Close all the control interface sockets and delete the client socket
files explicitly at the end of the test loop. This removes needs for
various workarounds that tried to force WpaSupplicant and Ctrl class
__del__() to remove the sockets.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This triggers a kmemleak scan between tests. This allows finding memory
leaks and doing this should attribute the leak to the correct test in
most cases. Note that it does add a sleep after each test, as such it is
most sensible when combined with UML time-travel.
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
If a tests hits a WARN_ON_ONCE then the retriggering will make it appear
to pass, since the warning will not happen again. Make this more
reliable by resetting the states at the beginning of each test.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Add SKB tracing (which shows now why/where a frame was dropped
in the stack), and also -T for stack trace at each event.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
When an in interface is added dynamically to hostapd with
HWSimRadio, it's not removed during device reset.
This requires to manually remove it, otherwise subsequent tests may
fail. Better do it during device reset.
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Unfortunately, some objects (WlantestCapture, WpaSupplicant
and wpaspy.Ctrl) use __del__ and actually have some logic
there. This is more or less wrong, and we should be using
context managers for it. However, cleaning that up is a
pretty large task.
Unfortunately, __del__ can cause reentrant logging which is
wrong too, because it might be invoked while in the middle
of a logging call, and the __del__ of these objects closes
connections and logs while doing that.
Since we're (likely) using cpython, we can work around this
by explicitly calling gc.collect() in a context where the
logging and close is fine, not only ensuring that all the
connections are closed properly before the next test, but
also fixing the issue with reentrant logging.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
If the current country code cannot be fetched, do not try to report the
current value in the log since that will result in TypeError exception.
Signed-off-by: Jouni Malinen <j@w1.fi>
Test names can now use wildcard in the end (e.g., ap_wpa2_psk*) to match
all test cases with the specified prefix.
Signed-off-by: Jouni Malinen <j@w1.fi>
The earlier change from using apdev[1] to using a no-interface
ifname=as-erp in the ERP related test cases ended up leaving the
separate authentication server interface running after the end of the
test case. This left UDP port 18128 open and any consecutive test case
needing this port failed to start another authentication server.
This can be reproduced with following test case sequence:
fils_auth_gtk_rekey authsrv_unknown_user
Fixes: e374def207 ("tests: Start ERP authentication server without AP")
Signed-off-by: Jouni Malinen <j@w1.fi>
The change to python3 broke trace-cmd recording, two
strings need to be bytes instead. Fix that.
Also add a flush() that seems to be needed now.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Use a helper function to perform this common sequence to disconnect and
stop any possibly started reconnection attempt.
Signed-off-by: Jouni Malinen <j@w1.fi>
Some of the test cases left behind attached control interface monitor
sockets that could result in hitting the wpa_supplicant socket TX queue
limit. Try to be a bit more careful about detaching and closing the
sockets to avoid this.
Signed-off-by: Jouni Malinen <j@w1.fi>
Use more readable "foo not in bar" construction for the couple of places
that did "not foo in bar".
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Prior cleanup broke the indentation here and made the addition of test
cases unreachable.
Fixes: 0663ae22ff ("tests: Do not use tabs for indentation")
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
This seems to be needed when using python3 in VM for the ssid_utf8 test
case debug prints from the control interface requests. This breaks
python2 support for the same logging entries, but there does not seem to
be any easy way of addressing this in a manner that works for both
python versions, so move ahead with the python3-only support from now
on.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
This allows more accurate logging of failures related to the cfg80211
country=98 (intersection) case. This version is trying to give some more
time to allow the country code to clear, but that does not seem to be
sufficient with the current cfg80211 implementation for country=98 (but
might be for other cases).
The additional check for country=98 at the beginning of each test case
is an attempt to force cfg80211 to restore world roaming state with a
new association and disconnection at the station side detected after the
AP side has already stopped. This is needed after the Linux kernel
commit 113f3aaa81bd ("cfg80211: Prevent regulatory restore during STA
disconnect in concurrent interfaces").
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes it easier to synchronize log entries in the kernel log
(seconds from boot) and wpa_supplicant/hostapd (UNIX timestamp).
Signed-off-by: Jouni Malinen <j@w1.fi>
There is no need to attach the monitor interface was events when issuing
only a single INTERFACES command.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This moves the wpa_supplicant debug entries from the end of a test case
using a dynamically added wlan5 interface to the correct test case,
i.e., the test case that added the interface instead of whatever test
case happens to follow this.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use a monitor interface given in the command line that is not also a
station or an AP as a monitor running wlantest on the channel used by
the test. This makes all the tests that use wlantest available for
execution on real hardware on remote hosts.
Signed-off-by: Jonathan Afek <jonathanx.afek@intel.com>
This is needed to avoid issues in some cases where 8-bit bytestrings may
be present in the otherwise text debug log.
Signed-off-by: Jouni Malinen <j@w1.fi>
Without this, the run-tests.py socket could have been left in attached
to receive all hostapd global events during a test case. This could hit
the limit of pending messages on the socket since there is nothing
clearing this socket during the execution of a test case. Fix this by
explicitly closing the socket after having completed the RELOG command.
Signed-off-by: Jouni Malinen <j@w1.fi>
It's somewhat annoying that you can only run parallel-vm.py as
./parallel-vm.py, not from elsewhere by giving the full path,
so fix that by resolving the paths correctly in the scripts where
needed.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This allows multiple dmesg files to be saved if a test case is executed
multiple times similarly to the other logfiles.
Signed-off-by: Jouni Malinen <j@w1.fi>
run-tests.py is running as root, so sudo does not need to be used
anymore from within each test case.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
There were couple of common cases where the control interface for the
dynamic wpa_supplicant instance could have been left in attached state
until Python ends up cleaning up the instance. This could result in
issues if many monitor interface events were queued for that attached
socket. Make this less likely to cause issues by explicitly detaching
and closing control interfaces before moving to the next test case.
Signed-off-by: Jouni Malinen <j@w1.fi>
Instead of returning "skip" from the test function, raise the new
HwsimSkip exception to indicate a test case was skipped.
Signed-off-by: Jouni Malinen <j@w1.fi>
