blob: b9b9aabfb4d2403e67565ef1b49992e3ae565407 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040023#include <linux/vmalloc.h>
Mingming Caodab291a2006-10-11 01:21:01 -070024#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070025#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070030#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040037#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040038#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040039#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040040#include <linux/crc16.h>
Dan Magenheimer7abc52c2011-05-26 10:02:03 -060041#include <linux/cleancache.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070042#include <asm/uaccess.h>
43
Lukas Czernerbfff6872010-10-27 21:30:05 -040044#include <linux/kthread.h>
45#include <linux/freezer.h>
46
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040047#include "ext4.h"
Theodore Ts'o4a092d72012-11-28 13:03:30 -050048#include "ext4_extents.h" /* Needed for trace points definition */
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040049#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070050#include "xattr.h"
51#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040052#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070053
Theodore Ts'o9bffad12009-06-17 11:48:11 -040054#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
Theodore Ts'o1f109d52010-10-27 21:30:14 -040057static struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040058static struct kset *ext4_kset;
Lukas Czerner0b75a842011-02-23 12:22:49 -050059static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
T Makphaibulchoke9c191f72014-03-18 19:24:49 -040062static int ext4_mballoc_ready;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040063
Mingming Cao617ba132006-10-11 01:20:53 -070064static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070065 unsigned long journal_devnum);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -050066static int ext4_show_options(struct seq_file *seq, struct dentry *root);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040067static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040068static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070072static int ext4_sync_fs(struct super_block *sb, int wait);
Dmitry Monakhov06a407f2013-06-12 22:25:07 -040073static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040074static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080076static int ext4_unfreeze(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080077static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040078static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
Theodore Ts'o2035e772011-04-18 17:29:14 -040080static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
Amir Goldsteind39195c2011-02-28 00:53:45 -050082static int ext4_feature_set_ok(struct super_block *sb, int readonly);
Lukas Czernerbfff6872010-10-27 21:30:05 -040083static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050085static void ext4_clear_request_list(void);
Lukas Czerner27dd4382013-04-09 22:11:22 -040086static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070087
Theodore Ts'o2035e772011-04-18 17:29:14 -040088#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95};
Eric W. Biederman7f78e032013-03-02 19:39:14 -080096MODULE_ALIAS_FS("ext2");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -070097MODULE_ALIAS("ext2");
Theodore Ts'o2035e772011-04-18 17:29:14 -040098#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99#else
100#define IS_EXT2_SB(sb) (0)
101#endif
102
103
Jan Karaba69f9a2010-03-24 20:18:37 -0400104#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
105static struct file_system_type ext3_fs_type = {
106 .owner = THIS_MODULE,
107 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +0400108 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -0400109 .kill_sb = kill_block_super,
110 .fs_flags = FS_REQUIRES_DEV,
111};
Eric W. Biederman7f78e032013-03-02 19:39:14 -0800112MODULE_ALIAS_FS("ext3");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -0700113MODULE_ALIAS("ext3");
Jan Karaba69f9a2010-03-24 20:18:37 -0400114#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115#else
116#define IS_EXT3_SB(sb) (0)
117#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700118
Darrick J. Wongd25425f2012-04-29 18:25:10 -0400119static int ext4_verify_csum_type(struct super_block *sb,
120 struct ext4_super_block *es)
121{
122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return 1;
125
126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127}
128
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400129static __le32 ext4_superblock_csum(struct super_block *sb,
130 struct ext4_super_block *es)
131{
132 struct ext4_sb_info *sbi = EXT4_SB(sb);
133 int offset = offsetof(struct ext4_super_block, s_checksum);
134 __u32 csum;
135
136 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137
138 return cpu_to_le32(csum);
139}
140
Stephen Hemmingerc1978552014-05-12 10:50:23 -0400141static int ext4_superblock_csum_verify(struct super_block *sb,
142 struct ext4_super_block *es)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400143{
144 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
146 return 1;
147
148 return es->s_checksum == ext4_superblock_csum(sb, es);
149}
150
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400151void ext4_superblock_csum_set(struct super_block *sb)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400152{
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400153 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
154
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
157 return;
158
159 es->s_checksum = ext4_superblock_csum(sb, es);
160}
161
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400162void *ext4_kvmalloc(size_t size, gfp_t flags)
163{
164 void *ret;
165
Joe Perches8be04b92013-06-19 12:15:53 -0700166 ret = kmalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400167 if (!ret)
168 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 return ret;
170}
171
172void *ext4_kvzalloc(size_t size, gfp_t flags)
173{
174 void *ret;
175
Joe Perches8be04b92013-06-19 12:15:53 -0700176 ret = kzalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400177 if (!ret)
178 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 return ret;
180}
181
182void ext4_kvfree(void *ptr)
183{
184 if (is_vmalloc_addr(ptr))
185 vfree(ptr);
186 else
187 kfree(ptr);
188
189}
190
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700191ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
192 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700193{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400194 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400196 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700197}
198
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700199ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700201{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400202 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400204 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700205}
206
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700207ext4_fsblk_t ext4_inode_table(struct super_block *sb,
208 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700209{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400210 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700213}
214
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400215__u32 ext4_free_group_clusters(struct super_block *sb,
216 struct ext4_group_desc *bg)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500217{
218 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400220 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500221}
222
223__u32 ext4_free_inodes_count(struct super_block *sb,
224 struct ext4_group_desc *bg)
225{
226 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400228 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500229}
230
231__u32 ext4_used_dirs_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
233{
234 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400236 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500237}
238
239__u32 ext4_itable_unused_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241{
242 return le16_to_cpu(bg->bg_itable_unused_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400244 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500245}
246
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700247void ext4_block_bitmap_set(struct super_block *sb,
248 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700249{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400250 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700253}
254
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700255void ext4_inode_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700257{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400258 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700261}
262
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700263void ext4_inode_table_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700265{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400266 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700269}
270
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400271void ext4_free_group_clusters_set(struct super_block *sb,
272 struct ext4_group_desc *bg, __u32 count)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500273{
274 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
277}
278
279void ext4_free_inodes_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
281{
282 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
285}
286
287void ext4_used_dirs_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
289{
290 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
293}
294
295void ext4_itable_unused_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297{
298 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
301}
302
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400303
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400304static void __save_error_info(struct super_block *sb, const char *func,
305 unsigned int line)
306{
307 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
308
309 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
311 es->s_last_error_time = cpu_to_le32(get_seconds());
312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
313 es->s_last_error_line = cpu_to_le32(line);
314 if (!es->s_first_error_time) {
315 es->s_first_error_time = es->s_last_error_time;
316 strncpy(es->s_first_error_func, func,
317 sizeof(es->s_first_error_func));
318 es->s_first_error_line = cpu_to_le32(line);
319 es->s_first_error_ino = es->s_last_error_ino;
320 es->s_first_error_block = es->s_last_error_block;
321 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400322 /*
323 * Start the daily error reporting function if it hasn't been
324 * started already
325 */
326 if (!es->s_error_count)
327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Wei Yongjunba39ebb2012-09-27 09:37:53 -0400328 le32_add_cpu(&es->s_error_count, 1);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400329}
330
331static void save_error_info(struct super_block *sb, const char *func,
332 unsigned int line)
333{
334 __save_error_info(sb, func, line);
335 ext4_commit_super(sb, 1);
336}
337
Theodore Ts'o7c2e7082011-09-09 18:28:51 -0400338/*
339 * The del_gendisk() function uninitializes the disk-specific data
340 * structures, including the bdi structure, without telling anyone
341 * else. Once this happens, any attempt to call mark_buffer_dirty()
342 * (for example, by ext4_commit_super), will cause a kernel OOPS.
343 * This is a kludge to prevent these oops until we can put in a proper
344 * hook in del_gendisk() to inform the VFS and file system layers.
345 */
346static int block_device_ejected(struct super_block *sb)
347{
348 struct inode *bd_inode = sb->s_bdev->bd_inode;
349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
350
351 return bdi->dev == NULL;
352}
353
Bobi Jam18aadd42012-02-20 17:53:02 -0500354static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
355{
356 struct super_block *sb = journal->j_private;
357 struct ext4_sb_info *sbi = EXT4_SB(sb);
358 int error = is_journal_aborted(journal);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400359 struct ext4_journal_cb_entry *jce;
Bobi Jam18aadd42012-02-20 17:53:02 -0500360
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400361 BUG_ON(txn->t_state == T_FINISHED);
Bobi Jam18aadd42012-02-20 17:53:02 -0500362 spin_lock(&sbi->s_md_lock);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400363 while (!list_empty(&txn->t_private_list)) {
364 jce = list_entry(txn->t_private_list.next,
365 struct ext4_journal_cb_entry, jce_list);
Bobi Jam18aadd42012-02-20 17:53:02 -0500366 list_del_init(&jce->jce_list);
367 spin_unlock(&sbi->s_md_lock);
368 jce->jce_func(sb, jce, error);
369 spin_lock(&sbi->s_md_lock);
370 }
371 spin_unlock(&sbi->s_md_lock);
372}
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400373
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700374/* Deal with the reporting of failure conditions on a filesystem such as
375 * inconsistencies detected or read IO failures.
376 *
377 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700378 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700379 * write ordering constraints on the superblock which prevent us from
380 * writing it out straight away; and given that the journal is about to
381 * be aborted, we can't rely on the current, or future, transactions to
382 * write out the superblock safely.
383 *
Mingming Caodab291a2006-10-11 01:21:01 -0700384 * 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 -0200385 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700386 * that error until we've noted it down and cleared it.
387 */
388
Mingming Cao617ba132006-10-11 01:20:53 -0700389static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700390{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700391 if (sb->s_flags & MS_RDONLY)
392 return;
393
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400394 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700395 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700396
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700398 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700399 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700400 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400401 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400403 /*
404 * Make sure updated value of ->s_mount_flags will be visible
405 * before ->s_flags update
406 */
407 smp_wmb();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700408 sb->s_flags |= MS_RDONLY;
409 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700410 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700411 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700412 sb->s_id);
413}
414
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400415#define ext4_error_ratelimit(sb) \
416 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
417 "EXT4-fs error")
418
Eric Sandeen12062dd2010-02-15 14:19:27 -0500419void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400420 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700421{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500422 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700423 va_list args;
424
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400425 if (ext4_error_ratelimit(sb)) {
426 va_start(args, fmt);
427 vaf.fmt = fmt;
428 vaf.va = &args;
429 printk(KERN_CRIT
430 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
431 sb->s_id, function, line, current->comm, &vaf);
432 va_end(args);
433 }
Theodore Ts'of3fc02102012-05-30 23:00:16 -0400434 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700435 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700436}
437
Joe Perchese7c96e82013-07-01 08:12:37 -0400438void __ext4_error_inode(struct inode *inode, const char *function,
439 unsigned int line, ext4_fsblk_t block,
440 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500441{
442 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500443 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400444 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500445
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400446 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
447 es->s_last_error_block = cpu_to_le64(block);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400448 if (ext4_error_ratelimit(inode->i_sb)) {
449 va_start(args, fmt);
450 vaf.fmt = fmt;
451 vaf.va = &args;
452 if (block)
453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
454 "inode #%lu: block %llu: comm %s: %pV\n",
455 inode->i_sb->s_id, function, line, inode->i_ino,
456 block, current->comm, &vaf);
457 else
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
459 "inode #%lu: comm %s: %pV\n",
460 inode->i_sb->s_id, function, line, inode->i_ino,
461 current->comm, &vaf);
462 va_end(args);
463 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400464 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500465 ext4_handle_error(inode->i_sb);
466}
467
Joe Perchese7c96e82013-07-01 08:12:37 -0400468void __ext4_error_file(struct file *file, const char *function,
469 unsigned int line, ext4_fsblk_t block,
470 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500471{
472 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500473 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400474 struct ext4_super_block *es;
Al Viro496ad9a2013-01-23 17:07:38 -0500475 struct inode *inode = file_inode(file);
Frank Mayhar273df552010-03-02 11:46:09 -0500476 char pathname[80], *path;
477
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400478 es = EXT4_SB(inode->i_sb)->s_es;
479 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400480 if (ext4_error_ratelimit(inode->i_sb)) {
481 path = d_path(&(file->f_path), pathname, sizeof(pathname));
482 if (IS_ERR(path))
483 path = "(unknown)";
484 va_start(args, fmt);
485 vaf.fmt = fmt;
486 vaf.va = &args;
487 if (block)
488 printk(KERN_CRIT
489 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
490 "block %llu: comm %s: path %s: %pV\n",
491 inode->i_sb->s_id, function, line, inode->i_ino,
492 block, current->comm, path, &vaf);
493 else
494 printk(KERN_CRIT
495 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
496 "comm %s: path %s: %pV\n",
497 inode->i_sb->s_id, function, line, inode->i_ino,
498 current->comm, path, &vaf);
499 va_end(args);
500 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400501 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500502 ext4_handle_error(inode->i_sb);
503}
504
Theodore Ts'o722887d2013-02-08 13:00:31 -0500505const char *ext4_decode_error(struct super_block *sb, int errno,
506 char nbuf[16])
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700507{
508 char *errstr = NULL;
509
510 switch (errno) {
511 case -EIO:
512 errstr = "IO failure";
513 break;
514 case -ENOMEM:
515 errstr = "Out of memory";
516 break;
517 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400518 if (!sb || (EXT4_SB(sb)->s_journal &&
519 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700520 errstr = "Journal has aborted";
521 else
522 errstr = "Readonly filesystem";
523 break;
524 default:
525 /* If the caller passed in an extra buffer for unknown
526 * errors, textualise them now. Else we just return
527 * NULL. */
528 if (nbuf) {
529 /* Check for truncated error codes... */
530 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
531 errstr = nbuf;
532 }
533 break;
534 }
535
536 return errstr;
537}
538
Mingming Cao617ba132006-10-11 01:20:53 -0700539/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700540 * automatically and invokes the appropriate error response. */
541
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400542void __ext4_std_error(struct super_block *sb, const char *function,
543 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700544{
545 char nbuf[16];
546 const char *errstr;
547
548 /* Special case: if the error is EROFS, and we're not already
549 * inside a transaction, then there's really no point in logging
550 * an error. */
551 if (errno == -EROFS && journal_current_handle() == NULL &&
552 (sb->s_flags & MS_RDONLY))
553 return;
554
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400555 if (ext4_error_ratelimit(sb)) {
556 errstr = ext4_decode_error(sb, errno, nbuf);
557 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
558 sb->s_id, function, line, errstr);
559 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700560
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400561 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700562 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700563}
564
565/*
Mingming Cao617ba132006-10-11 01:20:53 -0700566 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700567 * abort function may be used to deal with unrecoverable failures such
568 * as journal IO errors or ENOMEM at a critical moment in log management.
569 *
570 * We unconditionally force the filesystem into an ABORT|READONLY state,
571 * unless the error response on the fs has been set to panic in which
572 * case we take the easy way out and panic immediately.
573 */
574
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400575void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400576 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700577{
578 va_list args;
579
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400580 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700581 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
583 function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700584 vprintk(fmt, args);
585 printk("\n");
586 va_end(args);
587
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400588 if ((sb->s_flags & MS_RDONLY) == 0) {
589 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400590 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400591 /*
592 * Make sure updated value of ->s_mount_flags will be visible
593 * before ->s_flags update
594 */
595 smp_wmb();
596 sb->s_flags |= MS_RDONLY;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400597 if (EXT4_SB(sb)->s_journal)
598 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
599 save_error_info(sb, function, line);
600 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700602 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700603}
604
Joe Perchese7c96e82013-07-01 08:12:37 -0400605void __ext4_msg(struct super_block *sb,
606 const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400607{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500608 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400609 va_list args;
610
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400611 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
612 return;
613
Eric Sandeenb31e1552009-06-04 17:36:36 -0400614 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500615 vaf.fmt = fmt;
616 vaf.va = &args;
617 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400618 va_end(args);
619}
620
Eric Sandeen12062dd2010-02-15 14:19:27 -0500621void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400622 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700623{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500624 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700625 va_list args;
626
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400627 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
628 "EXT4-fs warning"))
629 return;
630
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700631 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500632 vaf.fmt = fmt;
633 vaf.va = &args;
634 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
635 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700636 va_end(args);
637}
638
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400639void __ext4_grp_locked_error(const char *function, unsigned int line,
640 struct super_block *sb, ext4_group_t grp,
641 unsigned long ino, ext4_fsblk_t block,
642 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500643__releases(bitlock)
644__acquires(bitlock)
645{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500646 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500647 va_list args;
648 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
649
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400650 es->s_last_error_ino = cpu_to_le32(ino);
651 es->s_last_error_block = cpu_to_le64(block);
652 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500653
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400654 if (ext4_error_ratelimit(sb)) {
655 va_start(args, fmt);
656 vaf.fmt = fmt;
657 vaf.va = &args;
658 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
659 sb->s_id, function, line, grp);
660 if (ino)
661 printk(KERN_CONT "inode %lu: ", ino);
662 if (block)
663 printk(KERN_CONT "block %llu:",
664 (unsigned long long) block);
665 printk(KERN_CONT "%pV\n", &vaf);
666 va_end(args);
667 }
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500668
669 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400670 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500671 return;
672 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400673
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500674 ext4_unlock_group(sb, grp);
675 ext4_handle_error(sb);
676 /*
677 * We only get here in the ERRORS_RO case; relocking the group
678 * may be dangerous, but nothing bad will happen since the
679 * filesystem will have already been marked read/only and the
680 * journal has been aborted. We return 1 as a hint to callers
681 * who might what to use the return value from
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300682 * ext4_grp_locked_error() to distinguish between the
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500683 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
684 * aggressively from the ext4 function in question, with a
685 * more appropriate error code.
686 */
687 ext4_lock_group(sb, grp);
688 return;
689}
690
Mingming Cao617ba132006-10-11 01:20:53 -0700691void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700692{
Mingming Cao617ba132006-10-11 01:20:53 -0700693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700694
Mingming Cao617ba132006-10-11 01:20:53 -0700695 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700696 return;
697
Eric Sandeen12062dd2010-02-15 14:19:27 -0500698 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700699 "updating to rev %d because of new feature flag, "
700 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700701 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700702
Mingming Cao617ba132006-10-11 01:20:53 -0700703 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
704 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
705 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700706 /* leave es->s_feature_*compat flags alone */
707 /* es->s_uuid will be set by e2fsck if empty */
708
709 /*
710 * The rest of the superblock fields should be zero, and if not it
711 * means they are likely already in use, so leave them alone. We
712 * can leave it up to e2fsck to clean up any inconsistencies there.
713 */
714}
715
716/*
717 * Open the external journal device
718 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400719static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700720{
721 struct block_device *bdev;
722 char b[BDEVNAME_SIZE];
723
Tejun Heod4d77622010-11-13 11:55:18 +0100724 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700725 if (IS_ERR(bdev))
726 goto fail;
727 return bdev;
728
729fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400730 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700731 __bdevname(dev, b), PTR_ERR(bdev));
732 return NULL;
733}
734
735/*
736 * Release the journal device
737 */
Al Viro4385bab2013-05-05 22:11:03 -0400738static void ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700739{
Al Viro4385bab2013-05-05 22:11:03 -0400740 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700741}
742
Al Viro4385bab2013-05-05 22:11:03 -0400743static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700744{
745 struct block_device *bdev;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700746 bdev = sbi->journal_bdev;
747 if (bdev) {
Al Viro4385bab2013-05-05 22:11:03 -0400748 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700749 sbi->journal_bdev = NULL;
750 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700751}
752
753static inline struct inode *orphan_list_entry(struct list_head *l)
754{
Mingming Cao617ba132006-10-11 01:20:53 -0700755 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700756}
757
Mingming Cao617ba132006-10-11 01:20:53 -0700758static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700759{
760 struct list_head *l;
761
Eric Sandeenb31e1552009-06-04 17:36:36 -0400762 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
763 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700764
765 printk(KERN_ERR "sb_info orphan list:\n");
766 list_for_each(l, &sbi->s_orphan) {
767 struct inode *inode = orphan_list_entry(l);
768 printk(KERN_ERR " "
769 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
770 inode->i_sb->s_id, inode->i_ino, inode,
771 inode->i_mode, inode->i_nlink,
772 NEXT_ORPHAN(inode));
773 }
774}
775
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400776static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700777{
Mingming Cao617ba132006-10-11 01:20:53 -0700778 struct ext4_sb_info *sbi = EXT4_SB(sb);
779 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400780 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700781
Lukas Czerner857ac882010-10-27 21:30:05 -0400782 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400783 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
784
Jan Kara2e8fa542013-06-04 14:21:02 -0400785 flush_workqueue(sbi->rsv_conversion_wq);
Jan Kara2e8fa542013-06-04 14:21:02 -0400786 destroy_workqueue(sbi->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -0400787
Frank Mayhar03901312009-01-07 00:06:22 -0500788 if (sbi->s_journal) {
789 err = jbd2_journal_destroy(sbi->s_journal);
790 sbi->s_journal = NULL;
791 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400792 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500793 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500794
Zheng Liud3922a72013-07-01 08:12:37 -0400795 ext4_es_unregister_shrinker(sbi);
Al Viro9105bb12013-12-08 20:52:31 -0500796 del_timer_sync(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500797 ext4_release_system_zone(sb);
798 ext4_mb_release(sb);
799 ext4_ext_release(sb);
800 ext4_xattr_put_super(sb);
801
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700802 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700803 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700804 es->s_state = cpu_to_le16(sbi->s_mount_state);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700805 }
Artem Bityutskiy58c58732012-07-22 20:33:31 -0400806 if (!(sb->s_flags & MS_RDONLY))
Artem Bityutskiya8e25a82012-03-21 22:29:15 -0400807 ext4_commit_super(sb, 1);
808
Theodore Ts'o240799c2008-10-09 23:53:47 -0400809 if (sbi->s_proc) {
Theodore Ts'o66acdcf2012-03-04 20:21:38 -0500810 remove_proc_entry("options", sbi->s_proc);
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400811 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400812 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400813 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700814
815 for (i = 0; i < sbi->s_gdb_count; i++)
816 brelse(sbi->s_group_desc[i]);
Theodore Ts'of18a5f22011-08-01 08:45:38 -0400817 ext4_kvfree(sbi->s_group_desc);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400818 ext4_kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -0400819 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
821 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -0400822 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Theodore Ts'o1ac64662013-03-02 10:27:46 -0500823 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700824 brelse(sbi->s_sbh);
825#ifdef CONFIG_QUOTA
826 for (i = 0; i < MAXQUOTAS; i++)
827 kfree(sbi->s_qf_names[i]);
828#endif
829
830 /* Debugging code just in case the in-memory inode orphan list
831 * isn't empty. The on-disk one can be non-empty if we've
832 * detected an error and taken the fs readonly, but the
833 * in-memory list had better be clean by this point. */
834 if (!list_empty(&sbi->s_orphan))
835 dump_orphan_list(sb, sbi);
836 J_ASSERT(list_empty(&sbi->s_orphan));
837
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700838 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700839 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
840 /*
841 * Invalidate the journal device's buffers. We don't want them
842 * floating about in memory - the physical journal device may
843 * hotswapped, and it breaks the `ro-after' testing code.
