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);
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;
135 kfree(sbi->s_blockgroup_lock);
141 static struct kmem_cache * ext2_inode_cachep;
143 static struct inode *ext2_alloc_inode(struct super_block *sb)
145 struct ext2_inode_info *ei;
146 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
149 #ifdef CONFIG_EXT2_FS_POSIX_ACL
150 ei->i_acl = EXT2_ACL_NOT_CACHED;
151 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
153 ei->i_block_alloc_info = NULL;
154 ei->vfs_inode.i_version = 1;
155 return &ei->vfs_inode;
158 static void ext2_destroy_inode(struct inode *inode)
160 remove_extent_mappings(&EXT2_I(inode)->extent_tree, 0, (u64)-1);
161 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
164 static void init_once(void *foo)
166 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
168 rwlock_init(&ei->i_meta_lock);
169 #ifdef CONFIG_EXT2_FS_XATTR
170 init_rwsem(&ei->xattr_sem);
172 mutex_init(&ei->truncate_mutex);
173 extent_map_tree_init(&ei->extent_tree);
174 inode_init_once(&ei->vfs_inode);
177 static int init_inodecache(void)
179 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
180 sizeof(struct ext2_inode_info),
181 0, (SLAB_RECLAIM_ACCOUNT|
184 if (ext2_inode_cachep == NULL)
189 static void destroy_inodecache(void)
191 kmem_cache_destroy(ext2_inode_cachep);
194 static void ext2_clear_inode(struct inode *inode)
196 struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
197 #ifdef CONFIG_EXT2_FS_POSIX_ACL
198 struct ext2_inode_info *ei = EXT2_I(inode);
200 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
201 posix_acl_release(ei->i_acl);
202 ei->i_acl = EXT2_ACL_NOT_CACHED;
204 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
205 posix_acl_release(ei->i_default_acl);
206 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
209 ext2_discard_reservation(inode);
210 EXT2_I(inode)->i_block_alloc_info = NULL;
215 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
217 struct super_block *sb = vfs->mnt_sb;
218 struct ext2_sb_info *sbi = EXT2_SB(sb);
219 struct ext2_super_block *es = sbi->s_es;
220 unsigned long def_mount_opts;
222 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
224 if (sbi->s_sb_block != 1)
225 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
226 if (test_opt(sb, MINIX_DF))
227 seq_puts(seq, ",minixdf");
228 if (test_opt(sb, GRPID))
229 seq_puts(seq, ",grpid");
230 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
231 seq_puts(seq, ",nogrpid");
232 if (sbi->s_resuid != EXT2_DEF_RESUID ||
233 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
234 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
236 if (sbi->s_resgid != EXT2_DEF_RESGID ||
237 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
238 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
240 if (test_opt(sb, ERRORS_RO)) {
241 int def_errors = le16_to_cpu(es->s_errors);
243 if (def_errors == EXT2_ERRORS_PANIC ||
244 def_errors == EXT2_ERRORS_CONTINUE) {
245 seq_puts(seq, ",errors=remount-ro");
248 if (test_opt(sb, ERRORS_CONT))
249 seq_puts(seq, ",errors=continue");
250 if (test_opt(sb, ERRORS_PANIC))
251 seq_puts(seq, ",errors=panic");
252 if (test_opt(sb, NO_UID32))
253 seq_puts(seq, ",nouid32");
254 if (test_opt(sb, DEBUG))
255 seq_puts(seq, ",debug");
256 if (test_opt(sb, OLDALLOC))
257 seq_puts(seq, ",oldalloc");
259 #ifdef CONFIG_EXT2_FS_XATTR
260 if (test_opt(sb, XATTR_USER))
261 seq_puts(seq, ",user_xattr");
262 if (!test_opt(sb, XATTR_USER) &&
263 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
264 seq_puts(seq, ",nouser_xattr");
268 #ifdef CONFIG_EXT2_FS_POSIX_ACL
269 if (test_opt(sb, POSIX_ACL))
270 seq_puts(seq, ",acl");
271 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
272 seq_puts(seq, ",noacl");
275 if (test_opt(sb, NOBH))
