blob: 735c20d5fd56565dcb53adbf2726acb9cc79fe8c [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
Frank Mayhar03901312009-01-07 00:06:22 -0500606 if (sbi->s_journal) {
607 err = jbd2_journal_destroy(sbi->s_journal);
608 sbi->s_journal = NULL;
609 if (err < 0)
610 ext4_abort(sb, __func__,
611 "Couldn't clean up the journal");
612 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500613
614 ext4_release_system_zone(sb);
615 ext4_mb_release(sb);
616 ext4_ext_release(sb);
617 ext4_xattr_put_super(sb);
618
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700619 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700620 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700621 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400622 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700623 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400624 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400625 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400626 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400627 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700628
629 for (i = 0; i < sbi->s_gdb_count; i++)
630 brelse(sbi->s_group_desc[i]);
631 kfree(sbi->s_group_desc);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -0400632 if (is_vmalloc_addr(sbi->s_flex_groups))
633 vfree(sbi->s_flex_groups);
634 else
635 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700636 percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400639 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700640 brelse(sbi->s_sbh);
641#ifdef CONFIG_QUOTA
642 for (i = 0; i < MAXQUOTAS; i++)
643 kfree(sbi->s_qf_names[i]);
644#endif
645
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi->s_orphan))
651 dump_orphan_list(sb, sbi);
652 J_ASSERT(list_empty(&sbi->s_orphan));
653
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700654 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700655 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
656 /*
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
660 */
661 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700662 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700663 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700664 }
665 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400666 /*
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
669 */
670 unlock_kernel();
671 unlock_super(sb);
672 kobject_put(&sbi->s_kobj);
673 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500674 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700675 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700676}
677
Christoph Lametere18b8902006-12-06 20:33:20 -0800678static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700679
680/*
681 * Called inside transaction, so use GFP_NOFS
682 */
Mingming Cao617ba132006-10-11 01:20:53 -0700683static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700684{
Mingming Cao617ba132006-10-11 01:20:53 -0700685 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700686
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800687 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700688 if (!ei)
689 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400690
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700691 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400692 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700693 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500694 INIT_LIST_HEAD(&ei->i_prealloc_list);
695 spin_lock_init(&ei->i_prealloc_lock);
Frank Mayhar03901312009-01-07 00:06:22 -0500696 /*
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
699 * jinode.
700 */
Jan Kara678aaf42008-07-11 19:27:31 -0400701 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
Mingming Caod2a17632008-07-14 17:52:37 -0400702 ei->i_reserved_data_blocks = 0;
703 ei->i_reserved_meta_blocks = 0;
704 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500705 ei->i_da_metadata_calc_len = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400706 ei->i_delalloc_reserved_flag = 0;
707 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300708#ifdef CONFIG_QUOTA
709 ei->i_reserved_quota = 0;
710#endif
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400711 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
712 ei->cur_aio_dio = NULL;
Jan Karab436b9b2009-12-08 23:51:10 -0500713 ei->i_sync_tid = 0;
714 ei->i_datasync_tid = 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400715
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700716 return &ei->vfs_inode;
717}
718
Mingming Cao617ba132006-10-11 01:20:53 -0700719static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700720{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700721 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400722 ext4_msg(inode->i_sb, KERN_ERR,
723 "Inode %lu (%p): orphan list check failed!",
724 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700725 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
726 EXT4_I(inode), sizeof(struct ext4_inode_info),
727 true);
728 dump_stack();
729 }
Mingming Cao617ba132006-10-11 01:20:53 -0700730 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700731}
732
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700733static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700734{
Mingming Cao617ba132006-10-11 01:20:53 -0700735 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700736
Christoph Lametera35afb82007-05-16 22:10:57 -0700737 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400738#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700739 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700740#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500741 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700742 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700743}
744
745static int init_inodecache(void)
746{
Mingming Cao617ba132006-10-11 01:20:53 -0700747 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
748 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700749 0, (SLAB_RECLAIM_ACCOUNT|
750 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900751 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700752 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700753 return -ENOMEM;
754 return 0;
755}
756
757static void destroy_inodecache(void)
758{
Mingming Cao617ba132006-10-11 01:20:53 -0700759 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700760}
761
Mingming Cao617ba132006-10-11 01:20:53 -0700762static void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700763{
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400764 ext4_discard_preallocations(inode);
Frank Mayhar03901312009-01-07 00:06:22 -0500765 if (EXT4_JOURNAL(inode))
766 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
Jan Kara678aaf42008-07-11 19:27:31 -0400767 &EXT4_I(inode)->jinode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700768}
769
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400770static inline void ext4_show_quota_options(struct seq_file *seq,
771 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700772{
773#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700774 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700775
Jan Kara5a20bdf2009-11-30 23:58:32 +0100776 if (sbi->s_jquota_fmt) {
777 char *fmtname = "";
778
779 switch (sbi->s_jquota_fmt) {
780 case QFMT_VFS_OLD:
781 fmtname = "vfsold";
782 break;
783 case QFMT_VFS_V0:
784 fmtname = "vfsv0";
785 break;
786 case QFMT_VFS_V1:
787 fmtname = "vfsv1";
788 break;
789 }
790 seq_printf(seq, ",jqfmt=%s", fmtname);
791 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700792
793 if (sbi->s_qf_names[USRQUOTA])
794 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
795
796 if (sbi->s_qf_names[GRPQUOTA])
797 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
798
Mingming Cao617ba132006-10-11 01:20:53 -0700799 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700800 seq_puts(seq, ",usrquota");
801
Mingming Cao617ba132006-10-11 01:20:53 -0700802 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700803 seq_puts(seq, ",grpquota");
804#endif
805}
806
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700807/*
808 * Show an option if
809 * - it's set to a non-default value OR
810 * - if the per-sb default is different from the global default
811 */
Mingming Cao617ba132006-10-11 01:20:53 -0700812static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700813{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500814 int def_errors;
815 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700816 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700817 struct ext4_sb_info *sbi = EXT4_SB(sb);
818 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700819
820 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500821 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700822
823 if (sbi->s_sb_block != 1)
824 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
825 if (test_opt(sb, MINIX_DF))
826 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500827 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700828 seq_puts(seq, ",grpid");
829 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
830 seq_puts(seq, ",nogrpid");
831 if (sbi->s_resuid != EXT4_DEF_RESUID ||
832 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
833 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
834 }
835 if (sbi->s_resgid != EXT4_DEF_RESGID ||
836 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
837 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
838 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500839 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700840 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500841 def_errors == EXT4_ERRORS_CONTINUE) {
842 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700843 }
844 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500845 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500846 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500847 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700848 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500849 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700850 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500851 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700852 seq_puts(seq, ",debug");
853 if (test_opt(sb, OLDALLOC))
854 seq_puts(seq, ",oldalloc");
Theodore Ts'o03010a32008-10-10 20:02:48 -0400855#ifdef CONFIG_EXT4_FS_XATTR
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500856 if (test_opt(sb, XATTR_USER) &&
857 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700858 seq_puts(seq, ",user_xattr");
859 if (!test_opt(sb, XATTR_USER) &&
860 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
861 seq_puts(seq, ",nouser_xattr");
862 }
863#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -0400864#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500865 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700866 seq_puts(seq, ",acl");
867 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
868 seq_puts(seq, ",noacl");
869#endif
Theodore Ts'o30773842009-01-03 20:27:38 -0500870 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700871 seq_printf(seq, ",commit=%u",
872 (unsigned) (sbi->s_commit_interval / HZ));
873 }
Theodore Ts'o30773842009-01-03 20:27:38 -0500874 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
875 seq_printf(seq, ",min_batch_time=%u",
876 (unsigned) sbi->s_min_batch_time);
877 }
878 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
879 seq_printf(seq, ",max_batch_time=%u",
880 (unsigned) sbi->s_min_batch_time);
881 }
882
Eric Sandeen571640c2008-05-26 12:29:46 -0400883 /*
884 * We're changing the default of barrier mount option, so
885 * let's always display its mount state so it's clear what its
886 * status is.
887 */
888 seq_puts(seq, ",barrier=");
889 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -0400890 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
891 seq_puts(seq, ",journal_async_commit");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700892 if (test_opt(sb, NOBH))
893 seq_puts(seq, ",nobh");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -0500894 if (test_opt(sb, I_VERSION))
895 seq_puts(seq, ",i_version");
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -0400896 if (!test_opt(sb, DELALLOC))
897 seq_puts(seq, ",nodelalloc");
898
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700899
Miklos Szeredicb45bbe2008-01-28 23:58:27 -0500900 if (sbi->s_stripe)
901 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500902 /*
903 * journal mode get enabled in different ways
904 * So just print the value even if we didn't specify it
905 */
Mingming Cao617ba132006-10-11 01:20:53 -0700906 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700907 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -0700908 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700909 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -0700910 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700911 seq_puts(seq, ",data=writeback");
912
Theodore Ts'o240799c2008-10-09 23:53:47 -0400913 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
914 seq_printf(seq, ",inode_readahead_blks=%u",
915 sbi->s_inode_readahead_blks);
916
Hidehiro Kawai5bf56832008-10-10 22:12:43 -0400917 if (test_opt(sb, DATA_ERR_ABORT))
918 seq_puts(seq, ",data_err=abort");
919
Theodore Ts'oafd46722009-03-16 23:12:23 -0400920 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -0400921 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -0400922
Eric Sandeen5328e632009-11-19 14:25:42 -0500923 if (test_opt(sb, DISCARD))
924 seq_puts(seq, ",discard");
925
Eric Sandeene3bb52a2009-11-19 14:28:50 -0500926 if (test_opt(sb, NOLOAD))
927 seq_puts(seq, ",norecovery");
928
Mingming Cao617ba132006-10-11 01:20:53 -0700929 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400930
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700931 return 0;
932}
933
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700934static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400935 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700936{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700937 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700938
Mingming Cao617ba132006-10-11 01:20:53 -0700939 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700940 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700941 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700942 return ERR_PTR(-ESTALE);
943
944 /* iget isn't really right if the inode is currently unallocated!!
945 *
Mingming Cao617ba132006-10-11 01:20:53 -0700946 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700947 * deleted, so we should be safe.
948 *
949 * Currently we don't know the generation for parent directory, so
950 * a generation of 0 means "accept any"
951 */
David Howells1d1fe1e2008-02-07 00:15:37 -0800952 inode = ext4_iget(sb, ino);
953 if (IS_ERR(inode))
954 return ERR_CAST(inode);
955 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700956 iput(inode);
957 return ERR_PTR(-ESTALE);
958 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700959
960 return inode;
961}
962
963static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400964 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700965{
966 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
967 ext4_nfs_get_inode);
968}
969
970static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400971 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700972{
973 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
974 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700975}
976
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -0500977/*
978 * Try to release metadata pages (indirect blocks, directories) which are
979 * mapped via the block device. Since these pages could have journal heads
980 * which would prevent try_to_free_buffers() from freeing them, we must use
981 * jbd2 layer's try_to_free_buffers() function to release them.
