[PATCH v3 1/4] drm/msm/mdp4: add lcdc-align-lsb flag to control lane alignment

[PATCH v3 1/4] drm/msm/mdp4: add lcdc-align-lsb flag to control lane alignment

[Jonathan Marek]

This allows controlling which of the 8 lanes are used for 6 bit color.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
---
v3: removed empty line and added documentation

 .../devicetree/bindings/display/msm/mdp4.txt  |  2 ++
 .../gpu/drm/msm/disp/mdp4/mdp4_lcdc_encoder.c | 21 ++++++++++++-------
 2 files changed, 15 insertions(+), 8 deletions(-)

diff --git a/Documentation/devicetree/bindings/display/msm/mdp4.txt b/Documentation/devicetree/bindings/display/msm/mdp4.txt
index 3c341a15c..b07eeb38f 100644
--- a/Documentation/devicetree/bindings/display/msm/mdp4.txt
+++ b/Documentation/devicetree/bindings/display/msm/mdp4.txt
@@ -38,6 +38,8 @@ Required properties:
 Optional properties:
 - clock-names: the following clocks are optional:
   * "lut_clk"
+- qcom,lcdc-align-lsb: Boolean value indicating that LSB alignment should be
+  used for LCDC. This is only valid for 18bpp panels.
 
 Example:
 
diff --git a/drivers/gpu/drm/msm/disp/mdp4/mdp4_lcdc_encoder.c b/drivers/gpu/drm/msm/disp/mdp4/mdp4_lcdc_encoder.c
index 9e08c2efa..c9e34501a 100644
--- a/drivers/gpu/drm/msm/disp/mdp4/mdp4_lcdc_encoder.c
+++ b/drivers/gpu/drm/msm/disp/mdp4/mdp4_lcdc_encoder.c
@@ -377,20 +377,25 @@ static void mdp4_lcdc_encoder_enable(struct drm_encoder *encoder)
 	unsigned long pc = mdp4_lcdc_encoder->pixclock;
 	struct mdp4_kms *mdp4_kms = get_kms(encoder);
 	struct drm_panel *panel;
+	uint32_t config;
 	int i, ret;
 
 	if (WARN_ON(mdp4_lcdc_encoder->enabled))
 		return;
 
 	/* TODO: hard-coded for 18bpp: */
-	mdp4_crtc_set_config(encoder->crtc,
-			MDP4_DMA_CONFIG_R_BPC(BPC6) |
-			MDP4_DMA_CONFIG_G_BPC(BPC6) |
-			MDP4_DMA_CONFIG_B_BPC(BPC6) |
-			MDP4_DMA_CONFIG_PACK_ALIGN_MSB |
-			MDP4_DMA_CONFIG_PACK(0x21) |
-			MDP4_DMA_CONFIG_DEFLKR_EN |
-			MDP4_DMA_CONFIG_DITHER_EN);
+	config =
+		MDP4_DMA_CONFIG_R_BPC(BPC6) |
+		MDP4_DMA_CONFIG_G_BPC(BPC6) |
+		MDP4_DMA_CONFIG_B_BPC(BPC6) |
+		MDP4_DMA_CONFIG_PACK(0x21) |
+		MDP4_DMA_CONFIG_DEFLKR_EN |
+		MDP4_DMA_CONFIG_DITHER_EN;
+
+	if (!of_property_read_bool(dev->dev->of_node, "qcom,lcdc-align-lsb"))
+		config |= MDP4_DMA_CONFIG_PACK_ALIGN_MSB;
+
+	mdp4_crtc_set_config(encoder->crtc, config);
 	mdp4_crtc_set_intf(encoder->crtc, INTF_LCDC_DTV, 0);
 
 	bs_set(mdp4_lcdc_encoder, 1);
-- 
2.17.1

[PATCH v3 2/4] drm/msm: add headless gpu device for imx5

[Jonathan Marek]

This patch allows using drm/msm without qcom display hardware. It adds a
amd,imageon compatible, which is used instead of qcom,adreno, but does
not require a top level msm node.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
---
v3: reworked to work with only a amd,imageon node

 drivers/gpu/drm/msm/Kconfig                |  4 +--
 drivers/gpu/drm/msm/adreno/adreno_device.c | 35 ++++++++++++++++++++--
 drivers/gpu/drm/msm/msm_debugfs.c          |  2 +-
 drivers/gpu/drm/msm/msm_drv.c              | 21 +++++++------
 4 files changed, 46 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/msm/Kconfig b/drivers/gpu/drm/msm/Kconfig
index 843a9d40c..cf549f1ed 100644
--- a/drivers/gpu/drm/msm/Kconfig
+++ b/drivers/gpu/drm/msm/Kconfig
@@ -2,7 +2,7 @@
 config DRM_MSM
 	tristate "MSM DRM"
 	depends on DRM
-	depends on ARCH_QCOM || (ARM && COMPILE_TEST)
+	depends on ARCH_QCOM || SOC_IMX5 || (ARM && COMPILE_TEST)
 	depends on OF && COMMON_CLK
 	depends on MMU
 	select QCOM_MDT_LOADER if ARCH_QCOM
@@ -11,7 +11,7 @@ config DRM_MSM
 	select DRM_PANEL
 	select SHMEM
 	select TMPFS
-	select QCOM_SCM
+	select QCOM_SCM if ARCH_QCOM
 	select WANT_DEV_COREDUMP
 	select SND_SOC_HDMI_CODEC if SND_SOC
 	select SYNC_FILE
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index adc442f73..99e363c3d 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -271,7 +271,8 @@ static int find_chipid(struct device *dev, struct adreno_rev *rev)
 	if (ret == 0) {
 		unsigned int r, patch;
 
