[PATCH 2/2] nvme-rdma: short-circuit connect retries

Hannes Reinecke hare at suse.de
Thu Jul 14 05:41:33 PDT 2022


When a reconnect attempt fails with a non-retryable status
(eg when the subsystem has been unprovisioned) there hardly
is any reason to retry the reconnect attempt.
So pass the actual error status to nvme_tcp_reconnect_or_remove()
and short-circuit retries if the DNR bit is set.

Signed-off-by: Hannes Reinecke <hare at suse.de>
---
 drivers/nvme/host/rdma.c | 25 ++++++++++++++++++-------
 1 file changed, 18 insertions(+), 7 deletions(-)

diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index 84ce3347d158..bcc84f181dcd 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -1065,8 +1065,10 @@ static void nvme_rdma_free_ctrl(struct nvme_ctrl *nctrl)
 	kfree(ctrl);
 }
 
-static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl)
+static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl, int status)
 {
+	bool recon = true;
+
 	/* If we are resetting/deleting then do nothing */
 	if (ctrl->ctrl.state != NVME_CTRL_CONNECTING) {
 		WARN_ON_ONCE(ctrl->ctrl.state == NVME_CTRL_NEW ||
@@ -1074,7 +1076,12 @@ static void nvme_rdma_reconnect_or_remove(struct nvme_rdma_ctrl *ctrl)
 		return;
 	}
 
-	if (nvmf_should_reconnect(&ctrl->ctrl)) {
+	if (status > 0 && (status & NVME_SC_DNR)) {
+		dev_info(ctrl->ctrl.device, "reconnect failure %d\n", status);
+		recon = false;
+	}
+
+	if (recon && nvmf_should_reconnect(&ctrl->ctrl)) {
 		dev_info(ctrl->ctrl.device, "Reconnecting in %d seconds...\n",
 			ctrl->ctrl.opts->reconnect_delay);
 		queue_delayed_work(nvme_wq, &ctrl->reconnect_work,
@@ -1173,10 +1180,12 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
 {
 	struct nvme_rdma_ctrl *ctrl = container_of(to_delayed_work(work),
 			struct nvme_rdma_ctrl, reconnect_work);
+	int ret;
 
 	++ctrl->ctrl.nr_reconnects;
 
-	if (nvme_rdma_setup_ctrl(ctrl, false))
+	ret = nvme_rdma_setup_ctrl(ctrl, false);
+	if (ret)
 		goto requeue;
 
 	dev_info(ctrl->ctrl.device, "Successfully reconnected (%d attempts)\n",
@@ -1189,7 +1198,7 @@ static void nvme_rdma_reconnect_ctrl_work(struct work_struct *work)
 requeue:
 	dev_info(ctrl->ctrl.device, "Failed reconnect attempt %d\n",
 			ctrl->ctrl.nr_reconnects);
-	nvme_rdma_reconnect_or_remove(ctrl);
+	nvme_rdma_reconnect_or_remove(ctrl, ret);
 }
 
 static void nvme_rdma_error_recovery_work(struct work_struct *work)
@@ -1212,7 +1221,7 @@ static void nvme_rdma_error_recovery_work(struct work_struct *work)
 		return;
 	}
 
-	nvme_rdma_reconnect_or_remove(ctrl);
+	nvme_rdma_reconnect_or_remove(ctrl, -ENOTCONN);
 }
 
 static void nvme_rdma_error_recovery(struct nvme_rdma_ctrl *ctrl)
@@ -2274,6 +2283,7 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
 {
 	struct nvme_rdma_ctrl *ctrl =
 		container_of(work, struct nvme_rdma_ctrl, ctrl.reset_work);
+	int ret;
 
 	nvme_stop_ctrl(&ctrl->ctrl);
 	nvme_rdma_shutdown_ctrl(ctrl, false);
@@ -2284,14 +2294,15 @@ static void nvme_rdma_reset_ctrl_work(struct work_struct *work)
 		return;
 	}
 
-	if (nvme_rdma_setup_ctrl(ctrl, false))
+	ret = nvme_rdma_setup_ctrl(ctrl, false);
+	if (ret)
 		goto out_fail;
 
 	return;
 
 out_fail:
 	++ctrl->ctrl.nr_reconnects;
-	nvme_rdma_reconnect_or_remove(ctrl);
+	nvme_rdma_reconnect_or_remove(ctrl, ret);
 }
 
 static const struct nvme_ctrl_ops nvme_rdma_ctrl_ops = {
-- 
2.29.2




More information about the Linux-nvme mailing list