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);
118 ext2_write_super(sb);
120 ext2_xattr_put_super(sb);
121 if (!(sb->s_flags & MS_RDONLY)) {
122 struct ext2_super_block *es = sbi->s_es;
124 es->s_state = cpu_to_le16(sbi->s_mount_state);
125 ext2_sync_super(sb, es);
127 db_count = sbi->s_gdb_count;
128 for (i = 0; i < db_count; i++)
129 if (sbi->s_group_desc[i])
130 brelse (sbi->s_group_desc[i]);
131 kfree(sbi->s_group_desc);
133 percpu_counter_destroy(&sbi->s_freeblocks_counter);
134 percpu_counter_destroy(&sbi->s_freeinodes_counter);
135 percpu_counter_destroy(&sbi->s_dirs_counter);
137 sb->s_fs_info = NULL;
138 kfree(sbi->s_blockgroup_lock);
144 static struct kmem_cache * ext2_inode_cachep;
146 static struct inode *ext2_alloc_inode(struct super_block *sb)
148 struct ext2_inode_info *ei;
149 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
152 #ifdef CONFIG_EXT2_FS_POSIX_ACL
153 ei->i_acl = EXT2_ACL_NOT_CACHED;
154 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
156 ei->i_block_alloc_info = NULL;
157 ei->vfs_inode.i_version = 1;
158 return &ei->vfs_inode;
161 static void ext2_destroy_inode(struct inode *inode)
163 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
166 static void init_once(void *foo)
168 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
170 rwlock_init(&ei->i_meta_lock);
171 #ifdef CONFIG_EXT2_FS_XATTR
172 init_rwsem(&ei->xattr_sem);
174 mutex_init(&ei->truncate_mutex);
175 inode_init_once(&ei->vfs_inode);
178 static int init_inodecache(void)
180 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
181 sizeof(struct ext2_inode_info),
182 0, (SLAB_RECLAIM_ACCOUNT|
185 if (ext2_inode_cachep == NULL)
190 static void destroy_inodecache(void)
192 kmem_cache_destroy(ext2_inode_cachep);
195 static void ext2_clear_inode(struct inode *inode)
197 struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
198 #ifdef CONFIG_EXT2_FS_POSIX_ACL
199 struct ext2_inode_info *ei = EXT2_I(inode);
201 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
202 posix_acl_release(ei->i_acl);
203 ei->i_acl = EXT2_ACL_NOT_CACHED;
205 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
206 posix_acl_release(ei->i_default_acl);
207 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
210 ext2_discard_reservation(inode);
211 EXT2_I(inode)->i_block_alloc_info = NULL;
216 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
218 struct super_block *sb = vfs->mnt_sb;
219 struct ext2_sb_info *sbi = EXT2_SB(sb);
220 struct ext2_super_block *es = sbi->s_es;
221 unsigned long def_mount_opts;
223 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
225 if (sbi->s_sb_block != 1)
226 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
227 if (test_opt(sb, MINIX_DF))
228 seq_puts(seq, ",minixdf");
229 if (test_opt(sb, GRPID))
230 seq_puts(seq, ",grpid");
231 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
232 seq_puts(seq, ",nogrpid");
233 if (sbi->s_resuid != EXT2_DEF_RESUID ||
234 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
235 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
237 if (sbi->s_resgid != EXT2_DEF_RESGID ||
238 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
239 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
241 if (test_opt(sb, ERRORS_RO)) {
242 int def_errors = le16_to_cpu(es->s_errors);
244 if (def_errors == EXT2_ERRORS_PANIC ||
245 def_errors == EXT2_ERRORS_CONTINUE) {
246 seq_puts(seq, ",errors=remount-ro");
249 if (test_opt(sb, ERRORS_CONT))
250 seq_puts(seq, ",errors=continue");
251 if (test_opt(sb, ERRORS_PANIC))
252 seq_puts(seq, ",errors=panic");
253 if (test_opt(sb, NO_UID32))
254 seq_puts(seq, ",nouid32");
255 if (test_opt(sb, DEBUG))
256 seq_puts(seq, ",debug");
257 if (test_opt(sb, OLDALLOC))
258 seq_puts(seq, ",oldalloc");
260 #ifdef CONFIG_EXT2_FS_XATTR
261 if (test_opt(sb, XATTR_USER))
262 seq_puts(seq, ",user_xattr");
263 if (!test_opt(sb, XATTR_USER) &&
