Hu Wang [Wed, 8 Aug 2018 03:21:05 +0000 (11:21 +0800)]
nl80211: Do not ignore disconnect event in case of !drv->associated
Commit 3f53c006c7d7362cf715ceaeda92c69d91ea7b63 ('nl80211: Ignore
disconnect event in case of locally generated request') made
wpa_supplicant ignore the next received disconnect event for cases where
wpa_supplicant itself requested a disconnection. This can result in
ignoring a disconnection notification in some cases.
Considering a P2P Client receiving disconnect event from the kernel
after a P2P group is started, drv->ignore_next_local_disconnect is
cleared to 0, then wpa_driver_nl80211_disconnect() will be called during
the removal of the group, in which drv->ignore_next_local_disconnect is
set to 1 by mistake.
Do not allow ignore_next_local_{disconnect,deauth} to be set to 1 if the
driver is not in associated state (drv->associated is 0) to avoid this
type of cases.
DPP: Set group id through DPP_AUTH_INIT or dpp_configurator_params
This enhances DPP_AUTH_INIT, DPP_CONFIGURATOR_SIGN, and SET
dpp_configurator_params to allow optional setting of the DPP groupId
string for a Connector. If the value is not set, the previously wildcard
value ("*") is used by default.
Nachiket Kukade [Fri, 3 Aug 2018 10:28:07 +0000 (15:58 +0530)]
Add new QCA vendor command and attributes to enhance NAN
The payload used for the existing NAN vendor command,
QCA_NL80211_VENDOR_SUBCMD_NAN is a binary blob of data. This
command is not extendable to send additional information. Hence
define a new vendor command QCA_NL80211_VENDOR_SUBCMD_NAN_EXT,
that can carry the binary blob encapsulated within an attribute
and can carry additional attributes to enhance the NAN command
interface. Define additional 3 new attributes for conveying
type of NAN subcmd and channel information.
Nachiket Kukade [Fri, 10 Aug 2018 11:24:09 +0000 (16:54 +0530)]
Correct the documentation of NAN vendor command
Present implementation of NAN vendor command does not use
attribute encapsulation for sending the command from userspace
to the driver, payload is directly sent as is. Attribute
QCA_WLAN_VENDOR_ATTR_NAN is used only for receiving vendor
events in the userspace from the driver. Update the doc as per
this implementation.
Change the ADDBA buffer size attribute type to U16 from U8
The max supported ADDBA buffer size value is 256, so change
the buffer size attribute type to U16 to configure the testbed
device to use the 256 buffer size in ADDBA negotiation in 11ax
testing.
This attribute is used only to configure a testbed device and the old
definition of this attribute was not used in any deployed implementation
hence it is still justifiable to change the definition.
nl80211: Fix sending of WDS STA event to the correct BSS context
The WDS-STA-INTERFACE-ADDED/WDS-STA-INTERFACE-REMOVED events were always
sent to the first BSS instead of the specific BSS that the STA was
connected to in multi-BSS cases. Fix this by using the BSS specific
context pointer.
Fixes: 1952b626ba57 ("hostapd: Add ctrl iface indications for WDS STA interface") Signed-off-by: Bhagavathi Perumal S <bperumal@codeaurora.org>
FILS: Fix FILS connect failures after ERP key invalidation
If the RADIUS authentication server dropped the cached ERP keys for any
reason, FILS authentication attempts with ERP fails and the previous
wpa_supplicant implementation ended up trying to use the same keys for
all consecutive attempts as well. This did not allow recovery from state
mismatch between the ERP server and peer using full EAP authentication.
Address this by trying to use full (non-FILS) authentication when trying
to connect to an AP using the same ERP realm with FILS-enabled network
profile if the previous authentication attempt had failed. This allows
new ERP keys to be established and FILS authentication to be used again
for the consecutive connections.
hostapd: SET ht_capab support for disabling 40 MHz bandwidth
'hostapd_cli SET ht_capab' only checked for [HT40+] or [HT40-] or both
to be present. Based on the offset + or -, secondary_channel is updated
but HT20/VHT20 mode can be brought up only from config file and can't be
done using the SET command when the current HT mode is HT40+ or HT40-.
