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 kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
163 static void init_once(void *foo)
165 struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
167 rwlock_init(&ei->i_meta_lock);
168 #ifdef CONFIG_EXT2_FS_XATTR
169 init_rwsem(&ei->xattr_sem);
171 mutex_init(&ei->truncate_mutex);
172 inode_init_once(&ei->vfs_inode);
175 static int init_inodecache(void)
177 ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
178 sizeof(struct ext2_inode_info),
179 0, (SLAB_RECLAIM_ACCOUNT|
182 if (ext2_inode_cachep == NULL)
187 static void destroy_inodecache(void)
189 kmem_cache_destroy(ext2_inode_cachep);
192 static void ext2_clear_inode(struct inode *inode)
194 struct ext2_block_alloc_info *rsv = EXT2_I(inode)->i_block_alloc_info;
195 #ifdef CONFIG_EXT2_FS_POSIX_ACL
196 struct ext2_inode_info *ei = EXT2_I(inode);
198 if (ei->i_acl && ei->i_acl != EXT2_ACL_NOT_CACHED) {
199 posix_acl_release(ei->i_acl);
200 ei->i_acl = EXT2_ACL_NOT_CACHED;
202 if (ei->i_default_acl && ei->i_default_acl != EXT2_ACL_NOT_CACHED) {
203 posix_acl_release(ei->i_default_acl);
204 ei->i_default_acl = EXT2_ACL_NOT_CACHED;
207 ext2_discard_reservation(inode);
208 EXT2_I(inode)->i_block_alloc_info = NULL;
213 static int ext2_show_options(struct seq_file *seq, struct vfsmount *vfs)
215 struct super_block *sb = vfs->mnt_sb;
216 struct ext2_sb_info *sbi = EXT2_SB(sb);
217 struct ext2_super_block *es = sbi->s_es;
218 unsigned long def_mount_opts;
220 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
222 if (sbi->s_sb_block != 1)
223 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
224 if (test_opt(sb, MINIX_DF))
225 seq_puts(seq, ",minixdf");
226 if (test_opt(sb, GRPID))
227 seq_puts(seq, ",grpid");
228 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
229 seq_puts(seq, ",nogrpid");
230 if (sbi->s_resuid != EXT2_DEF_RESUID ||
231 le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
232 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
234 if (sbi->s_resgid != EXT2_DEF_RESGID ||
235 le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
236 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
238 if (test_opt(sb, ERRORS_RO)) {
239 int def_errors = le16_to_cpu(es->s_errors);
241 if (def_errors == EXT2_ERRORS_PANIC ||
242 def_errors == EXT2_ERRORS_CONTINUE) {
243 seq_puts(seq, ",errors=remount-ro");
246 if (test_opt(sb, ERRORS_CONT))
247 seq_puts(seq, ",errors=continue");
248 if (test_opt(sb, ERRORS_PANIC))
249 seq_puts(seq, ",errors=panic");
250 if (test_opt(sb, NO_UID32))
251 seq_puts(seq, ",nouid32");
252 if (test_opt(sb, DEBUG))
253 seq_puts(seq, ",debug");
254 if (test_opt(sb, OLDALLOC))
255 seq_puts(seq, ",oldalloc");
257 #ifdef CONFIG_EXT2_FS_XATTR
258 if (test_opt(sb, XATTR_USER))
259 seq_puts(seq, ",user_xattr");
260 if (!test_opt(sb, XATTR_USER) &&
261 (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
262 seq_puts(seq, ",nouser_xattr");
266 #ifdef CONFIG_EXT2_FS_POSIX_ACL
267 if (test_opt(sb, POSIX_ACL))
268 seq_puts(seq, ",acl");
269 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
270 seq_puts(seq, ",noacl");
273 if (test_opt(sb, NOBH))
274 seq_puts(seq, ",nobh");
276 #if defined(CONFIG_QUOTA)
277 if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
278 seq_puts(seq, ",usrquota");
280 if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
281 seq_puts(seq, ",grpquota");
284 #if defined(CONFIG_EXT2_FS_XIP)
285 if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
286 seq_puts(seq, ",xip");
289 if (!test_opt(sb, RESERVATION))
290 seq_puts(seq, ",noreservation");
296 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
297 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
300 static const struct super_operations ext2_sops = {
301 .alloc_inode = ext2_alloc_inode,
302 .destroy_inode = ext2_destroy_inode,
303 .write_inode = ext2_write_inode,
304 .delete_inode = ext2_delete_inode,
305 .put_super = ext2_put_super,
306 .write_super = ext2_write_super,
307 .statfs = ext2_statfs,
308 .remount_fs = ext2_remount,
309 .clear_inode = ext2_clear_inode,
310 .show_options = ext2_show_options,
312 .quota_read = ext2_quota_read,
313 .quota_write = ext2_quota_write,
317 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
318 u64 ino, u32 generation)
322 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
323 return ERR_PTR(-ESTALE);
324 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
325 return ERR_PTR(-ESTALE);
327 /* iget isn't really right if the inode is currently unallocated!!
