lkml.org 
[lkml]   [2014]   [Jul]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 3/3] rcu: Use rcu_num_nodes instead of NUM_RCU_NODES
On Tue, Jul 15, 2014 at 06:31:49PM -0400, Pranith Kumar wrote:
> NUM_RCU_NODES is set at build time and is usually a huge number. We calculate the
> actual number of rcu nodes necessary at boot time based on nr_cpu_ids in
> rcu_init_geometry() and store it in rcu_num_nodes. We should use this variable
> instead of NUM_RCU_NODES.
>
> This commit changes all such uses of NUM_RCU_NODES to rcu_num_nodes.
>
> Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
> ---
> kernel/rcu/tree_plugin.h | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index cedb020..17ccb62 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -46,6 +46,8 @@ static bool __read_mostly rcu_nocb_poll; /* Offload kthread are to poll. */
> static char __initdata nocb_buf[NR_CPUS * 5];
> #endif /* #ifdef CONFIG_RCU_NOCB_CPU */
>
> +extern int rcu_num_nodes;

This should not be necessary given the existing declaration in
kernel/rcu/tree.c way before the #include of kernel/rcu/tree_plugin.h.

Or did you get a build failure without this? (And if you did get a build
failure, I would be really curious how that happened!)

So please send an updated patch or tell me how your build managed to fail.

> +
> /*
> * Check the RCU kernel configuration parameters and print informative
> * messages about anything out of the ordinary. If you like #ifdef, you
> @@ -885,7 +887,7 @@ void synchronize_rcu_expedited(void)
> /* Snapshot current state of ->blkd_tasks lists. */
> rcu_for_each_leaf_node(rsp, rnp)
> sync_rcu_preempt_exp_init(rsp, rnp);
> - if (NUM_RCU_NODES > 1)
> + if (rcu_num_nodes > 1)
> sync_rcu_preempt_exp_init(rsp, rcu_get_root(rsp));
>
> put_online_cpus();
> @@ -1475,7 +1477,7 @@ static void __init rcu_spawn_boost_kthreads(void)
> BUG_ON(smpboot_register_percpu_thread(&rcu_cpu_thread_spec));
> rnp = rcu_get_root(rcu_state_p);
> (void)rcu_spawn_one_boost_kthread(rcu_state_p, rnp);
> - if (NUM_RCU_NODES > 1) {
> + if (rcu_num_nodes > 1) {
> rcu_for_each_leaf_node(rcu_state_p, rnp)
> (void)rcu_spawn_one_boost_kthread(rcu_state_p, rnp);
> }

For whatever it is worth, the reason that this works is that it is a
perforamnce optimzation and NUM_RCU_NODES is always greater than or
equal to num_rcu_nodes. Still, your change is a good one.

Thanx, Paul



\
 
 \ /
  Last update: 2014-07-16 15:21    [W:0.128 / U:1.180 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site