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 <asm/uaccess.h>
39 static void ext2_sync_super(struct super_block *sb,
40 struct ext2_super_block *es);
41 static int ext2_remount (struct super_block * sb, int * flags, char * data);
42 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
44 void ext2_error (struct super_block * sb, const char * function,
45 const char * fmt, ...)
48 struct ext2_sb_info *sbi = EXT2_SB(sb);
49 struct ext2_super_block *es = sbi->s_es;
51 if (!(sb->s_flags & MS_RDONLY)) {
52 sbi->s_mount_state |= EXT2_ERROR_FS;
54 cpu_to_le16(le16_to_cpu(es->s_state) | EXT2_ERROR_FS);
55 ext2_sync_super(sb, es);
59 printk(KERN_CRIT "EXT2-fs error (device %s): %s: ",sb->s_id, function);
64 if (test_opt(sb, ERRORS_PANIC))
65 panic("EXT2-fs panic from previous error\n");
66 if (test_opt(sb, ERRORS_RO)) {
67 printk("Remounting filesystem read-only\n");
68 sb->s_flags |= MS_RDONLY;
72 void ext2_warning (struct super_block * sb, const char * function,
73 const char * fmt, ...)
78 printk(KERN_WARNING "EXT2-fs warning (device %s): %s: ",
85 void ext2_update_dynamic_rev(struct super_block *sb)
87 struct ext2_super_block *es = EXT2_SB(sb)->s_es;
89 if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
92 ext2_warning(sb, __FUNCTION__,
93 "updating to rev %d because of new feature flag, "
94 "running e2fsck is recommended",
97 es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
98 es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
99 es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
100 /* leave es->s_feature_*compat flags alone */
101 /* es->s_uuid will be set by e2fsck if empty */
104 * The rest of the superblock fields should be zero, and if not it
105 * means they are likely already in use, so leave them alone. We
106 * can leave it up to e2fsck to clean up any inconsistencies there.
110 static void ext2_put_super (struct super_block * sb)
114 struct ext2_sb_info *sbi = EXT2_SB(sb);
116 ext2_xattr_put_super(sb);
117 if (!(sb->s_flags & MS_RDONLY)) {
118 struct ext2_super_block *es = sbi->s_es;
120 es->s_state = cpu_to_le16(sbi->s_mount_state);
121 ext2_sync_super(sb, es);
123 db_count = sbi->s_gdb_count;
124 for (i = 0; i < db_count; i++)
125 if (sbi->s_group_desc[i])
126 brelse (sbi->s_group_desc[i]);
127 kfree(sbi->s_group_desc);
129 percpu_counter_destroy(&sbi->s_freeblocks_counter);
130 percpu_counter_destroy(&sbi->s_freeinodes_counter);
131 percpu_counter_destroy(&sbi->s_dirs_counter);
133 sb->s_fs_info = NULL;
139 static struct kmem_cache * ext2_inode_cachep;
141 static struct inode *ext2_alloc_inode(struct super_block *sb)
143 struct ext2_inode_info *ei;
144 ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
147 #ifdef CONFIG_EXT2_FS_POSIX_ACL
148 ei->i_acl = EXT2_ACL_NOT_CACHED;
149 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
151 ei->vfs_inode.i_version = 1;
152 return &ei->vfs_inode;
155 static void ext2_destroy_inode(struct inode *inode)
157 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
160 static void init_once(struct kmem_cache * cachep, void *foo)
162 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
164 rwlock_init(&ei->i_meta_lock);
165 #ifdef CONFIG_EXT2_FS_XATTR
166 init_rwsem(&ei->xattr_sem);
168 inode_init_once(&ei->vfs_inode);
171 static int init_inodecache(void)
173 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
174 sizeof(struct ext2_inode_info),
175 0, (SLAB_RECLAIM_ACCOUNT|
178 if (ext2_inode_cachep == NULL)
183 static void destroy_inodecache(void)
185 kmem_cache_destroy(ext2_inode_cachep);
188 static void ext2_clear_inode(struct inode *inode)
190 #ifdef CONFIG_EXT2_FS_POSIX_ACL
191 struct ext2_inode_info *ei = EXT2_I(inode);
193 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
