Re: [PATCH v5 2/4] ASoC: Intel - add Skylake HDA audio driver

[Vinod Koul <vinod.koul@intel.com>]

On Mon, Jun 15, 2015 at 06:46:33PM +0200, Takashi Iwai wrote:
> At Mon, 15 Jun 2015 17:42:02 +0100,
> Mark Brown wrote:
> > 
> > On Mon, Jun 15, 2015 at 06:35:16PM +0200, Takashi Iwai wrote:
> > > Mark Brown wrote:
> > > > On Thu, Jun 11, 2015 at 10:03:56PM +0530, Vinod Koul wrote:
> > 
> > > > > +	for (i = 0; i < num_stream; i++) {
> > > > > +		struct hdac_ext_stream *stream =
> > > > > +				kzalloc(sizeof(*stream), GFP_KERNEL);
> > 
> > > > Still not sure why these are Sky Lake specific?
> > 
> > > Currently the allocation and the free of each HDA(-ext) stream are
> > > left to each controller driver.  (See the stream object is embedded.)
> > 
> > It looks awfully like it's dynamically allocated here...
> > 
> > > And, yes, the allocation (especially the assignment of the stream tag)
> > > *is* SKL specific.  SKL has some twists in the interpretation of
> > > HD-audio spec.
> > 
> > I can see the thing calling these functions being driver specific but as
> > far as I can see all these are doing is allocating structs, initialising
> > them with passed in parameters, putting them on a list and calling a
> > core function on them.  It's not the bit taking the decisions, it's just
> > doing mechanical things.
> 
> Actually, this can be done a bit more cleanly -- if there will be no
> more additions to the stream object.  It wasn't clear, so the
> allocation and free are left to the driver, so far.
Stream object shouldnt be modfied, so in that case should I move this to
core and ext/ ?

-- 
~Vinod