Re: [PATCH] drm/modes: remove unused variables

[Daniel Vetter]

On Thu, Dec 05, 2019 at 10:23:51AM +0000, Benjamin GAIGNARD wrote:
> + Zhao Yakui
> 
> On 12/5/19 10:55 AM, Jani Nikula wrote:
> > On Wed, 04 Dec 2019, Benjamin GAIGNARD <benjamin.gaignard@xxxxxx> wrote:
> >> On 12/4/19 10:35 AM, Thomas Zimmermann wrote:
> >>> Hi
> >>>
> >>> Am 19.11.19 um 14:47 schrieb Benjamin Gaignard:
> >>>> When compiling with W=1 few warnings about unused variables show up.
> >>>> This patch removes all the involved variables.
> >>>>
> >>>> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@xxxxxx>
> >>>> ---
> >>>>    drivers/gpu/drm/drm_modes.c | 22 +++-------------------
> >>>>    1 file changed, 3 insertions(+), 19 deletions(-)
> >>>>
> >>>> diff --git a/drivers/gpu/drm/drm_modes.c b/drivers/gpu/drm/drm_modes.c
> >>>> index 88232698d7a0..aca901aff042 100644
> >>>> --- a/drivers/gpu/drm/drm_modes.c
> >>>> +++ b/drivers/gpu/drm/drm_modes.c
> >>>> @@ -233,7 +233,7 @@ struct drm_display_mode *drm_cvt_mode(struct drm_device *dev, int hdisplay,
> >>>>    		/* 3) Nominal HSync width (% of line period) - default 8 */
> >>>>    #define CVT_HSYNC_PERCENTAGE	8
> >>>>    		unsigned int hblank_percentage;
> >>>> -		int vsyncandback_porch, vback_porch, hblank;
> >>>> +		int vsyncandback_porch, hblank;
> >>>>    
> >>>>    		/* estimated the horizontal period */
> >>>>    		tmp1 = HV_FACTOR * 1000000  -
> >>>> @@ -249,7 +249,6 @@ struct drm_display_mode *drm_cvt_mode(struct drm_device *dev, int hdisplay,
> >>>>    		else
> >>>>    			vsyncandback_porch = tmp1;
> >>>>    		/* 10. Find number of lines in back porch */
> >>>> -		vback_porch = vsyncandback_porch - vsync;
> >>>>    		drm_mode->vtotal = vdisplay_rnd + 2 * vmargin +
> >>>>    				vsyncandback_porch + CVT_MIN_V_PORCH;
> >>>>    		/* 5) Definition of Horizontal blanking time limitation */
> >>>> @@ -386,9 +385,8 @@ drm_gtf_mode_complex(struct drm_device *dev, int hdisplay, int vdisplay,
> >>>>    	int top_margin, bottom_margin;
> >>>>    	int interlace;
> >>>>    	unsigned int hfreq_est;
> >>>> -	int vsync_plus_bp, vback_porch;
> >>>> -	unsigned int vtotal_lines, vfieldrate_est, hperiod;
> >>>> -	unsigned int vfield_rate, vframe_rate;
> >>>> +	int vsync_plus_bp;
> >>>> +	unsigned int vtotal_lines;
> >>>>    	int left_margin, right_margin;
> >>>>    	unsigned int total_active_pixels, ideal_duty_cycle;
> >>>>    	unsigned int hblank, total_pixels, pixel_freq;
> >>>> @@ -451,23 +449,9 @@ drm_gtf_mode_complex(struct drm_device *dev, int hdisplay, int vdisplay,
> >>>>    	/* [V SYNC+BP] = RINT(([MIN VSYNC+BP] * hfreq_est / 1000000)) */
> >>>>    	vsync_plus_bp = MIN_VSYNC_PLUS_BP * hfreq_est / 1000;
> >>>>    	vsync_plus_bp = (vsync_plus_bp + 500) / 1000;
> >>>> -	/*  9. Find the number of lines in V back porch alone: */
> >>>> -	vback_porch = vsync_plus_bp - V_SYNC_RQD;
> >>>>    	/*  10. Find the total number of lines in Vertical field period: */
> >>>>    	vtotal_lines = vdisplay_rnd + top_margin + bottom_margin +
> >>>>    			vsync_plus_bp + GTF_MIN_V_PORCH;
> >>>> -	/*  11. Estimate the Vertical field frequency: */
> >>>> -	vfieldrate_est = hfreq_est / vtotal_lines;
> >>>> -	/*  12. Find the actual horizontal period: */
> >>>> -	hperiod = 1000000 / (vfieldrate_rqd * vtotal_lines);
> >>>> -
> >>>> -	/*  13. Find the actual Vertical field frequency: */
> >>>> -	vfield_rate = hfreq_est / vtotal_lines;
> >>>> -	/*  14. Find the Vertical frame frequency: */
> >>>> -	if (interlaced)
> >>>> -		vframe_rate = vfield_rate / 2;
> >>>> -	else
> >>>> -		vframe_rate = vfield_rate;
> >>> The amount of unused code is quite large, which makes me wonder if
> >>> there's something missing below where these variables are supposed to be
> >>> used.
> >>>
> >>> If these variables can be removed, comments should mention that steps 9
> >>> and 11 to 14 are being left out. After all, the function is fairly
> >>> explicit about implementing the GTF algorithm step by step.
> >>>
> >>> Best regards
> >>> Thomas
> >> If the goal is to keep all the steps then I could prefix all problematic
> >> variables with __maybe_unused macro.
> > The effect is the same; it hides a potential bug that should be analyzed
> > and fixed. If you have the time, please look at the code and figure out
> > what it's supposed to do, and why isn't it using the information. Look
> > at git blame and log, was it always so, or did something change?
> >
> > The warnings are about potential bugs. The objective or end goal is to
> > fix the bugs, not to silence the warnings.
> This code haven't change since it has been added by commit:
> 26bbdadad356e ("drm/mode: add the GTF algorithm in kernel space")
> The variables that I'm removing are not used anywhere else.
> The algorithm is copy from xserver/hw/xfree86/modes/xf86gtf.c where
> vframe_rate and v_back_porch are used with (void) calls:
> (void) v_back_porch;
> (void) v_frame_rate;
> It is another way avoid the warnings.
> Note that if you start removing v_frame_rate then vfield_rate becomes 
> unused, etc...

I'd say we should do the same thing then as xf86 and just mark them up as
unused. Might be good to double-check with the spec at least whether
they're really not unused, or whether we have some issue. But given that
it seems to have worked for ages, I suspect it's all good.

Removing the code otoh feels a bit icky.
-Daniel


> 
> Benjamin
> 
> >
> > BR,
> > Jani.
> >
> >
> >> Benjamin
> >>
> >>>>    	/*  15. Find number of pixels in left margin: */
> >>>>    	if (margins)
> >>>>    		left_margin = (hdisplay_rnd * GTF_MARGIN_PERCENTAGE + 500) /
> >>>>
> >> _______________________________________________
> >> dri-devel mailing list
> >> dri-devel@xxxxxxxxxxxxxxxxxxxxx
> >> https://lists.freedesktop.org/mailman/listinfo/dri-devel

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch