
1411 " not in group (block %lu)!",
1412 i, (unsigned long)
1413 le32_to_cpu(gdp->bg_block_bitmap));
1414 return 0;
1415 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001416 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1417 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 {
1419 ext3_error (sb, "ext3_check_descriptors",
1420 "Inode bitmap for group %d"
1421 " not in group (block %lu)!",
1422 i, (unsigned long)
1423 le32_to_cpu(gdp->bg_inode_bitmap));
1424 return 0;
1425 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001426 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
Eric Sandeen780dcdb2007-07-26 10:41:11 -07001427 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
Eric Sandeen855565e2006-09-27 01:49:29 -07001428 last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 {
1430 ext3_error (sb, "ext3_check_descriptors",
1431 "Inode table for group %d"
1432 " not in group (block %lu)!",
1433 i, (unsigned long)
1434 le32_to_cpu(gdp->bg_inode_table));
1435 return 0;
1436 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 }
1438
1439 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1440 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1441 return 1;
1442}
1443
1444
1445/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1446 * the superblock) which were deleted from all directories, but held open by
1447 * a process at the time of a crash. We walk the list and try to delete these
1448 * inodes at recovery time (only with a read-write filesystem).
1449 *
1450 * In order to keep the orphan inode chain consistent during traversal (in
1451 * case of crash during recovery), we link each inode into the superblock
1452 * orphan list_head and handle it the same way as an inode deletion during
1453 * normal operation (which journals the operations for us).
1454 *
1455 * We only do an iget() and an iput() on each inode, which is very safe if we
1456 * accidentally point at an in-use or already deleted inode. The worst that
1457 * can happen in this case is that we get a "bit already cleared" message from
1458 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1459 * e2fsck was run on this filesystem, and it must have already done the orphan
1460 * inode cleanup for us, so we can safely abort without any further action.
1461 */
1462static void ext3_orphan_cleanup (struct super_block * sb,
1463 struct ext3_super_block * es)
1464{
1465 unsigned int s_flags = sb->s_flags;
1466 int nr_orphans = 0, nr_truncates = 0;
1467#ifdef CONFIG_QUOTA
1468 int i;
1469#endif
1470 if (!es->s_last_orphan) {
1471 jbd_debug(4, "no orphan inodes to clean up\n");
1472 return;
1473 }
1474
Eric Sandeena8f48a92006-12-06 20:40:13 -08001475 if (bdev_read_only(sb->s_bdev)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001476 ext3_msg(sb, KERN_ERR, "error: write access "
1477 "unavailable, skipping orphan cleanup.");
Eric Sandeena8f48a92006-12-06 20:40:13 -08001478 return;
1479 }
1480
Amir Goldsteince654b32011-02-26 22:40:19 +02001481 /* Check if feature set allows readwrite operations */
1482 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1483 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1484 "unknown ROCOMPAT features");
1485 return;
1486 }
1487
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1489 if (es->s_last_orphan)
1490 jbd_debug(1, "Errors on filesystem, "
1491 "clearing orphan list.\n");
1492 es->s_last_orphan = 0;
1493 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1494 return;
1495 }
1496
1497 if (s_flags & MS_RDONLY) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001498 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 sb->s_flags &= ~MS_RDONLY;
1500 }
1501#ifdef CONFIG_QUOTA
1502 /* Needed for iput() to work correctly and not trash data */
1503 sb->s_flags |= MS_ACTIVE;
1504 /* Turn on quotas so that they are updated correctly */
1505 for (i = 0; i < MAXQUOTAS; i++) {
1506 if (EXT3_SB(sb)->s_qf_names[i]) {
1507 int ret = ext3_quota_on_mount(sb, i);
1508 if (ret < 0)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001509 ext3_msg(sb, KERN_ERR,
1510 "error: cannot turn on journaled "
1511 "quota: %d", ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 }
1513 }
1514#endif
1515
1516 while (es->s_last_orphan) {
1517 struct inode *inode;
1518
David Howells473043d2008-02-07 00:15:36 -08001519 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1520 if (IS_ERR(inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 es->s_last_orphan = 0;
1522 break;
1523 }
1524
1525 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05001526 dquot_initialize(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 if (inode->i_nlink) {
1528 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001529 "%s: truncating inode %lu to %Ld bytes\n",
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001530 __func__, inode->i_ino, inode->i_size);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001531 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 inode->i_ino, inode->i_size);
1533 ext3_truncate(inode);
1534 nr_truncates++;
1535 } else {
1536 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001537 "%s: deleting unreferenced inode %lu\n",
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001538 __func__, inode->i_ino);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001539 jbd_debug(2, "deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 inode->i_ino);
1541 nr_orphans++;
1542 }
1543 iput(inode); /* The delete magic happens here! */
1544 }
1545
1546#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1547
1548 if (nr_orphans)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001549 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1550 PLURAL(nr_orphans));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 if (nr_truncates)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001552 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1553 PLURAL(nr_truncates));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554#ifdef CONFIG_QUOTA
1555 /* Turn quotas off */
1556 for (i = 0; i < MAXQUOTAS; i++) {
1557 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04001558 dquot_quota_off(sb, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 }
1560#endif
1561 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1562}
1563
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564/*
1565 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1566 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1567 * We need to be 1 filesystem block less than the 2^32 sector limit.
1568 */
1569static loff_t ext3_max_size(int bits)
1570{
1571 loff_t res = EXT3_NDIR_BLOCKS;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001572 int meta_blocks;
1573 loff_t upper_limit;
1574
1575 /* This is calculated to be the largest file size for a
1576 * dense, file such that the total number of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 * sectors in the file, including data and all indirect blocks,
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001578 * does not exceed 2^32 -1
1579 * __u32 i_blocks representing the total number of
1580 * 512 bytes blocks of the file
1581 */
1582 upper_limit = (1LL << 32) - 1;
1583
1584 /* total blocks in file system block size */
1585 upper_limit >>= (bits - 9);
1586
1587
1588 /* indirect blocks */
1589 meta_blocks = 1;
1590 /* double indirect blocks */
1591 meta_blocks += 1 + (1LL << (bits-2));
1592 /* tripple indirect blocks */
1593 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1594
1595 upper_limit -= meta_blocks;
1596 upper_limit <<= bits;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
1598 res += 1LL << (bits-2);
1599 res += 1LL << (2*(bits-2));
1600 res += 1LL << (3*(bits-2));
1601 res <<= bits;
1602 if (res > upper_limit)
1603 res = upper_limit;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001604
1605 if (res > MAX_LFS_FILESIZE)
1606 res = MAX_LFS_FILESIZE;
1607
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 return res;
1609}
1610
Mingming Cao43d23f92006-06-25 05:48:07 -07001611static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1612 ext3_fsblk_t logic_sb_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 int nr)
1614{
1615 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001616 unsigned long bg, first_meta_bg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617 int has_super = 0;
1618
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1620
1621 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1622 nr < first_meta_bg)
1623 return (logic_sb_block + nr + 1);
1624 bg = sbi->s_desc_per_block * nr;
1625 if (ext3_bg_has_super(sb, bg))
1626 has_super = 1;
Mingming Cao43d23f92006-06-25 05:48:07 -07001627 return (has_super + ext3_group_first_block_no(sb, bg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628}
1629
1630
1631static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1632{
1633 struct buffer_head * bh;
1634 struct ext3_super_block *es = NULL;
1635 struct ext3_sb_info *sbi;
Mingming Cao43d23f92006-06-25 05:48:07 -07001636 ext3_fsblk_t block;
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001637 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001638 ext3_fsblk_t logic_sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 unsigned long offset = 0;
Eric Sandeeneee194e2006-09-27 01:49:30 -07001640 unsigned int journal_inum = 0;
Johann Lombardi71b96252006-01-08 01:03:20 -08001641 unsigned long journal_devnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 unsigned long def_mount_opts;
1643 struct inode *root;
1644 int blocksize;
1645 int hblock;
1646 int db_count;
1647 int i;
1648 int needs_recovery;
David Howells473043d2008-02-07 00:15:36 -08001649 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 __le32 features;
Peter Zijlstra833f4072007-10-16 23:25:45 -07001651 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652
Panagiotis Issarisf8314dc2006-09-27 01:49:37 -07001653 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 if (!sbi)
1655 return -ENOMEM;
Pekka Enberg5df096d2009-01-07 18:07:25 -08001656
1657 sbi->s_blockgroup_lock =
1658 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1659 if (!sbi->s_blockgroup_lock) {
1660 kfree(sbi);
1661 return -ENOMEM;
1662 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 sb->s_fs_info = sbi;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 sbi->s_mount_opt = 0;
1665 sbi->s_resuid = EXT3_DEF_RESUID;
1666 sbi->s_resgid = EXT3_DEF_RESGID;
Miklos Szeredi571beed2007-10-16 23:26:26 -07001667 sbi->s_sb_block = sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1670 if (!blocksize) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001671 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 goto out_fail;
1673 }
1674
1675 /*
1676 * The ext3 superblock will not be buffer aligned for other than 1kB
1677 * block sizes. We need to calculate the offset from buffer start.
