[AArch64][WinCFI] Fix a crash due to missing seh directives #123993
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
hjyamauchi
commented
Jan 22, 2025
| StackOffset::getFixed(CFAOffset)); | ||
|
|
||
| return std::prev(MBBI); | ||
| } | ||
|
|
||
| // Get rid of the SEH code associated with the old instruction. |
There was a problem hiding this comment.
Note: This seh removing code needed to be moved past the above if statement because it would otherwise early-return and leave the seh removed but without readding one back, which would cause a missing seh crash.
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
hjyamauchi
commented
Jan 22, 2025
| ++MBBI; | ||
| // Also skip the SEH instruction, if needed |
There was a problem hiding this comment.
Note: This additional skip logic was needed to skip two instructions (the load/store and its seh instruction pair, if any) instead of one. Otherwise, the following emitFrameOffset call would add the load/store + seh instructions at a wrong position (it should add it after the load/store and its seh, not between the two).
hjyamauchi
force-pushed
the
arm64seh
branch
2 times, most recently
from
7206ed9 to
c51e257
Compare
|
@llvm/pr-subscribers-platform-windows @llvm/pr-subscribers-backend-aarch64 Author: Hiroshi Yamauchi (hjyamauchi) ChangesFull diff: https://github.com/llvm/llvm-project/pull/123993.diff 3 Files Affected:
diff --git a/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp b/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
index eabe64361938b4..a082a1ebe95bf8 100644
--- a/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
@@ -1491,13 +1491,6 @@ static MachineBasicBlock::iterator convertCalleeSaveRestoreToSPPrePostIncDec(
NewOpc = AArch64::LDRQpost;
break;
}
- // Get rid of the SEH code associated with the old instruction.
- if (NeedsWinCFI) {
- auto SEH = std::next(MBBI);
- if (AArch64InstrInfo::isSEHInstruction(*SEH))
- SEH->eraseFromParent();
- }
-
TypeSize Scale = TypeSize::getFixed(1), Width = TypeSize::getFixed(0);
int64_t MinOffset, MaxOffset;
bool Success = static_cast<const AArch64InstrInfo *>(TII)->getMemOpInfo(
@@ -1512,16 +1505,27 @@ static MachineBasicBlock::iterator convertCalleeSaveRestoreToSPPrePostIncDec(
CSStackSizeInc > MaxOffset * (int64_t)Scale.getFixedValue()) {
// If we are destroying the frame, make sure we add the increment after the
// last frame operation.
- if (FrameFlag == MachineInstr::FrameDestroy)
+ if (FrameFlag == MachineInstr::FrameDestroy) {
++MBBI;
+ // Also skip the SEH instruction, if needed
+ if (NeedsWinCFI && AArch64InstrInfo::isSEHInstruction(*MBBI))
+ ++MBBI;
+ }
emitFrameOffset(MBB, MBBI, DL, AArch64::SP, AArch64::SP,
StackOffset::getFixed(CSStackSizeInc), TII, FrameFlag,
- false, false, nullptr, EmitCFI,
+ false, NeedsWinCFI, HasWinCFI, EmitCFI,
StackOffset::getFixed(CFAOffset));
return std::prev(MBBI);
}
+ // Get rid of the SEH code associated with the old instruction.
+ if (NeedsWinCFI) {
+ auto SEH = std::next(MBBI);
+ if (AArch64InstrInfo::isSEHInstruction(*SEH))
+ SEH->eraseFromParent();
+ }
+
MachineInstrBuilder MIB = BuildMI(MBB, MBBI, DL, TII->get(NewOpc));
MIB.addReg(AArch64::SP, RegState::Define);
diff --git a/llvm/test/CodeGen/AArch64/stack-hazard-windows.ll b/llvm/test/CodeGen/AArch64/stack-hazard-windows.ll
index 2a034fe5e5290c..927d8b68c46be3 100644
--- a/llvm/test/CodeGen/AArch64/stack-hazard-windows.ll
+++ b/llvm/test/CodeGen/AArch64/stack-hazard-windows.ll
