blob: dfe9bf503f8514c64f26de0f2a84259583b68f7b [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>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040023#include <linux/vmalloc.h>
Mingming Caodab291a2006-10-11 01:21:01 -070024#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070025#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070031#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070032#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040038#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040039#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040040#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040041#include <linux/crc16.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070042#include <asm/uaccess.h>
43
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040044#include "ext4.h"
45#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070046#include "xattr.h"
47#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040048#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070049
Theodore Ts'o9bffad12009-06-17 11:48:11 -040050#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040053struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040054static struct kset *ext4_kset;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040055
Mingming Cao617ba132006-10-11 01:20:53 -070056static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070057 unsigned long journal_devnum);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040058static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040059static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070063static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040064static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070065 char nbuf[16]);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040066static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080068static int ext4_unfreeze(struct super_block *sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static void ext4_write_super(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080070static int ext4_freeze(struct super_block *sb);
Jan Karaba69f9a2010-03-24 20:18:37 -040071static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070073
Jan Karaba69f9a2010-03-24 20:18:37 -040074#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75static struct file_system_type ext3_fs_type = {
76 .owner = THIS_MODULE,
77 .name = "ext3",
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
81};
82#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
83#else
84#define IS_EXT3_SB(sb) (0)
85#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070086
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070087ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070089{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040090 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070091 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -040092 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070093}
94
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070095ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070097{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040098 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070099 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700101}
102
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700103ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700105{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400106 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700109}
110
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500111__u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
113{
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500117}
118
119__u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500125}
126
127__u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500133}
134
135__u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500141}
142
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700143void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700145{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700149}
150
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700151void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700153{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700157}
158
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700159void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700161{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700165}
166
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500167void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
169{
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
173}
174
175void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
197}
198
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400199
200/* Just increment the non-pointer handle value */
201static handle_t *ext4_get_nojournal(void)
202{
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
205
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
207
208 ref_cnt++;
209 handle = (handle_t *)ref_cnt;
210
211 current->journal_info = handle;
212 return handle;
213}
214
215
216/* Decrement the non-pointer handle value */
217static void ext4_put_nojournal(handle_t *handle)
218{
219 unsigned long ref_cnt = (unsigned long)handle;
220
221 BUG_ON(ref_cnt == 0);
222
223 ref_cnt--;
224 handle = (handle_t *)ref_cnt;
225
226 current->journal_info = handle;
227}
228
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700229/*
Mingming Caodab291a2006-10-11 01:21:01 -0700230 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700231 *
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
235 * appropriate.
236 */
Mingming Cao617ba132006-10-11 01:20:53 -0700237handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700238{
239 journal_t *journal;
240
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
243
244 /* Special case here: if the journal has aborted behind our
245 * backs (eg. EIO in the commit thread), then we still need to
246 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700247 journal = EXT4_SB(sb)->s_journal;
Frank Mayhar03901312009-01-07 00:06:22 -0500248 if (journal) {
249 if (is_journal_aborted(journal)) {
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400250 ext4_abort(sb, __func__, "Detected aborted journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500251 return ERR_PTR(-EROFS);
252 }
253 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700254 }
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400255 return ext4_get_nojournal();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700256}
257
258/*
259 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700260 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700261 * that sync() will call the filesystem's write_super callback if
262 * appropriate.
263 */
Mingming Cao617ba132006-10-11 01:20:53 -0700264int __ext4_journal_stop(const char *where, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700265{
266 struct super_block *sb;
267 int err;
268 int rc;
269
Frank Mayhar03901312009-01-07 00:06:22 -0500270 if (!ext4_handle_valid(handle)) {
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400271 ext4_put_nojournal(handle);
Frank Mayhar03901312009-01-07 00:06:22 -0500272 return 0;
273 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700274 sb = handle->h_transaction->t_journal->j_private;
275 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700276 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700277
278 if (!err)
279 err = rc;
280 if (err)
Mingming Cao617ba132006-10-11 01:20:53 -0700281 __ext4_std_error(sb, where, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700282 return err;
283}
284
Mingming Cao617ba132006-10-11 01:20:53 -0700285void ext4_journal_abort_handle(const char *caller, const char *err_fn,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700286 struct buffer_head *bh, handle_t *handle, int err)
287{
288 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700289 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700290
Frank Mayhar03901312009-01-07 00:06:22 -0500291 BUG_ON(!ext4_handle_valid(handle));
292
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700293 if (bh)
294 BUFFER_TRACE(bh, "abort");
295
296 if (!handle->h_err)
297 handle->h_err = err;
298
299 if (is_handle_aborted(handle))
300 return;
301
302 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
303 caller, errstr, err_fn);
304
Mingming Caodab291a2006-10-11 01:21:01 -0700305 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700306}
307
308/* Deal with the reporting of failure conditions on a filesystem such as
309 * inconsistencies detected or read IO failures.
310 *
311 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700312 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700313 * write ordering constraints on the superblock which prevent us from
314 * writing it out straight away; and given that the journal is about to
315 * be aborted, we can't rely on the current, or future, transactions to
316 * write out the superblock safely.
317 *
Mingming Caodab291a2006-10-11 01:21:01 -0700318 * We'll just use the jbd2_journal_abort() error code to record an error in
Thadeu Lima de Souza Cascardod6b198b2010-01-17 19:10:07 -0200319 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700320 * that error until we've noted it down and cleared it.
321 */
322
Mingming Cao617ba132006-10-11 01:20:53 -0700323static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700324{
Mingming Cao617ba132006-10-11 01:20:53 -0700325 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700326
Mingming Cao617ba132006-10-11 01:20:53 -0700327 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
328 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700329
330 if (sb->s_flags & MS_RDONLY)
331 return;
332
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400333 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700334 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700335
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400336 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700337 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700338 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700339 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400340 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400341 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700342 sb->s_flags |= MS_RDONLY;
343 }
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400344 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700345 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700346 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700347 sb->s_id);
348}
349
Eric Sandeen12062dd2010-02-15 14:19:27 -0500350void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400351 const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700352{
353 va_list args;
354
355 va_start(args, fmt);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400356 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700357 vprintk(fmt, args);
358 printk("\n");
359 va_end(args);
360
Mingming Cao617ba132006-10-11 01:20:53 -0700361 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700362}
363
Frank Mayhar273df552010-03-02 11:46:09 -0500364void ext4_error_inode(const char *function, struct inode *inode,
365 const char *fmt, ...)
366{
367 va_list args;
368
369 va_start(args, fmt);
370 printk(KERN_CRIT "EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ",
371 inode->i_sb->s_id, function, inode->i_ino, current->comm);
372 vprintk(fmt, args);
373 printk("\n");
374 va_end(args);
375
376 ext4_handle_error(inode->i_sb);
377}
378
379void ext4_error_file(const char *function, struct file *file,
380 const char *fmt, ...)
381{
382 va_list args;
383 struct inode *inode = file->f_dentry->d_inode;
384 char pathname[80], *path;
385
386 va_start(args, fmt);
387 path = d_path(&(file->f_path), pathname, sizeof(pathname));
388 if (!path)
389 path = "(unknown)";
390 printk(KERN_CRIT
391 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ",
392 inode->i_sb->s_id, function, inode->i_ino, current->comm, path);
393 vprintk(fmt, args);
394 printk("\n");
395 va_end(args);
396
397 ext4_handle_error(inode->i_sb);
398}
399
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400400static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700401 char nbuf[16])
402{
403 char *errstr = NULL;
404
405 switch (errno) {
406 case -EIO:
407 errstr = "IO failure";
408 break;
409 case -ENOMEM:
410 errstr = "Out of memory";
411 break;
412 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400413 if (!sb || (EXT4_SB(sb)->s_journal &&
414 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700415 errstr = "Journal has aborted";
416 else
417 errstr = "Readonly filesystem";
418 break;
419 default:
420 /* If the caller passed in an extra buffer for unknown
421 * errors, textualise them now. Else we just return
422 * NULL. */
423 if (nbuf) {
424 /* Check for truncated error codes... */
425 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
426 errstr = nbuf;
427 }
428 break;
429 }
430
431 return errstr;
432}
433
Mingming Cao617ba132006-10-11 01:20:53 -0700434/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700435 * automatically and invokes the appropriate error response. */
436
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400437void __ext4_std_error(struct super_block *sb, const char *function, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700438{
439 char nbuf[16];
440 const char *errstr;
441
442 /* Special case: if the error is EROFS, and we're not already
443 * inside a transaction, then there's really no point in logging
444 * an error. */
445 if (errno == -EROFS && journal_current_handle() == NULL &&
446 (sb->s_flags & MS_RDONLY))
447 return;
448
Mingming Cao617ba132006-10-11 01:20:53 -0700449 errstr = ext4_decode_error(sb, errno, nbuf);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400450 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
451 sb->s_id, function, errstr);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700452
Mingming Cao617ba132006-10-11 01:20:53 -0700453 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700454}
455
456/*
Mingming Cao617ba132006-10-11 01:20:53 -0700457 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700458 * abort function may be used to deal with unrecoverable failures such
459 * as journal IO errors or ENOMEM at a critical moment in log management.
460 *
461 * We unconditionally force the filesystem into an ABORT|READONLY state,
462 * unless the error response on the fs has been set to panic in which
463 * case we take the easy way out and panic immediately.
464 */
465
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400466void ext4_abort(struct super_block *sb, const char *function,
467 const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700468{
469 va_list args;
470
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700471 va_start(args, fmt);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400472 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700473 vprintk(fmt, args);
474 printk("\n");
475 va_end(args);
476
477 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700478 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700479
480 if (sb->s_flags & MS_RDONLY)
481 return;
482
Eric Sandeenb31e1552009-06-04 17:36:36 -0400483 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Mingming Cao617ba132006-10-11 01:20:53 -0700484 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700485 sb->s_flags |= MS_RDONLY;
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400486 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400487 if (EXT4_SB(sb)->s_journal)
488 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700489}
490
Eric Sandeenb31e1552009-06-04 17:36:36 -0400491void ext4_msg (struct super_block * sb, const char *prefix,
492 const char *fmt, ...)
493{
494 va_list args;
495
496 va_start(args, fmt);
497 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
498 vprintk(fmt, args);
499 printk("\n");
500 va_end(args);
501}
502
Eric Sandeen12062dd2010-02-15 14:19:27 -0500503void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400504 const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700505{
506 va_list args;
507
508 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700509 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700510 sb->s_id, function);
511 vprintk(fmt, args);
512 printk("\n");
513 va_end(args);
514}
515
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500516void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400517 const char *function, const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500518__releases(bitlock)
519__acquires(bitlock)
520{
521 va_list args;
522 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
523
524 va_start(args, fmt);
525 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
526 vprintk(fmt, args);
527 printk("\n");
528 va_end(args);
529
530 if (test_opt(sb, ERRORS_CONT)) {
531 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
532 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400533 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500534 return;
535 }
536 ext4_unlock_group(sb, grp);
537 ext4_handle_error(sb);
538 /*
539 * We only get here in the ERRORS_RO case; relocking the group
540 * may be dangerous, but nothing bad will happen since the
541 * filesystem will have already been marked read/only and the
542 * journal has been aborted. We return 1 as a hint to callers
543 * who might what to use the return value from
544 * ext4_grp_locked_error() to distinguish beween the
545 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
546 * aggressively from the ext4 function in question, with a
547 * more appropriate error code.
548 */
549 ext4_lock_group(sb, grp);
550 return;
551}
552
Mingming Cao617ba132006-10-11 01:20:53 -0700553void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700554{
Mingming Cao617ba132006-10-11 01:20:53 -0700555 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700556
Mingming Cao617ba132006-10-11 01:20:53 -0700557 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700558 return;
559
Eric Sandeen12062dd2010-02-15 14:19:27 -0500560 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700561 "updating to rev %d because of new feature flag, "
562 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700563 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700564
Mingming Cao617ba132006-10-11 01:20:53 -0700565 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
566 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
567 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700568 /* leave es->s_feature_*compat flags alone */
569 /* es->s_uuid will be set by e2fsck if empty */
570
571 /*
572 * The rest of the superblock fields should be zero, and if not it
573 * means they are likely already in use, so leave them alone. We
574 * can leave it up to e2fsck to clean up any inconsistencies there.
