[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[PATCH v4 16/28] x86emul: support APX forms of {LD,ST}TILECFG


  • To: "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Wed, 11 Nov 2026 16:13:36 +0100
  • Authentication-results: eu.smtp.expurgate.cloud; dkim=pass header.s=google header.d=suse.com header.i="@suse.com" header.h="Content-Transfer-Encoding:Content-Type:In-Reply-To:Autocrypt:Content-Language:References:Cc:To:From:Subject:User-Agent:MIME-Version:Date:Message-ID"
  • Autocrypt: addr=jbeulich@xxxxxxxx; keydata= xsDiBFk3nEQRBADAEaSw6zC/EJkiwGPXbWtPxl2xCdSoeepS07jW8UgcHNurfHvUzogEq5xk hu507c3BarVjyWCJOylMNR98Yd8VqD9UfmX0Hb8/BrA+Hl6/DB/eqGptrf4BSRwcZQM32aZK 7Pj2XbGWIUrZrd70x1eAP9QE3P79Y2oLrsCgbZJfEwCgvz9JjGmQqQkRiTVzlZVCJYcyGGsD /0tbFCzD2h20ahe8rC1gbb3K3qk+LpBtvjBu1RY9drYk0NymiGbJWZgab6t1jM7sk2vuf0Py O9Hf9XBmK0uE9IgMaiCpc32XV9oASz6UJebwkX+zF2jG5I1BfnO9g7KlotcA/v5ClMjgo6Gl MDY4HxoSRu3i1cqqSDtVlt+AOVBJBACrZcnHAUSuCXBPy0jOlBhxPqRWv6ND4c9PH1xjQ3NP nxJuMBS8rnNg22uyfAgmBKNLpLgAGVRMZGaGoJObGf72s6TeIqKJo/LtggAS9qAUiuKVnygo 3wjfkS9A3DRO+SpU7JqWdsveeIQyeyEJ/8PTowmSQLakF+3fote9ybzd880fSmFuIEJldWxp Y2ggPGpiZXVsaWNoQHN1c2UuY29tPsJgBBMRAgAgBQJZN5xEAhsDBgsJCAcDAgQVAggDBBYC AwECHgECF4AACgkQoDSui/t3IH4J+wCfQ5jHdEjCRHj23O/5ttg9r9OIruwAn3103WUITZee e7Sbg12UgcQ5lv7SzsFNBFk3nEQQCACCuTjCjFOUdi5Nm244F+78kLghRcin/awv+IrTcIWF hUpSs1Y91iQQ7KItirz5uwCPlwejSJDQJLIS+QtJHaXDXeV6NI0Uef1hP20+y8qydDiVkv6l IreXjTb7DvksRgJNvCkWtYnlS3mYvQ9NzS9PhyALWbXnH6sIJd2O9lKS1Mrfq+y0IXCP10eS FFGg+Av3IQeFatkJAyju0PPthyTqxSI4lZYuJVPknzgaeuJv/2NccrPvmeDg6Coe7ZIeQ8Yj t0ARxu2xytAkkLCel1Lz1WLmwLstV30g80nkgZf/wr+/BXJW/oIvRlonUkxv+IbBM3dX2OV8 AmRv1ySWPTP7AAMFB/9PQK/VtlNUJvg8GXj9ootzrteGfVZVVT4XBJkfwBcpC/XcPzldjv+3 HYudvpdNK3lLujXeA5fLOH+Z/G9WBc5pFVSMocI71I8bT8lIAzreg0WvkWg5V2WZsUMlnDL9 mpwIGFhlbM3gfDMs7MPMu8YQRFVdUvtSpaAs8OFfGQ0ia3LGZcjA6Ik2+xcqscEJzNH+qh8V m5jjp28yZgaqTaRbg3M/+MTbMpicpZuqF4rnB0AQD12/3BNWDR6bmh+EkYSMcEIpQmBM51qM EKYTQGybRCjpnKHGOxG0rfFY1085mBDZCH5Kx0cl0HVJuQKC+dV2ZY5AqjcKwAxpE75MLFkr wkkEGBECAAkFAlk3nEQCGwwACgkQoDSui/t3IH7nnwCfcJWUDUFKdCsBH/E5d+0ZnMQi+G0A nAuWpQkjM1ASeQwSHEeAWPgskBQL
  • Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Teddy Astie <teddy.astie@xxxxxxxxxx>, Roger Pau Monné <roger@xxxxxxxxxxxxxx>
  • Delivery-date: Wed, 11 Nov 2026 15:13:40 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