844 */
845 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700846 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700847 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700848 }
T Makphaibulchoke9c191f72014-03-18 19:24:49 -0400849 if (sbi->s_mb_cache) {
850 ext4_xattr_destroy_cache(sbi->s_mb_cache);
851 sbi->s_mb_cache = NULL;
852 }
Johann Lombardic5e06d12011-05-24 18:31:25 -0400853 if (sbi->s_mmp_tsk)
854 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700855 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400856 /*
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
859 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400860 kobject_put(&sbi->s_kobj);
861 wait_for_completion(&sbi->s_kobj_unregister);
Darrick J. Wong04419842012-04-29 18:27:10 -0400862 if (sbi->s_chksum_driver)
863 crypto_free_shash(sbi->s_chksum_driver);
Pekka Enberg705895b2009-02-15 18:07:52 -0500864 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700865 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700866}
867
Christoph Lametere18b8902006-12-06 20:33:20 -0800868static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700869
870/*
871 * Called inside transaction, so use GFP_NOFS
872 */
Mingming Cao617ba132006-10-11 01:20:53 -0700873static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700874{
Mingming Cao617ba132006-10-11 01:20:53 -0700875 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700876
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800877 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700878 if (!ei)
879 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400880
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700881 ei->vfs_inode.i_version = 1;
Theodore Ts'o202ee5d2014-04-21 14:37:55 -0400882 spin_lock_init(&ei->i_raw_lock);
Alex Tomasc9de5602008-01-29 00:19:52 -0500883 INIT_LIST_HEAD(&ei->i_prealloc_list);
884 spin_lock_init(&ei->i_prealloc_lock);
Zheng Liu9a26b662012-11-08 21:57:30 -0500885 ext4_es_init_tree(&ei->i_es_tree);
886 rwlock_init(&ei->i_es_lock);
Zheng Liu74cd15c2013-02-18 00:32:55 -0500887 INIT_LIST_HEAD(&ei->i_es_lru);
888 ei->i_es_lru_nr = 0;
Zheng Liud3922a72013-07-01 08:12:37 -0400889 ei->i_touch_when = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400890 ei->i_reserved_data_blocks = 0;
891 ei->i_reserved_meta_blocks = 0;
892 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500893 ei->i_da_metadata_calc_len = 0;
Theodore Ts'o7e731bc2012-08-05 23:28:16 -0400894 ei->i_da_metadata_calc_last_lblock = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400895 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300896#ifdef CONFIG_QUOTA
897 ei->i_reserved_quota = 0;
898#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500899 ei->jinode = NULL;
Jan Kara2e8fa542013-06-04 14:21:02 -0400900 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500901 spin_lock_init(&ei->i_completed_io_lock);
Jan Karab436b9b2009-12-08 23:51:10 -0500902 ei->i_sync_tid = 0;
903 ei->i_datasync_tid = 0;
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500904 atomic_set(&ei->i_ioend_count, 0);
Dmitry Monakhove27f41e2012-09-28 23:24:52 -0400905 atomic_set(&ei->i_unwritten, 0);
Jan Kara2e8fa542013-06-04 14:21:02 -0400906 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400907
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700908 return &ei->vfs_inode;
909}
910
Theodore Ts'o7ff9c072010-11-08 13:51:33 -0500911static int ext4_drop_inode(struct inode *inode)
912{
913 int drop = generic_drop_inode(inode);
914
915 trace_ext4_drop_inode(inode, drop);
916 return drop;
917}
918
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100919static void ext4_i_callback(struct rcu_head *head)
920{
921 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100922 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
923}
924
Mingming Cao617ba132006-10-11 01:20:53 -0700925static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700926{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700927 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400928 ext4_msg(inode->i_sb, KERN_ERR,
929 "Inode %lu (%p): orphan list check failed!",
930 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700931 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
932 EXT4_I(inode), sizeof(struct ext4_inode_info),
933 true);
934 dump_stack();
935 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100936 call_rcu(&inode->i_rcu, ext4_i_callback);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700937}
938
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700939static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700940{
Mingming Cao617ba132006-10-11 01:20:53 -0700941 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700942
Christoph Lametera35afb82007-05-16 22:10:57 -0700943 INIT_LIST_HEAD(&ei->i_orphan);
Christoph Lametera35afb82007-05-16 22:10:57 -0700944 init_rwsem(&ei->xattr_sem);
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500945 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700946 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700947}
948
Fabian Fredericke67bc2b2014-02-17 20:34:53 -0500949static int __init init_inodecache(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700950{
Mingming Cao617ba132006-10-11 01:20:53 -0700951 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
952 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700953 0, (SLAB_RECLAIM_ACCOUNT|
954 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900955 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700956 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700957 return -ENOMEM;
958 return 0;
959}
960
961static void destroy_inodecache(void)
962{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +1000963 /*
964 * Make sure all delayed rcu free inodes are flushed before we
965 * destroy cache.
966 */
967 rcu_barrier();
Mingming Cao617ba132006-10-11 01:20:53 -0700968 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700969}
970
Al Viro0930fcc2010-06-07 13:16:22 -0400971void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700972{
Al Viro0930fcc2010-06-07 13:16:22 -0400973 invalidate_inode_buffers(inode);
Jan Karadbd57682012-05-03 14:48:02 +0200974 clear_inode(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -0500975 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400976 ext4_discard_preallocations(inode);
Zheng Liu51865fd2012-11-08 21:57:32 -0500977 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
Zheng Liu74cd15c2013-02-18 00:32:55 -0500978 ext4_es_lru_del(inode);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500979 if (EXT4_I(inode)->jinode) {
980 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
981 EXT4_I(inode)->jinode);
982 jbd2_free_inode(EXT4_I(inode)->jinode);
983 EXT4_I(inode)->jinode = NULL;
984 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700985}
986
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700987static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400988 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700989{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700990 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700991
Mingming Cao617ba132006-10-11 01:20:53 -0700992 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700993 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700994 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700995 return ERR_PTR(-ESTALE);
996
997 /* iget isn't really right if the inode is currently unallocated!!
998 *
Mingming Cao617ba132006-10-11 01:20:53 -0700999 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001000 * deleted, so we should be safe.
1001 *
1002 * Currently we don't know the generation for parent directory, so
1003 * a generation of 0 means "accept any"
1004 */
David Howells1d1fe1e2008-02-07 00:15:37 -08001005 inode = ext4_iget(sb, ino);
1006 if (IS_ERR(inode))
1007 return ERR_CAST(inode);
1008 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001009 iput(inode);
1010 return ERR_PTR(-ESTALE);
1011 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001012
1013 return inode;
1014}
1015
1016static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001017 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001018{
1019 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1020 ext4_nfs_get_inode);
1021}
1022
1023static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001024 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001025{
1026 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001028}
1029
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001030/*
1031 * Try to release metadata pages (indirect blocks, directories) which are
1032 * mapped via the block device. Since these pages could have journal heads
1033 * which would prevent try_to_free_buffers() from freeing them, we must use
1034 * jbd2 layer's try_to_free_buffers() function to release them.
1035 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001036static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1037 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001038{
1039 journal_t *journal = EXT4_SB(sb)->s_journal;
1040
1041 WARN_ON(PageChecked(page));
1042 if (!page_has_buffers(page))
1043 return 0;
1044 if (journal)
1045 return jbd2_journal_try_to_free_buffers(journal, page,
1046 wait & ~__GFP_WAIT);
1047 return try_to_free_buffers(page);
1048}
1049
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001050#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001051#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001052#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001053
Mingming Cao617ba132006-10-11 01:20:53 -07001054static int ext4_write_dquot(struct dquot *dquot);
1055static int ext4_acquire_dquot(struct dquot *dquot);
1056static int ext4_release_dquot(struct dquot *dquot);
1057static int ext4_mark_dquot_dirty(struct dquot *dquot);
1058static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001059static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02001060 struct path *path);
Aditya Kali7c319d32012-07-22 20:21:31 -04001061static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1062 int format_id);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001063static int ext4_quota_off(struct super_block *sb, int type);
Aditya Kali7c319d32012-07-22 20:21:31 -04001064static int ext4_quota_off_sysfile(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001065static int ext4_quota_on_mount(struct super_block *sb, int type);
1066static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001067 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001068static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001069 const char *data, size_t len, loff_t off);
Aditya Kali7c319d32012-07-22 20:21:31 -04001070static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1071 unsigned int flags);
1072static int ext4_enable_quotas(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001073
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001074static const struct dquot_operations ext4_quota_operations = {
Mingming Cao60e58e02009-01-22 18:13:05 +01001075 .get_reserved_space = ext4_get_reserved_space,
Mingming Cao617ba132006-10-11 01:20:53 -07001076 .write_dquot = ext4_write_dquot,
1077 .acquire_dquot = ext4_acquire_dquot,
1078 .release_dquot = ext4_release_dquot,
1079 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001080 .write_info = ext4_write_info,
1081 .alloc_dquot = dquot_alloc,
1082 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001083};
1084
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001085static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001086 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001087 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001088 .quota_sync = dquot_quota_sync,
1089 .get_info = dquot_get_dqinfo,
1090 .set_info = dquot_set_dqinfo,
1091 .get_dqblk = dquot_get_dqblk,
1092 .set_dqblk = dquot_set_dqblk
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001093};
Aditya Kali7c319d32012-07-22 20:21:31 -04001094
1095static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1096 .quota_on_meta = ext4_quota_on_sysfile,
1097 .quota_off = ext4_quota_off_sysfile,
1098 .quota_sync = dquot_quota_sync,
1099 .get_info = dquot_get_dqinfo,
1100 .set_info = dquot_set_dqinfo,
1101 .get_dqblk = dquot_get_dqblk,
1102 .set_dqblk = dquot_set_dqblk
1103};
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001104#endif
1105
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001106static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001111 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001112 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001113 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001114 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001115 .freeze_fs = ext4_freeze,
1116 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001119 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001120#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001121 .quota_read = ext4_quota_read,
1122 .quota_write = ext4_quota_write,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001123#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001124 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001125};
1126
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001127static const struct super_operations ext4_nojournal_sops = {
1128 .alloc_inode = ext4_alloc_inode,
1129 .destroy_inode = ext4_destroy_inode,
1130 .write_inode = ext4_write_inode,
1131 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001132 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001133 .evict_inode = ext4_evict_inode,
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04001134 .sync_fs = ext4_sync_fs_nojournal,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001135 .put_super = ext4_put_super,
1136 .statfs = ext4_statfs,
1137 .remount_fs = ext4_remount,
Theodore Ts'o9ca92382009-05-01 12:52:25 -04001138 .show_options = ext4_show_options,
1139#ifdef CONFIG_QUOTA
1140 .quota_read = ext4_quota_read,
1141 .quota_write = ext4_quota_write,
1142#endif
1143 .bdev_try_to_free_page = bdev_try_to_free_page,
1144};
1145
Christoph Hellwig39655162007-10-21 16:42:17 -07001146static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001147 .fh_to_dentry = ext4_fh_to_dentry,
1148 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001149 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001150};
1151
1152enum {
1153 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1154 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001155 Opt_nouid32, Opt_debug, Opt_removed,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001156 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001157 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
Eric Sandeenad4eec62013-08-28 19:05:07 -04001158 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1159 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001160 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o296c3552009-09-30 00:32:42 -04001161 Opt_data_err_abort, Opt_data_err_ignore,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001162 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001163 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
Theodore Ts'oee4a3fc2012-03-02 12:14:24 -05001164 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
Eric Sandeen661aa522012-02-20 17:53:04 -05001165 Opt_usrquota, Opt_grpquota, Opt_i_version,
Theodore Ts'o14490322010-12-14 15:27:50 -05001166 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1167 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001168 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001169 Opt_dioread_nolock, Opt_dioread_lock,
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001170 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
Theodore Ts'odf981d02012-08-17 09:48:17 -04001171 Opt_max_dir_size_kb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001172};
1173
Steven Whitehousea447c092008-10-13 10:46:57 +01001174static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001175 {Opt_bsd_df, "bsddf"},
1176 {Opt_minix_df, "minixdf"},
1177 {Opt_grpid, "grpid"},
1178 {Opt_grpid, "bsdgroups"},
1179 {Opt_nogrpid, "nogrpid"},
1180 {Opt_nogrpid, "sysvgroups"},
1181 {Opt_resgid, "resgid=%u"},
1182 {Opt_resuid, "resuid=%u"},
1183 {Opt_sb, "sb=%u"},
1184 {Opt_err_cont, "errors=continue"},
1185 {Opt_err_panic, "errors=panic"},
1186 {Opt_err_ro, "errors=remount-ro"},
1187 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001188 {Opt_debug, "debug"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001189 {Opt_removed, "oldalloc"},
1190 {Opt_removed, "orlov"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001191 {Opt_user_xattr, "user_xattr"},
1192 {Opt_nouser_xattr, "nouser_xattr"},
1193 {Opt_acl, "acl"},
1194 {Opt_noacl, "noacl"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001195 {Opt_noload, "norecovery"},
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001196 {Opt_noload, "noload"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001197 {Opt_removed, "nobh"},
1198 {Opt_removed, "bh"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001199 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001200 {Opt_min_batch_time, "min_batch_time=%u"},
1201 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001202 {Opt_journal_dev, "journal_dev=%u"},
Eric Sandeenad4eec62013-08-28 19:05:07 -04001203 {Opt_journal_path, "journal_path=%s"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001204 {Opt_journal_checksum, "journal_checksum"},
1205 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001206 {Opt_abort, "abort"},
1207 {Opt_data_journal, "data=journal"},
1208 {Opt_data_ordered, "data=ordered"},
1209 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001210 {Opt_data_err_abort, "data_err=abort"},
1211 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001212 {Opt_offusrjquota, "usrjquota="},
1213 {Opt_usrjquota, "usrjquota=%s"},
1214 {Opt_offgrpjquota, "grpjquota="},
1215 {Opt_grpjquota, "grpjquota=%s"},
1216 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1217 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001218 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001219 {Opt_grpquota, "grpquota"},
1220 {Opt_noquota, "noquota"},
1221 {Opt_quota, "quota"},
1222 {Opt_usrquota, "usrquota"},
1223 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001224 {Opt_barrier, "barrier"},
1225 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001226 {Opt_i_version, "i_version"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001227 {Opt_stripe, "stripe=%u"},
Alex Tomas64769242008-07-11 19:27:31 -04001228 {Opt_delalloc, "delalloc"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001229 {Opt_nodelalloc, "nodelalloc"},
Jan Kara36ade452013-01-28 09:30:52 -05001230 {Opt_removed, "mblk_io_submit"},
1231 {Opt_removed, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001232 {Opt_block_validity, "block_validity"},
1233 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001234 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001235 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001236 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001237 {Opt_auto_da_alloc, "auto_da_alloc"},
1238 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001239 {Opt_dioread_nolock, "dioread_nolock"},
1240 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001241 {Opt_discard, "discard"},
1242 {Opt_nodiscard, "nodiscard"},
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001243 {Opt_init_itable, "init_itable=%u"},
1244 {Opt_init_itable, "init_itable"},
1245 {Opt_noinit_itable, "noinit_itable"},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001246 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
Theodore Ts'oc7198b92012-03-04 22:00:53 -05001247 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1248 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1249 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1251 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001252 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001253};
1254
Mingming Cao617ba132006-10-11 01:20:53 -07001255static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001256{
Mingming Cao617ba132006-10-11 01:20:53 -07001257 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001258 char *options = (char *) *data;
1259
1260 if (!options || strncmp(options, "sb=", 3) != 0)
1261 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001262
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001263 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001264 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001265 sb_block = simple_strtoul(options, &options, 0);
1266 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001267 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001268 (char *) *data);
1269 return 1;
1270 }
1271 if (*options == ',')
1272 options++;
1273 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001274
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001275 return sb_block;
1276}
1277
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001278#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001279static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1280 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001281
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001282#ifdef CONFIG_QUOTA
1283static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1284{
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1286 char *qname;
Chen Gang03dafb52013-01-24 23:24:58 -05001287 int ret = -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001288
1289 if (sb_any_quota_loaded(sb) &&
1290 !sbi->s_qf_names[qtype]) {
1291 ext4_msg(sb, KERN_ERR,
1292 "Cannot change journaled "
1293 "quota options when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001294 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001295 }
Jan Kara262b4662013-03-02 17:57:08 -05001296 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1297 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1298 "when QUOTA feature is enabled");
1299 return -1;
1300 }
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001301 qname = match_strdup(args);
1302 if (!qname) {
1303 ext4_msg(sb, KERN_ERR,
1304 "Not enough memory for storing quotafile name");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001305 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001306 }
Chen Gang03dafb52013-01-24 23:24:58 -05001307 if (sbi->s_qf_names[qtype]) {
1308 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1309 ret = 1;
1310 else
1311 ext4_msg(sb, KERN_ERR,
1312 "%s quota file already specified",
1313 QTYPE2NAME(qtype));
1314 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001315 }
Chen Gang03dafb52013-01-24 23:24:58 -05001316 if (strchr(qname, '/')) {
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001317 ext4_msg(sb, KERN_ERR,
1318 "quotafile must be on filesystem root");
Chen Gang03dafb52013-01-24 23:24:58 -05001319 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001320 }
Chen Gang03dafb52013-01-24 23:24:58 -05001321 sbi->s_qf_names[qtype] = qname;
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001322 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001323 return 1;
Chen Gang03dafb52013-01-24 23:24:58 -05001324errout:
1325 kfree(qname);
1326 return ret;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001327}
1328
1329static int clear_qf_name(struct super_block *sb, int qtype)
1330{
1331
1332 struct ext4_sb_info *sbi = EXT4_SB(sb);
1333
1334 if (sb_any_quota_loaded(sb) &&
1335 sbi->s_qf_names[qtype]) {
1336 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1337 " when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001338 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001339 }
Chen Gang03dafb52013-01-24 23:24:58 -05001340 kfree(sbi->s_qf_names[qtype]);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001341 sbi->s_qf_names[qtype] = NULL;
1342 return 1;
1343}
1344#endif
1345
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001346#define MOPT_SET 0x0001
1347#define MOPT_CLEAR 0x0002
1348#define MOPT_NOSUPPORT 0x0004
1349#define MOPT_EXPLICIT 0x0008
1350#define MOPT_CLEAR_ERR 0x0010
1351#define MOPT_GTE0 0x0020
1352#ifdef CONFIG_QUOTA
1353#define MOPT_Q 0
1354#define MOPT_QFMT 0x0040
1355#else
1356#define MOPT_Q MOPT_NOSUPPORT
1357#define MOPT_QFMT MOPT_NOSUPPORT
1358#endif
1359#define MOPT_DATAJ 0x0080
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001360#define MOPT_NO_EXT2 0x0100
1361#define MOPT_NO_EXT3 0x0200
1362#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
Eric Sandeenad4eec62013-08-28 19:05:07 -04001363#define MOPT_STRING 0x0400
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001364
1365static const struct mount_opts {
1366 int token;
1367 int mount_opt;
1368 int flags;
1369} ext4_mount_opts[] = {
1370 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1371 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1372 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1373 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001374 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1375 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001376 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001380 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1381 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001382 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1384 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
Theodore Ts'o59d9fa52013-08-08 23:01:24 -04001385 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_stripe, 0, MOPT_GTE0},
Jan Kara0efb3b22013-02-02 22:52:19 -05001410 {Opt_resuid, 0, MOPT_GTE0},
1411 {Opt_resgid, 0, MOPT_GTE0},
1412 {Opt_journal_dev, 0, MOPT_GTE0},
Eric Sandeenad4eec62013-08-28 19:05:07 -04001413 {Opt_journal_path, 0, MOPT_STRING},
Jan Kara0efb3b22013-02-02 22:52:19 -05001414 {Opt_journal_ioprio, 0, MOPT_GTE0},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001415 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1416 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1418 MOPT_NO_EXT2 | MOPT_DATAJ},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001419 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1420 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001421#ifdef CONFIG_EXT4_FS_POSIX_ACL
1422 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1423 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1424#else
1425 {Opt_acl, 0, MOPT_NOSUPPORT},
1426 {Opt_noacl, 0, MOPT_NOSUPPORT},
1427#endif
1428 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1429 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1430 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1431 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1432 MOPT_SET | MOPT_Q},
1433 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1434 MOPT_SET | MOPT_Q},
1435 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1436 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1437 {Opt_usrjquota, 0, MOPT_Q},
1438 {Opt_grpjquota, 0, MOPT_Q},
1439 {Opt_offusrjquota, 0, MOPT_Q},
1440 {Opt_offgrpjquota, 0, MOPT_Q},
1441 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001444 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001445 {Opt_err, 0, 0}
1446};
1447
1448static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1449 substring_t *args, unsigned long *journal_devnum,
1450 unsigned int *journal_ioprio, int is_remount)
1451{
1452 struct ext4_sb_info *sbi = EXT4_SB(sb);
1453 const struct mount_opts *m;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001454 kuid_t uid;
1455 kgid_t gid;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001456 int arg = 0;
1457
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001458#ifdef CONFIG_QUOTA
1459 if (token == Opt_usrjquota)
1460 return set_qf_name(sb, USRQUOTA, &args[0]);
1461 else if (token == Opt_grpjquota)
1462 return set_qf_name(sb, GRPQUOTA, &args[0]);
1463 else if (token == Opt_offusrjquota)
1464 return clear_qf_name(sb, USRQUOTA);
1465 else if (token == Opt_offgrpjquota)
1466 return clear_qf_name(sb, GRPQUOTA);
1467#endif
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001468 switch (token) {
Theodore Ts'of7048602012-03-04 22:06:20 -05001469 case Opt_noacl:
1470 case Opt_nouser_xattr:
1471 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1472 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001473 case Opt_sb:
1474 return 1; /* handled by get_sb_block() */
1475 case Opt_removed:
Jan Kara5f3633e2013-02-02 23:09:36 -05001476 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001477 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001478 case Opt_abort:
1479 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1480 return 1;
1481 case Opt_i_version:
1482 sb->s_flags |= MS_I_VERSION;
1483 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001484 }
1485
Jan Kara5f3633e2013-02-02 23:09:36 -05001486 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 if (token == m->token)
1488 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001489
Jan Kara5f3633e2013-02-02 23:09:36 -05001490 if (m->token == Opt_err) {
