blob: 1e69f29a8c5582ba209ad7dbeb49e57396d7e2fe [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
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
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
Mingming Caodab291a2006-10-11 01:21:01 -070023#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070024#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
28#include <linux/smp_lock.h>
29#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070030#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
Vignesh Babu13305932007-07-18 09:11:02 -040037#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040038#include <linux/crc16.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070039#include <asm/uaccess.h>
40
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040041#include "ext4.h"
42#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070043#include "xattr.h"
44#include "acl.h"
45#include "namei.h"
Andreas Dilger717d50e2007-10-16 18:38:25 -040046#include "group.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070047
Mingming Cao617ba132006-10-11 01:20:53 -070048static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070049 unsigned long journal_devnum);
Mingming Cao617ba132006-10-11 01:20:53 -070050static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070051 unsigned int);
Mingming Cao617ba132006-10-11 01:20:53 -070052static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070054 int sync);
Mingming Cao617ba132006-10-11 01:20:53 -070055static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59static int ext4_sync_fs(struct super_block *sb, int wait);
60static const char *ext4_decode_error(struct super_block * sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070061 char nbuf[16]);
Mingming Cao617ba132006-10-11 01:20:53 -070062static int ext4_remount (struct super_block * sb, int * flags, char * data);
63static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64static void ext4_unlockfs(struct super_block *sb);
65static void ext4_write_super (struct super_block * sb);
66static void ext4_write_super_lockfs(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070067
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070068
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070069ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070071{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040072 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070073 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040074 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070075}
76
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070077ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070079{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040080 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070081 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040082 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070083}
84
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070085ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070087{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040088 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070089 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040090 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070091}
92
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070093void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070095{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040096 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070097 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070099}
100
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700101void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700103{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700107}
108
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700109void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700111{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700115}
116
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700117/*
Mingming Caodab291a2006-10-11 01:21:01 -0700118 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700119 *
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
123 * appropriate.
124 */
Mingming Cao617ba132006-10-11 01:20:53 -0700125handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700126{
127 journal_t *journal;
128
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
131
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700135 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700136 if (is_journal_aborted(journal)) {
Harvey Harrison46e665e2008-04-17 10:38:59 -0400137 ext4_abort(sb, __func__,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
140 }
141
Mingming Caodab291a2006-10-11 01:21:01 -0700142 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700143}
144
145/*
146 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700148 * that sync() will call the filesystem's write_super callback if
149 * appropriate.
150 */
Mingming Cao617ba132006-10-11 01:20:53 -0700151int __ext4_journal_stop(const char *where, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700152{
153 struct super_block *sb;
154 int err;
155 int rc;
156
157 sb = handle->h_transaction->t_journal->j_private;
158 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700159 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700160
161 if (!err)
162 err = rc;
163 if (err)
Mingming Cao617ba132006-10-11 01:20:53 -0700164 __ext4_std_error(sb, where, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700165 return err;
166}
167
Mingming Cao617ba132006-10-11 01:20:53 -0700168void ext4_journal_abort_handle(const char *caller, const char *err_fn,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700169 struct buffer_head *bh, handle_t *handle, int err)
170{
171 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700173
174 if (bh)
175 BUFFER_TRACE(bh, "abort");
176
177 if (!handle->h_err)
178 handle->h_err = err;
179
180 if (is_handle_aborted(handle))
181 return;
182
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
185
Mingming Caodab291a2006-10-11 01:21:01 -0700186 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700187}
188
189/* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
191 *
192 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700193 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
198 *
Mingming Caodab291a2006-10-11 01:21:01 -0700199 * We'll just use the jbd2_journal_abort() error code to record an error in
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
202 */
203
Mingming Cao617ba132006-10-11 01:20:53 -0700204static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700205{
Mingming Cao617ba132006-10-11 01:20:53 -0700206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700207
Mingming Cao617ba132006-10-11 01:20:53 -0700208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700210
211 if (sb->s_flags & MS_RDONLY)
212 return;
213
214 if (!test_opt (sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700215 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700216
Mingming Cao617ba132006-10-11 01:20:53 -0700217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700218 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700219 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700220 }
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
224 }
Mingming Cao617ba132006-10-11 01:20:53 -0700225 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700226 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700227 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700228 sb->s_id);
229}
230
Mingming Cao617ba132006-10-11 01:20:53 -0700231void ext4_error (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700232 const char * fmt, ...)
233{
234 va_list args;
235
236 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700238 vprintk(fmt, args);
239 printk("\n");
240 va_end(args);
241
Mingming Cao617ba132006-10-11 01:20:53 -0700242 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700243}
244
Mingming Cao617ba132006-10-11 01:20:53 -0700245static const char *ext4_decode_error(struct super_block * sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700246 char nbuf[16])
247{
248 char *errstr = NULL;
249
250 switch (errno) {
251 case -EIO:
252 errstr = "IO failure";
253 break;
254 case -ENOMEM:
255 errstr = "Out of memory";
256 break;
257 case -EROFS:
Mingming Caodab291a2006-10-11 01:21:01 -0700258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700259 errstr = "Journal has aborted";
260 else
261 errstr = "Readonly filesystem";
262 break;
263 default:
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
266 * NULL. */
267 if (nbuf) {
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
270 errstr = nbuf;
271 }
272 break;
273 }
274
275 return errstr;
276}
277
Mingming Cao617ba132006-10-11 01:20:53 -0700278/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700279 * automatically and invokes the appropriate error response. */
280
Mingming Cao617ba132006-10-11 01:20:53 -0700281void __ext4_std_error (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700282 int errno)
283{
284 char nbuf[16];
285 const char *errstr;
286
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
289 * an error. */
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
292 return;
293
Mingming Cao617ba132006-10-11 01:20:53 -0700294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700296 sb->s_id, function, errstr);
297
Mingming Cao617ba132006-10-11 01:20:53 -0700298 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700299}
300
301/*
Mingming Cao617ba132006-10-11 01:20:53 -0700302 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
305 *
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
309 */
310
Mingming Cao617ba132006-10-11 01:20:53 -0700311void ext4_abort (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700312 const char * fmt, ...)
313{
314 va_list args;
315
Mingming Cao617ba132006-10-11 01:20:53 -0700316 printk (KERN_CRIT "ext4_abort called.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700317
318 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700320 vprintk(fmt, args);
321 printk("\n");
322 va_end(args);
323
324 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700325 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700326
327 if (sb->s_flags & MS_RDONLY)
328 return;
329
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
Mingming Cao617ba132006-10-11 01:20:53 -0700331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700332 sb->s_flags |= MS_RDONLY;
Mingming Cao617ba132006-10-11 01:20:53 -0700333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
Mingming Caodab291a2006-10-11 01:21:01 -0700334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700335}
336
Mingming Cao617ba132006-10-11 01:20:53 -0700337void ext4_warning (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700338 const char * fmt, ...)
339{
340 va_list args;
341
342 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700344 sb->s_id, function);
345 vprintk(fmt, args);
346 printk("\n");
347 va_end(args);
348}
349
Mingming Cao617ba132006-10-11 01:20:53 -0700350void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700351{
Mingming Cao617ba132006-10-11 01:20:53 -0700352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700353
Mingming Cao617ba132006-10-11 01:20:53 -0700354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700355 return;
356
Harvey Harrison46e665e2008-04-17 10:38:59 -0400357 ext4_warning(sb, __func__,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700360 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700361
Mingming Cao617ba132006-10-11 01:20:53 -0700362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
367
368 /*
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
372 */
373}
374
Aneesh Kumar K.V99e6f822008-01-28 23:58:27 -0500375int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
377{
378 int err = 0;
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
381 EXT4_SB(sb)->s_sbh);
382 if (err)
383 return err;
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
385 sb->s_dirt = 1;
386 handle->h_sync = 1;
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
390 EXT4_SB(sb)->s_sbh);
391 }
392 return err;
393}
394
395int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
397{
398 int err = 0;
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
401 EXT4_SB(sb)->s_sbh);
402 if (err)
403 return err;
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 sb->s_dirt = 1;
406 handle->h_sync = 1;
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
410 EXT4_SB(sb)->s_sbh);
411 }
412 return err;
413}
414
415int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
417{
418 int err = 0;
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
421 EXT4_SB(sb)->s_sbh);
422 if (err)
423 return err;
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 sb->s_dirt = 1;
426 handle->h_sync = 1;
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
430 EXT4_SB(sb)->s_sbh);
431 }
432 return err;
433}
434
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700435/*
436 * Open the external journal device
437 */
Mingming Cao617ba132006-10-11 01:20:53 -0700438static struct block_device *ext4_blkdev_get(dev_t dev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700439{
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
442
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
444 if (IS_ERR(bdev))
445 goto fail;
446 return bdev;
447
448fail:
Mingming Cao617ba132006-10-11 01:20:53 -0700449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700450 __bdevname(dev, b), PTR_ERR(bdev));
451 return NULL;
452}
453
454/*
455 * Release the journal device
456 */
Mingming Cao617ba132006-10-11 01:20:53 -0700457static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700458{
459 bd_release(bdev);
460 return blkdev_put(bdev);
461}
462
Mingming Cao617ba132006-10-11 01:20:53 -0700463static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700464{
465 struct block_device *bdev;
466 int ret = -ENODEV;
467
468 bdev = sbi->journal_bdev;
469 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700470 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700471 sbi->journal_bdev = NULL;
472 }
473 return ret;
474}
475
476static inline struct inode *orphan_list_entry(struct list_head *l)
477{
Mingming Cao617ba132006-10-11 01:20:53 -0700478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700479}
480
Mingming Cao617ba132006-10-11 01:20:53 -0700481static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700482{
483 struct list_head *l;
484
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
487
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
491 printk(KERN_ERR " "
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
495 NEXT_ORPHAN(inode));
496 }
497}
498
Mingming Cao617ba132006-10-11 01:20:53 -0700499static void ext4_put_super (struct super_block * sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700500{
Mingming Cao617ba132006-10-11 01:20:53 -0700501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700503 int i;
504
Alex Tomasc9de5602008-01-29 00:19:52 -0500505 ext4_mb_release(sb);
Alex Tomasa86c6182006-10-11 01:21:03 -0700506 ext4_ext_release(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700507 ext4_xattr_put_super(sb);
Mingming Caodab291a2006-10-11 01:21:01 -0700508 jbd2_journal_destroy(sbi->s_journal);
Jan Kara47b4a502008-07-11 19:27:31 -0400509 sbi->s_journal = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700510 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700512 es->s_state = cpu_to_le16(sbi->s_mount_state);
513 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 mark_buffer_dirty(sbi->s_sbh);
Mingming Cao617ba132006-10-11 01:20:53 -0700515 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700516 }
517
518 for (i = 0; i < sbi->s_gdb_count; i++)
519 brelse(sbi->s_group_desc[i]);
520 kfree(sbi->s_group_desc);
Jose R. Santos772cb7c2008-07-11 19:27:31 -0400521 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700522 percpu_counter_destroy(&sbi->s_freeblocks_counter);
523 percpu_counter_destroy(&sbi->s_freeinodes_counter);
524 percpu_counter_destroy(&sbi->s_dirs_counter);
525 brelse(sbi->s_sbh);
526#ifdef CONFIG_QUOTA
527 for (i = 0; i < MAXQUOTAS; i++)
528 kfree(sbi->s_qf_names[i]);
529#endif
530
531 /* Debugging code just in case the in-memory inode orphan list
532 * isn't empty. The on-disk one can be non-empty if we've
533 * detected an error and taken the fs readonly, but the
534 * in-memory list had better be clean by this point. */
535 if (!list_empty(&sbi->s_orphan))
536 dump_orphan_list(sb, sbi);
537 J_ASSERT(list_empty(&sbi->s_orphan));
538
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700539 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700540 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
541 /*
542 * Invalidate the journal device's buffers. We don't want them
543 * floating about in memory - the physical journal device may
544 * hotswapped, and it breaks the `ro-after' testing code.
