Re: [PATCH] dax: set errors in mapping when writeback fails

[Jeff Layton <jlayton@redhat.com>]

On Fri, 2017-05-26 at 14:31 -0600, Ross Zwisler wrote:
> On Thu, May 25, 2017 at 06:47:46AM -0400, Jeff Layton wrote:
> > Jan's description for this patch is much better than mine, so I'm
> > quoting it verbatim here:
> > 
> > DAX currently doesn't set errors in the mapping when cache flushing
> > fails in dax_writeback_mapping_range(). Since this function can get
> > called only from fsync(2) or sync(2), this is actually as good as it can
> > currently get since we correctly propagate the error up from
> > dax_writeback_mapping_range() to filemap_fdatawrite(). However in the
> > future better writeback error handling will enable us to properly report
> > these errors on fsync(2) even if there are multiple file descriptors
> > open against the file or if sync(2) gets called before fsync(2). So
> > convert DAX to using standard error reporting through the mapping.
> > 
> > Signed-off-by: Jeff Layton <jlayton@redhat.com>
> > Reviewed-by: Jan Kara <jack@suse.cz>
> > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > Reviewed-and-Tested-by: Ross Zwisler <ross.zwisler@linux.intel.com>
> 
> Is this still part of a larger series, or are you trying to submit it on its
> own?

I was trying to submit it on its own. Trying to get a series merged that
sprawls over the tree like this is proving difficult, so I was hoping to
get maintainers to pick up individual patches for their subsystems.

>  On it's own this patch still suffers from the issue I reported here:
> 
> https://lkml.org/lkml/2017/3/6/976
> 

My bad. I remember seeing you mail this and then let it slip through the
cracks. Let's see, you said:

> I think maybe the missing piece is that our normal DAX fsync call stack
> doesn't include a call to filemap_check_errors() if we return -EIO.

I think the actual problem is in filemap_write_and_wait_range. When
there is an error from __filemap_fdatawrite_range, we probably ought to
go ahead and clear the AS_EIO/AS_ENOSPC flags as well even though we're
not waiting on writeback.

I think I had a patch to do that in an earlier iteration of this series,
but I dropped it when I started pursuing the errseq_t based error
tracking.

Now that we're going to have to do this in a more piecemeal way, it may
be worth resurrecting that patch as an interim step before merging this 
one. I'll see if I can dust that patch off after the holiday weekend.

Thanks,
-- 
Jeff Layton <jlayton@redhat.com>