aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ext4/super.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ext4/super.c')
-rw-r--r--fs/ext4/super.c263
1 files changed, 175 insertions, 88 deletions
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 44d0c8db2239..9953d80145ad 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -45,6 +45,7 @@
45#include <linux/freezer.h> 45#include <linux/freezer.h>
46 46
47#include "ext4.h" 47#include "ext4.h"
48#include "ext4_extents.h"
48#include "ext4_jbd2.h" 49#include "ext4_jbd2.h"
49#include "xattr.h" 50#include "xattr.h"
50#include "acl.h" 51#include "acl.h"
@@ -163,8 +164,8 @@ ext4_fsblk_t ext4_inode_table(struct super_block *sb,
163 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
164} 165}
165 166
166__u32 ext4_free_blks_count(struct super_block *sb, 167__u32 ext4_free_group_clusters(struct super_block *sb,
167 struct ext4_group_desc *bg) 168 struct ext4_group_desc *bg)
168{ 169{
169 return le16_to_cpu(bg->bg_free_blocks_count_lo) | 170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
170 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
@@ -219,8 +220,8 @@ void ext4_inode_table_set(struct super_block *sb,
219 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
220} 221}
221 222
222void ext4_free_blks_set(struct super_block *sb, 223void ext4_free_group_clusters_set(struct super_block *sb,
223 struct ext4_group_desc *bg, __u32 count) 224 struct ext4_group_desc *bg, __u32 count)
224{ 225{
225 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
226 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
@@ -414,6 +415,22 @@ static void save_error_info(struct super_block *sb, const char *func,
414 ext4_commit_super(sb, 1); 415 ext4_commit_super(sb, 1);
415} 416}
416 417
418/*
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
425 */
426static int block_device_ejected(struct super_block *sb)
427{
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
430
431 return bdi->dev == NULL;
432}
433
417 434
418/* Deal with the reporting of failure conditions on a filesystem such as 435/* Deal with the reporting of failure conditions on a filesystem such as
419 * inconsistencies detected or read IO failures. 436 * inconsistencies detected or read IO failures.
@@ -821,10 +838,10 @@ static void ext4_put_super(struct super_block *sb)
821 brelse(sbi->s_group_desc[i]); 838 brelse(sbi->s_group_desc[i]);
822 ext4_kvfree(sbi->s_group_desc); 839 ext4_kvfree(sbi->s_group_desc);
823 ext4_kvfree(sbi->s_flex_groups); 840 ext4_kvfree(sbi->s_flex_groups);
824 percpu_counter_destroy(&sbi->s_freeblocks_counter); 841 percpu_counter_destroy(&sbi->s_freeclusters_counter);
825 percpu_counter_destroy(&sbi->s_freeinodes_counter); 842 percpu_counter_destroy(&sbi->s_freeinodes_counter);
826 percpu_counter_destroy(&sbi->s_dirs_counter); 843 percpu_counter_destroy(&sbi->s_dirs_counter);
827 percpu_counter_destroy(&sbi->s_dirtyblocks_counter); 844 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
828 brelse(sbi->s_sbh); 845 brelse(sbi->s_sbh);
829#ifdef CONFIG_QUOTA 846#ifdef CONFIG_QUOTA
830 for (i = 0; i < MAXQUOTAS; i++) 847 for (i = 0; i < MAXQUOTAS; i++)
@@ -1057,8 +1074,6 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1057 seq_puts(seq, ",nouid32"); 1074 seq_puts(seq, ",nouid32");
1058 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG)) 1075 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1059 seq_puts(seq, ",debug"); 1076 seq_puts(seq, ",debug");
1060 if (test_opt(sb, OLDALLOC))
1061 seq_puts(seq, ",oldalloc");
1062#ifdef CONFIG_EXT4_FS_XATTR 1077#ifdef CONFIG_EXT4_FS_XATTR
1063 if (test_opt(sb, XATTR_USER)) 1078 if (test_opt(sb, XATTR_USER))
1064 seq_puts(seq, ",user_xattr"); 1079 seq_puts(seq, ",user_xattr");
@@ -1567,10 +1582,12 @@ static int parse_options(char *options, struct super_block *sb,
1567 set_opt(sb, DEBUG); 1582 set_opt(sb, DEBUG);
1568 break; 1583 break;
1569 case Opt_oldalloc: 1584 case Opt_oldalloc:
1570 set_opt(sb, OLDALLOC); 1585 ext4_msg(sb, KERN_WARNING,
1586 "Ignoring deprecated oldalloc option");
1571 break; 1587 break;
1572 case Opt_orlov: 1588 case Opt_orlov:
1573 clear_opt(sb, OLDALLOC); 1589 ext4_msg(sb, KERN_WARNING,
1590 "Ignoring deprecated orlov option");
1574 break; 1591 break;
1575#ifdef CONFIG_EXT4_FS_XATTR 1592#ifdef CONFIG_EXT4_FS_XATTR
1576 case Opt_user_xattr: 1593 case Opt_user_xattr:
@@ -1801,6 +1818,7 @@ set_qf_format:
1801 break; 1818 break;
1802 case Opt_nodelalloc: 1819 case Opt_nodelalloc:
1803 clear_opt(sb, DELALLOC); 1820 clear_opt(sb, DELALLOC);
1821 clear_opt2(sb, EXPLICIT_DELALLOC);
1804 break; 1822 break;
1805 case Opt_mblk_io_submit: 1823 case Opt_mblk_io_submit:
1806 set_opt(sb, MBLK_IO_SUBMIT); 1824 set_opt(sb, MBLK_IO_SUBMIT);
@@ -1817,6 +1835,7 @@ set_qf_format:
1817 break; 1835 break;
1818 case Opt_delalloc: 1836 case Opt_delalloc:
1819 set_opt(sb, DELALLOC); 1837 set_opt(sb, DELALLOC);
1838 set_opt2(sb, EXPLICIT_DELALLOC);
1820 break; 1839 break;
1821 case Opt_block_validity: 1840 case Opt_block_validity:
1822 set_opt(sb, BLOCK_VALIDITY); 1841 set_opt(sb, BLOCK_VALIDITY);
@@ -1935,7 +1954,7 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1935 res = MS_RDONLY; 1954 res = MS_RDONLY;
1936 } 1955 }
1937 if (read_only) 1956 if (read_only)
1938 return res; 1957 goto done;
1939 if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1958 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1940 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1941 "running e2fsck is recommended"); 1960 "running e2fsck is recommended");
@@ -1966,6 +1985,7 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1966 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1985 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1967 1986
1968 ext4_commit_super(sb, 1); 1987 ext4_commit_super(sb, 1);
1988done:
1969 if (test_opt(sb, DEBUG)) 1989 if (test_opt(sb, DEBUG))
1970 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1971 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
@@ -2015,8 +2035,8 @@ static int ext4_fill_flex_info(struct super_block *sb)
2015 flex_group = ext4_flex_group(sbi, i); 2035 flex_group = ext4_flex_group(sbi, i);
2016 atomic_add(ext4_free_inodes_count(sb, gdp), 2036 atomic_add(ext4_free_inodes_count(sb, gdp),
2017 &sbi->s_flex_groups[flex_group].free_inodes); 2037 &sbi->s_flex_groups[flex_group].free_inodes);
2018 atomic_add(ext4_free_blks_count(sb, gdp), 2038 atomic_add(ext4_free_group_clusters(sb, gdp),
2019 &sbi->s_flex_groups[flex_group].free_blocks); 2039 &sbi->s_flex_groups[flex_group].free_clusters);
2020 atomic_add(ext4_used_dirs_count(sb, gdp), 2040 atomic_add(ext4_used_dirs_count(sb, gdp),
2021 &sbi->s_flex_groups[flex_group].used_dirs); 2041 &sbi->s_flex_groups[flex_group].used_dirs);
2022 } 2042 }
@@ -2134,7 +2154,8 @@ static int ext4_check_descriptors(struct super_block *sb,
2134 if (NULL != first_not_zeroed) 2154 if (NULL != first_not_zeroed)
2135 *first_not_zeroed = grp; 2155 *first_not_zeroed = grp;
2136 2156
2137 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); 2157 ext4_free_blocks_count_set(sbi->s_es,
2158 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2138 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2159 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2139 return 1; 2160 return 1;
2140} 2161}
@@ -2454,7 +2475,8 @@ static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2454 char *buf) 2475 char *buf)