264 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
265 seq_puts(seq, ",nouser_xattr");
269 #ifdef CONFIG_EXT2_FS_POSIX_ACL
270 if (test_opt(sb, POSIX_ACL))
271 seq_puts(seq, ",acl");
272 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
273 seq_puts(seq, ",noacl");
276 if (test_opt(sb, NOBH))
277 seq_puts(seq, ",nobh");
279 #if defined(CONFIG_QUOTA)
280 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
281 seq_puts(seq, ",usrquota");
283 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
284 seq_puts(seq, ",grpquota");
287 #if defined(CONFIG_EXT2_FS_XIP)
288 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
289 seq_puts(seq, ",xip");
292 if (!test_opt(sb, RESERVATION))
293 seq_puts(seq, ",noreservation");
299 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
300 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
303 static const struct super_operations ext2_sops = {
304 .alloc_inode = ext2_alloc_inode,
305 .destroy_inode = ext2_destroy_inode,
306 .write_inode = ext2_write_inode,
307 .delete_inode = ext2_delete_inode,
308 .put_super = ext2_put_super,
309 .write_super = ext2_write_super,
310 .statfs = ext2_statfs,
311 .remount_fs = ext2_remount,
312 .clear_inode = ext2_clear_inode,
313 .show_options = ext2_show_options,
315 .quota_read = ext2_quota_read,
316 .quota_write = ext2_quota_write,
320 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
321 u64 ino, u32 generation)
325 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
326 return ERR_PTR(-ESTALE);
327 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
328 return ERR_PTR(-ESTALE);
330 /* iget isn't really right if the inode is currently unallocated!!
331 * ext2_read_inode currently does appropriate checks, but
332 * it might be "neater" to call ext2_get_inode first and check
333 * if the inode is valid.....
335 inode = ext2_iget(sb, ino);
337 return ERR_CAST(inode);
338 if (generation && inode->i_generation != generation) {
339 /* we didn't find the right inode.. */
341 return ERR_PTR(-ESTALE);
346 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
347 int fh_len, int fh_type)
349 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
353 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
354 int fh_len, int fh_type)
356 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
360 /* Yes, most of these are left as NULL!!
361 * A NULL value implies the default, which works with ext2-like file
362 * systems, but can be improved upon.
363 * Currently only get_parent is required.
365 static const struct export_operations ext2_export_ops = {
366 .fh_to_dentry = ext2_fh_to_dentry,
367 .fh_to_parent = ext2_fh_to_parent,
368 .get_parent = ext2_get_parent,
371 static unsigned long get_sb_block(void **data)
373 unsigned long sb_block;
374 char *options = (char *) *data;
376 if (!options || strncmp(options, "sb=", 3) != 0)
377 return 1; /* Default location */
379 sb_block = simple_strtoul(options, &options, 0);
380 if (*options && *options != ',') {
381 printk("EXT2-fs: Invalid sb specification: %s\n",
387 *data = (void *) options;
392 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
393 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
394 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
395 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
396 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
397 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
400 static const match_table_t tokens = {
401 {Opt_bsd_df, "bsddf"},
402 {Opt_minix_df, "minixdf"},
403 {Opt_grpid, "grpid"},
404 {Opt_grpid, "bsdgroups"},
405 {Opt_nogrpid, "nogrpid"},
406 {Opt_nogrpid, "sysvgroups"},
407 {Opt_resgid, "resgid=%u"},
408 {Opt_resuid, "resuid=%u"},
410 {Opt_err_cont, "errors=continue"},
411 {Opt_err_panic, "errors=panic"},
412 {Opt_err_ro, "errors=remount-ro"},
413 {Opt_nouid32, "nouid32"},
414 {Opt_nocheck, "check=none"},
