Em Thu, 18 Sep 2014 16:01:59 +0300
Antti Palosaari <crope@xxxxxx> escreveu:
>
>
> On 09/18/2014 03:31 PM, Mauro Carvalho Chehab wrote:
> > Em Sun, 7 Sep 2014 04:59:55 +0300
> > Antti Palosaari <crope@xxxxxx> escreveu:
> >
> >> Used tda18212 tuner is implemented as I2C driver. Implement I2C
> >> client to anysee and use it for tda18212.
>
> >> +static int anysee_add_i2c_dev(struct dvb_usb_device *d, char *type, u8 addr,
> >> + void *platform_data)
>
> >> +
> >> +static void anysee_del_i2c_dev(struct dvb_usb_device *d)
>
> >
> > Please, instead of adding a function to insert/remove an I2C driver on every
> > place, put them into a common place.
> >
> > I would actually very much prefer if you could reuse the same code that
> > are already at the media subsystem (see v4l2_i2c_new_subdev_board &
> > friends at drivers/media/v4l2-core/v4l2-common.c), eventually making it
> > more generic.
> >
> > Btw, as we want to use the media controller also for DVB, we'll end
> > by needing to use a call similar to v4l2_device_register_subdev().
> > So, having this code all on just one place will make easier for us to
> > go to this next step.
>
> I am just learning and finding out best practices to use I2C drivers.
> That was one implementation solution and IMHO not so bad even. Sure
> those 2 functions could be replaced some more common at some phase, but
> currently, when there is only few drivers, I don't see need for common
> implementation. Let it happen when best practices are clear.
> And I
> really wonder why there is no such general implementation provided by
> I2C framework?
It used to have on Kernel 2.4 and 2.6 (up to 2.6.1x or 2.6.2x - don't
remember the exact Kernel where this changed). The previous binding model
got deprecated and removed, as different subsystems might need to do
different things. So, it was opted to have one solution per subsystem.
>
> If you look how I have improved that in a long ran; 1st implementation
> is in em28xx driver, 2nd test was dd-bridge driver, then this anysee and
> eventually there is af9035 (which is almost same than this anysee).
So, now we have 3 different implementations (4, if we count the generic
one at v4l-core). That's bad. We should really stick with the original
idea of having just one per subsystem.
So, please write a patch unifying them into just one.
For now, I'll be merging it, but please let's avoid all those code
duplication, as this makes harder to maintain.
>
>
> >> @@ -939,46 +1025,63 @@ static int anysee_tuner_attach(struct dvb_usb_adapter *adap)
> >> * fails attach old simple PLL. */
> >>
> >> /* attach tuner */
> >> - fe = dvb_attach(tda18212_attach, adap->fe[0], &d->i2c_adap,
> >> - &anysee_tda18212_config);
> >> + if (state->has_tda18212) {
> >> + struct tda18212_config tda18212_config =
> >> + anysee_tda18212_config;
> >>
> >> - if (fe && adap->fe[1]) {
> >> - /* attach tuner for 2nd FE */
> >> - fe = dvb_attach(tda18212_attach, adap->fe[1],
> >> - &d->i2c_adap, &anysee_tda18212_config);
> >> - break;
> >> - } else if (fe) {
> >> - break;
> >> - }
> >> -
> >> - /* attach tuner */
> >> - fe = dvb_attach(dvb_pll_attach, adap->fe[0], (0xc0 >> 1),
> >> - &d->i2c_adap, DVB_PLL_SAMSUNG_DTOS403IH102A);
> >> + tda18212_config.fe = adap->fe[0];
> >> + ret = anysee_add_i2c_dev(d, "tda18212", 0x60,
> >> + &tda18212_config);
> >> + if (ret)
> >> + goto err;
> >> +
> >> + /* copy tuner ops for 2nd FE as tuner is shared */
> >> + if (adap->fe[1]) {
> >> + adap->fe[1]->tuner_priv =
> >> + adap->fe[0]->tuner_priv;
> >> + memcpy(&adap->fe[1]->ops.tuner_ops,
> >> + &adap->fe[0]->ops.tuner_ops,
> >> + sizeof(struct dvb_tuner_ops));
> >> + }
> >>
> >> - if (fe && adap->fe[1]) {
> >> - /* attach tuner for 2nd FE */
> >> - fe = dvb_attach(dvb_pll_attach, adap->fe[1],
> >> + return 0;
> >> + } else {
> >> + /* attach tuner */
> >> + fe = dvb_attach(dvb_pll_attach, adap->fe[0],
> >> (0xc0 >> 1), &d->i2c_adap,
> >> DVB_PLL_SAMSUNG_DTOS403IH102A);
> >
> > Please don't use dvb_attach() for those converted modules. The
> > dvb_attach() is a very dirty hack that was created as an alternative
> > to provide an abstraction similar to the one that the I2C core already
> > provides. See how V4L calls the subdev callbacks at
> > include/media/v4l2-subdev.h.
>
> You looked it wrong, it is dvb_pll_attach. tda18212 attach is replaced
> here with I2C driver. It is tda18212 which is converted here to I2C
> driver, whilst dvb-pll leaves old.
Ah, OK.
>
> >
> > One of the big disadvantages of the dvb_attach() is that it allows just
> > _one_ entry point function on a sub-device. This only works for very
> > simple demods that don't provide, for example, hardware filtering.
> >
> >> +
> >> + if (fe && adap->fe[1]) {
> >> + /* attach tuner for 2nd FE */
> >> + fe = dvb_attach(dvb_pll_attach, adap->fe[1],
> >> + (0xc0 >> 1), &d->i2c_adap,
> >> + DVB_PLL_SAMSUNG_DTOS403IH102A);
> >> + }
>
> That patch has nothing wrong as I explained :)
> It could be improved by implementing general I2C sub-driver loading
> functions, but these driver specific implementations are just fine until
> more common solution is developed.
>
> regards
> Antti
>
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
