blob: 12e726a7073fac69631f8d516c01dbce2babb2e2 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040023#include <linux/vmalloc.h>
Mingming Caodab291a2006-10-11 01:21:01 -070024#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070025#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070031#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070032#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040038#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040039#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040040#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040041#include <linux/crc16.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070042#include <asm/uaccess.h>
43
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040044#include "ext4.h"
45#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070046#include "xattr.h"
47#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040048#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070049
Theodore Ts'o9bffad12009-06-17 11:48:11 -040050#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040053struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040054static struct kset *ext4_kset;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040055
Mingming Cao617ba132006-10-11 01:20:53 -070056static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070057 unsigned long journal_devnum);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040058static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040059static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070063static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040064static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070065 char nbuf[16]);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040066static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080068static int ext4_unfreeze(struct super_block *sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static void ext4_write_super(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080070static int ext4_freeze(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070071
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070072
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070073ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070075{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040076 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070077 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -040078 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070079}
80
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070081ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070083{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040084 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070085 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -040086 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070087}
88
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070089ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070091{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -040092 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070093 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -040094 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070095}
96
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -050097__u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
99{
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500103}
104
105__u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
107{
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500111}
112
113__u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
115{
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500119}
120
121__u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
123{
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500127}
128
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700129void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700131{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700135}
136
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700137void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700139{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700143}
144
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700145void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700147{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700151}
152
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500153void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
155{
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159}
160
161void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
163{
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167}
168
169void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
171{
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175}
176
177void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
179{
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183}
184
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400185
186/* Just increment the non-pointer handle value */
187static handle_t *ext4_get_nojournal(void)
188{
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
191
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
193
194 ref_cnt++;
195 handle = (handle_t *)ref_cnt;
196
197 current->journal_info = handle;
198 return handle;
199}
200
201
202/* Decrement the non-pointer handle value */
203static void ext4_put_nojournal(handle_t *handle)
204{
205 unsigned long ref_cnt = (unsigned long)handle;
206
207 BUG_ON(ref_cnt == 0);
208
209 ref_cnt--;
210 handle = (handle_t *)ref_cnt;
211
212 current->journal_info = handle;
213}
214
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700215/*
Mingming Caodab291a2006-10-11 01:21:01 -0700216 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700217 *
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
221 * appropriate.
222 */
Mingming Cao617ba132006-10-11 01:20:53 -0700223handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700224{
225 journal_t *journal;
226
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
229
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700233 journal = EXT4_SB(sb)->s_journal;
Frank Mayhar03901312009-01-07 00:06:22 -0500234 if (journal) {
235 if (is_journal_aborted(journal)) {
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400236 ext4_abort(sb, __func__, "Detected aborted journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500237 return ERR_PTR(-EROFS);
238 }
239 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700240 }
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400241 return ext4_get_nojournal();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700242}
243
244/*
245 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700247 * that sync() will call the filesystem's write_super callback if
248 * appropriate.
249 */
Mingming Cao617ba132006-10-11 01:20:53 -0700250int __ext4_journal_stop(const char *where, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700251{
252 struct super_block *sb;
253 int err;
254 int rc;
255
Frank Mayhar03901312009-01-07 00:06:22 -0500256 if (!ext4_handle_valid(handle)) {
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400257 ext4_put_nojournal(handle);
Frank Mayhar03901312009-01-07 00:06:22 -0500258 return 0;
259 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700260 sb = handle->h_transaction->t_journal->j_private;
261 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700262 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700263
264 if (!err)
265 err = rc;
266 if (err)
Mingming Cao617ba132006-10-11 01:20:53 -0700267 __ext4_std_error(sb, where, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700268 return err;
269}
270
Mingming Cao617ba132006-10-11 01:20:53 -0700271void ext4_journal_abort_handle(const char *caller, const char *err_fn,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700272 struct buffer_head *bh, handle_t *handle, int err)
273{
274 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700276
Frank Mayhar03901312009-01-07 00:06:22 -0500277 BUG_ON(!ext4_handle_valid(handle));
278
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700279 if (bh)
280 BUFFER_TRACE(bh, "abort");
281
282 if (!handle->h_err)
283 handle->h_err = err;
284
285 if (is_handle_aborted(handle))
286 return;
287
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
290
Mingming Caodab291a2006-10-11 01:21:01 -0700291 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700292}
293
294/* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
296 *
297 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700298 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
303 *
Mingming Caodab291a2006-10-11 01:21:01 -0700304 * We'll just use the jbd2_journal_abort() error code to record an error in
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
307 */
308
Mingming Cao617ba132006-10-11 01:20:53 -0700309static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700310{
Mingming Cao617ba132006-10-11 01:20:53 -0700311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700312
Mingming Cao617ba132006-10-11 01:20:53 -0700313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700315
316 if (sb->s_flags & MS_RDONLY)
317 return;
318
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400319 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700320 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700321
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700323 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700324 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700325 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400326 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700328 sb->s_flags |= MS_RDONLY;
329 }
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400330 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700331 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700332 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700333 sb->s_id);
334}
335
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400336void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700338{
339 va_list args;
340
341 va_start(args, fmt);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346
Mingming Cao617ba132006-10-11 01:20:53 -0700347 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700348}
349
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400350static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700351 char nbuf[16])
352{
353 char *errstr = NULL;
354
355 switch (errno) {
356 case -EIO:
357 errstr = "IO failure";
358 break;
359 case -ENOMEM:
360 errstr = "Out of memory";
361 break;
362 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700365 errstr = "Journal has aborted";
366 else
367 errstr = "Readonly filesystem";
368 break;
369 default:
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
372 * NULL. */
373 if (nbuf) {
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
376 errstr = nbuf;
377 }
378 break;
379 }
380
381 return errstr;
382}
383
Mingming Cao617ba132006-10-11 01:20:53 -0700384/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700385 * automatically and invokes the appropriate error response. */
386
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400387void __ext4_std_error(struct super_block *sb, const char *function, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700388{
389 char nbuf[16];
390 const char *errstr;
391
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
394 * an error. */
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
397 return;
398
Mingming Cao617ba132006-10-11 01:20:53 -0700399 errstr = ext4_decode_error(sb, errno, nbuf);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700402
Mingming Cao617ba132006-10-11 01:20:53 -0700403 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700404}
405
406/*
Mingming Cao617ba132006-10-11 01:20:53 -0700407 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
410 *
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
414 */
415
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400416void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700418{
419 va_list args;
420
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700421 va_start(args, fmt);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700423 vprintk(fmt, args);
424 printk("\n");
425 va_end(args);
426
427 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700428 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700429
430 if (sb->s_flags & MS_RDONLY)
431 return;
432
Eric Sandeenb31e1552009-06-04 17:36:36 -0400433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Mingming Cao617ba132006-10-11 01:20:53 -0700434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700435 sb->s_flags |= MS_RDONLY;
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700439}
440
Eric Sandeenb31e1552009-06-04 17:36:36 -0400441void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
443{
444 va_list args;
445
446 va_start(args, fmt);
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451}
452
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400453void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700455{
456 va_list args;
457
458 va_start(args, fmt);
Mingming Cao617ba132006-10-11 01:20:53 -0700459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700460 sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464}
465
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500466void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400467 const char *function, const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500468__releases(bitlock)
469__acquires(bitlock)
470{
471 va_list args;
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473
474 va_start(args, fmt);
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
476 vprintk(fmt, args);
477 printk("\n");
478 va_end(args);
479
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400483 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500484 return;
485 }
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
488 /*
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
498 */
499 ext4_lock_group(sb, grp);
500 return;
501}
502
Mingming Cao617ba132006-10-11 01:20:53 -0700503void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700504{
Mingming Cao617ba132006-10-11 01:20:53 -0700505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700506
Mingming Cao617ba132006-10-11 01:20:53 -0700507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700508 return;
509
Harvey Harrison46e665e2008-04-17 10:38:59 -0400510 ext4_warning(sb, __func__,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700513 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700514
Mingming Cao617ba132006-10-11 01:20:53 -0700515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
520
521 /*
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
525 */
526}
527
528/*
529 * Open the external journal device
530 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400531static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700532{
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
535
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
537 if (IS_ERR(bdev))
538 goto fail;
539 return bdev;
540
541fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700543 __bdevname(dev, b), PTR_ERR(bdev));
544 return NULL;
545}
546
547/*
548 * Release the journal device
549 */
Mingming Cao617ba132006-10-11 01:20:53 -0700550static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700551{
552 bd_release(bdev);
Al Viro9a1c3542008-02-22 20:40:24 -0500553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700554}
555
Mingming Cao617ba132006-10-11 01:20:53 -0700556static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700557{
558 struct block_device *bdev;
559 int ret = -ENODEV;
560
561 bdev = sbi->journal_bdev;
562 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700563 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700564 sbi->journal_bdev = NULL;
565 }
566 return ret;
567}
568
569static inline struct inode *orphan_list_entry(struct list_head *l)
570{
Mingming Cao617ba132006-10-11 01:20:53 -0700571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700572}
573
Mingming Cao617ba132006-10-11 01:20:53 -0700574static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700575{
576 struct list_head *l;
577
Eric Sandeenb31e1552009-06-04 17:36:36 -0400578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700580
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
584 printk(KERN_ERR " "
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
588 NEXT_ORPHAN(inode));
589 }
590}
591
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400592static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700593{
Mingming Cao617ba132006-10-11 01:20:53 -0700594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400596 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700597
Mingming Cao4c0425f2009-09-28 15:48:41 -0400598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
600
Al Viroa9e220f2009-05-05 22:10:44 -0400601 lock_super(sb);
Christoph Hellwig6cfd0142009-05-05 15:40:36 +0200602 lock_kernel();
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200603 if (sb->s_dirt)
Christoph Hellwigebc1ac12009-05-11 23:35:03 +0200604 ext4_commit_super(sb, 1);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200605
Theodore Ts'o6fd058f2009-05-17 15:38:01 -0400606 ext4_release_system_zone(sb);
Alex Tomasc9de5602008-01-29 00:19:52 -0500607 ext4_mb_release(sb);
Alex Tomasa86c6182006-10-11 01:21:03 -0700608 ext4_ext_release(sb);
Mingming Cao617ba132006-10-11 01:20:53 -0700609 ext4_xattr_put_super(sb);
Frank Mayhar03901312009-01-07 00:06:22 -0500610 if (sbi->s_journal) {
611 err = jbd2_journal_destroy(sbi->s_journal);
612 sbi->s_journal = NULL;
613 if (err < 0)
614 ext4_abort(sb, __func__,
615 "Couldn't clean up the journal");
616 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700617 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700618 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700619 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400620 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700621 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400622 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400623 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400624 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400625 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700626
627 for (i = 0; i < sbi->s_gdb_count; i++)
628 brelse(sbi->s_group_desc[i]);
629 kfree(sbi->s_group_desc);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -0400630 if (is_vmalloc_addr(sbi->s_flex_groups))
631 vfree(sbi->s_flex_groups);
632 else
633 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700634 percpu_counter_destroy(&sbi->s_freeblocks_counter);
635 percpu_counter_destroy(&sbi->s_freeinodes_counter);
636 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400637 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700638 brelse(sbi->s_sbh);
639#ifdef CONFIG_QUOTA
640 for (i = 0; i < MAXQUOTAS; i++)
641 kfree(sbi->s_qf_names[i]);
642#endif
643
644 /* Debugging code just in case the in-memory inode orphan list
645 * isn't empty. The on-disk one can be non-empty if we've
646 * detected an error and taken the fs readonly, but the
647 * in-memory list had better be clean by this point. */
648 if (!list_empty(&sbi->s_orphan))
649 dump_orphan_list(sb, sbi);
650 J_ASSERT(list_empty(&sbi->s_orphan));
651
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700652 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700653 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
654 /*
655 * Invalidate the journal device's buffers. We don't want them
656 * floating about in memory - the physical journal device may
657 * hotswapped, and it breaks the `ro-after' testing code.
658 */
659 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700660 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700661 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700662 }
663 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400664 /*
665 * Now that we are completely done shutting down the
666 * superblock, we need to actually destroy the kobject.
667 */
668 unlock_kernel();
669 unlock_super(sb);
670 kobject_put(&sbi->s_kobj);
671 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500672 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700673 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700674}
675
Christoph Lametere18b8902006-12-06 20:33:20 -0800676static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700677
678/*
679 * Called inside transaction, so use GFP_NOFS
680 */
Mingming Cao617ba132006-10-11 01:20:53 -0700681static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700682{
Mingming Cao617ba132006-10-11 01:20:53 -0700683 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700684
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800685 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700686 if (!ei)
687 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400688
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700689 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400690 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700691 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500692 INIT_LIST_HEAD(&ei->i_prealloc_list);
693 spin_lock_init(&ei->i_prealloc_lock);
Frank Mayhar03901312009-01-07 00:06:22 -0500694 /*
695 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
696 * therefore it can be null here. Don't check it, just initialize
697 * jinode.
