Merge branch 'msm-next' of git://people.freedesktop.org/~robclark/linux into drm-next

Noteworthy changes this time:
1) 4k support for newer chips (ganging up hwpipes and mixers)
2) using OPP bindings for gpu
3) more prep work towards per-process pagetables

* 'msm-next' of git://people.freedesktop.org/~robclark/linux: (47 commits)
  msm/drm: gpu: Dynamically locate the clocks from the device tree
  drm/msm: gpu: Use OPP tables if we can
  drm/msm: Hard code the GPU "slow frequency"
  drm/msm: Add MSM_PARAM_GMEM_BASE
  drm/msm: Reference count address spaces
  drm/msm: Make sure to detach the MMU during GPU cleanup
  drm/msm/mdp5: Enable 3D mux in mdp5_ctl
  drm/msm/mdp5: Reset CTL blend registers before configuring them
  drm/msm/mdp5: Assign 'right' mixer to CRTC state
  drm/msm/mdp5: Stage border out on base stage if CRTC has 2 LMs
  drm/msm/mdp5: Stage right side hwpipes on Right-side Layer Mixer
  drm/msm/mdp5: Prepare Layer Mixers for source split
  drm/msm/mdp5: Configure 'right' hwpipe
  drm/msm/mdp5: Assign a 'right hwpipe' to plane state
  drm/msm/mdp5: Create mdp5_hwpipe_mode_set
  drm/msm/mdp5: Add optional 'right' Layer Mixer in CRTC state
  drm/msm/mdp5: Add a CAP for Source Split
  drm/msm/mdp5: Remove mixer/intf pointers from mdp5_ctl
  drm/msm/mdp5: Start using parameters from CRTC state
  drm/msm/mdp5: Add more stuff to CRTC state
  ...

35 files changed, 1569 insertions, 622 deletions

diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile
index 39055362da95..5241ac8803ba 100644
--- a/drivers/gpu/drm/msm/Makefile
+++ b/drivers/gpu/drm/msm/Makefile
@@ -40,6 +40,7 @@ msm-y := \
 	mdp/mdp5/mdp5_mdss.o \
 	mdp/mdp5/mdp5_kms.o \
 	mdp/mdp5/mdp5_pipe.o \
+	mdp/mdp5/mdp5_mixer.o \
 	mdp/mdp5/mdp5_plane.o \
 	mdp/mdp5/mdp5_smp.o \
 	msm_atomic.o \
diff --git a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
index b999349b7d2d..7fd77958a436 100644
--- a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
@@ -412,10 +412,8 @@ static const unsigned int a3xx_registers[] = {
 #ifdef CONFIG_DEBUG_FS
 static void a3xx_show(struct msm_gpu *gpu, struct seq_file *m)
 {
-	gpu->funcs->pm_resume(gpu);
 	seq_printf(m, "status:   %08x\n",
 			gpu_read(gpu, REG_A3XX_RBBM_STATUS));
-	gpu->funcs->pm_suspend(gpu);
 	adreno_show(gpu, m);
 }
 #endif
diff --git a/drivers/gpu/drm/msm/adreno/a4xx_gpu.c b/drivers/gpu/drm/msm/adreno/a4xx_gpu.c
index 511bc855cc7f..dfe0eceaae3b 100644
--- a/drivers/gpu/drm/msm/adreno/a4xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a4xx_gpu.c
@@ -456,12 +456,8 @@ static const unsigned int a4xx_registers[] = {
 #ifdef CONFIG_DEBUG_FS
 static void a4xx_show(struct msm_gpu *gpu, struct seq_file *m)
 {
-	gpu->funcs->pm_resume(gpu);
-
 	seq_printf(m, "status:   %08x\n",
 			gpu_read(gpu, REG_A4XX_RBBM_STATUS));
-	gpu->funcs->pm_suspend(gpu);
-
 	adreno_show(gpu, m);
 
 }
diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index 36602ac7e248..31a9bceed32c 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -638,10 +638,8 @@ static void a5xx_cp_err_irq(struct msm_gpu *gpu)
 	}
 }
 
-static void a5xx_rbbm_err_irq(struct msm_gpu *gpu)
+static void a5xx_rbbm_err_irq(struct msm_gpu *gpu, u32 status)
 {
-	u32 status = gpu_read(gpu, REG_A5XX_RBBM_INT_0_STATUS);
-
 	if (status & A5XX_RBBM_INT_0_MASK_RBBM_AHB_ERROR) {
 		u32 val = gpu_read(gpu, REG_A5XX_RBBM_AHB_ERROR_STATUS);
 
@@ -653,6 +651,10 @@ static void a5xx_rbbm_err_irq(struct msm_gpu *gpu)
 
 		/* Clear the error */
 		gpu_write(gpu, REG_A5XX_RBBM_AHB_CMD, (1 << 4));
+
+		/* Clear the interrupt */
+		gpu_write(gpu, REG_A5XX_RBBM_INT_CLEAR_CMD,
+			A5XX_RBBM_INT_0_MASK_RBBM_AHB_ERROR);
 	}
 
 	if (status & A5XX_RBBM_INT_0_MASK_RBBM_TRANSFER_TIMEOUT)
@@ -704,10 +706,16 @@ static irqreturn_t a5xx_irq(struct msm_gpu *gpu)
 {
 	u32 status = gpu_read(gpu, REG_A5XX_RBBM_INT_0_STATUS);
 
-	gpu_write(gpu, REG_A5XX_RBBM_INT_CLEAR_CMD, status);
+	/*
+	 * Clear all the interrupts except RBBM_AHB_ERROR - if we clear it
+	 * before the source is cleared the interrupt will storm.
+	 */
+	gpu_write(gpu, REG_A5XX_RBBM_INT_CLEAR_CMD,
+		status & ~A5XX_RBBM_INT_0_MASK_RBBM_AHB_ERROR);
 
+	/* Pass status to a5xx_rbbm_err_irq because we've already cleared it */
 	if (status & RBBM_ERROR_MASK)
-		a5xx_rbbm_err_irq(gpu);
+		a5xx_rbbm_err_irq(gpu, status);
 
 	if (status & A5XX_RBBM_INT_0_MASK_CP_HW_ERROR)
 		a5xx_cp_err_irq(gpu);
@@ -837,12 +845,8 @@ static int a5xx_get_timestamp(struct msm_gpu *gpu, uint64_t *value)
 #ifdef CONFIG_DEBUG_FS
 static void a5xx_show(struct msm_gpu *gpu, struct seq_file *m)
 {
-	gpu->funcs->pm_resume(gpu);
-
 	seq_printf(m, "status:   %08x\n",
 			gpu_read(gpu, REG_A5XX_RBBM_STATUS));
-	gpu->funcs->pm_suspend(gpu);
-
 	adreno_show(gpu, m);
 }
 #endif
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index ece39b16a864..c0fa5d1c75ff 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -2,7 +2,7 @@
  * Copyright (C) 2013-2014 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
- * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ * Copyright (c) 2014,2017 The Linux Foundation. All rights reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published by
@@ -17,6 +17,7 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <linux/pm_opp.h>
 #include "adreno_gpu.h"
 
 #define ANY_ID 0xff
@@ -155,21 +156,14 @@ struct msm_gpu *adreno_load_gpu(struct drm_device *dev)
 
 	if (gpu) {
 		int ret;
-		mutex_lock(&dev->struct_mutex);
-		gpu->funcs->pm_resume(gpu);
-		mutex_unlock(&dev->struct_mutex);
 
-		disable_irq(gpu->irq);
-
-		ret = gpu->funcs->hw_init(gpu);
+		pm_runtime_get_sync(&pdev->dev);
+		ret = msm_gpu_hw_init(gpu);
+		pm_runtime_put_sync(&pdev->dev);
 		if (ret) {
 			dev_err(dev->dev, "gpu hw init failed: %d\n", ret);
 			gpu->funcs->destroy(gpu);
 			gpu = NULL;
-		} else {
-			enable_irq(gpu->irq);
-			/* give inactive pm a chance to kick in: */
-			msm_gpu_retire(gpu);
 		}
 	}
 
@@ -220,10 +214,71 @@ static int find_chipid(struct device *dev, u32 *chipid)
 	return 0;
 }
 
+/* Get legacy powerlevels from qcom,gpu-pwrlevels and populate the opp table */
+static int adreno_get_legacy_pwrlevels(struct device *dev)
+{
+	struct device_node *child, *node;
+	int ret;
+
+	node = of_find_compatible_node(dev->of_node, NULL,
+		"qcom,gpu-pwrlevels");
+	if (!node) {
+		dev_err(dev, "Could not find the GPU powerlevels\n");
+		return -ENXIO;
+	}
+
+	for_each_child_of_node(node, child) {
+		unsigned int val;
+
+		ret = of_property_read_u32(child, "qcom,gpu-freq", &val);
+		if (ret)
+			continue;
+
+		/*
+		 * Skip the intentionally bogus clock value found at the bottom
+		 * of most legacy frequency tables
+		 */
+		if (val != 27000000)
+			dev_pm_opp_add(dev, val, 0);
+	}
+
+	return 0;
+}
+
+static int adreno_get_pwrlevels(struct device *dev,
+		struct adreno_platform_config *config)
+{
+	unsigned long freq = ULONG_MAX;
+	struct dev_pm_opp *opp;
+	int ret;
+
+	/* You down with OPP? */
+	if (!of_find_property(dev->of_node, "operating-points-v2", NULL))
+		ret = adreno_get_legacy_pwrlevels(dev);
+	else
+		ret = dev_pm_opp_of_add_table(dev);
+
+	if (ret)
+		return ret;
+
+	/* Find the fastest defined rate */
+	opp = dev_pm_opp_find_freq_floor(dev, &freq);
+	if (!IS_ERR(opp))
+		config->fast_rate = dev_pm_opp_get_freq(opp);
+
+	if (!config->fast_rate) {
+		DRM_DEV_INFO(dev,
+			"Could not find clock rate. Using default\n");
+		/* Pick a suitably safe clock speed for any target */
+		config->fast_rate = 200000000;
+	}
+
+	return 0;
+}
+
 static int adreno_bind(struct device *dev, struct device *master, void *data)
 {
 	static struct adreno_platform_config config = {};
-	struct device_node *child, *node = dev->of_node;
 	u32 val;
 	int ret;
 
@@ -238,28 +293,10 @@ static int adreno_bind(struct device *dev, struct device *master, void *data)
 
 	/* find clock rates: */
 	config.fast_rate = 0;
-	config.slow_rate = ~0;
-	for_each_child_of_node(node, child) {
-		if (of_device_is_compatible(child, "qcom,gpu-pwrlevels")) {
-			struct device_node *pwrlvl;
-			for_each_child_of_node(child, pwrlvl) {
-				ret = of_property_read_u32(pwrlvl, "qcom,gpu-freq", &val);
-				if (ret) {
-					dev_err(dev, "could not find gpu-freq: %d\n", ret);
-					return ret;
-				}
-				config.fast_rate = max(config.fast_rate, val);
-				config.slow_rate = min(config.slow_rate, val);
-			}
-		}
-	}
 
-	if (!config.fast_rate) {
-		dev_warn(dev, "could not find clk rates\n");
-		/* This is a safe low speed for all devices: */
-		config.fast_rate = 200000000;
-		config.slow_rate = 27000000;
-	}
+	ret = adreno_get_pwrlevels(dev, &config);
+	if (ret)
+		return ret;
 
 	dev->platform_data = &config;
 	set_gpu_pdev(dev_get_drvdata(master), to_platform_device(dev));
@@ -296,12 +333,35 @@ static const struct of_device_id dt_match[] = {
 	{}
 };
 
+#ifdef CONFIG_PM
+static int adreno_resume(struct device *dev)
+{
+	struct platform_device *pdev = to_platform_device(dev);
+	struct msm_gpu *gpu = platform_get_drvdata(pdev);
+
+	return gpu->funcs->pm_resume(gpu);
+}
+
+static int adreno_suspend(struct device *dev)
+{
+	struct platform_device *pdev = to_platform_device(dev);
+	struct msm_gpu *gpu = platform_get_drvdata(pdev);
+
+	return gpu->funcs->pm_suspend(gpu);
+}
+#endif
+
+static const struct dev_pm_ops adreno_pm_ops = {
+	SET_RUNTIME_PM_OPS(adreno_suspend, adreno_resume, NULL)
+};
+
 static struct platform_driver adreno_driver = {
 	.probe = adreno_probe,
 	.remove = adreno_remove,
 	.driver = {
 		.name = "adreno",
 		.of_match_table = dt_match,
+		.pm = &adreno_pm_ops,
 	},
 };
 
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index 5ae65426b4e5..5b63fc649dcc 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -35,6 +35,9 @@ int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value)
 	case MSM_PARAM_GMEM_SIZE:
 		*value = adreno_gpu->gmem;
 		return 0;
+	case MSM_PARAM_GMEM_BASE:
+		*value = 0x100000;
+		return 0;
 	case MSM_PARAM_CHIP_ID:
 		*value = adreno_gpu->rev.patchid |
 				(adreno_gpu->rev.minor << 8) |
@@ -68,6 +71,14 @@ int adreno_hw_init(struct msm_gpu *gpu)
 		return ret;
 	}
 
+	/* reset ringbuffer: */
+	gpu->rb->cur = gpu->rb->start;
+
+	/* reset completed fence seqno: */
+	adreno_gpu->memptrs->fence = gpu->fctx->completed_fence;
+	adreno_gpu->memptrs->rptr  = 0;
+	adreno_gpu->memptrs->wptr  = 0;
+
 	/* Setup REG_CP_RB_CNTL: */
 	adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_CNTL,
 			/* size is log2(quad-words): */
@@ -111,29 +122,20 @@ uint32_t adreno_last_fence(struct msm_gpu *gpu)
 
 void adreno_recover(struct msm_gpu *gpu)
 {
-	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
 	struct drm_device *dev = gpu->dev;
 	int ret;
 
-	gpu->funcs->pm_suspend(gpu);
-
-	/* reset ringbuffer: */
-	gpu->rb->cur = gpu->rb->start;
-
-	/* reset completed fence seqno: */
-	adreno_gpu->memptrs->fence = gpu->fctx->completed_fence;
-	adreno_gpu->memptrs->rptr  = 0;
-	adreno_gpu->memptrs->wptr  = 0;
+	// XXX pm-runtime??  we *need* the device to be off after this
+	// so maybe continuing to call ->pm_suspend/resume() is better?
 
+	gpu->funcs->pm_suspend(gpu);
 	gpu->funcs->pm_resume(gpu);
 
-	disable_irq(gpu->irq);
-	ret = gpu->funcs->hw_init(gpu);
+	ret = msm_gpu_hw_init(gpu);
 	if (ret) {
 		dev_err(dev->dev, "gpu hw init failed: %d\n", ret);
 		/* hmm, oh well? */
 	}
-	enable_irq(gpu->irq);
 }
 
 void adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
@@ -259,8 +261,6 @@ void adreno_show(struct msm_gpu *gpu, struct seq_file *m)
 	seq_printf(m, "wptr:     %d\n", adreno_gpu->memptrs->wptr);
 	seq_printf(m, "rb wptr:  %d\n", get_wptr(gpu->rb));
 
-	gpu->funcs->pm_resume(gpu);
-
 	/* dump these out in a form that can be parsed by demsm: */
 	seq_printf(m, "IO:region %s 00000000 00020000\n", gpu->name);
 	for (i = 0; adreno_gpu->registers[i] != ~0; i += 2) {
@@ -273,8 +273,6 @@ void adreno_show(struct msm_gpu *gpu, struct seq_file *m)
 			seq_printf(m, "IO:R %08x %08x\n", addr<<2, val);
 		}
 	}
-
-	gpu->funcs->pm_suspend(gpu);
 }
 #endif
 
@@ -354,14 +352,13 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	adreno_gpu->rev = config->rev;
 
 	gpu->fast_rate = config->fast_rate;
-	gpu->slow_rate = config->slow_rate;
 	gpu->bus_freq  = config->bus_freq;
 #ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
 	gpu->bus_scale_table = config->bus_scale_table;
 #endif
 
-	DBG("fast_rate=%u, slow_rate=%u, bus_freq=%u",
-			gpu->fast_rate, gpu->slow_rate, gpu->bus_freq);
+	DBG("fast_rate=%u, slow_rate=27000000, bus_freq=%u",
+			gpu->fast_rate, gpu->bus_freq);
 
 	ret = msm_gpu_init(drm, pdev, &adreno_gpu->base, &funcs->base,
 			adreno_gpu->info->name, "kgsl_3d0_reg_memory", "kgsl_3d0_irq",
@@ -369,6 +366,10 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	if (ret)
 		return ret;
 
+	pm_runtime_set_autosuspend_delay(&pdev->dev, DRM_MSM_INACTIVE_PERIOD);
+	pm_runtime_use_autosuspend(&pdev->dev);
+	pm_runtime_enable(&pdev->dev);
+
 	ret = request_firmware(&adreno_gpu->pm4, adreno_gpu->info->pm4fw, drm->dev);
 	if (ret) {
 		dev_err(drm->dev, "failed to load %s PM4 firmware: %d\n",
@@ -439,6 +440,6 @@ void adreno_gpu_cleanup(struct adreno_gpu *adreno_gpu)
 	if (gpu->aspace) {
 		gpu->aspace->mmu->funcs->detach(gpu->aspace->mmu,
 			iommu_ports, ARRAY_SIZE(iommu_ports));
-		msm_gem_address_space_destroy(gpu->aspace);
+		msm_gem_address_space_put(gpu->aspace);
 	}
 }
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
index 42e444a67630..fb4831f9f80b 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
@@ -123,7 +123,7 @@ struct adreno_gpu {
 /* platform config data (ie. from DT, or pdata) */
 struct adreno_platform_config {
 	struct adreno_rev rev;
-	uint32_t fast_rate, slow_rate, bus_freq;
+	uint32_t fast_rate, bus_freq;
 #ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
 	struct msm_bus_scale_pdata *bus_scale_table;
 #endif
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
index 1c29618f4ddb..f29194a74a19 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
@@ -114,15 +114,9 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
 	spin_lock_irqsave(&dev->event_lock, flags);
 	event = mdp4_crtc->event;
 	if (event) {
-		/* if regular vblank case (!file) or if cancel-flip from
-		 * preclose on file that requested flip, then send the
-		 * event:
-		 */
-		if (!file || (event->base.file_priv == file)) {
-			mdp4_crtc->event = NULL;
-			DBG("%s: send event: %p", mdp4_crtc->name, event);
-			drm_crtc_send_vblank_event(crtc, event);
-		}
+		mdp4_crtc->event = NULL;
+		DBG("%s: send event: %p", mdp4_crtc->name, event);
+		drm_crtc_send_vblank_event(crtc, event);
 	}
 	spin_unlock_irqrestore(&dev->event_lock, flags);
 }
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c
index 5b6516bb9d06..3d26d7774c08 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c
@@ -169,7 +169,7 @@ static void mdp4_destroy(struct msm_kms *kms)
 	if (aspace) {
 		aspace->mmu->funcs->detach(aspace->mmu,
 				iommu_ports, ARRAY_SIZE(iommu_ports));
-		msm_gem_address_space_destroy(aspace);
+		msm_gem_address_space_put(aspace);
 	}
 
 	if (mdp4_kms->rpm_enabled)
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c
index ba2d017f6591..c2bdad88447e 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c
@@ -70,6 +70,18 @@ const struct mdp5_cfg_hw msm8x74v1_config = {
 	.lm = {
 		.count = 5,
 		.base = { 0x03100, 0x03500, 0x03900, 0x03d00, 0x04100 },
+		.instances = {
+				{ .id = 0, .pp = 0, .dspp = 0,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 1, .pp = 1, .dspp = 1,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 2, .pp = 2, .dspp = 2,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 3, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB },
+				{ .id = 4, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB },
+			     },
 		.nb_stages = 5,
 	},
 	.dspp = {
@@ -134,6 +146,18 @@ const struct mdp5_cfg_hw msm8x74v2_config = {
 	.lm = {
 		.count = 5,
 		.base = { 0x03100, 0x03500, 0x03900, 0x03d00, 0x04100 },
+		.instances = {
+				{ .id = 0, .pp = 0, .dspp = 0,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 1, .pp = 1, .dspp = 1,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 2, .pp = 2, .dspp = 2,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 3, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 4, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+			     },
 		.nb_stages = 5,
 		.max_width = 2048,
 		.max_height = 0xFFFF,
@@ -167,6 +191,7 @@ const struct mdp5_cfg_hw apq8084_config = {
 	.mdp = {
 		.count = 1,
 		.caps = MDP_CAP_SMP |
+			MDP_CAP_SRC_SPLIT |
 			0,
 	},
 	.smp = {
@@ -211,6 +236,22 @@ const struct mdp5_cfg_hw apq8084_config = {
 	.lm = {
 		.count = 6,
 		.base = { 0x03900, 0x03d00, 0x04100, 0x04500, 0x04900, 0x04d00 },
+		.instances = {
+				{ .id = 0, .pp = 0, .dspp = 0,
+				  .caps = MDP_LM_CAP_DISPLAY |
+					  MDP_LM_CAP_PAIR, },
+				{ .id = 1, .pp = 1, .dspp = 1,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 2, .pp = 2, .dspp = 2,
+				  .caps = MDP_LM_CAP_DISPLAY |
+					  MDP_LM_CAP_PAIR, },
+				{ .id = 3, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 4, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 5, .pp = 3, .dspp = 3,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+			     },
 		.nb_stages = 5,
 		.max_width = 2048,
 		.max_height = 0xFFFF,
@@ -282,6 +323,12 @@ const struct mdp5_cfg_hw msm8x16_config = {
 	.lm = {
 		.count = 2, /* LM0 and LM3 */
 		.base = { 0x44000, 0x47000 },
+		.instances = {
+				{ .id = 0, .pp = 0, .dspp = 0,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 3, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB },
+			     },
 		.nb_stages = 8,
 		.max_width = 2048,
 		.max_height = 0xFFFF,
@@ -306,6 +353,7 @@ const struct mdp5_cfg_hw msm8x94_config = {
 	.mdp = {
 		.count = 1,
 		.caps = MDP_CAP_SMP |
+			MDP_CAP_SRC_SPLIT |
 			0,
 	},
 	.smp = {
@@ -350,6 +398,22 @@ const struct mdp5_cfg_hw msm8x94_config = {
 	.lm = {
 		.count = 6,
 		.base = { 0x44000, 0x45000, 0x46000, 0x47000, 0x48000, 0x49000 },
+		.instances = {
+				{ .id = 0, .pp = 0, .dspp = 0,
+				  .caps = MDP_LM_CAP_DISPLAY |
+					  MDP_LM_CAP_PAIR, },
+				{ .id = 1, .pp = 1, .dspp = 1,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 2, .pp = 2, .dspp = 2,
+				  .caps = MDP_LM_CAP_DISPLAY |
+					  MDP_LM_CAP_PAIR, },
+				{ .id = 3, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 4, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 5, .pp = 3, .dspp = 3,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+			     },
 		.nb_stages = 8,
 		.max_width = 2048,
 		.max_height = 0xFFFF,
@@ -385,6 +449,7 @@ const struct mdp5_cfg_hw msm8x96_config = {
 		.count = 1,
 		.caps = MDP_CAP_DSC |
 			MDP_CAP_CDM |
+			MDP_CAP_SRC_SPLIT |
 			0,
 	},
 	.ctl = {
@@ -434,6 +499,22 @@ const struct mdp5_cfg_hw msm8x96_config = {
 	.lm = {
 		.count = 6,
 		.base = { 0x44000, 0x45000, 0x46000, 0x47000, 0x48000, 0x49000 },
+		.instances = {
+				{ .id = 0, .pp = 0, .dspp = 0,
+				  .caps = MDP_LM_CAP_DISPLAY |
+					  MDP_LM_CAP_PAIR, },
+				{ .id = 1, .pp = 1, .dspp = 1,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+				{ .id = 2, .pp = 2, .dspp = -1,
+				  .caps = MDP_LM_CAP_DISPLAY |
+					  MDP_LM_CAP_PAIR, },
+				{ .id = 3, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 4, .pp = -1, .dspp = -1,
+				  .caps = MDP_LM_CAP_WB, },
+				{ .id = 5, .pp = 3, .dspp = -1,
+				  .caps = MDP_LM_CAP_DISPLAY, },
+			     },
 		.nb_stages = 8,
 		.max_width = 2560,
 		.max_height = 0xFFFF,
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h
index b1c7daaede86..75910d0f2f4c 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h
@@ -39,8 +39,16 @@ struct mdp5_sub_block {
 	MDP5_SUB_BLOCK_DEFINITION;
 };
 