982 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400983static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
984 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -0500985{
986 journal_t *journal = EXT4_SB(sb)->s_journal;
987
988 WARN_ON(PageChecked(page));
989 if (!page_has_buffers(page))
990 return 0;
991 if (journal)
992 return jbd2_journal_try_to_free_buffers(journal, page,
993 wait & ~__GFP_WAIT);
994 return try_to_free_buffers(page);
995}
996
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700997#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400998#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400999#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001000
Mingming Cao617ba132006-10-11 01:20:53 -07001001static int ext4_write_dquot(struct dquot *dquot);
1002static int ext4_acquire_dquot(struct dquot *dquot);
1003static int ext4_release_dquot(struct dquot *dquot);
1004static int ext4_mark_dquot_dirty(struct dquot *dquot);
1005static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001006static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1007 char *path, int remount);
Mingming Cao617ba132006-10-11 01:20:53 -07001008static int ext4_quota_on_mount(struct super_block *sb, int type);
1009static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001010 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001011static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001012 const char *data, size_t len, loff_t off);
1013
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001014static const struct dquot_operations ext4_quota_operations = {
Jan Karaedf72452009-01-12 19:05:26 +01001015 .initialize = dquot_initialize,
1016 .drop = dquot_drop,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001017 .alloc_space = dquot_alloc_space,
Mingming Cao60e58e02009-01-22 18:13:05 +01001018 .reserve_space = dquot_reserve_space,
1019 .claim_space = dquot_claim_space,
1020 .release_rsv = dquot_release_reserved_space,
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001021#ifdef CONFIG_QUOTA
Mingming Cao60e58e02009-01-22 18:13:05 +01001022 .get_reserved_space = ext4_get_reserved_space,
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001023#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001024 .alloc_inode = dquot_alloc_inode,
1025 .free_space = dquot_free_space,
1026 .free_inode = dquot_free_inode,
1027 .transfer = dquot_transfer,
Mingming Cao617ba132006-10-11 01:20:53 -07001028 .write_dquot = ext4_write_dquot,
1029 .acquire_dquot = ext4_acquire_dquot,
1030 .release_dquot = ext4_release_dquot,
1031 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001032 .write_info = ext4_write_info,
1033 .alloc_dquot = dquot_alloc,
1034 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001035};
1036
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001037static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001038 .quota_on = ext4_quota_on,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001039 .quota_off = vfs_quota_off,
1040 .quota_sync = vfs_quota_sync,
1041 .get_info = vfs_get_dqinfo,
1042 .set_info = vfs_set_dqinfo,
1043 .get_dqblk = vfs_get_dqblk,
1044 .set_dqblk = vfs_set_dqblk
1045};
1046#endif
1047
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001048static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001049 .alloc_inode = ext4_alloc_inode,
1050 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001051 .write_inode = ext4_write_inode,
1052 .dirty_inode = ext4_dirty_inode,
1053 .delete_inode = ext4_delete_inode,
1054 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001055 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001056 .freeze_fs = ext4_freeze,
1057 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001058 .statfs = ext4_statfs,
1059 .remount_fs = ext4_remount,
1060 .clear_inode = ext4_clear_inode,
1061 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001062#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001063 .quota_read = ext4_quota_read,
1064 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001065#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001066 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001067};
1068
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001069static const struct super_operations ext4_nojournal_sops = {
1070 .alloc_inode = ext4_alloc_inode,
1071 .destroy_inode = ext4_destroy_inode,
1072 .write_inode = ext4_write_inode,
1073 .dirty_inode = ext4_dirty_inode,
1074 .delete_inode = ext4_delete_inode,
1075 .write_super = ext4_write_super,
1076 .put_super = ext4_put_super,
1077 .statfs = ext4_statfs,
1078 .remount_fs = ext4_remount,
1079 .clear_inode = ext4_clear_inode,
1080 .show_options = ext4_show_options,
1081#ifdef CONFIG_QUOTA
1082 .quota_read = ext4_quota_read,
1083 .quota_write = ext4_quota_write,
1084#endif
1085 .bdev_try_to_free_page = bdev_try_to_free_page,
1086};
1087
Christoph Hellwig39655162007-10-21 16:42:17 -07001088static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001089 .fh_to_dentry = ext4_fh_to_dentry,
1090 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001091 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001092};
1093
1094enum {
1095 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1096 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001097 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001098 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001099 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001100 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001101 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001102 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001103 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001104 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001105 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001106 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1107 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1108 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001109 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001110 Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001111 Opt_inode_readahead_blks, Opt_journal_ioprio,
1112 Opt_discard, Opt_nodiscard,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001113};
1114
Steven Whitehousea447c092008-10-13 10:46:57 +01001115static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001116 {Opt_bsd_df, "bsddf"},
1117 {Opt_minix_df, "minixdf"},
1118 {Opt_grpid, "grpid"},
1119 {Opt_grpid, "bsdgroups"},
1120 {Opt_nogrpid, "nogrpid"},
1121 {Opt_nogrpid, "sysvgroups"},
1122 {Opt_resgid, "resgid=%u"},
1123 {Opt_resuid, "resuid=%u"},
1124 {Opt_sb, "sb=%u"},
1125 {Opt_err_cont, "errors=continue"},
1126 {Opt_err_panic, "errors=panic"},
1127 {Opt_err_ro, "errors=remount-ro"},
1128 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001129 {Opt_debug, "debug"},
1130 {Opt_oldalloc, "oldalloc"},
1131 {Opt_orlov, "orlov"},
1132 {Opt_user_xattr, "user_xattr"},
1133 {Opt_nouser_xattr, "nouser_xattr"},
1134 {Opt_acl, "acl"},
1135 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001136 {Opt_noload, "noload"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001137 {Opt_noload, "norecovery"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001138 {Opt_nobh, "nobh"},
1139 {Opt_bh, "bh"},
1140 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001141 {Opt_min_batch_time, "min_batch_time=%u"},
1142 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001143 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001144 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001145 {Opt_journal_checksum, "journal_checksum"},
1146 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001147 {Opt_abort, "abort"},
1148 {Opt_data_journal, "data=journal"},
1149 {Opt_data_ordered, "data=ordered"},
1150 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001151 {Opt_data_err_abort, "data_err=abort"},
1152 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001153 {Opt_offusrjquota, "usrjquota="},
1154 {Opt_usrjquota, "usrjquota=%s"},
1155 {Opt_offgrpjquota, "grpjquota="},
1156 {Opt_grpjquota, "grpjquota=%s"},
1157 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1158 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001159 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001160 {Opt_grpquota, "grpquota"},
1161 {Opt_noquota, "noquota"},
1162 {Opt_quota, "quota"},
1163 {Opt_usrquota, "usrquota"},
1164 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001165 {Opt_barrier, "barrier"},
1166 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001167 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001168 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001169 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001170 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001171 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001172 {Opt_block_validity, "block_validity"},
1173 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001174 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001175 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001176 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001177 {Opt_auto_da_alloc, "auto_da_alloc"},
1178 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001179 {Opt_discard, "discard"},
1180 {Opt_nodiscard, "nodiscard"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001181 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001182};
1183
Mingming Cao617ba132006-10-11 01:20:53 -07001184static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001185{
Mingming Cao617ba132006-10-11 01:20:53 -07001186 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001187 char *options = (char *) *data;
1188
1189 if (!options || strncmp(options, "sb=", 3) != 0)
1190 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001191
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001192 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001193 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001194 sb_block = simple_strtoul(options, &options, 0);
1195 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001196 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001197 (char *) *data);
1198 return 1;
1199 }
1200 if (*options == ',')
1201 options++;
1202 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001203
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001204 return sb_block;
1205}
1206
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001207#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1208
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001209static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001210 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001211 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001212 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001213{
Mingming Cao617ba132006-10-11 01:20:53 -07001214 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001215 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001216 substring_t args[MAX_OPT_ARGS];
1217 int data_opt = 0;
1218 int option;
1219#ifdef CONFIG_QUOTA
Jan Karadfc5d032008-05-13 19:11:51 -04001220 int qtype, qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001221 char *qname;
1222#endif
1223
1224 if (!options)
1225 return 1;
1226
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001227 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001228 int token;
1229 if (!*p)
1230 continue;
1231
1232 token = match_token(p, tokens, args);
1233 switch (token) {
1234 case Opt_bsd_df:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001235 clear_opt(sbi->s_mount_opt, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001236 break;
1237 case Opt_minix_df:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001238 set_opt(sbi->s_mount_opt, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001239 break;
1240 case Opt_grpid:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001241 set_opt(sbi->s_mount_opt, GRPID);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001242 break;
1243 case Opt_nogrpid:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001244 clear_opt(sbi->s_mount_opt, GRPID);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001245 break;
1246 case Opt_resuid:
1247 if (match_int(&args[0], &option))
1248 return 0;
1249 sbi->s_resuid = option;
1250 break;
1251 case Opt_resgid:
1252 if (match_int(&args[0], &option))
1253 return 0;
1254 sbi->s_resgid = option;
1255 break;
1256 case Opt_sb:
1257 /* handled by get_sb_block() instead of here */
1258 /* *sb_block = match_int(&args[0]); */
1259 break;
1260 case Opt_err_panic:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001261 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1262 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1263 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001264 break;
1265 case Opt_err_ro:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001266 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1267 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1268 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001269 break;
1270 case Opt_err_cont:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001271 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1272 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1273 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001274 break;
1275 case Opt_nouid32:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001276 set_opt(sbi->s_mount_opt, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001277 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001278 case Opt_debug:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001279 set_opt(sbi->s_mount_opt, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001280 break;
1281 case Opt_oldalloc:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001282 set_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001283 break;
1284 case Opt_orlov:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001285 clear_opt(sbi->s_mount_opt, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001286 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001287#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001288 case Opt_user_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001289 set_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001290 break;
1291 case Opt_nouser_xattr:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001292 clear_opt(sbi->s_mount_opt, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001293 break;
1294#else
1295 case Opt_user_xattr:
1296 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001297 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001298 break;
1299#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001300#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001301 case Opt_acl:
1302 set_opt(sbi->s_mount_opt, POSIX_ACL);
1303 break;
1304 case Opt_noacl:
1305 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1306 break;
1307#else
1308 case Opt_acl:
1309 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001310 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001311 break;
1312#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001313 case Opt_journal_update:
1314 /* @@@ FIXME */
1315 /* Eventually we will want to be able to create
1316 a journal file here. For now, only allow the
1317 user to specify an existing inode to be the
1318 journal file. */
1319 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001320 ext4_msg(sb, KERN_ERR,
1321 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001322 return 0;
1323 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001324 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001325 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001326 case Opt_journal_dev:
1327 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001328 ext4_msg(sb, KERN_ERR,
1329 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001330 return 0;
1331 }
1332 if (match_int(&args[0], &option))
1333 return 0;
1334 *journal_devnum = option;
1335 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001336 case Opt_journal_checksum:
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001337 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1338 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001339 case Opt_journal_async_commit:
1340 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001341 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001342 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001343 case Opt_noload:
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001344 set_opt(sbi->s_mount_opt, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001345 break;
1346 case Opt_commit:
1347 if (match_int(&args[0], &option))
1348 return 0;
1349 if (option < 0)
1350 return 0;
1351 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001352 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001353 sbi->s_commit_interval = HZ * option;
1354 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001355 case Opt_max_batch_time:
1356 if (match_int(&args[0], &option))
1357 return 0;
1358 if (option < 0)
1359 return 0;
1360 if (option == 0)
1361 option = EXT4_DEF_MAX_BATCH_TIME;
1362 sbi->s_max_batch_time = option;
1363 break;
1364 case Opt_min_batch_time:
1365 if (match_int(&args[0], &option))
1366 return 0;
1367 if (option < 0)
1368 return 0;
1369 sbi->s_min_batch_time = option;
1370 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001371 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001372 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001373 goto datacheck;
1374 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001375 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001376 goto datacheck;
1377 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001378 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001379 datacheck:
1380 if (is_remount) {
Mingming Cao617ba132006-10-11 01:20:53 -07001381 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001382 != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001383 ext4_msg(sb, KERN_ERR,
1384 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001385 return 0;
1386 }
1387 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07001388 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001389 sbi->s_mount_opt |= data_opt;
1390 }
1391 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001392 case Opt_data_err_abort:
1393 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1394 break;
1395 case Opt_data_err_ignore:
1396 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1397 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001398#ifdef CONFIG_QUOTA
1399 case Opt_usrjquota:
1400 qtype = USRQUOTA;
1401 goto set_qf_name;
1402 case Opt_grpjquota:
1403 qtype = GRPQUOTA;
1404set_qf_name:
Jan Kara17bd13b2008-08-20 18:14:35 +02001405 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001406 !sbi->s_qf_names[qtype]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001407 ext4_msg(sb, KERN_ERR,
1408 "Cannot change journaled "
1409 "quota options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001410 return 0;
1411 }
1412 qname = match_strdup(&args[0]);
1413 if (!qname) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001414 ext4_msg(sb, KERN_ERR,
1415 "Not enough memory for "
1416 "storing quotafile name");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001417 return 0;
1418 }
