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

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

Version: ~ [ linux-5.10-rc1 ] ~ [ linux-5.9.1 ] ~ [ linux-5.8.16 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.72 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.152 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.202 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.240 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.240 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ 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/jbd2.h>
 25 #include <linux/slab.h>
 26 #include <linux/init.h>
 27 #include <linux/blkdev.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/proc_fs.h>
 38 #include <linux/ctype.h>
 39 #include <linux/log2.h>
 40 #include <linux/crc16.h>
 41 #include <linux/cleancache.h>
 42 #include <asm/uaccess.h>
 43 
 44 #include <linux/kthread.h>
 45 #include <linux/freezer.h>
 46 
 47 #include "ext4.h"
 48 #include "ext4_extents.h"       /* Needed for trace points definition */
 49 #include "ext4_jbd2.h"
 50 #include "xattr.h"
 51 #include "acl.h"
 52 #include "mballoc.h"
 53 
 54 #define CREATE_TRACE_POINTS
 55 #include <trace/events/ext4.h>
 56 
 57 static struct proc_dir_entry *ext4_proc_root;
 58 static struct kset *ext4_kset;
 59 static struct ext4_lazy_init *ext4_li_info;
 60 static struct mutex ext4_li_mtx;
 61 static struct ext4_features *ext4_feat;
 62 
 63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
 64                              unsigned long journal_devnum);
 65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
 66 static int ext4_commit_super(struct super_block *sb, int sync);
 67 static void ext4_mark_recovery_complete(struct super_block *sb,
 68                                         struct ext4_super_block *es);
 69 static void ext4_clear_journal_err(struct super_block *sb,
 70                                    struct ext4_super_block *es);
 71 static int ext4_sync_fs(struct super_block *sb, int wait);
 72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
 73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
 74 static int ext4_unfreeze(struct super_block *sb);
 75 static int ext4_freeze(struct super_block *sb);
 76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
 77                        const char *dev_name, void *data);
 78 static inline int ext2_feature_set_ok(struct super_block *sb);
 79 static inline int ext3_feature_set_ok(struct super_block *sb);
 80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
 81 static void ext4_destroy_lazyinit_thread(void);
 82 static void ext4_unregister_li_request(struct super_block *sb);
 83 static void ext4_clear_request_list(void);
 84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
 85 
 86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
 87 static struct file_system_type ext2_fs_type = {
 88         .owner          = THIS_MODULE,
 89         .name           = "ext2",
 90         .mount          = ext4_mount,
 91         .kill_sb        = kill_block_super,
 92         .fs_flags       = FS_REQUIRES_DEV,
 93 };
 94 MODULE_ALIAS_FS("ext2");
 95 MODULE_ALIAS("ext2");
 96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
 97 #else
 98 #define IS_EXT2_SB(sb) (0)
 99 #endif
100 
101 
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104         .owner          = THIS_MODULE,
105         .name           = "ext3",
106         .mount          = ext4_mount,
107         .kill_sb        = kill_block_super,
108         .fs_flags       = FS_REQUIRES_DEV,
109 };
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113 #else
114 #define IS_EXT3_SB(sb) (0)
115 #endif
116 
117 static int ext4_verify_csum_type(struct super_block *sb,
118                                  struct ext4_super_block *es)
119 {
120         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121                                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122                 return 1;
123 
124         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125 }
126 
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128                                    struct ext4_super_block *es)
129 {
130         struct ext4_sb_info *sbi = EXT4_SB(sb);
131         int offset = offsetof(struct ext4_super_block, s_checksum);
132         __u32 csum;
133 
134         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135 
136         return cpu_to_le32(csum);
137 }
138 
139 int ext4_superblock_csum_verify(struct super_block *sb,
140                                 struct ext4_super_block *es)
141 {
142         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
144                 return 1;
145 
146         return es->s_checksum == ext4_superblock_csum(sb, es);
147 }
148 
149 void ext4_superblock_csum_set(struct super_block *sb)
150 {
151         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152 
153         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154                 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
155                 return;
156 
157         es->s_checksum = ext4_superblock_csum(sb, es);
158 }
159 
160 void *ext4_kvmalloc(size_t size, gfp_t flags)
161 {
162         void *ret;
163 
164         ret = kmalloc(size, flags);
165         if (!ret)
166                 ret = __vmalloc(size, flags, PAGE_KERNEL);
167         return ret;
168 }
169 
170 void *ext4_kvzalloc(size_t size, gfp_t flags)
171 {
172         void *ret;
173 
174         ret = kzalloc(size, flags);
175         if (!ret)
176                 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177         return ret;
178 }
179 
180 void ext4_kvfree(void *ptr)
181 {
182         if (is_vmalloc_addr(ptr))
183                 vfree(ptr);
184         else
185                 kfree(ptr);
186 
187 }
188 
189 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190                                struct ext4_group_desc *bg)
191 {
192         return le32_to_cpu(bg->bg_block_bitmap_lo) |
193                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
194                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
195 }
196 
197 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198                                struct ext4_group_desc *bg)
199 {
200         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
201                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
202                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
203 }
204 
205 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206                               struct ext4_group_desc *bg)
207 {
208         return le32_to_cpu(bg->bg_inode_table_lo) |
209                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
211 }
212 
213 __u32 ext4_free_group_clusters(struct super_block *sb,
214                                struct ext4_group_desc *bg)
215 {
216         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
219 }
220 
221 __u32 ext4_free_inodes_count(struct super_block *sb,
222                               struct ext4_group_desc *bg)
223 {
224         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
227 }
228 
229 __u32 ext4_used_dirs_count(struct super_block *sb,
230                               struct ext4_group_desc *bg)
231 {
232         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
235 }
236 
237 __u32 ext4_itable_unused_count(struct super_block *sb,
238                               struct ext4_group_desc *bg)
239 {
240         return le16_to_cpu(bg->bg_itable_unused_lo) |
241                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
243 }
244 
245 void ext4_block_bitmap_set(struct super_block *sb,
246                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
247 {
248         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
249         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
251 }
252 
253 void ext4_inode_bitmap_set(struct super_block *sb,
254                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
255 {
256         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
257         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
259 }
260 
261 void ext4_inode_table_set(struct super_block *sb,
262                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 {
264         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
265         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
267 }
268 
269 void ext4_free_group_clusters_set(struct super_block *sb,
270                                   struct ext4_group_desc *bg, __u32 count)
271 {
272         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
275 }
276 
277 void ext4_free_inodes_set(struct super_block *sb,
278                           struct ext4_group_desc *bg, __u32 count)
279 {
280         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
283 }
284 
285 void ext4_used_dirs_set(struct super_block *sb,
286                           struct ext4_group_desc *bg, __u32 count)
287 {
288         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
291 }
292 
293 void ext4_itable_unused_set(struct super_block *sb,
294                           struct ext4_group_desc *bg, __u32 count)
295 {
296         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299 }
300 
301 
302 static void __save_error_info(struct super_block *sb, const char *func,
303                             unsigned int line)
304 {
305         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306 
307         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309         es->s_last_error_time = cpu_to_le32(get_seconds());
310         strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311         es->s_last_error_line = cpu_to_le32(line);
312         if (!es->s_first_error_time) {
313                 es->s_first_error_time = es->s_last_error_time;
314                 strncpy(es->s_first_error_func, func,
315                         sizeof(es->s_first_error_func));
316                 es->s_first_error_line = cpu_to_le32(line);
317                 es->s_first_error_ino = es->s_last_error_ino;
318                 es->s_first_error_block = es->s_last_error_block;
319         }
320         /*
321          * Start the daily error reporting function if it hasn't been
322          * started already
323          */
324         if (!es->s_error_count)
325                 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
326         le32_add_cpu(&es->s_error_count, 1);
327 }
328 
329 static void save_error_info(struct super_block *sb, const char *func,
330                             unsigned int line)
331 {
332         __save_error_info(sb, func, line);
333         ext4_commit_super(sb, 1);
334 }
335 
336 /*
337  * The del_gendisk() function uninitializes the disk-specific data
338  * structures, including the bdi structure, without telling anyone
339  * else.  Once this happens, any attempt to call mark_buffer_dirty()
340  * (for example, by ext4_commit_super), will cause a kernel OOPS.
341  * This is a kludge to prevent these oops until we can put in a proper
342  * hook in del_gendisk() to inform the VFS and file system layers.
343  */
344 static int block_device_ejected(struct super_block *sb)
345 {
346         struct inode *bd_inode = sb->s_bdev->bd_inode;
347         struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348 
349         return bdi->dev == NULL;
350 }
351 
352 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353 {
354         struct super_block              *sb = journal->j_private;
355         struct ext4_sb_info             *sbi = EXT4_SB(sb);
356         int                             error = is_journal_aborted(journal);
357         struct ext4_journal_cb_entry    *jce;
358 
359         BUG_ON(txn->t_state == T_FINISHED);
360         spin_lock(&sbi->s_md_lock);
361         while (!list_empty(&txn->t_private_list)) {
362                 jce = list_entry(txn->t_private_list.next,
363                                  struct ext4_journal_cb_entry, jce_list);
364                 list_del_init(&jce->jce_list);
365                 spin_unlock(&sbi->s_md_lock);
366                 jce->jce_func(sb, jce, error);
367                 spin_lock(&sbi->s_md_lock);
368         }
369         spin_unlock(&sbi->s_md_lock);
370 }
371 
372 /* Deal with the reporting of failure conditions on a filesystem such as
373  * inconsistencies detected or read IO failures.
374  *
375  * On ext2, we can store the error state of the filesystem in the
376  * superblock.  That is not possible on ext4, because we may have other
377  * write ordering constraints on the superblock which prevent us from
378  * writing it out straight away; and given that the journal is about to
379  * be aborted, we can't rely on the current, or future, transactions to
380  * write out the superblock safely.
381  *
382  * We'll just use the jbd2_journal_abort() error code to record an error in
383  * the journal instead.  On recovery, the journal will complain about
384  * that error until we've noted it down and cleared it.
385  */
386 
387 static void ext4_handle_error(struct super_block *sb)
388 {
389         if (sb->s_flags & MS_RDONLY)
390                 return;
391 
392         if (!test_opt(sb, ERRORS_CONT)) {
393                 journal_t *journal = EXT4_SB(sb)->s_journal;
394 
395                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
396                 if (journal)
397                         jbd2_journal_abort(journal, -EIO);
398         }
399         if (test_opt(sb, ERRORS_RO)) {
400                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
401                 sb->s_flags |= MS_RDONLY;
402         }
403         if (test_opt(sb, ERRORS_PANIC)) {
404                 if (EXT4_SB(sb)->s_journal &&
405                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
406                         return;
407                 panic("EXT4-fs (device %s): panic forced after error\n",
408                         sb->s_id);
409         }
410 }
411 
412 void __ext4_error(struct super_block *sb, const char *function,
413                   unsigned int line, const char *fmt, ...)
414 {
415         struct va_format vaf;
416         va_list args;
417 
418         va_start(args, fmt);
419         vaf.fmt = fmt;
420         vaf.va = &args;
421         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
422                sb->s_id, function, line, current->comm, &vaf);
423         va_end(args);
424         save_error_info(sb, function, line);
425 
426         ext4_handle_error(sb);
427 }
428 
429 void ext4_error_inode(struct inode *inode, const char *function,
430                       unsigned int line, ext4_fsblk_t block,
431                       const char *fmt, ...)
432 {
433         va_list args;
434         struct va_format vaf;
435         struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
436 
437         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
438         es->s_last_error_block = cpu_to_le64(block);
439         save_error_info(inode->i_sb, function, line);
440         va_start(args, fmt);
441         vaf.fmt = fmt;
442         vaf.va = &args;
443         if (block)
444                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
445                        "inode #%lu: block %llu: comm %s: %pV\n",
446                        inode->i_sb->s_id, function, line, inode->i_ino,
447                        block, current->comm, &vaf);
448         else
449                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
450                        "inode #%lu: comm %s: %pV\n",
451                        inode->i_sb->s_id, function, line, inode->i_ino,
452                        current->comm, &vaf);
453         va_end(args);
454 
455         ext4_handle_error(inode->i_sb);
456 }
457 
458 void ext4_error_file(struct file *file, const char *function,
459                      unsigned int line, ext4_fsblk_t block,
460                      const char *fmt, ...)
461 {
462         va_list args;
463         struct va_format vaf;
464         struct ext4_super_block *es;
465         struct inode *inode = file_inode(file);
466         char pathname[80], *path;
467 
468         es = EXT4_SB(inode->i_sb)->s_es;
469         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
470         save_error_info(inode->i_sb, function, line);
471         path = d_path(&(file->f_path), pathname, sizeof(pathname));
472         if (IS_ERR(path))
473                 path = "(unknown)";
474         va_start(args, fmt);
475         vaf.fmt = fmt;
476         vaf.va = &args;
477         if (block)
478                 printk(KERN_CRIT
479                        "EXT4-fs error (device %s): %s:%d: inode #%lu: "
480                        "block %llu: comm %s: path %s: %pV\n",
481                        inode->i_sb->s_id, function, line, inode->i_ino,
482                        block, current->comm, path, &vaf);
483         else
484                 printk(KERN_CRIT
485                        "EXT4-fs error (device %s): %s:%d: inode #%lu: "
486                        "comm %s: path %s: %pV\n",
487                        inode->i_sb->s_id, function, line, inode->i_ino,
488                        current->comm, path, &vaf);
489         va_end(args);
490 
491         ext4_handle_error(inode->i_sb);
492 }
493 
494 const char *ext4_decode_error(struct super_block *sb, int errno,
495                               char nbuf[16])
496 {
497         char *errstr = NULL;
498 
499         switch (errno) {
500         case -EIO:
501                 errstr = "IO failure";
502                 break;
503         case -ENOMEM:
504                 errstr = "Out of memory";
505                 break;
506         case -EROFS:
507                 if (!sb || (EXT4_SB(sb)->s_journal &&
508                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
509                         errstr = "Journal has aborted";
510                 else
511                         errstr = "Readonly filesystem";
512                 break;
513         default:
514                 /* If the caller passed in an extra buffer for unknown
515                  * errors, textualise them now.  Else we just return
516                  * NULL. */
517                 if (nbuf) {
518                         /* Check for truncated error codes... */
519                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
520                                 errstr = nbuf;
521                 }
522                 break;
523         }
524 
525         return errstr;
526 }
527 
528 /* __ext4_std_error decodes expected errors from journaling functions
529  * automatically and invokes the appropriate error response.  */
530 
531 void __ext4_std_error(struct super_block *sb, const char *function,
532                       unsigned int line, int errno)
533 {
534         char nbuf[16];
535         const char *errstr;
536 
537         /* Special case: if the error is EROFS, and we're not already
538          * inside a transaction, then there's really no point in logging
539          * an error. */
540         if (errno == -EROFS && journal_current_handle() == NULL &&
541             (sb->s_flags & MS_RDONLY))
542                 return;
543 
544         errstr = ext4_decode_error(sb, errno, nbuf);
545         printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546                sb->s_id, function, line, errstr);
547         save_error_info(sb, function, line);
548 
549         ext4_handle_error(sb);
550 }
551 
552 /*
553  * ext4_abort is a much stronger failure handler than ext4_error.  The
554  * abort function may be used to deal with unrecoverable failures such
555  * as journal IO errors or ENOMEM at a critical moment in log management.
556  *
557  * We unconditionally force the filesystem into an ABORT|READONLY state,
558  * unless the error response on the fs has been set to panic in which
559  * case we take the easy way out and panic immediately.
560  */
561 
562 void __ext4_abort(struct super_block *sb, const char *function,
563                 unsigned int line, const char *fmt, ...)
564 {
565         va_list args;
566 
567         save_error_info(sb, function, line);
568         va_start(args, fmt);
569         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570                function, line);
571         vprintk(fmt, args);
572         printk("\n");
573         va_end(args);
574 
575         if ((sb->s_flags & MS_RDONLY) == 0) {
576                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
577                 sb->s_flags |= MS_RDONLY;
578                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
579                 if (EXT4_SB(sb)->s_journal)
580                         jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
581                 save_error_info(sb, function, line);
582         }
583         if (test_opt(sb, ERRORS_PANIC)) {
584                 if (EXT4_SB(sb)->s_journal &&
585                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
586                         return;
587                 panic("EXT4-fs panic from previous error\n");
588         }
589 }
590 
591 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
592 {
593         struct va_format vaf;
594         va_list args;
595 
596         va_start(args, fmt);
597         vaf.fmt = fmt;
598         vaf.va = &args;
599         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
600         va_end(args);
601 }
602 
603 void __ext4_warning(struct super_block *sb, const char *function,
604                     unsigned int line, const char *fmt, ...)
605 {
606         struct va_format vaf;
607         va_list args;
608 
609         va_start(args, fmt);
610         vaf.fmt = fmt;
611         vaf.va = &args;
612         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
613                sb->s_id, function, line, &vaf);
614         va_end(args);
615 }
616 
617 void __ext4_grp_locked_error(const char *function, unsigned int line,
618                              struct super_block *sb, ext4_group_t grp,
619                              unsigned long ino, ext4_fsblk_t block,
620                              const char *fmt, ...)
621 __releases(bitlock)
622 __acquires(bitlock)
623 {
624         struct va_format vaf;
625         va_list args;
626         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
627 
628         es->s_last_error_ino = cpu_to_le32(ino);
629         es->s_last_error_block = cpu_to_le64(block);
630         __save_error_info(sb, function, line);
631 
632         va_start(args, fmt);
633 
634         vaf.fmt = fmt;
635         vaf.va = &args;
636         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
637                sb->s_id, function, line, grp);
638         if (ino)
639                 printk(KERN_CONT "inode %lu: ", ino);
640         if (block)
641                 printk(KERN_CONT "block %llu:", (unsigned long long) block);
642         printk(KERN_CONT "%pV\n", &vaf);
643         va_end(args);
644 
645         if (test_opt(sb, ERRORS_CONT)) {
646                 ext4_commit_super(sb, 0);
647                 return;
648         }
649 
650         ext4_unlock_group(sb, grp);
651         ext4_handle_error(sb);
652         /*
653          * We only get here in the ERRORS_RO case; relocking the group
654          * may be dangerous, but nothing bad will happen since the
655          * filesystem will have already been marked read/only and the
656          * journal has been aborted.  We return 1 as a hint to callers
657          * who might what to use the return value from
658          * ext4_grp_locked_error() to distinguish between the
659          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
660          * aggressively from the ext4 function in question, with a
661          * more appropriate error code.
662          */
663         ext4_lock_group(sb, grp);
664         return;
665 }
666 
667 void ext4_update_dynamic_rev(struct super_block *sb)
668 {
669         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
670 
671         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
672                 return;
673 
674         ext4_warning(sb,
675                      "updating to rev %d because of new feature flag, "
676                      "running e2fsck is recommended",
677                      EXT4_DYNAMIC_REV);
678 
679         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
680         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
681         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
682         /* leave es->s_feature_*compat flags alone */
683         /* es->s_uuid will be set by e2fsck if empty */
684 
685         /*
686          * The rest of the superblock fields should be zero, and if not it
687          * means they are likely already in use, so leave them alone.  We
688          * can leave it up to e2fsck to clean up any inconsistencies there.
689          */
690 }
691 
692 /*
693  * Open the external journal device
694  */
695 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
696 {
697         struct block_device *bdev;
698         char b[BDEVNAME_SIZE];
699 
700         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
701         if (IS_ERR(bdev))
702                 goto fail;
703         return bdev;
704 
705 fail:
706         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
707                         __bdevname(dev, b), PTR_ERR(bdev));
708         return NULL;
709 }
710 
711 /*
712  * Release the journal device
713  */
714 static void ext4_blkdev_put(struct block_device *bdev)
715 {
716         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
717 }
718 
719 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
720 {
721         struct block_device *bdev;
722         bdev = sbi->journal_bdev;
723         if (bdev) {
724                 ext4_blkdev_put(bdev);
725                 sbi->journal_bdev = NULL;
726         }
727 }
728 
729 static inline struct inode *orphan_list_entry(struct list_head *l)
730 {
731         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
732 }
733 
734 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
735 {
736         struct list_head *l;
737 
738         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
739                  le32_to_cpu(sbi->s_es->s_last_orphan));
740 
741         printk(KERN_ERR "sb_info orphan list:\n");
742         list_for_each(l, &sbi->s_orphan) {
743                 struct inode *inode = orphan_list_entry(l);
744                 printk(KERN_ERR "  "
745                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
746                        inode->i_sb->s_id, inode->i_ino, inode,
747                        inode->i_mode, inode->i_nlink,
748                        NEXT_ORPHAN(inode));
749         }
750 }
751 
752 static void ext4_put_super(struct super_block *sb)
753 {
754         struct ext4_sb_info *sbi = EXT4_SB(sb);
755         struct ext4_super_block *es = sbi->s_es;
756         int aborted = 0;
757         int i, err;
758 
759         ext4_unregister_li_request(sb);
760         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
761 
762         flush_workqueue(sbi->dio_unwritten_wq);
763         destroy_workqueue(sbi->dio_unwritten_wq);
764 
765         if (sbi->s_journal) {
766                 aborted = is_journal_aborted(sbi->s_journal);
767                 err = jbd2_journal_destroy(sbi->s_journal);
768                 sbi->s_journal = NULL;
769                 if ((err < 0) && !aborted)
770                         ext4_abort(sb, "Couldn't clean up the journal");
771         }
772 
773         ext4_es_unregister_shrinker(sb);
774         del_timer(&sbi->s_err_report);
775         ext4_release_system_zone(sb);
776         ext4_mb_release(sb);
777         ext4_ext_release(sb);
778         ext4_xattr_put_super(sb);
779 
780         if (!(sb->s_flags & MS_RDONLY) && !aborted) {
781                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
782                 es->s_state = cpu_to_le16(sbi->s_mount_state);
783         }
784         if (!(sb->s_flags & MS_RDONLY))
785                 ext4_commit_super(sb, 1);
786 
787         if (sbi->s_proc) {
788                 remove_proc_entry("options", sbi->s_proc);
789                 remove_proc_entry(sb->s_id, ext4_proc_root);
790         }
791         kobject_del(&sbi->s_kobj);
792 
793         for (i = 0; i < sbi->s_gdb_count; i++)
794                 brelse(sbi->s_group_desc[i]);
795         ext4_kvfree(sbi->s_group_desc);
796         ext4_kvfree(sbi->s_flex_groups);
797         percpu_counter_destroy(&sbi->s_freeclusters_counter);
798         percpu_counter_destroy(&sbi->s_freeinodes_counter);
799         percpu_counter_destroy(&sbi->s_dirs_counter);
800         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
801         percpu_counter_destroy(&sbi->s_extent_cache_cnt);
802         brelse(sbi->s_sbh);
803 #ifdef CONFIG_QUOTA
804         for (i = 0; i < MAXQUOTAS; i++)
805                 kfree(sbi->s_qf_names[i]);
806 #endif
807 
808         /* Debugging code just in case the in-memory inode orphan list
809          * isn't empty.  The on-disk one can be non-empty if we've
810          * detected an error and taken the fs readonly, but the
811          * in-memory list had better be clean by this point. */
812         if (!list_empty(&sbi->s_orphan))
813                 dump_orphan_list(sb, sbi);
814         J_ASSERT(list_empty(&sbi->s_orphan));
815 
816         sync_blockdev(sb->s_bdev);
817         invalidate_bdev(sb->s_bdev);
818         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
819                 /*
820                  * Invalidate the journal device's buffers.  We don't want them
821                  * floating about in memory - the physical journal device may
822                  * hotswapped, and it breaks the `ro-after' testing code.
