x86emul: support VAES insns

As to the feature dependency adjustment, just like for VPCLMULQDQ while
strictly speaking AVX is a sufficient prereq (to have YMM registers),
256-bit vectors of integers have got fully introduced with AVX2 only.

A new test case (also covering AESNI) will be added to the harness by a
subsequent patch.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citirx.com>

diff --git a/tools/tests/x86_emulator/evex-disp8.c b/tools/tests/x86_emulator/evex-disp8.c
index cc10a563d328978424a98e8781750248ddd41b6f..5cc985a940735a217887bdd13de15ee33a52c1df 100644 (file)
--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -591,6 +591,18 @@ static const struct test avx512_vpopcntdq_all[] = {
     INSN(popcnt, 66, 0f38, 55, vl, dq, vl)
 };
 
+/*
+ * The uses of b in this table are simply (one of) the shortest form(s) of
+ * saying "no broadcast" without introducing a 128-bit granularity enumerator.
+ * Due to all of the insns being WIG, w, d_nb, and q_nb would all also fit.
+ */
+static const struct test vaes_all[] = {
+    INSN(aesdec,     66, 0f38, de, vl, b, vl),
+    INSN(aesdeclast, 66, 0f38, df, vl, b, vl),
+    INSN(aesenc,     66, 0f38, dc, vl, b, vl),
+    INSN(aesenclast, 66, 0f38, dd, vl, b, vl),
+};
+
 static const struct test vpclmulqdq_all[] = {
     INSN(pclmulqdq, 66, 0f3a, 44, vl, q_nb, vl)
 };
@@ -975,6 +987,7 @@ void evex_disp8_test(void *instr, struct x86_emulate_ctxt *ctxt,
 
     if ( cpu_has_avx512f )
     {
+        RUN(vaes, all);
         RUN(vpclmulqdq, all);
     }
 }

diff --git a/tools/tests/x86_emulator/x86-emulate.h b/tools/tests/x86_emulator/x86-emulate.h
index 2cc04f752b7f4b870a091f229d3e6cc8f6af813b..bbb60bc65e6ccb41f81d4a5ede5821a9b24b3f7b 100644 (file)
--- a/tools/tests/x86_emulator/x86-emulate.h
+++ b/tools/tests/x86_emulator/x86-emulate.h
@@ -144,6 +144,7 @@ static inline bool xcr0_mask(uint64_t mask)
 #define cpu_has_avx512vl  (cp.feat.avx512vl && xcr0_mask(0xe6))
 #define cpu_has_avx512_vbmi (cp.feat.avx512_vbmi && xcr0_mask(0xe6))
 #define cpu_has_avx512_vbmi2 (cp.feat.avx512_vbmi2 && xcr0_mask(0xe6))
+#define cpu_has_vaes      (cp.feat.vaes && xcr0_mask(6))
 #define cpu_has_vpclmulqdq (cp.feat.vpclmulqdq && xcr0_mask(6))
 #define cpu_has_avx512_vnni (cp.feat.avx512_vnni && xcr0_mask(0xe6))
 #define cpu_has_avx512_bitalg (cp.feat.avx512_bitalg && xcr0_mask(0xe6))

diff --git a/xen/arch/x86/x86_emulate/x86_emulate.c b/xen/arch/x86/x86_emulate/x86_emulate.c
index 89098520944119c2482508f80cb7c5aa3699db55..245c900351b1f1927c2e42e5e989df2ed82b8b04 100644 (file)
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -541,7 +541,7 @@ static const struct ext0f38_table {
     [0xcc] = { .simd_size = simd_packed_fp, .two_op = 1, .d8s = d8s_vl },
     [0xcd] = { .simd_size = simd_scalar_vexw, .d8s = d8s_dq },
     [0xdb] = { .simd_size = simd_packed_int, .two_op = 1 },
-    [0xdc ... 0xdf] = { .simd_size = simd_packed_int },
+    [0xdc ... 0xdf] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
     [0xf0] = { .two_op = 1 },
     [0xf1] = { .to_mem = 1, .two_op = 1 },
     [0xf2 ... 0xf3] = {},
@@ -1890,6 +1890,7 @@ in_protmode(
 #define vcpu_has_avx512vl()    (ctxt->cpuid->feat.avx512vl)
 #define vcpu_has_avx512_vbmi() (ctxt->cpuid->feat.avx512_vbmi)
 #define vcpu_has_avx512_vbmi2() (ctxt->cpuid->feat.avx512_vbmi2)
+#define vcpu_has_vaes()        (ctxt->cpuid->feat.vaes)
 #define vcpu_has_vpclmulqdq()  (ctxt->cpuid->feat.vpclmulqdq)
 #define vcpu_has_avx512_vnni() (ctxt->cpuid->feat.avx512_vnni)
 #define vcpu_has_avx512_bitalg() (ctxt->cpuid->feat.avx512_bitalg)
@@ -8911,13 +8912,9 @@ x86_emulate(
     case X86EMUL_OPC_66(0x0f38, 0xdb):     /* aesimc xmm/m128,xmm */
     case X86EMUL_OPC_VEX_66(0x0f38, 0xdb): /* vaesimc xmm/m128,xmm */
     case X86EMUL_OPC_66(0x0f38, 0xdc):     /* aesenc xmm/m128,xmm,xmm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0xdc): /* vaesenc xmm/m128,xmm,xmm */
     case X86EMUL_OPC_66(0x0f38, 0xdd):     /* aesenclast xmm/m128,xmm,xmm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0xdd): /* vaesenclast xmm/m128,xmm,xmm */
     case X86EMUL_OPC_66(0x0f38, 0xde):     /* aesdec xmm/m128,xmm,xmm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0xde): /* vaesdec xmm/m128,xmm,xmm */
     case X86EMUL_OPC_66(0x0f38, 0xdf):     /* aesdeclast xmm/m128,xmm,xmm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0xdf): /* vaesdeclast xmm/m128,xmm,xmm */
         host_and_vcpu_must_have(aesni);
         if ( vex.opcx == vex_none )
             goto simd_0f38_common;
@@ -9643,6 +9640,24 @@ x86_emulate(
         host_and_vcpu_must_have(avx512er);
         goto simd_zmm_scalar_sae;
 
+    case X86EMUL_OPC_VEX_66(0x0f38, 0xdc):  /* vaesenc {x,y}mm/mem,{x,y}mm,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0xdd):  /* vaesenclast {x,y}mm/mem,{x,y}mm,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0xde):  /* vaesdec {x,y}mm/mem,{x,y}mm,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0xdf):  /* vaesdeclast {x,y}mm/mem,{x,y}mm,{x,y}mm */
+        if ( !vex.l )
+            host_and_vcpu_must_have(aesni);
+        else
+            host_and_vcpu_must_have(vaes);
+        goto simd_0f_avx;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xdc): /* vaesenc [xyz]mm/mem,[xyz]mm,[xyz]mm */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xdd): /* vaesenclast [xyz]mm/mem,[xyz]mm,[xyz]mm */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xde): /* vaesdec [xyz]mm/mem,[xyz]mm,[xyz]mm */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xdf): /* vaesdeclast [xyz]mm/mem,[xyz]mm,[xyz]mm */
+        host_and_vcpu_must_have(vaes);
+        generate_exception_if(evex.brs || evex.opmsk, EXC_UD);
+        goto avx512f_no_sae;
+
     case X86EMUL_OPC(0x0f38, 0xf0): /* movbe m,r */
     case X86EMUL_OPC(0x0f38, 0xf1): /* movbe r,m */
         vcpu_must_have(movbe);

