blob: 4d7ef31eacb1e47d11dc51409cb5a02d275a7f70 [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>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070030#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040037#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040038#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040039#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040040#include <linux/crc16.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070041#include <asm/uaccess.h>
42
Lukas Czernerbfff6872010-10-27 21:30:05 -040043#include <linux/kthread.h>
44#include <linux/freezer.h>
45
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040046#include "ext4.h"
47#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070048#include "xattr.h"
49#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040050#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070051
Theodore Ts'o9bffad12009-06-17 11:48:11 -040052#define CREATE_TRACE_POINTS
53#include <trace/events/ext4.h>
54
Theodore Ts'o1f109d52010-10-27 21:30:14 -040055static struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040056static struct kset *ext4_kset;
Lukas Czernerbfff6872010-10-27 21:30:05 -040057struct ext4_lazy_init *ext4_li_info;
58struct mutex ext4_li_mtx;
Lukas Czerner857ac882010-10-27 21:30:05 -040059struct ext4_features *ext4_feat;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040060
Mingming Cao617ba132006-10-11 01:20:53 -070061static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070062 unsigned long journal_devnum);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040063static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040064static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070068static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070070 char nbuf[16]);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040071static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080073static int ext4_unfreeze(struct super_block *sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040074static void ext4_write_super(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080075static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040076static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
Lukas Czernerbfff6872010-10-27 21:30:05 -040078static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070080
Jan Karaba69f9a2010-03-24 20:18:37 -040081#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
82static struct file_system_type ext3_fs_type = {
83 .owner = THIS_MODULE,
84 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +040085 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -040086 .kill_sb = kill_block_super,
87 .fs_flags = FS_REQUIRES_DEV,
88};
89#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
90#else
91#define IS_EXT3_SB(sb) (0)
92#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070093
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070094ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
95 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070096{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040097 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070098 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -040099 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700100}
101
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700102ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
103 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700104{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400105 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700106 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400107 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700108}
109
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700110ext4_fsblk_t ext4_inode_table(struct super_block *sb,
111 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700112{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400113 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700114 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400115 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700116}
117
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500118__u32 ext4_free_blks_count(struct super_block *sb,
119 struct ext4_group_desc *bg)
120{
121 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
122 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400123 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500124}
125
126__u32 ext4_free_inodes_count(struct super_block *sb,
127 struct ext4_group_desc *bg)
128{
129 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
130 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400131 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500132}
133
134__u32 ext4_used_dirs_count(struct super_block *sb,
135 struct ext4_group_desc *bg)
136{
137 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
138 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400139 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500140}
141
142__u32 ext4_itable_unused_count(struct super_block *sb,
143 struct ext4_group_desc *bg)
144{
145 return le16_to_cpu(bg->bg_itable_unused_lo) |
146 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400147 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500148}
149
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700150void ext4_block_bitmap_set(struct super_block *sb,
151 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700152{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400153 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700154 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
155 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700156}
157
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700158void ext4_inode_bitmap_set(struct super_block *sb,
159 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700160{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400161 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700162 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
163 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700164}
165
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700166void ext4_inode_table_set(struct super_block *sb,
167 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700168{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400169 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700170 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
171 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700172}
173
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500174void ext4_free_blks_set(struct super_block *sb,
175 struct ext4_group_desc *bg, __u32 count)
176{
177 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
178 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
179 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
180}
181
182void ext4_free_inodes_set(struct super_block *sb,
183 struct ext4_group_desc *bg, __u32 count)
184{
185 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
186 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
187 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
188}
189
190void ext4_used_dirs_set(struct super_block *sb,
191 struct ext4_group_desc *bg, __u32 count)
192{
193 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
194 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
195 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
196}
197
198void ext4_itable_unused_set(struct super_block *sb,
199 struct ext4_group_desc *bg, __u32 count)
200{
201 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
202 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
203 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
204}
205
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400206
207/* Just increment the non-pointer handle value */
208static handle_t *ext4_get_nojournal(void)
209{
210 handle_t *handle = current->journal_info;
211 unsigned long ref_cnt = (unsigned long)handle;
212
213 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
214
215 ref_cnt++;
216 handle = (handle_t *)ref_cnt;
217
218 current->journal_info = handle;
219 return handle;
220}
221
222
223/* Decrement the non-pointer handle value */
224static void ext4_put_nojournal(handle_t *handle)
225{
226 unsigned long ref_cnt = (unsigned long)handle;
227
228 BUG_ON(ref_cnt == 0);
229
230 ref_cnt--;
231 handle = (handle_t *)ref_cnt;
232
233 current->journal_info = handle;
234}
235
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700236/*
Mingming Caodab291a2006-10-11 01:21:01 -0700237 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700238 *
239 * The only special thing we need to do here is to make sure that all
240 * journal_end calls result in the superblock being marked dirty, so
241 * that sync() will call the filesystem's write_super callback if
242 * appropriate.
243 */
Mingming Cao617ba132006-10-11 01:20:53 -0700244handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700245{
246 journal_t *journal;
247
248 if (sb->s_flags & MS_RDONLY)
249 return ERR_PTR(-EROFS);
250
Eric Sandeen437f88c2010-08-01 17:33:29 -0400251 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700252 /* Special case here: if the journal has aborted behind our
253 * backs (eg. EIO in the commit thread), then we still need to
254 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700255 journal = EXT4_SB(sb)->s_journal;
Frank Mayhar03901312009-01-07 00:06:22 -0500256 if (journal) {
257 if (is_journal_aborted(journal)) {
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400258 ext4_abort(sb, "Detected aborted journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500259 return ERR_PTR(-EROFS);
260 }
261 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700262 }
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400263 return ext4_get_nojournal();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700264}
265
266/*
267 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700268 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700269 * that sync() will call the filesystem's write_super callback if
270 * appropriate.
271 */
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400272int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700273{
274 struct super_block *sb;
275 int err;
276 int rc;
277
Frank Mayhar03901312009-01-07 00:06:22 -0500278 if (!ext4_handle_valid(handle)) {
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400279 ext4_put_nojournal(handle);
Frank Mayhar03901312009-01-07 00:06:22 -0500280 return 0;
281 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700282 sb = handle->h_transaction->t_journal->j_private;
283 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700284 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700285
286 if (!err)
287 err = rc;
288 if (err)
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400289 __ext4_std_error(sb, where, line, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700290 return err;
291}
292
Theodore Ts'o90c72012010-06-29 14:53:24 -0400293void ext4_journal_abort_handle(const char *caller, unsigned int line,
294 const char *err_fn, struct buffer_head *bh,
295 handle_t *handle, int err)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700296{
297 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700298 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700299
Frank Mayhar03901312009-01-07 00:06:22 -0500300 BUG_ON(!ext4_handle_valid(handle));
301
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700302 if (bh)
303 BUFFER_TRACE(bh, "abort");
304
305 if (!handle->h_err)
306 handle->h_err = err;
307
308 if (is_handle_aborted(handle))
309 return;
310
Theodore Ts'o90c72012010-06-29 14:53:24 -0400311 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
312 caller, line, errstr, err_fn);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700313
Mingming Caodab291a2006-10-11 01:21:01 -0700314 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700315}
316
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400317static void __save_error_info(struct super_block *sb, const char *func,
318 unsigned int line)
319{
320 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
321
322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
324 es->s_last_error_time = cpu_to_le32(get_seconds());
325 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
326 es->s_last_error_line = cpu_to_le32(line);
327 if (!es->s_first_error_time) {
328 es->s_first_error_time = es->s_last_error_time;
329 strncpy(es->s_first_error_func, func,
330 sizeof(es->s_first_error_func));
331 es->s_first_error_line = cpu_to_le32(line);
332 es->s_first_error_ino = es->s_last_error_ino;
333 es->s_first_error_block = es->s_last_error_block;
334 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400335 /*
336 * Start the daily error reporting function if it hasn't been
337 * started already
338 */
339 if (!es->s_error_count)
340 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400341 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
342}
343
344static void save_error_info(struct super_block *sb, const char *func,
345 unsigned int line)
346{
347 __save_error_info(sb, func, line);
348 ext4_commit_super(sb, 1);
349}
350
351
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700352/* Deal with the reporting of failure conditions on a filesystem such as
353 * inconsistencies detected or read IO failures.
354 *
355 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700356 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700357 * write ordering constraints on the superblock which prevent us from
358 * writing it out straight away; and given that the journal is about to
359 * be aborted, we can't rely on the current, or future, transactions to
360 * write out the superblock safely.
361 *
Mingming Caodab291a2006-10-11 01:21:01 -0700362 * 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 -0200363 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700364 * that error until we've noted it down and cleared it.
365 */
366
Mingming Cao617ba132006-10-11 01:20:53 -0700367static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700368{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700369 if (sb->s_flags & MS_RDONLY)
370 return;
371
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400372 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700373 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700374
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400375 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700376 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700377 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700378 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400379 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400380 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700381 sb->s_flags |= MS_RDONLY;
382 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700383 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700384 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700385 sb->s_id);
386}
387
Eric Sandeen12062dd2010-02-15 14:19:27 -0500388void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400389 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700390{
391 va_list args;
392
393 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400394 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
395 sb->s_id, function, line, current->comm);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700396 vprintk(fmt, args);
397 printk("\n");
398 va_end(args);
399
Mingming Cao617ba132006-10-11 01:20:53 -0700400 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700401}
402
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400403void ext4_error_inode(struct inode *inode, const char *function,
404 unsigned int line, ext4_fsblk_t block,
Frank Mayhar273df552010-03-02 11:46:09 -0500405 const char *fmt, ...)
406{
407 va_list args;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400408 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500409
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400410 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
411 es->s_last_error_block = cpu_to_le64(block);
412 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500413 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400414 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
415 inode->i_sb->s_id, function, line, inode->i_ino);
416 if (block)
417 printk("block %llu: ", block);
418 printk("comm %s: ", current->comm);
Frank Mayhar273df552010-03-02 11:46:09 -0500419 vprintk(fmt, args);
420 printk("\n");
421 va_end(args);
422
423 ext4_handle_error(inode->i_sb);
424}
425
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400426void ext4_error_file(struct file *file, const char *function,
427 unsigned int line, const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500428{
429 va_list args;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400430 struct ext4_super_block *es;
Frank Mayhar273df552010-03-02 11:46:09 -0500431 struct inode *inode = file->f_dentry->d_inode;
432 char pathname[80], *path;
433
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400434 es = EXT4_SB(inode->i_sb)->s_es;
435 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
436 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500437 va_start(args, fmt);
438 path = d_path(&(file->f_path), pathname, sizeof(pathname));
439 if (!path)
440 path = "(unknown)";
441 printk(KERN_CRIT
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400442 "EXT4-fs error (device %s): %s:%d: inode #%lu "
443 "(comm %s path %s): ",
444 inode->i_sb->s_id, function, line, inode->i_ino,
445 current->comm, path);
Frank Mayhar273df552010-03-02 11:46:09 -0500446 vprintk(fmt, args);
447 printk("\n");
448 va_end(args);
449
450 ext4_handle_error(inode->i_sb);
451}
452
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400453static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700454 char nbuf[16])
455{
456 char *errstr = NULL;
457
458 switch (errno) {
459 case -EIO:
460 errstr = "IO failure";
461 break;
462 case -ENOMEM:
463 errstr = "Out of memory";
464 break;
465 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400466 if (!sb || (EXT4_SB(sb)->s_journal &&
467 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700468 errstr = "Journal has aborted";
469 else
470 errstr = "Readonly filesystem";
471 break;
472 default:
473 /* If the caller passed in an extra buffer for unknown
474 * errors, textualise them now. Else we just return
475 * NULL. */
476 if (nbuf) {
477 /* Check for truncated error codes... */
478 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
479 errstr = nbuf;
480 }
481 break;
482 }
483
484 return errstr;
485}
486
Mingming Cao617ba132006-10-11 01:20:53 -0700487/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700488 * automatically and invokes the appropriate error response. */
489
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400490void __ext4_std_error(struct super_block *sb, const char *function,
491 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700492{
493 char nbuf[16];
494 const char *errstr;
495
496 /* Special case: if the error is EROFS, and we're not already
497 * inside a transaction, then there's really no point in logging
498 * an error. */
499 if (errno == -EROFS && journal_current_handle() == NULL &&
500 (sb->s_flags & MS_RDONLY))
501 return;
502
Mingming Cao617ba132006-10-11 01:20:53 -0700503 errstr = ext4_decode_error(sb, errno, nbuf);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400504 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
505 sb->s_id, function, line, errstr);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400506 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700507
Mingming Cao617ba132006-10-11 01:20:53 -0700508 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700509}
510
511/*
Mingming Cao617ba132006-10-11 01:20:53 -0700512 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700513 * abort function may be used to deal with unrecoverable failures such
514 * as journal IO errors or ENOMEM at a critical moment in log management.
515 *
516 * We unconditionally force the filesystem into an ABORT|READONLY state,
517 * unless the error response on the fs has been set to panic in which
518 * case we take the easy way out and panic immediately.
519 */
520
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400521void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400522 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700523{
524 va_list args;
525
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400526 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700527 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400528 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
529 function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700530 vprintk(fmt, args);
531 printk("\n");
532 va_end(args);
533
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400534 if ((sb->s_flags & MS_RDONLY) == 0) {
535 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
536 sb->s_flags |= MS_RDONLY;
537 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
538 if (EXT4_SB(sb)->s_journal)
539 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
540 save_error_info(sb, function, line);
541 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700542 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700543 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700544}
545
Eric Sandeenb31e1552009-06-04 17:36:36 -0400546void ext4_msg (struct super_block * sb, const char *prefix,
547 const char *fmt, ...)
548{
549 va_list args;
550
551 va_start(args, fmt);
552 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
553 vprintk(fmt, args);
554 printk("\n");
555 va_end(args);
556}
557
Eric Sandeen12062dd2010-02-15 14:19:27 -0500558void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400559 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700560{
561 va_list args;
562
563 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400564 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
565 sb->s_id, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700566 vprintk(fmt, args);
567 printk("\n");
568 va_end(args);
569}
570
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400571void __ext4_grp_locked_error(const char *function, unsigned int line,
572 struct super_block *sb, ext4_group_t grp,
573 unsigned long ino, ext4_fsblk_t block,
574 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500575__releases(bitlock)
576__acquires(bitlock)
577{
578 va_list args;
579 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
580
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400581 es->s_last_error_ino = cpu_to_le32(ino);
582 es->s_last_error_block = cpu_to_le64(block);
583 __save_error_info(sb, function, line);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500584 va_start(args, fmt);
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400585 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
586 sb->s_id, function, line, grp);
587 if (ino)
588 printk("inode %lu: ", ino);
589 if (block)
590 printk("block %llu:", (unsigned long long) block);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500591 vprintk(fmt, args);
592 printk("\n");
593 va_end(args);
594
595 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400596 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500597 return;
598 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400599
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500600 ext4_unlock_group(sb, grp);
601 ext4_handle_error(sb);
602 /*
603 * We only get here in the ERRORS_RO case; relocking the group
604 * may be dangerous, but nothing bad will happen since the
605 * filesystem will have already been marked read/only and the
606 * journal has been aborted. We return 1 as a hint to callers
607 * who might what to use the return value from
608 * ext4_grp_locked_error() to distinguish beween the
609 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
610 * aggressively from the ext4 function in question, with a
611 * more appropriate error code.
612 */
613 ext4_lock_group(sb, grp);
614 return;
615}
616
Mingming Cao617ba132006-10-11 01:20:53 -0700617void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700618{
Mingming Cao617ba132006-10-11 01:20:53 -0700619 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700620
Mingming Cao617ba132006-10-11 01:20:53 -0700621 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700622 return;
623
Eric Sandeen12062dd2010-02-15 14:19:27 -0500624 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700625 "updating to rev %d because of new feature flag, "
626 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700627 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700628
Mingming Cao617ba132006-10-11 01:20:53 -0700629 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
630 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
631 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700632 /* leave es->s_feature_*compat flags alone */
633 /* es->s_uuid will be set by e2fsck if empty */
634
635 /*
636 * The rest of the superblock fields should be zero, and if not it
637 * means they are likely already in use, so leave them alone. We
638 * can leave it up to e2fsck to clean up any inconsistencies there.