276 seq_puts(seq, ",nobh");
278 #if defined(CONFIG_QUOTA)
279 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
280 seq_puts(seq, ",usrquota");
282 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
283 seq_puts(seq, ",grpquota");
286 #if defined(CONFIG_EXT2_FS_XIP)
287 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
288 seq_puts(seq, ",xip");
291 if (!test_opt(sb, RESERVATION))
292 seq_puts(seq, ",noreservation");
298 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
299 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
302 static const struct super_operations ext2_sops = {
303 .alloc_inode = ext2_alloc_inode,
304 .destroy_inode = ext2_destroy_inode,
305 .write_inode = ext2_write_inode,
306 .delete_inode = ext2_delete_inode,
307 .put_super = ext2_put_super,
308 .write_super = ext2_write_super,
309 .statfs = ext2_statfs,
310 .remount_fs = ext2_remount,
311 .clear_inode = ext2_clear_inode,
312 .show_options = ext2_show_options,
314 .quota_read = ext2_quota_read,
315 .quota_write = ext2_quota_write,
319 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
320 u64 ino, u32 generation)
324 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
325 return ERR_PTR(-ESTALE);
326 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
327 return ERR_PTR(-ESTALE);
329 /* iget isn't really right if the inode is currently unallocated!!
330 * ext2_read_inode currently does appropriate checks, but
331 * it might be "neater" to call ext2_get_inode first and check
332 * if the inode is valid.....
334 inode = ext2_iget(sb, ino);
336 return ERR_CAST(inode);
337 if (generation && inode->i_generation != generation) {
338 /* we didn't find the right inode.. */
340 return ERR_PTR(-ESTALE);
345 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
346 int fh_len, int fh_type)
348 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
352 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
353 int fh_len, int fh_type)
355 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
359 /* Yes, most of these are left as NULL!!
360 * A NULL value implies the default, which works with ext2-like file
361 * systems, but can be improved upon.
362 * Currently only get_parent is required.
364 static const struct export_operations ext2_export_ops = {
365 .fh_to_dentry = ext2_fh_to_dentry,
366 .fh_to_parent = ext2_fh_to_parent,
367 .get_parent = ext2_get_parent,
370 static unsigned long get_sb_block(void **data)
372 unsigned long sb_block;
373 char *options = (char *) *data;
375 if (!options || strncmp(options, "sb=", 3) != 0)
376 return 1; /* Default location */
378 sb_block = simple_strtoul(options, &options, 0);
379 if (*options && *options != ',') {
380 printk("EXT2-fs: Invalid sb specification: %s\n",
386 *data = (void *) options;
391 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
392 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
393 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
394 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
395 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
396 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
399 static const match_table_t tokens = {
400 {Opt_bsd_df, "bsddf"},
401 {Opt_minix_df, "minixdf"},
402 {Opt_grpid, "grpid"},
403 {Opt_grpid, "bsdgroups"},
404 {Opt_nogrpid, "nogrpid"},
405 {Opt_nogrpid, "sysvgroups"},
406 {Opt_resgid, "resgid=%u"},
407 {Opt_resuid, "resuid=%u"},
409 {Opt_err_cont, "errors=continue"},
410 {Opt_err_panic, "errors=panic"},
411 {Opt_err_ro, "errors=remount-ro"},
412 {Opt_nouid32, "nouid32"},
413 {Opt_nocheck, "check=none"},
414 {Opt_nocheck, "nocheck"},
415 {Opt_debug, "debug"},
416 {Opt_oldalloc, "oldalloc"},
417 {Opt_orlov, "orlov"},
419 {Opt_user_xattr, "user_xattr"},
420 {Opt_nouser_xattr, "nouser_xattr"},
422 {Opt_noacl, "noacl"},
424 {Opt_grpquota, "grpquota"},
