~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/ext4/super.c

Version: ~ [ linux-5.14-rc1 ] ~ [ linux-5.13.1 ] ~ [ linux-5.12.16 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.49 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.131 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.197 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.239 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.275 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.275 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *  linux/fs/ext4/super.c
  3  *
  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>
 23 #include <linux/vmalloc.h>
 24 #include <linux/slab.h>
 25 #include <linux/init.h>
 26 #include <linux/blkdev.h>
 27 #include <linux/backing-dev.h>
 28 #include <linux/parser.h>
 29 #include <linux/buffer_head.h>
 30 #include <linux/exportfs.h>
 31 #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>
 37 #include <linux/ctype.h>
 38 #include <linux/log2.h>
 39 #include <linux/crc16.h>
 40 #include <linux/cleancache.h>
 41 #include <asm/uaccess.h>
 42 
 43 #include <linux/kthread.h>
 44 #include <linux/freezer.h>
 45 
 46 #include "ext4.h"
 47 #include "ext4_extents.h"       /* Needed for trace points definition */
 48 #include "ext4_jbd2.h"
 49 #include "xattr.h"
 50 #include "acl.h"
 51 #include "mballoc.h"
 52 
 53 #define CREATE_TRACE_POINTS
 54 #include <trace/events/ext4.h>
 55 
 56 static struct ext4_lazy_init *ext4_li_info;
 57 static struct mutex ext4_li_mtx;
 58 static struct ratelimit_state ext4_mount_msg_ratelimit;
 59 
 60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
 61                              unsigned long journal_devnum);
 62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
 63 static int ext4_commit_super(struct super_block *sb, int sync);
 64 static void ext4_mark_recovery_complete(struct super_block *sb,
 65                                         struct ext4_super_block *es);
 66 static void ext4_clear_journal_err(struct super_block *sb,
 67                                    struct ext4_super_block *es);
 68 static int ext4_sync_fs(struct super_block *sb, int wait);
 69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
 70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
 71 static int ext4_unfreeze(struct super_block *sb);
 72 static int ext4_freeze(struct super_block *sb);
 73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
 74                        const char *dev_name, void *data);
 75 static inline int ext2_feature_set_ok(struct super_block *sb);
 76 static inline int ext3_feature_set_ok(struct super_block *sb);
 77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
 78 static void ext4_destroy_lazyinit_thread(void);
 79 static void ext4_unregister_li_request(struct super_block *sb);
 80 static void ext4_clear_request_list(void);
 81 
 82 /*
 83  * Lock ordering
 84  *
 85  * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
 86  * i_mmap_rwsem (inode->i_mmap_rwsem)!
 87  *
 88  * page fault path:
 89  * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
 90  *   page lock -> i_data_sem (rw)
 91  *
 92  * buffered write path:
 93  * sb_start_write -> i_mutex -> mmap_sem
 94  * sb_start_write -> i_mutex -> transaction start -> page lock ->
 95  *   i_data_sem (rw)
 96  *
 97  * truncate:
 98  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
 99  *   i_mmap_rwsem (w) -> page lock
100  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101  *   transaction start -> i_data_sem (rw)
102  *
103  * direct IO:
104  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
105  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
106  *   transaction start -> i_data_sem (rw)
107  *
108  * writepages:
109  * transaction start -> page lock(s) -> i_data_sem (rw)
110  */
111 
112 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
113 static struct file_system_type ext2_fs_type = {
114         .owner          = THIS_MODULE,
115         .name           = "ext2",
116         .mount          = ext4_mount,
117         .kill_sb        = kill_block_super,
118         .fs_flags       = FS_REQUIRES_DEV,
119 };
120 MODULE_ALIAS_FS("ext2");
121 MODULE_ALIAS("ext2");
122 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
123 #else
124 #define IS_EXT2_SB(sb) (0)
125 #endif
126 
127 
128 static struct file_system_type ext3_fs_type = {
129         .owner          = THIS_MODULE,
130         .name           = "ext3",
131         .mount          = ext4_mount,
132         .kill_sb        = kill_block_super,
133         .fs_flags       = FS_REQUIRES_DEV,
134 };
135 MODULE_ALIAS_FS("ext3");
136 MODULE_ALIAS("ext3");
137 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
138 
139 static int ext4_verify_csum_type(struct super_block *sb,
140                                  struct ext4_super_block *es)
141 {
142         if (!ext4_has_feature_metadata_csum(sb))
143                 return 1;
144 
145         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
146 }
147 
148 static __le32 ext4_superblock_csum(struct super_block *sb,
149                                    struct ext4_super_block *es)
150 {
151         struct ext4_sb_info *sbi = EXT4_SB(sb);
152         int offset = offsetof(struct ext4_super_block, s_checksum);
153         __u32 csum;
154 
155         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
156 
157         return cpu_to_le32(csum);
158 }
159 
160 static int ext4_superblock_csum_verify(struct super_block *sb,
161                                        struct ext4_super_block *es)
162 {
163         if (!ext4_has_metadata_csum(sb))
164                 return 1;
165 
166         return es->s_checksum == ext4_superblock_csum(sb, es);
167 }
168 
169 void ext4_superblock_csum_set(struct super_block *sb)
170 {
171         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
172 
173         if (!ext4_has_metadata_csum(sb))
174                 return;
175 
176         es->s_checksum = ext4_superblock_csum(sb, es);
177 }
178 
179 void *ext4_kvmalloc(size_t size, gfp_t flags)
180 {
181         void *ret;
182 
183         ret = kmalloc(size, flags | __GFP_NOWARN);
184         if (!ret)
185                 ret = __vmalloc(size, flags, PAGE_KERNEL);
186         return ret;
187 }
188 
189 void *ext4_kvzalloc(size_t size, gfp_t flags)
190 {
191         void *ret;
192 
193         ret = kzalloc(size, flags | __GFP_NOWARN);
194         if (!ret)
195                 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
196         return ret;
197 }
198 
199 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
200                                struct ext4_group_desc *bg)
201 {
202         return le32_to_cpu(bg->bg_block_bitmap_lo) |
203                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
204                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
205 }
206 
207 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
208                                struct ext4_group_desc *bg)
209 {
210         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
211                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
213 }
214 
215 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
216                               struct ext4_group_desc *bg)
217 {
218         return le32_to_cpu(bg->bg_inode_table_lo) |
219                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
221 }
222 
223 __u32 ext4_free_group_clusters(struct super_block *sb,
224                                struct ext4_group_desc *bg)
225 {
226         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
227                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
229 }
230 
231 __u32 ext4_free_inodes_count(struct super_block *sb,
232                               struct ext4_group_desc *bg)
233 {
234         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
235                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
237 }
238 
239 __u32 ext4_used_dirs_count(struct super_block *sb,
240                               struct ext4_group_desc *bg)
241 {
242         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
243                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
245 }
246 
247 __u32 ext4_itable_unused_count(struct super_block *sb,
248                               struct ext4_group_desc *bg)
249 {
250         return le16_to_cpu(bg->bg_itable_unused_lo) |
251                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
252                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
253 }
254 
255 void ext4_block_bitmap_set(struct super_block *sb,
256                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
257 {
258         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
259         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
261 }
262 
263 void ext4_inode_bitmap_set(struct super_block *sb,
264                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
265 {
266         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
267         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
269 }
270 
271 void ext4_inode_table_set(struct super_block *sb,
272                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
273 {
274         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
275         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
277 }
278 
279 void ext4_free_group_clusters_set(struct super_block *sb,
280                                   struct ext4_group_desc *bg, __u32 count)
281 {
282         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
283         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
285 }
286 
287 void ext4_free_inodes_set(struct super_block *sb,
288                           struct ext4_group_desc *bg, __u32 count)
289 {
290         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
291         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
293 }
294 
295 void ext4_used_dirs_set(struct super_block *sb,
296                           struct ext4_group_desc *bg, __u32 count)
297 {
298         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
299         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
301 }
302 
303 void ext4_itable_unused_set(struct super_block *sb,
304                           struct ext4_group_desc *bg, __u32 count)
305 {
306         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
307         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
309 }
310 
311 
312 static void __save_error_info(struct super_block *sb, const char *func,
313                             unsigned int line)
314 {
315         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
316 
317         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
318         if (bdev_read_only(sb->s_bdev))
319                 return;
320         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
321         es->s_last_error_time = cpu_to_le32(get_seconds());
322         strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
323         es->s_last_error_line = cpu_to_le32(line);
324         if (!es->s_first_error_time) {
325                 es->s_first_error_time = es->s_last_error_time;
326                 strncpy(es->s_first_error_func, func,
327                         sizeof(es->s_first_error_func));
328                 es->s_first_error_line = cpu_to_le32(line);
329                 es->s_first_error_ino = es->s_last_error_ino;
330                 es->s_first_error_block = es->s_last_error_block;
331         }
332         /*
333          * Start the daily error reporting function if it hasn't been
334          * started already
335          */
336         if (!es->s_error_count)
337                 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
338         le32_add_cpu(&es->s_error_count, 1);
339 }
340 
341 static void save_error_info(struct super_block *sb, const char *func,
342                             unsigned int line)
343 {
344         __save_error_info(sb, func, line);
345         ext4_commit_super(sb, 1);
346 }
347 
348 /*
349  * The del_gendisk() function uninitializes the disk-specific data
350  * structures, including the bdi structure, without telling anyone
351  * else.  Once this happens, any attempt to call mark_buffer_dirty()
352  * (for example, by ext4_commit_super), will cause a kernel OOPS.
353  * This is a kludge to prevent these oops until we can put in a proper
354  * hook in del_gendisk() to inform the VFS and file system layers.
355  */
356 static int block_device_ejected(struct super_block *sb)
357 {
358         struct inode *bd_inode = sb->s_bdev->bd_inode;
359         struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
360 
361         return bdi->dev == NULL;
362 }
363 
364 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
365 {
366         struct super_block              *sb = journal->j_private;
367         struct ext4_sb_info             *sbi = EXT4_SB(sb);
368         int                             error = is_journal_aborted(journal);
369         struct ext4_journal_cb_entry    *jce;
370 
371         BUG_ON(txn->t_state == T_FINISHED);
372         spin_lock(&sbi->s_md_lock);
373         while (!list_empty(&txn->t_private_list)) {
374                 jce = list_entry(txn->t_private_list.next,
375                                  struct ext4_journal_cb_entry, jce_list);
376                 list_del_init(&jce->jce_list);
377                 spin_unlock(&sbi->s_md_lock);
378                 jce->jce_func(sb, jce, error);
379                 spin_lock(&sbi->s_md_lock);
380         }
381         spin_unlock(&sbi->s_md_lock);
382 }
383 
384 /* Deal with the reporting of failure conditions on a filesystem such as
385  * inconsistencies detected or read IO failures.
386  *
387  * On ext2, we can store the error state of the filesystem in the
388  * superblock.  That is not possible on ext4, because we may have other
389  * write ordering constraints on the superblock which prevent us from
390  * writing it out straight away; and given that the journal is about to
391  * be aborted, we can't rely on the current, or future, transactions to
392  * write out the superblock safely.
393  *
394  * We'll just use the jbd2_journal_abort() error code to record an error in
395  * the journal instead.  On recovery, the journal will complain about
396  * that error until we've noted it down and cleared it.
397  */
398 
399 static void ext4_handle_error(struct super_block *sb)
400 {
401         if (sb->s_flags & MS_RDONLY)
402                 return;
403 
404         if (!test_opt(sb, ERRORS_CONT)) {
405                 journal_t *journal = EXT4_SB(sb)->s_journal;
406 
407                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
408                 if (journal)
409                         jbd2_journal_abort(journal, -EIO);
410         }
411         if (test_opt(sb, ERRORS_RO)) {
412                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
413                 /*
414                  * Make sure updated value of ->s_mount_flags will be visible
415                  * before ->s_flags update
416                  */
417                 smp_wmb();
418                 sb->s_flags |= MS_RDONLY;
419         }
420         if (test_opt(sb, ERRORS_PANIC)) {
421                 if (EXT4_SB(sb)->s_journal &&
422                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
423                         return;
424                 panic("EXT4-fs (device %s): panic forced after error\n",
425                         sb->s_id);
426         }
427 }
428 
429 #define ext4_error_ratelimit(sb)                                        \
430                 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),     \
431                              "EXT4-fs error")
432 
433 void __ext4_error(struct super_block *sb, const char *function,
434                   unsigned int line, const char *fmt, ...)
435 {
436         struct va_format vaf;
437         va_list args;
438 
439         if (ext4_error_ratelimit(sb)) {
440                 va_start(args, fmt);
441                 vaf.fmt = fmt;
442                 vaf.va = &args;
443                 printk(KERN_CRIT
444                        "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
445                        sb->s_id, function, line, current->comm, &vaf);
446                 va_end(args);
447         }
448         save_error_info(sb, function, line);
449         ext4_handle_error(sb);
450 }
451 
452 void __ext4_error_inode(struct inode *inode, const char *function,
453                         unsigned int line, ext4_fsblk_t block,
454                         const char *fmt, ...)
455 {
456         va_list args;
457         struct va_format vaf;
458         struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
459 
460         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
461         es->s_last_error_block = cpu_to_le64(block);
462         if (ext4_error_ratelimit(inode->i_sb)) {
463                 va_start(args, fmt);
464                 vaf.fmt = fmt;
465                 vaf.va = &args;
466                 if (block)
467                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
468                                "inode #%lu: block %llu: comm %s: %pV\n",
469                                inode->i_sb->s_id, function, line, inode->i_ino,
470                                block, current->comm, &vaf);
471                 else
472                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
473                                "inode #%lu: comm %s: %pV\n",
474                                inode->i_sb->s_id, function, line, inode->i_ino,
475                                current->comm, &vaf);
476                 va_end(args);
477         }
478         save_error_info(inode->i_sb, function, line);
479         ext4_handle_error(inode->i_sb);
480 }
481 
482 void __ext4_error_file(struct file *file, const char *function,
483                        unsigned int line, ext4_fsblk_t block,
484                        const char *fmt, ...)
485 {
486         va_list args;
487         struct va_format vaf;
488         struct ext4_super_block *es;
489         struct inode *inode = file_inode(file);
490         char pathname[80], *path;
491 
492         es = EXT4_SB(inode->i_sb)->s_es;
493         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
494         if (ext4_error_ratelimit(inode->i_sb)) {
495                 path = file_path(file, pathname, sizeof(pathname));
496                 if (IS_ERR(path))
497                         path = "(unknown)";
498                 va_start(args, fmt);
499                 vaf.fmt = fmt;
500                 vaf.va = &args;
501                 if (block)
502                         printk(KERN_CRIT
503                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
504                                "block %llu: comm %s: path %s: %pV\n",
505                                inode->i_sb->s_id, function, line, inode->i_ino,
506                                block, current->comm, path, &vaf);
507                 else
508                         printk(KERN_CRIT
509                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
510                                "comm %s: path %s: %pV\n",
511                                inode->i_sb->s_id, function, line, inode->i_ino,
512                                current->comm, path, &vaf);
513                 va_end(args);
514         }
515         save_error_info(inode->i_sb, function, line);
516         ext4_handle_error(inode->i_sb);
517 }
518 
519 const char *ext4_decode_error(struct super_block *sb, int errno,
520                               char nbuf[16])
521 {
522         char *errstr = NULL;
523 
524         switch (errno) {
525         case -EFSCORRUPTED:
526                 errstr = "Corrupt filesystem";
527                 break;
528         case -EFSBADCRC:
529                 errstr = "Filesystem failed CRC";
530                 break;
531         case -EIO:
532                 errstr = "IO failure";
533                 break;
534         case -ENOMEM:
535                 errstr = "Out of memory";
536                 break;
537         case -EROFS:
538                 if (!sb || (EXT4_SB(sb)->s_journal &&
539                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
540                         errstr = "Journal has aborted";
541                 else
542                         errstr = "Readonly filesystem";
543                 break;
544         default:
545                 /* If the caller passed in an extra buffer for unknown
546                  * errors, textualise them now.  Else we just return
547                  * NULL. */
548                 if (nbuf) {
549                         /* Check for truncated error codes... */
550                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
551                                 errstr = nbuf;
552                 }
553                 break;
554         }
555 
556         return errstr;
557 }
558 
559 /* __ext4_std_error decodes expected errors from journaling functions
560  * automatically and invokes the appropriate error response.  */
561 
562 void __ext4_std_error(struct super_block *sb, const char *function,
563                       unsigned int line, int errno)
564 {
565         char nbuf[16];
566         const char *errstr;
567 
568         /* Special case: if the error is EROFS, and we're not already
569          * inside a transaction, then there's really no point in logging
570          * an error. */
571         if (errno == -EROFS && journal_current_handle() == NULL &&
572             (sb->s_flags & MS_RDONLY))
573                 return;
574 
575         if (ext4_error_ratelimit(sb)) {
576                 errstr = ext4_decode_error(sb, errno, nbuf);
577                 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
578                        sb->s_id, function, line, errstr);
579         }
580 
581         save_error_info(sb, function, line);
582         ext4_handle_error(sb);
583 }
584 
585 /*
586  * ext4_abort is a much stronger failure handler than ext4_error.  The
587  * abort function may be used to deal with unrecoverable failures such
588  * as journal IO errors or ENOMEM at a critical moment in log management.
589  *
590  * We unconditionally force the filesystem into an ABORT|READONLY state,
591  * unless the error response on the fs has been set to panic in which
592  * case we take the easy way out and panic immediately.
593  */
594 
595 void __ext4_abort(struct super_block *sb, const char *function,
596                 unsigned int line, const char *fmt, ...)
597 {
598         va_list args;
599 
600         save_error_info(sb, function, line);
601         va_start(args, fmt);
602         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
603                function, line);
604         vprintk(fmt, args);
605         printk("\n");
606         va_end(args);
607 
608         if ((sb->s_flags & MS_RDONLY) == 0) {
609                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
610                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
611                 /*
612                  * Make sure updated value of ->s_mount_flags will be visible
613                  * before ->s_flags update
614                  */
615                 smp_wmb();
616                 sb->s_flags |= MS_RDONLY;
617                 if (EXT4_SB(sb)->s_journal)
618                         jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
619                 save_error_info(sb, function, line);
620         }
621         if (test_opt(sb, ERRORS_PANIC)) {
622                 if (EXT4_SB(sb)->s_journal &&
623                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
624                         return;
625                 panic("EXT4-fs panic from previous error\n");
626         }
627 }
628 
629 void __ext4_msg(struct super_block *sb,
630                 const char *prefix, const char *fmt, ...)
631 {
632         struct va_format vaf;
633         va_list args;
634 
635         if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
636                 return;
637 
638         va_start(args, fmt);
639         vaf.fmt = fmt;
640         vaf.va = &args;
641         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
642         va_end(args);
643 }
644 
645 #define ext4_warning_ratelimit(sb)                                      \
646                 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
647                              "EXT4-fs warning")
648 
649 void __ext4_warning(struct super_block *sb, const char *function,
650                     unsigned int line, const char *fmt, ...)
651 {
652         struct va_format vaf;
653         va_list args;
654 
655         if (!ext4_warning_ratelimit(sb))
656                 return;
657 
658         va_start(args, fmt);
659         vaf.fmt = fmt;
660         vaf.va = &args;
661         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
662                sb->s_id, function, line, &vaf);
663         va_end(args);
664 }
665 
666 void __ext4_warning_inode(const struct inode *inode, const char *function,
667                           unsigned int line, const char *fmt, ...)
668 {
669         struct va_format vaf;
670         va_list args;
671 
672         if (!ext4_warning_ratelimit(inode->i_sb))
673                 return;
674 
675         va_start(args, fmt);
676         vaf.fmt = fmt;
677         vaf.va = &args;
678         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
679                "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
680                function, line, inode->i_ino, current->comm, &vaf);
681         va_end(args);
682 }
683 
684 void __ext4_grp_locked_error(const char *function, unsigned int line,
685                              struct super_block *sb, ext4_group_t grp,
686                              unsigned long ino, ext4_fsblk_t block,
687                              const char *fmt, ...)
688 __releases(bitlock)
689 __acquires(bitlock)
690 {
691         struct va_format vaf;
692         va_list args;
693         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
694 
695         es->s_last_error_ino = cpu_to_le32(ino);
696         es->s_last_error_block = cpu_to_le64(block);
697         __save_error_info(sb, function, line);
698 
699         if (ext4_error_ratelimit(sb)) {
700                 va_start(args, fmt);
701                 vaf.fmt = fmt;
702                 vaf.va = &args;
703                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
704                        sb->s_id, function, line, grp);
705                 if (ino)
706                         printk(KERN_CONT "inode %lu: ", ino);
707                 if (block)
708                         printk(KERN_CONT "block %llu:",
709                                (unsigned long long) block);
710                 printk(KERN_CONT "%pV\n", &vaf);
711                 va_end(args);
712         }
713 
714         if (test_opt(sb, ERRORS_CONT)) {
715                 ext4_commit_super(sb, 0);
716                 return;
717         }
718 
719         ext4_unlock_group(sb, grp);
720         ext4_handle_error(sb);
721         /*
722          * We only get here in the ERRORS_RO case; relocking the group
723          * may be dangerous, but nothing bad will happen since the
724          * filesystem will have already been marked read/only and the
725          * journal has been aborted.  We return 1 as a hint to callers
726          * who might what to use the return value from
727          * ext4_grp_locked_error() to distinguish between the
728          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
729          * aggressively from the ext4 function in question, with a
730          * more appropriate error code.
731          */
732         ext4_lock_group(sb, grp);
733         return;
734 }
735 
736 void ext4_update_dynamic_rev(struct super_block *sb)
737 {
738         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
739 
740         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
741                 return;
742 
743         ext4_warning(sb,
744                      "updating to rev %d because of new feature flag, "
745                      "running e2fsck is recommended",
746                      EXT4_DYNAMIC_REV);
747 
748         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
749         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
750         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
751         /* leave es->s_feature_*compat flags alone */
752         /* es->s_uuid will be set by e2fsck if empty */
753 
754         /*
755          * The rest of the superblock fields should be zero, and if not it
756          * means they are likely already in use, so leave them alone.  We
757          * can leave it up to e2fsck to clean up any inconsistencies there.
