On Fri, 21 Feb 2014 13:52:20 -0300
Paulo Zanoni <przanoni@xxxxxxxxx> wrote:
> From: Paulo Zanoni <paulo.r.zanoni@xxxxxxxxx>
>
> When we call gen6_gt_force_wake_put we don't actually put force_wake,
> we just schedule gen6_force_wake_work through mod_delayed_work, and
> that will eventually release force_wake.
>
> The problem is that we call intel_runtime_pm_put directly at
> gen6_gt_force_wake_put, so most of the times we put our runtime PM
> reference before the delayed work happens, so we may runtime suspend
> while force_wake is still supposed to be enabled if the graphics
> autosuspend_delay_ms is too small.
>
> Now the nice thing about the current code is that after it triggers
> the delayed work function it gets a refcount, and it only triggers the
> delayed work function if refcount is zero. This guarantees that when
> we schedule the funciton, it will run before we try to schedule it
> again, which simplifies the problem and allows for the current
> solution to work properly (hopefully!).
>
> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@xxxxxxxxx>
> ---
> drivers/gpu/drm/i915/intel_uncore.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c
> index c628414..1f7226f 100644
> --- a/drivers/gpu/drm/i915/intel_uncore.c
> +++ b/drivers/gpu/drm/i915/intel_uncore.c
> @@ -299,6 +299,8 @@ static void gen6_force_wake_work(struct work_struct *work)
> if (--dev_priv->uncore.forcewake_count == 0)
> dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
> spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
> +
> + intel_runtime_pm_put(dev_priv);
> }
>
> static void intel_uncore_forcewake_reset(struct drm_device *dev)
> @@ -393,6 +395,7 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
> void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
> {
> unsigned long irqflags;
> + bool delayed = false;
>
> if (!dev_priv->uncore.funcs.force_wake_put)
> return;
> @@ -405,13 +408,15 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
> spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
> if (--dev_priv->uncore.forcewake_count == 0) {
> dev_priv->uncore.forcewake_count++;
> + delayed = true;
> mod_delayed_work(dev_priv->wq,
> &dev_priv->uncore.force_wake_work,
> 1);
> }
> spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
>
> - intel_runtime_pm_put(dev_priv);
> + if (!delayed)
> + intel_runtime_pm_put(dev_priv);
> }
>
> /* We give fast paths for the really cool registers */
Do we need this for the VLV path too?
It's a little confusing that we do this delayed thing, incrementing the
count and then decrementing again in the work queue, but what you have
looks correct for both cases.
So with the VLV thing addressed:
Reviewed-by: Jesse Barnes <jbarnes@xxxxxxxxxxxxxxxx>
--
Jesse Barnes, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
