drm/i915: dynamic render p-state support for Sandy Bridge

Add an interrupt handler for switching graphics frequencies and handling
PM interrupts.  This should allow for increased performance when busy
and lower power consumption when idle.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drivers/gpu/drm/i915/i915_debugfs.c

@@ -797,15 +797,51 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused)
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
 	struct drm_device *dev = node->minor->dev;
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	u16 rgvswctl = I915_READ16(MEMSWCTL);
-	u16 rgvstat = I915_READ16(MEMSTAT_ILK);
-
-	seq_printf(m, "Requested P-state: %d\n", (rgvswctl >> 8) & 0xf);
-	seq_printf(m, "Requested VID: %d\n", rgvswctl & 0x3f);
-	seq_printf(m, "Current VID: %d\n", (rgvstat & MEMSTAT_VID_MASK) >>
-		   MEMSTAT_VID_SHIFT);
-	seq_printf(m, "Current P-state: %d\n",
-		   (rgvstat & MEMSTAT_PSTATE_MASK) >> MEMSTAT_PSTATE_SHIFT);
+
+	if (IS_GEN5(dev)) {
+		u16 rgvswctl = I915_READ16(MEMSWCTL);
+		u16 rgvstat = I915_READ16(MEMSTAT_ILK);
+
+		seq_printf(m, "Requested P-state: %d\n", (rgvswctl >> 8) & 0xf);
+		seq_printf(m, "Requested VID: %d\n", rgvswctl & 0x3f);
+		seq_printf(m, "Current VID: %d\n", (rgvstat & MEMSTAT_VID_MASK) >>
+			   MEMSTAT_VID_SHIFT);
+		seq_printf(m, "Current P-state: %d\n",
+			   (rgvstat & MEMSTAT_PSTATE_MASK) >> MEMSTAT_PSTATE_SHIFT);
+	} else if (IS_GEN6(dev)) {
+		u32 gt_perf_status = I915_READ(GEN6_GT_PERF_STATUS);
+		u32 rp_state_limits = I915_READ(GEN6_RP_STATE_LIMITS);
+		u32 rp_state_cap = I915_READ(GEN6_RP_STATE_CAP);
+		int max_freq;
+
+		/* RPSTAT1 is in the GT power well */
+		__gen6_force_wake_get(dev_priv);
+
+		seq_printf(m, "GT_PERF_STATUS: 0x%08x\n", gt_perf_status);
+		seq_printf(m, "RPSTAT1: 0x%08x\n", I915_READ(GEN6_RPSTAT1));
+		seq_printf(m, "Render p-state ratio: %d\n",
+			   (gt_perf_status & 0xff00) >> 8);
+		seq_printf(m, "Render p-state VID: %d\n",
+			   gt_perf_status & 0xff);
+		seq_printf(m, "Render p-state limit: %d\n",
+			   rp_state_limits & 0xff);
+
+		max_freq = (rp_state_cap & 0xff0000) >> 16;
+		seq_printf(m, "Lowest (RPN) frequency: %dMHz\n",
+			   max_freq * 100);
+
+		max_freq = (rp_state_cap & 0xff00) >> 8;
+		seq_printf(m, "Nominal (RP1) frequency: %dMHz\n",
+			   max_freq * 100);
+
+		max_freq = rp_state_cap & 0xff;
+		seq_printf(m, "Max non-overclocked (RP0) frequency: %dMHz\n",
+			   max_freq * 100);
+
+		__gen6_force_wake_put(dev_priv);
+	} else {
+		seq_printf(m, "no P-state info available\n");
+	}
 
 	return 0;
 }

drivers/gpu/drm/i915/i915_drv.h

@@ -1264,6 +1264,7 @@ extern void intel_disable_fbc(struct drm_device *dev);
 extern void intel_enable_fbc(struct drm_crtc *crtc, unsigned long interval);
 extern bool intel_fbc_enabled(struct drm_device *dev);
 extern bool ironlake_set_drps(struct drm_device *dev, u8 val);
+extern void gen6_set_rps(struct drm_device *dev, u8 val);
 extern void intel_detect_pch (struct drm_device *dev);
 extern int intel_trans_dp_port_sel (struct drm_crtc *crtc);
 

drivers/gpu/drm/i915/i915_irq.c

@@ -397,11 +397,49 @@ static void notify_ring(struct drm_device *dev,
 		  jiffies + msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD));
 }
 
