blob: 933900909ed0facb352e8d2c86c19a456a4fbb49 [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>
Dan Magenheimer7abc52c2011-05-26 10:02:03 -060041#include <linux/cleancache.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070042#include <asm/uaccess.h>
43
Lukas Czernerbfff6872010-10-27 21:30:05 -040044#include <linux/kthread.h>
45#include <linux/freezer.h>
46
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040047#include "ext4.h"
Eric Gouriou6f91bc52011-10-27 11:43:23 -040048#include "ext4_extents.h"
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040049#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070050#include "xattr.h"
51#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040052#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070053
Theodore Ts'o9bffad12009-06-17 11:48:11 -040054#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
Theodore Ts'o1f109d52010-10-27 21:30:14 -040057static struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040058static struct kset *ext4_kset;
Lukas Czerner0b75a842011-02-23 12:22:49 -050059static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040062
Mingming Cao617ba132006-10-11 01:20:53 -070063static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070064 unsigned long journal_devnum);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040065static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040066static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070070static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040071static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070072 char nbuf[16]);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040073static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080075static int ext4_unfreeze(struct super_block *sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040076static void ext4_write_super(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080077static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040078static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
Theodore Ts'o2035e772011-04-18 17:29:14 -040080static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
Amir Goldsteind39195c2011-02-28 00:53:45 -050082static int ext4_feature_set_ok(struct super_block *sb, int readonly);
Lukas Czernerbfff6872010-10-27 21:30:05 -040083static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050085static void ext4_clear_request_list(void);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070086
Theodore Ts'o2035e772011-04-18 17:29:14 -040087#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
Jan Karaba69f9a2010-03-24 20:18:37 -0400101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +0400105 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -0400106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700113
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400114void *ext4_kvmalloc(size_t size, gfp_t flags)
115{
116 void *ret;
117
118 ret = kmalloc(size, flags);
119 if (!ret)
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
121 return ret;
122}
123
124void *ext4_kvzalloc(size_t size, gfp_t flags)
125{
126 void *ret;
127
Mathias Krausedb9481c2011-08-03 14:57:11 -0400128 ret = kzalloc(size, flags);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400129 if (!ret)
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
131 return ret;
132}
133
134void ext4_kvfree(void *ptr)
135{
136 if (is_vmalloc_addr(ptr))
137 vfree(ptr);
138 else
139 kfree(ptr);
140
141}
142
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700143ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700145{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700149}
150
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700151ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700153{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700157}
158
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700159ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700161{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400162 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700165}
166
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400167__u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500169{
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500173}
174
175__u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
177{
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500181}
182
183__u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
185{
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500189}
190
191__u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
193{
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500197}
198
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700199void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700201{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700205}
206
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700207void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700209{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700213}
214
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700215void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700217{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700221}
222
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400223void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500225{
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
229}
230
231void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
233{
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
237}
238
239void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
241{
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
245}
246
247void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
249{
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
253}
254
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400255
256/* Just increment the non-pointer handle value */
257static handle_t *ext4_get_nojournal(void)
258{
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
261
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
263
264 ref_cnt++;
265 handle = (handle_t *)ref_cnt;
266
267 current->journal_info = handle;
268 return handle;
269}
270
271
272/* Decrement the non-pointer handle value */
273static void ext4_put_nojournal(handle_t *handle)
274{
275 unsigned long ref_cnt = (unsigned long)handle;
276
277 BUG_ON(ref_cnt == 0);
278
279 ref_cnt--;
280 handle = (handle_t *)ref_cnt;
281
282 current->journal_info = handle;
283}
284
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700285/*
Mingming Caodab291a2006-10-11 01:21:01 -0700286 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700287 *
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
291 * appropriate.
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -0400292 *
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700296 */
Mingming Cao617ba132006-10-11 01:20:53 -0700297handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700298{
299 journal_t *journal;
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -0400300 handle_t *handle;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700301
Theodore Ts'o12706392011-07-10 22:37:50 -0400302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
305
Mingming Cao617ba132006-10-11 01:20:53 -0700306 journal = EXT4_SB(sb)->s_journal;
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -0400307 handle = ext4_journal_current_handle();
308
309 /*
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
315 */
316 if (!handle)
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
318
319 if (!journal)
320 return ext4_get_nojournal();
321 /*
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
325 */
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700329 }
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -0400330 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700331}
332
333/*
334 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700336 * that sync() will call the filesystem's write_super callback if
337 * appropriate.
338 */
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400339int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700340{
341 struct super_block *sb;
342 int err;
343 int rc;
344
Frank Mayhar03901312009-01-07 00:06:22 -0500345 if (!ext4_handle_valid(handle)) {
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400346 ext4_put_nojournal(handle);
Frank Mayhar03901312009-01-07 00:06:22 -0500347 return 0;
348 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700349 sb = handle->h_transaction->t_journal->j_private;
350 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700351 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700352
353 if (!err)
354 err = rc;
355 if (err)
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400356 __ext4_std_error(sb, where, line, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700357 return err;
358}
359
Theodore Ts'o90c72012010-06-29 14:53:24 -0400360void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700363{
364 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700366
Frank Mayhar03901312009-01-07 00:06:22 -0500367 BUG_ON(!ext4_handle_valid(handle));
368
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700369 if (bh)
370 BUFFER_TRACE(bh, "abort");
371
372 if (!handle->h_err)
373 handle->h_err = err;
374
375 if (is_handle_aborted(handle))
376 return;
377
Theodore Ts'o90c72012010-06-29 14:53:24 -0400378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700380
Mingming Caodab291a2006-10-11 01:21:01 -0700381 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700382}
383
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400384static void __save_error_info(struct super_block *sb, const char *func,
385 unsigned int line)
386{
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
388
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
401 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400402 /*
403 * Start the daily error reporting function if it hasn't been
404 * started already
405 */
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
409}
410
411static void save_error_info(struct super_block *sb, const char *func,
412 unsigned int line)
413{
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
416}
417
Theodore Ts'o7c2e7082011-09-09 18:28:51 -0400418/*
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
425 */
426static int block_device_ejected(struct super_block *sb)
427{
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
430
431 return bdi->dev == NULL;
432}
433
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400434
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700435/* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
437 *
438 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700439 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
444 *
Mingming Caodab291a2006-10-11 01:21:01 -0700445 * 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 -0200446 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700447 * that error until we've noted it down and cleared it.
448 */
449
Mingming Cao617ba132006-10-11 01:20:53 -0700450static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700451{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700452 if (sb->s_flags & MS_RDONLY)
453 return;
454
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400455 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700456 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700457
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700459 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700460 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700461 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400462 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700464 sb->s_flags |= MS_RDONLY;
465 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700466 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700467 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700468 sb->s_id);
469}
470
Eric Sandeen12062dd2010-02-15 14:19:27 -0500471void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400472 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700473{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500474 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700475 va_list args;
476
477 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500478 vaf.fmt = fmt;
479 vaf.va = &args;
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
481 sb->s_id, function, line, current->comm, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700482 va_end(args);
483
Mingming Cao617ba132006-10-11 01:20:53 -0700484 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700485}
486
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400487void ext4_error_inode(struct inode *inode, const char *function,
488 unsigned int line, ext4_fsblk_t block,
Frank Mayhar273df552010-03-02 11:46:09 -0500489 const char *fmt, ...)
490{
491 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500492 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400493 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500494
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 es->s_last_error_block = cpu_to_le64(block);
497 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500498 va_start(args, fmt);
Theodore Ts'of7c21172011-01-10 12:10:55 -0500499 vaf.fmt = fmt;
500 vaf.va = &args;
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400501 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
502 inode->i_sb->s_id, function, line, inode->i_ino);
503 if (block)
Theodore Ts'of7c21172011-01-10 12:10:55 -0500504 printk(KERN_CONT "block %llu: ", block);
505 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
Frank Mayhar273df552010-03-02 11:46:09 -0500506 va_end(args);
507
508 ext4_handle_error(inode->i_sb);
509}
510
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400511void ext4_error_file(struct file *file, const char *function,
Theodore Ts'of7c21172011-01-10 12:10:55 -0500512 unsigned int line, ext4_fsblk_t block,
513 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500514{
515 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500516 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400517 struct ext4_super_block *es;
Frank Mayhar273df552010-03-02 11:46:09 -0500518 struct inode *inode = file->f_dentry->d_inode;
519 char pathname[80], *path;
520
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400521 es = EXT4_SB(inode->i_sb)->s_es;
522 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
523 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500524 path = d_path(&(file->f_path), pathname, sizeof(pathname));
Dan Carpenterf9a62d02011-01-10 12:10:50 -0500525 if (IS_ERR(path))
Frank Mayhar273df552010-03-02 11:46:09 -0500526 path = "(unknown)";
527 printk(KERN_CRIT
Theodore Ts'of7c21172011-01-10 12:10:55 -0500528 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
529 inode->i_sb->s_id, function, line, inode->i_ino);
530 if (block)
531 printk(KERN_CONT "block %llu: ", block);
532 va_start(args, fmt);
533 vaf.fmt = fmt;
534 vaf.va = &args;
535 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
Frank Mayhar273df552010-03-02 11:46:09 -0500536 va_end(args);
537
538 ext4_handle_error(inode->i_sb);
539}
540
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400541static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700542 char nbuf[16])
543{
544 char *errstr = NULL;
545
546 switch (errno) {
547 case -EIO:
548 errstr = "IO failure";
549 break;
550 case -ENOMEM:
551 errstr = "Out of memory";
552 break;
553 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700556 errstr = "Journal has aborted";
557 else
558 errstr = "Readonly filesystem";
559 break;
560 default:
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
563 * NULL. */
564 if (nbuf) {
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
567 errstr = nbuf;
568 }
569 break;
570 }
571
572 return errstr;
573}
574
Mingming Cao617ba132006-10-11 01:20:53 -0700575/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700576 * automatically and invokes the appropriate error response. */
577
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400578void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700580{
581 char nbuf[16];
582 const char *errstr;
583
584 /* Special case: if the error is EROFS, and we're not already
585 * inside a transaction, then there's really no point in logging
586 * an error. */
587 if (errno == -EROFS && journal_current_handle() == NULL &&
588 (sb->s_flags & MS_RDONLY))
589 return;
590
Mingming Cao617ba132006-10-11 01:20:53 -0700591 errstr = ext4_decode_error(sb, errno, nbuf);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400592 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
593 sb->s_id, function, line, errstr);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400594 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700595
Mingming Cao617ba132006-10-11 01:20:53 -0700596 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700597}
598
599/*
Mingming Cao617ba132006-10-11 01:20:53 -0700600 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
603 *
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
607 */
608
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400609void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400610 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700611{
612 va_list args;
613
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400614 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700615 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400616 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
617 function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700618 vprintk(fmt, args);
619 printk("\n");
620 va_end(args);
621
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400622 if ((sb->s_flags & MS_RDONLY) == 0) {
623 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
624 sb->s_flags |= MS_RDONLY;
625 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
626 if (EXT4_SB(sb)->s_journal)
627 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
628 save_error_info(sb, function, line);
629 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700630 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700631 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700632}
633
Joe Perches0ff2ea72010-12-19 22:43:19 -0500634void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400635{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500636 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400637 va_list args;
638
639 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500640 vaf.fmt = fmt;
641 vaf.va = &args;
642 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400643 va_end(args);
644}
645
Eric Sandeen12062dd2010-02-15 14:19:27 -0500646void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400647 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700648{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500649 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700650 va_list args;
651
652 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500653 vaf.fmt = fmt;
654 vaf.va = &args;
655 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
656 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700657 va_end(args);
658}
659
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400660void __ext4_grp_locked_error(const char *function, unsigned int line,
661 struct super_block *sb, ext4_group_t grp,
662 unsigned long ino, ext4_fsblk_t block,
663 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500664__releases(bitlock)
665__acquires(bitlock)
666{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500667 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500668 va_list args;
669 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
670
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400671 es->s_last_error_ino = cpu_to_le32(ino);
672 es->s_last_error_block = cpu_to_le64(block);
673 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500674
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500675 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500676
677 vaf.fmt = fmt;
678 vaf.va = &args;
Robin Dong21149d62011-03-21 20:39:22 -0400679 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400680 sb->s_id, function, line, grp);
681 if (ino)
Joe Perches0ff2ea72010-12-19 22:43:19 -0500682 printk(KERN_CONT "inode %lu: ", ino);
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400683 if (block)
Joe Perches0ff2ea72010-12-19 22:43:19 -0500684 printk(KERN_CONT "block %llu:", (unsigned long long) block);
685 printk(KERN_CONT "%pV\n", &vaf);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500686 va_end(args);
687
688 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400689 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500690 return;
691 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400692
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500693 ext4_unlock_group(sb, grp);
694 ext4_handle_error(sb);
695 /*
696 * We only get here in the ERRORS_RO case; relocking the group
697 * may be dangerous, but nothing bad will happen since the
698 * filesystem will have already been marked read/only and the
699 * journal has been aborted. We return 1 as a hint to callers
700 * who might what to use the return value from
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300701 * ext4_grp_locked_error() to distinguish between the
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500702 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
703 * aggressively from the ext4 function in question, with a
704 * more appropriate error code.
705 */
706 ext4_lock_group(sb, grp);
707 return;
708}
709
Mingming Cao617ba132006-10-11 01:20:53 -0700710void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700711{
Mingming Cao617ba132006-10-11 01:20:53 -0700712 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700713
Mingming Cao617ba132006-10-11 01:20:53 -0700714 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700715 return;
716
Eric Sandeen12062dd2010-02-15 14:19:27 -0500717 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700718 "updating to rev %d because of new feature flag, "
719 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700720 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700721
Mingming Cao617ba132006-10-11 01:20:53 -0700722 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
723 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
724 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700725 /* leave es->s_feature_*compat flags alone */
726 /* es->s_uuid will be set by e2fsck if empty */
727
728 /*
729 * The rest of the superblock fields should be zero, and if not it
730 * means they are likely already in use, so leave them alone. We
731 * can leave it up to e2fsck to clean up any inconsistencies there.
732 */
733}
734
735/*
736 * Open the external journal device
737 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400738static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700739{
740 struct block_device *bdev;
741 char b[BDEVNAME_SIZE];
742
Tejun Heod4d77622010-11-13 11:55:18 +0100743 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700744 if (IS_ERR(bdev))
745 goto fail;
746 return bdev;
747
748fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400749 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700750 __bdevname(dev, b), PTR_ERR(bdev));
751 return NULL;
752}
753
754/*
755 * Release the journal device
756 */
Mingming Cao617ba132006-10-11 01:20:53 -0700757static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700758{
Tejun Heoe525fd82010-11-13 11:55:17 +0100759 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700760}
761
Mingming Cao617ba132006-10-11 01:20:53 -0700762static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700763{
764 struct block_device *bdev;
765 int ret = -ENODEV;
766
767 bdev = sbi->journal_bdev;
768 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700769 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700770 sbi->journal_bdev = NULL;
771 }
772 return ret;
773}
774
775static inline struct inode *orphan_list_entry(struct list_head *l)
776{
Mingming Cao617ba132006-10-11 01:20:53 -0700777 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700778}
779
Mingming Cao617ba132006-10-11 01:20:53 -0700780static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700781{
782 struct list_head *l;
783
Eric Sandeenb31e1552009-06-04 17:36:36 -0400784 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
785 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700786
787 printk(KERN_ERR "sb_info orphan list:\n");
788 list_for_each(l, &sbi->s_orphan) {
789 struct inode *inode = orphan_list_entry(l);
790 printk(KERN_ERR " "
791 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
792 inode->i_sb->s_id, inode->i_ino, inode,
793 inode->i_mode, inode->i_nlink,
794 NEXT_ORPHAN(inode));
795 }
796}
797
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400798static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700799{
Mingming Cao617ba132006-10-11 01:20:53 -0700800 struct ext4_sb_info *sbi = EXT4_SB(sb);
801 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400802 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700803
Lukas Czerner857ac882010-10-27 21:30:05 -0400804 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400805 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
806
Mingming Cao4c0425f2009-09-28 15:48:41 -0400807 flush_workqueue(sbi->dio_unwritten_wq);
808 destroy_workqueue(sbi->dio_unwritten_wq);
809
Al Viroa9e220f2009-05-05 22:10:44 -0400810 lock_super(sb);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200811 if (sb->s_dirt)
Christoph Hellwigebc1ac12009-05-11 23:35:03 +0200812 ext4_commit_super(sb, 1);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200813
Frank Mayhar03901312009-01-07 00:06:22 -0500814 if (sbi->s_journal) {
815 err = jbd2_journal_destroy(sbi->s_journal);
816 sbi->s_journal = NULL;
817 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400818 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500819 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500820
Sergey Senozhatskya1c6c562010-10-27 21:30:04 -0400821 del_timer(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500822 ext4_release_system_zone(sb);
823 ext4_mb_release(sb);
824 ext4_ext_release(sb);
825 ext4_xattr_put_super(sb);
826
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700827 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700828 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700829 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400830 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700831 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400832 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400833 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400834 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400835 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700836
837 for (i = 0; i < sbi->s_gdb_count; i++)
838 brelse(sbi->s_group_desc[i]);
Theodore Ts'of18a5f22011-08-01 08:45:38 -0400839 ext4_kvfree(sbi->s_group_desc);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400840 ext4_kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -0400841 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700842 percpu_counter_destroy(&sbi->s_freeinodes_counter);
843 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -0400844 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700845 brelse(sbi->s_sbh);
846#ifdef CONFIG_QUOTA
847 for (i = 0; i < MAXQUOTAS; i++)
848 kfree(sbi->s_qf_names[i]);
849#endif
850
851 /* Debugging code just in case the in-memory inode orphan list
852 * isn't empty. The on-disk one can be non-empty if we've
853 * detected an error and taken the fs readonly, but the
854 * in-memory list had better be clean by this point. */
855 if (!list_empty(&sbi->s_orphan))
856 dump_orphan_list(sb, sbi);
857 J_ASSERT(list_empty(&sbi->s_orphan));
858
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700859 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700860 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
861 /*
862 * Invalidate the journal device's buffers. We don't want them
863 * floating about in memory - the physical journal device may
864 * hotswapped, and it breaks the `ro-after' testing code.
