blob: 1539cf5597847cefe9f1d10079464a4fcfa2dc7f [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040023#include <linux/vmalloc.h>
Mingming Caodab291a2006-10-11 01:21:01 -070024#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070025#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070030#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040037#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040038#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040039#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040040#include <linux/crc16.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070041#include <asm/uaccess.h>
42
Lukas Czernerbfff6872010-10-27 21:30:05 -040043#include <linux/kthread.h>
44#include <linux/freezer.h>
45
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040046#include "ext4.h"
47#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070048#include "xattr.h"
49#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040050#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070051
Theodore Ts'o9bffad12009-06-17 11:48:11 -040052#define CREATE_TRACE_POINTS
53#include <trace/events/ext4.h>
54
Theodore Ts'o1f109d52010-10-27 21:30:14 -040055static struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040056static struct kset *ext4_kset;
Lukas Czerner0b75a842011-02-23 12:22:49 -050057static struct ext4_lazy_init *ext4_li_info;
58static struct mutex ext4_li_mtx;
59static struct ext4_features *ext4_feat;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040060
Mingming Cao617ba132006-10-11 01:20:53 -070061static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070062 unsigned long journal_devnum);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040063static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040064static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070068static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070070 char nbuf[16]);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040071static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080073static int ext4_unfreeze(struct super_block *sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040074static void ext4_write_super(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080075static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040076static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
Lukas Czernerbfff6872010-10-27 21:30:05 -040078static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050080static void ext4_clear_request_list(void);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070081
Jan Karaba69f9a2010-03-24 20:18:37 -040082#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
83static struct file_system_type ext3_fs_type = {
84 .owner = THIS_MODULE,
85 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +040086 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -040087 .kill_sb = kill_block_super,
88 .fs_flags = FS_REQUIRES_DEV,
89};
90#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
91#else
92#define IS_EXT3_SB(sb) (0)
93#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070094
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070095ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070097{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040098 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070099 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400100 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700101}
102
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700103ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
104 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700105{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400106 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700109}
110
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700111ext4_fsblk_t ext4_inode_table(struct super_block *sb,
112 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700113{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400114 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400116 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700117}
118
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500119__u32 ext4_free_blks_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400124 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500125}
126
127__u32 ext4_free_inodes_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400132 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500133}
134
135__u32 ext4_used_dirs_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400140 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500141}
142
143__u32 ext4_itable_unused_count(struct super_block *sb,
144 struct ext4_group_desc *bg)
145{
146 return le16_to_cpu(bg->bg_itable_unused_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400148 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500149}
150
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700151void ext4_block_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700153{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400154 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700157}
158
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700159void ext4_inode_bitmap_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700161{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400162 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700165}
166
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700167void ext4_inode_table_set(struct super_block *sb,
168 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700169{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400170 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700173}
174
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500175void ext4_free_blks_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_free_inodes_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_used_dirs_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
197}
198
199void ext4_itable_unused_set(struct super_block *sb,
200 struct ext4_group_desc *bg, __u32 count)
201{
202 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
205}
206
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400207
208/* Just increment the non-pointer handle value */
209static handle_t *ext4_get_nojournal(void)
210{
211 handle_t *handle = current->journal_info;
212 unsigned long ref_cnt = (unsigned long)handle;
213
214 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
215
216 ref_cnt++;
217 handle = (handle_t *)ref_cnt;
218
219 current->journal_info = handle;
220 return handle;
221}
222
223
224/* Decrement the non-pointer handle value */
225static void ext4_put_nojournal(handle_t *handle)
226{
227 unsigned long ref_cnt = (unsigned long)handle;
228
229 BUG_ON(ref_cnt == 0);
230
231 ref_cnt--;
232 handle = (handle_t *)ref_cnt;
233
234 current->journal_info = handle;
235}
236
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700237/*
Mingming Caodab291a2006-10-11 01:21:01 -0700238 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700239 *
240 * The only special thing we need to do here is to make sure that all
241 * journal_end calls result in the superblock being marked dirty, so
242 * that sync() will call the filesystem's write_super callback if
243 * appropriate.
244 */
Mingming Cao617ba132006-10-11 01:20:53 -0700245handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700246{
247 journal_t *journal;
248
249 if (sb->s_flags & MS_RDONLY)
250 return ERR_PTR(-EROFS);
251
Eric Sandeen437f88c2010-08-01 17:33:29 -0400252 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700253 /* Special case here: if the journal has aborted behind our
254 * backs (eg. EIO in the commit thread), then we still need to
255 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700256 journal = EXT4_SB(sb)->s_journal;
Frank Mayhar03901312009-01-07 00:06:22 -0500257 if (journal) {
258 if (is_journal_aborted(journal)) {
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400259 ext4_abort(sb, "Detected aborted journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500260 return ERR_PTR(-EROFS);
261 }
262 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700263 }
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400264 return ext4_get_nojournal();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700265}
266
267/*
268 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700269 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700270 * that sync() will call the filesystem's write_super callback if
271 * appropriate.
272 */
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400273int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700274{
275 struct super_block *sb;
276 int err;
277 int rc;
278
Frank Mayhar03901312009-01-07 00:06:22 -0500279 if (!ext4_handle_valid(handle)) {
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400280 ext4_put_nojournal(handle);
Frank Mayhar03901312009-01-07 00:06:22 -0500281 return 0;
282 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700283 sb = handle->h_transaction->t_journal->j_private;
284 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700285 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700286
287 if (!err)
288 err = rc;
289 if (err)
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400290 __ext4_std_error(sb, where, line, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700291 return err;
292}
293
Theodore Ts'o90c72012010-06-29 14:53:24 -0400294void ext4_journal_abort_handle(const char *caller, unsigned int line,
295 const char *err_fn, struct buffer_head *bh,
296 handle_t *handle, int err)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700297{
298 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700299 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700300
Frank Mayhar03901312009-01-07 00:06:22 -0500301 BUG_ON(!ext4_handle_valid(handle));
302
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700303 if (bh)
304 BUFFER_TRACE(bh, "abort");
305
306 if (!handle->h_err)
307 handle->h_err = err;
308
309 if (is_handle_aborted(handle))
310 return;
311
Theodore Ts'o90c72012010-06-29 14:53:24 -0400312 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
313 caller, line, errstr, err_fn);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700314
Mingming Caodab291a2006-10-11 01:21:01 -0700315 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700316}
317
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400318static void __save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322
323 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
335 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400336 /*
337 * Start the daily error reporting function if it hasn't been
338 * started already
339 */
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400342 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
343}
344
345static void save_error_info(struct super_block *sb, const char *func,
346 unsigned int line)
347{
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
350}
351
352
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700353/* Deal with the reporting of failure conditions on a filesystem such as
354 * inconsistencies detected or read IO failures.
355 *
356 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700357 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700358 * write ordering constraints on the superblock which prevent us from
359 * writing it out straight away; and given that the journal is about to
360 * be aborted, we can't rely on the current, or future, transactions to
361 * write out the superblock safely.
362 *
Mingming Caodab291a2006-10-11 01:21:01 -0700363 * We'll just use the jbd2_journal_abort() error code to record an error in
Thadeu Lima de Souza Cascardod6b198b2010-01-17 19:10:07 -0200364 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700365 * that error until we've noted it down and cleared it.
366 */
367
Mingming Cao617ba132006-10-11 01:20:53 -0700368static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700369{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700370 if (sb->s_flags & MS_RDONLY)
371 return;
372
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400373 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700374 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700375
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400376 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700377 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700378 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700379 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400380 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400381 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700382 sb->s_flags |= MS_RDONLY;
383 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700384 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700385 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700386 sb->s_id);
387}
388
Eric Sandeen12062dd2010-02-15 14:19:27 -0500389void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400390 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700391{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500392 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700393 va_list args;
394
395 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500396 vaf.fmt = fmt;
397 vaf.va = &args;
398 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
399 sb->s_id, function, line, current->comm, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700400 va_end(args);
401
Mingming Cao617ba132006-10-11 01:20:53 -0700402 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700403}
404
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400405void ext4_error_inode(struct inode *inode, const char *function,
406 unsigned int line, ext4_fsblk_t block,
Frank Mayhar273df552010-03-02 11:46:09 -0500407 const char *fmt, ...)
408{
409 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500410 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400411 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500412
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400413 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
414 es->s_last_error_block = cpu_to_le64(block);
415 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500416 va_start(args, fmt);
Theodore Ts'of7c21172011-01-10 12:10:55 -0500417 vaf.fmt = fmt;
418 vaf.va = &args;
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400419 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
420 inode->i_sb->s_id, function, line, inode->i_ino);
421 if (block)
Theodore Ts'of7c21172011-01-10 12:10:55 -0500422 printk(KERN_CONT "block %llu: ", block);
423 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
Frank Mayhar273df552010-03-02 11:46:09 -0500424 va_end(args);
425
426 ext4_handle_error(inode->i_sb);
427}
428
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400429void ext4_error_file(struct file *file, const char *function,
Theodore Ts'of7c21172011-01-10 12:10:55 -0500430 unsigned int line, ext4_fsblk_t block,
431 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500432{
433 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500434 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400435 struct ext4_super_block *es;
Frank Mayhar273df552010-03-02 11:46:09 -0500436 struct inode *inode = file->f_dentry->d_inode;
437 char pathname[80], *path;
438
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400439 es = EXT4_SB(inode->i_sb)->s_es;
440 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
441 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500442 path = d_path(&(file->f_path), pathname, sizeof(pathname));
Dan Carpenterf9a62d02011-01-10 12:10:50 -0500443 if (IS_ERR(path))
Frank Mayhar273df552010-03-02 11:46:09 -0500444 path = "(unknown)";
445 printk(KERN_CRIT
Theodore Ts'of7c21172011-01-10 12:10:55 -0500446 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
447 inode->i_sb->s_id, function, line, inode->i_ino);
448 if (block)
449 printk(KERN_CONT "block %llu: ", block);
450 va_start(args, fmt);
451 vaf.fmt = fmt;
452 vaf.va = &args;
453 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
Frank Mayhar273df552010-03-02 11:46:09 -0500454 va_end(args);
455
456 ext4_handle_error(inode->i_sb);
457}
458
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400459static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700460 char nbuf[16])
461{
462 char *errstr = NULL;
463
464 switch (errno) {
465 case -EIO:
466 errstr = "IO failure";
467 break;
468 case -ENOMEM:
469 errstr = "Out of memory";
470 break;
471 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400472 if (!sb || (EXT4_SB(sb)->s_journal &&
473 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700474 errstr = "Journal has aborted";
475 else
476 errstr = "Readonly filesystem";
477 break;
478 default:
479 /* If the caller passed in an extra buffer for unknown
480 * errors, textualise them now. Else we just return
481 * NULL. */
482 if (nbuf) {
483 /* Check for truncated error codes... */
484 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
485 errstr = nbuf;
486 }
487 break;
488 }
489
490 return errstr;
491}
492
Mingming Cao617ba132006-10-11 01:20:53 -0700493/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700494 * automatically and invokes the appropriate error response. */
495
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400496void __ext4_std_error(struct super_block *sb, const char *function,
497 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700498{
499 char nbuf[16];
500 const char *errstr;
501
502 /* Special case: if the error is EROFS, and we're not already
503 * inside a transaction, then there's really no point in logging
504 * an error. */
505 if (errno == -EROFS && journal_current_handle() == NULL &&
506 (sb->s_flags & MS_RDONLY))
507 return;
508
Mingming Cao617ba132006-10-11 01:20:53 -0700509 errstr = ext4_decode_error(sb, errno, nbuf);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400510 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
511 sb->s_id, function, line, errstr);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400512 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700513
Mingming Cao617ba132006-10-11 01:20:53 -0700514 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700515}
516
517/*
Mingming Cao617ba132006-10-11 01:20:53 -0700518 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700519 * abort function may be used to deal with unrecoverable failures such
520 * as journal IO errors or ENOMEM at a critical moment in log management.
521 *
522 * We unconditionally force the filesystem into an ABORT|READONLY state,
523 * unless the error response on the fs has been set to panic in which
524 * case we take the easy way out and panic immediately.
525 */
526
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400527void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400528 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700529{
530 va_list args;
531
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400532 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700533 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
535 function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700536 vprintk(fmt, args);
537 printk("\n");
538 va_end(args);
539
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400540 if ((sb->s_flags & MS_RDONLY) == 0) {
541 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
542 sb->s_flags |= MS_RDONLY;
543 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
544 if (EXT4_SB(sb)->s_journal)
545 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
546 save_error_info(sb, function, line);
547 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700548 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700549 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700550}
551
Joe Perches0ff2ea72010-12-19 22:43:19 -0500552void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400553{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500554 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400555 va_list args;
556
557 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500558 vaf.fmt = fmt;
559 vaf.va = &args;
560 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400561 va_end(args);
562}
563
Eric Sandeen12062dd2010-02-15 14:19:27 -0500564void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400565 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700566{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500567 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700568 va_list args;
569
570 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500571 vaf.fmt = fmt;
572 vaf.va = &args;
573 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
574 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700575 va_end(args);
576}
577
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400578void __ext4_grp_locked_error(const char *function, unsigned int line,
579 struct super_block *sb, ext4_group_t grp,
580 unsigned long ino, ext4_fsblk_t block,
581 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500582__releases(bitlock)
583__acquires(bitlock)
584{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500585 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500586 va_list args;
587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
588
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400589 es->s_last_error_ino = cpu_to_le32(ino);
590 es->s_last_error_block = cpu_to_le64(block);
591 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500592
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500593 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500594
595 vaf.fmt = fmt;
596 vaf.va = &args;
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400597 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
598 sb->s_id, function, line, grp);
599 if (ino)
Joe Perches0ff2ea72010-12-19 22:43:19 -0500600 printk(KERN_CONT "inode %lu: ", ino);
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400601 if (block)
Joe Perches0ff2ea72010-12-19 22:43:19 -0500602 printk(KERN_CONT "block %llu:", (unsigned long long) block);
603 printk(KERN_CONT "%pV\n", &vaf);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500604 va_end(args);
605
606 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400607 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500608 return;
609 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400610
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500611 ext4_unlock_group(sb, grp);
612 ext4_handle_error(sb);
613 /*
614 * We only get here in the ERRORS_RO case; relocking the group
615 * may be dangerous, but nothing bad will happen since the
616 * filesystem will have already been marked read/only and the
617 * journal has been aborted. We return 1 as a hint to callers
618 * who might what to use the return value from
619 * ext4_grp_locked_error() to distinguish beween the
620 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
621 * aggressively from the ext4 function in question, with a
622 * more appropriate error code.
623 */
624 ext4_lock_group(sb, grp);
625 return;
626}
627
Mingming Cao617ba132006-10-11 01:20:53 -0700628void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700629{
Mingming Cao617ba132006-10-11 01:20:53 -0700630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700631
Mingming Cao617ba132006-10-11 01:20:53 -0700632 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700633 return;
634
Eric Sandeen12062dd2010-02-15 14:19:27 -0500635 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700636 "updating to rev %d because of new feature flag, "
637 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700638 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700639
Mingming Cao617ba132006-10-11 01:20:53 -0700640 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
641 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
642 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700643 /* leave es->s_feature_*compat flags alone */
644 /* es->s_uuid will be set by e2fsck if empty */
645
646 /*
647 * The rest of the superblock fields should be zero, and if not it
648 * means they are likely already in use, so leave them alone. We
649 * can leave it up to e2fsck to clean up any inconsistencies there.
