the nvme_handle_cqe() interrupt handler calls nvme_complete_async_event()
but the latter may call nvme_auth_stop() which is a blocking function.
Sleeping functions can't be called in interrupt context
 BUG: sleeping function called from invalid context
 in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 0, name: swapper/15
  Call Trace:
     <IRQ>
      __cancel_work_timer+0x31e/0x460
      ? nvme_change_ctrl_state+0xcf/0x3c0 [nvme_core]
      ? nvme_change_ctrl_state+0xcf/0x3c0 [nvme_core]
      nvme_complete_async_event+0x365/0x480 [nvme_core]
      nvme_poll_cq+0x262/0xe50 [nvme]
Fix the bug by moving nvme_auth_stop() to fw_act_work
(executed by the nvme_wq workqueue)
Fixes: f50fff73d620 ("nvme: implement In-Band authentication")
Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Keith Busch <kbusch@kernel.org>
                                struct nvme_ctrl, fw_act_work);
        unsigned long fw_act_timeout;
 
+       nvme_auth_stop(ctrl);
+
        if (ctrl->mtfa)
                fw_act_timeout = jiffies +
                                msecs_to_jiffies(ctrl->mtfa * 100);
                 * firmware activation.
                 */
                if (nvme_change_ctrl_state(ctrl, NVME_CTRL_RESETTING)) {
-                       nvme_auth_stop(ctrl);
                        requeue = false;
                        queue_work(nvme_wq, &ctrl->fw_act_work);
                }