-		if (sscanf(compat, "qcom,adreno-%u.%u", &r, &patch) == 2) {
+		if (sscanf(compat, "qcom,adreno-%u.%u", &r, &patch) == 2 ||
+		    sscanf(compat, "amd,imageon-%u.%u", &r, &patch) == 2) {
 			rev->core = r / 100;
 			r %= 100;
 			rev->major = r / 10;
@@ -356,9 +357,37 @@ static const struct component_ops a3xx_ops = {
 		.unbind = adreno_unbind,
 };
 
+static void adreno_device_register_headless(void)
+{
+	/* on imx5, we don't have a top-level mdp/dpu node
+	 * this creates a dummy node for the driver for that case
+	 */
+	struct platform_device_info dummy_info = {
+		.parent = NULL,
+		.name = "msm",
+		.id = -1,
+		.res = NULL,
+		.num_res = 0,
+		.data = NULL,
+		.size_data = 0,
+		.dma_mask = ~0,
+	};
+	platform_device_register_full(&dummy_info);
+}
+
 static int adreno_probe(struct platform_device *pdev)
 {
-	return component_add(&pdev->dev, &a3xx_ops);
+
+	int ret;
+
+	ret = component_add(&pdev->dev, &a3xx_ops);
+	if (ret)
+		return ret;
+
+	if (of_device_is_compatible(pdev->dev.of_node, "amd,imageon"))
+		adreno_device_register_headless();
+
+	return 0;
 }
 
 static int adreno_remove(struct platform_device *pdev)
@@ -370,6 +399,8 @@ static int adreno_remove(struct platform_device *pdev)
 static const struct of_device_id dt_match[] = {
 	{ .compatible = "qcom,adreno" },
 	{ .compatible = "qcom,adreno-3xx" },
+	/* for compatibility with imx5 gpu: */
+	{ .compatible = "amd,imageon" },
 	/* for backwards compat w/ downstream kgsl DT files: */
 	{ .compatible = "qcom,kgsl-3d0" },
 	{}
diff --git a/drivers/gpu/drm/msm/msm_debugfs.c b/drivers/gpu/drm/msm/msm_debugfs.c
index 9a7cf9fe2..fb423d309 100644
--- a/drivers/gpu/drm/msm/msm_debugfs.c
+++ b/drivers/gpu/drm/msm/msm_debugfs.c
@@ -242,7 +242,7 @@ int msm_debugfs_init(struct drm_minor *minor)
 	debugfs_create_file("gpu", S_IRUSR, minor->debugfs_root,
 		dev, &msm_gpu_fops);
 
-	if (priv->kms->funcs->debugfs_init) {
+	if (priv->kms && priv->kms->funcs->debugfs_init) {
 		ret = priv->kms->funcs->debugfs_init(priv->kms, minor);
 		if (ret)
 			return ret;
diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
index 81bfac744..7842518a9 100644
--- a/drivers/gpu/drm/msm/msm_drv.c
+++ b/drivers/gpu/drm/msm/msm_drv.c
@@ -510,17 +510,13 @@ static int msm_drm_init(struct device *dev, struct drm_driver *drv)
 		priv->kms = kms;
 		break;
 	default:
-		kms = ERR_PTR(-ENODEV);
+		/* valid only for the dummy headless case, where of_node=NULL */
+		WARN_ON(dev->of_node);
+		kms = NULL;
 		break;
 	}
 
 	if (IS_ERR(kms)) {
-		/*
-		 * NOTE: once we have GPU support, having no kms should not
-		 * be considered fatal.. ideally we would still support gpu
-		 * and (for example) use dmabuf/prime to share buffers with
-		 * imx drm driver on iMX5
-		 */
 		DRM_DEV_ERROR(dev, "failed to load kms\n");
 		ret = PTR_ERR(kms);
 		priv->kms = NULL;
@@ -620,7 +616,7 @@ static int msm_drm_init(struct device *dev, struct drm_driver *drv)
 	drm_mode_config_reset(ddev);
 
 #ifdef CONFIG_DRM_FBDEV_EMULATION
-	if (fbdev)
+	if (kms && fbdev)
 		priv->fbdev = msm_fbdev_init(ddev);
 #endif
 
@@ -1302,6 +1298,7 @@ static int add_display_components(struct device *dev,
 static const struct of_device_id msm_gpu_match[] = {
 	{ .compatible = "qcom,adreno" },
 	{ .compatible = "qcom,adreno-3xx" },
+	{ .compatible = "amd,imageon" },
 	{ .compatible = "qcom,kgsl-3d0" },
 	{ },
 };
@@ -1346,9 +1343,11 @@ static int msm_pdev_probe(struct platform_device *pdev)
 	struct component_match *match = NULL;
 	int ret;
 
-	ret = add_display_components(&pdev->dev, &match);
-	if (ret)
-		return ret;
+	if (get_mdp_ver(pdev)) {
+		ret = add_display_components(&pdev->dev, &match);
+		if (ret)
+			return ret;
+	}
 
 	ret = add_gpu_components(&pdev->dev, &match);
 	if (ret)
-- 
2.17.1

[PATCH v3 3/4] drm/msm: implement a2xx mmu

[Jonathan Marek]

A2XX has its own very simple MMU.

