[Intel-gfx] [PATCH] drm/i915: add hotplug activation period to hotplug update mask

[Jani Nikula]

On Wed, 21 Oct 2015, Ville Syrjälä <ville.syrjala at linux.intel.com> wrote:
> On Wed, Oct 21, 2015 at 05:22:43PM +0300, Jani Nikula wrote:
>> commit 0706f17c307b056ff6f1848320ba82d76945a6ff
>> Author: Egbert Eich <eich at suse.de>
>> Date:   Wed Sep 23 16:15:27 2015 +0200
>> 
>>     drm/i915: Avoid race of intel_crt_detect_hotplug() with HPD interrupt, v2
>> 
>> added a check with WARN to ensure only bits within the mask are
>> enabled. Turns out that doesn't hold for G4X, which spits out:
>> 
>> [    2.641439] ------------[ cut here ]------------
>> [    2.641444] WARNING: CPU: 0 PID: 1 at drivers/gpu/drm/i915/i915_irq.c:182 i915_hotplug_interrupt_update_locked+0x45/0x83()
>> [    2.641446] WARN_ON(bits & ~mask)
>> etc.
>> 
>> Add CRT_HOTPLUG_ACTIVATION_PERIOD_64 to the mask to fix the warning.
>> 
>> Reported-by: Oleksij Rempel <linux at rempel-privat.de>
>> References: https://bugzilla.kernel.org/show_bug.cgi?id=104991
>> Fixes: 0706f17c307b ("drm/i915: Avoid race of intel_crt_detect_hotplug() with HPD interrupt, v2")
>> Cc: Egbert Eich <eich at suse.de>
>> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
>> Signed-off-by: Jani Nikula <jani.nikula at intel.com>
>> ---
>>  drivers/gpu/drm/i915/i915_irq.c | 7 ++++---
>>  1 file changed, 4 insertions(+), 3 deletions(-)
>> 
>> diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
>> index 7f91f74961f4..5b9f63d4318b 100644
>> --- a/drivers/gpu/drm/i915/i915_irq.c
>> +++ b/drivers/gpu/drm/i915/i915_irq.c
>> @@ -4236,9 +4236,10 @@ static void i915_hpd_irq_setup(struct drm_device *dev)
>>  
>>  	/* Ignore TV since it's buggy */
>>  	i915_hotplug_interrupt_update_locked(dev_priv,
>> -				      (HOTPLUG_INT_EN_MASK
>> -				       | CRT_HOTPLUG_VOLTAGE_COMPARE_MASK),
>> -				      hotplug_en);
>> +					     HOTPLUG_INT_EN_MASK |
>> +					     CRT_HOTPLUG_VOLTAGE_COMPARE_MASK |
>> +					     CRT_HOTPLUG_ACTIVATION_PERIOD_64,
>> +					     hotplug_en);
>
> Or maybe just ~CRT_HOTPLUG_FORCE_DETECT ?

This patch already potentially changes behaviour by explicitly setting
the activation period to zero (CRT_HOTPLUG_ACTIVATION_PERIOD_32) on
non-g4x. Previously it was "don't care". Call me coward, but I am not
comfortable with setting the rest of the bits to zero, at least not in
the context of this fix, and risking regressions on old machines.

BR,
Jani.


>
>>  }
>>  
>>  static irqreturn_t i965_irq_handler(int irq, void *arg)
>> -- 
>> 2.1.4
>> 
>> _______________________________________________
>> Intel-gfx mailing list
>> Intel-gfx at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>
> -- 
> Ville Syrjälä
> Intel OTC

-- 
Jani Nikula, Intel Open Source Technology Center