758          */
759 }
760 
761 /*
762  * Open the external journal device
763  */
764 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
765 {
766         struct block_device *bdev;
767         char b[BDEVNAME_SIZE];
768 
769         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
770         if (IS_ERR(bdev))
771                 goto fail;
772         return bdev;
773 
774 fail:
775         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
776                         __bdevname(dev, b), PTR_ERR(bdev));
777         return NULL;
778 }
779 
780 /*
781  * Release the journal device
782  */
783 static void ext4_blkdev_put(struct block_device *bdev)
784 {
785         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
786 }
787 
788 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
789 {
790         struct block_device *bdev;
791         bdev = sbi->journal_bdev;
792         if (bdev) {
793                 ext4_blkdev_put(bdev);
794                 sbi->journal_bdev = NULL;
795         }
796 }
797 
798 static inline struct inode *orphan_list_entry(struct list_head *l)
799 {
800         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
801 }
802 
803 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
804 {
805         struct list_head *l;
806 
807         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
808                  le32_to_cpu(sbi->s_es->s_last_orphan));
809 
810         printk(KERN_ERR "sb_info orphan list:\n");
811         list_for_each(l, &sbi->s_orphan) {
812                 struct inode *inode = orphan_list_entry(l);
813                 printk(KERN_ERR "  "
814                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
815                        inode->i_sb->s_id, inode->i_ino, inode,
816                        inode->i_mode, inode->i_nlink,
817                        NEXT_ORPHAN(inode));
818         }
819 }
820 
821 static void ext4_put_super(struct super_block *sb)
822 {
823         struct ext4_sb_info *sbi = EXT4_SB(sb);
824         struct ext4_super_block *es = sbi->s_es;
825         int i, err;
826 
827         ext4_unregister_li_request(sb);
828         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
829 
830         flush_workqueue(sbi->rsv_conversion_wq);
831         destroy_workqueue(sbi->rsv_conversion_wq);
832 
833         if (sbi->s_journal) {
834                 err = jbd2_journal_destroy(sbi->s_journal);
835                 sbi->s_journal = NULL;
836                 if (err < 0)
837                         ext4_abort(sb, "Couldn't clean up the journal");
838         }
839 
840         ext4_unregister_sysfs(sb);
841         ext4_es_unregister_shrinker(sbi);
842         del_timer_sync(&sbi->s_err_report);
843         ext4_release_system_zone(sb);
844         ext4_mb_release(sb);
845         ext4_ext_release(sb);
846 
847         if (!(sb->s_flags & MS_RDONLY)) {
848                 ext4_clear_feature_journal_needs_recovery(sb);
849                 es->s_state = cpu_to_le16(sbi->s_mount_state);
850         }
851         if (!(sb->s_flags & MS_RDONLY))
852                 ext4_commit_super(sb, 1);
853 
854         for (i = 0; i < sbi->s_gdb_count; i++)
855                 brelse(sbi->s_group_desc[i]);
856         kvfree(sbi->s_group_desc);
857         kvfree(sbi->s_flex_groups);
858         percpu_counter_destroy(&sbi->s_freeclusters_counter);
859         percpu_counter_destroy(&sbi->s_freeinodes_counter);
860         percpu_counter_destroy(&sbi->s_dirs_counter);
861         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
862         percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
863         brelse(sbi->s_sbh);
864 #ifdef CONFIG_QUOTA
865         for (i = 0; i < EXT4_MAXQUOTAS; i++)
866                 kfree(sbi->s_qf_names[i]);
867 #endif
868 
869         /* Debugging code just in case the in-memory inode orphan list
870          * isn't empty.  The on-disk one can be non-empty if we've
871          * detected an error and taken the fs readonly, but the
872          * in-memory list had better be clean by this point. */
873         if (!list_empty(&sbi->s_orphan))
874                 dump_orphan_list(sb, sbi);
875         J_ASSERT(list_empty(&sbi->s_orphan));
876 
877         sync_blockdev(sb->s_bdev);
878         invalidate_bdev(sb->s_bdev);
879         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
880                 /*
881                  * Invalidate the journal device's buffers.  We don't want them
882                  * floating about in memory - the physical journal device may
883                  * hotswapped, and it breaks the `ro-after' testing code.
884                  */
885                 sync_blockdev(sbi->journal_bdev);
886                 invalidate_bdev(sbi->journal_bdev);
887                 ext4_blkdev_remove(sbi);
888         }
889         if (sbi->s_mb_cache) {
890                 ext4_xattr_destroy_cache(sbi->s_mb_cache);
891                 sbi->s_mb_cache = NULL;
892         }
893         if (sbi->s_mmp_tsk)
894                 kthread_stop(sbi->s_mmp_tsk);
895         sb->s_fs_info = NULL;
896         /*
897          * Now that we are completely done shutting down the
898          * superblock, we need to actually destroy the kobject.
899          */
900         kobject_put(&sbi->s_kobj);
901         wait_for_completion(&sbi->s_kobj_unregister);
902         if (sbi->s_chksum_driver)
903                 crypto_free_shash(sbi->s_chksum_driver);
904         kfree(sbi->s_blockgroup_lock);
905         kfree(sbi);
906 }
907 
908 static struct kmem_cache *ext4_inode_cachep;
909 
910 /*
911  * Called inside transaction, so use GFP_NOFS
912  */
913 static struct inode *ext4_alloc_inode(struct super_block *sb)
914 {
915         struct ext4_inode_info *ei;
916 
917         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
918         if (!ei)
919                 return NULL;
920 
921         ei->vfs_inode.i_version = 1;
922         spin_lock_init(&ei->i_raw_lock);
923         INIT_LIST_HEAD(&ei->i_prealloc_list);
924         spin_lock_init(&ei->i_prealloc_lock);
925         ext4_es_init_tree(&ei->i_es_tree);
926         rwlock_init(&ei->i_es_lock);
927         INIT_LIST_HEAD(&ei->i_es_list);
928         ei->i_es_all_nr = 0;
929         ei->i_es_shk_nr = 0;
930         ei->i_es_shrink_lblk = 0;
931         ei->i_reserved_data_blocks = 0;
932         ei->i_reserved_meta_blocks = 0;
933         ei->i_allocated_meta_blocks = 0;
934         ei->i_da_metadata_calc_len = 0;
935         ei->i_da_metadata_calc_last_lblock = 0;
936         spin_lock_init(&(ei->i_block_reservation_lock));
937 #ifdef CONFIG_QUOTA
938         ei->i_reserved_quota = 0;
939         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
940 #endif
941         ei->jinode = NULL;
942         INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
943         spin_lock_init(&ei->i_completed_io_lock);
944         ei->i_sync_tid = 0;
945         ei->i_datasync_tid = 0;
946         atomic_set(&ei->i_unwritten, 0);
947         INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
948 #ifdef CONFIG_EXT4_FS_ENCRYPTION
949         ei->i_crypt_info = NULL;
950 #endif
951         return &ei->vfs_inode;
952 }
953 
954 static 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 
962 static void ext4_i_callback(struct rcu_head *head)
963 {
964         struct inode *inode = container_of(head, struct inode, i_rcu);
965         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
966 }
967 
968 static void ext4_destroy_inode(struct inode *inode)
969 {
970         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
971                 ext4_msg(inode->i_sb, KERN_ERR,
972                          "Inode %lu (%p): orphan list check failed!",
973                          inode->i_ino, EXT4_I(inode));
974                 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         }
979         call_rcu(&inode->i_rcu, ext4_i_callback);
980 }
981 
982 static void init_once(void *foo)
983 {
984         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
985 
986         INIT_LIST_HEAD(&ei->i_orphan);
987         init_rwsem(&ei->xattr_sem);
988         init_rwsem(&ei->i_data_sem);
989         init_rwsem(&ei->i_mmap_sem);
990         inode_init_once(&ei->vfs_inode);
991 }
992 
993 static int __init init_inodecache(void)
994 {
995         ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996                                              sizeof(struct ext4_inode_info),
997                                              0, (SLAB_RECLAIM_ACCOUNT|
998                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
999                                              init_once);
1000         if (ext4_inode_cachep == NULL)
1001                 return -ENOMEM;
1002         return 0;
1003 }
1004 
1005 static void destroy_inodecache(void)
1006 {
1007         /*
1008          * Make sure all delayed rcu free inodes are flushed before we
1009          * destroy cache.
1010          */
1011         rcu_barrier();
1012         kmem_cache_destroy(ext4_inode_cachep);
1013 }
1014 
1015 void ext4_clear_inode(struct inode *inode)
1016 {
1017         invalidate_inode_buffers(inode);
1018         clear_inode(inode);
1019         dquot_drop(inode);
1020         ext4_discard_preallocations(inode);
1021         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1022         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         }
1028 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1029         if (EXT4_I(inode)->i_crypt_info)
1030                 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1031 #endif
1032 }
1033 
1034 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1035                                         u64 ino, u32 generation)
1036 {
1037         struct inode *inode;
1038 
1039         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1040                 return ERR_PTR(-ESTALE);
1041         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1042                 return ERR_PTR(-ESTALE);
1043 
1044         /* iget isn't really right if the inode is currently unallocated!!
1045          *
1046          * ext4_read_inode will return a bad_inode if the inode had been
1047          * deleted, so we should be safe.
1048          *
1049          * Currently we don't know the generation for parent directory, so
1050          * a generation of 0 means "accept any"
1051          */
1052         inode = ext4_iget_normal(sb, ino);
1053         if (IS_ERR(inode))
1054                 return ERR_CAST(inode);
1055         if (generation && inode->i_generation != generation) {
1056                 iput(inode);
1057                 return ERR_PTR(-ESTALE);
1058         }
1059 
1060         return inode;
1061 }
1062 
1063 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1064                                         int fh_len, int fh_type)
1065 {
1066         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1067                                     ext4_nfs_get_inode);
1068 }
1069 
1070 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1071                                         int fh_len, int fh_type)
1072 {
1073         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1074                                     ext4_nfs_get_inode);
1075 }
1076 
1077 /*
1078  * Try to release metadata pages (indirect blocks, directories) which are
1079  * mapped via the block device.  Since these pages could have journal heads
1080  * which would prevent try_to_free_buffers() from freeing them, we must use
1081  * jbd2 layer's try_to_free_buffers() function to release them.
1082  */
1083 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1084                                  gfp_t wait)
1085 {
1086         journal_t *journal = EXT4_SB(sb)->s_journal;
1087 
1088         WARN_ON(PageChecked(page));
1089         if (!page_has_buffers(page))
1090                 return 0;
1091         if (journal)
1092                 return jbd2_journal_try_to_free_buffers(journal, page,
1093                                                 wait & ~__GFP_DIRECT_RECLAIM);
1094         return try_to_free_buffers(page);
1095 }
1096 
1097 #ifdef CONFIG_QUOTA
1098 static char *quotatypes[] = INITQFNAMES;
1099 #define QTYPE2NAME(t) (quotatypes[t])
1100 
1101 static int ext4_write_dquot(struct dquot *dquot);
1102 static int ext4_acquire_dquot(struct dquot *dquot);
1103 static int ext4_release_dquot(struct dquot *dquot);
1104 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1105 static int ext4_write_info(struct super_block *sb, int type);
1106 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1107                          struct path *path);
1108 static int ext4_quota_off(struct super_block *sb, int type);
1109 static int ext4_quota_on_mount(struct super_block *sb, int type);
1110 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1111                                size_t len, loff_t off);
1112 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1113                                 const char *data, size_t len, loff_t off);
1114 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1115                              unsigned int flags);
1116 static int ext4_enable_quotas(struct super_block *sb);
1117 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1118 
1119 static struct dquot **ext4_get_dquots(struct inode *inode)
1120 {
1121         return EXT4_I(inode)->i_dquot;
1122 }
1123 
1124 static const struct dquot_operations ext4_quota_operations = {
1125         .get_reserved_space = ext4_get_reserved_space,
1126         .write_dquot    = ext4_write_dquot,
1127         .acquire_dquot  = ext4_acquire_dquot,
1128         .release_dquot  = ext4_release_dquot,
1129         .mark_dirty     = ext4_mark_dquot_dirty,
1130         .write_info     = ext4_write_info,
1131         .alloc_dquot    = dquot_alloc,
1132         .destroy_dquot  = dquot_destroy,
1133         .get_projid     = ext4_get_projid,
1134         .get_next_id    = ext4_get_next_id,
1135 };
1136 
1137 static const struct quotactl_ops ext4_qctl_operations = {
1138         .quota_on       = ext4_quota_on,
1139         .quota_off      = ext4_quota_off,
1140         .quota_sync     = dquot_quota_sync,
1141         .get_state      = dquot_get_state,
1142         .set_info       = dquot_set_dqinfo,
1143         .get_dqblk      = dquot_get_dqblk,
1144         .set_dqblk      = dquot_set_dqblk,
1145         .get_nextdqblk  = dquot_get_next_dqblk,
1146 };
1147 #endif
1148 
1149 static const struct super_operations ext4_sops = {
1150         .alloc_inode    = ext4_alloc_inode,
1151         .destroy_inode  = ext4_destroy_inode,
1152         .write_inode    = ext4_write_inode,
1153         .dirty_inode    = ext4_dirty_inode,
1154         .drop_inode     = ext4_drop_inode,
1155         .evict_inode    = ext4_evict_inode,
1156         .put_super      = ext4_put_super,
1157         .sync_fs        = ext4_sync_fs,
1158         .freeze_fs      = ext4_freeze,
1159         .unfreeze_fs    = ext4_unfreeze,
1160         .statfs         = ext4_statfs,
1161         .remount_fs     = ext4_remount,
1162         .show_options   = ext4_show_options,
1163 #ifdef CONFIG_QUOTA
1164         .quota_read     = ext4_quota_read,
1165         .quota_write    = ext4_quota_write,
1166         .get_dquots     = ext4_get_dquots,
1167 #endif
1168         .bdev_try_to_free_page = bdev_try_to_free_page,
1169 };
1170 
1171 static const struct export_operations ext4_export_ops = {
1172         .fh_to_dentry = ext4_fh_to_dentry,
1173         .fh_to_parent = ext4_fh_to_parent,
1174         .get_parent = ext4_get_parent,
1175 };
1176 
1177 enum {
1178         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1179         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1180         Opt_nouid32, Opt_debug, Opt_removed,
1181         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1182         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1183         Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1184         Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1185         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1186         Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1187         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1188         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1189         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1190         Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1191         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1192         Opt_lazytime, Opt_nolazytime,
1193         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1194         Opt_inode_readahead_blks, Opt_journal_ioprio,
1195         Opt_dioread_nolock, Opt_dioread_lock,
1196         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1197         Opt_max_dir_size_kb, Opt_nojournal_checksum,
1198 };
1199 
1200 static const match_table_t tokens = {
1201         {Opt_bsd_df, "bsddf"},
1202         {Opt_minix_df, "minixdf"},
1203         {Opt_grpid, "grpid"},
1204         {Opt_grpid, "bsdgroups"},
1205         {Opt_nogrpid, "nogrpid"},
1206         {Opt_nogrpid, "sysvgroups"},
1207         {Opt_resgid, "resgid=%u"},
1208         {Opt_resuid, "resuid=%u"},
1209         {Opt_sb, "sb=%u"},
1210         {Opt_err_cont, "errors=continue"},
1211         {Opt_err_panic, "errors=panic"},
1212         {Opt_err_ro, "errors=remount-ro"},
1213         {Opt_nouid32, "nouid32"},
1214         {Opt_debug, "debug"},
1215         {Opt_removed, "oldalloc"},
1216         {Opt_removed, "orlov"},
1217         {Opt_user_xattr, "user_xattr"},
1218         {Opt_nouser_xattr, "nouser_xattr"},
1219         {Opt_acl, "acl"},
1220         {Opt_noacl, "noacl"},
1221         {Opt_noload, "norecovery"},
1222         {Opt_noload, "noload"},
1223         {Opt_removed, "nobh"},
1224         {Opt_removed, "bh"},
1225         {Opt_commit, "commit=%u"},
1226         {Opt_min_batch_time, "min_batch_time=%u"},
1227         {Opt_max_batch_time, "max_batch_time=%u"},
1228         {Opt_journal_dev, "journal_dev=%u"},
1229         {Opt_journal_path, "journal_path=%s"},
1230         {Opt_journal_checksum, "journal_checksum"},
1231         {Opt_nojournal_checksum, "nojournal_checksum"},
1232         {Opt_journal_async_commit, "journal_async_commit"},
1233         {Opt_abort, "abort"},
1234         {Opt_data_journal, "data=journal"},
1235         {Opt_data_ordered, "data=ordered"},
1236         {Opt_data_writeback, "data=writeback"},
1237         {Opt_data_err_abort, "data_err=abort"},
1238         {Opt_data_err_ignore, "data_err=ignore"},
1239         {Opt_offusrjquota, "usrjquota="},
1240         {Opt_usrjquota, "usrjquota=%s"},
1241         {Opt_offgrpjquota, "grpjquota="},
1242         {Opt_grpjquota, "grpjquota=%s"},
1243         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1244         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1245         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1246         {Opt_grpquota, "grpquota"},
1247         {Opt_noquota, "noquota"},
1248         {Opt_quota, "quota"},
1249         {Opt_usrquota, "usrquota"},
1250         {Opt_barrier, "barrier=%u"},
1251         {Opt_barrier, "barrier"},
1252         {Opt_nobarrier, "nobarrier"},
1253         {Opt_i_version, "i_version"},
1254         {Opt_dax, "dax"},
1255         {Opt_stripe, "stripe=%u"},
1256         {Opt_delalloc, "delalloc"},
1257         {Opt_lazytime, "lazytime"},
1258         {Opt_nolazytime, "nolazytime"},
1259         {Opt_nodelalloc, "nodelalloc"},
1260         {Opt_removed, "mblk_io_submit"},
1261         {Opt_removed, "nomblk_io_submit"},
1262         {Opt_block_validity, "block_validity"},
1263         {Opt_noblock_validity, "noblock_validity"},
1264         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1265         {Opt_journal_ioprio, "journal_ioprio=%u"},
1266         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1267         {Opt_auto_da_alloc, "auto_da_alloc"},
1268         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1269         {Opt_dioread_nolock, "dioread_nolock"},
1270         {Opt_dioread_lock, "dioread_lock"},
1271         {Opt_discard, "discard"},
1272         {Opt_nodiscard, "nodiscard"},
1273         {Opt_init_itable, "init_itable=%u"},
1274         {Opt_init_itable, "init_itable"},
1275         {Opt_noinit_itable, "noinit_itable"},
1276         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1277         {Opt_test_dummy_encryption, "test_dummy_encryption"},
1278         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1279         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1280         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1281         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1282         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1283         {Opt_err, NULL},
1284 };
1285 
1286 static ext4_fsblk_t get_sb_block(void **data)
1287 {
1288         ext4_fsblk_t    sb_block;
1289         char            *options = (char *) *data;
1290 
1291         if (!options || strncmp(options, "sb=", 3) != 0)
1292                 return 1;       /* Default location */
1293 
1294         options += 3;
1295         /* TODO: use simple_strtoll with >32bit ext4 */
1296         sb_block = simple_strtoul(options, &options, 0);
1297         if (*options && *options != ',') {
1298                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1299                        (char *) *data);
1300                 return 1;
1301         }
1302         if (*options == ',')
1303                 options++;
1304         *data = (void *) options;
1305 
1306         return sb_block;
1307 }
1308 
1309 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1310 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1311         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1312 
1313 #ifdef CONFIG_QUOTA
1314 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1315 {
1316         struct ext4_sb_info *sbi = EXT4_SB(sb);
1317         char *qname;
1318         int ret = -1;
1319 
1320         if (sb_any_quota_loaded(sb) &&
1321                 !sbi->s_qf_names[qtype]) {
1322                 ext4_msg(sb, KERN_ERR,
1323                         "Cannot change journaled "
1324                         "quota options when quota turned on");
1325                 return -1;
1326         }
1327         if (ext4_has_feature_quota(sb)) {
1328                 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1329                          "ignored when QUOTA feature is enabled");
1330                 return 1;
1331         }
1332         qname = match_strdup(args);
1333         if (!qname) {
1334                 ext4_msg(sb, KERN_ERR,
1335                         "Not enough memory for storing quotafile name");
1336                 return -1;
1337         }
1338         if (sbi->s_qf_names[qtype]) {
1339                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1340                         ret = 1;
1341                 else
1342                         ext4_msg(sb, KERN_ERR,
1343                                  "%s quota file already specified",
1344                                  QTYPE2NAME(qtype));
1345                 goto errout;
1346         }
1347         if (strchr(qname, '/')) {
1348                 ext4_msg(sb, KERN_ERR,
1349                         "quotafile must be on filesystem root");
1350                 goto errout;
1351         }
1352         sbi->s_qf_names[qtype] = qname;
1353         set_opt(sb, QUOTA);
1354         return 1;
1355 errout:
1356         kfree(qname);
1357         return ret;
1358 }
1359 
1360 static int clear_qf_name(struct super_block *sb, int qtype)
1361 {
1362 
1363         struct ext4_sb_info *sbi = EXT4_SB(sb);
1364 
1365         if (sb_any_quota_loaded(sb) &&
1366                 sbi->s_qf_names[qtype]) {
1367                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1368                         " when quota turned on");
1369                 return -1;
1370         }
1371         kfree(sbi->s_qf_names[qtype]);
1372         sbi->s_qf_names[qtype] = NULL;
1373         return 1;
1374 }
1375 #endif
1376 
1377 #define MOPT_SET        0x0001
1378 #define MOPT_CLEAR      0x0002
1379 #define MOPT_NOSUPPORT  0x0004
1380 #define MOPT_EXPLICIT   0x0008
1381 #define MOPT_CLEAR_ERR  0x0010
1382 #define MOPT_GTE0       0x0020
1383 #ifdef CONFIG_QUOTA
1384 #define MOPT_Q          0
1385 #define MOPT_QFMT       0x0040
1386 #else
1387 #define MOPT_Q          MOPT_NOSUPPORT
1388 #define MOPT_QFMT       MOPT_NOSUPPORT
1389 #endif
1390 #define MOPT_DATAJ      0x0080
1391 #define MOPT_NO_EXT2    0x0100
1392 #define MOPT_NO_EXT3    0x0200
1393 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1394 #define MOPT_STRING     0x0400
1395 
1396 static const struct mount_opts {
1397         int     token;
1398         int     mount_opt;
1399         int     flags;
1400 } ext4_mount_opts[] = {
1401         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1402         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1403         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1404         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1405         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1406         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1407         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1408          MOPT_EXT4_ONLY | MOPT_SET},
1409         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1410          MOPT_EXT4_ONLY | MOPT_CLEAR},
1411         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1412         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1413         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1414          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1415         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1416          MOPT_EXT4_ONLY | MOPT_CLEAR},
1417         {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1418          MOPT_EXT4_ONLY | MOPT_CLEAR},
1419         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1420          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1421         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1422                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1423          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1424         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1425         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1426         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1427         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1428         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1429          MOPT_NO_EXT2},
1430         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1431          MOPT_NO_EXT2},
1432         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1433         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1434         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1435         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1436         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1437         {Opt_commit, 0, MOPT_GTE0},
1438         {Opt_max_batch_time, 0, MOPT_GTE0},
1439         {Opt_min_batch_time, 0, MOPT_GTE0},
1440         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1441         {Opt_init_itable, 0, MOPT_GTE0},
1442         {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1443         {Opt_stripe, 0, MOPT_GTE0},
1444         {Opt_resuid, 0, MOPT_GTE0},
1445         {Opt_resgid, 0, MOPT_GTE0},
1446         {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1447         {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1448         {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1449         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1450         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1451         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1452          MOPT_NO_EXT2 | MOPT_DATAJ},
1453         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1454         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1455 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1456         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1457         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1458 #else
1459         {Opt_acl, 0, MOPT_NOSUPPORT},
1460         {Opt_noacl, 0, MOPT_NOSUPPORT},
1461 #endif
1462         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1463         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1464         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1465         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1466                                                         MOPT_SET | MOPT_Q},
1467         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1468                                                         MOPT_SET | MOPT_Q},
1469         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1470                        EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1471         {Opt_usrjquota, 0, MOPT_Q},
1472         {Opt_grpjquota, 0, MOPT_Q},
1473         {Opt_offusrjquota, 0, MOPT_Q},
1474         {Opt_offgrpjquota, 0, MOPT_Q},
1475         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1476         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1477         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1478         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1479         {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1480         {Opt_err, 0, 0}
1481 };
1482 
1483 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1484                             substring_t *args, unsigned long *journal_devnum,
1485                             unsigned int *journal_ioprio, int is_remount)
1486 {
1487         struct ext4_sb_info *sbi = EXT4_SB(sb);
1488         const struct mount_opts *m;
1489         kuid_t uid;
1490         kgid_t gid;
1491         int arg = 0;
1492 
1493 #ifdef CONFIG_QUOTA
1494         if (token == Opt_usrjquota)
1495                 return set_qf_name(sb, USRQUOTA, &args[0]);
1496         else if (token == Opt_grpjquota)
1497                 return set_qf_name(sb, GRPQUOTA, &args[0]);
1498         else if (token == Opt_offusrjquota)
1499                 return clear_qf_name(sb, USRQUOTA);
1500         else if (token == Opt_offgrpjquota)
1501                 return clear_qf_name(sb, GRPQUOTA);
1502 #endif
1503         switch (token) {
1504         case Opt_noacl:
1505         case Opt_nouser_xattr:
1506                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1507                 break;
1508         case Opt_sb:
1509                 return 1;       /* handled by get_sb_block() */
1510         case Opt_removed:
1511                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1512                 return 1;
1513         case Opt_abort:
1514                 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1515                 return 1;
1516         case Opt_i_version:
1517                 sb->s_flags |= MS_I_VERSION;
1518                 return 1;
1519         case Opt_lazytime:
1520                 sb->s_flags |= MS_LAZYTIME;
1521                 return 1;
1522         case Opt_nolazytime:
1523                 sb->s_flags &= ~MS_LAZYTIME;
1524                 return 1;
1525         }
1526 
1527         for (m = ext4_mount_opts; m->token != Opt_err; m++)
1528                 if (token == m->token)
1529                         break;
