[MemorySanitizer] Generate check lines for some vararg tests (NFC)

Use UTC_ARGS: --disable to skip the tests with many arguments.

5 files changed, 484 insertions, 117 deletions

diff --git a/llvm/test/Instrumentation/MemorySanitizer/LoongArch/vararg-loongarch64.ll b/llvm/test/Instrumentation/MemorySanitizer/LoongArch/vararg-loongarch64.ll
index e6d3a4b..4d4fc1b 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/LoongArch/vararg-loongarch64.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/LoongArch/vararg-loongarch64.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
 ; RUN: opt < %s -S -passes=msan 2>&1 | FileCheck %s
 
 target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n64-S128"
@@ -9,12 +10,36 @@ declare void @llvm.va_start(ptr) #2
 declare void @llvm.va_end(ptr) #2
 declare void @llvm.lifetime.end.p0(ptr nocapture) #1
 define i32 @foo(i32 %guard, ...) {
-; CHECK-LABEL: @foo
-; CHECK:    [[TMP1:%.*]] = load {{.*}} @__msan_va_arg_overflow_size_tls
-; CHECK:    [[TMP3:%.*]] = alloca {{.*}} [[TMP1]]
-; CHECK:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP3]], i8 0, i64 [[TMP1]], i1 false)
-; CHECK:    [[TMP4:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 800)
-; CHECK:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP3]], ptr align 8 @__msan_va_arg_tls, i64 [[TMP4]], i1 false)
+; CHECK-LABEL: define i32 @foo(
+; CHECK-SAME: i32 [[GUARD:%.*]], ...) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = alloca i8, i64 [[TMP1]], align 8
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP2]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 800)
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP2]], ptr align 8 @__msan_va_arg_tls, i64 [[TMP3]], i1 false)
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    [[VL:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = xor i64 [[TMP4]], 87960930222080
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP6]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = xor i64 [[TMP7]], 87960930222080
+; CHECK-NEXT:    [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP9]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.va_start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP10:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[TMP11]], align 8
+; CHECK-NEXT:    [[TMP13:%.*]] = ptrtoint ptr [[TMP12]] to i64
+; CHECK-NEXT:    [[TMP14:%.*]] = xor i64 [[TMP13]], 87960930222080
+; CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP15]], ptr align 8 [[TMP2]], i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @llvm.va_end.p0(ptr [[VL]])
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(ptr [[VL]])
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 0
 ;
   %vl = alloca ptr, align 8
   call void @llvm.lifetime.start.p0(ptr %vl)
@@ -27,11 +52,22 @@ define i32 @foo(i32 %guard, ...) {
 ;; Save the incoming shadow value from the arguments in the __msan_va_arg_tls
 ;; array.
 define i32 @bar() {
-; CHECK-LABEL: @bar
-; CHECK:    store i32 0, ptr @__msan_va_arg_tls, align 8
-; CHECK:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
-; CHECK:    store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
+; CHECK-LABEL: define i32 @bar() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
   %1 = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
@@ -40,15 +76,28 @@ define i32 @bar() {
 ;; Check multiple fixed arguments.
 declare i32 @foo2(i32 %g1, i32 %g2, ...)
 define i32 @bar2() {
-; CHECK-LABEL: @bar2
-; CHECK:    store i64 0, ptr @__msan_va_arg_tls, align 8
-; CHECK:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK:    store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
+; CHECK-LABEL: define i32 @bar2() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, i32, ...) @foo2(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
   %1 = call i32 (i32, i32, ...) @foo2(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
 
+; UTC_ARGS: --disable
+
 ;; Test that MSan doesn't generate code overflowing __msan_va_arg_tls when too many arguments are
 ;; passed to a variadic function.
 declare i64 @sum(i64 %n, ...)
diff --git a/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64.ll b/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64.ll
index 69a74a3..9f3f10e 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64.ll
@@ -1,9 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
 ; RUN: opt < %s -S -passes=msan 2>&1 | FileCheck %s
 
 target datalayout = "E-m:m-i8:8:32-i16:16:32-i64:64-n32:64-S128"
 target triple = "mips64--linux"
 