328 * ext2_read_inode currently does appropriate checks, but
329 * it might be "neater" to call ext2_get_inode first and check
330 * if the inode is valid.....
332 inode = ext2_iget(sb, ino);
334 return ERR_CAST(inode);
335 if (generation && inode->i_generation != generation) {
336 /* we didn't find the right inode.. */
338 return ERR_PTR(-ESTALE);
343 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
344 int fh_len, int fh_type)
346 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
350 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
351 int fh_len, int fh_type)
353 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
357 /* Yes, most of these are left as NULL!!
358 * A NULL value implies the default, which works with ext2-like file
359 * systems, but can be improved upon.
360 * Currently only get_parent is required.
362 static const struct export_operations ext2_export_ops = {
363 .fh_to_dentry = ext2_fh_to_dentry,
364 .fh_to_parent = ext2_fh_to_parent,
365 .get_parent = ext2_get_parent,
368 static unsigned long get_sb_block(void **data)
370 unsigned long sb_block;
371 char *options = (char *) *data;
373 if (!options || strncmp(options, "sb=", 3) != 0)
374 return 1; /* Default location */
376 sb_block = simple_strtoul(options, &options, 0);
377 if (*options && *options != ',') {
378 printk("EXT2-fs: Invalid sb specification: %s\n",
384 *data = (void *) options;
389 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
390 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
391 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
392 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
393 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
394 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
397 static const match_table_t tokens = {
398 {Opt_bsd_df, "bsddf"},
399 {Opt_minix_df, "minixdf"},
400 {Opt_grpid, "grpid"},
401 {Opt_grpid, "bsdgroups"},
402 {Opt_nogrpid, "nogrpid"},
403 {Opt_nogrpid, "sysvgroups"},
404 {Opt_resgid, "resgid=%u"},
405 {Opt_resuid, "resuid=%u"},
407 {Opt_err_cont, "errors=continue"},
408 {Opt_err_panic, "errors=panic"},
409 {Opt_err_ro, "errors=remount-ro"},
410 {Opt_nouid32, "nouid32"},
411 {Opt_nocheck, "check=none"},
412 {Opt_nocheck, "nocheck"},
413 {Opt_debug, "debug"},
414 {Opt_oldalloc, "oldalloc"},
415 {Opt_orlov, "orlov"},
417 {Opt_user_xattr, "user_xattr"},
418 {Opt_nouser_xattr, "nouser_xattr"},
420 {Opt_noacl, "noacl"},
422 {Opt_grpquota, "grpquota"},
423 {Opt_ignore, "noquota"},
424 {Opt_quota, "quota"},
425 {Opt_usrquota, "usrquota"},
426 {Opt_reservation, "reservation"},
427 {Opt_noreservation, "noreservation"},
431 static int parse_options (char * options,
432 struct ext2_sb_info *sbi)
435 substring_t args[MAX_OPT_ARGS];
441 while ((p = strsep (&options, ",")) != NULL) {
446 token = match_token(p, tokens, args);
449 clear_opt (sbi->s_mount_opt, MINIX_DF);
452 set_opt (sbi->s_mount_opt, MINIX_DF);
455 set_opt (sbi->s_mount_opt, GRPID);
458 clear_opt (sbi->s_mount_opt, GRPID);
461 if (match_int(&args[0], &option))
463 sbi->s_resuid = option;
466 if (match_int(&args[0], &option))
468 sbi->s_resgid = option;
471 /* handled by get_sb_block() instead of here */
472 /* *sb_block = match_int(&args[0]); */
475 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
476 clear_opt (sbi->s_mount_opt, ERRORS_RO);
477 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
480 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
481 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
482 set_opt (sbi->s_mount_opt, ERRORS_RO);
485 clear_opt (sbi->s_mount_opt, ERRORS_RO);
486 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
487 set_opt (sbi->s_mount_opt, ERRORS_CONT);
490 set_opt (sbi->s_mount_opt, NO_UID32);
493 clear_opt (sbi->s_mount_opt, CHECK);
496 set_opt (sbi->s_mount_opt, DEBUG);
499 set_opt (sbi->s_mount_opt, OLDALLOC);
502 clear_opt (sbi->s_mount_opt, OLDALLOC);
505 set_opt (sbi->s_mount_opt, NOBH);
507 #ifdef CONFIG_EXT2_FS_XATTR
509 set_opt (sbi->s_mount_opt, XATTR_USER);
511 case Opt_nouser_xattr:
512 clear_opt (sbi->s_mount_opt, XATTR_USER);
516 case Opt_nouser_xattr:
517 printk("EXT2 (no)user_xattr options not supported\n");
520 #ifdef CONFIG_EXT2_FS_POSIX_ACL
