media: rkisp1: Fix Bayer demosaicing bypass

[PATCH] media: rkisp1: Fix Bayer demosaicing bypass

Posted by Jacopo Mondi 1 week, 6 days ago

The RKISP1_CIF_ISP_DEMOSAIC_BYPASS bit, when set, bypasses the
demosaicing block on the RkISP1 ISP.

The current implementation however clears the bit when demosaicing
have to be bypassed and sets it when demosaicing has to be enabled,
effectively inverting the bypass bit handling logic.

Fix this by setting the bypass bit when disabling the demosaicing block,
and by clearing it instead when demosaicing has to be performed.

The issue never manifested itself as libcamera hasn't an algorithm
to control Bayer demosaicing bypass yet.

Fixes: 6c53a7b68c5d ("media: rkisp1: Implement extensible params support")
Cc: stable@vger.kernel.org
Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
---
media: rkisp1: Fix demosaicing bypass
---
 drivers/media/platform/rockchip/rkisp1/rkisp1-params.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
index 042b759eba62..496381962f1b 100644
--- a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
+++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
@@ -1854,8 +1854,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
 	const struct rkisp1_ext_params_bdm_config *bdm = &block->bdm;
 
 	if (bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_DISABLE) {
-		rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
-					RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
+		rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
+				      RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
 		return;
 	}
 
@@ -1863,8 +1863,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
 
 	if ((bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_ENABLE) &&
 	    !(params->enabled_blocks & BIT(bdm->header.type)))
-		rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
-				      RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
+		rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
+					RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
 }
 
 static void

---
base-commit: 27953c044974baf7e24dee3e9342fe0103dea80c
change-id: 20260911-imx8mp-demosaicing-bypass-bb284d9cca60

Best regards,
--  
Jacopo Mondi <jacopo.mondi@ideasonboard.com>

Re: [PATCH] media: rkisp1: Fix Bayer demosaicing bypass

Posted by Kieran Bingham 1 week, 6 days ago

Quoting Jacopo Mondi (2026-09-11 14:56:19)
> The RKISP1_CIF_ISP_DEMOSAIC_BYPASS bit, when set, bypasses the
> demosaicing block on the RkISP1 ISP.
> 
> The current implementation however clears the bit when demosaicing
> have to be bypassed and sets it when demosaicing has to be enabled,
> effectively inverting the bypass bit handling logic.

Ouch.

> Fix this by setting the bypass bit when disabling the demosaicing block,
> and by clearing it instead when demosaicing has to be performed.
> 
> The issue never manifested itself as libcamera hasn't an algorithm
> to control Bayer demosaicing bypass yet.
> 
> Fixes: 6c53a7b68c5d ("media: rkisp1: Implement extensible params support")
> Cc: stable@vger.kernel.org
> Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> ---
> media: rkisp1: Fix demosaicing bypass
> ---
>  drivers/media/platform/rockchip/rkisp1/rkisp1-params.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> index 042b759eba62..496381962f1b 100644
> --- a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> +++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> @@ -1854,8 +1854,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
>         const struct rkisp1_ext_params_bdm_config *bdm = &block->bdm;
>  
>         if (bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_DISABLE) {
> -               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> -                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> +               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> +                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
>                 return;
>         }
>  
> @@ -1863,8 +1863,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
>  
>         if ((bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_ENABLE) &&
>             !(params->enabled_blocks & BIT(bdm->header.type)))
> -               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> -                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> +               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> +                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);


I think this also opens us up to add mono formats as explicitly
supported by the ISP and potentially set the demosaic defaulting to off
in that instance?

Anyway, Looks sane to me in this order.

Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>

>  }
>  
>  static void
> 
> ---
> base-commit: 27953c044974baf7e24dee3e9342fe0103dea80c
> change-id: 20260911-imx8mp-demosaicing-bypass-bb284d9cca60
> 
> Best regards,
> --  
> Jacopo Mondi <jacopo.mondi@ideasonboard.com>
>

