On Mon, Jan 09, 2012 at 12:11:10PM -0800, Greg KH wrote:
On Mon, Jan 09, 2012 at 07:52:13PM +0000, Mark Brown wrote:
Trying to make any sort of modification to code this fragile is risky, especially during what's supposed to be a stabalization phase (which is what Greg is requesting). It just seems completely irresponsible for something that isn't actually a practical problem.
I find it hard to believe that ignoring the driver model is not a "practical" problem :)
For details as to why this is a problem, please see the kobject.txt file.
Sure, I'm fully aware of the issue. The reason this is so painful to work with is that AC'97 is just generally doing a really bad job of using the driver model.
Please fix this up, as you have seen, people end up cutting-and-pasting bad code.
In my copious free time, but like I say trying to do this for 3.3 (you only posted *after* the merge window opened) is just nuts and I'd rather hope someone who cares about AC'97 systems will come forward and work on it (having one would be a real bonus). If people actually had problems we were fixing that'd be one thing but if they do they're being extremely quiet about it.
If it's code like the rtd devices which is reasonably sensible and stable that's one thing but wading into code we know is fragile for what's essentially a warning fix is completely disproportionate. It's not like we'd actually make a substantial improvement in the overall code quality or robustness, if anyone's using AC'97 as a reference for how to do this stuff they're going to have a bunch of other issues to work through.