Added a msm_use_mmu() function because we can't rely on iommu_present to
decide to use MMU or not.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
---
v3: rebased on msm-next-staging and moved is_a2xx initialization earlier

 drivers/gpu/drm/msm/Makefile               |   3 +-
 drivers/gpu/drm/msm/adreno/a2xx_gpu.c      |  50 ++++++++-
 drivers/gpu/drm/msm/adreno/adreno_device.c |   3 +
 drivers/gpu/drm/msm/adreno/adreno_gpu.c    |   3 +
 drivers/gpu/drm/msm/msm_drv.c              |  11 +-
 drivers/gpu/drm/msm/msm_drv.h              |   8 ++
 drivers/gpu/drm/msm/msm_gem.c              |   4 +-
 drivers/gpu/drm/msm/msm_gem_vma.c          |  23 ++++
 drivers/gpu/drm/msm/msm_gpu.c              |  31 ++++--
 drivers/gpu/drm/msm/msm_gpummu.c           | 122 +++++++++++++++++++++
 drivers/gpu/drm/msm/msm_mmu.h              |   3 +
 11 files changed, 241 insertions(+), 20 deletions(-)
 create mode 100644 drivers/gpu/drm/msm/msm_gpummu.c

diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile
index 61e76f87a..1b26c4105 100644
--- a/drivers/gpu/drm/msm/Makefile
+++ b/drivers/gpu/drm/msm/Makefile
@@ -93,7 +93,8 @@ msm-y := \
 	msm_rd.o \
 	msm_ringbuffer.o \
 	msm_submitqueue.o \
-	msm_gpu_tracepoints.o
+	msm_gpu_tracepoints.o \
+	msm_gpummu.o
 
 msm-$(CONFIG_DEBUG_FS) += adreno/a5xx_debugfs.o \
 			  disp/dpu1/dpu_dbg.o
diff --git a/drivers/gpu/drm/msm/adreno/a2xx_gpu.c b/drivers/gpu/drm/msm/adreno/a2xx_gpu.c
index 5eddcf14e..1f83bc18d 100644
--- a/drivers/gpu/drm/msm/adreno/a2xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a2xx_gpu.c
@@ -2,6 +2,8 @@
 /* Copyright (c) 2018 The Linux Foundation. All rights reserved. */
 
 #include "a2xx_gpu.h"
+#include "msm_gem.h"
+#include "msm_mmu.h"
 
 extern bool hang_debug;
 
@@ -58,9 +60,12 @@ static bool a2xx_me_init(struct msm_gpu *gpu)
 static int a2xx_hw_init(struct msm_gpu *gpu)
 {
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+	dma_addr_t pt_base, tran_error;
 	uint32_t *ptr, len;
 	int i, ret;
 
+	msm_gpummu_params(gpu->aspace->mmu, &pt_base, &tran_error);
+
 	DBG("%s", gpu->name);
 
 	/* halt ME to avoid ucode upload issues on a20x */
@@ -80,9 +85,34 @@ static int a2xx_hw_init(struct msm_gpu *gpu)
 	/* note: kgsl uses 0x0000ffff for a20x */
 	gpu_write(gpu, REG_A2XX_RBBM_CNTL, 0x00004442);
 
-	gpu_write(gpu, REG_A2XX_MH_MMU_CONFIG, 0);
-	gpu_write(gpu, REG_A2XX_MH_MMU_MPU_BASE, 0);
+	/* MPU: physical range */
+	gpu_write(gpu, REG_A2XX_MH_MMU_MPU_BASE, 0x00000000);
 	gpu_write(gpu, REG_A2XX_MH_MMU_MPU_END, 0xfffff000);
+
+	gpu_write(gpu, REG_A2XX_MH_MMU_CONFIG, A2XX_MH_MMU_CONFIG_MMU_ENABLE |
+		A2XX_MH_MMU_CONFIG_RB_W_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_CP_W_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_CP_R0_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_CP_R1_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_CP_R2_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_CP_R3_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_CP_R4_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_VGT_R0_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_VGT_R1_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_TC_R_CLNT_BEHAVIOR(BEH_TRAN_RNG) |
+		A2XX_MH_MMU_CONFIG_PA_W_CLNT_BEHAVIOR(BEH_TRAN_RNG));
+
+	/* same as parameters in adreno_gpu */
+	gpu_write(gpu, REG_A2XX_MH_MMU_VA_RANGE, SZ_16M |
+		A2XX_MH_MMU_VA_RANGE_NUM_64KB_REGIONS(0xfff));
+
+	gpu_write(gpu, REG_A2XX_MH_MMU_PT_BASE, pt_base);
+	gpu_write(gpu, REG_A2XX_MH_MMU_TRAN_ERROR, tran_error);
+
+	gpu_write(gpu, REG_A2XX_MH_MMU_INVALIDATE,
+		A2XX_MH_MMU_INVALIDATE_INVALIDATE_ALL |
+		A2XX_MH_MMU_INVALIDATE_INVALIDATE_TC);
+
 	gpu_write(gpu, REG_A2XX_MH_ARBITER_CONFIG,
 		A2XX_MH_ARBITER_CONFIG_SAME_PAGE_LIMIT(16) |
 		A2XX_MH_ARBITER_CONFIG_L1_ARB_ENABLE |
@@ -109,9 +139,21 @@ static int a2xx_hw_init(struct msm_gpu *gpu)
 	/* note: gsl doesn't set this */
 	gpu_write(gpu, REG_A2XX_RBBM_DEBUG, 0x00080000);
 
