blob: 1fa3bdc3c6711f1c73e2930141686c0fe2149365 [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>
Mingming Cao617ba132006-10-11 01:20:53 -070024#include <linux/ext4_fs.h>
Mingming Caodab291a2006-10-11 01:21:01 -070025#include <linux/ext4_jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070026#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
38
39#include <asm/uaccess.h>
40
41#include "xattr.h"
42#include "acl.h"
43#include "namei.h"
44
Mingming Cao617ba132006-10-11 01:20:53 -070045static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070046 unsigned long journal_devnum);
Mingming Cao617ba132006-10-11 01:20:53 -070047static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070048 unsigned int);
Mingming Cao617ba132006-10-11 01:20:53 -070049static void ext4_commit_super (struct super_block * sb,
50 struct ext4_super_block * es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070051 int sync);
Mingming Cao617ba132006-10-11 01:20:53 -070052static void ext4_mark_recovery_complete(struct super_block * sb,
53 struct ext4_super_block * es);
54static void ext4_clear_journal_err(struct super_block * sb,
55 struct ext4_super_block * es);
56static int ext4_sync_fs(struct super_block *sb, int wait);
57static const char *ext4_decode_error(struct super_block * sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070058 char nbuf[16]);
Mingming Cao617ba132006-10-11 01:20:53 -070059static int ext4_remount (struct super_block * sb, int * flags, char * data);
60static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
61static void ext4_unlockfs(struct super_block *sb);
62static void ext4_write_super (struct super_block * sb);
63static void ext4_write_super_lockfs(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070064
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070065
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070066ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
67 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070068{
69 return le32_to_cpu(bg->bg_block_bitmap) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070070 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
71 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070072}
73
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070074ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
75 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070076{
77 return le32_to_cpu(bg->bg_inode_bitmap) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070078 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
79 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070080}
81
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070082ext4_fsblk_t ext4_inode_table(struct super_block *sb,
83 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070084{
85 return le32_to_cpu(bg->bg_inode_table) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070086 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
87 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070088}
89
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070090void ext4_block_bitmap_set(struct super_block *sb,
91 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070092{
93 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070094 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
95 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070096}
97
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070098void ext4_inode_bitmap_set(struct super_block *sb,
99 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700100{
101 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700102 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
103 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700104}
105
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700106void ext4_inode_table_set(struct super_block *sb,
107 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700108{
109 bg->bg_inode_table = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700110 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
111 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700112}
113
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700114/*
Mingming Caodab291a2006-10-11 01:21:01 -0700115 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700116 *
117 * The only special thing we need to do here is to make sure that all
118 * journal_end calls result in the superblock being marked dirty, so
119 * that sync() will call the filesystem's write_super callback if
120 * appropriate.
121 */
Mingming Cao617ba132006-10-11 01:20:53 -0700122handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700123{
124 journal_t *journal;
125
126 if (sb->s_flags & MS_RDONLY)
127 return ERR_PTR(-EROFS);
128
129 /* Special case here: if the journal has aborted behind our
130 * backs (eg. EIO in the commit thread), then we still need to
131 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700132 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700133 if (is_journal_aborted(journal)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700134 ext4_abort(sb, __FUNCTION__,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700135 "Detected aborted journal");
136 return ERR_PTR(-EROFS);
137 }
138
Mingming Caodab291a2006-10-11 01:21:01 -0700139 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700140}
141
142/*
143 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700144 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700145 * that sync() will call the filesystem's write_super callback if
146 * appropriate.
147 */
Mingming Cao617ba132006-10-11 01:20:53 -0700148int __ext4_journal_stop(const char *where, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700149{
150 struct super_block *sb;
151 int err;
152 int rc;
153
154 sb = handle->h_transaction->t_journal->j_private;
155 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700156 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700157
158 if (!err)
159 err = rc;
160 if (err)
Mingming Cao617ba132006-10-11 01:20:53 -0700161 __ext4_std_error(sb, where, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700162 return err;
163}
164
Mingming Cao617ba132006-10-11 01:20:53 -0700165void ext4_journal_abort_handle(const char *caller, const char *err_fn,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700166 struct buffer_head *bh, handle_t *handle, int err)
167{
168 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700169 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700170
171 if (bh)
172 BUFFER_TRACE(bh, "abort");
173
174 if (!handle->h_err)
175 handle->h_err = err;
176
177 if (is_handle_aborted(handle))
178 return;
179
180 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
181 caller, errstr, err_fn);
182
Mingming Caodab291a2006-10-11 01:21:01 -0700183 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700184}
185
186/* Deal with the reporting of failure conditions on a filesystem such as
187 * inconsistencies detected or read IO failures.
188 *
189 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700190 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700191 * write ordering constraints on the superblock which prevent us from
192 * writing it out straight away; and given that the journal is about to
193 * be aborted, we can't rely on the current, or future, transactions to
194 * write out the superblock safely.
195 *
Mingming Caodab291a2006-10-11 01:21:01 -0700196 * We'll just use the jbd2_journal_abort() error code to record an error in
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700197 * the journal instead. On recovery, the journal will compain about
198 * that error until we've noted it down and cleared it.
199 */
200
Mingming Cao617ba132006-10-11 01:20:53 -0700201static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700202{
Mingming Cao617ba132006-10-11 01:20:53 -0700203 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700204
Mingming Cao617ba132006-10-11 01:20:53 -0700205 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
206 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700207
208 if (sb->s_flags & MS_RDONLY)
209 return;
210
211 if (!test_opt (sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700212 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700213
Mingming Cao617ba132006-10-11 01:20:53 -0700214 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700215 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700216 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700217 }
218 if (test_opt (sb, ERRORS_RO)) {
219 printk (KERN_CRIT "Remounting filesystem read-only\n");
220 sb->s_flags |= MS_RDONLY;
221 }
Mingming Cao617ba132006-10-11 01:20:53 -0700222 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700223 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700224 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700225 sb->s_id);
226}
227
Mingming Cao617ba132006-10-11 01:20:53 -0700228void ext4_error (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700229 const char * fmt, ...)
230{
231 va_list args;
232
233 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700234 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700235 vprintk(fmt, args);
236 printk("\n");
237 va_end(args);
238
Mingming Cao617ba132006-10-11 01:20:53 -0700239 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700240}
241
Mingming Cao617ba132006-10-11 01:20:53 -0700242static const char *ext4_decode_error(struct super_block * sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700243 char nbuf[16])
244{
245 char *errstr = NULL;
246
247 switch (errno) {
248 case -EIO:
249 errstr = "IO failure";
250 break;
251 case -ENOMEM:
252 errstr = "Out of memory";
253 break;
254 case -EROFS:
Mingming Caodab291a2006-10-11 01:21:01 -0700255 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700256 errstr = "Journal has aborted";
257 else
258 errstr = "Readonly filesystem";
259 break;
260 default:
261 /* If the caller passed in an extra buffer for unknown
262 * errors, textualise them now. Else we just return
263 * NULL. */
264 if (nbuf) {
265 /* Check for truncated error codes... */
266 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
267 errstr = nbuf;
268 }
269 break;
270 }
271
272 return errstr;
273}
274
Mingming Cao617ba132006-10-11 01:20:53 -0700275/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700276 * automatically and invokes the appropriate error response. */
277
Mingming Cao617ba132006-10-11 01:20:53 -0700278void __ext4_std_error (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700279 int errno)
280{
281 char nbuf[16];
282 const char *errstr;
283
284 /* Special case: if the error is EROFS, and we're not already
285 * inside a transaction, then there's really no point in logging
286 * an error. */
287 if (errno == -EROFS && journal_current_handle() == NULL &&
288 (sb->s_flags & MS_RDONLY))
289 return;
290
Mingming Cao617ba132006-10-11 01:20:53 -0700291 errstr = ext4_decode_error(sb, errno, nbuf);
292 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700293 sb->s_id, function, errstr);
294
Mingming Cao617ba132006-10-11 01:20:53 -0700295 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700296}
297
298/*
Mingming Cao617ba132006-10-11 01:20:53 -0700299 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700300 * abort function may be used to deal with unrecoverable failures such
301 * as journal IO errors or ENOMEM at a critical moment in log management.
302 *
303 * We unconditionally force the filesystem into an ABORT|READONLY state,
304 * unless the error response on the fs has been set to panic in which
305 * case we take the easy way out and panic immediately.
306 */
307
Mingming Cao617ba132006-10-11 01:20:53 -0700308void ext4_abort (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700309 const char * fmt, ...)
310{
311 va_list args;
312
Mingming Cao617ba132006-10-11 01:20:53 -0700313 printk (KERN_CRIT "ext4_abort called.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700314
315 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700316 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700317 vprintk(fmt, args);
318 printk("\n");
319 va_end(args);
320
321 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700322 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700323
324 if (sb->s_flags & MS_RDONLY)
325 return;
326
327 printk(KERN_CRIT "Remounting filesystem read-only\n");
Mingming Cao617ba132006-10-11 01:20:53 -0700328 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700329 sb->s_flags |= MS_RDONLY;
Mingming Cao617ba132006-10-11 01:20:53 -0700330 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
Mingming Caodab291a2006-10-11 01:21:01 -0700331 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700332}
333
Mingming Cao617ba132006-10-11 01:20:53 -0700334void ext4_warning (struct super_block * sb, const char * function,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700335 const char * fmt, ...)
336{
337 va_list args;
338
339 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700340 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700341 sb->s_id, function);
342 vprintk(fmt, args);
343 printk("\n");
344 va_end(args);
345}
346
Mingming Cao617ba132006-10-11 01:20:53 -0700347void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700348{
Mingming Cao617ba132006-10-11 01:20:53 -0700349 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700350
Mingming Cao617ba132006-10-11 01:20:53 -0700351 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700352 return;
353
Mingming Cao617ba132006-10-11 01:20:53 -0700354 ext4_warning(sb, __FUNCTION__,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700355 "updating to rev %d because of new feature flag, "
356 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700357 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700358
Mingming Cao617ba132006-10-11 01:20:53 -0700359 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
360 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
361 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700362 /* leave es->s_feature_*compat flags alone */
363 /* es->s_uuid will be set by e2fsck if empty */
364
365 /*
366 * The rest of the superblock fields should be zero, and if not it
367 * means they are likely already in use, so leave them alone. We
368 * can leave it up to e2fsck to clean up any inconsistencies there.
