2 * linux/fs/ext2/super.c
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)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/blkdev.h>
25 #include <linux/parser.h>
26 #include <linux/random.h>
27 #include <linux/buffer_head.h>
28 #include <linux/exportfs.h>
29 #include <linux/smp_lock.h>
30 #include <linux/vfs.h>
31 #include <linux/seq_file.h>
32 #include <linux/mount.h>
33 #include <linux/log2.h>
34 #include <asm/uaccess.h>
40 static void ext2_sync_super(struct super_block *sb,
41 struct ext2_super_block *es);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
45 void ext2_error (struct super_block * sb, const char * function,
46 const char * fmt, ...)
49 struct ext2_sb_info *sbi = EXT2_SB(sb);
50 struct ext2_super_block *es = sbi->s_es;
52 if (!(sb->s_flags & MS_RDONLY)) {
53 sbi->s_mount_state |= EXT2_ERROR_FS;
55 cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
56 ext2_sync_super(sb, es);
60 printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
65 if (test_opt(sb, ERRORS_PANIC))
66 panic("EXT2-fs panic from previous error\n");
67 if (test_opt(sb, ERRORS_RO)) {
68 printk("Remounting filesystem read-only\n");
69 sb->s_flags |= MS_RDONLY;
73 void ext2_warning (struct super_block * sb, const char * function,
74 const char * fmt, ...)
79 printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
86 void ext2_update_dynamic_rev(struct super_block *sb)
88 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
90 if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
93 ext2_warning(sb, __FUNCTION__,
94 "updating to rev %d because of new feature flag, "
95 "running e2fsck is recommended",
98 es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
99 es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
100 es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
101 /* leave es->s_feature_*compat flags alone */
102 /* es->s_uuid will be set by e2fsck if empty */
105 * The rest of the superblock fields should be zero, and if not it
106 * means they are likely already in use, so leave them alone. We
107 * can leave it up to e2fsck to clean up any inconsistencies there.
111 static void ext2_put_super (struct super_block * sb)
115 struct ext2_sb_info *sbi = EXT2_SB(sb);
117 ext2_xattr_put_super(sb);
118 if (!(sb->s_flags & MS_RDONLY)) {
119 struct ext2_super_block *es = sbi->s_es;
121 es->s_state = cpu_to_le16(sbi->s_mount_state);
122 ext2_sync_super(sb, es);
124 db_count = sbi->s_gdb_count;
125 for (i = 0; i < db_count; i++)
126 if (sbi->s_group_desc[i])
127 brelse (sbi->s_group_desc[i]);
128 kfree(sbi->s_group_desc);
130 percpu_counter_destroy(&sbi->s_freeblocks_counter);
131 percpu_counter_destroy(&sbi->s_freeinodes_counter);
132 percpu_counter_destroy(&sbi->s_dirs_counter);
134 sb->s_fs_info = NULL;
140 static struct kmem_cache * ext2_inode_cachep;
142 static struct inode *ext2_alloc_inode(struct super_block *sb)
144 struct ext2_inode_info *ei;
145 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
148 #ifdef CONFIG_EXT2_FS_POSIX_ACL
149 ei->i_acl = EXT2_ACL_NOT_CACHED;
150 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
152 ei->vfs_inode.i_version = 1;
153 return &ei->vfs_inode;
156 static void ext2_destroy_inode(struct inode *inode)
158 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
161 static void init_once(struct kmem_cache * cachep, void *foo)
163 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
165 rwlock_init(&ei->i_meta_lock);
166 #ifdef CONFIG_EXT2_FS_XATTR
167 init_rwsem(&ei->xattr_sem);
169 inode_init_once(&ei->vfs_inode);
172 static int init_inodecache(void)
174 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
175 sizeof(struct ext2_inode_info),
176 0, (SLAB_RECLAIM_ACCOUNT|
179 if (ext2_inode_cachep == NULL)
184 static void destroy_inodecache(void)
186 kmem_cache_destroy(ext2_inode_cachep);
189 static void ext2_clear_inode(struct inode *inode)
