blob: f06d0589ddba5db7226a38a6e0de3ef2cdee48d9 [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>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070024#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070028#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070029#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070030#include <linux/vfs.h>
31#include <linux/random.h>
32#include <linux/mount.h>
33#include <linux/namei.h>
34#include <linux/quotaops.h>
35#include <linux/seq_file.h>
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040036#include <linux/proc_fs.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040037#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040038#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040039#include <linux/crc16.h>
Dan Magenheimer7abc52c2011-05-26 10:02:03 -060040#include <linux/cleancache.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070041#include <asm/uaccess.h>
42
Lukas Czernerbfff6872010-10-27 21:30:05 -040043#include <linux/kthread.h>
44#include <linux/freezer.h>
45
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040046#include "ext4.h"
Theodore Ts'o4a092d72012-11-28 13:03:30 -050047#include "ext4_extents.h" /* Needed for trace points definition */
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040048#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070049#include "xattr.h"
50#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040051#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070052
Theodore Ts'o9bffad12009-06-17 11:48:11 -040053#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
Theodore Ts'o1f109d52010-10-27 21:30:14 -040056static struct proc_dir_entry *ext4_proc_root;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040057static struct kset *ext4_kset;
Lukas Czerner0b75a842011-02-23 12:22:49 -050058static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ext4_features *ext4_feat;
T Makphaibulchoke9c191f72014-03-18 19:24:49 -040061static int ext4_mballoc_ready;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040062
Mingming Cao617ba132006-10-11 01:20:53 -070063static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070064 unsigned long journal_devnum);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -050065static int ext4_show_options(struct seq_file *seq, struct dentry *root);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040066static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040067static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070071static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040072static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080074static int ext4_unfreeze(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080075static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040076static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
Theodore Ts'o2035e772011-04-18 17:29:14 -040078static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
Amir Goldsteind39195c2011-02-28 00:53:45 -050080static int ext4_feature_set_ok(struct super_block *sb, int readonly);
Lukas Czernerbfff6872010-10-27 21:30:05 -040081static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050083static void ext4_clear_request_list(void);
Lukas Czerner27dd4382013-04-09 22:11:22 -040084static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070085
Theodore Ts'o2035e772011-04-18 17:29:14 -040086#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
Eric W. Biederman7f78e032013-03-02 19:39:14 -080094MODULE_ALIAS_FS("ext2");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -070095MODULE_ALIAS("ext2");
Theodore Ts'o2035e772011-04-18 17:29:14 -040096#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
Jan Karaba69f9a2010-03-24 20:18:37 -0400102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +0400106 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -0400107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
Eric W. Biederman7f78e032013-03-02 19:39:14 -0800110MODULE_ALIAS_FS("ext3");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -0700111MODULE_ALIAS("ext3");
Jan Karaba69f9a2010-03-24 20:18:37 -0400112#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113#else
114#define IS_EXT3_SB(sb) (0)
115#endif
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700116
Darrick J. Wongd25425f2012-04-29 18:25:10 -0400117static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119{
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125}
126
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400127static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129{
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137}
138
Stephen Hemmingerc1978552014-05-12 10:50:23 -0400139static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400141{
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -0400142 if (!ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400143 return 1;
144
145 return es->s_checksum == ext4_superblock_csum(sb, es);
146}
147
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400148void ext4_superblock_csum_set(struct super_block *sb)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400149{
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -0400152 if (!ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400153 return;
154
155 es->s_checksum = ext4_superblock_csum(sb, es);
156}
157
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400158void *ext4_kvmalloc(size_t size, gfp_t flags)
159{
160 void *ret;
161
Joe Perches8be04b92013-06-19 12:15:53 -0700162 ret = kmalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400163 if (!ret)
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
165 return ret;
166}
167
168void *ext4_kvzalloc(size_t size, gfp_t flags)
169{
170 void *ret;
171
Joe Perches8be04b92013-06-19 12:15:53 -0700172 ret = kzalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400173 if (!ret)
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 return ret;
176}
177
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700178ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700180{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700184}
185
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700186ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700188{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700192}
193
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700194ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700196{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400197 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700200}
201
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400202__u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500204{
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500208}
209
210__u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212{
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500216}
217
218__u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500224}
225
226__u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500232}
233
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700234void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700236{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700240}
241
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700242void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700244{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700248}
249
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700250void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700252{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700256}
257
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400258void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500260{
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
264}
265
266void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268{
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
288}
289
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400290
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400291static void __save_error_info(struct super_block *sb, const char *func,
292 unsigned int line)
293{
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298 es->s_last_error_time = cpu_to_le32(get_seconds());
299 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
300 es->s_last_error_line = cpu_to_le32(line);
301 if (!es->s_first_error_time) {
302 es->s_first_error_time = es->s_last_error_time;
303 strncpy(es->s_first_error_func, func,
304 sizeof(es->s_first_error_func));
305 es->s_first_error_line = cpu_to_le32(line);
306 es->s_first_error_ino = es->s_last_error_ino;
307 es->s_first_error_block = es->s_last_error_block;
308 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400309 /*
310 * Start the daily error reporting function if it hasn't been
311 * started already
312 */
313 if (!es->s_error_count)
314 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Wei Yongjunba39ebb2012-09-27 09:37:53 -0400315 le32_add_cpu(&es->s_error_count, 1);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400316}
317
318static void save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 __save_error_info(sb, func, line);
322 ext4_commit_super(sb, 1);
323}
324
Bobi Jam18aadd42012-02-20 17:53:02 -0500325static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
326{
327 struct super_block *sb = journal->j_private;
328 struct ext4_sb_info *sbi = EXT4_SB(sb);
329 int error = is_journal_aborted(journal);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400330 struct ext4_journal_cb_entry *jce;
Bobi Jam18aadd42012-02-20 17:53:02 -0500331
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400332 BUG_ON(txn->t_state == T_FINISHED);
Bobi Jam18aadd42012-02-20 17:53:02 -0500333 spin_lock(&sbi->s_md_lock);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400334 while (!list_empty(&txn->t_private_list)) {
335 jce = list_entry(txn->t_private_list.next,
336 struct ext4_journal_cb_entry, jce_list);
Bobi Jam18aadd42012-02-20 17:53:02 -0500337 list_del_init(&jce->jce_list);
338 spin_unlock(&sbi->s_md_lock);
339 jce->jce_func(sb, jce, error);
340 spin_lock(&sbi->s_md_lock);
341 }
342 spin_unlock(&sbi->s_md_lock);
343}
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400344
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700345/* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
347 *
348 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700349 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
354 *
Mingming Caodab291a2006-10-11 01:21:01 -0700355 * 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 -0200356 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700357 * that error until we've noted it down and cleared it.
358 */
359
Mingming Cao617ba132006-10-11 01:20:53 -0700360static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700361{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700362 if (sb->s_flags & MS_RDONLY)
363 return;
364
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400365 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700366 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700367
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700369 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700370 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700371 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400372 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400374 /*
375 * Make sure updated value of ->s_mount_flags will be visible
376 * before ->s_flags update
377 */
378 smp_wmb();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700379 sb->s_flags |= MS_RDONLY;
380 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700381 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700382 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700383 sb->s_id);
384}
385
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400386#define ext4_error_ratelimit(sb) \
387 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
388 "EXT4-fs error")
389
Eric Sandeen12062dd2010-02-15 14:19:27 -0500390void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400391 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700392{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500393 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700394 va_list args;
395
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400396 if (ext4_error_ratelimit(sb)) {
397 va_start(args, fmt);
398 vaf.fmt = fmt;
399 vaf.va = &args;
400 printk(KERN_CRIT
401 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
402 sb->s_id, function, line, current->comm, &vaf);
403 va_end(args);
404 }
Theodore Ts'of3fc02102012-05-30 23:00:16 -0400405 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700406 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700407}
408
Joe Perchese7c96e82013-07-01 08:12:37 -0400409void __ext4_error_inode(struct inode *inode, const char *function,
410 unsigned int line, ext4_fsblk_t block,
411 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500412{
413 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500414 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400415 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500416
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400417 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
418 es->s_last_error_block = cpu_to_le64(block);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400419 if (ext4_error_ratelimit(inode->i_sb)) {
420 va_start(args, fmt);
421 vaf.fmt = fmt;
422 vaf.va = &args;
423 if (block)
424 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
425 "inode #%lu: block %llu: comm %s: %pV\n",
426 inode->i_sb->s_id, function, line, inode->i_ino,
427 block, current->comm, &vaf);
428 else
429 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
430 "inode #%lu: comm %s: %pV\n",
431 inode->i_sb->s_id, function, line, inode->i_ino,
432 current->comm, &vaf);
433 va_end(args);
434 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400435 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500436 ext4_handle_error(inode->i_sb);
437}
438
Joe Perchese7c96e82013-07-01 08:12:37 -0400439void __ext4_error_file(struct file *file, const char *function,
440 unsigned int line, ext4_fsblk_t block,
441 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500442{
443 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500444 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400445 struct ext4_super_block *es;
Al Viro496ad9a2013-01-23 17:07:38 -0500446 struct inode *inode = file_inode(file);
Frank Mayhar273df552010-03-02 11:46:09 -0500447 char pathname[80], *path;
448
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400449 es = EXT4_SB(inode->i_sb)->s_es;
450 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400451 if (ext4_error_ratelimit(inode->i_sb)) {
452 path = d_path(&(file->f_path), pathname, sizeof(pathname));
453 if (IS_ERR(path))
454 path = "(unknown)";
455 va_start(args, fmt);
456 vaf.fmt = fmt;
457 vaf.va = &args;
458 if (block)
459 printk(KERN_CRIT
460 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
461 "block %llu: comm %s: path %s: %pV\n",
462 inode->i_sb->s_id, function, line, inode->i_ino,
463 block, current->comm, path, &vaf);
464 else
465 printk(KERN_CRIT
466 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
467 "comm %s: path %s: %pV\n",
468 inode->i_sb->s_id, function, line, inode->i_ino,
469 current->comm, path, &vaf);
470 va_end(args);
471 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400472 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500473 ext4_handle_error(inode->i_sb);
474}
475
Theodore Ts'o722887d2013-02-08 13:00:31 -0500476const char *ext4_decode_error(struct super_block *sb, int errno,
477 char nbuf[16])
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700478{
479 char *errstr = NULL;
480
481 switch (errno) {
482 case -EIO:
483 errstr = "IO failure";
484 break;
485 case -ENOMEM:
486 errstr = "Out of memory";
487 break;
488 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400489 if (!sb || (EXT4_SB(sb)->s_journal &&
490 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700491 errstr = "Journal has aborted";
492 else
493 errstr = "Readonly filesystem";
494 break;
495 default:
496 /* If the caller passed in an extra buffer for unknown
497 * errors, textualise them now. Else we just return
498 * NULL. */
499 if (nbuf) {
500 /* Check for truncated error codes... */
501 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
502 errstr = nbuf;
503 }
504 break;
505 }
506
507 return errstr;
508}
509
Mingming Cao617ba132006-10-11 01:20:53 -0700510/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700511 * automatically and invokes the appropriate error response. */
512
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400513void __ext4_std_error(struct super_block *sb, const char *function,
514 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700515{
516 char nbuf[16];
517 const char *errstr;
518
519 /* Special case: if the error is EROFS, and we're not already
520 * inside a transaction, then there's really no point in logging
521 * an error. */
522 if (errno == -EROFS && journal_current_handle() == NULL &&
523 (sb->s_flags & MS_RDONLY))
524 return;
525
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400526 if (ext4_error_ratelimit(sb)) {
527 errstr = ext4_decode_error(sb, errno, nbuf);
528 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
529 sb->s_id, function, line, errstr);
530 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700531
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400532 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700533 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700534}
535
536/*
Mingming Cao617ba132006-10-11 01:20:53 -0700537 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700538 * abort function may be used to deal with unrecoverable failures such
539 * as journal IO errors or ENOMEM at a critical moment in log management.
540 *
541 * We unconditionally force the filesystem into an ABORT|READONLY state,
542 * unless the error response on the fs has been set to panic in which
543 * case we take the easy way out and panic immediately.
544 */
545
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400546void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400547 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700548{
549 va_list args;
550
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400551 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700552 va_start(args, fmt);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400553 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
554 function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700555 vprintk(fmt, args);
556 printk("\n");
557 va_end(args);
558
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400559 if ((sb->s_flags & MS_RDONLY) == 0) {
560 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400561 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400562 /*
563 * Make sure updated value of ->s_mount_flags will be visible
564 * before ->s_flags update
565 */
566 smp_wmb();
567 sb->s_flags |= MS_RDONLY;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400568 if (EXT4_SB(sb)->s_journal)
569 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
570 save_error_info(sb, function, line);
571 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700572 if (test_opt(sb, ERRORS_PANIC))
Mingming Cao617ba132006-10-11 01:20:53 -0700573 panic("EXT4-fs panic from previous error\n");
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700574}
575
Joe Perchese7c96e82013-07-01 08:12:37 -0400576void __ext4_msg(struct super_block *sb,
577 const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400578{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500579 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400580 va_list args;
581
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400582 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
583 return;
584
Eric Sandeenb31e1552009-06-04 17:36:36 -0400585 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500586 vaf.fmt = fmt;
587 vaf.va = &args;
588 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400589 va_end(args);
590}
591
Eric Sandeen12062dd2010-02-15 14:19:27 -0500592void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400593 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700594{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500595 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700596 va_list args;
597
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400598 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
599 "EXT4-fs warning"))
600 return;
601
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700602 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
606 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700607 va_end(args);
608}
609
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400610void __ext4_grp_locked_error(const char *function, unsigned int line,
611 struct super_block *sb, ext4_group_t grp,
612 unsigned long ino, ext4_fsblk_t block,
613 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500614__releases(bitlock)
615__acquires(bitlock)
616{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500617 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500618 va_list args;
619 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
620
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400621 es->s_last_error_ino = cpu_to_le32(ino);
622 es->s_last_error_block = cpu_to_le64(block);
623 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500624
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400625 if (ext4_error_ratelimit(sb)) {
626 va_start(args, fmt);
627 vaf.fmt = fmt;
628 vaf.va = &args;
629 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
630 sb->s_id, function, line, grp);
631 if (ino)
632 printk(KERN_CONT "inode %lu: ", ino);
633 if (block)
634 printk(KERN_CONT "block %llu:",
635 (unsigned long long) block);
636 printk(KERN_CONT "%pV\n", &vaf);
637 va_end(args);
638 }
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500639
640 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400641 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500642 return;
643 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400644
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500645 ext4_unlock_group(sb, grp);
646 ext4_handle_error(sb);
647 /*
648 * We only get here in the ERRORS_RO case; relocking the group
649 * may be dangerous, but nothing bad will happen since the
650 * filesystem will have already been marked read/only and the
651 * journal has been aborted. We return 1 as a hint to callers
652 * who might what to use the return value from
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300653 * ext4_grp_locked_error() to distinguish between the
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500654 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
655 * aggressively from the ext4 function in question, with a
656 * more appropriate error code.
657 */
658 ext4_lock_group(sb, grp);
659 return;
660}
661
Mingming Cao617ba132006-10-11 01:20:53 -0700662void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700663{
Mingming Cao617ba132006-10-11 01:20:53 -0700664 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700665
Mingming Cao617ba132006-10-11 01:20:53 -0700666 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700667 return;
668
Eric Sandeen12062dd2010-02-15 14:19:27 -0500669 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700670 "updating to rev %d because of new feature flag, "
671 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700672 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700673
Mingming Cao617ba132006-10-11 01:20:53 -0700674 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
675 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
676 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700677 /* leave es->s_feature_*compat flags alone */
678 /* es->s_uuid will be set by e2fsck if empty */
679
680 /*
681 * The rest of the superblock fields should be zero, and if not it
682 * means they are likely already in use, so leave them alone. We
683 * can leave it up to e2fsck to clean up any inconsistencies there.
684 */
685}
686
687/*
688 * Open the external journal device
689 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400690static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700691{
692 struct block_device *bdev;
693 char b[BDEVNAME_SIZE];
694
Tejun Heod4d77622010-11-13 11:55:18 +0100695 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700696 if (IS_ERR(bdev))
697 goto fail;
698 return bdev;
699
700fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400701 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700702 __bdevname(dev, b), PTR_ERR(bdev));
703 return NULL;
704}
705
706/*
707 * Release the journal device
708 */
Al Viro4385bab2013-05-05 22:11:03 -0400709static void ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700710{
Al Viro4385bab2013-05-05 22:11:03 -0400711 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700712}
713
Al Viro4385bab2013-05-05 22:11:03 -0400714static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700715{
716 struct block_device *bdev;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700717 bdev = sbi->journal_bdev;
718 if (bdev) {
Al Viro4385bab2013-05-05 22:11:03 -0400719 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700720 sbi->journal_bdev = NULL;
721 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700722}
723
724static inline struct inode *orphan_list_entry(struct list_head *l)
725{
Mingming Cao617ba132006-10-11 01:20:53 -0700726 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700727}
728
Mingming Cao617ba132006-10-11 01:20:53 -0700729static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700730{
731 struct list_head *l;
732
Eric Sandeenb31e1552009-06-04 17:36:36 -0400733 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
734 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700735
736 printk(KERN_ERR "sb_info orphan list:\n");
737 list_for_each(l, &sbi->s_orphan) {
738 struct inode *inode = orphan_list_entry(l);
739 printk(KERN_ERR " "
740 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
741 inode->i_sb->s_id, inode->i_ino, inode,
742 inode->i_mode, inode->i_nlink,
743 NEXT_ORPHAN(inode));
744 }
745}
746
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400747static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700748{
Mingming Cao617ba132006-10-11 01:20:53 -0700749 struct ext4_sb_info *sbi = EXT4_SB(sb);
750 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400751 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700752
Lukas Czerner857ac882010-10-27 21:30:05 -0400753 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400754 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
755
Jan Kara2e8fa542013-06-04 14:21:02 -0400756 flush_workqueue(sbi->rsv_conversion_wq);
Jan Kara2e8fa542013-06-04 14:21:02 -0400757 destroy_workqueue(sbi->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -0400758
Frank Mayhar03901312009-01-07 00:06:22 -0500759 if (sbi->s_journal) {
760 err = jbd2_journal_destroy(sbi->s_journal);
761 sbi->s_journal = NULL;
762 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400763 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500764 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500765
Zheng Liud3922a72013-07-01 08:12:37 -0400766 ext4_es_unregister_shrinker(sbi);
Al Viro9105bb12013-12-08 20:52:31 -0500767 del_timer_sync(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500768 ext4_release_system_zone(sb);
769 ext4_mb_release(sb);
770 ext4_ext_release(sb);
771 ext4_xattr_put_super(sb);
772
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700773 if (!(sb->s_flags & MS_RDONLY)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700774 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700775 es->s_state = cpu_to_le16(sbi->s_mount_state);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700776 }
Artem Bityutskiy58c58732012-07-22 20:33:31 -0400777 if (!(sb->s_flags & MS_RDONLY))
Artem Bityutskiya8e25a82012-03-21 22:29:15 -0400778 ext4_commit_super(sb, 1);
779
Theodore Ts'o240799c2008-10-09 23:53:47 -0400780 if (sbi->s_proc) {
Theodore Ts'o66acdcf2012-03-04 20:21:38 -0500781 remove_proc_entry("options", sbi->s_proc);
Theodore Ts'o9f6200b2008-09-23 09:18:24 -0400782 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -0400783 }
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400784 kobject_del(&sbi->s_kobj);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700785
786 for (i = 0; i < sbi->s_gdb_count; i++)
787 brelse(sbi->s_group_desc[i]);
Al Virob93b41d2014-11-20 12:19:11 -0500788 kvfree(sbi->s_group_desc);
789 kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -0400790 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700791 percpu_counter_destroy(&sbi->s_freeinodes_counter);
792 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -0400793 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700794 brelse(sbi->s_sbh);
795#ifdef CONFIG_QUOTA
Jan Karaa2d4a642014-09-11 11:15:15 -0400796 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700797 kfree(sbi->s_qf_names[i]);
798#endif
799
800 /* Debugging code just in case the in-memory inode orphan list
801 * isn't empty. The on-disk one can be non-empty if we've
802 * detected an error and taken the fs readonly, but the
803 * in-memory list had better be clean by this point. */
804 if (!list_empty(&sbi->s_orphan))
805 dump_orphan_list(sb, sbi);
806 J_ASSERT(list_empty(&sbi->s_orphan));
807
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700808 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700809 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
810 /*
811 * Invalidate the journal device's buffers. We don't want them
812 * floating about in memory - the physical journal device may
813 * hotswapped, and it breaks the `ro-after' testing code.
814 */
815 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700816 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700817 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700818 }
T Makphaibulchoke9c191f72014-03-18 19:24:49 -0400819 if (sbi->s_mb_cache) {
820 ext4_xattr_destroy_cache(sbi->s_mb_cache);
821 sbi->s_mb_cache = NULL;
822 }
Johann Lombardic5e06d12011-05-24 18:31:25 -0400823 if (sbi->s_mmp_tsk)
824 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700825 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400826 /*
827 * Now that we are completely done shutting down the
828 * superblock, we need to actually destroy the kobject.
829 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400830 kobject_put(&sbi->s_kobj);
831 wait_for_completion(&sbi->s_kobj_unregister);
Darrick J. Wong04419842012-04-29 18:27:10 -0400832 if (sbi->s_chksum_driver)
833 crypto_free_shash(sbi->s_chksum_driver);
Pekka Enberg705895b2009-02-15 18:07:52 -0500834 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700835 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700836}
837
Christoph Lametere18b8902006-12-06 20:33:20 -0800838static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700839
840/*
841 * Called inside transaction, so use GFP_NOFS
842 */
Mingming Cao617ba132006-10-11 01:20:53 -0700843static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700844{
Mingming Cao617ba132006-10-11 01:20:53 -0700845 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700846
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800847 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700848 if (!ei)
849 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400850
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700851 ei->vfs_inode.i_version = 1;
Theodore Ts'o202ee5d2014-04-21 14:37:55 -0400852 spin_lock_init(&ei->i_raw_lock);
Alex Tomasc9de5602008-01-29 00:19:52 -0500853 INIT_LIST_HEAD(&ei->i_prealloc_list);
854 spin_lock_init(&ei->i_prealloc_lock);
Zheng Liu9a26b662012-11-08 21:57:30 -0500855 ext4_es_init_tree(&ei->i_es_tree);
856 rwlock_init(&ei->i_es_lock);
Zheng Liuedaa53c2014-11-25 11:45:37 -0500857 INIT_LIST_HEAD(&ei->i_es_list);
Zheng Liueb68d0e2014-09-01 22:26:49 -0400858 ei->i_es_all_nr = 0;
Zheng Liuedaa53c2014-11-25 11:45:37 -0500859 ei->i_es_shk_nr = 0;
Jan Karadd475922014-11-25 11:51:23 -0500860 ei->i_es_shrink_lblk = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400861 ei->i_reserved_data_blocks = 0;
862 ei->i_reserved_meta_blocks = 0;
863 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500864 ei->i_da_metadata_calc_len = 0;
Theodore Ts'o7e731bc2012-08-05 23:28:16 -0400865 ei->i_da_metadata_calc_last_lblock = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400866 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300867#ifdef CONFIG_QUOTA
868 ei->i_reserved_quota = 0;
Jan Kara96c7e0d2014-09-29 14:58:25 +0200869 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300870#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500871 ei->jinode = NULL;
Jan Kara2e8fa542013-06-04 14:21:02 -0400872 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500873 spin_lock_init(&ei->i_completed_io_lock);
Jan Karab436b9b2009-12-08 23:51:10 -0500874 ei->i_sync_tid = 0;
875 ei->i_datasync_tid = 0;
Theodore Ts'of7ad6d22010-11-08 13:43:33 -0500876 atomic_set(&ei->i_ioend_count, 0);
Dmitry Monakhove27f41e2012-09-28 23:24:52 -0400877 atomic_set(&ei->i_unwritten, 0);
Jan Kara2e8fa542013-06-04 14:21:02 -0400878 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
Michael Halcrowb30ab0e2015-04-12 00:43:56 -0400879#ifdef CONFIG_EXT4_FS_ENCRYPTION
880 ei->i_encryption_key.mode = EXT4_ENCRYPTION_MODE_INVALID;
881#endif
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400882
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700883 return &ei->vfs_inode;
884}
885
Theodore Ts'o7ff9c072010-11-08 13:51:33 -0500886static int ext4_drop_inode(struct inode *inode)
887{
888 int drop = generic_drop_inode(inode);
889
890 trace_ext4_drop_inode(inode, drop);
891 return drop;
892}
893
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100894static void ext4_i_callback(struct rcu_head *head)
895{
896 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100897 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
898}
899
Mingming Cao617ba132006-10-11 01:20:53 -0700900static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700901{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700902 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400903 ext4_msg(inode->i_sb, KERN_ERR,
904 "Inode %lu (%p): orphan list check failed!",
905 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700906 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
907 EXT4_I(inode), sizeof(struct ext4_inode_info),
908 true);
909 dump_stack();
910 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100911 call_rcu(&inode->i_rcu, ext4_i_callback);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700912}
913
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700914static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700915{
Mingming Cao617ba132006-10-11 01:20:53 -0700916 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700917
Christoph Lametera35afb82007-05-16 22:10:57 -0700918 INIT_LIST_HEAD(&ei->i_orphan);
Christoph Lametera35afb82007-05-16 22:10:57 -0700919 init_rwsem(&ei->xattr_sem);
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500920 init_rwsem(&ei->i_data_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700921 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700922}
923
Fabian Fredericke67bc2b2014-02-17 20:34:53 -0500924static int __init init_inodecache(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700925{
Mingming Cao617ba132006-10-11 01:20:53 -0700926 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
927 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700928 0, (SLAB_RECLAIM_ACCOUNT|
929 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900930 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -0700931 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700932 return -ENOMEM;
933 return 0;
934}
935
936static void destroy_inodecache(void)
937{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +1000938 /*
939 * Make sure all delayed rcu free inodes are flushed before we
940 * destroy cache.
941 */
942 rcu_barrier();
Mingming Cao617ba132006-10-11 01:20:53 -0700943 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700944}
945
Al Viro0930fcc2010-06-07 13:16:22 -0400946void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700947{
Al Viro0930fcc2010-06-07 13:16:22 -0400948 invalidate_inode_buffers(inode);
Jan Karadbd57682012-05-03 14:48:02 +0200949 clear_inode(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -0500950 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -0400951 ext4_discard_preallocations(inode);
Zheng Liu51865fd2012-11-08 21:57:32 -0500952 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500953 if (EXT4_I(inode)->jinode) {
954 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
955 EXT4_I(inode)->jinode);
956 jbd2_free_inode(EXT4_I(inode)->jinode);
957 EXT4_I(inode)->jinode = NULL;
958 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700959}
960
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700961static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400962 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700963{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700964 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700965
Mingming Cao617ba132006-10-11 01:20:53 -0700966 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700967 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -0700968 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700969 return ERR_PTR(-ESTALE);
970
971 /* iget isn't really right if the inode is currently unallocated!!
972 *
Mingming Cao617ba132006-10-11 01:20:53 -0700973 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700974 * deleted, so we should be safe.
975 *
976 * Currently we don't know the generation for parent directory, so
977 * a generation of 0 means "accept any"
978 */
Theodore Ts'of4bb2982014-10-05 22:56:00 -0400979 inode = ext4_iget_normal(sb, ino);
David Howells1d1fe1e2008-02-07 00:15:37 -0800980 if (IS_ERR(inode))
981 return ERR_CAST(inode);
982 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700983 iput(inode);
984 return ERR_PTR(-ESTALE);
985 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700986
987 return inode;
988}
989
990static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400991 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700992{
993 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
994 ext4_nfs_get_inode);
995}
996
997static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400998 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -0700999{
1000 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1001 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001002}
1003
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001004/*
1005 * Try to release metadata pages (indirect blocks, directories) which are
1006 * mapped via the block device. Since these pages could have journal heads
1007 * which would prevent try_to_free_buffers() from freeing them, we must use
1008 * jbd2 layer's try_to_free_buffers() function to release them.
1009 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001010static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1011 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001012{
1013 journal_t *journal = EXT4_SB(sb)->s_journal;
1014
1015 WARN_ON(PageChecked(page));
1016 if (!page_has_buffers(page))
1017 return 0;
1018 if (journal)
1019 return jbd2_journal_try_to_free_buffers(journal, page,
1020 wait & ~__GFP_WAIT);
1021 return try_to_free_buffers(page);
1022}
1023
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001024#ifdef CONFIG_QUOTA
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04001025#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001026#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001027
Mingming Cao617ba132006-10-11 01:20:53 -07001028static int ext4_write_dquot(struct dquot *dquot);
1029static int ext4_acquire_dquot(struct dquot *dquot);
1030static int ext4_release_dquot(struct dquot *dquot);
1031static int ext4_mark_dquot_dirty(struct dquot *dquot);
1032static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001033static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02001034 struct path *path);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001035static int ext4_quota_off(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001036static int ext4_quota_on_mount(struct super_block *sb, int type);
1037static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001038 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001039static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001040 const char *data, size_t len, loff_t off);
Aditya Kali7c319d32012-07-22 20:21:31 -04001041static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1042 unsigned int flags);
1043static int ext4_enable_quotas(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001044
Jan Kara96c7e0d2014-09-29 14:58:25 +02001045static struct dquot **ext4_get_dquots(struct inode *inode)
1046{
1047 return EXT4_I(inode)->i_dquot;
1048}
1049
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001050static const struct dquot_operations ext4_quota_operations = {
Mingming Cao60e58e02009-01-22 18:13:05 +01001051 .get_reserved_space = ext4_get_reserved_space,
Mingming Cao617ba132006-10-11 01:20:53 -07001052 .write_dquot = ext4_write_dquot,
1053 .acquire_dquot = ext4_acquire_dquot,
1054 .release_dquot = ext4_release_dquot,
1055 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001056 .write_info = ext4_write_info,
1057 .alloc_dquot = dquot_alloc,
1058 .destroy_dquot = dquot_destroy,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001059};
1060
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001061static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001062 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001063 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001064 .quota_sync = dquot_quota_sync,
Jan Kara0a2403392014-11-19 00:42:09 +01001065 .get_state = dquot_get_state,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001066 .set_info = dquot_set_dqinfo,
1067 .get_dqblk = dquot_get_dqblk,
1068 .set_dqblk = dquot_set_dqblk
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001069};
1070#endif
1071
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001072static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001073 .alloc_inode = ext4_alloc_inode,
1074 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001075 .write_inode = ext4_write_inode,
1076 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001077 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001078 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001079 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001080 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001081 .freeze_fs = ext4_freeze,
1082 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001083 .statfs = ext4_statfs,
1084 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001085 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001086#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001087 .quota_read = ext4_quota_read,
1088 .quota_write = ext4_quota_write,
Jan Kara96c7e0d2014-09-29 14:58:25 +02001089 .get_dquots = ext4_get_dquots,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001090#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001091 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001092};
1093
Christoph Hellwig39655162007-10-21 16:42:17 -07001094static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001095 .fh_to_dentry = ext4_fh_to_dentry,
1096 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001097 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001098};
1099
1100enum {
1101 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1102 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001103 Opt_nouid32, Opt_debug, Opt_removed,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001104 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001105 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
Eric Sandeenad4eec62013-08-28 19:05:07 -04001106 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1107 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001108 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001109 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001110 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001111 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
Theodore Ts'oee4a3fc2012-03-02 12:14:24 -05001112 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001113 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
Theodore Ts'o14490322010-12-14 15:27:50 -05001114 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001115 Opt_lazytime, Opt_nolazytime,
Theodore Ts'o14490322010-12-14 15:27:50 -05001116 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001117 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001118 Opt_dioread_nolock, Opt_dioread_lock,
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001119 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001120 Opt_max_dir_size_kb, Opt_nojournal_checksum,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001121};
1122
Steven Whitehousea447c092008-10-13 10:46:57 +01001123static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001124 {Opt_bsd_df, "bsddf"},
1125 {Opt_minix_df, "minixdf"},
1126 {Opt_grpid, "grpid"},
1127 {Opt_grpid, "bsdgroups"},
1128 {Opt_nogrpid, "nogrpid"},
1129 {Opt_nogrpid, "sysvgroups"},
1130 {Opt_resgid, "resgid=%u"},
1131 {Opt_resuid, "resuid=%u"},
1132 {Opt_sb, "sb=%u"},
1133 {Opt_err_cont, "errors=continue"},
1134 {Opt_err_panic, "errors=panic"},
1135 {Opt_err_ro, "errors=remount-ro"},
1136 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001137 {Opt_debug, "debug"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001138 {Opt_removed, "oldalloc"},
1139 {Opt_removed, "orlov"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001140 {Opt_user_xattr, "user_xattr"},
1141 {Opt_nouser_xattr, "nouser_xattr"},
1142 {Opt_acl, "acl"},
1143 {Opt_noacl, "noacl"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001144 {Opt_noload, "norecovery"},
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001145 {Opt_noload, "noload"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001146 {Opt_removed, "nobh"},
1147 {Opt_removed, "bh"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001148 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001149 {Opt_min_batch_time, "min_batch_time=%u"},
1150 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001151 {Opt_journal_dev, "journal_dev=%u"},
Eric Sandeenad4eec62013-08-28 19:05:07 -04001152 {Opt_journal_path, "journal_path=%s"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001153 {Opt_journal_checksum, "journal_checksum"},
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001154 {Opt_nojournal_checksum, "nojournal_checksum"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001155 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001156 {Opt_abort, "abort"},
1157 {Opt_data_journal, "data=journal"},
1158 {Opt_data_ordered, "data=ordered"},
1159 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001160 {Opt_data_err_abort, "data_err=abort"},
1161 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001162 {Opt_offusrjquota, "usrjquota="},
1163 {Opt_usrjquota, "usrjquota=%s"},
1164 {Opt_offgrpjquota, "grpjquota="},
1165 {Opt_grpjquota, "grpjquota=%s"},
1166 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1167 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001168 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001169 {Opt_grpquota, "grpquota"},
1170 {Opt_noquota, "noquota"},
1171 {Opt_quota, "quota"},
1172 {Opt_usrquota, "usrquota"},
1173 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001174 {Opt_barrier, "barrier"},
1175 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001176 {Opt_i_version, "i_version"},
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001177 {Opt_dax, "dax"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001178 {Opt_stripe, "stripe=%u"},
Alex Tomas64769242008-07-11 19:27:31 -04001179 {Opt_delalloc, "delalloc"},
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001180 {Opt_lazytime, "lazytime"},
1181 {Opt_nolazytime, "nolazytime"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001182 {Opt_nodelalloc, "nodelalloc"},
Jan Kara36ade452013-01-28 09:30:52 -05001183 {Opt_removed, "mblk_io_submit"},
1184 {Opt_removed, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001185 {Opt_block_validity, "block_validity"},
1186 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001187 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001188 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001189 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001190 {Opt_auto_da_alloc, "auto_da_alloc"},
1191 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001192 {Opt_dioread_nolock, "dioread_nolock"},
1193 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001194 {Opt_discard, "discard"},
1195 {Opt_nodiscard, "nodiscard"},
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001196 {Opt_init_itable, "init_itable=%u"},
1197 {Opt_init_itable, "init_itable"},
1198 {Opt_noinit_itable, "noinit_itable"},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001199 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001200 {Opt_test_dummy_encryption, "test_dummy_encryption"},
Theodore Ts'oc7198b92012-03-04 22:00:53 -05001201 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1202 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1203 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1204 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1205 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001206 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001207};
1208
Mingming Cao617ba132006-10-11 01:20:53 -07001209static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001210{
Mingming Cao617ba132006-10-11 01:20:53 -07001211 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001212 char *options = (char *) *data;
1213
1214 if (!options || strncmp(options, "sb=", 3) != 0)
1215 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001216
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001217 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001218 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001219 sb_block = simple_strtoul(options, &options, 0);
1220 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001221 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001222 (char *) *data);
1223 return 1;
1224 }
1225 if (*options == ',')
1226 options++;
1227 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001228
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001229 return sb_block;
1230}
1231
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001232#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001233static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1234 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001235
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001236#ifdef CONFIG_QUOTA
1237static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1238{
1239 struct ext4_sb_info *sbi = EXT4_SB(sb);
1240 char *qname;
Chen Gang03dafb52013-01-24 23:24:58 -05001241 int ret = -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001242
1243 if (sb_any_quota_loaded(sb) &&
1244 !sbi->s_qf_names[qtype]) {
1245 ext4_msg(sb, KERN_ERR,
1246 "Cannot change journaled "
1247 "quota options when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001248 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001249 }
Jan Kara262b4662013-03-02 17:57:08 -05001250 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1251 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1252 "when QUOTA feature is enabled");
1253 return -1;
1254 }
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001255 qname = match_strdup(args);
1256 if (!qname) {
1257 ext4_msg(sb, KERN_ERR,
1258 "Not enough memory for storing quotafile name");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001259 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001260 }
Chen Gang03dafb52013-01-24 23:24:58 -05001261 if (sbi->s_qf_names[qtype]) {
1262 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1263 ret = 1;
1264 else
1265 ext4_msg(sb, KERN_ERR,
1266 "%s quota file already specified",
1267 QTYPE2NAME(qtype));
1268 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001269 }
Chen Gang03dafb52013-01-24 23:24:58 -05001270 if (strchr(qname, '/')) {
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001271 ext4_msg(sb, KERN_ERR,
1272 "quotafile must be on filesystem root");
Chen Gang03dafb52013-01-24 23:24:58 -05001273 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001274 }
Chen Gang03dafb52013-01-24 23:24:58 -05001275 sbi->s_qf_names[qtype] = qname;
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001276 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001277 return 1;
Chen Gang03dafb52013-01-24 23:24:58 -05001278errout:
1279 kfree(qname);
1280 return ret;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001281}
1282
1283static int clear_qf_name(struct super_block *sb, int qtype)
1284{
1285
1286 struct ext4_sb_info *sbi = EXT4_SB(sb);
1287
1288 if (sb_any_quota_loaded(sb) &&
1289 sbi->s_qf_names[qtype]) {
1290 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1291 " when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001292 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001293 }
Chen Gang03dafb52013-01-24 23:24:58 -05001294 kfree(sbi->s_qf_names[qtype]);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001295 sbi->s_qf_names[qtype] = NULL;
1296 return 1;
1297}
1298#endif
1299
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001300#define MOPT_SET 0x0001
1301#define MOPT_CLEAR 0x0002
1302#define MOPT_NOSUPPORT 0x0004
1303#define MOPT_EXPLICIT 0x0008
1304#define MOPT_CLEAR_ERR 0x0010
1305#define MOPT_GTE0 0x0020
1306#ifdef CONFIG_QUOTA
1307#define MOPT_Q 0
1308#define MOPT_QFMT 0x0040
1309#else
1310#define MOPT_Q MOPT_NOSUPPORT
1311#define MOPT_QFMT MOPT_NOSUPPORT
1312#endif
1313#define MOPT_DATAJ 0x0080
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001314#define MOPT_NO_EXT2 0x0100
1315#define MOPT_NO_EXT3 0x0200
1316#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
Eric Sandeenad4eec62013-08-28 19:05:07 -04001317#define MOPT_STRING 0x0400
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001318
1319static const struct mount_opts {
1320 int token;
1321 int mount_opt;
1322 int flags;
1323} ext4_mount_opts[] = {
1324 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1325 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1326 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1327 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001328 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1329 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001330 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1331 MOPT_EXT4_ONLY | MOPT_SET},
1332 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1333 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001334 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1335 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001336 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1337 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1338 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
Theodore Ts'o59d9fa52013-08-08 23:01:24 -04001339 MOPT_EXT4_ONLY | MOPT_CLEAR},
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001340 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1341 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001342 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1343 MOPT_EXT4_ONLY | MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001344 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001345 EXT4_MOUNT_JOURNAL_CHECKSUM),
1346 MOPT_EXT4_ONLY | MOPT_SET},
1347 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001348 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1349 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1350 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001351 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1352 MOPT_NO_EXT2 | MOPT_SET},
1353 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1354 MOPT_NO_EXT2 | MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001355 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1356 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1357 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1358 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1359 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1360 {Opt_commit, 0, MOPT_GTE0},
1361 {Opt_max_batch_time, 0, MOPT_GTE0},
1362 {Opt_min_batch_time, 0, MOPT_GTE0},
1363 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1364 {Opt_init_itable, 0, MOPT_GTE0},
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001365 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001366 {Opt_stripe, 0, MOPT_GTE0},
Jan Kara0efb3b22013-02-02 22:52:19 -05001367 {Opt_resuid, 0, MOPT_GTE0},
1368 {Opt_resgid, 0, MOPT_GTE0},
1369 {Opt_journal_dev, 0, MOPT_GTE0},
Eric Sandeenad4eec62013-08-28 19:05:07 -04001370 {Opt_journal_path, 0, MOPT_STRING},
Jan Kara0efb3b22013-02-02 22:52:19 -05001371 {Opt_journal_ioprio, 0, MOPT_GTE0},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001372 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1373 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1375 MOPT_NO_EXT2 | MOPT_DATAJ},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001376 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1377 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001378#ifdef CONFIG_EXT4_FS_POSIX_ACL
1379 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1380 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1381#else
1382 {Opt_acl, 0, MOPT_NOSUPPORT},
1383 {Opt_noacl, 0, MOPT_NOSUPPORT},
1384#endif
1385 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1386 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1387 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1388 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1389 MOPT_SET | MOPT_Q},
1390 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1391 MOPT_SET | MOPT_Q},
1392 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1393 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1394 {Opt_usrjquota, 0, MOPT_Q},
1395 {Opt_grpjquota, 0, MOPT_Q},
1396 {Opt_offusrjquota, 0, MOPT_Q},
1397 {Opt_offgrpjquota, 0, MOPT_Q},
1398 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1399 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1400 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001401 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001402 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001403 {Opt_err, 0, 0}
1404};
1405
1406static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1407 substring_t *args, unsigned long *journal_devnum,
1408 unsigned int *journal_ioprio, int is_remount)
1409{
1410 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 const struct mount_opts *m;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001412 kuid_t uid;
1413 kgid_t gid;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001414 int arg = 0;
1415
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001416#ifdef CONFIG_QUOTA
1417 if (token == Opt_usrjquota)
1418 return set_qf_name(sb, USRQUOTA, &args[0]);
1419 else if (token == Opt_grpjquota)
1420 return set_qf_name(sb, GRPQUOTA, &args[0]);
1421 else if (token == Opt_offusrjquota)
1422 return clear_qf_name(sb, USRQUOTA);
1423 else if (token == Opt_offgrpjquota)
1424 return clear_qf_name(sb, GRPQUOTA);
1425#endif
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001426 switch (token) {
Theodore Ts'of7048602012-03-04 22:06:20 -05001427 case Opt_noacl:
1428 case Opt_nouser_xattr:
1429 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1430 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001431 case Opt_sb:
1432 return 1; /* handled by get_sb_block() */
1433 case Opt_removed:
Jan Kara5f3633e2013-02-02 23:09:36 -05001434 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001435 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001436 case Opt_abort:
1437 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1438 return 1;
1439 case Opt_i_version:
1440 sb->s_flags |= MS_I_VERSION;
1441 return 1;
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001442 case Opt_lazytime:
1443 sb->s_flags |= MS_LAZYTIME;
1444 return 1;
1445 case Opt_nolazytime:
1446 sb->s_flags &= ~MS_LAZYTIME;
1447 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001448 }
1449
Jan Kara5f3633e2013-02-02 23:09:36 -05001450 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1451 if (token == m->token)
1452 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001453
Jan Kara5f3633e2013-02-02 23:09:36 -05001454 if (m->token == Opt_err) {
1455 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1456 "or missing value", opt);
1457 return -1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001458 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001459
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001460 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1461 ext4_msg(sb, KERN_ERR,
1462 "Mount option \"%s\" incompatible with ext2", opt);
1463 return -1;
1464 }
1465 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1466 ext4_msg(sb, KERN_ERR,
1467 "Mount option \"%s\" incompatible with ext3", opt);
1468 return -1;
1469 }
1470
Eric Sandeenad4eec62013-08-28 19:05:07 -04001471 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
Jan Kara5f3633e2013-02-02 23:09:36 -05001472 return -1;
1473 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1474 return -1;
1475 if (m->flags & MOPT_EXPLICIT)
1476 set_opt2(sb, EXPLICIT_DELALLOC);
1477 if (m->flags & MOPT_CLEAR_ERR)
1478 clear_opt(sb, ERRORS_MASK);
1479 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1480 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1481 "options when quota turned on");
1482 return -1;
1483 }
1484
1485 if (m->flags & MOPT_NOSUPPORT) {
1486 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1487 } else if (token == Opt_commit) {
1488 if (arg == 0)
1489 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1490 sbi->s_commit_interval = HZ * arg;
1491 } else if (token == Opt_max_batch_time) {
Jan Kara5f3633e2013-02-02 23:09:36 -05001492 sbi->s_max_batch_time = arg;
1493 } else if (token == Opt_min_batch_time) {
1494 sbi->s_min_batch_time = arg;
1495 } else if (token == Opt_inode_readahead_blks) {
Jan Karae33e60e2013-02-02 23:14:31 -05001496 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1497 ext4_msg(sb, KERN_ERR,
1498 "EXT4-fs: inode_readahead_blks must be "
1499 "0 or a power of 2 smaller than 2^31");
Jan Kara5f3633e2013-02-02 23:09:36 -05001500 return -1;
1501 }
1502 sbi->s_inode_readahead_blks = arg;
1503 } else if (token == Opt_init_itable) {
1504 set_opt(sb, INIT_INODE_TABLE);
1505 if (!args->from)
1506 arg = EXT4_DEF_LI_WAIT_MULT;
1507 sbi->s_li_wait_mult = arg;
1508 } else if (token == Opt_max_dir_size_kb) {
1509 sbi->s_max_dir_size_kb = arg;
1510 } else if (token == Opt_stripe) {
1511 sbi->s_stripe = arg;
1512 } else if (token == Opt_resuid) {
1513 uid = make_kuid(current_user_ns(), arg);
1514 if (!uid_valid(uid)) {
1515 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1516 return -1;
1517 }
1518 sbi->s_resuid = uid;
1519 } else if (token == Opt_resgid) {
1520 gid = make_kgid(current_user_ns(), arg);
1521 if (!gid_valid(gid)) {
1522 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1523 return -1;
1524 }
1525 sbi->s_resgid = gid;
1526 } else if (token == Opt_journal_dev) {
1527 if (is_remount) {
1528 ext4_msg(sb, KERN_ERR,
1529 "Cannot specify journal on remount");
1530 return -1;
1531 }
1532 *journal_devnum = arg;
Eric Sandeenad4eec62013-08-28 19:05:07 -04001533 } else if (token == Opt_journal_path) {
1534 char *journal_path;
1535 struct inode *journal_inode;
1536 struct path path;
1537 int error;
1538
1539 if (is_remount) {
1540 ext4_msg(sb, KERN_ERR,
1541 "Cannot specify journal on remount");
1542 return -1;
1543 }
1544 journal_path = match_strdup(&args[0]);
1545 if (!journal_path) {
1546 ext4_msg(sb, KERN_ERR, "error: could not dup "
1547 "journal device string");
1548 return -1;
1549 }
1550
1551 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1552 if (error) {
1553 ext4_msg(sb, KERN_ERR, "error: could not find "
1554 "journal device path: error %d", error);
1555 kfree(journal_path);
1556 return -1;
1557 }
1558
David Howells2b0143b2015-03-17 22:25:59 +00001559 journal_inode = d_inode(path.dentry);
Eric Sandeenad4eec62013-08-28 19:05:07 -04001560 if (!S_ISBLK(journal_inode->i_mode)) {
1561 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1562 "is not a block device", journal_path);
1563 path_put(&path);
1564 kfree(journal_path);
1565 return -1;
1566 }
1567
1568 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1569 path_put(&path);
1570 kfree(journal_path);
Jan Kara5f3633e2013-02-02 23:09:36 -05001571 } else if (token == Opt_journal_ioprio) {
1572 if (arg > 7) {
1573 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1574 " (must be 0-7)");
1575 return -1;
1576 }
1577 *journal_ioprio =
1578 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001579 } else if (token == Opt_test_dummy_encryption) {
1580#ifdef CONFIG_EXT4_FS_ENCRYPTION
1581 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1582 ext4_msg(sb, KERN_WARNING,
1583 "Test dummy encryption mode enabled");
1584#else
1585 ext4_msg(sb, KERN_WARNING,
1586 "Test dummy encryption mount option ignored");
1587#endif
Jan Kara5f3633e2013-02-02 23:09:36 -05001588 } else if (m->flags & MOPT_DATAJ) {
1589 if (is_remount) {
1590 if (!sbi->s_journal)
1591 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1592 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1593 ext4_msg(sb, KERN_ERR,
1594 "Cannot change data mode on remount");
1595 return -1;
1596 }
1597 } else {
1598 clear_opt(sb, DATA_FLAGS);
1599 sbi->s_mount_opt |= m->mount_opt;
1600 }
1601#ifdef CONFIG_QUOTA
1602 } else if (m->flags & MOPT_QFMT) {
1603 if (sb_any_quota_loaded(sb) &&
1604 sbi->s_jquota_fmt != m->mount_opt) {
1605 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1606 "quota options when quota turned on");
1607 return -1;
1608 }
Jan Kara262b4662013-03-02 17:57:08 -05001609 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1610 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot set journaled quota options "
1613 "when QUOTA feature is enabled");
1614 return -1;
1615 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001616 sbi->s_jquota_fmt = m->mount_opt;
1617#endif
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001618#ifndef CONFIG_FS_DAX
1619 } else if (token == Opt_dax) {
1620 ext4_msg(sb, KERN_INFO, "dax option not supported");
1621 return -1;
1622#endif
Jan Kara5f3633e2013-02-02 23:09:36 -05001623 } else {
1624 if (!args->from)
1625 arg = 1;
1626 if (m->flags & MOPT_CLEAR)
1627 arg = !arg;
1628 else if (unlikely(!(m->flags & MOPT_SET))) {
1629 ext4_msg(sb, KERN_WARNING,
1630 "buggy handling of option %s", opt);
1631 WARN_ON(1);
1632 return -1;
1633 }
1634 if (arg != 0)
1635 sbi->s_mount_opt |= m->mount_opt;
1636 else
1637 sbi->s_mount_opt &= ~m->mount_opt;
1638 }
1639 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001640}
1641
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001642static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001643 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001644 unsigned int *journal_ioprio,
Eric Sandeen661aa522012-02-20 17:53:04 -05001645 int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001646{
Mingming Cao617ba132006-10-11 01:20:53 -07001647 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001648 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001649 substring_t args[MAX_OPT_ARGS];
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001650 int token;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001651
1652 if (!options)
1653 return 1;
1654
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001655 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001656 if (!*p)
1657 continue;
Eric Sandeen15121c12010-02-15 20:17:55 -05001658 /*
1659 * Initialize args struct so we know whether arg was
1660 * found; some options take optional arguments.
