[alsa-devel] [PATCH] ASoC: wm_adsp: Add basic debugfs entries

[Richard Fitzgerald]

On Mon, Jun 08, 2015 at 06:40:41PM +0100, Mark Brown wrote:
> On Mon, Jun 08, 2015 at 03:37:02PM +0100, Richard Fitzgerald wrote:
> 
> > +++ b/sound/soc/codecs/wm5102.c
> > @@ -1875,6 +1875,8 @@ static int wm5102_codec_probe(struct snd_soc_codec *codec)
> >  	struct wm5102_priv *priv = snd_soc_codec_get_drvdata(codec);
> >  	int ret;
> >  
> > +	wm_adsp_init_debugfs(&priv->core.adsp[0], codec);
> > +
> 
> Why are we adding this init to every individual CODEC rather than doing
> it when we initialize the DSP (which there are calls for already)?
> 

Because we call the existing wm_adsp_init() early in probe and at that point we
haven't registered the codec yet.

> > +#ifdef CONFIG_DEBUG_FS
> > +static void wm_adsp_debugfs_save_wmfwname(struct wm_adsp *dsp, const char *s);
> > +static void wm_adsp_debugfs_save_binname(struct wm_adsp *dsp, const char *s);
> > +static void wm_adsp_debugfs_clear(struct wm_adsp *dsp);
> > +#else
> > +static inline void wm_adsp_debugfs_save_wmfwname(struct wm_adsp *dsp,
> > +						 const char *s)
> > +{
> > +}
> > +
> > +static inline void wm_adsp_debugfs_save_binname(struct wm_adsp *dsp,
> > +						const char *s)
> > +{
> > +}
> > +
> > +static inline void wm_adsp_debugfs_clear(struct wm_adsp *dsp)
> > +{
> > +}
> > +#endif
> > +
> 
> Why not just put the functions here rather than prototypes?
> 

It was just personal preference, I like to have the important code higher up in
source files and keep the clutter of debug code near the end where it's not in
the way but I can turn it around.

> > +static ssize_t wm_adsp_debugfs_string_read(struct wm_adsp *dsp,
> > +					  char __user *user_buf,
> > +					  size_t count, loff_t *ppos,
> > +					  const char *string)
> > +{
> > +	char *temp;
> > +	int len;
> > +	ssize_t ret;
> > +
> > +	if (!string || !dsp->running)
> > +		return 0;
> 
> Does debugfs ensure that the right thing happens and this gets treated
> as EOF rather than a "zero length read, please retry" (which something
> might decide to busy wait trying)?  I'd have expected either an error or
> substituting in an empty/informative string here.
> 
> > +	temp = kmalloc(PAGE_SIZE, GFP_KERNEL);
> > +	if (!temp)
> > +		return -ENOMEM;
> > +
> > +	len = snprintf(temp, PAGE_SIZE, "%s\n", string);
> 
> Given that we already have the string I don't understand why we're
> allocating the temporary buffer - if it's just the length we're looking
> for then strlen() should be enough?
> 
> > +} wm_adsp_debugfs_fops[] = {
> > +	{
> > +		.name = "wmfw_file",
> 
> > +		.name = "bin_file",
> 
> Bikeshedding but _name not _file perhaps?  It's not going to give you a
> copy of the firmware/coefficients.