blob: c50c76190373a06bdb1a7971a935e7a5d00d9888 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/blkdev.h>
21#include <linux/parser.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070022#include <linux/exportfs.h>
Al Viro4613ad12012-03-29 22:30:07 -040023#include <linux/statfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/random.h>
25#include <linux/mount.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/quotaops.h>
Mark Bellon8fc27512005-09-06 15:16:54 -070027#include <linux/seq_file.h>
vignesh babu3fc74262007-07-15 23:41:17 -070028#include <linux/log2.h>
Dan Magenheimerd71bc6d2011-05-26 10:01:49 -060029#include <linux/cleancache.h>
Eric Sandeencf7eff42013-07-31 14:33:00 -050030#include <linux/namei.h>
Ben Dooks381be2542005-10-30 15:02:56 -080031
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <asm/uaccess.h>
Ben Dooks381be2542005-10-30 15:02:56 -080033
Al Viro4613ad12012-03-29 22:30:07 -040034#define CREATE_TRACE_POINTS
35
36#include "ext3.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include "xattr.h"
38#include "acl.h"
Ben Dooks381be2542005-10-30 15:02:56 -080039#include "namei.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
Linus Torvaldsbbae8bc2009-04-06 17:16:47 -070041#ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
42 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
43#else
44 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
45#endif
46
Johann Lombardi71b96252006-01-08 01:03:20 -080047static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
Linus Torvalds1da177e2005-04-16 15:20:36 -070049static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
Eric Sandeeneee194e2006-09-27 01:49:30 -070050 unsigned int);
Takashi Satoc4be0c12009-01-09 16:40:58 -080051static int ext3_commit_super(struct super_block *sb,
52 struct ext3_super_block *es,
Linus Torvalds1da177e2005-04-16 15:20:36 -070053 int sync);
54static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58static int ext3_sync_fs(struct super_block *sb, int wait);
59static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61static int ext3_remount (struct super_block * sb, int * flags, char * data);
David Howells726c3342006-06-23 02:02:58 -070062static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080063static int ext3_unfreeze(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080064static int ext3_freeze(struct super_block *sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
Mingming Caoae6ddcc2006-09-27 01:49:27 -070066/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 * Wrappers for journal_start/end.
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 */
69handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
70{
71 journal_t *journal;
72
73 if (sb->s_flags & MS_RDONLY)
74 return ERR_PTR(-EROFS);
75
76 /* Special case here: if the journal has aborted behind our
77 * backs (eg. EIO in the commit thread), then we still need to
78 * take the FS itself readonly cleanly. */
79 journal = EXT3_SB(sb)->s_journal;
80 if (is_journal_aborted(journal)) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -070081 ext3_abort(sb, __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 "Detected aborted journal");
83 return ERR_PTR(-EROFS);
84 }
85
86 return journal_start(journal, nblocks);
87}
88
Linus Torvalds1da177e2005-04-16 15:20:36 -070089int __ext3_journal_stop(const char *where, handle_t *handle)
90{
91 struct super_block *sb;
92 int err;
93 int rc;
94
95 sb = handle->h_transaction->t_journal->j_private;
96 err = handle->h_err;
97 rc = journal_stop(handle);
98
99 if (!err)
100 err = rc;
101 if (err)
102 __ext3_std_error(sb, where, err);
103 return err;
104}
105
106void ext3_journal_abort_handle(const char *caller, const char *err_fn,
107 struct buffer_head *bh, handle_t *handle, int err)
108{
109 char nbuf[16];
110 const char *errstr = ext3_decode_error(NULL, err, nbuf);
111
112 if (bh)
113 BUFFER_TRACE(bh, "abort");
114
115 if (!handle->h_err)
116 handle->h_err = err;
117
118 if (is_handle_aborted(handle))
119 return;
120
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100121 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
122 caller, errstr, err_fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
124 journal_abort_handle(handle);
125}
126
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100127void ext3_msg(struct super_block *sb, const char *prefix,
128 const char *fmt, ...)
129{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800130 struct va_format vaf;
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100131 va_list args;
132
133 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800134
135 vaf.fmt = fmt;
136 vaf.va = &args;
137
138 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
139
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100140 va_end(args);
141}
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143/* Deal with the reporting of failure conditions on a filesystem such as
144 * inconsistencies detected or read IO failures.
145 *
146 * On ext2, we can store the error state of the filesystem in the
147 * superblock. That is not possible on ext3, because we may have other
148 * write ordering constraints on the superblock which prevent us from
149 * writing it out straight away; and given that the journal is about to
150 * be aborted, we can't rely on the current, or future, transactions to
151 * write out the superblock safely.
152 *
153 * We'll just use the journal_abort() error code to record an error in
Thadeu Lima de Souza Cascardod6b198b2010-01-17 19:10:07 -0200154 * the journal instead. On recovery, the journal will complain about
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 * that error until we've noted it down and cleared it.
156 */
157
158static void ext3_handle_error(struct super_block *sb)
159{
160 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
161
162 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
163 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
164
165 if (sb->s_flags & MS_RDONLY)
166 return;
167
Vasily Averin7543fc72006-09-27 01:49:33 -0700168 if (!test_opt (sb, ERRORS_CONT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 journal_t *journal = EXT3_SB(sb)->s_journal;
170
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300171 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 if (journal)
173 journal_abort(journal, -EIO);
174 }
Vasily Averin7543fc72006-09-27 01:49:33 -0700175 if (test_opt (sb, ERRORS_RO)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100176 ext3_msg(sb, KERN_CRIT,
177 "error: remounting filesystem read-only");
Dmitry Monakhov3df32192013-05-28 13:19:01 +0400178 /*
179 * Make sure updated value of ->s_mount_state will be visible
180 * before ->s_flags update.
181 */
182 smp_wmb();
Vasily Averin7543fc72006-09-27 01:49:33 -0700183 sb->s_flags |= MS_RDONLY;
184 }
185 ext3_commit_super(sb, es, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 if (test_opt(sb, ERRORS_PANIC))
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100187 panic("EXT3-fs (%s): panic forced after error\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189}
190
Joe Perches99fbb1e2010-11-09 10:18:05 -0800191void ext3_error(struct super_block *sb, const char *function,
192 const char *fmt, ...)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800194 struct va_format vaf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 va_list args;
196
197 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800198
199 vaf.fmt = fmt;
200 vaf.va = &args;
201
202 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
203 sb->s_id, function, &vaf);
204
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 va_end(args);
206
207 ext3_handle_error(sb);
208}
209
210static const char *ext3_decode_error(struct super_block * sb, int errno,
211 char nbuf[16])
212{
213 char *errstr = NULL;
214
215 switch (errno) {
216 case -EIO:
217 errstr = "IO failure";
218 break;
219 case -ENOMEM:
220 errstr = "Out of memory";
221 break;
222 case -EROFS:
223 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
224 errstr = "Journal has aborted";
225 else
226 errstr = "Readonly filesystem";
227 break;
228 default:
229 /* If the caller passed in an extra buffer for unknown
230 * errors, textualise them now. Else we just return
231 * NULL. */
232 if (nbuf) {
233 /* Check for truncated error codes... */
234 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
235 errstr = nbuf;
236 }
237 break;
238 }
239
240 return errstr;
241}
242
243/* __ext3_std_error decodes expected errors from journaling functions
244 * automatically and invokes the appropriate error response. */
245
246void __ext3_std_error (struct super_block * sb, const char * function,
247 int errno)
248{
249 char nbuf[16];
Stephen Tweedie30121622005-05-18 11:47:17 -0400250 const char *errstr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
Stephen Tweedie30121622005-05-18 11:47:17 -0400252 /* Special case: if the error is EROFS, and we're not already
253 * inside a transaction, then there's really no point in logging
254 * an error. */
255 if (errno == -EROFS && journal_current_handle() == NULL &&
256 (sb->s_flags & MS_RDONLY))
257 return;
258
259 errstr = ext3_decode_error(sb, errno, nbuf);
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100260 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
262 ext3_handle_error(sb);
263}
264
265/*
266 * ext3_abort is a much stronger failure handler than ext3_error. The
267 * abort function may be used to deal with unrecoverable failures such
268 * as journal IO errors or ENOMEM at a critical moment in log management.
269 *
270 * We unconditionally force the filesystem into an ABORT|READONLY state,
271 * unless the error response on the fs has been set to panic in which
272 * case we take the easy way out and panic immediately.
273 */
274
Joe Perches99fbb1e2010-11-09 10:18:05 -0800275void ext3_abort(struct super_block *sb, const char *function,
276 const char *fmt, ...)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800278 struct va_format vaf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 va_list args;
280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800282
283 vaf.fmt = fmt;
284 vaf.va = &args;
285
286 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
287 sb->s_id, function, &vaf);
288
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 va_end(args);
290
291 if (test_opt(sb, ERRORS_PANIC))
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100292 panic("EXT3-fs: panic from previous error\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
294 if (sb->s_flags & MS_RDONLY)
295 return;
296
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100297 ext3_msg(sb, KERN_CRIT,
298 "error: remounting filesystem read-only");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300300 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
Dmitry Monakhov3df32192013-05-28 13:19:01 +0400301 /*
302 * Make sure updated value of ->s_mount_state will be visible
303 * before ->s_flags update.
304 */
305 smp_wmb();
306 sb->s_flags |= MS_RDONLY;
307
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400308 if (EXT3_SB(sb)->s_journal)
309 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310}
311
Joe Perches99fbb1e2010-11-09 10:18:05 -0800312void ext3_warning(struct super_block *sb, const char *function,
313 const char *fmt, ...)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314{
Joe Perches99fbb1e2010-11-09 10:18:05 -0800315 struct va_format vaf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 va_list args;
317
318 va_start(args, fmt);
Joe Perches99fbb1e2010-11-09 10:18:05 -0800319
320 vaf.fmt = fmt;
321 vaf.va = &args;
322
323 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
324 sb->s_id, function, &vaf);
325
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 va_end(args);
327}
328
329void ext3_update_dynamic_rev(struct super_block *sb)
330{
331 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
332
333 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
334 return;
335
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100336 ext3_msg(sb, KERN_WARNING,
337 "warning: updating to rev %d because of "
338 "new feature flag, running e2fsck is recommended",
339 EXT3_DYNAMIC_REV);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
341 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
342 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
343 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
344 /* leave es->s_feature_*compat flags alone */
345 /* es->s_uuid will be set by e2fsck if empty */
346
347 /*
348 * The rest of the superblock fields should be zero, and if not it
349 * means they are likely already in use, so leave them alone. We
350 * can leave it up to e2fsck to clean up any inconsistencies there.
351 */
352}
353
354/*
355 * Open the external journal device
356 */
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100357static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358{
359 struct block_device *bdev;
360 char b[BDEVNAME_SIZE];
361
Tejun Heod4d77622010-11-13 11:55:18 +0100362 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 if (IS_ERR(bdev))
364 goto fail;
365 return bdev;
366
367fail:
Lars-Peter Clausen8d0c2d12013-03-09 15:28:44 +0100368 ext3_msg(sb, KERN_ERR, "error: failed to open journal device %s: %ld",
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100369 __bdevname(dev, b), PTR_ERR(bdev));
370
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 return NULL;
372}
373
374/*
375 * Release the journal device
376 */
Al Viro4385bab2013-05-05 22:11:03 -0400377static void ext3_blkdev_put(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378{
Al Viro4385bab2013-05-05 22:11:03 -0400379 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380}
381
Al Viro4385bab2013-05-05 22:11:03 -0400382static void ext3_blkdev_remove(struct ext3_sb_info *sbi)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383{
384 struct block_device *bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 bdev = sbi->journal_bdev;
386 if (bdev) {
Al Viro4385bab2013-05-05 22:11:03 -0400387 ext3_blkdev_put(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 sbi->journal_bdev = NULL;
389 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390}
391
392static inline struct inode *orphan_list_entry(struct list_head *l)
393{
394 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
395}
396
397static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
398{
399 struct list_head *l;
400
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100401 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 le32_to_cpu(sbi->s_es->s_last_orphan));
403
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100404 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 list_for_each(l, &sbi->s_orphan) {
406 struct inode *inode = orphan_list_entry(l);
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100407 ext3_msg(sb, KERN_ERR, " "
Eric Sandeeneee194e2006-09-27 01:49:30 -0700408 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 inode->i_sb->s_id, inode->i_ino, inode,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700410 inode->i_mode, inode->i_nlink,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 NEXT_ORPHAN(inode));
412 }
413}
414
415static void ext3_put_super (struct super_block * sb)
416{
417 struct ext3_sb_info *sbi = EXT3_SB(sb);
418 struct ext3_super_block *es = sbi->s_es;
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400419 int i, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400421 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 ext3_xattr_put_super(sb);
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400423 err = journal_destroy(sbi->s_journal);
424 sbi->s_journal = NULL;
425 if (err < 0)
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -0700426 ext3_abort(sb, __func__, "Couldn't clean up the journal");
Hidehiro Kawai44d6f782008-10-27 22:51:46 -0400427
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 if (!(sb->s_flags & MS_RDONLY)) {
429 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
430 es->s_state = cpu_to_le16(sbi->s_mount_state);
431 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
432 mark_buffer_dirty(sbi->s_sbh);
433 ext3_commit_super(sb, es, 1);
434 }
435
436 for (i = 0; i < sbi->s_gdb_count; i++)
437 brelse(sbi->s_group_desc[i]);
438 kfree(sbi->s_group_desc);
439 percpu_counter_destroy(&sbi->s_freeblocks_counter);
440 percpu_counter_destroy(&sbi->s_freeinodes_counter);
441 percpu_counter_destroy(&sbi->s_dirs_counter);
442 brelse(sbi->s_sbh);
443#ifdef CONFIG_QUOTA
444 for (i = 0; i < MAXQUOTAS; i++)
445 kfree(sbi->s_qf_names[i]);
446#endif
447
448 /* Debugging code just in case the in-memory inode orphan list
449 * isn't empty. The on-disk one can be non-empty if we've
450 * detected an error and taken the fs readonly, but the
451 * in-memory list had better be clean by this point. */
452 if (!list_empty(&sbi->s_orphan))
453 dump_orphan_list(sb, sbi);
454 J_ASSERT(list_empty(&sbi->s_orphan));
455
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700456 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
458 /*
459 * Invalidate the journal device's buffers. We don't want them
460 * floating about in memory - the physical journal device may
461 * hotswapped, and it breaks the `ro-after' testing code.