191 #ifdef CONFIG_EXT2_FS_POSIX_ACL
192 struct ext2_inode_info *ei = EXT2_I(inode);
194 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
195 posix_acl_release(ei->i_acl);
196 ei->i_acl = EXT2_ACL_NOT_CACHED;
198 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
199 posix_acl_release(ei->i_default_acl);
200 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
205 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
207 struct super_block *sb = vfs->mnt_sb;
208 struct ext2_sb_info *sbi = EXT2_SB(sb);
209 struct ext2_super_block *es = sbi->s_es;
210 unsigned long def_mount_opts;
212 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
214 if (sbi->s_sb_block != 1)
215 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
216 if (test_opt(sb, MINIX_DF))
217 seq_puts(seq, ",minixdf");
218 if (test_opt(sb, GRPID))
219 seq_puts(seq, ",grpid");
220 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
221 seq_puts(seq, ",nogrpid");
222 if (sbi->s_resuid != EXT2_DEF_RESUID ||
223 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
224 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
226 if (sbi->s_resgid != EXT2_DEF_RESGID ||
227 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
228 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
230 if (test_opt(sb, ERRORS_CONT)) {
231 int def_errors = le16_to_cpu(es->s_errors);
233 if (def_errors == EXT2_ERRORS_PANIC ||
234 def_errors == EXT2_ERRORS_RO) {
235 seq_puts(seq, ",errors=continue");
238 if (test_opt(sb, ERRORS_RO))
239 seq_puts(seq, ",errors=remount-ro");
240 if (test_opt(sb, ERRORS_PANIC))
241 seq_puts(seq, ",errors=panic");
242 if (test_opt(sb, NO_UID32))
243 seq_puts(seq, ",nouid32");
244 if (test_opt(sb, DEBUG))
245 seq_puts(seq, ",debug");
246 if (test_opt(sb, OLDALLOC))
247 seq_puts(seq, ",oldalloc");
249 #ifdef CONFIG_EXT2_FS_XATTR
250 if (test_opt(sb, XATTR_USER))
251 seq_puts(seq, ",user_xattr");
252 if (!test_opt(sb, XATTR_USER) &&
253 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
254 seq_puts(seq, ",nouser_xattr");
258 #ifdef CONFIG_EXT2_FS_POSIX_ACL
259 if (test_opt(sb, POSIX_ACL))
260 seq_puts(seq, ",acl");
261 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
262 seq_puts(seq, ",noacl");
265 if (test_opt(sb, NOBH))
266 seq_puts(seq, ",nobh");
268 #if defined(CONFIG_QUOTA)
269 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
270 seq_puts(seq, ",usrquota");
272 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
273 seq_puts(seq, ",grpquota");
276 #if defined(CONFIG_EXT2_FS_XIP)
277 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
278 seq_puts(seq, ",xip");
285 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
286 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
289 static const struct super_operations ext2_sops = {
290 .alloc_inode = ext2_alloc_inode,
291 .destroy_inode = ext2_destroy_inode,
292 .read_inode = ext2_read_inode,
293 .write_inode = ext2_write_inode,
294 .put_inode = ext2_put_inode,
295 .delete_inode = ext2_delete_inode,
296 .put_super = ext2_put_super,
297 .write_super = ext2_write_super,
298 .statfs = ext2_statfs,
299 .remount_fs = ext2_remount,
300 .clear_inode = ext2_clear_inode,
301 .show_options = ext2_show_options,
303 .quota_read = ext2_quota_read,
304 .quota_write = ext2_quota_write,
308 static struct dentry *ext2_get_dentry(struct super_block *sb, void *vobjp)
311 unsigned long ino = objp[0];
312 __u32 generation = objp[1];
314 struct dentry *result;
316 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
317 return ERR_PTR(-ESTALE);
318 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
319 return ERR_PTR(-ESTALE);
321 /* iget isn't really right if the inode is currently unallocated!!
322 * ext2_read_inode currently does appropriate checks, but
323 * it might be "neater" to call ext2_get_inode first and check
324 * if the inode is valid.....
