On Sunday, 13 May 2007 23:54, Oleg Nesterov wrote: > On 05/13, Rafael J. Wysocki wrote: > > > > On Sunday, 13 May 2007 23:34, Oleg Nesterov wrote: > > > On 05/13, Rafael J. Wysocki wrote: > > > > > > > > @@ -819,20 +843,31 @@ static int __devinit workqueue_cpu_callb > > > > > > > > + > > > > + case CPU_DEAD_FROZEN: > > > > + if (wq->freezeable) { > > > > + take_over_work(wq, cpu); > > > > + thaw_process(cwq->thread); > > > > > > Suppose that PF_NOFREEZE task T does flush_workqueue(), and CPU 1 has > > > pending > > > works. T does flush_cpu_workqueue(0), CPU_DEAD_FROZEN moves works from > > > CPU 1 > > > to CPU 0, T does flush_cpu_workqueue(1) and finds nothing. > > > > I don't think this is possible, because we've acquired workqueue_mutex in > > _cpu_down(). > > Yes, we did... but flush_workqueue() doesn't take it?
I was looking at the 2.6.21 code, sorry. Hmm, I guess we could add an additional mutex that would only be taken in flush_workqueue() and in _cpu_down()/_cpu_up() via workqueue_cpu_callback() with CPU_LOCK_ACQUIRE? It doesn't seem to be a good idea to run flush_workqueue() while CPUs are being taken up and down anyway. Greetings, Rafael - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/