462 */
463 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700464 invalidate_bdev(sbi->journal_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 ext3_blkdev_remove(sbi);
466 }
467 sb->s_fs_info = NULL;
Pekka Enberg5df096d2009-01-07 18:07:25 -0800468 kfree(sbi->s_blockgroup_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 kfree(sbi);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470}
471
Christoph Lametere18b8902006-12-06 20:33:20 -0800472static struct kmem_cache *ext3_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
474/*
475 * Called inside transaction, so use GFP_NOFS
476 */
477static struct inode *ext3_alloc_inode(struct super_block *sb)
478{
479 struct ext3_inode_info *ei;
480
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800481 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (!ei)
483 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 ei->i_block_alloc_info = NULL;
485 ei->vfs_inode.i_version = 1;
Jan Karafe8bc912009-10-16 19:26:15 +0200486 atomic_set(&ei->i_datasync_tid, 0);
487 atomic_set(&ei->i_sync_tid, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 return &ei->vfs_inode;
489}
490
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200491static int ext3_drop_inode(struct inode *inode)
492{
493 int drop = generic_drop_inode(inode);
494
495 trace_ext3_drop_inode(inode, drop);
496 return drop;
497}
498
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100499static void ext3_i_callback(struct rcu_head *head)
500{
501 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100502 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
503}
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505static void ext3_destroy_inode(struct inode *inode)
506{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700507 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
508 printk("EXT3 Inode %p: orphan list check failed!\n",
509 EXT3_I(inode));
510 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
511 EXT3_I(inode), sizeof(struct ext3_inode_info),
512 false);
513 dump_stack();
514 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100515 call_rcu(&inode->i_rcu, ext3_i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516}
517
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700518static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519{
520 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
521
Christoph Lametera35afb82007-05-16 22:10:57 -0700522 INIT_LIST_HEAD(&ei->i_orphan);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523#ifdef CONFIG_EXT3_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700524 init_rwsem(&ei->xattr_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700526 mutex_init(&ei->truncate_mutex);
527 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528}
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700529
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530static int init_inodecache(void)
531{
532 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
533 sizeof(struct ext3_inode_info),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800534 0, (SLAB_RECLAIM_ACCOUNT|
535 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900536 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 if (ext3_inode_cachep == NULL)
538 return -ENOMEM;
539 return 0;
540}
541
542static void destroy_inodecache(void)
543{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +1000544 /*
545 * Make sure all delayed rcu free inodes are flushed before we
546 * destroy cache.
547 */
548 rcu_barrier();
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -0700549 kmem_cache_destroy(ext3_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550}
551
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200552static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
Mark Bellon8fc27512005-09-06 15:16:54 -0700553{
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200554#if defined(CONFIG_QUOTA)
OGAWA Hirofumi275abf52005-09-22 21:44:03 -0700555 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
Jan Kara1aeec432009-11-30 22:22:41 +0100557 if (sbi->s_jquota_fmt) {
558 char *fmtname = "";
559
560 switch (sbi->s_jquota_fmt) {
561 case QFMT_VFS_OLD:
562 fmtname = "vfsold";
563 break;
564 case QFMT_VFS_V0:
565 fmtname = "vfsv0";
566 break;
567 case QFMT_VFS_V1:
568 fmtname = "vfsv1";
569 break;
570 }
571 seq_printf(seq, ",jqfmt=%s", fmtname);
572 }
Mark Bellon8fc27512005-09-06 15:16:54 -0700573
574 if (sbi->s_qf_names[USRQUOTA])
575 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
576
577 if (sbi->s_qf_names[GRPQUOTA])
578 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
579
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300580 if (test_opt(sb, USRQUOTA))
Mark Bellon8fc27512005-09-06 15:16:54 -0700581 seq_puts(seq, ",usrquota");
582
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300583 if (test_opt(sb, GRPQUOTA))
Mark Bellon8fc27512005-09-06 15:16:54 -0700584 seq_puts(seq, ",grpquota");
585#endif
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200586}
587
Jan Kara3c4cec62009-08-24 16:38:43 +0200588static char *data_mode_string(unsigned long mode)
589{
590 switch (mode) {
591 case EXT3_MOUNT_JOURNAL_DATA:
592 return "journal";
593 case EXT3_MOUNT_ORDERED_DATA:
594 return "ordered";
595 case EXT3_MOUNT_WRITEBACK_DATA:
596 return "writeback";
597 }
598 return "unknown";
599}
600
Miklos Szeredi571beed2007-10-16 23:26:26 -0700601/*
602 * Show an option if
603 * - it's set to a non-default value OR
604 * - if the per-sb default is different from the global default
605 */
Al Viro34c80b12011-12-08 21:32:45 -0500606static int ext3_show_options(struct seq_file *seq, struct dentry *root)
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200607{
Al Viro34c80b12011-12-08 21:32:45 -0500608 struct super_block *sb = root->d_sb;
Miklos Szeredi571beed2007-10-16 23:26:26 -0700609 struct ext3_sb_info *sbi = EXT3_SB(sb);
610 struct ext3_super_block *es = sbi->s_es;
611 unsigned long def_mount_opts;
612
613 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
614
615 if (sbi->s_sb_block != 1)
616 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
617 if (test_opt(sb, MINIX_DF))
618 seq_puts(seq, ",minixdf");
619 if (test_opt(sb, GRPID))
620 seq_puts(seq, ",grpid");
621 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
622 seq_puts(seq, ",nogrpid");
Eric W. Biederman15232992012-02-07 15:41:24 -0800623 if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT3_DEF_RESUID)) ||
Miklos Szeredi571beed2007-10-16 23:26:26 -0700624 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
Eric W. Biederman15232992012-02-07 15:41:24 -0800625 seq_printf(seq, ",resuid=%u",
626 from_kuid_munged(&init_user_ns, sbi->s_resuid));
Miklos Szeredi571beed2007-10-16 23:26:26 -0700627 }
Eric W. Biederman15232992012-02-07 15:41:24 -0800628 if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT3_DEF_RESGID)) ||
Miklos Szeredi571beed2007-10-16 23:26:26 -0700629 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
Eric W. Biederman15232992012-02-07 15:41:24 -0800630 seq_printf(seq, ",resgid=%u",
631 from_kgid_munged(&init_user_ns, sbi->s_resgid));
Miklos Szeredi571beed2007-10-16 23:26:26 -0700632 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800633 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredi571beed2007-10-16 23:26:26 -0700634 int def_errors = le16_to_cpu(es->s_errors);
635
636 if (def_errors == EXT3_ERRORS_PANIC ||
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800637 def_errors == EXT3_ERRORS_CONTINUE) {
638 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700639 }
640 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800641 if (test_opt(sb, ERRORS_CONT))
642 seq_puts(seq, ",errors=continue");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700643 if (test_opt(sb, ERRORS_PANIC))
644 seq_puts(seq, ",errors=panic");
645 if (test_opt(sb, NO_UID32))
646 seq_puts(seq, ",nouid32");
647 if (test_opt(sb, DEBUG))
648 seq_puts(seq, ",debug");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700649#ifdef CONFIG_EXT3_FS_XATTR
650 if (test_opt(sb, XATTR_USER))
651 seq_puts(seq, ",user_xattr");
652 if (!test_opt(sb, XATTR_USER) &&
653 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
654 seq_puts(seq, ",nouser_xattr");
655 }
656#endif
657#ifdef CONFIG_EXT3_FS_POSIX_ACL
658 if (test_opt(sb, POSIX_ACL))
659 seq_puts(seq, ",acl");
660 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
661 seq_puts(seq, ",noacl");
662#endif
663 if (!test_opt(sb, RESERVATION))
664 seq_puts(seq, ",noreservation");
665 if (sbi->s_commit_interval) {
666 seq_printf(seq, ",commit=%u",
667 (unsigned) (sbi->s_commit_interval / HZ));
668 }
Eric Sandeen0636c732010-04-30 11:09:34 -0500669
670 /*
671 * Always display barrier state so it's clear what the status is.
672 */
673 seq_puts(seq, ",barrier=");
674 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Dmitry Monakhove3c96432010-02-02 16:05:51 +0300675 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700676 if (test_opt(sb, DATA_ERR_ABORT))
677 seq_puts(seq, ",data_err=abort");
678
Eric Sandeendee1d3b2009-11-16 16:50:49 -0600679 if (test_opt(sb, NOLOAD))
680 seq_puts(seq, ",norecovery");
681
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200682 ext3_show_quota_options(seq, sb);
Mark Bellon8fc27512005-09-06 15:16:54 -0700683
684 return 0;
685}
686
NeilBrownfdb36672006-09-16 12:15:38 -0700687
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700688static struct inode *ext3_nfs_get_inode(struct super_block *sb,
689 u64 ino, u32 generation)
NeilBrownfdb36672006-09-16 12:15:38 -0700690{
NeilBrownfdb36672006-09-16 12:15:38 -0700691 struct inode *inode;
NeilBrownfdb36672006-09-16 12:15:38 -0700692
693 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
694 return ERR_PTR(-ESTALE);
695 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
696 return ERR_PTR(-ESTALE);
697
698 /* iget isn't really right if the inode is currently unallocated!!
699 *
700 * ext3_read_inode will return a bad_inode if the inode had been
701 * deleted, so we should be safe.
702 *
703 * Currently we don't know the generation for parent directory, so
704 * a generation of 0 means "accept any"
705 */
David Howells473043d2008-02-07 00:15:36 -0800706 inode = ext3_iget(sb, ino);
707 if (IS_ERR(inode))
708 return ERR_CAST(inode);
709 if (generation && inode->i_generation != generation) {
NeilBrownfdb36672006-09-16 12:15:38 -0700710 iput(inode);
711 return ERR_PTR(-ESTALE);
712 }
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700713
714 return inode;
715}
716
717static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
718 int fh_len, int fh_type)
719{
720 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
721 ext3_nfs_get_inode);
722}
723
724static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
725 int fh_len, int fh_type)
726{
727 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
728 ext3_nfs_get_inode);
NeilBrownfdb36672006-09-16 12:15:38 -0700729}
730
Toshiyuki Okajima6b082b52009-01-05 22:38:14 -0500731/*
732 * Try to release metadata pages (indirect blocks, directories) which are
733 * mapped via the block device. Since these pages could have journal heads
734 * which would prevent try_to_free_buffers() from freeing them, we must use
735 * jbd layer's try_to_free_buffers() function to release them.
