2 * linux/fs/ext2/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/blkdev.h>
25 #include <linux/parser.h>
26 #include <linux/random.h>
27 #include <linux/buffer_head.h>
28 #include <linux/exportfs.h>
29 #include <linux/smp_lock.h>
30 #include <linux/vfs.h>
31 #include <linux/seq_file.h>
32 #include <linux/mount.h>
33 #include <linux/log2.h>
34 #include <asm/uaccess.h>
40 static void ext2_sync_super(struct super_block *sb,
41 struct ext2_super_block *es);
42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
45 void ext2_error (struct super_block * sb, const char * function,
46 const char * fmt, ...)
49 struct ext2_sb_info *sbi = EXT2_SB(sb);
50 struct ext2_super_block *es = sbi->s_es;
52 if (!(sb->s_flags & MS_RDONLY)) {
53 sbi->s_mount_state |= EXT2_ERROR_FS;
55 cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
56 ext2_sync_super(sb, es);
60 printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
65 if (test_opt(sb, ERRORS_PANIC))
66 panic("EXT2-fs panic from previous error\n");
67 if (test_opt(sb, ERRORS_RO)) {
68 printk("Remounting filesystem read-only\n");
69 sb->s_flags |= MS_RDONLY;
73 void ext2_warning (struct super_block * sb, const char * function,
74 const char * fmt, ...)
79 printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
86 void ext2_update_dynamic_rev(struct super_block *sb)
88 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
90 if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
93 ext2_warning(sb, __FUNCTION__,
94 "updating to rev %d because of new feature flag, "
95 "running e2fsck is recommended",
98 es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
99 es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
100 es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
101 /* leave es->s_feature_*compat flags alone */
102 /* es->s_uuid will be set by e2fsck if empty */
105 * The rest of the superblock fields should be zero, and if not it
106 * means they are likely already in use, so leave them alone. We
107 * can leave it up to e2fsck to clean up any inconsistencies there.
111 static void ext2_put_super (struct super_block * sb)
115 struct ext2_sb_info *sbi = EXT2_SB(sb);
117 ext2_xattr_put_super(sb);
118 if (!(sb->s_flags & MS_RDONLY)) {
119 struct ext2_super_block *es = sbi->s_es;
121 es->s_state = cpu_to_le16(sbi->s_mount_state);
122 ext2_sync_super(sb, es);
124 db_count = sbi->s_gdb_count;
125 for (i = 0; i < db_count; i++)
126 if (sbi->s_group_desc[i])
127 brelse (sbi->s_group_desc[i]);
128 kfree(sbi->s_group_desc);
130 percpu_counter_destroy(&sbi->s_freeblocks_counter);
131 percpu_counter_destroy(&sbi->s_freeinodes_counter);
132 percpu_counter_destroy(&sbi->s_dirs_counter);
134 sb->s_fs_info = NULL;
140 static struct kmem_cache * ext2_inode_cachep;
142 static struct inode *ext2_alloc_inode(struct super_block *sb)
144 struct ext2_inode_info *ei;
145 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
148 #ifdef CONFIG_EXT2_FS_POSIX_ACL
149 ei->i_acl = EXT2_ACL_NOT_CACHED;
150 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
152 ei->i_block_alloc_info = NULL;
153 ei->vfs_inode.i_version = 1;
154 return &ei->vfs_inode;
157 static void ext2_destroy_inode(struct inode *inode)
159 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
162 static void init_once(struct kmem_cache * cachep, void *foo)
164 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
166 rwlock_init(&ei->i_meta_lock);
167 #ifdef CONFIG_EXT2_FS_XATTR
168 init_rwsem(&ei->xattr_sem);
170 mutex_init(&ei->truncate_mutex);
171 inode_init_once(&ei->vfs_inode);
174 static int init_inodecache(void)
176 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
177 sizeof(struct ext2_inode_info),
178 0, (SLAB_RECLAIM_ACCOUNT|
181 if (ext2_inode_cachep == NULL)
186 static void destroy_inodecache(void)
188 kmem_cache_destroy(ext2_inode_cachep);
191 static void ext2_clear_inode(struct inode *inode)
193 struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
194 #ifdef CONFIG_EXT2_FS_POSIX_ACL
195 struct ext2_inode_info *ei = EXT2_I(inode);
197 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
198 posix_acl_release(ei->i_acl);
199 ei->i_acl = EXT2_ACL_NOT_CACHED;
