On Fri, Oct 16, 2009 at 5:32 PM, Mark Brown <broonie@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > On Fri, Oct 16, 2009 at 05:14:31PM +0800, Barry Song wrote: >> On Fri, Oct 16, 2009 at 5:08 PM, Mark Brown > >> > This is the wrong fix - null DAI ops are supposed to be supported for >> > all devices since not all DAIs will be configurable. > >> But the problem is ac97 DAIs will have no path to get the null ops. >> Then many functions will panic. It's very easy to repeat the problem >> in your site, only by deleting ops field in any AC97 codec DAI. > > Sure, I see the problem. The issue is that this could affect any CODEC, > not just AC97 ones, so an AC97-specific fix isn't what's needed. Given > the idiom of the rest of the code the best thing would be to make all > the remaining callers check for ops before dereferencing it. At the beginning, I wanted to do like this. But I saw all other codec DAIs register themselves, then got a null ops. Just ac97 is forgot. Then I patched AC97 codec DAIs. Then you prefer to patch all ops access? > _______________________________________________ Alsa-devel mailing list Alsa-devel@xxxxxxxxxxxxxxxx http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