865 */
866 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700867 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700868 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700869 }
Johann Lombardic5e06d12011-05-24 18:31:25 -0400870 if (sbi->s_mmp_tsk)
871 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700872 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400873 /*
874 * Now that we are completely done shutting down the
875 * superblock, we need to actually destroy the kobject.
876 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400877 unlock_super(sb);
878 kobject_put(&sbi->s_kobj);
879 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500880 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700881 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700882}
883
Christoph Lametere18b8902006-12-06 20:33:20 -0800884static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700885
886/*
887 * Called inside transaction, so use GFP_NOFS
888 */
Mingming Cao617ba132006-10-11 01:20:53 -0700889static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700890{
Mingming Cao617ba132006-10-11 01:20:53 -0700891 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700892
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800893 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700894 if (!ei)
895 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400896
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700897 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400898 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700899 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500900 INIT_LIST_HEAD(&ei->i_prealloc_list);
901 spin_lock_init(&ei->i_prealloc_lock);
Mingming Caod2a17632008-07-14 17:52:37 -0400902 ei->i_reserved_data_blocks = 0;
903 ei->i_reserved_meta_blocks = 0;
904 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500905 ei->i_da_metadata_calc_len = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400906 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300907#ifdef CONFIG_QUOTA
908 ei->i_reserved_quota = 0;
909#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500910 ei->jinode = NULL;
Jiaying Zhangc7064ef2010-03-02 13:28:44 -0500911 INIT_LIST_HEAD(&ei->i_completed_io_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500912 spin_lock_init(&ei->i_completed_io_lock);
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400913 ei->cur_aio_dio = NULL;
Jan Karab436b9b2009-12-08 23:51:10 -0500914 ei->i_sync_tid = 0;
915 ei->i_datasync_tid = 0;
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500916 atomic_set(&ei->i_ioend_count, 0);
Eric Sandeene9e3bce2011-02-12 08:17:34 -0500917 atomic_set(&ei->i_aiodio_unwritten, 0);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400918
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700919 return &ei->vfs_inode;
920}
921
Theodore Ts'o7ff9c072010-11-08 13:51:33 -0500922static int ext4_drop_inode(struct inode *inode)
923{
924 int drop = generic_drop_inode(inode);
925
926 trace_ext4_drop_inode(inode, drop);
927 return drop;
928}
929
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100930static void ext4_i_callback(struct rcu_head *head)
931{
932 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100933 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
934}
935
Mingming Cao617ba132006-10-11 01:20:53 -0700936static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700937{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700938 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400939 ext4_msg(inode->i_sb, KERN_ERR,
940 "Inode %lu (%p): orphan list check failed!",
941 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700942 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
943 EXT4_I(inode), sizeof(struct ext4_inode_info),
944 true);
945 dump_stack();
946 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100947 call_rcu(&inode->i_rcu, ext4_i_callback);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700948}
949
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700950static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700951{
Mingming Cao617ba132006-10-11 01:20:53 -0700952 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700953
Christoph Lametera35afb82007-05-16 22:10:57 -0700954 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400955#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700956 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700957#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500958 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700959 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700960}
961
962static int init_inodecache(void)
963{
Mingming Cao617ba132006-10-11 01:20:53 -0700964 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
965 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700966 0, (SLAB_RECLAIM_ACCOUNT|
967 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900968 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700969 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700970 return -ENOMEM;
971 return 0;
972}
973
974static void destroy_inodecache(void)
975{
Mingming Cao617ba132006-10-11 01:20:53 -0700976 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700977}
978
Al Viro0930fcc2010-06-07 13:16:22 -0400979void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700980{
Al Viro0930fcc2010-06-07 13:16:22 -0400981 invalidate_inode_buffers(inode);
982 end_writeback(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -0500983 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400984 ext4_discard_preallocations(inode);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500985 if (EXT4_I(inode)->jinode) {
986 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
987 EXT4_I(inode)->jinode);
988 jbd2_free_inode(EXT4_I(inode)->jinode);
989 EXT4_I(inode)->jinode = NULL;
990 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700991}
992
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400993static inline void ext4_show_quota_options(struct seq_file *seq,
994 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700995{
996#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700997 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700998
Jan Kara5a20bdf2009-11-30 23:58:32 +0100999 if (sbi->s_jquota_fmt) {
1000 char *fmtname = "";
1001
1002 switch (sbi->s_jquota_fmt) {
1003 case QFMT_VFS_OLD:
1004 fmtname = "vfsold";
1005 break;
1006 case QFMT_VFS_V0:
1007 fmtname = "vfsv0";
1008 break;
1009 case QFMT_VFS_V1:
1010 fmtname = "vfsv1";
1011 break;
1012 }
1013 seq_printf(seq, ",jqfmt=%s", fmtname);
1014 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001015
1016 if (sbi->s_qf_names[USRQUOTA])
1017 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1018
1019 if (sbi->s_qf_names[GRPQUOTA])
1020 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1021
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001022 if (test_opt(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001023 seq_puts(seq, ",usrquota");
1024
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001025 if (test_opt(sb, GRPQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001026 seq_puts(seq, ",grpquota");
1027#endif
1028}
1029
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001030/*
1031 * Show an option if
1032 * - it's set to a non-default value OR
1033 * - if the per-sb default is different from the global default
1034 */
Al Viro34c80b12011-12-08 21:32:45 -05001035static int ext4_show_options(struct seq_file *seq, struct dentry *root)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001036{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001037 int def_errors;
1038 unsigned long def_mount_opts;
Al Viro34c80b12011-12-08 21:32:45 -05001039 struct super_block *sb = root->d_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001040 struct ext4_sb_info *sbi = EXT4_SB(sb);
1041 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001042
1043 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001044 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001045
1046 if (sbi->s_sb_block != 1)
1047 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1048 if (test_opt(sb, MINIX_DF))
1049 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001050 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001051 seq_puts(seq, ",grpid");
1052 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1053 seq_puts(seq, ",nogrpid");
1054 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1055 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1056 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1057 }
1058 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1059 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1060 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1061 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05001062 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001063 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05001064 def_errors == EXT4_ERRORS_CONTINUE) {
1065 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001066 }
1067 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001068 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05001069 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001070 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001071 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001072 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001073 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001074 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001075 seq_puts(seq, ",debug");
Theodore Ts'o03010a32008-10-10 20:02:48 -04001076#ifdef CONFIG_EXT4_FS_XATTR
Eric Sandeenea663332011-02-23 17:51:51 -05001077 if (test_opt(sb, XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001078 seq_puts(seq, ",user_xattr");
Eric Sandeenea663332011-02-23 17:51:51 -05001079 if (!test_opt(sb, XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001080 seq_puts(seq, ",nouser_xattr");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001081#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001082#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001083 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001084 seq_puts(seq, ",acl");
1085 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1086 seq_puts(seq, ",noacl");
1087#endif
Theodore Ts'o30773842009-01-03 20:27:38 -05001088 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001089 seq_printf(seq, ",commit=%u",
1090 (unsigned) (sbi->s_commit_interval / HZ));
1091 }
Theodore Ts'o30773842009-01-03 20:27:38 -05001092 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1093 seq_printf(seq, ",min_batch_time=%u",
1094 (unsigned) sbi->s_min_batch_time);
1095 }
1096 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1097 seq_printf(seq, ",max_batch_time=%u",
Ben Hutchings1d526fc2012-01-04 21:22:51 -05001098 (unsigned) sbi->s_max_batch_time);
Theodore Ts'o30773842009-01-03 20:27:38 -05001099 }
1100
Eric Sandeen571640c2008-05-26 12:29:46 -04001101 /*
1102 * We're changing the default of barrier mount option, so
1103 * let's always display its mount state so it's clear what its
1104 * status is.
1105 */
1106 seq_puts(seq, ",barrier=");
1107 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -04001108 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1109 seq_puts(seq, ",journal_async_commit");
Jan Kara39a4bad2010-05-16 17:00:00 -04001110 else if (test_opt(sb, JOURNAL_CHECKSUM))
1111 seq_puts(seq, ",journal_checksum");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001112 if (test_opt(sb, I_VERSION))
1113 seq_puts(seq, ",i_version");
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001114 if (!test_opt(sb, DELALLOC) &&
1115 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001116 seq_puts(seq, ",nodelalloc");
1117
Theodore Ts'o6fd7a462011-02-26 13:53:09 -05001118 if (!test_opt(sb, MBLK_IO_SUBMIT))
1119 seq_puts(seq, ",nomblk_io_submit");
Miklos Szeredicb45bbe2008-01-28 23:58:27 -05001120 if (sbi->s_stripe)
1121 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001122 /*
1123 * journal mode get enabled in different ways
1124 * So just print the value even if we didn't specify it
1125 */
Mingming Cao617ba132006-10-11 01:20:53 -07001126 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001127 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -07001128 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001129 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -07001130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001131 seq_puts(seq, ",data=writeback");
1132
Theodore Ts'o240799c2008-10-09 23:53:47 -04001133 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1134 seq_printf(seq, ",inode_readahead_blks=%u",
1135 sbi->s_inode_readahead_blks);
1136
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001137 if (test_opt(sb, DATA_ERR_ABORT))
1138 seq_puts(seq, ",data_err=abort");
1139
Theodore Ts'oafd46722009-03-16 23:12:23 -04001140 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001141 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -04001142
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001143 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
Eric Sandeen5328e632009-11-19 14:25:42 -05001144 seq_puts(seq, ",discard");
1145
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001146 if (test_opt(sb, NOLOAD))
1147 seq_puts(seq, ",norecovery");
1148
Jiaying Zhang744692d2010-03-04 16:14:02 -05001149 if (test_opt(sb, DIOREAD_NOLOCK))
1150 seq_puts(seq, ",dioread_nolock");
1151
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001152 if (test_opt(sb, BLOCK_VALIDITY) &&
1153 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1154 seq_puts(seq, ",block_validity");
1155
Lukas Czernerbfff6872010-10-27 21:30:05 -04001156 if (!test_opt(sb, INIT_INODE_TABLE))
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001157 seq_puts(seq, ",noinit_itable");
Lukas Czerner51ce6512011-05-20 13:55:16 -04001158 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001159 seq_printf(seq, ",init_itable=%u",
Lukas Czernerbfff6872010-10-27 21:30:05 -04001160 (unsigned) sbi->s_li_wait_mult);
1161
Mingming Cao617ba132006-10-11 01:20:53 -07001162 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001163
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001164 return 0;
1165}
1166
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001167static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001168 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001169{
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001170 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001171
Mingming Cao617ba132006-10-11 01:20:53 -07001172 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001173 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -07001174 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001175 return ERR_PTR(-ESTALE);
1176
1177 /* iget isn't really right if the inode is currently unallocated!!
1178 *
Mingming Cao617ba132006-10-11 01:20:53 -07001179 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001180 * deleted, so we should be safe.
1181 *
1182 * Currently we don't know the generation for parent directory, so
1183 * a generation of 0 means "accept any"
1184 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001185 inode = ext4_iget(sb, ino);
1186 if (IS_ERR(inode))
1187 return ERR_CAST(inode);
1188 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001189 iput(inode);
1190 return ERR_PTR(-ESTALE);
1191 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001192
1193 return inode;
1194}
1195
1196static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001197 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001198{
1199 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1200 ext4_nfs_get_inode);
1201}
1202
1203static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001204 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001205{
1206 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1207 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001208}
1209
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001210/*
1211 * Try to release metadata pages (indirect blocks, directories) which are
1212 * mapped via the block device. Since these pages could have journal heads
1213 * which would prevent try_to_free_buffers() from freeing them, we must use
1214 * jbd2 layer's try_to_free_buffers() function to release them.
1215 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001216static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1217 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001218{
1219 journal_t *journal = EXT4_SB(sb)->s_journal;
1220
1221 WARN_ON(PageChecked(page));
1222 if (!page_has_buffers(page))
1223 return 0;
1224 if (journal)
1225 return jbd2_journal_try_to_free_buffers(journal, page,
1226 wait & ~__GFP_WAIT);
1227 return try_to_free_buffers(page);
1228}
1229
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001230#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001231#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001232#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001233
Mingming Cao617ba132006-10-11 01:20:53 -07001234static int ext4_write_dquot(struct dquot *dquot);
1235static int ext4_acquire_dquot(struct dquot *dquot);
1236static int ext4_release_dquot(struct dquot *dquot);
1237static int ext4_mark_dquot_dirty(struct dquot *dquot);
1238static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001239static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02001240 struct path *path);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001241static int ext4_quota_off(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001242static int ext4_quota_on_mount(struct super_block *sb, int type);
1243static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001244 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001245static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001246 const char *data, size_t len, loff_t off);
1247
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001248static const struct dquot_operations ext4_quota_operations = {
Mingming Cao60e58e02009-01-22 18:13:05 +01001249 .get_reserved_space = ext4_get_reserved_space,
Mingming Cao617ba132006-10-11 01:20:53 -07001250 .write_dquot = ext4_write_dquot,
1251 .acquire_dquot = ext4_acquire_dquot,
1252 .release_dquot = ext4_release_dquot,
1253 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001254 .write_info = ext4_write_info,
1255 .alloc_dquot = dquot_alloc,
1256 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001257};
1258
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001259static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001260 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001261 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001262 .quota_sync = dquot_quota_sync,
1263 .get_info = dquot_get_dqinfo,
1264 .set_info = dquot_set_dqinfo,
1265 .get_dqblk = dquot_get_dqblk,
1266 .set_dqblk = dquot_set_dqblk
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001267};
1268#endif
1269
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001270static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001271 .alloc_inode = ext4_alloc_inode,
1272 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001273 .write_inode = ext4_write_inode,
1274 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001275 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001276 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001277 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001278 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001279 .freeze_fs = ext4_freeze,
1280 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001281 .statfs = ext4_statfs,
1282 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001283 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001284#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001285 .quota_read = ext4_quota_read,
1286 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001287#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001288 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001289};
1290
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001291static const struct super_operations ext4_nojournal_sops = {
1292 .alloc_inode = ext4_alloc_inode,
1293 .destroy_inode = ext4_destroy_inode,
1294 .write_inode = ext4_write_inode,
1295 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001296 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001297 .evict_inode = ext4_evict_inode,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001298 .write_super = ext4_write_super,
1299 .put_super = ext4_put_super,
1300 .statfs = ext4_statfs,
1301 .remount_fs = ext4_remount,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001302 .show_options = ext4_show_options,
1303#ifdef CONFIG_QUOTA
1304 .quota_read = ext4_quota_read,
1305 .quota_write = ext4_quota_write,
1306#endif
1307 .bdev_try_to_free_page = bdev_try_to_free_page,
1308};
1309
Christoph Hellwig39655162007-10-21 16:42:17 -07001310static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001311 .fh_to_dentry = ext4_fh_to_dentry,
1312 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001313 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001314};
1315
1316enum {
1317 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1318 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001319 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001320 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001321 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001322 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001323 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001324 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001325 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001326 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001327 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001328 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1329 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1330 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o14490322010-12-14 15:27:50 -05001331 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1332 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001333 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001334 Opt_dioread_nolock, Opt_dioread_lock,
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001335 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001336};
1337
Steven Whitehousea447c092008-10-13 10:46:57 +01001338static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001339 {Opt_bsd_df, "bsddf"},
1340 {Opt_minix_df, "minixdf"},
1341 {Opt_grpid, "grpid"},
1342 {Opt_grpid, "bsdgroups"},
1343 {Opt_nogrpid, "nogrpid"},
1344 {Opt_nogrpid, "sysvgroups"},
1345 {Opt_resgid, "resgid=%u"},
1346 {Opt_resuid, "resuid=%u"},
1347 {Opt_sb, "sb=%u"},
1348 {Opt_err_cont, "errors=continue"},
1349 {Opt_err_panic, "errors=panic"},
1350 {Opt_err_ro, "errors=remount-ro"},
1351 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001352 {Opt_debug, "debug"},
1353 {Opt_oldalloc, "oldalloc"},
1354 {Opt_orlov, "orlov"},
1355 {Opt_user_xattr, "user_xattr"},
1356 {Opt_nouser_xattr, "nouser_xattr"},
1357 {Opt_acl, "acl"},
1358 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001359 {Opt_noload, "noload"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001360 {Opt_noload, "norecovery"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001361 {Opt_nobh, "nobh"},
1362 {Opt_bh, "bh"},
1363 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001364 {Opt_min_batch_time, "min_batch_time=%u"},
1365 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001366 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001367 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001368 {Opt_journal_checksum, "journal_checksum"},
1369 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001370 {Opt_abort, "abort"},
1371 {Opt_data_journal, "data=journal"},
1372 {Opt_data_ordered, "data=ordered"},
1373 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001374 {Opt_data_err_abort, "data_err=abort"},
1375 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001376 {Opt_offusrjquota, "usrjquota="},
1377 {Opt_usrjquota, "usrjquota=%s"},
1378 {Opt_offgrpjquota, "grpjquota="},
1379 {Opt_grpjquota, "grpjquota=%s"},
1380 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1381 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001382 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001383 {Opt_grpquota, "grpquota"},
1384 {Opt_noquota, "noquota"},
1385 {Opt_quota, "quota"},
1386 {Opt_usrquota, "usrquota"},
1387 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001388 {Opt_barrier, "barrier"},
1389 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001390 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001391 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001392 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001393 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001394 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o14490322010-12-14 15:27:50 -05001395 {Opt_mblk_io_submit, "mblk_io_submit"},
1396 {Opt_nomblk_io_submit, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001397 {Opt_block_validity, "block_validity"},
1398 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001399 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001400 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001401 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001402 {Opt_auto_da_alloc, "auto_da_alloc"},
1403 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001404 {Opt_dioread_nolock, "dioread_nolock"},
1405 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001406 {Opt_discard, "discard"},
1407 {Opt_nodiscard, "nodiscard"},
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001408 {Opt_init_itable, "init_itable=%u"},
1409 {Opt_init_itable, "init_itable"},
1410 {Opt_noinit_itable, "noinit_itable"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001411 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001412};
1413
Mingming Cao617ba132006-10-11 01:20:53 -07001414static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001415{
Mingming Cao617ba132006-10-11 01:20:53 -07001416 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001417 char *options = (char *) *data;
1418
1419 if (!options || strncmp(options, "sb=", 3) != 0)
1420 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001421
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001422 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001423 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001424 sb_block = simple_strtoul(options, &options, 0);
1425 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001426 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001427 (char *) *data);
1428 return 1;
1429 }
1430 if (*options == ',')
1431 options++;
1432 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001433
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001434 return sb_block;
1435}
1436
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001437#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001438static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1439 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001440
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001441#ifdef CONFIG_QUOTA
1442static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1443{
1444 struct ext4_sb_info *sbi = EXT4_SB(sb);
1445 char *qname;
1446
1447 if (sb_any_quota_loaded(sb) &&
1448 !sbi->s_qf_names[qtype]) {
1449 ext4_msg(sb, KERN_ERR,
1450 "Cannot change journaled "
1451 "quota options when quota turned on");
1452 return 0;
1453 }
1454 qname = match_strdup(args);
1455 if (!qname) {
1456 ext4_msg(sb, KERN_ERR,
1457 "Not enough memory for storing quotafile name");
1458 return 0;
1459 }
1460 if (sbi->s_qf_names[qtype] &&
1461 strcmp(sbi->s_qf_names[qtype], qname)) {
1462 ext4_msg(sb, KERN_ERR,
1463 "%s quota file already specified", QTYPE2NAME(qtype));
1464 kfree(qname);
1465 return 0;
1466 }
1467 sbi->s_qf_names[qtype] = qname;
1468 if (strchr(sbi->s_qf_names[qtype], '/')) {
1469 ext4_msg(sb, KERN_ERR,
1470 "quotafile must be on filesystem root");
1471 kfree(sbi->s_qf_names[qtype]);
1472 sbi->s_qf_names[qtype] = NULL;
1473 return 0;
1474 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001475 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001476 return 1;
1477}
1478
1479static int clear_qf_name(struct super_block *sb, int qtype)
1480{
1481
1482 struct ext4_sb_info *sbi = EXT4_SB(sb);
1483
1484 if (sb_any_quota_loaded(sb) &&
1485 sbi->s_qf_names[qtype]) {
1486 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1487 " when quota turned on");
1488 return 0;
1489 }
1490 /*
1491 * The space will be released later when all options are confirmed
1492 * to be correct
1493 */
1494 sbi->s_qf_names[qtype] = NULL;
1495 return 1;
1496}
1497#endif
1498
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001499static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001500 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001501 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001502 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001503{
Mingming Cao617ba132006-10-11 01:20:53 -07001504 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001505 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001506 substring_t args[MAX_OPT_ARGS];
1507 int data_opt = 0;
1508 int option;
1509#ifdef CONFIG_QUOTA
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001510 int qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001511#endif
1512
1513 if (!options)
1514 return 1;
1515
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001516 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001517 int token;
1518 if (!*p)
1519 continue;
1520
Eric Sandeen15121c12010-02-15 20:17:55 -05001521 /*
1522 * Initialize args struct so we know whether arg was
1523 * found; some options take optional arguments.
