Fixed a bug where diassembling an instruction that had a prefix would cause LLVM to identify a 1-byte instruction, but then upon querying it for that 1-byte instruction would cause an undefined opcode.

llvm-svn: 189698

2 files changed, 68 insertions, 7 deletions

diff --git a/llvm/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c b/llvm/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
index bb195ee3140..b63fd5aee3c 100644
--- a/llvm/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
+++ b/llvm/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c
@@ -314,20 +314,22 @@ static int readPrefixes(struct InternalInstruction* insn) {
   while (isPrefix) {
     prefixLocation = insn->readerCursor;
 
+    /* If we fail reading prefixes, just stop here and let the opcode reader deal with it */
     if (consumeByte(insn, &byte))
-      return -1;
+      break;
 
     /*
      * If the byte is a LOCK/REP/REPNE prefix and not a part of the opcode, then
      * break and let it be disassembled as a normal "instruction".
      */
+    if (insn->readerCursor - 1 == insn->startLocation && byte == 0xf0)
+      break;
+
+    uint8_t nextByte;
     if (insn->readerCursor - 1 == insn->startLocation
-        && (byte == 0xf0 || byte == 0xf2 || byte == 0xf3)) {
-      uint8_t nextByte;
-      if (byte == 0xf0)
-        break;
-      if (lookAtByte(insn, &nextByte))
-        return -1;
+        && (byte == 0xf2 || byte == 0xf3)
+        && !lookAtByte(insn, &nextByte))
+    {
       /*
        * If the byte is 0xf2 or 0xf3, and any of the following conditions are
        * met:
diff --git a/llvm/test/MC/Disassembler/X86/prefixes.txt b/llvm/test/MC/Disassembler/X86/prefixes.txt
new file mode 100644
index 00000000000..56596e38751
--- /dev/null
+++ b/llvm/test/MC/Disassembler/X86/prefixes.txt
@@ -0,0 +1,59 @@
+# RUN: llvm-mc --disassemble %s -triple=x86_64 | FileCheck %s
+
+# CHECK: lock
+# CHECK-NEXT:	orl	$16, %fs:776
+0xf0 0x64 0x83 0x0c 0x25 0x08 0x03 0x00 0x00 0x10
+
+# CHECK: movq	%fs:768, %rdi
+0x64 0x48 0x8b 0x3c 0x25 0x00 0x03 0x00 0x00
+
+# CHECK: rep
+# CHECK-NEXT:		stosq
+0xf3 0x48 0xab
+
+# CHECK: rep
+# CHECK-NEXT:		stosl
+0xf3 0x67 0x48 0xab
+
+# CHECK: movl 32(%rbp), %eax
+0x8b 0x45 0x20
+
+# CHECK: movl %es:32(%rbp), %eax
+0x26 0x8b 0x45 0x20
+
+# CHECK: movl %es:32(%rbp), %eax
+0x2e 0x26 0x8b 0x45 0x20
+
+# Test that multiple prefixes stack.
+#    (todo- the correct disassembly is actually more like "es movl %cs:32(%rbp), %eax"
+#    but we don't support that)
+# CHECK: movl %cs:32(%rbp), %eax
+0x26 0x2e 0x8b 0x45 0x20
+
+# Test that 0xf3 as part of the opcode works.
+# CHECK: cvtdq2pd	(%rax), %xmm0
+0xf3 0x0f 0xe6 0x00
+
+# CHECK: pause
+0xf3 0x90
+
+# CHECK: nop
+0x90
+
+# CHECK: 		lock
+# CHECK-NEXT:	nop
+0xf0 0x90
+
+# Test that multiple redundant prefixes work (redundant, but valid x86).
+# CHECK: rep
+# CHECK-NEXT: rep
+# CHECK-NEXT: stosq
+0xf3 0xf3 0x48 0xab
+
+# Test that a prefix on it's own works. It's debatable as to if this is 
+# something that is considered valid, but however as LLVM's own disassembler
+# has decided to disassemble prefixes as being separate opcodes, it therefore 
+# should be capable of re-consuming it's own output.
+# CHECK: rep
+0xf3
+# ***IMPORTANT ^-- this must be at the end of the file to be a valid test ***