[PATCH v3 2/6] x86emul: support CMPccXADD

[Jan Beulich <jbeulich@xxxxxxxx>]

Unconditionally wire this through the ->rmw() hook. Since x86_emul_rmw()
now wants to construct and invoke a stub, make stub_exn available to it
via a new field in the emulator state structure.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
v3: Add dependency on LM. Re-base.
v2: Use X86_EXC_*. Move past introduction of stub_exn in struct
    x86_emulate_state. Keep feature at just "a" for now.
---
SDE: -grr or -srf

--- a/tools/misc/xen-cpuid.c
+++ b/tools/misc/xen-cpuid.c
@@ -186,6 +186,7 @@ static const char *const str_7d0[32] =
 static const char *const str_7a1[32] =
 {
     [ 4] = "avx-vnni",      [ 5] = "avx512-bf16",
+    /* 6 */                 [ 7] = "cmpccxadd",
 
     [10] = "fzrm",          [11] = "fsrs",
     [12] = "fsrcs",
--- a/tools/tests/x86_emulator/predicates.c
+++ b/tools/tests/x86_emulator/predicates.c
@@ -1403,6 +1403,22 @@ static const struct vex {
     { { 0xdd }, 2, T, R, pfx_66, WIG, Ln }, /* vaesenclast */
     { { 0xde }, 2, T, R, pfx_66, WIG, Ln }, /* vaesdec */
     { { 0xdf }, 2, T, R, pfx_66, WIG, Ln }, /* vaesdeclast */
+    { { 0xe0 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpoxadd */
+    { { 0xe1 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpnoxadd */
+    { { 0xe2 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpbxadd */
+    { { 0xe3 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpnbxadd */
+    { { 0xe4 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpexadd */
+    { { 0xe5 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpnexadd */
+    { { 0xe6 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpbexadd */
+    { { 0xe7 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpaxadd */
+    { { 0xe8 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpsxadd */
+    { { 0xe9 }, 2, F, W, pfx_66, Wn, L0 }, /* cmpnsxadd */
+    { { 0xea }, 2, F, W, pfx_66, Wn, L0 }, /* cmppxadd */
+    { { 0xeb }, 2, F, W, pfx_66, Wn, L0 }, /* cmpnpxadd */
+    { { 0xec }, 2, F, W, pfx_66, Wn, L0 }, /* cmplxadd */
+    { { 0xed }, 2, F, W, pfx_66, Wn, L0 }, /* cmpgexadd */
+    { { 0xee }, 2, F, W, pfx_66, Wn, L0 }, /* cmplexadd */
+    { { 0xef }, 2, F, W, pfx_66, Wn, L0 }, /* cmpgxadd */
     { { 0xf2 }, 2, T, R, pfx_no, Wn, L0 }, /* andn */
     { { 0xf3, 0x08 }, 2, T, R, pfx_no, Wn, L0 }, /* blsr */
     { { 0xf3, 0x10 }, 2, T, R, pfx_no, Wn, L0 }, /* blsmsk */
--- a/tools/tests/x86_emulator/test_x86_emulator.c
+++ b/tools/tests/x86_emulator/test_x86_emulator.c
@@ -1412,6 +1412,78 @@ int main(int argc, char **argv)
     }
     printf("okay\n");
 