639 */
640}
641
642/*
643 * Open the external journal device
644 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400645static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700646{
647 struct block_device *bdev;
648 char b[BDEVNAME_SIZE];
649
650 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
651 if (IS_ERR(bdev))
652 goto fail;
653 return bdev;
654
655fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400656 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700657 __bdevname(dev, b), PTR_ERR(bdev));
658 return NULL;
659}
660
661/*
662 * Release the journal device
663 */
Mingming Cao617ba132006-10-11 01:20:53 -0700664static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700665{
666 bd_release(bdev);
Al Viro9a1c3542008-02-22 20:40:24 -0500667 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700668}
669
Mingming Cao617ba132006-10-11 01:20:53 -0700670static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700671{
672 struct block_device *bdev;
673 int ret = -ENODEV;
674
675 bdev = sbi->journal_bdev;
676 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700677 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700678 sbi->journal_bdev = NULL;
679 }
680 return ret;
681}
682
683static inline struct inode *orphan_list_entry(struct list_head *l)
684{
Mingming Cao617ba132006-10-11 01:20:53 -0700685 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700686}
687
Mingming Cao617ba132006-10-11 01:20:53 -0700688static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700689{
690 struct list_head *l;
691
Eric Sandeenb31e1552009-06-04 17:36:36 -0400692 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
693 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700694
695 printk(KERN_ERR "sb_info orphan list:\n");
696 list_for_each(l, &sbi->s_orphan) {
697 struct inode *inode = orphan_list_entry(l);
698 printk(KERN_ERR " "
699 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
700 inode->i_sb->s_id, inode->i_ino, inode,
701 inode->i_mode, inode->i_nlink,
702 NEXT_ORPHAN(inode));
703 }
704}
705
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400706static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700707{
Mingming Cao617ba132006-10-11 01:20:53 -0700708 struct ext4_sb_info *sbi = EXT4_SB(sb);
709 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400710 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700711
Lukas Czerner857ac882010-10-27 21:30:05 -0400712 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400713 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
714
Mingming Cao4c0425f2009-09-28 15:48:41 -0400715 flush_workqueue(sbi->dio_unwritten_wq);
716 destroy_workqueue(sbi->dio_unwritten_wq);
717
Al Viroa9e220f2009-05-05 22:10:44 -0400718 lock_super(sb);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200719 if (sb->s_dirt)
Christoph Hellwigebc1ac12009-05-11 23:35:03 +0200720 ext4_commit_super(sb, 1);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200721
Frank Mayhar03901312009-01-07 00:06:22 -0500722 if (sbi->s_journal) {
723 err = jbd2_journal_destroy(sbi->s_journal);
724 sbi->s_journal = NULL;
725 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400726 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500727 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500728
Sergey Senozhatskya1c6c562010-10-27 21:30:04 -0400729 del_timer(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500730 ext4_release_system_zone(sb);
731 ext4_mb_release(sb);
732 ext4_ext_release(sb);
733 ext4_xattr_put_super(sb);
734
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700735 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700736 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700737 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400738 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700739 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400740 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400741 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400742 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400743 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700744
745 for (i = 0; i < sbi->s_gdb_count; i++)
746 brelse(sbi->s_group_desc[i]);
747 kfree(sbi->s_group_desc);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -0400748 if (is_vmalloc_addr(sbi->s_flex_groups))
749 vfree(sbi->s_flex_groups);
750 else
751 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700752 percpu_counter_destroy(&sbi->s_freeblocks_counter);
753 percpu_counter_destroy(&sbi->s_freeinodes_counter);
754 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400755 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700756 brelse(sbi->s_sbh);
757#ifdef CONFIG_QUOTA
758 for (i = 0; i < MAXQUOTAS; i++)
759 kfree(sbi->s_qf_names[i]);
760#endif
761
762 /* Debugging code just in case the in-memory inode orphan list
763 * isn't empty. The on-disk one can be non-empty if we've
764 * detected an error and taken the fs readonly, but the
765 * in-memory list had better be clean by this point. */
766 if (!list_empty(&sbi->s_orphan))
767 dump_orphan_list(sb, sbi);
768 J_ASSERT(list_empty(&sbi->s_orphan));
769
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700770 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700771 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
772 /*
773 * Invalidate the journal device's buffers. We don't want them
774 * floating about in memory - the physical journal device may
775 * hotswapped, and it breaks the `ro-after' testing code.
776 */
777 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700778 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700779 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700780 }
781 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400782 /*
783 * Now that we are completely done shutting down the
784 * superblock, we need to actually destroy the kobject.
785 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400786 unlock_super(sb);
787 kobject_put(&sbi->s_kobj);
788 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500789 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700790 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700791}
792
Christoph Lametere18b8902006-12-06 20:33:20 -0800793static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700794
795/*
796 * Called inside transaction, so use GFP_NOFS
797 */
Mingming Cao617ba132006-10-11 01:20:53 -0700798static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700799{
Mingming Cao617ba132006-10-11 01:20:53 -0700800 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700801
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800802 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700803 if (!ei)
804 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400805
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700806 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400807 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700808 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500809 INIT_LIST_HEAD(&ei->i_prealloc_list);
810 spin_lock_init(&ei->i_prealloc_lock);
Frank Mayhar03901312009-01-07 00:06:22 -0500811 /*
812 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
813 * therefore it can be null here. Don't check it, just initialize
814 * jinode.
815 */
Jan Kara678aaf42008-07-11 19:27:31 -0400816 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
Mingming Caod2a17632008-07-14 17:52:37 -0400817 ei->i_reserved_data_blocks = 0;
818 ei->i_reserved_meta_blocks = 0;
819 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500820 ei->i_da_metadata_calc_len = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400821 ei->i_delalloc_reserved_flag = 0;
822 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300823#ifdef CONFIG_QUOTA
824 ei->i_reserved_quota = 0;
825#endif
Jiaying Zhangc7064ef2010-03-02 13:28:44 -0500826 INIT_LIST_HEAD(&ei->i_completed_io_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500827 spin_lock_init(&ei->i_completed_io_lock);
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400828 ei->cur_aio_dio = NULL;
Jan Karab436b9b2009-12-08 23:51:10 -0500829 ei->i_sync_tid = 0;
830 ei->i_datasync_tid = 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400831
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700832 return &ei->vfs_inode;
833}
834
Mingming Cao617ba132006-10-11 01:20:53 -0700835static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700836{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700837 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400838 ext4_msg(inode->i_sb, KERN_ERR,
839 "Inode %lu (%p): orphan list check failed!",
840 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700841 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
842 EXT4_I(inode), sizeof(struct ext4_inode_info),
843 true);
844 dump_stack();
845 }
Mingming Cao617ba132006-10-11 01:20:53 -0700846 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700847}
848
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700849static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700850{
Mingming Cao617ba132006-10-11 01:20:53 -0700851 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700852
Christoph Lametera35afb82007-05-16 22:10:57 -0700853 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400854#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700855 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700856#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500857 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700858 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700859}
860
861static int init_inodecache(void)
862{
Mingming Cao617ba132006-10-11 01:20:53 -0700863 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
864 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700865 0, (SLAB_RECLAIM_ACCOUNT|
866 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900867 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700868 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700869 return -ENOMEM;
870 return 0;
871}
872
873static void destroy_inodecache(void)
874{
Mingming Cao617ba132006-10-11 01:20:53 -0700875 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700876}
877
Al Viro0930fcc2010-06-07 13:16:22 -0400878void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700879{
Al Viro0930fcc2010-06-07 13:16:22 -0400880 invalidate_inode_buffers(inode);
881 end_writeback(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -0500882 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400883 ext4_discard_preallocations(inode);
Frank Mayhar03901312009-01-07 00:06:22 -0500884 if (EXT4_JOURNAL(inode))
885 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
Jan Kara678aaf42008-07-11 19:27:31 -0400886 &EXT4_I(inode)->jinode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700887}
888
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400889static inline void ext4_show_quota_options(struct seq_file *seq,
890 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700891{
892#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700893 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700894
Jan Kara5a20bdf2009-11-30 23:58:32 +0100895 if (sbi->s_jquota_fmt) {
896 char *fmtname = "";
897
898 switch (sbi->s_jquota_fmt) {
899 case QFMT_VFS_OLD:
900 fmtname = "vfsold";
901 break;
902 case QFMT_VFS_V0:
903 fmtname = "vfsv0";
904 break;
905 case QFMT_VFS_V1:
906 fmtname = "vfsv1";
907 break;
908 }
909 seq_printf(seq, ",jqfmt=%s", fmtname);
910 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700911
912 if (sbi->s_qf_names[USRQUOTA])
913 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
914
915 if (sbi->s_qf_names[GRPQUOTA])
916 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
917
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500918 if (test_opt(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700919 seq_puts(seq, ",usrquota");
920
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500921 if (test_opt(sb, GRPQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700922 seq_puts(seq, ",grpquota");
923#endif
924}
925
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700926/*
927 * Show an option if
928 * - it's set to a non-default value OR
929 * - if the per-sb default is different from the global default
930 */
Mingming Cao617ba132006-10-11 01:20:53 -0700931static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700932{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500933 int def_errors;
934 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700935 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700936 struct ext4_sb_info *sbi = EXT4_SB(sb);
937 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700938
939 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500940 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700941
942 if (sbi->s_sb_block != 1)
943 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
944 if (test_opt(sb, MINIX_DF))
945 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500946 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700947 seq_puts(seq, ",grpid");
948 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
949 seq_puts(seq, ",nogrpid");
950 if (sbi->s_resuid != EXT4_DEF_RESUID ||
951 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
952 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
953 }
954 if (sbi->s_resgid != EXT4_DEF_RESGID ||
955 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
956 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
957 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500958 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700959 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500960 def_errors == EXT4_ERRORS_CONTINUE) {
961 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700962 }
963 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500964 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500965 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500966 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700967 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500968 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700969 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500970 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700971 seq_puts(seq, ",debug");
972 if (test_opt(sb, OLDALLOC))
973 seq_puts(seq, ",oldalloc");
Theodore Ts'o03010a32008-10-10 20:02:48 -0400974#ifdef CONFIG_EXT4_FS_XATTR
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500975 if (test_opt(sb, XATTR_USER) &&
976 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700977 seq_puts(seq, ",user_xattr");
978 if (!test_opt(sb, XATTR_USER) &&
979 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
980 seq_puts(seq, ",nouser_xattr");
981 }
982#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -0400983#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500984 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700985 seq_puts(seq, ",acl");
986 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
987 seq_puts(seq, ",noacl");
988#endif
Theodore Ts'o30773842009-01-03 20:27:38 -0500989 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700990 seq_printf(seq, ",commit=%u",
991 (unsigned) (sbi->s_commit_interval / HZ));
992 }
Theodore Ts'o30773842009-01-03 20:27:38 -0500993 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
994 seq_printf(seq, ",min_batch_time=%u",
995 (unsigned) sbi->s_min_batch_time);
996 }
997 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
998 seq_printf(seq, ",max_batch_time=%u",
999 (unsigned) sbi->s_min_batch_time);
1000 }
1001
Eric Sandeen571640c2008-05-26 12:29:46 -04001002 /*
1003 * We're changing the default of barrier mount option, so
1004 * let's always display its mount state so it's clear what its
1005 * status is.
1006 */
1007 seq_puts(seq, ",barrier=");
1008 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -04001009 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1010 seq_puts(seq, ",journal_async_commit");
Jan Kara39a4bad2010-05-16 17:00:00 -04001011 else if (test_opt(sb, JOURNAL_CHECKSUM))
1012 seq_puts(seq, ",journal_checksum");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001013 if (test_opt(sb, I_VERSION))
1014 seq_puts(seq, ",i_version");
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001015 if (!test_opt(sb, DELALLOC) &&
1016 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001017 seq_puts(seq, ",nodelalloc");
1018
Miklos Szeredicb45bbe2008-01-28 23:58:27 -05001019 if (sbi->s_stripe)
1020 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001021 /*
1022 * journal mode get enabled in different ways
1023 * So just print the value even if we didn't specify it
1024 */
Mingming Cao617ba132006-10-11 01:20:53 -07001025 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001026 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -07001027 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001028 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -07001029 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001030 seq_puts(seq, ",data=writeback");
1031
Theodore Ts'o240799c2008-10-09 23:53:47 -04001032 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1033 seq_printf(seq, ",inode_readahead_blks=%u",
1034 sbi->s_inode_readahead_blks);
1035
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001036 if (test_opt(sb, DATA_ERR_ABORT))
1037 seq_puts(seq, ",data_err=abort");
1038
Theodore Ts'oafd46722009-03-16 23:12:23 -04001039 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001040 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -04001041
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001042 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
Eric Sandeen5328e632009-11-19 14:25:42 -05001043 seq_puts(seq, ",discard");
1044
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001045 if (test_opt(sb, NOLOAD))
1046 seq_puts(seq, ",norecovery");
1047
Jiaying Zhang744692d2010-03-04 16:14:02 -05001048 if (test_opt(sb, DIOREAD_NOLOCK))
1049 seq_puts(seq, ",dioread_nolock");
1050
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001051 if (test_opt(sb, BLOCK_VALIDITY) &&
1052 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1053 seq_puts(seq, ",block_validity");
1054
Lukas Czernerbfff6872010-10-27 21:30:05 -04001055 if (!test_opt(sb, INIT_INODE_TABLE))
1056 seq_puts(seq, ",noinit_inode_table");
1057 else if (sbi->s_li_wait_mult)
1058 seq_printf(seq, ",init_inode_table=%u",
1059 (unsigned) sbi->s_li_wait_mult);
1060
Mingming Cao617ba132006-10-11 01:20:53 -07001061 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001062
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001063 return 0;
1064}
1065
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001066static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001067 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001068{
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001069 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001070
Mingming Cao617ba132006-10-11 01:20:53 -07001071 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001072 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -07001073 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001074 return ERR_PTR(-ESTALE);
1075
1076 /* iget isn't really right if the inode is currently unallocated!!
1077 *
Mingming Cao617ba132006-10-11 01:20:53 -07001078 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001079 * deleted, so we should be safe.
1080 *
1081 * Currently we don't know the generation for parent directory, so
1082 * a generation of 0 means "accept any"
1083 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001084 inode = ext4_iget(sb, ino);
1085 if (IS_ERR(inode))
1086 return ERR_CAST(inode);
1087 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001088 iput(inode);
1089 return ERR_PTR(-ESTALE);
1090 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001091
1092 return inode;
1093}
1094
1095static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001096 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001097{
1098 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1099 ext4_nfs_get_inode);
1100}
1101
1102static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001103 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001104{
1105 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1106 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001107}
1108
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001109/*
1110 * Try to release metadata pages (indirect blocks, directories) which are
1111 * mapped via the block device. Since these pages could have journal heads
1112 * which would prevent try_to_free_buffers() from freeing them, we must use
1113 * jbd2 layer's try_to_free_buffers() function to release them.