 define i32 @foo(i32 %guard, ...) {
+; CHECK-LABEL: define i32 @foo(
+; CHECK-SAME: i32 [[GUARD:%.*]], ...) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = alloca i8, i64 [[TMP1]], align 8
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP2]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 800)
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP2]], ptr align 8 @__msan_va_arg_tls, i64 [[TMP3]], i1 false)
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    [[VL:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = xor i64 [[TMP4]], 549755813888
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP6]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = xor i64 [[TMP7]], 549755813888
+; CHECK-NEXT:    [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP9]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.va_start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP10:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[TMP11]], align 8
+; CHECK-NEXT:    [[TMP13:%.*]] = ptrtoint ptr [[TMP12]] to i64
+; CHECK-NEXT:    [[TMP14:%.*]] = xor i64 [[TMP13]], 549755813888
+; CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP15]], ptr align 8 [[TMP2]], i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @llvm.va_end.p0(ptr [[VL]])
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(ptr [[VL]])
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 0
+;
   %vl = alloca ptr, align 8
   call void @llvm.lifetime.start.p0(ptr %vl)
   call void @llvm.va_start(ptr %vl)
@@ -12,23 +44,29 @@ define i32 @foo(i32 %guard, ...) {
   ret i32 0
 }
 
-; First, check allocation of the save area.
-
-; CHECK-LABEL: @foo
-; CHECK: [[A:%.*]] = load {{.*}} @__msan_va_arg_overflow_size_tls
-; CHECK: [[C:%.*]] = alloca {{.*}} [[A]]
-
-; CHECK: call void @llvm.memset.p0.i64(ptr align 8 [[C]], i8 0, i64 [[A]], i1 false)
-
-; CHECK: [[D:%.*]] = call i64 @llvm.umin.i64(i64 [[A]], i64 800)
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[C]], ptr align 8 @__msan_va_arg_tls, i64 [[D]], i1 false)
-
 declare void @llvm.lifetime.start.p0(ptr nocapture) #1
 declare void @llvm.va_start(ptr) #2
 declare void @llvm.va_end(ptr) #2
 declare void @llvm.lifetime.end.p0(ptr nocapture) #1
 
 define i32 @bar() {
+; CHECK-LABEL: define i32 @bar() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 4) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
@@ -36,23 +74,32 @@ define i32 @bar() {
 ; Save the incoming shadow value from the arguments in the __msan_va_arg_tls
 ; array.  The first argument is stored at position 4, since it's right
 ; justified.
-; CHECK-LABEL: @bar
-; CHECK: store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 4) to ptr), align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
-; CHECK: store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check multiple fixed arguments.
 declare i32 @foo2(i32 %g1, i32 %g2, ...)
 define i32 @bar2() {
+; CHECK-LABEL: define i32 @bar2() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, i32, ...) @foo2(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, i32, ...) @foo2(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar2
-; CHECK: store i64 0, ptr @__msan_va_arg_tls, align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
+
+; UTC_ARGS: --disable
 
 ; Test that MSan doesn't generate code overflowing __msan_va_arg_tls when too many arguments are
 ; passed to a variadic function.
diff --git a/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64el.ll b/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64el.ll
index b19da8e..41fb975 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64el.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/Mips/vararg-mips64el.ll
@@ -1,9 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
 ; RUN: opt < %s -S -passes=msan 2>&1 | FileCheck %s
 
 target datalayout = "e-m:m-i8:8:32-i16:16:32-i64:64-n32:64-S128"
 target triple = "mips64el--linux"
 
 define i32 @foo(i32 %guard, ...) {
+; CHECK-LABEL: define i32 @foo(
+; CHECK-SAME: i32 [[GUARD:%.*]], ...) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = alloca i8, i64 [[TMP1]], align 8
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP2]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 800)
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP2]], ptr align 8 @__msan_va_arg_tls, i64 [[TMP3]], i1 false)
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    [[VL:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = xor i64 [[TMP4]], 549755813888
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP6]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = xor i64 [[TMP7]], 549755813888
+; CHECK-NEXT:    [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP9]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.va_start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP10:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[TMP11]], align 8
+; CHECK-NEXT:    [[TMP13:%.*]] = ptrtoint ptr [[TMP12]] to i64
+; CHECK-NEXT:    [[TMP14:%.*]] = xor i64 [[TMP13]], 549755813888
+; CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP15]], ptr align 8 [[TMP2]], i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @llvm.va_end.p0(ptr [[VL]])
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(ptr [[VL]])
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 0
+;
   %vl = alloca ptr, align 8
   call void @llvm.lifetime.start.p0(ptr %vl)
   call void @llvm.va_start(ptr %vl)
@@ -12,46 +44,60 @@ define i32 @foo(i32 %guard, ...) {
   ret i32 0
 }
 
