On 02/28/2017 11:07 AM, Daniel Bristot de Oliveira wrote: > + if (!pi_se->dl_throttled && dl_is_constrained(pi_se)) > + dl_check_constrained_dl(pi_se); > +
me--, it should be &p->dl, not pi_se. This is not causing problems in the test case because pi_se = &p->dl, but this would be a problem if we were activating the task in a PI case: it would check the pi-waiter, not the task being awakened (p). In the dl_check_constrained_dl(), the patch is already avoiding throttling if the task is "dl_boosted." So it is ready to avoid throttling the task in the PI control. Cooking a v4, sorry for the noise. -- Daniel