369 */
370}
371
372/*
373 * Open the external journal device
374 */
Mingming Cao617ba132006-10-11 01:20:53 -0700375static struct block_device *ext4_blkdev_get(dev_t dev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700376{
377 struct block_device *bdev;
378 char b[BDEVNAME_SIZE];
379
380 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
381 if (IS_ERR(bdev))
382 goto fail;
383 return bdev;
384
385fail:
Mingming Cao617ba132006-10-11 01:20:53 -0700386 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700387 __bdevname(dev, b), PTR_ERR(bdev));
388 return NULL;
389}
390
391/*
392 * Release the journal device
393 */
Mingming Cao617ba132006-10-11 01:20:53 -0700394static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700395{
396 bd_release(bdev);
397 return blkdev_put(bdev);
398}
399
Mingming Cao617ba132006-10-11 01:20:53 -0700400static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700401{
402 struct block_device *bdev;
403 int ret = -ENODEV;
404
405 bdev = sbi->journal_bdev;
406 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700407 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700408 sbi->journal_bdev = NULL;
409 }
410 return ret;
411}
412
413static inline struct inode *orphan_list_entry(struct list_head *l)
414{
Mingming Cao617ba132006-10-11 01:20:53 -0700415 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700416}
417
Mingming Cao617ba132006-10-11 01:20:53 -0700418static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700419{
420 struct list_head *l;
421
422 printk(KERN_ERR "sb orphan head is %d\n",
423 le32_to_cpu(sbi->s_es->s_last_orphan));
424
425 printk(KERN_ERR "sb_info orphan list:\n");
426 list_for_each(l, &sbi->s_orphan) {
427 struct inode *inode = orphan_list_entry(l);
428 printk(KERN_ERR " "
429 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
430 inode->i_sb->s_id, inode->i_ino, inode,
431 inode->i_mode, inode->i_nlink,
432 NEXT_ORPHAN(inode));
433 }
434}
435
Mingming Cao617ba132006-10-11 01:20:53 -0700436static void ext4_put_super (struct super_block * sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700437{
Mingming Cao617ba132006-10-11 01:20:53 -0700438 struct ext4_sb_info *sbi = EXT4_SB(sb);
439 struct ext4_super_block *es = sbi->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700440 int i;
441
Alex Tomasa86c6182006-10-11 01:21:03 -0700442 ext4_ext_release(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700443 ext4_xattr_put_super(sb);
Mingming Caodab291a2006-10-11 01:21:01 -0700444 jbd2_journal_destroy(sbi->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700445 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700446 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700447 es->s_state = cpu_to_le16(sbi->s_mount_state);
448 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
449 mark_buffer_dirty(sbi->s_sbh);
Mingming Cao617ba132006-10-11 01:20:53 -0700450 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700451 }
452
453 for (i = 0; i < sbi->s_gdb_count; i++)
454 brelse(sbi->s_group_desc[i]);
455 kfree(sbi->s_group_desc);
456 percpu_counter_destroy(&sbi->s_freeblocks_counter);
457 percpu_counter_destroy(&sbi->s_freeinodes_counter);
458 percpu_counter_destroy(&sbi->s_dirs_counter);
459 brelse(sbi->s_sbh);
460#ifdef CONFIG_QUOTA
461 for (i = 0; i < MAXQUOTAS; i++)
462 kfree(sbi->s_qf_names[i]);
463#endif
464
465 /* Debugging code just in case the in-memory inode orphan list
466 * isn't empty. The on-disk one can be non-empty if we've
467 * detected an error and taken the fs readonly, but the
468 * in-memory list had better be clean by this point. */
469 if (!list_empty(&sbi->s_orphan))
470 dump_orphan_list(sb, sbi);
471 J_ASSERT(list_empty(&sbi->s_orphan));
472
473 invalidate_bdev(sb->s_bdev, 0);
474 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
475 /*
476 * Invalidate the journal device's buffers. We don't want them
477 * floating about in memory - the physical journal device may
478 * hotswapped, and it breaks the `ro-after' testing code.
479 */
480 sync_blockdev(sbi->journal_bdev);
481 invalidate_bdev(sbi->journal_bdev, 0);
Mingming Cao617ba132006-10-11 01:20:53 -0700482 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700483 }
484 sb->s_fs_info = NULL;
485 kfree(sbi);
486 return;
487}
488
Mingming Cao617ba132006-10-11 01:20:53 -0700489static kmem_cache_t *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700490
491/*
492 * Called inside transaction, so use GFP_NOFS
493 */
Mingming Cao617ba132006-10-11 01:20:53 -0700494static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700495{
Mingming Cao617ba132006-10-11 01:20:53 -0700496 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700497
Mingming Cao617ba132006-10-11 01:20:53 -0700498 ei = kmem_cache_alloc(ext4_inode_cachep, SLAB_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700499 if (!ei)
500 return NULL;
Mingming Cao617ba132006-10-11 01:20:53 -0700501#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
502 ei->i_acl = EXT4_ACL_NOT_CACHED;
503 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700504#endif
505 ei->i_block_alloc_info = NULL;
506 ei->vfs_inode.i_version = 1;
Alex Tomasa86c6182006-10-11 01:21:03 -0700507 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700508 return &ei->vfs_inode;
509}
510
Mingming Cao617ba132006-10-11 01:20:53 -0700511static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700512{
Mingming Cao617ba132006-10-11 01:20:53 -0700513 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700514}
515
516static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
517{
Mingming Cao617ba132006-10-11 01:20:53 -0700518 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700519
520 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
521 SLAB_CTOR_CONSTRUCTOR) {
522 INIT_LIST_HEAD(&ei->i_orphan);
Mingming Cao617ba132006-10-11 01:20:53 -0700523#ifdef CONFIG_EXT4DEV_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700524 init_rwsem(&ei->xattr_sem);
525#endif
526 mutex_init(&ei->truncate_mutex);
527 inode_init_once(&ei->vfs_inode);
528 }
529}
530
531static int init_inodecache(void)
532{
Mingming Cao617ba132006-10-11 01:20:53 -0700533 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
534 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700535 0, (SLAB_RECLAIM_ACCOUNT|
536 SLAB_MEM_SPREAD),
537 init_once, NULL);
Mingming Cao617ba132006-10-11 01:20:53 -0700538 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700539 return -ENOMEM;
540 return 0;
541}
542
543static void destroy_inodecache(void)
544{
Mingming Cao617ba132006-10-11 01:20:53 -0700545 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700546}
547
Mingming Cao617ba132006-10-11 01:20:53 -0700548static void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700549{
Mingming Cao617ba132006-10-11 01:20:53 -0700550 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
551#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
552 if (EXT4_I(inode)->i_acl &&
553 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
554 posix_acl_release(EXT4_I(inode)->i_acl);
555 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700556 }
Mingming Cao617ba132006-10-11 01:20:53 -0700557 if (EXT4_I(inode)->i_default_acl &&
558 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
559 posix_acl_release(EXT4_I(inode)->i_default_acl);
560 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700561 }
562#endif
Mingming Cao617ba132006-10-11 01:20:53 -0700563 ext4_discard_reservation(inode);
564 EXT4_I(inode)->i_block_alloc_info = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700565 if (unlikely(rsv))
566 kfree(rsv);
567}
568
Mingming Cao617ba132006-10-11 01:20:53 -0700569static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700570{
571#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700572 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700573
574 if (sbi->s_jquota_fmt)
575 seq_printf(seq, ",jqfmt=%s",
576 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
577
578 if (sbi->s_qf_names[USRQUOTA])
579 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
580
581 if (sbi->s_qf_names[GRPQUOTA])
582 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
583
Mingming Cao617ba132006-10-11 01:20:53 -0700584 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700585 seq_puts(seq, ",usrquota");
586
Mingming Cao617ba132006-10-11 01:20:53 -0700587 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700588 seq_puts(seq, ",grpquota");
589#endif
590}
591
Mingming Cao617ba132006-10-11 01:20:53 -0700592static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700593{
594 struct super_block *sb = vfs->mnt_sb;
595
Mingming Cao617ba132006-10-11 01:20:53 -0700596 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700597 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -0700598 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700599 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -0700600 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601 seq_puts(seq, ",data=writeback");
602
Mingming Cao617ba132006-10-11 01:20:53 -0700603 ext4_show_quota_options(seq, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700604
605 return 0;
606}
607
608
Mingming Cao617ba132006-10-11 01:20:53 -0700609static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700610{
611 __u32 *objp = vobjp;
612 unsigned long ino = objp[0];
613 __u32 generation = objp[1];
614 struct inode *inode;
615 struct dentry *result;
616
Mingming Cao617ba132006-10-11 01:20:53 -0700617 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700618 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700619 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700620 return ERR_PTR(-ESTALE);
621
622 /* iget isn't really right if the inode is currently unallocated!!
623 *
Mingming Cao617ba132006-10-11 01:20:53 -0700624 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700625 * deleted, so we should be safe.
626 *
627 * Currently we don't know the generation for parent directory, so
628 * a generation of 0 means "accept any"
629 */
630 inode = iget(sb, ino);
631 if (inode == NULL)
632 return ERR_PTR(-ENOMEM);
633 if (is_bad_inode(inode) ||
634 (generation && inode->i_generation != generation)) {
635 iput(inode);
636 return ERR_PTR(-ESTALE);
637 }
638 /* now to find a dentry.
639 * If possible, get a well-connected one
640 */
641 result = d_alloc_anon(inode);
642 if (!result) {
643 iput(inode);
644 return ERR_PTR(-ENOMEM);
645 }
646 return result;
647}
648
649#ifdef CONFIG_QUOTA
650#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
651#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
652
Mingming Cao617ba132006-10-11 01:20:53 -0700653static int ext4_dquot_initialize(struct inode *inode, int type);
654static int ext4_dquot_drop(struct inode *inode);
655static int ext4_write_dquot(struct dquot *dquot);
656static int ext4_acquire_dquot(struct dquot *dquot);
657static int ext4_release_dquot(struct dquot *dquot);
658static int ext4_mark_dquot_dirty(struct dquot *dquot);
659static int ext4_write_info(struct super_block *sb, int type);
660static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
661static int ext4_quota_on_mount(struct super_block *sb, int type);
662static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700663 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -0700664static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700665 const char *data, size_t len, loff_t off);
666
Mingming Cao617ba132006-10-11 01:20:53 -0700667static struct dquot_operations ext4_quota_operations = {
668 .initialize = ext4_dquot_initialize,
669 .drop = ext4_dquot_drop,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700670 .alloc_space = dquot_alloc_space,
671 .alloc_inode = dquot_alloc_inode,
672 .free_space = dquot_free_space,
673 .free_inode = dquot_free_inode,
674 .transfer = dquot_transfer,
Mingming Cao617ba132006-10-11 01:20:53 -0700675 .write_dquot = ext4_write_dquot,
676 .acquire_dquot = ext4_acquire_dquot,
677 .release_dquot = ext4_release_dquot,
678 .mark_dirty = ext4_mark_dquot_dirty,
679 .write_info = ext4_write_info
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700680};
681
Mingming Cao617ba132006-10-11 01:20:53 -0700682static struct quotactl_ops ext4_qctl_operations = {
683 .quota_on = ext4_quota_on,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700684 .quota_off = vfs_quota_off,
685 .quota_sync = vfs_quota_sync,
686 .get_info = vfs_get_dqinfo,
687 .set_info = vfs_set_dqinfo,
688 .get_dqblk = vfs_get_dqblk,
689 .set_dqblk = vfs_set_dqblk
690};
691#endif
692
Mingming Cao617ba132006-10-11 01:20:53 -0700693static struct super_operations ext4_sops = {
694 .alloc_inode = ext4_alloc_inode,
695 .destroy_inode = ext4_destroy_inode,
696 .read_inode = ext4_read_inode,
697 .write_inode = ext4_write_inode,
698 .dirty_inode = ext4_dirty_inode,
699 .delete_inode = ext4_delete_inode,
700 .put_super = ext4_put_super,
701 .write_super = ext4_write_super,
702 .sync_fs = ext4_sync_fs,
703 .write_super_lockfs = ext4_write_super_lockfs,
704 .unlockfs = ext4_unlockfs,
705 .statfs = ext4_statfs,
706 .remount_fs = ext4_remount,
707 .clear_inode = ext4_clear_inode,
708 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700709#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -0700710 .quota_read = ext4_quota_read,
711 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700712#endif
713};
714
Mingming Cao617ba132006-10-11 01:20:53 -0700715static struct export_operations ext4_export_ops = {
