On 01/07, Srivatsa Vaddagiri wrote: > > On Sun, Jan 07, 2007 at 05:22:46PM +0300, Oleg Nesterov wrote: > > On 01/07, Oleg Nesterov wrote: > > > > > > Thoughts? > > > > How about: > > > > CPU_DEAD does nothing. After __cpu_disable() cwq->thread runs on > > all CPUs and becomes idle when it flushes cwq->worklist: nobody > ^^^ > > all except dead cpus that is.
yes, of course. > > > will add work_struct on that list. > > If CPU_DEAD does nothing, then the dead cpu's workqueue list may be > non-empty. How will it be flushed, given that no thread can run on the > dead cpu? But cwq->thread is not bound to the dead CPU at this point, it was aleady migrated (like all other threads which had that CPU in ->cpus_allowed). > Finally, I am concerned about the (un)friendliness of this programming > model, where programmers are restricted in not having a stable access to > cpu_online_map at all -and- also requiring them to code in non-obvious > terms. Granted that writing hotplug-safe code is non-trivial, but the > absence of "safe access to online_map" will make it more complicated. I guess you misunderstood me, I meant CPU_DEAD does nothing only in workqueue.c:workqueue_cpu_callback(). Oleg. - 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/
