On 9/27/21 8:48 AM, Hannes Reinecke wrote:

On 9/27/21 12:53 AM, Sagi Grimberg wrote:

quoted

quoted

+/* Assumes that the controller is in state RESETTING */
+static void nvme_dhchap_auth_work(struct work_struct *work)
+{
+    struct nvme_ctrl *ctrl =
+        container_of(work, struct nvme_ctrl, dhchap_auth_work);
+    int ret, q;
+

Here I would print a single:
     dev_info(ctrl->device, "re-authenticating controller");

This is instead of all the queue re-authentication prints that
should be dev_dbg.

Let's avoid doing the per-queue print...

Hmm. Actually the spec allows to use different keys per queue, even 
though our implementation doesn't. And fmds has struggled to come up 
with a sane usecase for that.

We don't need to support it, but regardless we should not
info print per-queue.

But yes, okay, will be updating it.

Great...