ext4: fix potential race between online resizing and write operations

During an online resize an array of pointers to buffer heads gets
replaced so it can get enlarged.  If there is a racing block
allocation or deallocation which uses the old array, and the old array
has gotten reused this can lead to a GPF or some other random kernel
memory getting modified.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=206443
Link: https://lore.kernel.org/r/20200221053458.730016-2-tytso@mit.edu
Reported-by: Suraj Jitindar Singh <surajjs@amazon.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org

1 files changed, 44 insertions, 11 deletions

diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 86a2500ed292..536cc9f38091 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -17,6 +17,33 @@
 
 #include "ext4_jbd2.h"
 
+struct ext4_rcu_ptr {
+	struct rcu_head rcu;
+	void *ptr;
+};
+
+static void ext4_rcu_ptr_callback(struct rcu_head *head)
+{
+	struct ext4_rcu_ptr *ptr;
+
+	ptr = container_of(head, struct ext4_rcu_ptr, rcu);
+	kvfree(ptr->ptr);
+	kfree(ptr);
+}
+
+void ext4_kvfree_array_rcu(void *to_free)
+{
+	struct ext4_rcu_ptr *ptr = kzalloc(sizeof(*ptr), GFP_KERNEL);
+
+	if (ptr) {
+		ptr->ptr = to_free;
+		call_rcu(&ptr->rcu, ext4_rcu_ptr_callback);
+		return;
+	}
+	synchronize_rcu();
+	kvfree(to_free);
+}
+
 int ext4_resize_begin(struct super_block *sb)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
@@ -542,8 +569,8 @@ static int setup_new_flex_group_blocks(struct super_block *sb,
 				brelse(gdb);
 				goto out;
 			}
-			memcpy(gdb->b_data, sbi->s_group_desc[j]->b_data,
-			       gdb->b_size);
+			memcpy(gdb->b_data, sbi_array_rcu_deref(sbi,
+				s_group_desc, j)->b_data, gdb->b_size);
 			set_buffer_uptodate(gdb);
 
 			err = ext4_handle_dirty_metadata(handle, NULL, gdb);
@@ -860,13 +887,15 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
 	}
 	brelse(dind);
 
-	o_group_desc = EXT4_SB(sb)->s_group_desc;
+	rcu_read_lock();
+	o_group_desc = rcu_dereference(EXT4_SB(sb)->s_group_desc);
 	memcpy(n_group_desc, o_group_desc,
 	       EXT4_SB(sb)->s_gdb_count * sizeof(struct buffer_head *));
+	rcu_read_unlock();
 	n_group_desc[gdb_num] = gdb_bh;
-	EXT4_SB(sb)->s_group_desc = n_group_desc;
+	rcu_assign_pointer(EXT4_SB(sb)->s_group_desc, n_group_desc);
 	EXT4_SB(sb)->s_gdb_count++;
-	kvfree(o_group_desc);
+	ext4_kvfree_array_rcu(o_group_desc);
 
 	le16_add_cpu(&es->s_reserved_gdt_blocks, -1);
 	err = ext4_handle_dirty_super(handle, sb);
@@ -909,9 +938,11 @@ static int add_new_gdb_meta_bg(struct super_block *sb,
 		return err;
 	}
 
-	o_group_desc = EXT4_SB(sb)->s_group_desc;
+	rcu_read_lock();
+	o_group_desc = rcu_dereference(EXT4_SB(sb)->s_group_desc);
 	memcpy(n_group_desc, o_group_desc,
 	       EXT4_SB(sb)->s_gdb_count * sizeof(struct buffer_head *));
+	rcu_read_unlock();
 	n_group_desc[gdb_num] = gdb_bh;
 
 	BUFFER_TRACE(gdb_bh, "get_write_access");
@@ -922,9 +953,9 @@ static int add_new_gdb_meta_bg(struct super_block *sb,
 		return err;
 	}
 
-	EXT4_SB(sb)->s_group_desc = n_group_desc;
+	rcu_assign_pointer(EXT4_SB(sb)->s_group_desc, n_group_desc);
 	EXT4_SB(sb)->s_gdb_count++;
-	kvfree(o_group_desc);
+	ext4_kvfree_array_rcu(o_group_desc);
 	return err;
 }
 
@@ -1188,7 +1219,8 @@ static int ext4_add_new_descs(handle_t *handle, struct super_block *sb,
 		 * use non-sparse filesystems anymore.  This is already checked above.
 		 */
 		if (gdb_off) {
-			gdb_bh = sbi->s_group_desc[gdb_num];
+			gdb_bh = sbi_array_rcu_deref(sbi, s_group_desc,
+						     gdb_num);
 			BUFFER_TRACE(gdb_bh, "get_write_access");
 			err = ext4_journal_get_write_access(handle, gdb_bh);
 
@@ -1270,7 +1302,7 @@ static int ext4_setup_new_descs(handle_t *handle, struct super_block *sb,
 		/*
 		 * get_write_access() has been called on gdb_bh by ext4_add_new_desc().
 		 */
-		gdb_bh = sbi->s_group_desc[gdb_num];
+		gdb_bh = sbi_array_rcu_deref(sbi, s_group_desc, gdb_num);
 		/* Update group descriptor block for new group */
 		gdp = (struct ext4_group_desc *)(gdb_bh->b_data +
 						 gdb_off * EXT4_DESC_SIZE(sb));
@@ -1497,7 +1529,8 @@ exit_journal:
 		for (; gdb_num <= gdb_num_end; gdb_num++) {
 			struct buffer_head *gdb_bh;
 
-			gdb_bh = sbi->s_group_desc[gdb_num];
+			gdb_bh = sbi_array_rcu_deref(sbi, s_group_desc,
+						     gdb_num);
 			if (old_gdb == gdb_bh->b_blocknr)
 				continue;
 			update_backups(sb, gdb_bh->b_blocknr, gdb_bh->b_data,