| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20200619032347.GH2723@paulmck-ThinkPad-P72>
Date: Thu, 18 Jun 2020 20:23:47 -0700
From: "Paul E. McKenney" <paulmck@...nel.org>
To: Joel Fernandes <joel@...lfernandes.org>
Cc: linux-kernel@...r.kernel.org, Davidlohr Bueso <dave@...olabs.net>,
Ingo Molnar <mingo@...hat.com>,
Josh Triplett <josh@...htriplett.org>,
Lai Jiangshan <jiangshanlai@...il.com>,
Marco Elver <elver@...gle.com>,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
rcu@...r.kernel.org, Steven Rostedt <rostedt@...dmis.org>,
"Uladzislau Rezki (Sony)" <urezki@...il.com>
Subject: Re: [PATCH 6/7] rcutorture: Add support to get the number of wakeups
of main GP kthread
On Thu, Jun 18, 2020 at 09:00:12PM -0400, Joel Fernandes wrote:
> On Thu, Jun 18, 2020 at 05:12:44PM -0700, Paul E. McKenney wrote:
> > On Thu, Jun 18, 2020 at 08:01:56PM -0400, Joel Fernandes wrote:
> > > On Thu, Jun 18, 2020 at 03:40:58PM -0700, Paul E. McKenney wrote:
> > > > On Thu, Jun 18, 2020 at 04:29:54PM -0400, Joel Fernandes (Google) wrote:
> > > > > This is useful to check for any improvements or degradation related to
> > > > > number of GP kthread wakeups during testing.
> > > > >
> > > > > Signed-off-by: Joel Fernandes (Google) <joel@...lfernandes.org>
> > > >
> > > > This was a good way to collect the data for your testing, but
> > > > we can expect rcutorture to only do so much. ;-)
> > >
> > > np, I will push this one into a git tag for the next time I need it ;)
> >
> > Sounds like a plan!
>
> In case it is at all an option, I could put this as a statistic under
> rcu_torture_stats_print(), that way it all is on the same line. But I take it
> you are not too thrilled to have it for now.
The thing is that rcutorture is a stress test, not a performance or
energy-efficiency test. So rcutorture will continue to wake things
up just to brutalize RCU, which makes such a statistic less helpful.
It might be something for rcuperf/rcuscale, although it is not clear that
a pure performance/scalability test would give a useful read on wakeups.
I would expect better results from a more random real-world workload.
But I bet that people already have this sort of thing set up. For
example, doesn't kbuild test robot track this sort of thing?
Thanx, Paul
> thanks,
>
> - Joel
>
>
>
> > Thanx, Paul
> >
> > > thanks,
> > >
> > > - Joel
> > >
> > >
> > > > Thanx, Paul
> > > >
> > > > > ---
> > > > > kernel/rcu/Kconfig.debug | 1 +
> > > > > kernel/rcu/rcu.h | 2 ++
> > > > > kernel/rcu/rcutorture.c | 23 ++++++++++++++++++++++-
> > > > > kernel/rcu/tree.c | 7 +++++++
> > > > > 4 files changed, 32 insertions(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/kernel/rcu/Kconfig.debug b/kernel/rcu/Kconfig.debug
> > > > > index 3cf6132a4bb9f..3323e3378af5a 100644
> > > > > --- a/kernel/rcu/Kconfig.debug
> > > > > +++ b/kernel/rcu/Kconfig.debug
> > > > > @@ -50,6 +50,7 @@ config RCU_TORTURE_TEST
> > > > > select TASKS_RCU
> > > > > select TASKS_RUDE_RCU
> > > > > select TASKS_TRACE_RCU
> > > > > + select SCHEDSTATS
> > > > > default n
> > > > > help
> > > > > This option provides a kernel module that runs torture tests
> > > > > diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
> > > > > index cf66a3ccd7573..7e867e81d9738 100644
> > > > > --- a/kernel/rcu/rcu.h
> > > > > +++ b/kernel/rcu/rcu.h
> > > > > @@ -511,6 +511,7 @@ srcu_batches_completed(struct srcu_struct *sp) { return 0; }