1524 */
Peter Huewe7dc57612011-02-21 21:01:42 -05001525 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001526 token = match_token(p, tokens, args);
1527 switch (token) {
1528 case Opt_bsd_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001529 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001530 clear_opt(sb, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001531 break;
1532 case Opt_minix_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001533 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001534 set_opt(sb, MINIX_DF);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001535
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001536 break;
1537 case Opt_grpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001538 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001539 set_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001540
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001541 break;
1542 case Opt_nogrpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001543 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001544 clear_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001545
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001546 break;
1547 case Opt_resuid:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 sbi->s_resuid = option;
1551 break;
1552 case Opt_resgid:
1553 if (match_int(&args[0], &option))
1554 return 0;
1555 sbi->s_resgid = option;
1556 break;
1557 case Opt_sb:
1558 /* handled by get_sb_block() instead of here */
1559 /* *sb_block = match_int(&args[0]); */
1560 break;
1561 case Opt_err_panic:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001562 clear_opt(sb, ERRORS_CONT);
1563 clear_opt(sb, ERRORS_RO);
1564 set_opt(sb, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001565 break;
1566 case Opt_err_ro:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001567 clear_opt(sb, ERRORS_CONT);
1568 clear_opt(sb, ERRORS_PANIC);
1569 set_opt(sb, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001570 break;
1571 case Opt_err_cont:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001572 clear_opt(sb, ERRORS_RO);
1573 clear_opt(sb, ERRORS_PANIC);
1574 set_opt(sb, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001575 break;
1576 case Opt_nouid32:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001577 set_opt(sb, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001578 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001579 case Opt_debug:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001580 set_opt(sb, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001581 break;
1582 case Opt_oldalloc:
Lukas Czerner4113c4c2011-10-08 14:34:47 -04001583 ext4_msg(sb, KERN_WARNING,
1584 "Ignoring deprecated oldalloc option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001585 break;
1586 case Opt_orlov:
Lukas Czerner4113c4c2011-10-08 14:34:47 -04001587 ext4_msg(sb, KERN_WARNING,
1588 "Ignoring deprecated orlov option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001589 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001590#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001591 case Opt_user_xattr:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001592 set_opt(sb, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001593 break;
1594 case Opt_nouser_xattr:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001595 clear_opt(sb, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001596 break;
1597#else
1598 case Opt_user_xattr:
1599 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001600 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001601 break;
1602#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001603#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001604 case Opt_acl:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001605 set_opt(sb, POSIX_ACL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001606 break;
1607 case Opt_noacl:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001608 clear_opt(sb, POSIX_ACL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001609 break;
1610#else
1611 case Opt_acl:
1612 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001613 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001614 break;
1615#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001616 case Opt_journal_update:
1617 /* @@@ FIXME */
1618 /* Eventually we will want to be able to create
1619 a journal file here. For now, only allow the
1620 user to specify an existing inode to be the
1621 journal file. */
1622 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001623 ext4_msg(sb, KERN_ERR,
1624 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001625 return 0;
1626 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001627 set_opt(sb, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001628 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001629 case Opt_journal_dev:
1630 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001631 ext4_msg(sb, KERN_ERR,
1632 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001633 return 0;
1634 }
1635 if (match_int(&args[0], &option))
1636 return 0;
1637 *journal_devnum = option;
1638 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001639 case Opt_journal_checksum:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001640 set_opt(sb, JOURNAL_CHECKSUM);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001641 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001642 case Opt_journal_async_commit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001643 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1644 set_opt(sb, JOURNAL_CHECKSUM);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001645 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001646 case Opt_noload:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001647 set_opt(sb, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001648 break;
1649 case Opt_commit:
1650 if (match_int(&args[0], &option))
1651 return 0;
1652 if (option < 0)
1653 return 0;
1654 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001655 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001656 sbi->s_commit_interval = HZ * option;
1657 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001658 case Opt_max_batch_time:
1659 if (match_int(&args[0], &option))
1660 return 0;
1661 if (option < 0)
1662 return 0;
1663 if (option == 0)
1664 option = EXT4_DEF_MAX_BATCH_TIME;
1665 sbi->s_max_batch_time = option;
1666 break;
1667 case Opt_min_batch_time:
1668 if (match_int(&args[0], &option))
1669 return 0;
1670 if (option < 0)
1671 return 0;
1672 sbi->s_min_batch_time = option;
1673 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001674 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001675 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001676 goto datacheck;
1677 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001678 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001679 goto datacheck;
1680 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001681 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001682 datacheck:
1683 if (is_remount) {
Theodore Ts'oeb513682011-11-07 10:47:42 -05001684 if (!sbi->s_journal)
1685 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1686 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001687 ext4_msg(sb, KERN_ERR,
1688 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001689 return 0;
1690 }
1691 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001692 clear_opt(sb, DATA_FLAGS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001693 sbi->s_mount_opt |= data_opt;
1694 }
1695 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001696 case Opt_data_err_abort:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001697 set_opt(sb, DATA_ERR_ABORT);
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001698 break;
1699 case Opt_data_err_ignore:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001700 clear_opt(sb, DATA_ERR_ABORT);
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001701 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001702#ifdef CONFIG_QUOTA
1703 case Opt_usrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001704 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1705 return 0;
1706 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001707 case Opt_grpjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001708 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001709 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001710 break;
1711 case Opt_offusrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001712 if (!clear_qf_name(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001713 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001714 break;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001715 case Opt_offgrpjquota:
1716 if (!clear_qf_name(sb, GRPQUOTA))
1717 return 0;
1718 break;
1719
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001720 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001721 qfmt = QFMT_VFS_OLD;
1722 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001723 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001724 qfmt = QFMT_VFS_V0;
Jan Kara5a20bdf2009-11-30 23:58:32 +01001725 goto set_qf_format;
1726 case Opt_jqfmt_vfsv1:
1727 qfmt = QFMT_VFS_V1;
Jan Karadfc5d032008-05-13 19:11:51 -04001728set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001729 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001730 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001731 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001732 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001733 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001734 return 0;
1735 }
1736 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001737 break;
1738 case Opt_quota:
1739 case Opt_usrquota:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001740 set_opt(sb, QUOTA);
1741 set_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001742 break;
1743 case Opt_grpquota:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001744 set_opt(sb, QUOTA);
1745 set_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001746 break;
1747 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001748 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001749 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1750 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001751 return 0;
1752 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001753 clear_opt(sb, QUOTA);
1754 clear_opt(sb, USRQUOTA);
1755 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001756 break;
1757#else
1758 case Opt_quota:
1759 case Opt_usrquota:
1760 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001761 ext4_msg(sb, KERN_ERR,
1762 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001763 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001764 case Opt_usrjquota:
1765 case Opt_grpjquota:
1766 case Opt_offusrjquota:
1767 case Opt_offgrpjquota:
1768 case Opt_jqfmt_vfsold:
1769 case Opt_jqfmt_vfsv0:
Jan Kara5a20bdf2009-11-30 23:58:32 +01001770 case Opt_jqfmt_vfsv1:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001771 ext4_msg(sb, KERN_ERR,
1772 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001773 break;
1774 case Opt_noquota:
1775 break;
1776#endif
1777 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001778 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001779 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001780 case Opt_nobarrier:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001781 clear_opt(sb, BARRIER);
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001782 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001783 case Opt_barrier:
Eric Sandeen15121c12010-02-15 20:17:55 -05001784 if (args[0].from) {
1785 if (match_int(&args[0], &option))
1786 return 0;
1787 } else
1788 option = 1; /* No argument, default to 1 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001789 if (option)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001790 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001791 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001792 clear_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001793 break;
1794 case Opt_ignore:
1795 break;
1796 case Opt_resize:
1797 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001798 ext4_msg(sb, KERN_ERR,
1799 "resize option only available "
1800 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001801 return 0;
1802 }
1803 if (match_int(&args[0], &option) != 0)
1804 return 0;
1805 *n_blocks_count = option;
1806 break;
1807 case Opt_nobh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001808 ext4_msg(sb, KERN_WARNING,
1809 "Ignoring deprecated nobh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001810 break;
1811 case Opt_bh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001812 ext4_msg(sb, KERN_WARNING,
1813 "Ignoring deprecated bh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001814 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001815 case Opt_i_version:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001816 set_opt(sb, I_VERSION);
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001817 sb->s_flags |= MS_I_VERSION;
1818 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001819 case Opt_nodelalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001820 clear_opt(sb, DELALLOC);
Theodore Ts'o56889782011-09-03 18:22:38 -04001821 clear_opt2(sb, EXPLICIT_DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001822 break;
Theodore Ts'o14490322010-12-14 15:27:50 -05001823 case Opt_mblk_io_submit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001824 set_opt(sb, MBLK_IO_SUBMIT);
Theodore Ts'o14490322010-12-14 15:27:50 -05001825 break;
1826 case Opt_nomblk_io_submit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001827 clear_opt(sb, MBLK_IO_SUBMIT);
Theodore Ts'o14490322010-12-14 15:27:50 -05001828 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001829 case Opt_stripe:
1830 if (match_int(&args[0], &option))
1831 return 0;
1832 if (option < 0)
1833 return 0;
1834 sbi->s_stripe = option;
1835 break;
Alex Tomas64769242008-07-11 19:27:31 -04001836 case Opt_delalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001837 set_opt(sb, DELALLOC);
Theodore Ts'o56889782011-09-03 18:22:38 -04001838 set_opt2(sb, EXPLICIT_DELALLOC);
Alex Tomas64769242008-07-11 19:27:31 -04001839 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001840 case Opt_block_validity:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001841 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001842 break;
1843 case Opt_noblock_validity:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001844 clear_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001845 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001846 case Opt_inode_readahead_blks:
1847 if (match_int(&args[0], &option))
1848 return 0;
1849 if (option < 0 || option > (1 << 30))
1850 return 0;
Alexander V. Lukyanov5dbd5712011-02-21 21:33:21 -05001851 if (option && !is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001852 ext4_msg(sb, KERN_ERR,
1853 "EXT4-fs: inode_readahead_blks"
1854 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001855 return 0;
1856 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001857 sbi->s_inode_readahead_blks = option;
1858 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001859 case Opt_journal_ioprio:
1860 if (match_int(&args[0], &option))
1861 return 0;
1862 if (option < 0 || option > 7)
1863 break;
1864 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1865 option);
1866 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001867 case Opt_noauto_da_alloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001868 set_opt(sb, NO_AUTO_DA_ALLOC);
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001869 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001870 case Opt_auto_da_alloc:
Eric Sandeen15121c12010-02-15 20:17:55 -05001871 if (args[0].from) {
1872 if (match_int(&args[0], &option))
1873 return 0;
1874 } else
1875 option = 1; /* No argument, default to 1 */
Theodore Ts'oafd46722009-03-16 23:12:23 -04001876 if (option)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001877 clear_opt(sb, NO_AUTO_DA_ALLOC);
Theodore Ts'oafd46722009-03-16 23:12:23 -04001878 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001879 set_opt(sb,NO_AUTO_DA_ALLOC);
Theodore Ts'oafd46722009-03-16 23:12:23 -04001880 break;
Eric Sandeen5328e632009-11-19 14:25:42 -05001881 case Opt_discard:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001882 set_opt(sb, DISCARD);
Eric Sandeen5328e632009-11-19 14:25:42 -05001883 break;
1884 case Opt_nodiscard:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001885 clear_opt(sb, DISCARD);
Eric Sandeen5328e632009-11-19 14:25:42 -05001886 break;
Jiaying Zhang744692d2010-03-04 16:14:02 -05001887 case Opt_dioread_nolock:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001888 set_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001889 break;
1890 case Opt_dioread_lock:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001891 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001892 break;
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001893 case Opt_init_itable:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001894 set_opt(sb, INIT_INODE_TABLE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001895 if (args[0].from) {
1896 if (match_int(&args[0], &option))
1897 return 0;
1898 } else
1899 option = EXT4_DEF_LI_WAIT_MULT;
1900 if (option < 0)
1901 return 0;
1902 sbi->s_li_wait_mult = option;
1903 break;
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001904 case Opt_noinit_itable:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001905 clear_opt(sb, INIT_INODE_TABLE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001906 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001907 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001908 ext4_msg(sb, KERN_ERR,
1909 "Unrecognized mount option \"%s\" "
1910 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001911 return 0;
1912 }
1913 }
1914#ifdef CONFIG_QUOTA
1915 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001916 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001917 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001918
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001919 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001920 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001921
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001922 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001923 ext4_msg(sb, KERN_ERR, "old and new quota "
1924 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001925 return 0;
1926 }
1927
1928 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001929 ext4_msg(sb, KERN_ERR, "journaled quota format "
1930 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001931 return 0;
1932 }
1933 } else {
1934 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001935 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001936 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001937 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001938 return 0;
1939 }
1940 }
1941#endif
1942 return 1;
1943}
1944
Mingming Cao617ba132006-10-11 01:20:53 -07001945static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001946 int read_only)
1947{
Mingming Cao617ba132006-10-11 01:20:53 -07001948 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001949 int res = 0;
1950
Mingming Cao617ba132006-10-11 01:20:53 -07001951 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001952 ext4_msg(sb, KERN_ERR, "revision level too high, "
1953 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001954 res = MS_RDONLY;
1955 }
1956 if (read_only)
Theodore Ts'o281b5992011-09-09 18:34:51 -04001957 goto done;
Mingming Cao617ba132006-10-11 01:20:53 -07001958 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1960 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001961 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001962 ext4_msg(sb, KERN_WARNING,
1963 "warning: mounting fs with errors, "
1964 "running e2fsck is recommended");
Tao Maed3ce802011-05-18 13:29:57 -04001965 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001966 le16_to_cpu(es->s_mnt_count) >=
1967 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001968 ext4_msg(sb, KERN_WARNING,
1969 "warning: maximal mount count reached, "
1970 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001971 else if (le32_to_cpu(es->s_checkinterval) &&
1972 (le32_to_cpu(es->s_lastcheck) +
1973 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001974 ext4_msg(sb, KERN_WARNING,
1975 "warning: checktime reached, "
1976 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001977 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001978 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001979 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001980 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001981 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001982 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001983 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001984 if (sbi->s_journal)
1985 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001986
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001987 ext4_commit_super(sb, 1);
Theodore Ts'o281b5992011-09-09 18:34:51 -04001988done:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001989 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001992 sb->s_blocksize,
1993 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001994 EXT4_BLOCKS_PER_GROUP(sb),
1995 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001996 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001997
Dan Magenheimer7abc52c2011-05-26 10:02:03 -06001998 cleancache_init_fs(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001999 return res;
2000}
2001
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002002static int ext4_fill_flex_info(struct super_block *sb)
2003{
2004 struct ext4_sb_info *sbi = EXT4_SB(sb);
2005 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002006 ext4_group_t flex_group_count;
2007 ext4_group_t flex_group;
Xi Wangd50f2ab2012-01-10 11:51:10 -05002008 unsigned int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04002009 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002010 int i;
2011
Theodore Ts'o503358a2009-11-23 07:24:46 -05002012 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
Xi Wangd50f2ab2012-01-10 11:51:10 -05002013 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002014 sbi->s_log_groups_per_flex = 0;
2015 return 1;
2016 }
Xi Wangd50f2ab2012-01-10 11:51:10 -05002017 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002018
Frederic Bohec62a11f2008-09-08 10:20:24 -04002019 /* We allocate both existing and potentially added groups */
2020 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05002021 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2022 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04002023 size = flex_group_count * sizeof(struct flex_groups);
Theodore Ts'o9933fc02011-08-01 08:45:02 -04002024 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04002025 if (sbi->s_flex_groups == NULL) {
Theodore Ts'o9933fc02011-08-01 08:45:02 -04002026 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2027 flex_group_count);
2028 goto failed;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002029 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002030
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002031 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04002032 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002033
2034 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002035 atomic_add(ext4_free_inodes_count(sb, gdp),
2036 &sbi->s_flex_groups[flex_group].free_inodes);
Theodore Ts'o021b65b2011-09-09 19:08:51 -04002037 atomic_add(ext4_free_group_clusters(sb, gdp),
Theodore Ts'o24aaa8e2011-09-09 18:58:51 -04002038 &sbi->s_flex_groups[flex_group].free_clusters);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002039 atomic_add(ext4_used_dirs_count(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002041 }
2042
2043 return 1;
2044failed:
2045 return 0;
2046}
2047
Andreas Dilger717d50e2007-10-16 18:38:25 -04002048__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2050{
2051 __u16 crc = 0;
2052
2053 if (sbi->s_es->s_feature_ro_compat &
2054 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2055 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2056 __le32 le_group = cpu_to_le32(block_group);
2057
2058 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2059 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2060 crc = crc16(crc, (__u8 *)gdp, offset);
2061 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2062 /* for checksum of struct ext4_group_desc do the rest...*/
2063 if ((sbi->s_es->s_feature_incompat &
2064 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2065 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2066 crc = crc16(crc, (__u8 *)gdp + offset,
2067 le16_to_cpu(sbi->s_es->s_desc_size) -
2068 offset);
2069 }
2070
2071 return cpu_to_le16(crc);
2072}
2073
2074int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2075 struct ext4_group_desc *gdp)
2076{
2077 if ((sbi->s_es->s_feature_ro_compat &
2078 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2079 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2080 return 0;
2081
2082 return 1;
2083}
2084
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002085/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002086static int ext4_check_descriptors(struct super_block *sb,
2087 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002088{
Mingming Cao617ba132006-10-11 01:20:53 -07002089 struct ext4_sb_info *sbi = EXT4_SB(sb);
2090 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2091 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002092 ext4_fsblk_t block_bitmap;
2093 ext4_fsblk_t inode_bitmap;
2094 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002095 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002096 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002097
Jose R. Santosce421582007-10-16 18:38:25 -04002098 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2099 flexbg_flag = 1;
2100
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002101 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002102
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002103 for (i = 0; i < sbi->s_groups_count; i++) {
2104 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2105
Jose R. Santosce421582007-10-16 18:38:25 -04002106 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002107 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002108 else
2109 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002110 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002111
Lukas Czernerbfff6872010-10-27 21:30:05 -04002112 if ((grp == sbi->s_groups_count) &&
2113 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2114 grp = i;
2115
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002116 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002117 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002118 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002119 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002120 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002121 return 0;
2122 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002123 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002124 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002126 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002127 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002128 return 0;
2129 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002130 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002131 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002132 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002133 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002134 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002135 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002136 return 0;
2137 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002138 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002139 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002140 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2141 "Checksum for group %u failed (%u!=%u)",
2142 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2143 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002144 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002145 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002146 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002147 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002148 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002149 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002150 if (!flexbg_flag)
2151 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002152 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002153 if (NULL != first_not_zeroed)
2154 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002155
Theodore Ts'o5dee5432011-09-09 19:10:51 -04002156 ext4_free_blocks_count_set(sbi->s_es,
2157 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002158 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002159 return 1;
2160}
2161
Mingming Cao617ba132006-10-11 01:20:53 -07002162/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002163 * the superblock) which were deleted from all directories, but held open by
2164 * a process at the time of a crash. We walk the list and try to delete these
2165 * inodes at recovery time (only with a read-write filesystem).
