[Devel] [PATCH RHEL7 COMMIT] ms/ext4: fix potential race between s_flex_groups online resizing and access
Konstantin Khorenko
khorenko at virtuozzo.com
Fri Mar 6 18:28:56 MSK 2020
The commit is pushed to "branch-rh7-3.10.0-1062.12.1.vz7.131.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-1062.12.1.vz7.131.8
------>
commit 2edeb3779621453eed8a421f1e9fcb70ec80ebf0
Author: Suraj Jitindar Singh <surajjs at amazon.com>
Date: Fri Mar 6 18:28:56 2020 +0300
ms/ext4: fix potential race between s_flex_groups online resizing and access
ms commit 7c990728b99e
During an online resize an array of s_flex_groups structures gets replaced
so it can get enlarged. If there is a concurrent access to the array and
this memory has been reused then this can lead to an invalid memory access.
The s_flex_group array has been converted into an array of pointers rather
than an array of structures. This is to ensure that the information
contained in the structures cannot get out of sync during a resize due to
an accessor updating the value in the old structure after it has been
copied but before the array pointer is updated. Since the structures them-
selves are no longer copied but only the pointers to them this case is
mitigated.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=206443
Link: https://lore.kernel.org/r/20200221053458.730016-4-tytso@mit.edu
Signed-off-by: Suraj Jitindar Singh <surajjs at amazon.com>
Signed-off-by: Theodore Ts'o <tytso at mit.edu>
Cc: stable at kernel.org
https://jira.sw.ru/browse/PSBM-101798
[@ktkhai: adopted for our kernel]
Signed-off-by: Kirill Tkhai <ktkhai at virtuozzo.com>
---
fs/ext4/ext4.h | 2 +-
fs/ext4/ialloc.c | 23 +++++++++++-------
fs/ext4/mballoc.c | 9 ++++---
fs/ext4/resize.c | 7 ++++--
fs/ext4/super.c | 72 ++++++++++++++++++++++++++++++++++++++-----------------
5 files changed, 76 insertions(+), 37 deletions(-)
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 0d919cb545a65..06cf254096668 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1395,7 +1395,7 @@ struct ext4_sb_info {
unsigned int s_extent_max_zeroout_kb;
unsigned int s_log_groups_per_flex;
- struct flex_groups *s_flex_groups;
+ struct flex_groups * __rcu *s_flex_groups;
ext4_group_t s_flex_groups_allocated;
/* workqueue for reserved extent conversions (buffered io) */
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index b30a983ee9c84..5ac50d155c9ca 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -300,11 +300,13 @@ void ext4_free_inode(handle_t *handle, struct inode *inode)
percpu_counter_inc(&sbi->s_freeinodes_counter);
if (sbi->s_log_groups_per_flex) {
- ext4_group_t f = ext4_flex_group(sbi, block_group);
+ struct flex_groups *fg;
- atomic_inc(&sbi->s_flex_groups[f].free_inodes);
+ fg = sbi_array_rcu_deref(sbi, s_flex_groups,
+ ext4_flex_group(sbi, block_group));
+ atomic_inc(&fg->free_inodes);
if (is_directory)
- atomic_dec(&sbi->s_flex_groups[f].used_dirs);
+ atomic_dec(&fg->used_dirs);
}
BUFFER_TRACE(bh2, "call ext4_handle_dirty_metadata");
fatal = ext4_handle_dirty_metadata(handle, NULL, bh2);
@@ -345,12 +347,13 @@ static void get_orlov_stats(struct super_block *sb, ext4_group_t g,
int flex_size, struct orlov_stats *stats)
{
struct ext4_group_desc *desc;
- struct flex_groups *flex_group = EXT4_SB(sb)->s_flex_groups;
if (flex_size > 1) {
- stats->free_inodes = atomic_read(&flex_group[g].free_inodes);
- stats->free_clusters = atomic64_read(&flex_group[g].free_clusters);
- stats->used_dirs = atomic_read(&flex_group[g].used_dirs);
+ struct flex_groups *fg = sbi_array_rcu_deref(EXT4_SB(sb),
+ s_flex_groups, g);
+ stats->free_inodes = atomic_read(&fg->free_inodes);
+ stats->free_clusters = atomic64_read(&fg->free_clusters);
+ stats->used_dirs = atomic_read(&fg->used_dirs);
return;
}
@@ -932,7 +935,8 @@ struct inode *__ext4_new_inode(handle_t *handle, struct inode *dir,
if (sbi->s_log_groups_per_flex) {
ext4_group_t f = ext4_flex_group(sbi, group);
- atomic_inc(&sbi->s_flex_groups[f].used_dirs);
+ atomic_inc(&sbi_array_rcu_deref(sbi, s_flex_groups,
+ f)->used_dirs);
}
}
if (ext4_has_group_desc_csum(sb)) {
@@ -955,7 +959,8 @@ struct inode *__ext4_new_inode(handle_t *handle, struct inode *dir,
if (sbi->s_log_groups_per_flex) {
flex_group = ext4_flex_group(sbi, group);
- atomic_dec(&sbi->s_flex_groups[flex_group].free_inodes);
+ atomic_dec(&sbi_array_rcu_deref(sbi, s_flex_groups,
+ flex_group)->free_inodes);
}
inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb);
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 76956fce9e767..d837b9b4f10fd 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -2991,7 +2991,8 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
ext4_group_t flex_group = ext4_flex_group(sbi,
ac->ac_b_ex.fe_group);
atomic64_sub(ac->ac_b_ex.fe_len,
- &sbi->s_flex_groups[flex_group].free_clusters);
+ &sbi_array_rcu_deref(sbi, s_flex_groups,
+ flex_group)->free_clusters);
}
err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
@@ -4883,7 +4884,8 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
if (sbi->s_log_groups_per_flex) {
ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
atomic64_add(count_clusters,
- &sbi->s_flex_groups[flex_group].free_clusters);
+ &sbi_array_rcu_deref(sbi, s_flex_groups,
+ flex_group)->free_clusters);
}
if (!(flags & EXT4_FREE_BLOCKS_NO_QUOT_UPDATE))
@@ -5027,7 +5029,8 @@ int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
if (sbi->s_log_groups_per_flex) {
ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
atomic64_add(EXT4_NUM_B2C(sbi, blocks_freed),
- &sbi->s_flex_groups[flex_group].free_clusters);
+ &sbi_array_rcu_deref(sbi, s_flex_groups,
+ flex_group)->free_clusters);
}
ext4_mb_unload_buddy(&e4b);
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 821f30f83b4e7..fbb0640d0a3b0 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -1388,11 +1388,14 @@ static void ext4_update_super(struct super_block *sb,
EXT4_FEATURE_INCOMPAT_FLEX_BG) &&
sbi->s_log_groups_per_flex) {
ext4_group_t flex_group;
+ struct flex_groups *fg;
+
flex_group = ext4_flex_group(sbi, group_data[0].group);
+ fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
atomic64_add(EXT4_NUM_B2C(sbi, free_blocks),
- &sbi->s_flex_groups[flex_group].free_clusters);
+ &fg->free_clusters);
atomic_add(EXT4_INODES_PER_GROUP(sb) * flex_gd->count,
- &sbi->s_flex_groups[flex_group].free_inodes);
+ &fg->free_inodes);
}
/*
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index f752e0190cdf2..a6b2522655d3e 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -894,6 +894,7 @@ static void ext4_put_super(struct super_block *sb)
{
struct ext4_sb_info *sbi = EXT4_SB(sb);
struct buffer_head **group_desc;
+ struct flex_groups **flex_groups;
struct ext4_super_block *es = sbi->s_es;
int aborted = 0;
struct workqueue_struct * rsv_conversion_wq = sbi->rsv_conversion_wq;
@@ -941,8 +942,13 @@ static void ext4_put_super(struct super_block *sb)
for (i = 0; i < sbi->s_gdb_count; i++)
brelse(group_desc[i]);
kvfree(group_desc);
+ flex_groups = rcu_dereference(sbi->s_flex_groups);
+ if (flex_groups) {
+ for (i = 0; i < sbi->s_flex_groups_allocated; i++)
+ kvfree(flex_groups[i]);
+ kvfree(flex_groups);
+ }
rcu_read_unlock();
- kvfree(sbi->s_flex_groups);
percpu_counter_destroy(&sbi->s_freeclusters_counter);
percpu_counter_destroy(&sbi->s_freeinodes_counter);
percpu_counter_destroy(&sbi->s_dirs_counter);
