On Tue, Dec 19, 2023 at 12:19:13AM +0100, Frederic Weisbecker wrote:
> 5c0930ccaad5 ("hrtimers: Push pending hrtimers away from outgoing CPU earlier")
> has introduced an issue with RCU. This is a proposal to solve the
> situation after realizing that fixing that on the timers side wouldn't
> be pretty to say the least.
>
> Oh and the last patch is absolutely irrelevant to the issue...
I pulled all three in and started testing.
That last one will need some careful review, but you knew that already. ;-)
Thanx, Paul
> Frederic Weisbecker (3):
> hrtimer: Report offline hrtimer enqueue
> rcu: Defer RCU kthreads wakeup when CPU is dying
> rcu/exp: Remove full barrier upon main thread wakeup
>
> include/linux/hrtimer.h | 3 ++-
> kernel/rcu/tree.c | 34 +++++++++++++++++++++++++++++++++-
> kernel/rcu/tree_exp.h | 8 +++-----
> kernel/time/hrtimer.c | 3 +++
> 4 files changed, 41 insertions(+), 7 deletions(-)
>
> --
> 2.42.1
>