On 3/30/21 1:59 PM, Eric Dumazet wrote:
>
>
> On 3/25/21 5:16 PM, Taehee Yoo wrote:
>> The purpose of this lock is to avoid a bottleneck in the query/report
>> event handler logic.
>>
>> By previous patches, almost all mld data is protected by RTNL.
>> So, the query and report event handler, which is data path logic
>> acquires RTNL too. Therefore if a lot of query and report events
>> are received, it uses RTNL for a long time.
>> So it makes the control-plane bottleneck because of using RTNL.
>> In order to avoid this bottleneck, mc_lock is added.
>>
>> mc_lock protect only per-interface mld data and per-interface mld
>> data is used in the query/report event handler logic.
>> So, no longer rtnl_lock is needed in the query/report event handler logic.
>> Therefore bottleneck will be disappeared by mc_lock.
>>
>
> What testsuite have you run exactly to validate this monster patch ?
>
> Have you used CONFIG_LOCKDEP=y / CONFIG_DEBUG_ATOMIC_SLEEP=y ?
>
>> Suggested-by: Cong Wang <xiyou.wangcong@xxxxxxxxx>
>> Signed-off-by: Taehee Yoo <ap420073@xxxxxxxxx>
>
> [...]
>
>> /*
>> - * device multicast group del
>> + * device multicast group del
>> */
>> int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
>> {
>> @@ -943,8 +967,9 @@ int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
>>
>> ASSERT_RTNL();
>>
>> + mutex_lock(&idev->mc_lock);
>> for (map = &idev->mc_list;
>> - (ma = rtnl_dereference(*map));
>> + (ma = mc_dereference(*map, idev));
>> map = &ma->next) {
>> if (ipv6_addr_equal(&ma->mca_addr, addr)) {
>> if (--ma->mca_users == 0) {
>
> This can be called with rcu_bh held, thus :
>
> BUG: sleeping function called from invalid context at kernel/locking/mutex.c:928
> in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 4624, name: kworker/1:2
> 4 locks held by kworker/1:2/4624:
> #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: arch_atomic64_set arch/x86/include/asm/atomic64_64.h:34 [inline]
> #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: atomic64_set include/asm-generic/atomic-instrumented.h:856 [inline]
> #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: atomic_long_set include/asm-generic/atomic-long.h:41 [inline]
> #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: set_work_data kernel/workqueue.c:616 [inline]
> #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: set_work_pool_and_clear_pending kernel/workqueue.c:643 [inline]
> #0: ffff88802135d138 ((wq_completion)ipv6_addrconf){+.+.}-{0:0}, at: process_one_work+0x871/0x1600 kernel/workqueue.c:2246
> #1: ffffc90009adfda8 ((addr_chk_work).work){+.+.}-{0:0}, at: process_one_work+0x8a5/0x1600 kernel/workqueue.c:2250
> #2: ffffffff8d66d328 (rtnl_mutex){+.+.}-{3:3}, at: addrconf_verify_work+0xa/0x20 net/ipv6/addrconf.c:4572
> #3: ffffffff8bf74300 (rcu_read_lock_bh){....}-{1:2}, at: addrconf_verify_rtnl+0x2b/0x1150 net/ipv6/addrconf.c:4459
> Preemption disabled at:
> [<ffffffff87b39f41>] local_bh_disable include/linux/bottom_half.h:19 [inline]
> [<ffffffff87b39f41>] rcu_read_lock_bh include/linux/rcupdate.h:727 [inline]
> [<ffffffff87b39f41>] addrconf_verify_rtnl+0x41/0x1150 net/ipv6/addrconf.c:4461
> CPU: 1 PID: 4624 Comm: kworker/1:2 Not tainted 5.12.0-rc4-syzkaller #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> Workqueue: ipv6_addrconf addrconf_verify_work
> Call Trace:
> __dump_stack lib/dump_stack.c:79 [inline]
> dump_stack+0x141/0x1d7 lib/dump_stack.c:120
> ___might_sleep.cold+0x1f1/0x237 kernel/sched/core.c:8328
> __mutex_lock_common kernel/locking/mutex.c:928 [inline]
> __mutex_lock+0xa9/0x1120 kernel/locking/mutex.c:1096
> __ipv6_dev_mc_dec+0x5f/0x340 net/ipv6/mcast.c:970
> addrconf_leave_solict net/ipv6/addrconf.c:2182 [inline]
> addrconf_leave_solict net/ipv6/addrconf.c:2174 [inline]
> __ipv6_ifa_notify+0x5b6/0xa90 net/ipv6/addrconf.c:6077
> ipv6_ifa_notify net/ipv6/addrconf.c:6100 [inline]
> ipv6_del_addr+0x463/0xae0 net/ipv6/addrconf.c:1294
> addrconf_verify_rtnl+0xd59/0x1150 net/ipv6/addrconf.c:4488
> addrconf_verify_work+0xf/0x20 net/ipv6/addrconf.c:4573
> process_one_work+0x98d/0x1600 kernel/workqueue.c:2275
> worker_thread+0x64c/0x1120 kernel/workqueue.c:2421
> kthread+0x3b1/0x4a0 kernel/kthread.c:292
> ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294
>
I will test this fix:
diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index 120073ffb666b18678e3145d91dac59fa865a592..8f3883f4cb4a15a0749b8f0fe00061e483ea26ca 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -4485,7 +4485,9 @@ static void addrconf_verify_rtnl(void)
age >= ifp->valid_lft) {
spin_unlock(&ifp->lock);
in6_ifa_hold(ifp);
+ rcu_read_unlock_bh();
ipv6_del_addr(ifp);
+ rcu_read_lock_bh();
goto restart;
} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
spin_unlock(&ifp->lock);