1530 
1531         if (m->token == Opt_err) {
1532                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1533                          "or missing value", opt);
1534                 return -1;
1535         }
1536 
1537         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1538                 ext4_msg(sb, KERN_ERR,
1539                          "Mount option \"%s\" incompatible with ext2", opt);
1540                 return -1;
1541         }
1542         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1543                 ext4_msg(sb, KERN_ERR,
1544                          "Mount option \"%s\" incompatible with ext3", opt);
1545                 return -1;
1546         }
1547 
1548         if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1549                 return -1;
1550         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1551                 return -1;
1552         if (m->flags & MOPT_EXPLICIT) {
1553                 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1554                         set_opt2(sb, EXPLICIT_DELALLOC);
1555                 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1556                         set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1557                 } else
1558                         return -1;
1559         }
1560         if (m->flags & MOPT_CLEAR_ERR)
1561                 clear_opt(sb, ERRORS_MASK);
1562         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1563                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1564                          "options when quota turned on");
1565                 return -1;
1566         }
1567 
1568         if (m->flags & MOPT_NOSUPPORT) {
1569                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1570         } else if (token == Opt_commit) {
1571                 if (arg == 0)
1572                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1573                 sbi->s_commit_interval = HZ * arg;
1574         } else if (token == Opt_max_batch_time) {
1575                 sbi->s_max_batch_time = arg;
1576         } else if (token == Opt_min_batch_time) {
1577                 sbi->s_min_batch_time = arg;
1578         } else if (token == Opt_inode_readahead_blks) {
1579                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1580                         ext4_msg(sb, KERN_ERR,
1581                                  "EXT4-fs: inode_readahead_blks must be "
1582                                  "0 or a power of 2 smaller than 2^31");
1583                         return -1;
1584                 }
1585                 sbi->s_inode_readahead_blks = arg;
1586         } else if (token == Opt_init_itable) {
1587                 set_opt(sb, INIT_INODE_TABLE);
1588                 if (!args->from)
1589                         arg = EXT4_DEF_LI_WAIT_MULT;
1590                 sbi->s_li_wait_mult = arg;
1591         } else if (token == Opt_max_dir_size_kb) {
1592                 sbi->s_max_dir_size_kb = arg;
1593         } else if (token == Opt_stripe) {
1594                 sbi->s_stripe = arg;
1595         } else if (token == Opt_resuid) {
1596                 uid = make_kuid(current_user_ns(), arg);
1597                 if (!uid_valid(uid)) {
1598                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1599                         return -1;
1600                 }
1601                 sbi->s_resuid = uid;
1602         } else if (token == Opt_resgid) {
1603                 gid = make_kgid(current_user_ns(), arg);
1604                 if (!gid_valid(gid)) {
1605                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1606                         return -1;
1607                 }
1608                 sbi->s_resgid = gid;
1609         } else if (token == Opt_journal_dev) {
1610                 if (is_remount) {
1611                         ext4_msg(sb, KERN_ERR,
1612                                  "Cannot specify journal on remount");
1613                         return -1;
1614                 }
1615                 *journal_devnum = arg;
1616         } else if (token == Opt_journal_path) {
1617                 char *journal_path;
1618                 struct inode *journal_inode;
1619                 struct path path;
1620                 int error;
1621 
1622                 if (is_remount) {
1623                         ext4_msg(sb, KERN_ERR,
1624                                  "Cannot specify journal on remount");
1625                         return -1;
1626                 }
1627                 journal_path = match_strdup(&args[0]);
1628                 if (!journal_path) {
1629                         ext4_msg(sb, KERN_ERR, "error: could not dup "
1630                                 "journal device string");
1631                         return -1;
1632                 }
1633 
1634                 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1635                 if (error) {
1636                         ext4_msg(sb, KERN_ERR, "error: could not find "
1637                                 "journal device path: error %d", error);
1638                         kfree(journal_path);
1639                         return -1;
1640                 }
1641 
1642                 journal_inode = d_inode(path.dentry);
1643                 if (!S_ISBLK(journal_inode->i_mode)) {
1644                         ext4_msg(sb, KERN_ERR, "error: journal path %s "
1645                                 "is not a block device", journal_path);
1646                         path_put(&path);
1647                         kfree(journal_path);
1648                         return -1;
1649                 }
1650 
1651                 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1652                 path_put(&path);
1653                 kfree(journal_path);
1654         } else if (token == Opt_journal_ioprio) {
1655                 if (arg > 7) {
1656                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1657                                  " (must be 0-7)");
1658                         return -1;
1659                 }
1660                 *journal_ioprio =
1661                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1662         } else if (token == Opt_test_dummy_encryption) {
1663 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1664                 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1665                 ext4_msg(sb, KERN_WARNING,
1666                          "Test dummy encryption mode enabled");
1667 #else
1668                 ext4_msg(sb, KERN_WARNING,
1669                          "Test dummy encryption mount option ignored");
1670 #endif
1671         } else if (m->flags & MOPT_DATAJ) {
1672                 if (is_remount) {
1673                         if (!sbi->s_journal)
1674                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1675                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1676                                 ext4_msg(sb, KERN_ERR,
1677                                          "Cannot change data mode on remount");
1678                                 return -1;
1679                         }
1680                 } else {
1681                         clear_opt(sb, DATA_FLAGS);
1682                         sbi->s_mount_opt |= m->mount_opt;
1683                 }
1684 #ifdef CONFIG_QUOTA
1685         } else if (m->flags & MOPT_QFMT) {
1686                 if (sb_any_quota_loaded(sb) &&
1687                     sbi->s_jquota_fmt != m->mount_opt) {
1688                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1689                                  "quota options when quota turned on");
1690                         return -1;
1691                 }
1692                 if (ext4_has_feature_quota(sb)) {
1693                         ext4_msg(sb, KERN_INFO,
1694                                  "Quota format mount options ignored "
1695                                  "when QUOTA feature is enabled");
1696                         return 1;
1697                 }
1698                 sbi->s_jquota_fmt = m->mount_opt;
1699 #endif
1700         } else if (token == Opt_dax) {
1701 #ifdef CONFIG_FS_DAX
1702                 ext4_msg(sb, KERN_WARNING,
1703                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1704                         sbi->s_mount_opt |= m->mount_opt;
1705 #else
1706                 ext4_msg(sb, KERN_INFO, "dax option not supported");
1707                 return -1;
1708 #endif
1709         } else if (token == Opt_data_err_abort) {
1710                 sbi->s_mount_opt |= m->mount_opt;
1711         } else if (token == Opt_data_err_ignore) {
1712                 sbi->s_mount_opt &= ~m->mount_opt;
1713         } else {
1714                 if (!args->from)
1715                         arg = 1;
1716                 if (m->flags & MOPT_CLEAR)
1717                         arg = !arg;
1718                 else if (unlikely(!(m->flags & MOPT_SET))) {
1719                         ext4_msg(sb, KERN_WARNING,
1720                                  "buggy handling of option %s", opt);
1721                         WARN_ON(1);
1722                         return -1;
1723                 }
1724                 if (arg != 0)
1725                         sbi->s_mount_opt |= m->mount_opt;
1726                 else
1727                         sbi->s_mount_opt &= ~m->mount_opt;
1728         }
1729         return 1;
1730 }
1731 
1732 static int parse_options(char *options, struct super_block *sb,
1733                          unsigned long *journal_devnum,
1734                          unsigned int *journal_ioprio,
1735                          int is_remount)
1736 {
1737         struct ext4_sb_info *sbi = EXT4_SB(sb);
1738         char *p;
1739         substring_t args[MAX_OPT_ARGS];
1740         int token;
1741 
1742         if (!options)
1743                 return 1;
1744 
1745         while ((p = strsep(&options, ",")) != NULL) {
1746                 if (!*p)
1747                         continue;
1748                 /*
1749                  * Initialize args struct so we know whether arg was
1750                  * found; some options take optional arguments.
1751                  */
1752                 args[0].to = args[0].from = NULL;
1753                 token = match_token(p, tokens, args);
1754                 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1755                                      journal_ioprio, is_remount) < 0)
1756                         return 0;
1757         }
1758 #ifdef CONFIG_QUOTA
1759         if (ext4_has_feature_quota(sb) &&
1760             (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1761                 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1762                          "mount options ignored.");
1763                 clear_opt(sb, USRQUOTA);
1764                 clear_opt(sb, GRPQUOTA);
1765         } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1766                 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1767                         clear_opt(sb, USRQUOTA);
1768 
1769                 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1770                         clear_opt(sb, GRPQUOTA);
1771 
1772                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1773                         ext4_msg(sb, KERN_ERR, "old and new quota "
1774                                         "format mixing");
1775                         return 0;
1776                 }
1777 
1778                 if (!sbi->s_jquota_fmt) {
1779                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1780                                         "not specified");
1781                         return 0;
1782                 }
1783         }
1784 #endif
1785         if (test_opt(sb, DIOREAD_NOLOCK)) {
1786                 int blocksize =
1787                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1788 
1789                 if (blocksize < PAGE_SIZE) {
1790                         ext4_msg(sb, KERN_ERR, "can't mount with "
1791                                  "dioread_nolock if block size != PAGE_SIZE");
1792                         return 0;
1793                 }
1794         }
1795         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1796             test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1797                 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1798                          "in data=ordered mode");
1799                 return 0;
1800         }
1801         return 1;
1802 }
1803 
1804 static inline void ext4_show_quota_options(struct seq_file *seq,
1805                                            struct super_block *sb)
1806 {
1807 #if defined(CONFIG_QUOTA)
1808         struct ext4_sb_info *sbi = EXT4_SB(sb);
1809 
1810         if (sbi->s_jquota_fmt) {
1811                 char *fmtname = "";
1812 
1813                 switch (sbi->s_jquota_fmt) {
1814                 case QFMT_VFS_OLD:
1815                         fmtname = "vfsold";
1816                         break;
1817                 case QFMT_VFS_V0:
1818                         fmtname = "vfsv0";
1819                         break;
1820                 case QFMT_VFS_V1:
1821                         fmtname = "vfsv1";
1822                         break;
1823                 }
1824                 seq_printf(seq, ",jqfmt=%s", fmtname);
1825         }
1826 
1827         if (sbi->s_qf_names[USRQUOTA])
1828                 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1829 
1830         if (sbi->s_qf_names[GRPQUOTA])
1831                 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1832 #endif
1833 }
1834 
1835 static const char *token2str(int token)
1836 {
1837         const struct match_token *t;
1838 
1839         for (t = tokens; t->token != Opt_err; t++)
1840                 if (t->token == token && !strchr(t->pattern, '='))
1841                         break;
1842         return t->pattern;
1843 }
1844 
1845 /*
1846  * Show an option if
1847  *  - it's set to a non-default value OR
1848  *  - if the per-sb default is different from the global default
1849  */
1850 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1851                               int nodefs)
1852 {
1853         struct ext4_sb_info *sbi = EXT4_SB(sb);
1854         struct ext4_super_block *es = sbi->s_es;
1855         int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1856         const struct mount_opts *m;
1857         char sep = nodefs ? '\n' : ',';
1858 
1859 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1860 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1861 
1862         if (sbi->s_sb_block != 1)
1863                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1864 
1865         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1866                 int want_set = m->flags & MOPT_SET;
1867                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1868                     (m->flags & MOPT_CLEAR_ERR))
1869                         continue;
1870                 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1871                         continue; /* skip if same as the default */
1872                 if ((want_set &&
1873                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1874                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1875                         continue; /* select Opt_noFoo vs Opt_Foo */
1876                 SEQ_OPTS_PRINT("%s", token2str(m->token));
1877         }
1878 
1879         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1880             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1881                 SEQ_OPTS_PRINT("resuid=%u",
1882                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1883         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1884             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1885                 SEQ_OPTS_PRINT("resgid=%u",
1886                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1887         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1888         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1889                 SEQ_OPTS_PUTS("errors=remount-ro");
1890         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1891                 SEQ_OPTS_PUTS("errors=continue");
1892         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1893                 SEQ_OPTS_PUTS("errors=panic");
1894         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1895                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1896         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1897                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1898         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1899                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1900         if (sb->s_flags & MS_I_VERSION)
1901                 SEQ_OPTS_PUTS("i_version");
1902         if (nodefs || sbi->s_stripe)
1903                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1904         if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1905                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1906                         SEQ_OPTS_PUTS("data=journal");
1907                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1908                         SEQ_OPTS_PUTS("data=ordered");
1909                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1910                         SEQ_OPTS_PUTS("data=writeback");
1911         }
1912         if (nodefs ||
1913             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1914                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1915                                sbi->s_inode_readahead_blks);
1916 
1917         if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1918                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1919                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1920         if (nodefs || sbi->s_max_dir_size_kb)
1921                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1922         if (test_opt(sb, DATA_ERR_ABORT))
1923                 SEQ_OPTS_PUTS("data_err=abort");
1924 
1925         ext4_show_quota_options(seq, sb);
1926         return 0;
1927 }
1928 
1929 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1930 {
1931         return _ext4_show_options(seq, root->d_sb, 0);
1932 }
1933 
1934 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1935 {
1936         struct super_block *sb = seq->private;
1937         int rc;
1938 
1939         seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1940         rc = _ext4_show_options(seq, sb, 1);
1941         seq_puts(seq, "\n");
1942         return rc;
1943 }
1944 
1945 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1946                             int read_only)
1947 {
1948         struct ext4_sb_info *sbi = EXT4_SB(sb);
1949         int res = 0;
1950 
1951         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1952                 ext4_msg(sb, KERN_ERR, "revision level too high, "
1953                          "forcing read-only mode");
1954                 res = MS_RDONLY;
1955         }
1956         if (read_only)
1957                 goto done;
1958         if (!(sbi->s_mount_state & EXT4_VALID_FS))
1959                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1960                          "running e2fsck is recommended");
1961         else if (sbi->s_mount_state & EXT4_ERROR_FS)
1962                 ext4_msg(sb, KERN_WARNING,
1963                          "warning: mounting fs with errors, "
1964                          "running e2fsck is recommended");
1965         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1966                  le16_to_cpu(es->s_mnt_count) >=
1967                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1968                 ext4_msg(sb, KERN_WARNING,
1969                          "warning: maximal mount count reached, "
1970                          "running e2fsck is recommended");
1971         else if (le32_to_cpu(es->s_checkinterval) &&
1972                 (le32_to_cpu(es->s_lastcheck) +
1973                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1974                 ext4_msg(sb, KERN_WARNING,
1975                          "warning: checktime reached, "
1976                          "running e2fsck is recommended");
1977         if (!sbi->s_journal)
1978                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1979         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1980                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1981         le16_add_cpu(&es->s_mnt_count, 1);
1982         es->s_mtime = cpu_to_le32(get_seconds());
1983         ext4_update_dynamic_rev(sb);
1984         if (sbi->s_journal)
1985                 ext4_set_feature_journal_needs_recovery(sb);
1986 
1987         ext4_commit_super(sb, 1);
1988 done:
1989         if (test_opt(sb, DEBUG))
1990                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1991                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1992                         sb->s_blocksize,
1993                         sbi->s_groups_count,
1994                         EXT4_BLOCKS_PER_GROUP(sb),
1995                         EXT4_INODES_PER_GROUP(sb),
1996                         sbi->s_mount_opt, sbi->s_mount_opt2);
1997 
1998         cleancache_init_fs(sb);
1999         return res;
2000 }
2001 
2002 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2003 {
2004         struct ext4_sb_info *sbi = EXT4_SB(sb);
2005         struct flex_groups *new_groups;
2006         int size;
2007 
2008         if (!sbi->s_log_groups_per_flex)
2009                 return 0;
2010 
2011         size = ext4_flex_group(sbi, ngroup - 1) + 1;
2012         if (size <= sbi->s_flex_groups_allocated)
2013                 return 0;
2014 
2015         size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2016         new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2017         if (!new_groups) {
2018                 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2019                          size / (int) sizeof(struct flex_groups));
2020                 return -ENOMEM;
2021         }
2022 
2023         if (sbi->s_flex_groups) {
2024                 memcpy(new_groups, sbi->s_flex_groups,
2025                        (sbi->s_flex_groups_allocated *
2026                         sizeof(struct flex_groups)));
2027                 kvfree(sbi->s_flex_groups);
2028         }
2029         sbi->s_flex_groups = new_groups;
2030         sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2031         return 0;
2032 }
2033 
2034 static int ext4_fill_flex_info(struct super_block *sb)
2035 {
2036         struct ext4_sb_info *sbi = EXT4_SB(sb);
2037         struct ext4_group_desc *gdp = NULL;
2038         ext4_group_t flex_group;
2039         int i, err;
2040 
2041         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2042         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2043                 sbi->s_log_groups_per_flex = 0;
2044                 return 1;
2045         }
2046 
2047         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2048         if (err)
2049                 goto failed;
2050 
2051         for (i = 0; i < sbi->s_groups_count; i++) {
2052                 gdp = ext4_get_group_desc(sb, i, NULL);
2053 
2054                 flex_group = ext4_flex_group(sbi, i);
2055                 atomic_add(ext4_free_inodes_count(sb, gdp),
2056                            &sbi->s_flex_groups[flex_group].free_inodes);
2057                 atomic64_add(ext4_free_group_clusters(sb, gdp),
2058                              &sbi->s_flex_groups[flex_group].free_clusters);
2059                 atomic_add(ext4_used_dirs_count(sb, gdp),
2060                            &sbi->s_flex_groups[flex_group].used_dirs);
2061         }
2062 
2063         return 1;
2064 failed:
2065         return 0;
2066 }
2067 
2068 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2069                                    struct ext4_group_desc *gdp)
2070 {
2071         int offset = offsetof(struct ext4_group_desc, bg_checksum);
2072         __u16 crc = 0;
2073         __le32 le_group = cpu_to_le32(block_group);
2074         struct ext4_sb_info *sbi = EXT4_SB(sb);
2075 
2076         if (ext4_has_metadata_csum(sbi->s_sb)) {
2077                 /* Use new metadata_csum algorithm */
2078                 __u32 csum32;
2079                 __u16 dummy_csum = 0;
2080 
2081                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2082                                      sizeof(le_group));
2083                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2084                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2085                                      sizeof(dummy_csum));
2086                 offset += sizeof(dummy_csum);
2087                 if (offset < sbi->s_desc_size)
2088                         csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2089                                              sbi->s_desc_size - offset);
2090 
2091                 crc = csum32 & 0xFFFF;
2092                 goto out;
2093         }
2094 
2095         /* old crc16 code */
2096         if (!ext4_has_feature_gdt_csum(sb))
2097                 return 0;
2098 
2099         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2100         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2101         crc = crc16(crc, (__u8 *)gdp, offset);
2102         offset += sizeof(gdp->bg_checksum); /* skip checksum */
2103         /* for checksum of struct ext4_group_desc do the rest...*/
2104         if (ext4_has_feature_64bit(sb) &&
2105             offset < le16_to_cpu(sbi->s_es->s_desc_size))
2106                 crc = crc16(crc, (__u8 *)gdp + offset,
2107                             le16_to_cpu(sbi->s_es->s_desc_size) -
2108                                 offset);
2109 
2110 out:
2111         return cpu_to_le16(crc);
2112 }
2113 
2114 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2115                                 struct ext4_group_desc *gdp)
2116 {
2117         if (ext4_has_group_desc_csum(sb) &&
2118             (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2119                 return 0;
2120 
2121         return 1;
2122 }
2123 
2124 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2125                               struct ext4_group_desc *gdp)
2126 {
2127         if (!ext4_has_group_desc_csum(sb))
2128                 return;
2129         gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2130 }
2131 
2132 /* Called at mount-time, super-block is locked */
2133 static int ext4_check_descriptors(struct super_block *sb,
2134                                   ext4_fsblk_t sb_block,
2135                                   ext4_group_t *first_not_zeroed)
2136 {
2137         struct ext4_sb_info *sbi = EXT4_SB(sb);
2138         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2139         ext4_fsblk_t last_block;
2140         ext4_fsblk_t block_bitmap;
2141         ext4_fsblk_t inode_bitmap;
2142         ext4_fsblk_t inode_table;
2143         int flexbg_flag = 0;
2144         ext4_group_t i, grp = sbi->s_groups_count;
2145 
2146         if (ext4_has_feature_flex_bg(sb))
2147                 flexbg_flag = 1;
2148 
2149         ext4_debug("Checking group descriptors");
2150 
2151         for (i = 0; i < sbi->s_groups_count; i++) {
2152                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2153 
2154                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2155                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2156                 else
2157                         last_block = first_block +
2158                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2159 
2160                 if ((grp == sbi->s_groups_count) &&
2161                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2162                         grp = i;
2163 
2164                 block_bitmap = ext4_block_bitmap(sb, gdp);
2165                 if (block_bitmap == sb_block) {
2166                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2167                                  "Block bitmap for group %u overlaps "
2168                                  "superblock", i);
2169                 }
2170                 if (block_bitmap < first_block || block_bitmap > last_block) {
2171                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2172                                "Block bitmap for group %u not in group "
2173                                "(block %llu)!", i, block_bitmap);
2174                         return 0;
2175                 }
2176                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2177                 if (inode_bitmap == sb_block) {
2178                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2179                                  "Inode bitmap for group %u overlaps "
2180                                  "superblock", i);
2181                 }
2182                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2183                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2184                                "Inode bitmap for group %u not in group "
2185                                "(block %llu)!", i, inode_bitmap);
2186                         return 0;
2187                 }
2188                 inode_table = ext4_inode_table(sb, gdp);
2189                 if (inode_table == sb_block) {
2190                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2191                                  "Inode table for group %u overlaps "
2192                                  "superblock", i);
2193                 }
2194                 if (inode_table < first_block ||
2195                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2196                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2197                                "Inode table for group %u not in group "
2198                                "(block %llu)!", i, inode_table);
2199                         return 0;
2200                 }
2201                 ext4_lock_group(sb, i);
2202                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2203                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2204                                  "Checksum for group %u failed (%u!=%u)",
2205                                  i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2206                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2207                         if (!(sb->s_flags & MS_RDONLY)) {
2208                                 ext4_unlock_group(sb, i);
2209                                 return 0;
2210                         }
2211                 }
2212                 ext4_unlock_group(sb, i);
2213                 if (!flexbg_flag)
2214                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2215         }
2216         if (NULL != first_not_zeroed)
2217                 *first_not_zeroed = grp;
2218         return 1;
2219 }
2220 
2221 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2222  * the superblock) which were deleted from all directories, but held open by
2223  * a process at the time of a crash.  We walk the list and try to delete these
2224  * inodes at recovery time (only with a read-write filesystem).