326 inode = iget(sb, ino);
328 return ERR_PTR(-ENOMEM);
329 if (is_bad_inode(inode) ||
330 (generation && inode->i_generation != generation)) {
331 /* we didn't find the right inode.. */
333 return ERR_PTR(-ESTALE);
335 /* now to find a dentry.
336 * If possible, get a well-connected one
338 result = d_alloc_anon(inode);
341 return ERR_PTR(-ENOMEM);
346 /* Yes, most of these are left as NULL!!
347 * A NULL value implies the default, which works with ext2-like file
348 * systems, but can be improved upon.
349 * Currently only get_parent is required.
351 static struct export_operations ext2_export_ops = {
352 .get_parent = ext2_get_parent,
353 .get_dentry = ext2_get_dentry,
356 static unsigned long get_sb_block(void **data)
358 unsigned long sb_block;
359 char *options = (char *) *data;
361 if (!options || strncmp(options, "sb=", 3) != 0)
362 return 1; /* Default location */
364 sb_block = simple_strtoul(options, &options, 0);
365 if (*options && *options != ',') {
366 printk("EXT2-fs: Invalid sb specification: %s\n",
372 *data = (void *) options;
377 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
378 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
379 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
380 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
381 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
382 Opt_usrquota, Opt_grpquota
385 static match_table_t tokens = {
386 {Opt_bsd_df, "bsddf"},
387 {Opt_minix_df, "minixdf"},
388 {Opt_grpid, "grpid"},
389 {Opt_grpid, "bsdgroups"},
390 {Opt_nogrpid, "nogrpid"},
391 {Opt_nogrpid, "sysvgroups"},
392 {Opt_resgid, "resgid=%u"},
393 {Opt_resuid, "resuid=%u"},
395 {Opt_err_cont, "errors=continue"},
396 {Opt_err_panic, "errors=panic"},
397 {Opt_err_ro, "errors=remount-ro"},
398 {Opt_nouid32, "nouid32"},
399 {Opt_nocheck, "check=none"},
400 {Opt_nocheck, "nocheck"},
401 {Opt_debug, "debug"},
402 {Opt_oldalloc, "oldalloc"},
403 {Opt_orlov, "orlov"},
405 {Opt_user_xattr, "user_xattr"},
406 {Opt_nouser_xattr, "nouser_xattr"},
408 {Opt_noacl, "noacl"},
410 {Opt_grpquota, "grpquota"},
411 {Opt_ignore, "noquota"},
412 {Opt_quota, "quota"},
413 {Opt_usrquota, "usrquota"},
417 static int parse_options (char * options,
418 struct ext2_sb_info *sbi)
421 substring_t args[MAX_OPT_ARGS];
427 while ((p = strsep (&options, ",")) != NULL) {
432 token = match_token(p, tokens, args);
435 clear_opt (sbi->s_mount_opt, MINIX_DF);
438 set_opt (sbi->s_mount_opt, MINIX_DF);
441 set_opt (sbi->s_mount_opt, GRPID);
444 clear_opt (sbi->s_mount_opt, GRPID);
447 if (match_int(&args[0], &option))
449 sbi->s_resuid = option;
452 if (match_int(&args[0], &option))
454 sbi->s_resgid = option;
457 /* handled by get_sb_block() instead of here */
458 /* *sb_block = match_int(&args[0]); */
461 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
462 clear_opt (sbi->s_mount_opt, ERRORS_RO);
463 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
466 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
467 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
468 set_opt (sbi->s_mount_opt, ERRORS_RO);
471 clear_opt (sbi->s_mount_opt, ERRORS_RO);
472 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
473 set_opt (sbi->s_mount_opt, ERRORS_CONT);
476 set_opt (sbi->s_mount_opt, NO_UID32);
479 clear_opt (sbi->s_mount_opt, CHECK);
482 set_opt (sbi->s_mount_opt, DEBUG);
485 set_opt (sbi->s_mount_opt, OLDALLOC);
488 clear_opt (sbi->s_mount_opt, OLDALLOC);
491 set_opt (sbi->s_mount_opt, NOBH);
493 #ifdef CONFIG_EXT2_FS_XATTR
495 set_opt (sbi->s_mount_opt, XATTR_USER);
497 case Opt_nouser_xattr:
498 clear_opt (sbi->s_mount_opt, XATTR_USER);
502 case Opt_nouser_xattr:
503 printk("EXT2 (no)user_xattr options not supported\n");