1661 */
Sachin Kamatcaecd0a2012-08-18 22:29:18 -04001662 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001663 token = match_token(p, tokens, args);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001664 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 journal_ioprio, is_remount) < 0)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001666 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001667 }
1668#ifdef CONFIG_QUOTA
Jan Kara262b4662013-03-02 17:57:08 -05001669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1670 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1671 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1672 "feature is enabled");
1673 return 0;
1674 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001675 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001676 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001677 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001678
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001679 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001680 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001681
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001682 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001683 ext4_msg(sb, KERN_ERR, "old and new quota "
1684 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001685 return 0;
1686 }
1687
1688 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001689 ext4_msg(sb, KERN_ERR, "journaled quota format "
1690 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001691 return 0;
1692 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001693 }
1694#endif
Jan Kara261cb20c2012-12-20 00:07:18 -05001695 if (test_opt(sb, DIOREAD_NOLOCK)) {
1696 int blocksize =
1697 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1698
1699 if (blocksize < PAGE_CACHE_SIZE) {
1700 ext4_msg(sb, KERN_ERR, "can't mount with "
1701 "dioread_nolock if block size != PAGE_SIZE");
1702 return 0;
1703 }
1704 }
Jan Karad4f76102014-11-25 20:19:17 -05001705 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1706 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1707 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1708 "in data=ordered mode");
1709 return 0;
1710 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001711 return 1;
1712}
1713
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001714static inline void ext4_show_quota_options(struct seq_file *seq,
1715 struct super_block *sb)
1716{
1717#if defined(CONFIG_QUOTA)
1718 struct ext4_sb_info *sbi = EXT4_SB(sb);
1719
1720 if (sbi->s_jquota_fmt) {
1721 char *fmtname = "";
1722
1723 switch (sbi->s_jquota_fmt) {
1724 case QFMT_VFS_OLD:
1725 fmtname = "vfsold";
1726 break;
1727 case QFMT_VFS_V0:
1728 fmtname = "vfsv0";
1729 break;
1730 case QFMT_VFS_V1:
1731 fmtname = "vfsv1";
1732 break;
1733 }
1734 seq_printf(seq, ",jqfmt=%s", fmtname);
1735 }
1736
1737 if (sbi->s_qf_names[USRQUOTA])
1738 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739
1740 if (sbi->s_qf_names[GRPQUOTA])
1741 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001742#endif
1743}
1744
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001745static const char *token2str(int token)
1746{
Herton Ronaldo Krzesinski50df9fd2012-09-23 22:49:12 -04001747 const struct match_token *t;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001748
1749 for (t = tokens; t->token != Opt_err; t++)
1750 if (t->token == token && !strchr(t->pattern, '='))
1751 break;
1752 return t->pattern;
1753}
1754
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001755/*
1756 * Show an option if
1757 * - it's set to a non-default value OR
1758 * - if the per-sb default is different from the global default
1759 */
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001760static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1761 int nodefs)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001762{
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001763 struct ext4_sb_info *sbi = EXT4_SB(sb);
1764 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001765 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001766 const struct mount_opts *m;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001767 char sep = nodefs ? '\n' : ',';
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001768
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001769#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1770#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001771
1772 if (sbi->s_sb_block != 1)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001773 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1774
1775 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1776 int want_set = m->flags & MOPT_SET;
1777 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1778 (m->flags & MOPT_CLEAR_ERR))
1779 continue;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001780 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001781 continue; /* skip if same as the default */
1782 if ((want_set &&
1783 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1784 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1785 continue; /* select Opt_noFoo vs Opt_Foo */
1786 SEQ_OPTS_PRINT("%s", token2str(m->token));
1787 }
1788
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001789 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001790 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001791 SEQ_OPTS_PRINT("resuid=%u",
1792 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1793 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001794 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001795 SEQ_OPTS_PRINT("resgid=%u",
1796 from_kgid_munged(&init_user_ns, sbi->s_resgid));
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001797 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001798 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1799 SEQ_OPTS_PUTS("errors=remount-ro");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001800 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001801 SEQ_OPTS_PUTS("errors=continue");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001802 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001803 SEQ_OPTS_PUTS("errors=panic");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001804 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001805 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001806 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001807 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001808 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001809 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001810 if (sb->s_flags & MS_I_VERSION)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001811 SEQ_OPTS_PUTS("i_version");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001812 if (nodefs || sbi->s_stripe)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001813 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001814 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001815 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1816 SEQ_OPTS_PUTS("data=journal");
1817 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1818 SEQ_OPTS_PUTS("data=ordered");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1820 SEQ_OPTS_PUTS("data=writeback");
1821 }
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001822 if (nodefs ||
1823 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001824 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1825 sbi->s_inode_readahead_blks);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001826
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001827 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1828 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001829 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
Theodore Ts'odf981d02012-08-17 09:48:17 -04001830 if (nodefs || sbi->s_max_dir_size_kb)
1831 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001832
1833 ext4_show_quota_options(seq, sb);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001834 return 0;
1835}
1836
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001837static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1838{
1839 return _ext4_show_options(seq, root->d_sb, 0);
1840}
1841
1842static int options_seq_show(struct seq_file *seq, void *offset)
1843{
1844 struct super_block *sb = seq->private;
1845 int rc;
1846
1847 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1848 rc = _ext4_show_options(seq, sb, 1);
1849 seq_puts(seq, "\n");
1850 return rc;
1851}
1852
1853static int options_open_fs(struct inode *inode, struct file *file)
1854{
Al Virod9dda782013-03-31 18:16:14 -04001855 return single_open(file, options_seq_show, PDE_DATA(inode));
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001856}
1857
1858static const struct file_operations ext4_seq_options_fops = {
1859 .owner = THIS_MODULE,
1860 .open = options_open_fs,
1861 .read = seq_read,
1862 .llseek = seq_lseek,
1863 .release = single_release,
1864};
1865
Mingming Cao617ba132006-10-11 01:20:53 -07001866static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001867 int read_only)
1868{
Mingming Cao617ba132006-10-11 01:20:53 -07001869 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001870 int res = 0;
1871
Mingming Cao617ba132006-10-11 01:20:53 -07001872 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001873 ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001875 res = MS_RDONLY;
1876 }
1877 if (read_only)
Theodore Ts'o281b5992011-09-09 18:34:51 -04001878 goto done;
Mingming Cao617ba132006-10-11 01:20:53 -07001879 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001880 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 "running e2fsck is recommended");
Lukas Czernerc8b459f2014-05-12 12:55:07 -04001882 else if (sbi->s_mount_state & EXT4_ERROR_FS)
Eric Sandeenb31e1552009-06-04 17:36:36 -04001883 ext4_msg(sb, KERN_WARNING,
1884 "warning: mounting fs with errors, "
1885 "running e2fsck is recommended");
Tao Maed3ce802011-05-18 13:29:57 -04001886 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001887 le16_to_cpu(es->s_mnt_count) >=
1888 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001889 ext4_msg(sb, KERN_WARNING,
1890 "warning: maximal mount count reached, "
1891 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001892 else if (le32_to_cpu(es->s_checkinterval) &&
1893 (le32_to_cpu(es->s_lastcheck) +
1894 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04001895 ext4_msg(sb, KERN_WARNING,
1896 "warning: checktime reached, "
1897 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001898 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05001899 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001900 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07001901 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04001902 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001903 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07001904 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05001905 if (sbi->s_journal)
1906 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001907
Theodore Ts'oe2d67052009-05-01 00:33:44 -04001908 ext4_commit_super(sb, 1);
Theodore Ts'o281b5992011-09-09 18:34:51 -04001909done:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001910 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05001911 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001913 sb->s_blocksize,
1914 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07001915 EXT4_BLOCKS_PER_GROUP(sb),
1916 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05001917 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001918
Dan Magenheimer7abc52c2011-05-26 10:02:03 -06001919 cleancache_init_fs(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001920 return res;
1921}
1922
Theodore Ts'o117fff12012-09-05 01:29:50 -04001923int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1924{
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct flex_groups *new_groups;
1927 int size;
1928
1929 if (!sbi->s_log_groups_per_flex)
1930 return 0;
1931
1932 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1933 if (size <= sbi->s_flex_groups_allocated)
1934 return 0;
1935
1936 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1937 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1938 if (!new_groups) {
1939 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1940 size / (int) sizeof(struct flex_groups));
1941 return -ENOMEM;
1942 }
1943
1944 if (sbi->s_flex_groups) {
1945 memcpy(new_groups, sbi->s_flex_groups,
1946 (sbi->s_flex_groups_allocated *
1947 sizeof(struct flex_groups)));
Al Virob93b41d2014-11-20 12:19:11 -05001948 kvfree(sbi->s_flex_groups);
Theodore Ts'o117fff12012-09-05 01:29:50 -04001949 }
1950 sbi->s_flex_groups = new_groups;
1951 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1952 return 0;
1953}
1954
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001955static int ext4_fill_flex_info(struct super_block *sb)
1956{
1957 struct ext4_sb_info *sbi = EXT4_SB(sb);
1958 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001959 ext4_group_t flex_group;
Theodore Ts'o117fff12012-09-05 01:29:50 -04001960 int i, err;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001961
Theodore Ts'o503358a2009-11-23 07:24:46 -05001962 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
Xi Wangd50f2ab2012-01-10 11:51:10 -05001963 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001964 sbi->s_log_groups_per_flex = 0;
1965 return 1;
1966 }
1967
Theodore Ts'o117fff12012-09-05 01:29:50 -04001968 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1969 if (err)
Theodore Ts'o9933fc02011-08-01 08:45:02 -04001970 goto failed;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001971
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001972 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04001973 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001974
1975 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001976 atomic_add(ext4_free_inodes_count(sb, gdp),
1977 &sbi->s_flex_groups[flex_group].free_inodes);
Theodore Ts'o90ba9832013-03-11 23:39:59 -04001978 atomic64_add(ext4_free_group_clusters(sb, gdp),
1979 &sbi->s_flex_groups[flex_group].free_clusters);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04001980 atomic_add(ext4_used_dirs_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04001982 }
1983
1984 return 1;
1985failed:
1986 return 0;
1987}
1988
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001989static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1990 struct ext4_group_desc *gdp)
Andreas Dilger717d50e2007-10-16 18:38:25 -04001991{
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001992 int offset;
Andreas Dilger717d50e2007-10-16 18:38:25 -04001993 __u16 crc = 0;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001994 __le32 le_group = cpu_to_le32(block_group);
Andreas Dilger717d50e2007-10-16 18:38:25 -04001995
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -04001996 if (ext4_has_metadata_csum(sbi->s_sb)) {
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001997 /* Use new metadata_csum algorithm */
Theodore Ts'od6a77102013-04-09 23:59:55 -04001998 __le16 save_csum;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04001999 __u32 csum32;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002000
Theodore Ts'od6a77102013-04-09 23:59:55 -04002001 save_csum = gdp->bg_checksum;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002002 gdp->bg_checksum = 0;
2003 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2004 sizeof(le_group));
2005 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2006 sbi->s_desc_size);
Theodore Ts'od6a77102013-04-09 23:59:55 -04002007 gdp->bg_checksum = save_csum;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002008
2009 crc = csum32 & 0xFFFF;
2010 goto out;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002011 }
2012
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002013 /* old crc16 code */
Darrick J. Wong813d32f2014-10-14 02:35:49 -04002014 if (!(sbi->s_es->s_feature_ro_compat &
2015 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2016 return 0;
2017
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002018 offset = offsetof(struct ext4_group_desc, bg_checksum);
2019
2020 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2021 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2022 crc = crc16(crc, (__u8 *)gdp, offset);
2023 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2024 /* for checksum of struct ext4_group_desc do the rest...*/
2025 if ((sbi->s_es->s_feature_incompat &
2026 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2027 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2028 crc = crc16(crc, (__u8 *)gdp + offset,
2029 le16_to_cpu(sbi->s_es->s_desc_size) -
2030 offset);
2031
2032out:
Andreas Dilger717d50e2007-10-16 18:38:25 -04002033 return cpu_to_le16(crc);
2034}
2035
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002036int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
Andreas Dilger717d50e2007-10-16 18:38:25 -04002037 struct ext4_group_desc *gdp)
2038{
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002039 if (ext4_has_group_desc_csum(sb) &&
2040 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2041 block_group, gdp)))
Andreas Dilger717d50e2007-10-16 18:38:25 -04002042 return 0;
2043
2044 return 1;
2045}
2046
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002047void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2048 struct ext4_group_desc *gdp)
2049{
2050 if (!ext4_has_group_desc_csum(sb))
2051 return;
2052 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2053}
2054
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002055/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002056static int ext4_check_descriptors(struct super_block *sb,
2057 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002058{
Mingming Cao617ba132006-10-11 01:20:53 -07002059 struct ext4_sb_info *sbi = EXT4_SB(sb);
2060 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2061 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002062 ext4_fsblk_t block_bitmap;
2063 ext4_fsblk_t inode_bitmap;
2064 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002065 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002066 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002067
Jose R. Santosce421582007-10-16 18:38:25 -04002068 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2069 flexbg_flag = 1;
2070
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002071 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002072
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002073 for (i = 0; i < sbi->s_groups_count; i++) {
2074 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2075
Jose R. Santosce421582007-10-16 18:38:25 -04002076 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002077 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002078 else
2079 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002080 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002081
Lukas Czernerbfff6872010-10-27 21:30:05 -04002082 if ((grp == sbi->s_groups_count) &&
2083 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2084 grp = i;
2085
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002086 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002087 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002089 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002090 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002091 return 0;
2092 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002093 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002094 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002096 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002097 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002098 return 0;
2099 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002100 inode_table = ext4_inode_table(sb, gdp);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002101 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002102 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002103 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002104 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002105 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002106 return 0;
2107 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002108 ext4_lock_group(sb, i);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002109 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 "Checksum for group %u failed (%u!=%u)",
2112 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2113 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002114 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002115 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002116 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002117 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002118 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002119 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002120 if (!flexbg_flag)
2121 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002122 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002123 if (NULL != first_not_zeroed)
2124 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002125 return 1;
2126}
2127
Mingming Cao617ba132006-10-11 01:20:53 -07002128/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002129 * the superblock) which were deleted from all directories, but held open by
2130 * a process at the time of a crash. We walk the list and try to delete these
2131 * inodes at recovery time (only with a read-write filesystem).
2132 *
2133 * In order to keep the orphan inode chain consistent during traversal (in
2134 * case of crash during recovery), we link each inode into the superblock
2135 * orphan list_head and handle it the same way as an inode deletion during
2136 * normal operation (which journals the operations for us).
2137 *
2138 * We only do an iget() and an iput() on each inode, which is very safe if we
2139 * accidentally point at an in-use or already deleted inode. The worst that
2140 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002141 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002142 * e2fsck was run on this filesystem, and it must have already done the orphan
2143 * inode cleanup for us, so we can safely abort without any further action.
2144 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002145static void ext4_orphan_cleanup(struct super_block *sb,
2146 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002147{
2148 unsigned int s_flags = sb->s_flags;
2149 int nr_orphans = 0, nr_truncates = 0;
2150#ifdef CONFIG_QUOTA
2151 int i;
2152#endif
2153 if (!es->s_last_orphan) {
2154 jbd_debug(4, "no orphan inodes to clean up\n");
2155 return;
2156 }
2157
Eric Sandeena8f48a92006-12-06 20:40:13 -08002158 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002159 ext4_msg(sb, KERN_ERR, "write access "
2160 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002161 return;
2162 }
2163
Amir Goldsteind39195c2011-02-28 00:53:45 -05002164 /* Check if feature set would not allow a r/w mount */
2165 if (!ext4_feature_set_ok(sb, 0)) {
2166 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2167 "unknown ROCOMPAT features");
2168 return;
2169 }
2170
Mingming Cao617ba132006-10-11 01:20:53 -07002171 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002172 /* don't clear list on RO mount w/ errors */
2173 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
Dmitry Monakhov84474972014-09-16 14:52:03 -04002174 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002175 "clearing orphan list.\n");
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002176 es->s_last_orphan = 0;
2177 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002178 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2179 return;
2180 }
2181
2182 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002183 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002184 sb->s_flags &= ~MS_RDONLY;
2185 }
2186#ifdef CONFIG_QUOTA
2187 /* Needed for iput() to work correctly and not trash data */
2188 sb->s_flags |= MS_ACTIVE;
2189 /* Turn on quotas so that they are updated correctly */
Jan Karaa2d4a642014-09-11 11:15:15 -04002190 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002191 if (EXT4_SB(sb)->s_qf_names[i]) {
2192 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002193 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002194 ext4_msg(sb, KERN_ERR,
2195 "Cannot turn on journaled "
2196 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002197 }
2198 }
2199#endif
2200
2201 while (es->s_last_orphan) {
2202 struct inode *inode;
2203
Josef Bacik97bd42b2008-04-29 22:04:56 -04002204 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2205 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002206 es->s_last_orphan = 0;
2207 break;
2208 }
2209
Mingming Cao617ba132006-10-11 01:20:53 -07002210 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002211 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002212 if (inode->i_nlink) {
Paul Taysom566370a2013-05-28 07:51:21 -04002213 if (test_opt(sb, DEBUG))
2214 ext4_msg(sb, KERN_DEBUG,
2215 "%s: truncating inode %lu to %lld bytes",
2216 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002217 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002218 inode->i_ino, inode->i_size);
Theodore Ts'o721e3eb2012-12-27 01:42:48 -05002219 mutex_lock(&inode->i_mutex);
Lukas Czerner55f252c2013-05-27 23:32:35 -04002220 truncate_inode_pages(inode->i_mapping, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002221 ext4_truncate(inode);
Theodore Ts'o721e3eb2012-12-27 01:42:48 -05002222 mutex_unlock(&inode->i_mutex);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002223 nr_truncates++;
2224 } else {
Paul Taysom566370a2013-05-28 07:51:21 -04002225 if (test_opt(sb, DEBUG))
2226 ext4_msg(sb, KERN_DEBUG,
2227 "%s: deleting unreferenced inode %lu",
2228 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002229 jbd_debug(2, "deleting unreferenced inode %lu\n",
2230 inode->i_ino);
2231 nr_orphans++;
2232 }
2233 iput(inode); /* The delete magic happens here! */
2234 }
2235
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002236#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002237
2238 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002239 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2240 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002241 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002242 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2243 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002244#ifdef CONFIG_QUOTA
2245 /* Turn quotas off */
Jan Karaa2d4a642014-09-11 11:15:15 -04002246 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002247 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002248 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002249 }
2250#endif
2251 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2252}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002253
Eric Sandeencd2291a2008-01-28 23:58:27 -05002254/*
2255 * Maximal extent format file size.
2256 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2257 * extent format containers, within a sector_t, and within i_blocks
2258 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2259 * so that won't be a limiting factor.
2260 *
Lukas Czernerf17722f2011-06-06 00:05:17 -04002261 * However there is other limiting factor. We do store extents in the form
2262 * of starting block and length, hence the resulting length of the extent
2263 * covering maximum file size must fit into on-disk format containers as
2264 * well. Given that length is always by 1 unit bigger than max unit (because
2265 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2266 *
Eric Sandeencd2291a2008-01-28 23:58:27 -05002267 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2268 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002269static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002270{
2271 loff_t res;
2272 loff_t upper_limit = MAX_LFS_FILESIZE;
2273
2274 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002275 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002276 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002277 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002278 * i_block represent total blocks in 512 bytes
2279 * 32 == size of vfs inode i_blocks * 8
2280 */
2281 upper_limit = (1LL << 32) - 1;
2282
2283 /* total blocks in file system block size */
2284 upper_limit >>= (blkbits - 9);
2285 upper_limit <<= blkbits;
2286 }
2287
Lukas Czernerf17722f2011-06-06 00:05:17 -04002288 /*
2289 * 32-bit extent-start container, ee_block. We lower the maxbytes
2290 * by one fs block, so ee_len can cover the extent of maximum file
2291 * size
2292 */
2293 res = (1LL << 32) - 1;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002294 res <<= blkbits;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002295
2296 /* Sanity check against vm- & vfs- imposed limits */
2297 if (res > upper_limit)
2298 res = upper_limit;
2299
2300 return res;
2301}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002302
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002303/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002304 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002305 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2306 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002307 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002308static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002309{
Mingming Cao617ba132006-10-11 01:20:53 -07002310 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002311 int meta_blocks;
2312 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002313 /* This is calculated to be the largest file size for a dense, block
2314 * mapped file such that the file's total number of 512-byte sectors,
2315 * including data and all indirect blocks, does not exceed (2^48 - 1).
2316 *
2317 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2318 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002319 */
2320
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002321 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002322 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002323 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002324 * the inode i_block field represents total file blocks in
2325 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002326 */
2327 upper_limit = (1LL << 32) - 1;
2328
2329 /* total blocks in file system block size */
2330 upper_limit >>= (bits - 9);
2331
2332 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002333 /*
2334 * We use 48 bit ext4_inode i_blocks
2335 * With EXT4_HUGE_FILE_FL set the i_blocks
2336 * represent total number of blocks in
2337 * file system block size
2338 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002339 upper_limit = (1LL << 48) - 1;
2340
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002341 }
2342
2343 /* indirect blocks */
2344 meta_blocks = 1;
2345 /* double indirect blocks */
2346 meta_blocks += 1 + (1LL << (bits-2));
2347 /* tripple indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2349
2350 upper_limit -= meta_blocks;
2351 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002352
2353 res += 1LL << (bits-2);
2354 res += 1LL << (2*(bits-2));
2355 res += 1LL << (3*(bits-2));
2356 res <<= bits;
2357 if (res > upper_limit)
2358 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002359
2360 if (res > MAX_LFS_FILESIZE)
2361 res = MAX_LFS_FILESIZE;
2362
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002363 return res;
2364}
2365
Mingming Cao617ba132006-10-11 01:20:53 -07002366static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002367 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002368{
Mingming Cao617ba132006-10-11 01:20:53 -07002369 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002370 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002371 int has_super = 0;
2372
2373 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2374
Mingming Cao617ba132006-10-11 01:20:53 -07002375 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002376 nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002377 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002378 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002379 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002380 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002381
Darrick J. Wongbd63f6b2014-05-12 10:06:27 -04002382 /*
2383 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2384 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2385 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2386 * compensate.
2387 */
2388 if (sb->s_blocksize == 1024 && nr == 0 &&
2389 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2390 has_super++;
2391
Mingming Cao617ba132006-10-11 01:20:53 -07002392 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002393}
2394
Alex Tomasc9de5602008-01-29 00:19:52 -05002395/**
2396 * ext4_get_stripe_size: Get the stripe size.
2397 * @sbi: In memory super block info
2398 *
2399 * If we have specified it via mount option, then
2400 * use the mount option value. If the value specified at mount time is
2401 * greater than the blocks per group use the super block value.
2402 * If the super block value is greater than blocks per group return 0.
2403 * Allocator needs it be less than blocks per group.
2404 *
2405 */
2406static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2407{
2408 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2409 unsigned long stripe_width =
2410 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002411 int ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002412
2413 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002414 ret = sbi->s_stripe;
2415 else if (stripe_width <= sbi->s_blocks_per_group)
2416 ret = stripe_width;
2417 else if (stride <= sbi->s_blocks_per_group)
2418 ret = stride;
2419 else
2420 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002421
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002422 /*
2423 * If the stripe width is 1, this makes no sense and
2424 * we set it to 0 to turn off stripe handling code.
2425 */
2426 if (ret <= 1)
2427 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002428
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002429 return ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002430}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002431
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002432/* sysfs supprt */
2433
2434struct ext4_attr {
2435 struct attribute attr;
2436 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
Theodore Ts'o60e66792010-05-17 07:00:00 -04002437 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002438 const char *, size_t);
Jan Karaf2d50a62013-06-04 12:51:16 -04002439 union {
2440 int offset;
2441 int deprecated_val;
2442 } u;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002443};
2444
Lukas Czerner27dd4382013-04-09 22:11:22 -04002445static int parse_strtoull(const char *buf,
2446 unsigned long long max, unsigned long long *value)
2447{
2448 int ret;
2449
2450 ret = kstrtoull(skip_spaces(buf), 0, value);
2451 if (!ret && *value > max)
2452 ret = -EINVAL;
2453 return ret;
2454}
2455
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002456static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi,
2458 char *buf)
2459{
2460 return snprintf(buf, PAGE_SIZE, "%llu\n",
Aditya Kali7b415bf2011-09-09 19:04:51 -04002461 (s64) EXT4_C2B(sbi,
2462 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002463}
2464
2465static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2466 struct ext4_sb_info *sbi, char *buf)
2467{
2468 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2469
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002470 if (!sb->s_bdev->bd_part)
2471 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002472 return snprintf(buf, PAGE_SIZE, "%lu\n",
2473 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2474 sbi->s_sectors_written_start) >> 1);
2475}
2476
2477static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi, char *buf)
2479{
2480 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2481
Theodore Ts'of613dfc2010-07-27 11:56:08 -04002482 if (!sb->s_bdev->bd_part)
2483 return snprintf(buf, PAGE_SIZE, "0\n");
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002484 return snprintf(buf, PAGE_SIZE, "%llu\n",
Andrew Mortona6b43e32009-12-23 07:48:08 -05002485 (unsigned long long)(sbi->s_kbytes_written +
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002486 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
Andrew Mortona6b43e32009-12-23 07:48:08 -05002487 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002488}
2489
2490static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi,
2492 const char *buf, size_t count)
2493{
2494 unsigned long t;
Lukas Czernere1091b12013-04-11 23:37:19 -04002495 int ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002496
Lukas Czernere1091b12013-04-11 23:37:19 -04002497 ret = kstrtoul(skip_spaces(buf), 0, &t);
2498 if (ret)
2499 return ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002500
Lukas Czernere1091b12013-04-11 23:37:19 -04002501 if (t && (!is_power_of_2(t) || t > 0x40000000))
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002502 return -EINVAL;
2503
2504 sbi->s_inode_readahead_blks = t;
2505 return count;
2506}
2507
2508static ssize_t sbi_ui_show(struct ext4_attr *a,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002509 struct ext4_sb_info *sbi, char *buf)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002510{
Jan Karaf2d50a62013-06-04 12:51:16 -04002511 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002512
2513 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2514}
2515
2516static ssize_t sbi_ui_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519{
Jan Karaf2d50a62013-06-04 12:51:16 -04002520 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002521 unsigned long t;
Lukas Czernere1091b12013-04-11 23:37:19 -04002522 int ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002523
Lukas Czernere1091b12013-04-11 23:37:19 -04002524 ret = kstrtoul(skip_spaces(buf), 0, &t);
2525 if (ret)
2526 return ret;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002527 *ui = t;
2528 return count;
2529}
2530
Lukas Czerner52c198c2014-09-11 11:18:13 -04002531static ssize_t es_ui_show(struct ext4_attr *a,
2532 struct ext4_sb_info *sbi, char *buf)
2533{
2534
2535 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2536 a->u.offset);
2537
2538 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2539}
2540
Lukas Czerner27dd4382013-04-09 22:11:22 -04002541static ssize_t reserved_clusters_show(struct ext4_attr *a,
2542 struct ext4_sb_info *sbi, char *buf)
2543{
2544 return snprintf(buf, PAGE_SIZE, "%llu\n",
2545 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2546}
2547
2548static ssize_t reserved_clusters_store(struct ext4_attr *a,
2549 struct ext4_sb_info *sbi,
2550 const char *buf, size_t count)
2551{
2552 unsigned long long val;
2553 int ret;
2554
2555 if (parse_strtoull(buf, -1ULL, &val))
2556 return -EINVAL;
2557 ret = ext4_reserve_clusters(sbi, val);
2558
2559 return ret ? ret : count;
2560}
2561
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04002562static ssize_t trigger_test_error(struct ext4_attr *a,
2563 struct ext4_sb_info *sbi,
2564 const char *buf, size_t count)
2565{
2566 int len = count;
2567
2568 if (!capable(CAP_SYS_ADMIN))
2569 return -EPERM;
2570
2571 if (len && buf[len-1] == '\n')
2572 len--;
2573
2574 if (len)
2575 ext4_error(sbi->s_sb, "%.*s", len, buf);
2576 return count;
2577}
2578
Jan Karaf2d50a62013-06-04 12:51:16 -04002579static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2580 struct ext4_sb_info *sbi, char *buf)
2581{
2582 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2583}
2584
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002585#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2586static struct ext4_attr ext4_attr_##_name = { \
2587 .attr = {.name = __stringify(_name), .mode = _mode }, \
2588 .show = _show, \
2589 .store = _store, \
Jan Karaf2d50a62013-06-04 12:51:16 -04002590 .u = { \
2591 .offset = offsetof(struct ext4_sb_info, _elname),\
2592 }, \
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002593}
Lukas Czerner52c198c2014-09-11 11:18:13 -04002594
2595#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2596static struct ext4_attr ext4_attr_##_name = { \
2597 .attr = {.name = __stringify(_name), .mode = _mode }, \
2598 .show = _show, \
2599 .store = _store, \
2600 .u = { \
2601 .offset = offsetof(struct ext4_super_block, _elname), \
2602 }, \
2603}
2604
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002605#define EXT4_ATTR(name, mode, show, store) \
2606static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2607
Lukas Czerner857ac882010-10-27 21:30:05 -04002608#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002609#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2610#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
Lukas Czerner52c198c2014-09-11 11:18:13 -04002611
2612#define EXT4_RO_ATTR_ES_UI(name, elname) \
2613 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002614#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2615 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
Lukas Czerner52c198c2014-09-11 11:18:13 -04002616
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002617#define ATTR_LIST(name) &ext4_attr_##name.attr
Jan Karaf2d50a62013-06-04 12:51:16 -04002618#define EXT4_DEPRECATED_ATTR(_name, _val) \
2619static struct ext4_attr ext4_attr_##_name = { \
2620 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2621 .show = sbi_deprecated_show, \
2622 .u = { \
2623 .deprecated_val = _val, \
2624 }, \
2625}
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002626
2627EXT4_RO_ATTR(delayed_allocation_blocks);
2628EXT4_RO_ATTR(session_write_kbytes);
2629EXT4_RO_ATTR(lifetime_write_kbytes);
Lukas Czerner27dd4382013-04-09 22:11:22 -04002630EXT4_RW_ATTR(reserved_clusters);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002631EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2632 inode_readahead_blks_store, s_inode_readahead_blks);
Andreas Dilger11013912009-06-13 11:45:35 -04002633EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002634EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2635EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2636EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2637EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2638EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2639EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
Jan Karaf2d50a62013-06-04 12:51:16 -04002640EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
Zheng Liu67a5da52012-08-17 09:54:17 -04002641EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04002642EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04002643EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2644EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2645EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2646EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2647EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2648EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
Lukas Czerner52c198c2014-09-11 11:18:13 -04002649EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2650EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2651EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002652
2653static struct attribute *ext4_attrs[] = {
2654 ATTR_LIST(delayed_allocation_blocks),
2655 ATTR_LIST(session_write_kbytes),
2656 ATTR_LIST(lifetime_write_kbytes),
Lukas Czerner27dd4382013-04-09 22:11:22 -04002657 ATTR_LIST(reserved_clusters),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002658 ATTR_LIST(inode_readahead_blks),
Andreas Dilger11013912009-06-13 11:45:35 -04002659 ATTR_LIST(inode_goal),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002660 ATTR_LIST(mb_stats),
2661 ATTR_LIST(mb_max_to_scan),
2662 ATTR_LIST(mb_min_to_scan),
2663 ATTR_LIST(mb_order2_req),
2664 ATTR_LIST(mb_stream_req),
2665 ATTR_LIST(mb_group_prealloc),
Theodore Ts'o55138e02009-09-29 13:31:31 -04002666 ATTR_LIST(max_writeback_mb_bump),
Zheng Liu67a5da52012-08-17 09:54:17 -04002667 ATTR_LIST(extent_max_zeroout_kb),
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04002668 ATTR_LIST(trigger_fs_error),
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04002669 ATTR_LIST(err_ratelimit_interval_ms),
2670 ATTR_LIST(err_ratelimit_burst),
2671 ATTR_LIST(warning_ratelimit_interval_ms),
2672 ATTR_LIST(warning_ratelimit_burst),
2673 ATTR_LIST(msg_ratelimit_interval_ms),
2674 ATTR_LIST(msg_ratelimit_burst),
Lukas Czerner52c198c2014-09-11 11:18:13 -04002675 ATTR_LIST(errors_count),
2676 ATTR_LIST(first_error_time),
2677 ATTR_LIST(last_error_time),
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002678 NULL,
2679};
2680
Lukas Czerner857ac882010-10-27 21:30:05 -04002681/* Features this copy of ext4 supports */
2682EXT4_INFO_ATTR(lazy_itable_init);
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002683EXT4_INFO_ATTR(batched_discard);
Theodore Ts'o5e7bbef2012-09-13 12:11:40 -04002684EXT4_INFO_ATTR(meta_bg_resize);
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04002685EXT4_INFO_ATTR(encryption);
Lukas Czerner857ac882010-10-27 21:30:05 -04002686
2687static struct attribute *ext4_feat_attrs[] = {
2688 ATTR_LIST(lazy_itable_init),
Lukas Czerner27ee40d2010-10-27 21:30:12 -04002689 ATTR_LIST(batched_discard),
Theodore Ts'o5e7bbef2012-09-13 12:11:40 -04002690 ATTR_LIST(meta_bg_resize),
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04002691 ATTR_LIST(encryption),
Lukas Czerner857ac882010-10-27 21:30:05 -04002692 NULL,
2693};
2694
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002695static ssize_t ext4_attr_show(struct kobject *kobj,
2696 struct attribute *attr, char *buf)
2697{
2698 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2699 s_kobj);
2700 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2701
2702 return a->show ? a->show(a, sbi, buf) : 0;
2703}
2704
2705static ssize_t ext4_attr_store(struct kobject *kobj,
2706 struct attribute *attr,
2707 const char *buf, size_t len)
2708{
2709 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2710 s_kobj);
2711 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2712
2713 return a->store ? a->store(a, sbi, buf, len) : 0;
2714}
2715
2716static void ext4_sb_release(struct kobject *kobj)
2717{
2718 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2719 s_kobj);
2720 complete(&sbi->s_kobj_unregister);
2721}
2722
Emese Revfy52cf25d2010-01-19 02:58:23 +01002723static const struct sysfs_ops ext4_attr_ops = {
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04002724 .show = ext4_attr_show,
2725 .store = ext4_attr_store,
2726};
2727
2728static struct kobj_type ext4_ktype = {
2729 .default_attrs = ext4_attrs,
2730 .sysfs_ops = &ext4_attr_ops,
2731 .release = ext4_sb_release,
2732};
2733
Lukas Czerner857ac882010-10-27 21:30:05 -04002734static void ext4_feat_release(struct kobject *kobj)
2735{
2736 complete(&ext4_feat->f_kobj_unregister);
2737}
2738
Lukas Czernerc7f72542014-09-11 11:27:58 -04002739static ssize_t ext4_feat_show(struct kobject *kobj,
2740 struct attribute *attr, char *buf)
2741{
2742 return snprintf(buf, PAGE_SIZE, "supported\n");
2743}
2744
2745/*
2746 * We can not use ext4_attr_show/store because it relies on the kobject
2747 * being embedded in the ext4_sb_info structure which is definitely not
2748 * true in this case.
2749 */
2750static const struct sysfs_ops ext4_feat_ops = {
2751 .show = ext4_feat_show,
2752 .store = NULL,
2753};
2754
Lukas Czerner857ac882010-10-27 21:30:05 -04002755static struct kobj_type ext4_feat_ktype = {
2756 .default_attrs = ext4_feat_attrs,
Lukas Czernerc7f72542014-09-11 11:27:58 -04002757 .sysfs_ops = &ext4_feat_ops,
Lukas Czerner857ac882010-10-27 21:30:05 -04002758 .release = ext4_feat_release,
2759};
2760
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002761/*
2762 * Check whether this filesystem can be mounted based on
2763 * the features present and the RDONLY/RDWR mount requested.
2764 * Returns 1 if this filesystem can be mounted as requested,
2765 * 0 if it cannot be.
2766 */
2767static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2768{
2769 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2770 ext4_msg(sb, KERN_ERR,
2771 "Couldn't mount because of "
2772 "unsupported optional features (%x)",
2773 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2774 ~EXT4_FEATURE_INCOMPAT_SUPP));
2775 return 0;
2776 }
2777
2778 if (readonly)
2779 return 1;
2780
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05002781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2782 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2783 sb->s_flags |= MS_RDONLY;
2784 return 1;
2785 }
2786
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002787 /* Check that feature set is OK for a read-write mount */
2788 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2789 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2790 "unsupported optional features (%x)",
2791 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2792 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2793 return 0;
2794 }
2795 /*
2796 * Large file size enabled file system can only be mounted
2797 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2798 */
2799 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2800 if (sizeof(blkcnt_t) < sizeof(u64)) {
2801 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2802 "cannot be mounted RDWR without "
2803 "CONFIG_LBDAF");
2804 return 0;
2805 }
2806 }
Theodore Ts'obab08ab2011-09-09 18:36:51 -04002807 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2808 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2809 ext4_msg(sb, KERN_ERR,
2810 "Can't support bigalloc feature without "
2811 "extents feature\n");
2812 return 0;
2813 }
Aditya Kali7c319d32012-07-22 20:21:31 -04002814
2815#ifndef CONFIG_QUOTA
2816 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2817 !readonly) {
2818 ext4_msg(sb, KERN_ERR,
2819 "Filesystem with quota feature cannot be mounted RDWR "
2820 "without CONFIG_QUOTA");
2821 return 0;
2822 }
2823#endif /* CONFIG_QUOTA */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002824 return 1;
2825}
2826
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002827/*
2828 * This function is called once a day if we have errors logged
2829 * on the file system
2830 */
2831static void print_daily_error_info(unsigned long arg)
2832{
2833 struct super_block *sb = (struct super_block *) arg;
2834 struct ext4_sb_info *sbi;
2835 struct ext4_super_block *es;
2836
2837 sbi = EXT4_SB(sb);
2838 es = sbi->s_es;
2839
2840 if (es->s_error_count)
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002841 /* fsck newer than v1.41.13 is needed to clean this condition. */
2842 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002843 le32_to_cpu(es->s_error_count));
2844 if (es->s_first_error_time) {
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002845 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002846 sb->s_id, le32_to_cpu(es->s_first_error_time),
2847 (int) sizeof(es->s_first_error_func),
2848 es->s_first_error_func,
2849 le32_to_cpu(es->s_first_error_line));
2850 if (es->s_first_error_ino)
2851 printk(": inode %u",
2852 le32_to_cpu(es->s_first_error_ino));
2853 if (es->s_first_error_block)
2854 printk(": block %llu", (unsigned long long)
2855 le64_to_cpu(es->s_first_error_block));
2856 printk("\n");
2857 }
2858 if (es->s_last_error_time) {
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002859 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002860 sb->s_id, le32_to_cpu(es->s_last_error_time),
2861 (int) sizeof(es->s_last_error_func),
2862 es->s_last_error_func,
2863 le32_to_cpu(es->s_last_error_line));
2864 if (es->s_last_error_ino)
2865 printk(": inode %u",
2866 le32_to_cpu(es->s_last_error_ino));
2867 if (es->s_last_error_block)
2868 printk(": block %llu", (unsigned long long)
2869 le64_to_cpu(es->s_last_error_block));
2870 printk("\n");
2871 }
2872 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2873}
2874
Lukas Czernerbfff6872010-10-27 21:30:05 -04002875/* Find next suitable group and run ext4_init_inode_table */
2876static int ext4_run_li_request(struct ext4_li_request *elr)
2877{
2878 struct ext4_group_desc *gdp = NULL;
2879 ext4_group_t group, ngroups;
2880 struct super_block *sb;
2881 unsigned long timeout = 0;
2882 int ret = 0;
2883
2884 sb = elr->lr_super;
2885 ngroups = EXT4_SB(sb)->s_groups_count;
2886
Jan Kara8e8ad8a2012-06-12 16:20:38 +02002887 sb_start_write(sb);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002888 for (group = elr->lr_next_group; group < ngroups; group++) {
2889 gdp = ext4_get_group_desc(sb, group, NULL);
2890 if (!gdp) {
2891 ret = 1;
2892 break;
2893 }
2894
2895 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2896 break;
2897 }
2898
Theodore Ts'o7f511862013-01-13 08:41:45 -05002899 if (group >= ngroups)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002900 ret = 1;
2901
2902 if (!ret) {
2903 timeout = jiffies;
2904 ret = ext4_init_inode_table(sb, group,
2905 elr->lr_timeout ? 0 : 1);
2906 if (elr->lr_timeout == 0) {
Lukas Czerner51ce6512011-05-20 13:55:16 -04002907 timeout = (jiffies - timeout) *
2908 elr->lr_sbi->s_li_wait_mult;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002909 elr->lr_timeout = timeout;
2910 }
2911 elr->lr_next_sched = jiffies + elr->lr_timeout;
2912 elr->lr_next_group = group + 1;
2913 }
Jan Kara8e8ad8a2012-06-12 16:20:38 +02002914 sb_end_write(sb);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002915
2916 return ret;
2917}
2918
2919/*
2920 * Remove lr_request from the list_request and free the
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002921 * request structure. Should be called with li_list_mtx held
Lukas Czernerbfff6872010-10-27 21:30:05 -04002922 */
2923static void ext4_remove_li_request(struct ext4_li_request *elr)
2924{
2925 struct ext4_sb_info *sbi;
2926
2927 if (!elr)
2928 return;
2929
2930 sbi = elr->lr_sbi;
2931
2932 list_del(&elr->lr_request);
2933 sbi->s_li_request = NULL;
2934 kfree(elr);
2935}
2936
2937static void ext4_unregister_li_request(struct super_block *sb)
2938{
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002939 mutex_lock(&ext4_li_mtx);
2940 if (!ext4_li_info) {
2941 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002942 return;
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002943 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002944
2945 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002946 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002947 mutex_unlock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002948 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002949}
2950
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002951static struct task_struct *ext4_lazyinit_task;
2952
Lukas Czernerbfff6872010-10-27 21:30:05 -04002953/*
2954 * This is the function where ext4lazyinit thread lives. It walks
2955 * through the request list searching for next scheduled filesystem.
2956 * When such a fs is found, run the lazy initialization request
2957 * (ext4_rn_li_request) and keep track of the time spend in this
2958 * function. Based on that time we compute next schedule time of
2959 * the request. When walking through the list is complete, compute
2960 * next waking time and put itself into sleep.
2961 */
2962static int ext4_lazyinit_thread(void *arg)
2963{
2964 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2965 struct list_head *pos, *n;
2966 struct ext4_li_request *elr;
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002967 unsigned long next_wakeup, cur;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002968
2969 BUG_ON(NULL == eli);
2970
Lukas Czernerbfff6872010-10-27 21:30:05 -04002971cont_thread:
2972 while (true) {
2973 next_wakeup = MAX_JIFFY_OFFSET;
2974
2975 mutex_lock(&eli->li_list_mtx);
2976 if (list_empty(&eli->li_request_list)) {
2977 mutex_unlock(&eli->li_list_mtx);
2978 goto exit_thread;
2979 }
2980
2981 list_for_each_safe(pos, n, &eli->li_request_list) {
2982 elr = list_entry(pos, struct ext4_li_request,
2983 lr_request);
2984
Theodore Ts'ob2c78cd2010-11-02 14:19:30 -04002985 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2986 if (ext4_run_li_request(elr) != 0) {
2987 /* error, remove the lazy_init job */
2988 ext4_remove_li_request(elr);
2989 continue;
2990 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002991 }
2992
2993 if (time_before(elr->lr_next_sched, next_wakeup))
2994 next_wakeup = elr->lr_next_sched;
2995 }
2996 mutex_unlock(&eli->li_list_mtx);
2997
Tejun Heoa0acae02011-11-21 12:32:22 -08002998 try_to_freeze();
Lukas Czernerbfff6872010-10-27 21:30:05 -04002999
Lukas Czerner4ed5c032011-05-20 13:49:04 -04003000 cur = jiffies;
3001 if ((time_after_eq(cur, next_wakeup)) ||
Lukas Czernerf4245bd2010-11-02 14:07:17 -04003002 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04003003 cond_resched();
3004 continue;
3005 }
3006
Lukas Czerner4ed5c032011-05-20 13:49:04 -04003007 schedule_timeout_interruptible(next_wakeup - cur);
3008
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003009 if (kthread_should_stop()) {
3010 ext4_clear_request_list();
3011 goto exit_thread;
3012 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003013 }
3014
3015exit_thread:
3016 /*
3017 * It looks like the request list is empty, but we need
3018 * to check it under the li_list_mtx lock, to prevent any
3019 * additions into it, and of course we should lock ext4_li_mtx
3020 * to atomically free the list and ext4_li_info, because at
3021 * this point another ext4 filesystem could be registering
3022 * new one.
3023 */
3024 mutex_lock(&ext4_li_mtx);
3025 mutex_lock(&eli->li_list_mtx);
3026 if (!list_empty(&eli->li_request_list)) {
3027 mutex_unlock(&eli->li_list_mtx);
3028 mutex_unlock(&ext4_li_mtx);
3029 goto cont_thread;
3030 }
3031 mutex_unlock(&eli->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003032 kfree(ext4_li_info);
3033 ext4_li_info = NULL;
3034 mutex_unlock(&ext4_li_mtx);
3035
3036 return 0;
3037}
3038
3039static void ext4_clear_request_list(void)
3040{
3041 struct list_head *pos, *n;
3042 struct ext4_li_request *elr;
3043
3044 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003045 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3046 elr = list_entry(pos, struct ext4_li_request,
3047 lr_request);
3048 ext4_remove_li_request(elr);
3049 }
3050 mutex_unlock(&ext4_li_info->li_list_mtx);
3051}
3052
3053static int ext4_run_lazyinit_thread(void)
3054{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003055 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3056 ext4_li_info, "ext4lazyinit");
3057 if (IS_ERR(ext4_lazyinit_task)) {
3058 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003059 ext4_clear_request_list();
Lukas Czernerbfff6872010-10-27 21:30:05 -04003060 kfree(ext4_li_info);
3061 ext4_li_info = NULL;
Theodore Ts'o92b97812012-03-19 23:41:49 -04003062 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
Lukas Czernerbfff6872010-10-27 21:30:05 -04003063 "initialization thread\n",
3064 err);
3065 return err;
3066 }
3067 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003068 return 0;
3069}
3070
3071/*
3072 * Check whether it make sense to run itable init. thread or not.
3073 * If there is at least one uninitialized inode table, return
3074 * corresponding group number, else the loop goes through all
3075 * groups and return total number of groups.
3076 */
3077static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3078{
3079 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3080 struct ext4_group_desc *gdp = NULL;
3081
3082 for (group = 0; group < ngroups; group++) {
3083 gdp = ext4_get_group_desc(sb, group, NULL);
3084 if (!gdp)
3085 continue;
3086
3087 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3088 break;
3089 }
3090
3091 return group;
3092}
3093
3094static int ext4_li_info_new(void)
3095{
3096 struct ext4_lazy_init *eli = NULL;
3097
3098 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3099 if (!eli)
3100 return -ENOMEM;
3101
Lukas Czernerbfff6872010-10-27 21:30:05 -04003102 INIT_LIST_HEAD(&eli->li_request_list);
3103 mutex_init(&eli->li_list_mtx);
3104
Lukas Czernerbfff6872010-10-27 21:30:05 -04003105 eli->li_state |= EXT4_LAZYINIT_QUIT;
3106
3107 ext4_li_info = eli;
3108
3109 return 0;
3110}
3111
3112static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3113 ext4_group_t start)
3114{
3115 struct ext4_sb_info *sbi = EXT4_SB(sb);
3116 struct ext4_li_request *elr;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003117
3118 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3119 if (!elr)
3120 return NULL;
3121
3122 elr->lr_super = sb;
3123 elr->lr_sbi = sbi;
3124 elr->lr_next_group = start;
3125
3126 /*
3127 * Randomize first schedule time of the request to
3128 * spread the inode table initialization requests
3129 * better.
3130 */
Theodore Ts'odd1f7232013-11-08 00:14:53 -05003131 elr->lr_next_sched = jiffies + (prandom_u32() %
3132 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
Lukas Czernerbfff6872010-10-27 21:30:05 -04003133 return elr;
3134}
3135
Theodore Ts'o7f511862013-01-13 08:41:45 -05003136int ext4_register_li_request(struct super_block *sb,
3137 ext4_group_t first_not_zeroed)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003138{
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003140 struct ext4_li_request *elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003141 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05003142 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003143
Theodore Ts'o7f511862013-01-13 08:41:45 -05003144 mutex_lock(&ext4_li_mtx);
Lukas Czerner51ce6512011-05-20 13:55:16 -04003145 if (sbi->s_li_request != NULL) {
3146 /*
3147 * Reset timeout so it can be computed again, because
3148 * s_li_wait_mult might have changed.
3149 */
3150 sbi->s_li_request->lr_timeout = 0;
Theodore Ts'o7f511862013-01-13 08:41:45 -05003151 goto out;
Lukas Czerner51ce6512011-05-20 13:55:16 -04003152 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003153
3154 if (first_not_zeroed == ngroups ||
3155 (sb->s_flags & MS_RDONLY) ||
Tao Ma55ff3842011-05-09 10:28:41 -04003156 !test_opt(sb, INIT_INODE_TABLE))
Theodore Ts'o7f511862013-01-13 08:41:45 -05003157 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003158
3159 elr = ext4_li_request_new(sb, first_not_zeroed);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003160 if (!elr) {
3161 ret = -ENOMEM;
3162 goto out;
3163 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003164
3165 if (NULL == ext4_li_info) {
3166 ret = ext4_li_info_new();
3167 if (ret)
3168 goto out;
3169 }
3170
3171 mutex_lock(&ext4_li_info->li_list_mtx);
3172 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3173 mutex_unlock(&ext4_li_info->li_list_mtx);
3174
3175 sbi->s_li_request = elr;
Tao Ma46e46902011-04-04 16:00:49 -04003176 /*
3177 * set elr to NULL here since it has been inserted to
3178 * the request_list and the removal and free of it is
3179 * handled by ext4_clear_request_list from now on.
3180 */
3181 elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003182
3183 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3184 ret = ext4_run_lazyinit_thread();
3185 if (ret)
3186 goto out;
3187 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003188out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003189 mutex_unlock(&ext4_li_mtx);
3190 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003191 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003192 return ret;
3193}
3194
3195/*
3196 * We do not need to lock anything since this is called on
3197 * module unload.
3198 */
3199static void ext4_destroy_lazyinit_thread(void)
3200{
3201 /*
3202 * If thread exited earlier
3203 * there's nothing to be done.
3204 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003205 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003206 return;
3207
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003208 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003209}
3210
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003211static int set_journal_csum_feature_set(struct super_block *sb)
3212{
3213 int ret = 1;
3214 int compat, incompat;
3215 struct ext4_sb_info *sbi = EXT4_SB(sb);
3216
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -04003217 if (ext4_has_metadata_csum(sb)) {
Darrick J. Wongdb9ee222014-08-27 18:40:07 -04003218 /* journal checksum v3 */
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003219 compat = 0;
Darrick J. Wongdb9ee222014-08-27 18:40:07 -04003220 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003221 } else {
3222 /* journal checksum v1 */
3223 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3224 incompat = 0;
3225 }
3226
Darrick J. Wongfeb8c6d2014-09-11 11:38:21 -04003227 jbd2_journal_clear_features(sbi->s_journal,
3228 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3229 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3230 JBD2_FEATURE_INCOMPAT_CSUM_V2);
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003231 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3232 ret = jbd2_journal_set_features(sbi->s_journal,
3233 compat, 0,
3234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3235 incompat);
3236 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3237 ret = jbd2_journal_set_features(sbi->s_journal,
3238 compat, 0,
3239 incompat);
3240 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3241 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3242 } else {
Darrick J. Wongfeb8c6d2014-09-11 11:38:21 -04003243 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003245 }
3246
3247 return ret;
3248}
3249
Theodore Ts'o952fc182012-07-09 16:27:05 -04003250/*
3251 * Note: calculating the overhead so we can be compatible with
3252 * historical BSD practice is quite difficult in the face of
3253 * clusters/bigalloc. This is because multiple metadata blocks from
3254 * different block group can end up in the same allocation cluster.
3255 * Calculating the exact overhead in the face of clustered allocation
3256 * requires either O(all block bitmaps) in memory or O(number of block
3257 * groups**2) in time. We will still calculate the superblock for
3258 * older file systems --- and if we come across with a bigalloc file
3259 * system with zero in s_overhead_clusters the estimate will be close to
3260 * correct especially for very large cluster sizes --- but for newer
3261 * file systems, it's better to calculate this figure once at mkfs
3262 * time, and store it in the superblock. If the superblock value is
3263 * present (even for non-bigalloc file systems), we will use it.
3264 */
3265static int count_overhead(struct super_block *sb, ext4_group_t grp,
3266 char *buf)
3267{
3268 struct ext4_sb_info *sbi = EXT4_SB(sb);
3269 struct ext4_group_desc *gdp;
3270 ext4_fsblk_t first_block, last_block, b;
3271 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3272 int s, j, count = 0;
3273
Theodore Ts'o0548bbb2012-08-16 11:59:04 -04003274 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3275 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3276 sbi->s_itb_per_group + 2);
3277
Theodore Ts'o952fc182012-07-09 16:27:05 -04003278 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3279 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3280 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3281 for (i = 0; i < ngroups; i++) {
3282 gdp = ext4_get_group_desc(sb, i, NULL);
3283 b = ext4_block_bitmap(sb, gdp);
3284 if (b >= first_block && b <= last_block) {
3285 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3286 count++;
3287 }
3288 b = ext4_inode_bitmap(sb, gdp);
3289 if (b >= first_block && b <= last_block) {
3290 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3291 count++;
3292 }
3293 b = ext4_inode_table(sb, gdp);
3294 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3295 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3296 int c = EXT4_B2C(sbi, b - first_block);
3297 ext4_set_bit(c, buf);
3298 count++;
3299 }
3300 if (i != grp)
3301 continue;
3302 s = 0;
3303 if (ext4_bg_has_super(sb, grp)) {
3304 ext4_set_bit(s++, buf);
3305 count++;
3306 }
3307 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3308 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3309 count++;
3310 }
3311 }
3312 if (!count)
3313 return 0;
3314 return EXT4_CLUSTERS_PER_GROUP(sb) -
3315 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3316}
3317
3318/*
3319 * Compute the overhead and stash it in sbi->s_overhead
3320 */
3321int ext4_calculate_overhead(struct super_block *sb)
3322{
3323 struct ext4_sb_info *sbi = EXT4_SB(sb);
3324 struct ext4_super_block *es = sbi->s_es;
3325 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3326 ext4_fsblk_t overhead = 0;
Dmitry Monakhov4fdb5542014-11-25 13:08:04 -05003327 char *buf = (char *) get_zeroed_page(GFP_NOFS);
Theodore Ts'o952fc182012-07-09 16:27:05 -04003328
Theodore Ts'o952fc182012-07-09 16:27:05 -04003329 if (!buf)
3330 return -ENOMEM;
3331
3332 /*
3333 * Compute the overhead (FS structures). This is constant
3334 * for a given filesystem unless the number of block groups
3335 * changes so we cache the previous value until it does.
3336 */
3337
3338 /*
3339 * All of the blocks before first_data_block are overhead
3340 */
3341 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3342
3343 /*
3344 * Add the overhead found in each block group
3345 */
3346 for (i = 0; i < ngroups; i++) {
3347 int blks;
3348
3349 blks = count_overhead(sb, i, buf);
3350 overhead += blks;
3351 if (blks)
3352 memset(buf, 0, PAGE_SIZE);
3353 cond_resched();
3354 }
Eric Sandeenb003b522014-11-25 16:27:44 -05003355 /* Add the internal journal blocks as well */
3356 if (sbi->s_journal && !sbi->journal_bdev)
Lukas Czerner810da242013-03-02 17:18:58 -05003357 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
Eric Sandeen0875a2b2012-12-25 13:56:01 -05003358
Theodore Ts'o952fc182012-07-09 16:27:05 -04003359 sbi->s_overhead = overhead;
3360 smp_wmb();
3361 free_page((unsigned long) buf);
3362 return 0;
3363}
3364
Lukas Czerner27dd4382013-04-09 22:11:22 -04003365
Jan Kara30fac0f2013-12-08 21:11:59 -05003366static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
Lukas Czerner27dd4382013-04-09 22:11:22 -04003367{
3368 ext4_fsblk_t resv_clusters;
3369
3370 /*
Jan Kara30fac0f2013-12-08 21:11:59 -05003371 * There's no need to reserve anything when we aren't using extents.
3372 * The space estimates are exact, there are no unwritten extents,
3373 * hole punching doesn't need new metadata... This is needed especially
3374 * to keep ext2/3 backward compatibility.
3375 */
3376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3377 return 0;
3378 /*
Lukas Czerner27dd4382013-04-09 22:11:22 -04003379 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3380 * This should cover the situations where we can not afford to run
3381 * out of space like for example punch hole, or converting
Lukas Czerner556615d2014-04-20 23:45:47 -04003382 * unwritten extents in delalloc path. In most cases such
Lukas Czerner27dd4382013-04-09 22:11:22 -04003383 * allocation would require 1, or 2 blocks, higher numbers are
3384 * very rare.
3385 */
Jan Kara30fac0f2013-12-08 21:11:59 -05003386 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3387 EXT4_SB(sb)->s_cluster_bits;
Lukas Czerner27dd4382013-04-09 22:11:22 -04003388
3389 do_div(resv_clusters, 50);
3390 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3391
3392 return resv_clusters;
3393}
3394
3395
3396static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3397{
3398 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3399 sbi->s_cluster_bits;
3400
3401 if (count >= clusters)
3402 return -EINVAL;
3403
3404 atomic64_set(&sbi->s_resv_clusters, count);
3405 return 0;
3406}
3407
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003408static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003409{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003410 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003411 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003412 struct ext4_super_block *es = NULL;
3413 struct ext4_sb_info *sbi;
3414 ext4_fsblk_t block;
3415 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003416 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003417 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003418 unsigned long journal_devnum = 0;
3419 unsigned long def_mount_opts;
3420 struct inode *root;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003421 char *cp;
Frank Mayhar03901312009-01-07 00:06:22 -05003422 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003423 int ret = -ENOMEM;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003424 int blocksize, clustersize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003425 unsigned int db_count;
3426 unsigned int i;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003427 int needs_recovery, has_huge_files, has_bigalloc;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003428 __u64 blocks_count;
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003429 int err = 0;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003430 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003431 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003432
3433 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3434 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003435 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003436
3437 sbi->s_blockgroup_lock =
3438 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3439 if (!sbi->s_blockgroup_lock) {
3440 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003441 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003442 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003443 sb->s_fs_info = sbi;
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04003444 sbi->s_sb = sb;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003445 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003446 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003447 if (sb->s_bdev->bd_part)
3448 sbi->s_sectors_written_start =
3449 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Michael Halcrowb30ab0e2015-04-12 00:43:56 -04003450#ifdef CONFIG_EXT4_FS_ENCRYPTION
3451 /* Modes of operations for file and directory encryption. */
3452 sbi->s_file_encryption_mode = EXT4_ENCRYPTION_MODE_AES_256_XTS;
3453 sbi->s_dir_encryption_mode = EXT4_ENCRYPTION_MODE_INVALID;
3454#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003455
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003456 /* Cleanup superblock name */
3457 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3458 *cp = '!';
3459
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003460 /* -EINVAL is default */
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003461 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003462 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003463 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003464 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003465 goto out_fail;
3466 }
3467
3468 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003469 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003470 * block sizes. We need to calculate the offset from buffer start.
3471 */
Mingming Cao617ba132006-10-11 01:20:53 -07003472 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3474 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003475 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003476 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003477 }
3478
Gioh Kima8ac9002014-09-04 22:36:15 -04003479 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003480 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003481 goto out_fail;
3482 }
3483 /*
3484 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003485 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003486 */
Theodore Ts'o2716b802012-05-28 17:47:52 -04003487 es = (struct ext4_super_block *) (bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003488 sbi->s_es = es;
3489 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003490 if (sb->s_magic != EXT4_SUPER_MAGIC)
3491 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003492 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003493
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003494 /* Warn if metadata_csum and gdt_csum are both set. */
3495 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3497 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
Jakub Wilk363307e2015-01-02 15:31:14 -05003498 ext4_warning(sb, "metadata_csum and uninit_bg are "
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003499 "redundant flags; please run fsck.");
3500
Darrick J. Wongd25425f2012-04-29 18:25:10 -04003501 /* Check for a known checksum algorithm */
3502 if (!ext4_verify_csum_type(sb, es)) {
3503 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3504 "unknown checksum algorithm.");
3505 silent = 1;
3506 goto cantfind_ext4;
3507 }
3508
Darrick J. Wong04419842012-04-29 18:27:10 -04003509 /* Load the checksum driver */
3510 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3511 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3512 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3513 if (IS_ERR(sbi->s_chksum_driver)) {
3514 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3515 ret = PTR_ERR(sbi->s_chksum_driver);
3516 sbi->s_chksum_driver = NULL;
3517 goto failed_mount;
3518 }
3519 }
3520
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003521 /* Check superblock checksum */
3522 if (!ext4_superblock_csum_verify(sb, es)) {
3523 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3524 "invalid superblock checksum. Run e2fsck?");
3525 silent = 1;
3526 goto cantfind_ext4;
3527 }
3528
3529 /* Precompute checksum seed for all metadata */
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -04003530 if (ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003531 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3532 sizeof(es->s_uuid));
3533
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003534 /* Set defaults before we parse the mount options */
3535 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003536 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003537 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003538 set_opt(sb, DEBUG);
Theodore Ts'o87f26802012-03-02 00:03:21 -05003539 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003540 set_opt(sb, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07003541 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003542 set_opt(sb, NO_UID32);
Eric Sandeenea663332011-02-23 17:51:51 -05003543 /* xattr user namespace & acls are now defaulted on */
Eric Sandeenea663332011-02-23 17:51:51 -05003544 set_opt(sb, XATTR_USER);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003545#ifdef CONFIG_EXT4_FS_POSIX_ACL
Eric Sandeenea663332011-02-23 17:51:51 -05003546 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003547#endif
Darrick J. Wong98c1a752014-10-30 10:53:16 -04003548 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3549 if (ext4_has_metadata_csum(sb))
3550 set_opt(sb, JOURNAL_CHECKSUM);
3551
Mingming Cao617ba132006-10-11 01:20:53 -07003552 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003553 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003554 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003555 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003556 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003557 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003558
Mingming Cao617ba132006-10-11 01:20:53 -07003559 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003560 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003561 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003562 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003563 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003564 set_opt(sb, ERRORS_RO);
Darrick J. Wong45f1a9c2014-09-01 21:34:09 -04003565 /* block_validity enabled by default; disable with noblock_validity */
3566 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003567 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003568 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003569
Eric W. Biederman08cefc72012-02-07 15:41:49 -08003570 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3571 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
Theodore Ts'o30773842009-01-03 20:27:38 -05003572 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3573 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3574 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003575
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003576 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003577 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003578
Mingming Cao1e2462f2007-07-18 09:00:55 -04003579 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003580 * enable delayed allocation by default
3581 * Use -o nodelalloc to turn it off
3582 */
Theodore Ts'obc0b75f2012-09-17 22:54:36 -04003583 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003584 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003585 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003586
Lukas Czerner51ce6512011-05-20 13:55:16 -04003587 /*
3588 * set default s_li_wait_mult for lazyinit, for the case there is
3589 * no mount option specified.
