Quick review of RCU-related patches in v5.10.8-rt23

Quick review of RCU-related patches in v5.10.8-rt23

[Paul E. McKenney]

Hello, Sebastian,

Just doing my periodic (but decidedly non-real-time) scan of RCU-related
patches in -rt, in this case v5.10.8-rt23:

db93e2f1b4b0 ("rcu: Prevent false positive softirq warning on RT")
	Looks ready for mainline, given CONFIG_PREEMPT_RT.
f3541b467fbb ("sched: Do not account rcu_preempt_depth on RT in might_sleep()")
	If the scheduler maintainers are OK with their part of this patch,
	looks good to me, given CONFIG_PREEMPT_RT.  Feel free to add:
	Acked-by: Paul E. McKenney <paulmck@ekernel.org>
d8c5a7d75e08 ("rcutorture: Avoid problematic critical section nesting on RT")
	This one I need to understand better.  I do like the use of local
	variables to make the "if" conditions less unruly.

The rest are in -rcu already:

a163ef8687a1 ("rcu: make RCU_BOOST default on RT")
	Commit 2341bc4a0311 in -rcu.  In yesterday's pull request.
5ffd75a96828 ("rcu: Use rcuc threads on PREEMPT_RT as we did")
	Commit 8b9a0ecc7ef5 in -rcu.  In yesterday's pull request.
e0b671bca2e7 ("rcu: enable rcu_normal_after_boot by default for RT")
	Commit 36221e109eb2 in -rcu.  In yesterday's pull request.
e27ef68731a1 ("rcu: Don't invoke try_invoke_on_locked_down_task() with irqs disabled")
	This one is in v5.10 mainline.

Any reason I shouldn't pull in db93e2f1b4b0 ("rcu: Prevent false positive
softirq warning on RT") for v5.13?

							Thanx, Paul

Re: Quick review of RCU-related patches in v5.10.8-rt23

[Paul E. McKenney]

On Fri, Jan 29, 2021 at 05:11:37PM +0100, Sebastian Andrzej Siewior wrote:
> On 2021-01-28 11:50:37 [-0800], Paul E. McKenney wrote:
> > Hello, Sebastian,
> 
> Hi Paul,
> 
> > Just doing my periodic (but decidedly non-real-time) scan of RCU-related
> > patches in -rt, in this case v5.10.8-rt23:
> > 
> > f3541b467fbb ("sched: Do not account rcu_preempt_depth on RT in might_sleep()")
> > 	If the scheduler maintainers are OK with their part of this patch,
> > 	looks good to me, given CONFIG_PREEMPT_RT.  Feel free to add:
> > 	Acked-by: Paul E. McKenney <paulmck@ekernel.org>
> 
> Thank. I think we should pump it together with the rt-mutex part. But I
> add a note.
> 
> > d8c5a7d75e08 ("rcutorture: Avoid problematic critical section nesting on RT")
> > 	This one I need to understand better.  I do like the use of local
> > 	variables to make the "if" conditions less unruly.
> 
> This originated in
>   https://lkml.kernel.org/r/20190911165729.11178-6-swood@redhat.com
> 
> I planned to post it upstream last cycle but it appears that it broke
> apart and I did not yet look how to fix it.

I do recall the discussion, I just need to get up to speed on the
details.  ;-)

> > The rest are in -rcu already:
> > 
> > a163ef8687a1 ("rcu: make RCU_BOOST default on RT")
> > 	Commit 2341bc4a0311 in -rcu.  In yesterday's pull request.
> > 5ffd75a96828 ("rcu: Use rcuc threads on PREEMPT_RT as we did")
> > 	Commit 8b9a0ecc7ef5 in -rcu.  In yesterday's pull request.
> > e0b671bca2e7 ("rcu: enable rcu_normal_after_boot by default for RT")
> > 	Commit 36221e109eb2 in -rcu.  In yesterday's pull request.
> > e27ef68731a1 ("rcu: Don't invoke try_invoke_on_locked_down_task() with irqs disabled")
> > 	This one is in v5.10 mainline.
> 
>  \o/
>  
> > Any reason I shouldn't pull in db93e2f1b4b0 ("rcu: Prevent false positive
> > softirq warning on RT") for v5.13?
> 
> tglx has a version of that with your Reviewed-by tag on it in this
> softirq tree waiting. So I guess just sit it out ;)

Works for me!

							Thanx, Paul

> Thank you for looking Paul.
> > 							Thanx, Paul
> 
> Sebastian