drm/i915: Sharing platform specific sequence between runtime and system suspend/...
authorSagar Kamble <sagar.a.kamble@intel.com>
Wed, 13 Aug 2014 17:37:06 +0000 (23:07 +0530)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Thu, 14 Aug 2014 14:13:34 +0000 (16:13 +0200)
On VLV, post S0i3 during i915_drm_thaw following issue is observed during ring
initialization.

[ 335.604039] [drm:stop_ring] ERROR render ring :timed out trying to stop ring
[ 336.607340] [drm:stop_ring] ERROR render ring :timed out trying to stop ring
[ 336.607345] [drm:init_ring_common] ERROR failed to set render ring head to zero ctl 00000000 head 00000000 tail 00000000 start 00000000
[ 337.610645] [drm:stop_ring] ERROR bsd ring :timed out trying to stop ring
[ 338.613952] [drm:stop_ring] ERROR bsd ring :timed out trying to stop ring
[ 338.613956] [drm:init_ring_common] ERROR failed to set bsd ring head to zero ctl 00000000 head 00000000 tail 00000000 start 00000000
[ 339.617256] [drm:stop_ring] ERROR render ring :timed out trying to stop ring
[ 339.617258] -----------[ cut here ]-----------
[ 339.617267] WARNING: CPU: 0 PID: 6 at drivers/gpu/drm/i915/intel_ringbuffer.c:1666 intel_cleanup_ring+0xe6/0xf0()
[ 339.617396] --[ end trace 5ef5ed1a3c92e2a6 ]--
[ 339.617428] [drm:__i915_drm_thaw] ERROR failed to re-initialize GPU, declaring wedged!

This is happening since wake is not enabled and Gunit registers are not restored.
For this system suspend/resume paths need to follow save/restore and additional
platform specific setup in suspend_complete and resume_prepare.

suspend_complete is shared unconditionaly for VLV, HSW, BDW. resume_prepare for
HSW and BDW has pc8 disabling which is needed during thaw_early so sharing
uncondtionally. For VLV and SNB runtime resume specific sequence exists.

Cc: Imre Deak <imre.deak@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Goel, Akash <akash.goel@intel.com>
Signed-off-by: Sagar Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/i915/i915_drv.c

index b06e975dba39d8a9ff0b4876ffb7e5acf5ba5c54..2f112853c36f70db4283f6d18b077c8d1cc16121 100644 (file)
@@ -496,7 +496,8 @@ bool i915_semaphore_is_enabled(struct drm_device *dev)
 
 
 static int intel_suspend_complete(struct drm_i915_private *dev_priv);
-static int intel_resume_prepare(struct drm_i915_private *dev_priv);
+static int intel_resume_prepare(struct drm_i915_private *dev_priv,
+                               bool rpm_resume);
 
 static int i915_drm_freeze(struct drm_device *dev)
 {
@@ -604,15 +605,17 @@ int i915_suspend(struct drm_device *dev, pm_message_t state)
 static int i915_drm_thaw_early(struct drm_device *dev)
 {
        struct drm_i915_private *dev_priv = dev->dev_private;
+       int ret;
 
-       if (IS_HASWELL(dev) || IS_BROADWELL(dev))
-               hsw_disable_pc8(dev_priv);
+       ret = intel_resume_prepare(dev_priv, false);
+       if (ret)
+               DRM_ERROR("Resume prepare failed: %d,Continuing resume\n", ret);
 
        intel_uncore_early_sanitize(dev, true);
        intel_uncore_sanitize(dev);
        intel_power_domains_init_hw(dev_priv);
 
-       return 0;
+       return ret;
 }
 
 static int __i915_drm_thaw(struct drm_device *dev, bool restore_gtt_mappings)
@@ -888,6 +891,7 @@ static int i915_pm_suspend_late(struct device *dev)
        struct pci_dev *pdev = to_pci_dev(dev);
        struct drm_device *drm_dev = pci_get_drvdata(pdev);
        struct drm_i915_private *dev_priv = drm_dev->dev_private;
+       int ret;
 
        /*
         * We have a suspedn ordering issue with the snd-hda driver also
@@ -901,13 +905,16 @@ static int i915_pm_suspend_late(struct device *dev)
        if (drm_dev->switch_power_state == DRM_SWITCH_POWER_OFF)
                return 0;
 
