Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
On ma, 2015-03-23 at 11:10 +, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursu...@intel.com
90/270 rotated scanout needs a rotated GTT view of the framebuffer.
This is put in a separate VMA with a dedicated ggtt view and wired such that
it is created when a framebuffer is pinned to a 90/270 rotated plane.
Rotation is only possible with Yb/Yf buffers and error is propagated to
user space in case of a mismatch.
Special rotated page view is constructed at the VMA creation time by
borrowing the DMA addresses from obj-pages.
v2:
* Do not bother with pages for rotated sg list, just populate the DMA
addresses. (Daniel Vetter)
* Checkpatch cleanup.
v3:
* Rebased on top of new plane handling (create rotated mapping when
setting the rotation property).
* Unpin rotated VMA on unpinning from display plane.
* Simplify rotation check using bitwise AND. (Chris Wilson)
v4:
* Fix unpinning of optional rotated mapping so it is really considered
to be optional.
v5:
* Rebased for fb modifier changes.
* Rebased for atomic commit.
* Only pin needed view for display. (Ville Syrjälä, Daniel Vetter)
v6:
* Rebased after preparatory work has been extracted out. (Daniel Vetter)
v7:
* Slightly simplified tiling geometry calculation.
* Moved rotated GGTT view implementation into i915_gem_gtt.c (Daniel
Vetter)
v8:
* Do not use i915_gem_obj_size to get object size since that actually
returns the size of an VMA which may not exist.
* Rebased for ggtt view changes.
v9:
* Rebased after code review changes on the preceding patches.
* Tidy function definitions. (Joonas Lahtinen)
For: VIZ-4726
Signed-off-by: Tvrtko Ursulin tvrtko.ursu...@intel.com
Reviewed-by: Michel Thierry michel.thie...@intel.com (v4)
Reviewed-by: Joonas Lahtinen joonas.lahti...@linux.intel.com (v9)
---
drivers/gpu/drm/i915/i915_gem.c | 4 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 117
+--
drivers/gpu/drm/i915/i915_gem_gtt.h | 12
drivers/gpu/drm/i915/intel_display.c | 27 +++-
drivers/gpu/drm/i915/intel_drv.h | 4 ++
5 files changed, 158 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 73b2638..b6d4b07 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3908,7 +3908,9 @@ i915_gem_object_pin_to_display_plane(struct
drm_i915_gem_object *obj,
* (e.g. libkms for the bootup splash), we have to ensure that we
* always use map_and_fenceable for all scanout buffers.
*/
- ret = i915_gem_object_ggtt_pin(obj, view, alignment, PIN_MAPPABLE);
+ ret = i915_gem_object_ggtt_pin(obj, view, alignment,
+view-type == I915_GGTT_VIEW_NORMAL ?
+PIN_MAPPABLE : 0);
if (ret)
goto err_unpin_display;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c
b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 645c363..fc56c11 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2500,15 +2500,119 @@ i915_gem_obj_lookup_or_create_ggtt_vma(struct
drm_i915_gem_object *obj,
}
+static void
+rotate_pages(dma_addr_t *in, unsigned int width, unsigned int height,
+ struct sg_table *st)
+{
+ unsigned int column, row;
+ unsigned int src_idx;
+ struct scatterlist *sg = st-sgl;
+
+ st-nents = 0;
+
+ for (column = 0; column width; column++) {
+ src_idx = width * (height - 1) + column;
+ for (row = 0; row height; row++) {
+ st-nents++;
+ /* We don't need the pages, but need to initialize
+ * the entries so the sg list can be happily traversed.
+ * The only thing we need are DMA addresses.