716 .get_parent = ext4_get_parent,
717 .get_dentry = ext4_get_dentry,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700718};
719
720enum {
721 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
722 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
723 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
724 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
725 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
726 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
727 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
728 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
729 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
730 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
Alex Tomasa86c6182006-10-11 01:21:03 -0700731 Opt_grpquota, Opt_extents,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700732};
733
734static match_table_t tokens = {
735 {Opt_bsd_df, "bsddf"},
736 {Opt_minix_df, "minixdf"},
737 {Opt_grpid, "grpid"},
738 {Opt_grpid, "bsdgroups"},
739 {Opt_nogrpid, "nogrpid"},
740 {Opt_nogrpid, "sysvgroups"},
741 {Opt_resgid, "resgid=%u"},
742 {Opt_resuid, "resuid=%u"},
743 {Opt_sb, "sb=%u"},
744 {Opt_err_cont, "errors=continue"},
745 {Opt_err_panic, "errors=panic"},
746 {Opt_err_ro, "errors=remount-ro"},
747 {Opt_nouid32, "nouid32"},
748 {Opt_nocheck, "nocheck"},
749 {Opt_nocheck, "check=none"},
750 {Opt_debug, "debug"},
751 {Opt_oldalloc, "oldalloc"},
752 {Opt_orlov, "orlov"},
753 {Opt_user_xattr, "user_xattr"},
754 {Opt_nouser_xattr, "nouser_xattr"},
755 {Opt_acl, "acl"},
756 {Opt_noacl, "noacl"},
757 {Opt_reservation, "reservation"},
758 {Opt_noreservation, "noreservation"},
759 {Opt_noload, "noload"},
760 {Opt_nobh, "nobh"},
761 {Opt_bh, "bh"},
762 {Opt_commit, "commit=%u"},
763 {Opt_journal_update, "journal=update"},
764 {Opt_journal_inum, "journal=%u"},
765 {Opt_journal_dev, "journal_dev=%u"},
766 {Opt_abort, "abort"},
767 {Opt_data_journal, "data=journal"},
768 {Opt_data_ordered, "data=ordered"},
769 {Opt_data_writeback, "data=writeback"},
770 {Opt_offusrjquota, "usrjquota="},
771 {Opt_usrjquota, "usrjquota=%s"},
772 {Opt_offgrpjquota, "grpjquota="},
773 {Opt_grpjquota, "grpjquota=%s"},
774 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
775 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
776 {Opt_grpquota, "grpquota"},
777 {Opt_noquota, "noquota"},
778 {Opt_quota, "quota"},
779 {Opt_usrquota, "usrquota"},
780 {Opt_barrier, "barrier=%u"},
Alex Tomasa86c6182006-10-11 01:21:03 -0700781 {Opt_extents, "extents"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700782 {Opt_err, NULL},
783 {Opt_resize, "resize"},
784};
785
Mingming Cao617ba132006-10-11 01:20:53 -0700786static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700787{
Mingming Cao617ba132006-10-11 01:20:53 -0700788 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700789 char *options = (char *) *data;
790
791 if (!options || strncmp(options, "sb=", 3) != 0)
792 return 1; /* Default location */
793 options += 3;
Mingming Cao617ba132006-10-11 01:20:53 -0700794 /*todo: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700795 sb_block = simple_strtoul(options, &options, 0);
796 if (*options && *options != ',') {
Mingming Cao617ba132006-10-11 01:20:53 -0700797 printk("EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700798 (char *) *data);
799 return 1;
800 }
801 if (*options == ',')
802 options++;
803 *data = (void *) options;
804 return sb_block;
805}
806
807static int parse_options (char *options, struct super_block *sb,
808 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao617ba132006-10-11 01:20:53 -0700809 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700810{
Mingming Cao617ba132006-10-11 01:20:53 -0700811 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700812 char * p;
813 substring_t args[MAX_OPT_ARGS];
814 int data_opt = 0;
815 int option;
816#ifdef CONFIG_QUOTA
817 int qtype;
818 char *qname;
819#endif
820
821 if (!options)
822 return 1;
823
824 while ((p = strsep (&options, ",")) != NULL) {
825 int token;
826 if (!*p)
827 continue;
828
829 token = match_token(p, tokens, args);
830 switch (token) {
831 case Opt_bsd_df:
832 clear_opt (sbi->s_mount_opt, MINIX_DF);
833 break;
834 case Opt_minix_df:
835 set_opt (sbi->s_mount_opt, MINIX_DF);
836 break;
837 case Opt_grpid:
838 set_opt (sbi->s_mount_opt, GRPID);
839 break;
840 case Opt_nogrpid:
841 clear_opt (sbi->s_mount_opt, GRPID);
842 break;
843 case Opt_resuid:
844 if (match_int(&args[0], &option))
845 return 0;
846 sbi->s_resuid = option;
847 break;
848 case Opt_resgid:
849 if (match_int(&args[0], &option))
850 return 0;
851 sbi->s_resgid = option;
852 break;
853 case Opt_sb:
854 /* handled by get_sb_block() instead of here */
855 /* *sb_block = match_int(&args[0]); */
856 break;
857 case Opt_err_panic:
858 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
859 clear_opt (sbi->s_mount_opt, ERRORS_RO);
860 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
861 break;
862 case Opt_err_ro:
863 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
864 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
865 set_opt (sbi->s_mount_opt, ERRORS_RO);
866 break;
867 case Opt_err_cont:
868 clear_opt (sbi->s_mount_opt, ERRORS_RO);
869 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
870 set_opt (sbi->s_mount_opt, ERRORS_CONT);
871 break;
872 case Opt_nouid32:
873 set_opt (sbi->s_mount_opt, NO_UID32);
874 break;
875 case Opt_nocheck:
876 clear_opt (sbi->s_mount_opt, CHECK);
877 break;
878 case Opt_debug:
879 set_opt (sbi->s_mount_opt, DEBUG);
880 break;
881 case Opt_oldalloc:
882 set_opt (sbi->s_mount_opt, OLDALLOC);
883 break;
884 case Opt_orlov:
885 clear_opt (sbi->s_mount_opt, OLDALLOC);
886 break;
Mingming Cao617ba132006-10-11 01:20:53 -0700887#ifdef CONFIG_EXT4DEV_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700888 case Opt_user_xattr:
889 set_opt (sbi->s_mount_opt, XATTR_USER);
890 break;
891 case Opt_nouser_xattr:
892 clear_opt (sbi->s_mount_opt, XATTR_USER);
893 break;
894#else
895 case Opt_user_xattr:
896 case Opt_nouser_xattr:
Mingming Cao617ba132006-10-11 01:20:53 -0700897 printk("EXT4 (no)user_xattr options not supported\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700898 break;
899#endif
Mingming Cao617ba132006-10-11 01:20:53 -0700900#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700901 case Opt_acl:
902 set_opt(sbi->s_mount_opt, POSIX_ACL);
903 break;
904 case Opt_noacl:
905 clear_opt(sbi->s_mount_opt, POSIX_ACL);
906 break;
907#else
908 case Opt_acl:
909 case Opt_noacl:
Mingming Cao617ba132006-10-11 01:20:53 -0700910 printk("EXT4 (no)acl options not supported\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700911 break;
912#endif
913 case Opt_reservation:
914 set_opt(sbi->s_mount_opt, RESERVATION);
915 break;
916 case Opt_noreservation:
917 clear_opt(sbi->s_mount_opt, RESERVATION);
918 break;
919 case Opt_journal_update:
920 /* @@@ FIXME */
921 /* Eventually we will want to be able to create
922 a journal file here. For now, only allow the
923 user to specify an existing inode to be the
924 journal file. */
925 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -0700926 printk(KERN_ERR "EXT4-fs: cannot specify "
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700927 "journal on remount\n");
928 return 0;
929 }
930 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
931 break;
932 case Opt_journal_inum:
933 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -0700934 printk(KERN_ERR "EXT4-fs: cannot specify "
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700935 "journal on remount\n");
936 return 0;
937 }
938 if (match_int(&args[0], &option))
939 return 0;
940 *inum = option;
941 break;
942 case Opt_journal_dev:
943 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -0700944 printk(KERN_ERR "EXT4-fs: cannot specify "
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700945 "journal on remount\n");
946 return 0;
947 }
948 if (match_int(&args[0], &option))
949 return 0;
950 *journal_devnum = option;
951 break;
952 case Opt_noload:
953 set_opt (sbi->s_mount_opt, NOLOAD);
954 break;
955 case Opt_commit:
956 if (match_int(&args[0], &option))
957 return 0;
958 if (option < 0)
959 return 0;
960 if (option == 0)
961 option = JBD_DEFAULT_MAX_COMMIT_AGE;
962 sbi->s_commit_interval = HZ * option;
963 break;
964 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -0700965 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700966 goto datacheck;
967 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -0700968 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700969 goto datacheck;
970 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -0700971 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700972 datacheck:
973 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -0700974 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700975 != data_opt) {
976 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -0700977 "EXT4-fs: cannot change data "
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700978 "mode on remount\n");
979 return 0;
980 }
981 } else {
Mingming Cao617ba132006-10-11 01:20:53 -0700982 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700983 sbi->s_mount_opt |= data_opt;
984 }
985 break;
986#ifdef CONFIG_QUOTA
987 case Opt_usrjquota:
988 qtype = USRQUOTA;
989 goto set_qf_name;
990 case Opt_grpjquota:
991 qtype = GRPQUOTA;
992set_qf_name:
993 if (sb_any_quota_enabled(sb)) {
994 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -0700995 "EXT4-fs: Cannot change journalled "
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700996 "quota options when quota turned on.\n");
997 return 0;
998 }
999 qname = match_strdup(&args[0]);
1000 if (!qname) {
1001 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001002 "EXT4-fs: not enough memory for "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001003 "storing quotafile name.\n");
1004 return 0;
1005 }
1006 if (sbi->s_qf_names[qtype] &&
1007 strcmp(sbi->s_qf_names[qtype], qname)) {
1008 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001009 "EXT4-fs: %s quota file already "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001010 "specified.\n", QTYPE2NAME(qtype));
1011 kfree(qname);
1012 return 0;
1013 }
1014 sbi->s_qf_names[qtype] = qname;
1015 if (strchr(sbi->s_qf_names[qtype], '/')) {
1016 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001017 "EXT4-fs: quotafile must be on "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001018 "filesystem root.\n");
1019 kfree(sbi->s_qf_names[qtype]);
1020 sbi->s_qf_names[qtype] = NULL;
1021 return 0;
1022 }
1023 set_opt(sbi->s_mount_opt, QUOTA);
1024 break;
1025 case Opt_offusrjquota:
1026 qtype = USRQUOTA;
1027 goto clear_qf_name;
1028 case Opt_offgrpjquota:
1029 qtype = GRPQUOTA;
1030clear_qf_name:
1031 if (sb_any_quota_enabled(sb)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001032 printk(KERN_ERR "EXT4-fs: Cannot change "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001033 "journalled quota options when "
1034 "quota turned on.\n");
1035 return 0;
1036 }
1037 /*
1038 * The space will be released later when all options
1039 * are confirmed to be correct
1040 */
1041 sbi->s_qf_names[qtype] = NULL;
1042 break;
1043 case Opt_jqfmt_vfsold:
1044 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1045 break;
1046 case Opt_jqfmt_vfsv0:
1047 sbi->s_jquota_fmt = QFMT_VFS_V0;
1048 break;
1049 case Opt_quota:
1050 case Opt_usrquota:
1051 set_opt(sbi->s_mount_opt, QUOTA);
1052 set_opt(sbi->s_mount_opt, USRQUOTA);
1053 break;
1054 case Opt_grpquota:
1055 set_opt(sbi->s_mount_opt, QUOTA);
1056 set_opt(sbi->s_mount_opt, GRPQUOTA);
1057 break;
1058 case Opt_noquota:
1059 if (sb_any_quota_enabled(sb)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001060 printk(KERN_ERR "EXT4-fs: Cannot change quota "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001061 "options when quota turned on.\n");
1062 return 0;
1063 }
1064 clear_opt(sbi->s_mount_opt, QUOTA);
1065 clear_opt(sbi->s_mount_opt, USRQUOTA);
1066 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1067 break;
1068#else
1069 case Opt_quota:
1070 case Opt_usrquota:
1071 case Opt_grpquota:
1072 case Opt_usrjquota:
1073 case Opt_grpjquota:
1074 case Opt_offusrjquota:
1075 case Opt_offgrpjquota:
1076 case Opt_jqfmt_vfsold:
1077 case Opt_jqfmt_vfsv0:
1078 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001079 "EXT4-fs: journalled quota options not "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001080 "supported.\n");
1081 break;
1082 case Opt_noquota:
1083 break;
1084#endif
1085 case Opt_abort:
1086 set_opt(sbi->s_mount_opt, ABORT);
1087 break;
1088 case Opt_barrier:
1089 if (match_int(&args[0], &option))
1090 return 0;
1091 if (option)
1092 set_opt(sbi->s_mount_opt, BARRIER);
1093 else
1094 clear_opt(sbi->s_mount_opt, BARRIER);
1095 break;
1096 case Opt_ignore:
1097 break;
1098 case Opt_resize:
1099 if (!is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001100 printk("EXT4-fs: resize option only available "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001101 "for remount\n");
1102 return 0;
1103 }
1104 if (match_int(&args[0], &option) != 0)
1105 return 0;
1106 *n_blocks_count = option;
1107 break;
1108 case Opt_nobh:
1109 set_opt(sbi->s_mount_opt, NOBH);
1110 break;
1111 case Opt_bh:
1112 clear_opt(sbi->s_mount_opt, NOBH);
1113 break;
Alex Tomasa86c6182006-10-11 01:21:03 -07001114 case Opt_extents:
1115 set_opt (sbi->s_mount_opt, EXTENTS);