522 set_opt(sbi->s_mount_opt, POSIX_ACL);
525 clear_opt(sbi->s_mount_opt, POSIX_ACL);
530 printk("EXT2 (no)acl options not supported\n");
534 #ifdef CONFIG_EXT2_FS_XIP
535 set_opt (sbi->s_mount_opt, XIP);
537 printk("EXT2 xip option not supported\n");
541 #if defined(CONFIG_QUOTA)
544 set_opt(sbi->s_mount_opt, USRQUOTA);
548 set_opt(sbi->s_mount_opt, GRPQUOTA);
555 "EXT2-fs: quota operations not supported.\n");
560 case Opt_reservation:
561 set_opt(sbi->s_mount_opt, RESERVATION);
562 printk("reservations ON\n");
564 case Opt_noreservation:
565 clear_opt(sbi->s_mount_opt, RESERVATION);
566 printk("reservations OFF\n");
577 static int ext2_setup_super (struct super_block * sb,
578 struct ext2_super_block * es,
582 struct ext2_sb_info *sbi = EXT2_SB(sb);
584 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
585 printk ("EXT2-fs warning: revision level too high, "
586 "forcing read-only mode\n");
591 if (!(sbi->s_mount_state & EXT2_VALID_FS))
592 printk ("EXT2-fs warning: mounting unchecked fs, "
593 "running e2fsck is recommended\n");
594 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
595 printk ("EXT2-fs warning: mounting fs with errors, "
596 "running e2fsck is recommended\n");
597 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
598 le16_to_cpu(es->s_mnt_count) >=
599 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
600 printk ("EXT2-fs warning: maximal mount count reached, "
601 "running e2fsck is recommended\n");
602 else if (le32_to_cpu(es->s_checkinterval) &&
603 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
604 printk ("EXT2-fs warning: checktime reached, "
605 "running e2fsck is recommended\n");
606 if (!le16_to_cpu(es->s_max_mnt_count))
607 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
608 le16_add_cpu(&es->s_mnt_count, 1);
609 ext2_write_super(sb);
610 if (test_opt (sb, DEBUG))
611 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
612 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
613 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
616 EXT2_BLOCKS_PER_GROUP(sb),
617 EXT2_INODES_PER_GROUP(sb),
622 static int ext2_check_descriptors(struct super_block *sb)
625 struct ext2_sb_info *sbi = EXT2_SB(sb);
627 ext2_debug ("Checking group descriptors");
629 for (i = 0; i < sbi->s_groups_count; i++) {
630 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
631 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
632 ext2_fsblk_t last_block;
634 if (i == sbi->s_groups_count - 1)
635 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
637 last_block = first_block +
638 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
640 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
641 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
643 ext2_error (sb, "ext2_check_descriptors",
644 "Block bitmap for group %d"
645 " not in group (block %lu)!",
646 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
649 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
650 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
652 ext2_error (sb, "ext2_check_descriptors",
653 "Inode bitmap for group %d"
654 " not in group (block %lu)!",
655 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
658 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
659 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
662 ext2_error (sb, "ext2_check_descriptors",
663 "Inode table for group %d"
664 " not in group (block %lu)!",
665 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
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_meta_bg;
727 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
729 if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
731 return (logic_sb_block + nr + 1);
732 bg = sbi->s_desc_per_block * nr;
733 if (ext2_bg_has_super(sb, bg))
736 return ext2_group_first_block_no(sb, bg) + has_super;
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_blockgroup_lock =
762 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
763 if (!sbi->s_blockgroup_lock) {
768 sbi->s_sb_block = sb_block;
771 * See what the current blocksize for the device is, and
772 * use that as the blocksize. Otherwise (or if the blocksize
773 * is smaller than the default) use the default.
