On Wednesday 21 July 2004 01:25, Johannes Stezenbach wrote:
Kenneth Aafløy wrote:
On Tuesday 20 July 2004 21:08, Holger Waechtler wrote:
wouldn't it be better to replace the from-scratch list-and-device handling in dvb_i2c.[hc] by the driver/bus infrastructure in 2.6 that provides the same functionality
Holger, what exactly do you mean here?
I think he's talking about the general device/bus interface <linux/device.h>,
right.
which needs a fair amount of support code in order to work, as far as I know.
I have to admit that I didn't felt the urge to use this yet, but at a first glance over the API everything looks good, pretty cleanly implemented and provides all the power management and hotplug features the current dvb_i2c and the kernel i2c API are still missing.
If implemented in a wrapper API pretty similiar to dvb_i2c.[ch] then the driver even don't need to know anything about the device/bus API.
and to rename every dvb_i2c occurence by dvb_uC in order to mirror the additional flexibility of this code?
I'm not sure anymore..does anyone else have an oppinion on this before I do the rest of the conversions?
The goal was to get sysfs support for firmware loading. And people on lkml also perceived dvb_i2c as a duplication of code from the i2c driver layer, and wanted it to be removed.
Johannes, thank you for recapping that, as I've lost track in the mist :)
Theoretically, the kernel I2C layer provides exactly the same behavior as the old dvb_i2c layer, with added features. Who says we can't mimic/emulate another bus type on I2C as long as the drivers are aware of the protocol in use on the adapter.
?!?
you are kidding, are you?
Holger
