blob: 52b0530c5d65a95fa0512b29cb2b3645d0277b69 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/super.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
Theodore Ts'oc5ca7c72009-04-27 22:48:48 -040023#include <linux/vmalloc.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070024#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
Tejun Heo66114ca2015-05-22 17:13:32 -040027#include <linux/backing-dev.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070028#include <linux/parser.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070030#include <linux/exportfs.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
Theodore Ts'o3197ebd2009-03-31 09:10:09 -040037#include <linux/ctype.h>
Vignesh Babu13305932007-07-18 09:11:02 -040038#include <linux/log2.h>
Andreas Dilger717d50e2007-10-16 18:38:25 -040039#include <linux/crc16.h>
Dan Magenheimer7abc52c2011-05-26 10:02:03 -060040#include <linux/cleancache.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070041#include <asm/uaccess.h>
42
Lukas Czernerbfff6872010-10-27 21:30:05 -040043#include <linux/kthread.h>
44#include <linux/freezer.h>
45
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040046#include "ext4.h"
Theodore Ts'o4a092d72012-11-28 13:03:30 -050047#include "ext4_extents.h" /* Needed for trace points definition */
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040048#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070049#include "xattr.h"
50#include "acl.h"
Theodore Ts'o3661d282009-09-14 22:59:50 -040051#include "mballoc.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070052
Theodore Ts'o9bffad12009-06-17 11:48:11 -040053#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
Lukas Czerner0b75a842011-02-23 12:22:49 -050056static struct ext4_lazy_init *ext4_li_info;
57static struct mutex ext4_li_mtx;
Theodore Ts'oe294a532015-08-15 14:59:44 -040058static struct ratelimit_state ext4_mount_msg_ratelimit;
Theodore Ts'o9f6200b2008-09-23 09:18:24 -040059
Mingming Cao617ba132006-10-11 01:20:53 -070060static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070061 unsigned long journal_devnum);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -050062static int ext4_show_options(struct seq_file *seq, struct dentry *root);
Theodore Ts'oe2d67052009-05-01 00:33:44 -040063static int ext4_commit_super(struct super_block *sb, int sync);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040064static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
Mingming Cao617ba132006-10-11 01:20:53 -070068static int ext4_sync_fs(struct super_block *sb, int wait);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -040069static int ext4_remount(struct super_block *sb, int *flags, char *data);
70static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
Takashi Satoc4be0c12009-01-09 16:40:58 -080071static int ext4_unfreeze(struct super_block *sb);
Takashi Satoc4be0c12009-01-09 16:40:58 -080072static int ext4_freeze(struct super_block *sb);
Al Viro152a0832010-07-25 00:46:55 +040073static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
Theodore Ts'o2035e772011-04-18 17:29:14 -040075static inline int ext2_feature_set_ok(struct super_block *sb);
76static inline int ext3_feature_set_ok(struct super_block *sb);
Amir Goldsteind39195c2011-02-28 00:53:45 -050077static int ext4_feature_set_ok(struct super_block *sb, int readonly);
Lukas Czernerbfff6872010-10-27 21:30:05 -040078static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
Eric Sandeen8f1f7452011-02-03 14:33:15 -050080static void ext4_clear_request_list(void);
Eric Whitneyc6cb7e72016-09-30 02:05:09 -040081static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070083
Jan Karae74031f2015-12-07 14:35:49 -050084/*
85 * Lock ordering
86 *
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 *
90 * page fault path:
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
93 *
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
97 * i_data_sem (rw)
98 *
99 * truncate:
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
104 *
105 * direct IO:
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
109 *
110 * writepages:
111 * transaction start -> page lock(s) -> i_data_sem (rw)
112 */
113
Jan Karac290ea02015-06-18 16:52:29 +0200114#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
Theodore Ts'o2035e772011-04-18 17:29:14 -0400115static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
117 .name = "ext2",
118 .mount = ext4_mount,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
121};
Eric W. Biederman7f78e032013-03-02 19:39:14 -0800122MODULE_ALIAS_FS("ext2");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -0700123MODULE_ALIAS("ext2");
Theodore Ts'o2035e772011-04-18 17:29:14 -0400124#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125#else
126#define IS_EXT2_SB(sb) (0)
127#endif
128
129
Jan Karaba69f9a2010-03-24 20:18:37 -0400130static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
132 .name = "ext3",
Al Viro152a0832010-07-25 00:46:55 +0400133 .mount = ext4_mount,
Jan Karaba69f9a2010-03-24 20:18:37 -0400134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
136};
Eric W. Biederman7f78e032013-03-02 19:39:14 -0800137MODULE_ALIAS_FS("ext3");
Eric W. Biedermanfa7614dd2013-03-12 18:27:41 -0700138MODULE_ALIAS("ext3");
Jan Karaba69f9a2010-03-24 20:18:37 -0400139#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700140
Darrick J. Wongd25425f2012-04-29 18:25:10 -0400141static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
143{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400144 if (!ext4_has_feature_metadata_csum(sb))
Darrick J. Wongd25425f2012-04-29 18:25:10 -0400145 return 1;
146
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148}
149
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400150static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
152{
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
155 __u32 csum;
156
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158
159 return cpu_to_le32(csum);
160}
161
Stephen Hemmingerc1978552014-05-12 10:50:23 -0400162static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400164{
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -0400165 if (!ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400166 return 1;
167
168 return es->s_checksum == ext4_superblock_csum(sb, es);
169}
170
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400171void ext4_superblock_csum_set(struct super_block *sb)
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400172{
Theodore Ts'o06db49e2012-10-10 01:06:58 -0400173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -0400175 if (!ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -0400176 return;
177
178 es->s_checksum = ext4_superblock_csum(sb, es);
179}
180
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400181void *ext4_kvmalloc(size_t size, gfp_t flags)
182{
183 void *ret;
184
Joe Perches8be04b92013-06-19 12:15:53 -0700185 ret = kmalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400186 if (!ret)
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
188 return ret;
189}
190
191void *ext4_kvzalloc(size_t size, gfp_t flags)
192{
193 void *ret;
194
Joe Perches8be04b92013-06-19 12:15:53 -0700195 ret = kzalloc(size, flags | __GFP_NOWARN);
Theodore Ts'o9933fc02011-08-01 08:45:02 -0400196 if (!ret)
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198 return ret;
199}
200
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700201ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700203{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700207}
208
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700209ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700211{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700215}
216
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700217ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700219{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400220 return le32_to_cpu(bg->bg_inode_table_lo) |
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700223}
224
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400225__u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500227{
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500231}
232
233__u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235{
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500239}
240
241__u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
243{
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500247}
248
249__u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
251{
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500255}
256
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700257void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700259{
Aneesh Kumar K.V3a145892007-10-16 18:38:25 -0400260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700263}
264
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700265void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700267{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700271}
272
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700273void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700275{
Aneesh Kumar K.V5272f832007-10-16 18:38:25 -0400276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
Alexandre Ratchov8fadc142006-10-11 01:21:15 -0700277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
Laurent Vivierbd81d8e2006-10-11 01:21:10 -0700279}
280
Theodore Ts'o021b65b2011-09-09 19:08:51 -0400281void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
Aneesh Kumar K.V560671a2009-01-05 22:20:24 -0500283{
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287}
288
289void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291{
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295}
296
297void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
299{
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303}
304
305void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
307{
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
311}
312
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400313
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400314static void __save_error_info(struct super_block *sb, const char *func,
315 unsigned int line)
316{
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
Theodore Ts'o1b466172015-05-14 18:37:30 -0400320 if (bdev_read_only(sb->s_bdev))
321 return;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
333 }
Theodore Ts'o66e61a92010-07-27 11:56:04 -0400334 /*
335 * Start the daily error reporting function if it hasn't been
336 * started already
337 */
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
Wei Yongjunba39ebb2012-09-27 09:37:53 -0400340 le32_add_cpu(&es->s_error_count, 1);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400341}
342
343static void save_error_info(struct super_block *sb, const char *func,
344 unsigned int line)
345{
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
348}
349
Theodore Ts'obdfe0cbd2015-08-16 10:03:57 -0400350/*
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
357 */
358static int block_device_ejected(struct super_block *sb)
359{
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362
363 return bdi->dev == NULL;
364}
365
Bobi Jam18aadd42012-02-20 17:53:02 -0500366static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367{
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400371 struct ext4_journal_cb_entry *jce;
Bobi Jam18aadd42012-02-20 17:53:02 -0500372
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400373 BUG_ON(txn->t_state == T_FINISHED);
Bobi Jam18aadd42012-02-20 17:53:02 -0500374 spin_lock(&sbi->s_md_lock);
Dmitry Monakhov5d3ee202013-04-03 22:08:52 -0400375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
Bobi Jam18aadd42012-02-20 17:53:02 -0500378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
382 }
383 spin_unlock(&sbi->s_md_lock);
384}
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400385
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700386/* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
388 *
389 * On ext2, we can store the error state of the filesystem in the
Mingming Cao617ba132006-10-11 01:20:53 -0700390 * superblock. That is not possible on ext4, because we may have other
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
395 *
Mingming Caodab291a2006-10-11 01:21:01 -0700396 * 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 -0200397 * the journal instead. On recovery, the journal will complain about
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700398 * that error until we've noted it down and cleared it.
399 */
400
Mingming Cao617ba132006-10-11 01:20:53 -0700401static void ext4_handle_error(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700402{
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700403 if (sb->s_flags & MS_RDONLY)
404 return;
405
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400406 if (!test_opt(sb, ERRORS_CONT)) {
Mingming Cao617ba132006-10-11 01:20:53 -0700407 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700408
Theodore Ts'o4ab2f152009-06-13 10:09:36 -0400409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700410 if (journal)
Mingming Caodab291a2006-10-11 01:21:01 -0700411 jbd2_journal_abort(journal, -EIO);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700412 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400413 if (test_opt(sb, ERRORS_RO)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400415 /*
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
418 */
419 smp_wmb();
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700420 sb->s_flags |= MS_RDONLY;
421 }
Daeho Jeong4327ba52015-10-18 17:02:56 -0400422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 return;
Mingming Cao617ba132006-10-11 01:20:53 -0700426 panic("EXT4-fs (device %s): panic forced after error\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700427 sb->s_id);
Daeho Jeong4327ba52015-10-18 17:02:56 -0400428 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700429}
430
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400431#define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 "EXT4-fs error")
434
Eric Sandeen12062dd2010-02-15 14:19:27 -0500435void __ext4_error(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400436 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700437{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500438 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700439 va_list args;
440
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400441 if (ext4_error_ratelimit(sb)) {
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 printk(KERN_CRIT
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
448 va_end(args);
449 }
Theodore Ts'of3fc02102012-05-30 23:00:16 -0400450 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700451 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700452}
453
Joe Perchese7c96e82013-07-01 08:12:37 -0400454void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500457{
458 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500459 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
Frank Mayhar273df552010-03-02 11:46:09 -0500461
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400464 if (ext4_error_ratelimit(inode->i_sb)) {
465 va_start(args, fmt);
466 vaf.fmt = fmt;
467 vaf.va = &args;
468 if (block)
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
473 else
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
478 va_end(args);
479 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400480 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500481 ext4_handle_error(inode->i_sb);
482}
483
Joe Perchese7c96e82013-07-01 08:12:37 -0400484void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
Frank Mayhar273df552010-03-02 11:46:09 -0500487{
488 va_list args;
Theodore Ts'of7c21172011-01-10 12:10:55 -0500489 struct va_format vaf;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400490 struct ext4_super_block *es;
Al Viro496ad9a2013-01-23 17:07:38 -0500491 struct inode *inode = file_inode(file);
Frank Mayhar273df552010-03-02 11:46:09 -0500492 char pathname[80], *path;
493
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400496 if (ext4_error_ratelimit(inode->i_sb)) {
Miklos Szeredi9bf39ab2015-06-19 10:29:13 +0200497 path = file_path(file, pathname, sizeof(pathname));
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400498 if (IS_ERR(path))
499 path = "(unknown)";
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
509 else
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
515 va_end(args);
516 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400517 save_error_info(inode->i_sb, function, line);
Frank Mayhar273df552010-03-02 11:46:09 -0500518 ext4_handle_error(inode->i_sb);
519}
520
Theodore Ts'o722887d2013-02-08 13:00:31 -0500521const char *ext4_decode_error(struct super_block *sb, int errno,
522 char nbuf[16])
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700523{
524 char *errstr = NULL;
525
526 switch (errno) {
Darrick J. Wong6a797d22015-10-17 16:16:04 -0400527 case -EFSCORRUPTED:
528 errstr = "Corrupt filesystem";
529 break;
530 case -EFSBADCRC:
531 errstr = "Filesystem failed CRC";
532 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700533 case -EIO:
534 errstr = "IO failure";
535 break;
536 case -ENOMEM:
537 errstr = "Out of memory";
538 break;
539 case -EROFS:
Theodore Ts'o78f1ddb2009-07-27 23:09:47 -0400540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700542 errstr = "Journal has aborted";
543 else
544 errstr = "Readonly filesystem";
545 break;
546 default:
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
549 * NULL. */
550 if (nbuf) {
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
553 errstr = nbuf;
554 }
555 break;
556 }
557
558 return errstr;
559}
560
Mingming Cao617ba132006-10-11 01:20:53 -0700561/* __ext4_std_error decodes expected errors from journaling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700562 * automatically and invokes the appropriate error response. */
563
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400564void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700566{
567 char nbuf[16];
568 const char *errstr;
569
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
572 * an error. */
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
575 return;
576
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
581 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700582
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400583 save_error_info(sb, function, line);
Mingming Cao617ba132006-10-11 01:20:53 -0700584 ext4_handle_error(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700585}
586
587/*
Mingming Cao617ba132006-10-11 01:20:53 -0700588 * ext4_abort is a much stronger failure handler than ext4_error. The
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
591 *
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
595 */
596
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400597void __ext4_abort(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400598 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700599{
Joe Perches651e1c32016-10-12 23:12:53 -0400600 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700601 va_list args;
602
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400603 save_error_info(sb, function, line);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700604 va_start(args, fmt);
Joe Perches651e1c32016-10-12 23:12:53 -0400605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700609 va_end(args);
610
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
Dmitry Monakhov4418e1412013-06-12 22:38:04 -0400614 /*
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
617 */
618 smp_wmb();
619 sb->s_flags |= MS_RDONLY;
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
623 }
Daeho Jeong4327ba52015-10-18 17:02:56 -0400624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
627 return;
Mingming Cao617ba132006-10-11 01:20:53 -0700628 panic("EXT4-fs panic from previous error\n");
Daeho Jeong4327ba52015-10-18 17:02:56 -0400629 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700630}
631
Joe Perchese7c96e82013-07-01 08:12:37 -0400632void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
Eric Sandeenb31e1552009-06-04 17:36:36 -0400634{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500635 struct va_format vaf;
Eric Sandeenb31e1552009-06-04 17:36:36 -0400636 va_list args;
637
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
639 return;
640
Eric Sandeenb31e1552009-06-04 17:36:36 -0400641 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500642 vaf.fmt = fmt;
643 vaf.va = &args;
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
Eric Sandeenb31e1552009-06-04 17:36:36 -0400645 va_end(args);
646}
647
Andreas Dilgerb03a2f72015-06-15 14:50:26 -0400648#define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
650 "EXT4-fs warning")
651
Eric Sandeen12062dd2010-02-15 14:19:27 -0500652void __ext4_warning(struct super_block *sb, const char *function,
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400653 unsigned int line, const char *fmt, ...)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700654{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500655 struct va_format vaf;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700656 va_list args;
657
Andreas Dilgerb03a2f72015-06-15 14:50:26 -0400658 if (!ext4_warning_ratelimit(sb))
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400659 return;
660
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700661 va_start(args, fmt);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700666 va_end(args);
667}
668
Andreas Dilgerb03a2f72015-06-15 14:50:26 -0400669void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
671{
672 struct va_format vaf;
673 va_list args;
674
675 if (!ext4_warning_ratelimit(inode->i_sb))
676 return;
677
678 va_start(args, fmt);
679 vaf.fmt = fmt;
680 vaf.va = &args;
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
684 va_end(args);
685}
686
Theodore Ts'oe29136f2010-06-29 12:54:28 -0400687void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500691__releases(bitlock)
692__acquires(bitlock)
693{
Joe Perches0ff2ea72010-12-19 22:43:19 -0500694 struct va_format vaf;
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500695 va_list args;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
697
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
Joe Perches0ff2ea72010-12-19 22:43:19 -0500701
Theodore Ts'oefbed4d2013-10-17 21:11:01 -0400702 if (ext4_error_ratelimit(sb)) {
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
708 if (ino)
709 printk(KERN_CONT "inode %lu: ", ino);
710 if (block)
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
714 va_end(args);
715 }
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500716
717 if (test_opt(sb, ERRORS_CONT)) {
Theodore Ts'oe2d67052009-05-01 00:33:44 -0400718 ext4_commit_super(sb, 0);
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500719 return;
720 }
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -0400721
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500722 ext4_unlock_group(sb, grp);
723 ext4_handle_error(sb);
724 /*
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300730 * ext4_grp_locked_error() to distinguish between the
Aneesh Kumar K.V5d1b1b32009-01-05 22:19:52 -0500731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
734 */
735 ext4_lock_group(sb, grp);
736 return;
737}
738
Mingming Cao617ba132006-10-11 01:20:53 -0700739void ext4_update_dynamic_rev(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700740{
Mingming Cao617ba132006-10-11 01:20:53 -0700741 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700742
Mingming Cao617ba132006-10-11 01:20:53 -0700743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700744 return;
745
Eric Sandeen12062dd2010-02-15 14:19:27 -0500746 ext4_warning(sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
Mingming Cao617ba132006-10-11 01:20:53 -0700749 EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700750
Mingming Cao617ba132006-10-11 01:20:53 -0700751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
756
757 /*
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
761 */
762}
763
764/*
765 * Open the external journal device
766 */
Eric Sandeenb31e1552009-06-04 17:36:36 -0400767static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700768{
769 struct block_device *bdev;
770 char b[BDEVNAME_SIZE];
771
Tejun Heod4d77622010-11-13 11:55:18 +0100772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700773 if (IS_ERR(bdev))
774 goto fail;
775 return bdev;
776
777fail:
Eric Sandeenb31e1552009-06-04 17:36:36 -0400778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700779 __bdevname(dev, b), PTR_ERR(bdev));
780 return NULL;
781}
782
783/*
784 * Release the journal device
785 */
Al Viro4385bab2013-05-05 22:11:03 -0400786static void ext4_blkdev_put(struct block_device *bdev)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700787{
Al Viro4385bab2013-05-05 22:11:03 -0400788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700789}
790
Al Viro4385bab2013-05-05 22:11:03 -0400791static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700792{
793 struct block_device *bdev;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700794 bdev = sbi->journal_bdev;
795 if (bdev) {
Al Viro4385bab2013-05-05 22:11:03 -0400796 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700797 sbi->journal_bdev = NULL;
798 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700799}
800
801static inline struct inode *orphan_list_entry(struct list_head *l)
802{
Mingming Cao617ba132006-10-11 01:20:53 -0700803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700804}
805
Mingming Cao617ba132006-10-11 01:20:53 -0700806static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700807{
808 struct list_head *l;
809
Eric Sandeenb31e1552009-06-04 17:36:36 -0400810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
811 le32_to_cpu(sbi->s_es->s_last_orphan));
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700812
813 printk(KERN_ERR "sb_info orphan list:\n");
814 list_for_each(l, &sbi->s_orphan) {
815 struct inode *inode = orphan_list_entry(l);
816 printk(KERN_ERR " "
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode->i_sb->s_id, inode->i_ino, inode,
819 inode->i_mode, inode->i_nlink,
820 NEXT_ORPHAN(inode));
821 }
822}
823
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -0400824static void ext4_put_super(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700825{
Mingming Cao617ba132006-10-11 01:20:53 -0700826 struct ext4_sb_info *sbi = EXT4_SB(sb);
827 struct ext4_super_block *es = sbi->s_es;
Hidehiro Kawaief2cabf2008-10-27 22:53:05 -0400828 int i, err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700829
Lukas Czerner857ac882010-10-27 21:30:05 -0400830 ext4_unregister_li_request(sb);
Christoph Hellwige0ccfd92010-05-19 07:16:42 -0400831 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
832
Jan Kara2e8fa542013-06-04 14:21:02 -0400833 flush_workqueue(sbi->rsv_conversion_wq);
Jan Kara2e8fa542013-06-04 14:21:02 -0400834 destroy_workqueue(sbi->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -0400835
Frank Mayhar03901312009-01-07 00:06:22 -0500836 if (sbi->s_journal) {
837 err = jbd2_journal_destroy(sbi->s_journal);
838 sbi->s_journal = NULL;
839 if (err < 0)
Theodore Ts'oc67d8592010-06-29 11:07:07 -0400840 ext4_abort(sb, "Couldn't clean up the journal");
Frank Mayhar03901312009-01-07 00:06:22 -0500841 }
Josef Bacikd4edac32009-12-08 21:48:58 -0500842
Theodore Ts'oebd173b2015-09-23 12:46:17 -0400843 ext4_unregister_sysfs(sb);
Zheng Liud3922a72013-07-01 08:12:37 -0400844 ext4_es_unregister_shrinker(sbi);
Al Viro9105bb12013-12-08 20:52:31 -0500845 del_timer_sync(&sbi->s_err_report);
Josef Bacikd4edac32009-12-08 21:48:58 -0500846 ext4_release_system_zone(sb);
847 ext4_mb_release(sb);
848 ext4_ext_release(sb);
Josef Bacikd4edac32009-12-08 21:48:58 -0500849
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700850 if (!(sb->s_flags & MS_RDONLY)) {
Darrick J. Wonge2b911c2015-10-17 16:18:43 -0400851 ext4_clear_feature_journal_needs_recovery(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700852 es->s_state = cpu_to_le16(sbi->s_mount_state);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700853 }
Artem Bityutskiy58c58732012-07-22 20:33:31 -0400854 if (!(sb->s_flags & MS_RDONLY))
Artem Bityutskiya8e25a82012-03-21 22:29:15 -0400855 ext4_commit_super(sb, 1);
856
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700857 for (i = 0; i < sbi->s_gdb_count; i++)
858 brelse(sbi->s_group_desc[i]);
Al Virob93b41d2014-11-20 12:19:11 -0500859 kvfree(sbi->s_group_desc);
860 kvfree(sbi->s_flex_groups);
Theodore Ts'o57042652011-09-09 18:56:51 -0400861 percpu_counter_destroy(&sbi->s_freeclusters_counter);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700862 percpu_counter_destroy(&sbi->s_freeinodes_counter);
863 percpu_counter_destroy(&sbi->s_dirs_counter);
Theodore Ts'o57042652011-09-09 18:56:51 -0400864 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Daeho Jeongc8585c62016-04-25 23:22:35 -0400865 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700866 brelse(sbi->s_sbh);
867#ifdef CONFIG_QUOTA
Jan Karaa2d4a642014-09-11 11:15:15 -0400868 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700869 kfree(sbi->s_qf_names[i]);
870#endif
871
872 /* Debugging code just in case the in-memory inode orphan list
873 * isn't empty. The on-disk one can be non-empty if we've
874 * detected an error and taken the fs readonly, but the
875 * in-memory list had better be clean by this point. */
876 if (!list_empty(&sbi->s_orphan))
877 dump_orphan_list(sb, sbi);
878 J_ASSERT(list_empty(&sbi->s_orphan));
879
Theodore Ts'o89d96a62015-06-20 22:50:33 -0400880 sync_blockdev(sb->s_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700881 invalidate_bdev(sb->s_bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700882 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
883 /*
884 * Invalidate the journal device's buffers. We don't want them
885 * floating about in memory - the physical journal device may
886 * hotswapped, and it breaks the `ro-after' testing code.
887 */
888 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700889 invalidate_bdev(sbi->journal_bdev);
Mingming Cao617ba132006-10-11 01:20:53 -0700890 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700891 }
T Makphaibulchoke9c191f72014-03-18 19:24:49 -0400892 if (sbi->s_mb_cache) {
893 ext4_xattr_destroy_cache(sbi->s_mb_cache);
894 sbi->s_mb_cache = NULL;
895 }
Johann Lombardic5e06d12011-05-24 18:31:25 -0400896 if (sbi->s_mmp_tsk)
897 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700898 sb->s_fs_info = NULL;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400899 /*
900 * Now that we are completely done shutting down the
901 * superblock, we need to actually destroy the kobject.
902 */
Theodore Ts'o3197ebd2009-03-31 09:10:09 -0400903 kobject_put(&sbi->s_kobj);
904 wait_for_completion(&sbi->s_kobj_unregister);
Darrick J. Wong04419842012-04-29 18:27:10 -0400905 if (sbi->s_chksum_driver)
906 crypto_free_shash(sbi->s_chksum_driver);
Pekka Enberg705895b2009-02-15 18:07:52 -0500907 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700908 kfree(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700909}
910
Christoph Lametere18b8902006-12-06 20:33:20 -0800911static struct kmem_cache *ext4_inode_cachep;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700912
913/*
914 * Called inside transaction, so use GFP_NOFS
915 */
Mingming Cao617ba132006-10-11 01:20:53 -0700916static struct inode *ext4_alloc_inode(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700917{
Mingming Cao617ba132006-10-11 01:20:53 -0700918 struct ext4_inode_info *ei;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700919
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800920 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700921 if (!ei)
922 return NULL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -0400923
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700924 ei->vfs_inode.i_version = 1;
Theodore Ts'o202ee5d2014-04-21 14:37:55 -0400925 spin_lock_init(&ei->i_raw_lock);
Alex Tomasc9de5602008-01-29 00:19:52 -0500926 INIT_LIST_HEAD(&ei->i_prealloc_list);
927 spin_lock_init(&ei->i_prealloc_lock);
Zheng Liu9a26b662012-11-08 21:57:30 -0500928 ext4_es_init_tree(&ei->i_es_tree);
929 rwlock_init(&ei->i_es_lock);
Zheng Liuedaa53c2014-11-25 11:45:37 -0500930 INIT_LIST_HEAD(&ei->i_es_list);
Zheng Liueb68d0e2014-09-01 22:26:49 -0400931 ei->i_es_all_nr = 0;
Zheng Liuedaa53c2014-11-25 11:45:37 -0500932 ei->i_es_shk_nr = 0;
Jan Karadd475922014-11-25 11:51:23 -0500933 ei->i_es_shrink_lblk = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400934 ei->i_reserved_data_blocks = 0;
935 ei->i_reserved_meta_blocks = 0;
936 ei->i_allocated_meta_blocks = 0;
Theodore Ts'o9d0be502010-01-01 02:41:30 -0500937 ei->i_da_metadata_calc_len = 0;
Theodore Ts'o7e731bc2012-08-05 23:28:16 -0400938 ei->i_da_metadata_calc_last_lblock = 0;
Mingming Caod2a17632008-07-14 17:52:37 -0400939 spin_lock_init(&(ei->i_block_reservation_lock));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300940#ifdef CONFIG_QUOTA
941 ei->i_reserved_quota = 0;
Jan Kara96c7e0d2014-09-29 14:58:25 +0200942 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
Dmitry Monakhova9e7f442009-12-14 15:21:14 +0300943#endif
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500944 ei->jinode = NULL;
Jan Kara2e8fa542013-06-04 14:21:02 -0400945 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
Jiaying Zhang744692d2010-03-04 16:14:02 -0500946 spin_lock_init(&ei->i_completed_io_lock);
Jan Karab436b9b2009-12-08 23:51:10 -0500947 ei->i_sync_tid = 0;
948 ei->i_datasync_tid = 0;
Dmitry Monakhove27f41e2012-09-28 23:24:52 -0400949 atomic_set(&ei->i_unwritten, 0);
Jan Kara2e8fa542013-06-04 14:21:02 -0400950 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700951 return &ei->vfs_inode;
952}
953
Theodore Ts'o7ff9c072010-11-08 13:51:33 -0500954static int ext4_drop_inode(struct inode *inode)
955{
956 int drop = generic_drop_inode(inode);
957
958 trace_ext4_drop_inode(inode, drop);
959 return drop;
960}
961
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100962static void ext4_i_callback(struct rcu_head *head)
963{
964 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
966}
967
Mingming Cao617ba132006-10-11 01:20:53 -0700968static void ext4_destroy_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700969{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -0400971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
Vasily Averin9f7dd932007-07-15 23:40:45 -0700974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
976 true);
977 dump_stack();
978 }
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100979 call_rcu(&inode->i_rcu, ext4_i_callback);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700980}
981
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700982static void init_once(void *foo)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700983{
Mingming Cao617ba132006-10-11 01:20:53 -0700984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700985
Christoph Lametera35afb82007-05-16 22:10:57 -0700986 INIT_LIST_HEAD(&ei->i_orphan);
Christoph Lametera35afb82007-05-16 22:10:57 -0700987 init_rwsem(&ei->xattr_sem);
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -0500988 init_rwsem(&ei->i_data_sem);
Jan Karaea3d7202015-12-07 14:28:03 -0500989 init_rwsem(&ei->i_mmap_sem);
Christoph Lametera35afb82007-05-16 22:10:57 -0700990 inode_init_once(&ei->vfs_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700991}
992
Fabian Fredericke67bc2b2014-02-17 20:34:53 -0500993static int __init init_inodecache(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700994{
Mingming Cao617ba132006-10-11 01:20:53 -0700995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700997 0, (SLAB_RECLAIM_ACCOUNT|
Vladimir Davydov5d097052016-01-14 15:18:21 -0800998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
Paul Mundt20c2df82007-07-20 10:11:58 +0900999 init_once);
Mingming Cao617ba132006-10-11 01:20:53 -07001000 if (ext4_inode_cachep == NULL)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001001 return -ENOMEM;
1002 return 0;
1003}
1004
1005static void destroy_inodecache(void)
1006{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +10001007 /*
1008 * Make sure all delayed rcu free inodes are flushed before we
1009 * destroy cache.
1010 */
1011 rcu_barrier();
Mingming Cao617ba132006-10-11 01:20:53 -07001012 kmem_cache_destroy(ext4_inode_cachep);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001013}
1014
Al Viro0930fcc2010-06-07 13:16:22 -04001015void ext4_clear_inode(struct inode *inode)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001016{
Al Viro0930fcc2010-06-07 13:16:22 -04001017 invalidate_inode_buffers(inode);
Jan Karadbd57682012-05-03 14:48:02 +02001018 clear_inode(inode);
Christoph Hellwig9f754752010-03-03 09:05:05 -05001019 dquot_drop(inode);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -04001020 ext4_discard_preallocations(inode);
Zheng Liu51865fd2012-11-08 21:57:32 -05001021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -05001022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1027 }
Theodore Ts'ob7236e22015-05-18 13:17:47 -04001028#ifdef CONFIG_EXT4_FS_ENCRYPTION
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001029 fscrypt_put_encryption_info(inode, NULL);
Theodore Ts'ob7236e22015-05-18 13:17:47 -04001030#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001031}
1032
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001033static struct inode *ext4_nfs_get_inode(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001034 u64 ino, u32 generation)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001035{
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001036 struct inode *inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001037
Mingming Cao617ba132006-10-11 01:20:53 -07001038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001039 return ERR_PTR(-ESTALE);
Mingming Cao617ba132006-10-11 01:20:53 -07001040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001041 return ERR_PTR(-ESTALE);
1042
1043 /* iget isn't really right if the inode is currently unallocated!!
1044 *
Mingming Cao617ba132006-10-11 01:20:53 -07001045 * ext4_read_inode will return a bad_inode if the inode had been
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001046 * deleted, so we should be safe.
1047 *
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1050 */
Theodore Ts'of4bb2982014-10-05 22:56:00 -04001051 inode = ext4_iget_normal(sb, ino);
David Howells1d1fe1e2008-02-07 00:15:37 -08001052 if (IS_ERR(inode))
1053 return ERR_CAST(inode);
1054 if (generation && inode->i_generation != generation) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001055 iput(inode);
1056 return ERR_PTR(-ESTALE);
1057 }
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001058
1059 return inode;
1060}
1061
1062static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001063 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001064{
1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1067}
1068
1069static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001070 int fh_len, int fh_type)
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001071{
1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001074}
1075
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001076/*
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1081 */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001082static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1083 gfp_t wait)
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001084{
1085 journal_t *journal = EXT4_SB(sb)->s_journal;
1086
1087 WARN_ON(PageChecked(page));
1088 if (!page_has_buffers(page))
1089 return 0;
1090 if (journal)
1091 return jbd2_journal_try_to_free_buffers(journal, page,
Mel Gormand0164ad2015-11-06 16:28:21 -08001092 wait & ~__GFP_DIRECT_RECLAIM);
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001093 return try_to_free_buffers(page);
1094}
1095
Jaegeuk Kima7550b32016-07-10 14:01:03 -04001096#ifdef CONFIG_EXT4_FS_ENCRYPTION
1097static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1098{
1099 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1101}
1102
1103static int ext4_key_prefix(struct inode *inode, u8 **key)
1104{
1105 *key = EXT4_SB(inode->i_sb)->key_prefix;
1106 return EXT4_SB(inode->i_sb)->key_prefix_size;
1107}
1108
1109static int ext4_prepare_context(struct inode *inode)
1110{
1111 return ext4_convert_inline_data(inode);
1112}
1113
1114static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1115 void *fs_data)
1116{
1117 handle_t *handle;
1118 int res, res2;
1119
1120 /* fs_data is null when internally used. */
1121 if (fs_data) {
1122 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1123 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1124 len, 0);
1125 if (!res) {
1126 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1127 ext4_clear_inode_state(inode,
1128 EXT4_STATE_MAY_INLINE_DATA);
1129 }
1130 return res;
1131 }
1132
1133 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1134 ext4_jbd2_credits_xattr(inode));
1135 if (IS_ERR(handle))
1136 return PTR_ERR(handle);
1137
1138 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1139 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1140 len, 0);
1141 if (!res) {
1142 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1143 res = ext4_mark_inode_dirty(handle, inode);
1144 if (res)
1145 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1146 }
1147 res2 = ext4_journal_stop(handle);
1148 if (!res)
1149 res = res2;
1150 return res;
1151}
1152
1153static int ext4_dummy_context(struct inode *inode)
1154{
1155 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1156}
1157
1158static unsigned ext4_max_namelen(struct inode *inode)
1159{
1160 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1161 EXT4_NAME_LEN;
1162}
1163
1164static struct fscrypt_operations ext4_cryptops = {
1165 .get_context = ext4_get_context,
1166 .key_prefix = ext4_key_prefix,
1167 .prepare_context = ext4_prepare_context,
1168 .set_context = ext4_set_context,
1169 .dummy_context = ext4_dummy_context,
1170 .is_encrypted = ext4_encrypted_inode,
1171 .empty_dir = ext4_empty_dir,
1172 .max_namelen = ext4_max_namelen,
1173};
1174#else
1175static struct fscrypt_operations ext4_cryptops = {
1176 .is_encrypted = ext4_encrypted_inode,
1177};
1178#endif
1179
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001180#ifdef CONFIG_QUOTA
Li Xi689c9582016-01-08 16:01:22 -05001181static char *quotatypes[] = INITQFNAMES;
1182#define QTYPE2NAME(t) (quotatypes[t])
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001183
Mingming Cao617ba132006-10-11 01:20:53 -07001184static int ext4_write_dquot(struct dquot *dquot);
1185static int ext4_acquire_dquot(struct dquot *dquot);
1186static int ext4_release_dquot(struct dquot *dquot);
1187static int ext4_mark_dquot_dirty(struct dquot *dquot);
1188static int ext4_write_info(struct super_block *sb, int type);
Jan Kara6f28e082008-04-28 02:14:34 -07001189static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02001190 struct path *path);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001191static int ext4_quota_off(struct super_block *sb, int type);
Mingming Cao617ba132006-10-11 01:20:53 -07001192static int ext4_quota_on_mount(struct super_block *sb, int type);
1193static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001194 size_t len, loff_t off);
Mingming Cao617ba132006-10-11 01:20:53 -07001195static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001196 const char *data, size_t len, loff_t off);
Aditya Kali7c319d32012-07-22 20:21:31 -04001197static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1198 unsigned int flags);
1199static int ext4_enable_quotas(struct super_block *sb);
Theodore Ts'o8f0e8742016-04-01 12:00:03 -04001200static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001201
Jan Kara96c7e0d2014-09-29 14:58:25 +02001202static struct dquot **ext4_get_dquots(struct inode *inode)
1203{
1204 return EXT4_I(inode)->i_dquot;
1205}
1206
Alexey Dobriyan61e225d2009-09-21 17:01:08 -07001207static const struct dquot_operations ext4_quota_operations = {
Mingming Cao60e58e02009-01-22 18:13:05 +01001208 .get_reserved_space = ext4_get_reserved_space,
Mingming Cao617ba132006-10-11 01:20:53 -07001209 .write_dquot = ext4_write_dquot,
1210 .acquire_dquot = ext4_acquire_dquot,
1211 .release_dquot = ext4_release_dquot,
1212 .mark_dirty = ext4_mark_dquot_dirty,
Jan Karaa5b5ee32008-11-25 15:31:35 +01001213 .write_info = ext4_write_info,
1214 .alloc_dquot = dquot_alloc,
1215 .destroy_dquot = dquot_destroy,
Li Xi040cb372016-01-08 16:01:21 -05001216 .get_projid = ext4_get_projid,
Theodore Ts'o8f0e8742016-04-01 12:00:03 -04001217 .get_next_id = ext4_get_next_id,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001218};
1219
Alexey Dobriyan0d54b212009-09-21 17:01:09 -07001220static const struct quotactl_ops ext4_qctl_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -07001221 .quota_on = ext4_quota_on,
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04001222 .quota_off = ext4_quota_off,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001223 .quota_sync = dquot_quota_sync,
Jan Kara0a2403392014-11-19 00:42:09 +01001224 .get_state = dquot_get_state,
Christoph Hellwig287a8092010-05-19 07:16:45 -04001225 .set_info = dquot_set_dqinfo,
1226 .get_dqblk = dquot_get_dqblk,
Eric Sandeen6332b9b2016-02-19 19:19:01 +01001227 .set_dqblk = dquot_set_dqblk,
1228 .get_nextdqblk = dquot_get_next_dqblk,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001229};
1230#endif
1231
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -08001232static const struct super_operations ext4_sops = {
Mingming Cao617ba132006-10-11 01:20:53 -07001233 .alloc_inode = ext4_alloc_inode,
1234 .destroy_inode = ext4_destroy_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001235 .write_inode = ext4_write_inode,
1236 .dirty_inode = ext4_dirty_inode,
Theodore Ts'o7ff9c072010-11-08 13:51:33 -05001237 .drop_inode = ext4_drop_inode,
Al Viro0930fcc2010-06-07 13:16:22 -04001238 .evict_inode = ext4_evict_inode,
Mingming Cao617ba132006-10-11 01:20:53 -07001239 .put_super = ext4_put_super,
Mingming Cao617ba132006-10-11 01:20:53 -07001240 .sync_fs = ext4_sync_fs,
Takashi Satoc4be0c12009-01-09 16:40:58 -08001241 .freeze_fs = ext4_freeze,
1242 .unfreeze_fs = ext4_unfreeze,
Mingming Cao617ba132006-10-11 01:20:53 -07001243 .statfs = ext4_statfs,
1244 .remount_fs = ext4_remount,
Mingming Cao617ba132006-10-11 01:20:53 -07001245 .show_options = ext4_show_options,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001246#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07001247 .quota_read = ext4_quota_read,
1248 .quota_write = ext4_quota_write,
Jan Kara96c7e0d2014-09-29 14:58:25 +02001249 .get_dquots = ext4_get_dquots,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001250#endif
Toshiyuki Okajimac39a7f82009-01-05 22:38:48 -05001251 .bdev_try_to_free_page = bdev_try_to_free_page,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001252};
1253
Christoph Hellwig39655162007-10-21 16:42:17 -07001254static const struct export_operations ext4_export_ops = {
Christoph Hellwig1b961ac2007-10-21 16:42:08 -07001255 .fh_to_dentry = ext4_fh_to_dentry,
1256 .fh_to_parent = ext4_fh_to_parent,
Mingming Cao617ba132006-10-11 01:20:53 -07001257 .get_parent = ext4_get_parent,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001258};
1259
1260enum {
1261 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1262 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001263 Opt_nouid32, Opt_debug, Opt_removed,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001264 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Theodore Ts'o72578c32012-03-03 18:04:40 -05001265 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
Eric Sandeenad4eec62013-08-28 19:05:07 -04001266 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1267 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001268 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001269 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001270 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara5a20bdf2009-11-30 23:58:32 +01001271 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
Theodore Ts'oee4a3fc2012-03-02 12:14:24 -05001272 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
Jan Kara49da9392016-09-05 23:08:16 -04001273 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
Theodore Ts'o14490322010-12-14 15:27:50 -05001274 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001275 Opt_lazytime, Opt_nolazytime,
Theodore Ts'o14490322010-12-14 15:27:50 -05001276 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
Eric Sandeen5328e632009-11-19 14:25:42 -05001277 Opt_inode_readahead_blks, Opt_journal_ioprio,
Jiaying Zhang744692d2010-03-04 16:14:02 -05001278 Opt_dioread_nolock, Opt_dioread_lock,
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001279 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001280 Opt_max_dir_size_kb, Opt_nojournal_checksum,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001281};
1282
Steven Whitehousea447c092008-10-13 10:46:57 +01001283static const match_table_t tokens = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001284 {Opt_bsd_df, "bsddf"},
1285 {Opt_minix_df, "minixdf"},
1286 {Opt_grpid, "grpid"},
1287 {Opt_grpid, "bsdgroups"},
1288 {Opt_nogrpid, "nogrpid"},
1289 {Opt_nogrpid, "sysvgroups"},
1290 {Opt_resgid, "resgid=%u"},
1291 {Opt_resuid, "resuid=%u"},
1292 {Opt_sb, "sb=%u"},
1293 {Opt_err_cont, "errors=continue"},
1294 {Opt_err_panic, "errors=panic"},
1295 {Opt_err_ro, "errors=remount-ro"},
1296 {Opt_nouid32, "nouid32"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001297 {Opt_debug, "debug"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001298 {Opt_removed, "oldalloc"},
1299 {Opt_removed, "orlov"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001300 {Opt_user_xattr, "user_xattr"},
1301 {Opt_nouser_xattr, "nouser_xattr"},
1302 {Opt_acl, "acl"},
1303 {Opt_noacl, "noacl"},
Eric Sandeene3bb52a2009-11-19 14:28:50 -05001304 {Opt_noload, "norecovery"},
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001305 {Opt_noload, "noload"},
Theodore Ts'o72578c32012-03-03 18:04:40 -05001306 {Opt_removed, "nobh"},
1307 {Opt_removed, "bh"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001308 {Opt_commit, "commit=%u"},
Theodore Ts'o30773842009-01-03 20:27:38 -05001309 {Opt_min_batch_time, "min_batch_time=%u"},
1310 {Opt_max_batch_time, "max_batch_time=%u"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001311 {Opt_journal_dev, "journal_dev=%u"},
Eric Sandeenad4eec62013-08-28 19:05:07 -04001312 {Opt_journal_path, "journal_path=%s"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001313 {Opt_journal_checksum, "journal_checksum"},
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001314 {Opt_nojournal_checksum, "nojournal_checksum"},
Girish Shilamkar818d2762008-01-28 23:58:27 -05001315 {Opt_journal_async_commit, "journal_async_commit"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001316 {Opt_abort, "abort"},
1317 {Opt_data_journal, "data=journal"},
1318 {Opt_data_ordered, "data=ordered"},
1319 {Opt_data_writeback, "data=writeback"},
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001320 {Opt_data_err_abort, "data_err=abort"},
1321 {Opt_data_err_ignore, "data_err=ignore"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001322 {Opt_offusrjquota, "usrjquota="},
1323 {Opt_usrjquota, "usrjquota=%s"},
1324 {Opt_offgrpjquota, "grpjquota="},
1325 {Opt_grpjquota, "grpjquota=%s"},
1326 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1327 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Jan Kara5a20bdf2009-11-30 23:58:32 +01001328 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001329 {Opt_grpquota, "grpquota"},
1330 {Opt_noquota, "noquota"},
1331 {Opt_quota, "quota"},
1332 {Opt_usrquota, "usrquota"},
Jan Kara49da9392016-09-05 23:08:16 -04001333 {Opt_prjquota, "prjquota"},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001334 {Opt_barrier, "barrier=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001335 {Opt_barrier, "barrier"},
1336 {Opt_nobarrier, "nobarrier"},
Jean Noel Cordenner25ec56b2008-01-28 23:58:27 -05001337 {Opt_i_version, "i_version"},
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001338 {Opt_dax, "dax"},
Alex Tomasc9de5602008-01-29 00:19:52 -05001339 {Opt_stripe, "stripe=%u"},
Alex Tomas64769242008-07-11 19:27:31 -04001340 {Opt_delalloc, "delalloc"},
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001341 {Opt_lazytime, "lazytime"},
1342 {Opt_nolazytime, "nolazytime"},
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04001343 {Opt_nodelalloc, "nodelalloc"},
Jan Kara36ade452013-01-28 09:30:52 -05001344 {Opt_removed, "mblk_io_submit"},
1345 {Opt_removed, "nomblk_io_submit"},
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04001346 {Opt_block_validity, "block_validity"},
1347 {Opt_noblock_validity, "noblock_validity"},
Theodore Ts'o240799c2008-10-09 23:53:47 -04001348 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001349 {Opt_journal_ioprio, "journal_ioprio=%u"},
Theodore Ts'oafd46722009-03-16 23:12:23 -04001350 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
Theodore Ts'o06705bf2009-03-28 10:59:57 -04001351 {Opt_auto_da_alloc, "auto_da_alloc"},
1352 {Opt_noauto_da_alloc, "noauto_da_alloc"},
Jiaying Zhang744692d2010-03-04 16:14:02 -05001353 {Opt_dioread_nolock, "dioread_nolock"},
1354 {Opt_dioread_lock, "dioread_lock"},
Eric Sandeen5328e632009-11-19 14:25:42 -05001355 {Opt_discard, "discard"},
1356 {Opt_nodiscard, "nodiscard"},
Theodore Ts'ofc6cb1c2011-12-12 22:06:18 -05001357 {Opt_init_itable, "init_itable=%u"},
1358 {Opt_init_itable, "init_itable"},
1359 {Opt_noinit_itable, "noinit_itable"},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001360 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001361 {Opt_test_dummy_encryption, "test_dummy_encryption"},
Theodore Ts'oc7198b92012-03-04 22:00:53 -05001362 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1363 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1364 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1365 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1366 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
Josef Bacikf3f12fa2008-04-29 22:05:28 -04001367 {Opt_err, NULL},
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001368};
1369
Mingming Cao617ba132006-10-11 01:20:53 -07001370static ext4_fsblk_t get_sb_block(void **data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001371{
Mingming Cao617ba132006-10-11 01:20:53 -07001372 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001373 char *options = (char *) *data;
1374
1375 if (!options || strncmp(options, "sb=", 3) != 0)
1376 return 1; /* Default location */
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001377
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001378 options += 3;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001379 /* TODO: use simple_strtoll with >32bit ext4 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001380 sb_block = simple_strtoul(options, &options, 0);
1381 if (*options && *options != ',') {
Theodore Ts'o4776004f2008-09-08 23:00:52 -04001382 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001383 (char *) *data);
1384 return 1;
1385 }
1386 if (*options == ',')
1387 options++;
1388 *data = (void *) options;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04001389
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001390 return sb_block;
1391}
1392
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001393#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
Dmitry Monakhov437ca0f2010-03-01 22:29:21 -05001394static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1395 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001396
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001397#ifdef CONFIG_QUOTA
1398static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1399{
1400 struct ext4_sb_info *sbi = EXT4_SB(sb);
1401 char *qname;
Chen Gang03dafb52013-01-24 23:24:58 -05001402 int ret = -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001403
1404 if (sb_any_quota_loaded(sb) &&
1405 !sbi->s_qf_names[qtype]) {
1406 ext4_msg(sb, KERN_ERR,
1407 "Cannot change journaled "
1408 "quota options when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001409 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001410 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04001411 if (ext4_has_feature_quota(sb)) {
Theodore Ts'oc325a672016-04-03 17:03:37 -04001412 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1413 "ignored when QUOTA feature is enabled");
1414 return 1;
Jan Kara262b4662013-03-02 17:57:08 -05001415 }
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001416 qname = match_strdup(args);
1417 if (!qname) {
1418 ext4_msg(sb, KERN_ERR,
1419 "Not enough memory for storing quotafile name");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001420 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001421 }
Chen Gang03dafb52013-01-24 23:24:58 -05001422 if (sbi->s_qf_names[qtype]) {
1423 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1424 ret = 1;
1425 else
1426 ext4_msg(sb, KERN_ERR,
1427 "%s quota file already specified",
1428 QTYPE2NAME(qtype));
1429 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001430 }
Chen Gang03dafb52013-01-24 23:24:58 -05001431 if (strchr(qname, '/')) {
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001432 ext4_msg(sb, KERN_ERR,
1433 "quotafile must be on filesystem root");
Chen Gang03dafb52013-01-24 23:24:58 -05001434 goto errout;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001435 }
Chen Gang03dafb52013-01-24 23:24:58 -05001436 sbi->s_qf_names[qtype] = qname;
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001437 set_opt(sb, QUOTA);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001438 return 1;
Chen Gang03dafb52013-01-24 23:24:58 -05001439errout:
1440 kfree(qname);
1441 return ret;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001442}
1443
1444static int clear_qf_name(struct super_block *sb, int qtype)
1445{
1446
1447 struct ext4_sb_info *sbi = EXT4_SB(sb);
1448
1449 if (sb_any_quota_loaded(sb) &&
1450 sbi->s_qf_names[qtype]) {
1451 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1452 " when quota turned on");
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001453 return -1;
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001454 }
Chen Gang03dafb52013-01-24 23:24:58 -05001455 kfree(sbi->s_qf_names[qtype]);
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001456 sbi->s_qf_names[qtype] = NULL;
1457 return 1;
1458}
1459#endif
1460
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001461#define MOPT_SET 0x0001
1462#define MOPT_CLEAR 0x0002
1463#define MOPT_NOSUPPORT 0x0004
1464#define MOPT_EXPLICIT 0x0008
1465#define MOPT_CLEAR_ERR 0x0010
1466#define MOPT_GTE0 0x0020
1467#ifdef CONFIG_QUOTA
1468#define MOPT_Q 0
1469#define MOPT_QFMT 0x0040
1470#else
1471#define MOPT_Q MOPT_NOSUPPORT
1472#define MOPT_QFMT MOPT_NOSUPPORT
1473#endif
1474#define MOPT_DATAJ 0x0080
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001475#define MOPT_NO_EXT2 0x0100
1476#define MOPT_NO_EXT3 0x0200
1477#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
Eric Sandeenad4eec62013-08-28 19:05:07 -04001478#define MOPT_STRING 0x0400
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001479
1480static const struct mount_opts {
1481 int token;
1482 int mount_opt;
1483 int flags;
1484} ext4_mount_opts[] = {
1485 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1486 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1487 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1488 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001489 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1490 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001491 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1492 MOPT_EXT4_ONLY | MOPT_SET},
1493 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1494 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001495 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1496 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001497 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1498 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1499 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
Theodore Ts'o59d9fa52013-08-08 23:01:24 -04001500 MOPT_EXT4_ONLY | MOPT_CLEAR},
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05001501 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1502 MOPT_EXT4_ONLY | MOPT_CLEAR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001503 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04001504 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001505 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001506 EXT4_MOUNT_JOURNAL_CHECKSUM),
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04001507 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001508 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001509 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1510 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1511 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001512 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
Ales Novak7915a862016-03-12 21:55:50 -05001513 MOPT_NO_EXT2},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001514 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
Ales Novak7915a862016-03-12 21:55:50 -05001515 MOPT_NO_EXT2},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001516 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1517 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1518 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1519 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1520 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1521 {Opt_commit, 0, MOPT_GTE0},
1522 {Opt_max_batch_time, 0, MOPT_GTE0},
1523 {Opt_min_batch_time, 0, MOPT_GTE0},
1524 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1525 {Opt_init_itable, 0, MOPT_GTE0},
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001526 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001527 {Opt_stripe, 0, MOPT_GTE0},
Jan Kara0efb3b22013-02-02 22:52:19 -05001528 {Opt_resuid, 0, MOPT_GTE0},
1529 {Opt_resgid, 0, MOPT_GTE0},
Carlos Maiolino5ba92bc2015-07-21 23:57:59 -04001530 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1531 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1532 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001533 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1534 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1535 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1536 MOPT_NO_EXT2 | MOPT_DATAJ},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001537 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1538 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001539#ifdef CONFIG_EXT4_FS_POSIX_ACL
1540 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1541 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1542#else
1543 {Opt_acl, 0, MOPT_NOSUPPORT},
1544 {Opt_noacl, 0, MOPT_NOSUPPORT},
1545#endif
1546 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1547 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1548 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1549 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1550 MOPT_SET | MOPT_Q},
1551 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1552 MOPT_SET | MOPT_Q},
Jan Kara49da9392016-09-05 23:08:16 -04001553 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1554 MOPT_SET | MOPT_Q},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001555 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
Jan Kara49da9392016-09-05 23:08:16 -04001556 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1557 MOPT_CLEAR | MOPT_Q},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001558 {Opt_usrjquota, 0, MOPT_Q},
1559 {Opt_grpjquota, 0, MOPT_Q},
1560 {Opt_offusrjquota, 0, MOPT_Q},
1561 {Opt_offgrpjquota, 0, MOPT_Q},
1562 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1563 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1564 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
Theodore Ts'odf981d02012-08-17 09:48:17 -04001565 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001566 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001567 {Opt_err, 0, 0}
1568};
1569
1570static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1571 substring_t *args, unsigned long *journal_devnum,
1572 unsigned int *journal_ioprio, int is_remount)
1573{
1574 struct ext4_sb_info *sbi = EXT4_SB(sb);
1575 const struct mount_opts *m;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001576 kuid_t uid;
1577 kgid_t gid;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001578 int arg = 0;
1579
Theodore Ts'o57f73c22012-04-16 18:55:26 -04001580#ifdef CONFIG_QUOTA
1581 if (token == Opt_usrjquota)
1582 return set_qf_name(sb, USRQUOTA, &args[0]);
1583 else if (token == Opt_grpjquota)
1584 return set_qf_name(sb, GRPQUOTA, &args[0]);
1585 else if (token == Opt_offusrjquota)
1586 return clear_qf_name(sb, USRQUOTA);
1587 else if (token == Opt_offgrpjquota)
1588 return clear_qf_name(sb, GRPQUOTA);
1589#endif
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001590 switch (token) {
Theodore Ts'of7048602012-03-04 22:06:20 -05001591 case Opt_noacl:
1592 case Opt_nouser_xattr:
1593 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1594 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001595 case Opt_sb:
1596 return 1; /* handled by get_sb_block() */
1597 case Opt_removed:
Jan Kara5f3633e2013-02-02 23:09:36 -05001598 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001599 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001600 case Opt_abort:
1601 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1602 return 1;
1603 case Opt_i_version:
1604 sb->s_flags |= MS_I_VERSION;
1605 return 1;
Theodore Ts'oa26f4992015-02-02 00:37:02 -05001606 case Opt_lazytime:
1607 sb->s_flags |= MS_LAZYTIME;
1608 return 1;
1609 case Opt_nolazytime:
1610 sb->s_flags &= ~MS_LAZYTIME;
1611 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001612 }
1613
Jan Kara5f3633e2013-02-02 23:09:36 -05001614 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1615 if (token == m->token)
1616 break;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001617
Jan Kara5f3633e2013-02-02 23:09:36 -05001618 if (m->token == Opt_err) {
1619 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1620 "or missing value", opt);
1621 return -1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001622 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001623
Theodore Ts'o8dc0aa82013-02-02 23:38:39 -05001624 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1625 ext4_msg(sb, KERN_ERR,
1626 "Mount option \"%s\" incompatible with ext2", opt);
1627 return -1;
1628 }
1629 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1630 ext4_msg(sb, KERN_ERR,
1631 "Mount option \"%s\" incompatible with ext3", opt);
1632 return -1;
1633 }
1634
Eric Sandeenad4eec62013-08-28 19:05:07 -04001635 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
Jan Kara5f3633e2013-02-02 23:09:36 -05001636 return -1;
1637 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1638 return -1;
Dmitry Monakhovc93cf2d2015-10-18 23:35:32 -04001639 if (m->flags & MOPT_EXPLICIT) {
1640 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1641 set_opt2(sb, EXPLICIT_DELALLOC);
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04001642 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1643 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
Dmitry Monakhovc93cf2d2015-10-18 23:35:32 -04001644 } else
1645 return -1;
1646 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001647 if (m->flags & MOPT_CLEAR_ERR)
1648 clear_opt(sb, ERRORS_MASK);
1649 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1650 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1651 "options when quota turned on");
1652 return -1;
1653 }
1654
1655 if (m->flags & MOPT_NOSUPPORT) {
1656 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1657 } else if (token == Opt_commit) {
1658 if (arg == 0)
1659 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1660 sbi->s_commit_interval = HZ * arg;
1661 } else if (token == Opt_max_batch_time) {
Jan Kara5f3633e2013-02-02 23:09:36 -05001662 sbi->s_max_batch_time = arg;
1663 } else if (token == Opt_min_batch_time) {
1664 sbi->s_min_batch_time = arg;
1665 } else if (token == Opt_inode_readahead_blks) {
Jan Karae33e60e2013-02-02 23:14:31 -05001666 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1667 ext4_msg(sb, KERN_ERR,
1668 "EXT4-fs: inode_readahead_blks must be "
1669 "0 or a power of 2 smaller than 2^31");
Jan Kara5f3633e2013-02-02 23:09:36 -05001670 return -1;
1671 }
1672 sbi->s_inode_readahead_blks = arg;
1673 } else if (token == Opt_init_itable) {
1674 set_opt(sb, INIT_INODE_TABLE);
1675 if (!args->from)
1676 arg = EXT4_DEF_LI_WAIT_MULT;
1677 sbi->s_li_wait_mult = arg;
1678 } else if (token == Opt_max_dir_size_kb) {
1679 sbi->s_max_dir_size_kb = arg;
1680 } else if (token == Opt_stripe) {
1681 sbi->s_stripe = arg;
1682 } else if (token == Opt_resuid) {
1683 uid = make_kuid(current_user_ns(), arg);
1684 if (!uid_valid(uid)) {
1685 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1686 return -1;
1687 }
1688 sbi->s_resuid = uid;
1689 } else if (token == Opt_resgid) {
1690 gid = make_kgid(current_user_ns(), arg);
1691 if (!gid_valid(gid)) {
1692 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1693 return -1;
1694 }
1695 sbi->s_resgid = gid;
1696 } else if (token == Opt_journal_dev) {
1697 if (is_remount) {
1698 ext4_msg(sb, KERN_ERR,
1699 "Cannot specify journal on remount");
1700 return -1;
1701 }
1702 *journal_devnum = arg;
Eric Sandeenad4eec62013-08-28 19:05:07 -04001703 } else if (token == Opt_journal_path) {
1704 char *journal_path;
1705 struct inode *journal_inode;
1706 struct path path;
1707 int error;
1708
1709 if (is_remount) {
1710 ext4_msg(sb, KERN_ERR,
1711 "Cannot specify journal on remount");
1712 return -1;
1713 }
1714 journal_path = match_strdup(&args[0]);
1715 if (!journal_path) {
1716 ext4_msg(sb, KERN_ERR, "error: could not dup "
1717 "journal device string");
1718 return -1;
1719 }
1720
1721 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1722 if (error) {
1723 ext4_msg(sb, KERN_ERR, "error: could not find "
1724 "journal device path: error %d", error);
1725 kfree(journal_path);
1726 return -1;
1727 }
1728
David Howells2b0143b2015-03-17 22:25:59 +00001729 journal_inode = d_inode(path.dentry);
Eric Sandeenad4eec62013-08-28 19:05:07 -04001730 if (!S_ISBLK(journal_inode->i_mode)) {
1731 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1732 "is not a block device", journal_path);
1733 path_put(&path);
1734 kfree(journal_path);
1735 return -1;
1736 }
1737
1738 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1739 path_put(&path);
1740 kfree(journal_path);
Jan Kara5f3633e2013-02-02 23:09:36 -05001741 } else if (token == Opt_journal_ioprio) {
1742 if (arg > 7) {
1743 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1744 " (must be 0-7)");
1745 return -1;
1746 }
1747 *journal_ioprio =
1748 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04001749 } else if (token == Opt_test_dummy_encryption) {
1750#ifdef CONFIG_EXT4_FS_ENCRYPTION
1751 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1752 ext4_msg(sb, KERN_WARNING,
1753 "Test dummy encryption mode enabled");
1754#else
1755 ext4_msg(sb, KERN_WARNING,
1756 "Test dummy encryption mount option ignored");
1757#endif
Jan Kara5f3633e2013-02-02 23:09:36 -05001758 } else if (m->flags & MOPT_DATAJ) {
1759 if (is_remount) {
1760 if (!sbi->s_journal)
1761 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1762 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1763 ext4_msg(sb, KERN_ERR,
1764 "Cannot change data mode on remount");
1765 return -1;
1766 }
1767 } else {
1768 clear_opt(sb, DATA_FLAGS);
1769 sbi->s_mount_opt |= m->mount_opt;
1770 }
1771#ifdef CONFIG_QUOTA
1772 } else if (m->flags & MOPT_QFMT) {
1773 if (sb_any_quota_loaded(sb) &&
1774 sbi->s_jquota_fmt != m->mount_opt) {
1775 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1776 "quota options when quota turned on");
1777 return -1;
1778 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04001779 if (ext4_has_feature_quota(sb)) {
Theodore Ts'oc325a672016-04-03 17:03:37 -04001780 ext4_msg(sb, KERN_INFO,
1781 "Quota format mount options ignored "
Jan Kara262b4662013-03-02 17:57:08 -05001782 "when QUOTA feature is enabled");
Theodore Ts'oc325a672016-04-03 17:03:37 -04001783 return 1;
Jan Kara262b4662013-03-02 17:57:08 -05001784 }
Jan Kara5f3633e2013-02-02 23:09:36 -05001785 sbi->s_jquota_fmt = m->mount_opt;
1786#endif
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001787 } else if (token == Opt_dax) {
Dan Williamsef83b6e2015-09-29 15:48:11 -04001788#ifdef CONFIG_FS_DAX
1789 ext4_msg(sb, KERN_WARNING,
1790 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1791 sbi->s_mount_opt |= m->mount_opt;
1792#else
Ross Zwisler923ae0f2015-02-16 15:59:38 -08001793 ext4_msg(sb, KERN_INFO, "dax option not supported");
1794 return -1;
1795#endif
Ales Novak7915a862016-03-12 21:55:50 -05001796 } else if (token == Opt_data_err_abort) {
1797 sbi->s_mount_opt |= m->mount_opt;
1798 } else if (token == Opt_data_err_ignore) {
1799 sbi->s_mount_opt &= ~m->mount_opt;
Jan Kara5f3633e2013-02-02 23:09:36 -05001800 } else {
1801 if (!args->from)
1802 arg = 1;
1803 if (m->flags & MOPT_CLEAR)
1804 arg = !arg;
1805 else if (unlikely(!(m->flags & MOPT_SET))) {
1806 ext4_msg(sb, KERN_WARNING,
1807 "buggy handling of option %s", opt);
1808 WARN_ON(1);
1809 return -1;
1810 }
1811 if (arg != 0)
1812 sbi->s_mount_opt |= m->mount_opt;
1813 else
1814 sbi->s_mount_opt &= ~m->mount_opt;
1815 }
1816 return 1;
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001817}
1818
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001819static int parse_options(char *options, struct super_block *sb,
Theodore Ts'oc3191062009-01-06 11:14:25 -05001820 unsigned long *journal_devnum,
Theodore Ts'ob3881f72009-01-05 22:46:26 -05001821 unsigned int *journal_ioprio,
Eric Sandeen661aa522012-02-20 17:53:04 -05001822 int is_remount)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001823{
Mingming Cao617ba132006-10-11 01:20:53 -07001824 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001825 char *p;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001826 substring_t args[MAX_OPT_ARGS];
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001827 int token;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001828
1829 if (!options)
1830 return 1;
1831
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04001832 while ((p = strsep(&options, ",")) != NULL) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001833 if (!*p)
1834 continue;
Eric Sandeen15121c12010-02-15 20:17:55 -05001835 /*
1836 * Initialize args struct so we know whether arg was
1837 * found; some options take optional arguments.
1838 */
Sachin Kamatcaecd0a2012-08-18 22:29:18 -04001839 args[0].to = args[0].from = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001840 token = match_token(p, tokens, args);
Theodore Ts'o26092bf2012-03-03 23:20:47 -05001841 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1842 journal_ioprio, is_remount) < 0)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001843 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001844 }
1845#ifdef CONFIG_QUOTA
Jan Kara49da9392016-09-05 23:08:16 -04001846 /*
1847 * We do the test below only for project quotas. 'usrquota' and
1848 * 'grpquota' mount options are allowed even without quota feature
1849 * to support legacy quotas in quota files.
1850 */
1851 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1852 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1853 "Cannot enable project quota enforcement.");
1854 return 0;
1855 }
1856 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001857 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001858 clear_opt(sb, USRQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001859
Dmitry Monakhov482a7422010-02-24 11:35:32 -05001860 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05001861 clear_opt(sb, GRPQUOTA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001862
Dmitry Monakhov56c50f12010-03-01 23:28:41 -05001863 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001864 ext4_msg(sb, KERN_ERR, "old and new quota "
1865 "format mixing");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001866 return 0;
1867 }
1868
1869 if (!sbi->s_jquota_fmt) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04001870 ext4_msg(sb, KERN_ERR, "journaled quota format "
1871 "not specified");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001872 return 0;
1873 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001874 }
1875#endif
Jan Kara261cb20c2012-12-20 00:07:18 -05001876 if (test_opt(sb, DIOREAD_NOLOCK)) {
1877 int blocksize =
1878 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1879
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001880 if (blocksize < PAGE_SIZE) {
Jan Kara261cb20c2012-12-20 00:07:18 -05001881 ext4_msg(sb, KERN_ERR, "can't mount with "
1882 "dioread_nolock if block size != PAGE_SIZE");
1883 return 0;
1884 }
1885 }
Jan Karad4f76102014-11-25 20:19:17 -05001886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1887 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1888 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1889 "in data=ordered mode");
1890 return 0;
1891 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001892 return 1;
1893}
1894
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001895static inline void ext4_show_quota_options(struct seq_file *seq,
1896 struct super_block *sb)
1897{
1898#if defined(CONFIG_QUOTA)
1899 struct ext4_sb_info *sbi = EXT4_SB(sb);
1900
1901 if (sbi->s_jquota_fmt) {
1902 char *fmtname = "";
1903
1904 switch (sbi->s_jquota_fmt) {
1905 case QFMT_VFS_OLD:
1906 fmtname = "vfsold";
1907 break;
1908 case QFMT_VFS_V0:
1909 fmtname = "vfsv0";
1910 break;
1911 case QFMT_VFS_V1:
1912 fmtname = "vfsv1";
1913 break;
1914 }
1915 seq_printf(seq, ",jqfmt=%s", fmtname);
1916 }
1917
1918 if (sbi->s_qf_names[USRQUOTA])
Kees Cooka068acf2015-09-04 15:44:57 -07001919 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001920
1921 if (sbi->s_qf_names[GRPQUOTA])
Kees Cooka068acf2015-09-04 15:44:57 -07001922 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001923#endif
1924}
1925
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001926static const char *token2str(int token)
1927{
Herton Ronaldo Krzesinski50df9fd2012-09-23 22:49:12 -04001928 const struct match_token *t;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001929
1930 for (t = tokens; t->token != Opt_err; t++)
1931 if (t->token == token && !strchr(t->pattern, '='))
1932 break;
1933 return t->pattern;
1934}
1935
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001936/*
1937 * Show an option if
1938 * - it's set to a non-default value OR
1939 * - if the per-sb default is different from the global default
1940 */
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001941static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1942 int nodefs)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001943{
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001944 struct ext4_sb_info *sbi = EXT4_SB(sb);
1945 struct ext4_super_block *es = sbi->s_es;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001946 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001947 const struct mount_opts *m;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001948 char sep = nodefs ? '\n' : ',';
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001949
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001950#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1951#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001952
1953 if (sbi->s_sb_block != 1)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001954 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1955
1956 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1957 int want_set = m->flags & MOPT_SET;
1958 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1959 (m->flags & MOPT_CLEAR_ERR))
1960 continue;
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001961 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001962 continue; /* skip if same as the default */
1963 if ((want_set &&
1964 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1965 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1966 continue; /* select Opt_noFoo vs Opt_Foo */
1967 SEQ_OPTS_PRINT("%s", token2str(m->token));
1968 }
1969
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001970 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001971 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001972 SEQ_OPTS_PRINT("resuid=%u",
1973 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1974 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001975 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
Eric W. Biederman08cefc72012-02-07 15:41:49 -08001976 SEQ_OPTS_PRINT("resgid=%u",
1977 from_kgid_munged(&init_user_ns, sbi->s_resgid));
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001978 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001979 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1980 SEQ_OPTS_PUTS("errors=remount-ro");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001981 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001982 SEQ_OPTS_PUTS("errors=continue");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001983 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001984 SEQ_OPTS_PUTS("errors=panic");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001985 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001986 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001987 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001988 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001989 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001990 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05001991 if (sb->s_flags & MS_I_VERSION)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001992 SEQ_OPTS_PUTS("i_version");
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001993 if (nodefs || sbi->s_stripe)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001994 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05001995 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
Theodore Ts'o5a916be2012-03-04 19:27:31 -05001996 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1997 SEQ_OPTS_PUTS("data=journal");
1998 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1999 SEQ_OPTS_PUTS("data=ordered");
2000 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2001 SEQ_OPTS_PUTS("data=writeback");
2002 }
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002003 if (nodefs ||
2004 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002005 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2006 sbi->s_inode_readahead_blks);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002007
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002008 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2009 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
Theodore Ts'o5a916be2012-03-04 19:27:31 -05002010 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
Theodore Ts'odf981d02012-08-17 09:48:17 -04002011 if (nodefs || sbi->s_max_dir_size_kb)
2012 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
Ales Novak7915a862016-03-12 21:55:50 -05002013 if (test_opt(sb, DATA_ERR_ABORT))
2014 SEQ_OPTS_PUTS("data_err=abort");
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002015
2016 ext4_show_quota_options(seq, sb);
Theodore Ts'o2adf6da2012-03-03 23:20:50 -05002017 return 0;
2018}
2019
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002020static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2021{
2022 return _ext4_show_options(seq, root->d_sb, 0);
2023}
2024
Theodore Ts'oebd173b2015-09-23 12:46:17 -04002025int ext4_seq_options_show(struct seq_file *seq, void *offset)
Theodore Ts'o66acdcf2012-03-04 20:21:38 -05002026{
2027 struct super_block *sb = seq->private;
2028 int rc;
2029
2030 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2031 rc = _ext4_show_options(seq, sb, 1);
2032 seq_puts(seq, "\n");
2033 return rc;
2034}
2035
Mingming Cao617ba132006-10-11 01:20:53 -07002036static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002037 int read_only)
2038{
Mingming Cao617ba132006-10-11 01:20:53 -07002039 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002040 int res = 0;
2041
Mingming Cao617ba132006-10-11 01:20:53 -07002042 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002043 ext4_msg(sb, KERN_ERR, "revision level too high, "
2044 "forcing read-only mode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002045 res = MS_RDONLY;
2046 }
2047 if (read_only)
Theodore Ts'o281b5992011-09-09 18:34:51 -04002048 goto done;
Mingming Cao617ba132006-10-11 01:20:53 -07002049 if (!(sbi->s_mount_state & EXT4_VALID_FS))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002050 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2051 "running e2fsck is recommended");
Lukas Czernerc8b459f2014-05-12 12:55:07 -04002052 else if (sbi->s_mount_state & EXT4_ERROR_FS)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002053 ext4_msg(sb, KERN_WARNING,
2054 "warning: mounting fs with errors, "
2055 "running e2fsck is recommended");
Tao Maed3ce802011-05-18 13:29:57 -04002056 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002057 le16_to_cpu(es->s_mnt_count) >=
2058 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002059 ext4_msg(sb, KERN_WARNING,
2060 "warning: maximal mount count reached, "
2061 "running e2fsck is recommended");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002062 else if (le32_to_cpu(es->s_checkinterval) &&
2063 (le32_to_cpu(es->s_lastcheck) +
2064 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
Eric Sandeenb31e1552009-06-04 17:36:36 -04002065 ext4_msg(sb, KERN_WARNING,
2066 "warning: checktime reached, "
2067 "running e2fsck is recommended");
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002068 if (!sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05002069 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002070 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
Mingming Cao617ba132006-10-11 01:20:53 -07002071 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
Marcin Slusarze8546d02008-04-17 10:38:59 -04002072 le16_add_cpu(&es->s_mnt_count, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002073 es->s_mtime = cpu_to_le32(get_seconds());
Mingming Cao617ba132006-10-11 01:20:53 -07002074 ext4_update_dynamic_rev(sb);
Frank Mayhar03901312009-01-07 00:06:22 -05002075 if (sbi->s_journal)
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002076 ext4_set_feature_journal_needs_recovery(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002077
Theodore Ts'oe2d67052009-05-01 00:33:44 -04002078 ext4_commit_super(sb, 1);
Theodore Ts'o281b5992011-09-09 18:34:51 -04002079done:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002080 if (test_opt(sb, DEBUG))
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002081 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
Theodore Ts'oa2595b82010-12-15 20:30:48 -05002082 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002083 sb->s_blocksize,
2084 sbi->s_groups_count,
Mingming Cao617ba132006-10-11 01:20:53 -07002085 EXT4_BLOCKS_PER_GROUP(sb),
2086 EXT4_INODES_PER_GROUP(sb),
Theodore Ts'oa2595b82010-12-15 20:30:48 -05002087 sbi->s_mount_opt, sbi->s_mount_opt2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002088
Dan Magenheimer7abc52c2011-05-26 10:02:03 -06002089 cleancache_init_fs(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002090 return res;
2091}
2092
Theodore Ts'o117fff12012-09-05 01:29:50 -04002093int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2094{
2095 struct ext4_sb_info *sbi = EXT4_SB(sb);
2096 struct flex_groups *new_groups;
2097 int size;
2098
2099 if (!sbi->s_log_groups_per_flex)
2100 return 0;
2101
2102 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2103 if (size <= sbi->s_flex_groups_allocated)
2104 return 0;
2105
2106 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2107 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2108 if (!new_groups) {
2109 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2110 size / (int) sizeof(struct flex_groups));
2111 return -ENOMEM;
2112 }
2113
2114 if (sbi->s_flex_groups) {
2115 memcpy(new_groups, sbi->s_flex_groups,
2116 (sbi->s_flex_groups_allocated *
2117 sizeof(struct flex_groups)));
Al Virob93b41d2014-11-20 12:19:11 -05002118 kvfree(sbi->s_flex_groups);
Theodore Ts'o117fff12012-09-05 01:29:50 -04002119 }
2120 sbi->s_flex_groups = new_groups;
2121 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2122 return 0;
2123}
2124
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002125static int ext4_fill_flex_info(struct super_block *sb)
2126{
2127 struct ext4_sb_info *sbi = EXT4_SB(sb);
2128 struct ext4_group_desc *gdp = NULL;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002129 ext4_group_t flex_group;
Theodore Ts'o117fff12012-09-05 01:29:50 -04002130 int i, err;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002131
Theodore Ts'o503358a2009-11-23 07:24:46 -05002132 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
Xi Wangd50f2ab2012-01-10 11:51:10 -05002133 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002134 sbi->s_log_groups_per_flex = 0;
2135 return 1;
2136 }
2137
Theodore Ts'o117fff12012-09-05 01:29:50 -04002138 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2139 if (err)
Theodore Ts'o9933fc02011-08-01 08:45:02 -04002140 goto failed;
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002141
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002142 for (i = 0; i < sbi->s_groups_count; i++) {
Theodore Ts'o88b6edd2009-05-25 11:50:39 -04002143 gdp = ext4_get_group_desc(sb, i, NULL);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002144
2145 flex_group = ext4_flex_group(sbi, i);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002146 atomic_add(ext4_free_inodes_count(sb, gdp),
2147 &sbi->s_flex_groups[flex_group].free_inodes);
Theodore Ts'o90ba9832013-03-11 23:39:59 -04002148 atomic64_add(ext4_free_group_clusters(sb, gdp),
2149 &sbi->s_flex_groups[flex_group].free_clusters);
Theodore Ts'o7ad9bb62009-09-11 16:51:28 -04002150 atomic_add(ext4_used_dirs_count(sb, gdp),
2151 &sbi->s_flex_groups[flex_group].used_dirs);
Jose R. Santos772cb7c2008-07-11 19:27:31 -04002152 }
2153
2154 return 1;
2155failed:
2156 return 0;
2157}
2158
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002159static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002160 struct ext4_group_desc *gdp)
Andreas Dilger717d50e2007-10-16 18:38:25 -04002161{
Daeho Jeongb47820e2016-07-03 17:51:39 -04002162 int offset = offsetof(struct ext4_group_desc, bg_checksum);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002163 __u16 crc = 0;
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002164 __le32 le_group = cpu_to_le32(block_group);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002165 struct ext4_sb_info *sbi = EXT4_SB(sb);
Andreas Dilger717d50e2007-10-16 18:38:25 -04002166
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -04002167 if (ext4_has_metadata_csum(sbi->s_sb)) {
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002168 /* Use new metadata_csum algorithm */
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002169 __u32 csum32;
Daeho Jeongb47820e2016-07-03 17:51:39 -04002170 __u16 dummy_csum = 0;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002171
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002172 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2173 sizeof(le_group));
Daeho Jeongb47820e2016-07-03 17:51:39 -04002174 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2175 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2176 sizeof(dummy_csum));
2177 offset += sizeof(dummy_csum);
2178 if (offset < sbi->s_desc_size)
2179 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2180 sbi->s_desc_size - offset);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002181
2182 crc = csum32 & 0xFFFF;
2183 goto out;
Andreas Dilger717d50e2007-10-16 18:38:25 -04002184 }
2185
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002186 /* old crc16 code */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002187 if (!ext4_has_feature_gdt_csum(sb))
Darrick J. Wong813d32f2014-10-14 02:35:49 -04002188 return 0;
2189
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002190 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2191 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2192 crc = crc16(crc, (__u8 *)gdp, offset);
2193 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2194 /* for checksum of struct ext4_group_desc do the rest...*/
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002195 if (ext4_has_feature_64bit(sb) &&
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002196 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2197 crc = crc16(crc, (__u8 *)gdp + offset,
2198 le16_to_cpu(sbi->s_es->s_desc_size) -
2199 offset);
2200
2201out:
Andreas Dilger717d50e2007-10-16 18:38:25 -04002202 return cpu_to_le16(crc);
2203}
2204
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002205int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
Andreas Dilger717d50e2007-10-16 18:38:25 -04002206 struct ext4_group_desc *gdp)
2207{
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002208 if (ext4_has_group_desc_csum(sb) &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002209 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
Andreas Dilger717d50e2007-10-16 18:38:25 -04002210 return 0;
2211
2212 return 1;
2213}
2214
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002215void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2216 struct ext4_group_desc *gdp)
2217{
2218 if (!ext4_has_group_desc_csum(sb))
2219 return;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002220 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002221}
2222
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002223/* Called at mount-time, super-block is locked */
Lukas Czernerbfff6872010-10-27 21:30:05 -04002224static int ext4_check_descriptors(struct super_block *sb,
Theodore Ts'o829fa702016-08-01 00:51:02 -04002225 ext4_fsblk_t sb_block,
Lukas Czernerbfff6872010-10-27 21:30:05 -04002226 ext4_group_t *first_not_zeroed)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002227{
Mingming Cao617ba132006-10-11 01:20:53 -07002228 struct ext4_sb_info *sbi = EXT4_SB(sb);
2229 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2230 ext4_fsblk_t last_block;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002231 ext4_fsblk_t block_bitmap;
2232 ext4_fsblk_t inode_bitmap;
2233 ext4_fsblk_t inode_table;
Jose R. Santosce421582007-10-16 18:38:25 -04002234 int flexbg_flag = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002235 ext4_group_t i, grp = sbi->s_groups_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002236
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002237 if (ext4_has_feature_flex_bg(sb))
Jose R. Santosce421582007-10-16 18:38:25 -04002238 flexbg_flag = 1;
2239
Theodore Ts'oaf5bc922008-09-08 22:25:24 -04002240 ext4_debug("Checking group descriptors");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002241
Akinobu Mita197cd65a2008-02-06 01:40:16 -08002242 for (i = 0; i < sbi->s_groups_count; i++) {
2243 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2244
Jose R. Santosce421582007-10-16 18:38:25 -04002245 if (i == sbi->s_groups_count - 1 || flexbg_flag)
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002246 last_block = ext4_blocks_count(sbi->s_es) - 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002247 else
2248 last_block = first_block +
Mingming Cao617ba132006-10-11 01:20:53 -07002249 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002250
Lukas Czernerbfff6872010-10-27 21:30:05 -04002251 if ((grp == sbi->s_groups_count) &&
2252 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2253 grp = i;
2254
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002255 block_bitmap = ext4_block_bitmap(sb, gdp);
Theodore Ts'o829fa702016-08-01 00:51:02 -04002256 if (block_bitmap == sb_block) {
2257 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2258 "Block bitmap for group %u overlaps "
2259 "superblock", i);
2260 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002261 if (block_bitmap < first_block || block_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002262 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002263 "Block bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002264 "(block %llu)!", i, block_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002265 return 0;
2266 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002267 inode_bitmap = ext4_inode_bitmap(sb, gdp);
Theodore Ts'o829fa702016-08-01 00:51:02 -04002268 if (inode_bitmap == sb_block) {
2269 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2270 "Inode bitmap for group %u overlaps "
2271 "superblock", i);
2272 }
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002273 if (inode_bitmap < first_block || inode_bitmap > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002274 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002275 "Inode bitmap for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002276 "(block %llu)!", i, inode_bitmap);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002277 return 0;
2278 }
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07002279 inode_table = ext4_inode_table(sb, gdp);
Theodore Ts'o829fa702016-08-01 00:51:02 -04002280 if (inode_table == sb_block) {
2281 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2282 "Inode table for group %u overlaps "
2283 "superblock", i);
2284 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07002285 if (inode_table < first_block ||
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002286 inode_table + sbi->s_itb_per_group - 1 > last_block) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
Theodore Ts'oa9df9a42009-01-05 22:18:16 -05002288 "Inode table for group %u not in group "
Eric Sandeenb31e1552009-06-04 17:36:36 -04002289 "(block %llu)!", i, inode_table);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002290 return 0;
2291 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002292 ext4_lock_group(sb, i);
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04002293 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2295 "Checksum for group %u failed (%u!=%u)",
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002296 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
Eric Sandeenb31e1552009-06-04 17:36:36 -04002297 gdp)), le16_to_cpu(gdp->bg_checksum));
Li Zefan7ee1ec42008-09-08 10:47:19 -04002298 if (!(sb->s_flags & MS_RDONLY)) {
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002299 ext4_unlock_group(sb, i);
Theodore Ts'o8a266462008-07-26 14:34:21 -04002300 return 0;
Li Zefan7ee1ec42008-09-08 10:47:19 -04002301 }
Andreas Dilger717d50e2007-10-16 18:38:25 -04002302 }
Aneesh Kumar K.V955ce5f2009-05-02 20:35:09 -04002303 ext4_unlock_group(sb, i);
Jose R. Santosce421582007-10-16 18:38:25 -04002304 if (!flexbg_flag)
2305 first_block += EXT4_BLOCKS_PER_GROUP(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002306 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002307 if (NULL != first_not_zeroed)
2308 *first_not_zeroed = grp;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002309 return 1;
2310}
2311
Mingming Cao617ba132006-10-11 01:20:53 -07002312/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002313 * the superblock) which were deleted from all directories, but held open by
2314 * a process at the time of a crash. We walk the list and try to delete these
2315 * inodes at recovery time (only with a read-write filesystem).
2316 *
2317 * In order to keep the orphan inode chain consistent during traversal (in
2318 * case of crash during recovery), we link each inode into the superblock
2319 * orphan list_head and handle it the same way as an inode deletion during
2320 * normal operation (which journals the operations for us).
2321 *
2322 * We only do an iget() and an iput() on each inode, which is very safe if we
2323 * accidentally point at an in-use or already deleted inode. The worst that
2324 * can happen in this case is that we get a "bit already cleared" message from
Mingming Cao617ba132006-10-11 01:20:53 -07002325 * ext4_free_inode(). The only reason we would point at a wrong inode is if
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002326 * e2fsck was run on this filesystem, and it must have already done the orphan
2327 * inode cleanup for us, so we can safely abort without any further action.
2328 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002329static void ext4_orphan_cleanup(struct super_block *sb,
2330 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002331{
2332 unsigned int s_flags = sb->s_flags;
2333 int nr_orphans = 0, nr_truncates = 0;
2334#ifdef CONFIG_QUOTA
2335 int i;
2336#endif
2337 if (!es->s_last_orphan) {
2338 jbd_debug(4, "no orphan inodes to clean up\n");
2339 return;
2340 }
2341
Eric Sandeena8f48a92006-12-06 20:40:13 -08002342 if (bdev_read_only(sb->s_bdev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002343 ext4_msg(sb, KERN_ERR, "write access "
2344 "unavailable, skipping orphan cleanup");
Eric Sandeena8f48a92006-12-06 20:40:13 -08002345 return;
2346 }
2347
Amir Goldsteind39195c2011-02-28 00:53:45 -05002348 /* Check if feature set would not allow a r/w mount */
2349 if (!ext4_feature_set_ok(sb, 0)) {
2350 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2351 "unknown ROCOMPAT features");
2352 return;
2353 }
2354
Mingming Cao617ba132006-10-11 01:20:53 -07002355 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002356 /* don't clear list on RO mount w/ errors */
2357 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
Dmitry Monakhov84474972014-09-16 14:52:03 -04002358 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002359 "clearing orphan list.\n");
Eric Sandeenc25f9bc2012-09-26 23:30:12 -04002360 es->s_last_orphan = 0;
2361 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002362 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2363 return;
2364 }
2365
2366 if (s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04002367 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002368 sb->s_flags &= ~MS_RDONLY;
2369 }
2370#ifdef CONFIG_QUOTA
2371 /* Needed for iput() to work correctly and not trash data */
2372 sb->s_flags |= MS_ACTIVE;
2373 /* Turn on quotas so that they are updated correctly */
Jan Karaa2d4a642014-09-11 11:15:15 -04002374 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Mingming Cao617ba132006-10-11 01:20:53 -07002375 if (EXT4_SB(sb)->s_qf_names[i]) {
2376 int ret = ext4_quota_on_mount(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002377 if (ret < 0)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002378 ext4_msg(sb, KERN_ERR,
2379 "Cannot turn on journaled "
2380 "quota: error %d", ret);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002381 }
2382 }
2383#endif
2384
2385 while (es->s_last_orphan) {
2386 struct inode *inode;
2387
Vegard Nossumc65d5c62016-07-14 23:21:35 -04002388 /*
2389 * We may have encountered an error during cleanup; if
2390 * so, skip the rest.
