drm/i915: Defer enabling rc6 til after we submit the first batch/context
Some hardware requires a valid render context before it can initiate
rc6 power gating of the GPU; the default state of the GPU is not
sufficient and may lead to undefined behaviour. The first execution of
any batch will load the "golden render state", at which point it is safe
to enable rc6. As we do not forcibly load the kernel context at resume,
we have to hook into the batch submission to be sure that the render
state is setup before enabling rc6.
However, since we don't enable powersaving until that first batch, we
queued a delayed task in order to guarantee that the batch is indeed
submitted.
v2: Rearrange intel_disable_gt_powersave() to match.
v3: Apply user specified cur_freq (or idle_freq if not set).
v4: Give in, and supply a delayed work to autoenable rc6
v5: Mika suggested a couple of better names for delayed_resume_work
v6: Rebalance rpm_put around the autoenable task
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1468397438-21226-7-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index aab1e0b..c77ec10 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -6536,6 +6536,8 @@
dev_priv->rps.boost_freq = dev_priv->rps.max_freq;
mutex_unlock(&dev_priv->rps.hw_lock);
+
+ intel_autoenable_gt_powersave(dev_priv);
}
void intel_cleanup_gt_powersave(struct drm_i915_private *dev_priv)
@@ -6549,13 +6551,6 @@
intel_runtime_pm_put(dev_priv);
}
-static void gen6_suspend_rps(struct drm_i915_private *dev_priv)
-{
- flush_delayed_work(&dev_priv->rps.delayed_resume_work);
-
- gen6_disable_rps_interrupts(dev_priv);
-}
-
/**
* intel_suspend_gt_powersave - suspend PM work and helper threads
* @dev_priv: i915 device
@@ -6569,50 +6564,63 @@
if (INTEL_GEN(dev_priv) < 6)
return;
- gen6_suspend_rps(dev_priv);
+ if (cancel_delayed_work_sync(&dev_priv->rps.autoenable_work))
+ intel_runtime_pm_put(dev_priv);
- /* Force GPU to min freq during suspend */
- gen6_rps_idle(dev_priv);
+ /* gen6_rps_idle() will be called later to disable interrupts */
+}
+
+void intel_sanitize_gt_powersave(struct drm_i915_private *dev_priv)
+{
+ dev_priv->rps.enabled = true; /* force disabling */
+ intel_disable_gt_powersave(dev_priv);
+
+ gen6_reset_rps_interrupts(dev_priv);
}
void intel_disable_gt_powersave(struct drm_i915_private *dev_priv)
{
- if (IS_IRONLAKE_M(dev_priv)) {
- ironlake_disable_drps(dev_priv);
- } else if (INTEL_INFO(dev_priv)->gen >= 6) {
- intel_suspend_gt_powersave(dev_priv);
-
- mutex_lock(&dev_priv->rps.hw_lock);
- if (INTEL_INFO(dev_priv)->gen >= 9) {
- gen9_disable_rc6(dev_priv);
- gen9_disable_rps(dev_priv);
- } else if (IS_CHERRYVIEW(dev_priv))
- cherryview_disable_rps(dev_priv);
- else if (IS_VALLEYVIEW(dev_priv))
- valleyview_disable_rps(dev_priv);
- else
- gen6_disable_rps(dev_priv);
-
- dev_priv->rps.enabled = false;
- mutex_unlock(&dev_priv->rps.hw_lock);
- }
-}
-
-static void intel_gen6_powersave_work(struct work_struct *work)
-{
- struct drm_i915_private *dev_priv =
- container_of(work, struct drm_i915_private,
- rps.delayed_resume_work.work);
+ if (!READ_ONCE(dev_priv->rps.enabled))
+ return;
mutex_lock(&dev_priv->rps.hw_lock);
- gen6_reset_rps_interrupts(dev_priv);
+ if (INTEL_GEN(dev_priv) >= 9) {
+ gen9_disable_rc6(dev_priv);
+ gen9_disable_rps(dev_priv);
+ } else if (IS_CHERRYVIEW(dev_priv)) {
+ cherryview_disable_rps(dev_priv);
+ } else if (IS_VALLEYVIEW(dev_priv)) {
+ valleyview_disable_rps(dev_priv);
+ } else if (INTEL_GEN(dev_priv) >= 6) {
+ gen6_disable_rps(dev_priv);
+ } else if (IS_IRONLAKE_M(dev_priv)) {
+ ironlake_disable_drps(dev_priv);
+ }
+
+ dev_priv->rps.enabled = false;
+ mutex_unlock(&dev_priv->rps.hw_lock);
+}
+
+void intel_enable_gt_powersave(struct drm_i915_private *dev_priv)
+{
+ /* We shouldn't be disabling as we submit, so this should be less
+ * racy than it appears!
