Skip to content

Commit

Permalink
xen: do not set xenstored_ready before xenbus_probe on hvm
Browse files Browse the repository at this point in the history
Register_xenstore_notifier should guarantee that the caller gets
notified even if xenstore is already up.
Therefore we revert "do not notify callers from
register_xenstore_notifier" and set xenstored_read at the right time for
PV on HVM guests too.
In fact in case of PV on HVM guests xenstored is ready only after the
platform pci driver has completed the initialization, so do not set
xenstored_ready before the call to xenbus_probe().

This patch fixes a shutdown_event watcher registration bug that causes
"xm shutdown" not to work properly.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Jeremy Fitzhardinge <jeremy@goop.org>
  • Loading branch information
Stefano Stabellini committed Oct 5, 2010
1 parent 899611e commit a947f0f
Showing 1 changed file with 6 additions and 3 deletions.
9 changes: 6 additions & 3 deletions drivers/xen/xenbus/xenbus_probe.c
Original file line number Diff line number Diff line change
Expand Up @@ -755,7 +755,10 @@ int register_xenstore_notifier(struct notifier_block *nb)
{
int ret = 0;

blocking_notifier_chain_register(&xenstore_chain, nb);
if (xenstored_ready > 0)
ret = nb->notifier_call(nb, 0, NULL);
else
blocking_notifier_chain_register(&xenstore_chain, nb);

return ret;
}
Expand All @@ -769,7 +772,7 @@ EXPORT_SYMBOL_GPL(unregister_xenstore_notifier);

void xenbus_probe(struct work_struct *unused)
{
BUG_ON((xenstored_ready <= 0));
xenstored_ready = 1;

/* Enumerate devices in xenstore and watch for changes. */
xenbus_probe_devices(&xenbus_frontend);
Expand Down Expand Up @@ -835,8 +838,8 @@ static int __init xenbus_init(void)
xen_store_evtchn = xen_start_info->store_evtchn;
xen_store_mfn = xen_start_info->store_mfn;
xen_store_interface = mfn_to_virt(xen_store_mfn);
xenstored_ready = 1;
}
xenstored_ready = 1;
}

/* Initialize the interface to xenstore. */
Expand Down

0 comments on commit a947f0f

Please sign in to comment.