Re: [alsa-devel] [PATCH 6/6] [ALSA] portman2x4 - use new parport device model

[Sudip Mukherjee <sudipm.mukherjee@gmail.com>]

On Thu, Jan 07, 2016 at 11:50:15AM +0100, Takashi Iwai wrote:
> On Thu, 07 Jan 2016 11:44:34 +0100,
> Sudip Mukherjee wrote:
> > 
> > On Thu, Jan 07, 2016 at 11:26:44AM +0100, Takashi Iwai wrote:
> > > On Thu, 07 Jan 2016 08:15:51 +0100,
> > > Sudip Mukherjee wrote:
> > > > 
> > > > Modify portman driver to use the new parallel port device model.
> > > > 
> > > > Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> > > 
> > > Did you actually test this?
> > 
> > No. :(
> > I donot have the hardware. But since the only change is in the way it
> > registers with the parport so it should not break.
> > I was preparing v2 for this and the other one. I missed seeing some more
> > points.
> > > 
> > > Also about the changes:
> > > 
> > > > ---
> > > >  sound/drivers/portman2x4.c | 24 ++++++++++++++----------
> > > >  1 file changed, 14 insertions(+), 10 deletions(-)
> > > > 
> > > > diff --git a/sound/drivers/portman2x4.c b/sound/drivers/portman2x4.c
> > > > index 5fcde7d..88b25ca 100644
> > > > --- a/sound/drivers/portman2x4.c
> > > > +++ b/sound/drivers/portman2x4.c
> > > > @@ -704,9 +704,10 @@ static void snd_portman_detach(struct parport *p)
> > > >  }
> > > >  
> > > >  static struct parport_driver portman_parport_driver = {
> > > > -	.name   = "portman2x4",
> > > > -	.attach = snd_portman_attach,
> > > > -	.detach = snd_portman_detach
> > > > +	.name		= "portman2x4",
> > > > +	.match_port	= snd_portman_attach,
> > > > +	.detach		= snd_portman_detach,
> > > > +	.devmodel	= true,
> > > >  };
> > > >  
> > > >  /*********************************************************************
> > > > @@ -734,6 +735,7 @@ static int snd_portman_probe(struct platform_device *pdev)
> > > >  	struct snd_card *card = NULL;
> > > >  	struct portman *pm = NULL;
> > > >  	int err;
> > > > +	struct pardev_cb portman_cb;
> > > >  
> > > >  	p = platform_get_drvdata(pdev);
> > > >  	platform_set_drvdata(pdev, NULL);
> > > > @@ -758,13 +760,15 @@ static int snd_portman_probe(struct platform_device *pdev)
> > > >  	sprintf(card->longname,  "%s at 0x%lx, irq %i",
> > > >  		card->shortname, p->base, p->irq);
> > > >  
> > > > -	pardev = parport_register_device(p,                     /* port */
> > > > -					 DRIVER_NAME,           /* name */
> > > > -					 NULL,                  /* preempt */
> > > > -					 NULL,                  /* wakeup */
> > > > -					 snd_portman_interrupt, /* ISR */
> > > > -					 PARPORT_DEV_EXCL,      /* flags */
> > > > -					 (void *)card);         /* private */
> > > > +	memset(&portman_cb, 0, sizeof(portman_cb));
> > > > +	portman_cb.private = card;                      /* private */
> > > > +	portman_cb.irq_func = snd_portman_interrupt;    /* ISR */
> > > > +	portman_cb.flags = PARPORT_DEV_EXCL;            /* flags */
> > > 
> > > You can put them initializers except for private.  Then the explicit
> > > memset can be omitted.
> > > 
> > > > +
> > > > +	pardev = parport_register_dev_model(p,		   /* port */
> > > > +					    DRIVER_NAME,   /* name */
> > > > +					    &portman_cb,   /* callbacks */
> > > > +					    device_count); /* device number */
> > > 
> > > Does device_count really work similarly for
> > > parport_register_dev_model()?  I supposed the argument being the
> > > device id number while you're passing the number of devices to
> > > create.
> > 
> > This device_count is actually used for the device name in
> > /sys/bus/parport/devices. Something like DRIVER_NAME.device_count.
> 
> Well, but device_count is incremented in snd_portman_attach().  The
> management of device_count should be moved around the caller side, if
> we use this as the id (and use the assigned id instead of device_count
> in snd_portman_attach()).

But, snd_portman_attach() finally decides if the probe/attach was a
success or not. And it will save the device in
platform_devices[device_count] and then it will increment device_count
to prepare it for the next device. Ofcourse, we can do it in
snd_portman_probe() but isn't snd_portman_attach() the caller here?

If you want I can move the count to snd_portman_probe() but since I do
not have the hardware I tried to have the minimum possible change.

regards
sudip