On Tue, Jan 24, 2017 at 3:37 AM, Max Gurtovoy <maxg@xxxxxxxxxxxx> wrote: > Robert, > can you please try replacing the ib_drain_qp with ib_drain_rq ? or make sure > ib_drain_qp don't stuck. > in the past (before v4.6) we sent "recv_beacon" on the isert disconnect > flow. > I suspect that area. > > Max. It is getting stuck in target_wait_for_sess_cmds. ib_drain_qp is returning just fine. I put in a bunch of debug statements and here is a clip of them from a previous test. [Tue Jan 3 16:55:24 2017] Starting iscsit_close_connection. [Tue Jan 3 16:55:24 2017] Calling flush_workqueue. [Tue Jan 3 16:55:24 2017] Returned from flush_workqueue. [Tue Jan 3 16:55:25 2017] isert_wait_conn calling ib_close_qp/ib_drain_qp. [Tue Jan 3 16:55:25 2017] isert_wait_conn finished ib_close_qp/ib_drain_qp. [Tue Jan 3 16:55:25 2017] isert_wait_conn calling isert_put_unsol_pending_cmds. [Tue Jan 3 16:55:25 2017] isert_wait_conn returned from isert_put_unsol_pending_cmds. [Tue Jan 3 16:55:25 2017] isert_wait_conn calling isert_wait4cmds. [Tue Jan 3 16:55:25 2017] isert_wait4cmds calling target_sess_cmd_list_set_waiting. [Tue Jan 3 16:55:25 2017] isert_wait4cmds returned from target_sess_cmd_list_set_waiting. [Tue Jan 3 16:55:25 2017] isert_wait4cmds calling target_wait_for_sess_cmds. [Tue Jan 3 16:55:25 2017] Waiting for se_cmd: ffff887ebf88bd00 t_state: 6, fabric state: 29 [Tue Jan 3 16:55:25 2017] target_wait_for_sess_cmds calling spin_unlock_irqrestore. [Tue Jan 3 16:55:25 2017] target_wait_for_sess_cmds calling wait_for_completion. [Tue Jan 3 16:55:26 2017] Calling wait_for_common. [Tue Jan 3 16:55:27 2017] Starting __wait_for_common. [Tue Jan 3 16:55:27 2017] Calling do_wait_for_common. [Tue Jan 3 16:55:27 2017] Starting do_wait_for_common with x->done = 0. [Tue Jan 3 16:55:27 2017] x->done = 0. [Tue Jan 3 16:55:27 2017] Calling __set_current_state. [Tue Jan 3 16:55:27 2017] Calling spin_unlock_irq. [Tue Jan 3 16:55:27 2017] Calling action(7fffffffffffffff). [Tue Jan 3 16:55:27 2017] Starting schedule_timeout with timeout 7fffffffffffffff (7fffffffffffffff). [Tue Jan 3 16:55:27 2017] Starting schedule() in switch. [Tue Jan 3 16:55:27 2017] Finished schedule() in switch. [Tue Jan 3 16:55:27 2017] Calling spin_lock_irq. [Tue Jan 3 16:55:27 2017] Calling __remove_wait_queue. [Tue Jan 3 16:55:27 2017] Returned from __remove_wait_queue. [Tue Jan 3 16:55:27 2017] Finished do_wait_for_common with x->done = 0. [Tue Jan 3 16:55:27 2017] Returned from do_wait_for_common. [Tue Jan 3 16:55:27 2017] Finished __wait_for_common. [Tue Jan 3 16:55:27 2017] Returned from wait_for_common. [Tue Jan 3 16:55:27 2017] Calling wait_for_common. [Tue Jan 3 16:55:27 2017] Starting __wait_for_common. [Tue Jan 3 16:55:27 2017] Calling do_wait_for_common. [Tue Jan 3 16:55:27 2017] Starting do_wait_for_common with x->done = 0. [Tue Jan 3 16:55:27 2017] x->done = 0. [Tue Jan 3 16:55:27 2017] Calling __set_current_state. [Tue Jan 3 16:55:27 2017] Calling spin_unlock_irq. [Tue Jan 3 16:55:27 2017] Calling action(7fffffffffffffff). [Tue Jan 3 16:55:27 2017] Starting schedule_timeout with timeout 7fffffffffffffff (7fffffffffffffff). [Tue Jan 3 16:55:27 2017] Starting schedule() in switch. [Tue Jan 3 16:55:27 2017] Finished schedule() in switch. [Tue Jan 3 16:55:27 2017] Calling spin_lock_irq. [Tue Jan 3 16:55:27 2017] Calling __remove_wait_queue. If you think changing ib_drain_qp to ib_drain_rq might still help, I can do it. Thanks, ---------------- Robert LeBlanc PGP Fingerprint 79A2 9CA4 6CC4 45DD A904 C70E E654 3BB2 FA62 B9F1 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
