In wpa_supplicant_terminate_proc(), while iterating and
terminating interfaces, after an interface is terminated,
we can no longer access its memory as it is no longer valid
and causes a crash.
Fix this by saving the pointer to the next interface before freeing
the current one to avoid accessing an invalid memory location.
Signed-hostap: Moshe Benji <moshe.benji@intel.com>
If secondary channel is provided for CSA, advertise it in the Secondary
Channel Offset element in Beacon and Probe Response frames.
Signed-hostap: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
When CSA flow starts, store the entire struct hostapd_freq_params and
not only CS frequency as it was before. The additional freq_params are
required to advertise CS supplementary IEs such as secondary channel,
wide bandwidth CS, etc.
Signed-hostap: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
This allows use of structs (and not only pointers) defined in drivers.h.
Remove also some not needed forward declarations and redundant includes.
Signed-hostap: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
This just serves to check if there was a scan within
the last 5 seconds, hence it should use monotonic time.
While at it, also use os_reltime_expired().
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
This is only used for a debug message, but that message prints
the time since the last attempt, so it should use monotonic
time instead of wall clock.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The MMIC failure code should use monotonic time to check
whether 60 seconds have elapsed or not. For type-safety,
use struct os_reltime for the timestamp variable, and
also convert to using os_reltime_expired().
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
Temporarily disabled networks are disabled for a certain
duration, so the code should use monotonic time.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
GO activation can fail if the first client doesn't connect
within a certain time, but this should not be dependent on
wall time -- use monotonic time instead.
While at it, use os_reltime_expired().
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The reinit detection skips reinit when the time since the own
authentication frame TX is less than half a second, so it shouldn't
be affected by wall time and use monotonic time instead.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The event debouncing isn't very accurate (since it doesn't
take sub-second resolution into account), but it should use
monotonic time anyway since it doesn't care about the wall
clock.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The PBC ignore-start workaround just needs to check whether
the time is within 5 seconds, so should use monotonic time.
While at it, add a few more ifdefs to clearly separate the
code and variables needed.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
PBC sessions are just time-stamped when activated, and
eventually time out, so should use monotonic time.
While at it, make the code use os_reltime_expired().
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The EAPOL RX workaround checks that the events are less than 100 ms
apart, so only uses relative times and should use monotonic time.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The bgscan simple and learn algorithms should run regardless
of wall clock time jumps, so make them use monotonic time.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The eloop already tracks the expiration/lifetime, and the expiration
isn't really used, so remove it. It should otherwise have used monotonic
time, but since it's not actually used, we can remove it instead.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The RADIUS server needs to calculate uptime, which is relative
and thus should use monotonic time.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
Since the RADIUS client cares about relative time (retry timeout)
only, it should use monotonic time.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The BSS table, scan timeout, and related functionality should use
monotonic time since they care about relative values (age) only.
Unfortunately, these are all connected, so the patch can't be split
further. Another problem with this is that it changes the driver wrapper
API. Though, it seems only the test driver is using this.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The SA query timeout is just a regular timeout (currently
hard-coded to 1000 TU), so use monotonic time for it.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This should probably have used monotonic time for entry timestamps, but
as those aren't used at all right now, so just remove them entirely.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
The request retry loop only retries for 5 seconds, so any time
jumps would probably not affect it much, but it should be using
monotonic time nonetheless since it only cares about duration.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
Wall time jumps shouldn't affect MMIC failure/TKIP countermeasures,
so use monotonic time. Change the michael_mic_failure variable to
struct os_reltime for type-safety.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
Since the PMKSA cache only uses relative time, use the monotonic time
functions instead of wall time to be correct when the clock jumps.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
For type-safety, make sta->acct_session_start a struct os_reltime
and then use monotonic time for accounting. For RADIUS reporting,
continue to use wall clock time as specified by RFC 2869, but for
the session time use monotonic time.
Interestingly, RFC 2869 doesn't specify a timezone, so the value
is somewhat arbitrary.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
bgscan_learn_get_probe_freq() starts from returning the second entry in
the supp_freqs arrays. Change its logic a bit to make it start from the
first entry.
Signed-hostap: Eliad Peller <eliad@wizery.com>
In case the initial signal level of the associated BSS was above the
given threshold, bgscan_learn module would begin using the
short_interval but never switch to the long_interval as there would be
no signal change event. Make the init code poll for the current signal
level and set scan_interval accordingly. This logic exists in
bgscan_simple but was missing in bgscan_learn.
Signed-hostap: Eyal Shapira <eyal@wizery.com>
In order to handle regulatory domain requests, crda needs to be
installed on the host, but we also need to install a uevent helper in
the VM so that it gets executed (since we don't run udev).
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
Some hw modes (e.g., 11b and 11g) contain the same frequencies,
causing the supp_freqs array to be populated with redundant entries.
Check for the existence of the freq before adding it.
Signed-hostap: Eliad Peller <eliad@wizery.com>
Stopping bgscan on any state other than COMPLETED results
in bgscan reset (stop + start) on every rekeying operation.
Signed-hostap: Eliad Peller <eliad@wizery.com>
This verifies that QoS Map configuration and update can be completed.
Actual QoS mapping for Data frames within mac80211/cfg80211 is not yet
verified.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Do not assume the driver supports QoS Mapping, but instead, advertise
support for this only if CONFIG_INTERWORKING is defined and driver
indicates support for configuring QoS Map.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
The driver_nl80211.c changes are needed to avoid compiler warnings
with two frequency attributes pointing to the same value.
Signed-hostap: Jouni Malinen <j@w1.fi>
Commit 202dec2a94 introduced a never
ending loop for a case where a single wpa_supplicant process is used
with multiple radios. Fix this by advancing the iface pointer properly
to the next interface in the loop until a NULL pointer is hit.
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
If more BSSes are added in config file than are supported by the driver,
segmentation fault can appear. For this case, the interface_added flag
needs to be cleared if adding a new BSS fails.
Signed-hostap: Marek Kwaczynski <marek.kwaczynski@tieto.com>
Until now DFS was simply restarting the AP when radar was detected. Now
CSA is used to perform smooth switch to the new channel. Stations not
supporting CSA will behave as before.
Signed-hostap: Janusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
This is needed for AP CSA. Since CSA must happen immediately after radar
is detected there's no time to perform CAC. Thus, radar channels must be
disabled when looking for a new channel to escape to after a radar is
detected.
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
Adds support for VHT by parsing bandwidth and center_freq{1,2}.
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
Signed-hostap: Janusz Dziedzic <janusz.dziedzic@tieto.com>
Just the usual, with a new function os_reltime_initialized()
thrown in that checks whether time has ever been retrieved
(time can't be completely zero).
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