194 posix_acl_release(ei->i_acl);
195 ei->i_acl = EXT2_ACL_NOT_CACHED;
197 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
198 posix_acl_release(ei->i_default_acl);
199 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
204 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
206 struct super_block *sb = vfs->mnt_sb;
207 struct ext2_sb_info *sbi = EXT2_SB(sb);
208 struct ext2_super_block *es = sbi->s_es;
209 unsigned long def_mount_opts;
211 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
213 if (sbi->s_sb_block != 1)
214 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
215 if (test_opt(sb, MINIX_DF))
216 seq_puts(seq, ",minixdf");
217 if (test_opt(sb, GRPID))
218 seq_puts(seq, ",grpid");
219 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
220 seq_puts(seq, ",nogrpid");
221 if (sbi->s_resuid != EXT2_DEF_RESUID ||
222 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
223 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
225 if (sbi->s_resgid != EXT2_DEF_RESGID ||
226 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
227 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
229 if (test_opt(sb, ERRORS_CONT)) {
230 int def_errors = le16_to_cpu(es->s_errors);
232 if (def_errors == EXT2_ERRORS_PANIC ||
233 def_errors == EXT2_ERRORS_RO) {
234 seq_puts(seq, ",errors=continue");
237 if (test_opt(sb, ERRORS_RO))
238 seq_puts(seq, ",errors=remount-ro");
239 if (test_opt(sb, ERRORS_PANIC))
240 seq_puts(seq, ",errors=panic");
241 if (test_opt(sb, NO_UID32))
242 seq_puts(seq, ",nouid32");
243 if (test_opt(sb, DEBUG))
244 seq_puts(seq, ",debug");
245 if (test_opt(sb, OLDALLOC))
246 seq_puts(seq, ",oldalloc");
248 #ifdef CONFIG_EXT2_FS_XATTR
249 if (test_opt(sb, XATTR_USER))
250 seq_puts(seq, ",user_xattr");
251 if (!test_opt(sb, XATTR_USER) &&
252 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
253 seq_puts(seq, ",nouser_xattr");
257 #ifdef CONFIG_EXT2_FS_POSIX_ACL
258 if (test_opt(sb, POSIX_ACL))
259 seq_puts(seq, ",acl");
260 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
261 seq_puts(seq, ",noacl");
264 if (test_opt(sb, NOBH))
265 seq_puts(seq, ",nobh");
267 #if defined(CONFIG_QUOTA)
268 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
269 seq_puts(seq, ",usrquota");
271 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
272 seq_puts(seq, ",grpquota");
275 #if defined(CONFIG_EXT2_FS_XIP)
276 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
277 seq_puts(seq, ",xip");
284 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
285 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
288 static const struct super_operations ext2_sops = {
289 .alloc_inode = ext2_alloc_inode,
290 .destroy_inode = ext2_destroy_inode,
291 .read_inode = ext2_read_inode,
292 .write_inode = ext2_write_inode,
293 .put_inode = ext2_put_inode,
294 .delete_inode = ext2_delete_inode,
295 .put_super = ext2_put_super,
296 .write_super = ext2_write_super,
297 .statfs = ext2_statfs,
298 .remount_fs = ext2_remount,
299 .clear_inode = ext2_clear_inode,
300 .show_options = ext2_show_options,
302 .quota_read = ext2_quota_read,
303 .quota_write = ext2_quota_write,
307 static struct dentry *ext2_get_dentry(struct super_block *sb, void *vobjp)
310 unsigned long ino = objp[0];
311 __u32 generation = objp[1];
313 struct dentry *result;
315 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
316 return ERR_PTR(-ESTALE);
317 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
318 return ERR_PTR(-ESTALE);
320 /* iget isn't really right if the inode is currently unallocated!!
321 * ext2_read_inode currently does appropriate checks, but
322 * it might be "neater" to call ext2_get_inode first and check
323 * if the inode is valid.....
