(This was [PATCH 0/4] sched/idle: Fix missing need_resched() checks after rcu_idle_enter() v2)
I initially followed Peterz review but eventually I tried a different approach. Instead of handling the late wake up from rcu_idle_enter(), I've split the delayed rcuog wake up and moved it right before the last generic need_resched() check, it makes more sense and we don't need to fiddle with cpuidle core and drivers anymore. It's also less error prone.
I also fixed the nohz_full case and (hopefully) the guest case.
And this comes with debugging to prevent from that pattern to happen again.
Only lightly tested so far.
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks.git sched/idle-v3
HEAD: d95fc510e804a5c4658a823ff12d9caba1d906c7
Thanks, Frederic ---
Frederic Weisbecker (8): rcu: Remove superfluous rdp fetch rcu: Pull deferred rcuog wake up to rcu_eqs_enter() callers rcu/nocb: Perform deferred wake up before last idle's need_resched() check rcu/nocb: Trigger self-IPI on late deferred wake up before user resume entry: Explicitly flush pending rcuog wakeup before last rescheduling points sched: Report local wake up on resched blind zone within idle loop entry: Report local wake up on resched blind zone while resuming to user timer: Report ignored local enqueue in nohz mode
include/linux/rcupdate.h | 2 ++ include/linux/sched.h | 11 ++++++++ kernel/entry/common.c | 10 ++++++++ kernel/rcu/tree.c | 27 ++++++++++++++++++-- kernel/rcu/tree.h | 2 +- kernel/rcu/tree_plugin.h | 30 +++++++++++++++------- kernel/sched/core.c | 66 +++++++++++++++++++++++++++++++++++++++++++++++- kernel/sched/idle.c | 6 +++++ kernel/sched/sched.h | 3 +++ 9 files changed, 144 insertions(+), 13 deletions(-)