-       if (IS_HASWELL(drm_dev) || IS_BROADWELL(drm_dev))
-               hsw_enable_pc8(dev_priv);
+       ret = intel_suspend_complete(dev_priv);
 
-       pci_disable_device(pdev);
-       pci_set_power_state(pdev, PCI_D3hot);
+       if (ret)
+               DRM_ERROR("Suspend complete failed: %d\n", ret);
+       else {
+               pci_disable_device(pdev);
+               pci_set_power_state(pdev, PCI_D3hot);
+       }
 
-       return 0;
+       return ret;
 }
 
 static int i915_pm_resume_early(struct device *dev)
@@ -970,16 +977,19 @@ static int hsw_suspend_complete(struct drm_i915_private *dev_priv)
        return 0;
 }
 
-static int snb_resume_prepare(struct drm_i915_private *dev_priv)
+static int snb_resume_prepare(struct drm_i915_private *dev_priv,
+                               bool rpm_resume)
 {
        struct drm_device *dev = dev_priv->dev;
 
-       intel_init_pch_refclk(dev);
+       if (rpm_resume)
+               intel_init_pch_refclk(dev);
 
        return 0;
 }
 
-static int hsw_resume_prepare(struct drm_i915_private *dev_priv)
+static int hsw_resume_prepare(struct drm_i915_private *dev_priv,
+                               bool rpm_resume)
 {
        hsw_disable_pc8(dev_priv);
 
@@ -1315,7 +1325,8 @@ err1:
        return err;
 }
 
-static int vlv_resume_prepare(struct drm_i915_private *dev_priv)
+static int vlv_resume_prepare(struct drm_i915_private *dev_priv,
+                               bool rpm_resume)
 {
        struct drm_device *dev = dev_priv->dev;
        int err;
@@ -1340,8 +1351,10 @@ static int vlv_resume_prepare(struct drm_i915_private *dev_priv)
 
        vlv_check_no_gt_access(dev_priv);
 
-       intel_init_clock_gating(dev);
-       i915_gem_restore_fences(dev);
+       if (rpm_resume) {
+               intel_init_clock_gating(dev);
+               i915_gem_restore_fences(dev);
+       }
 
        return ret;
 }
@@ -1431,7 +1444,7 @@ static int intel_runtime_resume(struct device *device)
        intel_opregion_notify_adapter(dev, PCI_D0);
        dev_priv->pm.suspended = false;
 
-       ret = intel_resume_prepare(dev_priv);
+       ret = intel_resume_prepare(dev_priv, true);
        /*
         * No point of rolling back things in case of an error, as the best
         * we can do is to hope that things will still work (and disable RPM).
@@ -1450,6 +1463,10 @@ static int intel_runtime_resume(struct device *device)
        return ret;
 }
 
+/*
+ * This function implements common functionality of runtime and system
+ * suspend sequence.
+ */
 static int intel_suspend_complete(struct drm_i915_private *dev_priv)
 {
        struct drm_device *dev = dev_priv->dev;
@@ -1469,17 +1486,23 @@ static int intel_suspend_complete(struct drm_i915_private *dev_priv)
        return ret;
 }
 
-static int intel_resume_prepare(struct drm_i915_private *dev_priv)
+/*
+ * This function implements common functionality of runtime and system
+ * resume sequence. Variable rpm_resume used for implementing different
+ * code paths.
+ */
+static int intel_resume_prepare(struct drm_i915_private *dev_priv,
+                               bool rpm_resume)
 {
        struct drm_device *dev = dev_priv->dev;
        int ret;
 
        if (IS_GEN6(dev)) {
-               ret = snb_resume_prepare(dev_priv);
+               ret = snb_resume_prepare(dev_priv, rpm_resume);
        } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
-               ret = hsw_resume_prepare(dev_priv);
+               ret = hsw_resume_prepare(dev_priv, rpm_resume);
        } else if (IS_VALLEYVIEW(dev)) {
-               ret = vlv_resume_prepare(dev_priv);
+               ret = vlv_resume_prepare(dev_priv, rpm_resume);
        } else {
                WARN_ON(1);
                ret = -ENODEV;