774 * This is important for devices that have a hardware
775 * sectorsize that is larger than the default.
777 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
779 printk ("EXT2-fs: unable to set blocksize\n");
784 * If the superblock doesn't start on a hardware sector boundary,
785 * calculate the offset.
787 if (blocksize != BLOCK_SIZE) {
788 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
789 offset = (sb_block*BLOCK_SIZE) % blocksize;
791 logic_sb_block = sb_block;
794 if (!(bh = sb_bread(sb, logic_sb_block))) {
795 printk ("EXT2-fs: unable to read superblock\n");
799 * Note: s_es must be initialized as soon as possible because
800 * some ext2 macro-instructions depend on its value
802 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
804 sb->s_magic = le16_to_cpu(es->s_magic);
806 if (sb->s_magic != EXT2_SUPER_MAGIC)
809 /* Set defaults before we parse the mount options */
810 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
811 if (def_mount_opts & EXT2_DEFM_DEBUG)
812 set_opt(sbi->s_mount_opt, DEBUG);
813 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
814 set_opt(sbi->s_mount_opt, GRPID);
815 if (def_mount_opts & EXT2_DEFM_UID16)
816 set_opt(sbi->s_mount_opt, NO_UID32);
817 #ifdef CONFIG_EXT2_FS_XATTR
818 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
819 set_opt(sbi->s_mount_opt, XATTR_USER);
821 #ifdef CONFIG_EXT2_FS_POSIX_ACL
822 if (def_mount_opts & EXT2_DEFM_ACL)
823 set_opt(sbi->s_mount_opt, POSIX_ACL);
826 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
827 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
828 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
829 set_opt(sbi->s_mount_opt, ERRORS_CONT);
831 set_opt(sbi->s_mount_opt, ERRORS_RO);
833 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
834 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
836 set_opt(sbi->s_mount_opt, RESERVATION);
838 if (!parse_options ((char *) data, sbi))
841 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
842 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
845 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
846 EXT2_MOUNT_XIP if not */
848 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
849 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
850 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
851 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
852 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
853 "running e2fsck is recommended\n");
855 * Check feature flags regardless of the revision level, since we
856 * previously didn't change the revision level when setting the flags,
857 * so there is a chance incompat flags are set on a rev 0 filesystem.
