[PATCH v6 1/2] ASoC: fsl: Add S/PDIF CPU DAI driver

[Bhushan Bharat-R65777]

> -----Original Message-----
> From: Chen Guangyu-B42378
> Sent: Monday, August 19, 2013 11:55 AM
> To: Bhushan Bharat-R65777
> Cc: broonie at kernel.org; lars at metafoo.de; p.zabel at pengutronix.de;
> s.hauer at pengutronix.de; mark.rutland at arm.com; devicetree at vger.kernel.org; alsa-
> devel at alsa-project.org; swarren at wwwdotorg.org; festevam at gmail.com;
> timur at tabi.org; rob.herring at calxeda.com; tomasz.figa at gmail.com;
> shawn.guo at linaro.org; linuxppc-dev at lists.ozlabs.org
> Subject: Re: [PATCH v6 1/2] ASoC: fsl: Add S/PDIF CPU DAI driver
> 
> Hi Bhushan,
> 
>    I'll revise some as you suggest. Just a few replies here.
> 
> On Mon, Aug 19, 2013 at 12:38:11PM +0800, Bhushan Bharat-R65777 wrote:
> > > We here suppose the reset bit would be cleared -- "The software
> > > reset will last
> > > 8 cycles." from RM, so if this happened to be a failure, the whole
> > > IP module won't be normally working as well.
> >
> > Also add a comment describing this against why cycle = 1000 is selected.
> 
> If it is done in 8 cycles, 1000-cycle will be surely a safe value for it.
> As long as it finished in 8 cycles, it would quit anyway. Why against?

I am not against, I am saying why it was not 200 or 50 or 20 etc. I am saying that write a comment saying this much is sufficient as per specification and so keep 1000/etc as preservative.

-Bharat

> 
> 
> > > > > +static bool fsl_spdif_volatile_reg(struct device *dev, unsigned
> > > > > +int reg) {
> > > > > +	/* Sync all registers after reset */
> > > >
> > > > Where us sync :) ?
> > >
> > > The "return true" would do that. For volatile registers, if no "return true"
> > > here, the whole regmap would use the value in cache, while for some
> > > bits we need to trace its true value from the physical registers not from
> cache.
> >
> > Where will be device registers cached? Do not we program them to be non-
> cacheable in core?
> 
> regmap has a regcache for all the mapped registers. Set the regsiters as
> volatile will allow the driver to sync the regcache with physical memory each
> time when using regmap_read/write/update_bits().
> 
> But I think I can try to use the regcache_bypass instead.
> 
> 
> Thank you,
> Nicolin Chen