506 #ifdef CONFIG_EXT2_FS_POSIX_ACL
508 set_opt(sbi->s_mount_opt, POSIX_ACL);
511 clear_opt(sbi->s_mount_opt, POSIX_ACL);
516 printk("EXT2 (no)acl options not supported\n");
520 #ifdef CONFIG_EXT2_FS_XIP
521 set_opt (sbi->s_mount_opt, XIP);
523 printk("EXT2 xip option not supported\n");
527 #if defined(CONFIG_QUOTA)
530 set_opt(sbi->s_mount_opt, USRQUOTA);
534 set_opt(sbi->s_mount_opt, GRPQUOTA);
541 "EXT2-fs: quota operations not supported.\n");
555 static int ext2_setup_super (struct super_block * sb,
556 struct ext2_super_block * es,
560 struct ext2_sb_info *sbi = EXT2_SB(sb);
562 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
563 printk ("EXT2-fs warning: revision level too high, "
564 "forcing read-only mode\n");
569 if (!(sbi->s_mount_state & EXT2_VALID_FS))
570 printk ("EXT2-fs warning: mounting unchecked fs, "
571 "running e2fsck is recommended\n");
572 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
573 printk ("EXT2-fs warning: mounting fs with errors, "
574 "running e2fsck is recommended\n");
575 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
576 le16_to_cpu(es->s_mnt_count) >=
577 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
578 printk ("EXT2-fs warning: maximal mount count reached, "
579 "running e2fsck is recommended\n");
580 else if (le32_to_cpu(es->s_checkinterval) &&
581 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
582 printk ("EXT2-fs warning: checktime reached, "
583 "running e2fsck is recommended\n");
584 if (!le16_to_cpu(es->s_max_mnt_count))
585 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
586 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
587 ext2_write_super(sb);
588 if (test_opt (sb, DEBUG))
589 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
590 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
591 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
594 EXT2_BLOCKS_PER_GROUP(sb),
595 EXT2_INODES_PER_GROUP(sb),
600 static int ext2_check_descriptors (struct super_block * sb)
604 struct ext2_sb_info *sbi = EXT2_SB(sb);
605 unsigned long first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
606 unsigned long last_block;
607 struct ext2_group_desc * gdp = NULL;
609 ext2_debug ("Checking group descriptors");
611 for (i = 0; i < sbi->s_groups_count; i++)
613 if (i == sbi->s_groups_count - 1)
614 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
616 last_block = first_block +
617 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
619 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
620 gdp = (struct ext2_group_desc *) sbi->s_group_desc[desc_block++]->b_data;
621 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
622 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
624 ext2_error (sb, "ext2_check_descriptors",
625 "Block bitmap for group %d"
626 " not in group (block %lu)!",
627 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
630 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
631 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
633 ext2_error (sb, "ext2_check_descriptors",
634 "Inode bitmap for group %d"
635 " not in group (block %lu)!",
636 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
639 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
640 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
643 ext2_error (sb, "ext2_check_descriptors",
644 "Inode table for group %d"
645 " not in group (block %lu)!",
646 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
649 first_block += EXT2_BLOCKS_PER_GROUP(sb);
656 * Maximal file size. There is a direct, and {,double-,triple-}indirect
657 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
658 * We need to be 1 filesystem block less than the 2^32 sector limit.
