ext4: Avoid races caused by on-line resizing and SMP memory reordering
Ext4's on-line resizing adds a new block group and then, only at the
last step adjusts s_groups_count. However, it's possible on SMP
systems that another CPU could see the updated the s_group_count and
not see the newly initialized data structures for the just-added block
group. For this reason, it's important to insert a SMP read barrier
after reading s_groups_count and before reading any (for example) the
new block group descriptors allowed by the increased value of
s_groups_count.
Unfortunately, we rather blatently violate this locking protocol
documented in fs/ext4/resize.c. Fortunately, (1) on-line resizes
happen relatively rarely, and (2) it seems rare that the filesystem
code will immediately try to use just-added block group before any
memory ordering issues resolve themselves. So apparently problems
here are relatively hard to hit, since ext3 has been vulnerable to the
same issue for years with no one apparently complaining.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index f871677..c3af9e6 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -739,6 +739,7 @@
static int ext4_mb_init_cache(struct page *page, char *incore)
{
+ ext4_group_t ngroups;
int blocksize;
int blocks_per_page;
int groups_per_page;
@@ -757,6 +758,7 @@
inode = page->mapping->host;
sb = inode->i_sb;
+ ngroups = ext4_get_groups_count(sb);
blocksize = 1 << inode->i_blkbits;
blocks_per_page = PAGE_CACHE_SIZE / blocksize;
@@ -780,7 +782,7 @@
for (i = 0; i < groups_per_page; i++) {
struct ext4_group_desc *desc;
- if (first_group + i >= EXT4_SB(sb)->s_groups_count)
+ if (first_group + i >= ngroups)
break;
err = -EIO;
@@ -852,7 +854,7 @@
struct ext4_group_info *grinfo;
group = (first_block + i) >> 1;
- if (group >= EXT4_SB(sb)->s_groups_count)
+ if (group >= ngroups)
break;
/*
@@ -1788,6 +1790,7 @@
int block, pnum;
int blocks_per_page;
int groups_per_page;
+ ext4_group_t ngroups = ext4_get_groups_count(sb);
ext4_group_t first_group;
struct ext4_group_info *grp;
@@ -1807,7 +1810,7 @@
/* read all groups the page covers into the cache */
for (i = 0; i < groups_per_page; i++) {
- if ((first_group + i) >= EXT4_SB(sb)->s_groups_count)
+ if ((first_group + i) >= ngroups)
break;
grp = ext4_get_group_info(sb, first_group + i);
/* take all groups write allocation
@@ -1945,8 +1948,7 @@
static noinline_for_stack int
ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
{
- ext4_group_t group;
- ext4_group_t i;
+ ext4_group_t ngroups, group, i;
int cr;
int err = 0;
int bsbits;
@@ -1957,6 +1959,7 @@
sb = ac->ac_sb;
sbi = EXT4_SB(sb);
+ ngroups = ext4_get_groups_count(sb);
BUG_ON(ac->ac_status == AC_STATUS_FOUND);
/* first, try the goal */
@@ -2017,11 +2020,11 @@
*/
group = ac->ac_g_ex.fe_group;
- for (i = 0; i < EXT4_SB(sb)->s_groups_count; group++, i++) {
+ for (i = 0; i < ngroups; group++, i++) {
struct ext4_group_info *grp;
struct ext4_group_desc *desc;
- if (group == EXT4_SB(sb)->s_groups_count)
+ if (group == ngroups)
group = 0;
/* quick check to skip empty groups */
@@ -2315,12 +2318,10 @@
static void *ext4_mb_seq_groups_start(struct seq_file *seq, loff_t *pos)
{
struct super_block *sb = seq->private;
- struct ext4_sb_info *sbi = EXT4_SB(sb);
ext4_group_t group;
- if (*pos < 0 || *pos >= sbi->s_groups_count)
+ if (*pos < 0 || *pos >= ext4_get_groups_count(sb))
return NULL;
-
group = *pos + 1;
return (void *) ((unsigned long) group);
}
@@ -2328,11 +2329,10 @@
static void *ext4_mb_seq_groups_next(struct seq_file *seq, void *v, loff_t *pos)
{
struct super_block *sb = seq->private;
- struct ext4_sb_info *sbi = EXT4_SB(sb);
ext4_group_t group;
++*pos;
- if (*pos < 0 || *pos >= sbi->s_groups_count)
+ if (*pos < 0 || *pos >= ext4_get_groups_count(sb))
return NULL;
group = *pos + 1;
return (void *) ((unsigned long) group);
@@ -2587,6 +2587,7 @@
static int ext4_mb_init_backend(struct super_block *sb)
{
+ ext4_group_t ngroups = ext4_get_groups_count(sb);
ext4_group_t i;
int metalen;
struct ext4_sb_info *sbi = EXT4_SB(sb);
@@ -2598,7 +2599,7 @@
struct ext4_group_desc *desc;
/* This is the number of blocks used by GDT */
- num_meta_group_infos = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) -
+ num_meta_group_infos = (ngroups + EXT4_DESC_PER_BLOCK(sb) -
1) >> EXT4_DESC_PER_BLOCK_BITS(sb);
/*
@@ -2644,7 +2645,7 @@
for (i = 0; i < num_meta_group_infos; i++) {
if ((i + 1) == num_meta_group_infos)
metalen = sizeof(*meta_group_info) *
- (sbi->s_groups_count -
+ (ngroups -
(i << EXT4_DESC_PER_BLOCK_BITS(sb)));
meta_group_info = kmalloc(metalen, GFP_KERNEL);
if (meta_group_info == NULL) {
@@ -2655,7 +2656,7 @@
sbi->s_group_info[i] = meta_group_info;
}
- for (i = 0; i < sbi->s_groups_count; i++) {
+ for (i = 0; i < ngroups; i++) {
desc = ext4_get_group_desc(sb, i, NULL);
if (desc == NULL) {
printk(KERN_ERR
@@ -2781,13 +2782,14 @@
int ext4_mb_release(struct super_block *sb)
{
+ ext4_group_t ngroups = ext4_get_groups_count(sb);
ext4_group_t i;
int num_meta_group_infos;
struct ext4_group_info *grinfo;
struct ext4_sb_info *sbi = EXT4_SB(sb);
if (sbi->s_group_info) {
- for (i = 0; i < sbi->s_groups_count; i++) {
+ for (i = 0; i < ngroups; i++) {
grinfo = ext4_get_group_info(sb, i);
#ifdef DOUBLE_CHECK
kfree(grinfo->bb_bitmap);
@@ -2797,7 +2799,7 @@
ext4_unlock_group(sb, i);
kfree(grinfo);
}
- num_meta_group_infos = (sbi->s_groups_count +
+ num_meta_group_infos = (ngroups +
EXT4_DESC_PER_BLOCK(sb) - 1) >>
EXT4_DESC_PER_BLOCK_BITS(sb);
for (i = 0; i < num_meta_group_infos; i++)
@@ -4121,7 +4123,7 @@
static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
{
struct super_block *sb = ac->ac_sb;
- ext4_group_t i;
+ ext4_group_t ngroups, i;
printk(KERN_ERR "EXT4-fs: Can't allocate:"
" Allocation context details:\n");
@@ -4145,7 +4147,8 @@
printk(KERN_ERR "EXT4-fs: %lu scanned, %d found\n", ac->ac_ex_scanned,
ac->ac_found);
printk(KERN_ERR "EXT4-fs: groups: \n");
- for (i = 0; i < EXT4_SB(sb)->s_groups_count; i++) {
+ ngroups = ext4_get_groups_count(sb);
+ for (i = 0; i < ngroups; i++) {
struct ext4_group_info *grp = ext4_get_group_info(sb, i);
struct ext4_prealloc_space *pa;
ext4_grpblk_t start;
@@ -4469,13 +4472,13 @@
static int ext4_mb_discard_preallocations(struct super_block *sb, int needed)
{
- ext4_group_t i;
+ ext4_group_t i, ngroups = ext4_get_groups_count(sb);
int ret;
int freed = 0;
trace_mark(ext4_mb_discard_preallocations, "dev %s needed %d",
sb->s_id, needed);
- for (i = 0; i < EXT4_SB(sb)->s_groups_count && needed > 0; i++) {
+ for (i = 0; i < ngroups && needed > 0; i++) {
ret = ext4_mb_discard_group_preallocations(sb, i, needed);
freed += ret;
needed -= ret;