On Wed, 12 May 2010 08:51:05 +1000 Dave Airlie <airlied at gmail.com> wrote:
> On Wed, May 12, 2010 at 8:32 AM, Andrew Morton > <akpm at linux-foundation.org> wrote: > > On Wed, 12 May 2010 08:22:49 +1000 > > Dave Airlie <airlied at gmail.com> wrote: > > > >> On Wed, May 12, 2010 at 5:57 AM, Chris Wilson <chris at > >> chris-wilson.co.uk> wrote: > >> > On Tue, 11 May 2010 12:10:01 -0700, Andrew Morton <akpm at > >> > linux-foundation.org> wrote: > >> >> On Tue, 11 May 2010 19:52:31 +0100 > >> >> Chris Wilson <chris at chris-wilson.co.uk> wrote: > >> >> > >> >> > On Tue, 11 May 2010 11:35:55 -0400, Andrew Morton <akpm at > >> >> > linux-foundation.org> wrote: > >> >> > > No, io_mapping_map_atomic_wc() cannot be used from [soft]irq > >> >> > > context: > >> >> > > it hardwires use of KM_USER0. __I suggest that > >> >> > > io_mapping_create_wc(), > >> >> > > io_mapping_map_atomic_wc() etc be changed so that the caller passes > >> >> > > in the > >> >> > > KM_foo kmap slot index. > >> >> > > >> >> > Argh, sorry for the noise, read the mail in the wrong order. Thanks > >> >> > for > >> >> > the review. It would be sensible to go with your simpler patch whilst > >> >> > io_mapping_map_atomic_wc() is improved. > >> >> > >> >> OK. __I'll be sending a bunch of fixes Linuswards in an hour or two. > >> >> Should I include this? > >> > > >> > Yes. > >> > > >> > Acked-by: Chris Wilson <chris at chris-wilson.co.uk> > >> > > >> > >> I'm not sure pushing this in at this point is a good idea, if I'm > >> reading it correctly we've no idea what KM_IRQ is being used for, > > > > It's used for taking kmaps from IRQ contexts. > > > >> and > >> this codepath is called from non-irq contexts just as much as irq > >> contexts. > > > > That's fine. __As long as we do a local_irq_disable(), KM_IRQ0 can be > > used from both irq- and non-irq contexts. __All we need to do is to > > ensure that some interrupt cannot come along on this CPU and corrupt > > the slot. > > I don't think we do that in a lot of places, and I'd rather not add > that in to fix this problem at this point in the release cycle, as > we've no idea what it might break/regress. What is "that"? The switch to irq-protected KM_IRQ0? That won't break anything. > Its easier to just disable the hangcheck copy and try again for 2.6.35 > with a workqueue or slow work.
