Add the `sae_check_mfp` global option to limit SAE when PMF will
not be selected for the connection.
With this option SAE is avoided when the hardware is not capable
of PMF due to missing ciphers.
With this option SAE is avoided on capable hardware when the AP
does not enable PMF.
Allows falling back to PSK on drivers with the
WPA_DRIVER_FLAGS_SAE capability but do not support the BIP cipher
necessary for PMF. This enables configurations that can fall back
to WPA-PSK and avoid problems associating with APs configured
with `sae_require_mfp=1`.
Useful when `pmf=1` and `sae_check_mfp=1` are enabled and networks
are configured with ieee80211w=3 (default) and key_mgmt="WPA-PSK SAE".
In this configuration if the device is unable to use PMF due to
lacking BIP group ciphers it will avoid SAE and fallback to
WPA-PSK for that connection.
Signed-off-by: Jeffery Miller <jefferymiller@google.com>
Stop listen mode if there is not sufficient time remaining to complete
the Authentication exchange within the current remain-on-channel
operation. This speeds up the operation and avoids some timeouts that
could prevent the provisioning step from completing. This addresses an
issue that was found in the following test case sequence:
dpp_controller_relay_discover dpp_chirp_ap_5g
Similar mechanism was already used for Reconfig Announcement frames, so
reuse that for this case with Presence Announcement frames.
Signed-off-by: Jouni Malinen <j@w1.fi>
When scanning for a new connection, we currently optimize by scanning
all frequencies only when our MCC capabilities will allow an additional
operating frequency, and scan only the existing operating frequencies
otherwise. This is problematic when there the current operating
frequency singularly accounts for one of the shared radio frequencies
because we should be able to switch operating frequencies without adding
to the channel count. Fix this.
Signed-off-by: Matthew Wang <matthewmwang@chromium.org>
Connection attempt could have been intermittently drop when reconnecting
to the same ESS due the current BSS entry getting removed immediately
after the disconnection for the purpose of reconnecting to the same ESS.
Avoid this by not removing a BSS entry for the same ESS when in this
special state of trying to reconnect to the same ESS.
Signed-off-by : WooYong Kim <wykim@newratek.com>
Scans are allowed on the current interface for group formation only when
in the P2P_SEARCH, P2P_IDLE, or P2P_PROVISIONING state. But Extended
Listen timeout being enabled moves to the P2P_LISTEN_ONLY state. Due to
this P2P connection fails if the GO is not found with scans triggerred
before Extended Listen timeout. To avoid this skip Extended Listen
timeout to allow scans during group formation on the current interface.
Signed-off-by: Shivani Baranwal <quic_shivbara@quicinc.com>
Clone sae_pwe configuration value when creating a new P2P group
interface. This is needed to avoid P2P connection failure at the P2P
Client side due to SAE H2E being disabled in the 6 GHz band operation
when a new group interface is created for the connection.
Signed-off-by: Veerendranath Jakkam <quic_vjakkam@quicinc.com>
For the 6 GHz opclass 132, ch for loop only allows non-PSC channels { 1,
9, 17, 25, 33, 41, 49,... } to be used. This does not match the IEEE Std
802.11ax-2021 expectation of a 6 GHz-only AP "should set up the BSS with
a primary 20 MHz channel that coincides with a preferred scanning
channel".
Increase ch by 4 to allow PSC channel to be used for 6 GHz BW40.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
struct wpas_pasn is common to both initiator and responder, so rename it
to pasn_data to avoid the "wpas_" prefix that could be seen as a
reference to wpa_supplicant (PASN initiator).