When managing AP+STA mode from userspace doing hostapd_cli: "disable ->
set channel, ht_capab -> enable" sequence, channel switch from
HT40/VHT40 to HT20/VHT20 was not possible with this SET ht_capab
limitation.
Cover this additional case by resetting secondary_channel to 0 for
HT20/VHT20 when ht_capab has neither [HT40+] nor [HT40-] present.
Jouke Witteveen [Mon, 25 Jun 2018 20:28:50 +0000 (22:28 +0200)]
Fix QoS Mapping ext capab bit setting
Fix the typo in using WPA_DRIVER_FLAGS_QOS_MAPPING to set the QoS Map
bit in Extended Capabilities. The previous implementation ended up
adding this bit even if the driver did not actually indicate support for
the capability.
Jouni Malinen [Fri, 10 Aug 2018 10:51:04 +0000 (13:51 +0300)]
Clarify the TODO comment regarding PMKID KDE in EAPOL-Key msg 1/4
Make it clear that the consideration should be only for the IBSS case
and in infrastructure BSS case, PMKID KDE should not be added due to
risks involved with exposing this to stations that do not know the
passphrase.
WPA: Ignore unauthenticated encrypted EAPOL-Key data
Ignore unauthenticated encrypted EAPOL-Key data in supplicant
processing. When using WPA2, these are frames that have the Encrypted
flag set, but not the MIC flag.
When using WPA2, EAPOL-Key frames that had the Encrypted flag set but
not the MIC flag, had their data field decrypted without first verifying
the MIC. In case the data field was encrypted using RC4 (i.e., when
negotiating TKIP as the pairwise cipher), this meant that
unauthenticated but decrypted data would then be processed. An adversary
could abuse this as a decryption oracle to recover sensitive information
in the data field of EAPOL-Key messages (e.g., the group key).
(CVE-2018-14526)
Jouni Malinen [Thu, 2 Aug 2018 21:40:51 +0000 (00:40 +0300)]
tests: Check both FT and non-FT connections in ap_hs20_ft
Verify that the automatically generated network profile is able to
connect to a non-FT network automatically after having used FT for the
first connection.
Jouni Malinen [Thu, 2 Aug 2018 13:15:14 +0000 (16:15 +0300)]
HS 2.0: Fix hs20-osu-client handling of HomeSP/HomeOIList/<X+>/HomeOI
This node was mapped to a SET_CRED roaming_consortium command with
quotation marks even though this is a hexdump of the OI. Remove the
quotation marks to allow this to be set correctly in the wpa_supplicant
credential.
Define test config vendor attribute to override MU EDCA
Add a new test config QCA vendor attribute to override the MU EDCA
parameters to default values in the driver for test configuration. This
is used for configuring the testbed device.
HS 2.0: Allow CCMP as group cipher for OSEN single SSID case
When OSEN is used in the BSS that is shared both for production data and
OSU uses, the group cipher might be either GTK_NOT_USED (like in Rel 2
OSEN) or CCMP. Modify hs20-osu-client to allow both these group ciphers
to be used when requesting OSEN connection.
HS 2.0: Fix Terms and Conditions Server URL macro replacement
The last character of the STA MAC address got replaced with 0x00 due to
truncation if the @1@ macro was used at the end of the URL. Fix this
case by allocating larger buffer for the URL.
Fixes: d4e39c51f8bb ("HS 2.0: Move Terms and Conditions Server URL generation from AP to AS") Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Define test config QCA vendor attribute for HE MAC padding duration
Add a new wifi test config QCA vendor attribute to configure HE
trigger frame MAC padding duration value in the driver.
This is used for testbed configuration.