+    printf("%-40s", "Testing cmpbxadd %rbx,%r9,(%rdx)...");
+    if ( stack_exec && cpu_has_cmpccxadd )
+    {
+        instr[0] = 0xc4; instr[1] = 0x62; instr[2] = 0xe1; instr[3] = 0xe2; 
instr[4] = 0x0a;
+        regs.rip = (unsigned long)&instr[0];
+        regs.eflags = EFLAGS_ALWAYS_SET;
+        res[0] = 0x11223344;
+        res[1] = 0x01020304;
+        regs.rdx = (unsigned long)res;
+        regs.r9  = 0x0001020300112233UL;
+        regs.rbx = 0x0101010101010101UL;
+        rc = x86_emulate(&ctxt, &emulops);
+        if ( (rc != X86EMUL_OKAY) ||
+             (regs.eip != (unsigned long)&instr[5]) ||
+             (regs.r9 != 0x0102030411223344UL) ||
+             (regs.rbx != 0x0101010101010101UL) ||
+             ((regs.eflags & EFLAGS_MASK) !=
+              (X86_EFLAGS_PF | EFLAGS_ALWAYS_SET)) ||
+             (res[0] != 0x11223344) ||
+             (res[1] != 0x01020304) )
+            goto fail;
+
+        regs.rip = (unsigned long)&instr[0];
+        regs.r9 <<= 8;
+        rc = x86_emulate(&ctxt, &emulops);
+        if ( (rc != X86EMUL_OKAY) ||
+             (regs.eip != (unsigned long)&instr[5]) ||
+             (regs.r9 != 0x0102030411223344UL) ||
+             (regs.rbx != 0x0101010101010101UL) ||
+             ((regs.eflags & EFLAGS_MASK) !=
+              (X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_SF |
+               EFLAGS_ALWAYS_SET)) ||
+             (res[0] != 0x12233445) ||
+             (res[1] != 0x02030405) )
+            goto fail;
+        printf("okay\n");
+
+        printf("%-40s", "Testing cmpsxadd %r9d,%ebx,4(%r10)...");
+        instr[1] = 0xc2; instr[2] = 0x31; instr[3] = 0xe8; instr[4] = 0x5a; 
instr[5] = 0x04;
+        regs.rip = (unsigned long)&instr[0];
+        res[2] = res[0] = ~0;
+        regs.r10 = (unsigned long)res;
+        rc = x86_emulate(&ctxt, &emulops);
+        if ( (rc != X86EMUL_OKAY) ||
+             (regs.eip != (unsigned long)&instr[6]) ||
+             (regs.r9 != 0x0102030411223344UL) ||
+             (regs.rbx != 0x02030405) ||
+             ((regs.eflags & EFLAGS_MASK) != EFLAGS_ALWAYS_SET) ||
+             (res[0] + 1) ||
+             (res[1] != 0x02030405) ||
+             (res[2] + 1) )
+            goto fail;
+
+        regs.rip = (unsigned long)&instr[0];
+        regs.rbx <<= 8;
+        rc = x86_emulate(&ctxt, &emulops);
+        if ( (rc != X86EMUL_OKAY) ||
+             (regs.eip != (unsigned long)&instr[6]) ||
+             (regs.r9 != 0x0102030411223344UL) ||
+             (regs.rbx != 0x02030405) ||
+             ((regs.eflags & EFLAGS_MASK) !=
+              (X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_SF |
+               EFLAGS_ALWAYS_SET)) ||
+             (res[0] + 1) ||
+             (res[1] != 0x13253749) ||
+             (res[2] + 1) )
+            goto fail;
+        printf("okay\n");
+    }
+    else
+        printf("skipped\n");
+
     emulops.write_segment = write_segment;
     emulops.write_msr     = write_msr;
 
--- a/tools/tests/x86_emulator/x86-emulate.h
+++ b/tools/tests/x86_emulator/x86-emulate.h
@@ -178,6 +178,7 @@ void wrpkru(unsigned int val);
 #define cpu_has_avx512_fp16 (cp.feat.avx512_fp16 && xcr0_mask(0xe6))
 #define cpu_has_avx_vnni   (cp.feat.avx_vnni && xcr0_mask(6))
 #define cpu_has_avx512_bf16 (cp.feat.avx512_bf16 && xcr0_mask(0xe6))
+#define cpu_has_cmpccxadd  cp.feat.cmpccxadd
 #define cpu_has_avx_ifma   (cp.feat.avx_ifma && xcr0_mask(6))
 #define cpu_has_avx_vnni_int8 (cp.feat.avx_vnni_int8 && xcr0_mask(6))
 #define cpu_has_avx_ne_convert (cp.feat.avx_ne_convert && xcr0_mask(6))
--- a/xen/arch/x86/include/asm/cpufeature.h
+++ b/xen/arch/x86/include/asm/cpufeature.h
@@ -186,6 +186,7 @@ static inline bool boot_cpu_has(unsigned
 /* CPUID level 0x00000007:1.eax */
 #define cpu_has_avx_vnni        boot_cpu_has(X86_FEATURE_AVX_VNNI)
 #define cpu_has_avx512_bf16     boot_cpu_has(X86_FEATURE_AVX512_BF16)
+#define cpu_has_cmpccxadd       boot_cpu_has(X86_FEATURE_CMPCCXADD)
 #define cpu_has_avx_ifma        boot_cpu_has(X86_FEATURE_AVX_IFMA)
 
 /* CPUID level 0x00000007:1.edx */
--- a/xen/arch/x86/x86_emulate/decode.c
+++ b/xen/arch/x86/x86_emulate/decode.c
@@ -439,6 +439,7 @@ static const struct ext0f38_table {
     [0xd7] = { .simd_size = simd_scalar_vexw, .d8s = d8s_dq },
     [0xdb] = { .simd_size = simd_packed_int, .two_op = 1 },
     [0xdc ... 0xdf] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
+    [0xe0 ... 0xef] = { .to_mem = 1 },
     [0xf0] = { .two_op = 1 },
     [0xf1] = { .to_mem = 1, .two_op = 1 },
     [0xf2 ... 0xf3] = {},
@@ -931,6 +932,8 @@ decode_0f38(struct x86_emulate_state *s,
             ctxt->opcode |= MASK_INSR(s->vex.pfx, X86EMUL_OPC_PFX_MASK);
         break;
 
+    case X86EMUL_OPC_VEX_66(0, 0xe0) ...
+         X86EMUL_OPC_VEX_66(0, 0xef): /* cmp<cc>xadd */
     case X86EMUL_OPC_VEX(0, 0xf2):    /* andn */
     case X86EMUL_OPC_VEX(0, 0xf3):    /* Grp 17 */
     case X86EMUL_OPC_VEX(0, 0xf5):    /* bzhi */
--- a/xen/arch/x86/x86_emulate/private.h
+++ b/xen/arch/x86/x86_emulate/private.h
@@ -257,6 +257,7 @@ struct x86_emulate_state {
         rmw_btc,
         rmw_btr,
         rmw_bts,
+        rmw_cmpccxadd,
         rmw_dec,
         rmw_inc,
         rmw_neg,
@@ -589,6 +590,7 @@ amd_like(const struct x86_emulate_ctxt *
 #define vcpu_has_avx512_fp16() (ctxt->cpuid->feat.avx512_fp16)
 #define vcpu_has_avx_vnni()    (ctxt->cpuid->feat.avx_vnni)
 #define vcpu_has_avx512_bf16() (ctxt->cpuid->feat.avx512_bf16)
+#define vcpu_has_cmpccxadd()   (ctxt->cpuid->feat.cmpccxadd)
 #define vcpu_has_lkgs()        (ctxt->cpuid->feat.lkgs)
 #define vcpu_has_wrmsrns()     (ctxt->cpuid->feat.wrmsrns)
 #define vcpu_has_avx_ifma()    (ctxt->cpuid->feat.avx_ifma)
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -6918,6 +6918,15 @@ x86_emulate(
 
 #endif /* !X86EMUL_NO_SIMD */
 
+    case X86EMUL_OPC_VEX_66(0x0f38, 0xe0) ...
+         X86EMUL_OPC_VEX_66(0x0f38, 0xef): /* cmp<cc>xadd r,r,m */
+        generate_exception_if(!mode_64bit() || dst.type != OP_MEM || vex.l,
+                              X86_EXC_UD);
+        host_and_vcpu_must_have(cmpccxadd);
+        fail_if(!ops->rmw);
+        state->rmw = rmw_cmpccxadd;
+        break;
+
     case X86EMUL_OPC(0x0f38, 0xf0): /* movbe m,r */
     case X86EMUL_OPC(0x0f38, 0xf1): /* movbe r,m */
         vcpu_must_have(movbe);
@@ -8191,14 +8200,20 @@ x86_emulate(
     {
         ea.val = src.val;
         op_bytes = dst.bytes;
+        state->stub_exn = &stub_exn;
         rc = ops->rmw(dst.mem.seg, dst.mem.off, dst.bytes, &_regs.eflags,
                       state, ctxt);
+#ifdef __XEN__
+        if ( rc == X86EMUL_stub_failure )
+            goto emulation_stub_failure;
+#endif
         if ( rc != X86EMUL_OKAY )
             goto done;
 
         /* Some operations require a register to be written. */
         switch ( state->rmw )
         {
+        case rmw_cmpccxadd:
         case rmw_xchg:
         case rmw_xadd:
             switch ( dst.bytes )
@@ -8473,6 +8488,7 @@ int x86_emul_rmw(
     uint32_t *eflags,
     struct x86_emulate_state *s,
     struct x86_emulate_ctxt *ctxt)
+#define stub_exn (*s->stub_exn) /* for invoke_stub() */
 {
     unsigned long *dst = ptr;
 
@@ -8538,6 +8554,37 @@ int x86_emul_rmw(
 #undef BINOP
 #undef SHIFT
 
+#ifdef __x86_64__
+    case rmw_cmpccxadd:
+    {
+        struct x86_emulate_stub stub = {};
+        uint8_t *buf = get_stub(stub);
+        typeof(s->vex) *pvex = container_of(buf + 1, typeof(s->vex),
+                                            raw[0]);
+        unsigned long dummy;
+
+        buf[0] = 0xc4;
+        *pvex = s->vex;
+        pvex->b = 1;
+        pvex->r = 1;
+        pvex->reg = 0xf; /* rAX */
+        buf[3] = ctxt->opcode;
+        buf[4] = 0x11; /* reg=rDX r/m=(%RCX) */
+        buf[5] = 0xc3;
+
+        *eflags &= ~EFLAGS_MASK;
+        invoke_stub("",
+                    _POST_EFLAGS("[eflags]", "[mask]", "[tmp]"),
+                    "+m" (*dst), "+d" (s->ea.val),
+                    [tmp] "=&r" (dummy), [eflags] "+g" (*eflags)
+                    : "a" (*decode_vex_gpr(s->vex.reg, ctxt->regs, ctxt)),
+                      "c" (dst), [mask] "i" (EFLAGS_MASK));
+
+        put_stub(stub);
+        break;
+    }
+#endif
+
     case rmw_not:
         switch ( s->op_bytes )
         {
@@ -8633,7 +8680,13 @@ int x86_emul_rmw(
 #undef JCXZ
 
     return X86EMUL_OKAY;
+
+#if defined(__XEN__) && defined(__x86_64__)
+ emulation_stub_failure:
+    return X86EMUL_stub_failure;
+#endif
 }
+#undef stub_exn
 
 static void __init __maybe_unused build_assertions(void)
 {
--- a/xen/include/public/arch-x86/cpufeatureset.h
+++ b/xen/include/public/arch-x86/cpufeatureset.h
@@ -279,6 +279,7 @@ XEN_CPUFEATURE(SSBD,          9*32+31) /
 /* Intel-defined CPU features, CPUID level 0x00000007:1.eax, word 10 */
 XEN_CPUFEATURE(AVX_VNNI,     10*32+ 4) /*A  AVX-VNNI Instructions */
 XEN_CPUFEATURE(AVX512_BF16,  10*32+ 5) /*A  AVX512 BFloat16 Instructions */
+XEN_CPUFEATURE(CMPCCXADD,    10*32+ 7) /*a  CMPccXADD Instructions */
 XEN_CPUFEATURE(FZRM,         10*32+10) /*A  Fast Zero-length REP MOVSB */
 XEN_CPUFEATURE(FSRS,         10*32+11) /*A  Fast Short REP STOSB */
 XEN_CPUFEATURE(FSRCS,        10*32+12) /*A  Fast Short REP CMPSB/SCASB */
--- a/xen/tools/gen-cpuid.py
+++ b/xen/tools/gen-cpuid.py
@@ -274,7 +274,7 @@ def crunch_numbers(state):
         # superpages, PCID and PKU are only available in 4 level paging.
         # NO_LMSL indicates the absense of Long Mode Segment Limits, which
         # have been dropped in hardware.
-        LM: [CX16, PCID, LAHF_LM, PAGE1GB, PKU, NO_LMSL, LKGS],
+        LM: [CX16, PCID, LAHF_LM, PAGE1GB, PKU, NO_LMSL, LKGS, CMPCCXADD],
 
         # AMD K6-2+ and K6-III processors shipped with 3DNow+, beyond the
         # standard 3DNow in the earlier K6 processors.