2166 *
2167 * In order to keep the orphan inode chain consistent during traversal (in
2168 * case of crash during recovery), we link each inode into the superblock
2169 * orphan list_head and handle it the same way as an inode deletion during
2170 * normal operation (which journals the operations for us).
2171 *
2172 * We only do an iget() and an iput() on each inode, which is very safe if we
2173 * accidentally point at an in-use or already deleted inode. The worst that
2174 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002175 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002176 * e2fsck was run on this filesystem, and it must have already done the orphan
2177 * inode cleanup for us, so we can safely abort without any further action.
2178 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002179static void ext4_orphan_cleanup(struct super_block *sb,
2180 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002181{
2182 unsigned int s_flags = sb->s_flags;
2183 int nr_orphans = 0, nr_truncates = 0;
2184#ifdef CONFIG_QUOTA
2185 int i;
2186#endif
2187 if (!es->s_last_orphan) {
2188 jbd_debug(4, "no orphan inodes to clean up\n");
2189 return;
2190 }
2191
Eric Sandeena8f48a92006-12-06 20:40:13 -08002192 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002193 ext4_msg(sb, KERN_ERR, "write access "
2194 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002195 return;
2196 }
2197
Amir Goldsteind39195c2011-02-28 00:53:45 -05002198 /* Check if feature set would not allow a r/w mount */
2199 if (!ext4_feature_set_ok(sb, 0)) {
2200 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2201 "unknown ROCOMPAT features");
2202 return;
2203 }
2204
Mingming Cao617ba132006-10-11 01:20:53 -07002205 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002206 if (es->s_last_orphan)
2207 jbd_debug(1, "Errors on filesystem, "
2208 "clearing orphan list.\n");
2209 es->s_last_orphan = 0;
2210 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2211 return;
2212 }
2213
2214 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002215 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002216 sb->s_flags &= ~MS_RDONLY;
2217 }
2218#ifdef CONFIG_QUOTA
2219 /* Needed for iput() to work correctly and not trash data */
2220 sb->s_flags |= MS_ACTIVE;
2221 /* Turn on quotas so that they are updated correctly */
2222 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002223 if (EXT4_SB(sb)->s_qf_names[i]) {
2224 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002225 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002226 ext4_msg(sb, KERN_ERR,
2227 "Cannot turn on journaled "
2228 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002229 }
2230 }
2231#endif
2232
2233 while (es->s_last_orphan) {
2234 struct inode *inode;
2235
Josef Bacik97bd42b2008-04-29 22:04:56 -04002236 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2237 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002238 es->s_last_orphan = 0;
2239 break;
2240 }
2241
Mingming Cao617ba132006-10-11 01:20:53 -07002242 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002243 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002244 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002245 ext4_msg(sb, KERN_DEBUG,
2246 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002247 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002248 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002249 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002250 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002251 nr_truncates++;
2252 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002253 ext4_msg(sb, KERN_DEBUG,
2254 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002255 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002256 jbd_debug(2, "deleting unreferenced inode %lu\n",
2257 inode->i_ino);
2258 nr_orphans++;
2259 }
2260 iput(inode); /* The delete magic happens here! */
2261 }
2262
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002263#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002264
2265 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002266 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2267 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002268 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002269 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2270 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002271#ifdef CONFIG_QUOTA
2272 /* Turn quotas off */
2273 for (i = 0; i < MAXQUOTAS; i++) {
2274 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002275 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002276 }
2277#endif
2278 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2279}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002280
Eric Sandeencd2291a2008-01-28 23:58:27 -05002281/*
2282 * Maximal extent format file size.
2283 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2284 * extent format containers, within a sector_t, and within i_blocks
2285 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2286 * so that won't be a limiting factor.
2287 *
Lukas Czernerf17722f2011-06-06 00:05:17 -04002288 * However there is other limiting factor. We do store extents in the form
2289 * of starting block and length, hence the resulting length of the extent
2290 * covering maximum file size must fit into on-disk format containers as
2291 * well. Given that length is always by 1 unit bigger than max unit (because
2292 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2293 *
Eric Sandeencd2291a2008-01-28 23:58:27 -05002294 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2295 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002296static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002297{
2298 loff_t res;
2299 loff_t upper_limit = MAX_LFS_FILESIZE;
2300
2301 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002302 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002303 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002304 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002305 * i_block represent total blocks in 512 bytes
2306 * 32 == size of vfs inode i_blocks * 8
2307 */
2308 upper_limit = (1LL << 32) - 1;
2309
2310 /* total blocks in file system block size */
2311 upper_limit >>= (blkbits - 9);
2312 upper_limit <<= blkbits;
2313 }
2314
Lukas Czernerf17722f2011-06-06 00:05:17 -04002315 /*
2316 * 32-bit extent-start container, ee_block. We lower the maxbytes
2317 * by one fs block, so ee_len can cover the extent of maximum file
2318 * size
2319 */
2320 res = (1LL << 32) - 1;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002321 res <<= blkbits;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002322
2323 /* Sanity check against vm- & vfs- imposed limits */
2324 if (res > upper_limit)
2325 res = upper_limit;
2326
2327 return res;
2328}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002329
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002330/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002331 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002332 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2333 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002334 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002335static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002336{
Mingming Cao617ba132006-10-11 01:20:53 -07002337 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002338 int meta_blocks;
2339 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002340 /* This is calculated to be the largest file size for a dense, block
2341 * mapped file such that the file's total number of 512-byte sectors,
2342 * including data and all indirect blocks, does not exceed (2^48 - 1).
2343 *
2344 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2345 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002346 */
2347
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002348 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002349 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002350 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002351 * the inode i_block field represents total file blocks in
2352 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002353 */
2354 upper_limit = (1LL << 32) - 1;
2355
2356 /* total blocks in file system block size */
2357 upper_limit >>= (bits - 9);
2358
2359 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002360 /*
2361 * We use 48 bit ext4_inode i_blocks
2362 * With EXT4_HUGE_FILE_FL set the i_blocks
2363 * represent total number of blocks in
2364 * file system block size
2365 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002366 upper_limit = (1LL << 48) - 1;
2367
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002368 }
2369
2370 /* indirect blocks */
2371 meta_blocks = 1;
2372 /* double indirect blocks */
2373 meta_blocks += 1 + (1LL << (bits-2));
2374 /* tripple indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2376
2377 upper_limit -= meta_blocks;
2378 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002379
2380 res += 1LL << (bits-2);
2381 res += 1LL << (2*(bits-2));
2382 res += 1LL << (3*(bits-2));
2383 res <<= bits;
2384 if (res > upper_limit)
2385 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002386
2387 if (res > MAX_LFS_FILESIZE)
2388 res = MAX_LFS_FILESIZE;
2389
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002390 return res;
2391}
2392
Mingming Cao617ba132006-10-11 01:20:53 -07002393static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002394 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002395{
Mingming Cao617ba132006-10-11 01:20:53 -07002396 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002397 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002398 int has_super = 0;
2399
2400 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2401
Mingming Cao617ba132006-10-11 01:20:53 -07002402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002403 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002404 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002405 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002406 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002407 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002408
Mingming Cao617ba132006-10-11 01:20:53 -07002409 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002410}
2411
Alex Tomasc9de5602008-01-29 00:19:52 -05002412/**
2413 * ext4_get_stripe_size: Get the stripe size.
2414 * @sbi: In memory super block info
2415 *
2416 * If we have specified it via mount option, then
2417 * use the mount option value. If the value specified at mount time is
2418 * greater than the blocks per group use the super block value.
2419 * If the super block value is greater than blocks per group return 0.
2420 * Allocator needs it be less than blocks per group.
2421 *
2422 */
2423static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2424{
2425 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2426 unsigned long stripe_width =
2427 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002428 int ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002429
2430 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002431 ret = sbi->s_stripe;
2432 else if (stripe_width <= sbi->s_blocks_per_group)
2433 ret = stripe_width;
2434 else if (stride <= sbi->s_blocks_per_group)
2435 ret = stride;
2436 else
2437 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002438
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002439 /*
2440 * If the stripe width is 1, this makes no sense and
2441 * we set it to 0 to turn off stripe handling code.
2442 */
2443 if (ret <= 1)
2444 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002445
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002446 return ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002447}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002448
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002449/* sysfs supprt */
2450
2451struct ext4_attr {
2452 struct attribute attr;
2453 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
Theodore Ts'o60e66792010-05-17 07:00:00 -04002454 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002455 const char *, size_t);
2456 int offset;
2457};
2458
2459static int parse_strtoul(const char *buf,
2460 unsigned long max, unsigned long *value)
2461{
2462 char *endp;
2463
André Goddard Rosae7d28602009-12-14 18:01:06 -08002464 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2465 endp = skip_spaces(endp);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002466 if (*endp || *value > max)
2467 return -EINVAL;
2468
2469 return 0;
2470}
2471
2472static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi,
2474 char *buf)
2475{
2476 return snprintf(buf, PAGE_SIZE, "%llu\n",
Aditya Kali7b415bf2011-09-09 19:04:51 -04002477 (s64) EXT4_C2B(sbi,
2478 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002479}
2480
2481static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi, char *buf)
2483{
2484 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2485
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002486 if (!sb->s_bdev->bd_part)
2487 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002488 return snprintf(buf, PAGE_SIZE, "%lu\n",
2489 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2490 sbi->s_sectors_written_start) >> 1);
2491}
2492
2493static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2495{
2496 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2497
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002498 if (!sb->s_bdev->bd_part)
2499 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002500 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002501 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002502 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002503 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002504}
2505
Vivek Haldar77f41352011-05-22 21:24:16 -04002506static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi, char *buf)
2508{
2509 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2510}
2511
2512static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2513 struct ext4_sb_info *sbi, char *buf)
2514{
2515 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2516}
2517
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002518static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2521{
2522 unsigned long t;
2523
2524 if (parse_strtoul(buf, 0x40000000, &t))
2525 return -EINVAL;
2526
Alexander V. Lukyanov5dbd5712011-02-21 21:33:21 -05002527 if (t && !is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002528 return -EINVAL;
2529
2530 sbi->s_inode_readahead_blks = t;
2531 return count;
2532}
2533
2534static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002535 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002536{
2537 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2538
2539 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2540}
2541
2542static ssize_t sbi_ui_store(struct ext4_attr *a,
2543 struct ext4_sb_info *sbi,
2544 const char *buf, size_t count)
2545{
2546 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2547 unsigned long t;
2548
2549 if (parse_strtoul(buf, 0xffffffff, &t))
2550 return -EINVAL;
2551 *ui = t;
2552 return count;
2553}
2554
2555#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2556static struct ext4_attr ext4_attr_##_name = { \
2557 .attr = {.name = __stringify(_name), .mode = _mode }, \
2558 .show = _show, \
2559 .store = _store, \
2560 .offset = offsetof(struct ext4_sb_info, _elname), \
2561}
2562#define EXT4_ATTR(name, mode, show, store) \
2563static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2564
Lukas Czerner857ac882010-10-27 21:30:05 -04002565#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002566#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2567#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2568#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2569 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2570#define ATTR_LIST(name) &ext4_attr_##name.attr
2571
2572EXT4_RO_ATTR(delayed_allocation_blocks);
2573EXT4_RO_ATTR(session_write_kbytes);
2574EXT4_RO_ATTR(lifetime_write_kbytes);
Vivek Haldar77f41352011-05-22 21:24:16 -04002575EXT4_RO_ATTR(extent_cache_hits);
2576EXT4_RO_ATTR(extent_cache_misses);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002577EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2578 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002579EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002580EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2581EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2582EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2583EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2584EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2585EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002586EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002587
2588static struct attribute *ext4_attrs[] = {
2589 ATTR_LIST(delayed_allocation_blocks),
2590 ATTR_LIST(session_write_kbytes),
2591 ATTR_LIST(lifetime_write_kbytes),
Vivek Haldar77f41352011-05-22 21:24:16 -04002592 ATTR_LIST(extent_cache_hits),
2593 ATTR_LIST(extent_cache_misses),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002594 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002595 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002596 ATTR_LIST(mb_stats),
2597 ATTR_LIST(mb_max_to_scan),
2598 ATTR_LIST(mb_min_to_scan),
2599 ATTR_LIST(mb_order2_req),
2600 ATTR_LIST(mb_stream_req),
2601 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002602 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002603 NULL,
2604};
2605
Lukas Czerner857ac882010-10-27 21:30:05 -04002606/* Features this copy of ext4 supports */
2607EXT4_INFO_ATTR(lazy_itable_init);
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002608EXT4_INFO_ATTR(batched_discard);
Lukas Czerner857ac882010-10-27 21:30:05 -04002609
2610static struct attribute *ext4_feat_attrs[] = {
2611 ATTR_LIST(lazy_itable_init),
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002612 ATTR_LIST(batched_discard),
Lukas Czerner857ac882010-10-27 21:30:05 -04002613 NULL,
2614};
2615
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002616static ssize_t ext4_attr_show(struct kobject *kobj,
2617 struct attribute *attr, char *buf)
2618{
2619 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2620 s_kobj);
2621 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2622
2623 return a->show ? a->show(a, sbi, buf) : 0;
2624}
2625
2626static ssize_t ext4_attr_store(struct kobject *kobj,
2627 struct attribute *attr,
2628 const char *buf, size_t len)
2629{
2630 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2631 s_kobj);
2632 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2633
2634 return a->store ? a->store(a, sbi, buf, len) : 0;
2635}
2636
2637static void ext4_sb_release(struct kobject *kobj)
2638{
2639 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2640 s_kobj);
2641 complete(&sbi->s_kobj_unregister);
2642}
2643
Emese Revfy52cf25d2010-01-19 02:58:23 +01002644static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002645 .show = ext4_attr_show,
2646 .store = ext4_attr_store,
2647};
2648
2649static struct kobj_type ext4_ktype = {
2650 .default_attrs = ext4_attrs,
2651 .sysfs_ops = &ext4_attr_ops,
2652 .release = ext4_sb_release,
2653};
2654
Lukas Czerner857ac882010-10-27 21:30:05 -04002655static void ext4_feat_release(struct kobject *kobj)
2656{
2657 complete(&ext4_feat->f_kobj_unregister);
2658}
2659
2660static struct kobj_type ext4_feat_ktype = {
2661 .default_attrs = ext4_feat_attrs,
2662 .sysfs_ops = &ext4_attr_ops,
2663 .release = ext4_feat_release,
2664};
2665
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002666/*
2667 * Check whether this filesystem can be mounted based on
2668 * the features present and the RDONLY/RDWR mount requested.
