From: Suraj Jitindar Singh surajjs@amazon.com
commit 7c990728b99ed6fbe9c75fc202fce1172d9916da upstream.
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@amazon.com Signed-off-by: Theodore Ts'o tytso@mit.edu Cc: stable@kernel.org Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org Signed-off-by: Yang Yingliang yangyingliang@huawei.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 fe60352..75a7d78 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -1480,7 +1480,7 @@ 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 0abb72d..127ead1 100644 --- a/fs/ext4/ialloc.c +++ b/fs/ext4/ialloc.c @@ -330,11 +330,13 @@ 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); @@ -370,12 +372,13 @@ 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; }
@@ -1056,7 +1059,8 @@ 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)) { @@ -1079,7 +1083,8 @@ 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 e0aa7ec..1eb6684 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -3038,7 +3038,8 @@ void ext4_exit_mballoc(void) 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); @@ -4931,7 +4932,8 @@ 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)) @@ -5080,7 +5082,8 @@ 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(clusters_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 b965fd3..e9409ec 100644 --- a/fs/ext4/resize.c +++ b/fs/ext4/resize.c @@ -1432,11 +1432,14 @@ static void ext4_update_super(struct super_block *sb, percpu_counter_read(&sbi->s_freeclusters_counter)); if (ext4_has_feature_flex_bg(sb) && 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 d83a585..15f0f28 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -972,6 +972,7 @@ static void ext4_put_super(struct super_block *sb) struct ext4_sb_info *sbi = EXT4_SB(sb); struct ext4_super_block *es = sbi->s_es; struct buffer_head **group_desc; + struct flex_groups **flex_groups; int aborted = 0; int i, err;
@@ -1007,8 +1008,13 @@ 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); @@ -2300,8 +2306,8 @@ 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; @@ -2310,22 +2316,37 @@ 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; }
@@ -2333,6 +2354,7 @@ 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;
@@ -2350,12 +2372,11 @@ 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; @@ -3564,6 +3585,7 @@ 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; @@ -4570,8 +4592,14 @@ 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);