325 inode = iget(sb, ino);
327 return ERR_PTR(-ENOMEM);
328 if (is_bad_inode(inode) ||
329 (generation && inode->i_generation != generation)) {
330 /* we didn't find the right inode.. */
332 return ERR_PTR(-ESTALE);
334 /* now to find a dentry.
335 * If possible, get a well-connected one
337 result = d_alloc_anon(inode);
340 return ERR_PTR(-ENOMEM);
345 /* Yes, most of these are left as NULL!!
346 * A NULL value implies the default, which works with ext2-like file
347 * systems, but can be improved upon.
348 * Currently only get_parent is required.
350 static struct export_operations ext2_export_ops = {
351 .get_parent = ext2_get_parent,
352 .get_dentry = ext2_get_dentry,
355 static unsigned long get_sb_block(void **data)
357 unsigned long sb_block;
358 char *options = (char *) *data;
360 if (!options || strncmp(options, "sb=", 3) != 0)
361 return 1; /* Default location */
363 sb_block = simple_strtoul(options, &options, 0);
364 if (*options && *options != ',') {
365 printk("EXT2-fs: Invalid sb specification: %s\n",
371 *data = (void *) options;
376 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
377 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
378 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
379 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
380 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
381 Opt_usrquota, Opt_grpquota
384 static match_table_t tokens = {
385 {Opt_bsd_df, "bsddf"},
386 {Opt_minix_df, "minixdf"},
387 {Opt_grpid, "grpid"},
388 {Opt_grpid, "bsdgroups"},
389 {Opt_nogrpid, "nogrpid"},
390 {Opt_nogrpid, "sysvgroups"},
391 {Opt_resgid, "resgid=%u"},
392 {Opt_resuid, "resuid=%u"},
394 {Opt_err_cont, "errors=continue"},
395 {Opt_err_panic, "errors=panic"},
396 {Opt_err_ro, "errors=remount-ro"},
397 {Opt_nouid32, "nouid32"},
398 {Opt_nocheck, "check=none"},
399 {Opt_nocheck, "nocheck"},
400 {Opt_debug, "debug"},
401 {Opt_oldalloc, "oldalloc"},
402 {Opt_orlov, "orlov"},
404 {Opt_user_xattr, "user_xattr"},
405 {Opt_nouser_xattr, "nouser_xattr"},
407 {Opt_noacl, "noacl"},
409 {Opt_grpquota, "grpquota"},
410 {Opt_ignore, "noquota"},
411 {Opt_quota, "quota"},
412 {Opt_usrquota, "usrquota"},
416 static int parse_options (char * options,
417 struct ext2_sb_info *sbi)
420 substring_t args[MAX_OPT_ARGS];
426 while ((p = strsep (&options, ",")) != NULL) {
431 token = match_token(p, tokens, args);
434 clear_opt (sbi->s_mount_opt, MINIX_DF);
437 set_opt (sbi->s_mount_opt, MINIX_DF);
440 set_opt (sbi->s_mount_opt, GRPID);
443 clear_opt (sbi->s_mount_opt, GRPID);
446 if (match_int(&args[0], &option))
448 sbi->s_resuid = option;
451 if (match_int(&args[0], &option))
453 sbi->s_resgid = option;
456 /* handled by get_sb_block() instead of here */
457 /* *sb_block = match_int(&args[0]); */
460 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
461 clear_opt (sbi->s_mount_opt, ERRORS_RO);
462 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
465 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
466 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
467 set_opt (sbi->s_mount_opt, ERRORS_RO);
470 clear_opt (sbi->s_mount_opt, ERRORS_RO);
471 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
472 set_opt (sbi->s_mount_opt, ERRORS_CONT);
475 set_opt (sbi->s_mount_opt, NO_UID32);
478 clear_opt (sbi->s_mount_opt, CHECK);
481 set_opt (sbi->s_mount_opt, DEBUG);
484 set_opt (sbi->s_mount_opt, OLDALLOC);
487 clear_opt (sbi->s_mount_opt, OLDALLOC);
490 set_opt (sbi->s_mount_opt, NOBH);
492 #ifdef CONFIG_EXT2_FS_XATTR
494 set_opt (sbi->s_mount_opt, XATTR_USER);
496 case Opt_nouser_xattr:
497 clear_opt (sbi->s_mount_opt, XATTR_USER);
501 case Opt_nouser_xattr:
502 printk("EXT2 (no)user_xattr options not supported\n");
