awalls at radix
Oct 18, 2008, 6:58 AM
Post #3 of 3
(1732 views)
Permalink
|
|
Re: cx18: cx18_probe() calls cx18_init_memory() twice
[In reply to]
|
|
On Sat, 2008-10-18 at 15:48 +0200, Hans Verkuil wrote:
> On Saturday 18 October 2008 15:06:28 Andy Walls wrote:
> > Hans,
> >
> > cx18_probe() calls cx18_init_memory() twice to init the DDR. I
> > suspect I should remove the second instance, since that one is called
> > after cx18_init_scb(). Or do both need to stay for some reason?
> >
> > Regards,
> > Andy
>
> Yes, there is a reason: the memory settings depend on the card type. But
> for Hauppauge cards the card type is derived from the eeprom. That
> requires a working i2c bus. So you cannot set the correct memory
> settings until after init_cx18_i2c() and cx18_process_eeprom().
>
> Now from what I remember some of the functions leading up to
> init_cx18_i2c() use scb (or at least require it to be non-NULL). And
> before you can call cx18_init_scb() you need to setup the memory.
>
> At the time I just put in a cx_init_memory() call so that I could setup
> scb. Of course, this is just a hack. The real solution would be to
> structure the driver so that you can setup the i2c bus without having
> to setup memory/scb.
>
> In any case, the second cx18_init_memory is certainly correct, since
> cx18_process_eeprom() might discover that it is really a different card
> with different timings.
>
> The first init_memory/init_scb pair basically just works around a
> different bug: someone using scb too early. I guess that should be
> cleaned up. Try removing the first init_memory and moving the init_scb
> to after the second init_memory and see what happens. Perhaps it is no
> longer an issue after all the changes done since I wrote that code.
Ah, I didn't get an appreciation for all the subtleties. I thought the
early init may have been linked to early i2c use - but I didn't hunt
down where it might be needed.
I'll leave this one alone for now. I was going to ask Mauro to pull a
few minor cleanups before I'm away from my setup for a few weeks. I
noticed this and thought it may be an easy one to include - I guess not.
Thanks.
Andy
> Regards,
>
> Hans
_______________________________________________
ivtv-devel mailing list
ivtv-devel [at] ivtvdriver
http://ivtvdriver.org/mailman/listinfo/ivtv-devel
|
