Re: [PATCH net-next v11 13/14] dpll: zl3073x: Add support to get/set frequency on input pins

[Paolo Abeni <pabeni@redhat.com>]

On 6/19/25 2:15 PM, Vadim Fedorenko wrote:
> On 19/06/2025 12:15, Paolo Abeni wrote:
>> On 6/16/25 10:14 PM, Ivan Vecera wrote:
>>> +/**
>>> + * zl3073x_dpll_input_ref_frequency_get - get input reference frequency
>>> + * @zldpll: pointer to zl3073x_dpll
>>> + * @ref_id: reference id
>>> + * @frequency: pointer to variable to store frequency
>>> + *
>>> + * Reads frequency of given input reference.
>>> + *
>>> + * Return: 0 on success, <0 on error
>>> + */
>>> +static int
>>> +zl3073x_dpll_input_ref_frequency_get(struct zl3073x_dpll *zldpll, u8 ref_id,
>>> +				     u32 *frequency)
>>> +{
>>> +	struct zl3073x_dev *zldev = zldpll->dev;
>>> +	u16 base, mult, num, denom;
>>> +	int rc;
>>> +
>>> +	guard(mutex)(&zldev->multiop_lock);
>>> +
>>> +	/* Read reference configuration */
>>> +	rc = zl3073x_mb_op(zldev, ZL_REG_REF_MB_SEM, ZL_REF_MB_SEM_RD,
>>> +			   ZL_REG_REF_MB_MASK, BIT(ref_id));
>>> +	if (rc)
>>> +		return rc;
>>> +
>>> +	/* Read registers to compute resulting frequency */
>>> +	rc = zl3073x_read_u16(zldev, ZL_REG_REF_FREQ_BASE, &base);
>>> +	if (rc)
>>> +		return rc;
>>> +	rc = zl3073x_read_u16(zldev, ZL_REG_REF_FREQ_MULT, &mult);
>>> +	if (rc)
>>> +		return rc;
>>> +	rc = zl3073x_read_u16(zldev, ZL_REG_REF_RATIO_M, &num);
>>> +	if (rc)
>>> +		return rc;
>>> +	rc = zl3073x_read_u16(zldev, ZL_REG_REF_RATIO_N, &denom);
>>> +	if (rc)
>>> +		return rc;
>>> +
>>> +	/* Sanity check that HW has not returned zero denominator */
>>> +	if (!denom) {
>>> +		dev_err(zldev->dev,
>>> +			"Zero divisor for ref %u frequency got from device\n",
>>> +			ref_id);
>>> +		return -EINVAL;
>>> +	}
>>> +
>>> +	/* Compute the frequency */
>>> +	*frequency = base * mult * num / denom;
>>
>> As base, mult, num and denom are u16, the above looks like integer
>> overflow prone.
>>
>> I think you should explicitly cast to u64, and possibly use a u64 frequency.
> 
> I might be a good idea to use mul_u64_u32_div together with mul_u32_u32?
> These macroses will take care of overflow on 32bit platforms as well.

I guess such macros will work, but u64 is available on 32bits platform
as well - and possibly simpler/more readable.

/P