> > > > > static inline void rcu_force_quiescent_state(void) { }
> > > > > static inline void show_rcu_gp_kthreads(void) { }
> > > > > static inline int rcu_get_gp_kthreads_prio(void) { return 0; }
> > > > > +static inline struct task_struct *rcu_get_main_gp_kthread(void) { return 0; }
> > > > > static inline void rcu_fwd_progress_check(unsigned long j) { }
> > > > > #else /* #ifdef CONFIG_TINY_RCU */
> > > > > bool rcu_dynticks_zero_in_eqs(int cpu, int *vp);
> > > > > @@ -519,6 +520,7 @@ unsigned long rcu_exp_batches_completed(void);
> > > > > unsigned long srcu_batches_completed(struct srcu_struct *sp);
> > > > > void show_rcu_gp_kthreads(void);
> > > > > int rcu_get_gp_kthreads_prio(void);
> > > > > +struct task_struct *rcu_get_main_gp_kthread(void);
> > > > > void rcu_fwd_progress_check(unsigned long j);
> > > > > void rcu_force_quiescent_state(void);
> > > > > extern struct workqueue_struct *rcu_gp_wq;
> > > > > diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
> > > > > index d0d265304d147..959a1f84d6904 100644
> > > > > --- a/kernel/rcu/rcutorture.c
> > > > > +++ b/kernel/rcu/rcutorture.c
> > > > > @@ -23,6 +23,7 @@
> > > > > #include <linux/rcupdate_wait.h>
> > > > > #include <linux/interrupt.h>
> > > > > #include <linux/sched/signal.h>
> > > > > +#include <linux/sched/stat.h>
> > > > > #include <uapi/linux/sched/types.h>
> > > > > #include <linux/atomic.h>
> > > > > #include <linux/bitops.h>
> > > > > @@ -460,9 +461,29 @@ static void rcu_sync_torture_init(void)
> > > > > INIT_LIST_HEAD(&rcu_torture_removed);
> > > > > }
> > > > >
> > > > > +unsigned long rcu_gp_nr_wakeups;
> > > > > +
> > > > > +static void rcu_flavor_init(void)
> > > > > +{
> > > > > + rcu_sync_torture_init();
> > > > > +
> > > > > + /* Make sure schedstat is enabled for GP thread wakeup count. */
> > > > > + force_schedstat_enabled();
> > > > > + rcu_gp_nr_wakeups = rcu_get_main_gp_kthread()->se.statistics.nr_wakeups;
> > > > > +}
> > > > > +
> > > > > +static void rcu_flavor_cleanup(void)
> > > > > +{
> > > > > + unsigned long now_nr = rcu_get_main_gp_kthread()->se.statistics.nr_wakeups;
> > > > > +
> > > > > + pr_alert("End-test: Cleanup: Total GP-kthread wakeups: %lu\n",
> > > > > + now_nr - rcu_gp_nr_wakeups);
> > > > > +}
> > > > > +
> > > > > static struct rcu_torture_ops rcu_ops = {
> > > > > .ttype = RCU_FLAVOR,
> > > > > - .init = rcu_sync_torture_init,
> > > > > + .init = rcu_flavor_init,
> > > > > + .cleanup = rcu_flavor_cleanup,
> > > > > .readlock = rcu_torture_read_lock,
> > > > > .read_delay = rcu_read_delay,
> > > > > .readunlock = rcu_torture_read_unlock,
> > > > > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > > > > index c3bae7a83d792..a3a175feb310a 100644
> > > > > --- a/kernel/rcu/tree.c
> > > > > +++ b/kernel/rcu/tree.c
> > > > > @@ -175,6 +175,13 @@ int rcu_get_gp_kthreads_prio(void)
> > > > > }
> > > > > EXPORT_SYMBOL_GPL(rcu_get_gp_kthreads_prio);
> > > > >
> > > > > +/* Retrieve RCU's main GP kthread task_struct */
> > > > > +struct task_struct *rcu_get_main_gp_kthread(void)
> > > > > +{
> > > > > + return rcu_state.gp_kthread;
> > > > > +}
> > > > > +EXPORT_SYMBOL_GPL(rcu_get_main_gp_kthread);
> > > > > +
> > > > > /*
> > > > > * Number of grace periods between delays, normalized by the duration of
> > > > > * the delay. The longer the delay, the more the grace periods between
> > > > > --
> > > > > 2.27.0.111.gc72c7da667-goog
> > > > >
Powered by blists - more mailing lists