[PATCH v2 2/8] drm/kmb : W/A for 256B cache alignment for video
Anitha Chrisanthus
anitha.chrisanthus at intel.com
Tue Aug 3 22:04:46 UTC 2021
For B0 silicon, the media driver pads the decoded video dmabufs for 256B
alignment. This is the backing buffer of the framebuffer and info in the
drm frame buffer is not correct for these buffers as this is done
internally in the media driver. This change extracts the meta data info
from dmabuf priv structure and uses that info for programming stride and
offsets in kmb_plane_atomic_update().
v2: simplified logic
Fixes: 7f7b96a8a0a1 ("drm/kmb: Add support for KeemBay Display")
Signed-off-by: Edmund Dea <edmund.j.dea at intel.com>
Signed-off-by: Anitha Chrisanthus <anitha.chrisanthus at intel.com>
---
drivers/gpu/drm/kmb/kmb_plane.c | 58 ++++++++++++++++++++++++++------
drivers/gpu/drm/kmb/kmb_vidmem.h | 52 ++++++++++++++++++++++++++++
2 files changed, 99 insertions(+), 11 deletions(-)
create mode 100644 drivers/gpu/drm/kmb/kmb_vidmem.h
diff --git a/drivers/gpu/drm/kmb/kmb_plane.c b/drivers/gpu/drm/kmb/kmb_plane.c
index ecee6782612d..e61c2798c206 100644
--- a/drivers/gpu/drm/kmb/kmb_plane.c
+++ b/drivers/gpu/drm/kmb/kmb_plane.c
@@ -3,6 +3,8 @@
* Copyright © 2018-2020 Intel Corporation
*/
+#include <linux/dma-buf.h>
+
#include <drm/drm_atomic.h>
#include <drm/drm_atomic_helper.h>
#include <drm/drm_crtc.h>
@@ -11,12 +13,14 @@
#include <drm/drm_fb_helper.h>
#include <drm/drm_fourcc.h>
#include <drm/drm_gem_cma_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
#include <drm/drm_managed.h>
#include <drm/drm_plane_helper.h>
#include "kmb_drv.h"
#include "kmb_plane.h"
#include "kmb_regs.h"
+#include "kmb_vidmem.h"
const u32 layer_irqs[] = {
LCD_INT_VL0,
@@ -296,6 +300,9 @@ static void kmb_plane_atomic_update(struct drm_plane *plane,
unsigned char plane_id;
int num_planes;
static dma_addr_t addr[MAX_SUB_PLANES];
+ struct viv_vidmem_metadata *md = NULL;
+ struct drm_gem_object *gem_obj;
+ unsigned int cb_stride, cr_stride;
if (!plane || !new_plane_state || !old_plane_state)
return;
@@ -325,6 +332,16 @@ static void kmb_plane_atomic_update(struct drm_plane *plane,
drm_dbg(&kmb->drm,
"src_w=%d src_h=%d, fb->format->format=0x%x fb->flags=0x%x\n",
src_w, src_h, fb->format->format, fb->flags);
+ gem_obj = drm_gem_fb_get_obj(fb, plane_id);
+ if (gem_obj && gem_obj->import_attach &&
+ gem_obj->import_attach->dmabuf &&
+ gem_obj->import_attach->dmabuf->priv) {
+ md = gem_obj->import_attach->dmabuf->priv;
+
+ /* Check if metadata is coming from hantro driver */
+ if (md->magic != HANTRO_IMAGE_VIV_META_DATA_MAGIC)
+ md = NULL;
+ }
width = fb->width;
height = fb->height;
@@ -332,8 +349,13 @@ static void kmb_plane_atomic_update(struct drm_plane *plane,
drm_dbg(&kmb->drm, "dma_len=%d ", dma_len);
kmb_write_lcd(kmb, LCD_LAYERn_DMA_LEN(plane_id), dma_len);
kmb_write_lcd(kmb, LCD_LAYERn_DMA_LEN_SHADOW(plane_id), dma_len);
- kmb_write_lcd(kmb, LCD_LAYERn_DMA_LINE_VSTRIDE(plane_id),
- fb->pitches[0]);
+ if (md) {
+ kmb_write_lcd(kmb, LCD_LAYERn_DMA_LINE_VSTRIDE(plane_id),
+ md->plane[0].stride);
+ } else {
+ kmb_write_lcd(kmb, LCD_LAYERn_DMA_LINE_VSTRIDE(plane_id),
+ fb->pitches[0]);
+ }
kmb_write_lcd(kmb, LCD_LAYERn_DMA_LINE_WIDTH(plane_id),
(width * fb->format->cpp[0]));
@@ -344,13 +366,19 @@ static void kmb_plane_atomic_update(struct drm_plane *plane,
val |= get_bits_per_pixel(fb->format);