1114 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001115static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1116 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001117{
1118 journal_t *journal = EXT4_SB(sb)->s_journal;
1119
1120 WARN_ON(PageChecked(page));
1121 if (!page_has_buffers(page))
1122 return 0;
1123 if (journal)
1124 return jbd2_journal_try_to_free_buffers(journal, page,
1125 wait & ~__GFP_WAIT);
1126 return try_to_free_buffers(page);
1127}
1128
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001129#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001130#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001131#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001132
Mingming Cao617ba132006-10-11 01:20:53 -07001133static int ext4_write_dquot(struct dquot *dquot);
1134static int ext4_acquire_dquot(struct dquot *dquot);
1135static int ext4_release_dquot(struct dquot *dquot);
1136static int ext4_mark_dquot_dirty(struct dquot *dquot);
1137static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001138static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Christoph Hellwig307ae182010-05-19 07:16:43 -04001139 char *path);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001140static int ext4_quota_off(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001141static int ext4_quota_on_mount(struct super_block *sb, int type);
1142static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001143 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001144static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001145 const char *data, size_t len, loff_t off);
1146
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001147static const struct dquot_operations ext4_quota_operations = {
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001148#ifdef CONFIG_QUOTA
Mingming Cao60e58e02009-01-22 18:13:05 +01001149 .get_reserved_space = ext4_get_reserved_space,
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001150#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001151 .write_dquot = ext4_write_dquot,
1152 .acquire_dquot = ext4_acquire_dquot,
1153 .release_dquot = ext4_release_dquot,
1154 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001155 .write_info = ext4_write_info,
1156 .alloc_dquot = dquot_alloc,
1157 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001158};
1159
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001160static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001161 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001162 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001163 .quota_sync = dquot_quota_sync,
1164 .get_info = dquot_get_dqinfo,
1165 .set_info = dquot_set_dqinfo,
1166 .get_dqblk = dquot_get_dqblk,
1167 .set_dqblk = dquot_set_dqblk
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001168};
1169#endif
1170
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001171static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001172 .alloc_inode = ext4_alloc_inode,
1173 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001174 .write_inode = ext4_write_inode,
1175 .dirty_inode = ext4_dirty_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001176 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001177 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001178 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001179 .freeze_fs = ext4_freeze,
1180 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001181 .statfs = ext4_statfs,
1182 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001183 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001184#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001185 .quota_read = ext4_quota_read,
1186 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001187#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001188 .bdev_try_to_free_page = bdev_try_to_free_page,
Lukas Czerner7360d172010-10-27 21:30:12 -04001189 .trim_fs = ext4_trim_fs
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001190};
1191
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001192static const struct super_operations ext4_nojournal_sops = {
1193 .alloc_inode = ext4_alloc_inode,
1194 .destroy_inode = ext4_destroy_inode,
1195 .write_inode = ext4_write_inode,
1196 .dirty_inode = ext4_dirty_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001197 .evict_inode = ext4_evict_inode,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001198 .write_super = ext4_write_super,
1199 .put_super = ext4_put_super,
1200 .statfs = ext4_statfs,
1201 .remount_fs = ext4_remount,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001202 .show_options = ext4_show_options,
1203#ifdef CONFIG_QUOTA
1204 .quota_read = ext4_quota_read,
1205 .quota_write = ext4_quota_write,
1206#endif
1207 .bdev_try_to_free_page = bdev_try_to_free_page,
1208};
1209
Christoph Hellwig39655162007-10-21 16:42:17 -07001210static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001211 .fh_to_dentry = ext4_fh_to_dentry,
1212 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001213 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001214};
1215
1216enum {
1217 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1218 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001219 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001220 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001221 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001222 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001223 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001224 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001225 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001226 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001227 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001228 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1229 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1230 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001231 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001232 Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001233 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001234 Opt_dioread_nolock, Opt_dioread_lock,
Eric Sandeen5328e632009-11-19 14:25:42 -05001235 Opt_discard, Opt_nodiscard,
Lukas Czernerbfff6872010-10-27 21:30:05 -04001236 Opt_init_inode_table, Opt_noinit_inode_table,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001237};
1238
Steven Whitehousea447c092008-10-13 10:46:57 +01001239static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001240 {Opt_bsd_df, "bsddf"},
1241 {Opt_minix_df, "minixdf"},
1242 {Opt_grpid, "grpid"},
1243 {Opt_grpid, "bsdgroups"},
1244 {Opt_nogrpid, "nogrpid"},
1245 {Opt_nogrpid, "sysvgroups"},
1246 {Opt_resgid, "resgid=%u"},
1247 {Opt_resuid, "resuid=%u"},
1248 {Opt_sb, "sb=%u"},
1249 {Opt_err_cont, "errors=continue"},
1250 {Opt_err_panic, "errors=panic"},
1251 {Opt_err_ro, "errors=remount-ro"},
1252 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001253 {Opt_debug, "debug"},
1254 {Opt_oldalloc, "oldalloc"},
1255 {Opt_orlov, "orlov"},
1256 {Opt_user_xattr, "user_xattr"},
1257 {Opt_nouser_xattr, "nouser_xattr"},
1258 {Opt_acl, "acl"},
1259 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001260 {Opt_noload, "noload"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001261 {Opt_noload, "norecovery"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001262 {Opt_nobh, "nobh"},
1263 {Opt_bh, "bh"},
1264 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001265 {Opt_min_batch_time, "min_batch_time=%u"},
1266 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001267 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001268 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001269 {Opt_journal_checksum, "journal_checksum"},
1270 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001271 {Opt_abort, "abort"},
1272 {Opt_data_journal, "data=journal"},
1273 {Opt_data_ordered, "data=ordered"},
1274 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001275 {Opt_data_err_abort, "data_err=abort"},
1276 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001277 {Opt_offusrjquota, "usrjquota="},
1278 {Opt_usrjquota, "usrjquota=%s"},
1279 {Opt_offgrpjquota, "grpjquota="},
1280 {Opt_grpjquota, "grpjquota=%s"},
1281 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1282 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001283 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001284 {Opt_grpquota, "grpquota"},
1285 {Opt_noquota, "noquota"},
1286 {Opt_quota, "quota"},
1287 {Opt_usrquota, "usrquota"},
1288 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001289 {Opt_barrier, "barrier"},
1290 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001291 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001292 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001293 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001294 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001295 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001296 {Opt_block_validity, "block_validity"},
1297 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001298 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001299 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001300 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001301 {Opt_auto_da_alloc, "auto_da_alloc"},
1302 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001303 {Opt_dioread_nolock, "dioread_nolock"},
1304 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001305 {Opt_discard, "discard"},
1306 {Opt_nodiscard, "nodiscard"},
Lukas Czernerbfff6872010-10-27 21:30:05 -04001307 {Opt_init_inode_table, "init_itable=%u"},
1308 {Opt_init_inode_table, "init_itable"},
1309 {Opt_noinit_inode_table, "noinit_itable"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001310 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001311};
1312
Mingming Cao617ba132006-10-11 01:20:53 -07001313static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001314{
Mingming Cao617ba132006-10-11 01:20:53 -07001315 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001316 char *options = (char *) *data;
1317
1318 if (!options || strncmp(options, "sb=", 3) != 0)
1319 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001320
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001321 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001322 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001323 sb_block = simple_strtoul(options, &options, 0);
1324 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001325 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001326 (char *) *data);
1327 return 1;
1328 }
1329 if (*options == ',')
1330 options++;
1331 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001332
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001333 return sb_block;
1334}
1335
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001336#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001337static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1338 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001339
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001340#ifdef CONFIG_QUOTA
1341static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1342{
1343 struct ext4_sb_info *sbi = EXT4_SB(sb);
1344 char *qname;
1345
1346 if (sb_any_quota_loaded(sb) &&
1347 !sbi->s_qf_names[qtype]) {
1348 ext4_msg(sb, KERN_ERR,
1349 "Cannot change journaled "
1350 "quota options when quota turned on");
1351 return 0;
1352 }
1353 qname = match_strdup(args);
1354 if (!qname) {
1355 ext4_msg(sb, KERN_ERR,
1356 "Not enough memory for storing quotafile name");
1357 return 0;
1358 }
1359 if (sbi->s_qf_names[qtype] &&
1360 strcmp(sbi->s_qf_names[qtype], qname)) {
1361 ext4_msg(sb, KERN_ERR,
1362 "%s quota file already specified", QTYPE2NAME(qtype));
1363 kfree(qname);
1364 return 0;
1365 }
1366 sbi->s_qf_names[qtype] = qname;
1367 if (strchr(sbi->s_qf_names[qtype], '/')) {
1368 ext4_msg(sb, KERN_ERR,
1369 "quotafile must be on filesystem root");
1370 kfree(sbi->s_qf_names[qtype]);
1371 sbi->s_qf_names[qtype] = NULL;
1372 return 0;
1373 }
1374 set_opt(sbi->s_mount_opt, QUOTA);
1375 return 1;
1376}
1377
1378static int clear_qf_name(struct super_block *sb, int qtype)
1379{
1380
1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
1382
1383 if (sb_any_quota_loaded(sb) &&
1384 sbi->s_qf_names[qtype]) {
1385 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1386 " when quota turned on");
1387 return 0;
1388 }
1389 /*
1390 * The space will be released later when all options are confirmed
1391 * to be correct
1392 */
1393 sbi->s_qf_names[qtype] = NULL;
1394 return 1;
1395}
1396#endif
1397
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001398static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001399 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001400 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001401 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001402{
Mingming Cao617ba132006-10-11 01:20:53 -07001403 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001404 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001405 substring_t args[MAX_OPT_ARGS];
1406 int data_opt = 0;
1407 int option;
1408#ifdef CONFIG_QUOTA
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001409 int qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001410#endif
1411
1412 if (!options)
1413 return 1;
1414
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001415 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001416 int token;
1417 if (!*p)
1418 continue;
1419
Eric Sandeen15121c12010-02-15 20:17:55 -05001420 /*
1421 * Initialize args struct so we know whether arg was
1422 * found; some options take optional arguments.
1423 */
1424 args[0].to = args[0].from = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001425 token = match_token(p, tokens, args);
1426 switch (token) {
1427 case Opt_bsd_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001428 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001429 clear_opt(sbi->s_mount_opt, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001430 break;
1431 case Opt_minix_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001432 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001433 set_opt(sbi->s_mount_opt, MINIX_DF);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001434
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001435 break;
1436 case Opt_grpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001437 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001438 set_opt(sbi->s_mount_opt, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001439
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001440 break;
1441 case Opt_nogrpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001442 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001443 clear_opt(sbi->s_mount_opt, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001444
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001445 break;
1446 case Opt_resuid:
1447 if (match_int(&args[0], &option))
1448 return 0;
1449 sbi->s_resuid = option;
1450 break;
1451 case Opt_resgid:
1452 if (match_int(&args[0], &option))
1453 return 0;
1454 sbi->s_resgid = option;
1455 break;
1456 case Opt_sb:
1457 /* handled by get_sb_block() instead of here */
1458 /* *sb_block = match_int(&args[0]); */
1459 break;
1460 case Opt_err_panic:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001461 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1462 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1463 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001464 break;
1465 case Opt_err_ro:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001466 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1467 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1468 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001469 break;
1470 case Opt_err_cont:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001471 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1472 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1473 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001474 break;
1475 case Opt_nouid32:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001476 set_opt(sbi->s_mount_opt, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001477 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001478 case Opt_debug:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001479 set_opt(sbi->s_mount_opt, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001480 break;
1481 case Opt_oldalloc:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001482 set_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001483 break;
1484 case Opt_orlov:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001485 clear_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001486 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001487#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001488 case Opt_user_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001489 set_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001490 break;
1491 case Opt_nouser_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001492 clear_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001493 break;
1494#else
1495 case Opt_user_xattr:
1496 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001497 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001498 break;
1499#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001500#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001501 case Opt_acl:
1502 set_opt(sbi->s_mount_opt, POSIX_ACL);
1503 break;
1504 case Opt_noacl:
1505 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1506 break;
1507#else
1508 case Opt_acl:
1509 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001510 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001511 break;
1512#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001513 case Opt_journal_update:
1514 /* @@@ FIXME */
1515 /* Eventually we will want to be able to create
1516 a journal file here. For now, only allow the
1517 user to specify an existing inode to be the
1518 journal file. */
1519 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001520 ext4_msg(sb, KERN_ERR,
1521 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001522 return 0;
1523 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001524 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001525 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001526 case Opt_journal_dev:
1527 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001528 ext4_msg(sb, KERN_ERR,
1529 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001530 return 0;
1531 }
1532 if (match_int(&args[0], &option))
1533 return 0;
1534 *journal_devnum = option;
1535 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001536 case Opt_journal_checksum:
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001537 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1538 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001539 case Opt_journal_async_commit:
1540 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001541 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001542 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001543 case Opt_noload:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001544 set_opt(sbi->s_mount_opt, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001545 break;
1546 case Opt_commit:
1547 if (match_int(&args[0], &option))
1548 return 0;
1549 if (option < 0)
1550 return 0;
1551 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001552 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001553 sbi->s_commit_interval = HZ * option;
1554 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001555 case Opt_max_batch_time:
1556 if (match_int(&args[0], &option))
1557 return 0;
1558 if (option < 0)
1559 return 0;
1560 if (option == 0)
1561 option = EXT4_DEF_MAX_BATCH_TIME;
1562 sbi->s_max_batch_time = option;
1563 break;
1564 case Opt_min_batch_time:
1565 if (match_int(&args[0], &option))
1566 return 0;
1567 if (option < 0)
1568 return 0;
1569 sbi->s_min_batch_time = option;
1570 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001571 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001572 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001573 goto datacheck;
1574 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001575 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001576 goto datacheck;
1577 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001578 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001579 datacheck:
1580 if (is_remount) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001581 if (test_opt(sb, DATA_FLAGS) != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001582 ext4_msg(sb, KERN_ERR,
1583 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001584 return 0;
1585 }
1586 } else {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001587 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001588 sbi->s_mount_opt |= data_opt;
1589 }
1590 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001591 case Opt_data_err_abort:
1592 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1593 break;
1594 case Opt_data_err_ignore:
1595 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1596 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001597#ifdef CONFIG_QUOTA
1598 case Opt_usrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001599 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1600 return 0;
1601 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001602 case Opt_grpjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001603 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001604 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001605 break;
1606 case Opt_offusrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001607 if (!clear_qf_name(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001608 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001609 break;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001610 case Opt_offgrpjquota:
1611 if (!clear_qf_name(sb, GRPQUOTA))
1612 return 0;
1613 break;
1614
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001615 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001616 qfmt = QFMT_VFS_OLD;
1617 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001618 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001619 qfmt = QFMT_VFS_V0;
Jan Kara5a20bdf2009-11-30 23:58:32 +01001620 goto set_qf_format;
1621 case Opt_jqfmt_vfsv1:
1622 qfmt = QFMT_VFS_V1;
Jan Karadfc5d032008-05-13 19:11:51 -04001623set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001624 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001625 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001626 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001627 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001628 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001629 return 0;
1630 }
1631 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001632 break;
1633 case Opt_quota:
1634 case Opt_usrquota:
1635 set_opt(sbi->s_mount_opt, QUOTA);
1636 set_opt(sbi->s_mount_opt, USRQUOTA);
1637 break;
1638 case Opt_grpquota:
1639 set_opt(sbi->s_mount_opt, QUOTA);
1640 set_opt(sbi->s_mount_opt, GRPQUOTA);
1641 break;
1642 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001643 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001644 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1645 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001646 return 0;
1647 }
1648 clear_opt(sbi->s_mount_opt, QUOTA);
1649 clear_opt(sbi->s_mount_opt, USRQUOTA);
1650 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1651 break;
1652#else
1653 case Opt_quota:
1654 case Opt_usrquota:
1655 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001656 ext4_msg(sb, KERN_ERR,
1657 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001658 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001659 case Opt_usrjquota:
1660 case Opt_grpjquota:
1661 case Opt_offusrjquota:
1662 case Opt_offgrpjquota:
1663 case Opt_jqfmt_vfsold:
1664 case Opt_jqfmt_vfsv0:
Jan Kara5a20bdf2009-11-30 23:58:32 +01001665 case Opt_jqfmt_vfsv1:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001666 ext4_msg(sb, KERN_ERR,
1667 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001668 break;
1669 case Opt_noquota:
1670 break;
1671#endif
1672 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001673 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001674 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001675 case Opt_nobarrier:
1676 clear_opt(sbi->s_mount_opt, BARRIER);
1677 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001678 case Opt_barrier:
Eric Sandeen15121c12010-02-15 20:17:55 -05001679 if (args[0].from) {
1680 if (match_int(&args[0], &option))
1681 return 0;
1682 } else
1683 option = 1; /* No argument, default to 1 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001684 if (option)
1685 set_opt(sbi->s_mount_opt, BARRIER);
1686 else
1687 clear_opt(sbi->s_mount_opt, BARRIER);
1688 break;
1689 case Opt_ignore:
1690 break;
1691 case Opt_resize:
1692 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001693 ext4_msg(sb, KERN_ERR,
1694 "resize option only available "
1695 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001696 return 0;
1697 }
