On Thu, 2017-06-15 at 01:22 -0700, Christoph Hellwig wrote:

On Wed, Jun 14, 2017 at 01:24:43PM -0400, Jeff Layton wrote:

quoted

In this smaller set, it's only really used for DAX.

DAX only is implemented by three filesystems, please just fix them
up in one go.

Ok.

quoted

sync_file_range: ->fsync isn't called directly there, and I think we
probably want similar semantics to fsync() for it

sync_file_range is only supposed to sync data, so it should not call
->fsync.

Correct.

But if there is a data writeback error, should we report an error on all
open fds at that time (like we will for fsync)?

I think we probably do want to do that, but like you say...there is no
file op for sync_file_range. It'll need some way to figure out what sort
of error tracking is in play.

quoted

JBD2: will try to re-set the error after clearing it with
filemap_fdatawait. That's problematic with the new infrastructure so we
need some way to avoid it.

JBD2 only has two users, please fix them up in one go.

I came up with a fix yesterday that makes the flag unnecessary there.

Thanks,
-- 
Jeff Layton [off-list ref]