Cygwin: aarch64: Add runtime pseudo relocation

The patch adds runtime pseudo relocation handling for 12-bit and 21-bit
relocations. The 26-bit relocation is handled using a jump stub generated by
the linker.

Author: eukarpov

winsup/cygwin/pseudo-reloc.cc

@@ -199,6 +199,9 @@ do_pseudo_reloc (void * start, void * end, void * base)
   ptrdiff_t reloc_target = (ptrdiff_t) ((char *)end - (char*)start);
   runtime_pseudo_reloc_v2 *v2_hdr = (runtime_pseudo_reloc_v2 *) start;
   runtime_pseudo_reloc_item_v2 *r;
+#ifdef __aarch64__
+  uint32_t opcode;
+#endif
 
   /* A valid relocation list will contain at least one entry, and
    * one v1 data structure (the smallest one) requires two DWORDs.
@@ -307,6 +310,13 @@ do_pseudo_reloc (void * start, void * end, void * base)
 	  if ((reldata & 0x8000) != 0)
 	    reldata |= ~((ptrdiff_t) 0xffff);
 	  break;
+#ifdef __aarch64__
+	case 12:
+	case 21:
+	  opcode = (*((unsigned int *) reloc_target));
+	  reldata = 0;
+	  break;
+#endif
 	case 32:
 	  reldata = (ptrdiff_t) (*((unsigned int *)reloc_target));
 #if defined (__x86_64__) || defined (_WIN64)
@@ -339,6 +349,31 @@ do_pseudo_reloc (void * start, void * end, void * base)
 	case 16:
 	  __write_memory ((void *) reloc_target, &reldata, 2);
 	  break;
+#ifdef __aarch64__
+	case 12:
+	  /* Replace add Xn, Xn, :lo12:label with ldr Xn, [Xn, :lo12:__imp__func].
+	     That loads the address of _func into Xn.  */
+	  opcode = 0xf9400000 | (opcode & 0x3ff); // ldr
+	  reldata = ((ptrdiff_t) base + r->sym) & ((1 << 12) - 1);
+	  reldata >>= 3;
+	  opcode |= reldata << 10;
+	  __write_memory ((void *) reloc_target, &opcode, 4);
+	  break;
+	case 21:
+	  /* Replace adrp Xn, label with adrp Xn, __imp__func.  */
+	  opcode &= 0x9f00001f;
+	  reldata = (((ptrdiff_t) base + r->sym) >> 12)
+		    - (((ptrdiff_t) base + r->target) >> 12);
+	  reldata &= (1 << 21) - 1;
+	  opcode |= (reldata & 3) << 29;
+	  reldata >>= 2;
+	  opcode |= reldata << 5;
+	  __write_memory ((void *) reloc_target, &opcode, 4);
+	  break;
+	/* A note regarding 26 bits relocation.
+	   A single opcode is not sufficient for 26 bits relocation in dynamic linking.
+	   The linker generates a jump stub instead.  */
+#endif
 	case 32:
 #if defined (__CYGWIN__) && defined (__x86_64__)
 	  if (reldata > (ptrdiff_t) __INT32_MAX__