201 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
202 posix_acl_release(ei->i_default_acl);
203 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
206 ext2_discard_reservation(inode);
207 EXT2_I(inode)->i_block_alloc_info = NULL;
212 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
214 struct super_block *sb = vfs->mnt_sb;
215 struct ext2_sb_info *sbi = EXT2_SB(sb);
216 struct ext2_super_block *es = sbi->s_es;
217 unsigned long def_mount_opts;
219 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
221 if (sbi->s_sb_block != 1)
222 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
223 if (test_opt(sb, MINIX_DF))
224 seq_puts(seq, ",minixdf");
225 if (test_opt(sb, GRPID))
226 seq_puts(seq, ",grpid");
227 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
228 seq_puts(seq, ",nogrpid");
229 if (sbi->s_resuid != EXT2_DEF_RESUID ||
230 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
231 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
233 if (sbi->s_resgid != EXT2_DEF_RESGID ||
234 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
235 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
237 if (test_opt(sb, ERRORS_RO)) {
238 int def_errors = le16_to_cpu(es->s_errors);
240 if (def_errors == EXT2_ERRORS_PANIC ||
241 def_errors == EXT2_ERRORS_CONTINUE) {
242 seq_puts(seq, ",errors=remount-ro");
245 if (test_opt(sb, ERRORS_CONT))
246 seq_puts(seq, ",errors=continue");
247 if (test_opt(sb, ERRORS_PANIC))
248 seq_puts(seq, ",errors=panic");
249 if (test_opt(sb, NO_UID32))
250 seq_puts(seq, ",nouid32");
251 if (test_opt(sb, DEBUG))
252 seq_puts(seq, ",debug");
253 if (test_opt(sb, OLDALLOC))
254 seq_puts(seq, ",oldalloc");
256 #ifdef CONFIG_EXT2_FS_XATTR
257 if (test_opt(sb, XATTR_USER))
258 seq_puts(seq, ",user_xattr");
259 if (!test_opt(sb, XATTR_USER) &&
260 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
261 seq_puts(seq, ",nouser_xattr");
265 #ifdef CONFIG_EXT2_FS_POSIX_ACL
266 if (test_opt(sb, POSIX_ACL))
267 seq_puts(seq, ",acl");
268 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
269 seq_puts(seq, ",noacl");
272 if (test_opt(sb, NOBH))
273 seq_puts(seq, ",nobh");
275 #if defined(CONFIG_QUOTA)
276 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
277 seq_puts(seq, ",usrquota");
279 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
280 seq_puts(seq, ",grpquota");
283 #if defined(CONFIG_EXT2_FS_XIP)
284 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
285 seq_puts(seq, ",xip");
288 if (!test_opt(sb, RESERVATION))
289 seq_puts(seq, ",noreservation");
295 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
296 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
299 static const struct super_operations ext2_sops = {
300 .alloc_inode = ext2_alloc_inode,
301 .destroy_inode = ext2_destroy_inode,
302 .write_inode = ext2_write_inode,
303 .delete_inode = ext2_delete_inode,
304 .put_super = ext2_put_super,
305 .write_super = ext2_write_super,
306 .statfs = ext2_statfs,
307 .remount_fs = ext2_remount,
308 .clear_inode = ext2_clear_inode,
309 .show_options = ext2_show_options,
311 .quota_read = ext2_quota_read,
312 .quota_write = ext2_quota_write,
316 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
317 u64 ino, u32 generation)
321 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
322 return ERR_PTR(-ESTALE);
323 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
324 return ERR_PTR(-ESTALE);
326 /* iget isn't really right if the inode is currently unallocated!!
327 * ext2_read_inode currently does appropriate checks, but
328 * it might be "neater" to call ext2_get_inode first and check
329 * if the inode is valid.....
331 inode = ext2_iget(sb, ino);
333 return ERR_CAST(inode);
334 if (generation && inode->i_generation != generation) {
335 /* we didn't find the right inode.. */
337 return ERR_PTR(-ESTALE);
342 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
343 int fh_len, int fh_type)
345 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
349 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
350 int fh_len, int fh_type)
352 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
356 /* Yes, most of these are left as NULL!!