diff --git a/xen/include/asm-x86/cpufeature.h b/xen/include/asm-x86/cpufeature.h
index eab9ecaf763456a5b8c7f4ccba485fe61a14c7c6..85f8382eb0051bd4a8924ca12f58a44faea07f70 100644 (file)
--- a/xen/include/asm-x86/cpufeature.h
+++ b/xen/include/asm-x86/cpufeature.h
@@ -111,6 +111,7 @@
 /* CPUID level 0x00000007:0.ecx */
 #define cpu_has_avx512_vbmi     boot_cpu_has(X86_FEATURE_AVX512_VBMI)
 #define cpu_has_avx512_vbmi2    boot_cpu_has(X86_FEATURE_AVX512_VBMI2)
+#define cpu_has_vaes            boot_cpu_has(X86_FEATURE_VAES)
 #define cpu_has_vpclmulqdq      boot_cpu_has(X86_FEATURE_VPCLMULQDQ)
 #define cpu_has_avx512_vnni     boot_cpu_has(X86_FEATURE_AVX512_VNNI)
 #define cpu_has_avx512_bitalg   boot_cpu_has(X86_FEATURE_AVX512_BITALG)

diff --git a/xen/include/public/arch-x86/cpufeatureset.h b/xen/include/public/arch-x86/cpufeatureset.h
index 56c9a75f3c7ac93c763c5d6fa255e82574b96a57..676722fc5978cfa07f87e4a89231dcc451ed32a9 100644 (file)
--- a/xen/include/public/arch-x86/cpufeatureset.h
+++ b/xen/include/public/arch-x86/cpufeatureset.h
@@ -229,6 +229,7 @@ XEN_CPUFEATURE(UMIP,          6*32+ 2) /*S  User Mode Instruction Prevention */
 XEN_CPUFEATURE(PKU,           6*32+ 3) /*H  Protection Keys for Userspace */
 XEN_CPUFEATURE(OSPKE,         6*32+ 4) /*!  OS Protection Keys Enable */
 XEN_CPUFEATURE(AVX512_VBMI2,  6*32+ 6) /*A  Additional AVX-512 Vector Byte Manipulation Instrs */
+XEN_CPUFEATURE(VAES,          6*32+ 9) /*A  Vector AES Instrs */
 XEN_CPUFEATURE(VPCLMULQDQ,    6*32+10) /*A  Vector Carry-less Multiplication Instrs */
 XEN_CPUFEATURE(AVX512_VNNI,   6*32+11) /*A  Vector Neural Network Instrs */
 XEN_CPUFEATURE(AVX512_BITALG, 6*32+12) /*A  Support for VPOPCNT[B,W] and VPSHUFBITQMB */

diff --git a/xen/tools/gen-cpuid.py b/xen/tools/gen-cpuid.py
index 9df8ae528f4b5180c567238f0abb59aaffe9d23b..e5ac93304dcf5acfc4f460a4f89ca679cbd681f9 100755 (executable)
--- a/xen/tools/gen-cpuid.py
+++ b/xen/tools/gen-cpuid.py
@@ -256,7 +256,7 @@ def crunch_numbers(state):
         # feature flags.  If want to use AVX512, AVX2 must be supported and
         # enabled.  Certain later extensions, acting on 256-bit vectors of
         # integers, better depend on AVX2 than AVX.
-        AVX2: [AVX512F, VPCLMULQDQ],
+        AVX2: [AVX512F, VAES, VPCLMULQDQ],
 
         # AVX512F is taken to mean hardware support for 512bit registers
         # (which in practice depends on the EVEX prefix to encode) as well
@@ -274,6 +274,7 @@ def crunch_numbers(state):
         # Extensions with VEX/EVEX encodings keyed to a separate feature
         # flag are made dependents of their respective legacy feature.
         PCLMULQDQ: [VPCLMULQDQ],
+        AESNI: [VAES],
 
         # The features:
         #   * Single Thread Indirect Branch Predictors