PR 11217, PowerPC64 automatic multiple TOC and __thread problems

[Alan Modra <amodra at gmail dot com>]

Seen when bootstrapping a native statically linked powerpc64-linux gcc
on a system with a recent glibc:

.../lib64/libc.a(setlocale.o): In function `_nl_locale_subfreeres':
.../locale/categories.def:42: relocation truncated to fit: R_PPC64_GOT_TPREL16_DS against undefined symbol `_nl_current_LC_COLLATE'

Errors like the above occur when the TOC exceeds 64k and there are
some TLS accesses using a model other than local exec.  A large TOC
requires ld to partition the TOC into groups smaller than 64k with TOC
adjusting stubs inserted on calls between functions using differing
TOC groups.  That works reasonably well except that code generated by
gcc for all TLS models besides local exec uses @got relocs, and these
generate GOT entries that are placed in the first TOC group.  The net
result is that functions using a TOC group other than the first can't
access their GOT entries.  ld ought to generate multiple GOT entries,
one in each group, or gcc should use the Power Open sequences for TLS
code (ie. use the TOC rather than the GOT).  Either way, I guess it's
my bug since I wrote both the linker automatic multiple TOC support
and the powerpc gcc TLS support..

This patch avoids some of these errors by teaching the linker tls
optimization support to handle undefined weak symbols, so most
executables will only use local exec model tls.  A proper fix for the
PR will require quite a lot more work.

	PR ld/11217
	* elf64-ppc.c (ppc64_elf_tls_optimize): Optimize tls sequences
	with relocations against undefined weak symbols.
	(ppc64_elf_relocate_section): Don't optimize calls to undefined
	weak functions if the symbol is dynamic.
	(ppc64_elf_relocate_section): Edit tprel tls sequences.
	* elf32-ppc.c (ppc_elf_relocate_section): Likewise.
	(_bfd_elf_ppc_at_tprel_transform): New function.
	* bfd-in.h (_bfd_elf_ppc_at_tprel_transform): Declare.
	* bfd-in2.h: Regenerate.

Index: bfd/elf64-ppc.c
===================================================================
RCS file: /cvs/src/src/bfd/elf64-ppc.c,v
retrieving revision 1.307
diff -u -p -r1.307 elf64-ppc.c
--- bfd/elf64-ppc.c	8 Jan 2010 05:55:10 -0000	1.307
+++ bfd/elf64-ppc.c	25 Jan 2010 05:02:36 -0000
@@ -7409,10 +7409,13 @@ ppc64_elf_tls_optimize (bfd *obfd ATTRIB
 
 		  if (h != NULL)
 		    {
-		      if (h->root.type != bfd_link_hash_defined
-			  && h->root.type != bfd_link_hash_defweak)
+		      if (h->root.type == bfd_link_hash_defined
+			  || h->root.type == bfd_link_hash_defweak)
+			value = h->root.u.def.value;
+		      else if (h->root.type == bfd_link_hash_undefweak)
+			value = 0;
+		      else
 			continue;
-		      value = h->root.u.def.value;
 		    }
 		  else
 		    /* Symbols referenced by TLS relocs must be of type
@@ -7425,11 +7428,17 @@ ppc64_elf_tls_optimize (bfd *obfd ATTRIB
 		      || !h->def_dynamic)
 		    {
 		      is_local = TRUE;
-		      value += sym_sec->output_offset;
-		      value += sym_sec->output_section->vma;
-		      value -= htab->elf.tls_sec->vma;
-		      ok_tprel = (value + TP_OFFSET + ((bfd_vma) 1 << 31)
-				  < (bfd_vma) 1 << 32);
+		      if (h != NULL
+			  && h->root.type == bfd_link_hash_undefweak)
+			ok_tprel = TRUE;
+		      else
+			{
+			  value += sym_sec->output_offset;
+			  value += sym_sec->output_section->vma;
+			  value -= htab->elf.tls_sec->vma;
+			  ok_tprel = (value + TP_OFFSET + ((bfd_vma) 1 << 31)
+				      < (bfd_vma) 1 << 32);
+			}
 		    }
 
