[SSP, 2/2] Create llvm.stackguard() intrinsic and lower it to LOAD_STACK_GUARD
With this change, ideally IR pass can always generate llvm.stackguard
call to get the stack guard; but for now there are still IR form stack
guard customizations around (see getIRStackGuard()). Future SSP
customization should go through LOAD_STACK_GUARD.
There is a behavior change: stack guard values are not CSEed anymore,
since we should never reuse the value in case that it has been spilled (and
corrupted). See ssp-guard-spill.ll. This also cause the change of stack
size and codegen in X86 and AArch64 test cases.
Ideally we'd like to know if the guard created in llvm.stackprotector() gets
spilled or not. If the value is spilled, discard the value and reload
stack guard; otherwise reuse the value. This can be done by teaching
register allocator to know how to rematerialize LOAD_STACK_GUARD and
force a rematerialization (which seems hard), or check for spilling in
expandPostRAPseudo. It only makes sense when the stack guard is a global
variable, which requires more instructions to load. Anyway, this seems to go out
of the scope of the current patch.
llvm-svn: 266806
diff --git a/llvm/test/CodeGen/AArch64/stack-guard-remat-bitcast.ll b/llvm/test/CodeGen/AArch64/stack-guard-remat-bitcast.ll
index ca30854..6f1515a 100644
--- a/llvm/test/CodeGen/AArch64/stack-guard-remat-bitcast.ll
+++ b/llvm/test/CodeGen/AArch64/stack-guard-remat-bitcast.ll
@@ -6,10 +6,14 @@
; CHECK: adrp [[R0:x[0-9]+]], ___stack_chk_guard@GOTPAGE
; CHECK: ldr [[R1:x[0-9]+]], {{\[}}[[R0]], ___stack_chk_guard@GOTPAGEOFF{{\]}}
+; Load the stack guard for the second time, just in case the previous value gets spilled.
+; CHECK: adrp [[GUARD_PAGE:x[0-9]+]], ___stack_chk_guard@GOTPAGE
; CHECK: ldr [[R2:x[0-9]+]], {{\[}}[[R1]]{{\]}}
; CHECK: stur [[R2]], {{\[}}x29, [[SLOT0:[0-9#\-]+]]{{\]}}
; CHECK: ldur [[R3:x[0-9]+]], {{\[}}x29, [[SLOT0]]{{\]}}
-; CHECK: sub [[R4:x[0-9]+]], [[R2]], [[R3]]
+; CHECK: ldr [[GUARD_ADDR:x[0-9]+]], {{\[}}[[GUARD_PAGE]], ___stack_chk_guard@GOTPAGEOFF{{\]}}
+; CHECK: ldr [[GUARD:x[0-9]+]], {{\[}}[[GUARD_ADDR]]{{\]}}
+; CHECK: sub [[R4:x[0-9]+]], [[GUARD]], [[R3]]
; CHECK: cbnz [[R4]], LBB
define i32 @test_stack_guard_remat2() {
diff --git a/llvm/test/CodeGen/ARM/warn-stack.ll b/llvm/test/CodeGen/ARM/warn-stack.ll
index 474dc1d..6e819e4 100644
--- a/llvm/test/CodeGen/ARM/warn-stack.ll
+++ b/llvm/test/CodeGen/ARM/warn-stack.ll
@@ -12,7 +12,7 @@
ret void
}
-; CHECK: warning: stack size limit exceeded (96) in warn
+; CHECK: warning: stack size limit exceeded (92) in warn
define void @warn() nounwind ssp {
entry:
%buffer = alloca [80 x i8], align 1
diff --git a/llvm/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll b/llvm/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll
index c0b078a..8edaf3f 100644
--- a/llvm/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll
+++ b/llvm/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll
@@ -12,7 +12,7 @@
; movq %rax, %rsp
; CHECK-LABEL: @foo
-; CHECK: movq -40(%rbp), %rsp
+; CHECK: movq -{{[0-9]+}}(%rbp), %rsp
entry:
%s1_addr = alloca i8* ; <i8**> [#uses=2]
diff --git a/llvm/test/CodeGen/X86/StackColoring.ll b/llvm/test/CodeGen/X86/StackColoring.ll
index 634f66a..14778a5 100644
--- a/llvm/test/CodeGen/X86/StackColoring.ll
+++ b/llvm/test/CodeGen/X86/StackColoring.ll
@@ -243,8 +243,8 @@
; Adopt the test from Transforms/Inline/array_merge.ll'
;CHECK-LABEL: array_merge:
-;YESCOLOR: subq $816, %rsp
-;NOCOLOR: subq $1616, %rsp
+;YESCOLOR: subq $808, %rsp
+;NOCOLOR: subq $1608, %rsp
define void @array_merge() nounwind ssp {
entry:
%A.i1 = alloca [100 x i32], align 4
diff --git a/llvm/test/CodeGen/X86/dynamic-allocas-VLAs.ll b/llvm/test/CodeGen/X86/dynamic-allocas-VLAs.ll
index 7e30192..71e5892 100644
--- a/llvm/test/CodeGen/X86/dynamic-allocas-VLAs.ll
+++ b/llvm/test/CodeGen/X86/dynamic-allocas-VLAs.ll
