[X86,ARM] Retain split-stack prolog check for sibling calls

Summary:
If a routine with no stack frame makes a sibling call, we need to
preserve the stack space check even if the local stack frame is empty,
since the call target could be a "no-split" function (in which case
the linker needs to be able to fix up the prolog sequence in order to
switch to a larger stack).

This fixes PR37807.

Reviewers: cherry, javed.absar

Subscribers: srhines, llvm-commits

Differential Revision: https://reviews.llvm.org/D48444

llvm-svn: 335604

4 files changed, 72 insertions, 4 deletions

diff --git a/llvm/lib/Target/ARM/ARMFrameLowering.cpp b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
index f350f3e8b30..af983ce2606 100644
--- a/llvm/lib/Target/ARM/ARMFrameLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
@@ -2148,8 +2148,10 @@ void ARMFrameLowering::adjustForSegmentedStacks(
 
   uint64_t StackSize = MFI.getStackSize();
 
-  // Do not generate a prologue for functions with a stack of size zero
-  if (StackSize == 0)
+  // Do not generate a prologue for leaf functions with a stack of size zero.
+  // For non-leaf functions we have to allow for the possibility that the
+  // call is to a non-split function, as in PR37807.
+  if (StackSize == 0 && !MFI.hasTailCall())
     return;
 
   // Use R4 and R5 as scratch registers.
diff --git a/llvm/lib/Target/X86/X86FrameLowering.cpp b/llvm/lib/Target/X86/X86FrameLowering.cpp
index 18f62e06afa..a257ec41f75 100644
--- a/llvm/lib/Target/X86/X86FrameLowering.cpp
+++ b/llvm/lib/Target/X86/X86FrameLowering.cpp
@@ -2241,8 +2241,10 @@ void X86FrameLowering::adjustForSegmentedStacks(
   // prologue.
   StackSize = MFI.getStackSize();
 
-  // Do not generate a prologue for functions with a stack of size zero
-  if (StackSize == 0)
+  // Do not generate a prologue for leaf functions with a stack of size zero.
+  // For non-leaf functions we have to allow for the possibility that the
+  // call is to a non-split function, as in PR37807.
+  if (StackSize == 0 && !MFI.hasTailCall())
     return;
 
   MachineBasicBlock *allocMBB = MF.CreateMachineBasicBlock();
diff --git a/llvm/test/CodeGen/ARM/segmented-stacks.ll b/llvm/test/CodeGen/ARM/segmented-stacks.ll
index 4fe84faa17f..971fdd96606 100644
--- a/llvm/test/CodeGen/ARM/segmented-stacks.ll
+++ b/llvm/test/CodeGen/ARM/segmented-stacks.ll
@@ -246,4 +246,22 @@ define void @test_nostack() #0 {
 ; ARM-android-NOT:   bl __morestack
 }
 
+; Test to make sure that a morestack call is generated if there is a
+; sibling call, even if the function in question has no stack frame
+; (PR37807).
+
+declare i32 @callee(i32)
+
+define i32 @test_sibling_call_empty_frame(i32 %x) #0 {
+  %call = tail call i32 @callee(i32 %x) #0
+  ret i32 %call
+
+; ARM-linux:      test_sibling_call_empty_frame:
+; ARM-linux:      bl      __morestack
+
+; ARM-android:      test_sibling_call_empty_frame:
+; ARM-android:      bl      __morestack
+
+}
+
 attributes #0 = { "split-stack" }
diff --git a/llvm/test/CodeGen/X86/segmented-stacks.ll b/llvm/test/CodeGen/X86/segmented-stacks.ll
index a4861efac0e..588262bbb39 100644
--- a/llvm/test/CodeGen/X86/segmented-stacks.ll
+++ b/llvm/test/CodeGen/X86/segmented-stacks.ll
@@ -640,6 +640,52 @@ define void @test_nosplitstck() {
 	ret void
 }
 
+; Test to make sure that a morestack call is generated if there is a
+; sibling call, even if the function in question has no stack frame
+; (PR37807).
+
+declare i32 @callee(i32)
+
+define i32 @test_sibling_call_empty_frame(i32 %x) #0 {
+  %call = tail call i32 @callee(i32 %x) #0
+  ret i32 %call
+
+; X32-Linux-LABEL:       test_sibling_call_empty_frame:
+; X32-Linux:  calll __morestack
+
+; X64-Linux-LABEL:       test_sibling_call_empty_frame:
+; X64-Linux:  callq __morestack
+
+; X64-Linux-Large-LABEL:       test_sibling_call_empty_frame:
+; X64-Linux-Large:  callq *__morestack_addr(%rip)
+
+; X32ABI-LABEL:       test_sibling_call_empty_frame:
+; X32ABI:  callq __morestack
+
+; X32-Darwin-LABEL:      test_sibling_call_empty_frame:
+; X32-Darwin: calll ___morestack
+
+; X64-Darwin-LABEL:      test_sibling_call_empty_frame:
+; X64-Darwin: callq ___morestack
+
+; X32-MinGW-LABEL:       test_sibling_call_empty_frame:
+; X32-MinGW:  calll ___morestack
+
+; X64-MinGW-LABEL:       test_sibling_call_empty_frame:
+; X64-MinGW:  callq __morestack
+
+; X64-FreeBSD-LABEL:       test_sibling_call_empty_frame:
+; X64-FreeBSD:  callq __morestack
+
+; X32-DFlyBSD-LABEL:       test_sibling_call_empty_frame:
+; X32-DFlyBSD:  calll __morestack
+; X32-DFlyBSD-NEXT:  ret
+
+; X64-DFlyBSD-LABEL:       test_sibling_call_empty_frame:
+; X64-DFlyBSD:  callq __morestack
+
+}
+
 attributes #0 = { "split-stack" }
 
 ; X64-Linux-Large: .rodata