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");
292 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
293 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
296 static const struct super_operations ext2_sops = {
297 .alloc_inode = ext2_alloc_inode,
298 .destroy_inode = ext2_destroy_inode,
299 .read_inode = ext2_read_inode,
300 .write_inode = ext2_write_inode,
301 .delete_inode = ext2_delete_inode,
302 .put_super = ext2_put_super,
303 .write_super = ext2_write_super,
304 .statfs = ext2_statfs,
305 .remount_fs = ext2_remount,
306 .clear_inode = ext2_clear_inode,
307 .show_options = ext2_show_options,
309 .quota_read = ext2_quota_read,
310 .quota_write = ext2_quota_write,
314 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
315 u64 ino, u32 generation)
319 if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
320 return ERR_PTR(-ESTALE);
321 if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
322 return ERR_PTR(-ESTALE);
324 /* iget isn't really right if the inode is currently unallocated!!
325 * ext2_read_inode currently does appropriate checks, but
326 * it might be "neater" to call ext2_get_inode first and check
327 * if the inode is valid.....
329 inode = iget(sb, ino);
331 return ERR_PTR(-ENOMEM);
332 if (is_bad_inode(inode) ||
333 (generation && inode->i_generation != generation)) {
334 /* we didn't find the right inode.. */
336 return ERR_PTR(-ESTALE);
341 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
342 int fh_len, int fh_type)
344 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
348 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
349 int fh_len, int fh_type)
351 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
355 /* Yes, most of these are left as NULL!!
356 * A NULL value implies the default, which works with ext2-like file
357 * systems, but can be improved upon.
358 * Currently only get_parent is required.
360 static const struct export_operations ext2_export_ops = {
361 .fh_to_dentry = ext2_fh_to_dentry,
362 .fh_to_parent = ext2_fh_to_parent,
363 .get_parent = ext2_get_parent,
366 static unsigned long get_sb_block(void **data)
368 unsigned long sb_block;
369 char *options = (char *) *data;
371 if (!options || strncmp(options, "sb=", 3) != 0)
372 return 1; /* Default location */
374 sb_block = simple_strtoul(options, &options, 0);
375 if (*options && *options != ',') {
376 printk("EXT2-fs: Invalid sb specification: %s\n",
382 *data = (void *) options;
387 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
388 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
389 Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
390 Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
391 Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
392 Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
395 static match_table_t tokens = {
396 {Opt_bsd_df, "bsddf"},
397 {Opt_minix_df, "minixdf"},
398 {Opt_grpid, "grpid"},
399 {Opt_grpid, "bsdgroups"},
400 {Opt_nogrpid, "nogrpid"},
401 {Opt_nogrpid, "sysvgroups"},
402 {Opt_resgid, "resgid=%u"},
403 {Opt_resuid, "resuid=%u"},
405 {Opt_err_cont, "errors=continue"},
406 {Opt_err_panic, "errors=panic"},
407 {Opt_err_ro, "errors=remount-ro"},
408 {Opt_nouid32, "nouid32"},
409 {Opt_nocheck, "check=none"},
410 {Opt_nocheck, "nocheck"},
411 {Opt_debug, "debug"},
412 {Opt_oldalloc, "oldalloc"},
413 {Opt_orlov, "orlov"},
415 {Opt_user_xattr, "user_xattr"},
416 {Opt_nouser_xattr, "nouser_xattr"},
418 {Opt_noacl, "noacl"},
420 {Opt_grpquota, "grpquota"},
421 {Opt_ignore, "noquota"},
422 {Opt_quota, "quota"},
423 {Opt_usrquota, "usrquota"},
424 {Opt_reservation, "reservation"},
425 {Opt_noreservation, "noreservation"},
429 static int parse_options (char * options,