1116 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001117 default:
1118 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001119 "EXT4-fs: Unrecognized mount option \"%s\" "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001120 "or missing value\n", p);
1121 return 0;
1122 }
1123 }
1124#ifdef CONFIG_QUOTA
1125 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Mingming Cao617ba132006-10-11 01:20:53 -07001126 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001127 sbi->s_qf_names[USRQUOTA])
1128 clear_opt(sbi->s_mount_opt, USRQUOTA);
1129
Mingming Cao617ba132006-10-11 01:20:53 -07001130 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001131 sbi->s_qf_names[GRPQUOTA])
1132 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1133
1134 if ((sbi->s_qf_names[USRQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001135 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001136 (sbi->s_qf_names[GRPQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001137 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1138 printk(KERN_ERR "EXT4-fs: old and new quota "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001139 "format mixing.\n");
1140 return 0;
1141 }
1142
1143 if (!sbi->s_jquota_fmt) {
Mingming Cao617ba132006-10-11 01:20:53 -07001144 printk(KERN_ERR "EXT4-fs: journalled quota format "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001145 "not specified.\n");
1146 return 0;
1147 }
1148 } else {
1149 if (sbi->s_jquota_fmt) {
Mingming Cao617ba132006-10-11 01:20:53 -07001150 printk(KERN_ERR "EXT4-fs: journalled quota format "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001151 "specified with no journalling "
1152 "enabled.\n");
1153 return 0;
1154 }
1155 }
1156#endif
1157 return 1;
1158}
1159
Mingming Cao617ba132006-10-11 01:20:53 -07001160static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001161 int read_only)
1162{
Mingming Cao617ba132006-10-11 01:20:53 -07001163 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001164 int res = 0;
1165
Mingming Cao617ba132006-10-11 01:20:53 -07001166 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1167 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001168 "forcing read-only mode\n");
1169 res = MS_RDONLY;
1170 }
1171 if (read_only)
1172 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001173 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1174 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001175 "running e2fsck is recommended\n");
Mingming Cao617ba132006-10-11 01:20:53 -07001176 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001177 printk (KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001178 "EXT4-fs warning: mounting fs with errors, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001179 "running e2fsck is recommended\n");
1180 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1181 le16_to_cpu(es->s_mnt_count) >=
1182 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1183 printk (KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001184 "EXT4-fs warning: maximal mount count reached, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001185 "running e2fsck is recommended\n");
1186 else if (le32_to_cpu(es->s_checkinterval) &&
1187 (le32_to_cpu(es->s_lastcheck) +
1188 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1189 printk (KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001190 "EXT4-fs warning: checktime reached, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001191 "running e2fsck is recommended\n");
1192#if 0
1193 /* @@@ We _will_ want to clear the valid bit if we find
1194 inconsistencies, to force a fsck at reboot. But for
1195 a plain journaled filesystem we can keep it set as
1196 valid forever! :) */
Mingming Cao617ba132006-10-11 01:20:53 -07001197 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001198#endif
1199 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001200 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001201 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1202 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001203 ext4_update_dynamic_rev(sb);
1204 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001205
Mingming Cao617ba132006-10-11 01:20:53 -07001206 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001207 if (test_opt(sb, DEBUG))
Mingming Cao617ba132006-10-11 01:20:53 -07001208 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001209 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1210 sb->s_blocksize,
1211 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001212 EXT4_BLOCKS_PER_GROUP(sb),
1213 EXT4_INODES_PER_GROUP(sb),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001214 sbi->s_mount_opt);
1215
Mingming Cao617ba132006-10-11 01:20:53 -07001216 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1217 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001218 char b[BDEVNAME_SIZE];
1219
1220 printk("external journal on %s\n",
Mingming Cao617ba132006-10-11 01:20:53 -07001221 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001222 } else {
1223 printk("internal journal\n");
1224 }
1225 return res;
1226}
1227
1228/* Called at mount-time, super-block is locked */
Mingming Cao617ba132006-10-11 01:20:53 -07001229static int ext4_check_descriptors (struct super_block * sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001230{
Mingming Cao617ba132006-10-11 01:20:53 -07001231 struct ext4_sb_info *sbi = EXT4_SB(sb);
1232 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1233 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001234 ext4_fsblk_t block_bitmap;
1235 ext4_fsblk_t inode_bitmap;
1236 ext4_fsblk_t inode_table;
Mingming Cao617ba132006-10-11 01:20:53 -07001237 struct ext4_group_desc * gdp = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001238 int desc_block = 0;
1239 int i;
1240
Mingming Cao617ba132006-10-11 01:20:53 -07001241 ext4_debug ("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001242
1243 for (i = 0; i < sbi->s_groups_count; i++)
1244 {
1245 if (i == sbi->s_groups_count - 1)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001246 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001247 else
1248 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07001249 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001250
Mingming Cao617ba132006-10-11 01:20:53 -07001251 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1252 gdp = (struct ext4_group_desc *)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001253 sbi->s_group_desc[desc_block++]->b_data;
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001254 block_bitmap = ext4_block_bitmap(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001255 if (block_bitmap < first_block || block_bitmap > last_block)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001256 {
Mingming Cao617ba132006-10-11 01:20:53 -07001257 ext4_error (sb, "ext4_check_descriptors",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001258 "Block bitmap for group %d"
Mingming Cao2ae02102006-10-11 01:21:11 -07001259 " not in group (block %llu)!",
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001260 i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001261 return 0;
1262 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001263 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001264 if (inode_bitmap < first_block || inode_bitmap > last_block)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001265 {
Mingming Cao617ba132006-10-11 01:20:53 -07001266 ext4_error (sb, "ext4_check_descriptors",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001267 "Inode bitmap for group %d"
Mingming Cao2ae02102006-10-11 01:21:11 -07001268 " not in group (block %llu)!",
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001269 i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001270 return 0;
1271 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001272 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001273 if (inode_table < first_block ||
1274 inode_table + sbi->s_itb_per_group > last_block)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001275 {
Mingming Cao617ba132006-10-11 01:20:53 -07001276 ext4_error (sb, "ext4_check_descriptors",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001277 "Inode table for group %d"
Mingming Cao2ae02102006-10-11 01:21:11 -07001278 " not in group (block %llu)!",
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001279 i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001280 return 0;
1281 }
Mingming Cao617ba132006-10-11 01:20:53 -07001282 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07001283 gdp = (struct ext4_group_desc *)
1284 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001285 }
1286
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001287 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Mingming Cao617ba132006-10-11 01:20:53 -07001288 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001289 return 1;
1290}
1291
1292
Mingming Cao617ba132006-10-11 01:20:53 -07001293/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001294 * the superblock) which were deleted from all directories, but held open by
1295 * a process at the time of a crash. We walk the list and try to delete these
1296 * inodes at recovery time (only with a read-write filesystem).
1297 *
1298 * In order to keep the orphan inode chain consistent during traversal (in
1299 * case of crash during recovery), we link each inode into the superblock
1300 * orphan list_head and handle it the same way as an inode deletion during
1301 * normal operation (which journals the operations for us).
1302 *
1303 * We only do an iget() and an iput() on each inode, which is very safe if we
1304 * accidentally point at an in-use or already deleted inode. The worst that
1305 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07001306 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001307 * e2fsck was run on this filesystem, and it must have already done the orphan
1308 * inode cleanup for us, so we can safely abort without any further action.
1309 */
Mingming Cao617ba132006-10-11 01:20:53 -07001310static void ext4_orphan_cleanup (struct super_block * sb,
1311 struct ext4_super_block * es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001312{
1313 unsigned int s_flags = sb->s_flags;
1314 int nr_orphans = 0, nr_truncates = 0;
1315#ifdef CONFIG_QUOTA
1316 int i;
1317#endif
1318 if (!es->s_last_orphan) {
1319 jbd_debug(4, "no orphan inodes to clean up\n");
1320 return;
1321 }
1322
Mingming Cao617ba132006-10-11 01:20:53 -07001323 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001324 if (es->s_last_orphan)
1325 jbd_debug(1, "Errors on filesystem, "
1326 "clearing orphan list.\n");
1327 es->s_last_orphan = 0;
1328 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1329 return;
1330 }
1331
1332 if (s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07001333 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001334 sb->s_id);
1335 sb->s_flags &= ~MS_RDONLY;
1336 }
1337#ifdef CONFIG_QUOTA
1338 /* Needed for iput() to work correctly and not trash data */
1339 sb->s_flags |= MS_ACTIVE;
1340 /* Turn on quotas so that they are updated correctly */
1341 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07001342 if (EXT4_SB(sb)->s_qf_names[i]) {
1343 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001344 if (ret < 0)
1345 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001346 "EXT4-fs: Cannot turn on journalled "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001347 "quota: error %d\n", ret);
1348 }
1349 }
1350#endif
1351
1352 while (es->s_last_orphan) {
1353 struct inode *inode;
1354
1355 if (!(inode =
Mingming Cao617ba132006-10-11 01:20:53 -07001356 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001357 es->s_last_orphan = 0;
1358 break;
1359 }
1360
Mingming Cao617ba132006-10-11 01:20:53 -07001361 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001362 DQUOT_INIT(inode);
1363 if (inode->i_nlink) {
1364 printk(KERN_DEBUG
1365 "%s: truncating inode %lu to %Ld bytes\n",
1366 __FUNCTION__, inode->i_ino, inode->i_size);
1367 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1368 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07001369 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001370 nr_truncates++;
1371 } else {
1372 printk(KERN_DEBUG
1373 "%s: deleting unreferenced inode %lu\n",
1374 __FUNCTION__, inode->i_ino);
1375 jbd_debug(2, "deleting unreferenced inode %lu\n",
1376 inode->i_ino);
1377 nr_orphans++;
1378 }
1379 iput(inode); /* The delete magic happens here! */
1380 }
1381
1382#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1383
1384 if (nr_orphans)
Mingming Cao617ba132006-10-11 01:20:53 -07001385 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001386 sb->s_id, PLURAL(nr_orphans));
1387 if (nr_truncates)
Mingming Cao617ba132006-10-11 01:20:53 -07001388 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001389 sb->s_id, PLURAL(nr_truncates));
1390#ifdef CONFIG_QUOTA
1391 /* Turn quotas off */
1392 for (i = 0; i < MAXQUOTAS; i++) {
1393 if (sb_dqopt(sb)->files[i])
1394 vfs_quota_off(sb, i);
1395 }
1396#endif
1397 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1398}
1399
1400#define log2(n) ffz(~(n))
1401
1402/*
1403 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1404 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1405 * We need to be 1 filesystem block less than the 2^32 sector limit.