823                  */
824                 sync_blockdev(sbi->journal_bdev);
825                 invalidate_bdev(sbi->journal_bdev);
826                 ext4_blkdev_remove(sbi);
827         }
828         if (sbi->s_mmp_tsk)
829                 kthread_stop(sbi->s_mmp_tsk);
830         sb->s_fs_info = NULL;
831         /*
832          * Now that we are completely done shutting down the
833          * superblock, we need to actually destroy the kobject.
834          */
835         kobject_put(&sbi->s_kobj);
836         wait_for_completion(&sbi->s_kobj_unregister);
837         if (sbi->s_chksum_driver)
838                 crypto_free_shash(sbi->s_chksum_driver);
839         kfree(sbi->s_blockgroup_lock);
840         kfree(sbi);
841 }
842 
843 static struct kmem_cache *ext4_inode_cachep;
844 
845 /*
846  * Called inside transaction, so use GFP_NOFS
847  */
848 static struct inode *ext4_alloc_inode(struct super_block *sb)
849 {
850         struct ext4_inode_info *ei;
851 
852         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
853         if (!ei)
854                 return NULL;
855 
856         ei->vfs_inode.i_version = 1;
857         INIT_LIST_HEAD(&ei->i_prealloc_list);
858         spin_lock_init(&ei->i_prealloc_lock);
859         ext4_es_init_tree(&ei->i_es_tree);
860         rwlock_init(&ei->i_es_lock);
861         INIT_LIST_HEAD(&ei->i_es_lru);
862         ei->i_es_lru_nr = 0;
863         ei->i_reserved_data_blocks = 0;
864         ei->i_reserved_meta_blocks = 0;
865         ei->i_allocated_meta_blocks = 0;
866         ei->i_da_metadata_calc_len = 0;
867         ei->i_da_metadata_calc_last_lblock = 0;
868         spin_lock_init(&(ei->i_block_reservation_lock));
869 #ifdef CONFIG_QUOTA
870         ei->i_reserved_quota = 0;
871 #endif
872         ei->jinode = NULL;
873         INIT_LIST_HEAD(&ei->i_completed_io_list);
874         spin_lock_init(&ei->i_completed_io_lock);
875         ei->i_sync_tid = 0;
876         ei->i_datasync_tid = 0;
877         atomic_set(&ei->i_ioend_count, 0);
878         atomic_set(&ei->i_unwritten, 0);
879         INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
880 
881         return &ei->vfs_inode;
882 }
883 
884 static int ext4_drop_inode(struct inode *inode)
885 {
886         int drop = generic_drop_inode(inode);
887 
888         trace_ext4_drop_inode(inode, drop);
889         return drop;
890 }
891 
892 static void ext4_i_callback(struct rcu_head *head)
893 {
894         struct inode *inode = container_of(head, struct inode, i_rcu);
895         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
896 }
897 
898 static void ext4_destroy_inode(struct inode *inode)
899 {
900         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
901                 ext4_msg(inode->i_sb, KERN_ERR,
902                          "Inode %lu (%p): orphan list check failed!",
903                          inode->i_ino, EXT4_I(inode));
904                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
905                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
906                                 true);
907                 dump_stack();
908         }
909         call_rcu(&inode->i_rcu, ext4_i_callback);
910 }
911 
912 static void init_once(void *foo)
913 {
914         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
915 
916         INIT_LIST_HEAD(&ei->i_orphan);
917         init_rwsem(&ei->xattr_sem);
918         init_rwsem(&ei->i_data_sem);
919         inode_init_once(&ei->vfs_inode);
920 }
921 
922 static int init_inodecache(void)
923 {
924         ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
925                                              sizeof(struct ext4_inode_info),
926                                              0, (SLAB_RECLAIM_ACCOUNT|
927                                                 SLAB_MEM_SPREAD),
928                                              init_once);
929         if (ext4_inode_cachep == NULL)
930                 return -ENOMEM;
931         return 0;
932 }
933 
934 static void destroy_inodecache(void)
935 {
936         /*
937          * Make sure all delayed rcu free inodes are flushed before we
938          * destroy cache.
939          */
940         rcu_barrier();
941         kmem_cache_destroy(ext4_inode_cachep);
942 }
943 
944 void ext4_clear_inode(struct inode *inode)
945 {
946         invalidate_inode_buffers(inode);
947         clear_inode(inode);
948         dquot_drop(inode);
949         ext4_discard_preallocations(inode);
950         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
951         ext4_es_lru_del(inode);
952         if (EXT4_I(inode)->jinode) {
953                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
954                                                EXT4_I(inode)->jinode);
955                 jbd2_free_inode(EXT4_I(inode)->jinode);
956                 EXT4_I(inode)->jinode = NULL;
957         }
958 }
959 
960 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
961                                         u64 ino, u32 generation)
962 {
963         struct inode *inode;
964 
965         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
966                 return ERR_PTR(-ESTALE);
967         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
968                 return ERR_PTR(-ESTALE);
969 
970         /* iget isn't really right if the inode is currently unallocated!!
971          *
972          * ext4_read_inode will return a bad_inode if the inode had been
973          * deleted, so we should be safe.
974          *
975          * Currently we don't know the generation for parent directory, so
976          * a generation of 0 means "accept any"
977          */
978         inode = ext4_iget_normal(sb, ino);
979         if (IS_ERR(inode))
980                 return ERR_CAST(inode);
981         if (generation && inode->i_generation != generation) {
982                 iput(inode);
983                 return ERR_PTR(-ESTALE);
984         }
985 
986         return inode;
987 }
988 
989 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
990                                         int fh_len, int fh_type)
991 {
992         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
993                                     ext4_nfs_get_inode);
994 }
995 
996 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
997                                         int fh_len, int fh_type)
998 {
999         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1000                                     ext4_nfs_get_inode);
1001 }
1002 
1003 /*
1004  * Try to release metadata pages (indirect blocks, directories) which are
1005  * mapped via the block device.  Since these pages could have journal heads
1006  * which would prevent try_to_free_buffers() from freeing them, we must use
1007  * jbd2 layer's try_to_free_buffers() function to release them.
1008  */
1009 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1010                                  gfp_t wait)
1011 {
1012         journal_t *journal = EXT4_SB(sb)->s_journal;
1013 
1014         WARN_ON(PageChecked(page));
1015         if (!page_has_buffers(page))
1016                 return 0;
1017         if (journal)
1018                 return jbd2_journal_try_to_free_buffers(journal, page,
1019                                                         wait & ~__GFP_WAIT);
1020         return try_to_free_buffers(page);
1021 }
1022 
1023 #ifdef CONFIG_QUOTA
1024 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1025 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1026 
1027 static int ext4_write_dquot(struct dquot *dquot);
1028 static int ext4_acquire_dquot(struct dquot *dquot);
1029 static int ext4_release_dquot(struct dquot *dquot);
1030 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1031 static int ext4_write_info(struct super_block *sb, int type);
1032 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1033                          struct path *path);
1034 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1035                                  int format_id);
1036 static int ext4_quota_off(struct super_block *sb, int type);
1037 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1038 static int ext4_quota_on_mount(struct super_block *sb, int type);
1039 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1040                                size_t len, loff_t off);
1041 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1042                                 const char *data, size_t len, loff_t off);
1043 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1044                              unsigned int flags);
1045 static int ext4_enable_quotas(struct super_block *sb);
1046 
1047 static const struct dquot_operations ext4_quota_operations = {
1048         .get_reserved_space = ext4_get_reserved_space,
1049         .write_dquot    = ext4_write_dquot,
1050         .acquire_dquot  = ext4_acquire_dquot,
1051         .release_dquot  = ext4_release_dquot,
1052         .mark_dirty     = ext4_mark_dquot_dirty,
1053         .write_info     = ext4_write_info,
1054         .alloc_dquot    = dquot_alloc,
1055         .destroy_dquot  = dquot_destroy,
1056 };
1057 
1058 static const struct quotactl_ops ext4_qctl_operations = {
1059         .quota_on       = ext4_quota_on,
1060         .quota_off      = ext4_quota_off,
1061         .quota_sync     = dquot_quota_sync,
1062         .get_info       = dquot_get_dqinfo,
1063         .set_info       = dquot_set_dqinfo,
1064         .get_dqblk      = dquot_get_dqblk,
1065         .set_dqblk      = dquot_set_dqblk
1066 };
1067 
1068 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1069         .quota_on_meta  = ext4_quota_on_sysfile,
1070         .quota_off      = ext4_quota_off_sysfile,
1071         .quota_sync     = dquot_quota_sync,
1072         .get_info       = dquot_get_dqinfo,
1073         .set_info       = dquot_set_dqinfo,
1074         .get_dqblk      = dquot_get_dqblk,
1075         .set_dqblk      = dquot_set_dqblk
1076 };
1077 #endif
1078 
1079 static const struct super_operations ext4_sops = {
1080         .alloc_inode    = ext4_alloc_inode,
1081         .destroy_inode  = ext4_destroy_inode,
1082         .write_inode    = ext4_write_inode,
1083         .dirty_inode    = ext4_dirty_inode,
1084         .drop_inode     = ext4_drop_inode,
1085         .evict_inode    = ext4_evict_inode,
1086         .put_super      = ext4_put_super,
1087         .sync_fs        = ext4_sync_fs,
1088         .freeze_fs      = ext4_freeze,
1089         .unfreeze_fs    = ext4_unfreeze,
1090         .statfs         = ext4_statfs,
1091         .remount_fs     = ext4_remount,
1092         .show_options   = ext4_show_options,
1093 #ifdef CONFIG_QUOTA
1094         .quota_read     = ext4_quota_read,
1095         .quota_write    = ext4_quota_write,
1096 #endif
1097         .bdev_try_to_free_page = bdev_try_to_free_page,
1098 };
1099 
1100 static const struct super_operations ext4_nojournal_sops = {
1101         .alloc_inode    = ext4_alloc_inode,
1102         .destroy_inode  = ext4_destroy_inode,
1103         .write_inode    = ext4_write_inode,
1104         .dirty_inode    = ext4_dirty_inode,
1105         .drop_inode     = ext4_drop_inode,
1106         .evict_inode    = ext4_evict_inode,
1107         .put_super      = ext4_put_super,
1108         .statfs         = ext4_statfs,
1109         .remount_fs     = ext4_remount,
1110         .show_options   = ext4_show_options,
1111 #ifdef CONFIG_QUOTA
1112         .quota_read     = ext4_quota_read,
1113         .quota_write    = ext4_quota_write,
1114 #endif
1115         .bdev_try_to_free_page = bdev_try_to_free_page,
1116 };
1117 
1118 static const struct export_operations ext4_export_ops = {
1119         .fh_to_dentry = ext4_fh_to_dentry,
1120         .fh_to_parent = ext4_fh_to_parent,
1121         .get_parent = ext4_get_parent,
1122 };
1123 
1124 enum {
1125         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1126         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1127         Opt_nouid32, Opt_debug, Opt_removed,
1128         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1129         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1130         Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1131         Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1132         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1133         Opt_data_err_abort, Opt_data_err_ignore,
1134         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1135         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1136         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1137         Opt_usrquota, Opt_grpquota, Opt_i_version,
1138         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1139         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1140         Opt_inode_readahead_blks, Opt_journal_ioprio,
1141         Opt_dioread_nolock, Opt_dioread_lock,
1142         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1143         Opt_max_dir_size_kb,
1144 };
1145 
1146 static const match_table_t tokens = {
1147         {Opt_bsd_df, "bsddf"},
1148         {Opt_minix_df, "minixdf"},
1149         {Opt_grpid, "grpid"},
1150         {Opt_grpid, "bsdgroups"},
1151         {Opt_nogrpid, "nogrpid"},
1152         {Opt_nogrpid, "sysvgroups"},
1153         {Opt_resgid, "resgid=%u"},
1154         {Opt_resuid, "resuid=%u"},
1155         {Opt_sb, "sb=%u"},
1156         {Opt_err_cont, "errors=continue"},
1157         {Opt_err_panic, "errors=panic"},
1158         {Opt_err_ro, "errors=remount-ro"},
1159         {Opt_nouid32, "nouid32"},
1160         {Opt_debug, "debug"},
1161         {Opt_removed, "oldalloc"},
1162         {Opt_removed, "orlov"},
1163         {Opt_user_xattr, "user_xattr"},
1164         {Opt_nouser_xattr, "nouser_xattr"},
1165         {Opt_acl, "acl"},
1166         {Opt_noacl, "noacl"},
1167         {Opt_noload, "norecovery"},
1168         {Opt_noload, "noload"},
1169         {Opt_removed, "nobh"},
1170         {Opt_removed, "bh"},
1171         {Opt_commit, "commit=%u"},
1172         {Opt_min_batch_time, "min_batch_time=%u"},
1173         {Opt_max_batch_time, "max_batch_time=%u"},
1174         {Opt_journal_dev, "journal_dev=%u"},
1175         {Opt_journal_checksum, "journal_checksum"},
1176         {Opt_journal_async_commit, "journal_async_commit"},
1177         {Opt_abort, "abort"},
1178         {Opt_data_journal, "data=journal"},
1179         {Opt_data_ordered, "data=ordered"},
1180         {Opt_data_writeback, "data=writeback"},
1181         {Opt_data_err_abort, "data_err=abort"},
1182         {Opt_data_err_ignore, "data_err=ignore"},
1183         {Opt_offusrjquota, "usrjquota="},
1184         {Opt_usrjquota, "usrjquota=%s"},
1185         {Opt_offgrpjquota, "grpjquota="},
1186         {Opt_grpjquota, "grpjquota=%s"},
1187         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1188         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1189         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1190         {Opt_grpquota, "grpquota"},
1191         {Opt_noquota, "noquota"},
1192         {Opt_quota, "quota"},
1193         {Opt_usrquota, "usrquota"},
1194         {Opt_barrier, "barrier=%u"},
1195         {Opt_barrier, "barrier"},
1196         {Opt_nobarrier, "nobarrier"},
1197         {Opt_i_version, "i_version"},
1198         {Opt_stripe, "stripe=%u"},
1199         {Opt_delalloc, "delalloc"},
1200         {Opt_nodelalloc, "nodelalloc"},
1201         {Opt_removed, "mblk_io_submit"},
1202         {Opt_removed, "nomblk_io_submit"},
1203         {Opt_block_validity, "block_validity"},
1204         {Opt_noblock_validity, "noblock_validity"},
1205         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1206         {Opt_journal_ioprio, "journal_ioprio=%u"},
1207         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1208         {Opt_auto_da_alloc, "auto_da_alloc"},
1209         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1210         {Opt_dioread_nolock, "dioread_nolock"},
1211         {Opt_dioread_lock, "dioread_lock"},
1212         {Opt_discard, "discard"},
1213         {Opt_nodiscard, "nodiscard"},
1214         {Opt_init_itable, "init_itable=%u"},
1215         {Opt_init_itable, "init_itable"},
1216         {Opt_noinit_itable, "noinit_itable"},
1217         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1218         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1219         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1220         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1221         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1222         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1223         {Opt_err, NULL},
1224 };
1225 
1226 static ext4_fsblk_t get_sb_block(void **data)
1227 {
1228         ext4_fsblk_t    sb_block;
1229         char            *options = (char *) *data;
1230 
1231         if (!options || strncmp(options, "sb=", 3) != 0)
1232                 return 1;       /* Default location */
1233 
1234         options += 3;
1235         /* TODO: use simple_strtoll with >32bit ext4 */
1236         sb_block = simple_strtoul(options, &options, 0);
1237         if (*options && *options != ',') {
1238                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1239                        (char *) *data);
1240                 return 1;
1241         }
1242         if (*options == ',')
1243                 options++;
1244         *data = (void *) options;
1245 
1246         return sb_block;
1247 }
1248 
1249 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1250 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1251         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1252 
1253 #ifdef CONFIG_QUOTA
1254 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1255 {
1256         struct ext4_sb_info *sbi = EXT4_SB(sb);
1257         char *qname;
1258         int ret = -1;
1259 
1260         if (sb_any_quota_loaded(sb) &&
1261                 !sbi->s_qf_names[qtype]) {
1262                 ext4_msg(sb, KERN_ERR,
1263                         "Cannot change journaled "
1264                         "quota options when quota turned on");
1265                 return -1;
1266         }
1267         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1268                 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1269                          "when QUOTA feature is enabled");
1270                 return -1;
1271         }
1272         qname = match_strdup(args);
1273         if (!qname) {
1274                 ext4_msg(sb, KERN_ERR,
1275                         "Not enough memory for storing quotafile name");
1276                 return -1;
1277         }
1278         if (sbi->s_qf_names[qtype]) {
1279                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1280                         ret = 1;
1281                 else
1282                         ext4_msg(sb, KERN_ERR,
1283                                  "%s quota file already specified",
1284                                  QTYPE2NAME(qtype));
1285                 goto errout;
1286         }
1287         if (strchr(qname, '/')) {
1288                 ext4_msg(sb, KERN_ERR,
1289                         "quotafile must be on filesystem root");
1290                 goto errout;
1291         }
1292         sbi->s_qf_names[qtype] = qname;
1293         set_opt(sb, QUOTA);
1294         return 1;
1295 errout:
1296         kfree(qname);
1297         return ret;
1298 }
1299 
1300 static int clear_qf_name(struct super_block *sb, int qtype)
1301 {
1302 
1303         struct ext4_sb_info *sbi = EXT4_SB(sb);
1304 
1305         if (sb_any_quota_loaded(sb) &&
1306                 sbi->s_qf_names[qtype]) {
1307                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1308                         " when quota turned on");
1309                 return -1;
1310         }
1311         kfree(sbi->s_qf_names[qtype]);
1312         sbi->s_qf_names[qtype] = NULL;
1313         return 1;
1314 }
1315 #endif
1316 
1317 #define MOPT_SET        0x0001
1318 #define MOPT_CLEAR      0x0002
1319 #define MOPT_NOSUPPORT  0x0004
1320 #define MOPT_EXPLICIT   0x0008
1321 #define MOPT_CLEAR_ERR  0x0010
1322 #define MOPT_GTE0       0x0020
1323 #ifdef CONFIG_QUOTA
1324 #define MOPT_Q          0
1325 #define MOPT_QFMT       0x0040
1326 #else
1327 #define MOPT_Q          MOPT_NOSUPPORT
1328 #define MOPT_QFMT       MOPT_NOSUPPORT
1329 #endif
1330 #define MOPT_DATAJ      0x0080
1331 #define MOPT_NO_EXT2    0x0100
1332 #define MOPT_NO_EXT3    0x0200
1333 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1334 
1335 static const struct mount_opts {
1336         int     token;
1337         int     mount_opt;
1338         int     flags;
1339 } ext4_mount_opts[] = {
1340         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1341         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1342         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1343         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1344         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1345         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1346         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1347          MOPT_EXT4_ONLY | MOPT_SET},
1348         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1349          MOPT_EXT4_ONLY | MOPT_CLEAR},
1350         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1351         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1352         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1353          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1354         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1355          MOPT_EXT4_ONLY | MOPT_CLEAR},
1356         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1357          MOPT_EXT4_ONLY | MOPT_SET},
1358         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1359                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1360          MOPT_EXT4_ONLY | MOPT_SET},
1361         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1362         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1363         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1364         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1365         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1366          MOPT_NO_EXT2 | MOPT_SET},
1367         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1368          MOPT_NO_EXT2 | MOPT_CLEAR},
1369         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1370         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1371         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1372         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1373         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1374         {Opt_commit, 0, MOPT_GTE0},
1375         {Opt_max_batch_time, 0, MOPT_GTE0},
1376         {Opt_min_batch_time, 0, MOPT_GTE0},
1377         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1378         {Opt_init_itable, 0, MOPT_GTE0},
1379         {Opt_stripe, 0, MOPT_GTE0},
1380         {Opt_resuid, 0, MOPT_GTE0},
1381         {Opt_resgid, 0, MOPT_GTE0},
1382         {Opt_journal_dev, 0, MOPT_GTE0},
1383         {Opt_journal_ioprio, 0, MOPT_GTE0},
1384         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1385         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1386         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1387          MOPT_NO_EXT2 | MOPT_DATAJ},
1388         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1389         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1390 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1391         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1392         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1393 #else
1394         {Opt_acl, 0, MOPT_NOSUPPORT},
1395         {Opt_noacl, 0, MOPT_NOSUPPORT},
1396 #endif
1397         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1398         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1399         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1400         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1401                                                         MOPT_SET | MOPT_Q},
1402         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1403                                                         MOPT_SET | MOPT_Q},
1404         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1405                        EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1406         {Opt_usrjquota, 0, MOPT_Q},
1407         {Opt_grpjquota, 0, MOPT_Q},
1408         {Opt_offusrjquota, 0, MOPT_Q},
1409         {Opt_offgrpjquota, 0, MOPT_Q},
1410         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1411         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1412         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1413         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1414         {Opt_err, 0, 0}
1415 };
1416 
1417 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1418                             substring_t *args, unsigned long *journal_devnum,
1419                             unsigned int *journal_ioprio, int is_remount)
1420 {
1421         struct ext4_sb_info *sbi = EXT4_SB(sb);
1422         const struct mount_opts *m;
1423         kuid_t uid;
1424         kgid_t gid;
1425         int arg = 0;
1426 
1427 #ifdef CONFIG_QUOTA
1428         if (token == Opt_usrjquota)
1429                 return set_qf_name(sb, USRQUOTA, &args[0]);
1430         else if (token == Opt_grpjquota)
1431                 return set_qf_name(sb, GRPQUOTA, &args[0]);
1432         else if (token == Opt_offusrjquota)
1433                 return clear_qf_name(sb, USRQUOTA);
1434         else if (token == Opt_offgrpjquota)
1435                 return clear_qf_name(sb, GRPQUOTA);
1436 #endif
1437         switch (token) {
1438         case Opt_noacl:
1439         case Opt_nouser_xattr:
1440                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1441                 break;
1442         case Opt_sb:
1443                 return 1;       /* handled by get_sb_block() */
1444         case Opt_removed:
1445                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1446                 return 1;
1447         case Opt_abort:
1448                 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1449                 return 1;
1450         case Opt_i_version:
1451                 sb->s_flags |= MS_I_VERSION;
1452                 return 1;
1453         }
1454 
1455         for (m = ext4_mount_opts; m->token != Opt_err; m++)
1456                 if (token == m->token)
1457                         break;
1458 
1459         if (m->token == Opt_err) {
1460                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1461                          "or missing value", opt);
1462                 return -1;
1463         }
1464 
1465         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1466                 ext4_msg(sb, KERN_ERR,
1467                          "Mount option \"%s\" incompatible with ext2", opt);
1468                 return -1;
1469         }
1470         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1471                 ext4_msg(sb, KERN_ERR,
1472                          "Mount option \"%s\" incompatible with ext3", opt);
1473                 return -1;
1474         }
1475 
1476         if (args->from && match_int(args, &arg))
1477                 return -1;
1478         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1479                 return -1;
1480         if (m->flags & MOPT_EXPLICIT)
1481                 set_opt2(sb, EXPLICIT_DELALLOC);
1482         if (m->flags & MOPT_CLEAR_ERR)
1483                 clear_opt(sb, ERRORS_MASK);
1484         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1485                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1486                          "options when quota turned on");