430 struct ext2_sb_info *sbi)
433 substring_t args[MAX_OPT_ARGS];
439 while ((p = strsep (&options, ",")) != NULL) {
444 token = match_token(p, tokens, args);
447 clear_opt (sbi->s_mount_opt, MINIX_DF);
450 set_opt (sbi->s_mount_opt, MINIX_DF);
453 set_opt (sbi->s_mount_opt, GRPID);
456 clear_opt (sbi->s_mount_opt, GRPID);
459 if (match_int(&args[0], &option))
461 sbi->s_resuid = option;
464 if (match_int(&args[0], &option))
466 sbi->s_resgid = option;
469 /* handled by get_sb_block() instead of here */
470 /* *sb_block = match_int(&args[0]); */
473 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
474 clear_opt (sbi->s_mount_opt, ERRORS_RO);
475 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
478 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
479 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
480 set_opt (sbi->s_mount_opt, ERRORS_RO);
483 clear_opt (sbi->s_mount_opt, ERRORS_RO);
484 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
485 set_opt (sbi->s_mount_opt, ERRORS_CONT);
488 set_opt (sbi->s_mount_opt, NO_UID32);
491 clear_opt (sbi->s_mount_opt, CHECK);
494 set_opt (sbi->s_mount_opt, DEBUG);
497 set_opt (sbi->s_mount_opt, OLDALLOC);
500 clear_opt (sbi->s_mount_opt, OLDALLOC);
503 set_opt (sbi->s_mount_opt, NOBH);
505 #ifdef CONFIG_EXT2_FS_XATTR
507 set_opt (sbi->s_mount_opt, XATTR_USER);
509 case Opt_nouser_xattr:
510 clear_opt (sbi->s_mount_opt, XATTR_USER);
514 case Opt_nouser_xattr:
515 printk("EXT2 (no)user_xattr options not supported\n");
518 #ifdef CONFIG_EXT2_FS_POSIX_ACL
520 set_opt(sbi->s_mount_opt, POSIX_ACL);
523 clear_opt(sbi->s_mount_opt, POSIX_ACL);
528 printk("EXT2 (no)acl options not supported\n");
532 #ifdef CONFIG_EXT2_FS_XIP
533 set_opt (sbi->s_mount_opt, XIP);
535 printk("EXT2 xip option not supported\n");
539 #if defined(CONFIG_QUOTA)
542 set_opt(sbi->s_mount_opt, USRQUOTA);
546 set_opt(sbi->s_mount_opt, GRPQUOTA);
553 "EXT2-fs: quota operations not supported.\n");
558 case Opt_reservation:
559 set_opt(sbi->s_mount_opt, RESERVATION);
560 printk("reservations ON\n");
562 case Opt_noreservation:
563 clear_opt(sbi->s_mount_opt, RESERVATION);
564 printk("reservations OFF\n");
575 static int ext2_setup_super (struct super_block * sb,
576 struct ext2_super_block * es,
580 struct ext2_sb_info *sbi = EXT2_SB(sb);
582 if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
583 printk ("EXT2-fs warning: revision level too high, "
584 "forcing read-only mode\n");
589 if (!(sbi->s_mount_state & EXT2_VALID_FS))
590 printk ("EXT2-fs warning: mounting unchecked fs, "
591 "running e2fsck is recommended\n");
592 else if ((sbi->s_mount_state & EXT2_ERROR_FS))
593 printk ("EXT2-fs warning: mounting fs with errors, "
594 "running e2fsck is recommended\n");
595 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
596 le16_to_cpu(es->s_mnt_count) >=
597 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
598 printk ("EXT2-fs warning: maximal mount count reached, "
599 "running e2fsck is recommended\n");
600 else if (le32_to_cpu(es->s_checkinterval) &&
601 (le32_to_cpu(es->s_lastcheck) + le32_to_cpu(es->s_checkinterval) <= get_seconds()))
602 printk ("EXT2-fs warning: checktime reached, "
603 "running e2fsck is recommended\n");
604 if (!le16_to_cpu(es->s_max_mnt_count))
605 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
606 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
607 ext2_write_super(sb);
608 if (test_opt (sb, DEBUG))
609 printk ("[EXT II FS %s, %s, bs=%lu, fs=%lu, gc=%lu, "
610 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
611 EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
614 EXT2_BLOCKS_PER_GROUP(sb),
615 EXT2_INODES_PER_GROUP(sb),
620 static int ext2_check_descriptors(struct super_block *sb)