+struct mdp5_lm_instance {
+	int id;
+	int pp;
+	int dspp;
+	uint32_t caps;
+};
+
 struct mdp5_lm_block {
 	MDP5_SUB_BLOCK_DEFINITION;
+	struct mdp5_lm_instance instances[MAX_BASES];
 	uint32_t nb_stages;		/* number of stages per blender */
 	uint32_t max_width;		/* Maximum output resolution */
 	uint32_t max_height;
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cmd_encoder.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cmd_encoder.c
index df1c8adec3f3..8dafc7bdba48 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cmd_encoder.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cmd_encoder.c
@@ -51,7 +51,8 @@ static int pingpong_tearcheck_setup(struct drm_encoder *encoder,
 	struct device *dev = encoder->dev->dev;
 	u32 total_lines_x100, vclks_line, cfg;
 	long vsync_clk_speed;
-	int pp_id = GET_PING_PONG_ID(mdp5_crtc_get_lm(encoder->crtc));
+	struct mdp5_hw_mixer *mixer = mdp5_crtc_get_mixer(encoder->crtc);
+	int pp_id = mixer->pp;
 
 	if (IS_ERR_OR_NULL(mdp5_kms->vsync_clk)) {
 		dev_err(dev, "vsync_clk is not initialized\n");
@@ -94,7 +95,8 @@ static int pingpong_tearcheck_setup(struct drm_encoder *encoder,
 static int pingpong_tearcheck_enable(struct drm_encoder *encoder)
 {
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
-	int pp_id = GET_PING_PONG_ID(mdp5_crtc_get_lm(encoder->crtc));
+	struct mdp5_hw_mixer *mixer = mdp5_crtc_get_mixer(encoder->crtc);
+	int pp_id = mixer->pp;
 	int ret;
 
 	ret = clk_set_rate(mdp5_kms->vsync_clk,
@@ -119,7 +121,8 @@ static int pingpong_tearcheck_enable(struct drm_encoder *encoder)
 static void pingpong_tearcheck_disable(struct drm_encoder *encoder)
 {
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
-	int pp_id = GET_PING_PONG_ID(mdp5_crtc_get_lm(encoder->crtc));
+	struct mdp5_hw_mixer *mixer = mdp5_crtc_get_mixer(encoder->crtc);
+	int pp_id = mixer->pp;
 
 	mdp5_write(mdp5_kms, REG_MDP5_PP_TEAR_CHECK_EN(pp_id), 0);
 	clk_disable_unprepare(mdp5_kms->vsync_clk);
@@ -129,8 +132,6 @@ void mdp5_cmd_encoder_mode_set(struct drm_encoder *encoder,
 			       struct drm_display_mode *mode,
 			       struct drm_display_mode *adjusted_mode)
 {
-	struct mdp5_encoder *mdp5_cmd_enc = to_mdp5_encoder(encoder);
-
 	mode = adjusted_mode;
 
 	DBG("set mode: %d:\"%s\" %d %d %d %d %d %d %d %d %d %d 0x%x 0x%x",
@@ -142,23 +143,23 @@ void mdp5_cmd_encoder_mode_set(struct drm_encoder *encoder,
 			mode->vsync_end, mode->vtotal,
 			mode->type, mode->flags);
 	pingpong_tearcheck_setup(encoder, mode);
-	mdp5_crtc_set_pipeline(encoder->crtc, &mdp5_cmd_enc->intf,
-				mdp5_cmd_enc->ctl);
+	mdp5_crtc_set_pipeline(encoder->crtc);
 }
 
 void mdp5_cmd_encoder_disable(struct drm_encoder *encoder)
 {
 	struct mdp5_encoder *mdp5_cmd_enc = to_mdp5_encoder(encoder);
 	struct mdp5_ctl *ctl = mdp5_cmd_enc->ctl;
-	struct mdp5_interface *intf = &mdp5_cmd_enc->intf;
+	struct mdp5_interface *intf = mdp5_cmd_enc->intf;
+	struct mdp5_pipeline *pipeline = mdp5_crtc_get_pipeline(encoder->crtc);
 
 	if (WARN_ON(!mdp5_cmd_enc->enabled))
 		return;
 
 	pingpong_tearcheck_disable(encoder);
 
-	mdp5_ctl_set_encoder_state(ctl, false);
-	mdp5_ctl_commit(ctl, mdp_ctl_flush_mask_encoder(intf));
+	mdp5_ctl_set_encoder_state(ctl, pipeline, false);
+	mdp5_ctl_commit(ctl, pipeline, mdp_ctl_flush_mask_encoder(intf));
 
 	bs_set(mdp5_cmd_enc, 0);
 
@@ -169,7 +170,8 @@ void mdp5_cmd_encoder_enable(struct drm_encoder *encoder)
 {
 	struct mdp5_encoder *mdp5_cmd_enc = to_mdp5_encoder(encoder);
 	struct mdp5_ctl *ctl = mdp5_cmd_enc->ctl;
-	struct mdp5_interface *intf = &mdp5_cmd_enc->intf;
+	struct mdp5_interface *intf = mdp5_cmd_enc->intf;
+	struct mdp5_pipeline *pipeline = mdp5_crtc_get_pipeline(encoder->crtc);
 
 	if (WARN_ON(mdp5_cmd_enc->enabled))
 		return;
@@ -178,9 +180,9 @@ void mdp5_cmd_encoder_enable(struct drm_encoder *encoder)
 	if (pingpong_tearcheck_enable(encoder))
 		return;
 
-	mdp5_ctl_commit(ctl, mdp_ctl_flush_mask_encoder(intf));
+	mdp5_ctl_commit(ctl, pipeline, mdp_ctl_flush_mask_encoder(intf));
 
-	mdp5_ctl_set_encoder_state(ctl, true);
+	mdp5_ctl_set_encoder_state(ctl, pipeline, true);
 
 	mdp5_cmd_enc->enabled = true;
 }
@@ -197,7 +199,7 @@ int mdp5_cmd_encoder_set_split_display(struct drm_encoder *encoder,
 		return -EINVAL;
 
 	mdp5_kms = get_kms(encoder);
-	intf_num = mdp5_cmd_enc->intf.num;
+	intf_num = mdp5_cmd_enc->intf->num;
 
 	/* Switch slave encoder's trigger MUX, to use the master's
 	 * start signal for the slave encoder
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
index d0c8b38b96ce..9217e0d6e93e 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
@@ -32,13 +32,7 @@ struct mdp5_crtc {
 	int id;
 	bool enabled;
 
-	/* layer mixer used for this CRTC (+ its lock): */
-#define GET_LM_ID(crtc_id)	((crtc_id == 3) ? 5 : crtc_id)
-	int lm;
-	spinlock_t lm_lock;	/* protect REG_MDP5_LM_* registers */
-
-	/* CTL used for this CRTC: */
-	struct mdp5_ctl *ctl;
+	spinlock_t lm_lock;     /* protect REG_MDP5_LM_* registers */
 
 	/* if there is a pending flip, these will be non-null: */
 	struct drm_pending_vblank_event *event;
@@ -61,8 +55,6 @@ struct mdp5_crtc {
 
 	struct completion pp_completion;
 
-	bool cmd_mode;
-
 	struct {
 		/* protect REG_MDP5_LM_CURSOR* registers and cursor scanout_bo*/
 		spinlock_t lock;
@@ -97,10 +89,12 @@ static void request_pp_done_pending(struct drm_crtc *crtc)
 
 static u32 crtc_flush(struct drm_crtc *crtc, u32 flush_mask)
 {
-	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	struct mdp5_ctl *ctl = mdp5_cstate->ctl;
+	struct mdp5_pipeline *pipeline = &mdp5_cstate->pipeline;
 
 	DBG("%s: flush=%08x", crtc->name, flush_mask);
-	return mdp5_ctl_commit(mdp5_crtc->ctl, flush_mask);
+	return mdp5_ctl_commit(ctl, pipeline, flush_mask);
 }
 
 /*
@@ -110,19 +104,25 @@ static u32 crtc_flush(struct drm_crtc *crtc, u32 flush_mask)
  */
 static u32 crtc_flush_all(struct drm_crtc *crtc)
 {
-	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	struct mdp5_hw_mixer *mixer, *r_mixer;
 	struct drm_plane *plane;
 	uint32_t flush_mask = 0;
 
 	/* this should not happen: */
-	if (WARN_ON(!mdp5_crtc->ctl))
+	if (WARN_ON(!mdp5_cstate->ctl))
 		return 0;
 
 	drm_atomic_crtc_for_each_plane(plane, crtc) {
 		flush_mask |= mdp5_plane_get_flush(plane);
 	}
 
-	flush_mask |= mdp_ctl_flush_mask_lm(mdp5_crtc->lm);
+	mixer = mdp5_cstate->pipeline.mixer;
+	flush_mask |= mdp_ctl_flush_mask_lm(mixer->lm);
+
+	r_mixer = mdp5_cstate->pipeline.r_mixer;
+	if (r_mixer)
+		flush_mask |= mdp_ctl_flush_mask_lm(r_mixer->lm);
 
 	return crtc_flush(crtc, flush_mask);
 }
@@ -130,7 +130,10 @@ static u32 crtc_flush_all(struct drm_crtc *crtc)
 /* if file!=NULL, this is preclose potential cancel-flip path */
 static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
 {
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	struct mdp5_pipeline *pipeline = &mdp5_cstate->pipeline;
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_ctl *ctl = mdp5_cstate->ctl;
 	struct drm_device *dev = crtc->dev;
 	struct drm_pending_vblank_event *event;
 	unsigned long flags;
@@ -138,22 +141,17 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
 	spin_lock_irqsave(&dev->event_lock, flags);
 	event = mdp5_crtc->event;
 	if (event) {
-		/* if regular vblank case (!file) or if cancel-flip from
-		 * preclose on file that requested flip, then send the
-		 * event:
-		 */
-		if (!file || (event->base.file_priv == file)) {
-			mdp5_crtc->event = NULL;
-			DBG("%s: send event: %p", crtc->name, event);
-			drm_crtc_send_vblank_event(crtc, event);
-		}
+		mdp5_crtc->event = NULL;
+		DBG("%s: send event: %p", crtc->name, event);
+		drm_crtc_send_vblank_event(crtc, event);
 	}
 	spin_unlock_irqrestore(&dev->event_lock, flags);
 
-	if (mdp5_crtc->ctl && !crtc->state->enable) {
+	if (ctl && !crtc->state->enable) {
 		/* set STAGE_UNUSED for all layers */
-		mdp5_ctl_blend(mdp5_crtc->ctl, NULL, 0, 0);
-		mdp5_crtc->ctl = NULL;
+		mdp5_ctl_blend(ctl, pipeline, NULL, NULL, 0, 0);
+		/* XXX: What to do here? */
+		/* mdp5_crtc->ctl = NULL; */
 	}
 }
 
@@ -193,6 +191,12 @@ static inline u32 mdp5_lm_use_fg_alpha_mask(enum mdp_mixer_stage_id stage)
 }
 
 /*
+ * left/right pipe offsets for the stage array used in blend_setup()
+ */
+#define PIPE_LEFT	0
+#define PIPE_RIGHT	1
+
+/*
  * blend_setup() - blend all the planes of a CRTC
  *
  * If no base layer is available, border will be enabled as the base layer.
@@ -202,18 +206,26 @@ static inline u32 mdp5_lm_use_fg_alpha_mask(enum mdp_mixer_stage_id stage)
 static void blend_setup(struct drm_crtc *crtc)
 {
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	struct mdp5_pipeline *pipeline = &mdp5_cstate->pipeline;
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
 	struct drm_plane *plane;
 	const struct mdp5_cfg_hw *hw_cfg;
 	struct mdp5_plane_state *pstate, *pstates[STAGE_MAX + 1] = {NULL};
 	const struct mdp_format *format;
-	uint32_t lm = mdp5_crtc->lm;
+	struct mdp5_hw_mixer *mixer = pipeline->mixer;
+	uint32_t lm = mixer->lm;
+	struct mdp5_hw_mixer *r_mixer = pipeline->r_mixer;
+	uint32_t r_lm = r_mixer ? r_mixer->lm : 0;
+	struct mdp5_ctl *ctl = mdp5_cstate->ctl;
 	uint32_t blend_op, fg_alpha, bg_alpha, ctl_blend_flags = 0;
 	unsigned long flags;
-	enum mdp5_pipe stage[STAGE_MAX + 1] = { SSPP_NONE };
+	enum mdp5_pipe stage[STAGE_MAX + 1][MAX_PIPE_STAGE] = { SSPP_NONE };
+	enum mdp5_pipe r_stage[STAGE_MAX + 1][MAX_PIPE_STAGE] = { SSPP_NONE };
 	int i, plane_cnt = 0;
 	bool bg_alpha_enabled = false;
 	u32 mixer_op_mode = 0;
+	u32 val;
 #define blender(stage)	((stage) - STAGE0)
 
 	hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
@@ -221,14 +233,35 @@ static void blend_setup(struct drm_crtc *crtc)
 	spin_lock_irqsave(&mdp5_crtc->lm_lock, flags);
 
 	/* ctl could be released already when we are shutting down: */
-	if (!mdp5_crtc->ctl)
+	/* XXX: Can this happen now? */
+	if (!ctl)
 		goto out;
 
 	/* Collect all plane information */
 	drm_atomic_crtc_for_each_plane(plane, crtc) {
+		enum mdp5_pipe right_pipe;
+
 		pstate = to_mdp5_plane_state(plane->state);
 		pstates[pstate->stage] = pstate;
-		stage[pstate->stage] = mdp5_plane_pipe(plane);
+		stage[pstate->stage][PIPE_LEFT] = mdp5_plane_pipe(plane);
+		/*
+		 * if we have a right mixer, stage the same pipe as we
+		 * have on the left mixer
+		 */
+		if (r_mixer)
+			r_stage[pstate->stage][PIPE_LEFT] =
+						mdp5_plane_pipe(plane);
+		/*
+		 * if we have a right pipe (i.e, the plane comprises of 2
+		 * hwpipes, then stage the right pipe on the right side of both
+		 * the layer mixers
+		 */
+		right_pipe = mdp5_plane_right_pipe(plane);
+		if (right_pipe) {
+			stage[pstate->stage][PIPE_RIGHT] = right_pipe;
+			r_stage[pstate->stage][PIPE_RIGHT] = right_pipe;
+		}
+
 		plane_cnt++;
 	}
 
@@ -294,12 +327,27 @@ static void blend_setup(struct drm_crtc *crtc)
 				blender(i)), fg_alpha);
 		mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_BG_ALPHA(lm,
 				blender(i)), bg_alpha);
+		if (r_mixer) {
+			mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_OP_MODE(r_lm,
+					blender(i)), blend_op);
+			mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_FG_ALPHA(r_lm,
+					blender(i)), fg_alpha);
+			mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_BG_ALPHA(r_lm,
+					blender(i)), bg_alpha);
+		}
 	}
 
-	mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(lm), mixer_op_mode);
-
-	mdp5_ctl_blend(mdp5_crtc->ctl, stage, plane_cnt, ctl_blend_flags);
+	val = mdp5_read(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(lm));
+	mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(lm),
+		   val | mixer_op_mode);
+	if (r_mixer) {
+		val = mdp5_read(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(r_lm));
+		mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(r_lm),
+			   val | mixer_op_mode);
+	}
 
+	mdp5_ctl_blend(ctl, pipeline, stage, r_stage, plane_cnt,
+		       ctl_blend_flags);
 out:
 	spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags);
 }
@@ -307,7 +355,12 @@ out:
 static void mdp5_crtc_mode_set_nofb(struct drm_crtc *crtc)
 {
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
+	struct mdp5_hw_mixer *mixer = mdp5_cstate->pipeline.mixer;
+	struct mdp5_hw_mixer *r_mixer = mdp5_cstate->pipeline.r_mixer;
+	uint32_t lm = mixer->lm;
+	u32 mixer_width, val;
 	unsigned long flags;
 	struct drm_display_mode *mode;
 
@@ -325,16 +378,40 @@ static void mdp5_crtc_mode_set_nofb(struct drm_crtc *crtc)
 			mode->vsync_end, mode->vtotal,
 			mode->type, mode->flags);
 
+	mixer_width = mode->hdisplay;
+	if (r_mixer)
+		mixer_width /= 2;
+
 	spin_lock_irqsave(&mdp5_crtc->lm_lock, flags);
-	mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(mdp5_crtc->lm),
-			MDP5_LM_OUT_SIZE_WIDTH(mode->hdisplay) |
+	mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(lm),
+			MDP5_LM_OUT_SIZE_WIDTH(mixer_width) |
 			MDP5_LM_OUT_SIZE_HEIGHT(mode->vdisplay));
+
+	/* Assign mixer to LEFT side in source split mode */
+	val = mdp5_read(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(lm));
+	val &= ~MDP5_LM_BLEND_COLOR_OUT_SPLIT_LEFT_RIGHT;
+	mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(lm), val);
+
+	if (r_mixer) {
+		u32 r_lm = r_mixer->lm;
+
+		mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(r_lm),
+			   MDP5_LM_OUT_SIZE_WIDTH(mixer_width) |
+			   MDP5_LM_OUT_SIZE_HEIGHT(mode->vdisplay));
+
+		/* Assign mixer to RIGHT side in source split mode */
+		val = mdp5_read(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(r_lm));
+		val |= MDP5_LM_BLEND_COLOR_OUT_SPLIT_LEFT_RIGHT;
+		mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(r_lm), val);
+	}
+
 	spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags);
 }
 
 static void mdp5_crtc_disable(struct drm_crtc *crtc)
 {
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
 
 	DBG("%s", crtc->name);
@@ -342,7 +419,7 @@ static void mdp5_crtc_disable(struct drm_crtc *crtc)
 	if (WARN_ON(!mdp5_crtc->enabled))
 		return;
 
-	if (mdp5_crtc->cmd_mode)
+	if (mdp5_cstate->cmd_mode)
 		mdp_irq_unregister(&mdp5_kms->base, &mdp5_crtc->pp_done);
 
 	mdp_irq_unregister(&mdp5_kms->base, &mdp5_crtc->err);
@@ -354,6 +431,7 @@ static void mdp5_crtc_disable(struct drm_crtc *crtc)
 static void mdp5_crtc_enable(struct drm_crtc *crtc)
 {
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
 
 	DBG("%s", crtc->name);
@@ -364,12 +442,73 @@ static void mdp5_crtc_enable(struct drm_crtc *crtc)
 	mdp5_enable(mdp5_kms);
 	mdp_irq_register(&mdp5_kms->base, &mdp5_crtc->err);
 
