Index: lib/xray/xray_arm.cc
===================================================================
--- lib/xray/xray_arm.cc
+++ lib/xray/xray_arm.cc
@@ -148,9 +148,7 @@
 
 bool patchFunctionTailExit(const bool Enable, const uint32_t FuncId,
                            const XRaySledEntry &Sled) XRAY_NEVER_INSTRUMENT {
-  // FIXME: In the future we'd need to distinguish between non-tail exits and
-  // tail exits for better information preservation.
-  return patchSled(Enable, FuncId, Sled, __xray_FunctionExit);
+  return patchSled(Enable, FuncId, Sled, __xray_FunctionTailExit);
 }
 
 } // namespace __xray
Index: lib/xray/xray_trampoline_arm.S
===================================================================
--- lib/xray/xray_trampoline_arm.S
+++ lib/xray/xray_trampoline_arm.S
@@ -3,6 +3,7 @@
     .fpu vfpv2
     .code 32
     .global _ZN6__xray19XRayPatchedFunctionE
+
     @ Word-aligned function entry point
     .p2align 2
     @ Let C/C++ see the symbol
@@ -63,3 +64,35 @@
     @ Restore the floating-point return value of the instrumented function
     VPOP {d0}
     POP {r1-r3,pc}
+
+    @ Word-aligned function entry point
+    .p2align 2
+    @ Let C/C++ see the symbol
+    .global __xray_FunctionTailExit
+    @ It preserves all registers except r0, r12(ip), r14(lr) and r15(pc)
+    @ Assume that "q" part of the floating-point registers is not used
+    @   for passing parameters to C/C++ functions.
+    .type __xray_FunctionTailExit, %function
+    @ In C++ it is void extern "C" __xray_FunctionTailExit(uint32_t FuncId)
+    @   with FuncId passed in r0 register.
+__xray_FunctionTailExit:
+    PUSH {r1-r3,lr}
+    @ Save floating-point parameters of the instrumented function
+    VPUSH {d0-d7}
+    MOVW r1,#:lower16:_ZN6__xray19XRayPatchedFunctionE
+    MOVT r1,#:upper16:_ZN6__xray19XRayPatchedFunctionE
+    LDR r2, [r1]
+    @ Handler address is nullptr if handler is not set
+    CMP r2, #0
+    BEQ FunctionTailExit_restore
+    @ Function ID is already in r0 (the first parameter).
+    @ r1=2 means that we are tracing a tail exit event
+    @ But before the logging part of XRay is ready, we pretend that here a
+    @   normal function exit happens, so we give the handler code 1
+    MOV r1, #1
+    @ Call the handler with 2 parameters in r0 and r1
+    BLX r2
+FunctionTailExit_restore:
+    @ Restore floating-point parameters of the instrumented function
+    VPOP {d0-d7}
+    POP {r1-r3,pc}