blob: 33c398785e53c4a1c8be9435920178e6c6d4aa9b [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040023#include <linux/vmalloc.h>
Mingming Caodab291a2006-10-11 01:21:01 -070024#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070025#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070030#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040037#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040038#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040039#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040040#include <linux/crc16.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070041#include <asm/uaccess.h>
42
Lukas Czernerbfff6872010-10-27 21:30:05 -040043#include <linux/kthread.h>
44#include <linux/freezer.h>
45
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040046#include "ext4.h"
47#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070048#include "xattr.h"
49#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040050#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070051
Theodore Ts'o9bffad12009-06-17 11:48:11 -040052#define CREATE_TRACE_POINTS
53#include <trace/events/ext4.h>
54
Theodore Ts'o1f109d52010-10-27 21:30:14 -040055static struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040056static struct kset *ext4_kset;
Lukas Czerner0b75a842011-02-23 12:22:49 -050057static struct ext4_lazy_init *ext4_li_info;
58static struct mutex ext4_li_mtx;
59static struct ext4_features *ext4_feat;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040060
Mingming Cao617ba132006-10-11 01:20:53 -070061static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070062 unsigned long journal_devnum);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040063static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040064static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070068static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070070 char nbuf[16]);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040071static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080073static int ext4_unfreeze(struct super_block *sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040074static void ext4_write_super(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080075static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040076static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
Lukas Czernerbfff6872010-10-27 21:30:05 -040078static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050080static void ext4_clear_request_list(void);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070081
Jan Karaba69f9a2010-03-24 20:18:37 -040082#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
83static struct file_system_type ext3_fs_type = {
84 .owner = THIS_MODULE,
85 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +040086 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -040087 .kill_sb = kill_block_super,
88 .fs_flags = FS_REQUIRES_DEV,
89};
90#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
91#else
92#define IS_EXT3_SB(sb) (0)
93#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070094
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070095ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -070097{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -040098 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -070099 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400100 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700101}
102
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700103ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
104 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700105{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400106 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700109}
110
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700111ext4_fsblk_t ext4_inode_table(struct super_block *sb,
112 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700113{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400114 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400116 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700117}
118
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500119__u32 ext4_free_blks_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400124 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500125}
126
127__u32 ext4_free_inodes_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400132 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500133}
134
135__u32 ext4_used_dirs_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400140 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500141}
142
143__u32 ext4_itable_unused_count(struct super_block *sb,
144 struct ext4_group_desc *bg)
145{
146 return le16_to_cpu(bg->bg_itable_unused_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400148 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500149}
150
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700151void ext4_block_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700153{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400154 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700157}
158
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700159void ext4_inode_bitmap_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700161{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400162 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700165}
166
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700167void ext4_inode_table_set(struct super_block *sb,
168 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700169{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400170 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700173}
174
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500175void ext4_free_blks_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_free_inodes_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_used_dirs_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
197}
198
199void ext4_itable_unused_set(struct super_block *sb,
200 struct ext4_group_desc *bg, __u32 count)
201{
202 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
205}
206
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400207
208/* Just increment the non-pointer handle value */
209static handle_t *ext4_get_nojournal(void)
210{
211 handle_t *handle = current->journal_info;
212 unsigned long ref_cnt = (unsigned long)handle;
213
214 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
215
216 ref_cnt++;
217 handle = (handle_t *)ref_cnt;
218
219 current->journal_info = handle;
220 return handle;
221}
222
223
224/* Decrement the non-pointer handle value */
225static void ext4_put_nojournal(handle_t *handle)
226{
227 unsigned long ref_cnt = (unsigned long)handle;
228
229 BUG_ON(ref_cnt == 0);
230
231 ref_cnt--;
232 handle = (handle_t *)ref_cnt;
233
234 current->journal_info = handle;
235}
236
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700237/*
Mingming Caodab291a2006-10-11 01:21:01 -0700238 * Wrappers for jbd2_journal_start/end.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700239 *
240 * The only special thing we need to do here is to make sure that all
241 * journal_end calls result in the superblock being marked dirty, so
242 * that sync() will call the filesystem's write_super callback if
243 * appropriate.
244 */
Mingming Cao617ba132006-10-11 01:20:53 -0700245handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700246{
247 journal_t *journal;
248
249 if (sb->s_flags & MS_RDONLY)
250 return ERR_PTR(-EROFS);
251
Eric Sandeen437f88c2010-08-01 17:33:29 -0400252 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700253 /* Special case here: if the journal has aborted behind our
254 * backs (eg. EIO in the commit thread), then we still need to
255 * take the FS itself readonly cleanly. */
Mingming Cao617ba132006-10-11 01:20:53 -0700256 journal = EXT4_SB(sb)->s_journal;
Frank Mayhar03901312009-01-07 00:06:22 -0500257 if (journal) {
258 if (is_journal_aborted(journal)) {
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400259 ext4_abort(sb, "Detected aborted journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500260 return ERR_PTR(-EROFS);
261 }
262 return jbd2_journal_start(journal, nblocks);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700263 }
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400264 return ext4_get_nojournal();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700265}
266
267/*
268 * The only special thing we need to do here is to make sure that all
Mingming Caodab291a2006-10-11 01:21:01 -0700269 * jbd2_journal_stop calls result in the superblock being marked dirty, so
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700270 * that sync() will call the filesystem's write_super callback if
271 * appropriate.
272 */
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400273int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700274{
275 struct super_block *sb;
276 int err;
277 int rc;
278
Frank Mayhar03901312009-01-07 00:06:22 -0500279 if (!ext4_handle_valid(handle)) {
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400280 ext4_put_nojournal(handle);
Frank Mayhar03901312009-01-07 00:06:22 -0500281 return 0;
282 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700283 sb = handle->h_transaction->t_journal->j_private;
284 err = handle->h_err;
Mingming Caodab291a2006-10-11 01:21:01 -0700285 rc = jbd2_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700286
287 if (!err)
288 err = rc;
289 if (err)
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400290 __ext4_std_error(sb, where, line, err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700291 return err;
292}
293
Theodore Ts'o90c72012010-06-29 14:53:24 -0400294void ext4_journal_abort_handle(const char *caller, unsigned int line,
295 const char *err_fn, struct buffer_head *bh,
296 handle_t *handle, int err)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700297{
298 char nbuf[16];
Mingming Cao617ba132006-10-11 01:20:53 -0700299 const char *errstr = ext4_decode_error(NULL, err, nbuf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700300
Frank Mayhar03901312009-01-07 00:06:22 -0500301 BUG_ON(!ext4_handle_valid(handle));
302
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700303 if (bh)
304 BUFFER_TRACE(bh, "abort");
305
306 if (!handle->h_err)
307 handle->h_err = err;
308
309 if (is_handle_aborted(handle))
310 return;
311
Theodore Ts'o90c72012010-06-29 14:53:24 -0400312 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
313 caller, line, errstr, err_fn);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700314
Mingming Caodab291a2006-10-11 01:21:01 -0700315 jbd2_journal_abort_handle(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700316}
317
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400318static void __save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322
323 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
335 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400336 /*
337 * Start the daily error reporting function if it hasn't been
338 * started already
339 */
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400342 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
343}
344
345static void save_error_info(struct super_block *sb, const char *func,
346 unsigned int line)
347{
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
350}
351
352
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700353/* Deal with the reporting of failure conditions on a filesystem such as
354 * inconsistencies detected or read IO failures.
355 *
356 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700357 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700358 * write ordering constraints on the superblock which prevent us from
359 * writing it out straight away; and given that the journal is about to
360 * be aborted, we can't rely on the current, or future, transactions to
361 * write out the superblock safely.
362 *
Mingming Caodab291a2006-10-11 01:21:01 -0700363 * We'll just use the jbd2_journal_abort() error code to record an error in
Thadeu Lima de Souza Cascardod6b198b2010-01-17 19:10:07 -0200364 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700365 * that error until we've noted it down and cleared it.
366 */
367
Mingming Cao617ba132006-10-11 01:20:53 -0700368static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700369{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700370 if (sb->s_flags & MS_RDONLY)
371 return;
372
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400373 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700374 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700375
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400376 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700377 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700378 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700379 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400380 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400381 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700382 sb->s_flags |= MS_RDONLY;
383 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700384 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700385 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700386 sb->s_id);
387}
388
Eric Sandeen12062dd2010-02-15 14:19:27 -0500389void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400390 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700391{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500392 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700393 va_list args;
394
395 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500396 vaf.fmt = fmt;
397 vaf.va = &args;
398 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
399 sb->s_id, function, line, current->comm, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700400 va_end(args);
401
Mingming Cao617ba132006-10-11 01:20:53 -0700402 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700403}
404
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400405void ext4_error_inode(struct inode *inode, const char *function,
406 unsigned int line, ext4_fsblk_t block,
Frank Mayhar273df552010-03-02 11:46:09 -0500407 const char *fmt, ...)
408{
409 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500410 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400411 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500412
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400413 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
414 es->s_last_error_block = cpu_to_le64(block);
415 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500416 va_start(args, fmt);
Theodore Ts'of7c21172011-01-10 12:10:55 -0500417 vaf.fmt = fmt;
418 vaf.va = &args;
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400419 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
420 inode->i_sb->s_id, function, line, inode->i_ino);
421 if (block)
Theodore Ts'of7c21172011-01-10 12:10:55 -0500422 printk(KERN_CONT "block %llu: ", block);
423 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
Frank Mayhar273df552010-03-02 11:46:09 -0500424 va_end(args);
425
426 ext4_handle_error(inode->i_sb);
427}
428
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400429void ext4_error_file(struct file *file, const char *function,
Theodore Ts'of7c21172011-01-10 12:10:55 -0500430 unsigned int line, ext4_fsblk_t block,
431 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500432{
433 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500434 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400435 struct ext4_super_block *es;
Frank Mayhar273df552010-03-02 11:46:09 -0500436 struct inode *inode = file->f_dentry->d_inode;
437 char pathname[80], *path;
438
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400439 es = EXT4_SB(inode->i_sb)->s_es;
440 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
441 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500442 path = d_path(&(file->f_path), pathname, sizeof(pathname));
Dan Carpenterf9a62d02011-01-10 12:10:50 -0500443 if (IS_ERR(path))
Frank Mayhar273df552010-03-02 11:46:09 -0500444 path = "(unknown)";
445 printk(KERN_CRIT
Theodore Ts'of7c21172011-01-10 12:10:55 -0500446 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
447 inode->i_sb->s_id, function, line, inode->i_ino);
448 if (block)
449 printk(KERN_CONT "block %llu: ", block);
450 va_start(args, fmt);
451 vaf.fmt = fmt;
452 vaf.va = &args;
453 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
Frank Mayhar273df552010-03-02 11:46:09 -0500454 va_end(args);
455
456 ext4_handle_error(inode->i_sb);
457}
458
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400459static const char *ext4_decode_error(struct super_block *sb, int errno,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700460 char nbuf[16])
461{
462 char *errstr = NULL;
463
464 switch (errno) {
465 case -EIO:
466 errstr = "IO failure";
467 break;
468 case -ENOMEM:
469 errstr = "Out of memory";
470 break;
471 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400472 if (!sb || (EXT4_SB(sb)->s_journal &&
473 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700474 errstr = "Journal has aborted";
475 else
476 errstr = "Readonly filesystem";
477 break;
478 default:
479 /* If the caller passed in an extra buffer for unknown
480 * errors, textualise them now. Else we just return
481 * NULL. */
482 if (nbuf) {
483 /* Check for truncated error codes... */
484 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
485 errstr = nbuf;
486 }
487 break;
488 }
489
490 return errstr;
491}
492
Mingming Cao617ba132006-10-11 01:20:53 -0700493/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700494 * automatically and invokes the appropriate error response. */
495
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400496void __ext4_std_error(struct super_block *sb, const char *function,
497 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700498{
499 char nbuf[16];
500 const char *errstr;
501
502 /* Special case: if the error is EROFS, and we're not already
503 * inside a transaction, then there's really no point in logging
504 * an error. */
505 if (errno == -EROFS && journal_current_handle() == NULL &&
506 (sb->s_flags & MS_RDONLY))
507 return;
508
Mingming Cao617ba132006-10-11 01:20:53 -0700509 errstr = ext4_decode_error(sb, errno, nbuf);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400510 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
511 sb->s_id, function, line, errstr);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400512 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700513
Mingming Cao617ba132006-10-11 01:20:53 -0700514 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700515}
516
517/*
Mingming Cao617ba132006-10-11 01:20:53 -0700518 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700519 * abort function may be used to deal with unrecoverable failures such
520 * as journal IO errors or ENOMEM at a critical moment in log management.
521 *
522 * We unconditionally force the filesystem into an ABORT|READONLY state,
523 * unless the error response on the fs has been set to panic in which
524 * case we take the easy way out and panic immediately.
525 */
526
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400527void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400528 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700529{
530 va_list args;
531
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400532 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700533 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
535 function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700536 vprintk(fmt, args);
537 printk("\n");
538 va_end(args);
539
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400540 if ((sb->s_flags & MS_RDONLY) == 0) {
541 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
542 sb->s_flags |= MS_RDONLY;
543 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
544 if (EXT4_SB(sb)->s_journal)
545 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
546 save_error_info(sb, function, line);
547 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700548 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700549 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700550}
551
Joe Perches0ff2ea72010-12-19 22:43:19 -0500552void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400553{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500554 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400555 va_list args;
556
557 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500558 vaf.fmt = fmt;
559 vaf.va = &args;
560 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400561 va_end(args);
562}
563
Eric Sandeen12062dd2010-02-15 14:19:27 -0500564void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400565 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700566{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500567 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700568 va_list args;
569
570 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500571 vaf.fmt = fmt;
572 vaf.va = &args;
573 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
574 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700575 va_end(args);
576}
577
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400578void __ext4_grp_locked_error(const char *function, unsigned int line,
579 struct super_block *sb, ext4_group_t grp,
580 unsigned long ino, ext4_fsblk_t block,
581 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500582__releases(bitlock)
583__acquires(bitlock)
584{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500585 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500586 va_list args;
587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
588
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400589 es->s_last_error_ino = cpu_to_le32(ino);
590 es->s_last_error_block = cpu_to_le64(block);
591 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500592
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500593 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500594
595 vaf.fmt = fmt;
596 vaf.va = &args;
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400597 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
598 sb->s_id, function, line, grp);
599 if (ino)
Joe Perches0ff2ea72010-12-19 22:43:19 -0500600 printk(KERN_CONT "inode %lu: ", ino);
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400601 if (block)
Joe Perches0ff2ea72010-12-19 22:43:19 -0500602 printk(KERN_CONT "block %llu:", (unsigned long long) block);
603 printk(KERN_CONT "%pV\n", &vaf);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500604 va_end(args);
605
606 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400607 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500608 return;
609 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400610
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500611 ext4_unlock_group(sb, grp);
612 ext4_handle_error(sb);
613 /*
614 * We only get here in the ERRORS_RO case; relocking the group
615 * may be dangerous, but nothing bad will happen since the
616 * filesystem will have already been marked read/only and the
617 * journal has been aborted. We return 1 as a hint to callers
618 * who might what to use the return value from
619 * ext4_grp_locked_error() to distinguish beween the
620 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
621 * aggressively from the ext4 function in question, with a
622 * more appropriate error code.
623 */
624 ext4_lock_group(sb, grp);
625 return;
626}
627
Mingming Cao617ba132006-10-11 01:20:53 -0700628void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700629{
Mingming Cao617ba132006-10-11 01:20:53 -0700630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700631
Mingming Cao617ba132006-10-11 01:20:53 -0700632 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700633 return;
634
Eric Sandeen12062dd2010-02-15 14:19:27 -0500635 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700636 "updating to rev %d because of new feature flag, "
637 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700638 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700639
Mingming Cao617ba132006-10-11 01:20:53 -0700640 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
641 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
642 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700643 /* leave es->s_feature_*compat flags alone */
644 /* es->s_uuid will be set by e2fsck if empty */
645
646 /*
647 * The rest of the superblock fields should be zero, and if not it
648 * means they are likely already in use, so leave them alone. We
649 * can leave it up to e2fsck to clean up any inconsistencies there.
