There was a race condition in how the l2_packet sockets got read that
could result in the same socket (e.g., non-bridge) to process both the
EAP-Success and the immediately following EAPOL-Key msg 1/4 instead of
each frame going in alternative order between the bridge and non-bridge
sockets. This could be hit, e.g., if the wpa_supplicant process did not
have enough CPU to process all the incoming frames without them getting
buffered and both sockets reporting frames simultaneously.
This resulted in the duplicated EAP-Success frame getting delivered
twice for processing and likely also the EAPOL-Key msg 1/4 getting
processed twice. While the latter does not do much harm, the former did
clear the EAP authentication state and could result in issues.
Fix this by extended the l2_packet Linux packet socket workaround for
bridge to check for duplicates against the last two received frames
instead of just the last one.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
It was possible for the packet socket on the bridge interface to receive
own transmitted frames between the bridge and non-bridge sockets
receiving the same incoming frame from a foreign host. This resulted in
the hash checksum validation step failing to notice a duplicate RX due
to the own frame updating the store hash value.
The own frame did get dropping in RX EAPOL processing, but that was too
late to address the issue with duplicate RX. Fix this by dropping own
frames already in l2_packet layer before checking and updating the last
RX hash value.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit e6dd8196e5 ('Work around Linux
packet socket regression') forgot to add the l2_packet_init_bridge()
wrapper for l2_packet_pcap.c while updating all the other l2_packet
options. This resulted in wpa_supplicant build failing due to missing
l2_packet_init_bridge() function when using CONFIG_L2_PACKET=pcap in
wpa_supplicant/.config. Fix this by adding the wrapper function.
Signed-off-by: Jouni Malinen <j@w1.fi>
Linux packet socket workaround(*) has an impact in performance when the
workaround socket needs to be kept open to receive EAPOL frames. While
this is normally avoided with a kernel that has the issue addressed by
closing the workaround packet socket when detecting a frame through the
main socket, it is possible for that mechanism to not be sufficient,
e.g., when an open network connection (no EAPOL frames) is used.
Add a build option (CONFIG_NO_LINUX_PACKET_SOCKET_WAR=y) to disable the
workaround. This build option is disabled by default and can be enabled
explicitly on distributions which have an older kernel or a fix for the
kernel regression.
Also remove the unused variable num_rx.
(*) 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.
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
This program can be used to run fuzzing tests for areas related to EAPOL
frame parsing and processing on the supplicant side.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit e6dd8196e5 ('Work around Linux
packet socket regression') closed the workaround socket on the first
received EAPOL frame from the main packet socket. This can result in
closing the socket in cases where the kernel does not really work in the
expected way during the following initial association since
reauthentication/rekeying using EAPOL frames happens while operstate is
not dormant and as such, the frames can get delivered through the main
packet socket.
Fix this by closing the workaround socket only in case the first EAPOL
frame is received through the main packet socket. This case happens
while the interface is in dormant state and as such, is more likely to
show the more restricted case of kernel functionality.
In order to avoid processing the received EAPOL frames twice, verify a
checksum of the frame contents when receiving frames alternatively from
the main packet socket and the workaround socket.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit e6dd8196e5 ('Work around Linux
packet socket regression') added a mechanism to close the workaround
bridge socket in l2_packet_receive(). However, it did not take into
account the possibility of the l2->rx_callback() closing the l2_packet
socket altogether. This could result in use of freed memory when usin
RSN pre-authentication. Fix this by reordering the calls to clear the
workaround socket before calling the rx_callback.
Signed-off-by: Jouni Malinen <j@w1.fi>
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>
This converts most of the remaining perror() and printf() calls from
hostapd and wpa_supplicant to use wpa_printf().
Signed-off-by: Jouni Malinen <j@w1.fi>
The RA, NS, and NA packets may be longer, so do not arbitrarily limit
the packet socket capture size to 150 bytes in the socket filter.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The sample code here ended up trying to register an eloop socket with fd
== -1. This was not really ever supposed to be used, but it is now also
hitting an assert in eloop. Skip the unnecessary
eloop_register_read_sock() to avoid this.
This was causing issues for hostapd since CONFIG_L2_PACKET is not set by
default. If CONFIG_RSN_PREAUTH=y was not used for CONFIG_L2_PACKET was
not set in .config explicitly, the defaul use of l2_packet_none.c ended
up hitting the newly added assert() in eloop.
Signed-off-by: Jouni Malinen <j@w1.fi>
This updates these files to use the license notification that uses only
the BSD license. The changes were acknowledged by email (Sam Leffler
<sam@errno.com>, Sat, 30 Jun 2012 07:57:53 -0700).
Signed-hostap: Jouni Malinen <j@w1.fi>
When using overlapped write, we must have the provided memory
areas available during the operation and cannot just use stack
unless we wait for the completion within the function. In the case
of TX here, we can easily wait for the completion since it is likely
to happen immediately. In addition, this provides more reliable
success/failure return value for l2_packet_send(). [Bug 328]
This patch adds support for wired IEEE 802.1X client on the Solaris.
I have tested with these:
OS : OpenSolaris 2009.06
EAP : EAP-MD5
Switch : Cisco Catalyst 2950
There are no subdirectories in any of these directories or plans
for adding ones. As such, there is no point in running the loop
that does not do anything and can cause problems with some shells.
This patch allows wpa_supplicant to compile on Debian's kfreebsd
architectures.
Patch by Stefan Lippers-Hollmann based on work done by Petr Salinger
and Emmanuel Bouthenot for 0.6.X (http://bugs.debian.org/480572).
This makes it clearer which files are including header from src/common.
Some of these cases should probably be cleaned up in the future not to
do that.
In addition, src/common/nl80211_copy.h and wireless_copy.h were moved
into src/drivers since they are only used by driver wrappers and do not
need to live in src/common.
gcc 4.4 ends up generating strict-aliasing warnings about some very common
networking socket uses that do not really result in a real problem and
cannot be easily avoided with union-based type-punning due to struct
definitions including another struct in system header files. To avoid having
to fully disable strict-aliasing warnings, provide a mechanism to hide the
typecast from aliasing for now. A cleaner solution will hopefully be found
in the future to handle these cases.
This removes need for local configuration to ignore *.o and *~
and allows the src/*/.gitignore files to be removed (subdirectories
will inherit the rules from the root .gitignore).