Re: [PATCH] media: rkisp1: Fix Bayer demosaicing bypass

Posted by Jacopo Mondi 1 week, 6 days ago

Hi Kieran

On Fri, Sep 11, 2026 at 03:27:50PM +0100, Kieran Bingham wrote:
> Quoting Jacopo Mondi (2026-09-11 14:56:19)
> > The RKISP1_CIF_ISP_DEMOSAIC_BYPASS bit, when set, bypasses the
> > demosaicing block on the RkISP1 ISP.
> >
> > The current implementation however clears the bit when demosaicing
> > have to be bypassed and sets it when demosaicing has to be enabled,
> > effectively inverting the bypass bit handling logic.
>
> Ouch.
>
> > Fix this by setting the bypass bit when disabling the demosaicing block,
> > and by clearing it instead when demosaicing has to be performed.
> >
> > The issue never manifested itself as libcamera hasn't an algorithm
> > to control Bayer demosaicing bypass yet.
> >
> > Fixes: 6c53a7b68c5d ("media: rkisp1: Implement extensible params support")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> > ---
> > media: rkisp1: Fix demosaicing bypass
> > ---
> >  drivers/media/platform/rockchip/rkisp1/rkisp1-params.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > index 042b759eba62..496381962f1b 100644
> > --- a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > +++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > @@ -1854,8 +1854,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
> >         const struct rkisp1_ext_params_bdm_config *bdm = &block->bdm;
> >
> >         if (bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_DISABLE) {
> > -               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > -                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > +               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > +                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> >                 return;
> >         }
> >
> > @@ -1863,8 +1863,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
> >
> >         if ((bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_ENABLE) &&
> >             !(params->enabled_blocks & BIT(bdm->header.type)))
> > -               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > -                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > +               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > +                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
>
>
> I think this also opens us up to add mono formats as explicitly
> supported by the ISP and potentially set the demosaic defaulting to off
> in that instance?

mmm, I think userspace is in a better position to decide when to
bypass debayer instead of relying on auto-configuration of the ISP ?

I guess we'll discuss this when support for luma-only formats will be
added to the driver

>
> Anyway, Looks sane to me in this order.
>
> Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>

thanks

>
> >  }
> >
> >  static void
> >
> > ---
> > base-commit: 27953c044974baf7e24dee3e9342fe0103dea80c
> > change-id: 20260911-imx8mp-demosaicing-bypass-bb284d9cca60
> >
> > Best regards,
> > --
> > Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> >

Re: [PATCH] media: rkisp1: Fix Bayer demosaicing bypass

Posted by Laurent Pinchart 1 week, 6 days ago

On Fri, Sep 11, 2026 at 04:43:32PM +0200, Jacopo Mondi wrote:
> On Fri, Sep 11, 2026 at 03:27:50PM +0100, Kieran Bingham wrote:
> > Quoting Jacopo Mondi (2026-09-11 14:56:19)
> > > The RKISP1_CIF_ISP_DEMOSAIC_BYPASS bit, when set, bypasses the
> > > demosaicing block on the RkISP1 ISP.
> > >
> > > The current implementation however clears the bit when demosaicing
> > > have to be bypassed and sets it when demosaicing has to be enabled,
> > > effectively inverting the bypass bit handling logic.
> >
> > Ouch.
> >
> > > Fix this by setting the bypass bit when disabling the demosaicing block,
> > > and by clearing it instead when demosaicing has to be performed.
> > >
> > > The issue never manifested itself as libcamera hasn't an algorithm
> > > to control Bayer demosaicing bypass yet.
> > >
> > > Fixes: 6c53a7b68c5d ("media: rkisp1: Implement extensible params support")
> > > Cc: stable@vger.kernel.org
> > > Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> > > ---
> > > media: rkisp1: Fix demosaicing bypass
> > > ---
> > >  drivers/media/platform/rockchip/rkisp1/rkisp1-params.c | 8 ++++----
> > >  1 file changed, 4 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > > index 042b759eba62..496381962f1b 100644
> > > --- a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > > +++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > > @@ -1854,8 +1854,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
> > >         const struct rkisp1_ext_params_bdm_config *bdm = &block->bdm;
> > >
> > >         if (bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_DISABLE) {
> > > -               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > -                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > > +               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > +                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > >                 return;
> > >         }
> > >
> > > @@ -1863,8 +1863,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
> > >
> > >         if ((bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_ENABLE) &&
> > >             !(params->enabled_blocks & BIT(bdm->header.type)))
> > > -               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > -                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > > +               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > +                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> >
> > I think this also opens us up to add mono formats as explicitly
> > supported by the ISP and potentially set the demosaic defaulting to off
> > in that instance?
> 
> mmm, I think userspace is in a better position to decide when to
> bypass debayer instead of relying on auto-configuration of the ISP ?
> 
> I guess we'll discuss this when support for luma-only formats will be
> added to the driver

While I understand why auto-configuration is tempting, I've found that
more often than not it makes the life of both the driver and userspace
more difficult. Look for instance at the colourspace handling code in
rkisp1.

