On 07/19/2016 05:55 PM, Jean Christophe TROTIN wrote:
> Hi Hans,
>
> Thank you for your comments.
> I've started to take them into account.
> I've got a question about V4L2_FIELD_ANY in buf_prepare (please see below).
>
> [snip]
>
> >> +static int hva_buf_prepare(struct vb2_buffer *vb)
> >> +{
> >> + struct hva_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
> >> + struct device *dev = ctx_to_dev(ctx);
> >> + struct vb2_v4l2_buffer *vbuf = to_vb2_v4l2_buffer(vb);
> >> +
> >> + if (vb->vb2_queue->type == V4L2_BUF_TYPE_VIDEO_OUTPUT) {
> >> + struct hva_frame *frame = to_hva_frame(vbuf);
> >> +
> >> + if (vbuf->field == V4L2_FIELD_ANY)
> >> + vbuf->field = V4L2_FIELD_NONE;
> >
> > Anything other than FIELD_NONE should result in an error since no interlaced
> is supported.
> > FIELD_ANY is an incorrect value as well.
> >
>
> In videodev2.h, V4L2_FIELD_ANY is commented as "driver can choose from none,
> top, bottom, interlaced depending on whatever it thinks is approximate ...": I
> understand this comment as if vbuf->field is equal to V4L2_FIELD_ANY, then the
> driver can choose to force it to V4L2_FIELD_NONE. Furthermore, it's coded in the
> same way in vim2m.c (vim2m_buf_prepare).
> Finally, if I remove these 2 lines, I've got the following error with the
> v4l2-compliance:
> Streaming ioctls:
> VIDIOC_G_INPUT returned -1 (Inappropriate ioctl for device)
> VIDIOC_ENUMINPUT returned -1 (Inappropriate ioctl for device)
> test read/write: OK (Not Supported)
> VIDIOC_QUERYCAP returned 0 (Success)
> [snip]
> VIDIOC_QUERYBUF returned 0 (Success)
> VIDIOC_QBUF returned -1 (Invalid argument)
> fail:
> /local/home/frq08988/views/opensdk-2.1.4.1/sources/v4l-utils/utils/v4l2-compliance/v4l2-test-buffers.cpp(773):
> buf.qbuf(node)
> fail:
> /local/home/frq08988/views/opensdk-2.1.4.1/sources/v4l-utils/utils/v4l2-compliance/v4l2-test-buffers.cpp(971):
> setupM2M(node, m2m_q)
> test MMAP: FAIL
>
> Don't you think that I could keep these two lines?
Keep it for now, I dug into this a bit further and it is really a workaround for
poorly written applications that can't be bothered to set the field value to a
proper value. I think the documentation needs to be updated for this.
I might change my mind, though :-)
Regards,
Hans
--
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
