Re: [PATCH v5 06/13] KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Data race causing register corruption in multi-vCPU nested tests due to globally shared guest_regs structure.
--

--- Patch [6]: [PATCH v5 06/13] KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK ---
commit 9a9a1bcc6fc0c85786cd983d4fae11f1dfafdf51
Author: Yosry Ahmed <yosry@kernel.org>

KVM: selftests: Drop HORRIFIC_L2_UCALL_CLOBBER_HACK

This commit drops the hack in ucall_arch_do_ucall(), as the nVMX test
code now preserves GPRs across nested VM-Exits.

> diff --git a/tools/testing/selftests/kvm/lib/x86/ucall.c b/tools/testing/selftests/kvm/lib/x86/ucall.c
> index e7dd5791959ba..38050c60a0670 100644
> --- a/tools/testing/selftests/kvm/lib/x86/ucall.c
> +++ b/tools/testing/selftests/kvm/lib/x86/ucall.c
> @@ -10,36 +10,8 @@
>  
>  void ucall_arch_do_ucall(gva_t uc)
>  {
[ ... ]
> -		     "pop %%r15\n\t"
> -		     "pop %%rbp\n\t"
> -		: : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax", "memory",
> -		     HORRIFIC_L2_UCALL_CLOBBER_HACK);
> +	asm volatile("in %[port], %%al"
> +		: : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax", "memory");
>  }

[Severity: Medium]
Does dropping this hack introduce a regression for multi-vCPU nested tests
where L2 registers could get clobbered?

The nVMX code that now preserves GPRs appears to use a globally shared
structure:

tools/testing/selftests/kvm/lib/x86/processor.c:
struct guest_regs guest_regs;

The VMX_SWITCH_GPRS_ASM macro uses this struct during nested VM entry and
exit:

tools/testing/selftests/kvm/include/x86/vmx.h:
#define VMX_SWITCH_GPRS_ASM \
	GUEST_SWITCH_GPR_ASM(rax) \
	GUEST_SWITCH_GPR_ASM(rbx) \
...

Because guest_regs is not thread-local, wouldn't concurrent vCPU threads
running vmlaunch or vmresume clobber each other's register state when
accessing this shared struct?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260728174232.2423257-1-yosry@kernel.org?part=6