questions regarding drivers/staging/iio/accel/sca3000_core.c

questions regarding drivers/staging/iio/accel/sca3000_core.c

[Himangi Saraogi]

[-- Attachment #1: Type: text/plain, Size: 596 bytes --]

Hi,

I was looking at the possibility of using a managed interface for
iio_device_register in sca3000_probe. But this will lead to the
iio_device_unregisterfunction being called after sca3000_unconfigure_ring
in the remove function. I have a few queries:

1. Is it safe to move the unregister function over the
sca3000_unconfigure_ring?

2. Is it correct that on failure the probe function does not call
sca3000_unconfigure_ring?

3. If it is fine moving unregister after sca3000_unconfigure_ring, I
propose to add a devm counterpart of iio_buffer_unregister. Is it a good
idea?

Thanks,
Himangi

[-- Attachment #2: Type: text/html, Size: 696 bytes --]

Re: questions regarding drivers/staging/iio/accel/sca3000_core.c

[Jonathan Cameron]

On 21/07/14 09:24, Himangi Saraogi wrote:
> Hi,
>
> I was looking at the possibility of using a managed interface for
> iio_device_register in sca3000_probe. But this will lead to the
> iio_device_unregister function being called after
> sca3000_unconfigure_ring in the remove function. I have a few
> queries:
>
> 1. Is it safe to move the unregister function over the
> sca3000_unconfigure_ring?
No, in fact it should be before various other things
(though it isn't currently).  The device unregister
removes the userspace interfaces.  Hence it should always
be called before anything else happens in a remove function.
>
> 2. Is it correct that on failure the probe function does not call
> sca3000_unconfigure_ring?
Nope. That's a bug.  Actually the whole probe function ordering is
suspicious. The device register should probably be a lot later
(i.e. last). There might be some esoteric reason that isn't the case
but I honestly can't remember.  Whilst I have one of these parts
the wire bodge needs repairing and I haven't fired the board it plugs
into up for a good while.  If it wasn't such and interesting part I'd
be tempted to suggest dropping the driver entirely (and find out if
anyone else actually has one in the process!)
>
> 3. If it is fine moving unregister after sca3000_unconfigure_ring, I
> propose to add a devm counterpart of iio_buffer_unregister. Is it a
> good idea?
I'm not sure there enough users to bother and some of those should
probably use the triggered_buffer helper functions.
>
> Thanks, Himangi