Index: lib/Target/AArch64/AArch64AsmPrinter.cpp
===================================================================
--- lib/Target/AArch64/AArch64AsmPrinter.cpp
+++ lib/Target/AArch64/AArch64AsmPrinter.cpp
@@ -591,6 +591,7 @@
   // attributes (isCall, isReturn, etc.). We lower them to the real
   // instruction here.
   case AArch64::TCRETURNri:
+  case AArch64::TCRETURNriBTI:
   case AArch64::TCRETURNriALL: {
     MCInst TmpInst;
     TmpInst.setOpcode(AArch64::BR);
Index: lib/Target/AArch64/AArch64FrameLowering.cpp
===================================================================
--- lib/Target/AArch64/AArch64FrameLowering.cpp
+++ lib/Target/AArch64/AArch64FrameLowering.cpp
@@ -927,7 +927,8 @@
     DL = MBBI->getDebugLoc();
     unsigned RetOpcode = MBBI->getOpcode();
     IsTailCallReturn = RetOpcode == AArch64::TCRETURNdi ||
-      RetOpcode == AArch64::TCRETURNri;
+                       RetOpcode == AArch64::TCRETURNri ||
+                       RetOpcode == AArch64::TCRETURNriBTI;
   }
   int NumBytes = MFI.getStackSize();
   const AArch64FunctionInfo *AFI = MF.getInfo<AArch64FunctionInfo>();
Index: lib/Target/AArch64/AArch64InstrInfo.td
===================================================================
--- lib/Target/AArch64/AArch64InstrInfo.td
+++ lib/Target/AArch64/AArch64InstrInfo.td
@@ -360,6 +360,9 @@
   def NotForCodeSize   : Predicate<"!MF->getFunction().optForSize()">;
   // Avoid generating STRQro if it is slow, unless we're optimizing for code size.
   def UseSTRQro : Predicate<"!Subtarget->isSTRQroSlow() || MF->getFunction().optForSize()">;
+
+  def UseBTI : Predicate<[{ MF->getFunction().hasFnAttribute("branch-target-enforcement") }]>;
+  def NotUseBTI : Predicate<[{ !MF->getFunction().hasFnAttribute("branch-target-enforcement") }]>;
 }
 
 include "AArch64InstrFormats.td"
@@ -6641,10 +6644,18 @@
   // some verifier checks for outlined functions.
   def TCRETURNriALL : Pseudo<(outs), (ins GPR64:$dst, i32imm:$FPDiff), []>,
                       Sched<[WriteBrReg]>;
+  // Indirect tail-call limited to only use registers (x16 and x17) which are
+  // allowed to tail-call a "BTI c" instruction.
+  def TCRETURNriBTI : Pseudo<(outs), (ins rtcGPR64:$dst, i32imm:$FPDiff), []>,
+                      Sched<[WriteBrReg]>;
 }
 
 def : Pat<(AArch64tcret tcGPR64:$dst, (i32 timm:$FPDiff)),
-          (TCRETURNri tcGPR64:$dst, imm:$FPDiff)>;
+          (TCRETURNri tcGPR64:$dst, imm:$FPDiff)>,
+      Requires<[NotUseBTI]>;
+def : Pat<(AArch64tcret rtcGPR64:$dst, (i32 timm:$FPDiff)),
+          (TCRETURNriBTI rtcGPR64:$dst, imm:$FPDiff)>,
+      Requires<[UseBTI]>;
 def : Pat<(AArch64tcret tglobaladdr:$dst, (i32 timm:$FPDiff)),
           (TCRETURNdi texternalsym:$dst, imm:$FPDiff)>;
 def : Pat<(AArch64tcret texternalsym:$dst, (i32 timm:$FPDiff)),
Index: lib/Target/AArch64/AArch64RegisterInfo.td
===================================================================
--- lib/Target/AArch64/AArch64RegisterInfo.td
+++ lib/Target/AArch64/AArch64RegisterInfo.td
@@ -200,6 +200,12 @@
                                                      X22, X23, X24, X25, X26,
                                                      X27, X28, FP, LR)>;
 
+// Restricted set of tail call registers, for use when branch target
+// enforcement is enabled. These are the only registers which can be used to
+// indirectly branch (not call) to the "BTI c" instruction at the start of a
+// BTI-protected function.
+def rtcGPR64 : RegisterClass<"AArch64", [i64], 64, (add X16, X17)>;
+
 // GPR register classes for post increment amount of vector load/store that
 // has alternate printing when Rm=31 and prints a constant immediate value
 // equal to the total number of bytes transferred.
Index: test/CodeGen/AArch64/branch-target-enforcement-indirect-calls.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AArch64/branch-target-enforcement-indirect-calls.ll
@@ -0,0 +1,25 @@
+; RUN: llc -mtriple aarch64--none-eabi -mattr=+bti < %s | FileCheck %s
+
+target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
+target triple = "aarch64-arm-none-eabi"
+
+; When BTI is enabled, all indirect tail-calls must use x16 or x17 (the intra
+; procedure call scratch registers) to hold the address, as these instructions
+; are allowed to target the "BTI c" instruction at the start of the target
+; function. The alternative to this would be to start functions with "BTI jc",
+; which increases the number of potential ways they could be called, and
+; weakens the security protections.
+
+define void @bti_disabled(void ()* %p) {
+entry:
+  tail call void %p()
+; CHECK: br x0
+  ret void
+}
+
+define void @bti_enabled(void ()* %p) "branch-target-enforcement" {
+entry:
+  tail call void %p()
+; CHECK: br {{x16|x17}}
+  ret void
+}