On Wed, 23 Mar 2011, David Daney wrote: > On 03/21/2011 02:13 PM, Thomas Gleixner wrote: > > I can't imagine any other purpose for that. > > Modifying the affinity of non-per-cpu IRQs to use the new CPU? Hmm. Good point. Ok, go ahead with that name. > > Hmm. The offline fixup_irq() code is arch specific and usually calls > > desc->irq_data.chip->irq_set_affinity under desc->lock. I have not yet > > found an arch independent way to do that. Any ideas welcome. > > > > There are all the new affinity callbacks, and the things shown in procfs? Are > those handled properly if I call chip->irq_set_affinity? I think not. If the arch code updates irq_data.affinity then yes. I still need to go through all that maze and figure out what all the archs really want to do in that case. Thanks, tglx