650 */
651}
652
653/*
654 * Open the external journal device
655 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400656static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700657{
658 struct block_device *bdev;
659 char b[BDEVNAME_SIZE];
660
Tejun Heod4d77622010-11-13 11:55:18 +0100661 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700662 if (IS_ERR(bdev))
663 goto fail;
664 return bdev;
665
666fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400667 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700668 __bdevname(dev, b), PTR_ERR(bdev));
669 return NULL;
670}
671
672/*
673 * Release the journal device
674 */
Mingming Cao617ba132006-10-11 01:20:53 -0700675static int ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700676{
Tejun Heoe525fd82010-11-13 11:55:17 +0100677 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700678}
679
Mingming Cao617ba132006-10-11 01:20:53 -0700680static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700681{
682 struct block_device *bdev;
683 int ret = -ENODEV;
684
685 bdev = sbi->journal_bdev;
686 if (bdev) {
Mingming Cao617ba132006-10-11 01:20:53 -0700687 ret = ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700688 sbi->journal_bdev = NULL;
689 }
690 return ret;
691}
692
693static inline struct inode *orphan_list_entry(struct list_head *l)
694{
Mingming Cao617ba132006-10-11 01:20:53 -0700695 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700696}
697
Mingming Cao617ba132006-10-11 01:20:53 -0700698static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700699{
700 struct list_head *l;
701
Eric Sandeenb31e1552009-06-04 17:36:36 -0400702 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
703 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700704
705 printk(KERN_ERR "sb_info orphan list:\n");
706 list_for_each(l, &sbi->s_orphan) {
707 struct inode *inode = orphan_list_entry(l);
708 printk(KERN_ERR " "
709 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
710 inode->i_sb->s_id, inode->i_ino, inode,
711 inode->i_mode, inode->i_nlink,
712 NEXT_ORPHAN(inode));
713 }
714}
715
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400716static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700717{
Mingming Cao617ba132006-10-11 01:20:53 -0700718 struct ext4_sb_info *sbi = EXT4_SB(sb);
719 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400720 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700721
Lukas Czerner857ac882010-10-27 21:30:05 -0400722 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400723 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
724
Mingming Cao4c0425f2009-09-28 15:48:41 -0400725 flush_workqueue(sbi->dio_unwritten_wq);
726 destroy_workqueue(sbi->dio_unwritten_wq);
727
Al Viroa9e220f2009-05-05 22:10:44 -0400728 lock_super(sb);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200729 if (sb->s_dirt)
Christoph Hellwigebc1ac12009-05-11 23:35:03 +0200730 ext4_commit_super(sb, 1);
Christoph Hellwig8c85e122009-04-28 18:00:26 +0200731
Frank Mayhar03901312009-01-07 00:06:22 -0500732 if (sbi->s_journal) {
733 err = jbd2_journal_destroy(sbi->s_journal);
734 sbi->s_journal = NULL;
735 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400736 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500737 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500738
Sergey Senozhatskya1c6c562010-10-27 21:30:04 -0400739 del_timer(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500740 ext4_release_system_zone(sb);
741 ext4_mb_release(sb);
742 ext4_ext_release(sb);
743 ext4_xattr_put_super(sb);
744
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700745 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700746 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700747 es->s_state = cpu_to_le16(sbi->s_mount_state);
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400748 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700749 }
Theodore Ts'o240799c2008-10-09 23:53:47 -0400750 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400751 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400752 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400753 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700754
755 for (i = 0; i < sbi->s_gdb_count; i++)
756 brelse(sbi->s_group_desc[i]);
757 kfree(sbi->s_group_desc);
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -0400758 if (is_vmalloc_addr(sbi->s_flex_groups))
759 vfree(sbi->s_flex_groups);
760 else
761 kfree(sbi->s_flex_groups);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700762 percpu_counter_destroy(&sbi->s_freeblocks_counter);
763 percpu_counter_destroy(&sbi->s_freeinodes_counter);
764 percpu_counter_destroy(&sbi->s_dirs_counter);
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -0400765 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700766 brelse(sbi->s_sbh);
767#ifdef CONFIG_QUOTA
768 for (i = 0; i < MAXQUOTAS; i++)
769 kfree(sbi->s_qf_names[i]);
770#endif
771
772 /* Debugging code just in case the in-memory inode orphan list
773 * isn't empty. The on-disk one can be non-empty if we've
774 * detected an error and taken the fs readonly, but the
775 * in-memory list had better be clean by this point. */
776 if (!list_empty(&sbi->s_orphan))
777 dump_orphan_list(sb, sbi);
778 J_ASSERT(list_empty(&sbi->s_orphan));
779
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700780 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700781 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
782 /*
783 * Invalidate the journal device's buffers. We don't want them
784 * floating about in memory - the physical journal device may
785 * hotswapped, and it breaks the `ro-after' testing code.
786 */
787 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700788 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700789 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700790 }
791 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400792 /*
793 * Now that we are completely done shutting down the
794 * superblock, we need to actually destroy the kobject.
795 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400796 unlock_super(sb);
797 kobject_put(&sbi->s_kobj);
798 wait_for_completion(&sbi->s_kobj_unregister);
Pekka Enberg705895b2009-02-15 18:07:52 -0500799 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700800 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700801}
802
Christoph Lametere18b8902006-12-06 20:33:20 -0800803static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700804
805/*
806 * Called inside transaction, so use GFP_NOFS
807 */
Mingming Cao617ba132006-10-11 01:20:53 -0700808static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700809{
Mingming Cao617ba132006-10-11 01:20:53 -0700810 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700811
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800812 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700813 if (!ei)
814 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400815
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700816 ei->vfs_inode.i_version = 1;
Aneesh Kumar K.V91246c02008-08-19 21:14:52 -0400817 ei->vfs_inode.i_data.writeback_index = 0;
Alex Tomasa86c6182006-10-11 01:21:03 -0700818 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
Alex Tomasc9de5602008-01-29 00:19:52 -0500819 INIT_LIST_HEAD(&ei->i_prealloc_list);
820 spin_lock_init(&ei->i_prealloc_lock);
Mingming Caod2a17632008-07-14 17:52:37 -0400821 ei->i_reserved_data_blocks = 0;
822 ei->i_reserved_meta_blocks = 0;
823 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500824 ei->i_da_metadata_calc_len = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400825 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300826#ifdef CONFIG_QUOTA
827 ei->i_reserved_quota = 0;
828#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500829 ei->jinode = NULL;
Jiaying Zhangc7064ef2010-03-02 13:28:44 -0500830 INIT_LIST_HEAD(&ei->i_completed_io_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500831 spin_lock_init(&ei->i_completed_io_lock);
Mingming Cao8d5d02e2009-09-28 15:48:29 -0400832 ei->cur_aio_dio = NULL;
Jan Karab436b9b2009-12-08 23:51:10 -0500833 ei->i_sync_tid = 0;
834 ei->i_datasync_tid = 0;
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500835 atomic_set(&ei->i_ioend_count, 0);
Eric Sandeene9e3bce2011-02-12 08:17:34 -0500836 atomic_set(&ei->i_aiodio_unwritten, 0);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400837
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700838 return &ei->vfs_inode;
839}
840
Theodore Ts'o7ff9c072010-11-08 13:51:33 -0500841static int ext4_drop_inode(struct inode *inode)
842{
843 int drop = generic_drop_inode(inode);
844
845 trace_ext4_drop_inode(inode, drop);
846 return drop;
847}
848
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100849static void ext4_i_callback(struct rcu_head *head)
850{
851 struct inode *inode = container_of(head, struct inode, i_rcu);
852 INIT_LIST_HEAD(&inode->i_dentry);
853 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
854}
855
Mingming Cao617ba132006-10-11 01:20:53 -0700856static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700857{
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500858 ext4_ioend_wait(inode);
Vasily Averin9f7dd932007-07-15 23:40:45 -0700859 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400860 ext4_msg(inode->i_sb, KERN_ERR,
861 "Inode %lu (%p): orphan list check failed!",
862 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700863 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
864 EXT4_I(inode), sizeof(struct ext4_inode_info),
865 true);
866 dump_stack();
867 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100868 call_rcu(&inode->i_rcu, ext4_i_callback);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700869}
870
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700871static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700872{
Mingming Cao617ba132006-10-11 01:20:53 -0700873 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700874
Christoph Lametera35afb82007-05-16 22:10:57 -0700875 INIT_LIST_HEAD(&ei->i_orphan);
Theodore Ts'o03010a32008-10-10 20:02:48 -0400876#ifdef CONFIG_EXT4_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700877 init_rwsem(&ei->xattr_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700878#endif
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500879 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700880 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700881}
882
883static int init_inodecache(void)
884{
Mingming Cao617ba132006-10-11 01:20:53 -0700885 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
886 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700887 0, (SLAB_RECLAIM_ACCOUNT|
888 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900889 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700890 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700891 return -ENOMEM;
892 return 0;
893}
894
895static void destroy_inodecache(void)
896{
Mingming Cao617ba132006-10-11 01:20:53 -0700897 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700898}
899
Al Viro0930fcc2010-06-07 13:16:22 -0400900void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700901{
Al Viro0930fcc2010-06-07 13:16:22 -0400902 invalidate_inode_buffers(inode);
903 end_writeback(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -0500904 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400905 ext4_discard_preallocations(inode);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500906 if (EXT4_I(inode)->jinode) {
907 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
908 EXT4_I(inode)->jinode);
909 jbd2_free_inode(EXT4_I(inode)->jinode);
910 EXT4_I(inode)->jinode = NULL;
911 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700912}
913
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400914static inline void ext4_show_quota_options(struct seq_file *seq,
915 struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700916{
917#if defined(CONFIG_QUOTA)
Mingming Cao617ba132006-10-11 01:20:53 -0700918 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700919
Jan Kara5a20bdf2009-11-30 23:58:32 +0100920 if (sbi->s_jquota_fmt) {
921 char *fmtname = "";
922
923 switch (sbi->s_jquota_fmt) {
924 case QFMT_VFS_OLD:
925 fmtname = "vfsold";
926 break;
927 case QFMT_VFS_V0:
928 fmtname = "vfsv0";
929 break;
930 case QFMT_VFS_V1:
931 fmtname = "vfsv1";
932 break;
933 }
934 seq_printf(seq, ",jqfmt=%s", fmtname);
935 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700936
937 if (sbi->s_qf_names[USRQUOTA])
938 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
939
940 if (sbi->s_qf_names[GRPQUOTA])
941 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
942
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500943 if (test_opt(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700944 seq_puts(seq, ",usrquota");
945
Dmitry Monakhov482a7422010-02-24 11:35:32 -0500946 if (test_opt(sb, GRPQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700947 seq_puts(seq, ",grpquota");
948#endif
949}
950
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700951/*
952 * Show an option if
953 * - it's set to a non-default value OR
954 * - if the per-sb default is different from the global default
955 */
Mingming Cao617ba132006-10-11 01:20:53 -0700956static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700957{
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500958 int def_errors;
959 unsigned long def_mount_opts;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700960 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700961 struct ext4_sb_info *sbi = EXT4_SB(sb);
962 struct ext4_super_block *es = sbi->s_es;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700963
964 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500965 def_errors = le16_to_cpu(es->s_errors);
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700966
967 if (sbi->s_sb_block != 1)
968 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
969 if (test_opt(sb, MINIX_DF))
970 seq_puts(seq, ",minixdf");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500971 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700972 seq_puts(seq, ",grpid");
973 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
974 seq_puts(seq, ",nogrpid");
975 if (sbi->s_resuid != EXT4_DEF_RESUID ||
976 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
977 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
978 }
979 if (sbi->s_resgid != EXT4_DEF_RESGID ||
980 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
981 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
982 }
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500983 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700984 if (def_errors == EXT4_ERRORS_PANIC ||
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500985 def_errors == EXT4_ERRORS_CONTINUE) {
986 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700987 }
988 }
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500989 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -0500990 seq_puts(seq, ",errors=continue");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500991 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700992 seq_puts(seq, ",errors=panic");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500993 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700994 seq_puts(seq, ",nouid32");
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -0500995 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -0700996 seq_puts(seq, ",debug");
997 if (test_opt(sb, OLDALLOC))
998 seq_puts(seq, ",oldalloc");
Theodore Ts'o03010a32008-10-10 20:02:48 -0400999#ifdef CONFIG_EXT4_FS_XATTR
Eric Sandeenea663332011-02-23 17:51:51 -05001000 if (test_opt(sb, XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001001 seq_puts(seq, ",user_xattr");
Eric Sandeenea663332011-02-23 17:51:51 -05001002 if (!test_opt(sb, XATTR_USER))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001003 seq_puts(seq, ",nouser_xattr");
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001004#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001005#ifdef CONFIG_EXT4_FS_POSIX_ACL
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001006 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001007 seq_puts(seq, ",acl");
1008 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1009 seq_puts(seq, ",noacl");
1010#endif
Theodore Ts'o30773842009-01-03 20:27:38 -05001011 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07001012 seq_printf(seq, ",commit=%u",
1013 (unsigned) (sbi->s_commit_interval / HZ));
1014 }
Theodore Ts'o30773842009-01-03 20:27:38 -05001015 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1016 seq_printf(seq, ",min_batch_time=%u",
1017 (unsigned) sbi->s_min_batch_time);
1018 }
1019 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1020 seq_printf(seq, ",max_batch_time=%u",
1021 (unsigned) sbi->s_min_batch_time);
1022 }
1023
Eric Sandeen571640c2008-05-26 12:29:46 -04001024 /*
1025 * We're changing the default of barrier mount option, so
1026 * let's always display its mount state so it's clear what its
1027 * status is.
1028 */
1029 seq_puts(seq, ",barrier=");
1030 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
Theodore Ts'ocd0b6a32008-05-26 10:28:28 -04001031 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1032 seq_puts(seq, ",journal_async_commit");
Jan Kara39a4bad2010-05-16 17:00:00 -04001033 else if (test_opt(sb, JOURNAL_CHECKSUM))
1034 seq_puts(seq, ",journal_checksum");
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001035 if (test_opt(sb, I_VERSION))
1036 seq_puts(seq, ",i_version");
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001037 if (!test_opt(sb, DELALLOC) &&
1038 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001039 seq_puts(seq, ",nodelalloc");
1040
Theodore Ts'o6fd7a462011-02-26 13:53:09 -05001041 if (!test_opt(sb, MBLK_IO_SUBMIT))
1042 seq_puts(seq, ",nomblk_io_submit");
Miklos Szeredicb45bbe2008-01-28 23:58:27 -05001043 if (sbi->s_stripe)
1044 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
Aneesh Kumar K.Vaa22df22008-01-28 23:58:27 -05001045 /*
1046 * journal mode get enabled in different ways
1047 * So just print the value even if we didn't specify it
1048 */
Mingming Cao617ba132006-10-11 01:20:53 -07001049 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001050 seq_puts(seq, ",data=journal");
Mingming Cao617ba132006-10-11 01:20:53 -07001051 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001052 seq_puts(seq, ",data=ordered");
Mingming Cao617ba132006-10-11 01:20:53 -07001053 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001054 seq_puts(seq, ",data=writeback");
1055
Theodore Ts'o240799c2008-10-09 23:53:47 -04001056 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1057 seq_printf(seq, ",inode_readahead_blks=%u",
1058 sbi->s_inode_readahead_blks);
1059
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001060 if (test_opt(sb, DATA_ERR_ABORT))
1061 seq_puts(seq, ",data_err=abort");
1062
Theodore Ts'oafd46722009-03-16 23:12:23 -04001063 if (test_opt(sb, NO_AUTO_DA_ALLOC))
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001064 seq_puts(seq, ",noauto_da_alloc");
Theodore Ts'oafd46722009-03-16 23:12:23 -04001065
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001066 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
Eric Sandeen5328e632009-11-19 14:25:42 -05001067 seq_puts(seq, ",discard");
1068
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001069 if (test_opt(sb, NOLOAD))
1070 seq_puts(seq, ",norecovery");
1071
Jiaying Zhang744692d2010-03-04 16:14:02 -05001072 if (test_opt(sb, DIOREAD_NOLOCK))
1073 seq_puts(seq, ",dioread_nolock");
1074
Theodore Ts'o8b67f042010-08-01 23:14:20 -04001075 if (test_opt(sb, BLOCK_VALIDITY) &&
1076 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1077 seq_puts(seq, ",block_validity");
1078
Lukas Czernerbfff6872010-10-27 21:30:05 -04001079 if (!test_opt(sb, INIT_INODE_TABLE))
1080 seq_puts(seq, ",noinit_inode_table");
1081 else if (sbi->s_li_wait_mult)
1082 seq_printf(seq, ",init_inode_table=%u",
1083 (unsigned) sbi->s_li_wait_mult);
1084
Mingming Cao617ba132006-10-11 01:20:53 -07001085 ext4_show_quota_options(seq, sb);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001086
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001087 return 0;
1088}
1089
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001090static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001091 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001092{
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001093 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001094
Mingming Cao617ba132006-10-11 01:20:53 -07001095 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001096 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -07001097 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001098 return ERR_PTR(-ESTALE);
1099
1100 /* iget isn't really right if the inode is currently unallocated!!
1101 *
Mingming Cao617ba132006-10-11 01:20:53 -07001102 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001103 * deleted, so we should be safe.
1104 *
1105 * Currently we don't know the generation for parent directory, so
1106 * a generation of 0 means "accept any"
1107 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001108 inode = ext4_iget(sb, ino);
1109 if (IS_ERR(inode))
1110 return ERR_CAST(inode);
1111 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001112 iput(inode);
1113 return ERR_PTR(-ESTALE);
1114 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001115
1116 return inode;
1117}
1118
1119static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001120 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001121{
1122 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1123 ext4_nfs_get_inode);
1124}
1125
1126static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001127 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001128{
1129 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1130 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001131}
1132
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001133/*
1134 * Try to release metadata pages (indirect blocks, directories) which are
1135 * mapped via the block device. Since these pages could have journal heads
1136 * which would prevent try_to_free_buffers() from freeing them, we must use
1137 * jbd2 layer's try_to_free_buffers() function to release them.
