[PATCH 3/5] nvme-tcp: add clean action for failed reconnection

Chao Leng lengchao at huawei.com
Wed Jan 20 22:32:38 EST 2021


If reconnect failed after start io queues, the queues will be unquiesced
and new requests continue to be delivered. Reconnection error handling
process directly free queues without cancel suspend requests. The
suppend request will time out, and then crash due to use the queue
after free.

Add sync queues and cancel suppend requests for reconnection error
handling.

Signed-off-by: Chao Leng <lengchao at huawei.com>
---
 drivers/nvme/host/tcp.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 216619926563..7e139d7bd279 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1809,8 +1809,10 @@ static int nvme_tcp_configure_io_queues(struct nvme_ctrl *ctrl, bool new)
 
 out_wait_freeze_timed_out:
 	nvme_stop_queues(ctrl);
+	nvme_sync_io_queues(ctrl);
 	nvme_tcp_stop_io_queues(ctrl);
 out_cleanup_connect_q:
+	nvme_cancel_tagset(ctrl);
 	if (new)
 		blk_cleanup_queue(ctrl->connect_q);
 out_free_tag_set:
@@ -1872,12 +1874,16 @@ static int nvme_tcp_configure_admin_queue(struct nvme_ctrl *ctrl, bool new)
 
 	error = nvme_init_identify(ctrl);
 	if (error)
-		goto out_stop_queue;
+		goto out_quiesce_queue;
 
 	return 0;
 
+out_quiesce_queue:
+	blk_mq_quiesce_queue(ctrl->admin_q);
+	blk_sync_queue(ctrl->admin_q);
 out_stop_queue:
 	nvme_tcp_stop_queue(ctrl, 0);
+	nvme_cancel_admin_tagset(ctrl);
 out_cleanup_queue:
 	if (new)
 		blk_cleanup_queue(ctrl->admin_q);
@@ -1997,10 +2003,18 @@ static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new)
 	return 0;
 
 destroy_io:
-	if (ctrl->queue_count > 1)
+	if (ctrl->queue_count > 1) {
+		nvme_stop_queues(ctrl);
+		nvme_sync_io_queues(ctrl);
+		nvme_tcp_stop_io_queues(ctrl);
+		nvme_cancel_tagset(ctrl);
 		nvme_tcp_destroy_io_queues(ctrl, new);
+	}
 destroy_admin:
+	blk_mq_quiesce_queue(ctrl->admin_q);
+	blk_sync_queue(ctrl->admin_q);
 	nvme_tcp_stop_queue(ctrl, 0);
+	nvme_cancel_admin_tagset(ctrl);
 	nvme_tcp_destroy_admin_queue(ctrl, new);
 	return ret;
 }
-- 
2.16.4




More information about the Linux-nvme mailing list