357 * A NULL value implies the default, which works with ext2-like file
358 * systems, but can be improved upon.
359 * Currently only get_parent is required.
361 static const struct export_operations ext2_export_ops = {
362 .fh_to_dentry = ext2_fh_to_dentry,
363 .fh_to_parent = ext2_fh_to_parent,
364 .get_parent = ext2_get_parent,
367 static unsigned long get_sb_block(void **data)
369 unsigned long sb_block;
370 char *options = (char *) *data;
372 if (!options || strncmp(options, "sb=", 3) != 0)
373 return 1; /* Default location */
375 sb_block = simple_strtoul(options, &options, 0);
376 if (*options && *options != ',') {
377 printk("EXT2-fs: Invalid sb specification: %s\n",
383 *data = (void *) options;
388 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
389 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
390 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
391 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
392 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
393 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
396 static match_table_t tokens = {
397 {Opt_bsd_df, "bsddf"},
398 {Opt_minix_df, "minixdf"},
399 {Opt_grpid, "grpid"},
400 {Opt_grpid, "bsdgroups"},
401 {Opt_nogrpid, "nogrpid"},
402 {Opt_nogrpid, "sysvgroups"},
403 {Opt_resgid, "resgid=%u"},
404 {Opt_resuid, "resuid=%u"},
406 {Opt_err_cont, "errors=continue"},
407 {Opt_err_panic, "errors=panic"},
408 {Opt_err_ro, "errors=remount-ro"},
409 {Opt_nouid32, "nouid32"},
410 {Opt_nocheck, "check=none"},
411 {Opt_nocheck, "nocheck"},
412 {Opt_debug, "debug"},
413 {Opt_oldalloc, "oldalloc"},
414 {Opt_orlov, "orlov"},
416 {Opt_user_xattr, "user_xattr"},
417 {Opt_nouser_xattr, "nouser_xattr"},
419 {Opt_noacl, "noacl"},
421 {Opt_grpquota, "grpquota"},
422 {Opt_ignore, "noquota"},
423 {Opt_quota, "quota"},
424 {Opt_usrquota, "usrquota"},
425 {Opt_reservation, "reservation"},
426 {Opt_noreservation, "noreservation"},
430 static int parse_options (char * options,
431 struct ext2_sb_info *sbi)
434 substring_t args[MAX_OPT_ARGS];
440 while ((p = strsep (&options, ",")) != NULL) {
445 token = match_token(p, tokens, args);
448 clear_opt (sbi->s_mount_opt, MINIX_DF);
451 set_opt (sbi->s_mount_opt, MINIX_DF);
454 set_opt (sbi->s_mount_opt, GRPID);
457 clear_opt (sbi->s_mount_opt, GRPID);
460 if (match_int(&args[0], &option))
462 sbi->s_resuid = option;
465 if (match_int(&args[0], &option))
467 sbi->s_resgid = option;
470 /* handled by get_sb_block() instead of here */
471 /* *sb_block = match_int(&args[0]); */
474 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
475 clear_opt (sbi->s_mount_opt, ERRORS_RO);
476 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
479 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
480 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
481 set_opt (sbi->s_mount_opt, ERRORS_RO);
484 clear_opt (sbi->s_mount_opt, ERRORS_RO);
485 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
486 set_opt (sbi->s_mount_opt, ERRORS_CONT);
489 set_opt (sbi->s_mount_opt, NO_UID32);
492 clear_opt (sbi->s_mount_opt, CHECK);
495 set_opt (sbi->s_mount_opt, DEBUG);
498 set_opt (sbi->s_mount_opt, OLDALLOC);
501 clear_opt (sbi->s_mount_opt, OLDALLOC);
504 set_opt (sbi->s_mount_opt, NOBH);
506 #ifdef CONFIG_EXT2_FS_XATTR
508 set_opt (sbi->s_mount_opt, XATTR_USER);
510 case Opt_nouser_xattr:
511 clear_opt (sbi->s_mount_opt, XATTR_USER);
515 case Opt_nouser_xattr:
516 printk("EXT2 (no)user_xattr options not supported\n");
519 #ifdef CONFIG_EXT2_FS_POSIX_ACL
521 set_opt(sbi->s_mount_opt, POSIX_ACL);
524 clear_opt(sbi->s_mount_opt, POSIX_ACL);
529 printk("EXT2 (no)acl options not supported\n");
533 #ifdef CONFIG_EXT2_FS_XIP
534 set_opt (sbi->s_mount_opt, XIP);
536 printk("EXT2 xip option not supported\n");
540 #if defined(CONFIG_QUOTA)
543 set_opt(sbi->s_mount_opt, USRQUOTA);
547 set_opt(sbi->s_mount_opt, GRPQUOTA);
554 "EXT2-fs: quota operations not supported.\n");
559 case Opt_reservation:
560 set_opt(sbi->s_mount_opt, RESERVATION);
561 printk("reservations ON\n");
563 case Opt_noreservation:
564 clear_opt(sbi->s_mount_opt, RESERVATION);
565 printk("reservations OFF\n");
576 static int ext2_setup_super (struct super_block * sb,
577 struct ext2_super_block * es,
581 struct ext2_sb_info *sbi = EXT2_SB(sb);
583 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
584 printk ("EXT2-fs warning: revision level too high, "
585 "forcing read-only mode\n");
590 if (!(sbi->s_mount_state & EXT2_VALID_FS))
591 printk ("EXT2-fs warning: mounting unchecked fs, "
592 "running e2fsck is recommended\n");
593 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
594 printk ("EXT2-fs warning: mounting fs with errors, "
595 "running e2fsck is recommended\n");
596 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
597 le16_to_cpu(es->s_mnt_count) >=
598 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
599 printk ("EXT2-fs warning: maximal mount count reached, "
600 "running e2fsck is recommended\n");
601 else if (le32_to_cpu(es->s_checkinterval) &&
602 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
603 printk ("EXT2-fs warning: checktime reached, "
604 "running e2fsck is recommended\n");
605 if (!le16_to_cpu(es->s_max_mnt_count))
606 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
607 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
608 ext2_write_super(sb);
609 if (test_opt (sb, DEBUG))
610 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
611 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
612 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
615 EXT2_BLOCKS_PER_GROUP(sb),
616 EXT2_INODES_PER_GROUP(sb),
621 static int ext2_check_descriptors(struct super_block *sb)
624 struct ext2_sb_info *sbi = EXT2_SB(sb);
625 unsigned long first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
626 unsigned long last_block;
628 ext2_debug ("Checking group descriptors");
630 for (i = 0; i < sbi->s_groups_count; i++) {
631 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
633 if (i == sbi->s_groups_count - 1)
634 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
636 last_block = first_block +
637 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
639 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
640 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
642 ext2_error (sb, "ext2_check_descriptors",
643 "Block bitmap for group %d"
644 " not in group (block %lu)!",
645 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
648 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
649 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
651 ext2_error (sb, "ext2_check_descriptors",
652 "Inode bitmap for group %d"
653 " not in group (block %lu)!",
654 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
657 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
658 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
661 ext2_error (sb, "ext2_check_descriptors",
662 "Inode table for group %d"
663 " not in group (block %lu)!",
664 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
667 first_block += EXT2_BLOCKS_PER_GROUP(sb);
673 * Maximal file size. There is a direct, and {,double-,triple-}indirect
674 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
675 * We need to be 1 filesystem block less than the 2^32 sector limit.