736 */
737static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
738 gfp_t wait)
739{
740 journal_t *journal = EXT3_SB(sb)->s_journal;
741
742 WARN_ON(PageChecked(page));
743 if (!page_has_buffers(page))
744 return 0;
745 if (journal)
746 return journal_try_to_free_buffers(journal, page,
747 wait & ~__GFP_WAIT);
748 return try_to_free_buffers(page);
749}
750
Mark Bellon8fc27512005-09-06 15:16:54 -0700751#ifdef CONFIG_QUOTA
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
753#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
754
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755static int ext3_write_dquot(struct dquot *dquot);
756static int ext3_acquire_dquot(struct dquot *dquot);
757static int ext3_release_dquot(struct dquot *dquot);
758static int ext3_mark_dquot_dirty(struct dquot *dquot);
759static int ext3_write_info(struct super_block *sb, int type);
Jan Kara2fd83a42008-04-28 02:14:34 -0700760static int ext3_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +0200761 struct path *path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762static int ext3_quota_on_mount(struct super_block *sb, int type);
763static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
764 size_t len, loff_t off);
765static ssize_t ext3_quota_write(struct super_block *sb, int type,
766 const char *data, size_t len, loff_t off);
767
Alexey Dobriyan61e225d2009-09-21 17:01:08 -0700768static const struct dquot_operations ext3_quota_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 .write_dquot = ext3_write_dquot,
770 .acquire_dquot = ext3_acquire_dquot,
771 .release_dquot = ext3_release_dquot,
772 .mark_dirty = ext3_mark_dquot_dirty,
Jan Kara157091a2008-11-25 15:31:34 +0100773 .write_info = ext3_write_info,
774 .alloc_dquot = dquot_alloc,
775 .destroy_dquot = dquot_destroy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776};
777
Alexey Dobriyan0d54b212009-09-21 17:01:09 -0700778static const struct quotactl_ops ext3_qctl_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 .quota_on = ext3_quota_on,
Christoph Hellwig287a8092010-05-19 07:16:45 -0400780 .quota_off = dquot_quota_off,
781 .quota_sync = dquot_quota_sync,
782 .get_info = dquot_get_dqinfo,
783 .set_info = dquot_set_dqinfo,
784 .get_dqblk = dquot_get_dqblk,
785 .set_dqblk = dquot_set_dqblk
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786};
787#endif
788
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800789static const struct super_operations ext3_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 .alloc_inode = ext3_alloc_inode,
791 .destroy_inode = ext3_destroy_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 .write_inode = ext3_write_inode,
793 .dirty_inode = ext3_dirty_inode,
Lukas Czerner785c4bc2011-05-23 18:33:01 +0200794 .drop_inode = ext3_drop_inode,
Al Viroac14a952010-06-06 07:08:19 -0400795 .evict_inode = ext3_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 .put_super = ext3_put_super,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 .sync_fs = ext3_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -0800798 .freeze_fs = ext3_freeze,
799 .unfreeze_fs = ext3_unfreeze,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 .statfs = ext3_statfs,
801 .remount_fs = ext3_remount,
Mark Bellon8fc27512005-09-06 15:16:54 -0700802 .show_options = ext3_show_options,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803#ifdef CONFIG_QUOTA
804 .quota_read = ext3_quota_read,
805 .quota_write = ext3_quota_write,
806#endif
Toshiyuki Okajima6b082b52009-01-05 22:38:14 -0500807 .bdev_try_to_free_page = bdev_try_to_free_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808};
809
Christoph Hellwig39655162007-10-21 16:42:17 -0700810static const struct export_operations ext3_export_ops = {
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700811 .fh_to_dentry = ext3_fh_to_dentry,
812 .fh_to_parent = ext3_fh_to_parent,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 .get_parent = ext3_get_parent,
814};
815
816enum {
817 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
818 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Adrian Bunk2860b732005-11-08 21:34:59 -0800819 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Badari Pulavartyade1a292006-06-26 00:25:04 -0700821 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
Johann Lombardi71b96252006-01-08 01:03:20 -0800822 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
Eric Sandeencf7eff42013-07-31 14:33:00 -0500823 Opt_journal_path,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700825 Opt_data_err_abort, Opt_data_err_ignore,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara1aeec432009-11-30 22:22:41 +0100827 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
Eric Sandeen0636c732010-04-30 11:09:34 -0500828 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
829 Opt_resize, Opt_usrquota, Opt_grpquota
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830};
831
Steven Whitehousea447c092008-10-13 10:46:57 +0100832static const match_table_t tokens = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 {Opt_bsd_df, "bsddf"},
834 {Opt_minix_df, "minixdf"},
835 {Opt_grpid, "grpid"},
836 {Opt_grpid, "bsdgroups"},
837 {Opt_nogrpid, "nogrpid"},
838 {Opt_nogrpid, "sysvgroups"},
839 {Opt_resgid, "resgid=%u"},
840 {Opt_resuid, "resuid=%u"},
841 {Opt_sb, "sb=%u"},
842 {Opt_err_cont, "errors=continue"},
843 {Opt_err_panic, "errors=panic"},
844 {Opt_err_ro, "errors=remount-ro"},
845 {Opt_nouid32, "nouid32"},
846 {Opt_nocheck, "nocheck"},
847 {Opt_nocheck, "check=none"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 {Opt_debug, "debug"},
849 {Opt_oldalloc, "oldalloc"},
850 {Opt_orlov, "orlov"},
851 {Opt_user_xattr, "user_xattr"},
852 {Opt_nouser_xattr, "nouser_xattr"},
853 {Opt_acl, "acl"},
854 {Opt_noacl, "noacl"},
855 {Opt_reservation, "reservation"},
856 {Opt_noreservation, "noreservation"},
857 {Opt_noload, "noload"},
Eric Sandeendee1d3b2009-11-16 16:50:49 -0600858 {Opt_noload, "norecovery"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 {Opt_nobh, "nobh"},
Badari Pulavartyade1a292006-06-26 00:25:04 -0700860 {Opt_bh, "bh"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 {Opt_commit, "commit=%u"},
862 {Opt_journal_update, "journal=update"},
863 {Opt_journal_inum, "journal=%u"},
Johann Lombardi71b96252006-01-08 01:03:20 -0800864 {Opt_journal_dev, "journal_dev=%u"},
Eric Sandeencf7eff42013-07-31 14:33:00 -0500865 {Opt_journal_path, "journal_path=%s"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 {Opt_abort, "abort"},
867 {Opt_data_journal, "data=journal"},
868 {Opt_data_ordered, "data=ordered"},
869 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700870 {Opt_data_err_abort, "data_err=abort"},
871 {Opt_data_err_ignore, "data_err=ignore"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 {Opt_offusrjquota, "usrjquota="},
873 {Opt_usrjquota, "usrjquota=%s"},
874 {Opt_offgrpjquota, "grpjquota="},
875 {Opt_grpjquota, "grpjquota=%s"},
876 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
877 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara1aeec432009-11-30 22:22:41 +0100878 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700879 {Opt_grpquota, "grpquota"},
Jan Kara1f545872005-06-23 22:01:04 -0700880 {Opt_noquota, "noquota"},
881 {Opt_quota, "quota"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700882 {Opt_usrquota, "usrquota"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 {Opt_barrier, "barrier=%u"},
Eric Sandeen0636c732010-04-30 11:09:34 -0500884 {Opt_barrier, "barrier"},
885 {Opt_nobarrier, "nobarrier"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 {Opt_resize, "resize"},
Josef Bacik925872162008-03-04 14:29:43 -0800887 {Opt_err, NULL},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888};
889
Alexey Fisher4cf46b62009-11-22 20:38:55 +0100890static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891{
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700892 ext3_fsblk_t sb_block;
893 char *options = (char *) *data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
895 if (!options || strncmp(options, "sb=", 3) != 0)
896 return 1; /* Default location */
897 options += 3;
Mingming Cao43d23f92006-06-25 05:48:07 -0700898 /*todo: use simple_strtoll with >32bit ext3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 sb_block = simple_strtoul(options, &options, 0);
900 if (*options && *options != ',') {
Lars-Peter Clausen8d0c2d12013-03-09 15:28:44 +0100901 ext3_msg(sb, KERN_ERR, "error: invalid sb specification: %s",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 (char *) *data);
903 return 1;
904 }
905 if (*options == ',')
906 options++;
907 *data = (void *) options;
908 return sb_block;
909}
910
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300911#ifdef CONFIG_QUOTA
912static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
913{
914 struct ext3_sb_info *sbi = EXT3_SB(sb);
915 char *qname;
916
917 if (sb_any_quota_loaded(sb) &&
918 !sbi->s_qf_names[qtype]) {
919 ext3_msg(sb, KERN_ERR,
920 "Cannot change journaled "
921 "quota options when quota turned on");
922 return 0;
923 }
924 qname = match_strdup(args);
925 if (!qname) {
926 ext3_msg(sb, KERN_ERR,
927 "Not enough memory for storing quotafile name");
928 return 0;
929 }
Jan Karaf56426a2012-12-31 12:38:36 +0100930 if (sbi->s_qf_names[qtype]) {
931 int same = !strcmp(sbi->s_qf_names[qtype], qname);
932
933 kfree(qname);
934 if (!same) {
935 ext3_msg(sb, KERN_ERR,
936 "%s quota file already specified",
937 QTYPE2NAME(qtype));
938 }
939 return same;
940 }
941 if (strchr(qname, '/')) {
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300942 ext3_msg(sb, KERN_ERR,
Jan Karaf56426a2012-12-31 12:38:36 +0100943 "quotafile must be on filesystem root");
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300944 kfree(qname);
945 return 0;
946 }
947 sbi->s_qf_names[qtype] = qname;
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300948 set_opt(sbi->s_mount_opt, QUOTA);
949 return 1;
950}
951
952static int clear_qf_name(struct super_block *sb, int qtype) {
953
954 struct ext3_sb_info *sbi = EXT3_SB(sb);
955
956 if (sb_any_quota_loaded(sb) &&
957 sbi->s_qf_names[qtype]) {
958 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
959 " when quota turned on");
960 return 0;
961 }
Jan Karaf56426a2012-12-31 12:38:36 +0100962 if (sbi->s_qf_names[qtype]) {
963 kfree(sbi->s_qf_names[qtype]);
964 sbi->s_qf_names[qtype] = NULL;
965 }
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300966 return 1;
967}
968#endif
969
Johann Lombardi71b96252006-01-08 01:03:20 -0800970static int parse_options (char *options, struct super_block *sb,
Eric Sandeeneee194e2006-09-27 01:49:30 -0700971 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao43d23f92006-06-25 05:48:07 -0700972 ext3_fsblk_t *n_blocks_count, int is_remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973{
974 struct ext3_sb_info *sbi = EXT3_SB(sb);
975 char * p;
976 substring_t args[MAX_OPT_ARGS];
977 int data_opt = 0;
978 int option;
Eric W. Biederman15232992012-02-07 15:41:24 -0800979 kuid_t uid;
980 kgid_t gid;
Eric Sandeencf7eff42013-07-31 14:33:00 -0500981 char *journal_path;
982 struct inode *journal_inode;
983 struct path path;
984 int error;
985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986#ifdef CONFIG_QUOTA
Dmitry Monakhove1f5c672010-02-02 16:05:53 +0300987 int qfmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988#endif
989
990 if (!options)
991 return 1;
992
993 while ((p = strsep (&options, ",")) != NULL) {
994 int token;
995 if (!*p)
996 continue;
Eric Sandeen0636c732010-04-30 11:09:34 -0500997 /*
998 * Initialize args struct so we know whether arg was
999 * found; some options take optional arguments.
1000 */
Sachin Kamat76f59b32012-09-04 13:42:00 +05301001 args[0].to = args[0].from = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 token = match_token(p, tokens, args);
1003 switch (token) {
1004 case Opt_bsd_df:
1005 clear_opt (sbi->s_mount_opt, MINIX_DF);
1006 break;
1007 case Opt_minix_df:
1008 set_opt (sbi->s_mount_opt, MINIX_DF);
1009 break;
1010 case Opt_grpid:
1011 set_opt (sbi->s_mount_opt, GRPID);
1012 break;
1013 case Opt_nogrpid:
1014 clear_opt (sbi->s_mount_opt, GRPID);
1015 break;
1016 case Opt_resuid:
1017 if (match_int(&args[0], &option))
1018 return 0;
Eric W. Biederman15232992012-02-07 15:41:24 -08001019 uid = make_kuid(current_user_ns(), option);
1020 if (!uid_valid(uid)) {
1021 ext3_msg(sb, KERN_ERR, "Invalid uid value %d", option);
Zhao Hongjiang66415f62012-10-09 13:48:47 +08001022 return 0;
Eric W. Biederman15232992012-02-07 15:41:24 -08001023
1024 }
1025 sbi->s_resuid = uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 break;
1027 case Opt_resgid:
1028 if (match_int(&args[0], &option))
1029 return 0;
Eric W. Biederman15232992012-02-07 15:41:24 -08001030 gid = make_kgid(current_user_ns(), option);
1031 if (!gid_valid(gid)) {
1032 ext3_msg(sb, KERN_ERR, "Invalid gid value %d", option);
Zhao Hongjiang66415f62012-10-09 13:48:47 +08001033 return 0;
Eric W. Biederman15232992012-02-07 15:41:24 -08001034 }
1035 sbi->s_resgid = gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 break;
1037 case Opt_sb:
1038 /* handled by get_sb_block() instead of here */
1039 /* *sb_block = match_int(&args[0]); */
1040 break;
1041 case Opt_err_panic:
1042 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1043 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1044 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1045 break;
1046 case Opt_err_ro:
1047 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1048 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1049 set_opt (sbi->s_mount_opt, ERRORS_RO);
1050 break;
1051 case Opt_err_cont:
1052 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1053 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1054 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1055 break;
1056 case Opt_nouid32:
1057 set_opt (sbi->s_mount_opt, NO_UID32);
1058 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 case Opt_nocheck:
1060 clear_opt (sbi->s_mount_opt, CHECK);
1061 break;
1062 case Opt_debug:
1063 set_opt (sbi->s_mount_opt, DEBUG);
1064 break;
1065 case Opt_oldalloc:
Lukas Czernerfbc85402011-08-16 18:08:06 +02001066 ext3_msg(sb, KERN_WARNING,
1067 "Ignoring deprecated oldalloc option");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 break;
1069 case Opt_orlov:
Lukas Czernerfbc85402011-08-16 18:08:06 +02001070 ext3_msg(sb, KERN_WARNING,
1071 "Ignoring deprecated orlov option");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 break;
1073#ifdef CONFIG_EXT3_FS_XATTR
1074 case Opt_user_xattr:
1075 set_opt (sbi->s_mount_opt, XATTR_USER);
1076 break;
1077 case Opt_nouser_xattr:
1078 clear_opt (sbi->s_mount_opt, XATTR_USER);
1079 break;
1080#else
1081 case Opt_user_xattr:
1082 case Opt_nouser_xattr:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001083 ext3_msg(sb, KERN_INFO,
1084 "(no)user_xattr options not supported");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 break;
1086#endif
1087#ifdef CONFIG_EXT3_FS_POSIX_ACL
1088 case Opt_acl:
1089 set_opt(sbi->s_mount_opt, POSIX_ACL);
1090 break;
1091 case Opt_noacl:
1092 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1093 break;
1094#else
1095 case Opt_acl:
1096 case Opt_noacl:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001097 ext3_msg(sb, KERN_INFO,
1098 "(no)acl options not supported");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 break;
1100#endif
1101 case Opt_reservation:
1102 set_opt(sbi->s_mount_opt, RESERVATION);
1103 break;
1104 case Opt_noreservation:
1105 clear_opt(sbi->s_mount_opt, RESERVATION);
1106 break;
1107 case Opt_journal_update:
1108 /* @@@ FIXME */
1109 /* Eventually we will want to be able to create
1110 a journal file here. For now, only allow the
1111 user to specify an existing inode to be the
1112 journal file. */
1113 if (is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001114 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1115 "journal on remount");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 return 0;
1117 }
1118 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1119 break;
1120 case Opt_journal_inum:
1121 if (is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001122 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1123 "journal on remount");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 return 0;
1125 }
1126 if (match_int(&args[0], &option))
1127 return 0;
1128 *inum = option;
1129 break;
Johann Lombardi71b96252006-01-08 01:03:20 -08001130 case Opt_journal_dev:
1131 if (is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001132 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1133 "journal on remount");
Johann Lombardi71b96252006-01-08 01:03:20 -08001134 return 0;
1135 }
1136 if (match_int(&args[0], &option))
1137 return 0;
1138 *journal_devnum = option;
1139 break;
Eric Sandeencf7eff42013-07-31 14:33:00 -05001140 case Opt_journal_path:
1141 if (is_remount) {
1142 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1143 "journal on remount");
1144 return 0;
1145 }
1146
1147 journal_path = match_strdup(&args[0]);
1148 if (!journal_path) {
1149 ext3_msg(sb, KERN_ERR, "error: could not dup "
1150 "journal device string");
1151 return 0;
1152 }
1153
1154 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1155 if (error) {
1156 ext3_msg(sb, KERN_ERR, "error: could not find "
1157 "journal device path: error %d", error);
1158 kfree(journal_path);
1159 return 0;
1160 }
1161
1162 journal_inode = path.dentry->d_inode;
1163 if (!S_ISBLK(journal_inode->i_mode)) {
1164 ext3_msg(sb, KERN_ERR, "error: journal path %s "
1165 "is not a block device", journal_path);
1166 path_put(&path);
1167 kfree(journal_path);
1168 return 0;
1169 }
1170
1171 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1172 path_put(&path);
1173 kfree(journal_path);
1174 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 case Opt_noload:
1176 set_opt (sbi->s_mount_opt, NOLOAD);
1177 break;
1178 case Opt_commit:
1179 if (match_int(&args[0], &option))
1180 return 0;
1181 if (option < 0)
1182 return 0;
1183 if (option == 0)
1184 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1185 sbi->s_commit_interval = HZ * option;
1186 break;
1187 case Opt_data_journal:
1188 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1189 goto datacheck;
1190 case Opt_data_ordered:
1191 data_opt = EXT3_MOUNT_ORDERED_DATA;
1192 goto datacheck;
1193 case Opt_data_writeback:
1194 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1195 datacheck:
1196 if (is_remount) {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001197 if (test_opt(sb, DATA_FLAGS) == data_opt)
Jan Kara3c4cec62009-08-24 16:38:43 +02001198 break;
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001199 ext3_msg(sb, KERN_ERR,
1200 "error: cannot change "
Jan Kara3c4cec62009-08-24 16:38:43 +02001201 "data mode on remount. The filesystem "
1202 "is mounted in data=%s mode and you "
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001203 "try to remount it in data=%s mode.",
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001204 data_mode_string(test_opt(sb,
1205 DATA_FLAGS)),
Jan Kara3c4cec62009-08-24 16:38:43 +02001206 data_mode_string(data_opt));
1207 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 } else {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001209 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 sbi->s_mount_opt |= data_opt;
1211 }
1212 break;
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -07001213 case Opt_data_err_abort:
1214 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1215 break;
1216 case Opt_data_err_ignore:
1217 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1218 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219#ifdef CONFIG_QUOTA
1220 case Opt_usrjquota:
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001221 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1222 return 0;
1223 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 case Opt_grpjquota:
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001225 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 break;
1228 case Opt_offusrjquota:
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001229 if (!clear_qf_name(sb, USRQUOTA))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 return 0;
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001231 break;
1232 case Opt_offgrpjquota:
1233 if (!clear_qf_name(sb, GRPQUOTA))
1234 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 break;
1236 case Opt_jqfmt_vfsold:
Jan Karad06bf1d2008-07-25 01:46:18 -07001237 qfmt = QFMT_VFS_OLD;
1238 goto set_qf_format;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 case Opt_jqfmt_vfsv0:
Jan Karad06bf1d2008-07-25 01:46:18 -07001240 qfmt = QFMT_VFS_V0;
Jan Kara1aeec432009-11-30 22:22:41 +01001241 goto set_qf_format;
1242 case Opt_jqfmt_vfsv1:
1243 qfmt = QFMT_VFS_V1;
Jan Karad06bf1d2008-07-25 01:46:18 -07001244set_qf_format:
Jan Karaee0d5ff2008-08-20 18:11:50 +02001245 if (sb_any_quota_loaded(sb) &&
Jan Karad06bf1d2008-07-25 01:46:18 -07001246 sbi->s_jquota_fmt != qfmt) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001247 ext3_msg(sb, KERN_ERR, "error: cannot change "
Jan Karad06bf1d2008-07-25 01:46:18 -07001248 "journaled quota options when "
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001249 "quota turned on.");
Jan Karad06bf1d2008-07-25 01:46:18 -07001250 return 0;
1251 }
1252 sbi->s_jquota_fmt = qfmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 break;
Jan Kara1f545872005-06-23 22:01:04 -07001254 case Opt_quota:
Mark Bellon8fc27512005-09-06 15:16:54 -07001255 case Opt_usrquota:
Jan Kara1f545872005-06-23 22:01:04 -07001256 set_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001257 set_opt(sbi->s_mount_opt, USRQUOTA);
1258 break;
1259 case Opt_grpquota:
1260 set_opt(sbi->s_mount_opt, QUOTA);
1261 set_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001262 break;
1263 case Opt_noquota:
Jan Karaee0d5ff2008-08-20 18:11:50 +02001264 if (sb_any_quota_loaded(sb)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001265 ext3_msg(sb, KERN_ERR, "error: cannot change "
1266 "quota options when quota turned on.");
Jan Kara1f545872005-06-23 22:01:04 -07001267 return 0;
1268 }
1269 clear_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001270 clear_opt(sbi->s_mount_opt, USRQUOTA);
1271 clear_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001272 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273#else
Mark Bellon8fc27512005-09-06 15:16:54 -07001274 case Opt_quota:
1275 case Opt_usrquota:
1276 case Opt_grpquota:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001277 ext3_msg(sb, KERN_ERR,
1278 "error: quota options not supported.");
Jan Karafa1ff1e2008-04-28 02:16:14 -07001279 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 case Opt_usrjquota:
1281 case Opt_grpjquota:
1282 case Opt_offusrjquota:
1283 case Opt_offgrpjquota:
1284 case Opt_jqfmt_vfsold:
1285 case Opt_jqfmt_vfsv0:
Jan Kara1aeec432009-11-30 22:22:41 +01001286 case Opt_jqfmt_vfsv1:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001287 ext3_msg(sb, KERN_ERR,
1288 "error: journaled quota options not "
1289 "supported.");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 break;
Jan Kara1f545872005-06-23 22:01:04 -07001291 case Opt_noquota:
1292 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293#endif
1294 case Opt_abort:
1295 set_opt(sbi->s_mount_opt, ABORT);
1296 break;
Eric Sandeen0636c732010-04-30 11:09:34 -05001297 case Opt_nobarrier:
1298 clear_opt(sbi->s_mount_opt, BARRIER);
1299 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 case Opt_barrier:
Eric Sandeen0636c732010-04-30 11:09:34 -05001301 if (args[0].from) {
1302 if (match_int(&args[0], &option))
1303 return 0;
1304 } else
1305 option = 1; /* No argument, default to 1 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 if (option)
1307 set_opt(sbi->s_mount_opt, BARRIER);
1308 else
1309 clear_opt(sbi->s_mount_opt, BARRIER);
1310 break;
1311 case Opt_ignore:
1312 break;
1313 case Opt_resize:
Jan Kara08c6a962005-07-12 13:58:28 -07001314 if (!is_remount) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001315 ext3_msg(sb, KERN_ERR,
1316 "error: resize option only available "
1317 "for remount");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 return 0;
1319 }
KAMBAROV, ZAURc7f17212005-06-28 20:45:11 -07001320 if (match_int(&args[0], &option) != 0)
1321 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 *n_blocks_count = option;
1323 break;
1324 case Opt_nobh:
Christoph Hellwig4c4d3902010-06-07 10:20:39 +02001325 ext3_msg(sb, KERN_WARNING,
1326 "warning: ignoring deprecated nobh option");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 break;
Badari Pulavartyade1a292006-06-26 00:25:04 -07001328 case Opt_bh:
Christoph Hellwig4c4d3902010-06-07 10:20:39 +02001329 ext3_msg(sb, KERN_WARNING,
1330 "warning: ignoring deprecated bh option");
Badari Pulavartyade1a292006-06-26 00:25:04 -07001331 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 default:
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001333 ext3_msg(sb, KERN_ERR,
1334 "error: unrecognized mount option \"%s\" "
1335 "or missing value", p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 return 0;
1337 }
1338 }
1339#ifdef CONFIG_QUOTA
Mark Bellon8fc27512005-09-06 15:16:54 -07001340 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001341 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Mark Bellon8fc27512005-09-06 15:16:54 -07001342 clear_opt(sbi->s_mount_opt, USRQUOTA);
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001343 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Mark Bellon8fc27512005-09-06 15:16:54 -07001344 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1345
Dmitry Monakhove1f5c672010-02-02 16:05:53 +03001346 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001347 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1348 "format mixing.");
Mark Bellon8fc27512005-09-06 15:16:54 -07001349 return 0;
1350 }
1351
1352 if (!sbi->s_jquota_fmt) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001353 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1354 "not specified.");
Mark Bellon8fc27512005-09-06 15:16:54 -07001355 return 0;
1356 }
1357 } else {
1358 if (sbi->s_jquota_fmt) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001359 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
Jan Kara99aeaf62008-07-25 01:46:17 -07001360 "specified with no journaling "
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001361 "enabled.");
Mark Bellon8fc27512005-09-06 15:16:54 -07001362 return 0;
1363 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 }
1365#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 return 1;
1367}
1368
1369static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1370 int read_only)
1371{
1372 struct ext3_sb_info *sbi = EXT3_SB(sb);
1373 int res = 0;
1374
1375 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001376 ext3_msg(sb, KERN_ERR,
1377 "error: revision level too high, "
1378 "forcing read-only mode");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 res = MS_RDONLY;
1380 }
1381 if (read_only)
1382 return res;
1383 if (!(sbi->s_mount_state & EXT3_VALID_FS))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001384 ext3_msg(sb, KERN_WARNING,
1385 "warning: mounting unchecked fs, "
1386 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001388 ext3_msg(sb, KERN_WARNING,
1389 "warning: mounting fs with errors, "
1390 "running e2fsck is recommended");
Namhyung Kimdf0d6b82010-10-10 21:36:59 +09001391 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 le16_to_cpu(es->s_mnt_count) >=
Namhyung Kimdf0d6b82010-10-10 21:36:59 +09001393 le16_to_cpu(es->s_max_mnt_count))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001394 ext3_msg(sb, KERN_WARNING,
1395 "warning: maximal mount count reached, "
1396 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 else if (le32_to_cpu(es->s_checkinterval) &&
1398 (le32_to_cpu(es->s_lastcheck) +
1399 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001400 ext3_msg(sb, KERN_WARNING,
1401 "warning: checktime reached, "
1402 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403#if 0
1404 /* @@@ We _will_ want to clear the valid bit if we find
1405 inconsistencies, to force a fsck at reboot. But for
1406 a plain journaled filesystem we can keep it set as
1407 valid forever! :) */
Marcin Slusarze7f23eb2008-04-28 02:16:06 -07001408 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409#endif
Namhyung Kimdf0d6b82010-10-10 21:36:59 +09001410 if (!le16_to_cpu(es->s_max_mnt_count))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
Marcin Slusarz50e8a282008-02-08 04:20:13 -08001412 le16_add_cpu(&es->s_mnt_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 es->s_mtime = cpu_to_le32(get_seconds());
1414 ext3_update_dynamic_rev(sb);
1415 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1416
1417 ext3_commit_super(sb, es, 1);
1418 if (test_opt(sb, DEBUG))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001419 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1420 "bpg=%lu, ipg=%lu, mo=%04lx]",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 sb->s_blocksize,
1422 sbi->s_groups_count,
1423 EXT3_BLOCKS_PER_GROUP(sb),
1424 EXT3_INODES_PER_GROUP(sb),
1425 sbi->s_mount_opt);
1426
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1428 char b[BDEVNAME_SIZE];
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001429 ext3_msg(sb, KERN_INFO, "using external journal on %s",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1431 } else {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001432 ext3_msg(sb, KERN_INFO, "using internal journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 }
Dan Magenheimerd71bc6d2011-05-26 10:01:49 -06001434 cleancache_init_fs(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 return res;
1436}
1437
1438/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001439static int ext3_check_descriptors(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440{
1441 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 int i;
1443
1444 ext3_debug ("Checking group descriptors");
1445
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001446 for (i = 0; i < sbi->s_groups_count; i++) {
1447 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
Akinobu Mita1eaafea2008-04-28 02:16:07 -07001448 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1449 ext3_fsblk_t last_block;
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001450
Eric Sandeen855565e2006-09-27 01:49:29 -07001451 if (i == sbi->s_groups_count - 1)
1452 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1453 else
1454 last_block = first_block +
1455 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1456
Eric Sandeen855565e2006-09-27 01:49:29 -07001457 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1458 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 {
1460 ext3_error (sb, "ext3_check_descriptors",
1461 "Block bitmap for group %d"
1462 " not in group (block %lu)!",
1463 i, (unsigned long)
1464 le32_to_cpu(gdp->bg_block_bitmap));
1465 return 0;
1466 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001467 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1468 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 {
1470 ext3_error (sb, "ext3_check_descriptors",
1471 "Inode bitmap for group %d"
1472 " not in group (block %lu)!",
1473 i, (unsigned long)
1474 le32_to_cpu(gdp->bg_inode_bitmap));
1475 return 0;
1476 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001477 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
Eric Sandeen780dcdb2007-07-26 10:41:11 -07001478 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
Eric Sandeen855565e2006-09-27 01:49:29 -07001479 last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 {
1481 ext3_error (sb, "ext3_check_descriptors",
1482 "Inode table for group %d"
1483 " not in group (block %lu)!",
1484 i, (unsigned long)
1485 le32_to_cpu(gdp->bg_inode_table));
1486 return 0;
1487 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 }
1489
1490 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1491 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1492 return 1;
1493}
1494
1495
1496/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1497 * the superblock) which were deleted from all directories, but held open by
1498 * a process at the time of a crash. We walk the list and try to delete these
1499 * inodes at recovery time (only with a read-write filesystem).
1500 *
1501 * In order to keep the orphan inode chain consistent during traversal (in
1502 * case of crash during recovery), we link each inode into the superblock
1503 * orphan list_head and handle it the same way as an inode deletion during
1504 * normal operation (which journals the operations for us).
1505 *
1506 * We only do an iget() and an iput() on each inode, which is very safe if we
1507 * accidentally point at an in-use or already deleted inode. The worst that
1508 * can happen in this case is that we get a "bit already cleared" message from
1509 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1510 * e2fsck was run on this filesystem, and it must have already done the orphan
1511 * inode cleanup for us, so we can safely abort without any further action.