1138 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001139static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1140 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001141{
1142 journal_t *journal = EXT4_SB(sb)->s_journal;
1143
1144 WARN_ON(PageChecked(page));
1145 if (!page_has_buffers(page))
1146 return 0;
1147 if (journal)
1148 return jbd2_journal_try_to_free_buffers(journal, page,
1149 wait & ~__GFP_WAIT);
1150 return try_to_free_buffers(page);
1151}
1152
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001153#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001154#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001155#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001156
Mingming Cao617ba132006-10-11 01:20:53 -07001157static int ext4_write_dquot(struct dquot *dquot);
1158static int ext4_acquire_dquot(struct dquot *dquot);
1159static int ext4_release_dquot(struct dquot *dquot);
1160static int ext4_mark_dquot_dirty(struct dquot *dquot);
1161static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001162static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02001163 struct path *path);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001164static int ext4_quota_off(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001165static int ext4_quota_on_mount(struct super_block *sb, int type);
1166static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001167 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001168static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001169 const char *data, size_t len, loff_t off);
1170
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001171static const struct dquot_operations ext4_quota_operations = {
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001172#ifdef CONFIG_QUOTA
Mingming Cao60e58e02009-01-22 18:13:05 +01001173 .get_reserved_space = ext4_get_reserved_space,
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001174#endif
Mingming Cao617ba132006-10-11 01:20:53 -07001175 .write_dquot = ext4_write_dquot,
1176 .acquire_dquot = ext4_acquire_dquot,
1177 .release_dquot = ext4_release_dquot,
1178 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001179 .write_info = ext4_write_info,
1180 .alloc_dquot = dquot_alloc,
1181 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001182};
1183
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001184static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001185 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001186 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001187 .quota_sync = dquot_quota_sync,
1188 .get_info = dquot_get_dqinfo,
1189 .set_info = dquot_set_dqinfo,
1190 .get_dqblk = dquot_get_dqblk,
1191 .set_dqblk = dquot_set_dqblk
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001192};
1193#endif
1194
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001195static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001196 .alloc_inode = ext4_alloc_inode,
1197 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001198 .write_inode = ext4_write_inode,
1199 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001200 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001201 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001202 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001203 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001204 .freeze_fs = ext4_freeze,
1205 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001206 .statfs = ext4_statfs,
1207 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001208 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001209#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001210 .quota_read = ext4_quota_read,
1211 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001212#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001213 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001214};
1215
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001216static const struct super_operations ext4_nojournal_sops = {
1217 .alloc_inode = ext4_alloc_inode,
1218 .destroy_inode = ext4_destroy_inode,
1219 .write_inode = ext4_write_inode,
1220 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001221 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001222 .evict_inode = ext4_evict_inode,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001223 .write_super = ext4_write_super,
1224 .put_super = ext4_put_super,
1225 .statfs = ext4_statfs,
1226 .remount_fs = ext4_remount,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001227 .show_options = ext4_show_options,
1228#ifdef CONFIG_QUOTA
1229 .quota_read = ext4_quota_read,
1230 .quota_write = ext4_quota_write,
1231#endif
1232 .bdev_try_to_free_page = bdev_try_to_free_page,
1233};
1234
Christoph Hellwig39655162007-10-21 16:42:17 -07001235static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001236 .fh_to_dentry = ext4_fh_to_dentry,
1237 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001238 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001239};
1240
1241enum {
1242 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1243 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o01436ef2008-10-17 07:22:35 -04001244 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001245 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001246 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
Theodore Ts'o30773842009-01-03 20:27:38 -05001247 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001248 Opt_journal_update, Opt_journal_dev,
Girish Shilamkar818d2762008-01-28 23:58:27 -05001249 Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001250 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001251 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001252 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001253 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1254 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1255 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o14490322010-12-14 15:27:50 -05001256 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1257 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001258 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001259 Opt_dioread_nolock, Opt_dioread_lock,
Eric Sandeen5328e632009-11-19 14:25:42 -05001260 Opt_discard, Opt_nodiscard,
Lukas Czernerbfff6872010-10-27 21:30:05 -04001261 Opt_init_inode_table, Opt_noinit_inode_table,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001262};
1263
Steven Whitehousea447c092008-10-13 10:46:57 +01001264static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001265 {Opt_bsd_df, "bsddf"},
1266 {Opt_minix_df, "minixdf"},
1267 {Opt_grpid, "grpid"},
1268 {Opt_grpid, "bsdgroups"},
1269 {Opt_nogrpid, "nogrpid"},
1270 {Opt_nogrpid, "sysvgroups"},
1271 {Opt_resgid, "resgid=%u"},
1272 {Opt_resuid, "resuid=%u"},
1273 {Opt_sb, "sb=%u"},
1274 {Opt_err_cont, "errors=continue"},
1275 {Opt_err_panic, "errors=panic"},
1276 {Opt_err_ro, "errors=remount-ro"},
1277 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001278 {Opt_debug, "debug"},
1279 {Opt_oldalloc, "oldalloc"},
1280 {Opt_orlov, "orlov"},
1281 {Opt_user_xattr, "user_xattr"},
1282 {Opt_nouser_xattr, "nouser_xattr"},
1283 {Opt_acl, "acl"},
1284 {Opt_noacl, "noacl"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001285 {Opt_noload, "noload"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001286 {Opt_noload, "norecovery"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001287 {Opt_nobh, "nobh"},
1288 {Opt_bh, "bh"},
1289 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001290 {Opt_min_batch_time, "min_batch_time=%u"},
1291 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001292 {Opt_journal_update, "journal=update"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001293 {Opt_journal_dev, "journal_dev=%u"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001294 {Opt_journal_checksum, "journal_checksum"},
1295 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001296 {Opt_abort, "abort"},
1297 {Opt_data_journal, "data=journal"},
1298 {Opt_data_ordered, "data=ordered"},
1299 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001300 {Opt_data_err_abort, "data_err=abort"},
1301 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001302 {Opt_offusrjquota, "usrjquota="},
1303 {Opt_usrjquota, "usrjquota=%s"},
1304 {Opt_offgrpjquota, "grpjquota="},
1305 {Opt_grpjquota, "grpjquota=%s"},
1306 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1307 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001308 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001309 {Opt_grpquota, "grpquota"},
1310 {Opt_noquota, "noquota"},
1311 {Opt_quota, "quota"},
1312 {Opt_usrquota, "usrquota"},
1313 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001314 {Opt_barrier, "barrier"},
1315 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001316 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001317 {Opt_stripe, "stripe=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001318 {Opt_resize, "resize"},
Alex Tomas64769242008-07-11 19:27:31 -04001319 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001320 {Opt_nodelalloc, "nodelalloc"},
Theodore Ts'o14490322010-12-14 15:27:50 -05001321 {Opt_mblk_io_submit, "mblk_io_submit"},
1322 {Opt_nomblk_io_submit, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001323 {Opt_block_validity, "block_validity"},
1324 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001325 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001326 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001327 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001328 {Opt_auto_da_alloc, "auto_da_alloc"},
1329 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001330 {Opt_dioread_nolock, "dioread_nolock"},
1331 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001332 {Opt_discard, "discard"},
1333 {Opt_nodiscard, "nodiscard"},
Lukas Czernerbfff6872010-10-27 21:30:05 -04001334 {Opt_init_inode_table, "init_itable=%u"},
1335 {Opt_init_inode_table, "init_itable"},
1336 {Opt_noinit_inode_table, "noinit_itable"},
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001337 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001338};
1339
Mingming Cao617ba132006-10-11 01:20:53 -07001340static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001341{
Mingming Cao617ba132006-10-11 01:20:53 -07001342 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001343 char *options = (char *) *data;
1344
1345 if (!options || strncmp(options, "sb=", 3) != 0)
1346 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001347
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001348 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001349 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001350 sb_block = simple_strtoul(options, &options, 0);
1351 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001352 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001353 (char *) *data);
1354 return 1;
1355 }
1356 if (*options == ',')
1357 options++;
1358 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001359
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001360 return sb_block;
1361}
1362
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001363#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001364static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1365 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001366
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001367#ifdef CONFIG_QUOTA
1368static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1369{
1370 struct ext4_sb_info *sbi = EXT4_SB(sb);
1371 char *qname;
1372
1373 if (sb_any_quota_loaded(sb) &&
1374 !sbi->s_qf_names[qtype]) {
1375 ext4_msg(sb, KERN_ERR,
1376 "Cannot change journaled "
1377 "quota options when quota turned on");
1378 return 0;
1379 }
1380 qname = match_strdup(args);
1381 if (!qname) {
1382 ext4_msg(sb, KERN_ERR,
1383 "Not enough memory for storing quotafile name");
1384 return 0;
1385 }
1386 if (sbi->s_qf_names[qtype] &&
1387 strcmp(sbi->s_qf_names[qtype], qname)) {
1388 ext4_msg(sb, KERN_ERR,
1389 "%s quota file already specified", QTYPE2NAME(qtype));
1390 kfree(qname);
1391 return 0;
1392 }
1393 sbi->s_qf_names[qtype] = qname;
1394 if (strchr(sbi->s_qf_names[qtype], '/')) {
1395 ext4_msg(sb, KERN_ERR,
1396 "quotafile must be on filesystem root");
1397 kfree(sbi->s_qf_names[qtype]);
1398 sbi->s_qf_names[qtype] = NULL;
1399 return 0;
1400 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001401 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001402 return 1;
1403}
1404
1405static int clear_qf_name(struct super_block *sb, int qtype)
1406{
1407
1408 struct ext4_sb_info *sbi = EXT4_SB(sb);
1409
1410 if (sb_any_quota_loaded(sb) &&
1411 sbi->s_qf_names[qtype]) {
1412 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1413 " when quota turned on");
1414 return 0;
1415 }
1416 /*
1417 * The space will be released later when all options are confirmed
1418 * to be correct
1419 */
1420 sbi->s_qf_names[qtype] = NULL;
1421 return 1;
1422}
1423#endif
1424
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001425static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001426 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001427 unsigned int *journal_ioprio,
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001428 ext4_fsblk_t *n_blocks_count, int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001429{
Mingming Cao617ba132006-10-11 01:20:53 -07001430 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001431 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001432 substring_t args[MAX_OPT_ARGS];
1433 int data_opt = 0;
1434 int option;
1435#ifdef CONFIG_QUOTA
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001436 int qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001437#endif
1438
1439 if (!options)
1440 return 1;
1441
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001442 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001443 int token;
1444 if (!*p)
1445 continue;
1446
Eric Sandeen15121c12010-02-15 20:17:55 -05001447 /*
1448 * Initialize args struct so we know whether arg was
1449 * found; some options take optional arguments.
1450 */
Peter Huewe7dc57612011-02-21 21:01:42 -05001451 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001452 token = match_token(p, tokens, args);
1453 switch (token) {
1454 case Opt_bsd_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001455 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001456 clear_opt(sb, MINIX_DF);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001457 break;
1458 case Opt_minix_df:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001459 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001460 set_opt(sb, MINIX_DF);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001461
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001462 break;
1463 case Opt_grpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001464 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001465 set_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001466
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001467 break;
1468 case Opt_nogrpid:
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001469 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001470 clear_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001471
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001472 break;
1473 case Opt_resuid:
1474 if (match_int(&args[0], &option))
1475 return 0;
1476 sbi->s_resuid = option;
1477 break;
1478 case Opt_resgid:
1479 if (match_int(&args[0], &option))
1480 return 0;
1481 sbi->s_resgid = option;
1482 break;
1483 case Opt_sb:
1484 /* handled by get_sb_block() instead of here */
1485 /* *sb_block = match_int(&args[0]); */
1486 break;
1487 case Opt_err_panic:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001488 clear_opt(sb, ERRORS_CONT);
1489 clear_opt(sb, ERRORS_RO);
1490 set_opt(sb, ERRORS_PANIC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001491 break;
1492 case Opt_err_ro:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001493 clear_opt(sb, ERRORS_CONT);
1494 clear_opt(sb, ERRORS_PANIC);
1495 set_opt(sb, ERRORS_RO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001496 break;
1497 case Opt_err_cont:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001498 clear_opt(sb, ERRORS_RO);
1499 clear_opt(sb, ERRORS_PANIC);
1500 set_opt(sb, ERRORS_CONT);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001501 break;
1502 case Opt_nouid32:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001503 set_opt(sb, NO_UID32);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001504 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001505 case Opt_debug:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001506 set_opt(sb, DEBUG);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001507 break;
1508 case Opt_oldalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001509 set_opt(sb, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001510 break;
1511 case Opt_orlov:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001512 clear_opt(sb, OLDALLOC);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001513 break;
Theodore Ts'o03010a32008-10-10 20:02:48 -04001514#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001515 case Opt_user_xattr:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001516 set_opt(sb, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001517 break;
1518 case Opt_nouser_xattr:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001519 clear_opt(sb, XATTR_USER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001520 break;
1521#else
1522 case Opt_user_xattr:
1523 case Opt_nouser_xattr:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001524 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001525 break;
1526#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04001527#ifdef CONFIG_EXT4_FS_POSIX_ACL
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001528 case Opt_acl:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001529 set_opt(sb, POSIX_ACL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001530 break;
1531 case Opt_noacl:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001532 clear_opt(sb, POSIX_ACL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001533 break;
1534#else
1535 case Opt_acl:
1536 case Opt_noacl:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001537 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001538 break;
1539#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001540 case Opt_journal_update:
1541 /* @@@ FIXME */
1542 /* Eventually we will want to be able to create
1543 a journal file here. For now, only allow the
1544 user to specify an existing inode to be the
1545 journal file. */
1546 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001547 ext4_msg(sb, KERN_ERR,
1548 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001549 return 0;
1550 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001551 set_opt(sb, UPDATE_JOURNAL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001552 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001553 case Opt_journal_dev:
1554 if (is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001555 ext4_msg(sb, KERN_ERR,
1556 "Cannot specify journal on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001557 return 0;
1558 }
1559 if (match_int(&args[0], &option))
1560 return 0;
1561 *journal_devnum = option;
1562 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001563 case Opt_journal_checksum:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001564 set_opt(sb, JOURNAL_CHECKSUM);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08001565 break;
Girish Shilamkar818d2762008-01-28 23:58:27 -05001566 case Opt_journal_async_commit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001567 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1568 set_opt(sb, JOURNAL_CHECKSUM);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001569 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001570 case Opt_noload:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001571 set_opt(sb, NOLOAD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001572 break;
1573 case Opt_commit:
1574 if (match_int(&args[0], &option))
1575 return 0;
1576 if (option < 0)
1577 return 0;
1578 if (option == 0)
Mingming Caocd02ff02007-10-16 18:38:25 -04001579 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001580 sbi->s_commit_interval = HZ * option;
1581 break;
Theodore Ts'o30773842009-01-03 20:27:38 -05001582 case Opt_max_batch_time:
1583 if (match_int(&args[0], &option))
1584 return 0;
1585 if (option < 0)
1586 return 0;
1587 if (option == 0)
1588 option = EXT4_DEF_MAX_BATCH_TIME;
1589 sbi->s_max_batch_time = option;
1590 break;
1591 case Opt_min_batch_time:
1592 if (match_int(&args[0], &option))
1593 return 0;
1594 if (option < 0)
1595 return 0;
1596 sbi->s_min_batch_time = option;
1597 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001598 case Opt_data_journal:
Mingming Cao617ba132006-10-11 01:20:53 -07001599 data_opt = EXT4_MOUNT_JOURNAL_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001600 goto datacheck;
1601 case Opt_data_ordered:
Mingming Cao617ba132006-10-11 01:20:53 -07001602 data_opt = EXT4_MOUNT_ORDERED_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001603 goto datacheck;
1604 case Opt_data_writeback:
Mingming Cao617ba132006-10-11 01:20:53 -07001605 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001606 datacheck:
1607 if (is_remount) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001608 if (test_opt(sb, DATA_FLAGS) != data_opt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001609 ext4_msg(sb, KERN_ERR,
1610 "Cannot change data mode on remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001611 return 0;
1612 }
1613 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001614 clear_opt(sb, DATA_FLAGS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001615 sbi->s_mount_opt |= data_opt;
1616 }
1617 break;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001618 case Opt_data_err_abort:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001619 set_opt(sb, DATA_ERR_ABORT);
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001620 break;
1621 case Opt_data_err_ignore:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001622 clear_opt(sb, DATA_ERR_ABORT);
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001623 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001624#ifdef CONFIG_QUOTA
1625 case Opt_usrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001626 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1627 return 0;
1628 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001629 case Opt_grpjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001630 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001631 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001632 break;
1633 case Opt_offusrjquota:
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001634 if (!clear_qf_name(sb, USRQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001635 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001636 break;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001637 case Opt_offgrpjquota:
1638 if (!clear_qf_name(sb, GRPQUOTA))
1639 return 0;
1640 break;
1641
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001642 case Opt_jqfmt_vfsold:
Jan Karadfc5d032008-05-13 19:11:51 -04001643 qfmt = QFMT_VFS_OLD;
1644 goto set_qf_format;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001645 case Opt_jqfmt_vfsv0:
Jan Karadfc5d032008-05-13 19:11:51 -04001646 qfmt = QFMT_VFS_V0;
Jan Kara5a20bdf2009-11-30 23:58:32 +01001647 goto set_qf_format;
1648 case Opt_jqfmt_vfsv1:
1649 qfmt = QFMT_VFS_V1;
Jan Karadfc5d032008-05-13 19:11:51 -04001650set_qf_format:
Jan Kara17bd13b2008-08-20 18:14:35 +02001651 if (sb_any_quota_loaded(sb) &&
Jan Karadfc5d032008-05-13 19:11:51 -04001652 sbi->s_jquota_fmt != qfmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001653 ext4_msg(sb, KERN_ERR, "Cannot change "
Jan Karadfc5d032008-05-13 19:11:51 -04001654 "journaled quota options when "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001655 "quota turned on");
Jan Karadfc5d032008-05-13 19:11:51 -04001656 return 0;
1657 }
1658 sbi->s_jquota_fmt = qfmt;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001659 break;
1660 case Opt_quota:
1661 case Opt_usrquota:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001662 set_opt(sb, QUOTA);
1663 set_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001664 break;
1665 case Opt_grpquota:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001666 set_opt(sb, QUOTA);
1667 set_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001668 break;
1669 case Opt_noquota:
Jan Kara17bd13b2008-08-20 18:14:35 +02001670 if (sb_any_quota_loaded(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001671 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1672 "options when quota turned on");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001673 return 0;
1674 }
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001675 clear_opt(sb, QUOTA);
1676 clear_opt(sb, USRQUOTA);
1677 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001678 break;
1679#else
1680 case Opt_quota:
1681 case Opt_usrquota:
1682 case Opt_grpquota:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001683 ext4_msg(sb, KERN_ERR,
1684 "quota options not supported");
Jan Karacd59e7b2008-05-13 19:11:51 -04001685 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001686 case Opt_usrjquota:
1687 case Opt_grpjquota:
1688 case Opt_offusrjquota:
1689 case Opt_offgrpjquota:
1690 case Opt_jqfmt_vfsold:
1691 case Opt_jqfmt_vfsv0:
Jan Kara5a20bdf2009-11-30 23:58:32 +01001692 case Opt_jqfmt_vfsv1:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001693 ext4_msg(sb, KERN_ERR,
1694 "journaled quota options not supported");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001695 break;
1696 case Opt_noquota:
1697 break;
1698#endif
1699 case Opt_abort:
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04001700 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001701 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001702 case Opt_nobarrier:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001703 clear_opt(sb, BARRIER);
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001704 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001705 case Opt_barrier:
Eric Sandeen15121c12010-02-15 20:17:55 -05001706 if (args[0].from) {
1707 if (match_int(&args[0], &option))
1708 return 0;
1709 } else
1710 option = 1; /* No argument, default to 1 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001711 if (option)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001712 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001713 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001714 clear_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001715 break;
1716 case Opt_ignore:
1717 break;
1718 case Opt_resize:
1719 if (!is_remount) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001720 ext4_msg(sb, KERN_ERR,
1721 "resize option only available "