1678 */
1679 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1680 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1681 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1682 } else {
1683 logic_sb_block = sb_block;
1684 }
1685
1686 if (!(bh = sb_bread(sb, logic_sb_block))) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001687 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 goto out_fail;
1689 }
1690 /*
1691 * Note: s_es must be initialized as soon as possible because
1692 * some ext3 macro-instructions depend on its value
1693 */
Namhyung Kim57e94d82010-10-11 02:10:45 +09001694 es = (struct ext3_super_block *) (bh->b_data + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 sbi->s_es = es;
1696 sb->s_magic = le16_to_cpu(es->s_magic);
1697 if (sb->s_magic != EXT3_SUPER_MAGIC)
1698 goto cantfind_ext3;
1699
1700 /* Set defaults before we parse the mount options */
1701 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1702 if (def_mount_opts & EXT3_DEFM_DEBUG)
1703 set_opt(sbi->s_mount_opt, DEBUG);
1704 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1705 set_opt(sbi->s_mount_opt, GRPID);
1706 if (def_mount_opts & EXT3_DEFM_UID16)
1707 set_opt(sbi->s_mount_opt, NO_UID32);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001708#ifdef CONFIG_EXT3_FS_XATTR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1710 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001711#endif
1712#ifdef CONFIG_EXT3_FS_POSIX_ACL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 if (def_mount_opts & EXT3_DEFM_ACL)
1714 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001715#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001717 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001719 set_opt(sbi->s_mount_opt, ORDERED_DATA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001721 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722
1723 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1724 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001725 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
Dmitry Mishin2245d7c2006-10-11 01:21:49 -07001726 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001727 else
1728 set_opt(sbi->s_mount_opt, ERRORS_RO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729
1730 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1731 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1732
Christoph Hellwig00eacd62011-07-16 16:46:50 -04001733 /* enable barriers by default */
1734 set_opt(sbi->s_mount_opt, BARRIER);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 set_opt(sbi->s_mount_opt, RESERVATION);
1736
Johann Lombardi71b96252006-01-08 01:03:20 -08001737 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1738 NULL, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 goto failed_mount;
1740
1741 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001742 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
1744 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1745 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1746 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1747 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001748 ext3_msg(sb, KERN_WARNING,
1749 "warning: feature flags set on rev 0 fs, "
1750 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 /*
1752 * Check feature flags regardless of the revision level, since we
1753 * previously didn't change the revision level when setting the flags,
1754 * so there is a chance incompat flags are set on a rev 0 filesystem.
1755 */
1756 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1757 if (features) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001758 ext3_msg(sb, KERN_ERR,
1759 "error: couldn't mount because of unsupported "
1760 "optional features (%x)", le32_to_cpu(features));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 goto failed_mount;
1762 }
1763 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1764 if (!(sb->s_flags & MS_RDONLY) && features) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001765 ext3_msg(sb, KERN_ERR,
1766 "error: couldn't mount RDWR because of unsupported "
1767 "optional features (%x)", le32_to_cpu(features));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 goto failed_mount;
1769 }
1770 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1771
1772 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1773 blocksize > EXT3_MAX_BLOCK_SIZE) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001774 ext3_msg(sb, KERN_ERR,
1775 "error: couldn't mount because of unsupported "
1776 "filesystem blocksize %d", blocksize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 goto failed_mount;
1778 }
1779
Martin K. Petersene1defc42009-05-22 17:17:49 -04001780 hblock = bdev_logical_block_size(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 if (sb->s_blocksize != blocksize) {
1782 /*
1783 * Make sure the blocksize for the filesystem is larger
1784 * than the hardware sectorsize for the machine.
