NM connections might not be available immediately and in that case an activation of such connection will raise DBusException. This patch adds a check for the exception and retries to activate connection five times with 1 second step. After timeout the exception is passed further.
Signed-off-by: Jan Tluka jtluka@redhat.com --- lnst/Slave/NmConfigDevice.py | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-)
diff --git a/lnst/Slave/NmConfigDevice.py b/lnst/Slave/NmConfigDevice.py index 396dea8..fa8fed4 100644 --- a/lnst/Slave/NmConfigDevice.py +++ b/lnst/Slave/NmConfigDevice.py @@ -24,6 +24,7 @@ from lnst.Common.Utils import kmod_in_use, bool_it from lnst.Common.NetUtils import scan_netdevs from lnst.Common.Utils import check_process_running from lnst.Common.Config import lnst_config +from dbus.exceptions import DBusException
NM_BUS = "org.freedesktop.NetworkManager" OBJ_PRE = "/org/freedesktop/NetworkManager" @@ -221,9 +222,19 @@ class NmConfigDeviceGeneric(object): except: device_obj_path = "/"
- netdev["acon_obj_path"] = nm_if.ActivateConnection( - netdev["con_obj_path"], - device_obj_path, "/") + timeout = 0 + while timeout != 5: + try: + netdev["acon_obj_path"] = nm_if.ActivateConnection( + netdev["con_obj_path"], + device_obj_path, "/") + break + except DBusException as e: + # connection might not be available yet, wait 1 second + timeout += 1 + if timeout == 5: + raise e + time.sleep(1)
logging.debug("Device object path: %s" % device_obj_path) logging.debug("Connection object path: %s" % netdev["con_obj_path"])
Tue, Jan 14, 2014 at 03:12:56PM CET, jtluka@redhat.com wrote:
NM connections might not be available immediately and in that case an activation of such connection will raise DBusException. This patch adds a check for the exception and retries to activate connection five times with 1 second step. After timeout the exception is passed further.
This looks like one ugly workaround. Why is this really needed? Why there is a delay? Can't we listen to some dbus signal instead of this polling?
Signed-off-by: Jan Tluka jtluka@redhat.com
lnst/Slave/NmConfigDevice.py | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-)
diff --git a/lnst/Slave/NmConfigDevice.py b/lnst/Slave/NmConfigDevice.py index 396dea8..fa8fed4 100644 --- a/lnst/Slave/NmConfigDevice.py +++ b/lnst/Slave/NmConfigDevice.py @@ -24,6 +24,7 @@ from lnst.Common.Utils import kmod_in_use, bool_it from lnst.Common.NetUtils import scan_netdevs from lnst.Common.Utils import check_process_running from lnst.Common.Config import lnst_config +from dbus.exceptions import DBusException
NM_BUS = "org.freedesktop.NetworkManager" OBJ_PRE = "/org/freedesktop/NetworkManager" @@ -221,9 +222,19 @@ class NmConfigDeviceGeneric(object): except: device_obj_path = "/"
netdev["acon_obj_path"] = nm_if.ActivateConnection(netdev["con_obj_path"],device_obj_path, "/")
timeout = 0while timeout != 5:try:netdev["acon_obj_path"] = nm_if.ActivateConnection(netdev["con_obj_path"],device_obj_path, "/")breakexcept DBusException as e:# connection might not be available yet, wait 1 secondtimeout += 1if timeout == 5:raise etime.sleep(1) logging.debug("Device object path: %s" % device_obj_path) logging.debug("Connection object path: %s" % netdev["con_obj_path"])-- 1.8.1.4
LNST-developers mailing list LNST-developers@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/lnst-developers
Tue, Jan 14, 2014 at 09:46:02PM CET, jpirko@redhat.com wrote:
Tue, Jan 14, 2014 at 03:12:56PM CET, jtluka@redhat.com wrote:
NM connections might not be available immediately and in that case an activation of such connection will raise DBusException. This patch adds a check for the exception and retries to activate connection five times with 1 second step. After timeout the exception is passed further.
This looks like one ugly workaround. Why is this really needed? Why there is a delay? Can't we listen to some dbus signal instead of this polling?
Yes, you're more than right. I was thinking about it after I sent the patch and I agree that this should be done prettier.
I'll post another patch once I have it.
Signed-off-by: Jan Tluka jtluka@redhat.com
lnst/Slave/NmConfigDevice.py | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-)
diff --git a/lnst/Slave/NmConfigDevice.py b/lnst/Slave/NmConfigDevice.py index 396dea8..fa8fed4 100644 --- a/lnst/Slave/NmConfigDevice.py +++ b/lnst/Slave/NmConfigDevice.py @@ -24,6 +24,7 @@ from lnst.Common.Utils import kmod_in_use, bool_it from lnst.Common.NetUtils import scan_netdevs from lnst.Common.Utils import check_process_running from lnst.Common.Config import lnst_config +from dbus.exceptions import DBusException
NM_BUS = "org.freedesktop.NetworkManager" OBJ_PRE = "/org/freedesktop/NetworkManager" @@ -221,9 +222,19 @@ class NmConfigDeviceGeneric(object): except: device_obj_path = "/"
netdev["acon_obj_path"] = nm_if.ActivateConnection(netdev["con_obj_path"],device_obj_path, "/")
timeout = 0while timeout != 5:try:netdev["acon_obj_path"] = nm_if.ActivateConnection(netdev["con_obj_path"],device_obj_path, "/")breakexcept DBusException as e:# connection might not be available yet, wait 1 secondtimeout += 1if timeout == 5:raise etime.sleep(1) logging.debug("Device object path: %s" % device_obj_path) logging.debug("Connection object path: %s" % netdev["con_obj_path"])-- 1.8.1.4
LNST-developers mailing list LNST-developers@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/lnst-developers
On Tue, Jan 14, 2014 at 09:46:02PM +0100, Jiri Pirko wrote:
Tue, Jan 14, 2014 at 03:12:56PM CET, jtluka@redhat.com wrote:
NM connections might not be available immediately and in that case an activation of such connection will raise DBusException. This patch adds a check for the exception and retries to activate connection five times with 1 second step. After timeout the exception is passed further.
This looks like one ugly workaround. Why is this really needed? Why there is a delay? Can't we listen to some dbus signal instead of this polling?
DBus signals require creating a gtk main loop object and creating a signal handler that will interrupt that loop so that the lnst-slave process can continue. If, for whatever reason, we start that loop after the connection was added and we won't recieve that signal we can't stop the loop. This is also the reason why I didn't use signal handling for activating connections or adding connections.
Even though polling is an ugly hack, it works and using signals would require a complete rewrite of the lnst-slave process to be a gtk application with a main loop that works for everything.
Signed-off-by: Jan Tluka jtluka@redhat.com
lnst/Slave/NmConfigDevice.py | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-)
diff --git a/lnst/Slave/NmConfigDevice.py b/lnst/Slave/NmConfigDevice.py index 396dea8..fa8fed4 100644 --- a/lnst/Slave/NmConfigDevice.py +++ b/lnst/Slave/NmConfigDevice.py @@ -24,6 +24,7 @@ from lnst.Common.Utils import kmod_in_use, bool_it from lnst.Common.NetUtils import scan_netdevs from lnst.Common.Utils import check_process_running from lnst.Common.Config import lnst_config +from dbus.exceptions import DBusException
NM_BUS = "org.freedesktop.NetworkManager" OBJ_PRE = "/org/freedesktop/NetworkManager" @@ -221,9 +222,19 @@ class NmConfigDeviceGeneric(object): except: device_obj_path = "/"
netdev["acon_obj_path"] = nm_if.ActivateConnection(netdev["con_obj_path"],device_obj_path, "/")
timeout = 0while timeout != 5:try:netdev["acon_obj_path"] = nm_if.ActivateConnection(netdev["con_obj_path"],device_obj_path, "/")breakexcept DBusException as e:# connection might not be available yet, wait 1 secondtimeout += 1if timeout == 5:raise etime.sleep(1) logging.debug("Device object path: %s" % device_obj_path) logging.debug("Connection object path: %s" % netdev["con_obj_path"])-- 1.8.1.4
LNST-developers mailing list LNST-developers@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/lnst-developers
LNST-developers mailing list LNST-developers@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/lnst-developers
lnst-developers@lists.fedorahosted.org
-
Jan Tluka -
Jiri Pirko -
Ondrej Lichtner