All of lore.kernel.org
 help / color / mirror / Atom feed
* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-26 11:47 ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-26 11:47 UTC (permalink / raw)
  To: airlied, dri-devel
  Cc: linux-kernel, linux-arm-kernel, kyungmin.park, sw0312.kim,
	jy0922.shim, Inki Dae

This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
but we will add HDMI support also in the future.

this patch is based on git repository below:
git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
branch name: drm-next
commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922

you can refer to our working repository below:
http://git.infradead.org/users/kmpark/linux-2.6-samsung
branch name: samsung-drm

We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
based on Linux framebuffer) but couldn't so because lowlevel codes
of s3c-fb.c are included internally and so FIMD module of this driver has
its own lowlevel codes.

We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
for buffer allocation. by using DMA API, we could use CMA later.

Refer to this link for CMA(Continuous Memory Allocator):
http://lkml.org/lkml/2011/7/20/45

this driver supports only physically continuous memory(non-iommu).

Links to previous versions of the patchset:
v1: < https://lwn.net/Articles/454380/ >
v2: < http://www.spinics.net/lists/kernel/msg1224275.html >

Changelog v2:
DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.

     this feature maps user address space to physical memory region
     once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.

DRM: code clean and add exception codes.

Changelog v3:
DRM: Support multiple irq.

     FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
     this patch supports mutiple irq for Samsung SoC.

DRM: Consider modularization.

     each DRM, FIMD could be built as a module.

DRM: Have indenpendent crtc object.

     crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
     created crtc could be attached to any encoder object.

DRM: code clean and add exception codes.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
---
 drivers/gpu/drm/Kconfig                         |    2 +
 drivers/gpu/drm/Makefile                        |    1 +
 drivers/gpu/drm/samsung/Kconfig                 |   18 +
 drivers/gpu/drm/samsung/Makefile                |   11 +
 drivers/gpu/drm/samsung/samsung_drm_buf.c       |  140 ++++++
 drivers/gpu/drm/samsung/samsung_drm_buf.h       |   78 +++
 drivers/gpu/drm/samsung/samsung_drm_connector.c |  260 ++++++++++
 drivers/gpu/drm/samsung/samsung_drm_connector.h |   34 ++
 drivers/gpu/drm/samsung/samsung_drm_core.c      |  212 ++++++++
 drivers/gpu/drm/samsung/samsung_drm_crtc.c      |  334 +++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_crtc.h      |   36 ++
 drivers/gpu/drm/samsung/samsung_drm_drv.c       |  241 +++++++++
 drivers/gpu/drm/samsung/samsung_drm_encoder.c   |  250 +++++++++
 drivers/gpu/drm/samsung/samsung_drm_encoder.h   |   45 ++
 drivers/gpu/drm/samsung/samsung_drm_fb.c        |  248 +++++++++
 drivers/gpu/drm/samsung/samsung_drm_fb.h        |   47 ++
 drivers/gpu/drm/samsung/samsung_drm_fbdev.c     |  409 +++++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_fbdev.h     |   37 ++
 drivers/gpu/drm/samsung/samsung_drm_fimd.c      |  611 +++++++++++++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_gem.c       |  492 ++++++++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_gem.h       |   76 +++
 include/drm/samsung_drm.h                       |  267 ++++++++++
 22 files changed, 3849 insertions(+), 0 deletions(-)
 create mode 100644 drivers/gpu/drm/samsung/Kconfig
 create mode 100644 drivers/gpu/drm/samsung/Makefile
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_buf.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_buf.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_connector.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_connector.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_core.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_crtc.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_crtc.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_drv.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_encoder.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_encoder.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fb.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fb.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fbdev.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fbdev.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fimd.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_gem.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_gem.h
 create mode 100644 include/drm/samsung_drm.h

diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index b493663..ce6d3ec 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -158,3 +158,5 @@ config DRM_SAVAGE
 	help
 	  Choose this option if you have a Savage3D/4/SuperSavage/Pro/Twister
 	  chipset. If M is selected the module will be called savage.
+
+source "drivers/gpu/drm/samsung/Kconfig"
diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
index 89cf05a..0c6e773 100644
--- a/drivers/gpu/drm/Makefile
+++ b/drivers/gpu/drm/Makefile
@@ -35,4 +35,5 @@ obj-$(CONFIG_DRM_SAVAGE)+= savage/
 obj-$(CONFIG_DRM_VMWGFX)+= vmwgfx/
 obj-$(CONFIG_DRM_VIA)	+=via/
 obj-$(CONFIG_DRM_NOUVEAU) +=nouveau/
+obj-$(CONFIG_DRM_SAMSUNG) +=samsung/
 obj-y			+= i2c/
diff --git a/drivers/gpu/drm/samsung/Kconfig b/drivers/gpu/drm/samsung/Kconfig
new file mode 100644
index 0000000..34cedda
--- /dev/null
+++ b/drivers/gpu/drm/samsung/Kconfig
@@ -0,0 +1,18 @@
+config DRM_SAMSUNG
+	tristate "DRM Support for Samsung SoC EXYNOS Series"
+	depends on DRM && PLAT_SAMSUNG
+	select DRM_KMS_HELPER
+	select FB_CFB_FILLRECT
+	select FB_CFB_COPYAREA
+	select FB_CFB_IMAGEBLIT
+	select VT_HW_CONSOLE_BINDING if FRAMEBUFFER_CONSOLE
+	help
+	  Choose this option if you have a Samsung SoC EXYNOS chipset.
+	  If M is selected the module will be called samsungdrm.
+
+config DRM_SAMSUNG_FIMD
+	tristate "Samsung DRM FIMD"
+	depends on DRM_SAMSUNG
+	help
+	  Choose this option if you want to use Samsung FIMD for DRM.
+	  If M is selected, the module will be called samsung_drm_fimd
diff --git a/drivers/gpu/drm/samsung/Makefile b/drivers/gpu/drm/samsung/Makefile
new file mode 100644
index 0000000..70f89f3
--- /dev/null
+++ b/drivers/gpu/drm/samsung/Makefile
@@ -0,0 +1,11 @@
+#
+# Makefile for the drm device driver.  This driver provides support for the
+# Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher.
+
+ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/samsung
+samsungdrm-y := samsung_drm_drv.o samsung_drm_encoder.o samsung_drm_connector.o \
+		samsung_drm_crtc.o samsung_drm_fbdev.o samsung_drm_fb.o \
+		samsung_drm_buf.o samsung_drm_gem.o samsung_drm_core.o
+
+obj-$(CONFIG_DRM_SAMSUNG) += samsungdrm.o
+obj-$(CONFIG_DRM_SAMSUNG_FIMD) += samsung_drm_fimd.o
diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c b/drivers/gpu/drm/samsung/samsung_drm_buf.c
new file mode 100644
index 0000000..563d07e
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
@@ -0,0 +1,140 @@
+/* samsung_drm_buf.c
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_buf.h"
+
+static DEFINE_MUTEX(samsung_drm_buf_lock);
+
+static int lowlevel_buffer_allocate(struct drm_device *dev,
+		struct samsung_drm_buf_entry *entry)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
+			(dma_addr_t *)&entry->paddr, GFP_KERNEL);
+	if (!entry->paddr) {
+		DRM_ERROR("failed to allocate buffer.\n");
+		return -ENOMEM;
+	}
+
+	DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
+			(unsigned int)entry->vaddr, entry->paddr, entry->size);
+
+	return 0;
+}
+
+static void lowlevel_buffer_deallocate(struct drm_device *dev,
+		struct samsung_drm_buf_entry *entry)
+{
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	dma_free_writecombine(dev->dev, entry->size, entry->vaddr,
+			entry->paddr);
+
+	DRM_DEBUG_KMS("deallocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
+			(unsigned int)entry->vaddr, entry->paddr, entry->size);
+}
+
+static void  samsung_drm_buf_del(struct drm_device *dev,
+		struct samsung_drm_gem_obj *obj)
+{
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	lowlevel_buffer_deallocate(dev, obj->entry);
+
+	kfree(obj->entry);
+
+	kfree(obj);
+}
+
+struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
+		unsigned int size)
+{
+	struct samsung_drm_gem_obj *obj;
+	struct samsung_drm_buf_entry *entry;
+	int ret;
+
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+	DRM_DEBUG_KMS("desired size = 0x%x\n", size);
+
+	obj = kzalloc(sizeof(*obj), GFP_KERNEL);
+	if (!obj) {
+		DRM_ERROR("failed to allocate samsung_drm_gem_obj.\n");
+		return NULL;
+	}
+
+	/* use only one memory plane yet. */
+	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
+	if (!entry) {
+		DRM_ERROR("failed to allocate samsung_drm_buf_entry.\n");
+		return NULL;
+	}
+
+	entry->size = size;
+
+	/* allocate memory region and set it to vaddr and paddr. */
+	ret = lowlevel_buffer_allocate(dev, entry);
+	if (ret < 0)
+		return NULL;
+
+	obj->entry = entry;
+
+	return obj;
+}
+
+struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
+		unsigned int size)
+{
+	struct samsung_drm_gem_obj *obj;
+
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	obj = samsung_drm_buf_new(dev, size);
+	if (!obj)
+		return NULL;
+
+	DRM_DEBUG_KMS("buffer id : 0x%x\n", obj->id);
+
+	return obj;
+}
+
+int samsung_drm_buf_destroy(struct drm_device *dev,
+		struct samsung_drm_gem_obj *in_obj)
+{
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	samsung_drm_buf_del(dev, in_obj);
+
+	return 0;
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Buffer Management Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
new file mode 100644
index 0000000..d6a7e95
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
@@ -0,0 +1,78 @@
+/* samsung_drm_buf.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_BUF_H_
+#define _SAMSUNG_DRM_BUF_H_
+
+/**
+ * samsung drm buffer entry structure.
+ *
+ * @paddr: physical address of allocated memory.
+ * @vaddr: kernel virtual address of allocated memory.
+ * @size: size of allocated memory.
+ */
+struct samsung_drm_buf_entry {
+	unsigned int paddr;
+	void __iomem *vaddr;
+	unsigned int size;
+};
+
+/**
+ * samsung drm buffer structure.
+ *
+ * @entry: pointer to samsung drm buffer entry object.
+ * @flags: it means memory type to be alloated or cache attributes.
+ * @handle: pointer to specific buffer object.
+ * @id: unique id to specific buffer object.
+ *
+ * ps. this object would be transfered to user as kms_bo.handle so
+ *	user can access to memory through kms_bo.handle.
+ */
+struct samsung_drm_gem_obj {
+	struct drm_gem_object base;
+	struct samsung_drm_buf_entry *entry;
+	unsigned int flags;
+
+	unsigned int handle;
+	unsigned int id;
+};
+
+/* create new buffer object and memory region and add the object to list. */
+struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
+		unsigned int size);
+
+/* allocate physical memory and add its object to list. */
+struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
+		unsigned int size);
+
+/* remove allocated physical memory. */
+int samsung_drm_buf_destroy(struct drm_device *dev,
+		struct samsung_drm_gem_obj *in_obj);
+
+/* find object added to list. */
+struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device *dev,
+		struct samsung_drm_gem_obj *in_obj, unsigned int paddr);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c b/drivers/gpu/drm/samsung/samsung_drm_connector.c
new file mode 100644
index 0000000..987a629
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
@@ -0,0 +1,260 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_encoder.h"
+
+#define MAX_EDID 256
+#define to_samsung_connector(x)	container_of(x, struct samsung_drm_connector,\
+				drm_connector)
+
+struct samsung_drm_connector {
+	struct drm_connector	drm_connector;
+	struct drm_encoder	*encoder;
+};
+
+/* convert samsung_video_timings to drm_display_mode */
+static inline void
+convert_to_display_mode(struct fb_videomode *timing,
+			struct drm_display_mode *mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode->clock = timing->pixclock / 1000;
+
+	mode->hdisplay = timing->xres;
+	mode->hsync_start = mode->hdisplay + timing->left_margin;
+	mode->hsync_end = mode->hsync_start + timing->hsync_len;
+	mode->htotal = mode->hsync_end + timing->right_margin;
+
+	mode->vdisplay = timing->yres;
+	mode->vsync_start = mode->vdisplay + timing->upper_margin;
+	mode->vsync_end = mode->vsync_start + timing->vsync_len;
+	mode->vtotal = mode->vsync_end + timing->lower_margin;
+}
+
+/* convert drm_display_mode to samsung_video_timings */
+static inline void
+convert_to_video_timing(struct drm_display_mode *mode,
+			struct fb_videomode *timing)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	timing->pixclock = mode->clock * 1000;
+
+	timing->xres = mode->hdisplay;
+	timing->left_margin = mode->hsync_start - mode->hdisplay;
+	timing->hsync_len = mode->hsync_end - mode->hsync_start;
+	timing->right_margin = mode->htotal - mode->hsync_end;
+
+	timing->yres = mode->vdisplay;
+	timing->upper_margin = mode->vsync_start - mode->vdisplay;
+	timing->vsync_len = mode->vsync_end - mode->vsync_start;
+	timing->lower_margin = mode->vtotal - mode->vsync_end;
+}
+
+static int samsung_drm_connector_get_modes(struct drm_connector *connector)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+	struct samsung_drm_display *display =
+		samsung_drm_get_manager(samsung_connector->encoder)->display;
+	unsigned int count;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (display && display->get_edid) {
+		void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
+		if (!edid) {
+			DRM_ERROR("failed to allocate edid\n");
+			return 0;
+		}
+
+		display->get_edid(connector, edid, MAX_EDID);
+
+		drm_mode_connector_update_edid_property(connector, edid);
+		count = drm_add_edid_modes(connector, edid);
+
+		kfree(connector->display_info.raw_edid);
+		connector->display_info.raw_edid = edid;
+	} else {
+		struct drm_display_mode *mode = drm_mode_create(connector->dev);
+		struct fb_videomode *timing;
+
+		if (display && display->get_timing)
+			timing = display->get_timing();
+		else
+			return 0;
+
+		convert_to_display_mode(timing, mode);
+
+		mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
+		drm_mode_set_name(mode);
+		drm_mode_probed_add(connector, mode);
+
+		count = 1;
+	}
+
+	return count;
+}
+
+static int samsung_drm_connector_mode_valid(struct drm_connector *connector,
+					    struct drm_display_mode *mode)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+	struct samsung_drm_display *display =
+		samsung_drm_get_manager(samsung_connector->encoder)->display;
+	struct fb_videomode timing;
+	int ret = MODE_BAD;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	convert_to_video_timing(mode, &timing);
+
+	if (display && display->check_timing)
+		if (!display->check_timing((void *)&timing))
+			ret = MODE_OK;
+
+	return ret;
+}
+
+struct drm_encoder *samsung_drm_best_encoder(struct drm_connector *connector)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	return to_samsung_connector(connector)->encoder;
+}
+
+static struct drm_connector_helper_funcs samsung_connector_helper_funcs = {
+	.get_modes	= samsung_drm_connector_get_modes,
+	.mode_valid	= samsung_drm_connector_mode_valid,
+	.best_encoder	= samsung_drm_best_encoder,
+};
+
+/* get detection status of display device. */
+static enum drm_connector_status
+samsung_drm_connector_detect(struct drm_connector *connector, bool force)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+	struct samsung_drm_display *display =
+		samsung_drm_get_manager(samsung_connector->encoder)->display;
+	unsigned int ret = connector_status_unknown;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (display && display->is_connected) {
+		if (display->is_connected())
+			ret = connector_status_connected;
+		else
+			ret = connector_status_disconnected;
+	}
+
+	return ret;
+}
+
+static void samsung_drm_connector_destroy(struct drm_connector *connector)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	drm_sysfs_connector_remove(connector);
+	drm_connector_cleanup(connector);
+	connector->dev->mode_config.num_connector--;
+	kfree(samsung_connector);
+}
+
+static struct drm_connector_funcs samsung_connector_funcs = {
+	.dpms		= drm_helper_connector_dpms,
+	.fill_modes	= drm_helper_probe_single_connector_modes,
+	.detect		= samsung_drm_connector_detect,
+	.destroy	= samsung_drm_connector_destroy,
+};
+
+struct drm_connector *samsung_drm_connector_create(struct drm_device *dev,
+						   struct drm_encoder *encoder)
+{
+	struct samsung_drm_connector *samsung_connector;
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+	struct drm_connector *connector;
+	int type;
+	int err;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_connector = kzalloc(sizeof(*samsung_connector), GFP_KERNEL);
+	if (!samsung_connector) {
+		DRM_ERROR("failed to allocate connector\n");
+		return NULL;
+	}
+
+	samsung_connector->encoder = encoder;
+	connector = &samsung_connector->drm_connector;
+
+	switch (manager->display->type) {
+	case SAMSUNG_DISPLAY_TYPE_HDMI:
+		type = DRM_MODE_CONNECTOR_HDMIA;
+		break;
+	default:
+		type = DRM_MODE_CONNECTOR_Unknown;
+		break;
+	}
+
+	drm_connector_init(dev, connector, &samsung_connector_funcs, type);
+	drm_connector_helper_add(connector, &samsung_connector_helper_funcs);
+
+	err = drm_sysfs_connector_add(connector);
+	if (err)
+		goto err_connector;
+
+	err = drm_mode_connector_attach_encoder(connector, encoder);
+	if (err) {
+		DRM_ERROR("failed to attach a connector to a encoder\n");
+		goto err_sysfs;
+	}
+
+	DRM_DEBUG_KMS("connector has been created\n");
+
+	return connector;
+
+err_sysfs:
+	drm_sysfs_connector_remove(connector);
+err_connector:
+	drm_connector_cleanup(connector);
+	kfree(samsung_connector);
+	return NULL;
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Connector Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.h b/drivers/gpu/drm/samsung/samsung_drm_connector.h
new file mode 100644
index 0000000..638d2b3
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_connector.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_CONNECTOR_H_
+#define _SAMSUNG_DRM_CONNECTOR_H_
+
+struct drm_connector *samsung_drm_connector_create(struct drm_device *dev,
+						   struct drm_encoder *encoder);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_core.c b/drivers/gpu/drm/samsung/samsung_drm_core.c
new file mode 100644
index 0000000..6221922
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_core.c
@@ -0,0 +1,212 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "samsung_drm_encoder.h"
+#include "samsung_drm_connector.h"
+#include "samsung_drm_fbdev.h"
+
+#include <drm/samsung_drm.h>
+
+static DEFINE_MUTEX(samsung_drm_mutex);
+static LIST_HEAD(samsung_drm_subdrv_list);
+static struct drm_device *drm_dev;
+
+static int samsung_drm_subdrv_probe(struct drm_device *dev,
+				    struct samsung_drm_subdrv *subdrv)
+{
+	struct drm_encoder *encoder;
+	struct drm_connector *connector;
+	int ret;
+
+	if (subdrv->probe) {
+		ret = subdrv->probe(dev);
+		if (ret)
+			return ret;
+	}
+
+	/* all crtc is available */
+	encoder = samsung_drm_encoder_create(dev, &subdrv->manager,
+			(1 << MAX_CRTC) - 1);
+	if (!encoder) {
+		DRM_ERROR("failed to create encoder\n");
+		return -EFAULT;
+	}
+
+	connector = samsung_drm_connector_create(dev, encoder);
+	if (!connector) {
+		DRM_ERROR("failed to create connector\n");
+		encoder->funcs->destroy(encoder);
+		return -EFAULT;
+	}
+
+	subdrv->encoder = encoder;
+	subdrv->connector = connector;
+
+	return 0;
+}
+
+static void samsung_drm_subdrv_remove(struct drm_device *dev,
+				      struct samsung_drm_subdrv *subdrv)
+{
+	if (subdrv->remove)
+		subdrv->remove(dev);
+
+	if (subdrv->encoder) {
+		struct drm_encoder *encoder = subdrv->encoder;
+		encoder->funcs->destroy(encoder);
+		subdrv->encoder = NULL;
+	}
+
+	if (subdrv->connector) {
+		struct drm_connector *connector = subdrv->connector;
+		connector->funcs->destroy(connector);
+		subdrv->connector = NULL;
+	}
+}
+
+int samsung_drm_device_register(struct drm_device *dev)
+{
+	struct samsung_drm_subdrv *subdrv, *n;
+	int err;
+
+	if (!dev)
+		return -EINVAL;
+
+	if (drm_dev) {
+		DRM_ERROR("Already drm device were registered\n");
+		return -EBUSY;
+	}
+
+	mutex_lock(&samsung_drm_mutex);
+	list_for_each_entry_safe(subdrv, n, &samsung_drm_subdrv_list, list) {
+		err = samsung_drm_subdrv_probe(dev, subdrv);
+		if (err) {
+			DRM_ERROR("samsung drm subdrv probe failed\n");
+			list_del(&subdrv->list);
+		}
+	}
+
+	drm_dev = dev;
+	mutex_unlock(&samsung_drm_mutex);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(samsung_drm_device_register);
+
+void samsung_drm_device_unregister(struct drm_device *dev)
+{
+	struct samsung_drm_subdrv *subdrv;
+
+	if (!dev || dev != drm_dev) {
+		WARN(1, "Unexpected drm device unregister!\n");
+		return;
+	}
+
+	mutex_lock(&samsung_drm_mutex);
+	list_for_each_entry(subdrv, &samsung_drm_subdrv_list, list)
+		samsung_drm_subdrv_remove(dev, subdrv);
+
+	drm_dev = NULL;
+	mutex_unlock(&samsung_drm_mutex);
+}
+EXPORT_SYMBOL_GPL(samsung_drm_device_unregister);
+
+static int samsung_drm_mode_group_reinit(struct drm_device *dev)
+{
+	struct drm_mode_group *group = &dev->primary->mode_group;
+	uint32_t *id_list = group->id_list;
+	int ret;
+
+	ret = drm_mode_group_init_legacy_group(dev, group);
+	if (ret < 0)
+		return ret;
+
+	kfree(id_list);
+	return 0;
+}
+
+int samsung_drm_subdrv_register(struct samsung_drm_subdrv *subdrv)
+{
+	int err;
+
+	if (!subdrv)
+		return -EINVAL;
+
+	mutex_lock(&samsung_drm_mutex);
+	if (drm_dev) {
+		err = samsung_drm_subdrv_probe(drm_dev, subdrv);
+		if (err) {
+			DRM_ERROR("failed to probe samsung drm subdrv\n");
+			mutex_unlock(&samsung_drm_mutex);
+			return err;
+		}
+
+		err = samsung_drm_fbdev_reinit(drm_dev);
+		if (err) {
+			DRM_ERROR("failed to reinitialize samsung drm fbdev\n");
+			samsung_drm_subdrv_remove(drm_dev, subdrv);
+			mutex_unlock(&samsung_drm_mutex);
+			return err;
+		}
+
+		err = samsung_drm_mode_group_reinit(drm_dev);
+		if (err) {
+			DRM_ERROR("failed to reinitialize mode group\n");
+			samsung_drm_fbdev_fini(drm_dev);
+			samsung_drm_subdrv_remove(drm_dev, subdrv);
+			mutex_unlock(&samsung_drm_mutex);
+			return err;
+		}
+	}
+
+	list_add_tail(&subdrv->list, &samsung_drm_subdrv_list);
+	mutex_unlock(&samsung_drm_mutex);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(samsung_drm_subdrv_register);
+
+void samsung_drm_subdrv_unregister(struct samsung_drm_subdrv *subdrv)
+{
+	if (!subdrv) {
+		WARN(1, "Unexpected samsung drm subdrv unregister!\n");
+		return;
+	}
+
+	mutex_lock(&samsung_drm_mutex);
+	if (drm_dev) {
+		samsung_drm_subdrv_remove(drm_dev, subdrv);
+
+		/* FIXME: error check */
+		samsung_drm_fbdev_reinit(drm_dev);
+		samsung_drm_mode_group_reinit(drm_dev);
+	}
+
+	list_del(&subdrv->list);
+	mutex_unlock(&samsung_drm_mutex);
+}
+EXPORT_SYMBOL_GPL(samsung_drm_subdrv_unregister);
diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
new file mode 100644
index 0000000..1c5a70a
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
@@ -0,0 +1,334 @@
+/* samsung_drm_crtc.c
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_fb.h"
+#include "samsung_drm_encoder.h"
+
+#define to_samsung_crtc(x)	container_of(x, struct samsung_drm_crtc,\
+				drm_crtc)
+
+/*
+ * @fb_x: horizontal position from framebuffer base
+ * @fb_y: vertical position from framebuffer base
+ * @base_x: horizontal position from screen base
+ * @base_y: vertical position from screen base
+ * @crtc_w: width of crtc
+ * @crtc_h: height of crtc
+ */
+struct samsung_drm_crtc_pos {
+	unsigned int fb_x;
+	unsigned int fb_y;
+	unsigned int base_x;
+	unsigned int base_y;
+	unsigned int crtc_w;
+	unsigned int crtc_h;
+};
+
+struct samsung_drm_crtc {
+	struct drm_crtc			drm_crtc;
+	struct samsung_drm_overlay	overlay;
+	unsigned int			pipe;
+};
+
+static void samsung_drm_overlay_update(struct samsung_drm_overlay *overlay,
+				       struct drm_framebuffer *fb,
+				       struct drm_display_mode *mode,
+				       struct samsung_drm_crtc_pos *pos)
+{
+	struct samsung_drm_buffer_info buffer_info;
+	unsigned int actual_w = pos->crtc_w;
+	unsigned int actual_h = pos->crtc_h;
+	unsigned int hw_w;
+	unsigned int hw_h;
+
+	/* update buffer address of framebuffer. */
+	samsung_drm_fb_update_buf_off(fb, pos->fb_x, pos->fb_y, &buffer_info);
+	overlay->paddr = buffer_info.paddr;
+	overlay->vaddr = buffer_info.vaddr;
+
+	hw_w = mode->hdisplay - pos->base_x;
+	hw_h = mode->vdisplay - pos->base_y;
+
+	if (actual_w > hw_w)
+		actual_w = hw_w;
+	if (actual_h > hw_h)
+		actual_h = hw_h;
+
+	overlay->offset_x = pos->base_x;
+	overlay->offset_y = pos->base_y;
+	overlay->width = actual_w;
+	overlay->height = actual_h;
+
+	DRM_DEBUG_KMS("overlay : offset_x/y(%d,%d), width/height(%d,%d)",
+			overlay->offset_x, overlay->offset_y,
+			overlay->width, overlay->height);
+
+	overlay->buf_offsize = fb->width - actual_w;
+	overlay->line_size = actual_w;
+	overlay->end_buf_off = fb->width * actual_h;
+}
+
+static int samsung_drm_crtc_update(struct drm_crtc *crtc)
+{
+	struct samsung_drm_crtc *samsung_crtc;
+	struct samsung_drm_overlay *overlay;
+	struct samsung_drm_crtc_pos pos;
+	struct drm_display_mode *mode = &crtc->mode;
+	struct drm_framebuffer *fb = crtc->fb;
+
+	if (!mode || !fb)
+		return -EINVAL;
+
+	samsung_crtc = to_samsung_crtc(crtc);
+	overlay = &samsung_crtc->overlay;
+
+	memset(&pos, 0, sizeof(struct samsung_drm_crtc_pos));
+	pos.fb_x = crtc->x;
+	pos.fb_y = crtc->y;
+	pos.crtc_w = fb->width - crtc->x;
+	pos.crtc_h = fb->height - crtc->y;
+
+	samsung_drm_overlay_update(overlay, crtc->fb, mode, &pos);
+
+	return 0;
+}
+
+/* CRTC helper functions */
+static void samsung_drm_crtc_dpms(struct drm_crtc *crtc, int mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* TODO */
+}
+
+static void samsung_drm_crtc_prepare(struct drm_crtc *crtc)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* drm framework doesn't check NULL. */
+}
+
+static void samsung_drm_crtc_commit(struct drm_crtc *crtc)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_drm_fn_encoder(crtc, NULL, samsung_drm_encoder_crtc_commit);
+}
+
+static bool
+samsung_drm_crtc_mode_fixup(struct drm_crtc *crtc,
+			    struct drm_display_mode *mode,
+			    struct drm_display_mode *adjusted_mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* drm framework doesn't check NULL */
+	return true;
+}
+
+static int
+samsung_drm_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode,
+			  struct drm_display_mode *adjusted_mode, int x, int y,
+			  struct drm_framebuffer *old_fb)
+{
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_overlay *overlay = &samsung_crtc->overlay;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode = adjusted_mode;
+
+	ret = samsung_drm_crtc_update(crtc);
+	if (ret)
+		return ret;
+
+	samsung_drm_fn_encoder(crtc, overlay,
+			samsung_drm_encoder_crtc_mode_set);
+
+	return ret;
+}
+
+static int samsung_drm_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
+					  struct drm_framebuffer *old_fb)
+{
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_overlay *overlay = &samsung_crtc->overlay;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	ret = samsung_drm_crtc_update(crtc);
+	if (ret)
+		return ret;
+
+	samsung_drm_fn_encoder(crtc, overlay,
+			samsung_drm_encoder_crtc_mode_set);
+	samsung_drm_fn_encoder(crtc, NULL, samsung_drm_encoder_crtc_commit);
+
+	return ret;
+}
+
+static void samsung_drm_crtc_load_lut(struct drm_crtc *crtc)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* drm framework doesn't check NULL */
+}
+
+static struct drm_crtc_helper_funcs samsung_crtc_helper_funcs = {
+	.dpms		= samsung_drm_crtc_dpms,
+	.prepare	= samsung_drm_crtc_prepare,
+	.commit		= samsung_drm_crtc_commit,
+	.mode_fixup	= samsung_drm_crtc_mode_fixup,
+	.mode_set	= samsung_drm_crtc_mode_set,
+	.mode_set_base	= samsung_drm_crtc_mode_set_base,
+	.load_lut	= samsung_drm_crtc_load_lut,
+};
+
+/* CRTC functions */
+static int samsung_drm_crtc_page_flip(struct drm_crtc *crtc,
+				      struct drm_framebuffer *fb,
+				      struct drm_pending_vblank_event *event)
+{
+	struct drm_device *dev = crtc->dev;
+	struct samsung_drm_private *dev_priv = dev->dev_private;
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_overlay *overlay = &samsung_crtc->overlay;
+	struct drm_framebuffer *old_fb = crtc->fb;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mutex_lock(&dev->struct_mutex);
+
+	if (event && !dev_priv->pageflip_event) {
+		list_add_tail(&event->base.link,
+				&dev_priv->pageflip_event_list);
+
+		ret = drm_vblank_get(dev, samsung_crtc->pipe);
+		if (ret) {
+			DRM_DEBUG("failed to acquire vblank counter\n");
+			goto out;
+		}
+
+		dev_priv->pageflip_event = true;
+	}
+
+	crtc->fb = fb;
+
+	ret = samsung_drm_crtc_update(crtc);
+	if (ret) {
+		crtc->fb = old_fb;
+		if (event && dev_priv->pageflip_event) {
+			drm_vblank_put(dev, samsung_crtc->pipe);
+			dev_priv->pageflip_event = false;
+		}
+
+		goto out;
+	}
+
+	samsung_drm_fn_encoder(crtc, overlay,
+			samsung_drm_encoder_crtc_mode_set);
+	samsung_drm_fn_encoder(crtc, NULL, samsung_drm_encoder_crtc_commit);
+
+out:
+	mutex_unlock(&dev->struct_mutex);
+	return ret;
+}
+
+static void samsung_drm_crtc_destroy(struct drm_crtc *crtc)
+{
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_private *private = crtc->dev->dev_private;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	private->crtc[samsung_crtc->pipe] = NULL;
+
+	drm_crtc_cleanup(crtc);
+	kfree(samsung_crtc);
+}
+
+static struct drm_crtc_funcs samsung_crtc_funcs = {
+	.set_config	= drm_crtc_helper_set_config,
+	.page_flip	= samsung_drm_crtc_page_flip,
+	.destroy	= samsung_drm_crtc_destroy,
+};
+
+int samsung_drm_crtc_create(struct drm_device *dev, unsigned int nr)
+{
+	struct samsung_drm_crtc *samsung_crtc;
+	struct samsung_drm_private *private = dev->dev_private;
+	struct drm_crtc *crtc;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_crtc = kzalloc(sizeof(*samsung_crtc), GFP_KERNEL);
+	if (!samsung_crtc) {
+		DRM_ERROR("failed to allocate samsung crtc\n");
+		return -ENOMEM;
+	}
+
+	samsung_crtc->pipe = nr;
+	crtc = &samsung_crtc->drm_crtc;
+
+	private->crtc[nr] = crtc;
+
+	drm_crtc_init(dev, crtc, &samsung_crtc_funcs);
+	drm_crtc_helper_add(crtc, &samsung_crtc_helper_funcs);
+
+	return 0;
+}
+
+int samsung_drm_crtc_enable_vblank(struct drm_device *dev, int crtc)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_drm_fn_encoder(private->crtc[crtc], &crtc,
+			samsung_drm_enable_vblank);
+
+	return 0;
+}
+
+void samsung_drm_crtc_disable_vblank(struct drm_device *dev, int crtc)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_drm_fn_encoder(private->crtc[crtc], &crtc,
+			samsung_drm_disable_vblank);
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM CRTC Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.h b/drivers/gpu/drm/samsung/samsung_drm_crtc.h
new file mode 100644
index 0000000..9256c7e
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.h
@@ -0,0 +1,36 @@
+/* samsung_drm_crtc.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_CRTC_H_
+#define _SAMSUNG_DRM_CRTC_H_
+
+int samsung_drm_crtc_create(struct drm_device *dev, unsigned int nr);
+int samsung_drm_crtc_enable_vblank(struct drm_device *dev, int crtc);
+void samsung_drm_crtc_disable_vblank(struct drm_device *dev, int crtc);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_drv.c b/drivers/gpu/drm/samsung/samsung_drm_drv.c
new file mode 100644
index 0000000..eefb08b
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_drv.c
@@ -0,0 +1,241 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_crtc.h"
+#include "samsung_drm_fbdev.h"
+#include "samsung_drm_fb.h"
+#include "samsung_drm_gem.h"
+
+#define DRIVER_NAME	"samsung-drm"
+#define DRIVER_DESC	"Samsung SoC DRM"
+#define DRIVER_DATE	"20110530"
+#define DRIVER_MAJOR	1
+#define DRIVER_MINOR	0
+
+static int samsung_drm_load(struct drm_device *dev, unsigned long flags)
+{
+	struct samsung_drm_private *private;
+	int ret;
+	int nr;
+
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	private = kzalloc(sizeof(struct samsung_drm_private), GFP_KERNEL);
+	if (!private) {
+		DRM_ERROR("failed to allocate private\n");
+		return -ENOMEM;
+	}
+
+	INIT_LIST_HEAD(&private->pageflip_event_list);
+	dev->dev_private = (void *)private;
+
+	drm_mode_config_init(dev);
+
+	samsung_drm_mode_config_init(dev);
+
+	for (nr = 0; nr < MAX_CRTC; nr++) {
+		ret = samsung_drm_crtc_create(dev, nr);
+		if (ret)
+			goto err_crtc;
+	}
+
+	ret = drm_vblank_init(dev, MAX_CRTC);
+	if (ret)
+		goto err_crtc;
+
+	ret = samsung_drm_device_register(dev);
+	if (ret)
+		goto err_vblank;
+
+	ret = samsung_drm_fbdev_init(dev);
+	if (ret) {
+		DRM_ERROR("failed to initialize drm fbdev\n");
+		goto err_drm_device;
+	}
+
+	return 0;
+
+err_drm_device:
+	samsung_drm_device_unregister(dev);
+err_vblank:
+	drm_vblank_cleanup(dev);
+err_crtc:
+	drm_mode_config_cleanup(dev);
+	kfree(private);
+
+	return ret;
+}
+
+static int samsung_drm_unload(struct drm_device *dev)
+{
+	samsung_drm_fbdev_fini(dev);
+	samsung_drm_device_unregister(dev);
+	drm_vblank_cleanup(dev);
+	drm_mode_config_cleanup(dev);
+	kfree(dev->dev_private);
+
+	return 0;
+}
+
+static int samsung_drm_open(struct drm_device *dev, struct drm_file *file_priv)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	return 0;
+}
+
+static void samsung_drm_lastclose(struct drm_device *dev)
+{
+	samsung_drm_fbdev_restore_mode(dev);
+}
+
+static int samsung_drm_master_create(struct drm_device *dev,
+		struct drm_master *master)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	master->driver_priv = NULL;
+
+	return 0;
+}
+
+static int samsung_drm_master_set(struct drm_device *dev,
+				  struct drm_file *file_priv, bool from_open)
+{
+	struct drm_master *master = file_priv->master;
+
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	master->lock.hw_lock = kzalloc(sizeof(struct drm_hw_lock), GFP_KERNEL);
+	if (!master->lock.hw_lock) {
+		DRM_DEBUG("failed to allocate hw_lock\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static struct vm_operations_struct samsung_drm_gem_vm_ops = {
+	.fault = samsung_drm_gem_fault,
+	.open = drm_gem_vm_open,
+	.close = drm_gem_vm_close,
+};
+
+static struct drm_ioctl_desc samsung_ioctls[] = {
+	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
+			DRM_UNLOCKED),
+	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
+			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
+	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
+			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
+};
+
+static struct drm_driver samsung_drm_driver = {
+	.driver_features	= DRIVER_HAVE_IRQ | DRIVER_BUS_PLATFORM |
+				  DRIVER_MODESET | DRIVER_GEM,
+	.load			= samsung_drm_load,
+	.unload			= samsung_drm_unload,
+	.open			= samsung_drm_open,
+	.lastclose		= samsung_drm_lastclose,
+	.get_vblank_counter	= drm_vblank_count,
+	.enable_vblank		= samsung_drm_crtc_enable_vblank,
+	.disable_vblank		= samsung_drm_crtc_disable_vblank,
+	.master_create		= samsung_drm_master_create,
+	.master_set		= samsung_drm_master_set,
+	.gem_init_object	= samsung_drm_gem_init_object,
+	.gem_free_object	= samsung_drm_gem_free_object,
+	.gem_vm_ops		= &samsung_drm_gem_vm_ops,
+	.dumb_create		= samsung_drm_gem_dumb_create,
+	.dumb_map_offset	= samsung_drm_gem_dumb_map_offset,
+	.dumb_destroy		= samsung_drm_gem_dumb_destroy,
+	.ioctls			= samsung_ioctls,
+	.fops = {
+		.owner		= THIS_MODULE,
+		.open		= drm_open,
+		.mmap		= samsung_drm_gem_mmap,
+		.poll		= drm_poll,
+		.read		= drm_read,
+		.unlocked_ioctl	= drm_ioctl,
+		.release	= drm_release,
+	},
+	.name	= DRIVER_NAME,
+	.desc	= DRIVER_DESC,
+	.date	= DRIVER_DATE,
+	.major	= DRIVER_MAJOR,
+	.minor	= DRIVER_MINOR,
+};
+
+static int samsung_drm_platform_probe(struct platform_device *pdev)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	samsung_drm_driver.num_ioctls = DRM_ARRAY_SIZE(samsung_ioctls);
+
+	return drm_platform_init(&samsung_drm_driver, pdev);
+}
+
+static int samsung_drm_platform_remove(struct platform_device *pdev)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	drm_platform_exit(&samsung_drm_driver, pdev);
+
+	return 0;
+}
+
+static struct platform_driver samsung_drm_platform_driver = {
+	.probe		= samsung_drm_platform_probe,
+	.remove		= __devexit_p(samsung_drm_platform_remove),
+	.driver		= {
+		.owner	= THIS_MODULE,
+		.name	= DRIVER_NAME,
+	},
+};
+
+static int __init samsung_drm_init(void)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	return platform_driver_register(&samsung_drm_platform_driver);
+}
+
+static void __exit samsung_drm_exit(void)
+{
+	platform_driver_unregister(&samsung_drm_platform_driver);
+}
+
+module_init(samsung_drm_init);
+module_exit(samsung_drm_exit);
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_encoder.c b/drivers/gpu/drm/samsung/samsung_drm_encoder.c
new file mode 100644
index 0000000..d1cb04e
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_encoder.c
@@ -0,0 +1,250 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_encoder.h"
+
+#define to_samsung_encoder(x)	container_of(x, struct samsung_drm_encoder,\
+				drm_encoder)
+
+struct samsung_drm_encoder {
+	struct drm_encoder		drm_encoder;
+	struct samsung_drm_manager	*manager;
+};
+
+static void samsung_drm_encoder_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_connector *connector;
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+
+	DRM_DEBUG_KMS("%s, encoder dpms: %d\n", __FILE__, mode);
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder) {
+			struct samsung_drm_display *display = manager->display;
+
+			if (display && display->power_on)
+				display->power_on(mode);
+		}
+	}
+}
+
+static bool
+samsung_drm_encoder_mode_fixup(struct drm_encoder *encoder,
+			       struct drm_display_mode *mode,
+			       struct drm_display_mode *adjusted_mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return true;
+}
+
+static void samsung_drm_encoder_mode_set(struct drm_encoder *encoder,
+					 struct drm_display_mode *mode,
+					 struct drm_display_mode *adjusted_mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_connector *connector;
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode = adjusted_mode;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder)
+			if (manager_ops && manager_ops->mode_set)
+				manager_ops->mode_set(manager->dev, mode);
+	}
+}
+
+static void samsung_drm_encoder_prepare(struct drm_encoder *encoder)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+}
+
+static void samsung_drm_encoder_commit(struct drm_encoder *encoder)
+{
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (manager_ops && manager_ops->commit)
+		manager_ops->commit(manager->dev);
+}
+
+static struct drm_crtc *
+samsung_drm_encoder_get_crtc(struct drm_encoder *encoder)
+{
+	return encoder->crtc;
+}
+
+static struct drm_encoder_helper_funcs samsung_encoder_helper_funcs = {
+	.dpms		= samsung_drm_encoder_dpms,
+	.mode_fixup	= samsung_drm_encoder_mode_fixup,
+	.mode_set	= samsung_drm_encoder_mode_set,
+	.prepare	= samsung_drm_encoder_prepare,
+	.commit		= samsung_drm_encoder_commit,
+	.get_crtc	= samsung_drm_encoder_get_crtc,
+};
+
+static void samsung_drm_encoder_destroy(struct drm_encoder *encoder)
+{
+	struct samsung_drm_encoder *samsung_encoder =
+		to_samsung_encoder(encoder);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_encoder->manager->pipe = -1;
+
+	drm_encoder_cleanup(encoder);
+	encoder->dev->mode_config.num_encoder--;
+	kfree(samsung_encoder);
+}
+
+static struct drm_encoder_funcs samsung_encoder_funcs = {
+	.destroy = samsung_drm_encoder_destroy,
+};
+
+struct drm_encoder *
+samsung_drm_encoder_create(struct drm_device *dev,
+			   struct samsung_drm_manager *manager,
+			   unsigned int possible_crtcs)
+{
+	struct drm_encoder *encoder;
+	struct samsung_drm_encoder *samsung_encoder;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!manager || !possible_crtcs)
+		return NULL;
+
+	if (!manager->dev)
+		return NULL;
+
+	samsung_encoder = kzalloc(sizeof(*samsung_encoder), GFP_KERNEL);
+	if (!samsung_encoder) {
+		DRM_ERROR("failed to allocate encoder\n");
+		return NULL;
+	}
+
+	samsung_encoder->manager = manager;
+	encoder = &samsung_encoder->drm_encoder;
+	encoder->possible_crtcs = possible_crtcs;
+
+	DRM_DEBUG_KMS("possible_crtcs = 0x%x\n", encoder->possible_crtcs);
+
+	drm_encoder_init(dev, encoder, &samsung_encoder_funcs,
+			DRM_MODE_ENCODER_TMDS);
+
+	drm_encoder_helper_add(encoder, &samsung_encoder_helper_funcs);
+
+	DRM_DEBUG_KMS("encoder has been created\n");
+
+	return encoder;
+}
+
+struct samsung_drm_manager *samsung_drm_get_manager(struct drm_encoder *encoder)
+{
+	return to_samsung_encoder(encoder)->manager;
+}
+
+void samsung_drm_fn_encoder(struct drm_crtc *crtc, void *data,
+			    void (*fn)(struct drm_encoder *, void *))
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_encoder *encoder;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (encoder->crtc != crtc)
+			continue;
+
+		fn(encoder, data);
+	}
+}
+
+void samsung_drm_enable_vblank(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+	int crtc = *(int *)data;
+
+	if (manager->pipe == -1)
+		manager->pipe = crtc;
+
+	/* old_crtc checking needs? FIXME!!! */
+
+	if (manager_ops->enable_vblank)
+		manager_ops->enable_vblank(manager->dev);
+}
+
+void samsung_drm_disable_vblank(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+	int crtc = *(int *)data;
+
+	if (manager->pipe == -1)
+		manager->pipe = crtc;
+
+	/* old_crtc checking needs? FIXME!!! */
+
+	if (manager_ops->disable_vblank)
+		manager_ops->disable_vblank(manager->dev);
+}
+
+void samsung_drm_encoder_crtc_commit(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_overlay_ops *overlay_ops = manager->overlay_ops;
+
+	overlay_ops->commit(manager->dev);
+}
+
+void samsung_drm_encoder_crtc_mode_set(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_overlay_ops *overlay_ops = manager->overlay_ops;
+	struct samsung_drm_overlay *overlay = data;
+
+	overlay_ops->mode_set(manager->dev, overlay);
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Encoder Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_encoder.h b/drivers/gpu/drm/samsung/samsung_drm_encoder.h
new file mode 100644
index 0000000..99040b2
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_encoder.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_ENCODER_H_
+#define _SAMSUNG_DRM_ENCODER_H_
+
+struct samsung_drm_manager;
+
+struct drm_encoder *samsung_drm_encoder_create(struct drm_device *dev,
+					       struct samsung_drm_manager *mgr,
+					       unsigned int possible_crtcs);
+struct samsung_drm_manager *
+samsung_drm_get_manager(struct drm_encoder *encoder);
+void samsung_drm_fn_encoder(struct drm_crtc *crtc, void *data,
+			    void (*fn)(struct drm_encoder *, void *));
+void samsung_drm_enable_vblank(struct drm_encoder *encoder, void *data);
+void samsung_drm_disable_vblank(struct drm_encoder *encoder, void *data);
+void samsung_drm_encoder_crtc_commit(struct drm_encoder *encoder, void *data);
+void samsung_drm_encoder_crtc_mode_set(struct drm_encoder *encoder, void *data);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c b/drivers/gpu/drm/samsung/samsung_drm_fb.c
new file mode 100644
index 0000000..42096eb
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
@@ -0,0 +1,248 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc.h"
+#include "drm_crtc_helper.h"
+
+#include "samsung_drm_fb.h"
+#include "samsung_drm_buf.h"
+#include "samsung_drm_gem.h"
+
+#define to_samsung_fb(x)	container_of(x, struct samsung_drm_fb, fb)
+
+struct samsung_drm_fb {
+	struct drm_framebuffer		fb;
+	struct drm_file			*file_priv;
+	struct samsung_drm_gem_obj	*samsung_gem_obj;
+	unsigned int			is_default;
+
+	/* samsung gem object handle. */
+	unsigned int			gem_handle;
+	/* unique id to buffer object. */
+	unsigned int			id;
+
+	unsigned int			fb_size;
+	unsigned long			paddr;
+	void __iomem			*vaddr;
+};
+
+static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
+{
+	struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	drm_framebuffer_cleanup(fb);
+
+	if (samsung_fb->is_default) {
+		ret = drm_gem_handle_delete(samsung_fb->file_priv,
+				samsung_fb->gem_handle);
+		if (ret < 0)
+			DRM_ERROR("failed to delete drm_gem_handle.\n");
+	}
+
+	kfree(samsung_fb);
+}
+
+static int samsung_drm_fb_create_handle(struct drm_framebuffer *fb,
+					struct drm_file *file_priv,
+					unsigned int *handle)
+{
+	struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!samsung_fb->gem_handle) {
+		DRM_ERROR("can't get id to buffer object.\n");
+		return -EINVAL;
+	}
+
+	*handle = samsung_fb->gem_handle;
+
+	DRM_DEBUG_KMS("got buffer object id(%d)\n", *handle);
+
+	return 0;
+}
+
+static int samsung_drm_fb_dirty(struct drm_framebuffer *fb,
+				struct drm_file *file_priv, unsigned flags,
+				unsigned color, struct drm_clip_rect *clips,
+				unsigned num_clips)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	/*
+	 * update framebuffer and its hardware.
+	 * - this callback would be called by user application
+	 *	with DRM_IOCTL_MODE_DIRTYFB command.
+	 *
+	 * ps. Userspace can notify the driver via this callback
+	 * that an area of the framebuffer has been changed then should
+	 * be flushed to the display hardware.
+	 */
+
+	return 0;
+}
+
+static struct drm_framebuffer_funcs samsung_drm_fb_funcs = {
+	.destroy	= samsung_drm_fb_destroy,
+	.create_handle	= samsung_drm_fb_create_handle,
+	.dirty		= samsung_drm_fb_dirty,
+};
+
+static struct drm_framebuffer *
+samsung_drm_fb_init(struct drm_file *file_priv, struct drm_device *dev,
+		    struct drm_mode_fb_cmd *mode_cmd)
+{
+	struct samsung_drm_fb *samsung_fb;
+	struct drm_framebuffer *fb;
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+	unsigned int size, gem_handle = 0;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode_cmd->pitch = max(mode_cmd->pitch,
+			mode_cmd->width * (mode_cmd->bpp >> 3));
+
+	DRM_LOG_KMS("drm fb create(%dx%d)\n",
+			mode_cmd->width, mode_cmd->height);
+
+	samsung_fb = kzalloc(sizeof(*samsung_fb), GFP_KERNEL);
+	if (!samsung_fb) {
+		DRM_ERROR("failed to allocate samsung drm framebuffer.\n");
+		return ERR_PTR(-ENOMEM);
+	}
+
+	fb = &samsung_fb->fb;
+	ret = drm_framebuffer_init(dev, fb, &samsung_drm_fb_funcs);
+	if (ret) {
+		DRM_ERROR("failed to initialize framebuffer.\n");
+		goto fail;
+	}
+
+	DRM_LOG_KMS("create: fb id: %d\n", fb->base.id);
+
+	size = mode_cmd->pitch * mode_cmd->height;
+
+	/*
+	 * mode_cmd->handle could be pointer to a buffer allocated by user
+	 * application using KMS library.
+	 */
+	if (!mode_cmd->handle) {
+		samsung_gem_obj = samsung_drm_buf_create(dev, size);
+		if (!samsung_gem_obj)
+			return ERR_PTR(-ENOMEM);
+	} else {
+		samsung_gem_obj = find_samsung_drm_gem_object(file_priv, dev,
+				mode_cmd->handle);
+		if (!samsung_gem_obj)
+			return ERR_PTR(-EINVAL);
+
+		gem_handle = mode_cmd->handle;
+	}
+
+	/* first is default framebuffer */
+	if (dev->mode_config.num_fb == 1)
+		samsung_fb->is_default = 1;
+
+	samsung_fb->file_priv = file_priv;
+	samsung_fb->samsung_gem_obj = samsung_gem_obj;
+	samsung_fb->gem_handle = gem_handle;
+	samsung_fb->id = samsung_gem_obj->id;
+	samsung_fb->fb_size = size;
+	samsung_fb->vaddr = samsung_gem_obj->entry->vaddr;
+	samsung_fb->paddr = samsung_gem_obj->entry->paddr;
+
+	DRM_DEBUG_KMS("handle = 0x%x, id = %d\n",
+			samsung_gem_obj->handle, samsung_gem_obj->id);
+	DRM_DEBUG_KMS("fb: size = 0x%x, vaddr = 0x%x, paddr = 0x%x\n",
+			samsung_fb->fb_size, (unsigned int)samsung_fb->vaddr,
+			(unsigned int)samsung_fb->paddr);
+
+	drm_helper_mode_fill_fb_struct(fb, mode_cmd);
+
+	return fb;
+
+fail:
+	kfree(samsung_fb);
+
+	return ERR_PTR(ret);
+}
+
+struct drm_framebuffer *samsung_drm_fb_create(struct drm_device *dev,
+					      struct drm_file *file_priv,
+					      struct drm_mode_fb_cmd *mode_cmd)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return samsung_drm_fb_init(file_priv, dev, mode_cmd);
+}
+
+void samsung_drm_fb_update_buf_off(struct drm_framebuffer *fb,
+				   unsigned int x, unsigned int y,
+				   struct samsung_drm_buffer_info *info)
+{
+	struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
+	unsigned long offset;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	offset = x * (fb->bits_per_pixel >> 3);
+	offset += y * fb->pitch;
+
+	info->base_addr = samsung_fb->paddr;
+	info->vaddr = samsung_fb->vaddr + offset;
+	info->paddr = samsung_fb->paddr + offset;
+
+	DRM_DEBUG_KMS("updated vaddr = 0x%x, paddr = 0x%x, offset = 0x%x\n",
+			(unsigned int)info->vaddr,
+			(unsigned int)info->paddr,
+			(unsigned int)offset);
+}
+
+static struct drm_mode_config_funcs samsung_drm_mode_config_funcs = {
+	.fb_create = samsung_drm_fb_create,
+};
+
+void samsung_drm_mode_config_init(struct drm_device *dev)
+{
+	dev->mode_config.min_width = 0;
+	dev->mode_config.min_height = 0;
+
+	/*
+	 * It sets max width and height as default value(4096x4096).
+	 * this value would be used to check for framebuffer size limitation
+	 * at drm_mode_addfb().
+	 */
+	dev->mode_config.max_width = 4096;
+	dev->mode_config.max_height = 4096;
+
+	dev->mode_config.funcs = &samsung_drm_mode_config_funcs;
+}
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.h b/drivers/gpu/drm/samsung/samsung_drm_fb.h
new file mode 100644
index 0000000..5c54521
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fb.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_FB_H_
+#define _SAMSUNG_DRM_FB_H
+
+struct samsung_drm_buffer_info {
+	unsigned long base_addr;
+	unsigned long paddr;
+	void __iomem *vaddr;
+};
+
+void samsung_drm_fb_update_buf_off(struct drm_framebuffer *fb,
+				   unsigned int x, unsigned int y,
+				   struct samsung_drm_buffer_info *info);
+
+struct drm_framebuffer *samsung_drm_fb_create(struct drm_device *dev,
+					      struct drm_file *file_priv,
+					      struct drm_mode_fb_cmd *mode_cmd);
+
+void samsung_drm_mode_config_init(struct drm_device *dev);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fbdev.c b/drivers/gpu/drm/samsung/samsung_drm_fbdev.c
new file mode 100644
index 0000000..9349d20
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fbdev.c
@@ -0,0 +1,409 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc.h"
+#include "drm_fb_helper.h"
+#include "drm_crtc_helper.h"
+#include "samsung_drm_fb.h"
+
+#include <drm/samsung_drm.h>
+
+#define to_samsung_fbdev(x)	container_of(x, struct samsung_drm_fbdev,\
+				drm_fb_helper)
+
+struct samsung_drm_fbdev {
+	struct drm_fb_helper	drm_fb_helper;
+	struct drm_framebuffer	*fb;
+};
+
+static inline unsigned int chan_to_field(unsigned int chan,
+					 struct fb_bitfield *bf)
+{
+	chan &= 0xffff;
+	chan >>= 16 - bf->length;
+
+	return chan << bf->offset;
+}
+
+static int samsung_drm_fbdev_setcolreg(unsigned regno, unsigned red,
+				       unsigned green, unsigned blue,
+				       unsigned transp, struct fb_info *info)
+{
+	unsigned int val;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	switch (info->fix.visual) {
+	case FB_VISUAL_TRUECOLOR:
+		if (regno < 16) {
+			u32 *pal = info->pseudo_palette;
+
+			val = chan_to_field(red, &info->var.red);
+			val |= chan_to_field(green, &info->var.green);
+			val |= chan_to_field(blue, &info->var.blue);
+
+			pal[regno] = val;
+		}
+		break;
+	default:
+		return 1;
+	}
+
+	return 0;
+}
+
+static struct fb_ops samsung_drm_fb_ops = {
+	.owner		= THIS_MODULE,
+	.fb_fillrect	= cfb_fillrect,
+	.fb_copyarea	= cfb_copyarea,
+	.fb_imageblit	= cfb_imageblit,
+	.fb_check_var	= drm_fb_helper_check_var,
+	.fb_set_par	= drm_fb_helper_set_par,
+	.fb_setcolreg	= samsung_drm_fbdev_setcolreg,
+	.fb_blank	= drm_fb_helper_blank,
+	.fb_pan_display	= drm_fb_helper_pan_display,
+	.fb_setcmap	= drm_fb_helper_setcmap,
+};
+
+static void samsung_drm_fbdev_update(struct drm_fb_helper *helper,
+				     struct drm_framebuffer *fb,
+				     unsigned int fb_width,
+				     unsigned int fb_height)
+{
+	struct fb_info *fbi = helper->fbdev;
+	struct drm_device *dev = helper->dev;
+	struct samsung_drm_fbdev *samsung_fb = to_samsung_fbdev(helper);
+	struct samsung_drm_buffer_info buffer_info;
+	unsigned int size = fb_width * fb_height * (fb->bits_per_pixel >> 3);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_fb->fb = fb;
+
+	drm_fb_helper_fill_fix(fbi, fb->pitch, fb->depth);
+	drm_fb_helper_fill_var(fbi, helper, fb_width, fb_height);
+
+	samsung_drm_fb_update_buf_off(fb, fbi->var.xoffset, fbi->var.yoffset,
+			&buffer_info);
+
+	dev->mode_config.fb_base = buffer_info.base_addr;
+
+	fbi->screen_base = buffer_info.vaddr;
+	fbi->screen_size = size;
+	fbi->fix.smem_start = buffer_info.paddr;
+	fbi->fix.smem_len = size;
+}
+
+static int samsung_drm_fbdev_create(struct drm_fb_helper *helper,
+				    struct drm_fb_helper_surface_size *sizes)
+{
+	struct samsung_drm_fbdev *samsung_fbdev = to_samsung_fbdev(helper);
+	struct drm_device *dev = helper->dev;
+	struct fb_info *fbi;
+	struct drm_mode_fb_cmd mode_cmd = { 0 };
+	struct platform_device *pdev = dev->platformdev;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
+			sizes->surface_width, sizes->surface_height,
+			sizes->surface_bpp);
+
+	mode_cmd.width = sizes->surface_width;
+	mode_cmd.height = sizes->surface_height;
+	mode_cmd.bpp = sizes->surface_bpp;
+	mode_cmd.depth = sizes->surface_depth;
+
+	mutex_lock(&dev->struct_mutex);
+
+	fbi = framebuffer_alloc(0, &pdev->dev);
+	if (!fbi) {
+		DRM_ERROR("failed to allocate fb info.\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	samsung_fbdev->fb = samsung_drm_fb_create(dev, NULL, &mode_cmd);
+	if (IS_ERR(samsung_fbdev->fb)) {
+		DRM_ERROR("failed to create drm dramebuffer.\n");
+		ret = PTR_ERR(samsung_fbdev->fb);
+		goto out;
+	}
+
+	helper->fb = samsung_fbdev->fb;
+	helper->fbdev = fbi;
+
+	fbi->par = helper;
+	fbi->flags = FBINFO_FLAG_DEFAULT;
+	fbi->fbops = &samsung_drm_fb_ops;
+
+	ret = fb_alloc_cmap(&fbi->cmap, 256, 0);
+	if (ret) {
+		DRM_ERROR("failed to allocate cmap.\n");
+		goto out;
+	}
+
+	samsung_drm_fbdev_update(helper, helper->fb, sizes->fb_width,
+			sizes->fb_height);
+
+out:
+	mutex_unlock(&dev->struct_mutex);
+	return ret;
+}
+
+static bool
+samsung_drm_fbdev_is_samefb(struct drm_framebuffer *fb,
+			    struct drm_fb_helper_surface_size *sizes)
+{
+	if (fb->width != sizes->surface_width)
+		return false;
+	if (fb->height != sizes->surface_height)
+		return false;
+	if (fb->bits_per_pixel != sizes->surface_bpp)
+		return false;
+	if (fb->depth != sizes->surface_depth)
+		return false;
+
+	return true;
+}
+
+static int samsung_drm_fbdev_recreate(struct drm_fb_helper *helper,
+				      struct drm_fb_helper_surface_size *sizes)
+{
+	struct drm_device *dev = helper->dev;
+	struct samsung_drm_fbdev *samsung_fbdev = to_samsung_fbdev(helper);
+	struct drm_framebuffer *fb = samsung_fbdev->fb;
+	struct drm_mode_fb_cmd mode_cmd = { 0 };
+
+	if (helper->fb != fb) {
+		DRM_ERROR("drm framebuffer is different\n");
+		return -EINVAL;
+	}
+
+	if (samsung_drm_fbdev_is_samefb(fb, sizes))
+		return 0;
+
+	mode_cmd.width = sizes->surface_width;
+	mode_cmd.height = sizes->surface_height;
+	mode_cmd.bpp = sizes->surface_bpp;
+	mode_cmd.depth = sizes->surface_depth;
+
+	if (fb->funcs->destroy)
+		fb->funcs->destroy(fb);
+
+	samsung_fbdev->fb = samsung_drm_fb_create(dev, NULL, &mode_cmd);
+	if (IS_ERR(samsung_fbdev->fb)) {
+		DRM_ERROR("failed to allocate fb.\n");
+		return PTR_ERR(samsung_fbdev->fb);
+	}
+
+	helper->fb = samsung_fbdev->fb;
+	samsung_drm_fbdev_update(helper, helper->fb, sizes->fb_width,
+			sizes->fb_height);
+
+	return 0;
+}
+
+static int samsung_drm_fbdev_probe(struct drm_fb_helper *helper,
+				   struct drm_fb_helper_surface_size *sizes)
+{
+	int ret = 0;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!helper->fb) {
+		ret = samsung_drm_fbdev_create(helper, sizes);
+		if (ret < 0) {
+			DRM_ERROR("failed to create fbdev.\n");
+			return ret;
+		}
+
+		/*
+		 * fb_helper expects a value more than 1 if succeed
+		 * because register_framebuffer() should be called.
+		 */
+		ret = 1;
+	} else {
+		ret = samsung_drm_fbdev_recreate(helper, sizes);
+		if (ret < 0) {
+			DRM_ERROR("failed to reconfigure fbdev\n");
+			return ret;
+		}
+	}
+
+	return ret;
+}
+
+static struct drm_fb_helper_funcs samsung_drm_fb_helper_funcs = {
+	.fb_probe =	samsung_drm_fbdev_probe,
+};
+
+int samsung_drm_fbdev_init(struct drm_device *dev)
+{
+	struct samsung_drm_fbdev *fbdev;
+	struct samsung_drm_private *private = dev->dev_private;
+	struct drm_fb_helper *helper;
+	unsigned int num_crtc;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!dev->mode_config.num_crtc || !dev->mode_config.num_connector)
+		return 0;
+
+	fbdev = kzalloc(sizeof(*fbdev), GFP_KERNEL);
+	if (!fbdev) {
+		DRM_ERROR("failed to allocate drm fbdev.\n");
+		return -ENOMEM;
+	}
+
+	private->fb_helper = helper = &fbdev->drm_fb_helper;
+	helper->funcs = &samsung_drm_fb_helper_funcs;
+
+	num_crtc = dev->mode_config.num_crtc;
+
+	ret = drm_fb_helper_init(dev, helper, num_crtc, 4);
+	if (ret < 0) {
+		DRM_ERROR("failed to initialize drm fb helper.\n");
+		goto fail;
+	}
+
+	ret = drm_fb_helper_single_add_all_connectors(helper);
+	if (ret < 0) {
+		DRM_ERROR("failed to register drm_fb_helper_connector.\n");
+		goto fail;
+
+	}
+
+	ret = drm_fb_helper_initial_config(helper, 32);
+	if (ret < 0) {
+		DRM_ERROR("failed to set up hw configuration.\n");
+		goto fail;
+	}
+
+	return 0;
+
+fail:
+	private->fb_helper = NULL;
+	kfree(fbdev);
+
+	return ret;
+}
+
+static void samsung_drm_fbdev_destroy(struct drm_device *dev,
+				      struct drm_fb_helper *fb_helper)
+{
+	struct drm_framebuffer *fb;
+	struct fb_info *info;
+
+	/* release drm framebuffer and real buffer */
+	if (fb_helper->fb && fb_helper->fb->funcs) {
+		fb = fb_helper->fb;
+		if (fb->funcs->destroy)
+			fb->funcs->destroy(fb);
+	}
+
+	/* release linux framebuffer */
+	if (fb_helper->fbdev) {
+		info = fb_helper->fbdev;
+		unregister_framebuffer(info);
+		if (info->cmap.len)
+			fb_dealloc_cmap(&info->cmap);
+		framebuffer_release(info);
+	}
+
+	drm_fb_helper_fini(fb_helper);
+}
+
+void samsung_drm_fbdev_fini(struct drm_device *dev)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+	struct samsung_drm_fbdev *fbdev;
+
+	if (!private || !private->fb_helper)
+		return;
+
+	fbdev = to_samsung_fbdev(private->fb_helper);
+
+	samsung_drm_fbdev_destroy(dev, private->fb_helper);
+	kfree(fbdev);
+	private->fb_helper = NULL;
+}
+
+void samsung_drm_fbdev_restore_mode(struct drm_device *dev)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+
+	if (!private || !private->fb_helper)
+		return;
+
+	drm_fb_helper_restore_fbdev_mode(private->fb_helper);
+}
+
+int samsung_drm_fbdev_reinit(struct drm_device *dev)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+	int ret;
+
+	if (!private)
+		return -EINVAL;
+
+	if (!dev->mode_config.num_connector) {
+		samsung_drm_fbdev_fini(dev);
+		return 0;
+	}
+
+	if (private->fb_helper) {
+		struct drm_fb_helper *fb_helper = private->fb_helper;
+
+		drm_fb_helper_fini(fb_helper);
+
+		ret = drm_fb_helper_init(dev, fb_helper,
+				dev->mode_config.num_crtc, 4);
+		if (ret < 0) {
+			DRM_ERROR("failed to initialize drm fb helper\n");
+			return ret;
+		}
+
+		ret = drm_fb_helper_single_add_all_connectors(fb_helper);
+		if (ret < 0) {
+			DRM_ERROR("failed to add fb helper to connectors\n");
+			return ret;
+		}
+
+		ret = drm_fb_helper_initial_config(fb_helper, 32);
+		if (ret < 0) {
+			DRM_ERROR("failed to set up hw configuration.\n");
+			return ret;
+		}
+	} else
+		ret = samsung_drm_fbdev_init(dev);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fbdev.h b/drivers/gpu/drm/samsung/samsung_drm_fbdev.h
new file mode 100644
index 0000000..3ef4e0d
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fbdev.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ *
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_FBDEV_H_
+#define _SAMSUNG_DRM_FBDEV_H_
+
+int samsung_drm_fbdev_init(struct drm_device *dev);
+int samsung_drm_fbdev_reinit(struct drm_device *dev);
+void samsung_drm_fbdev_fini(struct drm_device *dev);
+void samsung_drm_fbdev_restore_mode(struct drm_device *dev);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fimd.c b/drivers/gpu/drm/samsung/samsung_drm_fimd.c
new file mode 100644
index 0000000..e4dadc1
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fimd.c
@@ -0,0 +1,611 @@
+/*
+ * Copyright (C) 2011 Samsung Electronics Co.Ltd
+ * Author: Joonyoung Shim <jy0922.shim@samsung.com>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+
+/*
+ * TODO list
+ *  - Code cleanup(FIXME and TODO parts)
+ *  - Support multiple driver instance
+ *  - Clock gating and Power management
+ *  - Writeback feature
+ */
+
+#include "drmP.h"
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/clk.h>
+
+#include <drm/samsung_drm.h>
+#include <plat/regs-fb-v4.h>
+
+#include "samsung_drm_fbdev.h"
+#include "samsung_drm_crtc.h"
+
+/* irq_flags bits */
+#define FIMD_VSYNC_IRQ_EN	0
+
+#define VIDOSD_A(win)	(VIDOSD_BASE + 0x00 + (win) * 16)
+#define VIDOSD_B(win)	(VIDOSD_BASE + 0x04 + (win) * 16)
+#define VIDOSD_C(win)	(VIDOSD_BASE + 0x08 + (win) * 16)
+#define VIDOSD_D(win)	(VIDOSD_BASE + 0x0C + (win) * 16)
+
+#define VIDWx_BUF_START(win, buf)	(VIDW_BUF_START(buf) + (win) * 8)
+#define VIDWx_BUF_END(win, buf)		(VIDW_BUF_END(buf) + (win) * 8)
+#define VIDWx_BUF_SIZE(win, buf)	(VIDW_BUF_SIZE(buf) + (win) * 4)
+
+#define WINDOWS_NR	5
+
+#define get_fimd_data(dev)	platform_get_drvdata(to_platform_device(dev))
+
+struct fimd_win_data {
+	unsigned int		offset_x;
+	unsigned int		offset_y;
+	unsigned int		width;
+	unsigned int		height;
+	unsigned int		paddr;
+	void __iomem		*vaddr;
+	unsigned int		end_buf_off;
+	unsigned int		buf_offsize;
+	unsigned int		line_size;	/* bytes */
+};
+
+struct fimd_data {
+	struct drm_crtc			*crtc;
+	struct clk			*bus_clk;
+	struct clk			*lcd_clk;
+	struct resource			*regs_res;
+	void __iomem			*regs;
+	struct fimd_win_data		win_data[WINDOWS_NR];
+	unsigned int			clkdiv;
+	unsigned int			default_win;
+	unsigned int			bpp;
+	unsigned long			irq_flags;
+	u32				vidcon0;
+	u32				vidcon1;
+};
+
+/* TODO: remove global variables */
+static struct samsung_drm_subdrv fimd_subdrv;
+static struct fb_videomode fimd_timing;
+
+static bool fimd_display_is_connected(void)
+{
+	return true;
+}
+
+static void *fimd_get_timing(void)
+{
+	return &fimd_timing;
+}
+
+static int fimd_check_timing(void *timing)
+{
+	return 0;
+}
+
+static int fimd_display_power_on(int mode)
+{
+	return 0;
+}
+
+static struct samsung_drm_display fimd_display = {
+	.type = SAMSUNG_DISPLAY_TYPE_LCD,
+	.is_connected = fimd_display_is_connected,
+	.get_timing = fimd_get_timing,
+	.check_timing = fimd_check_timing,
+	.power_on = fimd_display_power_on,
+};
+
+static void fimd_commit(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	/* vidcon0 */
+	val = data->vidcon0;
+	val &= ~(VIDCON0_CLKVAL_F_MASK | VIDCON0_CLKDIR);
+
+	if (data->clkdiv > 1)
+		val |= VIDCON0_CLKVAL_F(data->clkdiv - 1) | VIDCON0_CLKDIR;
+	else
+		val &= ~VIDCON0_CLKDIR;	/* 1:1 clock */
+
+	val |= VIDCON0_ENVID | VIDCON0_ENVID_F;
+	writel(val, regs + VIDCON0);
+
+	/* vidcon1 */
+	writel(data->vidcon1, regs + VIDCON1);
+
+	/* vidtcon0 */
+	val = VIDTCON0_VBPD(fimd_timing.upper_margin - 1) |
+	       VIDTCON0_VFPD(fimd_timing.lower_margin - 1) |
+	       VIDTCON0_VSPW(fimd_timing.vsync_len - 1);
+	writel(val, regs + VIDTCON0);
+
+	/* vidtcon1 */
+	val = VIDTCON1_HBPD(fimd_timing.left_margin - 1) |
+	       VIDTCON1_HFPD(fimd_timing.right_margin - 1) |
+	       VIDTCON1_HSPW(fimd_timing.hsync_len - 1);
+	writel(val, regs + VIDTCON1);
+
+	/* vidtcon2 */
+	val = VIDTCON2_LINEVAL(fimd_timing.yres - 1) |
+	       VIDTCON2_HOZVAL(fimd_timing.xres - 1);
+	writel(val, regs + VIDTCON2);
+}
+
+static int fimd_enable_vblank(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	if (!test_and_set_bit(FIMD_VSYNC_IRQ_EN, &data->irq_flags)) {
+		val = readl(regs + VIDINTCON0);
+
+		val |= VIDINTCON0_INT_ENABLE;
+		val |= VIDINTCON0_INT_FRAME;
+
+		val &= ~VIDINTCON0_FRAMESEL0_MASK;
+		val |= VIDINTCON0_FRAMESEL0_VSYNC;
+		val &= ~VIDINTCON0_FRAMESEL1_MASK;
+		val |= VIDINTCON0_FRAMESEL1_NONE;
+
+		writel(val, regs + VIDINTCON0);
+	}
+
+	return 0;
+}
+
+static void fimd_disable_vblank(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	if (test_and_clear_bit(FIMD_VSYNC_IRQ_EN, &data->irq_flags)) {
+		val = readl(regs + VIDINTCON0);
+
+		val &= ~VIDINTCON0_INT_FRAME;
+		val &= ~VIDINTCON0_INT_ENABLE;
+
+		writel(val, regs + VIDINTCON0);
+	}
+}
+
+static struct samsung_drm_manager_ops fimd_manager_ops = {
+	.commit = fimd_commit,
+	.enable_vblank = fimd_enable_vblank,
+	.disable_vblank = fimd_disable_vblank,
+};
+
+static void fimd_win_mode_set(struct device *dev,
+			      struct samsung_drm_overlay *overlay)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	struct fimd_win_data *win_data;
+
+	if (!overlay) {
+		dev_err(dev, "overlay is NULL\n");
+		return;
+	}
+
+	win_data = &data->win_data[data->default_win];
+
+	win_data->offset_x = overlay->offset_x;
+	win_data->offset_y = overlay->offset_y;
+	win_data->width = overlay->width;
+	win_data->height = overlay->height;
+	win_data->paddr = overlay->paddr;
+	win_data->vaddr = overlay->vaddr;
+	win_data->end_buf_off = overlay->end_buf_off * (data->bpp >> 3);
+	win_data->buf_offsize = overlay->buf_offsize * (data->bpp >> 3);
+	win_data->line_size = overlay->line_size * (data->bpp >> 3);
+}
+
+static void fimd_win_commit(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	struct fimd_win_data *win_data;
+	int win = data->default_win;
+	u32 val;
+
+	if (win < 0 || win > WINDOWS_NR)
+		return;
+
+	win_data = &data->win_data[win];
+
+	/* protect windows */
+	val = readl(regs + SHADOWCON);
+	val |= SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+
+	/* buffer start address */
+	val = win_data->paddr;
+	writel(val, regs + VIDWx_BUF_START(win, 0));
+
+	/* buffer end address */
+	val = win_data->paddr + win_data->end_buf_off;
+	writel(val, regs + VIDWx_BUF_END(win, 0));
+
+	/* buffer size */
+	val = VIDW_BUF_SIZE_OFFSET(win_data->buf_offsize) |
+		VIDW_BUF_SIZE_PAGEWIDTH(win_data->line_size);
+	writel(val, regs + VIDWx_BUF_SIZE(win, 0));
+
+	/* OSD position */
+	val = VIDOSDxA_TOPLEFT_X(win_data->offset_x) |
+		VIDOSDxA_TOPLEFT_Y(win_data->offset_y);
+	writel(val, regs + VIDOSD_A(win));
+
+	val = VIDOSDxB_BOTRIGHT_X(win_data->offset_x + win_data->width - 1) |
+		VIDOSDxB_BOTRIGHT_Y(win_data->offset_y + win_data->height - 1);
+	writel(val, regs + VIDOSD_B(win));
+
+	/* TODO: OSD alpha */
+
+	/* OSD size */
+	if (win != 3 && win != 4) {
+		u32 offset = VIDOSD_D(win);
+		if (win == 0)
+			offset = VIDOSD_C(win);
+		val = win_data->width * win_data->height;
+		writel(val, regs + offset);
+	}
+
+	/* FIXME: remove fixed values */
+	val = WINCONx_ENWIN;
+	val |= WINCON0_BPPMODE_24BPP_888;
+	val |= WINCONx_WSWP;
+	val |= WINCONx_BURSTLEN_16WORD;
+	writel(val, regs + WINCON(win));
+
+	/* TODO: colour key */
+
+	/* Enable DMA channel and unprotect windows */
+	val = readl(regs + SHADOWCON);
+	val |= SHADOWCON_CHx_ENABLE(win);
+	val &= ~SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+}
+
+static void fimd_win_disable(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	struct fimd_win_data *win_data;
+	int win = data->default_win;
+	u32 val;
+
+	if (win < 0 || win > WINDOWS_NR)
+		return;
+
+	win_data = &data->win_data[win];
+
+	/* protect windows */
+	val = readl(regs + SHADOWCON);
+	val |= SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+
+	/* wincon */
+	val = readl(regs + WINCON(win));
+	val &= ~WINCONx_ENWIN;
+	writel(val, regs + WINCON(win));
+
+	/* unprotect windows */
+	val = readl(regs + SHADOWCON);
+	val &= ~SHADOWCON_CHx_ENABLE(win);
+	val &= ~SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+}
+
+static struct samsung_drm_overlay_ops fimd_overlay_ops = {
+	.mode_set = fimd_win_mode_set,
+	.commit = fimd_win_commit,
+	.disable = fimd_win_disable,
+};
+
+/* for pageflip event */
+static void fimd_finish_pageflip(struct drm_device *drm_dev, int crtc)
+{
+	struct samsung_drm_private *dev_priv = drm_dev->dev_private;
+	struct drm_pending_vblank_event *e, *t;
+	struct timeval now;
+	unsigned long flags;
+
+	if (!dev_priv->pageflip_event)
+		return;
+
+	spin_lock_irqsave(&drm_dev->event_lock, flags);
+
+	list_for_each_entry_safe(e, t, &dev_priv->pageflip_event_list,
+			base.link) {
+		do_gettimeofday(&now);
+		e->event.sequence = 0;
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+
+		list_move_tail(&e->base.link, &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	}
+
+	drm_vblank_put(drm_dev, crtc);
+	dev_priv->pageflip_event = false;
+
+	spin_unlock_irqrestore(&drm_dev->event_lock, flags);
+}
+
+static irqreturn_t fimd_irq_handler(DRM_IRQ_ARGS)
+{
+	struct drm_device *drm_dev = (struct drm_device *)arg;
+	struct device *dev = fimd_subdrv.manager.dev;
+	struct samsung_drm_manager *manager = &fimd_subdrv.manager;
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	val = readl(regs + VIDINTCON1);
+
+	if (val & VIDINTCON1_INT_FRAME)
+		/* VSYNC interrupt */
+		writel(VIDINTCON1_INT_FRAME, regs + VIDINTCON1);
+
+	drm_handle_vblank(drm_dev, manager->pipe);
+	fimd_finish_pageflip(drm_dev, manager->pipe);
+
+	return IRQ_HANDLED;
+}
+
+static int fimd_subdrv_probe(struct drm_device *drm_dev)
+{
+	struct drm_driver *drm_driver = drm_dev->driver;
+	int ret;
+
+	drm_driver->irq_handler = fimd_irq_handler;
+
+	ret = drm_irq_install(drm_dev);
+	if (ret) {
+		drm_driver->irq_handler = NULL;
+		return ret;
+	}
+
+	/* FIXME */
+	drm_dev->vblank_disable_allowed = 1;
+
+	return 0;
+}
+
+static void fimd_subdrv_remove(struct drm_device *drm_dev)
+{
+	struct drm_driver *drm_driver = drm_dev->driver;
+
+	drm_irq_uninstall(drm_dev);
+
+	drm_driver->irq_handler = NULL;
+}
+
+static struct samsung_drm_subdrv fimd_subdrv = {
+	.probe = fimd_subdrv_probe,
+	.remove = fimd_subdrv_remove,
+	.manager = {
+		.pipe = -1,
+		.ops = &fimd_manager_ops,
+		.overlay_ops = &fimd_overlay_ops,
+		.display = &fimd_display,
+	},
+};
+
+static int fimd_calc_clkdiv(struct fimd_data *data,
+			    struct fb_videomode *timing)
+{
+	unsigned long clk = clk_get_rate(data->lcd_clk);
+	u32 retrace;
+	u32 clkdiv;
+	u32 best_framerate = 0;
+	u32 framerate;
+
+	retrace = timing->left_margin + timing->hsync_len +
+				timing->right_margin + timing->xres;
+	retrace *= timing->upper_margin + timing->vsync_len +
+				timing->lower_margin + timing->yres;
+
+	/* default framerate is 60Hz */
+	if (!timing->refresh)
+		timing->refresh = 60;
+
+	clk /= retrace;
+
+	for (clkdiv = 1; clkdiv < 0x100; clkdiv++) {
+		int tmp;
+
+		/* get best framerate */
+		framerate = clk / clkdiv;
+		tmp = timing->refresh - framerate;
+		if (tmp < 0) {
+			best_framerate = framerate;
+			continue;
+		} else {
+			if (!best_framerate)
+				best_framerate = framerate;
+			else if (tmp < (best_framerate - framerate))
+				best_framerate = framerate ;
+			break;
+		}
+	}
+
+	return clkdiv;
+}
+
+static void fimd_clear_win(struct fimd_data *data, int win)
+{
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	writel(0, regs + WINCON(win));
+	writel(0, regs + VIDOSD_A(win));
+	writel(0, regs + VIDOSD_B(win));
+	writel(0, regs + VIDOSD_C(win));
+
+	if (win == 1 || win == 2)
+		writel(0, regs + VIDOSD_D(win));
+
+	val = readl(regs + SHADOWCON);
+	val &= ~SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+}
+
+static int __devinit fimd_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct fimd_data *data;
+	struct samsung_drm_fimd_pdata *pdata;
+	struct resource *res;
+	int win;
+	int ret;
+
+	printk(KERN_DEBUG "[%d] %s\n", __LINE__, __func__);
+
+	pdata = pdev->dev.platform_data;
+	if (!pdata) {
+		dev_err(dev, "no platform data specified\n");
+		return -EINVAL;
+	}
+
+	data = kzalloc(sizeof(struct fimd_data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	data->bus_clk = clk_get(dev, "fimd");
+	if (IS_ERR(data->bus_clk)) {
+		dev_err(dev, "failed to get bus clock\n");
+		ret = PTR_ERR(data->bus_clk);
+		goto err_data;
+	}
+
+	clk_enable(data->bus_clk);
+
+	data->lcd_clk = clk_get(dev, "sclk_fimd");
+	if (IS_ERR(data->lcd_clk)) {
+		dev_err(dev, "failed to get lcd clock\n");
+		ret = PTR_ERR(data->lcd_clk);
+		goto err_bus_clk;
+	}
+
+	clk_enable(data->lcd_clk);
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res) {
+		dev_err(dev, "failed to find registers\n");
+		ret = -ENOENT;
+		goto err_clk;
+	}
+
+	data->regs_res = request_mem_region(res->start, resource_size(res),
+					   dev_name(dev));
+	if (!data->regs_res) {
+		dev_err(dev, "failed to claim register region\n");
+		ret = -ENOENT;
+		goto err_clk;
+	}
+
+	data->regs = ioremap(res->start, resource_size(res));
+	if (!data->regs) {
+		dev_err(dev, "failed to map registers\n");
+		ret = -ENXIO;
+		goto err_req_region;
+	}
+
+	for (win = 0; win < WINDOWS_NR; win++)
+		fimd_clear_win(data, win);
+
+	data->clkdiv = fimd_calc_clkdiv(data, &pdata->timing);
+	data->vidcon0 = pdata->vidcon0;
+	data->vidcon1 = pdata->vidcon1;
+	data->default_win = pdata->default_win;
+	data->bpp = pdata->bpp;
+
+	platform_set_drvdata(pdev, data);
+
+	memcpy(&fimd_timing, &pdata->timing,
+			sizeof(struct fb_videomode));
+	fimd_timing.pixclock = clk_get_rate(data->lcd_clk) / data->clkdiv;
+
+	fimd_subdrv.manager.dev = dev;
+
+	samsung_drm_subdrv_register(&fimd_subdrv);
+
+	return 0;
+
+err_req_region:
+	release_resource(data->regs_res);
+	kfree(data->regs_res);
+
+err_clk:
+	clk_disable(data->lcd_clk);
+	clk_put(data->lcd_clk);
+
+err_bus_clk:
+	clk_disable(data->bus_clk);
+	clk_put(data->bus_clk);
+
+err_data:
+	kfree(data);
+	return ret;
+}
+
+static int __devexit fimd_remove(struct platform_device *pdev)
+{
+	struct fimd_data *data = platform_get_drvdata(pdev);
+
+	samsung_drm_subdrv_unregister(&fimd_subdrv);
+
+	clk_disable(data->lcd_clk);
+	clk_disable(data->bus_clk);
+	clk_put(data->lcd_clk);
+	clk_put(data->bus_clk);
+
+	iounmap(data->regs);
+	release_resource(data->regs_res);
+	kfree(data->regs_res);
+
+	kfree(data);
+
+	return 0;
+}
+
+static struct platform_driver fimd_driver = {
+	.probe		= fimd_probe,
+	.remove		= __devexit_p(fimd_remove),
+	.driver		= {
+		.name	= "exynos4-fb",
+		.owner	= THIS_MODULE,
+	},
+};
+
+static int __init fimd_init(void)
+{
+	return platform_driver_register(&fimd_driver);
+}
+
+static void __exit fimd_exit(void)
+{
+	platform_driver_unregister(&fimd_driver);
+}
+
+module_init(fimd_init);
+module_exit(fimd_exit);
+
+MODULE_AUTHOR("Joonyoung Shim <jy0922.shim@samsung.com>");
+MODULE_DESCRIPTION("Samsung DRM FIMD Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c b/drivers/gpu/drm/samsung/samsung_drm_gem.c
new file mode 100644
index 0000000..1e167a6
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
@@ -0,0 +1,492 @@
+/* samsung_drm_gem.c
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "samsung_drm.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_gem.h"
+#include "samsung_drm_buf.h"
+
+static unsigned int convert_to_vm_err_msg(int msg)
+{
+	unsigned int out_msg;
+
+	switch (msg) {
+	case 0:
+	case -ERESTARTSYS:
+	case -EINTR:
+		out_msg = VM_FAULT_NOPAGE;
+		break;
+
+	case -ENOMEM:
+		out_msg = VM_FAULT_OOM;
+		break;
+
+	default:
+		out_msg = VM_FAULT_SIGBUS;
+		break;
+	}
+
+	return out_msg;
+}
+
+static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
+}
+
+/**
+ * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for an object
+ * @obj: obj in question
+ *
+ * GEM memory mapping works by handing back to userspace a fake mmap offset
+ * it can use in a subsequent mmap(2) call.  The DRM core code then looks
+ * up the object based on the offset and sets up the various memory mapping
+ * structures.
+ *
+ * This routine allocates and attaches a fake offset for @obj.
+ */
+static int
+samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)
+{
+	struct drm_device *dev = obj->dev;
+	struct drm_gem_mm *mm = dev->mm_private;
+	struct drm_map_list *list;
+	struct drm_local_map *map;
+	int ret = 0;
+
+	/* Set the object up for mmap'ing */
+	list = &obj->map_list;
+	list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
+	if (!list->map)
+		return -ENOMEM;
+
+	map = list->map;
+	map->type = _DRM_GEM;
+	map->size = obj->size;
+	map->handle = obj;
+
+	/* Get a DRM GEM mmap offset allocated... */
+	list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
+						    obj->size / PAGE_SIZE,
+						    0, 0);
+	if (!list->file_offset_node) {
+		DRM_ERROR("failed to allocate offset for bo %d\n",
+			  obj->name);
+		ret = -ENOSPC;
+		goto out_free_list;
+	}
+
+	list->file_offset_node = drm_mm_get_block(list->file_offset_node,
+						  obj->size / PAGE_SIZE,
+						  0);
+	if (!list->file_offset_node) {
+		ret = -ENOMEM;
+		goto out_free_list;
+	}
+
+	list->hash.key = list->file_offset_node->start;
+	ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
+	if (ret) {
+		DRM_ERROR("failed to add to map hash\n");
+		goto out_free_mm;
+	}
+
+	return 0;
+
+out_free_mm:
+	drm_mm_put_block(list->file_offset_node);
+out_free_list:
+	kfree(list->map);
+	list->map = NULL;
+
+	return ret;
+}
+
+static void
+samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
+{
+	struct drm_device *dev = obj->dev;
+	struct drm_gem_mm *mm = dev->mm_private;
+	struct drm_map_list *list = &obj->map_list;
+
+	drm_ht_remove_item(&mm->offset_hash, &list->hash);
+	drm_mm_put_block(list->file_offset_node);
+	kfree(list->map);
+	list->map = NULL;
+}
+
+static int samsung_drm_gem_create(struct drm_file *file, struct drm_device *dev,
+		unsigned int size, unsigned int *handle_p)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+	struct drm_gem_object *obj;
+	unsigned int handle;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	size = roundup(size, PAGE_SIZE);
+
+	/* allocate the new buffer object and memory region. */
+	samsung_gem_obj = samsung_drm_buf_create(dev, size);
+	if (!samsung_gem_obj)
+		return -ENOMEM;
+
+	obj = &samsung_gem_obj->base;
+
+	ret = drm_gem_object_init(dev, obj, size);
+	if (ret < 0) {
+		DRM_ERROR("failed to initailize gem object.\n");
+		goto out;
+	}
+
+	DRM_DEBUG_KMS("created file object = 0x%x\n", (unsigned int)obj->filp);
+
+	ret = samsung_drm_gem_create_mmap_offset(obj);
+	if (ret < 0) {
+		DRM_ERROR("failed to allocate mmap offset.\n");
+		goto out;
+	}
+
+	/**
+	 * allocate a id of idr table where the obj is registered
+	 * and handle has the id what user can see.
+	 */
+	ret = drm_gem_handle_create(file, obj, &handle);
+	if (ret) {
+		drm_gem_object_release(obj);
+		samsung_drm_buf_destroy(dev, samsung_gem_obj);
+		goto out;
+	}
+
+	DRM_DEBUG_KMS("gem handle = 0x%x\n", handle);
+
+	/* drop reference from allocate - handle holds it now. */
+	drm_gem_object_unreference_unlocked(obj);
+
+	*handle_p = handle;
+
+	return 0;
+
+out:
+	drm_gem_object_unreference_unlocked(obj);
+
+	samsung_drm_buf_destroy(dev, samsung_gem_obj);
+
+	kfree(samsung_gem_obj);
+
+	return ret;
+}
+
+int samsung_drm_gem_create_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file)
+{
+	struct drm_samsung_gem_create *args = data;
+
+	DRM_DEBUG_KMS("%s : size = 0x%x\n", __FILE__, args->size);
+
+	return samsung_drm_gem_create(file, dev, args->size, &args->handle);
+}
+
+int samsung_drm_gem_map_offset_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file)
+{
+	struct drm_samsung_gem_map_off *args = data;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	DRM_DEBUG_KMS("handle = 0x%x, offset = 0x%x\n",
+			args->handle, (u32)args->offset);
+
+	if (!(dev->driver->driver_features & DRIVER_GEM)) {
+		DRM_ERROR("not support GEM.\n");
+		return -ENODEV;
+	}
+
+	return samsung_drm_gem_dumb_map_offset(file, dev, args->handle,
+			&args->offset);
+}
+
+static int samsung_drm_gem_mmap_buffer(struct file *filp,
+		struct vm_area_struct *vma)
+{
+	struct drm_gem_object *obj = filp->private_data;
+	struct samsung_drm_gem_obj *samsung_gem_obj = to_samsung_gem_obj(obj);
+	struct samsung_drm_buf_entry *entry;
+	unsigned long pfn, size;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	vma->vm_flags |= (VM_IO | VM_RESERVED);
+	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+	vma->vm_file = filp;
+
+	size = vma->vm_end - vma->vm_start;
+	entry = samsung_gem_obj->entry;
+
+	/* check if the region to be mapped is valid or not. */
+	if ((entry->paddr + vma->vm_pgoff + size) >
+			(entry->paddr + entry->size)) {
+		drm_gem_object_unreference_unlocked(obj);
+		DRM_ERROR("desired size is bigger then real size.\n");
+		return -EINVAL;
+	}
+
+	pfn = (samsung_gem_obj->entry->paddr + vma->vm_pgoff) >> PAGE_SHIFT;
+
+	DRM_DEBUG_KMS("offset = 0x%x, pfn to be mapped = 0x%x\n",
+			(u32)vma->vm_pgoff, (u32)pfn);
+
+	if (remap_pfn_range(vma, vma->vm_start, pfn,
+				vma->vm_end - vma->vm_start,
+				vma->vm_page_prot)) {
+		DRM_ERROR("failed to remap pfn range.\n");
+		return -EAGAIN;
+	}
+
+	return 0;
+}
+
+static const struct file_operations samsung_drm_gem_fops = {
+	.mmap = samsung_drm_gem_mmap_buffer,
+};
+
+int samsung_drm_gem_mmap_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *filp)
+{
+	struct drm_samsung_gem_mmap *args = data;
+	struct drm_gem_object *obj;
+	unsigned int addr;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!(dev->driver->driver_features & DRIVER_GEM)) {
+		DRM_ERROR("not support GEM.\n");
+		return -ENODEV;
+	}
+
+	obj = drm_gem_object_lookup(dev, filp, args->handle);
+	if (!obj) {
+		DRM_ERROR("failed to lookup gem object.\n");
+		return -EINVAL;
+	}
+
+	obj->filp->f_op = &samsung_drm_gem_fops;
+	obj->filp->private_data = obj;
+
+	down_write(&current->mm->mmap_sem);
+	addr = do_mmap(obj->filp, 0, args->size,
+			PROT_READ | PROT_WRITE, MAP_SHARED, args->offset);
+	up_write(&current->mm->mmap_sem);
+
+	drm_gem_object_unreference_unlocked(obj);
+
+	if (IS_ERR((void *)addr))
+		return PTR_ERR((void *)addr);
+
+	args->mapped = addr;
+
+	DRM_DEBUG_KMS("mapped = 0x%x\n", args->mapped);
+
+	return 0;
+}
+
+int samsung_drm_gem_init_object(struct drm_gem_object *obj)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return 0;
+}
+
+void samsung_drm_gem_free_object(struct drm_gem_object *gem_obj)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	DRM_DEBUG_KMS("handle count = %d\n",
+			atomic_read(&gem_obj->handle_count));
+
+	if (gem_obj->map_list.map)
+		samsung_drm_gem_free_mmap_offset(gem_obj);
+
+	/* release file pointer to gem object. */
+	drm_gem_object_release(gem_obj);
+
+	samsung_gem_obj = to_samsung_gem_obj(gem_obj);
+
+	samsung_drm_buf_destroy(gem_obj->dev, samsung_gem_obj);
+}
+
+struct samsung_drm_gem_obj *
+		find_samsung_drm_gem_object(struct drm_file *file_priv,
+			struct drm_device *dev, unsigned int handle)
+{
+	struct drm_gem_object *gem_obj;
+
+	gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
+	if (!gem_obj) {
+		DRM_LOG_KMS("a invalid gem object not registered to lookup.\n");
+		return NULL;
+	}
+
+	/**
+	 * unreference refcount of the gem object.
+	 * at drm_gem_object_lookup(), the gem object was referenced.
+	 */
+	drm_gem_object_unreference(gem_obj);
+
+	return to_samsung_gem_obj(gem_obj);
+}
+
+int samsung_drm_gem_dumb_create(struct drm_file *file_priv,
+		struct drm_device *dev, struct drm_mode_create_dumb *args)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	/**
+	 * alocate memory to be used for framebuffer.
+	 * - this callback would be called by user application
+	 *	with DRM_IOCTL_MODE_CREATE_DUMB command.
+	 */
+
+	args->pitch = args->width * args->bpp >> 3;
+	args->size = args->pitch * args->height;
+
+	return samsung_drm_gem_create(file_priv, dev, args->size,
+			&args->handle);
+}
+
+int samsung_drm_gem_dumb_map_offset(struct drm_file *file_priv,
+		struct drm_device *dev, uint32_t handle, uint64_t *offset)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mutex_lock(&dev->struct_mutex);
+
+	/**
+	 * get offset of memory allocated for drm framebuffer.
+	 * - this callback would be called by user application
+	 *	with DRM_IOCTL_MODE_MAP_DUMB command.
+	 */
+
+	samsung_gem_obj = find_samsung_drm_gem_object(file_priv, dev, handle);
+	if (!samsung_gem_obj) {
+		DRM_ERROR("failed to get samsung_drm_get_obj.\n");
+		mutex_unlock(&dev->struct_mutex);
+		return -EINVAL;
+	}
+
+	*offset = get_gem_mmap_offset(&samsung_gem_obj->base);
+
+	DRM_DEBUG_KMS("offset = 0x%x\n", (unsigned int)*offset);
+
+	mutex_unlock(&dev->struct_mutex);
+
+	return 0;
+}
+
+int samsung_drm_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	struct drm_gem_object *obj = vma->vm_private_data;
+	struct samsung_drm_gem_obj *samsung_gem_obj = to_samsung_gem_obj(obj);
+	struct drm_device *dev = obj->dev;
+	unsigned long pfn;
+	pgoff_t page_offset;
+	int ret;
+
+	page_offset = ((unsigned long)vmf->virtual_address -
+			vma->vm_start) >> PAGE_SHIFT;
+
+	mutex_lock(&dev->struct_mutex);
+
+	pfn = (samsung_gem_obj->entry->paddr >> PAGE_SHIFT) + page_offset;
+
+	ret = vm_insert_mixed(vma, (unsigned long)vmf->virtual_address, pfn);
+
+	mutex_unlock(&dev->struct_mutex);
+
+	return convert_to_vm_err_msg(ret);
+}
+
+int samsung_drm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	/* set vm_area_struct. */
+	ret = drm_gem_mmap(filp, vma);
+	if (ret < 0) {
+		DRM_ERROR("failed to mmap.\n");
+		return ret;
+	}
+
+	vma->vm_flags &= ~VM_PFNMAP;
+	vma->vm_flags |= VM_MIXEDMAP;
+
+	return ret;
+}
+
+
+int samsung_drm_gem_dumb_destroy(struct drm_file *file_priv,
+		struct drm_device *dev, unsigned int handle)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_gem_obj = find_samsung_drm_gem_object(file_priv, dev, handle);
+	if (!samsung_gem_obj) {
+		DRM_ERROR("failed to get samsung_drm_get_obj.\n");
+		return -EINVAL;
+	}
+
+	/**
+	 * obj->refcount and obj->handle_count are decreased and
+	 * if both them are 0 then samsung_drm_gem_free_object()
+	 * would be called by callback to release resources.
+	 */
+	ret = drm_gem_handle_delete(file_priv, handle);
+	if (ret < 0) {
+		DRM_ERROR("failed to delete drm_gem_handle.\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM GEM Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.h b/drivers/gpu/drm/samsung/samsung_drm_gem.h
new file mode 100644
index 0000000..dbc0ab2
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_gem.h
@@ -0,0 +1,76 @@
+/* samsung_drm_gem.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authoer: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_GEM_H_
+#define _SAMSUNG_DRM_GEM_H_
+
+#define to_samsung_gem_obj(x)	container_of(x,\
+			struct samsung_drm_gem_obj, base)
+
+/* create a new mm object and get a handle to it. */
+int samsung_drm_gem_create_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file);
+
+/* get buffer offset to map to user space. */
+int samsung_drm_gem_map_offset_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file);
+
+/* unmap a buffer from user space. */
+int samsung_drm_gem_munmap_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file);
+
+/* initialize gem object. */
+int samsung_drm_gem_init_object(struct drm_gem_object *obj);
+
+/* free gem object. */
+void samsung_drm_gem_free_object(struct drm_gem_object *gem_obj);
+
+struct samsung_drm_gem_obj *
+		find_samsung_drm_gem_object(struct drm_file *file_priv,
+			struct drm_device *dev, unsigned int handle);
+
+/* create memory region for drm framebuffer. */
+int samsung_drm_gem_dumb_create(struct drm_file *file_priv,
+		struct drm_device *dev, struct drm_mode_create_dumb *args);
+
+/* map memory region for drm framebuffer to user space. */
+int samsung_drm_gem_dumb_map_offset(struct drm_file *file_priv,
+		struct drm_device *dev, uint32_t handle, uint64_t *offset);
+
+/* page fault handler and mmap fault address(virtual) to physical memory. */
+int samsung_drm_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
+
+/* mmap gem object. */
+int samsung_drm_gem_mmap_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *filp);
+
+/* set vm_flags and we can change vm attribute to other here. */
+int samsung_drm_gem_mmap(struct file *filp, struct vm_area_struct *vma);
+
+/* destroy memory region allocated. */
+int samsung_drm_gem_dumb_destroy(struct drm_file *file_priv,
+		struct drm_device *dev, unsigned int handle);
+
+#endif
diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
new file mode 100644
index 0000000..05dc460
--- /dev/null
+++ b/include/drm/samsung_drm.h
@@ -0,0 +1,267 @@
+/* samsung_drm.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_H_
+#define _SAMSUNG_DRM_H_
+
+#include "drm.h"
+
+#define MAX_CRTC	2
+
+struct drm_device;
+struct samsung_drm_overlay;
+struct drm_connector;
+
+enum samsung_drm_output_type {
+	SAMSUNG_DISPLAY_TYPE_NONE,
+	SAMSUNG_DISPLAY_TYPE_LCD,	/* RGB or CPU Interface. */
+	SAMSUNG_DISPLAY_TYPE_HDMI,	/* HDMI Interface. */
+};
+
+/**
+ * Samsung drm overlay ops structure.
+ *
+ * @mode_set: copy drm overlay info to hw specific overlay info.
+ * @commit: set hw specific overlay into to hw.
+ */
+struct samsung_drm_overlay_ops {
+	void (*mode_set)(struct device *subdrv_dev,
+			 struct samsung_drm_overlay *overlay);
+	void (*commit)(struct device *subdrv_dev);
+	void (*disable)(struct device *subdrv_dev);
+};
+
+/**
+ * Samsung drm common overlay structure.
+ *
+ * @win_num: window number.
+ * @offset_x: offset to x position.
+ * @offset_y: offset to y position.
+ * @pos_x: x position.
+ * @pos_y: y position.
+ * @width: window width.
+ * @height: window height.
+ * @bpp: bit per pixel.
+ * @paddr: physical memory address to this overlay.
+ * @vaddr: virtual memory addresss to this overlay.
+ * @buf_off: start offset of framebuffer to be displayed.
+ * @end_buf_off: end offset of framebuffer to be displayed.
+ * @buf_offsize: this value has result from
+ *			(framebuffer width - display width) * bpp.
+ * @line_size: line size to this overlay memory in bytes.
+ * @default_win: a window to be enabled.
+ * @color_key: color key on or off.
+ * @index_color: if using color key feature then this value would be used
+ *			as index color.
+ * @local_path: in case of lcd type, local path mode on or off.
+ * @transparency: transparency on or off.
+ * @activated: activated or not.
+ * @subdrv_dev: pointer to device object for subdrv device driver.
+ * @ops: pointer to samsung_drm_overlay_ops.
+ *
+ * this structure is common to Samsung SoC and would be copied
+ * to hardware specific overlay info.
+ */
+struct samsung_drm_overlay {
+	unsigned int offset_x;
+	unsigned int offset_y;
+	unsigned int width;
+	unsigned int height;
+	unsigned int paddr;
+	void __iomem *vaddr;
+	unsigned int buf_off;
+	unsigned int end_buf_off;
+	unsigned int buf_offsize;
+	unsigned int line_size;
+
+	bool default_win;
+	bool color_key;
+	unsigned int index_color;
+	bool local_path;
+	bool transparency;
+	bool activated;
+};
+
+/**
+ * Samsung DRM Display Structure.
+ *	- this structure is common to analog tv, digital tv and lcd panel.
+ *
+ * @dev: pointer to specific device object.
+ * @is_connected: check for that display is connected or not.
+ * @get_edid: get edid modes from display driver.
+ * @get_timing: get timing object from display driver.
+ * @check_timing: check if timing is valid or not.
+ * @power_on: display device on or off.
+ */
+struct samsung_drm_display {
+	unsigned int type;
+	bool (*is_connected)(void);
+	int (*get_edid)(struct drm_connector *connector, u8 *edid, int len);
+	void *(*get_timing)(void);
+	int (*check_timing)(void *timing);
+	int (*power_on)(int mode);
+};
+
+/**
+ * Samsung drm manager ops
+ *
+ * @mode_set: convert drm_display_mode to hw specific display mode and
+ *	      would be called by encoder->mode_set().
+ * @commit: set current hw specific display mode to hw.
+ * @enable_vblank: specific driver callback for enabling vblank interrupt.
+ * @disable_vblank: specific driver callback for disabling vblank interrupt.
+ */
+struct samsung_drm_manager_ops {
+	void (*mode_set)(struct device *subdrv_dev, void *mode);
+	void (*commit)(struct device *subdrv_dev);
+	int (*enable_vblank)(struct device *subdrv_dev);
+	void (*disable_vblank)(struct device *subdrv_dev);
+};
+
+/**
+ * Samsung drm common manager structure.
+ *
+ * @dev: pointer to device object for subdrv device driver.
+ * @ops: ops pointer to samsung drm common framebuffer.
+ *	 ops of fimd or hdmi driver should be set to this ones.
+ */
+struct samsung_drm_manager {
+	struct device *dev;
+	int pipe;
+	struct samsung_drm_manager_ops *ops;
+	struct samsung_drm_overlay_ops *overlay_ops;
+	struct samsung_drm_display *display;
+};
+
+/**
+ * Samsung drm private structure.
+ */
+struct samsung_drm_private {
+	struct drm_fb_helper *fb_helper;
+
+	/* for pageflip */
+	struct list_head pageflip_event_list;
+	bool pageflip_event;
+
+	struct drm_crtc *crtc[MAX_CRTC];
+
+	/* add some structures. */
+};
+
+struct samsung_drm_subdrv {
+	struct list_head list;
+
+	/* driver ops */
+	int (*probe)(struct drm_device *dev);
+	void (*remove)(struct drm_device *dev);
+
+	struct samsung_drm_manager manager;
+	struct drm_encoder *encoder;
+	struct drm_connector *connector;
+};
+
+int samsung_drm_device_register(struct drm_device *dev);
+void samsung_drm_device_unregister(struct drm_device *dev);
+int samsung_drm_subdrv_register(struct samsung_drm_subdrv *drm_subdrv);
+void samsung_drm_subdrv_unregister(struct samsung_drm_subdrv *drm_subdrv);
+
+/**
+ * User-desired buffer creation information structure.
+ *
+ * @usr_addr: an address allocated by user process and this address
+ *	would be mmapped to physical region by fault handler.
+ * @size: requested size for the object.
+ *	- this size value would be page-aligned internally.
+ * @flags: user request for setting memory type or cache attributes.
+ * @handle: returned handle for the object.
+ */
+struct drm_samsung_gem_create {
+	unsigned int usr_addr;
+	unsigned int size;
+	unsigned int flags;
+
+	unsigned int handle;
+};
+
+/**
+ * A structure for getting buffer offset.
+ *
+ * @handle: a pointer to gem object created.
+ * @offset: relatived offset value of the memory region allocated.
+ *	- this value should be set by user.
+ */
+struct drm_samsung_gem_map_off {
+	unsigned int handle;
+	uint64_t offset;
+};
+
+/**
+ * A structure for mapping buffer.
+ *
+ * @handle: a pointer to gem object created.
+ * @offset: relatived offset value of the memory region allocated.
+ *	- this value should be set by user.
+ * @size: memory size to be mapped.
+ * @mapped: user virtual address to be mapped.
+ */
+struct drm_samsung_gem_mmap {
+	unsigned int handle;
+	uint64_t offset;
+	unsigned int size;
+
+	unsigned int mapped;
+};
+
+#define DRM_SAMSUNG_GEM_CREATE		0x00
+#define DRM_SAMSUNG_GEM_MAP_OFFSET	0x01
+#define DRM_SAMSUNG_GEM_MMAP		0x02
+
+#define DRM_IOCTL_SAMSUNG_GEM_CREATE		DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_SAMSUNG_GEM_CREATE, struct drm_samsung_gem_create)
+
+#define DRM_IOCTL_SAMSUNG_GEM_MAP_OFFSET	DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_SAMSUNG_GEM_MAP_OFFSET, struct drm_samsung_gem_map_off)
+
+#define DRM_IOCTL_SAMSUNG_GEM_MMAP	DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_SAMSUNG_GEM_MMAP, struct drm_samsung_gem_mmap)
+
+/**
+ * Platform Specific Structure for DRM based FIMD.
+ *
+ * @timing: default video mode for initializing
+ * @default_win: default window layer number to be used for UI.
+ * @bpp: default bit per pixel.
+ */
+struct samsung_drm_fimd_pdata {
+	struct fb_videomode		timing;
+	u32				vidcon0;
+	u32				vidcon1;
+	unsigned int			default_win;
+	unsigned int			bpp;
+};
+
+#endif
-- 
1.7.0.4