-	gpu_write(gpu, REG_A2XX_RBBM_INT_CNTL, 0);
-	gpu_write(gpu, REG_AXXX_CP_INT_CNTL, 0x80000000); /* RB INT */
+	gpu_write(gpu, REG_A2XX_RBBM_INT_CNTL,
+		A2XX_RBBM_INT_CNTL_RDERR_INT_MASK);
+	gpu_write(gpu, REG_AXXX_CP_INT_CNTL,
+		AXXX_CP_INT_CNTL_T0_PACKET_IN_IB_MASK |
+		AXXX_CP_INT_CNTL_OPCODE_ERROR_MASK |
+		AXXX_CP_INT_CNTL_PROTECTED_MODE_ERROR_MASK |
+		AXXX_CP_INT_CNTL_RESERVED_BIT_ERROR_MASK |
+		AXXX_CP_INT_CNTL_IB_ERROR_MASK |
+		AXXX_CP_INT_CNTL_IB1_INT_MASK |
+		AXXX_CP_INT_CNTL_RB_INT_MASK);
 	gpu_write(gpu, REG_A2XX_SQ_INT_CNTL, 0);
+	gpu_write(gpu, REG_A2XX_MH_INTERRUPT_MASK,
+		A2XX_MH_INTERRUPT_MASK_AXI_READ_ERROR |
+		A2XX_MH_INTERRUPT_MASK_AXI_WRITE_ERROR |
+		A2XX_MH_INTERRUPT_MASK_MMU_PAGE_FAULT);
 
 	for (i = 3; i <= 5; i++)
 		if ((SZ_16K << i) == adreno_gpu->gmem)
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index 99e363c3d..714ed6505 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -308,6 +308,7 @@ static int adreno_bind(struct device *dev, struct device *master, void *data)
 	static struct adreno_platform_config config = {};
 	const struct adreno_info *info;
 	struct drm_device *drm = dev_get_drvdata(master);
+	struct msm_drm_private *priv = drm->dev_private;
 	struct msm_gpu *gpu;
 	int ret;
 
@@ -330,6 +331,8 @@ static int adreno_bind(struct device *dev, struct device *master, void *data)
 	DBG("Found GPU: %u.%u.%u.%u", config.rev.core, config.rev.major,
 		config.rev.minor, config.rev.patchid);
 
+	priv->is_a2xx = config.rev.core == 2;
+
 	gpu = info->init(drm);
 	if (IS_ERR(gpu)) {
 		dev_warn(drm->dev, "failed to load adreno gpu\n");
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index 9ab67dd1b..2e4372ef1 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -769,6 +769,9 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 	adreno_gpu_config.va_start = SZ_16M;
 	adreno_gpu_config.va_end = 0xffffffff;
+	/* maximum range of a2xx mmu */
+	if (adreno_is_a2xx(adreno_gpu))
+		adreno_gpu_config.va_end = SZ_16M + 0xfff * SZ_64K;
 
 	adreno_gpu_config.nr_rings = nr_rings;
 
diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
index 7842518a9..271f1df61 100644
--- a/drivers/gpu/drm/msm/msm_drv.c
+++ b/drivers/gpu/drm/msm/msm_drv.c
@@ -26,6 +26,7 @@
 #include "msm_gem.h"
 #include "msm_gpu.h"
 #include "msm_kms.h"
+#include "adreno/adreno_gpu.h"
 
 
 /*
@@ -360,6 +361,14 @@ static int get_mdp_ver(struct platform_device *pdev)
 
 #include <linux/of_address.h>
 
+bool msm_use_mmu(struct drm_device *dev)
+{
+	struct msm_drm_private *priv = dev->dev_private;
+
+	/* a2xx comes with its own MMU */
+	return priv->is_a2xx || iommu_present(&platform_bus_type);
+}
+
 static int msm_init_vram(struct drm_device *dev)
 {
 	struct msm_drm_private *priv = dev->dev_private;
@@ -398,7 +407,7 @@ static int msm_init_vram(struct drm_device *dev)
 		 * Grab the entire CMA chunk carved out in early startup in
 		 * mach-msm:
 		 */
-	} else if (!iommu_present(&platform_bus_type)) {
+	} else if (!msm_use_mmu(dev)) {
 		DRM_INFO("using %s VRAM carveout\n", vram);
 		size = memparse(vram, NULL);
 	}
diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h
index 71a03ce21..9cd6a96c6 100644
--- a/drivers/gpu/drm/msm/msm_drv.h
+++ b/drivers/gpu/drm/msm/msm_drv.h
@@ -179,6 +179,8 @@ struct msm_drm_private {
 	/* when we have more than one 'msm_gpu' these need to be an array: */
 	struct msm_gpu *gpu;
 	struct msm_file_private *lastctx;
+	/* gpu is only set on open(), but we need this info earlier */
+	bool is_a2xx;
 
 	struct drm_fb_helper *fbdev;
 
@@ -258,9 +260,15 @@ struct msm_gem_address_space *
 msm_gem_address_space_create(struct device *dev, struct iommu_domain *domain,
 		const char *name);
 
+struct msm_gem_address_space *
+msm_gem_address_space_create_a2xx(struct device *dev, struct msm_gpu *gpu,
+		const char *name, uint64_t va_start, uint64_t va_end);
+
 int msm_register_mmu(struct drm_device *dev, struct msm_mmu *mmu);
 void msm_unregister_mmu(struct drm_device *dev, struct msm_mmu *mmu);
 
+bool msm_use_mmu(struct drm_device *dev);
+
 void msm_gem_submit_free(struct msm_gem_submit *submit);
 int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 		struct drm_file *file);
diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c
index 38b7f4e9e..51a95da69 100644
--- a/drivers/gpu/drm/msm/msm_gem.c
+++ b/drivers/gpu/drm/msm/msm_gem.c
@@ -975,7 +975,7 @@ static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,
 
 	size = PAGE_ALIGN(size);
 
-	if (!iommu_present(&platform_bus_type))
+	if (!msm_use_mmu(dev))
 		use_vram = true;
 	else if ((flags & (MSM_BO_STOLEN | MSM_BO_SCANOUT)) && priv->vram.size)
 		use_vram = true;
@@ -1052,7 +1052,7 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
 	int ret, npages;
 
 	/* if we don't have IOMMU, don't bother pretending we can import: */
-	if (!iommu_present(&platform_bus_type)) {
+	if (!msm_use_mmu(dev)) {
 		DRM_DEV_ERROR(dev->dev, "cannot import without IOMMU\n");
 		return ERR_PTR(-EINVAL);
 	}
diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
index ee46d8321..557360788 100644
--- a/drivers/gpu/drm/msm/msm_gem_vma.c
+++ b/drivers/gpu/drm/msm/msm_gem_vma.c
@@ -159,3 +159,26 @@ msm_gem_address_space_create(struct device *dev, struct iommu_domain *domain,
 
 	return aspace;
 }
+
+struct msm_gem_address_space *
+msm_gem_address_space_create_a2xx(struct device *dev, struct msm_gpu *gpu,
+		const char *name, uint64_t va_start, uint64_t va_end)
+{
+	struct msm_gem_address_space *aspace;
+	u64 size = va_end - va_start;
+
+	aspace = kzalloc(sizeof(*aspace), GFP_KERNEL);
+	if (!aspace)
+		return ERR_PTR(-ENOMEM);
+
+	spin_lock_init(&aspace->lock);
+	aspace->name = name;
+	aspace->mmu = msm_gpummu_new(dev, gpu);
+
+	drm_mm_init(&aspace->mm, (va_start >> PAGE_SHIFT),
+		size >> PAGE_SHIFT);
+
+	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 c1968d63b..5f3eff304 100644
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -20,6 +20,7 @@
 #include "msm_mmu.h"
 #include "msm_fence.h"
 #include "msm_gpu_trace.h"
+#include "adreno/adreno_gpu.h"
 
 #include <generated/utsrelease.h>
 #include <linux/string_helpers.h>
@@ -825,7 +826,6 @@ static struct msm_gem_address_space *
 msm_gpu_create_address_space(struct msm_gpu *gpu, struct platform_device *pdev,
 		uint64_t va_start, uint64_t va_end)
 {
-	struct iommu_domain *iommu;
 	struct msm_gem_address_space *aspace;
 	int ret;
 
@@ -834,20 +834,27 @@ msm_gpu_create_address_space(struct msm_gpu *gpu, struct platform_device *pdev,
 	 * and have separate page tables per context.  For now, to keep things
 	 * simple and to get something working, just use a single address space:
 	 */
-	iommu = iommu_domain_alloc(&platform_bus_type);
-	if (!iommu)
-		return NULL;
-
-	iommu->geometry.aperture_start = va_start;
-	iommu->geometry.aperture_end = va_end;
-
-	DRM_DEV_INFO(gpu->dev->dev, "%s: using IOMMU\n", gpu->name);
+	if (!adreno_is_a2xx(to_adreno_gpu(gpu))) {
+		struct iommu_domain *iommu = iommu_domain_alloc(&platform_bus_type);
+		if (!iommu)
+			return NULL;
+
+		iommu->geometry.aperture_start = va_start;
+		iommu->geometry.aperture_end = va_end;
+
+		DRM_DEV_INFO(gpu->dev->dev, "%s: using IOMMU\n", gpu->name);
+
+		aspace = msm_gem_address_space_create(&pdev->dev, iommu, "gpu");
+		if (IS_ERR(aspace))
+			iommu_domain_free(iommu);
+	} else {
+		aspace = msm_gem_address_space_create_a2xx(&pdev->dev, gpu, "gpu",
+			va_start, va_end);
+	}
 
-	aspace = msm_gem_address_space_create(&pdev->dev, iommu, "gpu");
 	if (IS_ERR(aspace)) {
-		DRM_DEV_ERROR(gpu->dev->dev, "failed to init iommu: %ld\n",
+		DRM_DEV_ERROR(gpu->dev->dev, "failed to init mmu: %ld\n",
 			PTR_ERR(aspace));
-		iommu_domain_free(iommu);
 		return ERR_CAST(aspace);
 	}
 
