On Friday, 12 January 2007 at 19:33, Julian Stecklina wrote:
> Hello,
>
> in an application I develop, event channel semantics seem a bit odd to
> me. Consider the following scenario: A Mini-OS binds an unbound port for
> domain 0 and issues a notification every second:
>
> evtchn_alloc_unbound(0, msg_handler, NULL,
> &port);
>
> printk("%u %u %u\n", domid, port);
>
> while (1) {
> sleep(1000);
> notify_remote_via_evtchn(port);
> }
>
> A Linux program running on Domain 0 calls bind_interdomain using domid
> and port to obtain local_port and watches for pending notifications:
>
> evtchn_port_t local_port = xc_evtchn_bind_interdomain(ev, domid,
> port);
>
> while (1) {
> evtchn_port_t port;
>
> port = xc_evtchn_pending(ev);
> printf("%u\n", port);
> xc_evtchn_unmask(ev, port);
> }
>
> Now what I do not understand is, why the Linux program sees exactly one
> notification and then indefinitely blocks. Any help on what I may be
> doing wrong is greatly appreciated. This is with xen-3.0.4-testing.
According to the interface guide
http://www.cl.cam.ac.uk/research/srg/netos/xen/readmes/interface/interface.html#SECTION00600000000000000000
the event is masked until the receiver clears the notification. I
think you do this by sending a notify back along the same port:
while (1) {
evtchn_port_t port;
port = xc_evtchn_pending(ev);
printf("%u\n", port);
xc_evtchn_notify(ev, port);
xc_evtchn_unmask(ev, port);
}
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|