^ permalink raw reply related	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-26 11:47 ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-26 11:47 UTC (permalink / raw)
  To: airlied, dri-devel
  Cc: sw0312.kim, linux-kernel, Inki Dae, kyungmin.park, linux-arm-kernel

This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
but we will add HDMI support also in the future.

this patch is based on git repository below:
git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
branch name: drm-next
commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922

you can refer to our working repository below:
http://git.infradead.org/users/kmpark/linux-2.6-samsung
branch name: samsung-drm

We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
based on Linux framebuffer) but couldn't so because lowlevel codes
of s3c-fb.c are included internally and so FIMD module of this driver has
its own lowlevel codes.

We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
for buffer allocation. by using DMA API, we could use CMA later.

Refer to this link for CMA(Continuous Memory Allocator):
http://lkml.org/lkml/2011/7/20/45

this driver supports only physically continuous memory(non-iommu).

Links to previous versions of the patchset:
v1: < https://lwn.net/Articles/454380/ >
v2: < http://www.spinics.net/lists/kernel/msg1224275.html >

Changelog v2:
DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.

     this feature maps user address space to physical memory region
     once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.

DRM: code clean and add exception codes.

Changelog v3:
DRM: Support multiple irq.

     FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
     this patch supports mutiple irq for Samsung SoC.

DRM: Consider modularization.

     each DRM, FIMD could be built as a module.

DRM: Have indenpendent crtc object.

     crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
     created crtc could be attached to any encoder object.