859 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
861 printk("EXT2-fs: %s: couldn't mount because of "
862 "unsupported optional features (%x).\n",
863 sb->s_id, le32_to_cpu(features));
866 if (!(sb->s_flags & MS_RDONLY) &&
867 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
868 printk("EXT2-fs: %s: couldn't mount RDWR because of "
869 "unsupported optional features (%x).\n",
870 sb->s_id, le32_to_cpu(features));
874 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
876 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
878 printk("XIP: Unsupported blocksize\n");
882 /* If the blocksize doesn't match, re-read the thing.. */
883 if (sb->s_blocksize != blocksize) {
886 if (!sb_set_blocksize(sb, blocksize)) {
887 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
891 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
892 offset = (sb_block*BLOCK_SIZE) % blocksize;
893 bh = sb_bread(sb, logic_sb_block);
895 printk("EXT2-fs: Couldn't read superblock on "
899 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
901 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
902 printk ("EXT2-fs: Magic mismatch, very weird !\n");
907 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
909 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
910 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
911 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
913 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
914 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
915 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
916 !is_power_of_2(sbi->s_inode_size) ||
917 (sbi->s_inode_size > blocksize)) {
918 printk ("EXT2-fs: unsupported inode size: %d\n",
924 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
925 le32_to_cpu(es->s_log_frag_size);
926 if (sbi->s_frag_size == 0)
928 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
930 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
931 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
932 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
934 if (EXT2_INODE_SIZE(sb) == 0)
936 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
937 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
939 sbi->s_itb_per_group = sbi->s_inodes_per_group /
940 sbi->s_inodes_per_block;
941 sbi->s_desc_per_block = sb->s_blocksize /
942 sizeof (struct ext2_group_desc);
944 sbi->s_mount_state = le16_to_cpu(es->s_state);
945 sbi->s_addr_per_block_bits =
946 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
947 sbi->s_desc_per_block_bits =
948 ilog2 (EXT2_DESC_PER_BLOCK(sb));
950 if (sb->s_magic != EXT2_SUPER_MAGIC)
953 if (sb->s_blocksize != bh->b_size) {
955 printk ("VFS: Unsupported blocksize on dev "
960 if (sb->s_blocksize != sbi->s_frag_size) {
961 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
962 sbi->s_frag_size, sb->s_blocksize);
966 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
967 printk ("EXT2-fs: #blocks per group too big: %lu\n",
968 sbi->s_blocks_per_group);
971 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
972 printk ("EXT2-fs: #fragments per group too big: %lu\n",
973 sbi->s_frags_per_group);
976 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
977 printk ("EXT2-fs: #inodes per group too big: %lu\n",
978 sbi->s_inodes_per_group);
982 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
984 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
985 le32_to_cpu(es->s_first_data_block) - 1)
986 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
987 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
988 EXT2_DESC_PER_BLOCK(sb);
989 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
990 if (sbi->s_group_desc == NULL) {
991 printk ("EXT2-fs: not enough memory\n");
994 bgl_lock_init(sbi->s_blockgroup_lock);
995 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
997 printk ("EXT2-fs: not enough memory\n");
998 goto failed_mount_group_desc;
1000 for (i = 0; i < db_count; i++) {
1001 block = descriptor_loc(sb, logic_sb_block, i);
1002 sbi->s_group_desc[i] = sb_bread(sb, block);
1003 if (!sbi->s_group_desc[i]) {
1004 for (j = 0; j < i; j++)
1005 brelse (sbi->s_group_desc[j]);
1006 printk ("EXT2-fs: unable to read group descriptors\n");
1007 goto failed_mount_group_desc;
1010 if (!ext2_check_descriptors (sb)) {
1011 printk ("EXT2-fs: group descriptors corrupted!\n");
1014 sbi->s_gdb_count = db_count;
1015 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1016 spin_lock_init(&sbi->s_next_gen_lock);
1018 /* per fileystem reservation list head & lock */
1019 spin_lock_init(&sbi->s_rsv_window_lock);
1020 sbi->s_rsv_window_root = RB_ROOT;
1022 * Add a single, static dummy reservation to the start of the
1023 * reservation window list --- it gives us a placeholder for
1024 * append-at-start-of-list which makes the allocation logic
1027 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1028 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1029 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1030 sbi->s_rsv_window_head.rsv_goal_size = 0;
1031 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1033 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1034 ext2_count_free_blocks(sb));
1036 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1037 ext2_count_free_inodes(sb));
1040 err = percpu_counter_init(&sbi->s_dirs_counter,
1041 ext2_count_dirs(sb));
1044 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1048 * set up enough so that it can read an inode
1050 sb->s_op = &ext2_sops;
1051 sb->s_export_op = &ext2_export_ops;
1052 sb->s_xattr = ext2_xattr_handlers;
1053 root = ext2_iget(sb, EXT2_ROOT_INO);
1055 ret = PTR_ERR(root);
1058 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1060 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1064 sb->s_root = d_alloc_root(root);
1067 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1071 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1072 ext2_warning(sb, __func__,
1073 "mounting ext3 filesystem as ext2");
1074 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1079 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1083 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1084 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1085 percpu_counter_destroy(&sbi->s_dirs_counter);
1087 for (i = 0; i < db_count; i++)
1088 brelse(sbi->s_group_desc[i]);
1089 failed_mount_group_desc:
1090 kfree(sbi->s_group_desc);
1091 kfree(sbi->s_debts);
1095 sb->s_fs_info = NULL;
1100 static void ext2_commit_super (struct super_block * sb,
1101 struct ext2_super_block * es)
1103 es->s_wtime = cpu_to_le32(get_seconds());
1104 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1108 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1110 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1111 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1112 es->s_wtime = cpu_to_le32(get_seconds());
1113 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1114 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1119 * In the second extended file system, it is not necessary to
1120 * write the super block since we use a mapping of the
1121 * disk super block in a buffer.