1722 "for remount");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001723 return 0;
1724 }
1725 if (match_int(&args[0], &option) != 0)
1726 return 0;
1727 *n_blocks_count = option;
1728 break;
1729 case Opt_nobh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001730 ext4_msg(sb, KERN_WARNING,
1731 "Ignoring deprecated nobh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001732 break;
1733 case Opt_bh:
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04001734 ext4_msg(sb, KERN_WARNING,
1735 "Ignoring deprecated bh option");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001736 break;
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001737 case Opt_i_version:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001738 set_opt(sb, I_VERSION);
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001739 sb->s_flags |= MS_I_VERSION;
1740 break;
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001741 case Opt_nodelalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001742 clear_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001743 break;
Theodore Ts'o14490322010-12-14 15:27:50 -05001744 case Opt_mblk_io_submit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001745 set_opt(sb, MBLK_IO_SUBMIT);
Theodore Ts'o14490322010-12-14 15:27:50 -05001746 break;
1747 case Opt_nomblk_io_submit:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001748 clear_opt(sb, MBLK_IO_SUBMIT);
Theodore Ts'o14490322010-12-14 15:27:50 -05001749 break;
Alex Tomasc9de5602008-01-29 00:19:52 -05001750 case Opt_stripe:
1751 if (match_int(&args[0], &option))
1752 return 0;
1753 if (option < 0)
1754 return 0;
1755 sbi->s_stripe = option;
1756 break;
Alex Tomas64769242008-07-11 19:27:31 -04001757 case Opt_delalloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001758 set_opt(sb, DELALLOC);
Alex Tomas64769242008-07-11 19:27:31 -04001759 break;
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001760 case Opt_block_validity:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001761 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001762 break;
1763 case Opt_noblock_validity:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001764 clear_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001765 break;
Theodore Ts'o240799c2008-10-09 23:53:47 -04001766 case Opt_inode_readahead_blks:
1767 if (match_int(&args[0], &option))
1768 return 0;
1769 if (option < 0 || option > (1 << 30))
1770 return 0;
Alexander V. Lukyanov5dbd5712011-02-21 21:33:21 -05001771 if (option && !is_power_of_2(option)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001772 ext4_msg(sb, KERN_ERR,
1773 "EXT4-fs: inode_readahead_blks"
1774 " must be a power of 2");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001775 return 0;
1776 }
Theodore Ts'o240799c2008-10-09 23:53:47 -04001777 sbi->s_inode_readahead_blks = option;
1778 break;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001779 case Opt_journal_ioprio:
1780 if (match_int(&args[0], &option))
1781 return 0;
1782 if (option < 0 || option > 7)
1783 break;
1784 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1785 option);
1786 break;
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001787 case Opt_noauto_da_alloc:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001788 set_opt(sb, NO_AUTO_DA_ALLOC);
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001789 break;
Theodore Ts'oafd46722009-03-16 23:12:23 -04001790 case Opt_auto_da_alloc:
Eric Sandeen15121c12010-02-15 20:17:55 -05001791 if (args[0].from) {
1792 if (match_int(&args[0], &option))
1793 return 0;
1794 } else
1795 option = 1; /* No argument, default to 1 */
Theodore Ts'oafd46722009-03-16 23:12:23 -04001796 if (option)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001797 clear_opt(sb, NO_AUTO_DA_ALLOC);
Theodore Ts'oafd46722009-03-16 23:12:23 -04001798 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001799 set_opt(sb,NO_AUTO_DA_ALLOC);
Theodore Ts'oafd46722009-03-16 23:12:23 -04001800 break;
Eric Sandeen5328e632009-11-19 14:25:42 -05001801 case Opt_discard:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001802 set_opt(sb, DISCARD);
Eric Sandeen5328e632009-11-19 14:25:42 -05001803 break;
1804 case Opt_nodiscard:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001805 clear_opt(sb, DISCARD);
Eric Sandeen5328e632009-11-19 14:25:42 -05001806 break;
Jiaying Zhang744692d2010-03-04 16:14:02 -05001807 case Opt_dioread_nolock:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001808 set_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001809 break;
1810 case Opt_dioread_lock:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001811 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001812 break;
Lukas Czernerbfff6872010-10-27 21:30:05 -04001813 case Opt_init_inode_table:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001814 set_opt(sb, INIT_INODE_TABLE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001815 if (args[0].from) {
1816 if (match_int(&args[0], &option))
1817 return 0;
1818 } else
1819 option = EXT4_DEF_LI_WAIT_MULT;
1820 if (option < 0)
1821 return 0;
1822 sbi->s_li_wait_mult = option;
1823 break;
1824 case Opt_noinit_inode_table:
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001825 clear_opt(sb, INIT_INODE_TABLE);
Lukas Czernerbfff6872010-10-27 21:30:05 -04001826 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001827 default:
Eric Sandeenb31e1552009-06-04 17:36:36 -04001828 ext4_msg(sb, KERN_ERR,
1829 "Unrecognized mount option \"%s\" "
1830 "or missing value", p);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001831 return 0;
1832 }
1833 }
1834#ifdef CONFIG_QUOTA
1835 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001836 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001837 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001838
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001839 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001840 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001841
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001842 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001843 ext4_msg(sb, KERN_ERR, "old and new quota "
1844 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001845 return 0;
1846 }
1847
1848 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001849 ext4_msg(sb, KERN_ERR, "journaled quota format "
1850 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001851 return 0;
1852 }
1853 } else {
1854 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001855 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001856 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001857 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001858 return 0;
1859 }
1860 }
1861#endif
1862 return 1;
1863}
1864
Mingming Cao617ba132006-10-11 01:20:53 -07001865static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001866 int read_only)
1867{
Mingming Cao617ba132006-10-11 01:20:53 -07001868 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001869 int res = 0;
1870
Mingming Cao617ba132006-10-11 01:20:53 -07001871 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001872 ext4_msg(sb, KERN_ERR, "revision level too high, "
1873 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001874 res = MS_RDONLY;
1875 }
1876 if (read_only)
1877 return res;
Mingming Cao617ba132006-10-11 01:20:53 -07001878 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001879 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1880 "running e2fsck is recommended");
Mingming Cao617ba132006-10-11 01:20:53 -07001881 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001882 ext4_msg(sb, KERN_WARNING,
1883 "warning: mounting fs with errors, "
1884 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001885 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1886 le16_to_cpu(es->s_mnt_count) >=
1887 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001888 ext4_msg(sb, KERN_WARNING,
1889 "warning: maximal mount count reached, "
1890 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001891 else if (le32_to_cpu(es->s_checkinterval) &&
1892 (le32_to_cpu(es->s_lastcheck) +
1893 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001894 ext4_msg(sb, KERN_WARNING,
1895 "warning: checktime reached, "
1896 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001897 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001898 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001899 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001900 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001901 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001902 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001903 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001904 if (sbi->s_journal)
1905 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001906
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001907 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001908 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001909 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001910 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001911 sb->s_blocksize,
1912 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001913 EXT4_BLOCKS_PER_GROUP(sb),
1914 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001915 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001916
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001917 return res;
1918}
1919
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001920static int ext4_fill_flex_info(struct super_block *sb)
1921{
1922 struct ext4_sb_info *sbi = EXT4_SB(sb);
1923 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001924 ext4_group_t flex_group_count;
1925 ext4_group_t flex_group;
1926 int groups_per_flex = 0;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001927 size_t size;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001928 int i;
1929
Theodore Ts'o503358a2009-11-23 07:24:46 -05001930 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1931 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1932
1933 if (groups_per_flex < 2) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001934 sbi->s_log_groups_per_flex = 0;
1935 return 1;
1936 }
1937
Frederic Bohec62a11f2008-09-08 10:20:24 -04001938 /* We allocate both existing and potentially added groups */
1939 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
Aneesh Kumar K.Vd94e99a2008-11-04 09:11:26 -05001940 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1941 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04001942 size = flex_group_count * sizeof(struct flex_groups);
1943 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1944 if (sbi->s_flex_groups == NULL) {
Joe Perches94de56a2010-12-19 22:21:02 -05001945 sbi->s_flex_groups = vzalloc(size);
1946 if (sbi->s_flex_groups == NULL) {
1947 ext4_msg(sb, KERN_ERR,
1948 "not enough memory for %u flex groups",
1949 flex_group_count);
1950 goto failed;
1951 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001952 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001953
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001954 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001955 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001956
1957 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001958 atomic_add(ext4_free_inodes_count(sb, gdp),
1959 &sbi->s_flex_groups[flex_group].free_inodes);
1960 atomic_add(ext4_free_blks_count(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].free_blocks);
1962 atomic_add(ext4_used_dirs_count(sb, gdp),
1963 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001964 }
1965
1966 return 1;
1967failed:
1968 return 0;
1969}
1970
Andreas Dilger717d50e2007-10-16 18:38:25 -04001971__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1972 struct ext4_group_desc *gdp)
1973{
1974 __u16 crc = 0;
1975
1976 if (sbi->s_es->s_feature_ro_compat &
1977 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1978 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1979 __le32 le_group = cpu_to_le32(block_group);
1980
1981 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1982 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1983 crc = crc16(crc, (__u8 *)gdp, offset);
1984 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1985 /* for checksum of struct ext4_group_desc do the rest...*/
1986 if ((sbi->s_es->s_feature_incompat &
1987 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1988 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1989 crc = crc16(crc, (__u8 *)gdp + offset,
1990 le16_to_cpu(sbi->s_es->s_desc_size) -
1991 offset);
1992 }
1993
1994 return cpu_to_le16(crc);
1995}
1996
1997int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1998 struct ext4_group_desc *gdp)
1999{
2000 if ((sbi->s_es->s_feature_ro_compat &
2001 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2002 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2003 return 0;
2004
2005 return 1;
2006}
2007
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002008/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002009static int ext4_check_descriptors(struct super_block *sb,
2010 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002011{
Mingming Cao617ba132006-10-11 01:20:53 -07002012 struct ext4_sb_info *sbi = EXT4_SB(sb);
2013 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2014 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002015 ext4_fsblk_t block_bitmap;
2016 ext4_fsblk_t inode_bitmap;
2017 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002018 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002019 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002020
Jose R. Santosce421582007-10-16 18:38:25 -04002021 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2022 flexbg_flag = 1;
2023
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002024 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002025
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002026 for (i = 0; i < sbi->s_groups_count; i++) {
2027 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2028
Jose R. Santosce421582007-10-16 18:38:25 -04002029 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002030 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002031 else
2032 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002033 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002034
Lukas Czernerbfff6872010-10-27 21:30:05 -04002035 if ((grp == sbi->s_groups_count) &&
2036 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2037 grp = i;
2038
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002039 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002040 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002041 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002042 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002043 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002044 return 0;
2045 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002046 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002047 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002048 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002049 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002050 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002051 return 0;
2052 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002053 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002054 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002055 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002056 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002057 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002058 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002059 return 0;
2060 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002061 ext4_lock_group(sb, i);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002062 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002063 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2064 "Checksum for group %u failed (%u!=%u)",
2065 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2066 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002067 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002068 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002069 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002070 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002071 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002072 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002073 if (!flexbg_flag)
2074 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002075 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002076 if (NULL != first_not_zeroed)
2077 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002078
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002079 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002080 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002081 return 1;
2082}
2083
Mingming Cao617ba132006-10-11 01:20:53 -07002084/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002085 * the superblock) which were deleted from all directories, but held open by
2086 * a process at the time of a crash. We walk the list and try to delete these
2087 * inodes at recovery time (only with a read-write filesystem).
2088 *
2089 * In order to keep the orphan inode chain consistent during traversal (in
2090 * case of crash during recovery), we link each inode into the superblock
2091 * orphan list_head and handle it the same way as an inode deletion during
2092 * normal operation (which journals the operations for us).
2093 *
2094 * We only do an iget() and an iput() on each inode, which is very safe if we
2095 * accidentally point at an in-use or already deleted inode. The worst that
2096 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002097 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002098 * e2fsck was run on this filesystem, and it must have already done the orphan
2099 * inode cleanup for us, so we can safely abort without any further action.
2100 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002101static void ext4_orphan_cleanup(struct super_block *sb,
2102 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002103{
2104 unsigned int s_flags = sb->s_flags;
2105 int nr_orphans = 0, nr_truncates = 0;
2106#ifdef CONFIG_QUOTA
2107 int i;
2108#endif
2109 if (!es->s_last_orphan) {
2110 jbd_debug(4, "no orphan inodes to clean up\n");
2111 return;
2112 }
2113
Eric Sandeena8f48a92006-12-06 20:40:13 -08002114 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002115 ext4_msg(sb, KERN_ERR, "write access "
2116 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002117 return;
2118 }
2119
Mingming Cao617ba132006-10-11 01:20:53 -07002120 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002121 if (es->s_last_orphan)
2122 jbd_debug(1, "Errors on filesystem, "
2123 "clearing orphan list.\n");
2124 es->s_last_orphan = 0;
2125 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2126 return;
2127 }
2128
2129 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002130 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002131 sb->s_flags &= ~MS_RDONLY;
2132 }
2133#ifdef CONFIG_QUOTA
2134 /* Needed for iput() to work correctly and not trash data */
2135 sb->s_flags |= MS_ACTIVE;
2136 /* Turn on quotas so that they are updated correctly */
2137 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002138 if (EXT4_SB(sb)->s_qf_names[i]) {
2139 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002140 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002141 ext4_msg(sb, KERN_ERR,
2142 "Cannot turn on journaled "
2143 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002144 }
2145 }
2146#endif
2147
2148 while (es->s_last_orphan) {
2149 struct inode *inode;
2150
Josef Bacik97bd42b2008-04-29 22:04:56 -04002151 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2152 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002153 es->s_last_orphan = 0;
2154 break;
2155 }
2156
Mingming Cao617ba132006-10-11 01:20:53 -07002157 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002158 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002159 if (inode->i_nlink) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002160 ext4_msg(sb, KERN_DEBUG,
2161 "%s: truncating inode %lu to %lld bytes",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002162 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002163 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002164 inode->i_ino, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002165 ext4_truncate(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002166 nr_truncates++;
2167 } else {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002168 ext4_msg(sb, KERN_DEBUG,
2169 "%s: deleting unreferenced inode %lu",
Harvey Harrison46e665e2008-04-17 10:38:59 -04002170 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002171 jbd_debug(2, "deleting unreferenced inode %lu\n",
2172 inode->i_ino);
2173 nr_orphans++;
2174 }
2175 iput(inode); /* The delete magic happens here! */
2176 }
2177
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002178#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002179
2180 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002181 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2182 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002183 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002184 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2185 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002186#ifdef CONFIG_QUOTA
2187 /* Turn quotas off */
2188 for (i = 0; i < MAXQUOTAS; i++) {
2189 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002190 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002191 }
2192#endif
2193 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2194}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002195
Eric Sandeencd2291a2008-01-28 23:58:27 -05002196/*
2197 * Maximal extent format file size.
2198 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2199 * extent format containers, within a sector_t, and within i_blocks
2200 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2201 * so that won't be a limiting factor.
2202 *
2203 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2204 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002205static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002206{
2207 loff_t res;
2208 loff_t upper_limit = MAX_LFS_FILESIZE;
2209
2210 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002211 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002212 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002213 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002214 * i_block represent total blocks in 512 bytes
2215 * 32 == size of vfs inode i_blocks * 8
2216 */
2217 upper_limit = (1LL << 32) - 1;
2218
2219 /* total blocks in file system block size */
2220 upper_limit >>= (blkbits - 9);
2221 upper_limit <<= blkbits;
2222 }
2223
2224 /* 32-bit extent-start container, ee_block */
2225 res = 1LL << 32;
2226 res <<= blkbits;
2227 res -= 1;
2228
2229 /* Sanity check against vm- & vfs- imposed limits */
2230 if (res > upper_limit)
2231 res = upper_limit;
2232
2233 return res;
2234}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002235
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002236/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002237 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002238 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2239 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002240 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002241static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002242{
Mingming Cao617ba132006-10-11 01:20:53 -07002243 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002244 int meta_blocks;
2245 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002246 /* This is calculated to be the largest file size for a dense, block
2247 * mapped file such that the file's total number of 512-byte sectors,
2248 * including data and all indirect blocks, does not exceed (2^48 - 1).
2249 *
2250 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2251 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002252 */
2253
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002254 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002255 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002256 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002257 * the inode i_block field represents total file blocks in
2258 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002259 */
2260 upper_limit = (1LL << 32) - 1;
2261
2262 /* total blocks in file system block size */
2263 upper_limit >>= (bits - 9);
2264
2265 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002266 /*
2267 * We use 48 bit ext4_inode i_blocks
2268 * With EXT4_HUGE_FILE_FL set the i_blocks
2269 * represent total number of blocks in
2270 * file system block size
2271 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002272 upper_limit = (1LL << 48) - 1;
2273
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002274 }
2275
2276 /* indirect blocks */
2277 meta_blocks = 1;
2278 /* double indirect blocks */
2279 meta_blocks += 1 + (1LL << (bits-2));
2280 /* tripple indirect blocks */
2281 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2282
2283 upper_limit -= meta_blocks;
2284 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002285
2286 res += 1LL << (bits-2);
2287 res += 1LL << (2*(bits-2));
2288 res += 1LL << (3*(bits-2));
2289 res <<= bits;
2290 if (res > upper_limit)
2291 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002292
2293 if (res > MAX_LFS_FILESIZE)
2294 res = MAX_LFS_FILESIZE;
2295
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002296 return res;
2297}
2298
Mingming Cao617ba132006-10-11 01:20:53 -07002299static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002300 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002301{
Mingming Cao617ba132006-10-11 01:20:53 -07002302 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002303 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002304 int has_super = 0;
2305
2306 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2307
Mingming Cao617ba132006-10-11 01:20:53 -07002308 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002309 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002310 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002311 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002312 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002313 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002314
Mingming Cao617ba132006-10-11 01:20:53 -07002315 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002316}
2317
Alex Tomasc9de5602008-01-29 00:19:52 -05002318/**
2319 * ext4_get_stripe_size: Get the stripe size.
2320 * @sbi: In memory super block info
2321 *
2322 * If we have specified it via mount option, then
2323 * use the mount option value. If the value specified at mount time is
2324 * greater than the blocks per group use the super block value.
2325 * If the super block value is greater than blocks per group return 0.
2326 * Allocator needs it be less than blocks per group.
