lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250324155159.GH5113@pendragon.ideasonboard.com>
Date: Mon, 24 Mar 2025 17:51:59 +0200
From: Laurent Pinchart <laurent.pinchart@...asonboard.com>
To: niklas soderlund <niklas.soderlund@...natech.se>
Cc: Jacopo Mondi <jacopo.mondi@...asonboard.com>,
	Jacopo Mondi <jacopo.mondi+renesas@...asonboard.com>,
	Kieran Bingham <kieran.bingham+renesas@...asonboard.com>,
	linux-kernel@...r.kernel.org, linux-media@...r.kernel.org,
	linux-renesas-soc@...r.kernel.org
Subject: Re: [PATCH v6 7/7] media: vsp1: pipe: Add RAW Bayer formats mapping

On Mon, Mar 24, 2025 at 09:35:13AM +0100, Niklas Söderlund wrote:
> On 2025-03-24 09:27:56 +0100, Jacopo Mondi wrote:
> > On Fri, Mar 21, 2025 at 11:56:34PM +0200, Laurent Pinchart wrote:
> > > On Fri, Mar 21, 2025 at 04:45:39PM +0100, Jacopo Mondi wrote:
> > > > Add formats definition for RAW Bayer formats in vsp1_pipe.c.
> > > >
> > > > 8-bits RAW Bayer pixel formats map on VSP format RGB332.
> > >
> > > s/map on/map to/
> > >
> > > > 10, 12 and 16 bits RAW Bayer pixel formats map on RGB565 insted.
> > > >
> > > > Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@...natech.se>
> > > > Tested-by: Niklas Söderlund <niklas.soderlund+renesas@...natech.se>
> > > > Signed-off-by: Jacopo Mondi <jacopo.mondi+renesas@...asonboard.com>
> > > > ---
> > > > v3->v4:
> > > > - Fix SWAP bits for RAW 10, 12 and 16
> > > > ---
> > > >  drivers/media/platform/renesas/vsp1/vsp1_pipe.c | 72 ++++++++++++++++++++++++-
> > > >  1 file changed, 71 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/media/platform/renesas/vsp1/vsp1_pipe.c b/drivers/media/platform/renesas/vsp1/vsp1_pipe.c
> > > > index 8e9be3ec1b4d..a51061738edc 100644
> > > > --- a/drivers/media/platform/renesas/vsp1/vsp1_pipe.c
> > > > +++ b/drivers/media/platform/renesas/vsp1/vsp1_pipe.c
> > > > @@ -30,10 +30,80 @@
> > > >   */
> > > >
> > > >  static const struct vsp1_format_info vsp1_video_formats[] = {
> > > > -	{ V4L2_PIX_FMT_RGB332, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > > +	/* Raw Bayer 8-bit: Maps on RGB332 */
> > > > +	{ V4L2_PIX_FMT_SBGGR8, MEDIA_BUS_FMT_Y8_1X8,
> > > > +	  VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > > +	  1, { 8, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGBRG8, MEDIA_BUS_FMT_Y8_1X8,
> > > > +	  VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > > +	  1, { 8, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGRBG8, MEDIA_BUS_FMT_Y8_1X8,
> > > > +	  VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > > +	  1, { 8, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SRGGB8, MEDIA_BUS_FMT_Y8_1X8,
> > > >  	  VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > >  	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > >  	  1, { 8, 0, 0 }, false, false, 1, 1, false },
> > >
> > > Similarly to the media bus codes, could we have a single entry, using
> > > V4L2_PIX_FMT_GREY ? Same below with V4L2_PIX_FMT_Y10, V4L2_PIX_FMT_Y12
> > > and V4L2_PIX_FMT_Y16.
> > 
> > mmm, the SRGB mbus codes come from the R-Car ISP input image format.
> > I understand these are multiple identical entries, but having
> > somewhere a translation from SRGB->Y formats just to have fewer
> > entries here it feels a bit of an hack

handling the CFA pattern in the media bus code and pixel format was a
historical mistake that we're slowly addressing for new drivers. Sakari
has for instance posted patches to use the Y formats only on sensor
subdevs that will follow the new raw camera sensor subdev model. I think
we should use Y formats in the VSP and ISP drivers already.

> > > This would still duplicate entries, as V4L2_PIX_FMT_Y1[026] are
> > > essentially treated the same, and they are identical to
> > > V4L2_PIX_FMT_RGB565. We could ask the ISP driver to use
> > > V4L2_PIX_FMT_RGB565 (and V4L2_PIX_FMT_RGB332 for 8-bit raw) when
> > > configuring the VSPX, but that's a bit of a hack.
> > 
> > Indeed, but I don't think 3 "duplicated" entries are any bad, if
> > that's how the HW work.
> > 
> > > Another option would be to handle the translation in
> > > vsp1_vspx_rwpf_set_subdev_fmt(). I would still in that case only expect
> > > the V4L2_PIX_FMT_GREY and V4L2_PIX_FMT_Y* 4CCs from the ISP driver. This
> > 
> > Do you expect the ISP driver to translate SRGB to Y formats ?