1491 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 "or missing value", opt);
1493 return -1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001494 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001495
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001496 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext2", opt);
1499 return -1;
1500 }
1501 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Mount option \"%s\" incompatible with ext3", opt);
1504 return -1;
1505 }
1506
Eric Sandeenad4eec62013-08-28 19:05:07 -04001507 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
Jan Kara5f3633e2013-02-02 23:09:36 -05001508 return -1;
1509 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 return -1;
1511 if (m->flags & MOPT_EXPLICIT)
1512 set_opt2(sb, EXPLICIT_DELALLOC);
1513 if (m->flags & MOPT_CLEAR_ERR)
1514 clear_opt(sb, ERRORS_MASK);
1515 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 "options when quota turned on");
1518 return -1;
1519 }
1520
1521 if (m->flags & MOPT_NOSUPPORT) {
1522 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 } else if (token == Opt_commit) {
1524 if (arg == 0)
1525 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 sbi->s_commit_interval = HZ * arg;
1527 } else if (token == Opt_max_batch_time) {
1528 if (arg == 0)
1529 arg = EXT4_DEF_MAX_BATCH_TIME;
1530 sbi->s_max_batch_time = arg;
1531 } else if (token == Opt_min_batch_time) {
1532 sbi->s_min_batch_time = arg;
1533 } else if (token == Opt_inode_readahead_blks) {
Jan Karae33e60e2013-02-02 23:14:31 -05001534 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1535 ext4_msg(sb, KERN_ERR,
1536 "EXT4-fs: inode_readahead_blks must be "
1537 "0 or a power of 2 smaller than 2^31");
Jan Kara5f3633e2013-02-02 23:09:36 -05001538 return -1;
1539 }
1540 sbi->s_inode_readahead_blks = arg;
1541 } else if (token == Opt_init_itable) {
1542 set_opt(sb, INIT_INODE_TABLE);
1543 if (!args->from)
1544 arg = EXT4_DEF_LI_WAIT_MULT;
1545 sbi->s_li_wait_mult = arg;
1546 } else if (token == Opt_max_dir_size_kb) {
1547 sbi->s_max_dir_size_kb = arg;
1548 } else if (token == Opt_stripe) {
1549 sbi->s_stripe = arg;
1550 } else if (token == Opt_resuid) {
1551 uid = make_kuid(current_user_ns(), arg);
1552 if (!uid_valid(uid)) {
1553 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1554 return -1;
1555 }
1556 sbi->s_resuid = uid;
1557 } else if (token == Opt_resgid) {
1558 gid = make_kgid(current_user_ns(), arg);
1559 if (!gid_valid(gid)) {
1560 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1561 return -1;
1562 }
1563 sbi->s_resgid = gid;
1564 } else if (token == Opt_journal_dev) {
1565 if (is_remount) {
1566 ext4_msg(sb, KERN_ERR,
1567 "Cannot specify journal on remount");
1568 return -1;
1569 }
1570 *journal_devnum = arg;
Eric Sandeenad4eec62013-08-28 19:05:07 -04001571 } else if (token == Opt_journal_path) {
1572 char *journal_path;
1573 struct inode *journal_inode;
1574 struct path path;
1575 int error;
1576
1577 if (is_remount) {
1578 ext4_msg(sb, KERN_ERR,
1579 "Cannot specify journal on remount");
1580 return -1;
1581 }
1582 journal_path = match_strdup(&args[0]);
1583 if (!journal_path) {
1584 ext4_msg(sb, KERN_ERR, "error: could not dup "
1585 "journal device string");
1586 return -1;
1587 }
1588
1589 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1590 if (error) {
1591 ext4_msg(sb, KERN_ERR, "error: could not find "
1592 "journal device path: error %d", error);
1593 kfree(journal_path);
1594 return -1;
1595 }
1596
1597 journal_inode = path.dentry->d_inode;
1598 if (!S_ISBLK(journal_inode->i_mode)) {
1599 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1600 "is not a block device", journal_path);
1601 path_put(&path);
1602 kfree(journal_path);
1603 return -1;
1604 }
1605
1606 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1607 path_put(&path);
1608 kfree(journal_path);
Jan Kara5f3633e2013-02-02 23:09:36 -05001609 } else if (token == Opt_journal_ioprio) {
1610 if (arg > 7) {
1611 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1612 " (must be 0-7)");
1613 return -1;
1614 }
1615 *journal_ioprio =
1616 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1617 } else if (m->flags & MOPT_DATAJ) {
1618 if (is_remount) {
1619 if (!sbi->s_journal)
1620 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1621 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1622 ext4_msg(sb, KERN_ERR,
1623 "Cannot change data mode on remount");
1624 return -1;
1625 }
1626 } else {
1627 clear_opt(sb, DATA_FLAGS);
1628 sbi->s_mount_opt |= m->mount_opt;
1629 }
1630#ifdef CONFIG_QUOTA
1631 } else if (m->flags & MOPT_QFMT) {
1632 if (sb_any_quota_loaded(sb) &&
1633 sbi->s_jquota_fmt != m->mount_opt) {
1634 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1635 "quota options when quota turned on");
1636 return -1;
1637 }
Jan Kara262b4662013-03-02 17:57:08 -05001638 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1639 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1640 ext4_msg(sb, KERN_ERR,
1641 "Cannot set journaled quota options "
1642 "when QUOTA feature is enabled");
1643 return -1;
1644 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001645 sbi->s_jquota_fmt = m->mount_opt;
1646#endif
1647 } else {
1648 if (!args->from)
1649 arg = 1;
1650 if (m->flags & MOPT_CLEAR)
1651 arg = !arg;
1652 else if (unlikely(!(m->flags & MOPT_SET))) {
1653 ext4_msg(sb, KERN_WARNING,
1654 "buggy handling of option %s", opt);
1655 WARN_ON(1);
1656 return -1;
1657 }
1658 if (arg != 0)
1659 sbi->s_mount_opt |= m->mount_opt;
1660 else
1661 sbi->s_mount_opt &= ~m->mount_opt;
1662 }
1663 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001664}
1665
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001666static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001667 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001668 unsigned int *journal_ioprio,
Eric Sandeen661aa522012-02-20 17:53:04 -05001669 int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001670{
Mingming Cao617ba132006-10-11 01:20:53 -07001671 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001672 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001673 substring_t args[MAX_OPT_ARGS];
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001674 int token;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001675
1676 if (!options)
1677 return 1;
1678
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001679 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001680 if (!*p)
1681 continue;
Eric Sandeen15121c12010-02-15 20:17:55 -05001682 /*
1683 * Initialize args struct so we know whether arg was
1684 * found; some options take optional arguments.
1685 */
Sachin Kamatcaecd0a2012-08-18 22:29:18 -04001686 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001687 token = match_token(p, tokens, args);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001688 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1689 journal_ioprio, is_remount) < 0)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001690 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001691 }
1692#ifdef CONFIG_QUOTA
Jan Kara262b4662013-03-02 17:57:08 -05001693 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1694 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1695 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1696 "feature is enabled");
1697 return 0;
1698 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001699 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001700 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001701 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001702
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001703 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001704 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001705
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001706 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001707 ext4_msg(sb, KERN_ERR, "old and new quota "
1708 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001709 return 0;
1710 }
1711
1712 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001713 ext4_msg(sb, KERN_ERR, "journaled quota format "
1714 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001715 return 0;
1716 }
1717 } else {
1718 if (sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001719 ext4_msg(sb, KERN_ERR, "journaled quota format "
Jan Kara2c8be6b2008-05-13 21:27:55 -04001720 "specified with no journaling "
Eric Sandeenb31e1552009-06-04 17:36:36 -04001721 "enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001722 return 0;
1723 }
1724 }
1725#endif
Jan Kara261cb20c2012-12-20 00:07:18 -05001726 if (test_opt(sb, DIOREAD_NOLOCK)) {
1727 int blocksize =
1728 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1729
1730 if (blocksize < PAGE_CACHE_SIZE) {
1731 ext4_msg(sb, KERN_ERR, "can't mount with "
1732 "dioread_nolock if block size != PAGE_SIZE");
1733 return 0;
1734 }
1735 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001736 return 1;
1737}
1738
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001739static inline void ext4_show_quota_options(struct seq_file *seq,
1740 struct super_block *sb)
1741{
1742#if defined(CONFIG_QUOTA)
1743 struct ext4_sb_info *sbi = EXT4_SB(sb);
1744
1745 if (sbi->s_jquota_fmt) {
1746 char *fmtname = "";
1747
1748 switch (sbi->s_jquota_fmt) {
1749 case QFMT_VFS_OLD:
1750 fmtname = "vfsold";
1751 break;
1752 case QFMT_VFS_V0:
1753 fmtname = "vfsv0";
1754 break;
1755 case QFMT_VFS_V1:
1756 fmtname = "vfsv1";
1757 break;
1758 }
1759 seq_printf(seq, ",jqfmt=%s", fmtname);
1760 }
1761
1762 if (sbi->s_qf_names[USRQUOTA])
1763 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1764
1765 if (sbi->s_qf_names[GRPQUOTA])
1766 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001767#endif
1768}
1769
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001770static const char *token2str(int token)
1771{
Herton Ronaldo Krzesinski50df9fd2012-09-23 22:49:12 -04001772 const struct match_token *t;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001773
1774 for (t = tokens; t->token != Opt_err; t++)
1775 if (t->token == token && !strchr(t->pattern, '='))
1776 break;
1777 return t->pattern;
1778}
1779
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001780/*
1781 * Show an option if
1782 * - it's set to a non-default value OR
1783 * - if the per-sb default is different from the global default
1784 */
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001785static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1786 int nodefs)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001787{
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001788 struct ext4_sb_info *sbi = EXT4_SB(sb);
1789 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001790 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001791 const struct mount_opts *m;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001792 char sep = nodefs ? '\n' : ',';
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001793
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001794#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1795#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001796
1797 if (sbi->s_sb_block != 1)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001798 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1799
1800 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1801 int want_set = m->flags & MOPT_SET;
1802 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1803 (m->flags & MOPT_CLEAR_ERR))
1804 continue;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001805 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001806 continue; /* skip if same as the default */
1807 if ((want_set &&
1808 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1809 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1810 continue; /* select Opt_noFoo vs Opt_Foo */
1811 SEQ_OPTS_PRINT("%s", token2str(m->token));
1812 }
1813
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001814 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001815 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001816 SEQ_OPTS_PRINT("resuid=%u",
1817 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1818 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001819 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001820 SEQ_OPTS_PRINT("resgid=%u",
1821 from_kgid_munged(&init_user_ns, sbi->s_resgid));
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001822 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001823 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1824 SEQ_OPTS_PUTS("errors=remount-ro");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001825 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001826 SEQ_OPTS_PUTS("errors=continue");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001827 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001828 SEQ_OPTS_PUTS("errors=panic");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001829 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001830 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001831 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001832 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001833 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001834 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001835 if (sb->s_flags & MS_I_VERSION)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001836 SEQ_OPTS_PUTS("i_version");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001837 if (nodefs || sbi->s_stripe)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001838 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001839 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001840 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1841 SEQ_OPTS_PUTS("data=journal");
1842 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1843 SEQ_OPTS_PUTS("data=ordered");
1844 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1845 SEQ_OPTS_PUTS("data=writeback");
1846 }
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001847 if (nodefs ||
1848 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001849 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1850 sbi->s_inode_readahead_blks);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001851
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001852 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1853 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001854 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
Theodore Ts'odf981d02012-08-17 09:48:17 -04001855 if (nodefs || sbi->s_max_dir_size_kb)
1856 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001857
1858 ext4_show_quota_options(seq, sb);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001859 return 0;
1860}
1861
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001862static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1863{
1864 return _ext4_show_options(seq, root->d_sb, 0);
1865}
1866
1867static int options_seq_show(struct seq_file *seq, void *offset)
1868{
1869 struct super_block *sb = seq->private;
1870 int rc;
1871
1872 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1873 rc = _ext4_show_options(seq, sb, 1);
1874 seq_puts(seq, "\n");
1875 return rc;
1876}
1877
1878static int options_open_fs(struct inode *inode, struct file *file)
1879{
Al Virod9dda782013-03-31 18:16:14 -04001880 return single_open(file, options_seq_show, PDE_DATA(inode));
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001881}
1882
1883static const struct file_operations ext4_seq_options_fops = {
1884 .owner = THIS_MODULE,
1885 .open = options_open_fs,
1886 .read = seq_read,
1887 .llseek = seq_lseek,
1888 .release = single_release,
1889};
1890
Mingming Cao617ba132006-10-11 01:20:53 -07001891static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001892 int read_only)
1893{
Mingming Cao617ba132006-10-11 01:20:53 -07001894 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001895 int res = 0;
1896
Mingming Cao617ba132006-10-11 01:20:53 -07001897 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001898 ext4_msg(sb, KERN_ERR, "revision level too high, "
1899 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001900 res = MS_RDONLY;
1901 }
1902 if (read_only)
Theodore Ts'o281b5992011-09-09 18:34:51 -04001903 goto done;
Mingming Cao617ba132006-10-11 01:20:53 -07001904 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001905 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1906 "running e2fsck is recommended");
Lukas Czernerc8b459f2014-05-12 12:55:07 -04001907 else if (sbi->s_mount_state & EXT4_ERROR_FS)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001908 ext4_msg(sb, KERN_WARNING,
1909 "warning: mounting fs with errors, "
1910 "running e2fsck is recommended");
Tao Maed3ce802011-05-18 13:29:57 -04001911 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001912 le16_to_cpu(es->s_mnt_count) >=
1913 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001914 ext4_msg(sb, KERN_WARNING,
1915 "warning: maximal mount count reached, "
1916 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001917 else if (le32_to_cpu(es->s_checkinterval) &&
1918 (le32_to_cpu(es->s_lastcheck) +
1919 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001920 ext4_msg(sb, KERN_WARNING,
1921 "warning: checktime reached, "
1922 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001923 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001924 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001925 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001926 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001927 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001928 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001929 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001930 if (sbi->s_journal)
1931 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001932
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001933 ext4_commit_super(sb, 1);
Theodore Ts'o281b5992011-09-09 18:34:51 -04001934done:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001935 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001936 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001937 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001938 sb->s_blocksize,
1939 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001940 EXT4_BLOCKS_PER_GROUP(sb),
1941 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001942 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001943
Dan Magenheimer7abc52c2011-05-26 10:02:03 -06001944 cleancache_init_fs(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001945 return res;
1946}
1947
Theodore Ts'o117fff12012-09-05 01:29:50 -04001948int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1949{
1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 struct flex_groups *new_groups;
1952 int size;
1953
1954 if (!sbi->s_log_groups_per_flex)
1955 return 0;
1956
1957 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1958 if (size <= sbi->s_flex_groups_allocated)
1959 return 0;
1960
1961 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1962 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1963 if (!new_groups) {
1964 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1965 size / (int) sizeof(struct flex_groups));
1966 return -ENOMEM;
1967 }
1968
1969 if (sbi->s_flex_groups) {
1970 memcpy(new_groups, sbi->s_flex_groups,
1971 (sbi->s_flex_groups_allocated *
1972 sizeof(struct flex_groups)));
1973 ext4_kvfree(sbi->s_flex_groups);
1974 }
1975 sbi->s_flex_groups = new_groups;
1976 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1977 return 0;
1978}
1979
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001980static int ext4_fill_flex_info(struct super_block *sb)
1981{
1982 struct ext4_sb_info *sbi = EXT4_SB(sb);
1983 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001984 ext4_group_t flex_group;
Theodore Ts'o117fff12012-09-05 01:29:50 -04001985 int i, err;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001986
Theodore Ts'o503358a2009-11-23 07:24:46 -05001987 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
Xi Wangd50f2ab2012-01-10 11:51:10 -05001988 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001989 sbi->s_log_groups_per_flex = 0;
1990 return 1;
1991 }
1992
Theodore Ts'o117fff12012-09-05 01:29:50 -04001993 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1994 if (err)
Theodore Ts'o9933fc02011-08-01 08:45:02 -04001995 goto failed;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001996
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001997 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001998 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001999
2000 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002001 atomic_add(ext4_free_inodes_count(sb, gdp),
2002 &sbi->s_flex_groups[flex_group].free_inodes);
Theodore Ts'o90ba9832013-03-11 23:39:59 -04002003 atomic64_add(ext4_free_group_clusters(sb, gdp),
2004 &sbi->s_flex_groups[flex_group].free_clusters);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002005 atomic_add(ext4_used_dirs_count(sb, gdp),
2006 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002007 }
2008
2009 return 1;
2010failed:
2011 return 0;
2012}
2013
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002014static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2015 struct ext4_group_desc *gdp)
Andreas Dilger717d50e2007-10-16 18:38:25 -04002016{
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002017 int offset;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002018 __u16 crc = 0;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002019 __le32 le_group = cpu_to_le32(block_group);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002020
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002021 if ((sbi->s_es->s_feature_ro_compat &
2022 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2023 /* Use new metadata_csum algorithm */
Theodore Ts'od6a77102013-04-09 23:59:55 -04002024 __le16 save_csum;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002025 __u32 csum32;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002026
Theodore Ts'od6a77102013-04-09 23:59:55 -04002027 save_csum = gdp->bg_checksum;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002028 gdp->bg_checksum = 0;
2029 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2030 sizeof(le_group));
2031 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2032 sbi->s_desc_size);
Theodore Ts'od6a77102013-04-09 23:59:55 -04002033 gdp->bg_checksum = save_csum;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002034
2035 crc = csum32 & 0xFFFF;
2036 goto out;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002037 }
2038
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002039 /* old crc16 code */
2040 offset = offsetof(struct ext4_group_desc, bg_checksum);
2041
2042 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2043 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2044 crc = crc16(crc, (__u8 *)gdp, offset);
2045 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2046 /* for checksum of struct ext4_group_desc do the rest...*/
2047 if ((sbi->s_es->s_feature_incompat &
2048 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2049 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2050 crc = crc16(crc, (__u8 *)gdp + offset,
2051 le16_to_cpu(sbi->s_es->s_desc_size) -
2052 offset);
2053
2054out:
Andreas Dilger717d50e2007-10-16 18:38:25 -04002055 return cpu_to_le16(crc);
2056}
2057
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002058int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
Andreas Dilger717d50e2007-10-16 18:38:25 -04002059 struct ext4_group_desc *gdp)
2060{
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002061 if (ext4_has_group_desc_csum(sb) &&
2062 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2063 block_group, gdp)))
Andreas Dilger717d50e2007-10-16 18:38:25 -04002064 return 0;
2065
2066 return 1;
2067}
2068
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002069void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2070 struct ext4_group_desc *gdp)
2071{
2072 if (!ext4_has_group_desc_csum(sb))
2073 return;
2074 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2075}
2076
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002077/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002078static int ext4_check_descriptors(struct super_block *sb,
2079 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002080{
Mingming Cao617ba132006-10-11 01:20:53 -07002081 struct ext4_sb_info *sbi = EXT4_SB(sb);
2082 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2083 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002084 ext4_fsblk_t block_bitmap;
2085 ext4_fsblk_t inode_bitmap;
2086 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002087 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002088 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002089
Jose R. Santosce421582007-10-16 18:38:25 -04002090 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2091 flexbg_flag = 1;
2092
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002093 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002094
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002095 for (i = 0; i < sbi->s_groups_count; i++) {
2096 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2097
Jose R. Santosce421582007-10-16 18:38:25 -04002098 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002099 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002100 else
2101 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002102 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002103
Lukas Czernerbfff6872010-10-27 21:30:05 -04002104 if ((grp == sbi->s_groups_count) &&
2105 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2106 grp = i;
2107
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002108 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002109 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002111 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002112 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002113 return 0;
2114 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002115 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002116 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002117 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002118 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002119 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002120 return 0;
2121 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002122 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002123 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002124 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002126 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002127 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002128 return 0;
2129 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002130 ext4_lock_group(sb, i);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002131 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002132 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2133 "Checksum for group %u failed (%u!=%u)",
2134 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2135 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002136 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002137 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002138 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002139 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002140 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002141 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002142 if (!flexbg_flag)
2143 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002144 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002145 if (NULL != first_not_zeroed)
2146 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002147
Theodore Ts'o5dee5432011-09-09 19:10:51 -04002148 ext4_free_blocks_count_set(sbi->s_es,
2149 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002150 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002151 return 1;
2152}
2153
Mingming Cao617ba132006-10-11 01:20:53 -07002154/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002155 * the superblock) which were deleted from all directories, but held open by
2156 * a process at the time of a crash. We walk the list and try to delete these
2157 * inodes at recovery time (only with a read-write filesystem).
2158 *
2159 * In order to keep the orphan inode chain consistent during traversal (in
2160 * case of crash during recovery), we link each inode into the superblock
2161 * orphan list_head and handle it the same way as an inode deletion during
2162 * normal operation (which journals the operations for us).
2163 *
2164 * We only do an iget() and an iput() on each inode, which is very safe if we
2165 * accidentally point at an in-use or already deleted inode. The worst that
2166 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002167 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002168 * e2fsck was run on this filesystem, and it must have already done the orphan
2169 * inode cleanup for us, so we can safely abort without any further action.