1419 if (sbi->s_qf_names[qtype] &&
1420 strcmp(sbi->s_qf_names[qtype], qname)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001421 ext4_msg(sb, KERN_ERR,
1422 "%s quota file already "
1423 "specified", QTYPE2NAME(qtype));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001424 kfree(qname);
1425 return 0;
1426 }
1427 sbi->s_qf_names[qtype] = qname;
1428 if (strchr(sbi->s_qf_names[qtype], '/')) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001429 ext4_msg(sb, KERN_ERR,
1430 "quotafile must be on "
1431 "filesystem root");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001432 kfree(sbi->s_qf_names[qtype]);
1433 sbi->s_qf_names[qtype] = NULL;
1434 return 0;
1435 }
1436 set_opt(sbi->s_mount_opt, QUOTA);
1437 break;
1438 case Opt_offusrjquota:
1439 qtype = USRQUOTA;
1440 goto clear_qf_name;
1441 case Opt_offgrpjquota:
1442 qtype = GRPQUOTA;
1443clear_qf_name:
Jan Kara17bd13b2008-08-20 18:14:35 +02001444 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001445 sbi->s_qf_names[qtype]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001446 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001447 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001448 "quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001449 return 0;
1450 }
1451 /*
1452 * The space will be released later when all options
1453 * are confirmed to be correct
1454 */
1455 sbi->s_qf_names[qtype] = NULL;
1456 break;
1457 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001458 qfmt = QFMT_VFS_OLD;
1459 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001460 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001461 qfmt = QFMT_VFS_V0;
Jan Kara5a20bdf2009-11-30 23:58:32 +01001462 goto set_qf_format;
1463 case Opt_jqfmt_vfsv1:
1464 qfmt = QFMT_VFS_V1;
Jan Karadfc5d032008-05-13 19:11:51 -04001465set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001466 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001467 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001468 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001469 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001470 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001471 return 0;
1472 }
1473 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001474 break;
1475 case Opt_quota:
1476 case Opt_usrquota:
1477 set_opt(sbi->s_mount_opt, QUOTA);
1478 set_opt(sbi->s_mount_opt, USRQUOTA);
1479 break;
1480 case Opt_grpquota:
1481 set_opt(sbi->s_mount_opt, QUOTA);
1482 set_opt(sbi->s_mount_opt, GRPQUOTA);
1483 break;
1484 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001485 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001486 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1487 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001488 return 0;
1489 }
1490 clear_opt(sbi->s_mount_opt, QUOTA);
1491 clear_opt(sbi->s_mount_opt, USRQUOTA);
1492 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1493 break;
1494#else
1495 case Opt_quota:
1496 case Opt_usrquota:
1497 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001498 ext4_msg(sb, KERN_ERR,
1499 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001500 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001501 case Opt_usrjquota:
1502 case Opt_grpjquota:
1503 case Opt_offusrjquota:
1504 case Opt_offgrpjquota:
1505 case Opt_jqfmt_vfsold:
1506 case Opt_jqfmt_vfsv0:
Jan Kara5a20bdf2009-11-30 23:58:32 +01001507 case Opt_jqfmt_vfsv1:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001508 ext4_msg(sb, KERN_ERR,
1509 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001510 break;
1511 case Opt_noquota:
1512 break;
1513#endif
1514 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001515 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001516 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001517 case Opt_nobarrier:
1518 clear_opt(sbi->s_mount_opt, BARRIER);
1519 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001520 case Opt_barrier:
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001521 if (match_int(&args[0], &option)) {
1522 set_opt(sbi->s_mount_opt, BARRIER);
1523 break;
1524 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001525 if (option)
1526 set_opt(sbi->s_mount_opt, BARRIER);
1527 else
1528 clear_opt(sbi->s_mount_opt, BARRIER);
1529 break;
1530 case Opt_ignore:
1531 break;
1532 case Opt_resize:
1533 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001534 ext4_msg(sb, KERN_ERR,
1535 "resize option only available "
1536 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001537 return 0;
1538 }
1539 if (match_int(&args[0], &option) != 0)
1540 return 0;
1541 *n_blocks_count = option;
1542 break;
1543 case Opt_nobh:
1544 set_opt(sbi->s_mount_opt, NOBH);
1545 break;
1546 case Opt_bh:
1547 clear_opt(sbi->s_mount_opt, NOBH);
1548 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001549 case Opt_i_version:
1550 set_opt(sbi->s_mount_opt, I_VERSION);
1551 sb->s_flags |= MS_I_VERSION;
1552 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001553 case Opt_nodelalloc:
1554 clear_opt(sbi->s_mount_opt, DELALLOC);
1555 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001556 case Opt_stripe:
1557 if (match_int(&args[0], &option))
1558 return 0;
1559 if (option < 0)
1560 return 0;
1561 sbi->s_stripe = option;
1562 break;
Alex Tomas64769242008-07-11 19:27:31 -04001563 case Opt_delalloc:
1564 set_opt(sbi->s_mount_opt, DELALLOC);
1565 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001566 case Opt_block_validity:
1567 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1568 break;
1569 case Opt_noblock_validity:
1570 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1571 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001572 case Opt_inode_readahead_blks:
1573 if (match_int(&args[0], &option))
1574 return 0;
1575 if (option < 0 || option > (1 << 30))
1576 return 0;
Theodore Ts'of7c43952009-04-24 23:31:59 -04001577 if (!is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001578 ext4_msg(sb, KERN_ERR,
1579 "EXT4-fs: inode_readahead_blks"
1580 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001581 return 0;
1582 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001583 sbi->s_inode_readahead_blks = option;
1584 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001585 case Opt_journal_ioprio:
1586 if (match_int(&args[0], &option))
1587 return 0;
1588 if (option < 0 || option > 7)
1589 break;
1590 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1591 option);
1592 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001593 case Opt_noauto_da_alloc:
1594 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1595 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001596 case Opt_auto_da_alloc:
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001597 if (match_int(&args[0], &option)) {
1598 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1599 break;
1600 }
Theodore Ts'oafd46722009-03-16 23:12:23 -04001601 if (option)
1602 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1603 else
1604 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1605 break;
Eric Sandeen5328e632009-11-19 14:25:42 -05001606 case Opt_discard:
1607 set_opt(sbi->s_mount_opt, DISCARD);
1608 break;
1609 case Opt_nodiscard:
1610 clear_opt(sbi->s_mount_opt, DISCARD);
1611 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001612 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001613 ext4_msg(sb, KERN_ERR,
1614 "Unrecognized mount option \"%s\" "
1615 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001616 return 0;
1617 }
1618 }
1619#ifdef CONFIG_QUOTA
1620 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Mingming Cao617ba132006-10-11 01:20:53 -07001621 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001622 sbi->s_qf_names[USRQUOTA])
1623 clear_opt(sbi->s_mount_opt, USRQUOTA);
1624
Mingming Cao617ba132006-10-11 01:20:53 -07001625 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001626 sbi->s_qf_names[GRPQUOTA])
1627 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1628
1629 if ((sbi->s_qf_names[USRQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001630 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001631 (sbi->s_qf_names[GRPQUOTA] &&
Mingming Cao617ba132006-10-11 01:20:53 -07001632 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001633 ext4_msg(sb, KERN_ERR, "old and new quota "
1634 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001635 return 0;
1636 }
1637
1638 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001639 ext4_msg(sb, KERN_ERR, "journaled quota format "
1640 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001641 return 0;
1642 }
1643 } else {
1644 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001645 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001646 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001647 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001648 return 0;
1649 }
1650 }
1651#endif
1652 return 1;
1653}
1654
Mingming Cao617ba132006-10-11 01:20:53 -07001655static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001656 int read_only)
1657{
Mingming Cao617ba132006-10-11 01:20:53 -07001658 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001659 int res = 0;
1660
Mingming Cao617ba132006-10-11 01:20:53 -07001661 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001662 ext4_msg(sb, KERN_ERR, "revision level too high, "
1663 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001664 res = MS_RDONLY;
1665 }
1666 if (read_only)
1667 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001668 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001669 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1670 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001671 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001672 ext4_msg(sb, KERN_WARNING,
1673 "warning: mounting fs with errors, "
1674 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001675 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1676 le16_to_cpu(es->s_mnt_count) >=
1677 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001678 ext4_msg(sb, KERN_WARNING,
1679 "warning: maximal mount count reached, "
1680 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001681 else if (le32_to_cpu(es->s_checkinterval) &&
1682 (le32_to_cpu(es->s_lastcheck) +
1683 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001684 ext4_msg(sb, KERN_WARNING,
1685 "warning: checktime reached, "
1686 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001687 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001688 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001689 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001690 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001691 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001692 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001693 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001694 if (sbi->s_journal)
1695 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001696
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001697 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001698 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001699 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'o7f4520c2009-06-13 10:09:41 -04001700 "bpg=%lu, ipg=%lu, mo=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001701 sb->s_blocksize,
1702 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001703 EXT4_BLOCKS_PER_GROUP(sb),
1704 EXT4_INODES_PER_GROUP(sb),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001705 sbi->s_mount_opt);
1706
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001707 return res;
1708}
1709
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001710static int ext4_fill_flex_info(struct super_block *sb)
1711{
1712 struct ext4_sb_info *sbi = EXT4_SB(sb);
1713 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001714 ext4_group_t flex_group_count;
1715 ext4_group_t flex_group;
1716 int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001717 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001718 int i;
1719
Theodore Ts'o503358a2009-11-23 07:24:46 -05001720 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1721 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1722
1723 if (groups_per_flex < 2) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001724 sbi->s_log_groups_per_flex = 0;
1725 return 1;
1726 }
1727
Frederic Bohec62a11f2008-09-08 10:20:24 -04001728 /* We allocate both existing and potentially added groups */
1729 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05001730 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1731 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001732 size = flex_group_count * sizeof(struct flex_groups);
1733 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1734 if (sbi->s_flex_groups == NULL) {
1735 sbi->s_flex_groups = vmalloc(size);
1736 if (sbi->s_flex_groups)
1737 memset(sbi->s_flex_groups, 0, size);
1738 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001739 if (sbi->s_flex_groups == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001740 ext4_msg(sb, KERN_ERR, "not enough memory for "
1741 "%u flex groups", flex_group_count);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001742 goto failed;
1743 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001744
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001745 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001746 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001747
1748 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001749 atomic_add(ext4_free_inodes_count(sb, gdp),
1750 &sbi->s_flex_groups[flex_group].free_inodes);
1751 atomic_add(ext4_free_blks_count(sb, gdp),
1752 &sbi->s_flex_groups[flex_group].free_blocks);
1753 atomic_add(ext4_used_dirs_count(sb, gdp),
1754 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001755 }
1756
1757 return 1;
1758failed:
1759 return 0;
1760}
1761
Andreas Dilger717d50e2007-10-16 18:38:25 -04001762__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1763 struct ext4_group_desc *gdp)
1764{
1765 __u16 crc = 0;
1766
1767 if (sbi->s_es->s_feature_ro_compat &
1768 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1769 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1770 __le32 le_group = cpu_to_le32(block_group);
1771
1772 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1773 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1774 crc = crc16(crc, (__u8 *)gdp, offset);
1775 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1776 /* for checksum of struct ext4_group_desc do the rest...*/
1777 if ((sbi->s_es->s_feature_incompat &
1778 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1779 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1780 crc = crc16(crc, (__u8 *)gdp + offset,
1781 le16_to_cpu(sbi->s_es->s_desc_size) -
1782 offset);
1783 }
1784
1785 return cpu_to_le16(crc);
1786}
1787
1788int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1789 struct ext4_group_desc *gdp)
1790{
1791 if ((sbi->s_es->s_feature_ro_compat &
1792 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1793 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1794 return 0;
1795
1796 return 1;
1797}
1798
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001799/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001800static int ext4_check_descriptors(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001801{
Mingming Cao617ba132006-10-11 01:20:53 -07001802 struct ext4_sb_info *sbi = EXT4_SB(sb);
1803 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1804 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001805 ext4_fsblk_t block_bitmap;
1806 ext4_fsblk_t inode_bitmap;
1807 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04001808 int flexbg_flag = 0;
Avantika Mathurfd2d4292008-01-28 23:58:27 -05001809 ext4_group_t i;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001810
Jose R. Santosce421582007-10-16 18:38:25 -04001811 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1812 flexbg_flag = 1;
1813
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001814 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001815
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001816 for (i = 0; i < sbi->s_groups_count; i++) {
1817 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1818
Jose R. Santosce421582007-10-16 18:38:25 -04001819 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001820 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001821 else
1822 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07001823 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001824
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001825 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001826 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001827 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001828 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001829 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001830 return 0;
1831 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001832 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001833 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001834 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001835 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001836 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001837 return 0;
1838 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07001839 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001840 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001841 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001842 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001843 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001844 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001845 return 0;
1846 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001847 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001848 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001849 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1850 "Checksum for group %u failed (%u!=%u)",
1851 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1852 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04001853 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001854 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04001855 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04001856 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04001857 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04001858 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04001859 if (!flexbg_flag)
1860 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001861 }
1862
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07001863 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001864 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001865 return 1;
1866}
1867
Mingming Cao617ba132006-10-11 01:20:53 -07001868/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 * the superblock) which were deleted from all directories, but held open by
1870 * a process at the time of a crash. We walk the list and try to delete these
1871 * inodes at recovery time (only with a read-write filesystem).
1872 *
1873 * In order to keep the orphan inode chain consistent during traversal (in
1874 * case of crash during recovery), we link each inode into the superblock
1875 * orphan list_head and handle it the same way as an inode deletion during
1876 * normal operation (which journals the operations for us).
1877 *
1878 * We only do an iget() and an iput() on each inode, which is very safe if we
1879 * accidentally point at an in-use or already deleted inode. The worst that
1880 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07001881 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001882 * e2fsck was run on this filesystem, and it must have already done the orphan
1883 * inode cleanup for us, so we can safely abort without any further action.
1884 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001885static void ext4_orphan_cleanup(struct super_block *sb,
1886 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001887{
1888 unsigned int s_flags = sb->s_flags;
1889 int nr_orphans = 0, nr_truncates = 0;
1890#ifdef CONFIG_QUOTA
1891 int i;
1892#endif
1893 if (!es->s_last_orphan) {
1894 jbd_debug(4, "no orphan inodes to clean up\n");
1895 return;
1896 }
1897
Eric Sandeena8f48a92006-12-06 20:40:13 -08001898 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001899 ext4_msg(sb, KERN_ERR, "write access "
1900 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08001901 return;
1902 }
1903
Mingming Cao617ba132006-10-11 01:20:53 -07001904 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001905 if (es->s_last_orphan)
1906 jbd_debug(1, "Errors on filesystem, "
1907 "clearing orphan list.\n");
1908 es->s_last_orphan = 0;
1909 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1910 return;
1911 }
1912
1913 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001914 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001915 sb->s_flags &= ~MS_RDONLY;
1916 }
1917#ifdef CONFIG_QUOTA
1918 /* Needed for iput() to work correctly and not trash data */
1919 sb->s_flags |= MS_ACTIVE;
1920 /* Turn on quotas so that they are updated correctly */
1921 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07001922 if (EXT4_SB(sb)->s_qf_names[i]) {
1923 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001924 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001925 ext4_msg(sb, KERN_ERR,
1926 "Cannot turn on journaled "
1927 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001928 }
1929 }
1930#endif
1931
1932 while (es->s_last_orphan) {
1933 struct inode *inode;
1934
Josef Bacik97bd42b2008-04-29 22:04:56 -04001935 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1936 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001937 es->s_last_orphan = 0;
1938 break;
1939 }
1940
Mingming Cao617ba132006-10-11 01:20:53 -07001941 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Jan Karaa269eb12009-01-26 17:04:39 +01001942 vfs_dq_init(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001943 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001944 ext4_msg(sb, KERN_DEBUG,
1945 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04001946 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04001947 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001948 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07001949 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001950 nr_truncates++;
1951 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001952 ext4_msg(sb, KERN_DEBUG,
1953 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04001954 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001955 jbd_debug(2, "deleting unreferenced inode %lu\n",
1956 inode->i_ino);
1957 nr_orphans++;
1958 }
1959 iput(inode); /* The delete magic happens here! */
1960 }
1961
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001962#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001963
1964 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001965 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1966 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001967 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001968 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1969 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001970#ifdef CONFIG_QUOTA
1971 /* Turn quotas off */
1972 for (i = 0; i < MAXQUOTAS; i++) {
1973 if (sb_dqopt(sb)->files[i])
Jan Kara6f28e082008-04-28 02:14:34 -07001974 vfs_quota_off(sb, i, 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001975 }
1976#endif
1977 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1978}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001979
Eric Sandeencd2291a2008-01-28 23:58:27 -05001980/*
1981 * Maximal extent format file size.
1982 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1983 * extent format containers, within a sector_t, and within i_blocks
1984 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1985 * so that won't be a limiting factor.
1986 *
1987 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1988 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04001989static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05001990{
1991 loff_t res;
1992 loff_t upper_limit = MAX_LFS_FILESIZE;
1993
1994 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04001995 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05001996 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02001997 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05001998 * i_block represent total blocks in 512 bytes
1999 * 32 == size of vfs inode i_blocks * 8
2000 */
2001 upper_limit = (1LL << 32) - 1;
2002
2003 /* total blocks in file system block size */
2004 upper_limit >>= (blkbits - 9);
2005 upper_limit <<= blkbits;
2006 }
2007
2008 /* 32-bit extent-start container, ee_block */
2009 res = 1LL << 32;
2010 res <<= blkbits;
2011 res -= 1;
2012
2013 /* Sanity check against vm- & vfs- imposed limits */
2014 if (res > upper_limit)
2015 res = upper_limit;
2016
2017 return res;
2018}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002019
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002020/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002021 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002022 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2023 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002024 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002025static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002026{
Mingming Cao617ba132006-10-11 01:20:53 -07002027 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002028 int meta_blocks;
2029 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002030 /* This is calculated to be the largest file size for a dense, block
2031 * mapped file such that the file's total number of 512-byte sectors,
2032 * including data and all indirect blocks, does not exceed (2^48 - 1).
2033 *
2034 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2035 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002036 */
2037
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002038 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002039 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002040 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002041 * the inode i_block field represents total file blocks in
2042 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002043 */
2044 upper_limit = (1LL << 32) - 1;
2045
2046 /* total blocks in file system block size */
2047 upper_limit >>= (bits - 9);
2048
2049 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002050 /*
2051 * We use 48 bit ext4_inode i_blocks
2052 * With EXT4_HUGE_FILE_FL set the i_blocks
2053 * represent total number of blocks in
2054 * file system block size
2055 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002056 upper_limit = (1LL << 48) - 1;
2057
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002058 }
2059
2060 /* indirect blocks */
2061 meta_blocks = 1;
2062 /* double indirect blocks */
2063 meta_blocks += 1 + (1LL << (bits-2));
2064 /* tripple indirect blocks */
2065 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2066
2067 upper_limit -= meta_blocks;
2068 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002069
2070 res += 1LL << (bits-2);
2071 res += 1LL << (2*(bits-2));
2072 res += 1LL << (3*(bits-2));
2073 res <<= bits;
2074 if (res > upper_limit)
2075 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002076
2077 if (res > MAX_LFS_FILESIZE)
2078 res = MAX_LFS_FILESIZE;
2079
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002080 return res;
2081}
2082
Mingming Cao617ba132006-10-11 01:20:53 -07002083static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002084 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002085{
Mingming Cao617ba132006-10-11 01:20:53 -07002086 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002087 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002088 int has_super = 0;
2089
2090 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2091
Mingming Cao617ba132006-10-11 01:20:53 -07002092 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002093 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002094 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002095 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002096 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002097 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002098
Mingming Cao617ba132006-10-11 01:20:53 -07002099 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002100}
2101
Alex Tomasc9de5602008-01-29 00:19:52 -05002102/**
2103 * ext4_get_stripe_size: Get the stripe size.
2104 * @sbi: In memory super block info
2105 *
2106 * If we have specified it via mount option, then
2107 * use the mount option value. If the value specified at mount time is
2108 * greater than the blocks per group use the super block value.
2109 * If the super block value is greater than blocks per group return 0.
2110 * Allocator needs it be less than blocks per group.
2111 *
2112 */
2113static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2114{
2115 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2116 unsigned long stripe_width =
2117 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2118
2119 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2120 return sbi->s_stripe;
2121
2122 if (stripe_width <= sbi->s_blocks_per_group)
2123 return stripe_width;
2124
2125 if (stride <= sbi->s_blocks_per_group)
2126 return stride;
2127
2128 return 0;
2129}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002130
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002131/* sysfs supprt */
2132
2133struct ext4_attr {
2134 struct attribute attr;
2135 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2136 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2137 const char *, size_t);
2138 int offset;
2139};
2140
2141static int parse_strtoul(const char *buf,
2142 unsigned long max, unsigned long *value)
2143{
2144 char *endp;
2145
André Goddard Rosae7d28602009-12-14 18:01:06 -08002146 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2147 endp = skip_spaces(endp);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002148 if (*endp || *value > max)
2149 return -EINVAL;
2150
2151 return 0;
2152}
2153
2154static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2155 struct ext4_sb_info *sbi,
2156 char *buf)
2157{
2158 return snprintf(buf, PAGE_SIZE, "%llu\n",
2159 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2160}
2161
2162static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2163 struct ext4_sb_info *sbi, char *buf)
2164{
2165 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2166
2167 return snprintf(buf, PAGE_SIZE, "%lu\n",
2168 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2169 sbi->s_sectors_written_start) >> 1);
2170}
2171
2172static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2173 struct ext4_sb_info *sbi, char *buf)
2174{
2175 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2176
2177 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002178 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002179 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002180 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002181}
2182
2183static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2184 struct ext4_sb_info *sbi,
2185 const char *buf, size_t count)
2186{
2187 unsigned long t;
2188
2189 if (parse_strtoul(buf, 0x40000000, &t))
2190 return -EINVAL;
2191
Theodore Ts'of7c43952009-04-24 23:31:59 -04002192 if (!is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002193 return -EINVAL;
2194
2195 sbi->s_inode_readahead_blks = t;
2196 return count;
2197}
2198
2199static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002200 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002201{
2202 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2203
2204 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2205}
2206
2207static ssize_t sbi_ui_store(struct ext4_attr *a,
2208 struct ext4_sb_info *sbi,
2209 const char *buf, size_t count)
2210{
2211 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2212 unsigned long t;
2213
2214 if (parse_strtoul(buf, 0xffffffff, &t))
2215 return -EINVAL;
2216 *ui = t;
2217 return count;
2218}
2219
2220#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2221static struct ext4_attr ext4_attr_##_name = { \
2222 .attr = {.name = __stringify(_name), .mode = _mode }, \
2223 .show = _show, \
2224 .store = _store, \
2225 .offset = offsetof(struct ext4_sb_info, _elname), \
2226}
2227#define EXT4_ATTR(name, mode, show, store) \
2228static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2229
2230#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2231#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2232#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2233 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2234#define ATTR_LIST(name) &ext4_attr_##name.attr
2235
2236EXT4_RO_ATTR(delayed_allocation_blocks);
2237EXT4_RO_ATTR(session_write_kbytes);
2238EXT4_RO_ATTR(lifetime_write_kbytes);
2239EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2240 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002241EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002242EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2243EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2244EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2245EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2246EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2247EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002248EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002249
2250static struct attribute *ext4_attrs[] = {
2251 ATTR_LIST(delayed_allocation_blocks),
2252 ATTR_LIST(session_write_kbytes),
2253 ATTR_LIST(lifetime_write_kbytes),
2254 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002255 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002256 ATTR_LIST(mb_stats),
2257 ATTR_LIST(mb_max_to_scan),
2258 ATTR_LIST(mb_min_to_scan),
2259 ATTR_LIST(mb_order2_req),
2260 ATTR_LIST(mb_stream_req),
2261 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002262 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002263 NULL,
2264};
2265
2266static ssize_t ext4_attr_show(struct kobject *kobj,
2267 struct attribute *attr, char *buf)
2268{
2269 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2270 s_kobj);
2271 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2272
2273 return a->show ? a->show(a, sbi, buf) : 0;
2274}
2275
2276static ssize_t ext4_attr_store(struct kobject *kobj,
2277 struct attribute *attr,
2278 const char *buf, size_t len)
2279{
2280 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2281 s_kobj);
2282 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2283
2284 return a->store ? a->store(a, sbi, buf, len) : 0;
2285}
2286
2287static void ext4_sb_release(struct kobject *kobj)
2288{
2289 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2290 s_kobj);
2291 complete(&sbi->s_kobj_unregister);
2292}
2293
2294
2295static struct sysfs_ops ext4_attr_ops = {
2296 .show = ext4_attr_show,
2297 .store = ext4_attr_store,
2298};
2299
2300static struct kobj_type ext4_ktype = {
2301 .default_attrs = ext4_attrs,
2302 .sysfs_ops = &ext4_attr_ops,
2303 .release = ext4_sb_release,
2304};
2305
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002306/*
2307 * Check whether this filesystem can be mounted based on
2308 * the features present and the RDONLY/RDWR mount requested.
2309 * Returns 1 if this filesystem can be mounted as requested,
2310 * 0 if it cannot be.