+ */
+ sg_set_page(sg, NULL, PAGE_SIZE, 0);
+ sg_dma_address(sg) = in[src_idx];
+ sg_dma_len(sg) = PAGE_SIZE;
+ sg = sg_next(sg);
+ src_idx -= width;
+ }
+ }
+}
+
+static struct sg_table *
+intel_rotate_fb_obj_pages(struct i915_ggtt_view *ggtt_view,
+ struct drm_i915_gem_object *obj)
+{
+ struct drm_device *dev = obj-base.dev;
+ struct intel_rotation_info *rot_info = ggtt_view-rotation_info;
+ unsigned long size, pages, rot_pages;
+ struct sg_page_iter sg_iter;
+ unsigned long i;
+ dma_addr_t *page_addr_list;
+ struct sg_table *st;
+ unsigned int tile_pitch, tile_height;
+ unsigned int width_pages, height_pages;
+ int ret = ENOMEM;
+
+ pages = obj-base.size / PAGE_SIZE;
+
+ /* Calculate tiling geometry. */
+ tile_height = intel_tile_height(dev,
Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
Hi,
On pe, 2015-03-20 at 12:11 +, Tvrtko Ursulin wrote:
On 03/20/2015 12:01 PM, Joonas Lahtinen wrote:
On to, 2015-03-19 at 15:07 +, Tvrtko Ursulin wrote:
Hi,
On 03/19/2015 01:02 PM, Joonas Lahtinen wrote:
static inline
int i915_get_ggtt_vma_pages(struct i915_vma *vma)
Same rant about function signatures as on earlier patch, put all on the
same line like most of new the code has it.
Ok.
struct i915_ggtt_view {
enum i915_ggtt_view_type type;
struct sg_table *pages;
+
+union {
+struct intel_rotation_info rotation_info;
+};
In preparation for the memcmp way of comparing views, I would move this
be before the variable struct parts (namely sg_table *pages), and also
wrap it once more so the result would be like this:
[snip]
enum i915_ggtt_view_type type;
union {
struct {
struct intel_rotation_info info;
} rotated;
struct {
...
} partial;
};
// private bits go here, to be wrapped in their struct with view
// type comparing patches
struct sg_table *pages;
[snip]
That way it's clear which view owns what.
Hm, rotation info is not considered in comparing views, it is just a
bucket of data passed around between layers. So I suppose private data
under your design. Since there is no private union yet, maybe do this
later?
Why not? Isn't a 270 degree rotated view substantially different from a
90 degree rotated view (even when the difference technically is just
some bit flip somewhere else).
At least I would be pretty upset if I was returned the address for 90
degree rotated view when I wanted 270 rotated. If multiple rotated views
are not possible, then it is again an implicit thing.
There are quite a lot of hardware constraints like this that appear in
the code implicitly, which IMHO makes the code hard to follow at times.
So I'd try to make it more explicit that the views are not the same,
there just can be one rotated view at a time (if that is the case).
90 and 270 views are indeed the same page layout - same address for
scanout. And there can only be one such VMA for an object at a time.
But how this mapping needs to look like is determined by more than the
object itself - framebuffer geometry defines it. The private data in the
view is used to transfer that meta-data so the GTT core can build the
appropriate view.
Right, I think I understand your viewpoint now. I would still prefer it
to be even more explicit like I915_GGTT_VIEW_Yf_SCANOUT, because what it
really does is rearranges the pages to layout suitable for rotated
scanout, not making the view rotated in itself.
And in that case it would not be in the memcmp range, like it is not
currently.
Regards, Joonas
That was my argument in fact for not putting the page shuffling bit in
i915_gem_gtt.c since it is really display engine ownership.
Regards,
Tvrtko
___
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
On 03/20/2015 01:31 PM, Joonas Lahtinen wrote: [snip] Hm, rotation info is not considered in comparing views, it is just a bucket of data passed around between layers. So I suppose private data under your design. Since there is no private union yet, maybe do this later? Why not? Isn't a 270 degree rotated view substantially different from a 90 degree rotated view (even when the difference technically is just some bit flip somewhere else). At least I would be pretty upset if I was returned the address for 90 degree rotated view when I wanted 270 rotated. If multiple rotated views are not possible, then it is again an implicit thing. There are quite a lot of hardware constraints like this that appear in the code implicitly, which IMHO makes the code hard to follow at times. So I'd try to make it more explicit that the views are not the same, there just can be one rotated view at a time (if that is the case). 90 and 270 views are indeed the same page layout - same address for scanout. And there can only be one such VMA for an object at a time. But how this mapping needs to look like is determined by more than the object itself - framebuffer geometry defines it. The private data in the view is used to transfer that meta-data so the GTT core can build the appropriate view. Right, I think I understand your viewpoint now. I would still prefer it to be even more explicit like I915_GGTT_VIEW_Yf_SCANOUT, because what it really does is rearranges the pages to layout suitable for rotated scanout, not making the view rotated in itself. Ha, not sure what you mean, but Yf is definitely not the right name for it since it is a different thing altogether. It is rotated in a way that tiles (==pages) are rotated by 90 degrees. But content within the tile is not (display engine handles that bit). In your mind that would maybe be I915_GGTT_VIEW_ROTATED_PAGES_FOR_ROTATED_SCANOUT? :) I am guessing only, but I still think I915_GGTT_VIEW_ROTATED is fine. Regards, Tvrtko ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
On to, 2015-03-19 at 15:07 +, Tvrtko Ursulin wrote:
Hi,
On 03/19/2015 01:02 PM, Joonas Lahtinen wrote:
static inline
int i915_get_ggtt_vma_pages(struct i915_vma *vma)
Same rant about function signatures as on earlier patch, put all on the
same line like most of new the code has it.