575 */
576}
577
578/*
579 * Open the external journal device
580 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400581static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700582{
583 struct block_device *bdev;
584 char b[BDEVNAME_SIZE];
585
586 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
587 if (IS_ERR(bdev))
588 goto fail;
589 return bdev;
590
591fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400592 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700593 __bdevname(dev, b), PTR_ERR(bdev));
594 return NULL;
595}
596
597/*
598 * Release the journal device
599 */
Mingming Cao617ba132006-10-11 01:20:53 -0700600static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601{
602 bd_release(bdev);
Al Viro9a1c3542008-02-22 20:40:24 -0500603 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700604}
605
Mingming Cao617ba132006-10-11 01:20:53 -0700606static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700607{
608 struct block_device *bdev;
609 int ret = -ENODEV;
610
611 bdev = sbi->journal_bdev;
612 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700613 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700614 sbi->journal_bdev = NULL;
615 }
616 return ret;
617}
618
619static inline struct inode *orphan_list_entry(struct list_head *l)
620{
Mingming Cao617ba132006-10-11 01:20:53 -0700621 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700622}
623
Mingming Cao617ba132006-10-11 01:20:53 -0700624static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700625{
626 struct list_head *l;
627
Eric Sandeenb31e1552009-06-04 17:36:36 -0400628 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
629 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700630
631 printk(KERN_ERR "sb_info orphan list:\n");
632 list_for_each(l, &sbi->s_orphan) {
633 struct inode *inode = orphan_list_entry(l);
634 printk(KERN_ERR " "
635 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
636 inode->i_sb->s_id, inode->i_ino, inode,
637 inode->i_mode, inode->i_nlink,
638 NEXT_ORPHAN(inode));
639 }
640}
641
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400642static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700643{
Mingming Cao617ba132006-10-11 01:20:53 -0700644 struct ext4_sb_info *sbi = EXT4_SB(sb);
645 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400646 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700647
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400648 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
649
Mingming Cao4c0425f2009-09-28 15:48:41 -0400650 flush_workqueue(sbi->dio_unwritten_wq);
651 destroy_workqueue(sbi->dio_unwritten_wq);
652
Al Viroa9e220f2009-05-05 22:10:44 -0400653 lock_super(sb);
Christoph Hellwig6cfd0142009-05-05 15:40:36 +0200654 lock_kernel();
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200655 if (sb->s_dirt)
Christoph Hellwigebc1ac12009-05-11 23:35:03 +0200656 ext4_commit_super(sb, 1);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200657
Frank Mayhar03901312009-01-07 00:06:22 -0500658 if (sbi->s_journal) {
659 err = jbd2_journal_destroy(sbi->s_journal);
660 sbi->s_journal = NULL;
661 if (err < 0)
662 ext4_abort(sb, __func__,
663 "Couldn't clean up the journal");
664 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500665
666 ext4_release_system_zone(sb);
667 ext4_mb_release(sb);
668 ext4_ext_release(sb);
669 ext4_xattr_put_super(sb);
670
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700671 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700672 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700673 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400674 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700675 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400676 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400677 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400678 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400679 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700680
681 for (i = 0; i < sbi->s_gdb_count; i++)
682 brelse(sbi->s_group_desc[i]);
683 kfree(sbi->s_group_desc);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -0400684 if (is_vmalloc_addr(sbi->s_flex_groups))
685 vfree(sbi->s_flex_groups);
686 else
687 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700688 percpu_counter_destroy(&sbi->s_freeblocks_counter);
689 percpu_counter_destroy(&sbi->s_freeinodes_counter);
690 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400691 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700692 brelse(sbi->s_sbh);
693#ifdef CONFIG_QUOTA
694 for (i = 0; i < MAXQUOTAS; i++)
695 kfree(sbi->s_qf_names[i]);
696#endif
697
698 /* Debugging code just in case the in-memory inode orphan list
699 * isn't empty. The on-disk one can be non-empty if we've
700 * detected an error and taken the fs readonly, but the
701 * in-memory list had better be clean by this point. */
702 if (!list_empty(&sbi->s_orphan))
703 dump_orphan_list(sb, sbi);
704 J_ASSERT(list_empty(&sbi->s_orphan));
705
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700706 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700707 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
708 /*
709 * Invalidate the journal device's buffers. We don't want them
710 * floating about in memory - the physical journal device may
711 * hotswapped, and it breaks the `ro-after' testing code.
712 */
713 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700714 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700715 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700716 }
717 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400718 /*
719 * Now that we are completely done shutting down the
720 * superblock, we need to actually destroy the kobject.
721 */
722 unlock_kernel();
723 unlock_super(sb);
724 kobject_put(&sbi->s_kobj);
725 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500726 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700727 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700728}
729
Christoph Lametere18b8902006-12-06 20:33:20 -0800730static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700731
732/*
733 * Called inside transaction, so use GFP_NOFS
734 */
Mingming Cao617ba132006-10-11 01:20:53 -0700735static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700736{
Mingming Cao617ba132006-10-11 01:20:53 -0700737 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700738
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800739 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700740 if (!ei)
741 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400742
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700743 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400744 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700745 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500746 INIT_LIST_HEAD(&ei->i_prealloc_list);
747 spin_lock_init(&ei->i_prealloc_lock);
Frank Mayhar03901312009-01-07 00:06:22 -0500748 /*
749 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
750 * therefore it can be null here. Don't check it, just initialize
751 * jinode.
752 */
Jan Kara678aaf42008-07-11 19:27:31 -0400753 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
Mingming Caod2a17632008-07-14 17:52:37 -0400754 ei->i_reserved_data_blocks = 0;
755 ei->i_reserved_meta_blocks = 0;
756 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500757 ei->i_da_metadata_calc_len = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400758 ei->i_delalloc_reserved_flag = 0;
759 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300760#ifdef CONFIG_QUOTA
761 ei->i_reserved_quota = 0;
762#endif
Jiaying Zhangc7064ef2010-03-02 13:28:44 -0500763 INIT_LIST_HEAD(&ei->i_completed_io_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500764 spin_lock_init(&ei->i_completed_io_lock);
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400765 ei->cur_aio_dio = NULL;
Jan Karab436b9b2009-12-08 23:51:10 -0500766 ei->i_sync_tid = 0;
767 ei->i_datasync_tid = 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400768
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700769 return &ei->vfs_inode;
770}
771
Mingming Cao617ba132006-10-11 01:20:53 -0700772static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700773{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700774 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400775 ext4_msg(inode->i_sb, KERN_ERR,
776 "Inode %lu (%p): orphan list check failed!",
777 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700778 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
779 EXT4_I(inode), sizeof(struct ext4_inode_info),
780 true);
781 dump_stack();
782 }
Mingming Cao617ba132006-10-11 01:20:53 -0700783 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700784}
785
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700786static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700787{
Mingming Cao617ba132006-10-11 01:20:53 -0700788 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700789
Christoph Lametera35afb82007-05-16 22:10:57 -0700790 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400791#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700792 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700793#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500794 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700795 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700796}
797
798static int init_inodecache(void)
799{
Mingming Cao617ba132006-10-11 01:20:53 -0700800 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
801 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700802 0, (SLAB_RECLAIM_ACCOUNT|
803 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900804 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700805 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700806 return -ENOMEM;
807 return 0;
808}
809
810static void destroy_inodecache(void)
811{
Mingming Cao617ba132006-10-11 01:20:53 -0700812 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700813}
814
Mingming Cao617ba132006-10-11 01:20:53 -0700815static void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700816{
Christoph Hellwig9f754752010-03-03 09:05:05 -0500817 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400818 ext4_discard_preallocations(inode);
Frank Mayhar03901312009-01-07 00:06:22 -0500819 if (EXT4_JOURNAL(inode))
820 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
Jan Kara678aaf42008-07-11 19:27:31 -0400821 &EXT4_I(inode)->jinode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700822}
823
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400824static inline void ext4_show_quota_options(struct seq_file *seq,
825 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700826{
827#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700828 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700829
Jan Kara5a20bdf2009-11-30 23:58:32 +0100830 if (sbi->s_jquota_fmt) {
831 char *fmtname = "";
832
833 switch (sbi->s_jquota_fmt) {
834 case QFMT_VFS_OLD:
835 fmtname = "vfsold";
836 break;
837 case QFMT_VFS_V0:
838 fmtname = "vfsv0";
839 break;
840 case QFMT_VFS_V1:
841 fmtname = "vfsv1";
842 break;
843 }
844 seq_printf(seq, ",jqfmt=%s", fmtname);
845 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700846
847 if (sbi->s_qf_names[USRQUOTA])
848 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
849
850 if (sbi->s_qf_names[GRPQUOTA])
851 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
852
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500853 if (test_opt(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700854 seq_puts(seq, ",usrquota");
855
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500856 if (test_opt(sb, GRPQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700857 seq_puts(seq, ",grpquota");
858#endif
859}
860
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700861/*
862 * Show an option if
863 * - it's set to a non-default value OR
864 * - if the per-sb default is different from the global default
865 */
Mingming Cao617ba132006-10-11 01:20:53 -0700866static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700867{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500868 int def_errors;
869 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700870 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700871 struct ext4_sb_info *sbi = EXT4_SB(sb);
872 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700873
874 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500875 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700876
877 if (sbi->s_sb_block != 1)
878 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
879 if (test_opt(sb, MINIX_DF))
880 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500881 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700882 seq_puts(seq, ",grpid");
883 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
884 seq_puts(seq, ",nogrpid");
885 if (sbi->s_resuid != EXT4_DEF_RESUID ||
886 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
887 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
888 }
889 if (sbi->s_resgid != EXT4_DEF_RESGID ||
890 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
891 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
892 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500893 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700894 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500895 def_errors == EXT4_ERRORS_CONTINUE) {
896 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700897 }
898 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500899 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500900 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500901 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700902 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500903 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700904 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500905 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700906 seq_puts(seq, ",debug");
907 if (test_opt(sb, OLDALLOC))
908 seq_puts(seq, ",oldalloc");
Theodore Ts'o03010a32008-10-10 20:02:48 -0400909#ifdef CONFIG_EXT4_FS_XATTR
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500910 if (test_opt(sb, XATTR_USER) &&
911 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700912 seq_puts(seq, ",user_xattr");
913 if (!test_opt(sb, XATTR_USER) &&
914 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
915 seq_puts(seq, ",nouser_xattr");
916 }
917#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -0400918#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500919 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700920 seq_puts(seq, ",acl");
921 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
922 seq_puts(seq, ",noacl");
923#endif
Theodore Ts'o30773842009-01-03 20:27:38 -0500924 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700925 seq_printf(seq, ",commit=%u",
926 (unsigned) (sbi->s_commit_interval / HZ));
927 }
Theodore Ts'o30773842009-01-03 20:27:38 -0500928 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
929 seq_printf(seq, ",min_batch_time=%u",
930 (unsigned) sbi->s_min_batch_time);
931 }
932 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
933 seq_printf(seq, ",max_batch_time=%u",
934 (unsigned) sbi->s_min_batch_time);
935 }
936
Eric Sandeen571640c2008-05-26 12:29:46 -0400937 /*
938 * We're changing the default of barrier mount option, so
939 * let's always display its mount state so it's clear what its
940 * status is.
941 */
942 seq_puts(seq, ",barrier=");
943 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -0400944 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
945 seq_puts(seq, ",journal_async_commit");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700946 if (test_opt(sb, NOBH))
947 seq_puts(seq, ",nobh");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -0500948 if (test_opt(sb, I_VERSION))
949 seq_puts(seq, ",i_version");
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -0400950 if (!test_opt(sb, DELALLOC))
951 seq_puts(seq, ",nodelalloc");
952
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700953
Miklos Szeredicb45bbe2008-01-28 23:58:27 -0500954 if (sbi->s_stripe)
955 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500956 /*
957 * journal mode get enabled in different ways
958 * So just print the value even if we didn't specify it
959 */
Mingming Cao617ba132006-10-11 01:20:53 -0700960 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700961 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -0700962 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700963 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -0700964 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700965 seq_puts(seq, ",data=writeback");
966
Theodore Ts'o240799c2008-10-09 23:53:47 -0400967 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
968 seq_printf(seq, ",inode_readahead_blks=%u",
969 sbi->s_inode_readahead_blks);
970
Hidehiro Kawai5bf56832008-10-10 22:12:43 -0400971 if (test_opt(sb, DATA_ERR_ABORT))
972 seq_puts(seq, ",data_err=abort");
973
Theodore Ts'oafd46722009-03-16 23:12:23 -0400974 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -0400975 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -0400976
Eric Sandeen5328e632009-11-19 14:25:42 -0500977 if (test_opt(sb, DISCARD))
978 seq_puts(seq, ",discard");
979
Eric Sandeene3bb52a2009-11-19 14:28:50 -0500980 if (test_opt(sb, NOLOAD))
981 seq_puts(seq, ",norecovery");
982
Jiaying Zhang744692d2010-03-04 16:14:02 -0500983 if (test_opt(sb, DIOREAD_NOLOCK))
984 seq_puts(seq, ",dioread_nolock");
985
Mingming Cao617ba132006-10-11 01:20:53 -0700986 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400987
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700988 return 0;
989}
990
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700991static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400992 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700993{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700994 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700995
Mingming Cao617ba132006-10-11 01:20:53 -0700996 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700997 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700998 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700999 return ERR_PTR(-ESTALE);
1000
1001 /* iget isn't really right if the inode is currently unallocated!!
1002 *
Mingming Cao617ba132006-10-11 01:20:53 -07001003 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001004 * deleted, so we should be safe.
1005 *
1006 * Currently we don't know the generation for parent directory, so
1007 * a generation of 0 means "accept any"
1008 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001009 inode = ext4_iget(sb, ino);
1010 if (IS_ERR(inode))
1011 return ERR_CAST(inode);
1012 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001013 iput(inode);
1014 return ERR_PTR(-ESTALE);
1015 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001016
1017 return inode;
1018}
1019
1020static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001021 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001022{
1023 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1024 ext4_nfs_get_inode);
1025}
1026
1027static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001028 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001029{
1030 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1031 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001032}
1033
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001034/*
1035 * Try to release metadata pages (indirect blocks, directories) which are
1036 * mapped via the block device. Since these pages could have journal heads
1037 * which would prevent try_to_free_buffers() from freeing them, we must use
1038 * jbd2 layer's try_to_free_buffers() function to release them.
