On Sat, Apr 26, 2014 at 09:02:06PM +0400, Alexander Shiyan wrote:

Sat, 26 Apr 2014 17:55:19 +0100 от Mark Brown broonie@kernel.org:

...

On Sat, Apr 26, 2014 at 06:49:16PM +0400, Alexander Shiyan wrote:

...

These are cases where we can do configuration but an invalid configuration has been provided and we have spotted an error. Ignoring that should cause runtime problems - I guess it isn't for you though so there may be something else going on?

...

Especially for cases like DAI format we should probably enhance the framework so that these devices can set a variable in their struct saying what they do support but for now we don't do that.

Other solution to this problem is to check ret != EPROBE_DEFER in the generic sound card. All other cases should be treated as good for setup parameters. Is this better?

I wouldn't expect to be getting -EPROBE_DEFER outside of the probe path? Can we take a step back here, please - what's the problem you're seeing?