Re: [PATCH v9 2/8] drm/print: Fix and add support for NULL as first argument in drm_* macros

From: Laurent Pinchart
Date: Tue Jun 06 2023 - 10:49:30 EST


On Tue, Jun 06, 2023 at 08:04:39PM +0530, Siddh Raman Pant wrote:
> On Tue, 06 Jun 2023 19:35:12 +0530, Laurent Pinchart wrote:
> > Hi Siddh,
> >
> > Thank you for the patch.
>
> Anytime :)
>
> > On Tue, Jun 06, 2023 at 04:15:16PM +0530, Siddh Raman Pant wrote:
> > > Comments say macros DRM_DEBUG_* are deprecated in favor of
> > > drm_dbg_*(NULL, ...), but they have broken support for it,
> > > as the macro will result in `(NULL) ? (NULL)->dev : NULL`.
> >
> > What's the problem there ?
>
> (NULL)->dev is invalid C. It's a macro, so preprocessor substitutes
> that text directly, there is no evaluation. GCC will throw an error
> regarding dereferencing a void* pointer.
>
> > > /* Helper for struct drm_device based logging. */
> > > #define __drm_printk(drm, level, type, fmt, ...) \
> > > - dev_##level##type((drm)->dev, "[drm] " fmt, ##__VA_ARGS__)
> > > +({ \
> > > + struct device *__dev_ = __drm_dev_ptr(drm); \
> > > + if (__dev_) \
> > > + dev_##level##type(__dev_, "[drm] " fmt, ##__VA_ARGS__); \
> > > + else \
> > > + pr_##level##type("[drm] " fmt, ##__VA_ARGS__); \
> >
> > If I recall correctly, dev_*() handle a NULL dev pointer just fine. Do
> > we need to manually fall back to pr_*() ?
>
> I took drm_dev_printk (on line 261 of drm_print.c) as the reference,
> wherein it uses a conditional for determining whether dev_printk or
> printk should be called.
>
> I suppose it is to avoid printing "(NULL device *)", which dev_printk
> does if it gets a NULL device pointer (refer the definition on line
> 4831 of drivers/base/core.c). Though if I'm wrong, kindly let me know.

You're right, it's probably best to avoid the "(NULL device *)".

Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@xxxxxxxxxxxxxxxx>

--
Regards,

Laurent Pinchart