Hu Wang [Tue, 26 Jun 2018 02:56:29 +0000 (10:56 +0800)]
nl80211: Indicate interface up only for the main netdev
ifname is checked for interface down event in commit 106fa1e97e29e7514ae324020802c280f993ee48 ("nl80211: Indicate
interface-down event only for the main netdev"). Do the same for
interface up event to avoid unexpected notifications when the master
interface (e.g., a bridge interface) is set up while then WLAN interface
remains down.
nl80211: Allocate nl_sock for NETLINK_ROUTE when already_in_bridge
When we start hostapd having Hotspot 2.0 configuration with interface
already added to bridge interface, addition and deletion of new neighbor
to bridge ip neighbor table fails.
This is since 'bss->added_if_into_bridge' is not set which only allows
'drv->rtnl_sk' (nl_sock for NETLINK_ROUTE) allocation needed for bridge
ip neighbor table.
Add a new bit 'already_in_bridge' and set it when interface is already
added to bridge by some external component. Check this bit in addition
to 'bss->added_if_into_bridge' for 'drv->rtnl_sk' allocation done in
i802_init().
Now 'drv->rtnl_sk' is closed in wpa_driver_nl80211_deinit() regardless of
'bss->added_if_into_bridge' since when we have 'bss->already_in_bridge'
case too, this need to be removed.
brctl show
bridge name bridge id STP enabled interfaces
br0 8000.8efdf006b050 no ap
Peng Xu [Fri, 15 Jun 2018 21:39:32 +0000 (14:39 -0700)]
nl80211: Skip vendor specific DFS offload if upstream mechanism is used
Since the generic DFS offload support flag is added as
NL80211_EXT_FEATURE_DFS_OFFLOAD, only use the vendor command to check
DFS capablity if the flag is not already set.
Define test config QCA vendor attribute for Tx beamformee NSTS
Add a new wifi test config QCA vendor attribute to configure Tx
beamformee number of space-time streams value in the driver.
This is used for testbed configuration.
Jouni Malinen [Fri, 22 Jun 2018 17:22:40 +0000 (20:22 +0300)]
HS 2.0: CoA-Request from Terms and Conditions server
This extends the terms.php implementation of Hotspot 2.0 Terms and
Conditions server to allow it to interact with hostapd(AS) to clear the
filtering rules from the AP. After requesting hostapd to send out the
CoA-Request, terms.php waits for up to 10 seconds to see whether the
current_sessions table gets an update to indicate that filtering has
been successfully disabled.
Jouni Malinen [Fri, 22 Jun 2018 16:32:46 +0000 (19:32 +0300)]
RADIUS: Add DAC implementation in hostapd(AS)
The new DAC_REQUEST control interface command can now be used to request
hostapd to send out Disconnect-Request and CoA-Request packets for an
existing session.
Jouni Malinen [Thu, 21 Jun 2018 17:50:54 +0000 (20:50 +0300)]
HS 2.0: Move Terms and Conditions Server URL generation from AP to AS
This makes it more convenient to generate the URL in a way that
interoperates between different vendors. The AP is simply copying the
already constructed URL as-is from Access-Accept to WNM-Notification.
This means that the HO AAA can generate the URL in a manner that works
for the associated T&C Server without having to coordinate with each AP.
Jouni Malinen [Thu, 21 Jun 2018 16:16:26 +0000 (19:16 +0300)]
HS 2.0: Process Credential/UsernamePassword/EAPMethod nodes in PPS MO
This allows hs20-osu-client to configure wpa_supplicant credential with
a specific EAP method so that roaming consortium OI -based matching can
be used.
Jouni Malinen [Wed, 20 Jun 2018 10:07:38 +0000 (13:07 +0300)]
tests: Add more details on sae_bignum_failure logs
It is easier to understand the hostapd debug log here if each step is
noted there before starting the wpa_supplicant operation. In addition,
it looks safer to process all pending event messages between each step
to avoid running out of any buffer limits.
