On (07/14/11 18:41), Peter Zijlstra wrote: > On Thu, 2011-07-14 at 17:49 +0300, Sergey Senozhatsky wrote: > > Hello, > > > > [ 4172.371094] ======================================================= > > [ 4172.371097] [ INFO: possible circular locking dependency detected ] > > [ 4172.371103] 3.0.0-rc7-dbg-00565-g6b31efc #644 > > [ 4172.371105] ------------------------------------------------------- > > [ 4172.371108] make/11702 is trying to acquire lock: > > [ 4172.371111] (rcu_node_level_0){..-...}, at: [] __rcu_read_unlock+0xb9/0x217 > > [ 4172.371123] > > [ 4172.371123] but task is already holding lock: > > [ 4172.371125] (&p->pi_lock){-.-.-.}, at: [] try_to_wake_up+0x29/0x28b > > [ 4172.371135] > > [ 4172.371135] which lock already depends on the new lock. > > Happen to have a .config and means of reproduction? We're a little > stumped at trying to explain how this can actually happen. > > Hello, .config is attached. Not sure can reproduce this easily. Sorry, I'm leaving for a small holidays and unlikely will be able to compile and test within next 5-6 days. The above traces were spotted during kernel compilation process (make -j8). Sergey