On Mon, 26 Jan 2009, morimoto.kuninori@xxxxxxxxxxx wrote:
>
> Dear Guennadi
>
> > Have you tested with v4l-dvb/v4l2-apps/test/capture_example.c? I think it
> > wouldn't work, because it first calls S_CROP, and then S_FMT, and even
> > with this your patch you'd fail S_CROP if S_FMT hadn't been called before
> > (priv->fmt == NULL). Am I right?
>
> hmm.. I would like to ask you about S_CROP on soc_camera.
> There are a lot of problem for me.
>
> at first, sh_mobile_ceu (and pxa_camera too) :: set_fmt is this
> -----------------------------------------------
> xlate = soc_camera_xlate_by_fourcc(icd, pixfmt);
> if (!xlate) {
> dev_warn(&ici->dev, "Format %x not found\n", pixfmt);
> return -EINVAL;
> }
> -----------------------------------------------
>
> and soc_camera_xlate_by_fourcc is this
> -----------------------------------------------
> const struct soc_camera_format_xlate *soc_camera_xlate_by_fourcc(
> struct soc_camera_device *icd, unsigned int fourcc)
> {
> unsigned int i;
>
> for (i = 0; i < icd->num_user_formats; i++)
> if (icd->user_formats[i].host_fmt->fourcc == fourcc)
> return icd->user_formats + i;
> return NULL;
> }
> -----------------------------------------------
>
> If pixfmt is 0, it said "Format 0 not found" to me.
Yes, this will be fixed by this:
http://www.mail-archive.com/linux-media@xxxxxxxxxxxxxxx/msg00161.html
> even if I fixed this problem (and ov772x set_fmt),
> kernel will run BUG_ON.
>
> call trace is this.
> -------------
> soc_camera_qbuf
> -> videobuf_qbuf
> -> videobuf_next_field
> -------------
>
> kernel will run BUG_ON when I use capture_example.
> -------------------------------------------------
> /* Locking: Caller holds q->vb_lock */
> enum v4l2_field videobuf_next_field(struct videobuf_queue *q)
> {
> enum v4l2_field field = q->field;
>
> => BUG_ON(V4L2_FIELD_ANY == field);
> -------------------------------------------------
>
> If sh_mobile_ceu use V4L2_FIELD_NONE on sh_mobile_ceu_init_videobuf,
> kernel will not run this BUG_ON.
> But tw9910 and I want sh_mobile_ceu_init_videobuf to use V4L2_FIELD_ANY.
> Do you remember this problem ?
Well, I looked through the old emails, but chenged a few things since
then, so I am not sure what the behaviour would be like now, if we do
switch back to _NONE? It looks like you should be prepared to run without
a call to S_FMT, which also means, your ->field should have a value other
than _ANY. What would break if you use _ANY now?
> And capture_example doesn't works even If
> sh_mobile_ceu_init_videobuf use V4L2_FIELD_NONE.
> It said "select timeout" to me.
> I don't know why.
This is probably because you don't setup a default mode in
sh_mobile_ceu_camera.c or ov772x.c if S_FMT is never called. This should
be fixed. I'm not sure if other drivers would deliver anything meaningful
without a call to S_FMT - has to be tested / fixed...
> what is the best way to us ???
> or do I miss understanding ???
Fix behaviour if no S_FMT is done.
> > it first calls S_CROP, and then S_FMT
>
> In this case, shoud ov772x use defaul color format
> when S_CROP order ?
The thing is that older versions of capture.c always called S_FMT, newer
ones don't unless forced per "-f" switch.
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