623 struct ext2_sb_info *sbi = EXT2_SB(sb);
624 unsigned long first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
625 unsigned long last_block;
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);
632 if (i == sbi->s_groups_count - 1)
633 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
635 last_block = first_block +
636 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
638 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
639 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
641 ext2_error (sb, "ext2_check_descriptors",
642 "Block bitmap for group %d"
643 " not in group (block %lu)!",
644 i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
647 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
648 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
650 ext2_error (sb, "ext2_check_descriptors",
651 "Inode bitmap for group %d"
652 " not in group (block %lu)!",
653 i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
656 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
657 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
660 ext2_error (sb, "ext2_check_descriptors",
661 "Inode table for group %d"
662 " not in group (block %lu)!",
663 i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
666 first_block += EXT2_BLOCKS_PER_GROUP(sb);
672 * Maximal file size. There is a direct, and {,double-,triple-}indirect
673 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
674 * We need to be 1 filesystem block less than the 2^32 sector limit.
676 static loff_t ext2_max_size(int bits)
678 loff_t res = EXT2_NDIR_BLOCKS;
682 /* This is calculated to be the largest file size for a
683 * dense, file such that the total number of
684 * sectors in the file, including data and all indirect blocks,
685 * does not exceed 2^32 -1
686 * __u32 i_blocks representing the total number of
687 * 512 bytes blocks of the file
689 upper_limit = (1LL << 32) - 1;
691 /* total blocks in file system block size */
692 upper_limit >>= (bits - 9);
695 /* indirect blocks */
697 /* double indirect blocks */
698 meta_blocks += 1 + (1LL << (bits-2));
699 /* tripple indirect blocks */
700 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
702 upper_limit -= meta_blocks;
703 upper_limit <<= bits;
705 res += 1LL << (bits-2);
706 res += 1LL << (2*(bits-2));
707 res += 1LL << (3*(bits-2));
709 if (res > upper_limit)
712 if (res > MAX_LFS_FILESIZE)
713 res = MAX_LFS_FILESIZE;
718 static unsigned long descriptor_loc(struct super_block *sb,
719 unsigned long logic_sb_block,
722 struct ext2_sb_info *sbi = EXT2_SB(sb);
723 unsigned long bg, first_data_block, first_meta_bg;
726 first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
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))
735 return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
738 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
740 struct buffer_head * bh;
741 struct ext2_sb_info * sbi;
742 struct ext2_super_block * es;
745 unsigned long sb_block = get_sb_block(&data);
746 unsigned long logic_sb_block;
747 unsigned long offset = 0;
748 unsigned long def_mount_opts;
749 int blocksize = BLOCK_SIZE;
755 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
759 sbi->s_sb_block = sb_block;
762 * See what the current blocksize for the device is, and
763 * use that as the blocksize. Otherwise (or if the blocksize
764 * is smaller than the default) use the default.
765 * This is important for devices that have a hardware
766 * sectorsize that is larger than the default.
768 blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
770 printk ("EXT2-fs: unable to set blocksize\n");
775 * If the superblock doesn't start on a hardware sector boundary,
776 * calculate the offset.