Ok.
struct i915_ggtt_view {
enum i915_ggtt_view_type type;
struct sg_table *pages;
+
+ union {
+ struct intel_rotation_info rotation_info;
+ };
In preparation for the memcmp way of comparing views, I would move this
be before the variable struct parts (namely sg_table *pages), and also
wrap it once more so the result would be like this:
[snip]
enum i915_ggtt_view_type type;
union {
struct {
struct intel_rotation_info info;
} rotated;
struct {
...
} partial;
};
// private bits go here, to be wrapped in their struct with view
// type comparing patches
struct sg_table *pages;
[snip]
That way it's clear which view owns what.
Hm, rotation info is not considered in comparing views, it is just a
bucket of data passed around between layers. So I suppose private data
under your design. Since there is no private union yet, maybe do this later?
Why not? Isn't a 270 degree rotated view substantially different from a
90 degree rotated view (even when the difference technically is just
some bit flip somewhere else).
At least I would be pretty upset if I was returned the address for 90
degree rotated view when I wanted 270 rotated. If multiple rotated views
are not possible, then it is again an implicit thing.
There are quite a lot of hardware constraints like this that appear in
the code implicitly, which IMHO makes the code hard to follow at times.
So I'd try to make it more explicit that the views are not the same,
there just can be one rotated view at a time (if that is the case).
extern const struct i915_ggtt_view i915_ggtt_view_normal;
diff --git a/drivers/gpu/drm/i915/intel_display.c
b/drivers/gpu/drm/i915/intel_display.c
index fe11e99..c2c3a76 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -2194,7 +2194,7 @@ static bool need_vtd_wa(struct drm_device *dev)
return false;
}
-static unsigned int
+unsigned int
intel_tile_height(struct drm_device *dev, uint32_t pixel_format,
uint64_t fb_format_modifier)
{
@@ -2254,9 +2254,34 @@ int intel_fill_fb_ggtt_view(struct i915_ggtt_view
*view,
struct drm_framebuffer *fb,
const struct drm_plane_state *plane_state)
{
+ struct intel_rotation_info *info = view-rotation_info;
+ static const struct i915_ggtt_view rotated_view =
+ { .type = I915_GGTT_VIEW_ROTATED };
+
*view = i915_ggtt_view_normal;
- return 0;
+ if (!plane_state)
+ return 0;
+
+ if (!(plane_state-rotation
+ (BIT(DRM_ROTATE_90) | BIT(DRM_ROTATE_270
+ return 0;
Should the rotation checking helper introduced in previous patch be used
here too?
It's the other way round, following patch adds the helper and replaces
this with a call to it.
Yes, my bad, I got:)
Regards, Joonas
Regards,
Tvrtko
___
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
On 03/20/2015 12:01 PM, Joonas Lahtinen wrote:
On to, 2015-03-19 at 15:07 +, Tvrtko Ursulin wrote:
Hi,
On 03/19/2015 01:02 PM, Joonas Lahtinen wrote:
static inline
int i915_get_ggtt_vma_pages(struct i915_vma *vma)
Same rant about function signatures as on earlier patch, put all on the
same line like most of new the code has it.
Ok.
struct i915_ggtt_view {
enum i915_ggtt_view_type type;
struct sg_table *pages;
+
+ union {
+ struct intel_rotation_info rotation_info;
+ };
In preparation for the memcmp way of comparing views, I would move this
be before the variable struct parts (namely sg_table *pages), and also
wrap it once more so the result would be like this:
[snip]
enum i915_ggtt_view_type type;
union {
struct {
struct intel_rotation_info info;
} rotated;
struct {
...
} partial;
};
// private bits go here, to be wrapped in their struct with view
// type comparing patches
struct sg_table *pages;
[snip]
That way it's clear which view owns what.
Hm, rotation info is not considered in comparing views, it is just a
bucket of data passed around between layers. So I suppose private data
under your design. Since there is no private union yet, maybe do this later?
Why not? Isn't a 270 degree rotated view substantially different from a
90 degree rotated view (even when the difference technically is just
some bit flip somewhere else).
At least I would be pretty upset if I was returned the address for 90
degree rotated view when I wanted 270 rotated. If multiple rotated views
are not possible, then it is again an implicit thing.
There are quite a lot of hardware constraints like this that appear in
the code implicitly, which IMHO makes the code hard to follow at times.
So I'd try to make it more explicit that the views are not the same,
there just can be one rotated view at a time (if that is the case).
90 and 270 views are indeed the same page layout - same address for
scanout. And there can only be one such VMA for an object at a time.
But how this mapping needs to look like is determined by more than the
object itself - framebuffer geometry defines it. The private data in the
view is used to transfer that meta-data so the GTT core can build the
appropriate view.
That was my argument in fact for not putting the page shuffling bit in
i915_gem_gtt.c since it is really display engine ownership.
Regards,
Tvrtko
___
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
On ti, 2015-03-17 at 15:45 +, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursu...@intel.com
90/270 rotated scanout needs a rotated GTT view of the framebuffer.
This is put in a separate VMA with a dedicated ggtt view and wired such that
it is created when a framebuffer is pinned to a 90/270 rotated plane.
Rotation is only possible with Yb/Yf buffers and error is propagated to
user space in case of a mismatch.
Special rotated page view is constructed at the VMA creation time by
borrowing the DMA addresses from obj-pages.
v2:
* Do not bother with pages for rotated sg list, just populate the DMA
addresses. (Daniel Vetter)
* Checkpatch cleanup.
v3:
* Rebased on top of new plane handling (create rotated mapping when
setting the rotation property).
* Unpin rotated VMA on unpinning from display plane.
* Simplify rotation check using bitwise AND. (Chris Wilson)
v4:
* Fix unpinning of optional rotated mapping so it is really considered
to be optional.
v5:
* Rebased for fb modifier changes.
* Rebased for atomic commit.
* Only pin needed view for display. (Ville Syrjälä, Daniel Vetter)
v6:
* Rebased after preparatory work has been extracted out. (Daniel Vetter)
v7:
* Slightly simplified tiling geometry calculation.
* Moved rotated GGTT view implementation into i915_gem_gtt.c (Daniel
Vetter)
v8:
* Do not use i915_gem_obj_size to get object size since that actually
returns the size of an VMA which may not exist.
* Rebased for ggtt view changes.
For: VIZ-4726
Signed-off-by: Tvrtko Ursulin tvrtko.ursu...@intel.com
Reviewed-by: Michel Thierry michel.thie...@intel.com (v4)
Cc: Joonas Lahtinen joonas.lahti...@linux.intel.com
---
drivers/gpu/drm/i915/i915_gem_gtt.c | 113
++-
drivers/gpu/drm/i915/i915_gem_gtt.h | 12
drivers/gpu/drm/i915/intel_display.c | 29 -
drivers/gpu/drm/i915/intel_drv.h | 4 ++
4 files changed, 154 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c
b/drivers/gpu/drm/i915/i915_gem_gtt.c
index f1b9ea6..5381ebf 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2413,15 +2413,119 @@ i915_gem_obj_lookup_or_create_ggtt_vma(struct
drm_i915_gem_object *obj,
}
+static
+void rotate_pages(dma_addr_t *in, unsigned int width, unsigned int height,
+ struct sg_table *st)
+{
+ unsigned int column, row;
+ unsigned int src_idx;
+ struct scatterlist *sg = st-sgl;
+
+ st-nents = 0;
+
+ for (column = 0; column width; column++) {
+ src_idx = width * (height - 1) + column;
+ for (row = 0; row height; row++) {
+ st-nents++;
+ /* We don't need the pages, but need to initialize
+ * the entries so the sg list can be happily traversed.
+ * The only thing we need are DMA addresses.