-; First, check allocation of the save area.
-
-; CHECK-LABEL: @foo
-; CHECK: [[A:%.*]] = load {{.*}} @__msan_va_arg_overflow_size_tls
-; CHECK: [[C:%.*]] = alloca {{.*}} [[A]]
-
-; CHECK: call void @llvm.memset.p0.i64(ptr align 8 [[C]], i8 0, i64 [[A]], i1 false)
-
-; CHECK: [[D:%.*]] = call i64 @llvm.umin.i64(i64 [[A]], i64 800)
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[C]], ptr align 8 @__msan_va_arg_tls, i64 [[D]], i1 false)
-
 declare void @llvm.lifetime.start.p0(ptr nocapture) #1
 declare void @llvm.va_start(ptr) #2
 declare void @llvm.va_end(ptr) #2
 declare void @llvm.lifetime.end.p0(ptr nocapture) #1
 
 define i32 @bar() {
+; CHECK-LABEL: define i32 @bar() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
 
 ; Save the incoming shadow value from the arguments in the __msan_va_arg_tls
 ; array.
-; CHECK-LABEL: @bar
-; CHECK: store i32 0, ptr @__msan_va_arg_tls, align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
-; CHECK: store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check multiple fixed arguments.
 declare i32 @foo2(i32 %g1, i32 %g2, ...)
 define i32 @bar2() {
+; CHECK-LABEL: define i32 @bar2() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, i32, ...) @foo2(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, i32, ...) @foo2(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar2
-; CHECK: store i64 0, ptr @__msan_va_arg_tls, align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
+; UTC_ARGS: --disable
 
 ; Test that MSan doesn't generate code overflowing __msan_va_arg_tls when too many arguments are
 ; passed to a variadic function.
diff --git a/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64.ll b/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64.ll
index 9351067..19b07e1 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64.ll
@@ -1,9 +1,47 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
 ; RUN: opt < %s -S -passes=msan 2>&1 | FileCheck %s
 
 target datalayout = "E-m:e-i64:64-n32:64"
 target triple = "powerpc64--linux"
 
 define i32 @foo(i32 %guard, ...) {
+; CHECK-LABEL: define i32 @foo(
+; CHECK-SAME: i32 [[GUARD:%.*]], ...) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = alloca i8, i64 [[TMP1]], align 8
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP2]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 800)
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP2]], ptr align 8 @__msan_va_arg_tls, i64 [[TMP3]], i1 false)
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    [[VL:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], -246290604621825
+; CHECK-NEXT:    [[TMP6:%.*]] = xor i64 [[TMP5]], 17592186044416
+; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 8796093022208
+; CHECK-NEXT:    [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP8]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP9:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP10:%.*]] = and i64 [[TMP9]], -246290604621825
+; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 17592186044416
+; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 8796093022208
+; CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP13]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.va_start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP14:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
+; CHECK-NEXT:    [[TMP17:%.*]] = ptrtoint ptr [[TMP16]] to i64
+; CHECK-NEXT:    [[TMP18:%.*]] = and i64 [[TMP17]], -246290604621825
+; CHECK-NEXT:    [[TMP19:%.*]] = xor i64 [[TMP18]], 17592186044416
+; CHECK-NEXT:    [[TMP20:%.*]] = add i64 [[TMP19]], 8796093022208
+; CHECK-NEXT:    [[TMP21:%.*]] = inttoptr i64 [[TMP20]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP21]], ptr align 8 [[TMP2]], i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @llvm.va_end.p0(ptr [[VL]])
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(ptr [[VL]])
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 0
+;
   %vl = alloca ptr, align 8
   call void @llvm.lifetime.start.p0(ptr %vl)
   call void @llvm.va_start(ptr %vl)
@@ -12,23 +50,29 @@ define i32 @foo(i32 %guard, ...) {
   ret i32 0
 }
 