2455{ 2476{
2456 return snprintf(buf, PAGE_SIZE, "%llu\n", 2477 return snprintf(buf, PAGE_SIZE, "%llu\n",
2457 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter)); 2478 (s64) EXT4_C2B(sbi,
2479 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2458} 2480}
2459 2481
2460static ssize_t session_write_kbytes_show(struct ext4_attr *a, 2482static ssize_t session_write_kbytes_show(struct ext4_attr *a,
@@ -2682,6 +2704,13 @@ static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2682 return 0; 2704 return 0;
2683 } 2705 }
2684 } 2706 }
2707 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2708 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2709 ext4_msg(sb, KERN_ERR,
2710 "Can't support bigalloc feature without "
2711 "extents feature\n");
2712 return 0;
2713 }
2685 return 1; 2714 return 1;
2686} 2715}
2687 2716
@@ -3087,10 +3116,10 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3087 char *cp; 3116 char *cp;
3088 const char *descr; 3117 const char *descr;
3089 int ret = -ENOMEM; 3118 int ret = -ENOMEM;
3090 int blocksize; 3119 int blocksize, clustersize;
3091 unsigned int db_count; 3120 unsigned int db_count;
3092 unsigned int i; 3121 unsigned int i;
3093 int needs_recovery, has_huge_files; 3122 int needs_recovery, has_huge_files, has_bigalloc;
3094 __u64 blocks_count; 3123 __u64 blocks_count;
3095 int err; 3124 int err;
3096 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3125 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
@@ -3224,6 +3253,33 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3224 &journal_ioprio, NULL, 0)) 3253 &journal_ioprio, NULL, 0))
3225 goto failed_mount; 3254 goto failed_mount;
3226 3255
3256 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3257 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3258 "with data=journal disables delayed "
3259 "allocation and O_DIRECT support!\n");
3260 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3261 ext4_msg(sb, KERN_ERR, "can't mount with "
3262 "both data=journal and delalloc");
3263 goto failed_mount;
3264 }
3265 if (test_opt(sb, DIOREAD_NOLOCK)) {
3266 ext4_msg(sb, KERN_ERR, "can't mount with "
3267 "both data=journal and delalloc");
3268 goto failed_mount;
3269 }
3270 if (test_opt(sb, DELALLOC))
3271 clear_opt(sb, DELALLOC);
3272 }
3273
3274 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3275 if (test_opt(sb, DIOREAD_NOLOCK)) {
3276 if (blocksize < PAGE_SIZE) {
3277 ext4_msg(sb, KERN_ERR, "can't mount with "
3278 "dioread_nolock if block size != PAGE_SIZE");
3279 goto failed_mount;
3280 }
3281 }
3282
3227 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3283 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3228 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3284 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3229 3285
@@ -3265,8 +3321,6 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3265 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3321 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3266 goto failed_mount; 3322 goto failed_mount;
3267 3323
3268 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3269
3270 if (blocksize < EXT4_MIN_BLOCK_SIZE || 3324 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3271 blocksize > EXT4_MAX_BLOCK_SIZE) { 3325 blocksize > EXT4_MAX_BLOCK_SIZE) {
3272 ext4_msg(sb, KERN_ERR, 3326 ext4_msg(sb, KERN_ERR,
@@ -3369,12 +3423,53 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3369 sb->s_dirt = 1; 3423 sb->s_dirt = 1;
3370 } 3424 }
3371 3425
3372 if (sbi->s_blocks_per_group > blocksize * 8) { 3426 /* Handle clustersize */