diff --git a/drivers/gpu/drm/msm/msm_gpummu.c b/drivers/gpu/drm/msm/msm_gpummu.c
new file mode 100644
index 000000000..a04a64faa
--- /dev/null
+++ b/drivers/gpu/drm/msm/msm_gpummu.c
@@ -0,0 +1,122 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2018 The Linux Foundation. All rights reserved. */
+
+#include "msm_drv.h"
+#include "msm_mmu.h"
+#include "adreno/adreno_gpu.h"
+#include "adreno/a2xx.xml.h"
+
+struct msm_gpummu {
+	struct msm_mmu base;
+	struct msm_gpu *gpu;
+	dma_addr_t pt_base;
+	uint32_t *table;
+};
+#define to_msm_gpummu(x) container_of(x, struct msm_gpummu, base)
+
+#define VA_START SZ_16M
+#define VA_RANGE (0xfff * SZ_64K)
+#define MMU_PAGE_SIZE SZ_4K
+#define TABLE_SIZE (sizeof(uint32_t) * VA_RANGE / MMU_PAGE_SIZE)
+
+static int msm_gpummu_attach(struct msm_mmu *mmu, const char * const *names,
+		int cnt)
+{
+	return 0;
+}
+
+static void msm_gpummu_detach(struct msm_mmu *mmu, const char * const *names,
+		int cnt)
+{
+}
+
+static void update_pt(struct msm_mmu *mmu, uint64_t iova,
+		struct sg_table *sgt, unsigned len, unsigned prot)
+{
+	struct msm_gpummu *gpummu = to_msm_gpummu(mmu);
+	unsigned idx = (iova - VA_START) / MMU_PAGE_SIZE;
+	struct scatterlist *sg;
+	unsigned i, j;
+
+	for_each_sg(sgt->sgl, sg, sgt->nents, i) {
+		dma_addr_t addr = sg->dma_address;
+		for (j = 0; j < sg->length / MMU_PAGE_SIZE; j++, idx++) {
+			gpummu->table[idx] = prot ? addr | prot : 0;
+			addr += MMU_PAGE_SIZE;
+		}
+	}
+
+	/* flush the tlb everytime the page table is updated
+	 * we can probably do something smarter
+	 */
+	gpu_write(gpummu->gpu, REG_A2XX_MH_MMU_INVALIDATE,
+		A2XX_MH_MMU_INVALIDATE_INVALIDATE_ALL |
+		A2XX_MH_MMU_INVALIDATE_INVALIDATE_TC);
+}
+
+static int msm_gpummu_map(struct msm_mmu *mmu, uint64_t iova,
+		struct sg_table *sgt, unsigned len, int prot)
+{
+	unsigned prot_bits = 0;
+	if (prot & IOMMU_WRITE)
+		prot_bits |= 1;
+	if (prot & IOMMU_READ)
+		prot_bits |= 2;
+	update_pt(mmu, iova, sgt, len, prot_bits);
+	return 0;
+}
+
+static int msm_gpummu_unmap(struct msm_mmu *mmu, uint64_t iova,
+		struct sg_table *sgt, unsigned len)
+{
+	update_pt(mmu, iova, sgt, len, 0);
+	return 0;
+}
+
+static void msm_gpummu_destroy(struct msm_mmu *mmu)
+{
+	struct msm_gpummu *gpummu = to_msm_gpummu(mmu);
+
+	dma_free_attrs(mmu->dev, TABLE_SIZE, gpummu->table, gpummu->pt_base,
+		DMA_ATTR_FORCE_CONTIGUOUS);
+
+	kfree(gpummu);
+}
+
+static const struct msm_mmu_funcs funcs = {
+		.attach = msm_gpummu_attach,
+		.detach = msm_gpummu_detach,
+		.map = msm_gpummu_map,
+		.unmap = msm_gpummu_unmap,
+		.destroy = msm_gpummu_destroy,
+};
+
+struct msm_mmu *msm_gpummu_new(struct device *dev, struct msm_gpu *gpu)
+{
+	struct msm_gpummu *gpummu;
+
+	gpummu = kzalloc(sizeof(*gpummu), GFP_KERNEL);
+	if (!gpummu)
+		return ERR_PTR(-ENOMEM);
+
+	gpummu->table = dma_alloc_attrs(dev, TABLE_SIZE + 32, &gpummu->pt_base,
+		GFP_KERNEL | __GFP_ZERO, DMA_ATTR_FORCE_CONTIGUOUS);
+	if (!gpummu->table) {
+		kfree(gpummu);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	gpummu->gpu = gpu;
+	msm_mmu_init(&gpummu->base, dev, &funcs);
+
+	return &gpummu->base;
+}
+
+void msm_gpummu_params(struct msm_mmu *mmu, dma_addr_t *pt_base,
+		dma_addr_t *tran_error)
+{
+	dma_addr_t base = to_msm_gpummu(mmu)->pt_base;
+
+	*pt_base = base;
+	*tran_error = base + TABLE_SIZE; /* 32-byte aligned */
+}
diff --git a/drivers/gpu/drm/msm/msm_mmu.h b/drivers/gpu/drm/msm/msm_mmu.h
index 94c0b83d8..d21b26604 100644
--- a/drivers/gpu/drm/msm/msm_mmu.h
+++ b/drivers/gpu/drm/msm/msm_mmu.h
@@ -53,4 +53,7 @@ static inline void msm_mmu_set_fault_handler(struct msm_mmu *mmu, void *arg,
 	mmu->handler = handler;
 }
 
+void msm_gpummu_params(struct msm_mmu *mmu, dma_addr_t *pt_base,
+		dma_addr_t *tran_error);
+
 #endif /* __MSM_MMU_H__ */
-- 
2.17.1

[PATCH v3 4/4] ARM: dts: imx5: add gpu nodes

[Jonathan Marek]

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
---
 arch/arm/boot/dts/imx51.dtsi | 17 +++++++++++++++++
 arch/arm/boot/dts/imx53.dtsi | 17 +++++++++++++++++
 2 files changed, 34 insertions(+)

diff --git a/arch/arm/boot/dts/imx51.dtsi b/arch/arm/boot/dts/imx51.dtsi
index 67d462715..e9a7bbce9 100644
--- a/arch/arm/boot/dts/imx51.dtsi
+++ b/arch/arm/boot/dts/imx51.dtsi
@@ -628,5 +628,22 @@
 				clock-names = "ipg", "ahb";
 			};
 		};
+
+		gpu: gpu@30000000 {
+			compatible = "amd,imageon-200.1", "amd,imageon";
+			reg = <0x30000000 0x20000>;
+			reg-names = "kgsl_3d0_reg_memory";
+			interrupts = <12>;
+			interrupt-names = "kgsl_3d0_irq";
+			clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
+			clock-names = "core_clk", "mem_iface_clk";
+
+			qcom,gpu-pwrlevels {
+				compatible = "qcom,gpu-pwrlevels";
+				qcom,gpu-pwrlevel@0 {
+					qcom,gpu-freq = <166250000>;
+				};
+			};
+		};
 	};
 };
diff --git a/arch/arm/boot/dts/imx53.dtsi b/arch/arm/boot/dts/imx53.dtsi
index 207eb557c..586d45586 100644
--- a/arch/arm/boot/dts/imx53.dtsi
+++ b/arch/arm/boot/dts/imx53.dtsi
@@ -838,5 +838,22 @@
 			reg = <0xf8000000 0x20000>;
 			clocks = <&clks IMX5_CLK_OCRAM>;
 		};
+
+		gpu: gpu@30000000 {
+			compatible = "amd,imageon-200.0", "amd,imageon";
+			reg = <0x30000000 0x20000>;
+			reg-names = "kgsl_3d0_reg_memory";
+			interrupts = <12>;
+			interrupt-names = "kgsl_3d0_irq";
+			clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
+			clock-names = "core_clk", "mem_iface_clk";
+
+			qcom,gpu-pwrlevels {
+				compatible = "qcom,gpu-pwrlevels";
+				qcom,gpu-pwrlevel@0 {
+					qcom,gpu-freq = <200000000>;
+				};
+			};
+		};
 	};
 };
-- 
2.17.1