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
The wpa_s and hapd context are isolated from the PASN initiator and
responder code and this functionality is now in the separate src/pasn
directory. Add option to build libpasn.so with this functionality. This
library can be used by a Wi-Fi Aware component to support NAN pairing
with other devices.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
PASN responder validates auth 1 frame and sends auth 2 frame to the
initiator. It analyses the auth 3 frame and verifies successful
authentication. Wi-Fi Aware modules can reuse this functionality through
a shared library libpasn.so generated from this code. Move the PASN
functionality that is now decoupled from the hapd context into a
separate file in a common directory to make it easier to build such a
library.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
PASN initiator functionality builds auth 1 and auth 3 frames, and
processes auth 2 frame received from the responder. Wi-Fi Aware modules
can reuse this functionality through a shared library libpasn.so
generated from this code. Move the PASN functionality that is now
decoupled from the wpa_s context into a separate file in a common
directory to make it easier to build such a library.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This is helpful in being able to get the functionality needed for SAE
into a separate library (libpasn.so) without needing all of the
ieee802_11.c functionality.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Wi-Fi Aware R4 specification defines Beacon RSNE/RSNXE to be same as
RSNE/RSNXE present in Auth2 frame. So, MIC validation should be done
with the RSNE and RSNXE received in Auth2 frame.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Wi-Fi Aware R4 specification introduces a custom PMKID derived from
Nonce and TAG. This custom PMKID is included in PASN Authentication
frames during pairing verification. So, allow use of a custom PMKID in
PASN frames and validate it using a function handler. Wi-Fi Aware
component that uses libpasn.so should take care of validating the custom
PMKID.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
PTK is stored in the PTKSA cache following a successful PASN handshake,
however AKMP is removed upon a WPA PASN reset. The PASN handshake is
used in the Wi-Fi Aware R4 specification to define the pairing setup
process. KDK is used to generate a new set of keys, while AKMP is
required for key derivation for pairing. So, keep AKMP in the PTKSA
cache.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This makes hostapd use the struct defines from pasn_common.h so that the
same struct is shared with wpa_supplicant.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This is a step towards decoupling the PASN initiator and responder
implemenentation is decoupled from the wpa_s and hapd contexts and
moving to a common folder for better abstraction. Move the struct
wpas_pasn definition to a common file for initiator and responder. The
idea is to provide a library libpasn.so from PASN common code. Include
C++ compatibilty wrapper to extend libpasn.so support for modules using
cpp code base.
This library can be used in applications implementing protocols based on
the PASN handshake.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Introduce a function handler to transmit PASN Authentication frames to
the driver. This removes the wpa_s dependency from PASN functionality
for sending the frames.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Get ssid and derive required information for SAE and FILS tunneled PASN
authentication to remove struct wpa_ssid dependency for PASN initiator.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Before invoking wpas_pasn_start(), derive KDK and secure LTF flags.
Instead of using network id for pasn_start(), use a pointer to the
struct wpa_ssid.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Obtain RSNXE capabilities from the driver flags as well and save them in
the wpas_pasn structure.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Remove wpa_s dependency for processing PASN M2 frame and preparing M3
frame. To hold the context of wpa_s when transmitting M3 frame, add cb
ctx.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This is a step towards allowing the PASN implementation to be used
outside the context of wpa_supplicant.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Store PMKSA cache entry in wpas_pasn and remove wpa_sm dependency to add
an entry to PMKSA cache. This is a step towards allowing the PASN
implementation to be used outside the context of wpa_supplicant.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Use struct wpas_pasn (i.e., &wpa_s->pasn) as the function argument for
the SAE functionality that does not need the full struct wpa_supplicant
context. This is a step towards allowing PASN to be used outside the
context of wpa_supplicant.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Split wpas_pasn_reset() so that the main reset steps can be used without
an instance of struct wpa_supplicant. This is needed to allow PASN
implementation to be used for additional purposes outside the context of
wpa_supplicant.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
For wpa_supplicant_get_new_bss(), wpa_s->current_ssid can be NULL in
some cases. Add a NULL check before accessing it to avoid NULL pointer
dereference errors.
Fixes: 7784964cbe ("MLD STA: Fetch MLO connection info into core wpa_supplicant")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
The initial wpa_drv_get_mlo_info() implementation cleared only the
valid_links information within struct driver_sta_mlo_info before trying
to fetch the information from the driver. While this is likely going to
work fine in practice, this can result in static analyzer warnings on
use of uninitialized memory (e.g., mlo.assoc_link_id could have been
read if wpa_s->valid_links was set to a nonzero value). In any case, it
is better to avoid such unnecessary warnings by clearing the full data
structure before using it.
Fixes: 7784964cbe ("MLD STA: Fetch MLO connection info into core wpa_supplicant")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
The Linux kernel expects to use the AP MLD address as destination and
BSSID addresses for non-Public Action frames when the current
association is MLO capable. Use the MLD addresses to support this.
Signed-off-by: Veerendranath Jakkam <quic_vjakkam@quicinc.com>
EAPOL-Key msg 1/4 indication can be received before association
indication from the driver. For MLO connection, the source address check
of such frames should be against the AP MLD address instead of the
associated link BSSID.
