[PATCH 0/2] posix-timers: Reduce spinlock contention

[Eric Dumazet]

We had incidents involving gazillions of concurrent posix timers.

Benjamin Segall is planning to add a cond_resched() in exit_itimers()

In this small series, I am reducing hash_lock contention in posix_timer_add()

Eric Dumazet (2):
  posix-timers: Make next_posix_timer_id an atomic_t
  posix-timers: Use RCU in posix_timer_add()

 include/linux/sched/signal.h |  2 +-
 kernel/time/posix-timers.c   | 21 +++++++++++++++++----
 2 files changed, 18 insertions(+), 5 deletions(-)

-- 
2.48.1.601.g30ceb7b040-goog