Re: [Freedreno] [PATCH v3 4/4] ARM: dts: imx5: add gpu nodes

[Jordan Crouse]

On Mon, Dec 03, 2018 at 04:18:16PM -0500, Jonathan Marek wrote:
> Signed-off-by: Jonathan Marek <jonathan@marek.ca>
> ---
>  arch/arm/boot/dts/imx51.dtsi | 17 +++++++++++++++++
>  arch/arm/boot/dts/imx53.dtsi | 17 +++++++++++++++++
>  2 files changed, 34 insertions(+)
> 
> diff --git a/arch/arm/boot/dts/imx51.dtsi b/arch/arm/boot/dts/imx51.dtsi
> index 67d462715..e9a7bbce9 100644
> --- a/arch/arm/boot/dts/imx51.dtsi
> +++ b/arch/arm/boot/dts/imx51.dtsi
> @@ -628,5 +628,22 @@
>  				clock-names = "ipg", "ahb";
>  			};
>  		};
> +
> +		gpu: gpu@30000000 {
> +			compatible = "amd,imageon-200.1", "amd,imageon";
> +			reg = <0x30000000 0x20000>;
> +			reg-names = "kgsl_3d0_reg_memory";
> +			interrupts = <12>;
> +			interrupt-names = "kgsl_3d0_irq";
> +			clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
> +			clock-names = "core_clk", "mem_iface_clk";
> +
> +			qcom,gpu-pwrlevels {
> +				compatible = "qcom,gpu-pwrlevels";
> +				qcom,gpu-pwrlevel@0 {
> +					qcom,gpu-freq = <166250000>;
> +				};
> +			};

There shouldn't be any incremental cost in the source code to use OPP; it should
just work. And then this won't give us a further reason to keep the legacy
code around when we decide to dump any pretense of downstream "compatibility".

> +		};
>  	};
>  };
> diff --git a/arch/arm/boot/dts/imx53.dtsi b/arch/arm/boot/dts/imx53.dtsi
> index 207eb557c..586d45586 100644
> --- a/arch/arm/boot/dts/imx53.dtsi
> +++ b/arch/arm/boot/dts/imx53.dtsi
> @@ -838,5 +838,22 @@
>  			reg = <0xf8000000 0x20000>;
>  			clocks = <&clks IMX5_CLK_OCRAM>;
>  		};
> +
> +		gpu: gpu@30000000 {
> +			compatible = "amd,imageon-200.0", "amd,imageon";
> +			reg = <0x30000000 0x20000>;
> +			reg-names = "kgsl_3d0_reg_memory";
> +			interrupts = <12>;
> +			interrupt-names = "kgsl_3d0_irq";
> +			clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
> +			clock-names = "core_clk", "mem_iface_clk";
> +
> +			qcom,gpu-pwrlevels {
> +				compatible = "qcom,gpu-pwrlevels";
> +				qcom,gpu-pwrlevel@0 {
> +					qcom,gpu-freq = <200000000>;
> +				};

Same.

> +			};

> +		};
>  	};
>  };
> -- 
> 2.17.1

-- 
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

Re: [PATCH v3 4/4] ARM: dts: imx5: add gpu nodes

[Fabio Estevam]

Hi Jonathan,

Thanks for working on this. Really nice to see GPU support for mx51/mx53!

On Mon, Dec 3, 2018 at 7:21 PM Jonathan Marek <jonathan@marek.ca> wrote:

Please add a commit log.

> Signed-off-by: Jonathan Marek <jonathan@marek.ca>

> ---
>  arch/arm/boot/dts/imx51.dtsi | 17 +++++++++++++++++
>  arch/arm/boot/dts/imx53.dtsi | 17 +++++++++++++++++
>  2 files changed, 34 insertions(+)
>
> diff --git a/arch/arm/boot/dts/imx51.dtsi b/arch/arm/boot/dts/imx51.dtsi
> index 67d462715..e9a7bbce9 100644
> --- a/arch/arm/boot/dts/imx51.dtsi
> +++ b/arch/arm/boot/dts/imx51.dtsi
> @@ -628,5 +628,22 @@
>                                 clock-names = "ipg", "ahb";
>                         };
>                 };
> +
> +               gpu: gpu@30000000 {

We put the peripheral nodes in address order, so this one should go
prior to the IPU node.

> +                       compatible = "amd,imageon-200.1", "amd,imageon";

I can't find the dt-bindings for these compatible entries. Have you
documented them?

> +                       reg = <0x30000000 0x20000>;
> +                       reg-names = "kgsl_3d0_reg_memory";
> +                       interrupts = <12>;
> +                       interrupt-names = "kgsl_3d0_irq";
> +                       clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
> +                       clock-names = "core_clk", "mem_iface_clk";
> +
> +                       qcom,gpu-pwrlevels {
> +                               compatible = "qcom,gpu-pwrlevels";
> +                               qcom,gpu-pwrlevel@0 {

You could drop this @0

Re: [PATCH v3 4/4] ARM: dts: imx5: add gpu nodes

[Jonathan marek]

