[PATCH] nvmet-rdma: Invoke fatal error on error completion

[hch@infradead.org (Christoph Hellwig)]

On Thu, Jun 23, 2016@08:01:30PM +0300, Sagi Grimberg wrote:
> In case we got an error completion the rdma queue pair
> is in error state, teardown the entire controller. Note
> that in recv or read error completion we might not have
> a controller yet, so check for the controller exsistence.
> 
> Signed-off-by: Sagi Grimberg <sagi at grimberg.me>

This looks fine minus a few minor codingstyle nitpicks that I'd be
happy to fix up:

> +static void nvmet_rdma_error_comp(struct nvmet_rdma_queue *queue)
> +{
> +	if (queue->nvme_sq.ctrl)
> +		nvmet_ctrl_fatal_error(queue->nvme_sq.ctrl);
> +	else
> +		/*
> +		 * we didn't setup the controller yet in case
> +		 * of admin connect error, just disconnect and
> +		 * cleanup the queue
> +		 */
> +		nvmet_rdma_queue_disconnect(queue);
> +}

With such a long comment I'd prefer to have curly braces just to make
the else visually more obvious
> +
> +	if (unlikely(wc->status != IB_WC_SUCCESS &&
> +		wc->status != IB_WC_WR_FLUSH_ERR)) {

Indenting the second line of a condition by a single tab is always wrong,
either indent it with two tabs, or so that it aligns with first line.
The second is probably nicer here:

	if (unlikely(wc->status != IB_WC_SUCCESS &&
		     wc->status != IB_WC_WR_FLUSH_ERR)) {

Given how many !success not !flush_err conditionals we have in various
drivers I wonder if we should have a helper in the RDMA core, though.