diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index b2e173dec9fe1a..bd4fd84441d891 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -288,16 +288,20 @@ Value *InstCombinerImpl::simplifyMaskedLoad(IntrinsicInst &II) {
 
   // If the mask is all ones or undefs, this is a plain vector load of the 1st
   // argument.
-  if (maskIsAllOneOrUndef(II.getArgOperand(2)))
-    return Builder.CreateAlignedLoad(II.getType(), LoadPtr, Alignment,
-                                     "unmaskedload");
+  if (maskIsAllOneOrUndef(II.getArgOperand(2))) {
+    LoadInst *L = Builder.CreateAlignedLoad(II.getType(), LoadPtr, Alignment,
+                                            "unmaskedload");
+    L->copyMetadata(II);
+    return L;
+  }
 
   // If we can unconditionally load from this address, replace with a
   // load/select idiom. TODO: use DT for context sensitive query
   if (isDereferenceablePointer(LoadPtr, II.getType(),
                                II.getModule()->getDataLayout(), &II, nullptr)) {
-    Value *LI = Builder.CreateAlignedLoad(II.getType(), LoadPtr, Alignment,
-                                         "unmaskedload");
+    LoadInst *LI = Builder.CreateAlignedLoad(II.getType(), LoadPtr, Alignment,
+                                             "unmaskedload");
+    LI->copyMetadata(II);
     return Builder.CreateSelect(II.getArgOperand(2), LI, II.getArgOperand(3));
   }
 
@@ -320,7 +324,10 @@ Instruction *InstCombinerImpl::simplifyMaskedStore(IntrinsicInst &II) {
   if (ConstMask->isAllOnesValue()) {
     Value *StorePtr = II.getArgOperand(1);
     Align Alignment = cast<ConstantInt>(II.getArgOperand(2))->getAlignValue();
-    return new StoreInst(II.getArgOperand(0), StorePtr, false, Alignment);
+    StoreInst *S =
+        new StoreInst(II.getArgOperand(0), StorePtr, false, Alignment);
+    S->copyMetadata(II);
+    return S;
   }
 
   if (isa<ScalableVectorType>(ConstMask->getType()))
diff --git a/llvm/test/Transforms/InstCombine/masked_intrinsics_keep_metadata.ll b/llvm/test/Transforms/InstCombine/masked_intrinsics_keep_metadata.ll
new file mode 100644
index 00000000000000..d2ab4e8cc59108
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/masked_intrinsics_keep_metadata.ll
@@ -0,0 +1,48 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S -instcombine < %s | FileCheck %s
+
+@g0 = global <4 x i32> zeroinitializer, align 16
+
+define inreg <4 x i32> @mload1(<4 x i32>* nocapture readonly %a0) #0 {
+; CHECK-LABEL: @mload1(
+; CHECK-NEXT:  b0:
+; CHECK-NEXT:    [[UNMASKEDLOAD:%.*]] = load <4 x i32>, <4 x i32>* [[A0:%.*]], align 16, !tbaa [[TBAA0:![0-9]+]]
+; CHECK-NEXT:    ret <4 x i32> [[UNMASKEDLOAD]]
+;
+b0:
+  %v0 = call <4 x i32> @llvm.masked.load.v4i1.p0v4i1(<4 x i32>* %a0, i32 16, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> undef), !tbaa !0
+  ret <4 x i32> %v0
+}
+
+define inreg <4 x i32> @mload2() #0 {
+; CHECK-LABEL: @mload2(
+; CHECK-NEXT:  b0:
+; CHECK-NEXT:    [[UNMASKEDLOAD:%.*]] = load <4 x i32>, <4 x i32>* @g0, align 16, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> [[UNMASKEDLOAD]], i32 0, i32 0
+; CHECK-NEXT:    ret <4 x i32> [[TMP0]]
+;
+b0:
+  %v0 = call <4 x i32> @llvm.masked.load.v4i1.p0v4i1(<4 x i32>* @g0, i32 16, <4 x i1> <i1 false, i1 true, i1 true, i1 true>, <4 x i32> zeroinitializer), !tbaa !0
+  ret <4 x i32> %v0
+}
+
+define void @mstore(<4 x i32> %a0, <4 x i32>* nocapture readonly %a1) #0 {
+; CHECK-LABEL: @mstore(
+; CHECK-NEXT:  b0:
+; CHECK-NEXT:    store <4 x i32> [[A0:%.*]], <4 x i32>* [[A1:%.*]], align 16, !tbaa [[TBAA0]]
+; CHECK-NEXT:    ret void
+;
+b0:
+  call void @llvm.masked.store.v4i1.p0v4i1(<4 x i32> %a0, <4 x i32>* %a1, i32 16, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa !0
+  ret void
+}
+
+attributes #0 = { norecurse nounwind }
+
+declare <4 x i32> @llvm.masked.load.v4i1.p0v4i1(<4 x i32>*, i32, <4 x i1>, <4 x i32>)
+declare void @llvm.masked.store.v4i1.p0v4i1(<4 x i32>, <4 x i32>*, i32, <4 x i1>)
+
+!0 = !{!1, !1, i64 0}
+!1 = !{!"omnipotent char", !2, i64 0}
+!2 = !{!"Simple C/C++ TBAA"}
+