linux-kernel - Re: [PATCH v2 2/7] KVM: x86: Extract VMXON and EFER.SVME enablement to kernel

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <aUS06wE6IvFti8Le@yilunxu-OptiPlex-7050>
Date: Fri, 19 Dec 2025 10:14:03 +0800
From: Xu Yilun <yilun.xu@...ux.intel.com>
To: Sean Christopherson <seanjc@...gle.com>
Cc: Chao Gao <chao.gao@...el.com>, Thomas Gleixner <tglx@...utronix.de>,
	Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
	Dave Hansen <dave.hansen@...ux.intel.com>, x86@...nel.org,
	Kiryl Shutsemau <kas@...nel.org>,
	Paolo Bonzini <pbonzini@...hat.com>, linux-kernel@...r.kernel.org,
	linux-coco@...ts.linux.dev, kvm@...r.kernel.org,
	Dan Williams <dan.j.williams@...el.com>
Subject: Re: [PATCH v2 2/7] KVM: x86: Extract VMXON and EFER.SVME enablement
 to kernel

On Wed, Dec 17, 2025 at 11:01:59AM -0800, Sean Christopherson wrote:
> On Wed, Dec 17, 2025, Xu Yilun wrote:
> > > >+#define x86_virt_call(fn)				\
> > > >+({							\
> > > >+	int __r;					\
> > > >+							\
> > > >+	if (IS_ENABLED(CONFIG_KVM_INTEL) &&		\
> > > >+	    cpu_feature_enabled(X86_FEATURE_VMX))	\
> > > >+		__r = x86_vmx_##fn();			\
> > > >+	else if (IS_ENABLED(CONFIG_KVM_AMD) &&		\
> > > >+		 cpu_feature_enabled(X86_FEATURE_SVM))	\
> > > >+		__r = x86_svm_##fn();			\
> > > >+	else						\
> > > >+		__r = -EOPNOTSUPP;			\
> > > >+							\
> > > >+	__r;						\
> > > >+})
> > > >+
> > > >+int x86_virt_get_cpu(int feat)
> > > >+{
> > > >+	int r;
> > > >+
> > > >+	if (!x86_virt_feature || x86_virt_feature != feat)
> > > >+		return -EOPNOTSUPP;
> > > >+
> > > >+	if (this_cpu_inc_return(virtualization_nr_users) > 1)
> > > >+		return 0;
> > > 
> > > Should we assert that preemption is disabled? Calling this API when preemption
> > > is enabled is wrong.
> > > 
> > > Maybe use __this_cpu_inc_return(), which already verifies preemption status.
> 
> I always forget that the double-underscores have the checks.  
> 
> > Is it better we explicitly assert the preemption for x86_virt_get_cpu()
> > rather than embed the check in __this_cpu_inc_return()? We are not just
> > protecting the racing for the reference counter. We should ensure the
> > "counter increase + x86_virt_call(get_cpu)" can't be preempted.
> 
> I don't have a strong preference.  Using __this_cpu_inc_return() without any
> nearby preemption_{enable,disable}() calls makes it quite clears that preemption
> is expected to be disabled by the caller.  But I'm also ok being explicit.

Looking into __this_cpu_inc_return(), it finally calls
check_preemption_disabled() which doesn't strictly requires preemption.
It only ensures the context doesn't switch to another CPU. If the caller
is in cpuhp context, preemption is possible.

But in this x86_virt_get_cpu(), we need to ensure preemption disabled,
otherwise caller A increases counter but hasn't do actual VMXON yet and
get preempted. Caller B opts in and get the wrong info that VMX is
already on, and fails on following vmx operations.

On a second thought, maybe we disable preemption inside
x86_virt_get_cpu() to protect the counter-vmxon racing, this is pure
internal thing for this kAPI.

Thanks,
Yilun