[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250919223258.1604852-21-seanjc@google.com>
Date: Fri, 19 Sep 2025 15:32:27 -0700
From: Sean Christopherson <seanjc@...gle.com>
To: Paolo Bonzini <pbonzini@...hat.com>, Sean Christopherson <seanjc@...gle.com>
Cc: kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
Tom Lendacky <thomas.lendacky@....com>, Mathias Krause <minipli@...ecurity.net>,
John Allen <john.allen@....com>, Rick Edgecombe <rick.p.edgecombe@...el.com>,
Chao Gao <chao.gao@...el.com>, Binbin Wu <binbin.wu@...ux.intel.com>,
Xiaoyao Li <xiaoyao.li@...el.com>, Maxim Levitsky <mlevitsk@...hat.com>,
Zhang Yi Z <yi.z.zhang@...ux.intel.com>, Xin Li <xin@...or.com>
Subject: [PATCH v16 20/51] KVM: x86: Emulate SSP[63:32]!=0 #GP(0) for FAR JMP
to 32-bit mode
Emulate the Shadow Stack restriction that the current SSP must be a 32-bit
value on a FAR JMP from 64-bit mode to compatibility mode. From the SDM's
pseudocode for FAR JMP:
IF ShadowStackEnabled(CPL)
IF (IA32_EFER.LMA and DEST(segment selector).L) = 0
(* If target is legacy or compatibility mode then the SSP must be in low 4GB *)
IF (SSP & 0xFFFFFFFF00000000 != 0); THEN
#GP(0);
FI;
FI;
FI;
Note, only the current CPL needs to be considered, as FAR JMP can't be
used for inter-privilege level transfers, and KVM rejects emulation of all
other far branch instructions when Shadow Stacks are enabled.
To give the emulator access to GUEST_SSP, special case handling
MSR_KVM_INTERNAL_GUEST_SSP in emulator_get_msr() to treat the access as a
host access (KVM doesn't allow guest accesses to internal "MSRs"). The
->get_msr() API is only used for implicit accesses from the emulator, i.e.
is only used with hardcoded MSR indices, and so any access to
MSR_KVM_INTERNAL_GUEST_SSP is guaranteed to be from KVM, i.e. not from the
guest via RDMSR.
Signed-off-by: Sean Christopherson <seanjc@...gle.com>
---
arch/x86/kvm/emulate.c | 35 +++++++++++++++++++++++++++++++++++
arch/x86/kvm/x86.c | 9 +++++++++
2 files changed, 44 insertions(+)
diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index dc0249929cbf..5c5fb6a6f7f9 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -1605,6 +1605,37 @@ static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
return linear_write_system(ctxt, addr, desc, sizeof(*desc));
}
+static bool emulator_is_ssp_invalid(struct x86_emulate_ctxt *ctxt, u8 cpl)
+{
+ const u32 MSR_IA32_X_CET = cpl == 3 ? MSR_IA32_U_CET : MSR_IA32_S_CET;
+ u64 efer = 0, cet = 0, ssp = 0;
+
+ if (!(ctxt->ops->get_cr(ctxt, 4) & X86_CR4_CET))
+ return false;
+
+ if (ctxt->ops->get_msr(ctxt, MSR_EFER, &efer))
+ return true;
+
+ /* SSP is guaranteed to be valid if the vCPU was already in 32-bit mode. */
+ if (!(efer & EFER_LMA))
+ return false;
+
+ if (ctxt->ops->get_msr(ctxt, MSR_IA32_X_CET, &cet))
+ return true;
+
+ if (!(cet & CET_SHSTK_EN))
+ return false;
+
+ if (ctxt->ops->get_msr(ctxt, MSR_KVM_INTERNAL_GUEST_SSP, &ssp))
+ return true;
+
+ /*
+ * On transfer from 64-bit mode to compatibility mode, SSP[63:32] must
+ * be 0, i.e. SSP must be a 32-bit value outside of 64-bit mode.
+ */
+ return ssp >> 32;
+}
+
static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
u16 selector, int seg, u8 cpl,
enum x86_transfer_type transfer,
@@ -1745,6 +1776,10 @@ static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
if (efer & EFER_LMA)
goto exception;
}
+ if (!seg_desc.l && emulator_is_ssp_invalid(ctxt, cpl)) {
+ err_code = 0;
+ goto exception;
+ }
/* CS(RPL) <- CPL */
selector = (selector & 0xfffc) | cpl;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 0c060e506f9d..40596fc5142e 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -8741,6 +8741,15 @@ static int emulator_set_msr_with_filter(struct x86_emulate_ctxt *ctxt,
static int emulator_get_msr(struct x86_emulate_ctxt *ctxt,
u32 msr_index, u64 *pdata)
{
+ /*
+ * Treat emulator accesses to the current shadow stack pointer as host-
+ * initiated, as they aren't true MSR accesses (SSP is a "just a reg"),
+ * and this API is used only for implicit accesses, i.e. not RDMSR, and
+ * so the index is fully KVM-controlled.
+ */
+ if (unlikely(msr_index == MSR_KVM_INTERNAL_GUEST_SSP))
+ return kvm_msr_read(emul_to_vcpu(ctxt), msr_index, pdata);
+
return __kvm_emulate_msr_read(emul_to_vcpu(ctxt), msr_index, pdata);
}
--
2.51.0.470.ga7dc726c21-goog
Powered by blists - more mailing lists