2225  *
2226  * In order to keep the orphan inode chain consistent during traversal (in
2227  * case of crash during recovery), we link each inode into the superblock
2228  * orphan list_head and handle it the same way as an inode deletion during
2229  * normal operation (which journals the operations for us).
2230  *
2231  * We only do an iget() and an iput() on each inode, which is very safe if we
2232  * accidentally point at an in-use or already deleted inode.  The worst that
2233  * can happen in this case is that we get a "bit already cleared" message from
2234  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2235  * e2fsck was run on this filesystem, and it must have already done the orphan
2236  * inode cleanup for us, so we can safely abort without any further action.
2237  */
2238 static void ext4_orphan_cleanup(struct super_block *sb,
2239                                 struct ext4_super_block *es)
2240 {
2241         unsigned int s_flags = sb->s_flags;
2242         int nr_orphans = 0, nr_truncates = 0;
2243 #ifdef CONFIG_QUOTA
2244         int i;
2245 #endif
2246         if (!es->s_last_orphan) {
2247                 jbd_debug(4, "no orphan inodes to clean up\n");
2248                 return;
2249         }
2250 
2251         if (bdev_read_only(sb->s_bdev)) {
2252                 ext4_msg(sb, KERN_ERR, "write access "
2253                         "unavailable, skipping orphan cleanup");
2254                 return;
2255         }
2256 
2257         /* Check if feature set would not allow a r/w mount */
2258         if (!ext4_feature_set_ok(sb, 0)) {
2259                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2260                          "unknown ROCOMPAT features");
2261                 return;
2262         }
2263 
2264         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2265                 /* don't clear list on RO mount w/ errors */
2266                 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2267                         ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2268                                   "clearing orphan list.\n");
2269                         es->s_last_orphan = 0;
2270                 }
2271                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2272                 return;
2273         }
2274 
2275         if (s_flags & MS_RDONLY) {
2276                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2277                 sb->s_flags &= ~MS_RDONLY;
2278         }
2279 #ifdef CONFIG_QUOTA
2280         /* Needed for iput() to work correctly and not trash data */
2281         sb->s_flags |= MS_ACTIVE;
2282         /* Turn on quotas so that they are updated correctly */
2283         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2284                 if (EXT4_SB(sb)->s_qf_names[i]) {
2285                         int ret = ext4_quota_on_mount(sb, i);
2286                         if (ret < 0)
2287                                 ext4_msg(sb, KERN_ERR,
2288                                         "Cannot turn on journaled "
2289                                         "quota: error %d", ret);
2290                 }
2291         }
2292 #endif
2293 
2294         while (es->s_last_orphan) {
2295                 struct inode *inode;
2296 
2297                 /*
2298                  * We may have encountered an error during cleanup; if
2299                  * so, skip the rest.
2300                  */
2301                 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2302                         jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2303                         es->s_last_orphan = 0;
2304                         break;
2305                 }
2306 
2307                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2308                 if (IS_ERR(inode)) {
2309                         es->s_last_orphan = 0;
2310                         break;
2311                 }
2312 
2313                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2314                 dquot_initialize(inode);
2315                 if (inode->i_nlink) {
2316                         if (test_opt(sb, DEBUG))
2317                                 ext4_msg(sb, KERN_DEBUG,
2318                                         "%s: truncating inode %lu to %lld bytes",
2319                                         __func__, inode->i_ino, inode->i_size);
2320                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2321                                   inode->i_ino, inode->i_size);
2322                         inode_lock(inode);
2323                         truncate_inode_pages(inode->i_mapping, inode->i_size);
2324                         ext4_truncate(inode);
2325                         inode_unlock(inode);
2326                         nr_truncates++;
2327                 } else {
2328                         if (test_opt(sb, DEBUG))
2329                                 ext4_msg(sb, KERN_DEBUG,
2330                                         "%s: deleting unreferenced inode %lu",
2331                                         __func__, inode->i_ino);
2332                         jbd_debug(2, "deleting unreferenced inode %lu\n",
2333                                   inode->i_ino);
2334                         nr_orphans++;
2335                 }
2336                 iput(inode);  /* The delete magic happens here! */
2337         }
2338 
2339 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2340 
2341         if (nr_orphans)
2342                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2343                        PLURAL(nr_orphans));
2344         if (nr_truncates)
2345                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2346                        PLURAL(nr_truncates));
2347 #ifdef CONFIG_QUOTA
2348         /* Turn quotas off */
2349         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2350                 if (sb_dqopt(sb)->files[i])
2351                         dquot_quota_off(sb, i);
2352         }
2353 #endif
2354         sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2355 }
2356 
2357 /*
2358  * Maximal extent format file size.
2359  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2360  * extent format containers, within a sector_t, and within i_blocks
2361  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2362  * so that won't be a limiting factor.
2363  *
2364  * However there is other limiting factor. We do store extents in the form
2365  * of starting block and length, hence the resulting length of the extent
2366  * covering maximum file size must fit into on-disk format containers as
2367  * well. Given that length is always by 1 unit bigger than max unit (because
2368  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2369  *
2370  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2371  */
2372 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2373 {
2374         loff_t res;
2375         loff_t upper_limit = MAX_LFS_FILESIZE;
2376 
2377         /* small i_blocks in vfs inode? */
2378         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2379                 /*
2380                  * CONFIG_LBDAF is not enabled implies the inode
2381                  * i_block represent total blocks in 512 bytes
2382                  * 32 == size of vfs inode i_blocks * 8
2383                  */
2384                 upper_limit = (1LL << 32) - 1;
2385 
2386                 /* total blocks in file system block size */
2387                 upper_limit >>= (blkbits - 9);
2388                 upper_limit <<= blkbits;
2389         }
2390 
2391         /*
2392          * 32-bit extent-start container, ee_block. We lower the maxbytes
2393          * by one fs block, so ee_len can cover the extent of maximum file
2394          * size
2395          */
2396         res = (1LL << 32) - 1;
2397         res <<= blkbits;
2398 
2399         /* Sanity check against vm- & vfs- imposed limits */
2400         if (res > upper_limit)
2401                 res = upper_limit;
2402 
2403         return res;
2404 }
2405 
2406 /*
2407  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2408  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2409  * We need to be 1 filesystem block less than the 2^48 sector limit.
2410  */
2411 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2412 {
2413         loff_t res = EXT4_NDIR_BLOCKS;
2414         int meta_blocks;
2415         loff_t upper_limit;
2416         /* This is calculated to be the largest file size for a dense, block
2417          * mapped file such that the file's total number of 512-byte sectors,
2418          * including data and all indirect blocks, does not exceed (2^48 - 1).
2419          *
2420          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2421          * number of 512-byte sectors of the file.
2422          */
2423 
2424         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2425                 /*
2426                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
2427                  * the inode i_block field represents total file blocks in
2428                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2429                  */
2430                 upper_limit = (1LL << 32) - 1;
2431 
2432                 /* total blocks in file system block size */
2433                 upper_limit >>= (bits - 9);
2434 
2435         } else {
2436                 /*
2437                  * We use 48 bit ext4_inode i_blocks
2438                  * With EXT4_HUGE_FILE_FL set the i_blocks
2439                  * represent total number of blocks in
2440                  * file system block size
2441                  */
2442                 upper_limit = (1LL << 48) - 1;
2443 
2444         }
2445 
2446         /* indirect blocks */
2447         meta_blocks = 1;
2448         /* double indirect blocks */
2449         meta_blocks += 1 + (1LL << (bits-2));
2450         /* tripple indirect blocks */
2451         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2452 
2453         upper_limit -= meta_blocks;
2454         upper_limit <<= bits;
2455 
2456         res += 1LL << (bits-2);
2457         res += 1LL << (2*(bits-2));
2458         res += 1LL << (3*(bits-2));
2459         res <<= bits;
2460         if (res > upper_limit)
2461                 res = upper_limit;
2462 
2463         if (res > MAX_LFS_FILESIZE)
2464                 res = MAX_LFS_FILESIZE;
2465 
2466         return res;
2467 }
2468 
2469 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2470                                    ext4_fsblk_t logical_sb_block, int nr)
2471 {
2472         struct ext4_sb_info *sbi = EXT4_SB(sb);
2473         ext4_group_t bg, first_meta_bg;
2474         int has_super = 0;
2475 
2476         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2477 
2478         if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2479                 return logical_sb_block + nr + 1;
2480         bg = sbi->s_desc_per_block * nr;
2481         if (ext4_bg_has_super(sb, bg))
2482                 has_super = 1;
2483 
2484         /*
2485          * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2486          * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2487          * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2488          * compensate.
2489          */
2490         if (sb->s_blocksize == 1024 && nr == 0 &&
2491             le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2492                 has_super++;
2493 
2494         return (has_super + ext4_group_first_block_no(sb, bg));
2495 }
2496 
2497 /**
2498  * ext4_get_stripe_size: Get the stripe size.
2499  * @sbi: In memory super block info
2500  *
2501  * If we have specified it via mount option, then
2502  * use the mount option value. If the value specified at mount time is
2503  * greater than the blocks per group use the super block value.
2504  * If the super block value is greater than blocks per group return 0.
2505  * Allocator needs it be less than blocks per group.
2506  *
2507  */
2508 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2509 {
2510         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2511         unsigned long stripe_width =
2512                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2513         int ret;
2514 
2515         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2516                 ret = sbi->s_stripe;
2517         else if (stripe_width <= sbi->s_blocks_per_group)
2518                 ret = stripe_width;
2519         else if (stride <= sbi->s_blocks_per_group)
2520                 ret = stride;
2521         else
2522                 ret = 0;
2523 
2524         /*
2525          * If the stripe width is 1, this makes no sense and
2526          * we set it to 0 to turn off stripe handling code.
2527          */
2528         if (ret <= 1)
2529                 ret = 0;
2530 
2531         return ret;
2532 }
2533 
2534 /*
2535  * Check whether this filesystem can be mounted based on
2536  * the features present and the RDONLY/RDWR mount requested.
2537  * Returns 1 if this filesystem can be mounted as requested,
2538  * 0 if it cannot be.
2539  */
2540 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2541 {
2542         if (ext4_has_unknown_ext4_incompat_features(sb)) {
2543                 ext4_msg(sb, KERN_ERR,
2544                         "Couldn't mount because of "
2545                         "unsupported optional features (%x)",
2546                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2547                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2548                 return 0;
2549         }
2550 
2551         if (readonly)
2552                 return 1;
2553 
2554         if (ext4_has_feature_readonly(sb)) {
2555                 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2556                 sb->s_flags |= MS_RDONLY;
2557                 return 1;
2558         }
2559 
2560         /* Check that feature set is OK for a read-write mount */
2561         if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2562                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2563                          "unsupported optional features (%x)",
2564                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2565                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2566                 return 0;
2567         }
2568         /*
2569          * Large file size enabled file system can only be mounted
2570          * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2571          */
2572         if (ext4_has_feature_huge_file(sb)) {
2573                 if (sizeof(blkcnt_t) < sizeof(u64)) {
2574                         ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2575                                  "cannot be mounted RDWR without "
2576                                  "CONFIG_LBDAF");
2577                         return 0;
2578                 }
2579         }
2580         if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2581                 ext4_msg(sb, KERN_ERR,
2582                          "Can't support bigalloc feature without "
2583                          "extents feature\n");
2584                 return 0;
2585         }
2586 
2587 #ifndef CONFIG_QUOTA
2588         if (ext4_has_feature_quota(sb) && !readonly) {
2589                 ext4_msg(sb, KERN_ERR,
2590                          "Filesystem with quota feature cannot be mounted RDWR "
2591                          "without CONFIG_QUOTA");
2592                 return 0;
2593         }
2594         if (ext4_has_feature_project(sb) && !readonly) {
2595                 ext4_msg(sb, KERN_ERR,
2596                          "Filesystem with project quota feature cannot be mounted RDWR "
2597                          "without CONFIG_QUOTA");
2598                 return 0;
2599         }
2600 #endif  /* CONFIG_QUOTA */
2601         return 1;
2602 }
2603 
2604 /*
2605  * This function is called once a day if we have errors logged
2606  * on the file system
2607  */
2608 static void print_daily_error_info(unsigned long arg)
2609 {
2610         struct super_block *sb = (struct super_block *) arg;
2611         struct ext4_sb_info *sbi;
2612         struct ext4_super_block *es;
2613 
2614         sbi = EXT4_SB(sb);
2615         es = sbi->s_es;
2616 
2617         if (es->s_error_count)
2618                 /* fsck newer than v1.41.13 is needed to clean this condition. */
2619                 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2620                          le32_to_cpu(es->s_error_count));
2621         if (es->s_first_error_time) {
2622                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2623                        sb->s_id, le32_to_cpu(es->s_first_error_time),
2624                        (int) sizeof(es->s_first_error_func),
2625                        es->s_first_error_func,
2626                        le32_to_cpu(es->s_first_error_line));
2627                 if (es->s_first_error_ino)
2628                         printk(": inode %u",
2629                                le32_to_cpu(es->s_first_error_ino));
2630                 if (es->s_first_error_block)
2631                         printk(": block %llu", (unsigned long long)
2632                                le64_to_cpu(es->s_first_error_block));
2633                 printk("\n");
2634         }
2635         if (es->s_last_error_time) {
2636                 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2637                        sb->s_id, le32_to_cpu(es->s_last_error_time),
2638                        (int) sizeof(es->s_last_error_func),
2639                        es->s_last_error_func,
2640                        le32_to_cpu(es->s_last_error_line));
2641                 if (es->s_last_error_ino)
2642                         printk(": inode %u",
2643                                le32_to_cpu(es->s_last_error_ino));
2644                 if (es->s_last_error_block)
2645                         printk(": block %llu", (unsigned long long)
2646                                le64_to_cpu(es->s_last_error_block));
2647                 printk("\n");
2648         }
2649         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2650 }
2651 
2652 /* Find next suitable group and run ext4_init_inode_table */
2653 static int ext4_run_li_request(struct ext4_li_request *elr)
2654 {
2655         struct ext4_group_desc *gdp = NULL;
2656         ext4_group_t group, ngroups;
2657         struct super_block *sb;
2658         unsigned long timeout = 0;
2659         int ret = 0;
2660 
2661         sb = elr->lr_super;
2662         ngroups = EXT4_SB(sb)->s_groups_count;
2663 
2664         sb_start_write(sb);
2665         for (group = elr->lr_next_group; group < ngroups; group++) {
2666                 gdp = ext4_get_group_desc(sb, group, NULL);
2667                 if (!gdp) {
2668                         ret = 1;
2669                         break;
2670                 }
2671 
2672                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2673                         break;
2674         }
2675 
2676         if (group >= ngroups)
2677                 ret = 1;
2678 
2679         if (!ret) {
2680                 timeout = jiffies;
2681                 ret = ext4_init_inode_table(sb, group,
2682                                             elr->lr_timeout ? 0 : 1);
2683                 if (elr->lr_timeout == 0) {
2684                         timeout = (jiffies - timeout) *
2685                                   elr->lr_sbi->s_li_wait_mult;
2686                         elr->lr_timeout = timeout;
2687                 }
2688                 elr->lr_next_sched = jiffies + elr->lr_timeout;
2689                 elr->lr_next_group = group + 1;
2690         }
2691         sb_end_write(sb);
2692 
2693         return ret;
2694 }
2695 
2696 /*
2697  * Remove lr_request from the list_request and free the
2698  * request structure. Should be called with li_list_mtx held
2699  */
2700 static void ext4_remove_li_request(struct ext4_li_request *elr)
2701 {
2702         struct ext4_sb_info *sbi;
2703 
2704         if (!elr)
2705                 return;
2706 
2707         sbi = elr->lr_sbi;
2708 
2709         list_del(&elr->lr_request);
2710         sbi->s_li_request = NULL;
2711         kfree(elr);
2712 }
2713 
2714 static void ext4_unregister_li_request(struct super_block *sb)
2715 {
2716         mutex_lock(&ext4_li_mtx);
2717         if (!ext4_li_info) {
2718                 mutex_unlock(&ext4_li_mtx);
2719                 return;
2720         }
2721 
2722         mutex_lock(&ext4_li_info->li_list_mtx);
2723         ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2724         mutex_unlock(&ext4_li_info->li_list_mtx);
2725         mutex_unlock(&ext4_li_mtx);
2726 }
2727 
2728 static struct task_struct *ext4_lazyinit_task;
2729 
2730 /*
2731  * This is the function where ext4lazyinit thread lives. It walks
2732  * through the request list searching for next scheduled filesystem.
2733  * When such a fs is found, run the lazy initialization request
2734  * (ext4_rn_li_request) and keep track of the time spend in this
2735  * function. Based on that time we compute next schedule time of
2736  * the request. When walking through the list is complete, compute
2737  * next waking time and put itself into sleep.
