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 <linux/quotaops.h>
35 #include <asm/uaccess.h>
41 static void ext2_sync_super(struct super_block *sb,
42 struct ext2_super_block *es);
43 static int ext2_remount (struct super_block * sb, int * flags, char * data);
44 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
46 void ext2_error (struct super_block * sb, const char * function,
47 const char * fmt, ...)
50 struct ext2_sb_info *sbi = EXT2_SB(sb);
51 struct ext2_super_block *es = sbi->s_es;
53 if (!(sb->s_flags & MS_RDONLY)) {
54 sbi->s_mount_state |= EXT2_ERROR_FS;
55 es->s_state |= cpu_to_le16(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, __func__,
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);
120 ext2_write_super(sb);
122 ext2_xattr_put_super(sb);
123 if (!(sb->s_flags & MS_RDONLY)) {
124 struct ext2_super_block *es = sbi->s_es;
126 es->s_state = cpu_to_le16(sbi->s_mount_state);
127 ext2_sync_super(sb, es);
129 db_count = sbi->s_gdb_count;
130 for (i = 0; i < db_count; i++)
131 if (sbi->s_group_desc[i])
132 brelse (sbi->s_group_desc[i]);
133 kfree(sbi->s_group_desc);
135 percpu_counter_destroy(&sbi->s_freeblocks_counter);
136 percpu_counter_destroy(&sbi->s_freeinodes_counter);
137 percpu_counter_destroy(&sbi->s_dirs_counter);
139 sb->s_fs_info = NULL;
140 kfree(sbi->s_blockgroup_lock);
146 static struct kmem_cache * ext2_inode_cachep;
148 static struct inode *ext2_alloc_inode(struct super_block *sb)
150 struct ext2_inode_info *ei;
151 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
154 #ifdef CONFIG_EXT2_FS_POSIX_ACL
155 ei->i_acl = EXT2_ACL_NOT_CACHED;
156 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
158 ei->i_block_alloc_info = NULL;
159 ei->vfs_inode.i_version = 1;
160 return &ei->vfs_inode;
163 static void ext2_destroy_inode(struct inode *inode)
165 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
168 static void init_once(void *foo)
170 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
172 rwlock_init(&ei->i_meta_lock);
173 #ifdef CONFIG_EXT2_FS_XATTR
174 init_rwsem(&ei->xattr_sem);
176 mutex_init(&ei->truncate_mutex);
177 inode_init_once(&ei->vfs_inode);
180 static int init_inodecache(void)
182 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
183 sizeof(struct ext2_inode_info),
184 0, (SLAB_RECLAIM_ACCOUNT|
187 if (ext2_inode_cachep == NULL)
192 static void destroy_inodecache(void)
194 kmem_cache_destroy(ext2_inode_cachep);
197 static void ext2_clear_inode(struct inode *inode)
199 struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
200 #ifdef CONFIG_EXT2_FS_POSIX_ACL
201 struct ext2_inode_info *ei = EXT2_I(inode);
203 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
204 posix_acl_release(ei->i_acl);
205 ei->i_acl = EXT2_ACL_NOT_CACHED;
207 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
208 posix_acl_release(ei->i_default_acl);
209 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
212 ext2_discard_reservation(inode);
213 EXT2_I(inode)->i_block_alloc_info = NULL;
218 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
220 struct super_block *sb = vfs->mnt_sb;
221 struct ext2_sb_info *sbi = EXT2_SB(sb);
222 struct ext2_super_block *es = sbi->s_es;
223 unsigned long def_mount_opts;
225 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
227 if (sbi->s_sb_block != 1)
228 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
229 if (test_opt(sb, MINIX_DF))
230 seq_puts(seq, ",minixdf");
231 if (test_opt(sb, GRPID))
232 seq_puts(seq, ",grpid");
233 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