505 #ifdef CONFIG_EXT2_FS_POSIX_ACL
507 set_opt(sbi->s_mount_opt, POSIX_ACL);
510 clear_opt(sbi->s_mount_opt, POSIX_ACL);
515 printk("EXT2 (no)acl options not supported\n");
519 #ifdef CONFIG_EXT2_FS_XIP
520 set_opt (sbi->s_mount_opt, XIP);
522 printk("EXT2 xip option not supported\n");
526 #if defined(CONFIG_QUOTA)
529 set_opt(sbi->s_mount_opt, USRQUOTA);
533 set_opt(sbi->s_mount_opt, GRPQUOTA);
540 "EXT2-fs: quota operations not supported.\n");
554 static int ext2_setup_super (struct super_block * sb,
555 struct ext2_super_block * es,
559 struct ext2_sb_info *sbi = EXT2_SB(sb);
561 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
562 printk ("EXT2-fs warning: revision level too high, "
563 "forcing read-only mode\n");
568 if (!(sbi->s_mount_state & EXT2_VALID_FS))
569 printk ("EXT2-fs warning: mounting unchecked fs, "
570 "running e2fsck is recommended\n");
571 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
572 printk ("EXT2-fs warning: mounting fs with errors, "
573 "running e2fsck is recommended\n");
574 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
575 le16_to_cpu(es->s_mnt_count) >=
576 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
577 printk ("EXT2-fs warning: maximal mount count reached, "
578 "running e2fsck is recommended\n");
579 else if (le32_to_cpu(es->s_checkinterval) &&
580 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
581 printk ("EXT2-fs warning: checktime reached, "
582 "running e2fsck is recommended\n");
583 if (!le16_to_cpu(es->s_max_mnt_count))
584 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
585 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
586 ext2_write_super(sb);
587 if (test_opt (sb, DEBUG))
588 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
589 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
590 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
593 EXT2_BLOCKS_PER_GROUP(sb),
594 EXT2_INODES_PER_GROUP(sb),
599 static int ext2_check_descriptors (struct super_block * sb)
603 struct ext2_sb_info *sbi = EXT2_SB(sb);
604 unsigned long first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
605 unsigned long last_block;
606 struct ext2_group_desc * gdp = NULL;
608 ext2_debug ("Checking group descriptors");
610 for (i = 0; i < sbi->s_groups_count; i++)
612 if (i == sbi->s_groups_count - 1)
613 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
615 last_block = first_block +
616 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
618 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0)
619 gdp = (struct ext2_group_desc *) sbi->s_group_desc[desc_block++]->b_data;
620 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
621 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
623 ext2_error (sb, "ext2_check_descriptors",
624 "Block bitmap for group %d"
625 " not in group (block %lu)!",
626 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
629 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
630 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
632 ext2_error (sb, "ext2_check_descriptors",
633 "Inode bitmap for group %d"
634 " not in group (block %lu)!",
635 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
638 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
639 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
642 ext2_error (sb, "ext2_check_descriptors",
643 "Inode table for group %d"
644 " not in group (block %lu)!",
645 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
648 first_block += EXT2_BLOCKS_PER_GROUP(sb);
655 * Maximal file size. There is a direct, and {,double-,triple-}indirect
656 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
657 * We need to be 1 filesystem block less than the 2^32 sector limit.
