This introduces a P2P module callback function that will be used to
create the pending P2PS group after sending PD Response and receiving
ACK status for it.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
When the peer device is trying to form a new group despite having
old persistent group with same roles, remove the stale persistent
group.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This adds new wpa_supplicant control interface commands
P2P_ASP_PROVISION and P2P_ASP_PROVISION_RESP.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This extends wpas_p2p_prov_disc() implementation to accept P2PS
parameters. None of the callers are yet using this functionality; the
following commit introduces a user.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This adds a data structure for storing P2PS PD information and code to
add the related attributes into PD Request. The actual operation to
trigger this behavior will be added in a separate commit.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Add support to parse received GAS requests for ASP services and
prepare GAS responses accordingly.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This extends the P2P_SERV_DISC_REQ control interface command with a new
"asp" service type. This takes service id, service string, and optional
service info as arguments.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
In addition, add a new P2P_SERVICE_REP command that can be used to
replace existing ASP advertisements.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This provides additional WPS definitions and rules for negotiating use
of P2PS default PIN configuration method.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
P2PS seek strings can now be specified in the P2P_FIND control interface
command with one or more optional "seek=<str>" parameters.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
It was possible for a Action frame sequence completion to stop an
ongoing offchannel remain-on-channel operation unexpectedly in cases
where TX operation was offloaded to the driver and such an operation
happened to occur during a previously started remain-on-channel (e.g.,
for P2P listen state).
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
It was possible for the previously requested remain-on-channel for P2P
listen state to get canceled before having received driver event
indicating start of that remain-on-channel operation. In such a case,
the event was able to trigger P2P module to start processing listen
timeout even though there was not supposed to be a following listen
operation anymore. Skip the driver event if we are not waiting for a new
listen state.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
When noise floor measurements are not available, compute SNR
using default values for the noise floor. This helps steer us
towards 5 GHz BSSes in high signal strength environments.
In more detail...
Existing code prefers a 5 GHz BSS when the 5 GHz BSS's signal
strength is "close" to that of the 2.4 GHz BSS, or when both SNRs
are large. However, the mwifiex driver does not provide noise
floor measurements, so we can't compute SNRs.
Because mwifiex doesn't provide NF measurements, the "large SNR"
code wasn't effective. By using default values for the noise floor,
we can again compute SNRs, and decide that the SNR is high enough
that we shouldn't worry about the exact difference in SNR.
The default noise floor values (one for 2.4 GHz, and one for 5 GHz)
were chosen by measurement in a noisy environment, so they should be
conservative.
Note that while this patch is motivated by mwifiex, it affects
ath9k as well. Although ath9k provides noise floor measurements
in general, it will sometimes fail to provide a measurement for
one or more specific channels.
As a result of this patch, we'll always compare BSSes based on SNR
(either measured or estimated), rather than sometimes comparing
based on signal strength. ("Always" assumes that the
WPA_SCAN_LEVEL_DBM flag is set. It is for mwifiex and ath9k.)
While there:
- fix a whitespace issue (spaces -> tab)
- clean up existing comments
- update dump_scan_res to indicate whether the noise floor is
measured, or default
Signed-hostap: mukesh agrawal <quiche@chromium.org>
For wired IEEE 802.1X authentication, phase1="allow_canned_success=1"
can now be used to configure a mode that allows EAP-Success (and
EAP-Failure) without going through authentication step. Some switches
use such sequence when forcing the port to be authorized/unauthorized or
as a fallback option if the authentication server is unreachable. By
default, wpa_supplicant discards such frames to protect against
potential attacks by rogue devices, but this option can be used to
disable that protection for cases where the server/authenticator does
not need to be authenticated.
When enabled, this mode allows EAP-Success/EAP-Failure as an immediate
response to EAPOL-Start (or even without EAPOL-Start) and EAP-Success is
also allowed immediately after EAP-Identity exchange (fallback case for
authenticator not being able to connect to authentication server).
Signed-off-by: Jouni Malinen <j@w1.fi>
wpa_supplicant used to request user to re-enter username/password if the
server indicated that EAP-MSCHAPv2 (e.g., in PEAP Phase 2)
authentication failed (E=691), but retry is allowed (R=1). This is a
reasonable default behavior, but there may be cases where it is more
convenient to close the authentication session immediately rather than
wait for user to do something.
Add a new "mschapv2_retry=0" option to the phase2 field to allow the
retry behavior to be disabled. This will make wpa_supplicant abort
authentication attempt on E=691 regardless of whether the server allows
retry.
