linux-kernel - Re: [PATCH 06/39] drm: renesas: shmobile: Add support for Runtime PM

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20230623151109.GL2112@pendragon.ideasonboard.com>
Date:   Fri, 23 Jun 2023 18:11:09 +0300
From:   Laurent Pinchart <laurent.pinchart@...asonboard.com>
To:     Geert Uytterhoeven <geert+renesas@...der.be>
Cc:     Kieran Bingham <kieran.bingham+renesas@...asonboard.com>,
        David Airlie <airlied@...il.com>,
        Daniel Vetter <daniel@...ll.ch>,
        Thomas Zimmermann <tzimmermann@...e.de>,
        Magnus Damm <magnus.damm@...il.com>,
        dri-devel@...ts.freedesktop.org, linux-renesas-soc@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 06/39] drm: renesas: shmobile: Add support for Runtime PM

On Fri, Jun 23, 2023 at 06:07:44PM +0300, Laurent Pinchart wrote:
> Hi Geert,
> 
> Thank you for the patch.
> 
> On Thu, Jun 22, 2023 at 11:21:18AM +0200, Geert Uytterhoeven wrote:
> > The SH-Mobile LCD Controller is part of a PM Domain on all relevant SoCs
> > (clock domain on all, power domain on some).  Hence it may not be
> > sufficient to manage the LCDC module clock explicitly (e.g. if the
> > selected clock source differs from SHMOB_DRM_CLK_BUS).
> > 
> > Fix this by using Runtime PM instead.
> > 
> > Signed-off-by: Geert Uytterhoeven <geert+renesas@...der.be>
> > ---
> >  drivers/gpu/drm/renesas/shmobile/shmob_drm_crtc.c | 11 ++++++++++-
> >  drivers/gpu/drm/renesas/shmobile/shmob_drm_drv.c  |  5 +++++
> >  2 files changed, 15 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/renesas/shmobile/shmob_drm_crtc.c b/drivers/gpu/drm/renesas/shmobile/shmob_drm_crtc.c
> > index fbfd906844da490c..84dbf35025d7be63 100644
> > --- a/drivers/gpu/drm/renesas/shmobile/shmob_drm_crtc.c
> > +++ b/drivers/gpu/drm/renesas/shmobile/shmob_drm_crtc.c
> > @@ -9,6 +9,7 @@
> >  
> >  #include <linux/backlight.h>
> >  #include <linux/clk.h>
> > +#include <linux/pm_runtime.h>
> >  
> >  #include <drm/drm_crtc.h>
> >  #include <drm/drm_crtc_helper.h>
> > @@ -170,10 +171,16 @@ static void shmob_drm_crtc_start(struct shmob_drm_crtc *scrtc)
> >  	if (WARN_ON(format == NULL))
> >  		return;
> >  
> > +	ret = pm_runtime_resume_and_get(sdev->dev);
> > +	if (ret)
> > +		return;
> > +
> >  	/* Enable clocks before accessing the hardware. */
> >  	ret = shmob_drm_clk_on(sdev);
> 
> This would be best located in the runtime PM resume handler. Same for
> disabling clocks in the runtime PM suspend handler.

The driver should then depend on CONFIG_PM. There's no indirect
dependency through CONFIG_DRM as far as I can tell, but there's one
through ARCH_SHMOBILE. This then got me puzzled, as ARCH_SHMOBILE is
defined in arch/sh/Kconfig, and this driver depends on ARM. Am I missing
something ?

> > -	if (ret < 0)
> > +	if (ret < 0) {
> > +		pm_runtime_put(sdev->dev);
> >  		return;
> > +	}
> >  
> >  	/* Reset and enable the LCDC. */
> >  	lcdc_write(sdev, LDCNT2R, lcdc_read(sdev, LDCNT2R) | LDCNT2R_BR);
> > @@ -271,6 +278,8 @@ static void shmob_drm_crtc_stop(struct shmob_drm_crtc *scrtc)
> >  	/* Stop clocks. */
> >  	shmob_drm_clk_off(sdev);
> >  
> > +	pm_runtime_put(sdev->dev);
> > +
> >  	scrtc->started = false;
> >  }
> >  
> > diff --git a/drivers/gpu/drm/renesas/shmobile/shmob_drm_drv.c b/drivers/gpu/drm/renesas/shmobile/shmob_drm_drv.c
> > index 30493ce874192e3e..4f01caa119637032 100644
> > --- a/drivers/gpu/drm/renesas/shmobile/shmob_drm_drv.c
> > +++ b/drivers/gpu/drm/renesas/shmobile/shmob_drm_drv.c
> > @@ -13,6 +13,7 @@
> >  #include <linux/module.h>
> >  #include <linux/platform_device.h>
> >  #include <linux/pm.h>
> > +#include <linux/pm_runtime.h>
> >  #include <linux/slab.h>
> >  
> >  #include <drm/drm_drv.h>
> > @@ -216,6 +217,10 @@ static int shmob_drm_probe(struct platform_device *pdev)
> >  	if (IS_ERR(sdev->mmio))
> >  		return PTR_ERR(sdev->mmio);
> >  
> > +	ret = devm_pm_runtime_enable(&pdev->dev);
> > +	if (ret)
> > +		return ret;
> > +
> 
> I would move this after shmob_drm_setup_clocks(), to ensure that the
> runtime PM suspend and resume handlers will have access to clocks.
> 
> >  	ret = shmob_drm_setup_clocks(sdev, pdata->clk_source);
> >  	if (ret < 0)
> >  		return ret;

-- 
Regards,

Laurent Pinchart