545 */
546 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700547 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700548 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700549 }
550 sb->s_fs_info = NULL;
551 kfree(sbi);
552 return;
553}
554
Christoph Lametere18b8902006-12-06 20:33:20 -0800555static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700556
557/*
558 * Called inside transaction, so use GFP_NOFS
559 */
Mingming Cao617ba132006-10-11 01:20:53 -0700560static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700561{
Mingming Cao617ba132006-10-11 01:20:53 -0700562 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700563
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800564 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700565 if (!ei)
566 return NULL;
Mingming Cao617ba132006-10-11 01:20:53 -0700567#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
568 ei->i_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700570#endif
571 ei->i_block_alloc_info = NULL;
572 ei->vfs_inode.i_version = 1;
Alex Tomasa86c6182006-10-11 01:21:03 -0700573 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500574 INIT_LIST_HEAD(&ei->i_prealloc_list);
575 spin_lock_init(&ei->i_prealloc_lock);
Jan Kara678aaf42008-07-11 19:27:31 -0400576 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
Mingming Caod2a17632008-07-14 17:52:37 -0400577 ei->i_reserved_data_blocks = 0;
578 ei->i_reserved_meta_blocks = 0;
579 ei->i_allocated_meta_blocks = 0;
580 ei->i_delalloc_reserved_flag = 0;
581 spin_lock_init(&(ei->i_block_reservation_lock));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700582 return &ei->vfs_inode;
583}
584
Mingming Cao617ba132006-10-11 01:20:53 -0700585static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700586{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700587 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
588 printk("EXT4 Inode %p: orphan list check failed!\n",
589 EXT4_I(inode));
590 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
591 EXT4_I(inode), sizeof(struct ext4_inode_info),
592 true);
593 dump_stack();
594 }
Mingming Cao617ba132006-10-11 01:20:53 -0700595 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700596}
597
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700598static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700599{
Mingming Cao617ba132006-10-11 01:20:53 -0700600 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601
Christoph Lametera35afb82007-05-16 22:10:57 -0700602 INIT_LIST_HEAD(&ei->i_orphan);
Mingming Cao617ba132006-10-11 01:20:53 -0700603#ifdef CONFIG_EXT4DEV_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700604 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700605#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500606 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700607 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700608}
609
610static int init_inodecache(void)
611{
Mingming Cao617ba132006-10-11 01:20:53 -0700612 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
613 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700614 0, (SLAB_RECLAIM_ACCOUNT|
615 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900616 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700617 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700618 return -ENOMEM;
619 return 0;
620}
621
622static void destroy_inodecache(void)
623{
Mingming Cao617ba132006-10-11 01:20:53 -0700624 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700625}
626
Mingming Cao617ba132006-10-11 01:20:53 -0700627static void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700628{
Mingming Cao617ba132006-10-11 01:20:53 -0700629 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
630#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
631 if (EXT4_I(inode)->i_acl &&
632 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
633 posix_acl_release(EXT4_I(inode)->i_acl);
634 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700635 }
Mingming Cao617ba132006-10-11 01:20:53 -0700636 if (EXT4_I(inode)->i_default_acl &&
637 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_default_acl);
639 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700640 }
641#endif
Mingming Cao617ba132006-10-11 01:20:53 -0700642 ext4_discard_reservation(inode);
643 EXT4_I(inode)->i_block_alloc_info = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700644 if (unlikely(rsv))
645 kfree(rsv);
Jan Kara678aaf42008-07-11 19:27:31 -0400646 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
647 &EXT4_I(inode)->jinode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700648}
649
Mingming Cao617ba132006-10-11 01:20:53 -0700650static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700651{
652#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700653 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700654
655 if (sbi->s_jquota_fmt)
656 seq_printf(seq, ",jqfmt=%s",
657 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
658
659 if (sbi->s_qf_names[USRQUOTA])
660 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
661
662 if (sbi->s_qf_names[GRPQUOTA])
663 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
664
Mingming Cao617ba132006-10-11 01:20:53 -0700665 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700666 seq_puts(seq, ",usrquota");
667
Mingming Cao617ba132006-10-11 01:20:53 -0700668 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700669 seq_puts(seq, ",grpquota");
670#endif
671}
672
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700673/*
674 * Show an option if
675 * - it's set to a non-default value OR
676 * - if the per-sb default is different from the global default
677 */
Mingming Cao617ba132006-10-11 01:20:53 -0700678static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700679{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500680 int def_errors;
681 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700682 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700683 struct ext4_sb_info *sbi = EXT4_SB(sb);
684 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700685
686 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500687 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700688
689 if (sbi->s_sb_block != 1)
690 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
691 if (test_opt(sb, MINIX_DF))
692 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500693 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700694 seq_puts(seq, ",grpid");
695 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
696 seq_puts(seq, ",nogrpid");
697 if (sbi->s_resuid != EXT4_DEF_RESUID ||
698 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
699 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
700 }
701 if (sbi->s_resgid != EXT4_DEF_RESGID ||
702 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
703 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
704 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500705 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700706 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500707 def_errors == EXT4_ERRORS_CONTINUE) {
708 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700709 }
710 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500711 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500712 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500713 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700714 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500715 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700716 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500717 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700718 seq_puts(seq, ",debug");
719 if (test_opt(sb, OLDALLOC))
720 seq_puts(seq, ",oldalloc");
Adrian Bunk07620f62008-01-28 23:58:27 -0500721#ifdef CONFIG_EXT4DEV_FS_XATTR
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500722 if (test_opt(sb, XATTR_USER) &&
723 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700724 seq_puts(seq, ",user_xattr");
725 if (!test_opt(sb, XATTR_USER) &&
726 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
727 seq_puts(seq, ",nouser_xattr");
728 }
729#endif
Adrian Bunk07620f62008-01-28 23:58:27 -0500730#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500731 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700732 seq_puts(seq, ",acl");
733 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
734 seq_puts(seq, ",noacl");
735#endif
736 if (!test_opt(sb, RESERVATION))
737 seq_puts(seq, ",noreservation");
738 if (sbi->s_commit_interval) {
739 seq_printf(seq, ",commit=%u",
740 (unsigned) (sbi->s_commit_interval / HZ));
741 }
Eric Sandeen571640c2008-05-26 12:29:46 -0400742 /*
743 * We're changing the default of barrier mount option, so
744 * let's always display its mount state so it's clear what its
745 * status is.
746 */
747 seq_puts(seq, ",barrier=");
748 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -0400749 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
750 seq_puts(seq, ",journal_async_commit");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700751 if (test_opt(sb, NOBH))
752 seq_puts(seq, ",nobh");
753 if (!test_opt(sb, EXTENTS))
754 seq_puts(seq, ",noextents");
Aneesh Kumar K.V3dbd0ed2008-01-28 23:58:26 -0500755 if (!test_opt(sb, MBALLOC))
756 seq_puts(seq, ",nomballoc");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -0500757 if (test_opt(sb, I_VERSION))
758 seq_puts(seq, ",i_version");
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -0400759 if (!test_opt(sb, DELALLOC))
760 seq_puts(seq, ",nodelalloc");
761
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700762
Miklos Szeredicb45bbe2008-01-28 23:58:27 -0500763 if (sbi->s_stripe)
764 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500765 /*
766 * journal mode get enabled in different ways
767 * So just print the value even if we didn't specify it
768 */
Mingming Cao617ba132006-10-11 01:20:53 -0700769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700770 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -0700771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700772 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -0700773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700774 seq_puts(seq, ",data=writeback");
775
Mingming Cao617ba132006-10-11 01:20:53 -0700776 ext4_show_quota_options(seq, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700777 return 0;
778}
779
780
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700781static struct inode *ext4_nfs_get_inode(struct super_block *sb,
782 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700783{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700784 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700785
Mingming Cao617ba132006-10-11 01:20:53 -0700786 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700787 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700788 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700789 return ERR_PTR(-ESTALE);
790
791 /* iget isn't really right if the inode is currently unallocated!!
792 *
Mingming Cao617ba132006-10-11 01:20:53 -0700793 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700794 * deleted, so we should be safe.
795 *
796 * Currently we don't know the generation for parent directory, so
797 * a generation of 0 means "accept any"
798 */
David Howells1d1fe1e2008-02-07 00:15:37 -0800799 inode = ext4_iget(sb, ino);
800 if (IS_ERR(inode))
801 return ERR_CAST(inode);
802 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700803 iput(inode);
804 return ERR_PTR(-ESTALE);
805 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700806
807 return inode;
808}
809
810static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
811 int fh_len, int fh_type)
812{
813 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
814 ext4_nfs_get_inode);
815}
816
817static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
818 int fh_len, int fh_type)
819{
820 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
821 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700822}
823
824#ifdef CONFIG_QUOTA
825#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
826#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
827
Mingming Cao617ba132006-10-11 01:20:53 -0700828static int ext4_dquot_initialize(struct inode *inode, int type);
829static int ext4_dquot_drop(struct inode *inode);
830static int ext4_write_dquot(struct dquot *dquot);
831static int ext4_acquire_dquot(struct dquot *dquot);
832static int ext4_release_dquot(struct dquot *dquot);
833static int ext4_mark_dquot_dirty(struct dquot *dquot);
834static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -0700835static int ext4_quota_on(struct super_block *sb, int type, int format_id,
836 char *path, int remount);
Mingming Cao617ba132006-10-11 01:20:53 -0700837static int ext4_quota_on_mount(struct super_block *sb, int type);
838static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700839 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -0700840static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700841 const char *data, size_t len, loff_t off);
842
Mingming Cao617ba132006-10-11 01:20:53 -0700843static struct dquot_operations ext4_quota_operations = {
844 .initialize = ext4_dquot_initialize,
845 .drop = ext4_dquot_drop,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700846 .alloc_space = dquot_alloc_space,
847 .alloc_inode = dquot_alloc_inode,
848 .free_space = dquot_free_space,
849 .free_inode = dquot_free_inode,
850 .transfer = dquot_transfer,
Mingming Cao617ba132006-10-11 01:20:53 -0700851 .write_dquot = ext4_write_dquot,
852 .acquire_dquot = ext4_acquire_dquot,
853 .release_dquot = ext4_release_dquot,
854 .mark_dirty = ext4_mark_dquot_dirty,
855 .write_info = ext4_write_info
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700856};
857
Mingming Cao617ba132006-10-11 01:20:53 -0700858static struct quotactl_ops ext4_qctl_operations = {
859 .quota_on = ext4_quota_on,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700860 .quota_off = vfs_quota_off,
861 .quota_sync = vfs_quota_sync,
862 .get_info = vfs_get_dqinfo,
863 .set_info = vfs_set_dqinfo,
864 .get_dqblk = vfs_get_dqblk,
865 .set_dqblk = vfs_set_dqblk
866};
867#endif
868
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800869static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -0700870 .alloc_inode = ext4_alloc_inode,
871 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -0700872 .write_inode = ext4_write_inode,
873 .dirty_inode = ext4_dirty_inode,
874 .delete_inode = ext4_delete_inode,
875 .put_super = ext4_put_super,
876 .write_super = ext4_write_super,
877 .sync_fs = ext4_sync_fs,
878 .write_super_lockfs = ext4_write_super_lockfs,
879 .unlockfs = ext4_unlockfs,
880 .statfs = ext4_statfs,
881 .remount_fs = ext4_remount,
882 .clear_inode = ext4_clear_inode,
883 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700884#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -0700885 .quota_read = ext4_quota_read,
886 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700887#endif
888};
889
Christoph Hellwig39655162007-10-21 16:42:17 -0700890static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700891 .fh_to_dentry = ext4_fh_to_dentry,
892 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -0700893 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700894};
895
896enum {
897 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
898 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
899 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
900 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
901 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
902 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -0500903 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700904 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
905 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
906 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
907 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -0500908 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -0400909 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700910};
911
912static match_table_t tokens = {
913 {Opt_bsd_df, "bsddf"},
914 {Opt_minix_df, "minixdf"},
915 {Opt_grpid, "grpid"},
916 {Opt_grpid, "bsdgroups"},
917 {Opt_nogrpid, "nogrpid"},
918 {Opt_nogrpid, "sysvgroups"},
919 {Opt_resgid, "resgid=%u"},
920 {Opt_resuid, "resuid=%u"},
921 {Opt_sb, "sb=%u"},
922 {Opt_err_cont, "errors=continue"},
923 {Opt_err_panic, "errors=panic"},
924 {Opt_err_ro, "errors=remount-ro"},
925 {Opt_nouid32, "nouid32"},
926 {Opt_nocheck, "nocheck"},
927 {Opt_nocheck, "check=none"},
928 {Opt_debug, "debug"},
929 {Opt_oldalloc, "oldalloc"},
930 {Opt_orlov, "orlov"},
931 {Opt_user_xattr, "user_xattr"},
932 {Opt_nouser_xattr, "nouser_xattr"},
933 {Opt_acl, "acl"},
934 {Opt_noacl, "noacl"},
935 {Opt_reservation, "reservation"},
936 {Opt_noreservation, "noreservation"},
937 {Opt_noload, "noload"},
938 {Opt_nobh, "nobh"},
939 {Opt_bh, "bh"},
940 {Opt_commit, "commit=%u"},
941 {Opt_journal_update, "journal=update"},
942 {Opt_journal_inum, "journal=%u"},
943 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -0500944 {Opt_journal_checksum, "journal_checksum"},
945 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700946 {Opt_abort, "abort"},
947 {Opt_data_journal, "data=journal"},
948 {Opt_data_ordered, "data=ordered"},
949 {Opt_data_writeback, "data=writeback"},
950 {Opt_offusrjquota, "usrjquota="},
951 {Opt_usrjquota, "usrjquota=%s"},
952 {Opt_offgrpjquota, "grpjquota="},
953 {Opt_grpjquota, "grpjquota=%s"},
954 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
955 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
956 {Opt_grpquota, "grpquota"},
957 {Opt_noquota, "noquota"},
958 {Opt_quota, "quota"},
959 {Opt_usrquota, "usrquota"},
960 {Opt_barrier, "barrier=%u"},
