- Add RETGUARD to clang for amd64. This security mechanism uses per-function
  random cookies to protect access to function return instructions, with the
  effect that the integrity of the return address is protected, and function
  return instructions are harder to use in ROP gadgets.

  On function entry the return address is combined with a per-function random
  cookie and stored in the stack frame. The integrity of this value is verified
  before function return, and if this check fails, the program aborts. In this way
  RETGUARD is an improved stack protector, since the cookies are per-function. The
  verification routine is constructed such that the binary space immediately
  before each ret instruction is padded with int03 instructions, which makes these
  return instructions difficult to use in ROP gadgets. In the kernel, this has the
  effect of removing approximately 50% of total ROP gadgets, and 15% of unique
  ROP gadgets compared to the 6.3 release kernel. Function epilogues are
  essentially gadget free, leaving only the polymorphic gadgets that result from
  jumping into the instruction stream partway through other instructions. Work to
  remove these gadgets will continue through other mechanisms.
- Improve the X86FixupGadgets pass

Index: lib/Target/X86/X86InstrCompiler.td
--- lib/Target/X86/X86InstrCompiler.td.orig
+++ lib/Target/X86/X86InstrCompiler.td
@@ -277,6 +277,25 @@ def MORESTACK_RET_RESTORE_R10 : I<0, Pseudo, (outs), (
 }
 
 //===----------------------------------------------------------------------===//
+// Pseudo instruction used by retguard
+
+// This is lowered to a JE 2; INT3; INT3. Prior to this pseudo should be a
+// compare instruction to ensure the retguard cookie is correct.
+// We use a pseudo here in order to avoid splitting the BB just before the return.
+// Splitting the BB and inserting a JE_1 over a new INT3 BB occasionally
+// resulted in incorrect code when a value from a byte register (CL) was
+// used as a return value. When emitted as a split BB, the single byte
+// register would sometimes be widened to 4 bytes, which would corrupt
+// the return value (ie mov %ecx, %eax instead of mov %cl, %al).
+let isCodeGenOnly = 1, hasNoSchedulingInfo = 1, Uses = [EFLAGS] in {
+def RETGUARD_JMP_TRAP: I<0, Pseudo, (outs), (ins), "", []>;
+}
+
+let isCodeGenOnly = 1, hasNoSchedulingInfo = 1 in {
+def JMP_TRAP: I<0, Pseudo, (outs), (ins), "", []>;
+}
+
+//===----------------------------------------------------------------------===//
 // Alias Instructions
 //===----------------------------------------------------------------------===//