DRM: code clean and add exception codes.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
---
 drivers/gpu/drm/Kconfig                         |    2 +
 drivers/gpu/drm/Makefile                        |    1 +
 drivers/gpu/drm/samsung/Kconfig                 |   18 +
 drivers/gpu/drm/samsung/Makefile                |   11 +
 drivers/gpu/drm/samsung/samsung_drm_buf.c       |  140 ++++++
 drivers/gpu/drm/samsung/samsung_drm_buf.h       |   78 +++
 drivers/gpu/drm/samsung/samsung_drm_connector.c |  260 ++++++++++
 drivers/gpu/drm/samsung/samsung_drm_connector.h |   34 ++
 drivers/gpu/drm/samsung/samsung_drm_core.c      |  212 ++++++++
 drivers/gpu/drm/samsung/samsung_drm_crtc.c      |  334 +++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_crtc.h      |   36 ++
 drivers/gpu/drm/samsung/samsung_drm_drv.c       |  241 +++++++++
 drivers/gpu/drm/samsung/samsung_drm_encoder.c   |  250 +++++++++
 drivers/gpu/drm/samsung/samsung_drm_encoder.h   |   45 ++
 drivers/gpu/drm/samsung/samsung_drm_fb.c        |  248 +++++++++
 drivers/gpu/drm/samsung/samsung_drm_fb.h        |   47 ++
 drivers/gpu/drm/samsung/samsung_drm_fbdev.c     |  409 +++++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_fbdev.h     |   37 ++
 drivers/gpu/drm/samsung/samsung_drm_fimd.c      |  611 +++++++++++++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_gem.c       |  492 ++++++++++++++++++
 drivers/gpu/drm/samsung/samsung_drm_gem.h       |   76 +++
 include/drm/samsung_drm.h                       |  267 ++++++++++
 22 files changed, 3849 insertions(+), 0 deletions(-)
 create mode 100644 drivers/gpu/drm/samsung/Kconfig
 create mode 100644 drivers/gpu/drm/samsung/Makefile
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_buf.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_buf.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_connector.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_connector.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_core.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_crtc.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_crtc.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_drv.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_encoder.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_encoder.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fb.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fb.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fbdev.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fbdev.h
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_fimd.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_gem.c
 create mode 100644 drivers/gpu/drm/samsung/samsung_drm_gem.h
 create mode 100644 include/drm/samsung_drm.h

diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index b493663..ce6d3ec 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -158,3 +158,5 @@ config DRM_SAVAGE
 	help
 	  Choose this option if you have a Savage3D/4/SuperSavage/Pro/Twister
 	  chipset. If M is selected the module will be called savage.
+
+source "drivers/gpu/drm/samsung/Kconfig"
diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile
index 89cf05a..0c6e773 100644
--- a/drivers/gpu/drm/Makefile
+++ b/drivers/gpu/drm/Makefile
@@ -35,4 +35,5 @@ obj-$(CONFIG_DRM_SAVAGE)+= savage/
 obj-$(CONFIG_DRM_VMWGFX)+= vmwgfx/
 obj-$(CONFIG_DRM_VIA)	+=via/
 obj-$(CONFIG_DRM_NOUVEAU) +=nouveau/
+obj-$(CONFIG_DRM_SAMSUNG) +=samsung/
 obj-y			+= i2c/
diff --git a/drivers/gpu/drm/samsung/Kconfig b/drivers/gpu/drm/samsung/Kconfig
new file mode 100644
index 0000000..34cedda
--- /dev/null
+++ b/drivers/gpu/drm/samsung/Kconfig
@@ -0,0 +1,18 @@
+config DRM_SAMSUNG
+	tristate "DRM Support for Samsung SoC EXYNOS Series"
+	depends on DRM && PLAT_SAMSUNG
+	select DRM_KMS_HELPER
+	select FB_CFB_FILLRECT
+	select FB_CFB_COPYAREA
+	select FB_CFB_IMAGEBLIT
+	select VT_HW_CONSOLE_BINDING if FRAMEBUFFER_CONSOLE
+	help
+	  Choose this option if you have a Samsung SoC EXYNOS chipset.
+	  If M is selected the module will be called samsungdrm.
+
+config DRM_SAMSUNG_FIMD
+	tristate "Samsung DRM FIMD"
+	depends on DRM_SAMSUNG
+	help
+	  Choose this option if you want to use Samsung FIMD for DRM.
+	  If M is selected, the module will be called samsung_drm_fimd
diff --git a/drivers/gpu/drm/samsung/Makefile b/drivers/gpu/drm/samsung/Makefile
new file mode 100644
index 0000000..70f89f3
--- /dev/null
+++ b/drivers/gpu/drm/samsung/Makefile
@@ -0,0 +1,11 @@
+#
+# Makefile for the drm device driver.  This driver provides support for the
+# Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher.
+
+ccflags-y := -Iinclude/drm -Idrivers/gpu/drm/samsung
+samsungdrm-y := samsung_drm_drv.o samsung_drm_encoder.o samsung_drm_connector.o \
+		samsung_drm_crtc.o samsung_drm_fbdev.o samsung_drm_fb.o \
+		samsung_drm_buf.o samsung_drm_gem.o samsung_drm_core.o
+
+obj-$(CONFIG_DRM_SAMSUNG) += samsungdrm.o
+obj-$(CONFIG_DRM_SAMSUNG_FIMD) += samsung_drm_fimd.o
diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c b/drivers/gpu/drm/samsung/samsung_drm_buf.c
new file mode 100644
index 0000000..563d07e
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
@@ -0,0 +1,140 @@
+/* samsung_drm_buf.c
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_buf.h"
+
+static DEFINE_MUTEX(samsung_drm_buf_lock);
+
+static int lowlevel_buffer_allocate(struct drm_device *dev,
+		struct samsung_drm_buf_entry *entry)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
+			(dma_addr_t *)&entry->paddr, GFP_KERNEL);
+	if (!entry->paddr) {
+		DRM_ERROR("failed to allocate buffer.\n");
+		return -ENOMEM;
+	}
+
+	DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
+			(unsigned int)entry->vaddr, entry->paddr, entry->size);
+
+	return 0;
+}
+
+static void lowlevel_buffer_deallocate(struct drm_device *dev,
+		struct samsung_drm_buf_entry *entry)
+{
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	dma_free_writecombine(dev->dev, entry->size, entry->vaddr,
+			entry->paddr);
+
+	DRM_DEBUG_KMS("deallocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
+			(unsigned int)entry->vaddr, entry->paddr, entry->size);
+}
+
+static void  samsung_drm_buf_del(struct drm_device *dev,
+		struct samsung_drm_gem_obj *obj)
+{
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	lowlevel_buffer_deallocate(dev, obj->entry);
+
+	kfree(obj->entry);
+
+	kfree(obj);
+}
+
+struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
+		unsigned int size)
+{
+	struct samsung_drm_gem_obj *obj;
+	struct samsung_drm_buf_entry *entry;
+	int ret;
+
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+	DRM_DEBUG_KMS("desired size = 0x%x\n", size);
+
+	obj = kzalloc(sizeof(*obj), GFP_KERNEL);
+	if (!obj) {
+		DRM_ERROR("failed to allocate samsung_drm_gem_obj.\n");
+		return NULL;
+	}
+
+	/* use only one memory plane yet. */
+	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
+	if (!entry) {
+		DRM_ERROR("failed to allocate samsung_drm_buf_entry.\n");
+		return NULL;
+	}
+
+	entry->size = size;
+
+	/* allocate memory region and set it to vaddr and paddr. */
+	ret = lowlevel_buffer_allocate(dev, entry);
+	if (ret < 0)
+		return NULL;
+
+	obj->entry = entry;
+
+	return obj;
+}
+
+struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
+		unsigned int size)
+{
+	struct samsung_drm_gem_obj *obj;
+
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	obj = samsung_drm_buf_new(dev, size);
+	if (!obj)
+		return NULL;
+
+	DRM_DEBUG_KMS("buffer id : 0x%x\n", obj->id);
+
+	return obj;
+}
+
+int samsung_drm_buf_destroy(struct drm_device *dev,
+		struct samsung_drm_gem_obj *in_obj)
+{
+	DRM_DEBUG_KMS("%s.\n", __FILE__);
+
+	samsung_drm_buf_del(dev, in_obj);
+
+	return 0;
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Buffer Management Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
new file mode 100644
index 0000000..d6a7e95
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
@@ -0,0 +1,78 @@
+/* samsung_drm_buf.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_BUF_H_
+#define _SAMSUNG_DRM_BUF_H_
+
+/**
+ * samsung drm buffer entry structure.
+ *
+ * @paddr: physical address of allocated memory.
+ * @vaddr: kernel virtual address of allocated memory.
+ * @size: size of allocated memory.
+ */
+struct samsung_drm_buf_entry {
+	unsigned int paddr;
+	void __iomem *vaddr;
+	unsigned int size;
+};
+
+/**
+ * samsung drm buffer structure.
+ *
+ * @entry: pointer to samsung drm buffer entry object.
+ * @flags: it means memory type to be alloated or cache attributes.
+ * @handle: pointer to specific buffer object.
+ * @id: unique id to specific buffer object.
+ *
+ * ps. this object would be transfered to user as kms_bo.handle so
+ *	user can access to memory through kms_bo.handle.
+ */
+struct samsung_drm_gem_obj {
+	struct drm_gem_object base;
+	struct samsung_drm_buf_entry *entry;
+	unsigned int flags;
+
+	unsigned int handle;
+	unsigned int id;
+};
+
+/* create new buffer object and memory region and add the object to list. */
+struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
+		unsigned int size);
+
+/* allocate physical memory and add its object to list. */
+struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
+		unsigned int size);
+
+/* remove allocated physical memory. */
+int samsung_drm_buf_destroy(struct drm_device *dev,
+		struct samsung_drm_gem_obj *in_obj);
+
+/* find object added to list. */
+struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device *dev,
+		struct samsung_drm_gem_obj *in_obj, unsigned int paddr);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c b/drivers/gpu/drm/samsung/samsung_drm_connector.c
new file mode 100644
index 0000000..987a629
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
@@ -0,0 +1,260 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_encoder.h"
+
+#define MAX_EDID 256
+#define to_samsung_connector(x)	container_of(x, struct samsung_drm_connector,\
+				drm_connector)
+
+struct samsung_drm_connector {
+	struct drm_connector	drm_connector;
+	struct drm_encoder	*encoder;
+};
+
+/* convert samsung_video_timings to drm_display_mode */
+static inline void
+convert_to_display_mode(struct fb_videomode *timing,
+			struct drm_display_mode *mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode->clock = timing->pixclock / 1000;
+
+	mode->hdisplay = timing->xres;
+	mode->hsync_start = mode->hdisplay + timing->left_margin;
+	mode->hsync_end = mode->hsync_start + timing->hsync_len;
+	mode->htotal = mode->hsync_end + timing->right_margin;
+
+	mode->vdisplay = timing->yres;
+	mode->vsync_start = mode->vdisplay + timing->upper_margin;
+	mode->vsync_end = mode->vsync_start + timing->vsync_len;
+	mode->vtotal = mode->vsync_end + timing->lower_margin;
+}
+
+/* convert drm_display_mode to samsung_video_timings */
+static inline void
+convert_to_video_timing(struct drm_display_mode *mode,
+			struct fb_videomode *timing)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	timing->pixclock = mode->clock * 1000;
+
+	timing->xres = mode->hdisplay;
+	timing->left_margin = mode->hsync_start - mode->hdisplay;
+	timing->hsync_len = mode->hsync_end - mode->hsync_start;
+	timing->right_margin = mode->htotal - mode->hsync_end;
+
+	timing->yres = mode->vdisplay;
+	timing->upper_margin = mode->vsync_start - mode->vdisplay;
+	timing->vsync_len = mode->vsync_end - mode->vsync_start;
+	timing->lower_margin = mode->vtotal - mode->vsync_end;
+}
+
+static int samsung_drm_connector_get_modes(struct drm_connector *connector)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+	struct samsung_drm_display *display =
+		samsung_drm_get_manager(samsung_connector->encoder)->display;
+	unsigned int count;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (display && display->get_edid) {
+		void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
+		if (!edid) {
+			DRM_ERROR("failed to allocate edid\n");
+			return 0;
+		}
+
+		display->get_edid(connector, edid, MAX_EDID);
+
+		drm_mode_connector_update_edid_property(connector, edid);
+		count = drm_add_edid_modes(connector, edid);
+
+		kfree(connector->display_info.raw_edid);
+		connector->display_info.raw_edid = edid;
+	} else {
+		struct drm_display_mode *mode = drm_mode_create(connector->dev);
+		struct fb_videomode *timing;
+
+		if (display && display->get_timing)
+			timing = display->get_timing();
+		else
+			return 0;
+
+		convert_to_display_mode(timing, mode);
+
+		mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
+		drm_mode_set_name(mode);
+		drm_mode_probed_add(connector, mode);
+
+		count = 1;
+	}
+
+	return count;
+}
+
+static int samsung_drm_connector_mode_valid(struct drm_connector *connector,
+					    struct drm_display_mode *mode)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+	struct samsung_drm_display *display =
+		samsung_drm_get_manager(samsung_connector->encoder)->display;
+	struct fb_videomode timing;
+	int ret = MODE_BAD;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	convert_to_video_timing(mode, &timing);
+
+	if (display && display->check_timing)
+		if (!display->check_timing((void *)&timing))
+			ret = MODE_OK;
+
+	return ret;
+}
+
+struct drm_encoder *samsung_drm_best_encoder(struct drm_connector *connector)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	return to_samsung_connector(connector)->encoder;
+}
+
+static struct drm_connector_helper_funcs samsung_connector_helper_funcs = {
+	.get_modes	= samsung_drm_connector_get_modes,
+	.mode_valid	= samsung_drm_connector_mode_valid,
+	.best_encoder	= samsung_drm_best_encoder,
+};
+
+/* get detection status of display device. */
+static enum drm_connector_status
+samsung_drm_connector_detect(struct drm_connector *connector, bool force)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+	struct samsung_drm_display *display =
+		samsung_drm_get_manager(samsung_connector->encoder)->display;
+	unsigned int ret = connector_status_unknown;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (display && display->is_connected) {
+		if (display->is_connected())
+			ret = connector_status_connected;
+		else
+			ret = connector_status_disconnected;
+	}
+
+	return ret;
+}
+
+static void samsung_drm_connector_destroy(struct drm_connector *connector)
+{
+	struct samsung_drm_connector *samsung_connector =
+		to_samsung_connector(connector);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	drm_sysfs_connector_remove(connector);
+	drm_connector_cleanup(connector);
+	connector->dev->mode_config.num_connector--;
+	kfree(samsung_connector);
+}
+
+static struct drm_connector_funcs samsung_connector_funcs = {
+	.dpms		= drm_helper_connector_dpms,
+	.fill_modes	= drm_helper_probe_single_connector_modes,
+	.detect		= samsung_drm_connector_detect,
+	.destroy	= samsung_drm_connector_destroy,
+};
+
+struct drm_connector *samsung_drm_connector_create(struct drm_device *dev,
+						   struct drm_encoder *encoder)
+{
+	struct samsung_drm_connector *samsung_connector;
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+	struct drm_connector *connector;
+	int type;
+	int err;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_connector = kzalloc(sizeof(*samsung_connector), GFP_KERNEL);
+	if (!samsung_connector) {
+		DRM_ERROR("failed to allocate connector\n");
+		return NULL;
+	}
+
+	samsung_connector->encoder = encoder;
+	connector = &samsung_connector->drm_connector;
+
+	switch (manager->display->type) {
+	case SAMSUNG_DISPLAY_TYPE_HDMI:
+		type = DRM_MODE_CONNECTOR_HDMIA;
+		break;
+	default:
+		type = DRM_MODE_CONNECTOR_Unknown;
+		break;
+	}
+
+	drm_connector_init(dev, connector, &samsung_connector_funcs, type);
+	drm_connector_helper_add(connector, &samsung_connector_helper_funcs);
+
+	err = drm_sysfs_connector_add(connector);
+	if (err)
+		goto err_connector;
+
+	err = drm_mode_connector_attach_encoder(connector, encoder);
+	if (err) {
+		DRM_ERROR("failed to attach a connector to a encoder\n");
+		goto err_sysfs;
+	}
+
+	DRM_DEBUG_KMS("connector has been created\n");
+
+	return connector;
+
+err_sysfs:
+	drm_sysfs_connector_remove(connector);
+err_connector:
+	drm_connector_cleanup(connector);
+	kfree(samsung_connector);
+	return NULL;
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Connector Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.h b/drivers/gpu/drm/samsung/samsung_drm_connector.h
new file mode 100644
index 0000000..638d2b3
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_connector.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_CONNECTOR_H_
+#define _SAMSUNG_DRM_CONNECTOR_H_
+
+struct drm_connector *samsung_drm_connector_create(struct drm_device *dev,
+						   struct drm_encoder *encoder);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_core.c b/drivers/gpu/drm/samsung/samsung_drm_core.c
new file mode 100644
index 0000000..6221922
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_core.c
@@ -0,0 +1,212 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "samsung_drm_encoder.h"
+#include "samsung_drm_connector.h"
+#include "samsung_drm_fbdev.h"
+
+#include <drm/samsung_drm.h>
+
+static DEFINE_MUTEX(samsung_drm_mutex);
+static LIST_HEAD(samsung_drm_subdrv_list);
+static struct drm_device *drm_dev;
+
+static int samsung_drm_subdrv_probe(struct drm_device *dev,
+				    struct samsung_drm_subdrv *subdrv)
+{
+	struct drm_encoder *encoder;
+	struct drm_connector *connector;
+	int ret;
+
+	if (subdrv->probe) {
+		ret = subdrv->probe(dev);
+		if (ret)
+			return ret;
+	}
+
+	/* all crtc is available */
+	encoder = samsung_drm_encoder_create(dev, &subdrv->manager,
+			(1 << MAX_CRTC) - 1);
+	if (!encoder) {
+		DRM_ERROR("failed to create encoder\n");
+		return -EFAULT;
+	}
+
+	connector = samsung_drm_connector_create(dev, encoder);
+	if (!connector) {
+		DRM_ERROR("failed to create connector\n");
+		encoder->funcs->destroy(encoder);
+		return -EFAULT;
+	}
+
+	subdrv->encoder = encoder;
+	subdrv->connector = connector;
+
+	return 0;
+}
+
+static void samsung_drm_subdrv_remove(struct drm_device *dev,
+				      struct samsung_drm_subdrv *subdrv)
+{
+	if (subdrv->remove)
+		subdrv->remove(dev);
+
+	if (subdrv->encoder) {
+		struct drm_encoder *encoder = subdrv->encoder;
+		encoder->funcs->destroy(encoder);
+		subdrv->encoder = NULL;
+	}
+
+	if (subdrv->connector) {
+		struct drm_connector *connector = subdrv->connector;
+		connector->funcs->destroy(connector);
+		subdrv->connector = NULL;
+	}
+}
+
+int samsung_drm_device_register(struct drm_device *dev)
+{
+	struct samsung_drm_subdrv *subdrv, *n;
+	int err;
+
+	if (!dev)
+		return -EINVAL;
+
+	if (drm_dev) {
+		DRM_ERROR("Already drm device were registered\n");
+		return -EBUSY;
+	}
+
+	mutex_lock(&samsung_drm_mutex);
+	list_for_each_entry_safe(subdrv, n, &samsung_drm_subdrv_list, list) {
+		err = samsung_drm_subdrv_probe(dev, subdrv);
+		if (err) {
+			DRM_ERROR("samsung drm subdrv probe failed\n");
+			list_del(&subdrv->list);
+		}
+	}
+
+	drm_dev = dev;
+	mutex_unlock(&samsung_drm_mutex);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(samsung_drm_device_register);
+
+void samsung_drm_device_unregister(struct drm_device *dev)
+{
+	struct samsung_drm_subdrv *subdrv;
+
+	if (!dev || dev != drm_dev) {
+		WARN(1, "Unexpected drm device unregister!\n");
+		return;
+	}
+
+	mutex_lock(&samsung_drm_mutex);
+	list_for_each_entry(subdrv, &samsung_drm_subdrv_list, list)
+		samsung_drm_subdrv_remove(dev, subdrv);
+
+	drm_dev = NULL;
+	mutex_unlock(&samsung_drm_mutex);
+}
+EXPORT_SYMBOL_GPL(samsung_drm_device_unregister);
+
+static int samsung_drm_mode_group_reinit(struct drm_device *dev)
+{
+	struct drm_mode_group *group = &dev->primary->mode_group;
+	uint32_t *id_list = group->id_list;
+	int ret;
+
+	ret = drm_mode_group_init_legacy_group(dev, group);
+	if (ret < 0)
+		return ret;
+
+	kfree(id_list);
+	return 0;
+}
+
+int samsung_drm_subdrv_register(struct samsung_drm_subdrv *subdrv)
+{
+	int err;
+
+	if (!subdrv)
+		return -EINVAL;
+
+	mutex_lock(&samsung_drm_mutex);
+	if (drm_dev) {
+		err = samsung_drm_subdrv_probe(drm_dev, subdrv);
+		if (err) {
+			DRM_ERROR("failed to probe samsung drm subdrv\n");
+			mutex_unlock(&samsung_drm_mutex);
+			return err;
+		}
+
+		err = samsung_drm_fbdev_reinit(drm_dev);
+		if (err) {
+			DRM_ERROR("failed to reinitialize samsung drm fbdev\n");
+			samsung_drm_subdrv_remove(drm_dev, subdrv);
+			mutex_unlock(&samsung_drm_mutex);
+			return err;
+		}
+
+		err = samsung_drm_mode_group_reinit(drm_dev);
+		if (err) {
+			DRM_ERROR("failed to reinitialize mode group\n");
+			samsung_drm_fbdev_fini(drm_dev);
+			samsung_drm_subdrv_remove(drm_dev, subdrv);
+			mutex_unlock(&samsung_drm_mutex);
+			return err;
+		}
+	}
+
+	list_add_tail(&subdrv->list, &samsung_drm_subdrv_list);
+	mutex_unlock(&samsung_drm_mutex);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(samsung_drm_subdrv_register);
+
+void samsung_drm_subdrv_unregister(struct samsung_drm_subdrv *subdrv)
+{
+	if (!subdrv) {
+		WARN(1, "Unexpected samsung drm subdrv unregister!\n");
+		return;
+	}
+
+	mutex_lock(&samsung_drm_mutex);
+	if (drm_dev) {
+		samsung_drm_subdrv_remove(drm_dev, subdrv);
+
+		/* FIXME: error check */
+		samsung_drm_fbdev_reinit(drm_dev);
+		samsung_drm_mode_group_reinit(drm_dev);
+	}
+
+	list_del(&subdrv->list);
+	mutex_unlock(&samsung_drm_mutex);
+}
+EXPORT_SYMBOL_GPL(samsung_drm_subdrv_unregister);
diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
new file mode 100644
index 0000000..1c5a70a
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
@@ -0,0 +1,334 @@
+/* samsung_drm_crtc.c
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_fb.h"
+#include "samsung_drm_encoder.h"
+
+#define to_samsung_crtc(x)	container_of(x, struct samsung_drm_crtc,\
+				drm_crtc)
+
+/*
+ * @fb_x: horizontal position from framebuffer base
+ * @fb_y: vertical position from framebuffer base
+ * @base_x: horizontal position from screen base
+ * @base_y: vertical position from screen base
+ * @crtc_w: width of crtc
+ * @crtc_h: height of crtc
+ */
+struct samsung_drm_crtc_pos {
+	unsigned int fb_x;
+	unsigned int fb_y;
+	unsigned int base_x;
+	unsigned int base_y;
+	unsigned int crtc_w;
+	unsigned int crtc_h;
+};
+
+struct samsung_drm_crtc {
+	struct drm_crtc			drm_crtc;
+	struct samsung_drm_overlay	overlay;
+	unsigned int			pipe;
+};
+
+static void samsung_drm_overlay_update(struct samsung_drm_overlay *overlay,
+				       struct drm_framebuffer *fb,
+				       struct drm_display_mode *mode,
+				       struct samsung_drm_crtc_pos *pos)
+{
+	struct samsung_drm_buffer_info buffer_info;
+	unsigned int actual_w = pos->crtc_w;
+	unsigned int actual_h = pos->crtc_h;
+	unsigned int hw_w;
+	unsigned int hw_h;
+
+	/* update buffer address of framebuffer. */
+	samsung_drm_fb_update_buf_off(fb, pos->fb_x, pos->fb_y, &buffer_info);
+	overlay->paddr = buffer_info.paddr;
+	overlay->vaddr = buffer_info.vaddr;
+
+	hw_w = mode->hdisplay - pos->base_x;
+	hw_h = mode->vdisplay - pos->base_y;
+
+	if (actual_w > hw_w)
+		actual_w = hw_w;
+	if (actual_h > hw_h)
+		actual_h = hw_h;
+
+	overlay->offset_x = pos->base_x;
+	overlay->offset_y = pos->base_y;
+	overlay->width = actual_w;
+	overlay->height = actual_h;
+
+	DRM_DEBUG_KMS("overlay : offset_x/y(%d,%d), width/height(%d,%d)",
+			overlay->offset_x, overlay->offset_y,
+			overlay->width, overlay->height);
+
+	overlay->buf_offsize = fb->width - actual_w;
+	overlay->line_size = actual_w;
+	overlay->end_buf_off = fb->width * actual_h;
+}
+
+static int samsung_drm_crtc_update(struct drm_crtc *crtc)
+{
+	struct samsung_drm_crtc *samsung_crtc;
+	struct samsung_drm_overlay *overlay;
+	struct samsung_drm_crtc_pos pos;
+	struct drm_display_mode *mode = &crtc->mode;
+	struct drm_framebuffer *fb = crtc->fb;
+
+	if (!mode || !fb)
+		return -EINVAL;
+
+	samsung_crtc = to_samsung_crtc(crtc);
+	overlay = &samsung_crtc->overlay;
+
+	memset(&pos, 0, sizeof(struct samsung_drm_crtc_pos));
+	pos.fb_x = crtc->x;
+	pos.fb_y = crtc->y;
+	pos.crtc_w = fb->width - crtc->x;
+	pos.crtc_h = fb->height - crtc->y;
+
+	samsung_drm_overlay_update(overlay, crtc->fb, mode, &pos);
+
+	return 0;
+}
+
+/* CRTC helper functions */
+static void samsung_drm_crtc_dpms(struct drm_crtc *crtc, int mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* TODO */
+}
+
+static void samsung_drm_crtc_prepare(struct drm_crtc *crtc)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* drm framework doesn't check NULL. */
+}
+
+static void samsung_drm_crtc_commit(struct drm_crtc *crtc)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_drm_fn_encoder(crtc, NULL, samsung_drm_encoder_crtc_commit);
+}
+
+static bool
+samsung_drm_crtc_mode_fixup(struct drm_crtc *crtc,
+			    struct drm_display_mode *mode,
+			    struct drm_display_mode *adjusted_mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* drm framework doesn't check NULL */
+	return true;
+}
+
+static int
+samsung_drm_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *mode,
+			  struct drm_display_mode *adjusted_mode, int x, int y,
+			  struct drm_framebuffer *old_fb)
+{
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_overlay *overlay = &samsung_crtc->overlay;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode = adjusted_mode;
+
+	ret = samsung_drm_crtc_update(crtc);
+	if (ret)
+		return ret;
+
+	samsung_drm_fn_encoder(crtc, overlay,
+			samsung_drm_encoder_crtc_mode_set);
+
+	return ret;
+}
+
+static int samsung_drm_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
+					  struct drm_framebuffer *old_fb)
+{
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_overlay *overlay = &samsung_crtc->overlay;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	ret = samsung_drm_crtc_update(crtc);
+	if (ret)
+		return ret;
+
+	samsung_drm_fn_encoder(crtc, overlay,
+			samsung_drm_encoder_crtc_mode_set);
+	samsung_drm_fn_encoder(crtc, NULL, samsung_drm_encoder_crtc_commit);
+
+	return ret;
+}
+
+static void samsung_drm_crtc_load_lut(struct drm_crtc *crtc)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+	/* drm framework doesn't check NULL */
+}
+
+static struct drm_crtc_helper_funcs samsung_crtc_helper_funcs = {
+	.dpms		= samsung_drm_crtc_dpms,
+	.prepare	= samsung_drm_crtc_prepare,
+	.commit		= samsung_drm_crtc_commit,
+	.mode_fixup	= samsung_drm_crtc_mode_fixup,
+	.mode_set	= samsung_drm_crtc_mode_set,
+	.mode_set_base	= samsung_drm_crtc_mode_set_base,
+	.load_lut	= samsung_drm_crtc_load_lut,
+};
+
+/* CRTC functions */
+static int samsung_drm_crtc_page_flip(struct drm_crtc *crtc,
+				      struct drm_framebuffer *fb,
+				      struct drm_pending_vblank_event *event)
+{
+	struct drm_device *dev = crtc->dev;
+	struct samsung_drm_private *dev_priv = dev->dev_private;
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_overlay *overlay = &samsung_crtc->overlay;
+	struct drm_framebuffer *old_fb = crtc->fb;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mutex_lock(&dev->struct_mutex);
+
+	if (event && !dev_priv->pageflip_event) {
+		list_add_tail(&event->base.link,
+				&dev_priv->pageflip_event_list);
+
+		ret = drm_vblank_get(dev, samsung_crtc->pipe);
+		if (ret) {
+			DRM_DEBUG("failed to acquire vblank counter\n");
+			goto out;
+		}
+
+		dev_priv->pageflip_event = true;
+	}
+
+	crtc->fb = fb;
+
+	ret = samsung_drm_crtc_update(crtc);
+	if (ret) {
+		crtc->fb = old_fb;
+		if (event && dev_priv->pageflip_event) {
+			drm_vblank_put(dev, samsung_crtc->pipe);
+			dev_priv->pageflip_event = false;
+		}
+
+		goto out;
+	}
+
+	samsung_drm_fn_encoder(crtc, overlay,
+			samsung_drm_encoder_crtc_mode_set);
+	samsung_drm_fn_encoder(crtc, NULL, samsung_drm_encoder_crtc_commit);
+
+out:
+	mutex_unlock(&dev->struct_mutex);
+	return ret;
+}
+
+static void samsung_drm_crtc_destroy(struct drm_crtc *crtc)
+{
+	struct samsung_drm_crtc *samsung_crtc = to_samsung_crtc(crtc);
+	struct samsung_drm_private *private = crtc->dev->dev_private;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	private->crtc[samsung_crtc->pipe] = NULL;
+
+	drm_crtc_cleanup(crtc);
+	kfree(samsung_crtc);
+}
+
+static struct drm_crtc_funcs samsung_crtc_funcs = {
+	.set_config	= drm_crtc_helper_set_config,
+	.page_flip	= samsung_drm_crtc_page_flip,
+	.destroy	= samsung_drm_crtc_destroy,
+};
+
+int samsung_drm_crtc_create(struct drm_device *dev, unsigned int nr)
+{
+	struct samsung_drm_crtc *samsung_crtc;
+	struct samsung_drm_private *private = dev->dev_private;
+	struct drm_crtc *crtc;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_crtc = kzalloc(sizeof(*samsung_crtc), GFP_KERNEL);
+	if (!samsung_crtc) {
+		DRM_ERROR("failed to allocate samsung crtc\n");
+		return -ENOMEM;
+	}
+
+	samsung_crtc->pipe = nr;
+	crtc = &samsung_crtc->drm_crtc;
+
+	private->crtc[nr] = crtc;
+
+	drm_crtc_init(dev, crtc, &samsung_crtc_funcs);
+	drm_crtc_helper_add(crtc, &samsung_crtc_helper_funcs);
+
+	return 0;
+}
+
+int samsung_drm_crtc_enable_vblank(struct drm_device *dev, int crtc)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_drm_fn_encoder(private->crtc[crtc], &crtc,
+			samsung_drm_enable_vblank);
+
+	return 0;
+}
+
+void samsung_drm_crtc_disable_vblank(struct drm_device *dev, int crtc)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_drm_fn_encoder(private->crtc[crtc], &crtc,
+			samsung_drm_disable_vblank);
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM CRTC Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.h b/drivers/gpu/drm/samsung/samsung_drm_crtc.h
new file mode 100644
index 0000000..9256c7e
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.h
@@ -0,0 +1,36 @@
+/* samsung_drm_crtc.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_CRTC_H_
+#define _SAMSUNG_DRM_CRTC_H_
+
+int samsung_drm_crtc_create(struct drm_device *dev, unsigned int nr);
+int samsung_drm_crtc_enable_vblank(struct drm_device *dev, int crtc);
+void samsung_drm_crtc_disable_vblank(struct drm_device *dev, int crtc);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_drv.c b/drivers/gpu/drm/samsung/samsung_drm_drv.c
new file mode 100644
index 0000000..eefb08b
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_drv.c
@@ -0,0 +1,241 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_crtc.h"
+#include "samsung_drm_fbdev.h"
+#include "samsung_drm_fb.h"
+#include "samsung_drm_gem.h"
+
+#define DRIVER_NAME	"samsung-drm"
+#define DRIVER_DESC	"Samsung SoC DRM"
+#define DRIVER_DATE	"20110530"
+#define DRIVER_MAJOR	1
+#define DRIVER_MINOR	0
+
+static int samsung_drm_load(struct drm_device *dev, unsigned long flags)
+{
+	struct samsung_drm_private *private;
+	int ret;
+	int nr;
+
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	private = kzalloc(sizeof(struct samsung_drm_private), GFP_KERNEL);
+	if (!private) {
+		DRM_ERROR("failed to allocate private\n");
+		return -ENOMEM;
+	}
+
+	INIT_LIST_HEAD(&private->pageflip_event_list);
+	dev->dev_private = (void *)private;
+
+	drm_mode_config_init(dev);
+
+	samsung_drm_mode_config_init(dev);
+
+	for (nr = 0; nr < MAX_CRTC; nr++) {
+		ret = samsung_drm_crtc_create(dev, nr);
+		if (ret)
+			goto err_crtc;
+	}
+
+	ret = drm_vblank_init(dev, MAX_CRTC);
+	if (ret)
+		goto err_crtc;
+
+	ret = samsung_drm_device_register(dev);
+	if (ret)
+		goto err_vblank;
+
+	ret = samsung_drm_fbdev_init(dev);
+	if (ret) {
+		DRM_ERROR("failed to initialize drm fbdev\n");
+		goto err_drm_device;
+	}
+
+	return 0;
+
+err_drm_device:
+	samsung_drm_device_unregister(dev);
+err_vblank:
+	drm_vblank_cleanup(dev);
+err_crtc:
+	drm_mode_config_cleanup(dev);
+	kfree(private);
+
+	return ret;
+}
+
+static int samsung_drm_unload(struct drm_device *dev)
+{
+	samsung_drm_fbdev_fini(dev);
+	samsung_drm_device_unregister(dev);
+	drm_vblank_cleanup(dev);
+	drm_mode_config_cleanup(dev);
+	kfree(dev->dev_private);
+
+	return 0;
+}
+
+static int samsung_drm_open(struct drm_device *dev, struct drm_file *file_priv)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	return 0;
+}
+
+static void samsung_drm_lastclose(struct drm_device *dev)
+{
+	samsung_drm_fbdev_restore_mode(dev);
+}
+
+static int samsung_drm_master_create(struct drm_device *dev,
+		struct drm_master *master)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	master->driver_priv = NULL;
+
+	return 0;
+}
+
+static int samsung_drm_master_set(struct drm_device *dev,
+				  struct drm_file *file_priv, bool from_open)
+{
+	struct drm_master *master = file_priv->master;
+
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	master->lock.hw_lock = kzalloc(sizeof(struct drm_hw_lock), GFP_KERNEL);
+	if (!master->lock.hw_lock) {
+		DRM_DEBUG("failed to allocate hw_lock\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static struct vm_operations_struct samsung_drm_gem_vm_ops = {
+	.fault = samsung_drm_gem_fault,
+	.open = drm_gem_vm_open,
+	.close = drm_gem_vm_close,
+};
+
+static struct drm_ioctl_desc samsung_ioctls[] = {
+	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
+			DRM_UNLOCKED),
+	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
+			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
+	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
+			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
+};
+
+static struct drm_driver samsung_drm_driver = {
+	.driver_features	= DRIVER_HAVE_IRQ | DRIVER_BUS_PLATFORM |
+				  DRIVER_MODESET | DRIVER_GEM,
+	.load			= samsung_drm_load,
+	.unload			= samsung_drm_unload,
+	.open			= samsung_drm_open,
+	.lastclose		= samsung_drm_lastclose,
+	.get_vblank_counter	= drm_vblank_count,
+	.enable_vblank		= samsung_drm_crtc_enable_vblank,
+	.disable_vblank		= samsung_drm_crtc_disable_vblank,
+	.master_create		= samsung_drm_master_create,
+	.master_set		= samsung_drm_master_set,
+	.gem_init_object	= samsung_drm_gem_init_object,
+	.gem_free_object	= samsung_drm_gem_free_object,
+	.gem_vm_ops		= &samsung_drm_gem_vm_ops,
+	.dumb_create		= samsung_drm_gem_dumb_create,
+	.dumb_map_offset	= samsung_drm_gem_dumb_map_offset,
+	.dumb_destroy		= samsung_drm_gem_dumb_destroy,
+	.ioctls			= samsung_ioctls,
+	.fops = {
+		.owner		= THIS_MODULE,
+		.open		= drm_open,
+		.mmap		= samsung_drm_gem_mmap,
+		.poll		= drm_poll,
+		.read		= drm_read,
+		.unlocked_ioctl	= drm_ioctl,
+		.release	= drm_release,
+	},
+	.name	= DRIVER_NAME,
+	.desc	= DRIVER_DESC,
+	.date	= DRIVER_DATE,
+	.major	= DRIVER_MAJOR,
+	.minor	= DRIVER_MINOR,
+};
+
+static int samsung_drm_platform_probe(struct platform_device *pdev)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	samsung_drm_driver.num_ioctls = DRM_ARRAY_SIZE(samsung_ioctls);
+
+	return drm_platform_init(&samsung_drm_driver, pdev);
+}
+
+static int samsung_drm_platform_remove(struct platform_device *pdev)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	drm_platform_exit(&samsung_drm_driver, pdev);
+
+	return 0;
+}
+
+static struct platform_driver samsung_drm_platform_driver = {
+	.probe		= samsung_drm_platform_probe,
+	.remove		= __devexit_p(samsung_drm_platform_remove),
+	.driver		= {
+		.owner	= THIS_MODULE,
+		.name	= DRIVER_NAME,
+	},
+};
+
+static int __init samsung_drm_init(void)
+{
+	DRM_DEBUG_DRIVER("%s\n", __FILE__);
+
+	return platform_driver_register(&samsung_drm_platform_driver);
+}
+
+static void __exit samsung_drm_exit(void)
+{
+	platform_driver_unregister(&samsung_drm_platform_driver);
+}
+
+module_init(samsung_drm_init);
+module_exit(samsung_drm_exit);
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_encoder.c b/drivers/gpu/drm/samsung/samsung_drm_encoder.c
new file mode 100644
index 0000000..d1cb04e
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_encoder.c
@@ -0,0 +1,250 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc_helper.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_encoder.h"
+
+#define to_samsung_encoder(x)	container_of(x, struct samsung_drm_encoder,\
+				drm_encoder)
+
+struct samsung_drm_encoder {
+	struct drm_encoder		drm_encoder;
+	struct samsung_drm_manager	*manager;
+};
+
+static void samsung_drm_encoder_dpms(struct drm_encoder *encoder, int mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_connector *connector;
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+
+	DRM_DEBUG_KMS("%s, encoder dpms: %d\n", __FILE__, mode);
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder) {
+			struct samsung_drm_display *display = manager->display;
+
+			if (display && display->power_on)
+				display->power_on(mode);
+		}
+	}
+}
+
+static bool
+samsung_drm_encoder_mode_fixup(struct drm_encoder *encoder,
+			       struct drm_display_mode *mode,
+			       struct drm_display_mode *adjusted_mode)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return true;
+}
+
+static void samsung_drm_encoder_mode_set(struct drm_encoder *encoder,
+					 struct drm_display_mode *mode,
+					 struct drm_display_mode *adjusted_mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct drm_connector *connector;
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode = adjusted_mode;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder)
+			if (manager_ops && manager_ops->mode_set)
+				manager_ops->mode_set(manager->dev, mode);
+	}
+}
+
+static void samsung_drm_encoder_prepare(struct drm_encoder *encoder)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+}
+
+static void samsung_drm_encoder_commit(struct drm_encoder *encoder)
+{
+	struct samsung_drm_manager *manager = samsung_drm_get_manager(encoder);
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (manager_ops && manager_ops->commit)
+		manager_ops->commit(manager->dev);
+}
+
+static struct drm_crtc *
+samsung_drm_encoder_get_crtc(struct drm_encoder *encoder)
+{
+	return encoder->crtc;
+}
+
+static struct drm_encoder_helper_funcs samsung_encoder_helper_funcs = {
+	.dpms		= samsung_drm_encoder_dpms,
+	.mode_fixup	= samsung_drm_encoder_mode_fixup,
+	.mode_set	= samsung_drm_encoder_mode_set,
+	.prepare	= samsung_drm_encoder_prepare,
+	.commit		= samsung_drm_encoder_commit,
+	.get_crtc	= samsung_drm_encoder_get_crtc,
+};
+
+static void samsung_drm_encoder_destroy(struct drm_encoder *encoder)
+{
+	struct samsung_drm_encoder *samsung_encoder =
+		to_samsung_encoder(encoder);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_encoder->manager->pipe = -1;
+
+	drm_encoder_cleanup(encoder);
+	encoder->dev->mode_config.num_encoder--;
+	kfree(samsung_encoder);
+}
+
+static struct drm_encoder_funcs samsung_encoder_funcs = {
+	.destroy = samsung_drm_encoder_destroy,
+};
+
+struct drm_encoder *
+samsung_drm_encoder_create(struct drm_device *dev,
+			   struct samsung_drm_manager *manager,
+			   unsigned int possible_crtcs)
+{
+	struct drm_encoder *encoder;
+	struct samsung_drm_encoder *samsung_encoder;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!manager || !possible_crtcs)
+		return NULL;
+
+	if (!manager->dev)
+		return NULL;
+
+	samsung_encoder = kzalloc(sizeof(*samsung_encoder), GFP_KERNEL);
+	if (!samsung_encoder) {
+		DRM_ERROR("failed to allocate encoder\n");
+		return NULL;
+	}
+
+	samsung_encoder->manager = manager;
+	encoder = &samsung_encoder->drm_encoder;
+	encoder->possible_crtcs = possible_crtcs;
+
+	DRM_DEBUG_KMS("possible_crtcs = 0x%x\n", encoder->possible_crtcs);
+
+	drm_encoder_init(dev, encoder, &samsung_encoder_funcs,
+			DRM_MODE_ENCODER_TMDS);
+
+	drm_encoder_helper_add(encoder, &samsung_encoder_helper_funcs);
+
+	DRM_DEBUG_KMS("encoder has been created\n");
+
+	return encoder;
+}
+
+struct samsung_drm_manager *samsung_drm_get_manager(struct drm_encoder *encoder)
+{
+	return to_samsung_encoder(encoder)->manager;
+}
+
+void samsung_drm_fn_encoder(struct drm_crtc *crtc, void *data,
+			    void (*fn)(struct drm_encoder *, void *))
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_encoder *encoder;
+
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (encoder->crtc != crtc)
+			continue;
+
+		fn(encoder, data);
+	}
+}
+
+void samsung_drm_enable_vblank(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+	int crtc = *(int *)data;
+
+	if (manager->pipe == -1)
+		manager->pipe = crtc;
+
+	/* old_crtc checking needs? FIXME!!! */
+
+	if (manager_ops->enable_vblank)
+		manager_ops->enable_vblank(manager->dev);
+}
+
+void samsung_drm_disable_vblank(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_manager_ops *manager_ops = manager->ops;
+	int crtc = *(int *)data;
+
+	if (manager->pipe == -1)
+		manager->pipe = crtc;
+
+	/* old_crtc checking needs? FIXME!!! */
+
+	if (manager_ops->disable_vblank)
+		manager_ops->disable_vblank(manager->dev);
+}
+
+void samsung_drm_encoder_crtc_commit(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_overlay_ops *overlay_ops = manager->overlay_ops;
+
+	overlay_ops->commit(manager->dev);
+}
+
+void samsung_drm_encoder_crtc_mode_set(struct drm_encoder *encoder, void *data)
+{
+	struct samsung_drm_manager *manager =
+		to_samsung_encoder(encoder)->manager;
+	struct samsung_drm_overlay_ops *overlay_ops = manager->overlay_ops;
+	struct samsung_drm_overlay *overlay = data;
+
+	overlay_ops->mode_set(manager->dev, overlay);
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM Encoder Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_encoder.h b/drivers/gpu/drm/samsung/samsung_drm_encoder.h
new file mode 100644
index 0000000..99040b2
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_encoder.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_ENCODER_H_
+#define _SAMSUNG_DRM_ENCODER_H_
+
+struct samsung_drm_manager;
+
+struct drm_encoder *samsung_drm_encoder_create(struct drm_device *dev,
+					       struct samsung_drm_manager *mgr,
+					       unsigned int possible_crtcs);
+struct samsung_drm_manager *
+samsung_drm_get_manager(struct drm_encoder *encoder);
+void samsung_drm_fn_encoder(struct drm_crtc *crtc, void *data,
+			    void (*fn)(struct drm_encoder *, void *));
+void samsung_drm_enable_vblank(struct drm_encoder *encoder, void *data);
+void samsung_drm_disable_vblank(struct drm_encoder *encoder, void *data);
+void samsung_drm_encoder_crtc_commit(struct drm_encoder *encoder, void *data);
+void samsung_drm_encoder_crtc_mode_set(struct drm_encoder *encoder, void *data);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c b/drivers/gpu/drm/samsung/samsung_drm_fb.c
new file mode 100644
index 0000000..42096eb
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
@@ -0,0 +1,248 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc.h"
+#include "drm_crtc_helper.h"
+
+#include "samsung_drm_fb.h"
+#include "samsung_drm_buf.h"
+#include "samsung_drm_gem.h"
+
+#define to_samsung_fb(x)	container_of(x, struct samsung_drm_fb, fb)
+
+struct samsung_drm_fb {
+	struct drm_framebuffer		fb;
+	struct drm_file			*file_priv;
+	struct samsung_drm_gem_obj	*samsung_gem_obj;
+	unsigned int			is_default;
+
+	/* samsung gem object handle. */
+	unsigned int			gem_handle;
+	/* unique id to buffer object. */
+	unsigned int			id;
+
+	unsigned int			fb_size;
+	unsigned long			paddr;
+	void __iomem			*vaddr;
+};
+
+static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
+{
+	struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	drm_framebuffer_cleanup(fb);
+
+	if (samsung_fb->is_default) {
+		ret = drm_gem_handle_delete(samsung_fb->file_priv,
+				samsung_fb->gem_handle);
+		if (ret < 0)
+			DRM_ERROR("failed to delete drm_gem_handle.\n");
+	}
+
+	kfree(samsung_fb);
+}
+
+static int samsung_drm_fb_create_handle(struct drm_framebuffer *fb,
+					struct drm_file *file_priv,
+					unsigned int *handle)
+{
+	struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!samsung_fb->gem_handle) {
+		DRM_ERROR("can't get id to buffer object.\n");
+		return -EINVAL;
+	}
+
+	*handle = samsung_fb->gem_handle;
+
+	DRM_DEBUG_KMS("got buffer object id(%d)\n", *handle);
+
+	return 0;
+}
+
+static int samsung_drm_fb_dirty(struct drm_framebuffer *fb,
+				struct drm_file *file_priv, unsigned flags,
+				unsigned color, struct drm_clip_rect *clips,
+				unsigned num_clips)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	/*
+	 * update framebuffer and its hardware.
+	 * - this callback would be called by user application
+	 *	with DRM_IOCTL_MODE_DIRTYFB command.
+	 *
+	 * ps. Userspace can notify the driver via this callback
+	 * that an area of the framebuffer has been changed then should
+	 * be flushed to the display hardware.
+	 */
+
+	return 0;
+}
+
+static struct drm_framebuffer_funcs samsung_drm_fb_funcs = {
+	.destroy	= samsung_drm_fb_destroy,
+	.create_handle	= samsung_drm_fb_create_handle,
+	.dirty		= samsung_drm_fb_dirty,
+};
+
+static struct drm_framebuffer *
+samsung_drm_fb_init(struct drm_file *file_priv, struct drm_device *dev,
+		    struct drm_mode_fb_cmd *mode_cmd)
+{
+	struct samsung_drm_fb *samsung_fb;
+	struct drm_framebuffer *fb;
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+	unsigned int size, gem_handle = 0;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mode_cmd->pitch = max(mode_cmd->pitch,
+			mode_cmd->width * (mode_cmd->bpp >> 3));
+
+	DRM_LOG_KMS("drm fb create(%dx%d)\n",
+			mode_cmd->width, mode_cmd->height);
+
+	samsung_fb = kzalloc(sizeof(*samsung_fb), GFP_KERNEL);
+	if (!samsung_fb) {
+		DRM_ERROR("failed to allocate samsung drm framebuffer.\n");
+		return ERR_PTR(-ENOMEM);
+	}
+
+	fb = &samsung_fb->fb;
+	ret = drm_framebuffer_init(dev, fb, &samsung_drm_fb_funcs);
+	if (ret) {
+		DRM_ERROR("failed to initialize framebuffer.\n");
+		goto fail;
+	}
+
+	DRM_LOG_KMS("create: fb id: %d\n", fb->base.id);
+
+	size = mode_cmd->pitch * mode_cmd->height;
+
+	/*
+	 * mode_cmd->handle could be pointer to a buffer allocated by user
+	 * application using KMS library.
+	 */
+	if (!mode_cmd->handle) {
+		samsung_gem_obj = samsung_drm_buf_create(dev, size);
+		if (!samsung_gem_obj)
+			return ERR_PTR(-ENOMEM);
+	} else {
+		samsung_gem_obj = find_samsung_drm_gem_object(file_priv, dev,
+				mode_cmd->handle);
+		if (!samsung_gem_obj)
+			return ERR_PTR(-EINVAL);
+
+		gem_handle = mode_cmd->handle;
+	}
+
+	/* first is default framebuffer */
+	if (dev->mode_config.num_fb == 1)
+		samsung_fb->is_default = 1;
+
+	samsung_fb->file_priv = file_priv;
+	samsung_fb->samsung_gem_obj = samsung_gem_obj;
+	samsung_fb->gem_handle = gem_handle;
+	samsung_fb->id = samsung_gem_obj->id;
+	samsung_fb->fb_size = size;
+	samsung_fb->vaddr = samsung_gem_obj->entry->vaddr;
+	samsung_fb->paddr = samsung_gem_obj->entry->paddr;
+
+	DRM_DEBUG_KMS("handle = 0x%x, id = %d\n",
+			samsung_gem_obj->handle, samsung_gem_obj->id);
+	DRM_DEBUG_KMS("fb: size = 0x%x, vaddr = 0x%x, paddr = 0x%x\n",
+			samsung_fb->fb_size, (unsigned int)samsung_fb->vaddr,
+			(unsigned int)samsung_fb->paddr);
+
+	drm_helper_mode_fill_fb_struct(fb, mode_cmd);
+
+	return fb;
+
+fail:
+	kfree(samsung_fb);
+
+	return ERR_PTR(ret);
+}
+
+struct drm_framebuffer *samsung_drm_fb_create(struct drm_device *dev,
+					      struct drm_file *file_priv,
+					      struct drm_mode_fb_cmd *mode_cmd)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return samsung_drm_fb_init(file_priv, dev, mode_cmd);
+}
+
+void samsung_drm_fb_update_buf_off(struct drm_framebuffer *fb,
+				   unsigned int x, unsigned int y,
+				   struct samsung_drm_buffer_info *info)
+{
+	struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
+	unsigned long offset;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	offset = x * (fb->bits_per_pixel >> 3);
+	offset += y * fb->pitch;
+
+	info->base_addr = samsung_fb->paddr;
+	info->vaddr = samsung_fb->vaddr + offset;
+	info->paddr = samsung_fb->paddr + offset;
+
+	DRM_DEBUG_KMS("updated vaddr = 0x%x, paddr = 0x%x, offset = 0x%x\n",
+			(unsigned int)info->vaddr,
+			(unsigned int)info->paddr,
+			(unsigned int)offset);
+}
+
+static struct drm_mode_config_funcs samsung_drm_mode_config_funcs = {
+	.fb_create = samsung_drm_fb_create,
+};
+
+void samsung_drm_mode_config_init(struct drm_device *dev)
+{
+	dev->mode_config.min_width = 0;
+	dev->mode_config.min_height = 0;
+
+	/*
+	 * It sets max width and height as default value(4096x4096).
+	 * this value would be used to check for framebuffer size limitation
+	 * at drm_mode_addfb().
+	 */
+	dev->mode_config.max_width = 4096;
+	dev->mode_config.max_height = 4096;
+
+	dev->mode_config.funcs = &samsung_drm_mode_config_funcs;
+}
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.h b/drivers/gpu/drm/samsung/samsung_drm_fb.h
new file mode 100644
index 0000000..5c54521
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fb.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_FB_H_
+#define _SAMSUNG_DRM_FB_H
+
+struct samsung_drm_buffer_info {
+	unsigned long base_addr;
+	unsigned long paddr;
+	void __iomem *vaddr;
+};
+
+void samsung_drm_fb_update_buf_off(struct drm_framebuffer *fb,
+				   unsigned int x, unsigned int y,
+				   struct samsung_drm_buffer_info *info);
+
+struct drm_framebuffer *samsung_drm_fb_create(struct drm_device *dev,
+					      struct drm_file *file_priv,
+					      struct drm_mode_fb_cmd *mode_cmd);
+
+void samsung_drm_mode_config_init(struct drm_device *dev);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fbdev.c b/drivers/gpu/drm/samsung/samsung_drm_fbdev.c
new file mode 100644
index 0000000..9349d20
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fbdev.c
@@ -0,0 +1,409 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm_crtc.h"
+#include "drm_fb_helper.h"
+#include "drm_crtc_helper.h"
+#include "samsung_drm_fb.h"
+
+#include <drm/samsung_drm.h>
+
+#define to_samsung_fbdev(x)	container_of(x, struct samsung_drm_fbdev,\
+				drm_fb_helper)
+
+struct samsung_drm_fbdev {
+	struct drm_fb_helper	drm_fb_helper;
+	struct drm_framebuffer	*fb;
+};
+
+static inline unsigned int chan_to_field(unsigned int chan,
+					 struct fb_bitfield *bf)
+{
+	chan &= 0xffff;
+	chan >>= 16 - bf->length;
+
+	return chan << bf->offset;
+}
+
+static int samsung_drm_fbdev_setcolreg(unsigned regno, unsigned red,
+				       unsigned green, unsigned blue,
+				       unsigned transp, struct fb_info *info)
+{
+	unsigned int val;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	switch (info->fix.visual) {
+	case FB_VISUAL_TRUECOLOR:
+		if (regno < 16) {
+			u32 *pal = info->pseudo_palette;
+
+			val = chan_to_field(red, &info->var.red);
+			val |= chan_to_field(green, &info->var.green);
+			val |= chan_to_field(blue, &info->var.blue);
+
+			pal[regno] = val;
+		}
+		break;
+	default:
+		return 1;
+	}
+
+	return 0;
+}
+
+static struct fb_ops samsung_drm_fb_ops = {
+	.owner		= THIS_MODULE,
+	.fb_fillrect	= cfb_fillrect,
+	.fb_copyarea	= cfb_copyarea,
+	.fb_imageblit	= cfb_imageblit,
+	.fb_check_var	= drm_fb_helper_check_var,
+	.fb_set_par	= drm_fb_helper_set_par,
+	.fb_setcolreg	= samsung_drm_fbdev_setcolreg,
+	.fb_blank	= drm_fb_helper_blank,
+	.fb_pan_display	= drm_fb_helper_pan_display,
+	.fb_setcmap	= drm_fb_helper_setcmap,
+};
+
+static void samsung_drm_fbdev_update(struct drm_fb_helper *helper,
+				     struct drm_framebuffer *fb,
+				     unsigned int fb_width,
+				     unsigned int fb_height)
+{
+	struct fb_info *fbi = helper->fbdev;
+	struct drm_device *dev = helper->dev;
+	struct samsung_drm_fbdev *samsung_fb = to_samsung_fbdev(helper);
+	struct samsung_drm_buffer_info buffer_info;
+	unsigned int size = fb_width * fb_height * (fb->bits_per_pixel >> 3);
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_fb->fb = fb;
+
+	drm_fb_helper_fill_fix(fbi, fb->pitch, fb->depth);
+	drm_fb_helper_fill_var(fbi, helper, fb_width, fb_height);
+
+	samsung_drm_fb_update_buf_off(fb, fbi->var.xoffset, fbi->var.yoffset,
+			&buffer_info);
+
+	dev->mode_config.fb_base = buffer_info.base_addr;
+
+	fbi->screen_base = buffer_info.vaddr;
+	fbi->screen_size = size;
+	fbi->fix.smem_start = buffer_info.paddr;
+	fbi->fix.smem_len = size;
+}
+
+static int samsung_drm_fbdev_create(struct drm_fb_helper *helper,
+				    struct drm_fb_helper_surface_size *sizes)
+{
+	struct samsung_drm_fbdev *samsung_fbdev = to_samsung_fbdev(helper);
+	struct drm_device *dev = helper->dev;
+	struct fb_info *fbi;
+	struct drm_mode_fb_cmd mode_cmd = { 0 };
+	struct platform_device *pdev = dev->platformdev;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	DRM_DEBUG_KMS("surface width(%d), height(%d) and bpp(%d\n",
+			sizes->surface_width, sizes->surface_height,
+			sizes->surface_bpp);
+
+	mode_cmd.width = sizes->surface_width;
+	mode_cmd.height = sizes->surface_height;
+	mode_cmd.bpp = sizes->surface_bpp;
+	mode_cmd.depth = sizes->surface_depth;
+
+	mutex_lock(&dev->struct_mutex);
+
+	fbi = framebuffer_alloc(0, &pdev->dev);
+	if (!fbi) {
+		DRM_ERROR("failed to allocate fb info.\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	samsung_fbdev->fb = samsung_drm_fb_create(dev, NULL, &mode_cmd);
+	if (IS_ERR(samsung_fbdev->fb)) {
+		DRM_ERROR("failed to create drm dramebuffer.\n");
+		ret = PTR_ERR(samsung_fbdev->fb);
+		goto out;
+	}
+
+	helper->fb = samsung_fbdev->fb;
+	helper->fbdev = fbi;
+
+	fbi->par = helper;
+	fbi->flags = FBINFO_FLAG_DEFAULT;
+	fbi->fbops = &samsung_drm_fb_ops;
+
+	ret = fb_alloc_cmap(&fbi->cmap, 256, 0);
+	if (ret) {
+		DRM_ERROR("failed to allocate cmap.\n");
+		goto out;
+	}
+
+	samsung_drm_fbdev_update(helper, helper->fb, sizes->fb_width,
+			sizes->fb_height);
+
+out:
+	mutex_unlock(&dev->struct_mutex);
+	return ret;
+}
+
+static bool
+samsung_drm_fbdev_is_samefb(struct drm_framebuffer *fb,
+			    struct drm_fb_helper_surface_size *sizes)
+{
+	if (fb->width != sizes->surface_width)
+		return false;
+	if (fb->height != sizes->surface_height)
+		return false;
+	if (fb->bits_per_pixel != sizes->surface_bpp)
+		return false;
+	if (fb->depth != sizes->surface_depth)
+		return false;
+
+	return true;
+}
+
+static int samsung_drm_fbdev_recreate(struct drm_fb_helper *helper,
+				      struct drm_fb_helper_surface_size *sizes)
+{
+	struct drm_device *dev = helper->dev;
+	struct samsung_drm_fbdev *samsung_fbdev = to_samsung_fbdev(helper);
+	struct drm_framebuffer *fb = samsung_fbdev->fb;
+	struct drm_mode_fb_cmd mode_cmd = { 0 };
+
+	if (helper->fb != fb) {
+		DRM_ERROR("drm framebuffer is different\n");
+		return -EINVAL;
+	}
+
+	if (samsung_drm_fbdev_is_samefb(fb, sizes))
+		return 0;
+
+	mode_cmd.width = sizes->surface_width;
+	mode_cmd.height = sizes->surface_height;
+	mode_cmd.bpp = sizes->surface_bpp;
+	mode_cmd.depth = sizes->surface_depth;
+
+	if (fb->funcs->destroy)
+		fb->funcs->destroy(fb);
+
+	samsung_fbdev->fb = samsung_drm_fb_create(dev, NULL, &mode_cmd);
+	if (IS_ERR(samsung_fbdev->fb)) {
+		DRM_ERROR("failed to allocate fb.\n");
+		return PTR_ERR(samsung_fbdev->fb);
+	}
+
+	helper->fb = samsung_fbdev->fb;
+	samsung_drm_fbdev_update(helper, helper->fb, sizes->fb_width,
+			sizes->fb_height);
+
+	return 0;
+}
+
+static int samsung_drm_fbdev_probe(struct drm_fb_helper *helper,
+				   struct drm_fb_helper_surface_size *sizes)
+{
+	int ret = 0;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!helper->fb) {
+		ret = samsung_drm_fbdev_create(helper, sizes);
+		if (ret < 0) {
+			DRM_ERROR("failed to create fbdev.\n");
+			return ret;
+		}
+
+		/*
+		 * fb_helper expects a value more than 1 if succeed
+		 * because register_framebuffer() should be called.
+		 */
+		ret = 1;
+	} else {
+		ret = samsung_drm_fbdev_recreate(helper, sizes);
+		if (ret < 0) {
+			DRM_ERROR("failed to reconfigure fbdev\n");
+			return ret;
+		}
+	}
+
+	return ret;
+}
+
+static struct drm_fb_helper_funcs samsung_drm_fb_helper_funcs = {
+	.fb_probe =	samsung_drm_fbdev_probe,
+};
+
+int samsung_drm_fbdev_init(struct drm_device *dev)
+{
+	struct samsung_drm_fbdev *fbdev;
+	struct samsung_drm_private *private = dev->dev_private;
+	struct drm_fb_helper *helper;
+	unsigned int num_crtc;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!dev->mode_config.num_crtc || !dev->mode_config.num_connector)
+		return 0;
+
+	fbdev = kzalloc(sizeof(*fbdev), GFP_KERNEL);
+	if (!fbdev) {
+		DRM_ERROR("failed to allocate drm fbdev.\n");
+		return -ENOMEM;
+	}
+
+	private->fb_helper = helper = &fbdev->drm_fb_helper;
+	helper->funcs = &samsung_drm_fb_helper_funcs;
+
+	num_crtc = dev->mode_config.num_crtc;
+
+	ret = drm_fb_helper_init(dev, helper, num_crtc, 4);
+	if (ret < 0) {
+		DRM_ERROR("failed to initialize drm fb helper.\n");
+		goto fail;
+	}
+
+	ret = drm_fb_helper_single_add_all_connectors(helper);
+	if (ret < 0) {
+		DRM_ERROR("failed to register drm_fb_helper_connector.\n");
+		goto fail;
+
+	}
+
+	ret = drm_fb_helper_initial_config(helper, 32);
+	if (ret < 0) {
+		DRM_ERROR("failed to set up hw configuration.\n");
+		goto fail;
+	}
+
+	return 0;
+
+fail:
+	private->fb_helper = NULL;
+	kfree(fbdev);
+
+	return ret;
+}
+
+static void samsung_drm_fbdev_destroy(struct drm_device *dev,
+				      struct drm_fb_helper *fb_helper)
+{
+	struct drm_framebuffer *fb;
+	struct fb_info *info;
+
+	/* release drm framebuffer and real buffer */
+	if (fb_helper->fb && fb_helper->fb->funcs) {
+		fb = fb_helper->fb;
+		if (fb->funcs->destroy)
+			fb->funcs->destroy(fb);
+	}
+
+	/* release linux framebuffer */
+	if (fb_helper->fbdev) {
+		info = fb_helper->fbdev;
+		unregister_framebuffer(info);
+		if (info->cmap.len)
+			fb_dealloc_cmap(&info->cmap);
+		framebuffer_release(info);
+	}
+
+	drm_fb_helper_fini(fb_helper);
+}
+
+void samsung_drm_fbdev_fini(struct drm_device *dev)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+	struct samsung_drm_fbdev *fbdev;
+
+	if (!private || !private->fb_helper)
+		return;
+
+	fbdev = to_samsung_fbdev(private->fb_helper);
+
+	samsung_drm_fbdev_destroy(dev, private->fb_helper);
+	kfree(fbdev);
+	private->fb_helper = NULL;
+}
+
+void samsung_drm_fbdev_restore_mode(struct drm_device *dev)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+
+	if (!private || !private->fb_helper)
+		return;
+
+	drm_fb_helper_restore_fbdev_mode(private->fb_helper);
+}
+
+int samsung_drm_fbdev_reinit(struct drm_device *dev)
+{
+	struct samsung_drm_private *private = dev->dev_private;
+	int ret;
+
+	if (!private)
+		return -EINVAL;
+
+	if (!dev->mode_config.num_connector) {
+		samsung_drm_fbdev_fini(dev);
+		return 0;
+	}
+
+	if (private->fb_helper) {
+		struct drm_fb_helper *fb_helper = private->fb_helper;
+
+		drm_fb_helper_fini(fb_helper);
+
+		ret = drm_fb_helper_init(dev, fb_helper,
+				dev->mode_config.num_crtc, 4);
+		if (ret < 0) {
+			DRM_ERROR("failed to initialize drm fb helper\n");
+			return ret;
+		}
+
+		ret = drm_fb_helper_single_add_all_connectors(fb_helper);
+		if (ret < 0) {
+			DRM_ERROR("failed to add fb helper to connectors\n");
+			return ret;
+		}
+
+		ret = drm_fb_helper_initial_config(fb_helper, 32);
+		if (ret < 0) {
+			DRM_ERROR("failed to set up hw configuration.\n");
+			return ret;
+		}
+	} else
+		ret = samsung_drm_fbdev_init(dev);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fbdev.h b/drivers/gpu/drm/samsung/samsung_drm_fbdev.h
new file mode 100644
index 0000000..3ef4e0d
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fbdev.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ *
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *	SeungWoo Kim <sw0312.kim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_FBDEV_H_
+#define _SAMSUNG_DRM_FBDEV_H_
+
+int samsung_drm_fbdev_init(struct drm_device *dev);
+int samsung_drm_fbdev_reinit(struct drm_device *dev);
+void samsung_drm_fbdev_fini(struct drm_device *dev);
+void samsung_drm_fbdev_restore_mode(struct drm_device *dev);
+
+#endif
diff --git a/drivers/gpu/drm/samsung/samsung_drm_fimd.c b/drivers/gpu/drm/samsung/samsung_drm_fimd.c
new file mode 100644
index 0000000..e4dadc1
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_fimd.c
@@ -0,0 +1,611 @@
+/*
+ * Copyright (C) 2011 Samsung Electronics Co.Ltd
+ * Author: Joonyoung Shim <jy0922.shim@samsung.com>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+
+/*
+ * TODO list
+ *  - Code cleanup(FIXME and TODO parts)
+ *  - Support multiple driver instance
+ *  - Clock gating and Power management
+ *  - Writeback feature
+ */
+
+#include "drmP.h"
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/clk.h>
+
+#include <drm/samsung_drm.h>
+#include <plat/regs-fb-v4.h>
+
+#include "samsung_drm_fbdev.h"
+#include "samsung_drm_crtc.h"
+
+/* irq_flags bits */
+#define FIMD_VSYNC_IRQ_EN	0
+
+#define VIDOSD_A(win)	(VIDOSD_BASE + 0x00 + (win) * 16)
+#define VIDOSD_B(win)	(VIDOSD_BASE + 0x04 + (win) * 16)
+#define VIDOSD_C(win)	(VIDOSD_BASE + 0x08 + (win) * 16)
+#define VIDOSD_D(win)	(VIDOSD_BASE + 0x0C + (win) * 16)
+
+#define VIDWx_BUF_START(win, buf)	(VIDW_BUF_START(buf) + (win) * 8)
+#define VIDWx_BUF_END(win, buf)		(VIDW_BUF_END(buf) + (win) * 8)
+#define VIDWx_BUF_SIZE(win, buf)	(VIDW_BUF_SIZE(buf) + (win) * 4)
+
+#define WINDOWS_NR	5
+
+#define get_fimd_data(dev)	platform_get_drvdata(to_platform_device(dev))
+
+struct fimd_win_data {
+	unsigned int		offset_x;
+	unsigned int		offset_y;
+	unsigned int		width;
+	unsigned int		height;
+	unsigned int		paddr;
+	void __iomem		*vaddr;
+	unsigned int		end_buf_off;
+	unsigned int		buf_offsize;
+	unsigned int		line_size;	/* bytes */
+};
+
+struct fimd_data {
+	struct drm_crtc			*crtc;
+	struct clk			*bus_clk;
+	struct clk			*lcd_clk;
+	struct resource			*regs_res;
+	void __iomem			*regs;
+	struct fimd_win_data		win_data[WINDOWS_NR];
+	unsigned int			clkdiv;
+	unsigned int			default_win;
+	unsigned int			bpp;
+	unsigned long			irq_flags;
+	u32				vidcon0;
+	u32				vidcon1;
+};
+
+/* TODO: remove global variables */
+static struct samsung_drm_subdrv fimd_subdrv;
+static struct fb_videomode fimd_timing;
+
+static bool fimd_display_is_connected(void)
+{
+	return true;
+}
+
+static void *fimd_get_timing(void)
+{
+	return &fimd_timing;
+}
+
+static int fimd_check_timing(void *timing)
+{
+	return 0;
+}
+
+static int fimd_display_power_on(int mode)
+{
+	return 0;
+}
+
+static struct samsung_drm_display fimd_display = {
+	.type = SAMSUNG_DISPLAY_TYPE_LCD,
+	.is_connected = fimd_display_is_connected,
+	.get_timing = fimd_get_timing,
+	.check_timing = fimd_check_timing,
+	.power_on = fimd_display_power_on,
+};
+
+static void fimd_commit(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	/* vidcon0 */
+	val = data->vidcon0;
+	val &= ~(VIDCON0_CLKVAL_F_MASK | VIDCON0_CLKDIR);
+
+	if (data->clkdiv > 1)
+		val |= VIDCON0_CLKVAL_F(data->clkdiv - 1) | VIDCON0_CLKDIR;
+	else
+		val &= ~VIDCON0_CLKDIR;	/* 1:1 clock */
+
+	val |= VIDCON0_ENVID | VIDCON0_ENVID_F;
+	writel(val, regs + VIDCON0);
+
+	/* vidcon1 */
+	writel(data->vidcon1, regs + VIDCON1);
+
+	/* vidtcon0 */
+	val = VIDTCON0_VBPD(fimd_timing.upper_margin - 1) |
+	       VIDTCON0_VFPD(fimd_timing.lower_margin - 1) |
+	       VIDTCON0_VSPW(fimd_timing.vsync_len - 1);
+	writel(val, regs + VIDTCON0);
+
+	/* vidtcon1 */
+	val = VIDTCON1_HBPD(fimd_timing.left_margin - 1) |
+	       VIDTCON1_HFPD(fimd_timing.right_margin - 1) |
+	       VIDTCON1_HSPW(fimd_timing.hsync_len - 1);
+	writel(val, regs + VIDTCON1);
+
+	/* vidtcon2 */
+	val = VIDTCON2_LINEVAL(fimd_timing.yres - 1) |
+	       VIDTCON2_HOZVAL(fimd_timing.xres - 1);
+	writel(val, regs + VIDTCON2);
+}
+
+static int fimd_enable_vblank(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	if (!test_and_set_bit(FIMD_VSYNC_IRQ_EN, &data->irq_flags)) {
+		val = readl(regs + VIDINTCON0);
+
+		val |= VIDINTCON0_INT_ENABLE;
+		val |= VIDINTCON0_INT_FRAME;
+
+		val &= ~VIDINTCON0_FRAMESEL0_MASK;
+		val |= VIDINTCON0_FRAMESEL0_VSYNC;
+		val &= ~VIDINTCON0_FRAMESEL1_MASK;
+		val |= VIDINTCON0_FRAMESEL1_NONE;
+
+		writel(val, regs + VIDINTCON0);
+	}
+
+	return 0;
+}
+
+static void fimd_disable_vblank(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	if (test_and_clear_bit(FIMD_VSYNC_IRQ_EN, &data->irq_flags)) {
+		val = readl(regs + VIDINTCON0);
+
+		val &= ~VIDINTCON0_INT_FRAME;
+		val &= ~VIDINTCON0_INT_ENABLE;
+
+		writel(val, regs + VIDINTCON0);
+	}
+}
+
+static struct samsung_drm_manager_ops fimd_manager_ops = {
+	.commit = fimd_commit,
+	.enable_vblank = fimd_enable_vblank,
+	.disable_vblank = fimd_disable_vblank,
+};
+
+static void fimd_win_mode_set(struct device *dev,
+			      struct samsung_drm_overlay *overlay)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	struct fimd_win_data *win_data;
+
+	if (!overlay) {
+		dev_err(dev, "overlay is NULL\n");
+		return;
+	}
+
+	win_data = &data->win_data[data->default_win];
+
+	win_data->offset_x = overlay->offset_x;
+	win_data->offset_y = overlay->offset_y;
+	win_data->width = overlay->width;
+	win_data->height = overlay->height;
+	win_data->paddr = overlay->paddr;
+	win_data->vaddr = overlay->vaddr;
+	win_data->end_buf_off = overlay->end_buf_off * (data->bpp >> 3);
+	win_data->buf_offsize = overlay->buf_offsize * (data->bpp >> 3);
+	win_data->line_size = overlay->line_size * (data->bpp >> 3);
+}
+
+static void fimd_win_commit(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	struct fimd_win_data *win_data;
+	int win = data->default_win;
+	u32 val;
+
+	if (win < 0 || win > WINDOWS_NR)
+		return;
+
+	win_data = &data->win_data[win];
+
+	/* protect windows */
+	val = readl(regs + SHADOWCON);
+	val |= SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+
+	/* buffer start address */
+	val = win_data->paddr;
+	writel(val, regs + VIDWx_BUF_START(win, 0));
+
+	/* buffer end address */
+	val = win_data->paddr + win_data->end_buf_off;
+	writel(val, regs + VIDWx_BUF_END(win, 0));
+
+	/* buffer size */
+	val = VIDW_BUF_SIZE_OFFSET(win_data->buf_offsize) |
+		VIDW_BUF_SIZE_PAGEWIDTH(win_data->line_size);
+	writel(val, regs + VIDWx_BUF_SIZE(win, 0));
+
+	/* OSD position */
+	val = VIDOSDxA_TOPLEFT_X(win_data->offset_x) |
+		VIDOSDxA_TOPLEFT_Y(win_data->offset_y);
+	writel(val, regs + VIDOSD_A(win));
+
+	val = VIDOSDxB_BOTRIGHT_X(win_data->offset_x + win_data->width - 1) |
+		VIDOSDxB_BOTRIGHT_Y(win_data->offset_y + win_data->height - 1);
+	writel(val, regs + VIDOSD_B(win));
+
+	/* TODO: OSD alpha */
+
+	/* OSD size */
+	if (win != 3 && win != 4) {
+		u32 offset = VIDOSD_D(win);
+		if (win == 0)
+			offset = VIDOSD_C(win);
+		val = win_data->width * win_data->height;
+		writel(val, regs + offset);
+	}
+
+	/* FIXME: remove fixed values */
+	val = WINCONx_ENWIN;
+	val |= WINCON0_BPPMODE_24BPP_888;
+	val |= WINCONx_WSWP;
+	val |= WINCONx_BURSTLEN_16WORD;
+	writel(val, regs + WINCON(win));
+
+	/* TODO: colour key */
+
+	/* Enable DMA channel and unprotect windows */
+	val = readl(regs + SHADOWCON);
+	val |= SHADOWCON_CHx_ENABLE(win);
+	val &= ~SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+}
+
+static void fimd_win_disable(struct device *dev)
+{
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	struct fimd_win_data *win_data;
+	int win = data->default_win;
+	u32 val;
+
+	if (win < 0 || win > WINDOWS_NR)
+		return;
+
+	win_data = &data->win_data[win];
+
+	/* protect windows */
+	val = readl(regs + SHADOWCON);
+	val |= SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+
+	/* wincon */
+	val = readl(regs + WINCON(win));
+	val &= ~WINCONx_ENWIN;
+	writel(val, regs + WINCON(win));
+
+	/* unprotect windows */
+	val = readl(regs + SHADOWCON);
+	val &= ~SHADOWCON_CHx_ENABLE(win);
+	val &= ~SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+}
+
+static struct samsung_drm_overlay_ops fimd_overlay_ops = {
+	.mode_set = fimd_win_mode_set,
+	.commit = fimd_win_commit,
+	.disable = fimd_win_disable,
+};
+
+/* for pageflip event */
+static void fimd_finish_pageflip(struct drm_device *drm_dev, int crtc)
+{
+	struct samsung_drm_private *dev_priv = drm_dev->dev_private;
+	struct drm_pending_vblank_event *e, *t;
+	struct timeval now;
+	unsigned long flags;
+
+	if (!dev_priv->pageflip_event)
+		return;
+
+	spin_lock_irqsave(&drm_dev->event_lock, flags);
+
+	list_for_each_entry_safe(e, t, &dev_priv->pageflip_event_list,
+			base.link) {
+		do_gettimeofday(&now);
+		e->event.sequence = 0;
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+
+		list_move_tail(&e->base.link, &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	}
+
+	drm_vblank_put(drm_dev, crtc);
+	dev_priv->pageflip_event = false;
+
+	spin_unlock_irqrestore(&drm_dev->event_lock, flags);
+}
+
+static irqreturn_t fimd_irq_handler(DRM_IRQ_ARGS)
+{
+	struct drm_device *drm_dev = (struct drm_device *)arg;
+	struct device *dev = fimd_subdrv.manager.dev;
+	struct samsung_drm_manager *manager = &fimd_subdrv.manager;
+	struct fimd_data *data = get_fimd_data(dev);
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	val = readl(regs + VIDINTCON1);
+
+	if (val & VIDINTCON1_INT_FRAME)
+		/* VSYNC interrupt */
+		writel(VIDINTCON1_INT_FRAME, regs + VIDINTCON1);
+
+	drm_handle_vblank(drm_dev, manager->pipe);
+	fimd_finish_pageflip(drm_dev, manager->pipe);
+
+	return IRQ_HANDLED;
+}
+
+static int fimd_subdrv_probe(struct drm_device *drm_dev)
+{
+	struct drm_driver *drm_driver = drm_dev->driver;
+	int ret;
+
+	drm_driver->irq_handler = fimd_irq_handler;
+
+	ret = drm_irq_install(drm_dev);
+	if (ret) {
+		drm_driver->irq_handler = NULL;
+		return ret;
+	}
+
+	/* FIXME */
+	drm_dev->vblank_disable_allowed = 1;
+
+	return 0;
+}
+
+static void fimd_subdrv_remove(struct drm_device *drm_dev)
+{
+	struct drm_driver *drm_driver = drm_dev->driver;
+
+	drm_irq_uninstall(drm_dev);
+
+	drm_driver->irq_handler = NULL;
+}
+
+static struct samsung_drm_subdrv fimd_subdrv = {
+	.probe = fimd_subdrv_probe,
+	.remove = fimd_subdrv_remove,
+	.manager = {
+		.pipe = -1,
+		.ops = &fimd_manager_ops,
+		.overlay_ops = &fimd_overlay_ops,
+		.display = &fimd_display,
+	},
+};
+
+static int fimd_calc_clkdiv(struct fimd_data *data,
+			    struct fb_videomode *timing)
+{
+	unsigned long clk = clk_get_rate(data->lcd_clk);
+	u32 retrace;
+	u32 clkdiv;
+	u32 best_framerate = 0;
+	u32 framerate;
+
+	retrace = timing->left_margin + timing->hsync_len +
+				timing->right_margin + timing->xres;
+	retrace *= timing->upper_margin + timing->vsync_len +
+				timing->lower_margin + timing->yres;
+
+	/* default framerate is 60Hz */
+	if (!timing->refresh)
+		timing->refresh = 60;
+
+	clk /= retrace;
+
+	for (clkdiv = 1; clkdiv < 0x100; clkdiv++) {
+		int tmp;
+
+		/* get best framerate */
+		framerate = clk / clkdiv;
+		tmp = timing->refresh - framerate;
+		if (tmp < 0) {
+			best_framerate = framerate;
+			continue;
+		} else {
+			if (!best_framerate)
+				best_framerate = framerate;
+			else if (tmp < (best_framerate - framerate))
+				best_framerate = framerate ;
+			break;
+		}
+	}
+
+	return clkdiv;
+}
+
+static void fimd_clear_win(struct fimd_data *data, int win)
+{
+	void __iomem *regs = data->regs;
+	u32 val;
+
+	writel(0, regs + WINCON(win));
+	writel(0, regs + VIDOSD_A(win));
+	writel(0, regs + VIDOSD_B(win));
+	writel(0, regs + VIDOSD_C(win));
+
+	if (win == 1 || win == 2)
+		writel(0, regs + VIDOSD_D(win));
+
+	val = readl(regs + SHADOWCON);
+	val &= ~SHADOWCON_WINx_PROTECT(win);
+	writel(val, regs + SHADOWCON);
+}
+
+static int __devinit fimd_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct fimd_data *data;
+	struct samsung_drm_fimd_pdata *pdata;
+	struct resource *res;
+	int win;
+	int ret;
+
+	printk(KERN_DEBUG "[%d] %s\n", __LINE__, __func__);
+
+	pdata = pdev->dev.platform_data;
+	if (!pdata) {
+		dev_err(dev, "no platform data specified\n");
+		return -EINVAL;
+	}
+
+	data = kzalloc(sizeof(struct fimd_data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	data->bus_clk = clk_get(dev, "fimd");
+	if (IS_ERR(data->bus_clk)) {
+		dev_err(dev, "failed to get bus clock\n");
+		ret = PTR_ERR(data->bus_clk);
+		goto err_data;
+	}
+
+	clk_enable(data->bus_clk);
+
+	data->lcd_clk = clk_get(dev, "sclk_fimd");
+	if (IS_ERR(data->lcd_clk)) {
+		dev_err(dev, "failed to get lcd clock\n");
+		ret = PTR_ERR(data->lcd_clk);
+		goto err_bus_clk;
+	}
+
+	clk_enable(data->lcd_clk);
+
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res) {
+		dev_err(dev, "failed to find registers\n");
+		ret = -ENOENT;
+		goto err_clk;
+	}
+
+	data->regs_res = request_mem_region(res->start, resource_size(res),
+					   dev_name(dev));
+	if (!data->regs_res) {
+		dev_err(dev, "failed to claim register region\n");
+		ret = -ENOENT;
+		goto err_clk;
+	}
+
+	data->regs = ioremap(res->start, resource_size(res));
+	if (!data->regs) {
+		dev_err(dev, "failed to map registers\n");
+		ret = -ENXIO;
+		goto err_req_region;
+	}
+
+	for (win = 0; win < WINDOWS_NR; win++)
+		fimd_clear_win(data, win);
+
+	data->clkdiv = fimd_calc_clkdiv(data, &pdata->timing);
+	data->vidcon0 = pdata->vidcon0;
+	data->vidcon1 = pdata->vidcon1;
+	data->default_win = pdata->default_win;
+	data->bpp = pdata->bpp;
+
+	platform_set_drvdata(pdev, data);
+
+	memcpy(&fimd_timing, &pdata->timing,
+			sizeof(struct fb_videomode));
+	fimd_timing.pixclock = clk_get_rate(data->lcd_clk) / data->clkdiv;
+
+	fimd_subdrv.manager.dev = dev;
+
+	samsung_drm_subdrv_register(&fimd_subdrv);
+
+	return 0;
+
+err_req_region:
+	release_resource(data->regs_res);
+	kfree(data->regs_res);
+
+err_clk:
+	clk_disable(data->lcd_clk);
+	clk_put(data->lcd_clk);
+
+err_bus_clk:
+	clk_disable(data->bus_clk);
+	clk_put(data->bus_clk);
+
+err_data:
+	kfree(data);
+	return ret;
+}
+
+static int __devexit fimd_remove(struct platform_device *pdev)
+{
+	struct fimd_data *data = platform_get_drvdata(pdev);
+
+	samsung_drm_subdrv_unregister(&fimd_subdrv);
+
+	clk_disable(data->lcd_clk);
+	clk_disable(data->bus_clk);
+	clk_put(data->lcd_clk);
+	clk_put(data->bus_clk);
+
+	iounmap(data->regs);
+	release_resource(data->regs_res);
+	kfree(data->regs_res);
+
+	kfree(data);
+
+	return 0;
+}
+
+static struct platform_driver fimd_driver = {
+	.probe		= fimd_probe,
+	.remove		= __devexit_p(fimd_remove),
+	.driver		= {
+		.name	= "exynos4-fb",
+		.owner	= THIS_MODULE,
+	},
+};
+
+static int __init fimd_init(void)
+{
+	return platform_driver_register(&fimd_driver);
+}
+
+static void __exit fimd_exit(void)
+{
+	platform_driver_unregister(&fimd_driver);
+}
+
+module_init(fimd_init);
+module_exit(fimd_exit);
+
+MODULE_AUTHOR("Joonyoung Shim <jy0922.shim@samsung.com>");
+MODULE_DESCRIPTION("Samsung DRM FIMD Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c b/drivers/gpu/drm/samsung/samsung_drm_gem.c
new file mode 100644
index 0000000..1e167a6
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
@@ -0,0 +1,492 @@
+/* samsung_drm_gem.c
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Author: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "drmP.h"
+#include "drm.h"
+#include "samsung_drm.h"
+
+#include <drm/samsung_drm.h>
+
+#include "samsung_drm_gem.h"
+#include "samsung_drm_buf.h"
+
+static unsigned int convert_to_vm_err_msg(int msg)
+{
+	unsigned int out_msg;
+
+	switch (msg) {
+	case 0:
+	case -ERESTARTSYS:
+	case -EINTR:
+		out_msg = VM_FAULT_NOPAGE;
+		break;
+
+	case -ENOMEM:
+		out_msg = VM_FAULT_OOM;
+		break;
+
+	default:
+		out_msg = VM_FAULT_SIGBUS;
+		break;
+	}
+
+	return out_msg;
+}
+
+static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
+}
+
+/**
+ * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for an object
+ * @obj: obj in question
+ *
+ * GEM memory mapping works by handing back to userspace a fake mmap offset
+ * it can use in a subsequent mmap(2) call.  The DRM core code then looks
+ * up the object based on the offset and sets up the various memory mapping
+ * structures.
+ *
+ * This routine allocates and attaches a fake offset for @obj.
+ */
+static int
+samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)
+{
+	struct drm_device *dev = obj->dev;
+	struct drm_gem_mm *mm = dev->mm_private;
+	struct drm_map_list *list;
+	struct drm_local_map *map;
+	int ret = 0;
+
+	/* Set the object up for mmap'ing */
+	list = &obj->map_list;
+	list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
+	if (!list->map)
+		return -ENOMEM;
+
+	map = list->map;
+	map->type = _DRM_GEM;
+	map->size = obj->size;
+	map->handle = obj;
+
+	/* Get a DRM GEM mmap offset allocated... */
+	list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
+						    obj->size / PAGE_SIZE,
+						    0, 0);
+	if (!list->file_offset_node) {
+		DRM_ERROR("failed to allocate offset for bo %d\n",
+			  obj->name);
+		ret = -ENOSPC;
+		goto out_free_list;
+	}
+
+	list->file_offset_node = drm_mm_get_block(list->file_offset_node,
+						  obj->size / PAGE_SIZE,
+						  0);
+	if (!list->file_offset_node) {
+		ret = -ENOMEM;
+		goto out_free_list;
+	}
+
+	list->hash.key = list->file_offset_node->start;
+	ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
+	if (ret) {
+		DRM_ERROR("failed to add to map hash\n");
+		goto out_free_mm;
+	}
+
+	return 0;
+
+out_free_mm:
+	drm_mm_put_block(list->file_offset_node);
+out_free_list:
+	kfree(list->map);
+	list->map = NULL;
+
+	return ret;
+}
+
+static void
+samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
+{
+	struct drm_device *dev = obj->dev;
+	struct drm_gem_mm *mm = dev->mm_private;
+	struct drm_map_list *list = &obj->map_list;
+
+	drm_ht_remove_item(&mm->offset_hash, &list->hash);
+	drm_mm_put_block(list->file_offset_node);
+	kfree(list->map);
+	list->map = NULL;
+}
+
+static int samsung_drm_gem_create(struct drm_file *file, struct drm_device *dev,
+		unsigned int size, unsigned int *handle_p)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+	struct drm_gem_object *obj;
+	unsigned int handle;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	size = roundup(size, PAGE_SIZE);
+
+	/* allocate the new buffer object and memory region. */
+	samsung_gem_obj = samsung_drm_buf_create(dev, size);
+	if (!samsung_gem_obj)
+		return -ENOMEM;
+
+	obj = &samsung_gem_obj->base;
+
+	ret = drm_gem_object_init(dev, obj, size);
+	if (ret < 0) {
+		DRM_ERROR("failed to initailize gem object.\n");
+		goto out;
+	}
+
+	DRM_DEBUG_KMS("created file object = 0x%x\n", (unsigned int)obj->filp);
+
+	ret = samsung_drm_gem_create_mmap_offset(obj);
+	if (ret < 0) {
+		DRM_ERROR("failed to allocate mmap offset.\n");
+		goto out;
+	}
+
+	/**
+	 * allocate a id of idr table where the obj is registered
+	 * and handle has the id what user can see.
+	 */
+	ret = drm_gem_handle_create(file, obj, &handle);
+	if (ret) {
+		drm_gem_object_release(obj);
+		samsung_drm_buf_destroy(dev, samsung_gem_obj);
+		goto out;
+	}
+
+	DRM_DEBUG_KMS("gem handle = 0x%x\n", handle);
+
+	/* drop reference from allocate - handle holds it now. */
+	drm_gem_object_unreference_unlocked(obj);
+
+	*handle_p = handle;
+
+	return 0;
+
+out:
+	drm_gem_object_unreference_unlocked(obj);
+
+	samsung_drm_buf_destroy(dev, samsung_gem_obj);
+
+	kfree(samsung_gem_obj);
+
+	return ret;
+}
+
+int samsung_drm_gem_create_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file)
+{
+	struct drm_samsung_gem_create *args = data;
+
+	DRM_DEBUG_KMS("%s : size = 0x%x\n", __FILE__, args->size);
+
+	return samsung_drm_gem_create(file, dev, args->size, &args->handle);
+}
+
+int samsung_drm_gem_map_offset_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file)
+{
+	struct drm_samsung_gem_map_off *args = data;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	DRM_DEBUG_KMS("handle = 0x%x, offset = 0x%x\n",
+			args->handle, (u32)args->offset);
+
+	if (!(dev->driver->driver_features & DRIVER_GEM)) {
+		DRM_ERROR("not support GEM.\n");
+		return -ENODEV;
+	}
+
+	return samsung_drm_gem_dumb_map_offset(file, dev, args->handle,
+			&args->offset);
+}
+
+static int samsung_drm_gem_mmap_buffer(struct file *filp,
+		struct vm_area_struct *vma)
+{
+	struct drm_gem_object *obj = filp->private_data;
+	struct samsung_drm_gem_obj *samsung_gem_obj = to_samsung_gem_obj(obj);
+	struct samsung_drm_buf_entry *entry;
+	unsigned long pfn, size;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	vma->vm_flags |= (VM_IO | VM_RESERVED);
+	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+	vma->vm_file = filp;
+
+	size = vma->vm_end - vma->vm_start;
+	entry = samsung_gem_obj->entry;
+
+	/* check if the region to be mapped is valid or not. */
+	if ((entry->paddr + vma->vm_pgoff + size) >
+			(entry->paddr + entry->size)) {
+		drm_gem_object_unreference_unlocked(obj);
+		DRM_ERROR("desired size is bigger then real size.\n");
+		return -EINVAL;
+	}
+
+	pfn = (samsung_gem_obj->entry->paddr + vma->vm_pgoff) >> PAGE_SHIFT;
+
+	DRM_DEBUG_KMS("offset = 0x%x, pfn to be mapped = 0x%x\n",
+			(u32)vma->vm_pgoff, (u32)pfn);
+
+	if (remap_pfn_range(vma, vma->vm_start, pfn,
+				vma->vm_end - vma->vm_start,
+				vma->vm_page_prot)) {
+		DRM_ERROR("failed to remap pfn range.\n");
+		return -EAGAIN;
+	}
+
+	return 0;
+}
+
+static const struct file_operations samsung_drm_gem_fops = {
+	.mmap = samsung_drm_gem_mmap_buffer,
+};
+
+int samsung_drm_gem_mmap_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *filp)
+{
+	struct drm_samsung_gem_mmap *args = data;
+	struct drm_gem_object *obj;
+	unsigned int addr;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	if (!(dev->driver->driver_features & DRIVER_GEM)) {
+		DRM_ERROR("not support GEM.\n");
+		return -ENODEV;
+	}
+
+	obj = drm_gem_object_lookup(dev, filp, args->handle);
+	if (!obj) {
+		DRM_ERROR("failed to lookup gem object.\n");
+		return -EINVAL;
+	}
+
+	obj->filp->f_op = &samsung_drm_gem_fops;
+	obj->filp->private_data = obj;
+
+	down_write(&current->mm->mmap_sem);
+	addr = do_mmap(obj->filp, 0, args->size,
+			PROT_READ | PROT_WRITE, MAP_SHARED, args->offset);
+	up_write(&current->mm->mmap_sem);
+
+	drm_gem_object_unreference_unlocked(obj);
+
+	if (IS_ERR((void *)addr))
+		return PTR_ERR((void *)addr);
+
+	args->mapped = addr;
+
+	DRM_DEBUG_KMS("mapped = 0x%x\n", args->mapped);
+
+	return 0;
+}
+
+int samsung_drm_gem_init_object(struct drm_gem_object *obj)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	return 0;
+}
+
+void samsung_drm_gem_free_object(struct drm_gem_object *gem_obj)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	DRM_DEBUG_KMS("handle count = %d\n",
+			atomic_read(&gem_obj->handle_count));
+
+	if (gem_obj->map_list.map)
+		samsung_drm_gem_free_mmap_offset(gem_obj);
+
+	/* release file pointer to gem object. */
+	drm_gem_object_release(gem_obj);
+
+	samsung_gem_obj = to_samsung_gem_obj(gem_obj);
+
+	samsung_drm_buf_destroy(gem_obj->dev, samsung_gem_obj);
+}
+
+struct samsung_drm_gem_obj *
+		find_samsung_drm_gem_object(struct drm_file *file_priv,
+			struct drm_device *dev, unsigned int handle)
+{
+	struct drm_gem_object *gem_obj;
+
+	gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
+	if (!gem_obj) {
+		DRM_LOG_KMS("a invalid gem object not registered to lookup.\n");
+		return NULL;
+	}
+
+	/**
+	 * unreference refcount of the gem object.
+	 * at drm_gem_object_lookup(), the gem object was referenced.
+	 */
+	drm_gem_object_unreference(gem_obj);
+
+	return to_samsung_gem_obj(gem_obj);
+}
+
+int samsung_drm_gem_dumb_create(struct drm_file *file_priv,
+		struct drm_device *dev, struct drm_mode_create_dumb *args)
+{
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	/**
+	 * alocate memory to be used for framebuffer.
+	 * - this callback would be called by user application
+	 *	with DRM_IOCTL_MODE_CREATE_DUMB command.
+	 */
+
+	args->pitch = args->width * args->bpp >> 3;
+	args->size = args->pitch * args->height;
+
+	return samsung_drm_gem_create(file_priv, dev, args->size,
+			&args->handle);
+}
+
+int samsung_drm_gem_dumb_map_offset(struct drm_file *file_priv,
+		struct drm_device *dev, uint32_t handle, uint64_t *offset)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	mutex_lock(&dev->struct_mutex);
+
+	/**
+	 * get offset of memory allocated for drm framebuffer.
+	 * - this callback would be called by user application
+	 *	with DRM_IOCTL_MODE_MAP_DUMB command.
+	 */
+
+	samsung_gem_obj = find_samsung_drm_gem_object(file_priv, dev, handle);
+	if (!samsung_gem_obj) {
+		DRM_ERROR("failed to get samsung_drm_get_obj.\n");
+		mutex_unlock(&dev->struct_mutex);
+		return -EINVAL;
+	}
+
+	*offset = get_gem_mmap_offset(&samsung_gem_obj->base);
+
+	DRM_DEBUG_KMS("offset = 0x%x\n", (unsigned int)*offset);
+
+	mutex_unlock(&dev->struct_mutex);
+
+	return 0;
+}
+
+int samsung_drm_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	struct drm_gem_object *obj = vma->vm_private_data;
+	struct samsung_drm_gem_obj *samsung_gem_obj = to_samsung_gem_obj(obj);
+	struct drm_device *dev = obj->dev;
+	unsigned long pfn;
+	pgoff_t page_offset;
+	int ret;
+
+	page_offset = ((unsigned long)vmf->virtual_address -
+			vma->vm_start) >> PAGE_SHIFT;
+
+	mutex_lock(&dev->struct_mutex);
+
+	pfn = (samsung_gem_obj->entry->paddr >> PAGE_SHIFT) + page_offset;
+
+	ret = vm_insert_mixed(vma, (unsigned long)vmf->virtual_address, pfn);
+
+	mutex_unlock(&dev->struct_mutex);
+
+	return convert_to_vm_err_msg(ret);
+}
+
+int samsung_drm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	/* set vm_area_struct. */
+	ret = drm_gem_mmap(filp, vma);
+	if (ret < 0) {
+		DRM_ERROR("failed to mmap.\n");
+		return ret;
+	}
+
+	vma->vm_flags &= ~VM_PFNMAP;
+	vma->vm_flags |= VM_MIXEDMAP;
+
+	return ret;
+}
+
+
+int samsung_drm_gem_dumb_destroy(struct drm_file *file_priv,
+		struct drm_device *dev, unsigned int handle)
+{
+	struct samsung_drm_gem_obj *samsung_gem_obj;
+	int ret;
+
+	DRM_DEBUG_KMS("%s\n", __FILE__);
+
+	samsung_gem_obj = find_samsung_drm_gem_object(file_priv, dev, handle);
+	if (!samsung_gem_obj) {
+		DRM_ERROR("failed to get samsung_drm_get_obj.\n");
+		return -EINVAL;
+	}
+
+	/**
+	 * obj->refcount and obj->handle_count are decreased and
+	 * if both them are 0 then samsung_drm_gem_free_object()
+	 * would be called by callback to release resources.
+	 */
+	ret = drm_gem_handle_delete(file_priv, handle);
+	if (ret < 0) {
+		DRM_ERROR("failed to delete drm_gem_handle.\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+MODULE_AUTHOR("Inki Dae <inki.dae@samsung.com>");
+MODULE_DESCRIPTION("Samsung SoC DRM GEM Module");
+MODULE_LICENSE("GPL");
diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.h b/drivers/gpu/drm/samsung/samsung_drm_gem.h
new file mode 100644
index 0000000..dbc0ab2
--- /dev/null
+++ b/drivers/gpu/drm/samsung/samsung_drm_gem.h
@@ -0,0 +1,76 @@
+/* samsung_drm_gem.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authoer: Inki Dae <inki.dae@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_GEM_H_
+#define _SAMSUNG_DRM_GEM_H_
+
+#define to_samsung_gem_obj(x)	container_of(x,\
+			struct samsung_drm_gem_obj, base)
+
+/* create a new mm object and get a handle to it. */
+int samsung_drm_gem_create_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file);
+
+/* get buffer offset to map to user space. */
+int samsung_drm_gem_map_offset_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file);
+
+/* unmap a buffer from user space. */
+int samsung_drm_gem_munmap_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *file);
+
+/* initialize gem object. */
+int samsung_drm_gem_init_object(struct drm_gem_object *obj);
+
+/* free gem object. */
+void samsung_drm_gem_free_object(struct drm_gem_object *gem_obj);
+
+struct samsung_drm_gem_obj *
+		find_samsung_drm_gem_object(struct drm_file *file_priv,
+			struct drm_device *dev, unsigned int handle);
+
+/* create memory region for drm framebuffer. */
+int samsung_drm_gem_dumb_create(struct drm_file *file_priv,
+		struct drm_device *dev, struct drm_mode_create_dumb *args);
+
+/* map memory region for drm framebuffer to user space. */
+int samsung_drm_gem_dumb_map_offset(struct drm_file *file_priv,
+		struct drm_device *dev, uint32_t handle, uint64_t *offset);
+
+/* page fault handler and mmap fault address(virtual) to physical memory. */
+int samsung_drm_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
+
+/* mmap gem object. */
+int samsung_drm_gem_mmap_ioctl(struct drm_device *dev, void *data,
+		struct drm_file *filp);
+
+/* set vm_flags and we can change vm attribute to other here. */
+int samsung_drm_gem_mmap(struct file *filp, struct vm_area_struct *vma);
+
+/* destroy memory region allocated. */
+int samsung_drm_gem_dumb_destroy(struct drm_file *file_priv,
+		struct drm_device *dev, unsigned int handle);
+
+#endif
diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
new file mode 100644
index 0000000..05dc460
--- /dev/null
+++ b/include/drm/samsung_drm.h
@@ -0,0 +1,267 @@
+/* samsung_drm.h
+ *
+ * Copyright (c) 2011 Samsung Electronics Co., Ltd.
+ * Authors:
+ *	Inki Dae <inki.dae@samsung.com>
+ *	Joonyoung Shim <jy0922.shim@samsung.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _SAMSUNG_DRM_H_
+#define _SAMSUNG_DRM_H_
+
+#include "drm.h"
+
+#define MAX_CRTC	2
+
+struct drm_device;
+struct samsung_drm_overlay;
+struct drm_connector;
+
+enum samsung_drm_output_type {
+	SAMSUNG_DISPLAY_TYPE_NONE,
+	SAMSUNG_DISPLAY_TYPE_LCD,	/* RGB or CPU Interface. */
+	SAMSUNG_DISPLAY_TYPE_HDMI,	/* HDMI Interface. */
+};
+
+/**
+ * Samsung drm overlay ops structure.
+ *
+ * @mode_set: copy drm overlay info to hw specific overlay info.
+ * @commit: set hw specific overlay into to hw.
+ */
+struct samsung_drm_overlay_ops {
+	void (*mode_set)(struct device *subdrv_dev,
+			 struct samsung_drm_overlay *overlay);
+	void (*commit)(struct device *subdrv_dev);
+	void (*disable)(struct device *subdrv_dev);
+};
+
+/**
+ * Samsung drm common overlay structure.
+ *
+ * @win_num: window number.
+ * @offset_x: offset to x position.
+ * @offset_y: offset to y position.
+ * @pos_x: x position.
+ * @pos_y: y position.
+ * @width: window width.
+ * @height: window height.
+ * @bpp: bit per pixel.
+ * @paddr: physical memory address to this overlay.
+ * @vaddr: virtual memory addresss to this overlay.
+ * @buf_off: start offset of framebuffer to be displayed.
+ * @end_buf_off: end offset of framebuffer to be displayed.
+ * @buf_offsize: this value has result from
+ *			(framebuffer width - display width) * bpp.
+ * @line_size: line size to this overlay memory in bytes.
+ * @default_win: a window to be enabled.
+ * @color_key: color key on or off.
+ * @index_color: if using color key feature then this value would be used
+ *			as index color.
+ * @local_path: in case of lcd type, local path mode on or off.
+ * @transparency: transparency on or off.
+ * @activated: activated or not.
+ * @subdrv_dev: pointer to device object for subdrv device driver.
+ * @ops: pointer to samsung_drm_overlay_ops.
+ *
+ * this structure is common to Samsung SoC and would be copied
+ * to hardware specific overlay info.
+ */
+struct samsung_drm_overlay {
+	unsigned int offset_x;
+	unsigned int offset_y;
+	unsigned int width;
+	unsigned int height;
+	unsigned int paddr;
+	void __iomem *vaddr;
+	unsigned int buf_off;
+	unsigned int end_buf_off;
+	unsigned int buf_offsize;
+	unsigned int line_size;
+
+	bool default_win;
+	bool color_key;
+	unsigned int index_color;
+	bool local_path;
+	bool transparency;
+	bool activated;
+};
+
+/**
+ * Samsung DRM Display Structure.
+ *	- this structure is common to analog tv, digital tv and lcd panel.
+ *
+ * @dev: pointer to specific device object.
+ * @is_connected: check for that display is connected or not.
+ * @get_edid: get edid modes from display driver.
+ * @get_timing: get timing object from display driver.
+ * @check_timing: check if timing is valid or not.
+ * @power_on: display device on or off.
+ */
+struct samsung_drm_display {
+	unsigned int type;
+	bool (*is_connected)(void);
+	int (*get_edid)(struct drm_connector *connector, u8 *edid, int len);
+	void *(*get_timing)(void);
+	int (*check_timing)(void *timing);
+	int (*power_on)(int mode);
+};
+
+/**
+ * Samsung drm manager ops
+ *
+ * @mode_set: convert drm_display_mode to hw specific display mode and
+ *	      would be called by encoder->mode_set().
+ * @commit: set current hw specific display mode to hw.
+ * @enable_vblank: specific driver callback for enabling vblank interrupt.
+ * @disable_vblank: specific driver callback for disabling vblank interrupt.
+ */
+struct samsung_drm_manager_ops {
+	void (*mode_set)(struct device *subdrv_dev, void *mode);
+	void (*commit)(struct device *subdrv_dev);
+	int (*enable_vblank)(struct device *subdrv_dev);
+	void (*disable_vblank)(struct device *subdrv_dev);
+};
+
+/**
+ * Samsung drm common manager structure.
+ *
+ * @dev: pointer to device object for subdrv device driver.
+ * @ops: ops pointer to samsung drm common framebuffer.
+ *	 ops of fimd or hdmi driver should be set to this ones.
+ */
+struct samsung_drm_manager {
+	struct device *dev;
+	int pipe;
+	struct samsung_drm_manager_ops *ops;
+	struct samsung_drm_overlay_ops *overlay_ops;
+	struct samsung_drm_display *display;
+};
+
+/**
+ * Samsung drm private structure.
+ */
+struct samsung_drm_private {
+	struct drm_fb_helper *fb_helper;
+
+	/* for pageflip */
+	struct list_head pageflip_event_list;
+	bool pageflip_event;
+
+	struct drm_crtc *crtc[MAX_CRTC];
+
+	/* add some structures. */
+};
+
+struct samsung_drm_subdrv {
+	struct list_head list;
+
+	/* driver ops */
+	int (*probe)(struct drm_device *dev);
+	void (*remove)(struct drm_device *dev);
+
+	struct samsung_drm_manager manager;
+	struct drm_encoder *encoder;
+	struct drm_connector *connector;
+};
+
+int samsung_drm_device_register(struct drm_device *dev);
+void samsung_drm_device_unregister(struct drm_device *dev);
+int samsung_drm_subdrv_register(struct samsung_drm_subdrv *drm_subdrv);
+void samsung_drm_subdrv_unregister(struct samsung_drm_subdrv *drm_subdrv);
+
+/**
+ * User-desired buffer creation information structure.
+ *
+ * @usr_addr: an address allocated by user process and this address
+ *	would be mmapped to physical region by fault handler.
+ * @size: requested size for the object.
+ *	- this size value would be page-aligned internally.
+ * @flags: user request for setting memory type or cache attributes.
+ * @handle: returned handle for the object.
+ */
+struct drm_samsung_gem_create {
+	unsigned int usr_addr;
+	unsigned int size;
+	unsigned int flags;
+
+	unsigned int handle;
+};
+
+/**
+ * A structure for getting buffer offset.
+ *
+ * @handle: a pointer to gem object created.
+ * @offset: relatived offset value of the memory region allocated.
+ *	- this value should be set by user.
+ */
+struct drm_samsung_gem_map_off {
+	unsigned int handle;
+	uint64_t offset;
+};
+
+/**
+ * A structure for mapping buffer.
+ *
+ * @handle: a pointer to gem object created.
+ * @offset: relatived offset value of the memory region allocated.
+ *	- this value should be set by user.
+ * @size: memory size to be mapped.
+ * @mapped: user virtual address to be mapped.
+ */
+struct drm_samsung_gem_mmap {
+	unsigned int handle;
+	uint64_t offset;
+	unsigned int size;
+
+	unsigned int mapped;
+};
+
+#define DRM_SAMSUNG_GEM_CREATE		0x00
+#define DRM_SAMSUNG_GEM_MAP_OFFSET	0x01
+#define DRM_SAMSUNG_GEM_MMAP		0x02
+
+#define DRM_IOCTL_SAMSUNG_GEM_CREATE		DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_SAMSUNG_GEM_CREATE, struct drm_samsung_gem_create)
+
+#define DRM_IOCTL_SAMSUNG_GEM_MAP_OFFSET	DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_SAMSUNG_GEM_MAP_OFFSET, struct drm_samsung_gem_map_off)
+
+#define DRM_IOCTL_SAMSUNG_GEM_MMAP	DRM_IOWR(DRM_COMMAND_BASE + \
+		DRM_SAMSUNG_GEM_MMAP, struct drm_samsung_gem_mmap)
+
+/**
+ * Platform Specific Structure for DRM based FIMD.
+ *
+ * @timing: default video mode for initializing
+ * @default_win: default window layer number to be used for UI.
+ * @bpp: default bit per pixel.
+ */
+struct samsung_drm_fimd_pdata {
+	struct fb_videomode		timing;
+	u32				vidcon0;
+	u32				vidcon1;
+	unsigned int			default_win;
+	unsigned int			bpp;
+};
+
+#endif
-- 
1.7.0.4

^ permalink raw reply related	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-26 11:47 ` Inki Dae
  (?)
@ 2011-08-31  1:57   ` Rob Clark
  -1 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-08-31  1:57 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

Hi Inki,

Sorry for slightly overdue review.. it took a little while to go
through the whole thing

comments in-line below


On Fri, Aug 26, 2011 at 6:47 AM, Inki Dae <inki.dae@samsung.com> wrote:
> This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
> but we will add HDMI support also in the future.
>
> this patch is based on git repository below:
> git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> branch name: drm-next
> commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
>
> you can refer to our working repository below:
> http://git.infradead.org/users/kmpark/linux-2.6-samsung
> branch name: samsung-drm
>
> We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> based on Linux framebuffer) but couldn't so because lowlevel codes
> of s3c-fb.c are included internally and so FIMD module of this driver has
> its own lowlevel codes.
>
> We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> for buffer allocation. by using DMA API, we could use CMA later.
>
> Refer to this link for CMA(Continuous Memory Allocator):
> http://lkml.org/lkml/2011/7/20/45
>
> this driver supports only physically continuous memory(non-iommu).
>
> Links to previous versions of the patchset:
> v1: < https://lwn.net/Articles/454380/ >
> v2: < http://www.spinics.net/lists/kernel/msg1224275.html >
>
> Changelog v2:
> DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
>
>     this feature maps user address space to physical memory region
>     once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
>
> DRM: code clean and add exception codes.
>
> Changelog v3:
> DRM: Support multiple irq.
>
>     FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
>     this patch supports mutiple irq for Samsung SoC.
>
> DRM: Consider modularization.
>
>     each DRM, FIMD could be built as a module.
>
> DRM: Have indenpendent crtc object.
>
>     crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
>     created crtc could be attached to any encoder object.
>
> DRM: code clean and add exception codes.
>
> Signed-off-by: Inki Dae <inki.dae@samsung.com>
> Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
> Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
> Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
> ---
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> new file mode 100644
> index 0000000..563d07e
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
[snip]
> +static int lowlevel_buffer_allocate(struct drm_device *dev,
> +               struct samsung_drm_buf_entry *entry)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);

very minor point, but DRM_DEBUG_KMS() already includes the function
name.. not sure if __FILE__ is needed everywhere

> +
> +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> +       if (!entry->paddr) {
> +               DRM_ERROR("failed to allocate buffer.\n");
> +               return -ENOMEM;
> +       }
> +
> +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
> +                       (unsigned int)entry->vaddr, entry->paddr, entry->size);
> +
> +       return 0;
> +}
> +
[snip]

> diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> new file mode 100644
> index 0000000..d6a7e95
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
[snip]
> +/**
> + * samsung drm buffer entry structure.
> + *
> + * @paddr: physical address of allocated memory.
> + * @vaddr: kernel virtual address of allocated memory.
> + * @size: size of allocated memory.
> + */
> +struct samsung_drm_buf_entry {
> +       unsigned int paddr;
> +       void __iomem *vaddr;
> +       unsigned int size;
> +};

any reason not to combine this w/ samsung_drm_gem_obj to avoid extra
levels of indirection?  (This would let you collapse some of the
buffer allocation/deletion fxns into one level too.)

> +
> +/**
> + * samsung drm buffer structure.
> + *
> + * @entry: pointer to samsung drm buffer entry object.
> + * @flags: it means memory type to be alloated or cache attributes.
> + * @handle: pointer to specific buffer object.
> + * @id: unique id to specific buffer object.
> + *
> + * ps. this object would be transfered to user as kms_bo.handle so
> + *     user can access to memory through kms_bo.handle.
> + */
> +struct samsung_drm_gem_obj {
> +       struct drm_gem_object base;
> +       struct samsung_drm_buf_entry *entry;
> +       unsigned int flags;
> +
> +       unsigned int handle;
> +       unsigned int id;
> +};
> +
> +/* create new buffer object and memory region and add the object to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
> +               unsigned int size);
> +
> +/* allocate physical memory and add its object to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
> +               unsigned int size);
> +
> +/* remove allocated physical memory. */
> +int samsung_drm_buf_destroy(struct drm_device *dev,
> +               struct samsung_drm_gem_obj *in_obj);
> +
> +/* find object added to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device *dev,
> +               struct samsung_drm_gem_obj *in_obj, unsigned int paddr);

this doesn't appear to be used anywhere

> +
> +#endif
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> new file mode 100644
> index 0000000..987a629
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
[snip]
> +
> +/* convert samsung_video_timings to drm_display_mode */
> +static inline void
> +convert_to_display_mode(struct fb_videomode *timing,
> +                       struct drm_display_mode *mode)

I sort of prefer copy functions to copy right to left, Ie.

convert_foo(a, b)

copies from b to a.. somehow this feels more natural.  (Think a = b,
the receiving arg is on the left.. same as w/ memcpy()).  Same applies
w/ the inverse of this fxn.

> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       mode->clock = timing->pixclock / 1000;
> +
> +       mode->hdisplay = timing->xres;
> +       mode->hsync_start = mode->hdisplay + timing->left_margin;
> +       mode->hsync_end = mode->hsync_start + timing->hsync_len;
> +       mode->htotal = mode->hsync_end + timing->right_margin;
> +
> +       mode->vdisplay = timing->yres;
> +       mode->vsync_start = mode->vdisplay + timing->upper_margin;
> +       mode->vsync_end = mode->vsync_start + timing->vsync_len;
> +       mode->vtotal = mode->vsync_end + timing->lower_margin;
> +}
> +
> +/* convert drm_display_mode to samsung_video_timings */
> +static inline void
> +convert_to_video_timing(struct drm_display_mode *mode,
> +                       struct fb_videomode *timing)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       timing->pixclock = mode->clock * 1000;
> +
> +       timing->xres = mode->hdisplay;
> +       timing->left_margin = mode->hsync_start - mode->hdisplay;
> +       timing->hsync_len = mode->hsync_end - mode->hsync_start;
> +       timing->right_margin = mode->htotal - mode->hsync_end;
> +
> +       timing->yres = mode->vdisplay;
> +       timing->upper_margin = mode->vsync_start - mode->vdisplay;
> +       timing->vsync_len = mode->vsync_end - mode->vsync_start;
> +       timing->lower_margin = mode->vtotal - mode->vsync_end;
> +}
> +
> +static int samsung_drm_connector_get_modes(struct drm_connector *connector)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +       struct samsung_drm_display *display =
> +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> +       unsigned int count;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       if (display && display->get_edid) {
> +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> +               if (!edid) {
> +                       DRM_ERROR("failed to allocate edid\n");
> +                       return 0;
> +               }
> +
> +               display->get_edid(connector, edid, MAX_EDID);
> +
> +               drm_mode_connector_update_edid_property(connector, edid);
> +               count = drm_add_edid_modes(connector, edid);
> +
> +               kfree(connector->display_info.raw_edid);
> +               connector->display_info.raw_edid = edid;
> +       } else {
> +               struct drm_display_mode *mode = drm_mode_create(connector->dev);
> +               struct fb_videomode *timing;
> +
> +               if (display && display->get_timing)
> +                       timing = display->get_timing();

also seems a bit weird to not do: display->get_timings(display)..

maybe you don't have the case of multiple instances of the same
display time.. but maybe someday you need that.  (Maybe this is just
an artifact of how the API of your current lower layer driver is.. but
maybe now is a good time to think about those APIs)

Same comment about not passing the display instance ptr to some of the
other display fxn ptrs elsewhere in the code.

> +               else
> +                       return 0;
> +
> +               convert_to_display_mode(timing, mode);
> +
> +               mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
> +               drm_mode_set_name(mode);
> +               drm_mode_probed_add(connector, mode);
> +
> +               count = 1;
> +       }
> +
> +       return count;
> +}
> +
> +static int samsung_drm_connector_mode_valid(struct drm_connector *connector,
> +                                           struct drm_display_mode *mode)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +       struct samsung_drm_display *display =
> +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> +       struct fb_videomode timing;
> +       int ret = MODE_BAD;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       convert_to_video_timing(mode, &timing);
> +
> +       if (display && display->check_timing)
> +               if (!display->check_timing((void *)&timing))
> +                       ret = MODE_OK;
> +
> +       return ret;
> +}
> +
> +struct drm_encoder *samsung_drm_best_encoder(struct drm_connector *connector)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +       return to_samsung_connector(connector)->encoder;
> +}
> +
> +static struct drm_connector_helper_funcs samsung_connector_helper_funcs = {
> +       .get_modes      = samsung_drm_connector_get_modes,
> +       .mode_valid     = samsung_drm_connector_mode_valid,
> +       .best_encoder   = samsung_drm_best_encoder,
> +};
> +
> +/* get detection status of display device. */
> +static enum drm_connector_status
> +samsung_drm_connector_detect(struct drm_connector *connector, bool force)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +       struct samsung_drm_display *display =
> +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> +       unsigned int ret = connector_status_unknown;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       if (display && display->is_connected) {
> +               if (display->is_connected())
> +                       ret = connector_status_connected;
> +               else
> +                       ret = connector_status_disconnected;
> +       }
> +
> +       return ret;
> +}
> +
> +static void samsung_drm_connector_destroy(struct drm_connector *connector)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       drm_sysfs_connector_remove(connector);
> +       drm_connector_cleanup(connector);
> +       connector->dev->mode_config.num_connector--;

I wonder if num_connector-- should be in drm_connector_cleanup()..

I missed this in OMAP driver, but it seems that none of the other drm
drivers are directly decrementing this.. and it would seem more
symmetric for it to be in drm_connector_cleanup().  But would be
interesting to see what others think.  (I guess no one has really
dealt w/ dynamic creation/deletion of connectors yet?)

> +       kfree(samsung_connector);
> +}
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> new file mode 100644
> index 0000000..1c5a70a
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
[snip]
> +
> +struct samsung_drm_crtc {
> +       struct drm_crtc                 drm_crtc;
> +       struct samsung_drm_overlay      overlay;


I guess you are looking fwd to drm_plane support too :-)


> +       unsigned int                    pipe;
> +};
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> new file mode 100644
> index 0000000..42096eb
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> @@ -0,0 +1,248 @@
> +/*
> + * Copyright (c) 2011 Samsung Electronics Co., Ltd.
> + * Authors:
> + *     Inki Dae <inki.dae@samsung.com>
> + *     Joonyoung Shim <jy0922.shim@samsung.com>
> + *     SeungWoo Kim <sw0312.kim@samsung.com>
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
> + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
> + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
> + * OTHER DEALINGS IN THE SOFTWARE.
> + */
> +
> +#include "drmP.h"
> +#include "drm_crtc.h"
> +#include "drm_crtc_helper.h"
> +
> +#include "samsung_drm_fb.h"
> +#include "samsung_drm_buf.h"
> +#include "samsung_drm_gem.h"
> +
> +#define to_samsung_fb(x)       container_of(x, struct samsung_drm_fb, fb)
> +
> +struct samsung_drm_fb {
> +       struct drm_framebuffer          fb;
> +       struct drm_file                 *file_priv;
> +       struct samsung_drm_gem_obj      *samsung_gem_obj;
> +       unsigned int                    is_default;
> +
> +       /* samsung gem object handle. */
> +       unsigned int                    gem_handle;
> +       /* unique id to buffer object. */
> +       unsigned int                    id;
> +
> +       unsigned int                    fb_size;
> +       unsigned long                   paddr;
> +       void __iomem                    *vaddr;
> +};
> +
> +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> +{
> +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> +       int ret;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       drm_framebuffer_cleanup(fb);
> +
> +       if (samsung_fb->is_default) {
> +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> +                               samsung_fb->gem_handle);

why not keep the gem buffer ptr, and do something like:

  drm_gem_object_unreference_unlocked(samsung_fb->bo)..

this way, you get the right behavior if someone somewhere else took a
ref to the gem buffer object?  And it avoids needing to keep the
file_priv ptr in the fb (which seems a bit strange)


> +               if (ret < 0)
> +                       DRM_ERROR("failed to delete drm_gem_handle.\n");
> +       }
> +
> +       kfree(samsung_fb);
> +}
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> new file mode 100644
> index 0000000..1e167a6
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
[snip]
> +
> +static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
> +}

fwiw, candidate to move to drm_gem helper fxn (I have same in my
driver.. but this wasn't included in my earlier patch to add a couple
common drm_gem helper fxns like create/free_mmap_offset)

> +
> +/**
> + * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for an object
> + * @obj: obj in question
> + *
> + * GEM memory mapping works by handing back to userspace a fake mmap offset
> + * it can use in a subsequent mmap(2) call.  The DRM core code then looks
> + * up the object based on the offset and sets up the various memory mapping
> + * structures.
> + *
> + * This routine allocates and attaches a fake offset for @obj.
> + */
> +static int
> +samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)

some of these other fxns you might at least want to include a comment
to replace w/ drm_gem helpers once that patch lands.. your's would be
the 4th driver w/ nearly identical code (well, if you can count
non-mainline drivers) ;-)

> +{
> +       struct drm_device *dev = obj->dev;
> +       struct drm_gem_mm *mm = dev->mm_private;
> +       struct drm_map_list *list;
> +       struct drm_local_map *map;
> +       int ret = 0;
> +
> +       /* Set the object up for mmap'ing */
> +       list = &obj->map_list;
> +       list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
> +       if (!list->map)
> +               return -ENOMEM;
> +
> +       map = list->map;
> +       map->type = _DRM_GEM;
> +       map->size = obj->size;
> +       map->handle = obj;
> +
> +       /* Get a DRM GEM mmap offset allocated... */
> +       list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
> +                                                   obj->size / PAGE_SIZE,
> +                                                   0, 0);
> +       if (!list->file_offset_node) {
> +               DRM_ERROR("failed to allocate offset for bo %d\n",
> +                         obj->name);
> +               ret = -ENOSPC;
> +               goto out_free_list;
> +       }
> +
> +       list->file_offset_node = drm_mm_get_block(list->file_offset_node,
> +                                                 obj->size / PAGE_SIZE,
> +                                                 0);
> +       if (!list->file_offset_node) {
> +               ret = -ENOMEM;
> +               goto out_free_list;
> +       }
> +
> +       list->hash.key = list->file_offset_node->start;
> +       ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
> +       if (ret) {
> +               DRM_ERROR("failed to add to map hash\n");
> +               goto out_free_mm;
> +       }
> +
> +       return 0;
> +
> +out_free_mm:
> +       drm_mm_put_block(list->file_offset_node);
> +out_free_list:
> +       kfree(list->map);
> +       list->map = NULL;
> +
> +       return ret;
> +}
> +
> +static void
> +samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
> +{
> +       struct drm_device *dev = obj->dev;
> +       struct drm_gem_mm *mm = dev->mm_private;
> +       struct drm_map_list *list = &obj->map_list;
> +
> +       drm_ht_remove_item(&mm->offset_hash, &list->hash);
> +       drm_mm_put_block(list->file_offset_node);
> +       kfree(list->map);
> +       list->map = NULL;
> +}
[snip]
> +struct samsung_drm_gem_obj *
> +               find_samsung_drm_gem_object(struct drm_file *file_priv,
> +                       struct drm_device *dev, unsigned int handle)
> +{
> +       struct drm_gem_object *gem_obj;
> +
> +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> +       if (!gem_obj) {
> +               DRM_LOG_KMS("a invalid gem object not registered to lookup.\n");
> +               return NULL;
> +       }
> +
> +       /**
> +        * unreference refcount of the gem object.
> +        * at drm_gem_object_lookup(), the gem object was referenced.
> +        */
> +       drm_gem_object_unreference(gem_obj);

this doesn't seem right, to drop the reference before you use the
buffer elsewhere..

