linux-kernel - [PATCH v2 0/2] KVM: SVM: Fix a bug where TSC

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-next>] [day] [month] [year] [list]

Message-ID: <20250919213806.1582673-1-seanjc@google.com>
Date: Fri, 19 Sep 2025 14:38:04 -0700
From: Sean Christopherson <seanjc@...gle.com>
To: Sean Christopherson <seanjc@...gle.com>, Paolo Bonzini <pbonzini@...hat.com>
Cc: kvm@...r.kernel.org, linux-kernel@...r.kernel.org, 
	Hou Wenlong <houwenlong.hwl@...group.com>, Lai Jiangshan <jiangshan.ljs@...group.com>
Subject: [PATCH v2 0/2] KVM: SVM: Fix a bug where TSC_AUX can get clobbered

v2 of Hou's series to fix a bug where an SEV-ES vCPU running on the same
pCPU as a non-SEV-ES vCPU could clobber TSC_AUX due to loading the host's
TSC_AUX on #VMEXIT, as opposed to restoring whatever was in hardware at the
time of VMRUN.

I tried to test this by hacking sev_smoke_test, but unfortunately I don't
have a machine that has SEV-ES *and* TSC_AUX virtualization.  *sigh*

diff --git a/tools/testing/selftests/kvm/x86/sev_smoke_test.c b/tools/testing/selftests/kvm/x86/sev_smoke_test.c
index 77256c89bb8d..73530a01a3b5 100644
--- a/tools/testing/selftests/kvm/x86/sev_smoke_test.c
+++ b/tools/testing/selftests/kvm/x86/sev_smoke_test.c
@@ -16,6 +16,12 @@
 
 #define XFEATURE_MASK_X87_AVX (XFEATURE_MASK_FP | XFEATURE_MASK_SSE | XFEATURE_MASK_YMM)
 
+static uint64_t guest_sev_es_get_info(void)
+{
+       wrmsr(MSR_AMD64_SEV_ES_GHCB, GHCB_MSR_TERM_REQ);
+       return rdmsr(MSR_AMD64_SEV_ES_GHCB);
+}
+
 static void guest_snp_code(void)
 {
        uint64_t sev_msr = rdmsr(MSR_AMD64_SEV);
@@ -34,6 +40,10 @@ static void guest_sev_es_code(void)
        GUEST_ASSERT(rdmsr(MSR_AMD64_SEV) & MSR_AMD64_SEV_ENABLED);
        GUEST_ASSERT(rdmsr(MSR_AMD64_SEV) & MSR_AMD64_SEV_ES_ENABLED);
 
+       wrmsr(MSR_TSC_AUX, 0x12345678);
+       guest_sev_es_get_info();
+       GUEST_ASSERT(rdmsr(MSR_TSC_AUX) == 0x12345678);
+
        /*
         * TODO: Add GHCB and ucall support for SEV-ES guests.  For now, simply
         * force "termination" to signal "done" via the GHCB MSR protocol.

v2:
 - Drop "cache" from the user_return API.
 - Handle the SEV-ES case in SEV-ES code.
 - Tag everything for stable@.
 - Massage changelog to avoid talking about the host's value and instead
   focus on failing to restore what KVM thinks is in hardware.

v1: https://lore.kernel.org/all/05a018a6997407080b3b7921ba692aa69a720f07.1758166596.git.houwenlong.hwl@antgroup.com

Hou Wenlong (2):
  KVM: x86: Add helper to retrieve current value of user return MSR
  KVM: SVM: Re-load current, not host, TSC_AUX on #VMEXIT from SEV-ES
    guest

 arch/x86/include/asm/kvm_host.h |  1 +
 arch/x86/kvm/svm/sev.c          | 14 +++++++++++++-
 arch/x86/kvm/svm/svm.c          | 26 +++++++-------------------
 arch/x86/kvm/svm/svm.h          |  4 +++-
 arch/x86/kvm/x86.c              |  6 ++++++
 5 files changed, 30 insertions(+), 21 deletions(-)


base-commit: c8fbf7ceb2ae3f64b0c377c8c21f6df577a13eb4
-- 
2.51.0.470.ga7dc726c21-goog