2327 *
2328 */
2329static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2330{
2331 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2332 unsigned long stripe_width =
2333 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2334
2335 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2336 return sbi->s_stripe;
2337
2338 if (stripe_width <= sbi->s_blocks_per_group)
2339 return stripe_width;
2340
2341 if (stride <= sbi->s_blocks_per_group)
2342 return stride;
2343
2344 return 0;
2345}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002346
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002347/* sysfs supprt */
2348
2349struct ext4_attr {
2350 struct attribute attr;
2351 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
Theodore Ts'o60e66792010-05-17 07:00:00 -04002352 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002353 const char *, size_t);
2354 int offset;
2355};
2356
2357static int parse_strtoul(const char *buf,
2358 unsigned long max, unsigned long *value)
2359{
2360 char *endp;
2361
André Goddard Rosae7d28602009-12-14 18:01:06 -08002362 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2363 endp = skip_spaces(endp);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002364 if (*endp || *value > max)
2365 return -EINVAL;
2366
2367 return 0;
2368}
2369
2370static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2371 struct ext4_sb_info *sbi,
2372 char *buf)
2373{
2374 return snprintf(buf, PAGE_SIZE, "%llu\n",
2375 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2376}
2377
2378static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2379 struct ext4_sb_info *sbi, char *buf)
2380{
2381 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2382
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002383 if (!sb->s_bdev->bd_part)
2384 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002385 return snprintf(buf, PAGE_SIZE, "%lu\n",
2386 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2387 sbi->s_sectors_written_start) >> 1);
2388}
2389
2390static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2391 struct ext4_sb_info *sbi, char *buf)
2392{
2393 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2394
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002395 if (!sb->s_bdev->bd_part)
2396 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002397 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002398 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002399 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002400 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002401}
2402
2403static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2404 struct ext4_sb_info *sbi,
2405 const char *buf, size_t count)
2406{
2407 unsigned long t;
2408
2409 if (parse_strtoul(buf, 0x40000000, &t))
2410 return -EINVAL;
2411
Alexander V. Lukyanov5dbd5712011-02-21 21:33:21 -05002412 if (t && !is_power_of_2(t))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002413 return -EINVAL;
2414
2415 sbi->s_inode_readahead_blks = t;
2416 return count;
2417}
2418
2419static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002420 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002421{
2422 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2423
2424 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2425}
2426
2427static ssize_t sbi_ui_store(struct ext4_attr *a,
2428 struct ext4_sb_info *sbi,
2429 const char *buf, size_t count)
2430{
2431 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2432 unsigned long t;
2433
2434 if (parse_strtoul(buf, 0xffffffff, &t))
2435 return -EINVAL;
2436 *ui = t;
2437 return count;
2438}
2439
2440#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2441static struct ext4_attr ext4_attr_##_name = { \
2442 .attr = {.name = __stringify(_name), .mode = _mode }, \
2443 .show = _show, \
2444 .store = _store, \
2445 .offset = offsetof(struct ext4_sb_info, _elname), \
2446}
2447#define EXT4_ATTR(name, mode, show, store) \
2448static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2449
Lukas Czerner857ac882010-10-27 21:30:05 -04002450#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002451#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2452#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2453#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2454 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2455#define ATTR_LIST(name) &ext4_attr_##name.attr
2456
2457EXT4_RO_ATTR(delayed_allocation_blocks);
2458EXT4_RO_ATTR(session_write_kbytes);
2459EXT4_RO_ATTR(lifetime_write_kbytes);
2460EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2461 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002462EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002463EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2464EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2465EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2466EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2467EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2468EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Theodore Ts'o55138e02009-09-29 13:31:31 -04002469EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002470
2471static struct attribute *ext4_attrs[] = {
2472 ATTR_LIST(delayed_allocation_blocks),
2473 ATTR_LIST(session_write_kbytes),
2474 ATTR_LIST(lifetime_write_kbytes),
2475 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002476 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002477 ATTR_LIST(mb_stats),
2478 ATTR_LIST(mb_max_to_scan),
2479 ATTR_LIST(mb_min_to_scan),
2480 ATTR_LIST(mb_order2_req),
2481 ATTR_LIST(mb_stream_req),
2482 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002483 ATTR_LIST(max_writeback_mb_bump),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002484 NULL,
2485};
2486
Lukas Czerner857ac882010-10-27 21:30:05 -04002487/* Features this copy of ext4 supports */
2488EXT4_INFO_ATTR(lazy_itable_init);
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002489EXT4_INFO_ATTR(batched_discard);
Lukas Czerner857ac882010-10-27 21:30:05 -04002490
2491static struct attribute *ext4_feat_attrs[] = {
2492 ATTR_LIST(lazy_itable_init),
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002493 ATTR_LIST(batched_discard),
Lukas Czerner857ac882010-10-27 21:30:05 -04002494 NULL,
2495};
2496
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002497static ssize_t ext4_attr_show(struct kobject *kobj,
2498 struct attribute *attr, char *buf)
2499{
2500 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2501 s_kobj);
2502 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2503
2504 return a->show ? a->show(a, sbi, buf) : 0;
2505}
2506
2507static ssize_t ext4_attr_store(struct kobject *kobj,
2508 struct attribute *attr,
2509 const char *buf, size_t len)
2510{
2511 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2512 s_kobj);
2513 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2514
2515 return a->store ? a->store(a, sbi, buf, len) : 0;
2516}
2517
2518static void ext4_sb_release(struct kobject *kobj)
2519{
2520 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2521 s_kobj);
2522 complete(&sbi->s_kobj_unregister);
2523}
2524
Emese Revfy52cf25d2010-01-19 02:58:23 +01002525static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002526 .show = ext4_attr_show,
2527 .store = ext4_attr_store,
2528};
2529
2530static struct kobj_type ext4_ktype = {
2531 .default_attrs = ext4_attrs,
2532 .sysfs_ops = &ext4_attr_ops,
2533 .release = ext4_sb_release,
2534};
2535
Lukas Czerner857ac882010-10-27 21:30:05 -04002536static void ext4_feat_release(struct kobject *kobj)
2537{
2538 complete(&ext4_feat->f_kobj_unregister);
2539}
2540
2541static struct kobj_type ext4_feat_ktype = {
2542 .default_attrs = ext4_feat_attrs,
2543 .sysfs_ops = &ext4_attr_ops,
2544 .release = ext4_feat_release,
2545};
2546
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002547/*
2548 * Check whether this filesystem can be mounted based on
2549 * the features present and the RDONLY/RDWR mount requested.
2550 * Returns 1 if this filesystem can be mounted as requested,
2551 * 0 if it cannot be.
2552 */
2553static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2554{
2555 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2556 ext4_msg(sb, KERN_ERR,
2557 "Couldn't mount because of "
2558 "unsupported optional features (%x)",
2559 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2560 ~EXT4_FEATURE_INCOMPAT_SUPP));
2561 return 0;
2562 }
2563
2564 if (readonly)
2565 return 1;
2566
2567 /* Check that feature set is OK for a read-write mount */
2568 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2569 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2570 "unsupported optional features (%x)",
2571 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2572 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2573 return 0;
2574 }
2575 /*
2576 * Large file size enabled file system can only be mounted
2577 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2578 */
2579 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2580 if (sizeof(blkcnt_t) < sizeof(u64)) {
2581 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2582 "cannot be mounted RDWR without "
2583 "CONFIG_LBDAF");
2584 return 0;
2585 }
2586 }
2587 return 1;
2588}
2589
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002590/*
2591 * This function is called once a day if we have errors logged
2592 * on the file system
2593 */
2594static void print_daily_error_info(unsigned long arg)
2595{
2596 struct super_block *sb = (struct super_block *) arg;
2597 struct ext4_sb_info *sbi;
2598 struct ext4_super_block *es;
2599
2600 sbi = EXT4_SB(sb);
2601 es = sbi->s_es;
2602
2603 if (es->s_error_count)
2604 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2605 le32_to_cpu(es->s_error_count));
2606 if (es->s_first_error_time) {
2607 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2608 sb->s_id, le32_to_cpu(es->s_first_error_time),
2609 (int) sizeof(es->s_first_error_func),
2610 es->s_first_error_func,
2611 le32_to_cpu(es->s_first_error_line));
2612 if (es->s_first_error_ino)
2613 printk(": inode %u",
2614 le32_to_cpu(es->s_first_error_ino));
2615 if (es->s_first_error_block)
2616 printk(": block %llu", (unsigned long long)
2617 le64_to_cpu(es->s_first_error_block));
2618 printk("\n");
2619 }
2620 if (es->s_last_error_time) {
2621 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2622 sb->s_id, le32_to_cpu(es->s_last_error_time),
2623 (int) sizeof(es->s_last_error_func),
2624 es->s_last_error_func,
2625 le32_to_cpu(es->s_last_error_line));
2626 if (es->s_last_error_ino)
2627 printk(": inode %u",
2628 le32_to_cpu(es->s_last_error_ino));
2629 if (es->s_last_error_block)
2630 printk(": block %llu", (unsigned long long)
2631 le64_to_cpu(es->s_last_error_block));
2632 printk("\n");
2633 }
2634 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2635}
2636
Lukas Czernerbfff6872010-10-27 21:30:05 -04002637static void ext4_lazyinode_timeout(unsigned long data)
2638{
2639 struct task_struct *p = (struct task_struct *)data;
2640 wake_up_process(p);
2641}
2642
2643/* Find next suitable group and run ext4_init_inode_table */
2644static int ext4_run_li_request(struct ext4_li_request *elr)
2645{
2646 struct ext4_group_desc *gdp = NULL;
2647 ext4_group_t group, ngroups;
2648 struct super_block *sb;
2649 unsigned long timeout = 0;
2650 int ret = 0;
2651
2652 sb = elr->lr_super;
2653 ngroups = EXT4_SB(sb)->s_groups_count;
2654
2655 for (group = elr->lr_next_group; group < ngroups; group++) {
2656 gdp = ext4_get_group_desc(sb, group, NULL);
2657 if (!gdp) {
2658 ret = 1;
2659 break;
2660 }
2661
2662 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2663 break;
2664 }
2665
2666 if (group == ngroups)
2667 ret = 1;
2668
2669 if (!ret) {
2670 timeout = jiffies;
2671 ret = ext4_init_inode_table(sb, group,
2672 elr->lr_timeout ? 0 : 1);
2673 if (elr->lr_timeout == 0) {
2674 timeout = jiffies - timeout;
2675 if (elr->lr_sbi->s_li_wait_mult)
2676 timeout *= elr->lr_sbi->s_li_wait_mult;
2677 else
2678 timeout *= 20;
2679 elr->lr_timeout = timeout;
2680 }
2681 elr->lr_next_sched = jiffies + elr->lr_timeout;
2682 elr->lr_next_group = group + 1;
2683 }
2684
2685 return ret;
2686}
2687
2688/*
2689 * Remove lr_request from the list_request and free the
2690 * request tructure. Should be called with li_list_mtx held
2691 */
2692static void ext4_remove_li_request(struct ext4_li_request *elr)
2693{
2694 struct ext4_sb_info *sbi;
2695
2696 if (!elr)
2697 return;
2698
2699 sbi = elr->lr_sbi;
2700
2701 list_del(&elr->lr_request);
2702 sbi->s_li_request = NULL;
2703 kfree(elr);
2704}
2705
2706static void ext4_unregister_li_request(struct super_block *sb)
2707{
2708 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2709
2710 if (!ext4_li_info)
2711 return;
2712
2713 mutex_lock(&ext4_li_info->li_list_mtx);
2714 ext4_remove_li_request(elr);
2715 mutex_unlock(&ext4_li_info->li_list_mtx);
2716}
2717
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002718static struct task_struct *ext4_lazyinit_task;
2719
Lukas Czernerbfff6872010-10-27 21:30:05 -04002720/*
2721 * This is the function where ext4lazyinit thread lives. It walks
2722 * through the request list searching for next scheduled filesystem.
2723 * When such a fs is found, run the lazy initialization request
2724 * (ext4_rn_li_request) and keep track of the time spend in this
2725 * function. Based on that time we compute next schedule time of
2726 * the request. When walking through the list is complete, compute
2727 * next waking time and put itself into sleep.
2728 */
2729static int ext4_lazyinit_thread(void *arg)
2730{
2731 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2732 struct list_head *pos, *n;
2733 struct ext4_li_request *elr;
2734 unsigned long next_wakeup;
2735 DEFINE_WAIT(wait);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002736
2737 BUG_ON(NULL == eli);
2738
2739 eli->li_timer.data = (unsigned long)current;
2740 eli->li_timer.function = ext4_lazyinode_timeout;
2741
2742 eli->li_task = current;
2743 wake_up(&eli->li_wait_task);
2744
2745cont_thread:
2746 while (true) {
2747 next_wakeup = MAX_JIFFY_OFFSET;
2748
2749 mutex_lock(&eli->li_list_mtx);
2750 if (list_empty(&eli->li_request_list)) {
2751 mutex_unlock(&eli->li_list_mtx);
2752 goto exit_thread;
2753 }
2754
2755 list_for_each_safe(pos, n, &eli->li_request_list) {
2756 elr = list_entry(pos, struct ext4_li_request,
2757 lr_request);
2758
Theodore Ts'ob2c78cd2010-11-02 14:19:30 -04002759 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2760 if (ext4_run_li_request(elr) != 0) {
2761 /* error, remove the lazy_init job */
2762 ext4_remove_li_request(elr);
2763 continue;
2764 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002765 }
2766
2767 if (time_before(elr->lr_next_sched, next_wakeup))
2768 next_wakeup = elr->lr_next_sched;
2769 }
2770 mutex_unlock(&eli->li_list_mtx);
2771
2772 if (freezing(current))
2773 refrigerator();
2774
Lukas Czernerf4245bd2010-11-02 14:07:17 -04002775 if ((time_after_eq(jiffies, next_wakeup)) ||
2776 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04002777 cond_resched();
2778 continue;
2779 }
2780
2781 eli->li_timer.expires = next_wakeup;
2782 add_timer(&eli->li_timer);
2783 prepare_to_wait(&eli->li_wait_daemon, &wait,
2784 TASK_INTERRUPTIBLE);
2785 if (time_before(jiffies, next_wakeup))
2786 schedule();
2787 finish_wait(&eli->li_wait_daemon, &wait);
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002788 if (kthread_should_stop()) {
2789 ext4_clear_request_list();
2790 goto exit_thread;
2791 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002792 }
2793
2794exit_thread:
2795 /*
2796 * It looks like the request list is empty, but we need
2797 * to check it under the li_list_mtx lock, to prevent any
2798 * additions into it, and of course we should lock ext4_li_mtx
2799 * to atomically free the list and ext4_li_info, because at
2800 * this point another ext4 filesystem could be registering
2801 * new one.
2802 */
2803 mutex_lock(&ext4_li_mtx);
2804 mutex_lock(&eli->li_list_mtx);
2805 if (!list_empty(&eli->li_request_list)) {
2806 mutex_unlock(&eli->li_list_mtx);
2807 mutex_unlock(&ext4_li_mtx);
2808 goto cont_thread;
2809 }
2810 mutex_unlock(&eli->li_list_mtx);
2811 del_timer_sync(&ext4_li_info->li_timer);
2812 eli->li_task = NULL;
2813 wake_up(&eli->li_wait_task);
2814
2815 kfree(ext4_li_info);
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002816 ext4_lazyinit_task = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002817 ext4_li_info = NULL;
2818 mutex_unlock(&ext4_li_mtx);
2819
2820 return 0;
2821}
2822
2823static void ext4_clear_request_list(void)
2824{
2825 struct list_head *pos, *n;
2826 struct ext4_li_request *elr;
2827
2828 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002829 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2830 elr = list_entry(pos, struct ext4_li_request,
2831 lr_request);
2832 ext4_remove_li_request(elr);
2833 }
2834 mutex_unlock(&ext4_li_info->li_list_mtx);
2835}
2836
2837static int ext4_run_lazyinit_thread(void)
2838{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002839 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2840 ext4_li_info, "ext4lazyinit");
2841 if (IS_ERR(ext4_lazyinit_task)) {
2842 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002843 ext4_clear_request_list();
2844 del_timer_sync(&ext4_li_info->li_timer);
2845 kfree(ext4_li_info);
2846 ext4_li_info = NULL;
2847 printk(KERN_CRIT "EXT4: error %d creating inode table "
2848 "initialization thread\n",
2849 err);
2850 return err;
2851 }
2852 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2853
2854 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2855 return 0;
2856}
2857
2858/*
2859 * Check whether it make sense to run itable init. thread or not.
2860 * If there is at least one uninitialized inode table, return
2861 * corresponding group number, else the loop goes through all
2862 * groups and return total number of groups.
2863 */
2864static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2865{
2866 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2867 struct ext4_group_desc *gdp = NULL;
2868
2869 for (group = 0; group < ngroups; group++) {
2870 gdp = ext4_get_group_desc(sb, group, NULL);
2871 if (!gdp)
2872 continue;
2873
2874 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2875 break;
2876 }
2877
2878 return group;
2879}
2880
2881static int ext4_li_info_new(void)
2882{
2883 struct ext4_lazy_init *eli = NULL;
2884
2885 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2886 if (!eli)
2887 return -ENOMEM;
2888
2889 eli->li_task = NULL;
2890 INIT_LIST_HEAD(&eli->li_request_list);
2891 mutex_init(&eli->li_list_mtx);
2892
2893 init_waitqueue_head(&eli->li_wait_daemon);
2894 init_waitqueue_head(&eli->li_wait_task);
2895 init_timer(&eli->li_timer);
2896 eli->li_state |= EXT4_LAZYINIT_QUIT;
2897
2898 ext4_li_info = eli;
2899
2900 return 0;
2901}
2902
2903static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2904 ext4_group_t start)
2905{
2906 struct ext4_sb_info *sbi = EXT4_SB(sb);
2907 struct ext4_li_request *elr;
2908 unsigned long rnd;
2909
2910 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2911 if (!elr)
2912 return NULL;
2913
2914 elr->lr_super = sb;
2915 elr->lr_sbi = sbi;
2916 elr->lr_next_group = start;
2917
2918 /*
2919 * Randomize first schedule time of the request to
2920 * spread the inode table initialization requests
2921 * better.
2922 */
2923 get_random_bytes(&rnd, sizeof(rnd));
2924 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2925 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2926
2927 return elr;
2928}
2929
2930static int ext4_register_li_request(struct super_block *sb,
2931 ext4_group_t first_not_zeroed)
2932{
2933 struct ext4_sb_info *sbi = EXT4_SB(sb);
2934 struct ext4_li_request *elr;
2935 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05002936 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002937
2938 if (sbi->s_li_request != NULL)
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002939 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002940
2941 if (first_not_zeroed == ngroups ||
2942 (sb->s_flags & MS_RDONLY) ||
2943 !test_opt(sb, INIT_INODE_TABLE)) {
2944 sbi->s_li_request = NULL;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002945 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002946 }
2947
2948 if (first_not_zeroed == ngroups) {
2949 sbi->s_li_request = NULL;
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002950 return 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002951 }
2952
2953 elr = ext4_li_request_new(sb, first_not_zeroed);
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002954 if (!elr)
2955 return -ENOMEM;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002956
2957 mutex_lock(&ext4_li_mtx);
2958
2959 if (NULL == ext4_li_info) {
2960 ret = ext4_li_info_new();
2961 if (ret)
2962 goto out;
2963 }
2964
2965 mutex_lock(&ext4_li_info->li_list_mtx);
2966 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2967 mutex_unlock(&ext4_li_info->li_list_mtx);
2968
2969 sbi->s_li_request = elr;
2970
2971 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2972 ret = ext4_run_lazyinit_thread();
2973 if (ret)
2974 goto out;
2975 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002976out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04002977 mutex_unlock(&ext4_li_mtx);
2978 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002979 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002980 return ret;
2981}
2982
2983/*
2984 * We do not need to lock anything since this is called on
2985 * module unload.