2170 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002171static void ext4_orphan_cleanup(struct super_block *sb,
2172 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002173{
2174 unsigned int s_flags = sb->s_flags;
2175 int nr_orphans = 0, nr_truncates = 0;
2176#ifdef CONFIG_QUOTA
2177 int i;
2178#endif
2179 if (!es->s_last_orphan) {
2180 jbd_debug(4, "no orphan inodes to clean up\n");
2181 return;
2182 }
2183
Eric Sandeena8f48a92006-12-06 20:40:13 -08002184 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002185 ext4_msg(sb, KERN_ERR, "write access "
2186 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002187 return;
2188 }
2189
Amir Goldsteind39195c2011-02-28 00:53:45 -05002190 /* Check if feature set would not allow a r/w mount */
2191 if (!ext4_feature_set_ok(sb, 0)) {
2192 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2193 "unknown ROCOMPAT features");
2194 return;
2195 }
2196
Mingming Cao617ba132006-10-11 01:20:53 -07002197 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002198 /* don't clear list on RO mount w/ errors */
2199 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002200 jbd_debug(1, "Errors on filesystem, "
2201 "clearing orphan list.\n");
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002202 es->s_last_orphan = 0;
2203 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002204 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2205 return;
2206 }
2207
2208 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002209 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002210 sb->s_flags &= ~MS_RDONLY;
2211 }
2212#ifdef CONFIG_QUOTA
2213 /* Needed for iput() to work correctly and not trash data */
2214 sb->s_flags |= MS_ACTIVE;
2215 /* Turn on quotas so that they are updated correctly */
2216 for (i = 0; i < MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002217 if (EXT4_SB(sb)->s_qf_names[i]) {
2218 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002219 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002220 ext4_msg(sb, KERN_ERR,
2221 "Cannot turn on journaled "
2222 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002223 }
2224 }
2225#endif
2226
2227 while (es->s_last_orphan) {
2228 struct inode *inode;
2229
Josef Bacik97bd42b2008-04-29 22:04:56 -04002230 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2231 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002232 es->s_last_orphan = 0;
2233 break;
2234 }
2235
Mingming Cao617ba132006-10-11 01:20:53 -07002236 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002237 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002238 if (inode->i_nlink) {
Paul Taysom566370a2013-05-28 07:51:21 -04002239 if (test_opt(sb, DEBUG))
2240 ext4_msg(sb, KERN_DEBUG,
2241 "%s: truncating inode %lu to %lld bytes",
2242 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002243 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002244 inode->i_ino, inode->i_size);
Theodore Ts'o721e3eb2012-12-27 01:42:48 -05002245 mutex_lock(&inode->i_mutex);
Lukas Czerner55f252c2013-05-27 23:32:35 -04002246 truncate_inode_pages(inode->i_mapping, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002247 ext4_truncate(inode);
Theodore Ts'o721e3eb2012-12-27 01:42:48 -05002248 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002249 nr_truncates++;
2250 } else {
Paul Taysom566370a2013-05-28 07:51:21 -04002251 if (test_opt(sb, DEBUG))
2252 ext4_msg(sb, KERN_DEBUG,
2253 "%s: deleting unreferenced inode %lu",
2254 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002255 jbd_debug(2, "deleting unreferenced inode %lu\n",
2256 inode->i_ino);
2257 nr_orphans++;
2258 }
2259 iput(inode); /* The delete magic happens here! */
2260 }
2261
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002262#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002263
2264 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002265 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2266 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002267 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002268 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2269 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002270#ifdef CONFIG_QUOTA
2271 /* Turn quotas off */
2272 for (i = 0; i < MAXQUOTAS; i++) {
2273 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002274 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002275 }
2276#endif
2277 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2278}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002279
Eric Sandeencd2291a2008-01-28 23:58:27 -05002280/*
2281 * Maximal extent format file size.
2282 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2283 * extent format containers, within a sector_t, and within i_blocks
2284 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2285 * so that won't be a limiting factor.
2286 *
Lukas Czernerf17722f2011-06-06 00:05:17 -04002287 * However there is other limiting factor. We do store extents in the form
2288 * of starting block and length, hence the resulting length of the extent
2289 * covering maximum file size must fit into on-disk format containers as
2290 * well. Given that length is always by 1 unit bigger than max unit (because
2291 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2292 *
Eric Sandeencd2291a2008-01-28 23:58:27 -05002293 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2294 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002295static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002296{
2297 loff_t res;
2298 loff_t upper_limit = MAX_LFS_FILESIZE;
2299
2300 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002301 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002302 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002303 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002304 * i_block represent total blocks in 512 bytes
2305 * 32 == size of vfs inode i_blocks * 8
2306 */
2307 upper_limit = (1LL << 32) - 1;
2308
2309 /* total blocks in file system block size */
2310 upper_limit >>= (blkbits - 9);
2311 upper_limit <<= blkbits;
2312 }
2313
Lukas Czernerf17722f2011-06-06 00:05:17 -04002314 /*
2315 * 32-bit extent-start container, ee_block. We lower the maxbytes
2316 * by one fs block, so ee_len can cover the extent of maximum file
2317 * size
2318 */
2319 res = (1LL << 32) - 1;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002320 res <<= blkbits;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002321
2322 /* Sanity check against vm- & vfs- imposed limits */
2323 if (res > upper_limit)
2324 res = upper_limit;
2325
2326 return res;
2327}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002328
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002329/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002330 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002331 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2332 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002333 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002334static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002335{
Mingming Cao617ba132006-10-11 01:20:53 -07002336 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002337 int meta_blocks;
2338 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002339 /* This is calculated to be the largest file size for a dense, block
2340 * mapped file such that the file's total number of 512-byte sectors,
2341 * including data and all indirect blocks, does not exceed (2^48 - 1).
2342 *
2343 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2344 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002345 */
2346
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002347 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002348 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002349 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002350 * the inode i_block field represents total file blocks in
2351 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002352 */
2353 upper_limit = (1LL << 32) - 1;
2354
2355 /* total blocks in file system block size */
2356 upper_limit >>= (bits - 9);
2357
2358 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002359 /*
2360 * We use 48 bit ext4_inode i_blocks
2361 * With EXT4_HUGE_FILE_FL set the i_blocks
2362 * represent total number of blocks in
2363 * file system block size
2364 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002365 upper_limit = (1LL << 48) - 1;
2366
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002367 }
2368
2369 /* indirect blocks */
2370 meta_blocks = 1;
2371 /* double indirect blocks */
2372 meta_blocks += 1 + (1LL << (bits-2));
2373 /* tripple indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2375
2376 upper_limit -= meta_blocks;
2377 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002378
2379 res += 1LL << (bits-2);
2380 res += 1LL << (2*(bits-2));
2381 res += 1LL << (3*(bits-2));
2382 res <<= bits;
2383 if (res > upper_limit)
2384 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002385
2386 if (res > MAX_LFS_FILESIZE)
2387 res = MAX_LFS_FILESIZE;
2388
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002389 return res;
2390}
2391
Mingming Cao617ba132006-10-11 01:20:53 -07002392static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002393 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002394{
Mingming Cao617ba132006-10-11 01:20:53 -07002395 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002396 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002397 int has_super = 0;
2398
2399 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2400
Mingming Cao617ba132006-10-11 01:20:53 -07002401 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002402 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002403 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002404 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002405 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002406 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002407
Darrick J. Wongbd63f6b2014-05-12 10:06:27 -04002408 /*
2409 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2410 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2411 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2412 * compensate.
2413 */
2414 if (sb->s_blocksize == 1024 && nr == 0 &&
2415 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2416 has_super++;
2417
Mingming Cao617ba132006-10-11 01:20:53 -07002418 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002419}
2420
Alex Tomasc9de5602008-01-29 00:19:52 -05002421/**
2422 * ext4_get_stripe_size: Get the stripe size.
2423 * @sbi: In memory super block info
2424 *
2425 * If we have specified it via mount option, then
2426 * use the mount option value. If the value specified at mount time is
2427 * greater than the blocks per group use the super block value.
2428 * If the super block value is greater than blocks per group return 0.
2429 * Allocator needs it be less than blocks per group.
2430 *
2431 */
2432static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2433{
2434 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2435 unsigned long stripe_width =
2436 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002437 int ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002438
2439 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002440 ret = sbi->s_stripe;
2441 else if (stripe_width <= sbi->s_blocks_per_group)
2442 ret = stripe_width;
2443 else if (stride <= sbi->s_blocks_per_group)
2444 ret = stride;
2445 else
2446 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002447
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002448 /*
2449 * If the stripe width is 1, this makes no sense and
2450 * we set it to 0 to turn off stripe handling code.
2451 */
2452 if (ret <= 1)
2453 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002454
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002455 return ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002456}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002457
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002458/* sysfs supprt */
2459
2460struct ext4_attr {
2461 struct attribute attr;
2462 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
Theodore Ts'o60e66792010-05-17 07:00:00 -04002463 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002464 const char *, size_t);
Jan Karaf2d50a62013-06-04 12:51:16 -04002465 union {
2466 int offset;
2467 int deprecated_val;
2468 } u;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002469};
2470
Lukas Czerner27dd4382013-04-09 22:11:22 -04002471static int parse_strtoull(const char *buf,
2472 unsigned long long max, unsigned long long *value)
2473{
2474 int ret;
2475
2476 ret = kstrtoull(skip_spaces(buf), 0, value);
2477 if (!ret && *value > max)
2478 ret = -EINVAL;
2479 return ret;
2480}
2481
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002482static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi,
2484 char *buf)
2485{
2486 return snprintf(buf, PAGE_SIZE, "%llu\n",
Aditya Kali7b415bf2011-09-09 19:04:51 -04002487 (s64) EXT4_C2B(sbi,
2488 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002489}
2490
2491static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi, char *buf)
2493{
2494 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2495
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002496 if (!sb->s_bdev->bd_part)
2497 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002498 return snprintf(buf, PAGE_SIZE, "%lu\n",
2499 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2500 sbi->s_sectors_written_start) >> 1);
2501}
2502
2503static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi, char *buf)
2505{
2506 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2507
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002508 if (!sb->s_bdev->bd_part)
2509 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002510 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002511 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002512 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002513 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002514}
2515
2516static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519{
2520 unsigned long t;
Lukas Czernere1091b12013-04-11 23:37:19 -04002521 int ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002522
Lukas Czernere1091b12013-04-11 23:37:19 -04002523 ret = kstrtoul(skip_spaces(buf), 0, &t);
2524 if (ret)
2525 return ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002526
Lukas Czernere1091b12013-04-11 23:37:19 -04002527 if (t && (!is_power_of_2(t) || t > 0x40000000))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002528 return -EINVAL;
2529
2530 sbi->s_inode_readahead_blks = t;
2531 return count;
2532}
2533
2534static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002535 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002536{
Jan Karaf2d50a62013-06-04 12:51:16 -04002537 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002538
2539 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2540}
2541
2542static ssize_t sbi_ui_store(struct ext4_attr *a,
2543 struct ext4_sb_info *sbi,
2544 const char *buf, size_t count)
2545{
Jan Karaf2d50a62013-06-04 12:51:16 -04002546 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002547 unsigned long t;
Lukas Czernere1091b12013-04-11 23:37:19 -04002548 int ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002549
Lukas Czernere1091b12013-04-11 23:37:19 -04002550 ret = kstrtoul(skip_spaces(buf), 0, &t);
2551 if (ret)
2552 return ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002553 *ui = t;
2554 return count;
2555}
2556
Lukas Czerner27dd4382013-04-09 22:11:22 -04002557static ssize_t reserved_clusters_show(struct ext4_attr *a,
2558 struct ext4_sb_info *sbi, char *buf)
2559{
2560 return snprintf(buf, PAGE_SIZE, "%llu\n",
2561 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2562}
2563
2564static ssize_t reserved_clusters_store(struct ext4_attr *a,
2565 struct ext4_sb_info *sbi,
2566 const char *buf, size_t count)
2567{
2568 unsigned long long val;
2569 int ret;
2570
2571 if (parse_strtoull(buf, -1ULL, &val))
2572 return -EINVAL;
2573 ret = ext4_reserve_clusters(sbi, val);
2574
2575 return ret ? ret : count;
2576}
2577
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04002578static ssize_t trigger_test_error(struct ext4_attr *a,
2579 struct ext4_sb_info *sbi,
2580 const char *buf, size_t count)
2581{
2582 int len = count;
2583
2584 if (!capable(CAP_SYS_ADMIN))
2585 return -EPERM;
2586
2587 if (len && buf[len-1] == '\n')
2588 len--;
2589
2590 if (len)
2591 ext4_error(sbi->s_sb, "%.*s", len, buf);
2592 return count;
2593}
2594
Jan Karaf2d50a62013-06-04 12:51:16 -04002595static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2596 struct ext4_sb_info *sbi, char *buf)
2597{
2598 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2599}
2600
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002601#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2602static struct ext4_attr ext4_attr_##_name = { \
2603 .attr = {.name = __stringify(_name), .mode = _mode }, \
2604 .show = _show, \
2605 .store = _store, \
Jan Karaf2d50a62013-06-04 12:51:16 -04002606 .u = { \
2607 .offset = offsetof(struct ext4_sb_info, _elname),\
2608 }, \
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002609}
2610#define EXT4_ATTR(name, mode, show, store) \
2611static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2612
Lukas Czerner857ac882010-10-27 21:30:05 -04002613#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002614#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2615#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2616#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2617 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2618#define ATTR_LIST(name) &ext4_attr_##name.attr
Jan Karaf2d50a62013-06-04 12:51:16 -04002619#define EXT4_DEPRECATED_ATTR(_name, _val) \
2620static struct ext4_attr ext4_attr_##_name = { \
2621 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2622 .show = sbi_deprecated_show, \
2623 .u = { \
2624 .deprecated_val = _val, \
2625 }, \
2626}
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002627
2628EXT4_RO_ATTR(delayed_allocation_blocks);
2629EXT4_RO_ATTR(session_write_kbytes);
2630EXT4_RO_ATTR(lifetime_write_kbytes);
Lukas Czerner27dd4382013-04-09 22:11:22 -04002631EXT4_RW_ATTR(reserved_clusters);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002632EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2633 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002634EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002635EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2636EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2637EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2638EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2639EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2640EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Jan Karaf2d50a62013-06-04 12:51:16 -04002641EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
Zheng Liu67a5da52012-08-17 09:54:17 -04002642EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04002643EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04002644EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2645EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2646EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2647EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2648EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2649EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002650
2651static struct attribute *ext4_attrs[] = {
2652 ATTR_LIST(delayed_allocation_blocks),
2653 ATTR_LIST(session_write_kbytes),
2654 ATTR_LIST(lifetime_write_kbytes),
Lukas Czerner27dd4382013-04-09 22:11:22 -04002655 ATTR_LIST(reserved_clusters),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002656 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002657 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002658 ATTR_LIST(mb_stats),
2659 ATTR_LIST(mb_max_to_scan),
2660 ATTR_LIST(mb_min_to_scan),
2661 ATTR_LIST(mb_order2_req),
2662 ATTR_LIST(mb_stream_req),
2663 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002664 ATTR_LIST(max_writeback_mb_bump),
Zheng Liu67a5da52012-08-17 09:54:17 -04002665 ATTR_LIST(extent_max_zeroout_kb),
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04002666 ATTR_LIST(trigger_fs_error),
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04002667 ATTR_LIST(err_ratelimit_interval_ms),
2668 ATTR_LIST(err_ratelimit_burst),
2669 ATTR_LIST(warning_ratelimit_interval_ms),
2670 ATTR_LIST(warning_ratelimit_burst),
2671 ATTR_LIST(msg_ratelimit_interval_ms),
2672 ATTR_LIST(msg_ratelimit_burst),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002673 NULL,
2674};
2675
Lukas Czerner857ac882010-10-27 21:30:05 -04002676/* Features this copy of ext4 supports */
2677EXT4_INFO_ATTR(lazy_itable_init);
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002678EXT4_INFO_ATTR(batched_discard);
Theodore Ts'o5e7bbef2012-09-13 12:11:40 -04002679EXT4_INFO_ATTR(meta_bg_resize);
Lukas Czerner857ac882010-10-27 21:30:05 -04002680
2681static struct attribute *ext4_feat_attrs[] = {
2682 ATTR_LIST(lazy_itable_init),
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002683 ATTR_LIST(batched_discard),
Theodore Ts'o5e7bbef2012-09-13 12:11:40 -04002684 ATTR_LIST(meta_bg_resize),
Lukas Czerner857ac882010-10-27 21:30:05 -04002685 NULL,
2686};
2687
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002688static ssize_t ext4_attr_show(struct kobject *kobj,
2689 struct attribute *attr, char *buf)
2690{
2691 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2692 s_kobj);
2693 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2694
2695 return a->show ? a->show(a, sbi, buf) : 0;
2696}
2697
2698static ssize_t ext4_attr_store(struct kobject *kobj,
2699 struct attribute *attr,
2700 const char *buf, size_t len)
2701{
2702 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2703 s_kobj);
2704 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2705
2706 return a->store ? a->store(a, sbi, buf, len) : 0;
2707}
2708
2709static void ext4_sb_release(struct kobject *kobj)
2710{
2711 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2712 s_kobj);
2713 complete(&sbi->s_kobj_unregister);
2714}
2715
Emese Revfy52cf25d2010-01-19 02:58:23 +01002716static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002717 .show = ext4_attr_show,
2718 .store = ext4_attr_store,
2719};
2720
2721static struct kobj_type ext4_ktype = {
2722 .default_attrs = ext4_attrs,
2723 .sysfs_ops = &ext4_attr_ops,
2724 .release = ext4_sb_release,
2725};
2726
Lukas Czerner857ac882010-10-27 21:30:05 -04002727static void ext4_feat_release(struct kobject *kobj)
2728{
2729 complete(&ext4_feat->f_kobj_unregister);
2730}
2731
2732static struct kobj_type ext4_feat_ktype = {
2733 .default_attrs = ext4_feat_attrs,
2734 .sysfs_ops = &ext4_attr_ops,
2735 .release = ext4_feat_release,
2736};
2737
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002738/*
2739 * Check whether this filesystem can be mounted based on
2740 * the features present and the RDONLY/RDWR mount requested.
2741 * Returns 1 if this filesystem can be mounted as requested,
2742 * 0 if it cannot be.
2743 */
2744static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2745{
2746 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2747 ext4_msg(sb, KERN_ERR,
2748 "Couldn't mount because of "
2749 "unsupported optional features (%x)",
2750 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2751 ~EXT4_FEATURE_INCOMPAT_SUPP));
2752 return 0;
2753 }
2754
2755 if (readonly)
2756 return 1;
2757
2758 /* Check that feature set is OK for a read-write mount */
2759 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2760 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2761 "unsupported optional features (%x)",
2762 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2763 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2764 return 0;
2765 }
2766 /*
2767 * Large file size enabled file system can only be mounted
2768 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2769 */
2770 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2771 if (sizeof(blkcnt_t) < sizeof(u64)) {
2772 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2773 "cannot be mounted RDWR without "
2774 "CONFIG_LBDAF");
2775 return 0;
2776 }
2777 }
Theodore Ts'obab08ab2011-09-09 18:36:51 -04002778 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2779 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2780 ext4_msg(sb, KERN_ERR,
2781 "Can't support bigalloc feature without "
2782 "extents feature\n");
2783 return 0;
2784 }
Aditya Kali7c319d32012-07-22 20:21:31 -04002785
2786#ifndef CONFIG_QUOTA
2787 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2788 !readonly) {
2789 ext4_msg(sb, KERN_ERR,
2790 "Filesystem with quota feature cannot be mounted RDWR "
2791 "without CONFIG_QUOTA");
2792 return 0;
2793 }
2794#endif /* CONFIG_QUOTA */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002795 return 1;
2796}
2797
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002798/*
2799 * This function is called once a day if we have errors logged
2800 * on the file system
2801 */
2802static void print_daily_error_info(unsigned long arg)
2803{
2804 struct super_block *sb = (struct super_block *) arg;
2805 struct ext4_sb_info *sbi;
2806 struct ext4_super_block *es;
2807
2808 sbi = EXT4_SB(sb);
2809 es = sbi->s_es;
2810
2811 if (es->s_error_count)
2812 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2813 le32_to_cpu(es->s_error_count));
2814 if (es->s_first_error_time) {
2815 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2816 sb->s_id, le32_to_cpu(es->s_first_error_time),
2817 (int) sizeof(es->s_first_error_func),
2818 es->s_first_error_func,
2819 le32_to_cpu(es->s_first_error_line));
2820 if (es->s_first_error_ino)
2821 printk(": inode %u",
2822 le32_to_cpu(es->s_first_error_ino));
2823 if (es->s_first_error_block)
2824 printk(": block %llu", (unsigned long long)
2825 le64_to_cpu(es->s_first_error_block));
2826 printk("\n");
2827 }
2828 if (es->s_last_error_time) {
2829 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2830 sb->s_id, le32_to_cpu(es->s_last_error_time),
2831 (int) sizeof(es->s_last_error_func),
2832 es->s_last_error_func,
2833 le32_to_cpu(es->s_last_error_line));
2834 if (es->s_last_error_ino)
2835 printk(": inode %u",
2836 le32_to_cpu(es->s_last_error_ino));
2837 if (es->s_last_error_block)
2838 printk(": block %llu", (unsigned long long)
2839 le64_to_cpu(es->s_last_error_block));
2840 printk("\n");
2841 }
2842 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2843}
2844
Lukas Czernerbfff6872010-10-27 21:30:05 -04002845/* Find next suitable group and run ext4_init_inode_table */
2846static int ext4_run_li_request(struct ext4_li_request *elr)
2847{
2848 struct ext4_group_desc *gdp = NULL;
2849 ext4_group_t group, ngroups;
2850 struct super_block *sb;
2851 unsigned long timeout = 0;
2852 int ret = 0;
2853
2854 sb = elr->lr_super;
2855 ngroups = EXT4_SB(sb)->s_groups_count;
2856
Jan Kara8e8ad8a2012-06-12 16:20:38 +02002857 sb_start_write(sb);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002858 for (group = elr->lr_next_group; group < ngroups; group++) {
2859 gdp = ext4_get_group_desc(sb, group, NULL);
2860 if (!gdp) {
2861 ret = 1;
2862 break;
2863 }
2864
2865 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2866 break;
2867 }
2868
Theodore Ts'o7f511862013-01-13 08:41:45 -05002869 if (group >= ngroups)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002870 ret = 1;
2871
2872 if (!ret) {
2873 timeout = jiffies;
2874 ret = ext4_init_inode_table(sb, group,
2875 elr->lr_timeout ? 0 : 1);
2876 if (elr->lr_timeout == 0) {
Lukas Czerner51ce6512011-05-20 13:55:16 -04002877 timeout = (jiffies - timeout) *
2878 elr->lr_sbi->s_li_wait_mult;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002879 elr->lr_timeout = timeout;
2880 }
2881 elr->lr_next_sched = jiffies + elr->lr_timeout;
2882 elr->lr_next_group = group + 1;
2883 }
Jan Kara8e8ad8a2012-06-12 16:20:38 +02002884 sb_end_write(sb);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002885
2886 return ret;
2887}
2888
2889/*
2890 * Remove lr_request from the list_request and free the
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002891 * request structure. Should be called with li_list_mtx held
Lukas Czernerbfff6872010-10-27 21:30:05 -04002892 */
2893static void ext4_remove_li_request(struct ext4_li_request *elr)
2894{
2895 struct ext4_sb_info *sbi;
2896
2897 if (!elr)
2898 return;
2899
2900 sbi = elr->lr_sbi;
2901
2902 list_del(&elr->lr_request);
2903 sbi->s_li_request = NULL;
2904 kfree(elr);
2905}
2906
2907static void ext4_unregister_li_request(struct super_block *sb)
2908{
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002909 mutex_lock(&ext4_li_mtx);
2910 if (!ext4_li_info) {
2911 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002912 return;
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002913 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002914
2915 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002916 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002917 mutex_unlock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002918 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002919}
2920
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002921static struct task_struct *ext4_lazyinit_task;
2922
Lukas Czernerbfff6872010-10-27 21:30:05 -04002923/*
2924 * This is the function where ext4lazyinit thread lives. It walks
2925 * through the request list searching for next scheduled filesystem.