1785 */
1786 if (blocksize < hblock) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001787 ext3_msg(sb, KERN_ERR,
1788 "error: fsblocksize %d too small for "
1789 "hardware sectorsize %d", blocksize, hblock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 goto failed_mount;
1791 }
1792
1793 brelse (bh);
Takashi Sato0f0a89e2007-10-18 03:06:56 -07001794 if (!sb_set_blocksize(sb, blocksize)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001795 ext3_msg(sb, KERN_ERR,
1796 "error: bad blocksize %d", blocksize);
Takashi Sato0f0a89e2007-10-18 03:06:56 -07001797 goto out_fail;
1798 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1800 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1801 bh = sb_bread(sb, logic_sb_block);
1802 if (!bh) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001803 ext3_msg(sb, KERN_ERR,
1804 "error: can't read superblock on 2nd try");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 goto failed_mount;
1806 }
Namhyung Kim57e94d82010-10-11 02:10:45 +09001807 es = (struct ext3_super_block *)(bh->b_data + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 sbi->s_es = es;
1809 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001810 ext3_msg(sb, KERN_ERR,
1811 "error: magic mismatch");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 goto failed_mount;
1813 }
1814 }
1815
1816 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1817
1818 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1819 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1820 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1821 } else {
1822 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1823 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1824 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
vignesh babu3fc74262007-07-15 23:41:17 -07001825 (!is_power_of_2(sbi->s_inode_size)) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 (sbi->s_inode_size > blocksize)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001827 ext3_msg(sb, KERN_ERR,
1828 "error: unsupported inode size: %d",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 sbi->s_inode_size);
1830 goto failed_mount;
1831 }
1832 }
1833 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1834 le32_to_cpu(es->s_log_frag_size);
1835 if (blocksize != sbi->s_frag_size) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001836 ext3_msg(sb, KERN_ERR,
1837 "error: fragsize %lu != blocksize %u (unsupported)",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 sbi->s_frag_size, blocksize);
1839 goto failed_mount;
1840 }
1841 sbi->s_frags_per_block = 1;
1842 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1843 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1844 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08001845 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 goto cantfind_ext3;
1847 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1848 if (sbi->s_inodes_per_block == 0)
1849 goto cantfind_ext3;
1850 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1851 sbi->s_inodes_per_block;
1852 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1853 sbi->s_sbh = bh;
1854 sbi->s_mount_state = le16_to_cpu(es->s_state);
David Howellsf0d1b0b2006-12-08 02:37:49 -08001855 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1856 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 for (i=0; i < 4; i++)
1858 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1859 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'o5e1f8c92008-10-28 13:21:55 -04001860 i = le32_to_cpu(es->s_flags);
1861 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1862 sbi->s_hash_unsigned = 3;
1863 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1864#ifdef __CHAR_UNSIGNED__
1865 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1866 sbi->s_hash_unsigned = 3;
1867#else
1868 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1869#endif
Theodore Ts'o5e1f8c92008-10-28 13:21:55 -04001870 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871
1872 if (sbi->s_blocks_per_group > blocksize * 8) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001873 ext3_msg(sb, KERN_ERR,
1874 "#blocks per group too big: %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 sbi->s_blocks_per_group);
1876 goto failed_mount;
1877 }
1878 if (sbi->s_frags_per_group > blocksize * 8) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001879 ext3_msg(sb, KERN_ERR,
1880 "error: #fragments per group too big: %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 sbi->s_frags_per_group);
1882 goto failed_mount;
1883 }
1884 if (sbi->s_inodes_per_group > blocksize * 8) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001885 ext3_msg(sb, KERN_ERR,
1886 "error: #inodes per group too big: %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 sbi->s_inodes_per_group);
1888 goto failed_mount;
1889 }
1890
Darrick J. Wongad692bf2010-11-16 22:57:44 +00001891 err = generic_check_addressable(sb->s_blocksize_bits,
1892 le32_to_cpu(es->s_blocks_count));
1893 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001894 ext3_msg(sb, KERN_ERR,
1895 "error: filesystem is too large to mount safely");
Mingming Caofcd5df32006-06-25 05:47:50 -07001896 if (sizeof(sector_t) < 8)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001897 ext3_msg(sb, KERN_ERR,
1898 "error: CONFIG_LBDAF not enabled");
Darrick J. Wongad692bf2010-11-16 22:57:44 +00001899 ret = err;
Mingming Caofcd5df32006-06-25 05:47:50 -07001900 goto failed_mount;
1901 }
1902
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1904 goto cantfind_ext3;
Eric Sandeen855565e2006-09-27 01:49:29 -07001905 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1906 le32_to_cpu(es->s_first_data_block) - 1)
1907 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
Namhyung Kim81a4e322010-10-11 19:38:39 +09001908 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1910 GFP_KERNEL);
1911 if (sbi->s_group_desc == NULL) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001912 ext3_msg(sb, KERN_ERR,
1913 "error: not enough memory");
Namhyung Kim4569cd12010-10-11 19:08:06 +09001914 ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 goto failed_mount;
1916 }
1917
Pekka Enberg5df096d2009-01-07 18:07:25 -08001918 bgl_lock_init(sbi->s_blockgroup_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
1920 for (i = 0; i < db_count; i++) {
1921 block = descriptor_loc(sb, logic_sb_block, i);
1922 sbi->s_group_desc[i] = sb_bread(sb, block);
1923 if (!sbi->s_group_desc[i]) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001924 ext3_msg(sb, KERN_ERR,
1925 "error: can't read group descriptor %d", i);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 db_count = i;
1927 goto failed_mount2;
1928 }
1929 }
1930 if (!ext3_check_descriptors (sb)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001931 ext3_msg(sb, KERN_ERR,
1932 "error: group descriptors corrupted");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 goto failed_mount2;
1934 }
1935 sbi->s_gdb_count = db_count;
1936 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1937 spin_lock_init(&sbi->s_next_gen_lock);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001938
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 /* per fileystem reservation list head & lock */
1940 spin_lock_init(&sbi->s_rsv_window_lock);
1941 sbi->s_rsv_window_root = RB_ROOT;
1942 /* Add a single, static dummy reservation to the start of the
1943 * reservation window list --- it gives us a placeholder for
1944 * append-at-start-of-list which makes the allocation logic
1945 * _much_ simpler. */
1946 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1947 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1948 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1949 sbi->s_rsv_window_head.rsv_goal_size = 0;
1950 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1951
1952 /*
1953 * set up enough so that it can read an inode
1954 */
1955 sb->s_op = &ext3_sops;
1956 sb->s_export_op = &ext3_export_ops;
1957 sb->s_xattr = ext3_xattr_handlers;
1958#ifdef CONFIG_QUOTA
1959 sb->s_qcop = &ext3_qctl_operations;
1960 sb->dq_op = &ext3_quota_operations;
1961#endif
Aneesh Kumar K.V03cb5f02011-01-29 18:43:39 +05301962 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Eric Sandeenb8a052d2009-12-14 13:00:30 -06001964 mutex_init(&sbi->s_orphan_lock);
Eric Sandeen96d2a492009-12-14 13:01:05 -06001965 mutex_init(&sbi->s_resize_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966
1967 sb->s_root = NULL;
1968
1969 needs_recovery = (es->s_last_orphan != 0 ||
1970 EXT3_HAS_INCOMPAT_FEATURE(sb,
1971 EXT3_FEATURE_INCOMPAT_RECOVER));
1972
1973 /*
1974 * The first inode we look at is the journal inode. Don't try
1975 * root first: it may be modified in the journal!
1976 */
1977 if (!test_opt(sb, NOLOAD) &&
1978 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
Johann Lombardi71b96252006-01-08 01:03:20 -08001979 if (ext3_load_journal(sb, es, journal_devnum))
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04001980 goto failed_mount2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 } else if (journal_inum) {
1982 if (ext3_create_journal(sb, es, journal_inum))
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04001983 goto failed_mount2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 } else {
1985 if (!silent)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001986 ext3_msg(sb, KERN_ERR,
1987 "error: no journal found. "
1988 "mounting ext3 over ext2?");
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04001989 goto failed_mount2;
1990 }
1991 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1992 ext3_count_free_blocks(sb));
1993 if (!err) {
1994 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1995 ext3_count_free_inodes(sb));
1996 }
1997 if (!err) {
1998 err = percpu_counter_init(&sbi->s_dirs_counter,
1999 ext3_count_dirs(sb));
2000 }
2001 if (err) {
2002 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
Namhyung Kim4569cd12010-10-11 19:08:06 +09002003 ret = err;
Mingming Cao0216bfc2006-06-23 02:05:41 -07002004 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 }
2006
2007 /* We have now updated the journal if required, so we can
2008 * validate the data journaling mode. */
2009 switch (test_opt(sb, DATA_FLAGS)) {
2010 case 0:
2011 /* No mode set, assume a default based on the journal
2012 capabilities: ORDERED_DATA if the journal can
2013 cope, else JOURNAL_DATA */
2014 if (journal_check_available_features
2015 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
Linus Torvaldsbbae8bc2009-04-06 17:16:47 -07002016 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 else
2018 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2019 break;
2020
2021 case EXT3_MOUNT_ORDERED_DATA:
2022 case EXT3_MOUNT_WRITEBACK_DATA:
2023 if (!journal_check_available_features
2024 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002025 ext3_msg(sb, KERN_ERR,
2026 "error: journal does not support "
2027 "requested data journaling mode");
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002028 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 }
2030 default:
2031 break;
2032 }
2033
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 /*
2035 * The journal_load will have done any necessary log recovery,
2036 * so we can safely mount the rest of the filesystem now.
2037 */
2038
David Howells473043d2008-02-07 00:15:36 -08002039 root = ext3_iget(sb, EXT3_ROOT_INO);
2040 if (IS_ERR(root)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002041 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
David Howells473043d2008-02-07 00:15:36 -08002042 ret = PTR_ERR(root);
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002043 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 }
2045 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells473043d2008-02-07 00:15:36 -08002046 iput(root);
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002047 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002048 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 }
David Howells473043d2008-02-07 00:15:36 -08002050 sb->s_root = d_alloc_root(root);
2051 if (!sb->s_root) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002052 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
David Howells473043d2008-02-07 00:15:36 -08002053 iput(root);
2054 ret = -ENOMEM;
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002055 goto failed_mount3;
David Howells473043d2008-02-07 00:15:36 -08002056 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057
2058 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
Eric Sandeened505ee2009-12-14 12:59:18 -06002059
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2061 ext3_orphan_cleanup(sb, es);
2062 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2063 if (needs_recovery)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002064 ext3_msg(sb, KERN_INFO, "recovery complete");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 ext3_mark_recovery_complete(sb, es);
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002066 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2068 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2069 "writeback");
2070
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 return 0;
2072
2073cantfind_ext3:
2074 if (!silent)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002075 ext3_msg(sb, KERN_INFO,
2076 "error: can't find ext3 filesystem on dev %s.",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077 sb->s_id);
2078 goto failed_mount;
2079
Mingming Cao0216bfc2006-06-23 02:05:41 -07002080failed_mount3:
2081 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2082 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2083 percpu_counter_destroy(&sbi->s_dirs_counter);
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002084 journal_destroy(sbi->s_journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085failed_mount2:
2086 for (i = 0; i < db_count; i++)
2087 brelse(sbi->s_group_desc[i]);
2088 kfree(sbi->s_group_desc);
2089failed_mount:
2090#ifdef CONFIG_QUOTA
2091 for (i = 0; i < MAXQUOTAS; i++)
2092 kfree(sbi->s_qf_names[i]);
2093#endif
2094 ext3_blkdev_remove(sbi);
2095 brelse(bh);
2096out_fail:
2097 sb->s_fs_info = NULL;
Manish Katiyarde5ce032009-05-17 23:52:47 -04002098 kfree(sbi->s_blockgroup_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 kfree(sbi);
David Howells473043d2008-02-07 00:15:36 -08002100 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101}
2102
2103/*
2104 * Setup any per-fs journal parameters now. We'll do this both on
2105 * initial mount, once the journal has been initialised but before we've
Mingming Caoae6ddcc2006-09-27 01:49:27 -07002106 * done any recovery; and again on any subsequent remount.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 */
2108static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2109{
2110 struct ext3_sb_info *sbi = EXT3_SB(sb);
2111
2112 if (sbi->s_commit_interval)
2113 journal->j_commit_interval = sbi->s_commit_interval;
2114 /* We could also set up an ext3-specific default for the commit
2115 * interval here, but for now we'll just fall back to the jbd
2116 * default. */
2117
2118 spin_lock(&journal->j_state_lock);
2119 if (test_opt(sb, BARRIER))
2120 journal->j_flags |= JFS_BARRIER;
2121 else
2122 journal->j_flags &= ~JFS_BARRIER;
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -07002123 if (test_opt(sb, DATA_ERR_ABORT))
2124 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2125 else
2126 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 spin_unlock(&journal->j_state_lock);
2128}
2129
Eric Sandeeneee194e2006-09-27 01:49:30 -07002130static journal_t *ext3_get_journal(struct super_block *sb,
2131 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132{
2133 struct inode *journal_inode;
2134 journal_t *journal;
2135
2136 /* First, test for the existence of a valid inode on disk. Bad
2137 * things happen if we iget() an unused inode, as the subsequent
2138 * iput() will try to delete it. */
2139
David Howells473043d2008-02-07 00:15:36 -08002140 journal_inode = ext3_iget(sb, journal_inum);
2141 if (IS_ERR(journal_inode)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002142 ext3_msg(sb, KERN_ERR, "error: no journal found");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 return NULL;
2144 }
2145 if (!journal_inode->i_nlink) {
2146 make_bad_inode(journal_inode);
2147 iput(journal_inode);
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002148 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 return NULL;
2150 }
2151
2152 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2153 journal_inode, journal_inode->i_size);
David Howells473043d2008-02-07 00:15:36 -08002154 if (!S_ISREG(journal_inode->i_mode)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002155 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 iput(journal_inode);
2157 return NULL;
2158 }
2159
2160 journal = journal_init_inode(journal_inode);
2161 if (!journal) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002162 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163 iput(journal_inode);
2164 return NULL;
2165 }
2166 journal->j_private = sb;
2167 ext3_init_journal_params(sb, journal);
2168 return journal;
2169}
2170
2171static journal_t *ext3_get_dev_journal(struct super_block *sb,
2172 dev_t j_dev)
2173{
2174 struct buffer_head * bh;
2175 journal_t *journal;
Mingming Cao43d23f92006-06-25 05:48:07 -07002176 ext3_fsblk_t start;
Mingming Cao1c2bf372006-06-25 05:48:06 -07002177 ext3_fsblk_t len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 int hblock, blocksize;
Mingming Cao43d23f92006-06-25 05:48:07 -07002179 ext3_fsblk_t sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180 unsigned long offset;
2181 struct ext3_super_block * es;
2182 struct block_device *bdev;
2183
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002184 bdev = ext3_blkdev_get(j_dev, sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185 if (bdev == NULL)
2186 return NULL;
2187
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04002189 hblock = bdev_logical_block_size(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 if (blocksize < hblock) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002191 ext3_msg(sb, KERN_ERR,
2192 "error: blocksize too small for journal device");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 goto out_bdev;
2194 }
2195
2196 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2197 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2198 set_blocksize(bdev, blocksize);
2199 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002200 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2201 "external journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 goto out_bdev;
2203 }
2204
Namhyung Kim57e94d82010-10-11 02:10:45 +09002205 es = (struct ext3_super_block *) (bh->b_data + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2207 !(le32_to_cpu(es->s_feature_incompat) &
2208 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002209 ext3_msg(sb, KERN_ERR, "error: external journal has "
2210 "bad superblock");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 brelse(bh);
2212 goto out_bdev;
2213 }
2214
2215 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002216 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 brelse(bh);
2218 goto out_bdev;
2219 }
2220
2221 len = le32_to_cpu(es->s_blocks_count);
2222 start = sb_block + 1;
2223 brelse(bh); /* we're done with the superblock */
2224
2225 journal = journal_init_dev(bdev, sb->s_bdev,
2226 start, len, blocksize);
2227 if (!journal) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002228 ext3_msg(sb, KERN_ERR,
2229 "error: failed to create device journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230 goto out_bdev;
2231 }
2232 journal->j_private = sb;
2233 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2234 wait_on_buffer(journal->j_sb_buffer);
2235 if (!buffer_uptodate(journal->j_sb_buffer)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002236 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237 goto out_journal;
2238 }
2239 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002240 ext3_msg(sb, KERN_ERR,
2241 "error: external journal has more than one "
2242 "user (unsupported) - %d",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243 be32_to_cpu(journal->j_superblock->s_nr_users));
2244 goto out_journal;
2245 }
2246 EXT3_SB(sb)->journal_bdev = bdev;
2247 ext3_init_journal_params(sb, journal);
2248 return journal;
2249out_journal:
2250 journal_destroy(journal);
2251out_bdev:
2252 ext3_blkdev_put(bdev);
2253 return NULL;
2254}
2255
Johann Lombardi71b96252006-01-08 01:03:20 -08002256static int ext3_load_journal(struct super_block *sb,
2257 struct ext3_super_block *es,
2258 unsigned long journal_devnum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259{
2260 journal_t *journal;
Eric Sandeeneee194e2006-09-27 01:49:30 -07002261 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
Johann Lombardi71b96252006-01-08 01:03:20 -08002262 dev_t journal_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263 int err = 0;
2264 int really_read_only;
2265
Johann Lombardi71b96252006-01-08 01:03:20 -08002266 if (journal_devnum &&
2267 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002268 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2269 "numbers have changed");
Johann Lombardi71b96252006-01-08 01:03:20 -08002270 journal_dev = new_decode_dev(journal_devnum);
2271 } else
2272 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2273
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 really_read_only = bdev_read_only(sb->s_bdev);
2275
2276 /*
2277 * Are we loading a blank journal or performing recovery after a
2278 * crash? For recovery, we need to check in advance whether we
2279 * can get read-write access to the device.