2986 */
2987static void ext4_destroy_lazyinit_thread(void)
2988{
2989 /*
2990 * If thread exited earlier
2991 * there's nothing to be done.
2992 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002993 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002994 return;
2995
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002996 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002997}
2998
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002999static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Aneesh Kumar K.V74778272008-07-11 19:27:31 -04003000 __releases(kernel_lock)
3001 __acquires(kernel_lock)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003002{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003003 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003004 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003005 struct ext4_super_block *es = NULL;
3006 struct ext4_sb_info *sbi;
3007 ext4_fsblk_t block;
3008 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003009 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003010 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003011 unsigned long journal_devnum = 0;
3012 unsigned long def_mount_opts;
3013 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003014 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05003015 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003016 int ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003017 int blocksize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003018 unsigned int db_count;
3019 unsigned int i;
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003020 int needs_recovery, has_huge_files;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003021 __u64 blocks_count;
Peter Zijlstra833f4072007-10-16 23:25:45 -07003022 int err;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003023 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003024 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003025
3026 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3027 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003028 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003029
3030 sbi->s_blockgroup_lock =
3031 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3032 if (!sbi->s_blockgroup_lock) {
3033 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003034 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003035 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003036 sb->s_fs_info = sbi;
3037 sbi->s_mount_opt = 0;
Mingming Cao617ba132006-10-11 01:20:53 -07003038 sbi->s_resuid = EXT4_DEF_RESUID;
3039 sbi->s_resgid = EXT4_DEF_RESGID;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003040 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003041 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003042 if (sb->s_bdev->bd_part)
3043 sbi->s_sectors_written_start =
3044 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003045
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003046 /* Cleanup superblock name */
3047 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3048 *cp = '!';
3049
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003050 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003051 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003052 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003053 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003054 goto out_fail;
3055 }
3056
3057 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003058 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003059 * block sizes. We need to calculate the offset from buffer start.
3060 */
Mingming Cao617ba132006-10-11 01:20:53 -07003061 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003062 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3063 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003064 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003065 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003066 }
3067
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003068 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003069 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003070 goto out_fail;
3071 }
3072 /*
3073 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003074 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003075 */
Mingming Cao617ba132006-10-11 01:20:53 -07003076 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003077 sbi->s_es = es;
3078 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003079 if (sb->s_magic != EXT4_SUPER_MAGIC)
3080 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003081 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003082
3083 /* Set defaults before we parse the mount options */
3084 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003085 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003086 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003087 set_opt(sb, DEBUG);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003088 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3089 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3090 "2.6.38");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003091 set_opt(sb, GRPID);
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05003092 }
Mingming Cao617ba132006-10-11 01:20:53 -07003093 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003094 set_opt(sb, NO_UID32);
Eric Sandeenea663332011-02-23 17:51:51 -05003095 /* xattr user namespace & acls are now defaulted on */
Theodore Ts'o03010a32008-10-10 20:02:48 -04003096#ifdef CONFIG_EXT4_FS_XATTR
Eric Sandeenea663332011-02-23 17:51:51 -05003097 set_opt(sb, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003098#endif
Theodore Ts'o03010a32008-10-10 20:02:48 -04003099#ifdef CONFIG_EXT4_FS_POSIX_ACL
Eric Sandeenea663332011-02-23 17:51:51 -05003100 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003101#endif
Theodore Ts'o6fd7a462011-02-26 13:53:09 -05003102 set_opt(sb, MBLK_IO_SUBMIT);
Mingming Cao617ba132006-10-11 01:20:53 -07003103 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003104 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003105 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003106 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003107 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003108 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003109
Mingming Cao617ba132006-10-11 01:20:53 -07003110 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003111 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003112 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003113 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003114 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003115 set_opt(sb, ERRORS_RO);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003116 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003117 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003118 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003119 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003120
3121 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3122 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
Theodore Ts'o30773842009-01-03 20:27:38 -05003123 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3124 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3125 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003126
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003127 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003128 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003129
Mingming Cao1e2462f2007-07-18 09:00:55 -04003130 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003131 * enable delayed allocation by default
3132 * Use -o nodelalloc to turn it off
3133 */
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003134 if (!IS_EXT3_SB(sb) &&
3135 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003136 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003137
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003138 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3139 &journal_devnum, &journal_ioprio, NULL, 0)) {
3140 ext4_msg(sb, KERN_WARNING,
3141 "failed to parse options in superblock: %s",
3142 sbi->s_es->s_mount_opts);
3143 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003144 if (!parse_options((char *) data, sb, &journal_devnum,
3145 &journal_ioprio, NULL, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003146 goto failed_mount;
3147
3148 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003149 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003150
Mingming Cao617ba132006-10-11 01:20:53 -07003151 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3152 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3153 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3154 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003155 ext4_msg(sb, KERN_WARNING,
3156 "feature flags set on rev 0 fs, "
3157 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003158
3159 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003160 * Check feature flags regardless of the revision level, since we
3161 * previously didn't change the revision level when setting the flags,
3162 * so there is a chance incompat flags are set on a rev 0 filesystem.
3163 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003164 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003165 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003166
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003167 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3168
Mingming Cao617ba132006-10-11 01:20:53 -07003169 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3170 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003171 ext4_msg(sb, KERN_ERR,
3172 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003173 goto failed_mount;
3174 }
3175
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003176 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003177 /* Validate the filesystem blocksize */
3178 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003179 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003180 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003181 goto failed_mount;
3182 }
3183
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003184 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003185 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3186 offset = do_div(logical_sb_block, blocksize);
3187 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003188 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003189 ext4_msg(sb, KERN_ERR,
3190 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003191 goto failed_mount;
3192 }
Mingming Cao617ba132006-10-11 01:20:53 -07003193 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003194 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003195 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003196 ext4_msg(sb, KERN_ERR,
3197 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003198 goto failed_mount;
3199 }
3200 }
3201
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003202 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3203 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003204 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3205 has_huge_files);
3206 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003207
Mingming Cao617ba132006-10-11 01:20:53 -07003208 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3209 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3210 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003211 } else {
3212 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3213 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003214 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003215 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003216 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003217 ext4_msg(sb, KERN_ERR,
3218 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003219 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003220 goto failed_mount;
3221 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003222 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3223 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003224 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003225
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003226 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3227 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003228 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003229 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003230 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003231 ext4_msg(sb, KERN_ERR,
3232 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003233 sbi->s_desc_size);
3234 goto failed_mount;
3235 }
3236 } else
3237 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003238
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003239 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003240 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003241 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003242 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003243
Mingming Cao617ba132006-10-11 01:20:53 -07003244 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003245 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003246 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003247 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3248 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003249 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003250 sbi->s_sbh = bh;
3251 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003252 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3253 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003254
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003255 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003256 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3257 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003258 i = le32_to_cpu(es->s_flags);
3259 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3260 sbi->s_hash_unsigned = 3;
3261 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3262#ifdef __CHAR_UNSIGNED__
3263 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3264 sbi->s_hash_unsigned = 3;
3265#else
3266 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3267#endif
3268 sb->s_dirt = 1;
3269 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003270
3271 if (sbi->s_blocks_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003272 ext4_msg(sb, KERN_ERR,
3273 "#blocks per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003274 sbi->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003275 goto failed_mount;
3276 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003277 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003278 ext4_msg(sb, KERN_ERR,
3279 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003280 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003281 goto failed_mount;
3282 }
3283
Eric Sandeenbf43d842009-08-17 23:48:51 -04003284 /*
3285 * Test whether we have more sectors than will fit in sector_t,
3286 * and whether the max offset is addressable by the page cache.
3287 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003288 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003289 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003290 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003291 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003292 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003293 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003294 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003295 ret = err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003296 goto failed_mount;
3297 }
3298
Mingming Cao617ba132006-10-11 01:20:53 -07003299 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3300 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003301
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003302 /* check blocks count against device size */
3303 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3304 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003305 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3306 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003307 ext4_blocks_count(es), blocks_count);
3308 goto failed_mount;
3309 }
3310
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003311 /*
3312 * It makes no sense for the first data block to be beyond the end
3313 * of the filesystem.
3314 */
3315 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003316 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3317 "block %u is beyond end of filesystem (%llu)",
3318 le32_to_cpu(es->s_first_data_block),
3319 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003320 goto failed_mount;
3321 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003322 blocks_count = (ext4_blocks_count(es) -
3323 le32_to_cpu(es->s_first_data_block) +
3324 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3325 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003326 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003327 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003328 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003329 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003330 ext4_blocks_count(es),
3331 le32_to_cpu(es->s_first_data_block),
3332 EXT4_BLOCKS_PER_GROUP(sb));
3333 goto failed_mount;
3334 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003335 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003336 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3337 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003338 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3339 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003340 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003341 GFP_KERNEL);
3342 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003343 ext4_msg(sb, KERN_ERR, "not enough memory");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003344 goto failed_mount;
3345 }
3346
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04003347#ifdef CONFIG_PROC_FS
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003348 if (ext4_proc_root)
3349 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Alexander Beregalov3244fcb2008-10-12 17:27:49 -04003350#endif
Theodore Ts'o240799c2008-10-09 23:53:47 -04003351
Pekka Enberg705895b2009-02-15 18:07:52 -05003352 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003353
3354 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003355 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003356 sbi->s_group_desc[i] = sb_bread(sb, block);
3357 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003358 ext4_msg(sb, KERN_ERR,
3359 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003360 db_count = i;
3361 goto failed_mount2;
3362 }
3363 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003364 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003365 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003366 goto failed_mount2;
3367 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003368 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3369 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003370 ext4_msg(sb, KERN_ERR,
3371 "unable to initialize "
3372 "flex_bg meta info!");
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003373 goto failed_mount2;
3374 }
3375
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003376 sbi->s_gdb_count = db_count;
3377 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3378 spin_lock_init(&sbi->s_next_gen_lock);
3379
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003380 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3381 ext4_count_free_blocks(sb));
3382 if (!err) {
3383 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3384 ext4_count_free_inodes(sb));
3385 }
3386 if (!err) {
3387 err = percpu_counter_init(&sbi->s_dirs_counter,
3388 ext4_count_dirs(sb));
3389 }
3390 if (!err) {
3391 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3392 }
3393 if (err) {
3394 ext4_msg(sb, KERN_ERR, "insufficient memory");
3395 goto failed_mount3;
3396 }
3397
Alex Tomasc9de5602008-01-29 00:19:52 -05003398 sbi->s_stripe = ext4_get_stripe_size(sbi);
Theodore Ts'o55138e02009-09-29 13:31:31 -04003399 sbi->s_max_writeback_mb_bump = 128;
Alex Tomasc9de5602008-01-29 00:19:52 -05003400
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003401 /*
3402 * set up enough so that it can read an inode
3403 */
Theodore Ts'o9ca92382009-05-01 12:52:25 -04003404 if (!test_opt(sb, NOLOAD) &&
3405 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3406 sb->s_op = &ext4_sops;
3407 else
3408 sb->s_op = &ext4_nojournal_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07003409 sb->s_export_op = &ext4_export_ops;
3410 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003411#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003412 sb->s_qcop = &ext4_qctl_operations;
3413 sb->dq_op = &ext4_quota_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003414#endif
3415 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003416 mutex_init(&sbi->s_orphan_lock);
Theodore Ts'o32ed5052009-04-25 22:53:39 -04003417 mutex_init(&sbi->s_resize_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003418
3419 sb->s_root = NULL;
3420
3421 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07003422 EXT4_HAS_INCOMPAT_FEATURE(sb,
3423 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003424
3425 /*
3426 * The first inode we look at is the journal inode. Don't try
3427 * root first: it may be modified in the journal!
3428 */
3429 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07003430 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3431 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003432 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05003433 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3434 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003435 ext4_msg(sb, KERN_ERR, "required journal recovery "
3436 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003437 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003438 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003439 clear_opt(sb, DATA_FLAGS);
3440 set_opt(sb, WRITEBACK_DATA);
Frank Mayhar03901312009-01-07 00:06:22 -05003441 sbi->s_journal = NULL;
3442 needs_recovery = 0;
3443 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003444 }
3445
Jose R. Santoseb40a092007-07-18 08:37:25 -04003446 if (ext4_blocks_count(es) > 0xffffffffULL &&
3447 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3448 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003449 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003450 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04003451 }
3452
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003453 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3454 jbd2_journal_set_features(sbi->s_journal,
3455 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
Girish Shilamkar818d2762008-01-28 23:58:27 -05003456 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003457 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3458 jbd2_journal_set_features(sbi->s_journal,
3459 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
Girish Shilamkar818d2762008-01-28 23:58:27 -05003460 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3461 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003462 } else {
3463 jbd2_journal_clear_features(sbi->s_journal,
3464 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3465 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3466 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05003467
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003468 /* We have now updated the journal if required, so we can
3469 * validate the data journaling mode. */
3470 switch (test_opt(sb, DATA_FLAGS)) {
3471 case 0:
3472 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07003473 * capabilities: ORDERED_DATA if the journal can
3474 * cope, else JOURNAL_DATA
3475 */
Mingming Caodab291a2006-10-11 01:21:01 -07003476 if (jbd2_journal_check_available_features
3477 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003478 set_opt(sb, ORDERED_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003479 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003480 set_opt(sb, JOURNAL_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003481 break;
3482
Mingming Cao617ba132006-10-11 01:20:53 -07003483 case EXT4_MOUNT_ORDERED_DATA:
3484 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07003485 if (!jbd2_journal_check_available_features
3486 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003487 ext4_msg(sb, KERN_ERR, "Journal does not support "
3488 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003489 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003490 }
3491 default:
3492 break;
3493 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003494 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003495
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003496 /*
3497 * The journal may have updated the bg summary counts, so we
3498 * need to update the global counters.
3499 */
3500 percpu_counter_set(&sbi->s_freeblocks_counter,
3501 ext4_count_free_blocks(sb));
3502 percpu_counter_set(&sbi->s_freeinodes_counter,
3503 ext4_count_free_inodes(sb));
3504 percpu_counter_set(&sbi->s_dirs_counter,
3505 ext4_count_dirs(sb));
3506 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04003507
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003508no_journal:
Mingming Cao4c0425f2009-09-28 15:48:41 -04003509 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3510 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3511 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3512 goto failed_mount_wq;
3513 }
3514
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003515 /*
Mingming Caodab291a2006-10-11 01:21:01 -07003516 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003517 * so we can safely mount the rest of the filesystem now.
3518 */
3519
David Howells1d1fe1e2008-02-07 00:15:37 -08003520 root = ext4_iget(sb, EXT4_ROOT_INO);
3521 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003522 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003523 ret = PTR_ERR(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003524 goto failed_mount4;
3525 }
3526 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells1d1fe1e2008-02-07 00:15:37 -08003527 iput(root);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003528 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003529 goto failed_mount4;
3530 }
David Howells1d1fe1e2008-02-07 00:15:37 -08003531 sb->s_root = d_alloc_root(root);
3532 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003533 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08003534 iput(root);
3535 ret = -ENOMEM;
3536 goto failed_mount4;
3537 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003538
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003539 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
Kalpak Shahef7f3832007-07-18 09:15:20 -04003540
3541 /* determine the minimum size of new large inodes, if present */
3542 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3543 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3544 EXT4_GOOD_OLD_INODE_SIZE;
3545 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3546 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3547 if (sbi->s_want_extra_isize <
3548 le16_to_cpu(es->s_want_extra_isize))
3549 sbi->s_want_extra_isize =
3550 le16_to_cpu(es->s_want_extra_isize);
3551 if (sbi->s_want_extra_isize <
3552 le16_to_cpu(es->s_min_extra_isize))
3553 sbi->s_want_extra_isize =
3554 le16_to_cpu(es->s_min_extra_isize);
3555 }
3556 }
3557 /* Check if enough inode space is available */
3558 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3559 sbi->s_inode_size) {
3560 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3561 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04003562 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3563 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04003564 }
3565
Theodore Ts'o90576c02009-09-29 15:51:30 -04003566 if (test_opt(sb, DELALLOC) &&
3567 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003568 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3569 "requested data journaling mode");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003570 clear_opt(sb, DELALLOC);
Theodore Ts'o90576c02009-09-29 15:51:30 -04003571 }
Jiaying Zhang744692d2010-03-04 16:14:02 -05003572 if (test_opt(sb, DIOREAD_NOLOCK)) {
3573 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3574 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3575 "option - requested data journaling mode");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003576 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05003577 }
3578 if (sb->s_blocksize < PAGE_SIZE) {
3579 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3580 "option - block size is too small");
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003581 clear_opt(sb, DIOREAD_NOLOCK);
Jiaying Zhang744692d2010-03-04 16:14:02 -05003582 }
3583 }
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003584
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003585 err = ext4_setup_system_zone(sb);
3586 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003587 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04003588 "zone (%d)", err);
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003589 goto failed_mount4;
3590 }
3591
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003592 ext4_ext_init(sb);
3593 err = ext4_mb_init(sb, needs_recovery);
3594 if (err) {
Uwe Kleine-König421f91d2010-06-11 12:17:00 +02003595 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
Eric Sandeenb31e1552009-06-04 17:36:36 -04003596 err);
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04003597 goto failed_mount4;
3598 }
3599
Lukas Czernerbfff6872010-10-27 21:30:05 -04003600 err = ext4_register_li_request(sb, first_not_zeroed);
3601 if (err)
3602 goto failed_mount4;
3603
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04003604 sbi->s_kobj.kset = ext4_kset;
3605 init_completion(&sbi->s_kobj_unregister);
3606 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3607 "%s", sb->s_id);
3608 if (err) {
3609 ext4_mb_release(sb);
3610 ext4_ext_release(sb);
3611 goto failed_mount4;
3612 };
3613
Mingming Cao617ba132006-10-11 01:20:53 -07003614 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3615 ext4_orphan_cleanup(sb, es);
3616 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05003617 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003618 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05003619 ext4_mark_recovery_complete(sb, es);
3620 }
3621 if (EXT4_SB(sb)->s_journal) {
3622 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3623 descr = " journalled data mode";
3624 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3625 descr = " ordered data mode";
3626 else
3627 descr = " writeback data mode";
3628 } else
3629 descr = "out journal";
3630
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003631 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003632 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3633 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003634
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003635 init_timer(&sbi->s_err_report);
3636 sbi->s_err_report.function = print_daily_error_info;
3637 sbi->s_err_report.data = (unsigned long) sb;
3638 if (es->s_error_count)
3639 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003640
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003641 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003642 return 0;
3643
Mingming Cao617ba132006-10-11 01:20:53 -07003644cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003645 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04003646 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003647 goto failed_mount;
3648
3649failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04003650 ext4_msg(sb, KERN_ERR, "mount failed");
Mingming Cao4c0425f2009-09-28 15:48:41 -04003651 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3652failed_mount_wq:
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04003653 ext4_release_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05003654 if (sbi->s_journal) {
3655 jbd2_journal_destroy(sbi->s_journal);
3656 sbi->s_journal = NULL;
3657 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003658failed_mount3:
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -04003659 if (sbi->s_flex_groups) {
3660 if (is_vmalloc_addr(sbi->s_flex_groups))
3661 vfree(sbi->s_flex_groups);
3662 else
3663 kfree(sbi->s_flex_groups);
3664 }
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003665 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3666 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3667 percpu_counter_destroy(&sbi->s_dirs_counter);
3668 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003669failed_mount2:
3670 for (i = 0; i < db_count; i++)
3671 brelse(sbi->s_group_desc[i]);
3672 kfree(sbi->s_group_desc);
3673failed_mount:
Theodore Ts'o240799c2008-10-09 23:53:47 -04003674 if (sbi->s_proc) {
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003675 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003676 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003677#ifdef CONFIG_QUOTA
3678 for (i = 0; i < MAXQUOTAS; i++)
3679 kfree(sbi->s_qf_names[i]);
3680#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003681 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003682 brelse(bh);
3683out_fail:
3684 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04003685 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003686 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003687out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003688 kfree(orig_data);
David Howells1d1fe1e2008-02-07 00:15:37 -08003689 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003690}
3691
3692/*
3693 * Setup any per-fs journal parameters now. We'll do this both on
3694 * initial mount, once the journal has been initialised but before we've
3695 * done any recovery; and again on any subsequent remount.