3590 */
3591 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3592
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003593 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
Eric Sandeen661aa522012-02-20 17:53:04 -05003594 &journal_devnum, &journal_ioprio, 0)) {
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003595 ext4_msg(sb, KERN_WARNING,
3596 "failed to parse options in superblock: %s",
3597 sbi->s_es->s_mount_opts);
3598 }
Theodore Ts'o5a916be2012-03-04 19:27:31 -05003599 sbi->s_def_mount_opt = sbi->s_mount_opt;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003600 if (!parse_options((char *) data, sb, &journal_devnum,
Eric Sandeen661aa522012-02-20 17:53:04 -05003601 &journal_ioprio, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003602 goto failed_mount;
3603
Theodore Ts'o56889782011-09-03 18:22:38 -04003604 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3605 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3606 "with data=journal disables delayed "
3607 "allocation and O_DIRECT support!\n");
3608 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3609 ext4_msg(sb, KERN_ERR, "can't mount with "
3610 "both data=journal and delalloc");
3611 goto failed_mount;
3612 }
3613 if (test_opt(sb, DIOREAD_NOLOCK)) {
3614 ext4_msg(sb, KERN_ERR, "can't mount with "
Piotr Sarna6ae65142013-08-08 23:02:24 -04003615 "both data=journal and dioread_nolock");
Theodore Ts'o56889782011-09-03 18:22:38 -04003616 goto failed_mount;
3617 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003618 if (test_opt(sb, DAX)) {
3619 ext4_msg(sb, KERN_ERR, "can't mount with "
3620 "both data=journal and dax");
3621 goto failed_mount;
3622 }
Theodore Ts'o56889782011-09-03 18:22:38 -04003623 if (test_opt(sb, DELALLOC))
3624 clear_opt(sb, DELALLOC);
3625 }
3626
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003627 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003628 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003629
Mingming Cao617ba132006-10-11 01:20:53 -07003630 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3631 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3632 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3633 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003634 ext4_msg(sb, KERN_WARNING,
3635 "feature flags set on rev 0 fs, "
3636 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003637
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003638 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3639 set_opt2(sb, HURD_COMPAT);
3640 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3641 EXT4_FEATURE_INCOMPAT_64BIT)) {
3642 ext4_msg(sb, KERN_ERR,
3643 "The Hurd can't support 64-bit file systems");
3644 goto failed_mount;
3645 }
3646 }
3647
Theodore Ts'o2035e772011-04-18 17:29:14 -04003648 if (IS_EXT2_SB(sb)) {
3649 if (ext2_feature_set_ok(sb))
3650 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3651 "using the ext4 subsystem");
3652 else {
3653 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3654 "to feature incompatibilities");
3655 goto failed_mount;
3656 }
3657 }
3658
3659 if (IS_EXT3_SB(sb)) {
3660 if (ext3_feature_set_ok(sb))
3661 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3662 "using the ext4 subsystem");
3663 else {
3664 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3665 "to feature incompatibilities");
3666 goto failed_mount;
3667 }
3668 }
3669
Theodore Tso469108f2008-02-10 01:11:44 -05003670 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003671 * Check feature flags regardless of the revision level, since we
3672 * previously didn't change the revision level when setting the flags,
3673 * so there is a chance incompat flags are set on a rev 0 filesystem.
3674 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003675 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003676 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003677
Jan Kara261cb20c2012-12-20 00:07:18 -05003678 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
Mingming Cao617ba132006-10-11 01:20:53 -07003679 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3680 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003681 ext4_msg(sb, KERN_ERR,
3682 "Unsupported filesystem blocksize %d", blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003683 goto failed_mount;
3684 }
3685
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003686 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3687 if (blocksize != PAGE_SIZE) {
3688 ext4_msg(sb, KERN_ERR,
3689 "error: unsupported blocksize for dax");
3690 goto failed_mount;
3691 }
3692 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3693 ext4_msg(sb, KERN_ERR,
3694 "error: device does not support dax");
3695 goto failed_mount;
3696 }
3697 }
3698
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04003699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3700 es->s_encryption_level) {
3701 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3702 es->s_encryption_level);
3703 goto failed_mount;
3704 }
3705
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003706 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003707 /* Validate the filesystem blocksize */
3708 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003709 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003710 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003711 goto failed_mount;
3712 }
3713
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003714 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003715 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3716 offset = do_div(logical_sb_block, blocksize);
Gioh Kima8ac9002014-09-04 22:36:15 -04003717 bh = sb_bread_unmovable(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003718 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003719 ext4_msg(sb, KERN_ERR,
3720 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003721 goto failed_mount;
3722 }
Theodore Ts'o2716b802012-05-28 17:47:52 -04003723 es = (struct ext4_super_block *)(bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003724 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003725 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003726 ext4_msg(sb, KERN_ERR,
3727 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003728 goto failed_mount;
3729 }
3730 }
3731
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003732 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3733 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003734 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3735 has_huge_files);
3736 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003737
Mingming Cao617ba132006-10-11 01:20:53 -07003738 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3739 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3740 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003741 } else {
3742 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3743 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003744 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003745 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003746 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003747 ext4_msg(sb, KERN_ERR,
3748 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003749 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003750 goto failed_mount;
3751 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3753 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003754 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003755
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003756 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003758 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003759 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003760 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003761 ext4_msg(sb, KERN_ERR,
3762 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003763 sbi->s_desc_size);
3764 goto failed_mount;
3765 }
3766 } else
3767 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003768
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003769 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003770 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003771 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003772 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003773
Mingming Cao617ba132006-10-11 01:20:53 -07003774 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003775 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003776 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003777 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3778 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003779 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003780 sbi->s_sbh = bh;
3781 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003782 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3783 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003784
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003785 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003786 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3787 sbi->s_def_hash_version = es->s_def_hash_version;
Theodore Ts'o23301412014-02-12 12:16:04 -05003788 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3789 i = le32_to_cpu(es->s_flags);
3790 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3791 sbi->s_hash_unsigned = 3;
3792 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
Theodore Ts'of99b2582008-10-28 13:21:44 -04003793#ifdef __CHAR_UNSIGNED__
Theodore Ts'o23301412014-02-12 12:16:04 -05003794 if (!(sb->s_flags & MS_RDONLY))
3795 es->s_flags |=
3796 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3797 sbi->s_hash_unsigned = 3;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003798#else
Theodore Ts'o23301412014-02-12 12:16:04 -05003799 if (!(sb->s_flags & MS_RDONLY))
3800 es->s_flags |=
3801 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
Theodore Ts'of99b2582008-10-28 13:21:44 -04003802#endif
Theodore Ts'o23301412014-02-12 12:16:04 -05003803 }
Theodore Ts'of99b2582008-10-28 13:21:44 -04003804 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003805
Theodore Ts'o281b5992011-09-09 18:34:51 -04003806 /* Handle clustersize */
3807 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3808 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3809 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3810 if (has_bigalloc) {
3811 if (clustersize < blocksize) {
3812 ext4_msg(sb, KERN_ERR,
3813 "cluster size (%d) smaller than "
3814 "block size (%d)", clustersize, blocksize);
3815 goto failed_mount;
3816 }
3817 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3818 le32_to_cpu(es->s_log_block_size);
3819 sbi->s_clusters_per_group =
3820 le32_to_cpu(es->s_clusters_per_group);
3821 if (sbi->s_clusters_per_group > blocksize * 8) {
3822 ext4_msg(sb, KERN_ERR,
3823 "#clusters per group too big: %lu",
3824 sbi->s_clusters_per_group);
3825 goto failed_mount;
3826 }
3827 if (sbi->s_blocks_per_group !=
3828 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3829 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3830 "clusters per group (%lu) inconsistent",
3831 sbi->s_blocks_per_group,
3832 sbi->s_clusters_per_group);
3833 goto failed_mount;
3834 }
3835 } else {
3836 if (clustersize != blocksize) {
3837 ext4_warning(sb, "fragment/cluster size (%d) != "
3838 "block size (%d)", clustersize,
3839 blocksize);
3840 clustersize = blocksize;
3841 }
3842 if (sbi->s_blocks_per_group > blocksize * 8) {
3843 ext4_msg(sb, KERN_ERR,
3844 "#blocks per group too big: %lu",
3845 sbi->s_blocks_per_group);
3846 goto failed_mount;
3847 }
3848 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3849 sbi->s_cluster_bits = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003850 }
Theodore Ts'o281b5992011-09-09 18:34:51 -04003851 sbi->s_cluster_ratio = clustersize / blocksize;
3852
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003853 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003854 ext4_msg(sb, KERN_ERR,
3855 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003856 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003857 goto failed_mount;
3858 }
3859
Theodore Ts'o960fd8562013-07-05 23:11:16 -04003860 /* Do we have standard group size of clustersize * 8 blocks ? */
3861 if (sbi->s_blocks_per_group == clustersize << 3)
3862 set_opt2(sb, STD_GROUP_SIZE);
3863
Eric Sandeenbf43d842009-08-17 23:48:51 -04003864 /*
3865 * Test whether we have more sectors than will fit in sector_t,
3866 * and whether the max offset is addressable by the page cache.
3867 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003868 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003869 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003870 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003871 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003872 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003873 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003874 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003875 goto failed_mount;
3876 }
3877
Mingming Cao617ba132006-10-11 01:20:53 -07003878 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3879 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003880
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003881 /* check blocks count against device size */
3882 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3883 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003884 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3885 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003886 ext4_blocks_count(es), blocks_count);
3887 goto failed_mount;
3888 }
3889
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003890 /*
3891 * It makes no sense for the first data block to be beyond the end
3892 * of the filesystem.
3893 */
3894 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Zheng Liu5635a622011-12-18 16:13:58 -05003895 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003896 "block %u is beyond end of filesystem (%llu)",
3897 le32_to_cpu(es->s_first_data_block),
3898 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003899 goto failed_mount;
3900 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003901 blocks_count = (ext4_blocks_count(es) -
3902 le32_to_cpu(es->s_first_data_block) +
3903 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3904 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003905 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003906 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003907 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003908 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003909 ext4_blocks_count(es),
3910 le32_to_cpu(es->s_first_data_block),
3911 EXT4_BLOCKS_PER_GROUP(sb));
3912 goto failed_mount;
3913 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003914 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003915 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3916 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003917 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3918 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'of18a5f22011-08-01 08:45:38 -04003919 sbi->s_group_desc = ext4_kvmalloc(db_count *
3920 sizeof(struct buffer_head *),
3921 GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003922 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003923 ext4_msg(sb, KERN_ERR, "not enough memory");
Theodore Ts'o2cde4172012-05-28 17:49:54 -04003924 ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003925 goto failed_mount;
3926 }
3927
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003928 if (ext4_proc_root)
3929 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04003930
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05003931 if (sbi->s_proc)
3932 proc_create_data("options", S_IRUGO, sbi->s_proc,
3933 &ext4_seq_options_fops, sb);
3934
Pekka Enberg705895b2009-02-15 18:07:52 -05003935 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003936
3937 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003938 block = descriptor_loc(sb, logical_sb_block, i);
Gioh Kima8ac9002014-09-04 22:36:15 -04003939 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003940 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003941 ext4_msg(sb, KERN_ERR,
3942 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003943 db_count = i;
3944 goto failed_mount2;
3945 }
3946 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003947 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003948 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04003949 goto failed_mount2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003950 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003951
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04003952 sbi->s_gdb_count = db_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003953 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3954 spin_lock_init(&sbi->s_next_gen_lock);
3955
Jan Mrazek04ecddb2015-01-26 14:42:31 -05003956 setup_timer(&sbi->s_err_report, print_daily_error_info,
3957 (unsigned long) sb);
Tao Ma04496412011-04-05 19:55:28 -04003958
Dmitry Monakhova75ae78f2013-04-03 22:10:52 -04003959 /* Register extent status tree shrinker */
Zheng Liueb68d0e2014-09-01 22:26:49 -04003960 if (ext4_es_register_shrinker(sbi))
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003961 goto failed_mount3;
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003962
Alex Tomasc9de5602008-01-29 00:19:52 -05003963 sbi->s_stripe = ext4_get_stripe_size(sbi);
Zheng Liu67a5da52012-08-17 09:54:17 -04003964 sbi->s_extent_max_zeroout_kb = 32;
Alex Tomasc9de5602008-01-29 00:19:52 -05003965
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04003966 /*
3967 * set up enough so that it can read an inode
3968 */
Theodore Ts'of6e63f92014-09-18 17:12:30 -04003969 sb->s_op = &ext4_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07003970 sb->s_export_op = &ext4_export_ops;
3971 sb->s_xattr = ext4_xattr_handlers;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003972#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003973 sb->dq_op = &ext4_quota_operations;
Jan Kara262b4662013-03-02 17:57:08 -05003974 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
Jan Kara1fa5efe2014-10-08 18:26:54 +02003975 sb->s_qcop = &dquot_quotactl_sysfile_ops;
Jan Kara262b4662013-03-02 17:57:08 -05003976 else
3977 sb->s_qcop = &ext4_qctl_operations;
Jan Kara96c7e0d2014-09-29 14:58:25 +02003978 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003979#endif
Aneesh Kumar K.Vf2fa2ff2011-01-29 18:43:40 +05303980 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3981
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003982 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003983 mutex_init(&sbi->s_orphan_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003984
3985 sb->s_root = NULL;
3986
3987 needs_recovery = (es->s_last_orphan != 0 ||
Mingming Cao617ba132006-10-11 01:20:53 -07003988 EXT4_HAS_INCOMPAT_FEATURE(sb,
3989 EXT4_FEATURE_INCOMPAT_RECOVER));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003990
Johann Lombardic5e06d12011-05-24 18:31:25 -04003991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3992 !(sb->s_flags & MS_RDONLY))
3993 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
Darrick J. Wong50460fe2014-10-30 10:53:16 -04003994 goto failed_mount3a;
Johann Lombardic5e06d12011-05-24 18:31:25 -04003995
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003996 /*
3997 * The first inode we look at is the journal inode. Don't try
3998 * root first: it may be modified in the journal!
3999 */
4000 if (!test_opt(sb, NOLOAD) &&
Mingming Cao617ba132006-10-11 01:20:53 -07004001 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4002 if (ext4_load_journal(sb, es, journal_devnum))
Darrick J. Wong50460fe2014-10-30 10:53:16 -04004003 goto failed_mount3a;
Frank Mayhar03901312009-01-07 00:06:22 -05004004 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4005 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004006 ext4_msg(sb, KERN_ERR, "required journal recovery "
4007 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004008 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004009 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004010 clear_opt(sb, DATA_FLAGS);
Frank Mayhar03901312009-01-07 00:06:22 -05004011 sbi->s_journal = NULL;
4012 needs_recovery = 0;
4013 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004014 }
4015
Theodore Ts'of32aaf22012-05-21 11:42:02 -04004016 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
Jose R. Santoseb40a092007-07-18 08:37:25 -04004017 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4018 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004019 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004020 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04004021 }
4022
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04004023 if (!set_journal_csum_feature_set(sb)) {
4024 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4025 "feature set");
4026 goto failed_mount_wq;
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08004027 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05004028
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004029 /* We have now updated the journal if required, so we can
4030 * validate the data journaling mode. */
4031 switch (test_opt(sb, DATA_FLAGS)) {
4032 case 0:
4033 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07004034 * capabilities: ORDERED_DATA if the journal can
4035 * cope, else JOURNAL_DATA
4036 */
Mingming Caodab291a2006-10-11 01:21:01 -07004037 if (jbd2_journal_check_available_features
4038 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004039 set_opt(sb, ORDERED_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004040 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004041 set_opt(sb, JOURNAL_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004042 break;
4043
Mingming Cao617ba132006-10-11 01:20:53 -07004044 case EXT4_MOUNT_ORDERED_DATA:
4045 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07004046 if (!jbd2_journal_check_available_features
4047 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004048 ext4_msg(sb, KERN_ERR, "Journal does not support "
4049 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004050 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004051 }
4052 default:
4053 break;
4054 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004055 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004056
Bobi Jam18aadd42012-02-20 17:53:02 -05004057 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4058
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004059no_journal:
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04004060 if (ext4_mballoc_ready) {
4061 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4062 if (!sbi->s_mb_cache) {
4063 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4064 goto failed_mount_wq;
4065 }
4066 }
4067
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04004068 if (unlikely(sbi->s_mount_flags & EXT4_MF_TEST_DUMMY_ENCRYPTION) &&
4069 !(sb->s_flags & MS_RDONLY) &&
4070 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4071 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4072 ext4_commit_super(sb, 1);
4073 }
4074
Tejun Heofd89d5f2011-02-01 11:42:42 +01004075 /*
Theodore Ts'o952fc182012-07-09 16:27:05 -04004076 * Get the # of file system overhead blocks from the
4077 * superblock if present.
4078 */
4079 if (es->s_overhead_clusters)
4080 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4081 else {
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004082 err = ext4_calculate_overhead(sb);
4083 if (err)
Theodore Ts'o952fc182012-07-09 16:27:05 -04004084 goto failed_mount_wq;
4085 }
4086
4087 /*
Tejun Heofd89d5f2011-02-01 11:42:42 +01004088 * The maximum number of concurrent works can be high and
4089 * concurrency isn't really necessary. Limit it to 1.
4090 */
Jan Kara2e8fa542013-06-04 14:21:02 -04004091 EXT4_SB(sb)->rsv_conversion_wq =
4092 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4093 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4094 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004095 ret = -ENOMEM;
Jan Kara2e8fa542013-06-04 14:21:02 -04004096 goto failed_mount4;
4097 }
4098
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004099 /*
Mingming Caodab291a2006-10-11 01:21:01 -07004100 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004101 * so we can safely mount the rest of the filesystem now.
4102 */
4103
David Howells1d1fe1e2008-02-07 00:15:37 -08004104 root = ext4_iget(sb, EXT4_ROOT_INO);
4105 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004106 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004107 ret = PTR_ERR(root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004108 root = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004109 goto failed_mount4;
4110 }
4111 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004112 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Al Viro94bf6082012-01-09 15:53:24 -05004113 iput(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004114 goto failed_mount4;
4115 }
Al Viro48fde702012-01-08 22:15:13 -05004116 sb->s_root = d_make_root(root);
David Howells1d1fe1e2008-02-07 00:15:37 -08004117 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004118 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004119 ret = -ENOMEM;
4120 goto failed_mount4;
4121 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004122
Eric Sandeen7e84b622012-05-28 14:17:25 -04004123 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4124 sb->s_flags |= MS_RDONLY;
Kalpak Shahef7f3832007-07-18 09:15:20 -04004125
4126 /* determine the minimum size of new large inodes, if present */
4127 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4128 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4129 EXT4_GOOD_OLD_INODE_SIZE;
4130 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4131 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4132 if (sbi->s_want_extra_isize <
4133 le16_to_cpu(es->s_want_extra_isize))
4134 sbi->s_want_extra_isize =
4135 le16_to_cpu(es->s_want_extra_isize);
4136 if (sbi->s_want_extra_isize <
4137 le16_to_cpu(es->s_min_extra_isize))
4138 sbi->s_want_extra_isize =
4139 le16_to_cpu(es->s_min_extra_isize);
4140 }
4141 }
4142 /* Check if enough inode space is available */
4143 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4144 sbi->s_inode_size) {
4145 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4146 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04004147 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4148 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04004149 }
4150
Jan Kara30fac0f2013-12-08 21:11:59 -05004151 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
Lukas Czerner27dd4382013-04-09 22:11:22 -04004152 if (err) {
4153 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
Jan Kara30fac0f2013-12-08 21:11:59 -05004154 "reserved pool", ext4_calculate_resv_clusters(sb));
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004155 goto failed_mount4a;
Lukas Czerner27dd4382013-04-09 22:11:22 -04004156 }
4157
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004158 err = ext4_setup_system_zone(sb);
4159 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004160 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04004161 "zone (%d)", err);
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004162 goto failed_mount4a;
4163 }
4164
4165 ext4_ext_init(sb);
4166 err = ext4_mb_init(sb);
4167 if (err) {
4168 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4169 err);
Tao Madcf2d802011-10-06 12:10:11 -04004170 goto failed_mount5;
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04004171 }
4172
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004173 block = ext4_count_free_clusters(sb);
4174 ext4_free_blocks_count_set(sbi->s_es,
4175 EXT4_C2B(sbi, block));
Tejun Heo908c7f12014-09-08 09:51:29 +09004176 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4177 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004178 if (!err) {
4179 unsigned long freei = ext4_count_free_inodes(sb);
4180 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
Tejun Heo908c7f12014-09-08 09:51:29 +09004181 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4182 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004183 }
4184 if (!err)
4185 err = percpu_counter_init(&sbi->s_dirs_counter,
Tejun Heo908c7f12014-09-08 09:51:29 +09004186 ext4_count_dirs(sb), GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004187 if (!err)
Tejun Heo908c7f12014-09-08 09:51:29 +09004188 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4189 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004190 if (err) {
4191 ext4_msg(sb, KERN_ERR, "insufficient memory");
4192 goto failed_mount6;
4193 }
4194
4195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4196 if (!ext4_fill_flex_info(sb)) {
4197 ext4_msg(sb, KERN_ERR,
4198 "unable to initialize "
4199 "flex_bg meta info!");
4200 goto failed_mount6;
4201 }
4202
Lukas Czernerbfff6872010-10-27 21:30:05 -04004203 err = ext4_register_li_request(sb, first_not_zeroed);
4204 if (err)
Tao Madcf2d802011-10-06 12:10:11 -04004205 goto failed_mount6;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004206
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004207 sbi->s_kobj.kset = ext4_kset;
4208 init_completion(&sbi->s_kobj_unregister);
4209 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4210 "%s", sb->s_id);
Tao Madcf2d802011-10-06 12:10:11 -04004211 if (err)
4212 goto failed_mount7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004213
Jan Kara9b2ff3572013-03-02 18:22:38 -05004214#ifdef CONFIG_QUOTA
4215 /* Enable quota usage during mount. */
4216 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4217 !(sb->s_flags & MS_RDONLY)) {
4218 err = ext4_enable_quotas(sb);
4219 if (err)
4220 goto failed_mount8;
4221 }
4222#endif /* CONFIG_QUOTA */
4223
Mingming Cao617ba132006-10-11 01:20:53 -07004224 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4225 ext4_orphan_cleanup(sb, es);
4226 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05004227 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004228 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05004229 ext4_mark_recovery_complete(sb, es);
4230 }
4231 if (EXT4_SB(sb)->s_journal) {
4232 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4233 descr = " journalled data mode";
4234 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4235 descr = " ordered data mode";
4236 else
4237 descr = " writeback data mode";
4238 } else
4239 descr = "out journal";
4240
Lukas Czerner79add3a2012-11-08 13:28:29 -05004241 if (test_opt(sb, DISCARD)) {
4242 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4243 if (!blk_queue_discard(q))
4244 ext4_msg(sb, KERN_WARNING,
4245 "mounting with \"discard\" option, but "
4246 "the device does not support discard");
4247 }
4248
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004249 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o8b67f042010-08-01 23:14:20 -04004250 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4251 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004252
Theodore Ts'o66e61a92010-07-27 11:56:04 -04004253 if (es->s_error_count)
4254 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004255
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04004256 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4257 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4258 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4259 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4260
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004261 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004262 return 0;
4263
Mingming Cao617ba132006-10-11 01:20:53 -07004264cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004265 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004266 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004267 goto failed_mount;
4268
Theodore Ts'o72ba7452013-01-24 23:24:54 -05004269#ifdef CONFIG_QUOTA
4270failed_mount8:
4271 kobject_del(&sbi->s_kobj);
4272#endif
Tao Madcf2d802011-10-06 12:10:11 -04004273failed_mount7:
4274 ext4_unregister_li_request(sb);
4275failed_mount6:
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004276 ext4_mb_release(sb);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004277 if (sbi->s_flex_groups)
Al Virob93b41d2014-11-20 12:19:11 -05004278 kvfree(sbi->s_flex_groups);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004279 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4280 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4281 percpu_counter_destroy(&sbi->s_dirs_counter);
4282 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Azat Khuzhin00764932014-04-07 10:54:20 -04004283failed_mount5:
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004284 ext4_ext_release(sb);
4285 ext4_release_system_zone(sb);
4286failed_mount4a:
Al Viro94bf6082012-01-09 15:53:24 -05004287 dput(sb->s_root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004288 sb->s_root = NULL;
Al Viro94bf6082012-01-09 15:53:24 -05004289failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04004290 ext4_msg(sb, KERN_ERR, "mount failed");
Jan Kara2e8fa542013-06-04 14:21:02 -04004291 if (EXT4_SB(sb)->rsv_conversion_wq)
4292 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -04004293failed_mount_wq:
Frank Mayhar03901312009-01-07 00:06:22 -05004294 if (sbi->s_journal) {
4295 jbd2_journal_destroy(sbi->s_journal);
4296 sbi->s_journal = NULL;
4297 }
Darrick J. Wong50460fe2014-10-30 10:53:16 -04004298failed_mount3a:
Zheng Liud3922a72013-07-01 08:12:37 -04004299 ext4_es_unregister_shrinker(sbi);
Zheng Liueb68d0e2014-09-01 22:26:49 -04004300failed_mount3:
Al Viro9105bb12013-12-08 20:52:31 -05004301 del_timer_sync(&sbi->s_err_report);
Johann Lombardic5e06d12011-05-24 18:31:25 -04004302 if (sbi->s_mmp_tsk)
4303 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004304failed_mount2:
4305 for (i = 0; i < db_count; i++)
4306 brelse(sbi->s_group_desc[i]);
Al Virob93b41d2014-11-20 12:19:11 -05004307 kvfree(sbi->s_group_desc);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004308failed_mount:
Darrick J. Wong04419842012-04-29 18:27:10 -04004309 if (sbi->s_chksum_driver)
4310 crypto_free_shash(sbi->s_chksum_driver);
Theodore Ts'o240799c2008-10-09 23:53:47 -04004311 if (sbi->s_proc) {
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05004312 remove_proc_entry("options", sbi->s_proc);
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04004313 remove_proc_entry(sb->s_id, ext4_proc_root);
Theodore Ts'o240799c2008-10-09 23:53:47 -04004314 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004315#ifdef CONFIG_QUOTA
Jan Karaa2d4a642014-09-11 11:15:15 -04004316 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004317 kfree(sbi->s_qf_names[i]);
4318#endif
Mingming Cao617ba132006-10-11 01:20:53 -07004319 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004320 brelse(bh);
4321out_fail:
4322 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04004323 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004324 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04004325out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004326 kfree(orig_data);
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004327 return err ? err : ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004328}
4329
4330/*
4331 * Setup any per-fs journal parameters now. We'll do this both on
4332 * initial mount, once the journal has been initialised but before we've
4333 * done any recovery; and again on any subsequent remount.
4334 */
Mingming Cao617ba132006-10-11 01:20:53 -07004335static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004336{
Mingming Cao617ba132006-10-11 01:20:53 -07004337 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004338
Theodore Ts'o30773842009-01-03 20:27:38 -05004339 journal->j_commit_interval = sbi->s_commit_interval;
4340 journal->j_min_batch_time = sbi->s_min_batch_time;
4341 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004342
Theodore Ts'oa931da62010-08-03 21:35:12 -04004343 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004344 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07004345 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004346 else
Mingming Caodab291a2006-10-11 01:21:01 -07004347 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04004348 if (test_opt(sb, DATA_ERR_ABORT))
4349 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4350 else
4351 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04004352 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004353}
4354
Mingming Cao617ba132006-10-11 01:20:53 -07004355static journal_t *ext4_get_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004356 unsigned int journal_inum)
4357{
4358 struct inode *journal_inode;
4359 journal_t *journal;
4360
Frank Mayhar03901312009-01-07 00:06:22 -05004361 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4362
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004363 /* First, test for the existence of a valid inode on disk. Bad
4364 * things happen if we iget() an unused inode, as the subsequent
4365 * iput() will try to delete it. */
4366
David Howells1d1fe1e2008-02-07 00:15:37 -08004367 journal_inode = ext4_iget(sb, journal_inum);
4368 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004369 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004370 return NULL;
4371 }
4372 if (!journal_inode->i_nlink) {
4373 make_bad_inode(journal_inode);
4374 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04004375 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004376 return NULL;
4377 }
4378
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04004379 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004380 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08004381 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004382 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004383 iput(journal_inode);
4384 return NULL;
4385 }
4386
Mingming Caodab291a2006-10-11 01:21:01 -07004387 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004388 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004389 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004390 iput(journal_inode);
4391 return NULL;
4392 }
4393 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004394 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004395 return journal;
4396}
4397
Mingming Cao617ba132006-10-11 01:20:53 -07004398static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004399 dev_t j_dev)
4400{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004401 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004402 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07004403 ext4_fsblk_t start;
4404 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004405 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07004406 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004407 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004408 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004409 struct block_device *bdev;
4410
Frank Mayhar03901312009-01-07 00:06:22 -05004411 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4412
Eric Sandeenb31e1552009-06-04 17:36:36 -04004413 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004414 if (bdev == NULL)
4415 return NULL;
4416
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004417 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04004418 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004419 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004420 ext4_msg(sb, KERN_ERR,
4421 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004422 goto out_bdev;
4423 }
4424
Mingming Cao617ba132006-10-11 01:20:53 -07004425 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4426 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004427 set_blocksize(bdev, blocksize);
4428 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004429 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4430 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004431 goto out_bdev;
4432 }
4433
Theodore Ts'o2716b802012-05-28 17:47:52 -04004434 es = (struct ext4_super_block *) (bh->b_data + offset);
Mingming Cao617ba132006-10-11 01:20:53 -07004435 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004436 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07004437 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004438 ext4_msg(sb, KERN_ERR, "external journal has "
4439 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004440 brelse(bh);
4441 goto out_bdev;
4442 }
4443
Darrick J. Wongdf4763b2014-09-11 11:44:36 -04004444 if ((le32_to_cpu(es->s_feature_ro_compat) &
4445 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4446 es->s_checksum != ext4_superblock_csum(sb, es)) {
4447 ext4_msg(sb, KERN_ERR, "external journal has "
4448 "corrupt superblock");
4449 brelse(bh);
4450 goto out_bdev;
4451 }
4452
Mingming Cao617ba132006-10-11 01:20:53 -07004453 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004454 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004455 brelse(bh);
4456 goto out_bdev;
4457 }
4458
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004459 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004460 start = sb_block + 1;
4461 brelse(bh); /* we're done with the superblock */
4462
Mingming Caodab291a2006-10-11 01:21:01 -07004463 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004464 start, len, blocksize);
4465 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004466 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004467 goto out_bdev;
4468 }
4469 journal->j_private = sb;
Theodore Ts'o9f203502013-04-20 15:46:17 -04004470 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004471 wait_on_buffer(journal->j_sb_buffer);
4472 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004473 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004474 goto out_journal;
4475 }
4476 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004477 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4478 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004479 be32_to_cpu(journal->j_superblock->s_nr_users));
4480 goto out_journal;
4481 }
Mingming Cao617ba132006-10-11 01:20:53 -07004482 EXT4_SB(sb)->journal_bdev = bdev;
4483 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004484 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004485
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004486out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07004487 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004488out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07004489 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004490 return NULL;
4491}
4492
Mingming Cao617ba132006-10-11 01:20:53 -07004493static int ext4_load_journal(struct super_block *sb,
4494 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004495 unsigned long journal_devnum)
4496{
4497 journal_t *journal;
4498 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4499 dev_t journal_dev;
4500 int err = 0;
4501 int really_read_only;
4502
Frank Mayhar03901312009-01-07 00:06:22 -05004503 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4504
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004505 if (journal_devnum &&
4506 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004507 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4508 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004509 journal_dev = new_decode_dev(journal_devnum);
4510 } else
4511 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4512
4513 really_read_only = bdev_read_only(sb->s_bdev);
4514
4515 /*
4516 * Are we loading a blank journal or performing recovery after a
4517 * crash? For recovery, we need to check in advance whether we
4518 * can get read-write access to the device.
4519 */
Mingming Cao617ba132006-10-11 01:20:53 -07004520 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004521 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004522 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4523 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004524 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004525 ext4_msg(sb, KERN_ERR, "write access "
4526 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004527 return -EROFS;
4528 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04004529 ext4_msg(sb, KERN_INFO, "write access will "
4530 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004531 }
4532 }
4533
4534 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004535 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4536 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004537 return -EINVAL;
4538 }
4539
4540 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07004541 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004542 return -EINVAL;
4543 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07004544 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004545 return -EINVAL;
4546 }
4547
Theodore Ts'o90576c02009-09-29 15:51:30 -04004548 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04004549 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04004550
Mingming Cao617ba132006-10-11 01:20:53 -07004551 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
Mingming Caodab291a2006-10-11 01:21:01 -07004552 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004553 if (!err) {
4554 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4555 if (save)
4556 memcpy(save, ((char *) es) +
4557 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07004558 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004559 if (save)
4560 memcpy(((char *) es) + EXT4_S_ERR_START,
4561 save, EXT4_S_ERR_LEN);
4562 kfree(save);
4563 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004564
4565 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004566 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07004567 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004568 return err;
4569 }
4570
Mingming Cao617ba132006-10-11 01:20:53 -07004571 EXT4_SB(sb)->s_journal = journal;
4572 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004573
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04004574 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004575 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4576 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004577
4578 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004579 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004580 }
4581
4582 return 0;
4583}
4584
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004585static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004586{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07004588 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08004589 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004590
Christoph Hellwig08439fe2015-04-02 23:56:32 -04004591 if (!sbh)
Takashi Satoc4be0c12009-01-09 16:40:58 -08004592 return error;
Theodore Ts'o914258b2008-10-06 21:35:40 -04004593 if (buffer_write_io_error(sbh)) {
4594 /*
4595 * Oh, dear. A previous attempt to write the
4596 * superblock failed. This could happen because the
4597 * USB device was yanked out. Or it could happen to
4598 * be a transient write error and maybe the block will
4599 * be remapped. Nothing we can do but to retry the
4600 * write and hope for the best.
4601 */
Eric Sandeenb31e1552009-06-04 17:36:36 -04004602 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4603 "superblock detected");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004604 clear_buffer_write_io_error(sbh);
4605 set_buffer_uptodate(sbh);
4606 }
Theodore Ts'o71290b32009-09-10 17:31:04 -04004607 /*
4608 * If the file system is mounted read-only, don't update the
4609 * superblock write time. This avoids updating the superblock
4610 * write time when we are mounting the root file system
4611 * read/only but we need to replay the journal; at that point,
4612 * for people who are east of GMT and who make their clock
4613 * tick in localtime for Windows bug-for-bug compatibility,
4614 * the clock is set in the future, and this will cause e2fsck
4615 * to complain and force a full file system check.
4616 */
4617 if (!(sb->s_flags & MS_RDONLY))
4618 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004619 if (sb->s_bdev->bd_part)
4620 es->s_kbytes_written =
4621 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05004622 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4623 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004624 else
4625 es->s_kbytes_written =
4626 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004627 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4628 ext4_free_blocks_count_set(es,
Theodore Ts'o57042652011-09-09 18:56:51 -04004629 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4630 &EXT4_SB(sb)->s_freeclusters_counter)));
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004631 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4632 es->s_free_inodes_count =
4633 cpu_to_le32(percpu_counter_sum_positive(
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004634 &EXT4_SB(sb)->s_freeinodes_counter));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004635 BUFFER_TRACE(sbh, "marking dirty");
Theodore Ts'o06db49e2012-10-10 01:06:58 -04004636 ext4_superblock_csum_set(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004637 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04004638 if (sync) {
Takashi Satoc4be0c12009-01-09 16:40:58 -08004639 error = sync_dirty_buffer(sbh);
4640 if (error)
4641 return error;
4642
4643 error = buffer_write_io_error(sbh);
4644 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004645 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4646 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004647 clear_buffer_write_io_error(sbh);
4648 set_buffer_uptodate(sbh);
4649 }
4650 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08004651 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004652}
4653
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004654/*
4655 * Have we just finished recovery? If so, and if we are mounting (or
4656 * remounting) the filesystem readonly, then we will end up with a
4657 * consistent fs on disk. Record that fact.
4658 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004659static void ext4_mark_recovery_complete(struct super_block *sb,
4660 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004661{
Mingming Cao617ba132006-10-11 01:20:53 -07004662 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004663
Frank Mayhar03901312009-01-07 00:06:22 -05004664 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4665 BUG_ON(journal != NULL);
4666 return;
4667 }
Mingming Caodab291a2006-10-11 01:21:01 -07004668 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004669 if (jbd2_journal_flush(journal) < 0)
4670 goto out;
4671
Mingming Cao617ba132006-10-11 01:20:53 -07004672 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004673 sb->s_flags & MS_RDONLY) {
Mingming Cao617ba132006-10-11 01:20:53 -07004674 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004675 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004676 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004677
4678out:
Mingming Caodab291a2006-10-11 01:21:01 -07004679 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004680}
4681
4682/*
4683 * If we are mounting (or read-write remounting) a filesystem whose journal
4684 * has recorded an error from a previous lifetime, move that error to the
4685 * main filesystem now.
4686 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004687static void ext4_clear_journal_err(struct super_block *sb,
4688 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004689{
4690 journal_t *journal;
4691 int j_errno;
4692 const char *errstr;
4693
Frank Mayhar03901312009-01-07 00:06:22 -05004694 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4695
Mingming Cao617ba132006-10-11 01:20:53 -07004696 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004697
4698 /*
4699 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004700 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004701 */
4702
Mingming Caodab291a2006-10-11 01:21:01 -07004703 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004704 if (j_errno) {
4705 char nbuf[16];
4706
Mingming Cao617ba132006-10-11 01:20:53 -07004707 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004708 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004709 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004710 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004711
Mingming Cao617ba132006-10-11 01:20:53 -07004712 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4713 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004714 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004715
Mingming Caodab291a2006-10-11 01:21:01 -07004716 jbd2_journal_clear_err(journal);
Theodore Ts'od796c522012-08-05 19:04:57 -04004717 jbd2_journal_update_sb_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004718 }
4719}
4720
4721/*
4722 * Force the running and committing transactions to commit,
4723 * and wait on the commit.
4724 */
Mingming Cao617ba132006-10-11 01:20:53 -07004725int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004726{
4727 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004728
4729 if (sb->s_flags & MS_RDONLY)
4730 return 0;
4731
Mingming Cao617ba132006-10-11 01:20:53 -07004732 journal = EXT4_SB(sb)->s_journal;
Guo Chaob1deefc2013-01-28 21:41:02 -05004733 return ext4_journal_force_commit(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004734}
4735
Mingming Cao617ba132006-10-11 01:20:53 -07004736static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004737{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004738 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004739 tid_t target;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004740 bool needs_barrier = false;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004741 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004742
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004743 trace_ext4_sync_fs(sb, wait);
Jan Kara2e8fa542013-06-04 14:21:02 -04004744 flush_workqueue(sbi->rsv_conversion_wq);
Jan Karaa1177822012-07-03 16:45:29 +02004745 /*
4746 * Writeback quota in non-journalled quota case - journalled quota has
4747 * no dirty dquots
4748 */
4749 dquot_writeback_dquots(sb, -1);
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004750 /*
4751 * Data writeback is possible w/o journal transaction, so barrier must
4752 * being sent at the end of the function. But we can skip it if
4753 * transaction_commit will do it for us.
4754 */
Theodore Ts'obda32532014-09-18 16:12:37 -04004755 if (sbi->s_journal) {
4756 target = jbd2_get_latest_transaction(sbi->s_journal);
4757 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4758 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4759 needs_barrier = true;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004760
Theodore Ts'obda32532014-09-18 16:12:37 -04004761 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4762 if (wait)
4763 ret = jbd2_log_wait_commit(sbi->s_journal,
4764 target);
4765 }
4766 } else if (wait && test_opt(sb, BARRIER))
4767 needs_barrier = true;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004768 if (needs_barrier) {
4769 int err;
4770 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4771 if (!ret)
4772 ret = err;
4773 }
4774
4775 return ret;
4776}
4777
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004778/*
4779 * LVM calls this function before a (read-only) snapshot is created. This
4780 * gives us a chance to flush the journal completely and mark the fs clean.
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -04004781 *
4782 * Note that only this function cannot bring a filesystem to be in a clean
Jan Kara8e8ad8a2012-06-12 16:20:38 +02004783 * state independently. It relies on upper layer to stop all data & metadata
4784 * modifications.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004785 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004786static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004787{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004788 int error = 0;
4789 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004790
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004791 if (sb->s_flags & MS_RDONLY)
4792 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004793
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004794 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004795
Theodore Ts'obb044572014-09-18 17:12:02 -04004796 if (journal) {
4797 /* Now we set up the journal barrier. */
4798 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004799
Theodore Ts'obb044572014-09-18 17:12:02 -04004800 /*
4801 * Don't clear the needs_recovery flag if we failed to
4802 * flush the journal.
4803 */
4804 error = jbd2_journal_flush(journal);
4805 if (error < 0)
4806 goto out;
4807 }
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004808
4809 /* Journal blocked and flushed, clear needs_recovery flag. */
4810 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4811 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004812out:
Theodore Ts'obb044572014-09-18 17:12:02 -04004813 if (journal)
4814 /* we rely on upper layer to stop further updates */
4815 jbd2_journal_unlock_updates(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004816 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004817}
4818
4819/*
4820 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4821 * flag here, even though the filesystem is not technically dirty yet.
4822 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004823static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004824{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004825 if (sb->s_flags & MS_RDONLY)
4826 return 0;
4827
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004828 /* Reset the needs_recovery flag before the fs is unlocked. */
4829 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4830 ext4_commit_super(sb, 1);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004831 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004832}
4833
Theodore Ts'o673c6102010-12-15 20:28:48 -05004834/*
4835 * Structure to save mount options for ext4_remount's benefit
4836 */
4837struct ext4_mount_options {
4838 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004839 unsigned long s_mount_opt2;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08004840 kuid_t s_resuid;
4841 kgid_t s_resgid;
Theodore Ts'o673c6102010-12-15 20:28:48 -05004842 unsigned long s_commit_interval;
4843 u32 s_min_batch_time, s_max_batch_time;
4844#ifdef CONFIG_QUOTA
4845 int s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04004846 char *s_qf_names[EXT4_MAXQUOTAS];
Theodore Ts'o673c6102010-12-15 20:28:48 -05004847#endif
4848};
4849
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004850static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004851{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004852 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004853 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004854 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004855 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004856 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004857 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004858 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004859 int err = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004860#ifdef CONFIG_QUOTA
Chen Gang03dafb52013-01-24 23:24:58 -05004861 int i, j;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004862#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004863 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004864
4865 /* Store the original options */
4866 old_sb_flags = sb->s_flags;
4867 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004868 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004869 old_opts.s_resuid = sbi->s_resuid;
4870 old_opts.s_resgid = sbi->s_resgid;
4871 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004872 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4873 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004874#ifdef CONFIG_QUOTA
4875 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04004876 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05004877 if (sbi->s_qf_names[i]) {
4878 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4879 GFP_KERNEL);
4880 if (!old_opts.s_qf_names[i]) {
4881 for (j = 0; j < i; j++)
4882 kfree(old_opts.s_qf_names[j]);
Wei Yongjun3e36a162013-03-02 17:13:55 -05004883 kfree(orig_data);
Chen Gang03dafb52013-01-24 23:24:58 -05004884 return -ENOMEM;
4885 }
4886 } else
4887 old_opts.s_qf_names[i] = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004888#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004889 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4890 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004891
Eric Sandeen661aa522012-02-20 17:53:04 -05004892 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004893 err = -EINVAL;
4894 goto restore_opts;
4895 }
4896
Darrick J. Wong6b992ff2014-10-30 10:53:16 -04004897 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4898 test_opt(sb, JOURNAL_CHECKSUM)) {
4899 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
Eric Sandeen2d5b86e2015-02-12 23:07:37 -05004900 "during remount not supported; ignoring");
4901 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05004902 }
4903
Piotr Sarna6ae65142013-08-08 23:02:24 -04004904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4905 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4906 ext4_msg(sb, KERN_ERR, "can't mount with "
4907 "both data=journal and delalloc");
4908 err = -EINVAL;
4909 goto restore_opts;
4910 }
4911 if (test_opt(sb, DIOREAD_NOLOCK)) {
4912 ext4_msg(sb, KERN_ERR, "can't mount with "
4913 "both data=journal and dioread_nolock");
4914 err = -EINVAL;
4915 goto restore_opts;
4916 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08004917 if (test_opt(sb, DAX)) {
4918 ext4_msg(sb, KERN_ERR, "can't mount with "
4919 "both data=journal and dax");
4920 err = -EINVAL;
4921 goto restore_opts;
4922 }
4923 }
4924
4925 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4926 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4927 "dax flag with busy inodes while remounting");
4928 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
Piotr Sarna6ae65142013-08-08 23:02:24 -04004929 }
4930
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004931 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004932 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004933
4934 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004935 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004936
4937 es = sbi->s_es;
4938
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004939 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004940 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004941 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4942 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004943
Eric Sandeen661aa522012-02-20 17:53:04 -05004944 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004945 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004946 err = -EROFS;
4947 goto restore_opts;
4948 }
4949
4950 if (*flags & MS_RDONLY) {
Theodore Ts'o38c03b32014-03-13 22:49:42 -04004951 err = sync_filesystem(sb);
4952 if (err < 0)
4953 goto restore_opts;
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004954 err = dquot_suspend(sb, -1);
4955 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004956 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004957
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004958 /*
4959 * First of all, the unconditional stuff we have to do
4960 * to disable replay of the journal when we next remount
4961 */
4962 sb->s_flags |= MS_RDONLY;
4963
4964 /*
4965 * OK, test if we are remounting a valid rw partition
4966 * readonly, and if so set the rdonly flag and then
4967 * mark the partition as valid again.