> +       return to_samsung_gem_obj(gem_obj);
> +}
> +
[snip]
> diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
> new file mode 100644
> index 0000000..05dc460
> --- /dev/null
> +++ b/include/drm/samsung_drm.h
[snip]
> +/**
> + * User-desired buffer creation information structure.
> + *
> + * @usr_addr: an address allocated by user process and this address
> + *     would be mmapped to physical region by fault handler.
> + * @size: requested size for the object.
> + *     - this size value would be page-aligned internally.
> + * @flags: user request for setting memory type or cache attributes.
> + * @handle: returned handle for the object.
> + */
> +struct drm_samsung_gem_create {
> +       unsigned int usr_addr;

usr_addr seems dangerous.. and unused?  Maybe can be removed?

also, I sort of prefer using stdint types for userspace<->kernel ioctl structs.

> +       unsigned int size;
> +       unsigned int flags;
> +
> +       unsigned int handle;
> +};
> +
> +/**
> + * A structure for getting buffer offset.
> + *
> + * @handle: a pointer to gem object created.
> + * @offset: relatived offset value of the memory region allocated.
> + *     - this value should be set by user.
> + */
> +struct drm_samsung_gem_map_off {
> +       unsigned int handle;
> +       uint64_t offset;
> +};
> +
> +/**
> + * A structure for mapping buffer.
> + *
> + * @handle: a pointer to gem object created.
> + * @offset: relatived offset value of the memory region allocated.
> + *     - this value should be set by user.
> + * @size: memory size to be mapped.
> + * @mapped: user virtual address to be mapped.
> + */
> +struct drm_samsung_gem_mmap {
> +       unsigned int handle;
> +       uint64_t offset;
> +       unsigned int size;
> +
> +       unsigned int mapped;

I guess this should be obtained by passing offset back to mmap() syscall


BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  1:57   ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-08-31  1:57 UTC (permalink / raw)
  To: linux-arm-kernel

Hi Inki,

Sorry for slightly overdue review.. it took a little while to go
through the whole thing

comments in-line below


On Fri, Aug 26, 2011 at 6:47 AM, Inki Dae <inki.dae@samsung.com> wrote:
> This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
> but we will add HDMI support also in the future.
>
> this patch is based on git repository below:
> git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> branch name: drm-next
> commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
>
> you can refer to our working repository below:
> http://git.infradead.org/users/kmpark/linux-2.6-samsung
> branch name: samsung-drm
>
> We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> based on Linux framebuffer) but couldn't so because lowlevel codes
> of s3c-fb.c are included internally and so FIMD module of this driver has
> its own lowlevel codes.
>
> We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> for buffer allocation. by using DMA API, we could use CMA later.
>
> Refer to this link for CMA(Continuous Memory Allocator):
> http://lkml.org/lkml/2011/7/20/45
>
> this driver supports only physically continuous memory(non-iommu).
>
> Links to previous versions of the patchset:
> v1: < https://lwn.net/Articles/454380/ >
> v2: < http://www.spinics.net/lists/kernel/msg1224275.html >
>
> Changelog v2:
> DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
>
> ? ? this feature maps user address space to physical memory region
> ? ? once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
>
> DRM: code clean and add exception codes.
>
> Changelog v3:
> DRM: Support multiple irq.
>
> ? ? FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
> ? ? this patch supports mutiple irq for Samsung SoC.
>
> DRM: Consider modularization.
>
> ? ? each DRM, FIMD could be built as a module.
>
> DRM: Have indenpendent crtc object.
>
> ? ? crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
> ? ? created crtc could be attached to any encoder object.
>
> DRM: code clean and add exception codes.
>
> Signed-off-by: Inki Dae <inki.dae@samsung.com>
> Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
> Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
> Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
> ---
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> new file mode 100644
> index 0000000..563d07e
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
[snip]
> +static int lowlevel_buffer_allocate(struct drm_device *dev,
> + ? ? ? ? ? ? ? struct samsung_drm_buf_entry *entry)
> +{
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);

very minor point, but DRM_DEBUG_KMS() already includes the function
name.. not sure if __FILE__ is needed everywhere

> +
> + ? ? ? entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> + ? ? ? ? ? ? ? ? ? ? ? (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> + ? ? ? if (!entry->paddr) {
> + ? ? ? ? ? ? ? DRM_ERROR("failed to allocate buffer.\n");
> + ? ? ? ? ? ? ? return -ENOMEM;
> + ? ? ? }
> +
> + ? ? ? DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
> + ? ? ? ? ? ? ? ? ? ? ? (unsigned int)entry->vaddr, entry->paddr, entry->size);
> +
> + ? ? ? return 0;
> +}
> +
[snip]

> diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> new file mode 100644
> index 0000000..d6a7e95
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
[snip]
> +/**
> + * samsung drm buffer entry structure.
> + *
> + * @paddr: physical address of allocated memory.
> + * @vaddr: kernel virtual address of allocated memory.
> + * @size: size of allocated memory.
> + */
> +struct samsung_drm_buf_entry {
> + ? ? ? unsigned int paddr;
> + ? ? ? void __iomem *vaddr;
> + ? ? ? unsigned int size;
> +};

any reason not to combine this w/ samsung_drm_gem_obj to avoid extra
levels of indirection?  (This would let you collapse some of the
buffer allocation/deletion fxns into one level too.)

> +
> +/**
> + * samsung drm buffer structure.
> + *
> + * @entry: pointer to samsung drm buffer entry object.
> + * @flags: it means memory type to be alloated or cache attributes.
> + * @handle: pointer to specific buffer object.
> + * @id: unique id to specific buffer object.
> + *
> + * ps. this object would be transfered to user as kms_bo.handle so
> + * ? ? user can access to memory through kms_bo.handle.
> + */
> +struct samsung_drm_gem_obj {
> + ? ? ? struct drm_gem_object base;
> + ? ? ? struct samsung_drm_buf_entry *entry;
> + ? ? ? unsigned int flags;
> +
> + ? ? ? unsigned int handle;
> + ? ? ? unsigned int id;
> +};
> +
> +/* create new buffer object and memory region and add the object to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
> + ? ? ? ? ? ? ? unsigned int size);
> +
> +/* allocate physical memory and add its object to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
> + ? ? ? ? ? ? ? unsigned int size);
> +
> +/* remove allocated physical memory. */
> +int samsung_drm_buf_destroy(struct drm_device *dev,
> + ? ? ? ? ? ? ? struct samsung_drm_gem_obj *in_obj);
> +
> +/* find object added to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device *dev,
> + ? ? ? ? ? ? ? struct samsung_drm_gem_obj *in_obj, unsigned int paddr);

this doesn't appear to be used anywhere

> +
> +#endif
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> new file mode 100644
> index 0000000..987a629
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
[snip]
> +
> +/* convert samsung_video_timings to drm_display_mode */
> +static inline void
> +convert_to_display_mode(struct fb_videomode *timing,
> + ? ? ? ? ? ? ? ? ? ? ? struct drm_display_mode *mode)

I sort of prefer copy functions to copy right to left, Ie.

convert_foo(a, b)

copies from b to a.. somehow this feels more natural.  (Think a = b,
the receiving arg is on the left.. same as w/ memcpy()).  Same applies
w/ the inverse of this fxn.

> +{
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? mode->clock = timing->pixclock / 1000;
> +
> + ? ? ? mode->hdisplay = timing->xres;
> + ? ? ? mode->hsync_start = mode->hdisplay + timing->left_margin;
> + ? ? ? mode->hsync_end = mode->hsync_start + timing->hsync_len;
> + ? ? ? mode->htotal = mode->hsync_end + timing->right_margin;
> +
> + ? ? ? mode->vdisplay = timing->yres;
> + ? ? ? mode->vsync_start = mode->vdisplay + timing->upper_margin;
> + ? ? ? mode->vsync_end = mode->vsync_start + timing->vsync_len;
> + ? ? ? mode->vtotal = mode->vsync_end + timing->lower_margin;
> +}
> +
> +/* convert drm_display_mode to samsung_video_timings */
> +static inline void
> +convert_to_video_timing(struct drm_display_mode *mode,
> + ? ? ? ? ? ? ? ? ? ? ? struct fb_videomode *timing)
> +{
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? timing->pixclock = mode->clock * 1000;
> +
> + ? ? ? timing->xres = mode->hdisplay;
> + ? ? ? timing->left_margin = mode->hsync_start - mode->hdisplay;
> + ? ? ? timing->hsync_len = mode->hsync_end - mode->hsync_start;
> + ? ? ? timing->right_margin = mode->htotal - mode->hsync_end;
> +
> + ? ? ? timing->yres = mode->vdisplay;
> + ? ? ? timing->upper_margin = mode->vsync_start - mode->vdisplay;
> + ? ? ? timing->vsync_len = mode->vsync_end - mode->vsync_start;
> + ? ? ? timing->lower_margin = mode->vtotal - mode->vsync_end;
> +}
> +
> +static int samsung_drm_connector_get_modes(struct drm_connector *connector)
> +{
> + ? ? ? struct samsung_drm_connector *samsung_connector =
> + ? ? ? ? ? ? ? to_samsung_connector(connector);
> + ? ? ? struct samsung_drm_display *display =
> + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)->display;
> + ? ? ? unsigned int count;
> +
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? if (display && display->get_edid) {
> + ? ? ? ? ? ? ? void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> + ? ? ? ? ? ? ? if (!edid) {
> + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to allocate edid\n");
> + ? ? ? ? ? ? ? ? ? ? ? return 0;
> + ? ? ? ? ? ? ? }
> +
> + ? ? ? ? ? ? ? display->get_edid(connector, edid, MAX_EDID);
> +
> + ? ? ? ? ? ? ? drm_mode_connector_update_edid_property(connector, edid);
> + ? ? ? ? ? ? ? count = drm_add_edid_modes(connector, edid);
> +
> + ? ? ? ? ? ? ? kfree(connector->display_info.raw_edid);
> + ? ? ? ? ? ? ? connector->display_info.raw_edid = edid;
> + ? ? ? } else {
> + ? ? ? ? ? ? ? struct drm_display_mode *mode = drm_mode_create(connector->dev);
> + ? ? ? ? ? ? ? struct fb_videomode *timing;
> +
> + ? ? ? ? ? ? ? if (display && display->get_timing)
> + ? ? ? ? ? ? ? ? ? ? ? timing = display->get_timing();

also seems a bit weird to not do: display->get_timings(display)..

maybe you don't have the case of multiple instances of the same
display time.. but maybe someday you need that.  (Maybe this is just
an artifact of how the API of your current lower layer driver is.. but
maybe now is a good time to think about those APIs)

Same comment about not passing the display instance ptr to some of the
other display fxn ptrs elsewhere in the code.

> + ? ? ? ? ? ? ? else
> + ? ? ? ? ? ? ? ? ? ? ? return 0;
> +
> + ? ? ? ? ? ? ? convert_to_display_mode(timing, mode);
> +
> + ? ? ? ? ? ? ? mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
> + ? ? ? ? ? ? ? drm_mode_set_name(mode);
> + ? ? ? ? ? ? ? drm_mode_probed_add(connector, mode);
> +
> + ? ? ? ? ? ? ? count = 1;
> + ? ? ? }
> +
> + ? ? ? return count;
> +}
> +
> +static int samsung_drm_connector_mode_valid(struct drm_connector *connector,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct drm_display_mode *mode)
> +{
> + ? ? ? struct samsung_drm_connector *samsung_connector =
> + ? ? ? ? ? ? ? to_samsung_connector(connector);
> + ? ? ? struct samsung_drm_display *display =
> + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)->display;
> + ? ? ? struct fb_videomode timing;
> + ? ? ? int ret = MODE_BAD;
> +
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? convert_to_video_timing(mode, &timing);
> +
> + ? ? ? if (display && display->check_timing)
> + ? ? ? ? ? ? ? if (!display->check_timing((void *)&timing))
> + ? ? ? ? ? ? ? ? ? ? ? ret = MODE_OK;
> +
> + ? ? ? return ret;
> +}
> +
> +struct drm_encoder *samsung_drm_best_encoder(struct drm_connector *connector)
> +{
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> + ? ? ? return to_samsung_connector(connector)->encoder;
> +}
> +
> +static struct drm_connector_helper_funcs samsung_connector_helper_funcs = {
> + ? ? ? .get_modes ? ? ?= samsung_drm_connector_get_modes,
> + ? ? ? .mode_valid ? ? = samsung_drm_connector_mode_valid,
> + ? ? ? .best_encoder ? = samsung_drm_best_encoder,
> +};
> +
> +/* get detection status of display device. */
> +static enum drm_connector_status
> +samsung_drm_connector_detect(struct drm_connector *connector, bool force)
> +{
> + ? ? ? struct samsung_drm_connector *samsung_connector =
> + ? ? ? ? ? ? ? to_samsung_connector(connector);
> + ? ? ? struct samsung_drm_display *display =
> + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)->display;
> + ? ? ? unsigned int ret = connector_status_unknown;
> +
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? if (display && display->is_connected) {
> + ? ? ? ? ? ? ? if (display->is_connected())
> + ? ? ? ? ? ? ? ? ? ? ? ret = connector_status_connected;
> + ? ? ? ? ? ? ? else
> + ? ? ? ? ? ? ? ? ? ? ? ret = connector_status_disconnected;
> + ? ? ? }
> +
> + ? ? ? return ret;
> +}
> +
> +static void samsung_drm_connector_destroy(struct drm_connector *connector)
> +{
> + ? ? ? struct samsung_drm_connector *samsung_connector =
> + ? ? ? ? ? ? ? to_samsung_connector(connector);
> +
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? drm_sysfs_connector_remove(connector);
> + ? ? ? drm_connector_cleanup(connector);
> + ? ? ? connector->dev->mode_config.num_connector--;

I wonder if num_connector-- should be in drm_connector_cleanup()..

I missed this in OMAP driver, but it seems that none of the other drm
drivers are directly decrementing this.. and it would seem more
symmetric for it to be in drm_connector_cleanup().  But would be
interesting to see what others think.  (I guess no one has really
dealt w/ dynamic creation/deletion of connectors yet?)

> + ? ? ? kfree(samsung_connector);
> +}
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> new file mode 100644
> index 0000000..1c5a70a
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
[snip]
> +
> +struct samsung_drm_crtc {
> + ? ? ? struct drm_crtc ? ? ? ? ? ? ? ? drm_crtc;
> + ? ? ? struct samsung_drm_overlay ? ? ?overlay;


I guess you are looking fwd to drm_plane support too :-)


> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?pipe;
> +};
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> new file mode 100644
> index 0000000..42096eb
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> @@ -0,0 +1,248 @@
> +/*
> + * Copyright (c) 2011 Samsung Electronics Co., Ltd.
> + * Authors:
> + * ? ? Inki Dae <inki.dae@samsung.com>
> + * ? ? Joonyoung Shim <jy0922.shim@samsung.com>
> + * ? ? SeungWoo Kim <sw0312.kim@samsung.com>
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. ?IN NO EVENT SHALL
> + * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
> + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
> + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
> + * OTHER DEALINGS IN THE SOFTWARE.
> + */
> +
> +#include "drmP.h"
> +#include "drm_crtc.h"
> +#include "drm_crtc_helper.h"
> +
> +#include "samsung_drm_fb.h"
> +#include "samsung_drm_buf.h"
> +#include "samsung_drm_gem.h"
> +
> +#define to_samsung_fb(x) ? ? ? container_of(x, struct samsung_drm_fb, fb)
> +
> +struct samsung_drm_fb {
> + ? ? ? struct drm_framebuffer ? ? ? ? ?fb;
> + ? ? ? struct drm_file ? ? ? ? ? ? ? ? *file_priv;
> + ? ? ? struct samsung_drm_gem_obj ? ? ?*samsung_gem_obj;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?is_default;
> +
> + ? ? ? /* samsung gem object handle. */
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?gem_handle;
> + ? ? ? /* unique id to buffer object. */
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?id;
> +
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?fb_size;
> + ? ? ? unsigned long ? ? ? ? ? ? ? ? ? paddr;
> + ? ? ? void __iomem ? ? ? ? ? ? ? ? ? ?*vaddr;
> +};
> +
> +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> +{
> + ? ? ? struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> + ? ? ? int ret;
> +
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? drm_framebuffer_cleanup(fb);
> +
> + ? ? ? if (samsung_fb->is_default) {
> + ? ? ? ? ? ? ? ret = drm_gem_handle_delete(samsung_fb->file_priv,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? samsung_fb->gem_handle);

why not keep the gem buffer ptr, and do something like:

  drm_gem_object_unreference_unlocked(samsung_fb->bo)..

this way, you get the right behavior if someone somewhere else took a
ref to the gem buffer object?  And it avoids needing to keep the
file_priv ptr in the fb (which seems a bit strange)


> + ? ? ? ? ? ? ? if (ret < 0)
> + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to delete drm_gem_handle.\n");
> + ? ? ? }
> +
> + ? ? ? kfree(samsung_fb);
> +}
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> new file mode 100644
> index 0000000..1e167a6
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
[snip]
> +
> +static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
> +{
> + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> + ? ? ? return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
> +}

fwiw, candidate to move to drm_gem helper fxn (I have same in my
driver.. but this wasn't included in my earlier patch to add a couple
common drm_gem helper fxns like create/free_mmap_offset)

> +
> +/**
> + * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for an object
> + * @obj: obj in question
> + *
> + * GEM memory mapping works by handing back to userspace a fake mmap offset
> + * it can use in a subsequent mmap(2) call. ?The DRM core code then looks
> + * up the object based on the offset and sets up the various memory mapping
> + * structures.
> + *
> + * This routine allocates and attaches a fake offset for @obj.
> + */
> +static int
> +samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)

some of these other fxns you might at least want to include a comment
to replace w/ drm_gem helpers once that patch lands.. your's would be
the 4th driver w/ nearly identical code (well, if you can count
non-mainline drivers) ;-)

> +{
> + ? ? ? struct drm_device *dev = obj->dev;
> + ? ? ? struct drm_gem_mm *mm = dev->mm_private;
> + ? ? ? struct drm_map_list *list;
> + ? ? ? struct drm_local_map *map;
> + ? ? ? int ret = 0;
> +
> + ? ? ? /* Set the object up for mmap'ing */
> + ? ? ? list = &obj->map_list;
> + ? ? ? list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
> + ? ? ? if (!list->map)
> + ? ? ? ? ? ? ? return -ENOMEM;
> +
> + ? ? ? map = list->map;
> + ? ? ? map->type = _DRM_GEM;
> + ? ? ? map->size = obj->size;
> + ? ? ? map->handle = obj;
> +
> + ? ? ? /* Get a DRM GEM mmap offset allocated... */
> + ? ? ? list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? obj->size / PAGE_SIZE,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 0, 0);
> + ? ? ? if (!list->file_offset_node) {
> + ? ? ? ? ? ? ? DRM_ERROR("failed to allocate offset for bo %d\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? obj->name);
> + ? ? ? ? ? ? ? ret = -ENOSPC;
> + ? ? ? ? ? ? ? goto out_free_list;
> + ? ? ? }
> +
> + ? ? ? list->file_offset_node = drm_mm_get_block(list->file_offset_node,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? obj->size / PAGE_SIZE,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 0);
> + ? ? ? if (!list->file_offset_node) {
> + ? ? ? ? ? ? ? ret = -ENOMEM;
> + ? ? ? ? ? ? ? goto out_free_list;
> + ? ? ? }
> +
> + ? ? ? list->hash.key = list->file_offset_node->start;
> + ? ? ? ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
> + ? ? ? if (ret) {
> + ? ? ? ? ? ? ? DRM_ERROR("failed to add to map hash\n");
> + ? ? ? ? ? ? ? goto out_free_mm;
> + ? ? ? }
> +
> + ? ? ? return 0;
> +
> +out_free_mm:
> + ? ? ? drm_mm_put_block(list->file_offset_node);
> +out_free_list:
> + ? ? ? kfree(list->map);
> + ? ? ? list->map = NULL;
> +
> + ? ? ? return ret;
> +}
> +
> +static void
> +samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
> +{
> + ? ? ? struct drm_device *dev = obj->dev;
> + ? ? ? struct drm_gem_mm *mm = dev->mm_private;
> + ? ? ? struct drm_map_list *list = &obj->map_list;
> +
> + ? ? ? drm_ht_remove_item(&mm->offset_hash, &list->hash);
> + ? ? ? drm_mm_put_block(list->file_offset_node);
> + ? ? ? kfree(list->map);
> + ? ? ? list->map = NULL;
> +}
[snip]
> +struct samsung_drm_gem_obj *
> + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file *file_priv,
> + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int handle)
> +{
> + ? ? ? struct drm_gem_object *gem_obj;
> +
> + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> + ? ? ? if (!gem_obj) {
> + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered to lookup.\n");
> + ? ? ? ? ? ? ? return NULL;
> + ? ? ? }
> +
> + ? ? ? /**
> + ? ? ? ?* unreference refcount of the gem object.
> + ? ? ? ?* at drm_gem_object_lookup(), the gem object was referenced.
> + ? ? ? ?*/
> + ? ? ? drm_gem_object_unreference(gem_obj);

this doesn't seem right, to drop the reference before you use the
buffer elsewhere..

> + ? ? ? return to_samsung_gem_obj(gem_obj);
> +}
> +
[snip]
> diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
> new file mode 100644
> index 0000000..05dc460
> --- /dev/null
> +++ b/include/drm/samsung_drm.h
[snip]
> +/**
> + * User-desired buffer creation information structure.
> + *
> + * @usr_addr: an address allocated by user process and this address
> + * ? ? would be mmapped to physical region by fault handler.
> + * @size: requested size for the object.
> + * ? ? - this size value would be page-aligned internally.
> + * @flags: user request for setting memory type or cache attributes.
> + * @handle: returned handle for the object.
> + */
> +struct drm_samsung_gem_create {
> + ? ? ? unsigned int usr_addr;

usr_addr seems dangerous.. and unused?  Maybe can be removed?

also, I sort of prefer using stdint types for userspace<->kernel ioctl structs.

> + ? ? ? unsigned int size;
> + ? ? ? unsigned int flags;
> +
> + ? ? ? unsigned int handle;
> +};
> +
> +/**
> + * A structure for getting buffer offset.
> + *
> + * @handle: a pointer to gem object created.
> + * @offset: relatived offset value of the memory region allocated.
> + * ? ? - this value should be set by user.
> + */
> +struct drm_samsung_gem_map_off {
> + ? ? ? unsigned int handle;
> + ? ? ? uint64_t offset;
> +};
> +
> +/**
> + * A structure for mapping buffer.
> + *
> + * @handle: a pointer to gem object created.
> + * @offset: relatived offset value of the memory region allocated.
> + * ? ? - this value should be set by user.
> + * @size: memory size to be mapped.
> + * @mapped: user virtual address to be mapped.
> + */
> +struct drm_samsung_gem_mmap {
> + ? ? ? unsigned int handle;
> + ? ? ? uint64_t offset;
> + ? ? ? unsigned int size;
> +
> + ? ? ? unsigned int mapped;

I guess this should be obtained by passing offset back to mmap() syscall


BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  1:57   ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-08-31  1:57 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

Hi Inki,

Sorry for slightly overdue review.. it took a little while to go
through the whole thing

comments in-line below


On Fri, Aug 26, 2011 at 6:47 AM, Inki Dae <inki.dae@samsung.com> wrote:
> This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
> but we will add HDMI support also in the future.
>
> this patch is based on git repository below:
> git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> branch name: drm-next
> commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
>
> you can refer to our working repository below:
> http://git.infradead.org/users/kmpark/linux-2.6-samsung
> branch name: samsung-drm
>
> We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> based on Linux framebuffer) but couldn't so because lowlevel codes
> of s3c-fb.c are included internally and so FIMD module of this driver has
> its own lowlevel codes.
>
> We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> for buffer allocation. by using DMA API, we could use CMA later.
>
> Refer to this link for CMA(Continuous Memory Allocator):
> http://lkml.org/lkml/2011/7/20/45
>
> this driver supports only physically continuous memory(non-iommu).
>
> Links to previous versions of the patchset:
> v1: < https://lwn.net/Articles/454380/ >
> v2: < http://www.spinics.net/lists/kernel/msg1224275.html >
>
> Changelog v2:
> DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
>
>     this feature maps user address space to physical memory region
>     once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
>
> DRM: code clean and add exception codes.
>
> Changelog v3:
> DRM: Support multiple irq.
>
>     FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
>     this patch supports mutiple irq for Samsung SoC.
>
> DRM: Consider modularization.
>
>     each DRM, FIMD could be built as a module.
>
> DRM: Have indenpendent crtc object.
>
>     crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
>     created crtc could be attached to any encoder object.
>
> DRM: code clean and add exception codes.
>
> Signed-off-by: Inki Dae <inki.dae@samsung.com>
> Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
> Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
> Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
> ---
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> new file mode 100644
> index 0000000..563d07e
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
[snip]
> +static int lowlevel_buffer_allocate(struct drm_device *dev,
> +               struct samsung_drm_buf_entry *entry)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);

very minor point, but DRM_DEBUG_KMS() already includes the function
name.. not sure if __FILE__ is needed everywhere

> +
> +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> +       if (!entry->paddr) {
> +               DRM_ERROR("failed to allocate buffer.\n");
> +               return -ENOMEM;
> +       }
> +
> +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
> +                       (unsigned int)entry->vaddr, entry->paddr, entry->size);
> +
> +       return 0;
> +}
> +
[snip]

> diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> new file mode 100644
> index 0000000..d6a7e95
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
[snip]
> +/**
> + * samsung drm buffer entry structure.
> + *
> + * @paddr: physical address of allocated memory.
> + * @vaddr: kernel virtual address of allocated memory.
> + * @size: size of allocated memory.
> + */
> +struct samsung_drm_buf_entry {
> +       unsigned int paddr;
> +       void __iomem *vaddr;
> +       unsigned int size;
> +};

any reason not to combine this w/ samsung_drm_gem_obj to avoid extra
levels of indirection?  (This would let you collapse some of the
buffer allocation/deletion fxns into one level too.)

> +
> +/**
> + * samsung drm buffer structure.
> + *
> + * @entry: pointer to samsung drm buffer entry object.
> + * @flags: it means memory type to be alloated or cache attributes.
> + * @handle: pointer to specific buffer object.
> + * @id: unique id to specific buffer object.
> + *
> + * ps. this object would be transfered to user as kms_bo.handle so
> + *     user can access to memory through kms_bo.handle.
> + */
> +struct samsung_drm_gem_obj {
> +       struct drm_gem_object base;
> +       struct samsung_drm_buf_entry *entry;
> +       unsigned int flags;
> +
> +       unsigned int handle;
> +       unsigned int id;
> +};
> +
> +/* create new buffer object and memory region and add the object to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
> +               unsigned int size);
> +
> +/* allocate physical memory and add its object to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device *dev,
> +               unsigned int size);
> +
> +/* remove allocated physical memory. */
> +int samsung_drm_buf_destroy(struct drm_device *dev,
> +               struct samsung_drm_gem_obj *in_obj);
> +
> +/* find object added to list. */
> +struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device *dev,
> +               struct samsung_drm_gem_obj *in_obj, unsigned int paddr);

this doesn't appear to be used anywhere

> +
> +#endif
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> new file mode 100644
> index 0000000..987a629
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
[snip]
> +
> +/* convert samsung_video_timings to drm_display_mode */
> +static inline void
> +convert_to_display_mode(struct fb_videomode *timing,
> +                       struct drm_display_mode *mode)

I sort of prefer copy functions to copy right to left, Ie.

convert_foo(a, b)

copies from b to a.. somehow this feels more natural.  (Think a = b,
the receiving arg is on the left.. same as w/ memcpy()).  Same applies
w/ the inverse of this fxn.

> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       mode->clock = timing->pixclock / 1000;
> +
> +       mode->hdisplay = timing->xres;
> +       mode->hsync_start = mode->hdisplay + timing->left_margin;
> +       mode->hsync_end = mode->hsync_start + timing->hsync_len;
> +       mode->htotal = mode->hsync_end + timing->right_margin;
> +
> +       mode->vdisplay = timing->yres;
> +       mode->vsync_start = mode->vdisplay + timing->upper_margin;
> +       mode->vsync_end = mode->vsync_start + timing->vsync_len;
> +       mode->vtotal = mode->vsync_end + timing->lower_margin;
> +}
> +
> +/* convert drm_display_mode to samsung_video_timings */
> +static inline void
> +convert_to_video_timing(struct drm_display_mode *mode,
> +                       struct fb_videomode *timing)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       timing->pixclock = mode->clock * 1000;
> +
> +       timing->xres = mode->hdisplay;
> +       timing->left_margin = mode->hsync_start - mode->hdisplay;
> +       timing->hsync_len = mode->hsync_end - mode->hsync_start;
> +       timing->right_margin = mode->htotal - mode->hsync_end;
> +
> +       timing->yres = mode->vdisplay;
> +       timing->upper_margin = mode->vsync_start - mode->vdisplay;
> +       timing->vsync_len = mode->vsync_end - mode->vsync_start;
> +       timing->lower_margin = mode->vtotal - mode->vsync_end;
> +}
> +
> +static int samsung_drm_connector_get_modes(struct drm_connector *connector)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +       struct samsung_drm_display *display =
> +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> +       unsigned int count;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       if (display && display->get_edid) {
> +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> +               if (!edid) {
> +                       DRM_ERROR("failed to allocate edid\n");
> +                       return 0;
> +               }
> +
> +               display->get_edid(connector, edid, MAX_EDID);
> +
> +               drm_mode_connector_update_edid_property(connector, edid);
> +               count = drm_add_edid_modes(connector, edid);
> +
> +               kfree(connector->display_info.raw_edid);
> +               connector->display_info.raw_edid = edid;
> +       } else {
> +               struct drm_display_mode *mode = drm_mode_create(connector->dev);
> +               struct fb_videomode *timing;
> +
> +               if (display && display->get_timing)
> +                       timing = display->get_timing();

also seems a bit weird to not do: display->get_timings(display)..

maybe you don't have the case of multiple instances of the same
display time.. but maybe someday you need that.  (Maybe this is just
an artifact of how the API of your current lower layer driver is.. but
maybe now is a good time to think about those APIs)

Same comment about not passing the display instance ptr to some of the
other display fxn ptrs elsewhere in the code.

> +               else
> +                       return 0;
> +
> +               convert_to_display_mode(timing, mode);
> +
> +               mode->type = DRM_MODE_TYPE_DRIVER | DRM_MODE_TYPE_PREFERRED;
> +               drm_mode_set_name(mode);
> +               drm_mode_probed_add(connector, mode);
> +
> +               count = 1;
> +       }
> +
> +       return count;
> +}
> +
> +static int samsung_drm_connector_mode_valid(struct drm_connector *connector,
> +                                           struct drm_display_mode *mode)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +       struct samsung_drm_display *display =
> +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> +       struct fb_videomode timing;
> +       int ret = MODE_BAD;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       convert_to_video_timing(mode, &timing);
> +
> +       if (display && display->check_timing)
> +               if (!display->check_timing((void *)&timing))
> +                       ret = MODE_OK;
> +
> +       return ret;
> +}
> +
> +struct drm_encoder *samsung_drm_best_encoder(struct drm_connector *connector)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +       return to_samsung_connector(connector)->encoder;
> +}
> +
> +static struct drm_connector_helper_funcs samsung_connector_helper_funcs = {
> +       .get_modes      = samsung_drm_connector_get_modes,
> +       .mode_valid     = samsung_drm_connector_mode_valid,
> +       .best_encoder   = samsung_drm_best_encoder,
> +};
> +
> +/* get detection status of display device. */
> +static enum drm_connector_status
> +samsung_drm_connector_detect(struct drm_connector *connector, bool force)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +       struct samsung_drm_display *display =
> +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> +       unsigned int ret = connector_status_unknown;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       if (display && display->is_connected) {
> +               if (display->is_connected())
> +                       ret = connector_status_connected;
> +               else
> +                       ret = connector_status_disconnected;
> +       }
> +
> +       return ret;
> +}
> +
> +static void samsung_drm_connector_destroy(struct drm_connector *connector)
> +{
> +       struct samsung_drm_connector *samsung_connector =
> +               to_samsung_connector(connector);
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       drm_sysfs_connector_remove(connector);
> +       drm_connector_cleanup(connector);
> +       connector->dev->mode_config.num_connector--;

I wonder if num_connector-- should be in drm_connector_cleanup()..

I missed this in OMAP driver, but it seems that none of the other drm
drivers are directly decrementing this.. and it would seem more
symmetric for it to be in drm_connector_cleanup().  But would be
interesting to see what others think.  (I guess no one has really
dealt w/ dynamic creation/deletion of connectors yet?)

> +       kfree(samsung_connector);
> +}
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> new file mode 100644
> index 0000000..1c5a70a
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
[snip]
> +
> +struct samsung_drm_crtc {
> +       struct drm_crtc                 drm_crtc;
> +       struct samsung_drm_overlay      overlay;


I guess you are looking fwd to drm_plane support too :-)


> +       unsigned int                    pipe;
> +};
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> new file mode 100644
> index 0000000..42096eb
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> @@ -0,0 +1,248 @@
> +/*
> + * Copyright (c) 2011 Samsung Electronics Co., Ltd.
> + * Authors:
> + *     Inki Dae <inki.dae@samsung.com>
> + *     Joonyoung Shim <jy0922.shim@samsung.com>
> + *     SeungWoo Kim <sw0312.kim@samsung.com>
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
> + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
> + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
> + * OTHER DEALINGS IN THE SOFTWARE.
> + */
> +
> +#include "drmP.h"
> +#include "drm_crtc.h"
> +#include "drm_crtc_helper.h"
> +
> +#include "samsung_drm_fb.h"
> +#include "samsung_drm_buf.h"
> +#include "samsung_drm_gem.h"
> +
> +#define to_samsung_fb(x)       container_of(x, struct samsung_drm_fb, fb)
> +
> +struct samsung_drm_fb {
> +       struct drm_framebuffer          fb;
> +       struct drm_file                 *file_priv;
> +       struct samsung_drm_gem_obj      *samsung_gem_obj;
> +       unsigned int                    is_default;
> +
> +       /* samsung gem object handle. */
> +       unsigned int                    gem_handle;
> +       /* unique id to buffer object. */
> +       unsigned int                    id;
> +
> +       unsigned int                    fb_size;
> +       unsigned long                   paddr;
> +       void __iomem                    *vaddr;
> +};
> +
> +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> +{
> +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> +       int ret;
> +
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       drm_framebuffer_cleanup(fb);
> +
> +       if (samsung_fb->is_default) {
> +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> +                               samsung_fb->gem_handle);

why not keep the gem buffer ptr, and do something like:

  drm_gem_object_unreference_unlocked(samsung_fb->bo)..

this way, you get the right behavior if someone somewhere else took a
ref to the gem buffer object?  And it avoids needing to keep the
file_priv ptr in the fb (which seems a bit strange)


> +               if (ret < 0)
> +                       DRM_ERROR("failed to delete drm_gem_handle.\n");
> +       }
> +
> +       kfree(samsung_fb);
> +}
> +
[snip]
> diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> new file mode 100644
> index 0000000..1e167a6
> --- /dev/null
> +++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
[snip]
> +
> +static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
> +{
> +       DRM_DEBUG_KMS("%s\n", __FILE__);
> +
> +       return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
> +}

fwiw, candidate to move to drm_gem helper fxn (I have same in my
driver.. but this wasn't included in my earlier patch to add a couple
common drm_gem helper fxns like create/free_mmap_offset)

> +
> +/**
> + * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for an object
> + * @obj: obj in question
> + *
> + * GEM memory mapping works by handing back to userspace a fake mmap offset
> + * it can use in a subsequent mmap(2) call.  The DRM core code then looks
> + * up the object based on the offset and sets up the various memory mapping
> + * structures.
> + *
> + * This routine allocates and attaches a fake offset for @obj.
> + */
> +static int
> +samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)

some of these other fxns you might at least want to include a comment
to replace w/ drm_gem helpers once that patch lands.. your's would be
the 4th driver w/ nearly identical code (well, if you can count
non-mainline drivers) ;-)

> +{
> +       struct drm_device *dev = obj->dev;
> +       struct drm_gem_mm *mm = dev->mm_private;
> +       struct drm_map_list *list;
> +       struct drm_local_map *map;
> +       int ret = 0;
> +
> +       /* Set the object up for mmap'ing */
> +       list = &obj->map_list;
> +       list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
> +       if (!list->map)
> +               return -ENOMEM;
> +
> +       map = list->map;
> +       map->type = _DRM_GEM;
> +       map->size = obj->size;
> +       map->handle = obj;
> +
> +       /* Get a DRM GEM mmap offset allocated... */
> +       list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
> +                                                   obj->size / PAGE_SIZE,
> +                                                   0, 0);
> +       if (!list->file_offset_node) {
> +               DRM_ERROR("failed to allocate offset for bo %d\n",
> +                         obj->name);
> +               ret = -ENOSPC;
> +               goto out_free_list;
> +       }
> +
> +       list->file_offset_node = drm_mm_get_block(list->file_offset_node,
> +                                                 obj->size / PAGE_SIZE,
> +                                                 0);
> +       if (!list->file_offset_node) {
> +               ret = -ENOMEM;
> +               goto out_free_list;
> +       }
> +
> +       list->hash.key = list->file_offset_node->start;
> +       ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
> +       if (ret) {
> +               DRM_ERROR("failed to add to map hash\n");
> +               goto out_free_mm;
> +       }
> +
> +       return 0;
> +
> +out_free_mm:
> +       drm_mm_put_block(list->file_offset_node);
> +out_free_list:
> +       kfree(list->map);
> +       list->map = NULL;
> +
> +       return ret;
> +}
> +
> +static void
> +samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
> +{
> +       struct drm_device *dev = obj->dev;
> +       struct drm_gem_mm *mm = dev->mm_private;
> +       struct drm_map_list *list = &obj->map_list;
> +
> +       drm_ht_remove_item(&mm->offset_hash, &list->hash);
> +       drm_mm_put_block(list->file_offset_node);
> +       kfree(list->map);
> +       list->map = NULL;
> +}
[snip]
> +struct samsung_drm_gem_obj *
> +               find_samsung_drm_gem_object(struct drm_file *file_priv,
> +                       struct drm_device *dev, unsigned int handle)
> +{
> +       struct drm_gem_object *gem_obj;
> +
> +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> +       if (!gem_obj) {
> +               DRM_LOG_KMS("a invalid gem object not registered to lookup.\n");
> +               return NULL;
> +       }
> +
> +       /**
> +        * unreference refcount of the gem object.
> +        * at drm_gem_object_lookup(), the gem object was referenced.
> +        */
> +       drm_gem_object_unreference(gem_obj);

this doesn't seem right, to drop the reference before you use the
buffer elsewhere..

> +       return to_samsung_gem_obj(gem_obj);
> +}
> +
[snip]
> diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
> new file mode 100644
> index 0000000..05dc460
> --- /dev/null
> +++ b/include/drm/samsung_drm.h
[snip]
> +/**
> + * User-desired buffer creation information structure.
> + *
> + * @usr_addr: an address allocated by user process and this address
> + *     would be mmapped to physical region by fault handler.
> + * @size: requested size for the object.
> + *     - this size value would be page-aligned internally.
> + * @flags: user request for setting memory type or cache attributes.
> + * @handle: returned handle for the object.
> + */
> +struct drm_samsung_gem_create {
> +       unsigned int usr_addr;

usr_addr seems dangerous.. and unused?  Maybe can be removed?

also, I sort of prefer using stdint types for userspace<->kernel ioctl structs.

> +       unsigned int size;
> +       unsigned int flags;
> +
> +       unsigned int handle;
> +};
> +
> +/**
> + * A structure for getting buffer offset.
> + *
> + * @handle: a pointer to gem object created.
> + * @offset: relatived offset value of the memory region allocated.
> + *     - this value should be set by user.
> + */
> +struct drm_samsung_gem_map_off {
> +       unsigned int handle;
> +       uint64_t offset;
> +};
> +
> +/**
> + * A structure for mapping buffer.
> + *
> + * @handle: a pointer to gem object created.
> + * @offset: relatived offset value of the memory region allocated.
> + *     - this value should be set by user.
> + * @size: memory size to be mapped.
> + * @mapped: user virtual address to be mapped.
> + */
> +struct drm_samsung_gem_mmap {
> +       unsigned int handle;
> +       uint64_t offset;
> +       unsigned int size;
> +
> +       unsigned int mapped;

I guess this should be obtained by passing offset back to mmap() syscall


BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31  1:57   ` Rob Clark
@ 2011-08-31  2:06     ` Konrad Rzeszutek Wilk
  -1 siblings, 0 replies; 38+ messages in thread
From: Konrad Rzeszutek Wilk @ 2011-08-31  2:06 UTC (permalink / raw)
  To: Rob Clark
  Cc: Inki Dae, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

> > +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > +       if (!entry->paddr) {
> > +               DRM_ERROR("failed to allocate buffer.\n");
> > +               return -ENOMEM;
> > +       }
> > +
> > +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
> > +                       (unsigned int)entry->vaddr, entry->paddr, entry->size);
> > +
> > +       return 0;
> > +}
> > +
> [snip]
> 
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > new file mode 100644
> > index 0000000..d6a7e95
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> [snip]
> > +/**
> > + * samsung drm buffer entry structure.
> > + *
> > + * @paddr: physical address of allocated memory.
> > + * @vaddr: kernel virtual address of allocated memory.
> > + * @size: size of allocated memory.
> > + */
> > +struct samsung_drm_buf_entry {
> > +       unsigned int paddr;

This could be made 'dma_addr_t' and then you can drop all of the
casts to (dma_addr_t *).

.. snip..
> > +static int samsung_drm_connector_get_modes(struct drm_connector *connector)

Why not make the return be 'unsigned int'? 

.. snip..
> > +/* get detection status of display device. */
> > +static enum drm_connector_status
> > +samsung_drm_connector_detect(struct drm_connector *connector, bool force)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> > +       unsigned int ret = connector_status_unknown;

Not 'enum drm_connector_status ret = connector_status_unknown' ?

> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       if (display && display->is_connected) {
> > +               if (display->is_connected())
> > +                       ret = connector_status_connected;
> > +               else
> > +                       ret = connector_status_disconnected;
> > +       }
> > +
> > +       return ret;
> > +}

.. snip..
> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > +{
> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > +       int ret;

Get rid of 'ret'
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       drm_framebuffer_cleanup(fb);
> > +
> > +       if (samsung_fb->is_default) {
> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > +                               samsung_fb->gem_handle);
> 
> why not keep the gem buffer ptr, and do something like:
> 
>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> 
> this way, you get the right behavior if someone somewhere else took a
> ref to the gem buffer object?  And it avoids needing to keep the
> file_priv ptr in the fb (which seems a bit strange)
> 
> 
> > +               if (ret < 0)
> > +                       DRM_ERROR("failed to delete drm_gem_handle.\n");

And just do the check on the function return value here. You are not using
the 'ret' for anything.
> > +       }
> > +
> > +       kfree(samsung_fb);
> > +}
> > +
> [snip]

Hm, so I stopped here - just realized that I am missing some of the code
and I should look at the original patch..


^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  2:06     ` Konrad Rzeszutek Wilk
  0 siblings, 0 replies; 38+ messages in thread
From: Konrad Rzeszutek Wilk @ 2011-08-31  2:06 UTC (permalink / raw)
  To: Rob Clark
  Cc: sw0312.kim, linux-kernel, dri-devel, Inki Dae, kyungmin.park,
	linux-arm-kernel

> > +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > +       if (!entry->paddr) {
> > +               DRM_ERROR("failed to allocate buffer.\n");
> > +               return -ENOMEM;
> > +       }
> > +
> > +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x), size(0x%x)\n",
> > +                       (unsigned int)entry->vaddr, entry->paddr, entry->size);
> > +
> > +       return 0;
> > +}
> > +
> [snip]
> 
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > new file mode 100644
> > index 0000000..d6a7e95
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> [snip]
> > +/**
> > + * samsung drm buffer entry structure.
> > + *
> > + * @paddr: physical address of allocated memory.
> > + * @vaddr: kernel virtual address of allocated memory.
> > + * @size: size of allocated memory.
> > + */
> > +struct samsung_drm_buf_entry {
> > +       unsigned int paddr;

This could be made 'dma_addr_t' and then you can drop all of the
casts to (dma_addr_t *).

.. snip..
> > +static int samsung_drm_connector_get_modes(struct drm_connector *connector)

Why not make the return be 'unsigned int'? 

.. snip..
> > +/* get detection status of display device. */
> > +static enum drm_connector_status
> > +samsung_drm_connector_detect(struct drm_connector *connector, bool force)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)->display;
> > +       unsigned int ret = connector_status_unknown;

Not 'enum drm_connector_status ret = connector_status_unknown' ?

> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       if (display && display->is_connected) {
> > +               if (display->is_connected())
> > +                       ret = connector_status_connected;
> > +               else
> > +                       ret = connector_status_disconnected;
> > +       }
> > +
> > +       return ret;
> > +}

.. snip..
> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > +{
> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > +       int ret;

Get rid of 'ret'
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       drm_framebuffer_cleanup(fb);
> > +
> > +       if (samsung_fb->is_default) {
> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > +                               samsung_fb->gem_handle);
> 
> why not keep the gem buffer ptr, and do something like:
> 
>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> 
> this way, you get the right behavior if someone somewhere else took a
> ref to the gem buffer object?  And it avoids needing to keep the
> file_priv ptr in the fb (which seems a bit strange)
> 
> 
> > +               if (ret < 0)
> > +                       DRM_ERROR("failed to delete drm_gem_handle.\n");

And just do the check on the function return value here. You are not using
the 'ret' for anything.
> > +       }
> > +
> > +       kfree(samsung_fb);
> > +}
> > +
> [snip]

Hm, so I stopped here - just realized that I am missing some of the code
and I should look at the original patch..

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31  1:57   ` Rob Clark
  (?)
@ 2011-08-31  2:28     ` Joonyoung Shim
  -1 siblings, 0 replies; 38+ messages in thread
From: Joonyoung Shim @ 2011-08-31  2:28 UTC (permalink / raw)
  To: Rob Clark
  Cc: Inki Dae, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

>> +static void samsung_drm_connector_destroy(struct drm_connector *connector)
>> +{
>> +       struct samsung_drm_connector *samsung_connector =
>> +               to_samsung_connector(connector);
>> +
>> +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> +
>> +       drm_sysfs_connector_remove(connector);
>> +       drm_connector_cleanup(connector);
>> +       connector->dev->mode_config.num_connector--;
>
> I wonder if num_connector-- should be in drm_connector_cleanup()..
>

Right, num_connector and num_encoder of mode_config should be decreased
in the cleanup functions, but currently drm driver is missing it.

I sent the patch.

Thanks.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  2:28     ` Joonyoung Shim
  0 siblings, 0 replies; 38+ messages in thread
From: Joonyoung Shim @ 2011-08-31  2:28 UTC (permalink / raw)
  To: linux-arm-kernel

>> +static void samsung_drm_connector_destroy(struct drm_connector *connector)
>> +{
>> +       struct samsung_drm_connector *samsung_connector =
>> +               to_samsung_connector(connector);
>> +
>> +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> +
>> +       drm_sysfs_connector_remove(connector);
>> +       drm_connector_cleanup(connector);
>> +       connector->dev->mode_config.num_connector--;
>
> I wonder if num_connector-- should be in drm_connector_cleanup()..
>

Right, num_connector and num_encoder of mode_config should be decreased
in the cleanup functions, but currently drm driver is missing it.

I sent the patch.

Thanks.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  2:28     ` Joonyoung Shim
  0 siblings, 0 replies; 38+ messages in thread
From: Joonyoung Shim @ 2011-08-31  2:28 UTC (permalink / raw)
  To: Rob Clark
  Cc: sw0312.kim, linux-kernel, dri-devel, Inki Dae, kyungmin.park,
	linux-arm-kernel

>> +static void samsung_drm_connector_destroy(struct drm_connector *connector)
>> +{
>> +       struct samsung_drm_connector *samsung_connector =
>> +               to_samsung_connector(connector);
>> +
>> +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> +
>> +       drm_sysfs_connector_remove(connector);
>> +       drm_connector_cleanup(connector);
>> +       connector->dev->mode_config.num_connector--;
>
> I wonder if num_connector-- should be in drm_connector_cleanup()..
>

Right, num_connector and num_encoder of mode_config should be decreased
in the cleanup functions, but currently drm driver is missing it.

I sent the patch.

Thanks.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31  1:57   ` Rob Clark
  (?)
@ 2011-08-31  6:51     ` Inki Dae
  -1 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-31  6:51 UTC (permalink / raw)
  To: 'Rob Clark'
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

Hello, Rob.
Below is my answers and questions. and could you please include me as CC
when you post your driver?

Thank you.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark@gmail.com]
> Sent: Wednesday, August 31, 2011 10:58 AM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> Hi Inki,
> 
> Sorry for slightly overdue review.. it took a little while to go
> through the whole thing
> 

I  will be always pleased you to give me your advices and comments anytime.
:)

> comments in-line below
> 
> 
> On Fri, Aug 26, 2011 at 6:47 AM, Inki Dae <inki.dae@samsung.com> wrote:
> > This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables
> only FIMD yet
> > but we will add HDMI support also in the future.
> >
> > this patch is based on git repository below:
> > git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> > branch name: drm-next
> > commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
> >
> > you can refer to our working repository below:
> > http://git.infradead.org/users/kmpark/linux-2.6-samsung
> > branch name: samsung-drm
> >
> > We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> > based on Linux framebuffer) but couldn't so because lowlevel codes
> > of s3c-fb.c are included internally and so FIMD module of this driver
> has
> > its own lowlevel codes.
> >
> > We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> > for buffer allocation. by using DMA API, we could use CMA later.
> >
> > Refer to this link for CMA(Continuous Memory Allocator):
> > http://lkml.org/lkml/2011/7/20/45
> >
> > this driver supports only physically continuous memory(non-iommu).
> >
> > Links to previous versions of the patchset:
> > v1: < https://lwn.net/Articles/454380/ >
> > v2: < http://www.spinics.net/lists/kernel/msg1224275.html >
> >
> > Changelog v2:
> > DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
> >
> >     this feature maps user address space to physical memory region
> >     once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
> >
> > DRM: code clean and add exception codes.
> >
> > Changelog v3:
> > DRM: Support multiple irq.
> >
> >     FIMD and HDMI have their own irq handler but DRM Framework can
> regiter only one irq handler
> >     this patch supports mutiple irq for Samsung SoC.
> >
> > DRM: Consider modularization.
> >
> >     each DRM, FIMD could be built as a module.
> >
> > DRM: Have indenpendent crtc object.
> >
> >     crtc isn't specific to SoC Platform so this patch gets a crtc to be
> used as common object.
> >     created crtc could be attached to any encoder object.
> >
> > DRM: code clean and add exception codes.
> >
> > Signed-off-by: Inki Dae <inki.dae@samsung.com>
> > Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
> > Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
> > Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
> > ---
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c
> b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> > new file mode 100644
> > index 0000000..563d07e
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> [snip]
> > +static int lowlevel_buffer_allocate(struct drm_device *dev,
> > +               struct samsung_drm_buf_entry *entry)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> 
> very minor point, but DRM_DEBUG_KMS() already includes the function
> name.. not sure if __FILE__ is needed everywhere
> 
> > +
> > +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > +       if (!entry->paddr) {
> > +               DRM_ERROR("failed to allocate buffer.\n");
> > +               return -ENOMEM;
> > +       }
> > +
> > +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x),
> size(0x%x)\n",
> > +                       (unsigned int)entry->vaddr, entry->paddr, entry-
> >size);
> > +
> > +       return 0;
> > +}
> > +
> [snip]
> 
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h
> b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > new file mode 100644
> > index 0000000..d6a7e95
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> [snip]
> > +/**
> > + * samsung drm buffer entry structure.
> > + *
> > + * @paddr: physical address of allocated memory.
> > + * @vaddr: kernel virtual address of allocated memory.
> > + * @size: size of allocated memory.
> > + */
> > +struct samsung_drm_buf_entry {
> > +       unsigned int paddr;
> > +       void __iomem *vaddr;
> > +       unsigned int size;
> > +};
> 
> any reason not to combine this w/ samsung_drm_gem_obj to avoid extra
> levels of indirection?  (This would let you collapse some of the
> buffer allocation/deletion fxns into one level too.)
> 
I wanted it separates buffer allocator from gem but now seeing it, it
doesn't need to do so as you pointed out. I will combine it to
samsung_drm_gem_obj. thank you.

> > +
> > +/**
> > + * samsung drm buffer structure.
> > + *
> > + * @entry: pointer to samsung drm buffer entry object.
> > + * @flags: it means memory type to be alloated or cache attributes.
> > + * @handle: pointer to specific buffer object.
> > + * @id: unique id to specific buffer object.
> > + *
> > + * ps. this object would be transfered to user as kms_bo.handle so
> > + *     user can access to memory through kms_bo.handle.
> > + */
> > +struct samsung_drm_gem_obj {
> > +       struct drm_gem_object base;
> > +       struct samsung_drm_buf_entry *entry;
> > +       unsigned int flags;
> > +
> > +       unsigned int handle;
> > +       unsigned int id;
> > +};
> > +
> > +/* create new buffer object and memory region and add the object to
> list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
> > +               unsigned int size);
> > +
> > +/* allocate physical memory and add its object to list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device
> *dev,
> > +               unsigned int size);
> > +
> > +/* remove allocated physical memory. */
> > +int samsung_drm_buf_destroy(struct drm_device *dev,
> > +               struct samsung_drm_gem_obj *in_obj);
> > +
> > +/* find object added to list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device
> *dev,
> > +               struct samsung_drm_gem_obj *in_obj, unsigned int paddr);
> 
> this doesn't appear to be used anywhere
> 
Oh, it’s my mistake. I will remove it. thank you.

> > +
> > +#endif
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c
> b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> > new file mode 100644
> > index 0000000..987a629
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> [snip]
> > +
> > +/* convert samsung_video_timings to drm_display_mode */
> > +static inline void
> > +convert_to_display_mode(struct fb_videomode *timing,
> > +                       struct drm_display_mode *mode)
> 
> I sort of prefer copy functions to copy right to left, Ie.
> 
> convert_foo(a, b)
> 
> copies from b to a.. somehow this feels more natural.  (Think a = b,
> the receiving arg is on the left.. same as w/ memcpy()).  Same applies
> w/ the inverse of this fxn.
> 
Ok, I will modify it like your saying. Thank you.

> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       mode->clock = timing->pixclock / 1000;
> > +
> > +       mode->hdisplay = timing->xres;
> > +       mode->hsync_start = mode->hdisplay + timing->left_margin;
> > +       mode->hsync_end = mode->hsync_start + timing->hsync_len;
> > +       mode->htotal = mode->hsync_end + timing->right_margin;
> > +
> > +       mode->vdisplay = timing->yres;
> > +       mode->vsync_start = mode->vdisplay + timing->upper_margin;
> > +       mode->vsync_end = mode->vsync_start + timing->vsync_len;
> > +       mode->vtotal = mode->vsync_end + timing->lower_margin;
> > +}
> > +
> > +/* convert drm_display_mode to samsung_video_timings */
> > +static inline void
> > +convert_to_video_timing(struct drm_display_mode *mode,
> > +                       struct fb_videomode *timing)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       timing->pixclock = mode->clock * 1000;
> > +
> > +       timing->xres = mode->hdisplay;
> > +       timing->left_margin = mode->hsync_start - mode->hdisplay;
> > +       timing->hsync_len = mode->hsync_end - mode->hsync_start;
> > +       timing->right_margin = mode->htotal - mode->hsync_end;
> > +
> > +       timing->yres = mode->vdisplay;
> > +       timing->upper_margin = mode->vsync_start - mode->vdisplay;
> > +       timing->vsync_len = mode->vsync_end - mode->vsync_start;
> > +       timing->lower_margin = mode->vtotal - mode->vsync_end;
> > +}
> > +
> > +static int samsung_drm_connector_get_modes(struct drm_connector
> *connector)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > +       unsigned int count;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       if (display && display->get_edid) {
> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> > +               if (!edid) {
> > +                       DRM_ERROR("failed to allocate edid\n");
> > +                       return 0;
> > +               }
> > +
> > +               display->get_edid(connector, edid, MAX_EDID);
> > +
> > +               drm_mode_connector_update_edid_property(connector,
edid);
> > +               count = drm_add_edid_modes(connector, edid);
> > +
> > +               kfree(connector->display_info.raw_edid);
> > +               connector->display_info.raw_edid = edid;
> > +       } else {
> > +               struct drm_display_mode *mode =
drm_mode_create(connector-
> >dev);
> > +               struct fb_videomode *timing;
> > +
> > +               if (display && display->get_timing)
> > +                       timing = display->get_timing();
> 
> also seems a bit weird to not do: display->get_timings(display)..
> 
> maybe you don't have the case of multiple instances of the same
> display time.. but maybe someday you need that.  (Maybe this is just
> an artifact of how the API of your current lower layer driver is.. but
> maybe now is a good time to think about those APIs)
> 

display is static object set by hardware specific driver such as display
controller or hdmi. each hardware specific driver has their own display
object statically. You can refer to the definition in samsung_drm_fimd.c.
and I understood a encoder and a connector is 1:1 relationship, when any
hardware specific driver is probed, they would be created with a manager
that includes their own display object as pointer. Could you please give me
more comments why display object should be considered for multiple
instances? I am afraid there is any part I don't care.

thank you.

> Same comment about not passing the display instance ptr to some of the
> other display fxn ptrs elsewhere in the code.
> 
> > +               else
> > +                       return 0;
> > +
> > +               convert_to_display_mode(timing, mode);
> > +
> > +               mode->type = DRM_MODE_TYPE_DRIVER |
> DRM_MODE_TYPE_PREFERRED;
> > +               drm_mode_set_name(mode);
> > +               drm_mode_probed_add(connector, mode);
> > +
> > +               count = 1;
> > +       }
> > +
> > +       return count;
> > +}
> > +
> > +static int samsung_drm_connector_mode_valid(struct drm_connector
> *connector,
> > +                                           struct drm_display_mode
*mode)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > +       struct fb_videomode timing;
> > +       int ret = MODE_BAD;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       convert_to_video_timing(mode, &timing);
> > +
> > +       if (display && display->check_timing)
> > +               if (!display->check_timing((void *)&timing))
> > +                       ret = MODE_OK;
> > +
> > +       return ret;
> > +}
> > +
> > +struct drm_encoder *samsung_drm_best_encoder(struct drm_connector
> *connector)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +       return to_samsung_connector(connector)->encoder;
> > +}
> > +
> > +static struct drm_connector_helper_funcs samsung_connector_helper_funcs
> = {
> > +       .get_modes      = samsung_drm_connector_get_modes,
> > +       .mode_valid     = samsung_drm_connector_mode_valid,
> > +       .best_encoder   = samsung_drm_best_encoder,
> > +};
> > +
> > +/* get detection status of display device. */
> > +static enum drm_connector_status
> > +samsung_drm_connector_detect(struct drm_connector *connector, bool
> force)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > +       unsigned int ret = connector_status_unknown;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       if (display && display->is_connected) {
> > +               if (display->is_connected())
> > +                       ret = connector_status_connected;
> > +               else
> > +                       ret = connector_status_disconnected;
> > +       }
> > +
> > +       return ret;
> > +}
> > +
> > +static void samsung_drm_connector_destroy(struct drm_connector
> *connector)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       drm_sysfs_connector_remove(connector);
> > +       drm_connector_cleanup(connector);
> > +       connector->dev->mode_config.num_connector--;
> 
> I wonder if num_connector-- should be in drm_connector_cleanup()..
> 
> I missed this in OMAP driver, but it seems that none of the other drm
> drivers are directly decrementing this.. and it would seem more
> symmetric for it to be in drm_connector_cleanup().  But would be

For this, it was already commented by Joonyoun Shim.

> interesting to see what others think.  (I guess no one has really
> dealt w/ dynamic creation/deletion of connectors yet?)
> 

I think our driver creates encoder and connector dynamically. please, see
samsung_drm_subdrv_probe() of samsung_drm_core.c. hardware specific drivers
such as display controller and hdmi are registered to
samsung_drm_subdrv_list and when drm driver is probed, once subdrv->probe
callback is called, a encoder and a connector would be created dynamically.
if there is any missing point, feel free to give me your comments please.

Thank you.

> > +       kfree(samsung_connector);
> > +}
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> > new file mode 100644
> > index 0000000..1c5a70a
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> [snip]
> > +
> > +struct samsung_drm_crtc {
> > +       struct drm_crtc                 drm_crtc;
> > +       struct samsung_drm_overlay      overlay;
> 
> 
> I guess you are looking fwd to drm_plane support too :-)
> 
> 
> > +       unsigned int                    pipe;
> > +};
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c
> b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> > new file mode 100644
> > index 0000000..42096eb
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> > @@ -0,0 +1,248 @@
> > +/*
> > + * Copyright (c) 2011 Samsung Electronics Co., Ltd.
> > + * Authors:
> > + *     Inki Dae <inki.dae@samsung.com>
> > + *     Joonyoung Shim <jy0922.shim@samsung.com>
> > + *     SeungWoo Kim <sw0312.kim@samsung.com>
> > + *
> > + * Permission is hereby granted, free of charge, to any person
> obtaining a
> > + * copy of this software and associated documentation files (the
> "Software"),
> > + * to deal in the Software without restriction, including without
> limitation
> > + * the rights to use, copy, modify, merge, publish, distribute,
> sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom
> the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice (including the
> next
> > + * paragraph) shall be included in all copies or substantial portions
> of the
> > + * Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT
> SHALL
> > + * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
> DAMAGES OR
> > + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
OTHERWISE,
> > + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE
> OR
> > + * OTHER DEALINGS IN THE SOFTWARE.
> > + */
> > +
> > +#include "drmP.h"
> > +#include "drm_crtc.h"
> > +#include "drm_crtc_helper.h"
> > +
> > +#include "samsung_drm_fb.h"
> > +#include "samsung_drm_buf.h"
> > +#include "samsung_drm_gem.h"
> > +
> > +#define to_samsung_fb(x)       container_of(x, struct samsung_drm_fb,
fb)
> > +
> > +struct samsung_drm_fb {
> > +       struct drm_framebuffer          fb;
> > +       struct drm_file                 *file_priv;
> > +       struct samsung_drm_gem_obj      *samsung_gem_obj;
> > +       unsigned int                    is_default;
> > +
> > +       /* samsung gem object handle. */
> > +       unsigned int                    gem_handle;
> > +       /* unique id to buffer object. */
> > +       unsigned int                    id;
> > +
> > +       unsigned int                    fb_size;
> > +       unsigned long                   paddr;
> > +       void __iomem                    *vaddr;
> > +};
> > +
> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > +{
> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > +       int ret;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       drm_framebuffer_cleanup(fb);
> > +
> > +       if (samsung_fb->is_default) {
> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > +                               samsung_fb->gem_handle);
> 
> why not keep the gem buffer ptr, and do something like:
> 
>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> 
> this way, you get the right behavior if someone somewhere else took a
> ref to the gem buffer object?  And it avoids needing to keep the
> file_priv ptr in the fb (which seems a bit strange)
> 
> 
Yes, at booting time, one gem object is created. this is for linux
framebuffer and used as default buffer. register_framebuffer function is
called one time at booting time by drm driver. but when this driver is built
as modules, this gem object should be released with rmmod modules. and the
reason fb has file point is that drm_gem_handle_delete requests it to
release a gem object. our driver considered modularization also. If there is
any point I missed, give me any comment please. Thank you.

> > +               if (ret < 0)
> > +                       DRM_ERROR("failed to delete drm_gem_handle.\n");
> > +       }
> > +
> > +       kfree(samsung_fb);
> > +}
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c
> b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> > new file mode 100644
> > index 0000000..1e167a6
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> [snip]
> > +
> > +static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
> > +}
> 
> fwiw, candidate to move to drm_gem helper fxn (I have same in my
> driver.. but this wasn't included in my earlier patch to add a couple
> common drm_gem helper fxns like create/free_mmap_offset)
> 

I referred to your code. :) and I wish this code is moved to mainline to be
used commonly.

> > +
> > +/**
> > + * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for
> an object
> > + * @obj: obj in question
> > + *
> > + * GEM memory mapping works by handing back to userspace a fake mmap
> offset
> > + * it can use in a subsequent mmap(2) call.  The DRM core code then
> looks
> > + * up the object based on the offset and sets up the various memory
> mapping
> > + * structures.
> > + *
> > + * This routine allocates and attaches a fake offset for @obj.
> > + */
> > +static int
> > +samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)
> 
> some of these other fxns you might at least want to include a comment
> to replace w/ drm_gem helpers once that patch lands.. your's would be
> the 4th driver w/ nearly identical code (well, if you can count
> non-mainline drivers) ;-)
> 
Ok, I know you had posted this patch to be used commonly. but it isn't
applied to dri git. If applied, I will change it to common fxn but for now I
will add a comment.
> > +{
> > +       struct drm_device *dev = obj->dev;
> > +       struct drm_gem_mm *mm = dev->mm_private;
> > +       struct drm_map_list *list;
> > +       struct drm_local_map *map;
> > +       int ret = 0;
> > +
> > +       /* Set the object up for mmap'ing */
> > +       list = &obj->map_list;
> > +       list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
> > +       if (!list->map)
> > +               return -ENOMEM;
> > +
> > +       map = list->map;
> > +       map->type = _DRM_GEM;
> > +       map->size = obj->size;
> > +       map->handle = obj;
> > +
> > +       /* Get a DRM GEM mmap offset allocated... */
> > +       list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
> > +                                                   obj->size /
PAGE_SIZE,
> > +                                                   0, 0);
> > +       if (!list->file_offset_node) {
> > +               DRM_ERROR("failed to allocate offset for bo %d\n",
> > +                         obj->name);
> > +               ret = -ENOSPC;
> > +               goto out_free_list;
> > +       }
> > +
> > +       list->file_offset_node =
drm_mm_get_block(list->file_offset_node,
> > +                                                 obj->size / PAGE_SIZE,
> > +                                                 0);
> > +       if (!list->file_offset_node) {
> > +               ret = -ENOMEM;
> > +               goto out_free_list;
> > +       }
> > +
> > +       list->hash.key = list->file_offset_node->start;
> > +       ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
> > +       if (ret) {
> > +               DRM_ERROR("failed to add to map hash\n");
> > +               goto out_free_mm;
> > +       }
> > +
> > +       return 0;
> > +
> > +out_free_mm:
> > +       drm_mm_put_block(list->file_offset_node);
> > +out_free_list:
> > +       kfree(list->map);
> > +       list->map = NULL;
> > +
> > +       return ret;
> > +}
> > +
> > +static void
> > +samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
> > +{
> > +       struct drm_device *dev = obj->dev;
> > +       struct drm_gem_mm *mm = dev->mm_private;
> > +       struct drm_map_list *list = &obj->map_list;
> > +
> > +       drm_ht_remove_item(&mm->offset_hash, &list->hash);
> > +       drm_mm_put_block(list->file_offset_node);
> > +       kfree(list->map);
> > +       list->map = NULL;
> > +}
> [snip]
> > +struct samsung_drm_gem_obj *
> > +               find_samsung_drm_gem_object(struct drm_file *file_priv,
> > +                       struct drm_device *dev, unsigned int handle)
> > +{
> > +       struct drm_gem_object *gem_obj;
> > +
> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> > +       if (!gem_obj) {
> > +               DRM_LOG_KMS("a invalid gem object not registered to
> lookup.\n");
> > +               return NULL;
> > +       }
> > +
> > +       /**
> > +        * unreference refcount of the gem object.
> > +        * at drm_gem_object_lookup(), the gem object was referenced.
> > +        */
> > +       drm_gem_object_unreference(gem_obj);
> 
> this doesn't seem right, to drop the reference before you use the
> buffer elsewhere..
> 
No, see drm_gem_object_lookup fxn. at this function, if there is a object
found then drm_gem_object_reference is called to increase refcount of this
object. if there is any missing point, give me any comment please. thank
you.

> > +       return to_samsung_gem_obj(gem_obj);
> > +}
> > +
> [snip]
> > diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
> > new file mode 100644
> > index 0000000..05dc460
> > --- /dev/null
> > +++ b/include/drm/samsung_drm.h
> [snip]
> > +/**
> > + * User-desired buffer creation information structure.
> > + *
> > + * @usr_addr: an address allocated by user process and this address
> > + *     would be mmapped to physical region by fault handler.
> > + * @size: requested size for the object.
> > + *     - this size value would be page-aligned internally.
> > + * @flags: user request for setting memory type or cache attributes.
> > + * @handle: returned handle for the object.
> > + */
> > +struct drm_samsung_gem_create {
> > +       unsigned int usr_addr;
> 
> usr_addr seems dangerous.. and unused?  Maybe can be removed?
> 
Actually, usr_addr isn't used anywhere, so I removed it already and Dave
pointed out it before.

> also, I sort of prefer using stdint types for userspace<->kernel ioctl
> structs.
> 
Ok, I will use stdint types for it. thank you.

> > +       unsigned int size;
> > +       unsigned int flags;
> > +
> > +       unsigned int handle;
> > +};
> > +
> > +/**
> > + * A structure for getting buffer offset.
> > + *
> > + * @handle: a pointer to gem object created.
> > + * @offset: relatived offset value of the memory region allocated.
> > + *     - this value should be set by user.
> > + */
> > +struct drm_samsung_gem_map_off {
> > +       unsigned int handle;
> > +       uint64_t offset;
> > +};
> > +
> > +/**
> > + * A structure for mapping buffer.
> > + *
> > + * @handle: a pointer to gem object created.
> > + * @offset: relatived offset value of the memory region allocated.
> > + *     - this value should be set by user.
> > + * @size: memory size to be mapped.
> > + * @mapped: user virtual address to be mapped.
> > + */
> > +struct drm_samsung_gem_mmap {
> > +       unsigned int handle;
> > +       uint64_t offset;
> > +       unsigned int size;
> > +
> > +       unsigned int mapped;
> 
> I guess this should be obtained by passing offset back to mmap() syscall
No, as I gave you the words about it before, the offset means physical
memory offset to be mapped to user space. and mapped has user virtual
address from do_mmap.
refer to previous answer please: <
http://www.spinics.net/lists/dri-devel/msg13880.html >
> 
> BR,
> -R

Thank you for your comments and advices again. it's been very useful.  :)



^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  6:51     ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-31  6:51 UTC (permalink / raw)
  To: linux-arm-kernel

Hello, Rob.
Below is my answers and questions. and could you please include me as CC
when you post your driver?

Thank you.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark at gmail.com]
> Sent: Wednesday, August 31, 2011 10:58 AM
> To: Inki Dae
> Cc: airlied at linux.ie; dri-devel at lists.freedesktop.org;
> sw0312.kim at samsung.com; linux-kernel at vger.kernel.org;
> kyungmin.park at samsung.com; linux-arm-kernel at lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> Hi Inki,
> 
> Sorry for slightly overdue review.. it took a little while to go
> through the whole thing
> 

I  will be always pleased you to give me your advices and comments anytime.
:)

> comments in-line below
> 
> 
> On Fri, Aug 26, 2011 at 6:47 AM, Inki Dae <inki.dae@samsung.com> wrote:
> > This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables
> only FIMD yet
> > but we will add HDMI support also in the future.
> >
> > this patch is based on git repository below:
> > git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> > branch name: drm-next
> > commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
> >
> > you can refer to our working repository below:
> > http://git.infradead.org/users/kmpark/linux-2.6-samsung
> > branch name: samsung-drm
> >
> > We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> > based on Linux framebuffer) but couldn't so because lowlevel codes
> > of s3c-fb.c are included internally and so FIMD module of this driver
> has
> > its own lowlevel codes.
> >
> > We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> > for buffer allocation. by using DMA API, we could use CMA later.
> >
> > Refer to this link for CMA(Continuous Memory Allocator):
> > http://lkml.org/lkml/2011/7/20/45
> >
> > this driver supports only physically continuous memory(non-iommu).
> >
> > Links to previous versions of the patchset:
> > v1: < https://lwn.net/Articles/454380/ >
> > v2: < http://www.spinics.net/lists/kernel/msg1224275.html >
> >
> > Changelog v2:
> > DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
> >
> > ? ? this feature maps user address space to physical memory region
> > ? ? once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
> >
> > DRM: code clean and add exception codes.
> >
> > Changelog v3:
> > DRM: Support multiple irq.
> >
> > ? ? FIMD and HDMI have their own irq handler but DRM Framework can
> regiter only one irq handler
> > ? ? this patch supports mutiple irq for Samsung SoC.
> >
> > DRM: Consider modularization.
> >
> > ? ? each DRM, FIMD could be built as a module.
> >
> > DRM: Have indenpendent crtc object.
> >
> > ? ? crtc isn't specific to SoC Platform so this patch gets a crtc to be
> used as common object.
> > ? ? created crtc could be attached to any encoder object.
> >
> > DRM: code clean and add exception codes.
> >
> > Signed-off-by: Inki Dae <inki.dae@samsung.com>
> > Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
> > Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
> > Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
> > ---
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c
> b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> > new file mode 100644
> > index 0000000..563d07e
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> [snip]
> > +static int lowlevel_buffer_allocate(struct drm_device *dev,
> > + ? ? ? ? ? ? ? struct samsung_drm_buf_entry *entry)
> > +{
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> 
> very minor point, but DRM_DEBUG_KMS() already includes the function
> name.. not sure if __FILE__ is needed everywhere
> 
> > +
> > + ? ? ? entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > + ? ? ? ? ? ? ? ? ? ? ? (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > + ? ? ? if (!entry->paddr) {
> > + ? ? ? ? ? ? ? DRM_ERROR("failed to allocate buffer.\n");
> > + ? ? ? ? ? ? ? return -ENOMEM;
> > + ? ? ? }
> > +
> > + ? ? ? DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x),
> size(0x%x)\n",
> > + ? ? ? ? ? ? ? ? ? ? ? (unsigned int)entry->vaddr, entry->paddr, entry-
> >size);
> > +
> > + ? ? ? return 0;
> > +}
> > +
> [snip]
> 
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h
> b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > new file mode 100644
> > index 0000000..d6a7e95
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> [snip]
> > +/**
> > + * samsung drm buffer entry structure.
> > + *
> > + * @paddr: physical address of allocated memory.
> > + * @vaddr: kernel virtual address of allocated memory.
> > + * @size: size of allocated memory.
> > + */
> > +struct samsung_drm_buf_entry {
> > + ? ? ? unsigned int paddr;
> > + ? ? ? void __iomem *vaddr;
> > + ? ? ? unsigned int size;
> > +};
> 
> any reason not to combine this w/ samsung_drm_gem_obj to avoid extra
> levels of indirection?  (This would let you collapse some of the
> buffer allocation/deletion fxns into one level too.)
> 
I wanted it separates buffer allocator from gem but now seeing it, it
doesn't need to do so as you pointed out. I will combine it to
samsung_drm_gem_obj. thank you.

> > +
> > +/**
> > + * samsung drm buffer structure.
> > + *
> > + * @entry: pointer to samsung drm buffer entry object.
> > + * @flags: it means memory type to be alloated or cache attributes.
> > + * @handle: pointer to specific buffer object.
> > + * @id: unique id to specific buffer object.
> > + *
> > + * ps. this object would be transfered to user as kms_bo.handle so
> > + * ? ? user can access to memory through kms_bo.handle.
> > + */
> > +struct samsung_drm_gem_obj {
> > + ? ? ? struct drm_gem_object base;
> > + ? ? ? struct samsung_drm_buf_entry *entry;
> > + ? ? ? unsigned int flags;
> > +
> > + ? ? ? unsigned int handle;
> > + ? ? ? unsigned int id;
> > +};
> > +
> > +/* create new buffer object and memory region and add the object to
> list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
> > + ? ? ? ? ? ? ? unsigned int size);
> > +
> > +/* allocate physical memory and add its object to list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device
> *dev,
> > + ? ? ? ? ? ? ? unsigned int size);
> > +
> > +/* remove allocated physical memory. */
> > +int samsung_drm_buf_destroy(struct drm_device *dev,
> > + ? ? ? ? ? ? ? struct samsung_drm_gem_obj *in_obj);
> > +
> > +/* find object added to list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device
> *dev,
> > + ? ? ? ? ? ? ? struct samsung_drm_gem_obj *in_obj, unsigned int paddr);
> 
> this doesn't appear to be used anywhere
> 
Oh, it?s my mistake. I will remove it. thank you.

> > +
> > +#endif
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c
> b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> > new file mode 100644
> > index 0000000..987a629
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> [snip]
> > +
> > +/* convert samsung_video_timings to drm_display_mode */
> > +static inline void
> > +convert_to_display_mode(struct fb_videomode *timing,
> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_display_mode *mode)
> 
> I sort of prefer copy functions to copy right to left, Ie.
> 
> convert_foo(a, b)
> 
> copies from b to a.. somehow this feels more natural.  (Think a = b,
> the receiving arg is on the left.. same as w/ memcpy()).  Same applies
> w/ the inverse of this fxn.
> 
Ok, I will modify it like your saying. Thank you.

> > +{
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? mode->clock = timing->pixclock / 1000;
> > +
> > + ? ? ? mode->hdisplay = timing->xres;
> > + ? ? ? mode->hsync_start = mode->hdisplay + timing->left_margin;
> > + ? ? ? mode->hsync_end = mode->hsync_start + timing->hsync_len;
> > + ? ? ? mode->htotal = mode->hsync_end + timing->right_margin;
> > +
> > + ? ? ? mode->vdisplay = timing->yres;
> > + ? ? ? mode->vsync_start = mode->vdisplay + timing->upper_margin;
> > + ? ? ? mode->vsync_end = mode->vsync_start + timing->vsync_len;
> > + ? ? ? mode->vtotal = mode->vsync_end + timing->lower_margin;
> > +}
> > +
> > +/* convert drm_display_mode to samsung_video_timings */
> > +static inline void
> > +convert_to_video_timing(struct drm_display_mode *mode,
> > + ? ? ? ? ? ? ? ? ? ? ? struct fb_videomode *timing)
> > +{
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? timing->pixclock = mode->clock * 1000;
> > +
> > + ? ? ? timing->xres = mode->hdisplay;
> > + ? ? ? timing->left_margin = mode->hsync_start - mode->hdisplay;
> > + ? ? ? timing->hsync_len = mode->hsync_end - mode->hsync_start;
> > + ? ? ? timing->right_margin = mode->htotal - mode->hsync_end;
> > +
> > + ? ? ? timing->yres = mode->vdisplay;
> > + ? ? ? timing->upper_margin = mode->vsync_start - mode->vdisplay;
> > + ? ? ? timing->vsync_len = mode->vsync_end - mode->vsync_start;
> > + ? ? ? timing->lower_margin = mode->vtotal - mode->vsync_end;
> > +}
> > +
> > +static int samsung_drm_connector_get_modes(struct drm_connector
> *connector)
> > +{
> > + ? ? ? struct samsung_drm_connector *samsung_connector =
> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> > + ? ? ? struct samsung_drm_display *display =
> > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > + ? ? ? unsigned int count;
> > +
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? if (display && display->get_edid) {
> > + ? ? ? ? ? ? ? void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> > + ? ? ? ? ? ? ? if (!edid) {
> > + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to allocate edid\n");
> > + ? ? ? ? ? ? ? ? ? ? ? return 0;
> > + ? ? ? ? ? ? ? }
> > +
> > + ? ? ? ? ? ? ? display->get_edid(connector, edid, MAX_EDID);
> > +
> > + ? ? ? ? ? ? ? drm_mode_connector_update_edid_property(connector,
edid);
> > + ? ? ? ? ? ? ? count = drm_add_edid_modes(connector, edid);
> > +
> > + ? ? ? ? ? ? ? kfree(connector->display_info.raw_edid);
> > + ? ? ? ? ? ? ? connector->display_info.raw_edid = edid;
> > + ? ? ? } else {
> > + ? ? ? ? ? ? ? struct drm_display_mode *mode =
drm_mode_create(connector-
> >dev);
> > + ? ? ? ? ? ? ? struct fb_videomode *timing;
> > +
> > + ? ? ? ? ? ? ? if (display && display->get_timing)
> > + ? ? ? ? ? ? ? ? ? ? ? timing = display->get_timing();
> 
> also seems a bit weird to not do: display->get_timings(display)..
> 
> maybe you don't have the case of multiple instances of the same
> display time.. but maybe someday you need that.  (Maybe this is just
> an artifact of how the API of your current lower layer driver is.. but
> maybe now is a good time to think about those APIs)
> 

display is static object set by hardware specific driver such as display
controller or hdmi. each hardware specific driver has their own display
object statically. You can refer to the definition in samsung_drm_fimd.c.
and I understood a encoder and a connector is 1:1 relationship, when any
hardware specific driver is probed, they would be created with a manager
that includes their own display object as pointer. Could you please give me
more comments why display object should be considered for multiple
instances? I am afraid there is any part I don't care.

thank you.

> Same comment about not passing the display instance ptr to some of the
> other display fxn ptrs elsewhere in the code.
> 
> > + ? ? ? ? ? ? ? else
> > + ? ? ? ? ? ? ? ? ? ? ? return 0;
> > +
> > + ? ? ? ? ? ? ? convert_to_display_mode(timing, mode);
> > +
> > + ? ? ? ? ? ? ? mode->type = DRM_MODE_TYPE_DRIVER |
> DRM_MODE_TYPE_PREFERRED;
> > + ? ? ? ? ? ? ? drm_mode_set_name(mode);
> > + ? ? ? ? ? ? ? drm_mode_probed_add(connector, mode);
> > +
> > + ? ? ? ? ? ? ? count = 1;
> > + ? ? ? }
> > +
> > + ? ? ? return count;
> > +}
> > +
> > +static int samsung_drm_connector_mode_valid(struct drm_connector
> *connector,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct drm_display_mode
*mode)
> > +{
> > + ? ? ? struct samsung_drm_connector *samsung_connector =
> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> > + ? ? ? struct samsung_drm_display *display =
> > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > + ? ? ? struct fb_videomode timing;
> > + ? ? ? int ret = MODE_BAD;
> > +
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? convert_to_video_timing(mode, &timing);
> > +
> > + ? ? ? if (display && display->check_timing)
> > + ? ? ? ? ? ? ? if (!display->check_timing((void *)&timing))
> > + ? ? ? ? ? ? ? ? ? ? ? ret = MODE_OK;
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +struct drm_encoder *samsung_drm_best_encoder(struct drm_connector
> *connector)
> > +{
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > + ? ? ? return to_samsung_connector(connector)->encoder;
> > +}
> > +
> > +static struct drm_connector_helper_funcs samsung_connector_helper_funcs
> = {
> > + ? ? ? .get_modes ? ? ?= samsung_drm_connector_get_modes,
> > + ? ? ? .mode_valid ? ? = samsung_drm_connector_mode_valid,
> > + ? ? ? .best_encoder ? = samsung_drm_best_encoder,
> > +};
> > +
> > +/* get detection status of display device. */
> > +static enum drm_connector_status
> > +samsung_drm_connector_detect(struct drm_connector *connector, bool
> force)
> > +{
> > + ? ? ? struct samsung_drm_connector *samsung_connector =
> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> > + ? ? ? struct samsung_drm_display *display =
> > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > + ? ? ? unsigned int ret = connector_status_unknown;
> > +
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? if (display && display->is_connected) {
> > + ? ? ? ? ? ? ? if (display->is_connected())
> > + ? ? ? ? ? ? ? ? ? ? ? ret = connector_status_connected;
> > + ? ? ? ? ? ? ? else
> > + ? ? ? ? ? ? ? ? ? ? ? ret = connector_status_disconnected;
> > + ? ? ? }
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static void samsung_drm_connector_destroy(struct drm_connector
> *connector)
> > +{
> > + ? ? ? struct samsung_drm_connector *samsung_connector =
> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> > +
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? drm_sysfs_connector_remove(connector);
> > + ? ? ? drm_connector_cleanup(connector);
> > + ? ? ? connector->dev->mode_config.num_connector--;
> 
> I wonder if num_connector-- should be in drm_connector_cleanup()..
> 
> I missed this in OMAP driver, but it seems that none of the other drm
> drivers are directly decrementing this.. and it would seem more
> symmetric for it to be in drm_connector_cleanup().  But would be

For this, it was already commented by Joonyoun Shim.

> interesting to see what others think.  (I guess no one has really
> dealt w/ dynamic creation/deletion of connectors yet?)
> 

I think our driver creates encoder and connector dynamically. please, see
samsung_drm_subdrv_probe() of samsung_drm_core.c. hardware specific drivers
such as display controller and hdmi are registered to
samsung_drm_subdrv_list and when drm driver is probed, once subdrv->probe
callback is called, a encoder and a connector would be created dynamically.
if there is any missing point, feel free to give me your comments please.

Thank you.

> > + ? ? ? kfree(samsung_connector);
> > +}
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> > new file mode 100644
> > index 0000000..1c5a70a
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> [snip]
> > +
> > +struct samsung_drm_crtc {
> > + ? ? ? struct drm_crtc ? ? ? ? ? ? ? ? drm_crtc;
> > + ? ? ? struct samsung_drm_overlay ? ? ?overlay;
> 
> 
> I guess you are looking fwd to drm_plane support too :-)
> 
> 
> > + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?pipe;
> > +};
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c
> b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> > new file mode 100644
> > index 0000000..42096eb
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> > @@ -0,0 +1,248 @@
> > +/*
> > + * Copyright (c) 2011 Samsung Electronics Co., Ltd.
> > + * Authors:
> > + * ? ? Inki Dae <inki.dae@samsung.com>
> > + * ? ? Joonyoung Shim <jy0922.shim@samsung.com>
> > + * ? ? SeungWoo Kim <sw0312.kim@samsung.com>
> > + *
> > + * Permission is hereby granted, free of charge, to any person
> obtaining a
> > + * copy of this software and associated documentation files (the
> "Software"),
> > + * to deal in the Software without restriction, including without
> limitation
> > + * the rights to use, copy, modify, merge, publish, distribute,
> sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom
> the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice (including the
> next
> > + * paragraph) shall be included in all copies or substantial portions
> of the
> > + * Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. ?IN NO EVENT
> SHALL
> > + * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
> DAMAGES OR
> > + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
OTHERWISE,
> > + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE
> OR
> > + * OTHER DEALINGS IN THE SOFTWARE.
> > + */
> > +
> > +#include "drmP.h"
> > +#include "drm_crtc.h"
> > +#include "drm_crtc_helper.h"
> > +
> > +#include "samsung_drm_fb.h"
> > +#include "samsung_drm_buf.h"
> > +#include "samsung_drm_gem.h"
> > +
> > +#define to_samsung_fb(x) ? ? ? container_of(x, struct samsung_drm_fb,
fb)
> > +
> > +struct samsung_drm_fb {
> > + ? ? ? struct drm_framebuffer ? ? ? ? ?fb;
> > + ? ? ? struct drm_file ? ? ? ? ? ? ? ? *file_priv;
> > + ? ? ? struct samsung_drm_gem_obj ? ? ?*samsung_gem_obj;
> > + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?is_default;
> > +
> > + ? ? ? /* samsung gem object handle. */
> > + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?gem_handle;
> > + ? ? ? /* unique id to buffer object. */
> > + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?id;
> > +
> > + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?fb_size;
> > + ? ? ? unsigned long ? ? ? ? ? ? ? ? ? paddr;
> > + ? ? ? void __iomem ? ? ? ? ? ? ? ? ? ?*vaddr;
> > +};
> > +
> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > +{
> > + ? ? ? struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > + ? ? ? int ret;
> > +
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? drm_framebuffer_cleanup(fb);
> > +
> > + ? ? ? if (samsung_fb->is_default) {
> > + ? ? ? ? ? ? ? ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? samsung_fb->gem_handle);
> 
> why not keep the gem buffer ptr, and do something like:
> 
>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> 
> this way, you get the right behavior if someone somewhere else took a
> ref to the gem buffer object?  And it avoids needing to keep the
> file_priv ptr in the fb (which seems a bit strange)
> 
> 
Yes, at booting time, one gem object is created. this is for linux
framebuffer and used as default buffer. register_framebuffer function is
called one time at booting time by drm driver. but when this driver is built
as modules, this gem object should be released with rmmod modules. and the
reason fb has file point is that drm_gem_handle_delete requests it to
release a gem object. our driver considered modularization also. If there is
any point I missed, give me any comment please. Thank you.

> > + ? ? ? ? ? ? ? if (ret < 0)
> > + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to delete drm_gem_handle.\n");
> > + ? ? ? }
> > +
> > + ? ? ? kfree(samsung_fb);
> > +}
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c
> b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> > new file mode 100644
> > index 0000000..1e167a6
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> [snip]
> > +
> > +static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
> > +{
> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > + ? ? ? return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
> > +}
> 
> fwiw, candidate to move to drm_gem helper fxn (I have same in my
> driver.. but this wasn't included in my earlier patch to add a couple
> common drm_gem helper fxns like create/free_mmap_offset)
> 

I referred to your code. :) and I wish this code is moved to mainline to be
used commonly.

> > +
> > +/**
> > + * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for
> an object
> > + * @obj: obj in question
> > + *
> > + * GEM memory mapping works by handing back to userspace a fake mmap
> offset
> > + * it can use in a subsequent mmap(2) call. ?The DRM core code then
> looks
> > + * up the object based on the offset and sets up the various memory
> mapping
> > + * structures.
> > + *
> > + * This routine allocates and attaches a fake offset for @obj.
> > + */
> > +static int
> > +samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)
> 
> some of these other fxns you might at least want to include a comment
> to replace w/ drm_gem helpers once that patch lands.. your's would be
> the 4th driver w/ nearly identical code (well, if you can count
> non-mainline drivers) ;-)
> 
Ok, I know you had posted this patch to be used commonly. but it isn't
applied to dri git. If applied, I will change it to common fxn but for now I
will add a comment.
> > +{
> > + ? ? ? struct drm_device *dev = obj->dev;
> > + ? ? ? struct drm_gem_mm *mm = dev->mm_private;
> > + ? ? ? struct drm_map_list *list;
> > + ? ? ? struct drm_local_map *map;
> > + ? ? ? int ret = 0;
> > +
> > + ? ? ? /* Set the object up for mmap'ing */
> > + ? ? ? list = &obj->map_list;
> > + ? ? ? list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
> > + ? ? ? if (!list->map)
> > + ? ? ? ? ? ? ? return -ENOMEM;
> > +
> > + ? ? ? map = list->map;
> > + ? ? ? map->type = _DRM_GEM;
> > + ? ? ? map->size = obj->size;
> > + ? ? ? map->handle = obj;
> > +
> > + ? ? ? /* Get a DRM GEM mmap offset allocated... */
> > + ? ? ? list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? obj->size /
PAGE_SIZE,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 0, 0);
> > + ? ? ? if (!list->file_offset_node) {
> > + ? ? ? ? ? ? ? DRM_ERROR("failed to allocate offset for bo %d\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? obj->name);
> > + ? ? ? ? ? ? ? ret = -ENOSPC;
> > + ? ? ? ? ? ? ? goto out_free_list;
> > + ? ? ? }
> > +
> > + ? ? ? list->file_offset_node =
drm_mm_get_block(list->file_offset_node,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? obj->size / PAGE_SIZE,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 0);
> > + ? ? ? if (!list->file_offset_node) {
> > + ? ? ? ? ? ? ? ret = -ENOMEM;
> > + ? ? ? ? ? ? ? goto out_free_list;
> > + ? ? ? }
> > +
> > + ? ? ? list->hash.key = list->file_offset_node->start;
> > + ? ? ? ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
> > + ? ? ? if (ret) {
> > + ? ? ? ? ? ? ? DRM_ERROR("failed to add to map hash\n");
> > + ? ? ? ? ? ? ? goto out_free_mm;
> > + ? ? ? }
> > +
> > + ? ? ? return 0;
> > +
> > +out_free_mm:
> > + ? ? ? drm_mm_put_block(list->file_offset_node);
> > +out_free_list:
> > + ? ? ? kfree(list->map);
> > + ? ? ? list->map = NULL;
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static void
> > +samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
> > +{
> > + ? ? ? struct drm_device *dev = obj->dev;
> > + ? ? ? struct drm_gem_mm *mm = dev->mm_private;
> > + ? ? ? struct drm_map_list *list = &obj->map_list;
> > +
> > + ? ? ? drm_ht_remove_item(&mm->offset_hash, &list->hash);
> > + ? ? ? drm_mm_put_block(list->file_offset_node);
> > + ? ? ? kfree(list->map);
> > + ? ? ? list->map = NULL;
> > +}
> [snip]
> > +struct samsung_drm_gem_obj *
> > + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file *file_priv,
> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int handle)
> > +{
> > + ? ? ? struct drm_gem_object *gem_obj;
> > +
> > + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> > + ? ? ? if (!gem_obj) {
> > + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered to
> lookup.\n");
> > + ? ? ? ? ? ? ? return NULL;
> > + ? ? ? }
> > +
> > + ? ? ? /**
> > + ? ? ? ?* unreference refcount of the gem object.
> > + ? ? ? ?* at drm_gem_object_lookup(), the gem object was referenced.
> > + ? ? ? ?*/
> > + ? ? ? drm_gem_object_unreference(gem_obj);
> 
> this doesn't seem right, to drop the reference before you use the
> buffer elsewhere..
> 
No, see drm_gem_object_lookup fxn. at this function, if there is a object
found then drm_gem_object_reference is called to increase refcount of this
object. if there is any missing point, give me any comment please. thank
you.

> > + ? ? ? return to_samsung_gem_obj(gem_obj);
> > +}
> > +
> [snip]
> > diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
> > new file mode 100644
> > index 0000000..05dc460
> > --- /dev/null
> > +++ b/include/drm/samsung_drm.h
> [snip]
> > +/**
> > + * User-desired buffer creation information structure.
> > + *
> > + * @usr_addr: an address allocated by user process and this address
> > + * ? ? would be mmapped to physical region by fault handler.
> > + * @size: requested size for the object.
> > + * ? ? - this size value would be page-aligned internally.
> > + * @flags: user request for setting memory type or cache attributes.
> > + * @handle: returned handle for the object.
> > + */
> > +struct drm_samsung_gem_create {
> > + ? ? ? unsigned int usr_addr;
> 
> usr_addr seems dangerous.. and unused?  Maybe can be removed?
> 
Actually, usr_addr isn't used anywhere, so I removed it already and Dave
pointed out it before.

> also, I sort of prefer using stdint types for userspace<->kernel ioctl
> structs.
> 
Ok, I will use stdint types for it. thank you.

> > + ? ? ? unsigned int size;
> > + ? ? ? unsigned int flags;
> > +
> > + ? ? ? unsigned int handle;
> > +};
> > +
> > +/**
> > + * A structure for getting buffer offset.
> > + *
> > + * @handle: a pointer to gem object created.
> > + * @offset: relatived offset value of the memory region allocated.
> > + * ? ? - this value should be set by user.
> > + */
> > +struct drm_samsung_gem_map_off {
> > + ? ? ? unsigned int handle;
> > + ? ? ? uint64_t offset;
> > +};
> > +
> > +/**
> > + * A structure for mapping buffer.
> > + *
> > + * @handle: a pointer to gem object created.
> > + * @offset: relatived offset value of the memory region allocated.
> > + * ? ? - this value should be set by user.
> > + * @size: memory size to be mapped.
> > + * @mapped: user virtual address to be mapped.
> > + */
> > +struct drm_samsung_gem_mmap {
> > + ? ? ? unsigned int handle;
> > + ? ? ? uint64_t offset;
> > + ? ? ? unsigned int size;
> > +
> > + ? ? ? unsigned int mapped;
> 
> I guess this should be obtained by passing offset back to mmap() syscall
No, as I gave you the words about it before, the offset means physical
memory offset to be mapped to user space. and mapped has user virtual
address from do_mmap.
refer to previous answer please: <
http://www.spinics.net/lists/dri-devel/msg13880.html >
> 
> BR,
> -R

Thank you for your comments and advices again. it's been very useful.  :)

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  6:51     ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-31  6:51 UTC (permalink / raw)
  To: 'Rob Clark'
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

Hello, Rob.
Below is my answers and questions. and could you please include me as CC
when you post your driver?

Thank you.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark@gmail.com]
> Sent: Wednesday, August 31, 2011 10:58 AM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> Hi Inki,
> 
> Sorry for slightly overdue review.. it took a little while to go
> through the whole thing
> 

I  will be always pleased you to give me your advices and comments anytime.
:)

> comments in-line below
> 
> 
> On Fri, Aug 26, 2011 at 6:47 AM, Inki Dae <inki.dae@samsung.com> wrote:
> > This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables
> only FIMD yet
> > but we will add HDMI support also in the future.
> >
> > this patch is based on git repository below:
> > git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> > branch name: drm-next
> > commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
> >
> > you can refer to our working repository below:
> > http://git.infradead.org/users/kmpark/linux-2.6-samsung
> > branch name: samsung-drm
> >
> > We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> > based on Linux framebuffer) but couldn't so because lowlevel codes
> > of s3c-fb.c are included internally and so FIMD module of this driver
> has
> > its own lowlevel codes.
> >
> > We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> > for buffer allocation. by using DMA API, we could use CMA later.
> >
> > Refer to this link for CMA(Continuous Memory Allocator):
> > http://lkml.org/lkml/2011/7/20/45
> >
> > this driver supports only physically continuous memory(non-iommu).
> >
> > Links to previous versions of the patchset:
> > v1: < https://lwn.net/Articles/454380/ >
> > v2: < http://www.spinics.net/lists/kernel/msg1224275.html >
> >
> > Changelog v2:
> > DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
> >
> >     this feature maps user address space to physical memory region
> >     once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
> >
> > DRM: code clean and add exception codes.
> >
> > Changelog v3:
> > DRM: Support multiple irq.
> >
> >     FIMD and HDMI have their own irq handler but DRM Framework can
> regiter only one irq handler
> >     this patch supports mutiple irq for Samsung SoC.
> >
> > DRM: Consider modularization.
> >
> >     each DRM, FIMD could be built as a module.
> >
> > DRM: Have indenpendent crtc object.
> >
> >     crtc isn't specific to SoC Platform so this patch gets a crtc to be
> used as common object.
> >     created crtc could be attached to any encoder object.
> >
> > DRM: code clean and add exception codes.
> >
> > Signed-off-by: Inki Dae <inki.dae@samsung.com>
> > Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
> > Signed-off-by: SeungWoo Kim <sw0312.kim@samsung.com>
> > Signed-off-by: kyungmin.park <kyungmin.park@samsung.com>
> > ---
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.c
> b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> > new file mode 100644
> > index 0000000..563d07e
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.c
> [snip]
> > +static int lowlevel_buffer_allocate(struct drm_device *dev,
> > +               struct samsung_drm_buf_entry *entry)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> 
> very minor point, but DRM_DEBUG_KMS() already includes the function
> name.. not sure if __FILE__ is needed everywhere
> 
> > +
> > +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > +       if (!entry->paddr) {
> > +               DRM_ERROR("failed to allocate buffer.\n");
> > +               return -ENOMEM;
> > +       }
> > +
> > +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x),
> size(0x%x)\n",
> > +                       (unsigned int)entry->vaddr, entry->paddr, entry-
> >size);
> > +
> > +       return 0;
> > +}
> > +
> [snip]
> 
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h
> b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > new file mode 100644
> > index 0000000..d6a7e95
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> [snip]
> > +/**
> > + * samsung drm buffer entry structure.
> > + *
> > + * @paddr: physical address of allocated memory.
> > + * @vaddr: kernel virtual address of allocated memory.
> > + * @size: size of allocated memory.
> > + */
> > +struct samsung_drm_buf_entry {
> > +       unsigned int paddr;
> > +       void __iomem *vaddr;
> > +       unsigned int size;
> > +};
> 
> any reason not to combine this w/ samsung_drm_gem_obj to avoid extra
> levels of indirection?  (This would let you collapse some of the
> buffer allocation/deletion fxns into one level too.)
> 
I wanted it separates buffer allocator from gem but now seeing it, it
doesn't need to do so as you pointed out. I will combine it to
samsung_drm_gem_obj. thank you.

> > +
> > +/**
> > + * samsung drm buffer structure.
> > + *
> > + * @entry: pointer to samsung drm buffer entry object.
> > + * @flags: it means memory type to be alloated or cache attributes.
> > + * @handle: pointer to specific buffer object.
> > + * @id: unique id to specific buffer object.
> > + *
> > + * ps. this object would be transfered to user as kms_bo.handle so
> > + *     user can access to memory through kms_bo.handle.
> > + */
> > +struct samsung_drm_gem_obj {
> > +       struct drm_gem_object base;
> > +       struct samsung_drm_buf_entry *entry;
> > +       unsigned int flags;
> > +
> > +       unsigned int handle;
> > +       unsigned int id;
> > +};
> > +
> > +/* create new buffer object and memory region and add the object to
> list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buf_new(struct drm_device *dev,
> > +               unsigned int size);
> > +
> > +/* allocate physical memory and add its object to list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buf_create(struct drm_device
> *dev,
> > +               unsigned int size);
> > +
> > +/* remove allocated physical memory. */
> > +int samsung_drm_buf_destroy(struct drm_device *dev,
> > +               struct samsung_drm_gem_obj *in_obj);
> > +
> > +/* find object added to list. */
> > +struct samsung_drm_gem_obj *samsung_drm_buffer_find(struct drm_device
> *dev,
> > +               struct samsung_drm_gem_obj *in_obj, unsigned int paddr);
> 
> this doesn't appear to be used anywhere
> 
Oh, it’s my mistake. I will remove it. thank you.

> > +
> > +#endif
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_connector.c
> b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> > new file mode 100644
> > index 0000000..987a629
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_connector.c
> [snip]
> > +
> > +/* convert samsung_video_timings to drm_display_mode */
> > +static inline void
> > +convert_to_display_mode(struct fb_videomode *timing,
> > +                       struct drm_display_mode *mode)
> 
> I sort of prefer copy functions to copy right to left, Ie.
> 
> convert_foo(a, b)
> 
> copies from b to a.. somehow this feels more natural.  (Think a = b,
> the receiving arg is on the left.. same as w/ memcpy()).  Same applies
> w/ the inverse of this fxn.
> 
Ok, I will modify it like your saying. Thank you.

> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       mode->clock = timing->pixclock / 1000;
> > +
> > +       mode->hdisplay = timing->xres;
> > +       mode->hsync_start = mode->hdisplay + timing->left_margin;
> > +       mode->hsync_end = mode->hsync_start + timing->hsync_len;
> > +       mode->htotal = mode->hsync_end + timing->right_margin;
> > +
> > +       mode->vdisplay = timing->yres;
> > +       mode->vsync_start = mode->vdisplay + timing->upper_margin;
> > +       mode->vsync_end = mode->vsync_start + timing->vsync_len;
> > +       mode->vtotal = mode->vsync_end + timing->lower_margin;
> > +}
> > +
> > +/* convert drm_display_mode to samsung_video_timings */
> > +static inline void
> > +convert_to_video_timing(struct drm_display_mode *mode,
> > +                       struct fb_videomode *timing)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       timing->pixclock = mode->clock * 1000;
> > +
> > +       timing->xres = mode->hdisplay;
> > +       timing->left_margin = mode->hsync_start - mode->hdisplay;
> > +       timing->hsync_len = mode->hsync_end - mode->hsync_start;
> > +       timing->right_margin = mode->htotal - mode->hsync_end;
> > +
> > +       timing->yres = mode->vdisplay;
> > +       timing->upper_margin = mode->vsync_start - mode->vdisplay;
> > +       timing->vsync_len = mode->vsync_end - mode->vsync_start;
> > +       timing->lower_margin = mode->vtotal - mode->vsync_end;
> > +}
> > +
> > +static int samsung_drm_connector_get_modes(struct drm_connector
> *connector)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > +       unsigned int count;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       if (display && display->get_edid) {
> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> > +               if (!edid) {
> > +                       DRM_ERROR("failed to allocate edid\n");
> > +                       return 0;
> > +               }
> > +
> > +               display->get_edid(connector, edid, MAX_EDID);
> > +
> > +               drm_mode_connector_update_edid_property(connector,
edid);
> > +               count = drm_add_edid_modes(connector, edid);
> > +
> > +               kfree(connector->display_info.raw_edid);
> > +               connector->display_info.raw_edid = edid;
> > +       } else {
> > +               struct drm_display_mode *mode =
drm_mode_create(connector-
> >dev);
> > +               struct fb_videomode *timing;
> > +
> > +               if (display && display->get_timing)
> > +                       timing = display->get_timing();
> 
> also seems a bit weird to not do: display->get_timings(display)..
> 
> maybe you don't have the case of multiple instances of the same
> display time.. but maybe someday you need that.  (Maybe this is just
> an artifact of how the API of your current lower layer driver is.. but
> maybe now is a good time to think about those APIs)
> 

display is static object set by hardware specific driver such as display
controller or hdmi. each hardware specific driver has their own display
object statically. You can refer to the definition in samsung_drm_fimd.c.
and I understood a encoder and a connector is 1:1 relationship, when any
hardware specific driver is probed, they would be created with a manager
that includes their own display object as pointer. Could you please give me
more comments why display object should be considered for multiple
instances? I am afraid there is any part I don't care.

thank you.

> Same comment about not passing the display instance ptr to some of the
> other display fxn ptrs elsewhere in the code.
> 
> > +               else
> > +                       return 0;
> > +
> > +               convert_to_display_mode(timing, mode);
> > +
> > +               mode->type = DRM_MODE_TYPE_DRIVER |
> DRM_MODE_TYPE_PREFERRED;
> > +               drm_mode_set_name(mode);
> > +               drm_mode_probed_add(connector, mode);
> > +
> > +               count = 1;
> > +       }
> > +
> > +       return count;
> > +}
> > +
> > +static int samsung_drm_connector_mode_valid(struct drm_connector
> *connector,
> > +                                           struct drm_display_mode
*mode)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > +       struct fb_videomode timing;
> > +       int ret = MODE_BAD;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       convert_to_video_timing(mode, &timing);
> > +
> > +       if (display && display->check_timing)
> > +               if (!display->check_timing((void *)&timing))
> > +                       ret = MODE_OK;
> > +
> > +       return ret;
> > +}
> > +
> > +struct drm_encoder *samsung_drm_best_encoder(struct drm_connector
> *connector)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +       return to_samsung_connector(connector)->encoder;
> > +}
> > +
> > +static struct drm_connector_helper_funcs samsung_connector_helper_funcs
> = {
> > +       .get_modes      = samsung_drm_connector_get_modes,
> > +       .mode_valid     = samsung_drm_connector_mode_valid,
> > +       .best_encoder   = samsung_drm_best_encoder,
> > +};
> > +
> > +/* get detection status of display device. */
> > +static enum drm_connector_status
> > +samsung_drm_connector_detect(struct drm_connector *connector, bool
> force)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +       struct samsung_drm_display *display =
> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > +       unsigned int ret = connector_status_unknown;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       if (display && display->is_connected) {
> > +               if (display->is_connected())
> > +                       ret = connector_status_connected;
> > +               else
> > +                       ret = connector_status_disconnected;
> > +       }
> > +
> > +       return ret;
> > +}
> > +
> > +static void samsung_drm_connector_destroy(struct drm_connector
> *connector)
> > +{
> > +       struct samsung_drm_connector *samsung_connector =
> > +               to_samsung_connector(connector);
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       drm_sysfs_connector_remove(connector);
> > +       drm_connector_cleanup(connector);
> > +       connector->dev->mode_config.num_connector--;
> 
> I wonder if num_connector-- should be in drm_connector_cleanup()..
> 
> I missed this in OMAP driver, but it seems that none of the other drm
> drivers are directly decrementing this.. and it would seem more
> symmetric for it to be in drm_connector_cleanup().  But would be

For this, it was already commented by Joonyoun Shim.

> interesting to see what others think.  (I guess no one has really
> dealt w/ dynamic creation/deletion of connectors yet?)
> 

I think our driver creates encoder and connector dynamically. please, see
samsung_drm_subdrv_probe() of samsung_drm_core.c. hardware specific drivers
such as display controller and hdmi are registered to
samsung_drm_subdrv_list and when drm driver is probed, once subdrv->probe
callback is called, a encoder and a connector would be created dynamically.
if there is any missing point, feel free to give me your comments please.

Thank you.

> > +       kfree(samsung_connector);
> > +}
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> > new file mode 100644
> > index 0000000..1c5a70a
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_crtc.c
> [snip]
> > +
> > +struct samsung_drm_crtc {
> > +       struct drm_crtc                 drm_crtc;
> > +       struct samsung_drm_overlay      overlay;
> 
> 
> I guess you are looking fwd to drm_plane support too :-)
> 
> 
> > +       unsigned int                    pipe;
> > +};
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_fb.c
> b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> > new file mode 100644
> > index 0000000..42096eb
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_fb.c
> > @@ -0,0 +1,248 @@
> > +/*
> > + * Copyright (c) 2011 Samsung Electronics Co., Ltd.
> > + * Authors:
> > + *     Inki Dae <inki.dae@samsung.com>
> > + *     Joonyoung Shim <jy0922.shim@samsung.com>
> > + *     SeungWoo Kim <sw0312.kim@samsung.com>
> > + *
> > + * Permission is hereby granted, free of charge, to any person
> obtaining a
> > + * copy of this software and associated documentation files (the
> "Software"),
> > + * to deal in the Software without restriction, including without
> limitation
> > + * the rights to use, copy, modify, merge, publish, distribute,
> sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom
> the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice (including the
> next
> > + * paragraph) shall be included in all copies or substantial portions
> of the
> > + * Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT
> SHALL
> > + * VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
> DAMAGES OR
> > + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
OTHERWISE,
> > + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE
> OR
> > + * OTHER DEALINGS IN THE SOFTWARE.
> > + */
> > +
> > +#include "drmP.h"
> > +#include "drm_crtc.h"
> > +#include "drm_crtc_helper.h"
> > +
> > +#include "samsung_drm_fb.h"
> > +#include "samsung_drm_buf.h"
> > +#include "samsung_drm_gem.h"
> > +
> > +#define to_samsung_fb(x)       container_of(x, struct samsung_drm_fb,
fb)
> > +
> > +struct samsung_drm_fb {
> > +       struct drm_framebuffer          fb;
> > +       struct drm_file                 *file_priv;
> > +       struct samsung_drm_gem_obj      *samsung_gem_obj;
> > +       unsigned int                    is_default;
> > +
> > +       /* samsung gem object handle. */
> > +       unsigned int                    gem_handle;
> > +       /* unique id to buffer object. */
> > +       unsigned int                    id;
> > +
> > +       unsigned int                    fb_size;
> > +       unsigned long                   paddr;
> > +       void __iomem                    *vaddr;
> > +};
> > +
> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > +{
> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > +       int ret;
> > +
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       drm_framebuffer_cleanup(fb);
> > +
> > +       if (samsung_fb->is_default) {
> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > +                               samsung_fb->gem_handle);
> 
> why not keep the gem buffer ptr, and do something like:
> 
>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> 
> this way, you get the right behavior if someone somewhere else took a
> ref to the gem buffer object?  And it avoids needing to keep the
> file_priv ptr in the fb (which seems a bit strange)
> 
> 
Yes, at booting time, one gem object is created. this is for linux
framebuffer and used as default buffer. register_framebuffer function is
called one time at booting time by drm driver. but when this driver is built
as modules, this gem object should be released with rmmod modules. and the
reason fb has file point is that drm_gem_handle_delete requests it to
release a gem object. our driver considered modularization also. If there is
any point I missed, give me any comment please. Thank you.

> > +               if (ret < 0)
> > +                       DRM_ERROR("failed to delete drm_gem_handle.\n");
> > +       }
> > +
> > +       kfree(samsung_fb);
> > +}
> > +
> [snip]
> > diff --git a/drivers/gpu/drm/samsung/samsung_drm_gem.c
> b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> > new file mode 100644
> > index 0000000..1e167a6
> > --- /dev/null
> > +++ b/drivers/gpu/drm/samsung/samsung_drm_gem.c
> [snip]
> > +
> > +static unsigned int get_gem_mmap_offset(struct drm_gem_object *obj)
> > +{
> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > +
> > +       return (unsigned int)obj->map_list.hash.key << PAGE_SHIFT;
> > +}
> 
> fwiw, candidate to move to drm_gem helper fxn (I have same in my
> driver.. but this wasn't included in my earlier patch to add a couple
> common drm_gem helper fxns like create/free_mmap_offset)
> 

I referred to your code. :) and I wish this code is moved to mainline to be
used commonly.

> > +
> > +/**
> > + * samsung_drm_gem_create_mmap_offset - create a fake mmap offset for
> an object
> > + * @obj: obj in question
> > + *
> > + * GEM memory mapping works by handing back to userspace a fake mmap
> offset
> > + * it can use in a subsequent mmap(2) call.  The DRM core code then
> looks
> > + * up the object based on the offset and sets up the various memory
> mapping
> > + * structures.
> > + *
> > + * This routine allocates and attaches a fake offset for @obj.
> > + */
> > +static int
> > +samsung_drm_gem_create_mmap_offset(struct drm_gem_object *obj)
> 
> some of these other fxns you might at least want to include a comment
> to replace w/ drm_gem helpers once that patch lands.. your's would be
> the 4th driver w/ nearly identical code (well, if you can count
> non-mainline drivers) ;-)
> 
Ok, I know you had posted this patch to be used commonly. but it isn't
applied to dri git. If applied, I will change it to common fxn but for now I
will add a comment.
> > +{
> > +       struct drm_device *dev = obj->dev;
> > +       struct drm_gem_mm *mm = dev->mm_private;
> > +       struct drm_map_list *list;
> > +       struct drm_local_map *map;
> > +       int ret = 0;
> > +
> > +       /* Set the object up for mmap'ing */
> > +       list = &obj->map_list;
> > +       list->map = kzalloc(sizeof(struct drm_map_list), GFP_KERNEL);
> > +       if (!list->map)
> > +               return -ENOMEM;
> > +
> > +       map = list->map;
> > +       map->type = _DRM_GEM;
> > +       map->size = obj->size;
> > +       map->handle = obj;
> > +
> > +       /* Get a DRM GEM mmap offset allocated... */
> > +       list->file_offset_node = drm_mm_search_free(&mm->offset_manager,
> > +                                                   obj->size /
PAGE_SIZE,
> > +                                                   0, 0);
> > +       if (!list->file_offset_node) {
> > +               DRM_ERROR("failed to allocate offset for bo %d\n",
> > +                         obj->name);
> > +               ret = -ENOSPC;
> > +               goto out_free_list;
> > +       }
> > +
> > +       list->file_offset_node =
drm_mm_get_block(list->file_offset_node,
> > +                                                 obj->size / PAGE_SIZE,
> > +                                                 0);
> > +       if (!list->file_offset_node) {
> > +               ret = -ENOMEM;
> > +               goto out_free_list;
> > +       }
> > +
> > +       list->hash.key = list->file_offset_node->start;
> > +       ret = drm_ht_insert_item(&mm->offset_hash, &list->hash);
> > +       if (ret) {
> > +               DRM_ERROR("failed to add to map hash\n");
> > +               goto out_free_mm;
> > +       }
> > +
> > +       return 0;
> > +
> > +out_free_mm:
> > +       drm_mm_put_block(list->file_offset_node);
> > +out_free_list:
> > +       kfree(list->map);
> > +       list->map = NULL;
> > +
> > +       return ret;
> > +}
> > +
> > +static void
> > +samsung_drm_gem_free_mmap_offset(struct drm_gem_object *obj)
> > +{
> > +       struct drm_device *dev = obj->dev;
> > +       struct drm_gem_mm *mm = dev->mm_private;
> > +       struct drm_map_list *list = &obj->map_list;
> > +
> > +       drm_ht_remove_item(&mm->offset_hash, &list->hash);
> > +       drm_mm_put_block(list->file_offset_node);
> > +       kfree(list->map);
> > +       list->map = NULL;
> > +}
> [snip]
> > +struct samsung_drm_gem_obj *
> > +               find_samsung_drm_gem_object(struct drm_file *file_priv,
> > +                       struct drm_device *dev, unsigned int handle)
> > +{
> > +       struct drm_gem_object *gem_obj;
> > +
> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> > +       if (!gem_obj) {
> > +               DRM_LOG_KMS("a invalid gem object not registered to
> lookup.\n");
> > +               return NULL;
> > +       }
> > +
> > +       /**
> > +        * unreference refcount of the gem object.
> > +        * at drm_gem_object_lookup(), the gem object was referenced.
> > +        */
> > +       drm_gem_object_unreference(gem_obj);
> 
> this doesn't seem right, to drop the reference before you use the
> buffer elsewhere..
> 
No, see drm_gem_object_lookup fxn. at this function, if there is a object
found then drm_gem_object_reference is called to increase refcount of this
object. if there is any missing point, give me any comment please. thank
you.