Signed-off-by: Jouni Malinen <j@w1.fi>
Some packages don't install its headers in the default directory
(e.g.: In Arch Linux libiberty and libn13 includes are installed)
in their own subdirectory under /usr/include) and the build fails
trying to find the headers.
This patch will allow passing extra CFLAGS values without discarding
the assignments made in the Makefile. The CFLAGS values in the Makefile
are ignored, if defined directly in the make command line.
Signed-off-by: Roger Zanoni <roger.zanoni@openbossa.org>
Linux kernel commit 576eb62598f10c8c7fd75703fe89010cdcfff596 ('bridge:
respect RFC2863 operational state') from 2012 introduced a regression
for using wpa_supplicant with EAPOL frames and a station interface in a
bridge. Since it does not look like this regression is going to get
fixed any time soon (it is already two years from that commit and over
1.5 from a discussion pointing out the regression), add a workaround in
wpa_supplicant to avoid this issue.
The wpa_supplicant workaround uses a secondary packet socket to capture
all frames (ETH_P_ALL) from the netdev that is in a bridge. This is
needed to avoid the kernel regression. However, this comes at the price
of more CPU load. Some of this is avoided with use of Linux socket
filter, but still, this is less efficient than a packet socket bound to
the specific EAPOL ethertype. The workaround gets disabled
automatically, if the main packet socket interface on the bridge
interface turns out to be working for RX (e.g., due to an old kernel
version being used or a new kernel version having a fix for the
regression). In addition, this workaround is only taken into use for the
special case of running wpa_supplicant with an interface in a bridge.
Signed-off-by: Jouni Malinen <j@w1.fi>
Currently, there is a race in open mesh networks where mesh STA A
receives a beacon from B and sends a peering open frame to initiate
peering. STA B, having not yet received a beacon from A and thus
created the corresponding station entry, will ignore all such open
frames. If the beacon interval is sufficiently long then peering
will not succeed as a result.
In fact B can simply create the station entry when the popen is
received, as is done in Linux's in-kernel MPM, avoiding the issue.
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Previously, we would only free the station entry if a peering close
frame was received (freeing the station entry causes the kernel to
start sending peer candidate events again when suitable beacons are
received, triggering peering or authentication to restart).
The end result is the same in any case regardless of close reason:
if we leave holding state then peering has started again, so go
ahead and remove the station in all cases.
Signed-off-by: Bob Copeland <me@bobcopeland.com>
The plink_state exists both wpa_supplicant and kernel. Synchronize them
with wpa_mesh_set_plink_state().
Signed-off-by: Kenzoh Nishikawa <Kenzoh.Nishikawa@jp.sony.com>
Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
Drop the unneeded 'attempt' argument. This was originally used for
indicating an aborted PMKID caching attempt, but a fix in 2006 removed
the only such user and since that time, only attempt == 1 has been used.
Signed-off-by: Jouni Malinen <j@w1.fi>
This replaces the implementation in aes-wrap.c and aes-unwrap.c with
OpenSSL AES_wrap_key() and AES_unwrap_key() functions when building
hostapd or wpa_supplicant with OpenSSL.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use OpenSSL HMAC_* functions to implement HMAC-MD5 instead of depending
on the src/crypto/md5.c implementation.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Connection with a PMF enabled AP will fail if we try to negotiate PMF
while the local driver does not support this. Since pmf=1 does not
require PMF for a successful connection, it can be ignored in such a
case to avoid connectivity issues with invalid configuration. This makes
it somewhat easier to allow upper layer programs to use pmf=1 default
regardless of driver capabilities.
Signed-off-by: Jouni Malinen <j@w1.fi>
This modifies struct wpa_ptk to allow the length of KCK and KEK to be
stored. This is needed to allow longer keys to be used, e.g., with
Suite B 192-bit level.
Signed-off-by: Jouni Malinen <j@w1.fi>
This extends GET_CAPABILITY command to allow the supported group
management frame cipher suites to be listed.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
There are lots of reasons we can fail to match an EAP credential, so add
logging to help determine why it is happening in various cases.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signal strength was sorted lexically rather than numerically, which
put "-100 dBm" before "-50 dBm" if sorted in descending order.
This change fixes that. It also treats frequency in the same
manner, preparing it for the IEEE 802.11ah.
Signed-off-by: Adrian Nowicki <adinowicki@gmail.com>
This new wpa_supplicant configuration parameter can be used to force
passive scanning to be used for most scanning cases at the cost of
increased latency and less reliably scans. This may be of use for both
testing purposes and somewhat increased privacy due to no Probe Request
frames with fixed MAC address being sent out.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The ANQP verification/parsing is done only after the GAS_DONE indication
is sent over the control interface. This means that in case the ANQP
parsing fails there is no indication to the upper layers. Add an
ANQP-QUERY-DONE event that reports the status of the ANQP parsing.
