[X86] Fix for offsets of CFA directives

`emitPrologue` may insert stack pointer adjustment in tail call optimized functions where the callee argument stack size is bigger than the caller's. In such a case, the adjustment must be taken into account when generating CFA directives.

Reviewed By: pengfei

Differential Revision: https://reviews.llvm.org/D143618

llvm/lib/Target/X86/X86FrameLowering.cpp

@@ -1645,14 +1645,16 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF,
       // Define the current CFA rule to use the provided offset.
       assert(StackSize);
       BuildCFI(MBB, MBBI, DL,
-               MCCFIInstruction::cfiDefCfaOffset(nullptr, -2 * stackGrowth),
+               MCCFIInstruction::cfiDefCfaOffset(
+                   nullptr, -2 * stackGrowth + (int)TailCallArgReserveSize),
                MachineInstr::FrameSetup);
 
       // Change the rule for the FramePtr to be an "offset" rule.
       unsigned DwarfFramePtr = TRI->getDwarfRegNum(MachineFramePtr, true);
       BuildCFI(MBB, MBBI, DL,
                MCCFIInstruction::createOffset(nullptr, DwarfFramePtr,
-                                              2 * stackGrowth),
+                                              2 * stackGrowth -
+                                                  (int)TailCallArgReserveSize),
                MachineInstr::FrameSetup);
     }
 

llvm/test/CodeGen/X86/tailcc-dwarf.ll

@@ -0,0 +1,47 @@
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -O0 --frame-pointer=non-leaf %s -o - | FileCheck %s
+
+%block = type { %blockheader, [0 x i64*] }
+%blockheader = type { i64 }
+
+define void @scanStackRoots(i32) {
+  ret void
+}
+
+define i32 @main(i32 %argc, i8** %argv) {
+entry:
+  %0 = call tailcc %block* @apply_rule_6870(%block* null, %block* null)
+  ret i32 0
+}
+
+define internal tailcc %block* @apply_rule_6870(%block* %0, %block* %1) {
+entry:
+  %2 = tail call tailcc %block* @sender12(%block* %0, %block* %1)
+  ret %block* null
+}
+
+define internal tailcc %block* @sender12(%block* %0, %block* %1) {
+; CHECK-LABEL: sender12:
+; CHECK: .cfi_startproc
+; CHECK: subq $8160, %rsp
+; CHECK: pushq %rbp
+; CHECK: .cfi_def_cfa_offset 8176
+; CHECK: .cfi_offset %rbp, -8176
+entry:
+  %a = alloca [1024 x i32]
+  %b = load [1024 x i32], [1024 x i32]* %a
+  call void @scanStackRoots(i32 1)
+  %2 = tail call tailcc %block* @apply_rule_6300(%block* %0, %block* %1, [1024 x i32] %b)
+  ret %block* %2
+}
+
+define internal tailcc %block* @apply_rule_6300(%block* %0, %block* %1, [1024 x i32] %2) {
+entry:
+  %3 = tail call tailcc %block* @sender4(%block* %0, %block* %1)
+  ret %block* %3
+}
+
+define internal tailcc %block* @sender4(%block* %0, %block* %1) {
+entry:
+  call void @scanStackRoots(i32 2)
+  ret %block* null
+}