Re: [PATCH v2 1/1] gdb: Preserve IFUNC marker when finding inferior functions

[Muhammad Kamran <muhammad.kamran@arm.com>]

Hi Simon,

On 03/07/2026 13:56, Simon Marchi wrote:
> On 6/30/26 11:29 AM, Wilco Dijkstra wrote:
>> Hi Muhammad,
>>
>>> Yes.  GDB should prefer the cache/GOT path and avoid resolver calls
>>> where possible.  The patch does not make resolver calls safer or add a
>>> new resolver-calling mechanism; it just preserves the IFUNC marker so
>>> the existing IFUNC resolution path is used instead of calling the
>>> resolver address as though it were malloc itself.
>>
>> Indeed - while the patch works for now as a workaround, directly calling
>> ifunc resolvers without following the ifunc ABI does not work. GDB only
>> passes HWCAP as the first argument, so the resolver may return different
>> functions due to the set of HWCAP values given not being identical between
>> GDB and GLIBC (ie. It could call the wrong malloc implementation that was
>> not selected or initialized).
>>
>> So the best option is to call dlsym() as Florian suggested.
> 
> To track this, could either of you file a GDB bug with the necessary
> background (why it would be good to switch to using to dlsym to resolve
> ifuncs)?

A GDB bug has been created on bugzilla and can be seen here:
https://sourceware.org/bugzilla/show_bug.cgi?id=34367

> 
> Thanks,
> 
> Simon