2669 * Returns 1 if this filesystem can be mounted as requested,
2670 * 0 if it cannot be.
2671 */
2672static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2673{
2674 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2675 ext4_msg(sb, KERN_ERR,
2676 "Couldn't mount because of "
2677 "unsupported optional features (%x)",
2678 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2679 ~EXT4_FEATURE_INCOMPAT_SUPP));
2680 return 0;
2681 }
2682
2683 if (readonly)
2684 return 1;
2685
2686 /* Check that feature set is OK for a read-write mount */
2687 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2688 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2689 "unsupported optional features (%x)",
2690 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2691 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2692 return 0;
2693 }
2694 /*
2695 * Large file size enabled file system can only be mounted
2696 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2697 */
2698 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2699 if (sizeof(blkcnt_t) < sizeof(u64)) {
2700 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2701 "cannot be mounted RDWR without "
2702 "CONFIG_LBDAF");
2703 return 0;
2704 }
2705 }
Theodore Ts'obab08ab2011-09-09 18:36:51 -04002706 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2707 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2708 ext4_msg(sb, KERN_ERR,
2709 "Can't support bigalloc feature without "
2710 "extents feature\n");
2711 return 0;
2712 }
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002713 return 1;
2714}
2715
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002716/*
2717 * This function is called once a day if we have errors logged
2718 * on the file system
2719 */
2720static void print_daily_error_info(unsigned long arg)
2721{
2722 struct super_block *sb = (struct super_block *) arg;
2723 struct ext4_sb_info *sbi;
2724 struct ext4_super_block *es;
2725
2726 sbi = EXT4_SB(sb);
2727 es = sbi->s_es;
2728
2729 if (es->s_error_count)
2730 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2731 le32_to_cpu(es->s_error_count));
2732 if (es->s_first_error_time) {
2733 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2734 sb->s_id, le32_to_cpu(es->s_first_error_time),
2735 (int) sizeof(es->s_first_error_func),
2736 es->s_first_error_func,
2737 le32_to_cpu(es->s_first_error_line));
2738 if (es->s_first_error_ino)
2739 printk(": inode %u",
2740 le32_to_cpu(es->s_first_error_ino));
2741 if (es->s_first_error_block)
2742 printk(": block %llu", (unsigned long long)
2743 le64_to_cpu(es->s_first_error_block));
2744 printk("\n");
2745 }
2746 if (es->s_last_error_time) {
2747 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2748 sb->s_id, le32_to_cpu(es->s_last_error_time),
2749 (int) sizeof(es->s_last_error_func),
2750 es->s_last_error_func,
2751 le32_to_cpu(es->s_last_error_line));
2752 if (es->s_last_error_ino)
2753 printk(": inode %u",
2754 le32_to_cpu(es->s_last_error_ino));
2755 if (es->s_last_error_block)
2756 printk(": block %llu", (unsigned long long)
2757 le64_to_cpu(es->s_last_error_block));
2758 printk("\n");
2759 }
2760 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2761}
2762
Lukas Czernerbfff6872010-10-27 21:30:05 -04002763/* Find next suitable group and run ext4_init_inode_table */
2764static int ext4_run_li_request(struct ext4_li_request *elr)
2765{
2766 struct ext4_group_desc *gdp = NULL;
2767 ext4_group_t group, ngroups;
2768 struct super_block *sb;
2769 unsigned long timeout = 0;
2770 int ret = 0;
2771
2772 sb = elr->lr_super;
2773 ngroups = EXT4_SB(sb)->s_groups_count;
2774
2775 for (group = elr->lr_next_group; group < ngroups; group++) {
2776 gdp = ext4_get_group_desc(sb, group, NULL);
2777 if (!gdp) {
2778 ret = 1;
2779 break;
2780 }
2781
2782 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2783 break;
2784 }
2785
2786 if (group == ngroups)
2787 ret = 1;
2788
2789 if (!ret) {
2790 timeout = jiffies;
2791 ret = ext4_init_inode_table(sb, group,
2792 elr->lr_timeout ? 0 : 1);
2793 if (elr->lr_timeout == 0) {
Lukas Czerner51ce6512011-05-20 13:55:16 -04002794 timeout = (jiffies - timeout) *
2795 elr->lr_sbi->s_li_wait_mult;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002796 elr->lr_timeout = timeout;
2797 }
2798 elr->lr_next_sched = jiffies + elr->lr_timeout;
2799 elr->lr_next_group = group + 1;
2800 }
2801
2802 return ret;
2803}
2804
2805/*
2806 * Remove lr_request from the list_request and free the
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002807 * request structure. Should be called with li_list_mtx held
Lukas Czernerbfff6872010-10-27 21:30:05 -04002808 */
2809static void ext4_remove_li_request(struct ext4_li_request *elr)
2810{
2811 struct ext4_sb_info *sbi;
2812
2813 if (!elr)
2814 return;
2815
2816 sbi = elr->lr_sbi;
2817
2818 list_del(&elr->lr_request);
2819 sbi->s_li_request = NULL;
2820 kfree(elr);
2821}
2822
2823static void ext4_unregister_li_request(struct super_block *sb)
2824{
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002825 mutex_lock(&ext4_li_mtx);
2826 if (!ext4_li_info) {
2827 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002828 return;
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002829 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002830
2831 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002832 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002833 mutex_unlock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002834 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002835}
2836
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002837static struct task_struct *ext4_lazyinit_task;
2838
Lukas Czernerbfff6872010-10-27 21:30:05 -04002839/*
2840 * This is the function where ext4lazyinit thread lives. It walks
2841 * through the request list searching for next scheduled filesystem.
2842 * When such a fs is found, run the lazy initialization request
2843 * (ext4_rn_li_request) and keep track of the time spend in this
2844 * function. Based on that time we compute next schedule time of
2845 * the request. When walking through the list is complete, compute
2846 * next waking time and put itself into sleep.
2847 */
2848static int ext4_lazyinit_thread(void *arg)
2849{
2850 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2851 struct list_head *pos, *n;
2852 struct ext4_li_request *elr;
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002853 unsigned long next_wakeup, cur;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002854
2855 BUG_ON(NULL == eli);
2856
Lukas Czernerbfff6872010-10-27 21:30:05 -04002857cont_thread:
2858 while (true) {
2859 next_wakeup = MAX_JIFFY_OFFSET;
2860
2861 mutex_lock(&eli->li_list_mtx);
2862 if (list_empty(&eli->li_request_list)) {
2863 mutex_unlock(&eli->li_list_mtx);
2864 goto exit_thread;
2865 }
2866
2867 list_for_each_safe(pos, n, &eli->li_request_list) {
2868 elr = list_entry(pos, struct ext4_li_request,
2869 lr_request);
2870
Theodore Ts'ob2c78cd2010-11-02 14:19:30 -04002871 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2872 if (ext4_run_li_request(elr) != 0) {
2873 /* error, remove the lazy_init job */
2874 ext4_remove_li_request(elr);
2875 continue;
2876 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002877 }
2878
2879 if (time_before(elr->lr_next_sched, next_wakeup))
2880 next_wakeup = elr->lr_next_sched;
2881 }
2882 mutex_unlock(&eli->li_list_mtx);
2883
Tejun Heoa0acae02011-11-21 12:32:22 -08002884 try_to_freeze();
Lukas Czernerbfff6872010-10-27 21:30:05 -04002885
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002886 cur = jiffies;
2887 if ((time_after_eq(cur, next_wakeup)) ||
Lukas Czernerf4245bd2010-11-02 14:07:17 -04002888 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04002889 cond_resched();
2890 continue;
2891 }
2892
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002893 schedule_timeout_interruptible(next_wakeup - cur);
2894
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002895 if (kthread_should_stop()) {
2896 ext4_clear_request_list();
2897 goto exit_thread;
2898 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002899 }
2900
2901exit_thread:
2902 /*
2903 * It looks like the request list is empty, but we need
2904 * to check it under the li_list_mtx lock, to prevent any
2905 * additions into it, and of course we should lock ext4_li_mtx
2906 * to atomically free the list and ext4_li_info, because at
2907 * this point another ext4 filesystem could be registering
2908 * new one.
2909 */
2910 mutex_lock(&ext4_li_mtx);
2911 mutex_lock(&eli->li_list_mtx);
2912 if (!list_empty(&eli->li_request_list)) {
2913 mutex_unlock(&eli->li_list_mtx);
2914 mutex_unlock(&ext4_li_mtx);
2915 goto cont_thread;
2916 }
2917 mutex_unlock(&eli->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002918 kfree(ext4_li_info);
2919 ext4_li_info = NULL;
2920 mutex_unlock(&ext4_li_mtx);
2921
2922 return 0;
2923}
2924
2925static void ext4_clear_request_list(void)
2926{
2927 struct list_head *pos, *n;
2928 struct ext4_li_request *elr;
2929
2930 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002931 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2932 elr = list_entry(pos, struct ext4_li_request,
2933 lr_request);
2934 ext4_remove_li_request(elr);
2935 }
2936 mutex_unlock(&ext4_li_info->li_list_mtx);
2937}
2938
2939static int ext4_run_lazyinit_thread(void)
2940{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002941 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2942 ext4_li_info, "ext4lazyinit");
2943 if (IS_ERR(ext4_lazyinit_task)) {
2944 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002945 ext4_clear_request_list();
Lukas Czernerbfff6872010-10-27 21:30:05 -04002946 kfree(ext4_li_info);
2947 ext4_li_info = NULL;
2948 printk(KERN_CRIT "EXT4: error %d creating inode table "
2949 "initialization thread\n",
2950 err);
2951 return err;
2952 }
2953 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002954 return 0;
2955}
2956
2957/*
2958 * Check whether it make sense to run itable init. thread or not.
2959 * If there is at least one uninitialized inode table, return
2960 * corresponding group number, else the loop goes through all
2961 * groups and return total number of groups.
2962 */
2963static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2964{
2965 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2966 struct ext4_group_desc *gdp = NULL;
2967
2968 for (group = 0; group < ngroups; group++) {
2969 gdp = ext4_get_group_desc(sb, group, NULL);
2970 if (!gdp)
2971 continue;
2972
2973 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2974 break;
2975 }
2976
2977 return group;
2978}
2979
2980static int ext4_li_info_new(void)
2981{
2982 struct ext4_lazy_init *eli = NULL;
2983
2984 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2985 if (!eli)
2986 return -ENOMEM;
2987
Lukas Czernerbfff6872010-10-27 21:30:05 -04002988 INIT_LIST_HEAD(&eli->li_request_list);
2989 mutex_init(&eli->li_list_mtx);
2990
Lukas Czernerbfff6872010-10-27 21:30:05 -04002991 eli->li_state |= EXT4_LAZYINIT_QUIT;
2992
2993 ext4_li_info = eli;
2994
2995 return 0;
2996}
2997
2998static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2999 ext4_group_t start)
3000{
3001 struct ext4_sb_info *sbi = EXT4_SB(sb);
3002 struct ext4_li_request *elr;
3003 unsigned long rnd;
3004
3005 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3006 if (!elr)
3007 return NULL;
3008
3009 elr->lr_super = sb;
3010 elr->lr_sbi = sbi;
3011 elr->lr_next_group = start;
3012
3013 /*
3014 * Randomize first schedule time of the request to
3015 * spread the inode table initialization requests
3016 * better.
3017 */
3018 get_random_bytes(&rnd, sizeof(rnd));
3019 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3020 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3021
3022 return elr;
3023}
3024
3025static int ext4_register_li_request(struct super_block *sb,
3026 ext4_group_t first_not_zeroed)
3027{
3028 struct ext4_sb_info *sbi = EXT4_SB(sb);
3029 struct ext4_li_request *elr;
3030 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05003031 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003032
Lukas Czerner51ce6512011-05-20 13:55:16 -04003033 if (sbi->s_li_request != NULL) {
3034 /*
3035 * Reset timeout so it can be computed again, because
3036 * s_li_wait_mult might have changed.
3037 */
3038 sbi->s_li_request->lr_timeout = 0;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003039 return 0;
Lukas Czerner51ce6512011-05-20 13:55:16 -04003040 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003041
3042 if (first_not_zeroed == ngroups ||
3043 (sb->s_flags & MS_RDONLY) ||
Tao Ma55ff3842011-05-09 10:28:41 -04003044 !test_opt(sb, INIT_INODE_TABLE))
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003045 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003046
3047 elr = ext4_li_request_new(sb, first_not_zeroed);
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003048 if (!elr)
3049 return -ENOMEM;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003050
3051 mutex_lock(&ext4_li_mtx);
3052
3053 if (NULL == ext4_li_info) {
3054 ret = ext4_li_info_new();
3055 if (ret)
3056 goto out;
3057 }
3058
3059 mutex_lock(&ext4_li_info->li_list_mtx);
3060 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3061 mutex_unlock(&ext4_li_info->li_list_mtx);
3062
3063 sbi->s_li_request = elr;
Tao Ma46e46902011-04-04 16:00:49 -04003064 /*
3065 * set elr to NULL here since it has been inserted to
3066 * the request_list and the removal and free of it is
3067 * handled by ext4_clear_request_list from now on.
3068 */
3069 elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003070
3071 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3072 ret = ext4_run_lazyinit_thread();
3073 if (ret)
3074 goto out;
3075 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003076out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003077 mutex_unlock(&ext4_li_mtx);
3078 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003079 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003080 return ret;
3081}
3082
3083/*
3084 * We do not need to lock anything since this is called on
3085 * module unload.
3086 */
3087static void ext4_destroy_lazyinit_thread(void)
3088{
3089 /*
3090 * If thread exited earlier
3091 * there's nothing to be done.
3092 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003093 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003094 return;
3095
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003096 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003097}
3098
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003099static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003100{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003101 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003102 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003103 struct ext4_super_block *es = NULL;
3104 struct ext4_sb_info *sbi;
3105 ext4_fsblk_t block;
3106 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003107 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003108 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003109 unsigned long journal_devnum = 0;
3110 unsigned long def_mount_opts;
3111 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003112 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05003113 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003114 int ret = -ENOMEM;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003115 int blocksize, clustersize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003116 unsigned int db_count;
3117 unsigned int i;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003118 int needs_recovery, has_huge_files, has_bigalloc;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003119 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07003120 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003121 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003122 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003123
3124 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3125 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003126 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003127
3128 sbi->s_blockgroup_lock =
3129 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3130 if (!sbi->s_blockgroup_lock) {
3131 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003132 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003133 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003134 sb->s_fs_info = sbi;
3135 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07003136 sbi->s_resuid = EXT4_DEF_RESUID;
3137 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003138 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003139 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003140 if (sb->s_bdev->bd_part)
3141 sbi->s_sectors_written_start =
3142 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003143
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003144 /* Cleanup superblock name */
3145 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3146 *cp = '!';
3147
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003148 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003149 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003150 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003151 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003152 goto out_fail;
3153 }
3154
3155 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003156 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003157 * block sizes. We need to calculate the offset from buffer start.
3158 */
Mingming Cao617ba132006-10-11 01:20:53 -07003159 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003160 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3161 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003162 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003163 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003164 }
3165
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003166 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003167 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003168 goto out_fail;
3169 }
3170 /*
3171 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003172 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003173 */
Mingming Cao617ba132006-10-11 01:20:53 -07003174 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003175 sbi->s_es = es;
3176 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003177 if (sb->s_magic != EXT4_SUPER_MAGIC)
3178 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003179 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003180
3181 /* Set defaults before we parse the mount options */
3182 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003183 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003184 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003185 set_opt(sb, DEBUG);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003186 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3187 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3188 "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003189 set_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003190 }
Mingming Cao617ba132006-10-11 01:20:53 -07003191 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003192 set_opt(sb, NO_UID32);
Eric Sandeenea663332011-02-23 17:51:51 -05003193 /* xattr user namespace & acls are now defaulted on */
Theodore Ts'o03010a32008-10-10 20:02:48 -04003194#ifdef CONFIG_EXT4_FS_XATTR
Eric Sandeenea663332011-02-23 17:51:51 -05003195 set_opt(sb, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003196#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04003197#ifdef CONFIG_EXT4_FS_POSIX_ACL
Eric Sandeenea663332011-02-23 17:51:51 -05003198 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003199#endif
Theodore Ts'o6fd7a462011-02-26 13:53:09 -05003200 set_opt(sb, MBLK_IO_SUBMIT);
Mingming Cao617ba132006-10-11 01:20:53 -07003201 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003202 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003203 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003204 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003205 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003206 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003207
Mingming Cao617ba132006-10-11 01:20:53 -07003208 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003209 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003210 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003211 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003212 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003213 set_opt(sb, ERRORS_RO);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003214 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003215 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003216 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003217 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003218
3219 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3220 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05003221 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3222 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3223 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003224
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003225 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003226 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003227
Mingming Cao1e2462f2007-07-18 09:00:55 -04003228 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003229 * enable delayed allocation by default
3230 * Use -o nodelalloc to turn it off
3231 */
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003232 if (!IS_EXT3_SB(sb) &&
3233 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003234 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003235
Lukas Czerner51ce6512011-05-20 13:55:16 -04003236 /*
3237 * set default s_li_wait_mult for lazyinit, for the case there is
3238 * no mount option specified.
