Avraham Stern [Wed, 17 Feb 2021 10:14:33 +0000 (12:14 +0200)]
nl80211: Use process_bss_event() for the nl_connect handler
The nl_connect is initialized with the process_bss_event() handler.
However, it is used several times with the default valid handler. As a
result, if a message that is only valid for process_bss_event() is
received while the default handler is used, it will be dropped.
This has been observed in a case where during the 4-way handshake, a
Beacon frame is received on the AP side, which triggers a beacon update,
just before receiving the next EAPOL. When send_and_recv_msgs_owner() is
called for sending the NL80211_CMD_SET_BEACON command, the
NL80211_CMD_CONTROL_PORT_FRAME event is already pending. As a result, it
is received with the default handler, which drops it. Since the EAPOL
frame is dropped, the connection attempt fails.
Fix it by using the process_bss_event() handler when the nl_connect
handler is used.
Jouni Malinen [Sat, 6 Mar 2021 09:56:00 +0000 (11:56 +0200)]
DPP: Clear hapd->gas pointer on deinit
While it does not look like the stale pointer could have been
dereferenced in practice, it is better not to leave the stale pointer to
freed memory in place to avoid accidental uses.
Michael Braun [Mon, 1 Mar 2021 21:27:46 +0000 (23:27 +0200)]
Fix use after free with hapd->time_adv on interface restart
When an interface is disabled, e.g. due to radar detected,
hapd->time_adv is freed by hostapd_free_hapd_data(), but later
used by ieee802_11_build_ap_params() calling hostapd_eid_time_adv().
Thus hapd->time_adv needs to be cleared as well.
Fixes: 39b97072b2a4 ("Add support for Time Advertisement") Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
Matthew Wang [Sat, 6 Mar 2021 01:43:44 +0000 (17:43 -0800)]
Reject authentication start during explicit roam requests
The roam D-Bus and ROAM control itnerface commands flip the reassociate
bit before calling wpa_supplicant_connect(). wpa_supplicant connect
eventually aborts ongoing scans (if any), which causes scan results to
be reported. Since the reassociate bit is set, this will trigger a
connection attempt based on the aborted scan's scan results and cancel
the initial connetion request. This often causes wpa_supplicant to
reassociate to the same AP it is currently associated to instead of the
explicitly requested roaming target.
Add a roam_in_progress flag to indicate that we're currently attempting
to roam via an explicitly request to a specific BSS so that we don't
initiate another connection attempt based on the possibly received scan
results from a scan that was in progress at the time the roam command
was received.
Signed-off-by: Matthew Wang <matthewmwang@chromium.org>
Jouni Malinen [Mon, 1 Mar 2021 10:51:20 +0000 (12:51 +0200)]
wpaspy: Do not mark not-existing UNIX domain socket as UDP
os.stat(path) failure is an ambigious indication of the control
interface "path" type (UDP hostname vs. UNIX domain socket path). The
path may be a valid UNIX domain socket path, but that socket could have
been removed just before reaching here. At least the hwsim test case
concurrent_p2pcli managed to hit the "connect exception" print below
from UDP handling even when using a UNIX domain socket.
Work around incorrect determination of control interface socket type by
assuming anything starting with '/' is a UNIX domain socket and not a
hostname.
Fixes: a2c88a8025b2 ("wpaspy: Add support for UDP connection") Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Jouni Malinen [Mon, 1 Mar 2021 10:34:05 +0000 (12:34 +0200)]
tests: Fix openssl_systemwide_policy cleanup
Need to close the WpaSupplicant instance on the extra radio before
returning from this test case since that interface is going to be
removed and WpaSupplicant.__del__() can time out on trying to detach the
monitor connection after that.
Jouni Malinen [Sun, 28 Feb 2021 21:27:13 +0000 (23:27 +0200)]
OCV: Fix OCV-FAILURE event address for FT Reassociation Response frame
sm->bssid is still the BSSID of the previous AP at this point in the FT
protocol, so need to show the target AP's BSSID instead in the failure
message.
Jouni Malinen [Sun, 28 Feb 2021 18:20:38 +0000 (20:20 +0200)]
tests: Display unexpected stdout and stderr prints in parallel-vm.py
Make it more difficult to miss issues that were previously only printed
out in /tmp/hwsim-test-logs/*-parallel.log. This covers things like
memory leaks and test script failures or forgotten development time
prints to stdout.