2391 */
2392 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2393 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2394 es->s_last_orphan = 0;
2395 break;
2396 }
2397
Josef Bacik97bd42b2008-04-29 22:04:56 -04002398 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2399 if (IS_ERR(inode)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002400 es->s_last_orphan = 0;
2401 break;
2402 }
2403
Mingming Cao617ba132006-10-11 01:20:53 -07002404 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
Christoph Hellwig871a2932010-03-03 09:05:07 -05002405 dquot_initialize(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002406 if (inode->i_nlink) {
Paul Taysom566370a2013-05-28 07:51:21 -04002407 if (test_opt(sb, DEBUG))
2408 ext4_msg(sb, KERN_DEBUG,
2409 "%s: truncating inode %lu to %lld bytes",
2410 __func__, inode->i_ino, inode->i_size);
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04002411 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002412 inode->i_ino, inode->i_size);
Al Viro59551022016-01-22 15:40:57 -05002413 inode_lock(inode);
Lukas Czerner55f252c2013-05-27 23:32:35 -04002414 truncate_inode_pages(inode->i_mapping, inode->i_size);
Mingming Cao617ba132006-10-11 01:20:53 -07002415 ext4_truncate(inode);
Al Viro59551022016-01-22 15:40:57 -05002416 inode_unlock(inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002417 nr_truncates++;
2418 } else {
Paul Taysom566370a2013-05-28 07:51:21 -04002419 if (test_opt(sb, DEBUG))
2420 ext4_msg(sb, KERN_DEBUG,
2421 "%s: deleting unreferenced inode %lu",
2422 __func__, inode->i_ino);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002423 jbd_debug(2, "deleting unreferenced inode %lu\n",
2424 inode->i_ino);
2425 nr_orphans++;
2426 }
2427 iput(inode); /* The delete magic happens here! */
2428 }
2429
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04002430#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002431
2432 if (nr_orphans)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002433 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2434 PLURAL(nr_orphans));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002435 if (nr_truncates)
Eric Sandeenb31e1552009-06-04 17:36:36 -04002436 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2437 PLURAL(nr_truncates));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002438#ifdef CONFIG_QUOTA
2439 /* Turn quotas off */
Jan Karaa2d4a642014-09-11 11:15:15 -04002440 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002441 if (sb_dqopt(sb)->files[i])
Christoph Hellwig287a8092010-05-19 07:16:45 -04002442 dquot_quota_off(sb, i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002443 }
2444#endif
2445 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2446}
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002447
Eric Sandeencd2291a2008-01-28 23:58:27 -05002448/*
2449 * Maximal extent format file size.
2450 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2451 * extent format containers, within a sector_t, and within i_blocks
2452 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2453 * so that won't be a limiting factor.
2454 *
Lukas Czernerf17722f2011-06-06 00:05:17 -04002455 * However there is other limiting factor. We do store extents in the form
2456 * of starting block and length, hence the resulting length of the extent
2457 * covering maximum file size must fit into on-disk format containers as
2458 * well. Given that length is always by 1 unit bigger than max unit (because
2459 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2460 *
Eric Sandeencd2291a2008-01-28 23:58:27 -05002461 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2462 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002463static loff_t ext4_max_size(int blkbits, int has_huge_files)
Eric Sandeencd2291a2008-01-28 23:58:27 -05002464{
2465 loff_t res;
2466 loff_t upper_limit = MAX_LFS_FILESIZE;
2467
2468 /* small i_blocks in vfs inode? */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002469 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Eric Sandeencd2291a2008-01-28 23:58:27 -05002470 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002471 * CONFIG_LBDAF is not enabled implies the inode
Eric Sandeencd2291a2008-01-28 23:58:27 -05002472 * i_block represent total blocks in 512 bytes
2473 * 32 == size of vfs inode i_blocks * 8
2474 */
2475 upper_limit = (1LL << 32) - 1;
2476
2477 /* total blocks in file system block size */
2478 upper_limit >>= (blkbits - 9);
2479 upper_limit <<= blkbits;
2480 }
2481
Lukas Czernerf17722f2011-06-06 00:05:17 -04002482 /*
2483 * 32-bit extent-start container, ee_block. We lower the maxbytes
2484 * by one fs block, so ee_len can cover the extent of maximum file
2485 * size
2486 */
2487 res = (1LL << 32) - 1;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002488 res <<= blkbits;
Eric Sandeencd2291a2008-01-28 23:58:27 -05002489
2490 /* Sanity check against vm- & vfs- imposed limits */
2491 if (res > upper_limit)
2492 res = upper_limit;
2493
2494 return res;
2495}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002496
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002497/*
Eric Sandeencd2291a2008-01-28 23:58:27 -05002498 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002499 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2500 * We need to be 1 filesystem block less than the 2^48 sector limit.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002501 */
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002502static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002503{
Mingming Cao617ba132006-10-11 01:20:53 -07002504 loff_t res = EXT4_NDIR_BLOCKS;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002505 int meta_blocks;
2506 loff_t upper_limit;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002507 /* This is calculated to be the largest file size for a dense, block
2508 * mapped file such that the file's total number of 512-byte sectors,
2509 * including data and all indirect blocks, does not exceed (2^48 - 1).
2510 *
2511 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2512 * number of 512-byte sectors of the file.
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002513 */
2514
Theodore Ts'of287a1a2008-10-16 22:50:48 -04002515 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002516 /*
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02002517 * !has_huge_files or CONFIG_LBDAF not enabled implies that
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002518 * the inode i_block field represents total file blocks in
2519 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002520 */
2521 upper_limit = (1LL << 32) - 1;
2522
2523 /* total blocks in file system block size */
2524 upper_limit >>= (bits - 9);
2525
2526 } else {
Aneesh Kumar K.V8180a562008-01-28 23:58:27 -05002527 /*
2528 * We use 48 bit ext4_inode i_blocks
2529 * With EXT4_HUGE_FILE_FL set the i_blocks
2530 * represent total number of blocks in
2531 * file system block size
2532 */
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002533 upper_limit = (1LL << 48) - 1;
2534
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002535 }
2536
2537 /* indirect blocks */
2538 meta_blocks = 1;
2539 /* double indirect blocks */
2540 meta_blocks += 1 + (1LL << (bits-2));
2541 /* tripple indirect blocks */
2542 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2543
2544 upper_limit -= meta_blocks;
2545 upper_limit <<= bits;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002546
2547 res += 1LL << (bits-2);
2548 res += 1LL << (2*(bits-2));
2549 res += 1LL << (3*(bits-2));
2550 res <<= bits;
2551 if (res > upper_limit)
2552 res = upper_limit;
Aneesh Kumar K.V0fc1b452008-01-28 23:58:26 -05002553
2554 if (res > MAX_LFS_FILESIZE)
2555 res = MAX_LFS_FILESIZE;
2556
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002557 return res;
2558}
2559
Mingming Cao617ba132006-10-11 01:20:53 -07002560static ext4_fsblk_t descriptor_loc(struct super_block *sb,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002561 ext4_fsblk_t logical_sb_block, int nr)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002562{
Mingming Cao617ba132006-10-11 01:20:53 -07002563 struct ext4_sb_info *sbi = EXT4_SB(sb);
Avantika Mathurfd2d4292008-01-28 23:58:27 -05002564 ext4_group_t bg, first_meta_bg;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002565 int has_super = 0;
2566
2567 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2568
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002569 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
Andrew Morton70bbb3e2006-10-11 01:21:20 -07002570 return logical_sb_block + nr + 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002571 bg = sbi->s_desc_per_block * nr;
Mingming Cao617ba132006-10-11 01:20:53 -07002572 if (ext4_bg_has_super(sb, bg))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002573 has_super = 1;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04002574
Darrick J. Wongbd63f6b2014-05-12 10:06:27 -04002575 /*
2576 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2577 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2578 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2579 * compensate.
2580 */
2581 if (sb->s_blocksize == 1024 && nr == 0 &&
2582 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2583 has_super++;
2584
Mingming Cao617ba132006-10-11 01:20:53 -07002585 return (has_super + ext4_group_first_block_no(sb, bg));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002586}
2587
Alex Tomasc9de5602008-01-29 00:19:52 -05002588/**
2589 * ext4_get_stripe_size: Get the stripe size.
2590 * @sbi: In memory super block info
2591 *
2592 * If we have specified it via mount option, then
2593 * use the mount option value. If the value specified at mount time is
2594 * greater than the blocks per group use the super block value.
2595 * If the super block value is greater than blocks per group return 0.
2596 * Allocator needs it be less than blocks per group.
2597 *
2598 */
2599static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2600{
2601 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2602 unsigned long stripe_width =
2603 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002604 int ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002605
2606 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002607 ret = sbi->s_stripe;
2608 else if (stripe_width <= sbi->s_blocks_per_group)
2609 ret = stripe_width;
2610 else if (stride <= sbi->s_blocks_per_group)
2611 ret = stride;
2612 else
2613 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002614
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002615 /*
2616 * If the stripe width is 1, this makes no sense and
2617 * we set it to 0 to turn off stripe handling code.
2618 */
2619 if (ret <= 1)
2620 ret = 0;
Alex Tomasc9de5602008-01-29 00:19:52 -05002621
Dan Ehrenberg3eb08652011-07-17 21:18:51 -04002622 return ret;
Alex Tomasc9de5602008-01-29 00:19:52 -05002623}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07002624
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002625/*
2626 * Check whether this filesystem can be mounted based on
2627 * the features present and the RDONLY/RDWR mount requested.
2628 * Returns 1 if this filesystem can be mounted as requested,
2629 * 0 if it cannot be.
2630 */
2631static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2632{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002633 if (ext4_has_unknown_ext4_incompat_features(sb)) {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002634 ext4_msg(sb, KERN_ERR,
2635 "Couldn't mount because of "
2636 "unsupported optional features (%x)",
2637 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2638 ~EXT4_FEATURE_INCOMPAT_SUPP));
2639 return 0;
2640 }
2641
2642 if (readonly)
2643 return 1;
2644
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002645 if (ext4_has_feature_readonly(sb)) {
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05002646 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2647 sb->s_flags |= MS_RDONLY;
2648 return 1;
2649 }
2650
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002651 /* Check that feature set is OK for a read-write mount */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002652 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002653 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2654 "unsupported optional features (%x)",
2655 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2656 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2657 return 0;
2658 }
2659 /*
2660 * Large file size enabled file system can only be mounted
2661 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2662 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002663 if (ext4_has_feature_huge_file(sb)) {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002664 if (sizeof(blkcnt_t) < sizeof(u64)) {
2665 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2666 "cannot be mounted RDWR without "
2667 "CONFIG_LBDAF");
2668 return 0;
2669 }
2670 }
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002671 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
Theodore Ts'obab08ab2011-09-09 18:36:51 -04002672 ext4_msg(sb, KERN_ERR,
2673 "Can't support bigalloc feature without "
2674 "extents feature\n");
2675 return 0;
2676 }
Aditya Kali7c319d32012-07-22 20:21:31 -04002677
2678#ifndef CONFIG_QUOTA
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04002679 if (ext4_has_feature_quota(sb) && !readonly) {
Aditya Kali7c319d32012-07-22 20:21:31 -04002680 ext4_msg(sb, KERN_ERR,
2681 "Filesystem with quota feature cannot be mounted RDWR "
2682 "without CONFIG_QUOTA");
2683 return 0;
2684 }
Li Xi689c9582016-01-08 16:01:22 -05002685 if (ext4_has_feature_project(sb) && !readonly) {
2686 ext4_msg(sb, KERN_ERR,
2687 "Filesystem with project quota feature cannot be mounted RDWR "
2688 "without CONFIG_QUOTA");
2689 return 0;
2690 }
Aditya Kali7c319d32012-07-22 20:21:31 -04002691#endif /* CONFIG_QUOTA */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04002692 return 1;
2693}
2694
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002695/*
2696 * This function is called once a day if we have errors logged
2697 * on the file system
2698 */
2699static void print_daily_error_info(unsigned long arg)
2700{
2701 struct super_block *sb = (struct super_block *) arg;
2702 struct ext4_sb_info *sbi;
2703 struct ext4_super_block *es;
2704
2705 sbi = EXT4_SB(sb);
2706 es = sbi->s_es;
2707
2708 if (es->s_error_count)
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002709 /* fsck newer than v1.41.13 is needed to clean this condition. */
2710 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002711 le32_to_cpu(es->s_error_count));
2712 if (es->s_first_error_time) {
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002713 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002714 sb->s_id, le32_to_cpu(es->s_first_error_time),
2715 (int) sizeof(es->s_first_error_func),
2716 es->s_first_error_func,
2717 le32_to_cpu(es->s_first_error_line));
2718 if (es->s_first_error_ino)
Joe Perches651e1c32016-10-12 23:12:53 -04002719 printk(KERN_CONT ": inode %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002720 le32_to_cpu(es->s_first_error_ino));
2721 if (es->s_first_error_block)
Joe Perches651e1c32016-10-12 23:12:53 -04002722 printk(KERN_CONT ": block %llu", (unsigned long long)
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002723 le64_to_cpu(es->s_first_error_block));
Joe Perches651e1c32016-10-12 23:12:53 -04002724 printk(KERN_CONT "\n");
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002725 }
2726 if (es->s_last_error_time) {
Theodore Ts'oae0f78d2014-07-05 18:40:52 -04002727 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002728 sb->s_id, le32_to_cpu(es->s_last_error_time),
2729 (int) sizeof(es->s_last_error_func),
2730 es->s_last_error_func,
2731 le32_to_cpu(es->s_last_error_line));
2732 if (es->s_last_error_ino)
Joe Perches651e1c32016-10-12 23:12:53 -04002733 printk(KERN_CONT ": inode %u",
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002734 le32_to_cpu(es->s_last_error_ino));
2735 if (es->s_last_error_block)
Joe Perches651e1c32016-10-12 23:12:53 -04002736 printk(KERN_CONT ": block %llu", (unsigned long long)
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002737 le64_to_cpu(es->s_last_error_block));
Joe Perches651e1c32016-10-12 23:12:53 -04002738 printk(KERN_CONT "\n");
Theodore Ts'o66e61a92010-07-27 11:56:04 -04002739 }
2740 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2741}
2742
Lukas Czernerbfff6872010-10-27 21:30:05 -04002743/* Find next suitable group and run ext4_init_inode_table */
2744static int ext4_run_li_request(struct ext4_li_request *elr)
2745{
2746 struct ext4_group_desc *gdp = NULL;
2747 ext4_group_t group, ngroups;
2748 struct super_block *sb;
2749 unsigned long timeout = 0;
2750 int ret = 0;
2751
2752 sb = elr->lr_super;
2753 ngroups = EXT4_SB(sb)->s_groups_count;
2754
2755 for (group = elr->lr_next_group; group < ngroups; group++) {
2756 gdp = ext4_get_group_desc(sb, group, NULL);
2757 if (!gdp) {
2758 ret = 1;
2759 break;
2760 }
2761
2762 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2763 break;
2764 }
2765
Theodore Ts'o7f511862013-01-13 08:41:45 -05002766 if (group >= ngroups)
Lukas Czernerbfff6872010-10-27 21:30:05 -04002767 ret = 1;
2768
2769 if (!ret) {
2770 timeout = jiffies;
2771 ret = ext4_init_inode_table(sb, group,
2772 elr->lr_timeout ? 0 : 1);
2773 if (elr->lr_timeout == 0) {
Lukas Czerner51ce6512011-05-20 13:55:16 -04002774 timeout = (jiffies - timeout) *
2775 elr->lr_sbi->s_li_wait_mult;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002776 elr->lr_timeout = timeout;
2777 }
2778 elr->lr_next_sched = jiffies + elr->lr_timeout;
2779 elr->lr_next_group = group + 1;
2780 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002781 return ret;
2782}
2783
2784/*
2785 * Remove lr_request from the list_request and free the
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002786 * request structure. Should be called with li_list_mtx held
Lukas Czernerbfff6872010-10-27 21:30:05 -04002787 */
2788static void ext4_remove_li_request(struct ext4_li_request *elr)
2789{
2790 struct ext4_sb_info *sbi;
2791
2792 if (!elr)
2793 return;
2794
2795 sbi = elr->lr_sbi;
2796
2797 list_del(&elr->lr_request);
2798 sbi->s_li_request = NULL;
2799 kfree(elr);
2800}
2801
2802static void ext4_unregister_li_request(struct super_block *sb)
2803{
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002804 mutex_lock(&ext4_li_mtx);
2805 if (!ext4_li_info) {
2806 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002807 return;
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002808 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002809
2810 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002811 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002812 mutex_unlock(&ext4_li_info->li_list_mtx);
Lukas Czerner1bb933f2011-05-20 13:55:29 -04002813 mutex_unlock(&ext4_li_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002814}
2815
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002816static struct task_struct *ext4_lazyinit_task;
2817
Lukas Czernerbfff6872010-10-27 21:30:05 -04002818/*
2819 * This is the function where ext4lazyinit thread lives. It walks
2820 * through the request list searching for next scheduled filesystem.
2821 * When such a fs is found, run the lazy initialization request
2822 * (ext4_rn_li_request) and keep track of the time spend in this
2823 * function. Based on that time we compute next schedule time of
2824 * the request. When walking through the list is complete, compute
2825 * next waking time and put itself into sleep.
2826 */
2827static int ext4_lazyinit_thread(void *arg)
2828{
2829 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2830 struct list_head *pos, *n;
2831 struct ext4_li_request *elr;
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002832 unsigned long next_wakeup, cur;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002833
2834 BUG_ON(NULL == eli);
2835
Lukas Czernerbfff6872010-10-27 21:30:05 -04002836cont_thread:
2837 while (true) {
2838 next_wakeup = MAX_JIFFY_OFFSET;
2839
2840 mutex_lock(&eli->li_list_mtx);
2841 if (list_empty(&eli->li_request_list)) {
2842 mutex_unlock(&eli->li_list_mtx);
2843 goto exit_thread;
2844 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002845 list_for_each_safe(pos, n, &eli->li_request_list) {
Dmitry Monakhove22834f2016-09-05 23:38:36 -04002846 int err = 0;
2847 int progress = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002848 elr = list_entry(pos, struct ext4_li_request,
2849 lr_request);
2850
Dmitry Monakhove22834f2016-09-05 23:38:36 -04002851 if (time_before(jiffies, elr->lr_next_sched)) {
2852 if (time_before(elr->lr_next_sched, next_wakeup))
2853 next_wakeup = elr->lr_next_sched;
2854 continue;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002855 }
Dmitry Monakhove22834f2016-09-05 23:38:36 -04002856 if (down_read_trylock(&elr->lr_super->s_umount)) {
2857 if (sb_start_write_trylock(elr->lr_super)) {
2858 progress = 1;
2859 /*
2860 * We hold sb->s_umount, sb can not
2861 * be removed from the list, it is
2862 * now safe to drop li_list_mtx
2863 */
2864 mutex_unlock(&eli->li_list_mtx);
2865 err = ext4_run_li_request(elr);
2866 sb_end_write(elr->lr_super);
2867 mutex_lock(&eli->li_list_mtx);
2868 n = pos->next;
2869 }
2870 up_read((&elr->lr_super->s_umount));
2871 }
2872 /* error, remove the lazy_init job */
2873 if (err) {
2874 ext4_remove_li_request(elr);
2875 continue;
2876 }
2877 if (!progress) {
2878 elr->lr_next_sched = jiffies +
2879 (prandom_u32()
2880 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2881 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002882 if (time_before(elr->lr_next_sched, next_wakeup))
2883 next_wakeup = elr->lr_next_sched;
2884 }
2885 mutex_unlock(&eli->li_list_mtx);
2886
Tejun Heoa0acae02011-11-21 12:32:22 -08002887 try_to_freeze();
Lukas Czernerbfff6872010-10-27 21:30:05 -04002888
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002889 cur = jiffies;
2890 if ((time_after_eq(cur, next_wakeup)) ||
Lukas Czernerf4245bd2010-11-02 14:07:17 -04002891 (MAX_JIFFY_OFFSET == next_wakeup)) {
Lukas Czernerbfff6872010-10-27 21:30:05 -04002892 cond_resched();
2893 continue;
2894 }
2895
Lukas Czerner4ed5c032011-05-20 13:49:04 -04002896 schedule_timeout_interruptible(next_wakeup - cur);
2897
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002898 if (kthread_should_stop()) {
2899 ext4_clear_request_list();
2900 goto exit_thread;
2901 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04002902 }
2903
2904exit_thread:
2905 /*
2906 * It looks like the request list is empty, but we need
2907 * to check it under the li_list_mtx lock, to prevent any
2908 * additions into it, and of course we should lock ext4_li_mtx
2909 * to atomically free the list and ext4_li_info, because at
2910 * this point another ext4 filesystem could be registering
2911 * new one.
2912 */
2913 mutex_lock(&ext4_li_mtx);
2914 mutex_lock(&eli->li_list_mtx);
2915 if (!list_empty(&eli->li_request_list)) {
2916 mutex_unlock(&eli->li_list_mtx);
2917 mutex_unlock(&ext4_li_mtx);
2918 goto cont_thread;
2919 }
2920 mutex_unlock(&eli->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002921 kfree(ext4_li_info);
2922 ext4_li_info = NULL;
2923 mutex_unlock(&ext4_li_mtx);
2924
2925 return 0;
2926}
2927
2928static void ext4_clear_request_list(void)
2929{
2930 struct list_head *pos, *n;
2931 struct ext4_li_request *elr;
2932
2933 mutex_lock(&ext4_li_info->li_list_mtx);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002934 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2935 elr = list_entry(pos, struct ext4_li_request,
2936 lr_request);
2937 ext4_remove_li_request(elr);
2938 }
2939 mutex_unlock(&ext4_li_info->li_list_mtx);
2940}
2941
2942static int ext4_run_lazyinit_thread(void)
2943{
Eric Sandeen8f1f7452011-02-03 14:33:15 -05002944 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2945 ext4_li_info, "ext4lazyinit");
2946 if (IS_ERR(ext4_lazyinit_task)) {
2947 int err = PTR_ERR(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04002948 ext4_clear_request_list();
Lukas Czernerbfff6872010-10-27 21:30:05 -04002949 kfree(ext4_li_info);
2950 ext4_li_info = NULL;
Theodore Ts'o92b97812012-03-19 23:41:49 -04002951 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
Lukas Czernerbfff6872010-10-27 21:30:05 -04002952 "initialization thread\n",
2953 err);
2954 return err;
2955 }
2956 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
Lukas Czernerbfff6872010-10-27 21:30:05 -04002957 return 0;
2958}
2959
2960/*
2961 * Check whether it make sense to run itable init. thread or not.
2962 * If there is at least one uninitialized inode table, return
2963 * corresponding group number, else the loop goes through all
2964 * groups and return total number of groups.
2965 */
2966static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2967{
2968 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2969 struct ext4_group_desc *gdp = NULL;
2970
2971 for (group = 0; group < ngroups; group++) {
2972 gdp = ext4_get_group_desc(sb, group, NULL);
2973 if (!gdp)
2974 continue;
2975
2976 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2977 break;
2978 }
2979
2980 return group;
2981}
2982
2983static int ext4_li_info_new(void)
2984{
2985 struct ext4_lazy_init *eli = NULL;
2986
2987 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2988 if (!eli)
2989 return -ENOMEM;
2990
Lukas Czernerbfff6872010-10-27 21:30:05 -04002991 INIT_LIST_HEAD(&eli->li_request_list);
2992 mutex_init(&eli->li_list_mtx);
2993
Lukas Czernerbfff6872010-10-27 21:30:05 -04002994 eli->li_state |= EXT4_LAZYINIT_QUIT;
2995
2996 ext4_li_info = eli;
2997
2998 return 0;
2999}
3000
3001static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3002 ext4_group_t start)
3003{
3004 struct ext4_sb_info *sbi = EXT4_SB(sb);
3005 struct ext4_li_request *elr;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003006
3007 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3008 if (!elr)
3009 return NULL;
3010
3011 elr->lr_super = sb;
3012 elr->lr_sbi = sbi;
3013 elr->lr_next_group = start;
3014
3015 /*
3016 * Randomize first schedule time of the request to
3017 * spread the inode table initialization requests
3018 * better.
3019 */
Theodore Ts'odd1f7232013-11-08 00:14:53 -05003020 elr->lr_next_sched = jiffies + (prandom_u32() %
3021 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
Lukas Czernerbfff6872010-10-27 21:30:05 -04003022 return elr;
3023}
3024
Theodore Ts'o7f511862013-01-13 08:41:45 -05003025int ext4_register_li_request(struct super_block *sb,
3026 ext4_group_t first_not_zeroed)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003027{
3028 struct ext4_sb_info *sbi = EXT4_SB(sb);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003029 struct ext4_li_request *elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003030 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
Andrew Morton6c5a6cb2011-01-10 12:30:17 -05003031 int ret = 0;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003032
Theodore Ts'o7f511862013-01-13 08:41:45 -05003033 mutex_lock(&ext4_li_mtx);
Lukas Czerner51ce6512011-05-20 13:55:16 -04003034 if (sbi->s_li_request != NULL) {
3035 /*
3036 * Reset timeout so it can be computed again, because
3037 * s_li_wait_mult might have changed.
3038 */
3039 sbi->s_li_request->lr_timeout = 0;
Theodore Ts'o7f511862013-01-13 08:41:45 -05003040 goto out;
Lukas Czerner51ce6512011-05-20 13:55:16 -04003041 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003042
3043 if (first_not_zeroed == ngroups ||
3044 (sb->s_flags & MS_RDONLY) ||
Tao Ma55ff3842011-05-09 10:28:41 -04003045 !test_opt(sb, INIT_INODE_TABLE))
Theodore Ts'o7f511862013-01-13 08:41:45 -05003046 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003047
3048 elr = ext4_li_request_new(sb, first_not_zeroed);
Theodore Ts'o7f511862013-01-13 08:41:45 -05003049 if (!elr) {
3050 ret = -ENOMEM;
3051 goto out;
3052 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003053
3054 if (NULL == ext4_li_info) {
3055 ret = ext4_li_info_new();
3056 if (ret)
3057 goto out;
3058 }
3059
3060 mutex_lock(&ext4_li_info->li_list_mtx);
3061 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3062 mutex_unlock(&ext4_li_info->li_list_mtx);
3063
3064 sbi->s_li_request = elr;
Tao Ma46e46902011-04-04 16:00:49 -04003065 /*
3066 * set elr to NULL here since it has been inserted to
3067 * the request_list and the removal and free of it is
3068 * handled by ext4_clear_request_list from now on.
3069 */
3070 elr = NULL;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003071
3072 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3073 ret = ext4_run_lazyinit_thread();
3074 if (ret)
3075 goto out;
3076 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04003077out:
Nicolas Kaiserbeed5ecb2010-10-27 22:08:42 -04003078 mutex_unlock(&ext4_li_mtx);
3079 if (ret)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003080 kfree(elr);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003081 return ret;
3082}
3083
3084/*
3085 * We do not need to lock anything since this is called on
3086 * module unload.
3087 */
3088static void ext4_destroy_lazyinit_thread(void)
3089{
3090 /*
3091 * If thread exited earlier
3092 * there's nothing to be done.