Jouni Malinen [Wed, 20 Jun 2018 10:04:31 +0000 (13:04 +0300)]
SAE: Allow authentication restart on AP
In the infrastructure BSS case, the AP needs to clear pending SAE state
if a new Commit message is received when already in Committed state.
This allows the non-AP STA to negotiate a new group if it ends up trying
to go through SAE authentication again before the AP side has removed
the previous STA entry.
This fixes an issue where a kernel update changed something in SAE
timing or authentication sequence and started failing the
sae_bignum_failure hwsim test case.
Jouni Malinen [Fri, 15 Jun 2018 18:56:21 +0000 (21:56 +0300)]
tests: P2P and avoid frequencies driver event on 5 GHz
This test case verifies that wpa_supplicant is able to perform CSA to a
VHT80 channel when having to move the GO due to an avoid-frequencies
driver event.
Jouni Malinen [Fri, 15 Jun 2018 18:47:29 +0000 (21:47 +0300)]
P2P: Improve common group channel selection if GO needs to be moved
Prefer channels that support VHT80 (and secondarily, HT40 on the same
band) over other common group channels. If no such channel is found,
prefer any channel that uses the same band so that CSA can be used. This
improves the case where a P2P GO needs to move to another channel and
there is no other reason (e.g., preferred channel from the driver or an
already used channel from a virtual interface sharing the same radio) to
pick a specific channel.
When removing an interface, hostapd_bss_deinit() frees all associated
STAs. If any of the stations is 40MHz intolerant, the cleanup invokes
ht40_intolerant_remove(), that in turn registers a 20->40MHz transition
timer for the last station (ap_ht2040_timeout() function). That timer is
never canceled; once it executes, the interface is gone, most likely
resulting in a segfault when referencing it.
While hostapd_interface_deinit() cancels the transition timer, it does
so before cleaning up STAs. Move the cancellation after STA cleanup to
cancel any timer that was registered during that operation.
Signed-off-by: Martin Willi <martin@strongswan.org>
Handle NL80211_CMD_WIPHY_REG_CHANGE the same way as the
NL80211_CMD_REG_CHANGE event. The wiphy-specific event is generated by
the cfg80211 reg_process_self_managed_hints() function, e.g., when going
through regulatory_set_wiphy_regd_sync_rtnl(). Previously, such events
were ignored completely in hostapd/wpa_supplicant.
Jouni Malinen [Wed, 6 Jun 2018 18:57:58 +0000 (21:57 +0300)]
FT: Derive PMKR0Name/PMKR1Name using SHA-384 with AKM 00-0F-AC:13
The AKM 00-0F-AC:13 is supposed to use cryptographic algorithms
consistently, but the current IEEE 802.11 standard is not doing so for
the key names: PMKID (uses SHA-1), PMKR0Name/PMKR1Name (uses SHA-256).
The PMKID case was already implemented with SHA-384 and this commit
replaces use of SHA-256 with SHA-384 for PMKR0Name/PMKR1Name derivation
to be consistent in SHA-384. While this is not compliant with the
current IEEE 802.11 standard, this is clearly needed to meet CNSA Suite
requirements. Matching change is being proposed in REVmd to get the IEEE
802.11 standard to meet the use case requirements.
Jouni Malinen [Tue, 5 Jun 2018 22:22:01 +0000 (01:22 +0300)]
SAE: Do not drop STA entry on reauthentication in infrastructure BSS
A new SAE Commit message should not be allowed to drop an existing STA
entry since the sender of that Commit message cannot be authenticated
before receiving the Confirm message. This is important in particular
when PMF is used since this would provide a potential new path for
forcing a connection to be dropped.
Fix this by allowing a new SAE Authentication instance to be started
when the old instance is in Accepted state and the new Commit message
does not use the same peer-scalar value (checked in
sae_parse_commit_scalar()). When PMF is used, the AP will use SA Query
procedure when receiving the (Re)Association Request frame. In theory,
that step could be skipped in case of SAE Authentication since the
non-AP STA is demonstrating knowledge of the password. Anyway, there is
no allowance for that exception in the IEEE 802.11 standard, so at least
for now, leave this using SA Query procedure just like any other PMF
case.
Jouni Malinen [Tue, 5 Jun 2018 18:36:56 +0000 (21:36 +0300)]
FT: Support 256-bit IGTK in AP case
hostapd was hardcoded to use 128-bit IGTK in FT protocol (IGTK
subelement in FTE). Extend that to allow 256-bit IGTK (i.e.,
BIP-CMAC-256 and BIP-GMAC-256) to be used as well.
Jouni Malinen [Tue, 5 Jun 2018 18:48:27 +0000 (21:48 +0300)]
tests: Split ap_ft_ap_oom7 into separate test cases
The implementation changes in hostapd FT error path handling in the
follow commit would result in ap_ft_ap_oom7 test case failing. This is
triggered partially by PMF protections and SA Query attempts, so it
looks like it is easier to split each failure case into a separate test
case.
Jouni Malinen [Tue, 5 Jun 2018 17:13:34 +0000 (20:13 +0300)]
FILS: Fix KEK2 use in FT-FILS use cases
When support for KCK2 and KEK2 was added, both keys were derived for
FT-FILS cases, but only KCK2 was actually used. Add similar changes to
use KEK2 to protect GTK/IGTK in FTE with using FT-FILS AKMs.
This fixes AES key wrapping to use the correct key. The change is not
backwards compatible.
Fixes: 2f37387812a5 ("FILS: Add more complete support for FT-FILS use cases") Signed-off-by: Jouni Malinen <j@w1.fi>
Jouni Malinen [Tue, 5 Jun 2018 11:18:41 +0000 (14:18 +0300)]
FT: Fix RRB error path handling
When building an RRB message, a failure in wpa_ft_rrb_lin() calls could
have resulted in trying to free an uninitialized pointer. Fix this by
initializing *packet to NULL before going through the initial steps.
Jouni Malinen [Mon, 4 Jun 2018 12:16:54 +0000 (15:16 +0300)]
FT: Add support for variable length PMK-R1 receiving in RRB
SHA384-based FT AKM uses longer keys, so the RRB receive processing for
push and pull response messages needs to be able to accept variable
length PMK-R1.
Sven Eckelmann [Mon, 7 May 2018 13:24:29 +0000 (15:24 +0200)]
wpa_supplicant: Fix parsing of max_oper_chwidth
The max_oper_chwidth is parsed in wpa_config_set as INT_RANGE (see
ssid_fields). The actual parsing for INT_RANGE is done by
wpa_config_parse_int which can only store the result as full integer.
max_oper_chwidth is stored as u8 (a single byte) in wpa_ssid. This means
that on little endian systems, the least significant byte of the parsed
value are really stored in the max_oper_chwidth. But on big endian
system, the only most significant byte is stored as max_oper_chwidth.
This means that 0 is always stored because the provided range doesn't
allow any other value for systems with multi-byte-wide integers.
This also means that for common systems with 4-byte-wide integers, the
remaining 3 bytes were written after the actual member of the struct.
This should not have influenced the behavior of succeeding members
because these bytes would have been part of the padding between the
members on most systems.
Increasing its size to a full int fixes the write operations outside of
the member and allows to use the max_oper_chwidth setting on big endian
systems.
Fixes: 0f29bc68d18e ("IBSS/mesh: Add support for VHT80P80 configuration") Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Felix Fietkau [Thu, 3 May 2018 10:36:15 +0000 (12:36 +0200)]
mesh: Fix crash with CONFIG_TAXONOMY enabled
wpa_s->ifmsh needs to be allocated using hostapd_alloc_iface() instead
of a direct call to os_zalloc(), otherwise the linked list for station
taxonomy items remains uninitialized, leading to a crash on the first
attempt to traverse that list