650 */
651}
652
653/*
654 * Open the external journal device
655 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400656static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700657{
658 struct block_device *bdev;
659 char b[BDEVNAME_SIZE];
660
Tejun Heod4d77622010-11-13 11:55:18 +0100661 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700662 if (IS_ERR(bdev))
663 goto fail;
664 return bdev;
665
666fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400667 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700668 __bdevname(dev, b), PTR_ERR(bdev));
669 return NULL;
670}
671
672/*
673 * Release the journal device
674 */
Mingming Cao617ba132006-10-11 01:20:53 -0700675static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700676{
Tejun Heoe525fd82010-11-13 11:55:17 +0100677 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700678}
679
Mingming Cao617ba132006-10-11 01:20:53 -0700680static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700681{
682 struct block_device *bdev;
683 int ret = -ENODEV;
684
685 bdev = sbi->journal_bdev;
686 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700687 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700688 sbi->journal_bdev = NULL;
689 }
690 return ret;
691}
692
693static inline struct inode *orphan_list_entry(struct list_head *l)
694{
Mingming Cao617ba132006-10-11 01:20:53 -0700695 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700696}
697
Mingming Cao617ba132006-10-11 01:20:53 -0700698static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700699{
700 struct list_head *l;
701
Eric Sandeenb31e1552009-06-04 17:36:36 -0400702 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
703 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700704
705 printk(KERN_ERR "sb_info orphan list:\n");
706 list_for_each(l, &sbi->s_orphan) {
707 struct inode *inode = orphan_list_entry(l);
708 printk(KERN_ERR " "
709 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
710 inode->i_sb->s_id, inode->i_ino, inode,
711 inode->i_mode, inode->i_nlink,
712 NEXT_ORPHAN(inode));
713 }
714}
715
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400716static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700717{
Mingming Cao617ba132006-10-11 01:20:53 -0700718 struct ext4_sb_info *sbi = EXT4_SB(sb);
719 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400720 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700721
Lukas Czerner857ac882010-10-27 21:30:05 -0400722 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400723 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
724
Mingming Cao4c0425f2009-09-28 15:48:41 -0400725 flush_workqueue(sbi->dio_unwritten_wq);
726 destroy_workqueue(sbi->dio_unwritten_wq);
727
Al Viroa9e220f2009-05-05 22:10:44 -0400728 lock_super(sb);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200729 if (sb->s_dirt)
Christoph Hellwigebc1ac12009-05-11 23:35:03 +0200730 ext4_commit_super(sb, 1);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200731
Frank Mayhar03901312009-01-07 00:06:22 -0500732 if (sbi->s_journal) {
733 err = jbd2_journal_destroy(sbi->s_journal);
734 sbi->s_journal = NULL;
735 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400736 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500737 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500738
Sergey Senozhatskya1c6c562010-10-27 21:30:04 -0400739 del_timer(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500740 ext4_release_system_zone(sb);
741 ext4_mb_release(sb);
742 ext4_ext_release(sb);
743 ext4_xattr_put_super(sb);
744
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700745 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700746 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700747 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400748 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700749 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400750 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400751 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400752 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400753 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700754
755 for (i = 0; i < sbi->s_gdb_count; i++)
756 brelse(sbi->s_group_desc[i]);
757 kfree(sbi->s_group_desc);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -0400758 if (is_vmalloc_addr(sbi->s_flex_groups))
759 vfree(sbi->s_flex_groups);
760 else
761 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700762 percpu_counter_destroy(&sbi->s_freeblocks_counter);
763 percpu_counter_destroy(&sbi->s_freeinodes_counter);
764 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400765 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700766 brelse(sbi->s_sbh);
767#ifdef CONFIG_QUOTA
768 for (i = 0; i < MAXQUOTAS; i++)
769 kfree(sbi->s_qf_names[i]);
770#endif
771
772 /* Debugging code just in case the in-memory inode orphan list
773 * isn't empty. The on-disk one can be non-empty if we've
774 * detected an error and taken the fs readonly, but the
775 * in-memory list had better be clean by this point. */
776 if (!list_empty(&sbi->s_orphan))
777 dump_orphan_list(sb, sbi);
778 J_ASSERT(list_empty(&sbi->s_orphan));
779
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700780 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700781 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
782 /*
783 * Invalidate the journal device's buffers. We don't want them
784 * floating about in memory - the physical journal device may
785 * hotswapped, and it breaks the `ro-after' testing code.
786 */
787 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700788 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700789 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700790 }
791 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400792 /*
793 * Now that we are completely done shutting down the
794 * superblock, we need to actually destroy the kobject.
795 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400796 unlock_super(sb);
797 kobject_put(&sbi->s_kobj);
798 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500799 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700800 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700801}
802
Christoph Lametere18b8902006-12-06 20:33:20 -0800803static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700804
805/*
806 * Called inside transaction, so use GFP_NOFS
807 */
Mingming Cao617ba132006-10-11 01:20:53 -0700808static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700809{
Mingming Cao617ba132006-10-11 01:20:53 -0700810 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700811
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800812 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700813 if (!ei)
814 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400815
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700816 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400817 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700818 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500819 INIT_LIST_HEAD(&ei->i_prealloc_list);
820 spin_lock_init(&ei->i_prealloc_lock);
Mingming Caod2a17632008-07-14 17:52:37 -0400821 ei->i_reserved_data_blocks = 0;
822 ei->i_reserved_meta_blocks = 0;
823 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500824 ei->i_da_metadata_calc_len = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400825 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300826#ifdef CONFIG_QUOTA
827 ei->i_reserved_quota = 0;
828#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500829 ei->jinode = NULL;
Jiaying Zhangc7064ef2010-03-02 13:28:44 -0500830 INIT_LIST_HEAD(&ei->i_completed_io_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500831 spin_lock_init(&ei->i_completed_io_lock);
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400832 ei->cur_aio_dio = NULL;
Jan Karab436b9b2009-12-08 23:51:10 -0500833 ei->i_sync_tid = 0;
834 ei->i_datasync_tid = 0;
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500835 atomic_set(&ei->i_ioend_count, 0);
Eric Sandeene9e3bce2011-02-12 08:17:34 -0500836 atomic_set(&ei->i_aiodio_unwritten, 0);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400837
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700838 return &ei->vfs_inode;
839}
840
Theodore Ts'o7ff9c072010-11-08 13:51:33 -0500841static int ext4_drop_inode(struct inode *inode)
842{
843 int drop = generic_drop_inode(inode);
844
845 trace_ext4_drop_inode(inode, drop);
846 return drop;
847}
848
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100849static void ext4_i_callback(struct rcu_head *head)
850{
851 struct inode *inode = container_of(head, struct inode, i_rcu);
852 INIT_LIST_HEAD(&inode->i_dentry);
853 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
854}
855
Mingming Cao617ba132006-10-11 01:20:53 -0700856static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700857{
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500858 ext4_ioend_wait(inode);
Vasily Averin9f7dd932007-07-15 23:40:45 -0700859 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400860 ext4_msg(inode->i_sb, KERN_ERR,
861 "Inode %lu (%p): orphan list check failed!",
862 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700863 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
864 EXT4_I(inode), sizeof(struct ext4_inode_info),
865 true);
866 dump_stack();
867 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100868 call_rcu(&inode->i_rcu, ext4_i_callback);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700869}
870
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700871static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700872{
Mingming Cao617ba132006-10-11 01:20:53 -0700873 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700874
Christoph Lametera35afb82007-05-16 22:10:57 -0700875 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400876#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700877 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700878#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500879 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700880 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700881}
882
883static int init_inodecache(void)
884{
Mingming Cao617ba132006-10-11 01:20:53 -0700885 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
886 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700887 0, (SLAB_RECLAIM_ACCOUNT|
888 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900889 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700890 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700891 return -ENOMEM;
892 return 0;
893}
894
895static void destroy_inodecache(void)
896{
Mingming Cao617ba132006-10-11 01:20:53 -0700897 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700898}
899
Al Viro0930fcc2010-06-07 13:16:22 -0400900void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700901{
Al Viro0930fcc2010-06-07 13:16:22 -0400902 invalidate_inode_buffers(inode);
903 end_writeback(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -0500904 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400905 ext4_discard_preallocations(inode);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500906 if (EXT4_I(inode)->jinode) {
907 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
908 EXT4_I(inode)->jinode);
909 jbd2_free_inode(EXT4_I(inode)->jinode);
910 EXT4_I(inode)->jinode = NULL;
911 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700912}
913
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400914static inline void ext4_show_quota_options(struct seq_file *seq,
915 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700916{
917#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700918 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700919
Jan Kara5a20bdf2009-11-30 23:58:32 +0100920 if (sbi->s_jquota_fmt) {
921 char *fmtname = "";
922
923 switch (sbi->s_jquota_fmt) {
924 case QFMT_VFS_OLD:
925 fmtname = "vfsold";
926 break;
927 case QFMT_VFS_V0:
928 fmtname = "vfsv0";
929 break;
930 case QFMT_VFS_V1:
931 fmtname = "vfsv1";
932 break;
933 }
934 seq_printf(seq, ",jqfmt=%s", fmtname);
935 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700936
937 if (sbi->s_qf_names[USRQUOTA])
938 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
939
940 if (sbi->s_qf_names[GRPQUOTA])
941 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
942
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500943 if (test_opt(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700944 seq_puts(seq, ",usrquota");
945
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500946 if (test_opt(sb, GRPQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700947 seq_puts(seq, ",grpquota");
948#endif
949}
950
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700951/*
952 * Show an option if
953 * - it's set to a non-default value OR
954 * - if the per-sb default is different from the global default
955 */
Mingming Cao617ba132006-10-11 01:20:53 -0700956static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700957{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500958 int def_errors;
959 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700960 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700961 struct ext4_sb_info *sbi = EXT4_SB(sb);
962 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700963
964 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500965 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700966
967 if (sbi->s_sb_block != 1)
968 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
969 if (test_opt(sb, MINIX_DF))
970 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500971 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700972 seq_puts(seq, ",grpid");
973 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
974 seq_puts(seq, ",nogrpid");
975 if (sbi->s_resuid != EXT4_DEF_RESUID ||
976 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
977 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
978 }
979 if (sbi->s_resgid != EXT4_DEF_RESGID ||
980 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
981 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
982 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500983 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700984 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500985 def_errors == EXT4_ERRORS_CONTINUE) {
986 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700987 }
988 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500989 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500990 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500991 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700992 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500993 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700994 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500995 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700996 seq_puts(seq, ",debug");
997 if (test_opt(sb, OLDALLOC))
998 seq_puts(seq, ",oldalloc");
Theodore Ts'o03010a32008-10-10 20:02:48 -0400999#ifdef CONFIG_EXT4_FS_XATTR
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001000 if (test_opt(sb, XATTR_USER) &&
1001 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001002 seq_puts(seq, ",user_xattr");
1003 if (!test_opt(sb, XATTR_USER) &&
1004 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
1005 seq_puts(seq, ",nouser_xattr");
1006 }
1007#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001008#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001009 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001010 seq_puts(seq, ",acl");
1011 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1012 seq_puts(seq, ",noacl");
1013#endif
Theodore Ts'o30773842009-01-03 20:27:38 -05001014 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001015 seq_printf(seq, ",commit=%u",
1016 (unsigned) (sbi->s_commit_interval / HZ));
1017 }
Theodore Ts'o30773842009-01-03 20:27:38 -05001018 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1019 seq_printf(seq, ",min_batch_time=%u",
1020 (unsigned) sbi->s_min_batch_time);
1021 }
1022 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1023 seq_printf(seq, ",max_batch_time=%u",
1024 (unsigned) sbi->s_min_batch_time);
1025 }
1026
Eric Sandeen571640c2008-05-26 12:29:46 -04001027 /*
1028 * We're changing the default of barrier mount option, so
1029 * let's always display its mount state so it's clear what its
1030 * status is.
1031 */
1032 seq_puts(seq, ",barrier=");
1033 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -04001034 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1035 seq_puts(seq, ",journal_async_commit");
Jan Kara39a4bad2010-05-16 17:00:00 -04001036 else if (test_opt(sb, JOURNAL_CHECKSUM))
1037 seq_puts(seq, ",journal_checksum");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001038 if (test_opt(sb, I_VERSION))
1039 seq_puts(seq, ",i_version");
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001040 if (!test_opt(sb, DELALLOC) &&
1041 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001042 seq_puts(seq, ",nodelalloc");
1043
Theodore Ts'o14490322010-12-14 15:27:50 -05001044 if (test_opt(sb, MBLK_IO_SUBMIT))
1045 seq_puts(seq, ",mblk_io_submit");
Miklos Szeredicb45bbe2008-01-28 23:58:27 -05001046 if (sbi->s_stripe)
1047 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001048 /*
1049 * journal mode get enabled in different ways
1050 * So just print the value even if we didn't specify it
1051 */
Mingming Cao617ba132006-10-11 01:20:53 -07001052 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001053 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -07001054 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001055 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -07001056 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001057 seq_puts(seq, ",data=writeback");
1058
Theodore Ts'o240799c2008-10-09 23:53:47 -04001059 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1060 seq_printf(seq, ",inode_readahead_blks=%u",
1061 sbi->s_inode_readahead_blks);
1062
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001063 if (test_opt(sb, DATA_ERR_ABORT))
1064 seq_puts(seq, ",data_err=abort");
1065
Theodore Ts'oafd46722009-03-16 23:12:23 -04001066 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001067 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -04001068
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001069 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
Eric Sandeen5328e632009-11-19 14:25:42 -05001070 seq_puts(seq, ",discard");
1071
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001072 if (test_opt(sb, NOLOAD))
1073 seq_puts(seq, ",norecovery");
1074
Jiaying Zhang744692d2010-03-04 16:14:02 -05001075 if (test_opt(sb, DIOREAD_NOLOCK))
1076 seq_puts(seq, ",dioread_nolock");
1077
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001078 if (test_opt(sb, BLOCK_VALIDITY) &&
1079 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1080 seq_puts(seq, ",block_validity");
1081
Lukas Czernerbfff6872010-10-27 21:30:05 -04001082 if (!test_opt(sb, INIT_INODE_TABLE))
1083 seq_puts(seq, ",noinit_inode_table");
1084 else if (sbi->s_li_wait_mult)
1085 seq_printf(seq, ",init_inode_table=%u",
1086 (unsigned) sbi->s_li_wait_mult);
1087
Mingming Cao617ba132006-10-11 01:20:53 -07001088 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001089
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001090 return 0;
1091}
1092
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001093static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001094 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001095{
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001096 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001097
Mingming Cao617ba132006-10-11 01:20:53 -07001098 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001099 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -07001100 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001101 return ERR_PTR(-ESTALE);
1102
1103 /* iget isn't really right if the inode is currently unallocated!!
1104 *
Mingming Cao617ba132006-10-11 01:20:53 -07001105 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001106 * deleted, so we should be safe.
1107 *
1108 * Currently we don't know the generation for parent directory, so
1109 * a generation of 0 means "accept any"
1110 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001111 inode = ext4_iget(sb, ino);
1112 if (IS_ERR(inode))
1113 return ERR_CAST(inode);
1114 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001115 iput(inode);
1116 return ERR_PTR(-ESTALE);
1117 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001118
1119 return inode;
1120}
1121
1122static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001123 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001124{
1125 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1126 ext4_nfs_get_inode);
1127}
1128
1129static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001130 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001131{
1132 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1133 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001134}
1135
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001136/*
1137 * Try to release metadata pages (indirect blocks, directories) which are
1138 * mapped via the block device. Since these pages could have journal heads
1139 * which would prevent try_to_free_buffers() from freeing them, we must use
1140 * jbd2 layer's try_to_free_buffers() function to release them.
