Re: [RFC PATCH v1 2/2] rcu: Ignore RCU in nohz_full cpus if it was running a guest recently

[Paul E. McKenney]

On Thu, Mar 28, 2024 at 02:19:47PM -0300, Leonardo Bras wrote:
> In current code, we can ignore the RCU request on a nohz_full cpu for up to
> a second if it has interrupted idle or userspace tasks, since those are
> quiescent states, and will probably return to it soon thus not requiring
> to run a softirq or a rcuc thread.
> 
> Running a guest is also considered to be a quiescent state, and will
> follow the same logic, so it makes sense to also ignore the RCU request in
> this case.
> 
> This solves a latency issue of a latency-sensitive workload running on a
> guest pinned in nohz_full cpu: if the guest goes out for any reason, and a
> synchronize_rcu() is requested between guest exit and a timer interrupt,
> then invoke_rcu_core() is called, and introduce latency due to either a
> softirq, or a reschedule to run rcuc, if the host is a PREEMPT_RT kernel.
> 
> Suggested-by: Marcelo Tosatti <mtosatti@xxxxxxxxxx>
> Signed-off-by: Leonardo Bras <leobras@xxxxxxxxxx>

Looks plausible to me!

Acked-by: Paul E. McKenney <paulmck@xxxxxxxxxx>

Or let me know if you would rather these go through -rcu.

							Thanx, Paul

> ---
>  kernel/rcu/tree_plugin.h | 14 ++++++++++++++
>  kernel/rcu/tree.c        |  4 +++-
>  2 files changed, 17 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 36a8b5dbf5b5..16f3cf2e15df 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -5,20 +5,21 @@
>   * or preemptible semantics.
>   *
>   * Copyright Red Hat, 2009
>   * Copyright IBM Corporation, 2009
>   *
>   * Author: Ingo Molnar <mingo@xxxxxxx>
>   *	   Paul E. McKenney <paulmck@xxxxxxxxxxxxx>
>   */
>  
>  #include "../locking/rtmutex_common.h"
> +#include "linux/kvm_host.h"
>  
>  static bool rcu_rdp_is_offloaded(struct rcu_data *rdp)
>  {
>  	/*
>  	 * In order to read the offloaded state of an rdp in a safe
>  	 * and stable way and prevent from its value to be changed
>  	 * under us, we must either hold the barrier mutex, the cpu
>  	 * hotplug lock (read or write) or the nocb lock. Local
>  	 * non-preemptible reads are also safe. NOCB kthreads and
>  	 * timers have their own means of synchronization against the
> @@ -1260,10 +1261,23 @@ static bool rcu_nohz_full_cpu(void)
>  
>  /*
>   * Bind the RCU grace-period kthreads to the housekeeping CPU.
>   */
>  static void rcu_bind_gp_kthread(void)
>  {
>  	if (!tick_nohz_full_enabled())
>  		return;
>  	housekeeping_affine(current, HK_TYPE_RCU);
>  }
> +
> +/*
> + * true if for this cpu guest exit is at most over a second ago,
> + * false otherwise
> + */
> +static bool rcu_recent_guest_exit(void)
> +{
> +#ifdef CONFIG_KVM
> +	return time_before(jiffies, guest_exit_last_time() + HZ);
> +#else
> +	return false;
> +#endif
> +}
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index d9642dd06c25..e5ce00bf1898 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -148,20 +148,21 @@ static void rcu_report_qs_rnp(unsigned long mask, struct rcu_node *rnp,
>  static struct task_struct *rcu_boost_task(struct rcu_node *rnp);
>  static void invoke_rcu_core(void);
>  static void rcu_report_exp_rdp(struct rcu_data *rdp);
>  static void sync_sched_exp_online_cleanup(int cpu);
>  static void check_cb_ovld_locked(struct rcu_data *rdp, struct rcu_node *rnp);
>  static bool rcu_rdp_is_offloaded(struct rcu_data *rdp);
>  static bool rcu_rdp_cpu_online(struct rcu_data *rdp);
>  static bool rcu_init_invoked(void);
>  static void rcu_cleanup_dead_rnp(struct rcu_node *rnp_leaf);
>  static void rcu_init_new_rnp(struct rcu_node *rnp_leaf);
> +static bool rcu_recent_guest_exit(void);
>  
>  /*
>   * rcuc/rcub/rcuop kthread realtime priority. The "rcuop"
>   * real-time priority(enabling/disabling) is controlled by
>   * the extra CONFIG_RCU_NOCB_CPU_CB_BOOST configuration.
>   */
>  static int kthread_prio = IS_ENABLED(CONFIG_RCU_BOOST) ? 1 : 0;
>  module_param(kthread_prio, int, 0444);
>  
>  /* Delay in jiffies for grace-period initialization delays, debug only. */
> @@ -3931,21 +3932,22 @@ static int rcu_pending(int user)
>  	lockdep_assert_irqs_disabled();
>  
>  	/* Check for CPU stalls, if enabled. */
>  	check_cpu_stall(rdp);
>  
>  	/* Does this CPU need a deferred NOCB wakeup? */
>  	if (rcu_nocb_need_deferred_wakeup(rdp, RCU_NOCB_WAKE))
>  		return 1;
>  
>  	/* Is this a nohz_full CPU in userspace or idle?  (Ignore RCU if so.) */
> -	if ((user || rcu_is_cpu_rrupt_from_idle()) && rcu_nohz_full_cpu())
> +	if ((user || rcu_is_cpu_rrupt_from_idle() || rcu_recent_guest_exit()) &&
> +	    rcu_nohz_full_cpu())
>  		return 0;
>  
>  	/* Is the RCU core waiting for a quiescent state from this CPU? */
>  	gp_in_progress = rcu_gp_in_progress();
>  	if (rdp->core_needs_qs && !rdp->cpu_no_qs.b.norm && gp_in_progress)
>  		return 1;
>  
>  	/* Does this CPU have callbacks ready to invoke? */
>  	if (!rcu_rdp_is_offloaded(rdp) &&
>  	    rcu_segcblist_ready_cbs(&rdp->cblist))
> -- 
> 2.44.0
>