+ */
+ if (READ_ONCE(dev_priv->rps.enabled))
+ return;
+
+ /* Powersaving is controlled by the host when inside a VM */
+ if (intel_vgpu_active(dev_priv))
+ return;
+
+ mutex_lock(&dev_priv->rps.hw_lock);
if (IS_CHERRYVIEW(dev_priv)) {
cherryview_enable_rps(dev_priv);
} else if (IS_VALLEYVIEW(dev_priv)) {
valleyview_enable_rps(dev_priv);
- } else if (INTEL_INFO(dev_priv)->gen >= 9) {
+ } else if (INTEL_GEN(dev_priv) >= 9) {
gen9_enable_rc6(dev_priv);
gen9_enable_rps(dev_priv);
if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
@@ -6620,9 +6628,12 @@
} else if (IS_BROADWELL(dev_priv)) {
gen8_enable_rps(dev_priv);
__gen6_update_ring_freq(dev_priv);
- } else {
+ } else if (INTEL_GEN(dev_priv) >= 6) {
gen6_enable_rps(dev_priv);
__gen6_update_ring_freq(dev_priv);
+ } else if (IS_IRONLAKE_M(dev_priv)) {
+ ironlake_enable_drps(dev_priv);
+ intel_init_emon(dev_priv);
}
WARN_ON(dev_priv->rps.max_freq < dev_priv->rps.min_freq);
@@ -6632,18 +6643,47 @@
WARN_ON(dev_priv->rps.efficient_freq > dev_priv->rps.max_freq);
dev_priv->rps.enabled = true;
-
- gen6_enable_rps_interrupts(dev_priv);
-
mutex_unlock(&dev_priv->rps.hw_lock);
+}
+static void __intel_autoenable_gt_powersave(struct work_struct *work)
+{
+ struct drm_i915_private *dev_priv =
+ container_of(work, typeof(*dev_priv), rps.autoenable_work.work);
+ struct intel_engine_cs *rcs;
+ struct drm_i915_gem_request *req;
+
+ if (READ_ONCE(dev_priv->rps.enabled))
+ goto out;
+
+ rcs = &dev_priv->engine[RCS];
+ if (rcs->last_context)
+ goto out;
+
+ if (!rcs->init_context)
+ goto out;
+
+ mutex_lock(&dev_priv->drm.struct_mutex);
+
+ req = i915_gem_request_alloc(rcs, dev_priv->kernel_context);
+ if (IS_ERR(req))
+ goto unlock;
+
+ if (!i915.enable_execlists && i915_switch_context(req) == 0)
+ rcs->init_context(req);
+
+ /* Mark the device busy, calling intel_enable_gt_powersave() */
+ i915_add_request_no_flush(req);
+
+unlock:
+ mutex_unlock(&dev_priv->drm.struct_mutex);
+out:
intel_runtime_pm_put(dev_priv);
}
-void intel_enable_gt_powersave(struct drm_i915_private *dev_priv)
+void intel_autoenable_gt_powersave(struct drm_i915_private *dev_priv)
{
- /* Powersaving is controlled by the host when inside a VM */
- if (intel_vgpu_active(dev_priv))
+ if (READ_ONCE(dev_priv->rps.enabled))
return;
if (IS_IRONLAKE_M(dev_priv)) {
@@ -6664,21 +6704,13 @@
* paths, so the _noresume version is enough (and in case of
* runtime resume it's necessary).
*/
- if (schedule_delayed_work(&dev_priv->rps.delayed_resume_work,
- round_jiffies_up_relative(HZ)))
+ if (queue_delayed_work(dev_priv->wq,
+ &dev_priv->rps.autoenable_work,
+ round_jiffies_up_relative(HZ)))
intel_runtime_pm_get_noresume(dev_priv);
}
}
-void intel_reset_gt_powersave(struct drm_i915_private *dev_priv)
-{
- if (INTEL_INFO(dev_priv)->gen < 6)
- return;
-
- gen6_suspend_rps(dev_priv);
- dev_priv->rps.enabled = false;
-}
-
static void ibx_init_clock_gating(struct drm_device *dev)
{
struct drm_i915_private *dev_priv = to_i915(dev);
@@ -7785,8 +7817,8 @@
mutex_init(&dev_priv->rps.hw_lock);
spin_lock_init(&dev_priv->rps.client_lock);
- INIT_DELAYED_WORK(&dev_priv->rps.delayed_resume_work,
- intel_gen6_powersave_work);
+ INIT_DELAYED_WORK(&dev_priv->rps.autoenable_work,
+ __intel_autoenable_gt_powersave);
INIT_LIST_HEAD(&dev_priv->rps.clients);
INIT_LIST_HEAD(&dev_priv->rps.semaphores.link);
INIT_LIST_HEAD(&dev_priv->rps.mmioflips.link);