-	if (mdp5_crtc->cmd_mode)
+	if (mdp5_cstate->cmd_mode)
 		mdp_irq_register(&mdp5_kms->base, &mdp5_crtc->pp_done);
 
 	mdp5_crtc->enabled = true;
 }
 
+int mdp5_crtc_setup_pipeline(struct drm_crtc *crtc,
+			     struct drm_crtc_state *new_crtc_state,
+			     bool need_right_mixer)
+{
+	struct mdp5_crtc_state *mdp5_cstate =
+			to_mdp5_crtc_state(new_crtc_state);
+	struct mdp5_pipeline *pipeline = &mdp5_cstate->pipeline;
+	struct mdp5_interface *intf;
+	bool new_mixer = false;
+
+	new_mixer = !pipeline->mixer;
+
+	if ((need_right_mixer && !pipeline->r_mixer) ||
+	    (!need_right_mixer && pipeline->r_mixer))
+		new_mixer = true;
+
+	if (new_mixer) {
+		struct mdp5_hw_mixer *old_mixer = pipeline->mixer;
+		struct mdp5_hw_mixer *old_r_mixer = pipeline->r_mixer;
+		u32 caps;
+		int ret;
+
+		caps = MDP_LM_CAP_DISPLAY;
+		if (need_right_mixer)
+			caps |= MDP_LM_CAP_PAIR;
+
+		ret = mdp5_mixer_assign(new_crtc_state->state, crtc, caps,
+					&pipeline->mixer, need_right_mixer ?
+					&pipeline->r_mixer : NULL);
+		if (ret)
+			return ret;
+
+		mdp5_mixer_release(new_crtc_state->state, old_mixer);
+		if (old_r_mixer) {
+			mdp5_mixer_release(new_crtc_state->state, old_r_mixer);
+			if (!need_right_mixer)
+				pipeline->r_mixer = NULL;
+		}
+	}
+
+	/*
+	 * these should have been already set up in the encoder's atomic
+	 * check (called by drm_atomic_helper_check_modeset)
+	 */
+	intf = pipeline->intf;
+
+	mdp5_cstate->err_irqmask = intf2err(intf->num);
+	mdp5_cstate->vblank_irqmask = intf2vblank(pipeline->mixer, intf);
+
+	if ((intf->type == INTF_DSI) &&
+	    (intf->mode == MDP5_INTF_DSI_MODE_COMMAND)) {
+		mdp5_cstate->pp_done_irqmask = lm2ppdone(pipeline->mixer);
+		mdp5_cstate->cmd_mode = true;
+	} else {
+		mdp5_cstate->pp_done_irqmask = 0;
+		mdp5_cstate->cmd_mode = false;
+	}
+
+	return 0;
+}
+
 struct plane_state {
 	struct drm_plane *plane;
 	struct mdp5_plane_state *state;
@@ -391,6 +530,29 @@ static bool is_fullscreen(struct drm_crtc_state *cstate,
 		((pstate->crtc_y + pstate->crtc_h) >= cstate->mode.vdisplay);
 }
 
+enum mdp_mixer_stage_id get_start_stage(struct drm_crtc *crtc,
+					struct drm_crtc_state *new_crtc_state,
+					struct drm_plane_state *bpstate)
+{
+	struct mdp5_crtc_state *mdp5_cstate =
+			to_mdp5_crtc_state(new_crtc_state);
+
+	/*
+	 * if we're in source split mode, it's mandatory to have
+	 * border out on the base stage
+	 */
+	if (mdp5_cstate->pipeline.r_mixer)
+		return STAGE0;
+
+	/* if the bottom-most layer is not fullscreen, we need to use
+	 * it for solid-color:
+	 */
+	if (!is_fullscreen(new_crtc_state, bpstate))
+		return STAGE0;
+
+	return STAGE_BASE;
+}
+
 static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,
 		struct drm_crtc_state *state)
 {
@@ -400,8 +562,12 @@ static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,
 	struct plane_state pstates[STAGE_MAX + 1];
 	const struct mdp5_cfg_hw *hw_cfg;
 	const struct drm_plane_state *pstate;
+	const struct drm_display_mode *mode = &state->adjusted_mode;
 	bool cursor_plane = false;
-	int cnt = 0, base = 0, i;
+	bool need_right_mixer = false;
+	int cnt = 0, i;
+	int ret;
+	enum mdp_mixer_stage_id start;
 
 	DBG("%s: check", crtc->name);
 
@@ -409,32 +575,52 @@ static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,
 		pstates[cnt].plane = plane;
 		pstates[cnt].state = to_mdp5_plane_state(pstate);
 
+		/*
+		 * if any plane on this crtc uses 2 hwpipes, then we need
+		 * the crtc to have a right hwmixer.
+		 */
+		if (pstates[cnt].state->r_hwpipe)
+			need_right_mixer = true;
 		cnt++;
 
 		if (plane->type == DRM_PLANE_TYPE_CURSOR)
 			cursor_plane = true;
 	}
 
-	/* assign a stage based on sorted zpos property */
-	sort(pstates, cnt, sizeof(pstates[0]), pstate_cmp, NULL);
+	/* bail out early if there aren't any planes */
+	if (!cnt)
+		return 0;
 
-	/* if the bottom-most layer is not fullscreen, we need to use
-	 * it for solid-color:
+	hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
+
+	/*
+	 * we need a right hwmixer if the mode's width is greater than a single
+	 * LM's max width
 	 */
-	if ((cnt > 0) && !is_fullscreen(state, &pstates[0].state->base))
-		base++;
+	if (mode->hdisplay > hw_cfg->lm.max_width)
+		need_right_mixer = true;
+
+	ret = mdp5_crtc_setup_pipeline(crtc, state, need_right_mixer);
+	if (ret) {
+		dev_err(dev->dev, "couldn't assign mixers %d\n", ret);
+		return ret;
+	}
+
+	/* assign a stage based on sorted zpos property */
+	sort(pstates, cnt, sizeof(pstates[0]), pstate_cmp, NULL);
 
 	/* trigger a warning if cursor isn't the highest zorder */
 	WARN_ON(cursor_plane &&
 		(pstates[cnt - 1].plane->type != DRM_PLANE_TYPE_CURSOR));
 
+	start = get_start_stage(crtc, state, &pstates[0].state->base);
+
 	/* verify that there are not too many planes attached to crtc
 	 * and that we don't have conflicting mixer stages:
 	 */
-	hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
-
-	if ((cnt + base) >= hw_cfg->lm.nb_stages) {
-		dev_err(dev->dev, "too many planes! cnt=%d, base=%d\n", cnt, base);
+	if ((cnt + start - 1) >= hw_cfg->lm.nb_stages) {
+		dev_err(dev->dev, "too many planes! cnt=%d, start stage=%d\n",
+			cnt, start);
 		return -EINVAL;
 	}
 
@@ -442,7 +628,7 @@ static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,
 		if (cursor_plane && (i == (cnt - 1)))
 			pstates[i].state->stage = hw_cfg->lm.nb_stages;
 		else
-			pstates[i].state->stage = STAGE_BASE + i + base;
+			pstates[i].state->stage = start + i;
 		DBG("%s: assign pipe %s on stage=%d", crtc->name,
 				pstates[i].plane->name,
 				pstates[i].state->stage);
@@ -461,6 +647,7 @@ static void mdp5_crtc_atomic_flush(struct drm_crtc *crtc,
 				   struct drm_crtc_state *old_crtc_state)
 {
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 	struct drm_device *dev = crtc->dev;
 	unsigned long flags;
 
@@ -477,7 +664,8 @@ static void mdp5_crtc_atomic_flush(struct drm_crtc *crtc,
 	 * it means we are trying to flush a CRTC whose state is disabled:
 	 * nothing else needs to be done.
 	 */
-	if (unlikely(!mdp5_crtc->ctl))
+	/* XXX: Can this happen now ? */
+	if (unlikely(!mdp5_cstate->ctl))
 		return;
 
 	blend_setup(crtc);
@@ -488,11 +676,16 @@ static void mdp5_crtc_atomic_flush(struct drm_crtc *crtc,
 	 * This is safe because no pp_done will happen before SW trigger
 	 * in command mode.
 	 */
-	if (mdp5_crtc->cmd_mode)
+	if (mdp5_cstate->cmd_mode)
 		request_pp_done_pending(crtc);
 
 	mdp5_crtc->flushed_mask = crtc_flush_all(crtc);
 
+	/* XXX are we leaking out state here? */
+	mdp5_crtc->vblank.irqmask = mdp5_cstate->vblank_irqmask;
+	mdp5_crtc->err.irqmask = mdp5_cstate->err_irqmask;
+	mdp5_crtc->pp_done.irqmask = mdp5_cstate->pp_done_irqmask;
+
 	request_pending(crtc, PENDING_FLIP);
 }
 
@@ -527,11 +720,14 @@ static int mdp5_crtc_cursor_set(struct drm_crtc *crtc,
 		uint32_t width, uint32_t height)
 {
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	struct mdp5_pipeline *pipeline = &mdp5_cstate->pipeline;
 	struct drm_device *dev = crtc->dev;
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
 	struct drm_gem_object *cursor_bo, *old_bo = NULL;
 	uint32_t blendcfg, stride;
 	uint64_t cursor_addr;
+	struct mdp5_ctl *ctl;
 	int ret, lm;
 	enum mdp5_cursor_alpha cur_alpha = CURSOR_ALPHA_PER_PIXEL;
 	uint32_t flush_mask = mdp_ctl_flush_mask_cursor(0);
@@ -544,7 +740,12 @@ static int mdp5_crtc_cursor_set(struct drm_crtc *crtc,
 		return -EINVAL;
 	}
 
-	if (NULL == mdp5_crtc->ctl)
+	ctl = mdp5_cstate->ctl;
+	if (!ctl)
+		return -EINVAL;
+
+	/* don't support LM cursors when we we have source split enabled */
+	if (mdp5_cstate->pipeline.r_mixer)
 		return -EINVAL;
 
 	if (!handle) {
@@ -561,7 +762,7 @@ static int mdp5_crtc_cursor_set(struct drm_crtc *crtc,
 	if (ret)
 		return -EINVAL;
 
-	lm = mdp5_crtc->lm;
+	lm = mdp5_cstate->pipeline.mixer->lm;
 	stride = width * drm_format_plane_cpp(DRM_FORMAT_ARGB8888, 0);
 
 	spin_lock_irqsave(&mdp5_crtc->cursor.lock, flags);
@@ -591,7 +792,7 @@ static int mdp5_crtc_cursor_set(struct drm_crtc *crtc,
 	spin_unlock_irqrestore(&mdp5_crtc->cursor.lock, flags);
 
 set_cursor:
-	ret = mdp5_ctl_set_cursor(mdp5_crtc->ctl, 0, cursor_enable);
+	ret = mdp5_ctl_set_cursor(ctl, pipeline, 0, cursor_enable);
 	if (ret) {
 		dev_err(dev->dev, "failed to %sable cursor: %d\n",
 				cursor_enable ? "en" : "dis", ret);
@@ -613,11 +814,17 @@ static int mdp5_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
 {
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	uint32_t lm = mdp5_cstate->pipeline.mixer->lm;
 	uint32_t flush_mask = mdp_ctl_flush_mask_cursor(0);
 	uint32_t roi_w;
 	uint32_t roi_h;
 	unsigned long flags;
 
+	/* don't support LM cursors when we we have source split enabled */
+	if (mdp5_cstate->pipeline.r_mixer)
+		return -EINVAL;
+
 	/* In case the CRTC is disabled, just drop the cursor update */
 	if (unlikely(!crtc->state->enable))
 		return 0;
@@ -628,10 +835,10 @@ static int mdp5_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
 	get_roi(crtc, &roi_w, &roi_h);
 
 	spin_lock_irqsave(&mdp5_crtc->cursor.lock, flags);
-	mdp5_write(mdp5_kms, REG_MDP5_LM_CURSOR_SIZE(mdp5_crtc->lm),
+	mdp5_write(mdp5_kms, REG_MDP5_LM_CURSOR_SIZE(lm),
 			MDP5_LM_CURSOR_SIZE_ROI_H(roi_h) |
 			MDP5_LM_CURSOR_SIZE_ROI_W(roi_w));
-	mdp5_write(mdp5_kms, REG_MDP5_LM_CURSOR_START_XY(mdp5_crtc->lm),
+	mdp5_write(mdp5_kms, REG_MDP5_LM_CURSOR_START_XY(lm),
 			MDP5_LM_CURSOR_START_XY_Y_START(y) |
 			MDP5_LM_CURSOR_START_XY_X_START(x));
 	spin_unlock_irqrestore(&mdp5_crtc->cursor.lock, flags);
@@ -641,16 +848,80 @@ static int mdp5_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
 	return 0;
 }
 
+static void
+mdp5_crtc_atomic_print_state(struct drm_printer *p,
+			     const struct drm_crtc_state *state)
+{
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(state);
+	struct mdp5_pipeline *pipeline = &mdp5_cstate->pipeline;
+	struct mdp5_kms *mdp5_kms = get_kms(state->crtc);
+
+	if (WARN_ON(!pipeline))
+		return;
+
+	drm_printf(p, "\thwmixer=%s\n", pipeline->mixer ?
+			pipeline->mixer->name : "(null)");
+
+	if (mdp5_kms->caps & MDP_CAP_SRC_SPLIT)
+		drm_printf(p, "\tright hwmixer=%s\n", pipeline->r_mixer ?
+			   pipeline->r_mixer->name : "(null)");
+}
+
+static void mdp5_crtc_reset(struct drm_crtc *crtc)
+{
+	struct mdp5_crtc_state *mdp5_cstate;
+
+	if (crtc->state) {
+		__drm_atomic_helper_crtc_destroy_state(crtc->state);
+		kfree(to_mdp5_crtc_state(crtc->state));
+	}
+
+	mdp5_cstate = kzalloc(sizeof(*mdp5_cstate), GFP_KERNEL);
+
+	if (mdp5_cstate) {
+		mdp5_cstate->base.crtc = crtc;
+		crtc->state = &mdp5_cstate->base;
+	}
+}
+
+static struct drm_crtc_state *
+mdp5_crtc_duplicate_state(struct drm_crtc *crtc)
+{
+	struct mdp5_crtc_state *mdp5_cstate;
+
+	if (WARN_ON(!crtc->state))
+		return NULL;
+
+	mdp5_cstate = kmemdup(to_mdp5_crtc_state(crtc->state),
+			      sizeof(*mdp5_cstate), GFP_KERNEL);
+	if (!mdp5_cstate)
+		return NULL;
+
+	__drm_atomic_helper_crtc_duplicate_state(crtc, &mdp5_cstate->base);
+
+	return &mdp5_cstate->base;
+}
+
+static void mdp5_crtc_destroy_state(struct drm_crtc *crtc, struct drm_crtc_state *state)
+{
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(state);
+
+	__drm_atomic_helper_crtc_destroy_state(state);
+
+	kfree(mdp5_cstate);
+}
+
 static const struct drm_crtc_funcs mdp5_crtc_funcs = {
 	.set_config = drm_atomic_helper_set_config,
 	.destroy = mdp5_crtc_destroy,
 	.page_flip = drm_atomic_helper_page_flip,
 	.set_property = drm_atomic_helper_crtc_set_property,
-	.reset = drm_atomic_helper_crtc_reset,
-	.atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
-	.atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
+	.reset = mdp5_crtc_reset,
+	.atomic_duplicate_state = mdp5_crtc_duplicate_state,
+	.atomic_destroy_state = mdp5_crtc_destroy_state,
 	.cursor_set = mdp5_crtc_cursor_set,
 	.cursor_move = mdp5_crtc_cursor_move,
+	.atomic_print_state = mdp5_crtc_atomic_print_state,
 };
 
 static const struct drm_crtc_funcs mdp5_crtc_no_lm_cursor_funcs = {
@@ -658,9 +929,10 @@ static const struct drm_crtc_funcs mdp5_crtc_no_lm_cursor_funcs = {
 	.destroy = mdp5_crtc_destroy,
 	.page_flip = drm_atomic_helper_page_flip,
 	.set_property = drm_atomic_helper_crtc_set_property,
-	.reset = drm_atomic_helper_crtc_reset,
-	.atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
-	.atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
+	.reset = mdp5_crtc_reset,
+	.atomic_duplicate_state = mdp5_crtc_duplicate_state,
+	.atomic_destroy_state = mdp5_crtc_destroy_state,
+	.atomic_print_state = mdp5_crtc_atomic_print_state,
 };
 
 static const struct drm_crtc_helper_funcs mdp5_crtc_helper_funcs = {
@@ -710,22 +982,26 @@ static void mdp5_crtc_wait_for_pp_done(struct drm_crtc *crtc)
 {
 	struct drm_device *dev = crtc->dev;
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 	int ret;
 
 	ret = wait_for_completion_timeout(&mdp5_crtc->pp_completion,
 						msecs_to_jiffies(50));
 	if (ret == 0)
-		dev_warn(dev->dev, "pp done time out, lm=%d\n", mdp5_crtc->lm);
+		dev_warn(dev->dev, "pp done time out, lm=%d\n",
+			 mdp5_cstate->pipeline.mixer->lm);
 }
 
 static void mdp5_crtc_wait_for_flush_done(struct drm_crtc *crtc)
 {
 	struct drm_device *dev = crtc->dev;
 	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+	struct mdp5_ctl *ctl = mdp5_cstate->ctl;
 	int ret;
 
 	/* Should not call this function if crtc is disabled. */
-	if (!mdp5_crtc->ctl)
+	if (!ctl)
 		return;
 
 	ret = drm_crtc_vblank_get(crtc);
@@ -733,7 +1009,7 @@ static void mdp5_crtc_wait_for_flush_done(struct drm_crtc *crtc)
 		return;
 
 	ret = wait_event_timeout(dev->vblank[drm_crtc_index(crtc)].queue,
-		((mdp5_ctl_get_commit_status(mdp5_crtc->ctl) &
+		((mdp5_ctl_get_commit_status(ctl) &
 		mdp5_crtc->flushed_mask) == 0),
 		msecs_to_jiffies(50));
 	if (ret <= 0)
@@ -750,52 +1026,54 @@ uint32_t mdp5_crtc_vblank(struct drm_crtc *crtc)
 	return mdp5_crtc->vblank.irqmask;
 }
 
-void mdp5_crtc_set_pipeline(struct drm_crtc *crtc,
-		struct mdp5_interface *intf, struct mdp5_ctl *ctl)
+void mdp5_crtc_set_pipeline(struct drm_crtc *crtc)
 {
-	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 	struct mdp5_kms *mdp5_kms = get_kms(crtc);
-	int lm = mdp5_crtc_get_lm(crtc);
-
-	/* now that we know what irq's we want: */
-	mdp5_crtc->err.irqmask = intf2err(intf->num);
-	mdp5_crtc->vblank.irqmask = intf2vblank(lm, intf);
-
-	if ((intf->type == INTF_DSI) &&
-		(intf->mode == MDP5_INTF_DSI_MODE_COMMAND)) {
-		mdp5_crtc->pp_done.irqmask = lm2ppdone(lm);
-		mdp5_crtc->pp_done.irq = mdp5_crtc_pp_done_irq;
-		mdp5_crtc->cmd_mode = true;
-	} else {
-		mdp5_crtc->pp_done.irqmask = 0;
-		mdp5_crtc->pp_done.irq = NULL;
-		mdp5_crtc->cmd_mode = false;
-	}
 
+	/* should this be done elsewhere ? */
 	mdp_irq_update(&mdp5_kms->base);
 
-	mdp5_crtc->ctl = ctl;
-	mdp5_ctl_set_pipeline(ctl, intf, lm);
+	mdp5_ctl_set_pipeline(mdp5_cstate->ctl, &mdp5_cstate->pipeline);
 }
 
 struct mdp5_ctl *mdp5_crtc_get_ctl(struct drm_crtc *crtc)
 {
-	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 
-	return mdp5_crtc->ctl;
+	return mdp5_cstate->ctl;
 }
 
-int mdp5_crtc_get_lm(struct drm_crtc *crtc)
+struct mdp5_hw_mixer *mdp5_crtc_get_mixer(struct drm_crtc *crtc)
 {
-	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
-	return WARN_ON(!crtc) ? -EINVAL : mdp5_crtc->lm;
+	struct mdp5_crtc_state *mdp5_cstate;
+
+	if (WARN_ON(!crtc))
+		return ERR_PTR(-EINVAL);
+
+	mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+
+	return WARN_ON(!mdp5_cstate->pipeline.mixer) ?
+		ERR_PTR(-EINVAL) : mdp5_cstate->pipeline.mixer;
+}
+
+struct mdp5_pipeline *mdp5_crtc_get_pipeline(struct drm_crtc *crtc)
+{
+	struct mdp5_crtc_state *mdp5_cstate;
+
+	if (WARN_ON(!crtc))
+		return ERR_PTR(-EINVAL);
+
+	mdp5_cstate = to_mdp5_crtc_state(crtc->state);
+
+	return &mdp5_cstate->pipeline;
 }
 
 void mdp5_crtc_wait_for_commit_done(struct drm_crtc *crtc)
 {
-	struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc->state);
 
-	if (mdp5_crtc->cmd_mode)
+	if (mdp5_cstate->cmd_mode)
 		mdp5_crtc_wait_for_pp_done(crtc);
 	else
 		mdp5_crtc_wait_for_flush_done(crtc);
@@ -816,7 +1094,6 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev,
 	crtc = &mdp5_crtc->base;
 
 	mdp5_crtc->id = id;
-	mdp5_crtc->lm = GET_LM_ID(id);
 
 	spin_lock_init(&mdp5_crtc->lm_lock);
 	spin_lock_init(&mdp5_crtc->cursor.lock);
@@ -824,6 +1101,7 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev,
 
 	mdp5_crtc->vblank.irq = mdp5_crtc_vblank_irq;
 	mdp5_crtc->err.irq = mdp5_crtc_err_irq;
+	mdp5_crtc->pp_done.irq = mdp5_crtc_pp_done_irq;
 
 	if (cursor_plane)
 		drm_crtc_init_with_planes(dev, crtc, plane, cursor_plane,
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c
index 8b93f7e13200..439e0a300e25 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c
@@ -32,24 +32,16 @@
 #define CTL_STAT_BUSY		0x1
 #define CTL_STAT_BOOKED	0x2
 
-struct op_mode {
-	struct mdp5_interface intf;
-
-	bool encoder_enabled;
-	uint32_t start_mask;
-};
-
 struct mdp5_ctl {
 	struct mdp5_ctl_manager *ctlm;
 
 	u32 id;
-	int lm;
 
 	/* CTL status bitmask */
 	u32 status;
 