3373 ext4_msg(sb, KERN_ERR, 3427 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3374 "#blocks per group too big: %lu", 3428 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3375 sbi->s_blocks_per_group); 3429 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3376 goto failed_mount; 3430 if (has_bigalloc) {
3431 if (clustersize < blocksize) {
3432 ext4_msg(sb, KERN_ERR,
3433 "cluster size (%d) smaller than "
3434 "block size (%d)", clustersize, blocksize);
3435 goto failed_mount;
3436 }
3437 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3438 le32_to_cpu(es->s_log_block_size);
3439 sbi->s_clusters_per_group =
3440 le32_to_cpu(es->s_clusters_per_group);
3441 if (sbi->s_clusters_per_group > blocksize * 8) {
3442 ext4_msg(sb, KERN_ERR,
3443 "#clusters per group too big: %lu",
3444 sbi->s_clusters_per_group);
3445 goto failed_mount;
3446 }
3447 if (sbi->s_blocks_per_group !=
3448 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3449 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3450 "clusters per group (%lu) inconsistent",
3451 sbi->s_blocks_per_group,
3452 sbi->s_clusters_per_group);
3453 goto failed_mount;
3454 }
3455 } else {
3456 if (clustersize != blocksize) {
3457 ext4_warning(sb, "fragment/cluster size (%d) != "
3458 "block size (%d)", clustersize,
3459 blocksize);
3460 clustersize = blocksize;
3461 }
3462 if (sbi->s_blocks_per_group > blocksize * 8) {
3463 ext4_msg(sb, KERN_ERR,
3464 "#blocks per group too big: %lu",
3465 sbi->s_blocks_per_group);
3466 goto failed_mount;
3467 }
3468 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3469 sbi->s_cluster_bits = 0;
3377 } 3470 }
3471 sbi->s_cluster_ratio = clustersize / blocksize;
3472
3378 if (sbi->s_inodes_per_group > blocksize * 8) { 3473 if (sbi->s_inodes_per_group > blocksize * 8) {
3379 ext4_msg(sb, KERN_ERR, 3474 ext4_msg(sb, KERN_ERR,
3380 "#inodes per group too big: %lu", 3475 "#inodes per group too big: %lu",
@@ -3446,10 +3541,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3446 goto failed_mount; 3541 goto failed_mount;
3447 } 3542 }
3448 3543
3449#ifdef CONFIG_PROC_FS
3450 if (ext4_proc_root) 3544 if (ext4_proc_root)
3451 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); 3545 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3452#endif
3453 3546
3454 bgl_lock_init(sbi->s_blockgroup_lock); 3547 bgl_lock_init(sbi->s_blockgroup_lock);
3455 3548
@@ -3483,8 +3576,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3483 sbi->s_err_report.function = print_daily_error_info; 3576 sbi->s_err_report.function = print_daily_error_info;
3484 sbi->s_err_report.data = (unsigned long) sb; 3577 sbi->s_err_report.data = (unsigned long) sb;
3485 3578
3486 err = percpu_counter_init(&sbi->s_freeblocks_counter, 3579 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3487 ext4_count_free_blocks(sb)); 3580 ext4_count_free_clusters(sb));
3488 if (!err) { 3581 if (!err) {
3489 err = percpu_counter_init(&sbi->s_freeinodes_counter, 3582 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3490 ext4_count_free_inodes(sb)); 3583 ext4_count_free_inodes(sb));
@@ -3494,7 +3587,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3494 ext4_count_dirs(sb)); 3587 ext4_count_dirs(sb));
3495 } 3588 }
3496 if (!err) { 3589 if (!err) {
3497 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0); 3590 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3498 } 3591 }
3499 if (err) { 3592 if (err) {
3500 ext4_msg(sb, KERN_ERR, "insufficient memory"); 3593 ext4_msg(sb, KERN_ERR, "insufficient memory");
@@ -3609,13 +3702,13 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3609 * The journal may have updated the bg summary counts, so we 3702 * The journal may have updated the bg summary counts, so we
3610 * need to update the global counters. 3703 * need to update the global counters.