415 {Opt_nocheck, "nocheck"},
416 {Opt_debug, "debug"},
417 {Opt_oldalloc, "oldalloc"},
418 {Opt_orlov, "orlov"},
420 {Opt_user_xattr, "user_xattr"},
421 {Opt_nouser_xattr, "nouser_xattr"},
423 {Opt_noacl, "noacl"},
425 {Opt_grpquota, "grpquota"},
426 {Opt_ignore, "noquota"},
427 {Opt_quota, "quota"},
428 {Opt_usrquota, "usrquota"},
429 {Opt_reservation, "reservation"},
430 {Opt_noreservation, "noreservation"},
434 static int parse_options (char * options,
435 struct ext2_sb_info *sbi)
438 substring_t args[MAX_OPT_ARGS];
444 while ((p = strsep (&options, ",")) != NULL) {
449 token = match_token(p, tokens, args);
452 clear_opt (sbi->s_mount_opt, MINIX_DF);
455 set_opt (sbi->s_mount_opt, MINIX_DF);
458 set_opt (sbi->s_mount_opt, GRPID);
461 clear_opt (sbi->s_mount_opt, GRPID);
464 if (match_int(&args[0], &option))
466 sbi->s_resuid = option;
469 if (match_int(&args[0], &option))
471 sbi->s_resgid = option;
474 /* handled by get_sb_block() instead of here */
475 /* *sb_block = match_int(&args[0]); */
478 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
479 clear_opt (sbi->s_mount_opt, ERRORS_RO);
480 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
483 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
484 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
485 set_opt (sbi->s_mount_opt, ERRORS_RO);
488 clear_opt (sbi->s_mount_opt, ERRORS_RO);
489 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
490 set_opt (sbi->s_mount_opt, ERRORS_CONT);
493 set_opt (sbi->s_mount_opt, NO_UID32);
496 clear_opt (sbi->s_mount_opt, CHECK);
499 set_opt (sbi->s_mount_opt, DEBUG);
502 set_opt (sbi->s_mount_opt, OLDALLOC);
505 clear_opt (sbi->s_mount_opt, OLDALLOC);
508 set_opt (sbi->s_mount_opt, NOBH);
510 #ifdef CONFIG_EXT2_FS_XATTR
512 set_opt (sbi->s_mount_opt, XATTR_USER);
514 case Opt_nouser_xattr:
515 clear_opt (sbi->s_mount_opt, XATTR_USER);
519 case Opt_nouser_xattr:
520 printk("EXT2 (no)user_xattr options not supported\n");
523 #ifdef CONFIG_EXT2_FS_POSIX_ACL
525 set_opt(sbi->s_mount_opt, POSIX_ACL);
528 clear_opt(sbi->s_mount_opt, POSIX_ACL);
533 printk("EXT2 (no)acl options not supported\n");
537 #ifdef CONFIG_EXT2_FS_XIP
538 set_opt (sbi->s_mount_opt, XIP);
540 printk("EXT2 xip option not supported\n");
544 #if defined(CONFIG_QUOTA)
547 set_opt(sbi->s_mount_opt, USRQUOTA);
551 set_opt(sbi->s_mount_opt, GRPQUOTA);
558 "EXT2-fs: quota operations not supported.\n");
563 case Opt_reservation:
564 set_opt(sbi->s_mount_opt, RESERVATION);
565 printk("reservations ON\n");
567 case Opt_noreservation:
568 clear_opt(sbi->s_mount_opt, RESERVATION);
569 printk("reservations OFF\n");
580 static int ext2_setup_super (struct super_block * sb,
581 struct ext2_super_block * es,
585 struct ext2_sb_info *sbi = EXT2_SB(sb);
587 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
588 printk ("EXT2-fs warning: revision level too high, "
589 "forcing read-only mode\n");
594 if (!(sbi->s_mount_state & EXT2_VALID_FS))
595 printk ("EXT2-fs warning: mounting unchecked fs, "
596 "running e2fsck is recommended\n");
597 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
598 printk ("EXT2-fs warning: mounting fs with errors, "
599 "running e2fsck is recommended\n");
600 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
601 le16_to_cpu(es->s_mnt_count) >=
602 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
603 printk ("EXT2-fs warning: maximal mount count reached, "
604 "running e2fsck is recommended\n");
605 else if (le32_to_cpu(es->s_checkinterval) &&
606 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
607 printk ("EXT2-fs warning: checktime reached, "
608 "running e2fsck is recommended\n");
609 if (!le16_to_cpu(es->s_max_mnt_count))
610 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
611 le16_add_cpu(&es->s_mnt_count, 1);
612 ext2_write_super(sb);
613 if (test_opt (sb, DEBUG))
614 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
