Re: [PATCH] media: sunxi: sun6i_mipi_csi2.c/sun8i_a83t_mipi_csi2.c: clarify error handling

[Paul Kocialkowski <paul.kocialkowski@bootlin.com>]

[-- Attachment #1: Type: text/plain, Size: 5997 bytes --]

Hans,

On Tue 19 Jul 22, 12:04, Hans Verkuil wrote:
> On 7/19/22 11:54, Paul Kocialkowski wrote:
> > Hi Hans,
> > 
> > On Mon 18 Jul 22, 18:08, Hans Verkuil wrote:
> >> Both sun6i_mipi_csi2.c and sun8i_a83t_mipi_csi2.c have the same issue:
> >> the comment before the ret = 0 assignment is incorrect, drop it and
> >> always assign the result of the v4l2_subdev_call(..., 0) to ret.
> > 
> > Not sure what is incorrect about it. I agree that it's good to return the
> > return code of v4l2_subdev_call instead of always 0 on the !on path though.
> > 
> >> In the disable label check for !on and set ret to 0 in that case.
> >>
> >> This fixes two smatch warnings:
> >>
> >> drivers/media/platform/sunxi/sun6i-mipi-csi2/sun6i_mipi_csi2.c:193 sun6i_mipi_csi2_s_stream() warn: missing error code 'ret'
> > 
> > I don't think this is legit: the initialization of ret is specifically there
> > to avoid an undefined ret there.
> 
> I believe so as well, but I need to get rid of the smatch warning.

Yes fair enough, I was just a bit confused by the commit message.

> >> drivers/media/platform/sunxi/sun8i-a83t-mipi-csi2/sun8i_a83t_mipi_csi2.c:225 sun8i_a83t_mipi_csi2_s_stream() warn: missing error code 'ret'
> >>
> >> Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
> >> ---
> >>  .../media/platform/sunxi/sun6i-mipi-csi2/sun6i_mipi_csi2.c | 7 ++++---
> >>  .../sunxi/sun8i-a83t-mipi-csi2/sun8i_a83t_mipi_csi2.c      | 7 ++++---
> >>  2 files changed, 8 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/drivers/media/platform/sunxi/sun6i-mipi-csi2/sun6i_mipi_csi2.c b/drivers/media/platform/sunxi/sun6i-mipi-csi2/sun6i_mipi_csi2.c
> >> index 31e12f1506cb..a4e3f9a6b2ff 100644
> >> --- a/drivers/media/platform/sunxi/sun6i-mipi-csi2/sun6i_mipi_csi2.c
> >> +++ b/drivers/media/platform/sunxi/sun6i-mipi-csi2/sun6i_mipi_csi2.c
> >> @@ -182,14 +182,13 @@ static int sun6i_mipi_csi2_s_stream(struct v4l2_subdev *subdev, int on)
> >>  	unsigned int lanes_count =
> >>  		csi2_dev->bridge.endpoint.bus.mipi_csi2.num_data_lanes;
> >>  	unsigned long pixel_rate;
> >> -	/* Initialize to 0 to use both in disable label (ret != 0) and off. */
> >> -	int ret = 0;
> >> +	int ret;
> >>  
> >>  	if (!source_subdev)
> >>  		return -ENODEV;
> >>  
> >>  	if (!on) {
> >> -		v4l2_subdev_call(source_subdev, video, s_stream, 0);
> >> +		ret = v4l2_subdev_call(source_subdev, video, s_stream, 0);
> >>  		goto disable;
> >>  	}
> >>  
> >> @@ -281,6 +280,8 @@ static int sun6i_mipi_csi2_s_stream(struct v4l2_subdev *subdev, int on)
> >>  	return 0;
> >>  
> >>  disable:
> >> +	if (!on)
> >> +		ret = 0;
> > 
> > Then this will always overwrite the ret value from v4l2_subdev_call with the
> > !on disable path. Looks like this can be removed.
> 
> Returning an error when stopping streaming is kind of pointless, since
> what are you going to do with that error?

Yeah that's true and it's probably safer to pretend that things went well
just in case returning != 0 somehow ends up halting the turn-off sequence.

> I assumed that's why the original code did this, and I didn't want to
> change the behavior.

Yeah I think it was my initial intent to discard the v4l2_subdev_call return
code for this reason.

> If you think it is better to just return the error, then that can be a
> patch on top.

Nah I'm fine with returning zero after all. But instead of assigning ret from
v4l2_subdev_call and then overwriting it conditionally in the disable label,
you could just as well set ret = 0; before goto disable; which seems less
confusing to me. I think this should also satisfy the smatch warning.

Otherwise it would be nice to at least have a comment about it or to reflect
the ret overwrite in the commit message (and maybe add a blank like after the
conditional block while at it).

> BTW, I think it would have been better to have separate streamon and streamoff
> ops instead of a single s_stream. Starting streaming would return an int, while
> stopping streaming would be a void function. In general when you stop, release,
> close, etc. something there is no point in an error code since there is nothing
> you can do with it.

Agreed, maybe this can be somehow helpful in sensor drivers where you just have
one common bit to start/halt the stream but for more complex subdevs it's not
very helpful to have a single op.

Cheers,

Paul

> Regards,
> 
> 	Hans
> 
> > 
> > What do you think?
> > 
> > Paul
> > 
> >>  	phy_power_off(dphy);
> >>  	sun6i_mipi_csi2_disable(csi2_dev);
> >>  
> >> diff --git a/drivers/media/platform/sunxi/sun8i-a83t-mipi-csi2/sun8i_a83t_mipi_csi2.c b/drivers/media/platform/sunxi/sun8i-a83t-mipi-csi2/sun8i_a83t_mipi_csi2.c
> >> index c82ede3ca0ff..d052ee77ef0a 100644
> >> --- a/drivers/media/platform/sunxi/sun8i-a83t-mipi-csi2/sun8i_a83t_mipi_csi2.c
> >> +++ b/drivers/media/platform/sunxi/sun8i-a83t-mipi-csi2/sun8i_a83t_mipi_csi2.c
> >> @@ -214,14 +214,13 @@ static int sun8i_a83t_mipi_csi2_s_stream(struct v4l2_subdev *subdev, int on)
> >>  	unsigned int lanes_count =
> >>  		csi2_dev->bridge.endpoint.bus.mipi_csi2.num_data_lanes;
> >>  	unsigned long pixel_rate;
> >> -	/* Initialize to 0 to use both in disable label (ret != 0) and off. */
> >> -	int ret = 0;
> >> +	int ret;
> >>  
> >>  	if (!source_subdev)
> >>  		return -ENODEV;
> >>  
> >>  	if (!on) {
> >> -		v4l2_subdev_call(source_subdev, video, s_stream, 0);
> >> +		ret = v4l2_subdev_call(source_subdev, video, s_stream, 0);
> >>  		goto disable;
> >>  	}
> >>  
> >> @@ -313,6 +312,8 @@ static int sun8i_a83t_mipi_csi2_s_stream(struct v4l2_subdev *subdev, int on)
> >>  	return 0;
> >>  
> >>  disable:
> >> +	if (!on)
> >> +		ret = 0;
> >>  	phy_power_off(dphy);
> >>  	sun8i_a83t_mipi_csi2_disable(csi2_dev);
> >>  
> >> -- 
> >> 2.35.1
> >>
> > 

-- 
Paul Kocialkowski, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]