+static void gen6_pm_irq_handler(struct drm_device *dev)
+{
+	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	u8 new_delay = dev_priv->cur_delay;
+	u32 pm_iir;
+
+	pm_iir = I915_READ(GEN6_PMIIR);
+	if (!pm_iir)
+		return;
+
+	if (pm_iir & GEN6_PM_RP_UP_THRESHOLD) {
+		if (dev_priv->cur_delay != dev_priv->max_delay)
+			new_delay = dev_priv->cur_delay + 1;
+		if (new_delay > dev_priv->max_delay)
+			new_delay = dev_priv->max_delay;
+	} else if (pm_iir & (GEN6_PM_RP_DOWN_THRESHOLD | GEN6_PM_RP_DOWN_TIMEOUT)) {
+		if (dev_priv->cur_delay != dev_priv->min_delay)
+			new_delay = dev_priv->cur_delay - 1;
+		if (new_delay < dev_priv->min_delay) {
+			new_delay = dev_priv->min_delay;
+			I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,
+				   I915_READ(GEN6_RP_INTERRUPT_LIMITS) |
+				   ((new_delay << 16) & 0x3f0000));
+		} else {
+			/* Make sure we continue to get down interrupts
+			 * until we hit the minimum frequency */
+			I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,
+				   I915_READ(GEN6_RP_INTERRUPT_LIMITS) & ~0x3f0000);
+		}
+
+	}
+
+	gen6_set_rps(dev, new_delay);
+	dev_priv->cur_delay = new_delay;
+
+	I915_WRITE(GEN6_PMIIR, pm_iir);
+}
+
 static irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
 	int ret = IRQ_NONE;
-	u32 de_iir, gt_iir, de_ier, pch_iir;
+	u32 de_iir, gt_iir, de_ier, pch_iir, pm_iir;
 	u32 hotplug_mask;
 	struct drm_i915_master_private *master_priv;
 	u32 bsd_usr_interrupt = GT_BSD_USER_INTERRUPT;
@@ -417,8 +455,10 @@ static irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 	de_iir = I915_READ(DEIIR);
 	gt_iir = I915_READ(GTIIR);
 	pch_iir = I915_READ(SDEIIR);
+	pm_iir = I915_READ(GEN6_PMIIR);
 
-	if (de_iir == 0 && gt_iir == 0 && pch_iir == 0)
+	if (de_iir == 0 && gt_iir == 0 && pch_iir == 0 &&
+	    (!IS_GEN6(dev) || pm_iir == 0))
 		goto done;
 
 	if (HAS_PCH_CPT(dev))
@@ -470,6 +510,9 @@ static irqreturn_t ironlake_irq_handler(struct drm_device *dev)
 		i915_handle_rps_change(dev);
 	}
 
+	if (IS_GEN6(dev))
+		gen6_pm_irq_handler(dev);
+
 	/* should clear PCH hotplug event before clear CPU irq */
 	I915_WRITE(SDEIIR, pch_iir);
 	I915_WRITE(GTIIR, gt_iir);

drivers/gpu/drm/i915/i915_reg.h

@@ -1188,6 +1188,10 @@
 #define DDRMPLL1		0X12c20
 #define PEG_BAND_GAP_DATA	0x14d68
 
+#define GEN6_GT_PERF_STATUS	0x145948
+#define GEN6_RP_STATE_LIMITS	0x145994
+#define GEN6_RP_STATE_CAP	0x145998
+
 /*
  * Logical Context regs
  */
@@ -3169,7 +3173,7 @@
 #define  FORCEWAKE				0xA18C
 #define  FORCEWAKE_ACK				0x130090
 
-#define GEN6_RC_NORMAL_FREQ			0xA008
+#define GEN6_RPNSWREQ				0xA008
 #define   GEN6_TURBO_DISABLE			(1<<31)
 #define   GEN6_FREQUENCY(x)			((x)<<25)
 #define   GEN6_OFFSET(x)			((x)<<19)
@@ -3185,6 +3189,7 @@
 #define   GEN6_RC_CTL_HW_ENABLE			(1<<31)
 #define GEN6_RP_DOWN_TIMEOUT			0xA010
 #define GEN6_RP_INTERRUPT_LIMITS		0xA014
+#define GEN6_RPSTAT1				0xA01C
 #define GEN6_RP_CONTROL				0xA024
 #define   GEN6_RP_MEDIA_TURBO			(1<<11)
 #define   GEN6_RP_USE_NORMAL_FREQ		(1<<9)
@@ -3208,6 +3213,7 @@
 #define GEN6_RC6_THRESHOLD			0xA0B8
 #define GEN6_RC6p_THRESHOLD			0xA0BC
 #define GEN6_RC6pp_THRESHOLD			0xA0C0
+#define GEN6_PMINTRMSK				0xA168
 
 #define GEN6_PMISR				0x44020
 #define GEN6_PMIMR				0x44024

drivers/gpu/drm/i915/i915_suspend.c

@@ -817,8 +817,10 @@ int i915_save_state(struct drm_device *dev)
 		dev_priv->saveIMR = I915_READ(IMR);
 	}
 
-	if (HAS_PCH_SPLIT(dev))
+	if (IS_IRONLAKE_M(dev))
 		ironlake_disable_drps(dev);
+	if (IS_GEN6(dev))
+		gen6_disable_rps(dev);
 
 	intel_disable_clock_gating(dev);
 
@@ -867,11 +869,14 @@ int i915_restore_state(struct drm_device *dev)
 	/* Clock gating state */
 	intel_enable_clock_gating(dev);
 
-	if (HAS_PCH_SPLIT(dev)) {
+	if (IS_IRONLAKE_M(dev)) {
 		ironlake_enable_drps(dev);
 		intel_init_emon(dev);
 	}
 