3696 */
Mingming Cao617ba132006-10-11 01:20:53 -07003697static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003698{
Mingming Cao617ba132006-10-11 01:20:53 -07003699 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003700
Theodore Ts'o30773842009-01-03 20:27:38 -05003701 journal->j_commit_interval = sbi->s_commit_interval;
3702 journal->j_min_batch_time = sbi->s_min_batch_time;
3703 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003704
Theodore Ts'oa931da62010-08-03 21:35:12 -04003705 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003706 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07003707 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003708 else
Mingming Caodab291a2006-10-11 01:21:01 -07003709 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04003710 if (test_opt(sb, DATA_ERR_ABORT))
3711 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3712 else
3713 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04003714 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003715}
3716
Mingming Cao617ba132006-10-11 01:20:53 -07003717static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003718 unsigned int journal_inum)
3719{
3720 struct inode *journal_inode;
3721 journal_t *journal;
3722
Frank Mayhar03901312009-01-07 00:06:22 -05003723 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3724
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003725 /* First, test for the existence of a valid inode on disk. Bad
3726 * things happen if we iget() an unused inode, as the subsequent
3727 * iput() will try to delete it. */
3728
David Howells1d1fe1e2008-02-07 00:15:37 -08003729 journal_inode = ext4_iget(sb, journal_inum);
3730 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003731 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003732 return NULL;
3733 }
3734 if (!journal_inode->i_nlink) {
3735 make_bad_inode(journal_inode);
3736 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04003737 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003738 return NULL;
3739 }
3740
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04003741 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003742 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08003743 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003744 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003745 iput(journal_inode);
3746 return NULL;
3747 }
3748
Mingming Caodab291a2006-10-11 01:21:01 -07003749 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003750 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003751 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003752 iput(journal_inode);
3753 return NULL;
3754 }
3755 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07003756 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003757 return journal;
3758}
3759
Mingming Cao617ba132006-10-11 01:20:53 -07003760static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003761 dev_t j_dev)
3762{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003763 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003764 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07003765 ext4_fsblk_t start;
3766 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003767 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07003768 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003769 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003770 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003771 struct block_device *bdev;
3772
Frank Mayhar03901312009-01-07 00:06:22 -05003773 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3774
Eric Sandeenb31e1552009-06-04 17:36:36 -04003775 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003776 if (bdev == NULL)
3777 return NULL;
3778
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003779 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04003780 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003781 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003782 ext4_msg(sb, KERN_ERR,
3783 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003784 goto out_bdev;
3785 }
3786
Mingming Cao617ba132006-10-11 01:20:53 -07003787 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3788 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003789 set_blocksize(bdev, blocksize);
3790 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003791 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3792 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003793 goto out_bdev;
3794 }
3795
Mingming Cao617ba132006-10-11 01:20:53 -07003796 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3797 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003798 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07003799 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003800 ext4_msg(sb, KERN_ERR, "external journal has "
3801 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003802 brelse(bh);
3803 goto out_bdev;
3804 }
3805
Mingming Cao617ba132006-10-11 01:20:53 -07003806 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003807 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003808 brelse(bh);
3809 goto out_bdev;
3810 }
3811
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003812 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003813 start = sb_block + 1;
3814 brelse(bh); /* we're done with the superblock */
3815
Mingming Caodab291a2006-10-11 01:21:01 -07003816 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003817 start, len, blocksize);
3818 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003819 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003820 goto out_bdev;
3821 }
3822 journal->j_private = sb;
3823 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3824 wait_on_buffer(journal->j_sb_buffer);
3825 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003826 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003827 goto out_journal;
3828 }
3829 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003830 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3831 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003832 be32_to_cpu(journal->j_superblock->s_nr_users));
3833 goto out_journal;
3834 }
Mingming Cao617ba132006-10-11 01:20:53 -07003835 EXT4_SB(sb)->journal_bdev = bdev;
3836 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003837 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003838
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003839out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07003840 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003841out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07003842 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003843 return NULL;
3844}
3845
Mingming Cao617ba132006-10-11 01:20:53 -07003846static int ext4_load_journal(struct super_block *sb,
3847 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003848 unsigned long journal_devnum)
3849{
3850 journal_t *journal;
3851 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3852 dev_t journal_dev;
3853 int err = 0;
3854 int really_read_only;
3855
Frank Mayhar03901312009-01-07 00:06:22 -05003856 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3857
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003858 if (journal_devnum &&
3859 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003860 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3861 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003862 journal_dev = new_decode_dev(journal_devnum);
3863 } else
3864 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3865
3866 really_read_only = bdev_read_only(sb->s_bdev);
3867
3868 /*
3869 * Are we loading a blank journal or performing recovery after a
3870 * crash? For recovery, we need to check in advance whether we
3871 * can get read-write access to the device.
3872 */
Mingming Cao617ba132006-10-11 01:20:53 -07003873 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003874 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003875 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3876 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003877 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003878 ext4_msg(sb, KERN_ERR, "write access "
3879 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003880 return -EROFS;
3881 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04003882 ext4_msg(sb, KERN_INFO, "write access will "
3883 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003884 }
3885 }
3886
3887 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003888 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3889 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003890 return -EINVAL;
3891 }
3892
3893 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07003894 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003895 return -EINVAL;
3896 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07003897 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003898 return -EINVAL;
3899 }
3900
Theodore Ts'o90576c02009-09-29 15:51:30 -04003901 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003902 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04003903
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003904 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
Mingming Caodab291a2006-10-11 01:21:01 -07003905 err = jbd2_journal_update_format(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003906 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003907 ext4_msg(sb, KERN_ERR, "error updating journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003908 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003909 return err;
3910 }
3911 }
3912
Mingming Cao617ba132006-10-11 01:20:53 -07003913 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07003914 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04003915 if (!err) {
3916 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3917 if (save)
3918 memcpy(save, ((char *) es) +
3919 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07003920 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04003921 if (save)
3922 memcpy(((char *) es) + EXT4_S_ERR_START,
3923 save, EXT4_S_ERR_LEN);
3924 kfree(save);
3925 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003926
3927 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003928 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07003929 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003930 return err;
3931 }
3932
Mingming Cao617ba132006-10-11 01:20:53 -07003933 EXT4_SB(sb)->s_journal = journal;
3934 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003935
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04003936 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003937 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3938 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003939
3940 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003941 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003942 }
3943
3944 return 0;
3945}
3946
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003947static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003948{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04003949 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07003950 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08003951 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003952
3953 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08003954 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04003955 if (buffer_write_io_error(sbh)) {
3956 /*
3957 * Oh, dear. A previous attempt to write the
3958 * superblock failed. This could happen because the
3959 * USB device was yanked out. Or it could happen to
3960 * be a transient write error and maybe the block will
3961 * be remapped. Nothing we can do but to retry the
3962 * write and hope for the best.
3963 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04003964 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3965 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04003966 clear_buffer_write_io_error(sbh);
3967 set_buffer_uptodate(sbh);
3968 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04003969 /*
3970 * If the file system is mounted read-only, don't update the
3971 * superblock write time. This avoids updating the superblock
3972 * write time when we are mounting the root file system
3973 * read/only but we need to replay the journal; at that point,
3974 * for people who are east of GMT and who make their clock
3975 * tick in localtime for Windows bug-for-bug compatibility,
3976 * the clock is set in the future, and this will cause e2fsck
3977 * to complain and force a full file system check.
3978 */
3979 if (!(sb->s_flags & MS_RDONLY))
3980 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003981 if (sb->s_bdev->bd_part)
3982 es->s_kbytes_written =
3983 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003984 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3985 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003986 else
3987 es->s_kbytes_written =
3988 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003989 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3990 &EXT4_SB(sb)->s_freeblocks_counter));
3991 es->s_free_inodes_count =
3992 cpu_to_le32(percpu_counter_sum_positive(
3993 &EXT4_SB(sb)->s_freeinodes_counter));
Theodore Ts'o7234ab22009-04-30 21:24:04 -04003994 sb->s_dirt = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003995 BUFFER_TRACE(sbh, "marking dirty");
3996 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04003997 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08003998 error = sync_dirty_buffer(sbh);
3999 if (error)
4000 return error;
4001
4002 error = buffer_write_io_error(sbh);
4003 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004004 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4005 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004006 clear_buffer_write_io_error(sbh);
4007 set_buffer_uptodate(sbh);
4008 }
4009 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08004010 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004011}
4012
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004013/*
4014 * Have we just finished recovery? If so, and if we are mounting (or
4015 * remounting) the filesystem readonly, then we will end up with a
4016 * consistent fs on disk. Record that fact.
4017 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004018static void ext4_mark_recovery_complete(struct super_block *sb,
4019 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004020{
Mingming Cao617ba132006-10-11 01:20:53 -07004021 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004022
Frank Mayhar03901312009-01-07 00:06:22 -05004023 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4024 BUG_ON(journal != NULL);
4025 return;
4026 }
Mingming Caodab291a2006-10-11 01:21:01 -07004027 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004028 if (jbd2_journal_flush(journal) < 0)
4029 goto out;
4030
Mingming Cao617ba132006-10-11 01:20:53 -07004031 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004032 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07004033 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004034 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004035 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004036
4037out:
Mingming Caodab291a2006-10-11 01:21:01 -07004038 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004039}
4040
4041/*
4042 * If we are mounting (or read-write remounting) a filesystem whose journal
4043 * has recorded an error from a previous lifetime, move that error to the
4044 * main filesystem now.
4045 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004046static void ext4_clear_journal_err(struct super_block *sb,
4047 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004048{
4049 journal_t *journal;
4050 int j_errno;
4051 const char *errstr;
4052
Frank Mayhar03901312009-01-07 00:06:22 -05004053 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4054
Mingming Cao617ba132006-10-11 01:20:53 -07004055 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004056
4057 /*
4058 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004059 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004060 */
4061
Mingming Caodab291a2006-10-11 01:21:01 -07004062 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004063 if (j_errno) {
4064 char nbuf[16];
4065
Mingming Cao617ba132006-10-11 01:20:53 -07004066 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004067 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004068 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004069 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004070
Mingming Cao617ba132006-10-11 01:20:53 -07004071 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4072 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004073 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004074
Mingming Caodab291a2006-10-11 01:21:01 -07004075 jbd2_journal_clear_err(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004076 }
4077}
4078
4079/*
4080 * Force the running and committing transactions to commit,
4081 * and wait on the commit.
4082 */
Mingming Cao617ba132006-10-11 01:20:53 -07004083int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004084{
4085 journal_t *journal;
Frank Mayhar03901312009-01-07 00:06:22 -05004086 int ret = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004087
4088 if (sb->s_flags & MS_RDONLY)
4089 return 0;
4090
Mingming Cao617ba132006-10-11 01:20:53 -07004091 journal = EXT4_SB(sb)->s_journal;
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004092 if (journal) {
Eric Sandeen437f88c2010-08-01 17:33:29 -04004093 vfs_check_frozen(sb, SB_FREEZE_TRANS);
Frank Mayhar03901312009-01-07 00:06:22 -05004094 ret = ext4_journal_force_commit(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004095 }
Frank Mayhar03901312009-01-07 00:06:22 -05004096
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004097 return ret;
4098}
4099
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004100static void ext4_write_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004101{
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004102 lock_super(sb);
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004103 ext4_commit_super(sb, 1);
Christoph Hellwigebc1ac12009-05-11 23:35:03 +02004104 unlock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004105}
4106
Mingming Cao617ba132006-10-11 01:20:53 -07004107static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004108{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004109 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004110 tid_t target;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004111 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004112
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004113 trace_ext4_sync_fs(sb, wait);
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004114 flush_workqueue(sbi->dio_unwritten_wq);
4115 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004116 if (wait)
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004117 jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05004118 }
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004119 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004120}
4121
4122/*
4123 * LVM calls this function before a (read-only) snapshot is created. This
4124 * gives us a chance to flush the journal completely and mark the fs clean.
4125 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004126static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004127{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004128 int error = 0;
4129 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004130
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004131 if (sb->s_flags & MS_RDONLY)
4132 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004133
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004134 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004135
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004136 /* Now we set up the journal barrier. */
4137 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004138
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004139 /*
4140 * Don't clear the needs_recovery flag if we failed to flush
4141 * the journal.
4142 */
4143 error = jbd2_journal_flush(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004144 if (error < 0)
4145 goto out;
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004146
4147 /* Journal blocked and flushed, clear needs_recovery flag. */
4148 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4149 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004150out:
4151 /* we rely on s_frozen to stop further updates */
4152 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4153 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004154}
4155
4156/*
4157 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4158 * flag here, even though the filesystem is not technically dirty yet.
4159 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004160static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004161{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004162 if (sb->s_flags & MS_RDONLY)
4163 return 0;
4164
4165 lock_super(sb);
4166 /* Reset the needs_recovery flag before the fs is unlocked. */
4167 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4168 ext4_commit_super(sb, 1);
4169 unlock_super(sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004170 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004171}
4172
Theodore Ts'o673c6102010-12-15 20:28:48 -05004173/*
4174 * Structure to save mount options for ext4_remount's benefit
4175 */
4176struct ext4_mount_options {
4177 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004178 unsigned long s_mount_opt2;
Theodore Ts'o673c6102010-12-15 20:28:48 -05004179 uid_t s_resuid;
4180 gid_t s_resgid;
4181 unsigned long s_commit_interval;
4182 u32 s_min_batch_time, s_max_batch_time;
4183#ifdef CONFIG_QUOTA
4184 int s_jquota_fmt;
4185 char *s_qf_names[MAXQUOTAS];
4186#endif
4187};
4188
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004189static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004190{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004191 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004192 struct ext4_sb_info *sbi = EXT4_SB(sb);
4193 ext4_fsblk_t n_blocks_count = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004194 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004195 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004196 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004197 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004198 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004199 int err;
4200#ifdef CONFIG_QUOTA
4201 int i;
4202#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004203 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004204
4205 /* Store the original options */
Al Virobbd68512009-05-06 10:43:07 -04004206 lock_super(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004207 old_sb_flags = sb->s_flags;
4208 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004209 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004210 old_opts.s_resuid = sbi->s_resuid;
4211 old_opts.s_resgid = sbi->s_resgid;
4212 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004213 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4214 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004215#ifdef CONFIG_QUOTA
4216 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4217 for (i = 0; i < MAXQUOTAS; i++)
4218 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4219#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004220 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4221 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004222
4223 /*
4224 * Allow the "check" option to be passed as a remount option.
4225 */
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004226 if (!parse_options(data, sb, NULL, &journal_ioprio,
4227 &n_blocks_count, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004228 err = -EINVAL;
4229 goto restore_opts;
4230 }
4231
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004232 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004233 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004234
4235 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004236 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004237
4238 es = sbi->s_es;
4239
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004240 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004241 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004242 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4243 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004244
4245 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004246 n_blocks_count > ext4_blocks_count(es)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004247 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004248 err = -EROFS;
4249 goto restore_opts;
4250 }
4251
4252 if (*flags & MS_RDONLY) {
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004253 err = dquot_suspend(sb, -1);
4254 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004255 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004256
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004257 /*
4258 * First of all, the unconditional stuff we have to do
4259 * to disable replay of the journal when we next remount
4260 */
4261 sb->s_flags |= MS_RDONLY;
4262
4263 /*
4264 * OK, test if we are remounting a valid rw partition
4265 * readonly, and if so set the rdonly flag and then
4266 * mark the partition as valid again.
4267 */
Mingming Cao617ba132006-10-11 01:20:53 -07004268 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4269 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004270 es->s_state = cpu_to_le16(sbi->s_mount_state);
4271
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004272 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004273 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004274 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004275 /* Make sure we can mount this feature set readwrite */
4276 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004277 err = -EROFS;
4278 goto restore_opts;
4279 }
Eric Sandeenead65962007-02-10 01:46:08 -08004280 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004281 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004282 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004283 */
4284 for (g = 0; g < sbi->s_groups_count; g++) {
4285 struct ext4_group_desc *gdp =
4286 ext4_get_group_desc(sb, g, NULL);
4287
4288 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004289 ext4_msg(sb, KERN_ERR,
4290 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04004291 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4292 le16_to_cpu(gdp->bg_checksum));
4293 err = -EINVAL;
4294 goto restore_opts;
4295 }
4296 }
4297
4298 /*
Eric Sandeenead65962007-02-10 01:46:08 -08004299 * If we have an unprocessed orphan list hanging
4300 * around from a previously readonly bdev mount,
4301 * require a full umount/remount for now.
