From ae646f1a0bb97401bac0044bbe2a179a1e38b408 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 14 Jan 2025 20:55:30 +0000 Subject: [PATCH] net: reduce RTNL hold duration in unregister_netdevice_many_notify() (part 1) Two synchronize_net() calls are currently done while holding RTNL. This is source of RTNL contention in workloads adding and deleting many network namespaces per second, because synchronize_rcu() and synchronize_rcu_expedited() can use 60+ ms in some cases. For cleanup_net() use, temporarily release RTNL while calling the last synchronize_net(). This should be safe, because devices are no longer visible to other threads at this point. In any case, the new netdev_lock() / netdev_unlock() infrastructure that we are adding should allow to fix potential issues, with a combination of a per-device mutex and dev->reg_state awareness. Signed-off-by: Eric Dumazet Reviewed-by: Jesse Brandeburg Link: https://patch.msgid.link/20250114205531.967841-5-edumazet@google.com Signed-off-by: Jakub Kicinski --- net/core/dev.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/net/core/dev.c b/net/core/dev.c index d652ea66e5755..3924a4af68b81 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -11629,6 +11629,7 @@ void unregister_netdevice_many_notify(struct list_head *head, rtnl_drop_if_cleanup_net(); flush_all_backlogs(); rtnl_acquire_if_cleanup_net(); + /* TODO: move this before the prior rtnl_acquire_if_cleanup_net() */ synchronize_net(); list_for_each_entry(dev, head, unreg_list) { @@ -11689,7 +11690,9 @@ void unregister_netdevice_many_notify(struct list_head *head, #endif } + rtnl_drop_if_cleanup_net(); synchronize_net(); + rtnl_acquire_if_cleanup_net(); list_for_each_entry(dev, head, unreg_list) { netdev_put(dev, &dev->dev_registered_tracker);