659 static loff_t ext2_max_size(int bits)
661 loff_t res = EXT2_NDIR_BLOCKS;
662 /* This constant is calculated to be the largest file size for a
663 * dense, 4k-blocksize file such that the total number of
664 * sectors in the file, including data and all indirect blocks,
665 * does not exceed 2^32. */
666 const loff_t upper_limit = 0x1ff7fffd000LL;
668 res += 1LL << (bits-2);
669 res += 1LL << (2*(bits-2));
670 res += 1LL << (3*(bits-2));
672 if (res > upper_limit)
677 static unsigned long descriptor_loc(struct super_block *sb,
678 unsigned long logic_sb_block,
681 struct ext2_sb_info *sbi = EXT2_SB(sb);
682 unsigned long bg, first_data_block, first_meta_bg;
685 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
686 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
688 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
690 return (logic_sb_block + nr + 1);
691 bg = sbi->s_desc_per_block * nr;
692 if (ext2_bg_has_super(sb, bg))
694 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
697 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
699 struct buffer_head * bh;
700 struct ext2_sb_info * sbi;
701 struct ext2_super_block * es;
704 unsigned long sb_block = get_sb_block(&data);
705 unsigned long logic_sb_block;
706 unsigned long offset = 0;
707 unsigned long def_mount_opts;
708 int blocksize = BLOCK_SIZE;
714 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
718 sbi->s_sb_block = sb_block;
721 * See what the current blocksize for the device is, and
722 * use that as the blocksize. Otherwise (or if the blocksize
723 * is smaller than the default) use the default.
724 * This is important for devices that have a hardware
725 * sectorsize that is larger than the default.
727 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
729 printk ("EXT2-fs: unable to set blocksize\n");
734 * If the superblock doesn't start on a hardware sector boundary,
735 * calculate the offset.
737 if (blocksize != BLOCK_SIZE) {
738 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
739 offset = (sb_block*BLOCK_SIZE) % blocksize;
741 logic_sb_block = sb_block;
744 if (!(bh = sb_bread(sb, logic_sb_block))) {
745 printk ("EXT2-fs: unable to read superblock\n");
749 * Note: s_es must be initialized as soon as possible because
750 * some ext2 macro-instructions depend on its value
752 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
754 sb->s_magic = le16_to_cpu(es->s_magic);
756 if (sb->s_magic != EXT2_SUPER_MAGIC)
759 /* Set defaults before we parse the mount options */
760 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
761 if (def_mount_opts & EXT2_DEFM_DEBUG)
762 set_opt(sbi->s_mount_opt, DEBUG);
763 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
764 set_opt(sbi->s_mount_opt, GRPID);
765 if (def_mount_opts & EXT2_DEFM_UID16)
766 set_opt(sbi->s_mount_opt, NO_UID32);
767 #ifdef CONFIG_EXT2_FS_XATTR
768 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
769 set_opt(sbi->s_mount_opt, XATTR_USER);
771 #ifdef CONFIG_EXT2_FS_POSIX_ACL
772 if (def_mount_opts & EXT2_DEFM_ACL)
773 set_opt(sbi->s_mount_opt, POSIX_ACL);
776 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
777 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
778 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO)
779 set_opt(sbi->s_mount_opt, ERRORS_RO);
781 set_opt(sbi->s_mount_opt, ERRORS_CONT);
783 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
784 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
786 if (!parse_options ((char *) data, sbi))
789 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
790 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
793 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
794 EXT2_MOUNT_XIP if not */
796 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
797 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
798 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
799 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
800 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
801 "running e2fsck is recommended\n");
803 * Check feature flags regardless of the revision level, since we
804 * previously didn't change the revision level when setting the flags,
805 * so there is a chance incompat flags are set on a rev 0 filesystem.