1039 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001040static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1041 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001042{
1043 journal_t *journal = EXT4_SB(sb)->s_journal;
1044
1045 WARN_ON(PageChecked(page));
1046 if (!page_has_buffers(page))
1047 return 0;
1048 if (journal)
1049 return jbd2_journal_try_to_free_buffers(journal, page,
1050 wait & ~__GFP_WAIT);
1051 return try_to_free_buffers(page);
1052}
1053
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001054#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001055#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001056#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001057
Mingming Cao617ba132006-10-11 01:20:53 -07001058static int ext4_write_dquot(struct dquot *dquot);
1059static int ext4_acquire_dquot(struct dquot *dquot);
1060static int ext4_release_dquot(struct dquot *dquot);
1061static int ext4_mark_dquot_dirty(struct dquot *dquot);
1062static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001063static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Christoph Hellwig307ae182010-05-19 07:16:43 -04001064 char *path);
Mingming Cao617ba132006-10-11 01:20:53 -07001065static int ext4_quota_on_mount(struct super_block *sb, int type);
1066static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001067 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001068static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001069 const char *data, size_t len, loff_t off);
1070
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001071static const struct dquot_operations ext4_quota_operations = {
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001072#ifdef CONFIG_QUOTA
Mingming Cao60e58e02009-01-22 18:13:05 +01001073 .get_reserved_space = ext4_get_reserved_space,
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001074#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001075 .write_dquot = ext4_write_dquot,
1076 .acquire_dquot = ext4_acquire_dquot,
1077 .release_dquot = ext4_release_dquot,
1078 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001079 .write_info = ext4_write_info,
1080 .alloc_dquot = dquot_alloc,
1081 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001082};
1083
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001084static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001085 .quota_on = ext4_quota_on,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001086 .quota_off = vfs_quota_off,
1087 .quota_sync = vfs_quota_sync,
1088 .get_info = vfs_get_dqinfo,
1089 .set_info = vfs_set_dqinfo,
1090 .get_dqblk = vfs_get_dqblk,
1091 .set_dqblk = vfs_set_dqblk
1092};
1093#endif
1094
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001095static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001096 .alloc_inode = ext4_alloc_inode,
1097 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001098 .write_inode = ext4_write_inode,
1099 .dirty_inode = ext4_dirty_inode,
1100 .delete_inode = ext4_delete_inode,
1101 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001102 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001103 .freeze_fs = ext4_freeze,
1104 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001105 .statfs = ext4_statfs,
1106 .remount_fs = ext4_remount,
1107 .clear_inode = ext4_clear_inode,
1108 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001109#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001110 .quota_read = ext4_quota_read,
1111 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001112#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001113 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001114};
1115
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001116static const struct super_operations ext4_nojournal_sops = {
1117 .alloc_inode = ext4_alloc_inode,
1118 .destroy_inode = ext4_destroy_inode,
1119 .write_inode = ext4_write_inode,
1120 .dirty_inode = ext4_dirty_inode,
1121 .delete_inode = ext4_delete_inode,
1122 .write_super = ext4_write_super,
1123 .put_super = ext4_put_super,
1124 .statfs = ext4_statfs,
1125 .remount_fs = ext4_remount,
1126 .clear_inode = ext4_clear_inode,
1127 .show_options = ext4_show_options,
1128#ifdef CONFIG_QUOTA
1129 .quota_read = ext4_quota_read,
1130 .quota_write = ext4_quota_write,
1131#endif
1132 .bdev_try_to_free_page = bdev_try_to_free_page,
1133};
1134
Christoph Hellwig39655162007-10-21 16:42:17 -07001135static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001136 .fh_to_dentry = ext4_fh_to_dentry,
1137 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001138 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001139};
1140
1141enum {
1142 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1143 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001144 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001145 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001146 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001147 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001148 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001149 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001150 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001151 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001152 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001153 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1154 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1155 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001156 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001157 Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001158 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001159 Opt_dioread_nolock, Opt_dioread_lock,
Eric Sandeen5328e632009-11-19 14:25:42 -05001160 Opt_discard, Opt_nodiscard,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001161};
1162
Steven Whitehousea447c092008-10-13 10:46:57 +01001163static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001164 {Opt_bsd_df, "bsddf"},
1165 {Opt_minix_df, "minixdf"},
1166 {Opt_grpid, "grpid"},
1167 {Opt_grpid, "bsdgroups"},
1168 {Opt_nogrpid, "nogrpid"},
1169 {Opt_nogrpid, "sysvgroups"},
1170 {Opt_resgid, "resgid=%u"},
1171 {Opt_resuid, "resuid=%u"},
1172 {Opt_sb, "sb=%u"},
1173 {Opt_err_cont, "errors=continue"},
1174 {Opt_err_panic, "errors=panic"},
1175 {Opt_err_ro, "errors=remount-ro"},
1176 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001177 {Opt_debug, "debug"},
1178 {Opt_oldalloc, "oldalloc"},
1179 {Opt_orlov, "orlov"},
1180 {Opt_user_xattr, "user_xattr"},
1181 {Opt_nouser_xattr, "nouser_xattr"},
1182 {Opt_acl, "acl"},
1183 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001184 {Opt_noload, "noload"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001185 {Opt_noload, "norecovery"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001186 {Opt_nobh, "nobh"},
1187 {Opt_bh, "bh"},
1188 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001189 {Opt_min_batch_time, "min_batch_time=%u"},
1190 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001191 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001192 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001193 {Opt_journal_checksum, "journal_checksum"},
1194 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001195 {Opt_abort, "abort"},
1196 {Opt_data_journal, "data=journal"},
1197 {Opt_data_ordered, "data=ordered"},
1198 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001199 {Opt_data_err_abort, "data_err=abort"},
1200 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001201 {Opt_offusrjquota, "usrjquota="},
1202 {Opt_usrjquota, "usrjquota=%s"},
1203 {Opt_offgrpjquota, "grpjquota="},
1204 {Opt_grpjquota, "grpjquota=%s"},
1205 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1206 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001207 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001208 {Opt_grpquota, "grpquota"},
1209 {Opt_noquota, "noquota"},
1210 {Opt_quota, "quota"},
1211 {Opt_usrquota, "usrquota"},
1212 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001213 {Opt_barrier, "barrier"},
1214 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001215 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001216 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001217 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001218 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001219 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001220 {Opt_block_validity, "block_validity"},
1221 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001222 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001223 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001224 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001225 {Opt_auto_da_alloc, "auto_da_alloc"},
1226 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001227 {Opt_dioread_nolock, "dioread_nolock"},
1228 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001229 {Opt_discard, "discard"},
1230 {Opt_nodiscard, "nodiscard"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001231 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001232};
1233
Mingming Cao617ba132006-10-11 01:20:53 -07001234static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001235{
Mingming Cao617ba132006-10-11 01:20:53 -07001236 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001237 char *options = (char *) *data;
1238
1239 if (!options || strncmp(options, "sb=", 3) != 0)
1240 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001241
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001242 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001243 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001244 sb_block = simple_strtoul(options, &options, 0);
1245 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001246 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001247 (char *) *data);
1248 return 1;
1249 }
1250 if (*options == ',')
1251 options++;
1252 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001253
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001254 return sb_block;
1255}
1256
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001257#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001258static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1259 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001260
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001261#ifdef CONFIG_QUOTA
1262static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1263{
1264 struct ext4_sb_info *sbi = EXT4_SB(sb);
1265 char *qname;
1266
1267 if (sb_any_quota_loaded(sb) &&
1268 !sbi->s_qf_names[qtype]) {
1269 ext4_msg(sb, KERN_ERR,
1270 "Cannot change journaled "
1271 "quota options when quota turned on");
1272 return 0;
1273 }
1274 qname = match_strdup(args);
1275 if (!qname) {
1276 ext4_msg(sb, KERN_ERR,
1277 "Not enough memory for storing quotafile name");
1278 return 0;
1279 }
1280 if (sbi->s_qf_names[qtype] &&
1281 strcmp(sbi->s_qf_names[qtype], qname)) {
1282 ext4_msg(sb, KERN_ERR,
1283 "%s quota file already specified", QTYPE2NAME(qtype));
1284 kfree(qname);
1285 return 0;
1286 }
1287 sbi->s_qf_names[qtype] = qname;
1288 if (strchr(sbi->s_qf_names[qtype], '/')) {
1289 ext4_msg(sb, KERN_ERR,
1290 "quotafile must be on filesystem root");
1291 kfree(sbi->s_qf_names[qtype]);
1292 sbi->s_qf_names[qtype] = NULL;
1293 return 0;
1294 }
1295 set_opt(sbi->s_mount_opt, QUOTA);
1296 return 1;
1297}
1298
1299static int clear_qf_name(struct super_block *sb, int qtype)
1300{
1301
1302 struct ext4_sb_info *sbi = EXT4_SB(sb);
1303
1304 if (sb_any_quota_loaded(sb) &&
1305 sbi->s_qf_names[qtype]) {
1306 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1307 " when quota turned on");
1308 return 0;
1309 }
1310 /*
1311 * The space will be released later when all options are confirmed
1312 * to be correct
1313 */
1314 sbi->s_qf_names[qtype] = NULL;
1315 return 1;
1316}
1317#endif
1318
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001319static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001320 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001321 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001322 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001323{
Mingming Cao617ba132006-10-11 01:20:53 -07001324 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001325 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001326 substring_t args[MAX_OPT_ARGS];
1327 int data_opt = 0;
1328 int option;
1329#ifdef CONFIG_QUOTA
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001330 int qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001331#endif
1332
1333 if (!options)
1334 return 1;
1335
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001336 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001337 int token;
1338 if (!*p)
1339 continue;
1340
Eric Sandeen15121c12010-02-15 20:17:55 -05001341 /*
1342 * Initialize args struct so we know whether arg was
1343 * found; some options take optional arguments.
1344 */
1345 args[0].to = args[0].from = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001346 token = match_token(p, tokens, args);
1347 switch (token) {
1348 case Opt_bsd_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001349 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001350 clear_opt(sbi->s_mount_opt, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001351 break;
1352 case Opt_minix_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001353 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001354 set_opt(sbi->s_mount_opt, MINIX_DF);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001355
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001356 break;
1357 case Opt_grpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001358 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001359 set_opt(sbi->s_mount_opt, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001360
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001361 break;
1362 case Opt_nogrpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001363 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001364 clear_opt(sbi->s_mount_opt, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001365
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001366 break;
1367 case Opt_resuid:
1368 if (match_int(&args[0], &option))
1369 return 0;
1370 sbi->s_resuid = option;
1371 break;
1372 case Opt_resgid:
1373 if (match_int(&args[0], &option))
1374 return 0;
1375 sbi->s_resgid = option;
1376 break;
1377 case Opt_sb:
1378 /* handled by get_sb_block() instead of here */
1379 /* *sb_block = match_int(&args[0]); */
1380 break;
1381 case Opt_err_panic:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001382 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1383 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1384 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001385 break;
1386 case Opt_err_ro:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001387 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1388 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1389 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001390 break;
1391 case Opt_err_cont:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001392 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1393 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1394 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001395 break;
1396 case Opt_nouid32:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001397 set_opt(sbi->s_mount_opt, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001398 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001399 case Opt_debug:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001400 set_opt(sbi->s_mount_opt, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001401 break;
1402 case Opt_oldalloc:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001403 set_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001404 break;
1405 case Opt_orlov:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001406 clear_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001407 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001408#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001409 case Opt_user_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001410 set_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001411 break;
1412 case Opt_nouser_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001413 clear_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001414 break;
1415#else
1416 case Opt_user_xattr:
1417 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001418 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001419 break;
1420#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001421#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001422 case Opt_acl:
1423 set_opt(sbi->s_mount_opt, POSIX_ACL);
1424 break;
1425 case Opt_noacl:
1426 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1427 break;
1428#else
1429 case Opt_acl:
1430 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001431 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001432 break;
1433#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001434 case Opt_journal_update:
1435 /* @@@ FIXME */
1436 /* Eventually we will want to be able to create
1437 a journal file here. For now, only allow the
1438 user to specify an existing inode to be the
1439 journal file. */
1440 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001441 ext4_msg(sb, KERN_ERR,
1442 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001443 return 0;
1444 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001445 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001446 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001447 case Opt_journal_dev:
1448 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001449 ext4_msg(sb, KERN_ERR,
1450 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001451 return 0;
1452 }
1453 if (match_int(&args[0], &option))
1454 return 0;
1455 *journal_devnum = option;
1456 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001457 case Opt_journal_checksum:
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001458 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1459 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001460 case Opt_journal_async_commit:
1461 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001462 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001463 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001464 case Opt_noload:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001465 set_opt(sbi->s_mount_opt, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001466 break;
1467 case Opt_commit:
1468 if (match_int(&args[0], &option))
1469 return 0;
1470 if (option < 0)
1471 return 0;
1472 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001473 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001474 sbi->s_commit_interval = HZ * option;
1475 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001476 case Opt_max_batch_time:
1477 if (match_int(&args[0], &option))
1478 return 0;
1479 if (option < 0)
1480 return 0;
1481 if (option == 0)
1482 option = EXT4_DEF_MAX_BATCH_TIME;
1483 sbi->s_max_batch_time = option;
1484 break;
1485 case Opt_min_batch_time:
1486 if (match_int(&args[0], &option))
1487 return 0;
1488 if (option < 0)
1489 return 0;
1490 sbi->s_min_batch_time = option;
1491 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001492 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001493 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001494 goto datacheck;
1495 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001496 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001497 goto datacheck;
1498 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001499 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001500 datacheck:
1501 if (is_remount) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001502 if (test_opt(sb, DATA_FLAGS) != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001503 ext4_msg(sb, KERN_ERR,
1504 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001505 return 0;
1506 }
1507 } else {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001508 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001509 sbi->s_mount_opt |= data_opt;
1510 }
1511 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001512 case Opt_data_err_abort:
1513 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1514 break;
1515 case Opt_data_err_ignore:
1516 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1517 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001518#ifdef CONFIG_QUOTA
1519 case Opt_usrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001520 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1521 return 0;
1522 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001523 case Opt_grpjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001524 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001525 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001526 break;
1527 case Opt_offusrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001528 if (!clear_qf_name(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001529 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001530 break;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001531 case Opt_offgrpjquota:
1532 if (!clear_qf_name(sb, GRPQUOTA))
1533 return 0;
1534 break;
1535
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001536 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001537 qfmt = QFMT_VFS_OLD;
1538 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001539 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001540 qfmt = QFMT_VFS_V0;
Jan Kara5a20bdf2009-11-30 23:58:32 +01001541 goto set_qf_format;
1542 case Opt_jqfmt_vfsv1:
1543 qfmt = QFMT_VFS_V1;
Jan Karadfc5d032008-05-13 19:11:51 -04001544set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001545 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001546 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001547 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001548 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001549 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001550 return 0;
1551 }
1552 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001553 break;
1554 case Opt_quota:
1555 case Opt_usrquota:
1556 set_opt(sbi->s_mount_opt, QUOTA);
1557 set_opt(sbi->s_mount_opt, USRQUOTA);
1558 break;
1559 case Opt_grpquota:
1560 set_opt(sbi->s_mount_opt, QUOTA);
1561 set_opt(sbi->s_mount_opt, GRPQUOTA);
1562 break;
1563 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001564 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001565 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1566 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001567 return 0;
1568 }
1569 clear_opt(sbi->s_mount_opt, QUOTA);
1570 clear_opt(sbi->s_mount_opt, USRQUOTA);
1571 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1572 break;