2738  */
2739 static int ext4_lazyinit_thread(void *arg)
2740 {
2741         struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2742         struct list_head *pos, *n;
2743         struct ext4_li_request *elr;
2744         unsigned long next_wakeup, cur;
2745 
2746         BUG_ON(NULL == eli);
2747 
2748 cont_thread:
2749         while (true) {
2750                 next_wakeup = MAX_JIFFY_OFFSET;
2751 
2752                 mutex_lock(&eli->li_list_mtx);
2753                 if (list_empty(&eli->li_request_list)) {
2754                         mutex_unlock(&eli->li_list_mtx);
2755                         goto exit_thread;
2756                 }
2757 
2758                 list_for_each_safe(pos, n, &eli->li_request_list) {
2759                         elr = list_entry(pos, struct ext4_li_request,
2760                                          lr_request);
2761 
2762                         if (time_after_eq(jiffies, elr->lr_next_sched)) {
2763                                 if (ext4_run_li_request(elr) != 0) {
2764                                         /* error, remove the lazy_init job */
2765                                         ext4_remove_li_request(elr);
2766                                         continue;
2767                                 }
2768                         }
2769 
2770                         if (time_before(elr->lr_next_sched, next_wakeup))
2771                                 next_wakeup = elr->lr_next_sched;
2772                 }
2773                 mutex_unlock(&eli->li_list_mtx);
2774 
2775                 try_to_freeze();
2776 
2777                 cur = jiffies;
2778                 if ((time_after_eq(cur, next_wakeup)) ||
2779                     (MAX_JIFFY_OFFSET == next_wakeup)) {
2780                         cond_resched();
2781                         continue;
2782                 }
2783 
2784                 schedule_timeout_interruptible(next_wakeup - cur);
2785 
2786                 if (kthread_should_stop()) {
2787                         ext4_clear_request_list();
2788                         goto exit_thread;
2789                 }
2790         }
2791 
2792 exit_thread:
2793         /*
2794          * It looks like the request list is empty, but we need
2795          * to check it under the li_list_mtx lock, to prevent any
2796          * additions into it, and of course we should lock ext4_li_mtx
2797          * to atomically free the list and ext4_li_info, because at
2798          * this point another ext4 filesystem could be registering
2799          * new one.
2800          */
2801         mutex_lock(&ext4_li_mtx);
2802         mutex_lock(&eli->li_list_mtx);
2803         if (!list_empty(&eli->li_request_list)) {
2804                 mutex_unlock(&eli->li_list_mtx);
2805                 mutex_unlock(&ext4_li_mtx);
2806                 goto cont_thread;
2807         }
2808         mutex_unlock(&eli->li_list_mtx);
2809         kfree(ext4_li_info);
2810         ext4_li_info = NULL;
2811         mutex_unlock(&ext4_li_mtx);
2812 
2813         return 0;
2814 }
2815 
2816 static void ext4_clear_request_list(void)
2817 {
2818         struct list_head *pos, *n;
2819         struct ext4_li_request *elr;
2820 
2821         mutex_lock(&ext4_li_info->li_list_mtx);
2822         list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2823                 elr = list_entry(pos, struct ext4_li_request,
2824                                  lr_request);
2825                 ext4_remove_li_request(elr);
2826         }
2827         mutex_unlock(&ext4_li_info->li_list_mtx);
2828 }
2829 
2830 static int ext4_run_lazyinit_thread(void)
2831 {
2832         ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2833                                          ext4_li_info, "ext4lazyinit");
2834         if (IS_ERR(ext4_lazyinit_task)) {
2835                 int err = PTR_ERR(ext4_lazyinit_task);
2836                 ext4_clear_request_list();
2837                 kfree(ext4_li_info);
2838                 ext4_li_info = NULL;
2839                 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2840                                  "initialization thread\n",
2841                                  err);
2842                 return err;
2843         }
2844         ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2845         return 0;
2846 }
2847 
2848 /*
2849  * Check whether it make sense to run itable init. thread or not.
2850  * If there is at least one uninitialized inode table, return
2851  * corresponding group number, else the loop goes through all
2852  * groups and return total number of groups.
2853  */
2854 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2855 {
2856         ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2857         struct ext4_group_desc *gdp = NULL;
2858 
2859         for (group = 0; group < ngroups; group++) {
2860                 gdp = ext4_get_group_desc(sb, group, NULL);
2861                 if (!gdp)
2862                         continue;
2863 
2864                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2865                         break;
2866         }
2867 
2868         return group;
2869 }
2870 
2871 static int ext4_li_info_new(void)
2872 {
2873         struct ext4_lazy_init *eli = NULL;
2874 
2875         eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2876         if (!eli)
2877                 return -ENOMEM;
2878 
2879         INIT_LIST_HEAD(&eli->li_request_list);
2880         mutex_init(&eli->li_list_mtx);
2881 
2882         eli->li_state |= EXT4_LAZYINIT_QUIT;
2883 
2884         ext4_li_info = eli;
2885 
2886         return 0;
2887 }
2888 
2889 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2890                                             ext4_group_t start)
2891 {
2892         struct ext4_sb_info *sbi = EXT4_SB(sb);
2893         struct ext4_li_request *elr;
2894 
2895         elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2896         if (!elr)
2897                 return NULL;
2898 
2899         elr->lr_super = sb;
2900         elr->lr_sbi = sbi;
2901         elr->lr_next_group = start;
2902 
2903         /*
2904          * Randomize first schedule time of the request to
2905          * spread the inode table initialization requests
2906          * better.
2907          */
2908         elr->lr_next_sched = jiffies + (prandom_u32() %
2909                                 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2910         return elr;
2911 }
2912 
2913 int ext4_register_li_request(struct super_block *sb,
2914                              ext4_group_t first_not_zeroed)
2915 {
2916         struct ext4_sb_info *sbi = EXT4_SB(sb);
2917         struct ext4_li_request *elr = NULL;
2918         ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2919         int ret = 0;
2920 
2921         mutex_lock(&ext4_li_mtx);
2922         if (sbi->s_li_request != NULL) {
2923                 /*
2924                  * Reset timeout so it can be computed again, because
2925                  * s_li_wait_mult might have changed.
2926                  */
2927                 sbi->s_li_request->lr_timeout = 0;
2928                 goto out;
2929         }
2930 
2931         if (first_not_zeroed == ngroups ||
2932             (sb->s_flags & MS_RDONLY) ||
2933             !test_opt(sb, INIT_INODE_TABLE))
2934                 goto out;
2935 
2936         elr = ext4_li_request_new(sb, first_not_zeroed);
2937         if (!elr) {
2938                 ret = -ENOMEM;
2939                 goto out;
2940         }
2941 
2942         if (NULL == ext4_li_info) {
2943                 ret = ext4_li_info_new();
2944                 if (ret)
2945                         goto out;
2946         }
2947 
2948         mutex_lock(&ext4_li_info->li_list_mtx);
2949         list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2950         mutex_unlock(&ext4_li_info->li_list_mtx);
2951 
2952         sbi->s_li_request = elr;
2953         /*
2954          * set elr to NULL here since it has been inserted to
2955          * the request_list and the removal and free of it is
2956          * handled by ext4_clear_request_list from now on.
2957          */
2958         elr = NULL;
2959 
2960         if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2961                 ret = ext4_run_lazyinit_thread();
2962                 if (ret)
2963                         goto out;
2964         }
2965 out:
2966         mutex_unlock(&ext4_li_mtx);
2967         if (ret)
2968                 kfree(elr);
2969         return ret;
2970 }
2971 
2972 /*
2973  * We do not need to lock anything since this is called on
2974  * module unload.
2975  */
2976 static void ext4_destroy_lazyinit_thread(void)
2977 {
2978         /*
2979          * If thread exited earlier
2980          * there's nothing to be done.
2981          */
2982         if (!ext4_li_info || !ext4_lazyinit_task)
2983                 return;
2984 
2985         kthread_stop(ext4_lazyinit_task);
2986 }
2987 
2988 static int set_journal_csum_feature_set(struct super_block *sb)
2989 {
2990         int ret = 1;
2991         int compat, incompat;
2992         struct ext4_sb_info *sbi = EXT4_SB(sb);
2993 
2994         if (ext4_has_metadata_csum(sb)) {
2995                 /* journal checksum v3 */
2996                 compat = 0;
2997                 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2998         } else {
2999                 /* journal checksum v1 */
3000                 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3001                 incompat = 0;
3002         }
3003 
3004         jbd2_journal_clear_features(sbi->s_journal,
3005                         JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3006                         JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3007                         JBD2_FEATURE_INCOMPAT_CSUM_V2);
3008         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3009                 ret = jbd2_journal_set_features(sbi->s_journal,
3010                                 compat, 0,
3011                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3012                                 incompat);
3013         } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3014                 ret = jbd2_journal_set_features(sbi->s_journal,
3015                                 compat, 0,
3016                                 incompat);
3017                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3018                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3019         } else {
3020                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3021                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3022         }
3023 
3024         return ret;
3025 }
3026 
3027 /*
3028  * Note: calculating the overhead so we can be compatible with
3029  * historical BSD practice is quite difficult in the face of
3030  * clusters/bigalloc.  This is because multiple metadata blocks from
3031  * different block group can end up in the same allocation cluster.
3032  * Calculating the exact overhead in the face of clustered allocation
3033  * requires either O(all block bitmaps) in memory or O(number of block
3034  * groups**2) in time.  We will still calculate the superblock for
3035  * older file systems --- and if we come across with a bigalloc file
3036  * system with zero in s_overhead_clusters the estimate will be close to
3037  * correct especially for very large cluster sizes --- but for newer
3038  * file systems, it's better to calculate this figure once at mkfs
3039  * time, and store it in the superblock.  If the superblock value is
3040  * present (even for non-bigalloc file systems), we will use it.
3041  */
3042 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3043                           char *buf)
3044 {
3045         struct ext4_sb_info     *sbi = EXT4_SB(sb);
3046         struct ext4_group_desc  *gdp;
3047         ext4_fsblk_t            first_block, last_block, b;
3048         ext4_group_t            i, ngroups = ext4_get_groups_count(sb);
3049         int                     s, j, count = 0;
3050 
3051         if (!ext4_has_feature_bigalloc(sb))
3052                 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3053                         sbi->s_itb_per_group + 2);
3054 
3055         first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3056                 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3057         last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3058         for (i = 0; i < ngroups; i++) {
3059                 gdp = ext4_get_group_desc(sb, i, NULL);
3060                 b = ext4_block_bitmap(sb, gdp);
3061                 if (b >= first_block && b <= last_block) {
3062                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3063                         count++;
3064                 }
3065                 b = ext4_inode_bitmap(sb, gdp);
3066                 if (b >= first_block && b <= last_block) {
3067                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3068                         count++;
3069                 }
3070                 b = ext4_inode_table(sb, gdp);
3071                 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3072                         for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3073                                 int c = EXT4_B2C(sbi, b - first_block);
3074                                 ext4_set_bit(c, buf);
3075                                 count++;
3076                         }
3077                 if (i != grp)
3078                         continue;
3079                 s = 0;
3080                 if (ext4_bg_has_super(sb, grp)) {
3081                         ext4_set_bit(s++, buf);
3082                         count++;
3083                 }
3084                 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3085                         ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3086                         count++;
3087                 }
3088         }
3089         if (!count)
3090                 return 0;
3091         return EXT4_CLUSTERS_PER_GROUP(sb) -
3092                 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3093 }
3094 
3095 /*
3096  * Compute the overhead and stash it in sbi->s_overhead
3097  */
3098 int ext4_calculate_overhead(struct super_block *sb)
3099 {
3100         struct ext4_sb_info *sbi = EXT4_SB(sb);
3101         struct ext4_super_block *es = sbi->s_es;
3102         ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3103         ext4_fsblk_t overhead = 0;
3104         char *buf = (char *) get_zeroed_page(GFP_NOFS);
3105 
3106         if (!buf)
3107                 return -ENOMEM;
3108 
3109         /*
3110          * Compute the overhead (FS structures).  This is constant
3111          * for a given filesystem unless the number of block groups
3112          * changes so we cache the previous value until it does.
3113          */
3114 
3115         /*
3116          * All of the blocks before first_data_block are overhead
3117          */
3118         overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3119 
3120         /*
3121          * Add the overhead found in each block group
3122          */
3123         for (i = 0; i < ngroups; i++) {
3124                 int blks;
3125 
3126                 blks = count_overhead(sb, i, buf);
3127                 overhead += blks;
3128                 if (blks)
3129                         memset(buf, 0, PAGE_SIZE);
3130                 cond_resched();
3131         }
3132         /* Add the internal journal blocks as well */
3133         if (sbi->s_journal && !sbi->journal_bdev)
3134                 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3135 
3136         sbi->s_overhead = overhead;
3137         smp_wmb();
3138         free_page((unsigned long) buf);
3139         return 0;
3140 }
3141 
3142 static void ext4_set_resv_clusters(struct super_block *sb)
3143 {
3144         ext4_fsblk_t resv_clusters;
3145         struct ext4_sb_info *sbi = EXT4_SB(sb);
3146 
3147         /*
3148          * There's no need to reserve anything when we aren't using extents.
3149          * The space estimates are exact, there are no unwritten extents,
3150          * hole punching doesn't need new metadata... This is needed especially
3151          * to keep ext2/3 backward compatibility.
3152          */
3153         if (!ext4_has_feature_extents(sb))
3154                 return;
3155         /*
3156          * By default we reserve 2% or 4096 clusters, whichever is smaller.
3157          * This should cover the situations where we can not afford to run
3158          * out of space like for example punch hole, or converting
3159          * unwritten extents in delalloc path. In most cases such
3160          * allocation would require 1, or 2 blocks, higher numbers are
3161          * very rare.
3162          */
3163         resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3164                          sbi->s_cluster_bits);
3165 
3166         do_div(resv_clusters, 50);
3167         resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3168 
3169         atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3170 }
3171 
3172 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3173 {
3174         char *orig_data = kstrdup(data, GFP_KERNEL);
3175         struct buffer_head *bh;
3176         struct ext4_super_block *es = NULL;
3177         struct ext4_sb_info *sbi;
3178         ext4_fsblk_t block;
3179         ext4_fsblk_t sb_block = get_sb_block(&data);
3180         ext4_fsblk_t logical_sb_block;
3181         unsigned long offset = 0;
3182         unsigned long journal_devnum = 0;
3183         unsigned long def_mount_opts;
3184         struct inode *root;
3185         const char *descr;
3186         int ret = -ENOMEM;
3187         int blocksize, clustersize;
3188         unsigned int db_count;
3189         unsigned int i;
3190         int needs_recovery, has_huge_files, has_bigalloc;
3191         __u64 blocks_count;
3192         int err = 0;
3193         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3194         ext4_group_t first_not_zeroed;
3195 
3196         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3197         if (!sbi)
3198                 goto out_free_orig;
3199 
3200         sbi->s_blockgroup_lock =
3201                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3202         if (!sbi->s_blockgroup_lock) {
3203                 kfree(sbi);
3204                 goto out_free_orig;
3205         }
3206         sb->s_fs_info = sbi;
3207         sbi->s_sb = sb;
3208         sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3209         sbi->s_sb_block = sb_block;
3210         if (sb->s_bdev->bd_part)
3211                 sbi->s_sectors_written_start =
3212                         part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3213 
3214         /* Cleanup superblock name */
3215         strreplace(sb->s_id, '/', '!');
3216 
3217         /* -EINVAL is default */
3218         ret = -EINVAL;
3219         blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3220         if (!blocksize) {
3221                 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3222                 goto out_fail;
3223         }
3224 
3225         /*
3226          * The ext4 superblock will not be buffer aligned for other than 1kB
3227          * block sizes.  We need to calculate the offset from buffer start.
3228          */
3229         if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3230                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3231                 offset = do_div(logical_sb_block, blocksize);
3232         } else {
3233                 logical_sb_block = sb_block;
3234         }
3235 
3236         if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3237                 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3238                 goto out_fail;
3239         }
3240         /*
3241          * Note: s_es must be initialized as soon as possible because
3242          *       some ext4 macro-instructions depend on its value
3243          */
3244         es = (struct ext4_super_block *) (bh->b_data + offset);
3245         sbi->s_es = es;
3246         sb->s_magic = le16_to_cpu(es->s_magic);
3247         if (sb->s_magic != EXT4_SUPER_MAGIC)
3248                 goto cantfind_ext4;
3249         sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3250 
3251         /* Warn if metadata_csum and gdt_csum are both set. */
3252         if (ext4_has_feature_metadata_csum(sb) &&
3253             ext4_has_feature_gdt_csum(sb))
3254                 ext4_warning(sb, "metadata_csum and uninit_bg are "
3255                              "redundant flags; please run fsck.");
3256 
3257         /* Check for a known checksum algorithm */
3258         if (!ext4_verify_csum_type(sb, es)) {
3259                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3260                          "unknown checksum algorithm.");
3261                 silent = 1;
3262                 goto cantfind_ext4;
3263         }
3264 
3265         /* Load the checksum driver */
3266         if (ext4_has_feature_metadata_csum(sb)) {
3267                 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3268                 if (IS_ERR(sbi->s_chksum_driver)) {
3269                         ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3270                         ret = PTR_ERR(sbi->s_chksum_driver);
3271                         sbi->s_chksum_driver = NULL;
3272                         goto failed_mount;
3273                 }
3274         }
3275 
3276         /* Check superblock checksum */
3277         if (!ext4_superblock_csum_verify(sb, es)) {
3278                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3279                          "invalid superblock checksum.  Run e2fsck?");
3280                 silent = 1;
3281                 ret = -EFSBADCRC;
3282                 goto cantfind_ext4;
3283         }
3284 
3285         /* Precompute checksum seed for all metadata */
3286         if (ext4_has_feature_csum_seed(sb))
3287                 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3288         else if (ext4_has_metadata_csum(sb))
3289                 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3290                                                sizeof(es->s_uuid));
3291 
3292         /* Set defaults before we parse the mount options */
3293         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3294         set_opt(sb, INIT_INODE_TABLE);
3295         if (def_mount_opts & EXT4_DEFM_DEBUG)
3296                 set_opt(sb, DEBUG);
3297         if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3298                 set_opt(sb, GRPID);
3299         if (def_mount_opts & EXT4_DEFM_UID16)
3300                 set_opt(sb, NO_UID32);
3301         /* xattr user namespace & acls are now defaulted on */
3302         set_opt(sb, XATTR_USER);
3303 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3304         set_opt(sb, POSIX_ACL);
3305 #endif
3306         /* don't forget to enable journal_csum when metadata_csum is enabled. */
3307         if (ext4_has_metadata_csum(sb))
3308                 set_opt(sb, JOURNAL_CHECKSUM);
3309 
3310         if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3311                 set_opt(sb, JOURNAL_DATA);
3312         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3313                 set_opt(sb, ORDERED_DATA);
3314         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3315                 set_opt(sb, WRITEBACK_DATA);
3316 
3317         if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3318                 set_opt(sb, ERRORS_PANIC);
3319         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3320                 set_opt(sb, ERRORS_CONT);
3321         else
3322                 set_opt(sb, ERRORS_RO);
3323         /* block_validity enabled by default; disable with noblock_validity */
3324         set_opt(sb, BLOCK_VALIDITY);
3325         if (def_mount_opts & EXT4_DEFM_DISCARD)
3326                 set_opt(sb, DISCARD);
3327 
3328         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3329         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3330         sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3331         sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3332         sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3333 
3334         if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3335                 set_opt(sb, BARRIER);
3336 
3337         /*
3338          * enable delayed allocation by default
3339          * Use -o nodelalloc to turn it off
3340          */
3341         if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3342             ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3343                 set_opt(sb, DELALLOC);
3344 
3345         /*
3346          * set default s_li_wait_mult for lazyinit, for the case there is
3347          * no mount option specified.
3348          */
3349         sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3350 
3351         if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3352                            &journal_devnum, &journal_ioprio, 0)) {
3353                 ext4_msg(sb, KERN_WARNING,
3354                          "failed to parse options in superblock: %s",
3355                          sbi->s_es->s_mount_opts);
3356         }
3357         sbi->s_def_mount_opt = sbi->s_mount_opt;
3358         if (!parse_options((char *) data, sb, &journal_devnum,
3359                            &journal_ioprio, 0))
3360                 goto failed_mount;
3361 
3362         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3363                 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3364                             "with data=journal disables delayed "
3365                             "allocation and O_DIRECT support!\n");
3366                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3367                         ext4_msg(sb, KERN_ERR, "can't mount with "
3368                                  "both data=journal and delalloc");
3369                         goto failed_mount;
3370                 }
3371                 if (test_opt(sb, DIOREAD_NOLOCK)) {
3372                         ext4_msg(sb, KERN_ERR, "can't mount with "
3373                                  "both data=journal and dioread_nolock");
3374                         goto failed_mount;
3375                 }
3376                 if (test_opt(sb, DAX)) {
3377                         ext4_msg(sb, KERN_ERR, "can't mount with "
3378                                  "both data=journal and dax");
3379                         goto failed_mount;
3380                 }
3381                 if (test_opt(sb, DELALLOC))
3382                         clear_opt(sb, DELALLOC);
3383         } else {
3384                 sb->s_iflags |= SB_I_CGROUPWB;
3385         }
3386 
3387         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3388                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3389 
3390         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3391             (ext4_has_compat_features(sb) ||
3392              ext4_has_ro_compat_features(sb) ||
3393              ext4_has_incompat_features(sb)))
3394                 ext4_msg(sb, KERN_WARNING,
3395                        "feature flags set on rev 0 fs, "
3396                        "running e2fsck is recommended");
3397 
3398         if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3399                 set_opt2(sb, HURD_COMPAT);
3400                 if (ext4_has_feature_64bit(sb)) {
3401                         ext4_msg(sb, KERN_ERR,
3402                                  "The Hurd can't support 64-bit file systems");
3403                         goto failed_mount;
3404                 }
3405         }
3406 
3407         if (IS_EXT2_SB(sb)) {
3408                 if (ext2_feature_set_ok(sb))
3409                         ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3410                                  "using the ext4 subsystem");
3411                 else {
3412                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3413                                  "to feature incompatibilities");
3414                         goto failed_mount;
3415                 }
3416         }
3417 
3418         if (IS_EXT3_SB(sb)) {
3419                 if (ext3_feature_set_ok(sb))
3420                         ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3421                                  "using the ext4 subsystem");
3422                 else {
3423                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3424                                  "to feature incompatibilities");
3425                         goto failed_mount;
3426                 }
3427         }
3428 
3429         /*
3430          * Check feature flags regardless of the revision level, since we
3431          * previously didn't change the revision level when setting the flags,
3432          * so there is a chance incompat flags are set on a rev 0 filesystem.