1406 */
Mingming Cao617ba132006-10-11 01:20:53 -07001407static loff_t ext4_max_size(int bits)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001408{
Mingming Cao617ba132006-10-11 01:20:53 -07001409 loff_t res = EXT4_NDIR_BLOCKS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001410 /* This constant is calculated to be the largest file size for a
1411 * dense, 4k-blocksize file such that the total number of
1412 * sectors in the file, including data and all indirect blocks,
1413 * does not exceed 2^32. */
1414 const loff_t upper_limit = 0x1ff7fffd000LL;
1415
1416 res += 1LL << (bits-2);
1417 res += 1LL << (2*(bits-2));
1418 res += 1LL << (3*(bits-2));
1419 res <<= bits;
1420 if (res > upper_limit)
1421 res = upper_limit;
1422 return res;
1423}
1424
Mingming Cao617ba132006-10-11 01:20:53 -07001425static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001426 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001427{
Mingming Cao617ba132006-10-11 01:20:53 -07001428 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001429 unsigned long bg, first_meta_bg;
1430 int has_super = 0;
1431
1432 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1433
Mingming Cao617ba132006-10-11 01:20:53 -07001434 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001435 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001436 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001437 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07001438 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001439 has_super = 1;
Mingming Cao617ba132006-10-11 01:20:53 -07001440 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001441}
1442
1443
Mingming Cao617ba132006-10-11 01:20:53 -07001444static int ext4_fill_super (struct super_block *sb, void *data, int silent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001445{
1446 struct buffer_head * bh;
Mingming Cao617ba132006-10-11 01:20:53 -07001447 struct ext4_super_block *es = NULL;
1448 struct ext4_sb_info *sbi;
1449 ext4_fsblk_t block;
1450 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001451 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001452 unsigned long offset = 0;
1453 unsigned int journal_inum = 0;
1454 unsigned long journal_devnum = 0;
1455 unsigned long def_mount_opts;
1456 struct inode *root;
1457 int blocksize;
1458 int hblock;
1459 int db_count;
1460 int i;
1461 int needs_recovery;
1462 __le32 features;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001463 __u64 blocks_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001464
1465 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1466 if (!sbi)
1467 return -ENOMEM;
1468 sb->s_fs_info = sbi;
1469 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07001470 sbi->s_resuid = EXT4_DEF_RESUID;
1471 sbi->s_resgid = EXT4_DEF_RESGID;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001472
1473 unlock_kernel();
1474
Mingming Cao617ba132006-10-11 01:20:53 -07001475 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001476 if (!blocksize) {
Mingming Cao617ba132006-10-11 01:20:53 -07001477 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001478 goto out_fail;
1479 }
1480
1481 /*
Mingming Cao617ba132006-10-11 01:20:53 -07001482 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001483 * block sizes. We need to calculate the offset from buffer start.
1484 */
Mingming Cao617ba132006-10-11 01:20:53 -07001485 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001486 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1487 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001488 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001489 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001490 }
1491
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001492 if (!(bh = sb_bread(sb, logical_sb_block))) {
Mingming Cao617ba132006-10-11 01:20:53 -07001493 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001494 goto out_fail;
1495 }
1496 /*
1497 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07001498 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001499 */
Mingming Cao617ba132006-10-11 01:20:53 -07001500 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001501 sbi->s_es = es;
1502 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07001503 if (sb->s_magic != EXT4_SUPER_MAGIC)
1504 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001505
1506 /* Set defaults before we parse the mount options */
1507 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Mingming Cao617ba132006-10-11 01:20:53 -07001508 if (def_mount_opts & EXT4_DEFM_DEBUG)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001509 set_opt(sbi->s_mount_opt, DEBUG);
Mingming Cao617ba132006-10-11 01:20:53 -07001510 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001511 set_opt(sbi->s_mount_opt, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07001512 if (def_mount_opts & EXT4_DEFM_UID16)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001513 set_opt(sbi->s_mount_opt, NO_UID32);
Mingming Cao617ba132006-10-11 01:20:53 -07001514 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001515 set_opt(sbi->s_mount_opt, XATTR_USER);
Mingming Cao617ba132006-10-11 01:20:53 -07001516 if (def_mount_opts & EXT4_DEFM_ACL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001517 set_opt(sbi->s_mount_opt, POSIX_ACL);
Mingming Cao617ba132006-10-11 01:20:53 -07001518 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1519 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1520 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1521 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1522 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1523 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001524
Mingming Cao617ba132006-10-11 01:20:53 -07001525 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001526 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Mingming Cao617ba132006-10-11 01:20:53 -07001527 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001528 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dmitry Mishinceea16b2006-10-11 01:21:21 -07001529 else
1530 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001531
1532 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1533 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1534
1535 set_opt(sbi->s_mount_opt, RESERVATION);
1536
1537 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1538 NULL, 0))
1539 goto failed_mount;
1540
1541 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07001542 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001543
Mingming Cao617ba132006-10-11 01:20:53 -07001544 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1545 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1546 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1547 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001548 printk(KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07001549 "EXT4-fs warning: feature flags set on rev 0 fs, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001550 "running e2fsck is recommended\n");
1551 /*
1552 * Check feature flags regardless of the revision level, since we
1553 * previously didn't change the revision level when setting the flags,
1554 * so there is a chance incompat flags are set on a rev 0 filesystem.
1555 */
Mingming Cao617ba132006-10-11 01:20:53 -07001556 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001557 if (features) {
Mingming Cao617ba132006-10-11 01:20:53 -07001558 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001559 "unsupported optional features (%x).\n",
1560 sb->s_id, le32_to_cpu(features));
1561 goto failed_mount;
1562 }
Mingming Cao617ba132006-10-11 01:20:53 -07001563 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001564 if (!(sb->s_flags & MS_RDONLY) && features) {
Mingming Cao617ba132006-10-11 01:20:53 -07001565 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001566 "unsupported optional features (%x).\n",
1567 sb->s_id, le32_to_cpu(features));
1568 goto failed_mount;
1569 }
1570 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1571
Mingming Cao617ba132006-10-11 01:20:53 -07001572 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1573 blocksize > EXT4_MAX_BLOCK_SIZE) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001574 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001575 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001576 blocksize, sb->s_id);
1577 goto failed_mount;
1578 }
1579
1580 hblock = bdev_hardsect_size(sb->s_bdev);
1581 if (sb->s_blocksize != blocksize) {
1582 /*
1583 * Make sure the blocksize for the filesystem is larger
1584 * than the hardware sectorsize for the machine.
1585 */
1586 if (blocksize < hblock) {
Mingming Cao617ba132006-10-11 01:20:53 -07001587 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001588 "device blocksize %d.\n", blocksize, hblock);
1589 goto failed_mount;
1590 }
1591
1592 brelse (bh);
1593 sb_set_blocksize(sb, blocksize);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001594 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1595 offset = do_div(logical_sb_block, blocksize);
1596 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001597 if (!bh) {
1598 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001599 "EXT4-fs: Can't read superblock on 2nd try.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001600 goto failed_mount;
1601 }
Mingming Cao617ba132006-10-11 01:20:53 -07001602 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001603 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07001604 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001605 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001606 "EXT4-fs: Magic mismatch, very weird !\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001607 goto failed_mount;
1608 }
1609 }
1610
Mingming Cao617ba132006-10-11 01:20:53 -07001611 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001612
Mingming Cao617ba132006-10-11 01:20:53 -07001613 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1614 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1615 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001616 } else {
1617 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1618 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07001619 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001620 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
1621 (sbi->s_inode_size > blocksize)) {
1622 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001623 "EXT4-fs: unsupported inode size: %d\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001624 sbi->s_inode_size);
1625 goto failed_mount;
1626 }
1627 }
Mingming Cao617ba132006-10-11 01:20:53 -07001628 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001629 le32_to_cpu(es->s_log_frag_size);
1630 if (blocksize != sbi->s_frag_size) {
1631 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001632 "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001633 sbi->s_frag_size, blocksize);
1634 goto failed_mount;
1635 }
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07001636 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1637 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001638 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07001639 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1640 sbi->s_desc_size & (sbi->s_desc_size - 1)) {
1641 printk(KERN_ERR
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001642 "EXT4-fs: unsupported descriptor size %lu\n",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07001643 sbi->s_desc_size);
1644 goto failed_mount;
1645 }
1646 } else
1647 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001648 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1649 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1650 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Mingming Cao617ba132006-10-11 01:20:53 -07001651 if (EXT4_INODE_SIZE(sb) == 0)
1652 goto cantfind_ext4;
1653 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001654 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07001655 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001656 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1657 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07001658 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001659 sbi->s_sbh = bh;
1660 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07001661 sbi->s_addr_per_block_bits = log2(EXT4_ADDR_PER_BLOCK(sb));
1662 sbi->s_desc_per_block_bits = log2(EXT4_DESC_PER_BLOCK(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001663 for (i=0; i < 4; i++)
1664 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1665 sbi->s_def_hash_version = es->s_def_hash_version;
1666
1667 if (sbi->s_blocks_per_group > blocksize * 8) {
1668 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001669 "EXT4-fs: #blocks per group too big: %lu\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001670 sbi->s_blocks_per_group);
1671 goto failed_mount;
1672 }
1673 if (sbi->s_frags_per_group > blocksize * 8) {
1674 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001675 "EXT4-fs: #fragments per group too big: %lu\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001676 sbi->s_frags_per_group);
1677 goto failed_mount;
1678 }
1679 if (sbi->s_inodes_per_group > blocksize * 8) {
1680 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001681 "EXT4-fs: #inodes per group too big: %lu\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001682 sbi->s_inodes_per_group);
1683 goto failed_mount;
1684 }
1685
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001686 if (ext4_blocks_count(es) >
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001687 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001688 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001689 " too large to mount safely\n", sb->s_id);
1690 if (sizeof(sector_t) < 8)
Mingming Cao617ba132006-10-11 01:20:53 -07001691 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001692 "enabled\n");
1693 goto failed_mount;
1694 }
1695
Mingming Cao617ba132006-10-11 01:20:53 -07001696 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1697 goto cantfind_ext4;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001698 blocks_count = (ext4_blocks_count(es) -
1699 le32_to_cpu(es->s_first_data_block) +
1700 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1701 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1702 sbi->s_groups_count = blocks_count;
Mingming Cao617ba132006-10-11 01:20:53 -07001703 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1704 EXT4_DESC_PER_BLOCK(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001705 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1706 GFP_KERNEL);
1707 if (sbi->s_group_desc == NULL) {
Mingming Cao617ba132006-10-11 01:20:53 -07001708 printk (KERN_ERR "EXT4-fs: not enough memory\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001709 goto failed_mount;
1710 }
1711
1712 bgl_lock_init(&sbi->s_blockgroup_lock);
1713
1714 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07001715 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001716 sbi->s_group_desc[i] = sb_bread(sb, block);
1717 if (!sbi->s_group_desc[i]) {
Mingming Cao617ba132006-10-11 01:20:53 -07001718 printk (KERN_ERR "EXT4-fs: "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001719 "can't read group descriptor %d\n", i);
1720 db_count = i;
1721 goto failed_mount2;
1722 }
1723 }
Mingming Cao617ba132006-10-11 01:20:53 -07001724 if (!ext4_check_descriptors (sb)) {
1725 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001726 goto failed_mount2;
1727 }
1728 sbi->s_gdb_count = db_count;
1729 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1730 spin_lock_init(&sbi->s_next_gen_lock);
1731
1732 percpu_counter_init(&sbi->s_freeblocks_counter,
Mingming Cao617ba132006-10-11 01:20:53 -07001733 ext4_count_free_blocks(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001734 percpu_counter_init(&sbi->s_freeinodes_counter,
Mingming Cao617ba132006-10-11 01:20:53 -07001735 ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001736 percpu_counter_init(&sbi->s_dirs_counter,
Mingming Cao617ba132006-10-11 01:20:53 -07001737 ext4_count_dirs(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001738
1739 /* per fileystem reservation list head & lock */
1740 spin_lock_init(&sbi->s_rsv_window_lock);
1741 sbi->s_rsv_window_root = RB_ROOT;
1742 /* Add a single, static dummy reservation to the start of the
1743 * reservation window list --- it gives us a placeholder for
1744 * append-at-start-of-list which makes the allocation logic
1745 * _much_ simpler. */
Mingming Cao617ba132006-10-11 01:20:53 -07001746 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1747 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001748 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1749 sbi->s_rsv_window_head.rsv_goal_size = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07001750 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001751
1752 /*
1753 * set up enough so that it can read an inode
1754 */
Mingming Cao617ba132006-10-11 01:20:53 -07001755 sb->s_op = &ext4_sops;
1756 sb->s_export_op = &ext4_export_ops;
1757 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001758#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001759 sb->s_qcop = &ext4_qctl_operations;
1760 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001761#endif
1762 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1763
1764 sb->s_root = NULL;
1765
1766 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07001767 EXT4_HAS_INCOMPAT_FEATURE(sb,
1768 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001769
1770 /*
1771 * The first inode we look at is the journal inode. Don't try
1772 * root first: it may be modified in the journal!