3093 */
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003094 if (!ext4_li_info || !ext4_lazyinit_task)
Lukas Czernerbfff6872010-10-27 21:30:05 -04003095 return;
3096
Eric Sandeen8f1f7452011-02-03 14:33:15 -05003097 kthread_stop(ext4_lazyinit_task);
Lukas Czernerbfff6872010-10-27 21:30:05 -04003098}
3099
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003100static int set_journal_csum_feature_set(struct super_block *sb)
3101{
3102 int ret = 1;
3103 int compat, incompat;
3104 struct ext4_sb_info *sbi = EXT4_SB(sb);
3105
Dmitry Monakhov9aa5d322014-10-13 03:36:16 -04003106 if (ext4_has_metadata_csum(sb)) {
Darrick J. Wongdb9ee222014-08-27 18:40:07 -04003107 /* journal checksum v3 */
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003108 compat = 0;
Darrick J. Wongdb9ee222014-08-27 18:40:07 -04003109 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003110 } else {
3111 /* journal checksum v1 */
3112 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3113 incompat = 0;
3114 }
3115
Darrick J. Wongfeb8c6d2014-09-11 11:38:21 -04003116 jbd2_journal_clear_features(sbi->s_journal,
3117 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3118 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3119 JBD2_FEATURE_INCOMPAT_CSUM_V2);
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003120 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3121 ret = jbd2_journal_set_features(sbi->s_journal,
3122 compat, 0,
3123 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3124 incompat);
3125 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3126 ret = jbd2_journal_set_features(sbi->s_journal,
3127 compat, 0,
3128 incompat);
3129 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3130 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3131 } else {
Darrick J. Wongfeb8c6d2014-09-11 11:38:21 -04003132 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3133 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003134 }
3135
3136 return ret;
3137}
3138
Theodore Ts'o952fc182012-07-09 16:27:05 -04003139/*
3140 * Note: calculating the overhead so we can be compatible with
3141 * historical BSD practice is quite difficult in the face of
3142 * clusters/bigalloc. This is because multiple metadata blocks from
3143 * different block group can end up in the same allocation cluster.
3144 * Calculating the exact overhead in the face of clustered allocation
3145 * requires either O(all block bitmaps) in memory or O(number of block
3146 * groups**2) in time. We will still calculate the superblock for
3147 * older file systems --- and if we come across with a bigalloc file
3148 * system with zero in s_overhead_clusters the estimate will be close to
3149 * correct especially for very large cluster sizes --- but for newer
3150 * file systems, it's better to calculate this figure once at mkfs
3151 * time, and store it in the superblock. If the superblock value is
3152 * present (even for non-bigalloc file systems), we will use it.
3153 */
3154static int count_overhead(struct super_block *sb, ext4_group_t grp,
3155 char *buf)
3156{
3157 struct ext4_sb_info *sbi = EXT4_SB(sb);
3158 struct ext4_group_desc *gdp;
3159 ext4_fsblk_t first_block, last_block, b;
3160 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3161 int s, j, count = 0;
3162
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003163 if (!ext4_has_feature_bigalloc(sb))
Theodore Ts'o0548bbb2012-08-16 11:59:04 -04003164 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3165 sbi->s_itb_per_group + 2);
3166
Theodore Ts'o952fc182012-07-09 16:27:05 -04003167 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3168 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3169 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3170 for (i = 0; i < ngroups; i++) {
3171 gdp = ext4_get_group_desc(sb, i, NULL);
3172 b = ext4_block_bitmap(sb, gdp);
3173 if (b >= first_block && b <= last_block) {
3174 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3175 count++;
3176 }
3177 b = ext4_inode_bitmap(sb, gdp);
3178 if (b >= first_block && b <= last_block) {
3179 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3180 count++;
3181 }
3182 b = ext4_inode_table(sb, gdp);
3183 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3184 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3185 int c = EXT4_B2C(sbi, b - first_block);
3186 ext4_set_bit(c, buf);
3187 count++;
3188 }
3189 if (i != grp)
3190 continue;
3191 s = 0;
3192 if (ext4_bg_has_super(sb, grp)) {
3193 ext4_set_bit(s++, buf);
3194 count++;
3195 }
3196 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3197 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3198 count++;
3199 }
3200 }
3201 if (!count)
3202 return 0;
3203 return EXT4_CLUSTERS_PER_GROUP(sb) -
3204 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3205}
3206
3207/*
3208 * Compute the overhead and stash it in sbi->s_overhead
3209 */
3210int ext4_calculate_overhead(struct super_block *sb)
3211{
3212 struct ext4_sb_info *sbi = EXT4_SB(sb);
3213 struct ext4_super_block *es = sbi->s_es;
Eric Whitney3c816de2016-09-30 02:08:49 -04003214 struct inode *j_inode;
3215 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
Theodore Ts'o952fc182012-07-09 16:27:05 -04003216 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3217 ext4_fsblk_t overhead = 0;
Dmitry Monakhov4fdb5542014-11-25 13:08:04 -05003218 char *buf = (char *) get_zeroed_page(GFP_NOFS);
Theodore Ts'o952fc182012-07-09 16:27:05 -04003219
Theodore Ts'o952fc182012-07-09 16:27:05 -04003220 if (!buf)
3221 return -ENOMEM;
3222
3223 /*
3224 * Compute the overhead (FS structures). This is constant
3225 * for a given filesystem unless the number of block groups
3226 * changes so we cache the previous value until it does.
3227 */
3228
3229 /*
3230 * All of the blocks before first_data_block are overhead
3231 */
3232 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3233
3234 /*
3235 * Add the overhead found in each block group
3236 */
3237 for (i = 0; i < ngroups; i++) {
3238 int blks;
3239
3240 blks = count_overhead(sb, i, buf);
3241 overhead += blks;
3242 if (blks)
3243 memset(buf, 0, PAGE_SIZE);
3244 cond_resched();
3245 }
Eric Whitney3c816de2016-09-30 02:08:49 -04003246
3247 /*
3248 * Add the internal journal blocks whether the journal has been
3249 * loaded or not
3250 */
Eric Sandeenb003b522014-11-25 16:27:44 -05003251 if (sbi->s_journal && !sbi->journal_bdev)
Lukas Czerner810da242013-03-02 17:18:58 -05003252 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
Eric Whitney3c816de2016-09-30 02:08:49 -04003253 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3254 j_inode = ext4_get_journal_inode(sb, j_inum);
3255 if (j_inode) {
3256 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3257 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3258 iput(j_inode);
3259 } else {
3260 ext4_msg(sb, KERN_ERR, "can't get journal size");
3261 }
3262 }
Theodore Ts'o952fc182012-07-09 16:27:05 -04003263 sbi->s_overhead = overhead;
3264 smp_wmb();
3265 free_page((unsigned long) buf);
3266 return 0;
3267}
3268
Theodore Ts'ob5799012015-09-23 12:44:17 -04003269static void ext4_set_resv_clusters(struct super_block *sb)
Lukas Czerner27dd4382013-04-09 22:11:22 -04003270{
3271 ext4_fsblk_t resv_clusters;
Theodore Ts'ob5799012015-09-23 12:44:17 -04003272 struct ext4_sb_info *sbi = EXT4_SB(sb);
Lukas Czerner27dd4382013-04-09 22:11:22 -04003273
3274 /*
Jan Kara30fac0f2013-12-08 21:11:59 -05003275 * There's no need to reserve anything when we aren't using extents.
3276 * The space estimates are exact, there are no unwritten extents,
3277 * hole punching doesn't need new metadata... This is needed especially
3278 * to keep ext2/3 backward compatibility.
3279 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003280 if (!ext4_has_feature_extents(sb))
Theodore Ts'ob5799012015-09-23 12:44:17 -04003281 return;
Jan Kara30fac0f2013-12-08 21:11:59 -05003282 /*
Lukas Czerner27dd4382013-04-09 22:11:22 -04003283 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3284 * This should cover the situations where we can not afford to run
3285 * out of space like for example punch hole, or converting
Lukas Czerner556615d2014-04-20 23:45:47 -04003286 * unwritten extents in delalloc path. In most cases such
Lukas Czerner27dd4382013-04-09 22:11:22 -04003287 * allocation would require 1, or 2 blocks, higher numbers are
3288 * very rare.
3289 */
Theodore Ts'ob5799012015-09-23 12:44:17 -04003290 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3291 sbi->s_cluster_bits);
Lukas Czerner27dd4382013-04-09 22:11:22 -04003292
3293 do_div(resv_clusters, 50);
3294 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3295
Theodore Ts'ob5799012015-09-23 12:44:17 -04003296 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
Lukas Czerner27dd4382013-04-09 22:11:22 -04003297}
3298
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003299static int ext4_fill_super(struct super_block *sb, void *data, int silent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003300{
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04003301 char *orig_data = kstrdup(data, GFP_KERNEL);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003302 struct buffer_head *bh;
Mingming Cao617ba132006-10-11 01:20:53 -07003303 struct ext4_super_block *es = NULL;
3304 struct ext4_sb_info *sbi;
3305 ext4_fsblk_t block;
3306 ext4_fsblk_t sb_block = get_sb_block(&data);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003307 ext4_fsblk_t logical_sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003308 unsigned long offset = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003309 unsigned long journal_devnum = 0;
3310 unsigned long def_mount_opts;
3311 struct inode *root;
Frank Mayhar03901312009-01-07 00:06:22 -05003312 const char *descr;
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003313 int ret = -ENOMEM;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003314 int blocksize, clustersize;
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003315 unsigned int db_count;
3316 unsigned int i;
Theodore Ts'o281b5992011-09-09 18:34:51 -04003317 int needs_recovery, has_huge_files, has_bigalloc;
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003318 __u64 blocks_count;
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003319 int err = 0;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003320 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Lukas Czernerbfff6872010-10-27 21:30:05 -04003321 ext4_group_t first_not_zeroed;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003322
3323 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3324 if (!sbi)
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003325 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003326
3327 sbi->s_blockgroup_lock =
3328 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3329 if (!sbi->s_blockgroup_lock) {
3330 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003331 goto out_free_orig;
Pekka Enberg705895b2009-02-15 18:07:52 -05003332 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003333 sb->s_fs_info = sbi;
Theodore Ts'o2c0544b2012-05-30 22:56:46 -04003334 sbi->s_sb = sb;
Theodore Ts'o240799c2008-10-09 23:53:47 -04003335 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
Miklos Szeredid9c9bef12007-10-16 23:26:27 -07003336 sbi->s_sb_block = sb_block;
Theodore Ts'of613dfc2010-07-27 11:56:08 -04003337 if (sb->s_bdev->bd_part)
3338 sbi->s_sectors_written_start =
3339 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003340
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003341 /* Cleanup superblock name */
Rasmus Villemoesec3904d2015-06-25 15:02:41 -07003342 strreplace(sb->s_id, '/', '!');
Theodore Ts'o9f6200b2008-09-23 09:18:24 -04003343
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05003344 /* -EINVAL is default */
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04003345 ret = -EINVAL;
Mingming Cao617ba132006-10-11 01:20:53 -07003346 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003347 if (!blocksize) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003348 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003349 goto out_fail;
3350 }
3351
3352 /*
Mingming Cao617ba132006-10-11 01:20:53 -07003353 * The ext4 superblock will not be buffer aligned for other than 1kB
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003354 * block sizes. We need to calculate the offset from buffer start.
3355 */
Mingming Cao617ba132006-10-11 01:20:53 -07003356 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003357 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3358 offset = do_div(logical_sb_block, blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003359 } else {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003360 logical_sb_block = sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003361 }
3362
Gioh Kima8ac9002014-09-04 22:36:15 -04003363 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003364 ext4_msg(sb, KERN_ERR, "unable to read superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003365 goto out_fail;
3366 }
3367 /*
3368 * Note: s_es must be initialized as soon as possible because
Mingming Cao617ba132006-10-11 01:20:53 -07003369 * some ext4 macro-instructions depend on its value
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003370 */
Theodore Ts'o2716b802012-05-28 17:47:52 -04003371 es = (struct ext4_super_block *) (bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003372 sbi->s_es = es;
3373 sb->s_magic = le16_to_cpu(es->s_magic);
Mingming Cao617ba132006-10-11 01:20:53 -07003374 if (sb->s_magic != EXT4_SUPER_MAGIC)
3375 goto cantfind_ext4;
Theodore Ts'oafc32f72009-02-28 19:39:58 -05003376 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003377
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003378 /* Warn if metadata_csum and gdt_csum are both set. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003379 if (ext4_has_feature_metadata_csum(sb) &&
3380 ext4_has_feature_gdt_csum(sb))
Jakub Wilk363307e2015-01-02 15:31:14 -05003381 ext4_warning(sb, "metadata_csum and uninit_bg are "
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04003382 "redundant flags; please run fsck.");
3383
Darrick J. Wongd25425f2012-04-29 18:25:10 -04003384 /* Check for a known checksum algorithm */
3385 if (!ext4_verify_csum_type(sb, es)) {
3386 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3387 "unknown checksum algorithm.");
3388 silent = 1;
3389 goto cantfind_ext4;
3390 }
3391
Darrick J. Wong04419842012-04-29 18:27:10 -04003392 /* Load the checksum driver */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003393 if (ext4_has_feature_metadata_csum(sb)) {
Darrick J. Wong04419842012-04-29 18:27:10 -04003394 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3395 if (IS_ERR(sbi->s_chksum_driver)) {
3396 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3397 ret = PTR_ERR(sbi->s_chksum_driver);
3398 sbi->s_chksum_driver = NULL;
3399 goto failed_mount;
3400 }
3401 }
3402
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003403 /* Check superblock checksum */
3404 if (!ext4_superblock_csum_verify(sb, es)) {
3405 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3406 "invalid superblock checksum. Run e2fsck?");
3407 silent = 1;
Darrick J. Wong6a797d22015-10-17 16:16:04 -04003408 ret = -EFSBADCRC;
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003409 goto cantfind_ext4;
3410 }
3411
3412 /* Precompute checksum seed for all metadata */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003413 if (ext4_has_feature_csum_seed(sb))
Darrick J. Wong8c81bd82015-10-17 16:16:02 -04003414 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3415 else if (ext4_has_metadata_csum(sb))
Darrick J. Wonga9c473172012-04-29 18:29:10 -04003416 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3417 sizeof(es->s_uuid));
3418
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003419 /* Set defaults before we parse the mount options */
3420 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003421 set_opt(sb, INIT_INODE_TABLE);
Mingming Cao617ba132006-10-11 01:20:53 -07003422 if (def_mount_opts & EXT4_DEFM_DEBUG)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003423 set_opt(sb, DEBUG);
Theodore Ts'o87f26802012-03-02 00:03:21 -05003424 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003425 set_opt(sb, GRPID);
Mingming Cao617ba132006-10-11 01:20:53 -07003426 if (def_mount_opts & EXT4_DEFM_UID16)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003427 set_opt(sb, NO_UID32);
Eric Sandeenea663332011-02-23 17:51:51 -05003428 /* xattr user namespace & acls are now defaulted on */
Eric Sandeenea663332011-02-23 17:51:51 -05003429 set_opt(sb, XATTR_USER);
Theodore Ts'o03010a32008-10-10 20:02:48 -04003430#ifdef CONFIG_EXT4_FS_POSIX_ACL
Eric Sandeenea663332011-02-23 17:51:51 -05003431 set_opt(sb, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08003432#endif
Darrick J. Wong98c1a752014-10-30 10:53:16 -04003433 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3434 if (ext4_has_metadata_csum(sb))
3435 set_opt(sb, JOURNAL_CHECKSUM);
3436
Mingming Cao617ba132006-10-11 01:20:53 -07003437 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003438 set_opt(sb, JOURNAL_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003439 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003440 set_opt(sb, ORDERED_DATA);
Mingming Cao617ba132006-10-11 01:20:53 -07003441 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003442 set_opt(sb, WRITEBACK_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003443
Mingming Cao617ba132006-10-11 01:20:53 -07003444 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003445 set_opt(sb, ERRORS_PANIC);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003446 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003447 set_opt(sb, ERRORS_CONT);
Aneesh Kumar K.Vbb4f3972008-01-28 23:58:26 -05003448 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003449 set_opt(sb, ERRORS_RO);
Darrick J. Wong45f1a9c2014-09-01 21:34:09 -04003450 /* block_validity enabled by default; disable with noblock_validity */
3451 set_opt(sb, BLOCK_VALIDITY);
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003452 if (def_mount_opts & EXT4_DEFM_DISCARD)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003453 set_opt(sb, DISCARD);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003454
Eric W. Biederman08cefc72012-02-07 15:41:49 -08003455 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3456 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
Theodore Ts'o30773842009-01-03 20:27:38 -05003457 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3458 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3459 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003460
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003461 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003462 set_opt(sb, BARRIER);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003463
Mingming Cao1e2462f2007-07-18 09:00:55 -04003464 /*
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003465 * enable delayed allocation by default
3466 * Use -o nodelalloc to turn it off
3467 */
Theodore Ts'obc0b75f2012-09-17 22:54:36 -04003468 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003469 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003470 set_opt(sb, DELALLOC);
Aneesh Kumar K.Vdd919b92008-07-11 19:27:31 -04003471
Lukas Czerner51ce6512011-05-20 13:55:16 -04003472 /*
3473 * set default s_li_wait_mult for lazyinit, for the case there is
3474 * no mount option specified.
3475 */
3476 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3477
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003478 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
Eric Sandeen661aa522012-02-20 17:53:04 -05003479 &journal_devnum, &journal_ioprio, 0)) {
Theodore Ts'o8b67f042010-08-01 23:14:20 -04003480 ext4_msg(sb, KERN_WARNING,
3481 "failed to parse options in superblock: %s",
3482 sbi->s_es->s_mount_opts);
3483 }
Theodore Ts'o5a916be2012-03-04 19:27:31 -05003484 sbi->s_def_mount_opt = sbi->s_mount_opt;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003485 if (!parse_options((char *) data, sb, &journal_devnum,
Eric Sandeen661aa522012-02-20 17:53:04 -05003486 &journal_ioprio, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003487 goto failed_mount;
3488
Theodore Ts'o56889782011-09-03 18:22:38 -04003489 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3490 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3491 "with data=journal disables delayed "
3492 "allocation and O_DIRECT support!\n");
3493 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3494 ext4_msg(sb, KERN_ERR, "can't mount with "
3495 "both data=journal and delalloc");
3496 goto failed_mount;
3497 }
3498 if (test_opt(sb, DIOREAD_NOLOCK)) {
3499 ext4_msg(sb, KERN_ERR, "can't mount with "
Piotr Sarna6ae65142013-08-08 23:02:24 -04003500 "both data=journal and dioread_nolock");
Theodore Ts'o56889782011-09-03 18:22:38 -04003501 goto failed_mount;
3502 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003503 if (test_opt(sb, DAX)) {
3504 ext4_msg(sb, KERN_ERR, "can't mount with "
3505 "both data=journal and dax");
3506 goto failed_mount;
3507 }
Theodore Ts'o56889782011-09-03 18:22:38 -04003508 if (test_opt(sb, DELALLOC))
3509 clear_opt(sb, DELALLOC);
Tejun Heo001e4a82015-07-21 23:51:26 -04003510 } else {
3511 sb->s_iflags |= SB_I_CGROUPWB;
Theodore Ts'o56889782011-09-03 18:22:38 -04003512 }
3513
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003514 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05003515 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003516
Mingming Cao617ba132006-10-11 01:20:53 -07003517 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003518 (ext4_has_compat_features(sb) ||
3519 ext4_has_ro_compat_features(sb) ||
3520 ext4_has_incompat_features(sb)))
Eric Sandeenb31e1552009-06-04 17:36:36 -04003521 ext4_msg(sb, KERN_WARNING,
3522 "feature flags set on rev 0 fs, "
3523 "running e2fsck is recommended");
Theodore Tso469108f2008-02-10 01:11:44 -05003524
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003525 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3526 set_opt2(sb, HURD_COMPAT);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003527 if (ext4_has_feature_64bit(sb)) {
Theodore Ts'oed3654e2014-03-24 14:09:06 -04003528 ext4_msg(sb, KERN_ERR,
3529 "The Hurd can't support 64-bit file systems");
3530 goto failed_mount;
3531 }
3532 }
3533
Theodore Ts'o2035e772011-04-18 17:29:14 -04003534 if (IS_EXT2_SB(sb)) {
3535 if (ext2_feature_set_ok(sb))
3536 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3537 "using the ext4 subsystem");
3538 else {
3539 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3540 "to feature incompatibilities");
3541 goto failed_mount;
3542 }
3543 }
3544
3545 if (IS_EXT3_SB(sb)) {
3546 if (ext3_feature_set_ok(sb))
3547 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3548 "using the ext4 subsystem");
3549 else {
3550 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3551 "to feature incompatibilities");
3552 goto failed_mount;
3553 }
3554 }
3555
Theodore Tso469108f2008-02-10 01:11:44 -05003556 /*
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003557 * Check feature flags regardless of the revision level, since we
3558 * previously didn't change the revision level when setting the flags,
3559 * so there is a chance incompat flags are set on a rev 0 filesystem.
3560 */
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003561 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003562 goto failed_mount;
Eric Sandeena13fb1a2009-08-18 00:20:23 -04003563
Jan Kara261cb20c2012-12-20 00:07:18 -05003564 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
Mingming Cao617ba132006-10-11 01:20:53 -07003565 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3566 blocksize > EXT4_MAX_BLOCK_SIZE) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003567 ext4_msg(sb, KERN_ERR,
Theodore Ts'o8cdf3372016-11-18 13:00:24 -05003568 "Unsupported filesystem blocksize %d (%d log_block_size)",
3569 blocksize, le32_to_cpu(es->s_log_block_size));
3570 goto failed_mount;
3571 }
3572 if (le32_to_cpu(es->s_log_block_size) >
3573 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3574 ext4_msg(sb, KERN_ERR,
3575 "Invalid log block size: %u",
3576 le32_to_cpu(es->s_log_block_size));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003577 goto failed_mount;
3578 }
3579
Theodore Ts'o5b9554d2016-07-05 20:01:52 -04003580 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3581 ext4_msg(sb, KERN_ERR,
3582 "Number of reserved GDT blocks insanely large: %d",
3583 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3584 goto failed_mount;
3585 }
3586
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003587 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
Toshi Kani87eefeb2016-05-10 10:23:54 -06003588 err = bdev_dax_supported(sb, blocksize);
3589 if (err)
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003590 goto failed_mount;
Ross Zwisler923ae0f2015-02-16 15:59:38 -08003591 }
3592
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003593 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04003594 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3595 es->s_encryption_level);
3596 goto failed_mount;
3597 }
3598
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003599 if (sb->s_blocksize != blocksize) {
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003600 /* Validate the filesystem blocksize */
3601 if (!sb_set_blocksize(sb, blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003602 ext4_msg(sb, KERN_ERR, "bad block size %d",
Aneesh Kumar K.Vce407332008-01-28 23:58:27 -05003603 blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003604 goto failed_mount;
3605 }
3606
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003607 brelse(bh);
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003608 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3609 offset = do_div(logical_sb_block, blocksize);
Gioh Kima8ac9002014-09-04 22:36:15 -04003610 bh = sb_bread_unmovable(sb, logical_sb_block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003611 if (!bh) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003612 ext4_msg(sb, KERN_ERR,
3613 "Can't read superblock on 2nd try");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003614 goto failed_mount;
3615 }
Theodore Ts'o2716b802012-05-28 17:47:52 -04003616 es = (struct ext4_super_block *)(bh->b_data + offset);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003617 sbi->s_es = es;
Mingming Cao617ba132006-10-11 01:20:53 -07003618 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003619 ext4_msg(sb, KERN_ERR,
3620 "Magic mismatch, very weird!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003621 goto failed_mount;
3622 }
3623 }
3624
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003625 has_huge_files = ext4_has_feature_huge_file(sb);
Theodore Ts'of287a1a2008-10-16 22:50:48 -04003626 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3627 has_huge_files);
3628 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003629
Mingming Cao617ba132006-10-11 01:20:53 -07003630 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3631 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3632 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003633 } else {
3634 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3635 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
Mingming Cao617ba132006-10-11 01:20:53 -07003636 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
Vignesh Babu13305932007-07-18 09:11:02 -04003637 (!is_power_of_2(sbi->s_inode_size)) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003638 (sbi->s_inode_size > blocksize)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003639 ext4_msg(sb, KERN_ERR,
3640 "unsupported inode size: %d",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003641 sbi->s_inode_size);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003642 goto failed_mount;
3643 }
Kalpak Shahef7f3832007-07-18 09:15:20 -04003644 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3645 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003646 }
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003647
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003648 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003649 if (ext4_has_feature_64bit(sb)) {
Alexandre Ratchov8fadc142006-10-11 01:21:15 -07003650 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003651 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
vignesh babud8ea6cf2007-10-16 23:27:14 -07003652 !is_power_of_2(sbi->s_desc_size)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003653 ext4_msg(sb, KERN_ERR,
3654 "unsupported descriptor size %lu",
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003655 sbi->s_desc_size);
3656 goto failed_mount;
3657 }
3658 } else
3659 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003660
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003661 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003662 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08003663 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003664 goto cantfind_ext4;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003665
Mingming Cao617ba132006-10-11 01:20:53 -07003666 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003667 if (sbi->s_inodes_per_block == 0)
Mingming Cao617ba132006-10-11 01:20:53 -07003668 goto cantfind_ext4;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003669 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3670 sbi->s_inodes_per_block;
Alexandre Ratchov0d1ee422006-10-11 01:21:14 -07003671 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003672 sbi->s_sbh = bh;
3673 sbi->s_mount_state = le16_to_cpu(es->s_state);
Fengguang Wue57aa832007-10-16 23:26:25 -07003674 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3675 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003676
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003677 for (i = 0; i < 4; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003678 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3679 sbi->s_def_hash_version = es->s_def_hash_version;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003680 if (ext4_has_feature_dir_index(sb)) {
Theodore Ts'o23301412014-02-12 12:16:04 -05003681 i = le32_to_cpu(es->s_flags);
3682 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3683 sbi->s_hash_unsigned = 3;
3684 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
Theodore Ts'of99b2582008-10-28 13:21:44 -04003685#ifdef __CHAR_UNSIGNED__
Theodore Ts'o23301412014-02-12 12:16:04 -05003686 if (!(sb->s_flags & MS_RDONLY))
3687 es->s_flags |=
3688 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3689 sbi->s_hash_unsigned = 3;
Theodore Ts'of99b2582008-10-28 13:21:44 -04003690#else
Theodore Ts'o23301412014-02-12 12:16:04 -05003691 if (!(sb->s_flags & MS_RDONLY))
3692 es->s_flags |=
3693 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
Theodore Ts'of99b2582008-10-28 13:21:44 -04003694#endif
Theodore Ts'o23301412014-02-12 12:16:04 -05003695 }
Theodore Ts'of99b2582008-10-28 13:21:44 -04003696 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003697
Theodore Ts'o281b5992011-09-09 18:34:51 -04003698 /* Handle clustersize */
3699 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003700 has_bigalloc = ext4_has_feature_bigalloc(sb);
Theodore Ts'o281b5992011-09-09 18:34:51 -04003701 if (has_bigalloc) {
3702 if (clustersize < blocksize) {
3703 ext4_msg(sb, KERN_ERR,
3704 "cluster size (%d) smaller than "
3705 "block size (%d)", clustersize, blocksize);
3706 goto failed_mount;
3707 }
Theodore Ts'o8cdf3372016-11-18 13:00:24 -05003708 if (le32_to_cpu(es->s_log_cluster_size) >
3709 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3710 ext4_msg(sb, KERN_ERR,
3711 "Invalid log cluster size: %u",
3712 le32_to_cpu(es->s_log_cluster_size));
3713 goto failed_mount;
3714 }
Theodore Ts'o281b5992011-09-09 18:34:51 -04003715 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3716 le32_to_cpu(es->s_log_block_size);
3717 sbi->s_clusters_per_group =
3718 le32_to_cpu(es->s_clusters_per_group);
3719 if (sbi->s_clusters_per_group > blocksize * 8) {
3720 ext4_msg(sb, KERN_ERR,
3721 "#clusters per group too big: %lu",
3722 sbi->s_clusters_per_group);
3723 goto failed_mount;
3724 }
3725 if (sbi->s_blocks_per_group !=
3726 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3727 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3728 "clusters per group (%lu) inconsistent",
3729 sbi->s_blocks_per_group,
3730 sbi->s_clusters_per_group);
3731 goto failed_mount;
3732 }
3733 } else {
3734 if (clustersize != blocksize) {
3735 ext4_warning(sb, "fragment/cluster size (%d) != "
3736 "block size (%d)", clustersize,
3737 blocksize);
3738 clustersize = blocksize;
3739 }
3740 if (sbi->s_blocks_per_group > blocksize * 8) {
3741 ext4_msg(sb, KERN_ERR,
3742 "#blocks per group too big: %lu",
3743 sbi->s_blocks_per_group);
3744 goto failed_mount;
3745 }
3746 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3747 sbi->s_cluster_bits = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003748 }
Theodore Ts'o281b5992011-09-09 18:34:51 -04003749 sbi->s_cluster_ratio = clustersize / blocksize;
3750
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003751 if (sbi->s_inodes_per_group > blocksize * 8) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003752 ext4_msg(sb, KERN_ERR,
3753 "#inodes per group too big: %lu",
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04003754 sbi->s_inodes_per_group);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003755 goto failed_mount;
3756 }
3757
Theodore Ts'o960fd8562013-07-05 23:11:16 -04003758 /* Do we have standard group size of clustersize * 8 blocks ? */
3759 if (sbi->s_blocks_per_group == clustersize << 3)
3760 set_opt2(sb, STD_GROUP_SIZE);
3761
Eric Sandeenbf43d842009-08-17 23:48:51 -04003762 /*
3763 * Test whether we have more sectors than will fit in sector_t,
3764 * and whether the max offset is addressable by the page cache.
