| Message ID | 20240426064222.1152209-1-edumazet@google.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | cd42ba1c8ac9deb9032add6adf491110e7442040 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next] net: give more chances to rcu in netdev_wait_allrefs_any() | expand |
Fri, Apr 26, 2024 at 08:42:22AM CEST, edumazet@google.com wrote: >This came while reviewing commit c4e86b4363ac ("net: add two more >call_rcu_hurry()"). > >Paolo asked if adding one synchronize_rcu() would help. > >While synchronize_rcu() does not help, making sure to call >rcu_barrier() before msleep(wait) is definitely helping >to make sure lazy call_rcu() are completed. > >Instead of waiting ~100 seconds in my tests, the ref_tracker >splats occurs one time only, and netdev_wait_allrefs_any() >latency is reduced to the strict minimum. > >Ideally we should audit our call_rcu() users to make sure >no refcount (or cascading call_rcu()) is held too long, >because rcu_barrier() is quite expensive. > >Fixes: 0e4be9e57e8c ("net: use exponential backoff in netdev_wait_allrefs") >Signed-off-by: Eric Dumazet <edumazet@google.com> >Link: https://lore.kernel.org/all/28bbf698-befb-42f6-b561-851c67f464aa@kernel.org/T/#m76d73ed6b03cd930778ac4d20a777f22a08d6824 Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Hello: This patch was applied to netdev/net-next.git (main) by David S. Miller <davem@davemloft.net>: On Fri, 26 Apr 2024 06:42:22 +0000 you wrote: > This came while reviewing commit c4e86b4363ac ("net: add two more > call_rcu_hurry()"). > > Paolo asked if adding one synchronize_rcu() would help. > > While synchronize_rcu() does not help, making sure to call > rcu_barrier() before msleep(wait) is definitely helping > to make sure lazy call_rcu() are completed. > > [...] Here is the summary with links: - [net-next] net: give more chances to rcu in netdev_wait_allrefs_any() https://git.kernel.org/netdev/net-next/c/cd42ba1c8ac9 You are awesome, thank you!
diff --git a/net/core/dev.c b/net/core/dev.c index e09aa3785c159b4ab0fe7eb3546f9dd6797ebce2..c9e59eff8ec841f6267c2749489fdc7fe0d03430 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -10566,8 +10566,9 @@ static struct net_device *netdev_wait_allrefs_any(struct list_head *list) rebroadcast_time = jiffies; } + rcu_barrier(); + if (!wait) { - rcu_barrier(); wait = WAIT_REFS_MIN_MSECS; } else { msleep(wait);
This came while reviewing commit c4e86b4363ac ("net: add two more call_rcu_hurry()"). Paolo asked if adding one synchronize_rcu() would help. While synchronize_rcu() does not help, making sure to call rcu_barrier() before msleep(wait) is definitely helping to make sure lazy call_rcu() are completed. Instead of waiting ~100 seconds in my tests, the ref_tracker splats occurs one time only, and netdev_wait_allrefs_any() latency is reduced to the strict minimum. Ideally we should audit our call_rcu() users to make sure no refcount (or cascading call_rcu()) is held too long, because rcu_barrier() is quite expensive. Fixes: 0e4be9e57e8c ("net: use exponential backoff in netdev_wait_allrefs") Signed-off-by: Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/all/28bbf698-befb-42f6-b561-851c67f464aa@kernel.org/T/#m76d73ed6b03cd930778ac4d20a777f22a08d6824 --- net/core/dev.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)