> > +       return to_samsung_gem_obj(gem_obj);
> > +}
> > +
> [snip]
> > diff --git a/include/drm/samsung_drm.h b/include/drm/samsung_drm.h
> > new file mode 100644
> > index 0000000..05dc460
> > --- /dev/null
> > +++ b/include/drm/samsung_drm.h
> [snip]
> > +/**
> > + * User-desired buffer creation information structure.
> > + *
> > + * @usr_addr: an address allocated by user process and this address
> > + *     would be mmapped to physical region by fault handler.
> > + * @size: requested size for the object.
> > + *     - this size value would be page-aligned internally.
> > + * @flags: user request for setting memory type or cache attributes.
> > + * @handle: returned handle for the object.
> > + */
> > +struct drm_samsung_gem_create {
> > +       unsigned int usr_addr;
> 
> usr_addr seems dangerous.. and unused?  Maybe can be removed?
> 
Actually, usr_addr isn't used anywhere, so I removed it already and Dave
pointed out it before.

> also, I sort of prefer using stdint types for userspace<->kernel ioctl
> structs.
> 
Ok, I will use stdint types for it. thank you.

> > +       unsigned int size;
> > +       unsigned int flags;
> > +
> > +       unsigned int handle;
> > +};
> > +
> > +/**
> > + * A structure for getting buffer offset.
> > + *
> > + * @handle: a pointer to gem object created.
> > + * @offset: relatived offset value of the memory region allocated.
> > + *     - this value should be set by user.
> > + */
> > +struct drm_samsung_gem_map_off {
> > +       unsigned int handle;
> > +       uint64_t offset;
> > +};
> > +
> > +/**
> > + * A structure for mapping buffer.
> > + *
> > + * @handle: a pointer to gem object created.
> > + * @offset: relatived offset value of the memory region allocated.
> > + *     - this value should be set by user.
> > + * @size: memory size to be mapped.
> > + * @mapped: user virtual address to be mapped.
> > + */
> > +struct drm_samsung_gem_mmap {
> > +       unsigned int handle;
> > +       uint64_t offset;
> > +       unsigned int size;
> > +
> > +       unsigned int mapped;
> 
> I guess this should be obtained by passing offset back to mmap() syscall
No, as I gave you the words about it before, the offset means physical
memory offset to be mapped to user space. and mapped has user virtual
address from do_mmap.
refer to previous answer please: <
http://www.spinics.net/lists/dri-devel/msg13880.html >
> 
> BR,
> -R

Thank you for your comments and advices again. it's been very useful.  :)

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31  2:06     ` Konrad Rzeszutek Wilk
  (?)
@ 2011-08-31  7:33       ` Inki Dae
  -1 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-31  7:33 UTC (permalink / raw)
  To: 'Konrad Rzeszutek Wilk', 'Rob Clark'
  Cc: sw0312.kim, linux-kernel, dri-devel, kyungmin.park, linux-arm-kernel

Hello, Konrad Rzeszutek Wilk.

> -----Original Message-----
> From: Konrad Rzeszutek Wilk [mailto:konrad.wilk@oracle.com]
> Sent: Wednesday, August 31, 2011 11:07 AM
> To: Rob Clark
> Cc: Inki Dae; sw0312.kim@samsung.com; linux-kernel@vger.kernel.org; dri-
> devel@lists.freedesktop.org; kyungmin.park@samsung.com; linux-arm-
> kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> > > +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > > +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > > +       if (!entry->paddr) {
> > > +               DRM_ERROR("failed to allocate buffer.\n");
> > > +               return -ENOMEM;
> > > +       }
> > > +
> > > +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x),
> size(0x%x)\n",
> > > +                       (unsigned int)entry->vaddr, entry->paddr,
entry-
> >size);
> > > +
> > > +       return 0;
> > > +}
> > > +
> > [snip]
> >
> > > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h
> b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > > new file mode 100644
> > > index 0000000..d6a7e95
> > > --- /dev/null
> > > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > [snip]
> > > +/**
> > > + * samsung drm buffer entry structure.
> > > + *
> > > + * @paddr: physical address of allocated memory.
> > > + * @vaddr: kernel virtual address of allocated memory.
> > > + * @size: size of allocated memory.
> > > + */
> > > +struct samsung_drm_buf_entry {
> > > +       unsigned int paddr;
> 
> This could be made 'dma_addr_t' and then you can drop all of the
> casts to (dma_addr_t *).
> 

Ok, I will correct it right now. thank you.

> .. snip..
> > > +static int samsung_drm_connector_get_modes(struct drm_connector
> *connector)
> 
> Why not make the return be 'unsigned int'?
> 

Yes, I think so, but please, see drm_connector_helper_funcs structure of
drm_crtc_helper.h. get_modes callback has int type as return.

> .. snip..
> > > +/* get detection status of display device. */
> > > +static enum drm_connector_status
> > > +samsung_drm_connector_detect(struct drm_connector *connector, bool
> force)
> > > +{
> > > +       struct samsung_drm_connector *samsung_connector =
> > > +               to_samsung_connector(connector);
> > > +       struct samsung_drm_display *display =
> > > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > > +       unsigned int ret = connector_status_unknown;
> 
> Not 'enum drm_connector_status ret = connector_status_unknown' ?
> 

Oh, you are right, I will fix up it. thank you.

> > > +
> > > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > > +
> > > +       if (display && display->is_connected) {
> > > +               if (display->is_connected())
> > > +                       ret = connector_status_connected;
> > > +               else
> > > +                       ret = connector_status_disconnected;
> > > +       }
> > > +
> > > +       return ret;
> > > +}
> 
> .. snip..
> > > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > > +{
> > > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > > +       int ret;
> 
> Get rid of 'ret'

It seems that it doesn't need 'ret' but it needs to check 'ret' because of
drm_gem_handle_delete().

> > > +
> > > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > > +
> > > +       drm_framebuffer_cleanup(fb);
> > > +
> > > +       if (samsung_fb->is_default) {
> > > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > > +                               samsung_fb->gem_handle);
> >
> > why not keep the gem buffer ptr, and do something like:
> >
> >   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> >
> > this way, you get the right behavior if someone somewhere else took a
> > ref to the gem buffer object?  And it avoids needing to keep the
> > file_priv ptr in the fb (which seems a bit strange)
> >
> >
> > > +               if (ret < 0)
> > > +                       DRM_ERROR("failed to delete
drm_gem_handle.\n");
> 
> And just do the check on the function return value here. You are not using
> the 'ret' for anything.

Yes, right. it just prints out error message because drm_gem_handle_delete
function which is mainline function doesn't leave any error message. anyway
using 'ret' is not clear. so I will remove 'ret' and check
drm_gem_handle_delete function directly instead to print out error message.

> > > +       }
> > > +
> > > +       kfree(samsung_fb);
> > > +}
> > > +
> > [snip]
> 
> Hm, so I stopped here - just realized that I am missing some of the code
> and I should look at the original patch..

Thank you for your comments. it's been very useful. please give me your
comments and advices anytime then I will be pleased.



^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  7:33       ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-31  7:33 UTC (permalink / raw)
  To: linux-arm-kernel

Hello, Konrad Rzeszutek Wilk.

> -----Original Message-----
> From: Konrad Rzeszutek Wilk [mailto:konrad.wilk at oracle.com]
> Sent: Wednesday, August 31, 2011 11:07 AM
> To: Rob Clark
> Cc: Inki Dae; sw0312.kim at samsung.com; linux-kernel at vger.kernel.org; dri-
> devel at lists.freedesktop.org; kyungmin.park at samsung.com; linux-arm-
> kernel at lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> > > + ? ? ? entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > > + ? ? ? ? ? ? ? ? ? ? ? (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > > + ? ? ? if (!entry->paddr) {
> > > + ? ? ? ? ? ? ? DRM_ERROR("failed to allocate buffer.\n");
> > > + ? ? ? ? ? ? ? return -ENOMEM;
> > > + ? ? ? }
> > > +
> > > + ? ? ? DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x),
> size(0x%x)\n",
> > > + ? ? ? ? ? ? ? ? ? ? ? (unsigned int)entry->vaddr, entry->paddr,
entry-
> >size);
> > > +
> > > + ? ? ? return 0;
> > > +}
> > > +
> > [snip]
> >
> > > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h
> b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > > new file mode 100644
> > > index 0000000..d6a7e95
> > > --- /dev/null
> > > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > [snip]
> > > +/**
> > > + * samsung drm buffer entry structure.
> > > + *
> > > + * @paddr: physical address of allocated memory.
> > > + * @vaddr: kernel virtual address of allocated memory.
> > > + * @size: size of allocated memory.
> > > + */
> > > +struct samsung_drm_buf_entry {
> > > + ? ? ? unsigned int paddr;
> 
> This could be made 'dma_addr_t' and then you can drop all of the
> casts to (dma_addr_t *).
> 

Ok, I will correct it right now. thank you.

> .. snip..
> > > +static int samsung_drm_connector_get_modes(struct drm_connector
> *connector)
> 
> Why not make the return be 'unsigned int'?
> 

Yes, I think so, but please, see drm_connector_helper_funcs structure of
drm_crtc_helper.h. get_modes callback has int type as return.

> .. snip..
> > > +/* get detection status of display device. */
> > > +static enum drm_connector_status
> > > +samsung_drm_connector_detect(struct drm_connector *connector, bool
> force)
> > > +{
> > > + ? ? ? struct samsung_drm_connector *samsung_connector =
> > > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> > > + ? ? ? struct samsung_drm_display *display =
> > > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > > + ? ? ? unsigned int ret = connector_status_unknown;
> 
> Not 'enum drm_connector_status ret = connector_status_unknown' ?
> 

Oh, you are right, I will fix up it. thank you.

> > > +
> > > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > > +
> > > + ? ? ? if (display && display->is_connected) {
> > > + ? ? ? ? ? ? ? if (display->is_connected())
> > > + ? ? ? ? ? ? ? ? ? ? ? ret = connector_status_connected;
> > > + ? ? ? ? ? ? ? else
> > > + ? ? ? ? ? ? ? ? ? ? ? ret = connector_status_disconnected;
> > > + ? ? ? }
> > > +
> > > + ? ? ? return ret;
> > > +}
> 
> .. snip..
> > > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > > +{
> > > + ? ? ? struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > > + ? ? ? int ret;
> 
> Get rid of 'ret'

It seems that it doesn't need 'ret' but it needs to check 'ret' because of
drm_gem_handle_delete().

> > > +
> > > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> > > +
> > > + ? ? ? drm_framebuffer_cleanup(fb);
> > > +
> > > + ? ? ? if (samsung_fb->is_default) {
> > > + ? ? ? ? ? ? ? ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? samsung_fb->gem_handle);
> >
> > why not keep the gem buffer ptr, and do something like:
> >
> >   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> >
> > this way, you get the right behavior if someone somewhere else took a
> > ref to the gem buffer object?  And it avoids needing to keep the
> > file_priv ptr in the fb (which seems a bit strange)
> >
> >
> > > + ? ? ? ? ? ? ? if (ret < 0)
> > > + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to delete
drm_gem_handle.\n");
> 
> And just do the check on the function return value here. You are not using
> the 'ret' for anything.

Yes, right. it just prints out error message because drm_gem_handle_delete
function which is mainline function doesn't leave any error message. anyway
using 'ret' is not clear. so I will remove 'ret' and check
drm_gem_handle_delete function directly instead to print out error message.

> > > + ? ? ? }
> > > +
> > > + ? ? ? kfree(samsung_fb);
> > > +}
> > > +
> > [snip]
> 
> Hm, so I stopped here - just realized that I am missing some of the code
> and I should look at the original patch..

Thank you for your comments. it's been very useful. please give me your
comments and advices anytime then I will be pleased.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  7:33       ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-08-31  7:33 UTC (permalink / raw)
  To: 'Konrad Rzeszutek Wilk', 'Rob Clark'
  Cc: linux-arm-kernel, kyungmin.park, sw0312.kim, linux-kernel, dri-devel

Hello, Konrad Rzeszutek Wilk.

> -----Original Message-----
> From: Konrad Rzeszutek Wilk [mailto:konrad.wilk@oracle.com]
> Sent: Wednesday, August 31, 2011 11:07 AM
> To: Rob Clark
> Cc: Inki Dae; sw0312.kim@samsung.com; linux-kernel@vger.kernel.org; dri-
> devel@lists.freedesktop.org; kyungmin.park@samsung.com; linux-arm-
> kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> > > +       entry->vaddr = dma_alloc_writecombine(dev->dev, entry->size,
> > > +                       (dma_addr_t *)&entry->paddr, GFP_KERNEL);
> > > +       if (!entry->paddr) {
> > > +               DRM_ERROR("failed to allocate buffer.\n");
> > > +               return -ENOMEM;
> > > +       }
> > > +
> > > +       DRM_DEBUG_KMS("allocated : vaddr(0x%x), paddr(0x%x),
> size(0x%x)\n",
> > > +                       (unsigned int)entry->vaddr, entry->paddr,
entry-
> >size);
> > > +
> > > +       return 0;
> > > +}
> > > +
> > [snip]
> >
> > > diff --git a/drivers/gpu/drm/samsung/samsung_drm_buf.h
> b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > > new file mode 100644
> > > index 0000000..d6a7e95
> > > --- /dev/null
> > > +++ b/drivers/gpu/drm/samsung/samsung_drm_buf.h
> > [snip]
> > > +/**
> > > + * samsung drm buffer entry structure.
> > > + *
> > > + * @paddr: physical address of allocated memory.
> > > + * @vaddr: kernel virtual address of allocated memory.
> > > + * @size: size of allocated memory.
> > > + */
> > > +struct samsung_drm_buf_entry {
> > > +       unsigned int paddr;
> 
> This could be made 'dma_addr_t' and then you can drop all of the
> casts to (dma_addr_t *).
> 

Ok, I will correct it right now. thank you.

> .. snip..
> > > +static int samsung_drm_connector_get_modes(struct drm_connector
> *connector)
> 
> Why not make the return be 'unsigned int'?
> 

Yes, I think so, but please, see drm_connector_helper_funcs structure of
drm_crtc_helper.h. get_modes callback has int type as return.

> .. snip..
> > > +/* get detection status of display device. */
> > > +static enum drm_connector_status
> > > +samsung_drm_connector_detect(struct drm_connector *connector, bool
> force)
> > > +{
> > > +       struct samsung_drm_connector *samsung_connector =
> > > +               to_samsung_connector(connector);
> > > +       struct samsung_drm_display *display =
> > > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >display;
> > > +       unsigned int ret = connector_status_unknown;
> 
> Not 'enum drm_connector_status ret = connector_status_unknown' ?
> 

Oh, you are right, I will fix up it. thank you.

> > > +
> > > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > > +
> > > +       if (display && display->is_connected) {
> > > +               if (display->is_connected())
> > > +                       ret = connector_status_connected;
> > > +               else
> > > +                       ret = connector_status_disconnected;
> > > +       }
> > > +
> > > +       return ret;
> > > +}
> 
> .. snip..
> > > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> > > +{
> > > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> > > +       int ret;
> 
> Get rid of 'ret'

It seems that it doesn't need 'ret' but it needs to check 'ret' because of
drm_gem_handle_delete().

> > > +
> > > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> > > +
> > > +       drm_framebuffer_cleanup(fb);
> > > +
> > > +       if (samsung_fb->is_default) {
> > > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> > > +                               samsung_fb->gem_handle);
> >
> > why not keep the gem buffer ptr, and do something like:
> >
> >   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> >
> > this way, you get the right behavior if someone somewhere else took a
> > ref to the gem buffer object?  And it avoids needing to keep the
> > file_priv ptr in the fb (which seems a bit strange)
> >
> >
> > > +               if (ret < 0)
> > > +                       DRM_ERROR("failed to delete
drm_gem_handle.\n");
> 
> And just do the check on the function return value here. You are not using
> the 'ret' for anything.

Yes, right. it just prints out error message because drm_gem_handle_delete
function which is mainline function doesn't leave any error message. anyway
using 'ret' is not clear. so I will remove 'ret' and check
drm_gem_handle_delete function directly instead to print out error message.

> > > +       }
> > > +
> > > +       kfree(samsung_fb);
> > > +}
> > > +
> > [snip]
> 
> Hm, so I stopped here - just realized that I am missing some of the code
> and I should look at the original patch..

Thank you for your comments. it's been very useful. please give me your
comments and advices anytime then I will be pleased.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-26 11:47 ` Inki Dae
@ 2011-08-31  8:38   ` Thomas Hellstrom
  -1 siblings, 0 replies; 38+ messages in thread
From: Thomas Hellstrom @ 2011-08-31  8:38 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

On 08/26/2011 01:47 PM, Inki Dae wrote:
> This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
> but we will add HDMI support also in the future.
>
> this patch is based on git repository below:
> git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> branch name: drm-next
> commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
>
> you can refer to our working repository below:
> http://git.infradead.org/users/kmpark/linux-2.6-samsung
> branch name: samsung-drm
>
> We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> based on Linux framebuffer) but couldn't so because lowlevel codes
> of s3c-fb.c are included internally and so FIMD module of this driver has
> its own lowlevel codes.
>
> We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> for buffer allocation. by using DMA API, we could use CMA later.
>
> Refer to this link for CMA(Continuous Memory Allocator):
> http://lkml.org/lkml/2011/7/20/45
>
> this driver supports only physically continuous memory(non-iommu).
>
> Links to previous versions of the patchset:
> v1:<  https://lwn.net/Articles/454380/>
> v2:<  http://www.spinics.net/lists/kernel/msg1224275.html>
>
> Changelog v2:
> DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
>
>       this feature maps user address space to physical memory region
>       once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
>
> DRM: code clean and add exception codes.
>
> Changelog v3:
> DRM: Support multiple irq.
>
>       FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
>       this patch supports mutiple irq for Samsung SoC.
>
> DRM: Consider modularization.
>
>       each DRM, FIMD could be built as a module.
>
> DRM: Have indenpendent crtc object.
>
>       crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
>       created crtc could be attached to any encoder object.
>
> DRM: code clean and add exception codes.
>
> S
>    

...

> +static struct drm_ioctl_desc samsung_ioctls[] = {
> +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
> +			DRM_UNLOCKED),
> +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
> +			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
> +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
> +			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
> +};
>    

What about security here? It looks to me like *any* user-space process 
can create a gem object and quickly exhaust available DMA memory space, 
potentially bringing the system down?

Likewise, there seems to be no owner check in the SAMSUNG_GEM_MMAP 
ioctl, allowing any user-space process unlimited graphics buffer access?

/Thomas




^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31  8:38   ` Thomas Hellstrom
  0 siblings, 0 replies; 38+ messages in thread
From: Thomas Hellstrom @ 2011-08-31  8:38 UTC (permalink / raw)
  To: linux-arm-kernel

On 08/26/2011 01:47 PM, Inki Dae wrote:
> This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables only FIMD yet
> but we will add HDMI support also in the future.
>
> this patch is based on git repository below:
> git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> branch name: drm-next
> commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
>
> you can refer to our working repository below:
> http://git.infradead.org/users/kmpark/linux-2.6-samsung
> branch name: samsung-drm
>
> We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> based on Linux framebuffer) but couldn't so because lowlevel codes
> of s3c-fb.c are included internally and so FIMD module of this driver has
> its own lowlevel codes.
>
> We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> for buffer allocation. by using DMA API, we could use CMA later.
>
> Refer to this link for CMA(Continuous Memory Allocator):
> http://lkml.org/lkml/2011/7/20/45
>
> this driver supports only physically continuous memory(non-iommu).
>
> Links to previous versions of the patchset:
> v1:<  https://lwn.net/Articles/454380/>
> v2:<  http://www.spinics.net/lists/kernel/msg1224275.html>
>
> Changelog v2:
> DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
>
>       this feature maps user address space to physical memory region
>       once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
>
> DRM: code clean and add exception codes.
>
> Changelog v3:
> DRM: Support multiple irq.
>
>       FIMD and HDMI have their own irq handler but DRM Framework can regiter only one irq handler
>       this patch supports mutiple irq for Samsung SoC.
>
> DRM: Consider modularization.
>
>       each DRM, FIMD could be built as a module.
>
> DRM: Have indenpendent crtc object.
>
>       crtc isn't specific to SoC Platform so this patch gets a crtc to be used as common object.
>       created crtc could be attached to any encoder object.
>
> DRM: code clean and add exception codes.
>
> S
>    

...

> +static struct drm_ioctl_desc samsung_ioctls[] = {
> +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
> +			DRM_UNLOCKED),
> +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
> +			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
> +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
> +			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
> +};
>    

What about security here? It looks to me like *any* user-space process 
can create a gem object and quickly exhaust available DMA memory space, 
potentially bringing the system down?

Likewise, there seems to be no owner check in the SAMSUNG_GEM_MMAP 
ioctl, allowing any user-space process unlimited graphics buffer access?

/Thomas

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31  6:51     ` Inki Dae
@ 2011-08-31 16:02       ` Rob Clark
  -1 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-08-31 16:02 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
> Hello, Rob.
> Below is my answers and questions. and could you please include me as CC
> when you post your driver?

sure thing


>> > +static int samsung_drm_connector_get_modes(struct drm_connector
>> *connector)
>> > +{
>> > +       struct samsung_drm_connector *samsung_connector =
>> > +               to_samsung_connector(connector);
>> > +       struct samsung_drm_display *display =
>> > +               samsung_drm_get_manager(samsung_connector->encoder)-
>> >display;
>> > +       unsigned int count;
>> > +
>> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> > +
>> > +       if (display && display->get_edid) {
>> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
>> > +               if (!edid) {
>> > +                       DRM_ERROR("failed to allocate edid\n");
>> > +                       return 0;
>> > +               }
>> > +
>> > +               display->get_edid(connector, edid, MAX_EDID);
>> > +
>> > +               drm_mode_connector_update_edid_property(connector,
> edid);
>> > +               count = drm_add_edid_modes(connector, edid);
>> > +
>> > +               kfree(connector->display_info.raw_edid);
>> > +               connector->display_info.raw_edid = edid;
>> > +       } else {
>> > +               struct drm_display_mode *mode =
> drm_mode_create(connector-
>> >dev);
>> > +               struct fb_videomode *timing;
>> > +
>> > +               if (display && display->get_timing)
>> > +                       timing = display->get_timing();
>>
>> also seems a bit weird to not do: display->get_timings(display)..
>>
>> maybe you don't have the case of multiple instances of the same
>> display time.. but maybe someday you need that.  (Maybe this is just
>> an artifact of how the API of your current lower layer driver is.. but
>> maybe now is a good time to think about those APIs)
>>
>
> display is static object set by hardware specific driver such as display
> controller or hdmi. each hardware specific driver has their own display
> object statically. You can refer to the definition in samsung_drm_fimd.c.
> and I understood a encoder and a connector is 1:1 relationship, when any
> hardware specific driver is probed, they would be created with a manager
> that includes their own display object as pointer. Could you please give me
> more comments why display object should be considered for multiple
> instances? I am afraid there is any part I don't care.
>
> thank you.

Just thinking hypothetically.. what if some future device had two hdmi
controllers.  Then you'd want two instances of the same display
object.

Although it seems this API is just internal to the DRM driver (which I
had not realized earlier), so it should be pretty easy to change later
if needed.


>> > +static void samsung_drm_connector_destroy(struct drm_connector
>> *connector)
>> > +{
>> > +       struct samsung_drm_connector *samsung_connector =
>> > +               to_samsung_connector(connector);
>> > +
>> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> > +
>> > +       drm_sysfs_connector_remove(connector);
>> > +       drm_connector_cleanup(connector);
>> > +       connector->dev->mode_config.num_connector--;
>>
>> I wonder if num_connector-- should be in drm_connector_cleanup()..
>>
>> I missed this in OMAP driver, but it seems that none of the other drm
>> drivers are directly decrementing this.. and it would seem more
>> symmetric for it to be in drm_connector_cleanup().  But would be
>
> For this, it was already commented by Joonyoun Shim.

And it seems (which I hadn't noticed earlier) already merged :-)

So I think this (and similar bit in encoder destructor) can be removed


>> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
>> > +{
>> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
>> > +       int ret;
>> > +
>> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> > +
>> > +       drm_framebuffer_cleanup(fb);
>> > +
>> > +       if (samsung_fb->is_default) {
>> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
>> > +                               samsung_fb->gem_handle);
>>
>> why not keep the gem buffer ptr, and do something like:
>>
>>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
>>
>> this way, you get the right behavior if someone somewhere else took a
>> ref to the gem buffer object?  And it avoids needing to keep the
>> file_priv ptr in the fb (which seems a bit strange)
>>
>>
> Yes, at booting time, one gem object is created. this is for linux
> framebuffer and used as default buffer. register_framebuffer function is
> called one time at booting time by drm driver. but when this driver is built
> as modules, this gem object should be released with rmmod modules. and the
> reason fb has file point is that drm_gem_handle_delete requests it to
> release a gem object. our driver considered modularization also. If there is
> any point I missed, give me any comment please. Thank you.

Oh, I missed the point that drm_gem_handle_delete() is calling
drm_gem_object_unreference_unlocked().

But this still seems a bit odd, but I guess the difference is you are
keeping the buffer handle, rather than the 'struct drm_gem_object'
ptr.  I'm not sure if there is a point to keeping track of the buffer
in handle form on the kernel side.  If instead you just keep a GEM
object ptr, you can just drm_gem_object_unreference{_unlocked}() when
you are done with it without having to special-case is_default stuff.

>> > +struct samsung_drm_gem_obj *
>> > +               find_samsung_drm_gem_object(struct drm_file *file_priv,
>> > +                       struct drm_device *dev, unsigned int handle)
>> > +{
>> > +       struct drm_gem_object *gem_obj;
>> > +
>> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> > +       if (!gem_obj) {
>> > +               DRM_LOG_KMS("a invalid gem object not registered to
>> lookup.\n");
>> > +               return NULL;
>> > +       }
>> > +
>> > +       /**
>> > +        * unreference refcount of the gem object.
>> > +        * at drm_gem_object_lookup(), the gem object was referenced.
>> > +        */
>> > +       drm_gem_object_unreference(gem_obj);
>>
>> this doesn't seem right, to drop the reference before you use the
>> buffer elsewhere..
>>
> No, see drm_gem_object_lookup fxn. at this function, if there is a object
> found then drm_gem_object_reference is called to increase refcount of this
> object. if there is any missing point, give me any comment please. thank
> you.


Right, but I think there is a reason it takes a reference... so that
the object doesn't get free'd from under your feet.  So pattern
should, I think, be:

  obj = lookup(...);
  ... do stuff w/ obj ...
  unreference(obj)

so the caller who is using the looked up obj should unref it when done

Instead, you have:

  obj = lookup(...);
  unreference(obj);
  ... do stuff w/ obj ...


BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-08-31 16:02       ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-08-31 16:02 UTC (permalink / raw)
  To: linux-arm-kernel

On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
> Hello, Rob.
> Below is my answers and questions. and could you please include me as CC
> when you post your driver?

sure thing


>> > +static int samsung_drm_connector_get_modes(struct drm_connector
>> *connector)
>> > +{
>> > + ? ? ? struct samsung_drm_connector *samsung_connector =
>> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
>> > + ? ? ? struct samsung_drm_display *display =
>> > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
>> >display;
>> > + ? ? ? unsigned int count;
>> > +
>> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
>> > +
>> > + ? ? ? if (display && display->get_edid) {
>> > + ? ? ? ? ? ? ? void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
>> > + ? ? ? ? ? ? ? if (!edid) {
>> > + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to allocate edid\n");
>> > + ? ? ? ? ? ? ? ? ? ? ? return 0;
>> > + ? ? ? ? ? ? ? }
>> > +
>> > + ? ? ? ? ? ? ? display->get_edid(connector, edid, MAX_EDID);
>> > +
>> > + ? ? ? ? ? ? ? drm_mode_connector_update_edid_property(connector,
> edid);
>> > + ? ? ? ? ? ? ? count = drm_add_edid_modes(connector, edid);
>> > +
>> > + ? ? ? ? ? ? ? kfree(connector->display_info.raw_edid);
>> > + ? ? ? ? ? ? ? connector->display_info.raw_edid = edid;
>> > + ? ? ? } else {
>> > + ? ? ? ? ? ? ? struct drm_display_mode *mode =
> drm_mode_create(connector-
>> >dev);
>> > + ? ? ? ? ? ? ? struct fb_videomode *timing;
>> > +
>> > + ? ? ? ? ? ? ? if (display && display->get_timing)
>> > + ? ? ? ? ? ? ? ? ? ? ? timing = display->get_timing();
>>
>> also seems a bit weird to not do: display->get_timings(display)..
>>
>> maybe you don't have the case of multiple instances of the same
>> display time.. but maybe someday you need that. ?(Maybe this is just
>> an artifact of how the API of your current lower layer driver is.. but
>> maybe now is a good time to think about those APIs)
>>
>
> display is static object set by hardware specific driver such as display
> controller or hdmi. each hardware specific driver has their own display
> object statically. You can refer to the definition in samsung_drm_fimd.c.
> and I understood a encoder and a connector is 1:1 relationship, when any
> hardware specific driver is probed, they would be created with a manager
> that includes their own display object as pointer. Could you please give me
> more comments why display object should be considered for multiple
> instances? I am afraid there is any part I don't care.
>
> thank you.

Just thinking hypothetically.. what if some future device had two hdmi
controllers.  Then you'd want two instances of the same display
object.

Although it seems this API is just internal to the DRM driver (which I
had not realized earlier), so it should be pretty easy to change later
if needed.


>> > +static void samsung_drm_connector_destroy(struct drm_connector
>> *connector)
>> > +{
>> > + ? ? ? struct samsung_drm_connector *samsung_connector =
>> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
>> > +
>> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
>> > +
>> > + ? ? ? drm_sysfs_connector_remove(connector);
>> > + ? ? ? drm_connector_cleanup(connector);
>> > + ? ? ? connector->dev->mode_config.num_connector--;
>>
>> I wonder if num_connector-- should be in drm_connector_cleanup()..
>>
>> I missed this in OMAP driver, but it seems that none of the other drm
>> drivers are directly decrementing this.. and it would seem more
>> symmetric for it to be in drm_connector_cleanup(). ?But would be
>
> For this, it was already commented by Joonyoun Shim.

And it seems (which I hadn't noticed earlier) already merged :-)

So I think this (and similar bit in encoder destructor) can be removed


>> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
>> > +{
>> > + ? ? ? struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
>> > + ? ? ? int ret;
>> > +
>> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
>> > +
>> > + ? ? ? drm_framebuffer_cleanup(fb);
>> > +
>> > + ? ? ? if (samsung_fb->is_default) {
>> > + ? ? ? ? ? ? ? ret = drm_gem_handle_delete(samsung_fb->file_priv,
>> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? samsung_fb->gem_handle);
>>
>> why not keep the gem buffer ptr, and do something like:
>>
>> ? drm_gem_object_unreference_unlocked(samsung_fb->bo)..
>>
>> this way, you get the right behavior if someone somewhere else took a
>> ref to the gem buffer object? ?And it avoids needing to keep the
>> file_priv ptr in the fb (which seems a bit strange)
>>
>>
> Yes, at booting time, one gem object is created. this is for linux
> framebuffer and used as default buffer. register_framebuffer function is
> called one time at booting time by drm driver. but when this driver is built
> as modules, this gem object should be released with rmmod modules. and the
> reason fb has file point is that drm_gem_handle_delete requests it to
> release a gem object. our driver considered modularization also. If there is
> any point I missed, give me any comment please. Thank you.

Oh, I missed the point that drm_gem_handle_delete() is calling
drm_gem_object_unreference_unlocked().

But this still seems a bit odd, but I guess the difference is you are
keeping the buffer handle, rather than the 'struct drm_gem_object'
ptr.  I'm not sure if there is a point to keeping track of the buffer
in handle form on the kernel side.  If instead you just keep a GEM
object ptr, you can just drm_gem_object_unreference{_unlocked}() when
you are done with it without having to special-case is_default stuff.

>> > +struct samsung_drm_gem_obj *
>> > + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file *file_priv,
>> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int handle)
>> > +{
>> > + ? ? ? struct drm_gem_object *gem_obj;
>> > +
>> > + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> > + ? ? ? if (!gem_obj) {
>> > + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered to
>> lookup.\n");
>> > + ? ? ? ? ? ? ? return NULL;
>> > + ? ? ? }
>> > +
>> > + ? ? ? /**
>> > + ? ? ? ?* unreference refcount of the gem object.
>> > + ? ? ? ?* at drm_gem_object_lookup(), the gem object was referenced.
>> > + ? ? ? ?*/
>> > + ? ? ? drm_gem_object_unreference(gem_obj);
>>
>> this doesn't seem right, to drop the reference before you use the
>> buffer elsewhere..
>>
> No, see drm_gem_object_lookup fxn. at this function, if there is a object
> found then drm_gem_object_reference is called to increase refcount of this
> object. if there is any missing point, give me any comment please. thank
> you.


Right, but I think there is a reason it takes a reference... so that
the object doesn't get free'd from under your feet.  So pattern
should, I think, be:

  obj = lookup(...);
  ... do stuff w/ obj ...
  unreference(obj)

so the caller who is using the looked up obj should unref it when done

Instead, you have:

  obj = lookup(...);
  unreference(obj);
  ... do stuff w/ obj ...


BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31  8:38   ` Thomas Hellstrom
  (?)
@ 2011-09-01  3:57     ` Inki Dae
  -1 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-01  3:57 UTC (permalink / raw)
  To: 'Thomas Hellstrom'
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

Hello Thomas.

> -----Original Message-----
> From: Thomas Hellstrom [mailto:thomas@shipmail.org]
> Sent: Wednesday, August 31, 2011 5:39 PM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On 08/26/2011 01:47 PM, Inki Dae wrote:
> > This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables
> only FIMD yet
> > but we will add HDMI support also in the future.
> >
> > this patch is based on git repository below:
> > git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> > branch name: drm-next
> > commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
> >
> > you can refer to our working repository below:
> > http://git.infradead.org/users/kmpark/linux-2.6-samsung
> > branch name: samsung-drm
> >
> > We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> > based on Linux framebuffer) but couldn't so because lowlevel codes
> > of s3c-fb.c are included internally and so FIMD module of this driver
> has
> > its own lowlevel codes.
> >
> > We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> > for buffer allocation. by using DMA API, we could use CMA later.
> >
> > Refer to this link for CMA(Continuous Memory Allocator):
> > http://lkml.org/lkml/2011/7/20/45
> >
> > this driver supports only physically continuous memory(non-iommu).
> >
> > Links to previous versions of the patchset:
> > v1:<  https://lwn.net/Articles/454380/>
> > v2:<  http://www.spinics.net/lists/kernel/msg1224275.html>
> >
> > Changelog v2:
> > DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
> >
> >       this feature maps user address space to physical memory region
> >       once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
> >
> > DRM: code clean and add exception codes.
> >
> > Changelog v3:
> > DRM: Support multiple irq.
> >
> >       FIMD and HDMI have their own irq handler but DRM Framework can
> regiter only one irq handler
> >       this patch supports mutiple irq for Samsung SoC.
> >
> > DRM: Consider modularization.
> >
> >       each DRM, FIMD could be built as a module.
> >
> > DRM: Have indenpendent crtc object.
> >
> >       crtc isn't specific to SoC Platform so this patch gets a crtc to
be
> used as common object.
> >       created crtc could be attached to any encoder object.
> >
> > DRM: code clean and add exception codes.
> >
> > S
> >
> 
> ...
> 
> > +static struct drm_ioctl_desc samsung_ioctls[] = {
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
> > +			DRM_UNLOCKED),
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
> > +			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
> > +			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
> > +};
> >
> 
> What about security here? It looks to me like *any* user-space process
> can create a gem object and quickly exhaust available DMA memory space,
> potentially bringing the system down?
> 
> Likewise, there seems to be no owner check in the SAMSUNG_GEM_MMAP
> ioctl, allowing any user-space process unlimited graphics buffer access?
> 
> /Thomas
> 

Right, we should consider security issue also as you mentioned above. thank
you for your pointing out.



^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-01  3:57     ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-01  3:57 UTC (permalink / raw)
  To: linux-arm-kernel

Hello Thomas.

> -----Original Message-----
> From: Thomas Hellstrom [mailto:thomas at shipmail.org]
> Sent: Wednesday, August 31, 2011 5:39 PM
> To: Inki Dae
> Cc: airlied at linux.ie; dri-devel at lists.freedesktop.org;
> sw0312.kim at samsung.com; linux-kernel at vger.kernel.org;
> kyungmin.park at samsung.com; linux-arm-kernel at lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On 08/26/2011 01:47 PM, Inki Dae wrote:
> > This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables
> only FIMD yet
> > but we will add HDMI support also in the future.
> >
> > this patch is based on git repository below:
> > git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> > branch name: drm-next
> > commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
> >
> > you can refer to our working repository below:
> > http://git.infradead.org/users/kmpark/linux-2.6-samsung
> > branch name: samsung-drm
> >
> > We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> > based on Linux framebuffer) but couldn't so because lowlevel codes
> > of s3c-fb.c are included internally and so FIMD module of this driver
> has
> > its own lowlevel codes.
> >
> > We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> > for buffer allocation. by using DMA API, we could use CMA later.
> >
> > Refer to this link for CMA(Continuous Memory Allocator):
> > http://lkml.org/lkml/2011/7/20/45
> >
> > this driver supports only physically continuous memory(non-iommu).
> >
> > Links to previous versions of the patchset:
> > v1:<  https://lwn.net/Articles/454380/>
> > v2:<  http://www.spinics.net/lists/kernel/msg1224275.html>
> >
> > Changelog v2:
> > DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
> >
> >       this feature maps user address space to physical memory region
> >       once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
> >
> > DRM: code clean and add exception codes.
> >
> > Changelog v3:
> > DRM: Support multiple irq.
> >
> >       FIMD and HDMI have their own irq handler but DRM Framework can
> regiter only one irq handler
> >       this patch supports mutiple irq for Samsung SoC.
> >
> > DRM: Consider modularization.
> >
> >       each DRM, FIMD could be built as a module.
> >
> > DRM: Have indenpendent crtc object.
> >
> >       crtc isn't specific to SoC Platform so this patch gets a crtc to
be
> used as common object.
> >       created crtc could be attached to any encoder object.
> >
> > DRM: code clean and add exception codes.
> >
> > S
> >
> 
> ...
> 
> > +static struct drm_ioctl_desc samsung_ioctls[] = {
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
> > +			DRM_UNLOCKED),
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
> > +			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
> > +			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
> > +};
> >
> 
> What about security here? It looks to me like *any* user-space process
> can create a gem object and quickly exhaust available DMA memory space,
> potentially bringing the system down?
> 
> Likewise, there seems to be no owner check in the SAMSUNG_GEM_MMAP
> ioctl, allowing any user-space process unlimited graphics buffer access?
> 
> /Thomas
> 

Right, we should consider security issue also as you mentioned above. thank
you for your pointing out.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-01  3:57     ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-01  3:57 UTC (permalink / raw)
  To: 'Thomas Hellstrom'
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

Hello Thomas.

> -----Original Message-----
> From: Thomas Hellstrom [mailto:thomas@shipmail.org]
> Sent: Wednesday, August 31, 2011 5:39 PM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On 08/26/2011 01:47 PM, Inki Dae wrote:
> > This patch is a DRM Driver for Samsung SoC Exynos4210 and now enables
> only FIMD yet
> > but we will add HDMI support also in the future.
> >
> > this patch is based on git repository below:
> > git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git,
> > branch name: drm-next
> > commit-id: bcc65fd8e929a9d9d34d814d6efc1d2793546922
> >
> > you can refer to our working repository below:
> > http://git.infradead.org/users/kmpark/linux-2.6-samsung
> > branch name: samsung-drm
> >
> > We tried to re-use lowlevel codes of the FIMD driver(s3c-fb.c
> > based on Linux framebuffer) but couldn't so because lowlevel codes
> > of s3c-fb.c are included internally and so FIMD module of this driver
> has
> > its own lowlevel codes.
> >
> > We used GEM framework for buffer management and DMA APIs(dma_alloc_*)
> > for buffer allocation. by using DMA API, we could use CMA later.
> >
> > Refer to this link for CMA(Continuous Memory Allocator):
> > http://lkml.org/lkml/2011/7/20/45
> >
> > this driver supports only physically continuous memory(non-iommu).
> >
> > Links to previous versions of the patchset:
> > v1:<  https://lwn.net/Articles/454380/>
> > v2:<  http://www.spinics.net/lists/kernel/msg1224275.html>
> >
> > Changelog v2:
> > DRM: add DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl command.
> >
> >       this feature maps user address space to physical memory region
> >       once user application requests DRM_IOCTL_SAMSUNG_GEM_MMAP ioctl.
> >
> > DRM: code clean and add exception codes.
> >
> > Changelog v3:
> > DRM: Support multiple irq.
> >
> >       FIMD and HDMI have their own irq handler but DRM Framework can
> regiter only one irq handler
> >       this patch supports mutiple irq for Samsung SoC.
> >
> > DRM: Consider modularization.
> >
> >       each DRM, FIMD could be built as a module.
> >
> > DRM: Have indenpendent crtc object.
> >
> >       crtc isn't specific to SoC Platform so this patch gets a crtc to
be
> used as common object.
> >       created crtc could be attached to any encoder object.
> >
> > DRM: code clean and add exception codes.
> >
> > S
> >
> 
> ...
> 
> > +static struct drm_ioctl_desc samsung_ioctls[] = {
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_CREATE, samsung_drm_gem_create_ioctl,
> > +			DRM_UNLOCKED),
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MAP_OFFSET,
> > +			samsung_drm_gem_map_offset_ioctl, DRM_UNLOCKED),
> > +	DRM_IOCTL_DEF_DRV(SAMSUNG_GEM_MMAP,
> > +			samsung_drm_gem_mmap_ioctl, DRM_UNLOCKED),
> > +};
> >
> 
> What about security here? It looks to me like *any* user-space process
> can create a gem object and quickly exhaust available DMA memory space,
> potentially bringing the system down?
> 
> Likewise, there seems to be no owner check in the SAMSUNG_GEM_MMAP
> ioctl, allowing any user-space process unlimited graphics buffer access?
> 
> /Thomas
> 

Right, we should consider security issue also as you mentioned above. thank
you for your pointing out.

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-08-31 16:02       ` Rob Clark
  (?)
@ 2011-09-01 13:06         ` Inki Dae
  -1 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-01 13:06 UTC (permalink / raw)
  To: 'Rob Clark'
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

Hello Rob.
Below is my comments. thank you.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark@gmail.com]
> Sent: Thursday, September 01, 2011 1:02 AM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
> > Hello, Rob.
> > Below is my answers and questions. and could you please include me as CC
> > when you post your driver?
> 
> sure thing
> 
> 
> >> > +static int samsung_drm_connector_get_modes(struct drm_connector
> >> *connector)
> >> > +{
> >> > +       struct samsung_drm_connector *samsung_connector =
> >> > +               to_samsung_connector(connector);
> >> > +       struct samsung_drm_display *display =
> >> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >> >display;
> >> > +       unsigned int count;
> >> > +
> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > +       if (display && display->get_edid) {
> >> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> >> > +               if (!edid) {
> >> > +                       DRM_ERROR("failed to allocate edid\n");
> >> > +                       return 0;
> >> > +               }
> >> > +
> >> > +               display->get_edid(connector, edid, MAX_EDID);
> >> > +
> >> > +               drm_mode_connector_update_edid_property(connector,
> > edid);
> >> > +               count = drm_add_edid_modes(connector, edid);
> >> > +
> >> > +               kfree(connector->display_info.raw_edid);
> >> > +               connector->display_info.raw_edid = edid;
> >> > +       } else {
> >> > +               struct drm_display_mode *mode =
> > drm_mode_create(connector-
> >> >dev);
> >> > +               struct fb_videomode *timing;
> >> > +
> >> > +               if (display && display->get_timing)
> >> > +                       timing = display->get_timing();
> >>
> >> also seems a bit weird to not do: display->get_timings(display)..
> >>
> >> maybe you don't have the case of multiple instances of the same
> >> display time.. but maybe someday you need that.  (Maybe this is just
> >> an artifact of how the API of your current lower layer driver is.. but
> >> maybe now is a good time to think about those APIs)
> >>
> >
> > display is static object set by hardware specific driver such as display
> > controller or hdmi. each hardware specific driver has their own display
> > object statically. You can refer to the definition in
samsung_drm_fimd.c.
> > and I understood a encoder and a connector is 1:1 relationship, when any
> > hardware specific driver is probed, they would be created with a manager
> > that includes their own display object as pointer. Could you please give
> me
> > more comments why display object should be considered for multiple
> > instances? I am afraid there is any part I don't care.
> >
> > thank you.
> 
> Just thinking hypothetically.. what if some future device had two hdmi
> controllers.  Then you'd want two instances of the same display
> object.
> 
> Although it seems this API is just internal to the DRM driver (which I
> had not realized earlier), so it should be pretty easy to change later
> if needed.
> 

You are right. I guess we should consider multiple instances to display
object because the Exynos hardware has two display controllers and also each
display controller needs one display panel with sharing same driver. thank
you for your pointing. :)

> 
> >> > +static void samsung_drm_connector_destroy(struct drm_connector
> >> *connector)
> >> > +{
> >> > +       struct samsung_drm_connector *samsung_connector =
> >> > +               to_samsung_connector(connector);
> >> > +
> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > +       drm_sysfs_connector_remove(connector);
> >> > +       drm_connector_cleanup(connector);
> >> > +       connector->dev->mode_config.num_connector--;
> >>
> >> I wonder if num_connector-- should be in drm_connector_cleanup()..
> >>
> >> I missed this in OMAP driver, but it seems that none of the other drm
> >> drivers are directly decrementing this.. and it would seem more
> >> symmetric for it to be in drm_connector_cleanup().  But would be
> >
> > For this, it was already commented by Joonyoun Shim.
> 
> And it seems (which I hadn't noticed earlier) already merged :-)
> 
> So I think this (and similar bit in encoder destructor) can be removed
> 
> 
> >> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> >> > +{
> >> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> >> > +       int ret;
> >> > +
> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > +       drm_framebuffer_cleanup(fb);
> >> > +
> >> > +       if (samsung_fb->is_default) {
> >> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> >> > +                               samsung_fb->gem_handle);
> >>
> >> why not keep the gem buffer ptr, and do something like:
> >>
> >>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> >>
> >> this way, you get the right behavior if someone somewhere else took a
> >> ref to the gem buffer object?  And it avoids needing to keep the
> >> file_priv ptr in the fb (which seems a bit strange)
> >>
> >>
> > Yes, at booting time, one gem object is created. this is for linux
> > framebuffer and used as default buffer. register_framebuffer function is
> > called one time at booting time by drm driver. but when this driver is
> built
> > as modules, this gem object should be released with rmmod modules. and
> the
> > reason fb has file point is that drm_gem_handle_delete requests it to
> > release a gem object. our driver considered modularization also. If
> there is
> > any point I missed, give me any comment please. Thank you.
> 
> Oh, I missed the point that drm_gem_handle_delete() is calling
> drm_gem_object_unreference_unlocked().
> 
> But this still seems a bit odd, but I guess the difference is you are
> keeping the buffer handle, rather than the 'struct drm_gem_object'
> ptr.  I'm not sure if there is a point to keeping track of the buffer
> in handle form on the kernel side.  If instead you just keep a GEM
> object ptr, you can just drm_gem_object_unreference{_unlocked}() when
> you are done with it without having to special-case is_default stuff.

Ah, I found a problem through you. default framebuffer never has a gem
handle but only a buffer object. and as you mentioned, that isn't clear to
me also. I will consider better way. Thank you.

> 
> >> > +struct samsung_drm_gem_obj *
> >> > +               find_samsung_drm_gem_object(struct drm_file
*file_priv,
> >> > +                       struct drm_device *dev, unsigned int handle)
> >> > +{
> >> > +       struct drm_gem_object *gem_obj;
> >> > +
> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> >> > +       if (!gem_obj) {
> >> > +               DRM_LOG_KMS("a invalid gem object not registered to
> >> lookup.\n");
> >> > +               return NULL;
> >> > +       }
> >> > +
> >> > +       /**
> >> > +        * unreference refcount of the gem object.
> >> > +        * at drm_gem_object_lookup(), the gem object was referenced.
> >> > +        */
> >> > +       drm_gem_object_unreference(gem_obj);
> >>
> >> this doesn't seem right, to drop the reference before you use the
> >> buffer elsewhere..
> >>
> > No, see drm_gem_object_lookup fxn. at this function, if there is a
> object
> > found then drm_gem_object_reference is called to increase refcount of
> this
> > object. if there is any missing point, give me any comment please. thank
> > you.
> 
> 
> Right, but I think there is a reason it takes a reference... so that
> the object doesn't get free'd from under your feet.  So pattern
> should, I think, be:
> 
>   obj = lookup(...);
>   ... do stuff w/ obj ...
>   unreference(obj)
> 
> so the caller who is using the looked up obj should unref it when done
> 
> Instead, you have:
> 
>   obj = lookup(...);
>   unreference(obj);
>   ... do stuff w/ obj ...
> 
> 

Generally right, but in this case, it is just used to get specific gem
object through find_samsung_drm_gem_object() so doesn't reference this gem
object anywhere. 
therefore reference and unreference should be done within
find_samsung_drm_gem_object(). if there is any point I missed then let me
know please. thank you.

> BR,
> -R


^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-01 13:06         ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-01 13:06 UTC (permalink / raw)
  To: linux-arm-kernel

Hello Rob.
Below is my comments. thank you.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark at gmail.com]
> Sent: Thursday, September 01, 2011 1:02 AM
> To: Inki Dae
> Cc: airlied at linux.ie; dri-devel at lists.freedesktop.org;
> sw0312.kim at samsung.com; linux-kernel at vger.kernel.org;
> kyungmin.park at samsung.com; linux-arm-kernel at lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
> > Hello, Rob.
> > Below is my answers and questions. and could you please include me as CC
> > when you post your driver?
> 
> sure thing
> 
> 
> >> > +static int samsung_drm_connector_get_modes(struct drm_connector
> >> *connector)
> >> > +{
> >> > + ? ? ? struct samsung_drm_connector *samsung_connector =
> >> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> >> > + ? ? ? struct samsung_drm_display *display =
> >> > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
> >> >display;
> >> > + ? ? ? unsigned int count;
> >> > +
> >> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > + ? ? ? if (display && display->get_edid) {
> >> > + ? ? ? ? ? ? ? void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> >> > + ? ? ? ? ? ? ? if (!edid) {
> >> > + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to allocate edid\n");
> >> > + ? ? ? ? ? ? ? ? ? ? ? return 0;
> >> > + ? ? ? ? ? ? ? }
> >> > +
> >> > + ? ? ? ? ? ? ? display->get_edid(connector, edid, MAX_EDID);
> >> > +
> >> > + ? ? ? ? ? ? ? drm_mode_connector_update_edid_property(connector,
> > edid);
> >> > + ? ? ? ? ? ? ? count = drm_add_edid_modes(connector, edid);
> >> > +
> >> > + ? ? ? ? ? ? ? kfree(connector->display_info.raw_edid);
> >> > + ? ? ? ? ? ? ? connector->display_info.raw_edid = edid;
> >> > + ? ? ? } else {
> >> > + ? ? ? ? ? ? ? struct drm_display_mode *mode =
> > drm_mode_create(connector-
> >> >dev);
> >> > + ? ? ? ? ? ? ? struct fb_videomode *timing;
> >> > +
> >> > + ? ? ? ? ? ? ? if (display && display->get_timing)
> >> > + ? ? ? ? ? ? ? ? ? ? ? timing = display->get_timing();
> >>
> >> also seems a bit weird to not do: display->get_timings(display)..
> >>
> >> maybe you don't have the case of multiple instances of the same
> >> display time.. but maybe someday you need that. ?(Maybe this is just
> >> an artifact of how the API of your current lower layer driver is.. but
> >> maybe now is a good time to think about those APIs)
> >>
> >
> > display is static object set by hardware specific driver such as display
> > controller or hdmi. each hardware specific driver has their own display
> > object statically. You can refer to the definition in
samsung_drm_fimd.c.
> > and I understood a encoder and a connector is 1:1 relationship, when any
> > hardware specific driver is probed, they would be created with a manager
> > that includes their own display object as pointer. Could you please give
> me
> > more comments why display object should be considered for multiple
> > instances? I am afraid there is any part I don't care.
> >
> > thank you.
> 
> Just thinking hypothetically.. what if some future device had two hdmi
> controllers.  Then you'd want two instances of the same display
> object.
> 
> Although it seems this API is just internal to the DRM driver (which I
> had not realized earlier), so it should be pretty easy to change later
> if needed.
> 

You are right. I guess we should consider multiple instances to display
object because the Exynos hardware has two display controllers and also each
display controller needs one display panel with sharing same driver. thank
you for your pointing. :)

> 
> >> > +static void samsung_drm_connector_destroy(struct drm_connector
> >> *connector)
> >> > +{
> >> > + ? ? ? struct samsung_drm_connector *samsung_connector =
> >> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
> >> > +
> >> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > + ? ? ? drm_sysfs_connector_remove(connector);
> >> > + ? ? ? drm_connector_cleanup(connector);
> >> > + ? ? ? connector->dev->mode_config.num_connector--;
> >>
> >> I wonder if num_connector-- should be in drm_connector_cleanup()..
> >>
> >> I missed this in OMAP driver, but it seems that none of the other drm
> >> drivers are directly decrementing this.. and it would seem more
> >> symmetric for it to be in drm_connector_cleanup(). ?But would be
> >
> > For this, it was already commented by Joonyoun Shim.
> 
> And it seems (which I hadn't noticed earlier) already merged :-)
> 
> So I think this (and similar bit in encoder destructor) can be removed
> 
> 
> >> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> >> > +{
> >> > + ? ? ? struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> >> > + ? ? ? int ret;
> >> > +
> >> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > + ? ? ? drm_framebuffer_cleanup(fb);
> >> > +
> >> > + ? ? ? if (samsung_fb->is_default) {
> >> > + ? ? ? ? ? ? ? ret = drm_gem_handle_delete(samsung_fb->file_priv,
> >> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? samsung_fb->gem_handle);
> >>
> >> why not keep the gem buffer ptr, and do something like:
> >>
> >> ? drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> >>
> >> this way, you get the right behavior if someone somewhere else took a
> >> ref to the gem buffer object? ?And it avoids needing to keep the
> >> file_priv ptr in the fb (which seems a bit strange)
> >>
> >>
> > Yes, at booting time, one gem object is created. this is for linux
> > framebuffer and used as default buffer. register_framebuffer function is
> > called one time at booting time by drm driver. but when this driver is
> built
> > as modules, this gem object should be released with rmmod modules. and
> the
> > reason fb has file point is that drm_gem_handle_delete requests it to
> > release a gem object. our driver considered modularization also. If
> there is
> > any point I missed, give me any comment please. Thank you.
> 
> Oh, I missed the point that drm_gem_handle_delete() is calling
> drm_gem_object_unreference_unlocked().
> 
> But this still seems a bit odd, but I guess the difference is you are
> keeping the buffer handle, rather than the 'struct drm_gem_object'
> ptr.  I'm not sure if there is a point to keeping track of the buffer
> in handle form on the kernel side.  If instead you just keep a GEM
> object ptr, you can just drm_gem_object_unreference{_unlocked}() when
> you are done with it without having to special-case is_default stuff.

Ah, I found a problem through you. default framebuffer never has a gem
handle but only a buffer object. and as you mentioned, that isn't clear to
me also. I will consider better way. Thank you.

> 
> >> > +struct samsung_drm_gem_obj *
> >> > + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file
*file_priv,
> >> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int handle)
> >> > +{
> >> > + ? ? ? struct drm_gem_object *gem_obj;
> >> > +
> >> > + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> >> > + ? ? ? if (!gem_obj) {
> >> > + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered to
> >> lookup.\n");
> >> > + ? ? ? ? ? ? ? return NULL;
> >> > + ? ? ? }
> >> > +
> >> > + ? ? ? /**
> >> > + ? ? ? ?* unreference refcount of the gem object.
> >> > + ? ? ? ?* at drm_gem_object_lookup(), the gem object was referenced.
> >> > + ? ? ? ?*/
> >> > + ? ? ? drm_gem_object_unreference(gem_obj);
> >>
> >> this doesn't seem right, to drop the reference before you use the
> >> buffer elsewhere..
> >>
> > No, see drm_gem_object_lookup fxn. at this function, if there is a
> object
> > found then drm_gem_object_reference is called to increase refcount of
> this
> > object. if there is any missing point, give me any comment please. thank
> > you.
> 
> 
> Right, but I think there is a reason it takes a reference... so that
> the object doesn't get free'd from under your feet.  So pattern
> should, I think, be:
> 
>   obj = lookup(...);
>   ... do stuff w/ obj ...
>   unreference(obj)
> 
> so the caller who is using the looked up obj should unref it when done
> 
> Instead, you have:
> 
>   obj = lookup(...);
>   unreference(obj);
>   ... do stuff w/ obj ...
> 
> 

Generally right, but in this case, it is just used to get specific gem
object through find_samsung_drm_gem_object() so doesn't reference this gem
object anywhere. 
therefore reference and unreference should be done within
find_samsung_drm_gem_object(). if there is any point I missed then let me
know please. thank you.

