blob: 1d82b56d9b11f362b35582ab0f804a5ca44f0189 [file] [log] [blame]
Theodore Ts'of5166762017-12-17 22:00:59 -05001// SPDX-License-Identifier: GPL-2.0
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002/*
Mingming Cao617ba132006-10-11 01:20:53 -07003 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040024#include <linux/vmalloc.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070025#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
Tejun Heo66114ca2015-05-22 17:13:32 -040028#include <linux/backing-dev.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include <linux/parser.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070030#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070031#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070032#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040038#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040039#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040040#include <linux/crc16.h>
Dan Williamsef510422017-05-08 10:55:27 -070041#include <linux/dax.h>
Dan Magenheimer7abc52c2011-05-26 10:02:03 -060042#include <linux/cleancache.h>
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080043#include <linux/uaccess.h>
Jeff Laytonee73f9a2018-01-09 08:21:39 -050044#include <linux/iversion.h>
Gabriel Krisman Bertazic83ad552019-04-25 14:05:42 -040045#include <linux/unicode.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070046
Lukas Czernerbfff6872010-10-27 21:30:05 -040047#include <linux/kthread.h>
48#include <linux/freezer.h>
49
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040050#include "ext4.h"
Theodore Ts'o4a092d72012-11-28 13:03:30 -050051#include "ext4_extents.h" /* Needed for trace points definition */
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040052#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070053#include "xattr.h"
54#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040055#include "mballoc.h"
Darrick J. Wong0c9ec4b2017-04-30 00:36:53 -040056#include "fsmap.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070057
Theodore Ts'o9bffad12009-06-17 11:48:11 -040058#define CREATE_TRACE_POINTS
59#include <trace/events/ext4.h>
60
Lukas Czerner0b75a842011-02-23 12:22:49 -050061static struct ext4_lazy_init *ext4_li_info;
62static struct mutex ext4_li_mtx;
Theodore Ts'oe294a532015-08-15 14:59:44 -040063static struct ratelimit_state ext4_mount_msg_ratelimit;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040064
Mingming Cao617ba132006-10-11 01:20:53 -070065static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070066 unsigned long journal_devnum);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -050067static int ext4_show_options(struct seq_file *seq, struct dentry *root);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040068static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71static void ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070073static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040074static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080076static int ext4_unfreeze(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080077static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040078static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
Theodore Ts'o2035e772011-04-18 17:29:14 -040080static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
Amir Goldsteind39195c2011-02-28 00:53:45 -050082static int ext4_feature_set_ok(struct super_block *sb, int readonly);
Lukas Czernerbfff6872010-10-27 21:30:05 -040083static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050085static void ext4_clear_request_list(void);
Eric Whitneyc6cb7e72016-09-30 02:05:09 -040086static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070088
Jan Karae74031f2015-12-07 14:35:49 -050089/*
90 * Lock ordering
91 *
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 *
95 * page fault path:
96 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
98 *
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_sem
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * i_data_sem (rw)
103 *
104 * truncate:
Nikolay Borisov1d398342018-03-22 11:52:10 -0400105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
107 * i_data_sem (rw)
Jan Karae74031f2015-12-07 14:35:49 -0500108 *
109 * direct IO:
Nikolay Borisov1d398342018-03-22 11:52:10 -0400110 * sb_start_write -> i_mutex -> mmap_sem
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
Jan Karae74031f2015-12-07 14:35:49 -0500112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
Jan Karac290ea02015-06-18 16:52:29 +0200117#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
Theodore Ts'o2035e772011-04-18 17:29:14 -0400118static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
120 .name = "ext2",
121 .mount = ext4_mount,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
124};
Eric W. Biederman7f78e032013-03-02 19:39:14 -0800125MODULE_ALIAS_FS("ext2");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -0700126MODULE_ALIAS("ext2");
Theodore Ts'o2035e772011-04-18 17:29:14 -0400127#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128#else
129#define IS_EXT2_SB(sb) (0)
130#endif
131
132
Jan Karaba69f9a2010-03-24 20:18:37 -0400133static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
135 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +0400136 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -0400137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
139};
Eric W. Biederman7f78e032013-03-02 19:39:14 -0800140MODULE_ALIAS_FS("ext3");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -0700141MODULE_ALIAS("ext3");
Jan Karaba69f9a2010-03-24 20:18:37 -0400142#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700143
Theodore Ts'ofb265c9c2018-11-25 17:20:31 -0500144/*
145 * This works like sb_bread() except it uses ERR_PTR for error
146 * returns. Currently with sb_bread it's impossible to distinguish
147 * between ENOMEM and EIO situations (since both result in a NULL
148 * return.
149 */
150struct buffer_head *
151ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
152{
153 struct buffer_head *bh = sb_getblk(sb, block);
154
155 if (bh == NULL)
156 return ERR_PTR(-ENOMEM);
157 if (buffer_uptodate(bh))
158 return bh;
159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
160 wait_on_buffer(bh);
161 if (buffer_uptodate(bh))
162 return bh;
163 put_bh(bh);
164 return ERR_PTR(-EIO);
165}
166
Darrick J. Wongd25425f2012-04-29 18:25:10 -0400167static int ext4_verify_csum_type(struct super_block *sb,
168 struct ext4_super_block *es)
169{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400170 if (!ext4_has_feature_metadata_csum(sb))
Darrick J. Wongd25425f2012-04-29 18:25:10 -0400171 return 1;
172
173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
174}
175
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400176static __le32 ext4_superblock_csum(struct super_block *sb,
177 struct ext4_super_block *es)
178{
179 struct ext4_sb_info *sbi = EXT4_SB(sb);
180 int offset = offsetof(struct ext4_super_block, s_checksum);
181 __u32 csum;
182
183 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
184
185 return cpu_to_le32(csum);
186}
187
Stephen Hemmingerc1978552014-05-12 10:50:23 -0400188static int ext4_superblock_csum_verify(struct super_block *sb,
189 struct ext4_super_block *es)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400190{
Dmitry Monakhov9aa5d32b2014-10-13 03:36:16 -0400191 if (!ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400192 return 1;
193
194 return es->s_checksum == ext4_superblock_csum(sb, es);
195}
196
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400197void ext4_superblock_csum_set(struct super_block *sb)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400198{
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400199 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
200
Dmitry Monakhov9aa5d32b2014-10-13 03:36:16 -0400201 if (!ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400202 return;
203
204 es->s_checksum = ext4_superblock_csum(sb, es);
205}
206
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400207void *ext4_kvmalloc(size_t size, gfp_t flags)
208{
209 void *ret;
210
Joe Perches8be04b92013-06-19 12:15:53 -0700211 ret = kmalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400212 if (!ret)
213 ret = __vmalloc(size, flags, PAGE_KERNEL);
214 return ret;
215}
216
217void *ext4_kvzalloc(size_t size, gfp_t flags)
218{
219 void *ret;
220
Joe Perches8be04b92013-06-19 12:15:53 -0700221 ret = kzalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400222 if (!ret)
223 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
224 return ret;
225}
226
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700227ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
228 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700229{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400230 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400232 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700233}
234
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700235ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
236 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700237{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400238 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400240 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700241}
242
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700243ext4_fsblk_t ext4_inode_table(struct super_block *sb,
244 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700245{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400246 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400248 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700249}
250
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400251__u32 ext4_free_group_clusters(struct super_block *sb,
252 struct ext4_group_desc *bg)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500253{
254 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400256 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500257}
258
259__u32 ext4_free_inodes_count(struct super_block *sb,
260 struct ext4_group_desc *bg)
261{
262 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
263 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400264 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500265}
266
267__u32 ext4_used_dirs_count(struct super_block *sb,
268 struct ext4_group_desc *bg)
269{
270 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
271 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400272 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500273}
274
275__u32 ext4_itable_unused_count(struct super_block *sb,
276 struct ext4_group_desc *bg)
277{
278 return le16_to_cpu(bg->bg_itable_unused_lo) |
279 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400280 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500281}
282
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700283void ext4_block_bitmap_set(struct super_block *sb,
284 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700285{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400286 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700289}
290
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700291void ext4_inode_bitmap_set(struct super_block *sb,
292 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700293{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400294 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700297}
298
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700299void ext4_inode_table_set(struct super_block *sb,
300 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700301{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400302 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700305}
306
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400307void ext4_free_group_clusters_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500309{
310 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
313}
314
315void ext4_free_inodes_set(struct super_block *sb,
316 struct ext4_group_desc *bg, __u32 count)
317{
318 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
319 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
320 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
321}
322
323void ext4_used_dirs_set(struct super_block *sb,
324 struct ext4_group_desc *bg, __u32 count)
325{
326 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
327 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
328 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
329}
330
331void ext4_itable_unused_set(struct super_block *sb,
332 struct ext4_group_desc *bg, __u32 count)
333{
334 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
335 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
336 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
337}
338
Arnd Bergmann6a0678a2018-07-29 15:51:48 -0400339static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
340{
341 time64_t now = ktime_get_real_seconds();
342
343 now = clamp_val(now, 0, (1ull << 40) - 1);
344
345 *lo = cpu_to_le32(lower_32_bits(now));
346 *hi = upper_32_bits(now);
347}
348
349static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
350{
351 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
352}
353#define ext4_update_tstamp(es, tstamp) \
354 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
355#define ext4_get_tstamp(es, tstamp) \
356 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400357
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400358static void __save_error_info(struct super_block *sb, const char *func,
359 unsigned int line)
360{
361 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
362
363 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
Theodore Ts'o1b466172015-05-14 18:37:30 -0400364 if (bdev_read_only(sb->s_bdev))
365 return;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400366 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Arnd Bergmann6a0678a2018-07-29 15:51:48 -0400367 ext4_update_tstamp(es, s_last_error_time);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400368 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
369 es->s_last_error_line = cpu_to_le32(line);
370 if (!es->s_first_error_time) {
371 es->s_first_error_time = es->s_last_error_time;
Arnd Bergmann6a0678a2018-07-29 15:51:48 -0400372 es->s_first_error_time_hi = es->s_last_error_time_hi;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400373 strncpy(es->s_first_error_func, func,
374 sizeof(es->s_first_error_func));
375 es->s_first_error_line = cpu_to_le32(line);
376 es->s_first_error_ino = es->s_last_error_ino;
377 es->s_first_error_block = es->s_last_error_block;
378 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400379 /*
380 * Start the daily error reporting function if it hasn't been
381 * started already
382 */
383 if (!es->s_error_count)
384 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Wei Yongjunba39ebb2012-09-27 09:37:53 -0400385 le32_add_cpu(&es->s_error_count, 1);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400386}
387
388static void save_error_info(struct super_block *sb, const char *func,
389 unsigned int line)
390{
391 __save_error_info(sb, func, line);
392 ext4_commit_super(sb, 1);
393}
394
Theodore Ts'obdfe0cbd2015-08-16 10:03:57 -0400395/*
396 * The del_gendisk() function uninitializes the disk-specific data
397 * structures, including the bdi structure, without telling anyone
398 * else. Once this happens, any attempt to call mark_buffer_dirty()
399 * (for example, by ext4_commit_super), will cause a kernel OOPS.
400 * This is a kludge to prevent these oops until we can put in a proper
401 * hook in del_gendisk() to inform the VFS and file system layers.
402 */
403static int block_device_ejected(struct super_block *sb)
404{
405 struct inode *bd_inode = sb->s_bdev->bd_inode;
406 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
407
408 return bdi->dev == NULL;
409}
410
Bobi Jam18aadd42012-02-20 17:53:02 -0500411static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
412{
413 struct super_block *sb = journal->j_private;
414 struct ext4_sb_info *sbi = EXT4_SB(sb);
415 int error = is_journal_aborted(journal);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400416 struct ext4_journal_cb_entry *jce;
Bobi Jam18aadd42012-02-20 17:53:02 -0500417
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400418 BUG_ON(txn->t_state == T_FINISHED);
Daeho Jeonga0154342017-06-22 23:54:33 -0400419
420 ext4_process_freed_data(sb, txn->t_tid);
421
Bobi Jam18aadd42012-02-20 17:53:02 -0500422 spin_lock(&sbi->s_md_lock);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400423 while (!list_empty(&txn->t_private_list)) {
424 jce = list_entry(txn->t_private_list.next,
425 struct ext4_journal_cb_entry, jce_list);
Bobi Jam18aadd42012-02-20 17:53:02 -0500426 list_del_init(&jce->jce_list);
427 spin_unlock(&sbi->s_md_lock);
428 jce->jce_func(sb, jce, error);
429 spin_lock(&sbi->s_md_lock);
430 }
431 spin_unlock(&sbi->s_md_lock);
432}
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400433
Jan Kara1dc10972019-03-14 23:46:05 -0400434static bool system_going_down(void)
435{
436 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
437 || system_state == SYSTEM_RESTART;
438}
439
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700440/* Deal with the reporting of failure conditions on a filesystem such as
441 * inconsistencies detected or read IO failures.
442 *
443 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700444 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700445 * write ordering constraints on the superblock which prevent us from
446 * writing it out straight away; and given that the journal is about to
447 * be aborted, we can't rely on the current, or future, transactions to
448 * write out the superblock safely.
449 *
Mingming Caodab291a2006-10-11 01:21:01 -0700450 * 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 -0200451 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700452 * that error until we've noted it down and cleared it.
453 */
454
Mingming Cao617ba132006-10-11 01:20:53 -0700455static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700456{
Theodore Ts'o327eaf72018-06-12 23:34:57 -0400457 if (test_opt(sb, WARN_ON_ERROR))
458 WARN_ON_ONCE(1);
459
David Howellsbc98a422017-07-17 08:45:34 +0100460 if (sb_rdonly(sb))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700461 return;
462
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400463 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700464 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700465
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400466 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700467 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700468 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700469 }
Jan Kara1dc10972019-03-14 23:46:05 -0400470 /*
471 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
472 * could panic during 'reboot -f' as the underlying device got already
473 * disabled.
474 */
475 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400476 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400477 /*
478 * Make sure updated value of ->s_mount_flags will be visible
479 * before ->s_flags update
480 */
481 smp_wmb();
Linus Torvalds1751e8a2017-11-27 13:05:09 -0800482 sb->s_flags |= SB_RDONLY;
Jan Kara1dc10972019-03-14 23:46:05 -0400483 } else if (test_opt(sb, ERRORS_PANIC)) {
Daeho Jeong4327ba52015-10-18 17:02:56 -0400484 if (EXT4_SB(sb)->s_journal &&
485 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
486 return;
Mingming Cao617ba132006-10-11 01:20:53 -0700487 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700488 sb->s_id);
Daeho Jeong4327ba52015-10-18 17:02:56 -0400489 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700490}
491
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400492#define ext4_error_ratelimit(sb) \
493 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
494 "EXT4-fs error")
495
Eric Sandeen12062dd2010-02-15 14:19:27 -0500496void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400497 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700498{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500499 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700500 va_list args;
501
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500502 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
503 return;
504
Theodore Ts'occf0f322018-02-18 20:53:23 -0500505 trace_ext4_error(sb, function, line);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400506 if (ext4_error_ratelimit(sb)) {
507 va_start(args, fmt);
508 vaf.fmt = fmt;
509 vaf.va = &args;
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
512 sb->s_id, function, line, current->comm, &vaf);
513 va_end(args);
514 }
Theodore Ts'of3fc02102012-05-30 23:00:16 -0400515 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700516 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700517}
518
Joe Perchese7c96e82013-07-01 08:12:37 -0400519void __ext4_error_inode(struct inode *inode, const char *function,
520 unsigned int line, ext4_fsblk_t block,
521 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500522{
523 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500524 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400525 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500526
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500527 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
528 return;
529
Theodore Ts'occf0f322018-02-18 20:53:23 -0500530 trace_ext4_error(inode->i_sb, function, line);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400531 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
532 es->s_last_error_block = cpu_to_le64(block);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400533 if (ext4_error_ratelimit(inode->i_sb)) {
534 va_start(args, fmt);
535 vaf.fmt = fmt;
536 vaf.va = &args;
537 if (block)
538 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
539 "inode #%lu: block %llu: comm %s: %pV\n",
540 inode->i_sb->s_id, function, line, inode->i_ino,
541 block, current->comm, &vaf);
542 else
543 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
544 "inode #%lu: comm %s: %pV\n",
545 inode->i_sb->s_id, function, line, inode->i_ino,
546 current->comm, &vaf);
547 va_end(args);
548 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400549 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500550 ext4_handle_error(inode->i_sb);
551}
552
Joe Perchese7c96e82013-07-01 08:12:37 -0400553void __ext4_error_file(struct file *file, const char *function,
554 unsigned int line, ext4_fsblk_t block,
555 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500556{
557 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500558 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400559 struct ext4_super_block *es;
Al Viro496ad9a2013-01-23 17:07:38 -0500560 struct inode *inode = file_inode(file);
Frank Mayhar273df552010-03-02 11:46:09 -0500561 char pathname[80], *path;
562
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500563 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
564 return;
565
Theodore Ts'occf0f322018-02-18 20:53:23 -0500566 trace_ext4_error(inode->i_sb, function, line);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400567 es = EXT4_SB(inode->i_sb)->s_es;
568 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400569 if (ext4_error_ratelimit(inode->i_sb)) {
Miklos Szeredi9bf39ab2015-06-19 10:29:13 +0200570 path = file_path(file, pathname, sizeof(pathname));
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400571 if (IS_ERR(path))
572 path = "(unknown)";
573 va_start(args, fmt);
574 vaf.fmt = fmt;
575 vaf.va = &args;
576 if (block)
577 printk(KERN_CRIT
578 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
579 "block %llu: comm %s: path %s: %pV\n",
580 inode->i_sb->s_id, function, line, inode->i_ino,
581 block, current->comm, path, &vaf);
582 else
583 printk(KERN_CRIT
584 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
585 "comm %s: path %s: %pV\n",
586 inode->i_sb->s_id, function, line, inode->i_ino,
587 current->comm, path, &vaf);
588 va_end(args);
589 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400590 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500591 ext4_handle_error(inode->i_sb);
592}
593
Theodore Ts'o722887d2013-02-08 13:00:31 -0500594const char *ext4_decode_error(struct super_block *sb, int errno,
595 char nbuf[16])
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700596{
597 char *errstr = NULL;
598
599 switch (errno) {
Darrick J. Wong6a797d22015-10-17 16:16:04 -0400600 case -EFSCORRUPTED:
601 errstr = "Corrupt filesystem";
602 break;
603 case -EFSBADCRC:
604 errstr = "Filesystem failed CRC";
605 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700606 case -EIO:
607 errstr = "IO failure";
608 break;
609 case -ENOMEM:
610 errstr = "Out of memory";
611 break;
612 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400613 if (!sb || (EXT4_SB(sb)->s_journal &&
614 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700615 errstr = "Journal has aborted";
616 else
617 errstr = "Readonly filesystem";
618 break;
619 default:
620 /* If the caller passed in an extra buffer for unknown
621 * errors, textualise them now. Else we just return
622 * NULL. */
623 if (nbuf) {
624 /* Check for truncated error codes... */
625 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
626 errstr = nbuf;
627 }
628 break;
629 }
630
631 return errstr;
632}
633
Mingming Cao617ba132006-10-11 01:20:53 -0700634/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700635 * automatically and invokes the appropriate error response. */
636
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400637void __ext4_std_error(struct super_block *sb, const char *function,
638 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700639{
640 char nbuf[16];
641 const char *errstr;
642
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500643 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
644 return;
645
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700646 /* Special case: if the error is EROFS, and we're not already
647 * inside a transaction, then there's really no point in logging
648 * an error. */
David Howellsbc98a422017-07-17 08:45:34 +0100649 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700650 return;
651
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400652 if (ext4_error_ratelimit(sb)) {
653 errstr = ext4_decode_error(sb, errno, nbuf);
654 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
655 sb->s_id, function, line, errstr);
656 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700657
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400658 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700659 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700660}
661
662/*
Mingming Cao617ba132006-10-11 01:20:53 -0700663 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700664 * abort function may be used to deal with unrecoverable failures such
665 * as journal IO errors or ENOMEM at a critical moment in log management.
666 *
667 * We unconditionally force the filesystem into an ABORT|READONLY state,
668 * unless the error response on the fs has been set to panic in which
669 * case we take the easy way out and panic immediately.
670 */
671
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400672void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400673 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700674{
Joe Perches651e1c32016-10-12 23:12:53 -0400675 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700676 va_list args;
677
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500678 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
679 return;
680
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400681 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700682 va_start(args, fmt);
Joe Perches651e1c32016-10-12 23:12:53 -0400683 vaf.fmt = fmt;
684 vaf.va = &args;
685 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
686 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700687 va_end(args);
688
David Howellsbc98a422017-07-17 08:45:34 +0100689 if (sb_rdonly(sb) == 0) {
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400690 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400691 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400692 /*
693 * Make sure updated value of ->s_mount_flags will be visible
694 * before ->s_flags update
695 */
696 smp_wmb();
Linus Torvalds1751e8a2017-11-27 13:05:09 -0800697 sb->s_flags |= SB_RDONLY;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400698 if (EXT4_SB(sb)->s_journal)
699 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
700 save_error_info(sb, function, line);
701 }
Jan Kara2c1d0e32019-05-17 17:37:18 -0400702 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
Daeho Jeong4327ba52015-10-18 17:02:56 -0400703 if (EXT4_SB(sb)->s_journal &&
704 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
705 return;
Mingming Cao617ba132006-10-11 01:20:53 -0700706 panic("EXT4-fs panic from previous error\n");
Daeho Jeong4327ba52015-10-18 17:02:56 -0400707 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700708}
709
Joe Perchese7c96e82013-07-01 08:12:37 -0400710void __ext4_msg(struct super_block *sb,
711 const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400712{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500713 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400714 va_list args;
715
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400716 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
717 return;
718
Eric Sandeenb31e1552009-06-04 17:36:36 -0400719 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500720 vaf.fmt = fmt;
721 vaf.va = &args;
722 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400723 va_end(args);
724}
725
Andreas Dilgerb03a2f72015-06-15 14:50:26 -0400726#define ext4_warning_ratelimit(sb) \
727 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
728 "EXT4-fs warning")
729
Eric Sandeen12062dd2010-02-15 14:19:27 -0500730void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400731 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700732{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500733 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700734 va_list args;
735
Andreas Dilgerb03a2f72015-06-15 14:50:26 -0400736 if (!ext4_warning_ratelimit(sb))
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400737 return;
738
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700739 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500740 vaf.fmt = fmt;
741 vaf.va = &args;
742 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
743 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700744 va_end(args);
745}
746
Andreas Dilgerb03a2f72015-06-15 14:50:26 -0400747void __ext4_warning_inode(const struct inode *inode, const char *function,
748 unsigned int line, const char *fmt, ...)
749{
750 struct va_format vaf;
751 va_list args;
752
753 if (!ext4_warning_ratelimit(inode->i_sb))
754 return;
755
756 va_start(args, fmt);
757 vaf.fmt = fmt;
758 vaf.va = &args;
759 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
760 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
761 function, line, inode->i_ino, current->comm, &vaf);
762 va_end(args);
763}
764
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400765void __ext4_grp_locked_error(const char *function, unsigned int line,
766 struct super_block *sb, ext4_group_t grp,
767 unsigned long ino, ext4_fsblk_t block,
768 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500769__releases(bitlock)
770__acquires(bitlock)
771{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500772 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500773 va_list args;
774 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
775
Theodore Ts'o0db1ff22017-02-05 01:28:48 -0500776 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
777 return;
778
Theodore Ts'occf0f322018-02-18 20:53:23 -0500779 trace_ext4_error(sb, function, line);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400780 es->s_last_error_ino = cpu_to_le32(ino);
781 es->s_last_error_block = cpu_to_le64(block);
782 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500783
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400784 if (ext4_error_ratelimit(sb)) {
785 va_start(args, fmt);
786 vaf.fmt = fmt;
787 vaf.va = &args;
788 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
789 sb->s_id, function, line, grp);
790 if (ino)
791 printk(KERN_CONT "inode %lu: ", ino);
792 if (block)
793 printk(KERN_CONT "block %llu:",
794 (unsigned long long) block);
795 printk(KERN_CONT "%pV\n", &vaf);
796 va_end(args);
797 }
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500798
Theodore Ts'o327eaf72018-06-12 23:34:57 -0400799 if (test_opt(sb, WARN_ON_ERROR))
800 WARN_ON_ONCE(1);
801
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500802 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400803 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500804 return;
805 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400806
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500807 ext4_unlock_group(sb, grp);
Zhouyi Zhou06f29cc2018-01-10 00:34:19 -0500808 ext4_commit_super(sb, 1);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500809 ext4_handle_error(sb);
810 /*
811 * We only get here in the ERRORS_RO case; relocking the group
812 * may be dangerous, but nothing bad will happen since the
813 * filesystem will have already been marked read/only and the
814 * journal has been aborted. We return 1 as a hint to callers
815 * who might what to use the return value from
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300816 * ext4_grp_locked_error() to distinguish between the
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500817 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
818 * aggressively from the ext4 function in question, with a
819 * more appropriate error code.
820 */
821 ext4_lock_group(sb, grp);
822 return;
823}
824
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400825void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
826 ext4_group_t group,
827 unsigned int flags)
828{
829 struct ext4_sb_info *sbi = EXT4_SB(sb);
830 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
831 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
Wang Shilong9af0b3d2018-07-29 17:27:45 -0400832 int ret;
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400833
Wang Shilong9af0b3d2018-07-29 17:27:45 -0400834 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
835 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
836 &grp->bb_state);
837 if (!ret)
838 percpu_counter_sub(&sbi->s_freeclusters_counter,
839 grp->bb_free);
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400840 }
841
Wang Shilong9af0b3d2018-07-29 17:27:45 -0400842 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
843 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
844 &grp->bb_state);
845 if (!ret && gdp) {
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400846 int count;
847
848 count = ext4_free_inodes_count(sb, gdp);
849 percpu_counter_sub(&sbi->s_freeinodes_counter,
850 count);
851 }
Wang Shilongdb79e6d2018-05-12 11:39:40 -0400852 }
853}
854
Mingming Cao617ba132006-10-11 01:20:53 -0700855void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700856{
Mingming Cao617ba132006-10-11 01:20:53 -0700857 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700858
Mingming Cao617ba132006-10-11 01:20:53 -0700859 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700860 return;
861
Eric Sandeen12062dd2010-02-15 14:19:27 -0500862 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700863 "updating to rev %d because of new feature flag, "
864 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700865 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700866
Mingming Cao617ba132006-10-11 01:20:53 -0700867 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
868 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
869 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700870 /* leave es->s_feature_*compat flags alone */
871 /* es->s_uuid will be set by e2fsck if empty */
872
873 /*
874 * The rest of the superblock fields should be zero, and if not it
875 * means they are likely already in use, so leave them alone. We
876 * can leave it up to e2fsck to clean up any inconsistencies there.
877 */
878}
879
880/*
881 * Open the external journal device
882 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400883static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700884{
885 struct block_device *bdev;
886 char b[BDEVNAME_SIZE];
887
Tejun Heod4d77622010-11-13 11:55:18 +0100888 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700889 if (IS_ERR(bdev))
890 goto fail;
891 return bdev;
892
893fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400894 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700895 __bdevname(dev, b), PTR_ERR(bdev));
896 return NULL;
897}
898
899/*
900 * Release the journal device
901 */
Al Viro4385bab2013-05-05 22:11:03 -0400902static void ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700903{
Al Viro4385bab2013-05-05 22:11:03 -0400904 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700905}
906
Al Viro4385bab2013-05-05 22:11:03 -0400907static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700908{
909 struct block_device *bdev;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700910 bdev = sbi->journal_bdev;
911 if (bdev) {
Al Viro4385bab2013-05-05 22:11:03 -0400912 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700913 sbi->journal_bdev = NULL;
914 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700915}
916
917static inline struct inode *orphan_list_entry(struct list_head *l)
918{
Mingming Cao617ba132006-10-11 01:20:53 -0700919 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700920}
921
Mingming Cao617ba132006-10-11 01:20:53 -0700922static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700923{
924 struct list_head *l;
925
Eric Sandeenb31e1552009-06-04 17:36:36 -0400926 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
927 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700928
929 printk(KERN_ERR "sb_info orphan list:\n");
930 list_for_each(l, &sbi->s_orphan) {
931 struct inode *inode = orphan_list_entry(l);
932 printk(KERN_ERR " "
933 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
934 inode->i_sb->s_id, inode->i_ino, inode,
935 inode->i_mode, inode->i_nlink,
936 NEXT_ORPHAN(inode));
937 }
938}
939
Jan Kara957153f2017-04-06 15:40:06 +0200940#ifdef CONFIG_QUOTA
941static int ext4_quota_off(struct super_block *sb, int type);
942
943static inline void ext4_quota_off_umount(struct super_block *sb)
944{
945 int type;
946
Jan Kara964edf62017-05-21 22:31:23 -0400947 /* Use our quota_off function to clear inode flags etc. */
948 for (type = 0; type < EXT4_MAXQUOTAS; type++)
949 ext4_quota_off(sb, type);
Jan Kara957153f2017-04-06 15:40:06 +0200950}
Theodore Ts'o33458ea2018-10-12 09:28:09 -0400951
952/*
953 * This is a helper function which is used in the mount/remount
954 * codepaths (which holds s_umount) to fetch the quota file name.
