On Tue, Nov 20, 2012 at 03:14:30PM +0000, Liam Girdwood wrote:
On 20/11/12 03:27, Mark Brown wrote:
With coefficeints we have been moving towards a model where we just expose them as binary controls rather than doing the enumeration thing.
For ABE we have different coefficients tied to each enumerated EQU kcontrol setting and just configure the ABE coefficient depending on the kcontrol value. We load the coefficients and EQU kcontrol at the same time since it's enumerated text strings are coupled to the coefficients.
So, the text strings for coefficients were what drivers used to be using
We need to store on file, either an array of strings or an array of u32 values for the enumerated kcontrols (based on soc_enum). This gives us a fixed size file entry which is easy to read, write and count. But maybe I'm missing something from your question...
The issue is that when we have the u32 array we also need the strings to be able to present the control - I would suggest just having the format always include both the numbers and the strings so we don't need to worry about the special case where the numbers are 0, 1, 2... or matching things up if the idea is that we have two separate records.
- dev_info(sfw->dev, "ASoC: no handler specified for ext widget %s\n",
w->name);- return 0;
Shouldn't this be returning an error? The same is true for a bunch of other functions. Or shouild it be a dev_dbg()?
It should probably have a comment. This is an optional callback to component drivers in case they want to use set the widgets private data at widget init time.
In that case the log message should definitely be dev_dbg().
This one is going to be especially bad for example.
- if (se->dvalues)
kfree(se->dvalues);- else {
Braces on both sides if they're needed on one.
The last three points are personal taste, I don't mind too much so I'll change them ;)
They're all kernel coding style rules (the issue with breaking log messages is a particular bugbear of Linus IIRC, rmk also gets very upset).