The new optimized do_div implementation (now in asm-generic/next) exposes a
glitch in the brownstone audio driver by producing a compile-time warning:
sound/soc/pxa/brownstone.c: In function 'brownstone_wm8994_hw_params':
sound/soc/pxa/brownstone.c:67:85: warning: comparison of distinct pointer types lacks a cast
sound/soc/pxa/brownstone.c:67:10125: warning: right shift count >= width of type [-Wshift-count-overflow]
sound/soc/pxa/brownstone.c:67:10254: warning: passing argument 1 of '__div64_32' from incompatible pointer type [-Wincompatible-pointer-types]
The driver just divides two plain integer values, so it should not
use do_div to start with, but has apparently done so ever since the
code was first merged. This replaces do_div with a simple division
operator.
Signed-off-by: Arnd Bergmann <[email protected]>
---
I think this one is harmless, so we only need it in 4.5 to avoid the warning,
but not backported to earlier kernels.
Found today on ARM allmodconfig
diff --git a/sound/soc/pxa/brownstone.c b/sound/soc/pxa/brownstone.c
index 6147e86e9b0f..416ea646c3b1 100644
--- a/sound/soc/pxa/brownstone.c
+++ b/sound/soc/pxa/brownstone.c
@@ -63,8 +63,7 @@ static int brownstone_wm8994_hw_params(struct snd_pcm_substream *substream,
sysclk = params_rate(params) * 512;
sspa_mclk = params_rate(params) * 64;
}
- sspa_div = freq_out;
- do_div(sspa_div, sspa_mclk);
+ sspa_div = freq_out / sspa_mclk;
snd_soc_dai_set_sysclk(cpu_dai, MMP_SSPA_CLK_AUDIO, freq_out, 0);
snd_soc_dai_set_pll(cpu_dai, MMP_SYSCLK, 0, freq_out, sysclk);
On 11/20/2015 05:32 PM, Arnd Bergmann wrote:
> The new optimized do_div implementation (now in asm-generic/next) exposes a
> glitch in the brownstone audio driver by producing a compile-time warning:
>
> sound/soc/pxa/brownstone.c: In function 'brownstone_wm8994_hw_params':
> sound/soc/pxa/brownstone.c:67:85: warning: comparison of distinct pointer types lacks a cast
> sound/soc/pxa/brownstone.c:67:10125: warning: right shift count >= width of type [-Wshift-count-overflow]
> sound/soc/pxa/brownstone.c:67:10254: warning: passing argument 1 of '__div64_32' from incompatible pointer type [-Wincompatible-pointer-types]
>
> The driver just divides two plain integer values, so it should not
> use do_div to start with, but has apparently done so ever since the
> code was first merged. This replaces do_div with a simple division
> operator.
>
> Signed-off-by: Arnd Bergmann <[email protected]>
> ---
> I think this one is harmless, so we only need it in 4.5 to avoid the warning,
> but not backported to earlier kernels.
>
> Found today on ARM allmodconfig
>
> diff --git a/sound/soc/pxa/brownstone.c b/sound/soc/pxa/brownstone.c
> index 6147e86e9b0f..416ea646c3b1 100644
> --- a/sound/soc/pxa/brownstone.c
> +++ b/sound/soc/pxa/brownstone.c
> @@ -63,8 +63,7 @@ static int brownstone_wm8994_hw_params(struct snd_pcm_substream *substream,
> sysclk = params_rate(params) * 512;
> sspa_mclk = params_rate(params) * 64;
> }
> - sspa_div = freq_out;
> - do_div(sspa_div, sspa_mclk);
> + sspa_div = freq_out / sspa_mclk;
>
> snd_soc_dai_set_sysclk(cpu_dai, MMP_SSPA_CLK_AUDIO, freq_out, 0);
> snd_soc_dai_set_pll(cpu_dai, MMP_SYSCLK, 0, freq_out, sysclk);
>
Thanks Arnd.
Looks sspa_div is not used at all.
So we can remove sspa_div as well as do_div.
What do you think.
Thanks
On Monday 23 November 2015 16:10:49 zhangfei wrote:
> On 11/20/2015 05:32 PM, Arnd Bergmann wrote:
> > diff --git a/sound/soc/pxa/brownstone.c b/sound/soc/pxa/brownstone.c
> > index 6147e86e9b0f..416ea646c3b1 100644
> > --- a/sound/soc/pxa/brownstone.c
> > +++ b/sound/soc/pxa/brownstone.c
> > @@ -63,8 +63,7 @@ static int brownstone_wm8994_hw_params(struct snd_pcm_substream *substream,
> > sysclk = params_rate(params) * 512;
> > sspa_mclk = params_rate(params) * 64;
> > }
> > - sspa_div = freq_out;
> > - do_div(sspa_div, sspa_mclk);
> > + sspa_div = freq_out / sspa_mclk;
> >
> > snd_soc_dai_set_sysclk(cpu_dai, MMP_SSPA_CLK_AUDIO, freq_out, 0);
> > snd_soc_dai_set_pll(cpu_dai, MMP_SYSCLK, 0, freq_out, sysclk);
> >
>
> Thanks Arnd.
>
> Looks sspa_div is not used at all.
> So we can remove sspa_div as well as do_div.
>
> What do you think.
Good idea. I see that Mark has already applied the original patch, and with
my new version, gcc should be smart enough to completely optimize out the
division and the unused variable. This means the effect of the patch
is very small, but if Mark wants it, he can apply the below.
Arnd
8<---
Subject: ASoC: remove unused variable
As pointed out by Zhangfei Gao, the sspa_div variable in
brownstone_wm8994_hw_params() is completely unused, so as a cleanup
following a prior patch, this removes both the variable and the division.
Signed-off-by: Arnd Bergmann <[email protected]>
diff --git a/sound/soc/pxa/brownstone.c b/sound/soc/pxa/brownstone.c
index 416ea646c3b1..ec522e94b0e2 100644
--- a/sound/soc/pxa/brownstone.c
+++ b/sound/soc/pxa/brownstone.c
@@ -52,7 +52,6 @@ static int brownstone_wm8994_hw_params(struct snd_pcm_substream *substream,
struct snd_soc_dai *codec_dai = rtd->codec_dai;
struct snd_soc_dai *cpu_dai = rtd->cpu_dai;
int freq_out, sspa_mclk, sysclk;
- int sspa_div;
if (params_rate(params) > 11025) {
freq_out = params_rate(params) * 512;
@@ -63,7 +62,6 @@ static int brownstone_wm8994_hw_params(struct snd_pcm_substream *substream,
sysclk = params_rate(params) * 512;
sspa_mclk = params_rate(params) * 64;
}
- sspa_div = freq_out / sspa_mclk;
snd_soc_dai_set_sysclk(cpu_dai, MMP_SSPA_CLK_AUDIO, freq_out, 0);
snd_soc_dai_set_pll(cpu_dai, MMP_SYSCLK, 0, freq_out, sysclk);