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

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