[net 1/1] tipc: fix unsafe rcu locking when accessing publication list

[net 1/1] tipc: fix unsafe rcu locking when accessing publication list

[Jon Maloy]

From: Tung Nguyen <tung.q.nguyen@dektech.com.au>

The binding table's 'cluster_scope' list is rcu protected to handle
races between threads changing the list and those traversing the list at
the same moment. We have now found that the function named_distribute()
uses the regular list_for_each() macro to traverse the said list.
Likewise, the function tipc_named_withdraw() is removing items from the
same list using the regular list_del() call. When these two functions
execute in parallel we see occasional crashes.

This commit fixes this by adding the missing _rcu() suffixes.

Signed-off-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
---
 net/tipc/name_distr.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/tipc/name_distr.c b/net/tipc/name_distr.c
index 51b4b96..3cfeb9d 100644
--- a/net/tipc/name_distr.c
+++ b/net/tipc/name_distr.c
@@ -115,7 +115,7 @@ struct sk_buff *tipc_named_withdraw(struct net *net, struct publication *publ)
 	struct sk_buff *buf;
 	struct distr_item *item;
 
-	list_del(&publ->binding_node);
+	list_del_rcu(&publ->binding_node);
 
 	if (publ->scope == TIPC_NODE_SCOPE)
 		return NULL;
@@ -147,7 +147,7 @@ static void named_distribute(struct net *net, struct sk_buff_head *list,
 			ITEM_SIZE) * ITEM_SIZE;
 	u32 msg_rem = msg_dsz;
 
-	list_for_each_entry(publ, pls, binding_node) {
+	list_for_each_entry_rcu(publ, pls, binding_node) {
 		/* Prepare next buffer: */
 		if (!skb) {
 			skb = named_prepare_buf(net, PUBLICATION, msg_rem,
-- 
2.1.4

Re: [net 1/1] tipc: fix unsafe rcu locking when accessing publication list

[David Miller]

From: Jon Maloy <jon.maloy@ericsson.com>
Date: Fri, 12 Oct 2018 22:46:55 +0200

> From: Tung Nguyen <tung.q.nguyen@dektech.com.au>
> 
> The binding table's 'cluster_scope' list is rcu protected to handle
> races between threads changing the list and those traversing the list at
> the same moment. We have now found that the function named_distribute()
> uses the regular list_for_each() macro to traverse the said list.
> Likewise, the function tipc_named_withdraw() is removing items from the
> same list using the regular list_del() call. When these two functions
> execute in parallel we see occasional crashes.
> 
> This commit fixes this by adding the missing _rcu() suffixes.
> 
> Signed-off-by: Tung Nguyen <tung.q.nguyen@dektech.com.au>
> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>

Applied and queued up for -stable.