-	/* Operation Mode Configuration for the Pipeline */
-	struct op_mode pipeline;
+	bool encoder_enabled;
+	uint32_t start_mask;
 
 	/* REG_MDP5_CTL_*(<id>) registers access info + lock: */
 	spinlock_t hw_lock;
@@ -146,9 +138,10 @@ static void set_display_intf(struct mdp5_kms *mdp5_kms,
 	spin_unlock_irqrestore(&mdp5_kms->resource_lock, flags);
 }
 
-static void set_ctl_op(struct mdp5_ctl *ctl, struct mdp5_interface *intf)
+static void set_ctl_op(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline)
 {
 	unsigned long flags;
+	struct mdp5_interface *intf = pipeline->intf;
 	u32 ctl_op = 0;
 
 	if (!mdp5_cfg_intf_is_virtual(intf->type))
@@ -169,52 +162,50 @@ static void set_ctl_op(struct mdp5_ctl *ctl, struct mdp5_interface *intf)
 		break;
 	}
 
+	if (pipeline->r_mixer)
+		ctl_op |= MDP5_CTL_OP_PACK_3D_ENABLE |
+			  MDP5_CTL_OP_PACK_3D(1);
+
 	spin_lock_irqsave(&ctl->hw_lock, flags);
 	ctl_write(ctl, REG_MDP5_CTL_OP(ctl->id), ctl_op);
 	spin_unlock_irqrestore(&ctl->hw_lock, flags);
 }
 
-int mdp5_ctl_set_pipeline(struct mdp5_ctl *ctl,
-		struct mdp5_interface *intf, int lm)
+int mdp5_ctl_set_pipeline(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline)
 {
 	struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
 	struct mdp5_kms *mdp5_kms = get_kms(ctl_mgr);
+	struct mdp5_interface *intf = pipeline->intf;
+	struct mdp5_hw_mixer *mixer = pipeline->mixer;
+	struct mdp5_hw_mixer *r_mixer = pipeline->r_mixer;
 
-	if (unlikely(WARN_ON(intf->num != ctl->pipeline.intf.num))) {
-		dev_err(mdp5_kms->dev->dev,
-			"CTL %d is allocated by INTF %d, but used by INTF %d\n",
-			ctl->id, ctl->pipeline.intf.num, intf->num);
-		return -EINVAL;
-	}
-
-	ctl->lm = lm;
-
-	memcpy(&ctl->pipeline.intf, intf, sizeof(*intf));
-
-	ctl->pipeline.start_mask = mdp_ctl_flush_mask_lm(ctl->lm) |
-				   mdp_ctl_flush_mask_encoder(intf);
+	ctl->start_mask = mdp_ctl_flush_mask_lm(mixer->lm) |
+			  mdp_ctl_flush_mask_encoder(intf);
+	if (r_mixer)
+		ctl->start_mask |= mdp_ctl_flush_mask_lm(r_mixer->lm);
 
 	/* Virtual interfaces need not set a display intf (e.g.: Writeback) */
 	if (!mdp5_cfg_intf_is_virtual(intf->type))
 		set_display_intf(mdp5_kms, intf);
 
-	set_ctl_op(ctl, intf);
+	set_ctl_op(ctl, pipeline);
 
 	return 0;
 }
 
-static bool start_signal_needed(struct mdp5_ctl *ctl)
+static bool start_signal_needed(struct mdp5_ctl *ctl,
+				struct mdp5_pipeline *pipeline)
 {
-	struct op_mode *pipeline = &ctl->pipeline;
+	struct mdp5_interface *intf = pipeline->intf;
 
-	if (!pipeline->encoder_enabled || pipeline->start_mask != 0)
+	if (!ctl->encoder_enabled || ctl->start_mask != 0)
 		return false;
 
-	switch (pipeline->intf.type) {
+	switch (intf->type) {
 	case INTF_WB:
 		return true;
 	case INTF_DSI:
-		return pipeline->intf.mode == MDP5_INTF_DSI_MODE_COMMAND;
+		return intf->mode == MDP5_INTF_DSI_MODE_COMMAND;
 	default:
 		return false;
 	}
@@ -236,19 +227,23 @@ static void send_start_signal(struct mdp5_ctl *ctl)
 	spin_unlock_irqrestore(&ctl->hw_lock, flags);
 }
 
-static void refill_start_mask(struct mdp5_ctl *ctl)
+static void refill_start_mask(struct mdp5_ctl *ctl,
+			      struct mdp5_pipeline *pipeline)
 {
-	struct op_mode *pipeline = &ctl->pipeline;
-	struct mdp5_interface *intf = &ctl->pipeline.intf;
+	struct mdp5_interface *intf = pipeline->intf;
+	struct mdp5_hw_mixer *mixer = pipeline->mixer;
+	struct mdp5_hw_mixer *r_mixer = pipeline->r_mixer;
 
-	pipeline->start_mask = mdp_ctl_flush_mask_lm(ctl->lm);
+	ctl->start_mask = mdp_ctl_flush_mask_lm(mixer->lm);
+	if (r_mixer)
+		ctl->start_mask |= mdp_ctl_flush_mask_lm(r_mixer->lm);
 
 	/*
 	 * Writeback encoder needs to program & flush
 	 * address registers for each page flip..
 	 */
 	if (intf->type == INTF_WB)
-		pipeline->start_mask |= mdp_ctl_flush_mask_encoder(intf);
+		ctl->start_mask |= mdp_ctl_flush_mask_encoder(intf);
 }
 
 /**
@@ -259,17 +254,21 @@ static void refill_start_mask(struct mdp5_ctl *ctl)
  * Note:
  * This encoder state is needed to trigger START signal (data path kickoff).
  */
-int mdp5_ctl_set_encoder_state(struct mdp5_ctl *ctl, bool enabled)
+int mdp5_ctl_set_encoder_state(struct mdp5_ctl *ctl,
+			       struct mdp5_pipeline *pipeline,
+			       bool enabled)
 {
+	struct mdp5_interface *intf = pipeline->intf;
+
 	if (WARN_ON(!ctl))
 		return -EINVAL;
 
-	ctl->pipeline.encoder_enabled = enabled;
-	DBG("intf_%d: %s", ctl->pipeline.intf.num, enabled ? "on" : "off");
+	ctl->encoder_enabled = enabled;
+	DBG("intf_%d: %s", intf->num, enabled ? "on" : "off");
 
-	if (start_signal_needed(ctl)) {
+	if (start_signal_needed(ctl, pipeline)) {
 		send_start_signal(ctl);
-		refill_start_mask(ctl);
+		refill_start_mask(ctl, pipeline);
 	}
 
 	return 0;
@@ -280,29 +279,35 @@ int mdp5_ctl_set_encoder_state(struct mdp5_ctl *ctl, bool enabled)
  * CTL registers need to be flushed after calling this function
  * (call mdp5_ctl_commit() with mdp_ctl_flush_mask_ctl() mask)
  */
-int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, int cursor_id, bool enable)
+int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline,
+			int cursor_id, bool enable)
 {
 	struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
 	unsigned long flags;
 	u32 blend_cfg;
-	int lm = ctl->lm;
+	struct mdp5_hw_mixer *mixer = pipeline->mixer;
+
+	if (unlikely(WARN_ON(!mixer))) {
+		dev_err(ctl_mgr->dev->dev, "CTL %d cannot find LM",
+			ctl->id);
+		return -EINVAL;
+	}
 
-	if (unlikely(WARN_ON(lm < 0))) {
-		dev_err(ctl_mgr->dev->dev, "CTL %d cannot find LM: %d",
-				ctl->id, lm);
+	if (pipeline->r_mixer) {
+		dev_err(ctl_mgr->dev->dev, "unsupported configuration");
 		return -EINVAL;
 	}
 
 	spin_lock_irqsave(&ctl->hw_lock, flags);
 
-	blend_cfg = ctl_read(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm));
+	blend_cfg = ctl_read(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, mixer->lm));
 
 	if (enable)
 		blend_cfg |=  MDP5_CTL_LAYER_REG_CURSOR_OUT;
 	else
 		blend_cfg &= ~MDP5_CTL_LAYER_REG_CURSOR_OUT;
 
-	ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm), blend_cfg);
+	ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, mixer->lm), blend_cfg);
 	ctl->cursor_on = enable;
 
 	spin_unlock_irqrestore(&ctl->hw_lock, flags);
@@ -355,37 +360,88 @@ static u32 mdp_ctl_blend_ext_mask(enum mdp5_pipe pipe,
 	}
 }
 
-int mdp5_ctl_blend(struct mdp5_ctl *ctl, enum mdp5_pipe *stage, u32 stage_cnt,
-		   u32 ctl_blend_op_flags)
+static void mdp5_ctl_reset_blend_regs(struct mdp5_ctl *ctl)
+{
+	unsigned long flags;
+	struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
+	int i;
+
+	spin_lock_irqsave(&ctl->hw_lock, flags);
+
+	for (i = 0; i < ctl_mgr->nlm; i++) {
+		ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, i), 0x0);
+		ctl_write(ctl, REG_MDP5_CTL_LAYER_EXT_REG(ctl->id, i), 0x0);
+	}
+
+	spin_unlock_irqrestore(&ctl->hw_lock, flags);
+}
+
+#define PIPE_LEFT	0
+#define PIPE_RIGHT	1
+int mdp5_ctl_blend(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline,
+		   enum mdp5_pipe stage[][MAX_PIPE_STAGE],
+		   enum mdp5_pipe r_stage[][MAX_PIPE_STAGE],
+		   u32 stage_cnt, u32 ctl_blend_op_flags)
 {
+	struct mdp5_hw_mixer *mixer = pipeline->mixer;
+	struct mdp5_hw_mixer *r_mixer = pipeline->r_mixer;
 	unsigned long flags;
 	u32 blend_cfg = 0, blend_ext_cfg = 0;
+	u32 r_blend_cfg = 0, r_blend_ext_cfg = 0;
 	int i, start_stage;
 
+	mdp5_ctl_reset_blend_regs(ctl);
+
 	if (ctl_blend_op_flags & MDP5_CTL_BLEND_OP_FLAG_BORDER_OUT) {
 		start_stage = STAGE0;
 		blend_cfg |= MDP5_CTL_LAYER_REG_BORDER_COLOR;
+		if (r_mixer)
+			r_blend_cfg |= MDP5_CTL_LAYER_REG_BORDER_COLOR;
 	} else {
 		start_stage = STAGE_BASE;
 	}
 
 	for (i = start_stage; stage_cnt && i <= STAGE_MAX; i++) {
-		blend_cfg |= mdp_ctl_blend_mask(stage[i], i);
-		blend_ext_cfg |= mdp_ctl_blend_ext_mask(stage[i], i);
+		blend_cfg |=
+			mdp_ctl_blend_mask(stage[i][PIPE_LEFT], i) |
+			mdp_ctl_blend_mask(stage[i][PIPE_RIGHT], i);
+		blend_ext_cfg |=
+			mdp_ctl_blend_ext_mask(stage[i][PIPE_LEFT], i) |
+			mdp_ctl_blend_ext_mask(stage[i][PIPE_RIGHT], i);
+		if (r_mixer) {
+			r_blend_cfg |=
+				mdp_ctl_blend_mask(r_stage[i][PIPE_LEFT], i) |
+				mdp_ctl_blend_mask(r_stage[i][PIPE_RIGHT], i);
+			r_blend_ext_cfg |=
+			     mdp_ctl_blend_ext_mask(r_stage[i][PIPE_LEFT], i) |
+			     mdp_ctl_blend_ext_mask(r_stage[i][PIPE_RIGHT], i);
+		}
 	}
 
 	spin_lock_irqsave(&ctl->hw_lock, flags);
 	if (ctl->cursor_on)
 		blend_cfg |=  MDP5_CTL_LAYER_REG_CURSOR_OUT;
 
-	ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, ctl->lm), blend_cfg);
-	ctl_write(ctl, REG_MDP5_CTL_LAYER_EXT_REG(ctl->id, ctl->lm), blend_ext_cfg);
+	ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, mixer->lm), blend_cfg);
+	ctl_write(ctl, REG_MDP5_CTL_LAYER_EXT_REG(ctl->id, mixer->lm),
+		  blend_ext_cfg);
+	if (r_mixer) {
+		ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, r_mixer->lm),
+			  r_blend_cfg);
+		ctl_write(ctl, REG_MDP5_CTL_LAYER_EXT_REG(ctl->id, r_mixer->lm),
+			  r_blend_ext_cfg);
+	}
 	spin_unlock_irqrestore(&ctl->hw_lock, flags);
 
-	ctl->pending_ctl_trigger = mdp_ctl_flush_mask_lm(ctl->lm);
+	ctl->pending_ctl_trigger = mdp_ctl_flush_mask_lm(mixer->lm);
+	if (r_mixer)
+		ctl->pending_ctl_trigger |= mdp_ctl_flush_mask_lm(r_mixer->lm);
 
-	DBG("lm%d: blend config = 0x%08x. ext_cfg = 0x%08x", ctl->lm,
+	DBG("lm%d: blend config = 0x%08x. ext_cfg = 0x%08x", mixer->lm,
 		blend_cfg, blend_ext_cfg);
+	if (r_mixer)
+		DBG("lm%d: blend config = 0x%08x. ext_cfg = 0x%08x",
+		    r_mixer->lm, r_blend_cfg, r_blend_ext_cfg);
 
 	return 0;
 }
@@ -443,7 +499,8 @@ u32 mdp_ctl_flush_mask_lm(int lm)
 	}
 }
 
-static u32 fix_sw_flush(struct mdp5_ctl *ctl, u32 flush_mask)
+static u32 fix_sw_flush(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline,
+			u32 flush_mask)
 {
 	struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
 	u32 sw_mask = 0;
@@ -452,7 +509,7 @@ static u32 fix_sw_flush(struct mdp5_ctl *ctl, u32 flush_mask)
 
 	/* for some targets, cursor bit is the same as LM bit */
 	if (BIT_NEEDS_SW_FIX(MDP5_CTL_FLUSH_CURSOR_0))
-		sw_mask |= mdp_ctl_flush_mask_lm(ctl->lm);
+		sw_mask |= mdp_ctl_flush_mask_lm(pipeline->mixer->lm);
 
 	return sw_mask;
 }
@@ -498,25 +555,26 @@ static void fix_for_single_flush(struct mdp5_ctl *ctl, u32 *flush_mask,
  *
  * Return H/W flushed bit mask.
  */
-u32 mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask)
+u32 mdp5_ctl_commit(struct mdp5_ctl *ctl,
+		    struct mdp5_pipeline *pipeline,
+		    u32 flush_mask)
 {
 	struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
-	struct op_mode *pipeline = &ctl->pipeline;
 	unsigned long flags;
 	u32 flush_id = ctl->id;
 	u32 curr_ctl_flush_mask;
 
-	pipeline->start_mask &= ~flush_mask;
+	ctl->start_mask &= ~flush_mask;
 
 	VERB("flush_mask=%x, start_mask=%x, trigger=%x", flush_mask,
-			pipeline->start_mask, ctl->pending_ctl_trigger);
+			ctl->start_mask, ctl->pending_ctl_trigger);
 
 	if (ctl->pending_ctl_trigger & flush_mask) {
 		flush_mask |= MDP5_CTL_FLUSH_CTL;
 		ctl->pending_ctl_trigger = 0;
 	}
 
-	flush_mask |= fix_sw_flush(ctl, flush_mask);
+	flush_mask |= fix_sw_flush(ctl, pipeline, flush_mask);
 
 	flush_mask &= ctl_mgr->flush_hw_mask;
 
@@ -530,9 +588,9 @@ u32 mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask)
 		spin_unlock_irqrestore(&ctl->hw_lock, flags);
 	}
 
-	if (start_signal_needed(ctl)) {
+	if (start_signal_needed(ctl, pipeline)) {
 		send_start_signal(ctl);
-		refill_start_mask(ctl);
+		refill_start_mask(ctl, pipeline);
 	}
 
 	return curr_ctl_flush_mask;
@@ -619,8 +677,6 @@ struct mdp5_ctl *mdp5_ctlm_request(struct mdp5_ctl_manager *ctl_mgr,
 
 found:
 	ctl = &ctl_mgr->ctls[c];
-	ctl->pipeline.intf.num = intf_num;
-	ctl->lm = -1;
 	ctl->status |= CTL_STAT_BUSY;
 	ctl->pending_ctl_trigger = 0;
 	DBG("CTL %d allocated", ctl->id);
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h
index fda00d33e4db..b63120388dc6 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h
@@ -37,13 +37,17 @@ struct mdp5_ctl *mdp5_ctlm_request(struct mdp5_ctl_manager *ctlm, int intf_num);
 int mdp5_ctl_get_ctl_id(struct mdp5_ctl *ctl);
 
 struct mdp5_interface;
-int mdp5_ctl_set_pipeline(struct mdp5_ctl *ctl, struct mdp5_interface *intf,
-				int lm);
-int mdp5_ctl_set_encoder_state(struct mdp5_ctl *ctl, bool enabled);
+struct mdp5_pipeline;
+int mdp5_ctl_set_pipeline(struct mdp5_ctl *ctl, struct mdp5_pipeline *p);
+int mdp5_ctl_set_encoder_state(struct mdp5_ctl *ctl, struct mdp5_pipeline *p,
+			       bool enabled);
 
-int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, int cursor_id, bool enable);
+int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline,
+			int cursor_id, bool enable);
 int mdp5_ctl_pair(struct mdp5_ctl *ctlx, struct mdp5_ctl *ctly, bool enable);
 
+#define MAX_PIPE_STAGE		2
+
 /*
  * mdp5_ctl_blend() - Blend multiple layers on a Layer Mixer (LM)
  *
@@ -56,8 +60,10 @@ int mdp5_ctl_pair(struct mdp5_ctl *ctlx, struct mdp5_ctl *ctly, bool enable);
  * (call mdp5_ctl_commit() with mdp_ctl_flush_mask_ctl() mask)
  */
 #define MDP5_CTL_BLEND_OP_FLAG_BORDER_OUT	BIT(0)
-int mdp5_ctl_blend(struct mdp5_ctl *ctl, enum mdp5_pipe *stage, u32 stage_cnt,
-		   u32 ctl_blend_op_flags);
+int mdp5_ctl_blend(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline,
+		   enum mdp5_pipe stage[][MAX_PIPE_STAGE],
+		   enum mdp5_pipe r_stage[][MAX_PIPE_STAGE],
+		   u32 stage_cnt, u32 ctl_blend_op_flags);
 
 /**
  * mdp_ctl_flush_mask...() - Register FLUSH masks
@@ -71,7 +77,8 @@ u32 mdp_ctl_flush_mask_cursor(int cursor_id);
 u32 mdp_ctl_flush_mask_encoder(struct mdp5_interface *intf);
 
 /* @flush_mask: see CTL flush masks definitions below */
-u32 mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask);
+u32 mdp5_ctl_commit(struct mdp5_ctl *ctl, struct mdp5_pipeline *pipeline,
+		    u32 flush_mask);
 u32 mdp5_ctl_get_commit_status(struct mdp5_ctl *ctl);
 
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c
index 80fa482ae8ed..c2ab0f033031 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c
@@ -109,7 +109,7 @@ static void mdp5_vid_encoder_mode_set(struct drm_encoder *encoder,
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
 	struct drm_device *dev = encoder->dev;
 	struct drm_connector *connector;
-	int intf = mdp5_encoder->intf.num;
+	int intf = mdp5_encoder->intf->num;
 	uint32_t dtv_hsync_skew, vsync_period, vsync_len, ctrl_pol;
 	uint32_t display_v_start, display_v_end;
 	uint32_t hsync_start_x, hsync_end_x;
@@ -130,7 +130,7 @@ static void mdp5_vid_encoder_mode_set(struct drm_encoder *encoder,
 	ctrl_pol = 0;
 
 	/* DSI controller cannot handle active-low sync signals. */
-	if (mdp5_encoder->intf.type != INTF_DSI) {
+	if (mdp5_encoder->intf->type != INTF_DSI) {
 		if (mode->flags & DRM_MODE_FLAG_NHSYNC)
 			ctrl_pol |= MDP5_INTF_POLARITY_CTL_HSYNC_LOW;
 		if (mode->flags & DRM_MODE_FLAG_NVSYNC)
@@ -175,7 +175,7 @@ static void mdp5_vid_encoder_mode_set(struct drm_encoder *encoder,
 	 * DISPLAY_V_START = (VBP * HCYCLE) + HBP
 	 * DISPLAY_V_END = (VBP + VACTIVE) * HCYCLE - 1 - HFP
 	 */
-	if (mdp5_encoder->intf.type == INTF_eDP) {
+	if (mdp5_encoder->intf->type == INTF_eDP) {
 		display_v_start += mode->htotal - mode->hsync_start;
 		display_v_end -= mode->hsync_start - mode->hdisplay;
 	}
@@ -206,8 +206,7 @@ static void mdp5_vid_encoder_mode_set(struct drm_encoder *encoder,
 
 	spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags);
 
-	mdp5_crtc_set_pipeline(encoder->crtc, &mdp5_encoder->intf,
-				mdp5_encoder->ctl);
+	mdp5_crtc_set_pipeline(encoder->crtc);
 }
 
 static void mdp5_vid_encoder_disable(struct drm_encoder *encoder)
@@ -215,20 +214,21 @@ static void mdp5_vid_encoder_disable(struct drm_encoder *encoder)
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
 	struct mdp5_ctl *ctl = mdp5_encoder->ctl;
-	int lm = mdp5_crtc_get_lm(encoder->crtc);
-	struct mdp5_interface *intf = &mdp5_encoder->intf;
-	int intfn = mdp5_encoder->intf.num;
+	struct mdp5_pipeline *pipeline = mdp5_crtc_get_pipeline(encoder->crtc);
+	struct mdp5_hw_mixer *mixer = mdp5_crtc_get_mixer(encoder->crtc);
+	struct mdp5_interface *intf = mdp5_encoder->intf;
+	int intfn = mdp5_encoder->intf->num;
 	unsigned long flags;
 
 	if (WARN_ON(!mdp5_encoder->enabled))
 		return;
 
