ld: aarch64: fix TLS relaxation where TCB_SIZE is used

TCB_SIZE is 2*sizeof(void *), which is 0x10 for lp64, and 0x8 for
ilp32. During relaxation, ld goes to do a replace:
bl   __tls_get_addr => add R0, R0, TCB_SIZE

But actual implementation is:
bfd_putl32 (0x91004000, contents + rel->r_offset + 4);

Which is equivalent of add x0, x0, 0x10. This is wrong for ilp32.

The possible fix for it is:
bfd_putl32 (0x91000000 | (TCB_SIZE<<10), contents + rel->r_offset + 4);

But ilp32 also needs w-registers, so it's simpler to put proper
instruction in #if/#else condition.

THere are 2 such relaxations in elfNN_aarch64_tls_relax(), and so 2 new
tests added for ilp32 mode to test it.

Yury

bfd/elfnn-aarch64.c

@@ -5951,7 +5951,9 @@ elfNN_aarch64_tls_relax (struct elf_aarch64_link_hash_table *globals,
     case BFD_RELOC_AARCH64_TLSLD_ADR_PREL21:
       /* LD->LE relaxation (tiny):
 	 adr  x0, :tlsldm:x  => mrs x0, tpidr_el0
-	 bl   __tls_get_addr => add x0, x0, TCB_SIZE
+	 bl   __tls_get_addr => add R0, R0, TCB_SIZE
+
+	 Where R is x for lp64 mode, and w for ilp32 mode.
        */
       if (is_local)
 	{
@@ -5960,7 +5962,11 @@ elfNN_aarch64_tls_relax (struct elf_aarch64_link_hash_table *globals,
 	  /* No need of CALL26 relocation for tls_get_addr.  */
 	  rel[1].r_info = ELFNN_R_INFO (STN_UNDEF, R_AARCH64_NONE);
 	  bfd_putl32 (0xd53bd040, contents + rel->r_offset + 0);
+#if ARCH_SIZE ==64
 	  bfd_putl32 (0x91004000, contents + rel->r_offset + 4);
+#else
+	  bfd_putl32 (0x11002000, contents + rel->r_offset + 4);
+#endif
 	  return bfd_reloc_ok;
 	}
       return bfd_reloc_continue;
@@ -5978,8 +5984,10 @@ elfNN_aarch64_tls_relax (struct elf_aarch64_link_hash_table *globals,
 
     case BFD_RELOC_AARCH64_TLSLD_ADD_LO12_NC:
       /* LD->LE relaxation (small):
-	 add   x0, #:tlsldm_lo12:x => add x0, x0, TCB_SIZE
+	 add   x0, #:tlsldm_lo12:x => add R0, R0, TCB_SIZE
 	 bl   __tls_get_addr       => nop
+
+	 Where R is x for lp64 mode, and w for ilp32 mode.
        */
       if (is_local)
 	{
@@ -5987,8 +5995,12 @@ elfNN_aarch64_tls_relax (struct elf_aarch64_link_hash_table *globals,
 	  BFD_ASSERT (ELFNN_R_TYPE (rel[1].r_info) == AARCH64_R (CALL26));
 	  /* No need of CALL26 relocation for tls_get_addr.  */
 	  rel[1].r_info = ELFNN_R_INFO (STN_UNDEF, R_AARCH64_NONE);
+#if ARCH_SIZE ==64
 	  bfd_putl32 (0x91004000, contents + rel->r_offset + 0);
-	  bfd_putl32 (0xd503201f, contents + rel->r_offset + 4);
+#else
+	  bfd_putl32 (0x11002000, contents + rel->r_offset + 0);
+#endif
+	  bfd_putl32 (INSN_NOP, contents + rel->r_offset + 4);
 	  return bfd_reloc_ok;
 	}
       return bfd_reloc_continue;

ld/testsuite/ld-aarch64/aarch64-elf.exp

@@ -228,7 +228,9 @@ run_dump_test_lp64 "tls-relax-large-desc-le-be"
 run_dump_test "tls-relax-gdesc-ie"
 run_dump_test "tls-relax-ie-le"
 run_dump_test "tls-relax-ld-le-small"
+run_dump_test "tls-relax-ld-le-small-ilp32"
 run_dump_test "tls-relax-ld-le-tiny"
+run_dump_test "tls-relax-ld-le-tiny-ilp32"
 run_dump_test "tls-desc-ie"
 run_dump_test "tls-relax-gdesc-ie-2"
 run_dump_test "tls-relax-gdesc-le-2"

ld/testsuite/ld-aarch64/tls-relax-ld-le-small-ilp32.d

@@ -0,0 +1,14 @@
+#source: tls-relax-ld-le-small.s
+#as: -mabi=ilp32
+#ld: -m [aarch64_choose_ilp32_emul] -T relocs-ilp32.ld -e0
+#objdump: -dr
+#...
+ +10000:	910003fd 	mov	x29, sp
+ +10004:	d53bd040 	mrs	x0, tpidr_el0
+ +10008:	11002000 	add	w0, w0, #0x8
+ +1000c:	d503201f 	nop
+ +10010:	d503201f 	nop
+ +10014:	91400001 	add	x1, x0, #0x0, lsl #12
+ +10018:	91000021 	add	x1, x1, #0x0
+ +1001c:	90000000 	adrp	x0, 10000 <.*>
+ +10020:	d65f03c0 	ret

ld/testsuite/ld-aarch64/tls-relax-ld-le-tiny-ilp32.d

@@ -0,0 +1,13 @@
+#source: tls-relax-ld-le-tiny.s
+#as: -mabi=ilp32
+#ld: -m [aarch64_choose_ilp32_emul] -T relocs-ilp32.ld -e0
+#objdump: -dr
+#...
+ +10000:	910003fd 	mov	x29, sp
+ +10004:	d53bd040 	mrs	x0, tpidr_el0
+ +10008:	11002000 	add	w0, w0, #0x8
+ +1000c:	d503201f 	nop
+ +10010:	91400001 	add	x1, x0, #0x0, lsl #12
+ +10014:	91000021 	add	x1, x1, #0x0
+ +10018:	90000000 	adrp	x0, 10000 <main>
+ +1001c:	d65f03c0 	ret