2280 */
2281
2282 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2283 if (sb->s_flags & MS_RDONLY) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002284 ext3_msg(sb, KERN_INFO,
2285 "recovery required on readonly filesystem");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 if (really_read_only) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002287 ext3_msg(sb, KERN_ERR, "error: write access "
2288 "unavailable, cannot proceed");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 return -EROFS;
2290 }
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002291 ext3_msg(sb, KERN_INFO,
2292 "write access will be enabled during recovery");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293 }
2294 }
2295
2296 if (journal_inum && journal_dev) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002297 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2298 "and inode journals");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299 return -EINVAL;
2300 }
2301
2302 if (journal_inum) {
2303 if (!(journal = ext3_get_journal(sb, journal_inum)))
2304 return -EINVAL;
2305 } else {
2306 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2307 return -EINVAL;
2308 }
2309
Eric Sandeen0636c732010-04-30 11:09:34 -05002310 if (!(journal->j_flags & JFS_BARRIER))
2311 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2312
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2314 err = journal_update_format(journal);
2315 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002316 ext3_msg(sb, KERN_ERR, "error updating journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 journal_destroy(journal);
2318 return err;
2319 }
2320 }
2321
2322 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2323 err = journal_wipe(journal, !really_read_only);
2324 if (!err)
2325 err = journal_load(journal);
2326
2327 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002328 ext3_msg(sb, KERN_ERR, "error loading journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 journal_destroy(journal);
2330 return err;
2331 }
2332
2333 EXT3_SB(sb)->s_journal = journal;
2334 ext3_clear_journal_err(sb, es);
Johann Lombardi71b96252006-01-08 01:03:20 -08002335
Maciej Żenczykowski31d710a2010-09-26 12:38:28 +00002336 if (!really_read_only && journal_devnum &&
Johann Lombardi71b96252006-01-08 01:03:20 -08002337 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2338 es->s_journal_dev = cpu_to_le32(journal_devnum);
Johann Lombardi71b96252006-01-08 01:03:20 -08002339
2340 /* Make sure we flush the recovery flag to disk. */
2341 ext3_commit_super(sb, es, 1);
2342 }
2343
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 return 0;
2345}
2346
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002347static int ext3_create_journal(struct super_block *sb,
2348 struct ext3_super_block *es,
Eric Sandeeneee194e2006-09-27 01:49:30 -07002349 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350{
2351 journal_t *journal;
Borislav Petkov952d9de2007-07-15 23:41:45 -07002352 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353
2354 if (sb->s_flags & MS_RDONLY) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002355 ext3_msg(sb, KERN_ERR,
2356 "error: readonly filesystem when trying to "
2357 "create journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 return -EROFS;
2359 }
2360
Borislav Petkov952d9de2007-07-15 23:41:45 -07002361 journal = ext3_get_journal(sb, journal_inum);
2362 if (!journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363 return -EINVAL;
2364
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002365 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366 journal_inum);
2367
Borislav Petkov952d9de2007-07-15 23:41:45 -07002368 err = journal_create(journal);
2369 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002370 ext3_msg(sb, KERN_ERR, "error creating journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 journal_destroy(journal);
2372 return -EIO;
2373 }
2374
2375 EXT3_SB(sb)->s_journal = journal;
2376
2377 ext3_update_dynamic_rev(sb);
2378 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2379 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2380
2381 es->s_journal_inum = cpu_to_le32(journal_inum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382
2383 /* Make sure we flush the recovery flag to disk. */
2384 ext3_commit_super(sb, es, 1);
2385
2386 return 0;
2387}
2388
Takashi Satoc4be0c12009-01-09 16:40:58 -08002389static int ext3_commit_super(struct super_block *sb,
2390 struct ext3_super_block *es,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391 int sync)
2392{
2393 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08002394 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395
2396 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08002397 return error;
Darrick J. Wongdff68252010-10-04 12:35:05 -07002398
2399 if (buffer_write_io_error(sbh)) {
2400 /*
2401 * Oh, dear. A previous attempt to write the
2402 * superblock failed. This could happen because the
2403 * USB device was yanked out. Or it could happen to
2404 * be a transient write error and maybe the block will
2405 * be remapped. Nothing we can do but to retry the
2406 * write and hope for the best.
2407 */
2408 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2409 "superblock detected");
2410 clear_buffer_write_io_error(sbh);
2411 set_buffer_uptodate(sbh);
2412 }
Theodore Ts'o96ec2e02009-09-16 11:21:13 -04002413 /*
2414 * If the file system is mounted read-only, don't update the
2415 * superblock write time. This avoids updating the superblock
2416 * write time when we are mounting the root file system
2417 * read/only but we need to replay the journal; at that point,
2418 * for people who are east of GMT and who make their clock
2419 * tick in localtime for Windows bug-for-bug compatibility,
2420 * the clock is set in the future, and this will cause e2fsck
2421 * to complain and force a full file system check.
2422 */
2423 if (!(sb->s_flags & MS_RDONLY))
2424 es->s_wtime = cpu_to_le32(get_seconds());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2426 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2427 BUFFER_TRACE(sbh, "marking dirty");
2428 mark_buffer_dirty(sbh);
Darrick J. Wongdff68252010-10-04 12:35:05 -07002429 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08002430 error = sync_dirty_buffer(sbh);
Darrick J. Wongdff68252010-10-04 12:35:05 -07002431 if (buffer_write_io_error(sbh)) {
2432 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2433 "superblock");
2434 clear_buffer_write_io_error(sbh);
2435 set_buffer_uptodate(sbh);
2436 }
2437 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08002438 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439}
2440
2441
2442/*
2443 * Have we just finished recovery? If so, and if we are mounting (or
2444 * remounting) the filesystem readonly, then we will end up with a
2445 * consistent fs on disk. Record that fact.
2446 */
2447static void ext3_mark_recovery_complete(struct super_block * sb,
2448 struct ext3_super_block * es)
2449{
2450 journal_t *journal = EXT3_SB(sb)->s_journal;
2451
2452 journal_lock_updates(journal);
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002453 if (journal_flush(journal) < 0)
2454 goto out;
2455
Linus Torvalds1da177e2005-04-16 15:20:36 -07002456 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2457 sb->s_flags & MS_RDONLY) {
2458 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459 ext3_commit_super(sb, es, 1);
2460 }
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002461
2462out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 journal_unlock_updates(journal);
2464}
2465
2466/*
2467 * If we are mounting (or read-write remounting) a filesystem whose journal
2468 * has recorded an error from a previous lifetime, move that error to the
2469 * main filesystem now.