 		  r_type = ELF64_R_TYPE (rel->r_info);
@@ -11498,6 +11507,7 @@ ppc64_elf_relocate_section (bfd *output_
 	     checking whether the function is defined.  */
 	  else if (h != NULL
 		   && h->elf.root.type == bfd_link_hash_undefweak
+		   && h->elf.dynindx == -1
 		   && r_type == R_PPC64_REL24
 		   && relocation == 0
 		   && addend == 0)
@@ -11833,6 +11843,22 @@ ppc64_elf_relocate_section (bfd *output_
 	case R_PPC64_TPREL16_HIGHERA:
 	case R_PPC64_TPREL16_HIGHEST:
 	case R_PPC64_TPREL16_HIGHESTA:
+	  if (h != NULL
+	      && h->elf.root.type == bfd_link_hash_undefweak
+	      && h->elf.dynindx == -1)
+	    {
+	      /* Make this relocation against an undefined weak symbol
+		 resolve to zero.  This is really just a tweak, since
+		 code using weak externs ought to check that they are
+		 defined before using them.  */
+	      bfd_byte *p = contents + rel->r_offset - d_offset;
+
+	      insn = bfd_get_32 (output_bfd, p);
+	      insn = _bfd_elf_ppc_at_tprel_transform (insn, 13);
+	      if (insn != 0)
+		bfd_put_32 (output_bfd, insn, p);
+	      break;
+	    }
 	  addend -= htab->elf.tls_sec->vma + TP_OFFSET;
 	  if (info->shared)
 	    /* The TPREL16 relocs shouldn't really be used in shared
Index: bfd/elf32-ppc.c
===================================================================
RCS file: /cvs/src/src/bfd/elf32-ppc.c,v
retrieving revision 1.273
diff -u -p -r1.273 elf32-ppc.c
--- bfd/elf32-ppc.c	17 Dec 2009 05:45:25 -0000	1.273
+++ bfd/elf32-ppc.c	25 Jan 2010 05:02:40 -0000
@@ -1,6 +1,6 @@
 /* PowerPC-specific support for 32-bit ELF
    Copyright 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
-   2004, 2005, 2006, 2007, 2008, 2009 Free Software Foundation, Inc.
+   2004, 2005, 2006, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
    Written by Ian Lance Taylor, Cygnus Support.
 
    This file is part of BFD, the Binary File Descriptor library.
@@ -6669,6 +6669,51 @@ _bfd_elf_ppc_at_tls_transform (unsigned 
   return insn;
 }
 
+/* If INSN is an opcode that may be used with an @tprel operand, return
+   the transformed insn for an undefined weak symbol, ie. with the
+   thread pointer REG operand removed.  Otherwise return 0.  */
+
+unsigned int
+_bfd_elf_ppc_at_tprel_transform (unsigned int insn, unsigned int reg)
+{
+  if ((insn & (0x1f << 16)) == reg << 16
+      && ((insn & (0x3f << 26)) == 14u << 26 /* addi */
+	  || (insn & (0x3f << 26)) == 15u << 26 /* addis */
+	  || (insn & (0x3f << 26)) == 32u << 26 /* lwz */
+	  || (insn & (0x3f << 26)) == 34u << 26 /* lbz */
+	  || (insn & (0x3f << 26)) == 36u << 26 /* stw */
+	  || (insn & (0x3f << 26)) == 38u << 26 /* stb */
+	  || (insn & (0x3f << 26)) == 40u << 26 /* lhz */
+	  || (insn & (0x3f << 26)) == 42u << 26 /* lha */
+	  || (insn & (0x3f << 26)) == 44u << 26 /* sth */
+	  || (insn & (0x3f << 26)) == 46u << 26 /* lmw */
+	  || (insn & (0x3f << 26)) == 47u << 26 /* stmw */
+	  || (insn & (0x3f << 26)) == 48u << 26 /* lfs */
+	  || (insn & (0x3f << 26)) == 50u << 26 /* lfd */
+	  || (insn & (0x3f << 26)) == 52u << 26 /* stfs */
+	  || (insn & (0x3f << 26)) == 54u << 26 /* stfd */
+	  || ((insn & (0x3f << 26)) == 58u << 26 /* lwa,ld,lmd */
+	      && (insn & 3) != 1)
+	  || ((insn & (0x3f << 26)) == 62u << 26 /* std, stmd */
+	      && ((insn & 3) == 0 || (insn & 3) == 3))))
+    {
+      insn &= ~(0x1f << 16);
+    }
+  else if ((insn & (0x1f << 21)) == reg << 21
+	   && ((insn & (0x3e << 26)) == 24u << 26 /* ori, oris */
+	       || (insn & (0x3e << 26)) == 26u << 26 /* xori,xoris */
+	       || (insn & (0x3e << 26)) == 28u << 26 /* andi,andis */))
+    {
+      insn &= ~(0x1f << 21);
+      insn |= (insn & (0x1f << 16)) << 5;
+      if ((insn & (0x3e << 26)) == 26 << 26 /* xori,xoris */)
+	insn -= 2 >> 26;  /* convert to ori,oris */
+    }
+  else
+    insn = 0;
+  return insn;
+}
+
 /* The RELOCATE_SECTION function is called by the ELF backend linker
    to handle the relocations for a section.
 
@@ -7471,6 +7516,21 @@ ppc_elf_relocate_section (bfd *output_bf
 	case R_PPC_TPREL16_LO:
 	case R_PPC_TPREL16_HI:
 	case R_PPC_TPREL16_HA:
+	  if (h != NULL
+	      && h->root.type == bfd_link_hash_undefweak
+	      && h->dynindx == -1)
+	    {
+	      /* Make this relocation against an undefined weak symbol
+		 resolve to zero.  This is really just a tweak, since
+		 code using weak externs ought to check that they are
+		 defined before using them.  */
+	      bfd_byte *p = contents + rel->r_offset - d_offset;
+	      unsigned int insn = bfd_get_32 (output_bfd, p);
+	      insn = _bfd_elf_ppc_at_tprel_transform (insn, 2);
+	      if (insn != 0)
+		bfd_put_32 (output_bfd, insn, p);
+	      break;
+	    }
 	  addend -= htab->elf.tls_sec->vma + TP_OFFSET;
 	  /* The TPREL16 relocs shouldn't really be used in shared
 	     libs as they will result in DT_TEXTREL being set, but
Index: bfd/bfd-in.h
===================================================================
RCS file: /cvs/src/src/bfd/bfd-in.h,v
retrieving revision 1.148
diff -u -p -r1.148 bfd-in.h
--- bfd/bfd-in.h	18 Nov 2009 12:42:51 -0000	1.148
+++ bfd/bfd-in.h	25 Jan 2010 05:02:40 -0000
@@ -1,7 +1,7 @@
 /* Main header file for the bfd library -- portable access to object files.
 
    Copyright 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
-   1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
+   1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
    Free Software Foundation, Inc.
 
    Contributed by Cygnus Support.
@@ -911,6 +911,9 @@ extern bfd_boolean elf32_arm_fix_exidx_c
 /* PowerPC @tls opcode transform/validate.  */
 extern unsigned int _bfd_elf_ppc_at_tls_transform
   (unsigned int, unsigned int);
+/* PowerPC @tprel opcode transform/validate.  */
+extern unsigned int _bfd_elf_ppc_at_tprel_transform
+  (unsigned int, unsigned int);
 
 /* TI COFF load page support.  */
 extern void bfd_ticoff_set_section_load_page

-- 
Alan Modra
Australia Development Lab, IBM