[PATCH] drm/i915/gt: Ensure sleep calculations in wait_for_freq() do not use negative value of timeout_ms

[Jani Nikula]

On Tue, 05 Aug 2025, Krzysztof Karas <krzysztof.karas at intel.com> wrote:
> wait_for_freq() allows timeout_ms to be negative to pass it down
> to msecs_to_jiffies(). If the argument to the latter function is
> indeed negative, then MAX_JIFFY_OFFSET is returned from it.
>
> However, later in the wait_for_freq()'s "do while" loop,
> "timeout_ms" is used to calculate a new value for "sleep", which
> is a plain integer. While "sleep" being negative does not lead
> to erroneous behavior, as the "sleep" is cast to unsigned type
> in usleep_range(), it does not seem intentional.
>
> Change the type of "sleep" variable to unsigned and ensure it
> does not use "timeout_ms", when it is a negative value in the
> calculations.

That's an elaborate explanation for a case that never
occurs. wait_for_freq() is used in three places with fixed timeout
values.

What's the real reason for the change?

BR,
Jani.


>
> Signed-off-by: Krzysztof Karas <krzysztof.karas at intel.com>
> ---
>  drivers/gpu/drm/i915/gt/selftest_rps.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/selftest_rps.c b/drivers/gpu/drm/i915/gt/selftest_rps.c
> index 73bc91c6ea07..20ec7c0c94dc 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_rps.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_rps.c
> @@ -152,7 +152,7 @@ static u8 wait_for_freq(struct intel_rps *rps, u8 freq, int timeout_ms)
>  {
>  	u8 history[64], i;
>  	unsigned long end;
> -	int sleep;
> +	unsigned int sleep;
>  
>  	i = 0;
>  	memset(history, freq, sizeof(history));
> @@ -180,7 +180,7 @@ static u8 wait_for_freq(struct intel_rps *rps, u8 freq, int timeout_ms)
>  
>  		usleep_range(sleep, 2 * sleep);
>  		sleep *= 2;
> -		if (sleep > timeout_ms * 20)
> +		if (sleep > timeout_ms * 20 && timeout_ms > 0)
>  			sleep = timeout_ms * 20;
>  	} while (1);
>  }
> -- 
> 2.34.1

-- 
Jani Nikula, Intel