Besides adding table entries and case labels, re-existing handling can
be re-used. The only special aspect once again is dealing with the
512-bit memory operands in EVEX Disp8 testing.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
v4: New.

--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -1192,6 +1192,12 @@ static const struct test popcnt[] = {
     INSN(popcntw,   66, map4, 88,    el,    w, no),
 };
 
+static const struct test amx_tile[] = {
+    /* Element size is set arbitrarily here. See test_one(). */
+    INSN(ldtilecfg,     , 0f38, 49, el, b, no),
+    INSN(sttilecfg,   66, 0f38, 49, el, b, no),
+};
+
 static const unsigned char vl_all[] = { VL_512, VL_128, VL_256 };
 static const unsigned char vl_128[] = { VL_128 };
 static const unsigned char vl_no128[] = { VL_512, VL_256 };
@@ -1452,6 +1458,13 @@ static void test_one(const struct test *
     {
         asm volatile ( "kxnorw %k1, %k1, %k1" );
         asm volatile ( "vxorps %xmm4, %xmm4, %xmm4" );
+
+        /*
+         * {LD,ST}TILECFG's operand size of 512 bits doesn't match up with
+         * EVEX.L'L being 0.
+         */
+        if ( test->spc == SPC_0f38 && test->opc == 0x49 )
+            vsz = 64;
     }
     else
     {
@@ -1796,6 +1809,7 @@ void evex_disp8_test(void *instr, struct
     } while ( false )
 
         RUN(adx);
+        RUN(amx_tile);
         RUN(avx512f);
         RUN(avx512dq);
         RUN(avx512bw);
--- a/tools/tests/x86_emulator/predicates.c
+++ b/tools/tests/x86_emulator/predicates.c
@@ -1937,6 +1937,8 @@ static const struct evex {
     { { 0x45 }, 2, T, R, pfx_66, Wn, Ln }, /* vpsrlv{d,q} */
     { { 0x46 }, 2, T, R, pfx_66, Wn, Ln }, /* vpsrav{d,q} */
     { { 0x47 }, 2, T, R, pfx_66, Wn, Ln }, /* vpsllv{d,q} */
+    { { 0x49, 0x00 }, 2, F, R, pfx_no, W0, L0 }, /* ldtilecfg */
+    { { 0x49, 0x00 }, 2, F, W, pfx_66, W0, L0 }, /* sttilecfg */
     { { 0x4c }, 2, T, R, pfx_66, Wn, Ln }, /* vrcp14p{s,d} */
     { { 0x4d }, 2, T, R, pfx_66, Wn, LIG }, /* vrcp14s{s,d} */
     { { 0x4e }, 2, T, R, pfx_66, Wn, Ln }, /* vrsqrt14p{s,d} */
--- a/tools/tests/x86_emulator/test_x86_emulator.c
+++ b/tools/tests/x86_emulator/test_x86_emulator.c
@@ -6218,6 +6218,45 @@ int main(int argc, char **argv)
     else
         printf("skipped\n");
 
+    if ( stack_exec && cpu_has_amx_tile && cpu_has_apx_f )
+    {
+        static const struct x86_tilecfg tilecfg_apx = {
+            .palette = 1,
+            .colsb = { 2, 3, 4, 5 },
+            .rows = { 5, 4, 3, 2 },
+        };
+        decl_insn(ldtilecfg_apx);
+        decl_insn(sttilecfg_apx);
+
+        printf("%-40s", "Testing ldtilecfg (%r26)...");
+
+        asm volatile ( put_insn(ldtilecfg_apx,
+                                /* ldtilecfg (%%r26) */
+                                ".byte 0x62, 0xda, 0x7c, 0x08, 0x49, 0x02") );
+
+        set_insn(ldtilecfg_apx);
+        load_egpr(26, (unsigned long)&tilecfg_apx);
+        rc = x86_emulate(&ctxt, &emulops);
+        if ( rc != X86EMUL_OKAY || !check_eip(ldtilecfg_apx) )
+            goto fail;
+        printf("pending\n");
+
+        printf("%-40s", "Testing sttilecfg -4(%r25)...");
+
+        asm volatile ( put_insn(sttilecfg_apx,
+                                /* sttilecfg -4(%%r25) */
+                                ".byte 0x62, 0xda, 0x7d, 0x08, 0x49, 0x41, 
0xfc") );
+
+        memset(res, ~0, 72);
+        set_insn(sttilecfg_apx);
+        load_egpr(25, (unsigned long)(res + 2));
+        rc = x86_emulate(&ctxt, &emulops);
+        if ( rc != X86EMUL_OKAY || !check_eip(sttilecfg_apx) ||
+             ~res[0] || ~res[17] || memcmp(res + 1, &tilecfg_apx, 64) )
+            goto fail;
+        printf("okay\n");
+    }
+
     printf("%-40s", "Testing vzeroupper (compat)...");
     if ( cpu_has_avx )
     {
--- a/xen/arch/x86/x86_emulate/decode.c
+++ b/xen/arch/x86/x86_emulate/decode.c
@@ -1271,6 +1271,9 @@ decode_0f38(struct x86_emulate_state *s,
         s->simd_size = simd_packed_fp;
         break;
 
+    case X86EMUL_OPC_EVEX_66(0, 0x49): /* sttilecfg */
+        s->promoted_evex = true;
+        fallthrough;
     case X86EMUL_OPC_VEX_66(0, 0x49): /* sttilecfg */
     case X86EMUL_OPC_VEX_F3(0, 0x4b): /* tilestored */
         s->desc = DstMem | SrcImplicit | Mov;
@@ -1308,6 +1311,7 @@ decode_0f38(struct x86_emulate_state *s,
     case X86EMUL_OPC_VEX_F2(0, 0xf7): /* shrx */
         break;
 
+    case X86EMUL_OPC_EVEX(0, 0x49):    /* ldtilecfg */
     case X86EMUL_OPC_EVEX_66(0, 0xe0) ...
          X86EMUL_OPC_EVEX_66(0, 0xef): /* cmp<cc>xadd */
     case X86EMUL_OPC_EVEX(0, 0xf2):    /* andn */
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -6650,10 +6650,16 @@ x86_emulate(
         generate_exception_if(vex.l, X86_EXC_UD);
         goto simd_0f_avx;
 
+    case X86EMUL_OPC_EVEX(0x0f38, 0x49):
+        generate_exception_if(evex.lr || evex.brs || evex.opmsk, X86_EXC_UD);
+        fallthrough;
     case X86EMUL_OPC_VEX(0x0f38, 0x49):
         generate_exception_if(!mode_64bit() || vex.l || vex.w, X86_EXC_UD);
         if ( ea.type == OP_REG )
         {
+            if ( evex_encoded() )
+                goto unrecognized_insn;
+
             switch ( modrm )
             {
             case 0xc0: /* tilerelease */
@@ -6725,6 +6731,9 @@ x86_emulate(
         }
         goto unrecognized_insn;
 
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x49):
+        generate_exception_if(evex.lr || evex.brs || evex.opmsk, X86_EXC_UD);
+        fallthrough;
     case X86EMUL_OPC_VEX_66(0x0f38, 0x49):
         generate_exception_if(!mode_64bit() || vex.l || vex.w, X86_EXC_UD);
         if ( ea.type == OP_REG )




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.