615 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
616 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
619 EXT2_BLOCKS_PER_GROUP(sb),
620 EXT2_INODES_PER_GROUP(sb),
625 static int ext2_check_descriptors(struct super_block *sb)
628 struct ext2_sb_info *sbi = EXT2_SB(sb);
630 ext2_debug ("Checking group descriptors");
632 for (i = 0; i < sbi->s_groups_count; i++) {
633 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
634 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
635 ext2_fsblk_t last_block;
637 if (i == sbi->s_groups_count - 1)
638 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
640 last_block = first_block +
641 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
643 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
644 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
646 ext2_error (sb, "ext2_check_descriptors",
647 "Block bitmap for group %d"
648 " not in group (block %lu)!",
649 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
652 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
653 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
655 ext2_error (sb, "ext2_check_descriptors",
656 "Inode bitmap for group %d"
657 " not in group (block %lu)!",
658 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
661 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
662 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
665 ext2_error (sb, "ext2_check_descriptors",
666 "Inode table for group %d"
667 " not in group (block %lu)!",
668 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
676 * Maximal file size. There is a direct, and {,double-,triple-}indirect
677 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
678 * We need to be 1 filesystem block less than the 2^32 sector limit.
680 static loff_t ext2_max_size(int bits)
682 loff_t res = EXT2_NDIR_BLOCKS;
686 /* This is calculated to be the largest file size for a
687 * dense, file such that the total number of
688 * sectors in the file, including data and all indirect blocks,
689 * does not exceed 2^32 -1
690 * __u32 i_blocks representing the total number of
691 * 512 bytes blocks of the file
693 upper_limit = (1LL << 32) - 1;
695 /* total blocks in file system block size */
696 upper_limit >>= (bits - 9);
699 /* indirect blocks */
701 /* double indirect blocks */
702 meta_blocks += 1 + (1LL << (bits-2));
703 /* tripple indirect blocks */
704 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
706 upper_limit -= meta_blocks;
707 upper_limit <<= bits;
709 res += 1LL << (bits-2);
710 res += 1LL << (2*(bits-2));
711 res += 1LL << (3*(bits-2));
713 if (res > upper_limit)
716 if (res > MAX_LFS_FILESIZE)
717 res = MAX_LFS_FILESIZE;
722 static unsigned long descriptor_loc(struct super_block *sb,
723 unsigned long logic_sb_block,
726 struct ext2_sb_info *sbi = EXT2_SB(sb);
727 unsigned long bg, first_meta_bg;
730 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
732 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
734 return (logic_sb_block + nr + 1);
735 bg = sbi->s_desc_per_block * nr;
736 if (ext2_bg_has_super(sb, bg))
739 return ext2_group_first_block_no(sb, bg) + has_super;
742 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
744 struct buffer_head * bh;
745 struct ext2_sb_info * sbi;
746 struct ext2_super_block * es;
749 unsigned long sb_block = get_sb_block(&data);
750 unsigned long logic_sb_block;
751 unsigned long offset = 0;
752 unsigned long def_mount_opts;
754 int blocksize = BLOCK_SIZE;
760 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
764 sbi->s_blockgroup_lock =
765 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
766 if (!sbi->s_blockgroup_lock) {
771 sbi->s_sb_block = sb_block;
774 * See what the current blocksize for the device is, and
775 * use that as the blocksize. Otherwise (or if the blocksize
776 * is smaller than the default) use the default.