3611 */ 3704 */
3612 percpu_counter_set(&sbi->s_freeblocks_counter, 3705 percpu_counter_set(&sbi->s_freeclusters_counter,
3613 ext4_count_free_blocks(sb)); 3706 ext4_count_free_clusters(sb));
3614 percpu_counter_set(&sbi->s_freeinodes_counter, 3707 percpu_counter_set(&sbi->s_freeinodes_counter,
3615 ext4_count_free_inodes(sb)); 3708 ext4_count_free_inodes(sb));
3616 percpu_counter_set(&sbi->s_dirs_counter, 3709 percpu_counter_set(&sbi->s_dirs_counter,
3617 ext4_count_dirs(sb)); 3710 ext4_count_dirs(sb));
3618 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0); 3711 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3619 3712
3620no_journal: 3713no_journal:
3621 /* 3714 /*
@@ -3679,25 +3772,6 @@ no_journal:
3679 "available"); 3772 "available");
3680 } 3773 }
3681 3774
3682 if (test_opt(sb, DELALLOC) &&
3683 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3684 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3685 "requested data journaling mode");
3686 clear_opt(sb, DELALLOC);
3687 }
3688 if (test_opt(sb, DIOREAD_NOLOCK)) {
3689 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3690 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3691 "option - requested data journaling mode");
3692 clear_opt(sb, DIOREAD_NOLOCK);
3693 }
3694 if (sb->s_blocksize < PAGE_SIZE) {
3695 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3696 "option - block size is too small");
3697 clear_opt(sb, DIOREAD_NOLOCK);
3698 }
3699 }
3700
3701 err = ext4_setup_system_zone(sb); 3775 err = ext4_setup_system_zone(sb);
3702 if (err) { 3776 if (err) {
3703 ext4_msg(sb, KERN_ERR, "failed to initialize system " 3777 ext4_msg(sb, KERN_ERR, "failed to initialize system "
@@ -3710,22 +3784,19 @@ no_journal:
3710 if (err) { 3784 if (err) {
3711 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 3785 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3712 err); 3786 err);
3713 goto failed_mount4; 3787 goto failed_mount5;
3714 } 3788 }
3715 3789
3716 err = ext4_register_li_request(sb, first_not_zeroed); 3790 err = ext4_register_li_request(sb, first_not_zeroed);
3717 if (err) 3791 if (err)
3718 goto failed_mount4; 3792 goto failed_mount6;
3719 3793
3720 sbi->s_kobj.kset = ext4_kset; 3794 sbi->s_kobj.kset = ext4_kset;
3721 init_completion(&sbi->s_kobj_unregister); 3795 init_completion(&sbi->s_kobj_unregister);
3722 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 3796 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3723 "%s", sb->s_id); 3797 "%s", sb->s_id);
3724 if (err) { 3798 if (err)
3725 ext4_mb_release(sb); 3799 goto failed_mount7;
3726 ext4_ext_release(sb);
3727 goto failed_mount4;
3728 };
3729 3800
3730 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 3801 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3731 ext4_orphan_cleanup(sb, es); 3802 ext4_orphan_cleanup(sb, es);
@@ -3759,13 +3830,19 @@ cantfind_ext4:
3759 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 3830 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3760 goto failed_mount; 3831 goto failed_mount;
3761 3832
3833failed_mount7:
3834 ext4_unregister_li_request(sb);
3835failed_mount6:
3836 ext4_ext_release(sb);
3837failed_mount5:
3838 ext4_mb_release(sb);
3839 ext4_release_system_zone(sb);
3762failed_mount4: 3840failed_mount4:
3763 iput(root); 3841 iput(root);
3764 sb->s_root = NULL; 3842 sb->s_root = NULL;
3765 ext4_msg(sb, KERN_ERR, "mount failed"); 3843 ext4_msg(sb, KERN_ERR, "mount failed");
3766 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq); 3844 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3767failed_mount_wq: 3845failed_mount_wq:
3768 ext4_release_system_zone(sb);
3769 if (sbi->s_journal) { 3846 if (sbi->s_journal) {
3770 jbd2_journal_destroy(sbi->s_journal); 3847 jbd2_journal_destroy(sbi->s_journal);
3771 sbi->s_journal = NULL; 3848 sbi->s_journal = NULL;
@@ -3774,10 +3851,10 @@ failed_mount3:
3774 del_timer(&sbi->s_err_report); 3851 del_timer(&sbi->s_err_report);
3775 if (sbi->s_flex_groups) 3852 if (sbi->s_flex_groups)
3776 ext4_kvfree(sbi->s_flex_groups); 3853 ext4_kvfree(sbi->s_flex_groups);
3777 percpu_counter_destroy(&sbi->s_freeblocks_counter); 3854 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3778 percpu_counter_destroy(&sbi->s_freeinodes_counter); 3855 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3779 percpu_counter_destroy(&sbi->s_dirs_counter); 3856 percpu_counter_destroy(&sbi->s_dirs_counter);