3433          */
3434         if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3435                 goto failed_mount;
3436 
3437         blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3438         if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3439             blocksize > EXT4_MAX_BLOCK_SIZE) {
3440                 ext4_msg(sb, KERN_ERR,
3441                        "Unsupported filesystem blocksize %d", blocksize);
3442                 goto failed_mount;
3443         }
3444 
3445         if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3446                 ext4_msg(sb, KERN_ERR,
3447                          "Number of reserved GDT blocks insanely large: %d",
3448                          le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3449                 goto failed_mount;
3450         }
3451 
3452         if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3453                 err = bdev_dax_supported(sb, blocksize);
3454                 if (err)
3455                         goto failed_mount;
3456         }
3457 
3458         if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3459                 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3460                          es->s_encryption_level);
3461                 goto failed_mount;
3462         }
3463 
3464         if (sb->s_blocksize != blocksize) {
3465                 /* Validate the filesystem blocksize */
3466                 if (!sb_set_blocksize(sb, blocksize)) {
3467                         ext4_msg(sb, KERN_ERR, "bad block size %d",
3468                                         blocksize);
3469                         goto failed_mount;
3470                 }
3471 
3472                 brelse(bh);
3473                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3474                 offset = do_div(logical_sb_block, blocksize);
3475                 bh = sb_bread_unmovable(sb, logical_sb_block);
3476                 if (!bh) {
3477                         ext4_msg(sb, KERN_ERR,
3478                                "Can't read superblock on 2nd try");
3479                         goto failed_mount;
3480                 }
3481                 es = (struct ext4_super_block *)(bh->b_data + offset);
3482                 sbi->s_es = es;
3483                 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3484                         ext4_msg(sb, KERN_ERR,
3485                                "Magic mismatch, very weird!");
3486                         goto failed_mount;
3487                 }
3488         }
3489 
3490         has_huge_files = ext4_has_feature_huge_file(sb);
3491         sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3492                                                       has_huge_files);
3493         sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3494 
3495         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3496                 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3497                 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3498         } else {
3499                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3500                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3501                 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3502                     (!is_power_of_2(sbi->s_inode_size)) ||
3503                     (sbi->s_inode_size > blocksize)) {
3504                         ext4_msg(sb, KERN_ERR,
3505                                "unsupported inode size: %d",
3506                                sbi->s_inode_size);
3507                         goto failed_mount;
3508                 }
3509                 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3510                         sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3511         }
3512 
3513         sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3514         if (ext4_has_feature_64bit(sb)) {
3515                 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3516                     sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3517                     !is_power_of_2(sbi->s_desc_size)) {
3518                         ext4_msg(sb, KERN_ERR,
3519                                "unsupported descriptor size %lu",
3520                                sbi->s_desc_size);
3521                         goto failed_mount;
3522                 }
3523         } else
3524                 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3525 
3526         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3527         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3528         if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3529                 goto cantfind_ext4;
3530 
3531         sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3532         if (sbi->s_inodes_per_block == 0)
3533                 goto cantfind_ext4;
3534         sbi->s_itb_per_group = sbi->s_inodes_per_group /
3535                                         sbi->s_inodes_per_block;
3536         sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3537         sbi->s_sbh = bh;
3538         sbi->s_mount_state = le16_to_cpu(es->s_state);
3539         sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3540         sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3541 
3542         for (i = 0; i < 4; i++)
3543                 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3544         sbi->s_def_hash_version = es->s_def_hash_version;
3545         if (ext4_has_feature_dir_index(sb)) {
3546                 i = le32_to_cpu(es->s_flags);
3547                 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3548                         sbi->s_hash_unsigned = 3;
3549                 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3550 #ifdef __CHAR_UNSIGNED__
3551                         if (!(sb->s_flags & MS_RDONLY))
3552                                 es->s_flags |=
3553                                         cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3554                         sbi->s_hash_unsigned = 3;
3555 #else
3556                         if (!(sb->s_flags & MS_RDONLY))
3557                                 es->s_flags |=
3558                                         cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3559 #endif
3560                 }
3561         }
3562 
3563         /* Handle clustersize */
3564         clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3565         has_bigalloc = ext4_has_feature_bigalloc(sb);
3566         if (has_bigalloc) {
3567                 if (clustersize < blocksize) {
3568                         ext4_msg(sb, KERN_ERR,
3569                                  "cluster size (%d) smaller than "
3570                                  "block size (%d)", clustersize, blocksize);
3571                         goto failed_mount;
3572                 }
3573                 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3574                         le32_to_cpu(es->s_log_block_size);
3575                 sbi->s_clusters_per_group =
3576                         le32_to_cpu(es->s_clusters_per_group);
3577                 if (sbi->s_clusters_per_group > blocksize * 8) {
3578                         ext4_msg(sb, KERN_ERR,
3579                                  "#clusters per group too big: %lu",
3580                                  sbi->s_clusters_per_group);
3581                         goto failed_mount;
3582                 }
3583                 if (sbi->s_blocks_per_group !=
3584                     (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3585                         ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3586                                  "clusters per group (%lu) inconsistent",
3587                                  sbi->s_blocks_per_group,
3588                                  sbi->s_clusters_per_group);
3589                         goto failed_mount;
3590                 }
3591         } else {
3592                 if (clustersize != blocksize) {
3593                         ext4_warning(sb, "fragment/cluster size (%d) != "
3594                                      "block size (%d)", clustersize,
3595                                      blocksize);
3596                         clustersize = blocksize;
3597                 }
3598                 if (sbi->s_blocks_per_group > blocksize * 8) {
3599                         ext4_msg(sb, KERN_ERR,
3600                                  "#blocks per group too big: %lu",
3601                                  sbi->s_blocks_per_group);
3602                         goto failed_mount;
3603                 }
3604                 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3605                 sbi->s_cluster_bits = 0;
3606         }
3607         sbi->s_cluster_ratio = clustersize / blocksize;
3608 
3609         if (sbi->s_inodes_per_group > blocksize * 8) {
3610                 ext4_msg(sb, KERN_ERR,
3611                        "#inodes per group too big: %lu",
3612                        sbi->s_inodes_per_group);
3613                 goto failed_mount;
3614         }
3615 
3616         /* Do we have standard group size of clustersize * 8 blocks ? */
3617         if (sbi->s_blocks_per_group == clustersize << 3)
3618                 set_opt2(sb, STD_GROUP_SIZE);
3619 
3620         /*
3621          * Test whether we have more sectors than will fit in sector_t,
3622          * and whether the max offset is addressable by the page cache.
3623          */
3624         err = generic_check_addressable(sb->s_blocksize_bits,
3625                                         ext4_blocks_count(es));
3626         if (err) {
3627                 ext4_msg(sb, KERN_ERR, "filesystem"
3628                          " too large to mount safely on this system");
3629                 if (sizeof(sector_t) < 8)
3630                         ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3631                 goto failed_mount;
3632         }
3633 
3634         if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3635                 goto cantfind_ext4;
3636 
3637         /* check blocks count against device size */
3638         blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3639         if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3640                 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3641                        "exceeds size of device (%llu blocks)",
3642                        ext4_blocks_count(es), blocks_count);
3643                 goto failed_mount;
3644         }
3645 
3646         /*
3647          * It makes no sense for the first data block to be beyond the end
3648          * of the filesystem.
3649          */
3650         if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3651                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3652                          "block %u is beyond end of filesystem (%llu)",
3653                          le32_to_cpu(es->s_first_data_block),
3654                          ext4_blocks_count(es));
3655                 goto failed_mount;
3656         }
3657         blocks_count = (ext4_blocks_count(es) -
3658                         le32_to_cpu(es->s_first_data_block) +
3659                         EXT4_BLOCKS_PER_GROUP(sb) - 1);
3660         do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3661         if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3662                 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3663                        "(block count %llu, first data block %u, "
3664                        "blocks per group %lu)", sbi->s_groups_count,
3665                        ext4_blocks_count(es),
3666                        le32_to_cpu(es->s_first_data_block),
3667                        EXT4_BLOCKS_PER_GROUP(sb));
3668                 goto failed_mount;
3669         }
3670         sbi->s_groups_count = blocks_count;
3671         sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3672                         (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3673         db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3674                    EXT4_DESC_PER_BLOCK(sb);
3675         sbi->s_group_desc = ext4_kvmalloc(db_count *
3676                                           sizeof(struct buffer_head *),
3677                                           GFP_KERNEL);
3678         if (sbi->s_group_desc == NULL) {
3679                 ext4_msg(sb, KERN_ERR, "not enough memory");
3680                 ret = -ENOMEM;
3681                 goto failed_mount;
3682         }
3683 
3684         bgl_lock_init(sbi->s_blockgroup_lock);
3685 
3686         for (i = 0; i < db_count; i++) {
3687                 block = descriptor_loc(sb, logical_sb_block, i);
3688                 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3689                 if (!sbi->s_group_desc[i]) {
3690                         ext4_msg(sb, KERN_ERR,
3691                                "can't read group descriptor %d", i);
3692                         db_count = i;
3693                         goto failed_mount2;
3694                 }
3695         }
3696         if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3697                 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3698                 ret = -EFSCORRUPTED;
3699                 goto failed_mount2;
3700         }
3701 
3702         sbi->s_gdb_count = db_count;
3703         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3704         spin_lock_init(&sbi->s_next_gen_lock);
3705 
3706         setup_timer(&sbi->s_err_report, print_daily_error_info,
3707                 (unsigned long) sb);
3708 
3709         /* Register extent status tree shrinker */
3710         if (ext4_es_register_shrinker(sbi))
3711                 goto failed_mount3;
3712 
3713         sbi->s_stripe = ext4_get_stripe_size(sbi);
3714         sbi->s_extent_max_zeroout_kb = 32;
3715 
3716         /*
3717          * set up enough so that it can read an inode
3718          */
3719         sb->s_op = &ext4_sops;
3720         sb->s_export_op = &ext4_export_ops;
3721         sb->s_xattr = ext4_xattr_handlers;
3722 #ifdef CONFIG_QUOTA
3723         sb->dq_op = &ext4_quota_operations;
3724         if (ext4_has_feature_quota(sb))
3725                 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3726         else
3727                 sb->s_qcop = &ext4_qctl_operations;
3728         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3729 #endif
3730         memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3731 
3732         INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3733         mutex_init(&sbi->s_orphan_lock);
3734 
3735         sb->s_root = NULL;
3736 
3737         needs_recovery = (es->s_last_orphan != 0 ||
3738                           ext4_has_feature_journal_needs_recovery(sb));
3739 
3740         if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3741                 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3742                         goto failed_mount3a;
3743 
3744         /*
3745          * The first inode we look at is the journal inode.  Don't try
3746          * root first: it may be modified in the journal!
3747          */
3748         if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3749                 if (ext4_load_journal(sb, es, journal_devnum))
3750                         goto failed_mount3a;
3751         } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3752                    ext4_has_feature_journal_needs_recovery(sb)) {
3753                 ext4_msg(sb, KERN_ERR, "required journal recovery "
3754                        "suppressed and not mounted read-only");
3755                 goto failed_mount_wq;
3756         } else {
3757                 /* Nojournal mode, all journal mount options are illegal */
3758                 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3759                         ext4_msg(sb, KERN_ERR, "can't mount with "
3760                                  "journal_checksum, fs mounted w/o journal");
3761                         goto failed_mount_wq;
3762                 }
3763                 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3764                         ext4_msg(sb, KERN_ERR, "can't mount with "
3765                                  "journal_async_commit, fs mounted w/o journal");
3766                         goto failed_mount_wq;
3767                 }
3768                 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3769                         ext4_msg(sb, KERN_ERR, "can't mount with "
3770                                  "commit=%lu, fs mounted w/o journal",
3771                                  sbi->s_commit_interval / HZ);
3772                         goto failed_mount_wq;
3773                 }
3774                 if (EXT4_MOUNT_DATA_FLAGS &
3775                     (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3776                         ext4_msg(sb, KERN_ERR, "can't mount with "
3777                                  "data=, fs mounted w/o journal");
3778                         goto failed_mount_wq;
3779                 }
3780                 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3781                 clear_opt(sb, JOURNAL_CHECKSUM);
3782                 clear_opt(sb, DATA_FLAGS);
3783                 sbi->s_journal = NULL;
3784                 needs_recovery = 0;
3785                 goto no_journal;
3786         }
3787 
3788         if (ext4_has_feature_64bit(sb) &&
3789             !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3790                                        JBD2_FEATURE_INCOMPAT_64BIT)) {
3791                 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3792                 goto failed_mount_wq;
3793         }
3794 
3795         if (!set_journal_csum_feature_set(sb)) {
3796                 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3797                          "feature set");
3798                 goto failed_mount_wq;
3799         }
3800 
3801         /* We have now updated the journal if required, so we can
3802          * validate the data journaling mode. */
3803         switch (test_opt(sb, DATA_FLAGS)) {
3804         case 0:
3805                 /* No mode set, assume a default based on the journal
3806                  * capabilities: ORDERED_DATA if the journal can
3807                  * cope, else JOURNAL_DATA
3808                  */
3809                 if (jbd2_journal_check_available_features
3810                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3811                         set_opt(sb, ORDERED_DATA);
3812                 else
3813                         set_opt(sb, JOURNAL_DATA);
3814                 break;
3815 
3816         case EXT4_MOUNT_ORDERED_DATA:
3817         case EXT4_MOUNT_WRITEBACK_DATA:
3818                 if (!jbd2_journal_check_available_features
3819                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3820                         ext4_msg(sb, KERN_ERR, "Journal does not support "
3821                                "requested data journaling mode");
3822                         goto failed_mount_wq;
3823                 }
3824         default:
3825                 break;
3826         }
3827         set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3828 
3829         sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3830 
3831 no_journal:
3832         sbi->s_mb_cache = ext4_xattr_create_cache();
3833         if (!sbi->s_mb_cache) {
3834                 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3835                 goto failed_mount_wq;
3836         }
3837 
3838         if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3839             (blocksize != PAGE_SIZE)) {
3840                 ext4_msg(sb, KERN_ERR,
3841                          "Unsupported blocksize for fs encryption");
3842                 goto failed_mount_wq;
3843         }
3844 
3845         if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3846             !ext4_has_feature_encrypt(sb)) {
3847                 ext4_set_feature_encrypt(sb);
3848                 ext4_commit_super(sb, 1);
3849         }
3850 
3851         /*
3852          * Get the # of file system overhead blocks from the
3853          * superblock if present.
3854          */
3855         if (es->s_overhead_clusters)
3856                 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3857         else {
3858                 err = ext4_calculate_overhead(sb);
3859                 if (err)
3860                         goto failed_mount_wq;
3861         }
3862 
3863         /*
3864          * The maximum number of concurrent works can be high and
3865          * concurrency isn't really necessary.  Limit it to 1.
3866          */
3867         EXT4_SB(sb)->rsv_conversion_wq =
3868                 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3869         if (!EXT4_SB(sb)->rsv_conversion_wq) {
3870                 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3871                 ret = -ENOMEM;
3872                 goto failed_mount4;
3873         }
3874 
3875         /*
3876          * The jbd2_journal_load will have done any necessary log recovery,
3877          * so we can safely mount the rest of the filesystem now.
3878          */
3879 
3880         root = ext4_iget(sb, EXT4_ROOT_INO);
3881         if (IS_ERR(root)) {
3882                 ext4_msg(sb, KERN_ERR, "get root inode failed");
3883                 ret = PTR_ERR(root);
3884                 root = NULL;
3885                 goto failed_mount4;
3886         }
3887         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3888                 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3889                 iput(root);
3890                 goto failed_mount4;
3891         }
3892         sb->s_root = d_make_root(root);
3893         if (!sb->s_root) {
3894                 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3895                 ret = -ENOMEM;
3896                 goto failed_mount4;
3897         }
3898 
3899         if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3900                 sb->s_flags |= MS_RDONLY;
3901 
3902         /* determine the minimum size of new large inodes, if present */
3903         if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3904                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3905                                                      EXT4_GOOD_OLD_INODE_SIZE;
3906                 if (ext4_has_feature_extra_isize(sb)) {
3907                         if (sbi->s_want_extra_isize <
3908                             le16_to_cpu(es->s_want_extra_isize))
3909                                 sbi->s_want_extra_isize =
3910                                         le16_to_cpu(es->s_want_extra_isize);
3911                         if (sbi->s_want_extra_isize <
3912                             le16_to_cpu(es->s_min_extra_isize))
3913                                 sbi->s_want_extra_isize =
3914                                         le16_to_cpu(es->s_min_extra_isize);
3915                 }
3916         }
3917         /* Check if enough inode space is available */
3918         if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3919                                                         sbi->s_inode_size) {
3920                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3921                                                        EXT4_GOOD_OLD_INODE_SIZE;
3922                 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3923                          "available");
3924         }
3925 
3926         ext4_set_resv_clusters(sb);
3927 
3928         err = ext4_setup_system_zone(sb);
3929         if (err) {
3930                 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3931                          "zone (%d)", err);
3932                 goto failed_mount4a;
3933         }
3934 
3935         ext4_ext_init(sb);
3936         err = ext4_mb_init(sb);
3937         if (err) {
3938                 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3939                          err);
3940                 goto failed_mount5;
3941         }
3942 
3943         block = ext4_count_free_clusters(sb);
3944         ext4_free_blocks_count_set(sbi->s_es, 
3945                                    EXT4_C2B(sbi, block));
3946         err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3947                                   GFP_KERNEL);
3948         if (!err) {
3949                 unsigned long freei = ext4_count_free_inodes(sb);
3950                 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3951                 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3952                                           GFP_KERNEL);
3953         }
3954         if (!err)
3955                 err = percpu_counter_init(&sbi->s_dirs_counter,
3956                                           ext4_count_dirs(sb), GFP_KERNEL);
3957         if (!err)
3958                 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3959                                           GFP_KERNEL);
3960         if (!err)
3961                 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
3962 
3963         if (err) {
3964                 ext4_msg(sb, KERN_ERR, "insufficient memory");
3965                 goto failed_mount6;
3966         }
3967 
3968         if (ext4_has_feature_flex_bg(sb))
3969                 if (!ext4_fill_flex_info(sb)) {
3970                         ext4_msg(sb, KERN_ERR,
3971                                "unable to initialize "
3972                                "flex_bg meta info!");
3973                         goto failed_mount6;
3974                 }
3975 
3976         err = ext4_register_li_request(sb, first_not_zeroed);
3977         if (err)
3978                 goto failed_mount6;
3979 
3980         err = ext4_register_sysfs(sb);
3981         if (err)
3982                 goto failed_mount7;
3983 
3984 #ifdef CONFIG_QUOTA
3985         /* Enable quota usage during mount. */
3986         if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3987                 err = ext4_enable_quotas(sb);
3988                 if (err)
3989                         goto failed_mount8;
3990         }
3991 #endif  /* CONFIG_QUOTA */
3992 
3993         EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3994         ext4_orphan_cleanup(sb, es);
3995         EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3996         if (needs_recovery) {
3997                 ext4_msg(sb, KERN_INFO, "recovery complete");
3998                 ext4_mark_recovery_complete(sb, es);
3999         }
4000         if (EXT4_SB(sb)->s_journal) {
4001                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4002                         descr = " journalled data mode";
4003                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4004                         descr = " ordered data mode";
4005                 else
4006                         descr = " writeback data mode";
4007         } else
4008                 descr = "out journal";
4009 
4010         if (test_opt(sb, DISCARD)) {
4011                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4012                 if (!blk_queue_discard(q))
4013                         ext4_msg(sb, KERN_WARNING,
4014                                  "mounting with \"discard\" option, but "
4015                                  "the device does not support discard");
4016         }
4017 
4018         if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4019                 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4020                          "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4021                          *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4022 
4023         if (es->s_error_count)
4024                 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4025 
4026         /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4027         ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4028         ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4029         ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4030 
4031         kfree(orig_data);
4032         return 0;
4033 
4034 cantfind_ext4:
4035         if (!silent)
4036                 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4037         goto failed_mount;
4038 
4039 #ifdef CONFIG_QUOTA
4040 failed_mount8:
4041         ext4_unregister_sysfs(sb);
4042 #endif
4043 failed_mount7:
4044         ext4_unregister_li_request(sb);
4045 failed_mount6:
4046         ext4_mb_release(sb);
4047         if (sbi->s_flex_groups)
4048                 kvfree(sbi->s_flex_groups);
4049         percpu_counter_destroy(&sbi->s_freeclusters_counter);
4050         percpu_counter_destroy(&sbi->s_freeinodes_counter);
4051         percpu_counter_destroy(&sbi->s_dirs_counter);
4052         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4053 failed_mount5:
4054         ext4_ext_release(sb);
4055         ext4_release_system_zone(sb);
4056 failed_mount4a:
4057         dput(sb->s_root);
4058         sb->s_root = NULL;
4059 failed_mount4:
4060         ext4_msg(sb, KERN_ERR, "mount failed");
4061         if (EXT4_SB(sb)->rsv_conversion_wq)
4062                 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4063 failed_mount_wq:
4064         if (sbi->s_mb_cache) {
4065                 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4066                 sbi->s_mb_cache = NULL;
4067         }
4068         if (sbi->s_journal) {
4069                 jbd2_journal_destroy(sbi->s_journal);
4070                 sbi->s_journal = NULL;
4071         }
4072 failed_mount3a:
4073         ext4_es_unregister_shrinker(sbi);
4074 failed_mount3:
4075         del_timer_sync(&sbi->s_err_report);
4076         if (sbi->s_mmp_tsk)
4077                 kthread_stop(sbi->s_mmp_tsk);
4078 failed_mount2:
4079         for (i = 0; i < db_count; i++)
4080                 brelse(sbi->s_group_desc[i]);
4081         kvfree(sbi->s_group_desc);
4082 failed_mount:
4083         if (sbi->s_chksum_driver)
4084                 crypto_free_shash(sbi->s_chksum_driver);
4085 #ifdef CONFIG_QUOTA
4086         for (i = 0; i < EXT4_MAXQUOTAS; i++)
4087                 kfree(sbi->s_qf_names[i]);
4088 #endif
4089         ext4_blkdev_remove(sbi);
4090         brelse(bh);
4091 out_fail:
4092         sb->s_fs_info = NULL;
4093         kfree(sbi->s_blockgroup_lock);
4094         kfree(sbi);
4095 out_free_orig:
4096         kfree(orig_data);
4097         return err ? err : ret;
4098 }
4099 
4100 /*
4101  * Setup any per-fs journal parameters now.  We'll do this both on
4102  * initial mount, once the journal has been initialised but before we've
4103  * done any recovery; and again on any subsequent remount.
