[PATCH] readelf: Handle 0 sh_entsize of GOT sections for x86

[Alan Modra]

On Wed, Sep 03, 2025 at 04:28:30PM -0700, H.J. Lu wrote:
> Some linker sets sh_entsize on .got and .got.plt sections, for example
> x86-64 Google chrome browser:
> 
> $ readelf -SW /opt/google/chrome/chrome | grep got
>   [27] .got              PROGBITS        000000000fb29fc0 fb27fc0 001be0 00  WA  0   0  8
>   [28] .got.plt          PROGBITS        000000000fb2bba0 fb29ba0 0029a8 00  WA  0   0  8
> $
> 
> If sh_entsize of GOT sections is 0 for x86, assume 4 for i386 and 8 for
> x86-64.
> 
> 	PR binutils/33368
> 	* readelf.c (process_got_section_contents): Handle 0 sh_entsize
> 	of GOT sections for i386 and x86-64.
> 
> Signed-off-by: H.J. Lu <hjl.tools@gmail.com>
> ---
>  binutils/readelf.c | 18 ++++++++++++++++++
>  1 file changed, 18 insertions(+)
> 
> diff --git a/binutils/readelf.c b/binutils/readelf.c
> index 4b1a26602c2..c7a66d2a8e0 100644
> --- a/binutils/readelf.c
> +++ b/binutils/readelf.c
> @@ -21150,6 +21150,24 @@ process_got_section_contents (Filedata * filedata)
>  	  }
>  
>  	uint32_t entsz = section->sh_entsize;
> +	/* NB: Some linker sets sh_entsize on .got and .got.plt

You mean, "Some linkers don't set sh_entsize..."

> +	   sections, for example x86-64 Google chrome browser.  */
> +	if (entsz == 0)
> +	  switch (filedata->file_header.e_machine)

Why only a fix for x86?  It seems reasonable to choose entsz based on
file_header.e_ident[EI_CLASS].

> +	    {
> +	    default:
> +	      warn (_("Can't determine section '%s' sh_entsize\n"),
> +		    name);
> +	      res = false;
> +	      goto out;
> +	    case EM_386:
> +	      entsz = 4;
> +	      break;
> +	    case EM_X86_64:
> +	      entsz = 8;
> +	      break;
> +	    }
> +
>  	entries = section->sh_size / entsz;
>  	if (entries == 1)
>  	  printf (_("\nGlobal Offset Table '%s' contains 1 entry:\n"),
> -- 
> 2.51.0

-- 
Alan Modra