-	mdp5_ctl_set_encoder_state(ctl, false);
+	mdp5_ctl_set_encoder_state(ctl, pipeline, false);
 
 	spin_lock_irqsave(&mdp5_encoder->intf_lock, flags);
 	mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intfn), 0);
 	spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags);
-	mdp5_ctl_commit(ctl, mdp_ctl_flush_mask_encoder(intf));
+	mdp5_ctl_commit(ctl, pipeline, mdp_ctl_flush_mask_encoder(intf));
 
 	/*
 	 * Wait for a vsync so we know the ENABLE=0 latched before
@@ -238,7 +238,7 @@ static void mdp5_vid_encoder_disable(struct drm_encoder *encoder)
 	 * the settings changes for the new modeset (like new
 	 * scanout buffer) don't latch properly..
 	 */
-	mdp_irq_wait(&mdp5_kms->base, intf2vblank(lm, intf));
+	mdp_irq_wait(&mdp5_kms->base, intf2vblank(mixer, intf));
 
 	bs_set(mdp5_encoder, 0);
 
@@ -250,8 +250,9 @@ static void mdp5_vid_encoder_enable(struct drm_encoder *encoder)
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
 	struct mdp5_ctl *ctl = mdp5_encoder->ctl;
-	struct mdp5_interface *intf = &mdp5_encoder->intf;
-	int intfn = mdp5_encoder->intf.num;
+	struct mdp5_interface *intf = mdp5_encoder->intf;
+	struct mdp5_pipeline *pipeline = mdp5_crtc_get_pipeline(encoder->crtc);
+	int intfn = intf->num;
 	unsigned long flags;
 
 	if (WARN_ON(mdp5_encoder->enabled))
@@ -261,9 +262,9 @@ static void mdp5_vid_encoder_enable(struct drm_encoder *encoder)
 	spin_lock_irqsave(&mdp5_encoder->intf_lock, flags);
 	mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intfn), 1);
 	spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags);
-	mdp5_ctl_commit(ctl, mdp_ctl_flush_mask_encoder(intf));
+	mdp5_ctl_commit(ctl, pipeline, mdp_ctl_flush_mask_encoder(intf));
 
-	mdp5_ctl_set_encoder_state(ctl, true);
+	mdp5_ctl_set_encoder_state(ctl, pipeline, true);
 
 	mdp5_encoder->enabled = true;
 }
@@ -273,7 +274,7 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder,
 				  struct drm_display_mode *adjusted_mode)
 {
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
-	struct mdp5_interface *intf = &mdp5_encoder->intf;
+	struct mdp5_interface *intf = mdp5_encoder->intf;
 
 	if (intf->mode == MDP5_INTF_DSI_MODE_COMMAND)
 		mdp5_cmd_encoder_mode_set(encoder, mode, adjusted_mode);
@@ -284,7 +285,7 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder,
 static void mdp5_encoder_disable(struct drm_encoder *encoder)
 {
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
-	struct mdp5_interface *intf = &mdp5_encoder->intf;
+	struct mdp5_interface *intf = mdp5_encoder->intf;
 
 	if (intf->mode == MDP5_INTF_DSI_MODE_COMMAND)
 		mdp5_cmd_encoder_disable(encoder);
@@ -295,7 +296,7 @@ static void mdp5_encoder_disable(struct drm_encoder *encoder)
 static void mdp5_encoder_enable(struct drm_encoder *encoder)
 {
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
-	struct mdp5_interface *intf = &mdp5_encoder->intf;
+	struct mdp5_interface *intf = mdp5_encoder->intf;
 
 	if (intf->mode == MDP5_INTF_DSI_MODE_COMMAND)
 		mdp5_cmd_encoder_disable(encoder);
@@ -303,17 +304,33 @@ static void mdp5_encoder_enable(struct drm_encoder *encoder)
 		mdp5_vid_encoder_enable(encoder);
 }
 
+static int mdp5_encoder_atomic_check(struct drm_encoder *encoder,
+				     struct drm_crtc_state *crtc_state,
+				     struct drm_connector_state *conn_state)
+{
+	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
+	struct mdp5_crtc_state *mdp5_cstate = to_mdp5_crtc_state(crtc_state);
+	struct mdp5_interface *intf = mdp5_encoder->intf;
+	struct mdp5_ctl *ctl = mdp5_encoder->ctl;
+
+	mdp5_cstate->ctl = ctl;
+	mdp5_cstate->pipeline.intf = intf;
+
+	return 0;
+}
+
 static const struct drm_encoder_helper_funcs mdp5_encoder_helper_funcs = {
 	.mode_set = mdp5_encoder_mode_set,
 	.disable = mdp5_encoder_disable,
 	.enable = mdp5_encoder_enable,
+	.atomic_check = mdp5_encoder_atomic_check,
 };
 
 int mdp5_encoder_get_linecount(struct drm_encoder *encoder)
 {
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
-	int intf = mdp5_encoder->intf.num;
+	int intf = mdp5_encoder->intf->num;
 
 	return mdp5_read(mdp5_kms, REG_MDP5_INTF_LINE_COUNT(intf));
 }
@@ -322,7 +339,7 @@ u32 mdp5_encoder_get_framecount(struct drm_encoder *encoder)
 {
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
 	struct mdp5_kms *mdp5_kms = get_kms(encoder);
-	int intf = mdp5_encoder->intf.num;
+	int intf = mdp5_encoder->intf->num;
 
 	return mdp5_read(mdp5_kms, REG_MDP5_INTF_FRAME_COUNT(intf));
 }
@@ -340,7 +357,7 @@ int mdp5_vid_encoder_set_split_display(struct drm_encoder *encoder,
 		return -EINVAL;
 
 	mdp5_kms = get_kms(encoder);
-	intf_num = mdp5_encoder->intf.num;
+	intf_num = mdp5_encoder->intf->num;
 
 	/* Switch slave encoder's TimingGen Sync mode,
 	 * to use the master's enable signal for the slave encoder.
@@ -369,7 +386,7 @@ int mdp5_vid_encoder_set_split_display(struct drm_encoder *encoder,
 void mdp5_encoder_set_intf_mode(struct drm_encoder *encoder, bool cmd_mode)
 {
 	struct mdp5_encoder *mdp5_encoder = to_mdp5_encoder(encoder);
-	struct mdp5_interface *intf = &mdp5_encoder->intf;
+	struct mdp5_interface *intf = mdp5_encoder->intf;
 
 	/* TODO: Expand this to set writeback modes too */
 	if (cmd_mode) {
@@ -385,7 +402,8 @@ void mdp5_encoder_set_intf_mode(struct drm_encoder *encoder, bool cmd_mode)
 
 /* initialize encoder */
 struct drm_encoder *mdp5_encoder_init(struct drm_device *dev,
-			struct mdp5_interface *intf, struct mdp5_ctl *ctl)
+				      struct mdp5_interface *intf,
+				      struct mdp5_ctl *ctl)
 {
 	struct drm_encoder *encoder = NULL;
 	struct mdp5_encoder *mdp5_encoder;
@@ -399,9 +417,9 @@ struct drm_encoder *mdp5_encoder_init(struct drm_device *dev,
 		goto fail;
 	}
 
-	memcpy(&mdp5_encoder->intf, intf, sizeof(mdp5_encoder->intf));
 	encoder = &mdp5_encoder->base;
 	mdp5_encoder->ctl = ctl;
+	mdp5_encoder->intf = intf;
 
 	spin_lock_init(&mdp5_encoder->intf_lock);
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c
index 41ccd2a15d3c..d3d6b4cae1e6 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c
@@ -93,6 +93,7 @@ struct mdp5_state *mdp5_get_state(struct drm_atomic_state *s)
 
 	/* Copy state: */
 	new_state->hwpipe = mdp5_kms->state->hwpipe;
+	new_state->hwmixer = mdp5_kms->state->hwmixer;
 	if (mdp5_kms->smp)
 		new_state->smp = mdp5_kms->state->smp;
 
@@ -165,13 +166,16 @@ static void mdp5_kms_destroy(struct msm_kms *kms)
 	struct msm_gem_address_space *aspace = mdp5_kms->aspace;
 	int i;
 
+	for (i = 0; i < mdp5_kms->num_hwmixers; i++)
+		mdp5_mixer_destroy(mdp5_kms->hwmixers[i]);
+
 	for (i = 0; i < mdp5_kms->num_hwpipes; i++)
 		mdp5_pipe_destroy(mdp5_kms->hwpipes[i]);
 
 	if (aspace) {
 		aspace->mmu->funcs->detach(aspace->mmu,
 				iommu_ports, ARRAY_SIZE(iommu_ports));
-		msm_gem_address_space_destroy(aspace);
+		msm_gem_address_space_put(aspace);
 	}
 }
 
@@ -268,19 +272,14 @@ int mdp5_enable(struct mdp5_kms *mdp5_kms)
 }
 
 static struct drm_encoder *construct_encoder(struct mdp5_kms *mdp5_kms,
-		enum mdp5_intf_type intf_type, int intf_num,
-		struct mdp5_ctl *ctl)
+					     struct mdp5_interface *intf,
+					     struct mdp5_ctl *ctl)
 {
 	struct drm_device *dev = mdp5_kms->dev;
 	struct msm_drm_private *priv = dev->dev_private;
 	struct drm_encoder *encoder;
-	struct mdp5_interface intf = {
-			.num	= intf_num,
-			.type	= intf_type,
-			.mode	= MDP5_INTF_MODE_NONE,
-	};
 
-	encoder = mdp5_encoder_init(dev, &intf, ctl);
+	encoder = mdp5_encoder_init(dev, intf, ctl);
 	if (IS_ERR(encoder)) {
 		dev_err(dev->dev, "failed to construct encoder\n");
 		return encoder;
@@ -309,32 +308,28 @@ static int get_dsi_id_from_intf(const struct mdp5_cfg_hw *hw_cfg, int intf_num)
 	return -EINVAL;
 }
 
-static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)
+static int modeset_init_intf(struct mdp5_kms *mdp5_kms,
+			     struct mdp5_interface *intf)
 {
 	struct drm_device *dev = mdp5_kms->dev;
 	struct msm_drm_private *priv = dev->dev_private;
-	const struct mdp5_cfg_hw *hw_cfg =
-					mdp5_cfg_get_hw_config(mdp5_kms->cfg);
-	enum mdp5_intf_type intf_type = hw_cfg->intf.connect[intf_num];
 	struct mdp5_ctl_manager *ctlm = mdp5_kms->ctlm;
 	struct mdp5_ctl *ctl;
 	struct drm_encoder *encoder;
 	int ret = 0;
 
-	switch (intf_type) {
-	case INTF_DISABLED:
-		break;
+	switch (intf->type) {
 	case INTF_eDP:
 		if (!priv->edp)
 			break;
 
-		ctl = mdp5_ctlm_request(ctlm, intf_num);
+		ctl = mdp5_ctlm_request(ctlm, intf->num);
 		if (!ctl) {
 			ret = -EINVAL;
 			break;
 		}
 
-		encoder = construct_encoder(mdp5_kms, INTF_eDP, intf_num, ctl);
+		encoder = construct_encoder(mdp5_kms, intf, ctl);
 		if (IS_ERR(encoder)) {
 			ret = PTR_ERR(encoder);
 			break;
@@ -346,13 +341,13 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)
 		if (!priv->hdmi)
 			break;
 
-		ctl = mdp5_ctlm_request(ctlm, intf_num);
+		ctl = mdp5_ctlm_request(ctlm, intf->num);
 		if (!ctl) {
 			ret = -EINVAL;
 			break;
 		}
 
-		encoder = construct_encoder(mdp5_kms, INTF_HDMI, intf_num, ctl);
+		encoder = construct_encoder(mdp5_kms, intf, ctl);
 		if (IS_ERR(encoder)) {
 			ret = PTR_ERR(encoder);
 			break;
@@ -362,11 +357,13 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)
 		break;
 	case INTF_DSI:
 	{
-		int dsi_id = get_dsi_id_from_intf(hw_cfg, intf_num);
+		const struct mdp5_cfg_hw *hw_cfg =
+					mdp5_cfg_get_hw_config(mdp5_kms->cfg);
+		int dsi_id = get_dsi_id_from_intf(hw_cfg, intf->num);
 
 		if ((dsi_id >= ARRAY_SIZE(priv->dsi)) || (dsi_id < 0)) {
 			dev_err(dev->dev, "failed to find dsi from intf %d\n",
-				intf_num);
+				intf->num);
 			ret = -EINVAL;
 			break;
 		}
@@ -374,13 +371,13 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)
 		if (!priv->dsi[dsi_id])
 			break;
 
-		ctl = mdp5_ctlm_request(ctlm, intf_num);
+		ctl = mdp5_ctlm_request(ctlm, intf->num);
 		if (!ctl) {
 			ret = -EINVAL;
 			break;
 		}
 
-		encoder = construct_encoder(mdp5_kms, INTF_DSI, intf_num, ctl);
+		encoder = construct_encoder(mdp5_kms, intf, ctl);
 		if (IS_ERR(encoder)) {
 			ret = PTR_ERR(encoder);
 			break;
@@ -390,7 +387,7 @@ static int modeset_init_intf(struct mdp5_kms *mdp5_kms, int intf_num)
 		break;
 	}
 	default:
-		dev_err(dev->dev, "unknown intf: %d\n", intf_type);
+		dev_err(dev->dev, "unknown intf: %d\n", intf->type);
 		ret = -EINVAL;
 		break;
 	}
@@ -414,8 +411,8 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)
 	 * Construct encoders and modeset initialize connector devices
 	 * for each external display interface.
 	 */
-	for (i = 0; i < ARRAY_SIZE(hw_cfg->intf.connect); i++) {
-		ret = modeset_init_intf(mdp5_kms, i);
+	for (i = 0; i < mdp5_kms->num_intfs; i++) {
+		ret = modeset_init_intf(mdp5_kms, mdp5_kms->intfs[i]);
 		if (ret)
 			goto fail;
 	}
@@ -425,7 +422,7 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)
 	 * the MDP5 interfaces) than the number of layer mixers present in HW,
 	 * but let's be safe here anyway
 	 */
-	num_crtcs = min(priv->num_encoders, mdp5_cfg->lm.count);
+	num_crtcs = min(priv->num_encoders, mdp5_kms->num_hwmixers);
 
 	/*
 	 * Construct planes equaling the number of hw pipes, and CRTCs for the
@@ -744,6 +741,7 @@ fail:
 static void mdp5_destroy(struct platform_device *pdev)
 {
 	struct mdp5_kms *mdp5_kms = platform_get_drvdata(pdev);
+	int i;
 
 	if (mdp5_kms->ctlm)
 		mdp5_ctlm_destroy(mdp5_kms->ctlm);
@@ -752,6 +750,9 @@ static void mdp5_destroy(struct platform_device *pdev)
 	if (mdp5_kms->cfg)
 		mdp5_cfg_destroy(mdp5_kms->cfg);
 
+	for (i = 0; i < mdp5_kms->num_intfs; i++)
+		kfree(mdp5_kms->intfs[i]);
+
 	if (mdp5_kms->rpm_enabled)
 		pm_runtime_disable(&pdev->dev);
 
@@ -829,6 +830,64 @@ static int hwpipe_init(struct mdp5_kms *mdp5_kms)
 	return 0;
 }
 
+static int hwmixer_init(struct mdp5_kms *mdp5_kms)
+{
+	struct drm_device *dev = mdp5_kms->dev;
+	const struct mdp5_cfg_hw *hw_cfg;
+	int i, ret;
+
+	hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
+
+	for (i = 0; i < hw_cfg->lm.count; i++) {
+		struct mdp5_hw_mixer *mixer;
+
+		mixer = mdp5_mixer_init(&hw_cfg->lm.instances[i]);
+		if (IS_ERR(mixer)) {
+			ret = PTR_ERR(mixer);
+			dev_err(dev->dev, "failed to construct LM%d (%d)\n",
+				i, ret);
+			return ret;
+		}
+
+		mixer->idx = mdp5_kms->num_hwmixers;
+		mdp5_kms->hwmixers[mdp5_kms->num_hwmixers++] = mixer;
+	}
+
+	return 0;
+}
+
+static int interface_init(struct mdp5_kms *mdp5_kms)
+{
+	struct drm_device *dev = mdp5_kms->dev;
+	const struct mdp5_cfg_hw *hw_cfg;
+	const enum mdp5_intf_type *intf_types;
+	int i;
+
+	hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
+	intf_types = hw_cfg->intf.connect;
+
+	for (i = 0; i < ARRAY_SIZE(hw_cfg->intf.connect); i++) {
+		struct mdp5_interface *intf;
+
+		if (intf_types[i] == INTF_DISABLED)
+			continue;
+
+		intf = kzalloc(sizeof(*intf), GFP_KERNEL);
+		if (!intf) {
+			dev_err(dev->dev, "failed to construct INTF%d\n", i);
+			return -ENOMEM;
+		}
+
+		intf->num = i;
+		intf->type = intf_types[i];
+		intf->mode = MDP5_INTF_MODE_NONE;
+		intf->idx = mdp5_kms->num_intfs;
+		mdp5_kms->intfs[mdp5_kms->num_intfs++] = intf;
+	}
+
+	return 0;
+}
+
 static int mdp5_init(struct platform_device *pdev, struct drm_device *dev)
 {
 	struct msm_drm_private *priv = dev->dev_private;
@@ -929,6 +988,14 @@ static int mdp5_init(struct platform_device *pdev, struct drm_device *dev)
 	if (ret)
 		goto fail;
 
+	ret = hwmixer_init(mdp5_kms);
+	if (ret)
+		goto fail;
+
+	ret = interface_init(mdp5_kms);
+	if (ret)
+		goto fail;
+
 	/* set uninit-ed kms */
 	priv->kms = &mdp5_kms->base.base;
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h
index 9de471191eba..8bdb7ee4983b 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h
@@ -23,8 +23,9 @@
 #include "mdp/mdp_kms.h"
 #include "mdp5_cfg.h"	/* must be included before mdp5.xml.h */
 #include "mdp5.xml.h"
-#include "mdp5_ctl.h"
 #include "mdp5_pipe.h"
+#include "mdp5_mixer.h"
+#include "mdp5_ctl.h"
 #include "mdp5_smp.h"
 
 struct mdp5_state;
@@ -39,6 +40,12 @@ struct mdp5_kms {
 	unsigned num_hwpipes;
 	struct mdp5_hw_pipe *hwpipes[SSPP_MAX];
 
+	unsigned num_hwmixers;
+	struct mdp5_hw_mixer *hwmixers[8];
+
+	unsigned num_intfs;
+	struct mdp5_interface *intfs[5];
+
 	struct mdp5_cfg_handler *cfg;
 	uint32_t caps;	/* MDP capabilities (MDP_CAP_XXX bits) */
 
@@ -83,6 +90,7 @@ struct mdp5_kms {
  */
 struct mdp5_state {
 	struct mdp5_hw_pipe_state hwpipe;
+	struct mdp5_hw_mixer_state hwmixer;
 	struct mdp5_smp_state smp;
 };
 
@@ -96,6 +104,7 @@ struct mdp5_plane_state {
 	struct drm_plane_state base;
 
 	struct mdp5_hw_pipe *hwpipe;
+	struct mdp5_hw_pipe *r_hwpipe;	/* right hwpipe */
 
 	/* aligned with property */
 	uint8_t premultiplied;
@@ -108,6 +117,28 @@ struct mdp5_plane_state {
 #define to_mdp5_plane_state(x) \
 		container_of(x, struct mdp5_plane_state, base)
 
+struct mdp5_pipeline {
+	struct mdp5_interface *intf;
+	struct mdp5_hw_mixer *mixer;
+	struct mdp5_hw_mixer *r_mixer;	/* right mixer */
+};
+
+struct mdp5_crtc_state {
+	struct drm_crtc_state base;
+
+	struct mdp5_ctl *ctl;
+	struct mdp5_pipeline pipeline;
+
+	/* these are derivatives of intf/mixer state in mdp5_pipeline */
+	u32 vblank_irqmask;
+	u32 err_irqmask;
+	u32 pp_done_irqmask;
+
+	bool cmd_mode;
+};
+#define to_mdp5_crtc_state(x) \
+		container_of(x, struct mdp5_crtc_state, base)
+
 enum mdp5_intf_mode {
 	MDP5_INTF_MODE_NONE = 0,
 
@@ -121,6 +152,7 @@ enum mdp5_intf_mode {
 };
 
 struct mdp5_interface {
+	int idx;
 	int num; /* display interface number */
 	enum mdp5_intf_type type;
 	enum mdp5_intf_mode mode;
@@ -128,11 +160,11 @@ struct mdp5_interface {
 
 struct mdp5_encoder {
 	struct drm_encoder base;
-	struct mdp5_interface intf;
 	spinlock_t intf_lock;	/* protect REG_MDP5_INTF_* registers */
 	bool enabled;
 	uint32_t bsc;
 
+	struct mdp5_interface *intf;
 	struct mdp5_ctl *ctl;
 };
 #define to_mdp5_encoder(x) container_of(x, struct mdp5_encoder, base)
@@ -197,8 +229,8 @@ static inline uint32_t intf2err(int intf_num)
 	}
 }
 
-#define GET_PING_PONG_ID(layer_mixer)	((layer_mixer == 5) ? 3 : layer_mixer)
-static inline uint32_t intf2vblank(int lm, struct mdp5_interface *intf)
+static inline uint32_t intf2vblank(struct mdp5_hw_mixer *mixer,
+				   struct mdp5_interface *intf)
 {
 	/*
 	 * In case of DSI Command Mode, the Ping Pong's read pointer IRQ
@@ -208,7 +240,7 @@ static inline uint32_t intf2vblank(int lm, struct mdp5_interface *intf)
 
 	if ((intf->type == INTF_DSI) &&
 			(intf->mode == MDP5_INTF_DSI_MODE_COMMAND))
-		return MDP5_IRQ_PING_PONG_0_RD_PTR << GET_PING_PONG_ID(lm);
+		return MDP5_IRQ_PING_PONG_0_RD_PTR << mixer->pp;
 
 	if (intf->type == INTF_WB)
 		return MDP5_IRQ_WB_2_DONE;
@@ -222,9 +254,9 @@ static inline uint32_t intf2vblank(int lm, struct mdp5_interface *intf)
 	}
 }
 