Signed-off-by: ASHUTOSH NARAYAN <ashutoshx.narayan@intel.com>
The icon files and the osu-providers.txt that are generated may not have
proper permission for external programs to access. Set the access
permissions to the same as the permissions for osu_dir.
Signed-off-by: ASHUTOSH NARAYAN <ashutoshx.narayan@intel.com>
In wpa_cli_cmd_interface(), try_connection(), and main(), ctrl_ifname
may be NULL and could be dereferenced depending on the printf()
implementation.
Signed-off-by: Eytan Lifshitz <eytan.lifshitz@intel.com>
Stop any ongoing P2P listen/find flow before starting invitation flow.
This was partially handled in p2p_invite() that called p2p_find(), but
this did not cleanly handle cases such as long_listen.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
A control interface call to flush the current state used the
current wpa_s to clear the P2P state even though it might not
be the interface controlling the P2P state.
Fix it by using the correct interface to flush the P2P state.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Publish support for Tx STBC in the HT capabilities of a GO and AP
controlled by wpa_supplicant in case the hardware supports it.
Signed-off-by: Eyal Shapira <eyalx.shapira@intel.com>
external_scan_running should be common to all interfaces that share a
radio. This fixes a case where external_scan_running was set on a single
interface, but did not block scan on other interfaces.
Signed-off-by: Ben Rosenfeld <ben.rosenfeld@intel.com>
Clear the reattach flags, in case a connection request did not trigger a
scan. This needs to be done to avoid leaving the reattach flag set for
the next scan operation which may not have anything to do with the
specific request that could have been optimized using the single-channel
single-SSID scan.
Signed-off-by: Ben <ben.rosenfeld@intel.com>
Use interface's own NFC configuration instead of parent's one
to support a P2P device dedicated interface.
Signed-off-by: David Spinadel <david.spinadel@intel.com>
The length should be 14 and not 10.
The current situation causes failure during parsing of the command.
Signed-off-by: Matti Gottlieb <matti.gottlieb@intel.com>
This restores some of the pre-radio work behavior for scanning by
retrying scan trigger if the driver rejects it (most likely returning
EBUSY in case of nl80211-drivers). Retry is indicated in the
CTRL-EVENT-SCAN-FAILED event with "retry=1".
For manual scans (e.g., triggered through "SCAN" control interface
command), no additional retries are performed. In other words, if upper
layers want to retry, they can do so based on the CTRL-EVENT-SCAN-FAILED
event.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
For CONFIG_TESTING_OPTIONS=y builds, add a new test parameter than can
be used to trigger various error cases within wpa_supplicant operations
to make it easier to test error path processing. "SET test_failure
<val>" is used to set which operation fails. For now, 0 = no failures
and 1 = scan trigger fails with EBUSY. More operations can be added in
the future to extend coverage.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Current mesh code uses ap_max_inactivity as inactivity timer. This patch
makes it configurable.
There is another mesh inactivity timer in mac80211. The timer works even
if user_mpm=1. So this patch sets the max value to the timer for
workaround.
Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
It is easier to debug issues related to the wpa_supplicant control
interfaces being left behind in attached state when the debug log file
can be used to determine whether a specific monitor socket was a global
or per-interface one.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit 17b8995cf5 ('Interworking: Try to
use same BSS entry for storing GAS results') added a mechanism to try to
pair GAS request and response to a single BSS entry to cover cases where
multiple BSS entries may exists for the same BSSID. However, that commit
did not cover the Hotspot 2.0 ANQP elements. Extend this mechanism to
all ANQP elements. This can help in cases where information in the
Hotspot 2.0 specific ANQP elements got lost if a hidden SSID or some
other reason of duplicated BSS entries was present while doing ANQP
fetches.
Signed-off-by: Jouni Malinen <j@w1.fi>
Move the wpa_bss_flush() call to the end of the function to allow any
pending user of a BSS entry to be cleared before removing the unused
entries. There were number of cases where BSS entries could have been
left in the list and this resulted in some hwsim test failures.
Signed-off-by: Jouni Malinen <j@w1.fi>
This can be helpful in figuring out why the driver was requested to
flush its scan results prior to starting a new scan.
Signed-off-by: Jouni Malinen <j@w1.fi>
It was possible for interworking_find_network_match() to find a possible
BSS match in a case where more thorough checks in
wpa_supplicant_select_bss() reject network. This itself is fine, in
general, but when combined with wpa_supplicant_fast_associate()
optimization and auto_interworking=1, this resulted in a busy loop of up
to five seconds and a possible stack overflow due to recursion in that
loop.