1141 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001142static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1143 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001144{
1145 journal_t *journal = EXT4_SB(sb)->s_journal;
1146
1147 WARN_ON(PageChecked(page));
1148 if (!page_has_buffers(page))
1149 return 0;
1150 if (journal)
1151 return jbd2_journal_try_to_free_buffers(journal, page,
1152 wait & ~__GFP_WAIT);
1153 return try_to_free_buffers(page);
1154}
1155
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001156#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001157#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001158#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001159
Mingming Cao617ba132006-10-11 01:20:53 -07001160static int ext4_write_dquot(struct dquot *dquot);
1161static int ext4_acquire_dquot(struct dquot *dquot);
1162static int ext4_release_dquot(struct dquot *dquot);
1163static int ext4_mark_dquot_dirty(struct dquot *dquot);
1164static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001165static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02001166 struct path *path);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001167static int ext4_quota_off(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001168static int ext4_quota_on_mount(struct super_block *sb, int type);
1169static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001170 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001171static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001172 const char *data, size_t len, loff_t off);
1173
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001174static const struct dquot_operations ext4_quota_operations = {
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001175#ifdef CONFIG_QUOTA
Mingming Cao60e58e02009-01-22 18:13:05 +01001176 .get_reserved_space = ext4_get_reserved_space,
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001177#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001178 .write_dquot = ext4_write_dquot,
1179 .acquire_dquot = ext4_acquire_dquot,
1180 .release_dquot = ext4_release_dquot,
1181 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001182 .write_info = ext4_write_info,
1183 .alloc_dquot = dquot_alloc,
1184 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001185};
1186
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001187static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001188 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001189 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001190 .quota_sync = dquot_quota_sync,
1191 .get_info = dquot_get_dqinfo,
1192 .set_info = dquot_set_dqinfo,
1193 .get_dqblk = dquot_get_dqblk,
1194 .set_dqblk = dquot_set_dqblk
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001195};
1196#endif
1197
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001198static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001199 .alloc_inode = ext4_alloc_inode,
1200 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001201 .write_inode = ext4_write_inode,
1202 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001203 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001204 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001205 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001206 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001207 .freeze_fs = ext4_freeze,
1208 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001209 .statfs = ext4_statfs,
1210 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001211 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001212#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001213 .quota_read = ext4_quota_read,
1214 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001215#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001216 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001217};
1218
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001219static const struct super_operations ext4_nojournal_sops = {
1220 .alloc_inode = ext4_alloc_inode,
1221 .destroy_inode = ext4_destroy_inode,
1222 .write_inode = ext4_write_inode,
1223 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001224 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001225 .evict_inode = ext4_evict_inode,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001226 .write_super = ext4_write_super,
1227 .put_super = ext4_put_super,
1228 .statfs = ext4_statfs,
1229 .remount_fs = ext4_remount,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001230 .show_options = ext4_show_options,
1231#ifdef CONFIG_QUOTA
1232 .quota_read = ext4_quota_read,
1233 .quota_write = ext4_quota_write,
1234#endif
1235 .bdev_try_to_free_page = bdev_try_to_free_page,
1236};
1237
Christoph Hellwig39655162007-10-21 16:42:17 -07001238static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001239 .fh_to_dentry = ext4_fh_to_dentry,
1240 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001241 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001242};
1243
1244enum {
1245 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1246 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001247 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001248 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001249 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001250 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001251 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001252 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001253 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001254 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001255 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001256 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1257 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1258 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o14490322010-12-14 15:27:50 -05001259 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1260 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001261 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001262 Opt_dioread_nolock, Opt_dioread_lock,
Eric Sandeen5328e632009-11-19 14:25:42 -05001263 Opt_discard, Opt_nodiscard,
Lukas Czernerbfff6872010-10-27 21:30:05 -04001264 Opt_init_inode_table, Opt_noinit_inode_table,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001265};
1266
Steven Whitehousea447c092008-10-13 10:46:57 +01001267static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001268 {Opt_bsd_df, "bsddf"},
1269 {Opt_minix_df, "minixdf"},
1270 {Opt_grpid, "grpid"},
1271 {Opt_grpid, "bsdgroups"},
1272 {Opt_nogrpid, "nogrpid"},
1273 {Opt_nogrpid, "sysvgroups"},
1274 {Opt_resgid, "resgid=%u"},
1275 {Opt_resuid, "resuid=%u"},
1276 {Opt_sb, "sb=%u"},
1277 {Opt_err_cont, "errors=continue"},
1278 {Opt_err_panic, "errors=panic"},
1279 {Opt_err_ro, "errors=remount-ro"},
1280 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001281 {Opt_debug, "debug"},
1282 {Opt_oldalloc, "oldalloc"},
1283 {Opt_orlov, "orlov"},
1284 {Opt_user_xattr, "user_xattr"},
1285 {Opt_nouser_xattr, "nouser_xattr"},
1286 {Opt_acl, "acl"},
1287 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001288 {Opt_noload, "noload"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001289 {Opt_noload, "norecovery"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001290 {Opt_nobh, "nobh"},
1291 {Opt_bh, "bh"},
1292 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001293 {Opt_min_batch_time, "min_batch_time=%u"},
1294 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001295 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001296 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001297 {Opt_journal_checksum, "journal_checksum"},
1298 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001299 {Opt_abort, "abort"},
1300 {Opt_data_journal, "data=journal"},
1301 {Opt_data_ordered, "data=ordered"},
1302 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001303 {Opt_data_err_abort, "data_err=abort"},
1304 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001305 {Opt_offusrjquota, "usrjquota="},
1306 {Opt_usrjquota, "usrjquota=%s"},
1307 {Opt_offgrpjquota, "grpjquota="},
1308 {Opt_grpjquota, "grpjquota=%s"},
1309 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1310 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001311 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001312 {Opt_grpquota, "grpquota"},
1313 {Opt_noquota, "noquota"},
1314 {Opt_quota, "quota"},
1315 {Opt_usrquota, "usrquota"},
1316 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001317 {Opt_barrier, "barrier"},
1318 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001319 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001320 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001321 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001322 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001323 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o14490322010-12-14 15:27:50 -05001324 {Opt_mblk_io_submit, "mblk_io_submit"},
1325 {Opt_nomblk_io_submit, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001326 {Opt_block_validity, "block_validity"},
1327 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001328 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001329 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001330 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001331 {Opt_auto_da_alloc, "auto_da_alloc"},
1332 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001333 {Opt_dioread_nolock, "dioread_nolock"},
1334 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001335 {Opt_discard, "discard"},
1336 {Opt_nodiscard, "nodiscard"},
Lukas Czernerbfff6872010-10-27 21:30:05 -04001337 {Opt_init_inode_table, "init_itable=%u"},
1338 {Opt_init_inode_table, "init_itable"},
1339 {Opt_noinit_inode_table, "noinit_itable"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001340 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001341};
1342
Mingming Cao617ba132006-10-11 01:20:53 -07001343static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001344{
Mingming Cao617ba132006-10-11 01:20:53 -07001345 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001346 char *options = (char *) *data;
1347
1348 if (!options || strncmp(options, "sb=", 3) != 0)
1349 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001350
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001351 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001352 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001353 sb_block = simple_strtoul(options, &options, 0);
1354 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001355 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001356 (char *) *data);
1357 return 1;
1358 }
1359 if (*options == ',')
1360 options++;
1361 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001362
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001363 return sb_block;
1364}
1365
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001366#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001367static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1368 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001369
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001370#ifdef CONFIG_QUOTA
1371static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1372{
1373 struct ext4_sb_info *sbi = EXT4_SB(sb);
1374 char *qname;
1375
1376 if (sb_any_quota_loaded(sb) &&
1377 !sbi->s_qf_names[qtype]) {
1378 ext4_msg(sb, KERN_ERR,
1379 "Cannot change journaled "
1380 "quota options when quota turned on");
1381 return 0;
1382 }
1383 qname = match_strdup(args);
1384 if (!qname) {
1385 ext4_msg(sb, KERN_ERR,
1386 "Not enough memory for storing quotafile name");
1387 return 0;
1388 }
1389 if (sbi->s_qf_names[qtype] &&
1390 strcmp(sbi->s_qf_names[qtype], qname)) {
1391 ext4_msg(sb, KERN_ERR,
1392 "%s quota file already specified", QTYPE2NAME(qtype));
1393 kfree(qname);
1394 return 0;
1395 }
1396 sbi->s_qf_names[qtype] = qname;
1397 if (strchr(sbi->s_qf_names[qtype], '/')) {
1398 ext4_msg(sb, KERN_ERR,
1399 "quotafile must be on filesystem root");
1400 kfree(sbi->s_qf_names[qtype]);
1401 sbi->s_qf_names[qtype] = NULL;
1402 return 0;
1403 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001404 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001405 return 1;
1406}
1407
1408static int clear_qf_name(struct super_block *sb, int qtype)
1409{
1410
1411 struct ext4_sb_info *sbi = EXT4_SB(sb);
1412
1413 if (sb_any_quota_loaded(sb) &&
1414 sbi->s_qf_names[qtype]) {
1415 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1416 " when quota turned on");
1417 return 0;
1418 }
1419 /*
1420 * The space will be released later when all options are confirmed
1421 * to be correct
1422 */
1423 sbi->s_qf_names[qtype] = NULL;
1424 return 1;
1425}
1426#endif
1427
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001428static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001429 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001430 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001431 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001432{
Mingming Cao617ba132006-10-11 01:20:53 -07001433 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001434 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001435 substring_t args[MAX_OPT_ARGS];
1436 int data_opt = 0;
1437 int option;
1438#ifdef CONFIG_QUOTA
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001439 int qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001440#endif
1441
1442 if (!options)
1443 return 1;
1444
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001445 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001446 int token;
1447 if (!*p)
1448 continue;
1449
Eric Sandeen15121c12010-02-15 20:17:55 -05001450 /*
1451 * Initialize args struct so we know whether arg was
1452 * found; some options take optional arguments.
1453 */
Peter Huewe7dc57612011-02-21 21:01:42 -05001454 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001455 token = match_token(p, tokens, args);
1456 switch (token) {
1457 case Opt_bsd_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001458 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001459 clear_opt(sb, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001460 break;
1461 case Opt_minix_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001462 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001463 set_opt(sb, MINIX_DF);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001464
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001465 break;
1466 case Opt_grpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001467 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001468 set_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001469
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001470 break;
1471 case Opt_nogrpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001472 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001473 clear_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001474
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001475 break;
1476 case Opt_resuid:
1477 if (match_int(&args[0], &option))
1478 return 0;
1479 sbi->s_resuid = option;
1480 break;
1481 case Opt_resgid:
1482 if (match_int(&args[0], &option))
1483 return 0;
1484 sbi->s_resgid = option;
1485 break;
1486 case Opt_sb:
1487 /* handled by get_sb_block() instead of here */
1488 /* *sb_block = match_int(&args[0]); */
1489 break;
1490 case Opt_err_panic:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001491 clear_opt(sb, ERRORS_CONT);
1492 clear_opt(sb, ERRORS_RO);
1493 set_opt(sb, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001494 break;
1495 case Opt_err_ro:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001496 clear_opt(sb, ERRORS_CONT);
1497 clear_opt(sb, ERRORS_PANIC);
1498 set_opt(sb, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001499 break;
1500 case Opt_err_cont:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001501 clear_opt(sb, ERRORS_RO);
1502 clear_opt(sb, ERRORS_PANIC);
1503 set_opt(sb, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001504 break;
1505 case Opt_nouid32:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001506 set_opt(sb, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001507 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001508 case Opt_debug:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001509 set_opt(sb, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001510 break;
1511 case Opt_oldalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001512 set_opt(sb, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001513 break;
1514 case Opt_orlov:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001515 clear_opt(sb, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001516 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001517#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001518 case Opt_user_xattr:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001519 set_opt(sb, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001520 break;
1521 case Opt_nouser_xattr:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001522 clear_opt(sb, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001523 break;
1524#else
1525 case Opt_user_xattr:
1526 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001527 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001528 break;
1529#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001530#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001531 case Opt_acl:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001532 set_opt(sb, POSIX_ACL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001533 break;
1534 case Opt_noacl:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001535 clear_opt(sb, POSIX_ACL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001536 break;
1537#else
1538 case Opt_acl:
1539 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001540 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001541 break;
1542#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001543 case Opt_journal_update:
1544 /* @@@ FIXME */
1545 /* Eventually we will want to be able to create
1546 a journal file here. For now, only allow the
1547 user to specify an existing inode to be the
1548 journal file. */
1549 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001550 ext4_msg(sb, KERN_ERR,
1551 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001552 return 0;
1553 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001554 set_opt(sb, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001555 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001556 case Opt_journal_dev:
1557 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001558 ext4_msg(sb, KERN_ERR,
1559 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001560 return 0;
1561 }
1562 if (match_int(&args[0], &option))
1563 return 0;
1564 *journal_devnum = option;
1565 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001566 case Opt_journal_checksum:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001567 set_opt(sb, JOURNAL_CHECKSUM);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001568 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001569 case Opt_journal_async_commit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001570 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1571 set_opt(sb, JOURNAL_CHECKSUM);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001572 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001573 case Opt_noload:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001574 set_opt(sb, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001575 break;
1576 case Opt_commit:
1577 if (match_int(&args[0], &option))
1578 return 0;
1579 if (option < 0)
1580 return 0;
1581 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001582 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001583 sbi->s_commit_interval = HZ * option;
1584 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001585 case Opt_max_batch_time:
1586 if (match_int(&args[0], &option))
1587 return 0;
1588 if (option < 0)
1589 return 0;
1590 if (option == 0)
1591 option = EXT4_DEF_MAX_BATCH_TIME;
1592 sbi->s_max_batch_time = option;
1593 break;
1594 case Opt_min_batch_time:
1595 if (match_int(&args[0], &option))
1596 return 0;
1597 if (option < 0)
1598 return 0;
1599 sbi->s_min_batch_time = option;
1600 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001601 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001602 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001603 goto datacheck;
1604 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001605 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001606 goto datacheck;
1607 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001608 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001609 datacheck:
1610 if (is_remount) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001611 if (test_opt(sb, DATA_FLAGS) != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001612 ext4_msg(sb, KERN_ERR,
1613 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001614 return 0;
1615 }
1616 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001617 clear_opt(sb, DATA_FLAGS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001618 sbi->s_mount_opt |= data_opt;
1619 }
1620 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001621 case Opt_data_err_abort:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001622 set_opt(sb, DATA_ERR_ABORT);
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001623 break;
1624 case Opt_data_err_ignore:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001625 clear_opt(sb, DATA_ERR_ABORT);
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001626 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001627#ifdef CONFIG_QUOTA
1628 case Opt_usrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001629 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1630 return 0;
1631 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001632 case Opt_grpjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001633 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001634 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001635 break;
1636 case Opt_offusrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001637 if (!clear_qf_name(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001638 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001639 break;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001640 case Opt_offgrpjquota:
1641 if (!clear_qf_name(sb, GRPQUOTA))
1642 return 0;
1643 break;
1644
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001645 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001646 qfmt = QFMT_VFS_OLD;
1647 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001648 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001649 qfmt = QFMT_VFS_V0;
Jan Kara5a20bdf2009-11-30 23:58:32 +01001650 goto set_qf_format;
1651 case Opt_jqfmt_vfsv1:
1652 qfmt = QFMT_VFS_V1;
Jan Karadfc5d032008-05-13 19:11:51 -04001653set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001654 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001655 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001656 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001657 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001658 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001659 return 0;
1660 }
1661 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001662 break;
1663 case Opt_quota:
1664 case Opt_usrquota:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001665 set_opt(sb, QUOTA);
1666 set_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001667 break;
1668 case Opt_grpquota:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001669 set_opt(sb, QUOTA);
1670 set_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001671 break;
1672 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001673 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001674 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1675 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001676 return 0;
1677 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001678 clear_opt(sb, QUOTA);
1679 clear_opt(sb, USRQUOTA);
1680 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001681 break;
1682#else
1683 case Opt_quota:
1684 case Opt_usrquota:
1685 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001686 ext4_msg(sb, KERN_ERR,
1687 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001688 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001689 case Opt_usrjquota:
1690 case Opt_grpjquota:
1691 case Opt_offusrjquota:
1692 case Opt_offgrpjquota:
1693 case Opt_jqfmt_vfsold:
1694 case Opt_jqfmt_vfsv0:
Jan Kara5a20bdf2009-11-30 23:58:32 +01001695 case Opt_jqfmt_vfsv1:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001696 ext4_msg(sb, KERN_ERR,
1697 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001698 break;
1699 case Opt_noquota:
1700 break;
1701#endif
1702 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001703 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001704 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001705 case Opt_nobarrier:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001706 clear_opt(sb, BARRIER);
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001707 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001708 case Opt_barrier:
Eric Sandeen15121c12010-02-15 20:17:55 -05001709 if (args[0].from) {
1710 if (match_int(&args[0], &option))
1711 return 0;
1712 } else
1713 option = 1; /* No argument, default to 1 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001714 if (option)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001715 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001716 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001717 clear_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001718 break;
1719 case Opt_ignore:
1720 break;
1721 case Opt_resize:
1722 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001723 ext4_msg(sb, KERN_ERR,
1724 "resize option only available "
1725 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001726 return 0;
1727 }
1728 if (match_int(&args[0], &option) != 0)
1729 return 0;
1730 *n_blocks_count = option;
1731 break;
