nvmet: switch loopback target state to connecting when resetting

After commit bb06ec31452f ("nvme: expand nvmf_check_if_ready checks")
resetting of the loopback nvme target failed as we forgot to switch
it's state to NVME_CTRL_CONNECTING before we reconnect the admin
queues. Therefore the checks in nvmf_check_if_ready() choose to go to
the reject_io case and thus we couldn't sent out an identify
controller command to reconnect.

Change the controller state to NVME_CTRL_CONNECTING after tearing down
the old connection and before re-establishing the connection.

Fixes: bb06ec31452f ("nvme: expand nvmf_check_if_ready checks")
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/drivers/nvme/target/loop.c b/drivers/nvme/target/loop.c
index 31fdfba556a844958ee19933f33f2b5f3ed378f3..27a8561c0cb97596d25b53620493978040082124 100644 (file)
--- a/drivers/nvme/target/loop.c
+++ b/drivers/nvme/target/loop.c
@@ -469,6 +469,12 @@ static void nvme_loop_reset_ctrl_work(struct work_struct *work)
        nvme_stop_ctrl(&ctrl->ctrl);
        nvme_loop_shutdown_ctrl(ctrl);
 
+       if (!nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_CONNECTING)) {
+               /* state change failure should never happen */
+               WARN_ON_ONCE(1);
+               return;
+       }
+
        ret = nvme_loop_configure_admin_queue(ctrl);
        if (ret)
                goto out_disable;