Re: [PATCH] media: rcar-csi2: Add checking to rcsi2_start_receiver().

["Niklas Söderlund" <niklas.soderlund@ragnatech.se>]

Hi Nadezda,

Thanks for your work.

On 2021-08-11 16:31:42 +0300, Nadezda Lutovinova wrote:
> If rcsi2_code_to_fmt() return NULL,
> then null pointer dereference occurs in the next cycle.
> The patch adds checking if format is NULL.
> 
> Found by Linux Driver Verification project (linuxtesting.org).

Please drop the '.' at the end of the patch subject. Also the commit 
message could be better line wrapped.

> 
> Signed-off-by: Nadezda Lutovinova <lutovinova@ispras.ru>
> ---
>  drivers/media/platform/rcar-vin/rcar-csi2.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/drivers/media/platform/rcar-vin/rcar-csi2.c b/drivers/media/platform/rcar-vin/rcar-csi2.c
> index e28eff039688..55bb584d2a13 100644
> --- a/drivers/media/platform/rcar-vin/rcar-csi2.c
> +++ b/drivers/media/platform/rcar-vin/rcar-csi2.c
> @@ -553,6 +553,12 @@ static int rcsi2_start_receiver(struct rcar_csi2 *priv)
>  
>  	/* Code is validated in set_fmt. */
>  	format = rcsi2_code_to_fmt(priv->mf.code);
> +	if (!format) {

This can never happen ;-)

The only place priv->mf.code is set (after probe) is in  
rcsi2_set_pad_format() and there it explicitly checks if 
rcsi2_code_to_fmt() returns NULl and if so sets it to something that 
guarantees it will not.

Think of it as the verification is done at format configuration time so 
we don't have to have check it at start time. The reason for this is 
that we can't do much about a failure here other then fail the start 
while at configure time we can try to correct it and inform the user of 
the change.

That being said, I'm not oppose to fail the start here if we ever do 
introduce a bug here where rcsi2_code_to_fmt() would return NULL here.  
But I would like to drop the dev_err() here and just return -EINVAL.

I would mention in the commit message that this protects from future 
bugs.

> +		dev_err(priv->dev,
> +			"Incorrect mbus frame format code %u\n",
> +			priv->mf.code);
> +		return -EINVAL;
> +	}
>  
>  	/*
>  	 * Enable all supported CSI-2 channels with virtual channel and
> -- 
> 2.17.1
> 

-- 
Regards,
Niklas Söderlund