Hi,

On Fri, 24 Sep 2021 13:46:15 +0200
Philipp Zabel [off-list ref] wrote:

Hi Dorota,

On Fri, Sep 24, 2021 at 12:06:31PM +0200, Dorota Czaplejewicz wrote:

quoted

Change rounding to the minimal burst size from 2^n to n.

This fixes images with sizes that are a multiple of 8 pixels.  

Could you elaborate on what is currently wrong with images that are a
multiple of 8 pixels wide? Or are you rather trying to add support
for images that are not a multiple of 8 pixels wide?

quoted

See section 13.7.6.13 CSI Image Parameter Register of the
i.MX 8M Quad Applications Processors Reference Manual.

Fixes: 451a7b7815d0b ("media: imx: lift CSI and PRP ENC/VF width
alignment restriction")
Signed-off-by: Dorota Czaplejewicz <redacted>
---
Hi,

I tested this patch on the Librem 5 with the main camera.

--Dorota
 drivers/staging/media/imx/imx-media-utils.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/staging/media/imx/imx-media-utils.c b/drivers/staging/media/imx/imx-media-utils.c
index 5128915a5d6f..a2d8fab32a39 100644
--- a/drivers/staging/media/imx/imx-media-utils.c
+++ b/drivers/staging/media/imx/imx-media-utils.c

@@ -545,13 +545,13 @@ int imx_media_mbus_fmt_to_pix_fmt(struct v4l2_pix_format *pix,
 	}
 
 	/* Round up width for minimum burst size */
-	width = round_up(mbus->width, 8);
+	width = round_up(mbus->width, 3);  

That is not a valid use of the round_up() macro anymore.
The second parameter must be a power of 2.

regards
Philipp

Thanks for catching that. It turns out I was fixing an image that was a multiple of 4, but not a multiple of 8.

The problem that occurred has to do with an image of size 1052x780, which covers the entire sensor in use at 1/4 scale.

The sensor was configured to emit 1052*780 8-bit pixels, whereas the CSI portion here rounded up the image, expecting 1056*780. As a result, the userspace received a buffer sized according to the latter, each frame padded with some pixels from the next. As a result, each frame was composed of two or more, with a broken TV effect.

The data sheet says that the width parameter "should" be a multiple of 8:

If the input data from the sensor is 8-bit/pixel format, the IMAGE_WIDTH should be a multiple of 8 pixels.

Nevertheless, I double-tested this, and changing the relevant rounding from "8" to "4" works flawlessly here. Lower divisors were tested previously and did not work.

Does it make sense to accept 4 as a valid divisor, given that rounding is likely not going to help but result in an invalid image anyway (as it did for me)?

Cheers,
Dorota