re: __atomic_test_and_set() and mips o32 - help wanted

["Maciej W. Rozycki" <macro%orcam.me.uk@localhost>]

On Sat, 22 Nov 2025, matthew green wrote:

> FWIW, one problem with the RAS solution is that we disable RAS
> on SMP mips kernels, and i'm not sure how it would be fixed..

 Even if booted on a UP system?  That seems suboptimal to me.

 Also I gather a vDSO is not used to implement this after all, because 
then it would be entirely under the kernel control what mechanism to 
export to the userland for abstracted atomics support (also to possibly 
make use of the 58x0/ISIS interlock).

> someone else could totally implement a better solution -- my
> goal is to keep mips working with newer GCC, and for something
> that is only rarely used (so far, only GDB has triggered this,
> though anything using modern C++ test_and_set() would also
> trigger the same issue -- this is how GDB does), i'm not going
> to worry about better myself.

 FWIW I find it a pragmatic approach: better is the enemy of good enough, 
so make things work first and only then consider improvements.  Perhaps 
someone jumps in who has the enthusiasm and resources to work on the more 
sophisticated solution.

  Maciej