Re: [PATCH 1/2] ASoC: cs4270: introduce CS4270_I2C_INCR

[Daniel Mack <daniel@caiaq.de>]

On Tue, May 05, 2009 at 07:09:01PM +0100, Mark Brown wrote:
> > I planned to use that macro a second time for the register write-back at
> > resume time. But unfortunately, the i2c stack is so badly broken that it
> > does not allow the write of a whole data block without having the length
> > itself as part of the message (which is wrong for the codec).
> 
> Could you clarify what you mean when you say that the I2C block requires
> the length be part of the message?  The I2C stack needs to know the
> amount of data to write but I'm not aware of any restrictions it places
> on the content of the data.

All high-level i2c functions to write block data, namely
i2c_smbus_write_block_data() and i2c_smbus_write_i2c_block_data(), put
the length of the data being sent inside the block sent on the wire. I
couldn't believe it myself, but even my hardware I2C analyzer clearly
shows that. The API seems to assume that communication to I2C devices
always wants data to be sent with a leading command, followed by the
number of data bytes attached and then the data itself. Correct me if
I'm wrong on that.

Daniel