Always descend into output section statements in lang_do_assignments

Alan Modra amodra@gmail.com
Tue Oct 11 07:42:00 GMT 2016


On Fri, Jul 08, 2016 at 03:35:59PM +0930, Alan Modra wrote:
> On Thu, Jul 07, 2016 at 02:47:45PM +0200, Sebastian Huber wrote:
> > thanks for your analysis. It looks complicated, is this a bug or a feature?
> > If its a bug, is it worth to fix or is my use case too exotic?
> 
> It is a bug, and I wrote a patch for the lang_do_assignments fix early
> last month when looking into another problem.  I'll dust it off.

See https://sourceware.org/ml/binutils/2016-07/msg00091.html
This patch stop --gc-sections elf_gc_sweep_symbol localizing symbols
that ought to remain global.

The difficulty with always descending into output section statements
is that symbols defined by the script in such statements don't have
a bfd section when lang_do_assignments runs early in the link process.
There are two approaches to curing this problem.  Either we can
create the bfd section early, or we can use a special section.  This
patch takes the latter approach and uses bfd_und_section.  (Creating
bfd sections early results in changed output section order, and thus
lots of testsuite failures.  You can't create all output sections
early to ensure proper ordering as KEEP then stops empty sections
from being stripped.)

The wrinkle with this approach is that some code that runs at
gc-sections time needs to be made aware of the odd defined symbols
using bfd_und_section.

bfd/
	* elf64-x86-64.c (elf_x86_64_convert_load_reloc): Handle symbols
	defined temporarily with bfd_und_section.
	* elflink.c (_bfd_elf_gc_keep): Don't set SEC_KEEP for bfd_und_section.
	* elfxx-mips.c (mips_elf_local_pic_function_p): Exclude defined
	symbols with bfd_und_section.
ld/
	* ldlang.c (lang_do_assignments_1): Descend into output section
	statements that do not yet have bfd sections.  Set symbol section
	temporarily for symbols defined in such statements to the undefined
	section.  Don't error on data or reloc statements until final phase.
	* ldexp.c (exp_fold_tree_1 <etree_assign>): Handle bfd_und_section
	in expld.section.
	* testsuite/ld-mmix/bpo-10.d: Adjust.
	* testsuite/ld-mmix/bpo-11.d: Adjust.

diff --git a/bfd/elf64-x86-64.c b/bfd/elf64-x86-64.c
index 2a372a6..79b8f1c 100644
--- a/bfd/elf64-x86-64.c
+++ b/bfd/elf64-x86-64.c
@@ -1877,7 +1877,10 @@ elf_x86_64_convert_load_reloc (bfd *abfd, asection *sec,
 	     bfd_elf_record_link_assignment.   */
 	  if (h->def_regular
 	      && (h->root.type == bfd_link_hash_new
-		  || h->root.type == bfd_link_hash_undefined))
+		  || h->root.type == bfd_link_hash_undefined
+		  || ((h->root.type == bfd_link_hash_defined
+		       || h->root.type == bfd_link_hash_defweak)
+		      && h->root.u.def.section == bfd_und_section_ptr)))
 	    {
 	      /* Skip since R_X86_64_32/R_X86_64_32S may overflow.  */
 	      if (require_reloc_pc32)
diff --git a/bfd/elflink.c b/bfd/elflink.c
index 1591682..2f40eb0 100644
--- a/bfd/elflink.c
+++ b/bfd/elflink.c
@@ -13037,7 +13037,8 @@ _bfd_elf_gc_keep (struct bfd_link_info *info)
       if (h != NULL
 	  && (h->root.type == bfd_link_hash_defined
 	      || h->root.type == bfd_link_hash_defweak)
-	  && !bfd_is_abs_section (h->root.u.def.section))
+	  && !bfd_is_abs_section (h->root.u.def.section)
+	  && !bfd_is_und_section (h->root.u.def.section))
 	h->root.u.def.section->flags |= SEC_KEEP;
     }
 }
diff --git a/bfd/elfxx-mips.c b/bfd/elfxx-mips.c
index d618e54..cc29c0e 100644
--- a/bfd/elfxx-mips.c
+++ b/bfd/elfxx-mips.c
@@ -1808,6 +1808,7 @@ mips_elf_local_pic_function_p (struct mips_elf_link_hash_entry *h)
 	   || h->root.root.type == bfd_link_hash_defweak)
 	  && h->root.def_regular
 	  && !bfd_is_abs_section (h->root.root.u.def.section)
+	  && !bfd_is_und_section (h->root.root.u.def.section)
 	  && (!ELF_ST_IS_MIPS16 (h->root.other)
 	      || (h->fn_stub && h->need_fn_stub))
 	  && (PIC_OBJECT_P (h->root.root.u.def.section->owner)
diff --git a/ld/ldexp.c b/ld/ldexp.c
index b2c7620..cb5e093 100644
--- a/ld/ldexp.c
+++ b/ld/ldexp.c
@@ -1077,6 +1077,7 @@ exp_fold_tree_1 (etree_type *tree)
 		 before relaxation and so be stripped incorrectly.  */
 	      if (expld.phase == lang_mark_phase_enum
 		  && expld.section != bfd_abs_section_ptr
+		  && expld.section != bfd_und_section_ptr
 		  && !(expld.result.valid_p
 		       && expld.result.value == 0
 		       && (is_value (tree->assign.src, 0)
@@ -1085,7 +1086,8 @@ exp_fold_tree_1 (etree_type *tree)
 			   || is_align_conditional (tree->assign.src))))
 		expld.section->flags |= SEC_KEEP;
 
-	      if (!expld.result.valid_p)
+	      if (!expld.result.valid_p
+		  || expld.section == bfd_und_section_ptr)
 		{
 		  if (expld.phase != lang_mark_phase_enum)
 		    einfo (_("%F%S invalid assignment to"
diff --git a/ld/ldlang.c b/ld/ldlang.c
index fe0b844..af2ca99 100644
--- a/ld/ldlang.c
+++ b/ld/ldlang.c
@@ -5618,6 +5618,7 @@ lang_do_assignments_1 (lang_statement_union_type *s,
 	case lang_output_section_statement_enum:
 	  {
 	    lang_output_section_statement_type *os;
+	    bfd_vma newdot;
 
 	    os = &(s->output_section_statement);
 	    os->after_end = *found_end;
@@ -5629,18 +5630,24 @@ lang_do_assignments_1 (lang_statement_union_type *s,
 		    prefer_next_section = FALSE;
 		  }
 		dot = os->bfd_section->vma;
-
-		lang_do_assignments_1 (os->children.head,
-				       os, os->fill, dot, found_end);
-
-		/* .tbss sections effectively have zero size.  */
-		if (!IS_TBSS (os->bfd_section)
-		    || bfd_link_relocatable (&link_info))
-		  dot += TO_ADDR (os->bfd_section->size);
-
-		if (os->update_dot_tree != NULL)
-		  exp_fold_tree (os->update_dot_tree, bfd_abs_section_ptr,
-				 &dot);
+	      }
+	    newdot = lang_do_assignments_1 (os->children.head,
+					    os, os->fill, dot, found_end);
+	    if (!os->ignored)
+	      {
+		if (os->bfd_section != NULL)
+		  {
+		    /* .tbss sections effectively have zero size.  */
+		    if (!IS_TBSS (os->bfd_section)
+			|| bfd_link_relocatable (&link_info))
+		      dot += TO_ADDR (os->bfd_section->size);
+
+		    if (os->update_dot_tree != NULL)
+		      exp_fold_tree (os->update_dot_tree,
+				     bfd_abs_section_ptr, &dot);
+		  }
+		else
+		  dot = newdot;
 	      }
 	  }
 	  break;
@@ -5664,7 +5671,7 @@ lang_do_assignments_1 (lang_statement_union_type *s,
 	      if (expld.result.section != NULL)
 		s->data_statement.value += expld.result.section->vma;
 	    }
-	  else
+	  else if (expld.phase == lang_final_phase_enum)
 	    einfo (_("%F%P: invalid data statement\n"));
 	  {
 	    unsigned int size;
@@ -5697,7 +5704,7 @@ lang_do_assignments_1 (lang_statement_union_type *s,
 			 bfd_abs_section_ptr, &dot);
 	  if (expld.result.valid_p)
 	    s->reloc_statement.addend_value = expld.result.value;
-	  else
+	  else if (expld.phase == lang_final_phase_enum)
 	    einfo (_("%F%P: invalid reloc statement\n"));
 	  dot += TO_ADDR (bfd_get_reloc_size (s->reloc_statement.howto));
 	  break;
@@ -5733,7 +5740,8 @@ lang_do_assignments_1 (lang_statement_union_type *s,
 		*found_end = TRUE;
 	    }
 	  exp_fold_tree (s->assignment_statement.exp,
-			 current_os->bfd_section,
+			 (current_os->bfd_section != NULL
+			  ? current_os->bfd_section : bfd_und_section_ptr),
 			 &dot);
 	  break;
 
diff --git a/ld/testsuite/ld-mmix/bpo-10.d b/ld/testsuite/ld-mmix/bpo-10.d
index 1cb7d61..6f49287 100644
--- a/ld/testsuite/ld-mmix/bpo-10.d
+++ b/ld/testsuite/ld-mmix/bpo-10.d
@@ -14,11 +14,10 @@ SYMBOL TABLE:
 0+7f8 l +d  \.MMIX.reg_contents	0+ (|\.MMIX\.reg_contents)
 0+ l    df \*ABS\*	0+ .*
 0+ l       \.init	0+ _start
-0+ l    df \*ABS\*	0+ .*
-0+4 l       \.init	0+ _start\.
 2000000000000000 g       \.init	0+ __bss_start
 2000000000000000 g       \.init	0+ _edata
 2000000000000000 g       \.init	0+ _end
+0+4 g       \.init	0+ _start\.
 
 Contents of section \.init:
  0000 e37704a6                             .*
diff --git a/ld/testsuite/ld-mmix/bpo-11.d b/ld/testsuite/ld-mmix/bpo-11.d
index 9b38be8..1abf381 100644
--- a/ld/testsuite/ld-mmix/bpo-11.d
+++ b/ld/testsuite/ld-mmix/bpo-11.d
@@ -16,13 +16,12 @@ SYMBOL TABLE:
 0+7e8 l    d  \.MMIX\.reg_contents	0+ (|\.MMIX\.reg_contents)
 0+ l    df \*ABS\*	0+ .*
 0+ l       \.init	0+ _start
-0+ l    df \*ABS\*	0+ .*
-0+10 l       \.text	0+ _start\.
 0+14 g       \.text	0+ x
 0+10 g       \.text	0+ x2
 2000000000000000 g       \.text	0+ __bss_start
 2000000000000000 g       \.text	0+ _edata
 2000000000000000 g       \.text	0+ _end
+0+10 g       \.text	0+ _start\.
 
 Contents of section \.init:
  0000 00000000 0000003d 00000000 0000003a  .*

-- 
Alan Modra
Australia Development Lab, IBM



More information about the Binutils mailing list