On 12/03/2018 05:16 PM, Fabio Estevam wrote:
> Hi Jonathan,
> 
> Thanks for working on this. Really nice to see GPU support for mx51/mx53!
> 
> On Mon, Dec 3, 2018 at 7:21 PM Jonathan Marek <jonathan@marek.ca> wrote:
> 
> Please add a commit log.
> 
>> Signed-off-by: Jonathan Marek <jonathan@marek.ca>
> 
>> ---
>>   arch/arm/boot/dts/imx51.dtsi | 17 +++++++++++++++++
>>   arch/arm/boot/dts/imx53.dtsi | 17 +++++++++++++++++
>>   2 files changed, 34 insertions(+)
>>
>> diff --git a/arch/arm/boot/dts/imx51.dtsi b/arch/arm/boot/dts/imx51.dtsi
>> index 67d462715..e9a7bbce9 100644
>> --- a/arch/arm/boot/dts/imx51.dtsi
>> +++ b/arch/arm/boot/dts/imx51.dtsi
>> @@ -628,5 +628,22 @@
>>                                  clock-names = "ipg", "ahb";
>>                          };
>>                  };
>> +
>> +               gpu: gpu@30000000 {
> 
> We put the peripheral nodes in address order, so this one should go
> prior to the IPU node.
> 
>> +                       compatible = "amd,imageon-200.1", "amd,imageon";
> 
> I can't find the dt-bindings for these compatible entries. Have you
> documented them?
> 

It is the same as qcom,adreno which is documented here:

Documentation/devicetree/bindings/display/msm/gpu.txt

I guess I should add amd,imageon there.

>> +                       reg = <0x30000000 0x20000>;
>> +                       reg-names = "kgsl_3d0_reg_memory";
>> +                       interrupts = <12>;
>> +                       interrupt-names = "kgsl_3d0_irq";
>> +                       clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
>> +                       clock-names = "core_clk", "mem_iface_clk";
>> +
>> +                       qcom,gpu-pwrlevels {
>> +                               compatible = "qcom,gpu-pwrlevels";
>> +                               qcom,gpu-pwrlevel@0 {
> 
> You could drop this @0
> 

Re: [Freedreno] [PATCH v3 4/4] ARM: dts: imx5: add gpu nodes

[Jonathan marek]

On 12/03/2018 05:10 PM, Jordan Crouse wrote:
> On Mon, Dec 03, 2018 at 04:18:16PM -0500, Jonathan Marek wrote:
>> Signed-off-by: Jonathan Marek <jonathan@marek.ca>
>> ---
>>   arch/arm/boot/dts/imx51.dtsi | 17 +++++++++++++++++
>>   arch/arm/boot/dts/imx53.dtsi | 17 +++++++++++++++++
>>   2 files changed, 34 insertions(+)
>>
>> diff --git a/arch/arm/boot/dts/imx51.dtsi b/arch/arm/boot/dts/imx51.dtsi
>> index 67d462715..e9a7bbce9 100644
>> --- a/arch/arm/boot/dts/imx51.dtsi
>> +++ b/arch/arm/boot/dts/imx51.dtsi
>> @@ -628,5 +628,22 @@
>>   				clock-names = "ipg", "ahb";
>>   			};
>>   		};
>> +
>> +		gpu: gpu@30000000 {
>> +			compatible = "amd,imageon-200.1", "amd,imageon";
>> +			reg = <0x30000000 0x20000>;
>> +			reg-names = "kgsl_3d0_reg_memory";
>> +			interrupts = <12>;
>> +			interrupt-names = "kgsl_3d0_irq";
>> +			clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
>> +			clock-names = "core_clk", "mem_iface_clk";
>> +
>> +			qcom,gpu-pwrlevels {
>> +				compatible = "qcom,gpu-pwrlevels";
>> +				qcom,gpu-pwrlevel@0 {
>> +					qcom,gpu-freq = <166250000>;
>> +				};
>> +			};
> 
> There shouldn't be any incremental cost in the source code to use OPP; it should
> just work. And then this won't give us a further reason to keep the legacy
> code around when we decide to dump any pretense of downstream "compatibility".
> 

I will switch to OPP.

>> +		};
>>   	};
>>   };
>> diff --git a/arch/arm/boot/dts/imx53.dtsi b/arch/arm/boot/dts/imx53.dtsi
>> index 207eb557c..586d45586 100644
>> --- a/arch/arm/boot/dts/imx53.dtsi
>> +++ b/arch/arm/boot/dts/imx53.dtsi
>> @@ -838,5 +838,22 @@
>>   			reg = <0xf8000000 0x20000>;
>>   			clocks = <&clks IMX5_CLK_OCRAM>;
>>   		};
>> +
>> +		gpu: gpu@30000000 {
>> +			compatible = "amd,imageon-200.0", "amd,imageon";
>> +			reg = <0x30000000 0x20000>;
>> +			reg-names = "kgsl_3d0_reg_memory";
>> +			interrupts = <12>;
>> +			interrupt-names = "kgsl_3d0_irq";
>> +			clocks = <&clks IMX5_CLK_GPU3D_GATE>, <&clks IMX5_CLK_GARB_GATE>;
>> +			clock-names = "core_clk", "mem_iface_clk";
>> +
>> +			qcom,gpu-pwrlevels {
>> +				compatible = "qcom,gpu-pwrlevels";
>> +				qcom,gpu-pwrlevel@0 {
>> +					qcom,gpu-freq = <200000000>;
>> +				};
> 
> Same.
> 
>> +			};
> 
>> +		};
>>   	};
>>   };
>> -- 
>> 2.17.1
> 

Re: [PATCH v3 4/4] ARM: dts: imx5: add gpu nodes

[Fabio Estevam]

On Mon, Dec 3, 2018 at 8:24 PM Jonathan marek <jonathan@marek.ca> wrote:

> > I can't find the dt-bindings for these compatible entries. Have you
> > documented them?
> >
>
> It is the same as qcom,adreno which is documented here:
>
> Documentation/devicetree/bindings/display/msm/gpu.txt
>
> I guess I should add amd,imageon there.

Yes, please.