[PATCH 0/3] irqchip/gic-v4: Fix VMAPP/VMOVP races

[Marc Zyngier]

On Wed, 17 Jul 2024 09:41:30 +0100,
Tangnianyao <tangnianyao at huawei.com> wrote:
> 
> Hi, Marc
> 
> I meet another problem while fixing this in kernel 5.10.
> 
> Kernel 5.10 does not support guard and we replace it with raw_spin_lock/unlock.
> When guest insmod nic drivers, it trigger host cpu power off somehow. The same
> testcase runs quite good in kernel 6.6(both host and guest).

This suggests that you got the locking order wrong, that one or
several CPUs deadlock, and that a watchdog fires.

> Would you fix this on long-term kernel 5.10? Or any sugguestion?

Sorry, but I have no intention to maintain such an ancient kernel,
which is why I did not Cc stable on these patches. I have no bandwidth
for this, nor any interest in it.

	M.

-- 
Without deviation from the norm, progress is not possible.