Re: [PATCH 5/7] qemu-kvm: Move gsi bits from kvm_msix_vector_add to kvm_msi_add_message

["Michael S. Tsirkin" <mst@redhat.com>]

On Wed, Apr 27, 2011 at 03:29:15PM +0200, Jan Kiszka wrote:
> On 2011-04-27 14:54, Michael S. Tsirkin wrote:
> > On Sat, Apr 23, 2011 at 12:23:38PM +0200, Jan Kiszka wrote:
> >> From: Jan Kiszka <jan.kiszka@siemens.com>
> >>
> >> Testing support and allocating a GSI for an MSI message is required both
> >> for MSI and MSI-X. At this chance, drop the aging version warning.
> >>
> >> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> > 
> > No objection, but I do note that this means running on an old
> > kernel will lead to a silent failure.
> 
> Shouldn't be silent: The caller of msi_vector_use should check and
> process the error (virtio should even forward it to guest IIUC).

It does, but the msi/msix spec does not allow this so not
all guest OSes can use this extended reporting.

> > stderr output is not in fact much better: I think we should
> > check the capability in msix_init. Care coding this up?
> 
> I think the motivation to check on vector activation is that devices and
> guests without a need for MSI should not cause a failure if MSI is
> unsupported by KVM.
> 
> Jan

IMO it's better not to report MSI capability in this setup.


> -- 
> Siemens AG, Corporate Technology, CT T DE IT 1
> Corporate Competence Center Embedded Linux