diff --git a/llvm/lib/Transforms/Scalar/LICM.cpp b/llvm/lib/Transforms/Scalar/LICM.cpp
--- a/llvm/lib/Transforms/Scalar/LICM.cpp
+++ b/llvm/lib/Transforms/Scalar/LICM.cpp
@@ -1502,8 +1502,8 @@
     }
   }
 
-  // Build LCSSA PHI nodes for any in-loop operands. Note that this is
-  // particularly cheap because we can rip off the PHI node that we're
+  // Build LCSSA PHI nodes for any in-loop operands (if legal).  Note that
+  // this is particularly cheap because we can rip off the PHI node that we're
   // replacing for the number and blocks of the predecessors.
   // OPT: If this shows up in a profile, we can instead finish sinking all
   // invariant instructions, and then walk their operands to re-establish
@@ -1512,7 +1512,7 @@
   for (Use &Op : New->operands())
     if (Instruction *OInst = dyn_cast<Instruction>(Op))
       if (Loop *OLoop = LI->getLoopFor(OInst->getParent()))
-        if (!OLoop->contains(&PN)) {
+        if (!OLoop->contains(&PN) && !Op->getType()->isTokenTy()) {
           PHINode *OpPN =
               PHINode::Create(OInst->getType(), PN.getNumIncomingValues(),
                               OInst->getName() + ".lcssa", &ExitBlock.front());
@@ -1856,10 +1856,13 @@
   AAMDNodes AATags;
   ICFLoopSafetyInfo &SafetyInfo;
 
+  // We're about to add a use of V in a loop exit block.  Insert an LCSSA phi
+  // (if legal) if doing so would add an out-of-loop use to an instruction
+  // defined in-loop.
   Value *maybeInsertLCSSAPHI(Value *V, BasicBlock *BB) const {
     if (Instruction *I = dyn_cast<Instruction>(V))
       if (Loop *L = LI.getLoopFor(I->getParent()))
-        if (!L->contains(BB)) {
+        if (!L->contains(BB) && !I->getType()->isTokenTy()) {
           // We need to create an LCSSA PHI node for the incoming value and
           // store that.
           PHINode *PN = PHINode::Create(I->getType(), PredCache.size(BB),
diff --git a/llvm/test/Transforms/LICM/gc-relocate.ll b/llvm/test/Transforms/LICM/gc-relocate.ll
--- a/llvm/test/Transforms/LICM/gc-relocate.ll
+++ b/llvm/test/Transforms/LICM/gc-relocate.ll
@@ -10,18 +10,17 @@
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
 ; CHECK-NEXT:    [[TMP:%.*]] = call token (i64, i32, i32 (i32, i8 addrspace(1)*, i32, i32, i32)*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_i32i32p1i8i32i32i32f(i64 1, i32 16, i32 (i32, i8 addrspace(1)*, i32, i32, i32)* nonnull @zot, i32 5, i32 0, i32 undef, i8 addrspace(1)* undef, i32 undef, i32 undef, i32 undef, i32 0, i32 0) [ "deopt"(i32 0, i32 0, i32 0, i32 235, i32 3, i32 32, i32 0, i32 0, i8 addrspace(1)* undef, i32 3, i32 undef, i32 3, float undef, i32 0, i8 addrspace(1)* undef, i32 7, i8* null, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 0, i8 addrspace(1)* undef, i32 4, double undef, i32 7, i8* null, i32 0, i8 addrspace(1)* undef, i32 3, float undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* null, i32 3, i32 -15108, i32 7, i8* null, i32 7, i8* null, i32 7, i8* null, i32 7, i8* null, i32 7, i8* null), "gc-live"(i8 addrspace(1)* [[ARG:%.*]]) ]
-; CHECK-NEXT:    [[RES:%.*]] = call coldcc i8 addrspace(1)* @llvm.experimental.gc.relocate.p1i8(token [[TMP]], i32 0, i32 0)
 ; CHECK-NEXT:    br i1 false, label [[BB1]], label [[BB2:%.*]]
 ; CHECK:       bb2:
-; CHECK-NEXT:    [[RES_LCSSA:%.*]] = phi i8 addrspace(1)* [ [[RES]], [[BB1]] ]
-; CHECK-NEXT:    ret i8 addrspace(1)* [[RES_LCSSA]]
+; CHECK-NEXT:    [[RES_LE:%.*]] = call coldcc i8 addrspace(1)* @llvm.experimental.gc.relocate.p1i8(token [[TMP]], i32 0, i32 0) #[[ATTR2:[0-9]+]]
+; CHECK-NEXT:    ret i8 addrspace(1)* [[RES_LE]]
 ;
 bb:
   br label %bb1
 
 bb1:
   %tmp = call token (i64, i32, i32 (i32, i8 addrspace(1)*, i32, i32, i32)*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_i32i32p1i8i32i32i32f(i64 1, i32 16, i32 (i32, i8 addrspace(1)*, i32, i32, i32)* nonnull @zot, i32 5, i32 0, i32 undef, i8 addrspace(1)* undef, i32 undef, i32 undef, i32 undef, i32 0, i32 0) [ "deopt"(i32 0, i32 0, i32 0, i32 235, i32 3, i32 32, i32 0, i32 0, i8 addrspace(1)* undef, i32 3, i32 undef, i32 3, float undef, i32 0, i8 addrspace(1)* undef, i32 7, i8* null, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 3, i32 undef, i32 0, i8 addrspace(1)* undef, i32 4, double undef, i32 7, i8* null, i32 0, i8 addrspace(1)* undef, i32 3, float undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* undef, i32 0, i8 addrspace(1)* null, i32 3, i32 -15108, i32 7, i8* null, i32 7, i8* null, i32 7, i8* null, i32 7, i8* null, i32 7, i8* null), "gc-live"(i8 addrspace(1)* %arg) ]
-  %res = call coldcc i8 addrspace(1)* @llvm.experimental.gc.relocate.p1i8(token %tmp, i32 0, i32 0)
+  %res = call coldcc i8 addrspace(1)* @llvm.experimental.gc.relocate.p1i8(token %tmp, i32 0, i32 0) readnone
   br i1 undef, label %bb1, label %bb2
 
 bb2: