From a91c1c50a39ff68a7e73cef70829e8a5b08b4b6b Mon Sep 17 00:00:00 2001
From: iximeow <me@iximeow.net>
Date: Sun, 17 Mar 2024 02:44:57 +0000
Subject: be more explicit about undefined system instructions being invalid

---
 differential-tests/tests/capstone-differential.rs | 9 ++++++++-
 src/armv8/a64.rs                                  | 6 ++++++
 tests/armv8/a64.rs                                | 5 +++++
 3 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/differential-tests/tests/capstone-differential.rs b/differential-tests/tests/capstone-differential.rs
index 3db6fdd..cbed59f 100644
--- a/differential-tests/tests/capstone-differential.rs
+++ b/differential-tests/tests/capstone-differential.rs
@@ -368,7 +368,14 @@ fn capstone_differential() {
                         stats.missed_incomplete.fetch_add(1, Ordering::SeqCst);
                         continue;
                     } else {
-                        panic!("yax errored where capstone succeeded. cs text: '{}', bytes: {:x?}", cs_text, bytes);
+                        // capstone dedodes the UNDEFINED encodings in C5.1.2 as "mrs", yax returns
+                        // a decode error.
+                        if cs_text.starts_with("mrs ") {
+                            stats.yax_reject.fetch_add(1, Ordering::SeqCst);
+                            continue;
+                        } else {
+                            panic!("yax errored where capstone succeeded. cs text: '{}', bytes: {:x?}", cs_text, bytes);
+                        }
                     };
 
                     fn acceptable_match(yax_text: &str, cs_text: &str) -> bool {
diff --git a/src/armv8/a64.rs b/src/armv8/a64.rs
index 19440ba..a228027 100644
--- a/src/armv8/a64.rs
+++ b/src/armv8/a64.rs
@@ -10454,7 +10454,12 @@ impl Decoder<ARMv8> for InstDecoder {
                                     ];
                                 }
                                 0b100 => {
+                                    // quote C5.1.2:
+                                    // All encodings with L==1 and op0==0b0x are UNDEFINED, except
+                                    // for encodings in the area reserved for IMPLEMENTATION
+                                    // DEFINED use
                                     inst.opcode = Opcode::Invalid;
+                                    return Err(DecodeError::InvalidOpcode);
                                 }
                                 0b101 => {
                                     let Rt = word & 0b11111;
@@ -10484,6 +10489,7 @@ impl Decoder<ARMv8> for InstDecoder {
                                 }
                                 _ => {
                                     inst.opcode = Opcode::Invalid;
+                                    return Err(DecodeError::InvalidOpcode);
                                 }
                             }
                         }
diff --git a/tests/armv8/a64.rs b/tests/armv8/a64.rs
index a3ec96e..00f53e6 100644
--- a/tests/armv8/a64.rs
+++ b/tests/armv8/a64.rs
@@ -4753,6 +4753,11 @@ fn test_vec_shift() {
 }
 
 #[test]
+fn test_reserved() {
+    test_err([0x00, 0x00, 0x20, 0xd5], DecodeError::InvalidOpcode);
+}
+
+#[test]
 fn test_system() {
     const TESTS: &[([u8; 4], &'static str)] = &[
         ([0x00, 0x00, 0x08, 0xd5], "sys #0x0, c0, c0, #0x0, x0"),
-- 
cgit v1.1