Re: [Qemu-devel] [PATCH v2] 9pfs: fix dependencies

[Cornelia Huck]

On Wed, 9 Aug 2017 10:24:13 +0100
"Daniel P. Berrange" <address@hidden> wrote:

> On Wed, Aug 09, 2017 at 11:07:38AM +0200, Thomas Huth wrote:
> > On 09.08.2017 10:27, Cornelia Huck wrote:  
> > > On Wed, 9 Aug 2017 10:23:04 +0200
> > > Thomas Huth <address@hidden> wrote:
> > >   
> > >> On 09.08.2017 09:17, Cornelia Huck wrote:  
> > >>> Nothing in fsdev/ or hw/9pfs/ depends on pci; it should rather depend
> > >>> on CONFIG_VIRTFS and on the presence of an appropriate virtio transport
> > >>> device.
> > >>>
> > >>> Let's introduce CONFIG_VIRTIO_CCW to cover s390x and check for
> > >>> CONFIG_VIRTFS && (CONFIG_VIRTIO_PCI || CONFIG_VIRTIO_CCW).
> > >>>
> > >>> Signed-off-by: Cornelia Huck <address@hidden>
> > >>> ---
> > >>>
> > >>> Changes v1->v2: drop extraneous spaces, fix build on cris
> > >>>
> > >>> ---
> > >>>  default-configs/s390x-softmmu.mak | 1 +
> > >>>  fsdev/Makefile.objs               | 9 +++------
> > >>>  hw/Makefile.objs                  | 2 +-
> > >>>  3 files changed, 5 insertions(+), 7 deletions(-)  
> > [...]  
> > >>
> > >> Patch should be fine now, I think...
> > >>
> > >> But thinking about this again, I wonder whether it would be enough to
> > >> simply check for CONFIG_VIRTIO=y here instead. CONFIG_VIRTIO=y should be
> > >> sufficient to assert that there is also at least one kind of virtio
> > >> transport available, right?
> > >> Otherwise this will look really horrible as soon as somebody also tries
> > >> to add support for virtio-mmio here later ;-)  
> > > 
> > > Do all virtio transports have support for 9p, though? I thought it was
> > > only virtio-pci and virtio-ccw...  
> > 
> > While virtio-pci and virtio-ccw seem to require separate dedicated
> > devices (e.g. virtio-9p-pci and virtio-9p-ccw) for everything,
> > virtio-mmio seems to work different. As far as I can see, there are no
> > dedicated virtio-xxx-mmio devices in the code at all. According to
> > https://www.redhat.com/archives/libvir-list/2013-August/msg00009.html
> > you simply have to use virtio-xxx-device here instead. And a
> > virtio-9p-device is available. So theoretically, the 9p code should work
> > with virtio-mmio, too, or is there a problem that I did not see yet?
> > 
> > Anyway, we likely should not blindly enable this, so unless somebody has
> > a setup to test it, we should go with your current patch instead, I think.  
> 
> qemu-system-arm supports virtio-mmio so you can use that to test it

Hm, the default config for arm enables CONFIG_PCI, so machines using
virtio-mmio and 9p would be broken with that patch... should we rather
depend on PCI || VIRTIO_CCW?

(Any arches not enabling PCI that use virtio-mmio? Or is arm the only
user of virtio-mmio?)