234 seq_puts(seq, ",nogrpid");
235 if (sbi->s_resuid != EXT2_DEF_RESUID ||
236 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
237 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
239 if (sbi->s_resgid != EXT2_DEF_RESGID ||
240 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
241 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
243 if (test_opt(sb, ERRORS_RO)) {
244 int def_errors = le16_to_cpu(es->s_errors);
246 if (def_errors == EXT2_ERRORS_PANIC ||
247 def_errors == EXT2_ERRORS_CONTINUE) {
248 seq_puts(seq, ",errors=remount-ro");
251 if (test_opt(sb, ERRORS_CONT))
252 seq_puts(seq, ",errors=continue");
253 if (test_opt(sb, ERRORS_PANIC))
254 seq_puts(seq, ",errors=panic");
255 if (test_opt(sb, NO_UID32))
256 seq_puts(seq, ",nouid32");
257 if (test_opt(sb, DEBUG))
258 seq_puts(seq, ",debug");
259 if (test_opt(sb, OLDALLOC))
260 seq_puts(seq, ",oldalloc");
262 #ifdef CONFIG_EXT2_FS_XATTR
263 if (test_opt(sb, XATTR_USER))
264 seq_puts(seq, ",user_xattr");
265 if (!test_opt(sb, XATTR_USER) &&
266 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
267 seq_puts(seq, ",nouser_xattr");
271 #ifdef CONFIG_EXT2_FS_POSIX_ACL
272 if (test_opt(sb, POSIX_ACL))
273 seq_puts(seq, ",acl");
274 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
275 seq_puts(seq, ",noacl");
278 if (test_opt(sb, NOBH))
279 seq_puts(seq, ",nobh");
281 #if defined(CONFIG_QUOTA)
282 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
283 seq_puts(seq, ",usrquota");
285 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
286 seq_puts(seq, ",grpquota");
289 #if defined(CONFIG_EXT2_FS_XIP)
290 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
291 seq_puts(seq, ",xip");
294 if (!test_opt(sb, RESERVATION))
295 seq_puts(seq, ",noreservation");
301 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
302 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
305 static const struct super_operations ext2_sops = {
306 .alloc_inode = ext2_alloc_inode,
307 .destroy_inode = ext2_destroy_inode,
308 .write_inode = ext2_write_inode,
309 .delete_inode = ext2_delete_inode,
310 .put_super = ext2_put_super,
311 .write_super = ext2_write_super,
312 .statfs = ext2_statfs,
313 .remount_fs = ext2_remount,
314 .clear_inode = ext2_clear_inode,
315 .show_options = ext2_show_options,
317 .quota_read = ext2_quota_read,
318 .quota_write = ext2_quota_write,
322 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
323 u64 ino, u32 generation)
327 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
328 return ERR_PTR(-ESTALE);
329 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
330 return ERR_PTR(-ESTALE);
332 /* iget isn't really right if the inode is currently unallocated!!
333 * ext2_read_inode currently does appropriate checks, but
334 * it might be "neater" to call ext2_get_inode first and check
335 * if the inode is valid.....
337 inode = ext2_iget(sb, ino);
339 return ERR_CAST(inode);
340 if (generation && inode->i_generation != generation) {
341 /* we didn't find the right inode.. */
343 return ERR_PTR(-ESTALE);
348 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
349 int fh_len, int fh_type)
351 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
355 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
356 int fh_len, int fh_type)
358 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
362 /* Yes, most of these are left as NULL!!
363 * A NULL value implies the default, which works with ext2-like file
364 * systems, but can be improved upon.
365 * Currently only get_parent is required.