698 */
Jan Kara678aaf42008-07-11 19:27:31 -0400699 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
Mingming Caod2a17632008-07-14 17:52:37 -0400700 ei->i_reserved_data_blocks = 0;
701 ei->i_reserved_meta_blocks = 0;
702 ei->i_allocated_meta_blocks = 0;
703 ei->i_delalloc_reserved_flag = 0;
704 spin_lock_init(&(ei->i_block_reservation_lock));
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400705 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
706 ei->cur_aio_dio = NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400707
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700708 return &ei->vfs_inode;
709}
710
Mingming Cao617ba132006-10-11 01:20:53 -0700711static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700712{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700713 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400714 ext4_msg(inode->i_sb, KERN_ERR,
715 "Inode %lu (%p): orphan list check failed!",
716 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700717 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
718 EXT4_I(inode), sizeof(struct ext4_inode_info),
719 true);
720 dump_stack();
721 }
Mingming Cao617ba132006-10-11 01:20:53 -0700722 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700723}
724
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700725static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700726{
Mingming Cao617ba132006-10-11 01:20:53 -0700727 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700728
Christoph Lametera35afb82007-05-16 22:10:57 -0700729 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400730#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700731 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700732#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500733 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700734 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700735}
736
737static int init_inodecache(void)
738{
Mingming Cao617ba132006-10-11 01:20:53 -0700739 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
740 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700741 0, (SLAB_RECLAIM_ACCOUNT|
742 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900743 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700744 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700745 return -ENOMEM;
746 return 0;
747}
748
749static void destroy_inodecache(void)
750{
Mingming Cao617ba132006-10-11 01:20:53 -0700751 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700752}
753
Mingming Cao617ba132006-10-11 01:20:53 -0700754static void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700755{
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400756 ext4_discard_preallocations(inode);
Frank Mayhar03901312009-01-07 00:06:22 -0500757 if (EXT4_JOURNAL(inode))
758 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
Jan Kara678aaf42008-07-11 19:27:31 -0400759 &EXT4_I(inode)->jinode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700760}
761
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400762static inline void ext4_show_quota_options(struct seq_file *seq,
763 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700764{
765#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700766 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700767
768 if (sbi->s_jquota_fmt)
769 seq_printf(seq, ",jqfmt=%s",
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400770 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700771
772 if (sbi->s_qf_names[USRQUOTA])
773 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
774
775 if (sbi->s_qf_names[GRPQUOTA])
776 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
777
Mingming Cao617ba132006-10-11 01:20:53 -0700778 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700779 seq_puts(seq, ",usrquota");
780
Mingming Cao617ba132006-10-11 01:20:53 -0700781 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700782 seq_puts(seq, ",grpquota");
783#endif
784}
785
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700786/*
787 * Show an option if
788 * - it's set to a non-default value OR
789 * - if the per-sb default is different from the global default
790 */
Mingming Cao617ba132006-10-11 01:20:53 -0700791static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700792{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500793 int def_errors;
794 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700795 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700796 struct ext4_sb_info *sbi = EXT4_SB(sb);
797 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700798
799 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500800 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700801
802 if (sbi->s_sb_block != 1)
803 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
804 if (test_opt(sb, MINIX_DF))
805 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500806 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700807 seq_puts(seq, ",grpid");
808 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
809 seq_puts(seq, ",nogrpid");
810 if (sbi->s_resuid != EXT4_DEF_RESUID ||
811 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
812 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
813 }
814 if (sbi->s_resgid != EXT4_DEF_RESGID ||
815 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
816 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
817 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500818 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700819 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500820 def_errors == EXT4_ERRORS_CONTINUE) {
821 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700822 }
823 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500824 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500825 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500826 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700827 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500828 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700829 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500830 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700831 seq_puts(seq, ",debug");
832 if (test_opt(sb, OLDALLOC))
833 seq_puts(seq, ",oldalloc");
Theodore Ts'o03010a32008-10-10 20:02:48 -0400834#ifdef CONFIG_EXT4_FS_XATTR
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500835 if (test_opt(sb, XATTR_USER) &&
836 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700837 seq_puts(seq, ",user_xattr");
838 if (!test_opt(sb, XATTR_USER) &&
839 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
840 seq_puts(seq, ",nouser_xattr");
841 }
842#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -0400843#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500844 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700845 seq_puts(seq, ",acl");
846 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
847 seq_puts(seq, ",noacl");
848#endif
Theodore Ts'o30773842009-01-03 20:27:38 -0500849 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700850 seq_printf(seq, ",commit=%u",
851 (unsigned) (sbi->s_commit_interval / HZ));
852 }
Theodore Ts'o30773842009-01-03 20:27:38 -0500853 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
854 seq_printf(seq, ",min_batch_time=%u",
855 (unsigned) sbi->s_min_batch_time);
856 }
857 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
858 seq_printf(seq, ",max_batch_time=%u",
859 (unsigned) sbi->s_min_batch_time);
860 }
861
Eric Sandeen571640c2008-05-26 12:29:46 -0400862 /*
863 * We're changing the default of barrier mount option, so
864 * let's always display its mount state so it's clear what its
865 * status is.
866 */
867 seq_puts(seq, ",barrier=");
868 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -0400869 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
870 seq_puts(seq, ",journal_async_commit");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700871 if (test_opt(sb, NOBH))
872 seq_puts(seq, ",nobh");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -0500873 if (test_opt(sb, I_VERSION))
874 seq_puts(seq, ",i_version");
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -0400875 if (!test_opt(sb, DELALLOC))
876 seq_puts(seq, ",nodelalloc");
877
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700878
Miklos Szeredicb45bbe2008-01-28 23:58:27 -0500879 if (sbi->s_stripe)
880 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500881 /*
882 * journal mode get enabled in different ways
883 * So just print the value even if we didn't specify it
884 */
Mingming Cao617ba132006-10-11 01:20:53 -0700885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700886 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -0700887 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700888 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -0700889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700890 seq_puts(seq, ",data=writeback");
891
Theodore Ts'o240799c2008-10-09 23:53:47 -0400892 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
893 seq_printf(seq, ",inode_readahead_blks=%u",
894 sbi->s_inode_readahead_blks);
895
Hidehiro Kawai5bf56832008-10-10 22:12:43 -0400896 if (test_opt(sb, DATA_ERR_ABORT))
897 seq_puts(seq, ",data_err=abort");
898
Theodore Ts'oafd46722009-03-16 23:12:23 -0400899 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -0400900 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -0400901
Mingming Cao617ba132006-10-11 01:20:53 -0700902 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400903
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700904 return 0;
905}
906
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700907static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400908 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700909{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700910 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700911
Mingming Cao617ba132006-10-11 01:20:53 -0700912 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700913 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700914 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700915 return ERR_PTR(-ESTALE);
916
917 /* iget isn't really right if the inode is currently unallocated!!
918 *
Mingming Cao617ba132006-10-11 01:20:53 -0700919 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700920 * deleted, so we should be safe.
921 *
922 * Currently we don't know the generation for parent directory, so
923 * a generation of 0 means "accept any"
924 */
David Howells1d1fe1e2008-02-07 00:15:37 -0800925 inode = ext4_iget(sb, ino);
926 if (IS_ERR(inode))
927 return ERR_CAST(inode);
928 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700929 iput(inode);
930 return ERR_PTR(-ESTALE);
931 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700932
933 return inode;
934}
935
936static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400937 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700938{
939 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
940 ext4_nfs_get_inode);
941}
942
943static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400944 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700945{
946 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
947 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700948}
949
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -0500950/*
951 * Try to release metadata pages (indirect blocks, directories) which are
952 * mapped via the block device. Since these pages could have journal heads
953 * which would prevent try_to_free_buffers() from freeing them, we must use
954 * jbd2 layer's try_to_free_buffers() function to release them.
955 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400956static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
957 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -0500958{
959 journal_t *journal = EXT4_SB(sb)->s_journal;
960
961 WARN_ON(PageChecked(page));
962 if (!page_has_buffers(page))
963 return 0;
964 if (journal)
965 return jbd2_journal_try_to_free_buffers(journal, page,
966 wait & ~__GFP_WAIT);
967 return try_to_free_buffers(page);
968}
969
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700970#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400971#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400972#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700973
Mingming Cao617ba132006-10-11 01:20:53 -0700974static int ext4_write_dquot(struct dquot *dquot);
975static int ext4_acquire_dquot(struct dquot *dquot);
976static int ext4_release_dquot(struct dquot *dquot);
977static int ext4_mark_dquot_dirty(struct dquot *dquot);
978static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -0700979static int ext4_quota_on(struct super_block *sb, int type, int format_id,
980 char *path, int remount);
Mingming Cao617ba132006-10-11 01:20:53 -0700981static int ext4_quota_on_mount(struct super_block *sb, int type);
982static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700983 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -0700984static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700985 const char *data, size_t len, loff_t off);
986
Alexey Dobriyan61e225d2009-09-21 17:01:08 -0700987static const struct dquot_operations ext4_quota_operations = {
Jan Karaedf72452009-01-12 19:05:26 +0100988 .initialize = dquot_initialize,
989 .drop = dquot_drop,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700990 .alloc_space = dquot_alloc_space,
Mingming Cao60e58e02009-01-22 18:13:05 +0100991 .reserve_space = dquot_reserve_space,
992 .claim_space = dquot_claim_space,
993 .release_rsv = dquot_release_reserved_space,
994 .get_reserved_space = ext4_get_reserved_space,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700995 .alloc_inode = dquot_alloc_inode,
996 .free_space = dquot_free_space,
997 .free_inode = dquot_free_inode,
998 .transfer = dquot_transfer,
Mingming Cao617ba132006-10-11 01:20:53 -0700999 .write_dquot = ext4_write_dquot,
1000 .acquire_dquot = ext4_acquire_dquot,
1001 .release_dquot = ext4_release_dquot,
1002 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001003 .write_info = ext4_write_info,
1004 .alloc_dquot = dquot_alloc,
1005 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001006};
1007
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001008static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001009 .quota_on = ext4_quota_on,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001010 .quota_off = vfs_quota_off,
1011 .quota_sync = vfs_quota_sync,
1012 .get_info = vfs_get_dqinfo,
1013 .set_info = vfs_set_dqinfo,
1014 .get_dqblk = vfs_get_dqblk,
1015 .set_dqblk = vfs_set_dqblk
1016};
1017#endif
1018
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001019static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001020 .alloc_inode = ext4_alloc_inode,
1021 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001022 .write_inode = ext4_write_inode,
1023 .dirty_inode = ext4_dirty_inode,
1024 .delete_inode = ext4_delete_inode,
1025 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001026 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001027 .freeze_fs = ext4_freeze,
1028 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001029 .statfs = ext4_statfs,
1030 .remount_fs = ext4_remount,
1031 .clear_inode = ext4_clear_inode,
1032 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001033#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001034 .quota_read = ext4_quota_read,
1035 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001036#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001037 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001038};
1039
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001040static const struct super_operations ext4_nojournal_sops = {
1041 .alloc_inode = ext4_alloc_inode,
1042 .destroy_inode = ext4_destroy_inode,
1043 .write_inode = ext4_write_inode,
1044 .dirty_inode = ext4_dirty_inode,
1045 .delete_inode = ext4_delete_inode,
1046 .write_super = ext4_write_super,
1047 .put_super = ext4_put_super,
1048 .statfs = ext4_statfs,
1049 .remount_fs = ext4_remount,
1050 .clear_inode = ext4_clear_inode,
1051 .show_options = ext4_show_options,
1052#ifdef CONFIG_QUOTA
1053 .quota_read = ext4_quota_read,
1054 .quota_write = ext4_quota_write,
1055#endif
1056 .bdev_try_to_free_page = bdev_try_to_free_page,
1057};
1058
Christoph Hellwig39655162007-10-21 16:42:17 -07001059static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001060 .fh_to_dentry = ext4_fh_to_dentry,
1061 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001062 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001063};
1064
1065enum {
1066 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1067 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001068 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001069 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001070 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001071 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001072 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001073 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001074 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001075 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001076 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1077 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001078 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1079 Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001080 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001081 Opt_block_validity, Opt_noblock_validity,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001082 Opt_inode_readahead_blks, Opt_journal_ioprio
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001083};
1084
Steven Whitehousea447c092008-10-13 10:46:57 +01001085static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001086 {Opt_bsd_df, "bsddf"},
1087 {Opt_minix_df, "minixdf"},
1088 {Opt_grpid, "grpid"},
1089 {Opt_grpid, "bsdgroups"},
1090 {Opt_nogrpid, "nogrpid"},
1091 {Opt_nogrpid, "sysvgroups"},
1092 {Opt_resgid, "resgid=%u"},
1093 {Opt_resuid, "resuid=%u"},
1094 {Opt_sb, "sb=%u"},
1095 {Opt_err_cont, "errors=continue"},
1096 {Opt_err_panic, "errors=panic"},
1097 {Opt_err_ro, "errors=remount-ro"},
1098 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001099 {Opt_debug, "debug"},
1100 {Opt_oldalloc, "oldalloc"},
1101 {Opt_orlov, "orlov"},
1102 {Opt_user_xattr, "user_xattr"},
1103 {Opt_nouser_xattr, "nouser_xattr"},
1104 {Opt_acl, "acl"},
1105 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001106 {Opt_noload, "noload"},
1107 {Opt_nobh, "nobh"},
1108 {Opt_bh, "bh"},
1109 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001110 {Opt_min_batch_time, "min_batch_time=%u"},
1111 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001112 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001113 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001114 {Opt_journal_checksum, "journal_checksum"},
1115 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001116 {Opt_abort, "abort"},
1117 {Opt_data_journal, "data=journal"},
1118 {Opt_data_ordered, "data=ordered"},
1119 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001120 {Opt_data_err_abort, "data_err=abort"},
1121 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001122 {Opt_offusrjquota, "usrjquota="},
1123 {Opt_usrjquota, "usrjquota=%s"},
1124 {Opt_offgrpjquota, "grpjquota="},
1125 {Opt_grpjquota, "grpjquota=%s"},
1126 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1127 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1128 {Opt_grpquota, "grpquota"},
1129 {Opt_noquota, "noquota"},
1130 {Opt_quota, "quota"},