3780 percpu_counter_destroy(&sbi->s_dirtyblocks_counter); 3857 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3781 if (sbi->s_mmp_tsk) 3858 if (sbi->s_mmp_tsk)
3782 kthread_stop(sbi->s_mmp_tsk); 3859 kthread_stop(sbi->s_mmp_tsk);
3783failed_mount2: 3860failed_mount2:
@@ -4064,7 +4141,7 @@ static int ext4_commit_super(struct super_block *sb, int sync)
4064 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4141 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4065 int error = 0; 4142 int error = 0;
4066 4143
4067 if (!sbh) 4144 if (!sbh || block_device_ejected(sb))
4068 return error; 4145 return error;
4069 if (buffer_write_io_error(sbh)) { 4146 if (buffer_write_io_error(sbh)) {
4070 /* 4147 /*
@@ -4100,8 +4177,9 @@ static int ext4_commit_super(struct super_block *sb, int sync)
4100 else 4177 else
4101 es->s_kbytes_written = 4178 es->s_kbytes_written =
4102 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4179 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4103 ext4_free_blocks_count_set(es, percpu_counter_sum_positive( 4180 ext4_free_blocks_count_set(es,
4104 &EXT4_SB(sb)->s_freeblocks_counter)); 4181 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4182 &EXT4_SB(sb)->s_freeclusters_counter)));
4105 es->s_free_inodes_count = 4183 es->s_free_inodes_count =
4106 cpu_to_le32(percpu_counter_sum_positive( 4184 cpu_to_le32(percpu_counter_sum_positive(
4107 &EXT4_SB(sb)->s_freeinodes_counter)); 4185 &EXT4_SB(sb)->s_freeinodes_counter));
@@ -4506,16 +4584,34 @@ restore_opts:
4506 return err; 4584 return err;
4507} 4585}
4508 4586
4587/*
4588 * Note: calculating the overhead so we can be compatible with
4589 * historical BSD practice is quite difficult in the face of
4590 * clusters/bigalloc. This is because multiple metadata blocks from
4591 * different block group can end up in the same allocation cluster.
4592 * Calculating the exact overhead in the face of clustered allocation
4593 * requires either O(all block bitmaps) in memory or O(number of block
4594 * groups**2) in time. We will still calculate the superblock for
4595 * older file systems --- and if we come across with a bigalloc file
4596 * system with zero in s_overhead_clusters the estimate will be close to
4597 * correct especially for very large cluster sizes --- but for newer
4598 * file systems, it's better to calculate this figure once at mkfs
4599 * time, and store it in the superblock. If the superblock value is
4600 * present (even for non-bigalloc file systems), we will use it.
4601 */
4509static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 4602static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4510{ 4603{
4511 struct super_block *sb = dentry->d_sb; 4604 struct super_block *sb = dentry->d_sb;
4512 struct ext4_sb_info *sbi = EXT4_SB(sb); 4605 struct ext4_sb_info *sbi = EXT4_SB(sb);
4513 struct ext4_super_block *es = sbi->s_es; 4606 struct ext4_super_block *es = sbi->s_es;
4607 struct ext4_group_desc *gdp;
4514 u64 fsid; 4608 u64 fsid;
4515 s64 bfree; 4609 s64 bfree;
4516 4610
4517 if (test_opt(sb, MINIX_DF)) { 4611 if (test_opt(sb, MINIX_DF)) {
4518 sbi->s_overhead_last = 0; 4612 sbi->s_overhead_last = 0;
4613 } else if (es->s_overhead_clusters) {
4614 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4519 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) { 4615 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4520 ext4_group_t i, ngroups = ext4_get_groups_count(sb); 4616 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4521 ext4_fsblk_t overhead = 0; 4617 ext4_fsblk_t overhead = 0;
@@ -4530,24 +4626,16 @@ static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4530 * All of the blocks before first_data_block are 4626 * All of the blocks before first_data_block are
4531 * overhead 4627 * overhead
4532 */ 4628 */
4533 overhead = le32_to_cpu(es->s_first_data_block); 4629 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4534 4630
4535 /* 4631 /*
4536 * Add the overhead attributed to the superblock and 4632 * Add the overhead found in each block group
4537 * block group descriptors. If the sparse superblocks
4538 * feature is turned on, then not all groups have this.