778 if (blocksize != BLOCK_SIZE) {
779 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
780 offset = (sb_block*BLOCK_SIZE) % blocksize;
782 logic_sb_block = sb_block;
785 if (!(bh = sb_bread(sb, logic_sb_block))) {
786 printk ("EXT2-fs: unable to read superblock\n");
790 * Note: s_es must be initialized as soon as possible because
791 * some ext2 macro-instructions depend on its value
793 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
795 sb->s_magic = le16_to_cpu(es->s_magic);
797 if (sb->s_magic != EXT2_SUPER_MAGIC)
800 /* Set defaults before we parse the mount options */
801 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
802 if (def_mount_opts & EXT2_DEFM_DEBUG)
803 set_opt(sbi->s_mount_opt, DEBUG);
804 if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
805 set_opt(sbi->s_mount_opt, GRPID);
806 if (def_mount_opts & EXT2_DEFM_UID16)
807 set_opt(sbi->s_mount_opt, NO_UID32);
808 #ifdef CONFIG_EXT2_FS_XATTR
809 if (def_mount_opts & EXT2_DEFM_XATTR_USER)
810 set_opt(sbi->s_mount_opt, XATTR_USER);
812 #ifdef CONFIG_EXT2_FS_POSIX_ACL
813 if (def_mount_opts & EXT2_DEFM_ACL)
814 set_opt(sbi->s_mount_opt, POSIX_ACL);
817 if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
818 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
819 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
820 set_opt(sbi->s_mount_opt, ERRORS_CONT);
822 set_opt(sbi->s_mount_opt, ERRORS_RO);
824 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
825 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
827 set_opt(sbi->s_mount_opt, RESERVATION);
829 if (!parse_options ((char *) data, sbi))
832 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
833 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
836 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
837 EXT2_MOUNT_XIP if not */
839 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
840 (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
841 EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
842 EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
843 printk("EXT2-fs warning: feature flags set on rev 0 fs, "
844 "running e2fsck is recommended\n");
846 * Check feature flags regardless of the revision level, since we
847 * previously didn't change the revision level when setting the flags,
848 * so there is a chance incompat flags are set on a rev 0 filesystem.
850 features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
852 printk("EXT2-fs: %s: couldn't mount because of "
853 "unsupported optional features (%x).\n",
854 sb->s_id, le32_to_cpu(features));
857 if (!(sb->s_flags & MS_RDONLY) &&
858 (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
859 printk("EXT2-fs: %s: couldn't mount RDWR because of "
860 "unsupported optional features (%x).\n",
861 sb->s_id, le32_to_cpu(features));
865 blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
867 if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
869 printk("XIP: Unsupported blocksize\n");
873 /* If the blocksize doesn't match, re-read the thing.. */
874 if (sb->s_blocksize != blocksize) {
877 if (!sb_set_blocksize(sb, blocksize)) {
878 printk(KERN_ERR "EXT2-fs: blocksize too small for device.\n");
882 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
883 offset = (sb_block*BLOCK_SIZE) % blocksize;
884 bh = sb_bread(sb, logic_sb_block);
886 printk("EXT2-fs: Couldn't read superblock on "
890 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
892 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
893 printk ("EXT2-fs: Magic mismatch, very weird !\n");
898 sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
900 if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
901 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
902 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
904 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
905 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
906 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
907 !is_power_of_2(sbi->s_inode_size) ||
908 (sbi->s_inode_size > blocksize)) {
909 printk ("EXT2-fs: unsupported inode size: %d\n",
915 sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
916 le32_to_cpu(es->s_log_frag_size);
917 if (sbi->s_frag_size == 0)
919 sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
921 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
922 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
923 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
925 if (EXT2_INODE_SIZE(sb) == 0)
927 sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
928 if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
930 sbi->s_itb_per_group = sbi->s_inodes_per_group /
931 sbi->s_inodes_per_block;
932 sbi->s_desc_per_block = sb->s_blocksize /
933 sizeof (struct ext2_group_desc);
