When the key is invalid there is no point in retrying. Because the auth
code returns kernel error codes only, we can't test on the DNR bit.

Signed-off-by: Daniel Wagner <dwagner@suse.de>
---
 drivers/nvme/host/auth.c    | 6 +++---
 drivers/nvme/host/fabrics.c | 6 ++++++
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/nvme/host/auth.c b/drivers/nvme/host/auth.c
index a264b3ae078b..371e14f0a203 100644
--- a/drivers/nvme/host/auth.c
+++ b/drivers/nvme/host/auth.c
@@ -730,7 +730,7 @@ static void nvme_queue_auth_work(struct work_struct *work)
 					 NVME_AUTH_DHCHAP_MESSAGE_CHALLENGE);
 	if (ret) {
 		chap->status = ret;
-		chap->error = -ECONNREFUSED;
+		chap->error = -EKEYREJECTED;
 		return;
 	}
 
@@ -797,7 +797,7 @@ static void nvme_queue_auth_work(struct work_struct *work)
 					 NVME_AUTH_DHCHAP_MESSAGE_SUCCESS1);
 	if (ret) {
 		chap->status = ret;
-		chap->error = -ECONNREFUSED;
+		chap->error = -EKEYREJECTED;
 		return;
 	}
 
@@ -818,7 +818,7 @@ static void nvme_queue_auth_work(struct work_struct *work)
 	ret = nvme_auth_process_dhchap_success1(ctrl, chap);
 	if (ret) {
 		/* Controller authentication failed */
-		chap->error = -ECONNREFUSED;
+		chap->error = -EKEYREJECTED;
 		goto fail2;
 	}
 
diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c
index 36d3e2ff27f3..c6ad2148c2e0 100644
--- a/drivers/nvme/host/fabrics.c
+++ b/drivers/nvme/host/fabrics.c
@@ -567,12 +567,18 @@ EXPORT_SYMBOL_GPL(nvmf_connect_io_queue);
  *
  * - the DNR bit is set and the specification states no further connect
  *   attempts with the same set of paramenters should be attempted.
+ *
+ * - when the authentication attempt fails, because the key was invalid.
+ *   This error code is set on the host side.
  */
 bool nvmf_should_reconnect(struct nvme_ctrl *ctrl, int status)
 {
 	if (status > 0 && (status & NVME_SC_DNR))
 		return false;
 
+	if (status == -EKEYREJECTED)
+		return false;
+
 	if (ctrl->opts->max_reconnects == -1 ||
 	    ctrl->nr_reconnects < ctrl->opts->max_reconnects)
 		return true;
-- 
2.44.0

On 30/04/2024 16:19, Daniel Wagner wrote:
> When the key is invalid there is no point in retrying. Because the auth
> code returns kernel error codes only, we can't test on the DNR bit.

I still think that this patch has issues in the sense that in certain cases
the controller can suddenly disappear.