807 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
809 printk("EXT2-fs: %s: couldn't mount because of "
810 "unsupported optional features (%x).\n",
811 sb->s_id, le32_to_cpu(features));
814 if (!(sb->s_flags & MS_RDONLY) &&
815 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
816 printk("EXT2-fs: %s: couldn't mount RDWR because of "
817 "unsupported optional features (%x).\n",
818 sb->s_id, le32_to_cpu(features));
822 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
824 if ((ext2_use_xip(sb)) && ((blocksize != PAGE_SIZE) ||
825 (sb->s_blocksize != blocksize))) {
827 printk("XIP: Unsupported blocksize\n");
831 /* If the blocksize doesn't match, re-read the thing.. */
832 if (sb->s_blocksize != blocksize) {
835 if (!sb_set_blocksize(sb, blocksize)) {
836 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
840 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
841 offset = (sb_block*BLOCK_SIZE) % blocksize;
842 bh = sb_bread(sb, logic_sb_block);
844 printk("EXT2-fs: Couldn't read superblock on "
848 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
850 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
851 printk ("EXT2-fs: Magic mismatch, very weird !\n");
856 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
858 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
859 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
860 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
862 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
863 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
864 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
865 (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
866 (sbi->s_inode_size > blocksize)) {
867 printk ("EXT2-fs: unsupported inode size: %d\n",
873 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
874 le32_to_cpu(es->s_log_frag_size);
875 if (sbi->s_frag_size == 0)
877 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
879 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
880 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
881 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
883 if (EXT2_INODE_SIZE(sb) == 0)
885 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
886 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
888 sbi->s_itb_per_group = sbi->s_inodes_per_group /
889 sbi->s_inodes_per_block;
890 sbi->s_desc_per_block = sb->s_blocksize /
891 sizeof (struct ext2_group_desc);
893 sbi->s_mount_state = le16_to_cpu(es->s_state);
894 sbi->s_addr_per_block_bits =
895 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
896 sbi->s_desc_per_block_bits =
897 ilog2 (EXT2_DESC_PER_BLOCK(sb));
899 if (sb->s_magic != EXT2_SUPER_MAGIC)
902 if (sb->s_blocksize != bh->b_size) {
904 printk ("VFS: Unsupported blocksize on dev "
909 if (sb->s_blocksize != sbi->s_frag_size) {
910 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
911 sbi->s_frag_size, sb->s_blocksize);
915 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
916 printk ("EXT2-fs: #blocks per group too big: %lu\n",
917 sbi->s_blocks_per_group);
920 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
921 printk ("EXT2-fs: #fragments per group too big: %lu\n",
922 sbi->s_frags_per_group);
925 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
926 printk ("EXT2-fs: #inodes per group too big: %lu\n",
927 sbi->s_inodes_per_group);
931 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
933 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
934 le32_to_cpu(es->s_first_data_block) - 1)
935 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
936 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
937 EXT2_DESC_PER_BLOCK(sb);
938 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
939 if (sbi->s_group_desc == NULL) {
940 printk ("EXT2-fs: not enough memory\n");
943 bgl_lock_init(&sbi->s_blockgroup_lock);
944 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
946 printk ("EXT2-fs: not enough memory\n");
947 goto failed_mount_group_desc;
949 for (i = 0; i < db_count; i++) {
950 block = descriptor_loc(sb, logic_sb_block, i);
951 sbi->s_group_desc[i] = sb_bread(sb, block);
952 if (!sbi->s_group_desc[i]) {
953 for (j = 0; j < i; j++)
954 brelse (sbi->s_group_desc[j]);
955 printk ("EXT2-fs: unable to read group descriptors\n");
956 goto failed_mount_group_desc;
959 if (!ext2_check_descriptors (sb)) {
960 printk ("EXT2-fs: group descriptors corrupted!\n");
963 sbi->s_gdb_count = db_count;
964 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
965 spin_lock_init(&sbi->s_next_gen_lock);
967 err = percpu_counter_init(&sbi->s_freeblocks_counter,
968 ext2_count_free_blocks(sb));
970 err = percpu_counter_init(&sbi->s_freeinodes_counter,
971 ext2_count_free_inodes(sb));
974 err = percpu_counter_init(&sbi->s_dirs_counter,
975 ext2_count_dirs(sb));
978 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
982 * set up enough so that it can read an inode
984 sb->s_op = &ext2_sops;
985 sb->s_export_op = &ext2_export_ops;
986 sb->s_xattr = ext2_xattr_handlers;
987 root = iget(sb, EXT2_ROOT_INO);
988 sb->s_root = d_alloc_root(root);
991 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
994 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
997 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1000 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1001 ext2_warning(sb, __FUNCTION__,
1002 "mounting ext3 filesystem as ext2");
1003 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1008 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1012 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1013 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1014 percpu_counter_destroy(&sbi->s_dirs_counter);
1016 for (i = 0; i < db_count; i++)
1017 brelse(sbi->s_group_desc[i]);
1018 failed_mount_group_desc:
1019 kfree(sbi->s_group_desc);
1020 kfree(sbi->s_debts);
1024 sb->s_fs_info = NULL;
1029 static void ext2_commit_super (struct super_block * sb,
1030 struct ext2_super_block * es)
1032 es->s_wtime = cpu_to_le32(get_seconds());
1033 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1037 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1039 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1040 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1041 es->s_wtime = cpu_to_le32(get_seconds());
1042 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1043 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1048 * In the second extended file system, it is not necessary to
1049 * write the super block since we use a mapping of the
1050 * disk super block in a buffer.
1052 * However, this function is still used to set the fs valid
1053 * flags to 0. We need to set this flag to 0 since the fs
1054 * may have been checked while mounted and e2fsck may have
1055 * set s_state to EXT2_VALID_FS after some corrections.
1058 void ext2_write_super (struct super_block * sb)
1060 struct ext2_super_block * es;
1062 if (!(sb->s_flags & MS_RDONLY)) {
1063 es = EXT2_SB(sb)->s_es;
1065 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1066 ext2_debug ("setting valid to 0\n");
1067 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1069 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1070 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1071 es->s_mtime = cpu_to_le32(get_seconds());
1072 ext2_sync_super(sb, es);
1074 ext2_commit_super (sb, es);
1080 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1082 struct ext2_sb_info * sbi = EXT2_SB(sb);
1083 struct ext2_super_block * es;
1084 unsigned long old_mount_opt = sbi->s_mount_opt;
1085 struct ext2_mount_options old_opts;
1086 unsigned long old_sb_flags;
1089 /* Store the old options */
1090 old_sb_flags = sb->s_flags;
1091 old_opts.s_mount_opt = sbi->s_mount_opt;
1092 old_opts.s_resuid = sbi->s_resuid;
1093 old_opts.s_resgid = sbi->s_resgid;
1096 * Allow the "check" option to be passed as a remount option.
1098 if (!parse_options (data, sbi)) {
1103 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1104 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1106 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1107 EXT2_MOUNT_XIP if not */
1109 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1110 printk("XIP: Unsupported blocksize\n");
1116 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1117 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1118 invalidate_inodes(sb))
1119 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1120 "xip remain in cache (no functional problem)");
1121 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1123 if (*flags & MS_RDONLY) {
1124 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1125 !(sbi->s_mount_state & EXT2_VALID_FS))
1128 * OK, we are remounting a valid rw partition rdonly, so set
1129 * the rdonly flag and then mark the partition as valid again.
1131 es->s_state = cpu_to_le16(sbi->s_mount_state);
1132 es->s_mtime = cpu_to_le32(get_seconds());
1134 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1135 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1137 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1138 "unsupported optional features (%x).\n",
1139 sb->s_id, le32_to_cpu(ret));
1144 * Mounting a RDONLY partition read-write, so reread and
1145 * store the current valid flag. (It may have been changed
1146 * by e2fsck since we originally mounted the partition.)
1148 sbi->s_mount_state = le16_to_cpu(es->s_state);
1149 if (!ext2_setup_super (sb, es, 0))
1150 sb->s_flags &= ~MS_RDONLY;
1152 ext2_sync_super(sb, es);
1155 sbi->s_mount_opt = old_opts.s_mount_opt;
1156 sbi->s_resuid = old_opts.s_resuid;
1157 sbi->s_resgid = old_opts.s_resgid;
1158 sb->s_flags = old_sb_flags;
1162 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1164 struct super_block *sb = dentry->d_sb;
1165 struct ext2_sb_info *sbi = EXT2_SB(sb);
1166 struct ext2_super_block *es = sbi->s_es;
1169 if (test_opt (sb, MINIX_DF))
1170 sbi->s_overhead_last = 0;
1171 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1172 unsigned long i, overhead = 0;
1176 * Compute the overhead (FS structures). This is constant
1177 * for a given filesystem unless the number of block groups
1178 * changes so we cache the previous value until it does.