2926 * When such a fs is found, run the lazy initialization request
2927 * (ext4_rn_li_request) and keep track of the time spend in this
2928 * function. Based on that time we compute next schedule time of
2929 * the request. When walking through the list is complete, compute
2930 * next waking time and put itself into sleep.
2931 */
2932static int ext4_lazyinit_thread(void *arg)
2933{
2934 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2935 struct list_head *pos, *n;
2936 struct ext4_li_request *elr;
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002937 unsigned long next_wakeup, cur;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002938
2939 BUG_ON(NULL == eli);
2940
Lukas Czernerbfff6872010-10-27 21:30:05 -04002941cont_thread:
2942 while (true) {
2943 next_wakeup = MAX_JIFFY_OFFSET;
2944
2945 mutex_lock(&eli->li_list_mtx);
2946 if (list_empty(&eli->li_request_list)) {
2947 mutex_unlock(&eli->li_list_mtx);
2948 goto exit_thread;
2949 }
2950
2951 list_for_each_safe(pos, n, &eli->li_request_list) {
2952 elr = list_entry(pos, struct ext4_li_request,
2953 lr_request);
2954
Theodore Ts'ob2c78cd2010-11-02 14:19:30 -04002955 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2956 if (ext4_run_li_request(elr) != 0) {
2957 /* error, remove the lazy_init job */
2958 ext4_remove_li_request(elr);
2959 continue;
2960 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002961 }
2962
2963 if (time_before(elr->lr_next_sched, next_wakeup))
2964 next_wakeup = elr->lr_next_sched;
2965 }
2966 mutex_unlock(&eli->li_list_mtx);
2967
Tejun Heoa0acae02011-11-21 12:32:22 -08002968 try_to_freeze();
Lukas Czernerbfff6872010-10-27 21:30:05 -04002969
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002970 cur = jiffies;
2971 if ((time_after_eq(cur, next_wakeup)) ||
Lukas Czernerf4245bd2010-11-02 14:07:17 -04002972 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04002973 cond_resched();
2974 continue;
2975 }
2976
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002977 schedule_timeout_interruptible(next_wakeup - cur);
2978
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002979 if (kthread_should_stop()) {
2980 ext4_clear_request_list();
2981 goto exit_thread;
2982 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002983 }
2984
2985exit_thread:
2986 /*
2987 * It looks like the request list is empty, but we need
2988 * to check it under the li_list_mtx lock, to prevent any
2989 * additions into it, and of course we should lock ext4_li_mtx
2990 * to atomically free the list and ext4_li_info, because at
2991 * this point another ext4 filesystem could be registering
2992 * new one.
2993 */
2994 mutex_lock(&ext4_li_mtx);
2995 mutex_lock(&eli->li_list_mtx);
2996 if (!list_empty(&eli->li_request_list)) {
2997 mutex_unlock(&eli->li_list_mtx);
2998 mutex_unlock(&ext4_li_mtx);
2999 goto cont_thread;
3000 }
3001 mutex_unlock(&eli->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003002 kfree(ext4_li_info);
3003 ext4_li_info = NULL;
3004 mutex_unlock(&ext4_li_mtx);
3005
3006 return 0;
3007}
3008
3009static void ext4_clear_request_list(void)
3010{
3011 struct list_head *pos, *n;
3012 struct ext4_li_request *elr;
3013
3014 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003015 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3016 elr = list_entry(pos, struct ext4_li_request,
3017 lr_request);
3018 ext4_remove_li_request(elr);
3019 }
3020 mutex_unlock(&ext4_li_info->li_list_mtx);
3021}
3022
3023static int ext4_run_lazyinit_thread(void)
3024{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003025 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3026 ext4_li_info, "ext4lazyinit");
3027 if (IS_ERR(ext4_lazyinit_task)) {
3028 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003029 ext4_clear_request_list();
Lukas Czernerbfff6872010-10-27 21:30:05 -04003030 kfree(ext4_li_info);
3031 ext4_li_info = NULL;
Theodore Ts'o92b97812012-03-19 23:41:49 -04003032 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
Lukas Czernerbfff6872010-10-27 21:30:05 -04003033 "initialization thread\n",
3034 err);
3035 return err;
3036 }
3037 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003038 return 0;
3039}
3040
3041/*
3042 * Check whether it make sense to run itable init. thread or not.
3043 * If there is at least one uninitialized inode table, return
3044 * corresponding group number, else the loop goes through all
3045 * groups and return total number of groups.
3046 */
3047static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3048{
3049 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3050 struct ext4_group_desc *gdp = NULL;
3051
3052 for (group = 0; group < ngroups; group++) {
3053 gdp = ext4_get_group_desc(sb, group, NULL);
3054 if (!gdp)
3055 continue;
3056
3057 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3058 break;
3059 }
3060
3061 return group;
3062}
3063
3064static int ext4_li_info_new(void)
3065{
3066 struct ext4_lazy_init *eli = NULL;
3067
3068 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3069 if (!eli)
3070 return -ENOMEM;
3071
Lukas Czernerbfff6872010-10-27 21:30:05 -04003072 INIT_LIST_HEAD(&eli->li_request_list);
3073 mutex_init(&eli->li_list_mtx);
3074
Lukas Czernerbfff6872010-10-27 21:30:05 -04003075 eli->li_state |= EXT4_LAZYINIT_QUIT;
3076
3077 ext4_li_info = eli;
3078
3079 return 0;
3080}
3081
3082static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3083 ext4_group_t start)
3084{
3085 struct ext4_sb_info *sbi = EXT4_SB(sb);
3086 struct ext4_li_request *elr;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003087
3088 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3089 if (!elr)
3090 return NULL;
3091
3092 elr->lr_super = sb;
3093 elr->lr_sbi = sbi;
3094 elr->lr_next_group = start;
3095
3096 /*
3097 * Randomize first schedule time of the request to
3098 * spread the inode table initialization requests
3099 * better.
3100 */
Theodore Ts'odd1f7232013-11-08 00:14:53 -05003101 elr->lr_next_sched = jiffies + (prandom_u32() %
3102 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
Lukas Czernerbfff6872010-10-27 21:30:05 -04003103 return elr;
3104}
3105
Theodore Ts'o7f511862013-01-13 08:41:45 -05003106int ext4_register_li_request(struct super_block *sb,
3107 ext4_group_t first_not_zeroed)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003108{
3109 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003110 struct ext4_li_request *elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003111 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05003112 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003113
Theodore Ts'o7f511862013-01-13 08:41:45 -05003114 mutex_lock(&ext4_li_mtx);
Lukas Czerner51ce6512011-05-20 13:55:16 -04003115 if (sbi->s_li_request != NULL) {
3116 /*
3117 * Reset timeout so it can be computed again, because
3118 * s_li_wait_mult might have changed.
3119 */
3120 sbi->s_li_request->lr_timeout = 0;
Theodore Ts'o7f511862013-01-13 08:41:45 -05003121 goto out;
Lukas Czerner51ce6512011-05-20 13:55:16 -04003122 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003123
3124 if (first_not_zeroed == ngroups ||
3125 (sb->s_flags & MS_RDONLY) ||
Tao Ma55ff3842011-05-09 10:28:41 -04003126 !test_opt(sb, INIT_INODE_TABLE))
Theodore Ts'o7f511862013-01-13 08:41:45 -05003127 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003128
3129 elr = ext4_li_request_new(sb, first_not_zeroed);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003130 if (!elr) {
3131 ret = -ENOMEM;
3132 goto out;
3133 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003134
3135 if (NULL == ext4_li_info) {
3136 ret = ext4_li_info_new();
3137 if (ret)
3138 goto out;
3139 }
3140
3141 mutex_lock(&ext4_li_info->li_list_mtx);
3142 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3143 mutex_unlock(&ext4_li_info->li_list_mtx);
3144
3145 sbi->s_li_request = elr;
Tao Ma46e46902011-04-04 16:00:49 -04003146 /*
3147 * set elr to NULL here since it has been inserted to
3148 * the request_list and the removal and free of it is
3149 * handled by ext4_clear_request_list from now on.
3150 */
3151 elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003152
3153 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3154 ret = ext4_run_lazyinit_thread();
3155 if (ret)
3156 goto out;
3157 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003158out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003159 mutex_unlock(&ext4_li_mtx);
3160 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003161 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003162 return ret;
3163}
3164
3165/*
3166 * We do not need to lock anything since this is called on
3167 * module unload.
3168 */
3169static void ext4_destroy_lazyinit_thread(void)
3170{
3171 /*
3172 * If thread exited earlier
3173 * there's nothing to be done.
3174 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003175 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003176 return;
3177
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003178 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003179}
3180
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003181static int set_journal_csum_feature_set(struct super_block *sb)
3182{
3183 int ret = 1;
3184 int compat, incompat;
3185 struct ext4_sb_info *sbi = EXT4_SB(sb);
3186
3187 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3188 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3189 /* journal checksum v2 */
3190 compat = 0;
3191 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3192 } else {
3193 /* journal checksum v1 */
3194 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3195 incompat = 0;
3196 }
3197
3198 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3199 ret = jbd2_journal_set_features(sbi->s_journal,
3200 compat, 0,
3201 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3202 incompat);
3203 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3204 ret = jbd2_journal_set_features(sbi->s_journal,
3205 compat, 0,
3206 incompat);
3207 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3208 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3209 } else {
3210 jbd2_journal_clear_features(sbi->s_journal,
3211 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3212 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3213 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3214 }
3215
3216 return ret;
3217}
3218
Theodore Ts'o952fc182012-07-09 16:27:05 -04003219/*
3220 * Note: calculating the overhead so we can be compatible with
3221 * historical BSD practice is quite difficult in the face of
3222 * clusters/bigalloc. This is because multiple metadata blocks from
3223 * different block group can end up in the same allocation cluster.
3224 * Calculating the exact overhead in the face of clustered allocation
3225 * requires either O(all block bitmaps) in memory or O(number of block
3226 * groups**2) in time. We will still calculate the superblock for
3227 * older file systems --- and if we come across with a bigalloc file
3228 * system with zero in s_overhead_clusters the estimate will be close to
3229 * correct especially for very large cluster sizes --- but for newer
3230 * file systems, it's better to calculate this figure once at mkfs
3231 * time, and store it in the superblock. If the superblock value is
3232 * present (even for non-bigalloc file systems), we will use it.
3233 */
3234static int count_overhead(struct super_block *sb, ext4_group_t grp,
3235 char *buf)
3236{
3237 struct ext4_sb_info *sbi = EXT4_SB(sb);
3238 struct ext4_group_desc *gdp;
3239 ext4_fsblk_t first_block, last_block, b;
3240 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3241 int s, j, count = 0;
3242
Theodore Ts'o0548bbb2012-08-16 11:59:04 -04003243 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3244 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3245 sbi->s_itb_per_group + 2);
3246
Theodore Ts'o952fc182012-07-09 16:27:05 -04003247 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3248 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3249 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3250 for (i = 0; i < ngroups; i++) {
3251 gdp = ext4_get_group_desc(sb, i, NULL);
3252 b = ext4_block_bitmap(sb, gdp);
3253 if (b >= first_block && b <= last_block) {
3254 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3255 count++;
3256 }
3257 b = ext4_inode_bitmap(sb, gdp);
3258 if (b >= first_block && b <= last_block) {
3259 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3260 count++;
3261 }
3262 b = ext4_inode_table(sb, gdp);
3263 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3264 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3265 int c = EXT4_B2C(sbi, b - first_block);
3266 ext4_set_bit(c, buf);
3267 count++;
3268 }
3269 if (i != grp)
3270 continue;
3271 s = 0;
3272 if (ext4_bg_has_super(sb, grp)) {
3273 ext4_set_bit(s++, buf);
3274 count++;
3275 }
3276 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3277 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3278 count++;
3279 }
3280 }
3281 if (!count)
3282 return 0;
3283 return EXT4_CLUSTERS_PER_GROUP(sb) -
3284 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3285}
3286
3287/*
3288 * Compute the overhead and stash it in sbi->s_overhead
3289 */
3290int ext4_calculate_overhead(struct super_block *sb)
3291{
3292 struct ext4_sb_info *sbi = EXT4_SB(sb);
3293 struct ext4_super_block *es = sbi->s_es;
3294 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3295 ext4_fsblk_t overhead = 0;
3296 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3297
Theodore Ts'o952fc182012-07-09 16:27:05 -04003298 if (!buf)
3299 return -ENOMEM;
3300
3301 /*
3302 * Compute the overhead (FS structures). This is constant
3303 * for a given filesystem unless the number of block groups
3304 * changes so we cache the previous value until it does.
3305 */
3306
3307 /*
3308 * All of the blocks before first_data_block are overhead
3309 */
3310 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3311
3312 /*
3313 * Add the overhead found in each block group
3314 */
3315 for (i = 0; i < ngroups; i++) {
3316 int blks;
3317
3318 blks = count_overhead(sb, i, buf);
3319 overhead += blks;
3320 if (blks)
3321 memset(buf, 0, PAGE_SIZE);
3322 cond_resched();
3323 }
Eric Sandeen0875a2b2012-12-25 13:56:01 -05003324 /* Add the journal blocks as well */
3325 if (sbi->s_journal)
Lukas Czerner810da242013-03-02 17:18:58 -05003326 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
Eric Sandeen0875a2b2012-12-25 13:56:01 -05003327
Theodore Ts'o952fc182012-07-09 16:27:05 -04003328 sbi->s_overhead = overhead;
3329 smp_wmb();
3330 free_page((unsigned long) buf);
3331 return 0;
3332}
3333
Lukas Czerner27dd4382013-04-09 22:11:22 -04003334
Jan Kara30fac0f2013-12-08 21:11:59 -05003335static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
Lukas Czerner27dd4382013-04-09 22:11:22 -04003336{
3337 ext4_fsblk_t resv_clusters;
3338
3339 /*
Jan Kara30fac0f2013-12-08 21:11:59 -05003340 * There's no need to reserve anything when we aren't using extents.
3341 * The space estimates are exact, there are no unwritten extents,
3342 * hole punching doesn't need new metadata... This is needed especially
3343 * to keep ext2/3 backward compatibility.
3344 */
3345 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3346 return 0;
3347 /*
Lukas Czerner27dd4382013-04-09 22:11:22 -04003348 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3349 * This should cover the situations where we can not afford to run
3350 * out of space like for example punch hole, or converting
Lukas Czerner556615d2014-04-20 23:45:47 -04003351 * unwritten extents in delalloc path. In most cases such
Lukas Czerner27dd4382013-04-09 22:11:22 -04003352 * allocation would require 1, or 2 blocks, higher numbers are
3353 * very rare.
3354 */
Jan Kara30fac0f2013-12-08 21:11:59 -05003355 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3356 EXT4_SB(sb)->s_cluster_bits;
Lukas Czerner27dd4382013-04-09 22:11:22 -04003357
3358 do_div(resv_clusters, 50);
3359 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3360
3361 return resv_clusters;
3362}
3363
3364
3365static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3366{
3367 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3368 sbi->s_cluster_bits;
3369
3370 if (count >= clusters)
3371 return -EINVAL;
3372
3373 atomic64_set(&sbi->s_resv_clusters, count);
3374 return 0;
3375}
3376
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003377static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003378{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003379 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003380 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003381 struct ext4_super_block *es = NULL;
3382 struct ext4_sb_info *sbi;
3383 ext4_fsblk_t block;
3384 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003385 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003386 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003387 unsigned long journal_devnum = 0;
3388 unsigned long def_mount_opts;
3389 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003390 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05003391 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003392 int ret = -ENOMEM;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003393 int blocksize, clustersize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003394 unsigned int db_count;
3395 unsigned int i;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003396 int needs_recovery, has_huge_files, has_bigalloc;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003397 __u64 blocks_count;
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003398 int err = 0;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003399 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003400 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003401
3402 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3403 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003404 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003405
3406 sbi->s_blockgroup_lock =
3407 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3408 if (!sbi->s_blockgroup_lock) {
3409 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003410 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003411 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003412 sb->s_fs_info = sbi;
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04003413 sbi->s_sb = sb;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003414 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003415 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003416 if (sb->s_bdev->bd_part)
3417 sbi->s_sectors_written_start =
3418 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003419
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003420 /* Cleanup superblock name */
3421 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3422 *cp = '!';
3423
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003424 /* -EINVAL is default */
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003425 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003426 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003427 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003428 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003429 goto out_fail;
3430 }
3431
3432 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003433 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003434 * block sizes. We need to calculate the offset from buffer start.
3435 */
Mingming Cao617ba132006-10-11 01:20:53 -07003436 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003437 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3438 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003439 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003440 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003441 }
3442
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003443 if (!(bh = sb_bread(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003444 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003445 goto out_fail;
3446 }
3447 /*
3448 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003449 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003450 */
Theodore Ts'o2716b802012-05-28 17:47:52 -04003451 es = (struct ext4_super_block *) (bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003452 sbi->s_es = es;
3453 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003454 if (sb->s_magic != EXT4_SUPER_MAGIC)
3455 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003456 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003457
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003458 /* Warn if metadata_csum and gdt_csum are both set. */
3459 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3460 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3461 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3462 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3463 "redundant flags; please run fsck.");
3464
Darrick J. Wongd25425f2012-04-29 18:25:10 -04003465 /* Check for a known checksum algorithm */
3466 if (!ext4_verify_csum_type(sb, es)) {
3467 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3468 "unknown checksum algorithm.");
3469 silent = 1;
3470 goto cantfind_ext4;
3471 }
3472
Darrick J. Wong04419842012-04-29 18:27:10 -04003473 /* Load the checksum driver */
3474 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3475 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3476 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3477 if (IS_ERR(sbi->s_chksum_driver)) {
3478 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3479 ret = PTR_ERR(sbi->s_chksum_driver);
3480 sbi->s_chksum_driver = NULL;
3481 goto failed_mount;
3482 }
3483 }
3484
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003485 /* Check superblock checksum */
3486 if (!ext4_superblock_csum_verify(sb, es)) {
3487 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3488 "invalid superblock checksum. Run e2fsck?");
3489 silent = 1;
3490 goto cantfind_ext4;
3491 }
3492
3493 /* Precompute checksum seed for all metadata */
3494 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3495 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3496 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3497 sizeof(es->s_uuid));
3498
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003499 /* Set defaults before we parse the mount options */
3500 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003501 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003502 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003503 set_opt(sb, DEBUG);
Theodore Ts'o87f26802012-03-02 00:03:21 -05003504 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003505 set_opt(sb, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07003506 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003507 set_opt(sb, NO_UID32);
Eric Sandeenea663332011-02-23 17:51:51 -05003508 /* xattr user namespace & acls are now defaulted on */
Eric Sandeenea663332011-02-23 17:51:51 -05003509 set_opt(sb, XATTR_USER);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003510#ifdef CONFIG_EXT4_FS_POSIX_ACL
Eric Sandeenea663332011-02-23 17:51:51 -05003511 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003512#endif
Mingming Cao617ba132006-10-11 01:20:53 -07003513 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003514 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003515 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003516 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003517 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003518 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003519
Mingming Cao617ba132006-10-11 01:20:53 -07003520 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003521 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003522 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003523 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003524 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003525 set_opt(sb, ERRORS_RO);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003526 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003527 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003528 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003529 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003530
Eric W. Biederman08cefc72012-02-07 15:41:49 -08003531 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3532 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
Theodore Ts'o30773842009-01-03 20:27:38 -05003533 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3534 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3535 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003536
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003537 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003538 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003539
Mingming Cao1e2462f2007-07-18 09:00:55 -04003540 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003541 * enable delayed allocation by default
3542 * Use -o nodelalloc to turn it off
3543 */
Theodore Ts'obc0b75f2012-09-17 22:54:36 -04003544 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003545 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003546 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003547
Lukas Czerner51ce6512011-05-20 13:55:16 -04003548 /*
3549 * set default s_li_wait_mult for lazyinit, for the case there is
3550 * no mount option specified.
3551 */
3552 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3553
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003554 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
Eric Sandeen661aa522012-02-20 17:53:04 -05003555 &journal_devnum, &journal_ioprio, 0)) {
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003556 ext4_msg(sb, KERN_WARNING,
3557 "failed to parse options in superblock: %s",
3558 sbi->s_es->s_mount_opts);
3559 }
Theodore Ts'o5a916be2012-03-04 19:27:31 -05003560 sbi->s_def_mount_opt = sbi->s_mount_opt;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003561 if (!parse_options((char *) data, sb, &journal_devnum,
Eric Sandeen661aa522012-02-20 17:53:04 -05003562 &journal_ioprio, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003563 goto failed_mount;
3564
Theodore Ts'o56889782011-09-03 18:22:38 -04003565 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3566 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3567 "with data=journal disables delayed "
3568 "allocation and O_DIRECT support!\n");
3569 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3570 ext4_msg(sb, KERN_ERR, "can't mount with "
3571 "both data=journal and delalloc");
3572 goto failed_mount;
3573 }
3574 if (test_opt(sb, DIOREAD_NOLOCK)) {
3575 ext4_msg(sb, KERN_ERR, "can't mount with "
Piotr Sarna6ae65142013-08-08 23:02:24 -04003576 "both data=journal and dioread_nolock");
Theodore Ts'o56889782011-09-03 18:22:38 -04003577 goto failed_mount;
3578 }
3579 if (test_opt(sb, DELALLOC))
3580 clear_opt(sb, DELALLOC);
3581 }
3582
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003583 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003584 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003585
Mingming Cao617ba132006-10-11 01:20:53 -07003586 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3587 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3588 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3589 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003590 ext4_msg(sb, KERN_WARNING,
3591 "feature flags set on rev 0 fs, "
3592 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003593
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003594 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3595 set_opt2(sb, HURD_COMPAT);
3596 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3597 EXT4_FEATURE_INCOMPAT_64BIT)) {
3598 ext4_msg(sb, KERN_ERR,
3599 "The Hurd can't support 64-bit file systems");
3600 goto failed_mount;
3601 }
3602 }
3603
Theodore Ts'o2035e772011-04-18 17:29:14 -04003604 if (IS_EXT2_SB(sb)) {
3605 if (ext2_feature_set_ok(sb))
3606 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3607 "using the ext4 subsystem");
3608 else {
3609 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3610 "to feature incompatibilities");
3611 goto failed_mount;
3612 }
3613 }
3614
3615 if (IS_EXT3_SB(sb)) {
3616 if (ext3_feature_set_ok(sb))
3617 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3618 "using the ext4 subsystem");
3619 else {
3620 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3621 "to feature incompatibilities");
3622 goto failed_mount;
3623 }
3624 }
3625
Theodore Tso469108f2008-02-10 01:11:44 -05003626 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003627 * Check feature flags regardless of the revision level, since we
3628 * previously didn't change the revision level when setting the flags,
3629 * so there is a chance incompat flags are set on a rev 0 filesystem.
3630 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003631 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003632 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003633
Jan Kara261cb20c2012-12-20 00:07:18 -05003634 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
Mingming Cao617ba132006-10-11 01:20:53 -07003635 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3636 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003637 ext4_msg(sb, KERN_ERR,
3638 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003639 goto failed_mount;
3640 }
3641
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003642 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003643 /* Validate the filesystem blocksize */
3644 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003645 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003646 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003647 goto failed_mount;
3648 }
3649
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003650 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003651 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3652 offset = do_div(logical_sb_block, blocksize);
3653 bh = sb_bread(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003654 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003655 ext4_msg(sb, KERN_ERR,
3656 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003657 goto failed_mount;
3658 }
Theodore Ts'o2716b802012-05-28 17:47:52 -04003659 es = (struct ext4_super_block *)(bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003660 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003661 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003662 ext4_msg(sb, KERN_ERR,
3663 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003664 goto failed_mount;
3665 }
3666 }
3667
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003668 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3669 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003670 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3671 has_huge_files);
3672 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003673
Mingming Cao617ba132006-10-11 01:20:53 -07003674 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3675 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3676 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003677 } else {
3678 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3679 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003680 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003681 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003682 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003683 ext4_msg(sb, KERN_ERR,
3684 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003685 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003686 goto failed_mount;
3687 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003688 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3689 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003690 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003691
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003692 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3693 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003694 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003695 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003696 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003697 ext4_msg(sb, KERN_ERR,
3698 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003699 sbi->s_desc_size);
3700 goto failed_mount;
3701 }
3702 } else
3703 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003704
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003705 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003706 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003707 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003708 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003709
Mingming Cao617ba132006-10-11 01:20:53 -07003710 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003711 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003712 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003713 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3714 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003715 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003716 sbi->s_sbh = bh;
3717 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003718 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3719 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003720
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003721 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003722 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3723 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'o23301412014-02-12 12:16:04 -05003724 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3725 i = le32_to_cpu(es->s_flags);
3726 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3727 sbi->s_hash_unsigned = 3;
3728 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
Theodore Ts'of99b2582008-10-28 13:21:44 -04003729#ifdef __CHAR_UNSIGNED__
Theodore Ts'o23301412014-02-12 12:16:04 -05003730 if (!(sb->s_flags & MS_RDONLY))
3731 es->s_flags |=
3732 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3733 sbi->s_hash_unsigned = 3;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003734#else
Theodore Ts'o23301412014-02-12 12:16:04 -05003735 if (!(sb->s_flags & MS_RDONLY))
3736 es->s_flags |=
3737 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
Theodore Ts'of99b2582008-10-28 13:21:44 -04003738#endif
Theodore Ts'o23301412014-02-12 12:16:04 -05003739 }
Theodore Ts'of99b2582008-10-28 13:21:44 -04003740 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003741
Theodore Ts'o281b5992011-09-09 18:34:51 -04003742 /* Handle clustersize */
3743 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3744 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3745 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3746 if (has_bigalloc) {
3747 if (clustersize < blocksize) {
3748 ext4_msg(sb, KERN_ERR,
3749 "cluster size (%d) smaller than "
3750 "block size (%d)", clustersize, blocksize);
3751 goto failed_mount;
3752 }
3753 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3754 le32_to_cpu(es->s_log_block_size);
3755 sbi->s_clusters_per_group =
3756 le32_to_cpu(es->s_clusters_per_group);
3757 if (sbi->s_clusters_per_group > blocksize * 8) {
3758 ext4_msg(sb, KERN_ERR,
3759 "#clusters per group too big: %lu",
3760 sbi->s_clusters_per_group);
3761 goto failed_mount;
3762 }
3763 if (sbi->s_blocks_per_group !=
3764 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3765 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3766 "clusters per group (%lu) inconsistent",
3767 sbi->s_blocks_per_group,
3768 sbi->s_clusters_per_group);
3769 goto failed_mount;
3770 }
3771 } else {
3772 if (clustersize != blocksize) {
3773 ext4_warning(sb, "fragment/cluster size (%d) != "
3774 "block size (%d)", clustersize,
3775 blocksize);
3776 clustersize = blocksize;
3777 }
3778 if (sbi->s_blocks_per_group > blocksize * 8) {
3779 ext4_msg(sb, KERN_ERR,
3780 "#blocks per group too big: %lu",
3781 sbi->s_blocks_per_group);
3782 goto failed_mount;
3783 }
3784 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3785 sbi->s_cluster_bits = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003786 }
Theodore Ts'o281b5992011-09-09 18:34:51 -04003787 sbi->s_cluster_ratio = clustersize / blocksize;
3788
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003789 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003790 ext4_msg(sb, KERN_ERR,
3791 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003792 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003793 goto failed_mount;
3794 }
3795
Theodore Ts'o960fd8562013-07-05 23:11:16 -04003796 /* Do we have standard group size of clustersize * 8 blocks ? */
3797 if (sbi->s_blocks_per_group == clustersize << 3)
3798 set_opt2(sb, STD_GROUP_SIZE);
3799
Eric Sandeenbf43d842009-08-17 23:48:51 -04003800 /*
3801 * Test whether we have more sectors than will fit in sector_t,
3802 * and whether the max offset is addressable by the page cache.
3803 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003804 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003805 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003806 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003807 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003808 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003809 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003810 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003811 goto failed_mount;
3812 }
3813
Mingming Cao617ba132006-10-11 01:20:53 -07003814 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3815 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003816
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003817 /* check blocks count against device size */
3818 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3819 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003820 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3821 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003822 ext4_blocks_count(es), blocks_count);
3823 goto failed_mount;
3824 }
3825
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003826 /*
3827 * It makes no sense for the first data block to be beyond the end
3828 * of the filesystem.
3829 */
3830 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Zheng Liu5635a622011-12-18 16:13:58 -05003831 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003832 "block %u is beyond end of filesystem (%llu)",
3833 le32_to_cpu(es->s_first_data_block),
3834 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003835 goto failed_mount;
3836 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003837 blocks_count = (ext4_blocks_count(es) -
3838 le32_to_cpu(es->s_first_data_block) +
3839 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3840 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003841 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003842 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003843 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003844 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003845 ext4_blocks_count(es),
3846 le32_to_cpu(es->s_first_data_block),
3847 EXT4_BLOCKS_PER_GROUP(sb));
3848 goto failed_mount;
3849 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003850 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003851 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3852 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003853 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3854 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'of18a5f22011-08-01 08:45:38 -04003855 sbi->s_group_desc = ext4_kvmalloc(db_count *
3856 sizeof(struct buffer_head *),
3857 GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003858 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003859 ext4_msg(sb, KERN_ERR, "not enough memory");
Theodore Ts'o2cde4172012-05-28 17:49:54 -04003860 ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003861 goto failed_mount;
3862 }
3863
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003864 if (ext4_proc_root)
3865 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003866
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05003867 if (sbi->s_proc)
3868 proc_create_data("options", S_IRUGO, sbi->s_proc,
3869 &ext4_seq_options_fops, sb);
3870
Pekka Enberg705895b2009-02-15 18:07:52 -05003871 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003872
3873 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003874 block = descriptor_loc(sb, logical_sb_block, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003875 sbi->s_group_desc[i] = sb_bread(sb, block);
3876 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003877 ext4_msg(sb, KERN_ERR,
3878 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003879 db_count = i;
3880 goto failed_mount2;
3881 }
3882 }
Azat Khuzhin00764932014-04-07 10:54:20 -04003883
3884 /*
3885 * set up enough so that it can read an inode,
3886 * and create new inode for buddy allocator
3887 */
3888 sbi->s_gdb_count = db_count;
3889 if (!test_opt(sb, NOLOAD) &&
3890 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3891 sb->s_op = &ext4_sops;
3892 else
3893 sb->s_op = &ext4_nojournal_sops;
3894
3895 ext4_ext_init(sb);
3896 err = ext4_mb_init(sb);
3897 if (err) {
3898 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3899 err);
3900 goto failed_mount2;
3901 }
3902
Lukas Czernerbfff6872010-10-27 21:30:05 -04003903 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003904 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Azat Khuzhin00764932014-04-07 10:54:20 -04003905 goto failed_mount2a;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003906 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003907 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3908 if (!ext4_fill_flex_info(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003909 ext4_msg(sb, KERN_ERR,
3910 "unable to initialize "
3911 "flex_bg meta info!");
Azat Khuzhin00764932014-04-07 10:54:20 -04003912 goto failed_mount2a;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003913 }
3914
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003915 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3916 spin_lock_init(&sbi->s_next_gen_lock);
3917
Tao Ma04496412011-04-05 19:55:28 -04003918 init_timer(&sbi->s_err_report);
3919 sbi->s_err_report.function = print_daily_error_info;
3920 sbi->s_err_report.data = (unsigned long) sb;
3921
Dmitry Monakhova75ae78f2013-04-03 22:10:52 -04003922 /* Register extent status tree shrinker */
Zheng Liud3922a72013-07-01 08:12:37 -04003923 ext4_es_register_shrinker(sbi);
Dmitry Monakhova75ae78f2013-04-03 22:10:52 -04003924
Theodore Ts'o57042652011-09-09 18:56:51 -04003925 err = percpu_counter_init(&sbi->s_freeclusters_counter,
Theodore Ts'o5dee5432011-09-09 19:10:51 -04003926 ext4_count_free_clusters(sb));
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003927 if (!err) {
3928 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3929 ext4_count_free_inodes(sb));
3930 }
3931 if (!err) {
3932 err = percpu_counter_init(&sbi->s_dirs_counter,
3933 ext4_count_dirs(sb));
3934 }
3935 if (!err) {
Theodore Ts'o57042652011-09-09 18:56:51 -04003936 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003937 }
Theodore Ts'o1ac64662013-03-02 10:27:46 -05003938 if (!err) {
3939 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3940 }
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003941 if (err) {
3942 ext4_msg(sb, KERN_ERR, "insufficient memory");
3943 goto failed_mount3;
3944 }
3945
Alex Tomasc9de5602008-01-29 00:19:52 -05003946 sbi->s_stripe = ext4_get_stripe_size(sbi);
Zheng Liu67a5da52012-08-17 09:54:17 -04003947 sbi->s_extent_max_zeroout_kb = 32;
Alex Tomasc9de5602008-01-29 00:19:52 -05003948
Mingming Cao617ba132006-10-11 01:20:53 -07003949 sb->s_export_op = &ext4_export_ops;
3950 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003951#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003952 sb->dq_op = &ext4_quota_operations;
Jan Kara262b4662013-03-02 17:57:08 -05003953 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
Aditya Kali7c319d32012-07-22 20:21:31 -04003954 sb->s_qcop = &ext4_qctl_sysfile_operations;
Jan Kara262b4662013-03-02 17:57:08 -05003955 else
3956 sb->s_qcop = &ext4_qctl_operations;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003957#endif
Aneesh Kumar K.Vf2fa2ff2011-01-29 18:43:40 +05303958 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3959
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003960 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003961 mutex_init(&sbi->s_orphan_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003962
3963 sb->s_root = NULL;
3964
3965 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07003966 EXT4_HAS_INCOMPAT_FEATURE(sb,
3967 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003968
Johann Lombardic5e06d12011-05-24 18:31:25 -04003969 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3970 !(sb->s_flags & MS_RDONLY))
3971 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3972 goto failed_mount3;
3973
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003974 /*
3975 * The first inode we look at is the journal inode. Don't try
3976 * root first: it may be modified in the journal!
3977 */
3978 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07003979 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3980 if (ext4_load_journal(sb, es, journal_devnum))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003981 goto failed_mount3;
Frank Mayhar03901312009-01-07 00:06:22 -05003982 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3983 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003984 ext4_msg(sb, KERN_ERR, "required journal recovery "
3985 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003986 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003987 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003988 clear_opt(sb, DATA_FLAGS);
Frank Mayhar03901312009-01-07 00:06:22 -05003989 sbi->s_journal = NULL;
3990 needs_recovery = 0;
3991 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003992 }
3993
Theodore Ts'of32aaf22012-05-21 11:42:02 -04003994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
Jose R. Santoseb40a092007-07-18 08:37:25 -04003995 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3996 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003997 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003998 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04003999 }
4000
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04004001 if (!set_journal_csum_feature_set(sb)) {
4002 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4003 "feature set");
4004 goto failed_mount_wq;
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08004005 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05004006
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004007 /* We have now updated the journal if required, so we can
4008 * validate the data journaling mode. */
4009 switch (test_opt(sb, DATA_FLAGS)) {
4010 case 0:
4011 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07004012 * capabilities: ORDERED_DATA if the journal can
4013 * cope, else JOURNAL_DATA
4014 */
Mingming Caodab291a2006-10-11 01:21:01 -07004015 if (jbd2_journal_check_available_features
4016 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004017 set_opt(sb, ORDERED_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004018 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004019 set_opt(sb, JOURNAL_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004020 break;
4021
Mingming Cao617ba132006-10-11 01:20:53 -07004022 case EXT4_MOUNT_ORDERED_DATA:
4023 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07004024 if (!jbd2_journal_check_available_features
4025 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004026 ext4_msg(sb, KERN_ERR, "Journal does not support "
4027 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004028 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004029 }
4030 default:
4031 break;
4032 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004033 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004034
Bobi Jam18aadd42012-02-20 17:53:02 -05004035 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4036
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004037 /*
4038 * The journal may have updated the bg summary counts, so we
4039 * need to update the global counters.
4040 */
Theodore Ts'o57042652011-09-09 18:56:51 -04004041 percpu_counter_set(&sbi->s_freeclusters_counter,
Theodore Ts'o5dee5432011-09-09 19:10:51 -04004042 ext4_count_free_clusters(sb));
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004043 percpu_counter_set(&sbi->s_freeinodes_counter,
4044 ext4_count_free_inodes(sb));
4045 percpu_counter_set(&sbi->s_dirs_counter,
4046 ext4_count_dirs(sb));
Theodore Ts'o57042652011-09-09 18:56:51 -04004047 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
Christoph Hellwig206f7ab2010-06-14 14:42:49 -04004048
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004049no_journal:
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04004050 if (ext4_mballoc_ready) {
4051 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4052 if (!sbi->s_mb_cache) {
4053 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4054 goto failed_mount_wq;
4055 }
4056 }
4057
Tejun Heofd89d5f2011-02-01 11:42:42 +01004058 /*
Theodore Ts'o952fc182012-07-09 16:27:05 -04004059 * Get the # of file system overhead blocks from the
4060 * superblock if present.
4061 */
4062 if (es->s_overhead_clusters)
4063 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4064 else {
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004065 err = ext4_calculate_overhead(sb);
4066 if (err)
Theodore Ts'o952fc182012-07-09 16:27:05 -04004067 goto failed_mount_wq;
4068 }
4069
4070 /*
Tejun Heofd89d5f2011-02-01 11:42:42 +01004071 * The maximum number of concurrent works can be high and
4072 * concurrency isn't really necessary. Limit it to 1.
4073 */
Jan Kara2e8fa542013-06-04 14:21:02 -04004074 EXT4_SB(sb)->rsv_conversion_wq =
4075 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4076 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4077 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004078 ret = -ENOMEM;
Jan Kara2e8fa542013-06-04 14:21:02 -04004079 goto failed_mount4;
4080 }
4081
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004082 /*
Mingming Caodab291a2006-10-11 01:21:01 -07004083 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004084 * so we can safely mount the rest of the filesystem now.
4085 */
4086
David Howells1d1fe1e2008-02-07 00:15:37 -08004087 root = ext4_iget(sb, EXT4_ROOT_INO);
4088 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004089 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004090 ret = PTR_ERR(root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004091 root = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004092 goto failed_mount4;
4093 }
4094 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004095 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Al Viro94bf6082012-01-09 15:53:24 -05004096 iput(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004097 goto failed_mount4;
4098 }
Al Viro48fde702012-01-08 22:15:13 -05004099 sb->s_root = d_make_root(root);
David Howells1d1fe1e2008-02-07 00:15:37 -08004100 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004101 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004102 ret = -ENOMEM;
4103 goto failed_mount4;
4104 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004105
Eric Sandeen7e84b622012-05-28 14:17:25 -04004106 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4107 sb->s_flags |= MS_RDONLY;
Kalpak Shahef7f3832007-07-18 09:15:20 -04004108
4109 /* determine the minimum size of new large inodes, if present */
4110 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4111 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4112 EXT4_GOOD_OLD_INODE_SIZE;
4113 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4114 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4115 if (sbi->s_want_extra_isize <
4116 le16_to_cpu(es->s_want_extra_isize))
4117 sbi->s_want_extra_isize =
4118 le16_to_cpu(es->s_want_extra_isize);
4119 if (sbi->s_want_extra_isize <
4120 le16_to_cpu(es->s_min_extra_isize))
4121 sbi->s_want_extra_isize =
4122 le16_to_cpu(es->s_min_extra_isize);
4123 }
4124 }
4125 /* Check if enough inode space is available */
4126 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4127 sbi->s_inode_size) {
4128 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4129 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04004130 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4131 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04004132 }
4133
Jan Kara30fac0f2013-12-08 21:11:59 -05004134 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
Lukas Czerner27dd4382013-04-09 22:11:22 -04004135 if (err) {
4136 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
Jan Kara30fac0f2013-12-08 21:11:59 -05004137 "reserved pool", ext4_calculate_resv_clusters(sb));
Azat Khuzhin00764932014-04-07 10:54:20 -04004138 goto failed_mount5;
Lukas Czerner27dd4382013-04-09 22:11:22 -04004139 }
4140
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004141 err = ext4_setup_system_zone(sb);
4142 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004143 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04004144 "zone (%d)", err);
Tao Madcf2d802011-10-06 12:10:11 -04004145 goto failed_mount5;
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04004146 }
4147
Lukas Czernerbfff6872010-10-27 21:30:05 -04004148 err = ext4_register_li_request(sb, first_not_zeroed);
4149 if (err)
Tao Madcf2d802011-10-06 12:10:11 -04004150 goto failed_mount6;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004151
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004152 sbi->s_kobj.kset = ext4_kset;
4153 init_completion(&sbi->s_kobj_unregister);
4154 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4155 "%s", sb->s_id);
Tao Madcf2d802011-10-06 12:10:11 -04004156 if (err)
4157 goto failed_mount7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004158
Jan Kara9b2ff3572013-03-02 18:22:38 -05004159#ifdef CONFIG_QUOTA
4160 /* Enable quota usage during mount. */
4161 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4162 !(sb->s_flags & MS_RDONLY)) {
4163 err = ext4_enable_quotas(sb);
4164 if (err)
4165 goto failed_mount8;
4166 }
4167#endif /* CONFIG_QUOTA */
4168
Mingming Cao617ba132006-10-11 01:20:53 -07004169 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4170 ext4_orphan_cleanup(sb, es);
4171 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05004172 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004173 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05004174 ext4_mark_recovery_complete(sb, es);
4175 }
4176 if (EXT4_SB(sb)->s_journal) {
4177 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4178 descr = " journalled data mode";
4179 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4180 descr = " ordered data mode";
4181 else
4182 descr = " writeback data mode";
4183 } else
4184 descr = "out journal";
4185
Lukas Czerner79add3a2012-11-08 13:28:29 -05004186 if (test_opt(sb, DISCARD)) {
4187 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4188 if (!blk_queue_discard(q))
4189 ext4_msg(sb, KERN_WARNING,
4190 "mounting with \"discard\" option, but "
4191 "the device does not support discard");
4192 }
4193
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004194 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o8b67f042010-08-01 23:14:20 -04004195 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4196 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004197
Theodore Ts'o66e61a92010-07-27 11:56:04 -04004198 if (es->s_error_count)
4199 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004200
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04004201 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4202 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4203 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4204 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4205
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004206 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004207 return 0;
4208
Mingming Cao617ba132006-10-11 01:20:53 -07004209cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004210 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004211 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004212 goto failed_mount;
4213
Theodore Ts'o72ba7452013-01-24 23:24:54 -05004214#ifdef CONFIG_QUOTA
4215failed_mount8:
4216 kobject_del(&sbi->s_kobj);
4217#endif
Tao Madcf2d802011-10-06 12:10:11 -04004218failed_mount7:
4219 ext4_unregister_li_request(sb);
4220failed_mount6:
Tao Madcf2d802011-10-06 12:10:11 -04004221 ext4_release_system_zone(sb);
Azat Khuzhin00764932014-04-07 10:54:20 -04004222failed_mount5:
Al Viro94bf6082012-01-09 15:53:24 -05004223 dput(sb->s_root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004224 sb->s_root = NULL;
Al Viro94bf6082012-01-09 15:53:24 -05004225failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04004226 ext4_msg(sb, KERN_ERR, "mount failed");
Jan Kara2e8fa542013-06-04 14:21:02 -04004227 if (EXT4_SB(sb)->rsv_conversion_wq)
4228 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -04004229failed_mount_wq:
Frank Mayhar03901312009-01-07 00:06:22 -05004230 if (sbi->s_journal) {
4231 jbd2_journal_destroy(sbi->s_journal);
4232 sbi->s_journal = NULL;
4233 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004234failed_mount3:
Zheng Liud3922a72013-07-01 08:12:37 -04004235 ext4_es_unregister_shrinker(sbi);
Al Viro9105bb12013-12-08 20:52:31 -05004236 del_timer_sync(&sbi->s_err_report);
Theodore Ts'o9933fc02011-08-01 08:45:02 -04004237 if (sbi->s_flex_groups)
4238 ext4_kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -04004239 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004240 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4241 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -04004242 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Theodore Ts'o1ac64662013-03-02 10:27:46 -05004243 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
Johann Lombardic5e06d12011-05-24 18:31:25 -04004244 if (sbi->s_mmp_tsk)
4245 kthread_stop(sbi->s_mmp_tsk);
Azat Khuzhin00764932014-04-07 10:54:20 -04004246failed_mount2a:
4247 ext4_mb_release(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004248failed_mount2:
4249 for (i = 0; i < db_count; i++)
4250 brelse(sbi->s_group_desc[i]);
Theodore Ts'of18a5f22011-08-01 08:45:38 -04004251 ext4_kvfree(sbi->s_group_desc);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004252failed_mount:
Azat Khuzhin00764932014-04-07 10:54:20 -04004253 ext4_ext_release(sb);
Darrick J. Wong04419842012-04-29 18:27:10 -04004254 if (sbi->s_chksum_driver)
4255 crypto_free_shash(sbi->s_chksum_driver);
Theodore Ts'o240799c2008-10-09 23:53:47 -04004256 if (sbi->s_proc) {
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05004257 remove_proc_entry("options", sbi->s_proc);
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004258 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04004259 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004260#ifdef CONFIG_QUOTA
4261 for (i = 0; i < MAXQUOTAS; i++)
4262 kfree(sbi->s_qf_names[i]);
4263#endif
Mingming Cao617ba132006-10-11 01:20:53 -07004264 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004265 brelse(bh);
4266out_fail:
4267 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04004268 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004269 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04004270out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004271 kfree(orig_data);
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004272 return err ? err : ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004273}
4274
4275/*
4276 * Setup any per-fs journal parameters now. We'll do this both on
4277 * initial mount, once the journal has been initialised but before we've
4278 * done any recovery; and again on any subsequent remount.
