[RFC]xfs: using GFP_NOFS for blkdev_issue_flush

[Shaohua Li]

On Wed, Apr 18, 2012 at 11:37:46AM +1000, Dave Chinner wrote:
> On Mon, Apr 16, 2012 at 02:58:21PM +0800, Shaohua Li wrote:
> > On 4/16/12 2:49 PM, Shaohua Li wrote:
> > >
> > >flush request is issued in transaction commit code path usually, so
> > >looks using
> > >GFP_KERNEL to allocate memory for flush request bio falls into the classic
> > >deadlock issue (memory reclaim recursion). Use GFP_NOFS to avoid recursion
> > >from reclaim context. Per Dave Chinner, there is only blkdev_issue_flush
> > >might
> > >be buggy here. But using GFP_NOFS by default for all calls should not
> > >matter.
> 
> Can you update the commit message like I suggested previously?
Copied exactly.


Issuing a block device flush request in transaction context using GFP_KERNEL
directly can cause deadlocks due to memory reclaim recursion. Use GFP_NOFS to
avoid recursion from reclaim context.

Signed-off-by: Shaohua Li <shli at fusionio.com>
Reviewed-by: Mark Tinguely <tinguely at sgi.com>
---
 fs/xfs/xfs_super.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux/fs/xfs/xfs_super.c
===================================================================
--- linux.orig/fs/xfs/xfs_super.c	2012-04-13 10:08:26.095496072 +0800
+++ linux/fs/xfs/xfs_super.c	2012-04-13 10:08:42.555496865 +0800
@@ -622,7 +622,7 @@ void
 xfs_blkdev_issue_flush(
 	xfs_buftarg_t		*buftarg)
 {
-	blkdev_issue_flush(buftarg->bt_bdev, GFP_KERNEL, NULL);
+	blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
 }
 
 STATIC void