1123 * However, this function is still used to set the fs valid
1124 * flags to 0. We need to set this flag to 0 since the fs
1125 * may have been checked while mounted and e2fsck may have
1126 * set s_state to EXT2_VALID_FS after some corrections.
1129 void ext2_write_super (struct super_block * sb)
1131 struct ext2_super_block * es;
1133 if (!(sb->s_flags & MS_RDONLY)) {
1134 es = EXT2_SB(sb)->s_es;
1136 if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1137 ext2_debug ("setting valid to 0\n");
1138 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1139 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1140 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1141 es->s_mtime = cpu_to_le32(get_seconds());
1142 ext2_sync_super(sb, es);
1144 ext2_commit_super (sb, es);
1150 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1152 struct ext2_sb_info * sbi = EXT2_SB(sb);
1153 struct ext2_super_block * es;
1154 unsigned long old_mount_opt = sbi->s_mount_opt;
1155 struct ext2_mount_options old_opts;
1156 unsigned long old_sb_flags;
1159 /* Store the old options */
1160 old_sb_flags = sb->s_flags;
1161 old_opts.s_mount_opt = sbi->s_mount_opt;
1162 old_opts.s_resuid = sbi->s_resuid;
1163 old_opts.s_resgid = sbi->s_resgid;
1166 * Allow the "check" option to be passed as a remount option.
1168 if (!parse_options (data, sbi)) {
1173 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1174 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1176 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1177 EXT2_MOUNT_XIP if not */
1179 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1180 printk("XIP: Unsupported blocksize\n");
1186 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1187 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1188 invalidate_inodes(sb))
1189 ext2_warning(sb, __func__, "busy inodes while remounting "\
1190 "xip remain in cache (no functional problem)");
1191 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1193 if (*flags & MS_RDONLY) {
1194 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1195 !(sbi->s_mount_state & EXT2_VALID_FS))
1198 * OK, we are remounting a valid rw partition rdonly, so set
1199 * the rdonly flag and then mark the partition as valid again.
1201 es->s_state = cpu_to_le16(sbi->s_mount_state);
1202 es->s_mtime = cpu_to_le32(get_seconds());
1204 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1205 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1207 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1208 "unsupported optional features (%x).\n",
1209 sb->s_id, le32_to_cpu(ret));
1214 * Mounting a RDONLY partition read-write, so reread and
1215 * store the current valid flag. (It may have been changed
1216 * by e2fsck since we originally mounted the partition.)
1218 sbi->s_mount_state = le16_to_cpu(es->s_state);
1219 if (!ext2_setup_super (sb, es, 0))
1220 sb->s_flags &= ~MS_RDONLY;
1222 ext2_sync_super(sb, es);
1225 sbi->s_mount_opt = old_opts.s_mount_opt;
1226 sbi->s_resuid = old_opts.s_resuid;
1227 sbi->s_resgid = old_opts.s_resgid;
1228 sb->s_flags = old_sb_flags;
1232 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1234 struct super_block *sb = dentry->d_sb;
1235 struct ext2_sb_info *sbi = EXT2_SB(sb);
1236 struct ext2_super_block *es = sbi->s_es;
1239 if (test_opt (sb, MINIX_DF))
1240 sbi->s_overhead_last = 0;
1241 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1242 unsigned long i, overhead = 0;
1246 * Compute the overhead (FS structures). This is constant
1247 * for a given filesystem unless the number of block groups
1248 * changes so we cache the previous value until it does.
