On Tue, Mar 27, 2012 at 01:58:47AM +0800, Daniel Kurtz wrote:
On Mon, Mar 26, 2012 at 10:49 PM, Daniel Vetter daniel@ffwll.ch wrote:
On Mon, Mar 26, 2012 at 10:26:42PM +0800, Daniel Kurtz wrote:
Instead of rolling our own custom quirk_xfer function, use the bit_algo pre_xfer and post_xfer functions to setup and teardown bit-banged i2c transactions.
gmbus_xfer uses .force_bit to determine which i2c_algorithm to use, either i2c_bit_algo.master_xfer or its own. So, Similarly, let gmbus_func use .force_bit to determine which i2c functionalities are available, either i2c_bit_algo.functionality, or its own.
Please split this part of the patch into a separate patch. Furthermore I'm not sure what this should buy us, given that we might magically changes our i2c feature set once with gone to fallback mode. Can you please elaborate why we need this?
An i2c adapter's functionality is provided by its algorithm. Since these gmbus adapters can [for now] change their algorithm at runtime, I thought the functionality returned should match the currently selected algorithm at any given moment.
Arguably, the adapter actually sort of provides the union of the two functionalities since if a particular transfer fails using gmbus, it gets retried using bit-banged. But then again, this is a one-shot permanent switch, so perhaps we should return the union of the functionalities if force_bit == 0, and then only the bit-algo functionality after the switch?
In that case I guess we can drop it - current edid reading seems to work and without a good reason I'd like not to play clever tricks because it doesn't seem to be worth it. -Daniel