On Wed, Apr 29, 2015 at 04:33:44PM +0200, Cornelia Huck wrote: > On Wed, 29 Apr 2015 12:32:04 +0200 > "Michael S. Tsirkin" <m...@redhat.com> wrote: > > > On Wed, Apr 29, 2015 at 10:52:15AM +0200, Cornelia Huck wrote: > > > > And (surprise, surprise) virtio-blk now works - but it also works when > > > I back out the atomic.h change again. No barrier problems :) > > > > > > Good news is that we can change s390-virtio to be just like the other > > > transports. Although I'd like to understand why it was broken before. > > > Maybe a guest change? > > > > Or a compiler change? Try compiling some old release, see what happens. > > Nope, same compiler as back then (old setup). My current guess is some > change in other code.
bisect is you friend then :) > > Anyway, let's move DEFINE_VIRTIO_COMMON_FEATURES into the base class > > now. Can you send a patch pls? > > I think virtio-blk-s390 not working anymore will be easy to catch (it > is used in some of the iotests). I'll move the features.