Fix this by limiting the Interworking wpa_supplicant_fast_associate()
call to be used only once per scan iteration, so that new scan
operations can be completed before going through the scan results again.
Signed-off-by: Jouni Malinen <j@w1.fi>
Reduce maximum stack use by starting next ANQP fetch operation from an
eloop callback rather than calling interworking_next_anqp_fetch()
directly from interworking_start_fetch_anqp(). This avoids issues that
could potentially make the process run out of stack if long loops of
ANQP operations are executed in cases where automatic Interworking
network selection is used and scan results do not have a full match for
a network.
Signed-off-by: Jouni Malinen <j@w1.fi>
For Interworking connection to work, the SSID of the selected BSS needs
to be known to be able to associate with the AP. It was possible for the
scan results to include two BSS entries matching the BSSID when an
earlier scan with that AP has shown a hidden SSID configuration (e.g.,
when running hwsim test cases, but at least in theory, this could happen
with real use cases as well). When that happened, the incorrect BSS
entry may not have included RSN configuration and as such, it would get
rejected for Interworking connection.
Fix this by confirming that the selected BSS entry has a real SSID. If
not, try to find another BSS entry matching the same BSSID and use that,
if found with an SSID.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This is similar with domain_suffix_match, but required a full match of
the domain name rather than allowing suffix match (subdomains) or
wildcard certificates.
Signed-off-by: Jouni Malinen <j@w1.fi>
A new "CTRL-EVENT-EAP-PEER-ALT depth=<i> <alt name>" event is now used
to provide information about server certificate chain alternative
subject names for upper layers, e.g., to make it easier to configure
constraints on the server certificate. For example:
CTRL-EVENT-EAP-PEER-ALT depth=0 DNS:server.example.com
Currently, this includes DNS, EMAIL, and URI components from the
certificates. Similar information is priovided to D-Bus Certification
signal in the new altsubject argument which is a string array of these
items.
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes D-Bus network profile Set(Properties) clear cached EAP data
similarly to how SET_NETWORK does for control interface.
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes it easier for upper layer applications to get information
regarding the server certificate without having to use a special
certificate probing connection. This provides both the SHA256 hash of
the certificate (to be used with ca_cert="hash://server/sha256/<hash>",
if desired) and the full DER encoded X.509 certificate so that upper
layer applications can parse and display the certificate easily or
extract fields from it for purposes like configuring an altsubject_match
or domain_suffix_match.
The old behavior can be configured by adding cert_in_cb=0 to
wpa_supplicant configuration file.
Signed-off-by: Jouni Malinen <j@w1.fi>
The new VENDOR_ELEM value 13 can now be used to add a vendor element
into all (Re)Association Request frames, not just for P2P use cases like
the previous item was for.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This makes it easier for upper layer programs to figure out whether the
wpa_supplicant and and the driver supports mesh.
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes it easier for upper layer programs to figure out whether the
wpa_supplicant and and the driver supports SAE.
Signed-off-by: Jouni Malinen <j@w1.fi>
This new wpa_supplicant and hostapd control interface command can be
used to determine which TLS library is used in the build and what is the
version of that library.
Signed-off-by: Jouni Malinen <j@w1.fi>
At least MinGW did not have ENOTCONN, EOPNOTSUPP, ECANCELED, so define
these to allow the build to go through.
wpas_rrm_send_neighbor_rep_request() is not really used on Windows, so
the exact error code values do not make any difference here.
Signed-off-by: Jouni Malinen <j@w1.fi>
NSS as a TLS/crypto library alternative was never completed and this
barely functional code does not even build with the current NSS version.
Taken into account that there has not been much interest in working on
this crypto wrapper over the years, it is better to just remove this
code rather than try to get it into somewhat more functional state.
Signed-off-by: Jouni Malinen <j@w1.fi>
These were already covered in both README-HS20 for credentials and in
header files for developers' documentation, but the copy in
wpa_supplicant.conf did not include all the details. In addition, add a
clearer note pointing at subject_match not being suitable for suffix
matching domain names; domain_suffix_match must be used for that.
Signed-off-by: Jouni Malinen <j@w1.fi>
In many applications it is useful not just to enumerate a group of well
known access points, but to use a address/mask notation to match an
entire set of addresses (ca:ff:ee:00:00:00/ff:ff:ff:00:00:00).
This change expands the data structures used by MAC lists to include a
mask indicating the significant (non-masked) portions of an address and
extends the list parser to recognize mask suffixes.
Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de>
This change adds the configuration options "bssid_whitelist" and
"bssid_blacklist" used to limit the AP selection of a network to a
specified (finite) set or discard certain APs.
