[GOLD] Add --secure-plt option for ppc32

[GOLD] Add --secure-plt option for ppc32

[Alan Modra]

Added just to accept, and ignore.  gcc since 2015-10-21, when
configured with --enable-secureplt passes this option to the linker.
As powerpc gold cannot link --bss-plt code successfully, gold needs to
accept the option or the gcc specs file needs to be changed if you're
using gold.

The patch also make gold detect --bss-plt code and error out rather
than producing a binary that crashes.

OK?

	* options.h: Add --secure-plt option.
	* powerpc.cc (Target_powerpc::Scan::local): Detect and error
	on -fPIC bss-plt code.
	(Target_powerpc::Scan::global): Likewise.

diff --git a/gold/options.h b/gold/options.h
index d7ac6cd..905162b 100644
--- a/gold/options.h
+++ b/gold/options.h
@@ -1203,6 +1203,10 @@ class General_options
   DEFINE_special(section_start, options::TWO_DASHES, '\0',
 		 N_("Set address of section"), N_("SECTION=ADDRESS"));
 
+  DEFINE_bool(secure_plt, options::TWO_DASHES , '\0', true,
+	      N_("(PowerPC only) Use new-style PLT.  (default and forced)"),
+	      NULL);
+
   DEFINE_optional_string(sort_common, options::TWO_DASHES, '\0', NULL,
 			 N_("Sort common symbols by alignment"),
 			 N_("[={ascending,descending}]"));
diff --git a/gold/powerpc.cc b/gold/powerpc.cc
index 4b5026f..681fac7 100644
--- a/gold/powerpc.cc
+++ b/gold/powerpc.cc
@@ -5969,6 +5969,30 @@ Target_powerpc<size, big_endian>::Scan::local(
       break;
     }
 
+  if (size == 32)
+    {
+      switch (r_type)
+	{
+	case elfcpp::R_POWERPC_REL32:
+	  if (ppc_object->got2_shndx() != 0
+	      && parameters->options().output_is_position_independent())
+	    {
+	      unsigned int shndx = lsym.get_st_shndx();
+	      unsigned int r_sym = elfcpp::elf_r_sym<size>(reloc.get_r_info());
+	      bool is_ordinary;
+	      shndx = ppc_object->adjust_sym_shndx(r_sym, shndx, &is_ordinary);
+	      if (is_ordinary && shndx == ppc_object->got2_shndx()
+		  && (ppc_object->section_flags(data_shndx)
+		      & elfcpp::SHF_EXECINSTR) != 0)
+		gold_error(_("%s: unsupported -mbss-plt code"),
+			   ppc_object->name().c_str());
+	    }
+	  break;
+	default:
+	  break;
+	}
+    }
+
   switch (r_type)
     {
     case elfcpp::R_POWERPC_GOT_TLSLD16:
@@ -6470,6 +6494,20 @@ Target_powerpc<size, big_endian>::Scan::global(
       break;
     }
 
+  if (size == 32)
+    {
+      switch (r_type)
+	{
+	case elfcpp::R_PPC_LOCAL24PC:
+	  if (strcmp(gsym->name(), "_GLOBAL_OFFSET_TABLE_") == 0)
+	    gold_error(_("%s: unsupported -mbss-plt code"),
+		       ppc_object->name().c_str());
+	  break;
+	default:
+	  break;
+	}
+    }
+
   switch (r_type)
     {
     case elfcpp::R_POWERPC_GOT_TLSLD16:

-- 
Alan Modra
Australia Development Lab, IBM

Re: [GOLD] Add --secure-plt option for ppc32

[Cary Coutant]

> +  DEFINE_bool(secure_plt, options::TWO_DASHES , '\0', true,
> +             N_("(PowerPC only) Use new-style PLT.  (default and forced)"),
> +             NULL);

Gold's --help will now print "(default)" automatically for this
option, so it's redundant here. What do you mean by "forced"?

-cary

Re: [GOLD] Add --secure-plt option for ppc32

[Alan Modra]

On Mon, Jan 09, 2017 at 09:10:29AM -0800, Cary Coutant wrote:
> > +  DEFINE_bool(secure_plt, options::TWO_DASHES , '\0', true,
> > +             N_("(PowerPC only) Use new-style PLT.  (default and forced)"),
> > +             NULL);
> 
> Gold's --help will now print "(default)" automatically for this
> option, so it's redundant here. What do you mean by "forced"?

Oops, didn't even look at the help output.  By forced I meant that
there is no --no-secure-plt (or --bss-plt) but saying that isn't
necessary.  Will remove it and the unnecessary "default".

-- 
Alan Modra
Australia Development Lab, IBM

Re: [GOLD] Add --secure-plt option for ppc32

[Cary Coutant]

>> > +  DEFINE_bool(secure_plt, options::TWO_DASHES , '\0', true,
>> > +             N_("(PowerPC only) Use new-style PLT.  (default and forced)"),
>> > +             NULL);
>>
>> Gold's --help will now print "(default)" automatically for this
>> option, so it's redundant here. What do you mean by "forced"?
>
> Oops, didn't even look at the help output.  By forced I meant that
> there is no --no-secure-plt (or --bss-plt) but saying that isn't
> necessary.  Will remove it and the unnecessary "default".

OK with that change.

-cary