Jouni Malinen [Sun, 28 Feb 2021 16:39:49 +0000 (18:39 +0200)]
SAE: Use more explicit IE payload validation steps
This is an attempt of making the code easier to understand for static
analyzers. The helper functions were already verifying that these IEs
are fully within the memory buffer, but that may not have been clear
enough for automated analysis.
Jouni Malinen [Sun, 28 Feb 2021 09:51:16 +0000 (11:51 +0200)]
Use more consistent iface->conf checks
Commit f1df4fbfc7ad ("mesh: Use setup completion callback to complete
mesh join") added a check for iface->conf being NULL into a debug print.
However, it is not clear how that could be NULL here. In any case,
setup_interface() could end up dereferencing iface->conf in the call to
hostapd_validate_bssid_configuration(), so better be consistent with the
checks and not get warnings from static analyzers regardless of whether
this can happen in practice.
Jouni Malinen [Sun, 28 Feb 2021 09:37:09 +0000 (11:37 +0200)]
PASN: Avoid unreachable code with CONFIG_NO_RADIUS
There is no point in trying to build in rest of this function if in the
middle of it the CONFIG_NO_RADIUS case would unconditionally fail.
Simply make all of this be conditional on that build parameter not being
set to make things easier for static analyzers.
Jouni Malinen [Sun, 28 Feb 2021 09:27:42 +0000 (11:27 +0200)]
FILS: Fix RSN info in FD frame for no-group-addressed
The value from the initial RSN_CIPHER_SUITE_NO_GROUP_ADDRESSED check
ended up getting overridden with the following if. This was supposed to
be a single if statement to avoid that.
Fixes: 9c02a0f5a672 ("FILS: Add generation of FILS Discovery frame template") Signed-off-by: Jouni Malinen <j@w1.fi>
Jouni Malinen [Sat, 27 Feb 2021 21:42:21 +0000 (23:42 +0200)]
Fix dynamic EAP library building
Build eap_*.so into the wpa_supplicant similarly with the wpa_supplicant
binary and include the shared helper functions from additional files
into the builds. This got broken at some point with the build system
changes.
Jouni Malinen [Sat, 27 Feb 2021 19:07:54 +0000 (21:07 +0200)]
tests: Fix sigma_dut_ap_dpp_tcp_enrollee_init to stop hostapd
This test case was missing an explicit CAPI ap_reset_default and that
could result in hostapd being left running at the end of the test case.
This could result in issues with following test cases if they used a new
radio interface from HWSimRadio().
Jouni Malinen [Tue, 16 Feb 2021 09:34:50 +0000 (11:34 +0200)]
Ignore group-addressed SA Query frames
These frames are used for verifying that a specific SA and protected
link is in functional state between two devices. The IEEE 802.11
standard defines only a case that uses individual MAC address as the
destination. While there is no explicit rule on the receiver to ignore
other cases, it seems safer to make sure group-addressed frames do not
end up resulting in undesired behavior. As such, drop such frames
instead of interpreting them as valid SA Query Request/Response.
Jimmy Chen [Mon, 12 Oct 2020 02:23:40 +0000 (10:23 +0800)]
P2P: Pick a 5 GHz channel from more possible channels
For an autonomous P2P group on the 5 GHz band, a channel was picked only
from the operating class 115 which is not available in the EU region
anymore. As a result, an autonomous group creation would always fail in
this generic 5 GHz channel case.
There are more possible available channels for the 5 GHz currently.
Especially in the EU region, the operating class 115 channels are no
longer available, but SRD channels (the operating class 124) are
available. Allow them to be used here if they are marked as allowed for
P2P GO use.
In addition, iterate through all the potential options instead of just
checking the first randomly picked channel. Start this iteration from
random position to maintain some randomness in this process.
Signed-off-by: Jimmy Chen <jimmycmchen@google.com>
Jouni Malinen [Sat, 27 Feb 2021 16:10:41 +0000 (18:10 +0200)]
tests: Work around exception handling issue in wpas_config_file
Raising an exception while the wlan5 interface was remove (i.e., between
wpas.interface_remove() and .interface_add() calls) would result in the
cleanup code failing and generating yet another exception while the
first one was being processed. Work around this by re-adding the wlan5
interface back temporarily if the interface is not available for the
cleanup operations.
Determine if the TDLS peer is HE capable based on HE Capability element
received in the TDLS Setup Response frame. Indicate the peer's HE
capabilities to the driver through sta_add().
Jouni Malinen [Tue, 8 Dec 2020 21:52:50 +0000 (23:52 +0200)]
P2P: Fix a corner case in peer addition based on PD Request
p2p_add_device() may remove the oldest entry if there is no room in the
peer table for a new peer. This would result in any pointer to that
removed entry becoming stale. A corner case with an invalid PD Request
frame could result in such a case ending up using (read+write) freed
memory. This could only by triggered when the peer table has reached its
maximum size and the PD Request frame is received from the P2P Device
Address of the oldest remaining entry and the frame has incorrect P2P
Device Address in the payload.
Fix this by fetching the dev pointer again after having called
p2p_add_device() so that the stale pointer cannot be used.
Fixes: 17bef1e97a50 ("P2P: Add peer entry based on Provision Discovery Request") Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Reordering of code in handle_auth_cb() when adding support for full
station state messaged up frame length checks. The length was originally
tested before looking at the payload of the frame and that is obviously
the correct location for that check. The location after those full state
state changes was after having read six octets of the payload which did
not help at all since there was no addition accesses to the payload
after that check.
Move the payload length check to appropriate place to get this extra
level of protection behaving in the expected manner. Since this is a TX
status callback handler, the frame payload is from a locally generated
Authentication frame and as such, it will be long enough to include
these fields in production use cases. Anyway, better keep this check in
working condition.
Fixes: bb598c3bdd06 ("AP: Add support for full station state") Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Jouni Malinen [Sun, 21 Feb 2021 14:58:33 +0000 (16:58 +0200)]
tests: Fix the cleanup at the end of scan_setband
The iteration of WpaSupplicant instances used incorrect variable and
ended up cleaning up only the wlan5 interface. This left unexpected
setband parameter for wlan0/wlan1/wlan2 which could result in
consecutive test cases failing due to scan not finding the expected
BSSs.
Jouni Malinen [Sun, 21 Feb 2021 14:44:33 +0000 (16:44 +0200)]
DPP2: Accept Config Result before GAS response TX status
The TX event for the next frame in the sequence might be received before
the TX status for the final GAS response frame is processed. This used
to result in the Config Result getting discarded and the negotiation not
completing successfully on the Configurator side.
Accept the Config Result message as an indication of the final GAS
response frame having went through fine even if the TX status has not
yet been processed to avoid this issue from a potential race condition
on kernel events.
Jouni Malinen [Sun, 21 Feb 2021 10:07:38 +0000 (12:07 +0200)]
radiotap: Fix compiler issues with packed structures
Replace the Radiotap parser platform.h file with use of helper functions
from utils/common.h to avoid compiler issues with the updated design and
getting pointers to members of packet structs.
Silence the warning about _next_bitmap assignment. This pointer is
dereferenced only with operations that are safe for unaligned access, so
the compiler warning is not helpful here.
__packed might not be defined in this context, so use STRUCT_PACKED from
utils/common.h.
EAP: Extend Session-Id derivation with TLS 1.3 to PEAP and EAP-TTLS
This newer Session-Id/Method-Id derivation is used with PEAP and
EAP-TTLS when using TLS 1.3 per draft-ietf-emu-tls-eap-types-00, so do
not limit this to only EAP-TLS.
Signed-off-by: Alexander Clouter <alex@digriz.org.uk>
EAP-TTLS/PEAP peer: Fix failure when using session tickets under TLS 1.3
EAP peer does not expect data present when beginning the Phase 2 in
EAP-{TTLS,PEAP} but in TLS 1.3 session tickets are sent after the
handshake completes.
There are several strategies that can be used to handle this, but this
patch picks up from the discussion[1] and implements the proposed use of
SSL_MODE_AUTO_RETRY. SSL_MODE_AUTO_RETRY has already been enabled by
default in OpenSSL 1.1.1, but it needs to be enabled for older versions.
The main OpenSSL wrapper change in tls_connection_decrypt() takes care
of the new possible case with SSL_MODE_AUTO_RETRY for
SSL_ERROR_WANT_READ to indicate that a non-application_data was
processed. That is not really an error case with TLS 1.3, so allow it to
complete and return an empty decrypted application data buffer.
EAP-PEAP/TTLS processing can then use this to move ahead with starting
Phase 2.
Jouni Malinen [Sat, 20 Feb 2021 14:10:47 +0000 (16:10 +0200)]
tests: Use a longer timeout for connection in sigma_dut DPP test cases
It was possible for these test cases to fail if the first scan iteration
did not find the AP since the 10 second timeout was small enough to
terminate the second attempt before fetching the scan results. Increase
this timeout to allow at least two full scan iterations to be completed
before declaring failure.