4279 */
Mingming Cao617ba132006-10-11 01:20:53 -07004280static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004281{
Mingming Cao617ba132006-10-11 01:20:53 -07004282 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004283
Theodore Ts'o30773842009-01-03 20:27:38 -05004284 journal->j_commit_interval = sbi->s_commit_interval;
4285 journal->j_min_batch_time = sbi->s_min_batch_time;
4286 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004287
Theodore Ts'oa931da62010-08-03 21:35:12 -04004288 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004289 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07004290 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004291 else
Mingming Caodab291a2006-10-11 01:21:01 -07004292 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04004293 if (test_opt(sb, DATA_ERR_ABORT))
4294 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4295 else
4296 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04004297 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004298}
4299
Mingming Cao617ba132006-10-11 01:20:53 -07004300static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004301 unsigned int journal_inum)
4302{
4303 struct inode *journal_inode;
4304 journal_t *journal;
4305
Frank Mayhar03901312009-01-07 00:06:22 -05004306 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4307
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004308 /* First, test for the existence of a valid inode on disk. Bad
4309 * things happen if we iget() an unused inode, as the subsequent
4310 * iput() will try to delete it. */
4311
David Howells1d1fe1e2008-02-07 00:15:37 -08004312 journal_inode = ext4_iget(sb, journal_inum);
4313 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004314 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004315 return NULL;
4316 }
4317 if (!journal_inode->i_nlink) {
4318 make_bad_inode(journal_inode);
4319 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04004320 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004321 return NULL;
4322 }
4323
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04004324 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004325 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08004326 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004327 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004328 iput(journal_inode);
4329 return NULL;
4330 }
4331
Mingming Caodab291a2006-10-11 01:21:01 -07004332 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004333 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004334 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004335 iput(journal_inode);
4336 return NULL;
4337 }
4338 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004339 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004340 return journal;
4341}
4342
Mingming Cao617ba132006-10-11 01:20:53 -07004343static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004344 dev_t j_dev)
4345{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004346 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004347 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07004348 ext4_fsblk_t start;
4349 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004350 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07004351 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004352 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004353 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004354 struct block_device *bdev;
4355
Frank Mayhar03901312009-01-07 00:06:22 -05004356 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4357
Eric Sandeenb31e1552009-06-04 17:36:36 -04004358 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004359 if (bdev == NULL)
4360 return NULL;
4361
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004362 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04004363 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004364 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004365 ext4_msg(sb, KERN_ERR,
4366 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004367 goto out_bdev;
4368 }
4369
Mingming Cao617ba132006-10-11 01:20:53 -07004370 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4371 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004372 set_blocksize(bdev, blocksize);
4373 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004374 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4375 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004376 goto out_bdev;
4377 }
4378
Theodore Ts'o2716b802012-05-28 17:47:52 -04004379 es = (struct ext4_super_block *) (bh->b_data + offset);
Mingming Cao617ba132006-10-11 01:20:53 -07004380 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004381 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07004382 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004383 ext4_msg(sb, KERN_ERR, "external journal has "
4384 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004385 brelse(bh);
4386 goto out_bdev;
4387 }
4388
Mingming Cao617ba132006-10-11 01:20:53 -07004389 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004390 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004391 brelse(bh);
4392 goto out_bdev;
4393 }
4394
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004395 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004396 start = sb_block + 1;
4397 brelse(bh); /* we're done with the superblock */
4398
Mingming Caodab291a2006-10-11 01:21:01 -07004399 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004400 start, len, blocksize);
4401 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004402 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004403 goto out_bdev;
4404 }
4405 journal->j_private = sb;
Theodore Ts'o9f203502013-04-20 15:46:17 -04004406 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004407 wait_on_buffer(journal->j_sb_buffer);
4408 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004409 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004410 goto out_journal;
4411 }
4412 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004413 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4414 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004415 be32_to_cpu(journal->j_superblock->s_nr_users));
4416 goto out_journal;
4417 }
Mingming Cao617ba132006-10-11 01:20:53 -07004418 EXT4_SB(sb)->journal_bdev = bdev;
4419 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004420 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004421
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004422out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07004423 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004424out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07004425 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004426 return NULL;
4427}
4428
Mingming Cao617ba132006-10-11 01:20:53 -07004429static int ext4_load_journal(struct super_block *sb,
4430 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004431 unsigned long journal_devnum)
4432{
4433 journal_t *journal;
4434 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4435 dev_t journal_dev;
4436 int err = 0;
4437 int really_read_only;
4438
Frank Mayhar03901312009-01-07 00:06:22 -05004439 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4440
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004441 if (journal_devnum &&
4442 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004443 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4444 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004445 journal_dev = new_decode_dev(journal_devnum);
4446 } else
4447 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4448
4449 really_read_only = bdev_read_only(sb->s_bdev);
4450
4451 /*
4452 * Are we loading a blank journal or performing recovery after a
4453 * crash? For recovery, we need to check in advance whether we
4454 * can get read-write access to the device.
4455 */
Mingming Cao617ba132006-10-11 01:20:53 -07004456 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004457 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004458 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4459 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004460 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004461 ext4_msg(sb, KERN_ERR, "write access "
4462 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004463 return -EROFS;
4464 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04004465 ext4_msg(sb, KERN_INFO, "write access will "
4466 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004467 }
4468 }
4469
4470 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004471 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4472 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004473 return -EINVAL;
4474 }
4475
4476 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07004477 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004478 return -EINVAL;
4479 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07004480 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004481 return -EINVAL;
4482 }
4483
Theodore Ts'o90576c02009-09-29 15:51:30 -04004484 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04004485 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04004486
Mingming Cao617ba132006-10-11 01:20:53 -07004487 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07004488 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004489 if (!err) {
4490 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4491 if (save)
4492 memcpy(save, ((char *) es) +
4493 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07004494 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004495 if (save)
4496 memcpy(((char *) es) + EXT4_S_ERR_START,
4497 save, EXT4_S_ERR_LEN);
4498 kfree(save);
4499 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004500
4501 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004502 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07004503 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004504 return err;
4505 }
4506
Mingming Cao617ba132006-10-11 01:20:53 -07004507 EXT4_SB(sb)->s_journal = journal;
4508 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004509
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04004510 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004511 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4512 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004513
4514 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004515 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004516 }
4517
4518 return 0;
4519}
4520
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004521static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004522{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004523 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07004524 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08004525 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004526
Theodore Ts'o7c2e7082011-09-09 18:28:51 -04004527 if (!sbh || block_device_ejected(sb))
Takashi Satoc4be0c12009-01-09 16:40:58 -08004528 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04004529 if (buffer_write_io_error(sbh)) {
4530 /*
4531 * Oh, dear. A previous attempt to write the
4532 * superblock failed. This could happen because the
4533 * USB device was yanked out. Or it could happen to
4534 * be a transient write error and maybe the block will
4535 * be remapped. Nothing we can do but to retry the
4536 * write and hope for the best.
4537 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04004538 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4539 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004540 clear_buffer_write_io_error(sbh);
4541 set_buffer_uptodate(sbh);
4542 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04004543 /*
4544 * If the file system is mounted read-only, don't update the
4545 * superblock write time. This avoids updating the superblock
4546 * write time when we are mounting the root file system
4547 * read/only but we need to replay the journal; at that point,
4548 * for people who are east of GMT and who make their clock
4549 * tick in localtime for Windows bug-for-bug compatibility,
4550 * the clock is set in the future, and this will cause e2fsck
4551 * to complain and force a full file system check.
4552 */
4553 if (!(sb->s_flags & MS_RDONLY))
4554 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004555 if (sb->s_bdev->bd_part)
4556 es->s_kbytes_written =
4557 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05004558 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4559 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004560 else
4561 es->s_kbytes_written =
4562 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'o57042652011-09-09 18:56:51 -04004563 ext4_free_blocks_count_set(es,
4564 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4565 &EXT4_SB(sb)->s_freeclusters_counter)));
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004566 es->s_free_inodes_count =
4567 cpu_to_le32(percpu_counter_sum_positive(
4568 &EXT4_SB(sb)->s_freeinodes_counter));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004569 BUFFER_TRACE(sbh, "marking dirty");
Theodore Ts'o06db49e2012-10-10 01:06:58 -04004570 ext4_superblock_csum_set(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004571 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04004572 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08004573 error = sync_dirty_buffer(sbh);
4574 if (error)
4575 return error;
4576
4577 error = buffer_write_io_error(sbh);
4578 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004579 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4580 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004581 clear_buffer_write_io_error(sbh);
4582 set_buffer_uptodate(sbh);
4583 }
4584 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08004585 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004586}
4587
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004588/*
4589 * Have we just finished recovery? If so, and if we are mounting (or
4590 * remounting) the filesystem readonly, then we will end up with a
4591 * consistent fs on disk. Record that fact.
4592 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004593static void ext4_mark_recovery_complete(struct super_block *sb,
4594 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004595{
Mingming Cao617ba132006-10-11 01:20:53 -07004596 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004597
Frank Mayhar03901312009-01-07 00:06:22 -05004598 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4599 BUG_ON(journal != NULL);
4600 return;
4601 }
Mingming Caodab291a2006-10-11 01:21:01 -07004602 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004603 if (jbd2_journal_flush(journal) < 0)
4604 goto out;
4605
Mingming Cao617ba132006-10-11 01:20:53 -07004606 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004607 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07004608 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004609 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004610 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004611
4612out:
Mingming Caodab291a2006-10-11 01:21:01 -07004613 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004614}
4615
4616/*
4617 * If we are mounting (or read-write remounting) a filesystem whose journal
4618 * has recorded an error from a previous lifetime, move that error to the
4619 * main filesystem now.
4620 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004621static void ext4_clear_journal_err(struct super_block *sb,
4622 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004623{
4624 journal_t *journal;
4625 int j_errno;
4626 const char *errstr;
4627
Frank Mayhar03901312009-01-07 00:06:22 -05004628 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4629
Mingming Cao617ba132006-10-11 01:20:53 -07004630 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004631
4632 /*
4633 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004634 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004635 */
4636
Mingming Caodab291a2006-10-11 01:21:01 -07004637 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004638 if (j_errno) {
4639 char nbuf[16];
4640
Mingming Cao617ba132006-10-11 01:20:53 -07004641 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004642 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004643 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004644 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004645
Mingming Cao617ba132006-10-11 01:20:53 -07004646 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4647 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004648 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004649
Mingming Caodab291a2006-10-11 01:21:01 -07004650 jbd2_journal_clear_err(journal);
Theodore Ts'od796c522012-08-05 19:04:57 -04004651 jbd2_journal_update_sb_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004652 }
4653}
4654
4655/*
4656 * Force the running and committing transactions to commit,
4657 * and wait on the commit.
4658 */
Mingming Cao617ba132006-10-11 01:20:53 -07004659int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004660{
4661 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004662
4663 if (sb->s_flags & MS_RDONLY)
4664 return 0;
4665
Mingming Cao617ba132006-10-11 01:20:53 -07004666 journal = EXT4_SB(sb)->s_journal;
Guo Chaob1deefc2013-01-28 21:41:02 -05004667 return ext4_journal_force_commit(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004668}
4669
Mingming Cao617ba132006-10-11 01:20:53 -07004670static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004671{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004672 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004673 tid_t target;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004674 bool needs_barrier = false;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004675 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004676
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004677 trace_ext4_sync_fs(sb, wait);
Jan Kara2e8fa542013-06-04 14:21:02 -04004678 flush_workqueue(sbi->rsv_conversion_wq);
Jan Karaa1177822012-07-03 16:45:29 +02004679 /*
4680 * Writeback quota in non-journalled quota case - journalled quota has
4681 * no dirty dquots
4682 */
4683 dquot_writeback_dquots(sb, -1);
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004684 /*
4685 * Data writeback is possible w/o journal transaction, so barrier must
4686 * being sent at the end of the function. But we can skip it if
4687 * transaction_commit will do it for us.
4688 */
4689 target = jbd2_get_latest_transaction(sbi->s_journal);
4690 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4691 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4692 needs_barrier = true;
4693
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004694 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004695 if (wait)
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004696 ret = jbd2_log_wait_commit(sbi->s_journal, target);
Frank Mayhar03901312009-01-07 00:06:22 -05004697 }
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004698 if (needs_barrier) {
4699 int err;
4700 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4701 if (!ret)
4702 ret = err;
4703 }
4704
4705 return ret;
4706}
4707
4708static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4709{
4710 int ret = 0;
4711
4712 trace_ext4_sync_fs(sb, wait);
4713 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004714 dquot_writeback_dquots(sb, -1);
4715 if (wait && test_opt(sb, BARRIER))
4716 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4717
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004718 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004719}
4720
4721/*
4722 * LVM calls this function before a (read-only) snapshot is created. This
4723 * gives us a chance to flush the journal completely and mark the fs clean.
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -04004724 *
4725 * Note that only this function cannot bring a filesystem to be in a clean
Jan Kara8e8ad8a2012-06-12 16:20:38 +02004726 * state independently. It relies on upper layer to stop all data & metadata
4727 * modifications.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004728 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004729static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004730{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004731 int error = 0;
4732 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004733
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004734 if (sb->s_flags & MS_RDONLY)
4735 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004736
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004737 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004738
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004739 /* Now we set up the journal barrier. */
4740 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004741
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004742 /*
4743 * Don't clear the needs_recovery flag if we failed to flush
4744 * the journal.
4745 */
4746 error = jbd2_journal_flush(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004747 if (error < 0)
4748 goto out;
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004749
4750 /* Journal blocked and flushed, clear needs_recovery flag. */
4751 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4752 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004753out:
Jan Kara8e8ad8a2012-06-12 16:20:38 +02004754 /* we rely on upper layer to stop further updates */
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004755 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4756 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004757}
4758
4759/*
4760 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4761 * flag here, even though the filesystem is not technically dirty yet.
4762 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004763static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004764{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004765 if (sb->s_flags & MS_RDONLY)
4766 return 0;
4767
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004768 /* Reset the needs_recovery flag before the fs is unlocked. */
4769 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4770 ext4_commit_super(sb, 1);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004771 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004772}
4773
Theodore Ts'o673c6102010-12-15 20:28:48 -05004774/*
4775 * Structure to save mount options for ext4_remount's benefit
4776 */
4777struct ext4_mount_options {
4778 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004779 unsigned long s_mount_opt2;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08004780 kuid_t s_resuid;
4781 kgid_t s_resgid;
Theodore Ts'o673c6102010-12-15 20:28:48 -05004782 unsigned long s_commit_interval;
4783 u32 s_min_batch_time, s_max_batch_time;
4784#ifdef CONFIG_QUOTA
4785 int s_jquota_fmt;
4786 char *s_qf_names[MAXQUOTAS];
4787#endif
4788};
4789
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004790static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004791{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004792 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004793 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004794 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004795 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004796 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004797 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004798 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004799 int err = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004800#ifdef CONFIG_QUOTA
Chen Gang03dafb52013-01-24 23:24:58 -05004801 int i, j;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004802#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004803 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004804
4805 /* Store the original options */
4806 old_sb_flags = sb->s_flags;
4807 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004808 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004809 old_opts.s_resuid = sbi->s_resuid;
4810 old_opts.s_resgid = sbi->s_resgid;
4811 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004812 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4813 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004814#ifdef CONFIG_QUOTA
4815 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4816 for (i = 0; i < MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05004817 if (sbi->s_qf_names[i]) {
4818 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4819 GFP_KERNEL);
4820 if (!old_opts.s_qf_names[i]) {
4821 for (j = 0; j < i; j++)
4822 kfree(old_opts.s_qf_names[j]);
Wei Yongjun3e36a162013-03-02 17:13:55 -05004823 kfree(orig_data);
Chen Gang03dafb52013-01-24 23:24:58 -05004824 return -ENOMEM;
4825 }
4826 } else
4827 old_opts.s_qf_names[i] = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004828#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004829 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4830 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004831
4832 /*
4833 * Allow the "check" option to be passed as a remount option.
4834 */
Eric Sandeen661aa522012-02-20 17:53:04 -05004835 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004836 err = -EINVAL;
4837 goto restore_opts;
4838 }
4839
Piotr Sarna6ae65142013-08-08 23:02:24 -04004840 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4841 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4842 ext4_msg(sb, KERN_ERR, "can't mount with "
4843 "both data=journal and delalloc");
4844 err = -EINVAL;
4845 goto restore_opts;
4846 }
4847 if (test_opt(sb, DIOREAD_NOLOCK)) {
4848 ext4_msg(sb, KERN_ERR, "can't mount with "
4849 "both data=journal and dioread_nolock");
4850 err = -EINVAL;
4851 goto restore_opts;
4852 }
4853 }
4854
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004855 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004856 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004857
4858 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004859 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004860
4861 es = sbi->s_es;
4862
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004863 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004864 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004865 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4866 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004867
Eric Sandeen661aa522012-02-20 17:53:04 -05004868 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004869 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004870 err = -EROFS;
4871 goto restore_opts;
4872 }
4873
4874 if (*flags & MS_RDONLY) {
Theodore Ts'o38c03b32014-03-13 22:49:42 -04004875 err = sync_filesystem(sb);
4876 if (err < 0)
4877 goto restore_opts;
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004878 err = dquot_suspend(sb, -1);
4879 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004880 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004881
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004882 /*
4883 * First of all, the unconditional stuff we have to do
4884 * to disable replay of the journal when we next remount
4885 */
4886 sb->s_flags |= MS_RDONLY;
4887
4888 /*
4889 * OK, test if we are remounting a valid rw partition
4890 * readonly, and if so set the rdonly flag and then
4891 * mark the partition as valid again.
4892 */
Mingming Cao617ba132006-10-11 01:20:53 -07004893 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4894 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004895 es->s_state = cpu_to_le16(sbi->s_mount_state);
4896
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004897 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004898 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004899 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004900 /* Make sure we can mount this feature set readwrite */
4901 if (!ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004902 err = -EROFS;
4903 goto restore_opts;
4904 }
Eric Sandeenead65962007-02-10 01:46:08 -08004905 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004906 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004907 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004908 */
4909 for (g = 0; g < sbi->s_groups_count; g++) {
4910 struct ext4_group_desc *gdp =
4911 ext4_get_group_desc(sb, g, NULL);
4912
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04004913 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004914 ext4_msg(sb, KERN_ERR,
4915 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04004916 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4917 le16_to_cpu(gdp->bg_checksum));
4918 err = -EINVAL;
4919 goto restore_opts;
4920 }
4921 }
4922
4923 /*
Eric Sandeenead65962007-02-10 01:46:08 -08004924 * If we have an unprocessed orphan list hanging
4925 * around from a previously readonly bdev mount,
4926 * require a full umount/remount for now.
