If NL80211_CMD_AUTHENTICATE fails due to exiting authentication, forced
deauthentication is used to clear cfg80211 start and a new
authentication command is retried. This resulted in an extra
disconnection event getting delivered to the core wpa_supplicant code
and that could end up confusing state there, e.g., by clearing
connection parameters like wpa_s->key_mgmt and preventing the following
association from reaching proper COMPLETED state with open networks.
Fix this by hiding the unnecessary disconnection event when using the
deauth-for-auth workaround.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
On RF-kill, we should not request the kernel to start a P2P device. In
addition, we should call i802_set_iface_flags() both for a P2P Device
interface and any other interface instead of calling a dedicated
function for each.
Signed-off-by: Moshe Benji <Moshe.Benji@intel.com>
At least in nl80211, broadcast management frames like Probe Request
frames, may be processed multiple times per BSS if multi-BSS is active
and NL80211_CMD_FRAME event is used to deliver them. In the case of
Probe Request frames, hostapd will create multiple redundant Probe
Response frames which are problematic when many BSS are on one channel.
This problem is caused by driver_nl80211 generating an event for
wpa_supplicant_event() for each BSS, and hostapd_mgmt_rx() calls
ieee802_11_mgmt() for each BSS, too.
Fix this by processing broadcast events only for the BSS the driver
intended to. The behavior is not changed for drivers not setting a BSS.
Signed-hostap: Simon Wunderlich <simon@open-mesh.com>
While these were already available from elsewhere in the debug log, it
is convenient to have the values also available at the location where
the actual nl80211 command is issued.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
wpa_supplicant already allowed beacon interval to be configured for AP
mode operations, but this was not passed to the driver for IBSS even
though the same parameter can used for that case. Add this for the
nl80211 driver interface to allow beacon interval to be controlled for
IBSS as well.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This commit adds an option to optimize AP teardown by leaving the
deletion of keys (including group keys) and stations to the driver.
This optimization option should be used if the driver supports stations
and keys removal when stopping an AP.
For example, the optimization option will always be used for cfg80211
drivers since cfg80211 shall always remove stations and keys when
stopping an AP (in order to support cases where the AP is disabled
without the knowledge of wpa_supplicant/hostapd).
Signed-off-by: Moshe Benji <moshe.benji@intel.com>
CONFIG_TESTING_OPTIONS=y build of wpa_supplicant now allows arbitrary
cfg80211 commands to be performed through the new VENDOR ctrl_iface
command by using a special vendor_id ffffffff. The command identifier
(NL80211_CMD_*) is encoded as the subcmd and the attributes in the
hexformatted data area. Response attributes are returned as a hexdump.
For example, this shows a NL80211_CMD_FRAME and a response (cookie
attribute) on a little endian host:
wpa_cli -i wlan0 vendor ffffffff 59 080003004d0000000800260085090000....
0c00580000d7868c0388ffff
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
When an interface is disabled through rtm event, wpa_supplicant's
EVENT_INTERFACE_DISABLED is generated, which in turn, may
completely destroy wpa_driver_nl80211_data struct (drv). This
scenario happens now when P2P GO interface is disabled. Since this
struct may be used later in this function it causes segmentation fault.
Fix it by trying to find drv again in the interface list after
wpa_supplicant's event handling.
Signed-off-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Add a callback to the driver interface that allows vendor specific
commands to be sent. In addition, a control interface command is added
to expose this new interface outside wpa_supplicant:
Vendor command's format:
VENDOR <vendor id> <sub command id> [<hex formatted data>]
The 3rd argument will be converted to binary data and then passed as
argument to the sub command.
This interface is driver independent, but for now, this is only
implemented for the nl80211 driver interface using the cfg80211 vendor
commands.
Signed-off-by: Beni Lev <beni.lev@intel.com>
wpa_driver_nl80211_if_remove() checks bss->if_added before deleting an
interface, which is 0 for the first BSS. The only part of
wpa_driver_nl80211_if_remove() that should get called for WDS STA
interfaces is the call to nl80211_remove_iface(), which can be pulled in
here directly.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
no_offchannel_tx=1 driver parameter can now be used to force the older
remain-on-channel -based offchannel TX design to be used with
mac80211_hwsim. This can be used to increase test coverage with the
hwsim test cases.