367 static const struct export_operations ext2_export_ops = {
368 .fh_to_dentry = ext2_fh_to_dentry,
369 .fh_to_parent = ext2_fh_to_parent,
370 .get_parent = ext2_get_parent,
373 static unsigned long get_sb_block(void **data)
375 unsigned long sb_block;
376 char *options = (char *) *data;
378 if (!options || strncmp(options, "sb=", 3) != 0)
379 return 1; /* Default location */
381 sb_block = simple_strtoul(options, &options, 0);
382 if (*options && *options != ',') {
383 printk("EXT2-fs: Invalid sb specification: %s\n",
389 *data = (void *) options;
394 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
395 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
396 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
397 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
398 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
399 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
402 static const match_table_t tokens = {
403 {Opt_bsd_df, "bsddf"},
404 {Opt_minix_df, "minixdf"},
405 {Opt_grpid, "grpid"},
406 {Opt_grpid, "bsdgroups"},
407 {Opt_nogrpid, "nogrpid"},
408 {Opt_nogrpid, "sysvgroups"},
409 {Opt_resgid, "resgid=%u"},
410 {Opt_resuid, "resuid=%u"},
412 {Opt_err_cont, "errors=continue"},
413 {Opt_err_panic, "errors=panic"},
414 {Opt_err_ro, "errors=remount-ro"},
415 {Opt_nouid32, "nouid32"},
416 {Opt_nocheck, "check=none"},
417 {Opt_nocheck, "nocheck"},
418 {Opt_debug, "debug"},
419 {Opt_oldalloc, "oldalloc"},
420 {Opt_orlov, "orlov"},
422 {Opt_user_xattr, "user_xattr"},
423 {Opt_nouser_xattr, "nouser_xattr"},
425 {Opt_noacl, "noacl"},
427 {Opt_grpquota, "grpquota"},
428 {Opt_ignore, "noquota"},
429 {Opt_quota, "quota"},
430 {Opt_usrquota, "usrquota"},
431 {Opt_reservation, "reservation"},
432 {Opt_noreservation, "noreservation"},
436 static int parse_options (char * options,
437 struct ext2_sb_info *sbi)
440 substring_t args[MAX_OPT_ARGS];
446 while ((p = strsep (&options, ",")) != NULL) {
451 token = match_token(p, tokens, args);
454 clear_opt (sbi->s_mount_opt, MINIX_DF);
457 set_opt (sbi->s_mount_opt, MINIX_DF);
460 set_opt (sbi->s_mount_opt, GRPID);
463 clear_opt (sbi->s_mount_opt, GRPID);
466 if (match_int(&args[0], &option))
468 sbi->s_resuid = option;
471 if (match_int(&args[0], &option))
473 sbi->s_resgid = option;
476 /* handled by get_sb_block() instead of here */
477 /* *sb_block = match_int(&args[0]); */
480 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
481 clear_opt (sbi->s_mount_opt, ERRORS_RO);
482 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
485 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
486 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
487 set_opt (sbi->s_mount_opt, ERRORS_RO);
490 clear_opt (sbi->s_mount_opt, ERRORS_RO);
491 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
492 set_opt (sbi->s_mount_opt, ERRORS_CONT);
495 set_opt (sbi->s_mount_opt, NO_UID32);
498 clear_opt (sbi->s_mount_opt, CHECK);
501 set_opt (sbi->s_mount_opt, DEBUG);
504 set_opt (sbi->s_mount_opt, OLDALLOC);
507 clear_opt (sbi->s_mount_opt, OLDALLOC);
510 set_opt (sbi->s_mount_opt, NOBH);
512 #ifdef CONFIG_EXT2_FS_XATTR
514 set_opt (sbi->s_mount_opt, XATTR_USER);
516 case Opt_nouser_xattr:
517 clear_opt (sbi->s_mount_opt, XATTR_USER);
521 case Opt_nouser_xattr:
522 printk("EXT2 (no)user_xattr options not supported\n");
525 #ifdef CONFIG_EXT2_FS_POSIX_ACL
527 set_opt(sbi->s_mount_opt, POSIX_ACL);
530 clear_opt(sbi->s_mount_opt, POSIX_ACL);
535 printk("EXT2 (no)acl options not supported\n");
539 #ifdef CONFIG_EXT2_FS_XIP
540 set_opt (sbi->s_mount_opt, XIP);
542 printk("EXT2 xip option not supported\n");
546 #if defined(CONFIG_QUOTA)
549 set_opt(sbi->s_mount_opt, USRQUOTA);
553 set_opt(sbi->s_mount_opt, GRPQUOTA);
560 "EXT2-fs: quota operations not supported.\n");
565 case Opt_reservation:
566 set_opt(sbi->s_mount_opt, RESERVATION);
567 printk("reservations ON\n");
569 case Opt_noreservation:
570 clear_opt(sbi->s_mount_opt, RESERVATION);
571 printk("reservations OFF\n");
582 static int ext2_setup_super (struct super_block * sb,
583 struct ext2_super_block * es,
587 struct ext2_sb_info *sbi = EXT2_SB(sb);
589 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
590 printk ("EXT2-fs warning: revision level too high, "
591 "forcing read-only mode\n");
596 if (!(sbi->s_mount_state & EXT2_VALID_FS))
597 printk ("EXT2-fs warning: mounting unchecked fs, "
598 "running e2fsck is recommended\n");
599 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
600 printk ("EXT2-fs warning: mounting fs with errors, "
601 "running e2fsck is recommended\n");