660 static loff_t ext2_max_size(int bits)
662 loff_t res = EXT2_NDIR_BLOCKS;
663 /* This constant is calculated to be the largest file size for a
664 * dense, 4k-blocksize file such that the total number of
665 * sectors in the file, including data and all indirect blocks,
666 * does not exceed 2^32. */
667 const loff_t upper_limit = 0x1ff7fffd000LL;
669 res += 1LL << (bits-2);
670 res += 1LL << (2*(bits-2));
671 res += 1LL << (3*(bits-2));
673 if (res > upper_limit)
678 static unsigned long descriptor_loc(struct super_block *sb,
679 unsigned long logic_sb_block,
682 struct ext2_sb_info *sbi = EXT2_SB(sb);
683 unsigned long bg, first_data_block, first_meta_bg;
686 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
687 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
689 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
691 return (logic_sb_block + nr + 1);
692 bg = sbi->s_desc_per_block * nr;
693 if (ext2_bg_has_super(sb, bg))
695 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
698 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
700 struct buffer_head * bh;
701 struct ext2_sb_info * sbi;
702 struct ext2_super_block * es;
705 unsigned long sb_block = get_sb_block(&data);
706 unsigned long logic_sb_block;
707 unsigned long offset = 0;
708 unsigned long def_mount_opts;
709 int blocksize = BLOCK_SIZE;
715 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
719 sbi->s_sb_block = sb_block;
722 * See what the current blocksize for the device is, and
723 * use that as the blocksize. Otherwise (or if the blocksize
724 * is smaller than the default) use the default.
725 * This is important for devices that have a hardware
726 * sectorsize that is larger than the default.
728 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
730 printk ("EXT2-fs: unable to set blocksize\n");
735 * If the superblock doesn't start on a hardware sector boundary,
736 * calculate the offset.
738 if (blocksize != BLOCK_SIZE) {
739 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
740 offset = (sb_block*BLOCK_SIZE) % blocksize;
742 logic_sb_block = sb_block;
745 if (!(bh = sb_bread(sb, logic_sb_block))) {
746 printk ("EXT2-fs: unable to read superblock\n");
750 * Note: s_es must be initialized as soon as possible because
751 * some ext2 macro-instructions depend on its value
753 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
755 sb->s_magic = le16_to_cpu(es->s_magic);
757 if (sb->s_magic != EXT2_SUPER_MAGIC)
760 /* Set defaults before we parse the mount options */
761 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
762 if (def_mount_opts & EXT2_DEFM_DEBUG)
763 set_opt(sbi->s_mount_opt, DEBUG);
764 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
765 set_opt(sbi->s_mount_opt, GRPID);
766 if (def_mount_opts & EXT2_DEFM_UID16)
767 set_opt(sbi->s_mount_opt, NO_UID32);
768 #ifdef CONFIG_EXT2_FS_XATTR
769 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
770 set_opt(sbi->s_mount_opt, XATTR_USER);
772 #ifdef CONFIG_EXT2_FS_POSIX_ACL
773 if (def_mount_opts & EXT2_DEFM_ACL)
774 set_opt(sbi->s_mount_opt, POSIX_ACL);
777 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
778 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
779 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO)
780 set_opt(sbi->s_mount_opt, ERRORS_RO);
782 set_opt(sbi->s_mount_opt, ERRORS_CONT);
784 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
785 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
787 if (!parse_options ((char *) data, sbi))
790 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
791 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
794 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
795 EXT2_MOUNT_XIP if not */
797 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
798 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
799 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
800 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
801 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
802 "running e2fsck is recommended\n");
804 * Check feature flags regardless of the revision level, since we
805 * previously didn't change the revision level when setting the flags,
806 * so there is a chance incompat flags are set on a rev 0 filesystem.
808 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
810 printk("EXT2-fs: %s: couldn't mount because of "
811 "unsupported optional features (%x).\n",
812 sb->s_id, le32_to_cpu(features));
815 if (!(sb->s_flags & MS_RDONLY) &&
816 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
817 printk("EXT2-fs: %s: couldn't mount RDWR because of "
818 "unsupported optional features (%x).\n",
819 sb->s_id, le32_to_cpu(features));
823 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
825 if ((ext2_use_xip(sb)) && ((blocksize != PAGE_SIZE) ||
826 (sb->s_blocksize != blocksize))) {
828 printk("XIP: Unsupported blocksize\n");
832 /* If the blocksize doesn't match, re-read the thing.. */
833 if (sb->s_blocksize != blocksize) {
836 if (!sb_set_blocksize(sb, blocksize)) {
837 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
841 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
842 offset = (sb_block*BLOCK_SIZE) % blocksize;
843 bh = sb_bread(sb, logic_sb_block);
845 printk("EXT2-fs: Couldn't read superblock on "
849 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
851 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
852 printk ("EXT2-fs: Magic mismatch, very weird !\n");
857 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
859 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
860 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
861 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
863 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
864 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
865 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
866 !is_power_of_2(sbi->s_inode_size) ||
867 (sbi->s_inode_size > blocksize)) {
868 printk ("EXT2-fs: unsupported inode size: %d\n",
874 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
875 le32_to_cpu(es->s_log_frag_size);
876 if (sbi->s_frag_size == 0)
878 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
880 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
881 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
882 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
884 if (EXT2_INODE_SIZE(sb) == 0)
886 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
887 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
889 sbi->s_itb_per_group = sbi->s_inodes_per_group /
890 sbi->s_inodes_per_block;
891 sbi->s_desc_per_block = sb->s_blocksize /
892 sizeof (struct ext2_group_desc);
894 sbi->s_mount_state = le16_to_cpu(es->s_state);
895 sbi->s_addr_per_block_bits =
896 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
897 sbi->s_desc_per_block_bits =
898 ilog2 (EXT2_DESC_PER_BLOCK(sb));
900 if (sb->s_magic != EXT2_SUPER_MAGIC)
903 if (sb->s_blocksize != bh->b_size) {
905 printk ("VFS: Unsupported blocksize on dev "
910 if (sb->s_blocksize != sbi->s_frag_size) {
911 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
912 sbi->s_frag_size, sb->s_blocksize);
916 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
917 printk ("EXT2-fs: #blocks per group too big: %lu\n",
918 sbi->s_blocks_per_group);
921 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
922 printk ("EXT2-fs: #fragments per group too big: %lu\n",
923 sbi->s_frags_per_group);
926 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
927 printk ("EXT2-fs: #inodes per group too big: %lu\n",
928 sbi->s_inodes_per_group);
932 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
934 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
935 le32_to_cpu(es->s_first_data_block) - 1)
936 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
937 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
938 EXT2_DESC_PER_BLOCK(sb);
939 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
940 if (sbi->s_group_desc == NULL) {
941 printk ("EXT2-fs: not enough memory\n");
944 bgl_lock_init(&sbi->s_blockgroup_lock);
945 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
947 printk ("EXT2-fs: not enough memory\n");
948 goto failed_mount_group_desc;
950 for (i = 0; i < db_count; i++) {
951 block = descriptor_loc(sb, logic_sb_block, i);
952 sbi->s_group_desc[i] = sb_bread(sb, block);
953 if (!sbi->s_group_desc[i]) {
954 for (j = 0; j < i; j++)
955 brelse (sbi->s_group_desc[j]);
956 printk ("EXT2-fs: unable to read group descriptors\n");
957 goto failed_mount_group_desc;
960 if (!ext2_check_descriptors (sb)) {
961 printk ("EXT2-fs: group descriptors corrupted!\n");
964 sbi->s_gdb_count = db_count;
965 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
966 spin_lock_init(&sbi->s_next_gen_lock);
968 err = percpu_counter_init(&sbi->s_freeblocks_counter,
969 ext2_count_free_blocks(sb));
971 err = percpu_counter_init(&sbi->s_freeinodes_counter,
972 ext2_count_free_inodes(sb));
975 err = percpu_counter_init(&sbi->s_dirs_counter,
976 ext2_count_dirs(sb));
979 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
983 * set up enough so that it can read an inode
985 sb->s_op = &ext2_sops;
986 sb->s_export_op = &ext2_export_ops;
987 sb->s_xattr = ext2_xattr_handlers;
988 root = iget(sb, EXT2_ROOT_INO);
989 sb->s_root = d_alloc_root(root);
992 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
995 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
998 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1001 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1002 ext2_warning(sb, __FUNCTION__,
1003 "mounting ext3 filesystem as ext2");
1004 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1009 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1013 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1014 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1015 percpu_counter_destroy(&sbi->s_dirs_counter);
1017 for (i = 0; i < db_count; i++)
1018 brelse(sbi->s_group_desc[i]);
1019 failed_mount_group_desc:
1020 kfree(sbi->s_group_desc);
1021 kfree(sbi->s_debts);
1025 sb->s_fs_info = NULL;
1030 static void ext2_commit_super (struct super_block * sb,
1031 struct ext2_super_block * es)
1033 es->s_wtime = cpu_to_le32(get_seconds());
1034 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1038 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1040 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1041 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1042 es->s_wtime = cpu_to_le32(get_seconds());
1043 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1044 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1049 * In the second extended file system, it is not necessary to
1050 * write the super block since we use a mapping of the
1051 * disk super block in a buffer.