425 {Opt_ignore, "noquota"},
426 {Opt_quota, "quota"},
427 {Opt_usrquota, "usrquota"},
428 {Opt_reservation, "reservation"},
429 {Opt_noreservation, "noreservation"},
433 static int parse_options (char * options,
434 struct ext2_sb_info *sbi)
437 substring_t args[MAX_OPT_ARGS];
443 while ((p = strsep (&options, ",")) != NULL) {
448 token = match_token(p, tokens, args);
451 clear_opt (sbi->s_mount_opt, MINIX_DF);
454 set_opt (sbi->s_mount_opt, MINIX_DF);
457 set_opt (sbi->s_mount_opt, GRPID);
460 clear_opt (sbi->s_mount_opt, GRPID);
463 if (match_int(&args[0], &option))
465 sbi->s_resuid = option;
468 if (match_int(&args[0], &option))
470 sbi->s_resgid = option;
473 /* handled by get_sb_block() instead of here */
474 /* *sb_block = match_int(&args[0]); */
477 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
478 clear_opt (sbi->s_mount_opt, ERRORS_RO);
479 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
482 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
483 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
484 set_opt (sbi->s_mount_opt, ERRORS_RO);
487 clear_opt (sbi->s_mount_opt, ERRORS_RO);
488 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
489 set_opt (sbi->s_mount_opt, ERRORS_CONT);
492 set_opt (sbi->s_mount_opt, NO_UID32);
495 clear_opt (sbi->s_mount_opt, CHECK);
498 set_opt (sbi->s_mount_opt, DEBUG);
501 set_opt (sbi->s_mount_opt, OLDALLOC);
504 clear_opt (sbi->s_mount_opt, OLDALLOC);
507 set_opt (sbi->s_mount_opt, NOBH);
509 #ifdef CONFIG_EXT2_FS_XATTR
511 set_opt (sbi->s_mount_opt, XATTR_USER);
513 case Opt_nouser_xattr:
514 clear_opt (sbi->s_mount_opt, XATTR_USER);
518 case Opt_nouser_xattr:
519 printk("EXT2 (no)user_xattr options not supported\n");
522 #ifdef CONFIG_EXT2_FS_POSIX_ACL
524 set_opt(sbi->s_mount_opt, POSIX_ACL);
527 clear_opt(sbi->s_mount_opt, POSIX_ACL);
532 printk("EXT2 (no)acl options not supported\n");
536 #ifdef CONFIG_EXT2_FS_XIP
537 set_opt (sbi->s_mount_opt, XIP);
539 printk("EXT2 xip option not supported\n");
543 #if defined(CONFIG_QUOTA)
546 set_opt(sbi->s_mount_opt, USRQUOTA);
550 set_opt(sbi->s_mount_opt, GRPQUOTA);
557 "EXT2-fs: quota operations not supported.\n");
562 case Opt_reservation:
563 set_opt(sbi->s_mount_opt, RESERVATION);
564 printk("reservations ON\n");
566 case Opt_noreservation:
567 clear_opt(sbi->s_mount_opt, RESERVATION);
568 printk("reservations OFF\n");
579 static int ext2_setup_super (struct super_block * sb,
580 struct ext2_super_block * es,
584 struct ext2_sb_info *sbi = EXT2_SB(sb);
586 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
587 printk ("EXT2-fs warning: revision level too high, "
588 "forcing read-only mode\n");
593 if (!(sbi->s_mount_state & EXT2_VALID_FS))
594 printk ("EXT2-fs warning: mounting unchecked fs, "
595 "running e2fsck is recommended\n");
596 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
597 printk ("EXT2-fs warning: mounting fs with errors, "
598 "running e2fsck is recommended\n");
599 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
600 le16_to_cpu(es->s_mnt_count) >=
601 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
602 printk ("EXT2-fs warning: maximal mount count reached, "
603 "running e2fsck is recommended\n");
604 else if (le32_to_cpu(es->s_checkinterval) &&
605 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
606 printk ("EXT2-fs warning: checktime reached, "
607 "running e2fsck is recommended\n");
608 if (!le16_to_cpu(es->s_max_mnt_count))
609 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
610 le16_add_cpu(&es->s_mnt_count, 1);
611 ext2_write_super(sb);
612 if (test_opt (sb, DEBUG))
613 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
614 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
615 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
618 EXT2_BLOCKS_PER_GROUP(sb),