602 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
603 le16_to_cpu(es->s_mnt_count) >=
604 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
605 printk ("EXT2-fs warning: maximal mount count reached, "
606 "running e2fsck is recommended\n");
607 else if (le32_to_cpu(es->s_checkinterval) &&
608 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
609 printk ("EXT2-fs warning: checktime reached, "
610 "running e2fsck is recommended\n");
611 if (!le16_to_cpu(es->s_max_mnt_count))
612 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
613 le16_add_cpu(&es->s_mnt_count, 1);
614 ext2_write_super(sb);
615 if (test_opt (sb, DEBUG))
616 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
617 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
618 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
621 EXT2_BLOCKS_PER_GROUP(sb),
622 EXT2_INODES_PER_GROUP(sb),
627 static int ext2_check_descriptors(struct super_block *sb)
630 struct ext2_sb_info *sbi = EXT2_SB(sb);
632 ext2_debug ("Checking group descriptors");
634 for (i = 0; i < sbi->s_groups_count; i++) {
635 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
636 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
637 ext2_fsblk_t last_block;
639 if (i == sbi->s_groups_count - 1)
640 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
642 last_block = first_block +
643 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
645 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
646 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
648 ext2_error (sb, "ext2_check_descriptors",
649 "Block bitmap for group %d"
650 " not in group (block %lu)!",
651 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
654 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
655 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
657 ext2_error (sb, "ext2_check_descriptors",
658 "Inode bitmap for group %d"
659 " not in group (block %lu)!",
660 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
663 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
664 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
667 ext2_error (sb, "ext2_check_descriptors",
668 "Inode table for group %d"
669 " not in group (block %lu)!",
670 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
678 * Maximal file size. There is a direct, and {,double-,triple-}indirect
679 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
680 * We need to be 1 filesystem block less than the 2^32 sector limit.
682 static loff_t ext2_max_size(int bits)
684 loff_t res = EXT2_NDIR_BLOCKS;
688 /* This is calculated to be the largest file size for a
689 * dense, file such that the total number of
690 * sectors in the file, including data and all indirect blocks,
691 * does not exceed 2^32 -1
692 * __u32 i_blocks representing the total number of
693 * 512 bytes blocks of the file
695 upper_limit = (1LL << 32) - 1;
697 /* total blocks in file system block size */
698 upper_limit >>= (bits - 9);
701 /* indirect blocks */
703 /* double indirect blocks */
704 meta_blocks += 1 + (1LL << (bits-2));
705 /* tripple indirect blocks */
706 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
708 upper_limit -= meta_blocks;
709 upper_limit <<= bits;
711 res += 1LL << (bits-2);
712 res += 1LL << (2*(bits-2));
713 res += 1LL << (3*(bits-2));
715 if (res > upper_limit)
718 if (res > MAX_LFS_FILESIZE)
719 res = MAX_LFS_FILESIZE;
724 static unsigned long descriptor_loc(struct super_block *sb,
725 unsigned long logic_sb_block,
728 struct ext2_sb_info *sbi = EXT2_SB(sb);
729 unsigned long bg, first_meta_bg;
732 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
734 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
736 return (logic_sb_block + nr + 1);
737 bg = sbi->s_desc_per_block * nr;
738 if (ext2_bg_has_super(sb, bg))
741 return ext2_group_first_block_no(sb, bg) + has_super;
744 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
746 struct buffer_head * bh;
747 struct ext2_sb_info * sbi;
748 struct ext2_super_block * es;
751 unsigned long sb_block = get_sb_block(&data);
752 unsigned long logic_sb_block;
753 unsigned long offset = 0;
754 unsigned long def_mount_opts;
756 int blocksize = BLOCK_SIZE;
762 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
766 sbi->s_blockgroup_lock =
767 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
768 if (!sbi->s_blockgroup_lock) {
773 sbi->s_sb_block = sb_block;
776 * See what the current blocksize for the device is, and
777 * use that as the blocksize. Otherwise (or if the blocksize
778 * is smaller than the default) use the default.