2470 */
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002471static void ext3_clear_journal_err(struct super_block *sb,
2472 struct ext3_super_block *es)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473{
2474 journal_t *journal;
2475 int j_errno;
2476 const char *errstr;
2477
2478 journal = EXT3_SB(sb)->s_journal;
2479
2480 /*
2481 * Now check for any error status which may have been recorded in the
2482 * journal by a prior ext3_error() or ext3_abort()
2483 */
2484
2485 j_errno = journal_errno(journal);
2486 if (j_errno) {
2487 char nbuf[16];
2488
2489 errstr = ext3_decode_error(sb, j_errno, nbuf);
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002490 ext3_warning(sb, __func__, "Filesystem error recorded "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491 "from previous mount: %s", errstr);
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002492 ext3_warning(sb, __func__, "Marking fs in need of "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493 "filesystem check.");
2494
2495 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2496 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2497 ext3_commit_super (sb, es, 1);
2498
2499 journal_clear_err(journal);
2500 }
2501}
2502
2503/*
2504 * Force the running and committing transactions to commit,
2505 * and wait on the commit.
2506 */
2507int ext3_force_commit(struct super_block *sb)
2508{
2509 journal_t *journal;
2510 int ret;
2511
2512 if (sb->s_flags & MS_RDONLY)
2513 return 0;
2514
2515 journal = EXT3_SB(sb)->s_journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 ret = ext3_journal_force_commit(journal);
2517 return ret;
2518}
2519
Linus Torvalds1da177e2005-04-16 15:20:36 -07002520static int ext3_sync_fs(struct super_block *sb, int wait)
2521{
Jan Kara02ac5972009-02-11 13:04:26 -08002522 tid_t target;
Arthur Jonesc87591b2008-11-06 12:53:35 -08002523
Lukas Czerner785c4bc2011-05-23 18:33:01 +02002524 trace_ext3_sync_fs(sb, wait);
Jan Kara02ac5972009-02-11 13:04:26 -08002525 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2526 if (wait)
2527 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2528 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002529 return 0;
2530}
2531
2532/*
2533 * LVM calls this function before a (read-only) snapshot is created. This
2534 * gives us a chance to flush the journal completely and mark the fs clean.
2535 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08002536static int ext3_freeze(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537{
Takashi Satoc4be0c12009-01-09 16:40:58 -08002538 int error = 0;
2539 journal_t *journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540
2541 if (!(sb->s_flags & MS_RDONLY)) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08002542 journal = EXT3_SB(sb)->s_journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002543
2544 /* Now we set up the journal barrier. */
2545 journal_lock_updates(journal);
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002546
2547 /*
2548 * We don't want to clear needs_recovery flag when we failed
2549 * to flush the journal.
2550 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08002551 error = journal_flush(journal);
2552 if (error < 0)
2553 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554
2555 /* Journal blocked and flushed, clear needs_recovery flag. */
2556 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
Takashi Satoc4be0c12009-01-09 16:40:58 -08002557 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2558 if (error)
2559 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002560 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08002561 return 0;
2562
2563out:
2564 journal_unlock_updates(journal);
2565 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002566}
2567
2568/*
2569 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2570 * flag here, even though the filesystem is not technically dirty yet.
2571 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08002572static int ext3_unfreeze(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573{
2574 if (!(sb->s_flags & MS_RDONLY)) {
2575 lock_super(sb);
2576 /* Reser the needs_recovery flag before the fs is unlocked. */
2577 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2578 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2579 unlock_super(sb);
2580 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2581 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08002582 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002583}
2584
2585static int ext3_remount (struct super_block * sb, int * flags, char * data)
2586{
2587 struct ext3_super_block * es;
2588 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07002589 ext3_fsblk_t n_blocks_count = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002590 unsigned long old_sb_flags;
2591 struct ext3_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002592 int enable_quota = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002593 int err;
2594#ifdef CONFIG_QUOTA
2595 int i;
2596#endif
2597
2598 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04002599 lock_super(sb);
Jan Kara08c6a962005-07-12 13:58:28 -07002600 old_sb_flags = sb->s_flags;
2601 old_opts.s_mount_opt = sbi->s_mount_opt;
2602 old_opts.s_resuid = sbi->s_resuid;
2603 old_opts.s_resgid = sbi->s_resgid;
2604 old_opts.s_commit_interval = sbi->s_commit_interval;
2605#ifdef CONFIG_QUOTA
2606 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2607 for (i = 0; i < MAXQUOTAS; i++)
2608 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2609#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610
2611 /*
2612 * Allow the "check" option to be passed as a remount option.
2613 */
Johann Lombardi71b96252006-01-08 01:03:20 -08002614 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002615 err = -EINVAL;
2616 goto restore_opts;
2617 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618
Dmitry Monakhove3c96432010-02-02 16:05:51 +03002619 if (test_opt(sb, ABORT))
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002620 ext3_abort(sb, __func__, "Abort forced by user");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002621
2622 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhove3c96432010-02-02 16:05:51 +03002623 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624
2625 es = sbi->s_es;
2626
2627 ext3_init_journal_params(sb, sbi->s_journal);
2628
2629 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2630 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03002631 if (test_opt(sb, ABORT)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002632 err = -EROFS;
2633 goto restore_opts;
2634 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635
2636 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04002637 err = dquot_suspend(sb, -1);
2638 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002639 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002640
Linus Torvalds1da177e2005-04-16 15:20:36 -07002641 /*
2642 * First of all, the unconditional stuff we have to do
2643 * to disable replay of the journal when we next remount
2644 */
2645 sb->s_flags |= MS_RDONLY;
2646
2647 /*
2648 * OK, test if we are remounting a valid rw partition
2649 * readonly, and if so set the rdonly flag and then
2650 * mark the partition as valid again.
2651 */
2652 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2653 (sbi->s_mount_state & EXT3_VALID_FS))
2654 es->s_state = cpu_to_le16(sbi->s_mount_state);
2655
2656 ext3_mark_recovery_complete(sb, es);
2657 } else {
2658 __le32 ret;
2659 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2660 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002661 ext3_msg(sb, KERN_WARNING,
2662 "warning: couldn't remount RDWR "
2663 "because of unsupported optional "
2664 "features (%x)", le32_to_cpu(ret));
Jan Kara08c6a962005-07-12 13:58:28 -07002665 err = -EROFS;
2666 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002667 }
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002668
2669 /*
2670 * If we have an unprocessed orphan list hanging
2671 * around from a previously readonly bdev mount,
Toshiyuki Okajima1cde2012011-08-02 18:16:57 +09002672 * require a full umount & mount for now.
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002673 */
2674 if (es->s_last_orphan) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002675 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002676 "remount RDWR because of unprocessed "
2677 "orphan inode list. Please "
Toshiyuki Okajima1cde2012011-08-02 18:16:57 +09002678 "umount & mount instead.");
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002679 err = -EINVAL;
2680 goto restore_opts;
2681 }
2682
Linus Torvalds1da177e2005-04-16 15:20:36 -07002683 /*
2684 * Mounting a RDONLY partition read-write, so reread
2685 * and store the current valid flag. (It may have
2686 * been changed by e2fsck since we originally mounted
2687 * the partition.)
