For JDK8 & JDK9, there two redundant lsr instructions in jbyte_arraycopy and jbyte_disjoint_arraycopy stub as shown in the following code snippet:
StubRoutines::jbyte_disjoint_arraycopy [0x0000007f7c0bef80, 0x0000007f7c0bf020[ (160 bytes)
  0x0000007f7c0bef80: stp x29, x30, [sp,#-16]!
  0x0000007f7c0bef84: mov x29, sp
  0x0000007f7c0bef88: cmp x2, #0x10
  0x0000007f7c0bef8c: b.cc Stub::jbyte_disjoint_arraycopy+100 0x0x7f7c0befe4
  0x0000007f7c0bef90: neg x9, x0
  0x0000007f7c0bef94: and x9, x9, #0xf
  0x0000007f7c0bef98: cbz x9, Stub::jbyte_disjoint_arraycopy+92 0x0x7f7c0befdc
  0x0000007f7c0bef9c: lsr x9, x9, #0
  0x0000007f7c0befa0: sub x2, x2, x9
  0x0000007f7c0befa4: tbz w9, #3, Stub::jbyte_disjoint_arraycopy+48 0x0x7f7c0befb0
  0x0000007f7c0befa8: ldr x8, [x0],#8
  0x0000007f7c0befac: str x8, [x1],#8
  0x0000007f7c0befb0: tbz w9, #2, Stub::jbyte_disjoint_arraycopy+60 0x0x7f7c0befbc
  0x0000007f7c0befb4: ldr w8, [x0],#4
  0x0000007f7c0befb8: str w8, [x1],#4
  0x0000007f7c0befbc: tbz w9, #1, Stub::jbyte_disjoint_arraycopy+72 0x0x7f7c0befc8
  0x0000007f7c0befc0: ldrh w8, [x0],#2
  0x0000007f7c0befc4: strh w8, [x1],#2
  0x0000007f7c0befc8: tbz w9, #0, Stub::jbyte_disjoint_arraycopy+84 0x0x7f7c0befd4
  0x0000007f7c0befcc: ldrb w8, [x0],#1
  0x0000007f7c0befd0: strb w8, [x1],#1
  ........

A proposed patch to fix the issue:
diff -r 119702fc4dea src/cpu/aarch64/vm/stubGenerator_aarch64.cpp
--- a/src/cpu/aarch64/vm/stubGenerator_aarch64.cpp Tue Jan 26 17:13:18 2016 +0100
+++ b/src/cpu/aarch64/vm/stubGenerator_aarch64.cpp Wed Jan 27 12:11:24 2016 +0800
@@ -954,7 +954,7 @@
                    Register count, Register tmp, int step) {
     copy_direction direction = step < 0 ? copy_backwards : copy_forwards;
     bool is_backwards = step < 0;
- int granularity = uabs(step);
+ int shift, granularity = uabs(step);
     const Register t0 = r3, t1 = r4;

     if (is_backwards) {
@@ -962,7 +962,7 @@
       __ lea(d, Address(d, count, Address::lsl(exact_log2(-step))));
     }

- Label done, tail;
+ Label tail;

     __ cmp(count, 16/granularity);
     __ br(Assembler::LO, tail);
@@ -985,9 +985,12 @@
         __ neg(rscratch2, s);
         __ andr(rscratch2, rscratch2, 2 * wordSize - 1);
       }
+ shift = exact_log2(granularity);
       // rscratch2 is the byte adjustment needed to align s.
       __ cbz(rscratch2, aligned);
- __ lsr(rscratch2, rscratch2, exact_log2(granularity));
+ if (shift > 0) {
+ __ lsr(rscratch2, rscratch2, shift);
+ }
       __ sub(count, count, rscratch2);

 #if 0