Re: [Lse-tech] Re: [PATCH] cpusets+hotplug+preepmt broken

[Dipankar Sarma <dipankar@in.ibm.com>]

On Fri, May 13, 2005 at 12:59:53PM -0700, Paul Jackson wrote:
> Srivatsa, replying to Dinakar:
> > This in fact was the reason that we added lock_cpu_hotplug
> > in sched_setaffinity.
> 
> We do need to be clear about how these locks work, their semantics, what
> they require and what they insure, and their various interactions.
> 
> This is not easy stuff to get right.
> 
> I notice that the documentation for lock_cpu_hotplug() is a tad on
> the skimpy side:
> 
> 	/* Stop CPUs going up and down. */
> 
> That's it, so far as I can see.  Interaction of hotplug locking with
> the rest of the kernel has been, is now, and will continue to be, a
> difficult area.  More care than this needs to be put into explaining
> the use, semantics and interactions of any locking involved.
> 
> In particular, in my view, locks should guard data.  What data does
> lock_cpu_hotplug() guard?  I propose that it guards cpu_online_map.
> 
> I recommend considering a different name for this lock.  Perhaps
> 'cpu_online_sem', instead of 'cpucontrol'?   And perhaps the
> lock_cpu_hotplug() should be renamed, to say 'lock_cpu_online'?

No. CPU hotplug uses two different locking - see both lock_cpu_hotplug()
and __stop_machine_run(). Anyone reading cpu_online_map with
preemption disabled is safe from cpu hotplug even without taking
any lock.

Thanks
Dipankar