1487                 return -1;
1488         }
1489 
1490         if (m->flags & MOPT_NOSUPPORT) {
1491                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1492         } else if (token == Opt_commit) {
1493                 if (arg == 0)
1494                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1495                 sbi->s_commit_interval = HZ * arg;
1496         } else if (token == Opt_max_batch_time) {
1497                 sbi->s_max_batch_time = arg;
1498         } else if (token == Opt_min_batch_time) {
1499                 sbi->s_min_batch_time = arg;
1500         } else if (token == Opt_inode_readahead_blks) {
1501                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1502                         ext4_msg(sb, KERN_ERR,
1503                                  "EXT4-fs: inode_readahead_blks must be "
1504                                  "0 or a power of 2 smaller than 2^31");
1505                         return -1;
1506                 }
1507                 sbi->s_inode_readahead_blks = arg;
1508         } else if (token == Opt_init_itable) {
1509                 set_opt(sb, INIT_INODE_TABLE);
1510                 if (!args->from)
1511                         arg = EXT4_DEF_LI_WAIT_MULT;
1512                 sbi->s_li_wait_mult = arg;
1513         } else if (token == Opt_max_dir_size_kb) {
1514                 sbi->s_max_dir_size_kb = arg;
1515         } else if (token == Opt_stripe) {
1516                 sbi->s_stripe = arg;
1517         } else if (token == Opt_resuid) {
1518                 uid = make_kuid(current_user_ns(), arg);
1519                 if (!uid_valid(uid)) {
1520                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1521                         return -1;
1522                 }
1523                 sbi->s_resuid = uid;
1524         } else if (token == Opt_resgid) {
1525                 gid = make_kgid(current_user_ns(), arg);
1526                 if (!gid_valid(gid)) {
1527                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1528                         return -1;
1529                 }
1530                 sbi->s_resgid = gid;
1531         } else if (token == Opt_journal_dev) {
1532                 if (is_remount) {
1533                         ext4_msg(sb, KERN_ERR,
1534                                  "Cannot specify journal on remount");
1535                         return -1;
1536                 }
1537                 *journal_devnum = arg;
1538         } else if (token == Opt_journal_ioprio) {
1539                 if (arg > 7) {
1540                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1541                                  " (must be 0-7)");
1542                         return -1;
1543                 }
1544                 *journal_ioprio =
1545                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1546         } else if (m->flags & MOPT_DATAJ) {
1547                 if (is_remount) {
1548                         if (!sbi->s_journal)
1549                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1550                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1551                                 ext4_msg(sb, KERN_ERR,
1552                                          "Cannot change data mode on remount");
1553                                 return -1;
1554                         }
1555                 } else {
1556                         clear_opt(sb, DATA_FLAGS);
1557                         sbi->s_mount_opt |= m->mount_opt;
1558                 }
1559 #ifdef CONFIG_QUOTA
1560         } else if (m->flags & MOPT_QFMT) {
1561                 if (sb_any_quota_loaded(sb) &&
1562                     sbi->s_jquota_fmt != m->mount_opt) {
1563                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1564                                  "quota options when quota turned on");
1565                         return -1;
1566                 }
1567                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1568                                                EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1569                         ext4_msg(sb, KERN_ERR,
1570                                  "Cannot set journaled quota options "
1571                                  "when QUOTA feature is enabled");
1572                         return -1;
1573                 }
1574                 sbi->s_jquota_fmt = m->mount_opt;
1575 #endif
1576         } else {
1577                 if (!args->from)
1578                         arg = 1;
1579                 if (m->flags & MOPT_CLEAR)
1580                         arg = !arg;
1581                 else if (unlikely(!(m->flags & MOPT_SET))) {
1582                         ext4_msg(sb, KERN_WARNING,
1583                                  "buggy handling of option %s", opt);
1584                         WARN_ON(1);
1585                         return -1;
1586                 }
1587                 if (arg != 0)
1588                         sbi->s_mount_opt |= m->mount_opt;
1589                 else
1590                         sbi->s_mount_opt &= ~m->mount_opt;
1591         }
1592         return 1;
1593 }
1594 
1595 static int parse_options(char *options, struct super_block *sb,
1596                          unsigned long *journal_devnum,
1597                          unsigned int *journal_ioprio,
1598                          int is_remount)
1599 {
1600         struct ext4_sb_info *sbi = EXT4_SB(sb);
1601         char *p;
1602         substring_t args[MAX_OPT_ARGS];
1603         int token;
1604 
1605         if (!options)
1606                 return 1;
1607 
1608         while ((p = strsep(&options, ",")) != NULL) {
1609                 if (!*p)
1610                         continue;
1611                 /*
1612                  * Initialize args struct so we know whether arg was
1613                  * found; some options take optional arguments.
1614                  */
1615                 args[0].to = args[0].from = NULL;
1616                 token = match_token(p, tokens, args);
1617                 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1618                                      journal_ioprio, is_remount) < 0)
1619                         return 0;
1620         }
1621 #ifdef CONFIG_QUOTA
1622         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1623             (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1624                 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1625                          "feature is enabled");
1626                 return 0;
1627         }
1628         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1629                 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1630                         clear_opt(sb, USRQUOTA);
1631 
1632                 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1633                         clear_opt(sb, GRPQUOTA);
1634 
1635                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1636                         ext4_msg(sb, KERN_ERR, "old and new quota "
1637                                         "format mixing");
1638                         return 0;
1639                 }
1640 
1641                 if (!sbi->s_jquota_fmt) {
1642                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1643                                         "not specified");
1644                         return 0;
1645                 }
1646         }
1647 #endif
1648         if (test_opt(sb, DIOREAD_NOLOCK)) {
1649                 int blocksize =
1650                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1651 
1652                 if (blocksize < PAGE_CACHE_SIZE) {
1653                         ext4_msg(sb, KERN_ERR, "can't mount with "
1654                                  "dioread_nolock if block size != PAGE_SIZE");
1655                         return 0;
1656                 }
1657         }
1658         return 1;
1659 }
1660 
1661 static inline void ext4_show_quota_options(struct seq_file *seq,
1662                                            struct super_block *sb)
1663 {
1664 #if defined(CONFIG_QUOTA)
1665         struct ext4_sb_info *sbi = EXT4_SB(sb);
1666 
1667         if (sbi->s_jquota_fmt) {
1668                 char *fmtname = "";
1669 
1670                 switch (sbi->s_jquota_fmt) {
1671                 case QFMT_VFS_OLD:
1672                         fmtname = "vfsold";
1673                         break;
1674                 case QFMT_VFS_V0:
1675                         fmtname = "vfsv0";
1676                         break;
1677                 case QFMT_VFS_V1:
1678                         fmtname = "vfsv1";
1679                         break;
1680                 }
1681                 seq_printf(seq, ",jqfmt=%s", fmtname);
1682         }
1683 
1684         if (sbi->s_qf_names[USRQUOTA])
1685                 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1686 
1687         if (sbi->s_qf_names[GRPQUOTA])
1688                 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1689 #endif
1690 }
1691 
1692 static const char *token2str(int token)
1693 {
1694         const struct match_token *t;
1695 
1696         for (t = tokens; t->token != Opt_err; t++)
1697                 if (t->token == token && !strchr(t->pattern, '='))
1698                         break;
1699         return t->pattern;
1700 }
1701 
1702 /*
1703  * Show an option if
1704  *  - it's set to a non-default value OR
1705  *  - if the per-sb default is different from the global default
1706  */
1707 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1708                               int nodefs)
1709 {
1710         struct ext4_sb_info *sbi = EXT4_SB(sb);
1711         struct ext4_super_block *es = sbi->s_es;
1712         int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1713         const struct mount_opts *m;
1714         char sep = nodefs ? '\n' : ',';
1715 
1716 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1717 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1718 
1719         if (sbi->s_sb_block != 1)
1720                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1721 
1722         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1723                 int want_set = m->flags & MOPT_SET;
1724                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1725                     (m->flags & MOPT_CLEAR_ERR))
1726                         continue;
1727                 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1728                         continue; /* skip if same as the default */
1729                 if ((want_set &&
1730                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1731                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1732                         continue; /* select Opt_noFoo vs Opt_Foo */
1733                 SEQ_OPTS_PRINT("%s", token2str(m->token));
1734         }
1735 
1736         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1737             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1738                 SEQ_OPTS_PRINT("resuid=%u",
1739                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1740         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1741             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1742                 SEQ_OPTS_PRINT("resgid=%u",
1743                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1744         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1745         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1746                 SEQ_OPTS_PUTS("errors=remount-ro");
1747         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1748                 SEQ_OPTS_PUTS("errors=continue");
1749         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1750                 SEQ_OPTS_PUTS("errors=panic");
1751         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1752                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1753         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1754                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1755         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1756                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1757         if (sb->s_flags & MS_I_VERSION)
1758                 SEQ_OPTS_PUTS("i_version");
1759         if (nodefs || sbi->s_stripe)
1760                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1761         if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1762                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1763                         SEQ_OPTS_PUTS("data=journal");
1764                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1765                         SEQ_OPTS_PUTS("data=ordered");
1766                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1767                         SEQ_OPTS_PUTS("data=writeback");
1768         }
1769         if (nodefs ||
1770             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1771                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1772                                sbi->s_inode_readahead_blks);
1773 
1774         if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1775                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1776                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1777         if (nodefs || sbi->s_max_dir_size_kb)
1778                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1779 
1780         ext4_show_quota_options(seq, sb);
1781         return 0;
1782 }
1783 
1784 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1785 {
1786         return _ext4_show_options(seq, root->d_sb, 0);
1787 }
1788 
1789 static int options_seq_show(struct seq_file *seq, void *offset)
1790 {
1791         struct super_block *sb = seq->private;
1792         int rc;
1793 
1794         seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1795         rc = _ext4_show_options(seq, sb, 1);
1796         seq_puts(seq, "\n");
1797         return rc;
1798 }
1799 
1800 static int options_open_fs(struct inode *inode, struct file *file)
1801 {
1802         return single_open(file, options_seq_show, PDE_DATA(inode));
1803 }
1804 
1805 static const struct file_operations ext4_seq_options_fops = {
1806         .owner = THIS_MODULE,
1807         .open = options_open_fs,
1808         .read = seq_read,
1809         .llseek = seq_lseek,
1810         .release = single_release,
1811 };
1812 
1813 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1814                             int read_only)
1815 {
1816         struct ext4_sb_info *sbi = EXT4_SB(sb);
1817         int res = 0;
1818 
1819         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1820                 ext4_msg(sb, KERN_ERR, "revision level too high, "
1821                          "forcing read-only mode");
1822                 res = MS_RDONLY;
1823         }
1824         if (read_only)
1825                 goto done;
1826         if (!(sbi->s_mount_state & EXT4_VALID_FS))
1827                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1828                          "running e2fsck is recommended");
1829         else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1830                 ext4_msg(sb, KERN_WARNING,
1831                          "warning: mounting fs with errors, "
1832                          "running e2fsck is recommended");
1833         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1834                  le16_to_cpu(es->s_mnt_count) >=
1835                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1836                 ext4_msg(sb, KERN_WARNING,
1837                          "warning: maximal mount count reached, "
1838                          "running e2fsck is recommended");
1839         else if (le32_to_cpu(es->s_checkinterval) &&
1840                 (le32_to_cpu(es->s_lastcheck) +
1841                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1842                 ext4_msg(sb, KERN_WARNING,
1843                          "warning: checktime reached, "
1844                          "running e2fsck is recommended");
1845         if (!sbi->s_journal)
1846                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1847         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1848                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1849         le16_add_cpu(&es->s_mnt_count, 1);
1850         es->s_mtime = cpu_to_le32(get_seconds());
1851         ext4_update_dynamic_rev(sb);
1852         if (sbi->s_journal)
1853                 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1854 
1855         ext4_commit_super(sb, 1);
1856 done:
1857         if (test_opt(sb, DEBUG))
1858                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1859                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1860                         sb->s_blocksize,
1861                         sbi->s_groups_count,
1862                         EXT4_BLOCKS_PER_GROUP(sb),
1863                         EXT4_INODES_PER_GROUP(sb),
1864                         sbi->s_mount_opt, sbi->s_mount_opt2);
1865 
1866         cleancache_init_fs(sb);
1867         return res;
1868 }
1869 
1870 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1871 {
1872         struct ext4_sb_info *sbi = EXT4_SB(sb);
1873         struct flex_groups *new_groups;
1874         int size;
1875 
1876         if (!sbi->s_log_groups_per_flex)
1877                 return 0;
1878 
1879         size = ext4_flex_group(sbi, ngroup - 1) + 1;
1880         if (size <= sbi->s_flex_groups_allocated)
1881                 return 0;
1882 
1883         size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1884         new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1885         if (!new_groups) {
1886                 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1887                          size / (int) sizeof(struct flex_groups));
1888                 return -ENOMEM;
1889         }
1890 
1891         if (sbi->s_flex_groups) {
1892                 memcpy(new_groups, sbi->s_flex_groups,
1893                        (sbi->s_flex_groups_allocated *
1894                         sizeof(struct flex_groups)));
1895                 ext4_kvfree(sbi->s_flex_groups);
1896         }
1897         sbi->s_flex_groups = new_groups;
1898         sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1899         return 0;
1900 }
1901 
1902 static int ext4_fill_flex_info(struct super_block *sb)
1903 {
1904         struct ext4_sb_info *sbi = EXT4_SB(sb);
1905         struct ext4_group_desc *gdp = NULL;
1906         ext4_group_t flex_group;
1907         unsigned int groups_per_flex = 0;
1908         int i, err;
1909 
1910         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1911         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1912                 sbi->s_log_groups_per_flex = 0;
1913                 return 1;
1914         }
1915         groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1916 
1917         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1918         if (err)
1919                 goto failed;
1920 
1921         for (i = 0; i < sbi->s_groups_count; i++) {
1922                 gdp = ext4_get_group_desc(sb, i, NULL);
1923 
1924                 flex_group = ext4_flex_group(sbi, i);
1925                 atomic_add(ext4_free_inodes_count(sb, gdp),
1926                            &sbi->s_flex_groups[flex_group].free_inodes);
1927                 atomic64_add(ext4_free_group_clusters(sb, gdp),
1928                              &sbi->s_flex_groups[flex_group].free_clusters);
1929                 atomic_add(ext4_used_dirs_count(sb, gdp),
1930                            &sbi->s_flex_groups[flex_group].used_dirs);
1931         }
1932 
1933         return 1;
1934 failed:
1935         return 0;
1936 }
1937 
1938 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1939                                    struct ext4_group_desc *gdp)
1940 {
1941         int offset = offsetof(struct ext4_group_desc, bg_checksum);
1942         __u16 crc = 0;
1943         __le32 le_group = cpu_to_le32(block_group);
1944 
1945         if ((sbi->s_es->s_feature_ro_compat &
1946              cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1947                 /* Use new metadata_csum algorithm */
1948                 __u32 csum32;
1949                 __u16 dummy_csum = 0;
1950 
1951                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1952                                      sizeof(le_group));
1953                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
1954                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
1955                                      sizeof(dummy_csum));
1956                 offset += sizeof(dummy_csum);
1957                 if (offset < sbi->s_desc_size)
1958                         csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
1959                                              sbi->s_desc_size - offset);
1960 
1961                 crc = csum32 & 0xFFFF;
1962                 goto out;
1963         }
1964 
1965         /* old crc16 code */
1966         if (!(sbi->s_es->s_feature_ro_compat &
1967               cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
1968                 return 0;
1969 
1970         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1971         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1972         crc = crc16(crc, (__u8 *)gdp, offset);
1973         offset += sizeof(gdp->bg_checksum); /* skip checksum */
1974         /* for checksum of struct ext4_group_desc do the rest...*/
1975         if ((sbi->s_es->s_feature_incompat &
1976              cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1977             offset < le16_to_cpu(sbi->s_es->s_desc_size))
1978                 crc = crc16(crc, (__u8 *)gdp + offset,
1979                             le16_to_cpu(sbi->s_es->s_desc_size) -
1980                                 offset);
1981 
1982 out:
1983         return cpu_to_le16(crc);
1984 }
1985 
1986 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1987                                 struct ext4_group_desc *gdp)
1988 {
1989         if (ext4_has_group_desc_csum(sb) &&
1990             (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1991                                                       block_group, gdp)))
1992                 return 0;
1993 
1994         return 1;
1995 }
1996 
1997 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1998                               struct ext4_group_desc *gdp)
1999 {
2000         if (!ext4_has_group_desc_csum(sb))
2001                 return;
2002         gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2003 }
2004 
2005 /* Called at mount-time, super-block is locked */
2006 static int ext4_check_descriptors(struct super_block *sb,
2007                                   ext4_fsblk_t sb_block,
2008                                   ext4_group_t *first_not_zeroed)
2009 {
2010         struct ext4_sb_info *sbi = EXT4_SB(sb);
2011         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2012         ext4_fsblk_t last_block;
2013         ext4_fsblk_t block_bitmap;
2014         ext4_fsblk_t inode_bitmap;
2015         ext4_fsblk_t inode_table;
2016         int flexbg_flag = 0;
2017         ext4_group_t i, grp = sbi->s_groups_count;
2018 
2019         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2020                 flexbg_flag = 1;
2021 
2022         ext4_debug("Checking group descriptors");
2023 
2024         for (i = 0; i < sbi->s_groups_count; i++) {
2025                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2026 
2027                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2028                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2029                 else
2030                         last_block = first_block +
2031                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2032 
2033                 if ((grp == sbi->s_groups_count) &&
2034                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2035                         grp = i;
2036 
2037                 block_bitmap = ext4_block_bitmap(sb, gdp);
2038                 if (block_bitmap == sb_block) {
2039                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2040                                  "Block bitmap for group %u overlaps "
2041                                  "superblock", i);
2042                 }
2043                 if (block_bitmap < first_block || block_bitmap > last_block) {
2044                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2045                                "Block bitmap for group %u not in group "
2046                                "(block %llu)!", i, block_bitmap);
2047                         return 0;
2048                 }
2049                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2050                 if (inode_bitmap == sb_block) {
2051                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2052                                  "Inode bitmap for group %u overlaps "
2053                                  "superblock", i);
2054                 }
2055                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2056                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2057                                "Inode bitmap for group %u not in group "
2058                                "(block %llu)!", i, inode_bitmap);
2059                         return 0;
2060                 }
2061                 inode_table = ext4_inode_table(sb, gdp);
2062                 if (inode_table == sb_block) {
2063                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2064                                  "Inode table for group %u overlaps "
2065                                  "superblock", i);
2066                 }
2067                 if (inode_table < first_block ||
2068                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2069                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2070                                "Inode table for group %u not in group "
2071                                "(block %llu)!", i, inode_table);
2072                         return 0;
2073                 }
2074                 ext4_lock_group(sb, i);
2075                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2076                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2077                                  "Checksum for group %u failed (%u!=%u)",
2078                                  i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2079                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2080                         if (!(sb->s_flags & MS_RDONLY)) {
2081                                 ext4_unlock_group(sb, i);
2082                                 return 0;
2083                         }
2084                 }
2085                 ext4_unlock_group(sb, i);
2086                 if (!flexbg_flag)
2087                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2088         }
2089         if (NULL != first_not_zeroed)
2090                 *first_not_zeroed = grp;
2091 
2092         ext4_free_blocks_count_set(sbi->s_es,
2093                                    EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2094         sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2095         return 1;
2096 }
2097 
2098 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2099  * the superblock) which were deleted from all directories, but held open by
2100  * a process at the time of a crash.  We walk the list and try to delete these
2101  * inodes at recovery time (only with a read-write filesystem).