1732 case Opt_nobh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001733 ext4_msg(sb, KERN_WARNING,
1734 "Ignoring deprecated nobh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001735 break;
1736 case Opt_bh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001737 ext4_msg(sb, KERN_WARNING,
1738 "Ignoring deprecated bh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001739 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001740 case Opt_i_version:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001741 set_opt(sb, I_VERSION);
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001742 sb->s_flags |= MS_I_VERSION;
1743 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001744 case Opt_nodelalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001745 clear_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001746 break;
Theodore Ts'o14490322010-12-14 15:27:50 -05001747 case Opt_mblk_io_submit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001748 set_opt(sb, MBLK_IO_SUBMIT);
Theodore Ts'o14490322010-12-14 15:27:50 -05001749 break;
1750 case Opt_nomblk_io_submit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001751 clear_opt(sb, MBLK_IO_SUBMIT);
Theodore Ts'o14490322010-12-14 15:27:50 -05001752 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001753 case Opt_stripe:
1754 if (match_int(&args[0], &option))
1755 return 0;
1756 if (option < 0)
1757 return 0;
1758 sbi->s_stripe = option;
1759 break;
Alex Tomas64769242008-07-11 19:27:31 -04001760 case Opt_delalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001761 set_opt(sb, DELALLOC);
Alex Tomas64769242008-07-11 19:27:31 -04001762 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001763 case Opt_block_validity:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001764 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001765 break;
1766 case Opt_noblock_validity:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001767 clear_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001768 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001769 case Opt_inode_readahead_blks:
1770 if (match_int(&args[0], &option))
1771 return 0;
1772 if (option < 0 || option > (1 << 30))
1773 return 0;
Alexander V. Lukyanov5dbd5712011-02-21 21:33:21 -05001774 if (option && !is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001775 ext4_msg(sb, KERN_ERR,
1776 "EXT4-fs: inode_readahead_blks"
1777 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001778 return 0;
1779 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001780 sbi->s_inode_readahead_blks = option;
1781 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001782 case Opt_journal_ioprio:
1783 if (match_int(&args[0], &option))
1784 return 0;
1785 if (option < 0 || option > 7)
1786 break;
1787 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1788 option);
1789 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001790 case Opt_noauto_da_alloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001791 set_opt(sb, NO_AUTO_DA_ALLOC);
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001792 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001793 case Opt_auto_da_alloc:
Eric Sandeen15121c12010-02-15 20:17:55 -05001794 if (args[0].from) {
1795 if (match_int(&args[0], &option))
1796 return 0;
1797 } else
1798 option = 1; /* No argument, default to 1 */
Theodore Ts'oafd46722009-03-16 23:12:23 -04001799 if (option)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001800 clear_opt(sb, NO_AUTO_DA_ALLOC);
Theodore Ts'oafd46722009-03-16 23:12:23 -04001801 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001802 set_opt(sb,NO_AUTO_DA_ALLOC);
Theodore Ts'oafd46722009-03-16 23:12:23 -04001803 break;
Eric Sandeen5328e632009-11-19 14:25:42 -05001804 case Opt_discard:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001805 set_opt(sb, DISCARD);
Eric Sandeen5328e632009-11-19 14:25:42 -05001806 break;
1807 case Opt_nodiscard:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001808 clear_opt(sb, DISCARD);
Eric Sandeen5328e632009-11-19 14:25:42 -05001809 break;
Jiaying Zhang744692d2010-03-04 16:14:02 -05001810 case Opt_dioread_nolock:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001811 set_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001812 break;
1813 case Opt_dioread_lock:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001814 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001815 break;
Lukas Czernerbfff6872010-10-27 21:30:05 -04001816 case Opt_init_inode_table:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001817 set_opt(sb, INIT_INODE_TABLE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001818 if (args[0].from) {
1819 if (match_int(&args[0], &option))
1820 return 0;
1821 } else
1822 option = EXT4_DEF_LI_WAIT_MULT;
1823 if (option < 0)
1824 return 0;
1825 sbi->s_li_wait_mult = option;
1826 break;
1827 case Opt_noinit_inode_table:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001828 clear_opt(sb, INIT_INODE_TABLE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001829 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001830 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001831 ext4_msg(sb, KERN_ERR,
1832 "Unrecognized mount option \"%s\" "
1833 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001834 return 0;
1835 }
1836 }
1837#ifdef CONFIG_QUOTA
1838 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001839 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001840 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001841
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001842 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001843 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001844
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001845 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001846 ext4_msg(sb, KERN_ERR, "old and new quota "
1847 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001848 return 0;
1849 }
1850
1851 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001852 ext4_msg(sb, KERN_ERR, "journaled quota format "
1853 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001854 return 0;
1855 }
1856 } else {
1857 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001858 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001859 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001860 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001861 return 0;
1862 }
1863 }
1864#endif
1865 return 1;
1866}
1867
Mingming Cao617ba132006-10-11 01:20:53 -07001868static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 int read_only)
1870{
Mingming Cao617ba132006-10-11 01:20:53 -07001871 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001872 int res = 0;
1873
Mingming Cao617ba132006-10-11 01:20:53 -07001874 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001875 ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001877 res = MS_RDONLY;
1878 }
1879 if (read_only)
1880 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001881 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001882 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001884 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001885 ext4_msg(sb, KERN_WARNING,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001888 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1889 le16_to_cpu(es->s_mnt_count) >=
1890 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001891 ext4_msg(sb, KERN_WARNING,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001894 else if (le32_to_cpu(es->s_checkinterval) &&
1895 (le32_to_cpu(es->s_lastcheck) +
1896 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001897 ext4_msg(sb, KERN_WARNING,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001900 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001901 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001902 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001903 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001904 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001905 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001906 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001907 if (sbi->s_journal)
1908 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001909
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001910 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001911 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001912 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001913 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001914 sb->s_blocksize,
1915 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001916 EXT4_BLOCKS_PER_GROUP(sb),
1917 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001918 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001919
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001920 return res;
1921}
1922
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001923static int ext4_fill_flex_info(struct super_block *sb)
1924{
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001927 ext4_group_t flex_group_count;
1928 ext4_group_t flex_group;
1929 int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001930 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001931 int i;
1932
Theodore Ts'o503358a2009-11-23 07:24:46 -05001933 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1934 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1935
1936 if (groups_per_flex < 2) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001937 sbi->s_log_groups_per_flex = 0;
1938 return 1;
1939 }
1940
Frederic Bohec62a11f2008-09-08 10:20:24 -04001941 /* We allocate both existing and potentially added groups */
1942 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05001943 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1944 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001945 size = flex_group_count * sizeof(struct flex_groups);
1946 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1947 if (sbi->s_flex_groups == NULL) {
Joe Perches94de56a2010-12-19 22:21:02 -05001948 sbi->s_flex_groups = vzalloc(size);
1949 if (sbi->s_flex_groups == NULL) {
1950 ext4_msg(sb, KERN_ERR,
1951 "not enough memory for %u flex groups",
1952 flex_group_count);
1953 goto failed;
1954 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001955 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001956
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001957 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001958 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001959
1960 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001961 atomic_add(ext4_free_inodes_count(sb, gdp),
1962 &sbi->s_flex_groups[flex_group].free_inodes);
1963 atomic_add(ext4_free_blks_count(sb, gdp),
1964 &sbi->s_flex_groups[flex_group].free_blocks);
1965 atomic_add(ext4_used_dirs_count(sb, gdp),
1966 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001967 }
1968
1969 return 1;
1970failed:
1971 return 0;
1972}
1973
Andreas Dilger717d50e2007-10-16 18:38:25 -04001974__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1975 struct ext4_group_desc *gdp)
1976{
1977 __u16 crc = 0;
1978
1979 if (sbi->s_es->s_feature_ro_compat &
1980 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1981 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1982 __le32 le_group = cpu_to_le32(block_group);
1983
1984 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1985 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1986 crc = crc16(crc, (__u8 *)gdp, offset);
1987 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1988 /* for checksum of struct ext4_group_desc do the rest...*/
1989 if ((sbi->s_es->s_feature_incompat &
1990 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1991 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1992 crc = crc16(crc, (__u8 *)gdp + offset,
1993 le16_to_cpu(sbi->s_es->s_desc_size) -
1994 offset);
1995 }
1996
1997 return cpu_to_le16(crc);
1998}
1999
2000int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2001 struct ext4_group_desc *gdp)
2002{
2003 if ((sbi->s_es->s_feature_ro_compat &
2004 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2005 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2006 return 0;
2007
2008 return 1;
2009}
2010
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002011/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002012static int ext4_check_descriptors(struct super_block *sb,
2013 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002014{
Mingming Cao617ba132006-10-11 01:20:53 -07002015 struct ext4_sb_info *sbi = EXT4_SB(sb);
2016 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2017 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002018 ext4_fsblk_t block_bitmap;
2019 ext4_fsblk_t inode_bitmap;
2020 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002021 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002022 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002023
Jose R. Santosce421582007-10-16 18:38:25 -04002024 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2025 flexbg_flag = 1;
2026
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002027 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002028
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002029 for (i = 0; i < sbi->s_groups_count; i++) {
2030 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2031
Jose R. Santosce421582007-10-16 18:38:25 -04002032 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002033 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002034 else
2035 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002036 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002037
Lukas Czernerbfff6872010-10-27 21:30:05 -04002038 if ((grp == sbi->s_groups_count) &&
2039 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2040 grp = i;
2041
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002042 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002043 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002044 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002045 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002046 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002047 return 0;
2048 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002049 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002050 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002051 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002052 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002053 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002054 return 0;
2055 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002056 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002057 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002058 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002059 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002060 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002061 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002062 return 0;
2063 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002064 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002065 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002066 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2067 "Checksum for group %u failed (%u!=%u)",
2068 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2069 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002070 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002071 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002072 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002073 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002074 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002075 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002076 if (!flexbg_flag)
2077 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002078 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002079 if (NULL != first_not_zeroed)
2080 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002081
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002082 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002083 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002084 return 1;
2085}
2086
Mingming Cao617ba132006-10-11 01:20:53 -07002087/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002088 * the superblock) which were deleted from all directories, but held open by
2089 * a process at the time of a crash. We walk the list and try to delete these
2090 * inodes at recovery time (only with a read-write filesystem).
2091 *
2092 * In order to keep the orphan inode chain consistent during traversal (in
2093 * case of crash during recovery), we link each inode into the superblock
2094 * orphan list_head and handle it the same way as an inode deletion during
2095 * normal operation (which journals the operations for us).
2096 *
2097 * We only do an iget() and an iput() on each inode, which is very safe if we
2098 * accidentally point at an in-use or already deleted inode. The worst that
2099 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002100 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002101 * e2fsck was run on this filesystem, and it must have already done the orphan
2102 * inode cleanup for us, so we can safely abort without any further action.
2103 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002104static void ext4_orphan_cleanup(struct super_block *sb,
2105 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002106{
2107 unsigned int s_flags = sb->s_flags;
2108 int nr_orphans = 0, nr_truncates = 0;
2109#ifdef CONFIG_QUOTA
2110 int i;
2111#endif
2112 if (!es->s_last_orphan) {
2113 jbd_debug(4, "no orphan inodes to clean up\n");
2114 return;
2115 }
2116
Eric Sandeena8f48a92006-12-06 20:40:13 -08002117 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002118 ext4_msg(sb, KERN_ERR, "write access "
2119 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002120 return;
2121 }
2122
Mingming Cao617ba132006-10-11 01:20:53 -07002123 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002124 if (es->s_last_orphan)
2125 jbd_debug(1, "Errors on filesystem, "
2126 "clearing orphan list.\n");
2127 es->s_last_orphan = 0;
2128 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2129 return;
2130 }
2131
2132 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002133 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002134 sb->s_flags &= ~MS_RDONLY;
2135 }
2136#ifdef CONFIG_QUOTA
2137 /* Needed for iput() to work correctly and not trash data */
2138 sb->s_flags |= MS_ACTIVE;
2139 /* Turn on quotas so that they are updated correctly */
2140 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002141 if (EXT4_SB(sb)->s_qf_names[i]) {
2142 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002143 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002144 ext4_msg(sb, KERN_ERR,
2145 "Cannot turn on journaled "
2146 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002147 }
2148 }
2149#endif
2150
2151 while (es->s_last_orphan) {
2152 struct inode *inode;
2153
Josef Bacik97bd42b2008-04-29 22:04:56 -04002154 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2155 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002156 es->s_last_orphan = 0;
2157 break;
2158 }
2159
Mingming Cao617ba132006-10-11 01:20:53 -07002160 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002161 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002162 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002163 ext4_msg(sb, KERN_DEBUG,
2164 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002165 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002166 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002167 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002168 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002169 nr_truncates++;
2170 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002171 ext4_msg(sb, KERN_DEBUG,
2172 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002173 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002174 jbd_debug(2, "deleting unreferenced inode %lu\n",
2175 inode->i_ino);
2176 nr_orphans++;
2177 }
2178 iput(inode); /* The delete magic happens here! */
2179 }
2180
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002181#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002182
2183 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002184 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2185 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002186 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002187 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2188 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002189#ifdef CONFIG_QUOTA
2190 /* Turn quotas off */
2191 for (i = 0; i < MAXQUOTAS; i++) {
2192 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002193 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002194 }
2195#endif
2196 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2197}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002198
Eric Sandeencd2291a2008-01-28 23:58:27 -05002199/*
2200 * Maximal extent format file size.
2201 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2202 * extent format containers, within a sector_t, and within i_blocks
2203 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2204 * so that won't be a limiting factor.
2205 *
2206 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2207 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002208static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002209{
2210 loff_t res;
2211 loff_t upper_limit = MAX_LFS_FILESIZE;
2212
2213 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002214 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002215 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002216 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002217 * i_block represent total blocks in 512 bytes
2218 * 32 == size of vfs inode i_blocks * 8
2219 */
2220 upper_limit = (1LL << 32) - 1;
2221
2222 /* total blocks in file system block size */
2223 upper_limit >>= (blkbits - 9);
2224 upper_limit <<= blkbits;
2225 }
2226
2227 /* 32-bit extent-start container, ee_block */
2228 res = 1LL << 32;
2229 res <<= blkbits;
2230 res -= 1;
2231
2232 /* Sanity check against vm- & vfs- imposed limits */
2233 if (res > upper_limit)
2234 res = upper_limit;
2235
2236 return res;
2237}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002238
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002239/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002240 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002241 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2242 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002243 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002244static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002245{
Mingming Cao617ba132006-10-11 01:20:53 -07002246 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002247 int meta_blocks;
2248 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002249 /* This is calculated to be the largest file size for a dense, block
2250 * mapped file such that the file's total number of 512-byte sectors,
2251 * including data and all indirect blocks, does not exceed (2^48 - 1).
2252 *
2253 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2254 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002255 */
2256
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002257 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002258 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002259 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002260 * the inode i_block field represents total file blocks in
2261 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002262 */
2263 upper_limit = (1LL << 32) - 1;
2264
2265 /* total blocks in file system block size */
2266 upper_limit >>= (bits - 9);
2267
2268 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002269 /*
2270 * We use 48 bit ext4_inode i_blocks
2271 * With EXT4_HUGE_FILE_FL set the i_blocks
2272 * represent total number of blocks in
2273 * file system block size
2274 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002275 upper_limit = (1LL << 48) - 1;
2276
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002277 }
2278
2279 /* indirect blocks */
2280 meta_blocks = 1;
2281 /* double indirect blocks */
2282 meta_blocks += 1 + (1LL << (bits-2));
2283 /* tripple indirect blocks */
2284 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2285
2286 upper_limit -= meta_blocks;
2287 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002288
2289 res += 1LL << (bits-2);
2290 res += 1LL << (2*(bits-2));
2291 res += 1LL << (3*(bits-2));
2292 res <<= bits;
2293 if (res > upper_limit)
2294 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002295
2296 if (res > MAX_LFS_FILESIZE)
2297 res = MAX_LFS_FILESIZE;
2298
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002299 return res;
2300}
2301
Mingming Cao617ba132006-10-11 01:20:53 -07002302static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002303 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002304{
Mingming Cao617ba132006-10-11 01:20:53 -07002305 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002306 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002307 int has_super = 0;
2308
2309 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2310
Mingming Cao617ba132006-10-11 01:20:53 -07002311 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002312 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002313 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002314 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002315 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002316 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002317
Mingming Cao617ba132006-10-11 01:20:53 -07002318 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002319}
2320
Alex Tomasc9de5602008-01-29 00:19:52 -05002321/**
2322 * ext4_get_stripe_size: Get the stripe size.
2323 * @sbi: In memory super block info
2324 *
2325 * If we have specified it via mount option, then
2326 * use the mount option value. If the value specified at mount time is
2327 * greater than the blocks per group use the super block value.
2328 * If the super block value is greater than blocks per group return 0.
2329 * Allocator needs it be less than blocks per group.