777 * This is important for devices that have a hardware
778 * sectorsize that is larger than the default.
780 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
782 printk ("EXT2-fs: unable to set blocksize\n");
787 * If the superblock doesn't start on a hardware sector boundary,
788 * calculate the offset.
790 if (blocksize != BLOCK_SIZE) {
791 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
792 offset = (sb_block*BLOCK_SIZE) % blocksize;
794 logic_sb_block = sb_block;
797 if (!(bh = sb_bread(sb, logic_sb_block))) {
798 printk ("EXT2-fs: unable to read superblock\n");
802 * Note: s_es must be initialized as soon as possible because
803 * some ext2 macro-instructions depend on its value
805 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
807 sb->s_magic = le16_to_cpu(es->s_magic);
809 if (sb->s_magic != EXT2_SUPER_MAGIC)
812 /* Set defaults before we parse the mount options */
813 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
814 if (def_mount_opts & EXT2_DEFM_DEBUG)
815 set_opt(sbi->s_mount_opt, DEBUG);
816 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
817 set_opt(sbi->s_mount_opt, GRPID);
818 if (def_mount_opts & EXT2_DEFM_UID16)
819 set_opt(sbi->s_mount_opt, NO_UID32);
820 #ifdef CONFIG_EXT2_FS_XATTR
821 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
822 set_opt(sbi->s_mount_opt, XATTR_USER);
824 #ifdef CONFIG_EXT2_FS_POSIX_ACL
825 if (def_mount_opts & EXT2_DEFM_ACL)
826 set_opt(sbi->s_mount_opt, POSIX_ACL);
829 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
830 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
831 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
832 set_opt(sbi->s_mount_opt, ERRORS_CONT);
834 set_opt(sbi->s_mount_opt, ERRORS_RO);
836 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
837 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
839 set_opt(sbi->s_mount_opt, RESERVATION);
841 if (!parse_options ((char *) data, sbi))
844 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
845 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
848 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
849 EXT2_MOUNT_XIP if not */
851 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
852 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
853 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
854 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
855 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
856 "running e2fsck is recommended\n");
858 * Check feature flags regardless of the revision level, since we
859 * previously didn't change the revision level when setting the flags,
860 * so there is a chance incompat flags are set on a rev 0 filesystem.
862 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
864 printk("EXT2-fs: %s: couldn't mount because of "
865 "unsupported optional features (%x).\n",
866 sb->s_id, le32_to_cpu(features));
869 if (!(sb->s_flags & MS_RDONLY) &&
870 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
871 printk("EXT2-fs: %s: couldn't mount RDWR because of "
872 "unsupported optional features (%x).\n",
873 sb->s_id, le32_to_cpu(features));
877 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
879 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
881 printk("XIP: Unsupported blocksize\n");
885 /* If the blocksize doesn't match, re-read the thing.. */
886 if (sb->s_blocksize != blocksize) {
889 if (!sb_set_blocksize(sb, blocksize)) {
890 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
894 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
895 offset = (sb_block*BLOCK_SIZE) % blocksize;
896 bh = sb_bread(sb, logic_sb_block);
898 printk("EXT2-fs: Couldn't read superblock on "
902 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
904 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
905 printk ("EXT2-fs: Magic mismatch, very weird !\n");
910 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
912 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
913 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
914 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
916 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
917 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
918 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
919 !is_power_of_2(sbi->s_inode_size) ||
920 (sbi->s_inode_size > blocksize)) {
921 printk ("EXT2-fs: unsupported inode size: %d\n",
927 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
928 le32_to_cpu(es->s_log_frag_size);
929 if (sbi->s_frag_size == 0)
931 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
933 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
934 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
935 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
937 if (EXT2_INODE_SIZE(sb) == 0)
939 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
940 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
942 sbi->s_itb_per_group = sbi->s_inodes_per_group /
943 sbi->s_inodes_per_block;
944 sbi->s_desc_per_block = sb->s_blocksize /
945 sizeof (struct ext2_group_desc);
947 sbi->s_mount_state = le16_to_cpu(es->s_state);
948 sbi->s_addr_per_block_bits =
949 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
950 sbi->s_desc_per_block_bits =
951 ilog2 (EXT2_DESC_PER_BLOCK(sb));
953 if (sb->s_magic != EXT2_SUPER_MAGIC)
956 if (sb->s_blocksize != bh->b_size) {
958 printk ("VFS: Unsupported blocksize on dev "
963 if (sb->s_blocksize != sbi->s_frag_size) {
964 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
965 sbi->s_frag_size, sb->s_blocksize);
969 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
970 printk ("EXT2-fs: #blocks per group too big: %lu\n",
971 sbi->s_blocks_per_group);
974 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
975 printk ("EXT2-fs: #fragments per group too big: %lu\n",
976 sbi->s_frags_per_group);
979 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
980 printk ("EXT2-fs: #inodes per group too big: %lu\n",
981 sbi->s_inodes_per_group);
985 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
987 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
988 le32_to_cpu(es->s_first_data_block) - 1)
989 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
990 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
991 EXT2_DESC_PER_BLOCK(sb);
992 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
993 if (sbi->s_group_desc == NULL) {
994 printk ("EXT2-fs: not enough memory\n");
997 bgl_lock_init(sbi->s_blockgroup_lock);
998 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1000 printk ("EXT2-fs: not enough memory\n");
1001 goto failed_mount_group_desc;
1003 for (i = 0; i < db_count; i++) {
1004 block = descriptor_loc(sb, logic_sb_block, i);
1005 sbi->s_group_desc[i] = sb_bread(sb, block);
1006 if (!sbi->s_group_desc[i]) {
1007 for (j = 0; j < i; j++)
1008 brelse (sbi->s_group_desc[j]);
1009 printk ("EXT2-fs: unable to read group descriptors\n");
1010 goto failed_mount_group_desc;
1013 if (!ext2_check_descriptors (sb)) {
1014 printk ("EXT2-fs: group descriptors corrupted!\n");
1017 sbi->s_gdb_count = db_count;
1018 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1019 spin_lock_init(&sbi->s_next_gen_lock);
1021 /* per fileystem reservation list head & lock */
1022 spin_lock_init(&sbi->s_rsv_window_lock);
1023 sbi->s_rsv_window_root = RB_ROOT;
1025 * Add a single, static dummy reservation to the start of the
1026 * reservation window list --- it gives us a placeholder for
1027 * append-at-start-of-list which makes the allocation logic
1030 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1031 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1032 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1033 sbi->s_rsv_window_head.rsv_goal_size = 0;
1034 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1036 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1037 ext2_count_free_blocks(sb));
1039 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1040 ext2_count_free_inodes(sb));
1043 err = percpu_counter_init(&sbi->s_dirs_counter,
1044 ext2_count_dirs(sb));
1047 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1051 * set up enough so that it can read an inode
1053 sb->s_op = &ext2_sops;
1054 sb->s_export_op = &ext2_export_ops;
1055 sb->s_xattr = ext2_xattr_handlers;
1056 root = ext2_iget(sb, EXT2_ROOT_INO);
1058 ret = PTR_ERR(root);
1061 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1063 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1067 sb->s_root = d_alloc_root(root);
1070 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1074 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1075 ext2_warning(sb, __func__,
1076 "mounting ext3 filesystem as ext2");
1077 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1082 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1086 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1087 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1088 percpu_counter_destroy(&sbi->s_dirs_counter);
1090 for (i = 0; i < db_count; i++)
1091 brelse(sbi->s_group_desc[i]);
1092 failed_mount_group_desc:
1093 kfree(sbi->s_group_desc);
1094 kfree(sbi->s_debts);
1098 sb->s_fs_info = NULL;
1099 kfree(sbi->s_blockgroup_lock);
1104 static void ext2_commit_super (struct super_block * sb,
1105 struct ext2_super_block * es)
1107 es->s_wtime = cpu_to_le32(get_seconds());
1108 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1112 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1114 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1115 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1116 es->s_wtime = cpu_to_le32(get_seconds());
1117 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1118 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1123 * In the second extended file system, it is not necessary to
1124 * write the super block since we use a mapping of the
1125 * disk super block in a buffer.
1127 * However, this function is still used to set the fs valid
1128 * flags to 0. We need to set this flag to 0 since the fs
1129 * may have been checked while mounted and e2fsck may have
1130 * set s_state to EXT2_VALID_FS after some corrections.
1133 void ext2_write_super (struct super_block * sb)
1135 struct ext2_super_block * es;
1137 if (!(sb->s_flags & MS_RDONLY)) {
1138 es = EXT2_SB(sb)->s_es;
1140 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1141 ext2_debug ("setting valid to 0\n");
1142 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1143 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1144 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1145 es->s_mtime = cpu_to_le32(get_seconds());
1146 ext2_sync_super(sb, es);
1148 ext2_commit_super (sb, es);
1154 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1156 struct ext2_sb_info * sbi = EXT2_SB(sb);
1157 struct ext2_super_block * es;
1158 unsigned long old_mount_opt = sbi->s_mount_opt;
1159 struct ext2_mount_options old_opts;
1160 unsigned long old_sb_flags;
1163 /* Store the old options */
1164 old_sb_flags = sb->s_flags;
1165 old_opts.s_mount_opt = sbi->s_mount_opt;
1166 old_opts.s_resuid = sbi->s_resuid;
1167 old_opts.s_resgid = sbi->s_resgid;
1170 * Allow the "check" option to be passed as a remount option.
1172 if (!parse_options (data, sbi)) {
1177 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1178 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1180 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1181 EXT2_MOUNT_XIP if not */
1183 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1184 printk("XIP: Unsupported blocksize\n");
1190 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1191 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1192 invalidate_inodes(sb)) {
1193 ext2_warning(sb, __func__, "refusing change of xip flag "
1194 "with busy inodes while remounting");
1195 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1196 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1198 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1200 if (*flags & MS_RDONLY) {
1201 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1202 !(sbi->s_mount_state & EXT2_VALID_FS))
1205 * OK, we are remounting a valid rw partition rdonly, so set
1206 * the rdonly flag and then mark the partition as valid again.
1208 es->s_state = cpu_to_le16(sbi->s_mount_state);
1209 es->s_mtime = cpu_to_le32(get_seconds());
1211 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1212 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1214 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1215 "unsupported optional features (%x).\n",
1216 sb->s_id, le32_to_cpu(ret));
1221 * Mounting a RDONLY partition read-write, so reread and
1222 * store the current valid flag. (It may have been changed
1223 * by e2fsck since we originally mounted the partition.)
1225 sbi->s_mount_state = le16_to_cpu(es->s_state);
1226 if (!ext2_setup_super (sb, es, 0))
1227 sb->s_flags &= ~MS_RDONLY;
1229 ext2_sync_super(sb, es);
1232 sbi->s_mount_opt = old_opts.s_mount_opt;
1233 sbi->s_resuid = old_opts.s_resuid;
1234 sbi->s_resgid = old_opts.s_resgid;
1235 sb->s_flags = old_sb_flags;
1239 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1241 struct super_block *sb = dentry->d_sb;
1242 struct ext2_sb_info *sbi = EXT2_SB(sb);
1243 struct ext2_super_block *es = sbi->s_es;
1246 if (test_opt (sb, MINIX_DF))
1247 sbi->s_overhead_last = 0;
1248 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1249 unsigned long i, overhead = 0;
1253 * Compute the overhead (FS structures). This is constant
1254 * for a given filesystem unless the number of block groups
1255 * changes so we cache the previous value until it does.
1259 * All of the blocks before first_data_block are
1262 overhead = le32_to_cpu(es->s_first_data_block);
1265 * Add the overhead attributed to the superblock and
1266 * block group descriptors. If the sparse superblocks
1267 * feature is turned on, then not all groups have this.