1512 */
1513static void ext3_orphan_cleanup (struct super_block * sb,
1514 struct ext3_super_block * es)
1515{
1516 unsigned int s_flags = sb->s_flags;
1517 int nr_orphans = 0, nr_truncates = 0;
1518#ifdef CONFIG_QUOTA
1519 int i;
1520#endif
1521 if (!es->s_last_orphan) {
1522 jbd_debug(4, "no orphan inodes to clean up\n");
1523 return;
1524 }
1525
Eric Sandeena8f48a92006-12-06 20:40:13 -08001526 if (bdev_read_only(sb->s_bdev)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001527 ext3_msg(sb, KERN_ERR, "error: write access "
1528 "unavailable, skipping orphan cleanup.");
Eric Sandeena8f48a92006-12-06 20:40:13 -08001529 return;
1530 }
1531
Amir Goldsteince654b32011-02-26 22:40:19 +02001532 /* Check if feature set allows readwrite operations */
1533 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1534 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1535 "unknown ROCOMPAT features");
1536 return;
1537 }
1538
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
Eric Sandeene4230102012-08-27 14:30:40 -05001540 /* don't clear list on RO mount w/ errors */
1541 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 jbd_debug(1, "Errors on filesystem, "
1543 "clearing orphan list.\n");
Eric Sandeene4230102012-08-27 14:30:40 -05001544 es->s_last_orphan = 0;
1545 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1547 return;
1548 }
1549
1550 if (s_flags & MS_RDONLY) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001551 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 sb->s_flags &= ~MS_RDONLY;
1553 }
1554#ifdef CONFIG_QUOTA
1555 /* Needed for iput() to work correctly and not trash data */
1556 sb->s_flags |= MS_ACTIVE;
1557 /* Turn on quotas so that they are updated correctly */
1558 for (i = 0; i < MAXQUOTAS; i++) {
1559 if (EXT3_SB(sb)->s_qf_names[i]) {
1560 int ret = ext3_quota_on_mount(sb, i);
1561 if (ret < 0)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001562 ext3_msg(sb, KERN_ERR,
1563 "error: cannot turn on journaled "
1564 "quota: %d", ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 }
1566 }
1567#endif
1568
1569 while (es->s_last_orphan) {
1570 struct inode *inode;
1571
David Howells473043d2008-02-07 00:15:36 -08001572 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1573 if (IS_ERR(inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 es->s_last_orphan = 0;
1575 break;
1576 }
1577
1578 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05001579 dquot_initialize(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 if (inode->i_nlink) {
1581 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001582 "%s: truncating inode %lu to %Ld bytes\n",
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001583 __func__, inode->i_ino, inode->i_size);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001584 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 inode->i_ino, inode->i_size);
1586 ext3_truncate(inode);
1587 nr_truncates++;
1588 } else {
1589 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001590 "%s: deleting unreferenced inode %lu\n",
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001591 __func__, inode->i_ino);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001592 jbd_debug(2, "deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 inode->i_ino);
1594 nr_orphans++;
1595 }
1596 iput(inode); /* The delete magic happens here! */
1597 }
1598
1599#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1600
1601 if (nr_orphans)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001602 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1603 PLURAL(nr_orphans));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 if (nr_truncates)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001605 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1606 PLURAL(nr_truncates));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607#ifdef CONFIG_QUOTA
1608 /* Turn quotas off */
1609 for (i = 0; i < MAXQUOTAS; i++) {
1610 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04001611 dquot_quota_off(sb, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 }
1613#endif
1614 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1615}
1616
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617/*
1618 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1619 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1620 * We need to be 1 filesystem block less than the 2^32 sector limit.
1621 */
1622static loff_t ext3_max_size(int bits)
1623{
1624 loff_t res = EXT3_NDIR_BLOCKS;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001625 int meta_blocks;
1626 loff_t upper_limit;
1627
1628 /* This is calculated to be the largest file size for a
1629 * dense, file such that the total number of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 * sectors in the file, including data and all indirect blocks,
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001631 * does not exceed 2^32 -1
1632 * __u32 i_blocks representing the total number of
1633 * 512 bytes blocks of the file
1634 */
1635 upper_limit = (1LL << 32) - 1;
1636
1637 /* total blocks in file system block size */
1638 upper_limit >>= (bits - 9);
1639
1640
1641 /* indirect blocks */
1642 meta_blocks = 1;
1643 /* double indirect blocks */
1644 meta_blocks += 1 + (1LL << (bits-2));
1645 /* tripple indirect blocks */
1646 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1647
1648 upper_limit -= meta_blocks;
1649 upper_limit <<= bits;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650
1651 res += 1LL << (bits-2);
1652 res += 1LL << (2*(bits-2));
1653 res += 1LL << (3*(bits-2));
1654 res <<= bits;
1655 if (res > upper_limit)
1656 res = upper_limit;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001657
1658 if (res > MAX_LFS_FILESIZE)
1659 res = MAX_LFS_FILESIZE;
1660
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 return res;
1662}
1663
Mingming Cao43d23f92006-06-25 05:48:07 -07001664static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1665 ext3_fsblk_t logic_sb_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 int nr)
1667{
1668 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001669 unsigned long bg, first_meta_bg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 int has_super = 0;
1671
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1673
1674 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1675 nr < first_meta_bg)
1676 return (logic_sb_block + nr + 1);
1677 bg = sbi->s_desc_per_block * nr;
1678 if (ext3_bg_has_super(sb, bg))
1679 has_super = 1;
Mingming Cao43d23f92006-06-25 05:48:07 -07001680 return (has_super + ext3_group_first_block_no(sb, bg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681}
1682
1683
1684static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1685{
1686 struct buffer_head * bh;
1687 struct ext3_super_block *es = NULL;
1688 struct ext3_sb_info *sbi;
Mingming Cao43d23f92006-06-25 05:48:07 -07001689 ext3_fsblk_t block;
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001690 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001691 ext3_fsblk_t logic_sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 unsigned long offset = 0;
Eric Sandeeneee194e2006-09-27 01:49:30 -07001693 unsigned int journal_inum = 0;
Johann Lombardi71b96252006-01-08 01:03:20 -08001694 unsigned long journal_devnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 unsigned long def_mount_opts;
1696 struct inode *root;
1697 int blocksize;
1698 int hblock;
1699 int db_count;
1700 int i;
1701 int needs_recovery;
David Howells473043d2008-02-07 00:15:36 -08001702 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 __le32 features;
Peter Zijlstra833f4072007-10-16 23:25:45 -07001704 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Panagiotis Issarisf8314dc2006-09-27 01:49:37 -07001706 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 if (!sbi)
1708 return -ENOMEM;
Pekka Enberg5df096d2009-01-07 18:07:25 -08001709
1710 sbi->s_blockgroup_lock =
1711 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1712 if (!sbi->s_blockgroup_lock) {
1713 kfree(sbi);
1714 return -ENOMEM;
1715 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 sb->s_fs_info = sbi;
Miklos Szeredi571beed2007-10-16 23:26:26 -07001717 sbi->s_sb_block = sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1720 if (!blocksize) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001721 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 goto out_fail;
1723 }
1724
1725 /*
1726 * The ext3 superblock will not be buffer aligned for other than 1kB
1727 * block sizes. We need to calculate the offset from buffer start.
1728 */
1729 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1730 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1731 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1732 } else {
1733 logic_sb_block = sb_block;
1734 }
1735
1736 if (!(bh = sb_bread(sb, logic_sb_block))) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001737 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 goto out_fail;
1739 }
1740 /*
1741 * Note: s_es must be initialized as soon as possible because
1742 * some ext3 macro-instructions depend on its value
1743 */
Namhyung Kim57e94d82010-10-11 02:10:45 +09001744 es = (struct ext3_super_block *) (bh->b_data + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 sbi->s_es = es;
1746 sb->s_magic = le16_to_cpu(es->s_magic);
1747 if (sb->s_magic != EXT3_SUPER_MAGIC)
1748 goto cantfind_ext3;
1749
1750 /* Set defaults before we parse the mount options */
1751 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1752 if (def_mount_opts & EXT3_DEFM_DEBUG)
1753 set_opt(sbi->s_mount_opt, DEBUG);
1754 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1755 set_opt(sbi->s_mount_opt, GRPID);
1756 if (def_mount_opts & EXT3_DEFM_UID16)
1757 set_opt(sbi->s_mount_opt, NO_UID32);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001758#ifdef CONFIG_EXT3_FS_XATTR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1760 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001761#endif
1762#ifdef CONFIG_EXT3_FS_POSIX_ACL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 if (def_mount_opts & EXT3_DEFM_ACL)
1764 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001765#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001767 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001769 set_opt(sbi->s_mount_opt, ORDERED_DATA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001771 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
1773 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1774 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001775 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
Dmitry Mishin2245d7c2006-10-11 01:21:49 -07001776 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001777 else
1778 set_opt(sbi->s_mount_opt, ERRORS_RO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
Eric W. Biederman15232992012-02-07 15:41:24 -08001780 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
1781 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782
Christoph Hellwig00eacd62011-07-16 16:46:50 -04001783 /* enable barriers by default */
1784 set_opt(sbi->s_mount_opt, BARRIER);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 set_opt(sbi->s_mount_opt, RESERVATION);
1786
Johann Lombardi71b96252006-01-08 01:03:20 -08001787 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1788 NULL, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789 goto failed_mount;
1790
1791 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhove3c96432010-02-02 16:05:51 +03001792 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793
1794 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1795 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1796 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1797 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001798 ext3_msg(sb, KERN_WARNING,
1799 "warning: feature flags set on rev 0 fs, "
1800 "running e2fsck is recommended");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 /*
1802 * Check feature flags regardless of the revision level, since we
1803 * previously didn't change the revision level when setting the flags,
1804 * so there is a chance incompat flags are set on a rev 0 filesystem.
1805 */
1806 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1807 if (features) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001808 ext3_msg(sb, KERN_ERR,
1809 "error: couldn't mount because of unsupported "
1810 "optional features (%x)", le32_to_cpu(features));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 goto failed_mount;
1812 }
1813 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1814 if (!(sb->s_flags & MS_RDONLY) && features) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001815 ext3_msg(sb, KERN_ERR,
1816 "error: couldn't mount RDWR because of unsupported "
1817 "optional features (%x)", le32_to_cpu(features));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 goto failed_mount;
1819 }
1820 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1821
1822 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1823 blocksize > EXT3_MAX_BLOCK_SIZE) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001824 ext3_msg(sb, KERN_ERR,
1825 "error: couldn't mount because of unsupported "
1826 "filesystem blocksize %d", blocksize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 goto failed_mount;
1828 }
1829
Martin K. Petersene1defc42009-05-22 17:17:49 -04001830 hblock = bdev_logical_block_size(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 if (sb->s_blocksize != blocksize) {
1832 /*
1833 * Make sure the blocksize for the filesystem is larger
1834 * than the hardware sectorsize for the machine.
1835 */
1836 if (blocksize < hblock) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001837 ext3_msg(sb, KERN_ERR,
1838 "error: fsblocksize %d too small for "
1839 "hardware sectorsize %d", blocksize, hblock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 goto failed_mount;
1841 }
1842
1843 brelse (bh);
Takashi Sato0f0a89e2007-10-18 03:06:56 -07001844 if (!sb_set_blocksize(sb, blocksize)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001845 ext3_msg(sb, KERN_ERR,
1846 "error: bad blocksize %d", blocksize);
Takashi Sato0f0a89e2007-10-18 03:06:56 -07001847 goto out_fail;
1848 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1850 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1851 bh = sb_bread(sb, logic_sb_block);
1852 if (!bh) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001853 ext3_msg(sb, KERN_ERR,
1854 "error: can't read superblock on 2nd try");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 goto failed_mount;
1856 }
Namhyung Kim57e94d82010-10-11 02:10:45 +09001857 es = (struct ext3_super_block *)(bh->b_data + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 sbi->s_es = es;
1859 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001860 ext3_msg(sb, KERN_ERR,
1861 "error: magic mismatch");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 goto failed_mount;
1863 }
1864 }
1865
1866 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1867
1868 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1869 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1870 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1871 } else {
1872 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1873 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1874 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
vignesh babu3fc74262007-07-15 23:41:17 -07001875 (!is_power_of_2(sbi->s_inode_size)) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876 (sbi->s_inode_size > blocksize)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001877 ext3_msg(sb, KERN_ERR,
1878 "error: unsupported inode size: %d",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 sbi->s_inode_size);
1880 goto failed_mount;
1881 }
1882 }
1883 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1884 le32_to_cpu(es->s_log_frag_size);
1885 if (blocksize != sbi->s_frag_size) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001886 ext3_msg(sb, KERN_ERR,
1887 "error: fragsize %lu != blocksize %u (unsupported)",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 sbi->s_frag_size, blocksize);
1889 goto failed_mount;
1890 }
1891 sbi->s_frags_per_block = 1;
1892 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1893 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1894 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08001895 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 goto cantfind_ext3;
1897 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1898 if (sbi->s_inodes_per_block == 0)
1899 goto cantfind_ext3;
1900 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1901 sbi->s_inodes_per_block;
1902 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1903 sbi->s_sbh = bh;
1904 sbi->s_mount_state = le16_to_cpu(es->s_state);
David Howellsf0d1b0b2006-12-08 02:37:49 -08001905 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1906 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 for (i=0; i < 4; i++)
1908 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1909 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'o5e1f8c92008-10-28 13:21:55 -04001910 i = le32_to_cpu(es->s_flags);
1911 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1912 sbi->s_hash_unsigned = 3;
1913 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1914#ifdef __CHAR_UNSIGNED__
1915 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1916 sbi->s_hash_unsigned = 3;
1917#else
1918 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1919#endif
Theodore Ts'o5e1f8c92008-10-28 13:21:55 -04001920 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
1922 if (sbi->s_blocks_per_group > blocksize * 8) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001923 ext3_msg(sb, KERN_ERR,
1924 "#blocks per group too big: %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 sbi->s_blocks_per_group);
1926 goto failed_mount;
1927 }
1928 if (sbi->s_frags_per_group > blocksize * 8) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001929 ext3_msg(sb, KERN_ERR,
1930 "error: #fragments per group too big: %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 sbi->s_frags_per_group);
1932 goto failed_mount;
1933 }
1934 if (sbi->s_inodes_per_group > blocksize * 8) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001935 ext3_msg(sb, KERN_ERR,
1936 "error: #inodes per group too big: %lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 sbi->s_inodes_per_group);
1938 goto failed_mount;
1939 }
1940
Darrick J. Wongad692bf2010-11-16 22:57:44 +00001941 err = generic_check_addressable(sb->s_blocksize_bits,
1942 le32_to_cpu(es->s_blocks_count));
1943 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001944 ext3_msg(sb, KERN_ERR,
1945 "error: filesystem is too large to mount safely");
Mingming Caofcd5df32006-06-25 05:47:50 -07001946 if (sizeof(sector_t) < 8)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001947 ext3_msg(sb, KERN_ERR,
1948 "error: CONFIG_LBDAF not enabled");
Darrick J. Wongad692bf2010-11-16 22:57:44 +00001949 ret = err;
Mingming Caofcd5df32006-06-25 05:47:50 -07001950 goto failed_mount;
1951 }
1952
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1954 goto cantfind_ext3;
Eric Sandeen855565e2006-09-27 01:49:29 -07001955 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1956 le32_to_cpu(es->s_first_data_block) - 1)
1957 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
Namhyung Kim81a4e322010-10-11 19:38:39 +09001958 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1960 GFP_KERNEL);
1961 if (sbi->s_group_desc == NULL) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001962 ext3_msg(sb, KERN_ERR,
1963 "error: not enough memory");
Namhyung Kim4569cd12010-10-11 19:08:06 +09001964 ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 goto failed_mount;
1966 }
1967
Pekka Enberg5df096d2009-01-07 18:07:25 -08001968 bgl_lock_init(sbi->s_blockgroup_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969
1970 for (i = 0; i < db_count; i++) {
1971 block = descriptor_loc(sb, logic_sb_block, i);
1972 sbi->s_group_desc[i] = sb_bread(sb, block);
1973 if (!sbi->s_group_desc[i]) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001974 ext3_msg(sb, KERN_ERR,
1975 "error: can't read group descriptor %d", i);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 db_count = i;
1977 goto failed_mount2;
1978 }
1979 }
1980 if (!ext3_check_descriptors (sb)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01001981 ext3_msg(sb, KERN_ERR,
1982 "error: group descriptors corrupted");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 goto failed_mount2;
1984 }
1985 sbi->s_gdb_count = db_count;
1986 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1987 spin_lock_init(&sbi->s_next_gen_lock);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001988
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 /* per fileystem reservation list head & lock */
1990 spin_lock_init(&sbi->s_rsv_window_lock);
1991 sbi->s_rsv_window_root = RB_ROOT;
1992 /* Add a single, static dummy reservation to the start of the
1993 * reservation window list --- it gives us a placeholder for
1994 * append-at-start-of-list which makes the allocation logic
1995 * _much_ simpler. */
1996 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1997 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1998 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1999 sbi->s_rsv_window_head.rsv_goal_size = 0;
2000 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
2001
2002 /*
2003 * set up enough so that it can read an inode
2004 */
2005 sb->s_op = &ext3_sops;
2006 sb->s_export_op = &ext3_export_ops;
2007 sb->s_xattr = ext3_xattr_handlers;
2008#ifdef CONFIG_QUOTA
2009 sb->s_qcop = &ext3_qctl_operations;
2010 sb->dq_op = &ext3_quota_operations;
2011#endif
Aneesh Kumar K.V03cb5f02011-01-29 18:43:39 +05302012 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Eric Sandeenb8a052d2009-12-14 13:00:30 -06002014 mutex_init(&sbi->s_orphan_lock);
Eric Sandeen96d2a492009-12-14 13:01:05 -06002015 mutex_init(&sbi->s_resize_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
2017 sb->s_root = NULL;
2018
2019 needs_recovery = (es->s_last_orphan != 0 ||
2020 EXT3_HAS_INCOMPAT_FEATURE(sb,
2021 EXT3_FEATURE_INCOMPAT_RECOVER));
2022
2023 /*
2024 * The first inode we look at is the journal inode. Don't try
2025 * root first: it may be modified in the journal!
