When DFS channel state is shared across multiple radios on the system it
is possible that a CAC completion event is propagated from other radio
to us. When in enabled state, do not proceed with setup completion upon
processing CAC completion event with devices where DFS is not offloaded,
when in state other than enabled make sure the configured DFS channel is
in available state before start the AP.
Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qti.qualcomm.com>
The userspace may want to delay the the first scheduled scan.
This enhances sched_scan to add initial delay (in seconds) before
starting first scan cycle. The driver may optionally choose to
ignore this parameter and start immediately (or at any other time).
This uses NL80211_ATTR_SCHED_SCAN_DELAY to add this via user
global configurable option: sched_scan_start_delay.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The while (len) loop was updating the next pointer at the end even when
len == 0, i.e., when the new next value won't be used. This could result
in reading one octet beyond the end of the allocated response wpabuf.
While the read value is not really used in practice, this is not correct
behavior, so fix this by skipping the unnecessary next pointer update in
len == 0 case.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The maximum required size for each Beacon Report element is known in
advance: it is the size of the Beacon Report element fixed fields + the
size of the Reported Frame Body subelement.
Allocate the buffer used for constructing the Beacon Report element
dynamically with the maximum needed size, instead of using a very
large static buffer.
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Use a non-existing directory in the path to avoid SQLite from being able
to create a new database file. The previous design worked in the VM case
due to the host file system being read-only, but a bit more is needed
for the case when this is running on the host.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This leads to cleaner code overall, and also reduces the size
of the hostapd and wpa_supplicant binaries (in hwsim test build
on x86_64) by about 2.5 and 3.5KiB respectively.
The mechanical conversions all over the code were done with
the following spatch:
@@
expression SIZE, SRC;
expression a;
@@
-a = os_malloc(SIZE);
+a = os_memdup(SRC, SIZE);
<...
if (!a) {...}
...>
-os_memcpy(a, SRC, SIZE);
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This can be used to clean the code and reduce size by converting
os_malloc() followed by os_memcpy() cases to use a single function call.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Add support for rejecting a BSS transition request using MBO reject
reason codes. A candidate is selected or rejected based on whether it is
found acceptable by both wpa_supplicant and the driver. Also accept any
candidate meeting a certain threshold if disassoc imminent is set in BTM
Request frame.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Add driver interface command using the QCA vendor extensions to check
the driverr whether to accept or reject a BSS transition candidate. For
the reject case, report an MBO reject reason code.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use the new hostapd.add_ap() API (i.e., pass the ap device as a
parameter instead of the interface name) in beacon report tests to
make them remote compatible, and mark them appropriately.
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Use a local variable for the STA address instead of fetching it
separately for each operation. Dump control interface monitor events
between each test message to avoid increasing the socket output queue
unnecessarily.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The second AP is not really needed in this test case that verifies
parsing of various different BSs Transition Management Request frame
payloads.
Signed-off-by: Beni Lev <beni.lev@intel.com>
The parsing code expects non_pref_chan to be non-quoted.
Fix the example in wpa_supplicant.conf not to include
quotes.
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
radius_server_session_free() does not remove the session from the
session list and these radius_server_get_new_session() error paths ended
up leaving a pointer to freed memory into the session list. This
resulted in the following operations failing due to use of freed memory.
Fix this by using radius_server_session_remove() which removes the entry
from the list in addition to calling radius_server_session_free().
Signed-off-by: Jouni Malinen <j@w1.fi>
It was possible to hit a double-free on the l2_packet socket if
initialization of DHCP/NDISC snoop failed on a hostapd interface that
had previously had those enabled successfully. Fix this by clearing the
l2_packet pointers during deinit.
Signed-off-by: Jouni Malinen <j@w1.fi>
Add QCA_NL80211_VENDOR_SUBCMD_BRP_SET_ANT_LIMIT for setting the number
of antennas that will be active in different modes for each connection.
Signed-off-by: Alexei Avshalom Lazar <qca_ailizaro@qca.qualcomm.com>
Previously it was possible for wpa_s->scan_res_handler to remain set to
its old value in case wpa_drv_scan() failed and no retry for the scan
trigger was scheduled (i.e., when last_scan_req == MANUAL_SCAN_REQ).
This could result in getting stuck with the next connection attempt
after a failed "SCAN TYPE=ONLY" operation when wpa_s->scan_res_handler
was set to scan_only_handler().
Fix this by clearing wpa_s->scan_res_handler if wpa_drv_scan() fails and
no retry is scheduled.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Define a new attribute QCA_WLAN_VENDOR_ATTR_CONFIG_SUB20_CHAN_WIDTH.
This attribute can set a station device to work in 5 or 10 MHz channel
width while in disconnect state.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
wnm_bss_tm_global uses an unknown country code to use Table E-4. Extend
that with otherwise identical test case wnm_bss_tm_global4, but with the
country string explicitly indicating use of Table E-4 while using a
known country code.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The new hostapd.conf parameter country3 can now be used to configure the
third octet of the Country String that was previously hardcoded to ' '
(= 0x20).
For example:
All environments of the current frequency band and country (default)
country3=0x20
Outdoor environment only
country3=0x4f
Indoor environment only
country3=0x49
Noncountry entity (country_code=XX)
country3=0x58
IEEE 802.11 standard Annex E table indication: 0x01 .. 0x1f
Annex E, Table E-4 (Global operating classes)
country3=0x04
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
OpenSSL.SSL.Connection.state_string() was replaced with
get_state_string() in pyopenssl. Add workaround code to be able to use
either of these names.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
When ASLR is enabled, like it is by default on many distros now,
the trace code doesn't work right.
Fix this by looking up the start of the executable mapping and
subtracing it from all the lookups.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
The send_eapol() calls for delivering frames to wpa_supplicant had a
copy-paste bug from the earlier hostapd cases. These were supposed to
use the BSSID, not the address of the station, as the source address.
The local address worked for most cases since it was practically
ignored, but this could prevent the race condition workaround for
association event from working. Fix this by using the correct source
address (BSSID).
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
CONFIG_TLS=linux can now be used to select the crypto implementation
that uses the user space socket interface (AF_ALG) for the Linux kernel
crypto implementation. This commit includes some of the cipher, hash,
and HMAC functions. The functions that are not available through AF_ALG
(e.g., the actual TLS implementation) use the internal implementation
(CONFIG_TLS=internal).
Signed-off-by: Jouni Malinen <j@w1.fi>
These operations may fail with some crypto wrappers, so allow the
functions to report their results to the caller.
Signed-off-by: Jouni Malinen <j@w1.fi>
The reason detail field in non_pref_chan attribute was removed
from MBO draft v0.0_r25. Also oper_class can be 1 character for
few country codes (e.g., country code-UK, channel number-1). So the
shortest channel configuration is 7 characters.
This was missed in the earlier commit
4a83d4b686 ('MBO: Do not add reason_detail
in non_pref_chan attr (STA)') that took care of other changes related to
removal of the reason detail.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
When getting the channel number from a frequency, all supported modes
should be checked rather than just the current mode. This is needed when
hostapd switches to a channel in different band.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
