]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/commitdiff
media: vsp1: Support Interlaced display pipelines
authorKieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Fri, 3 Aug 2018 11:37:29 +0000 (07:37 -0400)
committerMauro Carvalho Chehab <mchehab+samsung@kernel.org>
Fri, 3 Aug 2018 20:02:27 +0000 (16:02 -0400)
Calculate the top and bottom fields for the interlaced frames and
utilise the extended display list command feature to implement the
auto-field operations. This allows the DU to update the VSP2 registers
dynamically based upon the currently processing field.

Signed-off-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
drivers/media/platform/vsp1/vsp1_dl.c
drivers/media/platform/vsp1/vsp1_drm.c
drivers/media/platform/vsp1/vsp1_pipe.h
drivers/media/platform/vsp1/vsp1_regs.h
drivers/media/platform/vsp1/vsp1_rpf.c
include/media/vsp1.h

index 6021c43c53b43e42135229e9982e6900de58f8a5..9255b5ee2cb877c1895d8821681f5503b82b2b0c 100644 (file)
@@ -947,6 +947,8 @@ void vsp1_dl_list_commit(struct vsp1_dl_list *dl, bool internal)
  */
 unsigned int vsp1_dlm_irq_frame_end(struct vsp1_dl_manager *dlm)
 {
+       struct vsp1_device *vsp1 = dlm->vsp1;
+       u32 status = vsp1_read(vsp1, VI6_STATUS);
        unsigned int flags = 0;
 
        spin_lock(&dlm->lock);
@@ -971,6 +973,14 @@ unsigned int vsp1_dlm_irq_frame_end(struct vsp1_dl_manager *dlm)
        if (vsp1_dl_list_hw_update_pending(dlm))
                goto done;
 
+       /*
+        * Progressive streams report only TOP fields. If we have a BOTTOM
+        * field, we are interlaced, and expect the frame to complete on the
+        * next frame end interrupt.
+        */
+       if (status & VI6_STATUS_FLD_STD(dlm->index))
+               goto done;
+
        /*
         * The device starts processing the queued display list right after the
         * frame end interrupt. The display list thus becomes active.
index a168568567894dcd99fd85fabaa308cc28563e07..87dc3ddbe2d387fb6dd5b60f7535cb255188a845 100644 (file)
@@ -670,9 +670,11 @@ int vsp1_du_setup_lif(struct device *dev, unsigned int pipe_index,
 
        drm_pipe->width = cfg->width;
        drm_pipe->height = cfg->height;
+       pipe->interlaced = cfg->interlaced;
 
-       dev_dbg(vsp1->dev, "%s: configuring LIF%u with format %ux%u\n",
-               __func__, pipe_index, cfg->width, cfg->height);
+       dev_dbg(vsp1->dev, "%s: configuring LIF%u with format %ux%u%s\n",
+               __func__, pipe_index, cfg->width, cfg->height,
+               pipe->interlaced ? "i" : "");
 
        mutex_lock(&vsp1->drm->lock);
 
index 743d8f0db45c87ba4aaa29bf47ff6a6c0f2ca16b..ae646c9ef337317ad1b7c1b589f3bfd080af363e 100644 (file)
@@ -104,6 +104,7 @@ struct vsp1_partition {
  * @entities: list of entities in the pipeline
  * @stream_config: cached stream configuration for video pipelines
  * @configured: when false the @stream_config shall be written to the hardware
+ * @interlaced: True when the pipeline is configured in interlaced mode
  * @partitions: The number of partitions used to process one frame
  * @partition: The current partition for configuration to process
  * @part_table: The pre-calculated partitions used by the pipeline
@@ -142,6 +143,7 @@ struct vsp1_pipeline {
 
        struct vsp1_dl_body *stream_config;
        bool configured;
+       bool interlaced;
 
        unsigned int partitions;
        struct vsp1_partition *partition;
index 5ea9f9070cf376fcb1e3d1a46032f5c7ade02c76..3738ff2f7b850052565d3b131d559fad048103d7 100644 (file)
@@ -28,6 +28,7 @@
 #define VI6_SRESET_SRTS(n)             (1 << (n))
 
 #define VI6_STATUS                     0x0038
+#define VI6_STATUS_FLD_STD(n)          (1 << ((n) + 28))
 #define VI6_STATUS_SYS_ACT(n)          (1 << ((n) + 8))
 
 #define VI6_WPF_IRQ_ENB(n)             (0x0048 + (n) * 12)
@@ -80,6 +81,8 @@
 #define VI6_DL_EXT_CTRL_EXPRI          (1 << 4)
 #define VI6_DL_EXT_CTRL_EXT            (1 << 0)
 
+#define VI6_DL_EXT_AUTOFLD_INT         BIT(0)
+
 #define VI6_DL_BODY_SIZE               0x0120
 #define VI6_DL_BODY_SIZE_UPD           (1 << 24)
 #define VI6_DL_BODY_SIZE_BS_MASK       (0x1ffff << 0)
index 69e5fe6e6b5072e7a60137c8e86a705e89a46a7b..f8005b60b9d239aa0fdd16a8b54bd1fc8903a5e9 100644 (file)
 #define RPF_MAX_WIDTH                          8190
 #define RPF_MAX_HEIGHT                         8190
 