2330 *
2331 */
2332static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2333{
2334 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2335 unsigned long stripe_width =
2336 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2337
2338 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2339 return sbi->s_stripe;
2340
2341 if (stripe_width <= sbi->s_blocks_per_group)
2342 return stripe_width;
2343
2344 if (stride <= sbi->s_blocks_per_group)
2345 return stride;
2346
2347 return 0;
2348}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002349
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002350/* sysfs supprt */
2351
2352struct ext4_attr {
2353 struct attribute attr;
2354 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
Theodore Ts'o60e66792010-05-17 07:00:00 -04002355 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002356 const char *, size_t);
2357 int offset;
2358};
2359
2360static int parse_strtoul(const char *buf,
2361 unsigned long max, unsigned long *value)
2362{
2363 char *endp;
2364
André Goddard Rosae7d28602009-12-14 18:01:06 -08002365 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2366 endp = skip_spaces(endp);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002367 if (*endp || *value > max)
2368 return -EINVAL;
2369
2370 return 0;
2371}
2372
2373static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2374 struct ext4_sb_info *sbi,
2375 char *buf)
2376{
2377 return snprintf(buf, PAGE_SIZE, "%llu\n",
2378 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2379}
2380
2381static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2382 struct ext4_sb_info *sbi, char *buf)
2383{
2384 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2385
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002386 if (!sb->s_bdev->bd_part)
2387 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002388 return snprintf(buf, PAGE_SIZE, "%lu\n",
2389 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2390 sbi->s_sectors_written_start) >> 1);
2391}
2392
2393static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2394 struct ext4_sb_info *sbi, char *buf)
2395{
2396 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2397
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002398 if (!sb->s_bdev->bd_part)
2399 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002400 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002401 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002402 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002403 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002404}
2405
2406static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2407 struct ext4_sb_info *sbi,
2408 const char *buf, size_t count)
2409{
2410 unsigned long t;
2411
2412 if (parse_strtoul(buf, 0x40000000, &t))
2413 return -EINVAL;
2414
Alexander V. Lukyanov5dbd5712011-02-21 21:33:21 -05002415 if (t && !is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002416 return -EINVAL;
2417
2418 sbi->s_inode_readahead_blks = t;
2419 return count;
2420}
2421
2422static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002423 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002424{
2425 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2426
2427 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2428}
2429
2430static ssize_t sbi_ui_store(struct ext4_attr *a,
2431 struct ext4_sb_info *sbi,
2432 const char *buf, size_t count)
2433{
2434 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2435 unsigned long t;
2436
2437 if (parse_strtoul(buf, 0xffffffff, &t))
2438 return -EINVAL;
2439 *ui = t;
2440 return count;
2441}
2442
2443#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2444static struct ext4_attr ext4_attr_##_name = { \
2445 .attr = {.name = __stringify(_name), .mode = _mode }, \
2446 .show = _show, \
2447 .store = _store, \
2448 .offset = offsetof(struct ext4_sb_info, _elname), \
2449}
2450#define EXT4_ATTR(name, mode, show, store) \
2451static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2452
Lukas Czerner857ac882010-10-27 21:30:05 -04002453#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002454#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2455#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2456#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2457 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2458#define ATTR_LIST(name) &ext4_attr_##name.attr
2459
2460EXT4_RO_ATTR(delayed_allocation_blocks);
2461EXT4_RO_ATTR(session_write_kbytes);
2462EXT4_RO_ATTR(lifetime_write_kbytes);
2463EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2464 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002465EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002466EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2467EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2468EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2469EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2470EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2471EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002472EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002473
2474static struct attribute *ext4_attrs[] = {
2475 ATTR_LIST(delayed_allocation_blocks),
2476 ATTR_LIST(session_write_kbytes),
2477 ATTR_LIST(lifetime_write_kbytes),
2478 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002479 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002480 ATTR_LIST(mb_stats),
2481 ATTR_LIST(mb_max_to_scan),
2482 ATTR_LIST(mb_min_to_scan),
2483 ATTR_LIST(mb_order2_req),
2484 ATTR_LIST(mb_stream_req),
2485 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002486 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002487 NULL,
2488};
2489
Lukas Czerner857ac882010-10-27 21:30:05 -04002490/* Features this copy of ext4 supports */
2491EXT4_INFO_ATTR(lazy_itable_init);
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002492EXT4_INFO_ATTR(batched_discard);
Lukas Czerner857ac882010-10-27 21:30:05 -04002493
2494static struct attribute *ext4_feat_attrs[] = {
2495 ATTR_LIST(lazy_itable_init),
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002496 ATTR_LIST(batched_discard),
Lukas Czerner857ac882010-10-27 21:30:05 -04002497 NULL,
2498};
2499
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002500static ssize_t ext4_attr_show(struct kobject *kobj,
2501 struct attribute *attr, char *buf)
2502{
2503 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2504 s_kobj);
2505 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2506
2507 return a->show ? a->show(a, sbi, buf) : 0;
2508}
2509
2510static ssize_t ext4_attr_store(struct kobject *kobj,
2511 struct attribute *attr,
2512 const char *buf, size_t len)
2513{
2514 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2515 s_kobj);
2516 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2517
2518 return a->store ? a->store(a, sbi, buf, len) : 0;
2519}
2520
2521static void ext4_sb_release(struct kobject *kobj)
2522{
2523 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2524 s_kobj);
2525 complete(&sbi->s_kobj_unregister);
2526}
2527
Emese Revfy52cf25d2010-01-19 02:58:23 +01002528static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002529 .show = ext4_attr_show,
2530 .store = ext4_attr_store,
2531};
2532
2533static struct kobj_type ext4_ktype = {
2534 .default_attrs = ext4_attrs,
2535 .sysfs_ops = &ext4_attr_ops,
2536 .release = ext4_sb_release,
2537};
2538
Lukas Czerner857ac882010-10-27 21:30:05 -04002539static void ext4_feat_release(struct kobject *kobj)
2540{
2541 complete(&ext4_feat->f_kobj_unregister);
2542}
2543
2544static struct kobj_type ext4_feat_ktype = {
2545 .default_attrs = ext4_feat_attrs,
2546 .sysfs_ops = &ext4_attr_ops,
2547 .release = ext4_feat_release,
2548};
2549
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002550/*
2551 * Check whether this filesystem can be mounted based on
2552 * the features present and the RDONLY/RDWR mount requested.
2553 * Returns 1 if this filesystem can be mounted as requested,
2554 * 0 if it cannot be.
2555 */
2556static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2557{
2558 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2559 ext4_msg(sb, KERN_ERR,
2560 "Couldn't mount because of "
2561 "unsupported optional features (%x)",
2562 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2563 ~EXT4_FEATURE_INCOMPAT_SUPP));
2564 return 0;
2565 }
2566
2567 if (readonly)
2568 return 1;
2569
2570 /* Check that feature set is OK for a read-write mount */
2571 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2572 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2573 "unsupported optional features (%x)",
2574 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2575 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2576 return 0;
2577 }
2578 /*
2579 * Large file size enabled file system can only be mounted
2580 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2581 */
2582 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2583 if (sizeof(blkcnt_t) < sizeof(u64)) {
2584 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2585 "cannot be mounted RDWR without "
2586 "CONFIG_LBDAF");
2587 return 0;
2588 }
2589 }
2590 return 1;
2591}
2592
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002593/*
2594 * This function is called once a day if we have errors logged
2595 * on the file system
2596 */
2597static void print_daily_error_info(unsigned long arg)
2598{
2599 struct super_block *sb = (struct super_block *) arg;
2600 struct ext4_sb_info *sbi;
2601 struct ext4_super_block *es;
2602
2603 sbi = EXT4_SB(sb);
2604 es = sbi->s_es;
2605
2606 if (es->s_error_count)
2607 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2608 le32_to_cpu(es->s_error_count));
2609 if (es->s_first_error_time) {
2610 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2611 sb->s_id, le32_to_cpu(es->s_first_error_time),
2612 (int) sizeof(es->s_first_error_func),
2613 es->s_first_error_func,
2614 le32_to_cpu(es->s_first_error_line));
2615 if (es->s_first_error_ino)
2616 printk(": inode %u",
2617 le32_to_cpu(es->s_first_error_ino));
2618 if (es->s_first_error_block)
2619 printk(": block %llu", (unsigned long long)
2620 le64_to_cpu(es->s_first_error_block));
2621 printk("\n");
2622 }
2623 if (es->s_last_error_time) {
2624 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2625 sb->s_id, le32_to_cpu(es->s_last_error_time),
2626 (int) sizeof(es->s_last_error_func),
2627 es->s_last_error_func,
2628 le32_to_cpu(es->s_last_error_line));
2629 if (es->s_last_error_ino)
2630 printk(": inode %u",
2631 le32_to_cpu(es->s_last_error_ino));
2632 if (es->s_last_error_block)
2633 printk(": block %llu", (unsigned long long)
2634 le64_to_cpu(es->s_last_error_block));
2635 printk("\n");
2636 }
2637 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2638}
2639
Lukas Czernerbfff6872010-10-27 21:30:05 -04002640static void ext4_lazyinode_timeout(unsigned long data)
2641{
2642 struct task_struct *p = (struct task_struct *)data;
2643 wake_up_process(p);
2644}
2645
2646/* Find next suitable group and run ext4_init_inode_table */
2647static int ext4_run_li_request(struct ext4_li_request *elr)
2648{
2649 struct ext4_group_desc *gdp = NULL;
2650 ext4_group_t group, ngroups;
2651 struct super_block *sb;
2652 unsigned long timeout = 0;
2653 int ret = 0;
2654
2655 sb = elr->lr_super;
2656 ngroups = EXT4_SB(sb)->s_groups_count;
2657
2658 for (group = elr->lr_next_group; group < ngroups; group++) {
2659 gdp = ext4_get_group_desc(sb, group, NULL);
2660 if (!gdp) {
2661 ret = 1;
2662 break;
2663 }
2664
2665 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2666 break;
2667 }
2668
2669 if (group == ngroups)
2670 ret = 1;
2671
2672 if (!ret) {
2673 timeout = jiffies;
2674 ret = ext4_init_inode_table(sb, group,
2675 elr->lr_timeout ? 0 : 1);
2676 if (elr->lr_timeout == 0) {
2677 timeout = jiffies - timeout;
2678 if (elr->lr_sbi->s_li_wait_mult)
2679 timeout *= elr->lr_sbi->s_li_wait_mult;
2680 else
2681 timeout *= 20;
2682 elr->lr_timeout = timeout;
2683 }
2684 elr->lr_next_sched = jiffies + elr->lr_timeout;
2685 elr->lr_next_group = group + 1;
2686 }
2687
2688 return ret;
2689}
2690
2691/*
2692 * Remove lr_request from the list_request and free the
2693 * request tructure. Should be called with li_list_mtx held
2694 */
2695static void ext4_remove_li_request(struct ext4_li_request *elr)
2696{
2697 struct ext4_sb_info *sbi;
2698
2699 if (!elr)
2700 return;
2701
2702 sbi = elr->lr_sbi;
2703
2704 list_del(&elr->lr_request);
2705 sbi->s_li_request = NULL;
2706 kfree(elr);
2707}
2708
2709static void ext4_unregister_li_request(struct super_block *sb)
2710{
2711 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2712
2713 if (!ext4_li_info)
2714 return;
2715
2716 mutex_lock(&ext4_li_info->li_list_mtx);
2717 ext4_remove_li_request(elr);
2718 mutex_unlock(&ext4_li_info->li_list_mtx);
2719}
2720
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002721static struct task_struct *ext4_lazyinit_task;
2722
Lukas Czernerbfff6872010-10-27 21:30:05 -04002723/*
2724 * This is the function where ext4lazyinit thread lives. It walks
2725 * through the request list searching for next scheduled filesystem.
2726 * When such a fs is found, run the lazy initialization request
2727 * (ext4_rn_li_request) and keep track of the time spend in this
2728 * function. Based on that time we compute next schedule time of
2729 * the request. When walking through the list is complete, compute
2730 * next waking time and put itself into sleep.
2731 */
2732static int ext4_lazyinit_thread(void *arg)
2733{
2734 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2735 struct list_head *pos, *n;
2736 struct ext4_li_request *elr;
2737 unsigned long next_wakeup;
2738 DEFINE_WAIT(wait);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002739
2740 BUG_ON(NULL == eli);
2741
2742 eli->li_timer.data = (unsigned long)current;
2743 eli->li_timer.function = ext4_lazyinode_timeout;
2744
2745 eli->li_task = current;
2746 wake_up(&eli->li_wait_task);
2747
2748cont_thread:
2749 while (true) {
2750 next_wakeup = MAX_JIFFY_OFFSET;
2751
2752 mutex_lock(&eli->li_list_mtx);
2753 if (list_empty(&eli->li_request_list)) {
2754 mutex_unlock(&eli->li_list_mtx);
2755 goto exit_thread;
2756 }
2757
2758 list_for_each_safe(pos, n, &eli->li_request_list) {
2759 elr = list_entry(pos, struct ext4_li_request,
2760 lr_request);
2761
Theodore Ts'ob2c78cd2010-11-02 14:19:30 -04002762 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2763 if (ext4_run_li_request(elr) != 0) {
2764 /* error, remove the lazy_init job */
2765 ext4_remove_li_request(elr);
2766 continue;
2767 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002768 }
2769
2770 if (time_before(elr->lr_next_sched, next_wakeup))
2771 next_wakeup = elr->lr_next_sched;
2772 }
2773 mutex_unlock(&eli->li_list_mtx);
2774
2775 if (freezing(current))
2776 refrigerator();
2777
Lukas Czernerf4245bd2010-11-02 14:07:17 -04002778 if ((time_after_eq(jiffies, next_wakeup)) ||
2779 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04002780 cond_resched();
2781 continue;
2782 }
2783
2784 eli->li_timer.expires = next_wakeup;
2785 add_timer(&eli->li_timer);
2786 prepare_to_wait(&eli->li_wait_daemon, &wait,
2787 TASK_INTERRUPTIBLE);
2788 if (time_before(jiffies, next_wakeup))
2789 schedule();
2790 finish_wait(&eli->li_wait_daemon, &wait);
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002791 if (kthread_should_stop()) {
2792 ext4_clear_request_list();
2793 goto exit_thread;
2794 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002795 }
2796
2797exit_thread:
2798 /*
2799 * It looks like the request list is empty, but we need
2800 * to check it under the li_list_mtx lock, to prevent any
2801 * additions into it, and of course we should lock ext4_li_mtx
2802 * to atomically free the list and ext4_li_info, because at
2803 * this point another ext4 filesystem could be registering
2804 * new one.
2805 */
2806 mutex_lock(&ext4_li_mtx);
2807 mutex_lock(&eli->li_list_mtx);
2808 if (!list_empty(&eli->li_request_list)) {
2809 mutex_unlock(&eli->li_list_mtx);
2810 mutex_unlock(&ext4_li_mtx);
2811 goto cont_thread;
2812 }
2813 mutex_unlock(&eli->li_list_mtx);
2814 del_timer_sync(&ext4_li_info->li_timer);
2815 eli->li_task = NULL;
2816 wake_up(&eli->li_wait_task);
2817
2818 kfree(ext4_li_info);
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002819 ext4_lazyinit_task = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002820 ext4_li_info = NULL;
2821 mutex_unlock(&ext4_li_mtx);
2822
2823 return 0;
2824}
2825
2826static void ext4_clear_request_list(void)
2827{
2828 struct list_head *pos, *n;
2829 struct ext4_li_request *elr;
2830
2831 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002832 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2833 elr = list_entry(pos, struct ext4_li_request,
2834 lr_request);
2835 ext4_remove_li_request(elr);
2836 }
2837 mutex_unlock(&ext4_li_info->li_list_mtx);
2838}
2839
2840static int ext4_run_lazyinit_thread(void)
2841{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002842 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2843 ext4_li_info, "ext4lazyinit");
2844 if (IS_ERR(ext4_lazyinit_task)) {
2845 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002846 ext4_clear_request_list();
2847 del_timer_sync(&ext4_li_info->li_timer);
2848 kfree(ext4_li_info);
2849 ext4_li_info = NULL;
2850 printk(KERN_CRIT "EXT4: error %d creating inode table "
2851 "initialization thread\n",
2852 err);
2853 return err;
2854 }
2855 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2856
2857 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2858 return 0;
2859}
2860
2861/*
2862 * Check whether it make sense to run itable init. thread or not.
2863 * If there is at least one uninitialized inode table, return
2864 * corresponding group number, else the loop goes through all
2865 * groups and return total number of groups.
2866 */
2867static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2868{
2869 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2870 struct ext4_group_desc *gdp = NULL;
2871
2872 for (group = 0; group < ngroups; group++) {
2873 gdp = ext4_get_group_desc(sb, group, NULL);
2874 if (!gdp)
2875 continue;
2876
2877 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2878 break;
2879 }
2880
2881 return group;
2882}
2883
2884static int ext4_li_info_new(void)
2885{
2886 struct ext4_lazy_init *eli = NULL;
2887
2888 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2889 if (!eli)
2890 return -ENOMEM;
2891
2892 eli->li_task = NULL;
2893 INIT_LIST_HEAD(&eli->li_request_list);
2894 mutex_init(&eli->li_list_mtx);
2895
2896 init_waitqueue_head(&eli->li_wait_daemon);
2897 init_waitqueue_head(&eli->li_wait_task);
2898 init_timer(&eli->li_timer);
2899 eli->li_state |= EXT4_LAZYINIT_QUIT;
2900
2901 ext4_li_info = eli;
2902
2903 return 0;
2904}
2905
2906static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2907 ext4_group_t start)
2908{
2909 struct ext4_sb_info *sbi = EXT4_SB(sb);
2910 struct ext4_li_request *elr;
2911 unsigned long rnd;
2912
2913 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2914 if (!elr)
2915 return NULL;
2916
2917 elr->lr_super = sb;
2918 elr->lr_sbi = sbi;
2919 elr->lr_next_group = start;
2920
2921 /*
2922 * Randomize first schedule time of the request to
2923 * spread the inode table initialization requests
2924 * better.
2925 */
2926 get_random_bytes(&rnd, sizeof(rnd));
2927 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2928 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2929
2930 return elr;
2931}
2932
2933static int ext4_register_li_request(struct super_block *sb,
2934 ext4_group_t first_not_zeroed)
2935{
2936 struct ext4_sb_info *sbi = EXT4_SB(sb);
2937 struct ext4_li_request *elr;
2938 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05002939 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002940
2941 if (sbi->s_li_request != NULL)
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002942 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002943
2944 if (first_not_zeroed == ngroups ||
2945 (sb->s_flags & MS_RDONLY) ||
2946 !test_opt(sb, INIT_INODE_TABLE)) {
2947 sbi->s_li_request = NULL;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002948 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002949 }
2950
2951 if (first_not_zeroed == ngroups) {
2952 sbi->s_li_request = NULL;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002953 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002954 }
2955
2956 elr = ext4_li_request_new(sb, first_not_zeroed);
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002957 if (!elr)
2958 return -ENOMEM;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002959
2960 mutex_lock(&ext4_li_mtx);
2961
2962 if (NULL == ext4_li_info) {
2963 ret = ext4_li_info_new();
2964 if (ret)
2965 goto out;
2966 }
2967
2968 mutex_lock(&ext4_li_info->li_list_mtx);
2969 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2970 mutex_unlock(&ext4_li_info->li_list_mtx);
2971
2972 sbi->s_li_request = elr;
2973
2974 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2975 ret = ext4_run_lazyinit_thread();
2976 if (ret)
2977 goto out;
2978 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002979out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002980 mutex_unlock(&ext4_li_mtx);
2981 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002982 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002983 return ret;
2984}
2985
2986/*
2987 * We do not need to lock anything since this is called on
2988 * module unload.