955 */
956static inline char *get_qf_name(struct super_block *sb,
957 struct ext4_sb_info *sbi,
958 int type)
959{
960 return rcu_dereference_protected(sbi->s_qf_names[type],
961 lockdep_is_held(&sb->s_umount));
962}
Jan Kara957153f2017-04-06 15:40:06 +0200963#else
964static inline void ext4_quota_off_umount(struct super_block *sb)
965{
966}
967#endif
968
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400969static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700970{
Mingming Cao617ba132006-10-11 01:20:53 -0700971 struct ext4_sb_info *sbi = EXT4_SB(sb);
972 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'o97abd7d2017-02-04 23:38:06 -0500973 int aborted = 0;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400974 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700975
Lukas Czerner857ac882010-10-27 21:30:05 -0400976 ext4_unregister_li_request(sb);
Jan Kara957153f2017-04-06 15:40:06 +0200977 ext4_quota_off_umount(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400978
Jan Kara2e8fa542013-06-04 14:21:02 -0400979 destroy_workqueue(sbi->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -0400980
Frank Mayhar03901312009-01-07 00:06:22 -0500981 if (sbi->s_journal) {
Theodore Ts'o97abd7d2017-02-04 23:38:06 -0500982 aborted = is_journal_aborted(sbi->s_journal);
Frank Mayhar03901312009-01-07 00:06:22 -0500983 err = jbd2_journal_destroy(sbi->s_journal);
984 sbi->s_journal = NULL;
Theodore Ts'o97abd7d2017-02-04 23:38:06 -0500985 if ((err < 0) && !aborted)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400986 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500987 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500988
Theodore Ts'oebd173b2015-09-23 12:46:17 -0400989 ext4_unregister_sysfs(sb);
Zheng Liud3922a72013-07-01 08:12:37 -0400990 ext4_es_unregister_shrinker(sbi);
Al Viro9105bb12013-12-08 20:52:31 -0500991 del_timer_sync(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500992 ext4_release_system_zone(sb);
993 ext4_mb_release(sb);
994 ext4_ext_release(sb);
Josef Bacikd4edac32009-12-08 21:48:58 -0500995
David Howellsbc98a422017-07-17 08:45:34 +0100996 if (!sb_rdonly(sb) && !aborted) {
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400997 ext4_clear_feature_journal_needs_recovery(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700998 es->s_state = cpu_to_le16(sbi->s_mount_state);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700999 }
David Howellsbc98a422017-07-17 08:45:34 +01001000 if (!sb_rdonly(sb))
Artem Bityutskiya8e25a82012-03-21 22:29:15 -04001001 ext4_commit_super(sb, 1);
1002
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001003 for (i = 0; i < sbi->s_gdb_count; i++)
1004 brelse(sbi->s_group_desc[i]);
Al Virob93b41d2014-11-20 12:19:11 -05001005 kvfree(sbi->s_group_desc);
1006 kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -04001007 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001008 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1009 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -04001010 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Daeho Jeongc8585c62016-04-25 23:22:35 -04001011 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001012#ifdef CONFIG_QUOTA
Jan Karaa2d4a642014-09-11 11:15:15 -04001013 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001014 kfree(get_qf_name(sb, sbi, i));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001015#endif
1016
1017 /* Debugging code just in case the in-memory inode orphan list
1018 * isn't empty. The on-disk one can be non-empty if we've
1019 * detected an error and taken the fs readonly, but the
1020 * in-memory list had better be clean by this point. */
1021 if (!list_empty(&sbi->s_orphan))
1022 dump_orphan_list(sb, sbi);
1023 J_ASSERT(list_empty(&sbi->s_orphan));
1024
Theodore Ts'o89d96a62015-06-20 22:50:33 -04001025 sync_blockdev(sb->s_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -07001026 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001027 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1028 /*
1029 * Invalidate the journal device's buffers. We don't want them
1030 * floating about in memory - the physical journal device may
1031 * hotswapped, and it breaks the `ro-after' testing code.
1032 */
1033 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -07001034 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -07001035 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001036 }
Chengguang Xu50c15df2018-12-04 00:24:42 -05001037
1038 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1039 sbi->s_ea_inode_cache = NULL;
1040
1041 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1042 sbi->s_ea_block_cache = NULL;
1043
Johann Lombardic5e06d12011-05-24 18:31:25 -04001044 if (sbi->s_mmp_tsk)
1045 kthread_stop(sbi->s_mmp_tsk);
Eric Sandeen9060dd22016-11-26 14:24:51 -05001046 brelse(sbi->s_sbh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001047 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001048 /*
1049 * Now that we are completely done shutting down the
1050 * superblock, we need to actually destroy the kobject.
1051 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04001052 kobject_put(&sbi->s_kobj);
1053 wait_for_completion(&sbi->s_kobj_unregister);
Darrick J. Wong04419842012-04-29 18:27:10 -04001054 if (sbi->s_chksum_driver)
1055 crypto_free_shash(sbi->s_chksum_driver);
Pekka Enberg705895b2009-02-15 18:07:52 -05001056 kfree(sbi->s_blockgroup_lock);
Dan Williams5e405592017-08-24 16:42:48 -07001057 fs_put_dax(sbi->s_daxdev);
Gabriel Krisman Bertazic83ad552019-04-25 14:05:42 -04001058#ifdef CONFIG_UNICODE
1059 utf8_unload(sbi->s_encoding);
1060#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001061 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001062}
1063
Christoph Lametere18b8902006-12-06 20:33:20 -08001064static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001065
1066/*
1067 * Called inside transaction, so use GFP_NOFS
1068 */
Mingming Cao617ba132006-10-11 01:20:53 -07001069static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001070{
Mingming Cao617ba132006-10-11 01:20:53 -07001071 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001072
Christoph Lametere6b4f8d2006-12-06 20:33:14 -08001073 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001074 if (!ei)
1075 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001076
Jeff Laytonee73f9a2018-01-09 08:21:39 -05001077 inode_set_iversion(&ei->vfs_inode, 1);
Theodore Ts'o202ee5d2014-04-21 14:37:55 -04001078 spin_lock_init(&ei->i_raw_lock);
Alex Tomasc9de5602008-01-29 00:19:52 -05001079 INIT_LIST_HEAD(&ei->i_prealloc_list);
1080 spin_lock_init(&ei->i_prealloc_lock);
Zheng Liu9a26b662012-11-08 21:57:30 -05001081 ext4_es_init_tree(&ei->i_es_tree);
1082 rwlock_init(&ei->i_es_lock);
Zheng Liuedaa53c2014-11-25 11:45:37 -05001083 INIT_LIST_HEAD(&ei->i_es_list);
Zheng Liueb68d0e2014-09-01 22:26:49 -04001084 ei->i_es_all_nr = 0;
Zheng Liuedaa53c2014-11-25 11:45:37 -05001085 ei->i_es_shk_nr = 0;
Jan Karadd475922014-11-25 11:51:23 -05001086 ei->i_es_shrink_lblk = 0;
Mingming Caod2a17632008-07-14 17:52:37 -04001087 ei->i_reserved_data_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -05001088 ei->i_da_metadata_calc_len = 0;
Theodore Ts'o7e731bc2012-08-05 23:28:16 -04001089 ei->i_da_metadata_calc_last_lblock = 0;
Mingming Caod2a17632008-07-14 17:52:37 -04001090 spin_lock_init(&(ei->i_block_reservation_lock));
Eric Whitney1dc0aa42018-10-01 14:17:41 -04001091 ext4_init_pending_tree(&ei->i_pending_tree);
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001092#ifdef CONFIG_QUOTA
1093 ei->i_reserved_quota = 0;
Jan Kara96c7e0d2014-09-29 14:58:25 +02001094 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +03001095#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -05001096 ei->jinode = NULL;
Jan Kara2e8fa542013-06-04 14:21:02 -04001097 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -05001098 spin_lock_init(&ei->i_completed_io_lock);
Jan Karab436b9b2009-12-08 23:51:10 -05001099 ei->i_sync_tid = 0;
1100 ei->i_datasync_tid = 0;
Dmitry Monakhove27f41e2012-09-28 23:24:52 -04001101 atomic_set(&ei->i_unwritten, 0);
Jan Kara2e8fa542013-06-04 14:21:02 -04001102 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001103 return &ei->vfs_inode;
1104}
1105
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001106static int ext4_drop_inode(struct inode *inode)
1107{
1108 int drop = generic_drop_inode(inode);
1109
Eric Biggers29b36922019-08-04 19:35:48 -07001110 if (!drop)
1111 drop = fscrypt_drop_inode(inode);
1112
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001113 trace_ext4_drop_inode(inode, drop);
1114 return drop;
1115}
1116
Al Viro94053132019-04-15 19:28:34 -04001117static void ext4_free_in_core_inode(struct inode *inode)
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +11001118{
Eric Biggers2c58d542019-04-10 13:21:15 -07001119 fscrypt_free_inode(inode);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +11001120 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1121}
1122
Mingming Cao617ba132006-10-11 01:20:53 -07001123static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001124{
Vasily Averin9f7dd932007-07-15 23:40:45 -07001125 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001126 ext4_msg(inode->i_sb, KERN_ERR,
1127 "Inode %lu (%p): orphan list check failed!",
1128 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -07001129 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1130 EXT4_I(inode), sizeof(struct ext4_inode_info),
1131 true);
1132 dump_stack();
1133 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001134}
1135
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07001136static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001137{
Mingming Cao617ba132006-10-11 01:20:53 -07001138 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001139
Christoph Lametera35afb82007-05-16 22:10:57 -07001140 INIT_LIST_HEAD(&ei->i_orphan);
Christoph Lametera35afb82007-05-16 22:10:57 -07001141 init_rwsem(&ei->xattr_sem);
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -05001142 init_rwsem(&ei->i_data_sem);
Jan Karaea3d7202015-12-07 14:28:03 -05001143 init_rwsem(&ei->i_mmap_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -07001144 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001145}
1146
Fabian Fredericke67bc2b2014-02-17 20:34:53 -05001147static int __init init_inodecache(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001148{
David Windsorf8dd7c72017-06-10 22:50:36 -04001149 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1150 sizeof(struct ext4_inode_info), 0,
1151 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1152 SLAB_ACCOUNT),
1153 offsetof(struct ext4_inode_info, i_data),
1154 sizeof_field(struct ext4_inode_info, i_data),
1155 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -07001156 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001157 return -ENOMEM;
1158 return 0;
1159}
1160
1161static void destroy_inodecache(void)
1162{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +10001163 /*
1164 * Make sure all delayed rcu free inodes are flushed before we
1165 * destroy cache.
1166 */
1167 rcu_barrier();
Mingming Cao617ba132006-10-11 01:20:53 -07001168 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001169}
1170
Al Viro0930fcc2010-06-07 13:16:22 -04001171void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001172{
Al Viro0930fcc2010-06-07 13:16:22 -04001173 invalidate_inode_buffers(inode);
Jan Karadbd57682012-05-03 14:48:02 +02001174 clear_inode(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -04001175 ext4_discard_preallocations(inode);
Zheng Liu51865fd2012-11-08 21:57:32 -05001176 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
Jan Karaf4c2d372b2019-11-08 12:45:11 +01001177 dquot_drop(inode);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -05001178 if (EXT4_I(inode)->jinode) {
1179 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1180 EXT4_I(inode)->jinode);
1181 jbd2_free_inode(EXT4_I(inode)->jinode);
1182 EXT4_I(inode)->jinode = NULL;
1183 }
Eric Biggers3d204e22018-01-11 23:30:13 -05001184 fscrypt_put_encryption_info(inode);
Eric Biggersc93d8f82019-07-22 09:26:24 -07001185 fsverity_cleanup_inode(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001186}
1187
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001188static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001189 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001190{
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001191 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001192
Theodore Ts'o8a363972018-12-19 12:29:13 -05001193 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001194 * Currently we don't know the generation for parent directory, so
1195 * a generation of 0 means "accept any"
1196 */
Theodore Ts'o8a363972018-12-19 12:29:13 -05001197 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
David Howells1d1fe1e2008-02-07 00:15:37 -08001198 if (IS_ERR(inode))
1199 return ERR_CAST(inode);
1200 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001201 iput(inode);
1202 return ERR_PTR(-ESTALE);
1203 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001204
1205 return inode;
1206}
1207
1208static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001209 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001210{
1211 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1212 ext4_nfs_get_inode);
1213}
1214
1215static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001216 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001217{
1218 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1219 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001220}
1221
Theodore Ts'ofde87262018-12-19 14:07:58 -05001222static int ext4_nfs_commit_metadata(struct inode *inode)
1223{
1224 struct writeback_control wbc = {
1225 .sync_mode = WB_SYNC_ALL
1226 };
1227
1228 trace_ext4_nfs_commit_metadata(inode);
1229 return ext4_write_inode(inode, &wbc);
1230}
1231
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001232/*
1233 * Try to release metadata pages (indirect blocks, directories) which are
1234 * mapped via the block device. Since these pages could have journal heads
1235 * which would prevent try_to_free_buffers() from freeing them, we must use
1236 * jbd2 layer's try_to_free_buffers() function to release them.
1237 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001238static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1239 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001240{
1241 journal_t *journal = EXT4_SB(sb)->s_journal;
1242
1243 WARN_ON(PageChecked(page));
1244 if (!page_has_buffers(page))
1245 return 0;
1246 if (journal)
1247 return jbd2_journal_try_to_free_buffers(journal, page,
Mel Gormand0164ad2015-11-06 16:28:21 -08001248 wait & ~__GFP_DIRECT_RECLAIM);
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001249 return try_to_free_buffers(page);
1250}
1251
Chandan Rajendra643fa962018-12-12 15:20:12 +05301252#ifdef CONFIG_FS_ENCRYPTION
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001253static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1254{
1255 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1256 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1257}
1258
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001259static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1260 void *fs_data)
1261{
Eric Biggers2f8f5e72016-11-21 11:52:44 -05001262 handle_t *handle = fs_data;
Tahsin Erdoganc1a5d5f2017-06-21 22:28:40 -04001263 int res, res2, credits, retries = 0;
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001264
Eric Biggers9ce01512017-06-23 00:10:36 -04001265 /*
1266 * Encrypting the root directory is not allowed because e2fsck expects
1267 * lost+found to exist and be unencrypted, and encrypting the root
1268 * directory would imply encrypting the lost+found directory as well as
1269 * the filename "lost+found" itself.
1270 */
1271 if (inode->i_ino == EXT4_ROOT_INO)
1272 return -EPERM;
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001273
Ross Zwisler7d3e06a2017-10-12 11:58:05 -04001274 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1275 return -EINVAL;
1276
Eric Biggers94840e32017-02-22 13:25:14 -08001277 res = ext4_convert_inline_data(inode);
1278 if (res)
1279 return res;
1280
Eric Biggers2f8f5e72016-11-21 11:52:44 -05001281 /*
1282 * If a journal handle was specified, then the encryption context is
1283 * being set on a new inode via inheritance and is part of a larger
1284 * transaction to create the inode. Otherwise the encryption context is
1285 * being set on an existing inode in its own transaction. Only in the
1286 * latter case should the "retry on ENOSPC" logic be used.
1287 */
1288
1289 if (handle) {
1290 res = ext4_xattr_set_handle(handle, inode,
1291 EXT4_XATTR_INDEX_ENCRYPTION,
1292 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1293 ctx, len, 0);
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001294 if (!res) {
1295 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1296 ext4_clear_inode_state(inode,
1297 EXT4_STATE_MAY_INLINE_DATA);
Jan Karaa3caa242016-11-20 17:32:59 -05001298 /*
Eric Biggers2ee6a572017-10-09 12:15:35 -07001299 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1300 * S_DAX may be disabled
Jan Karaa3caa242016-11-20 17:32:59 -05001301 */
1302 ext4_set_inode_flags(inode);
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001303 }
1304 return res;
1305 }
1306
Tahsin Erdoganb8cb5a52017-05-24 18:24:07 -04001307 res = dquot_initialize(inode);
1308 if (res)
1309 return res;
Eric Biggers2f8f5e72016-11-21 11:52:44 -05001310retry:
Tahsin Erdoganaf652072017-07-06 00:01:59 -04001311 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1312 &credits);
Tahsin Erdogandec214d2017-06-22 11:44:55 -04001313 if (res)
1314 return res;
1315
Tahsin Erdoganc1a5d5f2017-06-21 22:28:40 -04001316 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001317 if (IS_ERR(handle))
1318 return PTR_ERR(handle);
1319
Eric Biggers2f8f5e72016-11-21 11:52:44 -05001320 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1321 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1322 ctx, len, 0);
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001323 if (!res) {
1324 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
Eric Biggers2ee6a572017-10-09 12:15:35 -07001325 /*
1326 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1327 * S_DAX may be disabled
1328 */
Jan Karaa3caa242016-11-20 17:32:59 -05001329 ext4_set_inode_flags(inode);
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001330 res = ext4_mark_inode_dirty(handle, inode);
1331 if (res)
1332 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1333 }
1334 res2 = ext4_journal_stop(handle);
Eric Biggers2f8f5e72016-11-21 11:52:44 -05001335
1336 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1337 goto retry;
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001338 if (!res)
1339 res = res2;
1340 return res;
1341}
1342
Eric Biggersc250b7d2017-06-22 12:14:40 -07001343static bool ext4_dummy_context(struct inode *inode)
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001344{
1345 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1346}
1347
Eric Biggersb925acb2019-10-24 14:54:37 -07001348static bool ext4_has_stable_inodes(struct super_block *sb)
1349{
1350 return ext4_has_feature_stable_inodes(sb);
1351}
1352
1353static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1354 int *ino_bits_ret, int *lblk_bits_ret)
1355{
1356 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1357 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1358}
1359
Eric Biggers6f69f0e2017-02-07 12:42:10 -08001360static const struct fscrypt_operations ext4_cryptops = {
Eric Biggersa5d431e2017-01-05 13:51:18 -08001361 .key_prefix = "ext4:",
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001362 .get_context = ext4_get_context,
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001363 .set_context = ext4_set_context,
1364 .dummy_context = ext4_dummy_context,
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001365 .empty_dir = ext4_empty_dir,
Eric Biggerse12ee682018-04-30 15:51:44 -07001366 .max_namelen = EXT4_NAME_LEN,
Eric Biggersb925acb2019-10-24 14:54:37 -07001367 .has_stable_inodes = ext4_has_stable_inodes,
1368 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001369};
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001370#endif
1371
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001372#ifdef CONFIG_QUOTA
Eric Biggersd6006182017-04-29 23:47:50 -04001373static const char * const quotatypes[] = INITQFNAMES;
Li Xi689c958c2016-01-08 16:01:22 -05001374#define QTYPE2NAME(t) (quotatypes[t])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001375
Mingming Cao617ba132006-10-11 01:20:53 -07001376static int ext4_write_dquot(struct dquot *dquot);
1377static int ext4_acquire_dquot(struct dquot *dquot);
1378static int ext4_release_dquot(struct dquot *dquot);
1379static int ext4_mark_dquot_dirty(struct dquot *dquot);
1380static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001381static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Al Viro8c54ca92016-11-20 19:49:34 -05001382 const struct path *path);
Mingming Cao617ba132006-10-11 01:20:53 -07001383static int ext4_quota_on_mount(struct super_block *sb, int type);
1384static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001385 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001386static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001387 const char *data, size_t len, loff_t off);
Aditya Kali7c319d32012-07-22 20:21:31 -04001388static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1389 unsigned int flags);
1390static int ext4_enable_quotas(struct super_block *sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001391
Jan Kara96c7e0d2014-09-29 14:58:25 +02001392static struct dquot **ext4_get_dquots(struct inode *inode)
1393{
1394 return EXT4_I(inode)->i_dquot;
1395}
1396
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001397static const struct dquot_operations ext4_quota_operations = {
Tahsin Erdogan7a9ca532017-06-22 11:46:48 -04001398 .get_reserved_space = ext4_get_reserved_space,
1399 .write_dquot = ext4_write_dquot,
1400 .acquire_dquot = ext4_acquire_dquot,
1401 .release_dquot = ext4_release_dquot,
1402 .mark_dirty = ext4_mark_dquot_dirty,
1403 .write_info = ext4_write_info,
1404 .alloc_dquot = dquot_alloc,
1405 .destroy_dquot = dquot_destroy,
1406 .get_projid = ext4_get_projid,
1407 .get_inode_usage = ext4_get_inode_usage,
Chengguang Xuebc11f72019-10-06 18:30:28 +08001408 .get_next_id = dquot_get_next_id,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001409};
1410
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001411static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001412 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001413 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001414 .quota_sync = dquot_quota_sync,
Jan Kara0a2403392014-11-19 00:42:09 +01001415 .get_state = dquot_get_state,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001416 .set_info = dquot_set_dqinfo,
1417 .get_dqblk = dquot_get_dqblk,
Eric Sandeen6332b9b52016-02-19 19:19:01 +01001418 .set_dqblk = dquot_set_dqblk,
1419 .get_nextdqblk = dquot_get_next_dqblk,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001420};
1421#endif
1422
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001423static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001424 .alloc_inode = ext4_alloc_inode,
Al Viro94053132019-04-15 19:28:34 -04001425 .free_inode = ext4_free_in_core_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001426 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001427 .write_inode = ext4_write_inode,
1428 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001429 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001430 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001431 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001432 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001433 .freeze_fs = ext4_freeze,
1434 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001435 .statfs = ext4_statfs,
1436 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001437 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001438#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001439 .quota_read = ext4_quota_read,
1440 .quota_write = ext4_quota_write,
Jan Kara96c7e0d2014-09-29 14:58:25 +02001441 .get_dquots = ext4_get_dquots,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001442#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001443 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001444};
1445
Christoph Hellwig39655162007-10-21 16:42:17 -07001446static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001447 .fh_to_dentry = ext4_fh_to_dentry,
1448 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001449 .get_parent = ext4_get_parent,
Theodore Ts'ofde87262018-12-19 14:07:58 -05001450 .commit_metadata = ext4_nfs_commit_metadata,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001451};
1452
1453enum {
1454 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1455 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001456 Opt_nouid32, Opt_debug, Opt_removed,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001457 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001458 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
Eric Sandeenad4eec62013-08-28 19:05:07 -04001459 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1460 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001461 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001462 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001463 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001464 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
Theodore Ts'oee4a3fc2012-03-02 12:14:24 -05001465 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
Jan Kara49da9392016-09-05 23:08:16 -04001466 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
Theodore Ts'o327eaf72018-06-12 23:34:57 -04001467 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1468 Opt_nowarn_on_error, Opt_mblk_io_submit,
Theodore Ts'o670e9872017-01-11 15:32:22 -05001469 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
Theodore Ts'o14490322010-12-14 15:27:50 -05001470 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001471 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001472 Opt_dioread_nolock, Opt_dioread_lock,
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001473 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04001474 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001475};
1476
Steven Whitehousea447c092008-10-13 10:46:57 +01001477static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001478 {Opt_bsd_df, "bsddf"},
1479 {Opt_minix_df, "minixdf"},
1480 {Opt_grpid, "grpid"},
1481 {Opt_grpid, "bsdgroups"},
1482 {Opt_nogrpid, "nogrpid"},
1483 {Opt_nogrpid, "sysvgroups"},
1484 {Opt_resgid, "resgid=%u"},
1485 {Opt_resuid, "resuid=%u"},
1486 {Opt_sb, "sb=%u"},
1487 {Opt_err_cont, "errors=continue"},
1488 {Opt_err_panic, "errors=panic"},
1489 {Opt_err_ro, "errors=remount-ro"},
1490 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001491 {Opt_debug, "debug"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001492 {Opt_removed, "oldalloc"},
1493 {Opt_removed, "orlov"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001494 {Opt_user_xattr, "user_xattr"},
1495 {Opt_nouser_xattr, "nouser_xattr"},
1496 {Opt_acl, "acl"},
1497 {Opt_noacl, "noacl"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001498 {Opt_noload, "norecovery"},
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001499 {Opt_noload, "noload"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001500 {Opt_removed, "nobh"},
1501 {Opt_removed, "bh"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001502 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001503 {Opt_min_batch_time, "min_batch_time=%u"},
1504 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001505 {Opt_journal_dev, "journal_dev=%u"},
Eric Sandeenad4eec62013-08-28 19:05:07 -04001506 {Opt_journal_path, "journal_path=%s"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001507 {Opt_journal_checksum, "journal_checksum"},
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001508 {Opt_nojournal_checksum, "nojournal_checksum"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001509 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001510 {Opt_abort, "abort"},
1511 {Opt_data_journal, "data=journal"},
1512 {Opt_data_ordered, "data=ordered"},
1513 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001514 {Opt_data_err_abort, "data_err=abort"},
1515 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001516 {Opt_offusrjquota, "usrjquota="},
1517 {Opt_usrjquota, "usrjquota=%s"},
1518 {Opt_offgrpjquota, "grpjquota="},
1519 {Opt_grpjquota, "grpjquota=%s"},
1520 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1521 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001522 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001523 {Opt_grpquota, "grpquota"},
1524 {Opt_noquota, "noquota"},
1525 {Opt_quota, "quota"},
1526 {Opt_usrquota, "usrquota"},
Jan Kara49da9392016-09-05 23:08:16 -04001527 {Opt_prjquota, "prjquota"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001528 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001529 {Opt_barrier, "barrier"},
1530 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001531 {Opt_i_version, "i_version"},
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001532 {Opt_dax, "dax"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001533 {Opt_stripe, "stripe=%u"},
Alex Tomas64769242008-07-11 19:27:31 -04001534 {Opt_delalloc, "delalloc"},
Theodore Ts'o327eaf72018-06-12 23:34:57 -04001535 {Opt_warn_on_error, "warn_on_error"},
1536 {Opt_nowarn_on_error, "nowarn_on_error"},
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001537 {Opt_lazytime, "lazytime"},
1538 {Opt_nolazytime, "nolazytime"},
Theodore Ts'o670e9872017-01-11 15:32:22 -05001539 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001540 {Opt_nodelalloc, "nodelalloc"},
Jan Kara36ade452013-01-28 09:30:52 -05001541 {Opt_removed, "mblk_io_submit"},
1542 {Opt_removed, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001543 {Opt_block_validity, "block_validity"},
1544 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001545 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001546 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001547 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001548 {Opt_auto_da_alloc, "auto_da_alloc"},
1549 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001550 {Opt_dioread_nolock, "dioread_nolock"},
1551 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001552 {Opt_discard, "discard"},
1553 {Opt_nodiscard, "nodiscard"},
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001554 {Opt_init_itable, "init_itable=%u"},
1555 {Opt_init_itable, "init_itable"},
1556 {Opt_noinit_itable, "noinit_itable"},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001557 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001558 {Opt_test_dummy_encryption, "test_dummy_encryption"},
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04001559 {Opt_nombcache, "nombcache"},
1560 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
Theodore Ts'oc7198b92012-03-04 22:00:53 -05001561 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1562 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1563 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1564 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1565 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001566 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001567};
1568
Mingming Cao617ba132006-10-11 01:20:53 -07001569static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001570{
Mingming Cao617ba132006-10-11 01:20:53 -07001571 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001572 char *options = (char *) *data;
1573
1574 if (!options || strncmp(options, "sb=", 3) != 0)
1575 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001576
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001577 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001578 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001579 sb_block = simple_strtoul(options, &options, 0);
1580 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001581 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001582 (char *) *data);
1583 return 1;
1584 }
1585 if (*options == ',')
1586 options++;
1587 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001588
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001589 return sb_block;
1590}
1591
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001592#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Eric Biggersd6006182017-04-29 23:47:50 -04001593static const char deprecated_msg[] =
1594 "Mount option \"%s\" will be removed by %s\n"
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001595 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001596
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001597#ifdef CONFIG_QUOTA
1598static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1599{
1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001601 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
Chen Gang03dafb52013-01-24 23:24:58 -05001602 int ret = -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001603
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001604 if (sb_any_quota_loaded(sb) && !old_qname) {
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001605 ext4_msg(sb, KERN_ERR,
1606 "Cannot change journaled "
1607 "quota options when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001608 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001609 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04001610 if (ext4_has_feature_quota(sb)) {
Theodore Ts'oc325a672016-04-03 17:03:37 -04001611 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1612 "ignored when QUOTA feature is enabled");
1613 return 1;
Jan Kara262b4662013-03-02 17:57:08 -05001614 }
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001615 qname = match_strdup(args);
1616 if (!qname) {
1617 ext4_msg(sb, KERN_ERR,
1618 "Not enough memory for storing quotafile name");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001619 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001620 }
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001621 if (old_qname) {
1622 if (strcmp(old_qname, qname) == 0)
Chen Gang03dafb52013-01-24 23:24:58 -05001623 ret = 1;
1624 else
1625 ext4_msg(sb, KERN_ERR,
1626 "%s quota file already specified",
1627 QTYPE2NAME(qtype));
1628 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001629 }
Chen Gang03dafb52013-01-24 23:24:58 -05001630 if (strchr(qname, '/')) {
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001631 ext4_msg(sb, KERN_ERR,
1632 "quotafile must be on filesystem root");
Chen Gang03dafb52013-01-24 23:24:58 -05001633 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001634 }
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001635 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001636 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001637 return 1;
Chen Gang03dafb52013-01-24 23:24:58 -05001638errout:
1639 kfree(qname);
1640 return ret;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001641}
1642
1643static int clear_qf_name(struct super_block *sb, int qtype)
1644{
1645
1646 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001647 char *old_qname = get_qf_name(sb, sbi, qtype);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001648
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001649 if (sb_any_quota_loaded(sb) && old_qname) {
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001650 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1651 " when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001652 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001653 }
Theodore Ts'o33458ea2018-10-12 09:28:09 -04001654 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1655 synchronize_rcu();
1656 kfree(old_qname);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001657 return 1;
1658}
1659#endif
1660
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001661#define MOPT_SET 0x0001
1662#define MOPT_CLEAR 0x0002
1663#define MOPT_NOSUPPORT 0x0004
1664#define MOPT_EXPLICIT 0x0008
1665#define MOPT_CLEAR_ERR 0x0010
1666#define MOPT_GTE0 0x0020
1667#ifdef CONFIG_QUOTA
1668#define MOPT_Q 0
1669#define MOPT_QFMT 0x0040
1670#else
1671#define MOPT_Q MOPT_NOSUPPORT
1672#define MOPT_QFMT MOPT_NOSUPPORT
1673#endif
1674#define MOPT_DATAJ 0x0080
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001675#define MOPT_NO_EXT2 0x0100
1676#define MOPT_NO_EXT3 0x0200
1677#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
Eric Sandeenad4eec62013-08-28 19:05:07 -04001678#define MOPT_STRING 0x0400
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001679
1680static const struct mount_opts {
1681 int token;
1682 int mount_opt;
1683 int flags;
1684} ext4_mount_opts[] = {
1685 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1686 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1687 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1688 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001689 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1690 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001691 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1692 MOPT_EXT4_ONLY | MOPT_SET},
1693 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1694 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001695 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1696 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001697 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1698 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1699 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
Theodore Ts'o59d9fa52013-08-08 23:01:24 -04001700 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o327eaf72018-06-12 23:34:57 -04001701 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1702 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001703 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1704 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001705 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04001706 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001707 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001708 EXT4_MOUNT_JOURNAL_CHECKSUM),
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04001709 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001710 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001711 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1712 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1713 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001714 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
Ales Novak7915a862016-03-12 21:55:50 -05001715 MOPT_NO_EXT2},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001716 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
Ales Novak7915a862016-03-12 21:55:50 -05001717 MOPT_NO_EXT2},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001718 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1719 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1720 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1721 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1722 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1723 {Opt_commit, 0, MOPT_GTE0},
1724 {Opt_max_batch_time, 0, MOPT_GTE0},
1725 {Opt_min_batch_time, 0, MOPT_GTE0},
1726 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1727 {Opt_init_itable, 0, MOPT_GTE0},
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001728 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001729 {Opt_stripe, 0, MOPT_GTE0},
Jan Kara0efb3b22013-02-02 22:52:19 -05001730 {Opt_resuid, 0, MOPT_GTE0},
1731 {Opt_resgid, 0, MOPT_GTE0},
Carlos Maiolino5ba92bc2015-07-21 23:57:59 -04001732 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1733 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1734 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001735 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1736 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1737 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1738 MOPT_NO_EXT2 | MOPT_DATAJ},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001739 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1740 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001741#ifdef CONFIG_EXT4_FS_POSIX_ACL
1742 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1743 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1744#else
1745 {Opt_acl, 0, MOPT_NOSUPPORT},
1746 {Opt_noacl, 0, MOPT_NOSUPPORT},
1747#endif
1748 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1749 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
Theodore Ts'o670e9872017-01-11 15:32:22 -05001750 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001751 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1752 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1753 MOPT_SET | MOPT_Q},
1754 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1755 MOPT_SET | MOPT_Q},
Jan Kara49da9392016-09-05 23:08:16 -04001756 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1757 MOPT_SET | MOPT_Q},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001758 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
Jan Kara49da9392016-09-05 23:08:16 -04001759 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1760 MOPT_CLEAR | MOPT_Q},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001761 {Opt_usrjquota, 0, MOPT_Q},
1762 {Opt_grpjquota, 0, MOPT_Q},
1763 {Opt_offusrjquota, 0, MOPT_Q},
1764 {Opt_offgrpjquota, 0, MOPT_Q},
1765 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1766 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1767 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001768 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001769 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04001770 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001771 {Opt_err, 0, 0}
1772};
1773
Gabriel Krisman Bertazic83ad552019-04-25 14:05:42 -04001774#ifdef CONFIG_UNICODE
1775static const struct ext4_sb_encodings {
1776 __u16 magic;
1777 char *name;
1778 char *version;
1779} ext4_sb_encoding_map[] = {
1780 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1781};
1782
1783static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1784 const struct ext4_sb_encodings **encoding,
1785 __u16 *flags)
1786{
1787 __u16 magic = le16_to_cpu(es->s_encoding);
1788 int i;
1789
1790 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1791 if (magic == ext4_sb_encoding_map[i].magic)
1792 break;
1793
1794 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1795 return -EINVAL;
1796
1797 *encoding = &ext4_sb_encoding_map[i];
1798 *flags = le16_to_cpu(es->s_encoding_flags);
1799
1800 return 0;
1801}
1802#endif
1803
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001804static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1805 substring_t *args, unsigned long *journal_devnum,
1806 unsigned int *journal_ioprio, int is_remount)
1807{
1808 struct ext4_sb_info *sbi = EXT4_SB(sb);
1809 const struct mount_opts *m;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001810 kuid_t uid;
1811 kgid_t gid;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001812 int arg = 0;
1813
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001814#ifdef CONFIG_QUOTA
1815 if (token == Opt_usrjquota)
1816 return set_qf_name(sb, USRQUOTA, &args[0]);
1817 else if (token == Opt_grpjquota)
1818 return set_qf_name(sb, GRPQUOTA, &args[0]);
1819 else if (token == Opt_offusrjquota)
1820 return clear_qf_name(sb, USRQUOTA);
1821 else if (token == Opt_offgrpjquota)
1822 return clear_qf_name(sb, GRPQUOTA);
1823#endif
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001824 switch (token) {
Theodore Ts'of7048602012-03-04 22:06:20 -05001825 case Opt_noacl:
1826 case Opt_nouser_xattr:
1827 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1828 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001829 case Opt_sb:
1830 return 1; /* handled by get_sb_block() */
1831 case Opt_removed:
Jan Kara5f3633e2013-02-02 23:09:36 -05001832 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001833 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001834 case Opt_abort:
1835 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1836 return 1;
1837 case Opt_i_version:
Matthew Garrett357fdad2017-10-18 13:56:26 -07001838 sb->s_flags |= SB_I_VERSION;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001839 return 1;
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001840 case Opt_lazytime:
Linus Torvalds1751e8a2017-11-27 13:05:09 -08001841 sb->s_flags |= SB_LAZYTIME;
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001842 return 1;
1843 case Opt_nolazytime:
Linus Torvalds1751e8a2017-11-27 13:05:09 -08001844 sb->s_flags &= ~SB_LAZYTIME;
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001845 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001846 }
1847
Jan Kara5f3633e2013-02-02 23:09:36 -05001848 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1849 if (token == m->token)
1850 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001851
Jan Kara5f3633e2013-02-02 23:09:36 -05001852 if (m->token == Opt_err) {
1853 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1854 "or missing value", opt);
1855 return -1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001856 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001857
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001858 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1859 ext4_msg(sb, KERN_ERR,
1860 "Mount option \"%s\" incompatible with ext2", opt);
1861 return -1;
1862 }
1863 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1864 ext4_msg(sb, KERN_ERR,
1865 "Mount option \"%s\" incompatible with ext3", opt);
1866 return -1;
1867 }
1868
Eric Sandeenad4eec62013-08-28 19:05:07 -04001869 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
Jan Kara5f3633e2013-02-02 23:09:36 -05001870 return -1;
1871 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1872 return -1;
Dmitry Monakhovc93cf2d2015-10-18 23:35:32 -04001873 if (m->flags & MOPT_EXPLICIT) {
1874 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1875 set_opt2(sb, EXPLICIT_DELALLOC);
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04001876 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1877 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
Dmitry Monakhovc93cf2d2015-10-18 23:35:32 -04001878 } else
1879 return -1;
1880 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001881 if (m->flags & MOPT_CLEAR_ERR)
1882 clear_opt(sb, ERRORS_MASK);
1883 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1884 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1885 "options when quota turned on");
1886 return -1;
1887 }
1888
1889 if (m->flags & MOPT_NOSUPPORT) {
1890 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1891 } else if (token == Opt_commit) {
1892 if (arg == 0)
1893 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
zhangyi (F)9ba55542019-08-28 11:25:01 -04001894 else if (arg > INT_MAX / HZ) {
1895 ext4_msg(sb, KERN_ERR,
1896 "Invalid commit interval %d, "
1897 "must be smaller than %d",
1898 arg, INT_MAX / HZ);
1899 return -1;
1900 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001901 sbi->s_commit_interval = HZ * arg;
Theodore Ts'o670e9872017-01-11 15:32:22 -05001902 } else if (token == Opt_debug_want_extra_isize) {
1903 sbi->s_want_extra_isize = arg;
Jan Kara5f3633e2013-02-02 23:09:36 -05001904 } else if (token == Opt_max_batch_time) {
Jan Kara5f3633e2013-02-02 23:09:36 -05001905 sbi->s_max_batch_time = arg;
1906 } else if (token == Opt_min_batch_time) {
1907 sbi->s_min_batch_time = arg;
1908 } else if (token == Opt_inode_readahead_blks) {
Jan Karae33e60e2013-02-02 23:14:31 -05001909 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1910 ext4_msg(sb, KERN_ERR,
1911 "EXT4-fs: inode_readahead_blks must be "
1912 "0 or a power of 2 smaller than 2^31");
Jan Kara5f3633e2013-02-02 23:09:36 -05001913 return -1;
1914 }
1915 sbi->s_inode_readahead_blks = arg;
1916 } else if (token == Opt_init_itable) {
1917 set_opt(sb, INIT_INODE_TABLE);
1918 if (!args->from)
1919 arg = EXT4_DEF_LI_WAIT_MULT;
1920 sbi->s_li_wait_mult = arg;
1921 } else if (token == Opt_max_dir_size_kb) {
1922 sbi->s_max_dir_size_kb = arg;
1923 } else if (token == Opt_stripe) {
1924 sbi->s_stripe = arg;
1925 } else if (token == Opt_resuid) {
1926 uid = make_kuid(current_user_ns(), arg);
1927 if (!uid_valid(uid)) {
1928 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1929 return -1;
1930 }
1931 sbi->s_resuid = uid;
1932 } else if (token == Opt_resgid) {
1933 gid = make_kgid(current_user_ns(), arg);
1934 if (!gid_valid(gid)) {
1935 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1936 return -1;
1937 }
1938 sbi->s_resgid = gid;
1939 } else if (token == Opt_journal_dev) {
1940 if (is_remount) {
1941 ext4_msg(sb, KERN_ERR,
1942 "Cannot specify journal on remount");
1943 return -1;
1944 }
1945 *journal_devnum = arg;
Eric Sandeenad4eec62013-08-28 19:05:07 -04001946 } else if (token == Opt_journal_path) {
1947 char *journal_path;
1948 struct inode *journal_inode;
1949 struct path path;
1950 int error;
1951
1952 if (is_remount) {
1953 ext4_msg(sb, KERN_ERR,
1954 "Cannot specify journal on remount");
1955 return -1;
1956 }
1957 journal_path = match_strdup(&args[0]);
1958 if (!journal_path) {
1959 ext4_msg(sb, KERN_ERR, "error: could not dup "
1960 "journal device string");
1961 return -1;
1962 }
1963
1964 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1965 if (error) {
1966 ext4_msg(sb, KERN_ERR, "error: could not find "
1967 "journal device path: error %d", error);
1968 kfree(journal_path);
1969 return -1;
1970 }
1971
David Howells2b0143b2015-03-17 22:25:59 +00001972 journal_inode = d_inode(path.dentry);
Eric Sandeenad4eec62013-08-28 19:05:07 -04001973 if (!S_ISBLK(journal_inode->i_mode)) {
1974 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1975 "is not a block device", journal_path);
1976 path_put(&path);
1977 kfree(journal_path);
1978 return -1;
1979 }
1980
1981 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1982 path_put(&path);
1983 kfree(journal_path);
Jan Kara5f3633e2013-02-02 23:09:36 -05001984 } else if (token == Opt_journal_ioprio) {
1985 if (arg > 7) {
1986 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1987 " (must be 0-7)");
1988 return -1;
1989 }
1990 *journal_ioprio =
1991 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001992 } else if (token == Opt_test_dummy_encryption) {
Chandan Rajendra643fa962018-12-12 15:20:12 +05301993#ifdef CONFIG_FS_ENCRYPTION
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001994 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1995 ext4_msg(sb, KERN_WARNING,
1996 "Test dummy encryption mode enabled");
1997#else
1998 ext4_msg(sb, KERN_WARNING,
1999 "Test dummy encryption mount option ignored");
2000#endif
Jan Kara5f3633e2013-02-02 23:09:36 -05002001 } else if (m->flags & MOPT_DATAJ) {
2002 if (is_remount) {
2003 if (!sbi->s_journal)
2004 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2005 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2006 ext4_msg(sb, KERN_ERR,
2007 "Cannot change data mode on remount");
2008 return -1;
2009 }
2010 } else {
2011 clear_opt(sb, DATA_FLAGS);
2012 sbi->s_mount_opt |= m->mount_opt;
2013 }
2014#ifdef CONFIG_QUOTA
2015 } else if (m->flags & MOPT_QFMT) {
2016 if (sb_any_quota_loaded(sb) &&
2017 sbi->s_jquota_fmt != m->mount_opt) {
2018 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2019 "quota options when quota turned on");
2020 return -1;
2021 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002022 if (ext4_has_feature_quota(sb)) {
Theodore Ts'oc325a672016-04-03 17:03:37 -04002023 ext4_msg(sb, KERN_INFO,
2024 "Quota format mount options ignored "
Jan Kara262b4662013-03-02 17:57:08 -05002025 "when QUOTA feature is enabled");
Theodore Ts'oc325a672016-04-03 17:03:37 -04002026 return 1;
Jan Kara262b4662013-03-02 17:57:08 -05002027 }
Jan Kara5f3633e2013-02-02 23:09:36 -05002028 sbi->s_jquota_fmt = m->mount_opt;
2029#endif
Ross Zwisler923ae0f2015-02-16 15:59:38 -08002030 } else if (token == Opt_dax) {
Dan Williamsef83b6e2015-09-29 15:48:11 -04002031#ifdef CONFIG_FS_DAX
2032 ext4_msg(sb, KERN_WARNING,
2033 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
Colin Ian Kinga92abd72018-12-04 00:16:44 -05002034 sbi->s_mount_opt |= m->mount_opt;
Dan Williamsef83b6e2015-09-29 15:48:11 -04002035#else
Ross Zwisler923ae0f2015-02-16 15:59:38 -08002036 ext4_msg(sb, KERN_INFO, "dax option not supported");
2037 return -1;
2038#endif
Ales Novak7915a862016-03-12 21:55:50 -05002039 } else if (token == Opt_data_err_abort) {
2040 sbi->s_mount_opt |= m->mount_opt;
2041 } else if (token == Opt_data_err_ignore) {
2042 sbi->s_mount_opt &= ~m->mount_opt;
Jan Kara5f3633e2013-02-02 23:09:36 -05002043 } else {
2044 if (!args->from)
2045 arg = 1;
2046 if (m->flags & MOPT_CLEAR)
2047 arg = !arg;
2048 else if (unlikely(!(m->flags & MOPT_SET))) {
2049 ext4_msg(sb, KERN_WARNING,
2050 "buggy handling of option %s", opt);
2051 WARN_ON(1);
2052 return -1;
2053 }
2054 if (arg != 0)
2055 sbi->s_mount_opt |= m->mount_opt;
2056 else
2057 sbi->s_mount_opt &= ~m->mount_opt;
2058 }
2059 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05002060}
2061
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002062static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05002063 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05002064 unsigned int *journal_ioprio,
Eric Sandeen661aa522012-02-20 17:53:04 -05002065 int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002066{
Olof Johansson1e1a76e2019-11-10 18:25:23 -08002067 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
Theodore Ts'o33458ea2018-10-12 09:28:09 -04002068 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002069 substring_t args[MAX_OPT_ARGS];
Theodore Ts'o26092bf2012-03-03 23:20:47 -05002070 int token;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002071
2072 if (!options)
2073 return 1;
2074
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002075 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002076 if (!*p)
2077 continue;
Eric Sandeen15121c12010-02-15 20:17:55 -05002078 /*
2079 * Initialize args struct so we know whether arg was
2080 * found; some options take optional arguments.
2081 */
Sachin Kamatcaecd0a2012-08-18 22:29:18 -04002082 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002083 token = match_token(p, tokens, args);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05002084 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2085 journal_ioprio, is_remount) < 0)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002086 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002087 }
2088#ifdef CONFIG_QUOTA
Jan Kara49da9392016-09-05 23:08:16 -04002089 /*
2090 * We do the test below only for project quotas. 'usrquota' and
2091 * 'grpquota' mount options are allowed even without quota feature
2092 * to support legacy quotas in quota files.
2093 */
2094 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2095 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2096 "Cannot enable project quota enforcement.");
2097 return 0;
2098 }
Theodore Ts'o33458ea2018-10-12 09:28:09 -04002099 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2100 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2101 if (usr_qf_name || grp_qf_name) {
2102 if (test_opt(sb, USRQUOTA) && usr_qf_name)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05002103 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002104
Theodore Ts'o33458ea2018-10-12 09:28:09 -04002105 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05002106 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002107
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05002108 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002109 ext4_msg(sb, KERN_ERR, "old and new quota "
2110 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002111 return 0;
2112 }
2113
2114 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002115 ext4_msg(sb, KERN_ERR, "journaled quota format "
2116 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002117 return 0;
2118 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002119 }
2120#endif
2121 return 1;
2122}
2123
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002124static inline void ext4_show_quota_options(struct seq_file *seq,
2125 struct super_block *sb)
2126{
2127#if defined(CONFIG_QUOTA)
2128 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o33458ea2018-10-12 09:28:09 -04002129 char *usr_qf_name, *grp_qf_name;
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002130
2131 if (sbi->s_jquota_fmt) {
2132 char *fmtname = "";
2133
2134 switch (sbi->s_jquota_fmt) {
2135 case QFMT_VFS_OLD:
2136 fmtname = "vfsold";
2137 break;
2138 case QFMT_VFS_V0:
2139 fmtname = "vfsv0";
2140 break;
2141 case QFMT_VFS_V1:
2142 fmtname = "vfsv1";
2143 break;
2144 }
2145 seq_printf(seq, ",jqfmt=%s", fmtname);
2146 }
2147
Theodore Ts'o33458ea2018-10-12 09:28:09 -04002148 rcu_read_lock();
2149 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2150 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2151 if (usr_qf_name)
2152 seq_show_option(seq, "usrjquota", usr_qf_name);
2153 if (grp_qf_name)
2154 seq_show_option(seq, "grpjquota", grp_qf_name);
2155 rcu_read_unlock();
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002156#endif
2157}
2158
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002159static const char *token2str(int token)
2160{
Herton Ronaldo Krzesinski50df9fd2012-09-23 22:49:12 -04002161 const struct match_token *t;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002162
2163 for (t = tokens; t->token != Opt_err; t++)
2164 if (t->token == token && !strchr(t->pattern, '='))
2165 break;
2166 return t->pattern;
2167}
2168
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002169/*
2170 * Show an option if
2171 * - it's set to a non-default value OR
2172 * - if the per-sb default is different from the global default
2173 */
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002174static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2175 int nodefs)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002176{
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002177 struct ext4_sb_info *sbi = EXT4_SB(sb);
2178 struct ext4_super_block *es = sbi->s_es;
Tyson Nottingham68afa7e2018-03-30 00:51:10 -04002179 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002180 const struct mount_opts *m;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002181 char sep = nodefs ? '\n' : ',';
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002182
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002183#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2184#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002185
2186 if (sbi->s_sb_block != 1)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002187 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2188
2189 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2190 int want_set = m->flags & MOPT_SET;
2191 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2192 (m->flags & MOPT_CLEAR_ERR))
2193 continue;
Tyson Nottingham68afa7e2018-03-30 00:51:10 -04002194 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002195 continue; /* skip if same as the default */
2196 if ((want_set &&
2197 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2198 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2199 continue; /* select Opt_noFoo vs Opt_Foo */
2200 SEQ_OPTS_PRINT("%s", token2str(m->token));
2201 }
2202
Eric W. Biederman08cefc72012-02-07 15:41:49 -08002203 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002204 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08002205 SEQ_OPTS_PRINT("resuid=%u",
2206 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2207 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002208 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08002209 SEQ_OPTS_PRINT("resgid=%u",
2210 from_kgid_munged(&init_user_ns, sbi->s_resgid));
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002211 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002212 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2213 SEQ_OPTS_PUTS("errors=remount-ro");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002214 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002215 SEQ_OPTS_PUTS("errors=continue");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002216 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002217 SEQ_OPTS_PUTS("errors=panic");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002218 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002219 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002220 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002221 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002222 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002223 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
Matthew Garrett357fdad2017-10-18 13:56:26 -07002224 if (sb->s_flags & SB_I_VERSION)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002225 SEQ_OPTS_PUTS("i_version");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002226 if (nodefs || sbi->s_stripe)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002227 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
Tyson Nottingham68afa7e2018-03-30 00:51:10 -04002228 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2229 (sbi->s_mount_opt ^ def_mount_opt)) {
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002230 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2231 SEQ_OPTS_PUTS("data=journal");
2232 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2233 SEQ_OPTS_PUTS("data=ordered");
2234 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2235 SEQ_OPTS_PUTS("data=writeback");
2236 }
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002237 if (nodefs ||
2238 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002239 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2240 sbi->s_inode_readahead_blks);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002241
Tyson Nottinghamceec0372018-03-30 00:53:33 -04002242 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002243 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002244 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
Theodore Ts'odf981d02012-08-17 09:48:17 -04002245 if (nodefs || sbi->s_max_dir_size_kb)
2246 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
Ales Novak7915a862016-03-12 21:55:50 -05002247 if (test_opt(sb, DATA_ERR_ABORT))
2248 SEQ_OPTS_PUTS("data_err=abort");
Eric Biggers338affb2018-09-15 14:28:26 -04002249 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2250 SEQ_OPTS_PUTS("test_dummy_encryption");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002251
2252 ext4_show_quota_options(seq, sb);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002253 return 0;
2254}
2255
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002256static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2257{
2258 return _ext4_show_options(seq, root->d_sb, 0);
2259}
2260
Theodore Ts'oebd173b2015-09-23 12:46:17 -04002261int ext4_seq_options_show(struct seq_file *seq, void *offset)
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002262{
2263 struct super_block *sb = seq->private;
2264 int rc;
2265
David Howellsbc98a422017-07-17 08:45:34 +01002266 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002267 rc = _ext4_show_options(seq, sb, 1);
2268 seq_puts(seq, "\n");
2269 return rc;
2270}
2271
Mingming Cao617ba132006-10-11 01:20:53 -07002272static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002273 int read_only)
2274{
Mingming Cao617ba132006-10-11 01:20:53 -07002275 struct ext4_sb_info *sbi = EXT4_SB(sb);
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04002276 int err = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002277
Mingming Cao617ba132006-10-11 01:20:53 -07002278 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002279 ext4_msg(sb, KERN_ERR, "revision level too high, "
2280 "forcing read-only mode");
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04002281 err = -EROFS;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002282 }
2283 if (read_only)
Theodore Ts'o281b5992011-09-09 18:34:51 -04002284 goto done;
Mingming Cao617ba132006-10-11 01:20:53 -07002285 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002286 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2287 "running e2fsck is recommended");
Lukas Czernerc8b459f2014-05-12 12:55:07 -04002288 else if (sbi->s_mount_state & EXT4_ERROR_FS)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002289 ext4_msg(sb, KERN_WARNING,
2290 "warning: mounting fs with errors, "
2291 "running e2fsck is recommended");
Tao Maed3ce802011-05-18 13:29:57 -04002292 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002293 le16_to_cpu(es->s_mnt_count) >=
2294 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002295 ext4_msg(sb, KERN_WARNING,
2296 "warning: maximal mount count reached, "
2297 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002298 else if (le32_to_cpu(es->s_checkinterval) &&
Arnd Bergmann6a0678a2018-07-29 15:51:48 -04002299 (ext4_get_tstamp(es, s_lastcheck) +
2300 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002301 ext4_msg(sb, KERN_WARNING,
2302 "warning: checktime reached, "
2303 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002304 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05002305 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002306 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07002307 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04002308 le16_add_cpu(&es->s_mnt_count, 1);
Arnd Bergmann6a0678a2018-07-29 15:51:48 -04002309 ext4_update_tstamp(es, s_mtime);
Frank Mayhar03901312009-01-07 00:06:22 -05002310 if (sbi->s_journal)
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002311 ext4_set_feature_journal_needs_recovery(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002312
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04002313 err = ext4_commit_super(sb, 1);
Theodore Ts'o281b5992011-09-09 18:34:51 -04002314done:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002315 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002316 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05002317 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002318 sb->s_blocksize,
2319 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07002320 EXT4_BLOCKS_PER_GROUP(sb),
2321 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05002322 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002323
Dan Magenheimer7abc52c2011-05-26 10:02:03 -06002324 cleancache_init_fs(sb);
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04002325 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002326}
2327
Theodore Ts'o117fff12012-09-05 01:29:50 -04002328int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2329{
2330 struct ext4_sb_info *sbi = EXT4_SB(sb);
2331 struct flex_groups *new_groups;
2332 int size;
2333
2334 if (!sbi->s_log_groups_per_flex)
2335 return 0;
2336
2337 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2338 if (size <= sbi->s_flex_groups_allocated)
2339 return 0;
2340
2341 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
Michal Hockoa7c3e902017-05-08 15:57:09 -07002342 new_groups = kvzalloc(size, GFP_KERNEL);
Theodore Ts'o117fff12012-09-05 01:29:50 -04002343 if (!new_groups) {
2344 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2345 size / (int) sizeof(struct flex_groups));
2346 return -ENOMEM;
2347 }
2348
2349 if (sbi->s_flex_groups) {
2350 memcpy(new_groups, sbi->s_flex_groups,
2351 (sbi->s_flex_groups_allocated *
2352 sizeof(struct flex_groups)));
Al Virob93b41d2014-11-20 12:19:11 -05002353 kvfree(sbi->s_flex_groups);
Theodore Ts'o117fff12012-09-05 01:29:50 -04002354 }
2355 sbi->s_flex_groups = new_groups;
2356 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2357 return 0;
2358}
2359
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002360static int ext4_fill_flex_info(struct super_block *sb)
2361{
2362 struct ext4_sb_info *sbi = EXT4_SB(sb);
2363 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002364 ext4_group_t flex_group;
Theodore Ts'o117fff12012-09-05 01:29:50 -04002365 int i, err;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002366
Theodore Ts'o503358a2009-11-23 07:24:46 -05002367 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
Xi Wangd50f2ab2012-01-10 11:51:10 -05002368 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002369 sbi->s_log_groups_per_flex = 0;
2370 return 1;
2371 }
2372
Theodore Ts'o117fff12012-09-05 01:29:50 -04002373 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2374 if (err)
Theodore Ts'o9933fc02011-08-01 08:45:02 -04002375 goto failed;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002376
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002377 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04002378 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002379
2380 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002381 atomic_add(ext4_free_inodes_count(sb, gdp),
2382 &sbi->s_flex_groups[flex_group].free_inodes);
Theodore Ts'o90ba9832013-03-11 23:39:59 -04002383 atomic64_add(ext4_free_group_clusters(sb, gdp),
2384 &sbi->s_flex_groups[flex_group].free_clusters);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002385 atomic_add(ext4_used_dirs_count(sb, gdp),
2386 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002387 }
2388
2389 return 1;
2390failed:
2391 return 0;
2392}
2393
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002394static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002395 struct ext4_group_desc *gdp)
Andreas Dilger717d50e2007-10-16 18:38:25 -04002396{
Daeho Jeongb47820e2016-07-03 17:51:39 -04002397 int offset = offsetof(struct ext4_group_desc, bg_checksum);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002398 __u16 crc = 0;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002399 __le32 le_group = cpu_to_le32(block_group);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002400 struct ext4_sb_info *sbi = EXT4_SB(sb);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002401
Dmitry Monakhov9aa5d32b2014-10-13 03:36:16 -04002402 if (ext4_has_metadata_csum(sbi->s_sb)) {
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002403 /* Use new metadata_csum algorithm */
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002404 __u32 csum32;
Daeho Jeongb47820e2016-07-03 17:51:39 -04002405 __u16 dummy_csum = 0;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002406
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002407 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2408 sizeof(le_group));
Daeho Jeongb47820e2016-07-03 17:51:39 -04002409 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2410 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2411 sizeof(dummy_csum));
2412 offset += sizeof(dummy_csum);
2413 if (offset < sbi->s_desc_size)
2414 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2415 sbi->s_desc_size - offset);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002416
2417 crc = csum32 & 0xFFFF;
2418 goto out;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002419 }
2420
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002421 /* old crc16 code */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002422 if (!ext4_has_feature_gdt_csum(sb))
Darrick J. Wong813d32f2014-10-14 02:35:49 -04002423 return 0;
2424
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002425 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2426 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2427 crc = crc16(crc, (__u8 *)gdp, offset);
2428 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2429 /* for checksum of struct ext4_group_desc do the rest...*/
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002430 if (ext4_has_feature_64bit(sb) &&
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002431 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2432 crc = crc16(crc, (__u8 *)gdp + offset,
2433 le16_to_cpu(sbi->s_es->s_desc_size) -
2434 offset);
2435
2436out:
Andreas Dilger717d50e2007-10-16 18:38:25 -04002437 return cpu_to_le16(crc);
2438}
2439
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002440int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
Andreas Dilger717d50e2007-10-16 18:38:25 -04002441 struct ext4_group_desc *gdp)
2442{
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002443 if (ext4_has_group_desc_csum(sb) &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002444 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
Andreas Dilger717d50e2007-10-16 18:38:25 -04002445 return 0;
2446
2447 return 1;
2448}
2449
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002450void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2451 struct ext4_group_desc *gdp)
2452{
2453 if (!ext4_has_group_desc_csum(sb))
2454 return;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002455 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002456}
2457
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002458/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002459static int ext4_check_descriptors(struct super_block *sb,
Theodore Ts'o829fa702016-08-01 00:51:02 -04002460 ext4_fsblk_t sb_block,
Lukas Czernerbfff6872010-10-27 21:30:05 -04002461 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002462{
Mingming Cao617ba132006-10-11 01:20:53 -07002463 struct ext4_sb_info *sbi = EXT4_SB(sb);
2464 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2465 ext4_fsblk_t last_block;
Theodore Ts'o44de0222018-07-08 19:35:02 -04002466 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002467 ext4_fsblk_t block_bitmap;
2468 ext4_fsblk_t inode_bitmap;
2469 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002470 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002471 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002472
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002473 if (ext4_has_feature_flex_bg(sb))
Jose R. Santosce421582007-10-16 18:38:25 -04002474 flexbg_flag = 1;
2475
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002476 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002477
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002478 for (i = 0; i < sbi->s_groups_count; i++) {
2479 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2480
Jose R. Santosce421582007-10-16 18:38:25 -04002481 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002482 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002483 else
2484 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002485 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002486
Lukas Czernerbfff6872010-10-27 21:30:05 -04002487 if ((grp == sbi->s_groups_count) &&
2488 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2489 grp = i;
2490
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002491 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o829fa702016-08-01 00:51:02 -04002492 if (block_bitmap == sb_block) {
2493 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2494 "Block bitmap for group %u overlaps "
2495 "superblock", i);
Theodore Ts'o18db4b42018-03-29 22:10:35 -04002496 if (!sb_rdonly(sb))
2497 return 0;
Theodore Ts'o829fa702016-08-01 00:51:02 -04002498 }
Theodore Ts'o77260802018-06-13 23:08:26 -04002499 if (block_bitmap >= sb_block + 1 &&
2500 block_bitmap <= last_bg_block) {
2501 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2502 "Block bitmap for group %u overlaps "
2503 "block group descriptors", i);
2504 if (!sb_rdonly(sb))
2505 return 0;
2506 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002507 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002508 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002509 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002510 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002511 return 0;
2512 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002513 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o829fa702016-08-01 00:51:02 -04002514 if (inode_bitmap == sb_block) {
2515 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2516 "Inode bitmap for group %u overlaps "
2517 "superblock", i);
Theodore Ts'o18db4b42018-03-29 22:10:35 -04002518 if (!sb_rdonly(sb))
2519 return 0;
Theodore Ts'o829fa702016-08-01 00:51:02 -04002520 }
Theodore Ts'o77260802018-06-13 23:08:26 -04002521 if (inode_bitmap >= sb_block + 1 &&
2522 inode_bitmap <= last_bg_block) {
2523 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2524 "Inode bitmap for group %u overlaps "
2525 "block group descriptors", i);
2526 if (!sb_rdonly(sb))
2527 return 0;
2528 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002529 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002530 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002531 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002532 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002533 return 0;
2534 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002535 inode_table = ext4_inode_table(sb, gdp);
Theodore Ts'o829fa702016-08-01 00:51:02 -04002536 if (inode_table == sb_block) {
2537 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2538 "Inode table for group %u overlaps "
2539 "superblock", i);
Theodore Ts'o18db4b42018-03-29 22:10:35 -04002540 if (!sb_rdonly(sb))
2541 return 0;
Theodore Ts'o829fa702016-08-01 00:51:02 -04002542 }
Theodore Ts'o77260802018-06-13 23:08:26 -04002543 if (inode_table >= sb_block + 1 &&
2544 inode_table <= last_bg_block) {
2545 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2546 "Inode table for group %u overlaps "
2547 "block group descriptors", i);
2548 if (!sb_rdonly(sb))
2549 return 0;
2550 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002551 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002552 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002553 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002554 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002555 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002556 return 0;
2557 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002558 ext4_lock_group(sb, i);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002559 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002560 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2561 "Checksum for group %u failed (%u!=%u)",
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002562 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
Eric Sandeenb31e1552009-06-04 17:36:36 -04002563 gdp)), le16_to_cpu(gdp->bg_checksum));
David Howellsbc98a422017-07-17 08:45:34 +01002564 if (!sb_rdonly(sb)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002565 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002566 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002567 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002568 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002569 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002570 if (!flexbg_flag)
2571 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002572 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002573 if (NULL != first_not_zeroed)
2574 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002575 return 1;
2576}
2577
Mingming Cao617ba132006-10-11 01:20:53 -07002578/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002579 * the superblock) which were deleted from all directories, but held open by
2580 * a process at the time of a crash. We walk the list and try to delete these
2581 * inodes at recovery time (only with a read-write filesystem).
2582 *
2583 * In order to keep the orphan inode chain consistent during traversal (in
2584 * case of crash during recovery), we link each inode into the superblock
2585 * orphan list_head and handle it the same way as an inode deletion during
2586 * normal operation (which journals the operations for us).
2587 *
2588 * We only do an iget() and an iput() on each inode, which is very safe if we
2589 * accidentally point at an in-use or already deleted inode. The worst that
2590 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002591 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002592 * e2fsck was run on this filesystem, and it must have already done the orphan
2593 * inode cleanup for us, so we can safely abort without any further action.
2594 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002595static void ext4_orphan_cleanup(struct super_block *sb,
2596 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002597{
2598 unsigned int s_flags = sb->s_flags;
Theodore Ts'o2c98eb52016-11-13 22:02:26 -05002599 int ret, nr_orphans = 0, nr_truncates = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002600#ifdef CONFIG_QUOTA
zhangyi (F)95f1fda2017-08-24 15:21:50 -04002601 int quota_update = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002602 int i;
2603#endif
2604 if (!es->s_last_orphan) {
2605 jbd_debug(4, "no orphan inodes to clean up\n");
2606 return;
2607 }
2608
Eric Sandeena8f48a92006-12-06 20:40:13 -08002609 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002610 ext4_msg(sb, KERN_ERR, "write access "
2611 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002612 return;
2613 }
2614
Amir Goldsteind39195c2011-02-28 00:53:45 -05002615 /* Check if feature set would not allow a r/w mount */
2616 if (!ext4_feature_set_ok(sb, 0)) {
2617 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2618 "unknown ROCOMPAT features");
2619 return;
2620 }
2621
Mingming Cao617ba132006-10-11 01:20:53 -07002622 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002623 /* don't clear list on RO mount w/ errors */
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002624 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
Dmitry Monakhov84474972014-09-16 14:52:03 -04002625 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002626 "clearing orphan list.\n");
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002627 es->s_last_orphan = 0;
2628 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002629 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2630 return;
2631 }
2632
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002633 if (s_flags & SB_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002634 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002635 sb->s_flags &= ~SB_RDONLY;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002636 }
2637#ifdef CONFIG_QUOTA
2638 /* Needed for iput() to work correctly and not trash data */
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002639 sb->s_flags |= SB_ACTIVE;
zhangyi (F)95f1fda2017-08-24 15:21:50 -04002640
2641 /*
2642 * Turn on quotas which were not enabled for read-only mounts if
2643 * filesystem has quota feature, so that they are updated correctly.
2644 */
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002645 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
zhangyi (F)95f1fda2017-08-24 15:21:50 -04002646 int ret = ext4_enable_quotas(sb);
2647
2648 if (!ret)
2649 quota_update = 1;
2650 else
2651 ext4_msg(sb, KERN_ERR,
2652 "Cannot turn on quotas: error %d", ret);
2653 }
2654
2655 /* Turn on journaled quotas used for old sytle */
Jan Karaa2d4a642014-09-11 11:15:15 -04002656 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002657 if (EXT4_SB(sb)->s_qf_names[i]) {
2658 int ret = ext4_quota_on_mount(sb, i);
zhangyi (F)95f1fda2017-08-24 15:21:50 -04002659
2660 if (!ret)
2661 quota_update = 1;
2662 else
Eric Sandeenb31e1552009-06-04 17:36:36 -04002663 ext4_msg(sb, KERN_ERR,
2664 "Cannot turn on journaled "
zhangyi (F)95f1fda2017-08-24 15:21:50 -04002665 "quota: type %d: error %d", i, ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002666 }
2667 }
2668#endif
2669
2670 while (es->s_last_orphan) {
2671 struct inode *inode;
2672
Vegard Nossumc65d5c62016-07-14 23:21:35 -04002673 /*
2674 * We may have encountered an error during cleanup; if
2675 * so, skip the rest.
2676 */
2677 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2678 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2679 es->s_last_orphan = 0;
2680 break;
2681 }
2682
Josef Bacik97bd42b2008-04-29 22:04:56 -04002683 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2684 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002685 es->s_last_orphan = 0;
2686 break;
2687 }
2688
Mingming Cao617ba132006-10-11 01:20:53 -07002689 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002690 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002691 if (inode->i_nlink) {
Paul Taysom566370a2013-05-28 07:51:21 -04002692 if (test_opt(sb, DEBUG))
2693 ext4_msg(sb, KERN_DEBUG,
2694 "%s: truncating inode %lu to %lld bytes",
2695 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002696 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002697 inode->i_ino, inode->i_size);
Al Viro59551022016-01-22 15:40:57 -05002698 inode_lock(inode);
Lukas Czerner55f252c2013-05-27 23:32:35 -04002699 truncate_inode_pages(inode->i_mapping, inode->i_size);
Theodore Ts'o2c98eb52016-11-13 22:02:26 -05002700 ret = ext4_truncate(inode);
2701 if (ret)
2702 ext4_std_error(inode->i_sb, ret);
Al Viro59551022016-01-22 15:40:57 -05002703 inode_unlock(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002704 nr_truncates++;
2705 } else {
Paul Taysom566370a2013-05-28 07:51:21 -04002706 if (test_opt(sb, DEBUG))
2707 ext4_msg(sb, KERN_DEBUG,
2708 "%s: deleting unreferenced inode %lu",
2709 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002710 jbd_debug(2, "deleting unreferenced inode %lu\n",
2711 inode->i_ino);
2712 nr_orphans++;
2713 }
2714 iput(inode); /* The delete magic happens here! */
2715 }
2716
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002717#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002718
2719 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002720 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2721 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002722 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002723 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2724 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002725#ifdef CONFIG_QUOTA
zhangyi (F)95f1fda2017-08-24 15:21:50 -04002726 /* Turn off quotas if they were enabled for orphan cleanup */
2727 if (quota_update) {
2728 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2729 if (sb_dqopt(sb)->files[i])
2730 dquot_quota_off(sb, i);
2731 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002732 }
2733#endif
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002734 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002735}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002736
Eric Sandeencd2291a2008-01-28 23:58:27 -05002737/*
2738 * Maximal extent format file size.
2739 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2740 * extent format containers, within a sector_t, and within i_blocks
2741 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2742 * so that won't be a limiting factor.
2743 *
Lukas Czernerf17722f2011-06-06 00:05:17 -04002744 * However there is other limiting factor. We do store extents in the form
2745 * of starting block and length, hence the resulting length of the extent
2746 * covering maximum file size must fit into on-disk format containers as
2747 * well. Given that length is always by 1 unit bigger than max unit (because
2748 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2749 *
Eric Sandeencd2291a2008-01-28 23:58:27 -05002750 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2751 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002752static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002753{
2754 loff_t res;
2755 loff_t upper_limit = MAX_LFS_FILESIZE;
2756
Christoph Hellwig72deb452019-04-05 18:08:59 +02002757 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2758
2759 if (!has_huge_files) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002760 upper_limit = (1LL << 32) - 1;
2761
2762 /* total blocks in file system block size */
2763 upper_limit >>= (blkbits - 9);
2764 upper_limit <<= blkbits;
2765 }
2766
Lukas Czernerf17722f2011-06-06 00:05:17 -04002767 /*
2768 * 32-bit extent-start container, ee_block. We lower the maxbytes
2769 * by one fs block, so ee_len can cover the extent of maximum file
2770 * size
2771 */
2772 res = (1LL << 32) - 1;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002773 res <<= blkbits;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002774
2775 /* Sanity check against vm- & vfs- imposed limits */
2776 if (res > upper_limit)
2777 res = upper_limit;
2778
2779 return res;
2780}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002781
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002782/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002783 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002784 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2785 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002786 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002787static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002788{
Mingming Cao617ba132006-10-11 01:20:53 -07002789 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002790 int meta_blocks;
2791 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002792 /* This is calculated to be the largest file size for a dense, block
2793 * mapped file such that the file's total number of 512-byte sectors,
2794 * including data and all indirect blocks, does not exceed (2^48 - 1).
2795 *
2796 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2797 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002798 */
2799
Christoph Hellwig72deb452019-04-05 18:08:59 +02002800 if (!has_huge_files) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002801 /*
Christoph Hellwig72deb452019-04-05 18:08:59 +02002802 * !has_huge_files or implies that the inode i_block field
2803 * represents total file blocks in 2^32 512-byte sectors ==
2804 * size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002805 */
2806 upper_limit = (1LL << 32) - 1;
2807
2808 /* total blocks in file system block size */
2809 upper_limit >>= (bits - 9);
2810
2811 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002812 /*
2813 * We use 48 bit ext4_inode i_blocks
2814 * With EXT4_HUGE_FILE_FL set the i_blocks
2815 * represent total number of blocks in
2816 * file system block size
2817 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002818 upper_limit = (1LL << 48) - 1;
2819
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002820 }
2821
2822 /* indirect blocks */
2823 meta_blocks = 1;
2824 /* double indirect blocks */
2825 meta_blocks += 1 + (1LL << (bits-2));
2826 /* tripple indirect blocks */
2827 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2828
2829 upper_limit -= meta_blocks;
2830 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002831
2832 res += 1LL << (bits-2);
2833 res += 1LL << (2*(bits-2));
2834 res += 1LL << (3*(bits-2));
2835 res <<= bits;
2836 if (res > upper_limit)
2837 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002838
2839 if (res > MAX_LFS_FILESIZE)
2840 res = MAX_LFS_FILESIZE;
2841
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002842 return res;
2843}
2844
Mingming Cao617ba132006-10-11 01:20:53 -07002845static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002846 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002847{
Mingming Cao617ba132006-10-11 01:20:53 -07002848 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002849 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002850 int has_super = 0;
2851
2852 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2853
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002854 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002855 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002856 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002857 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002858 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002859
Darrick J. Wongbd63f6b2014-05-12 10:06:27 -04002860 /*
2861 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2862 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2863 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2864 * compensate.
2865 */
2866 if (sb->s_blocksize == 1024 && nr == 0 &&
Jun Piao49598e02018-01-11 13:17:49 -05002867 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
Darrick J. Wongbd63f6b2014-05-12 10:06:27 -04002868 has_super++;
2869
Mingming Cao617ba132006-10-11 01:20:53 -07002870 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002871}
2872
Alex Tomasc9de5602008-01-29 00:19:52 -05002873/**
2874 * ext4_get_stripe_size: Get the stripe size.
2875 * @sbi: In memory super block info
2876 *
2877 * If we have specified it via mount option, then
2878 * use the mount option value. If the value specified at mount time is
2879 * greater than the blocks per group use the super block value.
2880 * If the super block value is greater than blocks per group return 0.
2881 * Allocator needs it be less than blocks per group.
2882 *
2883 */
2884static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2885{
2886 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2887 unsigned long stripe_width =
2888 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002889 int ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002890
2891 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002892 ret = sbi->s_stripe;
Jan Kara5469d7c2017-02-10 00:56:09 -05002893 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002894 ret = stripe_width;
Jan Kara5469d7c2017-02-10 00:56:09 -05002895 else if (stride && stride <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002896 ret = stride;
2897 else
2898 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002899
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002900 /*
2901 * If the stripe width is 1, this makes no sense and
2902 * we set it to 0 to turn off stripe handling code.
2903 */
2904 if (ret <= 1)
2905 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002906
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002907 return ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002908}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002909
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002910/*
2911 * Check whether this filesystem can be mounted based on
2912 * the features present and the RDONLY/RDWR mount requested.
2913 * Returns 1 if this filesystem can be mounted as requested,
2914 * 0 if it cannot be.
2915 */
2916static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2917{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002918 if (ext4_has_unknown_ext4_incompat_features(sb)) {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002919 ext4_msg(sb, KERN_ERR,
2920 "Couldn't mount because of "
2921 "unsupported optional features (%x)",
2922 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2923 ~EXT4_FEATURE_INCOMPAT_SUPP));
2924 return 0;
2925 }
2926
Gabriel Krisman Bertazic83ad552019-04-25 14:05:42 -04002927#ifndef CONFIG_UNICODE
2928 if (ext4_has_feature_casefold(sb)) {
2929 ext4_msg(sb, KERN_ERR,
2930 "Filesystem with casefold feature cannot be "
2931 "mounted without CONFIG_UNICODE");
2932 return 0;
2933 }
2934#endif
2935
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002936 if (readonly)
2937 return 1;
2938
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002939 if (ext4_has_feature_readonly(sb)) {
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05002940 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
Linus Torvalds1751e8a2017-11-27 13:05:09 -08002941 sb->s_flags |= SB_RDONLY;
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05002942 return 1;
2943 }
2944
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002945 /* Check that feature set is OK for a read-write mount */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002946 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002947 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2948 "unsupported optional features (%x)",
2949 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2950 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2951 return 0;
2952 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002953 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
Theodore Ts'obab08ab2011-09-09 18:36:51 -04002954 ext4_msg(sb, KERN_ERR,
2955 "Can't support bigalloc feature without "
2956 "extents feature\n");
2957 return 0;
2958 }
Aditya Kali7c319d32012-07-22 20:21:31 -04002959
2960#ifndef CONFIG_QUOTA
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002961 if (ext4_has_feature_quota(sb) && !readonly) {
Aditya Kali7c319d32012-07-22 20:21:31 -04002962 ext4_msg(sb, KERN_ERR,
2963 "Filesystem with quota feature cannot be mounted RDWR "
2964 "without CONFIG_QUOTA");
2965 return 0;
2966 }
Li Xi689c958c2016-01-08 16:01:22 -05002967 if (ext4_has_feature_project(sb) && !readonly) {
2968 ext4_msg(sb, KERN_ERR,
2969 "Filesystem with project quota feature cannot be mounted RDWR "
2970 "without CONFIG_QUOTA");
2971 return 0;
2972 }
Aditya Kali7c319d32012-07-22 20:21:31 -04002973#endif /* CONFIG_QUOTA */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002974 return 1;
2975}
2976
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002977/*
2978 * This function is called once a day if we have errors logged
2979 * on the file system
2980 */
Kees Cook235699a2017-10-18 12:45:17 -04002981static void print_daily_error_info(struct timer_list *t)
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002982{
Kees Cook235699a2017-10-18 12:45:17 -04002983 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2984 struct super_block *sb = sbi->s_sb;
2985 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002986
2987 if (es->s_error_count)
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002988 /* fsck newer than v1.41.13 is needed to clean this condition. */
2989 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002990 le32_to_cpu(es->s_error_count));
2991 if (es->s_first_error_time) {
Arnd Bergmann6a0678a2018-07-29 15:51:48 -04002992 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2993 sb->s_id,
2994 ext4_get_tstamp(es, s_first_error_time),
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002995 (int) sizeof(es->s_first_error_func),
2996 es->s_first_error_func,
2997 le32_to_cpu(es->s_first_error_line));
2998 if (es->s_first_error_ino)
Joe Perches651e1c32016-10-12 23:12:53 -04002999 printk(KERN_CONT ": inode %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003000 le32_to_cpu(es->s_first_error_ino));
3001 if (es->s_first_error_block)
Joe Perches651e1c32016-10-12 23:12:53 -04003002 printk(KERN_CONT ": block %llu", (unsigned long long)
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003003 le64_to_cpu(es->s_first_error_block));
Joe Perches651e1c32016-10-12 23:12:53 -04003004 printk(KERN_CONT "\n");
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003005 }
3006 if (es->s_last_error_time) {
Arnd Bergmann6a0678a2018-07-29 15:51:48 -04003007 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3008 sb->s_id,
3009 ext4_get_tstamp(es, s_last_error_time),
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003010 (int) sizeof(es->s_last_error_func),
3011 es->s_last_error_func,
3012 le32_to_cpu(es->s_last_error_line));
3013 if (es->s_last_error_ino)
Joe Perches651e1c32016-10-12 23:12:53 -04003014 printk(KERN_CONT ": inode %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003015 le32_to_cpu(es->s_last_error_ino));
3016 if (es->s_last_error_block)
Joe Perches651e1c32016-10-12 23:12:53 -04003017 printk(KERN_CONT ": block %llu", (unsigned long long)
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003018 le64_to_cpu(es->s_last_error_block));
Joe Perches651e1c32016-10-12 23:12:53 -04003019 printk(KERN_CONT "\n");
Theodore Ts'o66e61a92010-07-27 11:56:04 -04003020 }
3021 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3022}
3023
Lukas Czernerbfff6872010-10-27 21:30:05 -04003024/* Find next suitable group and run ext4_init_inode_table */
3025static int ext4_run_li_request(struct ext4_li_request *elr)
3026{
3027 struct ext4_group_desc *gdp = NULL;
3028 ext4_group_t group, ngroups;
3029 struct super_block *sb;
3030 unsigned long timeout = 0;
3031 int ret = 0;
3032
3033 sb = elr->lr_super;
3034 ngroups = EXT4_SB(sb)->s_groups_count;
3035
3036 for (group = elr->lr_next_group; group < ngroups; group++) {
3037 gdp = ext4_get_group_desc(sb, group, NULL);
3038 if (!gdp) {
3039 ret = 1;
3040 break;
3041 }
3042
3043 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3044 break;
3045 }
3046
Theodore Ts'o7f511862013-01-13 08:41:45 -05003047 if (group >= ngroups)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003048 ret = 1;
3049
3050 if (!ret) {
3051 timeout = jiffies;
3052 ret = ext4_init_inode_table(sb, group,
3053 elr->lr_timeout ? 0 : 1);
3054 if (elr->lr_timeout == 0) {
Lukas Czerner51ce6512011-05-20 13:55:16 -04003055 timeout = (jiffies - timeout) *
3056 elr->lr_sbi->s_li_wait_mult;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003057 elr->lr_timeout = timeout;
3058 }
3059 elr->lr_next_sched = jiffies + elr->lr_timeout;
3060 elr->lr_next_group = group + 1;
3061 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003062 return ret;
3063}
3064
3065/*
3066 * Remove lr_request from the list_request and free the
Lukas Czerner4ed5c032011-05-20 13:49:04 -04003067 * request structure. Should be called with li_list_mtx held
Lukas Czernerbfff6872010-10-27 21:30:05 -04003068 */
3069static void ext4_remove_li_request(struct ext4_li_request *elr)
3070{
3071 struct ext4_sb_info *sbi;
3072
3073 if (!elr)
3074 return;
3075
3076 sbi = elr->lr_sbi;
3077
3078 list_del(&elr->lr_request);
3079 sbi->s_li_request = NULL;
3080 kfree(elr);
3081}
3082
3083static void ext4_unregister_li_request(struct super_block *sb)
3084{
Lukas Czerner1bb933f2011-05-20 13:55:29 -04003085 mutex_lock(&ext4_li_mtx);
3086 if (!ext4_li_info) {
3087 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003088 return;
Lukas Czerner1bb933f2011-05-20 13:55:29 -04003089 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003090
3091 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04003092 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003093 mutex_unlock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04003094 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003095}
3096
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003097static struct task_struct *ext4_lazyinit_task;
3098
Lukas Czernerbfff6872010-10-27 21:30:05 -04003099/*
3100 * This is the function where ext4lazyinit thread lives. It walks
3101 * through the request list searching for next scheduled filesystem.
3102 * When such a fs is found, run the lazy initialization request
3103 * (ext4_rn_li_request) and keep track of the time spend in this
3104 * function. Based on that time we compute next schedule time of
3105 * the request. When walking through the list is complete, compute
3106 * next waking time and put itself into sleep.
3107 */
3108static int ext4_lazyinit_thread(void *arg)
3109{
3110 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3111 struct list_head *pos, *n;
3112 struct ext4_li_request *elr;
Lukas Czerner4ed5c032011-05-20 13:49:04 -04003113 unsigned long next_wakeup, cur;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003114
3115 BUG_ON(NULL == eli);
3116
Lukas Czernerbfff6872010-10-27 21:30:05 -04003117cont_thread:
3118 while (true) {
3119 next_wakeup = MAX_JIFFY_OFFSET;
3120
3121 mutex_lock(&eli->li_list_mtx);
3122 if (list_empty(&eli->li_request_list)) {
3123 mutex_unlock(&eli->li_list_mtx);
3124 goto exit_thread;
3125 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003126 list_for_each_safe(pos, n, &eli->li_request_list) {
Dmitry Monakhove22834f2016-09-05 23:38:36 -04003127 int err = 0;
3128 int progress = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003129 elr = list_entry(pos, struct ext4_li_request,
3130 lr_request);
3131
Dmitry Monakhove22834f2016-09-05 23:38:36 -04003132 if (time_before(jiffies, elr->lr_next_sched)) {
3133 if (time_before(elr->lr_next_sched, next_wakeup))
3134 next_wakeup = elr->lr_next_sched;
3135 continue;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003136 }
Dmitry Monakhove22834f2016-09-05 23:38:36 -04003137 if (down_read_trylock(&elr->lr_super->s_umount)) {
3138 if (sb_start_write_trylock(elr->lr_super)) {
3139 progress = 1;
3140 /*
3141 * We hold sb->s_umount, sb can not
3142 * be removed from the list, it is
3143 * now safe to drop li_list_mtx
3144 */
3145 mutex_unlock(&eli->li_list_mtx);
3146 err = ext4_run_li_request(elr);
3147 sb_end_write(elr->lr_super);
3148 mutex_lock(&eli->li_list_mtx);
3149 n = pos->next;
3150 }
3151 up_read((&elr->lr_super->s_umount));
3152 }
3153 /* error, remove the lazy_init job */
3154 if (err) {
3155 ext4_remove_li_request(elr);
3156 continue;
3157 }
3158 if (!progress) {
3159 elr->lr_next_sched = jiffies +
3160 (prandom_u32()
3161 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3162 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003163 if (time_before(elr->lr_next_sched, next_wakeup))
3164 next_wakeup = elr->lr_next_sched;
3165 }
3166 mutex_unlock(&eli->li_list_mtx);
3167
Tejun Heoa0acae02011-11-21 12:32:22 -08003168 try_to_freeze();
Lukas Czernerbfff6872010-10-27 21:30:05 -04003169
Lukas Czerner4ed5c032011-05-20 13:49:04 -04003170 cur = jiffies;
3171 if ((time_after_eq(cur, next_wakeup)) ||
Lukas Czernerf4245bd2010-11-02 14:07:17 -04003172 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04003173 cond_resched();
3174 continue;
3175 }
3176
Lukas Czerner4ed5c032011-05-20 13:49:04 -04003177 schedule_timeout_interruptible(next_wakeup - cur);
3178
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003179 if (kthread_should_stop()) {
3180 ext4_clear_request_list();
3181 goto exit_thread;
3182 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003183 }
3184
3185exit_thread:
3186 /*
3187 * It looks like the request list is empty, but we need
3188 * to check it under the li_list_mtx lock, to prevent any
3189 * additions into it, and of course we should lock ext4_li_mtx
3190 * to atomically free the list and ext4_li_info, because at
3191 * this point another ext4 filesystem could be registering
3192 * new one.