> > Anyway, Looks sane to me in this order.
> >
> > Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

> thanks
> 
> > >  }
> > >
> > >  static void
> > >
> > > ---
> > > base-commit: 27953c044974baf7e24dee3e9342fe0103dea80c
> > > change-id: 20260911-imx8mp-demosaicing-bypass-bb284d9cca60

-- 
Regards,

Laurent Pinchart

Re: [PATCH] media: rkisp1: Fix Bayer demosaicing bypass

Posted by Laurent Pinchart 1 week, 6 days ago

On Fri, Sep 11, 2026 at 08:35:10PM +0300, Laurent Pinchart wrote:
> On Fri, Sep 11, 2026 at 04:43:32PM +0200, Jacopo Mondi wrote:
> > On Fri, Sep 11, 2026 at 03:27:50PM +0100, Kieran Bingham wrote:
> > > Quoting Jacopo Mondi (2026-09-11 14:56:19)
> > > > The RKISP1_CIF_ISP_DEMOSAIC_BYPASS bit, when set, bypasses the
> > > > demosaicing block on the RkISP1 ISP.
> > > >
> > > > The current implementation however clears the bit when demosaicing
> > > > have to be bypassed and sets it when demosaicing has to be enabled,
> > > > effectively inverting the bypass bit handling logic.
> > >
> > > Ouch.
> > >
> > > > Fix this by setting the bypass bit when disabling the demosaicing block,
> > > > and by clearing it instead when demosaicing has to be performed.
> > > >
> > > > The issue never manifested itself as libcamera hasn't an algorithm

s/hasn't/doesn't have/

I'll fix it when applying.

> > > > to control Bayer demosaicing bypass yet.
> > > >
> > > > Fixes: 6c53a7b68c5d ("media: rkisp1: Implement extensible params support")
> > > > Cc: stable@vger.kernel.org
> > > > Signed-off-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
> > > > ---
> > > > media: rkisp1: Fix demosaicing bypass
> > > > ---
> > > >  drivers/media/platform/rockchip/rkisp1/rkisp1-params.c | 8 ++++----
> > > >  1 file changed, 4 insertions(+), 4 deletions(-)
> > > >
> > > > diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > > > index 042b759eba62..496381962f1b 100644
> > > > --- a/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > > > +++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-params.c
> > > > @@ -1854,8 +1854,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
> > > >         const struct rkisp1_ext_params_bdm_config *bdm = &block->bdm;
> > > >
> > > >         if (bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_DISABLE) {
> > > > -               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > > -                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > > > +               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > > +                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > > >                 return;
> > > >         }
> > > >
> > > > @@ -1863,8 +1863,8 @@ rkisp1_ext_params_bdm(struct rkisp1_params *params,
> > > >
> > > >         if ((bdm->header.flags & RKISP1_EXT_PARAMS_FL_BLOCK_ENABLE) &&
> > > >             !(params->enabled_blocks & BIT(bdm->header.type)))
> > > > -               rkisp1_param_set_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > > -                                     RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > > > +               rkisp1_param_clear_bits(params, RKISP1_CIF_ISP_DEMOSAIC,
> > > > +                                       RKISP1_CIF_ISP_DEMOSAIC_BYPASS);
> > >
> > > I think this also opens us up to add mono formats as explicitly
> > > supported by the ISP and potentially set the demosaic defaulting to off
> > > in that instance?
> > 
> > mmm, I think userspace is in a better position to decide when to
> > bypass debayer instead of relying on auto-configuration of the ISP ?
> > 
> > I guess we'll discuss this when support for luma-only formats will be
> > added to the driver
> 
> While I understand why auto-configuration is tempting, I've found that
> more often than not it makes the life of both the driver and userspace
> more difficult. Look for instance at the colourspace handling code in
> rkisp1.
> 
> > > Anyway, Looks sane to me in this order.
> > >
> > > Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
> 
> Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> 
> > thanks
> > 
> > > >  }
> > > >
> > > >  static void
> > > >
> > > > ---
> > > > base-commit: 27953c044974baf7e24dee3e9342fe0103dea80c
> > > > change-id: 20260911-imx8mp-demosaicing-bypass-bb284d9cca60

-- 
Regards,

Laurent Pinchart