[linux-dvb] Re: mt352, site updated, diff

[Holger Waechtler <holger@qanu.de>]

Johannes Stezenbach wrote:
> Holger Waechtler wrote:
>
> > Wolfram Joost wrote:
> >
> > > -			.flags = I2C_M_NOSTART,
> > > +			.flags = 0,
> > If I understand the MT352 spec right i2c read transfers from the MT352 
> > have to get executed without STOP condition (this scenario is called 
> > RESTART condition in the Philips i2c spec). Even if the I2C_M_NOSTART 
> > name is misleading this flag achieves this requirement in most i2c drivers.
>
> To me it seems that the bit banging driver is about the only one
> which implements I2C_M_NOSTART -- the other i2c adapter drivers
> just ignore it.
most kernel i2c drivers implicitly don't insert STOP conditions between 
connected i2c messages. Unfortunally not all of them implement this bug 
consistently, this already caused headache once and was one of the 
reason that we introduced our own i2c subsystem.

Fixing the kernel drivers would mean to break the analog cards which 
rely on this behaviour -- and I suppose nobody of us has all the analog 
and digital TV cards available out there on his desk to test. Blindly 
changing expected behaviour by blindly fixing 'bugs' or 'features' would 
probably cause inconvenient problems for many users with old cards out 
there and is probably not worth the troubles, thus the tiny dvb-i2c 
subsystem.

The DVB drivers rely on correctly inserted STOP/RESTART conditions since 
almost all known cards use i2c bridges on the demodulators which close 
the connection to the PLL if this is not properly coded -- no idea 
whether analog cards are as sensitive.

Holger