diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -1718,11 +1718,10 @@
             // Figure out maximal valid memcpy alignment.
             const Align ArgAlign = DL.getValueOrABITypeAlignment(
                 MaybeAlign(FArg.getParamAlignment()), FArg.getParamByValType());
-            Value *CpShadowPtr =
+            Value *CpShadowPtr, *CpOriginPtr;
+            std::tie(CpShadowPtr, CpOriginPtr) =
                 getShadowOriginPtr(V, EntryIRB, EntryIRB.getInt8Ty(), ArgAlign,
-                                   /*isStore*/ true)
-                    .first;
-            // TODO(glider): need to copy origins.
+                                   /*isStore*/ true);
             if (!PropagateShadow || Overflow) {
               // ParamTLS overflow.
               EntryIRB.CreateMemSet(
@@ -1735,6 +1734,13 @@
                                                  CopyAlign, Size);
               LLVM_DEBUG(dbgs() << "  ByValCpy: " << *Cpy << "\n");
               (void)Cpy;
+
+              if (MS.TrackOrigins) {
+                Value *OriginPtr =
+                    getOriginPtrForArgument(&FArg, EntryIRB, ArgOffset);
+                EntryIRB.CreateMemCpy(CpOriginPtr, CopyAlign, OriginPtr,
+                                      CopyAlign, Size);
+              }
             }
           }
 
@@ -3701,7 +3707,6 @@
         insertShadowCheck(A, &CB);
         Size = DL.getTypeAllocSize(A->getType());
       } else {
-        bool ArgIsInitialized = false;
         Value *Store = nullptr;
         // Compute the Shadow for arg even if it is ByVal, because
         // in that case getShadow() will copy the actual arg shadow to
@@ -3722,14 +3727,19 @@
           MaybeAlign Alignment = llvm::None;
           if (ParamAlignment)
             Alignment = std::min(*ParamAlignment, kShadowTLSAlignment);
-          Value *AShadowPtr =
+          Value *AShadowPtr, *AOriginPtr;
+          std::tie(AShadowPtr, AOriginPtr) =
               getShadowOriginPtr(A, IRB, IRB.getInt8Ty(), Alignment,
-                                 /*isStore*/ false)
-                  .first;
+                                 /*isStore*/ false);
 
           Store = IRB.CreateMemCpy(ArgShadowBase, Alignment, AShadowPtr,
                                    Alignment, Size);
-          // TODO(glider): need to copy origins.
+          if (MS.TrackOrigins) {
+            Value *ArgOriginBase = getOriginPtrForArgument(A, IRB, ArgOffset);
+        
+            Store = IRB.CreateMemCpy(ArgOriginBase, Alignment, AOriginPtr,
+                                    Alignment, Size);
+          }
         } else {
           // Any other parameters mean we need bit-grained tracking of uninit
           // data
@@ -3739,12 +3749,11 @@
           Store = IRB.CreateAlignedStore(ArgShadow, ArgShadowBase,
                                          kShadowTLSAlignment);
           Constant *Cst = dyn_cast<Constant>(ArgShadow);
-          if (Cst && Cst->isNullValue())
-            ArgIsInitialized = true;
+          if (MS.TrackOrigins && !(Cst && Cst->isNullValue())) {
+            IRB.CreateStore(getOrigin(A),
+                            getOriginPtrForArgument(A, IRB, ArgOffset));
+          }
         }
-        if (MS.TrackOrigins && !ArgIsInitialized)
-          IRB.CreateStore(getOrigin(A),
-                          getOriginPtrForArgument(A, IRB, ArgOffset));
         (void)Store;
         assert(Store != nullptr);
         LLVM_DEBUG(dbgs() << "  Param:" << *Store << "\n");
diff --git a/llvm/test/Instrumentation/MemorySanitizer/byval.ll b/llvm/test/Instrumentation/MemorySanitizer/byval.ll
--- a/llvm/test/Instrumentation/MemorySanitizer/byval.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/byval.ll
@@ -11,6 +11,7 @@
 ; CHECK-LABEL: @ByValArgument(
 ; CHECK-NEXT:  entry:
 ; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64({{.*}} @__msan_param_tls to i64), i64 8) to i8*), i64 16, i1 false)
+; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64({{.*}} @__msan_param_origin_tls to i64), i64 8) to i8*), i64 16, i1 false)
 ; CHECK:         %x = load i128, i128* %p
 ; CHECK:         load i128
 ; CHECK:         load i32
@@ -37,11 +38,11 @@
   ret i128 %x
 }
 
-; FIXME: Origin of byval pointee is not propagated.
 define void @ByValForward(i32, i128* byval(i128) %p) sanitize_memory {
 ; CHECK-LABEL: @ByValForward(
 ; CHECK-NEXT:  entry:
 ; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64({{.*}} @__msan_param_tls to i64), i64 8) to i8*), i64 16, i1 false)
+; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64({{.*}} @__msan_param_origin_tls to i64), i64 8) to i8*), i64 16, i1 false)
 ; CHECK:         store i64 0, i64* getelementptr inbounds ([100 x i64], [100 x i64]* @__msan_param_tls, i32 0, i32 0)
 ; CHECK-NEXT:    call void @Fn(
 ; CHECK-NEXT:    ret void
@@ -64,13 +65,13 @@
   ret void
 }
 
-; FIXME: Origin of %p byval pointee is not propagated.
 define void @ByValForwardByVal(i32, i128* byval(i128) %p) sanitize_memory {
 ; CHECK-LABEL: @ByValForwardByVal(
 ; CHECK-NEXT:  entry:
 ; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64({{.*}} @__msan_param_tls to i64), i64 8) to i8*), i64 16, i1 false)
+; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64({{.*}} @__msan_param_origin_tls to i64), i64 8) to i8*), i64 16, i1 false)
 ; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64(i8* bitcast ([100 x i64]* @__msan_param_tls to i8*), i8* {{.*}}, i64 16, i1 false)
-; CHECK:         store i32 {{.*}}, i32* getelementptr inbounds ([200 x i32], [200 x i32]* @__msan_param_origin_tls, i32 0, i32 0)
+; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64(i8* bitcast ([200 x i32]* @__msan_param_origin_tls to i8*), i8* {{.*}}, i64 16, i1 false)
 ; CHECK-NEXT:    call void @FnByVal(
 ; CHECK-NEXT:    ret void
 ;
@@ -85,7 +86,7 @@
 ; CHECK-NEXT:  entry:
 ; CHECK:         call void @llvm.memset.p0i8.i64(i8* align 8 {{.*}}, i8 0, i64 16, i1 false)
 ; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64(i8* bitcast ([100 x i64]* @__msan_param_tls to i8*), i8* {{.*}}, i64 16, i1 false) 
-; CHECK:         store i32 0, i32* getelementptr inbounds ([200 x i32], [200 x i32]* @__msan_param_origin_tls, i32 0, i32 0)
+; CHECK:         call void @llvm.memcpy.p0i8.p0i8.i64(i8* bitcast ([200 x i32]* @__msan_param_origin_tls to i8*), i8*  {{.*}}, i64 16, i1 false)
 ; CHECK-NEXT:    call void @FnByVal(
 ; CHECK-NEXT:    ret void
 ;