3239 */
3240 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3241
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003242 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3243 &journal_devnum, &journal_ioprio, NULL, 0)) {
3244 ext4_msg(sb, KERN_WARNING,
3245 "failed to parse options in superblock: %s",
3246 sbi->s_es->s_mount_opts);
3247 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003248 if (!parse_options((char *) data, sb, &journal_devnum,
3249 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003250 goto failed_mount;
3251
Theodore Ts'o56889782011-09-03 18:22:38 -04003252 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3253 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3254 "with data=journal disables delayed "
3255 "allocation and O_DIRECT support!\n");
3256 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3257 ext4_msg(sb, KERN_ERR, "can't mount with "
3258 "both data=journal and delalloc");
3259 goto failed_mount;
3260 }
3261 if (test_opt(sb, DIOREAD_NOLOCK)) {
3262 ext4_msg(sb, KERN_ERR, "can't mount with "
3263 "both data=journal and delalloc");
3264 goto failed_mount;
3265 }
3266 if (test_opt(sb, DELALLOC))
3267 clear_opt(sb, DELALLOC);
3268 }
3269
3270 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3271 if (test_opt(sb, DIOREAD_NOLOCK)) {
3272 if (blocksize < PAGE_SIZE) {
3273 ext4_msg(sb, KERN_ERR, "can't mount with "
3274 "dioread_nolock if block size != PAGE_SIZE");
3275 goto failed_mount;
3276 }
3277 }
3278
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003279 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003280 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003281
Mingming Cao617ba132006-10-11 01:20:53 -07003282 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3283 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3284 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3285 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003286 ext4_msg(sb, KERN_WARNING,
3287 "feature flags set on rev 0 fs, "
3288 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003289
Theodore Ts'o2035e772011-04-18 17:29:14 -04003290 if (IS_EXT2_SB(sb)) {
3291 if (ext2_feature_set_ok(sb))
3292 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3293 "using the ext4 subsystem");
3294 else {
3295 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3296 "to feature incompatibilities");
3297 goto failed_mount;
3298 }
3299 }
3300
3301 if (IS_EXT3_SB(sb)) {
3302 if (ext3_feature_set_ok(sb))
3303 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3304 "using the ext4 subsystem");
3305 else {
3306 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3307 "to feature incompatibilities");
3308 goto failed_mount;
3309 }
3310 }
3311
Theodore Tso469108f2008-02-10 01:11:44 -05003312 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003313 * Check feature flags regardless of the revision level, since we
3314 * previously didn't change the revision level when setting the flags,
3315 * so there is a chance incompat flags are set on a rev 0 filesystem.
3316 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003317 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003318 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003319
Mingming Cao617ba132006-10-11 01:20:53 -07003320 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3321 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003322 ext4_msg(sb, KERN_ERR,
3323 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003324 goto failed_mount;
3325 }
3326
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003327 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003328 /* Validate the filesystem blocksize */
3329 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003330 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003331 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003332 goto failed_mount;
3333 }
3334
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003335 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003336 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3337 offset = do_div(logical_sb_block, blocksize);
3338 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003339 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003340 ext4_msg(sb, KERN_ERR,
3341 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003342 goto failed_mount;
3343 }
Mingming Cao617ba132006-10-11 01:20:53 -07003344 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003345 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003346 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003347 ext4_msg(sb, KERN_ERR,
3348 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003349 goto failed_mount;
3350 }
3351 }
3352
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003353 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3354 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003355 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3356 has_huge_files);
3357 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003358
Mingming Cao617ba132006-10-11 01:20:53 -07003359 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3360 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3361 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003362 } else {
3363 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3364 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003365 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003366 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003367 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003368 ext4_msg(sb, KERN_ERR,
3369 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003370 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003371 goto failed_mount;
3372 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003373 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3374 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003375 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003376
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003377 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3378 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003379 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003380 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003381 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003382 ext4_msg(sb, KERN_ERR,
3383 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003384 sbi->s_desc_size);
3385 goto failed_mount;
3386 }
3387 } else
3388 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003389
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003390 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003391 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003392 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003393 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003394
Mingming Cao617ba132006-10-11 01:20:53 -07003395 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003396 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003397 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003398 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3399 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003400 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003401 sbi->s_sbh = bh;
3402 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003403 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3404 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003405
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003406 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003407 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3408 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003409 i = le32_to_cpu(es->s_flags);
3410 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3411 sbi->s_hash_unsigned = 3;
3412 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3413#ifdef __CHAR_UNSIGNED__
3414 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3415 sbi->s_hash_unsigned = 3;
3416#else
3417 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3418#endif
3419 sb->s_dirt = 1;
3420 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003421
Theodore Ts'o281b5992011-09-09 18:34:51 -04003422 /* Handle clustersize */
3423 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3424 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3425 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3426 if (has_bigalloc) {
3427 if (clustersize < blocksize) {
3428 ext4_msg(sb, KERN_ERR,
3429 "cluster size (%d) smaller than "
3430 "block size (%d)", clustersize, blocksize);
3431 goto failed_mount;
3432 }
3433 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3434 le32_to_cpu(es->s_log_block_size);
3435 sbi->s_clusters_per_group =
3436 le32_to_cpu(es->s_clusters_per_group);
3437 if (sbi->s_clusters_per_group > blocksize * 8) {
3438 ext4_msg(sb, KERN_ERR,
3439 "#clusters per group too big: %lu",
3440 sbi->s_clusters_per_group);
3441 goto failed_mount;
3442 }
3443 if (sbi->s_blocks_per_group !=
3444 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3445 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3446 "clusters per group (%lu) inconsistent",
3447 sbi->s_blocks_per_group,
3448 sbi->s_clusters_per_group);
3449 goto failed_mount;
3450 }
3451 } else {
3452 if (clustersize != blocksize) {
3453 ext4_warning(sb, "fragment/cluster size (%d) != "
3454 "block size (%d)", clustersize,
3455 blocksize);
3456 clustersize = blocksize;
3457 }
3458 if (sbi->s_blocks_per_group > blocksize * 8) {
3459 ext4_msg(sb, KERN_ERR,
3460 "#blocks per group too big: %lu",
3461 sbi->s_blocks_per_group);
3462 goto failed_mount;
3463 }
3464 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3465 sbi->s_cluster_bits = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003466 }
Theodore Ts'o281b5992011-09-09 18:34:51 -04003467 sbi->s_cluster_ratio = clustersize / blocksize;
3468
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003469 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003470 ext4_msg(sb, KERN_ERR,
3471 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003472 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003473 goto failed_mount;
3474 }
3475
Eric Sandeenbf43d842009-08-17 23:48:51 -04003476 /*
3477 * Test whether we have more sectors than will fit in sector_t,
3478 * and whether the max offset is addressable by the page cache.
3479 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003480 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003481 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003482 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003483 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003484 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003485 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003486 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003487 ret = err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003488 goto failed_mount;
3489 }
3490
Mingming Cao617ba132006-10-11 01:20:53 -07003491 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3492 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003493
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003494 /* check blocks count against device size */
3495 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3496 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003497 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3498 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003499 ext4_blocks_count(es), blocks_count);
3500 goto failed_mount;
3501 }
3502
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003503 /*
3504 * It makes no sense for the first data block to be beyond the end
3505 * of the filesystem.
3506 */
3507 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Zheng Liu5635a622011-12-18 16:13:58 -05003508 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003509 "block %u is beyond end of filesystem (%llu)",
3510 le32_to_cpu(es->s_first_data_block),
3511 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003512 goto failed_mount;
3513 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003514 blocks_count = (ext4_blocks_count(es) -
3515 le32_to_cpu(es->s_first_data_block) +
3516 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3517 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003518 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003519 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003520 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003521 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003522 ext4_blocks_count(es),
3523 le32_to_cpu(es->s_first_data_block),
3524 EXT4_BLOCKS_PER_GROUP(sb));
3525 goto failed_mount;
3526 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003527 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003528 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3529 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003530 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3531 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'of18a5f22011-08-01 08:45:38 -04003532 sbi->s_group_desc = ext4_kvmalloc(db_count *
3533 sizeof(struct buffer_head *),
3534 GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003535 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003536 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003537 goto failed_mount;
3538 }
3539
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003540 if (ext4_proc_root)
3541 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003542
Pekka Enberg705895b2009-02-15 18:07:52 -05003543 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003544
3545 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003546 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003547 sbi->s_group_desc[i] = sb_bread(sb, block);
3548 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003549 ext4_msg(sb, KERN_ERR,
3550 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003551 db_count = i;
3552 goto failed_mount2;
3553 }
3554 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003555 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003556 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003557 goto failed_mount2;
3558 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003559 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3560 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003561 ext4_msg(sb, KERN_ERR,
3562 "unable to initialize "
3563 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003564 goto failed_mount2;
3565 }
3566
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003567 sbi->s_gdb_count = db_count;
3568 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3569 spin_lock_init(&sbi->s_next_gen_lock);
3570
Tao Ma04496412011-04-05 19:55:28 -04003571 init_timer(&sbi->s_err_report);
3572 sbi->s_err_report.function = print_daily_error_info;
3573 sbi->s_err_report.data = (unsigned long) sb;
3574
Theodore Ts'o57042652011-09-09 18:56:51 -04003575 err = percpu_counter_init(&sbi->s_freeclusters_counter,
Theodore Ts'o5dee5432011-09-09 19:10:51 -04003576 ext4_count_free_clusters(sb));
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003577 if (!err) {
3578 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3579 ext4_count_free_inodes(sb));
3580 }
3581 if (!err) {
3582 err = percpu_counter_init(&sbi->s_dirs_counter,
3583 ext4_count_dirs(sb));
3584 }
3585 if (!err) {
Theodore Ts'o57042652011-09-09 18:56:51 -04003586 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003587 }
3588 if (err) {
3589 ext4_msg(sb, KERN_ERR, "insufficient memory");
3590 goto failed_mount3;
3591 }
3592
Alex Tomasc9de5602008-01-29 00:19:52 -05003593 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04003594 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05003595
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003596 /*
3597 * set up enough so that it can read an inode
3598 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003599 if (!test_opt(sb, NOLOAD) &&
3600 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3601 sb->s_op = &ext4_sops;
3602 else
3603 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07003604 sb->s_export_op = &ext4_export_ops;
3605 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003606#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003607 sb->s_qcop = &ext4_qctl_operations;
3608 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003609#endif
Aneesh Kumar K.Vf2fa2ff2011-01-29 18:43:40 +05303610 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3611
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003612 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003613 mutex_init(&sbi->s_orphan_lock);
Yongqiang Yang8f82f842011-07-26 21:35:44 -04003614 sbi->s_resize_flags = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003615
3616 sb->s_root = NULL;
3617
3618 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07003619 EXT4_HAS_INCOMPAT_FEATURE(sb,
3620 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003621
Johann Lombardic5e06d12011-05-24 18:31:25 -04003622 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3623 !(sb->s_flags & MS_RDONLY))
3624 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3625 goto failed_mount3;
3626
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003627 /*
3628 * The first inode we look at is the journal inode. Don't try
3629 * root first: it may be modified in the journal!
3630 */
3631 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07003632 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3633 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003634 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05003635 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3636 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003637 ext4_msg(sb, KERN_ERR, "required journal recovery "
3638 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003639 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003640 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003641 clear_opt(sb, DATA_FLAGS);
Frank Mayhar03901312009-01-07 00:06:22 -05003642 sbi->s_journal = NULL;
3643 needs_recovery = 0;
3644 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003645 }
3646
Jose R. Santoseb40a092007-07-18 08:37:25 -04003647 if (ext4_blocks_count(es) > 0xffffffffULL &&
3648 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3649 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003650 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003651 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04003652 }
3653
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003654 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3655 jbd2_journal_set_features(sbi->s_journal,
3656 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05003657 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003658 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3659 jbd2_journal_set_features(sbi->s_journal,
3660 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
Girish Shilamkar818d2762008-01-28 23:58:27 -05003661 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3662 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003663 } else {
3664 jbd2_journal_clear_features(sbi->s_journal,
3665 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3666 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3667 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05003668
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003669 /* We have now updated the journal if required, so we can
3670 * validate the data journaling mode. */
3671 switch (test_opt(sb, DATA_FLAGS)) {
3672 case 0:
3673 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07003674 * capabilities: ORDERED_DATA if the journal can
3675 * cope, else JOURNAL_DATA
3676 */
Mingming Caodab291a2006-10-11 01:21:01 -07003677 if (jbd2_journal_check_available_features
3678 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003679 set_opt(sb, ORDERED_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003680 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003681 set_opt(sb, JOURNAL_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003682 break;
3683
Mingming Cao617ba132006-10-11 01:20:53 -07003684 case EXT4_MOUNT_ORDERED_DATA:
3685 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07003686 if (!jbd2_journal_check_available_features
3687 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003688 ext4_msg(sb, KERN_ERR, "Journal does not support "
3689 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003690 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003691 }
3692 default:
3693 break;
3694 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003695 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003696
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003697 /*
3698 * The journal may have updated the bg summary counts, so we
3699 * need to update the global counters.
3700 */
Theodore Ts'o57042652011-09-09 18:56:51 -04003701 percpu_counter_set(&sbi->s_freeclusters_counter,
Theodore Ts'o5dee5432011-09-09 19:10:51 -04003702 ext4_count_free_clusters(sb));
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003703 percpu_counter_set(&sbi->s_freeinodes_counter,
3704 ext4_count_free_inodes(sb));
3705 percpu_counter_set(&sbi->s_dirs_counter,
3706 ext4_count_dirs(sb));
Theodore Ts'o57042652011-09-09 18:56:51 -04003707 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04003708
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003709no_journal:
Tejun Heofd89d5f2011-02-01 11:42:42 +01003710 /*
3711 * The maximum number of concurrent works can be high and
3712 * concurrency isn't really necessary. Limit it to 1.
3713 */
3714 EXT4_SB(sb)->dio_unwritten_wq =
Linus Torvaldsae005cb2011-03-25 09:57:40 -07003715 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
Mingming Cao4c0425f2009-09-28 15:48:41 -04003716 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3717 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3718 goto failed_mount_wq;
3719 }
3720
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003721 /*
Mingming Caodab291a2006-10-11 01:21:01 -07003722 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003723 * so we can safely mount the rest of the filesystem now.
3724 */
3725
David Howells1d1fe1e2008-02-07 00:15:37 -08003726 root = ext4_iget(sb, EXT4_ROOT_INO);
3727 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003728 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003729 ret = PTR_ERR(root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05003730 root = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003731 goto failed_mount4;
3732 }
3733 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003734 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Al Viro94bf6082012-01-09 15:53:24 -05003735 iput(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003736 goto failed_mount4;
3737 }
Al Viro48fde702012-01-08 22:15:13 -05003738 sb->s_root = d_make_root(root);
David Howells1d1fe1e2008-02-07 00:15:37 -08003739 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003740 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003741 ret = -ENOMEM;
3742 goto failed_mount4;
3743 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003744
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003745 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04003746
3747 /* determine the minimum size of new large inodes, if present */
3748 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3749 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3750 EXT4_GOOD_OLD_INODE_SIZE;
3751 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3752 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3753 if (sbi->s_want_extra_isize <
3754 le16_to_cpu(es->s_want_extra_isize))
3755 sbi->s_want_extra_isize =
3756 le16_to_cpu(es->s_want_extra_isize);
3757 if (sbi->s_want_extra_isize <
3758 le16_to_cpu(es->s_min_extra_isize))
3759 sbi->s_want_extra_isize =
3760 le16_to_cpu(es->s_min_extra_isize);
3761 }
3762 }
3763 /* Check if enough inode space is available */
3764 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3765 sbi->s_inode_size) {
3766 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3767 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04003768 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3769 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04003770 }
3771
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003772 err = ext4_setup_system_zone(sb);
3773 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003774 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04003775 "zone (%d)", err);
Al Viro94bf6082012-01-09 15:53:24 -05003776 goto failed_mount4a;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003777 }
3778
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003779 ext4_ext_init(sb);
3780 err = ext4_mb_init(sb, needs_recovery);
3781 if (err) {
Uwe Kleine-König421f91d2010-06-11 12:17:00 +02003782 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
Eric Sandeenb31e1552009-06-04 17:36:36 -04003783 err);
Tao Madcf2d802011-10-06 12:10:11 -04003784 goto failed_mount5;
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003785 }
3786
Lukas Czernerbfff6872010-10-27 21:30:05 -04003787 err = ext4_register_li_request(sb, first_not_zeroed);
3788 if (err)
Tao Madcf2d802011-10-06 12:10:11 -04003789 goto failed_mount6;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003790
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003791 sbi->s_kobj.kset = ext4_kset;
3792 init_completion(&sbi->s_kobj_unregister);
3793 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3794 "%s", sb->s_id);
Tao Madcf2d802011-10-06 12:10:11 -04003795 if (err)
3796 goto failed_mount7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003797
Mingming Cao617ba132006-10-11 01:20:53 -07003798 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3799 ext4_orphan_cleanup(sb, es);
3800 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05003801 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003802 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05003803 ext4_mark_recovery_complete(sb, es);
3804 }
3805 if (EXT4_SB(sb)->s_journal) {
3806 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3807 descr = " journalled data mode";
3808 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3809 descr = " ordered data mode";
3810 else
3811 descr = " writeback data mode";
3812 } else
3813 descr = "out journal";
3814
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003815 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003816 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3817 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003818
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003819 if (es->s_error_count)
3820 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003821
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003822 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003823 return 0;
3824
Mingming Cao617ba132006-10-11 01:20:53 -07003825cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003826 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003827 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003828 goto failed_mount;
3829
Tao Madcf2d802011-10-06 12:10:11 -04003830failed_mount7:
3831 ext4_unregister_li_request(sb);
3832failed_mount6:
Tao Madcf2d802011-10-06 12:10:11 -04003833 ext4_mb_release(sb);
Al Viro94bf6082012-01-09 15:53:24 -05003834failed_mount5:
3835 ext4_ext_release(sb);
Tao Madcf2d802011-10-06 12:10:11 -04003836 ext4_release_system_zone(sb);
Al Viro94bf6082012-01-09 15:53:24 -05003837failed_mount4a:
3838 dput(sb->s_root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05003839 sb->s_root = NULL;
Al Viro94bf6082012-01-09 15:53:24 -05003840failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04003841 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04003842 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3843failed_mount_wq:
Frank Mayhar03901312009-01-07 00:06:22 -05003844 if (sbi->s_journal) {
3845 jbd2_journal_destroy(sbi->s_journal);
3846 sbi->s_journal = NULL;
3847 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003848failed_mount3:
Tao Ma04496412011-04-05 19:55:28 -04003849 del_timer(&sbi->s_err_report);
Theodore Ts'o9933fc02011-08-01 08:45:02 -04003850 if (sbi->s_flex_groups)
3851 ext4_kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -04003852 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003853 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3854 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -04003855 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Johann Lombardic5e06d12011-05-24 18:31:25 -04003856 if (sbi->s_mmp_tsk)
3857 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003858failed_mount2:
3859 for (i = 0; i < db_count; i++)
3860 brelse(sbi->s_group_desc[i]);
Theodore Ts'of18a5f22011-08-01 08:45:38 -04003861 ext4_kvfree(sbi->s_group_desc);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003862failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04003863 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003864 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003865 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003866#ifdef CONFIG_QUOTA
3867 for (i = 0; i < MAXQUOTAS; i++)
3868 kfree(sbi->s_qf_names[i]);
3869#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003870 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003871 brelse(bh);
3872out_fail:
3873 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04003874 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003875 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003876out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003877 kfree(orig_data);
David Howells1d1fe1e2008-02-07 00:15:37 -08003878 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003879}
3880
3881/*
3882 * Setup any per-fs journal parameters now. We'll do this both on
3883 * initial mount, once the journal has been initialised but before we've
3884 * done any recovery; and again on any subsequent remount.
