j*cxz/pusha/popa alternate size forms

these all existed since forever but the library did not distinguish them
and did not provide prefix information for users to tell which had been
decoded.

3 files changed, 64 insertions, 27 deletions

diff --git a/src/real_mode/behavior.rs b/src/real_mode/behavior.rs
index cc7f705..8941ee5 100644
--- a/src/real_mode/behavior.rs
+++ b/src/real_mode/behavior.rs
@@ -140,22 +140,6 @@ impl Instruction {
                         .set_operand(1, Access::Read)
                         .set_operand(2, Access::Read);
                 }
-            } else if self.opcode == Opcode::PUSHA {
-                if self.prefixes.operand_size() {
-                    behavior = behavior
-                        .set_implicit_ops(PUSHAD_IDX);
-                } else {
-                    behavior = behavior
-                        .set_implicit_ops(PUSHA_IDX);
-                }
-            } else if self.opcode == Opcode::POPA {
-                if self.prefixes.operand_size() {
-                    behavior = behavior
-                        .set_implicit_ops(POPAD_IDX);
-                } else {
-                    behavior = behavior
-                        .set_implicit_ops(POPA_IDX);
-                }
             } else if self.opcode == Opcode::DIV || self.opcode == Opcode::IDIV {
                 let op_width = if self.operands[0] == OperandSpec::RegMMM {
                     self.regs[1].width()
@@ -257,8 +241,7 @@ impl Instruction {
                 }
             } else if self.opcode() == Opcode::LOOPNZ
                 || self.opcode() == Opcode::LOOPZ
-                || self.opcode() == Opcode::LOOP
-                || self.opcode() == Opcode::JCXZ {
+                || self.opcode() == Opcode::LOOP {
                if !self.prefixes.operand_size() {
                    behavior = behavior
                        .set_implicit_ops(RW_CX_IDX);
@@ -3808,7 +3791,7 @@ fn behavior_table_size_is_right() {
 }
 
 /// this table MUST line up with Opcode declaration order in `mod.rs`.
-static TABLE: [BehaviorDigest; 1425] = [
+static TABLE: [BehaviorDigest; 1428] = [
     /* ADD => */ GENERAL_RW_R_FLAGWRITE,
     /* OR => */ GENERAL_RW_R_FLAGWRITE,
     /* ADC => */ GENERAL_RW_R_FLAGRW,
@@ -5424,14 +5407,14 @@ static TABLE: [BehaviorDigest; 1425] = [
     /* JCXZ => */ BehaviorDigest::empty()
             .set_pl_any()
             .set_operand(0, Access::Read)
-            .set_nontrivial(true),
+            .set_implicit_ops(RW_CX_IDX),
 
     /* PUSHA => */ BehaviorDigest::empty()
             .set_pl_any()
-            .set_nontrivial(true),
+            .set_implicit_ops(PUSHA_IDX),
     /* POPA => */ BehaviorDigest::empty()
             .set_pl_any()
-            .set_nontrivial(true),
+            .set_implicit_ops(POPA_IDX),
     /* BOUND => */ BehaviorDigest::empty()
             .set_pl_any()
             .set_operand(0, Access::Read)
@@ -6170,4 +6153,15 @@ static TABLE: [BehaviorDigest; 1425] = [
             .set_pl0()
             .set_flags_access(Access::Write)
             .set_complex(true),
+
+    /* PUSHAD => */ BehaviorDigest::empty()
+            .set_pl_any()
+            .set_implicit_ops(PUSHAD_IDX),
+    /* POPAD => */ BehaviorDigest::empty()
+            .set_pl_any()
+            .set_implicit_ops(POPAD_IDX),
+    /* JECXZ => */ BehaviorDigest::empty()
+            .set_pl_any()
+            .set_operand(0, Access::Read)
+            .set_implicit_ops(RW_ECX_IDX),
 ];
diff --git a/src/real_mode/display.rs b/src/real_mode/display.rs
index e686b0a..888b0b1 100644
--- a/src/real_mode/display.rs
+++ b/src/real_mode/display.rs
@@ -2094,6 +2094,10 @@ const MNEMONICS: &[&'static str] = &[
     "pvalidate",
     "rmpadjust",
     "rmpupdate",
+
+    "pushad",
+    "popad",
+    "jecxz",
 ];
 
 impl Opcode {
@@ -2572,6 +2576,10 @@ pub(crate) fn contextualize_c<T: DisplaySink>(instr: &Instruction, out: &mut T)
             out.write_str("jmp ")?;
             write_jmp_operand(instr.operand(0), out)?;
         },
+        Opcode::JECXZ => {
+            out.write_str("if ecx == 0 then jmp ")?;
+            write_jmp_operand(instr.operand(0), out)?;
+        },
         Opcode::JCXZ => {
             out.write_str("if cx == 0 then jmp ")?;
             write_jmp_operand(instr.operand(0), out)?;
@@ -2766,8 +2774,9 @@ impl <T: fmt::Write, Y: YaxColors> ShowContextual<u64, [Option<alloc::string::St
     }
 }
 
-static RELATIVE_BRANCHES: [Opcode; 22] = [
-    Opcode::JMP, Opcode::CALL, Opcode::JCXZ,
+static RELATIVE_BRANCHES: [Opcode; 23] = [
+    Opcode::JMP, Opcode::CALL,
+    Opcode::JECXZ, Opcode::JCXZ,
     Opcode::LOOP, Opcode::LOOPZ, Opcode::LOOPNZ,
     Opcode::JO, Opcode::JNO,
     Opcode::JB, Opcode::JNB,
diff --git a/src/real_mode/mod.rs b/src/real_mode/mod.rs
index f538fa5..8dd29b1 100644
--- a/src/real_mode/mod.rs
+++ b/src/real_mode/mod.rs
@@ -2508,6 +2508,13 @@ pub enum Opcode {
     PVALIDATE,
     RMPADJUST,
     RMPUPDATE,
+
+    // these were part of the base (i386) ISA but were reported as distinct opcodes relatively late.
+    PUSHAD,
+    POPAD,
+
+    // likewise with alternate-size jecxz/jcxz..
+    JECXZ,
 }
 
 impl PartialEq for Instruction {
@@ -3992,6 +3999,7 @@ enum OperandCase {
     ModRM_0xf30f38fa,
     ModRM_0xf30f38fb,
     ModRM_0xf30f3af0,
+    CXZ, // catch-all for "j*cxz and prefix handling"
 }
 
 #[allow(non_camel_case_types)]
@@ -4319,6 +4327,7 @@ enum OperandCode {
     ModRM_0xc4 = OperandCodeBuilder::new().operand_case(OperandCase::ModRM_0xc4).bits(),
     ModRM_0xc5 = OperandCodeBuilder::new().operand_case(OperandCase::ModRM_0xc5).bits(),
     MASKMOVDQU = OperandCodeBuilder::new().read_E().reg_mem().operand_case(OperandCase::MASKMOVDQU).bits(),
+    CXZ = OperandCodeBuilder::new().operand_case(OperandCase::CXZ).bits(),
 }
 
 fn base_opcode_map(v: u8) -> Opcode {
@@ -4622,7 +4631,7 @@ const OPCODES: [OpcodeRecord; 256] = [
     OpcodeRecord::new(Interpretation::Instruction(Opcode::LOOPNZ), OperandCode::Ibs),
     OpcodeRecord::new(Interpretation::Instruction(Opcode::LOOPZ), OperandCode::Ibs),
     OpcodeRecord::new(Interpretation::Instruction(Opcode::LOOP), OperandCode::Ibs),
-    OpcodeRecord::new(Interpretation::Instruction(Opcode::JCXZ), OperandCode::Ibs),
+    OpcodeRecord::new(Interpretation::Instruction(Opcode::JCXZ), OperandCode::CXZ),
     OpcodeRecord::new(Interpretation::Instruction(Opcode::IN), OperandCode::AL_Ib),
     OpcodeRecord::new(Interpretation::Instruction(Opcode::IN), OperandCode::AX_Ib),
     OpcodeRecord::new(Interpretation::Instruction(Opcode::OUT), OperandCode::Ib_AL),
@@ -6510,9 +6519,19 @@ fn read_operands<
             } else if instruction.opcode == Opcode::XLAT {
                 instruction.mem_size = 1;
             } else if instruction.opcode == Opcode::PUSHA {
-                instruction.mem_size = 2 * 8;
+                if !instruction.prefixes.operand_size() {
+                    instruction.mem_size = 2 * 8;
+                } else {
+                    instruction.opcode = Opcode::PUSHAD;
+                    instruction.mem_size = 4 * 8;
+                }
             } else if instruction.opcode == Opcode::POPA {
-                instruction.mem_size = 2 * 8;
+                if !instruction.prefixes.operand_size() {
+                    instruction.mem_size = 2 * 8;
+                } else {
+                    instruction.opcode = Opcode::POPAD;
+                    instruction.mem_size = 4 * 8;
+                }
             }
             instruction.operands[0] = OperandSpec::Nothing;
             instruction.operand_count = 0;
@@ -9050,6 +9069,21 @@ fn read_operands<
             instruction.mem_size = 2;
             instruction.operand_count = 2;
         },
+        OperandCase::CXZ => {
+            if instruction.prefixes.address_size() {
+                // address-size overridden from 16-bit to 32-bit
+                instruction.opcode = Opcode::JECXZ;
+            }
+            instruction.imm =
+                read_imm_signed(words, 1)? as u32;
+            sink.record(
+                words.offset() as u32 * 8 - 8,
+                words.offset() as u32 * 8 - 1,
+                InnerDescription::Number("1-byte immediate", instruction.imm as i64)
+                    .with_id(words.offset() as u32 * 8),
+            );
+            instruction.operands[0] = OperandSpec::ImmI8;
+        },
     };
     Ok(())
 }