779 * This is important for devices that have a hardware
780 * sectorsize that is larger than the default.
782 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
784 printk ("EXT2-fs: unable to set blocksize\n");
789 * If the superblock doesn't start on a hardware sector boundary,
790 * calculate the offset.
792 if (blocksize != BLOCK_SIZE) {
793 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
794 offset = (sb_block*BLOCK_SIZE) % blocksize;
796 logic_sb_block = sb_block;
799 if (!(bh = sb_bread(sb, logic_sb_block))) {
800 printk ("EXT2-fs: unable to read superblock\n");
804 * Note: s_es must be initialized as soon as possible because
805 * some ext2 macro-instructions depend on its value
807 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
809 sb->s_magic = le16_to_cpu(es->s_magic);
811 if (sb->s_magic != EXT2_SUPER_MAGIC)
814 /* Set defaults before we parse the mount options */
815 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
816 if (def_mount_opts & EXT2_DEFM_DEBUG)
817 set_opt(sbi->s_mount_opt, DEBUG);
818 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
819 set_opt(sbi->s_mount_opt, GRPID);
820 if (def_mount_opts & EXT2_DEFM_UID16)
821 set_opt(sbi->s_mount_opt, NO_UID32);
822 #ifdef CONFIG_EXT2_FS_XATTR
823 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
824 set_opt(sbi->s_mount_opt, XATTR_USER);
826 #ifdef CONFIG_EXT2_FS_POSIX_ACL
827 if (def_mount_opts & EXT2_DEFM_ACL)
828 set_opt(sbi->s_mount_opt, POSIX_ACL);
831 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
832 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
833 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
834 set_opt(sbi->s_mount_opt, ERRORS_CONT);
836 set_opt(sbi->s_mount_opt, ERRORS_RO);
838 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
839 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
841 set_opt(sbi->s_mount_opt, RESERVATION);
843 if (!parse_options ((char *) data, sbi))
846 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
847 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
850 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
851 EXT2_MOUNT_XIP if not */
853 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
854 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
855 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
856 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
857 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
858 "running e2fsck is recommended\n");
860 * Check feature flags regardless of the revision level, since we
861 * previously didn't change the revision level when setting the flags,
862 * so there is a chance incompat flags are set on a rev 0 filesystem.
864 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
866 printk("EXT2-fs: %s: couldn't mount because of "
867 "unsupported optional features (%x).\n",
868 sb->s_id, le32_to_cpu(features));
871 if (!(sb->s_flags & MS_RDONLY) &&
872 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
873 printk("EXT2-fs: %s: couldn't mount RDWR because of "
874 "unsupported optional features (%x).\n",
875 sb->s_id, le32_to_cpu(features));
879 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
881 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
883 printk("XIP: Unsupported blocksize\n");
887 /* If the blocksize doesn't match, re-read the thing.. */
888 if (sb->s_blocksize != blocksize) {
891 if (!sb_set_blocksize(sb, blocksize)) {
892 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
896 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
897 offset = (sb_block*BLOCK_SIZE) % blocksize;
898 bh = sb_bread(sb, logic_sb_block);
900 printk("EXT2-fs: Couldn't read superblock on "
904 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
906 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
907 printk ("EXT2-fs: Magic mismatch, very weird !\n");
912 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
914 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
915 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
916 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
918 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
919 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
920 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
921 !is_power_of_2(sbi->s_inode_size) ||
922 (sbi->s_inode_size > blocksize)) {
923 printk ("EXT2-fs: unsupported inode size: %d\n",
929 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
930 le32_to_cpu(es->s_log_frag_size);
931 if (sbi->s_frag_size == 0)
933 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
935 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
936 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
937 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
939 if (EXT2_INODE_SIZE(sb) == 0)
941 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
942 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
944 sbi->s_itb_per_group = sbi->s_inodes_per_group /