2989 */
2990static void ext4_destroy_lazyinit_thread(void)
2991{
2992 /*
2993 * If thread exited earlier
2994 * there's nothing to be done.
2995 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002996 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002997 return;
2998
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002999 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003000}
3001
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003002static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04003003 __releases(kernel_lock)
3004 __acquires(kernel_lock)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003005{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003006 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003007 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003008 struct ext4_super_block *es = NULL;
3009 struct ext4_sb_info *sbi;
3010 ext4_fsblk_t block;
3011 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003012 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003013 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003014 unsigned long journal_devnum = 0;
3015 unsigned long def_mount_opts;
3016 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003017 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05003018 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003019 int ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003020 int blocksize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003021 unsigned int db_count;
3022 unsigned int i;
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003023 int needs_recovery, has_huge_files;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003024 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07003025 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003026 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003027 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003028
3029 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3030 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003031 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003032
3033 sbi->s_blockgroup_lock =
3034 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3035 if (!sbi->s_blockgroup_lock) {
3036 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003037 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003038 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003039 sb->s_fs_info = sbi;
3040 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07003041 sbi->s_resuid = EXT4_DEF_RESUID;
3042 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003043 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003044 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003045 if (sb->s_bdev->bd_part)
3046 sbi->s_sectors_written_start =
3047 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003048
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003049 /* Cleanup superblock name */
3050 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3051 *cp = '!';
3052
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003053 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003054 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003055 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003056 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003057 goto out_fail;
3058 }
3059
3060 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003061 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003062 * block sizes. We need to calculate the offset from buffer start.
3063 */
Mingming Cao617ba132006-10-11 01:20:53 -07003064 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003065 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3066 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003067 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003068 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003069 }
3070
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003071 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003072 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003073 goto out_fail;
3074 }
3075 /*
3076 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003077 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003078 */
Mingming Cao617ba132006-10-11 01:20:53 -07003079 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003080 sbi->s_es = es;
3081 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003082 if (sb->s_magic != EXT4_SUPER_MAGIC)
3083 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003084 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003085
3086 /* Set defaults before we parse the mount options */
3087 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003088 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003089 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003090 set_opt(sb, DEBUG);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003091 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3092 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3093 "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003094 set_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003095 }
Mingming Cao617ba132006-10-11 01:20:53 -07003096 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003097 set_opt(sb, NO_UID32);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003098#ifdef CONFIG_EXT4_FS_XATTR
Mingming Cao617ba132006-10-11 01:20:53 -07003099 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003100 set_opt(sb, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003101#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04003102#ifdef CONFIG_EXT4_FS_POSIX_ACL
Mingming Cao617ba132006-10-11 01:20:53 -07003103 if (def_mount_opts & EXT4_DEFM_ACL)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003104 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003105#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003106 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003107 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003108 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003109 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003110 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003111 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003112
Mingming Cao617ba132006-10-11 01:20:53 -07003113 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003114 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003115 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003116 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003117 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003118 set_opt(sb, ERRORS_RO);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003119 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003120 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003121 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003122 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003123
3124 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3125 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05003126 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3127 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3128 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003129
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003130 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003131 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003132
Mingming Cao1e2462f2007-07-18 09:00:55 -04003133 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003134 * enable delayed allocation by default
3135 * Use -o nodelalloc to turn it off
3136 */
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003137 if (!IS_EXT3_SB(sb) &&
3138 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003139 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003140
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003141 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3142 &journal_devnum, &journal_ioprio, NULL, 0)) {
3143 ext4_msg(sb, KERN_WARNING,
3144 "failed to parse options in superblock: %s",
3145 sbi->s_es->s_mount_opts);
3146 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003147 if (!parse_options((char *) data, sb, &journal_devnum,
3148 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003149 goto failed_mount;
3150
3151 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003152 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003153
Mingming Cao617ba132006-10-11 01:20:53 -07003154 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3155 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3156 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3157 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003158 ext4_msg(sb, KERN_WARNING,
3159 "feature flags set on rev 0 fs, "
3160 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003161
3162 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003163 * Check feature flags regardless of the revision level, since we
3164 * previously didn't change the revision level when setting the flags,
3165 * so there is a chance incompat flags are set on a rev 0 filesystem.
3166 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003167 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003168 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003169
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003170 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3171
Mingming Cao617ba132006-10-11 01:20:53 -07003172 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3173 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003174 ext4_msg(sb, KERN_ERR,
3175 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003176 goto failed_mount;
3177 }
3178
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003179 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003180 /* Validate the filesystem blocksize */
3181 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003182 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003183 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003184 goto failed_mount;
3185 }
3186
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003187 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003188 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3189 offset = do_div(logical_sb_block, blocksize);
3190 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003191 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003192 ext4_msg(sb, KERN_ERR,
3193 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003194 goto failed_mount;
3195 }
Mingming Cao617ba132006-10-11 01:20:53 -07003196 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003197 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003198 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003199 ext4_msg(sb, KERN_ERR,
3200 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003201 goto failed_mount;
3202 }
3203 }
3204
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003205 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3206 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003207 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3208 has_huge_files);
3209 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003210
Mingming Cao617ba132006-10-11 01:20:53 -07003211 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3212 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3213 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003214 } else {
3215 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3216 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003217 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003218 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003219 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003220 ext4_msg(sb, KERN_ERR,
3221 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003222 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003223 goto failed_mount;
3224 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003225 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3226 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003227 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003228
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003229 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3230 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003231 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003232 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003233 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003234 ext4_msg(sb, KERN_ERR,
3235 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003236 sbi->s_desc_size);
3237 goto failed_mount;
3238 }
3239 } else
3240 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003241
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003242 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003243 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003244 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003245 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003246
Mingming Cao617ba132006-10-11 01:20:53 -07003247 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003248 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003249 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003250 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3251 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003252 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003253 sbi->s_sbh = bh;
3254 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003255 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3256 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003257
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003258 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003259 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3260 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003261 i = le32_to_cpu(es->s_flags);
3262 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3263 sbi->s_hash_unsigned = 3;
3264 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3265#ifdef __CHAR_UNSIGNED__
3266 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3267 sbi->s_hash_unsigned = 3;
3268#else
3269 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3270#endif
3271 sb->s_dirt = 1;
3272 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003273
3274 if (sbi->s_blocks_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003275 ext4_msg(sb, KERN_ERR,
3276 "#blocks per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003277 sbi->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003278 goto failed_mount;
3279 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003280 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003281 ext4_msg(sb, KERN_ERR,
3282 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003283 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003284 goto failed_mount;
3285 }
3286
Eric Sandeenbf43d842009-08-17 23:48:51 -04003287 /*
3288 * Test whether we have more sectors than will fit in sector_t,
3289 * and whether the max offset is addressable by the page cache.
3290 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003291 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003292 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003293 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003294 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003295 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003296 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003297 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003298 ret = err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003299 goto failed_mount;
3300 }
3301
Mingming Cao617ba132006-10-11 01:20:53 -07003302 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3303 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003304
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003305 /* check blocks count against device size */
3306 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3307 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003308 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3309 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003310 ext4_blocks_count(es), blocks_count);
3311 goto failed_mount;
3312 }
3313
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003314 /*
3315 * It makes no sense for the first data block to be beyond the end
3316 * of the filesystem.
3317 */
3318 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003319 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3320 "block %u is beyond end of filesystem (%llu)",
3321 le32_to_cpu(es->s_first_data_block),
3322 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003323 goto failed_mount;
3324 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003325 blocks_count = (ext4_blocks_count(es) -
3326 le32_to_cpu(es->s_first_data_block) +
3327 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3328 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003329 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003330 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003331 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003332 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003333 ext4_blocks_count(es),
3334 le32_to_cpu(es->s_first_data_block),
3335 EXT4_BLOCKS_PER_GROUP(sb));
3336 goto failed_mount;
3337 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003338 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003339 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3340 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003341 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3342 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003343 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003344 GFP_KERNEL);
3345 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003346 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003347 goto failed_mount;
3348 }
3349
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04003350#ifdef CONFIG_PROC_FS
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003351 if (ext4_proc_root)
3352 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04003353#endif
Theodore Ts'o240799c2008-10-09 23:53:47 -04003354
Pekka Enberg705895b2009-02-15 18:07:52 -05003355 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003356
3357 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003358 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003359 sbi->s_group_desc[i] = sb_bread(sb, block);
3360 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003361 ext4_msg(sb, KERN_ERR,
3362 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003363 db_count = i;
3364 goto failed_mount2;
3365 }
3366 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003367 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003368 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003369 goto failed_mount2;
3370 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003371 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3372 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003373 ext4_msg(sb, KERN_ERR,
3374 "unable to initialize "
3375 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003376 goto failed_mount2;
3377 }
3378
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003379 sbi->s_gdb_count = db_count;
3380 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3381 spin_lock_init(&sbi->s_next_gen_lock);
3382
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003383 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3384 ext4_count_free_blocks(sb));
3385 if (!err) {
3386 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3387 ext4_count_free_inodes(sb));
3388 }
3389 if (!err) {
3390 err = percpu_counter_init(&sbi->s_dirs_counter,
3391 ext4_count_dirs(sb));
3392 }
3393 if (!err) {
3394 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3395 }
3396 if (err) {
3397 ext4_msg(sb, KERN_ERR, "insufficient memory");
3398 goto failed_mount3;
3399 }
3400
Alex Tomasc9de5602008-01-29 00:19:52 -05003401 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04003402 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05003403
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003404 /*
3405 * set up enough so that it can read an inode
3406 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003407 if (!test_opt(sb, NOLOAD) &&
3408 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3409 sb->s_op = &ext4_sops;
3410 else
3411 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07003412 sb->s_export_op = &ext4_export_ops;
3413 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003414#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003415 sb->s_qcop = &ext4_qctl_operations;
3416 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003417#endif
3418 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003419 mutex_init(&sbi->s_orphan_lock);
Theodore Ts'o32ed5052009-04-25 22:53:39 -04003420 mutex_init(&sbi->s_resize_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003421
3422 sb->s_root = NULL;
3423
3424 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07003425 EXT4_HAS_INCOMPAT_FEATURE(sb,
3426 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003427
3428 /*
3429 * The first inode we look at is the journal inode. Don't try
3430 * root first: it may be modified in the journal!
3431 */
3432 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07003433 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3434 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003435 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05003436 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3437 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003438 ext4_msg(sb, KERN_ERR, "required journal recovery "
3439 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003440 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003441 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003442 clear_opt(sb, DATA_FLAGS);
3443 set_opt(sb, WRITEBACK_DATA);
Frank Mayhar03901312009-01-07 00:06:22 -05003444 sbi->s_journal = NULL;
3445 needs_recovery = 0;
3446 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003447 }
3448
Jose R. Santoseb40a092007-07-18 08:37:25 -04003449 if (ext4_blocks_count(es) > 0xffffffffULL &&
3450 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3451 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003452 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003453 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04003454 }
3455
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003456 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3457 jbd2_journal_set_features(sbi->s_journal,
3458 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05003459 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003460 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3461 jbd2_journal_set_features(sbi->s_journal,
3462 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
Girish Shilamkar818d2762008-01-28 23:58:27 -05003463 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3464 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003465 } else {
3466 jbd2_journal_clear_features(sbi->s_journal,
3467 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3468 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3469 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05003470
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003471 /* We have now updated the journal if required, so we can
3472 * validate the data journaling mode. */
3473 switch (test_opt(sb, DATA_FLAGS)) {
3474 case 0:
3475 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07003476 * capabilities: ORDERED_DATA if the journal can
3477 * cope, else JOURNAL_DATA
3478 */
Mingming Caodab291a2006-10-11 01:21:01 -07003479 if (jbd2_journal_check_available_features
3480 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003481 set_opt(sb, ORDERED_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003482 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003483 set_opt(sb, JOURNAL_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003484 break;
3485
Mingming Cao617ba132006-10-11 01:20:53 -07003486 case EXT4_MOUNT_ORDERED_DATA:
3487 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07003488 if (!jbd2_journal_check_available_features
3489 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003490 ext4_msg(sb, KERN_ERR, "Journal does not support "
3491 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003492 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003493 }
3494 default:
3495 break;
3496 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003497 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003498
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003499 /*
3500 * The journal may have updated the bg summary counts, so we
3501 * need to update the global counters.
3502 */
3503 percpu_counter_set(&sbi->s_freeblocks_counter,
3504 ext4_count_free_blocks(sb));
3505 percpu_counter_set(&sbi->s_freeinodes_counter,
3506 ext4_count_free_inodes(sb));
3507 percpu_counter_set(&sbi->s_dirs_counter,
3508 ext4_count_dirs(sb));
3509 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04003510
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003511no_journal:
Mingming Cao4c0425f2009-09-28 15:48:41 -04003512 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3513 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3514 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3515 goto failed_mount_wq;
3516 }
3517
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003518 /*
Mingming Caodab291a2006-10-11 01:21:01 -07003519 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003520 * so we can safely mount the rest of the filesystem now.
3521 */
3522
David Howells1d1fe1e2008-02-07 00:15:37 -08003523 root = ext4_iget(sb, EXT4_ROOT_INO);
3524 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003525 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003526 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003527 goto failed_mount4;
3528 }
3529 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08003530 iput(root);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003531 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003532 goto failed_mount4;
3533 }
David Howells1d1fe1e2008-02-07 00:15:37 -08003534 sb->s_root = d_alloc_root(root);
3535 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003536 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003537 iput(root);
3538 ret = -ENOMEM;
3539 goto failed_mount4;
3540 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003541
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003542 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04003543
3544 /* determine the minimum size of new large inodes, if present */
3545 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3546 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3547 EXT4_GOOD_OLD_INODE_SIZE;
3548 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3549 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3550 if (sbi->s_want_extra_isize <
3551 le16_to_cpu(es->s_want_extra_isize))
3552 sbi->s_want_extra_isize =
3553 le16_to_cpu(es->s_want_extra_isize);
3554 if (sbi->s_want_extra_isize <
3555 le16_to_cpu(es->s_min_extra_isize))
3556 sbi->s_want_extra_isize =
3557 le16_to_cpu(es->s_min_extra_isize);
3558 }
3559 }
3560 /* Check if enough inode space is available */
3561 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3562 sbi->s_inode_size) {
3563 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3564 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04003565 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3566 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04003567 }
3568
Theodore Ts'o90576c02009-09-29 15:51:30 -04003569 if (test_opt(sb, DELALLOC) &&
3570 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003571 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3572 "requested data journaling mode");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003573 clear_opt(sb, DELALLOC);
Theodore Ts'o90576c02009-09-29 15:51:30 -04003574 }
Jiaying Zhang744692d2010-03-04 16:14:02 -05003575 if (test_opt(sb, DIOREAD_NOLOCK)) {
3576 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3577 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3578 "option - requested data journaling mode");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003579 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05003580 }
3581 if (sb->s_blocksize < PAGE_SIZE) {
3582 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3583 "option - block size is too small");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003584 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05003585 }
3586 }
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003587
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003588 err = ext4_setup_system_zone(sb);
3589 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003590 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04003591 "zone (%d)", err);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003592 goto failed_mount4;
3593 }
3594
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003595 ext4_ext_init(sb);
3596 err = ext4_mb_init(sb, needs_recovery);
3597 if (err) {
Uwe Kleine-König421f91d2010-06-11 12:17:00 +02003598 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
Eric Sandeenb31e1552009-06-04 17:36:36 -04003599 err);
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003600 goto failed_mount4;
3601 }
3602
Lukas Czernerbfff6872010-10-27 21:30:05 -04003603 err = ext4_register_li_request(sb, first_not_zeroed);
3604 if (err)
3605 goto failed_mount4;
3606
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003607 sbi->s_kobj.kset = ext4_kset;
3608 init_completion(&sbi->s_kobj_unregister);
3609 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3610 "%s", sb->s_id);
3611 if (err) {
3612 ext4_mb_release(sb);
3613 ext4_ext_release(sb);
3614 goto failed_mount4;
3615 };
3616
Mingming Cao617ba132006-10-11 01:20:53 -07003617 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3618 ext4_orphan_cleanup(sb, es);
3619 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05003620 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003621 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05003622 ext4_mark_recovery_complete(sb, es);
3623 }
3624 if (EXT4_SB(sb)->s_journal) {
3625 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3626 descr = " journalled data mode";
3627 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3628 descr = " ordered data mode";
3629 else
3630 descr = " writeback data mode";
3631 } else
3632 descr = "out journal";
3633
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003634 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003635 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3636 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003637
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003638 init_timer(&sbi->s_err_report);
3639 sbi->s_err_report.function = print_daily_error_info;
3640 sbi->s_err_report.data = (unsigned long) sb;
3641 if (es->s_error_count)
3642 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003643
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003644 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003645 return 0;
3646
Mingming Cao617ba132006-10-11 01:20:53 -07003647cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003648 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003649 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003650 goto failed_mount;
3651
3652failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04003653 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04003654 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3655failed_mount_wq:
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003656 ext4_release_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003657 if (sbi->s_journal) {
3658 jbd2_journal_destroy(sbi->s_journal);
3659 sbi->s_journal = NULL;
3660 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003661failed_mount3:
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04003662 if (sbi->s_flex_groups) {
3663 if (is_vmalloc_addr(sbi->s_flex_groups))
3664 vfree(sbi->s_flex_groups);
3665 else
3666 kfree(sbi->s_flex_groups);
3667 }
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003668 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3669 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3670 percpu_counter_destroy(&sbi->s_dirs_counter);
3671 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003672failed_mount2:
3673 for (i = 0; i < db_count; i++)
3674 brelse(sbi->s_group_desc[i]);
3675 kfree(sbi->s_group_desc);
3676failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04003677 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003678 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003679 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003680#ifdef CONFIG_QUOTA
3681 for (i = 0; i < MAXQUOTAS; i++)
3682 kfree(sbi->s_qf_names[i]);
3683#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003684 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003685 brelse(bh);
3686out_fail:
3687 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04003688 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003689 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003690out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003691 kfree(orig_data);
David Howells1d1fe1e2008-02-07 00:15:37 -08003692 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003693}
3694
3695/*
3696 * Setup any per-fs journal parameters now. We'll do this both on
3697 * initial mount, once the journal has been initialised but before we've
3698 * done any recovery; and again on any subsequent remount.
