staging:iio:triggers introduce iio_trigger_ops to take const bits out of iio_trig_structure.

Hello Jonathan Cameron,

The patch d29f73db7910: "staging:iio:triggers introduce iio_trigger_ops to take const bits out of iio_trig_structure." from Aug 12, 2011, leads to the following static checker warning:

	drivers/iio/industrialio-trigger.c:376 iio_trigger_write_current()
	error: we previously assumed 'indio_dev->trig->ops' could be null (see line 367)

drivers/iio/industrialio-trigger.c
360

   361          if (trig && trig->ops && trig->ops->validate_device) {
                            ^^^^^^^^^

New check.

362 ret = trig->ops->validate_device(trig, indio_dev);
363 if (ret)
364 return ret;
365 }
366
367 indio_dev->trig = trig;
368
369 if (oldtrig) {
370 if (indio_dev->modes & INDIO_EVENT_TRIGGERED)
371 iio_trigger_detach_poll_func(oldtrig,
372 indio_dev->pollfunc_event);
373 iio_trigger_put(oldtrig);
374 }
375 if (indio_dev->trig) {

   376                  iio_trigger_get(indio_dev->trig);
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Dereferenced inside the function without checking.

377 if (indio_dev->modes & INDIO_EVENT_TRIGGERED)
378 iio_trigger_attach_poll_func(indio_dev->trig,
379 indio_dev->pollfunc_event);
380 }

regards,
dan carpenter

Good point - trig->ops is not optional at all (exactly as it has the module handle). Not exactly a new issue but will good to clean it up at some point.

Funnily enough this exact question got raised recently and I foolishly though it was just a documentation thing, whereas clearly quite a bit of the code in here makes it 'look' optional when it never was.

We probably should sanity check and fault out in any attempt to register a trigger without an ops structure though.

I'll put a patch together shortly to do this.

Jonathan