> BR,
> -R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-01 13:06         ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-01 13:06 UTC (permalink / raw)
  To: 'Rob Clark'
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

Hello Rob.
Below is my comments. thank you.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark@gmail.com]
> Sent: Thursday, September 01, 2011 1:02 AM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
> > Hello, Rob.
> > Below is my answers and questions. and could you please include me as CC
> > when you post your driver?
> 
> sure thing
> 
> 
> >> > +static int samsung_drm_connector_get_modes(struct drm_connector
> >> *connector)
> >> > +{
> >> > +       struct samsung_drm_connector *samsung_connector =
> >> > +               to_samsung_connector(connector);
> >> > +       struct samsung_drm_display *display =
> >> > +               samsung_drm_get_manager(samsung_connector->encoder)-
> >> >display;
> >> > +       unsigned int count;
> >> > +
> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > +       if (display && display->get_edid) {
> >> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
> >> > +               if (!edid) {
> >> > +                       DRM_ERROR("failed to allocate edid\n");
> >> > +                       return 0;
> >> > +               }
> >> > +
> >> > +               display->get_edid(connector, edid, MAX_EDID);
> >> > +
> >> > +               drm_mode_connector_update_edid_property(connector,
> > edid);
> >> > +               count = drm_add_edid_modes(connector, edid);
> >> > +
> >> > +               kfree(connector->display_info.raw_edid);
> >> > +               connector->display_info.raw_edid = edid;
> >> > +       } else {
> >> > +               struct drm_display_mode *mode =
> > drm_mode_create(connector-
> >> >dev);
> >> > +               struct fb_videomode *timing;
> >> > +
> >> > +               if (display && display->get_timing)
> >> > +                       timing = display->get_timing();
> >>
> >> also seems a bit weird to not do: display->get_timings(display)..
> >>
> >> maybe you don't have the case of multiple instances of the same
> >> display time.. but maybe someday you need that.  (Maybe this is just
> >> an artifact of how the API of your current lower layer driver is.. but
> >> maybe now is a good time to think about those APIs)
> >>
> >
> > display is static object set by hardware specific driver such as display
> > controller or hdmi. each hardware specific driver has their own display
> > object statically. You can refer to the definition in
samsung_drm_fimd.c.
> > and I understood a encoder and a connector is 1:1 relationship, when any
> > hardware specific driver is probed, they would be created with a manager
> > that includes their own display object as pointer. Could you please give
> me
> > more comments why display object should be considered for multiple
> > instances? I am afraid there is any part I don't care.
> >
> > thank you.
> 
> Just thinking hypothetically.. what if some future device had two hdmi
> controllers.  Then you'd want two instances of the same display
> object.
> 
> Although it seems this API is just internal to the DRM driver (which I
> had not realized earlier), so it should be pretty easy to change later
> if needed.
> 

You are right. I guess we should consider multiple instances to display
object because the Exynos hardware has two display controllers and also each
display controller needs one display panel with sharing same driver. thank
you for your pointing. :)

> 
> >> > +static void samsung_drm_connector_destroy(struct drm_connector
> >> *connector)
> >> > +{
> >> > +       struct samsung_drm_connector *samsung_connector =
> >> > +               to_samsung_connector(connector);
> >> > +
> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > +       drm_sysfs_connector_remove(connector);
> >> > +       drm_connector_cleanup(connector);
> >> > +       connector->dev->mode_config.num_connector--;
> >>
> >> I wonder if num_connector-- should be in drm_connector_cleanup()..
> >>
> >> I missed this in OMAP driver, but it seems that none of the other drm
> >> drivers are directly decrementing this.. and it would seem more
> >> symmetric for it to be in drm_connector_cleanup().  But would be
> >
> > For this, it was already commented by Joonyoun Shim.
> 
> And it seems (which I hadn't noticed earlier) already merged :-)
> 
> So I think this (and similar bit in encoder destructor) can be removed
> 
> 
> >> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
> >> > +{
> >> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
> >> > +       int ret;
> >> > +
> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
> >> > +
> >> > +       drm_framebuffer_cleanup(fb);
> >> > +
> >> > +       if (samsung_fb->is_default) {
> >> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
> >> > +                               samsung_fb->gem_handle);
> >>
> >> why not keep the gem buffer ptr, and do something like:
> >>
> >>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
> >>
> >> this way, you get the right behavior if someone somewhere else took a
> >> ref to the gem buffer object?  And it avoids needing to keep the
> >> file_priv ptr in the fb (which seems a bit strange)
> >>
> >>
> > Yes, at booting time, one gem object is created. this is for linux
> > framebuffer and used as default buffer. register_framebuffer function is
> > called one time at booting time by drm driver. but when this driver is
> built
> > as modules, this gem object should be released with rmmod modules. and
> the
> > reason fb has file point is that drm_gem_handle_delete requests it to
> > release a gem object. our driver considered modularization also. If
> there is
> > any point I missed, give me any comment please. Thank you.
> 
> Oh, I missed the point that drm_gem_handle_delete() is calling
> drm_gem_object_unreference_unlocked().
> 
> But this still seems a bit odd, but I guess the difference is you are
> keeping the buffer handle, rather than the 'struct drm_gem_object'
> ptr.  I'm not sure if there is a point to keeping track of the buffer
> in handle form on the kernel side.  If instead you just keep a GEM
> object ptr, you can just drm_gem_object_unreference{_unlocked}() when
> you are done with it without having to special-case is_default stuff.

Ah, I found a problem through you. default framebuffer never has a gem
handle but only a buffer object. and as you mentioned, that isn't clear to
me also. I will consider better way. Thank you.

> 
> >> > +struct samsung_drm_gem_obj *
> >> > +               find_samsung_drm_gem_object(struct drm_file
*file_priv,
> >> > +                       struct drm_device *dev, unsigned int handle)
> >> > +{
> >> > +       struct drm_gem_object *gem_obj;
> >> > +
> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> >> > +       if (!gem_obj) {
> >> > +               DRM_LOG_KMS("a invalid gem object not registered to
> >> lookup.\n");
> >> > +               return NULL;
> >> > +       }
> >> > +
> >> > +       /**
> >> > +        * unreference refcount of the gem object.
> >> > +        * at drm_gem_object_lookup(), the gem object was referenced.
> >> > +        */
> >> > +       drm_gem_object_unreference(gem_obj);
> >>
> >> this doesn't seem right, to drop the reference before you use the
> >> buffer elsewhere..
> >>
> > No, see drm_gem_object_lookup fxn. at this function, if there is a
> object
> > found then drm_gem_object_reference is called to increase refcount of
> this
> > object. if there is any missing point, give me any comment please. thank
> > you.
> 
> 
> Right, but I think there is a reason it takes a reference... so that
> the object doesn't get free'd from under your feet.  So pattern
> should, I think, be:
> 
>   obj = lookup(...);
>   ... do stuff w/ obj ...
>   unreference(obj)
> 
> so the caller who is using the looked up obj should unref it when done
> 
> Instead, you have:
> 
>   obj = lookup(...);
>   unreference(obj);
>   ... do stuff w/ obj ...
> 
> 

Generally right, but in this case, it is just used to get specific gem
object through find_samsung_drm_gem_object() so doesn't reference this gem
object anywhere. 
therefore reference and unreference should be done within
find_samsung_drm_gem_object(). if there is any point I missed then let me
know please. thank you.

> BR,
> -R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-09-01 13:06         ` Inki Dae
  (?)
@ 2011-09-02  1:02           ` Kyungmin Park
  -1 siblings, 0 replies; 38+ messages in thread
From: Kyungmin Park @ 2011-09-02  1:02 UTC (permalink / raw)
  To: Inki Dae
  Cc: Rob Clark, airlied, sw0312.kim, linux-kernel, dri-devel,
	linux-arm-kernel

On Thu, Sep 1, 2011 at 10:06 PM, Inki Dae <inki.dae@samsung.com> wrote:
> Hello Rob.
> Below is my comments. thank you.
>
>> -----Original Message-----
>> From: Rob Clark [mailto:robdclark@gmail.com]
>> Sent: Thursday, September 01, 2011 1:02 AM
>> To: Inki Dae
>> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
>> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
>> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
>> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
>> EXYNOS4210.
>>
>> On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
>> > Hello, Rob.
>> > Below is my answers and questions. and could you please include me as CC
>> > when you post your driver?
>>
>> sure thing
>>
>>
>> >> > +static int samsung_drm_connector_get_modes(struct drm_connector
>> >> *connector)
>> >> > +{
>> >> > +       struct samsung_drm_connector *samsung_connector =
>> >> > +               to_samsung_connector(connector);
>> >> > +       struct samsung_drm_display *display =
>> >> > +               samsung_drm_get_manager(samsung_connector->encoder)-
>> >> >display;
>> >> > +       unsigned int count;
>> >> > +
>> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > +       if (display && display->get_edid) {
>> >> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
>> >> > +               if (!edid) {
>> >> > +                       DRM_ERROR("failed to allocate edid\n");
>> >> > +                       return 0;
>> >> > +               }
>> >> > +
>> >> > +               display->get_edid(connector, edid, MAX_EDID);
>> >> > +
>> >> > +               drm_mode_connector_update_edid_property(connector,
>> > edid);
>> >> > +               count = drm_add_edid_modes(connector, edid);
>> >> > +
>> >> > +               kfree(connector->display_info.raw_edid);
>> >> > +               connector->display_info.raw_edid = edid;
>> >> > +       } else {
>> >> > +               struct drm_display_mode *mode =
>> > drm_mode_create(connector-
>> >> >dev);
>> >> > +               struct fb_videomode *timing;
>> >> > +
>> >> > +               if (display && display->get_timing)
>> >> > +                       timing = display->get_timing();
>> >>
>> >> also seems a bit weird to not do: display->get_timings(display)..
>> >>
>> >> maybe you don't have the case of multiple instances of the same
>> >> display time.. but maybe someday you need that.  (Maybe this is just
>> >> an artifact of how the API of your current lower layer driver is.. but
>> >> maybe now is a good time to think about those APIs)
>> >>
>> >
>> > display is static object set by hardware specific driver such as display
>> > controller or hdmi. each hardware specific driver has their own display
>> > object statically. You can refer to the definition in
> samsung_drm_fimd.c.
>> > and I understood a encoder and a connector is 1:1 relationship, when any
>> > hardware specific driver is probed, they would be created with a manager
>> > that includes their own display object as pointer. Could you please give
>> me
>> > more comments why display object should be considered for multiple
>> > instances? I am afraid there is any part I don't care.
>> >
>> > thank you.
>>
>> Just thinking hypothetically.. what if some future device had two hdmi
>> controllers.  Then you'd want two instances of the same display
>> object.
>>
>> Although it seems this API is just internal to the DRM driver (which I
>> had not realized earlier), so it should be pretty easy to change later
>> if needed.
>>
>
> You are right. I guess we should consider multiple instances to display
> object because the Exynos hardware has two display controllers and also each
> display controller needs one display panel with sharing same driver. thank
> you for your pointing. :)

Two HDMI and two FIMD is different. even though exynos4 supports the
two display controller. it has only one HDMI port and now there's no
device to use it. Of course if new device uses it. it will support it.
I mena currently make it simple and make it TODO list.

Thank you,
Kyungmin Park
>
>>
>> >> > +static void samsung_drm_connector_destroy(struct drm_connector
>> >> *connector)
>> >> > +{
>> >> > +       struct samsung_drm_connector *samsung_connector =
>> >> > +               to_samsung_connector(connector);
>> >> > +
>> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > +       drm_sysfs_connector_remove(connector);
>> >> > +       drm_connector_cleanup(connector);
>> >> > +       connector->dev->mode_config.num_connector--;
>> >>
>> >> I wonder if num_connector-- should be in drm_connector_cleanup()..
>> >>
>> >> I missed this in OMAP driver, but it seems that none of the other drm
>> >> drivers are directly decrementing this.. and it would seem more
>> >> symmetric for it to be in drm_connector_cleanup().  But would be
>> >
>> > For this, it was already commented by Joonyoun Shim.
>>
>> And it seems (which I hadn't noticed earlier) already merged :-)
>>
>> So I think this (and similar bit in encoder destructor) can be removed
>>
>>
>> >> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
>> >> > +{
>> >> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
>> >> > +       int ret;
>> >> > +
>> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > +       drm_framebuffer_cleanup(fb);
>> >> > +
>> >> > +       if (samsung_fb->is_default) {
>> >> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
>> >> > +                               samsung_fb->gem_handle);
>> >>
>> >> why not keep the gem buffer ptr, and do something like:
>> >>
>> >>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
>> >>
>> >> this way, you get the right behavior if someone somewhere else took a
>> >> ref to the gem buffer object?  And it avoids needing to keep the
>> >> file_priv ptr in the fb (which seems a bit strange)
>> >>
>> >>
>> > Yes, at booting time, one gem object is created. this is for linux
>> > framebuffer and used as default buffer. register_framebuffer function is
>> > called one time at booting time by drm driver. but when this driver is
>> built
>> > as modules, this gem object should be released with rmmod modules. and
>> the
>> > reason fb has file point is that drm_gem_handle_delete requests it to
>> > release a gem object. our driver considered modularization also. If
>> there is
>> > any point I missed, give me any comment please. Thank you.
>>
>> Oh, I missed the point that drm_gem_handle_delete() is calling
>> drm_gem_object_unreference_unlocked().
>>
>> But this still seems a bit odd, but I guess the difference is you are
>> keeping the buffer handle, rather than the 'struct drm_gem_object'
>> ptr.  I'm not sure if there is a point to keeping track of the buffer
>> in handle form on the kernel side.  If instead you just keep a GEM
>> object ptr, you can just drm_gem_object_unreference{_unlocked}() when
>> you are done with it without having to special-case is_default stuff.
>
> Ah, I found a problem through you. default framebuffer never has a gem
> handle but only a buffer object. and as you mentioned, that isn't clear to
> me also. I will consider better way. Thank you.
>
>>
>> >> > +struct samsung_drm_gem_obj *
>> >> > +               find_samsung_drm_gem_object(struct drm_file
> *file_priv,
>> >> > +                       struct drm_device *dev, unsigned int handle)
>> >> > +{
>> >> > +       struct drm_gem_object *gem_obj;
>> >> > +
>> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> >> > +       if (!gem_obj) {
>> >> > +               DRM_LOG_KMS("a invalid gem object not registered to
>> >> lookup.\n");
>> >> > +               return NULL;
>> >> > +       }
>> >> > +
>> >> > +       /**
>> >> > +        * unreference refcount of the gem object.
>> >> > +        * at drm_gem_object_lookup(), the gem object was referenced.
>> >> > +        */
>> >> > +       drm_gem_object_unreference(gem_obj);
>> >>
>> >> this doesn't seem right, to drop the reference before you use the
>> >> buffer elsewhere..
>> >>
>> > No, see drm_gem_object_lookup fxn. at this function, if there is a
>> object
>> > found then drm_gem_object_reference is called to increase refcount of
>> this
>> > object. if there is any missing point, give me any comment please. thank
>> > you.
>>
>>
>> Right, but I think there is a reason it takes a reference... so that
>> the object doesn't get free'd from under your feet.  So pattern
>> should, I think, be:
>>
>>   obj = lookup(...);
>>   ... do stuff w/ obj ...
>>   unreference(obj)
>>
>> so the caller who is using the looked up obj should unref it when done
>>
>> Instead, you have:
>>
>>   obj = lookup(...);
>>   unreference(obj);
>>   ... do stuff w/ obj ...
>>
>>
>
> Generally right, but in this case, it is just used to get specific gem
> object through find_samsung_drm_gem_object() so doesn't reference this gem
> object anywhere.
> therefore reference and unreference should be done within
> find_samsung_drm_gem_object(). if there is any point I missed then let me
> know please. thank you.
>
>> BR,
>> -R
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>

^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02  1:02           ` Kyungmin Park
  0 siblings, 0 replies; 38+ messages in thread
From: Kyungmin Park @ 2011-09-02  1:02 UTC (permalink / raw)
  To: linux-arm-kernel

On Thu, Sep 1, 2011 at 10:06 PM, Inki Dae <inki.dae@samsung.com> wrote:
> Hello Rob.
> Below is my comments. thank you.
>
>> -----Original Message-----
>> From: Rob Clark [mailto:robdclark at gmail.com]
>> Sent: Thursday, September 01, 2011 1:02 AM
>> To: Inki Dae
>> Cc: airlied at linux.ie; dri-devel at lists.freedesktop.org;
>> sw0312.kim at samsung.com; linux-kernel at vger.kernel.org;
>> kyungmin.park at samsung.com; linux-arm-kernel at lists.infradead.org
>> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
>> EXYNOS4210.
>>
>> On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
>> > Hello, Rob.
>> > Below is my answers and questions. and could you please include me as CC
>> > when you post your driver?
>>
>> sure thing
>>
>>
>> >> > +static int samsung_drm_connector_get_modes(struct drm_connector
>> >> *connector)
>> >> > +{
>> >> > + ? ? ? struct samsung_drm_connector *samsung_connector =
>> >> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
>> >> > + ? ? ? struct samsung_drm_display *display =
>> >> > + ? ? ? ? ? ? ? samsung_drm_get_manager(samsung_connector->encoder)-
>> >> >display;
>> >> > + ? ? ? unsigned int count;
>> >> > +
>> >> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > + ? ? ? if (display && display->get_edid) {
>> >> > + ? ? ? ? ? ? ? void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
>> >> > + ? ? ? ? ? ? ? if (!edid) {
>> >> > + ? ? ? ? ? ? ? ? ? ? ? DRM_ERROR("failed to allocate edid\n");
>> >> > + ? ? ? ? ? ? ? ? ? ? ? return 0;
>> >> > + ? ? ? ? ? ? ? }
>> >> > +
>> >> > + ? ? ? ? ? ? ? display->get_edid(connector, edid, MAX_EDID);
>> >> > +
>> >> > + ? ? ? ? ? ? ? drm_mode_connector_update_edid_property(connector,
>> > edid);
>> >> > + ? ? ? ? ? ? ? count = drm_add_edid_modes(connector, edid);
>> >> > +
>> >> > + ? ? ? ? ? ? ? kfree(connector->display_info.raw_edid);
>> >> > + ? ? ? ? ? ? ? connector->display_info.raw_edid = edid;
>> >> > + ? ? ? } else {
>> >> > + ? ? ? ? ? ? ? struct drm_display_mode *mode =
>> > drm_mode_create(connector-
>> >> >dev);
>> >> > + ? ? ? ? ? ? ? struct fb_videomode *timing;
>> >> > +
>> >> > + ? ? ? ? ? ? ? if (display && display->get_timing)
>> >> > + ? ? ? ? ? ? ? ? ? ? ? timing = display->get_timing();
>> >>
>> >> also seems a bit weird to not do: display->get_timings(display)..
>> >>
>> >> maybe you don't have the case of multiple instances of the same
>> >> display time.. but maybe someday you need that. ?(Maybe this is just
>> >> an artifact of how the API of your current lower layer driver is.. but
>> >> maybe now is a good time to think about those APIs)
>> >>
>> >
>> > display is static object set by hardware specific driver such as display
>> > controller or hdmi. each hardware specific driver has their own display
>> > object statically. You can refer to the definition in
> samsung_drm_fimd.c.
>> > and I understood a encoder and a connector is 1:1 relationship, when any
>> > hardware specific driver is probed, they would be created with a manager
>> > that includes their own display object as pointer. Could you please give
>> me
>> > more comments why display object should be considered for multiple
>> > instances? I am afraid there is any part I don't care.
>> >
>> > thank you.
>>
>> Just thinking hypothetically.. what if some future device had two hdmi
>> controllers. ?Then you'd want two instances of the same display
>> object.
>>
>> Although it seems this API is just internal to the DRM driver (which I
>> had not realized earlier), so it should be pretty easy to change later
>> if needed.
>>
>
> You are right. I guess we should consider multiple instances to display
> object because the Exynos hardware has two display controllers and also each
> display controller needs one display panel with sharing same driver. thank
> you for your pointing. :)

Two HDMI and two FIMD is different. even though exynos4 supports the
two display controller. it has only one HDMI port and now there's no
device to use it. Of course if new device uses it. it will support it.
I mena currently make it simple and make it TODO list.

Thank you,
Kyungmin Park
>
>>
>> >> > +static void samsung_drm_connector_destroy(struct drm_connector
>> >> *connector)
>> >> > +{
>> >> > + ? ? ? struct samsung_drm_connector *samsung_connector =
>> >> > + ? ? ? ? ? ? ? to_samsung_connector(connector);
>> >> > +
>> >> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > + ? ? ? drm_sysfs_connector_remove(connector);
>> >> > + ? ? ? drm_connector_cleanup(connector);
>> >> > + ? ? ? connector->dev->mode_config.num_connector--;
>> >>
>> >> I wonder if num_connector-- should be in drm_connector_cleanup()..
>> >>
>> >> I missed this in OMAP driver, but it seems that none of the other drm
>> >> drivers are directly decrementing this.. and it would seem more
>> >> symmetric for it to be in drm_connector_cleanup(). ?But would be
>> >
>> > For this, it was already commented by Joonyoun Shim.
>>
>> And it seems (which I hadn't noticed earlier) already merged :-)
>>
>> So I think this (and similar bit in encoder destructor) can be removed
>>
>>
>> >> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
>> >> > +{
>> >> > + ? ? ? struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
>> >> > + ? ? ? int ret;
>> >> > +
>> >> > + ? ? ? DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > + ? ? ? drm_framebuffer_cleanup(fb);
>> >> > +
>> >> > + ? ? ? if (samsung_fb->is_default) {
>> >> > + ? ? ? ? ? ? ? ret = drm_gem_handle_delete(samsung_fb->file_priv,
>> >> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? samsung_fb->gem_handle);
>> >>
>> >> why not keep the gem buffer ptr, and do something like:
>> >>
>> >> ? drm_gem_object_unreference_unlocked(samsung_fb->bo)..
>> >>
>> >> this way, you get the right behavior if someone somewhere else took a
>> >> ref to the gem buffer object? ?And it avoids needing to keep the
>> >> file_priv ptr in the fb (which seems a bit strange)
>> >>
>> >>
>> > Yes, at booting time, one gem object is created. this is for linux
>> > framebuffer and used as default buffer. register_framebuffer function is
>> > called one time at booting time by drm driver. but when this driver is
>> built
>> > as modules, this gem object should be released with rmmod modules. and
>> the
>> > reason fb has file point is that drm_gem_handle_delete requests it to
>> > release a gem object. our driver considered modularization also. If
>> there is
>> > any point I missed, give me any comment please. Thank you.
>>
>> Oh, I missed the point that drm_gem_handle_delete() is calling
>> drm_gem_object_unreference_unlocked().
>>
>> But this still seems a bit odd, but I guess the difference is you are
>> keeping the buffer handle, rather than the 'struct drm_gem_object'
>> ptr. ?I'm not sure if there is a point to keeping track of the buffer
>> in handle form on the kernel side. ?If instead you just keep a GEM
>> object ptr, you can just drm_gem_object_unreference{_unlocked}() when
>> you are done with it without having to special-case is_default stuff.
>
> Ah, I found a problem through you. default framebuffer never has a gem
> handle but only a buffer object. and as you mentioned, that isn't clear to
> me also. I will consider better way. Thank you.
>
>>
>> >> > +struct samsung_drm_gem_obj *
>> >> > + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file
> *file_priv,
>> >> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int handle)
>> >> > +{
>> >> > + ? ? ? struct drm_gem_object *gem_obj;
>> >> > +
>> >> > + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> >> > + ? ? ? if (!gem_obj) {
>> >> > + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered to
>> >> lookup.\n");
>> >> > + ? ? ? ? ? ? ? return NULL;
>> >> > + ? ? ? }
>> >> > +
>> >> > + ? ? ? /**
>> >> > + ? ? ? ?* unreference refcount of the gem object.
>> >> > + ? ? ? ?* at drm_gem_object_lookup(), the gem object was referenced.
>> >> > + ? ? ? ?*/
>> >> > + ? ? ? drm_gem_object_unreference(gem_obj);
>> >>
>> >> this doesn't seem right, to drop the reference before you use the
>> >> buffer elsewhere..
>> >>
>> > No, see drm_gem_object_lookup fxn. at this function, if there is a
>> object
>> > found then drm_gem_object_reference is called to increase refcount of
>> this
>> > object. if there is any missing point, give me any comment please. thank
>> > you.
>>
>>
>> Right, but I think there is a reason it takes a reference... so that
>> the object doesn't get free'd from under your feet. ?So pattern
>> should, I think, be:
>>
>> ? obj = lookup(...);
>> ? ... do stuff w/ obj ...
>> ? unreference(obj)
>>
>> so the caller who is using the looked up obj should unref it when done
>>
>> Instead, you have:
>>
>> ? obj = lookup(...);
>> ? unreference(obj);
>> ? ... do stuff w/ obj ...
>>
>>
>
> Generally right, but in this case, it is just used to get specific gem
> object through find_samsung_drm_gem_object() so doesn't reference this gem
> object anywhere.
> therefore reference and unreference should be done within
> find_samsung_drm_gem_object(). if there is any point I missed then let me
> know please. thank you.
>
>> BR,
>> -R
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02  1:02           ` Kyungmin Park
  0 siblings, 0 replies; 38+ messages in thread
From: Kyungmin Park @ 2011-09-02  1:02 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, Rob Clark,
	linux-arm-kernel

On Thu, Sep 1, 2011 at 10:06 PM, Inki Dae <inki.dae@samsung.com> wrote:
> Hello Rob.
> Below is my comments. thank you.
>
>> -----Original Message-----
>> From: Rob Clark [mailto:robdclark@gmail.com]
>> Sent: Thursday, September 01, 2011 1:02 AM
>> To: Inki Dae
>> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
>> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
>> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
>> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
>> EXYNOS4210.
>>
>> On Wed, Aug 31, 2011 at 1:51 AM, Inki Dae <inki.dae@samsung.com> wrote:
>> > Hello, Rob.
>> > Below is my answers and questions. and could you please include me as CC
>> > when you post your driver?
>>
>> sure thing
>>
>>
>> >> > +static int samsung_drm_connector_get_modes(struct drm_connector
>> >> *connector)
>> >> > +{
>> >> > +       struct samsung_drm_connector *samsung_connector =
>> >> > +               to_samsung_connector(connector);
>> >> > +       struct samsung_drm_display *display =
>> >> > +               samsung_drm_get_manager(samsung_connector->encoder)-
>> >> >display;
>> >> > +       unsigned int count;
>> >> > +
>> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > +       if (display && display->get_edid) {
>> >> > +               void *edid = kzalloc(MAX_EDID, GFP_KERNEL);
>> >> > +               if (!edid) {
>> >> > +                       DRM_ERROR("failed to allocate edid\n");
>> >> > +                       return 0;
>> >> > +               }
>> >> > +
>> >> > +               display->get_edid(connector, edid, MAX_EDID);
>> >> > +
>> >> > +               drm_mode_connector_update_edid_property(connector,
>> > edid);
>> >> > +               count = drm_add_edid_modes(connector, edid);
>> >> > +
>> >> > +               kfree(connector->display_info.raw_edid);
>> >> > +               connector->display_info.raw_edid = edid;
>> >> > +       } else {
>> >> > +               struct drm_display_mode *mode =
>> > drm_mode_create(connector-
>> >> >dev);
>> >> > +               struct fb_videomode *timing;
>> >> > +
>> >> > +               if (display && display->get_timing)
>> >> > +                       timing = display->get_timing();
>> >>
>> >> also seems a bit weird to not do: display->get_timings(display)..
>> >>
>> >> maybe you don't have the case of multiple instances of the same
>> >> display time.. but maybe someday you need that.  (Maybe this is just
>> >> an artifact of how the API of your current lower layer driver is.. but
>> >> maybe now is a good time to think about those APIs)
>> >>
>> >
>> > display is static object set by hardware specific driver such as display
>> > controller or hdmi. each hardware specific driver has their own display
>> > object statically. You can refer to the definition in
> samsung_drm_fimd.c.
>> > and I understood a encoder and a connector is 1:1 relationship, when any
>> > hardware specific driver is probed, they would be created with a manager
>> > that includes their own display object as pointer. Could you please give
>> me
>> > more comments why display object should be considered for multiple
>> > instances? I am afraid there is any part I don't care.
>> >
>> > thank you.
>>
>> Just thinking hypothetically.. what if some future device had two hdmi
>> controllers.  Then you'd want two instances of the same display
>> object.
>>
>> Although it seems this API is just internal to the DRM driver (which I
>> had not realized earlier), so it should be pretty easy to change later
>> if needed.
>>
>
> You are right. I guess we should consider multiple instances to display
> object because the Exynos hardware has two display controllers and also each
> display controller needs one display panel with sharing same driver. thank
> you for your pointing. :)

Two HDMI and two FIMD is different. even though exynos4 supports the
two display controller. it has only one HDMI port and now there's no
device to use it. Of course if new device uses it. it will support it.
I mena currently make it simple and make it TODO list.

Thank you,
Kyungmin Park
>
>>
>> >> > +static void samsung_drm_connector_destroy(struct drm_connector
>> >> *connector)
>> >> > +{
>> >> > +       struct samsung_drm_connector *samsung_connector =
>> >> > +               to_samsung_connector(connector);
>> >> > +
>> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > +       drm_sysfs_connector_remove(connector);
>> >> > +       drm_connector_cleanup(connector);
>> >> > +       connector->dev->mode_config.num_connector--;
>> >>
>> >> I wonder if num_connector-- should be in drm_connector_cleanup()..
>> >>
>> >> I missed this in OMAP driver, but it seems that none of the other drm
>> >> drivers are directly decrementing this.. and it would seem more
>> >> symmetric for it to be in drm_connector_cleanup().  But would be
>> >
>> > For this, it was already commented by Joonyoun Shim.
>>
>> And it seems (which I hadn't noticed earlier) already merged :-)
>>
>> So I think this (and similar bit in encoder destructor) can be removed
>>
>>
>> >> > +static void samsung_drm_fb_destroy(struct drm_framebuffer *fb)
>> >> > +{
>> >> > +       struct samsung_drm_fb *samsung_fb = to_samsung_fb(fb);
>> >> > +       int ret;
>> >> > +
>> >> > +       DRM_DEBUG_KMS("%s\n", __FILE__);
>> >> > +
>> >> > +       drm_framebuffer_cleanup(fb);
>> >> > +
>> >> > +       if (samsung_fb->is_default) {
>> >> > +               ret = drm_gem_handle_delete(samsung_fb->file_priv,
>> >> > +                               samsung_fb->gem_handle);
>> >>
>> >> why not keep the gem buffer ptr, and do something like:
>> >>
>> >>   drm_gem_object_unreference_unlocked(samsung_fb->bo)..
>> >>
>> >> this way, you get the right behavior if someone somewhere else took a
>> >> ref to the gem buffer object?  And it avoids needing to keep the
>> >> file_priv ptr in the fb (which seems a bit strange)
>> >>
>> >>
>> > Yes, at booting time, one gem object is created. this is for linux
>> > framebuffer and used as default buffer. register_framebuffer function is
>> > called one time at booting time by drm driver. but when this driver is
>> built
>> > as modules, this gem object should be released with rmmod modules. and
>> the
>> > reason fb has file point is that drm_gem_handle_delete requests it to
>> > release a gem object. our driver considered modularization also. If
>> there is
>> > any point I missed, give me any comment please. Thank you.
>>
>> Oh, I missed the point that drm_gem_handle_delete() is calling
>> drm_gem_object_unreference_unlocked().
>>
>> But this still seems a bit odd, but I guess the difference is you are
>> keeping the buffer handle, rather than the 'struct drm_gem_object'
>> ptr.  I'm not sure if there is a point to keeping track of the buffer
>> in handle form on the kernel side.  If instead you just keep a GEM
>> object ptr, you can just drm_gem_object_unreference{_unlocked}() when
>> you are done with it without having to special-case is_default stuff.
>
> Ah, I found a problem through you. default framebuffer never has a gem
> handle but only a buffer object. and as you mentioned, that isn't clear to
> me also. I will consider better way. Thank you.
>
>>
>> >> > +struct samsung_drm_gem_obj *
>> >> > +               find_samsung_drm_gem_object(struct drm_file
> *file_priv,
>> >> > +                       struct drm_device *dev, unsigned int handle)
>> >> > +{
>> >> > +       struct drm_gem_object *gem_obj;
>> >> > +
>> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> >> > +       if (!gem_obj) {
>> >> > +               DRM_LOG_KMS("a invalid gem object not registered to
>> >> lookup.\n");
>> >> > +               return NULL;
>> >> > +       }
>> >> > +
>> >> > +       /**
>> >> > +        * unreference refcount of the gem object.
>> >> > +        * at drm_gem_object_lookup(), the gem object was referenced.
>> >> > +        */
>> >> > +       drm_gem_object_unreference(gem_obj);
>> >>
>> >> this doesn't seem right, to drop the reference before you use the
>> >> buffer elsewhere..
>> >>
>> > No, see drm_gem_object_lookup fxn. at this function, if there is a
>> object
>> > found then drm_gem_object_reference is called to increase refcount of
>> this
>> > object. if there is any missing point, give me any comment please. thank
>> > you.
>>
>>
>> Right, but I think there is a reason it takes a reference... so that
>> the object doesn't get free'd from under your feet.  So pattern
>> should, I think, be:
>>
>>   obj = lookup(...);
>>   ... do stuff w/ obj ...
>>   unreference(obj)
>>
>> so the caller who is using the looked up obj should unref it when done
>>
>> Instead, you have:
>>
>>   obj = lookup(...);
>>   unreference(obj);
>>   ... do stuff w/ obj ...
>>
>>
>
> Generally right, but in this case, it is just used to get specific gem
> object through find_samsung_drm_gem_object() so doesn't reference this gem
> object anywhere.
> therefore reference and unreference should be done within
> find_samsung_drm_gem_object(). if there is any point I missed then let me
> know please. thank you.
>
>> BR,
>> -R
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-09-02  1:02           ` Kyungmin Park
  (?)
@ 2011-09-02  1:08             ` Rob Clark
  -1 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-09-02  1:08 UTC (permalink / raw)
  To: Kyungmin Park
  Cc: Inki Dae, airlied, sw0312.kim, linux-kernel, dri-devel, linux-arm-kernel

On Thu, Sep 1, 2011 at 8:02 PM, Kyungmin Park <kmpark@infradead.org> wrote:
>>> Just thinking hypothetically.. what if some future device had two hdmi
>>> controllers.  Then you'd want two instances of the same display
>>> object.
>>>
>>> Although it seems this API is just internal to the DRM driver (which I
>>> had not realized earlier), so it should be pretty easy to change later
>>> if needed.
>>>
>>
>> You are right. I guess we should consider multiple instances to display
>> object because the Exynos hardware has two display controllers and also each
>> display controller needs one display panel with sharing same driver. thank
>> you for your pointing. :)
>
> Two HDMI and two FIMD is different. even though exynos4 supports the
> two display controller. it has only one HDMI port and now there's no
> device to use it. Of course if new device uses it. it will support it.
> I mena currently make it simple and make it TODO list.

yeah.. since it is just an API internal to the driver, and not
external (as I initially mistakenly thought), I don't think it is an
immediate issue.  It should be easy enough to change later if you, for
example, later had hw w/ two HDMI ports.

BR,
-R

> Thank you,
> Kyungmin Park

^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02  1:08             ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-09-02  1:08 UTC (permalink / raw)
  To: linux-arm-kernel

On Thu, Sep 1, 2011 at 8:02 PM, Kyungmin Park <kmpark@infradead.org> wrote:
>>> Just thinking hypothetically.. what if some future device had two hdmi
>>> controllers. ?Then you'd want two instances of the same display
>>> object.
>>>
>>> Although it seems this API is just internal to the DRM driver (which I
>>> had not realized earlier), so it should be pretty easy to change later
>>> if needed.
>>>
>>
>> You are right. I guess we should consider multiple instances to display
>> object because the Exynos hardware has two display controllers and also each
>> display controller needs one display panel with sharing same driver. thank
>> you for your pointing. :)
>
> Two HDMI and two FIMD is different. even though exynos4 supports the
> two display controller. it has only one HDMI port and now there's no
> device to use it. Of course if new device uses it. it will support it.
> I mena currently make it simple and make it TODO list.

yeah.. since it is just an API internal to the driver, and not
external (as I initially mistakenly thought), I don't think it is an
immediate issue.  It should be easy enough to change later if you, for
example, later had hw w/ two HDMI ports.

BR,
-R

> Thank you,
> Kyungmin Park

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02  1:08             ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-09-02  1:08 UTC (permalink / raw)
  To: Kyungmin Park
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, Inki Dae, linux-arm-kernel

On Thu, Sep 1, 2011 at 8:02 PM, Kyungmin Park <kmpark@infradead.org> wrote:
>>> Just thinking hypothetically.. what if some future device had two hdmi
>>> controllers.  Then you'd want two instances of the same display
>>> object.
>>>
>>> Although it seems this API is just internal to the DRM driver (which I
>>> had not realized earlier), so it should be pretty easy to change later
>>> if needed.
>>>
>>
>> You are right. I guess we should consider multiple instances to display
>> object because the Exynos hardware has two display controllers and also each
>> display controller needs one display panel with sharing same driver. thank
>> you for your pointing. :)
>
> Two HDMI and two FIMD is different. even though exynos4 supports the
> two display controller. it has only one HDMI port and now there's no
> device to use it. Of course if new device uses it. it will support it.
> I mena currently make it simple and make it TODO list.

yeah.. since it is just an API internal to the driver, and not
external (as I initially mistakenly thought), I don't think it is an
immediate issue.  It should be easy enough to change later if you, for
example, later had hw w/ two HDMI ports.

BR,
-R

> Thank you,
> Kyungmin Park

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-09-01 13:06         ` Inki Dae
  (?)
@ 2011-09-02  1:18           ` Rob Clark
  -1 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-09-02  1:18 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

On Thu, Sep 1, 2011 at 8:06 AM, Inki Dae <inki.dae@samsung.com> wrote:
>> >> > +struct samsung_drm_gem_obj *
>> >> > +               find_samsung_drm_gem_object(struct drm_file
> *file_priv,
>> >> > +                       struct drm_device *dev, unsigned int handle)
>> >> > +{
>> >> > +       struct drm_gem_object *gem_obj;
>> >> > +
>> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> >> > +       if (!gem_obj) {
>> >> > +               DRM_LOG_KMS("a invalid gem object not registered to
>> >> lookup.\n");
>> >> > +               return NULL;
>> >> > +       }
>> >> > +
>> >> > +       /**
>> >> > +        * unreference refcount of the gem object.
>> >> > +        * at drm_gem_object_lookup(), the gem object was referenced.
>> >> > +        */
>> >> > +       drm_gem_object_unreference(gem_obj);
>> >>
>> >> this doesn't seem right, to drop the reference before you use the
>> >> buffer elsewhere..
>> >>
>> > No, see drm_gem_object_lookup fxn. at this function, if there is a
>> object
>> > found then drm_gem_object_reference is called to increase refcount of
>> this
>> > object. if there is any missing point, give me any comment please. thank
>> > you.
>>
>>
>> Right, but I think there is a reason it takes a reference... so that
>> the object doesn't get free'd from under your feet.  So pattern
>> should, I think, be:
>>
>>   obj = lookup(...);
>>   ... do stuff w/ obj ...
>>   unreference(obj)
>>
>> so the caller who is using the looked up obj should unref it when done
>>
>> Instead, you have:
>>
>>   obj = lookup(...);
>>   unreference(obj);
>>   ... do stuff w/ obj ...
>>
>>
>
> Generally right, but in this case, it is just used to get specific gem
> object through find_samsung_drm_gem_object() so doesn't reference this gem
> object anywhere.
> therefore reference and unreference should be done within
> find_samsung_drm_gem_object(). if there is any point I missed then let me
> know please. thank you.
>

Still, it seems like find_samsung_drm_gem_object() is encouraging the
wrong usage-pattern, even if it works fine today because you know
somewhere else is holding a reference to the object.  Later if you
expand your use of GEM objects, this fxn might come back to bite you.
There is a good reason that drm_gem_object_lookup() takes a reference
to the object, and it feels wrong to intentionally subvert that.

(I'm perfectly willing to be overridden on the subject.. there are
plenty of folks on this list who have been doing the GEM thing longer
than I have.  But it just seems better to use APIs like
drm_gem_object_lookup() the way they were intended.)

BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02  1:18           ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-09-02  1:18 UTC (permalink / raw)
  To: linux-arm-kernel

On Thu, Sep 1, 2011 at 8:06 AM, Inki Dae <inki.dae@samsung.com> wrote:
>> >> > +struct samsung_drm_gem_obj *
>> >> > + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file
> *file_priv,
>> >> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int handle)
>> >> > +{
>> >> > + ? ? ? struct drm_gem_object *gem_obj;
>> >> > +
>> >> > + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> >> > + ? ? ? if (!gem_obj) {
>> >> > + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered to
>> >> lookup.\n");
>> >> > + ? ? ? ? ? ? ? return NULL;
>> >> > + ? ? ? }
>> >> > +
>> >> > + ? ? ? /**
>> >> > + ? ? ? ?* unreference refcount of the gem object.
>> >> > + ? ? ? ?* at drm_gem_object_lookup(), the gem object was referenced.
>> >> > + ? ? ? ?*/
>> >> > + ? ? ? drm_gem_object_unreference(gem_obj);
>> >>
>> >> this doesn't seem right, to drop the reference before you use the
>> >> buffer elsewhere..
>> >>
>> > No, see drm_gem_object_lookup fxn. at this function, if there is a
>> object
>> > found then drm_gem_object_reference is called to increase refcount of
>> this
>> > object. if there is any missing point, give me any comment please. thank
>> > you.
>>
>>
>> Right, but I think there is a reason it takes a reference... so that
>> the object doesn't get free'd from under your feet. ?So pattern
>> should, I think, be:
>>
>> ? obj = lookup(...);
>> ? ... do stuff w/ obj ...
>> ? unreference(obj)
>>
>> so the caller who is using the looked up obj should unref it when done
>>
>> Instead, you have:
>>
>> ? obj = lookup(...);
>> ? unreference(obj);
>> ? ... do stuff w/ obj ...
>>
>>
>
> Generally right, but in this case, it is just used to get specific gem
> object through find_samsung_drm_gem_object() so doesn't reference this gem
> object anywhere.
> therefore reference and unreference should be done within
> find_samsung_drm_gem_object(). if there is any point I missed then let me
> know please. thank you.
>

Still, it seems like find_samsung_drm_gem_object() is encouraging the
wrong usage-pattern, even if it works fine today because you know
somewhere else is holding a reference to the object.  Later if you
expand your use of GEM objects, this fxn might come back to bite you.
There is a good reason that drm_gem_object_lookup() takes a reference
to the object, and it feels wrong to intentionally subvert that.

(I'm perfectly willing to be overridden on the subject.. there are
plenty of folks on this list who have been doing the GEM thing longer
than I have.  But it just seems better to use APIs like
drm_gem_object_lookup() the way they were intended.)

BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02  1:18           ` Rob Clark
  0 siblings, 0 replies; 38+ messages in thread
From: Rob Clark @ 2011-09-02  1:18 UTC (permalink / raw)
  To: Inki Dae
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

On Thu, Sep 1, 2011 at 8:06 AM, Inki Dae <inki.dae@samsung.com> wrote:
>> >> > +struct samsung_drm_gem_obj *
>> >> > +               find_samsung_drm_gem_object(struct drm_file
> *file_priv,
>> >> > +                       struct drm_device *dev, unsigned int handle)
>> >> > +{
>> >> > +       struct drm_gem_object *gem_obj;
>> >> > +
>> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
>> >> > +       if (!gem_obj) {
>> >> > +               DRM_LOG_KMS("a invalid gem object not registered to
>> >> lookup.\n");
>> >> > +               return NULL;
>> >> > +       }
>> >> > +
>> >> > +       /**
>> >> > +        * unreference refcount of the gem object.
>> >> > +        * at drm_gem_object_lookup(), the gem object was referenced.
>> >> > +        */
>> >> > +       drm_gem_object_unreference(gem_obj);
>> >>
>> >> this doesn't seem right, to drop the reference before you use the
>> >> buffer elsewhere..
>> >>
>> > No, see drm_gem_object_lookup fxn. at this function, if there is a
>> object
>> > found then drm_gem_object_reference is called to increase refcount of
>> this
>> > object. if there is any missing point, give me any comment please. thank
>> > you.
>>
>>
>> Right, but I think there is a reason it takes a reference... so that
>> the object doesn't get free'd from under your feet.  So pattern
>> should, I think, be:
>>
>>   obj = lookup(...);
>>   ... do stuff w/ obj ...
>>   unreference(obj)
>>
>> so the caller who is using the looked up obj should unref it when done
>>
>> Instead, you have:
>>
>>   obj = lookup(...);
>>   unreference(obj);
>>   ... do stuff w/ obj ...
>>
>>
>
> Generally right, but in this case, it is just used to get specific gem
> object through find_samsung_drm_gem_object() so doesn't reference this gem
> object anywhere.
> therefore reference and unreference should be done within
> find_samsung_drm_gem_object(). if there is any point I missed then let me
> know please. thank you.
>

Still, it seems like find_samsung_drm_gem_object() is encouraging the
wrong usage-pattern, even if it works fine today because you know
somewhere else is holding a reference to the object.  Later if you
expand your use of GEM objects, this fxn might come back to bite you.
There is a good reason that drm_gem_object_lookup() takes a reference
to the object, and it feels wrong to intentionally subvert that.

(I'm perfectly willing to be overridden on the subject.. there are
plenty of folks on this list who have been doing the GEM thing longer
than I have.  But it just seems better to use APIs like
drm_gem_object_lookup() the way they were intended.)

BR,
-R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
  2011-09-02  1:18           ` Rob Clark
  (?)
@ 2011-09-02 12:00             ` Inki Dae
  -1 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-02 12:00 UTC (permalink / raw)
  To: 'Rob Clark'
  Cc: airlied, dri-devel, sw0312.kim, linux-kernel, kyungmin.park,
	linux-arm-kernel

Hello Rob.
Below is my comments.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark@gmail.com]
> Sent: Friday, September 02, 2011 10:18 AM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On Thu, Sep 1, 2011 at 8:06 AM, Inki Dae <inki.dae@samsung.com> wrote:
> >> >> > +struct samsung_drm_gem_obj *
> >> >> > +               find_samsung_drm_gem_object(struct drm_file
> > *file_priv,
> >> >> > +                       struct drm_device *dev, unsigned int
handle)
> >> >> > +{
> >> >> > +       struct drm_gem_object *gem_obj;
> >> >> > +
> >> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> >> >> > +       if (!gem_obj) {
> >> >> > +               DRM_LOG_KMS("a invalid gem object not registered
to
> >> >> lookup.\n");
> >> >> > +               return NULL;
> >> >> > +       }
> >> >> > +
> >> >> > +       /**
> >> >> > +        * unreference refcount of the gem object.
> >> >> > +        * at drm_gem_object_lookup(), the gem object was
> referenced.
> >> >> > +        */
> >> >> > +       drm_gem_object_unreference(gem_obj);
> >> >>
> >> >> this doesn't seem right, to drop the reference before you use the
> >> >> buffer elsewhere..
> >> >>
> >> > No, see drm_gem_object_lookup fxn. at this function, if there is a
> >> object
> >> > found then drm_gem_object_reference is called to increase refcount of
> >> this
> >> > object. if there is any missing point, give me any comment please.
> thank
> >> > you.
> >>
> >>
> >> Right, but I think there is a reason it takes a reference... so that
> >> the object doesn't get free'd from under your feet.  So pattern
> >> should, I think, be:
> >>
> >>   obj = lookup(...);
> >>   ... do stuff w/ obj ...
> >>   unreference(obj)
> >>
> >> so the caller who is using the looked up obj should unref it when done
> >>
> >> Instead, you have:
> >>
> >>   obj = lookup(...);
> >>   unreference(obj);
> >>   ... do stuff w/ obj ...
> >>
> >>
> >
> > Generally right, but in this case, it is just used to get specific gem
> > object through find_samsung_drm_gem_object() so doesn't reference this
> gem
> > object anywhere.
> > therefore reference and unreference should be done within
> > find_samsung_drm_gem_object(). if there is any point I missed then let
> me
> > know please. thank you.
> >
> 
> Still, it seems like find_samsung_drm_gem_object() is encouraging the
> wrong usage-pattern, even if it works fine today because you know
> somewhere else is holding a reference to the object.  Later if you
> expand your use of GEM objects, this fxn might come back to bite you.
> There is a good reason that drm_gem_object_lookup() takes a reference
> to the object, and it feels wrong to intentionally subvert that.
> 
> (I'm perfectly willing to be overridden on the subject.. there are
> plenty of folks on this list who have been doing the GEM thing longer
> than I have.  But it just seems better to use APIs like
> drm_gem_object_lookup() the way they were intended.)
> 

Ah, you are right. I misunderstanded it. as you pointed out, a gem object
should be unreferenced after doing something with the gem object. so I will
remove find_samsung_drm_gem_object() and use drm_gem_object_lookup()
directly to get a gem object instead. of course, the gem object will be
unreferenced after doing something with it. thank you for your explanation.
:)

> BR,
> -R


^ permalink raw reply	[flat|nested] 38+ messages in thread

* [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02 12:00             ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-02 12:00 UTC (permalink / raw)
  To: linux-arm-kernel

Hello Rob.
Below is my comments.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark at gmail.com]
> Sent: Friday, September 02, 2011 10:18 AM
> To: Inki Dae
> Cc: airlied at linux.ie; dri-devel at lists.freedesktop.org;
> sw0312.kim at samsung.com; linux-kernel at vger.kernel.org;
> kyungmin.park at samsung.com; linux-arm-kernel at lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On Thu, Sep 1, 2011 at 8:06 AM, Inki Dae <inki.dae@samsung.com> wrote:
> >> >> > +struct samsung_drm_gem_obj *
> >> >> > + ? ? ? ? ? ? ? find_samsung_drm_gem_object(struct drm_file
> > *file_priv,
> >> >> > + ? ? ? ? ? ? ? ? ? ? ? struct drm_device *dev, unsigned int
handle)
> >> >> > +{
> >> >> > + ? ? ? struct drm_gem_object *gem_obj;
> >> >> > +
> >> >> > + ? ? ? gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> >> >> > + ? ? ? if (!gem_obj) {
> >> >> > + ? ? ? ? ? ? ? DRM_LOG_KMS("a invalid gem object not registered
to
> >> >> lookup.\n");
> >> >> > + ? ? ? ? ? ? ? return NULL;
> >> >> > + ? ? ? }
> >> >> > +
> >> >> > + ? ? ? /**
> >> >> > + ? ? ? ?* unreference refcount of the gem object.
> >> >> > + ? ? ? ?* at drm_gem_object_lookup(), the gem object was
> referenced.
> >> >> > + ? ? ? ?*/
> >> >> > + ? ? ? drm_gem_object_unreference(gem_obj);
> >> >>
> >> >> this doesn't seem right, to drop the reference before you use the
> >> >> buffer elsewhere..
> >> >>
> >> > No, see drm_gem_object_lookup fxn. at this function, if there is a
> >> object
> >> > found then drm_gem_object_reference is called to increase refcount of
> >> this
> >> > object. if there is any missing point, give me any comment please.
> thank
> >> > you.
> >>
> >>
> >> Right, but I think there is a reason it takes a reference... so that
> >> the object doesn't get free'd from under your feet. ?So pattern
> >> should, I think, be:
> >>
> >> ? obj = lookup(...);
> >> ? ... do stuff w/ obj ...
> >> ? unreference(obj)
> >>
> >> so the caller who is using the looked up obj should unref it when done
> >>
> >> Instead, you have:
> >>
> >> ? obj = lookup(...);
> >> ? unreference(obj);
> >> ? ... do stuff w/ obj ...
> >>
> >>
> >
> > Generally right, but in this case, it is just used to get specific gem
> > object through find_samsung_drm_gem_object() so doesn't reference this
> gem
> > object anywhere.
> > therefore reference and unreference should be done within
> > find_samsung_drm_gem_object(). if there is any point I missed then let
> me
> > know please. thank you.
> >
> 
> Still, it seems like find_samsung_drm_gem_object() is encouraging the
> wrong usage-pattern, even if it works fine today because you know
> somewhere else is holding a reference to the object.  Later if you
> expand your use of GEM objects, this fxn might come back to bite you.
> There is a good reason that drm_gem_object_lookup() takes a reference
> to the object, and it feels wrong to intentionally subvert that.
> 
> (I'm perfectly willing to be overridden on the subject.. there are
> plenty of folks on this list who have been doing the GEM thing longer
> than I have.  But it just seems better to use APIs like
> drm_gem_object_lookup() the way they were intended.)
> 

Ah, you are right. I misunderstanded it. as you pointed out, a gem object
should be unreferenced after doing something with the gem object. so I will
remove find_samsung_drm_gem_object() and use drm_gem_object_lookup()
directly to get a gem object instead. of course, the gem object will be
unreferenced after doing something with it. thank you for your explanation.
:)

> BR,
> -R

^ permalink raw reply	[flat|nested] 38+ messages in thread

* RE: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210.
@ 2011-09-02 12:00             ` Inki Dae
  0 siblings, 0 replies; 38+ messages in thread
From: Inki Dae @ 2011-09-02 12:00 UTC (permalink / raw)
  To: 'Rob Clark'
  Cc: airlied, sw0312.kim, linux-kernel, dri-devel, kyungmin.park,
	linux-arm-kernel

Hello Rob.
Below is my comments.

> -----Original Message-----
> From: Rob Clark [mailto:robdclark@gmail.com]
> Sent: Friday, September 02, 2011 10:18 AM
> To: Inki Dae
> Cc: airlied@linux.ie; dri-devel@lists.freedesktop.org;
> sw0312.kim@samsung.com; linux-kernel@vger.kernel.org;
> kyungmin.park@samsung.com; linux-arm-kernel@lists.infradead.org
> Subject: Re: [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC
> EXYNOS4210.
> 
> On Thu, Sep 1, 2011 at 8:06 AM, Inki Dae <inki.dae@samsung.com> wrote:
> >> >> > +struct samsung_drm_gem_obj *
> >> >> > +               find_samsung_drm_gem_object(struct drm_file
> > *file_priv,
> >> >> > +                       struct drm_device *dev, unsigned int
handle)
> >> >> > +{
> >> >> > +       struct drm_gem_object *gem_obj;
> >> >> > +
> >> >> > +       gem_obj = drm_gem_object_lookup(dev, file_priv, handle);
> >> >> > +       if (!gem_obj) {
> >> >> > +               DRM_LOG_KMS("a invalid gem object not registered
to
> >> >> lookup.\n");
> >> >> > +               return NULL;
> >> >> > +       }
> >> >> > +
> >> >> > +       /**
> >> >> > +        * unreference refcount of the gem object.
> >> >> > +        * at drm_gem_object_lookup(), the gem object was
> referenced.
> >> >> > +        */
> >> >> > +       drm_gem_object_unreference(gem_obj);
> >> >>
> >> >> this doesn't seem right, to drop the reference before you use the
> >> >> buffer elsewhere..
> >> >>
> >> > No, see drm_gem_object_lookup fxn. at this function, if there is a
> >> object
> >> > found then drm_gem_object_reference is called to increase refcount of
> >> this
> >> > object. if there is any missing point, give me any comment please.
> thank
> >> > you.
> >>
> >>
> >> Right, but I think there is a reason it takes a reference... so that
> >> the object doesn't get free'd from under your feet.  So pattern
> >> should, I think, be:
> >>
> >>   obj = lookup(...);
> >>   ... do stuff w/ obj ...
> >>   unreference(obj)
> >>
> >> so the caller who is using the looked up obj should unref it when done
> >>
> >> Instead, you have:
> >>
> >>   obj = lookup(...);
> >>   unreference(obj);
> >>   ... do stuff w/ obj ...
> >>
> >>
> >
> > Generally right, but in this case, it is just used to get specific gem
> > object through find_samsung_drm_gem_object() so doesn't reference this
> gem
> > object anywhere.
> > therefore reference and unreference should be done within
> > find_samsung_drm_gem_object(). if there is any point I missed then let
> me
> > know please. thank you.
> >
> 
> Still, it seems like find_samsung_drm_gem_object() is encouraging the
> wrong usage-pattern, even if it works fine today because you know
> somewhere else is holding a reference to the object.  Later if you
> expand your use of GEM objects, this fxn might come back to bite you.
> There is a good reason that drm_gem_object_lookup() takes a reference
> to the object, and it feels wrong to intentionally subvert that.
> 
> (I'm perfectly willing to be overridden on the subject.. there are
> plenty of folks on this list who have been doing the GEM thing longer
> than I have.  But it just seems better to use APIs like
> drm_gem_object_lookup() the way they were intended.)
> 

Ah, you are right. I misunderstanded it. as you pointed out, a gem object
should be unreferenced after doing something with the gem object. so I will
remove find_samsung_drm_gem_object() and use drm_gem_object_lookup()
directly to get a gem object instead. of course, the gem object will be
unreferenced after doing something with it. thank you for your explanation.
:)

> BR,
> -R

^ permalink raw reply	[flat|nested] 38+ messages in thread

end of thread, other threads:[~2011-09-02 12:00 UTC | newest]

Thread overview: 38+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-08-26 11:47 [RFC][PATCH v3] DRM: add DRM Driver for Samsung SoC EXYNOS4210 Inki Dae
2011-08-26 11:47 ` Inki Dae
2011-08-31  1:57 ` Rob Clark
2011-08-31  1:57   ` Rob Clark
2011-08-31  1:57   ` Rob Clark
2011-08-31  2:06   ` Konrad Rzeszutek Wilk
2011-08-31  2:06     ` Konrad Rzeszutek Wilk
2011-08-31  7:33     ` Inki Dae
2011-08-31  7:33       ` Inki Dae
2011-08-31  7:33       ` Inki Dae
2011-08-31  2:28   ` Joonyoung Shim
2011-08-31  2:28     ` Joonyoung Shim
2011-08-31  2:28     ` Joonyoung Shim
2011-08-31  6:51   ` Inki Dae
2011-08-31  6:51     ` Inki Dae
2011-08-31  6:51     ` Inki Dae
2011-08-31 16:02     ` Rob Clark
2011-08-31 16:02       ` Rob Clark
2011-09-01 13:06       ` Inki Dae
2011-09-01 13:06         ` Inki Dae
2011-09-01 13:06         ` Inki Dae
2011-09-02  1:02         ` Kyungmin Park
2011-09-02  1:02           ` Kyungmin Park
2011-09-02  1:02           ` Kyungmin Park
2011-09-02  1:08           ` Rob Clark
2011-09-02  1:08             ` Rob Clark
2011-09-02  1:08             ` Rob Clark
2011-09-02  1:18         ` Rob Clark
2011-09-02  1:18           ` Rob Clark
2011-09-02  1:18           ` Rob Clark
2011-09-02 12:00           ` Inki Dae
2011-09-02 12:00             ` Inki Dae
2011-09-02 12:00             ` Inki Dae
2011-08-31  8:38 ` Thomas Hellstrom
2011-08-31  8:38   ` Thomas Hellstrom
2011-09-01  3:57   ` Inki Dae
2011-09-01  3:57     ` Inki Dae
2011-09-01  3:57     ` Inki Dae

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.