619 EXT2_INODES_PER_GROUP(sb),
624 static int ext2_check_descriptors(struct super_block *sb)
627 struct ext2_sb_info *sbi = EXT2_SB(sb);
629 ext2_debug ("Checking group descriptors");
631 for (i = 0; i < sbi->s_groups_count; i++) {
632 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
633 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
634 ext2_fsblk_t last_block;
636 if (i == sbi->s_groups_count - 1)
637 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
639 last_block = first_block +
640 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
642 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
643 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
645 ext2_error (sb, "ext2_check_descriptors",
646 "Block bitmap for group %d"
647 " not in group (block %lu)!",
648 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
651 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
652 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
654 ext2_error (sb, "ext2_check_descriptors",
655 "Inode bitmap for group %d"
656 " not in group (block %lu)!",
657 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
660 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
661 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
664 ext2_error (sb, "ext2_check_descriptors",
665 "Inode table for group %d"
666 " not in group (block %lu)!",
667 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
675 * Maximal file size. There is a direct, and {,double-,triple-}indirect
676 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
677 * We need to be 1 filesystem block less than the 2^32 sector limit.
679 static loff_t ext2_max_size(int bits)
681 loff_t res = EXT2_NDIR_BLOCKS;
685 /* This is calculated to be the largest file size for a
686 * dense, file such that the total number of
687 * sectors in the file, including data and all indirect blocks,
688 * does not exceed 2^32 -1
689 * __u32 i_blocks representing the total number of
690 * 512 bytes blocks of the file
692 upper_limit = (1LL << 32) - 1;
694 /* total blocks in file system block size */
695 upper_limit >>= (bits - 9);
698 /* indirect blocks */
700 /* double indirect blocks */
701 meta_blocks += 1 + (1LL << (bits-2));
702 /* tripple indirect blocks */
703 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
705 upper_limit -= meta_blocks;
706 upper_limit <<= bits;
708 res += 1LL << (bits-2);
709 res += 1LL << (2*(bits-2));
710 res += 1LL << (3*(bits-2));
712 if (res > upper_limit)
715 if (res > MAX_LFS_FILESIZE)
716 res = MAX_LFS_FILESIZE;
721 static unsigned long descriptor_loc(struct super_block *sb,
722 unsigned long logic_sb_block,
725 struct ext2_sb_info *sbi = EXT2_SB(sb);
726 unsigned long bg, first_meta_bg;
729 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
731 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
733 return (logic_sb_block + nr + 1);
734 bg = sbi->s_desc_per_block * nr;
735 if (ext2_bg_has_super(sb, bg))
738 return ext2_group_first_block_no(sb, bg) + has_super;
741 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
743 struct buffer_head * bh;
744 struct ext2_sb_info * sbi;
745 struct ext2_super_block * es;
748 unsigned long sb_block = get_sb_block(&data);
749 unsigned long logic_sb_block;
750 unsigned long offset = 0;
751 unsigned long def_mount_opts;
753 int blocksize = BLOCK_SIZE;
759 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
763 sbi->s_blockgroup_lock =
764 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
765 if (!sbi->s_blockgroup_lock) {
770 sbi->s_sb_block = sb_block;
773 * See what the current blocksize for the device is, and
774 * use that as the blocksize. Otherwise (or if the blocksize
775 * is smaller than the default) use the default.
776 * This is important for devices that have a hardware
777 * sectorsize that is larger than the default.