1698 if (match_int(&args[0], &option) != 0)
1699 return 0;
1700 *n_blocks_count = option;
1701 break;
1702 case Opt_nobh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001703 ext4_msg(sb, KERN_WARNING,
1704 "Ignoring deprecated nobh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001705 break;
1706 case Opt_bh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001707 ext4_msg(sb, KERN_WARNING,
1708 "Ignoring deprecated bh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001709 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001710 case Opt_i_version:
1711 set_opt(sbi->s_mount_opt, I_VERSION);
1712 sb->s_flags |= MS_I_VERSION;
1713 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001714 case Opt_nodelalloc:
1715 clear_opt(sbi->s_mount_opt, DELALLOC);
1716 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001717 case Opt_stripe:
1718 if (match_int(&args[0], &option))
1719 return 0;
1720 if (option < 0)
1721 return 0;
1722 sbi->s_stripe = option;
1723 break;
Alex Tomas64769242008-07-11 19:27:31 -04001724 case Opt_delalloc:
1725 set_opt(sbi->s_mount_opt, DELALLOC);
1726 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001727 case Opt_block_validity:
1728 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1729 break;
1730 case Opt_noblock_validity:
1731 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1732 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001733 case Opt_inode_readahead_blks:
1734 if (match_int(&args[0], &option))
1735 return 0;
1736 if (option < 0 || option > (1 << 30))
1737 return 0;
Theodore Ts'of7c43952009-04-24 23:31:59 -04001738 if (!is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001739 ext4_msg(sb, KERN_ERR,
1740 "EXT4-fs: inode_readahead_blks"
1741 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001742 return 0;
1743 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001744 sbi->s_inode_readahead_blks = option;
1745 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001746 case Opt_journal_ioprio:
1747 if (match_int(&args[0], &option))
1748 return 0;
1749 if (option < 0 || option > 7)
1750 break;
1751 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1752 option);
1753 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001754 case Opt_noauto_da_alloc:
1755 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1756 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001757 case Opt_auto_da_alloc:
Eric Sandeen15121c12010-02-15 20:17:55 -05001758 if (args[0].from) {
1759 if (match_int(&args[0], &option))
1760 return 0;
1761 } else
1762 option = 1; /* No argument, default to 1 */
Theodore Ts'oafd46722009-03-16 23:12:23 -04001763 if (option)
1764 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1765 else
1766 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1767 break;
Eric Sandeen5328e632009-11-19 14:25:42 -05001768 case Opt_discard:
1769 set_opt(sbi->s_mount_opt, DISCARD);
1770 break;
1771 case Opt_nodiscard:
1772 clear_opt(sbi->s_mount_opt, DISCARD);
1773 break;
Jiaying Zhang744692d2010-03-04 16:14:02 -05001774 case Opt_dioread_nolock:
1775 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1776 break;
1777 case Opt_dioread_lock:
1778 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1779 break;
Lukas Czernerbfff6872010-10-27 21:30:05 -04001780 case Opt_init_inode_table:
1781 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1782 if (args[0].from) {
1783 if (match_int(&args[0], &option))
1784 return 0;
1785 } else
1786 option = EXT4_DEF_LI_WAIT_MULT;
1787 if (option < 0)
1788 return 0;
1789 sbi->s_li_wait_mult = option;
1790 break;
1791 case Opt_noinit_inode_table:
1792 clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1793 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001794 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001795 ext4_msg(sb, KERN_ERR,
1796 "Unrecognized mount option \"%s\" "
1797 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001798 return 0;
1799 }
1800 }
1801#ifdef CONFIG_QUOTA
1802 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001803 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001804 clear_opt(sbi->s_mount_opt, USRQUOTA);
1805
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001806 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001807 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1808
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001809 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001810 ext4_msg(sb, KERN_ERR, "old and new quota "
1811 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001812 return 0;
1813 }
1814
1815 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001816 ext4_msg(sb, KERN_ERR, "journaled quota format "
1817 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001818 return 0;
1819 }
1820 } else {
1821 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001822 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001823 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001824 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001825 return 0;
1826 }
1827 }
1828#endif
1829 return 1;
1830}
1831
Mingming Cao617ba132006-10-11 01:20:53 -07001832static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001833 int read_only)
1834{
Mingming Cao617ba132006-10-11 01:20:53 -07001835 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001836 int res = 0;
1837
Mingming Cao617ba132006-10-11 01:20:53 -07001838 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001839 ext4_msg(sb, KERN_ERR, "revision level too high, "
1840 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001841 res = MS_RDONLY;
1842 }
1843 if (read_only)
1844 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001845 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001846 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1847 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001848 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001849 ext4_msg(sb, KERN_WARNING,
1850 "warning: mounting fs with errors, "
1851 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001852 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1853 le16_to_cpu(es->s_mnt_count) >=
1854 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001855 ext4_msg(sb, KERN_WARNING,
1856 "warning: maximal mount count reached, "
1857 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001858 else if (le32_to_cpu(es->s_checkinterval) &&
1859 (le32_to_cpu(es->s_lastcheck) +
1860 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001861 ext4_msg(sb, KERN_WARNING,
1862 "warning: checktime reached, "
1863 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001864 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001865 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001866 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001867 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001868 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001870 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001871 if (sbi->s_journal)
1872 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001873
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001874 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001875 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001876 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'o7f4520c2009-06-13 10:09:41 -04001877 "bpg=%lu, ipg=%lu, mo=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001878 sb->s_blocksize,
1879 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001880 EXT4_BLOCKS_PER_GROUP(sb),
1881 EXT4_INODES_PER_GROUP(sb),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001882 sbi->s_mount_opt);
1883
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001884 return res;
1885}
1886
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001887static int ext4_fill_flex_info(struct super_block *sb)
1888{
1889 struct ext4_sb_info *sbi = EXT4_SB(sb);
1890 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001891 ext4_group_t flex_group_count;
1892 ext4_group_t flex_group;
1893 int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001894 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001895 int i;
1896
Theodore Ts'o503358a2009-11-23 07:24:46 -05001897 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1898 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1899
1900 if (groups_per_flex < 2) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001901 sbi->s_log_groups_per_flex = 0;
1902 return 1;
1903 }
1904
Frederic Bohec62a11f2008-09-08 10:20:24 -04001905 /* We allocate both existing and potentially added groups */
1906 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05001907 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1908 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001909 size = flex_group_count * sizeof(struct flex_groups);
1910 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1911 if (sbi->s_flex_groups == NULL) {
1912 sbi->s_flex_groups = vmalloc(size);
1913 if (sbi->s_flex_groups)
1914 memset(sbi->s_flex_groups, 0, size);
1915 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001916 if (sbi->s_flex_groups == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001917 ext4_msg(sb, KERN_ERR, "not enough memory for "
1918 "%u flex groups", flex_group_count);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001919 goto failed;
1920 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001921
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001922 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001923 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001924
1925 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001926 atomic_add(ext4_free_inodes_count(sb, gdp),
1927 &sbi->s_flex_groups[flex_group].free_inodes);
1928 atomic_add(ext4_free_blks_count(sb, gdp),
1929 &sbi->s_flex_groups[flex_group].free_blocks);
1930 atomic_add(ext4_used_dirs_count(sb, gdp),
1931 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001932 }
1933
1934 return 1;
1935failed:
1936 return 0;
1937}
1938
Andreas Dilger717d50e2007-10-16 18:38:25 -04001939__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1940 struct ext4_group_desc *gdp)
1941{
1942 __u16 crc = 0;
1943
1944 if (sbi->s_es->s_feature_ro_compat &
1945 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1946 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1947 __le32 le_group = cpu_to_le32(block_group);
1948
1949 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1950 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1951 crc = crc16(crc, (__u8 *)gdp, offset);
1952 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1953 /* for checksum of struct ext4_group_desc do the rest...*/
1954 if ((sbi->s_es->s_feature_incompat &
1955 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1956 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1957 crc = crc16(crc, (__u8 *)gdp + offset,
1958 le16_to_cpu(sbi->s_es->s_desc_size) -
1959 offset);
1960 }
1961
1962 return cpu_to_le16(crc);
1963}
1964
1965int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1966 struct ext4_group_desc *gdp)
1967{
1968 if ((sbi->s_es->s_feature_ro_compat &
1969 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1970 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1971 return 0;
1972
1973 return 1;
1974}
1975
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001976/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04001977static int ext4_check_descriptors(struct super_block *sb,
1978 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001979{
Mingming Cao617ba132006-10-11 01:20:53 -07001980 struct ext4_sb_info *sbi = EXT4_SB(sb);
1981 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1982 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001983 ext4_fsblk_t block_bitmap;
1984 ext4_fsblk_t inode_bitmap;
1985 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04001986 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04001987 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001988
Jose R. Santosce421582007-10-16 18:38:25 -04001989 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1990 flexbg_flag = 1;
1991
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001992 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001993
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001994 for (i = 0; i < sbi->s_groups_count; i++) {
1995 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1996
Jose R. Santosce421582007-10-16 18:38:25 -04001997 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001998 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001999 else
2000 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002001 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002002
Lukas Czernerbfff6872010-10-27 21:30:05 -04002003 if ((grp == sbi->s_groups_count) &&
2004 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2005 grp = i;
2006
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002007 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002008 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002009 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002010 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002011 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002012 return 0;
2013 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002014 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002015 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002016 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002017 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002018 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002019 return 0;
2020 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002021 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002022 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002023 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002024 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002025 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002026 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002027 return 0;
2028 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002029 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002030 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002031 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2032 "Checksum for group %u failed (%u!=%u)",
2033 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2034 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002035 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002036 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002037 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002038 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002039 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002040 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002041 if (!flexbg_flag)
2042 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002043 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002044 if (NULL != first_not_zeroed)
2045 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002046
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002047 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002048 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002049 return 1;
2050}
2051
Mingming Cao617ba132006-10-11 01:20:53 -07002052/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002053 * the superblock) which were deleted from all directories, but held open by
2054 * a process at the time of a crash. We walk the list and try to delete these
2055 * inodes at recovery time (only with a read-write filesystem).
2056 *
2057 * In order to keep the orphan inode chain consistent during traversal (in
2058 * case of crash during recovery), we link each inode into the superblock
2059 * orphan list_head and handle it the same way as an inode deletion during
2060 * normal operation (which journals the operations for us).
2061 *
2062 * We only do an iget() and an iput() on each inode, which is very safe if we
2063 * accidentally point at an in-use or already deleted inode. The worst that
2064 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002065 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002066 * e2fsck was run on this filesystem, and it must have already done the orphan
2067 * inode cleanup for us, so we can safely abort without any further action.
2068 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002069static void ext4_orphan_cleanup(struct super_block *sb,
2070 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002071{
2072 unsigned int s_flags = sb->s_flags;
2073 int nr_orphans = 0, nr_truncates = 0;
2074#ifdef CONFIG_QUOTA
2075 int i;
2076#endif
2077 if (!es->s_last_orphan) {
2078 jbd_debug(4, "no orphan inodes to clean up\n");
2079 return;
2080 }
2081
Eric Sandeena8f48a92006-12-06 20:40:13 -08002082 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002083 ext4_msg(sb, KERN_ERR, "write access "
2084 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002085 return;
2086 }
2087
Mingming Cao617ba132006-10-11 01:20:53 -07002088 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002089 if (es->s_last_orphan)
2090 jbd_debug(1, "Errors on filesystem, "
2091 "clearing orphan list.\n");
2092 es->s_last_orphan = 0;
2093 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2094 return;
2095 }
2096
2097 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002098 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002099 sb->s_flags &= ~MS_RDONLY;
2100 }
2101#ifdef CONFIG_QUOTA
2102 /* Needed for iput() to work correctly and not trash data */
2103 sb->s_flags |= MS_ACTIVE;
2104 /* Turn on quotas so that they are updated correctly */
2105 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002106 if (EXT4_SB(sb)->s_qf_names[i]) {
2107 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002108 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002109 ext4_msg(sb, KERN_ERR,
2110 "Cannot turn on journaled "
2111 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002112 }
2113 }
2114#endif
2115
2116 while (es->s_last_orphan) {
2117 struct inode *inode;
2118
Josef Bacik97bd42b2008-04-29 22:04:56 -04002119 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2120 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002121 es->s_last_orphan = 0;
2122 break;
2123 }
2124
Mingming Cao617ba132006-10-11 01:20:53 -07002125 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002126 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002127 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002128 ext4_msg(sb, KERN_DEBUG,
2129 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002130 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002131 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002132 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002133 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002134 nr_truncates++;
2135 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002136 ext4_msg(sb, KERN_DEBUG,
2137 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002138 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002139 jbd_debug(2, "deleting unreferenced inode %lu\n",
2140 inode->i_ino);
2141 nr_orphans++;
2142 }
2143 iput(inode); /* The delete magic happens here! */
2144 }
2145
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002146#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002147
2148 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002149 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2150 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002151 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002152 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2153 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002154#ifdef CONFIG_QUOTA
2155 /* Turn quotas off */
2156 for (i = 0; i < MAXQUOTAS; i++) {
2157 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002158 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002159 }
2160#endif
2161 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2162}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002163
Eric Sandeencd2291a2008-01-28 23:58:27 -05002164/*
2165 * Maximal extent format file size.
2166 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2167 * extent format containers, within a sector_t, and within i_blocks
2168 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2169 * so that won't be a limiting factor.
2170 *
2171 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2172 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002173static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002174{
2175 loff_t res;
2176 loff_t upper_limit = MAX_LFS_FILESIZE;
2177
2178 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002179 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002180 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002181 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002182 * i_block represent total blocks in 512 bytes
2183 * 32 == size of vfs inode i_blocks * 8
2184 */
2185 upper_limit = (1LL << 32) - 1;
2186
2187 /* total blocks in file system block size */
2188 upper_limit >>= (blkbits - 9);
2189 upper_limit <<= blkbits;
2190 }
2191
2192 /* 32-bit extent-start container, ee_block */
2193 res = 1LL << 32;
2194 res <<= blkbits;
2195 res -= 1;
2196
2197 /* Sanity check against vm- & vfs- imposed limits */
2198 if (res > upper_limit)
2199 res = upper_limit;
2200
2201 return res;
2202}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002203
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002204/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002205 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002206 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2207 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002208 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002209static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002210{
Mingming Cao617ba132006-10-11 01:20:53 -07002211 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002212 int meta_blocks;
2213 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002214 /* This is calculated to be the largest file size for a dense, block
2215 * mapped file such that the file's total number of 512-byte sectors,
2216 * including data and all indirect blocks, does not exceed (2^48 - 1).
2217 *
2218 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2219 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002220 */
2221
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002222 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002223 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002224 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002225 * the inode i_block field represents total file blocks in
2226 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002227 */
2228 upper_limit = (1LL << 32) - 1;
2229
2230 /* total blocks in file system block size */
2231 upper_limit >>= (bits - 9);
2232
2233 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002234 /*
2235 * We use 48 bit ext4_inode i_blocks
2236 * With EXT4_HUGE_FILE_FL set the i_blocks
2237 * represent total number of blocks in
2238 * file system block size
2239 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002240 upper_limit = (1LL << 48) - 1;
2241
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002242 }
2243
2244 /* indirect blocks */
2245 meta_blocks = 1;
2246 /* double indirect blocks */
2247 meta_blocks += 1 + (1LL << (bits-2));
2248 /* tripple indirect blocks */
2249 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2250
2251 upper_limit -= meta_blocks;
2252 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002253
2254 res += 1LL << (bits-2);
2255 res += 1LL << (2*(bits-2));
2256 res += 1LL << (3*(bits-2));
2257 res <<= bits;
2258 if (res > upper_limit)
2259 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002260
2261 if (res > MAX_LFS_FILESIZE)
2262 res = MAX_LFS_FILESIZE;
2263
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002264 return res;
2265}
2266
Mingming Cao617ba132006-10-11 01:20:53 -07002267static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002268 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002269{
Mingming Cao617ba132006-10-11 01:20:53 -07002270 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002271 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002272 int has_super = 0;
2273
2274 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2275
Mingming Cao617ba132006-10-11 01:20:53 -07002276 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002277 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002278 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002279 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002280 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002281 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002282
Mingming Cao617ba132006-10-11 01:20:53 -07002283 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002284}
2285
Alex Tomasc9de5602008-01-29 00:19:52 -05002286/**
2287 * ext4_get_stripe_size: Get the stripe size.
2288 * @sbi: In memory super block info
2289 *
2290 * If we have specified it via mount option, then
2291 * use the mount option value. If the value specified at mount time is
2292 * greater than the blocks per group use the super block value.
2293 * If the super block value is greater than blocks per group return 0.
2294 * Allocator needs it be less than blocks per group.