Alex Tomasa86c6182006-10-11 01:21:03 -0700961 {Opt_extents, "extents"},
Mingming Cao1e2462f2007-07-18 09:00:55 -0400962 {Opt_noextents, "noextents"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -0500963 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -0500964 {Opt_mballoc, "mballoc"},
965 {Opt_nomballoc, "nomballoc"},
966 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700967 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -0400968 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -0400969 {Opt_nodelalloc, "nodelalloc"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -0400970 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700971};
972
Mingming Cao617ba132006-10-11 01:20:53 -0700973static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700974{
Mingming Cao617ba132006-10-11 01:20:53 -0700975 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700976 char *options = (char *) *data;
977
978 if (!options || strncmp(options, "sb=", 3) != 0)
979 return 1; /* Default location */
980 options += 3;
Mingming Cao617ba132006-10-11 01:20:53 -0700981 /*todo: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700982 sb_block = simple_strtoul(options, &options, 0);
983 if (*options && *options != ',') {
Mingming Cao617ba132006-10-11 01:20:53 -0700984 printk("EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700985 (char *) *data);
986 return 1;
987 }
988 if (*options == ',')
989 options++;
990 *data = (void *) options;
991 return sb_block;
992}
993
994static int parse_options (char *options, struct super_block *sb,
995 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao617ba132006-10-11 01:20:53 -0700996 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700997{
Mingming Cao617ba132006-10-11 01:20:53 -0700998 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700999 char * p;
1000 substring_t args[MAX_OPT_ARGS];
1001 int data_opt = 0;
1002 int option;
1003#ifdef CONFIG_QUOTA
Jan Karadfc5d032008-05-13 19:11:51 -04001004 int qtype, qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001005 char *qname;
1006#endif
Aneesh Kumar K.Vc07651b2008-07-11 19:27:31 -04001007 ext4_fsblk_t last_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001008
1009 if (!options)
1010 return 1;
1011
1012 while ((p = strsep (&options, ",")) != NULL) {
1013 int token;
1014 if (!*p)
1015 continue;
1016
1017 token = match_token(p, tokens, args);
1018 switch (token) {
1019 case Opt_bsd_df:
1020 clear_opt (sbi->s_mount_opt, MINIX_DF);
1021 break;
1022 case Opt_minix_df:
1023 set_opt (sbi->s_mount_opt, MINIX_DF);
1024 break;
1025 case Opt_grpid:
1026 set_opt (sbi->s_mount_opt, GRPID);
1027 break;
1028 case Opt_nogrpid:
1029 clear_opt (sbi->s_mount_opt, GRPID);
1030 break;
1031 case Opt_resuid:
1032 if (match_int(&args[0], &option))
1033 return 0;
1034 sbi->s_resuid = option;
1035 break;
1036 case Opt_resgid:
1037 if (match_int(&args[0], &option))
1038 return 0;
1039 sbi->s_resgid = option;
1040 break;
1041 case Opt_sb:
1042 /* handled by get_sb_block() instead of here */
1043 /* *sb_block = match_int(&args[0]); */
1044 break;
1045 case Opt_err_panic:
1046 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1047 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1048 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1049 break;
1050 case Opt_err_ro:
1051 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1052 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1053 set_opt (sbi->s_mount_opt, ERRORS_RO);
1054 break;
1055 case Opt_err_cont:
1056 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1057 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1058 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1059 break;
1060 case Opt_nouid32:
1061 set_opt (sbi->s_mount_opt, NO_UID32);
1062 break;
1063 case Opt_nocheck:
1064 clear_opt (sbi->s_mount_opt, CHECK);
1065 break;
1066 case Opt_debug:
1067 set_opt (sbi->s_mount_opt, DEBUG);
1068 break;
1069 case Opt_oldalloc:
1070 set_opt (sbi->s_mount_opt, OLDALLOC);
1071 break;
1072 case Opt_orlov:
1073 clear_opt (sbi->s_mount_opt, OLDALLOC);
1074 break;
Mingming Cao617ba132006-10-11 01:20:53 -07001075#ifdef CONFIG_EXT4DEV_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001076 case Opt_user_xattr:
1077 set_opt (sbi->s_mount_opt, XATTR_USER);
1078 break;
1079 case Opt_nouser_xattr:
1080 clear_opt (sbi->s_mount_opt, XATTR_USER);
1081 break;
1082#else
1083 case Opt_user_xattr:
1084 case Opt_nouser_xattr:
Mingming Cao617ba132006-10-11 01:20:53 -07001085 printk("EXT4 (no)user_xattr options not supported\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001086 break;
1087#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001088#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001089 case Opt_acl:
1090 set_opt(sbi->s_mount_opt, POSIX_ACL);
1091 break;
1092 case Opt_noacl:
1093 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1094 break;
1095#else
1096 case Opt_acl:
1097 case Opt_noacl:
Mingming Cao617ba132006-10-11 01:20:53 -07001098 printk("EXT4 (no)acl options not supported\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -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) {
Mingming Cao617ba132006-10-11 01:20:53 -07001114 printk(KERN_ERR "EXT4-fs: cannot specify "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001115 "journal on remount\n");
1116 return 0;
1117 }
1118 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1119 break;
1120 case Opt_journal_inum:
1121 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001122 printk(KERN_ERR "EXT4-fs: cannot specify "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001123 "journal on remount\n");
1124 return 0;
1125 }
1126 if (match_int(&args[0], &option))
1127 return 0;
1128 *inum = option;
1129 break;
1130 case Opt_journal_dev:
1131 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001132 printk(KERN_ERR "EXT4-fs: cannot specify "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001133 "journal on remount\n");
1134 return 0;
1135 }
1136 if (match_int(&args[0], &option))
1137 return 0;
1138 *journal_devnum = option;
1139 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001140 case Opt_journal_checksum:
1141 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1142 break;
1143 case Opt_journal_async_commit:
1144 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1145 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1146 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001147 case Opt_noload:
1148 set_opt (sbi->s_mount_opt, NOLOAD);
1149 break;
1150 case Opt_commit:
1151 if (match_int(&args[0], &option))
1152 return 0;
1153 if (option < 0)
1154 return 0;
1155 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001156 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001157 sbi->s_commit_interval = HZ * option;
1158 break;
1159 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001160 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001161 goto datacheck;
1162 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001163 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001164 goto datacheck;
1165 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001166 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001167 datacheck:
1168 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001169 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001170 != data_opt) {
1171 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001172 "EXT4-fs: cannot change data "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001173 "mode on remount\n");
1174 return 0;
1175 }
1176 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07001177 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001178 sbi->s_mount_opt |= data_opt;
1179 }
1180 break;
1181#ifdef CONFIG_QUOTA
1182 case Opt_usrjquota:
1183 qtype = USRQUOTA;
1184 goto set_qf_name;
1185 case Opt_grpjquota:
1186 qtype = GRPQUOTA;
1187set_qf_name:
Jan Karadfc5d032008-05-13 19:11:51 -04001188 if ((sb_any_quota_enabled(sb) ||
1189 sb_any_quota_suspended(sb)) &&
1190 !sbi->s_qf_names[qtype]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001191 printk(KERN_ERR
Jan Kara2c8be6b2008-05-13 21:27:55 -04001192 "EXT4-fs: Cannot change journaled "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001193 "quota options when quota turned on.\n");
1194 return 0;
1195 }
1196 qname = match_strdup(&args[0]);
1197 if (!qname) {
1198 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001199 "EXT4-fs: not enough memory for "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001200 "storing quotafile name.\n");
1201 return 0;
1202 }
1203 if (sbi->s_qf_names[qtype] &&
1204 strcmp(sbi->s_qf_names[qtype], qname)) {
1205 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001206 "EXT4-fs: %s quota file already "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001207 "specified.\n", QTYPE2NAME(qtype));
1208 kfree(qname);
1209 return 0;
1210 }
1211 sbi->s_qf_names[qtype] = qname;
1212 if (strchr(sbi->s_qf_names[qtype], '/')) {
1213 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001214 "EXT4-fs: quotafile must be on "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001215 "filesystem root.\n");
1216 kfree(sbi->s_qf_names[qtype]);
1217 sbi->s_qf_names[qtype] = NULL;
1218 return 0;
1219 }
1220 set_opt(sbi->s_mount_opt, QUOTA);
1221 break;
1222 case Opt_offusrjquota:
1223 qtype = USRQUOTA;
1224 goto clear_qf_name;
1225 case Opt_offgrpjquota:
1226 qtype = GRPQUOTA;
1227clear_qf_name:
Jan Karadfc5d032008-05-13 19:11:51 -04001228 if ((sb_any_quota_enabled(sb) ||
1229 sb_any_quota_suspended(sb)) &&
1230 sbi->s_qf_names[qtype]) {
Mingming Cao617ba132006-10-11 01:20:53 -07001231 printk(KERN_ERR "EXT4-fs: Cannot change "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001232 "journaled quota options when "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001233 "quota turned on.\n");
1234 return 0;
1235 }
1236 /*
1237 * The space will be released later when all options
1238 * are confirmed to be correct
1239 */
1240 sbi->s_qf_names[qtype] = NULL;
1241 break;
1242 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001243 qfmt = QFMT_VFS_OLD;
1244 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001245 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001246 qfmt = QFMT_VFS_V0;
1247set_qf_format:
1248 if ((sb_any_quota_enabled(sb) ||
1249 sb_any_quota_suspended(sb)) &&
1250 sbi->s_jquota_fmt != qfmt) {
1251 printk(KERN_ERR "EXT4-fs: Cannot change "
1252 "journaled quota options when "
1253 "quota turned on.\n");
1254 return 0;
1255 }
1256 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001257 break;
1258 case Opt_quota:
1259 case Opt_usrquota:
1260 set_opt(sbi->s_mount_opt, QUOTA);
1261 set_opt(sbi->s_mount_opt, USRQUOTA);
1262 break;
1263 case Opt_grpquota:
1264 set_opt(sbi->s_mount_opt, QUOTA);
1265 set_opt(sbi->s_mount_opt, GRPQUOTA);
1266 break;
1267 case Opt_noquota:
1268 if (sb_any_quota_enabled(sb)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001269 printk(KERN_ERR "EXT4-fs: Cannot change quota "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001270 "options when quota turned on.\n");
1271 return 0;
1272 }
1273 clear_opt(sbi->s_mount_opt, QUOTA);
1274 clear_opt(sbi->s_mount_opt, USRQUOTA);
1275 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1276 break;
1277#else
1278 case Opt_quota:
1279 case Opt_usrquota:
1280 case Opt_grpquota:
Jan Karacd59e7b2008-05-13 19:11:51 -04001281 printk(KERN_ERR
1282 "EXT4-fs: quota options not supported.\n");
1283 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001284 case Opt_usrjquota:
1285 case Opt_grpjquota:
1286 case Opt_offusrjquota:
1287 case Opt_offgrpjquota:
1288 case Opt_jqfmt_vfsold:
1289 case Opt_jqfmt_vfsv0:
1290 printk(KERN_ERR
Jan Karacd59e7b2008-05-13 19:11:51 -04001291 "EXT4-fs: journaled quota options not "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001292 "supported.\n");
1293 break;
1294 case Opt_noquota:
1295 break;
1296#endif
1297 case Opt_abort:
1298 set_opt(sbi->s_mount_opt, ABORT);
1299 break;
1300 case Opt_barrier:
1301 if (match_int(&args[0], &option))
1302 return 0;
1303 if (option)
1304 set_opt(sbi->s_mount_opt, BARRIER);
1305 else
1306 clear_opt(sbi->s_mount_opt, BARRIER);
1307 break;
1308 case Opt_ignore:
1309 break;
1310 case Opt_resize:
1311 if (!is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001312 printk("EXT4-fs: resize option only available "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001313 "for remount\n");
1314 return 0;
1315 }
1316 if (match_int(&args[0], &option) != 0)
1317 return 0;
1318 *n_blocks_count = option;
1319 break;
1320 case Opt_nobh:
1321 set_opt(sbi->s_mount_opt, NOBH);
1322 break;
1323 case Opt_bh:
1324 clear_opt(sbi->s_mount_opt, NOBH);
1325 break;
Alex Tomasa86c6182006-10-11 01:21:03 -07001326 case Opt_extents:
Eric Sandeene4079a12008-07-11 19:27:31 -04001327 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1328 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1329 ext4_warning(sb, __func__,
1330 "extents feature not enabled "
1331 "on this filesystem, use tune2fs\n");
1332 return 0;
1333 }
Alex Tomasa86c6182006-10-11 01:21:03 -07001334 set_opt (sbi->s_mount_opt, EXTENTS);
1335 break;
Mingming Cao1e2462f2007-07-18 09:00:55 -04001336 case Opt_noextents:
Aneesh Kumar K.Vc07651b2008-07-11 19:27:31 -04001337 /*
1338 * When e2fsprogs support resizing an already existing
1339 * ext3 file system to greater than 2**32 we need to
1340 * add support to block allocator to handle growing
1341 * already existing block mapped inode so that blocks
1342 * allocated for them fall within 2**32
1343 */
1344 last_block = ext4_blocks_count(sbi->s_es) - 1;
1345 if (last_block > 0xffffffffULL) {
1346 printk(KERN_ERR "EXT4-fs: Filesystem too "
1347 "large to mount with "
1348 "-o noextents options\n");
1349 return 0;
1350 }
Mingming Cao1e2462f2007-07-18 09:00:55 -04001351 clear_opt (sbi->s_mount_opt, EXTENTS);
1352 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001353 case Opt_i_version:
1354 set_opt(sbi->s_mount_opt, I_VERSION);
1355 sb->s_flags |= MS_I_VERSION;
1356 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001357 case Opt_nodelalloc:
1358 clear_opt(sbi->s_mount_opt, DELALLOC);
1359 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001360 case Opt_mballoc:
1361 set_opt(sbi->s_mount_opt, MBALLOC);
1362 break;
1363 case Opt_nomballoc:
1364 clear_opt(sbi->s_mount_opt, MBALLOC);
1365 break;
1366 case Opt_stripe:
1367 if (match_int(&args[0], &option))
1368 return 0;
1369 if (option < 0)
1370 return 0;
1371 sbi->s_stripe = option;
1372 break;
Alex Tomas64769242008-07-11 19:27:31 -04001373 case Opt_delalloc:
1374 set_opt(sbi->s_mount_opt, DELALLOC);
1375 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001376 default:
1377 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001378 "EXT4-fs: Unrecognized mount option \"%s\" "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001379 "or missing value\n", p);
1380 return 0;
1381 }
1382 }
1383#ifdef CONFIG_QUOTA
1384 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Mingming Cao617ba132006-10-11 01:20:53 -07001385 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001386 sbi->s_qf_names[USRQUOTA])
1387 clear_opt(sbi->s_mount_opt, USRQUOTA);
1388
Mingming Cao617ba132006-10-11 01:20:53 -07001389 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001390 sbi->s_qf_names[GRPQUOTA])
1391 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1392
1393 if ((sbi->s_qf_names[USRQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001394 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001395 (sbi->s_qf_names[GRPQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001396 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1397 printk(KERN_ERR "EXT4-fs: old and new quota "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001398 "format mixing.\n");
1399 return 0;
1400 }
1401
1402 if (!sbi->s_jquota_fmt) {
Jan Kara2c8be6b2008-05-13 21:27:55 -04001403 printk(KERN_ERR "EXT4-fs: journaled quota format "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001404 "not specified.\n");
1405 return 0;
1406 }
1407 } else {
1408 if (sbi->s_jquota_fmt) {
Jan Kara2c8be6b2008-05-13 21:27:55 -04001409 printk(KERN_ERR "EXT4-fs: journaled quota format "
1410 "specified with no journaling "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001411 "enabled.\n");
1412 return 0;
1413 }
1414 }
1415#endif
1416 return 1;
1417}
1418
Mingming Cao617ba132006-10-11 01:20:53 -07001419static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001420 int read_only)
1421{
Mingming Cao617ba132006-10-11 01:20:53 -07001422 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001423 int res = 0;
1424
Mingming Cao617ba132006-10-11 01:20:53 -07001425 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1426 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001427 "forcing read-only mode\n");
1428 res = MS_RDONLY;
1429 }
1430 if (read_only)
1431 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001432 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1433 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001434 "running e2fsck is recommended\n");
Mingming Cao617ba132006-10-11 01:20:53 -07001435 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001436 printk (KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001437 "EXT4-fs warning: mounting fs with errors, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001438 "running e2fsck is recommended\n");
1439 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1440 le16_to_cpu(es->s_mnt_count) >=
1441 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1442 printk (KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001443 "EXT4-fs warning: maximal mount count reached, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001444 "running e2fsck is recommended\n");
1445 else if (le32_to_cpu(es->s_checkinterval) &&
1446 (le32_to_cpu(es->s_lastcheck) +
