Hi Guennadi,
On Thursday 26 January 2012 16:21:00 Guennadi Liakhovetski wrote:
> Hi Laurent
>
> Thanks for the patches. This one looks good mostly, a couple of questions
> though:
>
> On Wed, 25 Jan 2012, Laurent Pinchart wrote:
> > Instead of computing the buffer size manually in the videobuf queue
> > setup and buffer prepare callbacks, use the previously negotiated
> > soc_camera_device::sizeimage value.
> >
> > Signed-off-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > ---
> >
> > drivers/media/video/atmel-isi.c | 17 +++--------------
> > drivers/media/video/mx1_camera.c | 14 ++------------
> > drivers/media/video/mx2_camera.c | 14 ++------------
> > drivers/media/video/mx3_camera.c | 20 +++++++++-----------
> > drivers/media/video/omap1_camera.c | 14 ++------------
> > drivers/media/video/pxa_camera.c | 14 ++------------
> > drivers/media/video/sh_mobile_ceu_camera.c | 25
> > +++++++++---------------- 7 files changed, 29 insertions(+), 89
> > deletions(-)
>
> [snip]
>
> > diff --git a/drivers/media/video/mx2_camera.c
> > b/drivers/media/video/mx2_camera.c index a803d9e..e9b228d 100644
> > --- a/drivers/media/video/mx2_camera.c
> > +++ b/drivers/media/video/mx2_camera.c
> > @@ -433,15 +433,10 @@ static int mx2_videobuf_setup(struct videobuf_queue
> > *vq, unsigned int *count,
> >
> > unsigned int *size)
> >
> > {
> >
> > struct soc_camera_device *icd = vq->priv_data;
> >
> > - int bytes_per_line = soc_mbus_bytes_per_line(icd->user_width,
> > - icd->current_fmt->host_fmt);
> >
> > dev_dbg(icd->parent, "count=%d, size=%d\n", *count, *size);
> >
> > - if (bytes_per_line < 0)
> > - return bytes_per_line;
> > -
> > - *size = bytes_per_line * icd->user_height;
> > + *size = icd->sizeimage;
> >
> > if (0 == *count)
> >
> > *count = 32;
>
> I think, there is a bug in mx2_camera_try_fmt(), which also would affect
> these your calculations. On i.MX25 they restrict the image size based on
> maximum supported number of bytes. In such a case they recalculate
> .bytesperline, but they fail to update .sizeimage. The fix seems to be
> trivial, please add it, then your calculations should be fine.
Indeed. I'll add a patch for that, and I'll also modify mx2_camera_try_fmt()
to use the new soc_mbus_image_size() function.
> > @@ -476,16 +471,11 @@ static int mx2_videobuf_prepare(struct
> > videobuf_queue *vq,
> >
> > {
> >
> > struct soc_camera_device *icd = vq->priv_data;
> > struct mx2_buffer *buf = container_of(vb, struct mx2_buffer, vb);
> >
> > - int bytes_per_line = soc_mbus_bytes_per_line(icd->user_width,
> > - icd->current_fmt->host_fmt);
> >
> > int ret = 0;
> >
> > dev_dbg(icd->parent, "%s (vb=0x%p) 0x%08lx %d\n", __func__,
> >
> > vb, vb->baddr, vb->bsize);
> >
> > - if (bytes_per_line < 0)
> > - return bytes_per_line;
> > -
> >
> > #ifdef DEBUG
> >
> > /*
> >
> > * This can be useful if you want to see if we actually fill
> >
> > @@ -505,7 +495,7 @@ static int mx2_videobuf_prepare(struct videobuf_queue
> > *vq,
> >
> > vb->state = VIDEOBUF_NEEDS_INIT;
> >
> > }
> >
> > - vb->size = bytes_per_line * vb->height;
> > + vb->size = icd->sizeimage;
> >
> > if (vb->baddr && vb->bsize < vb->size) {
> >
> > ret = -EINVAL;
> > goto out;
> >
> > diff --git a/drivers/media/video/mx3_camera.c
> > b/drivers/media/video/mx3_camera.c index f96f92f..da45a89 100644
> > --- a/drivers/media/video/mx3_camera.c
> > +++ b/drivers/media/video/mx3_camera.c
> > @@ -199,8 +199,6 @@ static int mx3_videobuf_setup(struct vb2_queue *vq,
> >
> > struct soc_camera_device *icd = soc_camera_from_vb2q(vq);
> > struct soc_camera_host *ici = to_soc_camera_host(icd->parent);
> > struct mx3_camera_dev *mx3_cam = ici->priv;
> >
> > - int bytes_per_line;
> > - unsigned int height;
> >
> > if (!mx3_cam->idmac_channel[0])
> >
> > return -EINVAL;
> >
> > @@ -208,21 +206,21 @@ static int mx3_videobuf_setup(struct vb2_queue *vq,
> >
> > if (fmt) {
> >
> > const struct soc_camera_format_xlate *xlate =
> > soc_camera_xlate_by_fourcc(icd,
> >
> > fmt->fmt.pix.pixelformat);
> >
> > + int bytes_per_line;
> > +
> >
> > if (!xlate)
> >
> > return -EINVAL;
> >
> > +
> >
> > bytes_per_line = soc_mbus_bytes_per_line(fmt->fmt.pix.width,
> >
> > xlate->host_fmt);
> >
> > - height = fmt->fmt.pix.height;
> > + if (bytes_per_line < 0)
> > + return bytes_per_line;
> > +
> > + sizes[0] = bytes_per_line * fmt->fmt.pix.height;
> >
> > } else {
> >
> > /* Called from VIDIOC_REQBUFS or in compatibility mode */
> >
> > - bytes_per_line = soc_mbus_bytes_per_line(icd->user_width,
> > - icd->current_fmt->host_fmt);
> > - height = icd->user_height;
> > + sizes[0] = icd->sizeimage;
> >
> > }
> >
> > - if (bytes_per_line < 0)
> > - return bytes_per_line;
> > -
> > - sizes[0] = bytes_per_line * height;
> >
> > alloc_ctxs[0] = mx3_cam->alloc_ctx;
> >
> > @@ -274,7 +272,7 @@ static void mx3_videobuf_queue(struct vb2_buffer *vb)
> >
> > BUG_ON(bytes_per_line <= 0);
> >
> > - new_size = bytes_per_line * icd->user_height;
> > + new_size = icd->sizeimage;
>
> Don't you think, you could eliminate bytes_per_line too and just use
> icd->bytesperline here too?
Sure. That's why the next patch does just that :-)
> > if (vb2_plane_size(vb, 0) < new_size) {
> >
> > dev_err(icd->parent, "Buffer #%d too small (%lu < %zu)\n",
>
> [snip]
>
> > diff --git a/drivers/media/video/sh_mobile_ceu_camera.c
> > b/drivers/media/video/sh_mobile_ceu_camera.c index c51decf..f4eb9e1
> > 100644
> > --- a/drivers/media/video/sh_mobile_ceu_camera.c
> > +++ b/drivers/media/video/sh_mobile_ceu_camera.c
> > @@ -206,27 +206,25 @@ static int sh_mobile_ceu_videobuf_setup(struct
> > vb2_queue *vq,
> >
> > struct soc_camera_device *icd = container_of(vq, struct
> > soc_camera_device, vb2_vidq); struct soc_camera_host *ici =
> > to_soc_camera_host(icd->parent); struct sh_mobile_ceu_dev *pcdev =
> > ici->priv;
> >
> > - int bytes_per_line;
> > - unsigned int height;
> >
> > if (fmt) {
> >
> > const struct soc_camera_format_xlate *xlate =
> > soc_camera_xlate_by_fourcc(icd,
> >
> > fmt->fmt.pix.pixelformat);
> >
> > + int bytes_per_line;
> > +
> >
> > if (!xlate)
> >
> > return -EINVAL;
> >
> > +
> >
> > bytes_per_line = soc_mbus_bytes_per_line(fmt->fmt.pix.width,
> >
> > xlate->host_fmt);
> >
> > - height = fmt->fmt.pix.height;
> > + if (bytes_per_line < 0)
> > + return bytes_per_line;
> > +
> > + sizes[0] = bytes_per_line * fmt->fmt.pix.height;
> >
> > } else {
> >
> > /* Called from VIDIOC_REQBUFS or in compatibility mode */
> >
> > - bytes_per_line = soc_mbus_bytes_per_line(icd->user_width,
> > - icd->current_fmt->host_fmt);
> > - height = icd->user_height;
> > + sizes[0] = icd->sizeimage;
> >
> > }
> >
> > - if (bytes_per_line < 0)
> > - return bytes_per_line;
> > -
> > - sizes[0] = bytes_per_line * height;
> >
> > alloc_ctxs[0] = pcdev->alloc_ctx;
> >
> > @@ -373,13 +371,8 @@ static void sh_mobile_ceu_videobuf_queue(struct
> > vb2_buffer *vb)
> >
> > struct sh_mobile_ceu_dev *pcdev = ici->priv;
> > struct sh_mobile_ceu_buffer *buf = to_ceu_vb(vb);
> > unsigned long size;
> >
> > - int bytes_per_line = soc_mbus_bytes_per_line(icd->user_width,
> > - icd->current_fmt->host_fmt);
> > -
> > - if (bytes_per_line < 0)
> > - goto error;
> >
> > - size = icd->user_height * bytes_per_line;
> > + size = icd->sizeimage;
> >
> > if (vb2_plane_size(vb, 0) < size) {
> >
> > dev_err(icd->parent, "Buffer #%d too small (%lu < %lu)\n",
>
> Looks like sh_mobile_ceu_set_rect() can also be simplified, since there
> bytes_per_line is calculated for data-fetch mode, for which the
> ->bytesperline can also be used?
Is sh_mobile_ceu_set_rect() guaranteed to be called after try_fmt(), with the
->bytesperline value set to the correct value for the current format ?
--
Regards,
Laurent Pinchart
--
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
