On Wed, Oct 09, 2019 at 07:16:39PM +0200, Greg Kurz wrote: > On Wed, 9 Oct 2019 19:52:59 +1100 > David Gibson <da...@gibson.dropbear.id.au> wrote: > > > On Wed, Oct 09, 2019 at 10:37:38AM +0200, Greg Kurz wrote: > > > On Wed, 9 Oct 2019 17:08:18 +1100 > > > David Gibson <da...@gibson.dropbear.id.au> wrote: > > > > > > > Traditional PCI INTx for vfio devices can only perform well if using > > > > an in-kernel irqchip. Therefore, vfio_intx_update() issues a warning > > > > if an in kernel irqchip is not available. > > > > > > Can you elaborate on what doesn't "perform well" without an > > > in-kernel irqchip ? > > > > Not really, no, but Alex Williamson tells me it is soo. > > > > > Is it a matter of performance or is it > > > actually broken or something else ? > > > > I believe it's a matter of performance, but such a big one that it > > makes using it without kernel irqchip infeasible in many cases. > > > > > What is the impact on -machine accel=kvm,kernel-irqchip=off which > > > always spit this warning ? > > > > It should still spit that warning. > > > > Yeah of course it does, but it is expected that we cannot setup > the irqfd since we deliberately don't have an in-kernel irqchip, > isn't it ? Why spit a warning in this case ? Or is it just a not > very user friendly way of saying "you will have poor performance > if the VFIO device uses PCI INTx" ?
The last thing, AIUI.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
signature.asc
Description: PGP signature