779 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
781 printk ("EXT2-fs: unable to set blocksize\n");
786 * If the superblock doesn't start on a hardware sector boundary,
787 * calculate the offset.
789 if (blocksize != BLOCK_SIZE) {
790 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
791 offset = (sb_block*BLOCK_SIZE) % blocksize;
793 logic_sb_block = sb_block;
796 if (!(bh = sb_bread(sb, logic_sb_block))) {
797 printk ("EXT2-fs: unable to read superblock\n");
801 * Note: s_es must be initialized as soon as possible because
802 * some ext2 macro-instructions depend on its value
804 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
806 sb->s_magic = le16_to_cpu(es->s_magic);
808 if (sb->s_magic != EXT2_SUPER_MAGIC)
811 /* Set defaults before we parse the mount options */
812 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
813 if (def_mount_opts & EXT2_DEFM_DEBUG)
814 set_opt(sbi->s_mount_opt, DEBUG);
815 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
816 set_opt(sbi->s_mount_opt, GRPID);
817 if (def_mount_opts & EXT2_DEFM_UID16)
818 set_opt(sbi->s_mount_opt, NO_UID32);
819 #ifdef CONFIG_EXT2_FS_XATTR
820 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
821 set_opt(sbi->s_mount_opt, XATTR_USER);
823 #ifdef CONFIG_EXT2_FS_POSIX_ACL
824 if (def_mount_opts & EXT2_DEFM_ACL)
825 set_opt(sbi->s_mount_opt, POSIX_ACL);
828 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
829 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
830 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
831 set_opt(sbi->s_mount_opt, ERRORS_CONT);
833 set_opt(sbi->s_mount_opt, ERRORS_RO);
835 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
836 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
838 set_opt(sbi->s_mount_opt, RESERVATION);
840 if (!parse_options ((char *) data, sbi))
843 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
844 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
847 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
848 EXT2_MOUNT_XIP if not */
850 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
851 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
852 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
853 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
854 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
855 "running e2fsck is recommended\n");
857 * Check feature flags regardless of the revision level, since we
858 * previously didn't change the revision level when setting the flags,
859 * so there is a chance incompat flags are set on a rev 0 filesystem.
861 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
863 printk("EXT2-fs: %s: couldn't mount because of "
864 "unsupported optional features (%x).\n",
865 sb->s_id, le32_to_cpu(features));
868 if (!(sb->s_flags & MS_RDONLY) &&
869 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
870 printk("EXT2-fs: %s: couldn't mount RDWR because of "
871 "unsupported optional features (%x).\n",
872 sb->s_id, le32_to_cpu(features));
876 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
878 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
880 printk("XIP: Unsupported blocksize\n");
884 /* If the blocksize doesn't match, re-read the thing.. */
885 if (sb->s_blocksize != blocksize) {
888 if (!sb_set_blocksize(sb, blocksize)) {
889 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
893 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
894 offset = (sb_block*BLOCK_SIZE) % blocksize;
895 bh = sb_bread(sb, logic_sb_block);
897 printk("EXT2-fs: Couldn't read superblock on "
901 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
903 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
904 printk ("EXT2-fs: Magic mismatch, very weird !\n");
909 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
911 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
912 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
913 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
915 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
916 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
917 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
918 !is_power_of_2(sbi->s_inode_size) ||
919 (sbi->s_inode_size > blocksize)) {
920 printk ("EXT2-fs: unsupported inode size: %d\n",
926 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
927 le32_to_cpu(es->s_log_frag_size);
928 if (sbi->s_frag_size == 0)
930 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
932 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
933 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
934 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
936 if (EXT2_INODE_SIZE(sb) == 0)
938 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
939 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
941 sbi->s_itb_per_group = sbi->s_inodes_per_group /
942 sbi->s_inodes_per_block;