2295 *
2296 */
2297static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2298{
2299 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2300 unsigned long stripe_width =
2301 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2302
2303 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2304 return sbi->s_stripe;
2305
2306 if (stripe_width <= sbi->s_blocks_per_group)
2307 return stripe_width;
2308
2309 if (stride <= sbi->s_blocks_per_group)
2310 return stride;
2311
2312 return 0;
2313}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002314
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002315/* sysfs supprt */
2316
2317struct ext4_attr {
2318 struct attribute attr;
2319 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
Theodore Ts'o60e66792010-05-17 07:00:00 -04002320 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002321 const char *, size_t);
2322 int offset;
2323};
2324
2325static int parse_strtoul(const char *buf,
2326 unsigned long max, unsigned long *value)
2327{
2328 char *endp;
2329
André Goddard Rosae7d28602009-12-14 18:01:06 -08002330 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2331 endp = skip_spaces(endp);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002332 if (*endp || *value > max)
2333 return -EINVAL;
2334
2335 return 0;
2336}
2337
2338static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2339 struct ext4_sb_info *sbi,
2340 char *buf)
2341{
2342 return snprintf(buf, PAGE_SIZE, "%llu\n",
2343 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2344}
2345
2346static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2347 struct ext4_sb_info *sbi, char *buf)
2348{
2349 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2350
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002351 if (!sb->s_bdev->bd_part)
2352 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002353 return snprintf(buf, PAGE_SIZE, "%lu\n",
2354 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2355 sbi->s_sectors_written_start) >> 1);
2356}
2357
2358static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2359 struct ext4_sb_info *sbi, char *buf)
2360{
2361 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2362
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002363 if (!sb->s_bdev->bd_part)
2364 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002365 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002366 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002367 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002368 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002369}
2370
2371static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2372 struct ext4_sb_info *sbi,
2373 const char *buf, size_t count)
2374{
2375 unsigned long t;
2376
2377 if (parse_strtoul(buf, 0x40000000, &t))
2378 return -EINVAL;
2379
Theodore Ts'of7c43952009-04-24 23:31:59 -04002380 if (!is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002381 return -EINVAL;
2382
2383 sbi->s_inode_readahead_blks = t;
2384 return count;
2385}
2386
2387static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002388 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002389{
2390 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2391
2392 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2393}
2394
2395static ssize_t sbi_ui_store(struct ext4_attr *a,
2396 struct ext4_sb_info *sbi,
2397 const char *buf, size_t count)
2398{
2399 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2400 unsigned long t;
2401
2402 if (parse_strtoul(buf, 0xffffffff, &t))
2403 return -EINVAL;
2404 *ui = t;
2405 return count;
2406}
2407
2408#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2409static struct ext4_attr ext4_attr_##_name = { \
2410 .attr = {.name = __stringify(_name), .mode = _mode }, \
2411 .show = _show, \
2412 .store = _store, \
2413 .offset = offsetof(struct ext4_sb_info, _elname), \
2414}
2415#define EXT4_ATTR(name, mode, show, store) \
2416static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2417
Lukas Czerner857ac882010-10-27 21:30:05 -04002418#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002419#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2420#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2421#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2422 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2423#define ATTR_LIST(name) &ext4_attr_##name.attr
2424
2425EXT4_RO_ATTR(delayed_allocation_blocks);
2426EXT4_RO_ATTR(session_write_kbytes);
2427EXT4_RO_ATTR(lifetime_write_kbytes);
2428EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2429 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002430EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002431EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2432EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2433EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2434EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2435EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2436EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002437EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002438
2439static struct attribute *ext4_attrs[] = {
2440 ATTR_LIST(delayed_allocation_blocks),
2441 ATTR_LIST(session_write_kbytes),
2442 ATTR_LIST(lifetime_write_kbytes),
2443 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002444 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002445 ATTR_LIST(mb_stats),
2446 ATTR_LIST(mb_max_to_scan),
2447 ATTR_LIST(mb_min_to_scan),
2448 ATTR_LIST(mb_order2_req),
2449 ATTR_LIST(mb_stream_req),
2450 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002451 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002452 NULL,
2453};
2454
Lukas Czerner857ac882010-10-27 21:30:05 -04002455/* Features this copy of ext4 supports */
2456EXT4_INFO_ATTR(lazy_itable_init);
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002457EXT4_INFO_ATTR(batched_discard);
Lukas Czerner857ac882010-10-27 21:30:05 -04002458
2459static struct attribute *ext4_feat_attrs[] = {
2460 ATTR_LIST(lazy_itable_init),
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002461 ATTR_LIST(batched_discard),
Lukas Czerner857ac882010-10-27 21:30:05 -04002462 NULL,
2463};
2464
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002465static ssize_t ext4_attr_show(struct kobject *kobj,
2466 struct attribute *attr, char *buf)
2467{
2468 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2469 s_kobj);
2470 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2471
2472 return a->show ? a->show(a, sbi, buf) : 0;
2473}
2474
2475static ssize_t ext4_attr_store(struct kobject *kobj,
2476 struct attribute *attr,
2477 const char *buf, size_t len)
2478{
2479 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2480 s_kobj);
2481 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2482
2483 return a->store ? a->store(a, sbi, buf, len) : 0;
2484}
2485
2486static void ext4_sb_release(struct kobject *kobj)
2487{
2488 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2489 s_kobj);
2490 complete(&sbi->s_kobj_unregister);
2491}
2492
Emese Revfy52cf25d2010-01-19 02:58:23 +01002493static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002494 .show = ext4_attr_show,
2495 .store = ext4_attr_store,
2496};
2497
2498static struct kobj_type ext4_ktype = {
2499 .default_attrs = ext4_attrs,
2500 .sysfs_ops = &ext4_attr_ops,
2501 .release = ext4_sb_release,
2502};
2503
Lukas Czerner857ac882010-10-27 21:30:05 -04002504static void ext4_feat_release(struct kobject *kobj)
2505{
2506 complete(&ext4_feat->f_kobj_unregister);
2507}
2508
2509static struct kobj_type ext4_feat_ktype = {
2510 .default_attrs = ext4_feat_attrs,
2511 .sysfs_ops = &ext4_attr_ops,
2512 .release = ext4_feat_release,
2513};
2514
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002515/*
2516 * Check whether this filesystem can be mounted based on
2517 * the features present and the RDONLY/RDWR mount requested.
2518 * Returns 1 if this filesystem can be mounted as requested,
2519 * 0 if it cannot be.
2520 */
2521static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2522{
2523 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2524 ext4_msg(sb, KERN_ERR,
2525 "Couldn't mount because of "
2526 "unsupported optional features (%x)",
2527 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2528 ~EXT4_FEATURE_INCOMPAT_SUPP));
2529 return 0;
2530 }
2531
2532 if (readonly)
2533 return 1;
2534
2535 /* Check that feature set is OK for a read-write mount */
2536 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2537 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2538 "unsupported optional features (%x)",
2539 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2540 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2541 return 0;
2542 }
2543 /*
2544 * Large file size enabled file system can only be mounted
2545 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2546 */
2547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2548 if (sizeof(blkcnt_t) < sizeof(u64)) {
2549 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2550 "cannot be mounted RDWR without "
2551 "CONFIG_LBDAF");
2552 return 0;
2553 }
2554 }
2555 return 1;
2556}
2557
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002558/*
2559 * This function is called once a day if we have errors logged
2560 * on the file system
2561 */
2562static void print_daily_error_info(unsigned long arg)
2563{
2564 struct super_block *sb = (struct super_block *) arg;
2565 struct ext4_sb_info *sbi;
2566 struct ext4_super_block *es;
2567
2568 sbi = EXT4_SB(sb);
2569 es = sbi->s_es;
2570
2571 if (es->s_error_count)
2572 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2573 le32_to_cpu(es->s_error_count));
2574 if (es->s_first_error_time) {
2575 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2576 sb->s_id, le32_to_cpu(es->s_first_error_time),
2577 (int) sizeof(es->s_first_error_func),
2578 es->s_first_error_func,
2579 le32_to_cpu(es->s_first_error_line));
2580 if (es->s_first_error_ino)
2581 printk(": inode %u",
2582 le32_to_cpu(es->s_first_error_ino));
2583 if (es->s_first_error_block)
2584 printk(": block %llu", (unsigned long long)
2585 le64_to_cpu(es->s_first_error_block));
2586 printk("\n");
2587 }
2588 if (es->s_last_error_time) {
2589 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2590 sb->s_id, le32_to_cpu(es->s_last_error_time),
2591 (int) sizeof(es->s_last_error_func),
2592 es->s_last_error_func,
2593 le32_to_cpu(es->s_last_error_line));
2594 if (es->s_last_error_ino)
2595 printk(": inode %u",
2596 le32_to_cpu(es->s_last_error_ino));
2597 if (es->s_last_error_block)
2598 printk(": block %llu", (unsigned long long)
2599 le64_to_cpu(es->s_last_error_block));
2600 printk("\n");
2601 }
2602 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2603}
2604
Lukas Czernerbfff6872010-10-27 21:30:05 -04002605static void ext4_lazyinode_timeout(unsigned long data)
2606{
2607 struct task_struct *p = (struct task_struct *)data;
2608 wake_up_process(p);
2609}
2610
2611/* Find next suitable group and run ext4_init_inode_table */
2612static int ext4_run_li_request(struct ext4_li_request *elr)
2613{
2614 struct ext4_group_desc *gdp = NULL;
2615 ext4_group_t group, ngroups;
2616 struct super_block *sb;
2617 unsigned long timeout = 0;
2618 int ret = 0;
2619
2620 sb = elr->lr_super;
2621 ngroups = EXT4_SB(sb)->s_groups_count;
2622
2623 for (group = elr->lr_next_group; group < ngroups; group++) {
2624 gdp = ext4_get_group_desc(sb, group, NULL);
2625 if (!gdp) {
2626 ret = 1;
2627 break;
2628 }
2629
2630 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2631 break;
2632 }
2633
2634 if (group == ngroups)
2635 ret = 1;
2636
2637 if (!ret) {
2638 timeout = jiffies;
2639 ret = ext4_init_inode_table(sb, group,
2640 elr->lr_timeout ? 0 : 1);
2641 if (elr->lr_timeout == 0) {
2642 timeout = jiffies - timeout;
2643 if (elr->lr_sbi->s_li_wait_mult)
2644 timeout *= elr->lr_sbi->s_li_wait_mult;
2645 else
2646 timeout *= 20;
2647 elr->lr_timeout = timeout;
2648 }
2649 elr->lr_next_sched = jiffies + elr->lr_timeout;
2650 elr->lr_next_group = group + 1;
2651 }
2652
2653 return ret;
2654}
2655
2656/*
2657 * Remove lr_request from the list_request and free the
2658 * request tructure. Should be called with li_list_mtx held
2659 */
2660static void ext4_remove_li_request(struct ext4_li_request *elr)
2661{
2662 struct ext4_sb_info *sbi;
2663
2664 if (!elr)
2665 return;
2666
2667 sbi = elr->lr_sbi;
2668
2669 list_del(&elr->lr_request);
2670 sbi->s_li_request = NULL;
2671 kfree(elr);
2672}
2673
2674static void ext4_unregister_li_request(struct super_block *sb)
2675{
2676 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2677
2678 if (!ext4_li_info)
2679 return;
2680
2681 mutex_lock(&ext4_li_info->li_list_mtx);
2682 ext4_remove_li_request(elr);
2683 mutex_unlock(&ext4_li_info->li_list_mtx);
2684}
2685
2686/*
2687 * This is the function where ext4lazyinit thread lives. It walks
2688 * through the request list searching for next scheduled filesystem.
2689 * When such a fs is found, run the lazy initialization request
2690 * (ext4_rn_li_request) and keep track of the time spend in this
2691 * function. Based on that time we compute next schedule time of
2692 * the request. When walking through the list is complete, compute
2693 * next waking time and put itself into sleep.
2694 */
2695static int ext4_lazyinit_thread(void *arg)
2696{
2697 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2698 struct list_head *pos, *n;
2699 struct ext4_li_request *elr;
2700 unsigned long next_wakeup;
2701 DEFINE_WAIT(wait);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002702
2703 BUG_ON(NULL == eli);
2704
2705 eli->li_timer.data = (unsigned long)current;
2706 eli->li_timer.function = ext4_lazyinode_timeout;
2707
2708 eli->li_task = current;
2709 wake_up(&eli->li_wait_task);
2710
2711cont_thread:
2712 while (true) {
2713 next_wakeup = MAX_JIFFY_OFFSET;
2714
2715 mutex_lock(&eli->li_list_mtx);
2716 if (list_empty(&eli->li_request_list)) {
2717 mutex_unlock(&eli->li_list_mtx);
2718 goto exit_thread;
2719 }
2720
2721 list_for_each_safe(pos, n, &eli->li_request_list) {
2722 elr = list_entry(pos, struct ext4_li_request,
2723 lr_request);
2724
Theodore Ts'ob2c78cd2010-11-02 14:19:30 -04002725 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2726 if (ext4_run_li_request(elr) != 0) {
2727 /* error, remove the lazy_init job */
2728 ext4_remove_li_request(elr);
2729 continue;
2730 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002731 }
2732
2733 if (time_before(elr->lr_next_sched, next_wakeup))
2734 next_wakeup = elr->lr_next_sched;
2735 }
2736 mutex_unlock(&eli->li_list_mtx);
2737
2738 if (freezing(current))
2739 refrigerator();
2740
Lukas Czernerf4245bd2010-11-02 14:07:17 -04002741 if ((time_after_eq(jiffies, next_wakeup)) ||
2742 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04002743 cond_resched();
2744 continue;
2745 }
2746
2747 eli->li_timer.expires = next_wakeup;
2748 add_timer(&eli->li_timer);
2749 prepare_to_wait(&eli->li_wait_daemon, &wait,
2750 TASK_INTERRUPTIBLE);
2751 if (time_before(jiffies, next_wakeup))
2752 schedule();
2753 finish_wait(&eli->li_wait_daemon, &wait);
2754 }
2755
2756exit_thread:
2757 /*
2758 * It looks like the request list is empty, but we need
2759 * to check it under the li_list_mtx lock, to prevent any
2760 * additions into it, and of course we should lock ext4_li_mtx
2761 * to atomically free the list and ext4_li_info, because at
2762 * this point another ext4 filesystem could be registering
2763 * new one.
2764 */
2765 mutex_lock(&ext4_li_mtx);
2766 mutex_lock(&eli->li_list_mtx);
2767 if (!list_empty(&eli->li_request_list)) {
2768 mutex_unlock(&eli->li_list_mtx);
2769 mutex_unlock(&ext4_li_mtx);
2770 goto cont_thread;
2771 }
2772 mutex_unlock(&eli->li_list_mtx);
2773 del_timer_sync(&ext4_li_info->li_timer);
2774 eli->li_task = NULL;
2775 wake_up(&eli->li_wait_task);
2776
2777 kfree(ext4_li_info);
2778 ext4_li_info = NULL;
2779 mutex_unlock(&ext4_li_mtx);
2780
2781 return 0;
2782}
2783
2784static void ext4_clear_request_list(void)
2785{
2786 struct list_head *pos, *n;
2787 struct ext4_li_request *elr;
2788
2789 mutex_lock(&ext4_li_info->li_list_mtx);
2790 if (list_empty(&ext4_li_info->li_request_list))
2791 return;
2792
2793 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2794 elr = list_entry(pos, struct ext4_li_request,
2795 lr_request);
2796 ext4_remove_li_request(elr);
2797 }
2798 mutex_unlock(&ext4_li_info->li_list_mtx);
2799}
2800
2801static int ext4_run_lazyinit_thread(void)
2802{
2803 struct task_struct *t;
2804
2805 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2806 if (IS_ERR(t)) {
2807 int err = PTR_ERR(t);
2808 ext4_clear_request_list();
2809 del_timer_sync(&ext4_li_info->li_timer);
2810 kfree(ext4_li_info);
2811 ext4_li_info = NULL;
2812 printk(KERN_CRIT "EXT4: error %d creating inode table "
2813 "initialization thread\n",
2814 err);
2815 return err;
2816 }
2817 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2818
2819 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2820 return 0;
2821}
2822
2823/*
2824 * Check whether it make sense to run itable init. thread or not.
2825 * If there is at least one uninitialized inode table, return
2826 * corresponding group number, else the loop goes through all
2827 * groups and return total number of groups.
2828 */
2829static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2830{
2831 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2832 struct ext4_group_desc *gdp = NULL;
2833
2834 for (group = 0; group < ngroups; group++) {
2835 gdp = ext4_get_group_desc(sb, group, NULL);
2836 if (!gdp)
2837 continue;
2838
2839 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2840 break;
2841 }
2842
2843 return group;
2844}
2845
2846static int ext4_li_info_new(void)
2847{
2848 struct ext4_lazy_init *eli = NULL;
2849
2850 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2851 if (!eli)
2852 return -ENOMEM;
2853
2854 eli->li_task = NULL;
2855 INIT_LIST_HEAD(&eli->li_request_list);
2856 mutex_init(&eli->li_list_mtx);
2857
2858 init_waitqueue_head(&eli->li_wait_daemon);
2859 init_waitqueue_head(&eli->li_wait_task);
2860 init_timer(&eli->li_timer);
2861 eli->li_state |= EXT4_LAZYINIT_QUIT;
2862
2863 ext4_li_info = eli;
2864
2865 return 0;
2866}
2867
2868static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2869 ext4_group_t start)
2870{
2871 struct ext4_sb_info *sbi = EXT4_SB(sb);
2872 struct ext4_li_request *elr;
2873 unsigned long rnd;
2874
2875 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2876 if (!elr)
2877 return NULL;
2878
2879 elr->lr_super = sb;
2880 elr->lr_sbi = sbi;
2881 elr->lr_next_group = start;
2882
2883 /*
2884 * Randomize first schedule time of the request to
2885 * spread the inode table initialization requests
2886 * better.
2887 */
2888 get_random_bytes(&rnd, sizeof(rnd));
2889 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2890 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2891
2892 return elr;
2893}
2894
2895static int ext4_register_li_request(struct super_block *sb,
2896 ext4_group_t first_not_zeroed)
2897{
2898 struct ext4_sb_info *sbi = EXT4_SB(sb);
2899 struct ext4_li_request *elr;
2900 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002901 int ret;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002902
2903 if (sbi->s_li_request != NULL)
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002904 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002905
2906 if (first_not_zeroed == ngroups ||
2907 (sb->s_flags & MS_RDONLY) ||
2908 !test_opt(sb, INIT_INODE_TABLE)) {
2909 sbi->s_li_request = NULL;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002910 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002911 }
2912
2913 if (first_not_zeroed == ngroups) {
2914 sbi->s_li_request = NULL;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002915 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002916 }
2917
2918 elr = ext4_li_request_new(sb, first_not_zeroed);
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002919 if (!elr)
2920 return -ENOMEM;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002921
2922 mutex_lock(&ext4_li_mtx);
2923
2924 if (NULL == ext4_li_info) {
2925 ret = ext4_li_info_new();
2926 if (ret)
2927 goto out;
2928 }
2929
2930 mutex_lock(&ext4_li_info->li_list_mtx);
2931 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2932 mutex_unlock(&ext4_li_info->li_list_mtx);
2933
2934 sbi->s_li_request = elr;
2935
2936 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2937 ret = ext4_run_lazyinit_thread();
2938 if (ret)
2939 goto out;
2940 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002941out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002942 mutex_unlock(&ext4_li_mtx);
2943 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002944 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002945 return ret;
2946}
2947
2948/*
2949 * We do not need to lock anything since this is called on
2950 * module unload.