1053 * However, this function is still used to set the fs valid
1054 * flags to 0. We need to set this flag to 0 since the fs
1055 * may have been checked while mounted and e2fsck may have
1056 * set s_state to EXT2_VALID_FS after some corrections.
1059 void ext2_write_super (struct super_block * sb)
1061 struct ext2_super_block * es;
1063 if (!(sb->s_flags & MS_RDONLY)) {
1064 es = EXT2_SB(sb)->s_es;
1066 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1067 ext2_debug ("setting valid to 0\n");
1068 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1070 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1071 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1072 es->s_mtime = cpu_to_le32(get_seconds());
1073 ext2_sync_super(sb, es);
1075 ext2_commit_super (sb, es);
1081 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1083 struct ext2_sb_info * sbi = EXT2_SB(sb);
1084 struct ext2_super_block * es;
1085 unsigned long old_mount_opt = sbi->s_mount_opt;
1086 struct ext2_mount_options old_opts;
1087 unsigned long old_sb_flags;
1090 /* Store the old options */
1091 old_sb_flags = sb->s_flags;
1092 old_opts.s_mount_opt = sbi->s_mount_opt;
1093 old_opts.s_resuid = sbi->s_resuid;
1094 old_opts.s_resgid = sbi->s_resgid;
1097 * Allow the "check" option to be passed as a remount option.
1099 if (!parse_options (data, sbi)) {
1104 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1105 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1107 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1108 EXT2_MOUNT_XIP if not */
1110 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1111 printk("XIP: Unsupported blocksize\n");
1117 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1118 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1119 invalidate_inodes(sb))
1120 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1121 "xip remain in cache (no functional problem)");
1122 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1124 if (*flags & MS_RDONLY) {
1125 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1126 !(sbi->s_mount_state & EXT2_VALID_FS))
1129 * OK, we are remounting a valid rw partition rdonly, so set
1130 * the rdonly flag and then mark the partition as valid again.
1132 es->s_state = cpu_to_le16(sbi->s_mount_state);
1133 es->s_mtime = cpu_to_le32(get_seconds());
1135 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1136 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1138 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1139 "unsupported optional features (%x).\n",
1140 sb->s_id, le32_to_cpu(ret));
1145 * Mounting a RDONLY partition read-write, so reread and
1146 * store the current valid flag. (It may have been changed
1147 * by e2fsck since we originally mounted the partition.)
1149 sbi->s_mount_state = le16_to_cpu(es->s_state);
1150 if (!ext2_setup_super (sb, es, 0))
1151 sb->s_flags &= ~MS_RDONLY;
1153 ext2_sync_super(sb, es);
1156 sbi->s_mount_opt = old_opts.s_mount_opt;
1157 sbi->s_resuid = old_opts.s_resuid;
1158 sbi->s_resgid = old_opts.s_resgid;
1159 sb->s_flags = old_sb_flags;
1163 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1165 struct super_block *sb = dentry->d_sb;
1166 struct ext2_sb_info *sbi = EXT2_SB(sb);
1167 struct ext2_super_block *es = sbi->s_es;
1170 if (test_opt (sb, MINIX_DF))
1171 sbi->s_overhead_last = 0;
1172 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1173 unsigned long i, overhead = 0;
1177 * Compute the overhead (FS structures). This is constant
1178 * for a given filesystem unless the number of block groups
1179 * changes so we cache the previous value until it does.