Jouni Malinen [Sat, 20 Feb 2021 14:01:32 +0000 (16:01 +0200)]
tests: Clear scan results at the end of scan-ssid-list tests
These test cases use hidden SSIDs and left behind a BSS entry with no
SSID. That can cause issues for consecutive test cases where the BSSID
can be used as the key for finding a BSS entry. That could end up
picking the old hidden SSID BSS instead of the one that was meant to be
used in the test case.
Flush the scan cache at the end of the scan-ssid-list test cases to
reduce invalid test failures for the consecutive test cases.
Jouni Malinen [Sat, 20 Feb 2021 13:45:54 +0000 (15:45 +0200)]
tests: Make PASN tests more robust
The PASN_START command can fail if there is an old BSS entry for the
same BSSID from an earlier test case. Try to avoid this by flushing the
scan results before running these test cases.
Previously nl80211_nlmsg_clear() would be called under a special
condition when valid_handler is NULL and valid_data is -1. Such API is
not very convenient as it forces the handler to be NULL. Change the
send_and_recv() function to always clear the nl_msg, which will simplify
all this logic.
Signed-off-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Ilan Peer [Tue, 16 Feb 2021 13:23:57 +0000 (15:23 +0200)]
PASN: Correctly set RSNXE bits from STA
These defines are for the capability bit number, not the binary value
from the bit index. As such, need to use BIT() here to set the bitmap
appropriately.
Signed-off-by: Ilan Peer <ilan.peer@intel.com> Signed-off-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
wpa_supplicant: Fix potential memleak on an error path
extra_buf allocation was missed in one of the error cases.
Fixes: 170775232d61 ("ANQP: Add support to specify frequency in ANQP_GET command") Signed-off-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Anusha Datar [Wed, 17 Feb 2021 22:32:58 +0000 (17:32 -0500)]
Check for message truncation in RADIUS client
The RADIUS client currently determines if a radius message is longer
than the supported maximum length by checking whether the size of the
received buffer and the length of the buffer (as returned by recv()) is
equal. This method fails to detect if the buffer has actually been
truncated. This change modifies the RADIUS client to instead use the
recvmsg() call and then check the message header flags to determine
whether or not the received message has been truncated and drop the
message if that is the case.
Anusha Datar [Wed, 17 Feb 2021 22:32:57 +0000 (17:32 -0500)]
Set RADIUS message length to reflect RFC 2865
The current RADIUS server message maximum length limits the length of
each RADIUS message to 3000 bytes. As specified in RFC 2865 section 3
("Packet Format"), the RADIUS standard's maximum message size is 4096
bytes, so this change increases the RADIUS server's maximum message
size from 3000 to 4096 to match the standard.
Anusha Datar [Wed, 17 Feb 2021 22:32:56 +0000 (17:32 -0500)]
Create RADIUS_MAX_MSG_LEN param in the shared radius.h
The RADIUS client currently uses a hardcoded value of 3000 for the
maximum length of a RADIUS message, and the RADIUS server currently
defines a constant value for the maximum length of the RADIUS message
within its source. The client and the server should use the same
maximum length value, so this change creates a shared parameter
RADIUS_MAX_MSG_LEN within the header file radius.h and modifies
both the client and the server to use that parameter instead of
a locally set value.
Sachin Ahuja [Mon, 1 Feb 2021 11:51:47 +0000 (17:21 +0530)]
Add new attributes in get_sta_info QCA vendor command
Add additional attributes for the QCA vendor command
QCA_NL80211_VENDOR_SUBCMD_GET_STA_INFO to get finer details on roaming
behavior, TSF out of sync count, and the latest TX rate, Rate Index used
for the transmission.
AP: Check driver's capability to enable OCV when driver SME is used
When the driver SME is used, offloaded handshakes which need Operating
Channel Validation (OCV) such as SA Query procedure, etc. would fail if
hostapd enables OCV based on configuration but the driver doesn't
support OCV. To avoid this when driver SME is used, enable OCV from
hostapd only when the driver indicates support for OCV.
This commit also adds a capability flag to indicate whether driver SME
is used in AP mode.
STA: Check driver capability to enable OCV when driver SME is used
When the driver SME is used, offloaded RSN handshakes like SA Query, GTK
rekeying, FT authentication, etc. would fail if wpa_supplicant enables
OCV in initial connection based on configuration but the driver doesn't
support OCV. To avoid such failures check the driver's capability for
enabling OCV when the driver SME used.