677 static loff_t ext2_max_size(int bits)
679 loff_t res = EXT2_NDIR_BLOCKS;
683 /* This is calculated to be the largest file size for a
684 * dense, file such that the total number of
685 * sectors in the file, including data and all indirect blocks,
686 * does not exceed 2^32 -1
687 * __u32 i_blocks representing the total number of
688 * 512 bytes blocks of the file
690 upper_limit = (1LL << 32) - 1;
692 /* total blocks in file system block size */
693 upper_limit >>= (bits - 9);
696 /* indirect blocks */
698 /* double indirect blocks */
699 meta_blocks += 1 + (1LL << (bits-2));
700 /* tripple indirect blocks */
701 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
703 upper_limit -= meta_blocks;
704 upper_limit <<= bits;
706 res += 1LL << (bits-2);
707 res += 1LL << (2*(bits-2));
708 res += 1LL << (3*(bits-2));
710 if (res > upper_limit)
713 if (res > MAX_LFS_FILESIZE)
714 res = MAX_LFS_FILESIZE;
719 static unsigned long descriptor_loc(struct super_block *sb,
720 unsigned long logic_sb_block,
723 struct ext2_sb_info *sbi = EXT2_SB(sb);
724 unsigned long bg, first_data_block, first_meta_bg;
727 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
728 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
730 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
732 return (logic_sb_block + nr + 1);
733 bg = sbi->s_desc_per_block * nr;
734 if (ext2_bg_has_super(sb, bg))
736 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
739 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
741 struct buffer_head * bh;
742 struct ext2_sb_info * sbi;
743 struct ext2_super_block * es;
746 unsigned long sb_block = get_sb_block(&data);
747 unsigned long logic_sb_block;
748 unsigned long offset = 0;
749 unsigned long def_mount_opts;
751 int blocksize = BLOCK_SIZE;
757 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
761 sbi->s_sb_block = sb_block;
764 * See what the current blocksize for the device is, and
765 * use that as the blocksize. Otherwise (or if the blocksize
766 * is smaller than the default) use the default.
767 * This is important for devices that have a hardware
768 * sectorsize that is larger than the default.
770 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
772 printk ("EXT2-fs: unable to set blocksize\n");
777 * If the superblock doesn't start on a hardware sector boundary,
778 * calculate the offset.
780 if (blocksize != BLOCK_SIZE) {
781 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
782 offset = (sb_block*BLOCK_SIZE) % blocksize;
784 logic_sb_block = sb_block;
787 if (!(bh = sb_bread(sb, logic_sb_block))) {
788 printk ("EXT2-fs: unable to read superblock\n");
792 * Note: s_es must be initialized as soon as possible because
793 * some ext2 macro-instructions depend on its value
795 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
797 sb->s_magic = le16_to_cpu(es->s_magic);
799 if (sb->s_magic != EXT2_SUPER_MAGIC)
802 /* Set defaults before we parse the mount options */
803 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
804 if (def_mount_opts & EXT2_DEFM_DEBUG)
805 set_opt(sbi->s_mount_opt, DEBUG);
806 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
807 set_opt(sbi->s_mount_opt, GRPID);
808 if (def_mount_opts & EXT2_DEFM_UID16)
809 set_opt(sbi->s_mount_opt, NO_UID32);
810 #ifdef CONFIG_EXT2_FS_XATTR
811 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
812 set_opt(sbi->s_mount_opt, XATTR_USER);
814 #ifdef CONFIG_EXT2_FS_POSIX_ACL
815 if (def_mount_opts & EXT2_DEFM_ACL)
816 set_opt(sbi->s_mount_opt, POSIX_ACL);
819 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
820 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
821 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
822 set_opt(sbi->s_mount_opt, ERRORS_CONT);
824 set_opt(sbi->s_mount_opt, ERRORS_RO);
826 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
827 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
829 set_opt(sbi->s_mount_opt, RESERVATION);
831 if (!parse_options ((char *) data, sbi))
834 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
835 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
838 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
839 EXT2_MOUNT_XIP if not */
841 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
842 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
843 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
844 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
845 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
846 "running e2fsck is recommended\n");
848 * Check feature flags regardless of the revision level, since we
849 * previously didn't change the revision level when setting the flags,
850 * so there is a chance incompat flags are set on a rev 0 filesystem.