3765 */
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003766 err = generic_check_addressable(sb->s_blocksize_bits,
Patrick J. LoPresti30ca22c2010-07-22 15:03:41 -07003767 ext4_blocks_count(es));
Darrick J. Wong5a9ae682010-11-19 09:56:44 -05003768 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003769 ext4_msg(sb, KERN_ERR, "filesystem"
Eric Sandeenbf43d842009-08-17 23:48:51 -04003770 " too large to mount safely on this system");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003771 if (sizeof(sector_t) < 8)
Bartlomiej Zolnierkiewicz90c699a2009-06-19 08:08:50 +02003772 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003773 goto failed_mount;
3774 }
3775
Mingming Cao617ba132006-10-11 01:20:53 -07003776 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3777 goto cantfind_ext4;
Eric Sandeene7c95592008-01-28 23:58:27 -05003778
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003779 /* check blocks count against device size */
3780 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3781 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003782 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3783 "exceeds size of device (%llu blocks)",
From: Thiemo Nagel0f2ddca2009-04-07 14:07:47 -04003784 ext4_blocks_count(es), blocks_count);
3785 goto failed_mount;
3786 }
3787
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04003788 /*
3789 * It makes no sense for the first data block to be beyond the end
3790 * of the filesystem.
3791 */
3792 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
Zheng Liu5635a622011-12-18 16:13:58 -05003793 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003794 "block %u is beyond end of filesystem (%llu)",
3795 le32_to_cpu(es->s_first_data_block),
3796 ext4_blocks_count(es));
Eric Sandeene7c95592008-01-28 23:58:27 -05003797 goto failed_mount;
3798 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003799 blocks_count = (ext4_blocks_count(es) -
3800 le32_to_cpu(es->s_first_data_block) +
3801 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3802 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003803 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003804 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003805 "(block count %llu, first data block %u, "
Eric Sandeenb31e1552009-06-04 17:36:36 -04003806 "blocks per group %lu)", sbi->s_groups_count,
Theodore Ts'o4ec11022009-01-06 14:53:26 -05003807 ext4_blocks_count(es),
3808 le32_to_cpu(es->s_first_data_block),
3809 EXT4_BLOCKS_PER_GROUP(sb));
3810 goto failed_mount;
3811 }
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07003812 sbi->s_groups_count = blocks_count;
Eric Sandeenfb0a3872009-09-16 14:45:10 -04003813 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3814 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
Mingming Cao617ba132006-10-11 01:20:53 -07003815 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3816 EXT4_DESC_PER_BLOCK(sb);
Theodore Ts'of18a5f22011-08-01 08:45:38 -04003817 sbi->s_group_desc = ext4_kvmalloc(db_count *
3818 sizeof(struct buffer_head *),
3819 GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003820 if (sbi->s_group_desc == NULL) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003821 ext4_msg(sb, KERN_ERR, "not enough memory");
Theodore Ts'o2cde4172012-05-28 17:49:54 -04003822 ret = -ENOMEM;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003823 goto failed_mount;
3824 }
3825
Pekka Enberg705895b2009-02-15 18:07:52 -05003826 bgl_lock_init(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003827
3828 for (i = 0; i < db_count; i++) {
Andrew Morton70bbb3e2006-10-11 01:21:20 -07003829 block = descriptor_loc(sb, logical_sb_block, i);
Gioh Kima8ac9002014-09-04 22:36:15 -04003830 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003831 if (!sbi->s_group_desc[i]) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003832 ext4_msg(sb, KERN_ERR,
3833 "can't read group descriptor %d", i);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003834 db_count = i;
3835 goto failed_mount2;
3836 }
3837 }
Theodore Ts'o829fa702016-08-01 00:51:02 -04003838 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003839 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
Darrick J. Wong6a797d22015-10-17 16:16:04 -04003840 ret = -EFSCORRUPTED;
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04003841 goto failed_mount2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003842 }
Jose R. Santos772cb7c2008-07-11 19:27:31 -04003843
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04003844 sbi->s_gdb_count = db_count;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003845 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3846 spin_lock_init(&sbi->s_next_gen_lock);
3847
Jan Mrazek04ecddb2015-01-26 14:42:31 -05003848 setup_timer(&sbi->s_err_report, print_daily_error_info,
3849 (unsigned long) sb);
Tao Ma04496412011-04-05 19:55:28 -04003850
Dmitry Monakhova75ae78f2013-04-03 22:10:52 -04003851 /* Register extent status tree shrinker */
Zheng Liueb68d0e2014-09-01 22:26:49 -04003852 if (ext4_es_register_shrinker(sbi))
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003853 goto failed_mount3;
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003854
Alex Tomasc9de5602008-01-29 00:19:52 -05003855 sbi->s_stripe = ext4_get_stripe_size(sbi);
Zheng Liu67a5da52012-08-17 09:54:17 -04003856 sbi->s_extent_max_zeroout_kb = 32;
Alex Tomasc9de5602008-01-29 00:19:52 -05003857
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04003858 /*
3859 * set up enough so that it can read an inode
3860 */
Theodore Ts'of6e63f92014-09-18 17:12:30 -04003861 sb->s_op = &ext4_sops;
Mingming Cao617ba132006-10-11 01:20:53 -07003862 sb->s_export_op = &ext4_export_ops;
3863 sb->s_xattr = ext4_xattr_handlers;
Jaegeuk Kima7550b32016-07-10 14:01:03 -04003864 sb->s_cop = &ext4_cryptops;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003865#ifdef CONFIG_QUOTA
Mingming Cao617ba132006-10-11 01:20:53 -07003866 sb->dq_op = &ext4_quota_operations;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003867 if (ext4_has_feature_quota(sb))
Jan Kara1fa5efe2014-10-08 18:26:54 +02003868 sb->s_qcop = &dquot_quotactl_sysfile_ops;
Jan Kara262b4662013-03-02 17:57:08 -05003869 else
3870 sb->s_qcop = &ext4_qctl_operations;
Li Xi689c9582016-01-08 16:01:22 -05003871 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003872#endif
Aneesh Kumar K.Vf2fa2ff2011-01-29 18:43:40 +05303873 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3874
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003875 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
Theodore Ts'o3b9d4ed2009-04-25 22:54:04 -04003876 mutex_init(&sbi->s_orphan_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003877
3878 sb->s_root = NULL;
3879
3880 needs_recovery = (es->s_last_orphan != 0 ||
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003881 ext4_has_feature_journal_needs_recovery(sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003882
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003883 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
Johann Lombardic5e06d12011-05-24 18:31:25 -04003884 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
Darrick J. Wong50460fe2014-10-30 10:53:16 -04003885 goto failed_mount3a;
Johann Lombardic5e06d12011-05-24 18:31:25 -04003886
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003887 /*
3888 * The first inode we look at is the journal inode. Don't try
3889 * root first: it may be modified in the journal!
3890 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003891 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
Mingming Cao617ba132006-10-11 01:20:53 -07003892 if (ext4_load_journal(sb, es, journal_devnum))
Darrick J. Wong50460fe2014-10-30 10:53:16 -04003893 goto failed_mount3a;
Frank Mayhar03901312009-01-07 00:06:22 -05003894 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003895 ext4_has_feature_journal_needs_recovery(sb)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003896 ext4_msg(sb, KERN_ERR, "required journal recovery "
3897 "suppressed and not mounted read-only");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003898 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003899 } else {
Dmitry Monakhov1e381f62015-10-18 23:50:26 -04003900 /* Nojournal mode, all journal mount options are illegal */
3901 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3902 ext4_msg(sb, KERN_ERR, "can't mount with "
3903 "journal_checksum, fs mounted w/o journal");
3904 goto failed_mount_wq;
3905 }
3906 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3907 ext4_msg(sb, KERN_ERR, "can't mount with "
3908 "journal_async_commit, fs mounted w/o journal");
3909 goto failed_mount_wq;
3910 }
3911 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3912 ext4_msg(sb, KERN_ERR, "can't mount with "
3913 "commit=%lu, fs mounted w/o journal",
3914 sbi->s_commit_interval / HZ);
3915 goto failed_mount_wq;
3916 }
3917 if (EXT4_MOUNT_DATA_FLAGS &
3918 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3919 ext4_msg(sb, KERN_ERR, "can't mount with "
3920 "data=, fs mounted w/o journal");
3921 goto failed_mount_wq;
3922 }
3923 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3924 clear_opt(sb, JOURNAL_CHECKSUM);
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003925 clear_opt(sb, DATA_FLAGS);
Frank Mayhar03901312009-01-07 00:06:22 -05003926 sbi->s_journal = NULL;
3927 needs_recovery = 0;
3928 goto no_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003929 }
3930
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003931 if (ext4_has_feature_64bit(sb) &&
Jose R. Santoseb40a092007-07-18 08:37:25 -04003932 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3933 JBD2_FEATURE_INCOMPAT_64BIT)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003934 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003935 goto failed_mount_wq;
Jose R. Santoseb40a092007-07-18 08:37:25 -04003936 }
3937
Darrick J. Wong25ed6e82012-05-27 07:48:56 -04003938 if (!set_journal_csum_feature_set(sb)) {
3939 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3940 "feature set");
3941 goto failed_mount_wq;
Linus Torvaldsd4da6c92009-11-02 10:15:27 -08003942 }
Girish Shilamkar818d2762008-01-28 23:58:27 -05003943
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003944 /* We have now updated the journal if required, so we can
3945 * validate the data journaling mode. */
3946 switch (test_opt(sb, DATA_FLAGS)) {
3947 case 0:
3948 /* No mode set, assume a default based on the journal
Andrew Morton63f57932006-10-11 01:21:24 -07003949 * capabilities: ORDERED_DATA if the journal can
3950 * cope, else JOURNAL_DATA
3951 */
Mingming Caodab291a2006-10-11 01:21:01 -07003952 if (jbd2_journal_check_available_features
3953 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003954 set_opt(sb, ORDERED_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003955 else
Theodore Ts'ofd8c37e2010-12-15 20:26:48 -05003956 set_opt(sb, JOURNAL_DATA);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003957 break;
3958
Mingming Cao617ba132006-10-11 01:20:53 -07003959 case EXT4_MOUNT_ORDERED_DATA:
3960 case EXT4_MOUNT_WRITEBACK_DATA:
Mingming Caodab291a2006-10-11 01:21:01 -07003961 if (!jbd2_journal_check_available_features
3962 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04003963 ext4_msg(sb, KERN_ERR, "Journal does not support "
3964 "requested data journaling mode");
Jiaying Zhang744692d2010-03-04 16:14:02 -05003965 goto failed_mount_wq;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003966 }
3967 default:
3968 break;
3969 }
Theodore Ts'ob3881f72009-01-05 22:46:26 -05003970 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003971
Bobi Jam18aadd42012-02-20 17:53:02 -05003972 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3973
Theodore Ts'oce7e0102010-11-03 12:03:21 -04003974no_journal:
Andreas Gruenbacher2335d052016-02-22 22:41:05 -05003975 sbi->s_mb_cache = ext4_xattr_create_cache();
3976 if (!sbi->s_mb_cache) {
3977 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3978 goto failed_mount_wq;
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04003979 }
3980
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003981 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003982 (blocksize != PAGE_SIZE)) {
Theodore Ts'o1cb767c2015-06-12 23:44:33 -04003983 ext4_msg(sb, KERN_ERR,
3984 "Unsupported blocksize for fs encryption");
3985 goto failed_mount_wq;
3986 }
3987
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04003988 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3989 !ext4_has_feature_encrypt(sb)) {
3990 ext4_set_feature_encrypt(sb);
Theodore Ts'o6ddb2442015-04-16 01:56:00 -04003991 ext4_commit_super(sb, 1);
3992 }
3993
Tejun Heofd89d5f2011-02-01 11:42:42 +01003994 /*
Theodore Ts'o952fc182012-07-09 16:27:05 -04003995 * Get the # of file system overhead blocks from the
3996 * superblock if present.
3997 */
3998 if (es->s_overhead_clusters)
3999 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4000 else {
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004001 err = ext4_calculate_overhead(sb);
4002 if (err)
Theodore Ts'o952fc182012-07-09 16:27:05 -04004003 goto failed_mount_wq;
4004 }
4005
4006 /*
Tejun Heofd89d5f2011-02-01 11:42:42 +01004007 * The maximum number of concurrent works can be high and
4008 * concurrency isn't really necessary. Limit it to 1.
4009 */
Jan Kara2e8fa542013-06-04 14:21:02 -04004010 EXT4_SB(sb)->rsv_conversion_wq =
4011 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4012 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4013 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004014 ret = -ENOMEM;
Jan Kara2e8fa542013-06-04 14:21:02 -04004015 goto failed_mount4;
4016 }
4017
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004018 /*
Mingming Caodab291a2006-10-11 01:21:01 -07004019 * The jbd2_journal_load will have done any necessary log recovery,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004020 * so we can safely mount the rest of the filesystem now.
4021 */
4022
David Howells1d1fe1e2008-02-07 00:15:37 -08004023 root = ext4_iget(sb, EXT4_ROOT_INO);
4024 if (IS_ERR(root)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004025 ext4_msg(sb, KERN_ERR, "get root inode failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004026 ret = PTR_ERR(root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004027 root = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004028 goto failed_mount4;
4029 }
4030 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004031 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
Al Viro94bf6082012-01-09 15:53:24 -05004032 iput(root);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004033 goto failed_mount4;
4034 }
Al Viro48fde702012-01-08 22:15:13 -05004035 sb->s_root = d_make_root(root);
David Howells1d1fe1e2008-02-07 00:15:37 -08004036 if (!sb->s_root) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004037 ext4_msg(sb, KERN_ERR, "get root dentry failed");
David Howells1d1fe1e2008-02-07 00:15:37 -08004038 ret = -ENOMEM;
4039 goto failed_mount4;
4040 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004041
Eric Sandeen7e84b622012-05-28 14:17:25 -04004042 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4043 sb->s_flags |= MS_RDONLY;
Kalpak Shahef7f3832007-07-18 09:15:20 -04004044
4045 /* determine the minimum size of new large inodes, if present */
4046 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4047 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4048 EXT4_GOOD_OLD_INODE_SIZE;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004049 if (ext4_has_feature_extra_isize(sb)) {
Kalpak Shahef7f3832007-07-18 09:15:20 -04004050 if (sbi->s_want_extra_isize <
4051 le16_to_cpu(es->s_want_extra_isize))
4052 sbi->s_want_extra_isize =
4053 le16_to_cpu(es->s_want_extra_isize);
4054 if (sbi->s_want_extra_isize <
4055 le16_to_cpu(es->s_min_extra_isize))
4056 sbi->s_want_extra_isize =
4057 le16_to_cpu(es->s_min_extra_isize);
4058 }
4059 }
4060 /* Check if enough inode space is available */
4061 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4062 sbi->s_inode_size) {
4063 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4064 EXT4_GOOD_OLD_INODE_SIZE;
Eric Sandeenb31e1552009-06-04 17:36:36 -04004065 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4066 "available");
Kalpak Shahef7f3832007-07-18 09:15:20 -04004067 }
4068
Theodore Ts'ob5799012015-09-23 12:44:17 -04004069 ext4_set_resv_clusters(sb);
Lukas Czerner27dd4382013-04-09 22:11:22 -04004070
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004071 err = ext4_setup_system_zone(sb);
4072 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004073 ext4_msg(sb, KERN_ERR, "failed to initialize system "
Curt Wohlgemuthfbe845d2010-05-16 13:00:00 -04004074 "zone (%d)", err);
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004075 goto failed_mount4a;
4076 }
4077
4078 ext4_ext_init(sb);
4079 err = ext4_mb_init(sb);
4080 if (err) {
4081 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4082 err);
Tao Madcf2d802011-10-06 12:10:11 -04004083 goto failed_mount5;
Aneesh Kumar K.Vc2774d82008-10-10 20:07:20 -04004084 }
4085
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004086 block = ext4_count_free_clusters(sb);
4087 ext4_free_blocks_count_set(sbi->s_es,
4088 EXT4_C2B(sbi, block));
Tejun Heo908c7f12014-09-08 09:51:29 +09004089 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4090 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004091 if (!err) {
4092 unsigned long freei = ext4_count_free_inodes(sb);
4093 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
Tejun Heo908c7f12014-09-08 09:51:29 +09004094 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4095 GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004096 }
4097 if (!err)
4098 err = percpu_counter_init(&sbi->s_dirs_counter,
Tejun Heo908c7f12014-09-08 09:51:29 +09004099 ext4_count_dirs(sb), GFP_KERNEL);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004100 if (!err)
Tejun Heo908c7f12014-09-08 09:51:29 +09004101 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4102 GFP_KERNEL);
Daeho Jeongc8585c62016-04-25 23:22:35 -04004103 if (!err)
4104 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4105
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004106 if (err) {
4107 ext4_msg(sb, KERN_ERR, "insufficient memory");
4108 goto failed_mount6;
4109 }
4110
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004111 if (ext4_has_feature_flex_bg(sb))
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004112 if (!ext4_fill_flex_info(sb)) {
4113 ext4_msg(sb, KERN_ERR,
4114 "unable to initialize "
4115 "flex_bg meta info!");
4116 goto failed_mount6;
4117 }
4118
Lukas Czernerbfff6872010-10-27 21:30:05 -04004119 err = ext4_register_li_request(sb, first_not_zeroed);
4120 if (err)
Tao Madcf2d802011-10-06 12:10:11 -04004121 goto failed_mount6;
Lukas Czernerbfff6872010-10-27 21:30:05 -04004122
Theodore Ts'ob5799012015-09-23 12:44:17 -04004123 err = ext4_register_sysfs(sb);
Tao Madcf2d802011-10-06 12:10:11 -04004124 if (err)
4125 goto failed_mount7;
Theodore Ts'o3197ebd2009-03-31 09:10:09 -04004126
Jan Kara9b2ff3572013-03-02 18:22:38 -05004127#ifdef CONFIG_QUOTA
4128 /* Enable quota usage during mount. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004129 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
Jan Kara9b2ff3572013-03-02 18:22:38 -05004130 err = ext4_enable_quotas(sb);
4131 if (err)
4132 goto failed_mount8;
4133 }
4134#endif /* CONFIG_QUOTA */
4135
Mingming Cao617ba132006-10-11 01:20:53 -07004136 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4137 ext4_orphan_cleanup(sb, es);
4138 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
Frank Mayhar03901312009-01-07 00:06:22 -05004139 if (needs_recovery) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004140 ext4_msg(sb, KERN_INFO, "recovery complete");
Frank Mayhar03901312009-01-07 00:06:22 -05004141 ext4_mark_recovery_complete(sb, es);
4142 }
4143 if (EXT4_SB(sb)->s_journal) {
4144 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4145 descr = " journalled data mode";
4146 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4147 descr = " ordered data mode";
4148 else
4149 descr = " writeback data mode";
4150 } else
4151 descr = "out journal";
4152
Lukas Czerner79add3a2012-11-08 13:28:29 -05004153 if (test_opt(sb, DISCARD)) {
4154 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4155 if (!blk_queue_discard(q))
4156 ext4_msg(sb, KERN_WARNING,
4157 "mounting with \"discard\" option, but "
4158 "the device does not support discard");
4159 }
4160
Theodore Ts'oe294a532015-08-15 14:59:44 -04004161 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4162 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4163 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4164 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004165
Theodore Ts'o66e61a92010-07-27 11:56:04 -04004166 if (es->s_error_count)
4167 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004168
Theodore Ts'oefbed4d2013-10-17 21:11:01 -04004169 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4170 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4171 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4172 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4173
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004174 kfree(orig_data);
Jaegeuk Kima7550b32016-07-10 14:01:03 -04004175#ifdef CONFIG_EXT4_FS_ENCRYPTION
4176 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4177 EXT4_KEY_DESC_PREFIX_SIZE);
4178 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4179#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004180 return 0;
4181
Mingming Cao617ba132006-10-11 01:20:53 -07004182cantfind_ext4:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004183 if (!silent)
Eric Sandeenb31e1552009-06-04 17:36:36 -04004184 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004185 goto failed_mount;
4186
Theodore Ts'o72ba7452013-01-24 23:24:54 -05004187#ifdef CONFIG_QUOTA
4188failed_mount8:
Theodore Ts'oebd173b2015-09-23 12:46:17 -04004189 ext4_unregister_sysfs(sb);
Theodore Ts'o72ba7452013-01-24 23:24:54 -05004190#endif
Tao Madcf2d802011-10-06 12:10:11 -04004191failed_mount7:
4192 ext4_unregister_li_request(sb);
4193failed_mount6:
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004194 ext4_mb_release(sb);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004195 if (sbi->s_flex_groups)
Al Virob93b41d2014-11-20 12:19:11 -05004196 kvfree(sbi->s_flex_groups);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004197 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4198 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4199 percpu_counter_destroy(&sbi->s_dirs_counter);
4200 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
Azat Khuzhin00764932014-04-07 10:54:20 -04004201failed_mount5:
Theodore Ts'of9ae9cf2014-07-11 13:55:40 -04004202 ext4_ext_release(sb);
4203 ext4_release_system_zone(sb);
4204failed_mount4a:
Al Viro94bf6082012-01-09 15:53:24 -05004205 dput(sb->s_root);
Manish Katiyar32a9bb52011-02-27 20:42:06 -05004206 sb->s_root = NULL;
Al Viro94bf6082012-01-09 15:53:24 -05004207failed_mount4:
Eric Sandeenb31e1552009-06-04 17:36:36 -04004208 ext4_msg(sb, KERN_ERR, "mount failed");
Jan Kara2e8fa542013-06-04 14:21:02 -04004209 if (EXT4_SB(sb)->rsv_conversion_wq)
4210 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
Mingming Cao4c0425f2009-09-28 15:48:41 -04004211failed_mount_wq:
Jan Kara82939d72016-02-22 11:50:13 -05004212 if (sbi->s_mb_cache) {
4213 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4214 sbi->s_mb_cache = NULL;
4215 }
Frank Mayhar03901312009-01-07 00:06:22 -05004216 if (sbi->s_journal) {
4217 jbd2_journal_destroy(sbi->s_journal);
4218 sbi->s_journal = NULL;
4219 }
Darrick J. Wong50460fe2014-10-30 10:53:16 -04004220failed_mount3a:
Zheng Liud3922a72013-07-01 08:12:37 -04004221 ext4_es_unregister_shrinker(sbi);
Zheng Liueb68d0e2014-09-01 22:26:49 -04004222failed_mount3:
Al Viro9105bb12013-12-08 20:52:31 -05004223 del_timer_sync(&sbi->s_err_report);
Johann Lombardic5e06d12011-05-24 18:31:25 -04004224 if (sbi->s_mmp_tsk)
4225 kthread_stop(sbi->s_mmp_tsk);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004226failed_mount2:
4227 for (i = 0; i < db_count; i++)
4228 brelse(sbi->s_group_desc[i]);
Al Virob93b41d2014-11-20 12:19:11 -05004229 kvfree(sbi->s_group_desc);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004230failed_mount:
Darrick J. Wong04419842012-04-29 18:27:10 -04004231 if (sbi->s_chksum_driver)
4232 crypto_free_shash(sbi->s_chksum_driver);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004233#ifdef CONFIG_QUOTA
Jan Karaa2d4a642014-09-11 11:15:15 -04004234 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004235 kfree(sbi->s_qf_names[i]);
4236#endif
Mingming Cao617ba132006-10-11 01:20:53 -07004237 ext4_blkdev_remove(sbi);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004238 brelse(bh);
4239out_fail:
4240 sb->s_fs_info = NULL;
Manish Katiyarf6830162009-05-17 23:52:44 -04004241 kfree(sbi->s_blockgroup_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004242 kfree(sbi);
Cyrill Gorcunovdcc7dae2010-07-27 11:56:07 -04004243out_free_orig:
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004244 kfree(orig_data);
Lukas Czerner07aa2ea2012-11-08 15:16:54 -05004245 return err ? err : ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004246}
4247
4248/*
4249 * Setup any per-fs journal parameters now. We'll do this both on
4250 * initial mount, once the journal has been initialised but before we've
4251 * done any recovery; and again on any subsequent remount.
4252 */
Mingming Cao617ba132006-10-11 01:20:53 -07004253static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004254{
Mingming Cao617ba132006-10-11 01:20:53 -07004255 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004256
Theodore Ts'o30773842009-01-03 20:27:38 -05004257 journal->j_commit_interval = sbi->s_commit_interval;
4258 journal->j_min_batch_time = sbi->s_min_batch_time;
4259 journal->j_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004260
Theodore Ts'oa931da62010-08-03 21:35:12 -04004261 write_lock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004262 if (test_opt(sb, BARRIER))
Mingming Caodab291a2006-10-11 01:21:01 -07004263 journal->j_flags |= JBD2_BARRIER;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004264 else
Mingming Caodab291a2006-10-11 01:21:01 -07004265 journal->j_flags &= ~JBD2_BARRIER;
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04004266 if (test_opt(sb, DATA_ERR_ABORT))
4267 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4268 else
4269 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
Theodore Ts'oa931da62010-08-03 21:35:12 -04004270 write_unlock(&journal->j_state_lock);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004271}
4272
Eric Whitneyc6cb7e72016-09-30 02:05:09 -04004273static struct inode *ext4_get_journal_inode(struct super_block *sb,
4274 unsigned int journal_inum)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004275{
4276 struct inode *journal_inode;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004277
Eric Whitneyc6cb7e72016-09-30 02:05:09 -04004278 /*
4279 * Test for the existence of a valid inode on disk. Bad things
4280 * happen if we iget() an unused inode, as the subsequent iput()
4281 * will try to delete it.