1573#else
1574 case Opt_quota:
1575 case Opt_usrquota:
1576 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001577 ext4_msg(sb, KERN_ERR,
1578 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001579 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001580 case Opt_usrjquota:
1581 case Opt_grpjquota:
1582 case Opt_offusrjquota:
1583 case Opt_offgrpjquota:
1584 case Opt_jqfmt_vfsold:
1585 case Opt_jqfmt_vfsv0:
Jan Kara5a20bdf2009-11-30 23:58:32 +01001586 case Opt_jqfmt_vfsv1:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001587 ext4_msg(sb, KERN_ERR,
1588 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001589 break;
1590 case Opt_noquota:
1591 break;
1592#endif
1593 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001594 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001595 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001596 case Opt_nobarrier:
1597 clear_opt(sbi->s_mount_opt, BARRIER);
1598 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001599 case Opt_barrier:
Eric Sandeen15121c12010-02-15 20:17:55 -05001600 if (args[0].from) {
1601 if (match_int(&args[0], &option))
1602 return 0;
1603 } else
1604 option = 1; /* No argument, default to 1 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001605 if (option)
1606 set_opt(sbi->s_mount_opt, BARRIER);
1607 else
1608 clear_opt(sbi->s_mount_opt, BARRIER);
1609 break;
1610 case Opt_ignore:
1611 break;
1612 case Opt_resize:
1613 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001614 ext4_msg(sb, KERN_ERR,
1615 "resize option only available "
1616 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001617 return 0;
1618 }
1619 if (match_int(&args[0], &option) != 0)
1620 return 0;
1621 *n_blocks_count = option;
1622 break;
1623 case Opt_nobh:
1624 set_opt(sbi->s_mount_opt, NOBH);
1625 break;
1626 case Opt_bh:
1627 clear_opt(sbi->s_mount_opt, NOBH);
1628 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001629 case Opt_i_version:
1630 set_opt(sbi->s_mount_opt, I_VERSION);
1631 sb->s_flags |= MS_I_VERSION;
1632 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001633 case Opt_nodelalloc:
1634 clear_opt(sbi->s_mount_opt, DELALLOC);
1635 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001636 case Opt_stripe:
1637 if (match_int(&args[0], &option))
1638 return 0;
1639 if (option < 0)
1640 return 0;
1641 sbi->s_stripe = option;
1642 break;
Alex Tomas64769242008-07-11 19:27:31 -04001643 case Opt_delalloc:
1644 set_opt(sbi->s_mount_opt, DELALLOC);
1645 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001646 case Opt_block_validity:
1647 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1648 break;
1649 case Opt_noblock_validity:
1650 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1651 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001652 case Opt_inode_readahead_blks:
1653 if (match_int(&args[0], &option))
1654 return 0;
1655 if (option < 0 || option > (1 << 30))
1656 return 0;
Theodore Ts'of7c43952009-04-24 23:31:59 -04001657 if (!is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001658 ext4_msg(sb, KERN_ERR,
1659 "EXT4-fs: inode_readahead_blks"
1660 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001661 return 0;
1662 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001663 sbi->s_inode_readahead_blks = option;
1664 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001665 case Opt_journal_ioprio:
1666 if (match_int(&args[0], &option))
1667 return 0;
1668 if (option < 0 || option > 7)
1669 break;
1670 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1671 option);
1672 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001673 case Opt_noauto_da_alloc:
1674 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1675 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001676 case Opt_auto_da_alloc:
Eric Sandeen15121c12010-02-15 20:17:55 -05001677 if (args[0].from) {
1678 if (match_int(&args[0], &option))
1679 return 0;
1680 } else
1681 option = 1; /* No argument, default to 1 */
Theodore Ts'oafd46722009-03-16 23:12:23 -04001682 if (option)
1683 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1684 else
1685 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1686 break;
Eric Sandeen5328e632009-11-19 14:25:42 -05001687 case Opt_discard:
1688 set_opt(sbi->s_mount_opt, DISCARD);
1689 break;
1690 case Opt_nodiscard:
1691 clear_opt(sbi->s_mount_opt, DISCARD);
1692 break;
Jiaying Zhang744692d2010-03-04 16:14:02 -05001693 case Opt_dioread_nolock:
1694 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1695 break;
1696 case Opt_dioread_lock:
1697 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1698 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001699 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001700 ext4_msg(sb, KERN_ERR,
1701 "Unrecognized mount option \"%s\" "
1702 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001703 return 0;
1704 }
1705 }
1706#ifdef CONFIG_QUOTA
1707 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001708 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001709 clear_opt(sbi->s_mount_opt, USRQUOTA);
1710
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001711 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001712 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1713
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001714 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001715 ext4_msg(sb, KERN_ERR, "old and new quota "
1716 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001717 return 0;
1718 }
1719
1720 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001721 ext4_msg(sb, KERN_ERR, "journaled quota format "
1722 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001723 return 0;
1724 }
1725 } else {
1726 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001727 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001728 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001729 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001730 return 0;
1731 }
1732 }
1733#endif
1734 return 1;
1735}
1736
Mingming Cao617ba132006-10-11 01:20:53 -07001737static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001738 int read_only)
1739{
Mingming Cao617ba132006-10-11 01:20:53 -07001740 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001741 int res = 0;
1742
Mingming Cao617ba132006-10-11 01:20:53 -07001743 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001744 ext4_msg(sb, KERN_ERR, "revision level too high, "
1745 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001746 res = MS_RDONLY;
1747 }
1748 if (read_only)
1749 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001750 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001751 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1752 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001753 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001754 ext4_msg(sb, KERN_WARNING,
1755 "warning: mounting fs with errors, "
1756 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001757 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1758 le16_to_cpu(es->s_mnt_count) >=
1759 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001760 ext4_msg(sb, KERN_WARNING,
1761 "warning: maximal mount count reached, "
1762 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001763 else if (le32_to_cpu(es->s_checkinterval) &&
1764 (le32_to_cpu(es->s_lastcheck) +
1765 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001766 ext4_msg(sb, KERN_WARNING,
1767 "warning: checktime reached, "
1768 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001769 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001770 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001771 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001772 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001773 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001774 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001775 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001776 if (sbi->s_journal)
1777 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001778
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001779 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001780 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001781 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'o7f4520c2009-06-13 10:09:41 -04001782 "bpg=%lu, ipg=%lu, mo=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001783 sb->s_blocksize,
1784 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001785 EXT4_BLOCKS_PER_GROUP(sb),
1786 EXT4_INODES_PER_GROUP(sb),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001787 sbi->s_mount_opt);
1788
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001789 return res;
1790}
1791
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001792static int ext4_fill_flex_info(struct super_block *sb)
1793{
1794 struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001796 ext4_group_t flex_group_count;
1797 ext4_group_t flex_group;
1798 int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001799 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001800 int i;
1801
Theodore Ts'o503358a2009-11-23 07:24:46 -05001802 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1803 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1804
1805 if (groups_per_flex < 2) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001806 sbi->s_log_groups_per_flex = 0;
1807 return 1;
1808 }
1809
Frederic Bohec62a11f2008-09-08 10:20:24 -04001810 /* We allocate both existing and potentially added groups */
1811 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05001812 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1813 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001814 size = flex_group_count * sizeof(struct flex_groups);
1815 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1816 if (sbi->s_flex_groups == NULL) {
1817 sbi->s_flex_groups = vmalloc(size);
1818 if (sbi->s_flex_groups)
1819 memset(sbi->s_flex_groups, 0, size);
1820 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001821 if (sbi->s_flex_groups == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001822 ext4_msg(sb, KERN_ERR, "not enough memory for "
1823 "%u flex groups", flex_group_count);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001824 goto failed;
1825 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001826
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001827 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001828 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001829
1830 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001831 atomic_add(ext4_free_inodes_count(sb, gdp),
1832 &sbi->s_flex_groups[flex_group].free_inodes);
1833 atomic_add(ext4_free_blks_count(sb, gdp),
1834 &sbi->s_flex_groups[flex_group].free_blocks);
1835 atomic_add(ext4_used_dirs_count(sb, gdp),
1836 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001837 }
1838
1839 return 1;
1840failed:
1841 return 0;
1842}
1843
Andreas Dilger717d50e2007-10-16 18:38:25 -04001844__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1845 struct ext4_group_desc *gdp)
1846{
1847 __u16 crc = 0;
1848
1849 if (sbi->s_es->s_feature_ro_compat &
1850 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1851 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1852 __le32 le_group = cpu_to_le32(block_group);
1853
1854 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1855 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1856 crc = crc16(crc, (__u8 *)gdp, offset);
1857 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1858 /* for checksum of struct ext4_group_desc do the rest...*/
1859 if ((sbi->s_es->s_feature_incompat &
1860 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1861 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1862 crc = crc16(crc, (__u8 *)gdp + offset,
1863 le16_to_cpu(sbi->s_es->s_desc_size) -
1864 offset);
1865 }
1866
1867 return cpu_to_le16(crc);
1868}
1869
1870int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1871 struct ext4_group_desc *gdp)
1872{
1873 if ((sbi->s_es->s_feature_ro_compat &
1874 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1875 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1876 return 0;
1877
1878 return 1;
1879}
1880
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001881/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001882static int ext4_check_descriptors(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001883{
Mingming Cao617ba132006-10-11 01:20:53 -07001884 struct ext4_sb_info *sbi = EXT4_SB(sb);
1885 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1886 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001887 ext4_fsblk_t block_bitmap;
1888 ext4_fsblk_t inode_bitmap;
1889 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04001890 int flexbg_flag = 0;
Avantika Mathurfd2d4292008-01-28 23:58:27 -05001891 ext4_group_t i;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001892
Jose R. Santosce421582007-10-16 18:38:25 -04001893 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1894 flexbg_flag = 1;
1895
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001896 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001897
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001898 for (i = 0; i < sbi->s_groups_count; i++) {
1899 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1900
Jose R. Santosce421582007-10-16 18:38:25 -04001901 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001902 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001903 else
1904 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07001905 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001906
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001907 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001908 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001909 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001910 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001911 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001912 return 0;
1913 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001914 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001915 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001916 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001917 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001918 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001919 return 0;
1920 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001921 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001922 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001923 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001924 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001925 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001926 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001927 return 0;
1928 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001929 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001930 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001931 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1932 "Checksum for group %u failed (%u!=%u)",
1933 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1934 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04001935 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001936 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04001937 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04001938 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04001939 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001940 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04001941 if (!flexbg_flag)
1942 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001943 }
1944
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001945 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001946 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001947 return 1;
1948}
1949
Mingming Cao617ba132006-10-11 01:20:53 -07001950/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001951 * the superblock) which were deleted from all directories, but held open by
1952 * a process at the time of a crash. We walk the list and try to delete these
1953 * inodes at recovery time (only with a read-write filesystem).
1954 *
1955 * In order to keep the orphan inode chain consistent during traversal (in
1956 * case of crash during recovery), we link each inode into the superblock
1957 * orphan list_head and handle it the same way as an inode deletion during
1958 * normal operation (which journals the operations for us).
1959 *
1960 * We only do an iget() and an iput() on each inode, which is very safe if we
1961 * accidentally point at an in-use or already deleted inode. The worst that
1962 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07001963 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001964 * e2fsck was run on this filesystem, and it must have already done the orphan
1965 * inode cleanup for us, so we can safely abort without any further action.
1966 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001967static void ext4_orphan_cleanup(struct super_block *sb,
1968 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001969{
1970 unsigned int s_flags = sb->s_flags;
1971 int nr_orphans = 0, nr_truncates = 0;
1972#ifdef CONFIG_QUOTA
1973 int i;
1974#endif
1975 if (!es->s_last_orphan) {
1976 jbd_debug(4, "no orphan inodes to clean up\n");
1977 return;
1978 }
1979
Eric Sandeena8f48a92006-12-06 20:40:13 -08001980 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001981 ext4_msg(sb, KERN_ERR, "write access "
1982 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08001983 return;
1984 }
1985
Mingming Cao617ba132006-10-11 01:20:53 -07001986 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001987 if (es->s_last_orphan)
1988 jbd_debug(1, "Errors on filesystem, "
1989 "clearing orphan list.\n");
1990 es->s_last_orphan = 0;
1991 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1992 return;
1993 }
1994
1995 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001996 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001997 sb->s_flags &= ~MS_RDONLY;
1998 }
1999#ifdef CONFIG_QUOTA
2000 /* Needed for iput() to work correctly and not trash data */
2001 sb->s_flags |= MS_ACTIVE;
2002 /* Turn on quotas so that they are updated correctly */
2003 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002004 if (EXT4_SB(sb)->s_qf_names[i]) {
2005 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002006 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002007 ext4_msg(sb, KERN_ERR,
2008 "Cannot turn on journaled "
2009 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002010 }
2011 }
2012#endif
2013
2014 while (es->s_last_orphan) {
2015 struct inode *inode;
2016
Josef Bacik97bd42b2008-04-29 22:04:56 -04002017 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2018 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002019 es->s_last_orphan = 0;
2020 break;
2021 }
2022
Mingming Cao617ba132006-10-11 01:20:53 -07002023 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002024 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002025 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002026 ext4_msg(sb, KERN_DEBUG,
2027 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002028 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002029 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002030 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002031 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002032 nr_truncates++;
2033 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002034 ext4_msg(sb, KERN_DEBUG,
2035 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002036 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002037 jbd_debug(2, "deleting unreferenced inode %lu\n",
2038 inode->i_ino);
2039 nr_orphans++;
2040 }
2041 iput(inode); /* The delete magic happens here! */
2042 }
2043
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002044#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002045
2046 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002047 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2048 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002049 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002050 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2051 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002052#ifdef CONFIG_QUOTA
2053 /* Turn quotas off */
2054 for (i = 0; i < MAXQUOTAS; i++) {
2055 if (sb_dqopt(sb)->files[i])
Christoph Hellwig307ae182010-05-19 07:16:43 -04002056 vfs_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002057 }
2058#endif
2059 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2060}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002061
Eric Sandeencd2291a2008-01-28 23:58:27 -05002062/*
2063 * Maximal extent format file size.
2064 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2065 * extent format containers, within a sector_t, and within i_blocks
2066 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2067 * so that won't be a limiting factor.
2068 *
2069 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2070 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002071static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002072{
2073 loff_t res;
2074 loff_t upper_limit = MAX_LFS_FILESIZE;
2075
2076 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002077 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002078 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002079 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002080 * i_block represent total blocks in 512 bytes
2081 * 32 == size of vfs inode i_blocks * 8
2082 */
2083 upper_limit = (1LL << 32) - 1;
2084
2085 /* total blocks in file system block size */
2086 upper_limit >>= (blkbits - 9);
2087 upper_limit <<= blkbits;
2088 }
2089
2090 /* 32-bit extent-start container, ee_block */
2091 res = 1LL << 32;
2092 res <<= blkbits;
2093 res -= 1;
2094
2095 /* Sanity check against vm- & vfs- imposed limits */
2096 if (res > upper_limit)
2097 res = upper_limit;
2098
2099 return res;
2100}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002101
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002102/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002103 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002104 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2105 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002106 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002107static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002108{
Mingming Cao617ba132006-10-11 01:20:53 -07002109 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002110 int meta_blocks;
2111 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002112 /* This is calculated to be the largest file size for a dense, block
2113 * mapped file such that the file's total number of 512-byte sectors,
2114 * including data and all indirect blocks, does not exceed (2^48 - 1).