2688 */
2689 ext3_clear_journal_err(sb, es);
2690 sbi->s_mount_state = le16_to_cpu(es->s_state);
Dave Kleikampa4e4de32006-09-27 01:49:36 -07002691 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
Jan Kara08c6a962005-07-12 13:58:28 -07002692 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002693 if (!ext3_setup_super (sb, es, 0))
2694 sb->s_flags &= ~MS_RDONLY;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002695 enable_quota = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696 }
2697 }
Jan Kara08c6a962005-07-12 13:58:28 -07002698#ifdef CONFIG_QUOTA
2699 /* Release old quota file names */
2700 for (i = 0; i < MAXQUOTAS; i++)
2701 if (old_opts.s_qf_names[i] &&
2702 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2703 kfree(old_opts.s_qf_names[i]);
2704#endif
Al Virobbd68512009-05-06 10:43:07 -04002705 unlock_super(sb);
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002706
2707 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04002708 dquot_resume(sb, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002709 return 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002710restore_opts:
2711 sb->s_flags = old_sb_flags;
2712 sbi->s_mount_opt = old_opts.s_mount_opt;
2713 sbi->s_resuid = old_opts.s_resuid;
2714 sbi->s_resgid = old_opts.s_resgid;
2715 sbi->s_commit_interval = old_opts.s_commit_interval;
2716#ifdef CONFIG_QUOTA
2717 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2718 for (i = 0; i < MAXQUOTAS; i++) {
2719 if (sbi->s_qf_names[i] &&
2720 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2721 kfree(sbi->s_qf_names[i]);
2722 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2723 }
2724#endif
Al Virobbd68512009-05-06 10:43:07 -04002725 unlock_super(sb);
Jan Kara08c6a962005-07-12 13:58:28 -07002726 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002727}
2728
David Howells726c3342006-06-23 02:02:58 -07002729static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730{
David Howells726c3342006-06-23 02:02:58 -07002731 struct super_block *sb = dentry->d_sb;
Alex Tomas09fe3162006-03-24 03:16:16 -08002732 struct ext3_sb_info *sbi = EXT3_SB(sb);
2733 struct ext3_super_block *es = sbi->s_es;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002734 u64 fsid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002735
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002736 if (test_opt(sb, MINIX_DF)) {
2737 sbi->s_overhead_last = 0;
2738 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2739 unsigned long ngroups = sbi->s_groups_count, i;
2740 ext3_fsblk_t overhead = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741 smp_rmb();
2742
2743 /*
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002744 * Compute the overhead (FS structures). This is constant
2745 * for a given filesystem unless the number of block groups
2746 * changes so we cache the previous value until it does.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747 */
2748
2749 /*
2750 * All of the blocks before first_data_block are
2751 * overhead
2752 */
2753 overhead = le32_to_cpu(es->s_first_data_block);
2754
2755 /*
2756 * Add the overhead attributed to the superblock and
2757 * block group descriptors. If the sparse superblocks
2758 * feature is turned on, then not all groups have this.
2759 */
2760 for (i = 0; i < ngroups; i++) {
2761 overhead += ext3_bg_has_super(sb, i) +
2762 ext3_bg_num_gdb(sb, i);
2763 cond_resched();
2764 }
2765
2766 /*
2767 * Every block group has an inode bitmap, a block
2768 * bitmap, and an inode table.
2769 */
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002770 overhead += ngroups * (2 + sbi->s_itb_per_group);
2771 sbi->s_overhead_last = overhead;
2772 smp_wmb();
2773 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002774 }
2775
2776 buf->f_type = EXT3_SUPER_MAGIC;
2777 buf->f_bsize = sb->s_blocksize;
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002778 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002779 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2781 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2782 buf->f_bavail = 0;
2783 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002784 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002785 buf->f_namelen = EXT3_NAME_LEN;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002786 fsid = le64_to_cpup((void *)es->s_uuid) ^
2787 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2788 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2789 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002790 return 0;
2791}
2792
2793/* Helper function for writing quotas on sync - we need to start transaction before quota file
2794 * is locked for write. Otherwise the are possible deadlocks:
2795 * Process 1 Process 2
2796 * ext3_create() quota_sync()
2797 * journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05002798 * dquot_initialize() down(dqio_mutex)
Ingo Molnard3be9152006-03-23 03:00:29 -08002799 * down(dqio_mutex) journal_start()
Linus Torvalds1da177e2005-04-16 15:20:36 -07002800 *
2801 */
2802
2803#ifdef CONFIG_QUOTA
2804
2805static inline struct inode *dquot_to_inode(struct dquot *dquot)
2806{
2807 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2808}
2809
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810static int ext3_write_dquot(struct dquot *dquot)
2811{
2812 int ret, err;
2813 handle_t *handle;
2814 struct inode *inode;
2815
2816 inode = dquot_to_inode(dquot);
2817 handle = ext3_journal_start(inode,
Jan Kara1f545872005-06-23 22:01:04 -07002818 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819 if (IS_ERR(handle))
2820 return PTR_ERR(handle);
2821 ret = dquot_commit(dquot);
2822 err = ext3_journal_stop(handle);
2823 if (!ret)
2824 ret = err;
2825 return ret;
2826}
2827
2828static int ext3_acquire_dquot(struct dquot *dquot)
2829{
2830 int ret, err;
2831 handle_t *handle;
2832
2833 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002834 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835 if (IS_ERR(handle))
2836 return PTR_ERR(handle);
2837 ret = dquot_acquire(dquot);
2838 err = ext3_journal_stop(handle);
2839 if (!ret)
2840 ret = err;
2841 return ret;
2842}
2843
2844static int ext3_release_dquot(struct dquot *dquot)
2845{
2846 int ret, err;
2847 handle_t *handle;
2848
2849 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002850 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07002851 if (IS_ERR(handle)) {
2852 /* Release dquot anyway to avoid endless cycle in dqput() */
2853 dquot_release(dquot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002854 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07002855 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856 ret = dquot_release(dquot);
2857 err = ext3_journal_stop(handle);
2858 if (!ret)
2859 ret = err;
2860 return ret;
2861}
2862
2863static int ext3_mark_dquot_dirty(struct dquot *dquot)
2864{
Jan Kara99aeaf62008-07-25 01:46:17 -07002865 /* Are we journaling quotas? */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002866 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2867 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2868 dquot_mark_dquot_dirty(dquot);
2869 return ext3_write_dquot(dquot);
2870 } else {
2871 return dquot_mark_dquot_dirty(dquot);
2872 }
2873}
2874
2875static int ext3_write_info(struct super_block *sb, int type)
2876{
2877 int ret, err;
2878 handle_t *handle;
2879
2880 /* Data block + inode block */
2881 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2882 if (IS_ERR(handle))
2883 return PTR_ERR(handle);
2884 ret = dquot_commit_info(sb, type);
2885 err = ext3_journal_stop(handle);
2886 if (!ret)
2887 ret = err;
2888 return ret;
2889}
2890
2891/*
2892 * Turn on quotas during mount time - we need to find
2893 * the quota file and such...
