PowerPC TPREL16_HA/LO reloc optimization

In the TLS GD/LD to LE optimization, ld replaces a sequence like

 addi 3,2,x@got@tlsgd		R_PPC64_GOT_TLSGD16	x
 bl __tls_get_addr(x@tlsgd)	R_PPC64_TLSGD		x
				R_PPC64_REL24		__tls_get_addr
 nop

with

 addis 3,13,x@tprel@ha		R_PPC64_TPREL16_HA	x
 addi 3,3,x@tprel@l		R_PPC64_TPREL16_LO	x
 nop

When the tprel offset is small, this can be further optimized to

 nop
 addi 3,13,x@tprel
 nop

bfd/
	* elf64-ppc.c (struct ppc_link_hash_table): Add do_tls_opt.
	(ppc64_elf_tls_optimize): Set it.
	(ppc64_elf_relocate_section): Nop addis on TPREL16_HA, and convert
	insn on TPREL16_LO and TPREL16_LO_DS relocs to use r13 when
	addis would add zero.
	* elf32-ppc.c (struct ppc_elf_link_hash_table): Add do_tls_opt.
	(ppc_elf_tls_optimize): Set it.
	(ppc_elf_relocate_section): Nop addis on TPREL16_HA, and convert
	insn on TPREL16_LO relocs to use r2 when addis would add zero.
gold/
	* powerpc.cc (Target_powerpc::Relocate::relocate): Nop addis on
	TPREL16_HA, and convert insn on TPREL16_LO and TPREL16_LO_DS
	relocs to use r2/r13 when addis would add zero.
ld/
	* testsuite/ld-powerpc/tls.s: Add calls with tls markers.
	* testsuite/ld-powerpc/tls32.s: Likewise.
	* testsuite/ld-powerpc/powerpc.exp: Run tls marker tests.
	* testsuite/ld-powerpc/tls.d: Adjust for TPREL16_HA/LO optimization.
	* testsuite/ld-powerpc/tlsexe.d: Likewise.
	* testsuite/ld-powerpc/tlsexetoc.d: Likewise.
	* testsuite/ld-powerpc/tlsld.d: Likewise.
	* testsuite/ld-powerpc/tlsmark.d: Likewise.
	* testsuite/ld-powerpc/tlsopt4.d: Likewise.
	* testsuite/ld-powerpc/tlstoc.d: Likewise.

3 files changed, 84 insertions, 2 deletions

diff --git a/bfd/ChangeLog b/bfd/ChangeLog
index ec74d9d..205b016 100644
--- a/bfd/ChangeLog
+++ b/bfd/ChangeLog
@@ -1,5 +1,17 @@
 2017-08-30  Alan Modra  <amodra@gmail.com>
 
+	* elf64-ppc.c (struct ppc_link_hash_table): Add do_tls_opt.
+	(ppc64_elf_tls_optimize): Set it.
+	(ppc64_elf_relocate_section): Nop addis on TPREL16_HA, and convert
+	insn on TPREL16_LO and TPREL16_LO_DS relocs to use r13 when
+	addis would add zero.
+	* elf32-ppc.c (struct ppc_elf_link_hash_table): Add do_tls_opt.
+	(ppc_elf_tls_optimize): Set it.
+	(ppc_elf_relocate_section): Nop addis on TPREL16_HA, and convert
+	insn on TPREL16_LO relocs to use r2 when addis would add zero.
+
+2017-08-30  Alan Modra  <amodra@gmail.com>
+
 	* elf64-ppc.c (ppc64_elf_relocate_section): When optimizing
 	__tls_get_addr call sequences to LE, don't move the addi down
 	to the nop.  Replace the bl with addi and leave the nop alone.
diff --git a/bfd/elf32-ppc.c b/bfd/elf32-ppc.c
index a5d8193..2bdba9f 100644
--- a/bfd/elf32-ppc.c
+++ b/bfd/elf32-ppc.c
@@ -3345,6 +3345,9 @@ struct ppc_elf_link_hash_table
   unsigned int local_ifunc_resolver:1;
   unsigned int maybe_local_ifunc_resolver:1;
 
+  /* Set if tls optimization is enabled.  */
+  unsigned int do_tls_opt:1;
+
   /* The size of PLT entries.  */
   int plt_entry_size;
   /* The distance between adjacent PLT slots.  */
@@ -5631,6 +5634,7 @@ ppc_elf_tls_optimize (bfd *obfd ATTRIBUTE_UNUSED,
 	      symtab_hdr->contents = (unsigned char *) locsyms;
 	  }
       }
+  htab->do_tls_opt = 1;
   return TRUE;
 }
 
@@ -8426,10 +8430,44 @@ ppc_elf_relocate_section (bfd *output_bfd,
 	}
 
       addend = rel->r_addend;
-      tls_type = 0;
       howto = NULL;
       if (r_type < R_PPC_max)
 	howto = ppc_elf_howto_table[r_type];
+
+      switch (r_type)
+	{
+	default:
+	  break;
+
+	case R_PPC_TPREL16_HA:
+	  if (htab->do_tls_opt && relocation + addend + 0x8000 < 0x10000)
+	    {
+	      bfd_byte *p = contents + (rel->r_offset & ~3);
+	      unsigned int insn = bfd_get_32 (input_bfd, p);
+	      if ((insn & ((0x3f << 26) | 0x1f << 16))
+		  != ((15u << 26) | (2 << 16)) /* addis rt,2,imm */)
+		/* xgettext:c-format */
+		info->callbacks->minfo
+		  (_("%H: warning: %s unexpected insn %#x.\n"),
+		   input_bfd, input_section, rel->r_offset, howto->name, insn);
+	      else
+		bfd_put_32 (input_bfd, NOP, p);
+	    }
+	  break;
+
+	case R_PPC_TPREL16_LO:
+	  if (htab->do_tls_opt && relocation + addend + 0x8000 < 0x10000)
+	    {
+	      bfd_byte *p = contents + (rel->r_offset & ~3);
+	      unsigned int insn = bfd_get_32 (input_bfd, p);
+	      insn &= ~(0x1f << 16);
+	      insn |= 2 << 16;
+	      bfd_put_32 (input_bfd, insn, p);
+	    }
+	  break;
+	}
+
+      tls_type = 0;
       switch (r_type)
 	{
 	default:
diff --git a/bfd/elf64-ppc.c b/bfd/elf64-ppc.c
index 5b96a04..cf7c178 100644
--- a/bfd/elf64-ppc.c
+++ b/bfd/elf64-ppc.c
@@ -4120,6 +4120,9 @@ struct ppc_link_hash_table
   unsigned int second_toc_pass:1;
   unsigned int do_toc_opt:1;
 
+  /* Set if tls optimization is enabled.  */
+  unsigned int do_tls_opt:1;
+
   /* Set on error.  */
   unsigned int stub_error:1;
 
@@ -8925,6 +8928,7 @@ ppc64_elf_tls_optimize (struct bfd_link_info *info)
 
   if (toc_ref != NULL)
     free (toc_ref);
+  htab->do_tls_opt = 1;
   return TRUE;
 }
 
@@ -15152,6 +15156,7 @@ ppc64_elf_relocate_section (bfd *output_bfd,
       /* Multi-instruction sequences that access the TOC can be
 	 optimized, eg. addis ra,r2,0; addi rb,ra,x;
 	 to             nop;           addi rb,r2,x;  */
+      howto = ppc64_elf_howto_table[(int) r_type];
       switch (r_type)
 	{
 	default:
@@ -15213,10 +15218,37 @@ ppc64_elf_relocate_section (bfd *output_bfd,
 	      bfd_put_32 (input_bfd, insn, p);
 	    }
 	  break;
+
+	case R_PPC64_TPREL16_HA:
+	  if (htab->do_tls_opt && relocation + addend + 0x8000 < 0x10000)
+	    {
+	      bfd_byte *p = contents + (rel->r_offset & ~3);
+	      insn = bfd_get_32 (input_bfd, p);
+	      if ((insn & ((0x3f << 26) | 0x1f << 16))
+		  != ((15u << 26) | (13 << 16)) /* addis rt,13,imm */)
+		/* xgettext:c-format */
+		info->callbacks->minfo
+		  (_("%H: warning: %s unexpected insn %#x.\n"),
+		   input_bfd, input_section, rel->r_offset, howto->name, insn);
+	      else
+		bfd_put_32 (input_bfd, NOP, p);
+	    }
+	  break;
+
+	case R_PPC64_TPREL16_LO:
+	case R_PPC64_TPREL16_LO_DS:
+	  if (htab->do_tls_opt && relocation + addend + 0x8000 < 0x10000)
+	    {
+	      bfd_byte *p = contents + (rel->r_offset & ~3);
+	      insn = bfd_get_32 (input_bfd, p);
+	      insn &= ~(0x1f << 16);
+	      insn |= 13 << 16;
+	      bfd_put_32 (input_bfd, insn, p);
+	    }
+	  break;
 	}
 
       /* Do any further special processing.  */
-      howto = ppc64_elf_howto_table[(int) r_type];
       switch (r_type)
 	{
 	default: