Re: [PATCH 02/10] scsi: iscsi: Fix offload conn cleanup when iscsid restarts

[Chris Leech <cleech@redhat.com>]

On Thu, Apr 07, 2022 at 07:13:06PM -0500, Mike Christie wrote:
> When userspace restarts during boot or upgrades it won't know about the
> offload driver's endpoint and connection mappings. iscsid will start by
> cleaning up the old session by doing a stop_conn call. Later if we are
> able to create a new connection, we cleanup the old endpoint during the
> binding stage. The problem is that if we do stop_conn before doing the
> ep_disconnect call offload drivers can still be executing IO. We then
> might free tasks from the under the card/driver.
> 
> This moves the ep_disconnect call to before we do the stop_conn call for
> this case. It will then work and look like a normal recovery/cleanup
> procedure from the driver's point of view.
> 
> Signed-off-by: Mike Christie <michael.christie@oracle.com>
> ---
>  drivers/scsi/scsi_transport_iscsi.c | 48 +++++++++++++++++------------
>  1 file changed, 28 insertions(+), 20 deletions(-)

Reviewed-by: Chris Leech <cleech@redhat.com>
 
> diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c
> index 4e10457e3ab9..bf39fb5569b6 100644
> --- a/drivers/scsi/scsi_transport_iscsi.c
> +++ b/drivers/scsi/scsi_transport_iscsi.c
> @@ -2236,6 +2236,23 @@ static void iscsi_ep_disconnect(struct iscsi_cls_conn *conn, bool is_active)
>  	ISCSI_DBG_TRANS_CONN(conn, "disconnect ep done.\n");
>  }
>  
> +static void iscsi_if_disconnect_bound_ep(struct iscsi_cls_conn *conn,
> +					 struct iscsi_endpoint *ep,
> +					 bool is_active)
> +{
> +	/* Check if this was a conn error and the kernel took ownership */
> +	if (!test_bit(ISCSI_CLS_CONN_BIT_CLEANUP, &conn->flags)) {
> +		iscsi_ep_disconnect(conn, is_active);
> +	} else {
> +		ISCSI_DBG_TRANS_CONN(conn, "flush kernel conn cleanup.\n");
> +		mutex_unlock(&conn->ep_mutex);
> +
> +		flush_work(&conn->cleanup_work);
> +
> +		mutex_lock(&conn->ep_mutex);
> +	}
> +}
> +
>  static int iscsi_if_stop_conn(struct iscsi_transport *transport,
>  			      struct iscsi_uevent *ev)
>  {
> @@ -2256,6 +2273,16 @@ static int iscsi_if_stop_conn(struct iscsi_transport *transport,
>  		cancel_work_sync(&conn->cleanup_work);
>  		iscsi_stop_conn(conn, flag);
>  	} else {
> +		/*
> +		 * For offload, when iscsid is restarted it won't know about
> +		 * existing endpoints so it can't do a ep_disconnect. We clean
> +		 * it up here for userspace.
> +		 */
> +		mutex_lock(&conn->ep_mutex);
> +		if (conn->ep)
> +			iscsi_if_disconnect_bound_ep(conn, conn->ep, true);
> +		mutex_unlock(&conn->ep_mutex);
> +
>  		/*
>  		 * Figure out if it was the kernel or userspace initiating this.
>  		 */
> @@ -2984,16 +3011,7 @@ static int iscsi_if_ep_disconnect(struct iscsi_transport *transport,
>  	}
>  
>  	mutex_lock(&conn->ep_mutex);
> -	/* Check if this was a conn error and the kernel took ownership */
> -	if (test_bit(ISCSI_CLS_CONN_BIT_CLEANUP, &conn->flags)) {
> -		ISCSI_DBG_TRANS_CONN(conn, "flush kernel conn cleanup.\n");
> -		mutex_unlock(&conn->ep_mutex);
> -
> -		flush_work(&conn->cleanup_work);
> -		goto put_ep;
> -	}
> -
> -	iscsi_ep_disconnect(conn, false);
> +	iscsi_if_disconnect_bound_ep(conn, ep, false);
>  	mutex_unlock(&conn->ep_mutex);
>  put_ep:
>  	iscsi_put_endpoint(ep);
> @@ -3704,16 +3722,6 @@ static int iscsi_if_transport_conn(struct iscsi_transport *transport,
>  
>  	switch (nlh->nlmsg_type) {
>  	case ISCSI_UEVENT_BIND_CONN:
> -		if (conn->ep) {
> -			/*
> -			 * For offload boot support where iscsid is restarted
> -			 * during the pivot root stage, the ep will be intact
> -			 * here when the new iscsid instance starts up and
> -			 * reconnects.
> -			 */
> -			iscsi_ep_disconnect(conn, true);
> -		}
> -
>  		session = iscsi_session_lookup(ev->u.b_conn.sid);
>  		if (!session) {
>  			err = -EINVAL;
> -- 
> 2.25.1
>