Messages in this thread | | | Date | Tue, 4 Oct 2011 19:14:37 +0200 | From | Oleg Nesterov <> | Subject | Re: [RFC][PATCH 0/5] Signal scalability series |
| |
On 10/03, Peter Zijlstra wrote: > > On Mon, 2011-10-03 at 15:07 +0200, Oleg Nesterov wrote: > > On 10/01, Peter Zijlstra wrote: > > > But this series can't help afaics. At least in its current state. It > > only adds more locking to the sending paths. > > Right, so I was hoping Matt had a plan (TM)... :-)
Whatever we do with the locking, this can't remove O(nr_threads), although read_lock() could help to reduce the contention.
> > I think, the best solution would be: never send the signal from irq > > context, and ->siglock shouldn't disable irqs. > > Bit hard that, posix timers need to deliver signals which pretty much > mandates we do something from irq context
Of course. We should notify a thread even if it blocks the signal.
> (and the round-trip through > softirq context really isn't pretty nor good for performance).
No, no. I meant, it would be nice to offload the sending to the target. The process itself should take care.
> > Probably this is possible too. I was thinking anout this when > > set_current_blocked() was added. Unfortunately this needs a lot of > > complications. > > Right, so the thing Thomas and I have been promoting for a while now is > to update a signal target vector on every signal mask update. Mask > updates should be the slow path. This would leave us with a ready target > in O(1).
Yes. This is the "obvious" solution ;) Now that we have set_current_blocked() this is simple. Except, of course, this blows signal_struct and set_current_blocked() can't rely on TIF_SIGPENDING. But we can probably add TIF_YOU_ARE_LISTED_IN_CURR_TARGET_ARRAY.
And in fact this was _one_ of the reasons for set_current_blocked().
> > Agreed. But I am not sure how much we should split the locking when > > it comes to sending/dequeueing/etc signals. 5 locks seems too much. > > It doesn't need all 5 locks to send a signal, does it?
Depending on group/private 3 or 4. Plus ->ctrl_lock if the signal is fatal but I don't really understand this part... Fortunately I think this is not needed ;)
And mostly (afaics) this tries to help to dequeue the signal, not to send. At least currently.
Oleg.
| |