@@ -591,6 +591,18 @@ static const struct test avx512_vpopcntd
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
if ( cpu_has_avx512f )
{
+ RUN(vaes, all);
RUN(vpclmulqdq, all);
}
}
@@ -144,6 +144,7 @@ static inline bool xcr0_mask(uint64_t ma
#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))
@@ -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] = {},
@@ -1922,6 +1922,7 @@ static bool vcpu_has(
#define vcpu_has_avx512vl() vcpu_has( 7, EBX, 31, ctxt, ops)
#define vcpu_has_avx512_vbmi() vcpu_has( 7, ECX, 1, ctxt, ops)
#define vcpu_has_avx512_vbmi2() vcpu_has( 7, ECX, 6, ctxt, ops)
+#define vcpu_has_vaes() vcpu_has( 7, ECX, 9, ctxt, ops)
#define vcpu_has_vpclmulqdq() vcpu_has( 7, ECX, 10, ctxt, ops)
#define vcpu_has_avx512_vnni() vcpu_has( 7, ECX, 11, ctxt, ops)
#define vcpu_has_avx512_bitalg() vcpu_has( 7, ECX, 12, ctxt, ops)
@@ -8935,13 +8936,9 @@ 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;
@@ -9655,6 +9652,24 @@ 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);
@@ -112,6 +112,7 @@
/* 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)
@@ -229,6 +229,7 @@ XEN_CPUFEATURE(UMIP, 6*32+ 2) /
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 */
@@ -257,7 +257,7 @@ 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
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> --- TBD: Should VAES also depend on AESNI? --- v8: No need to set fault_suppression to false. v7: New.