| Date | Thu, 28 May 2020 14:28:46 +0200 | From | Frederic Weisbecker <> | Subject | Re: [RFC][PATCH 2/7] smp: Optimize flush_smp_call_function_queue() |
| |
On Tue, May 26, 2020 at 06:10:59PM +0200, Peter Zijlstra wrote: > The call_single_queue can contain (two) different callbacks, > synchronous and asynchronous. The current interrupt handler runs them > in-order, which means that remote CPUs that are waiting for their > synchronous call can be delayed by running asynchronous callbacks. > > Rework the interrupt handler to first run the synchonous callbacks. > > Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
I hope we are not unlucky enough to have dependencies between some async and sync callbacks that require the whole execution in order.
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
(Sweet llist dance, I think I need fresh air and coffee now).
|