4282 */
David Howells1d1fe1e2008-02-07 00:15:37 -08004283 journal_inode = ext4_iget(sb, journal_inum);
4284 if (IS_ERR(journal_inode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004285 ext4_msg(sb, KERN_ERR, "no journal found");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004286 return NULL;
4287 }
4288 if (!journal_inode->i_nlink) {
4289 make_bad_inode(journal_inode);
4290 iput(journal_inode);
Eric Sandeenb31e1552009-06-04 17:36:36 -04004291 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004292 return NULL;
4293 }
4294
Theodore Ts'oe5f8eab82008-09-08 22:25:04 -04004295 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004296 journal_inode, journal_inode->i_size);
David Howells1d1fe1e2008-02-07 00:15:37 -08004297 if (!S_ISREG(journal_inode->i_mode)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004298 ext4_msg(sb, KERN_ERR, "invalid journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004299 iput(journal_inode);
4300 return NULL;
4301 }
Eric Whitneyc6cb7e72016-09-30 02:05:09 -04004302 return journal_inode;
4303}
4304
4305static journal_t *ext4_get_journal(struct super_block *sb,
4306 unsigned int journal_inum)
4307{
4308 struct inode *journal_inode;
4309 journal_t *journal;
4310
4311 BUG_ON(!ext4_has_feature_journal(sb));
4312
4313 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4314 if (!journal_inode)
4315 return NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004316
Mingming Caodab291a2006-10-11 01:21:01 -07004317 journal = jbd2_journal_init_inode(journal_inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004318 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004319 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004320 iput(journal_inode);
4321 return NULL;
4322 }
4323 journal->j_private = sb;
Mingming Cao617ba132006-10-11 01:20:53 -07004324 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004325 return journal;
4326}
4327
Mingming Cao617ba132006-10-11 01:20:53 -07004328static journal_t *ext4_get_dev_journal(struct super_block *sb,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004329 dev_t j_dev)
4330{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004331 struct buffer_head *bh;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004332 journal_t *journal;
Mingming Cao617ba132006-10-11 01:20:53 -07004333 ext4_fsblk_t start;
4334 ext4_fsblk_t len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004335 int hblock, blocksize;
Mingming Cao617ba132006-10-11 01:20:53 -07004336 ext4_fsblk_t sb_block;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004337 unsigned long offset;
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004338 struct ext4_super_block *es;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004339 struct block_device *bdev;
4340
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004341 BUG_ON(!ext4_has_feature_journal(sb));
Frank Mayhar03901312009-01-07 00:06:22 -05004342
Eric Sandeenb31e1552009-06-04 17:36:36 -04004343 bdev = ext4_blkdev_get(j_dev, sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004344 if (bdev == NULL)
4345 return NULL;
4346
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004347 blocksize = sb->s_blocksize;
Martin K. Petersene1defc42009-05-22 17:17:49 -04004348 hblock = bdev_logical_block_size(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004349 if (blocksize < hblock) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004350 ext4_msg(sb, KERN_ERR,
4351 "blocksize too small for journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004352 goto out_bdev;
4353 }
4354
Mingming Cao617ba132006-10-11 01:20:53 -07004355 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4356 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004357 set_blocksize(bdev, blocksize);
4358 if (!(bh = __bread(bdev, sb_block, blocksize))) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004359 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4360 "external journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004361 goto out_bdev;
4362 }
4363
Theodore Ts'o2716b802012-05-28 17:47:52 -04004364 es = (struct ext4_super_block *) (bh->b_data + offset);
Mingming Cao617ba132006-10-11 01:20:53 -07004365 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004366 !(le32_to_cpu(es->s_feature_incompat) &
Mingming Cao617ba132006-10-11 01:20:53 -07004367 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004368 ext4_msg(sb, KERN_ERR, "external journal has "
4369 "bad superblock");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004370 brelse(bh);
4371 goto out_bdev;
4372 }
4373
Darrick J. Wongdf4763b2014-09-11 11:44:36 -04004374 if ((le32_to_cpu(es->s_feature_ro_compat) &
4375 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4376 es->s_checksum != ext4_superblock_csum(sb, es)) {
4377 ext4_msg(sb, KERN_ERR, "external journal has "
4378 "corrupt superblock");
4379 brelse(bh);
4380 goto out_bdev;
4381 }
4382
Mingming Cao617ba132006-10-11 01:20:53 -07004383 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004384 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004385 brelse(bh);
4386 goto out_bdev;
4387 }
4388
Laurent Vivierbd81d8e2006-10-11 01:21:10 -07004389 len = ext4_blocks_count(es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004390 start = sb_block + 1;
4391 brelse(bh); /* we're done with the superblock */
4392
Mingming Caodab291a2006-10-11 01:21:01 -07004393 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004394 start, len, blocksize);
4395 if (!journal) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004396 ext4_msg(sb, KERN_ERR, "failed to create device journal");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004397 goto out_bdev;
4398 }
4399 journal->j_private = sb;
Mike Christiedfec8a12016-06-05 14:31:44 -05004400 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004401 wait_on_buffer(journal->j_sb_buffer);
4402 if (!buffer_uptodate(journal->j_sb_buffer)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004403 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004404 goto out_journal;
4405 }
4406 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004407 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4408 "user (unsupported) - %d",
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004409 be32_to_cpu(journal->j_superblock->s_nr_users));
4410 goto out_journal;
4411 }
Mingming Cao617ba132006-10-11 01:20:53 -07004412 EXT4_SB(sb)->journal_bdev = bdev;
4413 ext4_init_journal_params(sb, journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004414 return journal;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004415
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004416out_journal:
Mingming Caodab291a2006-10-11 01:21:01 -07004417 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004418out_bdev:
Mingming Cao617ba132006-10-11 01:20:53 -07004419 ext4_blkdev_put(bdev);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004420 return NULL;
4421}
4422
Mingming Cao617ba132006-10-11 01:20:53 -07004423static int ext4_load_journal(struct super_block *sb,
4424 struct ext4_super_block *es,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004425 unsigned long journal_devnum)
4426{
4427 journal_t *journal;
4428 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4429 dev_t journal_dev;
4430 int err = 0;
4431 int really_read_only;
4432
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004433 BUG_ON(!ext4_has_feature_journal(sb));
Frank Mayhar03901312009-01-07 00:06:22 -05004434
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004435 if (journal_devnum &&
4436 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004437 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4438 "numbers have changed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004439 journal_dev = new_decode_dev(journal_devnum);
4440 } else
4441 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4442
4443 really_read_only = bdev_read_only(sb->s_bdev);
4444
4445 /*
4446 * Are we loading a blank journal or performing recovery after a
4447 * crash? For recovery, we need to check in advance whether we
4448 * can get read-write access to the device.
4449 */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004450 if (ext4_has_feature_journal_needs_recovery(sb)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004451 if (sb->s_flags & MS_RDONLY) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004452 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4453 "required on readonly filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004454 if (really_read_only) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004455 ext4_msg(sb, KERN_ERR, "write access "
4456 "unavailable, cannot proceed");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004457 return -EROFS;
4458 }
Eric Sandeenb31e1552009-06-04 17:36:36 -04004459 ext4_msg(sb, KERN_INFO, "write access will "
4460 "be enabled during recovery");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004461 }
4462 }
4463
4464 if (journal_inum && journal_dev) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004465 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4466 "and inode journals!");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004467 return -EINVAL;
4468 }
4469
4470 if (journal_inum) {
Mingming Cao617ba132006-10-11 01:20:53 -07004471 if (!(journal = ext4_get_journal(sb, journal_inum)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004472 return -EINVAL;
4473 } else {
Mingming Cao617ba132006-10-11 01:20:53 -07004474 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004475 return -EINVAL;
4476 }
4477
Theodore Ts'o90576c02009-09-29 15:51:30 -04004478 if (!(journal->j_flags & JBD2_BARRIER))
Eric Sandeenb31e1552009-06-04 17:36:36 -04004479 ext4_msg(sb, KERN_INFO, "barriers disabled");
Theodore Ts'o4776004f2008-09-08 23:00:52 -04004480
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004481 if (!ext4_has_feature_journal_needs_recovery(sb))
Mingming Caodab291a2006-10-11 01:21:01 -07004482 err = jbd2_journal_wipe(journal, !really_read_only);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004483 if (!err) {
4484 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4485 if (save)
4486 memcpy(save, ((char *) es) +
4487 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
Mingming Caodab291a2006-10-11 01:21:01 -07004488 err = jbd2_journal_load(journal);
Theodore Ts'o1c13d5c2010-07-27 11:56:03 -04004489 if (save)
4490 memcpy(((char *) es) + EXT4_S_ERR_START,
4491 save, EXT4_S_ERR_LEN);
4492 kfree(save);
4493 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004494
4495 if (err) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004496 ext4_msg(sb, KERN_ERR, "error loading journal");
Mingming Caodab291a2006-10-11 01:21:01 -07004497 jbd2_journal_destroy(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004498 return err;
4499 }
4500
Mingming Cao617ba132006-10-11 01:20:53 -07004501 EXT4_SB(sb)->s_journal = journal;
4502 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004503
Maciej Żenczykowskic41303c2010-10-27 21:30:06 -04004504 if (!really_read_only && journal_devnum &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004505 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4506 es->s_journal_dev = cpu_to_le32(journal_devnum);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004507
4508 /* Make sure we flush the recovery flag to disk. */
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004509 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004510 }
4511
4512 return 0;
4513}
4514
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004515static int ext4_commit_super(struct super_block *sb, int sync)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004516{
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004517 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
Mingming Cao617ba132006-10-11 01:20:53 -07004518 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
Takashi Satoc4be0c12009-01-09 16:40:58 -08004519 int error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004520
Theodore Ts'obdfe0cbd2015-08-16 10:03:57 -04004521 if (!sbh || block_device_ejected(sb))
Takashi Satoc4be0c12009-01-09 16:40:58 -08004522 return error;
Theodore Ts'o71290b32009-09-10 17:31:04 -04004523 /*
4524 * If the file system is mounted read-only, don't update the
4525 * superblock write time. This avoids updating the superblock
4526 * write time when we are mounting the root file system
4527 * read/only but we need to replay the journal; at that point,
4528 * for people who are east of GMT and who make their clock
4529 * tick in localtime for Windows bug-for-bug compatibility,
4530 * the clock is set in the future, and this will cause e2fsck
4531 * to complain and force a full file system check.
4532 */
4533 if (!(sb->s_flags & MS_RDONLY))
4534 es->s_wtime = cpu_to_le32(get_seconds());
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004535 if (sb->s_bdev->bd_part)
4536 es->s_kbytes_written =
4537 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
Theodore Ts'oafc32f72009-02-28 19:39:58 -05004538 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4539 EXT4_SB(sb)->s_sectors_written_start) >> 1));
Theodore Ts'of613dfc2010-07-27 11:56:08 -04004540 else
4541 es->s_kbytes_written =
4542 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004543 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4544 ext4_free_blocks_count_set(es,
Theodore Ts'o57042652011-09-09 18:56:51 -04004545 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4546 &EXT4_SB(sb)->s_freeclusters_counter)));
Theodore Ts'od5e03cb2014-07-15 06:01:38 -04004547 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4548 es->s_free_inodes_count =
4549 cpu_to_le32(percpu_counter_sum_positive(
Theodore Ts'oce7e0102010-11-03 12:03:21 -04004550 &EXT4_SB(sb)->s_freeinodes_counter));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004551 BUFFER_TRACE(sbh, "marking dirty");
Theodore Ts'o06db49e2012-10-10 01:06:58 -04004552 ext4_superblock_csum_set(sb);
Pranay Kr. Srivastava4743f832016-07-04 10:24:52 -04004553 lock_buffer(sbh);
4554 if (buffer_write_io_error(sbh)) {
4555 /*
4556 * Oh, dear. A previous attempt to write the
4557 * superblock failed. This could happen because the
4558 * USB device was yanked out. Or it could happen to
4559 * be a transient write error and maybe the block will
4560 * be remapped. Nothing we can do but to retry the
4561 * write and hope for the best.
4562 */
4563 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4564 "superblock detected");
4565 clear_buffer_write_io_error(sbh);
4566 set_buffer_uptodate(sbh);
4567 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004568 mark_buffer_dirty(sbh);
Pranay Kr. Srivastava4743f832016-07-04 10:24:52 -04004569 unlock_buffer(sbh);
Theodore Ts'o914258b2008-10-06 21:35:40 -04004570 if (sync) {
Daeho Jeong564bc402015-07-23 09:46:11 -04004571 error = __sync_dirty_buffer(sbh,
4572 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004573 if (error)
4574 return error;
4575
4576 error = buffer_write_io_error(sbh);
4577 if (error) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004578 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4579 "superblock");
Theodore Ts'o914258b2008-10-06 21:35:40 -04004580 clear_buffer_write_io_error(sbh);
4581 set_buffer_uptodate(sbh);
4582 }
4583 }
Takashi Satoc4be0c12009-01-09 16:40:58 -08004584 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004585}
4586
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004587/*
4588 * Have we just finished recovery? If so, and if we are mounting (or
4589 * remounting) the filesystem readonly, then we will end up with a
4590 * consistent fs on disk. Record that fact.
4591 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004592static void ext4_mark_recovery_complete(struct super_block *sb,
4593 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004594{
Mingming Cao617ba132006-10-11 01:20:53 -07004595 journal_t *journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004596
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004597 if (!ext4_has_feature_journal(sb)) {
Frank Mayhar03901312009-01-07 00:06:22 -05004598 BUG_ON(journal != NULL);
4599 return;
4600 }
Mingming Caodab291a2006-10-11 01:21:01 -07004601 jbd2_journal_lock_updates(journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004602 if (jbd2_journal_flush(journal) < 0)
4603 goto out;
4604
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004605 if (ext4_has_feature_journal_needs_recovery(sb) &&
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004606 sb->s_flags & MS_RDONLY) {
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004607 ext4_clear_feature_journal_needs_recovery(sb);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004608 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004609 }
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004610
4611out:
Mingming Caodab291a2006-10-11 01:21:01 -07004612 jbd2_journal_unlock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004613}
4614
4615/*
4616 * If we are mounting (or read-write remounting) a filesystem whose journal
4617 * has recorded an error from a previous lifetime, move that error to the
4618 * main filesystem now.
4619 */
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004620static void ext4_clear_journal_err(struct super_block *sb,
4621 struct ext4_super_block *es)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004622{
4623 journal_t *journal;
4624 int j_errno;
4625 const char *errstr;
4626
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004627 BUG_ON(!ext4_has_feature_journal(sb));
Frank Mayhar03901312009-01-07 00:06:22 -05004628
Mingming Cao617ba132006-10-11 01:20:53 -07004629 journal = EXT4_SB(sb)->s_journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004630
4631 /*
4632 * Now check for any error status which may have been recorded in the
Mingming Cao617ba132006-10-11 01:20:53 -07004633 * journal by a prior ext4_error() or ext4_abort()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004634 */
4635
Mingming Caodab291a2006-10-11 01:21:01 -07004636 j_errno = jbd2_journal_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004637 if (j_errno) {
4638 char nbuf[16];
4639
Mingming Cao617ba132006-10-11 01:20:53 -07004640 errstr = ext4_decode_error(sb, j_errno, nbuf);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004641 ext4_warning(sb, "Filesystem error recorded "
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004642 "from previous mount: %s", errstr);
Eric Sandeen12062dd2010-02-15 14:19:27 -05004643 ext4_warning(sb, "Marking fs in need of filesystem check.");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004644
Mingming Cao617ba132006-10-11 01:20:53 -07004645 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4646 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004647 ext4_commit_super(sb, 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004648
Mingming Caodab291a2006-10-11 01:21:01 -07004649 jbd2_journal_clear_err(journal);
Theodore Ts'od796c522012-08-05 19:04:57 -04004650 jbd2_journal_update_sb_errno(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004651 }
4652}
4653
4654/*
4655 * Force the running and committing transactions to commit,
4656 * and wait on the commit.
4657 */
Mingming Cao617ba132006-10-11 01:20:53 -07004658int ext4_force_commit(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004659{
4660 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004661
4662 if (sb->s_flags & MS_RDONLY)
4663 return 0;
4664
Mingming Cao617ba132006-10-11 01:20:53 -07004665 journal = EXT4_SB(sb)->s_journal;
Guo Chaob1deefc2013-01-28 21:41:02 -05004666 return ext4_journal_force_commit(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004667}
4668
Mingming Cao617ba132006-10-11 01:20:53 -07004669static int ext4_sync_fs(struct super_block *sb, int wait)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004670{
Theodore Ts'o14ce0cb2008-11-03 18:10:55 -05004671 int ret = 0;
Jan Kara9eddacf2009-02-10 06:46:05 -05004672 tid_t target;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004673 bool needs_barrier = false;
Mingming Cao8d5d02e2009-09-28 15:48:29 -04004674 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004675
Theodore Ts'o9bffad12009-06-17 11:48:11 -04004676 trace_ext4_sync_fs(sb, wait);
Jan Kara2e8fa542013-06-04 14:21:02 -04004677 flush_workqueue(sbi->rsv_conversion_wq);
Jan Karaa1177822012-07-03 16:45:29 +02004678 /*
4679 * Writeback quota in non-journalled quota case - journalled quota has
4680 * no dirty dquots
4681 */
4682 dquot_writeback_dquots(sb, -1);
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004683 /*
4684 * Data writeback is possible w/o journal transaction, so barrier must
4685 * being sent at the end of the function. But we can skip it if
4686 * transaction_commit will do it for us.
4687 */
Theodore Ts'obda32532014-09-18 16:12:37 -04004688 if (sbi->s_journal) {
4689 target = jbd2_get_latest_transaction(sbi->s_journal);
4690 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4691 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4692 needs_barrier = true;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004693
Theodore Ts'obda32532014-09-18 16:12:37 -04004694 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4695 if (wait)
4696 ret = jbd2_log_wait_commit(sbi->s_journal,
4697 target);
4698 }
4699 } else if (wait && test_opt(sb, BARRIER))
4700 needs_barrier = true;
Dmitry Monakhov06a407f2013-06-12 22:25:07 -04004701 if (needs_barrier) {
4702 int err;
4703 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4704 if (!ret)
4705 ret = err;
4706 }
4707
4708 return ret;
4709}
4710
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004711/*
4712 * LVM calls this function before a (read-only) snapshot is created. This
4713 * gives us a chance to flush the journal completely and mark the fs clean.
Yongqiang Yangbe4f27d2011-04-10 22:06:07 -04004714 *
4715 * Note that only this function cannot bring a filesystem to be in a clean
Jan Kara8e8ad8a2012-06-12 16:20:38 +02004716 * state independently. It relies on upper layer to stop all data & metadata
4717 * modifications.
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004718 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004719static int ext4_freeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004720{
Takashi Satoc4be0c12009-01-09 16:40:58 -08004721 int error = 0;
4722 journal_t *journal;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004723
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004724 if (sb->s_flags & MS_RDONLY)
4725 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004726
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004727 journal = EXT4_SB(sb)->s_journal;
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04004728
Theodore Ts'obb044572014-09-18 17:12:02 -04004729 if (journal) {
4730 /* Now we set up the journal barrier. */
4731 jbd2_journal_lock_updates(journal);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004732
Theodore Ts'obb044572014-09-18 17:12:02 -04004733 /*
4734 * Don't clear the needs_recovery flag if we failed to
4735 * flush the journal.
4736 */
4737 error = jbd2_journal_flush(journal);
4738 if (error < 0)
4739 goto out;
Eric Sandeenc642dc92015-08-15 10:45:06 -04004740
4741 /* Journal blocked and flushed, clear needs_recovery flag. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004742 ext4_clear_feature_journal_needs_recovery(sb);
Theodore Ts'obb044572014-09-18 17:12:02 -04004743 }
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004744
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004745 error = ext4_commit_super(sb, 1);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004746out:
Theodore Ts'obb044572014-09-18 17:12:02 -04004747 if (journal)
4748 /* we rely on upper layer to stop further updates */
4749 jbd2_journal_unlock_updates(journal);
Eric Sandeen6b0310f2010-05-16 02:00:00 -04004750 return error;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004751}
4752
4753/*
4754 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4755 * flag here, even though the filesystem is not technically dirty yet.
4756 */
Takashi Satoc4be0c12009-01-09 16:40:58 -08004757static int ext4_unfreeze(struct super_block *sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004758{
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004759 if (sb->s_flags & MS_RDONLY)
4760 return 0;
4761
Eric Sandeenc642dc92015-08-15 10:45:06 -04004762 if (EXT4_SB(sb)->s_journal) {
4763 /* Reset the needs_recovery flag before the fs is unlocked. */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004764 ext4_set_feature_journal_needs_recovery(sb);
Eric Sandeenc642dc92015-08-15 10:45:06 -04004765 }
4766
Theodore Ts'o9ca92382009-05-01 12:52:25 -04004767 ext4_commit_super(sb, 1);
Takashi Satoc4be0c12009-01-09 16:40:58 -08004768 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004769}
4770
Theodore Ts'o673c6102010-12-15 20:28:48 -05004771/*
4772 * Structure to save mount options for ext4_remount's benefit
4773 */
4774struct ext4_mount_options {
4775 unsigned long s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004776 unsigned long s_mount_opt2;
Eric W. Biederman08cefc72012-02-07 15:41:49 -08004777 kuid_t s_resuid;
4778 kgid_t s_resgid;
Theodore Ts'o673c6102010-12-15 20:28:48 -05004779 unsigned long s_commit_interval;
4780 u32 s_min_batch_time, s_max_batch_time;
4781#ifdef CONFIG_QUOTA
4782 int s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04004783 char *s_qf_names[EXT4_MAXQUOTAS];
Theodore Ts'o673c6102010-12-15 20:28:48 -05004784#endif
4785};
4786
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004787static int ext4_remount(struct super_block *sb, int *flags, char *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004788{
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004789 struct ext4_super_block *es;
Mingming Cao617ba132006-10-11 01:20:53 -07004790 struct ext4_sb_info *sbi = EXT4_SB(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004791 unsigned long old_sb_flags;
Mingming Cao617ba132006-10-11 01:20:53 -07004792 struct ext4_mount_options old_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004793 int enable_quota = 0;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004794 ext4_group_t g;
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004795 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
Johann Lombardic5e06d12011-05-24 18:31:25 -04004796 int err = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004797#ifdef CONFIG_QUOTA
Chen Gang03dafb52013-01-24 23:24:58 -05004798 int i, j;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004799#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04004800 char *orig_data = kstrdup(data, GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004801
4802 /* Store the original options */
4803 old_sb_flags = sb->s_flags;
4804 old_opts.s_mount_opt = sbi->s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05004805 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004806 old_opts.s_resuid = sbi->s_resuid;
4807 old_opts.s_resgid = sbi->s_resgid;
4808 old_opts.s_commit_interval = sbi->s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05004809 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4810 old_opts.s_max_batch_time = sbi->s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004811#ifdef CONFIG_QUOTA
4812 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04004813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05004814 if (sbi->s_qf_names[i]) {
4815 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4816 GFP_KERNEL);
4817 if (!old_opts.s_qf_names[i]) {
4818 for (j = 0; j < i; j++)
4819 kfree(old_opts.s_qf_names[j]);
Wei Yongjun3e36a162013-03-02 17:13:55 -05004820 kfree(orig_data);
Chen Gang03dafb52013-01-24 23:24:58 -05004821 return -ENOMEM;
4822 }
4823 } else
4824 old_opts.s_qf_names[i] = NULL;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004825#endif
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004826 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4827 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004828
Eric Sandeen661aa522012-02-20 17:53:04 -05004829 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004830 err = -EINVAL;
4831 goto restore_opts;
4832 }
4833
Darrick J. Wong6b992ff2014-10-30 10:53:16 -04004834 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4835 test_opt(sb, JOURNAL_CHECKSUM)) {
4836 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
Eric Sandeen2d5b86e2015-02-12 23:07:37 -05004837 "during remount not supported; ignoring");
4838 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
Darrick J. Wongc6d3d562014-11-25 16:20:50 -05004839 }
4840
Piotr Sarna6ae65142013-08-08 23:02:24 -04004841 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4842 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4843 ext4_msg(sb, KERN_ERR, "can't mount with "
4844 "both data=journal and delalloc");
4845 err = -EINVAL;
4846 goto restore_opts;
4847 }
4848 if (test_opt(sb, DIOREAD_NOLOCK)) {
4849 ext4_msg(sb, KERN_ERR, "can't mount with "
4850 "both data=journal and dioread_nolock");
4851 err = -EINVAL;
4852 goto restore_opts;
4853 }
Ross Zwisler923ae0f2015-02-16 15:59:38 -08004854 if (test_opt(sb, DAX)) {
4855 ext4_msg(sb, KERN_ERR, "can't mount with "
4856 "both data=journal and dax");
4857 err = -EINVAL;
4858 goto restore_opts;
4859 }
4860 }
4861
4862 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4863 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4864 "dax flag with busy inodes while remounting");
4865 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
Piotr Sarna6ae65142013-08-08 23:02:24 -04004866 }
4867
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004868 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
Theodore Ts'oc67d8592010-06-29 11:07:07 -04004869 ext4_abort(sb, "Abort forced by user");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004870
4871 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
Dmitry Monakhov482a7422010-02-24 11:35:32 -05004872 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004873
4874 es = sbi->s_es;
4875
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004876 if (sbi->s_journal) {
Frank Mayhar03901312009-01-07 00:06:22 -05004877 ext4_init_journal_params(sb, sbi->s_journal);
Theodore Ts'ob3881f72009-01-05 22:46:26 -05004878 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4879 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004880
Theodore Ts'oa2fd66d2015-06-23 11:03:54 -04004881 if (*flags & MS_LAZYTIME)
4882 sb->s_flags |= MS_LAZYTIME;
4883
Eric Sandeen661aa522012-02-20 17:53:04 -05004884 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
Theodore Ts'o4ab2f152009-06-13 10:09:36 -04004885 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004886 err = -EROFS;
4887 goto restore_opts;
4888 }
4889
4890 if (*flags & MS_RDONLY) {
Theodore Ts'o38c03b32014-03-13 22:49:42 -04004891 err = sync_filesystem(sb);
4892 if (err < 0)
4893 goto restore_opts;
Christoph Hellwig0f0dd622010-05-19 07:16:41 -04004894 err = dquot_suspend(sb, -1);
4895 if (err < 0)
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004896 goto restore_opts;
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004897
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004898 /*
4899 * First of all, the unconditional stuff we have to do
4900 * to disable replay of the journal when we next remount
4901 */
4902 sb->s_flags |= MS_RDONLY;
4903
4904 /*
4905 * OK, test if we are remounting a valid rw partition
4906 * readonly, and if so set the rdonly flag and then
4907 * mark the partition as valid again.
4908 */
Mingming Cao617ba132006-10-11 01:20:53 -07004909 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4910 (sbi->s_mount_state & EXT4_VALID_FS))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004911 es->s_state = cpu_to_le16(sbi->s_mount_state);
4912
Theodore Ts'oa63c9eb2009-05-01 01:59:42 -04004913 if (sbi->s_journal)
Frank Mayhar03901312009-01-07 00:06:22 -05004914 ext4_mark_recovery_complete(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004915 } else {
Eric Sandeena13fb1a2009-08-18 00:20:23 -04004916 /* Make sure we can mount this feature set readwrite */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004917 if (ext4_has_feature_readonly(sb) ||
Darrick J. Wong2cb5cc82015-02-12 22:31:21 -05004918 !ext4_feature_set_ok(sb, 0)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004919 err = -EROFS;
4920 goto restore_opts;
4921 }
Eric Sandeenead65962007-02-10 01:46:08 -08004922 /*
Theodore Ts'o8a266462008-07-26 14:34:21 -04004923 * Make sure the group descriptor checksums
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04004924 * are sane. If they aren't, refuse to remount r/w.
Theodore Ts'o8a266462008-07-26 14:34:21 -04004925 */
4926 for (g = 0; g < sbi->s_groups_count; g++) {
4927 struct ext4_group_desc *gdp =
4928 ext4_get_group_desc(sb, g, NULL);
4929
Darrick J. Wongfeb0ab32012-04-29 18:45:10 -04004930 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004931 ext4_msg(sb, KERN_ERR,
4932 "ext4_remount: Checksum for group %u failed (%u!=%u)",
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004933 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
Theodore Ts'o8a266462008-07-26 14:34:21 -04004934 le16_to_cpu(gdp->bg_checksum));
Darrick J. Wong6a797d22015-10-17 16:16:04 -04004935 err = -EFSBADCRC;
Theodore Ts'o8a266462008-07-26 14:34:21 -04004936 goto restore_opts;
4937 }
4938 }
4939
4940 /*
Eric Sandeenead65962007-02-10 01:46:08 -08004941 * If we have an unprocessed orphan list hanging
4942 * around from a previously readonly bdev mount,
4943 * require a full umount/remount for now.
4944 */
4945 if (es->s_last_orphan) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04004946 ext4_msg(sb, KERN_WARNING, "Couldn't "
Eric Sandeenead65962007-02-10 01:46:08 -08004947 "remount RDWR because of unprocessed "
4948 "orphan inode list. Please "
Eric Sandeenb31e1552009-06-04 17:36:36 -04004949 "umount/remount instead");
Eric Sandeenead65962007-02-10 01:46:08 -08004950 err = -EINVAL;
4951 goto restore_opts;
4952 }
4953
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004954 /*
4955 * Mounting a RDONLY partition read-write, so reread
4956 * and store the current valid flag. (It may have
4957 * been changed by e2fsck since we originally mounted
4958 * the partition.)
4959 */
Frank Mayhar03901312009-01-07 00:06:22 -05004960 if (sbi->s_journal)
4961 ext4_clear_journal_err(sb, es);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004962 sbi->s_mount_state = le16_to_cpu(es->s_state);
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04004963 if (!ext4_setup_super(sb, es, 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004964 sb->s_flags &= ~MS_RDONLY;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004965 if (ext4_has_feature_mmp(sb))
Johann Lombardic5e06d12011-05-24 18:31:25 -04004966 if (ext4_multi_mount_protect(sb,
4967 le64_to_cpu(es->s_mmp_block))) {
4968 err = -EROFS;
4969 goto restore_opts;
4970 }
Christoph Hellwigc79d9672010-05-19 07:16:40 -04004971 enable_quota = 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004972 }
4973 }
Lukas Czernerbfff6872010-10-27 21:30:05 -04004974
4975 /*
4976 * Reinitialize lazy itable initialization thread based on
4977 * current settings
4978 */
4979 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4980 ext4_unregister_li_request(sb);
4981 else {
4982 ext4_group_t first_not_zeroed;
4983 first_not_zeroed = ext4_has_uninit_itable(sb);
4984 ext4_register_li_request(sb, first_not_zeroed);
4985 }
4986
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04004987 ext4_setup_system_zone(sb);
Michael Tokarevd096ad02012-12-25 14:08:16 -05004988 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
Theodore Ts'oe2d67052009-05-01 00:33:44 -04004989 ext4_commit_super(sb, 1);
Frank Mayhar03901312009-01-07 00:06:22 -05004990
Dave Kleikampac27a0e2006-10-11 01:20:50 -07004991#ifdef CONFIG_QUOTA
4992 /* Release old quota file names */
Jan Karaa2d4a642014-09-11 11:15:15 -04004993 for (i = 0; i < EXT4_MAXQUOTAS; i++)
Chen Gang03dafb52013-01-24 23:24:58 -05004994 kfree(old_opts.s_qf_names[i]);
Aditya Kali7c319d32012-07-22 20:21:31 -04004995 if (enable_quota) {
4996 if (sb_any_quota_suspended(sb))
4997 dquot_resume(sb, -1);
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04004998 else if (ext4_has_feature_quota(sb)) {
Aditya Kali7c319d32012-07-22 20:21:31 -04004999 err = ext4_enable_quotas(sb);
Theodore Ts'o07724f92012-08-17 19:08:42 -04005000 if (err)
Aditya Kali7c319d32012-07-22 20:21:31 -04005001 goto restore_opts;
Aditya Kali7c319d32012-07-22 20:21:31 -04005002 }
5003 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005004#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005005
Theodore Ts'oa26f4992015-02-02 00:37:02 -05005006 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005007 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5008 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005009 return 0;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005010
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005011restore_opts:
5012 sb->s_flags = old_sb_flags;
5013 sbi->s_mount_opt = old_opts.s_mount_opt;
Theodore Ts'oa2595b82010-12-15 20:30:48 -05005014 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005015 sbi->s_resuid = old_opts.s_resuid;
5016 sbi->s_resgid = old_opts.s_resgid;
5017 sbi->s_commit_interval = old_opts.s_commit_interval;
Theodore Ts'o30773842009-01-03 20:27:38 -05005018 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5019 sbi->s_max_batch_time = old_opts.s_max_batch_time;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005020#ifdef CONFIG_QUOTA
5021 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
Jan Karaa2d4a642014-09-11 11:15:15 -04005022 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
Chen Gang03dafb52013-01-24 23:24:58 -05005023 kfree(sbi->s_qf_names[i]);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005024 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5025 }
5026#endif
Curt Wohlgemuthd4c402d2010-05-16 12:00:00 -04005027 kfree(orig_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005028 return err;
5029}
5030
Li Xi689c9582016-01-08 16:01:22 -05005031#ifdef CONFIG_QUOTA
5032static int ext4_statfs_project(struct super_block *sb,
5033 kprojid_t projid, struct kstatfs *buf)
5034{
5035 struct kqid qid;
5036 struct dquot *dquot;
5037 u64 limit;
5038 u64 curblock;
5039
5040 qid = make_kqid_projid(projid);
5041 dquot = dqget(sb, qid);
5042 if (IS_ERR(dquot))
5043 return PTR_ERR(dquot);
5044 spin_lock(&dq_data_lock);
5045
5046 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5047 dquot->dq_dqb.dqb_bsoftlimit :
5048 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5049 if (limit && buf->f_blocks > limit) {
5050 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5051 buf->f_blocks = limit;
5052 buf->f_bfree = buf->f_bavail =
5053 (buf->f_blocks > curblock) ?
5054 (buf->f_blocks - curblock) : 0;
5055 }
5056
5057 limit = dquot->dq_dqb.dqb_isoftlimit ?
5058 dquot->dq_dqb.dqb_isoftlimit :
5059 dquot->dq_dqb.dqb_ihardlimit;
5060 if (limit && buf->f_files > limit) {
5061 buf->f_files = limit;
5062 buf->f_ffree =
5063 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5064 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5065 }
5066
5067 spin_unlock(&dq_data_lock);
5068 dqput(dquot);
5069 return 0;
5070}
5071#endif
5072
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005073static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005074{
5075 struct super_block *sb = dentry->d_sb;
Mingming Cao617ba132006-10-11 01:20:53 -07005076 struct ext4_sb_info *sbi = EXT4_SB(sb);
5077 struct ext4_super_block *es = sbi->s_es;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005078 ext4_fsblk_t overhead = 0, resv_blocks;
Pekka Enberg960cc392006-12-06 20:35:29 -08005079 u64 fsid;
Kazuya Miod02a9392011-05-24 18:30:07 -04005080 s64 bfree;
Lukas Czerner27dd4382013-04-09 22:11:22 -04005081 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005082
Theodore Ts'o952fc182012-07-09 16:27:05 -04005083 if (!test_opt(sb, MINIX_DF))
5084 overhead = sbi->s_overhead;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005085
Mingming Cao617ba132006-10-11 01:20:53 -07005086 buf->f_type = EXT4_SUPER_MAGIC;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005087 buf->f_bsize = sb->s_blocksize;
Eric Sandeenb72f78c2012-11-08 10:33:36 -05005088 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
Theodore Ts'o57042652011-09-09 18:56:51 -04005089 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5090 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
Kazuya Miod02a9392011-05-24 18:30:07 -04005091 /* prevent underflow in case that few free space is available */
Theodore Ts'o57042652011-09-09 18:56:51 -04005092 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
Lukas Czerner27dd4382013-04-09 22:11:22 -04005093 buf->f_bavail = buf->f_bfree -
5094 (ext4_r_blocks_count(es) + resv_blocks);
5095 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005096 buf->f_bavail = 0;
5097 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07005098 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Mingming Cao617ba132006-10-11 01:20:53 -07005099 buf->f_namelen = EXT4_NAME_LEN;
Pekka Enberg960cc392006-12-06 20:35:29 -08005100 fsid = le64_to_cpup((void *)es->s_uuid) ^
5101 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5102 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5103 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005104
Li Xi689c9582016-01-08 16:01:22 -05005105#ifdef CONFIG_QUOTA
5106 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5107 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5108 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5109#endif
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005110 return 0;
5111}
5112
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005113/* Helper function for writing quotas on sync - we need to start transaction
5114 * before quota file is locked for write. Otherwise the are possible deadlocks:
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005115 * Process 1 Process 2
Mingming Cao617ba132006-10-11 01:20:53 -07005116 * ext4_create() quota_sync()
Jan Karaa269eb12009-01-26 17:04:39 +01005117 * jbd2_journal_start() write_dquot()
Christoph Hellwig871a2932010-03-03 09:05:07 -05005118 * dquot_initialize() down(dqio_mutex)
Mingming Caodab291a2006-10-11 01:21:01 -07005119 * down(dqio_mutex) jbd2_journal_start()
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005120 *
5121 */
5122
5123#ifdef CONFIG_QUOTA
5124
5125static inline struct inode *dquot_to_inode(struct dquot *dquot)
5126{
Eric W. Biederman4c376dc2012-09-16 03:56:19 -07005127 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005128}
5129
Mingming Cao617ba132006-10-11 01:20:53 -07005130static int ext4_write_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005131{
5132 int ret, err;
5133 handle_t *handle;
5134 struct inode *inode;
5135
5136 inode = dquot_to_inode(dquot);
Theodore Ts'o9924a922013-02-08 21:59:22 -05005137 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005138 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005139 if (IS_ERR(handle))
5140 return PTR_ERR(handle);
5141 ret = dquot_commit(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005142 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005143 if (!ret)
5144 ret = err;
5145 return ret;
5146}
5147
Mingming Cao617ba132006-10-11 01:20:53 -07005148static int ext4_acquire_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005149{
5150 int ret, err;
5151 handle_t *handle;
5152
Theodore Ts'o9924a922013-02-08 21:59:22 -05005153 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005154 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005155 if (IS_ERR(handle))
5156 return PTR_ERR(handle);
5157 ret = dquot_acquire(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005158 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005159 if (!ret)
5160 ret = err;
5161 return ret;
5162}
5163
Mingming Cao617ba132006-10-11 01:20:53 -07005164static int ext4_release_dquot(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005165{
5166 int ret, err;
5167 handle_t *handle;
5168
Theodore Ts'o9924a922013-02-08 21:59:22 -05005169 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
Andreas Dilger0b8e58a2009-06-03 17:59:28 -04005170 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07005171 if (IS_ERR(handle)) {
5172 /* Release dquot anyway to avoid endless cycle in dqput() */
5173 dquot_release(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005174 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07005175 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005176 ret = dquot_release(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005177 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005178 if (!ret)
5179 ret = err;
5180 return ret;
5181}
5182
Mingming Cao617ba132006-10-11 01:20:53 -07005183static int ext4_mark_dquot_dirty(struct dquot *dquot)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005184{
Jan Kara262b4662013-03-02 17:57:08 -05005185 struct super_block *sb = dquot->dq_sb;
5186 struct ext4_sb_info *sbi = EXT4_SB(sb);
5187
Jan Kara2c8be6b2008-05-13 21:27:55 -04005188 /* Are we journaling quotas? */
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005189 if (ext4_has_feature_quota(sb) ||
Jan Kara262b4662013-03-02 17:57:08 -05005190 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005191 dquot_mark_dquot_dirty(dquot);
Mingming Cao617ba132006-10-11 01:20:53 -07005192 return ext4_write_dquot(dquot);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005193 } else {
5194 return dquot_mark_dquot_dirty(dquot);
5195 }
5196}
5197
Mingming Cao617ba132006-10-11 01:20:53 -07005198static int ext4_write_info(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005199{
5200 int ret, err;
5201 handle_t *handle;
5202
5203 /* Data block + inode block */
David Howells2b0143b2015-03-17 22:25:59 +00005204 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005205 if (IS_ERR(handle))
5206 return PTR_ERR(handle);
5207 ret = dquot_commit_info(sb, type);
Mingming Cao617ba132006-10-11 01:20:53 -07005208 err = ext4_journal_stop(handle);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005209 if (!ret)
5210 ret = err;
5211 return ret;
5212}
5213
5214/*
5215 * Turn on quotas during mount time - we need to find
5216 * the quota file and such...
5217 */
Mingming Cao617ba132006-10-11 01:20:53 -07005218static int ext4_quota_on_mount(struct super_block *sb, int type)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005219{
Christoph Hellwig287a8092010-05-19 07:16:45 -04005220 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5221 EXT4_SB(sb)->s_jquota_fmt, type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005222}
5223
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005224static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5225{
5226 struct ext4_inode_info *ei = EXT4_I(inode);
5227
5228 /* The first argument of lockdep_set_subclass has to be
5229 * *exactly* the same as the argument to init_rwsem() --- in
5230 * this case, in init_once() --- or lockdep gets unhappy
5231 * because the name of the lock is set using the
5232 * stringification of the argument to init_rwsem().
5233 */
5234 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5235 lockdep_set_subclass(&ei->i_data_sem, subclass);
5236}
5237
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005238/*
5239 * Standard function to be called on quota_on
5240 */
Mingming Cao617ba132006-10-11 01:20:53 -07005241static int ext4_quota_on(struct super_block *sb, int type, int format_id,
Jan Karaf00c9e42010-09-15 17:38:58 +02005242 struct path *path)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005243{
5244 int err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005245
5246 if (!test_opt(sb, QUOTA))
5247 return -EINVAL;
Jan Kara06235432008-05-13 19:11:51 -04005248
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005249 /* Quotafile not on the same filesystem? */
Al Virod8c95842011-12-07 18:16:57 -05005250 if (path->dentry->d_sb != sb)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005251 return -EXDEV;
Jan Kara06235432008-05-13 19:11:51 -04005252 /* Journaling quota? */
5253 if (EXT4_SB(sb)->s_qf_names[type]) {
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005254 /* Quotafile not in fs root? */
Jan Karaf00c9e42010-09-15 17:38:58 +02005255 if (path->dentry->d_parent != sb->s_root)
Eric Sandeenb31e1552009-06-04 17:36:36 -04005256 ext4_msg(sb, KERN_WARNING,
5257 "Quota file not on filesystem root. "
5258 "Journaled quota will not work");
Theodore Ts'o2b2d6d02008-07-26 16:15:44 -04005259 }
Jan Kara06235432008-05-13 19:11:51 -04005260
5261 /*
5262 * When we journal data on quota file, we have to flush journal to see
5263 * all updates to the file when we bypass pagecache...
5264 */
Frank Mayhar03901312009-01-07 00:06:22 -05005265 if (EXT4_SB(sb)->s_journal &&
David Howells2b0143b2015-03-17 22:25:59 +00005266 ext4_should_journal_data(d_inode(path->dentry))) {
Jan Kara06235432008-05-13 19:11:51 -04005267 /*
5268 * We don't need to lock updates but journal_flush() could
5269 * otherwise be livelocked...
5270 */
5271 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005272 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
Jan Kara06235432008-05-13 19:11:51 -04005273 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
Jan Karaf00c9e42010-09-15 17:38:58 +02005274 if (err)
Hidehiro Kawai7ffe1ea2008-10-10 20:29:21 -04005275 return err;
Jan Kara06235432008-05-13 19:11:51 -04005276 }
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005277 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5278 err = dquot_quota_on(sb, type, format_id, path);
5279 if (err)
5280 lockdep_set_quota_inode(path->dentry->d_inode,
5281 I_DATA_SEM_NORMAL);
5282 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005283}
5284
Aditya Kali7c319d32012-07-22 20:21:31 -04005285static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5286 unsigned int flags)
5287{
5288 int err;
5289 struct inode *qf_inode;
Jan Karaa2d4a642014-09-11 11:15:15 -04005290 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
Aditya Kali7c319d32012-07-22 20:21:31 -04005291 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
Li Xi689c9582016-01-08 16:01:22 -05005292 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5293 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
Aditya Kali7c319d32012-07-22 20:21:31 -04005294 };
5295
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005296 BUG_ON(!ext4_has_feature_quota(sb));
Aditya Kali7c319d32012-07-22 20:21:31 -04005297
5298 if (!qf_inums[type])
5299 return -EPERM;
5300
5301 qf_inode = ext4_iget(sb, qf_inums[type]);
5302 if (IS_ERR(qf_inode)) {
5303 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5304 return PTR_ERR(qf_inode);
5305 }
5306
Jan Karabcb13852013-04-09 09:21:41 -04005307 /* Don't account quota for quota files to avoid recursion */
5308 qf_inode->i_flags |= S_NOQUOTA;
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005309 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
Aditya Kali7c319d32012-07-22 20:21:31 -04005310 err = dquot_enable(qf_inode, type, format_id, flags);
5311 iput(qf_inode);
Theodore Ts'odaf647d2016-04-01 01:31:28 -04005312 if (err)
5313 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
Aditya Kali7c319d32012-07-22 20:21:31 -04005314
5315 return err;
5316}
5317
5318/* Enable usage tracking for all quota types. */
5319static int ext4_enable_quotas(struct super_block *sb)
5320{
5321 int type, err = 0;
Jan Karaa2d4a642014-09-11 11:15:15 -04005322 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
Aditya Kali7c319d32012-07-22 20:21:31 -04005323 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
Li Xi689c9582016-01-08 16:01:22 -05005324 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5325 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
Aditya Kali7c319d32012-07-22 20:21:31 -04005326 };
Jan Kara49da9392016-09-05 23:08:16 -04005327 bool quota_mopt[EXT4_MAXQUOTAS] = {
5328 test_opt(sb, USRQUOTA),
5329 test_opt(sb, GRPQUOTA),
5330 test_opt(sb, PRJQUOTA),
5331 };
Aditya Kali7c319d32012-07-22 20:21:31 -04005332
5333 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
Jan Karaa2d4a642014-09-11 11:15:15 -04005334 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
Aditya Kali7c319d32012-07-22 20:21:31 -04005335 if (qf_inums[type]) {
5336 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
Jan Kara49da9392016-09-05 23:08:16 -04005337 DQUOT_USAGE_ENABLED |
5338 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
Aditya Kali7c319d32012-07-22 20:21:31 -04005339 if (err) {
5340 ext4_warning(sb,
Theodore Ts'o72ba7452013-01-24 23:24:54 -05005341 "Failed to enable quota tracking "
5342 "(type=%d, err=%d). Please run "
5343 "e2fsck to fix.", type, err);
Aditya Kali7c319d32012-07-22 20:21:31 -04005344 return err;
5345 }
5346 }
5347 }
5348 return 0;
5349}
5350
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005351static int ext4_quota_off(struct super_block *sb, int type)
5352{
Jan Kara21f97692011-04-04 15:33:39 -04005353 struct inode *inode = sb_dqopt(sb)->files[type];
5354 handle_t *handle;
5355
Dmitry Monakhov87009d82010-11-08 13:47:33 -05005356 /* Force all delayed allocation blocks to be allocated.
5357 * Caller already holds s_umount sem */
5358 if (test_opt(sb, DELALLOC))
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005359 sync_filesystem(sb);
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005360
Amir Goldstein0b268592011-05-16 09:59:13 -04005361 if (!inode)
5362 goto out;
5363
Jan Kara21f97692011-04-04 15:33:39 -04005364 /* Update modification times of quota files when userspace can
5365 * start looking at them */
Theodore Ts'o9924a922013-02-08 21:59:22 -05005366 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
Jan Kara21f97692011-04-04 15:33:39 -04005367 if (IS_ERR(handle))
5368 goto out;
5369 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5370 ext4_mark_inode_dirty(handle, inode);
5371 ext4_journal_stop(handle);
5372
5373out:
Dmitry Monakhovca0e05e2010-08-01 17:48:36 -04005374 return dquot_quota_off(sb, type);
5375}
5376
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005377/* Read data from quotafile - avoid pagecache and such because we cannot afford
5378 * acquiring the locks... As quota files are never truncated and quota code
Lucas De Marchi25985ed2011-03-30 22:57:33 -03005379 * itself serializes the operations (and no one else should touch the files)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005380 * we don't have to be afraid of races */
Mingming Cao617ba132006-10-11 01:20:53 -07005381static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005382 size_t len, loff_t off)
5383{
5384 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005385 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005386 int offset = off & (sb->s_blocksize - 1);
5387 int tocopy;
5388 size_t toread;
5389 struct buffer_head *bh;
5390 loff_t i_size = i_size_read(inode);
5391
5392 if (off > i_size)
5393 return 0;
5394 if (off+len > i_size)
5395 len = i_size-off;
5396 toread = len;
5397 while (toread > 0) {
5398 tocopy = sb->s_blocksize - offset < toread ?
5399 sb->s_blocksize - offset : toread;
Theodore Ts'o1c215022014-08-29 20:52:15 -04005400 bh = ext4_bread(NULL, inode, blk, 0);
5401 if (IS_ERR(bh))
5402 return PTR_ERR(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005403 if (!bh) /* A hole? */
5404 memset(data, 0, tocopy);
5405 else
5406 memcpy(data, bh->b_data+offset, tocopy);
5407 brelse(bh);
5408 offset = 0;
5409 toread -= tocopy;
5410 data += tocopy;
5411 blk++;
5412 }
5413 return len;
5414}
5415
5416/* Write to quotafile (we know the transaction is already started and has
5417 * enough credits) */
Mingming Cao617ba132006-10-11 01:20:53 -07005418static ssize_t ext4_quota_write(struct super_block *sb, int type,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005419 const char *data, size_t len, loff_t off)
5420{
5421 struct inode *inode = sb_dqopt(sb)->files[type];
Aneesh Kumar K.V725d26d2008-01-28 23:58:27 -05005422 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
Theodore Ts'o1c215022014-08-29 20:52:15 -04005423 int err, offset = off & (sb->s_blocksize - 1);
Theodore Ts'oc5e298a2015-06-21 01:25:29 -04005424 int retries = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005425 struct buffer_head *bh;
5426 handle_t *handle = journal_current_handle();
5427
Frank Mayhar03901312009-01-07 00:06:22 -05005428 if (EXT4_SB(sb)->s_journal && !handle) {
Eric Sandeenb31e1552009-06-04 17:36:36 -04005429 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5430 " cancelled because transaction is not started",
Jan Kara9c3013e2007-09-11 15:23:29 -07005431 (unsigned long long)off, (unsigned long long)len);
5432 return -EIO;
5433 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005434 /*
5435 * Since we account only one data block in transaction credits,
5436 * then it is impossible to cross a block boundary.
5437 */
5438 if (sb->s_blocksize - offset < len) {
5439 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5440 " cancelled because not block aligned",
5441 (unsigned long long)off, (unsigned long long)len);
5442 return -EIO;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005443 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005444
Theodore Ts'oc5e298a2015-06-21 01:25:29 -04005445 do {
5446 bh = ext4_bread(handle, inode, blk,
5447 EXT4_GET_BLOCKS_CREATE |
5448 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5449 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5450 ext4_should_retry_alloc(inode->i_sb, &retries));
Theodore Ts'o1c215022014-08-29 20:52:15 -04005451 if (IS_ERR(bh))
5452 return PTR_ERR(bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005453 if (!bh)
5454 goto out;
liang xie5d601252014-05-12 22:06:43 -04005455 BUFFER_TRACE(bh, "get write access");
Jan Kara62d2b5f2010-07-27 11:56:07 -04005456 err = ext4_journal_get_write_access(handle, bh);
5457 if (err) {
5458 brelse(bh);
Theodore Ts'o1c215022014-08-29 20:52:15 -04005459 return err;
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005460 }
5461 lock_buffer(bh);
5462 memcpy(bh->b_data+offset, data, len);
5463 flush_dcache_page(bh->b_page);
5464 unlock_buffer(bh);
Jan Kara62d2b5f2010-07-27 11:56:07 -04005465 err = ext4_handle_dirty_metadata(handle, NULL, bh);
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005466 brelse(bh);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005467out:
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005468 if (inode->i_size < off + len) {
5469 i_size_write(inode, off + len);
Mingming Cao617ba132006-10-11 01:20:53 -07005470 EXT4_I(inode)->i_disksize = inode->i_size;
Jan Kara21f97692011-04-04 15:33:39 -04005471 ext4_mark_inode_dirty(handle, inode);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005472 }
Dmitry Monakhov67eeb562010-03-02 08:08:51 -05005473 return len;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005474}
5475
Theodore Ts'o8f0e8742016-04-01 12:00:03 -04005476static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5477{
5478 const struct quota_format_ops *ops;
5479
5480 if (!sb_has_quota_loaded(sb, qid->type))
5481 return -ESRCH;
5482 ops = sb_dqopt(sb)->ops[qid->type];
5483 if (!ops || !ops->get_next_id)
5484 return -ENOSYS;
5485 return dquot_get_next_id(sb, qid);
5486}
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005487#endif
5488
Al Viro152a0832010-07-25 00:46:55 +04005489static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5490 const char *dev_name, void *data)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005491{
Al Viro152a0832010-07-25 00:46:55 +04005492 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005493}
5494
Jan Karac290ea02015-06-18 16:52:29 +02005495#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
Theodore Ts'o24b58422009-12-07 14:08:51 -05005496static inline void register_as_ext2(void)
5497{
5498 int err = register_filesystem(&ext2_fs_type);
5499 if (err)
5500 printk(KERN_WARNING
5501 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5502}
5503
5504static inline void unregister_as_ext2(void)
5505{
5506 unregister_filesystem(&ext2_fs_type);
5507}
Theodore Ts'o2035e772011-04-18 17:29:14 -04005508
5509static inline int ext2_feature_set_ok(struct super_block *sb)
5510{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005511 if (ext4_has_unknown_ext2_incompat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04005512 return 0;
5513 if (sb->s_flags & MS_RDONLY)
5514 return 1;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005515 if (ext4_has_unknown_ext2_ro_compat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04005516 return 0;
5517 return 1;
5518}
Theodore Ts'o24b58422009-12-07 14:08:51 -05005519#else
5520static inline void register_as_ext2(void) { }
5521static inline void unregister_as_ext2(void) { }
Theodore Ts'o2035e772011-04-18 17:29:14 -04005522static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
Theodore Ts'o24b58422009-12-07 14:08:51 -05005523#endif
5524
Theodore Ts'o24b58422009-12-07 14:08:51 -05005525static inline void register_as_ext3(void)
5526{
5527 int err = register_filesystem(&ext3_fs_type);
5528 if (err)
5529 printk(KERN_WARNING
5530 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5531}
5532
5533static inline void unregister_as_ext3(void)
5534{
5535 unregister_filesystem(&ext3_fs_type);
5536}
Theodore Ts'o2035e772011-04-18 17:29:14 -04005537
5538static inline int ext3_feature_set_ok(struct super_block *sb)
5539{
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005540 if (ext4_has_unknown_ext3_incompat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04005541 return 0;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005542 if (!ext4_has_feature_journal(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04005543 return 0;
5544 if (sb->s_flags & MS_RDONLY)
5545 return 1;
Darrick J. Wonge2b911c2015-10-17 16:18:43 -04005546 if (ext4_has_unknown_ext3_ro_compat_features(sb))
Theodore Ts'o2035e772011-04-18 17:29:14 -04005547 return 0;
5548 return 1;
5549}
Theodore Ts'o24b58422009-12-07 14:08:51 -05005550
Theodore Ts'o03010a32008-10-10 20:02:48 -04005551static struct file_system_type ext4_fs_type = {
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005552 .owner = THIS_MODULE,
Theodore Ts'o03010a32008-10-10 20:02:48 -04005553 .name = "ext4",
Al Viro152a0832010-07-25 00:46:55 +04005554 .mount = ext4_mount,
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005555 .kill_sb = kill_block_super,
5556 .fs_flags = FS_REQUIRES_DEV,
5557};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08005558MODULE_ALIAS_FS("ext4");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005559
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005560/* Shared across all ext4 file systems */
5561wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005562
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005563static int __init ext4_init_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005564{
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005565 int i, err;
Alex Tomasc9de5602008-01-29 00:19:52 -05005566
Theodore Ts'oe294a532015-08-15 14:59:44 -04005567 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
Al Viro07c0c5d82012-03-20 22:05:02 -04005568 ext4_li_info = NULL;
5569 mutex_init(&ext4_li_mtx);
5570
Carlos Maiolino9a4c8012012-12-10 16:30:45 -05005571 /* Build-time check for flags consistency */
Dmitry Monakhov12e9b892010-05-16 22:00:00 -04005572 ext4_check_flag_values();
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005573
Jan Karae142d052016-03-08 22:44:50 -05005574 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005575 init_waitqueue_head(&ext4__ioend_wq[i]);
Eric Sandeene9e3bce2011-02-12 08:17:34 -05005576
Zheng Liu51865fd2012-11-08 21:57:32 -05005577 err = ext4_init_es();
Theodore Ts'o6fd058f2009-05-17 15:38:01 -04005578 if (err)
5579 return err;
Zheng Liu51865fd2012-11-08 21:57:32 -05005580
5581 err = ext4_init_pageio();
5582 if (err)
Theodore Ts'ob5799012015-09-23 12:44:17 -04005583 goto out5;
Zheng Liu51865fd2012-11-08 21:57:32 -05005584
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005585 err = ext4_init_system_zone();
Theodore Ts'obd2d0212010-10-27 21:30:10 -04005586 if (err)
Theodore Ts'ob5799012015-09-23 12:44:17 -04005587 goto out4;
Lukas Czerner857ac882010-10-27 21:30:05 -04005588
Theodore Ts'ob5799012015-09-23 12:44:17 -04005589 err = ext4_init_sysfs();
Theodore Ts'odd683142011-02-03 14:33:49 -05005590 if (err)
Theodore Ts'ob5799012015-09-23 12:44:17 -04005591 goto out3;
Lukas Czerner857ac882010-10-27 21:30:05 -04005592
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005593 err = ext4_init_mballoc();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005594 if (err)
Alex Tomasc9de5602008-01-29 00:19:52 -05005595 goto out2;
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005596 err = init_inodecache();
5597 if (err)
5598 goto out1;
Theodore Ts'o24b58422009-12-07 14:08:51 -05005599 register_as_ext3();
Theodore Ts'o2035e772011-04-18 17:29:14 -04005600 register_as_ext2();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005601 err = register_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005602 if (err)
5603 goto out;
Lukas Czernerbfff6872010-10-27 21:30:05 -04005604
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005605 return 0;
5606out:
Theodore Ts'o24b58422009-12-07 14:08:51 -05005607 unregister_as_ext2();
5608 unregister_as_ext3();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005609 destroy_inodecache();
5610out1:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005611 ext4_exit_mballoc();
T Makphaibulchoke9c191f72014-03-18 19:24:49 -04005612out2:
Theodore Ts'ob5799012015-09-23 12:44:17 -04005613 ext4_exit_sysfs();
5614out3:
Theodore Ts'odd683142011-02-03 14:33:49 -05005615 ext4_exit_system_zone();
Theodore Ts'ob5799012015-09-23 12:44:17 -04005616out4:
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005617 ext4_exit_pageio();
Theodore Ts'ob5799012015-09-23 12:44:17 -04005618out5:
Zheng Liu51865fd2012-11-08 21:57:32 -05005619 ext4_exit_es();
5620
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005621 return err;
5622}
5623
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005624static void __exit ext4_exit_fs(void)
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005625{
Lukas Czernerbfff6872010-10-27 21:30:05 -04005626 ext4_destroy_lazyinit_thread();
Theodore Ts'o24b58422009-12-07 14:08:51 -05005627 unregister_as_ext2();
5628 unregister_as_ext3();
Theodore Ts'o03010a32008-10-10 20:02:48 -04005629 unregister_filesystem(&ext4_fs_type);
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005630 destroy_inodecache();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005631 ext4_exit_mballoc();
Theodore Ts'ob5799012015-09-23 12:44:17 -04005632 ext4_exit_sysfs();
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005633 ext4_exit_system_zone();
5634 ext4_exit_pageio();
Eric Sandeendd12ed12013-07-26 15:21:11 -04005635 ext4_exit_es();
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005636}
5637
5638MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
Theodore Ts'o83982b62009-01-06 14:53:16 -05005639MODULE_DESCRIPTION("Fourth Extended Filesystem");
Dave Kleikampac27a0e2006-10-11 01:20:50 -07005640MODULE_LICENSE("GPL");
Theodore Ts'o5dabfc72010-10-27 21:30:14 -04005641module_init(ext4_init_fs)
5642module_exit(ext4_exit_fs)