945 sbi->s_inodes_per_block;
946 sbi->s_desc_per_block = sb->s_blocksize /
947 sizeof (struct ext2_group_desc);
949 sbi->s_mount_state = le16_to_cpu(es->s_state);
950 sbi->s_addr_per_block_bits =
951 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
952 sbi->s_desc_per_block_bits =
953 ilog2 (EXT2_DESC_PER_BLOCK(sb));
955 if (sb->s_magic != EXT2_SUPER_MAGIC)
958 if (sb->s_blocksize != bh->b_size) {
960 printk ("VFS: Unsupported blocksize on dev "
965 if (sb->s_blocksize != sbi->s_frag_size) {
966 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
967 sbi->s_frag_size, sb->s_blocksize);
971 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
972 printk ("EXT2-fs: #blocks per group too big: %lu\n",
973 sbi->s_blocks_per_group);
976 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
977 printk ("EXT2-fs: #fragments per group too big: %lu\n",
978 sbi->s_frags_per_group);
981 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
982 printk ("EXT2-fs: #inodes per group too big: %lu\n",
983 sbi->s_inodes_per_group);
987 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
989 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
990 le32_to_cpu(es->s_first_data_block) - 1)
991 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
992 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
993 EXT2_DESC_PER_BLOCK(sb);
994 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
995 if (sbi->s_group_desc == NULL) {
996 printk ("EXT2-fs: not enough memory\n");
999 bgl_lock_init(sbi->s_blockgroup_lock);
1000 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1001 if (!sbi->s_debts) {
1002 printk ("EXT2-fs: not enough memory\n");
1003 goto failed_mount_group_desc;
1005 for (i = 0; i < db_count; i++) {
1006 block = descriptor_loc(sb, logic_sb_block, i);
1007 sbi->s_group_desc[i] = sb_bread(sb, block);
1008 if (!sbi->s_group_desc[i]) {
1009 for (j = 0; j < i; j++)
1010 brelse (sbi->s_group_desc[j]);
1011 printk ("EXT2-fs: unable to read group descriptors\n");
1012 goto failed_mount_group_desc;
1015 if (!ext2_check_descriptors (sb)) {
1016 printk ("EXT2-fs: group descriptors corrupted!\n");
1019 sbi->s_gdb_count = db_count;
1020 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1021 spin_lock_init(&sbi->s_next_gen_lock);
1023 /* per fileystem reservation list head & lock */
1024 spin_lock_init(&sbi->s_rsv_window_lock);
1025 sbi->s_rsv_window_root = RB_ROOT;
1027 * Add a single, static dummy reservation to the start of the
1028 * reservation window list --- it gives us a placeholder for
1029 * append-at-start-of-list which makes the allocation logic
1032 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1033 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1034 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1035 sbi->s_rsv_window_head.rsv_goal_size = 0;
1036 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1038 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1039 ext2_count_free_blocks(sb));
1041 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1042 ext2_count_free_inodes(sb));
1045 err = percpu_counter_init(&sbi->s_dirs_counter,
1046 ext2_count_dirs(sb));
1049 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1053 * set up enough so that it can read an inode
1055 sb->s_op = &ext2_sops;
1056 sb->s_export_op = &ext2_export_ops;
1057 sb->s_xattr = ext2_xattr_handlers;
1058 root = ext2_iget(sb, EXT2_ROOT_INO);
1060 ret = PTR_ERR(root);
1063 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1065 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1069 sb->s_root = d_alloc_root(root);
1072 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1076 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1077 ext2_warning(sb, __func__,
1078 "mounting ext3 filesystem as ext2");
1079 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1084 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1088 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1089 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1090 percpu_counter_destroy(&sbi->s_dirs_counter);
1092 for (i = 0; i < db_count; i++)
1093 brelse(sbi->s_group_desc[i]);
1094 failed_mount_group_desc:
1095 kfree(sbi->s_group_desc);
1096 kfree(sbi->s_debts);
1100 sb->s_fs_info = NULL;
1101 kfree(sbi->s_blockgroup_lock);
1106 static void ext2_commit_super (struct super_block * sb,
1107 struct ext2_super_block * es)
1109 es->s_wtime = cpu_to_le32(get_seconds());
1110 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1114 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1116 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1117 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1118 es->s_wtime = cpu_to_le32(get_seconds());
1119 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1120 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1125 * In the second extended file system, it is not necessary to
1126 * write the super block since we use a mapping of the
1127 * disk super block in a buffer.