This can be useful for environments where multiple networks operate
using the same SSID and roaming between those is not desired. It is also
useful to ignore a faulty or otherwise unwanted AP.
Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de>
This change generalizes the code used for parsing the configuration
option 'p2p_client_list' and makes it suitable to use it in other
contexts.
Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de>
There should not be a mesh-specific mechanism for setting up channel
parameters since that will just result in duplicated code. IBSS, mesh,
and AP mode can use the same data structures and parameters for setting
up such parameters.
Signed-off-by: Jouni Malinen <j@w1.fi>
Automatically enable HT20, HT40+, HT40-, or VHT, based on driver
capabilities. This obsoletes the mesh_ht_mode network block parameter
that was previously used to configure HT parameters.
Signed-off-by: Jouni Malinen <j@w1.fi>
Add WPA_DRIVER_FLAGS_HT_IBSS driver feature flag. Some drivers could not
set this feature and next could fail when we will enable HT support for
IBSS with error message: nl80211: Join IBSS failed: ret=-22 (Invalid
argument).
Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
When performing SAE authentication in mesh, one station may
initiate authentication by sending a COMMIT as soon as a peer
candidate is discovered. Previously we did this in mesh_rsn.c,
but this left some of the state initialization in a different
part of the code from the rest of the state machine, and we may
need to add other initializations here in the future, so move
that to a more central function.
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Try to be a bit more consistent by using NoMemory instead of InvalidArgs
if os_strdup() fails in the CreateInterface handler.
Signed-off-by: Jouni Malinen <j@w1.fi>
It was possible for the interface not to be marked in INTERFACE_DISABLED
state in case the event was processed for P2P GO because the wpa_s
instance could have been removed in case of a separate group interface.
Change the state first to avoid leaving different state for the case
where separate group interface is not used.
Mark scan to be a normal scan on INTERFACE_ENABLED so that scanning
rules (e.g., skip scan if no networks enabled) get used consistently.
Signed-off-by: Jouni Malinen <j@w1.fi>
There is a race condition between receiving an AP stopped event and
netdev down event. These resulted in different group removal reasons on
a GO device (UNAVAILABLE for stop AP event coming first and REQUESTED
for netdev event first). Make this more consistent by reporting
UNAVAILABLE for both possible cases.
Signed-off-by: Jouni Malinen <j@w1.fi>
The new control interface command TEST_ALLOC_FAIL and GET_ALLOC_FAIL can
now be used to trigger memory allocation failures for testing purposes.
TEST_ALLOC_FAIL sets a failure conditions with
<count>:func[;func][;func]... string and GET_ALLOC_FAIL returns the
current state using the same format. Whenever an allocation is made with
a matching backtrace of calling functions, the count is decremented by
one and once zero is reached, the allocation is forced to fail.
Function names can be prefixed with either '=' or '?' to get different
matching behavior. '=' requires this specific function to be the next
one in the backtrace (i.e., do not skip any other functions in the list
which is the default behavior). '?' allows the function to be optionally
present in the backtrace.
Signed-off-by: Jouni Malinen <j@w1.fi>
It is possible for this function to get called even if D-Bus
registration had failed, so the path can be NULL here.
Signed-off-by: Jouni Malinen <j@w1.fi>
It is possible for this function to be called even if D-Bus interface
registration has failed. Avoid a D-Bus assert in such a case.
Signed-off-by: Jouni Malinen <j@w1.fi>
entry->strarray_value was left to point to freed memory in case
os_realloc_array() failed. This resulted in the following
wpa_dbus_dict_entry_clear() trying to free an already freed memory area.
In addition, the separately allocated strings in the array would have
been leaked in such a case. Furthermore, wpa_dbus_dict_entry_clear() was
not prepared for the possibility of the initial os_calloc() call failing
and entry->strarray_value being NULL without array_len being cleared to
zero. That would have resulted in reading uninitialized memory and NULL
pointer dereference.
Signed-off-by: Jouni Malinen <j@w1.fi>
entry->bytearray_value was left to point to freed memory in case
os_realloc_array() failed. This resulted in the following
wpa_dbus_dict_entry_clear() trying to free an already freed memory area.
Signed-off-by: Jouni Malinen <j@w1.fi>
add_interface() did not check for os_strdup() return value and could end
up dereferencing a NULL pointer if memory allocation failed.
Signed-off-by: Jouni Malinen <j@w1.fi>
The error reply needs to be sent out as a response. The "out" label was
in incorrect place to allow that to happen; instead, it ended up leaking
memory for the generated reply message.
Signed-off-by: Jouni Malinen <j@w1.fi>