Signed-off-by: Jouni Malinen <j@w1.fi>
Subscription remediation notification WNM-Notification Request is now
shown in the following way in wpa_supplicant control interface:
<3>HS20-SUBSCRIPTION-REMEDIATION http://example.com/foo/
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Many drivers support operation without GTK configured, but most (if any)
today do not advertise this. Handle this by skipping GTK cipher suite
configuration if the driver did not advertise support in order to work
around cfg80211 validation steps.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Use nl80211_set_iface_id() in wpa_driver_nl80211_get_hw_feature_data(),
as otherwise the function fails for a P2P Device interface (which does
not have a netdev associated with it).
Signed-hostap: Ilan Peer <ilan.peer@intel.com>
This uses the new nl80211 attributes to allow the connect command to
provide bssid and freq hints to the driver without limiting roaming to
the specific BSS/frequency. This can be used by drivers that perform
internal BSS selection (WPA_DRIVER_FLAGS_BSS_SELECTION) as a candidate
for initial association.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
Commit b1f625e0d8 extended
ap_scan_as_station to be able to distinguish between AP and P2P GO
iftypes. However, it did this in a way that completely lost the original
mode because drv->nlmode had already been replaced with the station
mode. Fix this by storing the correct old mode.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Commit 795baf773f ('hostapd: Filter
channel list updated events after country code change') uses the
EVENT_CHANNEL_LIST_CHANGED data pointer, but it updated only one of the
callers to provide that data. NL80211_CMD_REG_BEACON_HINT event was
still sending the event without the initiator data and resulted in NULL
pointer dereference, e.g., if a scan was run while hostapd was running
and the driver was in world roaming state and enabled a channel for
active scans.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
The original RSSI filter semantics for scheduled scan were
really confusing - a separate matchset was created, but it
wasn't actually treated as a separate matchset in the kernel
but rather used as the global RSSI value. The RSSI matchset
thus behaved like an RSSI filter outside of the matchsets,
being ANDed rather than ORed (as normal between matchsets.)
To make this less confusing, I changed the kernel API a bit
to actually treat the RSSI inside each matchset properly,
but keeping it compatible with the old approach by using a
matchset with only an RSSI value as the default for all the
other matchsets, and adding it as a separate matchset only
if it's the only one.
The proper way for wpa_supplicant to do this then would be
to add the RSSI to each SSID matchset, and only add another
matchset without SSID if there's none with.
However, to keep compatibility with older kernels, always
keep the non-SSID matchset and only add the RSSI to all the
other matchsets. This gets close to the desired behaviour,
the only difference would be that we shouldn't add the RSSI
matchset if there are others, but stays compatible with old
and new kernels, as new ones ignore the RSSI-only matchset
if there are others and those others have an RSSI.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
It is possible for channel switch notification to be missing channel
type attribute. This is true at least for VHT80. This led to
iface->conf->secondary_channel being set to 0. This in turn made
subsequent DFS-triggered CSA to fail due to invalid frequency
parameters.
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
This checks if the NL80211_ATTR_IFINDEX attribute is present before
trying to get the value of interface index.
Signed-hostap: Shital Jaju <shitalj@broadcom.com>
This adds a QCA vendor specific nl80211 event to allow the driver to
indicate a list of frequency ranges that should be avoided due to
interference or possible known co-existance constraints. Such
frequencies are marked as not allowed for P2P use to force groups to be
formed on different channels.
If a P2P GO is operating on a channel that the driver recommended not to
use, a notification about this is sent on the control interface and
upper layer code may decide to tear down the group and optionally
restart it on another channel. As a TODO item, this could also be changed
to use CSA to avoid removing the group.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
This file is used as a registry of identifier assignments from the
Qualcomm Atheros OUI 00:13:74 for purposes other than MAC address
assignment. One of the first uses will be for nl80211 vendor
commands/events which is reason for the preparation change in
driver_nl80211.c
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
When GAS is used with PMF negotiated, Protected Dual of Public Action
frames are expected to be used instead of Public Action frames, i.e.,
the GAS/ANQP frames are expected to be encrypted. Conver Public Action
GAS queries to use Dual of Public Action frame if PMF has been
negotiated with the AP to which the frame is being sent.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Fix sizeof in a validity comparison of nl80211_vendor_cmd_info size. The
incorrect version happened to work on 64-bit builds due the structure
being eight octets, but this was incorrect and would not used with
32-bit builds.
Signed-hostap: Max Stepanov <Max.Stepanov@intel.com>
Replace channel_switch_supported flag of the
wpa_driver_nl80211_data by WPA_DRIVER_FLAGS_AP_CSA inside
wpa_driver_capa.flags. It makes more sense and also can
be accessed by wpa_supplicant.
