The WMM TSPEC processor used the input buffer for processing the request
and building the response. This was fine for the FT case, but for the
WMM Action frame case, the input buffer is marked const, so it should
not really be modified. This modification could not really cause any
noticeable harm, but it can result in error reports from fuzzing and
potentially even from some static analyzers.
Fix this by marking the input arguments const more consistently (the
parsed IE was able to drop the const) and copy the const input data to a
temporary buffer for processing and modification instead of allowing the
input data to be modified.
Credit to OSS-Fuzz: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=19050
Signed-off-by: Jouni Malinen <j@w1.fi>
The previous description of wmediumd_scan_only_one used multiple lines
and that resulted in parallel-vm.py miscounting total number of test
cases. Fix that by getting rid of the newlines from the description.
Signed-off-by: Jouni Malinen <j@w1.fi>
Now that parallel-vm.py is actually stopping VMs as soon as they are not
needed for retries, it is not really an unexpected exit to see a VM exit
while test cases remain in the queue as long as at least that many VMs
remain running. Get rid of confusing 'unexpected exit' status from the
UI in such cases.
Fixes: 4aaddecdd8 ("tests: Handle test retries through the same queue")
Signed-off-by: Jouni Malinen <j@w1.fi>
Now that wpa_supplicant was modified to use NAI realm query even if
roaming_consortium is set, this test case is actually able to find a
match. Update it to avoid cause incorrect test failures.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Previously ANQP fetch considered NAI realm query only when cred block
did not specify roaming_consortium to optimize ANQP operations (NAI
realm list can have a very long value). In certain cases, both NAI realm
and roaming_consortium are configured in credential block and this
resulted in ANQP fetch for NAI realm not being initiated. That could
result in not being able to select the highest priority available
credential/network.
Remove roaming_consortium check for NAI realm query inclusion in ANQP
request so that we will request NAI realm information whenever it can
result in matching additional networks. This makes the ANQP queries more
costly in some cases, but the additional information is needed for
correct behavior in network selection.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Commit 947b5a1532 ("P2P: Stop listen state if Action frame TX is
needed on another channel") added an optimization for P2P response
transmission in certain concurrent operation cases. However, it did not
take into account possibility of the driver not being in listen
state (p2p->drv_in_listen == 0) and could end up getting stuck with the
P2P state machine in a manner that made the device not listen for
following messages. This showed up in following manner in the debug log:
P2P: Starting short listen state (state=SEARCH)
P2P: Driver ended Listen state (freq=2437)
process received frame and send a response
P2P: Stop listen on 0 MHz to allow a frame to be sent immediately on 2437 MHz
P2P: Clear timeout (state=SEARCH)
--> state machine stuck
Fix this by adding drv_in_listen > 0 condition for the optimization to
stop the listen operation in send_action() resulting in scheduled TX.
Fixes: 947b5a1532 ("P2P: Stop listen state if Action frame TX is needed on another channel")
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Update center frequency and center frequency2's DFS channel status in
VHT80+80 mode. Otherwise it will cause AP failed to start on a DFS
channel.
Tested: qca9984 with firmware ver 10.4-3.10-00047
Signed-off-by: Rick Wu <rwu@codeaurora.org>
Signed-off-by: Lei Wang <leiwa@codeaurora.org>
wpa_validate_wpa_ie() was already extended to cover these cases with
WPA_INVALID_PMKID return value, but hostapd_notif_assoc() did not have
code for mapping this into the appropriate status code
(STATUS_INVALID_PMKID) and ended up using the default
(WLAN_STATUS_INVALID_IE) instead. This caused AP SME-in-driver cases
returning incorrect status code when the AP did not have a matching
PMKSA cache entry. This could result in unexpected station behavior
where the station could continue trying to use a PMKSA cache entry that
the AP does not have and not being able to recover this.
Fix this by adding the previously missed mapping of validation errors to
status/reason codes.
Fixes: 567da5bbd0 ("DPP: Add new AKM")
Fixes: 458d8984de ("SAE: Reject request with mismatching PMKID (no PMKSA cache entry)")
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
wpas_dbus_handler_scan() constructs a set of 'params' each time, but it
doesn't acknowledge the existing randomization settings when doing so.
That means that any D-Bus initiated scans weren't going to follow the
configured settings.
Signed-off-by: Eric Caruso <ejcaruso@chromium.org>
Add D-Bus property:
* MACAddressRandomizationMask: a{say}
which configure random MAC address functionality in the Wi-Fi
driver via netlink.
Signed-off-by: Eric Caruso <ejcaruso@chromium.org>
This array can be freed either from the scan parameters or from
clearing the MAC address randomization parameters from the
wpa_supplicant struct. To make this ownership more clear, we have
each struct own its own copy of the parameters.
Signed-off-by: Eric Caruso <ejcaruso@chromium.org>
These parameter use hexdump values over the control interface, so
mention them in the DPP example documentation.
Signed-off-by: Jouni Malinen <j@w1.fi>
The "\\u%04x" printf string did not really work in the correct way if
char is signed. Fix this by type casting this to unsigned char.
Signed-off-by: Jouni Malinen <j@w1.fi>
There was a bug in wmediumd in that it didn't set the
frequency of frames, and thus they were always received
by mac80211_hwsim, regardless of channel it was on.
Add two tests that verify we only find a single instance
of an AP if we only have that one, and run this both with
and without wmediumd.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Avoid race conditions in completing mesh group removal/re-addition steps
and starting connectivity check. It seemed to be possible to get to the
connectivity test in some cases before the mesh STA had rejoined the
mesh and completed key configuration.
Signed-off-by: Jouni Malinen <j@w1.fi>
Export a new 'owe' capability to indicate that wpa_supplicant was
built with OWE support and accepts 'key_mgmt=OWE'. Also, support 'owe'
in the array of BSS' available key managements.
Signed-off-by: Beniamino Galvani <bgalvani@redhat.com>
When the P2P management interface is deleted, P2P is then disabled and
global->p2p_init_wpa_s is set to NULL. After that, other interfaces can
still trigger P2P functions (like wpas_p2p_find()) using D-Bus. This
makes wpa_supplicant terminate with SIGSEGV, because it dereferences a
NULL pointer. Fix this by adding proper checks, like it's done with
wpa_cli.
CC: Beniamino Galvani <bgalvani@redhat.com>
CC: Benjamin Berg <benjamin@sipsolutions.net>
Reported-by: Vladimir Benes <vbenes@redhat.com>
Signed-off-by: Davide Caratti <davide.caratti@gmail.com>
SUITE_B_192 AKM capability was indicated for all devices using the
nl80211 driver (without the QCA vendor specific AKM capability
indication). However, some devices can't handle Suite B 192 due to
insufficient ciphers supported. Add a check for CCMP-256 or GCMP-256
cipher support and only indicate SUITE_B_192 capability when such cipher
is supported. This allows compiling with CONFIG_SUITEB192 and still get
proper response to the 'GET_CAPABILITY key_mgmt' command. Under Android
it can also serve as a dynamic way for HAL to query platform for
WPA3-Enterprise 192-bit support.
Signed-off-by: Mikael Kanstrup <mikael.kanstrup@sony.com>
Previously, an unexpected SAE commit message resulted in forcing
disconnection. While that allowed recovery by starting from scratch,
this is not really necessary. Ignore such unexpected SAE commit message
instead and allow SAE confirm message to be processed after this. This
is somewhat more robust way of handling the cases where SAE commit
message might be retransmitted either in STA->AP or AP->STA direction.
Signed-off-by: Jouni Malinen <j@w1.fi>
wpa_key_mgmt_wpa_psk() includes SAE AKMs. However, with SAE, there is no
way of reaching 4-way handshake without the password having already been
verified as part of SAE authentication. As such, a failure to complete
4-way handshake with SAE cannot indicate that the used password was
incorrect.
Signed-off-by: Jouni Malinen <j@w1.fi>
Addi STA node details in AP through QCA vendor subcommand
QCA_NL80211_VENDOR_SUBCMD_ADD_STA_NODE vendor when processing FT
protocol roaming.
Signed-off-by: Shiva Sankar Gajula <sgajula@codeaurora.org>
To config BT coex chain mode, add a new QCA sub command
QCA_NL80211_VENDOR_SUBCMD_BTC_CHAIN_MODE.
This new command has two attributes:
1. QCA_VENDOR_ATTR_COEX_BTC_CHAIN_MODE: u32 attribute. Indicates the
BT coex chain mode, are 32-bit values from enum qca_btc_chain_mode.
2. QCA_VENDOR_ATTR_COEX_BTC_CHAIN_MODE_RESTART: flag attribute.
If set, vdev should be restarted once BT coex chain mode is updated.
Signed-off-by: Yu Wang <yyuwang@codeaurora.org>
The QCA_NL80211_VENDOR_SUBCMD_OEM_DATA command was previously defined
only to pass data blobs from user space to kernel (application to
firmware) but there was no mechanism to send the data blobs from
firmware to application. Extend this to define use of the same
subcommand and attributes as vendor events.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
If hostapd had existing STA SAE state, e.g., from a previously completed
SAE authentication, a new start of a separate SAE authentication (i.e.,
receiving of a new SAE commit) ended up using some of the previous
state. This is problematic for determining whether to H2E vs. looping
since the STA is allowed (even if not really expected to) to change
between these two alternatives. This could result in trying to use H2E
when STA was using looping to derive PWE and that would result in SAE
confirm failing.
Fix this by determining whether to use H2E or looping for the restarted
authentication based on the Status Code in the new SAE commit message
instead of previously cached state information.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
This wrapper function is not used anymore, so drop it instead of trying
to figure out good way of implementing it in constant time with various
crypto libraries.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Remove support for performing full sqrt(), i.e., only support curves
that use prime with p = 3 mod 4. In practice, this drops only group 26
with SAE H2E. This seems acceptable since there does not seem to be any
strong use case for that group taken into account the limits being
placed on acceptable prime lengths.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
This is in preparation of an implementation to to remove the alternative
SSWU path that was needed only for this particular group.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
These test cases were all using group 21 even though they were supposed
to go through all the possible groups.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
When using void pointers in calculations, the behavior is undefined.
Arithmetic operations on 'void *' is a GNU C extension,
which defines the 'sizeof(void)' to be 1.
This change improves portability of the code.
Signed-off-by: Jesus Fernandez Manzano <jesus.manzano@galgus.net>
The IE minimum length determination in hostapd_eid_ext_capab() was not
fully up to date with the hostapd_ext_capab_byte() conditions. This
could result in omitting some of the capability octets depending on
configuration. Fix this by adding the missing conditions.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
If there are no post-provision discovery operations, we should continue
in find mode to avoid getting the p2p_find operation stopped (stuck in
SEARCH state) unexpectedly.
Signed-off-by: Jimmy Chen <jimmycmchen@google.com>
Otherwise, we may get linker failures:
ld.lld: error: unable to find library -lbfd
While we're at it, pull in the library selection fixes from commit
848905b12a ("Avoid undefined references with CONFIG_WPA_TRACE_BFD=y").
Signed-off-by: Brian Norris <briannorris@chromium.org>
wpa_s->dbus_groupobj_path is always NULL so suppress to show it.
This was found by gcc 9.2.1 warning.
dbus/dbus_new.c: In function ‘wpas_dbus_unregister_p2p_group’:
dbus/dbus_new.c:4793:3: warning: ‘%s’ directive argument is null [-Wformat-overflow=]
4793 | wpa_printf(MSG_DEBUG,
| ^~~~~~~~~~~~~~~~~~~~~
4794 | "%s: Group object '%s' already unregistered",
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4795 | __func__, wpa_s->dbus_groupobj_path);
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
It looks like the wpa_auth_get_seqnum() function might not always be
visible in the backtrace since the compiler may optimize that wrapper
away. Use the i802_get_seqnum() function instead as a trigger for the
get_seqnum operation failure trigger to avoid this.
Signed-off-by: Jouni Malinen <j@w1.fi>
Protect RSNXE, if present, in FT Reassociation Request/Response frames.
This is needed for SAE H2E with FT.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Use a single struct definition and a single shared implementation for
parsing EAPOL-Key KDEs and IEs instead of maintaining more or less
identical functionality separately for wpa_supplicant and hostapd.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
If a STA indicates support for SAE H2E in RSNXE and H2E is enabled in
the AP configuration, require H2E to be used.
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