4104  */
4105 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4106 {
4107         struct ext4_sb_info *sbi = EXT4_SB(sb);
4108 
4109         journal->j_commit_interval = sbi->s_commit_interval;
4110         journal->j_min_batch_time = sbi->s_min_batch_time;
4111         journal->j_max_batch_time = sbi->s_max_batch_time;
4112 
4113         write_lock(&journal->j_state_lock);
4114         if (test_opt(sb, BARRIER))
4115                 journal->j_flags |= JBD2_BARRIER;
4116         else
4117                 journal->j_flags &= ~JBD2_BARRIER;
4118         if (test_opt(sb, DATA_ERR_ABORT))
4119                 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4120         else
4121                 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4122         write_unlock(&journal->j_state_lock);
4123 }
4124 
4125 static journal_t *ext4_get_journal(struct super_block *sb,
4126                                    unsigned int journal_inum)
4127 {
4128         struct inode *journal_inode;
4129         journal_t *journal;
4130 
4131         BUG_ON(!ext4_has_feature_journal(sb));
4132 
4133         /* First, test for the existence of a valid inode on disk.  Bad
4134          * things happen if we iget() an unused inode, as the subsequent
4135          * iput() will try to delete it. */
4136 
4137         journal_inode = ext4_iget(sb, journal_inum);
4138         if (IS_ERR(journal_inode)) {
4139                 ext4_msg(sb, KERN_ERR, "no journal found");
4140                 return NULL;
4141         }
4142         if (!journal_inode->i_nlink) {
4143                 make_bad_inode(journal_inode);
4144                 iput(journal_inode);
4145                 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4146                 return NULL;
4147         }
4148 
4149         jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4150                   journal_inode, journal_inode->i_size);
4151         if (!S_ISREG(journal_inode->i_mode)) {
4152                 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4153                 iput(journal_inode);
4154                 return NULL;
4155         }
4156 
4157         journal = jbd2_journal_init_inode(journal_inode);
4158         if (!journal) {
4159                 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4160                 iput(journal_inode);
4161                 return NULL;
4162         }
4163         journal->j_private = sb;
4164         ext4_init_journal_params(sb, journal);
4165         return journal;
4166 }
4167 
4168 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4169                                        dev_t j_dev)
4170 {
4171         struct buffer_head *bh;
4172         journal_t *journal;
4173         ext4_fsblk_t start;
4174         ext4_fsblk_t len;
4175         int hblock, blocksize;
4176         ext4_fsblk_t sb_block;
4177         unsigned long offset;
4178         struct ext4_super_block *es;
4179         struct block_device *bdev;
4180 
4181         BUG_ON(!ext4_has_feature_journal(sb));
4182 
4183         bdev = ext4_blkdev_get(j_dev, sb);
4184         if (bdev == NULL)
4185                 return NULL;
4186 
4187         blocksize = sb->s_blocksize;
4188         hblock = bdev_logical_block_size(bdev);
4189         if (blocksize < hblock) {
4190                 ext4_msg(sb, KERN_ERR,
4191                         "blocksize too small for journal device");
4192                 goto out_bdev;
4193         }
4194 
4195         sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4196         offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4197         set_blocksize(bdev, blocksize);
4198         if (!(bh = __bread(bdev, sb_block, blocksize))) {
4199                 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4200                        "external journal");
4201                 goto out_bdev;
4202         }
4203 
4204         es = (struct ext4_super_block *) (bh->b_data + offset);
4205         if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4206             !(le32_to_cpu(es->s_feature_incompat) &
4207               EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4208                 ext4_msg(sb, KERN_ERR, "external journal has "
4209                                         "bad superblock");
4210                 brelse(bh);
4211                 goto out_bdev;
4212         }
4213 
4214         if ((le32_to_cpu(es->s_feature_ro_compat) &
4215              EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4216             es->s_checksum != ext4_superblock_csum(sb, es)) {
4217                 ext4_msg(sb, KERN_ERR, "external journal has "
4218                                        "corrupt superblock");
4219                 brelse(bh);
4220                 goto out_bdev;
4221         }
4222 
4223         if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4224                 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4225                 brelse(bh);
4226                 goto out_bdev;
4227         }
4228 
4229         len = ext4_blocks_count(es);
4230         start = sb_block + 1;
4231         brelse(bh);     /* we're done with the superblock */
4232 
4233         journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4234                                         start, len, blocksize);
4235         if (!journal) {
4236                 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4237                 goto out_bdev;
4238         }
4239         journal->j_private = sb;
4240         ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4241         wait_on_buffer(journal->j_sb_buffer);
4242         if (!buffer_uptodate(journal->j_sb_buffer)) {
4243                 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4244                 goto out_journal;
4245         }
4246         if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4247                 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4248                                         "user (unsupported) - %d",
4249                         be32_to_cpu(journal->j_superblock->s_nr_users));
4250                 goto out_journal;
4251         }
4252         EXT4_SB(sb)->journal_bdev = bdev;
4253         ext4_init_journal_params(sb, journal);
4254         return journal;
4255 
4256 out_journal:
4257         jbd2_journal_destroy(journal);
4258 out_bdev:
4259         ext4_blkdev_put(bdev);
4260         return NULL;
4261 }
4262 
4263 static int ext4_load_journal(struct super_block *sb,
4264                              struct ext4_super_block *es,
4265                              unsigned long journal_devnum)
4266 {
4267         journal_t *journal;
4268         unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4269         dev_t journal_dev;
4270         int err = 0;
4271         int really_read_only;
4272 
4273         BUG_ON(!ext4_has_feature_journal(sb));
4274 
4275         if (journal_devnum &&
4276             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4277                 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4278                         "numbers have changed");
4279                 journal_dev = new_decode_dev(journal_devnum);
4280         } else
4281                 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4282 
4283         really_read_only = bdev_read_only(sb->s_bdev);
4284 
4285         /*
4286          * Are we loading a blank journal or performing recovery after a
4287          * crash?  For recovery, we need to check in advance whether we
4288          * can get read-write access to the device.
4289          */
4290         if (ext4_has_feature_journal_needs_recovery(sb)) {
4291                 if (sb->s_flags & MS_RDONLY) {
4292                         ext4_msg(sb, KERN_INFO, "INFO: recovery "
4293                                         "required on readonly filesystem");
4294                         if (really_read_only) {
4295                                 ext4_msg(sb, KERN_ERR, "write access "
4296                                         "unavailable, cannot proceed");
4297                                 return -EROFS;
4298                         }
4299                         ext4_msg(sb, KERN_INFO, "write access will "
4300                                "be enabled during recovery");
4301                 }
4302         }
4303 
4304         if (journal_inum && journal_dev) {
4305                 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4306                        "and inode journals!");
4307                 return -EINVAL;
4308         }
4309 
4310         if (journal_inum) {
4311                 if (!(journal = ext4_get_journal(sb, journal_inum)))
4312                         return -EINVAL;
4313         } else {
4314                 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4315                         return -EINVAL;
4316         }
4317 
4318         if (!(journal->j_flags & JBD2_BARRIER))
4319                 ext4_msg(sb, KERN_INFO, "barriers disabled");
4320 
4321         if (!ext4_has_feature_journal_needs_recovery(sb))
4322                 err = jbd2_journal_wipe(journal, !really_read_only);
4323         if (!err) {
4324                 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4325                 if (save)
4326                         memcpy(save, ((char *) es) +
4327                                EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4328                 err = jbd2_journal_load(journal);
4329                 if (save)
4330                         memcpy(((char *) es) + EXT4_S_ERR_START,
4331                                save, EXT4_S_ERR_LEN);
4332                 kfree(save);
4333         }
4334 
4335         if (err) {
4336                 ext4_msg(sb, KERN_ERR, "error loading journal");
4337                 jbd2_journal_destroy(journal);
4338                 return err;
4339         }
4340 
4341         EXT4_SB(sb)->s_journal = journal;
4342         ext4_clear_journal_err(sb, es);
4343 
4344         if (!really_read_only && journal_devnum &&
4345             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4346                 es->s_journal_dev = cpu_to_le32(journal_devnum);
4347 
4348                 /* Make sure we flush the recovery flag to disk. */
4349                 ext4_commit_super(sb, 1);
4350         }
4351 
4352         return 0;
4353 }
4354 
4355 static int ext4_commit_super(struct super_block *sb, int sync)
4356 {
4357         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4358         struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4359         int error = 0;
4360 
4361         if (!sbh || block_device_ejected(sb))
4362                 return error;
4363         if (buffer_write_io_error(sbh)) {
4364                 /*
4365                  * Oh, dear.  A previous attempt to write the
4366                  * superblock failed.  This could happen because the
4367                  * USB device was yanked out.  Or it could happen to
4368                  * be a transient write error and maybe the block will
4369                  * be remapped.  Nothing we can do but to retry the
4370                  * write and hope for the best.
4371                  */
4372                 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4373                        "superblock detected");
4374                 clear_buffer_write_io_error(sbh);
4375                 set_buffer_uptodate(sbh);
4376         }
4377         /*
4378          * If the file system is mounted read-only, don't update the
4379          * superblock write time.  This avoids updating the superblock
4380          * write time when we are mounting the root file system
4381          * read/only but we need to replay the journal; at that point,
4382          * for people who are east of GMT and who make their clock
4383          * tick in localtime for Windows bug-for-bug compatibility,
4384          * the clock is set in the future, and this will cause e2fsck
4385          * to complain and force a full file system check.
4386          */
4387         if (!(sb->s_flags & MS_RDONLY))
4388                 es->s_wtime = cpu_to_le32(get_seconds());
4389         if (sb->s_bdev->bd_part)
4390                 es->s_kbytes_written =
4391                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4392                             ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4393                               EXT4_SB(sb)->s_sectors_written_start) >> 1));
4394         else
4395                 es->s_kbytes_written =
4396                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4397         if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4398                 ext4_free_blocks_count_set(es,
4399                         EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4400                                 &EXT4_SB(sb)->s_freeclusters_counter)));
4401         if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4402                 es->s_free_inodes_count =
4403                         cpu_to_le32(percpu_counter_sum_positive(
4404                                 &EXT4_SB(sb)->s_freeinodes_counter));
4405         BUFFER_TRACE(sbh, "marking dirty");
4406         ext4_superblock_csum_set(sb);
4407         mark_buffer_dirty(sbh);
4408         if (sync) {
4409                 error = __sync_dirty_buffer(sbh,
4410                         test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4411                 if (error)
4412                         return error;
4413 
4414                 error = buffer_write_io_error(sbh);
4415                 if (error) {
4416                         ext4_msg(sb, KERN_ERR, "I/O error while writing "
4417                                "superblock");
4418                         clear_buffer_write_io_error(sbh);
4419                         set_buffer_uptodate(sbh);
4420                 }
4421         }
4422         return error;
4423 }
4424 
4425 /*
4426  * Have we just finished recovery?  If so, and if we are mounting (or
4427  * remounting) the filesystem readonly, then we will end up with a
4428  * consistent fs on disk.  Record that fact.
4429  */
4430 static void ext4_mark_recovery_complete(struct super_block *sb,
4431                                         struct ext4_super_block *es)
4432 {
4433         journal_t *journal = EXT4_SB(sb)->s_journal;
4434 
4435         if (!ext4_has_feature_journal(sb)) {
4436                 BUG_ON(journal != NULL);
4437                 return;
4438         }
4439         jbd2_journal_lock_updates(journal);
4440         if (jbd2_journal_flush(journal) < 0)
4441                 goto out;
4442 
4443         if (ext4_has_feature_journal_needs_recovery(sb) &&
4444             sb->s_flags & MS_RDONLY) {
4445                 ext4_clear_feature_journal_needs_recovery(sb);
4446                 ext4_commit_super(sb, 1);
4447         }
4448 
4449 out:
4450         jbd2_journal_unlock_updates(journal);
4451 }
4452 
4453 /*
4454  * If we are mounting (or read-write remounting) a filesystem whose journal
4455  * has recorded an error from a previous lifetime, move that error to the
4456  * main filesystem now.
4457  */
4458 static void ext4_clear_journal_err(struct super_block *sb,
4459                                    struct ext4_super_block *es)
4460 {
4461         journal_t *journal;
4462         int j_errno;
4463         const char *errstr;
4464 
4465         BUG_ON(!ext4_has_feature_journal(sb));
4466 
4467         journal = EXT4_SB(sb)->s_journal;
4468 
4469         /*
4470          * Now check for any error status which may have been recorded in the
4471          * journal by a prior ext4_error() or ext4_abort()
4472          */
4473 
4474         j_errno = jbd2_journal_errno(journal);
4475         if (j_errno) {
4476                 char nbuf[16];
4477 
4478                 errstr = ext4_decode_error(sb, j_errno, nbuf);
4479                 ext4_warning(sb, "Filesystem error recorded "
4480                              "from previous mount: %s", errstr);
4481                 ext4_warning(sb, "Marking fs in need of filesystem check.");
4482 
4483                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4484                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4485                 ext4_commit_super(sb, 1);
4486 
4487                 jbd2_journal_clear_err(journal);
4488                 jbd2_journal_update_sb_errno(journal);
4489         }
4490 }
4491 
4492 /*
4493  * Force the running and committing transactions to commit,
4494  * and wait on the commit.
4495  */
4496 int ext4_force_commit(struct super_block *sb)
4497 {
4498         journal_t *journal;
4499 
4500         if (sb->s_flags & MS_RDONLY)
4501                 return 0;
4502 
4503         journal = EXT4_SB(sb)->s_journal;
4504         return ext4_journal_force_commit(journal);
4505 }
4506 
4507 static int ext4_sync_fs(struct super_block *sb, int wait)
4508 {
4509         int ret = 0;
4510         tid_t target;
4511         bool needs_barrier = false;
4512         struct ext4_sb_info *sbi = EXT4_SB(sb);
4513 
4514         trace_ext4_sync_fs(sb, wait);
4515         flush_workqueue(sbi->rsv_conversion_wq);
4516         /*
4517          * Writeback quota in non-journalled quota case - journalled quota has
4518          * no dirty dquots
4519          */
4520         dquot_writeback_dquots(sb, -1);
4521         /*
4522          * Data writeback is possible w/o journal transaction, so barrier must
4523          * being sent at the end of the function. But we can skip it if
4524          * transaction_commit will do it for us.
4525          */
4526         if (sbi->s_journal) {
4527                 target = jbd2_get_latest_transaction(sbi->s_journal);
4528                 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4529                     !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4530                         needs_barrier = true;
4531 
4532                 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4533                         if (wait)
4534                                 ret = jbd2_log_wait_commit(sbi->s_journal,
4535                                                            target);
4536                 }
4537         } else if (wait && test_opt(sb, BARRIER))
4538                 needs_barrier = true;
4539         if (needs_barrier) {
4540                 int err;
4541                 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4542                 if (!ret)
4543                         ret = err;
4544         }
4545 
4546         return ret;
4547 }
4548 
4549 /*
4550  * LVM calls this function before a (read-only) snapshot is created.  This
4551  * gives us a chance to flush the journal completely and mark the fs clean.
4552  *
4553  * Note that only this function cannot bring a filesystem to be in a clean
4554  * state independently. It relies on upper layer to stop all data & metadata
4555  * modifications.
4556  */
4557 static int ext4_freeze(struct super_block *sb)
4558 {
4559         int error = 0;
4560         journal_t *journal;
4561 
4562         if (sb->s_flags & MS_RDONLY)
4563                 return 0;
4564 
4565         journal = EXT4_SB(sb)->s_journal;
4566 
4567         if (journal) {
4568                 /* Now we set up the journal barrier. */
4569                 jbd2_journal_lock_updates(journal);
4570 
4571                 /*
4572                  * Don't clear the needs_recovery flag if we failed to
4573                  * flush the journal.
4574                  */
4575                 error = jbd2_journal_flush(journal);
4576                 if (error < 0)
4577                         goto out;
4578 
4579                 /* Journal blocked and flushed, clear needs_recovery flag. */
4580                 ext4_clear_feature_journal_needs_recovery(sb);
4581         }
4582 
4583         error = ext4_commit_super(sb, 1);
4584 out:
4585         if (journal)
4586                 /* we rely on upper layer to stop further updates */
4587                 jbd2_journal_unlock_updates(journal);
4588         return error;
4589 }
4590 
4591 /*
4592  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4593  * flag here, even though the filesystem is not technically dirty yet.
4594  */
4595 static int ext4_unfreeze(struct super_block *sb)
4596 {
4597         if (sb->s_flags & MS_RDONLY)
4598                 return 0;
4599 
4600         if (EXT4_SB(sb)->s_journal) {
4601                 /* Reset the needs_recovery flag before the fs is unlocked. */
4602                 ext4_set_feature_journal_needs_recovery(sb);
4603         }
4604 
4605         ext4_commit_super(sb, 1);
4606         return 0;
4607 }
4608 
4609 /*
4610  * Structure to save mount options for ext4_remount's benefit
4611  */
4612 struct ext4_mount_options {
4613         unsigned long s_mount_opt;
4614         unsigned long s_mount_opt2;
4615         kuid_t s_resuid;
4616         kgid_t s_resgid;
4617         unsigned long s_commit_interval;
4618         u32 s_min_batch_time, s_max_batch_time;
4619 #ifdef CONFIG_QUOTA
4620         int s_jquota_fmt;
4621         char *s_qf_names[EXT4_MAXQUOTAS];
4622 #endif
4623 };
4624 
4625 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4626 {
4627         struct ext4_super_block *es;
4628         struct ext4_sb_info *sbi = EXT4_SB(sb);
4629         unsigned long old_sb_flags;
4630         struct ext4_mount_options old_opts;
4631         int enable_quota = 0;
4632         ext4_group_t g;
4633         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4634         int err = 0;
4635 #ifdef CONFIG_QUOTA
4636         int i, j;
4637 #endif
4638         char *orig_data = kstrdup(data, GFP_KERNEL);
4639 
4640         /* Store the original options */
4641         old_sb_flags = sb->s_flags;
4642         old_opts.s_mount_opt = sbi->s_mount_opt;
4643         old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4644         old_opts.s_resuid = sbi->s_resuid;
4645         old_opts.s_resgid = sbi->s_resgid;
4646         old_opts.s_commit_interval = sbi->s_commit_interval;
4647         old_opts.s_min_batch_time = sbi->s_min_batch_time;
4648         old_opts.s_max_batch_time = sbi->s_max_batch_time;
4649 #ifdef CONFIG_QUOTA
4650         old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4651         for (i = 0; i < EXT4_MAXQUOTAS; i++)
4652                 if (sbi->s_qf_names[i]) {
4653                         old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4654                                                          GFP_KERNEL);
4655                         if (!old_opts.s_qf_names[i]) {
4656                                 for (j = 0; j < i; j++)
4657                                         kfree(old_opts.s_qf_names[j]);
4658                                 kfree(orig_data);
4659                                 return -ENOMEM;
4660                         }
4661                 } else
4662                         old_opts.s_qf_names[i] = NULL;
4663 #endif
4664         if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4665                 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4666 
4667         if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4668                 err = -EINVAL;
4669                 goto restore_opts;
4670         }
4671 
4672         if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4673             test_opt(sb, JOURNAL_CHECKSUM)) {
4674                 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4675                          "during remount not supported; ignoring");
4676                 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4677         }
4678 
4679         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4680                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4681                         ext4_msg(sb, KERN_ERR, "can't mount with "
4682                                  "both data=journal and delalloc");
4683                         err = -EINVAL;
4684                         goto restore_opts;
4685                 }
4686                 if (test_opt(sb, DIOREAD_NOLOCK)) {
4687                         ext4_msg(sb, KERN_ERR, "can't mount with "
4688                                  "both data=journal and dioread_nolock");
4689                         err = -EINVAL;
4690                         goto restore_opts;
4691                 }
4692                 if (test_opt(sb, DAX)) {
4693                         ext4_msg(sb, KERN_ERR, "can't mount with "
4694                                  "both data=journal and dax");
4695                         err = -EINVAL;
4696                         goto restore_opts;
4697                 }
4698         }
4699 
4700         if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4701                 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4702                         "dax flag with busy inodes while remounting");
4703                 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4704         }
4705 
4706         if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4707                 ext4_abort(sb, "Abort forced by user");
4708 
4709         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4710                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4711 
4712         es = sbi->s_es;
4713 
4714         if (sbi->s_journal) {
4715                 ext4_init_journal_params(sb, sbi->s_journal);
4716                 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4717         }
4718 
4719         if (*flags & MS_LAZYTIME)
4720                 sb->s_flags |= MS_LAZYTIME;
4721 
4722         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4723                 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4724                         err = -EROFS;
4725                         goto restore_opts;
4726                 }
4727 
4728                 if (*flags & MS_RDONLY) {
4729                         err = sync_filesystem(sb);
4730                         if (err < 0)
4731                                 goto restore_opts;
4732                         err = dquot_suspend(sb, -1);
4733                         if (err < 0)
4734                                 goto restore_opts;
4735 
4736                         /*
4737                          * First of all, the unconditional stuff we have to do
4738                          * to disable replay of the journal when we next remount
4739                          */
4740                         sb->s_flags |= MS_RDONLY;
4741 
4742                         /*
4743                          * OK, test if we are remounting a valid rw partition
4744                          * readonly, and if so set the rdonly flag and then
4745                          * mark the partition as valid again.
4746                          */
4747                         if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4748                             (sbi->s_mount_state & EXT4_VALID_FS))
4749                                 es->s_state = cpu_to_le16(sbi->s_mount_state);
4750 
4751                         if (sbi->s_journal)
4752                                 ext4_mark_recovery_complete(sb, es);
4753                 } else {
4754                         /* Make sure we can mount this feature set readwrite */
4755                         if (ext4_has_feature_readonly(sb) ||
4756                             !ext4_feature_set_ok(sb, 0)) {
4757                                 err = -EROFS;
4758                                 goto restore_opts;
4759                         }
4760                         /*
4761                          * Make sure the group descriptor checksums
4762                          * are sane.  If they aren't, refuse to remount r/w.