Signed-hostap: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
If the BSS table within wpa_supplicant is flushed, request the driver to
flush its own scan result table during the next scan. This can avoid
unexpected old BSS entries showing up after BSS_FLUSH or FLUSH command
in cases where the driver may maintain its internal cache of scan
results (e.g., cfg80211 BSS table persists at least for 15 seconds).
In addition to doing this automatically on BSS_FLUSH/FLUSH, a new SCAN
command argument, only_new=1, can be used to request a manual scan
request to do same. Though, it should be noted that this maintains the
BSS table within wpa_supplicant. BSS_FLUSH followed by SCAN command can
be used to clear all BSS entries from both the driver and
wpa_supplicant.
Signed-hostap: Jouni Malinen <j@w1.fi>
Print into the debug log the list of vendor commands and events that the
driver supports. In addition, add a generic handler for vendor events.
This can be extended for each vendor/subcmd.
Signed-hostap: Jouni Malinen <j@w1.fi>
use_monitor=1 and force_connect_cmd=1 driver parameters can now be used
to force older monitor interface design and the connect API (SME in
driver) to increase hwsim testing coverage.
Signed-hostap: Jouni Malinen <j@w1.fi>
If Action frame registration in nl80211_mgmt_subscribe_non_ap() failed
for any frame type, the previous implementation skipped
nl80211_mgmt_handle_register_eloop() call. This is not desirable since
none of the Action frame types could be received and even worse, the
following nl80211_destroy_eloop_handle() call for nl_mgmt would likely
result in crashing the process due to the ELOOP_SOCKET_INVALID XOR
operation. This could be triggered at least in a P2P group interface
startup failure case.
Signed-hostap: Jouni Malinen <j@w1.fi>
Previously, it was possible for bss_info_handler() to end up dropping a
newer scan result entry if there were two entries with the same BSSID
and SSID (i.e., only frequency was different) and we were not associated
with either. This could happen, e.g., in some P2P use cases where device
discovery may happen on different channels. Fix this by checking the age
of the scan entries as well to prefer the most recent response.
Signed-hostap: Jouni Malinen <j@w1.fi>
Most of the attributes to these commands are identical and there is no
need to maintain two copies of the same functionality.
Signed-hostap: Jouni Malinen <j@w1.fi>
WPA_CIPHER_* and CIPHER_* are used for the exact same set of cipher
suites with the main difference being that the WPA_CIPHER_* version is
suitable to be used as a bitfield. Similarly, WPA_KEY_MGMT_* and
KEY_MGMT_* have similar design for AKMs. There is no need to maintain
two separate copies of the definitions since the bitfield compatible
version can be used for both needs. Get rid of the CIPHER_* and
KEY_MGMT_* versions to clean up the implementation by getting rid of
unnecessary mapping functions.
Signed-hostap: Jouni Malinen <j@w1.fi>
NL80211_ATTR_CONTROL_PORT was previously set only for
NL80211_CMD_ASSOCIATE, but it should also be set when using
NL80211_CMD_COMMAND (driver-based SME) even though none of the current
non-mac80211 drivers use this.
Signed-hostap: Jouni Malinen <j@w1.fi>
This driver event was used separately for some Action frames, but all
the driver wrappers converted to this from information that would have
been enough to indicate an EVENT_RX_MGMT event. In addition, the
received event was then converted back to a full IEEE 802.11 management
frame for processing in most cases. This is unnecessary complexity, so
get rid of the extra path and use EVENT_RX_MGMT for Action frames as
well as other management frame subtypes.
Signed-hostap: Jouni Malinen <j@w1.fi>
Action frame RX report through EVENT_RX_ACTION did not indicate whether
the frame was protected or not even though that information is available
in mlme_event_mgmt(). hostapd_rx_action() has a workaround for setting
the protected flag for SA Query frames, but that did not apply for other
frames, like FT Action. This broke FT-over-DS when PMF is enabled with
newer kernel versions (i.e., the ones that do not use monitor interface
for receiving management frames).
Signed-hostap: Jouni Malinen <j@w1.fi>
This adds initial parts for supporting the new GCMP-256, CCMP-256,
BIP-GMAC-128, BIP-GMAC-256, and BIP-CMAC-256 cipher suites.
