Hi Thomas, On 09/02/2016 05:12 PM, Thomas Gleixner wrote: > On Mon, 1 Aug 2016, Sebastian Frias wrote: >> NOTE: While the proposed unmap() function attempts to undo as much things >> as done by the map() function, I did not find a way to undo the following: >> >> a) irq_gc_init_mask_cache(gc, dgc->gc_flags) > > You can't undo that. Because that represents the mask cache of the irq chip > and that is required to be consistent over the life time of the irq > chip. > > Unmapping does not make the generic chip and the underlying irqchip go > away. > >> b) irq_set_lockdep_class(virq, &irq_nested_lock_class) > > No point in undoing that. The irq descriptor is released on unmap. > >> c) irq_modify_status(virq, dgc->irq_flags_to_clear, dgc->irq_flags_to_set) > > See b) >
Thanks for your time, I gather from this that the patch is ok then? I submitted a few more patches, one of them a follow up of this one, and some more, here are pointers to them just in case: https://lkml.org/lkml/2016/8/1/302 https://lkml.org/lkml/2016/8/2/118 https://lkml.org/lkml/2016/8/19/598 Best regards, Sebastian