4968 */
Mingming Cao617ba132006-10-11 01:20:53 -07004969 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4970 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004971 es->s_state = cpu_to_le16(sbi->s_mount_state);
4972
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004973 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004974 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004975 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004976 /* Make sure we can mount this feature set readwrite */
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05004977 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4978 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4979 !ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004980 err = -EROFS;
4981 goto restore_opts;
4982 }
Eric Sandeenead65962007-02-10 01:46:08 -08004983 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004984 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004985 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004986 */
4987 for (g = 0; g < sbi->s_groups_count; g++) {
4988 struct ext4_group_desc *gdp =
4989 ext4_get_group_desc(sb, g, NULL);
4990
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04004991 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004992 ext4_msg(sb, KERN_ERR,
4993 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Theodore Ts'o8a266462008-07-26 14:34:21 -04004994 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4995 le16_to_cpu(gdp->bg_checksum));
4996 err = -EINVAL;
4997 goto restore_opts;
4998 }
4999 }
5000
5001 /*
Eric Sandeenead65962007-02-10 01:46:08 -08005002 * If we have an unprocessed orphan list hanging
5003 * around from a previously readonly bdev mount,
5004 * require a full umount/remount for now.
5005 */
5006 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005007 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08005008 "remount RDWR because of unprocessed "
5009 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04005010 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08005011 err = -EINVAL;
5012 goto restore_opts;
5013 }
5014
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005015 /*
5016 * Mounting a RDONLY partition read-write, so reread
5017 * and store the current valid flag. (It may have
5018 * been changed by e2fsck since we originally mounted
5019 * the partition.)
5020 */
Frank Mayhar03901312009-01-07 00:06:22 -05005021 if (sbi->s_journal)
5022 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005023 sbi->s_mount_state = le16_to_cpu(es->s_state);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005024 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005025 sb->s_flags &= ~MS_RDONLY;
Johann Lombardic5e06d12011-05-24 18:31:25 -04005026 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5027 EXT4_FEATURE_INCOMPAT_MMP))
5028 if (ext4_multi_mount_protect(sb,
5029 le64_to_cpu(es->s_mmp_block))) {
5030 err = -EROFS;
5031 goto restore_opts;
5032 }
Christoph Hellwigc79d9672010-05-19 07:16:40 -04005033 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005034 }
5035 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04005036
5037 /*
5038 * Reinitialize lazy itable initialization thread based on
5039 * current settings
5040 */
5041 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5042 ext4_unregister_li_request(sb);
5043 else {
5044 ext4_group_t first_not_zeroed;
5045 first_not_zeroed = ext4_has_uninit_itable(sb);
5046 ext4_register_li_request(sb, first_not_zeroed);
5047 }
5048
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005049 ext4_setup_system_zone(sb);
Michael Tokarevd096ad02012-12-25 14:08:16 -05005050 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
Theodore Ts'oe2d67052009-05-01 00:33:44 -04005051 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05005052
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005053#ifdef CONFIG_QUOTA
5054 /* Release old quota file names */
Jan Karaa2d4a642014-09-11 11:15:15 -04005055 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05005056 kfree(old_opts.s_qf_names[i]);
Aditya Kali7c319d32012-07-22 20:21:31 -04005057 if (enable_quota) {
5058 if (sb_any_quota_suspended(sb))
5059 dquot_resume(sb, -1);
5060 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5061 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5062 err = ext4_enable_quotas(sb);
Theodore Ts'o07724f92012-08-17 19:08:42 -04005063 if (err)
Aditya Kali7c319d32012-07-22 20:21:31 -04005064 goto restore_opts;
Aditya Kali7c319d32012-07-22 20:21:31 -04005065 }
5066 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005067#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005068
Theodore Ts'oa26f4992015-02-02 00:37:02 -05005069 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005070 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5071 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005072 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005073
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005074restore_opts:
5075 sb->s_flags = old_sb_flags;
5076 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05005077 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005078 sbi->s_resuid = old_opts.s_resuid;
5079 sbi->s_resgid = old_opts.s_resgid;
5080 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05005081 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5082 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005083#ifdef CONFIG_QUOTA
5084 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04005085 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Chen Gang03dafb52013-01-24 23:24:58 -05005086 kfree(sbi->s_qf_names[i]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005087 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5088 }
5089#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005090 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005091 return err;
5092}
5093
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005094static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005095{
5096 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07005097 struct ext4_sb_info *sbi = EXT4_SB(sb);
5098 struct ext4_super_block *es = sbi->s_es;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005099 ext4_fsblk_t overhead = 0, resv_blocks;
Pekka Enberg960cc392006-12-06 20:35:29 -08005100 u64 fsid;
Kazuya Miod02a9392011-05-24 18:30:07 -04005101 s64 bfree;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005102 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005103
Theodore Ts'o952fc182012-07-09 16:27:05 -04005104 if (!test_opt(sb, MINIX_DF))
5105 overhead = sbi->s_overhead;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005106
Mingming Cao617ba132006-10-11 01:20:53 -07005107 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005108 buf->f_bsize = sb->s_blocksize;
Eric Sandeenb72f78c2012-11-08 10:33:36 -05005109 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
Theodore Ts'o57042652011-09-09 18:56:51 -04005110 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5111 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
Kazuya Miod02a9392011-05-24 18:30:07 -04005112 /* prevent underflow in case that few free space is available */
Theodore Ts'o57042652011-09-09 18:56:51 -04005113 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
Lukas Czerner27dd4382013-04-09 22:11:22 -04005114 buf->f_bavail = buf->f_bfree -
5115 (ext4_r_blocks_count(es) + resv_blocks);
5116 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005117 buf->f_bavail = 0;
5118 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07005119 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07005120 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08005121 fsid = le64_to_cpup((void *)es->s_uuid) ^
5122 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5123 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5124 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005125
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005126 return 0;
5127}
5128
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005129/* Helper function for writing quotas on sync - we need to start transaction
5130 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005131 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07005132 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01005133 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05005134 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07005135 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005136 *
5137 */
5138
5139#ifdef CONFIG_QUOTA
5140
5141static inline struct inode *dquot_to_inode(struct dquot *dquot)
5142{
Eric W. Biederman4c376dc2012-09-16 03:56:19 -07005143 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005144}
5145
Mingming Cao617ba132006-10-11 01:20:53 -07005146static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005147{
5148 int ret, err;
5149 handle_t *handle;
5150 struct inode *inode;
5151
5152 inode = dquot_to_inode(dquot);
Theodore Ts'o9924a922013-02-08 21:59:22 -05005153 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005154 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005155 if (IS_ERR(handle))
5156 return PTR_ERR(handle);
5157 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005158 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005159 if (!ret)
5160 ret = err;
5161 return ret;
5162}
5163
Mingming Cao617ba132006-10-11 01:20:53 -07005164static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005165{
5166 int ret, err;
5167 handle_t *handle;
5168
Theodore Ts'o9924a922013-02-08 21:59:22 -05005169 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005170 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005171 if (IS_ERR(handle))
5172 return PTR_ERR(handle);
5173 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005174 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005175 if (!ret)
5176 ret = err;
5177 return ret;
5178}
5179
Mingming Cao617ba132006-10-11 01:20:53 -07005180static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005181{
5182 int ret, err;
5183 handle_t *handle;
5184
Theodore Ts'o9924a922013-02-08 21:59:22 -05005185 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005186 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07005187 if (IS_ERR(handle)) {
5188 /* Release dquot anyway to avoid endless cycle in dqput() */
5189 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005190 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07005191 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005192 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005193 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005194 if (!ret)
5195 ret = err;
5196 return ret;
5197}
5198
Mingming Cao617ba132006-10-11 01:20:53 -07005199static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005200{
Jan Kara262b4662013-03-02 17:57:08 -05005201 struct super_block *sb = dquot->dq_sb;
5202 struct ext4_sb_info *sbi = EXT4_SB(sb);
5203
Jan Kara2c8be6b2008-05-13 21:27:55 -04005204 /* Are we journaling quotas? */
Jan Kara262b4662013-03-02 17:57:08 -05005205 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5206 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005207 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005208 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005209 } else {
5210 return dquot_mark_dquot_dirty(dquot);
5211 }
5212}
5213
Mingming Cao617ba132006-10-11 01:20:53 -07005214static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005215{
5216 int ret, err;
5217 handle_t *handle;
5218
5219 /* Data block + inode block */
David Howells2b0143b2015-03-17 22:25:59 +00005220 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005221 if (IS_ERR(handle))
5222 return PTR_ERR(handle);
5223 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07005224 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005225 if (!ret)
5226 ret = err;
5227 return ret;
5228}
5229
5230/*
5231 * Turn on quotas during mount time - we need to find
5232 * the quota file and such...
5233 */
Mingming Cao617ba132006-10-11 01:20:53 -07005234static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005235{
Christoph Hellwig287a8092010-05-19 07:16:45 -04005236 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5237 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005238}
5239
5240/*
5241 * Standard function to be called on quota_on
5242 */
Mingming Cao617ba132006-10-11 01:20:53 -07005243static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02005244 struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005245{
5246 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005247
5248 if (!test_opt(sb, QUOTA))
5249 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04005250
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005251 /* Quotafile not on the same filesystem? */
Al Virod8c95842011-12-07 18:16:57 -05005252 if (path->dentry->d_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005253 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04005254 /* Journaling quota? */
5255 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005256 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02005257 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04005258 ext4_msg(sb, KERN_WARNING,
5259 "Quota file not on filesystem root. "
5260 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005261 }
Jan Kara06235432008-05-13 19:11:51 -04005262
5263 /*
5264 * When we journal data on quota file, we have to flush journal to see
5265 * all updates to the file when we bypass pagecache...
5266 */
Frank Mayhar03901312009-01-07 00:06:22 -05005267 if (EXT4_SB(sb)->s_journal &&
David Howells2b0143b2015-03-17 22:25:59 +00005268 ext4_should_journal_data(d_inode(path->dentry))) {
Jan Kara06235432008-05-13 19:11:51 -04005269 /*
5270 * We don't need to lock updates but journal_flush() could
5271 * otherwise be livelocked...
5272 */
5273 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005274 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04005275 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02005276 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005277 return err;
Jan Kara06235432008-05-13 19:11:51 -04005278 }
5279
Jan Karaf00c9e42010-09-15 17:38:58 +02005280 return dquot_quota_on(sb, type, format_id, path);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005281}
5282
Aditya Kali7c319d32012-07-22 20:21:31 -04005283static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5284 unsigned int flags)
5285{
5286 int err;
5287 struct inode *qf_inode;
Jan Karaa2d4a642014-09-11 11:15:15 -04005288 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
Aditya Kali7c319d32012-07-22 20:21:31 -04005289 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5290 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5291 };
5292
5293 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5294
5295 if (!qf_inums[type])
5296 return -EPERM;
5297
5298 qf_inode = ext4_iget(sb, qf_inums[type]);
5299 if (IS_ERR(qf_inode)) {
5300 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5301 return PTR_ERR(qf_inode);
5302 }
5303
Jan Karabcb13852013-04-09 09:21:41 -04005304 /* Don't account quota for quota files to avoid recursion */
5305 qf_inode->i_flags |= S_NOQUOTA;
Aditya Kali7c319d32012-07-22 20:21:31 -04005306 err = dquot_enable(qf_inode, type, format_id, flags);
5307 iput(qf_inode);
5308
5309 return err;
5310}
5311
5312/* Enable usage tracking for all quota types. */
5313static int ext4_enable_quotas(struct super_block *sb)
5314{
5315 int type, err = 0;
Jan Karaa2d4a642014-09-11 11:15:15 -04005316 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
Aditya Kali7c319d32012-07-22 20:21:31 -04005317 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5318 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5319 };
5320
5321 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
Jan Karaa2d4a642014-09-11 11:15:15 -04005322 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
Aditya Kali7c319d32012-07-22 20:21:31 -04005323 if (qf_inums[type]) {
5324 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5325 DQUOT_USAGE_ENABLED);
5326 if (err) {
5327 ext4_warning(sb,
Theodore Ts'o72ba7452013-01-24 23:24:54 -05005328 "Failed to enable quota tracking "
5329 "(type=%d, err=%d). Please run "
5330 "e2fsck to fix.", type, err);
Aditya Kali7c319d32012-07-22 20:21:31 -04005331 return err;
5332 }
5333 }
5334 }
5335 return 0;
5336}
5337
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005338static int ext4_quota_off(struct super_block *sb, int type)
5339{
Jan Kara21f97692011-04-04 15:33:39 -04005340 struct inode *inode = sb_dqopt(sb)->files[type];
5341 handle_t *handle;
5342
Dmitry Monakhov87009d82010-11-08 13:47:33 -05005343 /* Force all delayed allocation blocks to be allocated.
5344 * Caller already holds s_umount sem */
5345 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005346 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005347
Amir Goldstein0b268592011-05-16 09:59:13 -04005348 if (!inode)
5349 goto out;
5350
Jan Kara21f97692011-04-04 15:33:39 -04005351 /* Update modification times of quota files when userspace can
5352 * start looking at them */
Theodore Ts'o9924a922013-02-08 21:59:22 -05005353 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
Jan Kara21f97692011-04-04 15:33:39 -04005354 if (IS_ERR(handle))
5355 goto out;
5356 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5357 ext4_mark_inode_dirty(handle, inode);
5358 ext4_journal_stop(handle);
5359
5360out:
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005361 return dquot_quota_off(sb, type);
5362}
5363
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005364/* Read data from quotafile - avoid pagecache and such because we cannot afford
5365 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03005366 * itself serializes the operations (and no one else should touch the files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005367 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07005368static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005369 size_t len, loff_t off)
5370{
5371 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005372 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005373 int offset = off & (sb->s_blocksize - 1);
5374 int tocopy;
5375 size_t toread;
5376 struct buffer_head *bh;
5377 loff_t i_size = i_size_read(inode);
5378
5379 if (off > i_size)
5380 return 0;
5381 if (off+len > i_size)
5382 len = i_size-off;
5383 toread = len;
5384 while (toread > 0) {
5385 tocopy = sb->s_blocksize - offset < toread ?
5386 sb->s_blocksize - offset : toread;
Theodore Ts'o1c215022014-08-29 20:52:15 -04005387 bh = ext4_bread(NULL, inode, blk, 0);
5388 if (IS_ERR(bh))
5389 return PTR_ERR(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005390 if (!bh) /* A hole? */
5391 memset(data, 0, tocopy);
5392 else
5393 memcpy(data, bh->b_data+offset, tocopy);
5394 brelse(bh);
5395 offset = 0;
5396 toread -= tocopy;
5397 data += tocopy;
5398 blk++;
5399 }
5400 return len;
5401}
5402
5403/* Write to quotafile (we know the transaction is already started and has
5404 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07005405static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005406 const char *data, size_t len, loff_t off)
5407{
5408 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005409 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Theodore Ts'o1c215022014-08-29 20:52:15 -04005410 int err, offset = off & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005411 struct buffer_head *bh;
5412 handle_t *handle = journal_current_handle();
5413
Frank Mayhar03901312009-01-07 00:06:22 -05005414 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005415 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5416 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07005417 (unsigned long long)off, (unsigned long long)len);
5418 return -EIO;
5419 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005420 /*
5421 * Since we account only one data block in transaction credits,
5422 * then it is impossible to cross a block boundary.
5423 */
5424 if (sb->s_blocksize - offset < len) {
5425 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5426 " cancelled because not block aligned",
5427 (unsigned long long)off, (unsigned long long)len);
5428 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005429 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005430
Theodore Ts'o1c215022014-08-29 20:52:15 -04005431 bh = ext4_bread(handle, inode, blk, 1);
5432 if (IS_ERR(bh))
5433 return PTR_ERR(bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005434 if (!bh)
5435 goto out;
liang xie5d601252014-05-12 22:06:43 -04005436 BUFFER_TRACE(bh, "get write access");
Jan Kara62d2b5f2010-07-27 11:56:07 -04005437 err = ext4_journal_get_write_access(handle, bh);
5438 if (err) {
5439 brelse(bh);
Theodore Ts'o1c215022014-08-29 20:52:15 -04005440 return err;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005441 }
5442 lock_buffer(bh);
5443 memcpy(bh->b_data+offset, data, len);
5444 flush_dcache_page(bh->b_page);
5445 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04005446 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005447 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005448out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005449 if (inode->i_size < off + len) {
5450 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07005451 EXT4_I(inode)->i_disksize = inode->i_size;
Jan Kara21f97692011-04-04 15:33:39 -04005452 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005453 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005454 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005455}
5456
5457#endif
5458
Al Viro152a0832010-07-25 00:46:55 +04005459static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5460 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005461{
Al Viro152a0832010-07-25 00:46:55 +04005462 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005463}
5464
Theodore Ts'o37f328e2010-03-24 20:06:41 -04005465#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05005466static inline void register_as_ext2(void)
5467{
5468 int err = register_filesystem(&ext2_fs_type);
5469 if (err)
5470 printk(KERN_WARNING
5471 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5472}
5473
5474static inline void unregister_as_ext2(void)
5475{
5476 unregister_filesystem(&ext2_fs_type);
5477}
Theodore Ts'o2035e772011-04-18 17:29:14 -04005478
5479static inline int ext2_feature_set_ok(struct super_block *sb)
5480{
5481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5482 return 0;
5483 if (sb->s_flags & MS_RDONLY)
5484 return 1;
5485 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5486 return 0;
5487 return 1;
5488}
Theodore Ts'o24b58422009-12-07 14:08:51 -05005489#else
5490static inline void register_as_ext2(void) { }
5491static inline void unregister_as_ext2(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04005492static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05005493#endif
5494
Theodore Ts'o37f328e2010-03-24 20:06:41 -04005495#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
Theodore Ts'o24b58422009-12-07 14:08:51 -05005496static inline void register_as_ext3(void)
5497{
5498 int err = register_filesystem(&ext3_fs_type);
5499 if (err)
5500 printk(KERN_WARNING
5501 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5502}
5503
5504static inline void unregister_as_ext3(void)
5505{
5506 unregister_filesystem(&ext3_fs_type);
5507}
Theodore Ts'o2035e772011-04-18 17:29:14 -04005508
5509static inline int ext3_feature_set_ok(struct super_block *sb)
5510{
5511 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5512 return 0;
5513 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5514 return 0;
5515 if (sb->s_flags & MS_RDONLY)
5516 return 1;
5517 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5518 return 0;
5519 return 1;
5520}
Theodore Ts'o24b58422009-12-07 14:08:51 -05005521#else
5522static inline void register_as_ext3(void) { }
5523static inline void unregister_as_ext3(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04005524static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05005525#endif
5526
Theodore Ts'o03010a32008-10-10 20:02:48 -04005527static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005528 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04005529 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04005530 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005531 .kill_sb = kill_block_super,
5532 .fs_flags = FS_REQUIRES_DEV,
5533};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08005534MODULE_ALIAS_FS("ext4");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005535
Lukas Czerner8f021222011-02-03 14:33:33 -05005536static int __init ext4_init_feat_adverts(void)
Lukas Czerner857ac882010-10-27 21:30:05 -04005537{
5538 struct ext4_features *ef;
5539 int ret = -ENOMEM;
5540
5541 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5542 if (!ef)
5543 goto out;
5544
5545 ef->f_kobj.kset = ext4_kset;
5546 init_completion(&ef->f_kobj_unregister);
5547 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5548 "features");
5549 if (ret) {
5550 kfree(ef);
5551 goto out;
5552 }
5553
5554 ext4_feat = ef;
5555 ret = 0;
5556out:
5557 return ret;
5558}
5559
Lukas Czerner8f021222011-02-03 14:33:33 -05005560static void ext4_exit_feat_adverts(void)
5561{
5562 kobject_put(&ext4_feat->f_kobj);
5563 wait_for_completion(&ext4_feat->f_kobj_unregister);
5564 kfree(ext4_feat);
5565}
5566
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005567/* Shared across all ext4 file systems */
5568wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5569struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5570
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005571static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005572{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005573 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05005574
Al Viro07c0c5d82012-03-20 22:05:02 -04005575 ext4_li_info = NULL;
5576 mutex_init(&ext4_li_mtx);
5577
Carlos Maiolino9a4c8012012-12-10 16:30:45 -05005578 /* Build-time check for flags consistency */
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04005579 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005580
5581 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5582 mutex_init(&ext4__aio_mutex[i]);
5583 init_waitqueue_head(&ext4__ioend_wq[i]);
5584 }
5585
Zheng Liu51865fd2012-11-08 21:57:32 -05005586 err = ext4_init_es();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005587 if (err)
5588 return err;
Zheng Liu51865fd2012-11-08 21:57:32 -05005589
5590 err = ext4_init_pageio();
5591 if (err)
5592 goto out7;
5593
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005594 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04005595 if (err)
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005596 goto out6;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04005597 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
Theodore Ts'o0e376b12012-08-17 10:04:17 -04005598 if (!ext4_kset) {
5599 err = -ENOMEM;
Theodore Ts'odd683142011-02-03 14:33:49 -05005600 goto out5;
Theodore Ts'o0e376b12012-08-17 10:04:17 -04005601 }
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005602 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
Lukas Czerner857ac882010-10-27 21:30:05 -04005603
5604 err = ext4_init_feat_adverts();
Theodore Ts'odd683142011-02-03 14:33:49 -05005605 if (err)
5606 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04005607
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005608 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005609 if (err)
Alex Tomasc9de5602008-01-29 00:19:52 -05005610 goto out2;
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005611 else
5612 ext4_mballoc_ready = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005613 err = init_inodecache();
5614 if (err)
5615 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05005616 register_as_ext3();
Theodore Ts'o2035e772011-04-18 17:29:14 -04005617 register_as_ext2();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005618 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005619 if (err)
5620 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04005621
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005622 return 0;
5623out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05005624 unregister_as_ext2();
5625 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005626 destroy_inodecache();
5627out1:
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005628 ext4_mballoc_ready = 0;
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005629 ext4_exit_mballoc();
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005630out2:
Lukas Czerner8f021222011-02-03 14:33:33 -05005631 ext4_exit_feat_adverts();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005632out4:
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005633 if (ext4_proc_root)
5634 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'odd683142011-02-03 14:33:49 -05005635 kset_unregister(ext4_kset);
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005636out5:
Theodore Ts'odd683142011-02-03 14:33:49 -05005637 ext4_exit_system_zone();
Fabrice Jouhaudd44651d2011-10-08 16:26:03 -04005638out6:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005639 ext4_exit_pageio();
Zheng Liu51865fd2012-11-08 21:57:32 -05005640out7:
5641 ext4_exit_es();
5642
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005643 return err;
5644}
5645
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005646static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005647{
Lukas Czernerbfff6872010-10-27 21:30:05 -04005648 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05005649 unregister_as_ext2();
5650 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005651 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005652 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005653 ext4_exit_mballoc();
Lukas Czerner8f021222011-02-03 14:33:33 -05005654 ext4_exit_feat_adverts();
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04005655 remove_proc_entry("fs/ext4", NULL);
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04005656 kset_unregister(ext4_kset);
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005657 ext4_exit_system_zone();
5658 ext4_exit_pageio();
Eric Sandeendd12ed12013-07-26 15:21:11 -04005659 ext4_exit_es();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005660}
5661
5662MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05005663MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005664MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005665module_init(ext4_init_fs)
5666module_exit(ext4_exit_fs)