2311 */
2312static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2313{
2314 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2315 ext4_msg(sb, KERN_ERR,
2316 "Couldn't mount because of "
2317 "unsupported optional features (%x)",
2318 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2319 ~EXT4_FEATURE_INCOMPAT_SUPP));
2320 return 0;
2321 }
2322
2323 if (readonly)
2324 return 1;
2325
2326 /* Check that feature set is OK for a read-write mount */
2327 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2328 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2329 "unsupported optional features (%x)",
2330 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2331 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2332 return 0;
2333 }
2334 /*
2335 * Large file size enabled file system can only be mounted
2336 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2337 */
2338 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2339 if (sizeof(blkcnt_t) < sizeof(u64)) {
2340 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2341 "cannot be mounted RDWR without "
2342 "CONFIG_LBDAF");
2343 return 0;
2344 }
2345 }
2346 return 1;
2347}
2348
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002349static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04002350 __releases(kernel_lock)
2351 __acquires(kernel_lock)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002352{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002353 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07002354 struct ext4_super_block *es = NULL;
2355 struct ext4_sb_info *sbi;
2356 ext4_fsblk_t block;
2357 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002358 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002359 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002360 unsigned long journal_devnum = 0;
2361 unsigned long def_mount_opts;
2362 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002363 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05002364 const char *descr;
David Howells1d1fe1e2008-02-07 00:15:37 -08002365 int ret = -EINVAL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002366 int blocksize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002367 unsigned int db_count;
2368 unsigned int i;
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002369 int needs_recovery, has_huge_files;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002370 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07002371 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002372 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002373
2374 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2375 if (!sbi)
2376 return -ENOMEM;
Pekka Enberg705895b2009-02-15 18:07:52 -05002377
2378 sbi->s_blockgroup_lock =
2379 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2380 if (!sbi->s_blockgroup_lock) {
2381 kfree(sbi);
2382 return -ENOMEM;
2383 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002384 sb->s_fs_info = sbi;
2385 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07002386 sbi->s_resuid = EXT4_DEF_RESUID;
2387 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04002388 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07002389 sbi->s_sb_block = sb_block;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05002390 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2391 sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002392
2393 unlock_kernel();
2394
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002395 /* Cleanup superblock name */
2396 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2397 *cp = '!';
2398
Mingming Cao617ba132006-10-11 01:20:53 -07002399 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002400 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002401 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002402 goto out_fail;
2403 }
2404
2405 /*
Mingming Cao617ba132006-10-11 01:20:53 -07002406 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002407 * block sizes. We need to calculate the offset from buffer start.
2408 */
Mingming Cao617ba132006-10-11 01:20:53 -07002409 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002410 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2411 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002412 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002413 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002414 }
2415
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002416 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002417 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002418 goto out_fail;
2419 }
2420 /*
2421 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07002422 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002423 */
Mingming Cao617ba132006-10-11 01:20:53 -07002424 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002425 sbi->s_es = es;
2426 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07002427 if (sb->s_magic != EXT4_SUPER_MAGIC)
2428 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05002429 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002430
2431 /* Set defaults before we parse the mount options */
2432 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Mingming Cao617ba132006-10-11 01:20:53 -07002433 if (def_mount_opts & EXT4_DEFM_DEBUG)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002434 set_opt(sbi->s_mount_opt, DEBUG);
Mingming Cao617ba132006-10-11 01:20:53 -07002435 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002436 set_opt(sbi->s_mount_opt, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07002437 if (def_mount_opts & EXT4_DEFM_UID16)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002438 set_opt(sbi->s_mount_opt, NO_UID32);
Theodore Ts'o03010a32008-10-10 20:02:48 -04002439#ifdef CONFIG_EXT4_FS_XATTR
Mingming Cao617ba132006-10-11 01:20:53 -07002440 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002441 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08002442#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04002443#ifdef CONFIG_EXT4_FS_POSIX_ACL
Mingming Cao617ba132006-10-11 01:20:53 -07002444 if (def_mount_opts & EXT4_DEFM_ACL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002445 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08002446#endif
Mingming Cao617ba132006-10-11 01:20:53 -07002447 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2448 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2449 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2450 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2451 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2452 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002453
Mingming Cao617ba132006-10-11 01:20:53 -07002454 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002455 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05002456 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Dmitry Mishinceea16b2006-10-11 01:21:21 -07002457 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05002458 else
2459 set_opt(sbi->s_mount_opt, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002460
2461 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2462 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05002463 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2464 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2465 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002466
Eric Sandeen571640c2008-05-26 12:29:46 -04002467 set_opt(sbi->s_mount_opt, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002468
Mingming Cao1e2462f2007-07-18 09:00:55 -04002469 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04002470 * enable delayed allocation by default
2471 * Use -o nodelalloc to turn it off
2472 */
2473 set_opt(sbi->s_mount_opt, DELALLOC);
2474
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002475 if (!parse_options((char *) data, sb, &journal_devnum,
2476 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002477 goto failed_mount;
2478
2479 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07002480 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002481
Mingming Cao617ba132006-10-11 01:20:53 -07002482 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2483 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2484 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2485 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002486 ext4_msg(sb, KERN_WARNING,
2487 "feature flags set on rev 0 fs, "
2488 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05002489
2490 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002491 * Check feature flags regardless of the revision level, since we
2492 * previously didn't change the revision level when setting the flags,
2493 * so there is a chance incompat flags are set on a rev 0 filesystem.
2494 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002495 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002496 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002497
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002498 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2499
Mingming Cao617ba132006-10-11 01:20:53 -07002500 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2501 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002502 ext4_msg(sb, KERN_ERR,
2503 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002504 goto failed_mount;
2505 }
2506
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002507 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002508 /* Validate the filesystem blocksize */
2509 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002510 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05002511 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002512 goto failed_mount;
2513 }
2514
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002515 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002516 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2517 offset = do_div(logical_sb_block, blocksize);
2518 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002519 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002520 ext4_msg(sb, KERN_ERR,
2521 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002522 goto failed_mount;
2523 }
Mingming Cao617ba132006-10-11 01:20:53 -07002524 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002525 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07002526 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002527 ext4_msg(sb, KERN_ERR,
2528 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002529 goto failed_mount;
2530 }
2531 }
2532
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002533 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2534 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002535 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2536 has_huge_files);
2537 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002538
Mingming Cao617ba132006-10-11 01:20:53 -07002539 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2540 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2541 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002542 } else {
2543 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2544 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07002545 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04002546 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002547 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002548 ext4_msg(sb, KERN_ERR,
2549 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002550 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002551 goto failed_mount;
2552 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04002553 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2554 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002555 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002556
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002557 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2558 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002559 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002560 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07002561 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002562 ext4_msg(sb, KERN_ERR,
2563 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002564 sbi->s_desc_size);
2565 goto failed_mount;
2566 }
2567 } else
2568 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002569
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002570 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002571 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08002572 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002573 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002574
Mingming Cao617ba132006-10-11 01:20:53 -07002575 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002576 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07002577 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002578 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2579 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07002580 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002581 sbi->s_sbh = bh;
2582 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07002583 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2584 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002585
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002586 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002587 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2588 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04002589 i = le32_to_cpu(es->s_flags);
2590 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2591 sbi->s_hash_unsigned = 3;
2592 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2593#ifdef __CHAR_UNSIGNED__
2594 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2595 sbi->s_hash_unsigned = 3;
2596#else
2597 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2598#endif
2599 sb->s_dirt = 1;
2600 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002601
2602 if (sbi->s_blocks_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002603 ext4_msg(sb, KERN_ERR,
2604 "#blocks per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002605 sbi->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002606 goto failed_mount;
2607 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002608 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002609 ext4_msg(sb, KERN_ERR,
2610 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002611 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002612 goto failed_mount;
2613 }
2614
Eric Sandeenbf43d842009-08-17 23:48:51 -04002615 /*
2616 * Test whether we have more sectors than will fit in sector_t,
2617 * and whether the max offset is addressable by the page cache.
2618 */
2619 if ((ext4_blocks_count(es) >
2620 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2621 (ext4_blocks_count(es) >
2622 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002623 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04002624 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002625 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002626 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Eric Sandeenbf43d842009-08-17 23:48:51 -04002627 ret = -EFBIG;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002628 goto failed_mount;
2629 }
2630
Mingming Cao617ba132006-10-11 01:20:53 -07002631 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2632 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05002633
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04002634 /* check blocks count against device size */
2635 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2636 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002637 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2638 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04002639 ext4_blocks_count(es), blocks_count);
2640 goto failed_mount;
2641 }
2642
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002643 /*
2644 * It makes no sense for the first data block to be beyond the end
2645 * of the filesystem.
2646 */
2647 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002648 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2649 "block %u is beyond end of filesystem (%llu)",
2650 le32_to_cpu(es->s_first_data_block),
2651 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05002652 goto failed_mount;
2653 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002654 blocks_count = (ext4_blocks_count(es) -
2655 le32_to_cpu(es->s_first_data_block) +
2656 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2657 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002658 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002659 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002660 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002661 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05002662 ext4_blocks_count(es),
2663 le32_to_cpu(es->s_first_data_block),
2664 EXT4_BLOCKS_PER_GROUP(sb));
2665 goto failed_mount;
2666 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002667 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04002668 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2669 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07002670 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2671 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002672 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002673 GFP_KERNEL);
2674 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002675 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002676 goto failed_mount;
2677 }
2678
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04002679#ifdef CONFIG_PROC_FS
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002680 if (ext4_proc_root)
2681 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04002682#endif
Theodore Ts'o240799c2008-10-09 23:53:47 -04002683
Pekka Enberg705895b2009-02-15 18:07:52 -05002684 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002685
2686 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002687 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002688 sbi->s_group_desc[i] = sb_bread(sb, block);
2689 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002690 ext4_msg(sb, KERN_ERR,
2691 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002692 db_count = i;
2693 goto failed_mount2;
2694 }
2695 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002696 if (!ext4_check_descriptors(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002697 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002698 goto failed_mount2;
2699 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002700 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2701 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002702 ext4_msg(sb, KERN_ERR,
2703 "unable to initialize "
2704 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002705 goto failed_mount2;
2706 }
2707
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002708 sbi->s_gdb_count = db_count;
2709 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2710 spin_lock_init(&sbi->s_next_gen_lock);
2711
Peter Zijlstra833f4072007-10-16 23:25:45 -07002712 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2713 ext4_count_free_blocks(sb));
2714 if (!err) {
2715 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2716 ext4_count_free_inodes(sb));
2717 }
2718 if (!err) {
2719 err = percpu_counter_init(&sbi->s_dirs_counter,
2720 ext4_count_dirs(sb));
2721 }
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04002722 if (!err) {
2723 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2724 }
Peter Zijlstra833f4072007-10-16 23:25:45 -07002725 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002726 ext4_msg(sb, KERN_ERR, "insufficient memory");
Peter Zijlstra833f4072007-10-16 23:25:45 -07002727 goto failed_mount3;
2728 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002729
Alex Tomasc9de5602008-01-29 00:19:52 -05002730 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002731 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05002732
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002733 /*
2734 * set up enough so that it can read an inode
2735 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04002736 if (!test_opt(sb, NOLOAD) &&
2737 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2738 sb->s_op = &ext4_sops;
2739 else
2740 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07002741 sb->s_export_op = &ext4_export_ops;
2742 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002743#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07002744 sb->s_qcop = &ext4_qctl_operations;
2745 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002746#endif
2747 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04002748 mutex_init(&sbi->s_orphan_lock);
Theodore Ts'o32ed5052009-04-25 22:53:39 -04002749 mutex_init(&sbi->s_resize_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002750
2751 sb->s_root = NULL;
2752
2753 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07002754 EXT4_HAS_INCOMPAT_FEATURE(sb,
2755 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002756
2757 /*
2758 * The first inode we look at is the journal inode. Don't try
2759 * root first: it may be modified in the journal!
2760 */
2761 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07002762 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2763 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002764 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05002765 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2766 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002767 ext4_msg(sb, KERN_ERR, "required journal recovery "
2768 "suppressed and not mounted read-only");
Frank Mayhar03901312009-01-07 00:06:22 -05002769 goto failed_mount4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002770 } else {
Frank Mayhar03901312009-01-07 00:06:22 -05002771 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2772 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2773 sbi->s_journal = NULL;
2774 needs_recovery = 0;
2775 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002776 }
2777
Jose R. Santoseb40a092007-07-18 08:37:25 -04002778 if (ext4_blocks_count(es) > 0xffffffffULL &&
2779 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2780 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002781 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jose R. Santoseb40a092007-07-18 08:37:25 -04002782 goto failed_mount4;
2783 }
2784
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08002785 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2786 jbd2_journal_set_features(sbi->s_journal,
2787 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05002788 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08002789 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2790 jbd2_journal_set_features(sbi->s_journal,
2791 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
Girish Shilamkar818d2762008-01-28 23:58:27 -05002792 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2793 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08002794 } else {
2795 jbd2_journal_clear_features(sbi->s_journal,
2796 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2797 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2798 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05002799
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002800 /* We have now updated the journal if required, so we can
2801 * validate the data journaling mode. */
2802 switch (test_opt(sb, DATA_FLAGS)) {
2803 case 0:
2804 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07002805 * capabilities: ORDERED_DATA if the journal can
2806 * cope, else JOURNAL_DATA
2807 */
Mingming Caodab291a2006-10-11 01:21:01 -07002808 if (jbd2_journal_check_available_features
2809 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002810 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2811 else
2812 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2813 break;
2814
Mingming Cao617ba132006-10-11 01:20:53 -07002815 case EXT4_MOUNT_ORDERED_DATA:
2816 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07002817 if (!jbd2_journal_check_available_features
2818 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002819 ext4_msg(sb, KERN_ERR, "Journal does not support "
2820 "requested data journaling mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002821 goto failed_mount4;
2822 }
2823 default:
2824 break;
2825 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002826 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002827
Frank Mayhar03901312009-01-07 00:06:22 -05002828no_journal:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002829
2830 if (test_opt(sb, NOBH)) {
Mingming Cao617ba132006-10-11 01:20:53 -07002831 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002832 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2833 "its supported only with writeback mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002834 clear_opt(sbi->s_mount_opt, NOBH);
2835 }
2836 }
Mingming Cao4c0425f2009-09-28 15:48:41 -04002837 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2838 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2839 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2840 goto failed_mount_wq;
2841 }
2842
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002843 /*
Mingming Caodab291a2006-10-11 01:21:01 -07002844 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002845 * so we can safely mount the rest of the filesystem now.
2846 */
2847
David Howells1d1fe1e2008-02-07 00:15:37 -08002848 root = ext4_iget(sb, EXT4_ROOT_INO);
2849 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002850 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08002851 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002852 goto failed_mount4;
2853 }
2854 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08002855 iput(root);
Eric Sandeenb31e1552009-06-04 17:36:36 -04002856 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002857 goto failed_mount4;
2858 }
David Howells1d1fe1e2008-02-07 00:15:37 -08002859 sb->s_root = d_alloc_root(root);
2860 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002861 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08002862 iput(root);
2863 ret = -ENOMEM;
2864 goto failed_mount4;
2865 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002866
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002867 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04002868
2869 /* determine the minimum size of new large inodes, if present */
2870 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2871 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2872 EXT4_GOOD_OLD_INODE_SIZE;
2873 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2874 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2875 if (sbi->s_want_extra_isize <
2876 le16_to_cpu(es->s_want_extra_isize))
2877 sbi->s_want_extra_isize =
2878 le16_to_cpu(es->s_want_extra_isize);
2879 if (sbi->s_want_extra_isize <
2880 le16_to_cpu(es->s_min_extra_isize))
2881 sbi->s_want_extra_isize =
2882 le16_to_cpu(es->s_min_extra_isize);
2883 }
2884 }
2885 /* Check if enough inode space is available */
2886 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2887 sbi->s_inode_size) {
2888 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2889 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04002890 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2891 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04002892 }
2893
Theodore Ts'o90576c02009-09-29 15:51:30 -04002894 if (test_opt(sb, DELALLOC) &&
2895 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002896 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2897 "requested data journaling mode");
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002898 clear_opt(sbi->s_mount_opt, DELALLOC);
Theodore Ts'o90576c02009-09-29 15:51:30 -04002899 }
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002900
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04002901 err = ext4_setup_system_zone(sb);
2902 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002903 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2904 "zone (%d)\n", err);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04002905 goto failed_mount4;
2906 }
2907
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002908 ext4_ext_init(sb);
2909 err = ext4_mb_init(sb, needs_recovery);
2910 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002911 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2912 err);
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04002913 goto failed_mount4;
2914 }
2915
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002916 sbi->s_kobj.kset = ext4_kset;
2917 init_completion(&sbi->s_kobj_unregister);
2918 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2919 "%s", sb->s_id);
2920 if (err) {
2921 ext4_mb_release(sb);
2922 ext4_ext_release(sb);
2923 goto failed_mount4;
2924 };
2925
Mingming Cao617ba132006-10-11 01:20:53 -07002926 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2927 ext4_orphan_cleanup(sb, es);
2928 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05002929 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002930 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05002931 ext4_mark_recovery_complete(sb, es);
2932 }
2933 if (EXT4_SB(sb)->s_journal) {
2934 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2935 descr = " journalled data mode";
2936 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2937 descr = " ordered data mode";
2938 else
2939 descr = " writeback data mode";
2940 } else
2941 descr = "out journal";
2942
Eric Sandeenb31e1552009-06-04 17:36:36 -04002943 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002944
2945 lock_kernel();
2946 return 0;
2947
Mingming Cao617ba132006-10-11 01:20:53 -07002948cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002949 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002950 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002951 goto failed_mount;
2952
2953failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04002954 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04002955 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2956failed_mount_wq:
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04002957 ext4_release_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05002958 if (sbi->s_journal) {
2959 jbd2_journal_destroy(sbi->s_journal);
2960 sbi->s_journal = NULL;
2961 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002962failed_mount3:
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04002963 if (sbi->s_flex_groups) {
2964 if (is_vmalloc_addr(sbi->s_flex_groups))
2965 vfree(sbi->s_flex_groups);
2966 else
2967 kfree(sbi->s_flex_groups);
2968 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002969 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2970 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2971 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04002972 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002973failed_mount2:
2974 for (i = 0; i < db_count; i++)
2975 brelse(sbi->s_group_desc[i]);
2976 kfree(sbi->s_group_desc);
2977failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04002978 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04002979 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04002980 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002981#ifdef CONFIG_QUOTA
2982 for (i = 0; i < MAXQUOTAS; i++)
2983 kfree(sbi->s_qf_names[i]);
2984#endif
Mingming Cao617ba132006-10-11 01:20:53 -07002985 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002986 brelse(bh);
2987out_fail:
2988 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04002989 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002990 kfree(sbi);
2991 lock_kernel();
David Howells1d1fe1e2008-02-07 00:15:37 -08002992 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002993}
2994
2995/*
2996 * Setup any per-fs journal parameters now. We'll do this both on
2997 * initial mount, once the journal has been initialised but before we've
2998 * done any recovery; and again on any subsequent remount.
2999 */
Mingming Cao617ba132006-10-11 01:20:53 -07003000static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003001{
Mingming Cao617ba132006-10-11 01:20:53 -07003002 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003003
Theodore Ts'o30773842009-01-03 20:27:38 -05003004 journal->j_commit_interval = sbi->s_commit_interval;
3005 journal->j_min_batch_time = sbi->s_min_batch_time;
3006 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003007
3008 spin_lock(&journal->j_state_lock);
3009 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07003010 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003011 else
Mingming Caodab291a2006-10-11 01:21:01 -07003012 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04003013 if (test_opt(sb, DATA_ERR_ABORT))
3014 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3015 else
3016 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003017 spin_unlock(&journal->j_state_lock);
3018}
3019
Mingming Cao617ba132006-10-11 01:20:53 -07003020static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003021 unsigned int journal_inum)
3022{
3023 struct inode *journal_inode;
3024 journal_t *journal;
3025
Frank Mayhar03901312009-01-07 00:06:22 -05003026 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3027
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003028 /* First, test for the existence of a valid inode on disk. Bad
3029 * things happen if we iget() an unused inode, as the subsequent
3030 * iput() will try to delete it. */
3031
David Howells1d1fe1e2008-02-07 00:15:37 -08003032 journal_inode = ext4_iget(sb, journal_inum);
3033 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003034 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003035 return NULL;
3036 }
3037 if (!journal_inode->i_nlink) {
3038 make_bad_inode(journal_inode);
3039 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003040 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003041 return NULL;
3042 }
3043
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003044 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003045 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003046 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003047 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003048 iput(journal_inode);
3049 return NULL;
3050 }
3051
Mingming Caodab291a2006-10-11 01:21:01 -07003052 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003053 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003054 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003055 iput(journal_inode);
3056 return NULL;
3057 }
3058 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003059 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003060 return journal;
3061}
3062
Mingming Cao617ba132006-10-11 01:20:53 -07003063static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003064 dev_t j_dev)
3065{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003066 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003067 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003068 ext4_fsblk_t start;
3069 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003070 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003071 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003072 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003073 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003074 struct block_device *bdev;
3075
Frank Mayhar03901312009-01-07 00:06:22 -05003076 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3077
Eric Sandeenb31e1552009-06-04 17:36:36 -04003078 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003079 if (bdev == NULL)
3080 return NULL;
3081
3082 if (bd_claim(bdev, sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003083 ext4_msg(sb, KERN_ERR,
3084 "failed to claim external journal device");
Al Viro9a1c3542008-02-22 20:40:24 -05003085 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003086 return NULL;
3087 }
3088
3089 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003090 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003091 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003092 ext4_msg(sb, KERN_ERR,
3093 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003094 goto out_bdev;
3095 }
3096
Mingming Cao617ba132006-10-11 01:20:53 -07003097 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3098 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003099 set_blocksize(bdev, blocksize);
3100 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003101 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3102 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003103 goto out_bdev;
3104 }
3105
Mingming Cao617ba132006-10-11 01:20:53 -07003106 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3107 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003108 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003109 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003110 ext4_msg(sb, KERN_ERR, "external journal has "
3111 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003112 brelse(bh);
3113 goto out_bdev;
3114 }
3115
Mingming Cao617ba132006-10-11 01:20:53 -07003116 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003117 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003118 brelse(bh);
3119 goto out_bdev;
3120 }
3121
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003122 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003123 start = sb_block + 1;
3124 brelse(bh); /* we're done with the superblock */
3125
Mingming Caodab291a2006-10-11 01:21:01 -07003126 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003127 start, len, blocksize);
3128 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003129 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003130 goto out_bdev;
3131 }
3132 journal->j_private = sb;
3133 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3134 wait_on_buffer(journal->j_sb_buffer);
3135 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003136 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003137 goto out_journal;
3138 }
3139 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003140 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3141 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003142 be32_to_cpu(journal->j_superblock->s_nr_users));
3143 goto out_journal;
3144 }
Mingming Cao617ba132006-10-11 01:20:53 -07003145 EXT4_SB(sb)->journal_bdev = bdev;
3146 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003147 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003148
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003149out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07003150 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003151out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07003152 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003153 return NULL;
3154}
3155
Mingming Cao617ba132006-10-11 01:20:53 -07003156static int ext4_load_journal(struct super_block *sb,
3157 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003158 unsigned long journal_devnum)
3159{
3160 journal_t *journal;
3161 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3162 dev_t journal_dev;
3163 int err = 0;
3164 int really_read_only;
3165
Frank Mayhar03901312009-01-07 00:06:22 -05003166 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3167
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003168 if (journal_devnum &&
3169 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003170 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3171 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003172 journal_dev = new_decode_dev(journal_devnum);
3173 } else
3174 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3175
3176 really_read_only = bdev_read_only(sb->s_bdev);
3177
3178 /*
3179 * Are we loading a blank journal or performing recovery after a
3180 * crash? For recovery, we need to check in advance whether we
3181 * can get read-write access to the device.
3182 */
Mingming Cao617ba132006-10-11 01:20:53 -07003183 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003184 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003185 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3186 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003187 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003188 ext4_msg(sb, KERN_ERR, "write access "
3189 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003190 return -EROFS;
3191 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04003192 ext4_msg(sb, KERN_INFO, "write access will "
3193 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003194 }
3195 }
3196
3197 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003198 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3199 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003200 return -EINVAL;
3201 }
3202
3203 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07003204 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003205 return -EINVAL;
3206 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07003207 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003208 return -EINVAL;
3209 }
3210
Theodore Ts'o90576c02009-09-29 15:51:30 -04003211 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003212 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04003213
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003214 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07003215 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003216 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003217 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003218 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003219 return err;
3220 }
3221 }
3222
Mingming Cao617ba132006-10-11 01:20:53 -07003223 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07003224 err = jbd2_journal_wipe(journal, !really_read_only);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003225 if (!err)
Mingming Caodab291a2006-10-11 01:21:01 -07003226 err = jbd2_journal_load(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003227
3228 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003229 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003230 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003231 return err;
3232 }
3233
Mingming Cao617ba132006-10-11 01:20:53 -07003234 EXT4_SB(sb)->s_journal = journal;
3235 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003236
3237 if (journal_devnum &&
3238 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3239 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003240
3241 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003242 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003243 }
3244
3245 return 0;
3246}
3247
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003248static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003249{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003250 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07003251 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003252 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003253
3254 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08003255 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04003256 if (buffer_write_io_error(sbh)) {
3257 /*
3258 * Oh, dear. A previous attempt to write the
3259 * superblock failed. This could happen because the
3260 * USB device was yanked out. Or it could happen to
3261 * be a transient write error and maybe the block will
3262 * be remapped. Nothing we can do but to retry the
3263 * write and hope for the best.