This commit also adds a capability flag for indicating OCV support
by the driver.
Jouni Malinen [Mon, 15 Feb 2021 21:48:04 +0000 (23:48 +0200)]
Clean up RSN parameter setting for PASN
Set conf.force_kdk_derivation within the same if block as all the other
parameters. This is used only if ssid is not NULL, so no need to have
any special handling for this parameter.
Enable beacon protection only when driver indicates support
Enabling beacon protection will cause STA connection/AP setup failures
if the driver doesn't support beacon protection. To avoid this, check
the driver capability before enabling beacon protection.
This commit also adds a capability flag to indicate beacon protection
support in client mode only.
Hu Keping [Fri, 4 Sep 2020 14:04:55 +0000 (22:04 +0800)]
Update sgml to generate reproducible manpages
Prior to this patch, we failed to recreate bit-by-bit identical
copies of wpa_supplicant because it doesn't generate reproducible manpages.
Since the latest version(0.6.14-3 or new) of docbook-utils have already
support getting the date from sgml file [1], it is possible to make some
progress on the "reproducible builds" effort [2].
It was not easily possible to separate configuration of an interface and
credentials when using the configuration file instead of the control
interface or D-Bus interface for setting up the network profiles. This
makes it hard to distribute configuration across a set of nodes which
use wpa_supplicant without also having to store credentials in the same
file. While this can be solved via scripting, having a native way to
achieve this would be preferable.
Turns out there already is a framework to have external password
storages. It only had a single "test" backend though, which is kind of
an in-memory store which gets initialized with all passwords up front
and is mainly for testing purposes. This isn't really suitable for the
above use case: the backend cannot be initialized as part of the central
configuration given that it needs the credentials, and we want to avoid
scripting.
This commit thus extends the infrastructure to implement a new backend,
which instead uses a simple configuration file containing key-value
pairs. The file follows the format which wpa_supplicant.conf(5) uses:
empty lines and comments are ignored, while passwords can be specified
with simple `password-name=password-value` assignments.
With this new backend, splitting up credentials and configuration
becomes trivial:
Alternative approaches would be to support including other configuration
files in the main configuration, such that common configuration and
network declarations including credentials are split up into separate
files. But the implementation would probably have been more complex
compared to reusing the already-existing framework for external password
backends.
wpa_supplicant: Move wpa_config_get_line() into utils
The function wpa_config_get_line() is used by the wpa_supplicant config
file parser to retrieve the next non-comment non-blank line. We'll need
the same kind of functionality to implement the file-based external
password backend, so as a preparatory step this commit extracts the
function into its own standalone file in the utils package.
No functional changes are expected from this commit.
Jouni Malinen [Mon, 15 Feb 2021 16:24:37 +0000 (18:24 +0200)]
HE: Fall back to 20 MHz on 2.4 GHz if 40 MHz is not supported
At least the ACS case of an attempt to pick a 40 MHz channel on the 2.4
GHz band could fail if HE was enabled and the driver did not include
support for 40 MHz channel bandwidth on the 2.4 GHz band in HE
capabilities. This resulted in "40 MHz channel width is not supported in
2.4 GHz" message when trying to configure the channel and failure to
start the AP.
Avoid this by automatically falling back to using 20 MHz bandwidth as
part of channel parameter determination at the end of the ACS procedure.
Aloka Dixit [Thu, 14 May 2020 23:18:38 +0000 (16:18 -0700)]
ACS: Allow downgrading to 20 MHz based on OBSS results
When auto channel selection (ACS) is used for HE 40 MHz in the 2.4 GHz
band, AP sets center frequency after finding a 40 MHz channel and then
runs a scan for overlapping BSSes in neighboring channels. Upon OBSS
detection, AP should downgrade to 20 MHz bandwidth.
This was broken because allowed_ht40_channel_pair() returns true in this
case and the steps to reset center frequency are not executed causing
failure to bring interface up.
Fix the condition to allow rollback to 20 MHz.
Signed-off-by: Aloka Dixit <alokad@codeaurora.org>
Unsolicited broadcast Probe Response transmission is used for in-band
discovery in the 6 GHz band (IEEE P802.11ax/D8.0 26.17.2.3.2, AP
behavior for fast passive scanning). Add support for configuring the
parameters for such frames.
Signed-off-by: Aloka Dixit <alokad@codeaurora.org>
projects / thirdparty / hostap.git / log