@@ -76,7 +76,9 @@ define i32 @fpr_csr_stackobj(double %x) "aarch64_pstate_sm_compatible" "frame-po
; CHECK1024: .seh_proc fpr_csr_stackobj
; CHECK1024-NEXT: // %bb.0: // %entry
; CHECK1024-NEXT: sub sp, sp, #1072
+; CHECK1024-NEXT: .seh_stackalloc 1072
; CHECK1024-NEXT: str x23, [sp] // 8-byte Folded Spill
+; CHECK1024-NEXT: .seh_save_reg x23, 0
; CHECK1024-NEXT: str x29, [sp, #8] // 8-byte Folded Spill
; CHECK1024-NEXT: .seh_save_reg x29, 8
; CHECK1024-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
@@ -105,7 +107,9 @@ define i32 @fpr_csr_stackobj(double %x) "aarch64_pstate_sm_compatible" "frame-po
; CHECK1024-NEXT: ldr x29, [sp, #8] // 8-byte Folded Reload
; CHECK1024-NEXT: .seh_save_reg x29, 8
; CHECK1024-NEXT: ldr x23, [sp] // 8-byte Folded Reload
+; CHECK1024-NEXT: .seh_save_reg x23, 0
; CHECK1024-NEXT: add sp, sp, #1072
+; CHECK1024-NEXT: .seh_stackalloc 1072
; CHECK1024-NEXT: .seh_endepilogue
; CHECK1024-NEXT: ret
; CHECK1024-NEXT: .seh_endfunclet
diff --git a/llvm/test/CodeGen/AArch64/wincfi-missing-seh-directives.ll b/llvm/test/CodeGen/AArch64/wincfi-missing-seh-directives.ll
new file mode 100644
index 00000000000000..568161dd5b43f3
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/wincfi-missing-seh-directives.ll
@@ -0,0 +1,86 @@
+; RUN: llc -mtriple=aarch64-windows %s --filetype obj
+; RUN: llc -mtriple=aarch64-windows %s --filetype asm -o - | FileCheck %s
+
+; Check that it doesn't crash and that each instruction in the
+; prologue has a corresponding seh directive.
+;
+; CHECK-NOT: error: Incorrect size for
+; CHECK: foo:
+; CHECK: .seh_proc foo
+; CHECK: sub sp, sp, #288
+; CHECK: .seh_stackalloc 288
+; CHECK: str x19, [sp] // 8-byte Folded Spill
+; CHECK: .seh_save_reg x19, 0
+; CHECK: str x21, [sp, #8] // 8-byte Folded Spill
+; CHECK: .seh_save_reg x21, 8
+; CHECK: stp x23, x24, [sp, #16] // 16-byte Folded Spill
+; CHECK: .seh_save_regp x23, 16
+; CHECK: stp x25, x26, [sp, #32] // 16-byte Folded Spill
+; CHECK: .seh_save_regp x25, 32
+; CHECK: stp x27, x28, [sp, #48] // 16-byte Folded Spill
+; CHECK: .seh_save_regp x27, 48
+; CHECK: stp x29, x30, [sp, #64] // 16-byte Folded Spill
+; CHECK: .seh_save_fplr 64
+; CHECK: sub sp, sp, #224
+; CHECK: .seh_stackalloc 224
+; CHECK: .seh_endprologue
+
+target datalayout = "e-m:w-p:64:64-i32:32-i64:64-i128:128-n32:64-S128-Fn32"
+target triple = "aarch64-unknown-windows-msvc19.42.34436"
+
+%swift.refcounted = type { ptr, i64 }
+%TScA_pSg = type <{ [16 x i8] }>
+%T5repro4TestVSg = type <{ [32 x i8] }>
+%T5repro4TestV = type <{ %TSS, %TSS }>
+%TSS = type <{ %Ts11_StringGutsV }>
+%Ts11_StringGutsV = type <{ %Ts13_StringObjectV }>
+%Ts13_StringObjectV = type <{ %Ts6UInt64V, ptr }>
+%Ts6UInt64V = type <{ i64 }>
+
+declare swiftcc ptr @swift_task_alloc()
+
+declare swifttailcc void @bar(ptr, ptr, i64, i64, i64, ptr, i64, i64, i64, i64, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr, i64, ptr)
+
+define swifttailcc void @foo(ptr %0, ptr swiftasync %1, ptr swiftself %2, ptr %3, ptr %._guts2._object._object, ptr %.rid4._guts._object._object, ptr %4, ptr %.idx8, ptr %.idx8._guts._object._object, ptr %5, ptr %.rid9._guts._object._object, ptr %6) {
+entry:
+ %7 = load i64, ptr null, align 8
+ %8 = load i64, ptr %3, align 8
+ %9 = getelementptr <{ %swift.refcounted, %TScA_pSg, %TSS, %T5repro4TestVSg, %T5repro4TestV, %TSS, %TSS, %TSS, %T5repro4TestV, %TSS, %T5repro4TestV, %T5repro4TestV, %TSS }>, ptr %2, i32 0, i32 2
+ %10 = load i64, ptr %9, align 8
+ %11 = load ptr, ptr %1, align 8
+ %12 = getelementptr <{ %swift.refcounted, %TScA_pSg, %TSS, %T5repro4TestVSg, %T5repro4TestV, %TSS, %TSS, %TSS, %T5repro4TestV, %TSS, %T5repro4TestV, %T5repro4TestV, %TSS }>, ptr %2, i32 0, i32 3
+ %13 = load i64, ptr %.rid9._guts._object._object, align 8
+ %14 = load i64, ptr %.idx8._guts._object._object, align 8
+ %15 = load i64, ptr %5, align 8
+ %16 = getelementptr { i64, i64, i64, i64 }, ptr %12, i32 0, i32 3
+ %17 = load i64, ptr %16, align 8
+ %18 = getelementptr <{ %swift.refcounted, %TScA_pSg, %TSS, %T5repro4TestVSg, %T5repro4TestV, %TSS, %TSS, %TSS, %T5repro4TestV, %TSS, %T5repro4TestV, %T5repro4TestV, %TSS }>, ptr %2, i32 0, i32 4
+ %19 = load i64, ptr %18, align 8
+ %.rid._guts._object._object = getelementptr %Ts13_StringObjectV, ptr %18, i32 0, i32 1
+ %20 = load ptr, ptr %.rid._guts._object._object, align 8
+ %21 = load i64, ptr %.rid4._guts._object._object, align 8
+ %22 = load i64, ptr %0, align 8
+ %23 = load ptr, ptr %6, align 8
+ %24 = load i64, ptr %2, align 8
+ %25 = load ptr, ptr %._guts2._object._object, align 8
+ %26 = getelementptr <{ %swift.refcounted, %TScA_pSg, %TSS, %T5repro4TestVSg, %T5repro4TestV, %TSS, %TSS, %TSS, %T5repro4TestV, %TSS, %T5repro4TestV, %T5repro4TestV, %TSS }>, ptr %2, i32 0, i32 7
+ %27 = load i64, ptr %26, align 8
+ %._guts3._object._object = getelementptr %Ts13_StringObjectV, ptr %26, i32 0, i32 1
+ %28 = load ptr, ptr %._guts3._object._object, align 8
+ %29 = getelementptr <{ %swift.refcounted, %TScA_pSg, %TSS, %T5repro4TestVSg, %T5repro4TestV, %TSS, %TSS, %TSS, %T5repro4TestV, %TSS, %T5repro4TestV, %T5repro4TestV, %TSS }>, ptr %2, i32 0, i32 8
+ %30 = load i64, ptr %29, align 8
+ %.idx5 = getelementptr %T5repro4TestV, ptr %29, i32 0, i32 1
+ %31 = load i64, ptr %.idx5, align 8
+ %.idx5._guts._object._object = getelementptr %Ts13_StringObjectV, ptr %.idx5, i32 0, i32 1
+ %32 = load ptr, ptr %.idx5._guts._object._object, align 8
+ %33 = getelementptr <{ %swift.refcounted, %TScA_pSg, %TSS, %T5repro4TestVSg, %T5repro4TestV, %TSS, %TSS, %TSS, %T5repro4TestV, %TSS, %T5repro4TestV, %T5repro4TestV, %TSS }>, ptr %2, i32 0, i32 9
+ %34 = load i64, ptr %33, align 8
+ %35 = load i64, ptr %4, align 8
+ %36 = load i64, ptr %.idx8, align 8
+ %37 = load i64, ptr %1, align 8
+ %38 = call swiftcc ptr @swift_task_alloc()
+ store ptr null, ptr %3, align 8
+ store ptr null, ptr %4, align 8
+ musttail call swifttailcc void @bar(ptr null, ptr swiftasync %.rid4._guts._object._object, i64 %7, i64 %8, i64 %10, ptr %5, i64 %13, i64 %14, i64 %15, i64 %17, i64 %19, ptr %20, i64 %21, ptr %.idx8, i64 %22, ptr %23, i64 %24, ptr %25, i64 %27, ptr %28, i64 %30, ptr %.idx8._guts._object._object, i64 %31, ptr %32, i64 %34, ptr %._guts2._object._object, i64 %35, ptr %2, i64 %36, ptr %1, i64 %37, ptr %0, i64 0, ptr null, i64 0, ptr null)
+ ret void
+}
|
compnerd
approved these changes
Jan 22, 2025
efriedma-quic
approved these changes
Jan 23, 2025
efriedma-quic
approved these changes
Jan 24, 2025
hjyamauchi
added a commit
to hjyamauchi/llvm-project
that referenced
this pull request
Jan 24, 2025
Cherry pick commit llvm#123993
hjyamauchi
added a commit
to hjyamauchi/llvm-project
that referenced
this pull request
Jan 24, 2025
Cherry pick commit llvm#123993
hjyamauchi
added a commit
to hjyamauchi/llvm-project
that referenced
this pull request
Jan 24, 2025
Cherry pick commit llvm#123993
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
#123808