+	if (IS_GEN6(dev))
+		gen6_enable_rps(dev_priv);
+
 	/* Cache mode state */
 	I915_WRITE (CACHE_MODE_0, dev_priv->saveCACHE_MODE_0 | 0xffff0000);
 

drivers/gpu/drm/i915/intel_display.c

@@ -6021,6 +6021,25 @@ void ironlake_disable_drps(struct drm_device *dev)
 
 }
 
+void gen6_set_rps(struct drm_device *dev, u8 val)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	u32 swreq;
+
+	swreq = (val & 0x3ff) << 25;
+	I915_WRITE(GEN6_RPNSWREQ, swreq);
+}
+
+void gen6_disable_rps(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	I915_WRITE(GEN6_RPNSWREQ, 1 << 31);
+	I915_WRITE(GEN6_PMINTRMSK, 0xffffffff);
+	I915_WRITE(GEN6_PMIER, 0);
+	I915_WRITE(GEN6_PMIIR, I915_READ(GEN6_PMIIR));
+}
+
 static unsigned long intel_pxfreq(u32 vidfreq)
 {
 	unsigned long freq;
@@ -6107,7 +6126,7 @@ void intel_init_emon(struct drm_device *dev)
 	dev_priv->corr = (lcfuse & LCFUSE_HIV_MASK);
 }
 
-static void gen6_enable_rc6(struct drm_i915_private *dev_priv)
+void gen6_enable_rps(struct drm_i915_private *dev_priv)
 {
 	int i;
 
@@ -6120,7 +6139,7 @@ static void gen6_enable_rc6(struct drm_i915_private *dev_priv)
 	I915_WRITE(GEN6_RC_STATE, 0);
 	__gen6_force_wake_get(dev_priv);
 
-	/* disable the counters and set determistic thresholds */
+	/* disable the counters and set deterministic thresholds */
 	I915_WRITE(GEN6_RC_CONTROL, 0);
 
 	I915_WRITE(GEN6_RC1_WAKE_RATE_LIMIT, 1000 << 16);
@@ -6144,7 +6163,7 @@ static void gen6_enable_rc6(struct drm_i915_private *dev_priv)
 		   GEN6_RC_CTL_EI_MODE(1) |
 		   GEN6_RC_CTL_HW_ENABLE);
 
-	I915_WRITE(GEN6_RC_NORMAL_FREQ,
+	I915_WRITE(GEN6_RPNSWREQ,
 		   GEN6_FREQUENCY(10) |
 		   GEN6_OFFSET(0) |
 		   GEN6_AGGRESSIVE_TURBO);
@@ -6189,6 +6208,9 @@ static void gen6_enable_rc6(struct drm_i915_private *dev_priv)
 		   GEN6_PM_RP_DOWN_THRESHOLD |
 		   GEN6_PM_RP_UP_EI_EXPIRED |
 		   GEN6_PM_RP_DOWN_EI_EXPIRED);
+	I915_WRITE(GEN6_PMIMR, 0);
+	/* enable all PM interrupts */
+	I915_WRITE(GEN6_PMINTRMSK, 0);
 
 	__gen6_force_wake_put(dev_priv);
 }
@@ -6381,9 +6403,6 @@ void intel_enable_clock_gating(struct drm_device *dev)
 				   I915_READ(MCHBAR_RENDER_STANDBY) & ~RCX_SW_EXIT);
 		}
 	}
-
-	if (IS_GEN6(dev))
-		gen6_enable_rc6(dev_priv);
 }
 
 void intel_disable_clock_gating(struct drm_device *dev)
@@ -6657,6 +6676,9 @@ void intel_modeset_init(struct drm_device *dev)
 		intel_init_emon(dev);
 	}
 
+	if (IS_GEN6(dev))
+		gen6_enable_rps(dev_priv);
+
 	INIT_WORK(&dev_priv->idle_work, intel_idle_update);
 	setup_timer(&dev_priv->idle_timer, intel_gpu_idle_timer,
 		    (unsigned long)dev);
@@ -6690,6 +6712,8 @@ void intel_modeset_cleanup(struct drm_device *dev)
 
 	if (IS_IRONLAKE_M(dev))
 		ironlake_disable_drps(dev);
+	if (IS_GEN6(dev))
+		gen6_disable_rps(dev);
 
 	intel_disable_clock_gating(dev);
 

drivers/gpu/drm/i915/intel_drv.h

@@ -298,6 +298,8 @@ extern void intel_enable_clock_gating(struct drm_device *dev);
 extern void intel_disable_clock_gating(struct drm_device *dev);
 extern void ironlake_enable_drps(struct drm_device *dev);
 extern void ironlake_disable_drps(struct drm_device *dev);
+extern void gen6_enable_rps(struct drm_i915_private *dev_priv);
+extern void gen6_disable_rps(struct drm_device *dev);
 extern void intel_init_emon(struct drm_device *dev);
 
 extern int intel_pin_and_fence_fb_obj(struct drm_device *dev,