Re: [PATCH 3/6] xfs: don't stall cowblocks scan if we can't take locks

On Thu, Jan 14, 2021 at 01:54:53PM -0800, Darrick J. Wong wrote:

On Wed, Jan 13, 2021 at 03:43:57PM +0100, Christoph Hellwig wrote:

quoted

On Mon, Jan 11, 2021 at 03:22:52PM -0800, Darrick J. Wong wrote:

quoted

From: Darrick J. Wong <djwong@kernel.org>

Don't stall the cowblocks scan on a locked inode if we possibly can.
We'd much rather the background scanner keep moving.

Wouldn't it make more sense to move the logic to ignore the -EAGAIN
for not-sync calls into xfs_inode_walk_ag?

I'm not sure what you're asking here?  _free_cowblocks only returns
EAGAIN for sync calls.  Locking failure for a not-sync call results in a
return 0, which means that _walk_ag just moves on to the next inode.

What I mean is:

 - always return -EAGAIN when taking the locks fails
 - don't exit early on -EAGAIN in xfs_inode_walk at least for sync
   calls, although thinking loud I see no good reason to exit early
   even for non-sync invocations