linux-kernel - Re: [PATCH 4/4] KVM: x86: forcibly leave nested mode on vCPU reset

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]

Date:   Thu, 20 Oct 2022 15:33:50 +0000
From:   Sean Christopherson <seanjc@...gle.com>
To:     Maxim Levitsky <mlevitsk@...hat.com>
Cc:     kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
        Paolo Bonzini <pbonzini@...hat.com>,
        Borislav Petkov <bp@...en8.de>, Ingo Molnar <mingo@...hat.com>,
        x86@...nel.org, Thomas Gleixner <tglx@...utronix.de>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        "H. Peter Anvin" <hpa@...or.com>, stable@...r.kernel.org
Subject: Re: [PATCH 4/4] KVM: x86: forcibly leave nested mode on vCPU reset

On Thu, Oct 20, 2022, Maxim Levitsky wrote:
> While not obivous, kvm_vcpu_reset leaves the nested mode by

Please add () when referencing function, and wrap closer to ~75 chars.

> clearing 'vcpu->arch.hflags' but it does so without all the
> required housekeeping.
> 
> This makes SVM and VMX continue to use vmcs02/vmcb02 while

This bug should be impossible to hit on VMX as INIT and TRIPLE_FAULT unconditionally
cause VM-Exit, i.e. will always be forwarded to L1.

> the cpu is not in nested mode.

Can you add a blurb to call out exactly how this bug can be triggered?  Doesn't
take much effort to suss out the "how", but it'd be nice to capture that info in
the changelog.

> In particular, in SVM code, it makes the 'svm_free_nested'
> free the vmcb02, while still in use, which later triggers
> use after free and a kernel crash.
> 
> This issue is assigned CVE-2022-3344
> 
> Cc: stable@...r.kernel.org
> Signed-off-by: Maxim Levitsky <mlevitsk@...hat.com>
> ---
>  arch/x86/kvm/x86.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index d86a8aae1471d3..313c4a6dc65e45 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -11931,6 +11931,7 @@ void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
>  	WARN_ON_ONCE(!init_event &&
>  		     (old_cr0 || kvm_read_cr3(vcpu) || kvm_read_cr4(vcpu)));
>  
> +	kvm_leave_nested(vcpu);

Not a big deal, especially if/when nested_ops are turned into static_calls, but
at the same time it's quite easy to do:

	if (is_guest_mode(vcpu))
		kvm_leave_nested(vcpu);

I think it's worth adding a comment explaining how this can happen, and to also
call out that EFER is cleared on INIT, i.e. that virtualization is disabled due
to EFER.SVME=0.  Unsurprisingly, I don't see anything in the APM that explicitly
states what happens if INIT occurs in guest mode, i.e. it's not immediately obvious
that forcing the vCPU back to L1 is architecturally correct.


>  	kvm_lapic_reset(vcpu, init_event);
>  
>  	vcpu->arch.hflags = 0;

Maybe add a WARN above this to try and detect other potential issues?  Kinda silly,
but it'd at least help draw attention to the importance of hflags.

E.g. this?

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 4bd5f8a751de..c50fa0751a0b 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -11915,6 +11915,15 @@ void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
        unsigned long old_cr0 = kvm_read_cr0(vcpu);
        unsigned long new_cr0;
 
+       /*
+        * SVM doesn't unconditionally VM-Exit on INIT and SHUTDOWN, thus it's
+        * possible to INIT the vCPU while L2 is active.  Force the vCPU back
+        * into L1 as EFER.SVME is cleared on INIT (along with all other EFER
+        * bits), i.e. virtualization is disabled.
+        */
+       if (is_guest_mode(vcpu))
+               kvm_leave_nested(vcpu);
+
        /*
         * Several of the "set" flows, e.g. ->set_cr0(), read other registers
         * to handle side effects.  RESET emulation hits those flows and relies
@@ -11927,6 +11936,7 @@ void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
 
        kvm_lapic_reset(vcpu, init_event);
 
+       WARN_ON_ONCE(is_guest_mode(vcpu) || is_smm(vcpu));
        vcpu->arch.hflags = 0;
 
        vcpu->arch.smi_pending = 0;