Message-ID: <CAD=FV=XF65otS2S+6sg6qga6Le3xb1f5GC6R6qpf27zx49DQ6w@mail.gmail.com>
Date:   Fri, 15 Sep 2023 06:44:21 -0700
From:   Doug Anderson <dianders@...omium.org>
To:     suijingfeng <suijingfeng@...ngson.cn>
Cc:     dri-devel@...ts.freedesktop.org,
        Maxime Ripard <mripard@...nel.org>, airlied@...il.com,
        airlied@...hat.com, alain.volmat@...s.st.com,
        alexander.deucher@....com, alexandre.belloni@...tlin.com,
        alison.wang@....com, bbrezillon@...nel.org,
        christian.koenig@....com, claudiu.beznea@...rochip.com,
        daniel@...ll.ch, drawat.floss@...il.com, javierm@...hat.com,
        jernej.skrabec@...il.com, jfalempe@...hat.com, jstultz@...gle.com,
        kong.kongxinwei@...ilicon.com, kraxel@...hat.com,
        linus.walleij@...aro.org, linux-arm-kernel@...ts.infradead.org,
        linux-hyperv@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-sunxi@...ts.linux.dev, liviu.dudau@....com,
        nicolas.ferre@...rochip.com, paul.kocialkowski@...tlin.com,
        sam@...nborg.org, samuel@...lland.org,
        spice-devel@...ts.freedesktop.org, stefan@...er.ch,
        sumit.semwal@...aro.org, tiantao6@...ilicon.com,
        tomi.valkeinen@...asonboard.com, tzimmermann@...e.de,
        virtualization@...ts.linux-foundation.org, wens@...e.org,
        xinliang.liu@...aro.org, yongqin.liu@...aro.org, zackr@...are.com
Subject: Re: [RFT PATCH 2/6] drm: Call drm_atomic_helper_shutdown() at
 shutdown time for misc drivers

Hi,

On Fri, Sep 15, 2023 at 2:11 AM suijingfeng <suijingfeng@...ngson.cn> wrote:
>
> Hi,
>
>
> On 2023/9/2 07:39, Douglas Anderson wrote:
> > Based on grepping through the source code these drivers appear to be
> > missing a call to drm_atomic_helper_shutdown() at system shutdown
> > time. Among other things, this means that if a panel is in use that it
> > won't be cleanly powered off at system shutdown time.
> >
> > The fact that we should call drm_atomic_helper_shutdown() in the case
> > of OS shutdown/restart comes straight out of the kernel doc "driver
> > instance overview" in drm_drv.c.
> >
> > All of the drivers in this patch were fairly straightforward to fix
> > since they already had a call to drm_atomic_helper_shutdown() at
> > remove/unbind time but were just lacking one at system shutdown. The
> > only hitch is that some of these drivers use the component model to
> > register/unregister their DRM devices. The shutdown callback is part
> > of the original device. The typical solution here, based on how other
> > DRM drivers do this, is to keep track of whether the device is bound
> > based on drvdata. In most cases the drvdata is the drm_device, so we
> > can just make sure it is NULL when the device is not bound. In some
> > drivers, this required minor code changes. To make things simpler,
> > drm_atomic_helper_shutdown() has been modified to consider a NULL
> > drm_device as a noop in the patch ("drm/atomic-helper:
> > drm_atomic_helper_shutdown(NULL) should be a noop").
> >
> > Suggested-by: Maxime Ripard <mripard@...nel.org>
> > Signed-off-by: Douglas Anderson <dianders@...omium.org>
> > ---
>
>
> I have just tested the whole series, thanks for the patch. For drm/loongson only:
>
>
> Reviewed-by: Sui Jingfeng <suijingfeng@...ngson.cn>
> Tested-by: Sui Jingfeng <suijingfeng@...ngson.cn>

Thanks!


> By the way, I add 'pr_info("lsdc_pci_shutdown\n");' into the lsdc_pci_shutdown() function,
> And seeing that lsdc_pci_shutdown() will be called when reboot and shutdown the machine.
> I did not witness something weird happen at present. As you have said, this is useful for
> drm panels drivers. But for the rest(drm/hibmc, drm/ast, drm/mgag200 and drm/loongson etc)
> drivers, you didn't mention what's the benefit for those drivers.

I didn't mention it because I have no idea! I presume that for
non-drm_panel use cases it's not a huge deal, otherwise it wouldn't
have been missing from so many drivers. Thus, my "one sentence" reason
for the non-drm_panel case is just "we should do this because the
documentation of the API says we should", which is already in the
commit message. ;-)

If you have a specific other benefit you'd like me to list then I'm happy to.


> Probably, you can
> mention it with at least one sentence at the next version. I also prefer to alter the
> lsdc_pci_shutdown() function as the following pattern:
>
>
> static void lsdc_pci_shutdown(struct pci_dev *pdev)
> {
>
>      struct drm_device *ddev = pci_get_drvdata(pdev);
>
>      drm_atomic_helper_shutdown(ddev);
> }

I was hoping to land this patch without spinning it unless there's a
good reason. How strongly do you feel about needing to change the
above? I will note that I coded it the way I did specifically to try
to follow the style in the documentation in "drm_drv.c". In the
example "driver_shutdown()" function you can see that they combined it
into one line and so I followed that style. ;-) That being said, I
have no problem changing this if I spin the patch.

-Doug

Powered by blists - more mailing lists