Re: [PATCH 2/3] xfs: truncate delalloc extents when IO fails in writeback

[Dave Chinner <david@xxxxxxxxxxxxx>]

On Wed, Mar 10, 2010 at 04:12:37AM -0500, Christoph Hellwig wrote:
> Btw, I can trigger the new warnings from this patch in xfsqa after
> a few full XFSQA runs:
> 
> [39829.737139] Filesystem "vdb6": page discard on page c21d7480, inode
> 0x3f8, offset 0.
> [39829.739488] Filesystem "vdb6": page discard failed delalloc mapping
> lookup.
> [39829.744005] Filesystem "vdb6": page discard on page c206ba80, inode
> 0x3f9, offset 0.
> [39829.746496] Filesystem "vdb6": page discard failed delalloc mapping
> lookup.
> [39829.748655] Filesystem "vdb6": page discard on page c20e02a0, inode
> 0x3fa, offset 0.
> [39829.751000] Filesystem "vdb6": page discard failed delalloc mapping
> lookup.

The "page discard on page ...." are expected, but the failed mapping
warnings are not - that means xfs_bmapi() returned an error while
the inode still has delayed allocation buffers in memory. Did this
happen during a forced shutdown?  What test is running when this
occurs?

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx