qemu-kvm/kvm-i386-tdx-Add-TDX-fixed1-bits-to-supported-CPUIDs.patch

From 70f7099dfd895a7ffeee3f66e188c1853d3885d6 Mon Sep 17 00:00:00 2001
From: Paolo Bonzini <pbonzini@redhat.com>
Date: Fri, 18 Jul 2025 18:03:48 +0200
Subject: [PATCH 074/115] i386/tdx: Add TDX fixed1 bits to supported CPUIDs

RH-Author: Paolo Bonzini <pbonzini@redhat.com>
RH-MergeRequest: 391: TDX support, including attestation and device assignment
RH-Jira: RHEL-15710 RHEL-20798 RHEL-49728
RH-Acked-by: Yash Mankad <None>
RH-Acked-by: Peter Xu <peterx@redhat.com>
RH-Acked-by: David Hildenbrand <david@redhat.com>
RH-Commit: [74/115] 81fb39c826363f892e46386883c69c0e862d7850 (bonzini/rhel-qemu-kvm)

TDX architecture forcibly sets some CPUID bits for TD guest that VMM
cannot disable it. They are fixed1 bits.

Fixed1 bits are not covered by tdx_caps.cpuid (which only contains the
directly configurable bits), while fixed1 bits are supported for TD guest
obviously.

Add fixed1 bits to tdx_supported_cpuid. Besides, set all the fixed1
bits to the initial set of KVM's support since KVM might not report them
as supported.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Link: https://lore.kernel.org/r/20250508150002.689633-46-xiaoyao.li@intel.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 0ba06e46d09b84a2cb97a268da5576aaca3a24ca)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/cpu.h          |   2 +
 target/i386/kvm/kvm_i386.h |   7 ++
 target/i386/kvm/tdx.c      | 134 +++++++++++++++++++++++++++++++++++++
 target/i386/sev.c          |   8 ---
 4 files changed, 143 insertions(+), 8 deletions(-)

diff --git a/target/i386/cpu.h b/target/i386/cpu.h
index 601e828577..529f24df00 100644
--- a/target/i386/cpu.h
+++ b/target/i386/cpu.h
@@ -924,6 +924,8 @@ uint64_t x86_cpu_get_supported_feature_word(X86CPU *cpu, FeatureWord w);
 #define CPUID_7_0_EDX_FSRM              (1U << 4)
 /* AVX512 Vector Pair Intersection to a Pair of Mask Registers */
 #define CPUID_7_0_EDX_AVX512_VP2INTERSECT (1U << 8)
+ /* "md_clear" VERW clears CPU buffers */
+#define CPUID_7_0_EDX_MD_CLEAR          (1U << 10)
 /* SERIALIZE instruction */
 #define CPUID_7_0_EDX_SERIALIZE         (1U << 14)
 /* TSX Suspend Load Address Tracking instruction */
diff --git a/target/i386/kvm/kvm_i386.h b/target/i386/kvm/kvm_i386.h
index 797610496a..f1d55d5b75 100644
--- a/target/i386/kvm/kvm_i386.h
+++ b/target/i386/kvm/kvm_i386.h
@@ -44,6 +44,13 @@ void kvm_request_xsave_components(X86CPU *cpu, uint64_t mask);
 
 #ifdef CONFIG_KVM
 
+#include <linux/kvm.h>
+
+typedef struct KvmCpuidInfo {
+    struct kvm_cpuid2 cpuid;
+    struct kvm_cpuid_entry2 entries[KVM_MAX_CPUID_ENTRIES];
+} KvmCpuidInfo;
+
 bool kvm_is_vm_type_supported(int type);
 bool kvm_has_adjust_clock_stable(void);
 bool kvm_has_exception_payload(void);
diff --git a/target/i386/kvm/tdx.c b/target/i386/kvm/tdx.c
index 4949d01f22..6fa30c3ec4 100644
--- a/target/i386/kvm/tdx.c
+++ b/target/i386/kvm/tdx.c
@@ -367,6 +367,133 @@ static Notifier tdx_machine_done_notify = {
     .notify = tdx_finalize_vm,
 };
 
+/*
+ * Some CPUID bits change from fixed1 to configurable bits when TDX module
+ * supports TDX_FEATURES0.VE_REDUCTION. e.g., MCA/MCE/MTRR/CORE_CAPABILITY.
+ *
+ * To make QEMU work with all the versions of TDX module, keep the fixed1 bits
+ * here if they are ever fixed1 bits in any of the version though not fixed1 in
+ * the latest version. Otherwise, with the older version of TDX module, QEMU may
+ * treat the fixed1 bit as unsupported.
+ *
+ * For newer TDX module, it does no harm to keep them in tdx_fixed1_bits even
+ * though they changed to configurable bits. Because tdx_fixed1_bits is used to
+ * setup the supported bits.
+ */
+KvmCpuidInfo tdx_fixed1_bits = {
+    .cpuid.nent = 8,
+    .entries[0] = {
+        .function = 0x1,
+        .index = 0,
+        .ecx = CPUID_EXT_SSE3 | CPUID_EXT_PCLMULQDQ | CPUID_EXT_DTES64 |
+               CPUID_EXT_DSCPL | CPUID_EXT_SSSE3 | CPUID_EXT_CX16 |
+               CPUID_EXT_PDCM | CPUID_EXT_PCID | CPUID_EXT_SSE41 |
+               CPUID_EXT_SSE42 | CPUID_EXT_X2APIC | CPUID_EXT_MOVBE |
+               CPUID_EXT_POPCNT | CPUID_EXT_AES | CPUID_EXT_XSAVE |
+               CPUID_EXT_RDRAND | CPUID_EXT_HYPERVISOR,
+        .edx = CPUID_FP87 | CPUID_VME | CPUID_DE | CPUID_PSE | CPUID_TSC |
+               CPUID_MSR | CPUID_PAE | CPUID_MCE | CPUID_CX8 | CPUID_APIC |
+               CPUID_SEP | CPUID_MTRR | CPUID_PGE | CPUID_MCA | CPUID_CMOV |
+               CPUID_PAT | CPUID_CLFLUSH | CPUID_DTS | CPUID_MMX | CPUID_FXSR |
+               CPUID_SSE | CPUID_SSE2,
+    },
+    .entries[1] = {
+        .function = 0x6,
+        .index = 0,
+        .eax = CPUID_6_EAX_ARAT,
+    },
+    .entries[2] = {
+        .function = 0x7,
+        .index = 0,
+        .flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX,
+        .ebx = CPUID_7_0_EBX_FSGSBASE | CPUID_7_0_EBX_FDP_EXCPTN_ONLY |
+               CPUID_7_0_EBX_SMEP | CPUID_7_0_EBX_INVPCID |
+               CPUID_7_0_EBX_ZERO_FCS_FDS | CPUID_7_0_EBX_RDSEED |
+               CPUID_7_0_EBX_SMAP | CPUID_7_0_EBX_CLFLUSHOPT |
+               CPUID_7_0_EBX_CLWB | CPUID_7_0_EBX_SHA_NI,
+        .ecx = CPUID_7_0_ECX_BUS_LOCK_DETECT | CPUID_7_0_ECX_MOVDIRI |
+               CPUID_7_0_ECX_MOVDIR64B,
+        .edx = CPUID_7_0_EDX_MD_CLEAR | CPUID_7_0_EDX_SPEC_CTRL |
+               CPUID_7_0_EDX_STIBP | CPUID_7_0_EDX_FLUSH_L1D |
+               CPUID_7_0_EDX_ARCH_CAPABILITIES | CPUID_7_0_EDX_CORE_CAPABILITY |
+               CPUID_7_0_EDX_SPEC_CTRL_SSBD,
+    },
+    .entries[3] = {
+        .function = 0x7,
+        .index = 2,
+        .flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX,
+        .edx = CPUID_7_2_EDX_PSFD | CPUID_7_2_EDX_IPRED_CTRL |
+               CPUID_7_2_EDX_RRSBA_CTRL | CPUID_7_2_EDX_BHI_CTRL,
+    },
+    .entries[4] = {
+        .function = 0xD,
+        .index = 0,
+        .flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX,
+        .eax = XSTATE_FP_MASK | XSTATE_SSE_MASK,
+    },
+    .entries[5] = {
+        .function = 0xD,
+        .index = 1,
+        .flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX,
+        .eax = CPUID_XSAVE_XSAVEOPT | CPUID_XSAVE_XSAVEC|
+               CPUID_XSAVE_XGETBV1 | CPUID_XSAVE_XSAVES,
+    },
+    .entries[6] = {
+        .function = 0x80000001,
+        .index = 0,
+        .ecx = CPUID_EXT3_LAHF_LM | CPUID_EXT3_ABM | CPUID_EXT3_3DNOWPREFETCH,
+        /*
+         * Strictly speaking, SYSCALL is not fixed1 bit since it depends on
+         * the CPU to be in 64-bit mode. But here fixed1 is used to serve the
+         * purpose of supported bits for TDX. In this sense, SYACALL is always
+         * supported.
+         */
+        .edx = CPUID_EXT2_SYSCALL | CPUID_EXT2_NX | CPUID_EXT2_PDPE1GB |
+               CPUID_EXT2_RDTSCP | CPUID_EXT2_LM,
+    },
+    .entries[7] = {
+        .function = 0x80000007,
+        .index = 0,
+        .edx = CPUID_APM_INVTSC,
+    },
+};
+
+static struct kvm_cpuid_entry2 *find_in_supported_entry(uint32_t function,
+                                                        uint32_t index)
+{
+    struct kvm_cpuid_entry2 *e;
+
+    e = cpuid_find_entry(tdx_supported_cpuid, function, index);
+    if (!e) {
+        if (tdx_supported_cpuid->nent >= KVM_MAX_CPUID_ENTRIES) {
+            error_report("tdx_supported_cpuid requries more space than %d entries",
+                          KVM_MAX_CPUID_ENTRIES);
+            exit(1);
+        }
+        e = &tdx_supported_cpuid->entries[tdx_supported_cpuid->nent++];
+        e->function = function;
+        e->index = index;
+    }
+
+    return e;
+}
+
+static void tdx_add_supported_cpuid_by_fixed1_bits(void)
+{
+    struct kvm_cpuid_entry2 *e, *e1;
+    int i;
+
+    for (i = 0; i < tdx_fixed1_bits.cpuid.nent; i++) {
+        e = &tdx_fixed1_bits.entries[i];
+
+        e1 = find_in_supported_entry(e->function, e->index);
+        e1->eax |= e->eax;
+        e1->ebx |= e->ebx;
+        e1->ecx |= e->ecx;
+        e1->edx |= e->edx;
+    }
+}
+
 static void tdx_setup_supported_cpuid(void)
 {
     if (tdx_supported_cpuid) {
@@ -379,6 +506,8 @@ static void tdx_setup_supported_cpuid(void)
     memcpy(tdx_supported_cpuid->entries, tdx_caps->cpuid.entries,
            tdx_caps->cpuid.nent * sizeof(struct kvm_cpuid_entry2));
     tdx_supported_cpuid->nent = tdx_caps->cpuid.nent;
+
+    tdx_add_supported_cpuid_by_fixed1_bits();
 }
 