1131 {Opt_usrquota, "usrquota"},
1132 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001133 {Opt_barrier, "barrier"},
1134 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001135 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001136 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001137 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001138 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001139 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001140 {Opt_block_validity, "block_validity"},
1141 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001142 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001143 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001144 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001145 {Opt_auto_da_alloc, "auto_da_alloc"},
1146 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001147 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001148};
1149
Mingming Cao617ba132006-10-11 01:20:53 -07001150static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001151{
Mingming Cao617ba132006-10-11 01:20:53 -07001152 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001153 char *options = (char *) *data;
1154
1155 if (!options || strncmp(options, "sb=", 3) != 0)
1156 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001157
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001158 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001159 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001160 sb_block = simple_strtoul(options, &options, 0);
1161 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001162 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001163 (char *) *data);
1164 return 1;
1165 }
1166 if (*options == ',')
1167 options++;
1168 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001169
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001170 return sb_block;
1171}
1172
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001173#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1174
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001175static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001176 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001177 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001178 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001179{
Mingming Cao617ba132006-10-11 01:20:53 -07001180 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001181 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001182 substring_t args[MAX_OPT_ARGS];
1183 int data_opt = 0;
1184 int option;
1185#ifdef CONFIG_QUOTA
Jan Karadfc5d032008-05-13 19:11:51 -04001186 int qtype, qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001187 char *qname;
1188#endif
1189
1190 if (!options)
1191 return 1;
1192
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001193 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001194 int token;
1195 if (!*p)
1196 continue;
1197
1198 token = match_token(p, tokens, args);
1199 switch (token) {
1200 case Opt_bsd_df:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001201 clear_opt(sbi->s_mount_opt, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001202 break;
1203 case Opt_minix_df:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001204 set_opt(sbi->s_mount_opt, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001205 break;
1206 case Opt_grpid:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001207 set_opt(sbi->s_mount_opt, GRPID);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001208 break;
1209 case Opt_nogrpid:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001210 clear_opt(sbi->s_mount_opt, GRPID);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001211 break;
1212 case Opt_resuid:
1213 if (match_int(&args[0], &option))
1214 return 0;
1215 sbi->s_resuid = option;
1216 break;
1217 case Opt_resgid:
1218 if (match_int(&args[0], &option))
1219 return 0;
1220 sbi->s_resgid = option;
1221 break;
1222 case Opt_sb:
1223 /* handled by get_sb_block() instead of here */
1224 /* *sb_block = match_int(&args[0]); */
1225 break;
1226 case Opt_err_panic:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001227 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1228 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1229 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001230 break;
1231 case Opt_err_ro:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001232 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1233 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1234 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001235 break;
1236 case Opt_err_cont:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001237 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1238 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1239 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001240 break;
1241 case Opt_nouid32:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001242 set_opt(sbi->s_mount_opt, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001243 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001244 case Opt_debug:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001245 set_opt(sbi->s_mount_opt, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001246 break;
1247 case Opt_oldalloc:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001248 set_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001249 break;
1250 case Opt_orlov:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001251 clear_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001252 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001253#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001254 case Opt_user_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001255 set_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001256 break;
1257 case Opt_nouser_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001258 clear_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001259 break;
1260#else
1261 case Opt_user_xattr:
1262 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001263 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001264 break;
1265#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001266#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001267 case Opt_acl:
1268 set_opt(sbi->s_mount_opt, POSIX_ACL);
1269 break;
1270 case Opt_noacl:
1271 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1272 break;
1273#else
1274 case Opt_acl:
1275 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001276 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001277 break;
1278#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001279 case Opt_journal_update:
1280 /* @@@ FIXME */
1281 /* Eventually we will want to be able to create
1282 a journal file here. For now, only allow the
1283 user to specify an existing inode to be the
1284 journal file. */
1285 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001286 ext4_msg(sb, KERN_ERR,
1287 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001288 return 0;
1289 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001290 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001291 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001292 case Opt_journal_dev:
1293 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001294 ext4_msg(sb, KERN_ERR,
1295 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001296 return 0;
1297 }
1298 if (match_int(&args[0], &option))
1299 return 0;
1300 *journal_devnum = option;
1301 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001302 case Opt_journal_checksum:
Theodore Ts'od0646f72009-09-05 12:50:43 -04001303 break; /* Kept for backwards compatibility */
Girish Shilamkar818d2762008-01-28 23:58:27 -05001304 case Opt_journal_async_commit:
1305 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001306 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001307 case Opt_noload:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001308 set_opt(sbi->s_mount_opt, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001309 break;
1310 case Opt_commit:
1311 if (match_int(&args[0], &option))
1312 return 0;
1313 if (option < 0)
1314 return 0;
1315 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001316 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001317 sbi->s_commit_interval = HZ * option;
1318 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001319 case Opt_max_batch_time:
1320 if (match_int(&args[0], &option))
1321 return 0;
1322 if (option < 0)
1323 return 0;
1324 if (option == 0)
1325 option = EXT4_DEF_MAX_BATCH_TIME;
1326 sbi->s_max_batch_time = option;
1327 break;
1328 case Opt_min_batch_time:
1329 if (match_int(&args[0], &option))
1330 return 0;
1331 if (option < 0)
1332 return 0;
1333 sbi->s_min_batch_time = option;
1334 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001335 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001336 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001337 goto datacheck;
1338 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001339 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001340 goto datacheck;
1341 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001342 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001343 datacheck:
1344 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001345 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001346 != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001347 ext4_msg(sb, KERN_ERR,
1348 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001349 return 0;
1350 }
1351 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07001352 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001353 sbi->s_mount_opt |= data_opt;
1354 }
1355 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001356 case Opt_data_err_abort:
1357 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1358 break;
1359 case Opt_data_err_ignore:
1360 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1361 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001362#ifdef CONFIG_QUOTA
1363 case Opt_usrjquota:
1364 qtype = USRQUOTA;
1365 goto set_qf_name;
1366 case Opt_grpjquota:
1367 qtype = GRPQUOTA;
1368set_qf_name:
Jan Kara17bd13b2008-08-20 18:14:35 +02001369 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001370 !sbi->s_qf_names[qtype]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001371 ext4_msg(sb, KERN_ERR,
1372 "Cannot change journaled "
1373 "quota options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001374 return 0;
1375 }
1376 qname = match_strdup(&args[0]);
1377 if (!qname) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001378 ext4_msg(sb, KERN_ERR,
1379 "Not enough memory for "
1380 "storing quotafile name");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001381 return 0;
1382 }
1383 if (sbi->s_qf_names[qtype] &&
1384 strcmp(sbi->s_qf_names[qtype], qname)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001385 ext4_msg(sb, KERN_ERR,
1386 "%s quota file already "
1387 "specified", QTYPE2NAME(qtype));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001388 kfree(qname);
1389 return 0;
1390 }
1391 sbi->s_qf_names[qtype] = qname;
1392 if (strchr(sbi->s_qf_names[qtype], '/')) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001393 ext4_msg(sb, KERN_ERR,
1394 "quotafile must be on "
1395 "filesystem root");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001396 kfree(sbi->s_qf_names[qtype]);
1397 sbi->s_qf_names[qtype] = NULL;
1398 return 0;
1399 }
1400 set_opt(sbi->s_mount_opt, QUOTA);
1401 break;
1402 case Opt_offusrjquota:
1403 qtype = USRQUOTA;
1404 goto clear_qf_name;
1405 case Opt_offgrpjquota:
1406 qtype = GRPQUOTA;
1407clear_qf_name:
Jan Kara17bd13b2008-08-20 18:14:35 +02001408 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001409 sbi->s_qf_names[qtype]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001410 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001411 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001412 "quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001413 return 0;
1414 }
1415 /*
1416 * The space will be released later when all options
1417 * are confirmed to be correct
1418 */
1419 sbi->s_qf_names[qtype] = NULL;
1420 break;
1421 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001422 qfmt = QFMT_VFS_OLD;
1423 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001424 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001425 qfmt = QFMT_VFS_V0;
1426set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001427 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001428 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001429 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001430 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001431 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001432 return 0;
1433 }
1434 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001435 break;
1436 case Opt_quota:
1437 case Opt_usrquota:
1438 set_opt(sbi->s_mount_opt, QUOTA);
1439 set_opt(sbi->s_mount_opt, USRQUOTA);
1440 break;
1441 case Opt_grpquota:
1442 set_opt(sbi->s_mount_opt, QUOTA);
1443 set_opt(sbi->s_mount_opt, GRPQUOTA);
1444 break;
1445 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001446 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001447 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1448 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001449 return 0;
1450 }
1451 clear_opt(sbi->s_mount_opt, QUOTA);
1452 clear_opt(sbi->s_mount_opt, USRQUOTA);
1453 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1454 break;
1455#else
1456 case Opt_quota:
1457 case Opt_usrquota:
1458 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001459 ext4_msg(sb, KERN_ERR,
1460 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001461 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001462 case Opt_usrjquota:
1463 case Opt_grpjquota:
1464 case Opt_offusrjquota:
1465 case Opt_offgrpjquota:
1466 case Opt_jqfmt_vfsold:
1467 case Opt_jqfmt_vfsv0:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001468 ext4_msg(sb, KERN_ERR,
1469 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001470 break;
1471 case Opt_noquota:
1472 break;
1473#endif
1474 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001475 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001476 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001477 case Opt_nobarrier:
1478 clear_opt(sbi->s_mount_opt, BARRIER);
1479 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001480 case Opt_barrier:
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001481 if (match_int(&args[0], &option)) {
1482 set_opt(sbi->s_mount_opt, BARRIER);
1483 break;
1484 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001485 if (option)
1486 set_opt(sbi->s_mount_opt, BARRIER);
1487 else
1488 clear_opt(sbi->s_mount_opt, BARRIER);
1489 break;
1490 case Opt_ignore:
1491 break;
1492 case Opt_resize:
1493 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001494 ext4_msg(sb, KERN_ERR,
1495 "resize option only available "
1496 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001497 return 0;
1498 }
1499 if (match_int(&args[0], &option) != 0)
1500 return 0;
1501 *n_blocks_count = option;
1502 break;
1503 case Opt_nobh:
1504 set_opt(sbi->s_mount_opt, NOBH);
1505 break;
1506 case Opt_bh:
1507 clear_opt(sbi->s_mount_opt, NOBH);
1508 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001509 case Opt_i_version:
1510 set_opt(sbi->s_mount_opt, I_VERSION);
1511 sb->s_flags |= MS_I_VERSION;
1512 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001513 case Opt_nodelalloc:
1514 clear_opt(sbi->s_mount_opt, DELALLOC);
1515 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001516 case Opt_stripe:
1517 if (match_int(&args[0], &option))
1518 return 0;
1519 if (option < 0)
1520 return 0;
1521 sbi->s_stripe = option;
1522 break;
Alex Tomas64769242008-07-11 19:27:31 -04001523 case Opt_delalloc:
1524 set_opt(sbi->s_mount_opt, DELALLOC);
1525 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001526 case Opt_block_validity:
1527 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1528 break;
1529 case Opt_noblock_validity:
1530 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1531 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001532 case Opt_inode_readahead_blks:
1533 if (match_int(&args[0], &option))
1534 return 0;
1535 if (option < 0 || option > (1 << 30))
1536 return 0;
Theodore Ts'of7c43952009-04-24 23:31:59 -04001537 if (!is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001538 ext4_msg(sb, KERN_ERR,
1539 "EXT4-fs: inode_readahead_blks"
1540 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001541 return 0;
1542 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001543 sbi->s_inode_readahead_blks = option;
1544 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001545 case Opt_journal_ioprio:
1546 if (match_int(&args[0], &option))
1547 return 0;
1548 if (option < 0 || option > 7)
1549 break;
1550 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1551 option);
1552 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001553 case Opt_noauto_da_alloc:
1554 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1555 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001556 case Opt_auto_da_alloc:
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001557 if (match_int(&args[0], &option)) {
1558 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1559 break;
1560 }
Theodore Ts'oafd46722009-03-16 23:12:23 -04001561 if (option)
1562 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1563 else
1564 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1565 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001566 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001567 ext4_msg(sb, KERN_ERR,
1568 "Unrecognized mount option \"%s\" "
1569 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001570 return 0;
1571 }
1572 }
1573#ifdef CONFIG_QUOTA
1574 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Mingming Cao617ba132006-10-11 01:20:53 -07001575 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001576 sbi->s_qf_names[USRQUOTA])
1577 clear_opt(sbi->s_mount_opt, USRQUOTA);
1578
Mingming Cao617ba132006-10-11 01:20:53 -07001579 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001580 sbi->s_qf_names[GRPQUOTA])
1581 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1582
1583 if ((sbi->s_qf_names[USRQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001584 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001585 (sbi->s_qf_names[GRPQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001586 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001587 ext4_msg(sb, KERN_ERR, "old and new quota "
1588 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001589 return 0;
1590 }
1591
1592 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001593 ext4_msg(sb, KERN_ERR, "journaled quota format "
1594 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001595 return 0;
1596 }
1597 } else {
1598 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001599 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001600 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001601 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001602 return 0;
1603 }
1604 }
1605#endif
1606 return 1;
1607}
1608
Mingming Cao617ba132006-10-11 01:20:53 -07001609static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001610 int read_only)
1611{
Mingming Cao617ba132006-10-11 01:20:53 -07001612 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001613 int res = 0;
1614
Mingming Cao617ba132006-10-11 01:20:53 -07001615 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001616 ext4_msg(sb, KERN_ERR, "revision level too high, "