1447 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1448 printk (KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001449 "EXT4-fs warning: checktime reached, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001450 "running e2fsck is recommended\n");
1451#if 0
1452 /* @@@ We _will_ want to clear the valid bit if we find
Andrew Morton63f57932006-10-11 01:21:24 -07001453 * inconsistencies, to force a fsck at reboot. But for
1454 * a plain journaled filesystem we can keep it set as
1455 * valid forever! :)
1456 */
Marcin Slusarz216c34b2008-04-17 10:38:59 -04001457 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001458#endif
1459 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001460 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001461 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001462 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001463 ext4_update_dynamic_rev(sb);
1464 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001465
Mingming Cao617ba132006-10-11 01:20:53 -07001466 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001467 if (test_opt(sb, DEBUG))
Mingming Cao617ba132006-10-11 01:20:53 -07001468 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001469 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1470 sb->s_blocksize,
1471 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001472 EXT4_BLOCKS_PER_GROUP(sb),
1473 EXT4_INODES_PER_GROUP(sb),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001474 sbi->s_mount_opt);
1475
Mingming Cao617ba132006-10-11 01:20:53 -07001476 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1477 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001478 char b[BDEVNAME_SIZE];
1479
1480 printk("external journal on %s\n",
Mingming Cao617ba132006-10-11 01:20:53 -07001481 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001482 } else {
1483 printk("internal journal\n");
1484 }
1485 return res;
1486}
1487
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001488static int ext4_fill_flex_info(struct super_block *sb)
1489{
1490 struct ext4_sb_info *sbi = EXT4_SB(sb);
1491 struct ext4_group_desc *gdp = NULL;
1492 struct buffer_head *bh;
1493 ext4_group_t flex_group_count;
1494 ext4_group_t flex_group;
1495 int groups_per_flex = 0;
1496 __u64 block_bitmap = 0;
1497 int i;
1498
1499 if (!sbi->s_es->s_log_groups_per_flex) {
1500 sbi->s_log_groups_per_flex = 0;
1501 return 1;
1502 }
1503
1504 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1505 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1506
1507 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1508 groups_per_flex;
1509 sbi->s_flex_groups = kmalloc(flex_group_count *
1510 sizeof(struct flex_groups), GFP_KERNEL);
1511 if (sbi->s_flex_groups == NULL) {
1512 printk(KERN_ERR "EXT4-fs: not enough memory\n");
1513 goto failed;
1514 }
1515 memset(sbi->s_flex_groups, 0, flex_group_count *
1516 sizeof(struct flex_groups));
1517
1518 gdp = ext4_get_group_desc(sb, 1, &bh);
1519 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1520
1521 for (i = 0; i < sbi->s_groups_count; i++) {
1522 gdp = ext4_get_group_desc(sb, i, &bh);
1523
1524 flex_group = ext4_flex_group(sbi, i);
1525 sbi->s_flex_groups[flex_group].free_inodes +=
1526 le16_to_cpu(gdp->bg_free_inodes_count);
1527 sbi->s_flex_groups[flex_group].free_blocks +=
1528 le16_to_cpu(gdp->bg_free_blocks_count);
1529 }
1530
1531 return 1;
1532failed:
1533 return 0;
1534}
1535
Andreas Dilger717d50e2007-10-16 18:38:25 -04001536__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1537 struct ext4_group_desc *gdp)
1538{
1539 __u16 crc = 0;
1540
1541 if (sbi->s_es->s_feature_ro_compat &
1542 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1543 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1544 __le32 le_group = cpu_to_le32(block_group);
1545
1546 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1547 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1548 crc = crc16(crc, (__u8 *)gdp, offset);
1549 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1550 /* for checksum of struct ext4_group_desc do the rest...*/
1551 if ((sbi->s_es->s_feature_incompat &
1552 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1553 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1554 crc = crc16(crc, (__u8 *)gdp + offset,
1555 le16_to_cpu(sbi->s_es->s_desc_size) -
1556 offset);
1557 }
1558
1559 return cpu_to_le16(crc);
1560}
1561
1562int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1563 struct ext4_group_desc *gdp)
1564{
1565 if ((sbi->s_es->s_feature_ro_compat &
1566 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1567 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1568 return 0;
1569
1570 return 1;
1571}
1572
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001573/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001574static int ext4_check_descriptors(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001575{
Mingming Cao617ba132006-10-11 01:20:53 -07001576 struct ext4_sb_info *sbi = EXT4_SB(sb);
1577 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1578 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001579 ext4_fsblk_t block_bitmap;
1580 ext4_fsblk_t inode_bitmap;
1581 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04001582 int flexbg_flag = 0;
Avantika Mathurfd2d4292008-01-28 23:58:27 -05001583 ext4_group_t i;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001584
Jose R. Santosce421582007-10-16 18:38:25 -04001585 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1586 flexbg_flag = 1;
1587
Mingming Cao617ba132006-10-11 01:20:53 -07001588 ext4_debug ("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001589
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001590 for (i = 0; i < sbi->s_groups_count; i++) {
1591 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1592
Jose R. Santosce421582007-10-16 18:38:25 -04001593 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001594 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001595 else
1596 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07001597 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001598
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001599 block_bitmap = ext4_block_bitmap(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001600 if (block_bitmap < first_block || block_bitmap > last_block)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001601 {
Josef Bacikc19204b2008-04-29 22:00:28 -04001602 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1603 "Block bitmap for group %lu not in group "
1604 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001605 return 0;
1606 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001607 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001608 if (inode_bitmap < first_block || inode_bitmap > last_block)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001609 {
Josef Bacikc19204b2008-04-29 22:00:28 -04001610 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1611 "Inode bitmap for group %lu not in group "
1612 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001613 return 0;
1614 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001615 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001616 if (inode_table < first_block ||
Eric Sandeen780dcdb2007-07-26 10:41:11 -07001617 inode_table + sbi->s_itb_per_group - 1 > last_block)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001618 {
Josef Bacikc19204b2008-04-29 22:00:28 -04001619 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1620 "Inode table for group %lu not in group "
1621 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001622 return 0;
1623 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04001624 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Josef Bacikc19204b2008-04-29 22:00:28 -04001625 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1626 "Checksum for group %lu failed (%u!=%u)\n",
1627 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1628 gdp)), le16_to_cpu(gdp->bg_checksum));
Andreas Dilger717d50e2007-10-16 18:38:25 -04001629 return 0;
1630 }
Jose R. Santosce421582007-10-16 18:38:25 -04001631 if (!flexbg_flag)
1632 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001633 }
1634
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001635 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Mingming Cao617ba132006-10-11 01:20:53 -07001636 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001637 return 1;
1638}
1639
Mingming Cao617ba132006-10-11 01:20:53 -07001640/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001641 * the superblock) which were deleted from all directories, but held open by
1642 * a process at the time of a crash. We walk the list and try to delete these
1643 * inodes at recovery time (only with a read-write filesystem).
1644 *
1645 * In order to keep the orphan inode chain consistent during traversal (in
1646 * case of crash during recovery), we link each inode into the superblock
1647 * orphan list_head and handle it the same way as an inode deletion during
1648 * normal operation (which journals the operations for us).
1649 *
1650 * We only do an iget() and an iput() on each inode, which is very safe if we
1651 * accidentally point at an in-use or already deleted inode. The worst that
1652 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07001653 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001654 * e2fsck was run on this filesystem, and it must have already done the orphan
1655 * inode cleanup for us, so we can safely abort without any further action.
1656 */
Mingming Cao617ba132006-10-11 01:20:53 -07001657static void ext4_orphan_cleanup (struct super_block * sb,
1658 struct ext4_super_block * es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001659{
1660 unsigned int s_flags = sb->s_flags;
1661 int nr_orphans = 0, nr_truncates = 0;
1662#ifdef CONFIG_QUOTA
1663 int i;
1664#endif
1665 if (!es->s_last_orphan) {
1666 jbd_debug(4, "no orphan inodes to clean up\n");
1667 return;
1668 }
1669
Eric Sandeena8f48a92006-12-06 20:40:13 -08001670 if (bdev_read_only(sb->s_bdev)) {
1671 printk(KERN_ERR "EXT4-fs: write access "
1672 "unavailable, skipping orphan cleanup.\n");
1673 return;
1674 }
1675
Mingming Cao617ba132006-10-11 01:20:53 -07001676 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001677 if (es->s_last_orphan)
1678 jbd_debug(1, "Errors on filesystem, "
1679 "clearing orphan list.\n");
1680 es->s_last_orphan = 0;
1681 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1682 return;
1683 }
1684
1685 if (s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07001686 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001687 sb->s_id);
1688 sb->s_flags &= ~MS_RDONLY;
1689 }
1690#ifdef CONFIG_QUOTA
1691 /* Needed for iput() to work correctly and not trash data */
1692 sb->s_flags |= MS_ACTIVE;
1693 /* Turn on quotas so that they are updated correctly */
1694 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07001695 if (EXT4_SB(sb)->s_qf_names[i]) {
1696 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001697 if (ret < 0)
1698 printk(KERN_ERR
Jan Kara2c8be6b2008-05-13 21:27:55 -04001699 "EXT4-fs: Cannot turn on journaled "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001700 "quota: error %d\n", ret);
1701 }
1702 }
1703#endif
1704
1705 while (es->s_last_orphan) {
1706 struct inode *inode;
1707
Josef Bacik97bd42b2008-04-29 22:04:56 -04001708 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1709 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001710 es->s_last_orphan = 0;
1711 break;
1712 }
1713
Mingming Cao617ba132006-10-11 01:20:53 -07001714 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001715 DQUOT_INIT(inode);
1716 if (inode->i_nlink) {
1717 printk(KERN_DEBUG
1718 "%s: truncating inode %lu to %Ld bytes\n",
Harvey Harrison46e665e2008-04-17 10:38:59 -04001719 __func__, inode->i_ino, inode->i_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001720 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1721 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07001722 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001723 nr_truncates++;
1724 } else {
1725 printk(KERN_DEBUG
1726 "%s: deleting unreferenced inode %lu\n",
Harvey Harrison46e665e2008-04-17 10:38:59 -04001727 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001728 jbd_debug(2, "deleting unreferenced inode %lu\n",
1729 inode->i_ino);
1730 nr_orphans++;
1731 }
1732 iput(inode); /* The delete magic happens here! */
1733 }
1734
1735#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1736
1737 if (nr_orphans)
Mingming Cao617ba132006-10-11 01:20:53 -07001738 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001739 sb->s_id, PLURAL(nr_orphans));
1740 if (nr_truncates)
Mingming Cao617ba132006-10-11 01:20:53 -07001741 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001742 sb->s_id, PLURAL(nr_truncates));
1743#ifdef CONFIG_QUOTA
1744 /* Turn quotas off */
1745 for (i = 0; i < MAXQUOTAS; i++) {
1746 if (sb_dqopt(sb)->files[i])
Jan Kara6f28e082008-04-28 02:14:34 -07001747 vfs_quota_off(sb, i, 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001748 }
1749#endif
1750 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1751}
Eric Sandeencd2291a2008-01-28 23:58:27 -05001752/*
1753 * Maximal extent format file size.
1754 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1755 * extent format containers, within a sector_t, and within i_blocks
1756 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1757 * so that won't be a limiting factor.
1758 *
1759 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1760 */
1761static loff_t ext4_max_size(int blkbits)
1762{
1763 loff_t res;
1764 loff_t upper_limit = MAX_LFS_FILESIZE;
1765
1766 /* small i_blocks in vfs inode? */
1767 if (sizeof(blkcnt_t) < sizeof(u64)) {
1768 /*
1769 * CONFIG_LSF is not enabled implies the inode
1770 * i_block represent total blocks in 512 bytes
1771 * 32 == size of vfs inode i_blocks * 8
1772 */
1773 upper_limit = (1LL << 32) - 1;
1774
1775 /* total blocks in file system block size */
1776 upper_limit >>= (blkbits - 9);
1777 upper_limit <<= blkbits;
1778 }
1779
1780 /* 32-bit extent-start container, ee_block */
1781 res = 1LL << 32;
1782 res <<= blkbits;
1783 res -= 1;
1784
1785 /* Sanity check against vm- & vfs- imposed limits */
1786 if (res > upper_limit)
1787 res = upper_limit;
1788
1789 return res;
1790}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001791
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001792/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05001793 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001794 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1795 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001796 */
Eric Sandeencd2291a2008-01-28 23:58:27 -05001797static loff_t ext4_max_bitmap_size(int bits)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001798{
Mingming Cao617ba132006-10-11 01:20:53 -07001799 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001800 int meta_blocks;
1801 loff_t upper_limit;
1802 /* This is calculated to be the largest file size for a
Eric Sandeencd2291a2008-01-28 23:58:27 -05001803 * dense, bitmapped file such that the total number of
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001804 * sectors in the file, including data and all indirect blocks,
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001805 * does not exceed 2^48 -1
1806 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1807 * total number of 512 bytes blocks of the file
1808 */
1809
1810 if (sizeof(blkcnt_t) < sizeof(u64)) {
1811 /*
1812 * CONFIG_LSF is not enabled implies the inode
1813 * i_block represent total blocks in 512 bytes
1814 * 32 == size of vfs inode i_blocks * 8
1815 */
1816 upper_limit = (1LL << 32) - 1;
1817
1818 /* total blocks in file system block size */
1819 upper_limit >>= (bits - 9);
1820
1821 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05001822 /*
1823 * We use 48 bit ext4_inode i_blocks
1824 * With EXT4_HUGE_FILE_FL set the i_blocks
1825 * represent total number of blocks in
1826 * file system block size
1827 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001828 upper_limit = (1LL << 48) - 1;
1829
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001830 }
1831
1832 /* indirect blocks */
1833 meta_blocks = 1;
1834 /* double indirect blocks */
1835 meta_blocks += 1 + (1LL << (bits-2));
1836 /* tripple indirect blocks */
1837 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1838
1839 upper_limit -= meta_blocks;
1840 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001841
1842 res += 1LL << (bits-2);
1843 res += 1LL << (2*(bits-2));
1844 res += 1LL << (3*(bits-2));
1845 res <<= bits;
1846 if (res > upper_limit)
1847 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001848
1849 if (res > MAX_LFS_FILESIZE)
1850 res = MAX_LFS_FILESIZE;
1851
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001852 return res;
1853}
1854
Mingming Cao617ba132006-10-11 01:20:53 -07001855static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001856 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001857{
Mingming Cao617ba132006-10-11 01:20:53 -07001858 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05001859 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001860 int has_super = 0;
1861
1862 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1863
Mingming Cao617ba132006-10-11 01:20:53 -07001864 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001865 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001866 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001867 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07001868 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 has_super = 1;
Mingming Cao617ba132006-10-11 01:20:53 -07001870 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001871}
1872
Alex Tomasc9de5602008-01-29 00:19:52 -05001873/**
1874 * ext4_get_stripe_size: Get the stripe size.