2115 *
2116 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2117 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002118 */
2119
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002120 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002121 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002122 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002123 * the inode i_block field represents total file blocks in
2124 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002125 */
2126 upper_limit = (1LL << 32) - 1;
2127
2128 /* total blocks in file system block size */
2129 upper_limit >>= (bits - 9);
2130
2131 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002132 /*
2133 * We use 48 bit ext4_inode i_blocks
2134 * With EXT4_HUGE_FILE_FL set the i_blocks
2135 * represent total number of blocks in
2136 * file system block size
2137 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002138 upper_limit = (1LL << 48) - 1;
2139
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002140 }
2141
2142 /* indirect blocks */
2143 meta_blocks = 1;
2144 /* double indirect blocks */
2145 meta_blocks += 1 + (1LL << (bits-2));
2146 /* tripple indirect blocks */
2147 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2148
2149 upper_limit -= meta_blocks;
2150 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002151
2152 res += 1LL << (bits-2);
2153 res += 1LL << (2*(bits-2));
2154 res += 1LL << (3*(bits-2));
2155 res <<= bits;
2156 if (res > upper_limit)
2157 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002158
2159 if (res > MAX_LFS_FILESIZE)
2160 res = MAX_LFS_FILESIZE;
2161
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002162 return res;
2163}
2164
Mingming Cao617ba132006-10-11 01:20:53 -07002165static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002166 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002167{
Mingming Cao617ba132006-10-11 01:20:53 -07002168 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002169 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002170 int has_super = 0;
2171
2172 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2173
Mingming Cao617ba132006-10-11 01:20:53 -07002174 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002175 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002176 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002177 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002178 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002179 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002180
Mingming Cao617ba132006-10-11 01:20:53 -07002181 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002182}
2183
Alex Tomasc9de5602008-01-29 00:19:52 -05002184/**
2185 * ext4_get_stripe_size: Get the stripe size.
2186 * @sbi: In memory super block info
2187 *
2188 * If we have specified it via mount option, then
2189 * use the mount option value. If the value specified at mount time is
2190 * greater than the blocks per group use the super block value.
2191 * If the super block value is greater than blocks per group return 0.
2192 * Allocator needs it be less than blocks per group.
2193 *
2194 */
2195static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2196{
2197 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2198 unsigned long stripe_width =
2199 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2200
2201 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2202 return sbi->s_stripe;
2203
2204 if (stripe_width <= sbi->s_blocks_per_group)
2205 return stripe_width;
2206
2207 if (stride <= sbi->s_blocks_per_group)
2208 return stride;
2209
2210 return 0;
2211}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002212
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002213/* sysfs supprt */
2214
2215struct ext4_attr {
2216 struct attribute attr;
2217 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2218 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2219 const char *, size_t);
2220 int offset;
2221};
2222
2223static int parse_strtoul(const char *buf,
2224 unsigned long max, unsigned long *value)
2225{
2226 char *endp;
2227
André Goddard Rosae7d28602009-12-14 18:01:06 -08002228 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2229 endp = skip_spaces(endp);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002230 if (*endp || *value > max)
2231 return -EINVAL;
2232
2233 return 0;
2234}
2235
2236static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2237 struct ext4_sb_info *sbi,
2238 char *buf)
2239{
2240 return snprintf(buf, PAGE_SIZE, "%llu\n",
2241 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2242}
2243
2244static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2245 struct ext4_sb_info *sbi, char *buf)
2246{
2247 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2248
2249 return snprintf(buf, PAGE_SIZE, "%lu\n",
2250 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2251 sbi->s_sectors_written_start) >> 1);
2252}
2253
2254static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2255 struct ext4_sb_info *sbi, char *buf)
2256{
2257 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2258
2259 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002260 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002261 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002262 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002263}
2264
2265static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2266 struct ext4_sb_info *sbi,
2267 const char *buf, size_t count)
2268{
2269 unsigned long t;
2270
2271 if (parse_strtoul(buf, 0x40000000, &t))
2272 return -EINVAL;
2273
Theodore Ts'of7c43952009-04-24 23:31:59 -04002274 if (!is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002275 return -EINVAL;
2276
2277 sbi->s_inode_readahead_blks = t;
2278 return count;
2279}
2280
2281static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002282 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002283{
2284 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2285
2286 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2287}
2288
2289static ssize_t sbi_ui_store(struct ext4_attr *a,
2290 struct ext4_sb_info *sbi,
2291 const char *buf, size_t count)
2292{
2293 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2294 unsigned long t;
2295
2296 if (parse_strtoul(buf, 0xffffffff, &t))
2297 return -EINVAL;
2298 *ui = t;
2299 return count;
2300}
2301
2302#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2303static struct ext4_attr ext4_attr_##_name = { \
2304 .attr = {.name = __stringify(_name), .mode = _mode }, \
2305 .show = _show, \
2306 .store = _store, \
2307 .offset = offsetof(struct ext4_sb_info, _elname), \
2308}
2309#define EXT4_ATTR(name, mode, show, store) \
2310static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2311
2312#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2313#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2314#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2315 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2316#define ATTR_LIST(name) &ext4_attr_##name.attr
2317
2318EXT4_RO_ATTR(delayed_allocation_blocks);
2319EXT4_RO_ATTR(session_write_kbytes);
2320EXT4_RO_ATTR(lifetime_write_kbytes);
2321EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2322 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002323EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002324EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2325EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2326EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2327EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2328EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2329EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002330EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002331
2332static struct attribute *ext4_attrs[] = {
2333 ATTR_LIST(delayed_allocation_blocks),
2334 ATTR_LIST(session_write_kbytes),
2335 ATTR_LIST(lifetime_write_kbytes),
2336 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002337 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002338 ATTR_LIST(mb_stats),
2339 ATTR_LIST(mb_max_to_scan),
2340 ATTR_LIST(mb_min_to_scan),
2341 ATTR_LIST(mb_order2_req),
2342 ATTR_LIST(mb_stream_req),
2343 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002344 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002345 NULL,
2346};
2347
2348static ssize_t ext4_attr_show(struct kobject *kobj,
2349 struct attribute *attr, char *buf)
2350{
2351 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2352 s_kobj);
2353 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2354
2355 return a->show ? a->show(a, sbi, buf) : 0;
2356}
2357
2358static ssize_t ext4_attr_store(struct kobject *kobj,
2359 struct attribute *attr,
2360 const char *buf, size_t len)
2361{
2362 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2363 s_kobj);
2364 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2365
2366 return a->store ? a->store(a, sbi, buf, len) : 0;
2367}
2368
2369static void ext4_sb_release(struct kobject *kobj)
2370{
2371 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2372 s_kobj);
2373 complete(&sbi->s_kobj_unregister);
2374}
2375
2376
Emese Revfy52cf25d2010-01-19 02:58:23 +01002377static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002378 .show = ext4_attr_show,
2379 .store = ext4_attr_store,
2380};
2381
2382static struct kobj_type ext4_ktype = {
2383 .default_attrs = ext4_attrs,
2384 .sysfs_ops = &ext4_attr_ops,
2385 .release = ext4_sb_release,
2386};
2387
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002388/*
2389 * Check whether this filesystem can be mounted based on
2390 * the features present and the RDONLY/RDWR mount requested.
2391 * Returns 1 if this filesystem can be mounted as requested,
2392 * 0 if it cannot be.
2393 */
2394static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2395{
2396 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2397 ext4_msg(sb, KERN_ERR,
2398 "Couldn't mount because of "
2399 "unsupported optional features (%x)",
2400 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2401 ~EXT4_FEATURE_INCOMPAT_SUPP));
2402 return 0;
2403 }
2404
2405 if (readonly)
2406 return 1;
2407
2408 /* Check that feature set is OK for a read-write mount */
2409 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2410 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2411 "unsupported optional features (%x)",
2412 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2413 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2414 return 0;
2415 }
2416 /*
2417 * Large file size enabled file system can only be mounted
2418 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2419 */
2420 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2421 if (sizeof(blkcnt_t) < sizeof(u64)) {
2422 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2423 "cannot be mounted RDWR without "
2424 "CONFIG_LBDAF");
2425 return 0;
2426 }
2427 }
2428 return 1;
2429}
2430
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002431static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04002432 __releases(kernel_lock)
2433 __acquires(kernel_lock)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002434{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002435 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07002436 struct ext4_super_block *es = NULL;
2437 struct ext4_sb_info *sbi;
2438 ext4_fsblk_t block;
2439 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002440 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002441 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002442 unsigned long journal_devnum = 0;
2443 unsigned long def_mount_opts;
2444 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002445 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05002446 const char *descr;
David Howells1d1fe1e2008-02-07 00:15:37 -08002447 int ret = -EINVAL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002448 int blocksize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002449 unsigned int db_count;
2450 unsigned int i;
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002451 int needs_recovery, has_huge_files;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002452 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07002453 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002454 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002455
2456 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2457 if (!sbi)
2458 return -ENOMEM;
Pekka Enberg705895b2009-02-15 18:07:52 -05002459
2460 sbi->s_blockgroup_lock =
2461 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2462 if (!sbi->s_blockgroup_lock) {
2463 kfree(sbi);
2464 return -ENOMEM;
2465 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002466 sb->s_fs_info = sbi;
2467 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002468 sbi->s_resuid = EXT4_DEF_RESUID;
2469 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04002470 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07002471 sbi->s_sb_block = sb_block;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05002472 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2473 sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002474
2475 unlock_kernel();
2476
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002477 /* Cleanup superblock name */
2478 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2479 *cp = '!';
2480
Mingming Cao617ba132006-10-11 01:20:53 -07002481 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002482 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002483 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002484 goto out_fail;
2485 }
2486
2487 /*
Mingming Cao617ba132006-10-11 01:20:53 -07002488 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002489 * block sizes. We need to calculate the offset from buffer start.
2490 */
Mingming Cao617ba132006-10-11 01:20:53 -07002491 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002492 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2493 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002494 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002495 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002496 }
2497
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002498 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002499 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002500 goto out_fail;
2501 }
2502 /*
2503 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07002504 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002505 */
Mingming Cao617ba132006-10-11 01:20:53 -07002506 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002507 sbi->s_es = es;
2508 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07002509 if (sb->s_magic != EXT4_SUPER_MAGIC)
2510 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05002511 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002512
2513 /* Set defaults before we parse the mount options */
2514 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Mingming Cao617ba132006-10-11 01:20:53 -07002515 if (def_mount_opts & EXT4_DEFM_DEBUG)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002516 set_opt(sbi->s_mount_opt, DEBUG);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05002517 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2518 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2519 "2.6.38");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002520 set_opt(sbi->s_mount_opt, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05002521 }
Mingming Cao617ba132006-10-11 01:20:53 -07002522 if (def_mount_opts & EXT4_DEFM_UID16)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002523 set_opt(sbi->s_mount_opt, NO_UID32);
Theodore Ts'o03010a32008-10-10 20:02:48 -04002524#ifdef CONFIG_EXT4_FS_XATTR
Mingming Cao617ba132006-10-11 01:20:53 -07002525 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002526 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08002527#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04002528#ifdef CONFIG_EXT4_FS_POSIX_ACL
Mingming Cao617ba132006-10-11 01:20:53 -07002529 if (def_mount_opts & EXT4_DEFM_ACL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002530 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08002531#endif
Mingming Cao617ba132006-10-11 01:20:53 -07002532 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Dmitry Monakhov482a7422010-02-24 11:35:32 -05002533 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07002534 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Dmitry Monakhov482a7422010-02-24 11:35:32 -05002535 set_opt(sbi->s_mount_opt, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07002536 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Dmitry Monakhov482a7422010-02-24 11:35:32 -05002537 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002538
Mingming Cao617ba132006-10-11 01:20:53 -07002539 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002540 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05002541 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Dmitry Mishinceea16b2006-10-11 01:21:21 -07002542 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05002543 else
2544 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002545
2546 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2547 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05002548 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2549 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2550 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002551
Eric Sandeen571640c2008-05-26 12:29:46 -04002552 set_opt(sbi->s_mount_opt, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002553
Mingming Cao1e2462f2007-07-18 09:00:55 -04002554 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04002555 * enable delayed allocation by default
2556 * Use -o nodelalloc to turn it off
2557 */
Jan Karaba69f9a2010-03-24 20:18:37 -04002558 if (!IS_EXT3_SB(sb))
2559 set_opt(sbi->s_mount_opt, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04002560
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002561 if (!parse_options((char *) data, sb, &journal_devnum,
2562 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002563 goto failed_mount;
2564
2565 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05002566 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002567
Mingming Cao617ba132006-10-11 01:20:53 -07002568 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2569 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2570 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2571 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002572 ext4_msg(sb, KERN_WARNING,
2573 "feature flags set on rev 0 fs, "
2574 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05002575
2576 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002577 * Check feature flags regardless of the revision level, since we
2578 * previously didn't change the revision level when setting the flags,
2579 * so there is a chance incompat flags are set on a rev 0 filesystem.