-static inline uint32_t lm2ppdone(int lm)
+static inline uint32_t lm2ppdone(struct mdp5_hw_mixer *mixer)
 {
-	return MDP5_IRQ_PING_PONG_0_DONE << GET_PING_PONG_ID(lm);
+	return MDP5_IRQ_PING_PONG_0_DONE << mixer->pp;
 }
 
 int mdp5_disable(struct mdp5_kms *mdp5_kms);
@@ -243,15 +275,16 @@ void mdp5_irq_domain_fini(struct mdp5_kms *mdp5_kms);
 
 uint32_t mdp5_plane_get_flush(struct drm_plane *plane);
 enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane);
+enum mdp5_pipe mdp5_plane_right_pipe(struct drm_plane *plane);
 struct drm_plane *mdp5_plane_init(struct drm_device *dev,
 				  enum drm_plane_type type);
 
 struct mdp5_ctl *mdp5_crtc_get_ctl(struct drm_crtc *crtc);
 uint32_t mdp5_crtc_vblank(struct drm_crtc *crtc);
 
-int mdp5_crtc_get_lm(struct drm_crtc *crtc);
-void mdp5_crtc_set_pipeline(struct drm_crtc *crtc,
-		struct mdp5_interface *intf, struct mdp5_ctl *ctl);
+struct mdp5_hw_mixer *mdp5_crtc_get_mixer(struct drm_crtc *crtc);
+struct mdp5_pipeline *mdp5_crtc_get_pipeline(struct drm_crtc *crtc);
+void mdp5_crtc_set_pipeline(struct drm_crtc *crtc);
 void mdp5_crtc_wait_for_commit_done(struct drm_crtc *crtc);
 struct drm_crtc *mdp5_crtc_init(struct drm_device *dev,
 				struct drm_plane *plane,
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_mixer.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_mixer.c
new file mode 100644
index 000000000000..8a00991f03c7
--- /dev/null
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_mixer.c
@@ -0,0 +1,172 @@
+/*
+ * Copyright (C) 2017 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "mdp5_kms.h"
+
+/*
+ * As of now, there are only 2 combinations possible for source split:
+ *
+ * Left | Right
+ * -----|------
+ *  LM0 | LM1
+ *  LM2 | LM5
+ *
+ */
+static int lm_right_pair[] = { 1, -1, 5, -1, -1, -1 };
+
+static int get_right_pair_idx(struct mdp5_kms *mdp5_kms, int lm)
+{
+	int i;
+	int pair_lm;
+
+	pair_lm = lm_right_pair[lm];
+	if (pair_lm < 0)
+		return -EINVAL;
+
+	for (i = 0; i < mdp5_kms->num_hwmixers; i++) {
+		struct mdp5_hw_mixer *mixer = mdp5_kms->hwmixers[i];
+
+		if (mixer->lm == pair_lm)
+			return mixer->idx;
+	}
+
+	return -1;
+}
+
+int mdp5_mixer_assign(struct drm_atomic_state *s, struct drm_crtc *crtc,
+		      uint32_t caps, struct mdp5_hw_mixer **mixer,
+		      struct mdp5_hw_mixer **r_mixer)
+{
+	struct msm_drm_private *priv = s->dev->dev_private;
+	struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(priv->kms));
+	struct mdp5_state *state = mdp5_get_state(s);
+	struct mdp5_hw_mixer_state *new_state;
+	int i;
+
+	if (IS_ERR(state))
+		return PTR_ERR(state);
+
+	new_state = &state->hwmixer;
+
+	for (i = 0; i < mdp5_kms->num_hwmixers; i++) {
+		struct mdp5_hw_mixer *cur = mdp5_kms->hwmixers[i];
+
+		/*
+		 * skip if already in-use by a different CRTC. If there is a
+		 * mixer already assigned to this CRTC, it means this call is
+		 * a request to get an additional right mixer. Assume that the
+		 * existing mixer is the 'left' one, and try to see if we can
+		 * get its corresponding 'right' pair.
+		 */
+		if (new_state->hwmixer_to_crtc[cur->idx] &&
+		    new_state->hwmixer_to_crtc[cur->idx] != crtc)
+			continue;
+
+		/* skip if doesn't support some required caps: */
+		if (caps & ~cur->caps)
+			continue;
+
+		if (r_mixer) {
+			int pair_idx;
+
+			pair_idx = get_right_pair_idx(mdp5_kms, cur->lm);
+			if (pair_idx < 0)
+				return -EINVAL;
+
+			if (new_state->hwmixer_to_crtc[pair_idx])
+				continue;
+
+			*r_mixer = mdp5_kms->hwmixers[pair_idx];
+		}
+
+		/*
+		 * prefer a pair-able LM over an unpairable one. We can
+		 * switch the CRTC from Normal mode to Source Split mode
+		 * without requiring a full modeset if we had already
+		 * assigned this CRTC a pair-able LM.
+		 *
+		 * TODO: There will be assignment sequences which would
+		 * result in the CRTC requiring a full modeset, even
+		 * if we have the LM resources to prevent it. For a platform
+		 * with a few displays, we don't run out of pair-able LMs
+		 * so easily. For now, ignore the possibility of requiring
+		 * a full modeset.
+		 */
+		if (!(*mixer) || cur->caps & MDP_LM_CAP_PAIR)
+			*mixer = cur;
+	}
+
+	if (!(*mixer))
+		return -ENOMEM;
+
+	if (r_mixer && !(*r_mixer))
+		return -ENOMEM;
+
+	DBG("assigning Layer Mixer %d to crtc %s", (*mixer)->lm, crtc->name);
+
+	new_state->hwmixer_to_crtc[(*mixer)->idx] = crtc;
+	if (r_mixer) {
+		DBG("assigning Right Layer Mixer %d to crtc %s", (*r_mixer)->lm,
+		    crtc->name);
+		new_state->hwmixer_to_crtc[(*r_mixer)->idx] = crtc;
+	}
+
+	return 0;
+}
+
+void mdp5_mixer_release(struct drm_atomic_state *s, struct mdp5_hw_mixer *mixer)
+{
+	struct mdp5_state *state = mdp5_get_state(s);
+	struct mdp5_hw_mixer_state *new_state = &state->hwmixer;
+
+	if (!mixer)
+		return;
+
+	if (WARN_ON(!new_state->hwmixer_to_crtc[mixer->idx]))
+		return;
+
+	DBG("%s: release from crtc %s", mixer->name,
+	    new_state->hwmixer_to_crtc[mixer->idx]->name);
+
+	new_state->hwmixer_to_crtc[mixer->idx] = NULL;
+}
+
+void mdp5_mixer_destroy(struct mdp5_hw_mixer *mixer)
+{
+	kfree(mixer);
+}
+
+static const char * const mixer_names[] = {
+	"LM0", "LM1", "LM2", "LM3", "LM4", "LM5",
+};
+
+struct mdp5_hw_mixer *mdp5_mixer_init(const struct mdp5_lm_instance *lm)
+{
+	struct mdp5_hw_mixer *mixer;
+
+	mixer = kzalloc(sizeof(*mixer), GFP_KERNEL);
+	if (!mixer)
+		return ERR_PTR(-ENOMEM);
+
+	mixer->name = mixer_names[lm->id];
+	mixer->lm = lm->id;
+	mixer->caps = lm->caps;
+	mixer->pp = lm->pp;
+	mixer->dspp = lm->dspp;
+	mixer->flush_mask = mdp_ctl_flush_mask_lm(lm->id);
+
+	return mixer;
+}
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_mixer.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_mixer.h
new file mode 100644
index 000000000000..9be94f567fbd
--- /dev/null
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_mixer.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C) 2017 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __MDP5_LM_H__
+#define __MDP5_LM_H__
+
+/* represents a hw Layer Mixer, one (or more) is dynamically assigned to a crtc */
+struct mdp5_hw_mixer {
+	int idx;
+
+	const char *name;
+
+	int lm;			/* the LM instance # */
+	uint32_t caps;
+	int pp;
+	int dspp;
+
+	uint32_t flush_mask;      /* used to commit LM registers */
+};
+
+/* global atomic state of assignment between CRTCs and Layer Mixers: */
+struct mdp5_hw_mixer_state {
+	struct drm_crtc *hwmixer_to_crtc[8];
+};
+
+struct mdp5_hw_mixer *mdp5_mixer_init(const struct mdp5_lm_instance *lm);
+void mdp5_mixer_destroy(struct mdp5_hw_mixer *lm);
+int mdp5_mixer_assign(struct drm_atomic_state *s, struct drm_crtc *crtc,
+		      uint32_t caps, struct mdp5_hw_mixer **mixer,
+		      struct mdp5_hw_mixer **r_mixer);
+void mdp5_mixer_release(struct drm_atomic_state *s,
+			struct mdp5_hw_mixer *mixer);
+
+#endif /* __MDP5_LM_H__ */
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c
index 35c4dabb0c0c..2bfac3712685 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.c
@@ -135,7 +135,5 @@ struct mdp5_hw_pipe *mdp5_pipe_init(enum mdp5_pipe pipe,
 	hwpipe->caps = caps;
 	hwpipe->flush_mask = mdp_ctl_flush_mask_pipe(pipe);
 
-	spin_lock_init(&hwpipe->pipe_lock);
-
 	return hwpipe;
 }
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h
index 238901987e00..924c3e6f9517 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_pipe.h
@@ -28,7 +28,6 @@ struct mdp5_hw_pipe {
 	const char *name;
 	enum mdp5_pipe pipe;
 
-	spinlock_t pipe_lock;     /* protect REG_MDP5_PIPE_* registers */
 	uint32_t reg_offset;
 	uint32_t caps;
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c
index 60a5451ae0b9..a38c5fe6cc19 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c
@@ -22,6 +22,8 @@
 struct mdp5_plane {
 	struct drm_plane base;
 
+	spinlock_t pipe_lock;     /* protect REG_MDP5_PIPE_* registers */
+
 	uint32_t nformats;
 	uint32_t formats[32];
 };
@@ -40,9 +42,6 @@ static int mdp5_update_cursor_plane_legacy(struct drm_plane *plane,
 		uint32_t src_w, uint32_t src_h,
 		struct drm_modeset_acquire_ctx *ctx);
 
-static void set_scanout_locked(struct drm_plane *plane,
-		struct drm_framebuffer *fb);
-
 static struct mdp5_kms *get_kms(struct drm_plane *plane)
 {
 	struct msm_drm_private *priv = plane->dev->dev_private;
@@ -178,9 +177,14 @@ mdp5_plane_atomic_print_state(struct drm_printer *p,
 		const struct drm_plane_state *state)
 {
 	struct mdp5_plane_state *pstate = to_mdp5_plane_state(state);
+	struct mdp5_kms *mdp5_kms = get_kms(state->plane);
 
 	drm_printf(p, "\thwpipe=%s\n", pstate->hwpipe ?
 			pstate->hwpipe->name : "(null)");
+	if (mdp5_kms->caps & MDP_CAP_SRC_SPLIT)
+		drm_printf(p, "\tright-hwpipe=%s\n",
+			   pstate->r_hwpipe ? pstate->r_hwpipe->name :
+					      "(null)");
 	drm_printf(p, "\tpremultiplied=%u\n", pstate->premultiplied);
 	drm_printf(p, "\tzpos=%u\n", pstate->zpos);
 	drm_printf(p, "\talpha=%u\n", pstate->alpha);
@@ -300,7 +304,9 @@ static int mdp5_plane_atomic_check_with_state(struct drm_crtc_state *crtc_state,
 	struct drm_plane_state *old_state = plane->state;
 	struct mdp5_cfg *config = mdp5_cfg_get_config(get_kms(plane)->cfg);
 	bool new_hwpipe = false;
+	bool need_right_hwpipe = false;
 	uint32_t max_width, max_height;
+	bool out_of_bounds = false;
 	uint32_t caps = 0;
 	struct drm_rect clip;
 	int min_scale, max_scale;
@@ -313,7 +319,23 @@ static int mdp5_plane_atomic_check_with_state(struct drm_crtc_state *crtc_state,
 	max_height = config->hw->lm.max_height << 16;
 
 	/* Make sure source dimensions are within bounds. */
-	if ((state->src_w > max_width) || (state->src_h > max_height)) {
+	if (state->src_h > max_height)
+		out_of_bounds = true;
+
+	if (state->src_w > max_width) {
+		/* If source split is supported, we can go up to 2x
+		 * the max LM width, but we'd need to stage another
+		 * hwpipe to the right LM. So, the drm_plane would
+		 * consist of 2 hwpipes.
+		 */
+		if (config->hw->mdp.caps & MDP_CAP_SRC_SPLIT &&
+		    (state->src_w <= 2 * max_width))
+			need_right_hwpipe = true;
+		else
+			out_of_bounds = true;
+	}
+
+	if (out_of_bounds) {
 		struct drm_rect src = drm_plane_state_src(state);
 		DBG("Invalid source size "DRM_RECT_FP_FMT,
 				DRM_RECT_FP_ARG(&src));
@@ -364,6 +386,15 @@ static int mdp5_plane_atomic_check_with_state(struct drm_crtc_state *crtc_state,
 		if (!mdp5_state->hwpipe || (caps & ~mdp5_state->hwpipe->caps))
 			new_hwpipe = true;
 
+		/*
+		 * (re)allocte hw pipe if we're either requesting for 2 hw pipes
+		 * or we're switching from 2 hw pipes to 1 hw pipe because the
+		 * new src_w can be supported by 1 hw pipe itself.
+		 */
+		if ((need_right_hwpipe && !mdp5_state->r_hwpipe) ||
+		    (!need_right_hwpipe && mdp5_state->r_hwpipe))
+			new_hwpipe = true;
+
 		if (mdp5_kms->smp) {
 			const struct mdp_format *format =
 				to_mdp_format(msm_framebuffer_format(state->fb));
@@ -382,13 +413,36 @@ static int mdp5_plane_atomic_check_with_state(struct drm_crtc_state *crtc_state,
 			 * it available for other planes?
 			 */
 			struct mdp5_hw_pipe *old_hwpipe = mdp5_state->hwpipe;
+			struct mdp5_hw_pipe *old_right_hwpipe =
+							  mdp5_state->r_hwpipe;
+
 			mdp5_state->hwpipe = mdp5_pipe_assign(state->state,
 					plane, caps, blkcfg);
 			if (IS_ERR(mdp5_state->hwpipe)) {
 				DBG("%s: failed to assign hwpipe!", plane->name);
 				return PTR_ERR(mdp5_state->hwpipe);
 			}
+
+			if (need_right_hwpipe) {
+				mdp5_state->r_hwpipe =
+					mdp5_pipe_assign(state->state, plane,
+							 caps, blkcfg);
+				if (IS_ERR(mdp5_state->r_hwpipe)) {
+					DBG("%s: failed to assign right hwpipe",
+					    plane->name);
+					return PTR_ERR(mdp5_state->r_hwpipe);
+				}
+			} else {
+				/*
+				 * set it to NULL so that the driver knows we
+				 * don't have a right hwpipe when committing a
+				 * new state
+				 */
+				mdp5_state->r_hwpipe = NULL;
+			}
+
 			mdp5_pipe_release(state->state, old_hwpipe);
+			mdp5_pipe_release(state->state, old_right_hwpipe);
 		}
 	}
 
@@ -437,13 +491,10 @@ static const struct drm_plane_helper_funcs mdp5_plane_helper_funcs = {
 		.atomic_update = mdp5_plane_atomic_update,
 };
 
-static void set_scanout_locked(struct drm_plane *plane,
-		struct drm_framebuffer *fb)
+static void set_scanout_locked(struct mdp5_kms *mdp5_kms,
+			       enum mdp5_pipe pipe,
+			       struct drm_framebuffer *fb)
 {
-	struct mdp5_kms *mdp5_kms = get_kms(plane);
-	struct mdp5_hw_pipe *hwpipe = to_mdp5_plane_state(plane->state)->hwpipe;
-	enum mdp5_pipe pipe = hwpipe->pipe;
-
 	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_A(pipe),
 			MDP5_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) |
 			MDP5_PIPE_SRC_STRIDE_A_P1(fb->pitches[1]));
@@ -460,8 +511,6 @@ static void set_scanout_locked(struct drm_plane *plane,
 			msm_framebuffer_iova(fb, mdp5_kms->id, 2));
 	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe),
 			msm_framebuffer_iova(fb, mdp5_kms->id, 3));
-
-	plane->fb = fb;
 }
 
 /* Note: mdp5_plane->pipe_lock must be locked */
@@ -714,21 +763,129 @@ static void mdp5_write_pixel_ext(struct mdp5_kms *mdp5_kms, enum mdp5_pipe pipe,
 	}
 }
 
+struct pixel_ext {
+	int left[COMP_MAX];
+	int right[COMP_MAX];
+	int top[COMP_MAX];
+	int bottom[COMP_MAX];
+};
+
+struct phase_step {
+	u32 x[COMP_MAX];
+	u32 y[COMP_MAX];
+};
+
+static void mdp5_hwpipe_mode_set(struct mdp5_kms *mdp5_kms,
+				 struct mdp5_hw_pipe *hwpipe,
+				 struct drm_framebuffer *fb,
+				 struct phase_step *step,
+				 struct pixel_ext *pe,
+				 u32 scale_config, u32 hdecm, u32 vdecm,
+				 bool hflip, bool vflip,
+				 int crtc_x, int crtc_y,
+				 unsigned int crtc_w, unsigned int crtc_h,
+				 u32 src_img_w, u32 src_img_h,
+				 u32 src_x, u32 src_y,
+				 u32 src_w, u32 src_h)
+{
+	enum mdp5_pipe pipe = hwpipe->pipe;
+	bool has_pe = hwpipe->caps & MDP_PIPE_CAP_SW_PIX_EXT;
+	const struct mdp_format *format =
+			to_mdp_format(msm_framebuffer_format(fb));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_IMG_SIZE(pipe),
+			MDP5_PIPE_SRC_IMG_SIZE_WIDTH(src_img_w) |
+			MDP5_PIPE_SRC_IMG_SIZE_HEIGHT(src_img_h));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_SIZE(pipe),
+			MDP5_PIPE_SRC_SIZE_WIDTH(src_w) |
+			MDP5_PIPE_SRC_SIZE_HEIGHT(src_h));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_XY(pipe),
+			MDP5_PIPE_SRC_XY_X(src_x) |
+			MDP5_PIPE_SRC_XY_Y(src_y));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_OUT_SIZE(pipe),
+			MDP5_PIPE_OUT_SIZE_WIDTH(crtc_w) |
+			MDP5_PIPE_OUT_SIZE_HEIGHT(crtc_h));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_OUT_XY(pipe),
+			MDP5_PIPE_OUT_XY_X(crtc_x) |
+			MDP5_PIPE_OUT_XY_Y(crtc_y));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_FORMAT(pipe),
+			MDP5_PIPE_SRC_FORMAT_A_BPC(format->bpc_a) |
+			MDP5_PIPE_SRC_FORMAT_R_BPC(format->bpc_r) |
+			MDP5_PIPE_SRC_FORMAT_G_BPC(format->bpc_g) |
+			MDP5_PIPE_SRC_FORMAT_B_BPC(format->bpc_b) |
+			COND(format->alpha_enable, MDP5_PIPE_SRC_FORMAT_ALPHA_ENABLE) |
+			MDP5_PIPE_SRC_FORMAT_CPP(format->cpp - 1) |
+			MDP5_PIPE_SRC_FORMAT_UNPACK_COUNT(format->unpack_count - 1) |
+			COND(format->unpack_tight, MDP5_PIPE_SRC_FORMAT_UNPACK_TIGHT) |
+			MDP5_PIPE_SRC_FORMAT_FETCH_TYPE(format->fetch_type) |
+			MDP5_PIPE_SRC_FORMAT_CHROMA_SAMP(format->chroma_sample));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_UNPACK(pipe),
+			MDP5_PIPE_SRC_UNPACK_ELEM0(format->unpack[0]) |
+			MDP5_PIPE_SRC_UNPACK_ELEM1(format->unpack[1]) |
+			MDP5_PIPE_SRC_UNPACK_ELEM2(format->unpack[2]) |
+			MDP5_PIPE_SRC_UNPACK_ELEM3(format->unpack[3]));
+
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_OP_MODE(pipe),
+			(hflip ? MDP5_PIPE_SRC_OP_MODE_FLIP_LR : 0) |
+			(vflip ? MDP5_PIPE_SRC_OP_MODE_FLIP_UD : 0) |
+			COND(has_pe, MDP5_PIPE_SRC_OP_MODE_SW_PIX_EXT_OVERRIDE) |
+			MDP5_PIPE_SRC_OP_MODE_BWC(BWC_LOSSLESS));
+
+	/* not using secure mode: */
+	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_ADDR_SW_STATUS(pipe), 0);
+
+	if (hwpipe->caps & MDP_PIPE_CAP_SW_PIX_EXT)
+		mdp5_write_pixel_ext(mdp5_kms, pipe, format,
+				src_w, pe->left, pe->right,
+				src_h, pe->top, pe->bottom);
+
+	if (hwpipe->caps & MDP_PIPE_CAP_SCALE) {
+		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_PHASE_STEP_X(pipe),
+				step->x[COMP_0]);
+		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_PHASE_STEP_Y(pipe),
+				step->y[COMP_0]);
+		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_CR_PHASE_STEP_X(pipe),
+				step->x[COMP_1_2]);
+		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_CR_PHASE_STEP_Y(pipe),
+				step->y[COMP_1_2]);
+		mdp5_write(mdp5_kms, REG_MDP5_PIPE_DECIMATION(pipe),
+				MDP5_PIPE_DECIMATION_VERT(vdecm) |
+				MDP5_PIPE_DECIMATION_HORZ(hdecm));
+		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_CONFIG(pipe),
+			   scale_config);
+	}
+
+	if (hwpipe->caps & MDP_PIPE_CAP_CSC) {
+		if (MDP_FORMAT_IS_YUV(format))
+			csc_enable(mdp5_kms, pipe,
+					mdp_get_default_csc_cfg(CSC_YUV2RGB));
+		else
+			csc_disable(mdp5_kms, pipe);
+	}
+
+	set_scanout_locked(mdp5_kms, pipe, fb);
+}
 
 static int mdp5_plane_mode_set(struct drm_plane *plane,
 		struct drm_crtc *crtc, struct drm_framebuffer *fb,
 		struct drm_rect *src, struct drm_rect *dest)
 {
+	struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
 	struct drm_plane_state *pstate = plane->state;
 	struct mdp5_hw_pipe *hwpipe = to_mdp5_plane_state(pstate)->hwpipe;
 	struct mdp5_kms *mdp5_kms = get_kms(plane);
 	enum mdp5_pipe pipe = hwpipe->pipe;
+	struct mdp5_hw_pipe *right_hwpipe;
 	const struct mdp_format *format;
 	uint32_t nplanes, config = 0;
-	uint32_t phasex_step[COMP_MAX] = {0,}, phasey_step[COMP_MAX] = {0,};
-	bool pe = hwpipe->caps & MDP_PIPE_CAP_SW_PIX_EXT;
-	int pe_left[COMP_MAX], pe_right[COMP_MAX];
-	int pe_top[COMP_MAX], pe_bottom[COMP_MAX];
+	struct phase_step step = { 0 };
+	struct pixel_ext pe = { 0 };
 	uint32_t hdecm = 0, vdecm = 0;
 	uint32_t pix_format;
 	unsigned int rotation;
@@ -737,6 +894,9 @@ static int mdp5_plane_mode_set(struct drm_plane *plane,
 	unsigned int crtc_w, crtc_h;
 	uint32_t src_x, src_y;
 	uint32_t src_w, src_h;
+	uint32_t src_img_w, src_img_h;
+	uint32_t src_x_r;
+	int crtc_x_r;
 	unsigned long flags;
 	int ret;
 