3264 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04003265 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3266 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003267 clear_buffer_write_io_error(sbh);
3268 set_buffer_uptodate(sbh);
3269 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04003270 /*
3271 * If the file system is mounted read-only, don't update the
3272 * superblock write time. This avoids updating the superblock
3273 * write time when we are mounting the root file system
3274 * read/only but we need to replay the journal; at that point,
3275 * for people who are east of GMT and who make their clock
3276 * tick in localtime for Windows bug-for-bug compatibility,
3277 * the clock is set in the future, and this will cause e2fsck
3278 * to complain and force a full file system check.
3279 */
3280 if (!(sb->s_flags & MS_RDONLY))
3281 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003282 es->s_kbytes_written =
3283 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3284 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3285 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -05003286 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3287 &EXT4_SB(sb)->s_freeblocks_counter));
3288 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3289 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003290 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003291 BUFFER_TRACE(sbh, "marking dirty");
3292 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04003293 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08003294 error = sync_dirty_buffer(sbh);
3295 if (error)
3296 return error;
3297
3298 error = buffer_write_io_error(sbh);
3299 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003300 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3301 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003302 clear_buffer_write_io_error(sbh);
3303 set_buffer_uptodate(sbh);
3304 }
3305 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08003306 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003307}
3308
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003309/*
3310 * Have we just finished recovery? If so, and if we are mounting (or
3311 * remounting) the filesystem readonly, then we will end up with a
3312 * consistent fs on disk. Record that fact.
3313 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003314static void ext4_mark_recovery_complete(struct super_block *sb,
3315 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003316{
Mingming Cao617ba132006-10-11 01:20:53 -07003317 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003318
Frank Mayhar03901312009-01-07 00:06:22 -05003319 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3320 BUG_ON(journal != NULL);
3321 return;
3322 }
Mingming Caodab291a2006-10-11 01:21:01 -07003323 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003324 if (jbd2_journal_flush(journal) < 0)
3325 goto out;
3326
Mingming Cao617ba132006-10-11 01:20:53 -07003327 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003328 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07003329 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003330 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003331 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003332
3333out:
Mingming Caodab291a2006-10-11 01:21:01 -07003334 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003335}
3336
3337/*
3338 * If we are mounting (or read-write remounting) a filesystem whose journal
3339 * has recorded an error from a previous lifetime, move that error to the
3340 * main filesystem now.
3341 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003342static void ext4_clear_journal_err(struct super_block *sb,
3343 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003344{
3345 journal_t *journal;
3346 int j_errno;
3347 const char *errstr;
3348
Frank Mayhar03901312009-01-07 00:06:22 -05003349 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3350
Mingming Cao617ba132006-10-11 01:20:53 -07003351 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003352
3353 /*
3354 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07003355 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003356 */
3357
Mingming Caodab291a2006-10-11 01:21:01 -07003358 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003359 if (j_errno) {
3360 char nbuf[16];
3361
Mingming Cao617ba132006-10-11 01:20:53 -07003362 errstr = ext4_decode_error(sb, j_errno, nbuf);
Harvey Harrison46e665e2008-04-17 10:38:59 -04003363 ext4_warning(sb, __func__, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003364 "from previous mount: %s", errstr);
Harvey Harrison46e665e2008-04-17 10:38:59 -04003365 ext4_warning(sb, __func__, "Marking fs in need of "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003366 "filesystem check.");
3367
Mingming Cao617ba132006-10-11 01:20:53 -07003368 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3369 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003370 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003371
Mingming Caodab291a2006-10-11 01:21:01 -07003372 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003373 }
3374}
3375
3376/*
3377 * Force the running and committing transactions to commit,
3378 * and wait on the commit.
3379 */
Mingming Cao617ba132006-10-11 01:20:53 -07003380int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003381{
3382 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05003383 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003384
3385 if (sb->s_flags & MS_RDONLY)
3386 return 0;
3387
Mingming Cao617ba132006-10-11 01:20:53 -07003388 journal = EXT4_SB(sb)->s_journal;
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003389 if (journal)
Frank Mayhar03901312009-01-07 00:06:22 -05003390 ret = ext4_journal_force_commit(journal);
Frank Mayhar03901312009-01-07 00:06:22 -05003391
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003392 return ret;
3393}
3394
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003395static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003396{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02003397 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003398 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02003399 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003400}
3401
Mingming Cao617ba132006-10-11 01:20:53 -07003402static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003403{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05003404 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05003405 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003406 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003407
Theodore Ts'o9bffad12009-06-17 11:48:11 -04003408 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003409 flush_workqueue(sbi->dio_unwritten_wq);
3410 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003411 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04003412 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05003413 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05003414 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003415}
3416
3417/*
3418 * LVM calls this function before a (read-only) snapshot is created. This
3419 * gives us a chance to flush the journal completely and mark the fs clean.
3420 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08003421static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003422{
Takashi Satoc4be0c12009-01-09 16:40:58 -08003423 int error = 0;
3424 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003425
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003426 if (sb->s_flags & MS_RDONLY)
3427 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003428
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003429 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003430
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003431 /* Now we set up the journal barrier. */
3432 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003433
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003434 /*
3435 * Don't clear the needs_recovery flag if we failed to flush
3436 * the journal.
3437 */
3438 error = jbd2_journal_flush(journal);
3439 if (error < 0) {
3440 out:
3441 jbd2_journal_unlock_updates(journal);
3442 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003443 }
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003444
3445 /* Journal blocked and flushed, clear needs_recovery flag. */
3446 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3447 error = ext4_commit_super(sb, 1);
3448 if (error)
3449 goto out;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003450 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003451}
3452
3453/*
3454 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3455 * flag here, even though the filesystem is not technically dirty yet.
3456 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08003457static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003458{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003459 if (sb->s_flags & MS_RDONLY)
3460 return 0;
3461
3462 lock_super(sb);
3463 /* Reset the needs_recovery flag before the fs is unlocked. */
3464 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3465 ext4_commit_super(sb, 1);
3466 unlock_super(sb);
3467 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Takashi Satoc4be0c12009-01-09 16:40:58 -08003468 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003469}
3470
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003471static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003472{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003473 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07003474 struct ext4_sb_info *sbi = EXT4_SB(sb);
3475 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003476 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07003477 struct ext4_mount_options old_opts;
Theodore Ts'o8a266462008-07-26 14:34:21 -04003478 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003479 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003480 int err;
3481#ifdef CONFIG_QUOTA
3482 int i;
3483#endif
3484
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003485 lock_kernel();
3486
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003487 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04003488 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003489 old_sb_flags = sb->s_flags;
3490 old_opts.s_mount_opt = sbi->s_mount_opt;
3491 old_opts.s_resuid = sbi->s_resuid;
3492 old_opts.s_resgid = sbi->s_resgid;
3493 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05003494 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3495 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003496#ifdef CONFIG_QUOTA
3497 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3498 for (i = 0; i < MAXQUOTAS; i++)
3499 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3500#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003501 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3502 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003503
3504 /*
3505 * Allow the "check" option to be passed as a remount option.
3506 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003507 if (!parse_options(data, sb, NULL, &journal_ioprio,
3508 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003509 err = -EINVAL;
3510 goto restore_opts;
3511 }
3512
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04003513 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Harvey Harrison46e665e2008-04-17 10:38:59 -04003514 ext4_abort(sb, __func__, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003515
3516 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Mingming Cao617ba132006-10-11 01:20:53 -07003517 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003518
3519 es = sbi->s_es;
3520
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003521 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05003522 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003523 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3524 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003525
3526 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003527 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04003528 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003529 err = -EROFS;
3530 goto restore_opts;
3531 }
3532
3533 if (*flags & MS_RDONLY) {
3534 /*
3535 * First of all, the unconditional stuff we have to do
3536 * to disable replay of the journal when we next remount
3537 */
3538 sb->s_flags |= MS_RDONLY;
3539
3540 /*
3541 * OK, test if we are remounting a valid rw partition
3542 * readonly, and if so set the rdonly flag and then
3543 * mark the partition as valid again.
3544 */
Mingming Cao617ba132006-10-11 01:20:53 -07003545 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3546 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003547 es->s_state = cpu_to_le16(sbi->s_mount_state);
3548
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04003549 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05003550 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003551 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003552 /* Make sure we can mount this feature set readwrite */
3553 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003554 err = -EROFS;
3555 goto restore_opts;
3556 }
Eric Sandeenead65962007-02-10 01:46:08 -08003557 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04003558 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003559 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04003560 */
3561 for (g = 0; g < sbi->s_groups_count; g++) {
3562 struct ext4_group_desc *gdp =
3563 ext4_get_group_desc(sb, g, NULL);
3564
3565 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003566 ext4_msg(sb, KERN_ERR,
3567 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04003568 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3569 le16_to_cpu(gdp->bg_checksum));
3570 err = -EINVAL;
3571 goto restore_opts;
3572 }
3573 }
3574
3575 /*
Eric Sandeenead65962007-02-10 01:46:08 -08003576 * If we have an unprocessed orphan list hanging
3577 * around from a previously readonly bdev mount,
3578 * require a full umount/remount for now.
3579 */
3580 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003581 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08003582 "remount RDWR because of unprocessed "
3583 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003584 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08003585 err = -EINVAL;
3586 goto restore_opts;
3587 }
3588
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003589 /*
3590 * Mounting a RDONLY partition read-write, so reread
3591 * and store the current valid flag. (It may have
3592 * been changed by e2fsck since we originally mounted
3593 * the partition.)
3594 */
Frank Mayhar03901312009-01-07 00:06:22 -05003595 if (sbi->s_journal)
3596 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003597 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07003598 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003599 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003600 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003601 sb->s_flags &= ~MS_RDONLY;
3602 }
3603 }
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003604 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003605 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003606 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05003607
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003608#ifdef CONFIG_QUOTA
3609 /* Release old quota file names */
3610 for (i = 0; i < MAXQUOTAS; i++)
3611 if (old_opts.s_qf_names[i] &&
3612 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3613 kfree(old_opts.s_qf_names[i]);
3614#endif
Al Virobbd68512009-05-06 10:43:07 -04003615 unlock_super(sb);
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003616 unlock_kernel();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003617 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003618
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003619restore_opts:
3620 sb->s_flags = old_sb_flags;
3621 sbi->s_mount_opt = old_opts.s_mount_opt;
3622 sbi->s_resuid = old_opts.s_resuid;
3623 sbi->s_resgid = old_opts.s_resgid;
3624 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05003625 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3626 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003627#ifdef CONFIG_QUOTA
3628 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3629 for (i = 0; i < MAXQUOTAS; i++) {
3630 if (sbi->s_qf_names[i] &&
3631 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3632 kfree(sbi->s_qf_names[i]);
3633 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3634 }
3635#endif
Al Virobbd68512009-05-06 10:43:07 -04003636 unlock_super(sb);
Alessio Igor Bogani337eb002009-05-12 15:10:54 +02003637 unlock_kernel();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003638 return err;
3639}
3640
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003641static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003642{
3643 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003644 struct ext4_sb_info *sbi = EXT4_SB(sb);
3645 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08003646 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003647
Badari Pulavarty5e700302007-07-15 23:42:00 -07003648 if (test_opt(sb, MINIX_DF)) {
3649 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003650 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04003651 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07003652 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003653
3654 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07003655 * Compute the overhead (FS structures). This is constant
3656 * for a given filesystem unless the number of block groups
3657 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003658 */
3659
3660 /*
3661 * All of the blocks before first_data_block are
3662 * overhead
3663 */
3664 overhead = le32_to_cpu(es->s_first_data_block);
3665
3666 /*
3667 * Add the overhead attributed to the superblock and
3668 * block group descriptors. If the sparse superblocks
3669 * feature is turned on, then not all groups have this.
3670 */
3671 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07003672 overhead += ext4_bg_has_super(sb, i) +
3673 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003674 cond_resched();
3675 }
3676
3677 /*
3678 * Every block group has an inode bitmap, a block
3679 * bitmap, and an inode table.