2580 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002581 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002582 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002583
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002584 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2585
Mingming Cao617ba132006-10-11 01:20:53 -07002586 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2587 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002588 ext4_msg(sb, KERN_ERR,
2589 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002590 goto failed_mount;
2591 }
2592
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002593 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002594 /* Validate the filesystem blocksize */
2595 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002596 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002597 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002598 goto failed_mount;
2599 }
2600
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002601 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002602 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2603 offset = do_div(logical_sb_block, blocksize);
2604 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002605 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002606 ext4_msg(sb, KERN_ERR,
2607 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002608 goto failed_mount;
2609 }
Mingming Cao617ba132006-10-11 01:20:53 -07002610 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002611 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07002612 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002613 ext4_msg(sb, KERN_ERR,
2614 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002615 goto failed_mount;
2616 }
2617 }
2618
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002619 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2620 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002621 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2622 has_huge_files);
2623 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002624
Mingming Cao617ba132006-10-11 01:20:53 -07002625 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2626 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2627 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002628 } else {
2629 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2630 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07002631 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04002632 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002633 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002634 ext4_msg(sb, KERN_ERR,
2635 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002636 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002637 goto failed_mount;
2638 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04002639 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2640 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002641 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002642
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002643 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2644 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002645 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002646 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07002647 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002648 ext4_msg(sb, KERN_ERR,
2649 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002650 sbi->s_desc_size);
2651 goto failed_mount;
2652 }
2653 } else
2654 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002655
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002656 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002657 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08002658 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002659 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002660
Mingming Cao617ba132006-10-11 01:20:53 -07002661 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002662 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002663 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002664 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2665 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002666 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002667 sbi->s_sbh = bh;
2668 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07002669 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2670 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002671
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002672 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002673 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2674 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04002675 i = le32_to_cpu(es->s_flags);
2676 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2677 sbi->s_hash_unsigned = 3;
2678 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2679#ifdef __CHAR_UNSIGNED__
2680 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2681 sbi->s_hash_unsigned = 3;
2682#else
2683 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2684#endif
2685 sb->s_dirt = 1;
2686 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002687
2688 if (sbi->s_blocks_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002689 ext4_msg(sb, KERN_ERR,
2690 "#blocks per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002691 sbi->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002692 goto failed_mount;
2693 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002694 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002695 ext4_msg(sb, KERN_ERR,
2696 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002697 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002698 goto failed_mount;
2699 }
2700
Eric Sandeenbf43d842009-08-17 23:48:51 -04002701 /*
2702 * Test whether we have more sectors than will fit in sector_t,
2703 * and whether the max offset is addressable by the page cache.
2704 */
2705 if ((ext4_blocks_count(es) >
2706 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2707 (ext4_blocks_count(es) >
2708 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002709 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04002710 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002711 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002712 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Eric Sandeenbf43d842009-08-17 23:48:51 -04002713 ret = -EFBIG;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002714 goto failed_mount;
2715 }
2716
Mingming Cao617ba132006-10-11 01:20:53 -07002717 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2718 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05002719
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04002720 /* check blocks count against device size */
2721 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2722 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002723 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2724 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04002725 ext4_blocks_count(es), blocks_count);
2726 goto failed_mount;
2727 }
2728
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002729 /*
2730 * It makes no sense for the first data block to be beyond the end
2731 * of the filesystem.
2732 */
2733 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002734 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2735 "block %u is beyond end of filesystem (%llu)",
2736 le32_to_cpu(es->s_first_data_block),
2737 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05002738 goto failed_mount;
2739 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002740 blocks_count = (ext4_blocks_count(es) -
2741 le32_to_cpu(es->s_first_data_block) +
2742 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2743 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002744 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002745 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002746 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002747 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002748 ext4_blocks_count(es),
2749 le32_to_cpu(es->s_first_data_block),
2750 EXT4_BLOCKS_PER_GROUP(sb));
2751 goto failed_mount;
2752 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002753 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04002754 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2755 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07002756 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2757 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002758 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002759 GFP_KERNEL);
2760 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002761 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002762 goto failed_mount;
2763 }
2764
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04002765#ifdef CONFIG_PROC_FS
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002766 if (ext4_proc_root)
2767 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04002768#endif
Theodore Ts'o240799c2008-10-09 23:53:47 -04002769
Pekka Enberg705895b2009-02-15 18:07:52 -05002770 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002771
2772 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002773 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002774 sbi->s_group_desc[i] = sb_bread(sb, block);
2775 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002776 ext4_msg(sb, KERN_ERR,
2777 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002778 db_count = i;
2779 goto failed_mount2;
2780 }
2781 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002782 if (!ext4_check_descriptors(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002783 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002784 goto failed_mount2;
2785 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002786 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2787 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002788 ext4_msg(sb, KERN_ERR,
2789 "unable to initialize "
2790 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002791 goto failed_mount2;
2792 }
2793
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002794 sbi->s_gdb_count = db_count;
2795 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2796 spin_lock_init(&sbi->s_next_gen_lock);
2797
Peter Zijlstra833f4072007-10-16 23:25:45 -07002798 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2799 ext4_count_free_blocks(sb));
2800 if (!err) {
2801 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2802 ext4_count_free_inodes(sb));
2803 }
2804 if (!err) {
2805 err = percpu_counter_init(&sbi->s_dirs_counter,
2806 ext4_count_dirs(sb));
2807 }
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04002808 if (!err) {
2809 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2810 }
Peter Zijlstra833f4072007-10-16 23:25:45 -07002811 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002812 ext4_msg(sb, KERN_ERR, "insufficient memory");
Peter Zijlstra833f4072007-10-16 23:25:45 -07002813 goto failed_mount3;
2814 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002815
Alex Tomasc9de5602008-01-29 00:19:52 -05002816 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002817 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05002818
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002819 /*
2820 * set up enough so that it can read an inode
2821 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04002822 if (!test_opt(sb, NOLOAD) &&
2823 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2824 sb->s_op = &ext4_sops;
2825 else
2826 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07002827 sb->s_export_op = &ext4_export_ops;
2828 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002829#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07002830 sb->s_qcop = &ext4_qctl_operations;
2831 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002832#endif
2833 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04002834 mutex_init(&sbi->s_orphan_lock);
Theodore Ts'o32ed5052009-04-25 22:53:39 -04002835 mutex_init(&sbi->s_resize_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002836
2837 sb->s_root = NULL;
2838
2839 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07002840 EXT4_HAS_INCOMPAT_FEATURE(sb,
2841 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002842
2843 /*
2844 * The first inode we look at is the journal inode. Don't try
2845 * root first: it may be modified in the journal!
2846 */
2847 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07002848 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2849 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002850 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05002851 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2852 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002853 ext4_msg(sb, KERN_ERR, "required journal recovery "
2854 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05002855 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002856 } else {
Frank Mayhar03901312009-01-07 00:06:22 -05002857 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2858 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2859 sbi->s_journal = NULL;
2860 needs_recovery = 0;
2861 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002862 }
2863
Jose R. Santoseb40a092007-07-18 08:37:25 -04002864 if (ext4_blocks_count(es) > 0xffffffffULL &&
2865 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2866 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002867 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05002868 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04002869 }
2870
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08002871 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2872 jbd2_journal_set_features(sbi->s_journal,
2873 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05002874 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08002875 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2876 jbd2_journal_set_features(sbi->s_journal,
2877 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
Girish Shilamkar818d2762008-01-28 23:58:27 -05002878 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2879 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08002880 } else {
2881 jbd2_journal_clear_features(sbi->s_journal,
2882 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2883 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2884 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05002885
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002886 /* We have now updated the journal if required, so we can
2887 * validate the data journaling mode. */
2888 switch (test_opt(sb, DATA_FLAGS)) {
2889 case 0:
2890 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07002891 * capabilities: ORDERED_DATA if the journal can
2892 * cope, else JOURNAL_DATA
2893 */
Mingming Caodab291a2006-10-11 01:21:01 -07002894 if (jbd2_journal_check_available_features
2895 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002896 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2897 else
2898 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2899 break;
2900
Mingming Cao617ba132006-10-11 01:20:53 -07002901 case EXT4_MOUNT_ORDERED_DATA:
2902 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07002903 if (!jbd2_journal_check_available_features
2904 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002905 ext4_msg(sb, KERN_ERR, "Journal does not support "
2906 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05002907 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002908 }
2909 default:
2910 break;
2911 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002912 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002913
Frank Mayhar03901312009-01-07 00:06:22 -05002914no_journal:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002915 if (test_opt(sb, NOBH)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002916 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002917 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2918 "its supported only with writeback mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002919 clear_opt(sbi->s_mount_opt, NOBH);
2920 }
Jiaying Zhang744692d2010-03-04 16:14:02 -05002921 if (test_opt(sb, DIOREAD_NOLOCK)) {
2922 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2923 "not supported with nobh mode");
2924 goto failed_mount_wq;
2925 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002926 }
Mingming Cao4c0425f2009-09-28 15:48:41 -04002927 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2928 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2929 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2930 goto failed_mount_wq;
2931 }
2932
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002933 /*
Mingming Caodab291a2006-10-11 01:21:01 -07002934 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002935 * so we can safely mount the rest of the filesystem now.
2936 */
2937
David Howells1d1fe1e2008-02-07 00:15:37 -08002938 root = ext4_iget(sb, EXT4_ROOT_INO);
2939 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002940 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08002941 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002942 goto failed_mount4;
2943 }
2944 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08002945 iput(root);
Eric Sandeenb31e1552009-06-04 17:36:36 -04002946 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002947 goto failed_mount4;
2948 }
David Howells1d1fe1e2008-02-07 00:15:37 -08002949 sb->s_root = d_alloc_root(root);
2950 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002951 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08002952 iput(root);
2953 ret = -ENOMEM;
2954 goto failed_mount4;
2955 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002956
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002957 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04002958
2959 /* determine the minimum size of new large inodes, if present */
2960 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2961 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2962 EXT4_GOOD_OLD_INODE_SIZE;
2963 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2964 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2965 if (sbi->s_want_extra_isize <
2966 le16_to_cpu(es->s_want_extra_isize))
2967 sbi->s_want_extra_isize =
2968 le16_to_cpu(es->s_want_extra_isize);
2969 if (sbi->s_want_extra_isize <
2970 le16_to_cpu(es->s_min_extra_isize))
2971 sbi->s_want_extra_isize =
2972 le16_to_cpu(es->s_min_extra_isize);
2973 }
2974 }
2975 /* Check if enough inode space is available */
2976 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2977 sbi->s_inode_size) {
2978 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2979 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04002980 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2981 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04002982 }
2983
Theodore Ts'o90576c02009-09-29 15:51:30 -04002984 if (test_opt(sb, DELALLOC) &&
2985 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002986 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2987 "requested data journaling mode");
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002988 clear_opt(sbi->s_mount_opt, DELALLOC);
Theodore Ts'o90576c02009-09-29 15:51:30 -04002989 }
Jiaying Zhang744692d2010-03-04 16:14:02 -05002990 if (test_opt(sb, DIOREAD_NOLOCK)) {
2991 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2992 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2993 "option - requested data journaling mode");
2994 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2995 }
2996 if (sb->s_blocksize < PAGE_SIZE) {
2997 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2998 "option - block size is too small");
2999 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3000 }
3001 }
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003002
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003003 err = ext4_setup_system_zone(sb);
3004 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003005 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3006 "zone (%d)\n", err);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003007 goto failed_mount4;
3008 }
3009
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003010 ext4_ext_init(sb);
3011 err = ext4_mb_init(sb, needs_recovery);
3012 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003013 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3014 err);
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003015 goto failed_mount4;
3016 }
3017
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003018 sbi->s_kobj.kset = ext4_kset;
3019 init_completion(&sbi->s_kobj_unregister);
3020 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3021 "%s", sb->s_id);
3022 if (err) {
3023 ext4_mb_release(sb);
3024 ext4_ext_release(sb);
3025 goto failed_mount4;
3026 };
3027
Mingming Cao617ba132006-10-11 01:20:53 -07003028 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3029 ext4_orphan_cleanup(sb, es);
3030 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05003031 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003032 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05003033 ext4_mark_recovery_complete(sb, es);
3034 }
3035 if (EXT4_SB(sb)->s_journal) {
3036 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3037 descr = " journalled data mode";
3038 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3039 descr = " ordered data mode";
3040 else
3041 descr = " writeback data mode";
3042 } else
3043 descr = "out journal";
3044
Eric Sandeenb31e1552009-06-04 17:36:36 -04003045 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003046
3047 lock_kernel();
3048 return 0;
3049
Mingming Cao617ba132006-10-11 01:20:53 -07003050cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003051 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003052 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003053 goto failed_mount;
3054
3055failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04003056 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04003057 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3058failed_mount_wq:
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003059 ext4_release_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003060 if (sbi->s_journal) {
3061 jbd2_journal_destroy(sbi->s_journal);
3062 sbi->s_journal = NULL;
3063 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003064failed_mount3:
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04003065 if (sbi->s_flex_groups) {
3066 if (is_vmalloc_addr(sbi->s_flex_groups))
3067 vfree(sbi->s_flex_groups);
3068 else
3069 kfree(sbi->s_flex_groups);
3070 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003071 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3072 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3073 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04003074 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003075failed_mount2:
3076 for (i = 0; i < db_count; i++)
3077 brelse(sbi->s_group_desc[i]);
3078 kfree(sbi->s_group_desc);
3079failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04003080 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003081 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003082 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003083#ifdef CONFIG_QUOTA
3084 for (i = 0; i < MAXQUOTAS; i++)
3085 kfree(sbi->s_qf_names[i]);
3086#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003087 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003088 brelse(bh);
3089out_fail:
3090 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04003091 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003092 kfree(sbi);
3093 lock_kernel();
David Howells1d1fe1e2008-02-07 00:15:37 -08003094 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003095}
3096
3097/*
3098 * Setup any per-fs journal parameters now. We'll do this both on
3099 * initial mount, once the journal has been initialised but before we've
3100 * done any recovery; and again on any subsequent remount.
