On 2019.11.19 14:14 Rafael J. Wysocki wrote:
> On Tuesday, November 19, 2019 8:17:05 PM CET Rafael J. Wysocki wrote:
...
>> However, I now also see that freq_qos_remove_request() doesn't clear
>> the qos field in req which is should do, so freq_qos_add_request()
>> will complain and fail if the object pointed to by req is passed to it
>> again.
>>
>> I'll send a patch to test for this later today.
>>
>
> The patch is appended. Please test it (on top of 5.4-rc8) and report back.
>
> ---
> kernel/power/qos.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> Index: linux-pm/kernel/power/qos.c
> ===================================================================
> --- linux-pm.orig/kernel/power/qos.c
> +++ linux-pm/kernel/power/qos.c
> @@ -814,6 +814,8 @@ EXPORT_SYMBOL_GPL(freq_qos_update_reques
> */
> int freq_qos_remove_request(struct freq_qos_request *req)
> {
> + int ret;
> +
> if (!req)
> return -EINVAL;
>
> @@ -821,7 +823,11 @@ int freq_qos_remove_request(struct freq_
> "%s() called for unknown object\n", __func__))
> return -EINVAL;
>
> - return freq_qos_apply(req, PM_QOS_REMOVE_REQ, PM_QOS_DEFAULT_VALUE);
> + ret = freq_qos_apply(req, PM_QOS_REMOVE_REQ, PM_QOS_DEFAULT_VALUE);
> + req->qos = NULL;
> + req->type = 0;
> +
> + return ret;
> }
> EXPORT_SYMBOL_GPL(freq_qos_remove_request);
Yes the patch fixes the problem. Thanks.
I tested several hundred times switching between
passive and active modes with the intel_pstate driver,
including with various CPUs disabled and re-enabled.
... Doug
