Re: [PATCH v2] rcu: Provide a boot time parameter to control lazy RCU

[Paul E. McKenney]

On Fri, Dec 15, 2023 at 05:58:55PM +0100, Uladzislau Rezki wrote:
> Hello, Joel!
> 
> > [....]
> > > > > > +       Use rcutree.enable_rcu_lazy=0 to turn it off at boot time.
> > > > > > +
> > > > > > +config RCU_LAZY_DEFAULT_OFF
> > > > > > +     bool "Turn RCU lazy invocation off by default"
> > > > > > +     depends on RCU_LAZY
> > > > > > +     default n
> > > > > > +     help
> > > > > > +       Allows building the kernel with CONFIG_RCU_LAZY=y yet keep it default
> > > > > > +       off. Boot time param rcutree.enable_rcu_lazy=1 can be used to switch
> > > > > > +       it back on.
> > > > > > +
> > > > > >  config RCU_DOUBLE_CHECK_CB_TIME
> > > > > >       bool "RCU callback-batch backup time check"
> > > > > >       depends on RCU_EXPERT
> > > > > > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > > > > > index 3ac3c846105f..8b7675624815 100644
> > > > > > --- a/kernel/rcu/tree.c
> > > > > > +++ b/kernel/rcu/tree.c
> > > > > > @@ -2719,6 +2719,9 @@ __call_rcu_common(struct rcu_head *head, rcu_callback_t func, bool lazy_in)
> > > > > >  }
> > > > > >
> > > > > >  #ifdef CONFIG_RCU_LAZY
> > > > > > +static bool enable_rcu_lazy __read_mostly = !IS_ENABLED(CONFIG_RCU_LAZY_DEFAULT_OFF);
> > > > > > +module_param(enable_rcu_lazy, bool, 0444);
> > > > > > +
> > > > > >  /**
> > > > > >   * call_rcu_hurry() - Queue RCU callback for invocation after grace period, and
> > > > > >   * flush all lazy callbacks (including the new one) to the main ->cblist while
> > > > > > @@ -2744,6 +2747,8 @@ void call_rcu_hurry(struct rcu_head *head, rcu_callback_t func)
> > > > > >       __call_rcu_common(head, func, false);
> > > > > >  }
> > > > > >  EXPORT_SYMBOL_GPL(call_rcu_hurry);
> > > > > > +#else
> > > > > > +#define enable_rcu_lazy              false
> > > > > >  #endif
> > > > > >
> > > > > >  /**
> > > > > > @@ -2792,7 +2797,7 @@ EXPORT_SYMBOL_GPL(call_rcu_hurry);
> > > > > >   */
> > > > > >  void call_rcu(struct rcu_head *head, rcu_callback_t func)
> > > > > >  {
> > > > > > -     __call_rcu_common(head, func, IS_ENABLED(CONFIG_RCU_LAZY));
> > > > > > +     __call_rcu_common(head, func, enable_rcu_lazy);
> > > > > >  }
> > > > > >  EXPORT_SYMBOL_GPL(call_rcu);
> > > > > >
> > > > > I think, it makes sense. Especially for devices/systems where it is hard
> > > > > to recompile the kernel and deploy it. For example, Google and GKI approach.
> > > >
> > > > My concerns had nothing to do with recompiling the kernel. Passing a
> > > > boot parameter (without a kernel compile) can just as well
> > > > default-disable the feature.
> > > >
> > > > I think what Qais is saying is that passing a boot parameter is itself
> > > > a hassle in Android (something I did not know about) because of GKI
> > > > etc.
> > > >
> > > That is true. Doing:
> > >
> > > echo 1 > /sys/.../enable_lazy
> > >
> > > is a way how to make it easy and flexible.
> > 
> > Hey Vlad, are you suggesting that the boot parameter be made to
> > support runtime? We can keep that for later as it may get complicated.
> > Qais's boot parameter is designed only for boot time.
> > 
> No problem. Yes, i meant a runtime one. But as you stated there might
> be hidden issues witch we are not aware of yet.

My current thought is that Qais's version currently in -rcu for
the merge window after next (v6.9) suits our current situation.
But if we are eventually able to support runtime changes to this new
rcutree.enable_rcu_lazy module parameter via simplification to the
rcu_nocb_try_bypass() function (or maybe a better analysis of it),
then at that point it would be good to allow this module parameter to
be changed via sysfs at runtime.

Does that make sense, or am I missing some aspect or use case?

							Thanx, Paul