1129 * However, this function is still used to set the fs valid
1130 * flags to 0. We need to set this flag to 0 since the fs
1131 * may have been checked while mounted and e2fsck may have
1132 * set s_state to EXT2_VALID_FS after some corrections.
1135 void ext2_write_super (struct super_block * sb)
1137 struct ext2_super_block * es;
1139 if (!(sb->s_flags & MS_RDONLY)) {
1140 es = EXT2_SB(sb)->s_es;
1142 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1143 ext2_debug ("setting valid to 0\n");
1144 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1145 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1146 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1147 es->s_mtime = cpu_to_le32(get_seconds());
1148 ext2_sync_super(sb, es);
1150 ext2_commit_super (sb, es);
1156 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1158 struct ext2_sb_info * sbi = EXT2_SB(sb);
1159 struct ext2_super_block * es;
1160 unsigned long old_mount_opt = sbi->s_mount_opt;
1161 struct ext2_mount_options old_opts;
1162 unsigned long old_sb_flags;
1165 /* Store the old options */
1166 old_sb_flags = sb->s_flags;
1167 old_opts.s_mount_opt = sbi->s_mount_opt;
1168 old_opts.s_resuid = sbi->s_resuid;
1169 old_opts.s_resgid = sbi->s_resgid;
1172 * Allow the "check" option to be passed as a remount option.
1174 if (!parse_options (data, sbi)) {
1179 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1180 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1182 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1183 EXT2_MOUNT_XIP if not */
1185 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1186 printk("XIP: Unsupported blocksize\n");
1192 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1193 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1194 invalidate_inodes(sb)) {
1195 ext2_warning(sb, __func__, "refusing change of xip flag "
1196 "with busy inodes while remounting");
1197 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1198 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1200 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1202 if (*flags & MS_RDONLY) {
1203 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1204 !(sbi->s_mount_state & EXT2_VALID_FS))
1207 * OK, we are remounting a valid rw partition rdonly, so set
1208 * the rdonly flag and then mark the partition as valid again.
1210 es->s_state = cpu_to_le16(sbi->s_mount_state);
1211 es->s_mtime = cpu_to_le32(get_seconds());
1213 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1214 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1216 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1217 "unsupported optional features (%x).\n",
1218 sb->s_id, le32_to_cpu(ret));
1223 * Mounting a RDONLY partition read-write, so reread and
1224 * store the current valid flag. (It may have been changed
1225 * by e2fsck since we originally mounted the partition.)
1227 sbi->s_mount_state = le16_to_cpu(es->s_state);
1228 if (!ext2_setup_super (sb, es, 0))
1229 sb->s_flags &= ~MS_RDONLY;
1231 ext2_sync_super(sb, es);
1234 sbi->s_mount_opt = old_opts.s_mount_opt;
1235 sbi->s_resuid = old_opts.s_resuid;
1236 sbi->s_resgid = old_opts.s_resgid;
1237 sb->s_flags = old_sb_flags;
1241 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1243 struct super_block *sb = dentry->d_sb;
1244 struct ext2_sb_info *sbi = EXT2_SB(sb);
1245 struct ext2_super_block *es = sbi->s_es;
1248 if (test_opt (sb, MINIX_DF))
1249 sbi->s_overhead_last = 0;
1250 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1251 unsigned long i, overhead = 0;
1255 * Compute the overhead (FS structures). This is constant
1256 * for a given filesystem unless the number of block groups
1257 * changes so we cache the previous value until it does.
