[linux-dvb] Re: CX88 i2c issue w/ DVB tuners

[Michael Hunold <hunold-ml@web.de>]

Hi,

On 12.09.2004 16:08, Andrew de Quincey wrote:
> > > I wouldn't care when kernel i2c would be crappy code inside (which I
> > > thing it isn't)
> > :) (it is - really: I've rarely seen anything broken-by-concept like
> >
> > this. And I've rarely seen anybody fighting like a pitbull for such a
> > stupid concept.)
> You seem to have the wrong end of the stick here. I'm only fighting to FIX 
> i2c. I'm saying if we use i2c_transfer() as you suggest, then we have to 
> register an i2c_adapter. If we register an i2c_adapter, it will be autoprobed 
> as the *braindead* i2c stuff doesn't let you turn autoprobing OFF. If it is 
> autoprobed, the _other_ i2c devices (e.g. the tda9887) in the system can 
> stomp on us. The alternative is to stop using kernel i2c. 
I think it's easy to stop autoprobing by introducing a new .flags field 
to struct i2c_adapter, for example #define I2C_ADAP_NO_AUTOPROBE.

It should be possible for an i2c_adapter to tell which i2c_driver it is 
expecting, ie. it should specify an array of driver ids like 
I2C_DRIVERID_DVBFE_TDA1004X. This array can even be a member of struct 
i2c_adapter itself.

Basically i2c_add_driver() and i2c_add_adapter() hold the autoprobing 
logic. We can check for the new flag and prevent autoprobing for all 
devices, unless the devices that are present in the whitelist.

> Since we can see this problem coming, why not fix it now, instead of waiting 
> for it to bite us in the arse? :)
IMO the fix is in the i2c subsystem. Sure, it's more hard to get changes 
done in there.

CU
Michael.