4539 */ 4633 */
4540 for (i = 0; i < ngroups; i++) { 4634 for (i = 0; i < ngroups; i++) {
4541 overhead += ext4_bg_has_super(sb, i) + 4635 gdp = ext4_get_group_desc(sb, i, NULL);
4542 ext4_bg_num_gdb(sb, i); 4636 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4543 cond_resched(); 4637 cond_resched();
4544 } 4638 }
4545
4546 /*
4547 * Every block group has an inode bitmap, a block
4548 * bitmap, and an inode table.
4549 */
4550 overhead += ngroups * (2 + sbi->s_itb_per_group);
4551 sbi->s_overhead_last = overhead; 4639 sbi->s_overhead_last = overhead;
4552 smp_wmb(); 4640 smp_wmb();
4553 sbi->s_blocks_last = ext4_blocks_count(es); 4641 sbi->s_blocks_last = ext4_blocks_count(es);
@@ -4555,11 +4643,12 @@ static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4555 4643
4556 buf->f_type = EXT4_SUPER_MAGIC; 4644 buf->f_type = EXT4_SUPER_MAGIC;
4557 buf->f_bsize = sb->s_blocksize; 4645 buf->f_bsize = sb->s_blocksize;
4558 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last; 4646 buf->f_blocks = (ext4_blocks_count(es) -
4559 bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) - 4647 EXT4_C2B(sbi, sbi->s_overhead_last));
4560 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter); 4648 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4649 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4561 /* prevent underflow in case that few free space is available */ 4650 /* prevent underflow in case that few free space is available */
4562 buf->f_bfree = max_t(s64, bfree, 0); 4651 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4563 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es); 4652 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4564 if (buf->f_bfree < ext4_r_blocks_count(es)) 4653 if (buf->f_bfree < ext4_r_blocks_count(es))
4565 buf->f_bavail = 0; 4654 buf->f_bavail = 0;
@@ -4980,13 +5069,11 @@ static int __init ext4_init_fs(void)
4980 return err; 5069 return err;
4981 err = ext4_init_system_zone(); 5070 err = ext4_init_system_zone();
4982 if (err) 5071 if (err)
4983 goto out7; 5072 goto out6;
4984 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 5073 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4985 if (!ext4_kset) 5074 if (!ext4_kset)
4986 goto out6;
4987 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4988 if (!ext4_proc_root)
4989 goto out5; 5075 goto out5;
5076 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4990 5077
4991 err = ext4_init_feat_adverts(); 5078 err = ext4_init_feat_adverts();
4992 if (err) 5079 if (err)
@@ -5022,12 +5109,12 @@ out2:
5022out3: 5109out3:
5023 ext4_exit_feat_adverts(); 5110 ext4_exit_feat_adverts();
5024out4: 5111out4:
5025 remove_proc_entry("fs/ext4", NULL); 5112 if (ext4_proc_root)
5026out5: 5113 remove_proc_entry("fs/ext4", NULL);
5027 kset_unregister(ext4_kset); 5114 kset_unregister(ext4_kset);
5028out6: 5115out5:
5029 ext4_exit_system_zone(); 5116 ext4_exit_system_zone();
5030out7: 5117out6:
5031 ext4_exit_pageio(); 5118 ext4_exit_pageio();
5032 return err; 5119 return err;
5033} 5120}