2026 */
2027 if (!test_opt(sb, NOLOAD) &&
2028 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
Johann Lombardi71b96252006-01-08 01:03:20 -08002029 if (ext3_load_journal(sb, es, journal_devnum))
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002030 goto failed_mount2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 } else if (journal_inum) {
2032 if (ext3_create_journal(sb, es, journal_inum))
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002033 goto failed_mount2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 } else {
2035 if (!silent)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002036 ext3_msg(sb, KERN_ERR,
2037 "error: no journal found. "
2038 "mounting ext3 over ext2?");
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002039 goto failed_mount2;
2040 }
2041 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2042 ext3_count_free_blocks(sb));
2043 if (!err) {
2044 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2045 ext3_count_free_inodes(sb));
2046 }
2047 if (!err) {
2048 err = percpu_counter_init(&sbi->s_dirs_counter,
2049 ext3_count_dirs(sb));
2050 }
2051 if (err) {
2052 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
Namhyung Kim4569cd12010-10-11 19:08:06 +09002053 ret = err;
Mingming Cao0216bfc2006-06-23 02:05:41 -07002054 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 }
2056
2057 /* We have now updated the journal if required, so we can
2058 * validate the data journaling mode. */
2059 switch (test_opt(sb, DATA_FLAGS)) {
2060 case 0:
2061 /* No mode set, assume a default based on the journal
2062 capabilities: ORDERED_DATA if the journal can
2063 cope, else JOURNAL_DATA */
2064 if (journal_check_available_features
2065 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
Linus Torvaldsbbae8bc2009-04-06 17:16:47 -07002066 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067 else
2068 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2069 break;
2070
2071 case EXT3_MOUNT_ORDERED_DATA:
2072 case EXT3_MOUNT_WRITEBACK_DATA:
2073 if (!journal_check_available_features
2074 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002075 ext3_msg(sb, KERN_ERR,
2076 "error: journal does not support "
2077 "requested data journaling mode");
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002078 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 }
2080 default:
2081 break;
2082 }
2083
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 /*
2085 * The journal_load will have done any necessary log recovery,
2086 * so we can safely mount the rest of the filesystem now.
2087 */
2088
David Howells473043d2008-02-07 00:15:36 -08002089 root = ext3_iget(sb, EXT3_ROOT_INO);
2090 if (IS_ERR(root)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002091 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
David Howells473043d2008-02-07 00:15:36 -08002092 ret = PTR_ERR(root);
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002093 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 }
2095 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells473043d2008-02-07 00:15:36 -08002096 iput(root);
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002097 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002098 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 }
Al Viro48fde702012-01-08 22:15:13 -05002100 sb->s_root = d_make_root(root);
David Howells473043d2008-02-07 00:15:36 -08002101 if (!sb->s_root) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002102 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
David Howells473043d2008-02-07 00:15:36 -08002103 ret = -ENOMEM;
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002104 goto failed_mount3;
David Howells473043d2008-02-07 00:15:36 -08002105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106
Eric Sandeenf007dbf2012-05-24 12:00:37 -05002107 if (ext3_setup_super(sb, es, sb->s_flags & MS_RDONLY))
2108 sb->s_flags |= MS_RDONLY;
Eric Sandeened505ee2009-12-14 12:59:18 -06002109
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2111 ext3_orphan_cleanup(sb, es);
2112 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
Eryu Guan63894ab2011-11-01 10:06:19 +08002113 if (needs_recovery) {
2114 ext3_mark_recovery_complete(sb, es);
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002115 ext3_msg(sb, KERN_INFO, "recovery complete");
Eryu Guan63894ab2011-11-01 10:06:19 +08002116 }
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002117 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2119 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2120 "writeback");
2121
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 return 0;
2123
2124cantfind_ext3:
2125 if (!silent)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002126 ext3_msg(sb, KERN_INFO,
2127 "error: can't find ext3 filesystem on dev %s.",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 sb->s_id);
2129 goto failed_mount;
2130
Mingming Cao0216bfc2006-06-23 02:05:41 -07002131failed_mount3:
2132 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2133 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2134 percpu_counter_destroy(&sbi->s_dirs_counter);
Dmitry Monakhov41d1a632010-04-12 23:46:00 +04002135 journal_destroy(sbi->s_journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136failed_mount2:
2137 for (i = 0; i < db_count; i++)
2138 brelse(sbi->s_group_desc[i]);
2139 kfree(sbi->s_group_desc);
2140failed_mount:
2141#ifdef CONFIG_QUOTA
2142 for (i = 0; i < MAXQUOTAS; i++)
2143 kfree(sbi->s_qf_names[i]);
2144#endif
2145 ext3_blkdev_remove(sbi);
2146 brelse(bh);
2147out_fail:
2148 sb->s_fs_info = NULL;
Manish Katiyarde5ce032009-05-17 23:52:47 -04002149 kfree(sbi->s_blockgroup_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150 kfree(sbi);
David Howells473043d2008-02-07 00:15:36 -08002151 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152}
2153
2154/*
2155 * Setup any per-fs journal parameters now. We'll do this both on
2156 * initial mount, once the journal has been initialised but before we've
Mingming Caoae6ddcc2006-09-27 01:49:27 -07002157 * done any recovery; and again on any subsequent remount.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158 */
2159static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2160{
2161 struct ext3_sb_info *sbi = EXT3_SB(sb);
2162
2163 if (sbi->s_commit_interval)
2164 journal->j_commit_interval = sbi->s_commit_interval;
2165 /* We could also set up an ext3-specific default for the commit
2166 * interval here, but for now we'll just fall back to the jbd
2167 * default. */
2168
2169 spin_lock(&journal->j_state_lock);
2170 if (test_opt(sb, BARRIER))
2171 journal->j_flags |= JFS_BARRIER;
2172 else
2173 journal->j_flags &= ~JFS_BARRIER;
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -07002174 if (test_opt(sb, DATA_ERR_ABORT))
2175 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2176 else
2177 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 spin_unlock(&journal->j_state_lock);
2179}
2180
Eric Sandeeneee194e2006-09-27 01:49:30 -07002181static journal_t *ext3_get_journal(struct super_block *sb,
2182 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183{
2184 struct inode *journal_inode;
2185 journal_t *journal;
2186
2187 /* First, test for the existence of a valid inode on disk. Bad
2188 * things happen if we iget() an unused inode, as the subsequent
2189 * iput() will try to delete it. */
2190
David Howells473043d2008-02-07 00:15:36 -08002191 journal_inode = ext3_iget(sb, journal_inum);
2192 if (IS_ERR(journal_inode)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002193 ext3_msg(sb, KERN_ERR, "error: no journal found");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 return NULL;
2195 }
2196 if (!journal_inode->i_nlink) {
2197 make_bad_inode(journal_inode);
2198 iput(journal_inode);
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002199 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200 return NULL;
2201 }
2202
2203 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2204 journal_inode, journal_inode->i_size);
David Howells473043d2008-02-07 00:15:36 -08002205 if (!S_ISREG(journal_inode->i_mode)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002206 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207 iput(journal_inode);
2208 return NULL;
2209 }
2210
2211 journal = journal_init_inode(journal_inode);
2212 if (!journal) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002213 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 iput(journal_inode);
2215 return NULL;
2216 }
2217 journal->j_private = sb;
2218 ext3_init_journal_params(sb, journal);
2219 return journal;
2220}
2221
2222static journal_t *ext3_get_dev_journal(struct super_block *sb,
2223 dev_t j_dev)
2224{
2225 struct buffer_head * bh;
2226 journal_t *journal;
Mingming Cao43d23f92006-06-25 05:48:07 -07002227 ext3_fsblk_t start;
Mingming Cao1c2bf372006-06-25 05:48:06 -07002228 ext3_fsblk_t len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 int hblock, blocksize;
Mingming Cao43d23f92006-06-25 05:48:07 -07002230 ext3_fsblk_t sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 unsigned long offset;
2232 struct ext3_super_block * es;
2233 struct block_device *bdev;
2234
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002235 bdev = ext3_blkdev_get(j_dev, sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236 if (bdev == NULL)
2237 return NULL;
2238
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04002240 hblock = bdev_logical_block_size(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 if (blocksize < hblock) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002242 ext3_msg(sb, KERN_ERR,
2243 "error: blocksize too small for journal device");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244 goto out_bdev;
2245 }
2246
2247 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2248 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2249 set_blocksize(bdev, blocksize);
2250 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002251 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2252 "external journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 goto out_bdev;
2254 }
2255
Namhyung Kim57e94d82010-10-11 02:10:45 +09002256 es = (struct ext3_super_block *) (bh->b_data + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2258 !(le32_to_cpu(es->s_feature_incompat) &
2259 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002260 ext3_msg(sb, KERN_ERR, "error: external journal has "
2261 "bad superblock");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262 brelse(bh);
2263 goto out_bdev;
2264 }
2265
2266 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002267 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268 brelse(bh);
2269 goto out_bdev;
2270 }
2271
2272 len = le32_to_cpu(es->s_blocks_count);
2273 start = sb_block + 1;
2274 brelse(bh); /* we're done with the superblock */
2275
2276 journal = journal_init_dev(bdev, sb->s_bdev,
2277 start, len, blocksize);
2278 if (!journal) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002279 ext3_msg(sb, KERN_ERR,
2280 "error: failed to create device journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281 goto out_bdev;
2282 }
2283 journal->j_private = sb;
Zheng Liud03e1292011-12-05 15:55:11 +08002284 if (!bh_uptodate_or_lock(journal->j_sb_buffer)) {
2285 if (bh_submit_read(journal->j_sb_buffer)) {
2286 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2287 goto out_journal;
2288 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 }
2290 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002291 ext3_msg(sb, KERN_ERR,
2292 "error: external journal has more than one "
2293 "user (unsupported) - %d",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294 be32_to_cpu(journal->j_superblock->s_nr_users));
2295 goto out_journal;
2296 }
2297 EXT3_SB(sb)->journal_bdev = bdev;
2298 ext3_init_journal_params(sb, journal);
2299 return journal;
2300out_journal:
2301 journal_destroy(journal);
2302out_bdev:
2303 ext3_blkdev_put(bdev);
2304 return NULL;
2305}
2306
Johann Lombardi71b96252006-01-08 01:03:20 -08002307static int ext3_load_journal(struct super_block *sb,
2308 struct ext3_super_block *es,
2309 unsigned long journal_devnum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310{
2311 journal_t *journal;
Eric Sandeeneee194e2006-09-27 01:49:30 -07002312 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
Johann Lombardi71b96252006-01-08 01:03:20 -08002313 dev_t journal_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314 int err = 0;
2315 int really_read_only;
2316
Johann Lombardi71b96252006-01-08 01:03:20 -08002317 if (journal_devnum &&
2318 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002319 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2320 "numbers have changed");
Johann Lombardi71b96252006-01-08 01:03:20 -08002321 journal_dev = new_decode_dev(journal_devnum);
2322 } else
2323 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2324
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 really_read_only = bdev_read_only(sb->s_bdev);
2326
2327 /*
2328 * Are we loading a blank journal or performing recovery after a
2329 * crash? For recovery, we need to check in advance whether we
2330 * can get read-write access to the device.