1617 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001618 res = MS_RDONLY;
1619 }
1620 if (read_only)
1621 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001622 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001623 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1624 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001625 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001626 ext4_msg(sb, KERN_WARNING,
1627 "warning: mounting fs with errors, "
1628 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001629 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1630 le16_to_cpu(es->s_mnt_count) >=
1631 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001632 ext4_msg(sb, KERN_WARNING,
1633 "warning: maximal mount count reached, "
1634 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001635 else if (le32_to_cpu(es->s_checkinterval) &&
1636 (le32_to_cpu(es->s_lastcheck) +
1637 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001638 ext4_msg(sb, KERN_WARNING,
1639 "warning: checktime reached, "
1640 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001641 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001642 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001643 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001644 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001645 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001646 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001647 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001648 if (sbi->s_journal)
1649 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001650
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001651 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001652 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001653 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'o7f4520c2009-06-13 10:09:41 -04001654 "bpg=%lu, ipg=%lu, mo=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001655 sb->s_blocksize,
1656 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001657 EXT4_BLOCKS_PER_GROUP(sb),
1658 EXT4_INODES_PER_GROUP(sb),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001659 sbi->s_mount_opt);
1660
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001661 return res;
1662}
1663
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001664static int ext4_fill_flex_info(struct super_block *sb)
1665{
1666 struct ext4_sb_info *sbi = EXT4_SB(sb);
1667 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001668 ext4_group_t flex_group_count;
1669 ext4_group_t flex_group;
1670 int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001671 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001672 int i;
1673
1674 if (!sbi->s_es->s_log_groups_per_flex) {
1675 sbi->s_log_groups_per_flex = 0;
1676 return 1;
1677 }
1678
1679 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1680 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1681
Frederic Bohec62a11f2008-09-08 10:20:24 -04001682 /* We allocate both existing and potentially added groups */
1683 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05001684 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1685 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001686 size = flex_group_count * sizeof(struct flex_groups);
1687 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1688 if (sbi->s_flex_groups == NULL) {
1689 sbi->s_flex_groups = vmalloc(size);
1690 if (sbi->s_flex_groups)
1691 memset(sbi->s_flex_groups, 0, size);
1692 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001693 if (sbi->s_flex_groups == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001694 ext4_msg(sb, KERN_ERR, "not enough memory for "
1695 "%u flex groups", flex_group_count);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001696 goto failed;
1697 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001698
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001699 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001700 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001701
1702 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001703 atomic_add(ext4_free_inodes_count(sb, gdp),
1704 &sbi->s_flex_groups[flex_group].free_inodes);
1705 atomic_add(ext4_free_blks_count(sb, gdp),
1706 &sbi->s_flex_groups[flex_group].free_blocks);
1707 atomic_add(ext4_used_dirs_count(sb, gdp),
1708 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001709 }
1710
1711 return 1;
1712failed:
1713 return 0;
1714}
1715
Andreas Dilger717d50e2007-10-16 18:38:25 -04001716__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1717 struct ext4_group_desc *gdp)
1718{
1719 __u16 crc = 0;
1720
1721 if (sbi->s_es->s_feature_ro_compat &
1722 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1723 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1724 __le32 le_group = cpu_to_le32(block_group);
1725
1726 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1727 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1728 crc = crc16(crc, (__u8 *)gdp, offset);
1729 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1730 /* for checksum of struct ext4_group_desc do the rest...*/
1731 if ((sbi->s_es->s_feature_incompat &
1732 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1733 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1734 crc = crc16(crc, (__u8 *)gdp + offset,
1735 le16_to_cpu(sbi->s_es->s_desc_size) -
1736 offset);
1737 }
1738
1739 return cpu_to_le16(crc);
1740}
1741
1742int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1743 struct ext4_group_desc *gdp)
1744{
1745 if ((sbi->s_es->s_feature_ro_compat &
1746 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1747 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1748 return 0;
1749
1750 return 1;
1751}
1752
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001753/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001754static int ext4_check_descriptors(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001755{
Mingming Cao617ba132006-10-11 01:20:53 -07001756 struct ext4_sb_info *sbi = EXT4_SB(sb);
1757 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1758 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001759 ext4_fsblk_t block_bitmap;
1760 ext4_fsblk_t inode_bitmap;
1761 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04001762 int flexbg_flag = 0;
Avantika Mathurfd2d4292008-01-28 23:58:27 -05001763 ext4_group_t i;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001764
Jose R. Santosce421582007-10-16 18:38:25 -04001765 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1766 flexbg_flag = 1;
1767
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001768 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001769
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001770 for (i = 0; i < sbi->s_groups_count; i++) {
1771 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1772
Jose R. Santosce421582007-10-16 18:38:25 -04001773 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001774 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001775 else
1776 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07001777 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001778
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001779 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001780 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001781 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001782 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001783 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001784 return 0;
1785 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001786 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001787 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001788 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001789 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001790 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001791 return 0;
1792 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001793 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001794 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001795 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001796 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001797 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001798 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001799 return 0;
1800 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001801 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001802 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001803 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1804 "Checksum for group %u failed (%u!=%u)",
1805 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1806 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04001807 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001808 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04001809 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04001810 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04001811 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001812 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04001813 if (!flexbg_flag)
1814 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001815 }
1816
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001817 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001818 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001819 return 1;
1820}
1821
Mingming Cao617ba132006-10-11 01:20:53 -07001822/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001823 * the superblock) which were deleted from all directories, but held open by
1824 * a process at the time of a crash. We walk the list and try to delete these
1825 * inodes at recovery time (only with a read-write filesystem).
1826 *
1827 * In order to keep the orphan inode chain consistent during traversal (in
1828 * case of crash during recovery), we link each inode into the superblock
1829 * orphan list_head and handle it the same way as an inode deletion during
1830 * normal operation (which journals the operations for us).
1831 *
1832 * We only do an iget() and an iput() on each inode, which is very safe if we
1833 * accidentally point at an in-use or already deleted inode. The worst that
1834 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07001835 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001836 * e2fsck was run on this filesystem, and it must have already done the orphan
1837 * inode cleanup for us, so we can safely abort without any further action.
1838 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001839static void ext4_orphan_cleanup(struct super_block *sb,
1840 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001841{
1842 unsigned int s_flags = sb->s_flags;
1843 int nr_orphans = 0, nr_truncates = 0;
1844#ifdef CONFIG_QUOTA
1845 int i;
1846#endif
1847 if (!es->s_last_orphan) {
1848 jbd_debug(4, "no orphan inodes to clean up\n");
1849 return;
1850 }
1851
Eric Sandeena8f48a92006-12-06 20:40:13 -08001852 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001853 ext4_msg(sb, KERN_ERR, "write access "
1854 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08001855 return;
1856 }
1857
Mingming Cao617ba132006-10-11 01:20:53 -07001858 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001859 if (es->s_last_orphan)
1860 jbd_debug(1, "Errors on filesystem, "
1861 "clearing orphan list.\n");
1862 es->s_last_orphan = 0;
1863 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1864 return;
1865 }
1866
1867 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001868 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 sb->s_flags &= ~MS_RDONLY;
1870 }
1871#ifdef CONFIG_QUOTA
1872 /* Needed for iput() to work correctly and not trash data */
1873 sb->s_flags |= MS_ACTIVE;
1874 /* Turn on quotas so that they are updated correctly */
1875 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07001876 if (EXT4_SB(sb)->s_qf_names[i]) {
1877 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001878 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001879 ext4_msg(sb, KERN_ERR,
1880 "Cannot turn on journaled "
1881 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001882 }
1883 }
1884#endif
1885
1886 while (es->s_last_orphan) {
1887 struct inode *inode;
1888
Josef Bacik97bd42b2008-04-29 22:04:56 -04001889 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1890 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001891 es->s_last_orphan = 0;
1892 break;
1893 }
1894
Mingming Cao617ba132006-10-11 01:20:53 -07001895 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Jan Karaa269eb12009-01-26 17:04:39 +01001896 vfs_dq_init(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001897 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001898 ext4_msg(sb, KERN_DEBUG,
1899 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04001900 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04001901 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001902 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07001903 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001904 nr_truncates++;
1905 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001906 ext4_msg(sb, KERN_DEBUG,
1907 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04001908 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001909 jbd_debug(2, "deleting unreferenced inode %lu\n",
1910 inode->i_ino);
1911 nr_orphans++;
1912 }
1913 iput(inode); /* The delete magic happens here! */
1914 }
1915
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001916#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001917
1918 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001919 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1920 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001921 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001922 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1923 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001924#ifdef CONFIG_QUOTA
1925 /* Turn quotas off */
1926 for (i = 0; i < MAXQUOTAS; i++) {
1927 if (sb_dqopt(sb)->files[i])
Jan Kara6f28e082008-04-28 02:14:34 -07001928 vfs_quota_off(sb, i, 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001929 }
1930#endif
1931 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1932}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001933
Eric Sandeencd2291a2008-01-28 23:58:27 -05001934/*
1935 * Maximal extent format file size.
1936 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1937 * extent format containers, within a sector_t, and within i_blocks
1938 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1939 * so that won't be a limiting factor.
1940 *
1941 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1942 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04001943static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05001944{
1945 loff_t res;
1946 loff_t upper_limit = MAX_LFS_FILESIZE;
1947
1948 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04001949 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05001950 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02001951 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05001952 * i_block represent total blocks in 512 bytes
1953 * 32 == size of vfs inode i_blocks * 8
1954 */
1955 upper_limit = (1LL << 32) - 1;
1956
1957 /* total blocks in file system block size */
1958 upper_limit >>= (blkbits - 9);
1959 upper_limit <<= blkbits;
1960 }
1961
1962 /* 32-bit extent-start container, ee_block */
1963 res = 1LL << 32;
1964 res <<= blkbits;
1965 res -= 1;
1966
1967 /* Sanity check against vm- & vfs- imposed limits */
1968 if (res > upper_limit)
1969 res = upper_limit;
1970
1971 return res;
1972}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001973
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001974/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05001975 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001976 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1977 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001978 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04001979static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001980{
Mingming Cao617ba132006-10-11 01:20:53 -07001981 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001982 int meta_blocks;
1983 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001984 /* This is calculated to be the largest file size for a dense, block
1985 * mapped file such that the file's total number of 512-byte sectors,
1986 * including data and all indirect blocks, does not exceed (2^48 - 1).
1987 *
1988 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1989 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001990 */
1991
Theodore Ts'of287a1a2008-10-16 22:50:48 -04001992 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001993 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02001994 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001995 * the inode i_block field represents total file blocks in
1996 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05001997 */
1998 upper_limit = (1LL << 32) - 1;
1999
2000 /* total blocks in file system block size */
2001 upper_limit >>= (bits - 9);
2002
2003 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002004 /*
2005 * We use 48 bit ext4_inode i_blocks
2006 * With EXT4_HUGE_FILE_FL set the i_blocks
2007 * represent total number of blocks in
2008 * file system block size
2009 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002010 upper_limit = (1LL << 48) - 1;
2011
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002012 }
2013
2014 /* indirect blocks */
2015 meta_blocks = 1;
2016 /* double indirect blocks */
2017 meta_blocks += 1 + (1LL << (bits-2));
2018 /* tripple indirect blocks */
2019 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2020
2021 upper_limit -= meta_blocks;
2022 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002023
2024 res += 1LL << (bits-2);
2025 res += 1LL << (2*(bits-2));
2026 res += 1LL << (3*(bits-2));
2027 res <<= bits;
2028 if (res > upper_limit)
2029 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002030
2031 if (res > MAX_LFS_FILESIZE)
2032 res = MAX_LFS_FILESIZE;
2033
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002034 return res;
2035}
2036
Mingming Cao617ba132006-10-11 01:20:53 -07002037static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002038 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002039{
Mingming Cao617ba132006-10-11 01:20:53 -07002040 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002041 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002042 int has_super = 0;
2043
2044 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2045
Mingming Cao617ba132006-10-11 01:20:53 -07002046 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002047 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002048 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002049 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002050 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002051 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002052
Mingming Cao617ba132006-10-11 01:20:53 -07002053 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002054}
2055
Alex Tomasc9de5602008-01-29 00:19:52 -05002056/**
2057 * ext4_get_stripe_size: Get the stripe size.
2058 * @sbi: In memory super block info
2059 *
2060 * If we have specified it via mount option, then
2061 * use the mount option value. If the value specified at mount time is
2062 * greater than the blocks per group use the super block value.
2063 * If the super block value is greater than blocks per group return 0.
2064 * Allocator needs it be less than blocks per group.
2065 *
2066 */
2067static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2068{
2069 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2070 unsigned long stripe_width =
2071 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2072
2073 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2074 return sbi->s_stripe;
2075
2076 if (stripe_width <= sbi->s_blocks_per_group)
2077 return stripe_width;
2078
2079 if (stride <= sbi->s_blocks_per_group)
2080 return stride;
2081
2082 return 0;
2083}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002084
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002085/* sysfs supprt */
2086
2087struct ext4_attr {
2088 struct attribute attr;
2089 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2090 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2091 const char *, size_t);
2092 int offset;
2093};
2094
2095static int parse_strtoul(const char *buf,
2096 unsigned long max, unsigned long *value)
2097{
2098 char *endp;
2099
2100 while (*buf && isspace(*buf))
2101 buf++;
2102 *value = simple_strtoul(buf, &endp, 0);
2103 while (*endp && isspace(*endp))
2104 endp++;
2105 if (*endp || *value > max)
2106 return -EINVAL;
2107
2108 return 0;
2109}
2110
2111static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2112 struct ext4_sb_info *sbi,
2113 char *buf)
2114{
2115 return snprintf(buf, PAGE_SIZE, "%llu\n",
2116 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2117}
2118
2119static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2120 struct ext4_sb_info *sbi, char *buf)
2121{
2122 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2123
2124 return snprintf(buf, PAGE_SIZE, "%lu\n",
2125 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2126 sbi->s_sectors_written_start) >> 1);
2127}
2128
2129static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2130 struct ext4_sb_info *sbi, char *buf)
2131{
2132 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2133
2134 return snprintf(buf, PAGE_SIZE, "%llu\n",
2135 sbi->s_kbytes_written +
2136 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2137 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2138}
2139
2140static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2141 struct ext4_sb_info *sbi,
2142 const char *buf, size_t count)
2143{
2144 unsigned long t;
2145
2146 if (parse_strtoul(buf, 0x40000000, &t))
2147 return -EINVAL;
2148
Theodore Ts'of7c43952009-04-24 23:31:59 -04002149 if (!is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002150 return -EINVAL;
2151
2152 sbi->s_inode_readahead_blks = t;
2153 return count;
2154}
2155
2156static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002157 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002158{
2159 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2160
2161 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2162}
2163
2164static ssize_t sbi_ui_store(struct ext4_attr *a,
2165 struct ext4_sb_info *sbi,
2166 const char *buf, size_t count)
2167{
2168 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2169 unsigned long t;
2170
2171 if (parse_strtoul(buf, 0xffffffff, &t))
2172 return -EINVAL;
2173 *ui = t;
2174 return count;
2175}
2176
2177#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2178static struct ext4_attr ext4_attr_##_name = { \
2179 .attr = {.name = __stringify(_name), .mode = _mode }, \
2180 .show = _show, \
2181 .store = _store, \
2182 .offset = offsetof(struct ext4_sb_info, _elname), \
2183}
2184#define EXT4_ATTR(name, mode, show, store) \
2185static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2186
2187#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2188#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2189#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2190 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2191#define ATTR_LIST(name) &ext4_attr_##name.attr
2192
2193EXT4_RO_ATTR(delayed_allocation_blocks);
2194EXT4_RO_ATTR(session_write_kbytes);
2195EXT4_RO_ATTR(lifetime_write_kbytes);
2196EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2197 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002198EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002199EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2200EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2201EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2202EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2203EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2204EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002205EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002206
2207static struct attribute *ext4_attrs[] = {
2208 ATTR_LIST(delayed_allocation_blocks),
2209 ATTR_LIST(session_write_kbytes),
2210 ATTR_LIST(lifetime_write_kbytes),
2211 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002212 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002213 ATTR_LIST(mb_stats),
2214 ATTR_LIST(mb_max_to_scan),
2215 ATTR_LIST(mb_min_to_scan),
2216 ATTR_LIST(mb_order2_req),
2217 ATTR_LIST(mb_stream_req),
2218 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002219 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002220 NULL,
2221};
2222
2223static ssize_t ext4_attr_show(struct kobject *kobj,
2224 struct attribute *attr, char *buf)
2225{
2226 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2227 s_kobj);
2228 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2229
2230 return a->show ? a->show(a, sbi, buf) : 0;
2231}
2232
2233static ssize_t ext4_attr_store(struct kobject *kobj,
2234 struct attribute *attr,
2235 const char *buf, size_t len)
2236{
2237 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2238 s_kobj);
2239 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2240
2241 return a->store ? a->store(a, sbi, buf, len) : 0;
2242}
2243
2244static void ext4_sb_release(struct kobject *kobj)
2245{
2246 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2247 s_kobj);
2248 complete(&sbi->s_kobj_unregister);
2249}
2250
2251
2252static struct sysfs_ops ext4_attr_ops = {
2253 .show = ext4_attr_show,
2254 .store = ext4_attr_store,
2255};
2256
2257static struct kobj_type ext4_ktype = {
2258 .default_attrs = ext4_attrs,
2259 .sysfs_ops = &ext4_attr_ops,
2260 .release = ext4_sb_release,
2261};
2262
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002263/*
2264 * Check whether this filesystem can be mounted based on
2265 * the features present and the RDONLY/RDWR mount requested.
2266 * Returns 1 if this filesystem can be mounted as requested,
2267 * 0 if it cannot be.
2268 */
2269static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2270{
2271 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2272 ext4_msg(sb, KERN_ERR,
2273 "Couldn't mount because of "
2274 "unsupported optional features (%x)",
2275 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2276 ~EXT4_FEATURE_INCOMPAT_SUPP));
2277 return 0;
2278 }
2279
2280 if (readonly)
2281 return 1;
2282
2283 /* Check that feature set is OK for a read-write mount */
2284 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2285 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2286 "unsupported optional features (%x)",
2287 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2288 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2289 return 0;
2290 }
2291 /*
2292 * Large file size enabled file system can only be mounted
2293 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2294 */
2295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2296 if (sizeof(blkcnt_t) < sizeof(u64)) {
2297 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2298 "cannot be mounted RDWR without "
2299 "CONFIG_LBDAF");
2300 return 0;
2301 }
2302 }
2303 return 1;
2304}
2305
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002306static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04002307 __releases(kernel_lock)
2308 __acquires(kernel_lock)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002309{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002310 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07002311 struct ext4_super_block *es = NULL;
2312 struct ext4_sb_info *sbi;
2313 ext4_fsblk_t block;
2314 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002315 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002316 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002317 unsigned long journal_devnum = 0;
2318 unsigned long def_mount_opts;
2319 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002320 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05002321 const char *descr;
David Howells1d1fe1e2008-02-07 00:15:37 -08002322 int ret = -EINVAL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002323 int blocksize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002324 unsigned int db_count;
2325 unsigned int i;
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002326 int needs_recovery, has_huge_files;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002327 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07002328 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002329 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002330
2331 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2332 if (!sbi)
2333 return -ENOMEM;
Pekka Enberg705895b2009-02-15 18:07:52 -05002334
2335 sbi->s_blockgroup_lock =
2336 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2337 if (!sbi->s_blockgroup_lock) {
2338 kfree(sbi);
2339 return -ENOMEM;
2340 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002341 sb->s_fs_info = sbi;
2342 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002343 sbi->s_resuid = EXT4_DEF_RESUID;
2344 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04002345 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07002346 sbi->s_sb_block = sb_block;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05002347 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2348 sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002349
2350 unlock_kernel();
2351
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002352 /* Cleanup superblock name */
2353 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2354 *cp = '!';
2355
Mingming Cao617ba132006-10-11 01:20:53 -07002356 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002357 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002358 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002359 goto out_fail;
2360 }
2361
2362 /*
Mingming Cao617ba132006-10-11 01:20:53 -07002363 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002364 * block sizes. We need to calculate the offset from buffer start.
2365 */
Mingming Cao617ba132006-10-11 01:20:53 -07002366 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002367 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2368 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002369 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002370 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002371 }
2372
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002373 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002374 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002375 goto out_fail;
2376 }
2377 /*
2378 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07002379 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002380 */
Mingming Cao617ba132006-10-11 01:20:53 -07002381 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002382 sbi->s_es = es;
2383 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07002384 if (sb->s_magic != EXT4_SUPER_MAGIC)
2385 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05002386 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002387
2388 /* Set defaults before we parse the mount options */
2389 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Mingming Cao617ba132006-10-11 01:20:53 -07002390 if (def_mount_opts & EXT4_DEFM_DEBUG)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002391 set_opt(sbi->s_mount_opt, DEBUG);
Mingming Cao617ba132006-10-11 01:20:53 -07002392 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002393 set_opt(sbi->s_mount_opt, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07002394 if (def_mount_opts & EXT4_DEFM_UID16)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002395 set_opt(sbi->s_mount_opt, NO_UID32);
Theodore Ts'o03010a32008-10-10 20:02:48 -04002396#ifdef CONFIG_EXT4_FS_XATTR
Mingming Cao617ba132006-10-11 01:20:53 -07002397 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002398 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08002399#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04002400#ifdef CONFIG_EXT4_FS_POSIX_ACL
Mingming Cao617ba132006-10-11 01:20:53 -07002401 if (def_mount_opts & EXT4_DEFM_ACL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002402 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08002403#endif
Mingming Cao617ba132006-10-11 01:20:53 -07002404 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2405 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2406 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2407 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2408 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2409 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002410
Mingming Cao617ba132006-10-11 01:20:53 -07002411 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002412 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05002413 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Dmitry Mishinceea16b2006-10-11 01:21:21 -07002414 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05002415 else
2416 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002417
2418 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2419 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05002420 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2421 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2422 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002423
Eric Sandeen571640c2008-05-26 12:29:46 -04002424 set_opt(sbi->s_mount_opt, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002425
Mingming Cao1e2462f2007-07-18 09:00:55 -04002426 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04002427 * enable delayed allocation by default
2428 * Use -o nodelalloc to turn it off
2429 */
2430 set_opt(sbi->s_mount_opt, DELALLOC);
2431
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002432 if (!parse_options((char *) data, sb, &journal_devnum,
2433 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002434 goto failed_mount;
2435
2436 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07002437 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002438
Mingming Cao617ba132006-10-11 01:20:53 -07002439 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2440 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2441 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2442 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002443 ext4_msg(sb, KERN_WARNING,
2444 "feature flags set on rev 0 fs, "
2445 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05002446
2447 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002448 * Check feature flags regardless of the revision level, since we
2449 * previously didn't change the revision level when setting the flags,
2450 * so there is a chance incompat flags are set on a rev 0 filesystem.
2451 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002452 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002453 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002454
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002455 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2456
Mingming Cao617ba132006-10-11 01:20:53 -07002457 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2458 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002459 ext4_msg(sb, KERN_ERR,
2460 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002461 goto failed_mount;
2462 }
2463
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002464 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002465 /* Validate the filesystem blocksize */
2466 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002467 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002468 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002469 goto failed_mount;
2470 }
2471
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002472 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2474 offset = do_div(logical_sb_block, blocksize);
2475 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002476 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002477 ext4_msg(sb, KERN_ERR,
2478 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002479 goto failed_mount;
2480 }
Mingming Cao617ba132006-10-11 01:20:53 -07002481 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002482 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07002483 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002484 ext4_msg(sb, KERN_ERR,
2485 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002486 goto failed_mount;
2487 }
2488 }
2489
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002490 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2491 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002492 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2493 has_huge_files);
2494 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002495
Mingming Cao617ba132006-10-11 01:20:53 -07002496 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2497 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2498 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002499 } else {
2500 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2501 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07002502 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04002503 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002504 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002505 ext4_msg(sb, KERN_ERR,
2506 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002507 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002508 goto failed_mount;
2509 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04002510 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2511 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002512 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002513
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002514 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2515 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002516 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002517 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07002518 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002519 ext4_msg(sb, KERN_ERR,
2520 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002521 sbi->s_desc_size);
2522 goto failed_mount;
2523 }
2524 } else
2525 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002526
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002527 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002528 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08002529 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002530 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002531
Mingming Cao617ba132006-10-11 01:20:53 -07002532 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002533 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002534 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002535 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2536 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002537 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002538 sbi->s_sbh = bh;
2539 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07002540 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2541 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002542
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002543 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002544 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2545 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04002546 i = le32_to_cpu(es->s_flags);
2547 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2548 sbi->s_hash_unsigned = 3;
2549 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2550#ifdef __CHAR_UNSIGNED__
2551 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2552 sbi->s_hash_unsigned = 3;
2553#else
2554 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2555#endif
2556 sb->s_dirt = 1;
2557 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002558
2559 if (sbi->s_blocks_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002560 ext4_msg(sb, KERN_ERR,
2561 "#blocks per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002562 sbi->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002563 goto failed_mount;
2564 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002565 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002566 ext4_msg(sb, KERN_ERR,
2567 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002568 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002569 goto failed_mount;
2570 }
2571
Eric Sandeenbf43d842009-08-17 23:48:51 -04002572 /*
2573 * Test whether we have more sectors than will fit in sector_t,
2574 * and whether the max offset is addressable by the page cache.
2575 */
2576 if ((ext4_blocks_count(es) >
2577 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2578 (ext4_blocks_count(es) >
2579 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002580 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04002581 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002582 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002583 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Eric Sandeenbf43d842009-08-17 23:48:51 -04002584 ret = -EFBIG;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002585 goto failed_mount;
2586 }
2587
Mingming Cao617ba132006-10-11 01:20:53 -07002588 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2589 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05002590
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04002591 /* check blocks count against device size */
2592 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2593 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002594 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2595 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04002596 ext4_blocks_count(es), blocks_count);
2597 goto failed_mount;
2598 }
2599
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002600 /*
2601 * It makes no sense for the first data block to be beyond the end
2602 * of the filesystem.
2603 */
2604 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002605 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2606 "block %u is beyond end of filesystem (%llu)",
2607 le32_to_cpu(es->s_first_data_block),
2608 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05002609 goto failed_mount;
2610 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002611 blocks_count = (ext4_blocks_count(es) -
2612 le32_to_cpu(es->s_first_data_block) +
2613 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2614 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002615 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002616 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002617 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002618 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002619 ext4_blocks_count(es),
2620 le32_to_cpu(es->s_first_data_block),
2621 EXT4_BLOCKS_PER_GROUP(sb));
2622 goto failed_mount;
2623 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002624 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04002625 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2626 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07002627 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2628 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002629 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002630 GFP_KERNEL);
2631 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002632 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002633 goto failed_mount;
2634 }
2635
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04002636#ifdef CONFIG_PROC_FS
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002637 if (ext4_proc_root)
2638 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04002639#endif
Theodore Ts'o240799c2008-10-09 23:53:47 -04002640
Pekka Enberg705895b2009-02-15 18:07:52 -05002641 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002642
2643 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002644 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002645 sbi->s_group_desc[i] = sb_bread(sb, block);
2646 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002647 ext4_msg(sb, KERN_ERR,
2648 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002649 db_count = i;
2650 goto failed_mount2;
2651 }
2652 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002653 if (!ext4_check_descriptors(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002654 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002655 goto failed_mount2;
2656 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002657 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2658 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002659 ext4_msg(sb, KERN_ERR,
2660 "unable to initialize "
2661 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002662 goto failed_mount2;
2663 }
2664
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002665 sbi->s_gdb_count = db_count;
2666 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2667 spin_lock_init(&sbi->s_next_gen_lock);
2668
Peter Zijlstra833f4072007-10-16 23:25:45 -07002669 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2670 ext4_count_free_blocks(sb));
2671 if (!err) {
2672 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2673 ext4_count_free_inodes(sb));
2674 }
2675 if (!err) {
2676 err = percpu_counter_init(&sbi->s_dirs_counter,
2677 ext4_count_dirs(sb));
2678 }
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04002679 if (!err) {
2680 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2681 }
Peter Zijlstra833f4072007-10-16 23:25:45 -07002682 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002683 ext4_msg(sb, KERN_ERR, "insufficient memory");
Peter Zijlstra833f4072007-10-16 23:25:45 -07002684 goto failed_mount3;
2685 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002686
Alex Tomasc9de5602008-01-29 00:19:52 -05002687 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002688 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05002689
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002690 /*
2691 * set up enough so that it can read an inode
2692 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04002693 if (!test_opt(sb, NOLOAD) &&
2694 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2695 sb->s_op = &ext4_sops;
2696 else
2697 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07002698 sb->s_export_op = &ext4_export_ops;
2699 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002700#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07002701 sb->s_qcop = &ext4_qctl_operations;
2702 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002703#endif
2704 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04002705 mutex_init(&sbi->s_orphan_lock);
Theodore Ts'o32ed5052009-04-25 22:53:39 -04002706 mutex_init(&sbi->s_resize_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002707
2708 sb->s_root = NULL;
2709
2710 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07002711 EXT4_HAS_INCOMPAT_FEATURE(sb,
2712 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002713
2714 /*
2715 * The first inode we look at is the journal inode. Don't try
2716 * root first: it may be modified in the journal!