1252 * All of the blocks before first_data_block are
1255 overhead = le32_to_cpu(es->s_first_data_block);
1258 * Add the overhead attributed to the superblock and
1259 * block group descriptors. If the sparse superblocks
1260 * feature is turned on, then not all groups have this.
1262 for (i = 0; i < sbi->s_groups_count; i++)
1263 overhead += ext2_bg_has_super(sb, i) +
1264 ext2_bg_num_gdb(sb, i);
1267 * Every block group has an inode bitmap, a block
1268 * bitmap, and an inode table.
1270 overhead += (sbi->s_groups_count *
1271 (2 + sbi->s_itb_per_group));
1272 sbi->s_overhead_last = overhead;
1274 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1277 buf->f_type = EXT2_SUPER_MAGIC;
1278 buf->f_bsize = sb->s_blocksize;
1279 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1280 buf->f_bfree = ext2_count_free_blocks(sb);
1281 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1282 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1283 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1285 buf->f_files = le32_to_cpu(es->s_inodes_count);
1286 buf->f_ffree = ext2_count_free_inodes(sb);
1287 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1288 buf->f_namelen = EXT2_NAME_LEN;
1289 fsid = le64_to_cpup((void *)es->s_uuid) ^
1290 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1291 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1292 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1296 static int ext2_get_sb(struct file_system_type *fs_type,
1297 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1299 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1304 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1305 * acquiring the locks... As quota files are never truncated and quota code
1306 * itself serializes the operations (and noone else should touch the files)
1307 * we don't have to be afraid of races */
1308 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1309 size_t len, loff_t off)
1311 struct inode *inode = sb_dqopt(sb)->files[type];
1312 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1314 int offset = off & (sb->s_blocksize - 1);
1317 struct buffer_head tmp_bh;
1318 struct buffer_head *bh;
1319 loff_t i_size = i_size_read(inode);
1323 if (off+len > i_size)
1326 while (toread > 0) {
1327 tocopy = sb->s_blocksize - offset < toread ?
1328 sb->s_blocksize - offset : toread;
1331 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1334 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1335 memset(data, 0, tocopy);
1337 bh = sb_bread(sb, tmp_bh.b_blocknr);
1340 memcpy(data, bh->b_data+offset, tocopy);
1351 /* Write to quotafile */
1352 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1353 const char *data, size_t len, loff_t off)
1355 struct inode *inode = sb_dqopt(sb)->files[type];
1356 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1358 int offset = off & (sb->s_blocksize - 1);
1360 size_t towrite = len;
1361 struct buffer_head tmp_bh;
1362 struct buffer_head *bh;
1364 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1365 while (towrite > 0) {
1366 tocopy = sb->s_blocksize - offset < towrite ?
1367 sb->s_blocksize - offset : towrite;
1370 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1373 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1374 bh = sb_bread(sb, tmp_bh.b_blocknr);
1376 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1382 memcpy(bh->b_data+offset, data, tocopy);
1383 flush_dcache_page(bh->b_page);
1384 set_buffer_uptodate(bh);
1385 mark_buffer_dirty(bh);
1396 if (inode->i_size < off+len-towrite)
1397 i_size_write(inode, off+len-towrite);
1399 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1400 mark_inode_dirty(inode);
1401 mutex_unlock(&inode->i_mutex);
1402 return len - towrite;
1407 static struct file_system_type ext2_fs_type = {
1408 .owner = THIS_MODULE,
1410 .get_sb = ext2_get_sb,
1411 .kill_sb = kill_block_super,
1412 .fs_flags = FS_REQUIRES_DEV,
1415 static int __init init_ext2_fs(void)
1417 int err = init_ext2_xattr();
1420 err = init_inodecache();
1423 err = register_filesystem(&ext2_fs_type);
1428 destroy_inodecache();
1434 static void __exit exit_ext2_fs(void)
1436 unregister_filesystem(&ext2_fs_type);
1437 destroy_inodecache();
1441 module_init(init_ext2_fs)
1442 module_exit(exit_ext2_fs)