4763                          */
4764                         for (g = 0; g < sbi->s_groups_count; g++) {
4765                                 struct ext4_group_desc *gdp =
4766                                         ext4_get_group_desc(sb, g, NULL);
4767 
4768                                 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4769                                         ext4_msg(sb, KERN_ERR,
4770                "ext4_remount: Checksum for group %u failed (%u!=%u)",
4771                 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4772                                                le16_to_cpu(gdp->bg_checksum));
4773                                         err = -EFSBADCRC;
4774                                         goto restore_opts;
4775                                 }
4776                         }
4777 
4778                         /*
4779                          * If we have an unprocessed orphan list hanging
4780                          * around from a previously readonly bdev mount,
4781                          * require a full umount/remount for now.
4782                          */
4783                         if (es->s_last_orphan) {
4784                                 ext4_msg(sb, KERN_WARNING, "Couldn't "
4785                                        "remount RDWR because of unprocessed "
4786                                        "orphan inode list.  Please "
4787                                        "umount/remount instead");
4788                                 err = -EINVAL;
4789                                 goto restore_opts;
4790                         }
4791 
4792                         /*
4793                          * Mounting a RDONLY partition read-write, so reread
4794                          * and store the current valid flag.  (It may have
4795                          * been changed by e2fsck since we originally mounted
4796                          * the partition.)
4797                          */
4798                         if (sbi->s_journal)
4799                                 ext4_clear_journal_err(sb, es);
4800                         sbi->s_mount_state = le16_to_cpu(es->s_state);
4801                         if (!ext4_setup_super(sb, es, 0))
4802                                 sb->s_flags &= ~MS_RDONLY;
4803                         if (ext4_has_feature_mmp(sb))
4804                                 if (ext4_multi_mount_protect(sb,
4805                                                 le64_to_cpu(es->s_mmp_block))) {
4806                                         err = -EROFS;
4807                                         goto restore_opts;
4808                                 }
4809                         enable_quota = 1;
4810                 }
4811         }
4812 
4813         /*
4814          * Reinitialize lazy itable initialization thread based on
4815          * current settings
4816          */
4817         if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4818                 ext4_unregister_li_request(sb);
4819         else {
4820                 ext4_group_t first_not_zeroed;
4821                 first_not_zeroed = ext4_has_uninit_itable(sb);
4822                 ext4_register_li_request(sb, first_not_zeroed);
4823         }
4824 
4825         ext4_setup_system_zone(sb);
4826         if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4827                 ext4_commit_super(sb, 1);
4828 
4829 #ifdef CONFIG_QUOTA
4830         /* Release old quota file names */
4831         for (i = 0; i < EXT4_MAXQUOTAS; i++)
4832                 kfree(old_opts.s_qf_names[i]);
4833         if (enable_quota) {
4834                 if (sb_any_quota_suspended(sb))
4835                         dquot_resume(sb, -1);
4836                 else if (ext4_has_feature_quota(sb)) {
4837                         err = ext4_enable_quotas(sb);
4838                         if (err)
4839                                 goto restore_opts;
4840                 }
4841         }
4842 #endif
4843 
4844         *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4845         ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4846         kfree(orig_data);
4847         return 0;
4848 
4849 restore_opts:
4850         sb->s_flags = old_sb_flags;
4851         sbi->s_mount_opt = old_opts.s_mount_opt;
4852         sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4853         sbi->s_resuid = old_opts.s_resuid;
4854         sbi->s_resgid = old_opts.s_resgid;
4855         sbi->s_commit_interval = old_opts.s_commit_interval;
4856         sbi->s_min_batch_time = old_opts.s_min_batch_time;
4857         sbi->s_max_batch_time = old_opts.s_max_batch_time;
4858 #ifdef CONFIG_QUOTA
4859         sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4860         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4861                 kfree(sbi->s_qf_names[i]);
4862                 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4863         }
4864 #endif
4865         kfree(orig_data);
4866         return err;
4867 }
4868 
4869 #ifdef CONFIG_QUOTA
4870 static int ext4_statfs_project(struct super_block *sb,
4871                                kprojid_t projid, struct kstatfs *buf)
4872 {
4873         struct kqid qid;
4874         struct dquot *dquot;
4875         u64 limit;
4876         u64 curblock;
4877 
4878         qid = make_kqid_projid(projid);
4879         dquot = dqget(sb, qid);
4880         if (IS_ERR(dquot))
4881                 return PTR_ERR(dquot);
4882         spin_lock(&dq_data_lock);
4883 
4884         limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4885                  dquot->dq_dqb.dqb_bsoftlimit :
4886                  dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4887         if (limit && buf->f_blocks > limit) {
4888                 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4889                 buf->f_blocks = limit;
4890                 buf->f_bfree = buf->f_bavail =
4891                         (buf->f_blocks > curblock) ?
4892                          (buf->f_blocks - curblock) : 0;
4893         }
4894 
4895         limit = dquot->dq_dqb.dqb_isoftlimit ?
4896                 dquot->dq_dqb.dqb_isoftlimit :
4897                 dquot->dq_dqb.dqb_ihardlimit;
4898         if (limit && buf->f_files > limit) {
4899                 buf->f_files = limit;
4900                 buf->f_ffree =
4901                         (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4902                          (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4903         }
4904 
4905         spin_unlock(&dq_data_lock);
4906         dqput(dquot);
4907         return 0;
4908 }
4909 #endif
4910 
4911 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4912 {
4913         struct super_block *sb = dentry->d_sb;
4914         struct ext4_sb_info *sbi = EXT4_SB(sb);
4915         struct ext4_super_block *es = sbi->s_es;
4916         ext4_fsblk_t overhead = 0, resv_blocks;
4917         u64 fsid;
4918         s64 bfree;
4919         resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4920 
4921         if (!test_opt(sb, MINIX_DF))
4922                 overhead = sbi->s_overhead;
4923 
4924         buf->f_type = EXT4_SUPER_MAGIC;
4925         buf->f_bsize = sb->s_blocksize;
4926         buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4927         bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4928                 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4929         /* prevent underflow in case that few free space is available */
4930         buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4931         buf->f_bavail = buf->f_bfree -
4932                         (ext4_r_blocks_count(es) + resv_blocks);
4933         if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4934                 buf->f_bavail = 0;
4935         buf->f_files = le32_to_cpu(es->s_inodes_count);
4936         buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4937         buf->f_namelen = EXT4_NAME_LEN;
4938         fsid = le64_to_cpup((void *)es->s_uuid) ^
4939                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4940         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4941         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4942 
4943 #ifdef CONFIG_QUOTA
4944         if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
4945             sb_has_quota_limits_enabled(sb, PRJQUOTA))
4946                 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
4947 #endif
4948         return 0;
4949 }
4950 
4951 /* Helper function for writing quotas on sync - we need to start transaction
4952  * before quota file is locked for write. Otherwise the are possible deadlocks:
4953  * Process 1                         Process 2
4954  * ext4_create()                     quota_sync()
4955  *   jbd2_journal_start()                  write_dquot()
4956  *   dquot_initialize()                         down(dqio_mutex)
4957  *     down(dqio_mutex)                    jbd2_journal_start()
4958  *
4959  */
4960 
4961 #ifdef CONFIG_QUOTA
4962 
4963 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4964 {
4965         return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4966 }
4967 
4968 static int ext4_write_dquot(struct dquot *dquot)
4969 {
4970         int ret, err;
4971         handle_t *handle;
4972         struct inode *inode;
4973 
4974         inode = dquot_to_inode(dquot);
4975         handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4976                                     EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4977         if (IS_ERR(handle))
4978                 return PTR_ERR(handle);
4979         ret = dquot_commit(dquot);
4980         err = ext4_journal_stop(handle);
4981         if (!ret)
4982                 ret = err;
4983         return ret;
4984 }
4985 
4986 static int ext4_acquire_dquot(struct dquot *dquot)
4987 {
4988         int ret, err;
4989         handle_t *handle;
4990 
4991         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4992                                     EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4993         if (IS_ERR(handle))
4994                 return PTR_ERR(handle);
4995         ret = dquot_acquire(dquot);
4996         err = ext4_journal_stop(handle);
4997         if (!ret)
4998                 ret = err;
4999         return ret;
5000 }
5001 
5002 static int ext4_release_dquot(struct dquot *dquot)
5003 {
5004         int ret, err;
5005         handle_t *handle;
5006 
5007         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5008                                     EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5009         if (IS_ERR(handle)) {
5010                 /* Release dquot anyway to avoid endless cycle in dqput() */
5011                 dquot_release(dquot);
5012                 return PTR_ERR(handle);
5013         }
5014         ret = dquot_release(dquot);
5015         err = ext4_journal_stop(handle);
5016         if (!ret)
5017                 ret = err;
5018         return ret;
5019 }
5020 
5021 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5022 {
5023         struct super_block *sb = dquot->dq_sb;
5024         struct ext4_sb_info *sbi = EXT4_SB(sb);
5025 
5026         /* Are we journaling quotas? */
5027         if (ext4_has_feature_quota(sb) ||
5028             sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5029                 dquot_mark_dquot_dirty(dquot);
5030                 return ext4_write_dquot(dquot);
5031         } else {
5032                 return dquot_mark_dquot_dirty(dquot);
5033         }
5034 }
5035 
5036 static int ext4_write_info(struct super_block *sb, int type)
5037 {
5038         int ret, err;
5039         handle_t *handle;
5040 
5041         /* Data block + inode block */
5042         handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5043         if (IS_ERR(handle))
5044                 return PTR_ERR(handle);
5045         ret = dquot_commit_info(sb, type);
5046         err = ext4_journal_stop(handle);
5047         if (!ret)
5048                 ret = err;
5049         return ret;
5050 }
5051 
5052 /*
5053  * Turn on quotas during mount time - we need to find
5054  * the quota file and such...
5055  */
5056 static int ext4_quota_on_mount(struct super_block *sb, int type)
5057 {
5058         return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5059                                         EXT4_SB(sb)->s_jquota_fmt, type);
5060 }
5061 
5062 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5063 {
5064         struct ext4_inode_info *ei = EXT4_I(inode);
5065 
5066         /* The first argument of lockdep_set_subclass has to be
5067          * *exactly* the same as the argument to init_rwsem() --- in
5068          * this case, in init_once() --- or lockdep gets unhappy
5069          * because the name of the lock is set using the
5070          * stringification of the argument to init_rwsem().
5071          */
5072         (void) ei;      /* shut up clang warning if !CONFIG_LOCKDEP */
5073         lockdep_set_subclass(&ei->i_data_sem, subclass);
5074 }
5075 
5076 /*
5077  * Standard function to be called on quota_on
5078  */
5079 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5080                          struct path *path)
5081 {
5082         int err;
5083 
5084         if (!test_opt(sb, QUOTA))
5085                 return -EINVAL;
5086 
5087         /* Quotafile not on the same filesystem? */
5088         if (path->dentry->d_sb != sb)
5089                 return -EXDEV;
5090         /* Journaling quota? */
5091         if (EXT4_SB(sb)->s_qf_names[type]) {
5092                 /* Quotafile not in fs root? */
5093                 if (path->dentry->d_parent != sb->s_root)
5094                         ext4_msg(sb, KERN_WARNING,
5095                                 "Quota file not on filesystem root. "
5096                                 "Journaled quota will not work");
5097         }
5098 
5099         /*
5100          * When we journal data on quota file, we have to flush journal to see
5101          * all updates to the file when we bypass pagecache...
5102          */
5103         if (EXT4_SB(sb)->s_journal &&
5104             ext4_should_journal_data(d_inode(path->dentry))) {
5105                 /*
5106                  * We don't need to lock updates but journal_flush() could
5107                  * otherwise be livelocked...
5108                  */
5109                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5110                 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5111                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5112                 if (err)
5113                         return err;
5114         }
5115         lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5116         err = dquot_quota_on(sb, type, format_id, path);
5117         if (err)
5118                 lockdep_set_quota_inode(path->dentry->d_inode,
5119                                              I_DATA_SEM_NORMAL);
5120         return err;
5121 }
5122 
5123 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5124                              unsigned int flags)
5125 {
5126         int err;
5127         struct inode *qf_inode;
5128         unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5129                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5130                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5131                 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5132         };
5133 
5134         BUG_ON(!ext4_has_feature_quota(sb));
5135 
5136         if (!qf_inums[type])
5137                 return -EPERM;
5138 
5139         qf_inode = ext4_iget(sb, qf_inums[type]);
5140         if (IS_ERR(qf_inode)) {
5141                 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5142                 return PTR_ERR(qf_inode);
5143         }
5144 
5145         /* Don't account quota for quota files to avoid recursion */
5146         qf_inode->i_flags |= S_NOQUOTA;
5147         lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5148         err = dquot_enable(qf_inode, type, format_id, flags);
5149         iput(qf_inode);
5150         if (err)
5151                 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5152 
5153         return err;
5154 }
5155 
5156 /* Enable usage tracking for all quota types. */
5157 static int ext4_enable_quotas(struct super_block *sb)
5158 {
5159         int type, err = 0;
5160         unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5161                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5162                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5163                 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5164         };
5165 
5166         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5167         for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5168                 if (qf_inums[type]) {
5169                         err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5170                                                 DQUOT_USAGE_ENABLED);
5171                         if (err) {
5172                                 ext4_warning(sb,
5173                                         "Failed to enable quota tracking "
5174                                         "(type=%d, err=%d). Please run "
5175                                         "e2fsck to fix.", type, err);
5176                                 return err;
5177                         }
5178                 }
5179         }
5180         return 0;
5181 }
5182 
5183 static int ext4_quota_off(struct super_block *sb, int type)
5184 {
5185         struct inode *inode = sb_dqopt(sb)->files[type];
5186         handle_t *handle;
5187 
5188         /* Force all delayed allocation blocks to be allocated.
5189          * Caller already holds s_umount sem */
5190         if (test_opt(sb, DELALLOC))
5191                 sync_filesystem(sb);
5192 
5193         if (!inode)
5194                 goto out;
5195 
5196         /* Update modification times of quota files when userspace can
5197          * start looking at them */
5198         handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5199         if (IS_ERR(handle))
5200                 goto out;
5201         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5202         ext4_mark_inode_dirty(handle, inode);
5203         ext4_journal_stop(handle);
5204 
5205 out:
5206         return dquot_quota_off(sb, type);
5207 }
5208 
5209 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5210  * acquiring the locks... As quota files are never truncated and quota code
5211  * itself serializes the operations (and no one else should touch the files)
5212  * we don't have to be afraid of races */
5213 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5214                                size_t len, loff_t off)
5215 {
5216         struct inode *inode = sb_dqopt(sb)->files[type];
5217         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5218         int offset = off & (sb->s_blocksize - 1);
5219         int tocopy;
5220         size_t toread;
5221         struct buffer_head *bh;
5222         loff_t i_size = i_size_read(inode);
5223 
5224         if (off > i_size)
5225                 return 0;
5226         if (off+len > i_size)
5227                 len = i_size-off;
5228         toread = len;
5229         while (toread > 0) {
5230                 tocopy = sb->s_blocksize - offset < toread ?
5231                                 sb->s_blocksize - offset : toread;
5232                 bh = ext4_bread(NULL, inode, blk, 0);
5233                 if (IS_ERR(bh))
5234                         return PTR_ERR(bh);
5235                 if (!bh)        /* A hole? */
5236                         memset(data, 0, tocopy);
5237                 else
5238                         memcpy(data, bh->b_data+offset, tocopy);
5239                 brelse(bh);
5240                 offset = 0;
5241                 toread -= tocopy;
5242                 data += tocopy;
5243                 blk++;
5244         }
5245         return len;
5246 }
5247 
5248 /* Write to quotafile (we know the transaction is already started and has
5249  * enough credits) */
5250 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5251                                 const char *data, size_t len, loff_t off)
5252 {
5253         struct inode *inode = sb_dqopt(sb)->files[type];
5254         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5255         int err, offset = off & (sb->s_blocksize - 1);
5256         int retries = 0;
5257         struct buffer_head *bh;
5258         handle_t *handle = journal_current_handle();
5259 
5260         if (EXT4_SB(sb)->s_journal && !handle) {
5261                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5262                         " cancelled because transaction is not started",
5263                         (unsigned long long)off, (unsigned long long)len);
5264                 return -EIO;
5265         }
5266         /*
5267          * Since we account only one data block in transaction credits,
5268          * then it is impossible to cross a block boundary.
5269          */
5270         if (sb->s_blocksize - offset < len) {
5271                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5272                         " cancelled because not block aligned",
5273                         (unsigned long long)off, (unsigned long long)len);
5274                 return -EIO;
5275         }
5276 
5277         do {
5278                 bh = ext4_bread(handle, inode, blk,
5279                                 EXT4_GET_BLOCKS_CREATE |
5280                                 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5281         } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5282                  ext4_should_retry_alloc(inode->i_sb, &retries));
5283         if (IS_ERR(bh))
5284                 return PTR_ERR(bh);
5285         if (!bh)
5286                 goto out;
5287         BUFFER_TRACE(bh, "get write access");
5288         err = ext4_journal_get_write_access(handle, bh);
5289         if (err) {
5290                 brelse(bh);
5291                 return err;
5292         }
5293         lock_buffer(bh);
5294         memcpy(bh->b_data+offset, data, len);
5295         flush_dcache_page(bh->b_page);
5296         unlock_buffer(bh);
5297         err = ext4_handle_dirty_metadata(handle, NULL, bh);
5298         brelse(bh);
5299 out:
5300         if (inode->i_size < off + len) {
5301                 i_size_write(inode, off + len);
5302                 EXT4_I(inode)->i_disksize = inode->i_size;
5303                 ext4_mark_inode_dirty(handle, inode);
5304         }
5305         return len;
5306 }
5307 
5308 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5309 {
5310         const struct quota_format_ops   *ops;
5311 
5312         if (!sb_has_quota_loaded(sb, qid->type))
5313                 return -ESRCH;
5314         ops = sb_dqopt(sb)->ops[qid->type];
5315         if (!ops || !ops->get_next_id)
5316                 return -ENOSYS;
5317         return dquot_get_next_id(sb, qid);
5318 }
5319 #endif
5320 
5321 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5322                        const char *dev_name, void *data)
5323 {
5324         return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5325 }
5326 
5327 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5328 static inline void register_as_ext2(void)
5329 {
5330         int err = register_filesystem(&ext2_fs_type);
5331         if (err)
5332                 printk(KERN_WARNING
5333                        "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5334 }
5335 
5336 static inline void unregister_as_ext2(void)
5337 {
5338         unregister_filesystem(&ext2_fs_type);
5339 }
5340 
5341 static inline int ext2_feature_set_ok(struct super_block *sb)
5342 {
5343         if (ext4_has_unknown_ext2_incompat_features(sb))
5344                 return 0;
5345         if (sb->s_flags & MS_RDONLY)
5346                 return 1;
5347         if (ext4_has_unknown_ext2_ro_compat_features(sb))
5348                 return 0;
5349         return 1;
5350 }
5351 #else
5352 static inline void register_as_ext2(void) { }
5353 static inline void unregister_as_ext2(void) { }
5354 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5355 #endif
5356 
5357 static inline void register_as_ext3(void)
5358 {
5359         int err = register_filesystem(&ext3_fs_type);
5360         if (err)
5361                 printk(KERN_WARNING
5362                        "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5363 }
5364 
5365 static inline void unregister_as_ext3(void)
5366 {
5367         unregister_filesystem(&ext3_fs_type);
5368 }
5369 
5370 static inline int ext3_feature_set_ok(struct super_block *sb)
5371 {
5372         if (ext4_has_unknown_ext3_incompat_features(sb))
5373                 return 0;
5374         if (!ext4_has_feature_journal(sb))
5375                 return 0;
5376         if (sb->s_flags & MS_RDONLY)
5377                 return 1;
5378         if (ext4_has_unknown_ext3_ro_compat_features(sb))
5379                 return 0;
5380         return 1;
5381 }
5382 
5383 static struct file_system_type ext4_fs_type = {
5384         .owner          = THIS_MODULE,
5385         .name           = "ext4",
5386         .mount          = ext4_mount,
5387         .kill_sb        = kill_block_super,
5388         .fs_flags       = FS_REQUIRES_DEV,
5389 };
5390 MODULE_ALIAS_FS("ext4");
5391 
5392 /* Shared across all ext4 file systems */
5393 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5394 
5395 static int __init ext4_init_fs(void)
5396 {
5397         int i, err;
5398 
5399         ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5400         ext4_li_info = NULL;
5401         mutex_init(&ext4_li_mtx);
5402 
5403         /* Build-time check for flags consistency */
5404         ext4_check_flag_values();
5405 
5406         for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5407                 init_waitqueue_head(&ext4__ioend_wq[i]);
5408 
5409         err = ext4_init_es();
5410         if (err)
5411                 return err;
5412 
5413         err = ext4_init_pageio();
5414         if (err)
5415                 goto out5;
5416 
5417         err = ext4_init_system_zone();
5418         if (err)
5419                 goto out4;
5420 
5421         err = ext4_init_sysfs();
5422         if (err)
5423                 goto out3;
5424 
5425         err = ext4_init_mballoc();
5426         if (err)
5427                 goto out2;
5428         err = init_inodecache();
5429         if (err)
5430                 goto out1;
5431         register_as_ext3();
5432         register_as_ext2();
5433         err = register_filesystem(&ext4_fs_type);
5434         if (err)
5435                 goto out;
5436 
5437         return 0;
5438 out:
5439         unregister_as_ext2();
5440         unregister_as_ext3();
5441         destroy_inodecache();
5442 out1:
5443         ext4_exit_mballoc();
5444 out2:
5445         ext4_exit_sysfs();
5446 out3:
5447         ext4_exit_system_zone();
5448 out4:
5449         ext4_exit_pageio();
5450 out5:
5451         ext4_exit_es();
5452 
5453         return err;
5454 }
5455 
5456 static void __exit ext4_exit_fs(void)
5457 {
5458         ext4_exit_crypto();
5459         ext4_destroy_lazyinit_thread();
5460         unregister_as_ext2();
5461         unregister_as_ext3();
5462         unregister_filesystem(&ext4_fs_type);
5463         destroy_inodecache();
5464         ext4_exit_mballoc();
5465         ext4_exit_sysfs();
5466         ext4_exit_system_zone();
5467         ext4_exit_pageio();
5468         ext4_exit_es();
5469 }
5470 
5471 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5472 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5473 MODULE_LICENSE("GPL");
5474 module_init(ext4_init_fs)
5475 module_exit(ext4_exit_fs)
5476 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp