[bug report] xen: add the Xenbus sysfs and virtual device hotplug driver

[Dan Carpenter <dan.carpenter@xxxxxxxxxx>]

Hello Jeremy Fitzhardinge,

The patch 4bac07c993d0: "xen: add the Xenbus sysfs and virtual device
hotplug driver" from Jul 17, 2007, leads to the following static
checker warning:

        drivers/xen/xenbus/xenbus_xs.c:125 xs_request_enter()
        warn: sleeping in atomic context

drivers/xen/xenbus/xenbus_client.c
   413  /**
   414   * Allocate an event channel for the given xenbus_device, assigning the 
newly
   415   * created local port to *port.  Return 0 on success, or -errno on 
error.  On
   416   * error, the device will switch to XenbusStateClosing, and the error 
will be
   417   * saved in the store.
           
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The documentation is very clear how it's supposed to handle the error.

   418   */
   419  int xenbus_alloc_evtchn(struct xenbus_device *dev, evtchn_port_t *port)
   420  {
   421          struct evtchn_alloc_unbound alloc_unbound;
   422          int err;
   423  
   424          alloc_unbound.dom = DOMID_SELF;
   425          alloc_unbound.remote_dom = dev->otherend_id;
   426  
   427          err = HYPERVISOR_event_channel_op(EVTCHNOP_alloc_unbound,
   428                                            &alloc_unbound);
   429          if (err)
   430                  xenbus_dev_fatal(dev, err, "allocating event channel");
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The xenbus_dev_fatal() has a bunch of sleeping functions like xenbus_write()
and we could be holding a spinlock here.

pvcalls_front_connect() <-- holds &bedata->socket_lock
 --> create_active()
     --> xenbus_alloc_evtchn()

   431          else
   432                  *port = alloc_unbound.port;
   433  
   434          return err;
   435  }

regards,
dan carpenter