+ */
+ sg_set_page(sg, NULL, PAGE_SIZE, 0);
+ sg_dma_address(sg) = in[src_idx];
+ sg_dma_len(sg) = PAGE_SIZE;
+ sg = sg_next(sg);
+ src_idx -= width;
+ }
+ }
+}
+
+static
+struct sg_table *intel_rotate_fb_obj_pages(struct i915_ggtt_view *ggtt_view,
+struct drm_i915_gem_object *obj)
+{
+ struct drm_device *dev = obj-base.dev;
+ struct intel_rotation_info *rot_info = ggtt_view-rotation_info;
+ unsigned long size, pages, rot_pages;
+ struct sg_page_iter sg_iter;
+ unsigned long i;
+ dma_addr_t *page_addr_list;
+ struct sg_table *st;
+ unsigned int tile_pitch, tile_height;
+ unsigned int width_pages, height_pages;
+ int ret = ENOMEM;
+
+ pages = obj-base.size / PAGE_SIZE;
+
+ /* Calculate tiling geometry. */
+ tile_height = intel_tile_height(dev, rot_info-pixel_format,
+ rot_info-fb_modifier);
+ tile_pitch = PAGE_SIZE / tile_height;
+ width_pages = DIV_ROUND_UP(rot_info-pitch, tile_pitch);
+ height_pages = DIV_ROUND_UP(rot_info-height, tile_height);
+ rot_pages = width_pages * height_pages;
+ size = rot_pages * PAGE_SIZE;
+
+ /* Allocate a temporary list of source pages for random access. */
+ page_addr_list = drm_malloc_ab(pages, sizeof(dma_addr_t));
+ if (!page_addr_list)
+ return ERR_PTR(ret);
+
+ /* Allocate target SG list. */
+ st = kmalloc(sizeof(*st), GFP_KERNEL);
+ if (!st)
+ goto err_st_alloc;
+
+ ret = sg_alloc_table(st, rot_pages, GFP_KERNEL);
+ if (ret)
+ goto err_sg_alloc;
+
+ /* Populate source page list from the object. */
+ i = 0;
+ for_each_sg_page(obj-pages-sgl, sg_iter,
Re: [Intel-gfx] [PATCH 5/7] drm/i915/skl: Support secondary (rotated) frame buffer mapping
Hi,
On 03/19/2015 01:02 PM, Joonas Lahtinen wrote:
static inline
int i915_get_ggtt_vma_pages(struct i915_vma *vma)
Same rant about function signatures as on earlier patch, put all on the
same line like most of new the code has it.
Ok.
struct i915_ggtt_view {
enum i915_ggtt_view_type type;
struct sg_table *pages;
+
+ union {
+ struct intel_rotation_info rotation_info;
+ };
In preparation for the memcmp way of comparing views, I would move this
be before the variable struct parts (namely sg_table *pages), and also
wrap it once more so the result would be like this:
[snip]
enum i915_ggtt_view_type type;
union {
struct {
struct intel_rotation_info info;
} rotated;
struct {
...
} partial;
};
// private bits go here, to be wrapped in their struct with view
// type comparing patches
struct sg_table *pages;
[snip]
That way it's clear which view owns what.
Hm, rotation info is not considered in comparing views, it is just a
bucket of data passed around between layers. So I suppose private data
under your design. Since there is no private union yet, maybe do this later?
extern const struct i915_ggtt_view i915_ggtt_view_normal;
diff --git a/drivers/gpu/drm/i915/intel_display.c
b/drivers/gpu/drm/i915/intel_display.c
index fe11e99..c2c3a76 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -2194,7 +2194,7 @@ static bool need_vtd_wa(struct drm_device *dev)
return false;
}
-static unsigned int
+unsigned int
intel_tile_height(struct drm_device *dev, uint32_t pixel_format,
uint64_t fb_format_modifier)
{
@@ -2254,9 +2254,34 @@ int intel_fill_fb_ggtt_view(struct i915_ggtt_view *view,
struct drm_framebuffer *fb,
const struct drm_plane_state *plane_state)
{
+ struct intel_rotation_info *info = view-rotation_info;
+ static const struct i915_ggtt_view rotated_view =
+ { .type = I915_GGTT_VIEW_ROTATED };
+
*view = i915_ggtt_view_normal;
- return 0;
+ if (!plane_state)
+ return 0;
+
+ if (!(plane_state-rotation
+ (BIT(DRM_ROTATE_90) | BIT(DRM_ROTATE_270
+ return 0;
Should the rotation checking helper introduced in previous patch be used
here too?
It's the other way round, following patch adds the helper and replaces
this with a call to it.
Regards,
Tvrtko
___
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