1269 for (i = 0; i < sbi->s_groups_count; i++)
1270 overhead += ext2_bg_has_super(sb, i) +
1271 ext2_bg_num_gdb(sb, i);
1274 * Every block group has an inode bitmap, a block
1275 * bitmap, and an inode table.
1277 overhead += (sbi->s_groups_count *
1278 (2 + sbi->s_itb_per_group));
1279 sbi->s_overhead_last = overhead;
1281 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1284 buf->f_type = EXT2_SUPER_MAGIC;
1285 buf->f_bsize = sb->s_blocksize;
1286 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1287 buf->f_bfree = ext2_count_free_blocks(sb);
1288 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1289 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1290 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1292 buf->f_files = le32_to_cpu(es->s_inodes_count);
1293 buf->f_ffree = ext2_count_free_inodes(sb);
1294 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1295 buf->f_namelen = EXT2_NAME_LEN;
1296 fsid = le64_to_cpup((void *)es->s_uuid) ^
1297 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1298 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1299 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1303 static int ext2_get_sb(struct file_system_type *fs_type,
1304 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1306 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1311 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1312 * acquiring the locks... As quota files are never truncated and quota code
1313 * itself serializes the operations (and noone else should touch the files)
1314 * we don't have to be afraid of races */
1315 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1316 size_t len, loff_t off)
1318 struct inode *inode = sb_dqopt(sb)->files[type];
1319 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1321 int offset = off & (sb->s_blocksize - 1);
1324 struct buffer_head tmp_bh;
1325 struct buffer_head *bh;
1326 loff_t i_size = i_size_read(inode);
1330 if (off+len > i_size)
1333 while (toread > 0) {
1334 tocopy = sb->s_blocksize - offset < toread ?
1335 sb->s_blocksize - offset : toread;
1338 tmp_bh.b_size = sb->s_blocksize;
1339 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1342 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1343 memset(data, 0, tocopy);
1345 bh = sb_bread(sb, tmp_bh.b_blocknr);
1348 memcpy(data, bh->b_data+offset, tocopy);
1359 /* Write to quotafile */
1360 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1361 const char *data, size_t len, loff_t off)
1363 struct inode *inode = sb_dqopt(sb)->files[type];
1364 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1366 int offset = off & (sb->s_blocksize - 1);
1368 size_t towrite = len;
1369 struct buffer_head tmp_bh;
1370 struct buffer_head *bh;
1372 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1373 while (towrite > 0) {
1374 tocopy = sb->s_blocksize - offset < towrite ?
1375 sb->s_blocksize - offset : towrite;
1378 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1381 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1382 bh = sb_bread(sb, tmp_bh.b_blocknr);
1384 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1390 memcpy(bh->b_data+offset, data, tocopy);
1391 flush_dcache_page(bh->b_page);
1392 set_buffer_uptodate(bh);
1393 mark_buffer_dirty(bh);
1402 if (len == towrite) {
1403 mutex_unlock(&inode->i_mutex);
1406 if (inode->i_size < off+len-towrite)
1407 i_size_write(inode, off+len-towrite);
1409 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1410 mark_inode_dirty(inode);
1411 mutex_unlock(&inode->i_mutex);
1412 return len - towrite;
1417 static struct file_system_type ext2_fs_type = {
1418 .owner = THIS_MODULE,
1420 .get_sb = ext2_get_sb,
1421 .kill_sb = kill_block_super,
1422 .fs_flags = FS_REQUIRES_DEV,
1425 static int __init init_ext2_fs(void)
1427 int err = init_ext2_xattr();
1430 err = init_inodecache();
1433 err = register_filesystem(&ext2_fs_type);
1438 destroy_inodecache();
1444 static void __exit exit_ext2_fs(void)
1446 unregister_filesystem(&ext2_fs_type);
1447 destroy_inodecache();
1451 module_init(init_ext2_fs)
1452 module_exit(exit_ext2_fs)