3193 */
3194 mutex_lock(&ext4_li_mtx);
3195 mutex_lock(&eli->li_list_mtx);
3196 if (!list_empty(&eli->li_request_list)) {
3197 mutex_unlock(&eli->li_list_mtx);
3198 mutex_unlock(&ext4_li_mtx);
3199 goto cont_thread;
3200 }
3201 mutex_unlock(&eli->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003202 kfree(ext4_li_info);
3203 ext4_li_info = NULL;
3204 mutex_unlock(&ext4_li_mtx);
3205
3206 return 0;
3207}
3208
3209static void ext4_clear_request_list(void)
3210{
3211 struct list_head *pos, *n;
3212 struct ext4_li_request *elr;
3213
3214 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003215 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3216 elr = list_entry(pos, struct ext4_li_request,
3217 lr_request);
3218 ext4_remove_li_request(elr);
3219 }
3220 mutex_unlock(&ext4_li_info->li_list_mtx);
3221}
3222
3223static int ext4_run_lazyinit_thread(void)
3224{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003225 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3226 ext4_li_info, "ext4lazyinit");
3227 if (IS_ERR(ext4_lazyinit_task)) {
3228 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003229 ext4_clear_request_list();
Lukas Czernerbfff6872010-10-27 21:30:05 -04003230 kfree(ext4_li_info);
3231 ext4_li_info = NULL;
Theodore Ts'o92b978162012-03-19 23:41:49 -04003232 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
Lukas Czernerbfff6872010-10-27 21:30:05 -04003233 "initialization thread\n",
3234 err);
3235 return err;
3236 }
3237 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003238 return 0;
3239}
3240
3241/*
3242 * Check whether it make sense to run itable init. thread or not.
3243 * If there is at least one uninitialized inode table, return
3244 * corresponding group number, else the loop goes through all
3245 * groups and return total number of groups.
3246 */
3247static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3248{
3249 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3250 struct ext4_group_desc *gdp = NULL;
3251
Theodore Ts'o88446182018-06-14 00:58:00 -04003252 if (!ext4_has_group_desc_csum(sb))
3253 return ngroups;
3254
Lukas Czernerbfff6872010-10-27 21:30:05 -04003255 for (group = 0; group < ngroups; group++) {
3256 gdp = ext4_get_group_desc(sb, group, NULL);
3257 if (!gdp)
3258 continue;
3259
Theodore Ts'o50122842018-07-28 08:12:04 -04003260 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
Lukas Czernerbfff6872010-10-27 21:30:05 -04003261 break;
3262 }
3263
3264 return group;
3265}
3266
3267static int ext4_li_info_new(void)
3268{
3269 struct ext4_lazy_init *eli = NULL;
3270
3271 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3272 if (!eli)
3273 return -ENOMEM;
3274
Lukas Czernerbfff6872010-10-27 21:30:05 -04003275 INIT_LIST_HEAD(&eli->li_request_list);
3276 mutex_init(&eli->li_list_mtx);
3277
Lukas Czernerbfff6872010-10-27 21:30:05 -04003278 eli->li_state |= EXT4_LAZYINIT_QUIT;
3279
3280 ext4_li_info = eli;
3281
3282 return 0;
3283}
3284
3285static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3286 ext4_group_t start)
3287{
3288 struct ext4_sb_info *sbi = EXT4_SB(sb);
3289 struct ext4_li_request *elr;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003290
3291 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3292 if (!elr)
3293 return NULL;
3294
3295 elr->lr_super = sb;
3296 elr->lr_sbi = sbi;
3297 elr->lr_next_group = start;
3298
3299 /*
3300 * Randomize first schedule time of the request to
3301 * spread the inode table initialization requests
3302 * better.
3303 */
Theodore Ts'odd1f7232013-11-08 00:14:53 -05003304 elr->lr_next_sched = jiffies + (prandom_u32() %
3305 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
Lukas Czernerbfff6872010-10-27 21:30:05 -04003306 return elr;
3307}
3308
Theodore Ts'o7f511862013-01-13 08:41:45 -05003309int ext4_register_li_request(struct super_block *sb,
3310 ext4_group_t first_not_zeroed)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003311{
3312 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003313 struct ext4_li_request *elr = NULL;
Jun Piao49598e02018-01-11 13:17:49 -05003314 ext4_group_t ngroups = sbi->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05003315 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003316
Theodore Ts'o7f511862013-01-13 08:41:45 -05003317 mutex_lock(&ext4_li_mtx);
Lukas Czerner51ce6512011-05-20 13:55:16 -04003318 if (sbi->s_li_request != NULL) {
3319 /*
3320 * Reset timeout so it can be computed again, because
3321 * s_li_wait_mult might have changed.
3322 */
3323 sbi->s_li_request->lr_timeout = 0;
Theodore Ts'o7f511862013-01-13 08:41:45 -05003324 goto out;
Lukas Czerner51ce6512011-05-20 13:55:16 -04003325 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003326
David Howellsbc98a422017-07-17 08:45:34 +01003327 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
Tao Ma55ff3842011-05-09 10:28:41 -04003328 !test_opt(sb, INIT_INODE_TABLE))
Theodore Ts'o7f511862013-01-13 08:41:45 -05003329 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003330
3331 elr = ext4_li_request_new(sb, first_not_zeroed);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003332 if (!elr) {
3333 ret = -ENOMEM;
3334 goto out;
3335 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003336
3337 if (NULL == ext4_li_info) {
3338 ret = ext4_li_info_new();
3339 if (ret)
3340 goto out;
3341 }
3342
3343 mutex_lock(&ext4_li_info->li_list_mtx);
3344 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3345 mutex_unlock(&ext4_li_info->li_list_mtx);
3346
3347 sbi->s_li_request = elr;
Tao Ma46e46902011-04-04 16:00:49 -04003348 /*
3349 * set elr to NULL here since it has been inserted to
3350 * the request_list and the removal and free of it is
3351 * handled by ext4_clear_request_list from now on.
3352 */
3353 elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003354
3355 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3356 ret = ext4_run_lazyinit_thread();
3357 if (ret)
3358 goto out;
3359 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003360out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003361 mutex_unlock(&ext4_li_mtx);
3362 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003363 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003364 return ret;
3365}
3366
3367/*
3368 * We do not need to lock anything since this is called on
3369 * module unload.
3370 */
3371static void ext4_destroy_lazyinit_thread(void)
3372{
3373 /*
3374 * If thread exited earlier
3375 * there's nothing to be done.
3376 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003377 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003378 return;
3379
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003380 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003381}
3382
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003383static int set_journal_csum_feature_set(struct super_block *sb)
3384{
3385 int ret = 1;
3386 int compat, incompat;
3387 struct ext4_sb_info *sbi = EXT4_SB(sb);
3388
Dmitry Monakhov9aa5d32b2014-10-13 03:36:16 -04003389 if (ext4_has_metadata_csum(sb)) {
Darrick J. Wongdb9ee222014-08-27 18:40:07 -04003390 /* journal checksum v3 */
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003391 compat = 0;
Darrick J. Wongdb9ee222014-08-27 18:40:07 -04003392 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003393 } else {
3394 /* journal checksum v1 */
3395 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3396 incompat = 0;
3397 }
3398
Darrick J. Wongfeb8c6d2014-09-11 11:38:21 -04003399 jbd2_journal_clear_features(sbi->s_journal,
3400 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3401 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3402 JBD2_FEATURE_INCOMPAT_CSUM_V2);
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003403 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3404 ret = jbd2_journal_set_features(sbi->s_journal,
3405 compat, 0,
3406 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3407 incompat);
3408 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3409 ret = jbd2_journal_set_features(sbi->s_journal,
3410 compat, 0,
3411 incompat);
3412 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3413 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3414 } else {
Darrick J. Wongfeb8c6d2014-09-11 11:38:21 -04003415 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3416 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003417 }
3418
3419 return ret;
3420}
3421
Theodore Ts'o952fc182012-07-09 16:27:05 -04003422/*
3423 * Note: calculating the overhead so we can be compatible with
3424 * historical BSD practice is quite difficult in the face of
3425 * clusters/bigalloc. This is because multiple metadata blocks from
3426 * different block group can end up in the same allocation cluster.
3427 * Calculating the exact overhead in the face of clustered allocation
3428 * requires either O(all block bitmaps) in memory or O(number of block
3429 * groups**2) in time. We will still calculate the superblock for
3430 * older file systems --- and if we come across with a bigalloc file
3431 * system with zero in s_overhead_clusters the estimate will be close to
3432 * correct especially for very large cluster sizes --- but for newer
3433 * file systems, it's better to calculate this figure once at mkfs
3434 * time, and store it in the superblock. If the superblock value is
3435 * present (even for non-bigalloc file systems), we will use it.
3436 */
3437static int count_overhead(struct super_block *sb, ext4_group_t grp,
3438 char *buf)
3439{
3440 struct ext4_sb_info *sbi = EXT4_SB(sb);
3441 struct ext4_group_desc *gdp;
3442 ext4_fsblk_t first_block, last_block, b;
3443 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3444 int s, j, count = 0;
3445
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003446 if (!ext4_has_feature_bigalloc(sb))
Theodore Ts'o0548bbb2012-08-16 11:59:04 -04003447 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3448 sbi->s_itb_per_group + 2);
3449
Theodore Ts'o952fc182012-07-09 16:27:05 -04003450 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3451 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3452 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3453 for (i = 0; i < ngroups; i++) {
3454 gdp = ext4_get_group_desc(sb, i, NULL);
3455 b = ext4_block_bitmap(sb, gdp);
3456 if (b >= first_block && b <= last_block) {
3457 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3458 count++;
3459 }
3460 b = ext4_inode_bitmap(sb, gdp);
3461 if (b >= first_block && b <= last_block) {
3462 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3463 count++;
3464 }
3465 b = ext4_inode_table(sb, gdp);
3466 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3467 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3468 int c = EXT4_B2C(sbi, b - first_block);
3469 ext4_set_bit(c, buf);
3470 count++;
3471 }
3472 if (i != grp)
3473 continue;
3474 s = 0;
3475 if (ext4_bg_has_super(sb, grp)) {
3476 ext4_set_bit(s++, buf);
3477 count++;
3478 }
Theodore Ts'oc48ae412016-11-18 13:37:47 -05003479 j = ext4_bg_num_gdb(sb, grp);
3480 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3481 ext4_error(sb, "Invalid number of block group "
3482 "descriptor blocks: %d", j);
3483 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
Theodore Ts'o952fc182012-07-09 16:27:05 -04003484 }
Theodore Ts'oc48ae412016-11-18 13:37:47 -05003485 count += j;
3486 for (; j > 0; j--)
3487 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
Theodore Ts'o952fc182012-07-09 16:27:05 -04003488 }
3489 if (!count)
3490 return 0;
3491 return EXT4_CLUSTERS_PER_GROUP(sb) -
3492 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3493}
3494
3495/*
3496 * Compute the overhead and stash it in sbi->s_overhead
3497 */
3498int ext4_calculate_overhead(struct super_block *sb)
3499{
3500 struct ext4_sb_info *sbi = EXT4_SB(sb);
3501 struct ext4_super_block *es = sbi->s_es;
Eric Whitney3c816de2016-09-30 02:08:49 -04003502 struct inode *j_inode;
3503 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
Theodore Ts'o952fc182012-07-09 16:27:05 -04003504 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3505 ext4_fsblk_t overhead = 0;
Dmitry Monakhov4fdb5542014-11-25 13:08:04 -05003506 char *buf = (char *) get_zeroed_page(GFP_NOFS);
Theodore Ts'o952fc182012-07-09 16:27:05 -04003507
Theodore Ts'o952fc182012-07-09 16:27:05 -04003508 if (!buf)
3509 return -ENOMEM;
3510
3511 /*
3512 * Compute the overhead (FS structures). This is constant
3513 * for a given filesystem unless the number of block groups
3514 * changes so we cache the previous value until it does.
3515 */
3516
3517 /*
3518 * All of the blocks before first_data_block are overhead
3519 */
3520 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3521
3522 /*
3523 * Add the overhead found in each block group
3524 */
3525 for (i = 0; i < ngroups; i++) {
3526 int blks;
3527
3528 blks = count_overhead(sb, i, buf);
3529 overhead += blks;
3530 if (blks)
3531 memset(buf, 0, PAGE_SIZE);
3532 cond_resched();
3533 }
Eric Whitney3c816de2016-09-30 02:08:49 -04003534
3535 /*
3536 * Add the internal journal blocks whether the journal has been
3537 * loaded or not
3538 */
Eric Sandeenb003b522014-11-25 16:27:44 -05003539 if (sbi->s_journal && !sbi->journal_bdev)
Lukas Czerner810da242013-03-02 17:18:58 -05003540 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
Eric Whitney3c816de2016-09-30 02:08:49 -04003541 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3542 j_inode = ext4_get_journal_inode(sb, j_inum);
3543 if (j_inode) {
3544 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3545 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3546 iput(j_inode);
3547 } else {
3548 ext4_msg(sb, KERN_ERR, "can't get journal size");
3549 }
3550 }
Theodore Ts'o952fc182012-07-09 16:27:05 -04003551 sbi->s_overhead = overhead;
3552 smp_wmb();
3553 free_page((unsigned long) buf);
3554 return 0;
3555}
3556
Barret Rhoden7bc04c52019-04-25 11:55:50 -04003557static void ext4_clamp_want_extra_isize(struct super_block *sb)
3558{
3559 struct ext4_sb_info *sbi = EXT4_SB(sb);
3560 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'o4ea99932019-11-07 21:43:41 -05003561 unsigned def_extra_isize = sizeof(struct ext4_inode) -
3562 EXT4_GOOD_OLD_INODE_SIZE;
Barret Rhoden7bc04c52019-04-25 11:55:50 -04003563
Theodore Ts'o4ea99932019-11-07 21:43:41 -05003564 if (sbi->s_inode_size == EXT4_GOOD_OLD_INODE_SIZE) {
3565 sbi->s_want_extra_isize = 0;
3566 return;
3567 }
3568 if (sbi->s_want_extra_isize < 4) {
3569 sbi->s_want_extra_isize = def_extra_isize;
Barret Rhoden7bc04c52019-04-25 11:55:50 -04003570 if (ext4_has_feature_extra_isize(sb)) {
3571 if (sbi->s_want_extra_isize <
3572 le16_to_cpu(es->s_want_extra_isize))
3573 sbi->s_want_extra_isize =
3574 le16_to_cpu(es->s_want_extra_isize);
3575 if (sbi->s_want_extra_isize <
3576 le16_to_cpu(es->s_min_extra_isize))
3577 sbi->s_want_extra_isize =
3578 le16_to_cpu(es->s_min_extra_isize);
3579 }
3580 }
3581 /* Check if enough inode space is available */
Theodore Ts'o4ea99932019-11-07 21:43:41 -05003582 if ((sbi->s_want_extra_isize > sbi->s_inode_size) ||
3583 (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3584 sbi->s_inode_size)) {
3585 sbi->s_want_extra_isize = def_extra_isize;
Barret Rhoden7bc04c52019-04-25 11:55:50 -04003586 ext4_msg(sb, KERN_INFO,
3587 "required extra inode space not available");
3588 }
3589}
3590
Theodore Ts'ob5799012015-09-23 12:44:17 -04003591static void ext4_set_resv_clusters(struct super_block *sb)
Lukas Czerner27dd4382013-04-09 22:11:22 -04003592{
3593 ext4_fsblk_t resv_clusters;
Theodore Ts'ob5799012015-09-23 12:44:17 -04003594 struct ext4_sb_info *sbi = EXT4_SB(sb);
Lukas Czerner27dd4382013-04-09 22:11:22 -04003595
3596 /*
Jan Kara30fac0f2013-12-08 21:11:59 -05003597 * There's no need to reserve anything when we aren't using extents.
3598 * The space estimates are exact, there are no unwritten extents,
3599 * hole punching doesn't need new metadata... This is needed especially
3600 * to keep ext2/3 backward compatibility.
3601 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003602 if (!ext4_has_feature_extents(sb))
Theodore Ts'ob5799012015-09-23 12:44:17 -04003603 return;
Jan Kara30fac0f2013-12-08 21:11:59 -05003604 /*
Lukas Czerner27dd4382013-04-09 22:11:22 -04003605 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3606 * This should cover the situations where we can not afford to run
3607 * out of space like for example punch hole, or converting
Lukas Czerner556615d2014-04-20 23:45:47 -04003608 * unwritten extents in delalloc path. In most cases such
Lukas Czerner27dd4382013-04-09 22:11:22 -04003609 * allocation would require 1, or 2 blocks, higher numbers are
3610 * very rare.
3611 */
Theodore Ts'ob5799012015-09-23 12:44:17 -04003612 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3613 sbi->s_cluster_bits);
Lukas Czerner27dd4382013-04-09 22:11:22 -04003614
3615 do_div(resv_clusters, 50);
3616 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3617
Theodore Ts'ob5799012015-09-23 12:44:17 -04003618 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
Lukas Czerner27dd4382013-04-09 22:11:22 -04003619}
3620
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003621static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003622{
Dan Williams5e405592017-08-24 16:42:48 -07003623 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003624 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003625 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003626 struct ext4_super_block *es = NULL;
Theodore Ts'o5aee0f82016-11-18 13:24:26 -05003627 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
Mingming Cao617ba132006-10-11 01:20:53 -07003628 ext4_fsblk_t block;
3629 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003630 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003631 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003632 unsigned long journal_devnum = 0;
3633 unsigned long def_mount_opts;
3634 struct inode *root;
Frank Mayhar03901312009-01-07 00:06:22 -05003635 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003636 int ret = -ENOMEM;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003637 int blocksize, clustersize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003638 unsigned int db_count;
3639 unsigned int i;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003640 int needs_recovery, has_huge_files, has_bigalloc;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003641 __u64 blocks_count;
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003642 int err = 0;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003643 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003644 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003645
Theodore Ts'o5aee0f82016-11-18 13:24:26 -05003646 if ((data && !orig_data) || !sbi)
3647 goto out_free_base;
Pekka Enberg705895b2009-02-15 18:07:52 -05003648
Colin Ian Kingaed9eb12017-09-05 17:51:23 +01003649 sbi->s_daxdev = dax_dev;
Pekka Enberg705895b2009-02-15 18:07:52 -05003650 sbi->s_blockgroup_lock =
3651 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
Theodore Ts'o5aee0f82016-11-18 13:24:26 -05003652 if (!sbi->s_blockgroup_lock)
3653 goto out_free_base;
3654
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003655 sb->s_fs_info = sbi;
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04003656 sbi->s_sb = sb;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003657 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003658 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003659 if (sb->s_bdev->bd_part)
3660 sbi->s_sectors_written_start =
Michael Callahandbae2c52018-07-18 04:47:38 -07003661 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003662
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003663 /* Cleanup superblock name */
Rasmus Villemoesec3904d2015-06-25 15:02:41 -07003664 strreplace(sb->s_id, '/', '!');
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003665
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003666 /* -EINVAL is default */
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003667 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003668 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003669 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003670 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003671 goto out_fail;
3672 }
3673
3674 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003675 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003676 * block sizes. We need to calculate the offset from buffer start.
3677 */
Mingming Cao617ba132006-10-11 01:20:53 -07003678 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003679 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3680 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003681 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003682 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003683 }
3684
Gioh Kima8ac9002014-09-04 22:36:15 -04003685 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003686 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003687 goto out_fail;
3688 }
3689 /*
3690 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003691 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003692 */
Theodore Ts'o2716b802012-05-28 17:47:52 -04003693 es = (struct ext4_super_block *) (bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003694 sbi->s_es = es;
3695 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003696 if (sb->s_magic != EXT4_SUPER_MAGIC)
3697 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003698 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003699
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003700 /* Warn if metadata_csum and gdt_csum are both set. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003701 if (ext4_has_feature_metadata_csum(sb) &&
3702 ext4_has_feature_gdt_csum(sb))
Jakub Wilk363307e2015-01-02 15:31:14 -05003703 ext4_warning(sb, "metadata_csum and uninit_bg are "
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003704 "redundant flags; please run fsck.");
3705
Darrick J. Wongd25425f2012-04-29 18:25:10 -04003706 /* Check for a known checksum algorithm */
3707 if (!ext4_verify_csum_type(sb, es)) {
3708 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3709 "unknown checksum algorithm.");
3710 silent = 1;
3711 goto cantfind_ext4;
3712 }
3713
Darrick J. Wong04419842012-04-29 18:27:10 -04003714 /* Load the checksum driver */
Theodore Ts'oa45403b2018-03-29 22:10:31 -04003715 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3716 if (IS_ERR(sbi->s_chksum_driver)) {
3717 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3718 ret = PTR_ERR(sbi->s_chksum_driver);
3719 sbi->s_chksum_driver = NULL;
3720 goto failed_mount;
Darrick J. Wong04419842012-04-29 18:27:10 -04003721 }
3722
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003723 /* Check superblock checksum */
3724 if (!ext4_superblock_csum_verify(sb, es)) {
3725 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3726 "invalid superblock checksum. Run e2fsck?");
3727 silent = 1;
Darrick J. Wong6a797d22015-10-17 16:16:04 -04003728 ret = -EFSBADCRC;
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003729 goto cantfind_ext4;
3730 }
3731
3732 /* Precompute checksum seed for all metadata */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003733 if (ext4_has_feature_csum_seed(sb))
Darrick J. Wong8c81bd82015-10-17 16:16:02 -04003734 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
Tahsin Erdogandec214d2017-06-22 11:44:55 -04003735 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003736 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3737 sizeof(es->s_uuid));
3738
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003739 /* Set defaults before we parse the mount options */
3740 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003741 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003742 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003743 set_opt(sb, DEBUG);
Theodore Ts'o87f26802012-03-02 00:03:21 -05003744 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003745 set_opt(sb, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07003746 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003747 set_opt(sb, NO_UID32);
Eric Sandeenea663332011-02-23 17:51:51 -05003748 /* xattr user namespace & acls are now defaulted on */
Eric Sandeenea663332011-02-23 17:51:51 -05003749 set_opt(sb, XATTR_USER);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003750#ifdef CONFIG_EXT4_FS_POSIX_ACL
Eric Sandeenea663332011-02-23 17:51:51 -05003751 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003752#endif
Darrick J. Wong98c1a752014-10-30 10:53:16 -04003753 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3754 if (ext4_has_metadata_csum(sb))
3755 set_opt(sb, JOURNAL_CHECKSUM);
3756
Mingming Cao617ba132006-10-11 01:20:53 -07003757 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003758 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003759 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003760 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003761 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003762 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003763
Mingming Cao617ba132006-10-11 01:20:53 -07003764 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003765 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003766 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003767 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003768 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003769 set_opt(sb, ERRORS_RO);
Darrick J. Wong45f1a9c2014-09-01 21:34:09 -04003770 /* block_validity enabled by default; disable with noblock_validity */
3771 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003772 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003773 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003774
Eric W. Biederman08cefc72012-02-07 15:41:49 -08003775 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3776 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
Theodore Ts'o30773842009-01-03 20:27:38 -05003777 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3778 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3779 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003780
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003781 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003782 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003783
Mingming Cao1e2462f2007-07-18 09:00:55 -04003784 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003785 * enable delayed allocation by default
3786 * Use -o nodelalloc to turn it off
3787 */
Theodore Ts'obc0b75f2012-09-17 22:54:36 -04003788 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003789 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003790 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003791
Lukas Czerner51ce6512011-05-20 13:55:16 -04003792 /*
3793 * set default s_li_wait_mult for lazyinit, for the case there is
3794 * no mount option specified.
3795 */
3796 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3797
Theodore Ts'o5aee0f82016-11-18 13:24:26 -05003798 if (sbi->s_es->s_mount_opts[0]) {
3799 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3800 sizeof(sbi->s_es->s_mount_opts),
3801 GFP_KERNEL);
3802 if (!s_mount_opts)
3803 goto failed_mount;
3804 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3805 &journal_ioprio, 0)) {
3806 ext4_msg(sb, KERN_WARNING,
3807 "failed to parse options in superblock: %s",
3808 s_mount_opts);
3809 }
3810 kfree(s_mount_opts);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003811 }
Theodore Ts'o5a916be2012-03-04 19:27:31 -05003812 sbi->s_def_mount_opt = sbi->s_mount_opt;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003813 if (!parse_options((char *) data, sb, &journal_devnum,
Eric Sandeen661aa522012-02-20 17:53:04 -05003814 &journal_ioprio, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003815 goto failed_mount;
3816
Gabriel Krisman Bertazic83ad552019-04-25 14:05:42 -04003817#ifdef CONFIG_UNICODE
3818 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3819 const struct ext4_sb_encodings *encoding_info;
3820 struct unicode_map *encoding;
3821 __u16 encoding_flags;
3822
3823 if (ext4_has_feature_encrypt(sb)) {
3824 ext4_msg(sb, KERN_ERR,
3825 "Can't mount with encoding and encryption");
3826 goto failed_mount;
3827 }
3828
3829 if (ext4_sb_read_encoding(es, &encoding_info,
3830 &encoding_flags)) {
3831 ext4_msg(sb, KERN_ERR,
3832 "Encoding requested by superblock is unknown");
3833 goto failed_mount;
3834 }
3835
3836 encoding = utf8_load(encoding_info->version);
3837 if (IS_ERR(encoding)) {
3838 ext4_msg(sb, KERN_ERR,
3839 "can't mount with superblock charset: %s-%s "
3840 "not supported by the kernel. flags: 0x%x.",
3841 encoding_info->name, encoding_info->version,
3842 encoding_flags);
3843 goto failed_mount;
3844 }
3845 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3846 "%s-%s with flags 0x%hx", encoding_info->name,
3847 encoding_info->version?:"\b", encoding_flags);
3848
3849 sbi->s_encoding = encoding;
3850 sbi->s_encoding_flags = encoding_flags;
3851 }
3852#endif
3853
Theodore Ts'o56889782011-09-03 18:22:38 -04003854 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3855 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3856 "with data=journal disables delayed "
3857 "allocation and O_DIRECT support!\n");
3858 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3859 ext4_msg(sb, KERN_ERR, "can't mount with "
3860 "both data=journal and delalloc");
3861 goto failed_mount;
3862 }
3863 if (test_opt(sb, DIOREAD_NOLOCK)) {
3864 ext4_msg(sb, KERN_ERR, "can't mount with "
Piotr Sarna6ae65142013-08-08 23:02:24 -04003865 "both data=journal and dioread_nolock");
Theodore Ts'o56889782011-09-03 18:22:38 -04003866 goto failed_mount;
3867 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003868 if (test_opt(sb, DAX)) {
3869 ext4_msg(sb, KERN_ERR, "can't mount with "
3870 "both data=journal and dax");
3871 goto failed_mount;
3872 }
Sergey Karamov73b92a22016-12-10 17:54:58 -05003873 if (ext4_has_feature_encrypt(sb)) {
3874 ext4_msg(sb, KERN_WARNING,
3875 "encrypted files will use data=ordered "
3876 "instead of data journaling mode");
3877 }
Theodore Ts'o56889782011-09-03 18:22:38 -04003878 if (test_opt(sb, DELALLOC))
3879 clear_opt(sb, DELALLOC);
Tejun Heo001e4a82015-07-21 23:51:26 -04003880 } else {
3881 sb->s_iflags |= SB_I_CGROUPWB;
Theodore Ts'o56889782011-09-03 18:22:38 -04003882 }
3883
Linus Torvalds1751e8a2017-11-27 13:05:09 -08003884 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3885 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003886
Mingming Cao617ba132006-10-11 01:20:53 -07003887 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003888 (ext4_has_compat_features(sb) ||
3889 ext4_has_ro_compat_features(sb) ||
3890 ext4_has_incompat_features(sb)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003891 ext4_msg(sb, KERN_WARNING,
3892 "feature flags set on rev 0 fs, "
3893 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003894
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003895 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3896 set_opt2(sb, HURD_COMPAT);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003897 if (ext4_has_feature_64bit(sb)) {
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003898 ext4_msg(sb, KERN_ERR,
3899 "The Hurd can't support 64-bit file systems");
3900 goto failed_mount;
3901 }
Tahsin Erdogandec214d2017-06-22 11:44:55 -04003902
3903 /*
3904 * ea_inode feature uses l_i_version field which is not
3905 * available in HURD_COMPAT mode.
3906 */
3907 if (ext4_has_feature_ea_inode(sb)) {
3908 ext4_msg(sb, KERN_ERR,
3909 "ea_inode feature is not supported for Hurd");
3910 goto failed_mount;
3911 }
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003912 }
3913
Theodore Ts'o2035e772011-04-18 17:29:14 -04003914 if (IS_EXT2_SB(sb)) {
3915 if (ext2_feature_set_ok(sb))
3916 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3917 "using the ext4 subsystem");
3918 else {
Eric Sandeen0d9366d2018-03-22 11:59:00 -04003919 /*
3920 * If we're probing be silent, if this looks like
3921 * it's actually an ext[34] filesystem.
3922 */
3923 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3924 goto failed_mount;
Theodore Ts'o2035e772011-04-18 17:29:14 -04003925 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3926 "to feature incompatibilities");
3927 goto failed_mount;
3928 }
3929 }
3930
3931 if (IS_EXT3_SB(sb)) {
3932 if (ext3_feature_set_ok(sb))
3933 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3934 "using the ext4 subsystem");
3935 else {
Eric Sandeen0d9366d2018-03-22 11:59:00 -04003936 /*
3937 * If we're probing be silent, if this looks like
3938 * it's actually an ext4 filesystem.
