ext4: fix potential race between s_flex_groups online resizing and access

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

1 files changed, 50 insertions, 22 deletions

diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index e00bcc19099f..6b7e628b7903 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -1015,6 +1015,7 @@ 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;
 
@@ -1052,8 +1053,13 @@ 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);
@@ -2384,8 +2390,8 @@ done:
 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;
@@ -2394,22 +2400,37 @@ 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;
 }
 
@@ -2417,6 +2438,7 @@ 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;
 
@@ -2434,12 +2456,11 @@ 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;
@@ -3641,6 +3662,7 @@ 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;
@@ -4692,8 +4714,14 @@ failed_mount7:
 	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);