Re: [Qemu-devel] [PATCH for-4.0.1] q35: Revert to kernel irqchip

[Peter Xu]

On Tue, May 14, 2019 at 02:22:03PM -0600, Alex Williamson wrote:
> On Tue, 14 May 2019 13:03:31 -0600
> Alex Williamson <address@hidden> wrote:
> 
> > Commit b2fc91db8447 ("q35: set split kernel irqchip as default") changed
> > the default for the pc-q35-4.0 machine type to use split irqchip, which
> > turned out to have disasterous effects on vfio-pci INTx support.  KVM
> > resampling irqfds are registered for handling these interrupts, but
> > these are non-functional in split irqchip mode.  We can't simply test
> > for split irqchip in QEMU as userspace handling of this interrupt is a
> > significant performance regression versus KVM handling (GeForce GPUs
> > assigned to Windows VMs are non-functional without forcing MSI mode or
> > re-enabling kernel irqchip).
> > 
> > The resolution is to revert the change in default irqchip mode with a
> > new pc-q35-4.0.1 machine type for qemu-stable while the development
> > branch makes the same change in the pc-q35-4.1 machine type.  The
> > qemu-q35-4.0 machine type should not be used in vfio-pci configurations
> > for devices requiring legacy INTx support without explicitly modifying
> > the VM configuration to use KVM irqchip.  This new 4.0.1 machine type
> > makes this change automatically.
> > 
> > Link: https://bugs.launchpad.net/qemu/+bug/1826422
> > Link: https://lists.gnu.org/archive/html/qemu-devel/2019-05/msg03305.html
> 
> This link is superseded by a v2 of the mainline patch:
> 
> Link: https://lists.gnu.org/archive/html/qemu-devel/2019-05/msg03338.html
> 
> I believe this patch is still the proper stable backport though.  Also
> to clarify, this patch should be gated on mainline acceptance of the
> link above, but clearly there's no clean cherry-pick between mainline
> and stable for this, so I'm proposing them in parallel.  Thanks,

Agreed.  As long as the 4.1 patch can be accepted, this should be the
correct patch for 4.0-stable AFAICT:

Reviewed-by: Peter Xu <address@hidden>

Thanks,

-- 
Peter Xu