1875 * @sbi: In memory super block info
1876 *
1877 * If we have specified it via mount option, then
1878 * use the mount option value. If the value specified at mount time is
1879 * greater than the blocks per group use the super block value.
1880 * If the super block value is greater than blocks per group return 0.
1881 * Allocator needs it be less than blocks per group.
1882 *
1883 */
1884static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1885{
1886 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1887 unsigned long stripe_width =
1888 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1889
1890 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1891 return sbi->s_stripe;
1892
1893 if (stripe_width <= sbi->s_blocks_per_group)
1894 return stripe_width;
1895
1896 if (stride <= sbi->s_blocks_per_group)
1897 return stride;
1898
1899 return 0;
1900}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001901
Mingming Cao617ba132006-10-11 01:20:53 -07001902static int ext4_fill_super (struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04001903 __releases(kernel_lock)
1904 __acquires(kernel_lock)
Aneesh Kumar K.V1d03ec92008-01-28 23:58:27 -05001905
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001906{
1907 struct buffer_head * bh;
Mingming Cao617ba132006-10-11 01:20:53 -07001908 struct ext4_super_block *es = NULL;
1909 struct ext4_sb_info *sbi;
1910 ext4_fsblk_t block;
1911 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001912 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001913 unsigned long offset = 0;
1914 unsigned int journal_inum = 0;
1915 unsigned long journal_devnum = 0;
1916 unsigned long def_mount_opts;
1917 struct inode *root;
David Howells1d1fe1e2008-02-07 00:15:37 -08001918 int ret = -EINVAL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001919 int blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001920 int db_count;
1921 int i;
1922 int needs_recovery;
1923 __le32 features;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001924 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07001925 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001926
1927 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1928 if (!sbi)
1929 return -ENOMEM;
1930 sb->s_fs_info = sbi;
1931 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07001932 sbi->s_resuid = EXT4_DEF_RESUID;
1933 sbi->s_resgid = EXT4_DEF_RESGID;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001934 sbi->s_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001935
1936 unlock_kernel();
1937
Mingming Cao617ba132006-10-11 01:20:53 -07001938 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001939 if (!blocksize) {
Mingming Cao617ba132006-10-11 01:20:53 -07001940 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001941 goto out_fail;
1942 }
1943
1944 /*
Mingming Cao617ba132006-10-11 01:20:53 -07001945 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001946 * block sizes. We need to calculate the offset from buffer start.
1947 */
Mingming Cao617ba132006-10-11 01:20:53 -07001948 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001949 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1950 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001951 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001952 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001953 }
1954
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001955 if (!(bh = sb_bread(sb, logical_sb_block))) {
Mingming Cao617ba132006-10-11 01:20:53 -07001956 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001957 goto out_fail;
1958 }
1959 /*
1960 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07001961 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001962 */
Mingming Cao617ba132006-10-11 01:20:53 -07001963 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001964 sbi->s_es = es;
1965 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07001966 if (sb->s_magic != EXT4_SUPER_MAGIC)
1967 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001968
1969 /* Set defaults before we parse the mount options */
1970 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Mingming Cao617ba132006-10-11 01:20:53 -07001971 if (def_mount_opts & EXT4_DEFM_DEBUG)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001972 set_opt(sbi->s_mount_opt, DEBUG);
Mingming Cao617ba132006-10-11 01:20:53 -07001973 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001974 set_opt(sbi->s_mount_opt, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07001975 if (def_mount_opts & EXT4_DEFM_UID16)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001976 set_opt(sbi->s_mount_opt, NO_UID32);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001977#ifdef CONFIG_EXT4DEV_FS_XATTR
Mingming Cao617ba132006-10-11 01:20:53 -07001978 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001979 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001980#endif
1981#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
Mingming Cao617ba132006-10-11 01:20:53 -07001982 if (def_mount_opts & EXT4_DEFM_ACL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001983 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001984#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001985 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1986 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1987 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1988 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1989 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1990 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001991
Mingming Cao617ba132006-10-11 01:20:53 -07001992 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001993 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05001994 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Dmitry Mishinceea16b2006-10-11 01:21:21 -07001995 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05001996 else
1997 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001998
1999 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2000 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2001
2002 set_opt(sbi->s_mount_opt, RESERVATION);
Eric Sandeen571640c2008-05-26 12:29:46 -04002003 set_opt(sbi->s_mount_opt, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002004
Mingming Cao1e2462f2007-07-18 09:00:55 -04002005 /*
2006 * turn on extents feature by default in ext4 filesystem
Eric Sandeene4079a12008-07-11 19:27:31 -04002007 * only if feature flag already set by mkfs or tune2fs.
2008 * Use -o noextents to turn it off
Mingming Cao1e2462f2007-07-18 09:00:55 -04002009 */
Eric Sandeene4079a12008-07-11 19:27:31 -04002010 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2011 set_opt(sbi->s_mount_opt, EXTENTS);
2012 else
2013 ext4_warning(sb, __func__,
2014 "extents feature not enabled on this filesystem, "
2015 "use tune2fs.\n");
Aneesh Kumar K.V3dbd0ed2008-01-28 23:58:26 -05002016 /*
Eric Sandeene4079a12008-07-11 19:27:31 -04002017 * turn on mballoc code by default in ext4 filesystem
2018 * Use -o nomballoc to turn it off
Aneesh Kumar K.V3dbd0ed2008-01-28 23:58:26 -05002019 */
2020 set_opt(sbi->s_mount_opt, MBALLOC);
Mingming Cao1e2462f2007-07-18 09:00:55 -04002021
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04002022 /*
2023 * enable delayed allocation by default
2024 * Use -o nodelalloc to turn it off
2025 */
2026 set_opt(sbi->s_mount_opt, DELALLOC);
2027
2028
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002029 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
2030 NULL, 0))
2031 goto failed_mount;
2032
2033 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07002034 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002035
Mingming Cao617ba132006-10-11 01:20:53 -07002036 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2037 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2038 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2039 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002040 printk(KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07002041 "EXT4-fs warning: feature flags set on rev 0 fs, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002042 "running e2fsck is recommended\n");
Theodore Tso469108f2008-02-10 01:11:44 -05002043
2044 /*
2045 * Since ext4 is still considered development code, we require
2046 * that the TEST_FILESYS flag in s->flags be set.
2047 */
2048 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2049 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2050 "OK to use with test code.\n", sb->s_id);
2051 goto failed_mount;
2052 }
2053
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002054 /*
2055 * Check feature flags regardless of the revision level, since we
2056 * previously didn't change the revision level when setting the flags,
2057 * so there is a chance incompat flags are set on a rev 0 filesystem.
2058 */
Mingming Cao617ba132006-10-11 01:20:53 -07002059 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002060 if (features) {
Mingming Cao617ba132006-10-11 01:20:53 -07002061 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002062 "unsupported optional features (%x).\n",
2063 sb->s_id, le32_to_cpu(features));
2064 goto failed_mount;
2065 }
Mingming Cao617ba132006-10-11 01:20:53 -07002066 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002067 if (!(sb->s_flags & MS_RDONLY) && features) {
Mingming Cao617ba132006-10-11 01:20:53 -07002068 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002069 "unsupported optional features (%x).\n",
2070 sb->s_id, le32_to_cpu(features));
2071 goto failed_mount;
2072 }
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002073 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2074 /*
2075 * Large file size enabled file system can only be
2076 * mount if kernel is build with CONFIG_LSF
2077 */
2078 if (sizeof(root->i_blocks) < sizeof(u64) &&
2079 !(sb->s_flags & MS_RDONLY)) {
2080 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2081 "files cannot be mounted read-write "
2082 "without CONFIG_LSF.\n", sb->s_id);
2083 goto failed_mount;
2084 }
2085 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002086 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2087
Mingming Cao617ba132006-10-11 01:20:53 -07002088 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2089 blocksize > EXT4_MAX_BLOCK_SIZE) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002090 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002091 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002092 blocksize, sb->s_id);
2093 goto failed_mount;
2094 }
2095
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002096 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002097
2098 /* Validate the filesystem blocksize */
2099 if (!sb_set_blocksize(sb, blocksize)) {
2100 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2101 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002102 goto failed_mount;
2103 }
2104
2105 brelse (bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002106 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2107 offset = do_div(logical_sb_block, blocksize);
2108 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002109 if (!bh) {
2110 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002111 "EXT4-fs: Can't read superblock on 2nd try.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002112 goto failed_mount;
2113 }
Mingming Cao617ba132006-10-11 01:20:53 -07002114 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002115 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07002116 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002117 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002118 "EXT4-fs: Magic mismatch, very weird !\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002119 goto failed_mount;
2120 }
2121 }
2122
Eric Sandeene2b46572008-01-28 23:58:27 -05002123 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
Mingming Cao617ba132006-10-11 01:20:53 -07002124 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002125
Mingming Cao617ba132006-10-11 01:20:53 -07002126 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2127 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2128 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002129 } else {
2130 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2131 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07002132 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04002133 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002134 (sbi->s_inode_size > blocksize)) {
2135 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002136 "EXT4-fs: unsupported inode size: %d\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002137 sbi->s_inode_size);
2138 goto failed_mount;
2139 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04002140 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2141 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002142 }
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002143 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2144 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002145 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002146 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07002147 !is_power_of_2(sbi->s_desc_size)) {
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002148 printk(KERN_ERR
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002149 "EXT4-fs: unsupported descriptor size %lu\n",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002150 sbi->s_desc_size);
2151 goto failed_mount;
2152 }
2153 } else
2154 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002155 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002156 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08002157 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002158 goto cantfind_ext4;
2159 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002160 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002161 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002162 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2163 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002164 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002165 sbi->s_sbh = bh;
2166 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07002167 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2168 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002169 for (i=0; i < 4; i++)
2170 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2171 sbi->s_def_hash_version = es->s_def_hash_version;
2172
2173 if (sbi->s_blocks_per_group > blocksize * 8) {
2174 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002175 "EXT4-fs: #blocks per group too big: %lu\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002176 sbi->s_blocks_per_group);
2177 goto failed_mount;
2178 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002179 if (sbi->s_inodes_per_group > blocksize * 8) {
2180 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002181 "EXT4-fs: #inodes per group too big: %lu\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002182 sbi->s_inodes_per_group);
2183 goto failed_mount;
2184 }
2185
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002186 if (ext4_blocks_count(es) >
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002187 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002188 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002189 " too large to mount safely\n", sb->s_id);
2190 if (sizeof(sector_t) < 8)
Mingming Cao617ba132006-10-11 01:20:53 -07002191 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002192 "enabled\n");
2193 goto failed_mount;
2194 }
2195
Mingming Cao617ba132006-10-11 01:20:53 -07002196 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2197 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05002198
2199 /* ensure blocks_count calculation below doesn't sign-extend */
2200 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2201 le32_to_cpu(es->s_first_data_block) + 1) {
2202 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2203 "first data block %u, blocks per group %lu\n",
2204 ext4_blocks_count(es),
2205 le32_to_cpu(es->s_first_data_block),
2206 EXT4_BLOCKS_PER_GROUP(sb));
2207 goto failed_mount;
2208 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002209 blocks_count = (ext4_blocks_count(es) -
2210 le32_to_cpu(es->s_first_data_block) +
2211 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2212 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2213 sbi->s_groups_count = blocks_count;
Mingming Cao617ba132006-10-11 01:20:53 -07002214 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2215 EXT4_DESC_PER_BLOCK(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002216 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2217 GFP_KERNEL);
2218 if (sbi->s_group_desc == NULL) {
Mingming Cao617ba132006-10-11 01:20:53 -07002219 printk (KERN_ERR "EXT4-fs: not enough memory\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002220 goto failed_mount;
2221 }
2222
2223 bgl_lock_init(&sbi->s_blockgroup_lock);
2224
2225 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002226 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002227 sbi->s_group_desc[i] = sb_bread(sb, block);
2228 if (!sbi->s_group_desc[i]) {
Mingming Cao617ba132006-10-11 01:20:53 -07002229 printk (KERN_ERR "EXT4-fs: "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002230 "can't read group descriptor %d\n", i);
2231 db_count = i;
2232 goto failed_mount2;
2233 }
2234 }
Mingming Cao617ba132006-10-11 01:20:53 -07002235 if (!ext4_check_descriptors (sb)) {
2236 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002237 goto failed_mount2;
2238 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002239 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2240 if (!ext4_fill_flex_info(sb)) {
2241 printk(KERN_ERR
2242 "EXT4-fs: unable to initialize "
2243 "flex_bg meta info!\n");
2244 goto failed_mount2;
2245 }
2246
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002247 sbi->s_gdb_count = db_count;
2248 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2249 spin_lock_init(&sbi->s_next_gen_lock);
2250
Peter Zijlstra833f4072007-10-16 23:25:45 -07002251 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2252 ext4_count_free_blocks(sb));
2253 if (!err) {
2254 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2255 ext4_count_free_inodes(sb));
2256 }
2257 if (!err) {
2258 err = percpu_counter_init(&sbi->s_dirs_counter,
2259 ext4_count_dirs(sb));
2260 }
2261 if (err) {
2262 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2263 goto failed_mount3;
2264 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002265
2266 /* per fileystem reservation list head & lock */
2267 spin_lock_init(&sbi->s_rsv_window_lock);
2268 sbi->s_rsv_window_root = RB_ROOT;
2269 /* Add a single, static dummy reservation to the start of the
2270 * reservation window list --- it gives us a placeholder for
2271 * append-at-start-of-list which makes the allocation logic
2272 * _much_ simpler. */
Mingming Cao617ba132006-10-11 01:20:53 -07002273 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2274 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002275 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2276 sbi->s_rsv_window_head.rsv_goal_size = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002277 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002278
Alex Tomasc9de5602008-01-29 00:19:52 -05002279 sbi->s_stripe = ext4_get_stripe_size(sbi);
2280
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002281 /*
2282 * set up enough so that it can read an inode
2283 */
Mingming Cao617ba132006-10-11 01:20:53 -07002284 sb->s_op = &ext4_sops;
2285 sb->s_export_op = &ext4_export_ops;
2286 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002287#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07002288 sb->s_qcop = &ext4_qctl_operations;
2289 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002290#endif
2291 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2292
2293 sb->s_root = NULL;
2294
2295 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07002296 EXT4_HAS_INCOMPAT_FEATURE(sb,
2297 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002298
2299 /*
2300 * The first inode we look at is the journal inode. Don't try
2301 * root first: it may be modified in the journal!