3699 */
Mingming Cao617ba132006-10-11 01:20:53 -07003700static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003701{
Mingming Cao617ba132006-10-11 01:20:53 -07003702 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003703
Theodore Ts'o30773842009-01-03 20:27:38 -05003704 journal->j_commit_interval = sbi->s_commit_interval;
3705 journal->j_min_batch_time = sbi->s_min_batch_time;
3706 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003707
Theodore Ts'oa931da62010-08-03 21:35:12 -04003708 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003709 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07003710 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003711 else
Mingming Caodab291a2006-10-11 01:21:01 -07003712 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04003713 if (test_opt(sb, DATA_ERR_ABORT))
3714 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3715 else
3716 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04003717 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003718}
3719
Mingming Cao617ba132006-10-11 01:20:53 -07003720static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003721 unsigned int journal_inum)
3722{
3723 struct inode *journal_inode;
3724 journal_t *journal;
3725
Frank Mayhar03901312009-01-07 00:06:22 -05003726 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3727
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003728 /* First, test for the existence of a valid inode on disk. Bad
3729 * things happen if we iget() an unused inode, as the subsequent
3730 * iput() will try to delete it. */
3731
David Howells1d1fe1e2008-02-07 00:15:37 -08003732 journal_inode = ext4_iget(sb, journal_inum);
3733 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003734 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003735 return NULL;
3736 }
3737 if (!journal_inode->i_nlink) {
3738 make_bad_inode(journal_inode);
3739 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003740 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003741 return NULL;
3742 }
3743
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003744 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003745 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003746 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003747 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003748 iput(journal_inode);
3749 return NULL;
3750 }
3751
Mingming Caodab291a2006-10-11 01:21:01 -07003752 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003753 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003754 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003755 iput(journal_inode);
3756 return NULL;
3757 }
3758 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003759 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003760 return journal;
3761}
3762
Mingming Cao617ba132006-10-11 01:20:53 -07003763static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003764 dev_t j_dev)
3765{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003766 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003767 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003768 ext4_fsblk_t start;
3769 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003770 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003771 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003772 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003773 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003774 struct block_device *bdev;
3775
Frank Mayhar03901312009-01-07 00:06:22 -05003776 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3777
Eric Sandeenb31e1552009-06-04 17:36:36 -04003778 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003779 if (bdev == NULL)
3780 return NULL;
3781
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003782 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003783 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003784 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003785 ext4_msg(sb, KERN_ERR,
3786 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003787 goto out_bdev;
3788 }
3789
Mingming Cao617ba132006-10-11 01:20:53 -07003790 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3791 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003792 set_blocksize(bdev, blocksize);
3793 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003794 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3795 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003796 goto out_bdev;
3797 }
3798
Mingming Cao617ba132006-10-11 01:20:53 -07003799 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3800 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003801 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003802 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003803 ext4_msg(sb, KERN_ERR, "external journal has "
3804 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003805 brelse(bh);
3806 goto out_bdev;
3807 }
3808
Mingming Cao617ba132006-10-11 01:20:53 -07003809 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003810 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003811 brelse(bh);
3812 goto out_bdev;
3813 }
3814
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003815 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003816 start = sb_block + 1;
3817 brelse(bh); /* we're done with the superblock */
3818
Mingming Caodab291a2006-10-11 01:21:01 -07003819 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003820 start, len, blocksize);
3821 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003822 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003823 goto out_bdev;
3824 }
3825 journal->j_private = sb;
3826 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3827 wait_on_buffer(journal->j_sb_buffer);
3828 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003829 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003830 goto out_journal;
3831 }
3832 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003833 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3834 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003835 be32_to_cpu(journal->j_superblock->s_nr_users));
3836 goto out_journal;
3837 }
Mingming Cao617ba132006-10-11 01:20:53 -07003838 EXT4_SB(sb)->journal_bdev = bdev;
3839 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003840 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003841
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003842out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07003843 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003844out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07003845 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003846 return NULL;
3847}
3848
Mingming Cao617ba132006-10-11 01:20:53 -07003849static int ext4_load_journal(struct super_block *sb,
3850 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003851 unsigned long journal_devnum)
3852{
3853 journal_t *journal;
3854 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3855 dev_t journal_dev;
3856 int err = 0;
3857 int really_read_only;
3858
Frank Mayhar03901312009-01-07 00:06:22 -05003859 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3860
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003861 if (journal_devnum &&
3862 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003863 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3864 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003865 journal_dev = new_decode_dev(journal_devnum);
3866 } else
3867 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3868
3869 really_read_only = bdev_read_only(sb->s_bdev);
3870
3871 /*
3872 * Are we loading a blank journal or performing recovery after a
3873 * crash? For recovery, we need to check in advance whether we
3874 * can get read-write access to the device.
3875 */
Mingming Cao617ba132006-10-11 01:20:53 -07003876 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003877 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003878 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3879 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003880 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003881 ext4_msg(sb, KERN_ERR, "write access "
3882 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003883 return -EROFS;
3884 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04003885 ext4_msg(sb, KERN_INFO, "write access will "
3886 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003887 }
3888 }
3889
3890 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003891 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3892 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003893 return -EINVAL;
3894 }
3895
3896 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07003897 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003898 return -EINVAL;
3899 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07003900 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003901 return -EINVAL;
3902 }
3903
Theodore Ts'o90576c02009-09-29 15:51:30 -04003904 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003905 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04003906
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003907 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07003908 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003909 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003910 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003911 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003912 return err;
3913 }
3914 }
3915
Mingming Cao617ba132006-10-11 01:20:53 -07003916 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07003917 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04003918 if (!err) {
3919 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3920 if (save)
3921 memcpy(save, ((char *) es) +
3922 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07003923 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04003924 if (save)
3925 memcpy(((char *) es) + EXT4_S_ERR_START,
3926 save, EXT4_S_ERR_LEN);
3927 kfree(save);
3928 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003929
3930 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003931 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003932 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003933 return err;
3934 }
3935
Mingming Cao617ba132006-10-11 01:20:53 -07003936 EXT4_SB(sb)->s_journal = journal;
3937 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003938
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04003939 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003940 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3941 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003942
3943 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003944 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003945 }
3946
3947 return 0;
3948}
3949
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003950static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003951{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003952 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07003953 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003954 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003955
3956 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08003957 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04003958 if (buffer_write_io_error(sbh)) {
3959 /*
3960 * Oh, dear. A previous attempt to write the
3961 * superblock failed. This could happen because the
3962 * USB device was yanked out. Or it could happen to
3963 * be a transient write error and maybe the block will
3964 * be remapped. Nothing we can do but to retry the
3965 * write and hope for the best.
3966 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04003967 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3968 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003969 clear_buffer_write_io_error(sbh);
3970 set_buffer_uptodate(sbh);
3971 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04003972 /*
3973 * If the file system is mounted read-only, don't update the
3974 * superblock write time. This avoids updating the superblock
3975 * write time when we are mounting the root file system
3976 * read/only but we need to replay the journal; at that point,
3977 * for people who are east of GMT and who make their clock
3978 * tick in localtime for Windows bug-for-bug compatibility,
3979 * the clock is set in the future, and this will cause e2fsck
3980 * to complain and force a full file system check.
3981 */
3982 if (!(sb->s_flags & MS_RDONLY))
3983 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003984 if (sb->s_bdev->bd_part)
3985 es->s_kbytes_written =
3986 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003987 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3988 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003989 else
3990 es->s_kbytes_written =
3991 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003992 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3993 &EXT4_SB(sb)->s_freeblocks_counter));
3994 es->s_free_inodes_count =
3995 cpu_to_le32(percpu_counter_sum_positive(
3996 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003997 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003998 BUFFER_TRACE(sbh, "marking dirty");
3999 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04004000 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08004001 error = sync_dirty_buffer(sbh);
4002 if (error)
4003 return error;
4004
4005 error = buffer_write_io_error(sbh);
4006 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004007 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4008 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004009 clear_buffer_write_io_error(sbh);
4010 set_buffer_uptodate(sbh);
4011 }
4012 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08004013 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004014}
4015
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004016/*
4017 * Have we just finished recovery? If so, and if we are mounting (or
4018 * remounting) the filesystem readonly, then we will end up with a
4019 * consistent fs on disk. Record that fact.
4020 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004021static void ext4_mark_recovery_complete(struct super_block *sb,
4022 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004023{
Mingming Cao617ba132006-10-11 01:20:53 -07004024 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004025
Frank Mayhar03901312009-01-07 00:06:22 -05004026 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4027 BUG_ON(journal != NULL);
4028 return;
4029 }
Mingming Caodab291a2006-10-11 01:21:01 -07004030 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004031 if (jbd2_journal_flush(journal) < 0)
4032 goto out;
4033
Mingming Cao617ba132006-10-11 01:20:53 -07004034 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004035 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07004036 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004037 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004038 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004039
4040out:
Mingming Caodab291a2006-10-11 01:21:01 -07004041 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004042}
4043
4044/*
4045 * If we are mounting (or read-write remounting) a filesystem whose journal
4046 * has recorded an error from a previous lifetime, move that error to the
4047 * main filesystem now.
4048 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004049static void ext4_clear_journal_err(struct super_block *sb,
4050 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004051{
4052 journal_t *journal;
4053 int j_errno;
4054 const char *errstr;
4055
Frank Mayhar03901312009-01-07 00:06:22 -05004056 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4057
Mingming Cao617ba132006-10-11 01:20:53 -07004058 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004059
4060 /*
4061 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004062 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004063 */
4064
Mingming Caodab291a2006-10-11 01:21:01 -07004065 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004066 if (j_errno) {
4067 char nbuf[16];
4068
Mingming Cao617ba132006-10-11 01:20:53 -07004069 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004070 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004071 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004072 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004073
Mingming Cao617ba132006-10-11 01:20:53 -07004074 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4075 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004076 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004077
Mingming Caodab291a2006-10-11 01:21:01 -07004078 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004079 }
4080}
4081
4082/*
4083 * Force the running and committing transactions to commit,
4084 * and wait on the commit.
4085 */
Mingming Cao617ba132006-10-11 01:20:53 -07004086int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004087{
4088 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05004089 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004090
4091 if (sb->s_flags & MS_RDONLY)
4092 return 0;
4093
Mingming Cao617ba132006-10-11 01:20:53 -07004094 journal = EXT4_SB(sb)->s_journal;
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004095 if (journal) {
Eric Sandeen437f88c2010-08-01 17:33:29 -04004096 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Frank Mayhar03901312009-01-07 00:06:22 -05004097 ret = ext4_journal_force_commit(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004098 }
Frank Mayhar03901312009-01-07 00:06:22 -05004099
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004100 return ret;
4101}
4102
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004103static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004104{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004105 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004106 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004107 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004108}
4109
Mingming Cao617ba132006-10-11 01:20:53 -07004110static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004111{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004112 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004113 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004114 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004115
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004116 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004117 flush_workqueue(sbi->dio_unwritten_wq);
4118 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004119 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004120 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05004121 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004122 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004123}
4124
4125/*
4126 * LVM calls this function before a (read-only) snapshot is created. This
4127 * gives us a chance to flush the journal completely and mark the fs clean.
4128 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004129static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004130{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004131 int error = 0;
4132 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004133
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004134 if (sb->s_flags & MS_RDONLY)
4135 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004136
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004137 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004138
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004139 /* Now we set up the journal barrier. */
4140 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004141
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004142 /*
4143 * Don't clear the needs_recovery flag if we failed to flush
4144 * the journal.
4145 */
4146 error = jbd2_journal_flush(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004147 if (error < 0)
4148 goto out;
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004149
4150 /* Journal blocked and flushed, clear needs_recovery flag. */
4151 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4152 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004153out:
4154 /* we rely on s_frozen to stop further updates */
4155 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4156 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004157}
4158
4159/*
4160 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4161 * flag here, even though the filesystem is not technically dirty yet.
4162 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004163static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004164{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004165 if (sb->s_flags & MS_RDONLY)
4166 return 0;
4167
4168 lock_super(sb);
4169 /* Reset the needs_recovery flag before the fs is unlocked. */
4170 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4171 ext4_commit_super(sb, 1);
4172 unlock_super(sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004173 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004174}
4175
Theodore Ts'o673c6102010-12-15 20:28:48 -05004176/*
4177 * Structure to save mount options for ext4_remount's benefit
4178 */
4179struct ext4_mount_options {
4180 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004181 unsigned long s_mount_opt2;
Theodore Ts'o673c6102010-12-15 20:28:48 -05004182 uid_t s_resuid;
4183 gid_t s_resgid;
4184 unsigned long s_commit_interval;
4185 u32 s_min_batch_time, s_max_batch_time;
4186#ifdef CONFIG_QUOTA
4187 int s_jquota_fmt;
4188 char *s_qf_names[MAXQUOTAS];
4189#endif
4190};
4191
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004192static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004193{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004194 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004195 struct ext4_sb_info *sbi = EXT4_SB(sb);
4196 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004197 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004198 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004199 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004200 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004201 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004202 int err;
4203#ifdef CONFIG_QUOTA
4204 int i;
4205#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004206 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004207
4208 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04004209 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004210 old_sb_flags = sb->s_flags;
4211 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004212 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004213 old_opts.s_resuid = sbi->s_resuid;
4214 old_opts.s_resgid = sbi->s_resgid;
4215 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004216 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4217 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004218#ifdef CONFIG_QUOTA
4219 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4220 for (i = 0; i < MAXQUOTAS; i++)
4221 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4222#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004223 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4224 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004225
4226 /*
4227 * Allow the "check" option to be passed as a remount option.
4228 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004229 if (!parse_options(data, sb, NULL, &journal_ioprio,
4230 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004231 err = -EINVAL;
4232 goto restore_opts;
4233 }
4234
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004235 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004236 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004237
4238 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004239 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004240
4241 es = sbi->s_es;
4242
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004243 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004244 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004245 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4246 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004247
4248 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004249 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004250 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004251 err = -EROFS;
4252 goto restore_opts;
4253 }
4254
4255 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004256 err = dquot_suspend(sb, -1);
4257 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004258 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004259
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004260 /*
4261 * First of all, the unconditional stuff we have to do
4262 * to disable replay of the journal when we next remount
4263 */
4264 sb->s_flags |= MS_RDONLY;
4265
4266 /*
4267 * OK, test if we are remounting a valid rw partition
4268 * readonly, and if so set the rdonly flag and then
4269 * mark the partition as valid again.
