cafe_ccic/ov7670 hang on boot

[Erik Garrison]

On Tue, Jun 17, 2008 at 01:08:47PM -0600, Jonathan Corbet wrote:
> Daniel Drake <dsd at laptop.org> wrote:
> 
> > OK, thanks for the input. However there doesn't seem to be a timeout in
> > the case added by commit 6d77444aca29:
> > +	do {
> > +		prepare_to_wait(&cam->smbus_wait, &the_wait,
> > +				TASK_UNINTERRUPTIBLE);
> > +		schedule_timeout(1); /* even 1 jiffy is too long */
> > +		finish_wait(&cam->smbus_wait, &the_wait);
> > +	} while (!cafe_smbus_write_done(cam));
> > 
> > 
> > I'm only guessing, but I have a feeling we're looping infinitely there.
> 
> That, of course, would be easy enough to determine empirically.
> 
> > We're testing an XO with that patch reverted, so far so good but we
> > should leave it for longer before making conclusions.
> 
> But does that get you back to the long startup times that the patch went
> in to fix?  That won't make anybody happy.
> 
> If reverting this patch fixes the problem, then my earlier hypothesis of
> i2c problems is completely wrong.  But it would sure be nice to know
> what is really going on - the code in that patch is correct (though it
> could benefit from a loop counter to keep it from going forever).  I
> wonder if the behavior of the scheduler has changed enough that the
> first cafe_smbus_write_done() call is happening too soon?

Yes, the startup now takes ~3 seconds because of the frequent msleeps().
At the same time we are not meeting the prior failure mode.  After
reverting commit 6d77444aca29, we have run 39 boots without failure.

We will run overnight to verify that this commit can be invoked to
trigger the bug.

Erik