Skip to content

Commit 77f2430

Browse files
staticfloattstellar
staticfloat
authored and
tstellar
committed
[X86] Don't clobber EBX in stackprobes
On X86, the stackprobe emission code chooses the `R11D` register, which is illegal on i686. This ends up wrapping around to `EBX`, which does not get properly callee-saved within the stack probing prologue, clobbering the register for the callers. We fix this by explicitly using `EAX` as the stack probe register. Reviewed By: pengfei Differential Revision: https://reviews.llvm.org/D109203 (cherry picked from commit ae8507b)
1 parent 1ff9aa2 commit 77f2430

File tree

2 files changed

+64
-41
lines changed

2 files changed

+64
-41
lines changed

‎llvm/lib/Target/X86/X86FrameLowering.cpp

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -671,7 +671,9 @@ void X86FrameLowering::emitStackProbeInlineGenericLoop(
671671
MF.insert(MBBIter, testMBB);
672672
MF.insert(MBBIter, tailMBB);
673673

674-
Register FinalStackProbed = Uses64BitFramePtr ? X86::R11 : X86::R11D;
674+
Register FinalStackProbed = Uses64BitFramePtr ? X86::R11
675+
: Is64Bit ? X86::R11D
676+
: X86::EAX;
675677
BuildMI(MBB, MBBI, DL, TII.get(TargetOpcode::COPY), FinalStackProbed)
676678
.addReg(StackPtr)
677679
.setMIFlag(MachineInstr::FrameSetup);
@@ -1092,7 +1094,9 @@ void X86FrameLowering::BuildStackAlignAND(MachineBasicBlock &MBB,
10921094
MF.insert(MBBIter, bodyMBB);
10931095
MF.insert(MBBIter, footMBB);
10941096
const unsigned MovMIOpc = Is64Bit ? X86::MOV64mi32 : X86::MOV32mi;
1095-
Register FinalStackProbed = Uses64BitFramePtr ? X86::R11 : X86::R11D;
1097+
Register FinalStackProbed = Uses64BitFramePtr ? X86::R11
1098+
: Is64Bit ? X86::R11D
1099+
: X86::EAX;
10961100

10971101
// Setup entry block
10981102
{

‎llvm/test/CodeGen/X86/stack-clash-large.ll

Lines changed: 58 additions & 39 deletions
Original file line numberDiff line numberDiff line change
@@ -1,45 +1,64 @@
1-
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --no_x86_scrub_sp
2-
; RUN: llc -mtriple=x86_64-linux-android < %s | FileCheck -check-prefix=CHECK-X86-64 %s
3-
; RUN: llc -mtriple=i686-linux-android < %s | FileCheck -check-prefix=CHECK-X86-32 %s
1+
; RUN: llc -mtriple=x86_64-linux-android < %s | FileCheck -check-prefix=CHECK-X64 %s
2+
; RUN: llc -mtriple=i686-linux-android < %s | FileCheck -check-prefix=CHECK-X86 %s
3+
; RUN: llc -mtriple=x86_64-linux-gnux32 < %s | FileCheck -check-prefix=CHECK-X32 %s
44

55
define i32 @foo() local_unnamed_addr #0 {
6-
; CHECK-X86-64-LABEL: foo:
7-
; CHECK-X86-64: # %bb.0:
8-
; CHECK-X86-64-NEXT: movq %rsp, %r11
9-
; CHECK-X86-64-NEXT: subq $69632, %r11 # imm = 0x11000
10-
; CHECK-X86-64-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
11-
; CHECK-X86-64-NEXT: subq $4096, %rsp # imm = 0x1000
12-
; CHECK-X86-64-NEXT: movq $0, (%rsp)
13-
; CHECK-X86-64-NEXT: cmpq %r11, %rsp
14-
; CHECK-X86-64-NEXT: jne .LBB0_1
15-
; CHECK-X86-64-NEXT: # %bb.2:
16-
; CHECK-X86-64-NEXT: subq $2248, %rsp # imm = 0x8C8
17-
; CHECK-X86-64-NEXT: .cfi_def_cfa_offset 71888
18-
; CHECK-X86-64-NEXT: movl $1, 264(%rsp)
19-
; CHECK-X86-64-NEXT: movl $1, 28664(%rsp)
20-
; CHECK-X86-64-NEXT: movl -128(%rsp), %eax
21-
; CHECK-X86-64-NEXT: addq $71880, %rsp # imm = 0x118C8
22-
; CHECK-X86-64-NEXT: .cfi_def_cfa_offset 8
23-
; CHECK-X86-64-NEXT: retq
6+
; CHECK-X64-LABEL: foo:
7+
; CHECK-X64: # %bb.0:
8+
; CHECK-X64-NEXT: movq %rsp, %r11
9+
; CHECK-X64-NEXT: subq $69632, %r11 # imm = 0x11000
10+
; CHECK-X64-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
11+
; CHECK-X64-NEXT: subq $4096, %rsp # imm = 0x1000
12+
; CHECK-X64-NEXT: movq $0, (%rsp)
13+
; CHECK-X64-NEXT: cmpq %r11, %rsp
14+
; CHECK-X64-NEXT: jne .LBB0_1
15+
; CHECK-X64-NEXT: # %bb.2:
16+
; CHECK-X64-NEXT: subq $2248, %rsp # imm = 0x8C8
17+
; CHECK-X64-NEXT: .cfi_def_cfa_offset 71888
18+
; CHECK-X64-NEXT: movl $1, 264(%rsp)
19+
; CHECK-X64-NEXT: movl $1, 28664(%rsp)
20+
; CHECK-X64-NEXT: movl -128(%rsp), %eax
21+
; CHECK-X64-NEXT: addq $71880, %rsp # imm = 0x118C8
22+
; CHECK-X64-NEXT: .cfi_def_cfa_offset 8
23+
; CHECK-X64-NEXT: retq
2424
;
25-
; CHECK-X86-32-LABEL: foo:
26-
; CHECK-X86-32: # %bb.0:
27-
; CHECK-X86-32-NEXT: movl %esp, %r11d
28-
; CHECK-X86-32-NEXT: subl $69632, %r11d # imm = 0x11000
29-
; CHECK-X86-32-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
30-
; CHECK-X86-32-NEXT: subl $4096, %esp # imm = 0x1000
31-
; CHECK-X86-32-NEXT: movl $0, (%esp)
32-
; CHECK-X86-32-NEXT: cmpl %r11d, %esp
33-
; CHECK-X86-32-NEXT: jne .LBB0_1
34-
; CHECK-X86-32-NEXT: # %bb.2:
35-
; CHECK-X86-32-NEXT: subl $2380, %esp # imm = 0x94C
36-
; CHECK-X86-32-NEXT: .cfi_def_cfa_offset 72016
37-
; CHECK-X86-32-NEXT: movl $1, 392(%esp)
38-
; CHECK-X86-32-NEXT: movl $1, 28792(%esp)
39-
; CHECK-X86-32-NEXT: movl (%esp), %eax
40-
; CHECK-X86-32-NEXT: addl $72012, %esp # imm = 0x1194C
41-
; CHECK-X86-32-NEXT: .cfi_def_cfa_offset 4
42-
; CHECK-X86-32-NEXT: retl
25+
; CHECK-X86-LABEL: foo:
26+
; CHECK-X86: # %bb.0:
27+
; CHECK-X86-NEXT: movl %esp, %eax
28+
; CHECK-X86-NEXT: subl $69632, %eax # imm = 0x11000
29+
; CHECK-X86-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
30+
; CHECK-X86-NEXT: subl $4096, %esp # imm = 0x1000
31+
; CHECK-X86-NEXT: movl $0, (%esp)
32+
; CHECK-X86-NEXT: cmpl %eax, %esp
33+
; CHECK-X86-NEXT: jne .LBB0_1
34+
; CHECK-X86-NEXT: # %bb.2:
35+
; CHECK-X86-NEXT: subl $2380, %esp # imm = 0x94C
36+
; CHECK-X86-NEXT: .cfi_def_cfa_offset 72016
37+
; CHECK-X86-NEXT: movl $1, 392(%esp)
38+
; CHECK-X86-NEXT: movl $1, 28792(%esp)
39+
; CHECK-X86-NEXT: movl (%esp), %eax
40+
; CHECK-X86-NEXT: addl $72012, %esp # imm = 0x1194C
41+
; CHECK-X86-NEXT: .cfi_def_cfa_offset 4
42+
; CHECK-X86-NEXT: retl
43+
;
44+
; CHECK-X32-LABEL: foo:
45+
; CHECK-X32: # %bb.0:
46+
; CHECK-X32-NEXT: movl %esp, %r11d
47+
; CHECK-X32-NEXT: subl $69632, %r11d # imm = 0x11000
48+
; CHECK-X32-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
49+
; CHECK-X32-NEXT: subl $4096, %esp # imm = 0x1000
50+
; CHECK-X32-NEXT: movq $0, (%esp)
51+
; CHECK-X32-NEXT: cmpl %r11d, %esp
52+
; CHECK-X32-NEXT: jne .LBB0_1
53+
; CHECK-X32-NEXT: # %bb.2:
54+
; CHECK-X32-NEXT: subl $2248, %esp # imm = 0x8C8
55+
; CHECK-X32-NEXT: .cfi_def_cfa_offset 71888
56+
; CHECK-X32-NEXT: movl $1, 264(%esp)
57+
; CHECK-X32-NEXT: movl $1, 28664(%esp)
58+
; CHECK-X32-NEXT: movl -128(%esp), %eax
59+
; CHECK-X32-NEXT: addl $71880, %esp # imm = 0x118C8
60+
; CHECK-X32-NEXT: .cfi_def_cfa_offset 8
61+
; CHECK-X32-NEXT: retq
4362
%a = alloca i32, i64 18000, align 16
4463
%b0 = getelementptr inbounds i32, i32* %a, i64 98
4564
%b1 = getelementptr inbounds i32, i32* %a, i64 7198

0 commit comments

Comments
 (0)