/* Program Cb/Cr for planar formats */
if (num_planes > 1) {
- kmb_write_lcd(kmb, LCD_LAYERn_DMA_CB_LINE_VSTRIDE(plane_id),
- width * fb->format->cpp[0]);
- kmb_write_lcd(kmb, LCD_LAYERn_DMA_CB_LINE_WIDTH(plane_id),
- (width * fb->format->cpp[0]));
-
- addr[U_PLANE] = drm_fb_cma_get_gem_addr(fb, new_plane_state,
- U_PLANE);
+ cb_stride = md ? md->plane[1].stride :
+ width * fb->format->cpp[0];
+ kmb_write_lcd(kmb,
+ LCD_LAYERn_DMA_CB_LINE_VSTRIDE(plane_id),
+ cb_stride);
+
+ if (md) {
+ addr[U_PLANE] = addr[Y_PLANE] + md->plane[1].offset;
+ } else {
+ addr[U_PLANE] = drm_fb_cma_get_gem_addr(fb,
+ new_plane_state,
+ U_PLANE);
+ }
/* check if Cb/Cr is swapped*/
if (num_planes == 3 && (val & LCD_LAYER_CRCB_ORDER))
kmb_write_lcd(kmb,
@@ -362,17 +390,25 @@ static void kmb_plane_atomic_update(struct drm_plane *plane,
addr[U_PLANE]);
if (num_planes == 3) {
+ cr_stride = md ? md->plane[2].stride :
+ width * fb->format->cpp[0];
kmb_write_lcd(kmb,
LCD_LAYERn_DMA_CR_LINE_VSTRIDE(plane_id),
- ((width) * fb->format->cpp[0]));
+ cr_stride);
kmb_write_lcd(kmb,
LCD_LAYERn_DMA_CR_LINE_WIDTH(plane_id),
((width) * fb->format->cpp[0]));
- addr[V_PLANE] = drm_fb_cma_get_gem_addr(fb,
+ if (md) {
+ addr[V_PLANE] = addr[Y_PLANE] +
+ md->plane[2].offset;
+ } else {
+ addr[V_PLANE] =
+ drm_fb_cma_get_gem_addr(fb,
new_plane_state,
V_PLANE);
+ }
/* check if Cb/Cr is swapped*/
if (val & LCD_LAYER_CRCB_ORDER)
diff --git a/drivers/gpu/drm/kmb/kmb_vidmem.h b/drivers/gpu/drm/kmb/kmb_vidmem.h
new file mode 100644
index 000000000000..06198d413f50
--- /dev/null
+++ b/drivers/gpu/drm/kmb/kmb_vidmem.h
@@ -0,0 +1,52 @@
+/* SPDX-License-Identifier: GPL-2.0-only
+ *
+ * Copyright © 2018-2020 Intel Corporation
+ */
+
+#ifndef __KMB_VIDMEM_H__
+#define __KMB_VIDMEM_H__
+
+#define HANTRO_MAGIC(ch0, ch1, ch2, ch3) \
+ ((unsigned long)(unsigned char)(ch0) | \
+ ((unsigned long)(unsigned char)(ch1) << 8) | \
+ ((unsigned long)(unsigned char)(ch2) << 16) | \
+ ((unsigned long)(unsigned char)(ch3) << 24))
+
+#define HANTRO_IMAGE_VIV_META_DATA_MAGIC HANTRO_MAGIC('V', 'I', 'V', 'M')
+
+struct viv_vidmem_metadata {
+ u32 magic; // __FOURCC('v', 'i', 'v', 'm')
+ u32 dmabuf_size; // DMABUF buffer size in byte (Maximum 4GB)
+ u32 time_stamp; // time stamp for the DMABUF buffer
+
+ u32 image_format; // ImageFormat, determined plane number.
+ u32 compressed; // if DMABUF buffer is compressed by DEC400
+ struct {
+ u32 offset; // plane buffer address offset from DMABUF address
+ u32 stride; // pitch in bytes
+ u32 width; // width in pixels
+ u32 height; // height in pixels
+
+ u32 tile_format; // uncompressed tile format
+ u32 compress_format; // tile mode for DEC400
+
+ /** tile status buffer offset within this plane buffer. when it is 0，
+ * indicates using separate tile status buffer
+ */
+ u32 ts_offset;
+ /** fd of separate tile status buffer of the plane buffer */
+ u32 ts_fd;
+ /** valid fd of the ts buffer in consumer side. */
+ u32 ts_fd2;
+ /** the vpu virtual address for this ts data buffer */
+ u32 ts_vaddr;
+
+ /** gpu fastclear enabled for the plane buffer */
+ u32 fc_enabled;
+ /** gpu fastclear color value (lower 32 bits) for the plane buffer */
+ u32 fc_value_lower;
+ /** gpu fastclear color value (upper 32 bits) for the plane buffer */
+ u32 fc_value_upper;
+ } plane[3];
+};
+#endif /*__KMB_VIDMEM_H__*/
--
2.25.1
More information about the dri-devel
mailing list