3101 */
Mingming Cao617ba132006-10-11 01:20:53 -07003102static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003103{
Mingming Cao617ba132006-10-11 01:20:53 -07003104 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003105
Theodore Ts'o30773842009-01-03 20:27:38 -05003106 journal->j_commit_interval = sbi->s_commit_interval;
3107 journal->j_min_batch_time = sbi->s_min_batch_time;
3108 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003109
3110 spin_lock(&journal->j_state_lock);
3111 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07003112 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003113 else
Mingming Caodab291a2006-10-11 01:21:01 -07003114 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04003115 if (test_opt(sb, DATA_ERR_ABORT))
3116 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3117 else
3118 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003119 spin_unlock(&journal->j_state_lock);
3120}
3121
Mingming Cao617ba132006-10-11 01:20:53 -07003122static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003123 unsigned int journal_inum)
3124{
3125 struct inode *journal_inode;
3126 journal_t *journal;
3127
Frank Mayhar03901312009-01-07 00:06:22 -05003128 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3129
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003130 /* First, test for the existence of a valid inode on disk. Bad
3131 * things happen if we iget() an unused inode, as the subsequent
3132 * iput() will try to delete it. */
3133
David Howells1d1fe1e2008-02-07 00:15:37 -08003134 journal_inode = ext4_iget(sb, journal_inum);
3135 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003136 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003137 return NULL;
3138 }
3139 if (!journal_inode->i_nlink) {
3140 make_bad_inode(journal_inode);
3141 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003142 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003143 return NULL;
3144 }
3145
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003146 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003147 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003148 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003149 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003150 iput(journal_inode);
3151 return NULL;
3152 }
3153
Mingming Caodab291a2006-10-11 01:21:01 -07003154 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003155 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003156 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003157 iput(journal_inode);
3158 return NULL;
3159 }
3160 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003161 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003162 return journal;
3163}
3164
Mingming Cao617ba132006-10-11 01:20:53 -07003165static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003166 dev_t j_dev)
3167{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003168 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003169 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003170 ext4_fsblk_t start;
3171 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003172 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003173 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003174 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003175 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003176 struct block_device *bdev;
3177
Frank Mayhar03901312009-01-07 00:06:22 -05003178 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3179
Eric Sandeenb31e1552009-06-04 17:36:36 -04003180 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003181 if (bdev == NULL)
3182 return NULL;
3183
3184 if (bd_claim(bdev, sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003185 ext4_msg(sb, KERN_ERR,
3186 "failed to claim external journal device");
Al Viro9a1c3542008-02-22 20:40:24 -05003187 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003188 return NULL;
3189 }
3190
3191 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003192 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003193 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003194 ext4_msg(sb, KERN_ERR,
3195 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003196 goto out_bdev;
3197 }
3198
Mingming Cao617ba132006-10-11 01:20:53 -07003199 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3200 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003201 set_blocksize(bdev, blocksize);
3202 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003203 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3204 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003205 goto out_bdev;
3206 }
3207
Mingming Cao617ba132006-10-11 01:20:53 -07003208 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3209 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003210 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003211 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003212 ext4_msg(sb, KERN_ERR, "external journal has "
3213 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003214 brelse(bh);
3215 goto out_bdev;
3216 }
3217
Mingming Cao617ba132006-10-11 01:20:53 -07003218 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003219 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003220 brelse(bh);
3221 goto out_bdev;
3222 }
3223
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003224 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003225 start = sb_block + 1;
3226 brelse(bh); /* we're done with the superblock */
3227
Mingming Caodab291a2006-10-11 01:21:01 -07003228 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003229 start, len, blocksize);
3230 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003231 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003232 goto out_bdev;
3233 }
3234 journal->j_private = sb;
3235 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3236 wait_on_buffer(journal->j_sb_buffer);
3237 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003238 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003239 goto out_journal;
3240 }
3241 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003242 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3243 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003244 be32_to_cpu(journal->j_superblock->s_nr_users));
3245 goto out_journal;
3246 }
Mingming Cao617ba132006-10-11 01:20:53 -07003247 EXT4_SB(sb)->journal_bdev = bdev;
3248 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003249 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003250
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003251out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07003252 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003253out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07003254 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003255 return NULL;
3256}
3257
Mingming Cao617ba132006-10-11 01:20:53 -07003258static int ext4_load_journal(struct super_block *sb,
3259 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003260 unsigned long journal_devnum)
3261{
3262 journal_t *journal;
3263 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3264 dev_t journal_dev;
3265 int err = 0;
3266 int really_read_only;
3267
Frank Mayhar03901312009-01-07 00:06:22 -05003268 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3269
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003270 if (journal_devnum &&
3271 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003272 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3273 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003274 journal_dev = new_decode_dev(journal_devnum);
3275 } else
3276 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3277
3278 really_read_only = bdev_read_only(sb->s_bdev);
3279
3280 /*
3281 * Are we loading a blank journal or performing recovery after a
3282 * crash? For recovery, we need to check in advance whether we
3283 * can get read-write access to the device.
3284 */
Mingming Cao617ba132006-10-11 01:20:53 -07003285 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003286 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003287 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3288 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003289 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003290 ext4_msg(sb, KERN_ERR, "write access "
3291 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003292 return -EROFS;
3293 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04003294 ext4_msg(sb, KERN_INFO, "write access will "
3295 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003296 }
3297 }
3298
3299 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003300 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3301 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003302 return -EINVAL;
3303 }
3304
3305 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07003306 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003307 return -EINVAL;
3308 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07003309 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003310 return -EINVAL;
3311 }
3312
Theodore Ts'o90576c02009-09-29 15:51:30 -04003313 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003314 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04003315
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003316 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07003317 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003318 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003319 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003320 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003321 return err;
3322 }
3323 }
3324
Mingming Cao617ba132006-10-11 01:20:53 -07003325 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07003326 err = jbd2_journal_wipe(journal, !really_read_only);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003327 if (!err)
Mingming Caodab291a2006-10-11 01:21:01 -07003328 err = jbd2_journal_load(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003329
3330 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003331 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003332 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003333 return err;
3334 }
3335
Mingming Cao617ba132006-10-11 01:20:53 -07003336 EXT4_SB(sb)->s_journal = journal;
3337 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003338
3339 if (journal_devnum &&
3340 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3341 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003342
3343 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003344 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003345 }
3346
3347 return 0;
3348}
3349
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003350static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003351{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07003353 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003354 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003355
3356 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08003357 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04003358 if (buffer_write_io_error(sbh)) {
3359 /*
3360 * Oh, dear. A previous attempt to write the
3361 * superblock failed. This could happen because the
3362 * USB device was yanked out. Or it could happen to
3363 * be a transient write error and maybe the block will
3364 * be remapped. Nothing we can do but to retry the
3365 * write and hope for the best.
3366 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04003367 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3368 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003369 clear_buffer_write_io_error(sbh);
3370 set_buffer_uptodate(sbh);
3371 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04003372 /*
3373 * If the file system is mounted read-only, don't update the
3374 * superblock write time. This avoids updating the superblock
3375 * write time when we are mounting the root file system
3376 * read/only but we need to replay the journal; at that point,
3377 * for people who are east of GMT and who make their clock
3378 * tick in localtime for Windows bug-for-bug compatibility,
3379 * the clock is set in the future, and this will cause e2fsck
3380 * to complain and force a full file system check.
3381 */
3382 if (!(sb->s_flags & MS_RDONLY))
3383 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003384 es->s_kbytes_written =
3385 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3386 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3387 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -05003388 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3389 &EXT4_SB(sb)->s_freeblocks_counter));
3390 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3391 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003392 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003393 BUFFER_TRACE(sbh, "marking dirty");
3394 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04003395 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08003396 error = sync_dirty_buffer(sbh);
3397 if (error)
3398 return error;
3399
3400 error = buffer_write_io_error(sbh);
3401 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003402 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3403 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003404 clear_buffer_write_io_error(sbh);
3405 set_buffer_uptodate(sbh);
3406 }
3407 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08003408 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003409}
3410
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003411/*
3412 * Have we just finished recovery? If so, and if we are mounting (or
3413 * remounting) the filesystem readonly, then we will end up with a
3414 * consistent fs on disk. Record that fact.
3415 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003416static void ext4_mark_recovery_complete(struct super_block *sb,
3417 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003418{
Mingming Cao617ba132006-10-11 01:20:53 -07003419 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003420
Frank Mayhar03901312009-01-07 00:06:22 -05003421 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3422 BUG_ON(journal != NULL);
3423 return;
3424 }
Mingming Caodab291a2006-10-11 01:21:01 -07003425 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003426 if (jbd2_journal_flush(journal) < 0)
3427 goto out;
3428
Mingming Cao617ba132006-10-11 01:20:53 -07003429 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003430 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07003431 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003432 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003433 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003434
3435out:
Mingming Caodab291a2006-10-11 01:21:01 -07003436 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003437}
3438
3439/*
3440 * If we are mounting (or read-write remounting) a filesystem whose journal
3441 * has recorded an error from a previous lifetime, move that error to the
3442 * main filesystem now.
3443 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003444static void ext4_clear_journal_err(struct super_block *sb,
3445 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003446{
3447 journal_t *journal;
3448 int j_errno;
3449 const char *errstr;
3450
Frank Mayhar03901312009-01-07 00:06:22 -05003451 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3452
Mingming Cao617ba132006-10-11 01:20:53 -07003453 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003454
3455 /*
3456 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07003457 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003458 */
3459
Mingming Caodab291a2006-10-11 01:21:01 -07003460 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003461 if (j_errno) {
3462 char nbuf[16];
3463
Mingming Cao617ba132006-10-11 01:20:53 -07003464 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05003465 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003466 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05003467 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003468
Mingming Cao617ba132006-10-11 01:20:53 -07003469 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3470 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003471 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003472
Mingming Caodab291a2006-10-11 01:21:01 -07003473 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003474 }
3475}
3476
3477/*
3478 * Force the running and committing transactions to commit,
3479 * and wait on the commit.
3480 */
Mingming Cao617ba132006-10-11 01:20:53 -07003481int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003482{
3483 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05003484 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003485
3486 if (sb->s_flags & MS_RDONLY)
3487 return 0;
3488
Mingming Cao617ba132006-10-11 01:20:53 -07003489 journal = EXT4_SB(sb)->s_journal;
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003490 if (journal)
Frank Mayhar03901312009-01-07 00:06:22 -05003491 ret = ext4_journal_force_commit(journal);
Frank Mayhar03901312009-01-07 00:06:22 -05003492
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003493 return ret;
3494}
3495
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003496static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003497{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02003498 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003499 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02003500 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003501}
3502
Mingming Cao617ba132006-10-11 01:20:53 -07003503static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003504{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05003505 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05003506 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003507 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003508
Theodore Ts'o9bffad12009-06-17 11:48:11 -04003509 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003510 flush_workqueue(sbi->dio_unwritten_wq);
3511 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003512 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003513 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05003514 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05003515 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003516}
3517
3518/*
3519 * LVM calls this function before a (read-only) snapshot is created. This
3520 * gives us a chance to flush the journal completely and mark the fs clean.
3521 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08003522static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003523{
Takashi Satoc4be0c12009-01-09 16:40:58 -08003524 int error = 0;
3525 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003526
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003527 if (sb->s_flags & MS_RDONLY)
3528 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003529
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003530 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003531
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003532 /* Now we set up the journal barrier. */
3533 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003534
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003535 /*
3536 * Don't clear the needs_recovery flag if we failed to flush
3537 * the journal.
3538 */
3539 error = jbd2_journal_flush(journal);
3540 if (error < 0) {
3541 out:
3542 jbd2_journal_unlock_updates(journal);
3543 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003544 }
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003545
3546 /* Journal blocked and flushed, clear needs_recovery flag. */
3547 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3548 error = ext4_commit_super(sb, 1);
3549 if (error)
3550 goto out;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003551 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003552}
3553
3554/*
3555 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3556 * flag here, even though the filesystem is not technically dirty yet.
3557 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08003558static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003559{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003560 if (sb->s_flags & MS_RDONLY)
3561 return 0;
3562
3563 lock_super(sb);
3564 /* Reset the needs_recovery flag before the fs is unlocked. */
3565 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3566 ext4_commit_super(sb, 1);
3567 unlock_super(sb);
3568 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Takashi Satoc4be0c12009-01-09 16:40:58 -08003569 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003570}
3571
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003572static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003573{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003574 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07003575 struct ext4_sb_info *sbi = EXT4_SB(sb);
3576 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003577 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07003578 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04003579 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04003580 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003581 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003582 int err;
3583#ifdef CONFIG_QUOTA
3584 int i;
3585#endif
3586
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003587 lock_kernel();
3588
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003589 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04003590 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003591 old_sb_flags = sb->s_flags;
3592 old_opts.s_mount_opt = sbi->s_mount_opt;
3593 old_opts.s_resuid = sbi->s_resuid;
3594 old_opts.s_resgid = sbi->s_resgid;
3595 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05003596 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3597 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003598#ifdef CONFIG_QUOTA
3599 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3600 for (i = 0; i < MAXQUOTAS; i++)
3601 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3602#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003603 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3604 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003605
3606 /*
3607 * Allow the "check" option to be passed as a remount option.
3608 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003609 if (!parse_options(data, sb, NULL, &journal_ioprio,
3610 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003611 err = -EINVAL;
3612 goto restore_opts;
3613 }
3614
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04003615 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Harvey Harrison46e665e2008-04-17 10:38:59 -04003616 ext4_abort(sb, __func__, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003617
3618 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003619 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003620
3621 es = sbi->s_es;
3622
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003623 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05003624 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003625 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3626 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003627
3628 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003629 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04003630 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003631 err = -EROFS;
3632 goto restore_opts;
3633 }
3634
3635 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04003636 err = dquot_suspend(sb, -1);
3637 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04003638 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04003639
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003640 /*
3641 * First of all, the unconditional stuff we have to do
3642 * to disable replay of the journal when we next remount
3643 */
3644 sb->s_flags |= MS_RDONLY;
3645
3646 /*
3647 * OK, test if we are remounting a valid rw partition
3648 * readonly, and if so set the rdonly flag and then
3649 * mark the partition as valid again.
3650 */
Mingming Cao617ba132006-10-11 01:20:53 -07003651 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3652 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003653 es->s_state = cpu_to_le16(sbi->s_mount_state);
3654
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04003655 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05003656 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003657 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003658 /* Make sure we can mount this feature set readwrite */
3659 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003660 err = -EROFS;
3661 goto restore_opts;
3662 }
Eric Sandeenead65962007-02-10 01:46:08 -08003663 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04003664 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003665 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04003666 */
3667 for (g = 0; g < sbi->s_groups_count; g++) {
3668 struct ext4_group_desc *gdp =
3669 ext4_get_group_desc(sb, g, NULL);
3670
3671 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003672 ext4_msg(sb, KERN_ERR,
3673 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04003674 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3675 le16_to_cpu(gdp->bg_checksum));
3676 err = -EINVAL;
3677 goto restore_opts;
3678 }
3679 }
3680
3681 /*
Eric Sandeenead65962007-02-10 01:46:08 -08003682 * If we have an unprocessed orphan list hanging
3683 * around from a previously readonly bdev mount,
3684 * require a full umount/remount for now.
3685 */
3686 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003687 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08003688 "remount RDWR because of unprocessed "
3689 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003690 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08003691 err = -EINVAL;
3692 goto restore_opts;
3693 }
3694
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003695 /*
3696 * Mounting a RDONLY partition read-write, so reread
3697 * and store the current valid flag. (It may have
3698 * been changed by e2fsck since we originally mounted
3699 * the partition.)