1183 * All of the blocks before first_data_block are
1186 overhead = le32_to_cpu(es->s_first_data_block);
1189 * Add the overhead attributed to the superblock and
1190 * block group descriptors. If the sparse superblocks
1191 * feature is turned on, then not all groups have this.
1193 for (i = 0; i < sbi->s_groups_count; i++)
1194 overhead += ext2_bg_has_super(sb, i) +
1195 ext2_bg_num_gdb(sb, i);
1198 * Every block group has an inode bitmap, a block
1199 * bitmap, and an inode table.
1201 overhead += (sbi->s_groups_count *
1202 (2 + sbi->s_itb_per_group));
1203 sbi->s_overhead_last = overhead;
1205 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1208 buf->f_type = EXT2_SUPER_MAGIC;
1209 buf->f_bsize = sb->s_blocksize;
1210 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1211 buf->f_bfree = ext2_count_free_blocks(sb);
1212 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1213 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1214 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1216 buf->f_files = le32_to_cpu(es->s_inodes_count);
1217 buf->f_ffree = ext2_count_free_inodes(sb);
1218 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1219 buf->f_namelen = EXT2_NAME_LEN;
1220 fsid = le64_to_cpup((void *)es->s_uuid) ^
1221 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1222 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1223 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1227 static int ext2_get_sb(struct file_system_type *fs_type,
1228 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1230 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1235 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1236 * acquiring the locks... As quota files are never truncated and quota code
1237 * itself serializes the operations (and noone else should touch the files)
1238 * we don't have to be afraid of races */
1239 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1240 size_t len, loff_t off)
1242 struct inode *inode = sb_dqopt(sb)->files[type];
1243 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1245 int offset = off & (sb->s_blocksize - 1);
1248 struct buffer_head tmp_bh;
1249 struct buffer_head *bh;
1250 loff_t i_size = i_size_read(inode);
1254 if (off+len > i_size)
1257 while (toread > 0) {
1258 tocopy = sb->s_blocksize - offset < toread ?
1259 sb->s_blocksize - offset : toread;
1262 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1265 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1266 memset(data, 0, tocopy);
1268 bh = sb_bread(sb, tmp_bh.b_blocknr);
1271 memcpy(data, bh->b_data+offset, tocopy);
1282 /* Write to quotafile */
1283 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1284 const char *data, size_t len, loff_t off)
1286 struct inode *inode = sb_dqopt(sb)->files[type];
1287 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1289 int offset = off & (sb->s_blocksize - 1);
1291 size_t towrite = len;
1292 struct buffer_head tmp_bh;
1293 struct buffer_head *bh;
1295 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1296 while (towrite > 0) {
1297 tocopy = sb->s_blocksize - offset < towrite ?
1298 sb->s_blocksize - offset : towrite;
1301 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1304 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1305 bh = sb_bread(sb, tmp_bh.b_blocknr);
1307 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1313 memcpy(bh->b_data+offset, data, tocopy);
1314 flush_dcache_page(bh->b_page);
1315 set_buffer_uptodate(bh);
1316 mark_buffer_dirty(bh);
1327 if (inode->i_size < off+len-towrite)
1328 i_size_write(inode, off+len-towrite);
1330 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1331 mark_inode_dirty(inode);
1332 mutex_unlock(&inode->i_mutex);
1333 return len - towrite;
1338 static struct file_system_type ext2_fs_type = {
1339 .owner = THIS_MODULE,
1341 .get_sb = ext2_get_sb,
1342 .kill_sb = kill_block_super,
1343 .fs_flags = FS_REQUIRES_DEV,
1346 static int __init init_ext2_fs(void)
1348 int err = init_ext2_xattr();
1351 err = init_inodecache();
1354 err = register_filesystem(&ext2_fs_type);
1359 destroy_inodecache();
1365 static void __exit exit_ext2_fs(void)
1367 unregister_filesystem(&ext2_fs_type);
1368 destroy_inodecache();
1372 module_init(init_ext2_fs)
1373 module_exit(exit_ext2_fs)