| 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: <20190709124102.GR26519@linux.ibm.com>
Date: Tue, 9 Jul 2019 05:41:02 -0700
From: "Paul E. McKenney" <paulmck@...ux.ibm.com>
To: Byungchul Park <byungchul.park@....com>
Cc: Joel Fernandes <joel@...lfernandes.org>, josh@...htriplett.org,
rostedt@...dmis.org, mathieu.desnoyers@...icios.com,
jiangshanlai@...il.com, rcu@...r.kernel.org,
linux-kernel@...r.kernel.org, kernel-team@....com
Subject: Re: [PATCH] rcu: Make jiffies_till_sched_qs writable
On Tue, Jul 09, 2019 at 02:58:16PM +0900, Byungchul Park wrote:
> On Mon, Jul 08, 2019 at 09:03:59AM -0400, Joel Fernandes wrote:
> > > Actually, the intent was to only allow this to be changed at boot time.
> > > Of course, if there is now a good reason to adjust it, it needs
> > > to be adjustable. So what situation is making you want to change
> > > jiffies_till_sched_qs at runtime? To what values is it proving useful
> > > to adjust it? What (if any) relationships between this timeout and the
> > > various other RCU timeouts need to be maintained? What changes to
> > > rcutorture should be applied in order to test the ability to change
> > > this at runtime?
> >
> > I am also interested in the context, are you changing it at runtime for
> > experimentation? I recently was doing some performance experiments and it is
> > quite interesting how reducing this value can shorten grace period times :)
>
> Hi Joel,
>
> I've read a thread talking about your experiment to see how the grace
> periods change depending on the tunnable variables which was interesting
> to me. While reading it, I found out jiffies_till_sched_qs is not
> tunnable at runtime unlike jiffies_till_{first,next}_fqs which looks
> like non-sense to me that's why I tried this patch. :)
>
> Hi Paul,
>
> IMHO, as much as we want to tune the time for fqs to be initiated, we
> can also want to tune the time for the help from scheduler to start.
> I thought only difference between them is a level of urgency. I might be
> wrong. It would be appreciated if you let me know if I miss something.
Hello, Byungchul,
I understand that one hypothetically might want to tune this at runtime,
but have you had need to tune this at runtime on a real production
workload? If so, what problem was happening that caused you to want to
do this tuning?
> And it's ok even if the patch is turned down based on your criteria. :)
If there is a real need, something needs to be provided to meet that
need. But in the absence of a real need, past experience has shown
that speculative tuning knobs usually do more harm than good. ;-)
Hence my question to you about a real need.
Thanx, Paul
> Thanks,
> Byungchul
>
> > Joel
> >
> >
> > > Thanx, Paul
> > >
> > > > The function for setting jiffies_to_sched_qs,
> > > > adjust_jiffies_till_sched_qs() will be called only if
> > > > the value from sysfs != ULONG_MAX. And the value won't be adjusted
> > > > unlike first/next fqs jiffies.
> > > >
> > > > While at it, changed the positions of two module_param()s downward.
> > > >
> > > > Signed-off-by: Byungchul Park <byungchul.park@....com>
> > > > ---
> > > > kernel/rcu/tree.c | 22 ++++++++++++++++++++--
> > > > 1 file changed, 20 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > > > index a2f8ba2..a28e2fe 100644
> > > > --- a/kernel/rcu/tree.c
> > > > +++ b/kernel/rcu/tree.c
> > > > @@ -422,9 +422,7 @@ static int rcu_is_cpu_rrupt_from_idle(void)
> > > > * quiescent-state help from rcu_note_context_switch().
> > > > */
> > > > static ulong jiffies_till_sched_qs = ULONG_MAX;
> > > > -module_param(jiffies_till_sched_qs, ulong, 0444);
> > > > static ulong jiffies_to_sched_qs; /* See adjust_jiffies_till_sched_qs(). */
> > > > -module_param(jiffies_to_sched_qs, ulong, 0444); /* Display only! */
> > > >
> > > > /*
> > > > * Make sure that we give the grace-period kthread time to detect any
> > > > @@ -450,6 +448,18 @@ static void adjust_jiffies_till_sched_qs(void)
> > > > WRITE_ONCE(jiffies_to_sched_qs, j);
> > > > }
> > > >
> > > > +static int param_set_sched_qs_jiffies(const char *val, const struct kernel_param *kp)
> > > > +{
> > > > + ulong j;
> > > > + int ret = kstrtoul(val, 0, &j);
> > > > +
> > > > + if (!ret && j != ULONG_MAX) {
> > > > + WRITE_ONCE(*(ulong *)kp->arg, j);
> > > > + adjust_jiffies_till_sched_qs();
> > > > + }
> > > > + return ret;
> > > > +}
> > > > +
> > > > static int param_set_first_fqs_jiffies(const char *val, const struct kernel_param *kp)
> > > > {
> > > > ulong j;
> > > > @@ -474,6 +484,11 @@ static int param_set_next_fqs_jiffies(const char *val, const struct kernel_param
> > > > return ret;
> > > > }
> > > >
> > > > +static struct kernel_param_ops sched_qs_jiffies_ops = {
> > > > + .set = param_set_sched_qs_jiffies,
> > > > + .get = param_get_ulong,
> > > > +};
> > > > +
> > > > static struct kernel_param_ops first_fqs_jiffies_ops = {
> > > > .set = param_set_first_fqs_jiffies,
> > > > .get = param_get_ulong,
> > > > @@ -484,8 +499,11 @@ static int param_set_next_fqs_jiffies(const char *val, const struct kernel_param
> > > > .get = param_get_ulong,
> > > > };
> > > >
> > > > +module_param_cb(jiffies_till_sched_qs, &sched_qs_jiffies_ops, &jiffies_till_sched_qs, 0644);
> > > > module_param_cb(jiffies_till_first_fqs, &first_fqs_jiffies_ops, &jiffies_till_first_fqs, 0644);
> > > > module_param_cb(jiffies_till_next_fqs, &next_fqs_jiffies_ops, &jiffies_till_next_fqs, 0644);
> > > > +
> > > > +module_param(jiffies_to_sched_qs, ulong, 0444); /* Display only! */
> > > > module_param(rcu_kick_kthreads, bool, 0644);
> > > >
> > > > static void force_qs_rnp(int (*f)(struct rcu_data *rdp));
> > > > --
> > > > 1.9.1
> > > >
> > >
>
Powered by blists - more mailing lists