Re: [PATCH] pwm: mtk_disp: fix disp_pwm coverity issue

[AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>]

Il 06/07/23 14:29, Alexandre Mergnat ha scritto:
> 
> 
> On 06/07/2023 12:04, Shuijing Li wrote:
>> There is a coverity issue in the original mtk_disp_pwm_get_state()
>> function. In function call DIV64_U64_ROUND_UP, division by expression
>> Which may be zero has undefined behavior.
>> Fix this accordingly.
>>
>> Signed-off-by: Shuijing Li <shuijing.li@mediatek.com>
>> ---
>>   drivers/pwm/pwm-mtk-disp.c | 9 ++++++++-
>>   1 file changed, 8 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/pwm/pwm-mtk-disp.c b/drivers/pwm/pwm-mtk-disp.c
>> index 79e321e96f56..ca00058a6ef4 100644
>> --- a/drivers/pwm/pwm-mtk-disp.c
>> +++ b/drivers/pwm/pwm-mtk-disp.c
>> @@ -196,6 +196,14 @@ static int mtk_disp_pwm_get_state(struct pwm_chip *chip,
>>           return err;
>>       }
>> +    rate = clk_get_rate(mdp->clk_main);
>> +    if (rate <= 0) {
>> +        dev_err(chip->dev, "Can't get rate: %pe\n", ERR_PTR(rate));
>> +        clk_disable_unprepare(mdp->clk_mm);
>> +        clk_disable_unprepare(mdp->clk_main);
>> +        return err;
>> +    }
>> +
>>       /*
>>        * Apply DISP_PWM_DEBUG settings to choose whether to enable or disable
>>        * registers double buffer and manual commit to working register before
>> @@ -206,7 +214,6 @@ static int mtk_disp_pwm_get_state(struct pwm_chip *chip,
>>                        mdp->data->bls_debug_mask,
>>                        mdp->data->bls_debug_mask);
>> -    rate = clk_get_rate(mdp->clk_main);
>>       con0 = readl(mdp->base + mdp->data->con0);
>>       con1 = readl(mdp->base + mdp->data->con1);
>>       pwm_en = readl(mdp->base + DISP_PWM_EN);
> 
> IMHO, it should be done int the function `mtk_disp_pwm_apply` too.
> Do you agree ?
> 

I think that realistically this will never happen.

We're getting the clk_main clock's handle at probe time (or fail and get out),
then the PWM clock never has the CLK_GET_RATE_NOCACHE flag, as there wouldn't
be any reason to not cache the rates for this clock.
But even if we had the CLK_GET_RATE_NOCACHE flag, it wouldn't change much, as
our validation actually happens at probe time...

This means that our call to clk_get_rate is guaranteed to have a not NULL
pointer to this clock's `struct clk_core` and, unless the declaration of
this clock in the clock controller driver was wrong, for it, or for its
parent, the branch...

	if (!core->num_parents || core->parent)
		return core->rate;

...is always satisfied, so, in the end, this instance of clk_get_rate() can't
really return zero.
If you got such an issue, I suggest to check what the problem is, as it is
likely to be outside of this driver.


...that, besides the fact that the proposed code is incorrect, as clk_get_rate()
returns an `unsigned long`, so `rate` can never be less than zero, anyway.


Cheers,
Angelo