4302 */
4303 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004304 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08004305 "remount RDWR because of unprocessed "
4306 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004307 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08004308 err = -EINVAL;
4309 goto restore_opts;
4310 }
4311
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004312 /*
4313 * Mounting a RDONLY partition read-write, so reread
4314 * and store the current valid flag. (It may have
4315 * been changed by e2fsck since we originally mounted
4316 * the partition.)
4317 */
Frank Mayhar03901312009-01-07 00:06:22 -05004318 if (sbi->s_journal)
4319 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004320 sbi->s_mount_state = le16_to_cpu(es->s_state);
Mingming Cao617ba132006-10-11 01:20:53 -07004321 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004322 goto restore_opts;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004323 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004324 sb->s_flags &= ~MS_RDONLY;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004325 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004326 }
4327 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04004328
4329 /*
4330 * Reinitialize lazy itable initialization thread based on
4331 * current settings
4332 */
4333 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4334 ext4_unregister_li_request(sb);
4335 else {
4336 ext4_group_t first_not_zeroed;
4337 first_not_zeroed = ext4_has_uninit_itable(sb);
4338 ext4_register_li_request(sb, first_not_zeroed);
4339 }
4340
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004341 ext4_setup_system_zone(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05004342 if (sbi->s_journal == NULL)
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004343 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05004344
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004345#ifdef CONFIG_QUOTA
4346 /* Release old quota file names */
4347 for (i = 0; i < MAXQUOTAS; i++)
4348 if (old_opts.s_qf_names[i] &&
4349 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4350 kfree(old_opts.s_qf_names[i]);
4351#endif
Al Virobbd68512009-05-06 10:43:07 -04004352 unlock_super(sb);
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004353 if (enable_quota)
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004354 dquot_resume(sb, -1);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004355
4356 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4357 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004358 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004359
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004360restore_opts:
4361 sb->s_flags = old_sb_flags;
4362 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004363 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004364 sbi->s_resuid = old_opts.s_resuid;
4365 sbi->s_resgid = old_opts.s_resgid;
4366 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004367 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4368 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004369#ifdef CONFIG_QUOTA
4370 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4371 for (i = 0; i < MAXQUOTAS; i++) {
4372 if (sbi->s_qf_names[i] &&
4373 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4374 kfree(sbi->s_qf_names[i]);
4375 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4376 }
4377#endif
Al Virobbd68512009-05-06 10:43:07 -04004378 unlock_super(sb);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004379 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004380 return err;
4381}
4382
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004383static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004384{
4385 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004386 struct ext4_sb_info *sbi = EXT4_SB(sb);
4387 struct ext4_super_block *es = sbi->s_es;
Pekka Enberg960cc392006-12-06 20:35:29 -08004388 u64 fsid;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004389
Badari Pulavarty5e700302007-07-15 23:42:00 -07004390 if (test_opt(sb, MINIX_DF)) {
4391 sbi->s_overhead_last = 0;
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004392 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
Theodore Ts'o8df96752009-05-01 08:50:38 -04004393 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
Badari Pulavarty5e700302007-07-15 23:42:00 -07004394 ext4_fsblk_t overhead = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004395
4396 /*
Badari Pulavarty5e700302007-07-15 23:42:00 -07004397 * Compute the overhead (FS structures). This is constant
4398 * for a given filesystem unless the number of block groups
4399 * changes so we cache the previous value until it does.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004400 */
4401
4402 /*
4403 * All of the blocks before first_data_block are
4404 * overhead
4405 */
4406 overhead = le32_to_cpu(es->s_first_data_block);
4407
4408 /*
4409 * Add the overhead attributed to the superblock and
4410 * block group descriptors. If the sparse superblocks
4411 * feature is turned on, then not all groups have this.
4412 */
4413 for (i = 0; i < ngroups; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07004414 overhead += ext4_bg_has_super(sb, i) +
4415 ext4_bg_num_gdb(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004416 cond_resched();
4417 }
4418
4419 /*
4420 * Every block group has an inode bitmap, a block
4421 * bitmap, and an inode table.
4422 */
Badari Pulavarty5e700302007-07-15 23:42:00 -07004423 overhead += ngroups * (2 + sbi->s_itb_per_group);
4424 sbi->s_overhead_last = overhead;
4425 smp_wmb();
Aneesh Kumar K.V6bc9fef2007-10-16 18:38:25 -04004426 sbi->s_blocks_last = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004427 }
4428
Mingming Cao617ba132006-10-11 01:20:53 -07004429 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004430 buf->f_bsize = sb->s_blocksize;
Badari Pulavarty5e700302007-07-15 23:42:00 -07004431 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
Aneesh Kumar K.V6bc6e632008-10-10 09:39:00 -04004432 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4433 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004434 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4435 if (buf->f_bfree < ext4_r_blocks_count(es))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004436 buf->f_bavail = 0;
4437 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07004438 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07004439 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08004440 fsid = le64_to_cpup((void *)es->s_uuid) ^
4441 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4442 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4443 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004444
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004445 return 0;
4446}
4447
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004448/* Helper function for writing quotas on sync - we need to start transaction
4449 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004450 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07004451 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01004452 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05004453 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07004454 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004455 *
4456 */
4457
4458#ifdef CONFIG_QUOTA
4459
4460static inline struct inode *dquot_to_inode(struct dquot *dquot)
4461{
4462 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4463}
4464
Mingming Cao617ba132006-10-11 01:20:53 -07004465static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004466{
4467 int ret, err;
4468 handle_t *handle;
4469 struct inode *inode;
4470
4471 inode = dquot_to_inode(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004472 handle = ext4_journal_start(inode,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004473 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004474 if (IS_ERR(handle))
4475 return PTR_ERR(handle);
4476 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004477 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004478 if (!ret)
4479 ret = err;
4480 return ret;
4481}
4482
Mingming Cao617ba132006-10-11 01:20:53 -07004483static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004484{
4485 int ret, err;
4486 handle_t *handle;
4487
Mingming Cao617ba132006-10-11 01:20:53 -07004488 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004489 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004490 if (IS_ERR(handle))
4491 return PTR_ERR(handle);
4492 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004493 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004494 if (!ret)
4495 ret = err;
4496 return ret;
4497}
4498
Mingming Cao617ba132006-10-11 01:20:53 -07004499static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004500{
4501 int ret, err;
4502 handle_t *handle;
4503
Mingming Cao617ba132006-10-11 01:20:53 -07004504 handle = ext4_journal_start(dquot_to_inode(dquot),
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004505 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07004506 if (IS_ERR(handle)) {
4507 /* Release dquot anyway to avoid endless cycle in dqput() */
4508 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004509 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07004510 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004511 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004512 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004513 if (!ret)
4514 ret = err;
4515 return ret;
4516}
4517
Mingming Cao617ba132006-10-11 01:20:53 -07004518static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004519{
Jan Kara2c8be6b2008-05-13 21:27:55 -04004520 /* Are we journaling quotas? */
Mingming Cao617ba132006-10-11 01:20:53 -07004521 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4522 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004523 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07004524 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004525 } else {
4526 return dquot_mark_dquot_dirty(dquot);
4527 }
4528}
4529
Mingming Cao617ba132006-10-11 01:20:53 -07004530static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004531{
4532 int ret, err;
4533 handle_t *handle;
4534
4535 /* Data block + inode block */
Mingming Cao617ba132006-10-11 01:20:53 -07004536 handle = ext4_journal_start(sb->s_root->d_inode, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004537 if (IS_ERR(handle))
4538 return PTR_ERR(handle);
4539 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07004540 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004541 if (!ret)
4542 ret = err;
4543 return ret;
4544}
4545
4546/*
4547 * Turn on quotas during mount time - we need to find
4548 * the quota file and such...
4549 */
Mingming Cao617ba132006-10-11 01:20:53 -07004550static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004551{
Christoph Hellwig287a8092010-05-19 07:16:45 -04004552 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4553 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004554}
4555
4556/*
4557 * Standard function to be called on quota_on
4558 */
Mingming Cao617ba132006-10-11 01:20:53 -07004559static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02004560 struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004561{
4562 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004563
4564 if (!test_opt(sb, QUOTA))
4565 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04004566
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004567 /* Quotafile not on the same filesystem? */
Jan Karaf00c9e42010-09-15 17:38:58 +02004568 if (path->mnt->mnt_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004569 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04004570 /* Journaling quota? */
4571 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004572 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02004573 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004574 ext4_msg(sb, KERN_WARNING,
4575 "Quota file not on filesystem root. "
4576 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004577 }
Jan Kara06235432008-05-13 19:11:51 -04004578
4579 /*
4580 * When we journal data on quota file, we have to flush journal to see
4581 * all updates to the file when we bypass pagecache...
4582 */
Frank Mayhar03901312009-01-07 00:06:22 -05004583 if (EXT4_SB(sb)->s_journal &&
Jan Karaf00c9e42010-09-15 17:38:58 +02004584 ext4_should_journal_data(path->dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04004585 /*
4586 * We don't need to lock updates but journal_flush() could
4587 * otherwise be livelocked...
4588 */
4589 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004590 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04004591 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02004592 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004593 return err;
Jan Kara06235432008-05-13 19:11:51 -04004594 }
4595
Jan Karaf00c9e42010-09-15 17:38:58 +02004596 return dquot_quota_on(sb, type, format_id, path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004597}
4598
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004599static int ext4_quota_off(struct super_block *sb, int type)
4600{
Dmitry Monakhov87009d82010-11-08 13:47:33 -05004601 /* Force all delayed allocation blocks to be allocated.
4602 * Caller already holds s_umount sem */
4603 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004604 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04004605
4606 return dquot_quota_off(sb, type);
4607}
4608
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004609/* Read data from quotafile - avoid pagecache and such because we cannot afford
4610 * acquiring the locks... As quota files are never truncated and quota code
4611 * itself serializes the operations (and noone else should touch the files)
4612 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07004613static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004614 size_t len, loff_t off)
4615{
4616 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004617 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004618 int err = 0;
4619 int offset = off & (sb->s_blocksize - 1);
4620 int tocopy;
4621 size_t toread;
4622 struct buffer_head *bh;
4623 loff_t i_size = i_size_read(inode);
4624
4625 if (off > i_size)
4626 return 0;
4627 if (off+len > i_size)
4628 len = i_size-off;
4629 toread = len;
4630 while (toread > 0) {
4631 tocopy = sb->s_blocksize - offset < toread ?
4632 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07004633 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004634 if (err)
4635 return err;
4636 if (!bh) /* A hole? */
4637 memset(data, 0, tocopy);
4638 else
4639 memcpy(data, bh->b_data+offset, tocopy);
4640 brelse(bh);
4641 offset = 0;
4642 toread -= tocopy;
4643 data += tocopy;
4644 blk++;
4645 }
4646 return len;
4647}
4648
4649/* Write to quotafile (we know the transaction is already started and has
4650 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07004651static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004652 const char *data, size_t len, loff_t off)
4653{
4654 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05004655 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004656 int err = 0;
4657 int offset = off & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004658 struct buffer_head *bh;
4659 handle_t *handle = journal_current_handle();
4660
Frank Mayhar03901312009-01-07 00:06:22 -05004661 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004662 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4663 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07004664 (unsigned long long)off, (unsigned long long)len);
4665 return -EIO;
4666 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004667 /*
4668 * Since we account only one data block in transaction credits,
4669 * then it is impossible to cross a block boundary.
4670 */
4671 if (sb->s_blocksize - offset < len) {
4672 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4673 " cancelled because not block aligned",
4674 (unsigned long long)off, (unsigned long long)len);
4675 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004676 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004677
4678 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4679 bh = ext4_bread(handle, inode, blk, 1, &err);
4680 if (!bh)
4681 goto out;
Jan Kara62d2b5f2010-07-27 11:56:07 -04004682 err = ext4_journal_get_write_access(handle, bh);
4683 if (err) {
4684 brelse(bh);
4685 goto out;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004686 }
4687 lock_buffer(bh);
4688 memcpy(bh->b_data+offset, data, len);
4689 flush_dcache_page(bh->b_page);
4690 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04004691 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004692 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004693out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004694 if (err) {
Jan Kara4d04e4f2008-07-04 09:59:34 -07004695 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004696 return err;
Jan Kara4d04e4f2008-07-04 09:59:34 -07004697 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004698 if (inode->i_size < off + len) {
4699 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07004700 EXT4_I(inode)->i_disksize = inode->i_size;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004701 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004702 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Mingming Cao617ba132006-10-11 01:20:53 -07004703 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004704 mutex_unlock(&inode->i_mutex);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05004705 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004706}
4707
4708#endif
4709
Al Viro152a0832010-07-25 00:46:55 +04004710static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4711 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004712{
Al Viro152a0832010-07-25 00:46:55 +04004713 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004714}
4715
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004716#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004717static struct file_system_type ext2_fs_type = {
4718 .owner = THIS_MODULE,
4719 .name = "ext2",
Al Viro152a0832010-07-25 00:46:55 +04004720 .mount = ext4_mount,
Theodore Ts'o24b58422009-12-07 14:08:51 -05004721 .kill_sb = kill_block_super,
4722 .fs_flags = FS_REQUIRES_DEV,
4723};
4724
4725static inline void register_as_ext2(void)
4726{
4727 int err = register_filesystem(&ext2_fs_type);
4728 if (err)
4729 printk(KERN_WARNING
4730 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4731}
4732
4733static inline void unregister_as_ext2(void)
4734{
4735 unregister_filesystem(&ext2_fs_type);
4736}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004737MODULE_ALIAS("ext2");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004738#else
4739static inline void register_as_ext2(void) { }
4740static inline void unregister_as_ext2(void) { }
4741#endif
4742
Theodore Ts'o37f328e2010-03-24 20:06:41 -04004743#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05004744static inline void register_as_ext3(void)
4745{
4746 int err = register_filesystem(&ext3_fs_type);
4747 if (err)
4748 printk(KERN_WARNING
4749 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4750}
4751
4752static inline void unregister_as_ext3(void)
4753{
4754 unregister_filesystem(&ext3_fs_type);
4755}
Theodore Ts'o51b7e3c2009-12-21 10:56:09 -05004756MODULE_ALIAS("ext3");
Theodore Ts'o24b58422009-12-07 14:08:51 -05004757#else
4758static inline void register_as_ext3(void) { }
4759static inline void unregister_as_ext3(void) { }
4760#endif
4761
Theodore Ts'o03010a32008-10-10 20:02:48 -04004762static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004763 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04004764 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04004765 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004766 .kill_sb = kill_block_super,
4767 .fs_flags = FS_REQUIRES_DEV,
4768};
4769
Lukas Czerner8f021222011-02-03 14:33:33 -05004770static int __init ext4_init_feat_adverts(void)
Lukas Czerner857ac882010-10-27 21:30:05 -04004771{
4772 struct ext4_features *ef;
4773 int ret = -ENOMEM;
4774
4775 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4776 if (!ef)
4777 goto out;
4778
4779 ef->f_kobj.kset = ext4_kset;
4780 init_completion(&ef->f_kobj_unregister);
4781 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4782 "features");
4783 if (ret) {
4784 kfree(ef);
4785 goto out;
4786 }
4787
4788 ext4_feat = ef;
4789 ret = 0;
4790out:
4791 return ret;
4792}
4793
Lukas Czerner8f021222011-02-03 14:33:33 -05004794static void ext4_exit_feat_adverts(void)
4795{
4796 kobject_put(&ext4_feat->f_kobj);
4797 wait_for_completion(&ext4_feat->f_kobj_unregister);
4798 kfree(ext4_feat);
4799}
4800
Eric Sandeene9e3bce2011-02-12 08:17:34 -05004801/* Shared across all ext4 file systems */
4802wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4803struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4804
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004805static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004806{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05004807 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05004808
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04004809 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05004810
4811 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4812 mutex_init(&ext4__aio_mutex[i]);
4813 init_waitqueue_head(&ext4__ioend_wq[i]);
4814 }
4815
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004816 err = ext4_init_pageio();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004817 if (err)
4818 return err;
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004819 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04004820 if (err)
Theodore Ts'odd683142011-02-03 14:33:49 -05004821 goto out7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004822 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4823 if (!ext4_kset)
Theodore Ts'odd683142011-02-03 14:33:49 -05004824 goto out6;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004825 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Theodore Ts'odd683142011-02-03 14:33:49 -05004826 if (!ext4_proc_root)
4827 goto out5;
Lukas Czerner857ac882010-10-27 21:30:05 -04004828
4829 err = ext4_init_feat_adverts();
Theodore Ts'odd683142011-02-03 14:33:49 -05004830 if (err)
4831 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04004832
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004833 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004834 if (err)
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004835 goto out3;
Alex Tomasc9de5602008-01-29 00:19:52 -05004836
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004837 err = ext4_init_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004838 if (err)
4839 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004840 err = init_inodecache();
4841 if (err)
4842 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05004843 register_as_ext2();
4844 register_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004845 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004846 if (err)
4847 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004848
4849 ext4_li_info = NULL;
4850 mutex_init(&ext4_li_mtx);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004851 return 0;
4852out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05004853 unregister_as_ext2();
4854 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004855 destroy_inodecache();
4856out1:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004857 ext4_exit_xattr();
Alex Tomasc9de5602008-01-29 00:19:52 -05004858out2:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004859 ext4_exit_mballoc();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004860out3:
Lukas Czerner8f021222011-02-03 14:33:33 -05004861 ext4_exit_feat_adverts();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004862out4:
Theodore Ts'odd683142011-02-03 14:33:49 -05004863 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'obd2d0212010-10-27 21:30:10 -04004864out5:
Theodore Ts'odd683142011-02-03 14:33:49 -05004865 kset_unregister(ext4_kset);
4866out6:
4867 ext4_exit_system_zone();
4868out7:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004869 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004870 return err;
4871}
4872
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004873static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004874{
Lukas Czernerbfff6872010-10-27 21:30:05 -04004875 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05004876 unregister_as_ext2();
4877 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04004878 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004879 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004880 ext4_exit_xattr();
4881 ext4_exit_mballoc();
Lukas Czerner8f021222011-02-03 14:33:33 -05004882 ext4_exit_feat_adverts();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004883 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004884 kset_unregister(ext4_kset);
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004885 ext4_exit_system_zone();
4886 ext4_exit_pageio();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004887}
4888
4889MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05004890MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004891MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04004892module_init(ext4_init_fs)
4893module_exit(ext4_exit_fs)