+/* Pre extended display list command data structure. */
+struct vsp1_extcmd_auto_fld_body {
+       u32 top_y0;
+       u32 bottom_y0;
+       u32 top_c0;
+       u32 bottom_c0;
+       u32 top_c1;
+       u32 bottom_c1;
+       u32 reserved0;
+       u32 reserved1;
+} __packed;
+
 /* -----------------------------------------------------------------------------
  * Device Access
  */
@@ -64,6 +76,14 @@ static void rpf_configure_stream(struct vsp1_entity *entity,
                pstride |= format->plane_fmt[1].bytesperline
                        << VI6_RPF_SRCM_PSTRIDE_C_SHIFT;
 
+       /*
+        * pstride has both STRIDE_Y and STRIDE_C, but multiplying the whole
+        * of pstride by 2 is conveniently OK here as we are multiplying both
+        * values.
+        */
+       if (pipe->interlaced)
+               pstride *= 2;
+
        vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_PSTRIDE, pstride);
 
        /* Format */
@@ -100,6 +120,9 @@ static void rpf_configure_stream(struct vsp1_entity *entity,
                top = compose->top;
        }
 
+       if (pipe->interlaced)
+               top /= 2;
+
        vsp1_rpf_write(rpf, dlb, VI6_RPF_LOC,
                       (left << VI6_RPF_LOC_HCOORD_SHIFT) |
                       (top << VI6_RPF_LOC_VCOORD_SHIFT));
@@ -169,6 +192,36 @@ static void rpf_configure_stream(struct vsp1_entity *entity,
 
 }
 
+static void vsp1_rpf_configure_autofld(struct vsp1_rwpf *rpf,
+                                      struct vsp1_dl_list *dl)
+{
+       const struct v4l2_pix_format_mplane *format = &rpf->format;
+       struct vsp1_dl_ext_cmd *cmd;
+       struct vsp1_extcmd_auto_fld_body *auto_fld;
+       u32 offset_y, offset_c;
+
+       cmd = vsp1_dl_get_pre_cmd(dl);
+       if (WARN_ONCE(!cmd, "Failed to obtain an autofld cmd"))
+               return;
+
+       /* Re-index our auto_fld to match the current RPF. */
+       auto_fld = cmd->data;
+       auto_fld = &auto_fld[rpf->entity.index];
+
+       auto_fld->top_y0 = rpf->mem.addr[0];
+       auto_fld->top_c0 = rpf->mem.addr[1];
+       auto_fld->top_c1 = rpf->mem.addr[2];
+
+       offset_y = format->plane_fmt[0].bytesperline;
+       offset_c = format->plane_fmt[1].bytesperline;
+
+       auto_fld->bottom_y0 = rpf->mem.addr[0] + offset_y;
+       auto_fld->bottom_c0 = rpf->mem.addr[1] + offset_c;
+       auto_fld->bottom_c1 = rpf->mem.addr[2] + offset_c;
+
+       cmd->flags |= VI6_DL_EXT_AUTOFLD_INT | BIT(16 + rpf->entity.index);
+}
+
 static void rpf_configure_frame(struct vsp1_entity *entity,
                                struct vsp1_pipeline *pipe,
                                struct vsp1_dl_list *dl,
@@ -221,6 +274,11 @@ static void rpf_configure_partition(struct vsp1_entity *entity,
                crop.left += pipe->partition->rpf.left;
        }
 
+       if (pipe->interlaced) {
+               crop.height = round_down(crop.height / 2, fmtinfo->vsub);
+               crop.top = round_down(crop.top / 2, fmtinfo->vsub);
+       }
+
        vsp1_rpf_write(rpf, dlb, VI6_RPF_SRC_BSIZE,
                       (crop.width << VI6_RPF_SRC_BSIZE_BHSIZE_SHIFT) |
                       (crop.height << VI6_RPF_SRC_BSIZE_BVSIZE_SHIFT));
@@ -249,9 +307,17 @@ static void rpf_configure_partition(struct vsp1_entity *entity,
            fmtinfo->swap_uv)
                swap(mem.addr[1], mem.addr[2]);
 
-       vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_Y, mem.addr[0]);
-       vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C0, mem.addr[1]);
-       vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C1, mem.addr[2]);
+       /*
+        * Interlaced pipelines will use the extended pre-cmd to process
+        * SRCM_ADDR_{Y,C0,C1}
+        */
+       if (pipe->interlaced) {
+               vsp1_rpf_configure_autofld(rpf, dl);
+       } else {
+               vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_Y, mem.addr[0]);
+               vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C0, mem.addr[1]);
+               vsp1_rpf_write(rpf, dlb, VI6_RPF_SRCM_ADDR_C1, mem.addr[2]);
+       }
 }
 
 static void rpf_partition(struct vsp1_entity *entity,
index 678c24de1ac6368d7f2a786b4bee04b07ffeaef0..3093b9cb9067e0a9599fdfd5d081e1ff08029f7b 100644 (file)
@@ -25,6 +25,7 @@ int vsp1_du_init(struct device *dev);
  * struct vsp1_du_lif_config - VSP LIF configuration
  * @width: output frame width
  * @height: output frame height
+ * @interlaced: true for interlaced pipelines
  * @callback: frame completion callback function (optional). When a callback
  *           is provided, the VSP driver guarantees that it will be called once
  *           and only once for each vsp1_du_atomic_flush() call.
@@ -33,6 +34,7 @@ int vsp1_du_init(struct device *dev);
 struct vsp1_du_lif_config {
        unsigned int width;
        unsigned int height;
+       bool interlaced;
 
        void (*callback)(void *data, bool completed, u32 crc);
        void *callback_data;