On 31 January 2018 at 16:27, Daniel Lezcano [off-list ref] wrote:

On 31/01/2018 10:56, Vincent Guittot wrote:

quoted

On 31 January 2018 at 10:50, Daniel Lezcano [off-list ref] wrote:

quoted

On 31/01/2018 10:46, Vincent Guittot wrote:

quoted

On 31 January 2018 at 10:33, Daniel Lezcano [off-list ref] wrote:

quoted

On 31/01/2018 10:01, Vincent Guittot wrote:

quoted

Hi Daniel,

On 23 January 2018 at 16:34, Daniel Lezcano [off-list ref] wrote:

[ ... ] (please trim :)

quoted

quoted

+               /*
+                * Each cooling device is per package. Each package
+                * has a set of cpus where the physical number is
+                * duplicate in the kernel namespace. We need a way to
+                * address the waitq[] and tsk[] arrays with index
+                * which are not Linux cpu numbered.
+                *
+                * One solution is to use the
+                * topology_core_id(cpu). Other solution is to use the
+                * modulo.
+                *
+                * eg. 2 x cluster - 4 cores.
+                *
+                * Physical numbering -> Linux numbering -> % nr_cpus
+                *
+                * Pkg0 - Cpu0 -> 0 -> 0
+                * Pkg0 - Cpu1 -> 1 -> 1
+                * Pkg0 - Cpu2 -> 2 -> 2
+                * Pkg0 - Cpu3 -> 3 -> 3
+                *
+                * Pkg1 - Cpu0 -> 4 -> 0
+                * Pkg1 - Cpu1 -> 5 -> 1
+                * Pkg1 - Cpu2 -> 6 -> 2
+                * Pkg1 - Cpu3 -> 7 -> 3

I'm not sure that the assumption above for the CPU numbering is safe.
Can't you use a per cpu structure to point to resources that are per
cpu instead ? so you will not have to rely on CPU ordering

Can you elaborate ? I don't get the part with the percpu structure.

Something like:

struct cpuidle_cooling_cpu {
       struct task_struct *tsk;
       wait_queue_head_t waitq;
};

DECLARE_PER_CPU(struct cpuidle_cooling_cpu *, cpu_data);

I got this part but I don't get how that fixes the ordering thing.

Because you don't care of the CPU ordering to retrieve the data as
they are stored per cpu directly

That's what I did initially, but for consistency reasons with the
cpufreq cpu cooling device which is stored in a list and the combo cpu
cooling device, the cpuidle cooling device must be per cluster and
stored in a list.

I'm not sure to catch your problem. You can still have cpuidle cooling
device per cluster and stored in the list but keep per cpu data in a

AFAICT, you will not have more than one cpu cooling device registered
per CPU so one per cpu variable that will gathers cpu private data
should be enough ?

Alternatively I can do:

struct cpuidle_cooling_device {
        struct thermal_cooling_device *cdev;
-       struct task_struct **tsk;
+       struct task_struct __percpu *tsk;
        struct cpumask *cpumask;
        struct list_head node;
        struct hrtimer timer;
        struct kref kref;
-       wait_queue_head_t *waitq;
+       wait_queue_head_t __percpu waitq;
        atomic_t count;
        unsigned int idle_cycle;
        unsigned int state;
};

struct cpuidle_cooling_device {
         struct thermal_cooling_device *cdev;
         struct cpumask *cpumask;
         struct list_head node;
         struct hrtimer timer;
         struct kref kref;
         atomic_t count;
         unsigned int idle_cycle;
         unsigned int state;
};

struct cpuidle_cooling_cpu {
        struct task_struct *tsk;
        wait_queue_head_t waitq;
};
DECLARE_PER_CPU(struct cpuidle_cooling_cpu *, cpu_data);

You continue to have cpuidle_cooling_device allocated dynamically per
cluster and added in the list but task and waitq are stored per cpu

--
 <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog