[PATCH] Check corrupt VTENTRY entry in bfd_elf_gc_record_vtentry

[Alan Modra]

On Wed, Apr 10, 2019 at 08:32:08PM -0700, H.J. Lu wrote:
> 	* elf-m10300.c (mn10300_elf_check_relocs): Remove BFD_ASSERT of
> 	"h != NULL".  Don't check "h != NULL" before calling.
> 	bfd_elf_gc_record_vtentry.
> 	* elf32-arm.c (elf32_arm_check_relocs): Likewise.
> 	* elf32-bfin.c (bfin_check_relocs): Likewise.
> 	* elf32-cris.c (cris_elf_check_relocs): Likewise.
> 	* elf32-csky.c (csky_elf_check_relocs): Likewise.
> 	* elf32-d10v.c (elf32_d10v_check_relocs): Likewise.
> 	* elf32-dlx.c (elf32_dlx_check_relocs): Likewise.
> 	* elf32-fr30.c (fr30_elf_check_relocs): Likewise.
> 	* elf32-frv.c (elf32_frv_check_relocs): Likewise.
> 	* elf32-hppa.c (elf32_hppa_check_relocs): Likewise.
> 	* elf32-i386.c (elf_i386_check_relocs): Likewise.
> 	* elf32-iq2000.c (iq2000_elf_check_relocs): Likewise.
> 	* elf32-m32r.c (m32r_elf_check_relocs): Likewise.
> 	* elf32-m68hc1x.c (elf32_m68hc11_check_relocs): Likewise.
> 	* elf32-m68k.c (elf_m68k_check_relocs): Likewise.
> 	* elf32-mcore.c (mcore_elf_check_relocs): Likewise.
> 	* elf32-metag.c (elf_metag_check_relocs): Likewise.
> 	* elf32-or1k.c (or1k_elf_check_relocs): Likewise.
> 	* elf32-ppc.c (ppc_elf_check_relocs): Likewise.
> 	* elf32-s390.c (elf_s390_check_relocs): Likewise.
> 	* elf32-sh.c (sh_elf_check_relocs): Likewise.
> 	* elf32-v850.c (v850_elf_check_relocs): Likewise.
> 	* elf32-vax.c (elf_vax_check_relocs): Likewise.
> 	* elf32-xstormy16.c (xstormy16_elf_check_relocs): Likewise.
> 	* elf32-xtensa.c (elf_xtensa_check_relocs): Likewise.
> 	* elf64-mmix.c (mmix_elf_check_relocs): Likewise.
> 	* elf64-ppc.c (ppc64_elf_check_relocs): Likewise.
> 	* elf64-s390.c (elf_s390_check_relocs): Likewise.
> 	* elf64-x86-64.c (elf_s390_check_relocs): Likewise.
> 	* elfxx-mips.c (_bfd_mips_elf_check_relocs): Likewise.
> 	* elfxx-sparc.c (_bfd_sparc_elf_check_relocs): Likewise.
> 	* elflink.c (bfd_elf_gc_record_vtinherit): Check for corrupt
> 	VTENTRY entry.

OK, thanks, except

> --- a/bfd/elf32-hppa.c
> +++ b/bfd/elf32-hppa.c
> @@ -1273,9 +1273,9 @@ elf32_hppa_check_relocs (bfd *abfd,
>  	  /* This relocation describes which C++ vtable entries are actually
>  	     used.  Record for later use during GC.  */
>  	case R_PARISC_GNU_VTENTRY:
> -	  BFD_ASSERT (hh != NULL);
> -	  if (hh != NULL
> -	      && !bfd_elf_gc_record_vtentry (abfd, sec, &hh->eh, rela->r_addend))
> +	  if (!bfd_elf_gc_record_vtentry (abfd, sec,
> +					  hh ? &hh->eh : NULL,
> +					  rela->r_addend))
>  	    return FALSE;
>  	  continue;
>  

let's not special case hh being NULL here.  The call as written before
should be fine.  &hh->eh is equivalent to
(struct elf_link_hash_entry *) hh.

Is this a work-around for some nonsense compiler warning?  If so, how
does it fare with the typical offsetof macro?
#define offsetof(type, member) ((size_t) &((type *) 0)->member)

-- 
Alan Modra
Australia Development Lab, IBM