@@ -765,23 +925,41 @@ static int mdp5_plane_mode_set(struct drm_plane *plane,
 	src_w = src_w >> 16;
 	src_h = src_h >> 16;
 
+	src_img_w = min(fb->width, src_w);
+	src_img_h = min(fb->height, src_h);
+
 	DBG("%s: FB[%u] %u,%u,%u,%u -> CRTC[%u] %d,%d,%u,%u", plane->name,
 			fb->base.id, src_x, src_y, src_w, src_h,
 			crtc->base.id, crtc_x, crtc_y, crtc_w, crtc_h);
 
-	ret = calc_scalex_steps(plane, pix_format, src_w, crtc_w, phasex_step);
+	right_hwpipe = to_mdp5_plane_state(pstate)->r_hwpipe;
+	if (right_hwpipe) {
+		/*
+		 * if the plane comprises of 2 hw pipes, assume that the width
+		 * is split equally across them. The only parameters that varies
+		 * between the 2 pipes are src_x and crtc_x
+		 */
+		crtc_w /= 2;
+		src_w /= 2;
+		src_img_w /= 2;
+
+		crtc_x_r = crtc_x + crtc_w;
+		src_x_r = src_x + src_w;
+	}
+
+	ret = calc_scalex_steps(plane, pix_format, src_w, crtc_w, step.x);
 	if (ret)
 		return ret;
 
-	ret = calc_scaley_steps(plane, pix_format, src_h, crtc_h, phasey_step);
+	ret = calc_scaley_steps(plane, pix_format, src_h, crtc_h, step.y);
 	if (ret)
 		return ret;
 
 	if (hwpipe->caps & MDP_PIPE_CAP_SW_PIX_EXT) {
-		calc_pixel_ext(format, src_w, crtc_w, phasex_step,
-					 pe_left, pe_right, true);
-		calc_pixel_ext(format, src_h, crtc_h, phasey_step,
-					pe_top, pe_bottom, false);
+		calc_pixel_ext(format, src_w, crtc_w, step.x,
+			       pe.left, pe.right, true);
+		calc_pixel_ext(format, src_h, crtc_h, step.y,
+			       pe.top, pe.bottom, false);
 	}
 
 	/* TODO calc hdecm, vdecm */
@@ -798,86 +976,23 @@ static int mdp5_plane_mode_set(struct drm_plane *plane,
 	hflip = !!(rotation & DRM_REFLECT_X);
 	vflip = !!(rotation & DRM_REFLECT_Y);
 
-	spin_lock_irqsave(&hwpipe->pipe_lock, flags);
-
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_IMG_SIZE(pipe),
-			MDP5_PIPE_SRC_IMG_SIZE_WIDTH(min(fb->width, src_w)) |
-			MDP5_PIPE_SRC_IMG_SIZE_HEIGHT(min(fb->height, src_h)));
-
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_SIZE(pipe),
-			MDP5_PIPE_SRC_SIZE_WIDTH(src_w) |
-			MDP5_PIPE_SRC_SIZE_HEIGHT(src_h));
+	spin_lock_irqsave(&mdp5_plane->pipe_lock, flags);
 
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_XY(pipe),
-			MDP5_PIPE_SRC_XY_X(src_x) |
-			MDP5_PIPE_SRC_XY_Y(src_y));
+	mdp5_hwpipe_mode_set(mdp5_kms, hwpipe, fb, &step, &pe,
+			     config, hdecm, vdecm, hflip, vflip,
+			     crtc_x, crtc_y, crtc_w, crtc_h,
+			     src_img_w, src_img_h,
+			     src_x, src_y, src_w, src_h);
+	if (right_hwpipe)
+		mdp5_hwpipe_mode_set(mdp5_kms, right_hwpipe, fb, &step, &pe,
+				     config, hdecm, vdecm, hflip, vflip,
+				     crtc_x_r, crtc_y, crtc_w, crtc_h,
+				     src_img_w, src_img_h,
+				     src_x_r, src_y, src_w, src_h);
 
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_OUT_SIZE(pipe),
-			MDP5_PIPE_OUT_SIZE_WIDTH(crtc_w) |
-			MDP5_PIPE_OUT_SIZE_HEIGHT(crtc_h));
+	spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags);
 
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_OUT_XY(pipe),
-			MDP5_PIPE_OUT_XY_X(crtc_x) |
-			MDP5_PIPE_OUT_XY_Y(crtc_y));
-
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_FORMAT(pipe),
-			MDP5_PIPE_SRC_FORMAT_A_BPC(format->bpc_a) |
-			MDP5_PIPE_SRC_FORMAT_R_BPC(format->bpc_r) |
-			MDP5_PIPE_SRC_FORMAT_G_BPC(format->bpc_g) |
-			MDP5_PIPE_SRC_FORMAT_B_BPC(format->bpc_b) |
-			COND(format->alpha_enable, MDP5_PIPE_SRC_FORMAT_ALPHA_ENABLE) |
-			MDP5_PIPE_SRC_FORMAT_CPP(format->cpp - 1) |
-			MDP5_PIPE_SRC_FORMAT_UNPACK_COUNT(format->unpack_count - 1) |
-			COND(format->unpack_tight, MDP5_PIPE_SRC_FORMAT_UNPACK_TIGHT) |
-			MDP5_PIPE_SRC_FORMAT_FETCH_TYPE(format->fetch_type) |
-			MDP5_PIPE_SRC_FORMAT_CHROMA_SAMP(format->chroma_sample));
-
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_UNPACK(pipe),
-			MDP5_PIPE_SRC_UNPACK_ELEM0(format->unpack[0]) |
-			MDP5_PIPE_SRC_UNPACK_ELEM1(format->unpack[1]) |
-			MDP5_PIPE_SRC_UNPACK_ELEM2(format->unpack[2]) |
-			MDP5_PIPE_SRC_UNPACK_ELEM3(format->unpack[3]));
-
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_OP_MODE(pipe),
-			(hflip ? MDP5_PIPE_SRC_OP_MODE_FLIP_LR : 0) |
-			(vflip ? MDP5_PIPE_SRC_OP_MODE_FLIP_UD : 0) |
-			COND(pe, MDP5_PIPE_SRC_OP_MODE_SW_PIX_EXT_OVERRIDE) |
-			MDP5_PIPE_SRC_OP_MODE_BWC(BWC_LOSSLESS));
-
-	/* not using secure mode: */
-	mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_ADDR_SW_STATUS(pipe), 0);
-
-	if (hwpipe->caps & MDP_PIPE_CAP_SW_PIX_EXT)
-		mdp5_write_pixel_ext(mdp5_kms, pipe, format,
-				src_w, pe_left, pe_right,
-				src_h, pe_top, pe_bottom);
-
-	if (hwpipe->caps & MDP_PIPE_CAP_SCALE) {
-		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_PHASE_STEP_X(pipe),
-				phasex_step[COMP_0]);
-		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_PHASE_STEP_Y(pipe),
-				phasey_step[COMP_0]);
-		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_CR_PHASE_STEP_X(pipe),
-				phasex_step[COMP_1_2]);
-		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_CR_PHASE_STEP_Y(pipe),
-				phasey_step[COMP_1_2]);
-		mdp5_write(mdp5_kms, REG_MDP5_PIPE_DECIMATION(pipe),
-				MDP5_PIPE_DECIMATION_VERT(vdecm) |
-				MDP5_PIPE_DECIMATION_HORZ(hdecm));
-		mdp5_write(mdp5_kms, REG_MDP5_PIPE_SCALE_CONFIG(pipe), config);
-	}
-
-	if (hwpipe->caps & MDP_PIPE_CAP_CSC) {
-		if (MDP_FORMAT_IS_YUV(format))
-			csc_enable(mdp5_kms, pipe,
-					mdp_get_default_csc_cfg(CSC_YUV2RGB));
-		else
-			csc_disable(mdp5_kms, pipe);
-	}
-
-	set_scanout_locked(plane, fb);
-
-	spin_unlock_irqrestore(&hwpipe->pipe_lock, flags);
+	plane->fb = fb;
 
 	return ret;
 }
@@ -934,6 +1049,7 @@ static int mdp5_update_cursor_plane_legacy(struct drm_plane *plane,
 
 	if (new_plane_state->visible) {
 		struct mdp5_ctl *ctl;
+		struct mdp5_pipeline *pipeline = mdp5_crtc_get_pipeline(crtc);
 
 		ret = mdp5_plane_mode_set(plane, crtc, fb,
 					  &new_plane_state->src,
@@ -942,7 +1058,7 @@ static int mdp5_update_cursor_plane_legacy(struct drm_plane *plane,
 
 		ctl = mdp5_crtc_get_ctl(crtc);
 
-		mdp5_ctl_commit(ctl, mdp5_plane_get_flush(plane));
+		mdp5_ctl_commit(ctl, pipeline, mdp5_plane_get_flush(plane));
 	}
 
 	*to_mdp5_plane_state(plane_state) =
@@ -959,6 +1075,10 @@ slow:
 					      src_x, src_y, src_w, src_h, ctx);
 }
 
+/*
+ * Use this func and the one below only after the atomic state has been
+ * successfully swapped
+ */
 enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane)
 {
 	struct mdp5_plane_state *pstate = to_mdp5_plane_state(plane->state);
@@ -969,14 +1089,30 @@ enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane)
 	return pstate->hwpipe->pipe;
 }
 
+enum mdp5_pipe mdp5_plane_right_pipe(struct drm_plane *plane)
+{
+	struct mdp5_plane_state *pstate = to_mdp5_plane_state(plane->state);
+
+	if (!pstate->r_hwpipe)
+		return SSPP_NONE;
+
+	return pstate->r_hwpipe->pipe;
+}
+
 uint32_t mdp5_plane_get_flush(struct drm_plane *plane)
 {
 	struct mdp5_plane_state *pstate = to_mdp5_plane_state(plane->state);
+	u32 mask;
 
 	if (WARN_ON(!pstate->hwpipe))
 		return 0;
 
-	return pstate->hwpipe->flush_mask;
+	mask = pstate->hwpipe->flush_mask;
+
+	if (pstate->r_hwpipe)
+		mask |= pstate->r_hwpipe->flush_mask;
+
+	return mask;
 }
 
 /* initialize plane */
@@ -998,6 +1134,8 @@ struct drm_plane *mdp5_plane_init(struct drm_device *dev,
 	mdp5_plane->nformats = mdp_get_formats(mdp5_plane->formats,
 		ARRAY_SIZE(mdp5_plane->formats), false);
 
+	spin_lock_init(&mdp5_plane->pipe_lock);
+
 	if (type == DRM_PLANE_TYPE_CURSOR)
 		ret = drm_universal_plane_init(dev, plane, 0xff,
 				&mdp5_cursor_plane_funcs,
diff --git a/drivers/gpu/drm/msm/mdp/mdp_kms.h b/drivers/gpu/drm/msm/mdp/mdp_kms.h
index 7574cdfef418..1185487e7e5e 100644
--- a/drivers/gpu/drm/msm/mdp/mdp_kms.h
+++ b/drivers/gpu/drm/msm/mdp/mdp_kms.h
@@ -104,6 +104,7 @@ const struct msm_format *mdp_get_format(struct msm_kms *kms, uint32_t format);
 #define MDP_CAP_SMP		BIT(0)	/* Shared Memory Pool                 */
 #define MDP_CAP_DSC		BIT(1)	/* VESA Display Stream Compression    */
 #define MDP_CAP_CDM		BIT(2)	/* Chroma Down Module (HDMI 2.0 YUV)  */
+#define MDP_CAP_SRC_SPLIT	BIT(3)	/* Source Split of SSPPs */
 
 /* MDP pipe capabilities */
 #define MDP_PIPE_CAP_HFLIP			BIT(0)
@@ -114,6 +115,11 @@ const struct msm_format *mdp_get_format(struct msm_kms *kms, uint32_t format);
 #define MDP_PIPE_CAP_SW_PIX_EXT			BIT(5)
 #define MDP_PIPE_CAP_CURSOR			BIT(6)
 
+/* MDP layer mixer caps */
+#define MDP_LM_CAP_DISPLAY			BIT(0)
+#define MDP_LM_CAP_WB				BIT(1)
+#define MDP_LM_CAP_PAIR				BIT(2)
+
 static inline bool pipe_supports_yuv(uint32_t pipe_caps)
 {
 	return (pipe_caps & MDP_PIPE_CAP_SCALE) &&
diff --git a/drivers/gpu/drm/msm/msm_debugfs.c b/drivers/gpu/drm/msm/msm_debugfs.c
index 4f35d4eb85d0..1855182c76ce 100644
--- a/drivers/gpu/drm/msm/msm_debugfs.c
+++ b/drivers/gpu/drm/msm/msm_debugfs.c
@@ -28,7 +28,9 @@ static int msm_gpu_show(struct drm_device *dev, struct seq_file *m)
 
 	if (gpu) {
 		seq_printf(m, "%s Status:\n", gpu->name);
+		pm_runtime_get_sync(&gpu->pdev->dev);
 		gpu->funcs->show(gpu, m);
+		pm_runtime_put_sync(&gpu->pdev->dev);
 	}
 
 	return 0;
diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
index 9208e67be453..87b5695d4034 100644
--- a/drivers/gpu/drm/msm/msm_drv.c
+++ b/drivers/gpu/drm/msm/msm_drv.c
@@ -55,14 +55,13 @@ int msm_register_address_space(struct drm_device *dev,
 		struct msm_gem_address_space *aspace)
 {
 	struct msm_drm_private *priv = dev->dev_private;
-	int idx = priv->num_aspaces++;
 
-	if (WARN_ON(idx >= ARRAY_SIZE(priv->aspace)))
+	if (WARN_ON(priv->num_aspaces >= ARRAY_SIZE(priv->aspace)))
 		return -EINVAL;
 
-	priv->aspace[idx] = aspace;
+	priv->aspace[priv->num_aspaces] = aspace;
 
-	return idx;
+	return priv->num_aspaces++;
 }
 
 #ifdef CONFIG_DRM_MSM_REGISTER_LOGGING
@@ -265,6 +264,8 @@ static int msm_drm_uninit(struct device *dev)
 
 	if (gpu) {
 		mutex_lock(&ddev->struct_mutex);
+		// XXX what do we do here?
+		//pm_runtime_enable(&pdev->dev);
 		gpu->funcs->pm_suspend(gpu);
 		mutex_unlock(&ddev->struct_mutex);
 		gpu->funcs->destroy(gpu);
@@ -539,7 +540,7 @@ static int msm_open(struct drm_device *dev, struct drm_file *file)
 	return 0;
 }
 
-static void msm_preclose(struct drm_device *dev, struct drm_file *file)
+static void msm_postclose(struct drm_device *dev, struct drm_file *file)
 {
 	struct msm_drm_private *priv = dev->dev_private;
 	struct msm_file_private *ctx = file->driver_priv;
@@ -812,7 +813,7 @@ static struct drm_driver msm_driver = {
 				DRIVER_ATOMIC |
 				DRIVER_MODESET,
 	.open               = msm_open,
-	.preclose           = msm_preclose,
+	.postclose           = msm_postclose,
 	.lastclose          = msm_lastclose,
 	.irq_handler        = msm_irq,
 	.irq_preinstall     = msm_irq_preinstall,
diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h
index b885c3d5ae4d..28b6f9ba5066 100644
--- a/drivers/gpu/drm/msm/msm_drv.h
+++ b/drivers/gpu/drm/msm/msm_drv.h
@@ -191,7 +191,8 @@ void msm_gem_unmap_vma(struct msm_gem_address_space *aspace,
 int msm_gem_map_vma(struct msm_gem_address_space *aspace,
 		struct msm_gem_vma *vma, struct sg_table *sgt, int npages);
 
-void msm_gem_address_space_destroy(struct msm_gem_address_space *aspace);
+void msm_gem_address_space_put(struct msm_gem_address_space *aspace);
+
 struct msm_gem_address_space *
 msm_gem_address_space_create(struct device *dev, struct iommu_domain *domain,
 		const char *name);
diff --git a/drivers/gpu/drm/msm/msm_gem.h b/drivers/gpu/drm/msm/msm_gem.h
index 7d529516b332..1b4cf20043ea 100644
--- a/drivers/gpu/drm/msm/msm_gem.h
+++ b/drivers/gpu/drm/msm/msm_gem.h
@@ -18,6 +18,7 @@
 #ifndef __MSM_GEM_H__
 #define __MSM_GEM_H__
 
+#include <linux/kref.h>
 #include <linux/reservation.h>
 #include "msm_drv.h"
 
@@ -31,6 +32,7 @@ struct msm_gem_address_space {
 	 */
 	struct drm_mm mm;
 	struct msm_mmu *mmu;
+	struct kref kref;
 };
 
 struct msm_gem_vma {
diff --git a/drivers/gpu/drm/msm/msm_gem_submit.c b/drivers/gpu/drm/msm/msm_gem_submit.c
index 1172fe7a9252..1c545ebe6a5a 100644
--- a/drivers/gpu/drm/msm/msm_gem_submit.c
+++ b/drivers/gpu/drm/msm/msm_gem_submit.c
@@ -404,6 +404,24 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	if (MSM_PIPE_FLAGS(args->flags) & ~MSM_SUBMIT_FLAGS)
 		return -EINVAL;
 
+	if (args->flags & MSM_SUBMIT_FENCE_FD_IN) {
+		in_fence = sync_file_get_fence(args->fence_fd);
+
+		if (!in_fence)
+			return -EINVAL;
+
+		/* TODO if we get an array-fence due to userspace merging multiple
+		 * fences, we need a way to determine if all the backing fences
+		 * are from our own context..
+		 */
+
+		if (in_fence->context != gpu->fctx->context) {
+			ret = dma_fence_wait(in_fence, true);
+			if (ret)
+				return ret;
+		}
+	}
+
 	ret = mutex_lock_interruptible(&dev->struct_mutex);
 	if (ret)
 		return ret;
@@ -431,27 +449,6 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	if (ret)
 		goto out;
 
-	if (args->flags & MSM_SUBMIT_FENCE_FD_IN) {
-		in_fence = sync_file_get_fence(args->fence_fd);
-
-		if (!in_fence) {
-			ret = -EINVAL;
-			goto out;
-		}
-
-		/* TODO if we get an array-fence due to userspace merging multiple
-		 * fences, we need a way to determine if all the backing fences
-		 * are from our own context..
-		 */
-
-		if (in_fence->context != gpu->fctx->context) {
-			ret = dma_fence_wait(in_fence, true);
-			if (ret)
-				goto out;
-		}
-
-	}
-
 	if (!(args->fence & MSM_SUBMIT_NO_IMPLICIT)) {
 		ret = submit_fence_sync(submit);
 		if (ret)
diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
index b654eca7636a..f285d7e210db 100644
--- a/drivers/gpu/drm/msm/msm_gem_vma.c
+++ b/drivers/gpu/drm/msm/msm_gem_vma.c
@@ -19,6 +19,25 @@
 #include "msm_gem.h"
 #include "msm_mmu.h"
 
+static void
+msm_gem_address_space_destroy(struct kref *kref)
+{
+	struct msm_gem_address_space *aspace = container_of(kref,
+			struct msm_gem_address_space, kref);
+
+	drm_mm_takedown(&aspace->mm);
+	if (aspace->mmu)
+		aspace->mmu->funcs->destroy(aspace->mmu);
+	kfree(aspace);
+}
+
+
+void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
+{
+	if (aspace)
+		kref_put(&aspace->kref, msm_gem_address_space_destroy);
+}
+
 void
 msm_gem_unmap_vma(struct msm_gem_address_space *aspace,
 		struct msm_gem_vma *vma, struct sg_table *sgt)
@@ -34,6 +53,8 @@ msm_gem_unmap_vma(struct msm_gem_address_space *aspace,
 	drm_mm_remove_node(&vma->node);
 
 	vma->iova = 0;
+
+	msm_gem_address_space_put(aspace);
 }
 
 int
@@ -57,16 +78,10 @@ msm_gem_map_vma(struct msm_gem_address_space *aspace,
 				size, IOMMU_READ | IOMMU_WRITE);
 	}
 