2302 */
2303 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07002304 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2305 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002306 goto failed_mount3;
Theodore Ts'o624080e2008-06-06 17:50:40 -04002307 if (!(sb->s_flags & MS_RDONLY) &&
2308 EXT4_SB(sb)->s_journal->j_failed_commit) {
2309 printk(KERN_CRIT "EXT4-fs error (device %s): "
2310 "ext4_fill_super: Journal transaction "
2311 "%u is corrupt\n", sb->s_id,
2312 EXT4_SB(sb)->s_journal->j_failed_commit);
2313 if (test_opt (sb, ERRORS_RO)) {
2314 printk (KERN_CRIT
2315 "Mounting filesystem read-only\n");
2316 sb->s_flags |= MS_RDONLY;
2317 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2318 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2319 }
2320 if (test_opt(sb, ERRORS_PANIC)) {
2321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2323 ext4_commit_super(sb, es, 1);
2324 printk(KERN_CRIT
2325 "EXT4-fs (device %s): mount failed\n",
2326 sb->s_id);
2327 goto failed_mount4;
2328 }
2329 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002330 } else if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07002331 if (ext4_create_journal(sb, es, journal_inum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002332 goto failed_mount3;
2333 } else {
2334 if (!silent)
2335 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002336 "ext4: No journal on filesystem on %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002337 sb->s_id);
2338 goto failed_mount3;
2339 }
2340
Jose R. Santoseb40a092007-07-18 08:37:25 -04002341 if (ext4_blocks_count(es) > 0xffffffffULL &&
2342 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2343 JBD2_FEATURE_INCOMPAT_64BIT)) {
2344 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2345 goto failed_mount4;
2346 }
2347
Girish Shilamkar818d2762008-01-28 23:58:27 -05002348 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2349 jbd2_journal_set_features(sbi->s_journal,
2350 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2351 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2352 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2353 jbd2_journal_set_features(sbi->s_journal,
2354 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2355 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2356 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2357 } else {
2358 jbd2_journal_clear_features(sbi->s_journal,
2359 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2360 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2361 }
2362
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002363 /* We have now updated the journal if required, so we can
2364 * validate the data journaling mode. */
2365 switch (test_opt(sb, DATA_FLAGS)) {
2366 case 0:
2367 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07002368 * capabilities: ORDERED_DATA if the journal can
2369 * cope, else JOURNAL_DATA
2370 */
Mingming Caodab291a2006-10-11 01:21:01 -07002371 if (jbd2_journal_check_available_features
2372 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002373 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2374 else
2375 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2376 break;
2377
Mingming Cao617ba132006-10-11 01:20:53 -07002378 case EXT4_MOUNT_ORDERED_DATA:
2379 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07002380 if (!jbd2_journal_check_available_features
2381 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002382 printk(KERN_ERR "EXT4-fs: Journal does not support "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002383 "requested data journaling mode\n");
2384 goto failed_mount4;
2385 }
2386 default:
2387 break;
2388 }
2389
2390 if (test_opt(sb, NOBH)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002391 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2392 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002393 "its supported only with writeback mode\n");
2394 clear_opt(sbi->s_mount_opt, NOBH);
2395 }
2396 }
2397 /*
Mingming Caodab291a2006-10-11 01:21:01 -07002398 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002399 * so we can safely mount the rest of the filesystem now.
2400 */
2401
David Howells1d1fe1e2008-02-07 00:15:37 -08002402 root = ext4_iget(sb, EXT4_ROOT_INO);
2403 if (IS_ERR(root)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002404 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
David Howells1d1fe1e2008-02-07 00:15:37 -08002405 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002406 goto failed_mount4;
2407 }
2408 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08002409 iput(root);
Mingming Cao617ba132006-10-11 01:20:53 -07002410 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002411 goto failed_mount4;
2412 }
David Howells1d1fe1e2008-02-07 00:15:37 -08002413 sb->s_root = d_alloc_root(root);
2414 if (!sb->s_root) {
2415 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2416 iput(root);
2417 ret = -ENOMEM;
2418 goto failed_mount4;
2419 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002420
Mingming Cao617ba132006-10-11 01:20:53 -07002421 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04002422
2423 /* determine the minimum size of new large inodes, if present */
2424 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2425 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2426 EXT4_GOOD_OLD_INODE_SIZE;
2427 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2428 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2429 if (sbi->s_want_extra_isize <
2430 le16_to_cpu(es->s_want_extra_isize))
2431 sbi->s_want_extra_isize =
2432 le16_to_cpu(es->s_want_extra_isize);
2433 if (sbi->s_want_extra_isize <
2434 le16_to_cpu(es->s_min_extra_isize))
2435 sbi->s_want_extra_isize =
2436 le16_to_cpu(es->s_min_extra_isize);
2437 }
2438 }
2439 /* Check if enough inode space is available */
2440 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2441 sbi->s_inode_size) {
2442 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2443 EXT4_GOOD_OLD_INODE_SIZE;
2444 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2445 "available.\n");
2446 }
2447
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002448 /*
2449 * akpm: core read_super() calls in here with the superblock locked.
2450 * That deadlocks, because orphan cleanup needs to lock the superblock
2451 * in numerous places. Here we just pop the lock - it's relatively
2452 * harmless, because we are now ready to accept write_super() requests,
2453 * and aviro says that's the only reason for hanging onto the
2454 * superblock lock.
2455 */
Mingming Cao617ba132006-10-11 01:20:53 -07002456 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2457 ext4_orphan_cleanup(sb, es);
2458 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002459 if (needs_recovery)
Mingming Cao617ba132006-10-11 01:20:53 -07002460 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2461 ext4_mark_recovery_complete(sb, es);
2462 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2463 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2464 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002465 "writeback");
2466
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04002467 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2468 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2469 "requested data journaling mode\n");
2470 clear_opt(sbi->s_mount_opt, DELALLOC);
2471 } else if (test_opt(sb, DELALLOC))
2472 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2473
Alex Tomasa86c6182006-10-11 01:21:03 -07002474 ext4_ext_init(sb);
Alex Tomasc9de5602008-01-29 00:19:52 -05002475 ext4_mb_init(sb, needs_recovery);
Alex Tomasa86c6182006-10-11 01:21:03 -07002476
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002477 lock_kernel();
2478 return 0;
2479
Mingming Cao617ba132006-10-11 01:20:53 -07002480cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002481 if (!silent)
Mingming Cao617ba132006-10-11 01:20:53 -07002482 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002483 sb->s_id);
2484 goto failed_mount;
2485
2486failed_mount4:
Mingming Caodab291a2006-10-11 01:21:01 -07002487 jbd2_journal_destroy(sbi->s_journal);
Jan Kara47b4a502008-07-11 19:27:31 -04002488 sbi->s_journal = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002489failed_mount3:
2490 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2491 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2492 percpu_counter_destroy(&sbi->s_dirs_counter);
2493failed_mount2:
2494 for (i = 0; i < db_count; i++)
2495 brelse(sbi->s_group_desc[i]);
2496 kfree(sbi->s_group_desc);
2497failed_mount:
2498#ifdef CONFIG_QUOTA
2499 for (i = 0; i < MAXQUOTAS; i++)
2500 kfree(sbi->s_qf_names[i]);
2501#endif
Mingming Cao617ba132006-10-11 01:20:53 -07002502 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002503 brelse(bh);
2504out_fail:
2505 sb->s_fs_info = NULL;
2506 kfree(sbi);
2507 lock_kernel();
David Howells1d1fe1e2008-02-07 00:15:37 -08002508 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002509}
2510
2511/*
2512 * Setup any per-fs journal parameters now. We'll do this both on
2513 * initial mount, once the journal has been initialised but before we've
2514 * done any recovery; and again on any subsequent remount.
2515 */
Mingming Cao617ba132006-10-11 01:20:53 -07002516static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002517{
Mingming Cao617ba132006-10-11 01:20:53 -07002518 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002519
2520 if (sbi->s_commit_interval)
2521 journal->j_commit_interval = sbi->s_commit_interval;
Mingming Cao617ba132006-10-11 01:20:53 -07002522 /* We could also set up an ext4-specific default for the commit
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002523 * interval here, but for now we'll just fall back to the jbd
2524 * default. */
2525
2526 spin_lock(&journal->j_state_lock);
2527 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07002528 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002529 else
Mingming Caodab291a2006-10-11 01:21:01 -07002530 journal->j_flags &= ~JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002531 spin_unlock(&journal->j_state_lock);
2532}
2533
Mingming Cao617ba132006-10-11 01:20:53 -07002534static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002535 unsigned int journal_inum)
2536{
2537 struct inode *journal_inode;
2538 journal_t *journal;
2539
2540 /* First, test for the existence of a valid inode on disk. Bad
2541 * things happen if we iget() an unused inode, as the subsequent
2542 * iput() will try to delete it. */
2543
David Howells1d1fe1e2008-02-07 00:15:37 -08002544 journal_inode = ext4_iget(sb, journal_inum);
2545 if (IS_ERR(journal_inode)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002546 printk(KERN_ERR "EXT4-fs: no journal found.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002547 return NULL;
2548 }
2549 if (!journal_inode->i_nlink) {
2550 make_bad_inode(journal_inode);
2551 iput(journal_inode);
Mingming Cao617ba132006-10-11 01:20:53 -07002552 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002553 return NULL;
2554 }
2555
2556 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2557 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08002558 if (!S_ISREG(journal_inode->i_mode)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002559 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002560 iput(journal_inode);
2561 return NULL;
2562 }
2563
Mingming Caodab291a2006-10-11 01:21:01 -07002564 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002565 if (!journal) {
Mingming Cao617ba132006-10-11 01:20:53 -07002566 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002567 iput(journal_inode);
2568 return NULL;
2569 }
2570 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07002571 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002572 return journal;
2573}
2574
Mingming Cao617ba132006-10-11 01:20:53 -07002575static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002576 dev_t j_dev)
2577{
2578 struct buffer_head * bh;
2579 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07002580 ext4_fsblk_t start;
2581 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002582 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07002583 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002584 unsigned long offset;
Mingming Cao617ba132006-10-11 01:20:53 -07002585 struct ext4_super_block * es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002586 struct block_device *bdev;
2587
Mingming Cao617ba132006-10-11 01:20:53 -07002588 bdev = ext4_blkdev_get(j_dev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002589 if (bdev == NULL)
2590 return NULL;
2591
2592 if (bd_claim(bdev, sb)) {
2593 printk(KERN_ERR
Dave Kleikamp8c55e202007-05-24 13:04:54 -04002594 "EXT4: failed to claim external journal device.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002595 blkdev_put(bdev);
2596 return NULL;
2597 }
2598
2599 blocksize = sb->s_blocksize;
2600 hblock = bdev_hardsect_size(bdev);
2601 if (blocksize < hblock) {
2602 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07002603 "EXT4-fs: blocksize too small for journal device.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002604 goto out_bdev;
2605 }
2606
Mingming Cao617ba132006-10-11 01:20:53 -07002607 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2608 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002609 set_blocksize(bdev, blocksize);
2610 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Mingming Cao617ba132006-10-11 01:20:53 -07002611 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002612 "external journal\n");
2613 goto out_bdev;
2614 }
2615
Mingming Cao617ba132006-10-11 01:20:53 -07002616 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2617 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002618 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07002619 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2620 printk(KERN_ERR "EXT4-fs: external journal has "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002621 "bad superblock\n");
2622 brelse(bh);
2623 goto out_bdev;
2624 }
2625
Mingming Cao617ba132006-10-11 01:20:53 -07002626 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2627 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002628 brelse(bh);
2629 goto out_bdev;
2630 }
2631
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002632 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002633 start = sb_block + 1;
2634 brelse(bh); /* we're done with the superblock */
2635
Mingming Caodab291a2006-10-11 01:21:01 -07002636 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002637 start, len, blocksize);
2638 if (!journal) {
Mingming Cao617ba132006-10-11 01:20:53 -07002639 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002640 goto out_bdev;
2641 }
2642 journal->j_private = sb;
2643 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2644 wait_on_buffer(journal->j_sb_buffer);
2645 if (!buffer_uptodate(journal->j_sb_buffer)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002646 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002647 goto out_journal;
2648 }
2649 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Mingming Cao617ba132006-10-11 01:20:53 -07002650 printk(KERN_ERR "EXT4-fs: External journal has more than one "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002651 "user (unsupported) - %d\n",
2652 be32_to_cpu(journal->j_superblock->s_nr_users));
2653 goto out_journal;
2654 }
Mingming Cao617ba132006-10-11 01:20:53 -07002655 EXT4_SB(sb)->journal_bdev = bdev;
2656 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002657 return journal;
2658out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07002659 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002660out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07002661 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002662 return NULL;
2663}
2664
Mingming Cao617ba132006-10-11 01:20:53 -07002665static int ext4_load_journal(struct super_block *sb,
2666 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002667 unsigned long journal_devnum)
2668{
2669 journal_t *journal;
2670 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2671 dev_t journal_dev;
2672 int err = 0;
2673 int really_read_only;
2674
2675 if (journal_devnum &&
2676 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002677 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002678 "numbers have changed\n");
2679 journal_dev = new_decode_dev(journal_devnum);
2680 } else
2681 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2682
2683 really_read_only = bdev_read_only(sb->s_bdev);
2684
2685 /*
2686 * Are we loading a blank journal or performing recovery after a
2687 * crash? For recovery, we need to check in advance whether we
2688 * can get read-write access to the device.
