Am 19.01.2010 11:58, schrieb Christoph Hellwig: > On Mon, Jan 18, 2010 at 01:11:27PM +0100, Kevin Wolf wrote: >> Return the appropriate error value instead of always using EIO. Don't free >> the >> L1 table on errors, we still need it. >> >> Signed-off-by: Kevin Wolf <kw...@redhat.com> > > Looks good correct, but shouldn't we free the clusters for the new l1 > table if writing to it fails? At least dependend on whether it's EIO > in which case we could assume the sectos on disk to be worn out.
Agreed, we should try to free them. At least in RHEV, the most likely case for failure will be ENOSPC and we're going to succeed with the free in that case. This will become the first patch of the second series then. ;-) Kevin