2102  *
2103  * In order to keep the orphan inode chain consistent during traversal (in
2104  * case of crash during recovery), we link each inode into the superblock
2105  * orphan list_head and handle it the same way as an inode deletion during
2106  * normal operation (which journals the operations for us).
2107  *
2108  * We only do an iget() and an iput() on each inode, which is very safe if we
2109  * accidentally point at an in-use or already deleted inode.  The worst that
2110  * can happen in this case is that we get a "bit already cleared" message from
2111  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2112  * e2fsck was run on this filesystem, and it must have already done the orphan
2113  * inode cleanup for us, so we can safely abort without any further action.
2114  */
2115 static void ext4_orphan_cleanup(struct super_block *sb,
2116                                 struct ext4_super_block *es)
2117 {
2118         unsigned int s_flags = sb->s_flags;
2119         int nr_orphans = 0, nr_truncates = 0;
2120 #ifdef CONFIG_QUOTA
2121         int i;
2122 #endif
2123         if (!es->s_last_orphan) {
2124                 jbd_debug(4, "no orphan inodes to clean up\n");
2125                 return;
2126         }
2127 
2128         if (bdev_read_only(sb->s_bdev)) {
2129                 ext4_msg(sb, KERN_ERR, "write access "
2130                         "unavailable, skipping orphan cleanup");
2131                 return;
2132         }
2133 
2134         /* Check if feature set would not allow a r/w mount */
2135         if (!ext4_feature_set_ok(sb, 0)) {
2136                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2137                          "unknown ROCOMPAT features");
2138                 return;
2139         }
2140 
2141         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2142                 /* don't clear list on RO mount w/ errors */
2143                 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2144                         jbd_debug(1, "Errors on filesystem, "
2145                                   "clearing orphan list.\n");
2146                         es->s_last_orphan = 0;
2147                 }
2148                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2149                 return;
2150         }
2151 
2152         if (s_flags & MS_RDONLY) {
2153                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2154                 sb->s_flags &= ~MS_RDONLY;
2155         }
2156 #ifdef CONFIG_QUOTA
2157         /* Needed for iput() to work correctly and not trash data */
2158         sb->s_flags |= MS_ACTIVE;
2159         /* Turn on quotas so that they are updated correctly */
2160         for (i = 0; i < MAXQUOTAS; i++) {
2161                 if (EXT4_SB(sb)->s_qf_names[i]) {
2162                         int ret = ext4_quota_on_mount(sb, i);
2163                         if (ret < 0)
2164                                 ext4_msg(sb, KERN_ERR,
2165                                         "Cannot turn on journaled "
2166                                         "quota: error %d", ret);
2167                 }
2168         }
2169 #endif
2170 
2171         while (es->s_last_orphan) {
2172                 struct inode *inode;
2173 
2174                 /*
2175                  * We may have encountered an error during cleanup; if
2176                  * so, skip the rest.
2177                  */
2178                 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2179                         jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2180                         es->s_last_orphan = 0;
2181                         break;
2182                 }
2183 
2184                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2185                 if (IS_ERR(inode)) {
2186                         es->s_last_orphan = 0;
2187                         break;
2188                 }
2189 
2190                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2191                 dquot_initialize(inode);
2192                 if (inode->i_nlink) {
2193                         ext4_msg(sb, KERN_DEBUG,
2194                                 "%s: truncating inode %lu to %lld bytes",
2195                                 __func__, inode->i_ino, inode->i_size);
2196                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2197                                   inode->i_ino, inode->i_size);
2198                         mutex_lock(&inode->i_mutex);
2199                         ext4_truncate(inode);
2200                         mutex_unlock(&inode->i_mutex);
2201                         nr_truncates++;
2202                 } else {
2203                         ext4_msg(sb, KERN_DEBUG,
2204                                 "%s: deleting unreferenced inode %lu",
2205                                 __func__, inode->i_ino);
2206                         jbd_debug(2, "deleting unreferenced inode %lu\n",
2207                                   inode->i_ino);
2208                         nr_orphans++;
2209                 }
2210                 iput(inode);  /* The delete magic happens here! */
2211         }
2212 
2213 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2214 
2215         if (nr_orphans)
2216                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2217                        PLURAL(nr_orphans));
2218         if (nr_truncates)
2219                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2220                        PLURAL(nr_truncates));
2221 #ifdef CONFIG_QUOTA
2222         /* Turn quotas off */
2223         for (i = 0; i < MAXQUOTAS; i++) {
2224                 if (sb_dqopt(sb)->files[i])
2225                         dquot_quota_off(sb, i);
2226         }
2227 #endif
2228         sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2229 }
2230 
2231 /*
2232  * Maximal extent format file size.
2233  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2234  * extent format containers, within a sector_t, and within i_blocks
2235  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2236  * so that won't be a limiting factor.
2237  *
2238  * However there is other limiting factor. We do store extents in the form
2239  * of starting block and length, hence the resulting length of the extent
2240  * covering maximum file size must fit into on-disk format containers as
2241  * well. Given that length is always by 1 unit bigger than max unit (because
2242  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2243  *
2244  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2245  */
2246 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2247 {
2248         loff_t res;
2249         loff_t upper_limit = MAX_LFS_FILESIZE;
2250 
2251         /* small i_blocks in vfs inode? */
2252         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2253                 /*
2254                  * CONFIG_LBDAF is not enabled implies the inode
2255                  * i_block represent total blocks in 512 bytes
2256                  * 32 == size of vfs inode i_blocks * 8
2257                  */
2258                 upper_limit = (1LL << 32) - 1;
2259 
2260                 /* total blocks in file system block size */
2261                 upper_limit >>= (blkbits - 9);
2262                 upper_limit <<= blkbits;
2263         }
2264 
2265         /*
2266          * 32-bit extent-start container, ee_block. We lower the maxbytes
2267          * by one fs block, so ee_len can cover the extent of maximum file
2268          * size
2269          */
2270         res = (1LL << 32) - 1;
2271         res <<= blkbits;
2272 
2273         /* Sanity check against vm- & vfs- imposed limits */
2274         if (res > upper_limit)
2275                 res = upper_limit;
2276 
2277         return res;
2278 }
2279 
2280 /*
2281  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2282  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2283  * We need to be 1 filesystem block less than the 2^48 sector limit.
2284  */
2285 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2286 {
2287         loff_t res = EXT4_NDIR_BLOCKS;
2288         int meta_blocks;
2289         loff_t upper_limit;
2290         /* This is calculated to be the largest file size for a dense, block
2291          * mapped file such that the file's total number of 512-byte sectors,
2292          * including data and all indirect blocks, does not exceed (2^48 - 1).
2293          *
2294          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2295          * number of 512-byte sectors of the file.
2296          */
2297 
2298         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2299                 /*
2300                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
2301                  * the inode i_block field represents total file blocks in
2302                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2303                  */
2304                 upper_limit = (1LL << 32) - 1;
2305 
2306                 /* total blocks in file system block size */
2307                 upper_limit >>= (bits - 9);
2308 
2309         } else {
2310                 /*
2311                  * We use 48 bit ext4_inode i_blocks
2312                  * With EXT4_HUGE_FILE_FL set the i_blocks
2313                  * represent total number of blocks in
2314                  * file system block size
2315                  */
2316                 upper_limit = (1LL << 48) - 1;
2317 
2318         }
2319 
2320         /* indirect blocks */
2321         meta_blocks = 1;
2322         /* double indirect blocks */
2323         meta_blocks += 1 + (1LL << (bits-2));
2324         /* tripple indirect blocks */
2325         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2326 
2327         upper_limit -= meta_blocks;
2328         upper_limit <<= bits;
2329 
2330         res += 1LL << (bits-2);
2331         res += 1LL << (2*(bits-2));
2332         res += 1LL << (3*(bits-2));
2333         res <<= bits;
2334         if (res > upper_limit)
2335                 res = upper_limit;
2336 
2337         if (res > MAX_LFS_FILESIZE)
2338                 res = MAX_LFS_FILESIZE;
2339 
2340         return res;
2341 }
2342 
2343 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2344                                    ext4_fsblk_t logical_sb_block, int nr)
2345 {
2346         struct ext4_sb_info *sbi = EXT4_SB(sb);
2347         ext4_group_t bg, first_meta_bg;
2348         int has_super = 0;
2349 
2350         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2351 
2352         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2353             nr < first_meta_bg)
2354                 return logical_sb_block + nr + 1;
2355         bg = sbi->s_desc_per_block * nr;
2356         if (ext4_bg_has_super(sb, bg))
2357                 has_super = 1;
2358 
2359         return (has_super + ext4_group_first_block_no(sb, bg));
2360 }
2361 
2362 /**
2363  * ext4_get_stripe_size: Get the stripe size.
2364  * @sbi: In memory super block info
2365  *
2366  * If we have specified it via mount option, then
2367  * use the mount option value. If the value specified at mount time is
2368  * greater than the blocks per group use the super block value.
2369  * If the super block value is greater than blocks per group return 0.
2370  * Allocator needs it be less than blocks per group.
2371  *
2372  */
2373 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2374 {
2375         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2376         unsigned long stripe_width =
2377                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2378         int ret;
2379 
2380         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2381                 ret = sbi->s_stripe;
2382         else if (stripe_width <= sbi->s_blocks_per_group)
2383                 ret = stripe_width;
2384         else if (stride <= sbi->s_blocks_per_group)
2385                 ret = stride;
2386         else
2387                 ret = 0;
2388 
2389         /*
2390          * If the stripe width is 1, this makes no sense and
2391          * we set it to 0 to turn off stripe handling code.
2392          */
2393         if (ret <= 1)
2394                 ret = 0;
2395 
2396         return ret;
2397 }
2398 
2399 /* sysfs supprt */
2400 
2401 struct ext4_attr {
2402         struct attribute attr;
2403         ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2404         ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2405                          const char *, size_t);
2406         int offset;
2407 };
2408 
2409 static int parse_strtoull(const char *buf,
2410                 unsigned long long max, unsigned long long *value)
2411 {
2412         int ret;
2413 
2414         ret = kstrtoull(skip_spaces(buf), 0, value);
2415         if (!ret && *value > max)
2416                 ret = -EINVAL;
2417         return ret;
2418 }
2419 
2420 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2421                                               struct ext4_sb_info *sbi,
2422                                               char *buf)
2423 {
2424         return snprintf(buf, PAGE_SIZE, "%llu\n",
2425                 (s64) EXT4_C2B(sbi,
2426                         percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2427 }
2428 
2429 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2430                                          struct ext4_sb_info *sbi, char *buf)
2431 {
2432         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2433 
2434         if (!sb->s_bdev->bd_part)
2435                 return snprintf(buf, PAGE_SIZE, "\n");
2436         return snprintf(buf, PAGE_SIZE, "%lu\n",
2437                         (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2438                          sbi->s_sectors_written_start) >> 1);
2439 }
2440 
2441 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2442                                           struct ext4_sb_info *sbi, char *buf)
2443 {
2444         struct super_block *sb = sbi->s_buddy_cache->i_sb;
2445 
2446         if (!sb->s_bdev->bd_part)
2447                 return snprintf(buf, PAGE_SIZE, "\n");
2448         return snprintf(buf, PAGE_SIZE, "%llu\n",
2449                         (unsigned long long)(sbi->s_kbytes_written +
2450                         ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2451                           EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2452 }
2453 
2454 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2455                                           struct ext4_sb_info *sbi,
2456                                           const char *buf, size_t count)
2457 {
2458         unsigned long t;
2459         int ret;
2460 
2461         ret = kstrtoul(skip_spaces(buf), 0, &t);
2462         if (ret)
2463                 return ret;
2464 
2465         if (t && (!is_power_of_2(t) || t > 0x40000000))
2466                 return -EINVAL;
2467 
2468         sbi->s_inode_readahead_blks = t;
2469         return count;
2470 }
2471 
2472 static ssize_t sbi_ui_show(struct ext4_attr *a,
2473                            struct ext4_sb_info *sbi, char *buf)
2474 {
2475         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2476 
2477         return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2478 }
2479 
2480 static ssize_t sbi_ui_store(struct ext4_attr *a,
2481                             struct ext4_sb_info *sbi,
2482                             const char *buf, size_t count)
2483 {
2484         unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2485         unsigned long t;
2486         int ret;
2487 
2488         ret = kstrtoul(skip_spaces(buf), 0, &t);
2489         if (ret)
2490                 return ret;
2491         *ui = t;
2492         return count;
2493 }
2494 
2495 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2496                                   struct ext4_sb_info *sbi, char *buf)
2497 {
2498         return snprintf(buf, PAGE_SIZE, "%llu\n",
2499                 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2500 }
2501 
2502 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2503                                    struct ext4_sb_info *sbi,
2504                                    const char *buf, size_t count)
2505 {
2506         unsigned long long val;
2507         int ret;
2508 
2509         if (parse_strtoull(buf, -1ULL, &val))
2510                 return -EINVAL;
2511         ret = ext4_reserve_clusters(sbi, val);
2512 
2513         return ret ? ret : count;
2514 }
2515 
2516 static ssize_t trigger_test_error(struct ext4_attr *a,
2517                                   struct ext4_sb_info *sbi,
2518                                   const char *buf, size_t count)
2519 {
2520         int len = count;
2521 
2522         if (!capable(CAP_SYS_ADMIN))
2523                 return -EPERM;
2524 
2525         if (len && buf[len-1] == '\n')
2526                 len--;
2527 
2528         if (len)
2529                 ext4_error(sbi->s_sb, "%.*s", len, buf);
2530         return count;
2531 }
2532 
2533 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2534 static struct ext4_attr ext4_attr_##_name = {                   \
2535         .attr = {.name = __stringify(_name), .mode = _mode },   \
2536         .show   = _show,                                        \
2537         .store  = _store,                                       \
2538         .offset = offsetof(struct ext4_sb_info, _elname),       \
2539 }
2540 #define EXT4_ATTR(name, mode, show, store) \
2541 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2542 
2543 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2544 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2545 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2546 #define EXT4_RW_ATTR_SBI_UI(name, elname)       \
2547         EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2548 #define ATTR_LIST(name) &ext4_attr_##name.attr
2549 
2550 EXT4_RO_ATTR(delayed_allocation_blocks);
2551 EXT4_RO_ATTR(session_write_kbytes);
2552 EXT4_RO_ATTR(lifetime_write_kbytes);
2553 EXT4_RW_ATTR(reserved_clusters);
2554 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2555                  inode_readahead_blks_store, s_inode_readahead_blks);
2556 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2557 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2558 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2559 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2560 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2561 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2562 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2563 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2564 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2565 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2566 
2567 static struct attribute *ext4_attrs[] = {
2568         ATTR_LIST(delayed_allocation_blocks),
2569         ATTR_LIST(session_write_kbytes),
2570         ATTR_LIST(lifetime_write_kbytes),
2571         ATTR_LIST(reserved_clusters),
2572         ATTR_LIST(inode_readahead_blks),
2573         ATTR_LIST(inode_goal),
2574         ATTR_LIST(mb_stats),
2575         ATTR_LIST(mb_max_to_scan),
2576         ATTR_LIST(mb_min_to_scan),
2577         ATTR_LIST(mb_order2_req),
2578         ATTR_LIST(mb_stream_req),
2579         ATTR_LIST(mb_group_prealloc),
2580         ATTR_LIST(max_writeback_mb_bump),
2581         ATTR_LIST(extent_max_zeroout_kb),
2582         ATTR_LIST(trigger_fs_error),
2583         NULL,
2584 };
2585 
2586 /* Features this copy of ext4 supports */
2587 EXT4_INFO_ATTR(lazy_itable_init);
2588 EXT4_INFO_ATTR(batched_discard);
2589 EXT4_INFO_ATTR(meta_bg_resize);
2590 
2591 static struct attribute *ext4_feat_attrs[] = {
2592         ATTR_LIST(lazy_itable_init),
2593         ATTR_LIST(batched_discard),
2594         ATTR_LIST(meta_bg_resize),
2595         NULL,
2596 };
2597 
2598 static ssize_t ext4_attr_show(struct kobject *kobj,
2599                               struct attribute *attr, char *buf)
2600 {
2601         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2602                                                 s_kobj);
2603         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2604 
2605         return a->show ? a->show(a, sbi, buf) : 0;
2606 }
2607 
2608 static ssize_t ext4_attr_store(struct kobject *kobj,
2609                                struct attribute *attr,
2610                                const char *buf, size_t len)
2611 {
2612         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2613                                                 s_kobj);
2614         struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2615 
2616         return a->store ? a->store(a, sbi, buf, len) : 0;
2617 }
2618 
2619 static void ext4_sb_release(struct kobject *kobj)
2620 {
2621         struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2622                                                 s_kobj);
2623         complete(&sbi->s_kobj_unregister);
2624 }
2625 
2626 static const struct sysfs_ops ext4_attr_ops = {
2627         .show   = ext4_attr_show,
2628         .store  = ext4_attr_store,
2629 };
2630 
2631 static struct kobj_type ext4_ktype = {
2632         .default_attrs  = ext4_attrs,
2633         .sysfs_ops      = &ext4_attr_ops,
2634         .release        = ext4_sb_release,
2635 };
2636 
2637 static void ext4_feat_release(struct kobject *kobj)
2638 {
2639         complete(&ext4_feat->f_kobj_unregister);
2640 }
2641 
2642 static struct kobj_type ext4_feat_ktype = {
2643         .default_attrs  = ext4_feat_attrs,
2644         .sysfs_ops      = &ext4_attr_ops,
2645         .release        = ext4_feat_release,
2646 };
2647 
2648 /*
2649  * Check whether this filesystem can be mounted based on
2650  * the features present and the RDONLY/RDWR mount requested.
2651  * Returns 1 if this filesystem can be mounted as requested,
2652  * 0 if it cannot be.
2653  */
2654 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2655 {
2656         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2657                 ext4_msg(sb, KERN_ERR,
2658                         "Couldn't mount because of "
2659                         "unsupported optional features (%x)",
2660                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2661                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2662                 return 0;
2663         }
2664 
2665         if (readonly)
2666                 return 1;
2667 
2668         /* Check that feature set is OK for a read-write mount */
2669         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2670                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2671                          "unsupported optional features (%x)",
2672                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2673                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2674                 return 0;
2675         }
2676         /*
2677          * Large file size enabled file system can only be mounted
2678          * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2679          */
2680         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2681                 if (sizeof(blkcnt_t) < sizeof(u64)) {
2682                         ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2683                                  "cannot be mounted RDWR without "
2684                                  "CONFIG_LBDAF");
2685                         return 0;
2686                 }
2687         }
2688         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2689             !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2690                 ext4_msg(sb, KERN_ERR,
2691                          "Can't support bigalloc feature without "
2692                          "extents feature\n");
2693                 return 0;
2694         }
2695 
2696 #ifndef CONFIG_QUOTA
2697         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2698             !readonly) {
2699                 ext4_msg(sb, KERN_ERR,
2700                          "Filesystem with quota feature cannot be mounted RDWR "
2701                          "without CONFIG_QUOTA");
2702                 return 0;
2703         }
2704 #endif  /* CONFIG_QUOTA */
2705         return 1;
2706 }
2707 
2708 /*
2709  * This function is called once a day if we have errors logged
2710  * on the file system
2711  */
2712 static void print_daily_error_info(unsigned long arg)
2713 {
2714         struct super_block *sb = (struct super_block *) arg;
2715         struct ext4_sb_info *sbi;
2716         struct ext4_super_block *es;
2717 
2718         sbi = EXT4_SB(sb);
2719         es = sbi->s_es;
2720 
2721         if (es->s_error_count)
2722                 /* fsck newer than v1.41.13 is needed to clean this condition. */
2723                 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2724                          le32_to_cpu(es->s_error_count));
2725         if (es->s_first_error_time) {
2726                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2727                        sb->s_id, le32_to_cpu(es->s_first_error_time),
2728                        (int) sizeof(es->s_first_error_func),
2729                        es->s_first_error_func,
2730                        le32_to_cpu(es->s_first_error_line));
2731                 if (es->s_first_error_ino)
2732                         printk(": inode %u",
2733                                le32_to_cpu(es->s_first_error_ino));
2734                 if (es->s_first_error_block)
2735                         printk(": block %llu", (unsigned long long)
2736                                le64_to_cpu(es->s_first_error_block));
2737                 printk("\n");
2738         }
2739         if (es->s_last_error_time) {
2740                 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2741                        sb->s_id, le32_to_cpu(es->s_last_error_time),
2742                        (int) sizeof(es->s_last_error_func),
2743                        es->s_last_error_func,
2744                        le32_to_cpu(es->s_last_error_line));
2745                 if (es->s_last_error_ino)
2746                         printk(": inode %u",
2747                                le32_to_cpu(es->s_last_error_ino));
2748                 if (es->s_last_error_block)
2749                         printk(": block %llu", (unsigned long long)
2750                                le64_to_cpu(es->s_last_error_block));
2751                 printk("\n");
2752         }
2753         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2754 }
2755 
2756 /* Find next suitable group and run ext4_init_inode_table */
2757 static int ext4_run_li_request(struct ext4_li_request *elr)
2758 {
2759         struct ext4_group_desc *gdp = NULL;
2760         ext4_group_t group, ngroups;
2761         struct super_block *sb;
2762         unsigned long timeout = 0;
2763         int ret = 0;
2764 
2765         sb = elr->lr_super;
2766         ngroups = EXT4_SB(sb)->s_groups_count;
2767 
2768         sb_start_write(sb);
2769         for (group = elr->lr_next_group; group < ngroups; group++) {
2770                 gdp = ext4_get_group_desc(sb, group, NULL);
2771                 if (!gdp) {
2772                         ret = 1;
2773                         break;
2774                 }
2775 
2776                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2777                         break;
2778         }
2779 
2780         if (group >= ngroups)
2781                 ret = 1;
2782 
2783         if (!ret) {
2784                 timeout = jiffies;
2785                 ret = ext4_init_inode_table(sb, group,
2786                                             elr->lr_timeout ? 0 : 1);
2787                 if (elr->lr_timeout == 0) {
2788                         timeout = (jiffies - timeout) *
2789                                   elr->lr_sbi->s_li_wait_mult;
2790                         elr->lr_timeout = timeout;
2791                 }
2792                 elr->lr_next_sched = jiffies + elr->lr_timeout;
2793                 elr->lr_next_group = group + 1;
2794         }
2795         sb_end_write(sb);
2796 
2797         return ret;
2798 }
2799 
2800 /*
2801  * Remove lr_request from the list_request and free the
2802  * request structure. Should be called with li_list_mtx held
2803  */
2804 static void ext4_remove_li_request(struct ext4_li_request *elr)
2805 {
2806         struct ext4_sb_info *sbi;
2807 
2808         if (!elr)
2809                 return;
2810 
2811         sbi = elr->lr_sbi;
2812 
2813         list_del(&elr->lr_request);
2814         sbi->s_li_request = NULL;
2815         kfree(elr);
2816 }
2817 
2818 static void ext4_unregister_li_request(struct super_block *sb)
2819 {
2820         mutex_lock(&ext4_li_mtx);
2821         if (!ext4_li_info) {
2822                 mutex_unlock(&ext4_li_mtx);
2823                 return;
2824         }
2825 
2826         mutex_lock(&ext4_li_info->li_list_mtx);
2827         ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2828         mutex_unlock(&ext4_li_info->li_list_mtx);
2829         mutex_unlock(&ext4_li_mtx);
2830 }
2831 
2832 static struct task_struct *ext4_lazyinit_task;
2833 
2834 /*
2835  * This is the function where ext4lazyinit thread lives. It walks
2836  * through the request list searching for next scheduled filesystem.
