[V3 10/36] [SFrame-V3] include: libsframe: bfd: gas: testsuite: support for signed 64-bit offset in SFrame FDE

[Jens Remus]

On 1/16/2026 10:16 AM, Jan Beulich wrote:
> On 16.01.2026 09:46, Indu Bhagat wrote:
>> On 1/15/26 11:52 PM, Jan Beulich wrote:
>>> On 13.01.2026 12:12, Indu Bhagat via Binutils wrote:
>>>> --- a/include/sframe.h
>>>> +++ b/include/sframe.h
>>>> @@ -247,9 +247,9 @@ typedef struct sframe_func_desc_entry_v2
>>>>     typedef struct sframe_func_desc_entry_v3
>>>>   {
>>>> -  /* Function start address.  Encoded as a signed offset, relative to the
>>>> -     beginning of the current FDE.  */
>>>> -  int32_t sfde_func_start_address;
>>>> +  /* Offset to the function start address.  Encoded as a signed offset,
>>>> +     relative to the beginning of the current FDE.  */
>>>> +  int64_t sfde_func_start_offset;
>>>>     /* Size of the function in bytes.  */
>>>>     uint32_t sfde_func_size;
>>>
>>> I realize I'm late now with comments, seeing that the whole series was
>>> committed, but still: While the v2 cover letter mentions mitigations
>>> towards the size increase, using 64 bits uniformly still feels odd to
>>> me. Those other size reductions are beneficial indepenently, and for
>>> not overly large binaries it would imo be beneficial if they could get
>>> away with just 32 bits of storage per function start offset. (Likely
>>> this was discussed before,
>>>
>>
>> Allowing 32bits per function start offset was given a thought. (In fact initially I had implemented the 64-bit func start offset usage enabled only via an explicit --gsframe64)
>>
>> I think the main concern was support for static links.  ld will not be able to mix input objects where some have 32-bit func start offset and others have 64-bit func start offset.  Linker would need to "upgrade" 32-bit relocations to 64-bit if needed in that case, something that we thought wouldnt be easy/possible/recommended to do.
> 
> Why would that be? Can't each struct instance flag what width it uses?

This would cause two different FDE sizes, which would make binary search
impossible.  Remember that during final link all FDE are grouped
together and sorted on function start address.  Therefore if FDEs with
32-bit and 64-bit function start address fields would be mixed the
32-bit ones would need to be upgraded to 64-bit.

Regards,
Jens
-- 
Jens Remus
Linux on Z Development (D3303)
jremus@de.ibm.com / jremus@linux.ibm.com

IBM Deutschland Research & Development GmbH; Vorsitzender des Aufsichtsrats: Wolfgang Wendt; Geschäftsführung: David Faller; Sitz der Gesellschaft: Ehningen; Registergericht: Amtsgericht Stuttgart, HRB 243294
IBM Data Privacy Statement: https://www.ibm.com/privacy/