-; First, check allocation of the save area.
-
-; CHECK-LABEL: @foo
-; CHECK: [[A:%.*]] = load {{.*}} @__msan_va_arg_overflow_size_tls
-; CHECK: [[C:%.*]] = alloca {{.*}} [[A]]
-
-; CHECK: call void @llvm.memset.p0.i64(ptr align 8 [[C]], i8 0, i64 [[A]], i1 false)
-
-; CHECK: [[D:%.*]] = call i64 @llvm.umin.i64(i64 [[A]], i64 800)
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[C]], ptr align 8 @__msan_va_arg_tls, i64 [[D]], i1 false)
-
 declare void @llvm.lifetime.start.p0(ptr nocapture) #1
 declare void @llvm.va_start(ptr) #2
 declare void @llvm.va_end(ptr) #2
 declare void @llvm.lifetime.end.p0(ptr nocapture) #1
 
 define i32 @bar() {
+; CHECK-LABEL: define i32 @bar() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 4) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
@@ -36,14 +80,22 @@ define i32 @bar() {
 ; Save the incoming shadow value from the arguments in the __msan_va_arg_tls
 ; array.  The first argument is stored at position 4, since it's right
 ; justified.
-; CHECK-LABEL: @bar
-; CHECK: store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 4) to ptr), align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
-; CHECK: store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check vector argument.
 define i32 @bar2() {
+; CHECK-LABEL: define i32 @bar2() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store <2 x i64> zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store <2 x i64> zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, <2 x i64> <i64 1, i64 2>)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, <2 x i64> <i64 1, i64 2>)
   ret i32 %1
 }
@@ -51,50 +103,110 @@ define i32 @bar2() {
 ; The vector is at offset 16 of parameter save area, but __msan_va_arg_tls
 ; corresponds to offset 8+ of parameter save area - so the offset from
 ; __msan_va_arg_tls is actually misaligned.
-; CHECK-LABEL: @bar2
-; CHECK: store <2 x i64> zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check i64 array.
 define i32 @bar4() {
+; CHECK-LABEL: define i32 @bar4() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store [2 x i64] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store [2 x i64] zeroinitializer, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, [2 x i64] [i64 1, i64 2])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, [2 x i64] [i64 1, i64 2])
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar4
-; CHECK: store [2 x i64] zeroinitializer, ptr @__msan_va_arg_tls, align 8
-; CHECK: store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check i128 array.
 define i32 @bar5() {
+; CHECK-LABEL: define i32 @bar5() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store [2 x i128] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store [2 x i128] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 40, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, [2 x i128] [i128 1, i128 2])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, [2 x i128] [i128 1, i128 2])
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar5
-; CHECK: store [2 x i128] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store {{.*}} 40, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check 8-aligned byval.
 define i32 @bar6(ptr %arg) {
+; CHECK-LABEL: define i32 @bar6(
+; CHECK-SAME: ptr [[ARG:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = and i64 [[TMP2]], -246290604621825
+; CHECK-NEXT:    [[TMP4:%.*]] = xor i64 [[TMP3]], 17592186044416
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[TMP4]], 8796093022208
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), i8 0, i64 16, i1 false)
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = and i64 [[TMP7]], -246290604621825
+; CHECK-NEXT:    [[TMP9:%.*]] = xor i64 [[TMP8]], 17592186044416
+; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[TMP9]], 8796093022208
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 @__msan_va_arg_tls, ptr align 8 [[TMP11]], i64 16, i1 false)
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 (i32, ...) @foo(i32 0, ptr byval([2 x i64]) align 8 [[ARG]])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP12]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, ptr byval([2 x i64]) align 8 %arg)
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar6
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 @__msan_va_arg_tls, ptr align 8 {{.*}}, i64 16, i1 false)
-; CHECK: store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check 16-aligned byval.
 define i32 @bar7(ptr %arg) {
+; CHECK-LABEL: define i32 @bar7(
+; CHECK-SAME: ptr [[ARG:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = and i64 [[TMP2]], -246290604621825
+; CHECK-NEXT:    [[TMP4:%.*]] = xor i64 [[TMP3]], 17592186044416
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[TMP4]], 8796093022208
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), i8 0, i64 32, i1 false)
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = and i64 [[TMP7]], -246290604621825
+; CHECK-NEXT:    [[TMP9:%.*]] = xor i64 [[TMP8]], 17592186044416
+; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[TMP9]], 8796093022208
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), ptr align 8 [[TMP11]], i64 32, i1 false)
+; CHECK-NEXT:    store i64 40, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 (i32, ...) @foo(i32 0, ptr byval([4 x i64]) align 16 [[ARG]])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP12]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, ptr byval([4 x i64]) align 16 %arg)
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar7
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), ptr align 8 {{.*}}, i64 32, i1 false)
-; CHECK: store {{.*}} 40, {{.*}} @__msan_va_arg_overflow_size_tls
 
