Re: RFC [patch 13/34] PID Virtualization Define new task_pid api

[ebiederm@xmission.com (Eric W. Biederman)]

Arjan van de Ven <arjan@infradead.org> writes:

>> 
>> A pointer to a task_struct while it kind of sort of works.  Is not
>> a good solution.  The problem is that in a lot of cases we register
>> a pid to get a signal or something similar and then we never unregister
>> it.  So by using a pointer to a trask_struct you effectively hold the
>> process in memory forever.
>
> this is not right. Both the PID and the task struct have the exact same
> lifetime rules, they HAVE to, to guard against pid reuse problems.

Yes PIDs reserved for the lifetime of the task_struct (baring minor
details).  There are actually a few races in /proc where it can
see the task_struct after the pid has been freed (see the pid_alive macro
in sched.h)

However when used as a reference the number can live as long as you
want.  The classic example is a pid file that can exist even after
you reboot a machine.

So currently a use of a PID as a reference to processes or process
groups can last forever.  An example of this is the kernel is
the result of fcntl(fd, F_SETOWN).  The session of a tty is similar.

Since the in kernel references have a lifetime that is completely
different than the lifetime of a process or a PID.  It is
not safe to simply replace such references with a direct reference
to a task_struct (besides being technically impossible).  Adding
those references could potentially increase the lifespan of a task_struct
for to the life of the kernel depending on the reference.

Eric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/