852 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
854 printk("EXT2-fs: %s: couldn't mount because of "
855 "unsupported optional features (%x).\n",
856 sb->s_id, le32_to_cpu(features));
859 if (!(sb->s_flags & MS_RDONLY) &&
860 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
861 printk("EXT2-fs: %s: couldn't mount RDWR because of "
862 "unsupported optional features (%x).\n",
863 sb->s_id, le32_to_cpu(features));
867 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
869 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
871 printk("XIP: Unsupported blocksize\n");
875 /* If the blocksize doesn't match, re-read the thing.. */
876 if (sb->s_blocksize != blocksize) {
879 if (!sb_set_blocksize(sb, blocksize)) {
880 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
884 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
885 offset = (sb_block*BLOCK_SIZE) % blocksize;
886 bh = sb_bread(sb, logic_sb_block);
888 printk("EXT2-fs: Couldn't read superblock on "
892 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
894 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
895 printk ("EXT2-fs: Magic mismatch, very weird !\n");
900 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
902 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
903 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
904 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
906 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
907 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
908 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
909 !is_power_of_2(sbi->s_inode_size) ||
910 (sbi->s_inode_size > blocksize)) {
911 printk ("EXT2-fs: unsupported inode size: %d\n",
917 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
918 le32_to_cpu(es->s_log_frag_size);
919 if (sbi->s_frag_size == 0)
921 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
923 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
924 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
925 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
927 if (EXT2_INODE_SIZE(sb) == 0)
929 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
930 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
932 sbi->s_itb_per_group = sbi->s_inodes_per_group /
933 sbi->s_inodes_per_block;
934 sbi->s_desc_per_block = sb->s_blocksize /
935 sizeof (struct ext2_group_desc);
937 sbi->s_mount_state = le16_to_cpu(es->s_state);
938 sbi->s_addr_per_block_bits =
939 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
940 sbi->s_desc_per_block_bits =
941 ilog2 (EXT2_DESC_PER_BLOCK(sb));
943 if (sb->s_magic != EXT2_SUPER_MAGIC)
946 if (sb->s_blocksize != bh->b_size) {
948 printk ("VFS: Unsupported blocksize on dev "
953 if (sb->s_blocksize != sbi->s_frag_size) {
954 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
955 sbi->s_frag_size, sb->s_blocksize);
959 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
960 printk ("EXT2-fs: #blocks per group too big: %lu\n",
961 sbi->s_blocks_per_group);
964 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
965 printk ("EXT2-fs: #fragments per group too big: %lu\n",
966 sbi->s_frags_per_group);
969 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
970 printk ("EXT2-fs: #inodes per group too big: %lu\n",
971 sbi->s_inodes_per_group);
975 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
977 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
978 le32_to_cpu(es->s_first_data_block) - 1)
979 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
980 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
981 EXT2_DESC_PER_BLOCK(sb);
982 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
983 if (sbi->s_group_desc == NULL) {
984 printk ("EXT2-fs: not enough memory\n");
987 bgl_lock_init(&sbi->s_blockgroup_lock);
988 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
990 printk ("EXT2-fs: not enough memory\n");
991 goto failed_mount_group_desc;
993 for (i = 0; i < db_count; i++) {
994 block = descriptor_loc(sb, logic_sb_block, i);
995 sbi->s_group_desc[i] = sb_bread(sb, block);
996 if (!sbi->s_group_desc[i]) {
997 for (j = 0; j < i; j++)
998 brelse (sbi->s_group_desc[j]);
999 printk ("EXT2-fs: unable to read group descriptors\n");
1000 goto failed_mount_group_desc;
1003 if (!ext2_check_descriptors (sb)) {
1004 printk ("EXT2-fs: group descriptors corrupted!\n");
1007 sbi->s_gdb_count = db_count;
1008 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1009 spin_lock_init(&sbi->s_next_gen_lock);
1011 /* per fileystem reservation list head & lock */
1012 spin_lock_init(&sbi->s_rsv_window_lock);
1013 sbi->s_rsv_window_root = RB_ROOT;
1015 * Add a single, static dummy reservation to the start of the
1016 * reservation window list --- it gives us a placeholder for
1017 * append-at-start-of-list which makes the allocation logic
1020 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1021 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1022 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1023 sbi->s_rsv_window_head.rsv_goal_size = 0;
1024 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1026 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1027 ext2_count_free_blocks(sb));
1029 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1030 ext2_count_free_inodes(sb));
1033 err = percpu_counter_init(&sbi->s_dirs_counter,
1034 ext2_count_dirs(sb));
1037 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1041 * set up enough so that it can read an inode
1043 sb->s_op = &ext2_sops;
1044 sb->s_export_op = &ext2_export_ops;
1045 sb->s_xattr = ext2_xattr_handlers;
1046 root = ext2_iget(sb, EXT2_ROOT_INO);
1048 ret = PTR_ERR(root);
1051 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1053 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1057 sb->s_root = d_alloc_root(root);
1060 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1064 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1065 ext2_warning(sb, __FUNCTION__,
1066 "mounting ext3 filesystem as ext2");
1067 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1072 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1076 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1077 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1078 percpu_counter_destroy(&sbi->s_dirs_counter);
1080 for (i = 0; i < db_count; i++)
1081 brelse(sbi->s_group_desc[i]);
1082 failed_mount_group_desc:
1083 kfree(sbi->s_group_desc);
1084 kfree(sbi->s_debts);
1088 sb->s_fs_info = NULL;
1093 static void ext2_commit_super (struct super_block * sb,
1094 struct ext2_super_block * es)
1096 es->s_wtime = cpu_to_le32(get_seconds());
1097 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1101 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1103 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1104 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1105 es->s_wtime = cpu_to_le32(get_seconds());
1106 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1107 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1112 * In the second extended file system, it is not necessary to
1113 * write the super block since we use a mapping of the
1114 * disk super block in a buffer.
1116 * However, this function is still used to set the fs valid
1117 * flags to 0. We need to set this flag to 0 since the fs
1118 * may have been checked while mounted and e2fsck may have
1119 * set s_state to EXT2_VALID_FS after some corrections.
1122 void ext2_write_super (struct super_block * sb)
1124 struct ext2_super_block * es;
1126 if (!(sb->s_flags & MS_RDONLY)) {
1127 es = EXT2_SB(sb)->s_es;
1129 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1130 ext2_debug ("setting valid to 0\n");
1131 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1133 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1134 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1135 es->s_mtime = cpu_to_le32(get_seconds());
1136 ext2_sync_super(sb, es);
1138 ext2_commit_super (sb, es);
1144 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1146 struct ext2_sb_info * sbi = EXT2_SB(sb);
1147 struct ext2_super_block * es;
1148 unsigned long old_mount_opt = sbi->s_mount_opt;
1149 struct ext2_mount_options old_opts;
1150 unsigned long old_sb_flags;
1153 /* Store the old options */
1154 old_sb_flags = sb->s_flags;
1155 old_opts.s_mount_opt = sbi->s_mount_opt;
1156 old_opts.s_resuid = sbi->s_resuid;
1157 old_opts.s_resgid = sbi->s_resgid;
1160 * Allow the "check" option to be passed as a remount option.
1162 if (!parse_options (data, sbi)) {
1167 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1168 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1170 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1171 EXT2_MOUNT_XIP if not */
1173 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1174 printk("XIP: Unsupported blocksize\n");
1180 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1181 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1182 invalidate_inodes(sb))
1183 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1184 "xip remain in cache (no functional problem)");
1185 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1187 if (*flags & MS_RDONLY) {
1188 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1189 !(sbi->s_mount_state & EXT2_VALID_FS))
1192 * OK, we are remounting a valid rw partition rdonly, so set
1193 * the rdonly flag and then mark the partition as valid again.
1195 es->s_state = cpu_to_le16(sbi->s_mount_state);
1196 es->s_mtime = cpu_to_le32(get_seconds());
1198 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1199 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1201 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1202 "unsupported optional features (%x).\n",
1203 sb->s_id, le32_to_cpu(ret));
1208 * Mounting a RDONLY partition read-write, so reread and
1209 * store the current valid flag. (It may have been changed
1210 * by e2fsck since we originally mounted the partition.)
1212 sbi->s_mount_state = le16_to_cpu(es->s_state);
1213 if (!ext2_setup_super (sb, es, 0))
1214 sb->s_flags &= ~MS_RDONLY;
1216 ext2_sync_super(sb, es);
1219 sbi->s_mount_opt = old_opts.s_mount_opt;
1220 sbi->s_resuid = old_opts.s_resuid;
1221 sbi->s_resgid = old_opts.s_resgid;
1222 sb->s_flags = old_sb_flags;
1226 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1228 struct super_block *sb = dentry->d_sb;
1229 struct ext2_sb_info *sbi = EXT2_SB(sb);
1230 struct ext2_super_block *es = sbi->s_es;
1233 if (test_opt (sb, MINIX_DF))
1234 sbi->s_overhead_last = 0;
1235 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1236 unsigned long i, overhead = 0;
1240 * Compute the overhead (FS structures). This is constant
1241 * for a given filesystem unless the number of block groups
1242 * changes so we cache the previous value until it does.