Signed-off-by: Veerendranath Jakkam <quic_vjakkam@quicinc.com>
Fetch the MLO association Link ID info from the driver to the
wpa_supplicant instance of the corresponding MLD STA interface. This
info is needed when setting the MLO connection info to wpa_sm.
Signed-off-by: Veerendranath Jakkam <quic_vjakkam@quicinc.com>
Use a full list of AKM suite selectors that can use SAE authentication
when checking for authentication trigger from the driver.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This allows the P2P client to fast associate if the GO is already in the
scan cache without doing a full scan.
Signed-off-by: Matthew Wang <matthewmwang@chromium.org>
AKM suite selector check was initially implemented with incorrect byte
swapping of the value from the driver (nl80211). Fix this and leave a
workaround option for any potentially deployed device where the driver
might be using the swapped byte order.
Fixes: 5ff39c1380 ("SAE: Support external authentication offload for driver-SME cases")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This can be helpful in trying to figure out unexpected behavior even
though we do not currently really do anything with the inner method
type.
Signed-off-by: Chaoli Zhou <quic_zchaoli@quicinc.com>
Move from this type of constructions:
switch (val) {
case 1:
something;
break;
}
default-action;
into following:
switch (val) {
case 1:
something;
break;
default:
default-action;
break
}
for cases where the switch statement is not expected to contain a full
set of enum values and as such, does not lose value from not having the
default target.
This makes the intent of default behavior clearer for static analyzers like
gcc with -Wswitch-default.
Signed-off-by: Chaoli Zhou <quic_zchaoli@quicinc.com>
Since NetworkManager doesn't support setting ieee80211w to
wpa_supplicant and only support pmf, so override ieee80211w from pmf for
AP mode if ieee80211w not configurated. Do not change behavior for the
P2P GO cases.
Signed-off-by: Chaoli Zhou <quic_zchaoli@quicinc.com>
While the IEEE 802.11 standard does not require MFPR=1, WPA3-Personal
requires PMF to be used with SAE. Use the stronger MFPR=1 configuration
for SAE-without-PSK case, i.e., interpret that as "WPA3-Personal only"
configuration.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Do not try to use variables that are not defined without
CONFIG_IEEE80211R=y and add the forgotten "inline" for the function
wrapper.
Fixes: 5c65ad6c0b ("PASN: Support PASN with FT key derivation")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Compiler throws a warning for using bitwise operation in
wpa_s_setup_sae_pt() function instead of logical AND operation on two
boolean values.
Fixes: e81ec0962d ("SAE: Use H2E unconditionally with the new AKM suites")
Signed-off-by: Sunil Ravi <sunilravi@google.com>
IEEE Std 802.11-2020, 11.8.8 (Selecting and advertising a new channel)
doesn't restrict switching to a channel which is not in the same band as
the current band. But at the same time, there are some issues in
switching between the 2.4/5 GHz bands and the 6 GHz band. So limit the
check to consider the critical use case of switching from a 5 GHz channel
to a 2.4 GHz channel.
Signed-off-by: Sunil Ravi <sunilravi@google.com>
While setting up the P2P GO interface, wpa_supplicant sets the operating
channel width to CHANWIDTH_USE_HT in cases where it fails to set 80 MHz
bandwidth. In the same flow, update the operating class, too, according
to the channel width. This is to avoid setting up the operational
channel width back to 80 MHz from hostapd interface setup.
Signed-off-by: Sunil <sunilravi@google.com>
wpas_p2p_optimize_listen_channel() checks for the state and current ssid
of the interface calling this function. This check prevents the function
from setting the optimized listen channel. Since the listen channel is
stored in global P2P configuration data, do not check the state and
current interface of the caller.
Signed-off-by: Sunil Ravi <sunilravi@google.com>
If wpa_supplicant was build without CONFIG_IEEE80211R, the
exposed key-management capabilities should not include one of the
FT protocols. If someone would use a FT protocol in such situation,
it would fail anyway.
Signed-off-by: Clemens Famulla-Conrad <cfamullaconrad@suse.de>
It is possible to specify -t or -K multiple times. With this the
value isn't boolean anymore and we hit a assert in libdbus
function `dbus_message_iter_append_basic()`, which expect 0 or 1
for DBUS_TYPE_BOOLEAN.
Signed-off-by: Clemens Famulla-Conrad <cfamullaconrad@suse.de>