@@ -60,12 +60,10 @@
; CHECK: _t3
; CHECK: pushq %rbp
; CHECK: movq %rsp, %rbp
-; CHECK: pushq %rbx
; CHECK-NOT: andq $-{{[0-9]+}}, %rsp
; CHECK: subq ${{[0-9]+}}, %rsp
;
-; CHECK: leaq -{{[0-9]+}}(%rbp), %rsp
-; CHECK: popq %rbx
+; CHECK: movq %rbp, %rsp
; CHECK: popq %rbp
}
@@ -85,7 +83,6 @@
; CHECK: _t4
; CHECK: pushq %rbp
; CHECK: movq %rsp, %rbp
-; CHECK: pushq %r14
; CHECK: pushq %rbx
; CHECK: andq $-32, %rsp
; CHECK: subq ${{[0-9]+}}, %rsp
@@ -95,9 +92,8 @@
; CHECK: leaq {{[0-9]*}}(%rbx), %rdx
; CHECK: callq _t4_helper
;
-; CHECK: leaq -16(%rbp), %rsp
+; CHECK: leaq -{{[0-9]+}}(%rbp), %rsp
; CHECK: popq %rbx
-; CHECK: popq %r14
; CHECK: popq %rbp
}
diff --git a/llvm/test/CodeGen/X86/fast-isel-stackcheck.ll b/llvm/test/CodeGen/X86/fast-isel-stackcheck.ll
index 3b7318f..1398b30 100644
--- a/llvm/test/CodeGen/X86/fast-isel-stackcheck.ll
+++ b/llvm/test/CodeGen/X86/fast-isel-stackcheck.ll
@@ -8,7 +8,7 @@
; CHECK-LABEL: foo:
; CHECK: movq ___stack_chk_guard@GOTPCREL(%rip), %rax
-; CHECK-NOT: movq ___stack_chk_guard@GOTPCREL(%rip), %rax
+; CHECK: movq ___stack_chk_guard@GOTPCREL(%rip), %rax
define void @foo() #0 {
entry:
%_tags = alloca [3 x i32], align 4
@@ -16,8 +16,10 @@
}
; CHECK-LABEL: bar:
-; CHECK: movq ___stack_chk_guard@GOTPCREL(%rip), %rax
-; CHECK: movq ___stack_chk_guard@GOTPCREL(%rip), %rax
+; CHECK: movq ___stack_chk_guard@GOTPCREL(%rip), %{{r.x}}
+; CHECK-DAG: movq ___stack_chk_guard@GOTPCREL(%rip), %[[GUARD:r.x]]
+; CHECK-DAG: movq {{[0-9]+}}(%rsp), %[[CANARY:r.x]]
+; CHECK: subq %[[CANARY]], %[[GUARD]]
define void @bar() #1 {
entry:
%vt = alloca [2 x double], align 16
diff --git a/llvm/test/CodeGen/X86/ssp-guard-spill.ll b/llvm/test/CodeGen/X86/ssp-guard-spill.ll
new file mode 100644
index 0000000..7364dee
--- /dev/null
+++ b/llvm/test/CodeGen/X86/ssp-guard-spill.ll
@@ -0,0 +1,54 @@
+; RUN: llc < %s | FileCheck %s
+target triple = "x86_64-apple-macosx10.4.0"
+
+; __stack_chk_guard must be loaded for twice, once for setting up the canary,
+; another time for performing the check. It is because if we reuse the same
+; stack guard value, it may get spilled to the stack, then the for loop may
+; corrupt it.
+;
+; bool Bar(int*);
+; bool Foo(int n) {
+; int a[10];
+; for (int i = 0; i < n; i++) {
+; a[i] = 0;
+; }
+; return Bar(a);
+; }
+;
+; CHECK: movq ___stack_chk_guard
+; CHECK: movq ___stack_chk_guard
+define zeroext i1 @_Z3Fooi(i32 %n) sspstrong {
+entry:
+ %n.addr = alloca i32, align 4
+ %a = alloca [10 x i32], align 16
+ %i = alloca i32, align 4
+ store i32 %n, i32* %n.addr, align 4
+ store i32 0, i32* %i, align 4
+ br label %for.cond
+
+for.cond: ; preds = %for.inc, %entry
+ %tmp = load i32, i32* %i, align 4
+ %tmp1 = load i32, i32* %n.addr, align 4
+ %cmp = icmp slt i32 %tmp, %tmp1
+ br i1 %cmp, label %for.body, label %for.end
+
+for.body: ; preds = %for.cond
+ %tmp2 = load i32, i32* %i, align 4
+ %idxprom = sext i32 %tmp2 to i64
+ %arrayidx = getelementptr inbounds [10 x i32], [10 x i32]* %a, i64 0, i64 %idxprom
+ store i32 0, i32* %arrayidx, align 4
+ br label %for.inc
+
+for.inc: ; preds = %for.body
+ %tmp3 = load i32, i32* %i, align 4
+ %inc = add nsw i32 %tmp3, 1
+ store i32 %inc, i32* %i, align 4
+ br label %for.cond
+
+for.end: ; preds = %for.cond
+ %arraydecay = getelementptr inbounds [10 x i32], [10 x i32]* %a, i32 0, i32 0
+ %call = call zeroext i1 @_Z3BarPi(i32* %arraydecay)
+ ret i1 %call
+}
+
+declare zeroext i1 @_Z3BarPi(i32*)
diff --git a/llvm/test/CodeGen/X86/warn-stack.ll b/llvm/test/CodeGen/X86/warn-stack.ll
index aa09ad8..7353d07 100644
--- a/llvm/test/CodeGen/X86/warn-stack.ll
+++ b/llvm/test/CodeGen/X86/warn-stack.ll
@@ -12,7 +12,7 @@
ret void
}
-; CHECK: warning: stack size limit exceeded (104) in warn
+; CHECK: warning: stack size limit exceeded (88) in warn
define void @warn() nounwind ssp {
entry:
%buffer = alloca [80 x i8], align 1