2951 */
2952static void ext4_destroy_lazyinit_thread(void)
2953{
2954 /*
2955 * If thread exited earlier
2956 * there's nothing to be done.
2957 */
2958 if (!ext4_li_info)
2959 return;
2960
2961 ext4_clear_request_list();
2962
2963 while (ext4_li_info->li_task) {
2964 wake_up(&ext4_li_info->li_wait_daemon);
2965 wait_event(ext4_li_info->li_wait_task,
2966 ext4_li_info->li_task == NULL);
2967 }
2968}
2969
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002970static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04002971 __releases(kernel_lock)
2972 __acquires(kernel_lock)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002973{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04002974 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002975 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07002976 struct ext4_super_block *es = NULL;
2977 struct ext4_sb_info *sbi;
2978 ext4_fsblk_t block;
2979 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002980 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002981 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002982 unsigned long journal_devnum = 0;
2983 unsigned long def_mount_opts;
2984 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002985 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05002986 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04002987 int ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002988 int blocksize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002989 unsigned int db_count;
2990 unsigned int i;
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002991 int needs_recovery, has_huge_files;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002992 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07002993 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002994 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002995 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002996
2997 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2998 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04002999 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003000
3001 sbi->s_blockgroup_lock =
3002 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3003 if (!sbi->s_blockgroup_lock) {
3004 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003005 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003006 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003007 sb->s_fs_info = sbi;
3008 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07003009 sbi->s_resuid = EXT4_DEF_RESUID;
3010 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003011 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003012 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003013 if (sb->s_bdev->bd_part)
3014 sbi->s_sectors_written_start =
3015 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003016
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003017 /* Cleanup superblock name */
3018 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3019 *cp = '!';
3020
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003021 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003022 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003023 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003024 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003025 goto out_fail;
3026 }
3027
3028 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003029 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003030 * block sizes. We need to calculate the offset from buffer start.
3031 */
Mingming Cao617ba132006-10-11 01:20:53 -07003032 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003033 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3034 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003035 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003036 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003037 }
3038
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003039 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003040 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003041 goto out_fail;
3042 }
3043 /*
3044 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003045 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003046 */
Mingming Cao617ba132006-10-11 01:20:53 -07003047 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003048 sbi->s_es = es;
3049 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003050 if (sb->s_magic != EXT4_SUPER_MAGIC)
3051 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003052 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003053
3054 /* Set defaults before we parse the mount options */
3055 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003056 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003057 if (def_mount_opts & EXT4_DEFM_DEBUG)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003058 set_opt(sbi->s_mount_opt, DEBUG);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003059 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3060 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3061 "2.6.38");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003062 set_opt(sbi->s_mount_opt, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003063 }
Mingming Cao617ba132006-10-11 01:20:53 -07003064 if (def_mount_opts & EXT4_DEFM_UID16)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003065 set_opt(sbi->s_mount_opt, NO_UID32);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003066#ifdef CONFIG_EXT4_FS_XATTR
Mingming Cao617ba132006-10-11 01:20:53 -07003067 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003068 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003069#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04003070#ifdef CONFIG_EXT4_FS_POSIX_ACL
Mingming Cao617ba132006-10-11 01:20:53 -07003071 if (def_mount_opts & EXT4_DEFM_ACL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003072 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003073#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003074 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003075 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003076 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003077 set_opt(sbi->s_mount_opt, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003078 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003079 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003080
Mingming Cao617ba132006-10-11 01:20:53 -07003081 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003082 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003083 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Dmitry Mishinceea16b2006-10-11 01:21:21 -07003084 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003085 else
3086 set_opt(sbi->s_mount_opt, ERRORS_RO);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003087 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3088 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
3089 if (def_mount_opts & EXT4_DEFM_DISCARD)
3090 set_opt(sbi->s_mount_opt, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003091
3092 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3093 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05003094 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3095 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3096 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003097
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003098 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3099 set_opt(sbi->s_mount_opt, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003100
Mingming Cao1e2462f2007-07-18 09:00:55 -04003101 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003102 * enable delayed allocation by default
3103 * Use -o nodelalloc to turn it off
3104 */
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003105 if (!IS_EXT3_SB(sb) &&
3106 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Jan Karaba69f9a2010-03-24 20:18:37 -04003107 set_opt(sbi->s_mount_opt, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003108
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003109 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3110 &journal_devnum, &journal_ioprio, NULL, 0)) {
3111 ext4_msg(sb, KERN_WARNING,
3112 "failed to parse options in superblock: %s",
3113 sbi->s_es->s_mount_opts);
3114 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003115 if (!parse_options((char *) data, sb, &journal_devnum,
3116 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003117 goto failed_mount;
3118
3119 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003120 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003121
Mingming Cao617ba132006-10-11 01:20:53 -07003122 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3123 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3124 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3125 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003126 ext4_msg(sb, KERN_WARNING,
3127 "feature flags set on rev 0 fs, "
3128 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003129
3130 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003131 * Check feature flags regardless of the revision level, since we
3132 * previously didn't change the revision level when setting the flags,
3133 * so there is a chance incompat flags are set on a rev 0 filesystem.
3134 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003135 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003136 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003137
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003138 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3139
Mingming Cao617ba132006-10-11 01:20:53 -07003140 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3141 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003142 ext4_msg(sb, KERN_ERR,
3143 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003144 goto failed_mount;
3145 }
3146
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003147 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003148 /* Validate the filesystem blocksize */
3149 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003150 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003151 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003152 goto failed_mount;
3153 }
3154
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003155 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003156 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3157 offset = do_div(logical_sb_block, blocksize);
3158 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003159 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003160 ext4_msg(sb, KERN_ERR,
3161 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003162 goto failed_mount;
3163 }
Mingming Cao617ba132006-10-11 01:20:53 -07003164 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003165 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003166 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003167 ext4_msg(sb, KERN_ERR,
3168 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003169 goto failed_mount;
3170 }
3171 }
3172
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003173 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3174 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003175 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3176 has_huge_files);
3177 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003178
Mingming Cao617ba132006-10-11 01:20:53 -07003179 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3180 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3181 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003182 } else {
3183 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3184 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003185 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003186 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003187 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003188 ext4_msg(sb, KERN_ERR,
3189 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003190 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003191 goto failed_mount;
3192 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003193 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3194 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003195 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003196
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003197 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3198 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003199 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003200 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003201 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003202 ext4_msg(sb, KERN_ERR,
3203 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003204 sbi->s_desc_size);
3205 goto failed_mount;
3206 }
3207 } else
3208 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003209
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003210 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003211 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003212 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003213 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003214
Mingming Cao617ba132006-10-11 01:20:53 -07003215 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003216 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003217 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003218 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3219 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003220 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003221 sbi->s_sbh = bh;
3222 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003223 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3224 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003225
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003226 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003227 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3228 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003229 i = le32_to_cpu(es->s_flags);
3230 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3231 sbi->s_hash_unsigned = 3;
3232 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3233#ifdef __CHAR_UNSIGNED__
3234 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3235 sbi->s_hash_unsigned = 3;
3236#else
3237 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3238#endif
3239 sb->s_dirt = 1;
3240 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003241
3242 if (sbi->s_blocks_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003243 ext4_msg(sb, KERN_ERR,
3244 "#blocks per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003245 sbi->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003246 goto failed_mount;
3247 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003248 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003249 ext4_msg(sb, KERN_ERR,
3250 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003251 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003252 goto failed_mount;
3253 }
3254
Eric Sandeenbf43d842009-08-17 23:48:51 -04003255 /*
3256 * Test whether we have more sectors than will fit in sector_t,
3257 * and whether the max offset is addressable by the page cache.
3258 */
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003259 ret = generic_check_addressable(sb->s_blocksize_bits,
3260 ext4_blocks_count(es));
3261 if (ret) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003262 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003263 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003264 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003265 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003266 goto failed_mount;
3267 }
3268
Mingming Cao617ba132006-10-11 01:20:53 -07003269 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3270 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003271
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003272 /* check blocks count against device size */
3273 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3274 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003275 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3276 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003277 ext4_blocks_count(es), blocks_count);
3278 goto failed_mount;
3279 }
3280
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003281 /*
3282 * It makes no sense for the first data block to be beyond the end
3283 * of the filesystem.
3284 */
3285 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003286 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3287 "block %u is beyond end of filesystem (%llu)",
3288 le32_to_cpu(es->s_first_data_block),
3289 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003290 goto failed_mount;
3291 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003292 blocks_count = (ext4_blocks_count(es) -
3293 le32_to_cpu(es->s_first_data_block) +
3294 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3295 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003296 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003297 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003298 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003299 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003300 ext4_blocks_count(es),
3301 le32_to_cpu(es->s_first_data_block),
3302 EXT4_BLOCKS_PER_GROUP(sb));
3303 goto failed_mount;
3304 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003305 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003306 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3307 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003308 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3309 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003310 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003311 GFP_KERNEL);
3312 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003313 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003314 goto failed_mount;
3315 }
3316
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04003317#ifdef CONFIG_PROC_FS
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003318 if (ext4_proc_root)
3319 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04003320#endif
Theodore Ts'o240799c2008-10-09 23:53:47 -04003321
Pekka Enberg705895b2009-02-15 18:07:52 -05003322 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003323
3324 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003325 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003326 sbi->s_group_desc[i] = sb_bread(sb, block);
3327 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003328 ext4_msg(sb, KERN_ERR,
3329 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003330 db_count = i;
3331 goto failed_mount2;
3332 }
3333 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003334 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003335 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003336 goto failed_mount2;
3337 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003338 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3339 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003340 ext4_msg(sb, KERN_ERR,
3341 "unable to initialize "
3342 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003343 goto failed_mount2;
3344 }
3345
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003346 sbi->s_gdb_count = db_count;
3347 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3348 spin_lock_init(&sbi->s_next_gen_lock);
3349
Alex Tomasc9de5602008-01-29 00:19:52 -05003350 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04003351 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05003352
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003353 /*
3354 * set up enough so that it can read an inode
3355 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003356 if (!test_opt(sb, NOLOAD) &&
3357 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3358 sb->s_op = &ext4_sops;
3359 else
3360 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07003361 sb->s_export_op = &ext4_export_ops;
3362 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003363#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003364 sb->s_qcop = &ext4_qctl_operations;
3365 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003366#endif
3367 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003368 mutex_init(&sbi->s_orphan_lock);
Theodore Ts'o32ed5052009-04-25 22:53:39 -04003369 mutex_init(&sbi->s_resize_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003370
3371 sb->s_root = NULL;
3372
3373 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07003374 EXT4_HAS_INCOMPAT_FEATURE(sb,
3375 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003376
3377 /*
3378 * The first inode we look at is the journal inode. Don't try
3379 * root first: it may be modified in the journal!
3380 */
3381 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07003382 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3383 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003384 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05003385 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3386 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003387 ext4_msg(sb, KERN_ERR, "required journal recovery "
3388 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003389 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003390 } else {
Frank Mayhar03901312009-01-07 00:06:22 -05003391 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
3392 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3393 sbi->s_journal = NULL;
3394 needs_recovery = 0;
3395 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003396 }
3397
Jose R. Santoseb40a092007-07-18 08:37:25 -04003398 if (ext4_blocks_count(es) > 0xffffffffULL &&
3399 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3400 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003401 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003402 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04003403 }
3404
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003405 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3406 jbd2_journal_set_features(sbi->s_journal,
3407 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05003408 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003409 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3410 jbd2_journal_set_features(sbi->s_journal,
3411 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
Girish Shilamkar818d2762008-01-28 23:58:27 -05003412 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3413 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003414 } else {
3415 jbd2_journal_clear_features(sbi->s_journal,
3416 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3417 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3418 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05003419
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003420 /* We have now updated the journal if required, so we can
3421 * validate the data journaling mode. */
3422 switch (test_opt(sb, DATA_FLAGS)) {
3423 case 0:
3424 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07003425 * capabilities: ORDERED_DATA if the journal can
3426 * cope, else JOURNAL_DATA
3427 */
Mingming Caodab291a2006-10-11 01:21:01 -07003428 if (jbd2_journal_check_available_features
3429 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003430 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3431 else
3432 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3433 break;
3434
Mingming Cao617ba132006-10-11 01:20:53 -07003435 case EXT4_MOUNT_ORDERED_DATA:
3436 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07003437 if (!jbd2_journal_check_available_features
3438 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003439 ext4_msg(sb, KERN_ERR, "Journal does not support "
3440 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003441 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003442 }
3443 default:
3444 break;
3445 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003446 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003447
Frank Mayhar03901312009-01-07 00:06:22 -05003448no_journal:
Dmitry Monakhov84061e02010-05-16 08:00:00 -04003449 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3450 ext4_count_free_blocks(sb));
3451 if (!err)
3452 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3453 ext4_count_free_inodes(sb));
3454 if (!err)
3455 err = percpu_counter_init(&sbi->s_dirs_counter,
3456 ext4_count_dirs(sb));
3457 if (!err)
3458 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3459 if (err) {
3460 ext4_msg(sb, KERN_ERR, "insufficient memory");
3461 goto failed_mount_wq;
3462 }
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04003463
Mingming Cao4c0425f2009-09-28 15:48:41 -04003464 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3465 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3466 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3467 goto failed_mount_wq;
3468 }
3469
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003470 /*
Mingming Caodab291a2006-10-11 01:21:01 -07003471 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003472 * so we can safely mount the rest of the filesystem now.
3473 */
3474
David Howells1d1fe1e2008-02-07 00:15:37 -08003475 root = ext4_iget(sb, EXT4_ROOT_INO);
3476 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003477 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003478 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003479 goto failed_mount4;
3480 }
3481 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08003482 iput(root);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003483 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003484 goto failed_mount4;
3485 }
David Howells1d1fe1e2008-02-07 00:15:37 -08003486 sb->s_root = d_alloc_root(root);
3487 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003488 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003489 iput(root);
3490 ret = -ENOMEM;
3491 goto failed_mount4;
3492 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003493
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003494 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04003495
3496 /* determine the minimum size of new large inodes, if present */
3497 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3498 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3499 EXT4_GOOD_OLD_INODE_SIZE;
3500 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3501 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3502 if (sbi->s_want_extra_isize <
3503 le16_to_cpu(es->s_want_extra_isize))
3504 sbi->s_want_extra_isize =
3505 le16_to_cpu(es->s_want_extra_isize);
3506 if (sbi->s_want_extra_isize <
3507 le16_to_cpu(es->s_min_extra_isize))
3508 sbi->s_want_extra_isize =
3509 le16_to_cpu(es->s_min_extra_isize);
3510 }
3511 }
3512 /* Check if enough inode space is available */
3513 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3514 sbi->s_inode_size) {
3515 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3516 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04003517 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3518 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04003519 }
3520
Theodore Ts'o90576c02009-09-29 15:51:30 -04003521 if (test_opt(sb, DELALLOC) &&
3522 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003523 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3524 "requested data journaling mode");
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003525 clear_opt(sbi->s_mount_opt, DELALLOC);
Theodore Ts'o90576c02009-09-29 15:51:30 -04003526 }
Jiaying Zhang744692d2010-03-04 16:14:02 -05003527 if (test_opt(sb, DIOREAD_NOLOCK)) {
3528 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3529 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3530 "option - requested data journaling mode");
3531 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3532 }
3533 if (sb->s_blocksize < PAGE_SIZE) {
3534 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3535 "option - block size is too small");
3536 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3537 }
3538 }
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003539
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003540 err = ext4_setup_system_zone(sb);
3541 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003542 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04003543 "zone (%d)", err);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003544 goto failed_mount4;
3545 }
3546
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003547 ext4_ext_init(sb);
3548 err = ext4_mb_init(sb, needs_recovery);
3549 if (err) {
Uwe Kleine-König421f91d2010-06-11 12:17:00 +02003550 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
Eric Sandeenb31e1552009-06-04 17:36:36 -04003551 err);
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003552 goto failed_mount4;
3553 }
3554
Lukas Czernerbfff6872010-10-27 21:30:05 -04003555 err = ext4_register_li_request(sb, first_not_zeroed);
3556 if (err)
3557 goto failed_mount4;
3558
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003559 sbi->s_kobj.kset = ext4_kset;
3560 init_completion(&sbi->s_kobj_unregister);
3561 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3562 "%s", sb->s_id);
3563 if (err) {
3564 ext4_mb_release(sb);
3565 ext4_ext_release(sb);
3566 goto failed_mount4;
3567 };
3568
Mingming Cao617ba132006-10-11 01:20:53 -07003569 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3570 ext4_orphan_cleanup(sb, es);
3571 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05003572 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003573 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05003574 ext4_mark_recovery_complete(sb, es);
3575 }
3576 if (EXT4_SB(sb)->s_journal) {
3577 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3578 descr = " journalled data mode";
3579 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3580 descr = " ordered data mode";
3581 else
3582 descr = " writeback data mode";
3583 } else
3584 descr = "out journal";
3585
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003586 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003587 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3588 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003589
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003590 init_timer(&sbi->s_err_report);
3591 sbi->s_err_report.function = print_daily_error_info;
3592 sbi->s_err_report.data = (unsigned long) sb;
3593 if (es->s_error_count)
3594 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003595
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003596 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003597 return 0;
3598
Mingming Cao617ba132006-10-11 01:20:53 -07003599cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003600 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003601 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003602 goto failed_mount;
3603
3604failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04003605 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04003606 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3607failed_mount_wq:
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003608 ext4_release_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003609 if (sbi->s_journal) {
3610 jbd2_journal_destroy(sbi->s_journal);
3611 sbi->s_journal = NULL;
3612 }
Dmitry Monakhov84061e02010-05-16 08:00:00 -04003613 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3614 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3615 percpu_counter_destroy(&sbi->s_dirs_counter);
3616 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003617failed_mount3:
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04003618 if (sbi->s_flex_groups) {
3619 if (is_vmalloc_addr(sbi->s_flex_groups))
3620 vfree(sbi->s_flex_groups);
3621 else
3622 kfree(sbi->s_flex_groups);
3623 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003624failed_mount2:
3625 for (i = 0; i < db_count; i++)
3626 brelse(sbi->s_group_desc[i]);
3627 kfree(sbi->s_group_desc);
3628failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04003629 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003630 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003631 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003632#ifdef CONFIG_QUOTA
3633 for (i = 0; i < MAXQUOTAS; i++)
3634 kfree(sbi->s_qf_names[i]);
3635#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003636 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003637 brelse(bh);
3638out_fail:
3639 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04003640 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003641 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003642out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003643 kfree(orig_data);
David Howells1d1fe1e2008-02-07 00:15:37 -08003644 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003645}
3646
3647/*
3648 * Setup any per-fs journal parameters now. We'll do this both on
3649 * initial mount, once the journal has been initialised but before we've
3650 * done any recovery; and again on any subsequent remount.
3651 */
Mingming Cao617ba132006-10-11 01:20:53 -07003652static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003653{
Mingming Cao617ba132006-10-11 01:20:53 -07003654 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003655
Theodore Ts'o30773842009-01-03 20:27:38 -05003656 journal->j_commit_interval = sbi->s_commit_interval;
3657 journal->j_min_batch_time = sbi->s_min_batch_time;
3658 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003659
Theodore Ts'oa931da62010-08-03 21:35:12 -04003660 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003661 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07003662 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003663 else
Mingming Caodab291a2006-10-11 01:21:01 -07003664 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04003665 if (test_opt(sb, DATA_ERR_ABORT))
3666 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3667 else
3668 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04003669 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003670}
3671
Mingming Cao617ba132006-10-11 01:20:53 -07003672static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003673 unsigned int journal_inum)
3674{
3675 struct inode *journal_inode;
3676 journal_t *journal;
3677
Frank Mayhar03901312009-01-07 00:06:22 -05003678 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3679
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003680 /* First, test for the existence of a valid inode on disk. Bad
3681 * things happen if we iget() an unused inode, as the subsequent
3682 * iput() will try to delete it. */
3683
David Howells1d1fe1e2008-02-07 00:15:37 -08003684 journal_inode = ext4_iget(sb, journal_inum);
3685 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003686 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003687 return NULL;
3688 }
3689 if (!journal_inode->i_nlink) {
3690 make_bad_inode(journal_inode);
3691 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003692 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003693 return NULL;
3694 }
3695
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003696 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003697 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003698 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003699 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003700 iput(journal_inode);
3701 return NULL;
3702 }
3703
Mingming Caodab291a2006-10-11 01:21:01 -07003704 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003705 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003706 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003707 iput(journal_inode);
3708 return NULL;
3709 }
3710 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003711 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003712 return journal;
3713}
3714
Mingming Cao617ba132006-10-11 01:20:53 -07003715static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003716 dev_t j_dev)
3717{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003718 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003719 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003720 ext4_fsblk_t start;
3721 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003722 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003723 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003724 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003725 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003726 struct block_device *bdev;
3727
Frank Mayhar03901312009-01-07 00:06:22 -05003728 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3729
Eric Sandeenb31e1552009-06-04 17:36:36 -04003730 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003731 if (bdev == NULL)
3732 return NULL;
3733
3734 if (bd_claim(bdev, sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003735 ext4_msg(sb, KERN_ERR,
3736 "failed to claim external journal device");
Al Viro9a1c3542008-02-22 20:40:24 -05003737 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003738 return NULL;
3739 }
3740
3741 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003742 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003743 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003744 ext4_msg(sb, KERN_ERR,
3745 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003746 goto out_bdev;
3747 }
3748
Mingming Cao617ba132006-10-11 01:20:53 -07003749 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3750 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003751 set_blocksize(bdev, blocksize);
3752 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003753 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3754 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003755 goto out_bdev;
3756 }
3757
Mingming Cao617ba132006-10-11 01:20:53 -07003758 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3759 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003760 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003761 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003762 ext4_msg(sb, KERN_ERR, "external journal has "
3763 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003764 brelse(bh);
3765 goto out_bdev;
3766 }
3767
Mingming Cao617ba132006-10-11 01:20:53 -07003768 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003769 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003770 brelse(bh);
3771 goto out_bdev;
3772 }
3773
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003774 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003775 start = sb_block + 1;
3776 brelse(bh); /* we're done with the superblock */
3777
Mingming Caodab291a2006-10-11 01:21:01 -07003778 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003779 start, len, blocksize);
3780 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003781 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003782 goto out_bdev;
3783 }
3784 journal->j_private = sb;
3785 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3786 wait_on_buffer(journal->j_sb_buffer);
3787 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003788 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003789 goto out_journal;
3790 }
3791 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003792 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3793 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003794 be32_to_cpu(journal->j_superblock->s_nr_users));
3795 goto out_journal;
3796 }
Mingming Cao617ba132006-10-11 01:20:53 -07003797 EXT4_SB(sb)->journal_bdev = bdev;
3798 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003799 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003800
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003801out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07003802 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003803out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07003804 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003805 return NULL;
3806}
3807
Mingming Cao617ba132006-10-11 01:20:53 -07003808static int ext4_load_journal(struct super_block *sb,
3809 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003810 unsigned long journal_devnum)
3811{
3812 journal_t *journal;
3813 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3814 dev_t journal_dev;
3815 int err = 0;
3816 int really_read_only;
3817
Frank Mayhar03901312009-01-07 00:06:22 -05003818 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3819
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003820 if (journal_devnum &&
3821 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003822 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3823 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003824 journal_dev = new_decode_dev(journal_devnum);
3825 } else
3826 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3827
3828 really_read_only = bdev_read_only(sb->s_bdev);
3829
3830 /*
3831 * Are we loading a blank journal or performing recovery after a
3832 * crash? For recovery, we need to check in advance whether we
3833 * can get read-write access to the device.
3834 */
Mingming Cao617ba132006-10-11 01:20:53 -07003835 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003836 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003837 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3838 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003839 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003840 ext4_msg(sb, KERN_ERR, "write access "
3841 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003842 return -EROFS;
3843 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04003844 ext4_msg(sb, KERN_INFO, "write access will "
3845 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003846 }
3847 }
3848
3849 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003850 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3851 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003852 return -EINVAL;
3853 }
3854
3855 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07003856 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003857 return -EINVAL;
3858 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07003859 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003860 return -EINVAL;
3861 }
3862
Theodore Ts'o90576c02009-09-29 15:51:30 -04003863 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003864 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04003865
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003866 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07003867 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003868 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003869 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003870 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003871 return err;
3872 }
3873 }
3874
Mingming Cao617ba132006-10-11 01:20:53 -07003875 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07003876 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04003877 if (!err) {
3878 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3879 if (save)
3880 memcpy(save, ((char *) es) +
3881 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07003882 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04003883 if (save)
3884 memcpy(((char *) es) + EXT4_S_ERR_START,
3885 save, EXT4_S_ERR_LEN);
3886 kfree(save);
3887 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003888
3889 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003890 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003891 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003892 return err;
3893 }
3894
Mingming Cao617ba132006-10-11 01:20:53 -07003895 EXT4_SB(sb)->s_journal = journal;
3896 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003897
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04003898 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003899 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3900 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003901
3902 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003903 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003904 }
3905
3906 return 0;
3907}
3908
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003909static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003910{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003911 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07003912 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003913 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003914
3915 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08003916 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04003917 if (buffer_write_io_error(sbh)) {
3918 /*
3919 * Oh, dear. A previous attempt to write the
3920 * superblock failed. This could happen because the
3921 * USB device was yanked out. Or it could happen to
3922 * be a transient write error and maybe the block will
3923 * be remapped. Nothing we can do but to retry the
3924 * write and hope for the best.
3925 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04003926 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3927 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003928 clear_buffer_write_io_error(sbh);
3929 set_buffer_uptodate(sbh);
3930 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04003931 /*
3932 * If the file system is mounted read-only, don't update the
3933 * superblock write time. This avoids updating the superblock
3934 * write time when we are mounting the root file system
3935 * read/only but we need to replay the journal; at that point,
3936 * for people who are east of GMT and who make their clock
3937 * tick in localtime for Windows bug-for-bug compatibility,
3938 * the clock is set in the future, and this will cause e2fsck
3939 * to complain and force a full file system check.
3940 */
3941 if (!(sb->s_flags & MS_RDONLY))
3942 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003943 if (sb->s_bdev->bd_part)
3944 es->s_kbytes_written =
3945 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003946 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3947 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003948 else
3949 es->s_kbytes_written =
3950 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'o7f93cff2010-10-27 21:30:13 -04003951 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeblocks_counter))
3952 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -05003953 &EXT4_SB(sb)->s_freeblocks_counter));
Theodore Ts'o7f93cff2010-10-27 21:30:13 -04003954 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
3955 es->s_free_inodes_count =
3956 cpu_to_le32(percpu_counter_sum_positive(
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -05003957 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003958 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003959 BUFFER_TRACE(sbh, "marking dirty");
3960 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04003961 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08003962 error = sync_dirty_buffer(sbh);
3963 if (error)
3964 return error;
3965
3966 error = buffer_write_io_error(sbh);
3967 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003968 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3969 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003970 clear_buffer_write_io_error(sbh);
3971 set_buffer_uptodate(sbh);
3972 }
3973 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08003974 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003975}
3976
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003977/*
3978 * Have we just finished recovery? If so, and if we are mounting (or
3979 * remounting) the filesystem readonly, then we will end up with a
3980 * consistent fs on disk. Record that fact.
3981 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003982static void ext4_mark_recovery_complete(struct super_block *sb,
3983 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003984{
Mingming Cao617ba132006-10-11 01:20:53 -07003985 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003986
Frank Mayhar03901312009-01-07 00:06:22 -05003987 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3988 BUG_ON(journal != NULL);
3989 return;
3990 }
Mingming Caodab291a2006-10-11 01:21:01 -07003991 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003992 if (jbd2_journal_flush(journal) < 0)
3993 goto out;
3994
Mingming Cao617ba132006-10-11 01:20:53 -07003995 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003996 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07003997 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003998 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003999 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004000
4001out:
Mingming Caodab291a2006-10-11 01:21:01 -07004002 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004003}
4004
4005/*
4006 * If we are mounting (or read-write remounting) a filesystem whose journal
4007 * has recorded an error from a previous lifetime, move that error to the
4008 * main filesystem now.
4009 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004010static void ext4_clear_journal_err(struct super_block *sb,
4011 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004012{
4013 journal_t *journal;
4014 int j_errno;
4015 const char *errstr;
4016
Frank Mayhar03901312009-01-07 00:06:22 -05004017 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4018
Mingming Cao617ba132006-10-11 01:20:53 -07004019 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004020
4021 /*
4022 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004023 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004024 */
4025
Mingming Caodab291a2006-10-11 01:21:01 -07004026 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004027 if (j_errno) {
4028 char nbuf[16];
4029
Mingming Cao617ba132006-10-11 01:20:53 -07004030 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004031 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004032 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004033 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004034
Mingming Cao617ba132006-10-11 01:20:53 -07004035 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4036 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004037 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004038
Mingming Caodab291a2006-10-11 01:21:01 -07004039 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004040 }
4041}
4042
4043/*
4044 * Force the running and committing transactions to commit,
4045 * and wait on the commit.
4046 */
Mingming Cao617ba132006-10-11 01:20:53 -07004047int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004048{
4049 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05004050 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004051
4052 if (sb->s_flags & MS_RDONLY)
4053 return 0;
4054
Mingming Cao617ba132006-10-11 01:20:53 -07004055 journal = EXT4_SB(sb)->s_journal;
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004056 if (journal) {
Eric Sandeen437f88c2010-08-01 17:33:29 -04004057 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Frank Mayhar03901312009-01-07 00:06:22 -05004058 ret = ext4_journal_force_commit(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004059 }
Frank Mayhar03901312009-01-07 00:06:22 -05004060
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004061 return ret;
4062}
4063
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004064static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004065{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004066 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004067 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004068 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004069}
4070
Mingming Cao617ba132006-10-11 01:20:53 -07004071static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004072{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004073 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004074 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004075 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004076
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004077 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004078 flush_workqueue(sbi->dio_unwritten_wq);
4079 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004080 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004081 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05004082 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004083 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004084}
4085
4086/*
4087 * LVM calls this function before a (read-only) snapshot is created. This
4088 * gives us a chance to flush the journal completely and mark the fs clean.
4089 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004090static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004091{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004092 int error = 0;
4093 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004094
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004095 if (sb->s_flags & MS_RDONLY)
4096 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004097
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004098 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004099
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004100 /* Now we set up the journal barrier. */
4101 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004102
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004103 /*
4104 * Don't clear the needs_recovery flag if we failed to flush
4105 * the journal.
4106 */
4107 error = jbd2_journal_flush(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004108 if (error < 0)
4109 goto out;
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004110
4111 /* Journal blocked and flushed, clear needs_recovery flag. */
4112 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4113 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004114out:
4115 /* we rely on s_frozen to stop further updates */
4116 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4117 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004118}
4119
4120/*
4121 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4122 * flag here, even though the filesystem is not technically dirty yet.
4123 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004124static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004125{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004126 if (sb->s_flags & MS_RDONLY)
4127 return 0;
4128
4129 lock_super(sb);
4130 /* Reset the needs_recovery flag before the fs is unlocked. */
4131 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4132 ext4_commit_super(sb, 1);
4133 unlock_super(sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004134 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004135}
4136
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004137static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004138{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004139 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004140 struct ext4_sb_info *sbi = EXT4_SB(sb);
4141 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004142 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004143 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004144 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004145 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004146 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004147 int err;
4148#ifdef CONFIG_QUOTA
4149 int i;
4150#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004151 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004152
4153 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04004154 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004155 old_sb_flags = sb->s_flags;
4156 old_opts.s_mount_opt = sbi->s_mount_opt;
4157 old_opts.s_resuid = sbi->s_resuid;
4158 old_opts.s_resgid = sbi->s_resgid;
4159 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004160 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4161 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004162#ifdef CONFIG_QUOTA
4163 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4164 for (i = 0; i < MAXQUOTAS; i++)
4165 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4166#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004167 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4168 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004169
4170 /*
4171 * Allow the "check" option to be passed as a remount option.
4172 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004173 if (!parse_options(data, sb, NULL, &journal_ioprio,
4174 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004175 err = -EINVAL;
4176 goto restore_opts;
4177 }
4178
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004179 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004180 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004181
4182 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004183 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004184
4185 es = sbi->s_es;
4186
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004187 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004188 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004189 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4190 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004191
4192 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004193 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004194 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004195 err = -EROFS;
4196 goto restore_opts;
4197 }
4198
4199 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004200 err = dquot_suspend(sb, -1);
4201 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004202 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004203
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004204 /*
4205 * First of all, the unconditional stuff we have to do
4206 * to disable replay of the journal when we next remount
4207 */
4208 sb->s_flags |= MS_RDONLY;
4209
4210 /*
4211 * OK, test if we are remounting a valid rw partition
4212 * readonly, and if so set the rdonly flag and then
4213 * mark the partition as valid again.
