Hi Dan,
On 5/4/24 12:25, Dan Carpenter wrote:
We accidentally deleted the "i++" as part of a cleanup. Restore it.
Fixes: 3f7ced7ac9af ("drivers/thermal/cpufreq_cooling : Refactor thermal_power_cpu_get_power tracing")
Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
---
This is based on static analysis and not tested.
Thank you for the patch. I have analyzed the code and why it
haven't trigger an issue when I was testing it.
I looks like the function get_load() which is called above that 'i++'
and takes the 'i' as the last argument is compiled in 2 versions:
1. for SMP system and the last argument 'cpu_idx' is ignored
2. for !SMP where we use the last argument 'cpu_idx' which is 'i'
value. Although, for !SMP system we only have 1 cpu, thus the
initialized 'int i = 0' at the beginning of that
cpufreq_get_requested_power() is used correctly.
The loop for !SMP goes only once.
drivers/thermal/cpufreq_cooling.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/thermal/cpufreq_cooling.c b/drivers/thermal/cpufreq_cooling.c
index 280071be30b1..a074192896de 100644
--- a/drivers/thermal/cpufreq_cooling.c
+++ b/drivers/thermal/cpufreq_cooling.c
@@ -249,6 +249,7 @@ static int cpufreq_get_requested_power(struct thermal_cooling_device *cdev,
load = 0;
total_load += load;
+ i++;
}
cpufreq_cdev->last_load = total_load;
Would you agree that I will keep you as 'Reported-by' and send a
separate patch to change that !SMP code completely in that
get_load() function and get rid of the 'cpu_idx' argument?
Or I'm happy that you can develop such code and I can review it.
It's up to you.
Regards,
Lukasz
