[Solved]How to get Linux framebuffer fully displayed on an interlaced mode DVO?

Pls explain your data flow for video and the changes you did to display full video frame when showing 1080i60 in displayLink. Also mention what changes you did to the grpx driver

Badri Narayanan said:

Pls explain your data flow for video and the changes you did to display full video frame when showing 1080i60 in displayLink. Also mention what changes you did to the grpx driver

Hi Narayanan, thanks for you replay. Here are my data flows:

(YUVSP420)                     (1920x1080P YUYV420)

VIP0-------->Merge------->SwMs----------------------->Display(DVO2 with 1080I@60)

              ^                                         ^

(YUVSP420)    |                                         | (scale up to 1920x1080)

VIP1----------|                                                                                         linux fb0(ARGB 1280x720)

two captures will be merged to SwMs, which produces 1920x1080P video frame.

Linux fb0 is 1280x720. It will be scaled up to 1920x1080 by Graphic built-in SC.

For displayLink_drv.c I just made a few changes:

In DisplayLink_drvProcessData, I changed

pFrame->addr[1][0] =
    (UInt8 *) pFrame->addr[0][0] + pitch0;
pFrame->addr[1][1] =
    (UInt8 *) pFrame->addr[0][1] + pitch1;

to

pFrame->addr[1][0] =
    (UInt8 *) pFrame->addr[0][0];
pFrame->addr[1][1] =
    (UInt8 *) pFrame->addr[0][1];

This may be strange, because according to the document of FVID2,

the former one is right. But in actually the video will be displayed abnormally.

I read relative sources of hdvpss and know that when filedMerged member in FVID2_Format

is set, the VPDMA will work with a effective pitch with double.

So I try the latter one, to my surprise, the video was displayed normally.

I think maybe the Graphics have the same behaviour.

So I made a few changes in grpx.c in kernel.

The first one:

static int vps_grpx_start(struct vps_grpx_ctrl *gctrl)
{
    int r = 0;
    VPSSDBG("start grpx%d\n", gctrl->grpx_num);
    if ((gctrl == NULL) || (gctrl->handle == NULL))
        return -EINVAL;

    if (!gctrl->gstate.isstarted) {
        grpx_pre_start(gctrl);
        /*start everything over, set format,
          params, queue buffer*/
#if 1  /* xyf for test */
        gctrl->inputf->fieldmerged[0] = true;
        gctrl->inputf->fieldmerged[1] = true;
        gctrl->inputf->fieldmerged[2] = true;
        gctrl->inputf->scanformat = FVID2_SF_INTERLACED;
#endif

The second one:

static int vps_grpx_set_buffer(struct vps_grpx_ctrl *gctrl,
             u32 buffer_addr)
{
    int r = 0;

    VPSSDBG("(%d)- add buffer %#x\n", gctrl->grpx_num, buffer_addr);

    grpx_cs_lock(gctrl);
    gctrl->buffer_addr = buffer_addr;
    gctrl->frames->addr[FVID2_RGB_ADDR_IDX]
        [FVID2_RGB_ADDR_IDX] = (void *)buffer_addr;
    /* BEGIN: Added by xyf, 2013/09/02   PN:pn*/
    gctrl->frames->addr[1]
        [FVID2_RGB_ADDR_IDX] = (void *)(buffer_addr);
    /* END:   Added by xyf, 2013/09/02 */

    grpx_cs_free(gctrl);
    return r;
}
But the grpx couldn't be started.

Is your capture 1080P60 capture ? WHat is the swms outputFPS and are you setting displayLink create params. forceFieldSeparatedInputMode to TRUE.

Hi Json Hsui,

I am also using IPNC_RDK_3.8.0. I want to use framebuffer device nodes for QT applications on my board.

I am able to create /dev/fb0 node , when I insert ti81xxfb.ko kernel module. But when my application tries to open that node, it gives an error.

Driver says: ti81xxfb: failed to start

debugging the issue further, I could find that error comes from grpx.c file

function vps_fvid2_control is returning -1, when vps_fvid2_start is called.

Please assist me on this issue.

I am juggling through this, for last few days. Any help would be highly appreciated.

Thanks,

Krinali Shah