[Intel-gfx] [PATCH] drm/i915/backlight: Fix backlight takeover on LPT
Tolga Cakir
cevelnet at gmail.com
Wed Nov 21 12:37:22 UTC 2018
Am Mi., 21. Nov. 2018 um 11:45 Uhr schrieb Ville Syrjälä
<ville.syrjala at linux.intel.com>:
>
> On Wed, Nov 21, 2018 at 08:16:06AM +0100, Tolga Cakir wrote:
> > Am Di., 20. Nov. 2018 um 19:51 Uhr schrieb Ville Syrjälä
> > <ville.syrjala at linux.intel.com>:
> > >
> > > On Tue, Nov 20, 2018 at 11:39:26AM +0100, Maarten Lankhorst wrote:
> > > > On lynxpoint the bios sometimes sets up the backlight using the CPU
> > > > display, but the driver expects using the PWM PCH override register.
> > > >
> > > > Read the value from the CPU register, then convert it to the other
> > > > units by converting from the old duty cycle, to freq, to the new units.
> > > >
> > > > This value is then programmed in the override register, after which
> > > > we set the override and disable the CPU display control. This allows
> > > > us to switch the source without flickering, and make the backlight
> > > > controls work in the driver.
> > > >
> > > > Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108225
> > > > Cc: Basil Eric Rabi <ericbasil.rabi at gmail.com>
> > > > Cc: Hans de Goede <jwrdegoede at fedoraproject.org>
> > > > Cc: Tolga Cakir <cevelnet at gmail.com>
> > > > Tested-by: Tolga Cakir <cevelnet at gmail.com>
> > > > ---
> > > > drivers/gpu/drm/i915/intel_panel.c | 34 +++++++++++++++++++++++++++---
> > > > 1 file changed, 31 insertions(+), 3 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c
> > > > index e6cd7b55c018..3357232f1b0e 100644
> > > > --- a/drivers/gpu/drm/i915/intel_panel.c
> > > > +++ b/drivers/gpu/drm/i915/intel_panel.c
> > > > @@ -1485,7 +1485,7 @@ static int lpt_setup_backlight(struct intel_connector *connector, enum pipe unus
> > > > struct drm_i915_private *dev_priv = to_i915(connector->base.dev);
> > > > struct intel_panel *panel = &connector->panel;
> > > > u32 pch_ctl1, pch_ctl2, val;
> > > > - bool alt;
> > > > + bool alt, cpu_mode = false;
> > > >
> > > > if (HAS_PCH_LPT(dev_priv))
> > > > alt = I915_READ(SOUTH_CHICKEN2) & LPT_PWM_GRANULARITY;
> > > > @@ -1507,12 +1507,40 @@ static int lpt_setup_backlight(struct intel_connector *connector, enum pipe unus
> > > >
> > > > panel->backlight.min = get_backlight_min_vbt(connector);
> > > >
> > > > - val = lpt_get_backlight(connector);
> > > > + panel->backlight.enabled = pch_ctl1 & BLM_PCH_PWM_ENABLE;
> > > > + if (panel->backlight.enabled && HAS_PCH_LPT(dev_priv) &&
> > > > + (I915_READ(BLC_PWM_CPU_CTL2) & BLM_PWM_ENABLE) &&
> > > > + !WARN_ON(pch_ctl1 & BLM_PCH_OVERRIDE_ENABLE)) {
> > > > + u32 freq;
> > > > +
> > > > + cpu_mode = true;
> > > > + /*
> > > > + * We're in cpu mode, convert to PCH units.
> > > > + *
> > > > + * Convert CPU pwm tick back to hz, back to new PCH units again.
> > > > + * this is the same formula as pch_hz_to_pwm, but the other way
> > > > + * around..
> > > > + */
> > > > + val = pch_get_backlight(connector);
> > > > + freq = DIV_ROUND_CLOSEST(KHz(dev_priv->rawclk_freq), val * 128);
> > >
> > > The docs are telling me that HSW/BDW use cdclk for the CPU backlight,
> > > and the increment is configurable as well.
> > >
> > > Also what about S4 resume? Don't we need this trick there as well?
> > >
> >
> > Hi,
> >
> > I couldn't test this patch for S4 resume on my Dell M3800 laptop (Core
> > i7-4702HQ, Lynx Point), as fastboot=1 doesn't work for S4 resume, only
> > normal boot. Therefore, there are no issues on S4 wake after
> > hibernate. Backlight controls etc. work as expected; just like
> > fastboot=0.
>
> I think fastboot should work the same for s4 resume as boot, but
> only if your boot kernel doesn't load i915.ko. That way the
> resuming kernel will directly take over from the BIOS programmed
> display configuration. OTOH if the boot kernel loads i915 then
> it will shut down all displays before handing control over to
> the resuming kernel.
>
You're right, fastboot=1 worked for S4 resume, it was GNOME doing the
modeset after S4 resume for whatever reason. I have disabled GNOME for
further testing. Just hibernated from console and S4 resumed back to
console: no flickering. As you already suspected, backlight controls
didn't work, eventhough they worked on normal boot with fastboot=1.
So, this patch also needs to be applied to S4 resume, as you said.
Good catch!
I'm ready to test as soon as the patches are ready.
> >
> > I have applied "[1/2] drm/i915: Enable fastset for non-boot modesets"
> > by Maarten, but that doesn't give me fastboot for S4 resume aswell.
> > Same behavior as unpatched on my system.
> >
> > Please let me know, when there is more I can test.
> >
> > Cheers,
> > Tolga
> >
> > > > +
> > > > + val = lpt_hz_to_pwm(connector, freq);
> > > > + } else
> > > > + val = lpt_get_backlight(connector);
> > > > val = intel_panel_compute_brightness(connector, val);
> > > > panel->backlight.level = clamp(val, panel->backlight.min,
> > > > panel->backlight.max);
> > > >
> > > > - panel->backlight.enabled = pch_ctl1 & BLM_PCH_PWM_ENABLE;
> > > > + if (cpu_mode) {
> > > > + u32 tmp;
> > > > +
> > > > + /* Use PWM mode, instead of cpu clock */
> > > > + lpt_set_backlight(connector->base.state, panel->backlight.level);
> > > > + I915_WRITE(BLC_PWM_PCH_CTL1, pch_ctl1 | BLM_PCH_OVERRIDE_ENABLE);
> > > > +
> > > > + tmp = I915_READ(BLC_PWM_CPU_CTL2);
> > > > + I915_WRITE(BLC_PWM_CPU_CTL2, tmp & ~BLM_PWM_ENABLE);
> > > > + }
> > > >
> > > > return 0;
> > > > }
> > > > --
> > > > 2.19.1
> > > >
> > > > _______________________________________________
> > > > Intel-gfx mailing list
> > > > Intel-gfx at lists.freedesktop.org
> > > > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
> > >
> > > --
> > > Ville Syrjälä
> > > Intel
>
> --
> Ville Syrjälä
> Intel
More information about the Intel-gfx
mailing list