2837  * When such a fs is found, run the lazy initialization request
2838  * (ext4_rn_li_request) and keep track of the time spend in this
2839  * function. Based on that time we compute next schedule time of
2840  * the request. When walking through the list is complete, compute
2841  * next waking time and put itself into sleep.
2842  */
2843 static int ext4_lazyinit_thread(void *arg)
2844 {
2845         struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2846         struct list_head *pos, *n;
2847         struct ext4_li_request *elr;
2848         unsigned long next_wakeup, cur;
2849 
2850         BUG_ON(NULL == eli);
2851 
2852 cont_thread:
2853         while (true) {
2854                 next_wakeup = MAX_JIFFY_OFFSET;
2855 
2856                 mutex_lock(&eli->li_list_mtx);
2857                 if (list_empty(&eli->li_request_list)) {
2858                         mutex_unlock(&eli->li_list_mtx);
2859                         goto exit_thread;
2860                 }
2861 
2862                 list_for_each_safe(pos, n, &eli->li_request_list) {
2863                         elr = list_entry(pos, struct ext4_li_request,
2864                                          lr_request);
2865 
2866                         if (time_after_eq(jiffies, elr->lr_next_sched)) {
2867                                 if (ext4_run_li_request(elr) != 0) {
2868                                         /* error, remove the lazy_init job */
2869                                         ext4_remove_li_request(elr);
2870                                         continue;
2871                                 }
2872                         }
2873 
2874                         if (time_before(elr->lr_next_sched, next_wakeup))
2875                                 next_wakeup = elr->lr_next_sched;
2876                 }
2877                 mutex_unlock(&eli->li_list_mtx);
2878 
2879                 try_to_freeze();
2880 
2881                 cur = jiffies;
2882                 if ((time_after_eq(cur, next_wakeup)) ||
2883                     (MAX_JIFFY_OFFSET == next_wakeup)) {
2884                         cond_resched();
2885                         continue;
2886                 }
2887 
2888                 schedule_timeout_interruptible(next_wakeup - cur);
2889 
2890                 if (kthread_should_stop()) {
2891                         ext4_clear_request_list();
2892                         goto exit_thread;
2893                 }
2894         }
2895 
2896 exit_thread:
2897         /*
2898          * It looks like the request list is empty, but we need
2899          * to check it under the li_list_mtx lock, to prevent any
2900          * additions into it, and of course we should lock ext4_li_mtx
2901          * to atomically free the list and ext4_li_info, because at
2902          * this point another ext4 filesystem could be registering
2903          * new one.
2904          */
2905         mutex_lock(&ext4_li_mtx);
2906         mutex_lock(&eli->li_list_mtx);
2907         if (!list_empty(&eli->li_request_list)) {
2908                 mutex_unlock(&eli->li_list_mtx);
2909                 mutex_unlock(&ext4_li_mtx);
2910                 goto cont_thread;
2911         }
2912         mutex_unlock(&eli->li_list_mtx);
2913         kfree(ext4_li_info);
2914         ext4_li_info = NULL;
2915         mutex_unlock(&ext4_li_mtx);
2916 
2917         return 0;
2918 }
2919 
2920 static void ext4_clear_request_list(void)
2921 {
2922         struct list_head *pos, *n;
2923         struct ext4_li_request *elr;
2924 
2925         mutex_lock(&ext4_li_info->li_list_mtx);
2926         list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2927                 elr = list_entry(pos, struct ext4_li_request,
2928                                  lr_request);
2929                 ext4_remove_li_request(elr);
2930         }
2931         mutex_unlock(&ext4_li_info->li_list_mtx);
2932 }
2933 
2934 static int ext4_run_lazyinit_thread(void)
2935 {
2936         ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2937                                          ext4_li_info, "ext4lazyinit");
2938         if (IS_ERR(ext4_lazyinit_task)) {
2939                 int err = PTR_ERR(ext4_lazyinit_task);
2940                 ext4_clear_request_list();
2941                 kfree(ext4_li_info);
2942                 ext4_li_info = NULL;
2943                 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2944                                  "initialization thread\n",
2945                                  err);
2946                 return err;
2947         }
2948         ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2949         return 0;
2950 }
2951 
2952 /*
2953  * Check whether it make sense to run itable init. thread or not.
2954  * If there is at least one uninitialized inode table, return
2955  * corresponding group number, else the loop goes through all
2956  * groups and return total number of groups.
2957  */
2958 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2959 {
2960         ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2961         struct ext4_group_desc *gdp = NULL;
2962 
2963         for (group = 0; group < ngroups; group++) {
2964                 gdp = ext4_get_group_desc(sb, group, NULL);
2965                 if (!gdp)
2966                         continue;
2967 
2968                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2969                         break;
2970         }
2971 
2972         return group;
2973 }
2974 
2975 static int ext4_li_info_new(void)
2976 {
2977         struct ext4_lazy_init *eli = NULL;
2978 
2979         eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2980         if (!eli)
2981                 return -ENOMEM;
2982 
2983         INIT_LIST_HEAD(&eli->li_request_list);
2984         mutex_init(&eli->li_list_mtx);
2985 
2986         eli->li_state |= EXT4_LAZYINIT_QUIT;
2987 
2988         ext4_li_info = eli;
2989 
2990         return 0;
2991 }
2992 
2993 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2994                                             ext4_group_t start)
2995 {
2996         struct ext4_sb_info *sbi = EXT4_SB(sb);
2997         struct ext4_li_request *elr;
2998         unsigned long rnd;
2999 
3000         elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3001         if (!elr)
3002                 return NULL;
3003 
3004         elr->lr_super = sb;
3005         elr->lr_sbi = sbi;
3006         elr->lr_next_group = start;
3007 
3008         /*
3009          * Randomize first schedule time of the request to
3010          * spread the inode table initialization requests
3011          * better.
3012          */
3013         get_random_bytes(&rnd, sizeof(rnd));
3014         elr->lr_next_sched = jiffies + (unsigned long)rnd %
3015                              (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3016 
3017         return elr;
3018 }
3019 
3020 int ext4_register_li_request(struct super_block *sb,
3021                              ext4_group_t first_not_zeroed)
3022 {
3023         struct ext4_sb_info *sbi = EXT4_SB(sb);
3024         struct ext4_li_request *elr = NULL;
3025         ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3026         int ret = 0;
3027 
3028         mutex_lock(&ext4_li_mtx);
3029         if (sbi->s_li_request != NULL) {
3030                 /*
3031                  * Reset timeout so it can be computed again, because
3032                  * s_li_wait_mult might have changed.
3033                  */
3034                 sbi->s_li_request->lr_timeout = 0;
3035                 goto out;
3036         }
3037 
3038         if (first_not_zeroed == ngroups ||
3039             (sb->s_flags & MS_RDONLY) ||
3040             !test_opt(sb, INIT_INODE_TABLE))
3041                 goto out;
3042 
3043         elr = ext4_li_request_new(sb, first_not_zeroed);
3044         if (!elr) {
3045                 ret = -ENOMEM;
3046                 goto out;
3047         }
3048 
3049         if (NULL == ext4_li_info) {
3050                 ret = ext4_li_info_new();
3051                 if (ret)
3052                         goto out;
3053         }
3054 
3055         mutex_lock(&ext4_li_info->li_list_mtx);
3056         list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3057         mutex_unlock(&ext4_li_info->li_list_mtx);
3058 
3059         sbi->s_li_request = elr;
3060         /*
3061          * set elr to NULL here since it has been inserted to
3062          * the request_list and the removal and free of it is
3063          * handled by ext4_clear_request_list from now on.
3064          */
3065         elr = NULL;
3066 
3067         if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3068                 ret = ext4_run_lazyinit_thread();
3069                 if (ret)
3070                         goto out;
3071         }
3072 out:
3073         mutex_unlock(&ext4_li_mtx);
3074         if (ret)
3075                 kfree(elr);
3076         return ret;
3077 }
3078 
3079 /*
3080  * We do not need to lock anything since this is called on
3081  * module unload.
3082  */
3083 static void ext4_destroy_lazyinit_thread(void)
3084 {
3085         /*
3086          * If thread exited earlier
3087          * there's nothing to be done.
3088          */
3089         if (!ext4_li_info || !ext4_lazyinit_task)
3090                 return;
3091 
3092         kthread_stop(ext4_lazyinit_task);
3093 }
3094 
3095 static int set_journal_csum_feature_set(struct super_block *sb)
3096 {
3097         int ret = 1;
3098         int compat, incompat;
3099         struct ext4_sb_info *sbi = EXT4_SB(sb);
3100 
3101         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3102                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3103                 /* journal checksum v2 */
3104                 compat = 0;
3105                 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3106         } else {
3107                 /* journal checksum v1 */
3108                 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3109                 incompat = 0;
3110         }
3111 
3112         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3113                 ret = jbd2_journal_set_features(sbi->s_journal,
3114                                 compat, 0,
3115                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3116                                 incompat);
3117         } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3118                 ret = jbd2_journal_set_features(sbi->s_journal,
3119                                 compat, 0,
3120                                 incompat);
3121                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3122                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3123         } else {
3124                 jbd2_journal_clear_features(sbi->s_journal,
3125                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3126                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3127                                 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3128         }
3129 
3130         return ret;
3131 }
3132 
3133 /*
3134  * Note: calculating the overhead so we can be compatible with
3135  * historical BSD practice is quite difficult in the face of
3136  * clusters/bigalloc.  This is because multiple metadata blocks from
3137  * different block group can end up in the same allocation cluster.
3138  * Calculating the exact overhead in the face of clustered allocation
3139  * requires either O(all block bitmaps) in memory or O(number of block
3140  * groups**2) in time.  We will still calculate the superblock for
3141  * older file systems --- and if we come across with a bigalloc file
3142  * system with zero in s_overhead_clusters the estimate will be close to
3143  * correct especially for very large cluster sizes --- but for newer
3144  * file systems, it's better to calculate this figure once at mkfs
3145  * time, and store it in the superblock.  If the superblock value is
3146  * present (even for non-bigalloc file systems), we will use it.
3147  */
3148 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3149                           char *buf)
3150 {
3151         struct ext4_sb_info     *sbi = EXT4_SB(sb);
3152         struct ext4_group_desc  *gdp;
3153         ext4_fsblk_t            first_block, last_block, b;
3154         ext4_group_t            i, ngroups = ext4_get_groups_count(sb);
3155         int                     s, j, count = 0;
3156 
3157         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3158                 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3159                         sbi->s_itb_per_group + 2);
3160 
3161         first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3162                 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3163         last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3164         for (i = 0; i < ngroups; i++) {
3165                 gdp = ext4_get_group_desc(sb, i, NULL);
3166                 b = ext4_block_bitmap(sb, gdp);
3167                 if (b >= first_block && b <= last_block) {
3168                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3169                         count++;
3170                 }
3171                 b = ext4_inode_bitmap(sb, gdp);
3172                 if (b >= first_block && b <= last_block) {
3173                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3174                         count++;
3175                 }
3176                 b = ext4_inode_table(sb, gdp);
3177                 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3178                         for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3179                                 int c = EXT4_B2C(sbi, b - first_block);
3180                                 ext4_set_bit(c, buf);
3181                                 count++;
3182                         }
3183                 if (i != grp)
3184                         continue;
3185                 s = 0;
3186                 if (ext4_bg_has_super(sb, grp)) {
3187                         ext4_set_bit(s++, buf);
3188                         count++;
3189                 }
3190                 j = ext4_bg_num_gdb(sb, grp);
3191                 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3192                         ext4_error(sb, "Invalid number of block group "
3193                                    "descriptor blocks: %d", j);
3194                         j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3195                 }
3196                 count += j;
3197                 for (; j > 0; j--)
3198                         ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3199         }
3200         if (!count)
3201                 return 0;
3202         return EXT4_CLUSTERS_PER_GROUP(sb) -
3203                 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3204 }
3205 
3206 /*
3207  * Compute the overhead and stash it in sbi->s_overhead
3208  */
3209 int ext4_calculate_overhead(struct super_block *sb)
3210 {
3211         struct ext4_sb_info *sbi = EXT4_SB(sb);
3212         struct ext4_super_block *es = sbi->s_es;
3213         ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3214         ext4_fsblk_t overhead = 0;
3215         char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3216 
3217         if (!buf)
3218                 return -ENOMEM;
3219 
3220         /*
3221          * Compute the overhead (FS structures).  This is constant
3222          * for a given filesystem unless the number of block groups
3223          * changes so we cache the previous value until it does.
3224          */
3225 
3226         /*
3227          * All of the blocks before first_data_block are overhead
3228          */
3229         overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3230 
3231         /*
3232          * Add the overhead found in each block group
3233          */
3234         for (i = 0; i < ngroups; i++) {
3235                 int blks;
3236 
3237                 blks = count_overhead(sb, i, buf);
3238                 overhead += blks;
3239                 if (blks)
3240                         memset(buf, 0, PAGE_SIZE);
3241                 cond_resched();
3242         }
3243         /* Add the journal blocks as well */
3244         if (sbi->s_journal)
3245                 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3246 
3247         sbi->s_overhead = overhead;
3248         smp_wmb();
3249         free_page((unsigned long) buf);
3250         return 0;
3251 }
3252 
3253 
3254 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3255 {
3256         ext4_fsblk_t resv_clusters;
3257 
3258         /*
3259          * There's no need to reserve anything when we aren't using extents.
3260          * The space estimates are exact, there are no unwritten extents,
3261          * hole punching doesn't need new metadata... This is needed especially
3262          * to keep ext2/3 backward compatibility.
3263          */
3264         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3265                 return 0;
3266         /*
3267          * By default we reserve 2% or 4096 clusters, whichever is smaller.
3268          * This should cover the situations where we can not afford to run
3269          * out of space like for example punch hole, or converting
3270          * uninitialized extents in delalloc path. In most cases such
3271          * allocation would require 1, or 2 blocks, higher numbers are
3272          * very rare.
3273          */
3274         resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3275                         EXT4_SB(sb)->s_cluster_bits;
3276 
3277         do_div(resv_clusters, 50);
3278         resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3279 
3280         return resv_clusters;
3281 }
3282 
3283 
3284 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3285 {
3286         ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3287                                 sbi->s_cluster_bits;
3288 
3289         if (count >= clusters)
3290                 return -EINVAL;
3291 
3292         atomic64_set(&sbi->s_resv_clusters, count);
3293         return 0;
3294 }
3295 
3296 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3297 {
3298         char *orig_data = kstrdup(data, GFP_KERNEL);
3299         struct buffer_head *bh;
3300         struct ext4_super_block *es = NULL;
3301         struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3302         ext4_fsblk_t block;
3303         ext4_fsblk_t sb_block = get_sb_block(&data);
3304         ext4_fsblk_t logical_sb_block;
3305         unsigned long offset = 0;
3306         unsigned long journal_devnum = 0;
3307         unsigned long def_mount_opts;
3308         struct inode *root;
3309         char *cp;
3310         const char *descr;
3311         int ret = -ENOMEM;
3312         int blocksize, clustersize;
3313         unsigned int db_count;
3314         unsigned int i;
3315         int needs_recovery, has_huge_files, has_bigalloc;
3316         __u64 blocks_count;
3317         int err = 0;
3318         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3319         ext4_group_t first_not_zeroed;
3320 
3321         if ((data && !orig_data) || !sbi)
3322                 goto out_free_base;
3323 
3324         sbi->s_blockgroup_lock =
3325                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3326         if (!sbi->s_blockgroup_lock)
3327                 goto out_free_base;
3328 
3329         sb->s_fs_info = sbi;
3330         sbi->s_sb = sb;
3331         sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3332         sbi->s_sb_block = sb_block;
3333         if (sb->s_bdev->bd_part)
3334                 sbi->s_sectors_written_start =
3335                         part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3336 
3337         /* Cleanup superblock name */
3338         for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3339                 *cp = '!';
3340 
3341         /* -EINVAL is default */
3342         ret = -EINVAL;
3343         blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3344         if (!blocksize) {
3345                 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3346                 goto out_fail;
3347         }
3348 
3349         /*
3350          * The ext4 superblock will not be buffer aligned for other than 1kB
3351          * block sizes.  We need to calculate the offset from buffer start.
3352          */
3353         if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3354                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3355                 offset = do_div(logical_sb_block, blocksize);
3356         } else {
3357                 logical_sb_block = sb_block;
3358         }
3359 
3360         if (!(bh = sb_bread(sb, logical_sb_block))) {
3361                 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3362                 goto out_fail;
3363         }
3364         /*
3365          * Note: s_es must be initialized as soon as possible because
3366          *       some ext4 macro-instructions depend on its value
3367          */
3368         es = (struct ext4_super_block *) (bh->b_data + offset);
3369         sbi->s_es = es;
3370         sb->s_magic = le16_to_cpu(es->s_magic);
3371         if (sb->s_magic != EXT4_SUPER_MAGIC)
3372                 goto cantfind_ext4;
3373         sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3374 
3375         /* Warn if metadata_csum and gdt_csum are both set. */
3376         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3377                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3378             EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3379                 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3380                              "redundant flags; please run fsck.");
3381 
3382         /* Check for a known checksum algorithm */
3383         if (!ext4_verify_csum_type(sb, es)) {
3384                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3385                          "unknown checksum algorithm.");
3386                 silent = 1;
3387                 goto cantfind_ext4;
3388         }
3389 
3390         /* Load the checksum driver */
3391         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3392                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3393                 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3394                 if (IS_ERR(sbi->s_chksum_driver)) {
3395                         ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3396                         ret = PTR_ERR(sbi->s_chksum_driver);
3397                         sbi->s_chksum_driver = NULL;
3398                         goto failed_mount;
3399                 }
3400         }
3401 
3402         /* Check superblock checksum */
3403         if (!ext4_superblock_csum_verify(sb, es)) {
3404                 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3405                          "invalid superblock checksum.  Run e2fsck?");
3406                 silent = 1;
3407                 goto cantfind_ext4;
3408         }
3409 
3410         /* Precompute checksum seed for all metadata */
3411         if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3412                         EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3413                 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3414                                                sizeof(es->s_uuid));
3415 
3416         /* Set defaults before we parse the mount options */
3417         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3418         set_opt(sb, INIT_INODE_TABLE);
3419         if (def_mount_opts & EXT4_DEFM_DEBUG)
3420                 set_opt(sb, DEBUG);
3421         if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3422                 set_opt(sb, GRPID);
3423         if (def_mount_opts & EXT4_DEFM_UID16)
3424                 set_opt(sb, NO_UID32);
3425         /* xattr user namespace & acls are now defaulted on */
3426         set_opt(sb, XATTR_USER);
3427 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3428         set_opt(sb, POSIX_ACL);
3429 #endif
3430         if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3431                 set_opt(sb, JOURNAL_DATA);
3432         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3433                 set_opt(sb, ORDERED_DATA);
3434         else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3435                 set_opt(sb, WRITEBACK_DATA);
3436 
3437         if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3438                 set_opt(sb, ERRORS_PANIC);
3439         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3440                 set_opt(sb, ERRORS_CONT);
3441         else
3442                 set_opt(sb, ERRORS_RO);
3443         if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3444                 set_opt(sb, BLOCK_VALIDITY);
3445         if (def_mount_opts & EXT4_DEFM_DISCARD)
3446                 set_opt(sb, DISCARD);
3447 
3448         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3449         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3450         sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3451         sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3452         sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3453 
3454         if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3455                 set_opt(sb, BARRIER);
3456 
3457         /*
3458          * enable delayed allocation by default
3459          * Use -o nodelalloc to turn it off
3460          */
3461         if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3462             ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3463                 set_opt(sb, DELALLOC);
3464 
3465         /*
3466          * set default s_li_wait_mult for lazyinit, for the case there is
3467          * no mount option specified.
3468          */
3469         sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3470 
3471         if (sbi->s_es->s_mount_opts[0]) {
3472                 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3473                                               sizeof(sbi->s_es->s_mount_opts),
3474                                               GFP_KERNEL);
3475                 if (!s_mount_opts)
3476                         goto failed_mount;
3477                 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3478                                    &journal_ioprio, 0)) {
3479                         ext4_msg(sb, KERN_WARNING,
3480                                  "failed to parse options in superblock: %s",
3481                                  s_mount_opts);
3482                 }
3483                 kfree(s_mount_opts);
3484         }
3485         sbi->s_def_mount_opt = sbi->s_mount_opt;
3486         if (!parse_options((char *) data, sb, &journal_devnum,
3487                            &journal_ioprio, 0))
3488                 goto failed_mount;
3489 
3490         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3491                 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3492                             "with data=journal disables delayed "
3493                             "allocation and O_DIRECT support!\n");
3494                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3495                         ext4_msg(sb, KERN_ERR, "can't mount with "
3496                                  "both data=journal and delalloc");
3497                         goto failed_mount;
3498                 }
3499                 if (test_opt(sb, DIOREAD_NOLOCK)) {
3500                         ext4_msg(sb, KERN_ERR, "can't mount with "
3501                                  "both data=journal and dioread_nolock");
3502                         goto failed_mount;
3503                 }
3504                 if (test_opt(sb, DELALLOC))
3505                         clear_opt(sb, DELALLOC);
3506         }
3507 
3508         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3509                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3510 
3511         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3512             (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3513              EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3514              EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3515                 ext4_msg(sb, KERN_WARNING,
3516                        "feature flags set on rev 0 fs, "
3517                        "running e2fsck is recommended");
3518 
3519         if (IS_EXT2_SB(sb)) {
3520                 if (ext2_feature_set_ok(sb))
3521                         ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3522                                  "using the ext4 subsystem");
3523                 else {
3524                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3525                                  "to feature incompatibilities");
3526                         goto failed_mount;
3527                 }
3528         }
3529 
3530         if (IS_EXT3_SB(sb)) {
3531                 if (ext3_feature_set_ok(sb))
3532                         ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3533                                  "using the ext4 subsystem");
3534                 else {
3535                         ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3536                                  "to feature incompatibilities");
3537                         goto failed_mount;
3538                 }
3539         }
3540 
3541         /*
3542          * Check feature flags regardless of the revision level, since we
3543          * previously didn't change the revision level when setting the flags,
3544          * so there is a chance incompat flags are set on a rev 0 filesystem.
3545          */
3546         if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3547                 goto failed_mount;
3548 
3549         blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3550         if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3551             blocksize > EXT4_MAX_BLOCK_SIZE) {
3552                 ext4_msg(sb, KERN_ERR,
3553                        "Unsupported filesystem blocksize %d (%d log_block_size)",
3554                          blocksize, le32_to_cpu(es->s_log_block_size));
3555                 goto failed_mount;
3556         }
3557         if (le32_to_cpu(es->s_log_block_size) >
3558             (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3559                 ext4_msg(sb, KERN_ERR,
3560                          "Invalid log block size: %u",
3561                          le32_to_cpu(es->s_log_block_size));
3562                 goto failed_mount;
3563         }
3564 
3565         if (sb->s_blocksize != blocksize) {
3566                 /* Validate the filesystem blocksize */
3567                 if (!sb_set_blocksize(sb, blocksize)) {
3568                         ext4_msg(sb, KERN_ERR, "bad block size %d",
3569                                         blocksize);
3570                         goto failed_mount;
3571                 }
3572 
3573                 brelse(bh);
3574                 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3575                 offset = do_div(logical_sb_block, blocksize);
3576                 bh = sb_bread(sb, logical_sb_block);
3577                 if (!bh) {
3578                         ext4_msg(sb, KERN_ERR,
3579                                "Can't read superblock on 2nd try");
3580                         goto failed_mount;
3581                 }
3582                 es = (struct ext4_super_block *)(bh->b_data + offset);
3583                 sbi->s_es = es;
3584                 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3585                         ext4_msg(sb, KERN_ERR,
3586                                "Magic mismatch, very weird!");
3587                         goto failed_mount;
3588                 }
3589         }
3590 
3591         has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3592                                 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3593         sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3594                                                       has_huge_files);
3595         sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3596 
3597         if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3598                 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3599                 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3600         } else {
3601                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3602                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3603                 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3604                     (!is_power_of_2(sbi->s_inode_size)) ||
3605                     (sbi->s_inode_size > blocksize)) {
3606                         ext4_msg(sb, KERN_ERR,
3607                                "unsupported inode size: %d",
3608                                sbi->s_inode_size);
3609                         goto failed_mount;
3610                 }
3611                 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3612                         sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3613         }
3614 
3615         sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3616         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3617                 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3618                     sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3619                     !is_power_of_2(sbi->s_desc_size)) {
3620                         ext4_msg(sb, KERN_ERR,
3621                                "unsupported descriptor size %lu",
3622                                sbi->s_desc_size);
3623                         goto failed_mount;
3624                 }
3625         } else
3626                 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3627 
3628         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3629         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3630 
3631         sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3632         if (sbi->s_inodes_per_block == 0)
3633                 goto cantfind_ext4;
3634         if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3635             sbi->s_inodes_per_group > blocksize * 8) {
3636                 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3637                          sbi->s_blocks_per_group);
3638                 goto failed_mount;
3639         }
3640         sbi->s_itb_per_group = sbi->s_inodes_per_group /
3641                                         sbi->s_inodes_per_block;
3642         sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3643         sbi->s_sbh = bh;
3644         sbi->s_mount_state = le16_to_cpu(es->s_state);
3645         sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3646         sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3647 
3648         for (i = 0; i < 4; i++)
3649                 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3650         sbi->s_def_hash_version = es->s_def_hash_version;
3651         if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3652                 i = le32_to_cpu(es->s_flags);
3653                 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3654                         sbi->s_hash_unsigned = 3;
3655                 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3656 #ifdef __CHAR_UNSIGNED__
3657                         if (!(sb->s_flags & MS_RDONLY))
3658                                 es->s_flags |=
3659                                         cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3660                         sbi->s_hash_unsigned = 3;
3661 #else
3662                         if (!(sb->s_flags & MS_RDONLY))
3663                                 es->s_flags |=
3664                                         cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3665 #endif
3666                 }
3667         }
3668 
3669         /* Handle clustersize */
3670         clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3671         has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3672                                 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3673         if (has_bigalloc) {
3674                 if (clustersize < blocksize) {
3675                         ext4_msg(sb, KERN_ERR,
3676                                  "cluster size (%d) smaller than "
3677                                  "block size (%d)", clustersize, blocksize);
3678                         goto failed_mount;
3679                 }
3680                 if (le32_to_cpu(es->s_log_cluster_size) >
3681                     (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3682                         ext4_msg(sb, KERN_ERR,
3683                                  "Invalid log cluster size: %u",
3684                                  le32_to_cpu(es->s_log_cluster_size));
3685                         goto failed_mount;
3686                 }
3687                 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3688                         le32_to_cpu(es->s_log_block_size);
3689                 sbi->s_clusters_per_group =
3690                         le32_to_cpu(es->s_clusters_per_group);
3691                 if (sbi->s_clusters_per_group > blocksize * 8) {
3692                         ext4_msg(sb, KERN_ERR,
3693                                  "#clusters per group too big: %lu",
3694                                  sbi->s_clusters_per_group);
3695                         goto failed_mount;
3696                 }
3697                 if (sbi->s_blocks_per_group !=
3698                     (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3699                         ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3700                                  "clusters per group (%lu) inconsistent",
3701                                  sbi->s_blocks_per_group,
3702                                  sbi->s_clusters_per_group);
3703                         goto failed_mount;
3704                 }
3705         } else {
3706                 if (clustersize != blocksize) {
3707                         ext4_warning(sb, "fragment/cluster size (%d) != "
3708                                      "block size (%d)", clustersize,
3709                                      blocksize);
3710                         clustersize = blocksize;
3711                 }
3712                 if (sbi->s_blocks_per_group > blocksize * 8) {
3713                         ext4_msg(sb, KERN_ERR,
3714                                  "#blocks per group too big: %lu",
3715                                  sbi->s_blocks_per_group);
3716                         goto failed_mount;
3717                 }
3718                 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3719                 sbi->s_cluster_bits = 0;
3720         }
3721         sbi->s_cluster_ratio = clustersize / blocksize;
3722 
3723         /* Do we have standard group size of clustersize * 8 blocks ? */
3724         if (sbi->s_blocks_per_group == clustersize << 3)
3725                 set_opt2(sb, STD_GROUP_SIZE);
3726 
3727         /*
3728          * Test whether we have more sectors than will fit in sector_t,
3729          * and whether the max offset is addressable by the page cache.
3730          */
3731         err = generic_check_addressable(sb->s_blocksize_bits,
3732                                         ext4_blocks_count(es));
3733         if (err) {
3734                 ext4_msg(sb, KERN_ERR, "filesystem"
3735                          " too large to mount safely on this system");
3736                 if (sizeof(sector_t) < 8)
3737                         ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3738                 goto failed_mount;
3739         }
3740 
3741         if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3742                 goto cantfind_ext4;
3743 
3744         /* check blocks count against device size */
3745         blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3746         if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3747                 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3748                        "exceeds size of device (%llu blocks)",
3749                        ext4_blocks_count(es), blocks_count);
3750                 goto failed_mount;
3751         }
3752 
3753         /*
3754          * It makes no sense for the first data block to be beyond the end
3755          * of the filesystem.
3756          */
3757         if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3758                 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3759                          "block %u is beyond end of filesystem (%llu)",
3760                          le32_to_cpu(es->s_first_data_block),
3761                          ext4_blocks_count(es));
3762                 goto failed_mount;
3763         }
3764         blocks_count = (ext4_blocks_count(es) -
3765                         le32_to_cpu(es->s_first_data_block) +
3766                         EXT4_BLOCKS_PER_GROUP(sb) - 1);
3767         do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3768         if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3769                 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3770                        "(block count %llu, first data block %u, "
3771                        "blocks per group %lu)", sbi->s_groups_count,
3772                        ext4_blocks_count(es),
3773                        le32_to_cpu(es->s_first_data_block),
3774                        EXT4_BLOCKS_PER_GROUP(sb));
3775                 goto failed_mount;
3776         }
3777         sbi->s_groups_count = blocks_count;
3778         sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3779                         (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3780         db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3781                    EXT4_DESC_PER_BLOCK(sb);
3782         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG)) {
3783                 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3784                         ext4_msg(sb, KERN_WARNING,
3785                                  "first meta block group too large: %u "
3786                                  "(group descriptor block count %u)",
3787                                  le32_to_cpu(es->s_first_meta_bg), db_count);
3788                         goto failed_mount;
3789                 }
3790         }
3791         sbi->s_group_desc = ext4_kvmalloc(db_count *
3792                                           sizeof(struct buffer_head *),
3793                                           GFP_KERNEL);
3794         if (sbi->s_group_desc == NULL) {
3795                 ext4_msg(sb, KERN_ERR, "not enough memory");
3796                 ret = -ENOMEM;
3797                 goto failed_mount;
3798         }
3799 
3800         if (ext4_proc_root)
3801                 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3802 
3803         if (sbi->s_proc)
3804                 proc_create_data("options", S_IRUGO, sbi->s_proc,
3805                                  &ext4_seq_options_fops, sb);
3806 
3807         bgl_lock_init(sbi->s_blockgroup_lock);
3808 
3809         for (i = 0; i < db_count; i++) {
3810                 block = descriptor_loc(sb, logical_sb_block, i);
3811                 sbi->s_group_desc[i] = sb_bread(sb, block);
3812                 if (!sbi->s_group_desc[i]) {
3813                         ext4_msg(sb, KERN_ERR,
3814                                "can't read group descriptor %d", i);
3815                         db_count = i;
3816                         goto failed_mount2;
3817                 }
3818         }
3819         if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3820                 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3821                 goto failed_mount2;
3822         }
3823         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3824                 if (!ext4_fill_flex_info(sb)) {
3825                         ext4_msg(sb, KERN_ERR,
3826                                "unable to initialize "
3827                                "flex_bg meta info!");
3828                         goto failed_mount2;
3829                 }
3830 
3831         sbi->s_gdb_count = db_count;
3832         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3833         spin_lock_init(&sbi->s_next_gen_lock);
3834 
3835         init_timer(&sbi->s_err_report);
3836         sbi->s_err_report.function = print_daily_error_info;
3837         sbi->s_err_report.data = (unsigned long) sb;
3838 
3839         /* Register extent status tree shrinker */
3840         ext4_es_register_shrinker(sb);
3841 
3842         err = percpu_counter_init(&sbi->s_freeclusters_counter,
3843                         ext4_count_free_clusters(sb));
3844         if (!err) {
3845                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3846                                 ext4_count_free_inodes(sb));
3847         }
3848         if (!err) {
3849                 err = percpu_counter_init(&sbi->s_dirs_counter,
3850                                 ext4_count_dirs(sb));
3851         }
3852         if (!err) {
3853                 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3854         }
3855         if (!err) {
3856                 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3857         }
3858         if (err) {
3859                 ext4_msg(sb, KERN_ERR, "insufficient memory");
3860                 goto failed_mount3;
3861         }
3862 
3863         sbi->s_stripe = ext4_get_stripe_size(sbi);
3864         sbi->s_max_writeback_mb_bump = 128;
3865         sbi->s_extent_max_zeroout_kb = 32;
3866 
3867         /*
3868          * set up enough so that it can read an inode
3869          */
3870         if (!test_opt(sb, NOLOAD) &&
3871             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3872                 sb->s_op = &ext4_sops;
3873         else
3874                 sb->s_op = &ext4_nojournal_sops;
3875         sb->s_export_op = &ext4_export_ops;
3876         sb->s_xattr = ext4_xattr_handlers;
3877 #ifdef CONFIG_QUOTA
3878         sb->dq_op = &ext4_quota_operations;
3879         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3880                 sb->s_qcop = &ext4_qctl_sysfile_operations;
3881         else
3882                 sb->s_qcop = &ext4_qctl_operations;
3883 #endif
3884         memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3885 
3886         INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3887         mutex_init(&sbi->s_orphan_lock);
3888 
3889         sb->s_root = NULL;
3890 
3891         needs_recovery = (es->s_last_orphan != 0 ||
3892                           EXT4_HAS_INCOMPAT_FEATURE(sb,
3893                                     EXT4_FEATURE_INCOMPAT_RECOVER));
3894 
3895         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3896             !(sb->s_flags & MS_RDONLY))
3897                 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3898                         goto failed_mount3;
3899 
3900         /*
3901          * The first inode we look at is the journal inode.  Don't try
3902          * root first: it may be modified in the journal!
3903          */
3904         if (!test_opt(sb, NOLOAD) &&
3905             EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3906                 err = ext4_load_journal(sb, es, journal_devnum);
3907                 if (err)
3908                         goto failed_mount3;
3909         } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3910               EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3911                 ext4_msg(sb, KERN_ERR, "required journal recovery "
3912                        "suppressed and not mounted read-only");
3913                 goto failed_mount_wq;
3914         } else {
3915                 clear_opt(sb, DATA_FLAGS);
3916                 sbi->s_journal = NULL;
3917                 needs_recovery = 0;
3918                 goto no_journal;
3919         }
3920 
3921         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3922             !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3923                                        JBD2_FEATURE_INCOMPAT_64BIT)) {
3924                 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3925                 goto failed_mount_wq;
3926         }
3927 
3928         if (!set_journal_csum_feature_set(sb)) {
3929                 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3930                          "feature set");
3931                 goto failed_mount_wq;
3932         }
3933 
3934         /* We have now updated the journal if required, so we can
3935          * validate the data journaling mode. */
3936         switch (test_opt(sb, DATA_FLAGS)) {
3937         case 0:
3938                 /* No mode set, assume a default based on the journal
3939                  * capabilities: ORDERED_DATA if the journal can
3940                  * cope, else JOURNAL_DATA
3941                  */
3942                 if (jbd2_journal_check_available_features
3943                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3944                         set_opt(sb, ORDERED_DATA);
3945                 else
3946                         set_opt(sb, JOURNAL_DATA);
3947                 break;
3948 
3949         case EXT4_MOUNT_ORDERED_DATA:
3950         case EXT4_MOUNT_WRITEBACK_DATA:
3951                 if (!jbd2_journal_check_available_features
3952                     (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3953                         ext4_msg(sb, KERN_ERR, "Journal does not support "
3954                                "requested data journaling mode");
3955                         goto failed_mount_wq;
3956                 }
3957         default:
3958                 break;
3959         }
3960         set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3961 
3962         sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3963 
3964         /*
3965          * The journal may have updated the bg summary counts, so we
3966          * need to update the global counters.
3967          */
3968         percpu_counter_set(&sbi->s_freeclusters_counter,
3969                            ext4_count_free_clusters(sb));
3970         percpu_counter_set(&sbi->s_freeinodes_counter,
3971                            ext4_count_free_inodes(sb));
3972         percpu_counter_set(&sbi->s_dirs_counter,
3973                            ext4_count_dirs(sb));
3974         percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3975 
3976 no_journal:
3977         /*
3978          * Get the # of file system overhead blocks from the
3979          * superblock if present.
3980          */
3981         if (es->s_overhead_clusters)
3982                 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3983         else {
3984                 err = ext4_calculate_overhead(sb);
3985                 if (err)
3986                         goto failed_mount_wq;
3987         }
3988 
3989         /*
3990          * The maximum number of concurrent works can be high and
3991          * concurrency isn't really necessary.  Limit it to 1.
3992          */
3993         EXT4_SB(sb)->dio_unwritten_wq =
3994                 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3995         if (!EXT4_SB(sb)->dio_unwritten_wq) {
3996                 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3997                 ret = -ENOMEM;
3998                 goto failed_mount_wq;
3999         }
4000 
4001         /*
4002          * The jbd2_journal_load will have done any necessary log recovery,
4003          * so we can safely mount the rest of the filesystem now.
4004          */
4005 
4006         root = ext4_iget(sb, EXT4_ROOT_INO);
4007         if (IS_ERR(root)) {
4008                 ext4_msg(sb, KERN_ERR, "get root inode failed");
4009                 ret = PTR_ERR(root);
4010                 root = NULL;
4011                 goto failed_mount4;
4012         }
4013         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4014                 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4015                 iput(root);
4016                 goto failed_mount4;
4017         }
4018         sb->s_root = d_make_root(root);
4019         if (!sb->s_root) {
4020                 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4021                 ret = -ENOMEM;
4022                 goto failed_mount4;
4023         }
4024 
4025         if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4026                 sb->s_flags |= MS_RDONLY;
4027 
4028         /* determine the minimum size of new large inodes, if present */
4029         if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4030                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4031                                                      EXT4_GOOD_OLD_INODE_SIZE;
4032                 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4033                                        EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4034                         if (sbi->s_want_extra_isize <
4035                             le16_to_cpu(es->s_want_extra_isize))
4036                                 sbi->s_want_extra_isize =
4037                                         le16_to_cpu(es->s_want_extra_isize);
4038                         if (sbi->s_want_extra_isize <
4039                             le16_to_cpu(es->s_min_extra_isize))
4040                                 sbi->s_want_extra_isize =
4041                                         le16_to_cpu(es->s_min_extra_isize);
4042                 }
4043         }
4044         /* Check if enough inode space is available */
4045         if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4046                                                         sbi->s_inode_size) {
4047                 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4048                                                        EXT4_GOOD_OLD_INODE_SIZE;
4049                 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4050                          "available");
4051         }
4052 
4053         err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4054         if (err) {
4055                 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4056                          "reserved pool", ext4_calculate_resv_clusters(sb));
4057                 goto failed_mount4a;
4058         }
4059 
4060         err = ext4_setup_system_zone(sb);
4061         if (err) {
4062                 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4063                          "zone (%d)", err);
4064                 goto failed_mount4a;
4065         }
4066 
4067         ext4_ext_init(sb);
4068         err = ext4_mb_init(sb);
4069         if (err) {
4070                 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4071                          err);
4072                 goto failed_mount5;
4073         }
4074 
4075         err = ext4_register_li_request(sb, first_not_zeroed);
4076         if (err)
4077                 goto failed_mount6;
4078 
4079         sbi->s_kobj.kset = ext4_kset;
4080         init_completion(&sbi->s_kobj_unregister);
4081         err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4082                                    "%s", sb->s_id);
4083         if (err)
4084                 goto failed_mount7;
4085 
4086 #ifdef CONFIG_QUOTA
4087         /* Enable quota usage during mount. */
4088         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4089             !(sb->s_flags & MS_RDONLY)) {
4090                 err = ext4_enable_quotas(sb);
4091                 if (err)
4092                         goto failed_mount8;
4093         }
4094 #endif  /* CONFIG_QUOTA */
4095 
4096         EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4097         ext4_orphan_cleanup(sb, es);
4098         EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4099         if (needs_recovery) {
4100                 ext4_msg(sb, KERN_INFO, "recovery complete");
4101                 ext4_mark_recovery_complete(sb, es);
4102         }
4103         if (EXT4_SB(sb)->s_journal) {
4104                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4105                         descr = " journalled data mode";
4106                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4107                         descr = " ordered data mode";
4108                 else
4109                         descr = " writeback data mode";
4110         } else
4111                 descr = "out journal";
4112 
4113         if (test_opt(sb, DISCARD)) {
4114                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4115                 if (!blk_queue_discard(q))
4116                         ext4_msg(sb, KERN_WARNING,
4117                                  "mounting with \"discard\" option, but "
4118                                  "the device does not support discard");
4119         }
4120 
4121         ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4122                  "Opts: %.*s%s%s", descr,
4123                  (int) sizeof(sbi->s_es->s_mount_opts),
4124                  sbi->s_es->s_mount_opts,
4125                  *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4126 
4127         if (es->s_error_count)
4128                 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4129 
4130         kfree(orig_data);
4131         return 0;
4132 
4133 cantfind_ext4:
4134         if (!silent)
4135                 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4136         goto failed_mount;
4137 
4138 #ifdef CONFIG_QUOTA
4139 failed_mount8:
4140         kobject_del(&sbi->s_kobj);
4141 #endif
4142 failed_mount7:
4143         ext4_unregister_li_request(sb);
4144 failed_mount6:
4145         ext4_mb_release(sb);
4146 failed_mount5:
4147         ext4_ext_release(sb);
4148         ext4_release_system_zone(sb);
4149 failed_mount4a:
4150         dput(sb->s_root);
4151         sb->s_root = NULL;
4152 failed_mount4:
4153         ext4_msg(sb, KERN_ERR, "mount failed");
4154         destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4155 failed_mount_wq:
4156         if (sbi->s_journal) {
4157                 jbd2_journal_destroy(sbi->s_journal);
4158                 sbi->s_journal = NULL;
4159         }
4160 failed_mount3:
4161         ext4_es_unregister_shrinker(sb);
4162         del_timer(&sbi->s_err_report);
4163         if (sbi->s_flex_groups)
4164                 ext4_kvfree(sbi->s_flex_groups);
4165         percpu_counter_destroy(&sbi->s_freeclusters_counter);
4166         percpu_counter_destroy(&sbi->s_freeinodes_counter);
4167         percpu_counter_destroy(&sbi->s_dirs_counter);
4168         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4169         percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4170         if (sbi->s_mmp_tsk)
4171                 kthread_stop(sbi->s_mmp_tsk);
4172 failed_mount2:
4173         for (i = 0; i < db_count; i++)
4174                 brelse(sbi->s_group_desc[i]);
4175         ext4_kvfree(sbi->s_group_desc);
4176 failed_mount:
4177         if (sbi->s_chksum_driver)
4178                 crypto_free_shash(sbi->s_chksum_driver);
4179         if (sbi->s_proc) {
4180                 remove_proc_entry("options", sbi->s_proc);
4181                 remove_proc_entry(sb->s_id, ext4_proc_root);
4182         }
4183 #ifdef CONFIG_QUOTA
4184         for (i = 0; i < MAXQUOTAS; i++)
4185                 kfree(sbi->s_qf_names[i]);
4186 #endif
4187         ext4_blkdev_remove(sbi);
4188         brelse(bh);
4189 out_fail:
4190         sb->s_fs_info = NULL;
4191         kfree(sbi->s_blockgroup_lock);
4192 out_free_base:
4193         kfree(sbi);
4194         kfree(orig_data);
4195         return err ? err : ret;
4196 }
4197 
4198 /*
4199  * Setup any per-fs journal parameters now.  We'll do this both on
4200  * initial mount, once the journal has been initialised but before we've
4201  * done any recovery; and again on any subsequent remount.
4202  */
4203 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4204 {
4205         struct ext4_sb_info *sbi = EXT4_SB(sb);
4206 
4207         journal->j_commit_interval = sbi->s_commit_interval;
4208         journal->j_min_batch_time = sbi->s_min_batch_time;
4209         journal->j_max_batch_time = sbi->s_max_batch_time;
4210 
4211         write_lock(&journal->j_state_lock);
4212         if (test_opt(sb, BARRIER))
4213                 journal->j_flags |= JBD2_BARRIER;
4214         else
4215                 journal->j_flags &= ~JBD2_BARRIER;
4216         if (test_opt(sb, DATA_ERR_ABORT))
4217                 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4218         else
4219                 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4220         write_unlock(&journal->j_state_lock);
4221 }
4222 
4223 static journal_t *ext4_get_journal(struct super_block *sb,
4224                                    unsigned int journal_inum)
4225 {
4226         struct inode *journal_inode;
4227         journal_t *journal;
4228 
4229         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4230 
4231         /* First, test for the existence of a valid inode on disk.  Bad
4232          * things happen if we iget() an unused inode, as the subsequent
4233          * iput() will try to delete it. */
4234 
4235         journal_inode = ext4_iget(sb, journal_inum);
4236         if (IS_ERR(journal_inode)) {
4237                 ext4_msg(sb, KERN_ERR, "no journal found");
4238                 return NULL;
4239         }
4240         if (!journal_inode->i_nlink) {
4241                 make_bad_inode(journal_inode);
4242                 iput(journal_inode);
4243                 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4244                 return NULL;
4245         }
4246 
4247         jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4248                   journal_inode, journal_inode->i_size);
4249         if (!S_ISREG(journal_inode->i_mode)) {
4250                 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4251                 iput(journal_inode);
4252                 return NULL;
4253         }
4254 
4255         journal = jbd2_journal_init_inode(journal_inode);
4256         if (!journal) {
4257                 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4258                 iput(journal_inode);
4259                 return NULL;
4260         }
4261         journal->j_private = sb;
4262         ext4_init_journal_params(sb, journal);
4263         return journal;
4264 }
4265 
4266 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4267                                        dev_t j_dev)
4268 {
4269         struct buffer_head *bh;
4270         journal_t *journal;
4271         ext4_fsblk_t start;
4272         ext4_fsblk_t len;
4273         int hblock, blocksize;
4274         ext4_fsblk_t sb_block;
4275         unsigned long offset;
4276         struct ext4_super_block *es;
4277         struct block_device *bdev;
4278 
4279         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4280 
4281         bdev = ext4_blkdev_get(j_dev, sb);
4282         if (bdev == NULL)
4283                 return NULL;
4284 
4285         blocksize = sb->s_blocksize;
4286         hblock = bdev_logical_block_size(bdev);
4287         if (blocksize < hblock) {
4288                 ext4_msg(sb, KERN_ERR,
4289                         "blocksize too small for journal device");
4290                 goto out_bdev;
4291         }
4292 
4293         sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4294         offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4295         set_blocksize(bdev, blocksize);
4296         if (!(bh = __bread(bdev, sb_block, blocksize))) {
4297                 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4298                        "external journal");
4299                 goto out_bdev;
4300         }
4301 
4302         es = (struct ext4_super_block *) (bh->b_data + offset);
4303         if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4304             !(le32_to_cpu(es->s_feature_incompat) &
4305               EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4306                 ext4_msg(sb, KERN_ERR, "external journal has "
4307                                         "bad superblock");
4308                 brelse(bh);
4309                 goto out_bdev;
4310         }
4311 
4312         if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4313                 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4314                 brelse(bh);
4315                 goto out_bdev;
4316         }
4317 
4318         len = ext4_blocks_count(es);
4319         start = sb_block + 1;
4320         brelse(bh);     /* we're done with the superblock */
4321 
4322         journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4323                                         start, len, blocksize);
4324         if (!journal) {
4325                 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4326                 goto out_bdev;
4327         }
4328         journal->j_private = sb;
4329         ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4330         wait_on_buffer(journal->j_sb_buffer);
4331         if (!buffer_uptodate(journal->j_sb_buffer)) {
4332                 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4333                 goto out_journal;
4334         }
4335         if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4336                 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4337                                         "user (unsupported) - %d",
4338                         be32_to_cpu(journal->j_superblock->s_nr_users));
4339                 goto out_journal;
4340         }
4341         EXT4_SB(sb)->journal_bdev = bdev;
4342         ext4_init_journal_params(sb, journal);
4343         return journal;
4344 
4345 out_journal:
4346         jbd2_journal_destroy(journal);
4347 out_bdev:
4348         ext4_blkdev_put(bdev);
4349         return NULL;
4350 }
4351 
4352 static int ext4_load_journal(struct super_block *sb,
4353                              struct ext4_super_block *es,
4354                              unsigned long journal_devnum)
4355 {
4356         journal_t *journal;
4357         unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4358         dev_t journal_dev;
4359         int err = 0;
4360         int really_read_only;
4361 
4362         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4363 
4364         if (journal_devnum &&
4365             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4366                 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4367                         "numbers have changed");
4368                 journal_dev = new_decode_dev(journal_devnum);
4369         } else
4370                 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4371 
4372         really_read_only = bdev_read_only(sb->s_bdev);
4373 
4374         /*
4375          * Are we loading a blank journal or performing recovery after a
4376          * crash?  For recovery, we need to check in advance whether we
4377          * can get read-write access to the device.
4378          */
4379         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4380                 if (sb->s_flags & MS_RDONLY) {
4381                         ext4_msg(sb, KERN_INFO, "INFO: recovery "
4382                                         "required on readonly filesystem");
4383                         if (really_read_only) {
4384                                 ext4_msg(sb, KERN_ERR, "write access "
4385                                         "unavailable, cannot proceed");
4386                                 return -EROFS;
4387                         }
4388                         ext4_msg(sb, KERN_INFO, "write access will "
4389                                "be enabled during recovery");
4390                 }
4391         }
4392 
4393         if (journal_inum && journal_dev) {
4394                 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4395                        "and inode journals!");
4396                 return -EINVAL;
4397         }
4398 
4399         if (journal_inum) {
4400                 if (!(journal = ext4_get_journal(sb, journal_inum)))
4401                         return -EINVAL;
4402         } else {
4403                 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4404                         return -EINVAL;
4405         }
4406 
4407         if (!(journal->j_flags & JBD2_BARRIER))
4408                 ext4_msg(sb, KERN_INFO, "barriers disabled");
4409 
4410         if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4411                 err = jbd2_journal_wipe(journal, !really_read_only);
4412         if (!err) {
4413                 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4414                 if (save)
4415                         memcpy(save, ((char *) es) +
4416                                EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4417                 err = jbd2_journal_load(journal);
4418                 if (save)
4419                         memcpy(((char *) es) + EXT4_S_ERR_START,
4420                                save, EXT4_S_ERR_LEN);
4421                 kfree(save);
4422         }
4423 
4424         if (err) {
4425                 ext4_msg(sb, KERN_ERR, "error loading journal");
4426                 jbd2_journal_destroy(journal);
4427                 return err;
4428         }
4429 
4430         EXT4_SB(sb)->s_journal = journal;
4431         ext4_clear_journal_err(sb, es);
4432 
4433         if (!really_read_only && journal_devnum &&
4434             journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4435                 es->s_journal_dev = cpu_to_le32(journal_devnum);
4436 
4437                 /* Make sure we flush the recovery flag to disk. */
4438                 ext4_commit_super(sb, 1);
4439         }
4440 
4441         return 0;
4442 }
4443 
4444 static int ext4_commit_super(struct super_block *sb, int sync)
4445 {
4446         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4447         struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4448         int error = 0;
4449 
4450         if (!sbh || block_device_ejected(sb))
4451                 return error;
4452         if (buffer_write_io_error(sbh)) {
4453                 /*
4454                  * Oh, dear.  A previous attempt to write the
4455                  * superblock failed.  This could happen because the
4456                  * USB device was yanked out.  Or it could happen to
4457                  * be a transient write error and maybe the block will
4458                  * be remapped.  Nothing we can do but to retry the
4459                  * write and hope for the best.
4460                  */
4461                 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4462                        "superblock detected");
4463                 clear_buffer_write_io_error(sbh);
4464                 set_buffer_uptodate(sbh);
4465         }
4466         /*
4467          * If the file system is mounted read-only, don't update the
4468          * superblock write time.  This avoids updating the superblock
4469          * write time when we are mounting the root file system
4470          * read/only but we need to replay the journal; at that point,
4471          * for people who are east of GMT and who make their clock
4472          * tick in localtime for Windows bug-for-bug compatibility,
4473          * the clock is set in the future, and this will cause e2fsck
4474          * to complain and force a full file system check.
4475          */
4476         if (!(sb->s_flags & MS_RDONLY))
4477                 es->s_wtime = cpu_to_le32(get_seconds());
4478         if (sb->s_bdev->bd_part)
4479                 es->s_kbytes_written =
4480                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4481                             ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4482                               EXT4_SB(sb)->s_sectors_written_start) >> 1));
4483         else
4484                 es->s_kbytes_written =
4485                         cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4486         ext4_free_blocks_count_set(es,
4487                         EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4488                                 &EXT4_SB(sb)->s_freeclusters_counter)));
4489         es->s_free_inodes_count =
4490                 cpu_to_le32(percpu_counter_sum_positive(
4491                                 &EXT4_SB(sb)->s_freeinodes_counter));
4492         BUFFER_TRACE(sbh, "marking dirty");
4493         ext4_superblock_csum_set(sb);
4494         mark_buffer_dirty(sbh);
4495         if (sync) {
4496                 error = sync_dirty_buffer(sbh);
4497                 if (error)
4498                         return error;
4499 
4500                 error = buffer_write_io_error(sbh);
4501                 if (error) {
4502                         ext4_msg(sb, KERN_ERR, "I/O error while writing "
4503                                "superblock");
4504                         clear_buffer_write_io_error(sbh);
4505                         set_buffer_uptodate(sbh);
4506                 }
4507         }
4508         return error;
4509 }
4510 
4511 /*
4512  * Have we just finished recovery?  If so, and if we are mounting (or
4513  * remounting) the filesystem readonly, then we will end up with a
4514  * consistent fs on disk.  Record that fact.
4515  */
4516 static void ext4_mark_recovery_complete(struct super_block *sb,
4517                                         struct ext4_super_block *es)
4518 {
4519         journal_t *journal = EXT4_SB(sb)->s_journal;
4520 
4521         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4522                 BUG_ON(journal != NULL);
4523                 return;
4524         }
4525         jbd2_journal_lock_updates(journal);
4526         if (jbd2_journal_flush(journal) < 0)
4527                 goto out;
4528 
4529         if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4530             sb->s_flags & MS_RDONLY) {
4531                 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4532                 ext4_commit_super(sb, 1);
4533         }
4534 
4535 out:
4536         jbd2_journal_unlock_updates(journal);
4537 }
4538 
4539 /*
4540  * If we are mounting (or read-write remounting) a filesystem whose journal
4541  * has recorded an error from a previous lifetime, move that error to the
4542  * main filesystem now.
4543  */
4544 static void ext4_clear_journal_err(struct super_block *sb,
4545                                    struct ext4_super_block *es)
4546 {
4547         journal_t *journal;
4548         int j_errno;
4549         const char *errstr;
4550 
4551         BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4552 
4553         journal = EXT4_SB(sb)->s_journal;
4554 
4555         /*
4556          * Now check for any error status which may have been recorded in the
4557          * journal by a prior ext4_error() or ext4_abort()
4558          */
4559 
4560         j_errno = jbd2_journal_errno(journal);
4561         if (j_errno) {
4562                 char nbuf[16];
4563 
4564                 errstr = ext4_decode_error(sb, j_errno, nbuf);
4565                 ext4_warning(sb, "Filesystem error recorded "
4566                              "from previous mount: %s", errstr);
4567                 ext4_warning(sb, "Marking fs in need of filesystem check.");
4568 
4569                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4570                 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4571                 ext4_commit_super(sb, 1);
4572 
4573                 jbd2_journal_clear_err(journal);
4574                 jbd2_journal_update_sb_errno(journal);
4575         }
4576 }
4577 
4578 /*
4579  * Force the running and committing transactions to commit,
4580  * and wait on the commit.
4581  */
4582 int ext4_force_commit(struct super_block *sb)
4583 {
4584         journal_t *journal;
4585 
4586         if (sb->s_flags & MS_RDONLY)
4587                 return 0;
4588 
4589         journal = EXT4_SB(sb)->s_journal;
4590         return ext4_journal_force_commit(journal);
4591 }
4592 
4593 static int ext4_sync_fs(struct super_block *sb, int wait)
4594 {
4595         int ret = 0;
4596         tid_t target;
4597         struct ext4_sb_info *sbi = EXT4_SB(sb);
4598 
4599         trace_ext4_sync_fs(sb, wait);
4600         flush_workqueue(sbi->dio_unwritten_wq);
4601         /*
4602          * Writeback quota in non-journalled quota case - journalled quota has
4603          * no dirty dquots
4604          */
4605         dquot_writeback_dquots(sb, -1);
4606         if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4607                 if (wait)
4608                         jbd2_log_wait_commit(sbi->s_journal, target);
4609         }
4610         return ret;
4611 }
4612 
4613 /*
4614  * LVM calls this function before a (read-only) snapshot is created.  This
4615  * gives us a chance to flush the journal completely and mark the fs clean.
4616  *
4617  * Note that only this function cannot bring a filesystem to be in a clean
4618  * state independently. It relies on upper layer to stop all data & metadata
4619  * modifications.
4620  */
4621 static int ext4_freeze(struct super_block *sb)
4622 {
4623         int error = 0;
4624         journal_t *journal;
4625 
4626         if (sb->s_flags & MS_RDONLY)
4627                 return 0;
4628 
4629         journal = EXT4_SB(sb)->s_journal;
4630 
4631         /* Now we set up the journal barrier. */
4632         jbd2_journal_lock_updates(journal);
4633 
4634         /*
4635          * Don't clear the needs_recovery flag if we failed to flush
4636          * the journal.
4637          */
4638         error = jbd2_journal_flush(journal);
4639         if (error < 0)
4640                 goto out;
4641 
4642         /* Journal blocked and flushed, clear needs_recovery flag. */
4643         EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4644         error = ext4_commit_super(sb, 1);
4645 out:
4646         /* we rely on upper layer to stop further updates */
4647         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4648         return error;
4649 }
4650 
4651 /*
4652  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4653  * flag here, even though the filesystem is not technically dirty yet.
4654  */
4655 static int ext4_unfreeze(struct super_block *sb)
4656 {
4657         if (sb->s_flags & MS_RDONLY)
4658                 return 0;
4659 
4660         /* Reset the needs_recovery flag before the fs is unlocked. */
4661         EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4662         ext4_commit_super(sb, 1);
4663         return 0;
4664 }
4665 
4666 /*
4667  * Structure to save mount options for ext4_remount's benefit
4668  */
4669 struct ext4_mount_options {
4670         unsigned long s_mount_opt;
4671         unsigned long s_mount_opt2;
4672         kuid_t s_resuid;
4673         kgid_t s_resgid;
4674         unsigned long s_commit_interval;
4675         u32 s_min_batch_time, s_max_batch_time;
4676 #ifdef CONFIG_QUOTA
4677         int s_jquota_fmt;
4678         char *s_qf_names[MAXQUOTAS];
4679 #endif
4680 };
4681 
4682 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4683 {
4684         struct ext4_super_block *es;
4685         struct ext4_sb_info *sbi = EXT4_SB(sb);
4686         unsigned long old_sb_flags;
4687         struct ext4_mount_options old_opts;
4688         int enable_quota = 0;
4689         ext4_group_t g;
4690         unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4691         int err = 0;
4692 #ifdef CONFIG_QUOTA
4693         int i, j;
4694 #endif
4695         char *orig_data = kstrdup(data, GFP_KERNEL);
4696 
4697         /* Store the original options */
4698         old_sb_flags = sb->s_flags;
4699         old_opts.s_mount_opt = sbi->s_mount_opt;
4700         old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4701         old_opts.s_resuid = sbi->s_resuid;
4702         old_opts.s_resgid = sbi->s_resgid;
4703         old_opts.s_commit_interval = sbi->s_commit_interval;
4704         old_opts.s_min_batch_time = sbi->s_min_batch_time;
4705         old_opts.s_max_batch_time = sbi->s_max_batch_time;
4706 #ifdef CONFIG_QUOTA
4707         old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4708         for (i = 0; i < MAXQUOTAS; i++)
4709                 if (sbi->s_qf_names[i]) {
4710                         old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4711                                                          GFP_KERNEL);
4712                         if (!old_opts.s_qf_names[i]) {
4713                                 for (j = 0; j < i; j++)
4714                                         kfree(old_opts.s_qf_names[j]);
4715                                 kfree(orig_data);
4716                                 return -ENOMEM;
4717                         }
4718                 } else
4719                         old_opts.s_qf_names[i] = NULL;
4720 #endif
4721         if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4722                 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4723 
4724         /*
4725          * Allow the "check" option to be passed as a remount option.
4726          */
4727         if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4728                 err = -EINVAL;
4729                 goto restore_opts;
4730         }
4731 
4732         if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4733                 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4734                         ext4_msg(sb, KERN_ERR, "can't mount with "
4735                                  "both data=journal and delalloc");
4736                         err = -EINVAL;
4737                         goto restore_opts;
4738                 }
4739                 if (test_opt(sb, DIOREAD_NOLOCK)) {
4740                         ext4_msg(sb, KERN_ERR, "can't mount with "
4741                                  "both data=journal and dioread_nolock");
4742                         err = -EINVAL;
4743                         goto restore_opts;
4744                 }
4745         }
4746 
4747         if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4748                 ext4_abort(sb, "Abort forced by user");
4749 
4750         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4751                 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4752 
4753         es = sbi->s_es;
4754 
4755         if (sbi->s_journal) {
4756                 ext4_init_journal_params(sb, sbi->s_journal);
4757                 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4758         }
4759 
4760         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4761                 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4762                         err = -EROFS;
4763                         goto restore_opts;
4764                 }
4765 
4766                 if (*flags & MS_RDONLY) {
4767                         err = dquot_suspend(sb, -1);
4768                         if (err < 0)
4769                                 goto restore_opts;
4770 
4771                         /*
4772                          * First of all, the unconditional stuff we have to do
4773                          * to disable replay of the journal when we next remount
4774                          */
4775                         sb->s_flags |= MS_RDONLY;
4776 
4777                         /*
4778                          * OK, test if we are remounting a valid rw partition
4779                          * readonly, and if so set the rdonly flag and then
4780                          * mark the partition as valid again.
4781                          */
4782                         if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4783                             (sbi->s_mount_state & EXT4_VALID_FS))
4784                                 es->s_state = cpu_to_le16(sbi->s_mount_state);
4785 
4786                         if (sbi->s_journal)
4787                                 ext4_mark_recovery_complete(sb, es);
4788                 } else {
4789                         /* Make sure we can mount this feature set readwrite */
4790                         if (!ext4_feature_set_ok(sb, 0)) {
4791                                 err = -EROFS;
4792                                 goto restore_opts;
4793                         }
4794                         /*
4795                          * Make sure the group descriptor checksums
4796                          * are sane.  If they aren't, refuse to remount r/w.
