Re: [RFC v1 4/4] nvmet-discovery: do not use invalid port

[Daniel Wagner]

On Wed, Sep 13, 2023 at 01:35:19PM +0200, Christoph Hellwig wrote:
> So that's interesting.  But what I'm mostly worried about is how the
> nvmet kernel code allows a request without ->port to progress to the
> actual command handler.

nvmet_fc_handle_fcp_rqst()

	if (tgtport->pe)
		fod->req.port = tgtport->pe->port;

Not sure why this is there. Will test what happens when we just return
an error when we don't have pe set.

> We should never allow a command to get that
> far if ->port is NULL, and should not allow to clear ->port while
> commands are still handled.

Okay, makes sense. I'll test this when I have access to my rig again tomorrow.