2717 */
2718 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07002719 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2720 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002721 goto failed_mount3;
Theodore Ts'o624080e2008-06-06 17:50:40 -04002722 if (!(sb->s_flags & MS_RDONLY) &&
2723 EXT4_SB(sb)->s_journal->j_failed_commit) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002724 ext4_msg(sb, KERN_CRIT, "error: "
Theodore Ts'o624080e2008-06-06 17:50:40 -04002725 "ext4_fill_super: Journal transaction "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002726 "%u is corrupt",
Theodore Ts'o624080e2008-06-06 17:50:40 -04002727 EXT4_SB(sb)->s_journal->j_failed_commit);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002728 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002729 ext4_msg(sb, KERN_CRIT,
2730 "Mounting filesystem read-only");
Theodore Ts'o624080e2008-06-06 17:50:40 -04002731 sb->s_flags |= MS_RDONLY;
2732 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2733 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2734 }
2735 if (test_opt(sb, ERRORS_PANIC)) {
2736 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2737 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04002738 ext4_commit_super(sb, 1);
Theodore Ts'o624080e2008-06-06 17:50:40 -04002739 goto failed_mount4;
2740 }
2741 }
Frank Mayhar03901312009-01-07 00:06:22 -05002742 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2743 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002744 ext4_msg(sb, KERN_ERR, "required journal recovery "
2745 "suppressed and not mounted read-only");
Frank Mayhar03901312009-01-07 00:06:22 -05002746 goto failed_mount4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002747 } else {
Frank Mayhar03901312009-01-07 00:06:22 -05002748 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2749 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2750 sbi->s_journal = NULL;
2751 needs_recovery = 0;
2752 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002753 }
2754
Jose R. Santoseb40a092007-07-18 08:37:25 -04002755 if (ext4_blocks_count(es) > 0xffffffffULL &&
2756 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2757 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002758 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jose R. Santoseb40a092007-07-18 08:37:25 -04002759 goto failed_mount4;
2760 }
2761
Theodore Ts'od0646f72009-09-05 12:50:43 -04002762 jbd2_journal_set_features(sbi->s_journal,
2763 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2764 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2765 jbd2_journal_set_features(sbi->s_journal, 0, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05002766 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Theodore Ts'od0646f72009-09-05 12:50:43 -04002767 else
Girish Shilamkar818d2762008-01-28 23:58:27 -05002768 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2769 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Girish Shilamkar818d2762008-01-28 23:58:27 -05002770
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002771 /* We have now updated the journal if required, so we can
2772 * validate the data journaling mode. */
2773 switch (test_opt(sb, DATA_FLAGS)) {
2774 case 0:
2775 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07002776 * capabilities: ORDERED_DATA if the journal can
2777 * cope, else JOURNAL_DATA
2778 */
Mingming Caodab291a2006-10-11 01:21:01 -07002779 if (jbd2_journal_check_available_features
2780 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002781 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2782 else
2783 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2784 break;
2785
Mingming Cao617ba132006-10-11 01:20:53 -07002786 case EXT4_MOUNT_ORDERED_DATA:
2787 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07002788 if (!jbd2_journal_check_available_features
2789 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002790 ext4_msg(sb, KERN_ERR, "Journal does not support "
2791 "requested data journaling mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002792 goto failed_mount4;
2793 }
2794 default:
2795 break;
2796 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002797 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002798
Frank Mayhar03901312009-01-07 00:06:22 -05002799no_journal:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002800
2801 if (test_opt(sb, NOBH)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002802 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002803 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2804 "its supported only with writeback mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002805 clear_opt(sbi->s_mount_opt, NOBH);
2806 }
2807 }
Mingming Cao4c0425f2009-09-28 15:48:41 -04002808 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2809 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2810 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2811 goto failed_mount_wq;
2812 }
2813
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002814 /*
Mingming Caodab291a2006-10-11 01:21:01 -07002815 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002816 * so we can safely mount the rest of the filesystem now.
2817 */
2818
David Howells1d1fe1e2008-02-07 00:15:37 -08002819 root = ext4_iget(sb, EXT4_ROOT_INO);
2820 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002821 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08002822 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002823 goto failed_mount4;
2824 }
2825 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08002826 iput(root);
Eric Sandeenb31e1552009-06-04 17:36:36 -04002827 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002828 goto failed_mount4;
2829 }
David Howells1d1fe1e2008-02-07 00:15:37 -08002830 sb->s_root = d_alloc_root(root);
2831 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002832 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08002833 iput(root);
2834 ret = -ENOMEM;
2835 goto failed_mount4;
2836 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002837
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002838 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04002839
2840 /* determine the minimum size of new large inodes, if present */
2841 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2842 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2843 EXT4_GOOD_OLD_INODE_SIZE;
2844 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2845 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2846 if (sbi->s_want_extra_isize <
2847 le16_to_cpu(es->s_want_extra_isize))
2848 sbi->s_want_extra_isize =
2849 le16_to_cpu(es->s_want_extra_isize);
2850 if (sbi->s_want_extra_isize <
2851 le16_to_cpu(es->s_min_extra_isize))
2852 sbi->s_want_extra_isize =
2853 le16_to_cpu(es->s_min_extra_isize);
2854 }
2855 }
2856 /* Check if enough inode space is available */
2857 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2858 sbi->s_inode_size) {
2859 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2860 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04002861 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2862 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04002863 }
2864
Theodore Ts'o90576c02009-09-29 15:51:30 -04002865 if (test_opt(sb, DELALLOC) &&
2866 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002867 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2868 "requested data journaling mode");
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002869 clear_opt(sbi->s_mount_opt, DELALLOC);
Theodore Ts'o90576c02009-09-29 15:51:30 -04002870 }
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002871
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04002872 err = ext4_setup_system_zone(sb);
2873 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002874 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2875 "zone (%d)\n", err);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04002876 goto failed_mount4;
2877 }
2878
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002879 ext4_ext_init(sb);
2880 err = ext4_mb_init(sb, needs_recovery);
2881 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002882 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2883 err);
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002884 goto failed_mount4;
2885 }
2886
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002887 sbi->s_kobj.kset = ext4_kset;
2888 init_completion(&sbi->s_kobj_unregister);
2889 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2890 "%s", sb->s_id);
2891 if (err) {
2892 ext4_mb_release(sb);
2893 ext4_ext_release(sb);
2894 goto failed_mount4;
2895 };
2896
Mingming Cao617ba132006-10-11 01:20:53 -07002897 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2898 ext4_orphan_cleanup(sb, es);
2899 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05002900 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002901 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05002902 ext4_mark_recovery_complete(sb, es);
2903 }
2904 if (EXT4_SB(sb)->s_journal) {
2905 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2906 descr = " journalled data mode";
2907 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2908 descr = " ordered data mode";
2909 else
2910 descr = " writeback data mode";
2911 } else
2912 descr = "out journal";
2913
Eric Sandeenb31e1552009-06-04 17:36:36 -04002914 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002915
2916 lock_kernel();
2917 return 0;
2918
Mingming Cao617ba132006-10-11 01:20:53 -07002919cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002920 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002921 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002922 goto failed_mount;
2923
2924failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04002925 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04002926 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2927failed_mount_wq:
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04002928 ext4_release_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05002929 if (sbi->s_journal) {
2930 jbd2_journal_destroy(sbi->s_journal);
2931 sbi->s_journal = NULL;
2932 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002933failed_mount3:
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04002934 if (sbi->s_flex_groups) {
2935 if (is_vmalloc_addr(sbi->s_flex_groups))
2936 vfree(sbi->s_flex_groups);
2937 else
2938 kfree(sbi->s_flex_groups);
2939 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002940 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2941 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2942 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04002943 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002944failed_mount2:
2945 for (i = 0; i < db_count; i++)
2946 brelse(sbi->s_group_desc[i]);
2947 kfree(sbi->s_group_desc);
2948failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04002949 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002950 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04002951 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002952#ifdef CONFIG_QUOTA
2953 for (i = 0; i < MAXQUOTAS; i++)
2954 kfree(sbi->s_qf_names[i]);
2955#endif
Mingming Cao617ba132006-10-11 01:20:53 -07002956 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002957 brelse(bh);
2958out_fail:
2959 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04002960 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002961 kfree(sbi);
2962 lock_kernel();
David Howells1d1fe1e2008-02-07 00:15:37 -08002963 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002964}
2965
2966/*
2967 * Setup any per-fs journal parameters now. We'll do this both on
2968 * initial mount, once the journal has been initialised but before we've
2969 * done any recovery; and again on any subsequent remount.
2970 */
Mingming Cao617ba132006-10-11 01:20:53 -07002971static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002972{
Mingming Cao617ba132006-10-11 01:20:53 -07002973 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002974
Theodore Ts'o30773842009-01-03 20:27:38 -05002975 journal->j_commit_interval = sbi->s_commit_interval;
2976 journal->j_min_batch_time = sbi->s_min_batch_time;
2977 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002978
2979 spin_lock(&journal->j_state_lock);
2980 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07002981 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002982 else
Mingming Caodab291a2006-10-11 01:21:01 -07002983 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04002984 if (test_opt(sb, DATA_ERR_ABORT))
2985 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2986 else
2987 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002988 spin_unlock(&journal->j_state_lock);
2989}
2990
Mingming Cao617ba132006-10-11 01:20:53 -07002991static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002992 unsigned int journal_inum)
2993{
2994 struct inode *journal_inode;
2995 journal_t *journal;
2996
Frank Mayhar03901312009-01-07 00:06:22 -05002997 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2998
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002999 /* First, test for the existence of a valid inode on disk. Bad
3000 * things happen if we iget() an unused inode, as the subsequent
3001 * iput() will try to delete it. */
3002
David Howells1d1fe1e2008-02-07 00:15:37 -08003003 journal_inode = ext4_iget(sb, journal_inum);
3004 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003005 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003006 return NULL;
3007 }
3008 if (!journal_inode->i_nlink) {
3009 make_bad_inode(journal_inode);
3010 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003011 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003012 return NULL;
3013 }
3014
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003015 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003016 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003017 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003018 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003019 iput(journal_inode);
3020 return NULL;
3021 }
3022
Mingming Caodab291a2006-10-11 01:21:01 -07003023 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003024 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003025 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003026 iput(journal_inode);
3027 return NULL;
3028 }
3029 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003030 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003031 return journal;
3032}
3033
Mingming Cao617ba132006-10-11 01:20:53 -07003034static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003035 dev_t j_dev)
3036{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003037 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003038 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003039 ext4_fsblk_t start;
3040 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003041 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003042 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003043 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003044 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003045 struct block_device *bdev;
3046
Frank Mayhar03901312009-01-07 00:06:22 -05003047 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3048
Eric Sandeenb31e1552009-06-04 17:36:36 -04003049 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003050 if (bdev == NULL)
3051 return NULL;
3052
3053 if (bd_claim(bdev, sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003054 ext4_msg(sb, KERN_ERR,
3055 "failed to claim external journal device");
Al Viro9a1c3542008-02-22 20:40:24 -05003056 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003057 return NULL;
3058 }
3059
3060 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003061 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003062 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003063 ext4_msg(sb, KERN_ERR,
3064 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003065 goto out_bdev;
3066 }
3067
Mingming Cao617ba132006-10-11 01:20:53 -07003068 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3069 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003070 set_blocksize(bdev, blocksize);
3071 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003072 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3073 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003074 goto out_bdev;
3075 }
3076
Mingming Cao617ba132006-10-11 01:20:53 -07003077 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3078 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003079 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003080 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003081 ext4_msg(sb, KERN_ERR, "external journal has "
3082 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003083 brelse(bh);
3084 goto out_bdev;
3085 }
3086
Mingming Cao617ba132006-10-11 01:20:53 -07003087 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003088 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003089 brelse(bh);
3090 goto out_bdev;
3091 }
3092
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003093 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003094 start = sb_block + 1;
3095 brelse(bh); /* we're done with the superblock */
3096
Mingming Caodab291a2006-10-11 01:21:01 -07003097 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003098 start, len, blocksize);
3099 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003100 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003101 goto out_bdev;
3102 }
3103 journal->j_private = sb;
3104 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3105 wait_on_buffer(journal->j_sb_buffer);
3106 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003107 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003108 goto out_journal;
3109 }
3110 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003111 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3112 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003113 be32_to_cpu(journal->j_superblock->s_nr_users));
3114 goto out_journal;
3115 }
Mingming Cao617ba132006-10-11 01:20:53 -07003116 EXT4_SB(sb)->journal_bdev = bdev;
3117 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003118 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003119
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003120out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07003121 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003122out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07003123 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003124 return NULL;
3125}
3126
Mingming Cao617ba132006-10-11 01:20:53 -07003127static int ext4_load_journal(struct super_block *sb,
3128 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003129 unsigned long journal_devnum)
3130{
3131 journal_t *journal;
3132 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3133 dev_t journal_dev;
3134 int err = 0;
3135 int really_read_only;
3136
Frank Mayhar03901312009-01-07 00:06:22 -05003137 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3138
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003139 if (journal_devnum &&
3140 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003141 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3142 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003143 journal_dev = new_decode_dev(journal_devnum);
3144 } else
3145 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3146
3147 really_read_only = bdev_read_only(sb->s_bdev);
3148
3149 /*
3150 * Are we loading a blank journal or performing recovery after a
3151 * crash? For recovery, we need to check in advance whether we
3152 * can get read-write access to the device.
3153 */
Mingming Cao617ba132006-10-11 01:20:53 -07003154 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003155 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003156 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3157 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003158 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003159 ext4_msg(sb, KERN_ERR, "write access "
3160 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003161 return -EROFS;
3162 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04003163 ext4_msg(sb, KERN_INFO, "write access will "
3164 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003165 }
3166 }
3167
3168 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003169 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3170 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003171 return -EINVAL;
3172 }
3173
3174 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07003175 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003176 return -EINVAL;
3177 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07003178 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003179 return -EINVAL;
3180 }
3181
Theodore Ts'o90576c02009-09-29 15:51:30 -04003182 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003183 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04003184
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003185 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07003186 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003187 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003188 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003189 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003190 return err;
3191 }
3192 }
3193
Mingming Cao617ba132006-10-11 01:20:53 -07003194 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07003195 err = jbd2_journal_wipe(journal, !really_read_only);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003196 if (!err)
Mingming Caodab291a2006-10-11 01:21:01 -07003197 err = jbd2_journal_load(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003198
3199 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003200 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003201 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003202 return err;
3203 }
3204
Mingming Cao617ba132006-10-11 01:20:53 -07003205 EXT4_SB(sb)->s_journal = journal;
3206 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003207
3208 if (journal_devnum &&
3209 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3210 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003211
3212 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003213 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003214 }
3215
3216 return 0;
3217}
3218
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003219static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003220{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003221 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07003222 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003223 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003224
3225 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08003226 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04003227 if (buffer_write_io_error(sbh)) {
3228 /*
3229 * Oh, dear. A previous attempt to write the
3230 * superblock failed. This could happen because the
3231 * USB device was yanked out. Or it could happen to
3232 * be a transient write error and maybe the block will
3233 * be remapped. Nothing we can do but to retry the
3234 * write and hope for the best.
3235 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04003236 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3237 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003238 clear_buffer_write_io_error(sbh);
3239 set_buffer_uptodate(sbh);
3240 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04003241 /*
3242 * If the file system is mounted read-only, don't update the
3243 * superblock write time. This avoids updating the superblock
3244 * write time when we are mounting the root file system
3245 * read/only but we need to replay the journal; at that point,
3246 * for people who are east of GMT and who make their clock
3247 * tick in localtime for Windows bug-for-bug compatibility,
3248 * the clock is set in the future, and this will cause e2fsck
3249 * to complain and force a full file system check.
3250 */
3251 if (!(sb->s_flags & MS_RDONLY))
3252 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003253 es->s_kbytes_written =
3254 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3255 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3256 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -05003257 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3258 &EXT4_SB(sb)->s_freeblocks_counter));
3259 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3260 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003261 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003262 BUFFER_TRACE(sbh, "marking dirty");
3263 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04003264 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08003265 error = sync_dirty_buffer(sbh);
3266 if (error)
3267 return error;
3268
3269 error = buffer_write_io_error(sbh);
3270 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003271 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3272 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003273 clear_buffer_write_io_error(sbh);
3274 set_buffer_uptodate(sbh);
3275 }
3276 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08003277 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003278}
3279
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003280/*
3281 * Have we just finished recovery? If so, and if we are mounting (or
3282 * remounting) the filesystem readonly, then we will end up with a
3283 * consistent fs on disk. Record that fact.
3284 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003285static void ext4_mark_recovery_complete(struct super_block *sb,
3286 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003287{
Mingming Cao617ba132006-10-11 01:20:53 -07003288 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003289
Frank Mayhar03901312009-01-07 00:06:22 -05003290 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3291 BUG_ON(journal != NULL);
3292 return;
3293 }
Mingming Caodab291a2006-10-11 01:21:01 -07003294 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003295 if (jbd2_journal_flush(journal) < 0)
3296 goto out;
3297
Mingming Cao617ba132006-10-11 01:20:53 -07003298 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003299 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07003300 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003301 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003302 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003303
3304out:
Mingming Caodab291a2006-10-11 01:21:01 -07003305 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003306}
3307
3308/*
3309 * If we are mounting (or read-write remounting) a filesystem whose journal
3310 * has recorded an error from a previous lifetime, move that error to the
3311 * main filesystem now.
3312 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003313static void ext4_clear_journal_err(struct super_block *sb,
3314 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003315{
3316 journal_t *journal;
3317 int j_errno;
3318 const char *errstr;
3319
Frank Mayhar03901312009-01-07 00:06:22 -05003320 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3321
Mingming Cao617ba132006-10-11 01:20:53 -07003322 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003323
3324 /*
3325 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07003326 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003327 */
3328
Mingming Caodab291a2006-10-11 01:21:01 -07003329 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003330 if (j_errno) {
3331 char nbuf[16];
3332
Mingming Cao617ba132006-10-11 01:20:53 -07003333 errstr = ext4_decode_error(sb, j_errno, nbuf);
Harvey Harrison46e665e2008-04-17 10:38:59 -04003334 ext4_warning(sb, __func__, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003335 "from previous mount: %s", errstr);
Harvey Harrison46e665e2008-04-17 10:38:59 -04003336 ext4_warning(sb, __func__, "Marking fs in need of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003337 "filesystem check.");
3338
Mingming Cao617ba132006-10-11 01:20:53 -07003339 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3340 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003341 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003342
Mingming Caodab291a2006-10-11 01:21:01 -07003343 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003344 }
3345}
3346
3347/*
3348 * Force the running and committing transactions to commit,
3349 * and wait on the commit.
3350 */
Mingming Cao617ba132006-10-11 01:20:53 -07003351int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003352{
3353 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05003354 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003355
3356 if (sb->s_flags & MS_RDONLY)
3357 return 0;
3358
Mingming Cao617ba132006-10-11 01:20:53 -07003359 journal = EXT4_SB(sb)->s_journal;
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003360 if (journal)
Frank Mayhar03901312009-01-07 00:06:22 -05003361 ret = ext4_journal_force_commit(journal);
Frank Mayhar03901312009-01-07 00:06:22 -05003362
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003363 return ret;
3364}
3365
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003366static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003367{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02003368 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003369 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02003370 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003371}
3372
Mingming Cao617ba132006-10-11 01:20:53 -07003373static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003374{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05003375 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05003376 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003377 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003378
Theodore Ts'o9bffad12009-06-17 11:48:11 -04003379 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003380 flush_workqueue(sbi->dio_unwritten_wq);
3381 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003382 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003383 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05003384 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05003385 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003386}
3387
3388/*
3389 * LVM calls this function before a (read-only) snapshot is created. This
3390 * gives us a chance to flush the journal completely and mark the fs clean.
3391 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08003392static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003393{
Takashi Satoc4be0c12009-01-09 16:40:58 -08003394 int error = 0;
3395 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003396
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003397 if (sb->s_flags & MS_RDONLY)
3398 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003399
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003400 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003401
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003402 /* Now we set up the journal barrier. */
3403 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003404
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003405 /*
3406 * Don't clear the needs_recovery flag if we failed to flush
3407 * the journal.
3408 */
3409 error = jbd2_journal_flush(journal);
3410 if (error < 0) {
3411 out:
3412 jbd2_journal_unlock_updates(journal);
3413 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003414 }
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003415
3416 /* Journal blocked and flushed, clear needs_recovery flag. */
3417 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3418 error = ext4_commit_super(sb, 1);
3419 if (error)
3420 goto out;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003421 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003422}
3423
3424/*
3425 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3426 * flag here, even though the filesystem is not technically dirty yet.
3427 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08003428static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003429{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003430 if (sb->s_flags & MS_RDONLY)
3431 return 0;
3432
3433 lock_super(sb);
3434 /* Reset the needs_recovery flag before the fs is unlocked. */
3435 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3436 ext4_commit_super(sb, 1);
3437 unlock_super(sb);
3438 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Takashi Satoc4be0c12009-01-09 16:40:58 -08003439 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003440}
3441
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003442static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003443{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003444 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07003445 struct ext4_sb_info *sbi = EXT4_SB(sb);
3446 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003447 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07003448 struct ext4_mount_options old_opts;
Theodore Ts'o8a266462008-07-26 14:34:21 -04003449 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003450 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003451 int err;
3452#ifdef CONFIG_QUOTA
3453 int i;
3454#endif
3455
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003456 lock_kernel();
3457
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003458 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04003459 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003460 old_sb_flags = sb->s_flags;
3461 old_opts.s_mount_opt = sbi->s_mount_opt;
3462 old_opts.s_resuid = sbi->s_resuid;
3463 old_opts.s_resgid = sbi->s_resgid;
3464 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05003465 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3466 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003467#ifdef CONFIG_QUOTA
3468 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3469 for (i = 0; i < MAXQUOTAS; i++)
3470 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3471#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003472 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3473 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003474
3475 /*
3476 * Allow the "check" option to be passed as a remount option.
3477 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003478 if (!parse_options(data, sb, NULL, &journal_ioprio,
3479 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003480 err = -EINVAL;
3481 goto restore_opts;
3482 }
3483
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04003484 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Harvey Harrison46e665e2008-04-17 10:38:59 -04003485 ext4_abort(sb, __func__, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003486
3487 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07003488 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003489
3490 es = sbi->s_es;
3491
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003492 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05003493 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003494 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3495 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003496
3497 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003498 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04003499 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003500 err = -EROFS;
3501 goto restore_opts;
3502 }
3503
3504 if (*flags & MS_RDONLY) {
3505 /*
3506 * First of all, the unconditional stuff we have to do
3507 * to disable replay of the journal when we next remount
3508 */
3509 sb->s_flags |= MS_RDONLY;
3510
3511 /*
3512 * OK, test if we are remounting a valid rw partition
3513 * readonly, and if so set the rdonly flag and then
3514 * mark the partition as valid again.
3515 */
Mingming Cao617ba132006-10-11 01:20:53 -07003516 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3517 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003518 es->s_state = cpu_to_le16(sbi->s_mount_state);
3519
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04003520 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05003521 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003522 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003523 /* Make sure we can mount this feature set readwrite */
3524 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003525 err = -EROFS;
3526 goto restore_opts;
3527 }
Eric Sandeenead65962007-02-10 01:46:08 -08003528 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04003529 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003530 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04003531 */
3532 for (g = 0; g < sbi->s_groups_count; g++) {
3533 struct ext4_group_desc *gdp =
3534 ext4_get_group_desc(sb, g, NULL);
3535
3536 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003537 ext4_msg(sb, KERN_ERR,
3538 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04003539 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3540 le16_to_cpu(gdp->bg_checksum));
3541 err = -EINVAL;
3542 goto restore_opts;
3543 }
3544 }
3545
3546 /*
Eric Sandeenead65962007-02-10 01:46:08 -08003547 * If we have an unprocessed orphan list hanging
3548 * around from a previously readonly bdev mount,
3549 * require a full umount/remount for now.
3550 */
3551 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003552 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08003553 "remount RDWR because of unprocessed "
3554 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003555 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08003556 err = -EINVAL;
3557 goto restore_opts;
3558 }
3559
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003560 /*
3561 * Mounting a RDONLY partition read-write, so reread
3562 * and store the current valid flag. (It may have
3563 * been changed by e2fsck since we originally mounted
3564 * the partition.)
3565 */
Frank Mayhar03901312009-01-07 00:06:22 -05003566 if (sbi->s_journal)
3567 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003568 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07003569 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003570 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003571 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003572 sb->s_flags &= ~MS_RDONLY;
3573 }
3574 }
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003575 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003576 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003577 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05003578
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003579#ifdef CONFIG_QUOTA
3580 /* Release old quota file names */
3581 for (i = 0; i < MAXQUOTAS; i++)
3582 if (old_opts.s_qf_names[i] &&
3583 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3584 kfree(old_opts.s_qf_names[i]);
3585#endif
Al Virobbd68512009-05-06 10:43:07 -04003586 unlock_super(sb);
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003587 unlock_kernel();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003588 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003589
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003590restore_opts:
3591 sb->s_flags = old_sb_flags;
3592 sbi->s_mount_opt = old_opts.s_mount_opt;
3593 sbi->s_resuid = old_opts.s_resuid;
3594 sbi->s_resgid = old_opts.s_resgid;
3595 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05003596 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3597 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003598#ifdef CONFIG_QUOTA
3599 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3600 for (i = 0; i < MAXQUOTAS; i++) {
3601 if (sbi->s_qf_names[i] &&
3602 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3603 kfree(sbi->s_qf_names[i]);
3604 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3605 }
3606#endif
Al Virobbd68512009-05-06 10:43:07 -04003607 unlock_super(sb);
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003608 unlock_kernel();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003609 return err;
3610}
3611
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003612static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003613{
3614 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003615 struct ext4_sb_info *sbi = EXT4_SB(sb);
3616 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08003617 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003618
Badari Pulavarty5e700302007-07-15 23:42:00 -07003619 if (test_opt(sb, MINIX_DF)) {
3620 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003621 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04003622 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07003623 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003624
3625 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07003626 * Compute the overhead (FS structures). This is constant
3627 * for a given filesystem unless the number of block groups
3628 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003629 */
3630
3631 /*
3632 * All of the blocks before first_data_block are
3633 * overhead
3634 */
3635 overhead = le32_to_cpu(es->s_first_data_block);
3636
3637 /*
3638 * Add the overhead attributed to the superblock and
3639 * block group descriptors. If the sparse superblocks
3640 * feature is turned on, then not all groups have this.
