[PATCH] dmatest: do not allow to interrupt ongoing tests

Will Deacon will.deacon at arm.com
Fri Jun 7 04:44:58 EDT 2013


On Fri, Jun 07, 2013 at 01:45:25AM +0100, Vinod Koul wrote:
> On Wed, Jun 05, 2013 at 06:10:01PM +0100, Will Deacon wrote:
> > On Mon, May 27, 2013 at 08:11:07AM +0100, Andy Shevchenko wrote:
> > > On Fri, 2013-05-24 at 23:55 +0200, Guennadi Liakhovetski wrote: 
> > > > On Thu, 23 May 2013, Andy Shevchenko wrote:
> > > > 
> > > > > When user interrupts ongoing transfers the dmatest may end up with console
> > > > > lockup, oops, or data mismatch. This patch prevents user to abort any ongoing
> > > > > test.
> > > > 
> > > > Personally I would be against such a change. What about interrupting the 
> > > > test with rmmod?
> > > >  Is it still possible after this your patch or not? If not 
> > > > - this doesn't seem like a good idea to me. Why don't we just fix those 
> > > > bugs, that you're describing?
> > > 
> > > The behaviour of the module is returned to the same page by this patch
> > > as it was before (w/o debugfs).
> > > 
> > > The user can interrupt tests by rmmod, but it will take time up to
> > > timeout.
> > > 
> > > I appreciate if you can do a deeper analysis of what happened in
> > > case Will reported.
> > 
> > Did this query hold up the application of this patch? I'd really like to see
> > *something* in 3.10, otherwise dmatest will be broken.
> I though we had some data corruption on subsequent tests, or was that different
> one?

We ended up with something that worked in the end:

  https://lkml.org/lkml/2013/5/23/196

Will



More information about the linux-arm-kernel mailing list