2331 */
2332
2333 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2334 if (sb->s_flags & MS_RDONLY) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002335 ext3_msg(sb, KERN_INFO,
2336 "recovery required on readonly filesystem");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337 if (really_read_only) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002338 ext3_msg(sb, KERN_ERR, "error: write access "
2339 "unavailable, cannot proceed");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340 return -EROFS;
2341 }
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002342 ext3_msg(sb, KERN_INFO,
2343 "write access will be enabled during recovery");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 }
2345 }
2346
2347 if (journal_inum && journal_dev) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002348 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2349 "and inode journals");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350 return -EINVAL;
2351 }
2352
2353 if (journal_inum) {
2354 if (!(journal = ext3_get_journal(sb, journal_inum)))
2355 return -EINVAL;
2356 } else {
2357 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2358 return -EINVAL;
2359 }
2360
Eric Sandeen0636c732010-04-30 11:09:34 -05002361 if (!(journal->j_flags & JFS_BARRIER))
2362 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2363
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2365 err = journal_update_format(journal);
2366 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002367 ext3_msg(sb, KERN_ERR, "error updating journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368 journal_destroy(journal);
2369 return err;
2370 }
2371 }
2372
2373 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2374 err = journal_wipe(journal, !really_read_only);
2375 if (!err)
2376 err = journal_load(journal);
2377
2378 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002379 ext3_msg(sb, KERN_ERR, "error loading journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380 journal_destroy(journal);
2381 return err;
2382 }
2383
2384 EXT3_SB(sb)->s_journal = journal;
2385 ext3_clear_journal_err(sb, es);
Johann Lombardi71b96252006-01-08 01:03:20 -08002386
Maciej Żenczykowski31d710a2010-09-26 12:38:28 +00002387 if (!really_read_only && journal_devnum &&
Johann Lombardi71b96252006-01-08 01:03:20 -08002388 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2389 es->s_journal_dev = cpu_to_le32(journal_devnum);
Johann Lombardi71b96252006-01-08 01:03:20 -08002390
2391 /* Make sure we flush the recovery flag to disk. */
2392 ext3_commit_super(sb, es, 1);
2393 }
2394
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395 return 0;
2396}
2397
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002398static int ext3_create_journal(struct super_block *sb,
2399 struct ext3_super_block *es,
Eric Sandeeneee194e2006-09-27 01:49:30 -07002400 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002401{
2402 journal_t *journal;
Borislav Petkov952d9de2007-07-15 23:41:45 -07002403 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404
2405 if (sb->s_flags & MS_RDONLY) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002406 ext3_msg(sb, KERN_ERR,
2407 "error: readonly filesystem when trying to "
2408 "create journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409 return -EROFS;
2410 }
2411
Borislav Petkov952d9de2007-07-15 23:41:45 -07002412 journal = ext3_get_journal(sb, journal_inum);
2413 if (!journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414 return -EINVAL;
2415
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002416 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002417 journal_inum);
2418
Borislav Petkov952d9de2007-07-15 23:41:45 -07002419 err = journal_create(journal);
2420 if (err) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002421 ext3_msg(sb, KERN_ERR, "error creating journal");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422 journal_destroy(journal);
2423 return -EIO;
2424 }
2425
2426 EXT3_SB(sb)->s_journal = journal;
2427
2428 ext3_update_dynamic_rev(sb);
2429 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2430 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2431
2432 es->s_journal_inum = cpu_to_le32(journal_inum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433
2434 /* Make sure we flush the recovery flag to disk. */
2435 ext3_commit_super(sb, es, 1);
2436
2437 return 0;
2438}
2439
Takashi Satoc4be0c12009-01-09 16:40:58 -08002440static int ext3_commit_super(struct super_block *sb,
2441 struct ext3_super_block *es,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 int sync)
2443{
2444 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08002445 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446
2447 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08002448 return error;
Darrick J. Wongdff68252010-10-04 12:35:05 -07002449
2450 if (buffer_write_io_error(sbh)) {
2451 /*
2452 * Oh, dear. A previous attempt to write the
2453 * superblock failed. This could happen because the
2454 * USB device was yanked out. Or it could happen to
2455 * be a transient write error and maybe the block will
2456 * be remapped. Nothing we can do but to retry the
2457 * write and hope for the best.
2458 */
2459 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2460 "superblock detected");
2461 clear_buffer_write_io_error(sbh);
2462 set_buffer_uptodate(sbh);
2463 }
Theodore Ts'o96ec2e02009-09-16 11:21:13 -04002464 /*
2465 * If the file system is mounted read-only, don't update the
2466 * superblock write time. This avoids updating the superblock
2467 * write time when we are mounting the root file system
2468 * read/only but we need to replay the journal; at that point,
2469 * for people who are east of GMT and who make their clock
2470 * tick in localtime for Windows bug-for-bug compatibility,
2471 * the clock is set in the future, and this will cause e2fsck
2472 * to complain and force a full file system check.
2473 */
2474 if (!(sb->s_flags & MS_RDONLY))
2475 es->s_wtime = cpu_to_le32(get_seconds());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002476 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2477 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2478 BUFFER_TRACE(sbh, "marking dirty");
2479 mark_buffer_dirty(sbh);
Darrick J. Wongdff68252010-10-04 12:35:05 -07002480 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08002481 error = sync_dirty_buffer(sbh);
Darrick J. Wongdff68252010-10-04 12:35:05 -07002482 if (buffer_write_io_error(sbh)) {
2483 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2484 "superblock");
2485 clear_buffer_write_io_error(sbh);
2486 set_buffer_uptodate(sbh);
2487 }
2488 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08002489 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490}
2491
2492
2493/*
2494 * Have we just finished recovery? If so, and if we are mounting (or
2495 * remounting) the filesystem readonly, then we will end up with a
2496 * consistent fs on disk. Record that fact.
2497 */
2498static void ext3_mark_recovery_complete(struct super_block * sb,
2499 struct ext3_super_block * es)
2500{
2501 journal_t *journal = EXT3_SB(sb)->s_journal;
2502
2503 journal_lock_updates(journal);
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002504 if (journal_flush(journal) < 0)
2505 goto out;
2506
Linus Torvalds1da177e2005-04-16 15:20:36 -07002507 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2508 sb->s_flags & MS_RDONLY) {
2509 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510 ext3_commit_super(sb, es, 1);
2511 }
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002512
2513out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514 journal_unlock_updates(journal);
2515}
2516
2517/*
2518 * If we are mounting (or read-write remounting) a filesystem whose journal
2519 * has recorded an error from a previous lifetime, move that error to the
2520 * main filesystem now.
2521 */
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002522static void ext3_clear_journal_err(struct super_block *sb,
2523 struct ext3_super_block *es)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524{
2525 journal_t *journal;
2526 int j_errno;
2527 const char *errstr;
2528
2529 journal = EXT3_SB(sb)->s_journal;
2530
2531 /*
2532 * Now check for any error status which may have been recorded in the
2533 * journal by a prior ext3_error() or ext3_abort()
2534 */
2535
2536 j_errno = journal_errno(journal);
2537 if (j_errno) {
2538 char nbuf[16];
2539
2540 errstr = ext3_decode_error(sb, j_errno, nbuf);
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002541 ext3_warning(sb, __func__, "Filesystem error recorded "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542 "from previous mount: %s", errstr);
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002543 ext3_warning(sb, __func__, "Marking fs in need of "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002544 "filesystem check.");
2545
2546 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2547 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2548 ext3_commit_super (sb, es, 1);
2549
2550 journal_clear_err(journal);
2551 }
2552}
2553
2554/*
2555 * Force the running and committing transactions to commit,
2556 * and wait on the commit.
2557 */
2558int ext3_force_commit(struct super_block *sb)
2559{
2560 journal_t *journal;
2561 int ret;
2562
2563 if (sb->s_flags & MS_RDONLY)
2564 return 0;
2565
2566 journal = EXT3_SB(sb)->s_journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567 ret = ext3_journal_force_commit(journal);
2568 return ret;
2569}
2570
Linus Torvalds1da177e2005-04-16 15:20:36 -07002571static int ext3_sync_fs(struct super_block *sb, int wait)
2572{
Jan Kara02ac5972009-02-11 13:04:26 -08002573 tid_t target;
Arthur Jonesc87591b2008-11-06 12:53:35 -08002574
Lukas Czerner785c4bc2011-05-23 18:33:01 +02002575 trace_ext3_sync_fs(sb, wait);
Jan Karaa1177822012-07-03 16:45:29 +02002576 /*
2577 * Writeback quota in non-journalled quota case - journalled quota has
2578 * no dirty dquots
2579 */
2580 dquot_writeback_dquots(sb, -1);
Jan Kara02ac5972009-02-11 13:04:26 -08002581 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2582 if (wait)
2583 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2584 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002585 return 0;
2586}
2587
2588/*
2589 * LVM calls this function before a (read-only) snapshot is created. This
2590 * gives us a chance to flush the journal completely and mark the fs clean.
2591 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08002592static int ext3_freeze(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593{
Takashi Satoc4be0c12009-01-09 16:40:58 -08002594 int error = 0;
2595 journal_t *journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002596
2597 if (!(sb->s_flags & MS_RDONLY)) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08002598 journal = EXT3_SB(sb)->s_journal;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002599
2600 /* Now we set up the journal barrier. */
2601 journal_lock_updates(journal);
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002602
2603 /*
2604 * We don't want to clear needs_recovery flag when we failed
2605 * to flush the journal.
2606 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08002607 error = journal_flush(journal);
2608 if (error < 0)
2609 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610
2611 /* Journal blocked and flushed, clear needs_recovery flag. */
2612 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
Takashi Satoc4be0c12009-01-09 16:40:58 -08002613 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2614 if (error)
2615 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002616 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08002617 return 0;
2618
2619out:
2620 journal_unlock_updates(journal);
2621 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622}
2623
2624/*
2625 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2626 * flag here, even though the filesystem is not technically dirty yet.
2627 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08002628static int ext3_unfreeze(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629{
2630 if (!(sb->s_flags & MS_RDONLY)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002631 /* Reser the needs_recovery flag before the fs is unlocked. */
2632 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2633 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2635 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08002636 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002637}
2638
2639static int ext3_remount (struct super_block * sb, int * flags, char * data)
2640{
2641 struct ext3_super_block * es;
2642 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07002643 ext3_fsblk_t n_blocks_count = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002644 unsigned long old_sb_flags;
2645 struct ext3_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002646 int enable_quota = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002647 int err;
2648#ifdef CONFIG_QUOTA
2649 int i;
2650#endif
2651
2652 /* Store the original options */
2653 old_sb_flags = sb->s_flags;
2654 old_opts.s_mount_opt = sbi->s_mount_opt;
2655 old_opts.s_resuid = sbi->s_resuid;
2656 old_opts.s_resgid = sbi->s_resgid;
2657 old_opts.s_commit_interval = sbi->s_commit_interval;
2658#ifdef CONFIG_QUOTA
2659 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2660 for (i = 0; i < MAXQUOTAS; i++)
Jan Karaf56426a2012-12-31 12:38:36 +01002661 if (sbi->s_qf_names[i]) {
2662 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
2663 GFP_KERNEL);
2664 if (!old_opts.s_qf_names[i]) {
2665 int j;
2666
2667 for (j = 0; j < i; j++)
2668 kfree(old_opts.s_qf_names[j]);
2669 return -ENOMEM;
2670 }
2671 } else
2672 old_opts.s_qf_names[i] = NULL;
Jan Kara08c6a962005-07-12 13:58:28 -07002673#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002674
2675 /*
2676 * Allow the "check" option to be passed as a remount option.
2677 */
Johann Lombardi71b96252006-01-08 01:03:20 -08002678 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002679 err = -EINVAL;
2680 goto restore_opts;
2681 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002682
Dmitry Monakhove3c96432010-02-02 16:05:51 +03002683 if (test_opt(sb, ABORT))
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002684 ext3_abort(sb, __func__, "Abort forced by user");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002685
2686 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhove3c96432010-02-02 16:05:51 +03002687 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002688
2689 es = sbi->s_es;
2690
2691 ext3_init_journal_params(sb, sbi->s_journal);
2692
2693 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2694 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
Dmitry Monakhove3c96432010-02-02 16:05:51 +03002695 if (test_opt(sb, ABORT)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002696 err = -EROFS;
2697 goto restore_opts;
2698 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002699
2700 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04002701 err = dquot_suspend(sb, -1);
2702 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002703 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002704
Linus Torvalds1da177e2005-04-16 15:20:36 -07002705 /*
2706 * First of all, the unconditional stuff we have to do
2707 * to disable replay of the journal when we next remount
2708 */
2709 sb->s_flags |= MS_RDONLY;
2710
2711 /*
2712 * OK, test if we are remounting a valid rw partition
2713 * readonly, and if so set the rdonly flag and then
2714 * mark the partition as valid again.
2715 */
2716 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2717 (sbi->s_mount_state & EXT3_VALID_FS))
2718 es->s_state = cpu_to_le16(sbi->s_mount_state);
2719
2720 ext3_mark_recovery_complete(sb, es);
2721 } else {
2722 __le32 ret;
2723 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2724 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002725 ext3_msg(sb, KERN_WARNING,
2726 "warning: couldn't remount RDWR "
2727 "because of unsupported optional "
2728 "features (%x)", le32_to_cpu(ret));
Jan Kara08c6a962005-07-12 13:58:28 -07002729 err = -EROFS;
2730 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002731 }
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002732
2733 /*
2734 * If we have an unprocessed orphan list hanging
2735 * around from a previously readonly bdev mount,
Toshiyuki Okajima1cde2012011-08-02 18:16:57 +09002736 * require a full umount & mount for now.
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002737 */
2738 if (es->s_last_orphan) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002739 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002740 "remount RDWR because of unprocessed "
2741 "orphan inode list. Please "
Toshiyuki Okajima1cde2012011-08-02 18:16:57 +09002742 "umount & mount instead.");
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002743 err = -EINVAL;
2744 goto restore_opts;
2745 }
2746
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747 /*
2748 * Mounting a RDONLY partition read-write, so reread
2749 * and store the current valid flag. (It may have
2750 * been changed by e2fsck since we originally mounted
2751 * the partition.)
2752 */
2753 ext3_clear_journal_err(sb, es);
2754 sbi->s_mount_state = le16_to_cpu(es->s_state);
Dave Kleikampa4e4de32006-09-27 01:49:36 -07002755 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
Jan Kara08c6a962005-07-12 13:58:28 -07002756 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002757 if (!ext3_setup_super (sb, es, 0))
2758 sb->s_flags &= ~MS_RDONLY;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002759 enable_quota = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002760 }
2761 }
Jan Kara08c6a962005-07-12 13:58:28 -07002762#ifdef CONFIG_QUOTA
2763 /* Release old quota file names */
2764 for (i = 0; i < MAXQUOTAS; i++)
Jan Karaf56426a2012-12-31 12:38:36 +01002765 kfree(old_opts.s_qf_names[i]);
Jan Kara08c6a962005-07-12 13:58:28 -07002766#endif
Christoph Hellwigc79d9672010-05-19 07:16:40 -04002767 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04002768 dquot_resume(sb, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002769 return 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002770restore_opts:
2771 sb->s_flags = old_sb_flags;
2772 sbi->s_mount_opt = old_opts.s_mount_opt;
2773 sbi->s_resuid = old_opts.s_resuid;
2774 sbi->s_resgid = old_opts.s_resgid;
2775 sbi->s_commit_interval = old_opts.s_commit_interval;
2776#ifdef CONFIG_QUOTA
2777 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2778 for (i = 0; i < MAXQUOTAS; i++) {
Jan Karaf56426a2012-12-31 12:38:36 +01002779 kfree(sbi->s_qf_names[i]);
Jan Kara08c6a962005-07-12 13:58:28 -07002780 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2781 }
2782#endif
2783 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784}
2785
David Howells726c3342006-06-23 02:02:58 -07002786static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002787{
David Howells726c3342006-06-23 02:02:58 -07002788 struct super_block *sb = dentry->d_sb;
Alex Tomas09fe3162006-03-24 03:16:16 -08002789 struct ext3_sb_info *sbi = EXT3_SB(sb);
2790 struct ext3_super_block *es = sbi->s_es;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002791 u64 fsid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002792
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002793 if (test_opt(sb, MINIX_DF)) {
2794 sbi->s_overhead_last = 0;
2795 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2796 unsigned long ngroups = sbi->s_groups_count, i;
2797 ext3_fsblk_t overhead = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002798 smp_rmb();
2799
2800 /*
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002801 * Compute the overhead (FS structures). This is constant
2802 * for a given filesystem unless the number of block groups
2803 * changes so we cache the previous value until it does.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804 */
2805
2806 /*
2807 * All of the blocks before first_data_block are
2808 * overhead
2809 */
2810 overhead = le32_to_cpu(es->s_first_data_block);
2811
2812 /*
2813 * Add the overhead attributed to the superblock and
2814 * block group descriptors. If the sparse superblocks
2815 * feature is turned on, then not all groups have this.
2816 */
2817 for (i = 0; i < ngroups; i++) {
2818 overhead += ext3_bg_has_super(sb, i) +
2819 ext3_bg_num_gdb(sb, i);
2820 cond_resched();
2821 }
2822
2823 /*
2824 * Every block group has an inode bitmap, a block
2825 * bitmap, and an inode table.
2826 */
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002827 overhead += ngroups * (2 + sbi->s_itb_per_group);
2828 sbi->s_overhead_last = overhead;
2829 smp_wmb();
2830 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831 }
2832
2833 buf->f_type = EXT3_SUPER_MAGIC;
2834 buf->f_bsize = sb->s_blocksize;
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002835 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002836 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002837 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2838 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2839 buf->f_bavail = 0;
2840 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002841 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002842 buf->f_namelen = EXT3_NAME_LEN;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002843 fsid = le64_to_cpup((void *)es->s_uuid) ^
2844 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2845 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2846 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002847 return 0;
2848}
2849
2850/* Helper function for writing quotas on sync - we need to start transaction before quota file
2851 * is locked for write. Otherwise the are possible deadlocks:
2852 * Process 1 Process 2
2853 * ext3_create() quota_sync()
2854 * journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05002855 * dquot_initialize() down(dqio_mutex)
Ingo Molnard3be9152006-03-23 03:00:29 -08002856 * down(dqio_mutex) journal_start()
Linus Torvalds1da177e2005-04-16 15:20:36 -07002857 *
2858 */
2859
2860#ifdef CONFIG_QUOTA
2861
2862static inline struct inode *dquot_to_inode(struct dquot *dquot)
2863{
Eric W. Biederman4c376dc2012-09-16 03:56:19 -07002864 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865}
2866
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867static int ext3_write_dquot(struct dquot *dquot)
2868{
2869 int ret, err;
2870 handle_t *handle;
2871 struct inode *inode;
2872
2873 inode = dquot_to_inode(dquot);
2874 handle = ext3_journal_start(inode,
Jan Kara1f545872005-06-23 22:01:04 -07002875 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876 if (IS_ERR(handle))
2877 return PTR_ERR(handle);
2878 ret = dquot_commit(dquot);
2879 err = ext3_journal_stop(handle);
2880 if (!ret)
2881 ret = err;
2882 return ret;
2883}
2884
2885static int ext3_acquire_dquot(struct dquot *dquot)
2886{
2887 int ret, err;
2888 handle_t *handle;
2889
2890 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002891 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002892 if (IS_ERR(handle))
2893 return PTR_ERR(handle);
2894 ret = dquot_acquire(dquot);
2895 err = ext3_journal_stop(handle);
2896 if (!ret)
2897 ret = err;
2898 return ret;
2899}
2900
2901static int ext3_release_dquot(struct dquot *dquot)
2902{
2903 int ret, err;
2904 handle_t *handle;
2905
2906 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002907 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07002908 if (IS_ERR(handle)) {
2909 /* Release dquot anyway to avoid endless cycle in dqput() */
2910 dquot_release(dquot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07002912 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002913 ret = dquot_release(dquot);
2914 err = ext3_journal_stop(handle);
2915 if (!ret)
2916 ret = err;
2917 return ret;
2918}
2919
2920static int ext3_mark_dquot_dirty(struct dquot *dquot)
2921{
Jan Kara99aeaf62008-07-25 01:46:17 -07002922 /* Are we journaling quotas? */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002923 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2924 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2925 dquot_mark_dquot_dirty(dquot);
2926 return ext3_write_dquot(dquot);
2927 } else {
2928 return dquot_mark_dquot_dirty(dquot);
2929 }
2930}
2931
2932static int ext3_write_info(struct super_block *sb, int type)
2933{
2934 int ret, err;
2935 handle_t *handle;
2936
2937 /* Data block + inode block */
2938 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2939 if (IS_ERR(handle))
2940 return PTR_ERR(handle);
2941 ret = dquot_commit_info(sb, type);
2942 err = ext3_journal_stop(handle);
2943 if (!ret)
2944 ret = err;
2945 return ret;
2946}
2947
2948/*
2949 * Turn on quotas during mount time - we need to find
2950 * the quota file and such...
2951 */
2952static int ext3_quota_on_mount(struct super_block *sb, int type)
2953{
Christoph Hellwig287a8092010-05-19 07:16:45 -04002954 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2955 EXT3_SB(sb)->s_jquota_fmt, type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002956}
2957
2958/*
2959 * Standard function to be called on quota_on
2960 */
2961static int ext3_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02002962 struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002963{
2964 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002965
Jan Kara1f545872005-06-23 22:01:04 -07002966 if (!test_opt(sb, QUOTA))
2967 return -EINVAL;
Jan Kara9cfe7b92008-07-25 01:46:16 -07002968
Linus Torvalds1da177e2005-04-16 15:20:36 -07002969 /* Quotafile not on the same filesystem? */
Al Virod8c95842011-12-07 18:16:57 -05002970 if (path->dentry->d_sb != sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002971 return -EXDEV;
Jan Kara9cfe7b92008-07-25 01:46:16 -07002972 /* Journaling quota? */
2973 if (EXT3_SB(sb)->s_qf_names[type]) {
2974 /* Quotafile not of fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02002975 if (path->dentry->d_parent != sb->s_root)
Alexey Fisher4cf46b62009-11-22 20:38:55 +01002976 ext3_msg(sb, KERN_WARNING,
2977 "warning: Quota file not on filesystem root. "
2978 "Journaled quota will not work.");
Jan Kara9cfe7b92008-07-25 01:46:16 -07002979 }
2980
2981 /*
2982 * When we journal data on quota file, we have to flush journal to see
2983 * all updates to the file when we bypass pagecache...
2984 */
Jan Karaf00c9e42010-09-15 17:38:58 +02002985 if (ext3_should_journal_data(path->dentry->d_inode)) {
Jan Kara9cfe7b92008-07-25 01:46:16 -07002986 /*
2987 * We don't need to lock updates but journal_flush() could
2988 * otherwise be livelocked...
2989 */
2990 journal_lock_updates(EXT3_SB(sb)->s_journal);
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002991 err = journal_flush(EXT3_SB(sb)->s_journal);
Jan Kara9cfe7b92008-07-25 01:46:16 -07002992 journal_unlock_updates(EXT3_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02002993 if (err)
Hidehiro Kawai2d7c8202008-10-22 14:15:01 -07002994 return err;
Jan Kara9cfe7b92008-07-25 01:46:16 -07002995 }
2996
Jan Karaf00c9e42010-09-15 17:38:58 +02002997 return dquot_quota_on(sb, type, format_id, path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002998}
2999
3000/* Read data from quotafile - avoid pagecache and such because we cannot afford
3001 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003002 * itself serializes the operations (and no one else should touch the files)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003003 * we don't have to be afraid of races */
3004static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
3005 size_t len, loff_t off)
3006{
3007 struct inode *inode = sb_dqopt(sb)->files[type];
3008 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
3009 int err = 0;
3010 int offset = off & (sb->s_blocksize - 1);
3011 int tocopy;
3012 size_t toread;
3013 struct buffer_head *bh;
3014 loff_t i_size = i_size_read(inode);
3015
3016 if (off > i_size)
3017 return 0;
3018 if (off+len > i_size)
3019 len = i_size-off;
3020 toread = len;
3021 while (toread > 0) {
3022 tocopy = sb->s_blocksize - offset < toread ?
3023 sb->s_blocksize - offset : toread;
3024 bh = ext3_bread(NULL, inode, blk, 0, &err);
3025 if (err)
3026 return err;
3027 if (!bh) /* A hole? */
3028 memset(data, 0, tocopy);
3029 else
3030 memcpy(data, bh->b_data+offset, tocopy);
3031 brelse(bh);
3032 offset = 0;
3033 toread -= tocopy;
3034 data += tocopy;
3035 blk++;
3036 }
3037 return len;
3038}
3039
3040/* Write to quotafile (we know the transaction is already started and has
3041 * enough credits) */
3042static ssize_t ext3_quota_write(struct super_block *sb, int type,
3043 const char *data, size_t len, loff_t off)
3044{
3045 struct inode *inode = sb_dqopt(sb)->files[type];
3046 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
3047 int err = 0;
3048 int offset = off & (sb->s_blocksize - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003050 struct buffer_head *bh;
3051 handle_t *handle = journal_current_handle();
3052
Jan Kara9c3013e2007-09-11 15:23:29 -07003053 if (!handle) {
Alexey Fisher4cf46b62009-11-22 20:38:55 +01003054 ext3_msg(sb, KERN_WARNING,
3055 "warning: quota write (off=%llu, len=%llu)"
3056 " cancelled because transaction is not started.",
Jan Kara9c3013e2007-09-11 15:23:29 -07003057 (unsigned long long)off, (unsigned long long)len);
3058 return -EIO;
3059 }
Dmitry Monakhove5472142010-02-16 19:33:42 +03003060
3061 /*
3062 * Since we account only one data block in transaction credits,
3063 * then it is impossible to cross a block boundary.
3064 */
3065 if (sb->s_blocksize - offset < len) {
3066 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3067 " cancelled because not block aligned",
3068 (unsigned long long)off, (unsigned long long)len);
3069 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070 }
Dmitry Monakhove5472142010-02-16 19:33:42 +03003071 bh = ext3_bread(handle, inode, blk, 1, &err);
3072 if (!bh)
3073 goto out;
3074 if (journal_quota) {
3075 err = ext3_journal_get_write_access(handle, bh);
3076 if (err) {
3077 brelse(bh);
3078 goto out;
3079 }
3080 }
3081 lock_buffer(bh);
3082 memcpy(bh->b_data+offset, data, len);
3083 flush_dcache_page(bh->b_page);
3084 unlock_buffer(bh);
3085 if (journal_quota)
3086 err = ext3_journal_dirty_metadata(handle, bh);
3087 else {
3088 /* Always do at least ordered writes for quotas */
3089 err = ext3_journal_dirty_data(handle, bh);
3090 mark_buffer_dirty(bh);
3091 }
3092 brelse(bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003093out:
Jan Kara905c3932012-04-25 21:24:30 +02003094 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003095 return err;
Dmitry Monakhove5472142010-02-16 19:33:42 +03003096 if (inode->i_size < off + len) {
3097 i_size_write(inode, off + len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003098 EXT3_I(inode)->i_disksize = inode->i_size;
3099 }
3100 inode->i_version++;
3101 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3102 ext3_mark_inode_dirty(handle, inode);
Dmitry Monakhove5472142010-02-16 19:33:42 +03003103 return len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003104}
3105
3106#endif
3107
Al Viro152a0832010-07-25 00:46:55 +04003108static struct dentry *ext3_mount(struct file_system_type *fs_type,
3109 int flags, const char *dev_name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003110{
Al Viro152a0832010-07-25 00:46:55 +04003111 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003112}
3113
3114static struct file_system_type ext3_fs_type = {
3115 .owner = THIS_MODULE,
3116 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +04003117 .mount = ext3_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003118 .kill_sb = kill_block_super,
3119 .fs_flags = FS_REQUIRES_DEV,
3120};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08003121MODULE_ALIAS_FS("ext3");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003122
3123static int __init init_ext3_fs(void)
3124{
3125 int err = init_ext3_xattr();
3126 if (err)
3127 return err;
3128 err = init_inodecache();
3129 if (err)
3130 goto out1;
3131 err = register_filesystem(&ext3_fs_type);
3132 if (err)
3133 goto out;
3134 return 0;
3135out:
3136 destroy_inodecache();
3137out1:
Dave Kleikampe9ad5622006-09-27 01:49:35 -07003138 exit_ext3_xattr();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003139 return err;
3140}
3141
3142static void __exit exit_ext3_fs(void)
3143{
3144 unregister_filesystem(&ext3_fs_type);
3145 destroy_inodecache();
3146 exit_ext3_xattr();
3147}
3148
3149MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3150MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3151MODULE_LICENSE("GPL");
3152module_init(init_ext3_fs)
3153module_exit(exit_ext3_fs)