Date:   Fri, 22 Sep 2023 10:01:36 +0800
From:   Stephen Zhang <starzhangzsd@...il.com>
To:     Lee Jones <lee@...nel.org>
Cc:     linux-kernel@...r.kernel.org, zhangshida@...inos.cn,
        k2ci <kernel-bot@...inos.cn>
Subject: Re: [PATCH] mfd: cs42l43: fix defined but not used warnings

Lee Jones <lee@...nel.org> 于2023年9月21日周四 18:37写道：
>
> On Thu, 21 Sep 2023, Stephen Zhang wrote:
>
> > Lee Jones <lee@...nel.org> 于2023年9月20日周三 17:52写道：
> > >
> > > On Tue, 05 Sep 2023, zhangshida wrote:
> > >
> > > > From: Shida Zhang <zhangshida@...inos.cn>
> > > >
> > > > Warnings were generated during compiling for functions like
> > > > cs42l43_*_{resume,suspend}:
> > > >
> > > > ../drivers/mfd/cs42l43.c:1138:12: error: ‘cs42l43_runtime_resume’ defined but not used [-Werror=unused-function]
> > > >  1138 | static int cs42l43_runtime_resume(struct device *dev)
> > > >       |            ^~~~~~~~~~~~~~~~~~~~~~
> > > > ../drivers/mfd/cs42l43.c:1124:12: error: ‘cs42l43_runtime_suspend’ defined but not used [-Werror=unused-function]
> > > >  1124 | static int cs42l43_runtime_suspend(struct device *dev)
> > > >       |            ^~~~~~~~~~~~~~~~~~~~~~~
> > > > ../drivers/mfd/cs42l43.c:1106:12: error: ‘cs42l43_resume’ defined but not used [-Werror=unused-function]
> > > >  1106 | static int cs42l43_resume(struct device *dev)
> > > >       |            ^~~~~~~~~~~~~~
> > > > ../drivers/mfd/cs42l43.c:1076:12: error: ‘cs42l43_suspend’ defined but not used [-Werror=unused-function]
> > > >  1076 | static int cs42l43_suspend(struct device *dev)
> > > >
> > > > Fix it by guarding it with CONFIG_PM/CONFIG_PM_SLEEP.
> > > >
> > > > Reported-by: k2ci <kernel-bot@...inos.cn>
> > > > Signed-off-by: Shida Zhang <zhangshida@...inos.cn>
> > > > ---
> > > >  drivers/mfd/cs42l43.c | 4 ++++
> > > >  1 file changed, 4 insertions(+)
> > > >
> > > > diff --git a/drivers/mfd/cs42l43.c b/drivers/mfd/cs42l43.c
> > > > index 37b23e9bae82..e589a61c118d 100644
> > > > --- a/drivers/mfd/cs42l43.c
> > > > +++ b/drivers/mfd/cs42l43.c
> > > > @@ -1073,6 +1073,7 @@ void cs42l43_dev_remove(struct cs42l43 *cs42l43)
> > > >  }
> > > >  EXPORT_SYMBOL_NS_GPL(cs42l43_dev_remove, MFD_CS42L43);
> > > >
> > > > +#ifdef CONFIG_PM_SLEEP
> > > >  static int cs42l43_suspend(struct device *dev)
> > > >  {
> > > >       struct cs42l43 *cs42l43 = dev_get_drvdata(dev);
> > > > @@ -1120,7 +1121,9 @@ static int cs42l43_resume(struct device *dev)
> > > >
> > > >       return 0;
> > > >  }
> > > > +#endif
> > > >
> > > > +#ifdef CONFIG_PM
> > > >  static int cs42l43_runtime_suspend(struct device *dev)
> > > >  {
> > > >       struct cs42l43 *cs42l43 = dev_get_drvdata(dev);
> > > > @@ -1176,6 +1179,7 @@ static int cs42l43_runtime_resume(struct device *dev)
> > > >
> > > >       return ret;
> > > >  }
> > > > +#endif
> > > >
> > > >  EXPORT_NS_GPL_DEV_PM_OPS(cs42l43_pm_ops, MFD_CS42L43) = {
> > > >       SET_SYSTEM_SLEEP_PM_OPS(cs42l43_suspend, cs42l43_resume)
> > >
> > > I see a bunch of drivers using PM helpers and not many of them are
> > > are being guarded by ugly #ifery.  Please find out what they're doing to
> > > solve the same issue and replicate that instead.
> > >
> > > Here's a really big hint:
> > >
> > >   `git log --oneline 02313a90095fb`
> > >
> >
> > Thanks, I've learned something from the hint.
> >
> > And I also checked the code:
> > -----
> > EXPORT_NS_GPL_DEV_PM_OPS(cs42l43_pm_ops, MFD_CS42L43) = {
> >         SET_SYSTEM_SLEEP_PM_OPS(cs42l43_suspend, cs42l43_resume)
> >         SET_RUNTIME_PM_OPS(cs42l43_runtime_suspend,
> > cs42l43_runtime_resume, NULL)
> > };
> > ----
> > #ifdef CONFIG_PM_SLEEP
> > #define SET_SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn) \
> >         SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn)
> > #else
> > #define SET_SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn)
> > #endif
> > ----
> > #define SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn) \
> >         .suspend = pm_sleep_ptr(suspend_fn), \
> >         .resume = pm_sleep_ptr(resume_fn), \
> >         .freeze = pm_sleep_ptr(suspend_fn), \
> >         .thaw = pm_sleep_ptr(resume_fn), \
> >         .poweroff = pm_sleep_ptr(suspend_fn), \
> >         .restore = pm_sleep_ptr(resume_fn),
> > ----
> > The technique has already been used by the marcos, but it still
> > reports the defined-but-not-used warning.
>
> The MACROS can use #ifery since they are located in header files.
>
> #ifery in C files is to be avoided if at all possible.
>
> > Maybe some compilers still choose to compile these functions in...
> > Anyway, I will just leave it alone since it is really not a big problem...
>
> If you're seeing an error, it should be fixed.

<nod>

>
> Why is this not an issue anywhere else?
>
> Does the same build complain about all the other drivers too?
>

Nope, at least 88pm860x-core.c is good:
----
drivers/mfd/88pm860x-core.c
static int pm860x_suspend(struct device *dev)
static DEFINE_SIMPLE_DEV_PM_OPS(pm860x_pm_ops, pm860x_suspend, pm860x_resume);
 static struct i2c_driver pm860x_driver = {
        .driver = {
                .name   = "88PM860x",
               .pm     = pm_sleep_ptr(&pm860x_pm_ops),
                .of_match_table = pm860x_dt_ids,
        },
        .probe_new      = pm860x_probe,
----
...
  LD [M]  sound/pci/echoaudio/snd-layla20.o
  CC      drivers/mfd/88pm860x-core.o
  CC      drivers/misc/eeprom/max6875.o
  CC      arch/mips/vdso/vdso-image.o
...
----

It is generated by mips-linux-gnu-gcc, which always exhibits some
peculiar behavior.

Since it reports 'defined but not used,' it implies that the macro
successfully removed any usage of 'cs42l43_*.'
However, the compiler still chose not to optimize these functions out.
It's very weird.
Maybe the problem comes from the optimization process in mips compiler?

I will attach the kernel.config and the build.log for your reference.

> --
> Lee Jones [李琼斯]

View attachment "build.log" of type "text/x-log" (399168 bytes)

Download attachment "kernel.conifg" of type "application/octet-stream" (191288 bytes)

Powered by blists - more mailing lists