@@ -2162,8 +2168,8 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
{
struct ext4_sb_info *sbi = EXT4_SB(sb);
- struct flex_groups *new_groups;
- int size;
+ struct flex_groups **old_groups, **new_groups;
+ int size, i;
if (!sbi->s_log_groups_per_flex)
return 0;
@@ -2172,22 +2178,37 @@ int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
if (size <= sbi->s_flex_groups_allocated)
return 0;
- size = roundup_pow_of_two(size * sizeof(struct flex_groups));
- new_groups = kvzalloc(size, GFP_KERNEL);
+ new_groups = kvzalloc(roundup_pow_of_two(size *
+ sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
if (!new_groups) {
- ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
- size / (int) sizeof(struct flex_groups));
+ ext4_msg(sb, KERN_ERR,
+ "not enough memory for %d flex group pointers", size);
return -ENOMEM;
}
-
- if (sbi->s_flex_groups) {
- memcpy(new_groups, sbi->s_flex_groups,
- (sbi->s_flex_groups_allocated *
- sizeof(struct flex_groups)));
- kvfree(sbi->s_flex_groups);
+ for (i = sbi->s_flex_groups_allocated; i < size; i++) {
+ new_groups[i] = kvzalloc(roundup_pow_of_two(
+ sizeof(struct flex_groups)),
+ GFP_KERNEL);
+ if (!new_groups[i]) {
+ for (i--; i >= sbi->s_flex_groups_allocated; i--)
+ kvfree(new_groups[i]);
+ kvfree(new_groups);
+ ext4_msg(sb, KERN_ERR,
+ "not enough memory for %d flex groups", size);
+ return -ENOMEM;
+ }
}
- sbi->s_flex_groups = new_groups;
- sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
+ rcu_read_lock();
+ old_groups = rcu_dereference(sbi->s_flex_groups);
+ if (old_groups)
+ memcpy(new_groups, old_groups,
+ (sbi->s_flex_groups_allocated *
+ sizeof(struct flex_groups *)));
+ rcu_read_unlock();
+ rcu_assign_pointer(sbi->s_flex_groups, new_groups);
+ sbi->s_flex_groups_allocated = size;
+ if (old_groups)
+ ext4_kvfree_array_rcu(old_groups);
return 0;
}
@@ -2227,6 +2248,7 @@ static int ext4_fill_flex_info(struct super_block *sb)
{
struct ext4_sb_info *sbi = EXT4_SB(sb);
struct ext4_group_desc *gdp = NULL;
+ struct flex_groups *fg;
ext4_group_t flex_group;
int i, err;
@@ -2244,12 +2266,11 @@ static int ext4_fill_flex_info(struct super_block *sb)
gdp = ext4_get_group_desc(sb, i, NULL);
flex_group = ext4_flex_group(sbi, i);
- atomic_add(ext4_free_inodes_count(sb, gdp),
- &sbi->s_flex_groups[flex_group].free_inodes);
+ fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
+ atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
atomic64_add(ext4_free_group_clusters(sb, gdp),
- &sbi->s_flex_groups[flex_group].free_clusters);
- atomic_add(ext4_used_dirs_count(sb, gdp),
- &sbi->s_flex_groups[flex_group].used_dirs);
+ &fg->free_clusters);
+ atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
}
return 1;
@@ -3862,6 +3883,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
struct buffer_head *bh, **group_desc;
struct ext4_super_block *es = NULL;
struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
+ struct flex_groups **flex_groups;
ext4_fsblk_t block;
ext4_fsblk_t sb_block = get_sb_block(&data);
ext4_fsblk_t logical_sb_block;
@@ -4794,8 +4816,14 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ext4_unregister_li_request(sb);
failed_mount6:
ext4_mb_release(sb);
- if (sbi->s_flex_groups)
- kvfree(sbi->s_flex_groups);
+ rcu_read_lock();
+ flex_groups = rcu_dereference(sbi->s_flex_groups);
+ if (flex_groups) {
+ for (i = 0; i < sbi->s_flex_groups_allocated; i++)
+ kvfree(flex_groups[i]);
+ kvfree(flex_groups);
+ }
+ rcu_read_unlock();
percpu_counter_destroy(&sbi->s_freeclusters_counter);
percpu_counter_destroy(&sbi->s_freeinodes_counter);
percpu_counter_destroy(&sbi->s_dirs_counter);
More information about the Devel
mailing list