On Mon, Feb 15, 2016 at 11:31:48PM +0100, Takashi Iwai wrote:
> Mark Brown wrote:
> > On Fri, Feb 12, 2016 at 07:46:03AM +0530, Subhransu S. Prusty wrote:

> > > +	vendor_param |= INTEL_EN_ALL_PIN_CVTS;
> > > +	vendor_param = snd_hdac_codec_read(hdac, INTEL_VENDOR_NID, 0,
> > > +				INTEL_SET_VENDOR_VERB, vendor_param);
> > > +	if (vendor_param == -1)
> > > +		return;
> > > +}

> > So to enable the pins we do a read?  That seems...  innovative.  :/

> It's a weird nature of HD-audio verb handling.  While *_write() just
> sends the verb asynchronously, *_read() sends the verb, does sync and
> read-back the return value.  But both read and write may handle the
> same verb.

The above one seems especially odd, we do the read and then essentially
ignore the value (the difference in handling is nonexistant).