From 8b29dcc4be8f4f3a55543c27403f71cf33b493ec Mon Sep 17 00:00:00 2001
From: edeiana <edeiana@google.com>
Date: Wed, 6 Mar 2024 03:40:56 -0800
Subject: [PATCH] Using x86 encode/decode for synthetic encoding/decoding
 through setting of instr ISA mode.

---
 core/ir/instr.h      |  6 ++++++
 core/ir/x86/decode.c | 11 +++++++++++
 core/ir/x86/encode.c |  5 ++---
 core/ir/x86/instr.c  |  8 ++++++++
 4 files changed, 27 insertions(+), 3 deletions(-)

diff --git a/core/ir/instr.h b/core/ir/instr.h
index b26e83017cb..7e54a5d7ee1 100644
--- a/core/ir/instr.h
+++ b/core/ir/instr.h
@@ -215,6 +215,12 @@ enum {
 #endif
     /* PR 267260: distinguish our own mangling from client-added instrs */
     INSTR_OUR_MANGLING = 0x80000000,
+    /*
+     * Synthetic ISA mode.
+     * TOFIX: we're exceeding 4 bytes, this is an issue on 32 bits arch
+     * where uint flags of instr_t is only 4 bytes.
+     */
+    INSTR_SYNTH_MODE = 0x100000000,
 };
 
 #define DR_TUPLE_TYPE_BITS 4
diff --git a/core/ir/x86/decode.c b/core/ir/x86/decode.c
index f01137e3f4a..33fd9d37385 100644
--- a/core/ir/x86/decode.c
+++ b/core/ir/x86/decode.c
@@ -38,6 +38,7 @@
 /* decode.c -- a full x86 decoder */
 
 #include "../globals.h"
+#include "../synthetic/decode.h"
 #include "arch.h"
 #include "instr.h"
 #include "decode.h"
@@ -2579,6 +2580,16 @@ check_is_variable_size(opnd_t op)
 static byte *
 decode_common(dcontext_t *dcontext, byte *pc, byte *orig_pc, instr_t *instr)
 {
+    /*
+     * If we're dealing with decoding from synthetic ISA, we don't care about returning
+     * the pc of the next instruction (?), so we just write the encoding in final_pc and
+     * return it.
+     */
+    if (instr_get_isa_mode(instr) == DR_ISA_SYNTH) {
+        decode_from_synth(dcontext, orig_pc, instr);
+        return orig_pc;
+    }
+
     const instr_info_t *info;
     decode_info_t di;
     byte *next_pc;
diff --git a/core/ir/x86/encode.c b/core/ir/x86/encode.c
index 44ff55c8105..862149f24ee 100644
--- a/core/ir/x86/encode.c
+++ b/core/ir/x86/encode.c
@@ -2760,13 +2760,12 @@ instr_encode_arch(dcontext_t *dcontext, instr_t *instr, byte *copy_pc, byte *fin
                   bool *has_instr_opnds /*OUT OPTIONAL*/
                       _IF_DEBUG(bool assert_reachable))
 {
-
     /*
      * If we're dealing with encoding to synthetic ISA, we don't care about returning
-     * the pc of the next instruction, so we just write the encoding in final_pc and
+     * the pc of the next instruction (?), so we just write the encoding in final_pc and
      * return it.
      */
-    if (dr_get_isa_mode(dcontext) == DR_ISA_SYNTH) {
+    if (instr_get_isa_mode(instr) == DR_ISA_SYNTH) {
         encode_to_synth(dcontext, instr, final_pc);
         return final_pc;
     }
diff --git a/core/ir/x86/instr.c b/core/ir/x86/instr.c
index 9db609e7e81..b7a2e688f81 100644
--- a/core/ir/x86/instr.c
+++ b/core/ir/x86/instr.c
@@ -40,6 +40,7 @@
 #include "instr.h"
 #include "decode.h"
 #include "decode_private.h"
+#include "encode_api.h"
 #include "instr_create_shared.h"
 
 #ifdef X64
@@ -70,6 +71,10 @@ instr_get_x86_mode(instr_t *instr)
 bool
 instr_set_isa_mode(instr_t *instr, dr_isa_mode_t mode)
 {
+    if (mode == DR_ISA_SYNTH) {
+        instr->flags |= INSTR_SYNTH_MODE;
+        return true;
+    }
 #ifdef X64
     if (mode == DR_ISA_IA32)
         instr_set_x86_mode(instr, true);
@@ -87,6 +92,9 @@ instr_set_isa_mode(instr_t *instr, dr_isa_mode_t mode)
 dr_isa_mode_t
 instr_get_isa_mode(instr_t *instr)
 {
+    if (TEST(INSTR_SYNTH_MODE, instr->flags)) {
+        return DR_ISA_SYNTH;
+    }
 #ifdef X64
     return TEST(INSTR_X86_MODE, instr->flags) ? DR_ISA_IA32 : DR_ISA_AMD64;
 #else