Thread (3 messages) flat view 3 messages, 3 authors, 2021-09-23

Re: [PATCH net-next] net: rtnetlink: convert rcu_assign_pointer to RCU_INIT_POINTER

From: Eric Dumazet <hidden>
Date: 2021-09-23 17:09:31
Also in: lkml


On 9/17/21 11:36 PM, Yajun Deng wrote:
quoted hunk ↗ jump to hunk
It no need barrier when assigning a NULL value to an RCU protected
pointer. So use RCU_INIT_POINTER() instead for more fast.

Signed-off-by: Yajun Deng <redacted>
---
 net/core/rtnetlink.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 972c8cb303a5..327ca6bc6e6d 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -301,7 +301,7 @@ int rtnl_unregister(int protocol, int msgtype)
 	}
 
 	link = rtnl_dereference(tab[msgindex]);
-	rcu_assign_pointer(tab[msgindex], NULL);
+	RCU_INIT_POINTER(tab[msgindex], NULL);
 	rtnl_unlock();
 
 	kfree_rcu(link, rcu);
@@ -337,7 +337,7 @@ void rtnl_unregister_all(int protocol)
 		if (!link)
 			continue;
 
-		rcu_assign_pointer(tab[msgindex], NULL);
+		RCU_INIT_POINTER(tab[msgindex], NULL);
 		kfree_rcu(link, rcu);
 	}
 	rtnl_unlock();
FYI, there is no memory barrier involved in 

rcu_assign_pointer(tab[msgindex], NULL);

This has been the case for the last 5 years.

Your patch was not needed really.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help