4927 */
4928 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004929 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08004930 "remount RDWR because of unprocessed "
4931 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004932 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08004933 err = -EINVAL;
4934 goto restore_opts;
4935 }
4936
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004937 /*
4938 * Mounting a RDONLY partition read-write, so reread
4939 * and store the current valid flag. (It may have
4940 * been changed by e2fsck since we originally mounted
4941 * the partition.)
4942 */
Frank Mayhar03901312009-01-07 00:06:22 -05004943 if (sbi->s_journal)
4944 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004945 sbi->s_mount_state = le16_to_cpu(es->s_state);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004946 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004947 sb->s_flags &= ~MS_RDONLY;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004948 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4949 EXT4_FEATURE_INCOMPAT_MMP))
4950 if (ext4_multi_mount_protect(sb,
4951 le64_to_cpu(es->s_mmp_block))) {
4952 err = -EROFS;
4953 goto restore_opts;
4954 }
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004955 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004956 }
4957 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04004958
4959 /*
4960 * Reinitialize lazy itable initialization thread based on
4961 * current settings
4962 */
4963 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4964 ext4_unregister_li_request(sb);
4965 else {
4966 ext4_group_t first_not_zeroed;
4967 first_not_zeroed = ext4_has_uninit_itable(sb);
4968 ext4_register_li_request(sb, first_not_zeroed);
4969 }
4970
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004971 ext4_setup_system_zone(sb);
Michael Tokarevd096ad02012-12-25 14:08:16 -05004972 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004973 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05004974
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004975#ifdef CONFIG_QUOTA
4976 /* Release old quota file names */
4977 for (i = 0; i < MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05004978 kfree(old_opts.s_qf_names[i]);
Aditya Kali7c319d32012-07-22 20:21:31 -04004979 if (enable_quota) {
4980 if (sb_any_quota_suspended(sb))
4981 dquot_resume(sb, -1);
4982 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4983 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4984 err = ext4_enable_quotas(sb);
Theodore Ts'o07724f92012-08-17 19:08:42 -04004985 if (err)
Aditya Kali7c319d32012-07-22 20:21:31 -04004986 goto restore_opts;
Aditya Kali7c319d32012-07-22 20:21:31 -04004987 }
4988 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004989#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004990
4991 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4992 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004993 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004994
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004995restore_opts:
4996 sb->s_flags = old_sb_flags;
4997 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004998 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004999 sbi->s_resuid = old_opts.s_resuid;
5000 sbi->s_resgid = old_opts.s_resgid;
5001 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05005002 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5003 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005004#ifdef CONFIG_QUOTA
5005 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5006 for (i = 0; i < MAXQUOTAS; i++) {
Chen Gang03dafb52013-01-24 23:24:58 -05005007 kfree(sbi->s_qf_names[i]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005008 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5009 }
5010#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005011 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005012 return err;
5013}
5014
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005015static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005016{
5017 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07005018 struct ext4_sb_info *sbi = EXT4_SB(sb);
5019 struct ext4_super_block *es = sbi->s_es;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005020 ext4_fsblk_t overhead = 0, resv_blocks;
Pekka Enberg960cc392006-12-06 20:35:29 -08005021 u64 fsid;
Kazuya Miod02a9392011-05-24 18:30:07 -04005022 s64 bfree;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005023 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005024
Theodore Ts'o952fc182012-07-09 16:27:05 -04005025 if (!test_opt(sb, MINIX_DF))
5026 overhead = sbi->s_overhead;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005027
Mingming Cao617ba132006-10-11 01:20:53 -07005028 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005029 buf->f_bsize = sb->s_blocksize;
Eric Sandeenb72f78c2012-11-08 10:33:36 -05005030 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
Theodore Ts'o57042652011-09-09 18:56:51 -04005031 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5032 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
Kazuya Miod02a9392011-05-24 18:30:07 -04005033 /* prevent underflow in case that few free space is available */
Theodore Ts'o57042652011-09-09 18:56:51 -04005034 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
Lukas Czerner27dd4382013-04-09 22:11:22 -04005035 buf->f_bavail = buf->f_bfree -
5036 (ext4_r_blocks_count(es) + resv_blocks);
5037 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005038 buf->f_bavail = 0;
5039 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07005040 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07005041 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08005042 fsid = le64_to_cpup((void *)es->s_uuid) ^
5043 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5044 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5045 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005046
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005047 return 0;
5048}
5049
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005050/* Helper function for writing quotas on sync - we need to start transaction
5051 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005052 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07005053 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01005054 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05005055 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07005056 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005057 *
5058 */
5059
5060#ifdef CONFIG_QUOTA
5061
5062static inline struct inode *dquot_to_inode(struct dquot *dquot)
5063{
Eric W. Biederman4c376dc2012-09-16 03:56:19 -07005064 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005065}
5066
Mingming Cao617ba132006-10-11 01:20:53 -07005067static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005068{
5069 int ret, err;
5070 handle_t *handle;
5071 struct inode *inode;
5072
5073 inode = dquot_to_inode(dquot);
Theodore Ts'o9924a922013-02-08 21:59:22 -05005074 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005075 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005076 if (IS_ERR(handle))
5077 return PTR_ERR(handle);
5078 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005079 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005080 if (!ret)
5081 ret = err;
5082 return ret;
5083}
5084
Mingming Cao617ba132006-10-11 01:20:53 -07005085static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005086{
5087 int ret, err;
5088 handle_t *handle;
5089
Theodore Ts'o9924a922013-02-08 21:59:22 -05005090 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005091 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005092 if (IS_ERR(handle))
5093 return PTR_ERR(handle);
5094 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005095 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005096 if (!ret)
5097 ret = err;
5098 return ret;
5099}
5100
Mingming Cao617ba132006-10-11 01:20:53 -07005101static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005102{
5103 int ret, err;
5104 handle_t *handle;
5105
Theodore Ts'o9924a922013-02-08 21:59:22 -05005106 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005107 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07005108 if (IS_ERR(handle)) {
5109 /* Release dquot anyway to avoid endless cycle in dqput() */
5110 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005111 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07005112 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005113 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005114 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005115 if (!ret)
5116 ret = err;
5117 return ret;
5118}
5119
Mingming Cao617ba132006-10-11 01:20:53 -07005120static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005121{
Jan Kara262b4662013-03-02 17:57:08 -05005122 struct super_block *sb = dquot->dq_sb;
5123 struct ext4_sb_info *sbi = EXT4_SB(sb);
5124
Jan Kara2c8be6b2008-05-13 21:27:55 -04005125 /* Are we journaling quotas? */
Jan Kara262b4662013-03-02 17:57:08 -05005126 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5127 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005128 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005129 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005130 } else {
5131 return dquot_mark_dquot_dirty(dquot);
5132 }
5133}
5134
Mingming Cao617ba132006-10-11 01:20:53 -07005135static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005136{
5137 int ret, err;
5138 handle_t *handle;
5139
5140 /* Data block + inode block */
Theodore Ts'o9924a922013-02-08 21:59:22 -05005141 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005142 if (IS_ERR(handle))
5143 return PTR_ERR(handle);
5144 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07005145 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005146 if (!ret)
5147 ret = err;
5148 return ret;
5149}
5150
5151/*
5152 * Turn on quotas during mount time - we need to find
5153 * the quota file and such...
5154 */
Mingming Cao617ba132006-10-11 01:20:53 -07005155static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005156{
Christoph Hellwig287a8092010-05-19 07:16:45 -04005157 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5158 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005159}
5160
5161/*
5162 * Standard function to be called on quota_on
5163 */
Mingming Cao617ba132006-10-11 01:20:53 -07005164static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02005165 struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005166{
5167 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005168
5169 if (!test_opt(sb, QUOTA))
5170 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04005171
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005172 /* Quotafile not on the same filesystem? */
Al Virod8c95842011-12-07 18:16:57 -05005173 if (path->dentry->d_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005174 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04005175 /* Journaling quota? */
5176 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005177 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02005178 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04005179 ext4_msg(sb, KERN_WARNING,
5180 "Quota file not on filesystem root. "
5181 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005182 }
Jan Kara06235432008-05-13 19:11:51 -04005183
5184 /*
5185 * When we journal data on quota file, we have to flush journal to see
5186 * all updates to the file when we bypass pagecache...
5187 */
Frank Mayhar03901312009-01-07 00:06:22 -05005188 if (EXT4_SB(sb)->s_journal &&
Jan Karaf00c9e42010-09-15 17:38:58 +02005189 ext4_should_journal_data(path->dentry->d_inode)) {
Jan Kara06235432008-05-13 19:11:51 -04005190 /*
5191 * We don't need to lock updates but journal_flush() could
5192 * otherwise be livelocked...
5193 */
5194 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005195 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04005196 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02005197 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005198 return err;
Jan Kara06235432008-05-13 19:11:51 -04005199 }
5200
Jan Karaf00c9e42010-09-15 17:38:58 +02005201 return dquot_quota_on(sb, type, format_id, path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005202}
5203
Aditya Kali7c319d32012-07-22 20:21:31 -04005204static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5205 unsigned int flags)
5206{
5207 int err;
5208 struct inode *qf_inode;
5209 unsigned long qf_inums[MAXQUOTAS] = {
5210 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5211 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5212 };
5213
5214 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5215
5216 if (!qf_inums[type])
5217 return -EPERM;
5218
5219 qf_inode = ext4_iget(sb, qf_inums[type]);
5220 if (IS_ERR(qf_inode)) {
5221 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5222 return PTR_ERR(qf_inode);
5223 }
5224
Jan Karabcb13852013-04-09 09:21:41 -04005225 /* Don't account quota for quota files to avoid recursion */
5226 qf_inode->i_flags |= S_NOQUOTA;
Aditya Kali7c319d32012-07-22 20:21:31 -04005227 err = dquot_enable(qf_inode, type, format_id, flags);
5228 iput(qf_inode);
5229
5230 return err;
5231}
5232
5233/* Enable usage tracking for all quota types. */
5234static int ext4_enable_quotas(struct super_block *sb)
5235{
5236 int type, err = 0;
5237 unsigned long qf_inums[MAXQUOTAS] = {
5238 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5239 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5240 };
5241
5242 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5243 for (type = 0; type < MAXQUOTAS; type++) {
5244 if (qf_inums[type]) {
5245 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5246 DQUOT_USAGE_ENABLED);
5247 if (err) {
5248 ext4_warning(sb,
Theodore Ts'o72ba7452013-01-24 23:24:54 -05005249 "Failed to enable quota tracking "
5250 "(type=%d, err=%d). Please run "
5251 "e2fsck to fix.", type, err);
Aditya Kali7c319d32012-07-22 20:21:31 -04005252 return err;
5253 }
5254 }
5255 }
5256 return 0;
5257}
5258
5259/*
5260 * quota_on function that is used when QUOTA feature is set.
5261 */
5262static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5263 int format_id)
5264{
5265 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5266 return -EINVAL;
5267
5268 /*
5269 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5270 */
5271 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5272}
5273
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005274static int ext4_quota_off(struct super_block *sb, int type)
5275{
Jan Kara21f97692011-04-04 15:33:39 -04005276 struct inode *inode = sb_dqopt(sb)->files[type];
5277 handle_t *handle;
5278
Dmitry Monakhov87009d82010-11-08 13:47:33 -05005279 /* Force all delayed allocation blocks to be allocated.
5280 * Caller already holds s_umount sem */
5281 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005282 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005283
Amir Goldstein0b268592011-05-16 09:59:13 -04005284 if (!inode)
5285 goto out;
5286
Jan Kara21f97692011-04-04 15:33:39 -04005287 /* Update modification times of quota files when userspace can
5288 * start looking at them */
Theodore Ts'o9924a922013-02-08 21:59:22 -05005289 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
Jan Kara21f97692011-04-04 15:33:39 -04005290 if (IS_ERR(handle))
5291 goto out;
5292 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5293 ext4_mark_inode_dirty(handle, inode);
5294 ext4_journal_stop(handle);
5295
5296out:
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005297 return dquot_quota_off(sb, type);
5298}
5299
Aditya Kali7c319d32012-07-22 20:21:31 -04005300/*
5301 * quota_off function that is used when QUOTA feature is set.
5302 */
5303static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5304{
5305 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5306 return -EINVAL;
5307
5308 /* Disable only the limits. */
5309 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5310}
5311
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005312/* Read data from quotafile - avoid pagecache and such because we cannot afford
5313 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03005314 * itself serializes the operations (and no one else should touch the files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005315 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07005316static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005317 size_t len, loff_t off)
5318{
5319 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005320 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005321 int err = 0;
5322 int offset = off & (sb->s_blocksize - 1);
5323 int tocopy;
5324 size_t toread;
5325 struct buffer_head *bh;
5326 loff_t i_size = i_size_read(inode);
5327
5328 if (off > i_size)
5329 return 0;
5330 if (off+len > i_size)
5331 len = i_size-off;
5332 toread = len;
5333 while (toread > 0) {
5334 tocopy = sb->s_blocksize - offset < toread ?
5335 sb->s_blocksize - offset : toread;
Mingming Cao617ba132006-10-11 01:20:53 -07005336 bh = ext4_bread(NULL, inode, blk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005337 if (err)
5338 return err;
5339 if (!bh) /* A hole? */
5340 memset(data, 0, tocopy);
5341 else
5342 memcpy(data, bh->b_data+offset, tocopy);
5343 brelse(bh);
5344 offset = 0;
5345 toread -= tocopy;
5346 data += tocopy;
5347 blk++;
5348 }
5349 return len;
5350}
5351
5352/* Write to quotafile (we know the transaction is already started and has
5353 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07005354static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005355 const char *data, size_t len, loff_t off)
5356{
5357 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005358 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005359 int err = 0;
5360 int offset = off & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005361 struct buffer_head *bh;
5362 handle_t *handle = journal_current_handle();
5363
Frank Mayhar03901312009-01-07 00:06:22 -05005364 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005365 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5366 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07005367 (unsigned long long)off, (unsigned long long)len);
5368 return -EIO;
5369 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005370 /*
5371 * Since we account only one data block in transaction credits,
5372 * then it is impossible to cross a block boundary.
5373 */
5374 if (sb->s_blocksize - offset < len) {
5375 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5376 " cancelled because not block aligned",
5377 (unsigned long long)off, (unsigned long long)len);
5378 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005379 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005380
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005381 bh = ext4_bread(handle, inode, blk, 1, &err);
5382 if (!bh)
5383 goto out;
liang xie5d601252014-05-12 22:06:43 -04005384 BUFFER_TRACE(bh, "get write access");
Jan Kara62d2b5f2010-07-27 11:56:07 -04005385 err = ext4_journal_get_write_access(handle, bh);
5386 if (err) {
5387 brelse(bh);
5388 goto out;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005389 }
5390 lock_buffer(bh);
5391 memcpy(bh->b_data+offset, data, len);
5392 flush_dcache_page(bh->b_page);
5393 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04005394 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005395 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005396out:
Jan Kara0b7f7ce2012-04-25 21:26:54 +02005397 if (err)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005398 return err;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005399 if (inode->i_size < off + len) {
5400 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07005401 EXT4_I(inode)->i_disksize = inode->i_size;
Jan Kara21f97692011-04-04 15:33:39 -04005402 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005403 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005404 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005405}
5406
5407#endif
5408
Al Viro152a0832010-07-25 00:46:55 +04005409static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5410 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005411{
Al Viro152a0832010-07-25 00:46:55 +04005412 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005413}
5414
Theodore Ts'o37f328e2010-03-24 20:06:41 -04005415#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05005416static inline void register_as_ext2(void)
5417{
5418 int err = register_filesystem(&ext2_fs_type);
5419 if (err)
5420 printk(KERN_WARNING
5421 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5422}
5423
5424static inline void unregister_as_ext2(void)
5425{
5426 unregister_filesystem(&ext2_fs_type);
5427}
Theodore Ts'o2035e772011-04-18 17:29:14 -04005428
5429static inline int ext2_feature_set_ok(struct super_block *sb)
5430{
5431 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5432 return 0;
5433 if (sb->s_flags & MS_RDONLY)
5434 return 1;
5435 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5436 return 0;
5437 return 1;
5438}
Theodore Ts'o24b58422009-12-07 14:08:51 -05005439#else
5440static inline void register_as_ext2(void) { }
5441static inline void unregister_as_ext2(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04005442static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05005443#endif
5444
Theodore Ts'o37f328e2010-03-24 20:06:41 -04005445#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05005446static inline void register_as_ext3(void)
5447{
5448 int err = register_filesystem(&ext3_fs_type);
5449 if (err)
5450 printk(KERN_WARNING
5451 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5452}
5453
5454static inline void unregister_as_ext3(void)
5455{
5456 unregister_filesystem(&ext3_fs_type);
5457}
Theodore Ts'o2035e772011-04-18 17:29:14 -04005458
5459static inline int ext3_feature_set_ok(struct super_block *sb)
5460{
5461 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5462 return 0;
5463 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5464 return 0;
5465 if (sb->s_flags & MS_RDONLY)
5466 return 1;
5467 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5468 return 0;
5469 return 1;
5470}
Theodore Ts'o24b58422009-12-07 14:08:51 -05005471#else
5472static inline void register_as_ext3(void) { }
5473static inline void unregister_as_ext3(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04005474static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05005475#endif
5476
Theodore Ts'o03010a32008-10-10 20:02:48 -04005477static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005478 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04005479 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04005480 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005481 .kill_sb = kill_block_super,
5482 .fs_flags = FS_REQUIRES_DEV,
5483};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08005484MODULE_ALIAS_FS("ext4");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005485
Lukas Czerner8f021222011-02-03 14:33:33 -05005486static int __init ext4_init_feat_adverts(void)
Lukas Czerner857ac882010-10-27 21:30:05 -04005487{
5488 struct ext4_features *ef;
5489 int ret = -ENOMEM;
5490
5491 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5492 if (!ef)
5493 goto out;
5494
5495 ef->f_kobj.kset = ext4_kset;
5496 init_completion(&ef->f_kobj_unregister);
5497 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5498 "features");
5499 if (ret) {
5500 kfree(ef);
5501 goto out;
5502 }
5503
5504 ext4_feat = ef;
5505 ret = 0;
5506out:
5507 return ret;
5508}
5509
Lukas Czerner8f021222011-02-03 14:33:33 -05005510static void ext4_exit_feat_adverts(void)
5511{
5512 kobject_put(&ext4_feat->f_kobj);
5513 wait_for_completion(&ext4_feat->f_kobj_unregister);
5514 kfree(ext4_feat);
5515}
5516
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005517/* Shared across all ext4 file systems */
5518wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5519struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5520
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005521static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005522{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005523 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05005524
Al Viro07c0c5d82012-03-20 22:05:02 -04005525 ext4_li_info = NULL;
5526 mutex_init(&ext4_li_mtx);
5527
Carlos Maiolino9a4c8012012-12-10 16:30:45 -05005528 /* Build-time check for flags consistency */
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04005529 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005530
5531 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5532 mutex_init(&ext4__aio_mutex[i]);
5533 init_waitqueue_head(&ext4__ioend_wq[i]);
5534 }
5535
Zheng Liu51865fd2012-11-08 21:57:32 -05005536 err = ext4_init_es();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005537 if (err)
5538 return err;
Zheng Liu51865fd2012-11-08 21:57:32 -05005539
5540 err = ext4_init_pageio();
5541 if (err)
5542 goto out7;
5543
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005544 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04005545 if (err)
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005546 goto out6;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04005547 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
Theodore Ts'o0e376b12012-08-17 10:04:17 -04005548 if (!ext4_kset) {
5549 err = -ENOMEM;
Theodore Ts'odd683142011-02-03 14:33:49 -05005550 goto out5;
Theodore Ts'o0e376b12012-08-17 10:04:17 -04005551 }
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005552 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Lukas Czerner857ac882010-10-27 21:30:05 -04005553
5554 err = ext4_init_feat_adverts();
Theodore Ts'odd683142011-02-03 14:33:49 -05005555 if (err)
5556 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04005557
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005558 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005559 if (err)
Alex Tomasc9de5602008-01-29 00:19:52 -05005560 goto out2;
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005561 else
5562 ext4_mballoc_ready = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005563 err = init_inodecache();
5564 if (err)
5565 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05005566 register_as_ext3();
Theodore Ts'o2035e772011-04-18 17:29:14 -04005567 register_as_ext2();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005568 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005569 if (err)
5570 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04005571
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005572 return 0;
5573out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05005574 unregister_as_ext2();
5575 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005576 destroy_inodecache();
5577out1:
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005578 ext4_mballoc_ready = 0;
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005579 ext4_exit_mballoc();
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005580out2:
Lukas Czerner8f021222011-02-03 14:33:33 -05005581 ext4_exit_feat_adverts();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005582out4:
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005583 if (ext4_proc_root)
5584 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'odd683142011-02-03 14:33:49 -05005585 kset_unregister(ext4_kset);
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005586out5:
Theodore Ts'odd683142011-02-03 14:33:49 -05005587 ext4_exit_system_zone();
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005588out6:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005589 ext4_exit_pageio();
Zheng Liu51865fd2012-11-08 21:57:32 -05005590out7:
5591 ext4_exit_es();
5592
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005593 return err;
5594}
5595
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005596static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005597{
Lukas Czernerbfff6872010-10-27 21:30:05 -04005598 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05005599 unregister_as_ext2();
5600 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005601 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005602 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005603 ext4_exit_mballoc();
Lukas Czerner8f021222011-02-03 14:33:33 -05005604 ext4_exit_feat_adverts();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04005605 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04005606 kset_unregister(ext4_kset);
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005607 ext4_exit_system_zone();
5608 ext4_exit_pageio();
Eric Sandeendd12ed12013-07-26 15:21:11 -04005609 ext4_exit_es();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005610}
5611
5612MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05005613MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005614MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005615module_init(ext4_init_fs)
5616module_exit(ext4_exit_fs)