REGRESSION: Upgrading host kernel from 5.11 to 5.13 breaks QEMU guests - perf/fw_devlink/kvm

[Russell King (Oracle)]

On Sun, Sep 19, 2021 at 02:36:46PM +0100, Marc Zyngier wrote:
> Urgh. That's a bummer. T1he PMU driver only comes up once it has found
> its interrupt controller, which on the Armada 8040 is not the GIC, but
> some weird thing on the side that doesn't actually serve any real
> purpose. On HW where the PMU is directly wired into the GIC, it all
> works fine, though by luck rather than by design.
> 
> Anyway, rant over. This is a bug that needs addressing so that KVM can
> initialise correctly irrespective of the probing order. This probably
> means that the static key controlling KVM's behaviour wrt the PMU must
> be controlled by the PMU infrastructure itself, rather than KVM trying
> to probe for it.
> 
> Can you please give the following hack a go (on top of 5.15-rc1)? I've
> briefly tested it on my McBin, and it did the trick. I've also tested
> it on the M1 (which really doesn't have an architectural PMU) to
> verify that it was correctly failing.

My test program that derives the number of registers qemu uses now
reports 236 registers again and I see:

kvm [7]: PMU detected and enabled

in the kernel boot log.

Tested-by: Russell King (Oracle) <rmk+kernel at armlinux.org.uk>

Thanks.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!