1182 * All of the blocks before first_data_block are
1185 overhead = le32_to_cpu(es->s_first_data_block);
1188 * Add the overhead attributed to the superblock and
1189 * block group descriptors. If the sparse superblocks
1190 * feature is turned on, then not all groups have this.
1192 for (i = 0; i < sbi->s_groups_count; i++)
1193 overhead += ext2_bg_has_super(sb, i) +
1194 ext2_bg_num_gdb(sb, i);
1197 * Every block group has an inode bitmap, a block
1198 * bitmap, and an inode table.
1200 overhead += (sbi->s_groups_count *
1201 (2 + sbi->s_itb_per_group));
1202 sbi->s_overhead_last = overhead;
1204 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1207 buf->f_type = EXT2_SUPER_MAGIC;
1208 buf->f_bsize = sb->s_blocksize;
1209 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1210 buf->f_bfree = ext2_count_free_blocks(sb);
1211 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1212 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1213 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1215 buf->f_files = le32_to_cpu(es->s_inodes_count);
1216 buf->f_ffree = ext2_count_free_inodes(sb);
1217 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1218 buf->f_namelen = EXT2_NAME_LEN;
1219 fsid = le64_to_cpup((void *)es->s_uuid) ^
1220 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1221 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1222 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1226 static int ext2_get_sb(struct file_system_type *fs_type,
1227 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1229 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1234 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1235 * acquiring the locks... As quota files are never truncated and quota code
1236 * itself serializes the operations (and noone else should touch the files)
1237 * we don't have to be afraid of races */
1238 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1239 size_t len, loff_t off)
1241 struct inode *inode = sb_dqopt(sb)->files[type];
1242 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1244 int offset = off & (sb->s_blocksize - 1);
1247 struct buffer_head tmp_bh;
1248 struct buffer_head *bh;
1249 loff_t i_size = i_size_read(inode);
1253 if (off+len > i_size)
1256 while (toread > 0) {
1257 tocopy = sb->s_blocksize - offset < toread ?
1258 sb->s_blocksize - offset : toread;
1261 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1264 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1265 memset(data, 0, tocopy);
1267 bh = sb_bread(sb, tmp_bh.b_blocknr);
1270 memcpy(data, bh->b_data+offset, tocopy);
1281 /* Write to quotafile */
1282 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1283 const char *data, size_t len, loff_t off)
1285 struct inode *inode = sb_dqopt(sb)->files[type];
1286 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1288 int offset = off & (sb->s_blocksize - 1);
1290 size_t towrite = len;
1291 struct buffer_head tmp_bh;
1292 struct buffer_head *bh;
1294 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1295 while (towrite > 0) {
1296 tocopy = sb->s_blocksize - offset < towrite ?
1297 sb->s_blocksize - offset : towrite;
1300 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1303 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1304 bh = sb_bread(sb, tmp_bh.b_blocknr);
1306 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1312 memcpy(bh->b_data+offset, data, tocopy);
1313 flush_dcache_page(bh->b_page);
1314 set_buffer_uptodate(bh);
1315 mark_buffer_dirty(bh);
1326 if (inode->i_size < off+len-towrite)
1327 i_size_write(inode, off+len-towrite);
1329 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1330 mark_inode_dirty(inode);
1331 mutex_unlock(&inode->i_mutex);
1332 return len - towrite;
1337 static struct file_system_type ext2_fs_type = {
1338 .owner = THIS_MODULE,
1340 .get_sb = ext2_get_sb,
1341 .kill_sb = kill_block_super,
1342 .fs_flags = FS_REQUIRES_DEV,
1345 static int __init init_ext2_fs(void)
1347 int err = init_ext2_xattr();
1350 err = init_inodecache();
1353 err = register_filesystem(&ext2_fs_type);
1358 destroy_inodecache();
1364 static void __exit exit_ext2_fs(void)
1366 unregister_filesystem(&ext2_fs_type);
1367 destroy_inodecache();
1371 module_init(init_ext2_fs)
1372 module_exit(exit_ext2_fs)