2689 */
2690
Mingming Cao617ba132006-10-11 01:20:53 -07002691 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002692 if (sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07002693 printk(KERN_INFO "EXT4-fs: INFO: recovery "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002694 "required on readonly filesystem.\n");
2695 if (really_read_only) {
Mingming Cao617ba132006-10-11 01:20:53 -07002696 printk(KERN_ERR "EXT4-fs: write access "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002697 "unavailable, cannot proceed.\n");
2698 return -EROFS;
2699 }
Mingming Cao617ba132006-10-11 01:20:53 -07002700 printk (KERN_INFO "EXT4-fs: write access will "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002701 "be enabled during recovery.\n");
2702 }
2703 }
2704
2705 if (journal_inum && journal_dev) {
Mingming Cao617ba132006-10-11 01:20:53 -07002706 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002707 "and inode journals!\n");
2708 return -EINVAL;
2709 }
2710
2711 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07002712 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002713 return -EINVAL;
2714 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07002715 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002716 return -EINVAL;
2717 }
2718
2719 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07002720 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002721 if (err) {
Mingming Cao617ba132006-10-11 01:20:53 -07002722 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
Mingming Caodab291a2006-10-11 01:21:01 -07002723 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002724 return err;
2725 }
2726 }
2727
Mingming Cao617ba132006-10-11 01:20:53 -07002728 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07002729 err = jbd2_journal_wipe(journal, !really_read_only);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002730 if (!err)
Mingming Caodab291a2006-10-11 01:21:01 -07002731 err = jbd2_journal_load(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002732
2733 if (err) {
Mingming Cao617ba132006-10-11 01:20:53 -07002734 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
Mingming Caodab291a2006-10-11 01:21:01 -07002735 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002736 return err;
2737 }
2738
Mingming Cao617ba132006-10-11 01:20:53 -07002739 EXT4_SB(sb)->s_journal = journal;
2740 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002741
2742 if (journal_devnum &&
2743 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2744 es->s_journal_dev = cpu_to_le32(journal_devnum);
2745 sb->s_dirt = 1;
2746
2747 /* Make sure we flush the recovery flag to disk. */
Mingming Cao617ba132006-10-11 01:20:53 -07002748 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002749 }
2750
2751 return 0;
2752}
2753
Mingming Cao617ba132006-10-11 01:20:53 -07002754static int ext4_create_journal(struct super_block * sb,
2755 struct ext4_super_block * es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002756 unsigned int journal_inum)
2757{
2758 journal_t *journal;
Borislav Petkov6c675bd2007-07-15 23:41:45 -07002759 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002760
2761 if (sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07002762 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002763 "create journal.\n");
2764 return -EROFS;
2765 }
2766
Borislav Petkov6c675bd2007-07-15 23:41:45 -07002767 journal = ext4_get_journal(sb, journal_inum);
2768 if (!journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002769 return -EINVAL;
2770
Mingming Cao617ba132006-10-11 01:20:53 -07002771 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002772 journal_inum);
2773
Borislav Petkov6c675bd2007-07-15 23:41:45 -07002774 err = jbd2_journal_create(journal);
2775 if (err) {
Mingming Cao617ba132006-10-11 01:20:53 -07002776 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
Mingming Caodab291a2006-10-11 01:21:01 -07002777 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002778 return -EIO;
2779 }
2780
Mingming Cao617ba132006-10-11 01:20:53 -07002781 EXT4_SB(sb)->s_journal = journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002782
Mingming Cao617ba132006-10-11 01:20:53 -07002783 ext4_update_dynamic_rev(sb);
2784 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2785 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002786
2787 es->s_journal_inum = cpu_to_le32(journal_inum);
2788 sb->s_dirt = 1;
2789
2790 /* Make sure we flush the recovery flag to disk. */
Mingming Cao617ba132006-10-11 01:20:53 -07002791 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002792
2793 return 0;
2794}
2795
Mingming Cao617ba132006-10-11 01:20:53 -07002796static void ext4_commit_super (struct super_block * sb,
2797 struct ext4_super_block * es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002798 int sync)
2799{
Mingming Cao617ba132006-10-11 01:20:53 -07002800 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002801
2802 if (!sbh)
2803 return;
2804 es->s_wtime = cpu_to_le32(get_seconds());
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002805 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
Mingming Cao617ba132006-10-11 01:20:53 -07002806 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002807 BUFFER_TRACE(sbh, "marking dirty");
2808 mark_buffer_dirty(sbh);
2809 if (sync)
2810 sync_dirty_buffer(sbh);
2811}
2812
2813
2814/*
2815 * Have we just finished recovery? If so, and if we are mounting (or
2816 * remounting) the filesystem readonly, then we will end up with a
2817 * consistent fs on disk. Record that fact.
2818 */
Mingming Cao617ba132006-10-11 01:20:53 -07002819static void ext4_mark_recovery_complete(struct super_block * sb,
2820 struct ext4_super_block * es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002821{
Mingming Cao617ba132006-10-11 01:20:53 -07002822 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002823
Mingming Caodab291a2006-10-11 01:21:01 -07002824 jbd2_journal_lock_updates(journal);
2825 jbd2_journal_flush(journal);
Jan Kara32c37732007-07-15 23:41:09 -07002826 lock_super(sb);
Mingming Cao617ba132006-10-11 01:20:53 -07002827 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002828 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07002829 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002830 sb->s_dirt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002831 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002832 }
Jan Kara32c37732007-07-15 23:41:09 -07002833 unlock_super(sb);
Mingming Caodab291a2006-10-11 01:21:01 -07002834 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002835}
2836
2837/*
2838 * If we are mounting (or read-write remounting) a filesystem whose journal
2839 * has recorded an error from a previous lifetime, move that error to the
2840 * main filesystem now.
2841 */
Mingming Cao617ba132006-10-11 01:20:53 -07002842static void ext4_clear_journal_err(struct super_block * sb,
2843 struct ext4_super_block * es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002844{
2845 journal_t *journal;
2846 int j_errno;
2847 const char *errstr;
2848
Mingming Cao617ba132006-10-11 01:20:53 -07002849 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002850
2851 /*
2852 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07002853 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002854 */
2855
Mingming Caodab291a2006-10-11 01:21:01 -07002856 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002857 if (j_errno) {
2858 char nbuf[16];
2859
Mingming Cao617ba132006-10-11 01:20:53 -07002860 errstr = ext4_decode_error(sb, j_errno, nbuf);
Harvey Harrison46e665e2008-04-17 10:38:59 -04002861 ext4_warning(sb, __func__, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002862 "from previous mount: %s", errstr);
Harvey Harrison46e665e2008-04-17 10:38:59 -04002863 ext4_warning(sb, __func__, "Marking fs in need of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002864 "filesystem check.");
2865
Mingming Cao617ba132006-10-11 01:20:53 -07002866 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2867 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2868 ext4_commit_super (sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002869
Mingming Caodab291a2006-10-11 01:21:01 -07002870 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002871 }
2872}
2873
2874/*
2875 * Force the running and committing transactions to commit,
2876 * and wait on the commit.
2877 */
Mingming Cao617ba132006-10-11 01:20:53 -07002878int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002879{
2880 journal_t *journal;
2881 int ret;
2882
2883 if (sb->s_flags & MS_RDONLY)
2884 return 0;
2885
Mingming Cao617ba132006-10-11 01:20:53 -07002886 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002887 sb->s_dirt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002888 ret = ext4_journal_force_commit(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002889 return ret;
2890}
2891
2892/*
Mingming Cao617ba132006-10-11 01:20:53 -07002893 * Ext4 always journals updates to the superblock itself, so we don't
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002894 * have to propagate any other updates to the superblock on disk at this
2895 * point. Just start an async writeback to get the buffers on their way
2896 * to the disk.
2897 *
2898 * This implicitly triggers the writebehind on sync().
2899 */
2900
Mingming Cao617ba132006-10-11 01:20:53 -07002901static void ext4_write_super (struct super_block * sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002902{
2903 if (mutex_trylock(&sb->s_lock) != 0)
2904 BUG();
2905 sb->s_dirt = 0;
2906}
2907
Mingming Cao617ba132006-10-11 01:20:53 -07002908static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002909{
2910 tid_t target;
2911
2912 sb->s_dirt = 0;
Mingming Caodab291a2006-10-11 01:21:01 -07002913 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002914 if (wait)
Mingming Caodab291a2006-10-11 01:21:01 -07002915 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002916 }
2917 return 0;
2918}
2919
2920/*
2921 * LVM calls this function before a (read-only) snapshot is created. This
2922 * gives us a chance to flush the journal completely and mark the fs clean.
2923 */
Mingming Cao617ba132006-10-11 01:20:53 -07002924static void ext4_write_super_lockfs(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002925{
2926 sb->s_dirt = 0;
2927
2928 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002929 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002930
2931 /* Now we set up the journal barrier. */
Mingming Caodab291a2006-10-11 01:21:01 -07002932 jbd2_journal_lock_updates(journal);
2933 jbd2_journal_flush(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002934
2935 /* Journal blocked and flushed, clear needs_recovery flag. */
Mingming Cao617ba132006-10-11 01:20:53 -07002936 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2937 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002938 }
2939}
2940
2941/*
2942 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2943 * flag here, even though the filesystem is not technically dirty yet.
2944 */
Mingming Cao617ba132006-10-11 01:20:53 -07002945static void ext4_unlockfs(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002946{
2947 if (!(sb->s_flags & MS_RDONLY)) {
2948 lock_super(sb);
2949 /* Reser the needs_recovery flag before the fs is unlocked. */
Mingming Cao617ba132006-10-11 01:20:53 -07002950 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2951 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002952 unlock_super(sb);
Mingming Caodab291a2006-10-11 01:21:01 -07002953 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002954 }
2955}
2956
Mingming Cao617ba132006-10-11 01:20:53 -07002957static int ext4_remount (struct super_block * sb, int * flags, char * data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002958{
Mingming Cao617ba132006-10-11 01:20:53 -07002959 struct ext4_super_block * es;
2960 struct ext4_sb_info *sbi = EXT4_SB(sb);
2961 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002962 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07002963 struct ext4_mount_options old_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002964 int err;
2965#ifdef CONFIG_QUOTA
2966 int i;
2967#endif
2968
2969 /* Store the original options */
2970 old_sb_flags = sb->s_flags;
2971 old_opts.s_mount_opt = sbi->s_mount_opt;
2972 old_opts.s_resuid = sbi->s_resuid;
2973 old_opts.s_resgid = sbi->s_resgid;
2974 old_opts.s_commit_interval = sbi->s_commit_interval;
2975#ifdef CONFIG_QUOTA
2976 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2977 for (i = 0; i < MAXQUOTAS; i++)
2978 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2979#endif
2980
2981 /*
2982 * Allow the "check" option to be passed as a remount option.
2983 */
2984 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2985 err = -EINVAL;
2986 goto restore_opts;
2987 }
2988
Mingming Cao617ba132006-10-11 01:20:53 -07002989 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
Harvey Harrison46e665e2008-04-17 10:38:59 -04002990 ext4_abort(sb, __func__, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002991
2992 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07002993 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002994
2995 es = sbi->s_es;
2996
Mingming Cao617ba132006-10-11 01:20:53 -07002997 ext4_init_journal_params(sb, sbi->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002998
2999 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003000 n_blocks_count > ext4_blocks_count(es)) {
Mingming Cao617ba132006-10-11 01:20:53 -07003001 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003002 err = -EROFS;
3003 goto restore_opts;
3004 }
3005
3006 if (*flags & MS_RDONLY) {
3007 /*
3008 * First of all, the unconditional stuff we have to do
3009 * to disable replay of the journal when we next remount
3010 */
3011 sb->s_flags |= MS_RDONLY;
3012
3013 /*
3014 * OK, test if we are remounting a valid rw partition
3015 * readonly, and if so set the rdonly flag and then
3016 * mark the partition as valid again.
3017 */
Mingming Cao617ba132006-10-11 01:20:53 -07003018 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3019 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003020 es->s_state = cpu_to_le16(sbi->s_mount_state);
3021
Jan Kara32c37732007-07-15 23:41:09 -07003022 /*
3023 * We have to unlock super so that we can wait for
3024 * transactions.
3025 */
3026 unlock_super(sb);
Mingming Cao617ba132006-10-11 01:20:53 -07003027 ext4_mark_recovery_complete(sb, es);
Jan Kara32c37732007-07-15 23:41:09 -07003028 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003029 } else {
3030 __le32 ret;
Mingming Cao617ba132006-10-11 01:20:53 -07003031 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3032 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3033 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003034 "remount RDWR because of unsupported "
3035 "optional features (%x).\n",
3036 sb->s_id, le32_to_cpu(ret));
3037 err = -EROFS;
3038 goto restore_opts;
3039 }
Eric Sandeenead65962007-02-10 01:46:08 -08003040
3041 /*
3042 * If we have an unprocessed orphan list hanging
3043 * around from a previously readonly bdev mount,
3044 * require a full umount/remount for now.
3045 */
3046 if (es->s_last_orphan) {
3047 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3048 "remount RDWR because of unprocessed "
3049 "orphan inode list. Please "
3050 "umount/remount instead.\n",
3051 sb->s_id);
3052 err = -EINVAL;
3053 goto restore_opts;
3054 }
3055
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003056 /*
3057 * Mounting a RDONLY partition read-write, so reread
3058 * and store the current valid flag. (It may have
3059 * been changed by e2fsck since we originally mounted
3060 * the partition.)
3061 */
Mingming Cao617ba132006-10-11 01:20:53 -07003062 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003063 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07003064 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003065 goto restore_opts;
Mingming Cao617ba132006-10-11 01:20:53 -07003066 if (!ext4_setup_super (sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003067 sb->s_flags &= ~MS_RDONLY;
3068 }
3069 }
3070#ifdef CONFIG_QUOTA
3071 /* Release old quota file names */
3072 for (i = 0; i < MAXQUOTAS; i++)
3073 if (old_opts.s_qf_names[i] &&
3074 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3075 kfree(old_opts.s_qf_names[i]);
3076#endif
3077 return 0;
3078restore_opts:
3079 sb->s_flags = old_sb_flags;
3080 sbi->s_mount_opt = old_opts.s_mount_opt;
3081 sbi->s_resuid = old_opts.s_resuid;
3082 sbi->s_resgid = old_opts.s_resgid;
3083 sbi->s_commit_interval = old_opts.s_commit_interval;
3084#ifdef CONFIG_QUOTA
3085 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3086 for (i = 0; i < MAXQUOTAS; i++) {
3087 if (sbi->s_qf_names[i] &&
3088 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3089 kfree(sbi->s_qf_names[i]);
3090 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3091 }
3092#endif
3093 return err;
3094}
3095
Mingming Cao617ba132006-10-11 01:20:53 -07003096static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003097{
3098 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003099 struct ext4_sb_info *sbi = EXT4_SB(sb);
3100 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08003101 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003102
Badari Pulavarty5e700302007-07-15 23:42:00 -07003103 if (test_opt(sb, MINIX_DF)) {
3104 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003105 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Avantika Mathurfd2d4292008-01-28 23:58:27 -05003106 ext4_group_t ngroups = sbi->s_groups_count, i;
Badari Pulavarty5e700302007-07-15 23:42:00 -07003107 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003108 smp_rmb();
3109
3110 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07003111 * Compute the overhead (FS structures). This is constant
3112 * for a given filesystem unless the number of block groups
3113 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003114 */
3115
3116 /*
3117 * All of the blocks before first_data_block are
3118 * overhead
3119 */
3120 overhead = le32_to_cpu(es->s_first_data_block);
3121
3122 /*
3123 * Add the overhead attributed to the superblock and
3124 * block group descriptors. If the sparse superblocks
3125 * feature is turned on, then not all groups have this.
