From: Sagi Grimberg <sagi@grimberg.me>
Date: Wed, 11 Oct 2017 15:29:11 +0300
Subject: nvme-rdma: teardown admin/io queues once on error recovery
Patch-mainline: v4.15-rc1
Git-commit: 5e1fe61d4170b1f498e20de92c7ce4cd5e40c3c5
References: FATE#323952, FATE#322506
Relying on the queue state while tearing down on every reconnect
attempt is not a good design. We should do it once in err_work
and simply try to establish the queues for each reconnect attempt.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Johannes Thumshirn <jthumshirn@suse.de>
---
drivers/nvme/host/rdma.c | 27 ++++++++++++---------------
1 file changed, 12 insertions(+), 15 deletions(-)
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -927,10 +927,6 @@ static void nvme_rdma_reconnect_ctrl_wor
++ctrl->ctrl.nr_reconnects;
- if (ctrl->ctrl.queue_count > 1)
- nvme_rdma_destroy_io_queues(ctrl, false);
-
- nvme_rdma_destroy_admin_queue(ctrl, false);
ret = nvme_rdma_configure_admin_queue(ctrl, false);
if (ret)
goto requeue;
@@ -938,7 +934,7 @@ static void nvme_rdma_reconnect_ctrl_wor
if (ctrl->ctrl.queue_count > 1) {
ret = nvme_rdma_configure_io_queues(ctrl, false);
if (ret)
- goto requeue;
+ goto destroy_admin;
}
changed = nvme_change_ctrl_state(&ctrl->ctrl, NVME_CTRL_LIVE);
@@ -948,14 +944,17 @@ static void nvme_rdma_reconnect_ctrl_wor
return;
}
- ctrl->ctrl.nr_reconnects = 0;
-
nvme_start_ctrl(&ctrl->ctrl);
- dev_info(ctrl->ctrl.device, "Successfully reconnected\n");
+ dev_info(ctrl->ctrl.device, "Successfully reconnected (%d attempts)\n",
+ ctrl->ctrl.nr_reconnects);
+
+ ctrl->ctrl.nr_reconnects = 0;
return;
+destroy_admin:
+ nvme_rdma_destroy_admin_queue(ctrl, false);
requeue:
dev_info(ctrl->ctrl.device, "Failed reconnect attempt %d\n",
ctrl->ctrl.nr_reconnects);
@@ -971,17 +970,15 @@ static void nvme_rdma_error_recovery_wor
if (ctrl->ctrl.queue_count > 1) {
nvme_stop_queues(&ctrl->ctrl);
- nvme_rdma_stop_io_queues(ctrl);
- }
- blk_mq_quiesce_queue(ctrl->ctrl.admin_q);
- nvme_rdma_stop_queue(&ctrl->queues[0]);
-
- /* We must take care of fastfail/requeue all our inflight requests */
- if (ctrl->ctrl.queue_count > 1)
blk_mq_tagset_busy_iter(&ctrl->tag_set,
nvme_cancel_request, &ctrl->ctrl);
+ nvme_rdma_destroy_io_queues(ctrl, false);
+ }
+
+ blk_mq_quiesce_queue(ctrl->ctrl.admin_q);
blk_mq_tagset_busy_iter(&ctrl->admin_tag_set,
nvme_cancel_request, &ctrl->ctrl);
+ nvme_rdma_destroy_admin_queue(ctrl, false);
/*
* queues are not a live anymore, so restart the queues to fail fast