[PATCH] D23101: Align XRay return sleds to 2 byte boundaries

[Dean Michael Berris via llvm-commits]

dberris created this revision.
dberris added reviewers: rSerge, echristo, majnemer.
dberris added a subscriber: llvm-commits.
Herald added a subscriber: mehdi_amini.

This should ensure that we can atomically write two bytes (on top of the
retq and the one past it) and have those two bytes not straddle cache
lines.

https://reviews.llvm.org/D23101

Files:
  lib/Target/X86/X86MCInstLower.cpp
  test/CodeGen/X86/xray-attribute-instrumentation.ll

Index: test/CodeGen/X86/xray-attribute-instrumentation.ll
===================================================================
--- test/CodeGen/X86/xray-attribute-instrumentation.ll
+++ test/CodeGen/X86/xray-attribute-instrumentation.ll
@@ -8,6 +8,7 @@
 ; CHECK-LABEL: Ltmp0:
   ret i32 0
 ; CHECK-LABEL: Lxray_sled_1:
+; CHECK-NEXT:  .p2align 2, 0x90
 ; CHECK-NEXT:  retq
 ; CHECK-NEXT:  nopw %cs:512(%rax,%rax)
 }
Index: lib/Target/X86/X86MCInstLower.cpp
===================================================================
--- lib/Target/X86/X86MCInstLower.cpp
+++ lib/Target/X86/X86MCInstLower.cpp
@@ -1075,12 +1075,14 @@
   // We should emit the RET followed by sleds.
   //
   // .Lxray_sled_N:
+  //   .palign 2, ...
   //   ret  # or equivalent instruction
   //   # 10 bytes worth of noops
   //
   // This just makes sure that the alignment for the next instruction is 2.
   auto CurSled = OutContext.createTempSymbol("xray_sled_", true);
   OutStreamer->EmitLabel(CurSled);
+  OutStreamer->EmitCodeAlignment(4);
   unsigned OpCode = MI.getOperand(0).getImm();
   MCInst Ret;
   Ret.setOpcode(OpCode);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D23101.66610.patch
Type: text/x-patch
Size: 1110 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160803/de60d393/attachment.bin>