 static int tdx_kvm_init(ConfidentialGuestSupport *cgs, Error **errp)
@@ -463,6 +592,11 @@ static uint32_t tdx_adjust_cpuid_features(X86ConfidentialGuest *cg,
 {
     struct kvm_cpuid_entry2 *e;
 
+    e = cpuid_find_entry(&tdx_fixed1_bits.cpuid, feature, index);
+    if (e) {
+        value |= cpuid_entry_get_reg(e, reg);
+    }
+
     if (is_feature_word_cpuid(feature, index, reg)) {
         e = cpuid_find_entry(tdx_supported_cpuid, feature, index);
         if (e) {
diff --git a/target/i386/sev.c b/target/i386/sev.c
index 24fcd078fc..edbad9bb92 100644
--- a/target/i386/sev.c
+++ b/target/i386/sev.c
@@ -211,14 +211,6 @@ static const char *const sev_fw_errlist[] = {
 
 #define SEV_FW_MAX_ERROR      ARRAY_SIZE(sev_fw_errlist)
 
-/* <linux/kvm.h> doesn't expose this, so re-use the max from kvm.c */
-#define KVM_MAX_CPUID_ENTRIES 100
-
-typedef struct KvmCpuidInfo {
-    struct kvm_cpuid2 cpuid;
-    struct kvm_cpuid_entry2 entries[KVM_MAX_CPUID_ENTRIES];
-} KvmCpuidInfo;
-
 #define SNP_CPUID_FUNCTION_MAXCOUNT 64
 #define SNP_CPUID_FUNCTION_UNKNOWN 0xFFFFFFFF
 
-- 
2.50.1