Signed-hostap: Jouni Malinen <j@w1.fi>
Do not assume the driver supports QoS Mapping, but instead, advertise
support for this only if CONFIG_INTERWORKING is defined and driver
indicates support for configuring QoS Map.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
The driver_nl80211.c changes are needed to avoid compiler warnings
with two frequency attributes pointing to the same value.
Signed-hostap: Jouni Malinen <j@w1.fi>
Adds support for VHT by parsing bandwidth and center_freq{1,2}.
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
Signed-hostap: Janusz Dziedzic <janusz.dziedzic@tieto.com>
Commit 390e489c0d extended hostapd BSS
interface removal capabilities. However, it ended up causing a
regression in wpa_supplicant P2P group interface handling. That P2P
group interface is removed through another driver_nl80211.c BSS context
and as such, the bss->added_if is not set. Fix this by verifying whether
the request is for another ifindex and if so, removing the interface
even if added_if is not marked.
Signed-hostap: Jouni Malinen <j@w1.fi>
This is a mechanism used in Android to extend driver interface in vendor
specific ways. This is included only for the purpose of Android
compatibility. Proper interface commands should be used for any new
functionality.
Signed-hostap: Jouni Malinen <j@w1.fi>
Android uses a vendor specific library for implementing couple of driver
interface functions. Add the wrapper code to allow that mechanism to be
used.
Signed-hostap: Jouni Malinen <j@w1.fi>
We were not filtering the EVENT_CHANNEL_LIST_CHANGED events based on the
regulatory hint initiator. So wait for EVENT_CHANNEL_LIST_CHANGED event
after our own change was triggered even when regulatory hint initiator
was the driver. This could result in the wait for the channel list to be
updated to be terminated before the real change has occurred and as
such, old channel list remaining in use when configuring
hostapd/wpa_supplicant country parameter. Fix this by filtering the
hints according to the initiator and only regulatory hints initiated by
user will be used to stop the wait.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
If hostapd was started with the interface already in AP mode, leave the
interface in AP mode on deinit instead of unconditionally forcing it to
station mode.
Signed-hostap: Jouni Malinen <j@w1.fi>
It is common to build wpa_supplicant with AP mode support and it is
justifiable to clean up driver_nl80211.c by removing the conditional
build blocks based on hostapd vs. wpa_supplicant builds.
Signed-hostap: Jouni Malinen <j@w1.fi>
This gets rid of some ifdef HOSTAPD constructs and shares more
of the initialization code between hostapd and wpa_supplicant.
Signed-hostap: Jouni Malinen <j@w1.fi>
These variables were originally used in hostapd to clear HT channel
information when exiting. However, that functionality was lost with
commit f019981aee when moving to a common
code for setting the channel. Taken into account that no one seems to
have missed this functionality over the last four years, it seems safe
to drop this rather than try to fix the old hostapd behavior.
Signed-hostap: Jouni Malinen <j@w1.fi>
system/core/libnl_2 does not include nl_socket_set_nonblocking(), so
need to implement that to fix the build.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
If hostapd is requested to set the country code and the previous country
code differs from the new one, the channel list information from the
driver may change. This change may not be instant, so wait for an
EVENT_CHANNEL_LIST_CHANGED event before continuing interface setup with
fetching of the channel list information. This fixes issues where the
selected channel is not available based on the previous regulatory data
and update through CRDA takes some time.
Signed-hostap: Jouni Malinen <j@w1.fi>
While these attributes may be expected to be present always, this needs
to be verified within driver_nl80211.c since we cannot depend on the
kernel/driver working correctly.
Signed-hostap: Jouni Malinen <j@w1.fi>
When removing and re-adding the first wlan# netdev to hostapd
dynamically, the netdev is already present and should not be removed and
re-added to maintain its state as not-added-by-hostapd so that it does
not get removed automatically.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
This moves the vif added check from core hostapd to the driver wrapper
(only driver_nl80211.c uses this) and reorders operations a bit to allow
the first BSS (vif) to be removed from a multi-BSS setup.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
The first user of monitor interface was not counted and that could
result in the monitor interface getting removed if the initial interface
was removed from a multi-BSS setup.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Later on, we'll consider the availability of these
channels when starting P2P GO with VHT support.
Signed-hostap: Eliad Peller <eliadx.peller@intel.com>
Previously the Country element contained the max TX power the local
hardware was capable of. Change this to just use the regulatory limit.
Signed-hostap: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
To avoid a problem where the beacon socket occasionally
blocks, mark any sockets on the eloop as non-blocking.
The previous patch reordered the code to never send a
command after a socket was put on the eloop, but now also
invalidate the nl handle pointer while it's on there.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
Abstract the handling of sockets on the eloop to avoid
destroying sockets still on the eloop and also to allow
the next patch to mark the socket non-blocking.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The IBSS code registers the bss nl_mgmt socket for auth
frames when the join event happens, but that is too late
as then the socket is already on the eloop, which could
cause problems when other events are received at the
same time as the registration is done.
Move the auth frame registration to the initial setup
before the socket is put onto the eloop.
Signed-hostap: Johannes Berg <johannes.berg@intel.com>
The allocation of new_bss and its use was separated by a lot of code in
this function. This can be cleaned up by moving the allocation next to
the use, so that this all can be within a single #ifdef HOSTAPD block.
The i802_check_bridge() call was outside type == WPA_IF_AP_BSS case, but
in practice, it is only used for WPA_IF_AP_BSS (and if used for
something else, this would have resulted in NULL pointer dereference
anyway).
Signed-hostap: Jouni Malinen <j@w1.fi>
Incorrect nla_get variants were used to get event type and frequency.
Kernel passes both as u32. This caused issues on tinynl/big-endian hosts
- CAC finished was treated as radar detection and frequency was 0.
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
It is possible for a vif netdev to be removed by something else than
hostapd and if that happens for a virtual AP interface, if_remove()
handler should still free the local data structure to avoid memory leaks
if something external removes a netdev.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Previously, the send_mlme->send_frame->send_frame_cmd path that could be
used when a GO sends an offchannel Action frame ended up not updating
drv->send_action_cookie. This can result in an issue with not being able
to cancel wait for the response, e.g., in invitation-to-running-group
case.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Indicate support for QoS Mapping and configure driver to update the QoS
Map if QoS Map Set elements is received from the AP either in
(Re)Association Response or QoS Map Configure frame.
This commit adds support for receiving the frames with nl80211 drivers,
but the actual QoS Map configuration command is still missing.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
STATUS-DRIVER command can now be used to fetch driver interface status
information. This is mainly for exporting low-level driver interface
information for debug purposes.
Signed-hostap: Jouni Malinen <j@w1.fi>
TX frequency gets lost when going through the monitor send MLME option
and this resulted in P2P operations like invitation from a GO failing
when the driver needs monitor socket, but would support offchannel TX.
Fix this by using frame_cmd path instead in case the monitor socket
would have been hit for action frame TX.
Signed-hostap: Jouni Malinen <j@w1.fi>
Previously, IBSS mode (NL80211_IFTYPE_ADHOC) was left in drv->nlmode
when leaving IBSS. This causes issues for send_mlme() handler for P2P
Probe Response transmission in Listen state. Fix this by clearing nlmode
back to NL80211_IFTYPE_STATION on leaving IBSS so that following P2P
operations can be executed correctly. Previously, this was fixed only
when the next authentication/association attempt in station mode
occured.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
This can be useful in figuring out how drv->last_mgmt_freq gets set
to debug issues with P2P frames being sent on incorrect channel.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
This adds survey dump support for all frequencies
and for specific desired frequencies. This will later
be used by ACS code for spectrum heuristics.
Signed-hostap: Michal Kazior <michal.kazior@tieto.com>
It is possible that when trying to remove a dynamically added interface,
changing its type to station mode is not possible (since the kernel does
not support so in its interface combinations).
Since P2P interfaces are always dynamically added, avoid changing their
type to station in the deinit_ap() and deinit_p2p_client() nl80211
callbacks, assuming that the interface is about to be removed.
Signed-hostap: Ilan Peer <ilan.peer@intel.com>
Commit 8e12685c43 replaced call to
nl80211_mgmt_unsubscribe() on the deinit path with a
wpa_driver_nl80211_set_mode() call. This is not enough to unregister the
bss->nl_mgmt read socket in all cases. Fix this by unconditionally
unsubscribing from the nl80211 events after having change mode to
station.
Signed-hostap: Jouni Malinen <j@w1.fi>
drv->assoc_freq was not updated during the connect command (neither
during the command's invocation nor after getting the event) unlike with
auth/assoc case where assoc_freq is updated. This resulted in
nl80211_get_link_noise() (and any other function for that matter) using
the improper drv->assoc_freq value with drivers that use the connect
API. Fix this by updating drv->assoc_freq on connect command and when
fetching the frequency from the driver.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
