[PATCH v2] PE/COFF: raise normal PE section limit safely

[Jan Beulich]

On 01.07.2026 19:58, Oleg Tolmatcev wrote:
> --- a/bfd/peXXigen.c
> +++ b/bfd/peXXigen.c
> @@ -109,12 +109,70 @@
>  #define SetHighBit(val)      ((val) | 0x80000000)
>  #define WithoutHighBit(val)  ((val) & 0x7fffffff)
>  

> +static bool
> +pe_decode_sym_section_number (bfd *abfd, const char *raw_scnum, int *scnump)
> +{
> +  unsigned int scnum = H_GET_16 (abfd, raw_scnum);
> +
> +  switch (scnum)
> +    {
> +    case IMAGE_SYM_UNDEFINED:
> +      *scnump = N_UNDEF;
> +      return true;
> +    case IMAGE_SYM_ABSOLUTE:
> +      *scnump = N_ABS;
> +      return true;
> +    case IMAGE_SYM_DEBUG:
> +      *scnump = N_DEBUG;
> +      return true;
> +    }
> +
> +  if (scnum > IMAGE_SYM_SECTION_MAX)
> +    {
> +      _bfd_error_handler
> +	/* xgettext:c-format */
> +	(_("%pB: invalid PE symbol section number %#x"), abfd, scnum);

Mind me suggesting s/invalid/unrecognized/ ?

> +      bfd_set_error (bfd_error_bad_value);
> +      return false;
> +    }

While returning an error here may make sense, ...

> @@ -124,7 +182,8 @@ _bfd_XXi_swap_sym_in (bfd * abfd, void * ext1, void * in1)
>      memcpy (in->_n._n_name, ext->e.e_name, SYMNMLEN);
>  
>    in->n_value = H_GET_32 (abfd, ext->e_value);
> -  in->n_scnum = (short) H_GET_16 (abfd, ext->e_scnum);
> +  if (! pe_decode_sym_section_number (abfd, ext->e_scnum, &in->n_scnum))
> +    return;

... that's of limited use here when the caller can't be told of the error.
I don't think bfd_set_error() has any respective effect, unless that
caller actively checked via bfd_get_error(). IOW I think you still need to
set *scnump in pe_decode_sym_section_number(), and bailing out here isn't
helpful.

Oh, there is the ->read_only flag in struct bfd which looks to exist for
purposes like this one. It's for objcopy / strip only though, so in
particular ld could still be in trouble.

Jan