1773 */
1774 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07001775 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1776 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001777 goto failed_mount3;
1778 } else if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07001779 if (ext4_create_journal(sb, es, journal_inum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001780 goto failed_mount3;
1781 } else {
1782 if (!silent)
1783 printk (KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001784 "ext4: No journal on filesystem on %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001785 sb->s_id);
1786 goto failed_mount3;
1787 }
1788
1789 /* We have now updated the journal if required, so we can
1790 * validate the data journaling mode. */
1791 switch (test_opt(sb, DATA_FLAGS)) {
1792 case 0:
1793 /* No mode set, assume a default based on the journal
1794 capabilities: ORDERED_DATA if the journal can
1795 cope, else JOURNAL_DATA */
Mingming Caodab291a2006-10-11 01:21:01 -07001796 if (jbd2_journal_check_available_features
1797 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001798 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1799 else
1800 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1801 break;
1802
Mingming Cao617ba132006-10-11 01:20:53 -07001803 case EXT4_MOUNT_ORDERED_DATA:
1804 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07001805 if (!jbd2_journal_check_available_features
1806 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001807 printk(KERN_ERR "EXT4-fs: Journal does not support "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001808 "requested data journaling mode\n");
1809 goto failed_mount4;
1810 }
1811 default:
1812 break;
1813 }
1814
1815 if (test_opt(sb, NOBH)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001816 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1817 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001818 "its supported only with writeback mode\n");
1819 clear_opt(sbi->s_mount_opt, NOBH);
1820 }
1821 }
1822 /*
Mingming Caodab291a2006-10-11 01:21:01 -07001823 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001824 * so we can safely mount the rest of the filesystem now.
1825 */
1826
Mingming Cao617ba132006-10-11 01:20:53 -07001827 root = iget(sb, EXT4_ROOT_INO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001828 sb->s_root = d_alloc_root(root);
1829 if (!sb->s_root) {
Mingming Cao617ba132006-10-11 01:20:53 -07001830 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001831 iput(root);
1832 goto failed_mount4;
1833 }
1834 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1835 dput(sb->s_root);
1836 sb->s_root = NULL;
Mingming Cao617ba132006-10-11 01:20:53 -07001837 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001838 goto failed_mount4;
1839 }
1840
Mingming Cao617ba132006-10-11 01:20:53 -07001841 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001842 /*
1843 * akpm: core read_super() calls in here with the superblock locked.
1844 * That deadlocks, because orphan cleanup needs to lock the superblock
1845 * in numerous places. Here we just pop the lock - it's relatively
1846 * harmless, because we are now ready to accept write_super() requests,
1847 * and aviro says that's the only reason for hanging onto the
1848 * superblock lock.
1849 */
Mingming Cao617ba132006-10-11 01:20:53 -07001850 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1851 ext4_orphan_cleanup(sb, es);
1852 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001853 if (needs_recovery)
Mingming Cao617ba132006-10-11 01:20:53 -07001854 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1855 ext4_mark_recovery_complete(sb, es);
1856 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
1857 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
1858 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001859 "writeback");
1860
Alex Tomasa86c6182006-10-11 01:21:03 -07001861 ext4_ext_init(sb);
1862
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001863 lock_kernel();
1864 return 0;
1865
Mingming Cao617ba132006-10-11 01:20:53 -07001866cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001867 if (!silent)
Mingming Cao617ba132006-10-11 01:20:53 -07001868 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 sb->s_id);
1870 goto failed_mount;
1871
1872failed_mount4:
Mingming Caodab291a2006-10-11 01:21:01 -07001873 jbd2_journal_destroy(sbi->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001874failed_mount3:
1875 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1876 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1877 percpu_counter_destroy(&sbi->s_dirs_counter);
1878failed_mount2:
1879 for (i = 0; i < db_count; i++)
1880 brelse(sbi->s_group_desc[i]);
1881 kfree(sbi->s_group_desc);
1882failed_mount:
1883#ifdef CONFIG_QUOTA
1884 for (i = 0; i < MAXQUOTAS; i++)
1885 kfree(sbi->s_qf_names[i]);
1886#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001887 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001888 brelse(bh);
1889out_fail:
1890 sb->s_fs_info = NULL;
1891 kfree(sbi);
1892 lock_kernel();
1893 return -EINVAL;
1894}
1895
1896/*
1897 * Setup any per-fs journal parameters now. We'll do this both on
1898 * initial mount, once the journal has been initialised but before we've
1899 * done any recovery; and again on any subsequent remount.
1900 */
Mingming Cao617ba132006-10-11 01:20:53 -07001901static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001902{
Mingming Cao617ba132006-10-11 01:20:53 -07001903 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001904
1905 if (sbi->s_commit_interval)
1906 journal->j_commit_interval = sbi->s_commit_interval;
Mingming Cao617ba132006-10-11 01:20:53 -07001907 /* We could also set up an ext4-specific default for the commit
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001908 * interval here, but for now we'll just fall back to the jbd
1909 * default. */
1910
1911 spin_lock(&journal->j_state_lock);
1912 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07001913 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001914 else
Mingming Caodab291a2006-10-11 01:21:01 -07001915 journal->j_flags &= ~JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001916 spin_unlock(&journal->j_state_lock);
1917}
1918
Mingming Cao617ba132006-10-11 01:20:53 -07001919static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001920 unsigned int journal_inum)
1921{
1922 struct inode *journal_inode;
1923 journal_t *journal;
1924
1925 /* First, test for the existence of a valid inode on disk. Bad
1926 * things happen if we iget() an unused inode, as the subsequent
1927 * iput() will try to delete it. */
1928
1929 journal_inode = iget(sb, journal_inum);
1930 if (!journal_inode) {
Mingming Cao617ba132006-10-11 01:20:53 -07001931 printk(KERN_ERR "EXT4-fs: no journal found.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001932 return NULL;
1933 }
1934 if (!journal_inode->i_nlink) {
1935 make_bad_inode(journal_inode);
1936 iput(journal_inode);
Mingming Cao617ba132006-10-11 01:20:53 -07001937 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001938 return NULL;
1939 }
1940
1941 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
1942 journal_inode, journal_inode->i_size);
1943 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
Mingming Cao617ba132006-10-11 01:20:53 -07001944 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001945 iput(journal_inode);
1946 return NULL;
1947 }
1948
Mingming Caodab291a2006-10-11 01:21:01 -07001949 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001950 if (!journal) {
Mingming Cao617ba132006-10-11 01:20:53 -07001951 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001952 iput(journal_inode);
1953 return NULL;
1954 }
1955 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07001956 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001957 return journal;
1958}
1959
Mingming Cao617ba132006-10-11 01:20:53 -07001960static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001961 dev_t j_dev)
1962{
1963 struct buffer_head * bh;
1964 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07001965 ext4_fsblk_t start;
1966 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001967 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07001968 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001969 unsigned long offset;
Mingming Cao617ba132006-10-11 01:20:53 -07001970 struct ext4_super_block * es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001971 struct block_device *bdev;
1972
Mingming Cao617ba132006-10-11 01:20:53 -07001973 bdev = ext4_blkdev_get(j_dev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001974 if (bdev == NULL)
1975 return NULL;
1976
1977 if (bd_claim(bdev, sb)) {
1978 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001979 "EXT4: failed to claim external journal device.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001980 blkdev_put(bdev);
1981 return NULL;
1982 }
1983
1984 blocksize = sb->s_blocksize;
1985 hblock = bdev_hardsect_size(bdev);
1986 if (blocksize < hblock) {
1987 printk(KERN_ERR
Mingming Cao617ba132006-10-11 01:20:53 -07001988 "EXT4-fs: blocksize too small for journal device.\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001989 goto out_bdev;
1990 }
1991
Mingming Cao617ba132006-10-11 01:20:53 -07001992 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
1993 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001994 set_blocksize(bdev, blocksize);
1995 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Mingming Cao617ba132006-10-11 01:20:53 -07001996 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001997 "external journal\n");
1998 goto out_bdev;
1999 }
2000
Mingming Cao617ba132006-10-11 01:20:53 -07002001 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2002 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002003 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07002004 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2005 printk(KERN_ERR "EXT4-fs: external journal has "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002006 "bad superblock\n");
2007 brelse(bh);
2008 goto out_bdev;
2009 }
2010
Mingming Cao617ba132006-10-11 01:20:53 -07002011 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2012 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002013 brelse(bh);
2014 goto out_bdev;
2015 }
2016
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002017 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002018 start = sb_block + 1;
2019 brelse(bh); /* we're done with the superblock */
2020
Mingming Caodab291a2006-10-11 01:21:01 -07002021 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002022 start, len, blocksize);
2023 if (!journal) {
Mingming Cao617ba132006-10-11 01:20:53 -07002024 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002025 goto out_bdev;
2026 }
2027 journal->j_private = sb;
2028 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2029 wait_on_buffer(journal->j_sb_buffer);
2030 if (!buffer_uptodate(journal->j_sb_buffer)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002031 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002032 goto out_journal;
2033 }
2034 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Mingming Cao617ba132006-10-11 01:20:53 -07002035 printk(KERN_ERR "EXT4-fs: External journal has more than one "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002036 "user (unsupported) - %d\n",
2037 be32_to_cpu(journal->j_superblock->s_nr_users));
2038 goto out_journal;
2039 }
Mingming Cao617ba132006-10-11 01:20:53 -07002040 EXT4_SB(sb)->journal_bdev = bdev;
2041 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002042 return journal;
2043out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07002044 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002045out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07002046 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002047 return NULL;
2048}
2049
Mingming Cao617ba132006-10-11 01:20:53 -07002050static int ext4_load_journal(struct super_block *sb,
2051 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002052 unsigned long journal_devnum)
2053{
2054 journal_t *journal;
2055 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2056 dev_t journal_dev;
2057 int err = 0;
2058 int really_read_only;
2059
2060 if (journal_devnum &&
2061 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002062 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002063 "numbers have changed\n");
2064 journal_dev = new_decode_dev(journal_devnum);
2065 } else
2066 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2067
2068 really_read_only = bdev_read_only(sb->s_bdev);
2069
2070 /*
2071 * Are we loading a blank journal or performing recovery after a
2072 * crash? For recovery, we need to check in advance whether we
2073 * can get read-write access to the device.
2074 */
2075
Mingming Cao617ba132006-10-11 01:20:53 -07002076 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002077 if (sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07002078 printk(KERN_INFO "EXT4-fs: INFO: recovery "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002079 "required on readonly filesystem.\n");
2080 if (really_read_only) {
Mingming Cao617ba132006-10-11 01:20:53 -07002081 printk(KERN_ERR "EXT4-fs: write access "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002082 "unavailable, cannot proceed.\n");
2083 return -EROFS;
2084 }
Mingming Cao617ba132006-10-11 01:20:53 -07002085 printk (KERN_INFO "EXT4-fs: write access will "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002086 "be enabled during recovery.\n");
2087 }
2088 }
2089
2090 if (journal_inum && journal_dev) {
Mingming Cao617ba132006-10-11 01:20:53 -07002091 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002092 "and inode journals!\n");
2093 return -EINVAL;
2094 }
2095
2096 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07002097 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002098 return -EINVAL;
2099 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07002100 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002101 return -EINVAL;
2102 }
2103
2104 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07002105 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002106 if (err) {
Mingming Cao617ba132006-10-11 01:20:53 -07002107 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
Mingming Caodab291a2006-10-11 01:21:01 -07002108 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002109 return err;
2110 }
2111 }
2112
Mingming Cao617ba132006-10-11 01:20:53 -07002113 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07002114 err = jbd2_journal_wipe(journal, !really_read_only);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002115 if (!err)
Mingming Caodab291a2006-10-11 01:21:01 -07002116 err = jbd2_journal_load(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002117
2118 if (err) {
Mingming Cao617ba132006-10-11 01:20:53 -07002119 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
Mingming Caodab291a2006-10-11 01:21:01 -07002120 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002121 return err;
2122 }
2123
Mingming Cao617ba132006-10-11 01:20:53 -07002124 EXT4_SB(sb)->s_journal = journal;
2125 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002126
2127 if (journal_devnum &&
2128 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2129 es->s_journal_dev = cpu_to_le32(journal_devnum);
2130 sb->s_dirt = 1;
2131
2132 /* Make sure we flush the recovery flag to disk. */
Mingming Cao617ba132006-10-11 01:20:53 -07002133 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002134 }
2135
2136 return 0;
2137}
2138
Mingming Cao617ba132006-10-11 01:20:53 -07002139static int ext4_create_journal(struct super_block * sb,
2140 struct ext4_super_block * es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002141 unsigned int journal_inum)
2142{
2143 journal_t *journal;
2144
2145 if (sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07002146 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002147 "create journal.\n");
2148 return -EROFS;
2149 }
2150
Mingming Cao617ba132006-10-11 01:20:53 -07002151 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002152 return -EINVAL;
2153
Mingming Cao617ba132006-10-11 01:20:53 -07002154 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002155 journal_inum);
2156
Mingming Caodab291a2006-10-11 01:21:01 -07002157 if (jbd2_journal_create(journal)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002158 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
Mingming Caodab291a2006-10-11 01:21:01 -07002159 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002160 return -EIO;
2161 }
2162
Mingming Cao617ba132006-10-11 01:20:53 -07002163 EXT4_SB(sb)->s_journal = journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002164
Mingming Cao617ba132006-10-11 01:20:53 -07002165 ext4_update_dynamic_rev(sb);
2166 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2167 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002168
2169 es->s_journal_inum = cpu_to_le32(journal_inum);
2170 sb->s_dirt = 1;
2171
2172 /* Make sure we flush the recovery flag to disk. */
Mingming Cao617ba132006-10-11 01:20:53 -07002173 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002174
2175 return 0;
2176}
2177
Mingming Cao617ba132006-10-11 01:20:53 -07002178static void ext4_commit_super (struct super_block * sb,
2179 struct ext4_super_block * es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002180 int sync)
2181{
Mingming Cao617ba132006-10-11 01:20:53 -07002182 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002183
2184 if (!sbh)
2185 return;
2186 es->s_wtime = cpu_to_le32(get_seconds());
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002187 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
Mingming Cao617ba132006-10-11 01:20:53 -07002188 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002189 BUFFER_TRACE(sbh, "marking dirty");
2190 mark_buffer_dirty(sbh);
2191 if (sync)
2192 sync_dirty_buffer(sbh);
2193}
2194
2195
2196/*
2197 * Have we just finished recovery? If so, and if we are mounting (or
2198 * remounting) the filesystem readonly, then we will end up with a
2199 * consistent fs on disk. Record that fact.
2200 */
Mingming Cao617ba132006-10-11 01:20:53 -07002201static void ext4_mark_recovery_complete(struct super_block * sb,
2202 struct ext4_super_block * es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002203{
Mingming Cao617ba132006-10-11 01:20:53 -07002204 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002205
Mingming Caodab291a2006-10-11 01:21:01 -07002206 jbd2_journal_lock_updates(journal);
2207 jbd2_journal_flush(journal);
Mingming Cao617ba132006-10-11 01:20:53 -07002208 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002209 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07002210 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002211 sb->s_dirt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002212 ext4_commit_super(sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002213 }
Mingming Caodab291a2006-10-11 01:21:01 -07002214 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002215}
2216
2217/*
2218 * If we are mounting (or read-write remounting) a filesystem whose journal
2219 * has recorded an error from a previous lifetime, move that error to the
2220 * main filesystem now.
2221 */
Mingming Cao617ba132006-10-11 01:20:53 -07002222static void ext4_clear_journal_err(struct super_block * sb,
2223 struct ext4_super_block * es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002224{
2225 journal_t *journal;
2226 int j_errno;
2227 const char *errstr;
2228
Mingming Cao617ba132006-10-11 01:20:53 -07002229 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002230
2231 /*
2232 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07002233 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002234 */
2235
Mingming Caodab291a2006-10-11 01:21:01 -07002236 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002237 if (j_errno) {
2238 char nbuf[16];
2239
Mingming Cao617ba132006-10-11 01:20:53 -07002240 errstr = ext4_decode_error(sb, j_errno, nbuf);
2241 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002242 "from previous mount: %s", errstr);
Mingming Cao617ba132006-10-11 01:20:53 -07002243 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002244 "filesystem check.");
2245
Mingming Cao617ba132006-10-11 01:20:53 -07002246 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2247 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2248 ext4_commit_super (sb, es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002249
Mingming Caodab291a2006-10-11 01:21:01 -07002250 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002251 }
2252}
2253
2254/*
2255 * Force the running and committing transactions to commit,
2256 * and wait on the commit.
2257 */
Mingming Cao617ba132006-10-11 01:20:53 -07002258int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002259{
2260 journal_t *journal;
2261 int ret;
2262
2263 if (sb->s_flags & MS_RDONLY)
2264 return 0;
2265
Mingming Cao617ba132006-10-11 01:20:53 -07002266 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002267 sb->s_dirt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002268 ret = ext4_journal_force_commit(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002269 return ret;
2270}
2271
2272/*
Mingming Cao617ba132006-10-11 01:20:53 -07002273 * Ext4 always journals updates to the superblock itself, so we don't
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002274 * have to propagate any other updates to the superblock on disk at this
2275 * point. Just start an async writeback to get the buffers on their way
2276 * to the disk.
2277 *
2278 * This implicitly triggers the writebehind on sync().
2279 */
2280
Mingming Cao617ba132006-10-11 01:20:53 -07002281static void ext4_write_super (struct super_block * sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002282{
2283 if (mutex_trylock(&sb->s_lock) != 0)
2284 BUG();
2285 sb->s_dirt = 0;
2286}
2287
Mingming Cao617ba132006-10-11 01:20:53 -07002288static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002289{
2290 tid_t target;
2291
2292 sb->s_dirt = 0;
Mingming Caodab291a2006-10-11 01:21:01 -07002293 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002294 if (wait)
Mingming Caodab291a2006-10-11 01:21:01 -07002295 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002296 }
2297 return 0;
2298}
2299
2300/*
2301 * LVM calls this function before a (read-only) snapshot is created. This
2302 * gives us a chance to flush the journal completely and mark the fs clean.
2303 */
Mingming Cao617ba132006-10-11 01:20:53 -07002304static void ext4_write_super_lockfs(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002305{
2306 sb->s_dirt = 0;
2307
2308 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002309 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002310
2311 /* Now we set up the journal barrier. */
Mingming Caodab291a2006-10-11 01:21:01 -07002312 jbd2_journal_lock_updates(journal);
2313 jbd2_journal_flush(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002314
2315 /* Journal blocked and flushed, clear needs_recovery flag. */
Mingming Cao617ba132006-10-11 01:20:53 -07002316 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2317 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002318 }
2319}
2320
2321/*
2322 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2323 * flag here, even though the filesystem is not technically dirty yet.
2324 */
Mingming Cao617ba132006-10-11 01:20:53 -07002325static void ext4_unlockfs(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002326{
2327 if (!(sb->s_flags & MS_RDONLY)) {
2328 lock_super(sb);
2329 /* Reser the needs_recovery flag before the fs is unlocked. */
Mingming Cao617ba132006-10-11 01:20:53 -07002330 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2331 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002332 unlock_super(sb);
Mingming Caodab291a2006-10-11 01:21:01 -07002333 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002334 }
2335}
2336
Mingming Cao617ba132006-10-11 01:20:53 -07002337static int ext4_remount (struct super_block * sb, int * flags, char * data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002338{
Mingming Cao617ba132006-10-11 01:20:53 -07002339 struct ext4_super_block * es;
2340 struct ext4_sb_info *sbi = EXT4_SB(sb);
2341 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002342 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07002343 struct ext4_mount_options old_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002344 int err;
2345#ifdef CONFIG_QUOTA
2346 int i;
2347#endif
2348
2349 /* Store the original options */
2350 old_sb_flags = sb->s_flags;
2351 old_opts.s_mount_opt = sbi->s_mount_opt;
2352 old_opts.s_resuid = sbi->s_resuid;
2353 old_opts.s_resgid = sbi->s_resgid;
2354 old_opts.s_commit_interval = sbi->s_commit_interval;
2355#ifdef CONFIG_QUOTA
2356 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2357 for (i = 0; i < MAXQUOTAS; i++)
2358 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2359#endif
2360
2361 /*
2362 * Allow the "check" option to be passed as a remount option.
2363 */
2364 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2365 err = -EINVAL;
2366 goto restore_opts;
2367 }
2368
Mingming Cao617ba132006-10-11 01:20:53 -07002369 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2370 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002371
2372 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07002373 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002374
2375 es = sbi->s_es;
2376
Mingming Cao617ba132006-10-11 01:20:53 -07002377 ext4_init_journal_params(sb, sbi->s_journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002378
2379 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002380 n_blocks_count > ext4_blocks_count(es)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002381 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002382 err = -EROFS;
2383 goto restore_opts;
2384 }
2385
2386 if (*flags & MS_RDONLY) {
2387 /*
2388 * First of all, the unconditional stuff we have to do
2389 * to disable replay of the journal when we next remount
2390 */
2391 sb->s_flags |= MS_RDONLY;
2392
2393 /*
2394 * OK, test if we are remounting a valid rw partition
2395 * readonly, and if so set the rdonly flag and then
2396 * mark the partition as valid again.
2397 */
Mingming Cao617ba132006-10-11 01:20:53 -07002398 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2399 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002400 es->s_state = cpu_to_le16(sbi->s_mount_state);
2401
Mingming Cao617ba132006-10-11 01:20:53 -07002402 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002403 } else {
2404 __le32 ret;
Mingming Cao617ba132006-10-11 01:20:53 -07002405 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2406 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2407 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002408 "remount RDWR because of unsupported "
2409 "optional features (%x).\n",
2410 sb->s_id, le32_to_cpu(ret));
2411 err = -EROFS;
2412 goto restore_opts;
2413 }
2414 /*
2415 * Mounting a RDONLY partition read-write, so reread
2416 * and store the current valid flag. (It may have
2417 * been changed by e2fsck since we originally mounted
2418 * the partition.)
2419 */
Mingming Cao617ba132006-10-11 01:20:53 -07002420 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002421 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07002422 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002423 goto restore_opts;
Mingming Cao617ba132006-10-11 01:20:53 -07002424 if (!ext4_setup_super (sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002425 sb->s_flags &= ~MS_RDONLY;
2426 }
2427 }
2428#ifdef CONFIG_QUOTA
2429 /* Release old quota file names */
2430 for (i = 0; i < MAXQUOTAS; i++)
2431 if (old_opts.s_qf_names[i] &&
2432 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2433 kfree(old_opts.s_qf_names[i]);
2434#endif
2435 return 0;
2436restore_opts:
2437 sb->s_flags = old_sb_flags;
2438 sbi->s_mount_opt = old_opts.s_mount_opt;
2439 sbi->s_resuid = old_opts.s_resuid;
2440 sbi->s_resgid = old_opts.s_resgid;
2441 sbi->s_commit_interval = old_opts.s_commit_interval;
2442#ifdef CONFIG_QUOTA
2443 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2444 for (i = 0; i < MAXQUOTAS; i++) {
2445 if (sbi->s_qf_names[i] &&
2446 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2447 kfree(sbi->s_qf_names[i]);
2448 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2449 }
2450#endif
2451 return err;
2452}
2453
Mingming Cao617ba132006-10-11 01:20:53 -07002454static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002455{
2456 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07002457 struct ext4_sb_info *sbi = EXT4_SB(sb);
2458 struct ext4_super_block *es = sbi->s_es;
2459 ext4_fsblk_t overhead;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002460 int i;
2461
2462 if (test_opt (sb, MINIX_DF))
2463 overhead = 0;
2464 else {
2465 unsigned long ngroups;
Mingming Cao617ba132006-10-11 01:20:53 -07002466 ngroups = EXT4_SB(sb)->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002467 smp_rmb();
2468
2469 /*
2470 * Compute the overhead (FS structures)
2471 */
2472
2473 /*
2474 * All of the blocks before first_data_block are
2475 * overhead
2476 */
2477 overhead = le32_to_cpu(es->s_first_data_block);
2478
2479 /*
2480 * Add the overhead attributed to the superblock and
2481 * block group descriptors. If the sparse superblocks
2482 * feature is turned on, then not all groups have this.
2483 */
2484 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002485 overhead += ext4_bg_has_super(sb, i) +
2486 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002487 cond_resched();
2488 }
2489
2490 /*
2491 * Every block group has an inode bitmap, a block
2492 * bitmap, and an inode table.
2493 */
Mingming Cao617ba132006-10-11 01:20:53 -07002494 overhead += (ngroups * (2 + EXT4_SB(sb)->s_itb_per_group));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002495 }
2496
Mingming Cao617ba132006-10-11 01:20:53 -07002497 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002498 buf->f_bsize = sb->s_blocksize;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002499 buf->f_blocks = ext4_blocks_count(es) - overhead;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002500 buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002501 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2502 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002503 buf->f_bavail = 0;
2504 buf->f_files = le32_to_cpu(es->s_inodes_count);
2505 buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07002506 buf->f_namelen = EXT4_NAME_LEN;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002507 return 0;
2508}
2509
2510/* Helper function for writing quotas on sync - we need to start transaction before quota file
2511 * is locked for write. Otherwise the are possible deadlocks:
2512 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07002513 * ext4_create() quota_sync()
Mingming Caodab291a2006-10-11 01:21:01 -07002514 * jbd2_journal_start() write_dquot()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002515 * DQUOT_INIT() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07002516 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002517 *
2518 */
2519
2520#ifdef CONFIG_QUOTA
2521
2522static inline struct inode *dquot_to_inode(struct dquot *dquot)
2523{
2524 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2525}
2526
Mingming Cao617ba132006-10-11 01:20:53 -07002527static int ext4_dquot_initialize(struct inode *inode, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002528{
2529 handle_t *handle;
2530 int ret, err;
2531
2532 /* We may create quota structure so we need to reserve enough blocks */
Mingming Cao617ba132006-10-11 01:20:53 -07002533 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002534 if (IS_ERR(handle))
2535 return PTR_ERR(handle);
2536 ret = dquot_initialize(inode, type);
Mingming Cao617ba132006-10-11 01:20:53 -07002537 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002538 if (!ret)
2539 ret = err;
2540 return ret;
2541}
2542
Mingming Cao617ba132006-10-11 01:20:53 -07002543static int ext4_dquot_drop(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002544{
2545 handle_t *handle;
2546 int ret, err;
2547
2548 /* We may delete quota structure so we need to reserve enough blocks */
Mingming Cao617ba132006-10-11 01:20:53 -07002549 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002550 if (IS_ERR(handle))
2551 return PTR_ERR(handle);
2552 ret = dquot_drop(inode);
Mingming Cao617ba132006-10-11 01:20:53 -07002553 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002554 if (!ret)
2555 ret = err;
2556 return ret;
2557}
2558
Mingming Cao617ba132006-10-11 01:20:53 -07002559static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002560{
2561 int ret, err;
2562 handle_t *handle;
2563 struct inode *inode;
2564
2565 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07002566 handle = ext4_journal_start(inode,
2567 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002568 if (IS_ERR(handle))
2569 return PTR_ERR(handle);
2570 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07002571 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002572 if (!ret)
2573 ret = err;
2574 return ret;
2575}
2576
Mingming Cao617ba132006-10-11 01:20:53 -07002577static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002578{
2579 int ret, err;
2580 handle_t *handle;
2581
Mingming Cao617ba132006-10-11 01:20:53 -07002582 handle = ext4_journal_start(dquot_to_inode(dquot),
2583 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002584 if (IS_ERR(handle))
2585 return PTR_ERR(handle);
2586 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07002587 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002588 if (!ret)
2589 ret = err;
2590 return ret;
2591}
2592
Mingming Cao617ba132006-10-11 01:20:53 -07002593static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002594{
2595 int ret, err;
2596 handle_t *handle;
2597
Mingming Cao617ba132006-10-11 01:20:53 -07002598 handle = ext4_journal_start(dquot_to_inode(dquot),
2599 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002600 if (IS_ERR(handle))
2601 return PTR_ERR(handle);
2602 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07002603 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002604 if (!ret)
2605 ret = err;
2606 return ret;
2607}
2608
Mingming Cao617ba132006-10-11 01:20:53 -07002609static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002610{
2611 /* Are we journalling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07002612 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2613 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002614 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07002615 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002616 } else {
2617 return dquot_mark_dquot_dirty(dquot);
2618 }
2619}
2620
Mingming Cao617ba132006-10-11 01:20:53 -07002621static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002622{
2623 int ret, err;
2624 handle_t *handle;
2625
2626 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07002627 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002628 if (IS_ERR(handle))
2629 return PTR_ERR(handle);
2630 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07002631 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002632 if (!ret)
2633 ret = err;
2634 return ret;
2635}
2636
2637/*
2638 * Turn on quotas during mount time - we need to find
2639 * the quota file and such...
2640 */
Mingming Cao617ba132006-10-11 01:20:53 -07002641static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002642{
Mingming Cao617ba132006-10-11 01:20:53 -07002643 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2644 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002645}
2646
2647/*
2648 * Standard function to be called on quota_on
2649 */
Mingming Cao617ba132006-10-11 01:20:53 -07002650static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002651 char *path)
2652{
2653 int err;
2654 struct nameidata nd;
2655
2656 if (!test_opt(sb, QUOTA))
2657 return -EINVAL;
2658 /* Not journalling quota? */
Mingming Cao617ba132006-10-11 01:20:53 -07002659 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2660 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002661 return vfs_quota_on(sb, type, format_id, path);
2662 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2663 if (err)
2664 return err;
2665 /* Quotafile not on the same filesystem? */
2666 if (nd.mnt->mnt_sb != sb) {
2667 path_release(&nd);
2668 return -EXDEV;
2669 }
2670 /* Quotafile not of fs root? */
2671 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2672 printk(KERN_WARNING
Mingming Cao617ba132006-10-11 01:20:53 -07002673 "EXT4-fs: Quota file not on filesystem root. "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002674 "Journalled quota will not work.\n");
2675 path_release(&nd);
2676 return vfs_quota_on(sb, type, format_id, path);
2677}
2678
2679/* Read data from quotafile - avoid pagecache and such because we cannot afford
2680 * acquiring the locks... As quota files are never truncated and quota code
2681 * itself serializes the operations (and noone else should touch the files)
2682 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07002683static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002684 size_t len, loff_t off)
2685{
2686 struct inode *inode = sb_dqopt(sb)->files[type];
Mingming Cao617ba132006-10-11 01:20:53 -07002687 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002688 int err = 0;
2689 int offset = off & (sb->s_blocksize - 1);
2690 int tocopy;
2691 size_t toread;
2692 struct buffer_head *bh;
2693 loff_t i_size = i_size_read(inode);
2694
2695 if (off > i_size)
2696 return 0;
2697 if (off+len > i_size)
2698 len = i_size-off;
2699 toread = len;
2700 while (toread > 0) {
2701 tocopy = sb->s_blocksize - offset < toread ?
2702 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07002703 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002704 if (err)
2705 return err;
2706 if (!bh) /* A hole? */
2707 memset(data, 0, tocopy);
2708 else
2709 memcpy(data, bh->b_data+offset, tocopy);
2710 brelse(bh);
2711 offset = 0;
2712 toread -= tocopy;
2713 data += tocopy;
2714 blk++;
2715 }
2716 return len;
2717}
2718
2719/* Write to quotafile (we know the transaction is already started and has
2720 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07002721static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002722 const char *data, size_t len, loff_t off)
2723{
2724 struct inode *inode = sb_dqopt(sb)->files[type];
Mingming Cao617ba132006-10-11 01:20:53 -07002725 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002726 int err = 0;
2727 int offset = off & (sb->s_blocksize - 1);
2728 int tocopy;
Mingming Cao617ba132006-10-11 01:20:53 -07002729 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002730 size_t towrite = len;
2731 struct buffer_head *bh;
2732 handle_t *handle = journal_current_handle();
2733
2734 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2735 while (towrite > 0) {
2736 tocopy = sb->s_blocksize - offset < towrite ?
2737 sb->s_blocksize - offset : towrite;
Mingming Cao617ba132006-10-11 01:20:53 -07002738 bh = ext4_bread(handle, inode, blk, 1, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002739 if (!bh)
2740 goto out;
2741 if (journal_quota) {
Mingming Cao617ba132006-10-11 01:20:53 -07002742 err = ext4_journal_get_write_access(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002743 if (err) {
2744 brelse(bh);
2745 goto out;
2746 }
2747 }
2748 lock_buffer(bh);
2749 memcpy(bh->b_data+offset, data, tocopy);
2750 flush_dcache_page(bh->b_page);
2751 unlock_buffer(bh);
2752 if (journal_quota)
Mingming Cao617ba132006-10-11 01:20:53 -07002753 err = ext4_journal_dirty_metadata(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002754 else {
2755 /* Always do at least ordered writes for quotas */
Mingming Cao617ba132006-10-11 01:20:53 -07002756 err = ext4_journal_dirty_data(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002757 mark_buffer_dirty(bh);
2758 }
2759 brelse(bh);
2760 if (err)
2761 goto out;
2762 offset = 0;
2763 towrite -= tocopy;
2764 data += tocopy;
2765 blk++;
2766 }
2767out:
2768 if (len == towrite)
2769 return err;
2770 if (inode->i_size < off+len-towrite) {
2771 i_size_write(inode, off+len-towrite);
Mingming Cao617ba132006-10-11 01:20:53 -07002772 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002773 }
2774 inode->i_version++;
2775 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07002776 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002777 mutex_unlock(&inode->i_mutex);
2778 return len - towrite;
2779}
2780
2781#endif
2782
Mingming Cao617ba132006-10-11 01:20:53 -07002783static int ext4_get_sb(struct file_system_type *fs_type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002784 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2785{
Mingming Cao617ba132006-10-11 01:20:53 -07002786 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002787}
2788
Mingming Cao617ba132006-10-11 01:20:53 -07002789static struct file_system_type ext4dev_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002790 .owner = THIS_MODULE,
Mingming Cao617ba132006-10-11 01:20:53 -07002791 .name = "ext4dev",
2792 .get_sb = ext4_get_sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002793 .kill_sb = kill_block_super,
2794 .fs_flags = FS_REQUIRES_DEV,
2795};
2796
Mingming Cao617ba132006-10-11 01:20:53 -07002797static int __init init_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002798{
Mingming Cao617ba132006-10-11 01:20:53 -07002799 int err = init_ext4_xattr();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002800 if (err)
2801 return err;
2802 err = init_inodecache();
2803 if (err)
2804 goto out1;
Mingming Cao617ba132006-10-11 01:20:53 -07002805 err = register_filesystem(&ext4dev_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002806 if (err)
2807 goto out;
2808 return 0;
2809out:
2810 destroy_inodecache();
2811out1:
Mingming Cao617ba132006-10-11 01:20:53 -07002812 exit_ext4_xattr();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002813 return err;
2814}
2815
Mingming Cao617ba132006-10-11 01:20:53 -07002816static void __exit exit_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002817{
Mingming Cao617ba132006-10-11 01:20:53 -07002818 unregister_filesystem(&ext4dev_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002819 destroy_inodecache();
Mingming Cao617ba132006-10-11 01:20:53 -07002820 exit_ext4_xattr();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002821}
2822
2823MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Mingming Cao617ba132006-10-11 01:20:53 -07002824MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002825MODULE_LICENSE("GPL");
Mingming Cao617ba132006-10-11 01:20:53 -07002826module_init(init_ext4_fs)
2827module_exit(exit_ext4_fs)