935 sbi->s_mount_state = le16_to_cpu(es->s_state);
936 sbi->s_addr_per_block_bits =
937 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
938 sbi->s_desc_per_block_bits =
939 ilog2 (EXT2_DESC_PER_BLOCK(sb));
941 if (sb->s_magic != EXT2_SUPER_MAGIC)
944 if (sb->s_blocksize != bh->b_size) {
946 printk ("VFS: Unsupported blocksize on dev "
951 if (sb->s_blocksize != sbi->s_frag_size) {
952 printk ("EXT2-fs: fragsize %lu != blocksize %lu (not supported yet)\n",
953 sbi->s_frag_size, sb->s_blocksize);
957 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
958 printk ("EXT2-fs: #blocks per group too big: %lu\n",
959 sbi->s_blocks_per_group);
962 if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
963 printk ("EXT2-fs: #fragments per group too big: %lu\n",
964 sbi->s_frags_per_group);
967 if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
968 printk ("EXT2-fs: #inodes per group too big: %lu\n",
969 sbi->s_inodes_per_group);
973 if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
975 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
976 le32_to_cpu(es->s_first_data_block) - 1)
977 / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
978 db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
979 EXT2_DESC_PER_BLOCK(sb);
980 sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
981 if (sbi->s_group_desc == NULL) {
982 printk ("EXT2-fs: not enough memory\n");
985 bgl_lock_init(&sbi->s_blockgroup_lock);
986 sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
988 printk ("EXT2-fs: not enough memory\n");
989 goto failed_mount_group_desc;
991 for (i = 0; i < db_count; i++) {
992 block = descriptor_loc(sb, logic_sb_block, i);
993 sbi->s_group_desc[i] = sb_bread(sb, block);
994 if (!sbi->s_group_desc[i]) {
995 for (j = 0; j < i; j++)
996 brelse (sbi->s_group_desc[j]);
997 printk ("EXT2-fs: unable to read group descriptors\n");
998 goto failed_mount_group_desc;
1001 if (!ext2_check_descriptors (sb)) {
1002 printk ("EXT2-fs: group descriptors corrupted!\n");
1005 sbi->s_gdb_count = db_count;
1006 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1007 spin_lock_init(&sbi->s_next_gen_lock);
1009 /* per fileystem reservation list head & lock */
1010 spin_lock_init(&sbi->s_rsv_window_lock);
1011 sbi->s_rsv_window_root = RB_ROOT;
1013 * Add a single, static dummy reservation to the start of the
1014 * reservation window list --- it gives us a placeholder for
1015 * append-at-start-of-list which makes the allocation logic
1018 sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1019 sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1020 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1021 sbi->s_rsv_window_head.rsv_goal_size = 0;
1022 ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1024 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1025 ext2_count_free_blocks(sb));
1027 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1028 ext2_count_free_inodes(sb));
1031 err = percpu_counter_init(&sbi->s_dirs_counter,
1032 ext2_count_dirs(sb));
1035 printk(KERN_ERR "EXT2-fs: insufficient memory\n");
1039 * set up enough so that it can read an inode
1041 sb->s_op = &ext2_sops;
1042 sb->s_export_op = &ext2_export_ops;
1043 sb->s_xattr = ext2_xattr_handlers;
1044 root = iget(sb, EXT2_ROOT_INO);
1045 sb->s_root = d_alloc_root(root);
1048 printk(KERN_ERR "EXT2-fs: get root inode failed\n");
1051 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1054 printk(KERN_ERR "EXT2-fs: corrupt root inode, run e2fsck\n");
1057 if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1058 ext2_warning(sb, __FUNCTION__,
1059 "mounting ext3 filesystem as ext2");
1060 ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1065 printk("VFS: Can't find an ext2 filesystem on dev %s.\n",
1069 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1070 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1071 percpu_counter_destroy(&sbi->s_dirs_counter);
1073 for (i = 0; i < db_count; i++)
1074 brelse(sbi->s_group_desc[i]);
1075 failed_mount_group_desc:
1076 kfree(sbi->s_group_desc);
1077 kfree(sbi->s_debts);
1081 sb->s_fs_info = NULL;
1086 static void ext2_commit_super (struct super_block * sb,
1087 struct ext2_super_block * es)
1089 es->s_wtime = cpu_to_le32(get_seconds());
1090 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1094 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es)
1096 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1097 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1098 es->s_wtime = cpu_to_le32(get_seconds());
1099 mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1100 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1105 * In the second extended file system, it is not necessary to
1106 * write the super block since we use a mapping of the
1107 * disk super block in a buffer.