3885 */
Mingming Cao617ba132006-10-11 01:20:53 -07003886static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003887{
Mingming Cao617ba132006-10-11 01:20:53 -07003888 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003889
Theodore Ts'o30773842009-01-03 20:27:38 -05003890 journal->j_commit_interval = sbi->s_commit_interval;
3891 journal->j_min_batch_time = sbi->s_min_batch_time;
3892 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003893
Theodore Ts'oa931da62010-08-03 21:35:12 -04003894 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003895 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07003896 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003897 else
Mingming Caodab291a2006-10-11 01:21:01 -07003898 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04003899 if (test_opt(sb, DATA_ERR_ABORT))
3900 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3901 else
3902 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04003903 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003904}
3905
Mingming Cao617ba132006-10-11 01:20:53 -07003906static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003907 unsigned int journal_inum)
3908{
3909 struct inode *journal_inode;
3910 journal_t *journal;
3911
Frank Mayhar03901312009-01-07 00:06:22 -05003912 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3913
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003914 /* First, test for the existence of a valid inode on disk. Bad
3915 * things happen if we iget() an unused inode, as the subsequent
3916 * iput() will try to delete it. */
3917
David Howells1d1fe1e2008-02-07 00:15:37 -08003918 journal_inode = ext4_iget(sb, journal_inum);
3919 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003920 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003921 return NULL;
3922 }
3923 if (!journal_inode->i_nlink) {
3924 make_bad_inode(journal_inode);
3925 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003926 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003927 return NULL;
3928 }
3929
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003930 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003931 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003932 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003933 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003934 iput(journal_inode);
3935 return NULL;
3936 }
3937
Mingming Caodab291a2006-10-11 01:21:01 -07003938 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003939 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003940 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003941 iput(journal_inode);
3942 return NULL;
3943 }
3944 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003945 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003946 return journal;
3947}
3948
Mingming Cao617ba132006-10-11 01:20:53 -07003949static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003950 dev_t j_dev)
3951{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003952 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003953 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003954 ext4_fsblk_t start;
3955 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003956 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003957 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003958 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003959 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003960 struct block_device *bdev;
3961
Frank Mayhar03901312009-01-07 00:06:22 -05003962 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3963
Eric Sandeenb31e1552009-06-04 17:36:36 -04003964 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003965 if (bdev == NULL)
3966 return NULL;
3967
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003968 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003969 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003970 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003971 ext4_msg(sb, KERN_ERR,
3972 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003973 goto out_bdev;
3974 }
3975
Mingming Cao617ba132006-10-11 01:20:53 -07003976 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3977 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003978 set_blocksize(bdev, blocksize);
3979 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003980 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3981 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003982 goto out_bdev;
3983 }
3984
Mingming Cao617ba132006-10-11 01:20:53 -07003985 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3986 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003987 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003988 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003989 ext4_msg(sb, KERN_ERR, "external journal has "
3990 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003991 brelse(bh);
3992 goto out_bdev;
3993 }
3994
Mingming Cao617ba132006-10-11 01:20:53 -07003995 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003996 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003997 brelse(bh);
3998 goto out_bdev;
3999 }
4000
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004001 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004002 start = sb_block + 1;
4003 brelse(bh); /* we're done with the superblock */
4004
Mingming Caodab291a2006-10-11 01:21:01 -07004005 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004006 start, len, blocksize);
4007 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004008 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004009 goto out_bdev;
4010 }
4011 journal->j_private = sb;
4012 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4013 wait_on_buffer(journal->j_sb_buffer);
4014 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004015 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004016 goto out_journal;
4017 }
4018 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004019 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4020 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004021 be32_to_cpu(journal->j_superblock->s_nr_users));
4022 goto out_journal;
4023 }
Mingming Cao617ba132006-10-11 01:20:53 -07004024 EXT4_SB(sb)->journal_bdev = bdev;
4025 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004026 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004027
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004028out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07004029 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004030out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07004031 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004032 return NULL;
4033}
4034
Mingming Cao617ba132006-10-11 01:20:53 -07004035static int ext4_load_journal(struct super_block *sb,
4036 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004037 unsigned long journal_devnum)
4038{
4039 journal_t *journal;
4040 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4041 dev_t journal_dev;
4042 int err = 0;
4043 int really_read_only;
4044
Frank Mayhar03901312009-01-07 00:06:22 -05004045 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4046
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004047 if (journal_devnum &&
4048 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004049 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4050 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004051 journal_dev = new_decode_dev(journal_devnum);
4052 } else
4053 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4054
4055 really_read_only = bdev_read_only(sb->s_bdev);
4056
4057 /*
4058 * Are we loading a blank journal or performing recovery after a
4059 * crash? For recovery, we need to check in advance whether we
4060 * can get read-write access to the device.
4061 */
Mingming Cao617ba132006-10-11 01:20:53 -07004062 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004063 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004064 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4065 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004066 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004067 ext4_msg(sb, KERN_ERR, "write access "
4068 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004069 return -EROFS;
4070 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04004071 ext4_msg(sb, KERN_INFO, "write access will "
4072 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004073 }
4074 }
4075
4076 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004077 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4078 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004079 return -EINVAL;
4080 }
4081
4082 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07004083 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004084 return -EINVAL;
4085 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07004086 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004087 return -EINVAL;
4088 }
4089
Theodore Ts'o90576c02009-09-29 15:51:30 -04004090 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04004091 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04004092
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004093 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07004094 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004095 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004096 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07004097 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004098 return err;
4099 }
4100 }
4101
Mingming Cao617ba132006-10-11 01:20:53 -07004102 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07004103 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004104 if (!err) {
4105 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4106 if (save)
4107 memcpy(save, ((char *) es) +
4108 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07004109 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004110 if (save)
4111 memcpy(((char *) es) + EXT4_S_ERR_START,
4112 save, EXT4_S_ERR_LEN);
4113 kfree(save);
4114 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004115
4116 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004117 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07004118 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004119 return err;
4120 }
4121
Mingming Cao617ba132006-10-11 01:20:53 -07004122 EXT4_SB(sb)->s_journal = journal;
4123 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004124
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04004125 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004126 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4127 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004128
4129 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004130 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004131 }
4132
4133 return 0;
4134}
4135
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004136static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004137{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004138 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07004139 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08004140 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004141
Theodore Ts'o7c2e7082011-09-09 18:28:51 -04004142 if (!sbh || block_device_ejected(sb))
Takashi Satoc4be0c12009-01-09 16:40:58 -08004143 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04004144 if (buffer_write_io_error(sbh)) {
4145 /*
4146 * Oh, dear. A previous attempt to write the
4147 * superblock failed. This could happen because the
4148 * USB device was yanked out. Or it could happen to
4149 * be a transient write error and maybe the block will
4150 * be remapped. Nothing we can do but to retry the
4151 * write and hope for the best.
4152 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04004153 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4154 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004155 clear_buffer_write_io_error(sbh);
4156 set_buffer_uptodate(sbh);
4157 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04004158 /*
4159 * If the file system is mounted read-only, don't update the
4160 * superblock write time. This avoids updating the superblock
4161 * write time when we are mounting the root file system
4162 * read/only but we need to replay the journal; at that point,
4163 * for people who are east of GMT and who make their clock
4164 * tick in localtime for Windows bug-for-bug compatibility,
4165 * the clock is set in the future, and this will cause e2fsck
4166 * to complain and force a full file system check.
4167 */
4168 if (!(sb->s_flags & MS_RDONLY))
4169 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004170 if (sb->s_bdev->bd_part)
4171 es->s_kbytes_written =
4172 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05004173 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4174 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004175 else
4176 es->s_kbytes_written =
4177 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'o57042652011-09-09 18:56:51 -04004178 ext4_free_blocks_count_set(es,
4179 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4180 &EXT4_SB(sb)->s_freeclusters_counter)));
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004181 es->s_free_inodes_count =
4182 cpu_to_le32(percpu_counter_sum_positive(
4183 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04004184 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004185 BUFFER_TRACE(sbh, "marking dirty");
4186 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04004187 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08004188 error = sync_dirty_buffer(sbh);
4189 if (error)
4190 return error;
4191
4192 error = buffer_write_io_error(sbh);
4193 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004194 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4195 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004196 clear_buffer_write_io_error(sbh);
4197 set_buffer_uptodate(sbh);
4198 }
4199 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08004200 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004201}
4202
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004203/*
4204 * Have we just finished recovery? If so, and if we are mounting (or
4205 * remounting) the filesystem readonly, then we will end up with a
4206 * consistent fs on disk. Record that fact.
4207 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004208static void ext4_mark_recovery_complete(struct super_block *sb,
4209 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004210{
Mingming Cao617ba132006-10-11 01:20:53 -07004211 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004212
Frank Mayhar03901312009-01-07 00:06:22 -05004213 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4214 BUG_ON(journal != NULL);
4215 return;
4216 }
Mingming Caodab291a2006-10-11 01:21:01 -07004217 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004218 if (jbd2_journal_flush(journal) < 0)
4219 goto out;
4220
Mingming Cao617ba132006-10-11 01:20:53 -07004221 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004222 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07004223 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004224 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004225 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004226
4227out:
Mingming Caodab291a2006-10-11 01:21:01 -07004228 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004229}
4230
4231/*
4232 * If we are mounting (or read-write remounting) a filesystem whose journal
4233 * has recorded an error from a previous lifetime, move that error to the
4234 * main filesystem now.
4235 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004236static void ext4_clear_journal_err(struct super_block *sb,
4237 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004238{
4239 journal_t *journal;
4240 int j_errno;
4241 const char *errstr;
4242
Frank Mayhar03901312009-01-07 00:06:22 -05004243 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4244
Mingming Cao617ba132006-10-11 01:20:53 -07004245 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004246
4247 /*
4248 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004249 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004250 */
4251
Mingming Caodab291a2006-10-11 01:21:01 -07004252 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004253 if (j_errno) {
4254 char nbuf[16];
4255
Mingming Cao617ba132006-10-11 01:20:53 -07004256 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004257 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004258 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004259 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004260
Mingming Cao617ba132006-10-11 01:20:53 -07004261 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4262 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004263 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004264
Mingming Caodab291a2006-10-11 01:21:01 -07004265 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004266 }
4267}
4268
4269/*
4270 * Force the running and committing transactions to commit,
4271 * and wait on the commit.
4272 */
Mingming Cao617ba132006-10-11 01:20:53 -07004273int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004274{
4275 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05004276 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004277
4278 if (sb->s_flags & MS_RDONLY)
4279 return 0;
4280
Mingming Cao617ba132006-10-11 01:20:53 -07004281 journal = EXT4_SB(sb)->s_journal;
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004282 if (journal) {
Eric Sandeen437f88c2010-08-01 17:33:29 -04004283 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Frank Mayhar03901312009-01-07 00:06:22 -05004284 ret = ext4_journal_force_commit(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004285 }
Frank Mayhar03901312009-01-07 00:06:22 -05004286
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004287 return ret;
4288}
4289
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004290static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004291{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004292 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004293 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004294 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004295}
4296
Mingming Cao617ba132006-10-11 01:20:53 -07004297static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004298{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004299 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004300 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004301 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004302
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004303 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004304 flush_workqueue(sbi->dio_unwritten_wq);
4305 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004306 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004307 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05004308 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004309 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004310}
4311
4312/*
4313 * LVM calls this function before a (read-only) snapshot is created. This
4314 * gives us a chance to flush the journal completely and mark the fs clean.
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -04004315 *
4316 * Note that only this function cannot bring a filesystem to be in a clean
4317 * state independently, because ext4 prevents a new handle from being started
4318 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4319 * the upper layer.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004320 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004321static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004322{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004323 int error = 0;
4324 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004325
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004326 if (sb->s_flags & MS_RDONLY)
4327 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004328
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004329 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004330
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004331 /* Now we set up the journal barrier. */
4332 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004333
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004334 /*
4335 * Don't clear the needs_recovery flag if we failed to flush
4336 * the journal.
4337 */
4338 error = jbd2_journal_flush(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004339 if (error < 0)
4340 goto out;
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004341
4342 /* Journal blocked and flushed, clear needs_recovery flag. */
4343 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4344 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004345out:
4346 /* we rely on s_frozen to stop further updates */
4347 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4348 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004349}
4350
4351/*
4352 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4353 * flag here, even though the filesystem is not technically dirty yet.
4354 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004355static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004356{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004357 if (sb->s_flags & MS_RDONLY)
4358 return 0;
4359
4360 lock_super(sb);
4361 /* Reset the needs_recovery flag before the fs is unlocked. */
4362 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4363 ext4_commit_super(sb, 1);
4364 unlock_super(sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004365 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004366}
4367
Theodore Ts'o673c6102010-12-15 20:28:48 -05004368/*
4369 * Structure to save mount options for ext4_remount's benefit
4370 */
4371struct ext4_mount_options {
4372 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004373 unsigned long s_mount_opt2;
Theodore Ts'o673c6102010-12-15 20:28:48 -05004374 uid_t s_resuid;
4375 gid_t s_resgid;
4376 unsigned long s_commit_interval;
4377 u32 s_min_batch_time, s_max_batch_time;
4378#ifdef CONFIG_QUOTA
4379 int s_jquota_fmt;
4380 char *s_qf_names[MAXQUOTAS];
4381#endif
4382};
4383
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004384static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004385{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004386 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004387 struct ext4_sb_info *sbi = EXT4_SB(sb);
4388 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004389 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004390 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004391 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004392 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004393 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004394 int err = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004395#ifdef CONFIG_QUOTA
4396 int i;
4397#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004398 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004399
4400 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04004401 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004402 old_sb_flags = sb->s_flags;
4403 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004404 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004405 old_opts.s_resuid = sbi->s_resuid;
4406 old_opts.s_resgid = sbi->s_resgid;
4407 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004408 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4409 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004410#ifdef CONFIG_QUOTA
4411 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4412 for (i = 0; i < MAXQUOTAS; i++)
4413 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4414#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004415 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4416 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004417
4418 /*
4419 * Allow the "check" option to be passed as a remount option.
4420 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004421 if (!parse_options(data, sb, NULL, &journal_ioprio,
4422 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004423 err = -EINVAL;
4424 goto restore_opts;
4425 }
4426
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004427 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004428 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004429
4430 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004431 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004432
4433 es = sbi->s_es;
4434
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004435 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004436 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004437 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4438 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004439
4440 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004441 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004442 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004443 err = -EROFS;
4444 goto restore_opts;
4445 }
4446
4447 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004448 err = dquot_suspend(sb, -1);
4449 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004450 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004451
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004452 /*
4453 * First of all, the unconditional stuff we have to do
4454 * to disable replay of the journal when we next remount
4455 */
4456 sb->s_flags |= MS_RDONLY;
4457
4458 /*
4459 * OK, test if we are remounting a valid rw partition
4460 * readonly, and if so set the rdonly flag and then
4461 * mark the partition as valid again.
4462 */
Mingming Cao617ba132006-10-11 01:20:53 -07004463 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4464 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004465 es->s_state = cpu_to_le16(sbi->s_mount_state);
4466
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004467 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004468 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004469 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004470 /* Make sure we can mount this feature set readwrite */
4471 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004472 err = -EROFS;
4473 goto restore_opts;
4474 }
Eric Sandeenead65962007-02-10 01:46:08 -08004475 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004476 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004477 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004478 */
4479 for (g = 0; g < sbi->s_groups_count; g++) {
4480 struct ext4_group_desc *gdp =
4481 ext4_get_group_desc(sb, g, NULL);
4482
4483 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004484 ext4_msg(sb, KERN_ERR,
4485 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04004486 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4487 le16_to_cpu(gdp->bg_checksum));
4488 err = -EINVAL;
4489 goto restore_opts;
4490 }
4491 }
4492
4493 /*
Eric Sandeenead65962007-02-10 01:46:08 -08004494 * If we have an unprocessed orphan list hanging
4495 * around from a previously readonly bdev mount,
4496 * require a full umount/remount for now.
4497 */
4498 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004499 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08004500 "remount RDWR because of unprocessed "
4501 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004502 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08004503 err = -EINVAL;
4504 goto restore_opts;
4505 }
4506
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004507 /*
4508 * Mounting a RDONLY partition read-write, so reread
4509 * and store the current valid flag. (It may have
4510 * been changed by e2fsck since we originally mounted
4511 * the partition.)