1261 * All of the blocks before first_data_block are
1264 overhead = le32_to_cpu(es->s_first_data_block);
1267 * Add the overhead attributed to the superblock and
1268 * block group descriptors. If the sparse superblocks
1269 * feature is turned on, then not all groups have this.
1271 for (i = 0; i < sbi->s_groups_count; i++)
1272 overhead += ext2_bg_has_super(sb, i) +
1273 ext2_bg_num_gdb(sb, i);
1276 * Every block group has an inode bitmap, a block
1277 * bitmap, and an inode table.
1279 overhead += (sbi->s_groups_count *
1280 (2 + sbi->s_itb_per_group));
1281 sbi->s_overhead_last = overhead;
1283 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1286 buf->f_type = EXT2_SUPER_MAGIC;
1287 buf->f_bsize = sb->s_blocksize;
1288 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1289 buf->f_bfree = ext2_count_free_blocks(sb);
1290 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1291 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1292 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1294 buf->f_files = le32_to_cpu(es->s_inodes_count);
1295 buf->f_ffree = ext2_count_free_inodes(sb);
1296 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1297 buf->f_namelen = EXT2_NAME_LEN;
1298 fsid = le64_to_cpup((void *)es->s_uuid) ^
1299 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1300 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1301 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1305 static int ext2_get_sb(struct file_system_type *fs_type,
1306 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1308 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1313 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1314 * acquiring the locks... As quota files are never truncated and quota code
1315 * itself serializes the operations (and noone else should touch the files)
1316 * we don't have to be afraid of races */
1317 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1318 size_t len, loff_t off)
1320 struct inode *inode = sb_dqopt(sb)->files[type];
1321 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1323 int offset = off & (sb->s_blocksize - 1);
1326 struct buffer_head tmp_bh;
1327 struct buffer_head *bh;
1328 loff_t i_size = i_size_read(inode);
1332 if (off+len > i_size)
1335 while (toread > 0) {
1336 tocopy = sb->s_blocksize - offset < toread ?
1337 sb->s_blocksize - offset : toread;
1340 tmp_bh.b_size = sb->s_blocksize;
1341 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1344 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1345 memset(data, 0, tocopy);
1347 bh = sb_bread(sb, tmp_bh.b_blocknr);
1350 memcpy(data, bh->b_data+offset, tocopy);
1361 /* Write to quotafile */
1362 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1363 const char *data, size_t len, loff_t off)
1365 struct inode *inode = sb_dqopt(sb)->files[type];
1366 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1368 int offset = off & (sb->s_blocksize - 1);
1370 size_t towrite = len;
1371 struct buffer_head tmp_bh;
1372 struct buffer_head *bh;
1374 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1375 while (towrite > 0) {
1376 tocopy = sb->s_blocksize - offset < towrite ?
1377 sb->s_blocksize - offset : towrite;
1380 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1383 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1384 bh = sb_bread(sb, tmp_bh.b_blocknr);
1386 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1392 memcpy(bh->b_data+offset, data, tocopy);
1393 flush_dcache_page(bh->b_page);
1394 set_buffer_uptodate(bh);
1395 mark_buffer_dirty(bh);
1404 if (len == towrite) {
1405 mutex_unlock(&inode->i_mutex);
1408 if (inode->i_size < off+len-towrite)
1409 i_size_write(inode, off+len-towrite);
1411 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1412 mark_inode_dirty(inode);
1413 mutex_unlock(&inode->i_mutex);
1414 return len - towrite;
1419 static struct file_system_type ext2_fs_type = {
1420 .owner = THIS_MODULE,
1422 .get_sb = ext2_get_sb,
1423 .kill_sb = kill_block_super,
1424 .fs_flags = FS_REQUIRES_DEV,
1427 static int __init init_ext2_fs(void)
1429 int err = init_ext2_xattr();
1432 err = init_inodecache();
1435 err = register_filesystem(&ext2_fs_type);
1440 destroy_inodecache();
1446 static void __exit exit_ext2_fs(void)
1448 unregister_filesystem(&ext2_fs_type);
1449 destroy_inodecache();
1453 module_init(init_ext2_fs)
1454 module_exit(exit_ext2_fs)