1109 * However, this function is still used to set the fs valid
1110 * flags to 0. We need to set this flag to 0 since the fs
1111 * may have been checked while mounted and e2fsck may have
1112 * set s_state to EXT2_VALID_FS after some corrections.
1115 void ext2_write_super (struct super_block * sb)
1117 struct ext2_super_block * es;
1119 if (!(sb->s_flags & MS_RDONLY)) {
1120 es = EXT2_SB(sb)->s_es;
1122 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS) {
1123 ext2_debug ("setting valid to 0\n");
1124 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) &
1126 es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1127 es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1128 es->s_mtime = cpu_to_le32(get_seconds());
1129 ext2_sync_super(sb, es);
1131 ext2_commit_super (sb, es);
1137 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1139 struct ext2_sb_info * sbi = EXT2_SB(sb);
1140 struct ext2_super_block * es;
1141 unsigned long old_mount_opt = sbi->s_mount_opt;
1142 struct ext2_mount_options old_opts;
1143 unsigned long old_sb_flags;
1146 /* Store the old options */
1147 old_sb_flags = sb->s_flags;
1148 old_opts.s_mount_opt = sbi->s_mount_opt;
1149 old_opts.s_resuid = sbi->s_resuid;
1150 old_opts.s_resgid = sbi->s_resgid;
1153 * Allow the "check" option to be passed as a remount option.
1155 if (!parse_options (data, sbi)) {
1160 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1161 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1163 ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1164 EXT2_MOUNT_XIP if not */
1166 if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1167 printk("XIP: Unsupported blocksize\n");
1173 if (((sbi->s_mount_opt & EXT2_MOUNT_XIP) !=
1174 (old_mount_opt & EXT2_MOUNT_XIP)) &&
1175 invalidate_inodes(sb))
1176 ext2_warning(sb, __FUNCTION__, "busy inodes while remounting "\
1177 "xip remain in cache (no functional problem)");
1178 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1180 if (*flags & MS_RDONLY) {
1181 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1182 !(sbi->s_mount_state & EXT2_VALID_FS))
1185 * OK, we are remounting a valid rw partition rdonly, so set
1186 * the rdonly flag and then mark the partition as valid again.
1188 es->s_state = cpu_to_le16(sbi->s_mount_state);
1189 es->s_mtime = cpu_to_le32(get_seconds());
1191 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1192 ~EXT2_FEATURE_RO_COMPAT_SUPP);
1194 printk("EXT2-fs: %s: couldn't remount RDWR because of "
1195 "unsupported optional features (%x).\n",
1196 sb->s_id, le32_to_cpu(ret));
1201 * Mounting a RDONLY partition read-write, so reread and
1202 * store the current valid flag. (It may have been changed
1203 * by e2fsck since we originally mounted the partition.)
1205 sbi->s_mount_state = le16_to_cpu(es->s_state);
1206 if (!ext2_setup_super (sb, es, 0))
1207 sb->s_flags &= ~MS_RDONLY;
1209 ext2_sync_super(sb, es);
1212 sbi->s_mount_opt = old_opts.s_mount_opt;
1213 sbi->s_resuid = old_opts.s_resuid;
1214 sbi->s_resgid = old_opts.s_resgid;
1215 sb->s_flags = old_sb_flags;
1219 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1221 struct super_block *sb = dentry->d_sb;
1222 struct ext2_sb_info *sbi = EXT2_SB(sb);
1223 struct ext2_super_block *es = sbi->s_es;
1226 if (test_opt (sb, MINIX_DF))
1227 sbi->s_overhead_last = 0;
1228 else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1229 unsigned long i, overhead = 0;
1233 * Compute the overhead (FS structures). This is constant
1234 * for a given filesystem unless the number of block groups
1235 * changes so we cache the previous value until it does.
1239 * All of the blocks before first_data_block are
1242 overhead = le32_to_cpu(es->s_first_data_block);
1245 * Add the overhead attributed to the superblock and
1246 * block group descriptors. If the sparse superblocks
1247 * feature is turned on, then not all groups have this.