4797                          */
4798                         for (g = 0; g < sbi->s_groups_count; g++) {
4799                                 struct ext4_group_desc *gdp =
4800                                         ext4_get_group_desc(sb, g, NULL);
4801 
4802                                 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4803                                         ext4_msg(sb, KERN_ERR,
4804                "ext4_remount: Checksum for group %u failed (%u!=%u)",
4805                 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4806                                                le16_to_cpu(gdp->bg_checksum));
4807                                         err = -EINVAL;
4808                                         goto restore_opts;
4809                                 }
4810                         }
4811 
4812                         /*
4813                          * If we have an unprocessed orphan list hanging
4814                          * around from a previously readonly bdev mount,
4815                          * require a full umount/remount for now.
4816                          */
4817                         if (es->s_last_orphan) {
4818                                 ext4_msg(sb, KERN_WARNING, "Couldn't "
4819                                        "remount RDWR because of unprocessed "
4820                                        "orphan inode list.  Please "
4821                                        "umount/remount instead");
4822                                 err = -EINVAL;
4823                                 goto restore_opts;
4824                         }
4825 
4826                         /*
4827                          * Mounting a RDONLY partition read-write, so reread
4828                          * and store the current valid flag.  (It may have
4829                          * been changed by e2fsck since we originally mounted
4830                          * the partition.)
4831                          */
4832                         if (sbi->s_journal)
4833                                 ext4_clear_journal_err(sb, es);
4834                         sbi->s_mount_state = le16_to_cpu(es->s_state);
4835                         if (!ext4_setup_super(sb, es, 0))
4836                                 sb->s_flags &= ~MS_RDONLY;
4837                         if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4838                                                      EXT4_FEATURE_INCOMPAT_MMP))
4839                                 if (ext4_multi_mount_protect(sb,
4840                                                 le64_to_cpu(es->s_mmp_block))) {
4841                                         err = -EROFS;
4842                                         goto restore_opts;
4843                                 }
4844                         enable_quota = 1;
4845                 }
4846         }
4847 
4848         /*
4849          * Reinitialize lazy itable initialization thread based on
4850          * current settings
4851          */
4852         if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4853                 ext4_unregister_li_request(sb);
4854         else {
4855                 ext4_group_t first_not_zeroed;
4856                 first_not_zeroed = ext4_has_uninit_itable(sb);
4857                 ext4_register_li_request(sb, first_not_zeroed);
4858         }
4859 
4860         ext4_setup_system_zone(sb);
4861         if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4862                 ext4_commit_super(sb, 1);
4863 
4864 #ifdef CONFIG_QUOTA
4865         /* Release old quota file names */
4866         for (i = 0; i < MAXQUOTAS; i++)
4867                 kfree(old_opts.s_qf_names[i]);
4868         if (enable_quota) {
4869                 if (sb_any_quota_suspended(sb))
4870                         dquot_resume(sb, -1);
4871                 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4872                                         EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4873                         err = ext4_enable_quotas(sb);
4874                         if (err)
4875                                 goto restore_opts;
4876                 }
4877         }
4878 #endif
4879 
4880         ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4881         kfree(orig_data);
4882         return 0;
4883 
4884 restore_opts:
4885         sb->s_flags = old_sb_flags;
4886         sbi->s_mount_opt = old_opts.s_mount_opt;
4887         sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4888         sbi->s_resuid = old_opts.s_resuid;
4889         sbi->s_resgid = old_opts.s_resgid;
4890         sbi->s_commit_interval = old_opts.s_commit_interval;
4891         sbi->s_min_batch_time = old_opts.s_min_batch_time;
4892         sbi->s_max_batch_time = old_opts.s_max_batch_time;
4893 #ifdef CONFIG_QUOTA
4894         sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4895         for (i = 0; i < MAXQUOTAS; i++) {
4896                 kfree(sbi->s_qf_names[i]);
4897                 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4898         }
4899 #endif
4900         kfree(orig_data);
4901         return err;
4902 }
4903 
4904 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4905 {
4906         struct super_block *sb = dentry->d_sb;
4907         struct ext4_sb_info *sbi = EXT4_SB(sb);
4908         struct ext4_super_block *es = sbi->s_es;
4909         ext4_fsblk_t overhead = 0, resv_blocks;
4910         u64 fsid;
4911         s64 bfree;
4912         resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4913 
4914         if (!test_opt(sb, MINIX_DF))
4915                 overhead = sbi->s_overhead;
4916 
4917         buf->f_type = EXT4_SUPER_MAGIC;
4918         buf->f_bsize = sb->s_blocksize;
4919         buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4920         bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4921                 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4922         /* prevent underflow in case that few free space is available */
4923         buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4924         buf->f_bavail = buf->f_bfree -
4925                         (ext4_r_blocks_count(es) + resv_blocks);
4926         if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4927                 buf->f_bavail = 0;
4928         buf->f_files = le32_to_cpu(es->s_inodes_count);
4929         buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4930         buf->f_namelen = EXT4_NAME_LEN;
4931         fsid = le64_to_cpup((void *)es->s_uuid) ^
4932                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4933         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4934         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4935 
4936         return 0;
4937 }
4938 
4939 /* Helper function for writing quotas on sync - we need to start transaction
4940  * before quota file is locked for write. Otherwise the are possible deadlocks:
4941  * Process 1                         Process 2
4942  * ext4_create()                     quota_sync()
4943  *   jbd2_journal_start()                  write_dquot()
4944  *   dquot_initialize()                         down(dqio_mutex)
4945  *     down(dqio_mutex)                    jbd2_journal_start()
4946  *
4947  */
4948 
4949 #ifdef CONFIG_QUOTA
4950 
4951 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4952 {
4953         return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4954 }
4955 
4956 static int ext4_write_dquot(struct dquot *dquot)
4957 {
4958         int ret, err;
4959         handle_t *handle;
4960         struct inode *inode;
4961 
4962         inode = dquot_to_inode(dquot);
4963         handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4964                                     EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4965         if (IS_ERR(handle))
4966                 return PTR_ERR(handle);
4967         ret = dquot_commit(dquot);
4968         err = ext4_journal_stop(handle);
4969         if (!ret)
4970                 ret = err;
4971         return ret;
4972 }
4973 
4974 static int ext4_acquire_dquot(struct dquot *dquot)
4975 {
4976         int ret, err;
4977         handle_t *handle;
4978 
4979         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4980                                     EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4981         if (IS_ERR(handle))
4982                 return PTR_ERR(handle);
4983         ret = dquot_acquire(dquot);
4984         err = ext4_journal_stop(handle);
4985         if (!ret)
4986                 ret = err;
4987         return ret;
4988 }
4989 
4990 static int ext4_release_dquot(struct dquot *dquot)
4991 {
4992         int ret, err;
4993         handle_t *handle;
4994 
4995         handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4996                                     EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4997         if (IS_ERR(handle)) {
4998                 /* Release dquot anyway to avoid endless cycle in dqput() */
4999                 dquot_release(dquot);
5000                 return PTR_ERR(handle);
5001         }
5002         ret = dquot_release(dquot);
5003         err = ext4_journal_stop(handle);
5004         if (!ret)
5005                 ret = err;
5006         return ret;
5007 }
5008 
5009 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5010 {
5011         struct super_block *sb = dquot->dq_sb;
5012         struct ext4_sb_info *sbi = EXT4_SB(sb);
5013 
5014         /* Are we journaling quotas? */
5015         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5016             sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5017                 dquot_mark_dquot_dirty(dquot);
5018                 return ext4_write_dquot(dquot);
5019         } else {
5020                 return dquot_mark_dquot_dirty(dquot);
5021         }
5022 }
5023 
5024 static int ext4_write_info(struct super_block *sb, int type)
5025 {
5026         int ret, err;
5027         handle_t *handle;
5028 
5029         /* Data block + inode block */
5030         handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5031         if (IS_ERR(handle))
5032                 return PTR_ERR(handle);
5033         ret = dquot_commit_info(sb, type);
5034         err = ext4_journal_stop(handle);
5035         if (!ret)
5036                 ret = err;
5037         return ret;
5038 }
5039 
5040 /*
5041  * Turn on quotas during mount time - we need to find
5042  * the quota file and such...
5043  */
5044 static int ext4_quota_on_mount(struct super_block *sb, int type)
5045 {
5046         return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5047                                         EXT4_SB(sb)->s_jquota_fmt, type);
5048 }
5049 
5050 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5051 {
5052         struct ext4_inode_info *ei = EXT4_I(inode);
5053 
5054         /* The first argument of lockdep_set_subclass has to be
5055          * *exactly* the same as the argument to init_rwsem() --- in
5056          * this case, in init_once() --- or lockdep gets unhappy
5057          * because the name of the lock is set using the
5058          * stringification of the argument to init_rwsem().
5059          */
5060         (void) ei;      /* shut up clang warning if !CONFIG_LOCKDEP */
5061         lockdep_set_subclass(&ei->i_data_sem, subclass);
5062 }
5063 
5064 /*
5065  * Standard function to be called on quota_on
5066  */
5067 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5068                          struct path *path)
5069 {
5070         int err;
5071 
5072         if (!test_opt(sb, QUOTA))
5073                 return -EINVAL;
5074 
5075         /* Quotafile not on the same filesystem? */
5076         if (path->dentry->d_sb != sb)
5077                 return -EXDEV;
5078         /* Journaling quota? */
5079         if (EXT4_SB(sb)->s_qf_names[type]) {
5080                 /* Quotafile not in fs root? */
5081                 if (path->dentry->d_parent != sb->s_root)
5082                         ext4_msg(sb, KERN_WARNING,
5083                                 "Quota file not on filesystem root. "
5084                                 "Journaled quota will not work");
5085         }
5086 
5087         /*
5088          * When we journal data on quota file, we have to flush journal to see
5089          * all updates to the file when we bypass pagecache...
5090          */
5091         if (EXT4_SB(sb)->s_journal &&
5092             ext4_should_journal_data(path->dentry->d_inode)) {
5093                 /*
5094                  * We don't need to lock updates but journal_flush() could
5095                  * otherwise be livelocked...
5096                  */
5097                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5098                 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5099                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5100                 if (err)
5101                         return err;
5102         }
5103         lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5104         err = dquot_quota_on(sb, type, format_id, path);
5105         if (err)
5106                 lockdep_set_quota_inode(path->dentry->d_inode,
5107                                              I_DATA_SEM_NORMAL);
5108         return err;
5109 }
5110 
5111 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5112                              unsigned int flags)
5113 {
5114         int err;
5115         struct inode *qf_inode;
5116         unsigned long qf_inums[MAXQUOTAS] = {
5117                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5118                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5119         };
5120 
5121         BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5122 
5123         if (!qf_inums[type])
5124                 return -EPERM;
5125 
5126         qf_inode = ext4_iget(sb, qf_inums[type]);
5127         if (IS_ERR(qf_inode)) {
5128                 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5129                 return PTR_ERR(qf_inode);
5130         }
5131 
5132         /* Don't account quota for quota files to avoid recursion */
5133         qf_inode->i_flags |= S_NOQUOTA;
5134         lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5135         err = dquot_enable(qf_inode, type, format_id, flags);
5136         iput(qf_inode);
5137         if (err)
5138                 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5139 
5140         return err;
5141 }
5142 
5143 /* Enable usage tracking for all quota types. */
5144 static int ext4_enable_quotas(struct super_block *sb)
5145 {
5146         int type, err = 0;
5147         unsigned long qf_inums[MAXQUOTAS] = {
5148                 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5149                 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5150         };
5151 
5152         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5153         for (type = 0; type < MAXQUOTAS; type++) {
5154                 if (qf_inums[type]) {
5155                         err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5156                                                 DQUOT_USAGE_ENABLED);
5157                         if (err) {
5158                                 ext4_warning(sb,
5159                                         "Failed to enable quota tracking "
5160                                         "(type=%d, err=%d). Please run "
5161                                         "e2fsck to fix.", type, err);
5162                                 return err;
5163                         }
5164                 }
5165         }
5166         return 0;
5167 }
5168 
5169 /*
5170  * quota_on function that is used when QUOTA feature is set.
5171  */
5172 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5173                                  int format_id)
5174 {
5175         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5176                 return -EINVAL;
5177 
5178         /*
5179          * USAGE was enabled at mount time. Only need to enable LIMITS now.
5180          */
5181         return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5182 }
5183 
5184 static int ext4_quota_off(struct super_block *sb, int type)
5185 {
5186         struct inode *inode = sb_dqopt(sb)->files[type];
5187         handle_t *handle;
5188 
5189         /* Force all delayed allocation blocks to be allocated.
5190          * Caller already holds s_umount sem */
5191         if (test_opt(sb, DELALLOC))
5192                 sync_filesystem(sb);
5193 
5194         if (!inode)
5195                 goto out;
5196 
5197         /* Update modification times of quota files when userspace can
5198          * start looking at them */
5199         handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5200         if (IS_ERR(handle))
5201                 goto out;
5202         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5203         ext4_mark_inode_dirty(handle, inode);
5204         ext4_journal_stop(handle);
5205 
5206 out:
5207         return dquot_quota_off(sb, type);
5208 }
5209 
5210 /*
5211  * quota_off function that is used when QUOTA feature is set.
5212  */
5213 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5214 {
5215         if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5216                 return -EINVAL;
5217 
5218         /* Disable only the limits. */
5219         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5220 }
5221 
5222 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5223  * acquiring the locks... As quota files are never truncated and quota code
5224  * itself serializes the operations (and no one else should touch the files)
5225  * we don't have to be afraid of races */
5226 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5227                                size_t len, loff_t off)
5228 {
5229         struct inode *inode = sb_dqopt(sb)->files[type];
5230         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5231         int err = 0;
5232         int offset = off & (sb->s_blocksize - 1);
5233         int tocopy;
5234         size_t toread;
5235         struct buffer_head *bh;
5236         loff_t i_size = i_size_read(inode);
5237 
5238         if (off > i_size)
5239                 return 0;
5240         if (off+len > i_size)
5241                 len = i_size-off;
5242         toread = len;
5243         while (toread > 0) {
5244                 tocopy = sb->s_blocksize - offset < toread ?
5245                                 sb->s_blocksize - offset : toread;
5246                 bh = ext4_bread(NULL, inode, blk, 0, &err);
5247                 if (err)
5248                         return err;
5249                 if (!bh)        /* A hole? */
5250                         memset(data, 0, tocopy);
5251                 else
5252                         memcpy(data, bh->b_data+offset, tocopy);
5253                 brelse(bh);
5254                 offset = 0;
5255                 toread -= tocopy;
5256                 data += tocopy;
5257                 blk++;
5258         }
5259         return len;
5260 }
5261 
5262 /* Write to quotafile (we know the transaction is already started and has
5263  * enough credits) */
5264 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5265                                 const char *data, size_t len, loff_t off)
5266 {
5267         struct inode *inode = sb_dqopt(sb)->files[type];
5268         ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5269         int err = 0;
5270         int offset = off & (sb->s_blocksize - 1);
5271         struct buffer_head *bh;
5272         handle_t *handle = journal_current_handle();
5273 
5274         if (EXT4_SB(sb)->s_journal && !handle) {
5275                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5276                         " cancelled because transaction is not started",
5277                         (unsigned long long)off, (unsigned long long)len);
5278                 return -EIO;
5279         }
5280         /*
5281          * Since we account only one data block in transaction credits,
5282          * then it is impossible to cross a block boundary.
5283          */
5284         if (sb->s_blocksize - offset < len) {
5285                 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5286                         " cancelled because not block aligned",
5287                         (unsigned long long)off, (unsigned long long)len);
5288                 return -EIO;
5289         }
5290 
5291         bh = ext4_bread(handle, inode, blk, 1, &err);
5292         if (!bh)
5293                 goto out;
5294         err = ext4_journal_get_write_access(handle, bh);
5295         if (err) {
5296                 brelse(bh);
5297                 goto out;
5298         }
5299         lock_buffer(bh);
5300         memcpy(bh->b_data+offset, data, len);
5301         flush_dcache_page(bh->b_page);
5302         unlock_buffer(bh);
5303         err = ext4_handle_dirty_metadata(handle, NULL, bh);
5304         brelse(bh);
5305 out:
5306         if (err)
5307                 return err;
5308         if (inode->i_size < off + len) {
5309                 i_size_write(inode, off + len);
5310                 EXT4_I(inode)->i_disksize = inode->i_size;
5311                 ext4_mark_inode_dirty(handle, inode);
5312         }
5313         return len;
5314 }
5315 
5316 #endif
5317 
5318 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5319                        const char *dev_name, void *data)
5320 {
5321         return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5322 }
5323 
5324 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5325 static inline void register_as_ext2(void)
5326 {
5327         int err = register_filesystem(&ext2_fs_type);
5328         if (err)
5329                 printk(KERN_WARNING
5330                        "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5331 }
5332 
5333 static inline void unregister_as_ext2(void)
5334 {
5335         unregister_filesystem(&ext2_fs_type);
5336 }
5337 
5338 static inline int ext2_feature_set_ok(struct super_block *sb)
5339 {
5340         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5341                 return 0;
5342         if (sb->s_flags & MS_RDONLY)
5343                 return 1;
5344         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5345                 return 0;
5346         return 1;
5347 }
5348 #else
5349 static inline void register_as_ext2(void) { }
5350 static inline void unregister_as_ext2(void) { }
5351 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5352 #endif
5353 
5354 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5355 static inline void register_as_ext3(void)
5356 {
5357         int err = register_filesystem(&ext3_fs_type);
5358         if (err)
5359                 printk(KERN_WARNING
5360                        "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5361 }
5362 
5363 static inline void unregister_as_ext3(void)
5364 {
5365         unregister_filesystem(&ext3_fs_type);
5366 }
5367 
5368 static inline int ext3_feature_set_ok(struct super_block *sb)
5369 {
5370         if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5371                 return 0;
5372         if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5373                 return 0;
5374         if (sb->s_flags & MS_RDONLY)
5375                 return 1;
5376         if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5377                 return 0;
5378         return 1;
5379 }
5380 #else
5381 static inline void register_as_ext3(void) { }
5382 static inline void unregister_as_ext3(void) { }
5383 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5384 #endif
5385 
5386 static struct file_system_type ext4_fs_type = {
5387         .owner          = THIS_MODULE,
5388         .name           = "ext4",
5389         .mount          = ext4_mount,
5390         .kill_sb        = kill_block_super,
5391         .fs_flags       = FS_REQUIRES_DEV,
5392 };
5393 MODULE_ALIAS_FS("ext4");
5394 
5395 static int __init ext4_init_feat_adverts(void)
5396 {
5397         struct ext4_features *ef;
5398         int ret = -ENOMEM;
5399 
5400         ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5401         if (!ef)
5402                 goto out;
5403 
5404         ef->f_kobj.kset = ext4_kset;
5405         init_completion(&ef->f_kobj_unregister);
5406         ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5407                                    "features");
5408         if (ret) {
5409                 kfree(ef);
5410                 goto out;
5411         }
5412 
5413         ext4_feat = ef;
5414         ret = 0;
5415 out:
5416         return ret;
5417 }
5418 
5419 static void ext4_exit_feat_adverts(void)
5420 {
5421         kobject_put(&ext4_feat->f_kobj);
5422         wait_for_completion(&ext4_feat->f_kobj_unregister);
5423         kfree(ext4_feat);
5424 }
5425 
5426 /* Shared across all ext4 file systems */
5427 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5428 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5429 
5430 static int __init ext4_init_fs(void)
5431 {
5432         int i, err;
5433 
5434         ext4_li_info = NULL;
5435         mutex_init(&ext4_li_mtx);
5436 
5437         /* Build-time check for flags consistency */
5438         ext4_check_flag_values();
5439 
5440         for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5441                 mutex_init(&ext4__aio_mutex[i]);
5442                 init_waitqueue_head(&ext4__ioend_wq[i]);
5443         }
5444 
5445         err = ext4_init_es();
5446         if (err)
5447                 return err;
5448 
5449         err = ext4_init_pageio();
5450         if (err)
5451                 goto out7;
5452 
5453         err = ext4_init_system_zone();
5454         if (err)
5455                 goto out6;
5456         ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5457         if (!ext4_kset) {
5458                 err = -ENOMEM;
5459                 goto out5;
5460         }
5461         ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5462 
5463         err = ext4_init_feat_adverts();
5464         if (err)
5465                 goto out4;
5466 
5467         err = ext4_init_mballoc();
5468         if (err)
5469                 goto out3;
5470 
5471         err = ext4_init_xattr();
5472         if (err)
5473                 goto out2;
5474         err = init_inodecache();
5475         if (err)
5476                 goto out1;
5477         register_as_ext3();
5478         register_as_ext2();
5479         err = register_filesystem(&ext4_fs_type);
5480         if (err)
5481                 goto out;
5482 
5483         return 0;
5484 out:
5485         unregister_as_ext2();
5486         unregister_as_ext3();
5487         destroy_inodecache();
5488 out1:
5489         ext4_exit_xattr();
5490 out2:
5491         ext4_exit_mballoc();
5492 out3:
5493         ext4_exit_feat_adverts();
5494 out4:
5495         if (ext4_proc_root)
5496                 remove_proc_entry("fs/ext4", NULL);
5497         kset_unregister(ext4_kset);
5498 out5:
5499         ext4_exit_system_zone();
5500 out6:
5501         ext4_exit_pageio();
5502 out7:
5503         ext4_exit_es();
5504 
5505         return err;
5506 }
5507 
5508 static void __exit ext4_exit_fs(void)
5509 {
5510         ext4_destroy_lazyinit_thread();
5511         unregister_as_ext2();
5512         unregister_as_ext3();
5513         unregister_filesystem(&ext4_fs_type);
5514         destroy_inodecache();
5515         ext4_exit_xattr();
5516         ext4_exit_mballoc();
5517         ext4_exit_feat_adverts();
5518         remove_proc_entry("fs/ext4", NULL);
5519         kset_unregister(ext4_kset);
5520         ext4_exit_system_zone();
5521         ext4_exit_pageio();
5522         ext4_exit_es();
5523 }
5524 
5525 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5526 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5527 MODULE_LICENSE("GPL");
5528 module_init(ext4_init_fs)
5529 module_exit(ext4_exit_fs)
5530 

~ [ 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