lkml.org 
[lkml]   [2021]   [Jul]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] drm/bridge: anx7625: Use pm_runtime_force_{suspend,resume}
Hi Pi-Hsun,

On Fri, Jul 16, 2021 at 04:26:44PM +0800, Pi-Hsun Shih wrote:
> On Wed, Jul 14, 2021 at 6:32 PM Daniel Vetter <daniel@ffwll.ch> wrote:
> > On Wed, Jul 14, 2021 at 02:01:59PM +0800, Pi-Hsun Shih wrote:
> > > Use pm_runtime_force_suspend and pm_runtime_force_resume to ensure that
> > > anx7625 would always be powered off when suspended. Also update the
> > > bridge enable hook to always ensure that the anx7625 is powered on
> > > before starting DP operations.
> > >
> > > Fixes: 409776fa3c42 ("drm/bridge: anx7625: add suspend / resume hooks")
> > >
> > > Signed-off-by: Pi-Hsun Shih <pihsun@chromium.org>
> > >
> > > ---
> > >
> > > An issue was found that the anx7625 driver won't power off when used as
> > > eDP bridge on Asurada board if suspend is entered via VT2.
> > >
> > > The reason is that in this case, anx7625_suspend won't power off anx7625
> > > (since intp_irq is not set). And anx7625_bridge_disable is only called
> > > indirectly by other driver's (mediatek-drm) suspend.
> > > pm_runtime_put_sync won't do anything since it's already in system
> > > suspend.
> > >
> > > If not in VT2, the bridge disable is indirectly called when Chrome
> > > stops, so anx7625 will be powered off correctly.
> > >
> > > To fix the issue, the suspend resume hooks are changed to
> > > pm_runtime_force_{suspend,resume} to ensure the runtime suspend / resume
> > > is always called correctly when system suspend / resume.
> > > (Note that IRQ no longer needs to be disabled on suspend after commit
> > > f03ab6629c7b ("drm/bridge: anx7625: Make hpd workqueue freezable"))
> > >
> > > Since bridge disable is called indirectly by mediatek-drm driver's
> > > suspend, it might happens after anx7625 suspend is called. So a check
> > > if the driver is already suspended via pm_runtime_force_suspend is also
> > > added, to ensure that the anx7625_dp_stop won't be called when power
> > > is off. And also since bridge enable might happens before anx7625 resume
> > > is called, a check to that is also added, and would force resume the
> > > device in this case.
> > >
> > > I'm not sure if the approach to fix this is the most appropriate way,
> > > since using pm_runtime_force_resume in bridge enable kinda feels hacky
> > > to me. I'm open to any suggestions.
> >
> > I thought the real fix was to create device links between the bridge and
> > the other parts of the overall drm driver, so that the driver core can
> > resume devices in the right order.
> >
> > Unfortunately those device link patches haven't made it in yet. Quick
> > search on lore didn't find anything, maybe I was just dreaming, or maybe
> > the patches only existed for panels.
> >
> > Either way, this is a drm_bridge.c problem that needs to be fixed there,
> > not individually in each driver.
> > -Daniel
>
> Hi,
>
> Thanks for the response, I did find some discussion about this in 2018 for
> drm_panel
> (https://patchwork.kernel.org/project/dri-devel/patch/b53584fd988d045c13de22d81825395b0ae0aad7.1524727888.git.jsarha@ti.com/),
> which also mentioned drm_bridge.
>
> From that thread it seems that linking all bridges with the previous one would
> break some drivers, and there was no conclusion on how this should be done.
>
> I have some ideas on how to solve this issue for the anx7625 driver without
> affecting other drivers, are patches that do one of the following acceptable?
> * Add some opt-in flag to drm_bridge which, if set, would create a stateless
> device link between the bridge and the encoder in drm_bridge_attach. And use
> the flag in anx7625 driver.

On a side note, we're moving away from encoders, towards modelling all
devices after the CRTC as bridges. This would need to be adapted
accordingly.