3700 */
Frank Mayhar03901312009-01-07 00:06:22 -05003701 if (sbi->s_journal)
3702 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003703 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07003704 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003705 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003706 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003707 sb->s_flags &= ~MS_RDONLY;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04003708 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003709 }
3710 }
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003711 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003712 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003713 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05003714
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003715#ifdef CONFIG_QUOTA
3716 /* Release old quota file names */
3717 for (i = 0; i < MAXQUOTAS; i++)
3718 if (old_opts.s_qf_names[i] &&
3719 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3720 kfree(old_opts.s_qf_names[i]);
3721#endif
Al Virobbd68512009-05-06 10:43:07 -04003722 unlock_super(sb);
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003723 unlock_kernel();
Christoph Hellwigc79d9672010-05-19 07:16:40 -04003724 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04003725 dquot_resume(sb, -1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003726 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003727
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003728restore_opts:
3729 sb->s_flags = old_sb_flags;
3730 sbi->s_mount_opt = old_opts.s_mount_opt;
3731 sbi->s_resuid = old_opts.s_resuid;
3732 sbi->s_resgid = old_opts.s_resgid;
3733 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05003734 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3735 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003736#ifdef CONFIG_QUOTA
3737 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3738 for (i = 0; i < MAXQUOTAS; i++) {
3739 if (sbi->s_qf_names[i] &&
3740 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3741 kfree(sbi->s_qf_names[i]);
3742 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3743 }
3744#endif
Al Virobbd68512009-05-06 10:43:07 -04003745 unlock_super(sb);
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003746 unlock_kernel();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003747 return err;
3748}
3749
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003750static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003751{
3752 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003753 struct ext4_sb_info *sbi = EXT4_SB(sb);
3754 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08003755 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003756
Badari Pulavarty5e700302007-07-15 23:42:00 -07003757 if (test_opt(sb, MINIX_DF)) {
3758 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003759 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04003760 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07003761 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003762
3763 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07003764 * Compute the overhead (FS structures). This is constant
3765 * for a given filesystem unless the number of block groups
3766 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003767 */
3768
3769 /*
3770 * All of the blocks before first_data_block are
3771 * overhead
3772 */
3773 overhead = le32_to_cpu(es->s_first_data_block);
3774
3775 /*
3776 * Add the overhead attributed to the superblock and
3777 * block group descriptors. If the sparse superblocks
3778 * feature is turned on, then not all groups have this.
3779 */
3780 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07003781 overhead += ext4_bg_has_super(sb, i) +
3782 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003783 cond_resched();
3784 }
3785
3786 /*
3787 * Every block group has an inode bitmap, a block
3788 * bitmap, and an inode table.
3789 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07003790 overhead += ngroups * (2 + sbi->s_itb_per_group);
3791 sbi->s_overhead_last = overhead;
3792 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003793 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003794 }
3795
Mingming Cao617ba132006-10-11 01:20:53 -07003796 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003797 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07003798 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04003799 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3800 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003801 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3802 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003803 buf->f_bavail = 0;
3804 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07003805 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07003806 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08003807 fsid = le64_to_cpup((void *)es->s_uuid) ^
3808 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3809 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3810 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003811
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003812 return 0;
3813}
3814
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003815/* Helper function for writing quotas on sync - we need to start transaction
3816 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003817 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07003818 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01003819 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05003820 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07003821 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003822 *
3823 */
3824
3825#ifdef CONFIG_QUOTA
3826
3827static inline struct inode *dquot_to_inode(struct dquot *dquot)
3828{
3829 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3830}
3831
Mingming Cao617ba132006-10-11 01:20:53 -07003832static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003833{
3834 int ret, err;
3835 handle_t *handle;
3836 struct inode *inode;
3837
3838 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003839 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003840 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003841 if (IS_ERR(handle))
3842 return PTR_ERR(handle);
3843 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003844 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003845 if (!ret)
3846 ret = err;
3847 return ret;
3848}
3849
Mingming Cao617ba132006-10-11 01:20:53 -07003850static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003851{
3852 int ret, err;
3853 handle_t *handle;
3854
Mingming Cao617ba132006-10-11 01:20:53 -07003855 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003856 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003857 if (IS_ERR(handle))
3858 return PTR_ERR(handle);
3859 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003860 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003861 if (!ret)
3862 ret = err;
3863 return ret;
3864}
3865
Mingming Cao617ba132006-10-11 01:20:53 -07003866static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003867{
3868 int ret, err;
3869 handle_t *handle;
3870
Mingming Cao617ba132006-10-11 01:20:53 -07003871 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003872 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07003873 if (IS_ERR(handle)) {
3874 /* Release dquot anyway to avoid endless cycle in dqput() */
3875 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003876 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07003877 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003878 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003879 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003880 if (!ret)
3881 ret = err;
3882 return ret;
3883}
3884
Mingming Cao617ba132006-10-11 01:20:53 -07003885static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003886{
Jan Kara2c8be6b2008-05-13 21:27:55 -04003887 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07003888 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3889 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003890 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003891 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003892 } else {
3893 return dquot_mark_dquot_dirty(dquot);
3894 }
3895}
3896
Mingming Cao617ba132006-10-11 01:20:53 -07003897static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003898{
3899 int ret, err;
3900 handle_t *handle;
3901
3902 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07003903 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003904 if (IS_ERR(handle))
3905 return PTR_ERR(handle);
3906 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07003907 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003908 if (!ret)
3909 ret = err;
3910 return ret;
3911}
3912
3913/*
3914 * Turn on quotas during mount time - we need to find
3915 * the quota file and such...
3916 */
Mingming Cao617ba132006-10-11 01:20:53 -07003917static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003918{
Mingming Cao617ba132006-10-11 01:20:53 -07003919 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003920 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003921}
3922
3923/*
3924 * Standard function to be called on quota_on
3925 */
Mingming Cao617ba132006-10-11 01:20:53 -07003926static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Christoph Hellwig307ae182010-05-19 07:16:43 -04003927 char *name)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003928{
3929 int err;
Al Viro82646132008-08-02 00:57:06 -04003930 struct path path;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003931
3932 if (!test_opt(sb, QUOTA))
3933 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04003934
Al Viro82646132008-08-02 00:57:06 -04003935 err = kern_path(name, LOOKUP_FOLLOW, &path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003936 if (err)
3937 return err;
Jan Kara06235432008-05-13 19:11:51 -04003938
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003939 /* Quotafile not on the same filesystem? */
Al Viro82646132008-08-02 00:57:06 -04003940 if (path.mnt->mnt_sb != sb) {
3941 path_put(&path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003942 return -EXDEV;
3943 }
Jan Kara06235432008-05-13 19:11:51 -04003944 /* Journaling quota? */
3945 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003946 /* Quotafile not in fs root? */
Al Viro82646132008-08-02 00:57:06 -04003947 if (path.dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003948 ext4_msg(sb, KERN_WARNING,
3949 "Quota file not on filesystem root. "
3950 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003951 }
Jan Kara06235432008-05-13 19:11:51 -04003952
3953 /*
3954 * When we journal data on quota file, we have to flush journal to see
3955 * all updates to the file when we bypass pagecache...
3956 */
Frank Mayhar03901312009-01-07 00:06:22 -05003957 if (EXT4_SB(sb)->s_journal &&
3958 ext4_should_journal_data(path.dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04003959 /*
3960 * We don't need to lock updates but journal_flush() could
3961 * otherwise be livelocked...
3962 */
3963 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003964 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04003965 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003966 if (err) {
Al Viro82646132008-08-02 00:57:06 -04003967 path_put(&path);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003968 return err;
3969 }
Jan Kara06235432008-05-13 19:11:51 -04003970 }
3971
Al Viro82646132008-08-02 00:57:06 -04003972 err = vfs_quota_on_path(sb, type, format_id, &path);
3973 path_put(&path);
Al Viro77e69da2008-08-01 04:29:18 -04003974 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003975}
3976
3977/* Read data from quotafile - avoid pagecache and such because we cannot afford
3978 * acquiring the locks... As quota files are never truncated and quota code
3979 * itself serializes the operations (and noone else should touch the files)
3980 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07003981static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003982 size_t len, loff_t off)
3983{
3984 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003985 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003986 int err = 0;
3987 int offset = off & (sb->s_blocksize - 1);
3988 int tocopy;
3989 size_t toread;
3990 struct buffer_head *bh;
3991 loff_t i_size = i_size_read(inode);
3992
3993 if (off > i_size)
3994 return 0;
3995 if (off+len > i_size)
3996 len = i_size-off;
3997 toread = len;
3998 while (toread > 0) {
3999 tocopy = sb->s_blocksize - offset < toread ?
4000 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07004001 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004002 if (err)
4003 return err;
4004 if (!bh) /* A hole? */
4005 memset(data, 0, tocopy);
4006 else
4007 memcpy(data, bh->b_data+offset, tocopy);
4008 brelse(bh);
4009 offset = 0;
4010 toread -= tocopy;
4011 data += tocopy;
4012 blk++;
4013 }
4014 return len;
4015}
4016
4017/* Write to quotafile (we know the transaction is already started and has
4018 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07004019static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004020 const char *data, size_t len, loff_t off)
4021{
4022 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004023 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004024 int err = 0;
4025 int offset = off & (sb->s_blocksize - 1);
Mingming Cao617ba132006-10-11 01:20:53 -07004026 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004027 struct buffer_head *bh;
4028 handle_t *handle = journal_current_handle();
4029
Frank Mayhar03901312009-01-07 00:06:22 -05004030 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004031 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4032 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07004033 (unsigned long long)off, (unsigned long long)len);
4034 return -EIO;
4035 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004036 /*
4037 * Since we account only one data block in transaction credits,
4038 * then it is impossible to cross a block boundary.
4039 */
4040 if (sb->s_blocksize - offset < len) {
4041 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4042 " cancelled because not block aligned",
4043 (unsigned long long)off, (unsigned long long)len);
4044 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004045 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004046
4047 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4048 bh = ext4_bread(handle, inode, blk, 1, &err);
4049 if (!bh)
4050 goto out;
4051 if (journal_quota) {
4052 err = ext4_journal_get_write_access(handle, bh);
4053 if (err) {
4054 brelse(bh);
4055 goto out;
4056 }
4057 }
4058 lock_buffer(bh);
4059 memcpy(bh->b_data+offset, data, len);
4060 flush_dcache_page(bh->b_page);
4061 unlock_buffer(bh);
4062 if (journal_quota)
4063 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4064 else {
4065 /* Always do at least ordered writes for quotas */
4066 err = ext4_jbd2_file_inode(handle, inode);
4067 mark_buffer_dirty(bh);
4068 }
4069 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004070out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004071 if (err) {
Jan Kara4d04e4f2008-07-04 09:59:34 -07004072 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004073 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07004074 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004075 if (inode->i_size < off + len) {
4076 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07004077 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004078 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004079 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07004080 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004081 mutex_unlock(&inode->i_mutex);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004082 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004083}
4084
4085#endif
4086
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004087static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4088 const char *dev_name, void *data, struct vfsmount *mnt)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004089{
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004090 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004091}
4092
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004093#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004094static struct file_system_type ext2_fs_type = {
4095 .owner = THIS_MODULE,
4096 .name = "ext2",
4097 .get_sb = ext4_get_sb,
4098 .kill_sb = kill_block_super,
4099 .fs_flags = FS_REQUIRES_DEV,
4100};
4101
4102static inline void register_as_ext2(void)
4103{
4104 int err = register_filesystem(&ext2_fs_type);
4105 if (err)
4106 printk(KERN_WARNING
4107 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4108}
4109
4110static inline void unregister_as_ext2(void)
4111{
4112 unregister_filesystem(&ext2_fs_type);
4113}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004114MODULE_ALIAS("ext2");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004115#else
4116static inline void register_as_ext2(void) { }
4117static inline void unregister_as_ext2(void) { }
4118#endif
4119
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004120#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004121static inline void register_as_ext3(void)
4122{
4123 int err = register_filesystem(&ext3_fs_type);
4124 if (err)
4125 printk(KERN_WARNING
4126 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4127}
4128
4129static inline void unregister_as_ext3(void)
4130{
4131 unregister_filesystem(&ext3_fs_type);
4132}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004133MODULE_ALIAS("ext3");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004134#else
4135static inline void register_as_ext3(void) { }
4136static inline void unregister_as_ext3(void) { }
4137#endif
4138
Theodore Ts'o03010a32008-10-10 20:02:48 -04004139static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004140 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04004141 .name = "ext4",
Mingming Cao617ba132006-10-11 01:20:53 -07004142 .get_sb = ext4_get_sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004143 .kill_sb = kill_block_super,
4144 .fs_flags = FS_REQUIRES_DEV,
4145};
4146
Mingming Cao617ba132006-10-11 01:20:53 -07004147static int __init init_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004148{
Alex Tomasc9de5602008-01-29 00:19:52 -05004149 int err;
4150
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004151 err = init_ext4_system_zone();
4152 if (err)
4153 return err;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004154 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4155 if (!ext4_kset)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004156 goto out4;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004157 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Alex Tomasc9de5602008-01-29 00:19:52 -05004158 err = init_ext4_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004159 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004160 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05004161
4162 err = init_ext4_xattr();
4163 if (err)
4164 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004165 err = init_inodecache();
4166 if (err)
4167 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05004168 register_as_ext2();
4169 register_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004170 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004171 if (err)
4172 goto out;
4173 return 0;
4174out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05004175 unregister_as_ext2();
4176 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004177 destroy_inodecache();
4178out1:
Mingming Cao617ba132006-10-11 01:20:53 -07004179 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004180out2:
4181 exit_ext4_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004182out3:
4183 remove_proc_entry("fs/ext4", NULL);
4184 kset_unregister(ext4_kset);
4185out4:
4186 exit_ext4_system_zone();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004187 return err;
4188}
4189
Mingming Cao617ba132006-10-11 01:20:53 -07004190static void __exit exit_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004191{
Theodore Ts'o24b58422009-12-07 14:08:51 -05004192 unregister_as_ext2();
4193 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004194 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004195 destroy_inodecache();
Mingming Cao617ba132006-10-11 01:20:53 -07004196 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004197 exit_ext4_mballoc();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004198 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004199 kset_unregister(ext4_kset);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004200 exit_ext4_system_zone();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004201}
4202
4203MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05004204MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004205MODULE_LICENSE("GPL");
Mingming Cao617ba132006-10-11 01:20:53 -07004206module_init(init_ext4_fs)
4207module_exit(exit_ext4_fs)