1249 for (i = 0; i < sbi->s_groups_count; i++)
1250 overhead += ext2_bg_has_super(sb, i) +
1251 ext2_bg_num_gdb(sb, i);
1254 * Every block group has an inode bitmap, a block
1255 * bitmap, and an inode table.
1257 overhead += (sbi->s_groups_count *
1258 (2 + sbi->s_itb_per_group));
1259 sbi->s_overhead_last = overhead;
1261 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1264 buf->f_type = EXT2_SUPER_MAGIC;
1265 buf->f_bsize = sb->s_blocksize;
1266 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1267 buf->f_bfree = ext2_count_free_blocks(sb);
1268 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1269 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1270 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1272 buf->f_files = le32_to_cpu(es->s_inodes_count);
1273 buf->f_ffree = ext2_count_free_inodes(sb);
1274 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1275 buf->f_namelen = EXT2_NAME_LEN;
1276 fsid = le64_to_cpup((void *)es->s_uuid) ^
1277 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1278 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1279 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1283 static int ext2_get_sb(struct file_system_type *fs_type,
1284 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1286 return get_sb_bdev(fs_type, flags, dev_name, data, ext2_fill_super, mnt);
1291 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1292 * acquiring the locks... As quota files are never truncated and quota code
1293 * itself serializes the operations (and noone else should touch the files)
1294 * we don't have to be afraid of races */
1295 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1296 size_t len, loff_t off)
1298 struct inode *inode = sb_dqopt(sb)->files[type];
1299 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1301 int offset = off & (sb->s_blocksize - 1);
1304 struct buffer_head tmp_bh;
1305 struct buffer_head *bh;
1306 loff_t i_size = i_size_read(inode);
1310 if (off+len > i_size)
1313 while (toread > 0) {
1314 tocopy = sb->s_blocksize - offset < toread ?
1315 sb->s_blocksize - offset : toread;
1318 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1321 if (!buffer_mapped(&tmp_bh)) /* A hole? */
1322 memset(data, 0, tocopy);
1324 bh = sb_bread(sb, tmp_bh.b_blocknr);
1327 memcpy(data, bh->b_data+offset, tocopy);
1338 /* Write to quotafile */
1339 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1340 const char *data, size_t len, loff_t off)
1342 struct inode *inode = sb_dqopt(sb)->files[type];
1343 sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1345 int offset = off & (sb->s_blocksize - 1);
1347 size_t towrite = len;
1348 struct buffer_head tmp_bh;
1349 struct buffer_head *bh;
1351 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
1352 while (towrite > 0) {
1353 tocopy = sb->s_blocksize - offset < towrite ?
1354 sb->s_blocksize - offset : towrite;
1357 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1360 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1361 bh = sb_bread(sb, tmp_bh.b_blocknr);
1363 bh = sb_getblk(sb, tmp_bh.b_blocknr);
1369 memcpy(bh->b_data+offset, data, tocopy);
1370 flush_dcache_page(bh->b_page);
1371 set_buffer_uptodate(bh);
1372 mark_buffer_dirty(bh);
1383 if (inode->i_size < off+len-towrite)
1384 i_size_write(inode, off+len-towrite);
1386 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1387 mark_inode_dirty(inode);
1388 mutex_unlock(&inode->i_mutex);
1389 return len - towrite;
1394 static struct file_system_type ext2_fs_type = {
1395 .owner = THIS_MODULE,
1397 .get_sb = ext2_get_sb,
1398 .kill_sb = kill_block_super,
1399 .fs_flags = FS_REQUIRES_DEV,
1402 static int __init init_ext2_fs(void)
1404 int err = init_ext2_xattr();
1407 err = init_inodecache();
1410 err = register_filesystem(&ext2_fs_type);
1415 destroy_inodecache();
1421 static void __exit exit_ext2_fs(void)
1423 unregister_filesystem(&ext2_fs_type);
1424 destroy_inodecache();
1428 module_init(init_ext2_fs)
1429 module_exit(exit_ext2_fs)