Supporting CLONE_NEWPID namespaces and process-shared mutexes supported?

[Carlos O'Donell]

In 2014 we saw a case of the problems associated with CLONE_NEWPID:
https://sourceware.org/ml/libc-alpha/2014-11/msg00379.html

We use tid's in the internals of a mutex to identify owner. This is
considered unique for the system (though it has an ABA pid-reuse
problem).

Should the community be looking at what is required to support
process-shared synchronization objects across two namespaces
with distinct PID sets?

Does this boil down to a kernel syscall that gives us the untranslated
pid of the thread/process?

-- 
Cheers,
Carlos.