3939 */
3940 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3941 goto failed_mount;
Theodore Ts'o2035e772011-04-18 17:29:14 -04003942 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3943 "to feature incompatibilities");
3944 goto failed_mount;
3945 }
3946 }
3947
Theodore Tso469108f2008-02-10 01:11:44 -05003948 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003949 * Check feature flags regardless of the revision level, since we
3950 * previously didn't change the revision level when setting the flags,
3951 * so there is a chance incompat flags are set on a rev 0 filesystem.
3952 */
David Howellsbc98a422017-07-17 08:45:34 +01003953 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003954 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003955
Jan Kara261cb20c2012-12-20 00:07:18 -05003956 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
Mingming Cao617ba132006-10-11 01:20:53 -07003957 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3958 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003959 ext4_msg(sb, KERN_ERR,
Theodore Ts'o8cdf3372016-11-18 13:00:24 -05003960 "Unsupported filesystem blocksize %d (%d log_block_size)",
3961 blocksize, le32_to_cpu(es->s_log_block_size));
3962 goto failed_mount;
3963 }
3964 if (le32_to_cpu(es->s_log_block_size) >
3965 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3966 ext4_msg(sb, KERN_ERR,
3967 "Invalid log block size: %u",
3968 le32_to_cpu(es->s_log_block_size));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003969 goto failed_mount;
3970 }
Theodore Ts'obfe0a5f2018-06-17 18:11:20 -04003971 if (le32_to_cpu(es->s_log_cluster_size) >
3972 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3973 ext4_msg(sb, KERN_ERR,
3974 "Invalid log cluster size: %u",
3975 le32_to_cpu(es->s_log_cluster_size));
3976 goto failed_mount;
3977 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003978
Theodore Ts'o5b9554d2016-07-05 20:01:52 -04003979 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3980 ext4_msg(sb, KERN_ERR,
3981 "Number of reserved GDT blocks insanely large: %d",
3982 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3983 goto failed_mount;
3984 }
3985
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003986 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
Ross Zwisler559db4c2017-10-12 11:52:34 -04003987 if (ext4_has_feature_inline_data(sb)) {
3988 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3989 " that may contain inline data");
Eric Sandeen361d24d2018-12-04 00:46:39 -05003990 goto failed_mount;
Ross Zwisler559db4c2017-10-12 11:52:34 -04003991 }
Dave Jiang80660f202018-05-30 13:03:46 -07003992 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
Dan Williams24f34782017-12-21 17:04:07 -08003993 ext4_msg(sb, KERN_ERR,
Eric Sandeen361d24d2018-12-04 00:46:39 -05003994 "DAX unsupported by block device.");
3995 goto failed_mount;
Dan Williams24f34782017-12-21 17:04:07 -08003996 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003997 }
3998
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003999 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04004000 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4001 es->s_encryption_level);
4002 goto failed_mount;
4003 }
4004
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004005 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05004006 /* Validate the filesystem blocksize */
4007 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004008 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05004009 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004010 goto failed_mount;
4011 }
4012
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004013 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07004014 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4015 offset = do_div(logical_sb_block, blocksize);
Gioh Kima8ac9002014-09-04 22:36:15 -04004016 bh = sb_bread_unmovable(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004017 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004018 ext4_msg(sb, KERN_ERR,
4019 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004020 goto failed_mount;
4021 }
Theodore Ts'o2716b802012-05-28 17:47:52 -04004022 es = (struct ext4_super_block *)(bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004023 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07004024 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004025 ext4_msg(sb, KERN_ERR,
4026 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004027 goto failed_mount;
4028 }
4029 }
4030
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004031 has_huge_files = ext4_has_feature_huge_file(sb);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04004032 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4033 has_huge_files);
4034 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004035
Mingming Cao617ba132006-10-11 01:20:53 -07004036 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4037 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4038 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004039 } else {
4040 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4041 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Theodore Ts'oc37e9e02018-06-17 00:41:14 -04004042 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4043 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4044 sbi->s_first_ino);
4045 goto failed_mount;
4046 }
Mingming Cao617ba132006-10-11 01:20:53 -07004047 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04004048 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004049 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004050 ext4_msg(sb, KERN_ERR,
4051 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004052 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004053 goto failed_mount;
4054 }
Deepa Dinamani4881c492018-01-21 18:04:24 -08004055 /*
4056 * i_atime_extra is the last extra field available for [acm]times in
4057 * struct ext4_inode. Checking for that field should suffice to ensure
4058 * we have extra space for all three.
4059 */
4060 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4061 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4062 sb->s_time_gran = 1;
4063 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4064 } else {
4065 sb->s_time_gran = NSEC_PER_SEC;
4066 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4067 }
4068
4069 sb->s_time_min = EXT4_TIMESTAMP_MIN;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004070 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004071
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07004072 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004073 if (ext4_has_feature_64bit(sb)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07004074 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07004075 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07004076 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004077 ext4_msg(sb, KERN_ERR,
4078 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07004079 sbi->s_desc_size);
4080 goto failed_mount;
4081 }
4082 } else
4083 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004084
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004085 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004086 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004087
Mingming Cao617ba132006-10-11 01:20:53 -07004088 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004089 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07004090 goto cantfind_ext4;
Theodore Ts'ocd6bb352016-11-18 13:28:30 -05004091 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4092 sbi->s_inodes_per_group > blocksize * 8) {
4093 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4094 sbi->s_blocks_per_group);
4095 goto failed_mount;
4096 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004097 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4098 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07004099 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004100 sbi->s_sbh = bh;
4101 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07004102 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4103 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004104
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004105 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004106 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4107 sbi->s_def_hash_version = es->s_def_hash_version;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004108 if (ext4_has_feature_dir_index(sb)) {
Theodore Ts'o23301412014-02-12 12:16:04 -05004109 i = le32_to_cpu(es->s_flags);
4110 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4111 sbi->s_hash_unsigned = 3;
4112 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
Theodore Ts'of99b2582008-10-28 13:21:44 -04004113#ifdef __CHAR_UNSIGNED__
David Howellsbc98a422017-07-17 08:45:34 +01004114 if (!sb_rdonly(sb))
Theodore Ts'o23301412014-02-12 12:16:04 -05004115 es->s_flags |=
4116 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4117 sbi->s_hash_unsigned = 3;
Theodore Ts'of99b2582008-10-28 13:21:44 -04004118#else
David Howellsbc98a422017-07-17 08:45:34 +01004119 if (!sb_rdonly(sb))
Theodore Ts'o23301412014-02-12 12:16:04 -05004120 es->s_flags |=
4121 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
Theodore Ts'of99b2582008-10-28 13:21:44 -04004122#endif
Theodore Ts'o23301412014-02-12 12:16:04 -05004123 }
Theodore Ts'of99b2582008-10-28 13:21:44 -04004124 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004125
Theodore Ts'o281b5992011-09-09 18:34:51 -04004126 /* Handle clustersize */
4127 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004128 has_bigalloc = ext4_has_feature_bigalloc(sb);
Theodore Ts'o281b5992011-09-09 18:34:51 -04004129 if (has_bigalloc) {
4130 if (clustersize < blocksize) {
4131 ext4_msg(sb, KERN_ERR,
4132 "cluster size (%d) smaller than "
4133 "block size (%d)", clustersize, blocksize);
4134 goto failed_mount;
4135 }
4136 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4137 le32_to_cpu(es->s_log_block_size);
4138 sbi->s_clusters_per_group =
4139 le32_to_cpu(es->s_clusters_per_group);
4140 if (sbi->s_clusters_per_group > blocksize * 8) {
4141 ext4_msg(sb, KERN_ERR,
4142 "#clusters per group too big: %lu",
4143 sbi->s_clusters_per_group);
4144 goto failed_mount;
4145 }
4146 if (sbi->s_blocks_per_group !=
4147 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4148 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4149 "clusters per group (%lu) inconsistent",
4150 sbi->s_blocks_per_group,
4151 sbi->s_clusters_per_group);
4152 goto failed_mount;
4153 }
4154 } else {
4155 if (clustersize != blocksize) {
Theodore Ts'obfe0a5f2018-06-17 18:11:20 -04004156 ext4_msg(sb, KERN_ERR,
4157 "fragment/cluster size (%d) != "
4158 "block size (%d)", clustersize, blocksize);
4159 goto failed_mount;
Theodore Ts'o281b5992011-09-09 18:34:51 -04004160 }
4161 if (sbi->s_blocks_per_group > blocksize * 8) {
4162 ext4_msg(sb, KERN_ERR,
4163 "#blocks per group too big: %lu",
4164 sbi->s_blocks_per_group);
4165 goto failed_mount;
4166 }
4167 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4168 sbi->s_cluster_bits = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004169 }
Theodore Ts'o281b5992011-09-09 18:34:51 -04004170 sbi->s_cluster_ratio = clustersize / blocksize;
4171
Theodore Ts'o960fd8562013-07-05 23:11:16 -04004172 /* Do we have standard group size of clustersize * 8 blocks ? */
4173 if (sbi->s_blocks_per_group == clustersize << 3)
4174 set_opt2(sb, STD_GROUP_SIZE);
4175
Eric Sandeenbf43d842009-08-17 23:48:51 -04004176 /*
4177 * Test whether we have more sectors than will fit in sector_t,
4178 * and whether the max offset is addressable by the page cache.
4179 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05004180 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07004181 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05004182 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004183 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04004184 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004185 goto failed_mount;
4186 }
4187
Mingming Cao617ba132006-10-11 01:20:53 -07004188 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4189 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05004190
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04004191 /* check blocks count against device size */
4192 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4193 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004194 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4195 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04004196 ext4_blocks_count(es), blocks_count);
4197 goto failed_mount;
4198 }
4199
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004200 /*
4201 * It makes no sense for the first data block to be beyond the end
4202 * of the filesystem.
4203 */
4204 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Zheng Liu5635a622011-12-18 16:13:58 -05004205 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004206 "block %u is beyond end of filesystem (%llu)",
4207 le32_to_cpu(es->s_first_data_block),
4208 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05004209 goto failed_mount;
4210 }
Theodore Ts'obfe0a5f2018-06-17 18:11:20 -04004211 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4212 (sbi->s_cluster_ratio == 1)) {
4213 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4214 "block is 0 with a 1k block and cluster size");
4215 goto failed_mount;
4216 }
4217
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004218 blocks_count = (ext4_blocks_count(es) -
4219 le32_to_cpu(es->s_first_data_block) +
4220 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4221 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05004222 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004223 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05004224 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004225 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05004226 ext4_blocks_count(es),
4227 le32_to_cpu(es->s_first_data_block),
4228 EXT4_BLOCKS_PER_GROUP(sb));
4229 goto failed_mount;
4230 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004231 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04004232 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4233 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Theodore Ts'o9e463082018-11-07 10:32:53 -05004234 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4235 le32_to_cpu(es->s_inodes_count)) {
4236 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4237 le32_to_cpu(es->s_inodes_count),
4238 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4239 ret = -EINVAL;
4240 goto failed_mount;
4241 }
Mingming Cao617ba132006-10-11 01:20:53 -07004242 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4243 EXT4_DESC_PER_BLOCK(sb);
Eryu Guan3a4b77c2016-12-01 15:08:37 -05004244 if (ext4_has_feature_meta_bg(sb)) {
Theodore Ts'o2ba3e6e2017-02-15 01:26:39 -05004245 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
Eryu Guan3a4b77c2016-12-01 15:08:37 -05004246 ext4_msg(sb, KERN_WARNING,
4247 "first meta block group too large: %u "
4248 "(group descriptor block count %u)",
4249 le32_to_cpu(es->s_first_meta_bg), db_count);
4250 goto failed_mount;
4251 }
4252 }
Kees Cook344476e2018-06-12 14:04:32 -07004253 sbi->s_group_desc = kvmalloc_array(db_count,
4254 sizeof(struct buffer_head *),
4255 GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004256 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004257 ext4_msg(sb, KERN_ERR, "not enough memory");
Theodore Ts'o2cde4172012-05-28 17:49:54 -04004258 ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004259 goto failed_mount;
4260 }
4261
Pekka Enberg705895b2009-02-15 18:07:52 -05004262 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004263
Andrew Perepechko85c8f172017-04-30 00:46:35 -04004264 /* Pre-read the descriptors into the buffer cache */
4265 for (i = 0; i < db_count; i++) {
4266 block = descriptor_loc(sb, logical_sb_block, i);
4267 sb_breadahead(sb, block);
4268 }
4269
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004270 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07004271 block = descriptor_loc(sb, logical_sb_block, i);
Gioh Kima8ac9002014-09-04 22:36:15 -04004272 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004273 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004274 ext4_msg(sb, KERN_ERR,
4275 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004276 db_count = i;
4277 goto failed_mount2;
4278 }
4279 }
Theodore Ts'o44de0222018-07-08 19:35:02 -04004280 sbi->s_gdb_count = db_count;
Theodore Ts'o829fa702016-08-01 00:51:02 -04004281 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004282 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Darrick J. Wong6a797d22015-10-17 16:16:04 -04004283 ret = -EFSCORRUPTED;
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004284 goto failed_mount2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004285 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04004286
Kees Cook235699a2017-10-18 12:45:17 -04004287 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
Tao Ma04496412011-04-05 19:55:28 -04004288
Dmitry Monakhova75ae78f2013-04-03 22:10:52 -04004289 /* Register extent status tree shrinker */
Zheng Liueb68d0e2014-09-01 22:26:49 -04004290 if (ext4_es_register_shrinker(sbi))
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004291 goto failed_mount3;
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004292
Alex Tomasc9de5602008-01-29 00:19:52 -05004293 sbi->s_stripe = ext4_get_stripe_size(sbi);
Zheng Liu67a5da52012-08-17 09:54:17 -04004294 sbi->s_extent_max_zeroout_kb = 32;
Alex Tomasc9de5602008-01-29 00:19:52 -05004295
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004296 /*
4297 * set up enough so that it can read an inode
4298 */
Theodore Ts'of6e63f92014-09-18 17:12:30 -04004299 sb->s_op = &ext4_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07004300 sb->s_export_op = &ext4_export_ops;
4301 sb->s_xattr = ext4_xattr_handlers;
Chandan Rajendra643fa962018-12-12 15:20:12 +05304302#ifdef CONFIG_FS_ENCRYPTION
Jaegeuk Kima7550b32016-07-10 14:01:03 -04004303 sb->s_cop = &ext4_cryptops;
Eric Biggersffcc41822017-10-09 12:15:38 -07004304#endif
Eric Biggersc93d8f82019-07-22 09:26:24 -07004305#ifdef CONFIG_FS_VERITY
4306 sb->s_vop = &ext4_verityops;
4307#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004308#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07004309 sb->dq_op = &ext4_quota_operations;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004310 if (ext4_has_feature_quota(sb))
Jan Kara1fa5efe2014-10-08 18:26:54 +02004311 sb->s_qcop = &dquot_quotactl_sysfile_ops;
Jan Kara262b4662013-03-02 17:57:08 -05004312 else
4313 sb->s_qcop = &ext4_qctl_operations;
Li Xi689c958c2016-01-08 16:01:22 -05004314 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004315#endif
Christoph Hellwig85787092017-05-10 15:06:33 +02004316 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
Aneesh Kumar K.Vf2fa2ff2011-01-29 18:43:40 +05304317
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004318 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04004319 mutex_init(&sbi->s_orphan_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004320
4321 sb->s_root = NULL;
4322
4323 needs_recovery = (es->s_last_orphan != 0 ||
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004324 ext4_has_feature_journal_needs_recovery(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004325
David Howellsbc98a422017-07-17 08:45:34 +01004326 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
Johann Lombardic5e06d12011-05-24 18:31:25 -04004327 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
Darrick J. Wong50460fe2014-10-30 10:53:16 -04004328 goto failed_mount3a;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004329
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004330 /*
4331 * The first inode we look at is the journal inode. Don't try
4332 * root first: it may be modified in the journal!
4333 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004334 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
Theodore Ts'o4753d8a2017-02-05 01:26:48 -05004335 err = ext4_load_journal(sb, es, journal_devnum);
4336 if (err)
Darrick J. Wong50460fe2014-10-30 10:53:16 -04004337 goto failed_mount3a;
David Howellsbc98a422017-07-17 08:45:34 +01004338 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004339 ext4_has_feature_journal_needs_recovery(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004340 ext4_msg(sb, KERN_ERR, "required journal recovery "
4341 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004342 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004343 } else {
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04004344 /* Nojournal mode, all journal mount options are illegal */
4345 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4346 ext4_msg(sb, KERN_ERR, "can't mount with "
4347 "journal_checksum, fs mounted w/o journal");
4348 goto failed_mount_wq;
4349 }
4350 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4351 ext4_msg(sb, KERN_ERR, "can't mount with "
4352 "journal_async_commit, fs mounted w/o journal");
4353 goto failed_mount_wq;
4354 }
4355 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4356 ext4_msg(sb, KERN_ERR, "can't mount with "
4357 "commit=%lu, fs mounted w/o journal",
4358 sbi->s_commit_interval / HZ);
4359 goto failed_mount_wq;
4360 }
4361 if (EXT4_MOUNT_DATA_FLAGS &
4362 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4363 ext4_msg(sb, KERN_ERR, "can't mount with "
4364 "data=, fs mounted w/o journal");
4365 goto failed_mount_wq;
4366 }
Debabrata Banerjee50b29d82019-04-30 23:08:15 -04004367 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04004368 clear_opt(sb, JOURNAL_CHECKSUM);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004369 clear_opt(sb, DATA_FLAGS);
Frank Mayhar03901312009-01-07 00:06:22 -05004370 sbi->s_journal = NULL;
4371 needs_recovery = 0;
4372 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004373 }
4374
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004375 if (ext4_has_feature_64bit(sb) &&
Jose R. Santoseb40a092007-07-18 08:37:25 -04004376 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4377 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004378 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004379 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04004380 }
4381
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04004382 if (!set_journal_csum_feature_set(sb)) {
4383 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4384 "feature set");
4385 goto failed_mount_wq;
Linus Torvaldsd4da6c9c2009-11-02 10:15:27 -08004386 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05004387
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004388 /* We have now updated the journal if required, so we can
4389 * validate the data journaling mode. */
4390 switch (test_opt(sb, DATA_FLAGS)) {
4391 case 0:
4392 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07004393 * capabilities: ORDERED_DATA if the journal can
4394 * cope, else JOURNAL_DATA
4395 */
Mingming Caodab291a2006-10-11 01:21:01 -07004396 if (jbd2_journal_check_available_features
Tyson Nottingham27f394a72018-03-30 00:56:10 -04004397 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004398 set_opt(sb, ORDERED_DATA);
Tyson Nottingham27f394a72018-03-30 00:56:10 -04004399 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4400 } else {
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05004401 set_opt(sb, JOURNAL_DATA);
Tyson Nottingham27f394a72018-03-30 00:56:10 -04004402 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4403 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004404 break;
4405
Mingming Cao617ba132006-10-11 01:20:53 -07004406 case EXT4_MOUNT_ORDERED_DATA:
4407 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07004408 if (!jbd2_journal_check_available_features
4409 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004410 ext4_msg(sb, KERN_ERR, "Journal does not support "
4411 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05004412 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004413 }
4414 default:
4415 break;
4416 }
Jan Karaab04df72016-12-03 16:20:53 -05004417
4418 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4419 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4420 ext4_msg(sb, KERN_ERR, "can't mount with "
4421 "journal_async_commit in data=ordered mode");
4422 goto failed_mount_wq;
4423 }
4424
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004425 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004426
Bobi Jam18aadd42012-02-20 17:53:02 -05004427 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4428
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004429no_journal:
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04004430 if (!test_opt(sb, NO_MBCACHE)) {
4431 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4432 if (!sbi->s_ea_block_cache) {
Tahsin Erdogandec214d2017-06-22 11:44:55 -04004433 ext4_msg(sb, KERN_ERR,
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04004434 "Failed to create ea_block_cache");
Tahsin Erdogandec214d2017-06-22 11:44:55 -04004435 goto failed_mount_wq;
4436 }
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04004437
4438 if (ext4_has_feature_ea_inode(sb)) {
4439 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4440 if (!sbi->s_ea_inode_cache) {
4441 ext4_msg(sb, KERN_ERR,
4442 "Failed to create ea_inode_cache");
4443 goto failed_mount_wq;
4444 }
4445 }
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04004446 }
4447
Eric Biggersc93d8f82019-07-22 09:26:24 -07004448 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4449 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4450 goto failed_mount_wq;
4451 }
4452
David Howellsbc98a422017-07-17 08:45:34 +01004453 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004454 !ext4_has_feature_encrypt(sb)) {
4455 ext4_set_feature_encrypt(sb);
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04004456 ext4_commit_super(sb, 1);
4457 }
4458
Tejun Heofd89d5f2011-02-01 11:42:42 +01004459 /*
Theodore Ts'o952fc182012-07-09 16:27:05 -04004460 * Get the # of file system overhead blocks from the
4461 * superblock if present.
4462 */
4463 if (es->s_overhead_clusters)
4464 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4465 else {
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004466 err = ext4_calculate_overhead(sb);
4467 if (err)
Theodore Ts'o952fc182012-07-09 16:27:05 -04004468 goto failed_mount_wq;
4469 }
4470
4471 /*
Tejun Heofd89d5f2011-02-01 11:42:42 +01004472 * The maximum number of concurrent works can be high and
4473 * concurrency isn't really necessary. Limit it to 1.
4474 */
Jan Kara2e8fa542013-06-04 14:21:02 -04004475 EXT4_SB(sb)->rsv_conversion_wq =
4476 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4477 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4478 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004479 ret = -ENOMEM;
Jan Kara2e8fa542013-06-04 14:21:02 -04004480 goto failed_mount4;
4481 }
4482
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004483 /*
Mingming Caodab291a2006-10-11 01:21:01 -07004484 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004485 * so we can safely mount the rest of the filesystem now.
4486 */
4487
Theodore Ts'o8a363972018-12-19 12:29:13 -05004488 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
David Howells1d1fe1e2008-02-07 00:15:37 -08004489 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004490 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004491 ret = PTR_ERR(root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004492 root = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004493 goto failed_mount4;
4494 }
4495 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004496 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Al Viro94bf6082012-01-09 15:53:24 -05004497 iput(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004498 goto failed_mount4;
4499 }
Gabriel Krisman Bertazib886ee32019-04-25 14:12:08 -04004500
4501#ifdef CONFIG_UNICODE
4502 if (sbi->s_encoding)
4503 sb->s_d_op = &ext4_dentry_ops;
4504#endif
4505
Al Viro48fde702012-01-08 22:15:13 -05004506 sb->s_root = d_make_root(root);
David Howells1d1fe1e2008-02-07 00:15:37 -08004507 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004508 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004509 ret = -ENOMEM;
4510 goto failed_mount4;
4511 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004512
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04004513 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4514 if (ret == -EROFS) {
Linus Torvalds1751e8a2017-11-27 13:05:09 -08004515 sb->s_flags |= SB_RDONLY;
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04004516 ret = 0;
4517 } else if (ret)
4518 goto failed_mount4a;
Kalpak Shahef7f3832007-07-18 09:15:20 -04004519
Barret Rhoden7bc04c52019-04-25 11:55:50 -04004520 ext4_clamp_want_extra_isize(sb);
Kalpak Shahef7f3832007-07-18 09:15:20 -04004521
Theodore Ts'ob5799012015-09-23 12:44:17 -04004522 ext4_set_resv_clusters(sb);
Lukas Czerner27dd4382013-04-09 22:11:22 -04004523
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004524 err = ext4_setup_system_zone(sb);
4525 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004526 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04004527 "zone (%d)", err);
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004528 goto failed_mount4a;
4529 }
4530
4531 ext4_ext_init(sb);
4532 err = ext4_mb_init(sb);
4533 if (err) {
4534 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4535 err);
Tao Madcf2d802011-10-06 12:10:11 -04004536 goto failed_mount5;
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04004537 }
4538
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004539 block = ext4_count_free_clusters(sb);
4540 ext4_free_blocks_count_set(sbi->s_es,
4541 EXT4_C2B(sbi, block));
Theodore Ts'o4274f512018-09-01 14:42:14 -04004542 ext4_superblock_csum_set(sb);
Tejun Heo908c7f12014-09-08 09:51:29 +09004543 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4544 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004545 if (!err) {
4546 unsigned long freei = ext4_count_free_inodes(sb);
4547 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
Theodore Ts'o4274f512018-09-01 14:42:14 -04004548 ext4_superblock_csum_set(sb);
Tejun Heo908c7f12014-09-08 09:51:29 +09004549 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4550 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004551 }
4552 if (!err)
4553 err = percpu_counter_init(&sbi->s_dirs_counter,
Tejun Heo908c7f12014-09-08 09:51:29 +09004554 ext4_count_dirs(sb), GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004555 if (!err)
Tejun Heo908c7f12014-09-08 09:51:29 +09004556 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4557 GFP_KERNEL);
Daeho Jeongc8585c62016-04-25 23:22:35 -04004558 if (!err)
4559 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4560
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004561 if (err) {
4562 ext4_msg(sb, KERN_ERR, "insufficient memory");
4563 goto failed_mount6;
4564 }
4565
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004566 if (ext4_has_feature_flex_bg(sb))
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004567 if (!ext4_fill_flex_info(sb)) {
4568 ext4_msg(sb, KERN_ERR,
4569 "unable to initialize "
4570 "flex_bg meta info!");
4571 goto failed_mount6;
4572 }
4573
Lukas Czernerbfff6872010-10-27 21:30:05 -04004574 err = ext4_register_li_request(sb, first_not_zeroed);
4575 if (err)
Tao Madcf2d802011-10-06 12:10:11 -04004576 goto failed_mount6;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004577
Theodore Ts'ob5799012015-09-23 12:44:17 -04004578 err = ext4_register_sysfs(sb);
Tao Madcf2d802011-10-06 12:10:11 -04004579 if (err)
4580 goto failed_mount7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004581
Jan Kara9b2ff3572013-03-02 18:22:38 -05004582#ifdef CONFIG_QUOTA
4583 /* Enable quota usage during mount. */
David Howellsbc98a422017-07-17 08:45:34 +01004584 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
Jan Kara9b2ff3572013-03-02 18:22:38 -05004585 err = ext4_enable_quotas(sb);
4586 if (err)
4587 goto failed_mount8;
4588 }
4589#endif /* CONFIG_QUOTA */
4590
Mingming Cao617ba132006-10-11 01:20:53 -07004591 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4592 ext4_orphan_cleanup(sb, es);
4593 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05004594 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004595 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05004596 ext4_mark_recovery_complete(sb, es);
4597 }
4598 if (EXT4_SB(sb)->s_journal) {
4599 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4600 descr = " journalled data mode";
4601 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4602 descr = " ordered data mode";
4603 else
4604 descr = " writeback data mode";
4605 } else
4606 descr = "out journal";
4607
Lukas Czerner79add3a2012-11-08 13:28:29 -05004608 if (test_opt(sb, DISCARD)) {
4609 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4610 if (!blk_queue_discard(q))
4611 ext4_msg(sb, KERN_WARNING,
4612 "mounting with \"discard\" option, but "
4613 "the device does not support discard");
4614 }
4615
Theodore Ts'oe294a532015-08-15 14:59:44 -04004616 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4617 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
Theodore Ts'o5aee0f82016-11-18 13:24:26 -05004618 "Opts: %.*s%s%s", descr,
4619 (int) sizeof(sbi->s_es->s_mount_opts),
4620 sbi->s_es->s_mount_opts,
Theodore Ts'oe294a532015-08-15 14:59:44 -04004621 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004622
Theodore Ts'o66e61a92010-07-27 11:56:04 -04004623 if (es->s_error_count)
4624 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004625
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04004626 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4627 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4628 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4629 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4630
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004631 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004632 return 0;
4633
Mingming Cao617ba132006-10-11 01:20:53 -07004634cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004635 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004636 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004637 goto failed_mount;
4638
Theodore Ts'o72ba7452013-01-24 23:24:54 -05004639#ifdef CONFIG_QUOTA
4640failed_mount8:
Theodore Ts'oebd173b2015-09-23 12:46:17 -04004641 ext4_unregister_sysfs(sb);
Theodore Ts'o72ba7452013-01-24 23:24:54 -05004642#endif
Tao Madcf2d802011-10-06 12:10:11 -04004643failed_mount7:
4644 ext4_unregister_li_request(sb);
4645failed_mount6:
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004646 ext4_mb_release(sb);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004647 if (sbi->s_flex_groups)
Al Virob93b41d2014-11-20 12:19:11 -05004648 kvfree(sbi->s_flex_groups);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004649 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4650 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4651 percpu_counter_destroy(&sbi->s_dirs_counter);
4652 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Vasily Averinaf18e352018-11-07 10:56:28 -05004653 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
Azat Khuzhin00764932014-04-07 10:54:20 -04004654failed_mount5:
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004655 ext4_ext_release(sb);
4656 ext4_release_system_zone(sb);
4657failed_mount4a:
Al Viro94bf6082012-01-09 15:53:24 -05004658 dput(sb->s_root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004659 sb->s_root = NULL;
Al Viro94bf6082012-01-09 15:53:24 -05004660failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04004661 ext4_msg(sb, KERN_ERR, "mount failed");
Jan Kara2e8fa542013-06-04 14:21:02 -04004662 if (EXT4_SB(sb)->rsv_conversion_wq)
4663 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -04004664failed_mount_wq:
Chengguang Xu50c15df2018-12-04 00:24:42 -05004665 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4666 sbi->s_ea_inode_cache = NULL;
4667
4668 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4669 sbi->s_ea_block_cache = NULL;
4670
Frank Mayhar03901312009-01-07 00:06:22 -05004671 if (sbi->s_journal) {
4672 jbd2_journal_destroy(sbi->s_journal);
4673 sbi->s_journal = NULL;
4674 }
Darrick J. Wong50460fe2014-10-30 10:53:16 -04004675failed_mount3a:
Zheng Liud3922a72013-07-01 08:12:37 -04004676 ext4_es_unregister_shrinker(sbi);
Zheng Liueb68d0e2014-09-01 22:26:49 -04004677failed_mount3:
Al Viro9105bb12013-12-08 20:52:31 -05004678 del_timer_sync(&sbi->s_err_report);
Johann Lombardic5e06d12011-05-24 18:31:25 -04004679 if (sbi->s_mmp_tsk)
4680 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004681failed_mount2:
4682 for (i = 0; i < db_count; i++)
4683 brelse(sbi->s_group_desc[i]);
Al Virob93b41d2014-11-20 12:19:11 -05004684 kvfree(sbi->s_group_desc);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004685failed_mount:
Darrick J. Wong04419842012-04-29 18:27:10 -04004686 if (sbi->s_chksum_driver)
4687 crypto_free_shash(sbi->s_chksum_driver);
Gabriel Krisman Bertazic83ad552019-04-25 14:05:42 -04004688
4689#ifdef CONFIG_UNICODE
4690 utf8_unload(sbi->s_encoding);
4691#endif
4692
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004693#ifdef CONFIG_QUOTA
Jan Karaa2d4a642014-09-11 11:15:15 -04004694 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Theodore Ts'o0ba33fa2019-05-12 04:49:47 -04004695 kfree(get_qf_name(sb, sbi, i));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004696#endif
Mingming Cao617ba132006-10-11 01:20:53 -07004697 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004698 brelse(bh);
4699out_fail:
4700 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04004701 kfree(sbi->s_blockgroup_lock);
Theodore Ts'o5aee0f82016-11-18 13:24:26 -05004702out_free_base:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004703 kfree(sbi);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004704 kfree(orig_data);
Dan Williams5e405592017-08-24 16:42:48 -07004705 fs_put_dax(dax_dev);
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004706 return err ? err : ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004707}
4708
4709/*
4710 * Setup any per-fs journal parameters now. We'll do this both on
4711 * initial mount, once the journal has been initialised but before we've
4712 * done any recovery; and again on any subsequent remount.