943 sbi->s_desc_per_block = sb->s_blocksize /
944 sizeof (struct ext2_group_desc);
946 sbi->s_mount_state = le16_to_cpu(es->s_state);
947 sbi->s_addr_per_block_bits =
948 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
949 sbi->s_desc_per_block_bits =
950 ilog2 (EXT2_DESC_PER_BLOCK(sb));
952 if (sb->s_magic != EXT2_SUPER_MAGIC)
955 if (sb->s_blocksize != bh->b_size) {
957 printk ("VFS: Unsupported blocksize on dev "
962 if (sb->s_blocksize != sbi->s_frag_size) {
963 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
964 sbi->s_frag_size, sb->s_blocksize);
968 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
969 printk ("EXT2-fs: #blocks per group too big: %lu\n",
970 sbi->s_blocks_per_group);
973 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
974 printk ("EXT2-fs: #fragments per group too big: %lu\n",
975 sbi->s_frags_per_group);
978 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
979 printk ("EXT2-fs: #inodes per group too big: %lu\n",
980 sbi->s_inodes_per_group);
984 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
986 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
987 le32_to_cpu(es->s_first_data_block) - 1)
988 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
989 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
990 EXT2_DESC_PER_BLOCK(sb);
991 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
992 if (sbi->s_group_desc == NULL) {
993 printk ("EXT2-fs: not enough memory\n");
996 bgl_lock_init(sbi->s_blockgroup_lock);
997 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
999 printk ("EXT2-fs: not enough memory\n");
1000 goto failed_mount_group_desc;
1002 for (i = 0; i < db_count; i++) {
1003 block = descriptor_loc(sb, logic_sb_block, i);
1004 sbi->s_group_desc[i] = sb_bread(sb, block);
1005 if (!sbi->s_group_desc[i]) {
1006 for (j = 0; j < i; j++)
1007 brelse (sbi->s_group_desc[j]);
1008 printk ("EXT2-fs: unable to read group descriptors\n");
1009 goto failed_mount_group_desc;
1012 if (!ext2_check_descriptors (sb)) {
1013 printk ("EXT2-fs: group descriptors corrupted!\n");
1016 sbi->s_gdb_count = db_count;
1017 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1018 spin_lock_init(&sbi->s_next_gen_lock);
1020 /* per fileystem reservation list head & lock */
1021 spin_lock_init(&sbi->s_rsv_window_lock);
1022 sbi->s_rsv_window_root = RB_ROOT;
1024 * Add a single, static dummy reservation to the start of the
1025 * reservation window list --- it gives us a placeholder for
1026 * append-at-start-of-list which makes the allocation logic
1029 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1030 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1031 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1032 sbi->s_rsv_window_head.rsv_goal_size = 0;
1033 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1035 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1036 ext2_count_free_blocks(sb));
1038 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1039 ext2_count_free_inodes(sb));
1042 err = percpu_counter_init(&sbi->s_dirs_counter,
1043 ext2_count_dirs(sb));
1046 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1050 * set up enough so that it can read an inode
1052 sb->s_op = &ext2_sops;
1053 sb->s_export_op = &ext2_export_ops;
1054 sb->s_xattr = ext2_xattr_handlers;
1055 root = ext2_iget(sb, EXT2_ROOT_INO);
1057 ret = PTR_ERR(root);
1060 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1062 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1066 sb->s_root = d_alloc_root(root);
1069 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1073 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1074 ext2_warning(sb, __func__,
1075 "mounting ext3 filesystem as ext2");
1076 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1081 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1085 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1086 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1087 percpu_counter_destroy(&sbi->s_dirs_counter);
1089 for (i = 0; i < db_count; i++)
1090 brelse(sbi->s_group_desc[i]);
1091 failed_mount_group_desc:
1092 kfree(sbi->s_group_desc);
1093 kfree(sbi->s_debts);
1097 sb->s_fs_info = NULL;
1102 static void ext2_commit_super (struct super_block * sb,
1103 struct ext2_super_block * es)
1105 es->s_wtime = cpu_to_le32(get_seconds());
1106 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1110 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1112 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1113 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1114 es->s_wtime = cpu_to_le32(get_seconds());
1115 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1116 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1121 * In the second extended file system, it is not necessary to
1122 * write the super block since we use a mapping of the
1123 * disk super block in a buffer.
1125 * However, this function is still used to set the fs valid
1126 * flags to 0. We need to set this flag to 0 since the fs
1127 * may have been checked while mounted and e2fsck may have
1128 * set s_state to EXT2_VALID_FS after some corrections.
1131 void ext2_write_super (struct super_block * sb)
1133 struct ext2_super_block * es;
1135 if (!(sb->s_flags & MS_RDONLY)) {
1136 es = EXT2_SB(sb)->s_es;
1138 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1139 ext2_debug ("setting valid to 0\n");
1140 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1141 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1142 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1143 es->s_mtime = cpu_to_le32(get_seconds());
1144 ext2_sync_super(sb, es);
1146 ext2_commit_super (sb, es);
1152 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1154 struct ext2_sb_info * sbi = EXT2_SB(sb);
1155 struct ext2_super_block * es;
1156 unsigned long old_mount_opt = sbi->s_mount_opt;
1157 struct ext2_mount_options old_opts;
1158 unsigned long old_sb_flags;
1161 /* Store the old options */
1162 old_sb_flags = sb->s_flags;
1163 old_opts.s_mount_opt = sbi->s_mount_opt;
1164 old_opts.s_resuid = sbi->s_resuid;
1165 old_opts.s_resgid = sbi->s_resgid;
1168 * Allow the "check" option to be passed as a remount option.
1170 if (!parse_options (data, sbi)) {
1175 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1176 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1178 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1179 EXT2_MOUNT_XIP if not */
1181 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1182 printk("XIP: Unsupported blocksize\n");
1188 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1189 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1190 invalidate_inodes(sb)) {
1191 ext2_warning(sb, __func__, "refusing change of xip flag "
1192 "with busy inodes while remounting");
1193 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1194 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1196 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1198 if (*flags & MS_RDONLY) {
1199 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1200 !(sbi->s_mount_state & EXT2_VALID_FS))
1203 * OK, we are remounting a valid rw partition rdonly, so set
1204 * the rdonly flag and then mark the partition as valid again.
1206 es->s_state = cpu_to_le16(sbi->s_mount_state);
1207 es->s_mtime = cpu_to_le32(get_seconds());
1209 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1210 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1212 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1213 "unsupported optional features (%x).\n",
1214 sb->s_id, le32_to_cpu(ret));
1219 * Mounting a RDONLY partition read-write, so reread and
1220 * store the current valid flag. (It may have been changed
1221 * by e2fsck since we originally mounted the partition.)
1223 sbi->s_mount_state = le16_to_cpu(es->s_state);
1224 if (!ext2_setup_super (sb, es, 0))
1225 sb->s_flags &= ~MS_RDONLY;
1227 ext2_sync_super(sb, es);
1230 sbi->s_mount_opt = old_opts.s_mount_opt;
1231 sbi->s_resuid = old_opts.s_resuid;
1232 sbi->s_resgid = old_opts.s_resgid;
1233 sb->s_flags = old_sb_flags;
1237 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1239 struct super_block *sb = dentry->d_sb;
1240 struct ext2_sb_info *sbi = EXT2_SB(sb);
1241 struct ext2_super_block *es = sbi->s_es;
1244 if (test_opt (sb, MINIX_DF))
1245 sbi->s_overhead_last = 0;
1246 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1247 unsigned long i, overhead = 0;
1251 * Compute the overhead (FS structures). This is constant
1252 * for a given filesystem unless the number of block groups
1253 * changes so we cache the previous value until it does.
1257 * All of the blocks before first_data_block are
1260 overhead = le32_to_cpu(es->s_first_data_block);
1263 * Add the overhead attributed to the superblock and
1264 * block group descriptors. If the sparse superblocks
1265 * feature is turned on, then not all groups have this.
1267 for (i = 0; i < sbi->s_groups_count; i++)
1268 overhead += ext2_bg_has_super(sb, i) +
1269 ext2_bg_num_gdb(sb, i);
1272 * Every block group has an inode bitmap, a block
1273 * bitmap, and an inode table.
1275 overhead += (sbi->s_groups_count *
1276 (2 + sbi->s_itb_per_group));
1277 sbi->s_overhead_last = overhead;
1279 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1282 buf->f_type = EXT2_SUPER_MAGIC;
1283 buf->f_bsize = sb->s_blocksize;
1284 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1285 buf->f_bfree = ext2_count_free_blocks(sb);
1286 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1287 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1288 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1290 buf->f_files = le32_to_cpu(es->s_inodes_count);
1291 buf->f_ffree = ext2_count_free_inodes(sb);
1292 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1293 buf->f_namelen = EXT2_NAME_LEN;
1294 fsid = le64_to_cpup((void *)es->s_uuid) ^
1295 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1296 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1297 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1301 static int ext2_get_sb(struct file_system_type *fs_type,
1302 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1304 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1309 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1310 * acquiring the locks... As quota files are never truncated and quota code
1311 * itself serializes the operations (and noone else should touch the files)
1312 * we don't have to be afraid of races */
1313 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1314 size_t len, loff_t off)
1316 struct inode *inode = sb_dqopt(sb)->files[type];
1317 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1319 int offset = off & (sb->s_blocksize - 1);
1322 struct buffer_head tmp_bh;
1323 struct buffer_head *bh;
1324 loff_t i_size = i_size_read(inode);
1328 if (off+len > i_size)
1331 while (toread > 0) {
1332 tocopy = sb->s_blocksize - offset < toread ?
1333 sb->s_blocksize - offset : toread;
1336 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1339 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1340 memset(data, 0, tocopy);
1342 bh = sb_bread(sb, tmp_bh.b_blocknr);
1345 memcpy(data, bh->b_data+offset, tocopy);
1356 /* Write to quotafile */
1357 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1358 const char *data, size_t len, loff_t off)
1360 struct inode *inode = sb_dqopt(sb)->files[type];
1361 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1363 int offset = off & (sb->s_blocksize - 1);
1365 size_t towrite = len;
1366 struct buffer_head tmp_bh;
1367 struct buffer_head *bh;
1369 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1370 while (towrite > 0) {
1371 tocopy = sb->s_blocksize - offset < towrite ?
1372 sb->s_blocksize - offset : towrite;
1375 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1378 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1379 bh = sb_bread(sb, tmp_bh.b_blocknr);
1381 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1387 memcpy(bh->b_data+offset, data, tocopy);
1388 flush_dcache_page(bh->b_page);
1389 set_buffer_uptodate(bh);
1390 mark_buffer_dirty(bh);
1401 if (inode->i_size < off+len-towrite)
1402 i_size_write(inode, off+len-towrite);
1404 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1405 mark_inode_dirty(inode);
1406 mutex_unlock(&inode->i_mutex);
1407 return len - towrite;
1412 static struct file_system_type ext2_fs_type = {
1413 .owner = THIS_MODULE,
1415 .get_sb = ext2_get_sb,
1416 .kill_sb = kill_block_super,
1417 .fs_flags = FS_REQUIRES_DEV,
1420 static int __init init_ext2_fs(void)
1422 int err = init_ext2_xattr();
1425 err = init_inodecache();
1428 err = register_filesystem(&ext2_fs_type);
1433 destroy_inodecache();
1439 static void __exit exit_ext2_fs(void)
1441 unregister_filesystem(&ext2_fs_type);
1442 destroy_inodecache();
1446 module_init(init_ext2_fs)
1447 module_exit(exit_ext2_fs)