Re: CLONE_KILLVMCH flag proposal

[Alan Cox]

> "p->mm" is unreliable: theoritically, a thread could be in temporary lazy
> tlb mode (start_lazy_tlb() in kernel/exit.c), then p->mm will be NULL. AFAIK
> this feature is not yet used.

if you wanted to kill all the threads of a clone without any suprises then
probably you want to look at how p->parent_exec_id and p->self_exec_id
are used for parent signalling on thread death. I think there is enough info
to actually answer the 'who are the threads' question