4713 */
Mingming Cao617ba132006-10-11 01:20:53 -07004714static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004715{
Mingming Cao617ba132006-10-11 01:20:53 -07004716 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004717
Theodore Ts'o30773842009-01-03 20:27:38 -05004718 journal->j_commit_interval = sbi->s_commit_interval;
4719 journal->j_min_batch_time = sbi->s_min_batch_time;
4720 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004721
Theodore Ts'oa931da62010-08-03 21:35:12 -04004722 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004723 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07004724 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004725 else
Mingming Caodab291a2006-10-11 01:21:01 -07004726 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04004727 if (test_opt(sb, DATA_ERR_ABORT))
4728 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4729 else
4730 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04004731 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004732}
4733
Eric Whitneyc6cb7e72016-09-30 02:05:09 -04004734static struct inode *ext4_get_journal_inode(struct super_block *sb,
4735 unsigned int journal_inum)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004736{
4737 struct inode *journal_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004738
Eric Whitneyc6cb7e72016-09-30 02:05:09 -04004739 /*
4740 * Test for the existence of a valid inode on disk. Bad things
4741 * happen if we iget() an unused inode, as the subsequent iput()
4742 * will try to delete it.
4743 */
Theodore Ts'o8a363972018-12-19 12:29:13 -05004744 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
David Howells1d1fe1e2008-02-07 00:15:37 -08004745 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004746 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004747 return NULL;
4748 }
4749 if (!journal_inode->i_nlink) {
4750 make_bad_inode(journal_inode);
4751 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04004752 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004753 return NULL;
4754 }
4755
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04004756 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004757 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08004758 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004759 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004760 iput(journal_inode);
4761 return NULL;
4762 }
Eric Whitneyc6cb7e72016-09-30 02:05:09 -04004763 return journal_inode;
4764}
4765
4766static journal_t *ext4_get_journal(struct super_block *sb,
4767 unsigned int journal_inum)
4768{
4769 struct inode *journal_inode;
4770 journal_t *journal;
4771
4772 BUG_ON(!ext4_has_feature_journal(sb));
4773
4774 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4775 if (!journal_inode)
4776 return NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004777
Mingming Caodab291a2006-10-11 01:21:01 -07004778 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004779 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004780 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004781 iput(journal_inode);
4782 return NULL;
4783 }
4784 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004785 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004786 return journal;
4787}
4788
Mingming Cao617ba132006-10-11 01:20:53 -07004789static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004790 dev_t j_dev)
4791{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004792 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004793 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07004794 ext4_fsblk_t start;
4795 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004796 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07004797 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004798 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004799 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004800 struct block_device *bdev;
4801
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004802 BUG_ON(!ext4_has_feature_journal(sb));
Frank Mayhar03901312009-01-07 00:06:22 -05004803
Eric Sandeenb31e1552009-06-04 17:36:36 -04004804 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004805 if (bdev == NULL)
4806 return NULL;
4807
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004808 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04004809 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004810 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004811 ext4_msg(sb, KERN_ERR,
4812 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004813 goto out_bdev;
4814 }
4815
Mingming Cao617ba132006-10-11 01:20:53 -07004816 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4817 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004818 set_blocksize(bdev, blocksize);
4819 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004820 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4821 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004822 goto out_bdev;
4823 }
4824
Theodore Ts'o2716b802012-05-28 17:47:52 -04004825 es = (struct ext4_super_block *) (bh->b_data + offset);
Mingming Cao617ba132006-10-11 01:20:53 -07004826 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004827 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07004828 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004829 ext4_msg(sb, KERN_ERR, "external journal has "
4830 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004831 brelse(bh);
4832 goto out_bdev;
4833 }
4834
Darrick J. Wongdf4763b2014-09-11 11:44:36 -04004835 if ((le32_to_cpu(es->s_feature_ro_compat) &
4836 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4837 es->s_checksum != ext4_superblock_csum(sb, es)) {
4838 ext4_msg(sb, KERN_ERR, "external journal has "
4839 "corrupt superblock");
4840 brelse(bh);
4841 goto out_bdev;
4842 }
4843
Mingming Cao617ba132006-10-11 01:20:53 -07004844 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004845 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004846 brelse(bh);
4847 goto out_bdev;
4848 }
4849
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004850 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004851 start = sb_block + 1;
4852 brelse(bh); /* we're done with the superblock */
4853
Mingming Caodab291a2006-10-11 01:21:01 -07004854 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004855 start, len, blocksize);
4856 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004857 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004858 goto out_bdev;
4859 }
4860 journal->j_private = sb;
Mike Christiedfec8a12016-06-05 14:31:44 -05004861 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004862 wait_on_buffer(journal->j_sb_buffer);
4863 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004864 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004865 goto out_journal;
4866 }
4867 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004868 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4869 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004870 be32_to_cpu(journal->j_superblock->s_nr_users));
4871 goto out_journal;
4872 }
Mingming Cao617ba132006-10-11 01:20:53 -07004873 EXT4_SB(sb)->journal_bdev = bdev;
4874 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004875 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004876
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004877out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07004878 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004879out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07004880 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004881 return NULL;
4882}
4883
Mingming Cao617ba132006-10-11 01:20:53 -07004884static int ext4_load_journal(struct super_block *sb,
4885 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004886 unsigned long journal_devnum)
4887{
4888 journal_t *journal;
4889 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4890 dev_t journal_dev;
4891 int err = 0;
4892 int really_read_only;
4893
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004894 BUG_ON(!ext4_has_feature_journal(sb));
Frank Mayhar03901312009-01-07 00:06:22 -05004895
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004896 if (journal_devnum &&
4897 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004898 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4899 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004900 journal_dev = new_decode_dev(journal_devnum);
4901 } else
4902 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4903
4904 really_read_only = bdev_read_only(sb->s_bdev);
4905
4906 /*
4907 * Are we loading a blank journal or performing recovery after a
4908 * crash? For recovery, we need to check in advance whether we
4909 * can get read-write access to the device.
4910 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004911 if (ext4_has_feature_journal_needs_recovery(sb)) {
David Howellsbc98a422017-07-17 08:45:34 +01004912 if (sb_rdonly(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004913 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4914 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004915 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004916 ext4_msg(sb, KERN_ERR, "write access "
Simon Ruderichd98bf8cd2017-10-18 13:06:37 -04004917 "unavailable, cannot proceed "
4918 "(try mounting with noload)");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004919 return -EROFS;
4920 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04004921 ext4_msg(sb, KERN_INFO, "write access will "
4922 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004923 }
4924 }
4925
4926 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004927 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4928 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004929 return -EINVAL;
4930 }
4931
4932 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07004933 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004934 return -EINVAL;
4935 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07004936 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004937 return -EINVAL;
4938 }
4939
Theodore Ts'o90576c02009-09-29 15:51:30 -04004940 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04004941 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04004942
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004943 if (!ext4_has_feature_journal_needs_recovery(sb))
Mingming Caodab291a2006-10-11 01:21:01 -07004944 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004945 if (!err) {
4946 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4947 if (save)
4948 memcpy(save, ((char *) es) +
4949 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07004950 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004951 if (save)
4952 memcpy(((char *) es) + EXT4_S_ERR_START,
4953 save, EXT4_S_ERR_LEN);
4954 kfree(save);
4955 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004956
4957 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004958 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07004959 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004960 return err;
4961 }
4962
Mingming Cao617ba132006-10-11 01:20:53 -07004963 EXT4_SB(sb)->s_journal = journal;
4964 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004965
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04004966 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004967 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4968 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004969
4970 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004971 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004972 }
4973
4974 return 0;
4975}
4976
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004977static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004978{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004979 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07004980 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08004981 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004982
Theodore Ts'obdfe0cbd2015-08-16 10:03:57 -04004983 if (!sbh || block_device_ejected(sb))
Takashi Satoc4be0c12009-01-09 16:40:58 -08004984 return error;
Jon Derricka17712c2018-07-02 18:45:18 -04004985
4986 /*
4987 * The superblock bh should be mapped, but it might not be if the
4988 * device was hot-removed. Not much we can do but fail the I/O.
4989 */
4990 if (!buffer_mapped(sbh))
4991 return error;
4992
Theodore Ts'o71290b32009-09-10 17:31:04 -04004993 /*
4994 * If the file system is mounted read-only, don't update the
4995 * superblock write time. This avoids updating the superblock
4996 * write time when we are mounting the root file system
4997 * read/only but we need to replay the journal; at that point,
4998 * for people who are east of GMT and who make their clock
4999 * tick in localtime for Windows bug-for-bug compatibility,
5000 * the clock is set in the future, and this will cause e2fsck
5001 * to complain and force a full file system check.
5002 */
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005003 if (!(sb->s_flags & SB_RDONLY))
Arnd Bergmann6a0678a2018-07-29 15:51:48 -04005004 ext4_update_tstamp(es, s_wtime);
Theodore Ts'of613dfc2010-07-27 11:56:08 -04005005 if (sb->s_bdev->bd_part)
5006 es->s_kbytes_written =
5007 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Michael Callahandbae2c52018-07-18 04:47:38 -07005008 ((part_stat_read(sb->s_bdev->bd_part,
5009 sectors[STAT_WRITE]) -
Theodore Ts'oafc32f72009-02-28 19:39:58 -05005010 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04005011 else
5012 es->s_kbytes_written =
5013 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04005014 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5015 ext4_free_blocks_count_set(es,
Theodore Ts'o57042652011-09-09 18:56:51 -04005016 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5017 &EXT4_SB(sb)->s_freeclusters_counter)));
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04005018 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5019 es->s_free_inodes_count =
5020 cpu_to_le32(percpu_counter_sum_positive(
Theodore Ts'oce7e0102010-11-03 12:03:21 -04005021 &EXT4_SB(sb)->s_freeinodes_counter));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005022 BUFFER_TRACE(sbh, "marking dirty");
Theodore Ts'o06db49e2012-10-10 01:06:58 -04005023 ext4_superblock_csum_set(sb);
Theodore Ts'o1566a482016-11-13 22:02:29 -05005024 if (sync)
5025 lock_buffer(sbh);
Theodore Ts'oe8680782018-12-30 23:20:39 -05005026 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
Pranay Kr. Srivastava4743f832016-07-04 10:24:52 -04005027 /*
5028 * Oh, dear. A previous attempt to write the
5029 * superblock failed. This could happen because the
5030 * USB device was yanked out. Or it could happen to
5031 * be a transient write error and maybe the block will
5032 * be remapped. Nothing we can do but to retry the
5033 * write and hope for the best.
5034 */
5035 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5036 "superblock detected");
5037 clear_buffer_write_io_error(sbh);
5038 set_buffer_uptodate(sbh);
5039 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005040 mark_buffer_dirty(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04005041 if (sync) {
Theodore Ts'o1566a482016-11-13 22:02:29 -05005042 unlock_buffer(sbh);
Daeho Jeong564bc402015-07-23 09:46:11 -04005043 error = __sync_dirty_buffer(sbh,
Jan Kara00473372017-05-04 10:58:03 -04005044 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04005045 if (buffer_write_io_error(sbh)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005046 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5047 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04005048 clear_buffer_write_io_error(sbh);
5049 set_buffer_uptodate(sbh);
5050 }
5051 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08005052 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005053}
5054
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005055/*
5056 * Have we just finished recovery? If so, and if we are mounting (or
5057 * remounting) the filesystem readonly, then we will end up with a
5058 * consistent fs on disk. Record that fact.
5059 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005060static void ext4_mark_recovery_complete(struct super_block *sb,
5061 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005062{
Mingming Cao617ba132006-10-11 01:20:53 -07005063 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005064
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005065 if (!ext4_has_feature_journal(sb)) {
Frank Mayhar03901312009-01-07 00:06:22 -05005066 BUG_ON(journal != NULL);
5067 return;
5068 }
Mingming Caodab291a2006-10-11 01:21:01 -07005069 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005070 if (jbd2_journal_flush(journal) < 0)
5071 goto out;
5072
David Howellsbc98a422017-07-17 08:45:34 +01005073 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005074 ext4_clear_feature_journal_needs_recovery(sb);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04005075 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005076 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005077
5078out:
Mingming Caodab291a2006-10-11 01:21:01 -07005079 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005080}
5081
5082/*
5083 * If we are mounting (or read-write remounting) a filesystem whose journal
5084 * has recorded an error from a previous lifetime, move that error to the
5085 * main filesystem now.
5086 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005087static void ext4_clear_journal_err(struct super_block *sb,
5088 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005089{
5090 journal_t *journal;
5091 int j_errno;
5092 const char *errstr;
5093
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005094 BUG_ON(!ext4_has_feature_journal(sb));
Frank Mayhar03901312009-01-07 00:06:22 -05005095
Mingming Cao617ba132006-10-11 01:20:53 -07005096 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005097
5098 /*
5099 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07005100 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005101 */
5102
Mingming Caodab291a2006-10-11 01:21:01 -07005103 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005104 if (j_errno) {
5105 char nbuf[16];
5106
Mingming Cao617ba132006-10-11 01:20:53 -07005107 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05005108 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005109 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05005110 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005111
Mingming Cao617ba132006-10-11 01:20:53 -07005112 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5113 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04005114 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005115
Mingming Caodab291a2006-10-11 01:21:01 -07005116 jbd2_journal_clear_err(journal);
Theodore Ts'od796c522012-08-05 19:04:57 -04005117 jbd2_journal_update_sb_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005118 }
5119}
5120
5121/*
5122 * Force the running and committing transactions to commit,
5123 * and wait on the commit.
5124 */
Mingming Cao617ba132006-10-11 01:20:53 -07005125int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005126{
5127 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005128
David Howellsbc98a422017-07-17 08:45:34 +01005129 if (sb_rdonly(sb))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005130 return 0;
5131
Mingming Cao617ba132006-10-11 01:20:53 -07005132 journal = EXT4_SB(sb)->s_journal;
Guo Chaob1deefc2013-01-28 21:41:02 -05005133 return ext4_journal_force_commit(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005134}
5135
Mingming Cao617ba132006-10-11 01:20:53 -07005136static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005137{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05005138 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05005139 tid_t target;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04005140 bool needs_barrier = false;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04005141 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005142
Jun Piao49598e02018-01-11 13:17:49 -05005143 if (unlikely(ext4_forced_shutdown(sbi)))
Theodore Ts'o0db1ff22017-02-05 01:28:48 -05005144 return 0;
5145
Theodore Ts'o9bffad12009-06-17 11:48:11 -04005146 trace_ext4_sync_fs(sb, wait);
Jan Kara2e8fa542013-06-04 14:21:02 -04005147 flush_workqueue(sbi->rsv_conversion_wq);
Jan Karaa1177822012-07-03 16:45:29 +02005148 /*
5149 * Writeback quota in non-journalled quota case - journalled quota has
5150 * no dirty dquots
5151 */
5152 dquot_writeback_dquots(sb, -1);
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04005153 /*
5154 * Data writeback is possible w/o journal transaction, so barrier must
5155 * being sent at the end of the function. But we can skip it if
5156 * transaction_commit will do it for us.
5157 */
Theodore Ts'obda32532014-09-18 16:12:37 -04005158 if (sbi->s_journal) {
5159 target = jbd2_get_latest_transaction(sbi->s_journal);
5160 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5161 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5162 needs_barrier = true;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04005163
Theodore Ts'obda32532014-09-18 16:12:37 -04005164 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5165 if (wait)
5166 ret = jbd2_log_wait_commit(sbi->s_journal,
5167 target);
5168 }
5169 } else if (wait && test_opt(sb, BARRIER))
5170 needs_barrier = true;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04005171 if (needs_barrier) {
5172 int err;
5173 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5174 if (!ret)
5175 ret = err;
5176 }
5177
5178 return ret;
5179}
5180
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005181/*
5182 * LVM calls this function before a (read-only) snapshot is created. This
5183 * gives us a chance to flush the journal completely and mark the fs clean.
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -04005184 *
5185 * Note that only this function cannot bring a filesystem to be in a clean
Jan Kara8e8ad8a2012-06-12 16:20:38 +02005186 * state independently. It relies on upper layer to stop all data & metadata
5187 * modifications.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005188 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08005189static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005190{
Takashi Satoc4be0c12009-01-09 16:40:58 -08005191 int error = 0;
5192 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005193
David Howellsbc98a422017-07-17 08:45:34 +01005194 if (sb_rdonly(sb))
Theodore Ts'o9ca92382009-05-01 12:52:25 -04005195 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005196
Theodore Ts'o9ca92382009-05-01 12:52:25 -04005197 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005198
Theodore Ts'obb044572014-09-18 17:12:02 -04005199 if (journal) {
5200 /* Now we set up the journal barrier. */
5201 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005202
Theodore Ts'obb044572014-09-18 17:12:02 -04005203 /*
5204 * Don't clear the needs_recovery flag if we failed to
5205 * flush the journal.
5206 */
5207 error = jbd2_journal_flush(journal);
5208 if (error < 0)
5209 goto out;
Eric Sandeenc642dc92015-08-15 10:45:06 -04005210
5211 /* Journal blocked and flushed, clear needs_recovery flag. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005212 ext4_clear_feature_journal_needs_recovery(sb);
Theodore Ts'obb044572014-09-18 17:12:02 -04005213 }
Theodore Ts'o9ca92382009-05-01 12:52:25 -04005214
Theodore Ts'o9ca92382009-05-01 12:52:25 -04005215 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04005216out:
Theodore Ts'obb044572014-09-18 17:12:02 -04005217 if (journal)
5218 /* we rely on upper layer to stop further updates */
5219 jbd2_journal_unlock_updates(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04005220 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005221}
5222
5223/*
5224 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5225 * flag here, even though the filesystem is not technically dirty yet.
5226 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08005227static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005228{
David Howellsbc98a422017-07-17 08:45:34 +01005229 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
Theodore Ts'o9ca92382009-05-01 12:52:25 -04005230 return 0;
5231
Eric Sandeenc642dc92015-08-15 10:45:06 -04005232 if (EXT4_SB(sb)->s_journal) {
5233 /* Reset the needs_recovery flag before the fs is unlocked. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005234 ext4_set_feature_journal_needs_recovery(sb);
Eric Sandeenc642dc92015-08-15 10:45:06 -04005235 }
5236
Theodore Ts'o9ca92382009-05-01 12:52:25 -04005237 ext4_commit_super(sb, 1);
Takashi Satoc4be0c12009-01-09 16:40:58 -08005238 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005239}
5240
Theodore Ts'o673c6102010-12-15 20:28:48 -05005241/*
5242 * Structure to save mount options for ext4_remount's benefit
5243 */
5244struct ext4_mount_options {
5245 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05005246 unsigned long s_mount_opt2;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08005247 kuid_t s_resuid;
5248 kgid_t s_resgid;
Theodore Ts'o673c6102010-12-15 20:28:48 -05005249 unsigned long s_commit_interval;
5250 u32 s_min_batch_time, s_max_batch_time;
5251#ifdef CONFIG_QUOTA
5252 int s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04005253 char *s_qf_names[EXT4_MAXQUOTAS];
Theodore Ts'o673c6102010-12-15 20:28:48 -05005254#endif
5255};
5256
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005257static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005258{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005259 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07005260 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005261 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07005262 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04005263 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04005264 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05005265 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Johann Lombardic5e06d12011-05-24 18:31:25 -04005266 int err = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005267#ifdef CONFIG_QUOTA
Chen Gang03dafb52013-01-24 23:24:58 -05005268 int i, j;
Theodore Ts'o33458ea2018-10-12 09:28:09 -04005269 char *to_free[EXT4_MAXQUOTAS];
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005270#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005271 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005272
Chengguang Xu21ac7382018-07-29 15:51:54 -04005273 if (data && !orig_data)
5274 return -ENOMEM;
5275
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005276 /* Store the original options */
5277 old_sb_flags = sb->s_flags;
5278 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05005279 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005280 old_opts.s_resuid = sbi->s_resuid;
5281 old_opts.s_resgid = sbi->s_resgid;
5282 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05005283 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5284 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005285#ifdef CONFIG_QUOTA
5286 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04005287 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05005288 if (sbi->s_qf_names[i]) {
Theodore Ts'o33458ea2018-10-12 09:28:09 -04005289 char *qf_name = get_qf_name(sb, sbi, i);
5290
5291 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
Chen Gang03dafb52013-01-24 23:24:58 -05005292 if (!old_opts.s_qf_names[i]) {
5293 for (j = 0; j < i; j++)
5294 kfree(old_opts.s_qf_names[j]);
Wei Yongjun3e36a162013-03-02 17:13:55 -05005295 kfree(orig_data);
Chen Gang03dafb52013-01-24 23:24:58 -05005296 return -ENOMEM;
5297 }
5298 } else
5299 old_opts.s_qf_names[i] = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005300#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05005301 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5302 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005303
Eric Sandeen661aa522012-02-20 17:53:04 -05005304 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005305 err = -EINVAL;
5306 goto restore_opts;
5307 }
5308
Barret Rhoden7bc04c52019-04-25 11:55:50 -04005309 ext4_clamp_want_extra_isize(sb);
5310
Darrick J. Wong6b992ff2014-10-30 10:53:16 -04005311 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5312 test_opt(sb, JOURNAL_CHECKSUM)) {
5313 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
Eric Sandeen2d5b86e2015-02-12 23:07:37 -05005314 "during remount not supported; ignoring");
5315 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05005316 }
5317
Piotr Sarna6ae65142013-08-08 23:02:24 -04005318 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5319 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5320 ext4_msg(sb, KERN_ERR, "can't mount with "
5321 "both data=journal and delalloc");
5322 err = -EINVAL;
5323 goto restore_opts;
5324 }
5325 if (test_opt(sb, DIOREAD_NOLOCK)) {
5326 ext4_msg(sb, KERN_ERR, "can't mount with "
5327 "both data=journal and dioread_nolock");
5328 err = -EINVAL;
5329 goto restore_opts;
5330 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08005331 if (test_opt(sb, DAX)) {
5332 ext4_msg(sb, KERN_ERR, "can't mount with "
5333 "both data=journal and dax");
5334 err = -EINVAL;
5335 goto restore_opts;
5336 }
Jan Karaab04df72016-12-03 16:20:53 -05005337 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5338 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5339 ext4_msg(sb, KERN_ERR, "can't mount with "
5340 "journal_async_commit in data=ordered mode");
5341 err = -EINVAL;
5342 goto restore_opts;
5343 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08005344 }
5345
Tahsin Erdogancdb7ee42017-06-22 11:55:14 -04005346 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5347 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5348 err = -EINVAL;
5349 goto restore_opts;
5350 }
5351
Ross Zwisler923ae0f2015-02-16 15:59:38 -08005352 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5353 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5354 "dax flag with busy inodes while remounting");
5355 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
Piotr Sarna6ae65142013-08-08 23:02:24 -04005356 }
5357
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04005358 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04005359 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005360
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005361 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5362 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005363
5364 es = sbi->s_es;
5365
Theodore Ts'ob3881f72009-01-05 22:46:26 -05005366 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05005367 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05005368 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5369 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005370
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005371 if (*flags & SB_LAZYTIME)
5372 sb->s_flags |= SB_LAZYTIME;
Theodore Ts'oa2fd66d2015-06-23 11:03:54 -04005373
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005374 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04005375 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005376 err = -EROFS;
5377 goto restore_opts;
5378 }
5379
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005380 if (*flags & SB_RDONLY) {
Theodore Ts'o38c03b32014-03-13 22:49:42 -04005381 err = sync_filesystem(sb);
5382 if (err < 0)
5383 goto restore_opts;
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04005384 err = dquot_suspend(sb, -1);
5385 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04005386 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04005387
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005388 /*
5389 * First of all, the unconditional stuff we have to do
5390 * to disable replay of the journal when we next remount
5391 */
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005392 sb->s_flags |= SB_RDONLY;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005393
5394 /*
5395 * OK, test if we are remounting a valid rw partition
5396 * readonly, and if so set the rdonly flag and then
5397 * mark the partition as valid again.
5398 */
Mingming Cao617ba132006-10-11 01:20:53 -07005399 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5400 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005401 es->s_state = cpu_to_le16(sbi->s_mount_state);
5402
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04005403 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05005404 ext4_mark_recovery_complete(sb, es);
Theodore Ts'o2dca60d2018-07-08 19:36:02 -04005405 if (sbi->s_mmp_tsk)
5406 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005407 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04005408 /* Make sure we can mount this feature set readwrite */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005409 if (ext4_has_feature_readonly(sb) ||
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05005410 !ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005411 err = -EROFS;
5412 goto restore_opts;
5413 }
Eric Sandeenead6596b2007-02-10 01:46:08 -08005414 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04005415 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005416 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04005417 */
5418 for (g = 0; g < sbi->s_groups_count; g++) {
5419 struct ext4_group_desc *gdp =
5420 ext4_get_group_desc(sb, g, NULL);
5421
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04005422 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005423 ext4_msg(sb, KERN_ERR,
5424 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005425 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
Theodore Ts'o8a266462008-07-26 14:34:21 -04005426 le16_to_cpu(gdp->bg_checksum));
Darrick J. Wong6a797d22015-10-17 16:16:04 -04005427 err = -EFSBADCRC;
Theodore Ts'o8a266462008-07-26 14:34:21 -04005428 goto restore_opts;
5429 }
5430 }
5431
5432 /*
Eric Sandeenead6596b2007-02-10 01:46:08 -08005433 * If we have an unprocessed orphan list hanging
5434 * around from a previously readonly bdev mount,
5435 * require a full umount/remount for now.
5436 */
5437 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005438 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead6596b2007-02-10 01:46:08 -08005439 "remount RDWR because of unprocessed "
5440 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04005441 "umount/remount instead");
Eric Sandeenead6596b2007-02-10 01:46:08 -08005442 err = -EINVAL;
5443 goto restore_opts;
5444 }
5445
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005446 /*
5447 * Mounting a RDONLY partition read-write, so reread
5448 * and store the current valid flag. (It may have
5449 * been changed by e2fsck since we originally mounted
5450 * the partition.)