+; UTC_ARGS: --disable
 
 ; Test that MSan doesn't generate code overflowing __msan_va_arg_tls when too many arguments are
 ; passed to a variadic function.
diff --git a/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64le.ll b/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64le.ll
index 4151f3b..1fe6385 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64le.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/PowerPC/vararg-ppc64le.ll
@@ -1,9 +1,47 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
 ; RUN: opt < %s -S -passes=msan 2>&1 | FileCheck %s
 
 target datalayout = "e-m:e-i64:64-n32:64"
 target triple = "powerpc64le--linux"
 
 define i32 @foo(i32 %guard, ...) {
+; CHECK-LABEL: define i32 @foo(
+; CHECK-SAME: i32 [[GUARD:%.*]], ...) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = alloca i8, i64 [[TMP1]], align 8
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP2]], i8 0, i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 800)
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP2]], ptr align 8 @__msan_va_arg_tls, i64 [[TMP3]], i1 false)
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    [[VL:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], -246290604621825
+; CHECK-NEXT:    [[TMP6:%.*]] = xor i64 [[TMP5]], 17592186044416
+; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 8796093022208
+; CHECK-NEXT:    [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP8]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP9:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP10:%.*]] = and i64 [[TMP9]], -246290604621825
+; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 17592186044416
+; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 8796093022208
+; CHECK-NEXT:    [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP13]], i8 0, i64 8, i1 false)
+; CHECK-NEXT:    call void @llvm.va_start.p0(ptr [[VL]])
+; CHECK-NEXT:    [[TMP14:%.*]] = ptrtoint ptr [[VL]] to i64
+; CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
+; CHECK-NEXT:    [[TMP17:%.*]] = ptrtoint ptr [[TMP16]] to i64
+; CHECK-NEXT:    [[TMP18:%.*]] = and i64 [[TMP17]], -246290604621825
+; CHECK-NEXT:    [[TMP19:%.*]] = xor i64 [[TMP18]], 17592186044416
+; CHECK-NEXT:    [[TMP20:%.*]] = add i64 [[TMP19]], 8796093022208
+; CHECK-NEXT:    [[TMP21:%.*]] = inttoptr i64 [[TMP20]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP21]], ptr align 8 [[TMP2]], i64 [[TMP1]], i1 false)
+; CHECK-NEXT:    call void @llvm.va_end.p0(ptr [[VL]])
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(ptr [[VL]])
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 0
+;
   %vl = alloca ptr, align 8
   call void @llvm.lifetime.start.p0(ptr %vl)
   call void @llvm.va_start(ptr %vl)
@@ -12,37 +50,51 @@ define i32 @foo(i32 %guard, ...) {
   ret i32 0
 }
 
-; First, check allocation of the save area.
-
-; CHECK-LABEL: @foo
-; CHECK: [[A:%.*]] = load {{.*}} @__msan_va_arg_overflow_size_tls
-; CHECK: [[C:%.*]] = alloca {{.*}} [[A]]
-
-; CHECK: call void @llvm.memset.p0.i64(ptr align 8 [[C]], i8 0, i64 [[A]], i1 false)
-
-; CHECK: [[D:%.*]] = call i64 @llvm.umin.i64(i64 [[A]], i64 800)
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[C]], ptr align 8 @__msan_va_arg_tls, i64 [[D]], i1 false)
-
 declare void @llvm.lifetime.start.p0(ptr nocapture) #1
 declare void @llvm.va_start(ptr) #2
 declare void @llvm.va_end(ptr) #2
 declare void @llvm.lifetime.end.p0(ptr nocapture) #1
 
 define i32 @bar() {
+; CHECK-LABEL: define i32 @bar() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, i32 1, i64 2, double 3.000000e+00)
   ret i32 %1
 }
 
