Re: [PATCH v2 net-next 6/9] net: netdevsim: create a mock-up PTP Hardware Clock driver

[Dan Carpenter <dan.carpenter@linaro.org>]

On Wed, Jun 14, 2023 at 12:54:37AM +0300, Vladimir Oltean wrote:
> +
> +	spin_lock_init(&phc->lock);
> +	timecounter_init(&phc->tc, &phc->cc, 0);
> +
> +	phc->clock = ptp_clock_register(&phc->info, dev);
> +	if (IS_ERR_OR_NULL(phc->clock)) {
> +		err = PTR_ERR_OR_ZERO(phc->clock);
> +		goto out_free_phc;
> +	}
> +
> +	ptp_schedule_worker(phc->clock, MOCK_PHC_REFRESH_INTERVAL);
> +
> +	return phc;
> +
> +out_free_phc:
> +	kfree(phc);
> +out:
> +	return ERR_PTR(err);
> +}

Simon added me to the CC list because this code generates a Smatch
warning about passing zero to ERR_PTR() which is NULL.  I have written
a blog about this kind of warning.

https://staticthinking.wordpress.com/2022/08/01/mixing-error-pointers-and-null/

Returning NULL can be perfectly fine, but in this code here it will lead
to a crash.  The caller checks for error pointers but after that it
assumes that "phc" is a non-NULL pointer.

The IS_ERR_OR_NULL() check is not correct.  It should just be if
if (IS_ERR()).

However, the question is can this driver work without a phc->clock?
Depending on the answer you have to do one of two things.
If yes, then the correct thing is to add NULL checks throughout the
driver to prevent a NULL dereference.

If no, then the correct thing is to ensure that CONFIG_PTP_1588_CLOCK is
enabled using Kconfig.  We should never have a driver where we compile
it and then it can't probe.

regards,
dan carpenter