2894 */
2895static int ext3_quota_on_mount(struct super_block *sb, int type)
2896{
Christoph Hellwig287a8092010-05-19 07:16:45 -04002897 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2898 EXT3_SB(sb)->s_jquota_fmt, type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002899}
2900
2901/*
2902 * Standard function to be called on quota_on
2903 */
2904static int ext3_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02002905 struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002906{
2907 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908
Jan Kara1f545872005-06-23 22:01:04 -07002909 if (!test_opt(sb, QUOTA))
2910 return -EINVAL;
Jan Kara9cfe7b92008-07-25 01:46:16 -07002911
Linus Torvalds1da177e2005-04-16 15:20:36 -07002912 /* Quotafile not on the same filesystem? */
Jan Karaf00c9e42010-09-15 17:38:58 +02002913 if (path->mnt->mnt_sb != sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002914 return -EXDEV;
Jan Kara9cfe7b92008-07-25 01:46:16 -07002915 /* Journaling quota? */
2916 if (EXT3_SB(sb)->s_qf_names[type]) {
2917 /* Quotafile not of fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02002918 if (path->dentry->d_parent != sb->s_root)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002919 ext3_msg(sb, KERN_WARNING,
2920 "warning: Quota file not on filesystem root. "
2921 "Journaled quota will not work.");
Jan Kara9cfe7b92008-07-25 01:46:16 -07002922 }
2923
2924 /*
2925 * When we journal data on quota file, we have to flush journal to see
2926 * all updates to the file when we bypass pagecache...
2927 */
Jan Karaf00c9e42010-09-15 17:38:58 +02002928 if (ext3_should_journal_data(path->dentry->d_inode)) {
Jan Kara9cfe7b92008-07-25 01:46:16 -07002929 /*
2930 * We don't need to lock updates but journal_flush() could
2931 * otherwise be livelocked...
2932 */
2933 journal_lock_updates(EXT3_SB(sb)->s_journal);
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002934 err = journal_flush(EXT3_SB(sb)->s_journal);
Jan Kara9cfe7b92008-07-25 01:46:16 -07002935 journal_unlock_updates(EXT3_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02002936 if (err)
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002937 return err;
Jan Kara9cfe7b92008-07-25 01:46:16 -07002938 }
2939
Jan Karaf00c9e42010-09-15 17:38:58 +02002940 return dquot_quota_on(sb, type, format_id, path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002941}
2942
2943/* Read data from quotafile - avoid pagecache and such because we cannot afford
2944 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002945 * itself serializes the operations (and no one else should touch the files)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002946 * we don't have to be afraid of races */
2947static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2948 size_t len, loff_t off)
2949{
2950 struct inode *inode = sb_dqopt(sb)->files[type];
2951 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2952 int err = 0;
2953 int offset = off & (sb->s_blocksize - 1);
2954 int tocopy;
2955 size_t toread;
2956 struct buffer_head *bh;
2957 loff_t i_size = i_size_read(inode);
2958
2959 if (off > i_size)
2960 return 0;
2961 if (off+len > i_size)
2962 len = i_size-off;
2963 toread = len;
2964 while (toread > 0) {
2965 tocopy = sb->s_blocksize - offset < toread ?
2966 sb->s_blocksize - offset : toread;
2967 bh = ext3_bread(NULL, inode, blk, 0, &err);
2968 if (err)
2969 return err;
2970 if (!bh) /* A hole? */
2971 memset(data, 0, tocopy);
2972 else
2973 memcpy(data, bh->b_data+offset, tocopy);
2974 brelse(bh);
2975 offset = 0;
2976 toread -= tocopy;
2977 data += tocopy;
2978 blk++;
2979 }
2980 return len;
2981}
2982
2983/* Write to quotafile (we know the transaction is already started and has
2984 * enough credits) */
2985static ssize_t ext3_quota_write(struct super_block *sb, int type,
2986 const char *data, size_t len, loff_t off)
2987{
2988 struct inode *inode = sb_dqopt(sb)->files[type];
2989 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2990 int err = 0;
2991 int offset = off & (sb->s_blocksize - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002992 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002993 struct buffer_head *bh;
2994 handle_t *handle = journal_current_handle();
2995
Jan Kara9c3013e2007-09-11 15:23:29 -07002996 if (!handle) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002997 ext3_msg(sb, KERN_WARNING,
2998 "warning: quota write (off=%llu, len=%llu)"
2999 " cancelled because transaction is not started.",
Jan Kara9c3013e2007-09-11 15:23:29 -07003000 (unsigned long long)off, (unsigned long long)len);
3001 return -EIO;
3002 }
Dmitry Monakhove5472142010-02-16 19:33:42 +03003003
3004 /*
3005 * Since we account only one data block in transaction credits,
3006 * then it is impossible to cross a block boundary.
3007 */
3008 if (sb->s_blocksize - offset < len) {
3009 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3010 " cancelled because not block aligned",
3011 (unsigned long long)off, (unsigned long long)len);
3012 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003013 }
Dmitry Monakhove5472142010-02-16 19:33:42 +03003014 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3015 bh = ext3_bread(handle, inode, blk, 1, &err);
3016 if (!bh)
3017 goto out;
3018 if (journal_quota) {
3019 err = ext3_journal_get_write_access(handle, bh);
3020 if (err) {
3021 brelse(bh);
3022 goto out;
3023 }
3024 }
3025 lock_buffer(bh);
3026 memcpy(bh->b_data+offset, data, len);
3027 flush_dcache_page(bh->b_page);
3028 unlock_buffer(bh);
3029 if (journal_quota)
3030 err = ext3_journal_dirty_metadata(handle, bh);
3031 else {
3032 /* Always do at least ordered writes for quotas */
3033 err = ext3_journal_dirty_data(handle, bh);
3034 mark_buffer_dirty(bh);
3035 }
3036 brelse(bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003037out:
Dmitry Monakhove5472142010-02-16 19:33:42 +03003038 if (err) {
Jan Karaf5c8f7d2008-07-04 09:59:33 -07003039 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003040 return err;
Jan Karaf5c8f7d2008-07-04 09:59:33 -07003041 }
Dmitry Monakhove5472142010-02-16 19:33:42 +03003042 if (inode->i_size < off + len) {
3043 i_size_write(inode, off + len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003044 EXT3_I(inode)->i_disksize = inode->i_size;
3045 }
3046 inode->i_version++;
3047 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3048 ext3_mark_inode_dirty(handle, inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08003049 mutex_unlock(&inode->i_mutex);
Dmitry Monakhove5472142010-02-16 19:33:42 +03003050 return len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051}
3052
3053#endif
3054
Al Viro152a0832010-07-25 00:46:55 +04003055static struct dentry *ext3_mount(struct file_system_type *fs_type,
3056 int flags, const char *dev_name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003057{
Al Viro152a0832010-07-25 00:46:55 +04003058 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003059}
3060
3061static struct file_system_type ext3_fs_type = {
3062 .owner = THIS_MODULE,
3063 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +04003064 .mount = ext3_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003065 .kill_sb = kill_block_super,
3066 .fs_flags = FS_REQUIRES_DEV,
3067};
3068
3069static int __init init_ext3_fs(void)
3070{
3071 int err = init_ext3_xattr();
3072 if (err)
3073 return err;
3074 err = init_inodecache();
3075 if (err)
3076 goto out1;
3077 err = register_filesystem(&ext3_fs_type);
3078 if (err)
3079 goto out;
3080 return 0;
3081out:
3082 destroy_inodecache();
3083out1:
Dave Kleikampe9ad5622006-09-27 01:49:35 -07003084 exit_ext3_xattr();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003085 return err;
3086}
3087
3088static void __exit exit_ext3_fs(void)
3089{
3090 unregister_filesystem(&ext3_fs_type);
3091 destroy_inodecache();
3092 exit_ext3_xattr();
3093}
3094
3095MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3096MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3097MODULE_LICENSE("GPL");
3098module_init(init_ext3_fs)
3099module_exit(exit_ext3_fs)