 ; Save the incoming shadow value from the arguments in the __msan_va_arg_tls
 ; array.
-; CHECK-LABEL: @bar
-; CHECK: store i32 0, ptr @__msan_va_arg_tls, align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store i64 0, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 16) to ptr), align 8
-; CHECK: store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check vector argument.
 define i32 @bar2() {
+; CHECK-LABEL: define i32 @bar2() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store <2 x i64> zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store <2 x i64> zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 24, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, <2 x i64> <i64 1, i64 2>)
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, <2 x i64> <i64 1, i64 2>)
   ret i32 %1
 }
@@ -50,49 +102,110 @@ define i32 @bar2() {
 ; The vector is at offset 16 of parameter save area, but __msan_va_arg_tls
 ; corresponds to offset 8+ of parameter save area - so the offset from
 ; __msan_va_arg_tls is actually misaligned.
-; CHECK-LABEL: @bar2
-; CHECK: store <2 x i64> zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store {{.*}} 24, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check i64 array.
 define i32 @bar4() {
+; CHECK-LABEL: define i32 @bar4() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store [2 x i64] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store [2 x i64] zeroinitializer, ptr @__msan_va_arg_tls, align 8
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, [2 x i64] [i64 1, i64 2])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, [2 x i64] [i64 1, i64 2])
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar4
-; CHECK: store [2 x i64] zeroinitializer, ptr @__msan_va_arg_tls, align 8
-; CHECK: store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check i128 array.
 define i32 @bar5() {
+; CHECK-LABEL: define i32 @bar5() {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    store [2 x i128] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store [2 x i128] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT:    store i64 40, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 (i32, ...) @foo(i32 0, [2 x i128] [i128 1, i128 2])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, [2 x i128] [i128 1, i128 2])
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar5
-; CHECK: store [2 x i128] zeroinitializer, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), align 8
-; CHECK: store {{.*}} 40, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check 8-aligned byval.
 define i32 @bar6(ptr %arg) {
+; CHECK-LABEL: define i32 @bar6(
+; CHECK-SAME: ptr [[ARG:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = and i64 [[TMP2]], -246290604621825
+; CHECK-NEXT:    [[TMP4:%.*]] = xor i64 [[TMP3]], 17592186044416
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[TMP4]], 8796093022208
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), i8 0, i64 16, i1 false)
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = and i64 [[TMP7]], -246290604621825
+; CHECK-NEXT:    [[TMP9:%.*]] = xor i64 [[TMP8]], 17592186044416
+; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[TMP9]], 8796093022208
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 @__msan_va_arg_tls, ptr align 8 [[TMP11]], i64 16, i1 false)
+; CHECK-NEXT:    store i64 16, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 (i32, ...) @foo(i32 0, ptr byval([2 x i64]) align 8 [[ARG]])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP12]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, ptr byval([2 x i64]) align 8 %arg)
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar6
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 @__msan_va_arg_tls, ptr align 8 {{.*}}, i64 16, i1 false)
-; CHECK: store {{.*}} 16, {{.*}} @__msan_va_arg_overflow_size_tls
 
 ; Check 16-aligned byval.
 define i32 @bar7(ptr %arg) {
+; CHECK-LABEL: define i32 @bar7(
+; CHECK-SAME: ptr [[ARG:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    call void @llvm.donothing()
+; CHECK-NEXT:    store i32 0, ptr @__msan_param_tls, align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = and i64 [[TMP2]], -246290604621825
+; CHECK-NEXT:    [[TMP4:%.*]] = xor i64 [[TMP3]], 17592186044416
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[TMP4]], 8796093022208
+; CHECK-NEXT:    [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), i8 0, i64 32, i1 false)
+; CHECK-NEXT:    [[TMP7:%.*]] = ptrtoint ptr [[ARG]] to i64
+; CHECK-NEXT:    [[TMP8:%.*]] = and i64 [[TMP7]], -246290604621825
+; CHECK-NEXT:    [[TMP9:%.*]] = xor i64 [[TMP8]], 17592186044416
+; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[TMP9]], 8796093022208
+; CHECK-NEXT:    [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), ptr align 8 [[TMP11]], i64 32, i1 false)
+; CHECK-NEXT:    store i64 40, ptr @__msan_va_arg_overflow_size_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 (i32, ...) @foo(i32 0, ptr byval([4 x i64]) align 16 [[ARG]])
+; CHECK-NEXT:    [[_MSRET:%.*]] = load i32, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store i32 0, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    ret i32 [[TMP12]]
+;
   %1 = call i32 (i32, ...) @foo(i32 0, ptr byval([4 x i64]) align 16 %arg)
   ret i32 %1
 }
 
-; CHECK-LABEL: @bar7
-; CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 8 inttoptr (i64 add (i64 ptrtoint (ptr @__msan_va_arg_tls to i64), i64 8) to ptr), ptr align 8 {{.*}}, i64 32, i1 false)
-; CHECK: store {{.*}} 40, {{.*}} @__msan_va_arg_overflow_size_tls
+
+; UTC_ARGS: --disable
 
 ; Test that MSan doesn't generate code overflowing __msan_va_arg_tls when too many arguments are
 ; passed to a variadic function.