-	return ret;
-}
+	/* Get a reference to the aspace to keep it around */
+	kref_get(&aspace->kref);
 
-void
-msm_gem_address_space_destroy(struct msm_gem_address_space *aspace)
-{
-	drm_mm_takedown(&aspace->mm);
-	if (aspace->mmu)
-		aspace->mmu->funcs->destroy(aspace->mmu);
-	kfree(aspace);
+	return ret;
 }
 
 struct msm_gem_address_space *
@@ -85,5 +100,7 @@ msm_gem_address_space_create(struct device *dev, struct iommu_domain *domain,
 	drm_mm_init(&aspace->mm, (domain->geometry.aperture_start >> PAGE_SHIFT),
 			(domain->geometry.aperture_end >> PAGE_SHIFT) - 1);
 
+	kref_init(&aspace->kref);
+
 	return aspace;
 }
diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
index af5b6ba4095b..97b9c38c6b3f 100644
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -93,18 +93,18 @@ static int enable_clk(struct msm_gpu *gpu)
 {
 	int i;
 
-	if (gpu->grp_clks[0] && gpu->fast_rate)
-		clk_set_rate(gpu->grp_clks[0], gpu->fast_rate);
+	if (gpu->core_clk && gpu->fast_rate)
+		clk_set_rate(gpu->core_clk, gpu->fast_rate);
 
 	/* Set the RBBM timer rate to 19.2Mhz */
-	if (gpu->grp_clks[2])
-		clk_set_rate(gpu->grp_clks[2], 19200000);
+	if (gpu->rbbmtimer_clk)
+		clk_set_rate(gpu->rbbmtimer_clk, 19200000);
 
-	for (i = ARRAY_SIZE(gpu->grp_clks) - 1; i >= 0; i--)
+	for (i = gpu->nr_clocks - 1; i >= 0; i--)
 		if (gpu->grp_clks[i])
 			clk_prepare(gpu->grp_clks[i]);
 
-	for (i = ARRAY_SIZE(gpu->grp_clks) - 1; i >= 0; i--)
+	for (i = gpu->nr_clocks - 1; i >= 0; i--)
 		if (gpu->grp_clks[i])
 			clk_enable(gpu->grp_clks[i]);
 
@@ -115,19 +115,24 @@ static int disable_clk(struct msm_gpu *gpu)
 {
 	int i;
 
-	for (i = ARRAY_SIZE(gpu->grp_clks) - 1; i >= 0; i--)
+	for (i = gpu->nr_clocks - 1; i >= 0; i--)
 		if (gpu->grp_clks[i])
 			clk_disable(gpu->grp_clks[i]);
 
-	for (i = ARRAY_SIZE(gpu->grp_clks) - 1; i >= 0; i--)
+	for (i = gpu->nr_clocks - 1; i >= 0; i--)
 		if (gpu->grp_clks[i])
 			clk_unprepare(gpu->grp_clks[i]);
 
-	if (gpu->grp_clks[0] && gpu->slow_rate)
-		clk_set_rate(gpu->grp_clks[0], gpu->slow_rate);
+	/*
+	 * Set the clock to a deliberately low rate. On older targets the clock
+	 * speed had to be non zero to avoid problems. On newer targets this
+	 * will be rounded down to zero anyway so it all works out.
+	 */
+	if (gpu->core_clk)
+		clk_set_rate(gpu->core_clk, 27000000);
 
-	if (gpu->grp_clks[2])
-		clk_set_rate(gpu->grp_clks[2], 0);
+	if (gpu->rbbmtimer_clk)
+		clk_set_rate(gpu->rbbmtimer_clk, 0);
 
 	return 0;
 }
@@ -152,18 +157,9 @@ static int disable_axi(struct msm_gpu *gpu)
 
 int msm_gpu_pm_resume(struct msm_gpu *gpu)
 {
-	struct drm_device *dev = gpu->dev;
 	int ret;
 
-	DBG("%s: active_cnt=%d", gpu->name, gpu->active_cnt);
-
-	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
-
-	if (gpu->active_cnt++ > 0)
-		return 0;
-
-	if (WARN_ON(gpu->active_cnt <= 0))
-		return -EINVAL;
+	DBG("%s", gpu->name);
 
 	ret = enable_pwrrail(gpu);
 	if (ret)
@@ -177,23 +173,16 @@ int msm_gpu_pm_resume(struct msm_gpu *gpu)
 	if (ret)
 		return ret;
 
+	gpu->needs_hw_init = true;
+
 	return 0;
 }
 
 int msm_gpu_pm_suspend(struct msm_gpu *gpu)
 {
-	struct drm_device *dev = gpu->dev;
 	int ret;
 
-	DBG("%s: active_cnt=%d", gpu->name, gpu->active_cnt);
-
-	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
-
-	if (--gpu->active_cnt > 0)
-		return 0;
-
-	if (WARN_ON(gpu->active_cnt < 0))
-		return -EINVAL;
+	DBG("%s", gpu->name);
 
 	ret = disable_axi(gpu);
 	if (ret)
@@ -210,53 +199,20 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
 	return 0;
 }
 
-/*
- * Inactivity detection (for suspend):
- */
-
-static void inactive_worker(struct work_struct *work)
-{
-	struct msm_gpu *gpu = container_of(work, struct msm_gpu, inactive_work);
-	struct drm_device *dev = gpu->dev;
-
-	if (gpu->inactive)
-		return;
-
-	DBG("%s: inactive!\n", gpu->name);
-	mutex_lock(&dev->struct_mutex);
-	if (!(msm_gpu_active(gpu) || gpu->inactive)) {
-		disable_axi(gpu);
-		disable_clk(gpu);
-		gpu->inactive = true;
-	}
-	mutex_unlock(&dev->struct_mutex);
-}
-
-static void inactive_handler(unsigned long data)
+int msm_gpu_hw_init(struct msm_gpu *gpu)
 {
-	struct msm_gpu *gpu = (struct msm_gpu *)data;
-	struct msm_drm_private *priv = gpu->dev->dev_private;
+	int ret;
 
-	queue_work(priv->wq, &gpu->inactive_work);
-}
+	if (!gpu->needs_hw_init)
+		return 0;
 
-/* cancel inactive timer and make sure we are awake: */
-static void inactive_cancel(struct msm_gpu *gpu)
-{
-	DBG("%s", gpu->name);
-	del_timer(&gpu->inactive_timer);
-	if (gpu->inactive) {
-		enable_clk(gpu);
-		enable_axi(gpu);
-		gpu->inactive = false;
-	}
-}
+	disable_irq(gpu->irq);
+	ret = gpu->funcs->hw_init(gpu);
+	if (!ret)
+		gpu->needs_hw_init = false;
+	enable_irq(gpu->irq);
 
-static void inactive_start(struct msm_gpu *gpu)
-{
-	DBG("%s", gpu->name);
-	mod_timer(&gpu->inactive_timer,
-			round_jiffies_up(jiffies + DRM_MSM_INACTIVE_JIFFIES));
+	return ret;
 }
 
 /*
@@ -296,8 +252,9 @@ static void recover_worker(struct work_struct *work)
 		/* retire completed submits, plus the one that hung: */
 		retire_submits(gpu);
 
-		inactive_cancel(gpu);
+		pm_runtime_get_sync(&gpu->pdev->dev);
 		gpu->funcs->recover(gpu);
+		pm_runtime_put_sync(&gpu->pdev->dev);
 
 		/* replay the remaining submits after the one that hung: */
 		list_for_each_entry(submit, &gpu->submit_list, node) {
@@ -400,6 +357,8 @@ void msm_gpu_perfcntr_start(struct msm_gpu *gpu)
 {
 	unsigned long flags;
 
+	pm_runtime_get_sync(&gpu->pdev->dev);
+
 	spin_lock_irqsave(&gpu->perf_lock, flags);
 	/* we could dynamically enable/disable perfcntr registers too.. */
 	gpu->last_sample.active = msm_gpu_active(gpu);
@@ -413,6 +372,7 @@ void msm_gpu_perfcntr_start(struct msm_gpu *gpu)
 void msm_gpu_perfcntr_stop(struct msm_gpu *gpu)
 {
 	gpu->perfcntr_active = false;
+	pm_runtime_put_sync(&gpu->pdev->dev);
 }
 
 /* returns -errno or # of cntrs sampled */
@@ -458,6 +418,8 @@ static void retire_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)
 		drm_gem_object_unreference(&msm_obj->base);
 	}
 
+	pm_runtime_mark_last_busy(&gpu->pdev->dev);
+	pm_runtime_put_autosuspend(&gpu->pdev->dev);
 	msm_gem_submit_free(submit);
 }
 
@@ -492,9 +454,6 @@ static void retire_worker(struct work_struct *work)
 	mutex_lock(&dev->struct_mutex);
 	retire_submits(gpu);
 	mutex_unlock(&dev->struct_mutex);
-
-	if (!msm_gpu_active(gpu))
-		inactive_start(gpu);
 }
 
 /* call from irq handler to schedule work to retire bo's */
@@ -515,7 +474,9 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
 	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
 
-	inactive_cancel(gpu);
+	pm_runtime_get_sync(&gpu->pdev->dev);
+
+	msm_gpu_hw_init(gpu);
 
 	list_add_tail(&submit->node, &gpu->submit_list);
 
@@ -559,16 +520,52 @@ static irqreturn_t irq_handler(int irq, void *data)
 	return gpu->funcs->irq(gpu);
 }
 
-static const char *clk_names[] = {
-	"core", "iface", "rbbmtimer", "mem", "mem_iface", "alt_mem_iface",
-};
+static struct clk *get_clock(struct device *dev, const char *name)
+{
+	struct clk *clk = devm_clk_get(dev, name);
+
+	return IS_ERR(clk) ? NULL : clk;
+}
+
+static int get_clocks(struct platform_device *pdev, struct msm_gpu *gpu)
+{
+	struct device *dev = &pdev->dev;
+	struct property *prop;
+	const char *name;
+	int i = 0;
+
+	gpu->nr_clocks = of_property_count_strings(dev->of_node, "clock-names");
+	if (gpu->nr_clocks < 1) {
+		gpu->nr_clocks = 0;
+		return 0;
+	}
+
+	gpu->grp_clks = devm_kcalloc(dev, sizeof(struct clk *), gpu->nr_clocks,
+		GFP_KERNEL);
+	if (!gpu->grp_clks)
+		return -ENOMEM;
+
+	of_property_for_each_string(dev->of_node, "clock-names", prop, name) {
+		gpu->grp_clks[i] = get_clock(dev, name);
+
+		/* Remember the key clocks that we need to control later */
+		if (!strcmp(name, "core"))
+			gpu->core_clk = gpu->grp_clks[i];
+		else if (!strcmp(name, "rbbmtimer"))
+			gpu->rbbmtimer_clk = gpu->grp_clks[i];
+
+		++i;
+	}
+
+	return 0;
+}
 
 int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		struct msm_gpu *gpu, const struct msm_gpu_funcs *funcs,
 		const char *name, const char *ioname, const char *irqname, int ringsz)
 {
 	struct iommu_domain *iommu;
-	int i, ret;
+	int ret;
 
 	if (WARN_ON(gpu->num_perfcntrs > ARRAY_SIZE(gpu->last_cntrs)))
 		gpu->num_perfcntrs = ARRAY_SIZE(gpu->last_cntrs);
@@ -576,7 +573,6 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	gpu->dev = drm;
 	gpu->funcs = funcs;
 	gpu->name = name;
-	gpu->inactive = true;
 	gpu->fctx = msm_fence_context_alloc(drm, name);
 	if (IS_ERR(gpu->fctx)) {
 		ret = PTR_ERR(gpu->fctx);
@@ -586,19 +582,15 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 	INIT_LIST_HEAD(&gpu->active_list);
 	INIT_WORK(&gpu->retire_work, retire_worker);
-	INIT_WORK(&gpu->inactive_work, inactive_worker);
 	INIT_WORK(&gpu->recover_work, recover_worker);
 
 	INIT_LIST_HEAD(&gpu->submit_list);
 
-	setup_timer(&gpu->inactive_timer, inactive_handler,
-			(unsigned long)gpu);
 	setup_timer(&gpu->hangcheck_timer, hangcheck_handler,
 			(unsigned long)gpu);
 
 	spin_lock_init(&gpu->perf_lock);
 
-	BUG_ON(ARRAY_SIZE(clk_names) != ARRAY_SIZE(gpu->grp_clks));
 
 	/* Map registers: */
 	gpu->mmio = msm_ioremap(pdev, ioname, name);
@@ -622,13 +614,9 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		goto fail;
 	}
 
-	/* Acquire clocks: */
-	for (i = 0; i < ARRAY_SIZE(clk_names); i++) {
-		gpu->grp_clks[i] = msm_clk_get(pdev, clk_names[i]);
-		DBG("grp_clks[%s]: %p", clk_names[i], gpu->grp_clks[i]);
-		if (IS_ERR(gpu->grp_clks[i]))
-			gpu->grp_clks[i] = NULL;
-	}
+	ret = get_clocks(pdev, gpu);
+	if (ret)
+		goto fail;
 
 	gpu->ebi1_clk = msm_clk_get(pdev, "bus");
 	DBG("ebi1_clk: %p", gpu->ebi1_clk);
@@ -684,6 +672,9 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		goto fail;
 	}
 
+	gpu->pdev = pdev;
+	platform_set_drvdata(pdev, gpu);
+
 	bs_init(gpu);
 
 	return 0;
diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
index c4c39d3272c7..aa3241000455 100644
--- a/drivers/gpu/drm/msm/msm_gpu.h
+++ b/drivers/gpu/drm/msm/msm_gpu.h
@@ -64,6 +64,7 @@ struct msm_gpu_funcs {
 struct msm_gpu {
 	const char *name;
 	struct drm_device *dev;
+	struct platform_device *pdev;
 	const struct msm_gpu_funcs *funcs;
 
 	/* performance counters (hw & sw): */
@@ -88,9 +89,8 @@ struct msm_gpu {
 	/* fencing: */
 	struct msm_fence_context *fctx;
 
-	/* is gpu powered/active? */
-	int active_cnt;
-	bool inactive;
+	/* does gpu need hw_init? */
+	bool needs_hw_init;
 
 	/* worker for handling active-list retiring: */
 	struct work_struct retire_work;
@@ -103,8 +103,10 @@ struct msm_gpu {
 
 	/* Power Control: */
 	struct regulator *gpu_reg, *gpu_cx;
-	struct clk *ebi1_clk, *grp_clks[6];
-	uint32_t fast_rate, slow_rate, bus_freq;
+	struct clk **grp_clks;
+	int nr_clocks;
+	struct clk *ebi1_clk, *core_clk, *rbbmtimer_clk;
+	uint32_t fast_rate, bus_freq;
 
 #ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
 	struct msm_bus_scale_pdata *bus_scale_table;
@@ -114,9 +116,7 @@ struct msm_gpu {
 	/* Hang and Inactivity Detection:
 	 */
 #define DRM_MSM_INACTIVE_PERIOD   66 /* in ms (roughly four frames) */
-#define DRM_MSM_INACTIVE_JIFFIES  msecs_to_jiffies(DRM_MSM_INACTIVE_PERIOD)
-	struct timer_list inactive_timer;
-	struct work_struct inactive_work;
+
 #define DRM_MSM_HANGCHECK_PERIOD 500 /* in ms */
 #define DRM_MSM_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_MSM_HANGCHECK_PERIOD)
 	struct timer_list hangcheck_timer;
@@ -196,6 +196,8 @@ static inline void gpu_write64(struct msm_gpu *gpu, u32 lo, u32 hi, u64 val)
 int msm_gpu_pm_suspend(struct msm_gpu *gpu);
 int msm_gpu_pm_resume(struct msm_gpu *gpu);
 
+int msm_gpu_hw_init(struct msm_gpu *gpu);
+
 void msm_gpu_perfcntr_start(struct msm_gpu *gpu);
 void msm_gpu_perfcntr_stop(struct msm_gpu *gpu);
 int msm_gpu_perfcntr_sample(struct msm_gpu *gpu, uint32_t *activetime,
diff --git a/drivers/gpu/drm/msm/msm_iommu.c b/drivers/gpu/drm/msm/msm_iommu.c
index 7f5779daf5c8..b23d33622f37 100644
--- a/drivers/gpu/drm/msm/msm_iommu.c
+++ b/drivers/gpu/drm/msm/msm_iommu.c
@@ -38,78 +38,47 @@ static int msm_iommu_attach(struct msm_mmu *mmu, const char * const *names,
 			    int cnt)
 {
 	struct msm_iommu *iommu = to_msm_iommu(mmu);
-	return iommu_attach_device(iommu->domain, mmu->dev);
+	int ret;
+
+	pm_runtime_get_sync(mmu->dev);
+	ret = iommu_attach_device(iommu->domain, mmu->dev);
+	pm_runtime_put_sync(mmu->dev);
+
+	return ret;
 }
 
 static void msm_iommu_detach(struct msm_mmu *mmu, const char * const *names,
 			     int cnt)
 {
 	struct msm_iommu *iommu = to_msm_iommu(mmu);
+
+	pm_runtime_get_sync(mmu->dev);
 	iommu_detach_device(iommu->domain, mmu->dev);
+	pm_runtime_put_sync(mmu->dev);
 }
 
 static int msm_iommu_map(struct msm_mmu *mmu, uint64_t iova,
 		struct sg_table *sgt, unsigned len, int prot)
 {
 	struct msm_iommu *iommu = to_msm_iommu(mmu);
-	struct iommu_domain *domain = iommu->domain;
-	struct scatterlist *sg;
-	unsigned long da = iova;
-	unsigned int i, j;
-	int ret;
-
-	if (!domain || !sgt)
-		return -EINVAL;
-
-	for_each_sg(sgt->sgl, sg, sgt->nents, i) {
-		dma_addr_t pa = sg_phys(sg) - sg->offset;
-		size_t bytes = sg->length + sg->offset;
-
-		VERB("map[%d]: %08lx %08lx(%zx)", i, da, (unsigned long)pa, bytes);
+	size_t ret;
 
-		ret = iommu_map(domain, da, pa, bytes, prot);
-		if (ret)
-			goto fail;
+//	pm_runtime_get_sync(mmu->dev);
+	ret = iommu_map_sg(iommu->domain, iova, sgt->sgl, sgt->nents, prot);
+//	pm_runtime_put_sync(mmu->dev);
+	WARN_ON(ret < 0);
 
-		da += bytes;
-	}
-
-	return 0;
-
-fail:
-	da = iova;
-
-	for_each_sg(sgt->sgl, sg, i, j) {
-		size_t bytes = sg->length + sg->offset;
-		iommu_unmap(domain, da, bytes);
-		da += bytes;
-	}
-	return ret;
+	return (ret == len) ? 0 : -EINVAL;
 }
 
 static int msm_iommu_unmap(struct msm_mmu *mmu, uint64_t iova,
 		struct sg_table *sgt, unsigned len)
 {
 	struct msm_iommu *iommu = to_msm_iommu(mmu);
-	struct iommu_domain *domain = iommu->domain;
-	struct scatterlist *sg;
-	unsigned long da = iova;
-	int i;
-
-	for_each_sg(sgt->sgl, sg, sgt->nents, i) {
-		size_t bytes = sg->length + sg->offset;
-		size_t unmapped;
-
-		unmapped = iommu_unmap(domain, da, bytes);
-		if (unmapped < bytes)
-			return unmapped;
-
-		VERB("unmap[%d]: %08lx(%zx)", i, da, bytes);
-
-		BUG_ON(!PAGE_ALIGNED(bytes));
 
-		da += bytes;
-	}
+	pm_runtime_get_sync(mmu->dev);
+	iommu_unmap(iommu->domain, iova, len);
+	pm_runtime_put_sync(mmu->dev);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/msm/msm_rd.c b/drivers/gpu/drm/msm/msm_rd.c
index 3df7322fd74e..0e81faab2c50 100644
--- a/drivers/gpu/drm/msm/msm_rd.c
+++ b/drivers/gpu/drm/msm/msm_rd.c
@@ -322,7 +322,7 @@ void msm_rd_dump_submit(struct msm_gem_submit *submit)
 	}
 
 	for (i = 0; i < submit->nr_cmds; i++) {
-		uint32_t iova = submit->cmd[i].iova;
+		uint64_t iova = submit->cmd[i].iova;
 		uint32_t szd  = submit->cmd[i].size; /* in dwords */
 
 		/* snapshot cmdstream bo's (if we haven't already): */
@@ -341,7 +341,7 @@ void msm_rd_dump_submit(struct msm_gem_submit *submit)
 		case MSM_SUBMIT_CMD_CTX_RESTORE_BUF:
 		case MSM_SUBMIT_CMD_BUF:
 			rd_write_section(rd, RD_CMDSTREAM_ADDR,
-					(uint32_t[2]){ iova, szd }, 8);
+				(uint32_t[3]){ iova, szd, iova >> 32 }, 12);
 			break;
 		}
 	}
diff --git a/include/uapi/drm/msm_drm.h b/include/uapi/drm/msm_drm.h
index 4d5d6a2bc59e..a4a189a240d7 100644
--- a/include/uapi/drm/msm_drm.h
+++ b/include/uapi/drm/msm_drm.h
@@ -72,6 +72,7 @@ struct drm_msm_timespec {
 #define MSM_PARAM_CHIP_ID    0x03
 #define MSM_PARAM_MAX_FREQ   0x04
 #define MSM_PARAM_TIMESTAMP  0x05
+#define MSM_PARAM_GMEM_BASE  0x06
 
 struct drm_msm_param {
 	__u32 pipe;           /* in, MSM_PIPE_x */