4512 */
Frank Mayhar03901312009-01-07 00:06:22 -05004513 if (sbi->s_journal)
4514 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004515 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07004516 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004517 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004518 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004519 sb->s_flags &= ~MS_RDONLY;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004520 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4521 EXT4_FEATURE_INCOMPAT_MMP))
4522 if (ext4_multi_mount_protect(sb,
4523 le64_to_cpu(es->s_mmp_block))) {
4524 err = -EROFS;
4525 goto restore_opts;
4526 }
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004527 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004528 }
4529 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04004530
4531 /*
4532 * Reinitialize lazy itable initialization thread based on
4533 * current settings
4534 */
4535 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4536 ext4_unregister_li_request(sb);
4537 else {
4538 ext4_group_t first_not_zeroed;
4539 first_not_zeroed = ext4_has_uninit_itable(sb);
4540 ext4_register_li_request(sb, first_not_zeroed);
4541 }
4542
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004543 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05004544 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004545 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05004546
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004547#ifdef CONFIG_QUOTA
4548 /* Release old quota file names */
4549 for (i = 0; i < MAXQUOTAS; i++)
4550 if (old_opts.s_qf_names[i] &&
4551 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4552 kfree(old_opts.s_qf_names[i]);
4553#endif
Al Virobbd68512009-05-06 10:43:07 -04004554 unlock_super(sb);
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004555 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004556 dquot_resume(sb, -1);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004557
4558 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4559 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004560 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004561
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004562restore_opts:
4563 sb->s_flags = old_sb_flags;
4564 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004565 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004566 sbi->s_resuid = old_opts.s_resuid;
4567 sbi->s_resgid = old_opts.s_resgid;
4568 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004569 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4570 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004571#ifdef CONFIG_QUOTA
4572 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4573 for (i = 0; i < MAXQUOTAS; i++) {
4574 if (sbi->s_qf_names[i] &&
4575 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4576 kfree(sbi->s_qf_names[i]);
4577 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4578 }
4579#endif
Al Virobbd68512009-05-06 10:43:07 -04004580 unlock_super(sb);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004581 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004582 return err;
4583}
4584
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004585/*
4586 * Note: calculating the overhead so we can be compatible with
4587 * historical BSD practice is quite difficult in the face of
4588 * clusters/bigalloc. This is because multiple metadata blocks from
4589 * different block group can end up in the same allocation cluster.
4590 * Calculating the exact overhead in the face of clustered allocation
4591 * requires either O(all block bitmaps) in memory or O(number of block
4592 * groups**2) in time. We will still calculate the superblock for
4593 * older file systems --- and if we come across with a bigalloc file
4594 * system with zero in s_overhead_clusters the estimate will be close to
4595 * correct especially for very large cluster sizes --- but for newer
4596 * file systems, it's better to calculate this figure once at mkfs
4597 * time, and store it in the superblock. If the superblock value is
4598 * present (even for non-bigalloc file systems), we will use it.
4599 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004600static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004601{
4602 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004603 struct ext4_sb_info *sbi = EXT4_SB(sb);
4604 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004605 struct ext4_group_desc *gdp;
Pekka Enberg960cc392006-12-06 20:35:29 -08004606 u64 fsid;
Kazuya Miod02a9392011-05-24 18:30:07 -04004607 s64 bfree;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004608
Badari Pulavarty5e700302007-07-15 23:42:00 -07004609 if (test_opt(sb, MINIX_DF)) {
4610 sbi->s_overhead_last = 0;
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004611 } else if (es->s_overhead_clusters) {
4612 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004613 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04004614 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07004615 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004616
4617 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07004618 * Compute the overhead (FS structures). This is constant
4619 * for a given filesystem unless the number of block groups
4620 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004621 */
4622
4623 /*
4624 * All of the blocks before first_data_block are
4625 * overhead
4626 */
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004627 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004628
4629 /*
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004630 * Add the overhead found in each block group
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004631 */
4632 for (i = 0; i < ngroups; i++) {
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004633 gdp = ext4_get_group_desc(sb, i, NULL);
4634 overhead += ext4_num_overhead_clusters(sb, i, gdp);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004635 cond_resched();
4636 }
Badari Pulavarty5e700302007-07-15 23:42:00 -07004637 sbi->s_overhead_last = overhead;
4638 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004639 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004640 }
4641
Mingming Cao617ba132006-10-11 01:20:53 -07004642 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004643 buf->f_bsize = sb->s_blocksize;
Theodore Ts'of975d6b2011-09-09 19:00:51 -04004644 buf->f_blocks = (ext4_blocks_count(es) -
4645 EXT4_C2B(sbi, sbi->s_overhead_last));
Theodore Ts'o57042652011-09-09 18:56:51 -04004646 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4647 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
Kazuya Miod02a9392011-05-24 18:30:07 -04004648 /* prevent underflow in case that few free space is available */
Theodore Ts'o57042652011-09-09 18:56:51 -04004649 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004650 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4651 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004652 buf->f_bavail = 0;
4653 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07004654 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07004655 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08004656 fsid = le64_to_cpup((void *)es->s_uuid) ^
4657 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4658 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4659 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004660
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004661 return 0;
4662}
4663
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004664/* Helper function for writing quotas on sync - we need to start transaction
4665 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004666 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07004667 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01004668 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05004669 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07004670 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004671 *
4672 */
4673
4674#ifdef CONFIG_QUOTA
4675
4676static inline struct inode *dquot_to_inode(struct dquot *dquot)
4677{
4678 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4679}
4680
Mingming Cao617ba132006-10-11 01:20:53 -07004681static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004682{
4683 int ret, err;
4684 handle_t *handle;
4685 struct inode *inode;
4686
4687 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004688 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004689 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004690 if (IS_ERR(handle))
4691 return PTR_ERR(handle);
4692 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004693 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004694 if (!ret)
4695 ret = err;
4696 return ret;
4697}
4698
Mingming Cao617ba132006-10-11 01:20:53 -07004699static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004700{
4701 int ret, err;
4702 handle_t *handle;
4703
Mingming Cao617ba132006-10-11 01:20:53 -07004704 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004705 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004706 if (IS_ERR(handle))
4707 return PTR_ERR(handle);
4708 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004709 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004710 if (!ret)
4711 ret = err;
4712 return ret;
4713}
4714
Mingming Cao617ba132006-10-11 01:20:53 -07004715static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004716{
4717 int ret, err;
4718 handle_t *handle;
4719
Mingming Cao617ba132006-10-11 01:20:53 -07004720 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004721 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07004722 if (IS_ERR(handle)) {
4723 /* Release dquot anyway to avoid endless cycle in dqput() */
4724 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004725 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07004726 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004727 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004728 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004729 if (!ret)
4730 ret = err;
4731 return ret;
4732}
4733
Mingming Cao617ba132006-10-11 01:20:53 -07004734static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004735{
Jan Kara2c8be6b2008-05-13 21:27:55 -04004736 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07004737 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4738 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004739 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004740 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004741 } else {
4742 return dquot_mark_dquot_dirty(dquot);
4743 }
4744}
4745
Mingming Cao617ba132006-10-11 01:20:53 -07004746static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004747{
4748 int ret, err;
4749 handle_t *handle;
4750
4751 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07004752 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004753 if (IS_ERR(handle))
4754 return PTR_ERR(handle);
4755 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07004756 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004757 if (!ret)
4758 ret = err;
4759 return ret;
4760}
4761
4762/*
4763 * Turn on quotas during mount time - we need to find
4764 * the quota file and such...
4765 */
Mingming Cao617ba132006-10-11 01:20:53 -07004766static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004767{
Christoph Hellwig287a8092010-05-19 07:16:45 -04004768 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4769 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004770}
4771
4772/*
4773 * Standard function to be called on quota_on
4774 */
Mingming Cao617ba132006-10-11 01:20:53 -07004775static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02004776 struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004777{
4778 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004779
4780 if (!test_opt(sb, QUOTA))
4781 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04004782
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004783 /* Quotafile not on the same filesystem? */
Al Virod8c95842011-12-07 18:16:57 -05004784 if (path->dentry->d_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004785 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04004786 /* Journaling quota? */
4787 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004788 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02004789 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004790 ext4_msg(sb, KERN_WARNING,
4791 "Quota file not on filesystem root. "
4792 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004793 }
Jan Kara06235432008-05-13 19:11:51 -04004794
4795 /*
4796 * When we journal data on quota file, we have to flush journal to see
4797 * all updates to the file when we bypass pagecache...
4798 */
Frank Mayhar03901312009-01-07 00:06:22 -05004799 if (EXT4_SB(sb)->s_journal &&
Jan Karaf00c9e42010-09-15 17:38:58 +02004800 ext4_should_journal_data(path->dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04004801 /*
4802 * We don't need to lock updates but journal_flush() could
4803 * otherwise be livelocked...
4804 */
4805 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004806 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04004807 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02004808 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004809 return err;
Jan Kara06235432008-05-13 19:11:51 -04004810 }
4811
Jan Karaf00c9e42010-09-15 17:38:58 +02004812 return dquot_quota_on(sb, type, format_id, path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004813}
4814
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004815static int ext4_quota_off(struct super_block *sb, int type)
4816{
Jan Kara21f97692011-04-04 15:33:39 -04004817 struct inode *inode = sb_dqopt(sb)->files[type];
4818 handle_t *handle;
4819
Dmitry Monakhov87009d82010-11-08 13:47:33 -05004820 /* Force all delayed allocation blocks to be allocated.
4821 * Caller already holds s_umount sem */
4822 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004823 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004824
Amir Goldstein0b268592011-05-16 09:59:13 -04004825 if (!inode)
4826 goto out;
4827
Jan Kara21f97692011-04-04 15:33:39 -04004828 /* Update modification times of quota files when userspace can
4829 * start looking at them */
4830 handle = ext4_journal_start(inode, 1);
4831 if (IS_ERR(handle))
4832 goto out;
4833 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4834 ext4_mark_inode_dirty(handle, inode);
4835 ext4_journal_stop(handle);
4836
4837out:
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004838 return dquot_quota_off(sb, type);
4839}
4840
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004841/* Read data from quotafile - avoid pagecache and such because we cannot afford
4842 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03004843 * itself serializes the operations (and no one else should touch the files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004844 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07004845static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004846 size_t len, loff_t off)
4847{
4848 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004849 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004850 int err = 0;
4851 int offset = off & (sb->s_blocksize - 1);
4852 int tocopy;
4853 size_t toread;
4854 struct buffer_head *bh;
4855 loff_t i_size = i_size_read(inode);
4856
4857 if (off > i_size)
4858 return 0;
4859 if (off+len > i_size)
4860 len = i_size-off;
4861 toread = len;
4862 while (toread > 0) {
4863 tocopy = sb->s_blocksize - offset < toread ?
4864 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07004865 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004866 if (err)
4867 return err;
4868 if (!bh) /* A hole? */
4869 memset(data, 0, tocopy);
4870 else
4871 memcpy(data, bh->b_data+offset, tocopy);
4872 brelse(bh);
4873 offset = 0;
4874 toread -= tocopy;
4875 data += tocopy;
4876 blk++;
4877 }
4878 return len;
4879}
4880
4881/* Write to quotafile (we know the transaction is already started and has
4882 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07004883static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004884 const char *data, size_t len, loff_t off)
4885{
4886 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004887 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004888 int err = 0;
4889 int offset = off & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004890 struct buffer_head *bh;
4891 handle_t *handle = journal_current_handle();
4892
Frank Mayhar03901312009-01-07 00:06:22 -05004893 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004894 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4895 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07004896 (unsigned long long)off, (unsigned long long)len);
4897 return -EIO;
4898 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004899 /*
4900 * Since we account only one data block in transaction credits,
4901 * then it is impossible to cross a block boundary.
4902 */
4903 if (sb->s_blocksize - offset < len) {
4904 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4905 " cancelled because not block aligned",
4906 (unsigned long long)off, (unsigned long long)len);
4907 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004908 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004909
4910 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4911 bh = ext4_bread(handle, inode, blk, 1, &err);
4912 if (!bh)
4913 goto out;
Jan Kara62d2b5f2010-07-27 11:56:07 -04004914 err = ext4_journal_get_write_access(handle, bh);
4915 if (err) {
4916 brelse(bh);
4917 goto out;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004918 }
4919 lock_buffer(bh);
4920 memcpy(bh->b_data+offset, data, len);
4921 flush_dcache_page(bh->b_page);
4922 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04004923 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004924 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004925out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004926 if (err) {
Jan Kara4d04e4f2008-07-04 09:59:34 -07004927 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004928 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07004929 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004930 if (inode->i_size < off + len) {
4931 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07004932 EXT4_I(inode)->i_disksize = inode->i_size;
Jan Kara21f97692011-04-04 15:33:39 -04004933 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004934 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004935 mutex_unlock(&inode->i_mutex);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004936 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004937}
4938
4939#endif
4940
Al Viro152a0832010-07-25 00:46:55 +04004941static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4942 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004943{
Al Viro152a0832010-07-25 00:46:55 +04004944 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004945}
4946
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004947#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004948static inline void register_as_ext2(void)
4949{
4950 int err = register_filesystem(&ext2_fs_type);
4951 if (err)
4952 printk(KERN_WARNING
4953 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4954}
4955
4956static inline void unregister_as_ext2(void)
4957{
4958 unregister_filesystem(&ext2_fs_type);
4959}
Theodore Ts'o2035e772011-04-18 17:29:14 -04004960
4961static inline int ext2_feature_set_ok(struct super_block *sb)
4962{
4963 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4964 return 0;
4965 if (sb->s_flags & MS_RDONLY)
4966 return 1;
4967 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4968 return 0;
4969 return 1;
4970}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004971MODULE_ALIAS("ext2");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004972#else
4973static inline void register_as_ext2(void) { }
4974static inline void unregister_as_ext2(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04004975static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05004976#endif
4977
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004978#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004979static inline void register_as_ext3(void)
4980{
4981 int err = register_filesystem(&ext3_fs_type);
4982 if (err)
4983 printk(KERN_WARNING
4984 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4985}
4986
4987static inline void unregister_as_ext3(void)
4988{
4989 unregister_filesystem(&ext3_fs_type);
4990}
Theodore Ts'o2035e772011-04-18 17:29:14 -04004991
4992static inline int ext3_feature_set_ok(struct super_block *sb)
4993{
4994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4995 return 0;
4996 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4997 return 0;
4998 if (sb->s_flags & MS_RDONLY)
4999 return 1;
5000 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5001 return 0;
5002 return 1;
5003}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05005004MODULE_ALIAS("ext3");
Theodore Ts'o24b58422009-12-07 14:08:51 -05005005#else
5006static inline void register_as_ext3(void) { }
5007static inline void unregister_as_ext3(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04005008static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05005009#endif
5010
Theodore Ts'o03010a32008-10-10 20:02:48 -04005011static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005012 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04005013 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04005014 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005015 .kill_sb = kill_block_super,
5016 .fs_flags = FS_REQUIRES_DEV,
5017};
5018
Lukas Czerner8f021222011-02-03 14:33:33 -05005019static int __init ext4_init_feat_adverts(void)
Lukas Czerner857ac882010-10-27 21:30:05 -04005020{
5021 struct ext4_features *ef;
5022 int ret = -ENOMEM;
5023
5024 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5025 if (!ef)
5026 goto out;
5027
5028 ef->f_kobj.kset = ext4_kset;
5029 init_completion(&ef->f_kobj_unregister);
5030 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5031 "features");
5032 if (ret) {
5033 kfree(ef);
5034 goto out;
5035 }
5036
5037 ext4_feat = ef;
5038 ret = 0;
5039out:
5040 return ret;
5041}
5042
Lukas Czerner8f021222011-02-03 14:33:33 -05005043static void ext4_exit_feat_adverts(void)
5044{
5045 kobject_put(&ext4_feat->f_kobj);
5046 wait_for_completion(&ext4_feat->f_kobj_unregister);
5047 kfree(ext4_feat);
5048}
5049
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005050/* Shared across all ext4 file systems */
5051wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5052struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5053
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005054static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005055{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005056 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05005057
Al Viro07c0c5d82012-03-20 22:05:02 -04005058 ext4_li_info = NULL;
5059 mutex_init(&ext4_li_mtx);
5060
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04005061 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005062
5063 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5064 mutex_init(&ext4__aio_mutex[i]);
5065 init_waitqueue_head(&ext4__ioend_wq[i]);
5066 }
5067
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005068 err = ext4_init_pageio();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005069 if (err)
5070 return err;
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005071 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04005072 if (err)
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005073 goto out6;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04005074 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5075 if (!ext4_kset)
Theodore Ts'odd683142011-02-03 14:33:49 -05005076 goto out5;
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005077 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Lukas Czerner857ac882010-10-27 21:30:05 -04005078
5079 err = ext4_init_feat_adverts();
Theodore Ts'odd683142011-02-03 14:33:49 -05005080 if (err)
5081 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04005082
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005083 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005084 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005085 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05005086
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005087 err = ext4_init_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05005088 if (err)
5089 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005090 err = init_inodecache();
5091 if (err)
5092 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05005093 register_as_ext3();
Theodore Ts'o2035e772011-04-18 17:29:14 -04005094 register_as_ext2();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005095 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005096 if (err)
5097 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04005098
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005099 return 0;
5100out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05005101 unregister_as_ext2();
5102 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005103 destroy_inodecache();
5104out1:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005105 ext4_exit_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05005106out2:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005107 ext4_exit_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005108out3:
Lukas Czerner8f021222011-02-03 14:33:33 -05005109 ext4_exit_feat_adverts();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005110out4:
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005111 if (ext4_proc_root)
5112 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'odd683142011-02-03 14:33:49 -05005113 kset_unregister(ext4_kset);
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005114out5:
Theodore Ts'odd683142011-02-03 14:33:49 -05005115 ext4_exit_system_zone();
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005116out6:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005117 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005118 return err;
5119}
5120
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005121static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005122{
Lukas Czernerbfff6872010-10-27 21:30:05 -04005123 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05005124 unregister_as_ext2();
5125 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005126 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005127 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005128 ext4_exit_xattr();
5129 ext4_exit_mballoc();
Lukas Czerner8f021222011-02-03 14:33:33 -05005130 ext4_exit_feat_adverts();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04005131 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04005132 kset_unregister(ext4_kset);
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005133 ext4_exit_system_zone();
5134 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005135}
5136
5137MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05005138MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005139MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005140module_init(ext4_init_fs)
5141module_exit(ext4_exit_fs)