[i2c] [PATCH 8/12] drivers: PMC MSP71xx TWI driver

[Marc St-Jean]

Trilok Soni wrote:
> On 6/28/07, Marc St-Jean <Marc_St-Jean at pmc-sierra.com> wrote:
>  >
>  > Marc St-Jean wrote:
>  > >
>  >
>  > > I have a question related to this, after converting to a platform 
> driver
>  > > our chips driver will no longer attempts to find it's devices. The bus
>  > > driver first loads successfully. When the chip driver loads, it's
>  > > attach_adapter function calls i2c_probe which returns no error. 
> However
>  > > i2c_probe never calls the detect function pointer that's passed to it!
>  > >
>  > > There is no debugging output with all flags on. I've been looking at
>  > > other bus and chip drivers but see nothing obvious. I'm also working
>  > > through the list archive, too bad it's not searchable.
>  > >
>  > > Anyone have an idea if anything new is required once the bus driver
>  > > is a platform driver?
>  > > Must a custom addr_data structure be passed in to i2c_probe?
>  > > Must the .id field of the chip driver be specified?
>  >
>  > Found the issue. Jean you had asked to drop the 'probe' workaround
>  > but that won't work as it appears the i2c core probes with 0 length
>  > reads. Without the work around we return errors and it doesn't
>  > proceed to calling the chip drivers detect function.
>  >
> 
> OMAP to have zero length reads hack, but we maintain that in our OMAP
> GIT [1] tree, until all the chips convert to new I2C style bindings
> and fill i2c_board_info structure. So, if your chips convert to new
> I2C style bindings and register themselves with
> i2c_register_board_info, then I don't think you need zero length read
> hack.
> 
> -- 
> --Trilok Soni
> [1] http://source.mvista.com/git/gitweb.cgi?p=linux-omap-2.6.git;a=log
> 

My mistake, it was actually zero byte write for the probe, not read.
Do the new I2C bindings eliminate the zero byte writes as well?

Marc