3680 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07003681 overhead += ngroups * (2 + sbi->s_itb_per_group);
3682 sbi->s_overhead_last = overhead;
3683 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04003684 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003685 }
3686
Mingming Cao617ba132006-10-11 01:20:53 -07003687 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003688 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07003689 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04003690 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3691 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003692 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3693 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003694 buf->f_bavail = 0;
3695 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07003696 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07003697 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08003698 fsid = le64_to_cpup((void *)es->s_uuid) ^
3699 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3700 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3701 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003702
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003703 return 0;
3704}
3705
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003706/* Helper function for writing quotas on sync - we need to start transaction
3707 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003708 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07003709 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01003710 * jbd2_journal_start() write_dquot()
3711 * vfs_dq_init() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07003712 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003713 *
3714 */
3715
3716#ifdef CONFIG_QUOTA
3717
3718static inline struct inode *dquot_to_inode(struct dquot *dquot)
3719{
3720 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3721}
3722
Mingming Cao617ba132006-10-11 01:20:53 -07003723static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003724{
3725 int ret, err;
3726 handle_t *handle;
3727 struct inode *inode;
3728
3729 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003730 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003731 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003732 if (IS_ERR(handle))
3733 return PTR_ERR(handle);
3734 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003735 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003736 if (!ret)
3737 ret = err;
3738 return ret;
3739}
3740
Mingming Cao617ba132006-10-11 01:20:53 -07003741static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003742{
3743 int ret, err;
3744 handle_t *handle;
3745
Mingming Cao617ba132006-10-11 01:20:53 -07003746 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003747 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003748 if (IS_ERR(handle))
3749 return PTR_ERR(handle);
3750 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003751 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003752 if (!ret)
3753 ret = err;
3754 return ret;
3755}
3756
Mingming Cao617ba132006-10-11 01:20:53 -07003757static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003758{
3759 int ret, err;
3760 handle_t *handle;
3761
Mingming Cao617ba132006-10-11 01:20:53 -07003762 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003763 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07003764 if (IS_ERR(handle)) {
3765 /* Release dquot anyway to avoid endless cycle in dqput() */
3766 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003767 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07003768 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003769 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003770 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003771 if (!ret)
3772 ret = err;
3773 return ret;
3774}
3775
Mingming Cao617ba132006-10-11 01:20:53 -07003776static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003777{
Jan Kara2c8be6b2008-05-13 21:27:55 -04003778 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07003779 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3780 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003781 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07003782 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003783 } else {
3784 return dquot_mark_dquot_dirty(dquot);
3785 }
3786}
3787
Mingming Cao617ba132006-10-11 01:20:53 -07003788static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003789{
3790 int ret, err;
3791 handle_t *handle;
3792
3793 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07003794 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003795 if (IS_ERR(handle))
3796 return PTR_ERR(handle);
3797 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07003798 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003799 if (!ret)
3800 ret = err;
3801 return ret;
3802}
3803
3804/*
3805 * Turn on quotas during mount time - we need to find
3806 * the quota file and such...
3807 */
Mingming Cao617ba132006-10-11 01:20:53 -07003808static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003809{
Mingming Cao617ba132006-10-11 01:20:53 -07003810 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003811 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003812}
3813
3814/*
3815 * Standard function to be called on quota_on
3816 */
Mingming Cao617ba132006-10-11 01:20:53 -07003817static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Al Viro82646132008-08-02 00:57:06 -04003818 char *name, int remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003819{
3820 int err;
Al Viro82646132008-08-02 00:57:06 -04003821 struct path path;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003822
3823 if (!test_opt(sb, QUOTA))
3824 return -EINVAL;
Al Viro82646132008-08-02 00:57:06 -04003825 /* When remounting, no checks are needed and in fact, name is NULL */
Jan Kara06235432008-05-13 19:11:51 -04003826 if (remount)
Al Viro82646132008-08-02 00:57:06 -04003827 return vfs_quota_on(sb, type, format_id, name, remount);
Jan Kara06235432008-05-13 19:11:51 -04003828
Al Viro82646132008-08-02 00:57:06 -04003829 err = kern_path(name, LOOKUP_FOLLOW, &path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003830 if (err)
3831 return err;
Jan Kara06235432008-05-13 19:11:51 -04003832
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003833 /* Quotafile not on the same filesystem? */
Al Viro82646132008-08-02 00:57:06 -04003834 if (path.mnt->mnt_sb != sb) {
3835 path_put(&path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003836 return -EXDEV;
3837 }
Jan Kara06235432008-05-13 19:11:51 -04003838 /* Journaling quota? */
3839 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003840 /* Quotafile not in fs root? */
Al Viro82646132008-08-02 00:57:06 -04003841 if (path.dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003842 ext4_msg(sb, KERN_WARNING,
3843 "Quota file not on filesystem root. "
3844 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003845 }
Jan Kara06235432008-05-13 19:11:51 -04003846
3847 /*
3848 * When we journal data on quota file, we have to flush journal to see
3849 * all updates to the file when we bypass pagecache...
3850 */
Frank Mayhar03901312009-01-07 00:06:22 -05003851 if (EXT4_SB(sb)->s_journal &&
3852 ext4_should_journal_data(path.dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04003853 /*
3854 * We don't need to lock updates but journal_flush() could
3855 * otherwise be livelocked...
3856 */
3857 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003858 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04003859 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003860 if (err) {
Al Viro82646132008-08-02 00:57:06 -04003861 path_put(&path);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04003862 return err;
3863 }
Jan Kara06235432008-05-13 19:11:51 -04003864 }
3865
Al Viro82646132008-08-02 00:57:06 -04003866 err = vfs_quota_on_path(sb, type, format_id, &path);
3867 path_put(&path);
Al Viro77e69da2008-08-01 04:29:18 -04003868 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003869}
3870
3871/* Read data from quotafile - avoid pagecache and such because we cannot afford
3872 * acquiring the locks... As quota files are never truncated and quota code
3873 * itself serializes the operations (and noone else should touch the files)
3874 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07003875static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003876 size_t len, loff_t off)
3877{
3878 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003879 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003880 int err = 0;
3881 int offset = off & (sb->s_blocksize - 1);
3882 int tocopy;
3883 size_t toread;
3884 struct buffer_head *bh;
3885 loff_t i_size = i_size_read(inode);
3886
3887 if (off > i_size)
3888 return 0;
3889 if (off+len > i_size)
3890 len = i_size-off;
3891 toread = len;
3892 while (toread > 0) {
3893 tocopy = sb->s_blocksize - offset < toread ?
3894 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07003895 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003896 if (err)
3897 return err;
3898 if (!bh) /* A hole? */
3899 memset(data, 0, tocopy);
3900 else
3901 memcpy(data, bh->b_data+offset, tocopy);
3902 brelse(bh);
3903 offset = 0;
3904 toread -= tocopy;
3905 data += tocopy;
3906 blk++;
3907 }
3908 return len;
3909}
3910
3911/* Write to quotafile (we know the transaction is already started and has
3912 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07003913static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003914 const char *data, size_t len, loff_t off)
3915{
3916 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05003917 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003918 int err = 0;
3919 int offset = off & (sb->s_blocksize - 1);
3920 int tocopy;
Mingming Cao617ba132006-10-11 01:20:53 -07003921 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003922 size_t towrite = len;
3923 struct buffer_head *bh;
3924 handle_t *handle = journal_current_handle();
3925
Frank Mayhar03901312009-01-07 00:06:22 -05003926 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003927 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3928 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07003929 (unsigned long long)off, (unsigned long long)len);
3930 return -EIO;
3931 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003932 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3933 while (towrite > 0) {
3934 tocopy = sb->s_blocksize - offset < towrite ?
3935 sb->s_blocksize - offset : towrite;
Mingming Cao617ba132006-10-11 01:20:53 -07003936 bh = ext4_bread(handle, inode, blk, 1, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003937 if (!bh)
3938 goto out;
3939 if (journal_quota) {
Mingming Cao617ba132006-10-11 01:20:53 -07003940 err = ext4_journal_get_write_access(handle, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003941 if (err) {
3942 brelse(bh);
3943 goto out;
3944 }
3945 }
3946 lock_buffer(bh);
3947 memcpy(bh->b_data+offset, data, tocopy);
3948 flush_dcache_page(bh->b_page);
3949 unlock_buffer(bh);
3950 if (journal_quota)
Frank Mayhar03901312009-01-07 00:06:22 -05003951 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003952 else {
3953 /* Always do at least ordered writes for quotas */
Jan Kara678aaf42008-07-11 19:27:31 -04003954 err = ext4_jbd2_file_inode(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003955 mark_buffer_dirty(bh);
3956 }
3957 brelse(bh);
3958 if (err)
3959 goto out;
3960 offset = 0;
3961 towrite -= tocopy;
3962 data += tocopy;
3963 blk++;
3964 }
3965out:
Jan Kara4d04e4f2008-07-04 09:59:34 -07003966 if (len == towrite) {
3967 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003968 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07003969 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003970 if (inode->i_size < off+len-towrite) {
3971 i_size_write(inode, off+len-towrite);
Mingming Cao617ba132006-10-11 01:20:53 -07003972 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003973 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003974 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07003975 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003976 mutex_unlock(&inode->i_mutex);
3977 return len - towrite;
3978}
3979
3980#endif
3981
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003982static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3983 const char *dev_name, void *data, struct vfsmount *mnt)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003984{
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003985 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003986}
3987
Theodore Ts'oa2142382009-12-09 21:09:58 -05003988#if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05003989static struct file_system_type ext2_fs_type = {
3990 .owner = THIS_MODULE,
3991 .name = "ext2",
3992 .get_sb = ext4_get_sb,
3993 .kill_sb = kill_block_super,
3994 .fs_flags = FS_REQUIRES_DEV,
3995};
3996
3997static inline void register_as_ext2(void)
3998{
3999 int err = register_filesystem(&ext2_fs_type);
4000 if (err)
4001 printk(KERN_WARNING
4002 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4003}
4004
4005static inline void unregister_as_ext2(void)
4006{
4007 unregister_filesystem(&ext2_fs_type);
4008}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004009MODULE_ALIAS("ext2");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004010#else
4011static inline void register_as_ext2(void) { }
4012static inline void unregister_as_ext2(void) { }
4013#endif
4014
Theodore Ts'oa2142382009-12-09 21:09:58 -05004015#if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004016static struct file_system_type ext3_fs_type = {
4017 .owner = THIS_MODULE,
4018 .name = "ext3",
4019 .get_sb = ext4_get_sb,
4020 .kill_sb = kill_block_super,
4021 .fs_flags = FS_REQUIRES_DEV,
4022};
4023
4024static inline void register_as_ext3(void)
4025{
4026 int err = register_filesystem(&ext3_fs_type);
4027 if (err)
4028 printk(KERN_WARNING
4029 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4030}
4031
4032static inline void unregister_as_ext3(void)
4033{
4034 unregister_filesystem(&ext3_fs_type);
4035}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004036MODULE_ALIAS("ext3");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004037#else
4038static inline void register_as_ext3(void) { }
4039static inline void unregister_as_ext3(void) { }
4040#endif
4041
Theodore Ts'o03010a32008-10-10 20:02:48 -04004042static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004043 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04004044 .name = "ext4",
Mingming Cao617ba132006-10-11 01:20:53 -07004045 .get_sb = ext4_get_sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004046 .kill_sb = kill_block_super,
4047 .fs_flags = FS_REQUIRES_DEV,
4048};
4049
Mingming Cao617ba132006-10-11 01:20:53 -07004050static int __init init_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004051{
Alex Tomasc9de5602008-01-29 00:19:52 -05004052 int err;
4053
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004054 err = init_ext4_system_zone();
4055 if (err)
4056 return err;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004057 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4058 if (!ext4_kset)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004059 goto out4;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004060 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Alex Tomasc9de5602008-01-29 00:19:52 -05004061 err = init_ext4_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004062 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004063 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05004064
4065 err = init_ext4_xattr();
4066 if (err)
4067 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004068 err = init_inodecache();
4069 if (err)
4070 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05004071 register_as_ext2();
4072 register_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004073 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004074 if (err)
4075 goto out;
4076 return 0;
4077out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05004078 unregister_as_ext2();
4079 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004080 destroy_inodecache();
4081out1:
Mingming Cao617ba132006-10-11 01:20:53 -07004082 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004083out2:
4084 exit_ext4_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004085out3:
4086 remove_proc_entry("fs/ext4", NULL);
4087 kset_unregister(ext4_kset);
4088out4:
4089 exit_ext4_system_zone();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004090 return err;
4091}
4092
Mingming Cao617ba132006-10-11 01:20:53 -07004093static void __exit exit_ext4_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004094{
Theodore Ts'o24b58422009-12-07 14:08:51 -05004095 unregister_as_ext2();
4096 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004097 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004098 destroy_inodecache();
Mingming Cao617ba132006-10-11 01:20:53 -07004099 exit_ext4_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004100 exit_ext4_mballoc();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004101 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004102 kset_unregister(ext4_kset);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004103 exit_ext4_system_zone();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004104}
4105
4106MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05004107MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004108MODULE_LICENSE("GPL");
Mingming Cao617ba132006-10-11 01:20:53 -07004109module_init(init_ext4_fs)
4110module_exit(exit_ext4_fs)