[PATCH] btrfs: do nofs allocations when adding and removing qgroup relations

[PATCH] btrfs: do nofs allocations when adding and removing qgroup relations

[fdmanana]

From: Filipe Manana <fdmanana@suse.com>

When adding or removing a qgroup relation we are doing a GFP_KERNEL
allocation which is not safe because we are holding a transaction
handle open and that can make us deadlock if the allocator needs to
recurse into the filesystem. So just surround those calls with a
nofs context.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
---
 fs/btrfs/qgroup.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
index cbc0266af001..af00003d9756 100644
--- a/fs/btrfs/qgroup.c
+++ b/fs/btrfs/qgroup.c
@@ -11,6 +11,7 @@
 #include <linux/slab.h>
 #include <linux/workqueue.h>
 #include <linux/btrfs.h>
+#include <linux/sched/mm.h>
 
 #include "ctree.h"
 #include "transaction.h"
@@ -1363,13 +1364,17 @@ int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
 	struct btrfs_qgroup *member;
 	struct btrfs_qgroup_list *list;
 	struct ulist *tmp;
+	unsigned int nofs_flag;
 	int ret = 0;
 
 	/* Check the level of src and dst first */
 	if (btrfs_qgroup_level(src) >= btrfs_qgroup_level(dst))
 		return -EINVAL;
 
+	/* We hold a transaction handle open, must do a NOFS allocation. */
+	nofs_flag = memalloc_nofs_save();
 	tmp = ulist_alloc(GFP_KERNEL);
+	memalloc_nofs_restore(nofs_flag);
 	if (!tmp)
 		return -ENOMEM;
 
@@ -1426,10 +1431,14 @@ static int __del_qgroup_relation(struct btrfs_trans_handle *trans, u64 src,
 	struct btrfs_qgroup_list *list;
 	struct ulist *tmp;
 	bool found = false;
+	unsigned int nofs_flag;
 	int ret = 0;
 	int ret2;
 
+	/* We hold a transaction handle open, must do a NOFS allocation. */
+	nofs_flag = memalloc_nofs_save();
 	tmp = ulist_alloc(GFP_KERNEL);
+	memalloc_nofs_restore(nofs_flag);
 	if (!tmp)
 		return -ENOMEM;
 
-- 
2.28.0

Re: [PATCH] btrfs: do nofs allocations when adding and removing qgroup relations

[David Sterba]

On Mon, Nov 23, 2020 at 06:30:54PM +0000, fdmanana@kernel.org wrote:
> From: Filipe Manana <fdmanana@suse.com>
> 
> When adding or removing a qgroup relation we are doing a GFP_KERNEL
> allocation which is not safe because we are holding a transaction
> handle open and that can make us deadlock if the allocator needs to
> recurse into the filesystem. So just surround those calls with a
> nofs context.
> 
> Signed-off-by: Filipe Manana <fdmanana@suse.com>

Added to misc-next, thanks.