Re: [PATCH, RFC] reimplement flush_workqueue()

[Srivatsa Vaddagiri]

On Fri, Jan 05, 2007 at 03:42:46PM +0300, Oleg Nesterov wrote:
> preempt_disable() can't prevent cpu_up, but flush_workqueue() doesn't care
> _unless_ cpu_down also happened meantime (and hence a fresh CPU may have
> pending work_structs which were moved from a dead CPU).

Yes, that was what I had in mind.

> So you are right, we still need the patch above, but I think we don't have
> new problems with preempt_disable().

Right, preempt_disable() hasn't added any new problem. Its just
revealing the same problem as earlier, by opening up window for cpu
hotplug events to happen in the middle of flush_workqueue().

Ideally I would have liked a lock_cpu_hotplug() equivalent which blocks
all cpu hotplug events during the entire flush_workqueue(). In its
absence, I guess we just need to deal with all these ugly races ..

In summary, I think we need to go ahead with the preemp_disable() patch
in flush_workqueue() from Andrew and the race fix you posted here:

	http://lkml.org/lkml/2006/12/30/37

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