Re: [Patch 3/3] IB/ib_cm: hang in cm_destroy_id during PCI error injection

[Carol Soto]

On 4/23/2014 1:15 PM, Hefty, Sean wrote:
Index: b/drivers/infiniband/core/cm.c
===================================================================
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -161,6 +161,7 @@ struct cm_port {
        struct ib_mad_agent *mad_agent;
        struct kobject port_obj;
        u8 port_num;
+       u8 device_fatal;
        struct cm_counter_group counter_group[CM_COUNTER_GROUPS];
  };
@@ -168,6 +169,7 @@ struct cm_device {
        struct list_head list;
        struct ib_device *ib_device;
        struct device *device;
+       struct ib_event_handler event_handler;
        u8 ack_delay;
        struct cm_port *port[0];
  };
@@ -258,6 +260,10 @@ static int cm_alloc_msg(struct cm_id_pri
        struct ib_mad_agent *mad_agent;
        struct ib_mad_send_buf *m;
        struct ib_ah *ah;
+       struct cm_port *port = cm_id_priv->av.port;
+
+       if (port->device_fatal)
+               return -EIO;

        mad_agent = cm_id_priv->av.port->mad_agent;
        ah = ib_create_ah(mad_agent->qp->pd, &cm_id_priv->av.ah_attr);
@@ -290,6 +296,9 @@ static int cm_alloc_response_msg(struct
        struct ib_mad_send_buf *m;
        struct ib_ah *ah;

+       if (port->device_fatal)
+               return -EIO;
+
        ah = ib_create_ah_from_wc(port->mad_agent->qp->pd, mad_recv_wc->wc,
                                  mad_recv_wc->recv_buf.grh, port->port_num);
        if (IS_ERR(ah))
@@ -3764,6 +3773,33 @@ static void cm_remove_port_fs(struct cm_
        kobject_put(&port->port_obj);
  }

+static void ib_cm_event_handler(struct ib_event_handler *handler,
+                               struct ib_event *event)
+{
+       struct cm_device *cm_dev;
+       struct cm_port *port = NULL;
+
+       cm_dev = container_of(handler, struct cm_device, event_handler);
+       switch (event->event) {
+       case IB_EVENT_PORT_ACTIVE:
+               port = cm_dev->port[event->element.port_num - 1];
+               if (port == NULL)
+                       return;
+               if (port->port_num == event->element.port_num)
+                       port->device_fatal = 0;
+               break;
+       case IB_EVENT_DEVICE_FATAL:
+               port = cm_dev->port[event->element.port_num - 1];
+               if (port == NULL)
+                       return;
+               if (port->port_num == event->element.port_num)
+                       port->device_fatal = 1;
+               break;
This is a device level event, not a port event.  The port_num value may not be 
valid.
The first patch of the series(IB/mlx4: send a IB_EVENT_DEVICE_FATAL to 
users during PCI error injection), when mlx4 send the IB_EVENT_DEVICE_FATAL
event I am passing the port number to that field so in that way I was 
able to make it work. Any other suggestions are welcome.

Thanks
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html



--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html