Re: [syzbot ci] Re: ipv6: Protect multiple routes insertion by table lock.

[Kuniyuki Iwashima <kuniyu@google.com>]

On Sat, 24 Oct 2026 10:18:08 -0700, syzbot ci wrote:
> syzbot ci has tested the following series
> 
> [v1] ipv6: Protect multiple routes insertion by table lock.
> https://lore.kernel.org/all/20261023191517.607834-1-kuniyu@google.com
> * [PATCH v1 net] ipv6: Protect multiple routes insertion by table lock.
> 
> and found the following issue:
> WARNING: suspicious RCU usage in rt6_multipath_rebalance
> 
> Full report is available here:
> https://ci.syzbot.org/series/7d878fa2-f84a-4468-85e2-c2a5e263aeff
> 
> ***
> 
> WARNING: suspicious RCU usage in rt6_multipath_rebalance
> 
> tree:      net
> URL:       https://kernel.googlesource.com/pub/scm/linux/kernel/git/netdev/net.git
> base:      7076f684d5132e6a6145e1da2fa74771ecd052e5
> arch:      amd64
> compiler:  Debian clang version 22.1.8 (1~deb13u4), Debian LLD 22.1.8
> config:    https://ci.syzbot.org/builds/bbb3053f-f23f-4fe3-a9e9-85396428119d/config
> 
> IPv6: RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE
> IPv6: NLM_F_CREATE should be set when creating new route
> =============================
> WARNING: suspicious RCU usage
> syzkaller #0 Not tainted
> -----------------------------
> ./include/net/addrconf.h:351 suspicious rcu_dereference_check() usage!
> 
> other info that might help us debug this:
> 
> 
> rcu_scheduler_active = 2, debug_locks = 1
> locks held by syz.2.3386/18922: 1, last CPU#0:
>  #0: ffff888177df3430 (&tb->tb6_lock){+...}-{3:3}, at: inet6_rtm_newroute+0x3dc/0x1ab0
> 
> stack backtrace:
> CPU: 0 UID: 0 PID: 18922 Comm: syz.2.3386 Not tainted syzkaller #0 PREEMPT(full) 
> Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.2-debian-1.16.2-1 04/01/2014
> Call Trace:
>  <TASK>
>  dump_stack_lvl+0xe8/0x150
>  lockdep_rcu_suspicious+0x140/0x1d0
>  rt6_multipath_rebalance+0xf11/0x1080
>  fib6_del+0xada/0x15a0

Similar to 8e5f1bb81274 ("ipv6: Narrow down RCU critical section
in inet6_rtm_newroute()."), I had to wrap rt6_multipath_rebalance()
with RCU.

Will squash this:

---8<---
diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index 0dc6ebee6e56..f7c69c9df411 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c
@@ -1228,9 +1228,7 @@ static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
 			fib6_nsiblings++;
 		}
 		BUG_ON(fib6_nsiblings != rt->fib6_nsiblings);
-		rcu_read_lock();
 		rt6_multipath_rebalance(temp_sibling);
-		rcu_read_unlock();
 	}
 
 	/*
@@ -1274,9 +1272,7 @@ static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
 						   sibling->fib6_nsiblings - 1);
 				WRITE_ONCE(rt->fib6_nsiblings, 0);
 				list_del_rcu(&rt->fib6_siblings);
-				rcu_read_lock();
 				rt6_multipath_rebalance(next_sibling);
-				rcu_read_unlock();
 				return err;
 			}
 		}
diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index 5926ddbd4349..9618e7b635e9 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -4875,8 +4875,10 @@ void rt6_multipath_rebalance(struct fib6_info *rt)
 	if (WARN_ON_ONCE(!first))
 		return;
 
+	rcu_read_lock();
 	total = rt6_multipath_total_weight(first);
 	rt6_multipath_upper_bound_set(first, total);
+	rcu_read_unlock();
 }
 
 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
---8<---

Thanks