For the tasks waiting in exp_wq inside exp_funnel_lock(),
there is a chance that they might be indefinitely blocked
in below scenario:
1. There is a task waiting on exp sequence 0b'100' inside
exp_funnel_lock(). This task blocks at wq index 1.
synchronize_rcu_expedited()
s = 0b'100'
exp_funnel_lock()
wait_event(rnp->exp_wq[rcu_seq_ctr(s) & 0x3]
2. The expedited grace period (which above task blocks for)
completes and task (task1) holding exp_mutex queues
worker and schedules out.
synchronize_rcu_expedited()
s = 0b'100'
queue_work(rcu_gp_wq, &rew.rew_work)
wake_up_worker()
schedule()
3. kworker A picks up the queued work and completes the exp gp
sequence and then blocks on exp_wake_mutex, which is held
by another kworker, which is doing wakeups for expedited_sequence
0.
rcu_exp_wait_wake()
rcu_exp_wait_wake()
rcu_exp_gp_seq_end(rsp) // rsp->expedited_sequence is incremented
// to 0b'100'
mutex_lock(&rcu_state.exp_wake_mutex)
4. task1 does not enter wait queue, as sync_exp_work_done() returns true,
and releases exp_mutex.
wait_event(rnp->exp_wq[rcu_seq_ctr(s) & 0x3],
sync_exp_work_done(rsp, s));
mutex_unlock(&rsp->exp_mutex);
5. Next exp GP completes, and sequence number is incremented:
rcu_exp_wait_wake()
rcu_exp_wait_wake()
rcu_exp_gp_seq_end(rsp) // rsp->expedited_sequence = 0b'200'
6. kworker A acquires exp_wake_mutex. As it uses current
expedited_sequence, it wakes up workers from wrong wait queue
index - it should have worken wait queue corresponding to
0b'100' sequence, but wakes up the ones for 0b'200' sequence.
This results in task at step 1 indefinitely blocked.
rcu_exp_wait_wake()
wake_up_all(&rnp->exp_wq[rcu_seq_ctr(rsp->expedited_sequence) & 0x3]);
This issue manifested as DPM device timeout during suspend, as scsi
device was stuck in _synchronize_rcu_expedited().
schedule()
synchronize_rcu_expedited()
synchronize_rcu()
scsi_device_quiesce()
scsi_bus_suspend()
dpm_run_callback()
__device_suspend()
Fix this by using the correct exp sequence number, the one which
owner of the exp_mutex initiated and passed to kworker,
to index the wait queue, inside rcu_exp_wait_wake().
Signed-off-by: Neeraj Upadhyay <neeraju@xxxxxxxxxxxxxx>
---
Changes since v1:
- Updates the commit log with failure information.
kernel/rcu/tree_exp.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h
index e4b77d3..28979d3 100644
--- a/kernel/rcu/tree_exp.h
+++ b/kernel/rcu/tree_exp.h
@@ -557,7 +557,7 @@ static void rcu_exp_wait_wake(unsigned long s)
spin_unlock(&rnp->exp_lock);
}
smp_mb(); /* All above changes before wakeup. */
- wake_up_all(&rnp->exp_wq[rcu_seq_ctr(rcu_state.expedited_sequence) & 0x3]);
+ wake_up_all(&rnp->exp_wq[rcu_seq_ctr(s) & 0x3]);
}
trace_rcu_exp_grace_period(rcu_state.name, s, TPS("endwake"));
mutex_unlock(&rcu_state.exp_wake_mutex);
--
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a
member of the Code Aurora Forum, hosted by The Linux Foundation
