[PATCH,V4 7/9] ld: bfd: sframe: fix incorrect r_offset in RELA entries

[Jan Beulich]

On 19.06.2025 08:24, Indu Bhagat wrote:
> +/* Adjust an address in the .sframe section.  Given OFFSET within
> +   SEC, this returns the new offset in the merged .sframe section,
> +   or -1 if the address refers to an FDE which has been removed.
> +
> +   PS: This function assumes that _bfd_elf_merge_section_sframe has
> +   not been called on the input section SEC yet.  Note how it uses
> +   sframe_encoder_get_num_fidx () to figure out the offset of FDE
> +   in the output section.  */
> +
> +bfd_vma
> +_bfd_elf_sframe_section_offset (bfd *output_bfd ATTRIBUTE_UNUSED,
> +				struct bfd_link_info *info,
> +				asection *sec,
> +				bfd_vma offset)
> +{
> +  struct sframe_dec_info *sfd_info;
> +  struct sframe_enc_info *sfe_info;
> +  sframe_decoder_ctx *sfd_ctx;
> +  sframe_encoder_ctx *sfe_ctx;
> +  struct elf_link_hash_table *htab;
> +
> +  unsigned int sec_fde_idx, out_num_fdes;
> +  unsigned int sfd_num_fdes, sfe_num_fdes;
> +  uint32_t sfd_fde_offset;
> +  bfd_vma new_offset;
> +
> +  if (sec->sec_info_type != SEC_INFO_TYPE_SFRAME)
> +    return offset;
> +
> +  sfd_info = (struct sframe_dec_info *) elf_section_data (sec)->sec_info;

Oh, actually, one cosmetic thing: Looks like this cast isn't really needed.
Please generally omit casts when they're not needed. Yes, there are still
very many of them in the code base, but quite often they're simply leftovers
from long gone K&R days.

Jan