4214 */
Mingming Cao617ba132006-10-11 01:20:53 -07004215 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4216 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004217 es->s_state = cpu_to_le16(sbi->s_mount_state);
4218
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004219 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004220 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004221 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004222 /* Make sure we can mount this feature set readwrite */
4223 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004224 err = -EROFS;
4225 goto restore_opts;
4226 }
Eric Sandeenead65962007-02-10 01:46:08 -08004227 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004228 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004229 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004230 */
4231 for (g = 0; g < sbi->s_groups_count; g++) {
4232 struct ext4_group_desc *gdp =
4233 ext4_get_group_desc(sb, g, NULL);
4234
4235 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004236 ext4_msg(sb, KERN_ERR,
4237 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04004238 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4239 le16_to_cpu(gdp->bg_checksum));
4240 err = -EINVAL;
4241 goto restore_opts;
4242 }
4243 }
4244
4245 /*
Eric Sandeenead65962007-02-10 01:46:08 -08004246 * If we have an unprocessed orphan list hanging
4247 * around from a previously readonly bdev mount,
4248 * require a full umount/remount for now.
4249 */
4250 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004251 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08004252 "remount RDWR because of unprocessed "
4253 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004254 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08004255 err = -EINVAL;
4256 goto restore_opts;
4257 }
4258
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004259 /*
4260 * Mounting a RDONLY partition read-write, so reread
4261 * and store the current valid flag. (It may have
4262 * been changed by e2fsck since we originally mounted
4263 * the partition.)
4264 */
Frank Mayhar03901312009-01-07 00:06:22 -05004265 if (sbi->s_journal)
4266 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004267 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07004268 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004269 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004270 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004271 sb->s_flags &= ~MS_RDONLY;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004272 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004273 }
4274 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04004275
4276 /*
4277 * Reinitialize lazy itable initialization thread based on
4278 * current settings
4279 */
4280 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4281 ext4_unregister_li_request(sb);
4282 else {
4283 ext4_group_t first_not_zeroed;
4284 first_not_zeroed = ext4_has_uninit_itable(sb);
4285 ext4_register_li_request(sb, first_not_zeroed);
4286 }
4287
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004288 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05004289 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004290 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05004291
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004292#ifdef CONFIG_QUOTA
4293 /* Release old quota file names */
4294 for (i = 0; i < MAXQUOTAS; i++)
4295 if (old_opts.s_qf_names[i] &&
4296 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4297 kfree(old_opts.s_qf_names[i]);
4298#endif
Al Virobbd68512009-05-06 10:43:07 -04004299 unlock_super(sb);
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004300 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004301 dquot_resume(sb, -1);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004302
4303 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4304 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004305 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004306
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004307restore_opts:
4308 sb->s_flags = old_sb_flags;
4309 sbi->s_mount_opt = old_opts.s_mount_opt;
4310 sbi->s_resuid = old_opts.s_resuid;
4311 sbi->s_resgid = old_opts.s_resgid;
4312 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004313 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4314 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004315#ifdef CONFIG_QUOTA
4316 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4317 for (i = 0; i < MAXQUOTAS; i++) {
4318 if (sbi->s_qf_names[i] &&
4319 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4320 kfree(sbi->s_qf_names[i]);
4321 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4322 }
4323#endif
Al Virobbd68512009-05-06 10:43:07 -04004324 unlock_super(sb);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004325 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004326 return err;
4327}
4328
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004329static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004330{
4331 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004332 struct ext4_sb_info *sbi = EXT4_SB(sb);
4333 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08004334 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004335
Badari Pulavarty5e700302007-07-15 23:42:00 -07004336 if (test_opt(sb, MINIX_DF)) {
4337 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004338 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04004339 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07004340 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004341
4342 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07004343 * Compute the overhead (FS structures). This is constant
4344 * for a given filesystem unless the number of block groups
4345 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004346 */
4347
4348 /*
4349 * All of the blocks before first_data_block are
4350 * overhead
4351 */
4352 overhead = le32_to_cpu(es->s_first_data_block);
4353
4354 /*
4355 * Add the overhead attributed to the superblock and
4356 * block group descriptors. If the sparse superblocks
4357 * feature is turned on, then not all groups have this.
4358 */
4359 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07004360 overhead += ext4_bg_has_super(sb, i) +
4361 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004362 cond_resched();
4363 }
4364
4365 /*
4366 * Every block group has an inode bitmap, a block
4367 * bitmap, and an inode table.
4368 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07004369 overhead += ngroups * (2 + sbi->s_itb_per_group);
4370 sbi->s_overhead_last = overhead;
4371 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004372 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004373 }
4374
Mingming Cao617ba132006-10-11 01:20:53 -07004375 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004376 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07004377 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04004378 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4379 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004380 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4381 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004382 buf->f_bavail = 0;
4383 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07004384 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07004385 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08004386 fsid = le64_to_cpup((void *)es->s_uuid) ^
4387 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4388 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4389 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004390
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004391 return 0;
4392}
4393
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004394/* Helper function for writing quotas on sync - we need to start transaction
4395 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004396 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07004397 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01004398 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05004399 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07004400 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004401 *
4402 */
4403
4404#ifdef CONFIG_QUOTA
4405
4406static inline struct inode *dquot_to_inode(struct dquot *dquot)
4407{
4408 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4409}
4410
Mingming Cao617ba132006-10-11 01:20:53 -07004411static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004412{
4413 int ret, err;
4414 handle_t *handle;
4415 struct inode *inode;
4416
4417 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004418 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004419 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004420 if (IS_ERR(handle))
4421 return PTR_ERR(handle);
4422 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004423 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004424 if (!ret)
4425 ret = err;
4426 return ret;
4427}
4428
Mingming Cao617ba132006-10-11 01:20:53 -07004429static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004430{
4431 int ret, err;
4432 handle_t *handle;
4433
Mingming Cao617ba132006-10-11 01:20:53 -07004434 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004435 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004436 if (IS_ERR(handle))
4437 return PTR_ERR(handle);
4438 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004439 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004440 if (!ret)
4441 ret = err;
4442 return ret;
4443}
4444
Mingming Cao617ba132006-10-11 01:20:53 -07004445static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004446{
4447 int ret, err;
4448 handle_t *handle;
4449
Mingming Cao617ba132006-10-11 01:20:53 -07004450 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004451 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07004452 if (IS_ERR(handle)) {
4453 /* Release dquot anyway to avoid endless cycle in dqput() */
4454 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004455 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07004456 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004457 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004458 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004459 if (!ret)
4460 ret = err;
4461 return ret;
4462}
4463
Mingming Cao617ba132006-10-11 01:20:53 -07004464static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004465{
Jan Kara2c8be6b2008-05-13 21:27:55 -04004466 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07004467 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4468 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004469 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004470 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004471 } else {
4472 return dquot_mark_dquot_dirty(dquot);
4473 }
4474}
4475
Mingming Cao617ba132006-10-11 01:20:53 -07004476static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004477{
4478 int ret, err;
4479 handle_t *handle;
4480
4481 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07004482 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004483 if (IS_ERR(handle))
4484 return PTR_ERR(handle);
4485 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07004486 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004487 if (!ret)
4488 ret = err;
4489 return ret;
4490}
4491
4492/*
4493 * Turn on quotas during mount time - we need to find
4494 * the quota file and such...
4495 */
Mingming Cao617ba132006-10-11 01:20:53 -07004496static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004497{
Christoph Hellwig287a8092010-05-19 07:16:45 -04004498 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4499 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004500}
4501
4502/*
4503 * Standard function to be called on quota_on
4504 */
Mingming Cao617ba132006-10-11 01:20:53 -07004505static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Christoph Hellwig307ae182010-05-19 07:16:43 -04004506 char *name)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004507{
4508 int err;
Al Viro82646132008-08-02 00:57:06 -04004509 struct path path;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004510
4511 if (!test_opt(sb, QUOTA))
4512 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04004513
Al Viro82646132008-08-02 00:57:06 -04004514 err = kern_path(name, LOOKUP_FOLLOW, &path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004515 if (err)
4516 return err;
Jan Kara06235432008-05-13 19:11:51 -04004517
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004518 /* Quotafile not on the same filesystem? */
Al Viro82646132008-08-02 00:57:06 -04004519 if (path.mnt->mnt_sb != sb) {
4520 path_put(&path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004521 return -EXDEV;
4522 }
Jan Kara06235432008-05-13 19:11:51 -04004523 /* Journaling quota? */
4524 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004525 /* Quotafile not in fs root? */
Al Viro82646132008-08-02 00:57:06 -04004526 if (path.dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004527 ext4_msg(sb, KERN_WARNING,
4528 "Quota file not on filesystem root. "
4529 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004530 }
Jan Kara06235432008-05-13 19:11:51 -04004531
4532 /*
4533 * When we journal data on quota file, we have to flush journal to see
4534 * all updates to the file when we bypass pagecache...
4535 */
Frank Mayhar03901312009-01-07 00:06:22 -05004536 if (EXT4_SB(sb)->s_journal &&
4537 ext4_should_journal_data(path.dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04004538 /*
4539 * We don't need to lock updates but journal_flush() could
4540 * otherwise be livelocked...
4541 */
4542 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004543 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04004544 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004545 if (err) {
Al Viro82646132008-08-02 00:57:06 -04004546 path_put(&path);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004547 return err;
4548 }
Jan Kara06235432008-05-13 19:11:51 -04004549 }
4550
Christoph Hellwig287a8092010-05-19 07:16:45 -04004551 err = dquot_quota_on_path(sb, type, format_id, &path);
Al Viro82646132008-08-02 00:57:06 -04004552 path_put(&path);
Al Viro77e69da2008-08-01 04:29:18 -04004553 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004554}
4555
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004556static int ext4_quota_off(struct super_block *sb, int type)
4557{
4558 /* Force all delayed allocation blocks to be allocated */
4559 if (test_opt(sb, DELALLOC)) {
4560 down_read(&sb->s_umount);
4561 sync_filesystem(sb);
4562 up_read(&sb->s_umount);
4563 }
4564
4565 return dquot_quota_off(sb, type);
4566}
4567
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004568/* Read data from quotafile - avoid pagecache and such because we cannot afford
4569 * acquiring the locks... As quota files are never truncated and quota code
4570 * itself serializes the operations (and noone else should touch the files)
4571 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07004572static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004573 size_t len, loff_t off)
4574{
4575 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004576 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004577 int err = 0;
4578 int offset = off & (sb->s_blocksize - 1);
4579 int tocopy;
4580 size_t toread;
4581 struct buffer_head *bh;
4582 loff_t i_size = i_size_read(inode);
4583
4584 if (off > i_size)
4585 return 0;
4586 if (off+len > i_size)
4587 len = i_size-off;
4588 toread = len;
4589 while (toread > 0) {
4590 tocopy = sb->s_blocksize - offset < toread ?
4591 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07004592 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004593 if (err)
4594 return err;
4595 if (!bh) /* A hole? */
4596 memset(data, 0, tocopy);
4597 else
4598 memcpy(data, bh->b_data+offset, tocopy);
4599 brelse(bh);
4600 offset = 0;
4601 toread -= tocopy;
4602 data += tocopy;
4603 blk++;
4604 }
4605 return len;
4606}
4607
4608/* Write to quotafile (we know the transaction is already started and has
4609 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07004610static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004611 const char *data, size_t len, loff_t off)
4612{
4613 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004614 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004615 int err = 0;
4616 int offset = off & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004617 struct buffer_head *bh;
4618 handle_t *handle = journal_current_handle();
4619
Frank Mayhar03901312009-01-07 00:06:22 -05004620 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004621 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4622 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07004623 (unsigned long long)off, (unsigned long long)len);
4624 return -EIO;
4625 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004626 /*
4627 * Since we account only one data block in transaction credits,
4628 * then it is impossible to cross a block boundary.
4629 */
4630 if (sb->s_blocksize - offset < len) {
4631 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4632 " cancelled because not block aligned",
4633 (unsigned long long)off, (unsigned long long)len);
4634 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004635 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004636
4637 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4638 bh = ext4_bread(handle, inode, blk, 1, &err);
4639 if (!bh)
4640 goto out;
Jan Kara62d2b5f2010-07-27 11:56:07 -04004641 err = ext4_journal_get_write_access(handle, bh);
4642 if (err) {
4643 brelse(bh);
4644 goto out;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004645 }
4646 lock_buffer(bh);
4647 memcpy(bh->b_data+offset, data, len);
4648 flush_dcache_page(bh->b_page);
4649 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04004650 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004651 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004652out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004653 if (err) {
Jan Kara4d04e4f2008-07-04 09:59:34 -07004654 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004655 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07004656 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004657 if (inode->i_size < off + len) {
4658 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07004659 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004660 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004661 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07004662 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004663 mutex_unlock(&inode->i_mutex);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004664 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004665}
4666
4667#endif
4668
Al Viro152a0832010-07-25 00:46:55 +04004669static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4670 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004671{
Al Viro152a0832010-07-25 00:46:55 +04004672 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004673}
4674
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004675#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004676static struct file_system_type ext2_fs_type = {
4677 .owner = THIS_MODULE,
4678 .name = "ext2",
Al Viro152a0832010-07-25 00:46:55 +04004679 .mount = ext4_mount,
Theodore Ts'o24b58422009-12-07 14:08:51 -05004680 .kill_sb = kill_block_super,
4681 .fs_flags = FS_REQUIRES_DEV,
4682};
4683
4684static inline void register_as_ext2(void)
4685{
4686 int err = register_filesystem(&ext2_fs_type);
4687 if (err)
4688 printk(KERN_WARNING
4689 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4690}
4691
4692static inline void unregister_as_ext2(void)
4693{
4694 unregister_filesystem(&ext2_fs_type);
4695}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004696MODULE_ALIAS("ext2");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004697#else
4698static inline void register_as_ext2(void) { }
4699static inline void unregister_as_ext2(void) { }
4700#endif
4701
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004702#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004703static inline void register_as_ext3(void)
4704{
4705 int err = register_filesystem(&ext3_fs_type);
4706 if (err)
4707 printk(KERN_WARNING
4708 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4709}
4710
4711static inline void unregister_as_ext3(void)
4712{
4713 unregister_filesystem(&ext3_fs_type);
4714}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004715MODULE_ALIAS("ext3");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004716#else
4717static inline void register_as_ext3(void) { }
4718static inline void unregister_as_ext3(void) { }
4719#endif
4720
Theodore Ts'o03010a32008-10-10 20:02:48 -04004721static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004722 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04004723 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04004724 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004725 .kill_sb = kill_block_super,
4726 .fs_flags = FS_REQUIRES_DEV,
4727};
4728
Lukas Czerner857ac882010-10-27 21:30:05 -04004729int __init ext4_init_feat_adverts(void)
4730{
4731 struct ext4_features *ef;
4732 int ret = -ENOMEM;
4733
4734 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4735 if (!ef)
4736 goto out;
4737
4738 ef->f_kobj.kset = ext4_kset;
4739 init_completion(&ef->f_kobj_unregister);
4740 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4741 "features");
4742 if (ret) {
4743 kfree(ef);
4744 goto out;
4745 }
4746
4747 ext4_feat = ef;
4748 ret = 0;
4749out:
4750 return ret;
4751}
4752
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004753static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004754{
Alex Tomasc9de5602008-01-29 00:19:52 -05004755 int err;
4756
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04004757 ext4_check_flag_values();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004758 err = ext4_init_pageio();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004759 if (err)
4760 return err;
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004761 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04004762 if (err)
4763 goto out5;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004764 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4765 if (!ext4_kset)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004766 goto out4;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004767 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Lukas Czerner857ac882010-10-27 21:30:05 -04004768
4769 err = ext4_init_feat_adverts();
4770
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004771 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004772 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004773 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05004774
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004775 err = ext4_init_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004776 if (err)
4777 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004778 err = init_inodecache();
4779 if (err)
4780 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05004781 register_as_ext2();
4782 register_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004783 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004784 if (err)
4785 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004786
4787 ext4_li_info = NULL;
4788 mutex_init(&ext4_li_mtx);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004789 return 0;
4790out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05004791 unregister_as_ext2();
4792 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004793 destroy_inodecache();
4794out1:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004795 ext4_exit_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004796out2:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004797 ext4_exit_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004798out3:
Lukas Czerner857ac882010-10-27 21:30:05 -04004799 kfree(ext4_feat);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004800 remove_proc_entry("fs/ext4", NULL);
4801 kset_unregister(ext4_kset);
4802out4:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004803 ext4_exit_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04004804out5:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004805 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004806 return err;
4807}
4808
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004809static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004810{
Lukas Czernerbfff6872010-10-27 21:30:05 -04004811 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05004812 unregister_as_ext2();
4813 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004814 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004815 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004816 ext4_exit_xattr();
4817 ext4_exit_mballoc();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004818 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004819 kset_unregister(ext4_kset);
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004820 ext4_exit_system_zone();
4821 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004822}
4823
4824MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05004825MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004826MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004827module_init(ext4_init_fs)
4828module_exit(ext4_exit_fs)