3126 */
3127 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07003128 overhead += ext4_bg_has_super(sb, i) +
3129 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003130 cond_resched();
3131 }
3132
3133 /*
3134 * Every block group has an inode bitmap, a block
3135 * bitmap, and an inode table.
3136 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07003137 overhead += ngroups * (2 + sbi->s_itb_per_group);
3138 sbi->s_overhead_last = overhead;
3139 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003140 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003141 }
3142
Mingming Cao617ba132006-10-11 01:20:53 -07003143 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003144 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07003145 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07003146 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
Aneesh Kumar K.V308ba3e2007-10-16 18:38:25 -04003147 ext4_free_blocks_count_set(es, buf->f_bfree);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003148 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3149 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003150 buf->f_bavail = 0;
3151 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07003152 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Badari Pulavarty5e700302007-07-15 23:42:00 -07003153 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
Mingming Cao617ba132006-10-11 01:20:53 -07003154 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08003155 fsid = le64_to_cpup((void *)es->s_uuid) ^
3156 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3157 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3158 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003159 return 0;
3160}
3161
3162/* Helper function for writing quotas on sync - we need to start transaction before quota file
3163 * is locked for write. Otherwise the are possible deadlocks:
3164 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07003165 * ext4_create() quota_sync()
Mingming Caodab291a2006-10-11 01:21:01 -07003166 * jbd2_journal_start() write_dquot()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003167 * DQUOT_INIT() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07003168 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003169 *
3170 */
3171
3172#ifdef CONFIG_QUOTA
3173
3174static inline struct inode *dquot_to_inode(struct dquot *dquot)
3175{
3176 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3177}
3178
Mingming Cao617ba132006-10-11 01:20:53 -07003179static int ext4_dquot_initialize(struct inode *inode, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003180{
3181 handle_t *handle;
3182 int ret, err;
3183
3184 /* We may create quota structure so we need to reserve enough blocks */
Mingming Cao617ba132006-10-11 01:20:53 -07003185 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003186 if (IS_ERR(handle))
3187 return PTR_ERR(handle);
3188 ret = dquot_initialize(inode, type);
Mingming Cao617ba132006-10-11 01:20:53 -07003189 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003190 if (!ret)
3191 ret = err;
3192 return ret;
3193}
3194
Mingming Cao617ba132006-10-11 01:20:53 -07003195static int ext4_dquot_drop(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003196{
3197 handle_t *handle;
3198 int ret, err;
3199
3200 /* We may delete quota structure so we need to reserve enough blocks */
Mingming Cao617ba132006-10-11 01:20:53 -07003201 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
Jan Kara2887df12008-04-29 22:02:07 -04003202 if (IS_ERR(handle)) {
3203 /*
3204 * We call dquot_drop() anyway to at least release references
3205 * to quota structures so that umount does not hang.
3206 */
3207 dquot_drop(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003208 return PTR_ERR(handle);
Jan Kara2887df12008-04-29 22:02:07 -04003209 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003210 ret = dquot_drop(inode);
Mingming Cao617ba132006-10-11 01:20:53 -07003211 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003212 if (!ret)
3213 ret = err;
3214 return ret;
3215}
3216
Mingming Cao617ba132006-10-11 01:20:53 -07003217static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003218{
3219 int ret, err;
3220 handle_t *handle;
3221 struct inode *inode;
3222
3223 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003224 handle = ext4_journal_start(inode,
3225 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003226 if (IS_ERR(handle))
3227 return PTR_ERR(handle);
3228 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003229 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003230 if (!ret)
3231 ret = err;
3232 return ret;
3233}
3234
Mingming Cao617ba132006-10-11 01:20:53 -07003235static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003236{
3237 int ret, err;
3238 handle_t *handle;
3239
Mingming Cao617ba132006-10-11 01:20:53 -07003240 handle = ext4_journal_start(dquot_to_inode(dquot),
3241 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003242 if (IS_ERR(handle))
3243 return PTR_ERR(handle);
3244 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003245 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003246 if (!ret)
3247 ret = err;
3248 return ret;
3249}
3250
Mingming Cao617ba132006-10-11 01:20:53 -07003251static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003252{
3253 int ret, err;
3254 handle_t *handle;
3255
Mingming Cao617ba132006-10-11 01:20:53 -07003256 handle = ext4_journal_start(dquot_to_inode(dquot),
3257 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07003258 if (IS_ERR(handle)) {
3259 /* Release dquot anyway to avoid endless cycle in dqput() */
3260 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003261 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07003262 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003263 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003264 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003265 if (!ret)
3266 ret = err;
3267 return ret;
3268}
3269
Mingming Cao617ba132006-10-11 01:20:53 -07003270static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003271{
Jan Kara2c8be6b2008-05-13 21:27:55 -04003272 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07003273 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3274 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003275 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003276 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003277 } else {
3278 return dquot_mark_dquot_dirty(dquot);
3279 }
3280}
3281
Mingming Cao617ba132006-10-11 01:20:53 -07003282static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003283{
3284 int ret, err;
3285 handle_t *handle;
3286
3287 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07003288 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003289 if (IS_ERR(handle))
3290 return PTR_ERR(handle);
3291 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07003292 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003293 if (!ret)
3294 ret = err;
3295 return ret;
3296}
3297
3298/*
3299 * Turn on quotas during mount time - we need to find
3300 * the quota file and such...
3301 */
Mingming Cao617ba132006-10-11 01:20:53 -07003302static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003303{
Mingming Cao617ba132006-10-11 01:20:53 -07003304 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3305 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003306}
3307
3308/*
3309 * Standard function to be called on quota_on
3310 */
Mingming Cao617ba132006-10-11 01:20:53 -07003311static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Kara6f28e082008-04-28 02:14:34 -07003312 char *path, int remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003313{
3314 int err;
3315 struct nameidata nd;
3316
3317 if (!test_opt(sb, QUOTA))
3318 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04003319 /* When remounting, no checks are needed and in fact, path is NULL */
3320 if (remount)
Jan Kara6f28e082008-04-28 02:14:34 -07003321 return vfs_quota_on(sb, type, format_id, path, remount);
Jan Kara06235432008-05-13 19:11:51 -04003322
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003323 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3324 if (err)
3325 return err;
Jan Kara06235432008-05-13 19:11:51 -04003326
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003327 /* Quotafile not on the same filesystem? */
Jan Blunck4ac91372008-02-14 19:34:32 -08003328 if (nd.path.mnt->mnt_sb != sb) {
Jan Blunck1d957f92008-02-14 19:34:35 -08003329 path_put(&nd.path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003330 return -EXDEV;
3331 }
Jan Kara06235432008-05-13 19:11:51 -04003332 /* Journaling quota? */
3333 if (EXT4_SB(sb)->s_qf_names[type]) {
3334 /* Quotafile not of fs root? */
3335 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3336 printk(KERN_WARNING
3337 "EXT4-fs: Quota file not on filesystem root. "
3338 "Journaled quota will not work.\n");
3339 }
3340
3341 /*
3342 * When we journal data on quota file, we have to flush journal to see
3343 * all updates to the file when we bypass pagecache...
3344 */
3345 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3346 /*
3347 * We don't need to lock updates but journal_flush() could
3348 * otherwise be livelocked...
3349 */
3350 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3351 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3352 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3353 }
3354
Al Viro77e69da2008-08-01 04:29:18 -04003355 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
Jan Blunck1d957f92008-02-14 19:34:35 -08003356 path_put(&nd.path);
Al Viro77e69da2008-08-01 04:29:18 -04003357 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003358}
3359
3360/* Read data from quotafile - avoid pagecache and such because we cannot afford
3361 * acquiring the locks... As quota files are never truncated and quota code
3362 * itself serializes the operations (and noone else should touch the files)
3363 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07003364static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003365 size_t len, loff_t off)
3366{
3367 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003368 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003369 int err = 0;
3370 int offset = off & (sb->s_blocksize - 1);
3371 int tocopy;
3372 size_t toread;
3373 struct buffer_head *bh;
3374 loff_t i_size = i_size_read(inode);
3375
3376 if (off > i_size)
3377 return 0;
3378 if (off+len > i_size)
3379 len = i_size-off;
3380 toread = len;
3381 while (toread > 0) {
3382 tocopy = sb->s_blocksize - offset < toread ?
3383 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07003384 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003385 if (err)
3386 return err;
3387 if (!bh) /* A hole? */
3388 memset(data, 0, tocopy);
3389 else
3390 memcpy(data, bh->b_data+offset, tocopy);
3391 brelse(bh);
3392 offset = 0;
3393 toread -= tocopy;
3394 data += tocopy;
3395 blk++;
3396 }
3397 return len;
3398}
3399
3400/* Write to quotafile (we know the transaction is already started and has
3401 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07003402static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003403 const char *data, size_t len, loff_t off)
3404{
3405 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003406 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003407 int err = 0;
3408 int offset = off & (sb->s_blocksize - 1);
3409 int tocopy;
Mingming Cao617ba132006-10-11 01:20:53 -07003410 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003411 size_t towrite = len;
3412 struct buffer_head *bh;
3413 handle_t *handle = journal_current_handle();
3414
Jan Kara9c3013e2007-09-11 15:23:29 -07003415 if (!handle) {
3416 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3417 " cancelled because transaction is not started.\n",
3418 (unsigned long long)off, (unsigned long long)len);
3419 return -EIO;
3420 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003421 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3422 while (towrite > 0) {
3423 tocopy = sb->s_blocksize - offset < towrite ?
3424 sb->s_blocksize - offset : towrite;
Mingming Cao617ba132006-10-11 01:20:53 -07003425 bh = ext4_bread(handle, inode, blk, 1, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003426 if (!bh)
3427 goto out;
3428 if (journal_quota) {
Mingming Cao617ba132006-10-11 01:20:53 -07003429 err = ext4_journal_get_write_access(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003430 if (err) {
3431 brelse(bh);
3432 goto out;
3433 }
3434 }
3435 lock_buffer(bh);
3436 memcpy(bh->b_data+offset, data, tocopy);
3437 flush_dcache_page(bh->b_page);
3438 unlock_buffer(bh);
3439 if (journal_quota)
Mingming Cao617ba132006-10-11 01:20:53 -07003440 err = ext4_journal_dirty_metadata(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003441 else {
3442 /* Always do at least ordered writes for quotas */
Jan Kara678aaf42008-07-11 19:27:31 -04003443 err = ext4_jbd2_file_inode(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003444 mark_buffer_dirty(bh);
3445 }
3446 brelse(bh);
3447 if (err)
3448 goto out;
3449 offset = 0;
3450 towrite -= tocopy;
3451 data += tocopy;
3452 blk++;
3453 }
3454out:
Jan Kara4d04e4f2008-07-04 09:59:34 -07003455 if (len == towrite) {
3456 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003457 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07003458 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003459 if (inode->i_size < off+len-towrite) {
3460 i_size_write(inode, off+len-towrite);
Mingming Cao617ba132006-10-11 01:20:53 -07003461 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003462 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003463 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07003464 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003465 mutex_unlock(&inode->i_mutex);
3466 return len - towrite;
3467}
3468
3469#endif
3470
Mingming Cao617ba132006-10-11 01:20:53 -07003471static int ext4_get_sb(struct file_system_type *fs_type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003472 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3473{
Mingming Cao617ba132006-10-11 01:20:53 -07003474 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003475}
3476
Mingming Cao617ba132006-10-11 01:20:53 -07003477static struct file_system_type ext4dev_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003478 .owner = THIS_MODULE,
Mingming Cao617ba132006-10-11 01:20:53 -07003479 .name = "ext4dev",
3480 .get_sb = ext4_get_sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003481 .kill_sb = kill_block_super,
3482 .fs_flags = FS_REQUIRES_DEV,
3483};
3484
Mingming Cao617ba132006-10-11 01:20:53 -07003485static int __init init_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003486{
Alex Tomasc9de5602008-01-29 00:19:52 -05003487 int err;
3488
3489 err = init_ext4_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003490 if (err)
3491 return err;
Alex Tomasc9de5602008-01-29 00:19:52 -05003492
3493 err = init_ext4_xattr();
3494 if (err)
3495 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003496 err = init_inodecache();
3497 if (err)
3498 goto out1;
Andrew Morton63f57932006-10-11 01:21:24 -07003499 err = register_filesystem(&ext4dev_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003500 if (err)
3501 goto out;
3502 return 0;
3503out:
3504 destroy_inodecache();
3505out1:
Mingming Cao617ba132006-10-11 01:20:53 -07003506 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05003507out2:
3508 exit_ext4_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003509 return err;
3510}
3511
Mingming Cao617ba132006-10-11 01:20:53 -07003512static void __exit exit_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003513{
Mingming Cao617ba132006-10-11 01:20:53 -07003514 unregister_filesystem(&ext4dev_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003515 destroy_inodecache();
Mingming Cao617ba132006-10-11 01:20:53 -07003516 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05003517 exit_ext4_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003518}
3519
3520MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Mingming Cao617ba132006-10-11 01:20:53 -07003521MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003522MODULE_LICENSE("GPL");
Mingming Cao617ba132006-10-11 01:20:53 -07003523module_init(init_ext4_fs)
3524module_exit(exit_ext4_fs)