> * Add the stateless device link in the anx7625 driver inside
> anx7625_bridge_attach
> (We can remove the link if a general solution for drm_bridge comes out later).
>
> Or is it still preferred to have some general solution in drm_bridge without
> explicit opt-in?
>
> > > ---
> > > drivers/gpu/drm/bridge/analogix/anx7625.c | 55 +++++++++--------------
> > > 1 file changed, 20 insertions(+), 35 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/bridge/analogix/anx7625.c b/drivers/gpu/drm/bridge/analogix/anx7625.c
> > > index a3d82377066b..9d0f5dc88b16 100644
> > > --- a/drivers/gpu/drm/bridge/analogix/anx7625.c
> > > +++ b/drivers/gpu/drm/bridge/analogix/anx7625.c
> > > @@ -1559,7 +1559,20 @@ static void anx7625_bridge_enable(struct drm_bridge *bridge)
> > >
> > > DRM_DEV_DEBUG_DRIVER(dev, "drm enable\n");
> > >
> > > - pm_runtime_get_sync(dev);
> > > + /*
> > > + * The only case where pm_runtime is disabled here is when the function
> > > + * is called other driver's resume hook by
> > > + * drm_mode_config_helper_resume, but when the pm_runtime_force_resume
> > > + * hasn't been called on this device.
> > > + *
> > > + * pm_runtime_get_sync won't power on anx7625 in this case since we're
> > > + * in system resume, so instead we force resume anx7625 to make sure
> > > + * the following anx7625_dp_start would succeed.
> > > + */
> > > + if (pm_runtime_enabled(dev))
> > > + pm_runtime_get_sync(dev);
> > > + else
> > > + pm_runtime_force_resume(dev);
> > >
> > > anx7625_dp_start(ctx);
> > > }
> > > @@ -1571,9 +1584,10 @@ static void anx7625_bridge_disable(struct drm_bridge *bridge)
> > >
> > > DRM_DEV_DEBUG_DRIVER(dev, "drm disable\n");
> > >
> > > - anx7625_dp_stop(ctx);
> > > -
> > > - pm_runtime_put_sync(dev);
> > > + if (pm_runtime_enabled(dev)) {
> > > + anx7625_dp_stop(ctx);
> > > + pm_runtime_put_sync(dev);
> > > + }
> > > }
> > >
> > > static enum drm_connector_status
> > > @@ -1705,38 +1719,9 @@ static int __maybe_unused anx7625_runtime_pm_resume(struct device *dev)
> > > return 0;
> > > }
> > >
> > > -static int __maybe_unused anx7625_resume(struct device *dev)
> > > -{
> > > - struct anx7625_data *ctx = dev_get_drvdata(dev);
> > > -
> > > - if (!ctx->pdata.intp_irq)
> > > - return 0;
> > > -
> > > - if (!pm_runtime_enabled(dev) || !pm_runtime_suspended(dev)) {
> > > - enable_irq(ctx->pdata.intp_irq);
> > > - anx7625_runtime_pm_resume(dev);
> > > - }
> > > -
> > > - return 0;
> > > -}
> > > -
> > > -static int __maybe_unused anx7625_suspend(struct device *dev)
> > > -{
> > > - struct anx7625_data *ctx = dev_get_drvdata(dev);
> > > -
> > > - if (!ctx->pdata.intp_irq)
> > > - return 0;
> > > -
> > > - if (!pm_runtime_enabled(dev) || !pm_runtime_suspended(dev)) {
> > > - anx7625_runtime_pm_suspend(dev);
> > > - disable_irq(ctx->pdata.intp_irq);
> > > - }
> > > -
> > > - return 0;
> > > -}
> > > -
> > > static const struct dev_pm_ops anx7625_pm_ops = {
> > > - SET_SYSTEM_SLEEP_PM_OPS(anx7625_suspend, anx7625_resume)
> > > + SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
> > > + pm_runtime_force_resume)
> > > SET_RUNTIME_PM_OPS(anx7625_runtime_pm_suspend,
> > > anx7625_runtime_pm_resume, NULL)
> > > };
> > >
> > > base-commit: c0d438dbc0b74901f1901d97a6c84f38daa0c831

--
Regards,

Laurent Pinchart

\
 
 \ /
  Last update: 2021-07-16 11:15    [W:2.674 / U:1.096 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site