I actually expect the ISP driver to expose Y formats only to userspace,
and get the CFA pattern from the ISP parameters buffers.

> > > patch could then be dropped.
> > 
> > So are you suggesting to translate in the ISP driver
> > 
> >         SRGB8 -> RGB332
> > 
> >         SRGB10/12/16 -> RGB565

No quite, the ISP driver would give us Y formats, and the translation to
RGB332 and RGB565 would be handled within the VSP1 driver, in the
vsp1_vspx_rwpf_set_subdev_fmt() function.

> > Niklas, what do you think ?
> 
> I would rather keep the true formats in the API between the VSP and ISP, 
> that is keep it as is. If really needed maybe a translation in the VSP 
> driver prior to querying vsp1_video_formats[] could be added? But this 
> driver is complex enough as-is :-)
> 
> > > What's your preference ?
> > >
> > > > +
> > > > +	/* Raw Bayer 10/12/16-bit: Maps on RGB565 */
> > > > +	{ V4L2_PIX_FMT_SBGGR10, MEDIA_BUS_FMT_Y10_1X10,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 10, 0, 0 }, false, false, 1, 1, false },
> > >
> > > The bpp values are used to calculate memory offsets. Unless I'm
> > > mistaken, you should use 16 here, not 10.
> > 
> > I'm rounding up in the vspx driver. However it is true these formats
> > are sampled in 16bpp chunks, so I can use 16 here.
> > 
> > > > +	{ V4L2_PIX_FMT_SGBRG10, MEDIA_BUS_FMT_Y10_1X10,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 10, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGRBG10, MEDIA_BUS_FMT_Y10_1X10,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 10, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SRGGB10, MEDIA_BUS_FMT_Y10_1X10,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 10, 0, 0 }, false, false, 1, 1, false },
> > > > +
> > > > +	{ V4L2_PIX_FMT_SBGGR12, MEDIA_BUS_FMT_Y12_1X12,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 12, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGBRG12, MEDIA_BUS_FMT_Y12_1X12,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 12, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGRBG12, MEDIA_BUS_FMT_Y12_1X12,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 12, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SRGGB12, MEDIA_BUS_FMT_Y12_1X12,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 12, 0, 0 }, false, false, 1, 1, false },
> > > > +
> > > > +	{ V4L2_PIX_FMT_SBGGR16, MEDIA_BUS_FMT_Y16_1X16,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 16, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGBRG16, MEDIA_BUS_FMT_Y16_1X16,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 16, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SGRBG16, MEDIA_BUS_FMT_Y16_1X16,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 16, 0, 0 }, false, false, 1, 1, false },
> > > > +	{ V4L2_PIX_FMT_SRGGB16, MEDIA_BUS_FMT_Y16_1X16,
> > > > +	  VI6_FMT_RGB_565, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS,
> > > > +	  1, { 16, 0, 0 }, false, false, 1, 1, false },
> > > > +
> > > > +	{ V4L2_PIX_FMT_RGB332, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > > +	  VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > > +	  VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> > > > +	  1, { 10, 0, 0 }, false, false, 1, 1, false },
> > >
> > > This doesn't seem right, the patch is changing the V4L2_PIX_FMT_RGB332.
> > 
> > If I'm not mistaken V4L2_PIX_FMT_RGB332 was
> > 
> >         { V4L2_PIX_FMT_RGB332, MEDIA_BUS_FMT_ARGB8888_1X32,
> >           VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> >           VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> >           1, { 8, 0, 0 }, false, false, 1, 1, false }
> > 
> > and is now
> > 
> >         { V4L2_PIX_FMT_RGB332, MEDIA_BUS_FMT_ARGB8888_1X32,
> >           VI6_FMT_RGB_332, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> >           VI6_RPF_DSWAP_P_WDS | VI6_RPF_DSWAP_P_BTS,
> >           1, { 10, 0, 0 }, false, false, 1, 1, false },
> > 
> > Seems like I messed up the bpp
> > 
> > With that fixed the diff looks saner. Thanks for spotting.
> > 
> > > >  	{ V4L2_PIX_FMT_ARGB444, MEDIA_BUS_FMT_ARGB8888_1X32,
> > > >  	  VI6_FMT_ARGB_4444, VI6_RPF_DSWAP_P_LLS | VI6_RPF_DSWAP_P_LWS |
> > > >  	  VI6_RPF_DSWAP_P_WDS,

-- 
Regards,

Laurent Pinchart

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