3641 */
3642 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07003643 overhead += ext4_bg_has_super(sb, i) +
3644 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003645 cond_resched();
3646 }
3647
3648 /*
3649 * Every block group has an inode bitmap, a block
3650 * bitmap, and an inode table.
3651 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07003652 overhead += ngroups * (2 + sbi->s_itb_per_group);
3653 sbi->s_overhead_last = overhead;
3654 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003655 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003656 }
3657
Mingming Cao617ba132006-10-11 01:20:53 -07003658 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003659 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07003660 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04003661 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3662 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
Aneesh Kumar K.V308ba3e2007-10-16 18:38:25 -04003663 ext4_free_blocks_count_set(es, buf->f_bfree);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003664 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3665 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003666 buf->f_bavail = 0;
3667 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07003668 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Badari Pulavarty5e700302007-07-15 23:42:00 -07003669 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
Mingming Cao617ba132006-10-11 01:20:53 -07003670 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08003671 fsid = le64_to_cpup((void *)es->s_uuid) ^
3672 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3673 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3674 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003675
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003676 return 0;
3677}
3678
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003679/* Helper function for writing quotas on sync - we need to start transaction
3680 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003681 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07003682 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01003683 * jbd2_journal_start() write_dquot()
3684 * vfs_dq_init() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07003685 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003686 *
3687 */
3688
3689#ifdef CONFIG_QUOTA
3690
3691static inline struct inode *dquot_to_inode(struct dquot *dquot)
3692{
3693 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3694}
3695
Mingming Cao617ba132006-10-11 01:20:53 -07003696static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003697{
3698 int ret, err;
3699 handle_t *handle;
3700 struct inode *inode;
3701
3702 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003703 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003704 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003705 if (IS_ERR(handle))
3706 return PTR_ERR(handle);
3707 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003708 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003709 if (!ret)
3710 ret = err;
3711 return ret;
3712}
3713
Mingming Cao617ba132006-10-11 01:20:53 -07003714static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003715{
3716 int ret, err;
3717 handle_t *handle;
3718
Mingming Cao617ba132006-10-11 01:20:53 -07003719 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003720 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003721 if (IS_ERR(handle))
3722 return PTR_ERR(handle);
3723 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003724 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003725 if (!ret)
3726 ret = err;
3727 return ret;
3728}
3729
Mingming Cao617ba132006-10-11 01:20:53 -07003730static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003731{
3732 int ret, err;
3733 handle_t *handle;
3734
Mingming Cao617ba132006-10-11 01:20:53 -07003735 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003736 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07003737 if (IS_ERR(handle)) {
3738 /* Release dquot anyway to avoid endless cycle in dqput() */
3739 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003740 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07003741 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003742 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003743 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003744 if (!ret)
3745 ret = err;
3746 return ret;
3747}
3748
Mingming Cao617ba132006-10-11 01:20:53 -07003749static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003750{
Jan Kara2c8be6b2008-05-13 21:27:55 -04003751 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07003752 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3753 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003754 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003755 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003756 } else {
3757 return dquot_mark_dquot_dirty(dquot);
3758 }
3759}
3760
Mingming Cao617ba132006-10-11 01:20:53 -07003761static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003762{
3763 int ret, err;
3764 handle_t *handle;
3765
3766 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07003767 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003768 if (IS_ERR(handle))
3769 return PTR_ERR(handle);
3770 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07003771 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003772 if (!ret)
3773 ret = err;
3774 return ret;
3775}
3776
3777/*
3778 * Turn on quotas during mount time - we need to find
3779 * the quota file and such...
3780 */
Mingming Cao617ba132006-10-11 01:20:53 -07003781static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003782{
Mingming Cao617ba132006-10-11 01:20:53 -07003783 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003784 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003785}
3786
3787/*
3788 * Standard function to be called on quota_on
3789 */
Mingming Cao617ba132006-10-11 01:20:53 -07003790static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Al Viro82646132008-08-02 00:57:06 -04003791 char *name, int remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003792{
3793 int err;
Al Viro82646132008-08-02 00:57:06 -04003794 struct path path;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003795
3796 if (!test_opt(sb, QUOTA))
3797 return -EINVAL;
Al Viro82646132008-08-02 00:57:06 -04003798 /* When remounting, no checks are needed and in fact, name is NULL */
Jan Kara06235432008-05-13 19:11:51 -04003799 if (remount)
Al Viro82646132008-08-02 00:57:06 -04003800 return vfs_quota_on(sb, type, format_id, name, remount);
Jan Kara06235432008-05-13 19:11:51 -04003801
Al Viro82646132008-08-02 00:57:06 -04003802 err = kern_path(name, LOOKUP_FOLLOW, &path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003803 if (err)
3804 return err;
Jan Kara06235432008-05-13 19:11:51 -04003805
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003806 /* Quotafile not on the same filesystem? */
Al Viro82646132008-08-02 00:57:06 -04003807 if (path.mnt->mnt_sb != sb) {
3808 path_put(&path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003809 return -EXDEV;
3810 }
Jan Kara06235432008-05-13 19:11:51 -04003811 /* Journaling quota? */
3812 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003813 /* Quotafile not in fs root? */
Al Viro82646132008-08-02 00:57:06 -04003814 if (path.dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003815 ext4_msg(sb, KERN_WARNING,
3816 "Quota file not on filesystem root. "
3817 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003818 }
Jan Kara06235432008-05-13 19:11:51 -04003819
3820 /*
3821 * When we journal data on quota file, we have to flush journal to see
3822 * all updates to the file when we bypass pagecache...
3823 */
Frank Mayhar03901312009-01-07 00:06:22 -05003824 if (EXT4_SB(sb)->s_journal &&
3825 ext4_should_journal_data(path.dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04003826 /*
3827 * We don't need to lock updates but journal_flush() could
3828 * otherwise be livelocked...
3829 */
3830 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003831 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04003832 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003833 if (err) {
Al Viro82646132008-08-02 00:57:06 -04003834 path_put(&path);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003835 return err;
3836 }
Jan Kara06235432008-05-13 19:11:51 -04003837 }
3838
Al Viro82646132008-08-02 00:57:06 -04003839 err = vfs_quota_on_path(sb, type, format_id, &path);
3840 path_put(&path);
Al Viro77e69da2008-08-01 04:29:18 -04003841 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003842}
3843
3844/* Read data from quotafile - avoid pagecache and such because we cannot afford
3845 * acquiring the locks... As quota files are never truncated and quota code
3846 * itself serializes the operations (and noone else should touch the files)
3847 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07003848static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003849 size_t len, loff_t off)
3850{
3851 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003852 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003853 int err = 0;
3854 int offset = off & (sb->s_blocksize - 1);
3855 int tocopy;
3856 size_t toread;
3857 struct buffer_head *bh;
3858 loff_t i_size = i_size_read(inode);
3859
3860 if (off > i_size)
3861 return 0;
3862 if (off+len > i_size)
3863 len = i_size-off;
3864 toread = len;
3865 while (toread > 0) {
3866 tocopy = sb->s_blocksize - offset < toread ?
3867 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07003868 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003869 if (err)
3870 return err;
3871 if (!bh) /* A hole? */
3872 memset(data, 0, tocopy);
3873 else
3874 memcpy(data, bh->b_data+offset, tocopy);
3875 brelse(bh);
3876 offset = 0;
3877 toread -= tocopy;
3878 data += tocopy;
3879 blk++;
3880 }
3881 return len;
3882}
3883
3884/* Write to quotafile (we know the transaction is already started and has
3885 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07003886static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003887 const char *data, size_t len, loff_t off)
3888{
3889 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003890 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003891 int err = 0;
3892 int offset = off & (sb->s_blocksize - 1);
3893 int tocopy;
Mingming Cao617ba132006-10-11 01:20:53 -07003894 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003895 size_t towrite = len;
3896 struct buffer_head *bh;
3897 handle_t *handle = journal_current_handle();
3898
Frank Mayhar03901312009-01-07 00:06:22 -05003899 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003900 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3901 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07003902 (unsigned long long)off, (unsigned long long)len);
3903 return -EIO;
3904 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003905 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3906 while (towrite > 0) {
3907 tocopy = sb->s_blocksize - offset < towrite ?
3908 sb->s_blocksize - offset : towrite;
Mingming Cao617ba132006-10-11 01:20:53 -07003909 bh = ext4_bread(handle, inode, blk, 1, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003910 if (!bh)
3911 goto out;
3912 if (journal_quota) {
Mingming Cao617ba132006-10-11 01:20:53 -07003913 err = ext4_journal_get_write_access(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003914 if (err) {
3915 brelse(bh);
3916 goto out;
3917 }
3918 }
3919 lock_buffer(bh);
3920 memcpy(bh->b_data+offset, data, tocopy);
3921 flush_dcache_page(bh->b_page);
3922 unlock_buffer(bh);
3923 if (journal_quota)
Frank Mayhar03901312009-01-07 00:06:22 -05003924 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003925 else {
3926 /* Always do at least ordered writes for quotas */
Jan Kara678aaf42008-07-11 19:27:31 -04003927 err = ext4_jbd2_file_inode(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003928 mark_buffer_dirty(bh);
3929 }
3930 brelse(bh);
3931 if (err)
3932 goto out;
3933 offset = 0;
3934 towrite -= tocopy;
3935 data += tocopy;
3936 blk++;
3937 }
3938out:
Jan Kara4d04e4f2008-07-04 09:59:34 -07003939 if (len == towrite) {
3940 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003941 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07003942 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003943 if (inode->i_size < off+len-towrite) {
3944 i_size_write(inode, off+len-towrite);
Mingming Cao617ba132006-10-11 01:20:53 -07003945 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003946 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003947 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07003948 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003949 mutex_unlock(&inode->i_mutex);
3950 return len - towrite;
3951}
3952
3953#endif
3954
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003955static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3956 const char *dev_name, void *data, struct vfsmount *mnt)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003957{
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003958 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003959}
3960
Theodore Ts'o03010a32008-10-10 20:02:48 -04003961static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003962 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04003963 .name = "ext4",
Mingming Cao617ba132006-10-11 01:20:53 -07003964 .get_sb = ext4_get_sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003965 .kill_sb = kill_block_super,
3966 .fs_flags = FS_REQUIRES_DEV,
3967};
3968
Theodore Ts'o03010a32008-10-10 20:02:48 -04003969#ifdef CONFIG_EXT4DEV_COMPAT
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003970static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3971 const char *dev_name, void *data,struct vfsmount *mnt)
Theodore Ts'o03010a32008-10-10 20:02:48 -04003972{
Eric Sandeenb31e1552009-06-04 17:36:36 -04003973 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3974 "to mount using ext4\n", dev_name);
3975 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3976 "will go away by 2.6.31\n", dev_name);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003977 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003978}
3979
3980static struct file_system_type ext4dev_fs_type = {
3981 .owner = THIS_MODULE,
3982 .name = "ext4dev",
3983 .get_sb = ext4dev_get_sb,
3984 .kill_sb = kill_block_super,
3985 .fs_flags = FS_REQUIRES_DEV,
3986};
3987MODULE_ALIAS("ext4dev");
3988#endif
3989
Mingming Cao617ba132006-10-11 01:20:53 -07003990static int __init init_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003991{
Alex Tomasc9de5602008-01-29 00:19:52 -05003992 int err;
3993
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003994 err = init_ext4_system_zone();
3995 if (err)
3996 return err;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003997 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3998 if (!ext4_kset)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003999 goto out4;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004000 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Alex Tomasc9de5602008-01-29 00:19:52 -05004001 err = init_ext4_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004002 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004003 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05004004
4005 err = init_ext4_xattr();
4006 if (err)
4007 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004008 err = init_inodecache();
4009 if (err)
4010 goto out1;
Theodore Ts'o03010a32008-10-10 20:02:48 -04004011 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004012 if (err)
4013 goto out;
Theodore Ts'o03010a32008-10-10 20:02:48 -04004014#ifdef CONFIG_EXT4DEV_COMPAT
4015 err = register_filesystem(&ext4dev_fs_type);
4016 if (err) {
4017 unregister_filesystem(&ext4_fs_type);
4018 goto out;
4019 }
4020#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004021 return 0;
4022out:
4023 destroy_inodecache();
4024out1:
Mingming Cao617ba132006-10-11 01:20:53 -07004025 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004026out2:
4027 exit_ext4_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004028out3:
4029 remove_proc_entry("fs/ext4", NULL);
4030 kset_unregister(ext4_kset);
4031out4:
4032 exit_ext4_system_zone();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004033 return err;
4034}
4035
Mingming Cao617ba132006-10-11 01:20:53 -07004036static void __exit exit_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004037{
Theodore Ts'o03010a32008-10-10 20:02:48 -04004038 unregister_filesystem(&ext4_fs_type);
4039#ifdef CONFIG_EXT4DEV_COMPAT
Mingming Cao617ba132006-10-11 01:20:53 -07004040 unregister_filesystem(&ext4dev_fs_type);
Theodore Ts'o03010a32008-10-10 20:02:48 -04004041#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004042 destroy_inodecache();
Mingming Cao617ba132006-10-11 01:20:53 -07004043 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004044 exit_ext4_mballoc();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004045 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004046 kset_unregister(ext4_kset);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004047 exit_ext4_system_zone();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004048}
4049
4050MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05004051MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004052MODULE_LICENSE("GPL");
Mingming Cao617ba132006-10-11 01:20:53 -07004053module_init(init_ext4_fs)
4054module_exit(exit_ext4_fs)