x86emul: VCVT{,U}DQ2PD ignores embedded rounding

IOW we shouldn't raise #UD in that case. Be on the safe side though and
only encode fully legitimate forms into the stub to be executed.

Things weren't quite right for VCVT{,U}SI2SD either, in the attempt to
be on the safe side: Clearing EVEX.L'L isn't useful; it's EVEX.b which
primarily needs clearing. Also reflect the somewhat improved doc
situation in the comment there.

Fixes: ed806f373730 ("x86emul: support AVX512F legacy-equivalent packed int/FP conversion insns")
Fixes: baf4a376f550 ("x86emul: support AVX512F legacy-equivalent scalar int/FP conversion insns")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: d3709d1324aa140f064b9c68da37547f459f8e8d
master date: 2025-01-08 11:01:17 +0100

diff --git a/xen/arch/x86/x86_emulate/x86_emulate.c b/xen/arch/x86/x86_emulate/x86_emulate.c
index 09ab75d035f7f54d5c855d3a0d20f7fd1291e847..4ef868de3d7dd7d9ac918d6b8cfe6bf6a7065164 100644 (file)
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -3592,12 +3592,15 @@ x86_emulate(
         if ( !mode_64bit() )
             evex.w = 0;
         /*
-         * SDM version 067 claims that exception type E10NF implies #UD when
-         * EVEX.L'L is non-zero for 32-bit VCVT{,U}SI2SD. Experimentally this
-         * cannot be confirmed, but be on the safe side for the stub.
+         * While SDM version 085 has explicit wording towards embedded rounding
+         * being ignored, it's still not entirely unambiguous with the exception
+         * type referred to. Be on the safe side for the stub.
          */
         if ( !evex.w && evex.pfx == vex_f2 )
+        {
+            evex.brs = 0;
             evex.lr = 0;
+        }
         opc[1] = (modrm & 0x38) | 0xc0;
         insn_bytes = EVEX_PFX_BYTES + 2;
         opc[2] = 0xc3;
@@ -4815,7 +4818,16 @@ x86_emulate(
         else
         {
             host_and_vcpu_must_have(avx512f);
-            generate_exception_if(ea.type != OP_MEM && evex.brs, X86_EXC_UD);
+            /*
+             * While SDM version 085 has explicit wording towards embedded
+             * rounding being ignored, it's still not entirely unambiguous with
+             * the exception type referred to. Be on the safe side for the stub.
+             */
+            if ( ea.type != OP_MEM && evex.brs )
+            {
+                evex.brs = 0;
+                evex.lr = 2;
+            }
         }
         if ( ea.type != OP_REG || !evex.brs )
             avx512_vlen_check(false);