1246 * All of the blocks before first_data_block are
1249 overhead = le32_to_cpu(es->s_first_data_block);
1252 * Add the overhead attributed to the superblock and
1253 * block group descriptors. If the sparse superblocks
1254 * feature is turned on, then not all groups have this.
1256 for (i = 0; i < sbi->s_groups_count; i++)
1257 overhead += ext2_bg_has_super(sb, i) +
1258 ext2_bg_num_gdb(sb, i);
1261 * Every block group has an inode bitmap, a block
1262 * bitmap, and an inode table.
1264 overhead += (sbi->s_groups_count *
1265 (2 + sbi->s_itb_per_group));
1266 sbi->s_overhead_last = overhead;
1268 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1271 buf->f_type = EXT2_SUPER_MAGIC;
1272 buf->f_bsize = sb->s_blocksize;
1273 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1274 buf->f_bfree = ext2_count_free_blocks(sb);
1275 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1276 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1277 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1279 buf->f_files = le32_to_cpu(es->s_inodes_count);
1280 buf->f_ffree = ext2_count_free_inodes(sb);
1281 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1282 buf->f_namelen = EXT2_NAME_LEN;
1283 fsid = le64_to_cpup((void *)es->s_uuid) ^
1284 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1285 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1286 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1290 static int ext2_get_sb(struct file_system_type *fs_type,
1291 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1293 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1298 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1299 * acquiring the locks... As quota files are never truncated and quota code
1300 * itself serializes the operations (and noone else should touch the files)
1301 * we don't have to be afraid of races */
1302 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1303 size_t len, loff_t off)
1305 struct inode *inode = sb_dqopt(sb)->files[type];
1306 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1308 int offset = off & (sb->s_blocksize - 1);
1311 struct buffer_head tmp_bh;
1312 struct buffer_head *bh;
1313 loff_t i_size = i_size_read(inode);
1317 if (off+len > i_size)
1320 while (toread > 0) {
1321 tocopy = sb->s_blocksize - offset < toread ?
1322 sb->s_blocksize - offset : toread;
1325 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1328 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1329 memset(data, 0, tocopy);
1331 bh = sb_bread(sb, tmp_bh.b_blocknr);
1334 memcpy(data, bh->b_data+offset, tocopy);
1345 /* Write to quotafile */
1346 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1347 const char *data, size_t len, loff_t off)
1349 struct inode *inode = sb_dqopt(sb)->files[type];
1350 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1352 int offset = off & (sb->s_blocksize - 1);
1354 size_t towrite = len;
1355 struct buffer_head tmp_bh;
1356 struct buffer_head *bh;
1358 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1359 while (towrite > 0) {
1360 tocopy = sb->s_blocksize - offset < towrite ?
1361 sb->s_blocksize - offset : towrite;
1364 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1367 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1368 bh = sb_bread(sb, tmp_bh.b_blocknr);
1370 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1376 memcpy(bh->b_data+offset, data, tocopy);
1377 flush_dcache_page(bh->b_page);
1378 set_buffer_uptodate(bh);
1379 mark_buffer_dirty(bh);
1390 if (inode->i_size < off+len-towrite)
1391 i_size_write(inode, off+len-towrite);
1393 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1394 mark_inode_dirty(inode);
1395 mutex_unlock(&inode->i_mutex);
1396 return len - towrite;
1401 static struct file_system_type ext2_fs_type = {
1402 .owner = THIS_MODULE,
1404 .get_sb = ext2_get_sb,
1405 .kill_sb = kill_block_super,
1406 .fs_flags = FS_REQUIRES_DEV,
1409 static int __init init_ext2_fs(void)
1411 int err = init_ext2_xattr();
1414 err = init_inodecache();
1417 err = register_filesystem(&ext2_fs_type);
1422 destroy_inodecache();
1428 static void __exit exit_ext2_fs(void)
1430 unregister_filesystem(&ext2_fs_type);
1431 destroy_inodecache();
1435 module_init(init_ext2_fs)
1436 module_exit(exit_ext2_fs)