net: linkwatch: use system_unbound_wq

linkwatch_event() grabs possibly very contended RTNL mutex.

system_wq is not suitable for such work.

Inspired by many noisy syzbot reports.

3 locks held by kworker/0:7/5266:
 #0: ffff888015480948 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work kernel/workqueue.c:3206 [inline]
 #0: ffff888015480948 ((wq_completion)events){+.+.}-{0:0}, at: process_scheduled_works+0x90a/0x1830 kernel/workqueue.c:3312
 #1: ffffc90003f6fd00 ((linkwatch_work).work){+.+.}-{0:0}, at: process_one_work kernel/workqueue.c:3207 [inline]
 , at: process_scheduled_works+0x945/0x1830 kernel/workqueue.c:3312
 #2: ffffffff8fa6f208 (rtnl_mutex){+.+.}-{3:3}, at: linkwatch_event+0xe/0x60 net/core/link_watch.c:276

Reported-by: syzbot <[email protected]>
Fixes: 1da177e ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <[email protected]>
Reviewed-by: Kuniyuki Iwashima <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net/core/link_watch.c

@@ -148,9 +148,9 @@ static void linkwatch_schedule_work(int urgent)
 	 * override the existing timer.
 	 */
 	if (test_bit(LW_URGENT, &linkwatch_flags))
-		mod_delayed_work(system_wq, &linkwatch_work, 0);
+		mod_delayed_work(system_unbound_wq, &linkwatch_work, 0);
 	else
-		schedule_delayed_work(&linkwatch_work, delay);
+		queue_delayed_work(system_unbound_wq, &linkwatch_work, delay);
 }