4270 */
Mingming Cao617ba132006-10-11 01:20:53 -07004271 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4272 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004273 es->s_state = cpu_to_le16(sbi->s_mount_state);
4274
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004275 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004276 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004277 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004278 /* Make sure we can mount this feature set readwrite */
4279 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004280 err = -EROFS;
4281 goto restore_opts;
4282 }
Eric Sandeenead65962007-02-10 01:46:08 -08004283 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004284 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004285 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004286 */
4287 for (g = 0; g < sbi->s_groups_count; g++) {
4288 struct ext4_group_desc *gdp =
4289 ext4_get_group_desc(sb, g, NULL);
4290
4291 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004292 ext4_msg(sb, KERN_ERR,
4293 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04004294 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4295 le16_to_cpu(gdp->bg_checksum));
4296 err = -EINVAL;
4297 goto restore_opts;
4298 }
4299 }
4300
4301 /*
Eric Sandeenead65962007-02-10 01:46:08 -08004302 * If we have an unprocessed orphan list hanging
4303 * around from a previously readonly bdev mount,
4304 * require a full umount/remount for now.
4305 */
4306 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004307 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08004308 "remount RDWR because of unprocessed "
4309 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004310 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08004311 err = -EINVAL;
4312 goto restore_opts;
4313 }
4314
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004315 /*
4316 * Mounting a RDONLY partition read-write, so reread
4317 * and store the current valid flag. (It may have
4318 * been changed by e2fsck since we originally mounted
4319 * the partition.)
4320 */
Frank Mayhar03901312009-01-07 00:06:22 -05004321 if (sbi->s_journal)
4322 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004323 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07004324 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004325 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004326 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004327 sb->s_flags &= ~MS_RDONLY;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004328 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004329 }
4330 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04004331
4332 /*
4333 * Reinitialize lazy itable initialization thread based on
4334 * current settings
4335 */
4336 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4337 ext4_unregister_li_request(sb);
4338 else {
4339 ext4_group_t first_not_zeroed;
4340 first_not_zeroed = ext4_has_uninit_itable(sb);
4341 ext4_register_li_request(sb, first_not_zeroed);
4342 }
4343
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004344 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05004345 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004346 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05004347
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004348#ifdef CONFIG_QUOTA
4349 /* Release old quota file names */
4350 for (i = 0; i < MAXQUOTAS; i++)
4351 if (old_opts.s_qf_names[i] &&
4352 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4353 kfree(old_opts.s_qf_names[i]);
4354#endif
Al Virobbd68512009-05-06 10:43:07 -04004355 unlock_super(sb);
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004356 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004357 dquot_resume(sb, -1);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004358
4359 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4360 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004361 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004362
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004363restore_opts:
4364 sb->s_flags = old_sb_flags;
4365 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004366 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004367 sbi->s_resuid = old_opts.s_resuid;
4368 sbi->s_resgid = old_opts.s_resgid;
4369 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004370 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4371 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004372#ifdef CONFIG_QUOTA
4373 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4374 for (i = 0; i < MAXQUOTAS; i++) {
4375 if (sbi->s_qf_names[i] &&
4376 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4377 kfree(sbi->s_qf_names[i]);
4378 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4379 }
4380#endif
Al Virobbd68512009-05-06 10:43:07 -04004381 unlock_super(sb);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004382 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004383 return err;
4384}
4385
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004386static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004387{
4388 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004389 struct ext4_sb_info *sbi = EXT4_SB(sb);
4390 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08004391 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004392
Badari Pulavarty5e700302007-07-15 23:42:00 -07004393 if (test_opt(sb, MINIX_DF)) {
4394 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004395 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04004396 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07004397 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004398
4399 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07004400 * Compute the overhead (FS structures). This is constant
4401 * for a given filesystem unless the number of block groups
4402 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004403 */
4404
4405 /*
4406 * All of the blocks before first_data_block are
4407 * overhead
4408 */
4409 overhead = le32_to_cpu(es->s_first_data_block);
4410
4411 /*
4412 * Add the overhead attributed to the superblock and
4413 * block group descriptors. If the sparse superblocks
4414 * feature is turned on, then not all groups have this.
4415 */
4416 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07004417 overhead += ext4_bg_has_super(sb, i) +
4418 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004419 cond_resched();
4420 }
4421
4422 /*
4423 * Every block group has an inode bitmap, a block
4424 * bitmap, and an inode table.
4425 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07004426 overhead += ngroups * (2 + sbi->s_itb_per_group);
4427 sbi->s_overhead_last = overhead;
4428 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004429 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004430 }
4431
Mingming Cao617ba132006-10-11 01:20:53 -07004432 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004433 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07004434 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04004435 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4436 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004437 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4438 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004439 buf->f_bavail = 0;
4440 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07004441 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07004442 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08004443 fsid = le64_to_cpup((void *)es->s_uuid) ^
4444 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4445 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4446 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004447
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004448 return 0;
4449}
4450
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004451/* Helper function for writing quotas on sync - we need to start transaction
4452 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004453 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07004454 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01004455 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05004456 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07004457 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004458 *
4459 */
4460
4461#ifdef CONFIG_QUOTA
4462
4463static inline struct inode *dquot_to_inode(struct dquot *dquot)
4464{
4465 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4466}
4467
Mingming Cao617ba132006-10-11 01:20:53 -07004468static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004469{
4470 int ret, err;
4471 handle_t *handle;
4472 struct inode *inode;
4473
4474 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004475 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004476 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004477 if (IS_ERR(handle))
4478 return PTR_ERR(handle);
4479 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004480 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004481 if (!ret)
4482 ret = err;
4483 return ret;
4484}
4485
Mingming Cao617ba132006-10-11 01:20:53 -07004486static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004487{
4488 int ret, err;
4489 handle_t *handle;
4490
Mingming Cao617ba132006-10-11 01:20:53 -07004491 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004492 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004493 if (IS_ERR(handle))
4494 return PTR_ERR(handle);
4495 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004496 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004497 if (!ret)
4498 ret = err;
4499 return ret;
4500}
4501
Mingming Cao617ba132006-10-11 01:20:53 -07004502static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004503{
4504 int ret, err;
4505 handle_t *handle;
4506
Mingming Cao617ba132006-10-11 01:20:53 -07004507 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004508 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07004509 if (IS_ERR(handle)) {
4510 /* Release dquot anyway to avoid endless cycle in dqput() */
4511 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004512 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07004513 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004514 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004515 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004516 if (!ret)
4517 ret = err;
4518 return ret;
4519}
4520
Mingming Cao617ba132006-10-11 01:20:53 -07004521static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004522{
Jan Kara2c8be6b2008-05-13 21:27:55 -04004523 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07004524 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4525 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004526 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004527 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004528 } else {
4529 return dquot_mark_dquot_dirty(dquot);
4530 }
4531}
4532
Mingming Cao617ba132006-10-11 01:20:53 -07004533static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004534{
4535 int ret, err;
4536 handle_t *handle;
4537
4538 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07004539 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004540 if (IS_ERR(handle))
4541 return PTR_ERR(handle);
4542 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07004543 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004544 if (!ret)
4545 ret = err;
4546 return ret;
4547}
4548
4549/*
4550 * Turn on quotas during mount time - we need to find
4551 * the quota file and such...
4552 */
Mingming Cao617ba132006-10-11 01:20:53 -07004553static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004554{
Christoph Hellwig287a8092010-05-19 07:16:45 -04004555 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4556 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004557}
4558
4559/*
4560 * Standard function to be called on quota_on
4561 */
Mingming Cao617ba132006-10-11 01:20:53 -07004562static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02004563 struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004564{
4565 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004566
4567 if (!test_opt(sb, QUOTA))
4568 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04004569
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004570 /* Quotafile not on the same filesystem? */
Jan Karaf00c9e42010-09-15 17:38:58 +02004571 if (path->mnt->mnt_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004572 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04004573 /* Journaling quota? */
4574 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004575 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02004576 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004577 ext4_msg(sb, KERN_WARNING,
4578 "Quota file not on filesystem root. "
4579 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004580 }
Jan Kara06235432008-05-13 19:11:51 -04004581
4582 /*
4583 * When we journal data on quota file, we have to flush journal to see
4584 * all updates to the file when we bypass pagecache...
4585 */
Frank Mayhar03901312009-01-07 00:06:22 -05004586 if (EXT4_SB(sb)->s_journal &&
Jan Karaf00c9e42010-09-15 17:38:58 +02004587 ext4_should_journal_data(path->dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04004588 /*
4589 * We don't need to lock updates but journal_flush() could
4590 * otherwise be livelocked...
4591 */
4592 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004593 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04004594 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02004595 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004596 return err;
Jan Kara06235432008-05-13 19:11:51 -04004597 }
4598
Jan Karaf00c9e42010-09-15 17:38:58 +02004599 return dquot_quota_on(sb, type, format_id, path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004600}
4601
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004602static int ext4_quota_off(struct super_block *sb, int type)
4603{
Dmitry Monakhov87009d82010-11-08 13:47:33 -05004604 /* Force all delayed allocation blocks to be allocated.
4605 * Caller already holds s_umount sem */
4606 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004607 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004608
4609 return dquot_quota_off(sb, type);
4610}
4611
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004612/* Read data from quotafile - avoid pagecache and such because we cannot afford
4613 * acquiring the locks... As quota files are never truncated and quota code
4614 * itself serializes the operations (and noone else should touch the files)
4615 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07004616static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004617 size_t len, loff_t off)
4618{
4619 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004620 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004621 int err = 0;
4622 int offset = off & (sb->s_blocksize - 1);
4623 int tocopy;
4624 size_t toread;
4625 struct buffer_head *bh;
4626 loff_t i_size = i_size_read(inode);
4627
4628 if (off > i_size)
4629 return 0;
4630 if (off+len > i_size)
4631 len = i_size-off;
4632 toread = len;
4633 while (toread > 0) {
4634 tocopy = sb->s_blocksize - offset < toread ?
4635 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07004636 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004637 if (err)
4638 return err;
4639 if (!bh) /* A hole? */
4640 memset(data, 0, tocopy);
4641 else
4642 memcpy(data, bh->b_data+offset, tocopy);
4643 brelse(bh);
4644 offset = 0;
4645 toread -= tocopy;
4646 data += tocopy;
4647 blk++;
4648 }
4649 return len;
4650}
4651
4652/* Write to quotafile (we know the transaction is already started and has
4653 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07004654static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004655 const char *data, size_t len, loff_t off)
4656{
4657 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004658 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004659 int err = 0;
4660 int offset = off & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004661 struct buffer_head *bh;
4662 handle_t *handle = journal_current_handle();
4663
Frank Mayhar03901312009-01-07 00:06:22 -05004664 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004665 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4666 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07004667 (unsigned long long)off, (unsigned long long)len);
4668 return -EIO;
4669 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004670 /*
4671 * Since we account only one data block in transaction credits,
4672 * then it is impossible to cross a block boundary.
4673 */
4674 if (sb->s_blocksize - offset < len) {
4675 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4676 " cancelled because not block aligned",
4677 (unsigned long long)off, (unsigned long long)len);
4678 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004679 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004680
4681 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4682 bh = ext4_bread(handle, inode, blk, 1, &err);
4683 if (!bh)
4684 goto out;
Jan Kara62d2b5f2010-07-27 11:56:07 -04004685 err = ext4_journal_get_write_access(handle, bh);
4686 if (err) {
4687 brelse(bh);
4688 goto out;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004689 }
4690 lock_buffer(bh);
4691 memcpy(bh->b_data+offset, data, len);
4692 flush_dcache_page(bh->b_page);
4693 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04004694 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004695 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004696out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004697 if (err) {
Jan Kara4d04e4f2008-07-04 09:59:34 -07004698 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004699 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07004700 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004701 if (inode->i_size < off + len) {
4702 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07004703 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004704 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004705 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07004706 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004707 mutex_unlock(&inode->i_mutex);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004708 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004709}
4710
4711#endif
4712
Al Viro152a0832010-07-25 00:46:55 +04004713static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4714 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004715{
Al Viro152a0832010-07-25 00:46:55 +04004716 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004717}
4718
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004719#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004720static struct file_system_type ext2_fs_type = {
4721 .owner = THIS_MODULE,
4722 .name = "ext2",
Al Viro152a0832010-07-25 00:46:55 +04004723 .mount = ext4_mount,
Theodore Ts'o24b58422009-12-07 14:08:51 -05004724 .kill_sb = kill_block_super,
4725 .fs_flags = FS_REQUIRES_DEV,
4726};
4727
4728static inline void register_as_ext2(void)
4729{
4730 int err = register_filesystem(&ext2_fs_type);
4731 if (err)
4732 printk(KERN_WARNING
4733 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4734}
4735
4736static inline void unregister_as_ext2(void)
4737{
4738 unregister_filesystem(&ext2_fs_type);
4739}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004740MODULE_ALIAS("ext2");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004741#else
4742static inline void register_as_ext2(void) { }
4743static inline void unregister_as_ext2(void) { }
4744#endif
4745
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004746#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004747static inline void register_as_ext3(void)
4748{
4749 int err = register_filesystem(&ext3_fs_type);
4750 if (err)
4751 printk(KERN_WARNING
4752 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4753}
4754
4755static inline void unregister_as_ext3(void)
4756{
4757 unregister_filesystem(&ext3_fs_type);
4758}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004759MODULE_ALIAS("ext3");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004760#else
4761static inline void register_as_ext3(void) { }
4762static inline void unregister_as_ext3(void) { }
4763#endif
4764
Theodore Ts'o03010a32008-10-10 20:02:48 -04004765static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004766 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04004767 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04004768 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004769 .kill_sb = kill_block_super,
4770 .fs_flags = FS_REQUIRES_DEV,
4771};
4772
Lukas Czerner8f021222011-02-03 14:33:33 -05004773static int __init ext4_init_feat_adverts(void)
Lukas Czerner857ac882010-10-27 21:30:05 -04004774{
4775 struct ext4_features *ef;
4776 int ret = -ENOMEM;
4777
4778 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4779 if (!ef)
4780 goto out;
4781
4782 ef->f_kobj.kset = ext4_kset;
4783 init_completion(&ef->f_kobj_unregister);
4784 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4785 "features");
4786 if (ret) {
4787 kfree(ef);
4788 goto out;
4789 }
4790
4791 ext4_feat = ef;
4792 ret = 0;
4793out:
4794 return ret;
4795}
4796
Lukas Czerner8f021222011-02-03 14:33:33 -05004797static void ext4_exit_feat_adverts(void)
4798{
4799 kobject_put(&ext4_feat->f_kobj);
4800 wait_for_completion(&ext4_feat->f_kobj_unregister);
4801 kfree(ext4_feat);
4802}
4803
Eric Sandeene9e3bce2011-02-12 08:17:34 -05004804/* Shared across all ext4 file systems */
4805wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4806struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4807
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004808static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004809{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05004810 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05004811
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04004812 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05004813
4814 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4815 mutex_init(&ext4__aio_mutex[i]);
4816 init_waitqueue_head(&ext4__ioend_wq[i]);
4817 }
4818
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004819 err = ext4_init_pageio();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004820 if (err)
4821 return err;
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004822 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04004823 if (err)
Theodore Ts'odd683142011-02-03 14:33:49 -05004824 goto out7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004825 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4826 if (!ext4_kset)
Theodore Ts'odd683142011-02-03 14:33:49 -05004827 goto out6;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004828 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Theodore Ts'odd683142011-02-03 14:33:49 -05004829 if (!ext4_proc_root)
4830 goto out5;
Lukas Czerner857ac882010-10-27 21:30:05 -04004831
4832 err = ext4_init_feat_adverts();
Theodore Ts'odd683142011-02-03 14:33:49 -05004833 if (err)
4834 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04004835
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004836 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004837 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004838 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05004839
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004840 err = ext4_init_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004841 if (err)
4842 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004843 err = init_inodecache();
4844 if (err)
4845 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05004846 register_as_ext2();
4847 register_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004848 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004849 if (err)
4850 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004851
4852 ext4_li_info = NULL;
4853 mutex_init(&ext4_li_mtx);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004854 return 0;
4855out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05004856 unregister_as_ext2();
4857 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004858 destroy_inodecache();
4859out1:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004860 ext4_exit_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004861out2:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004862 ext4_exit_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004863out3:
Lukas Czerner8f021222011-02-03 14:33:33 -05004864 ext4_exit_feat_adverts();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004865out4:
Theodore Ts'odd683142011-02-03 14:33:49 -05004866 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'obd2d0212010-10-27 21:30:10 -04004867out5:
Theodore Ts'odd683142011-02-03 14:33:49 -05004868 kset_unregister(ext4_kset);
4869out6:
4870 ext4_exit_system_zone();
4871out7:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004872 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004873 return err;
4874}
4875
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004876static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004877{
Lukas Czernerbfff6872010-10-27 21:30:05 -04004878 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05004879 unregister_as_ext2();
4880 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004881 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004882 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004883 ext4_exit_xattr();
4884 ext4_exit_mballoc();
Lukas Czerner8f021222011-02-03 14:33:33 -05004885 ext4_exit_feat_adverts();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004886 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004887 kset_unregister(ext4_kset);
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004888 ext4_exit_system_zone();
4889 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004890}
4891
4892MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05004893MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004894MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004895module_init(ext4_init_fs)
4896module_exit(ext4_exit_fs)