5451 */
Frank Mayhar03901312009-01-07 00:06:22 -05005452 if (sbi->s_journal)
5453 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005454 sbi->s_mount_state = le16_to_cpu(es->s_state);
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04005455
5456 err = ext4_setup_super(sb, es, 0);
5457 if (err)
5458 goto restore_opts;
5459
5460 sb->s_flags &= ~SB_RDONLY;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005461 if (ext4_has_feature_mmp(sb))
Johann Lombardic5e06d12011-05-24 18:31:25 -04005462 if (ext4_multi_mount_protect(sb,
5463 le64_to_cpu(es->s_mmp_block))) {
5464 err = -EROFS;
5465 goto restore_opts;
5466 }
Christoph Hellwigc79d9672010-05-19 07:16:40 -04005467 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005468 }
5469 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04005470
5471 /*
5472 * Reinitialize lazy itable initialization thread based on
5473 * current settings
5474 */
David Howellsbc98a422017-07-17 08:45:34 +01005475 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
Lukas Czernerbfff6872010-10-27 21:30:05 -04005476 ext4_unregister_li_request(sb);
5477 else {
5478 ext4_group_t first_not_zeroed;
5479 first_not_zeroed = ext4_has_uninit_itable(sb);
5480 ext4_register_li_request(sb, first_not_zeroed);
5481 }
5482
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005483 ext4_setup_system_zone(sb);
Jaegeuk Kimc89128a2018-05-13 23:02:19 -04005484 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5485 err = ext4_commit_super(sb, 1);
5486 if (err)
5487 goto restore_opts;
5488 }
Frank Mayhar03901312009-01-07 00:06:22 -05005489
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005490#ifdef CONFIG_QUOTA
5491 /* Release old quota file names */
Jan Karaa2d4a642014-09-11 11:15:15 -04005492 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05005493 kfree(old_opts.s_qf_names[i]);
Aditya Kali7c319d32012-07-22 20:21:31 -04005494 if (enable_quota) {
5495 if (sb_any_quota_suspended(sb))
5496 dquot_resume(sb, -1);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005497 else if (ext4_has_feature_quota(sb)) {
Aditya Kali7c319d32012-07-22 20:21:31 -04005498 err = ext4_enable_quotas(sb);
Theodore Ts'o07724f92012-08-17 19:08:42 -04005499 if (err)
Aditya Kali7c319d32012-07-22 20:21:31 -04005500 goto restore_opts;
Aditya Kali7c319d32012-07-22 20:21:31 -04005501 }
5502 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005503#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005504
Linus Torvalds1751e8a2017-11-27 13:05:09 -08005505 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005506 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5507 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005508 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005509
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005510restore_opts:
5511 sb->s_flags = old_sb_flags;
5512 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05005513 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005514 sbi->s_resuid = old_opts.s_resuid;
5515 sbi->s_resgid = old_opts.s_resgid;
5516 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05005517 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5518 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005519#ifdef CONFIG_QUOTA
5520 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04005521 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Theodore Ts'o33458ea2018-10-12 09:28:09 -04005522 to_free[i] = get_qf_name(sb, sbi, i);
5523 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005524 }
Theodore Ts'o33458ea2018-10-12 09:28:09 -04005525 synchronize_rcu();
5526 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5527 kfree(to_free[i]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005528#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005529 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005530 return err;
5531}
5532
Li Xi689c958c2016-01-08 16:01:22 -05005533#ifdef CONFIG_QUOTA
5534static int ext4_statfs_project(struct super_block *sb,
5535 kprojid_t projid, struct kstatfs *buf)
5536{
5537 struct kqid qid;
5538 struct dquot *dquot;
5539 u64 limit;
5540 u64 curblock;
5541
5542 qid = make_kqid_projid(projid);
5543 dquot = dqget(sb, qid);
5544 if (IS_ERR(dquot))
5545 return PTR_ERR(dquot);
Jan Kara7b9ca4c2017-08-07 13:19:50 +02005546 spin_lock(&dquot->dq_dqb_lock);
Li Xi689c958c2016-01-08 16:01:22 -05005547
5548 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5549 dquot->dq_dqb.dqb_bsoftlimit :
5550 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5551 if (limit && buf->f_blocks > limit) {
Konstantin Khlebnikovf06925c2018-05-20 22:49:54 -04005552 curblock = (dquot->dq_dqb.dqb_curspace +
5553 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
Li Xi689c958c2016-01-08 16:01:22 -05005554 buf->f_blocks = limit;
5555 buf->f_bfree = buf->f_bavail =
5556 (buf->f_blocks > curblock) ?
5557 (buf->f_blocks - curblock) : 0;
5558 }
5559
5560 limit = dquot->dq_dqb.dqb_isoftlimit ?
5561 dquot->dq_dqb.dqb_isoftlimit :
5562 dquot->dq_dqb.dqb_ihardlimit;
5563 if (limit && buf->f_files > limit) {
5564 buf->f_files = limit;
5565 buf->f_ffree =
5566 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5567 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5568 }
5569
Jan Kara7b9ca4c2017-08-07 13:19:50 +02005570 spin_unlock(&dquot->dq_dqb_lock);
Li Xi689c958c2016-01-08 16:01:22 -05005571 dqput(dquot);
5572 return 0;
5573}
5574#endif
5575
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005576static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005577{
5578 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07005579 struct ext4_sb_info *sbi = EXT4_SB(sb);
5580 struct ext4_super_block *es = sbi->s_es;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005581 ext4_fsblk_t overhead = 0, resv_blocks;
Pekka Enberg960cc392006-12-06 20:35:29 -08005582 u64 fsid;
Kazuya Miod02a9392011-05-24 18:30:07 -04005583 s64 bfree;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005584 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005585
Theodore Ts'o952fc182012-07-09 16:27:05 -04005586 if (!test_opt(sb, MINIX_DF))
5587 overhead = sbi->s_overhead;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005588
Mingming Cao617ba132006-10-11 01:20:53 -07005589 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005590 buf->f_bsize = sb->s_blocksize;
Eric Sandeenb72f78c2012-11-08 10:33:36 -05005591 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
Theodore Ts'o57042652011-09-09 18:56:51 -04005592 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5593 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
Kazuya Miod02a9392011-05-24 18:30:07 -04005594 /* prevent underflow in case that few free space is available */
Theodore Ts'o57042652011-09-09 18:56:51 -04005595 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
Lukas Czerner27dd4382013-04-09 22:11:22 -04005596 buf->f_bavail = buf->f_bfree -
5597 (ext4_r_blocks_count(es) + resv_blocks);
5598 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005599 buf->f_bavail = 0;
5600 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07005601 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07005602 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08005603 fsid = le64_to_cpup((void *)es->s_uuid) ^
5604 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5605 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5606 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005607
Li Xi689c958c2016-01-08 16:01:22 -05005608#ifdef CONFIG_QUOTA
5609 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5610 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5611 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5612#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005613 return 0;
5614}
5615
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005616
5617#ifdef CONFIG_QUOTA
5618
Jan Karabc8230e2017-06-08 14:39:48 +02005619/*
5620 * Helper functions so that transaction is started before we acquire dqio_sem
5621 * to keep correct lock ordering of transaction > dqio_sem
5622 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005623static inline struct inode *dquot_to_inode(struct dquot *dquot)
5624{
Eric W. Biederman4c376dc2012-09-16 03:56:19 -07005625 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005626}
5627
Mingming Cao617ba132006-10-11 01:20:53 -07005628static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005629{
5630 int ret, err;
5631 handle_t *handle;
5632 struct inode *inode;
5633
5634 inode = dquot_to_inode(dquot);
Theodore Ts'o9924a922013-02-08 21:59:22 -05005635 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005636 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005637 if (IS_ERR(handle))
5638 return PTR_ERR(handle);
5639 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005640 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005641 if (!ret)
5642 ret = err;
5643 return ret;
5644}
5645
Mingming Cao617ba132006-10-11 01:20:53 -07005646static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005647{
5648 int ret, err;
5649 handle_t *handle;
5650
Theodore Ts'o9924a922013-02-08 21:59:22 -05005651 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005652 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005653 if (IS_ERR(handle))
5654 return PTR_ERR(handle);
5655 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005656 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005657 if (!ret)
5658 ret = err;
5659 return ret;
5660}
5661
Mingming Cao617ba132006-10-11 01:20:53 -07005662static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005663{
5664 int ret, err;
5665 handle_t *handle;
5666
Theodore Ts'o9924a922013-02-08 21:59:22 -05005667 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005668 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07005669 if (IS_ERR(handle)) {
5670 /* Release dquot anyway to avoid endless cycle in dqput() */
5671 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005672 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07005673 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005674 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005675 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005676 if (!ret)
5677 ret = err;
5678 return ret;
5679}
5680
Mingming Cao617ba132006-10-11 01:20:53 -07005681static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005682{
Jan Kara262b4662013-03-02 17:57:08 -05005683 struct super_block *sb = dquot->dq_sb;
5684 struct ext4_sb_info *sbi = EXT4_SB(sb);
5685
Jan Kara2c8be6b2008-05-13 21:27:55 -04005686 /* Are we journaling quotas? */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005687 if (ext4_has_feature_quota(sb) ||
Jan Kara262b4662013-03-02 17:57:08 -05005688 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005689 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005690 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005691 } else {
5692 return dquot_mark_dquot_dirty(dquot);
5693 }
5694}
5695
Mingming Cao617ba132006-10-11 01:20:53 -07005696static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005697{
5698 int ret, err;
5699 handle_t *handle;
5700
5701 /* Data block + inode block */
David Howells2b0143b2015-03-17 22:25:59 +00005702 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005703 if (IS_ERR(handle))
5704 return PTR_ERR(handle);
5705 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07005706 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005707 if (!ret)
5708 ret = err;
5709 return ret;
5710}
5711
5712/*
5713 * Turn on quotas during mount time - we need to find
5714 * the quota file and such...
5715 */
Mingming Cao617ba132006-10-11 01:20:53 -07005716static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005717{
Theodore Ts'o33458ea2018-10-12 09:28:09 -04005718 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
Christoph Hellwig287a8092010-05-19 07:16:45 -04005719 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005720}
5721
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005722static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5723{
5724 struct ext4_inode_info *ei = EXT4_I(inode);
5725
5726 /* The first argument of lockdep_set_subclass has to be
5727 * *exactly* the same as the argument to init_rwsem() --- in
5728 * this case, in init_once() --- or lockdep gets unhappy
5729 * because the name of the lock is set using the
5730 * stringification of the argument to init_rwsem().
5731 */
5732 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5733 lockdep_set_subclass(&ei->i_data_sem, subclass);
5734}
5735
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005736/*
5737 * Standard function to be called on quota_on
5738 */
Mingming Cao617ba132006-10-11 01:20:53 -07005739static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Al Viro8c54ca92016-11-20 19:49:34 -05005740 const struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005741{
5742 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005743
5744 if (!test_opt(sb, QUOTA))
5745 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04005746
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005747 /* Quotafile not on the same filesystem? */
Al Virod8c95842011-12-07 18:16:57 -05005748 if (path->dentry->d_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005749 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04005750 /* Journaling quota? */
5751 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005752 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02005753 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04005754 ext4_msg(sb, KERN_WARNING,
5755 "Quota file not on filesystem root. "
5756 "Journaled quota will not work");
Jan Kara91389242017-08-03 11:25:55 +02005757 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5758 } else {
5759 /*
5760 * Clear the flag just in case mount options changed since
5761 * last time.
5762 */
5763 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005764 }
Jan Kara06235432008-05-13 19:11:51 -04005765
5766 /*
5767 * When we journal data on quota file, we have to flush journal to see
5768 * all updates to the file when we bypass pagecache...
5769 */
Frank Mayhar03901312009-01-07 00:06:22 -05005770 if (EXT4_SB(sb)->s_journal &&
David Howells2b0143b2015-03-17 22:25:59 +00005771 ext4_should_journal_data(d_inode(path->dentry))) {
Jan Kara06235432008-05-13 19:11:51 -04005772 /*
5773 * We don't need to lock updates but journal_flush() could
5774 * otherwise be livelocked...
5775 */
5776 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005777 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04005778 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02005779 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005780 return err;
Jan Kara06235432008-05-13 19:11:51 -04005781 }
Jan Kara957153f2017-04-06 15:40:06 +02005782
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005783 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5784 err = dquot_quota_on(sb, type, format_id, path);
Jan Kara957153f2017-04-06 15:40:06 +02005785 if (err) {
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005786 lockdep_set_quota_inode(path->dentry->d_inode,
5787 I_DATA_SEM_NORMAL);
Jan Kara957153f2017-04-06 15:40:06 +02005788 } else {
5789 struct inode *inode = d_inode(path->dentry);
5790 handle_t *handle;
5791
Jan Kara61a92982017-04-24 16:49:16 +02005792 /*
5793 * Set inode flags to prevent userspace from messing with quota
5794 * files. If this fails, we return success anyway since quotas
5795 * are already enabled and this is not a hard failure.
5796 */
Jan Kara957153f2017-04-06 15:40:06 +02005797 inode_lock(inode);
5798 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5799 if (IS_ERR(handle))
5800 goto unlock_inode;
5801 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5802 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5803 S_NOATIME | S_IMMUTABLE);
5804 ext4_mark_inode_dirty(handle, inode);
5805 ext4_journal_stop(handle);
5806 unlock_inode:
5807 inode_unlock(inode);
5808 }
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005809 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005810}
5811
Aditya Kali7c319d32012-07-22 20:21:31 -04005812static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5813 unsigned int flags)
5814{
5815 int err;
5816 struct inode *qf_inode;
Jan Karaa2d4a642014-09-11 11:15:15 -04005817 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
Aditya Kali7c319d32012-07-22 20:21:31 -04005818 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
Li Xi689c958c2016-01-08 16:01:22 -05005819 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5820 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
Aditya Kali7c319d32012-07-22 20:21:31 -04005821 };
5822
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005823 BUG_ON(!ext4_has_feature_quota(sb));
Aditya Kali7c319d32012-07-22 20:21:31 -04005824
5825 if (!qf_inums[type])
5826 return -EPERM;
5827
Theodore Ts'o8a363972018-12-19 12:29:13 -05005828 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
Aditya Kali7c319d32012-07-22 20:21:31 -04005829 if (IS_ERR(qf_inode)) {
5830 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5831 return PTR_ERR(qf_inode);
5832 }
5833
Jan Karabcb13852013-04-09 09:21:41 -04005834 /* Don't account quota for quota files to avoid recursion */
5835 qf_inode->i_flags |= S_NOQUOTA;
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005836 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
Jan Kara7212b952019-11-01 18:55:38 +01005837 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005838 if (err)
5839 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
Pan Bian61157b22018-12-03 23:28:02 -05005840 iput(qf_inode);
Aditya Kali7c319d32012-07-22 20:21:31 -04005841
5842 return err;
5843}
5844
5845/* Enable usage tracking for all quota types. */
5846static int ext4_enable_quotas(struct super_block *sb)
5847{
5848 int type, err = 0;
Jan Karaa2d4a642014-09-11 11:15:15 -04005849 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
Aditya Kali7c319d32012-07-22 20:21:31 -04005850 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
Li Xi689c958c2016-01-08 16:01:22 -05005851 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5852 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
Aditya Kali7c319d32012-07-22 20:21:31 -04005853 };
Jan Kara49da9392016-09-05 23:08:16 -04005854 bool quota_mopt[EXT4_MAXQUOTAS] = {
5855 test_opt(sb, USRQUOTA),
5856 test_opt(sb, GRPQUOTA),
5857 test_opt(sb, PRJQUOTA),
5858 };
Aditya Kali7c319d32012-07-22 20:21:31 -04005859
Jan Kara91389242017-08-03 11:25:55 +02005860 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
Jan Karaa2d4a642014-09-11 11:15:15 -04005861 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
Aditya Kali7c319d32012-07-22 20:21:31 -04005862 if (qf_inums[type]) {
5863 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
Jan Kara49da9392016-09-05 23:08:16 -04005864 DQUOT_USAGE_ENABLED |
5865 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
Aditya Kali7c319d32012-07-22 20:21:31 -04005866 if (err) {
5867 ext4_warning(sb,
Theodore Ts'o72ba7452013-01-24 23:24:54 -05005868 "Failed to enable quota tracking "
5869 "(type=%d, err=%d). Please run "
5870 "e2fsck to fix.", type, err);
Junichi Uekawa7f144fd2018-07-29 15:51:52 -04005871 for (type--; type >= 0; type--)
5872 dquot_quota_off(sb, type);
5873
Aditya Kali7c319d32012-07-22 20:21:31 -04005874 return err;
5875 }
5876 }
5877 }
5878 return 0;
5879}
5880
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005881static int ext4_quota_off(struct super_block *sb, int type)
5882{
Jan Kara21f97692011-04-04 15:33:39 -04005883 struct inode *inode = sb_dqopt(sb)->files[type];
5884 handle_t *handle;
Jan Kara957153f2017-04-06 15:40:06 +02005885 int err;
Jan Kara21f97692011-04-04 15:33:39 -04005886
Dmitry Monakhov87009d82010-11-08 13:47:33 -05005887 /* Force all delayed allocation blocks to be allocated.
5888 * Caller already holds s_umount sem */
5889 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005890 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005891
Jan Kara957153f2017-04-06 15:40:06 +02005892 if (!inode || !igrab(inode))
Amir Goldstein0b268592011-05-16 09:59:13 -04005893 goto out;
5894
Jan Kara957153f2017-04-06 15:40:06 +02005895 err = dquot_quota_off(sb, type);
Jan Kara964edf62017-05-21 22:31:23 -04005896 if (err || ext4_has_feature_quota(sb))
Jan Kara957153f2017-04-06 15:40:06 +02005897 goto out_put;
5898
5899 inode_lock(inode);
Jan Kara61a92982017-04-24 16:49:16 +02005900 /*
5901 * Update modification times of quota files when userspace can
5902 * start looking at them. If we fail, we return success anyway since
5903 * this is not a hard failure and quotas are already disabled.
5904 */
Theodore Ts'o9924a922013-02-08 21:59:22 -05005905 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
Jan Kara21f97692011-04-04 15:33:39 -04005906 if (IS_ERR(handle))
Jan Kara957153f2017-04-06 15:40:06 +02005907 goto out_unlock;
5908 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5909 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
Deepa Dinamanieeca7ea2016-11-14 21:40:10 -05005910 inode->i_mtime = inode->i_ctime = current_time(inode);
Jan Kara21f97692011-04-04 15:33:39 -04005911 ext4_mark_inode_dirty(handle, inode);
5912 ext4_journal_stop(handle);
Jan Kara957153f2017-04-06 15:40:06 +02005913out_unlock:
5914 inode_unlock(inode);
5915out_put:
Jan Kara964edf62017-05-21 22:31:23 -04005916 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
Jan Kara957153f2017-04-06 15:40:06 +02005917 iput(inode);
5918 return err;
Jan Kara21f97692011-04-04 15:33:39 -04005919out:
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005920 return dquot_quota_off(sb, type);
5921}
5922
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005923/* Read data from quotafile - avoid pagecache and such because we cannot afford
5924 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03005925 * itself serializes the operations (and no one else should touch the files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005926 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07005927static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005928 size_t len, loff_t off)
5929{
5930 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005931 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005932 int offset = off & (sb->s_blocksize - 1);
5933 int tocopy;
5934 size_t toread;
5935 struct buffer_head *bh;
5936 loff_t i_size = i_size_read(inode);
5937
5938 if (off > i_size)
5939 return 0;
5940 if (off+len > i_size)
5941 len = i_size-off;
5942 toread = len;
5943 while (toread > 0) {
5944 tocopy = sb->s_blocksize - offset < toread ?
5945 sb->s_blocksize - offset : toread;
Theodore Ts'o1c215022014-08-29 20:52:15 -04005946 bh = ext4_bread(NULL, inode, blk, 0);
5947 if (IS_ERR(bh))
5948 return PTR_ERR(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005949 if (!bh) /* A hole? */
5950 memset(data, 0, tocopy);
5951 else
5952 memcpy(data, bh->b_data+offset, tocopy);
5953 brelse(bh);
5954 offset = 0;
5955 toread -= tocopy;
5956 data += tocopy;
5957 blk++;
5958 }
5959 return len;
5960}
5961
5962/* Write to quotafile (we know the transaction is already started and has
5963 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07005964static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005965 const char *data, size_t len, loff_t off)
5966{
5967 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005968 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Theodore Ts'o1c215022014-08-29 20:52:15 -04005969 int err, offset = off & (sb->s_blocksize - 1);
Theodore Ts'oc5e298a2015-06-21 01:25:29 -04005970 int retries = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005971 struct buffer_head *bh;
5972 handle_t *handle = journal_current_handle();
5973
Frank Mayhar03901312009-01-07 00:06:22 -05005974 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005975 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5976 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07005977 (unsigned long long)off, (unsigned long long)len);
5978 return -EIO;
5979 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005980 /*
5981 * Since we account only one data block in transaction credits,
5982 * then it is impossible to cross a block boundary.
5983 */
5984 if (sb->s_blocksize - offset < len) {
5985 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5986 " cancelled because not block aligned",
5987 (unsigned long long)off, (unsigned long long)len);
5988 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005989 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005990
Theodore Ts'oc5e298a2015-06-21 01:25:29 -04005991 do {
5992 bh = ext4_bread(handle, inode, blk,
5993 EXT4_GET_BLOCKS_CREATE |
5994 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5995 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5996 ext4_should_retry_alloc(inode->i_sb, &retries));
Theodore Ts'o1c215022014-08-29 20:52:15 -04005997 if (IS_ERR(bh))
5998 return PTR_ERR(bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005999 if (!bh)
6000 goto out;
liang xie5d601252014-05-12 22:06:43 -04006001 BUFFER_TRACE(bh, "get write access");
Jan Kara62d2b5f2010-07-27 11:56:07 -04006002 err = ext4_journal_get_write_access(handle, bh);
6003 if (err) {
6004 brelse(bh);
Theodore Ts'o1c215022014-08-29 20:52:15 -04006005 return err;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05006006 }
6007 lock_buffer(bh);
6008 memcpy(bh->b_data+offset, data, len);
6009 flush_dcache_page(bh->b_page);
6010 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04006011 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05006012 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006013out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05006014 if (inode->i_size < off + len) {
6015 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07006016 EXT4_I(inode)->i_disksize = inode->i_size;
Jan Kara21f97692011-04-04 15:33:39 -04006017 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006018 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05006019 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006020}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006021#endif
6022
Al Viro152a0832010-07-25 00:46:55 +04006023static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6024 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006025{
Al Viro152a0832010-07-25 00:46:55 +04006026 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006027}
6028
Jan Karac290ea02015-06-18 16:52:29 +02006029#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
Theodore Ts'o24b58422009-12-07 14:08:51 -05006030static inline void register_as_ext2(void)
6031{
6032 int err = register_filesystem(&ext2_fs_type);
6033 if (err)
6034 printk(KERN_WARNING
6035 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6036}
6037
6038static inline void unregister_as_ext2(void)
6039{
6040 unregister_filesystem(&ext2_fs_type);
6041}
Theodore Ts'o2035e772011-04-18 17:29:14 -04006042
6043static inline int ext2_feature_set_ok(struct super_block *sb)
6044{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04006045 if (ext4_has_unknown_ext2_incompat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006046 return 0;
David Howellsbc98a422017-07-17 08:45:34 +01006047 if (sb_rdonly(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006048 return 1;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04006049 if (ext4_has_unknown_ext2_ro_compat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006050 return 0;
6051 return 1;
6052}
Theodore Ts'o24b58422009-12-07 14:08:51 -05006053#else
6054static inline void register_as_ext2(void) { }
6055static inline void unregister_as_ext2(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04006056static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05006057#endif
6058
Theodore Ts'o24b58422009-12-07 14:08:51 -05006059static inline void register_as_ext3(void)
6060{
6061 int err = register_filesystem(&ext3_fs_type);
6062 if (err)
6063 printk(KERN_WARNING
6064 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6065}
6066
6067static inline void unregister_as_ext3(void)
6068{
6069 unregister_filesystem(&ext3_fs_type);
6070}
Theodore Ts'o2035e772011-04-18 17:29:14 -04006071
6072static inline int ext3_feature_set_ok(struct super_block *sb)
6073{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04006074 if (ext4_has_unknown_ext3_incompat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006075 return 0;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04006076 if (!ext4_has_feature_journal(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006077 return 0;
David Howellsbc98a422017-07-17 08:45:34 +01006078 if (sb_rdonly(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006079 return 1;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04006080 if (ext4_has_unknown_ext3_ro_compat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04006081 return 0;
6082 return 1;
6083}
Theodore Ts'o24b58422009-12-07 14:08:51 -05006084
Theodore Ts'o03010a32008-10-10 20:02:48 -04006085static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006086 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04006087 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04006088 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006089 .kill_sb = kill_block_super,
6090 .fs_flags = FS_REQUIRES_DEV,
6091};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08006092MODULE_ALIAS_FS("ext4");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006093
Eric Sandeene9e3bce2011-02-12 08:17:34 -05006094/* Shared across all ext4 file systems */
6095wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
Eric Sandeene9e3bce2011-02-12 08:17:34 -05006096
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006097static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006098{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05006099 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05006100
Theodore Ts'oe294a532015-08-15 14:59:44 -04006101 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
Al Viro07c0c5d82012-03-20 22:05:02 -04006102 ext4_li_info = NULL;
6103 mutex_init(&ext4_li_mtx);
6104
Carlos Maiolino9a4c8012012-12-10 16:30:45 -05006105 /* Build-time check for flags consistency */
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04006106 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05006107
Jan Karae142d052016-03-08 22:44:50 -05006108 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
Eric Sandeene9e3bce2011-02-12 08:17:34 -05006109 init_waitqueue_head(&ext4__ioend_wq[i]);
Eric Sandeene9e3bce2011-02-12 08:17:34 -05006110
Zheng Liu51865fd2012-11-08 21:57:32 -05006111 err = ext4_init_es();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04006112 if (err)
6113 return err;
Zheng Liu51865fd2012-11-08 21:57:32 -05006114
Eric Whitney1dc0aa42018-10-01 14:17:41 -04006115 err = ext4_init_pending();
6116 if (err)
Eric Biggers22cfe4b2019-07-22 09:26:24 -07006117 goto out7;
6118
6119 err = ext4_init_post_read_processing();
6120 if (err)
Eric Whitney1dc0aa42018-10-01 14:17:41 -04006121 goto out6;
6122
Zheng Liu51865fd2012-11-08 21:57:32 -05006123 err = ext4_init_pageio();
6124 if (err)
Theodore Ts'ob5799012015-09-23 12:44:17 -04006125 goto out5;
Zheng Liu51865fd2012-11-08 21:57:32 -05006126
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006127 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04006128 if (err)
Theodore Ts'ob5799012015-09-23 12:44:17 -04006129 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04006130
Theodore Ts'ob5799012015-09-23 12:44:17 -04006131 err = ext4_init_sysfs();
Theodore Ts'odd683142011-02-03 14:33:49 -05006132 if (err)
Theodore Ts'ob5799012015-09-23 12:44:17 -04006133 goto out3;
Lukas Czerner857ac882010-10-27 21:30:05 -04006134
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006135 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006136 if (err)
Alex Tomasc9de5602008-01-29 00:19:52 -05006137 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006138 err = init_inodecache();
6139 if (err)
6140 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05006141 register_as_ext3();
Theodore Ts'o2035e772011-04-18 17:29:14 -04006142 register_as_ext2();
Theodore Ts'o03010a32008-10-10 20:02:48 -04006143 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006144 if (err)
6145 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04006146
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006147 return 0;
6148out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05006149 unregister_as_ext2();
6150 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006151 destroy_inodecache();
6152out1:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006153 ext4_exit_mballoc();
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04006154out2:
Theodore Ts'ob5799012015-09-23 12:44:17 -04006155 ext4_exit_sysfs();
6156out3:
Theodore Ts'odd683142011-02-03 14:33:49 -05006157 ext4_exit_system_zone();
Theodore Ts'ob5799012015-09-23 12:44:17 -04006158out4:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006159 ext4_exit_pageio();
Theodore Ts'ob5799012015-09-23 12:44:17 -04006160out5:
Eric Biggers22cfe4b2019-07-22 09:26:24 -07006161 ext4_exit_post_read_processing();
Eric Whitney1dc0aa42018-10-01 14:17:41 -04006162out6:
Eric Biggers22cfe4b2019-07-22 09:26:24 -07006163 ext4_exit_pending();
6164out7:
Zheng Liu51865fd2012-11-08 21:57:32 -05006165 ext4_exit_es();
6166
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006167 return err;
6168}
6169
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006170static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006171{
Lukas Czernerbfff6872010-10-27 21:30:05 -04006172 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05006173 unregister_as_ext2();
6174 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04006175 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006176 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006177 ext4_exit_mballoc();
Theodore Ts'ob5799012015-09-23 12:44:17 -04006178 ext4_exit_sysfs();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006179 ext4_exit_system_zone();
6180 ext4_exit_pageio();
Eric Biggers22cfe4b2019-07-22 09:26:24 -07006181 ext4_exit_post_read_processing();
Eric Sandeendd12ed142013-07-26 15:21:11 -04006182 ext4_exit_es();
Eric Whitney1dc0aa42018-10-01 14:17:41 -04006183 ext4_exit_pending();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006184}
6185
6186MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05006187MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07006188MODULE_LICENSE("GPL");
Theodore Ts'o7ef79ad2018-04-26 00:44:46 -04006189MODULE_SOFTDEP("pre: crc32c");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04006190module_init(ext4_init_fs)
6191module_exit(ext4_exit_fs)