This was refactored to set the mtu via __eap_set_config rather than
passing the MTU into eap_init. This makes eap work in a similar fashion
as eapol (i.e. __eapol_set_config).
If __eap_set_config is not used, the MTU will be set to 1020, which is
the same as previously passing 0 to eap_init.
Since iwd_modules_init is now defered until nl80211_appeared, we can
assume the nl80211 object is available. This removes the need for
netdev_set_nl80211 completely.
In preparation for integrating IWD_MODULE into modules which require
nl80211 we move the module init into the nl80211_appeared callback.
This will guarentee that the nl80211 is available during module init
and allow modules to get their own copy of nl80211 rather than needing
a set function (e.g. netdev_set_nl80211).
Since the dbus name request callback happens before this as well any
dbus module can also use IWD_MODULE and simply assume the dbus object
is ready.
plugin_init was also deferred to nl80211_appeared since some plugins
depend on modules being initialized.
Converts agent into an IWD module. This removes the dbus dependency
on agent. Since dbus is initialized very early we can assume
dbus_get_bus is going to return a valid object.
Previously, station state 'connected' used to identify an interface associated
with AP. With the introduction of netconfig, an interface is assumed to be
connected after the IP addresses have been assigned to it. If netconfig is
disabled, the behavior remains unchanged.
Refactoring was required to allow for embedded certs. The existing
eap_tls_state object was changed to hold the cert types (l_queue,
l_certchain, l_key) rather than the file path, since there may not
actually be separate PEM files.
Care was taken to properly manage the memory of these objects.
Since the TLS object takes ownership when setting auth data or the
CA certs all error cases must be handled properly to free these
objects after they are loaded and in addition they must be set to
NULL so that the cleanup doesn't double free them.
If everything goes to plan, we load all the PEMs in settings_load,
provide these objects to the TLS APIs, and then NULL out the
pointers (TLS now owns this memory). If anything fails between
settings_load and l_tls_start we must free these objects.
A special format must be used to indicate that a PEM is embedded
inside the settings file. First, the l_settings format should be
followed for the PEM itself, e.g.
[@pem@my_ca_cert]
<CA Cert data>
This PEM can then be referenced by "embed:my_ca_cert", e.g.
EAP-TLS-CACert=embed:my_ca_cert
Any other value not starting with "embed:" will be treated as a file
path.
The IPv6 default route needs to be explicitly revoked. Unlike in IPv4,
there is no SRC address associated with the route and it will not be
removed on address removal.
The network configuration options for IPv6 are grouped under [IPv6]
and include the following:
ip= ADDRESS/PREFIX
gateway=ADDRESS
dns=ADDRESS
The placeholders for DHCPv6 are placed along the way and marked
as TODO items.
Previously, netconfig_ipv4_select_and_install was used to install
addresses on initial connection to a network and after we have roamed.
Now for the after roaming connection scenario we have
netconfig_reconfigure. Remove roaming related code from
netconfig_ipv4_select_and_install
As part of the de-coupling from station object, switch all of
the network settings inquiries to use active_settings. active_settings
are set with netconfig_configure by the owner of netconfig object
and removed with netconfig_reset once network disconnects.
Instead of relying on station state changed signal, netconfig
introduces three new API calls to configure, re-configure and
reset the network configurations. The owner of netconfig object
is responsible for initiating the re-configuration of the device
depending on its state.
As a first step to enable the usage of netconfig in ead and
prospective transition to be a part of ell, the public API for
creation and destruction of the netconfig objects has been
renamed and changed. Instead of hiding the netconfig objects inside
of netconfig module, the object is now passed back to the caller.
The internal queue of netconfig objects remains untouched, due
to limitations in ell’s implementation of rtnl. After the proper
changes are done to ell, netconfig_list is expected to be removed
from netconfig module.
A NEW_WIPHY event may not always contain all the information about a
given phy, but GET_WIPHY will. In order to get everything we must
mimic the behavior done during initalization and dump both wiphy
and interfaces when a NEW_WIPHY comes in.
Now, any NEW_WIPHY event will initialize a wiphy, but then do a
GET_WIPHY/GET_INTERFACE to obtain all the information. Because of
this we can ignore any NEW_INTERFACE notifications since we are
dumping the interface anyways.
Once some kernel changes get merged we wont need to do this anymore
so long as the 'full' NEW_WIPHY feature is supported.
If the AP sent us the plain passphrase we can now store that rather
than generating the PSK. This will allow WPA3 to work properly when
WPA3 + WSC is implemented.
This lets other modules (like WSC) to set a plain text passphrase
as opposed to only allowing a PSK to be set. network_get_psk was
also updated to generate a PSK on-the-fly if required. Since WPA3
requires the raw passphrase to work, it makes sense to just store
the passphrase if we have it.
If neighbor reports are unavailable, or the report yielded no
results we can quickly scan for only known frequencies. This
changes the original behavior where we would do a full scan
in this case.
This password key was deprecated in favor of the common EAP-Password
key. Its been about a year so we are now removing support entirely
for EAP-PWD-Password.
Gets a newly created scan_freq_set containing the most recent
frequencies for the network. The currently connected BSS frequency
(passed as a parameters) will not be included in the set.
Since the UUID was being generated purely on the file path, it
would never change for a given network (unless the SSID/name changed).
In the future we would like to use this unique UUID to generate a
MAC per-SSID, and if that network is forgotten we also want the UUID
to change next time the network is connected to.
Rather than only using the file path, the mtime can also be fed into
the UUID generation. Since the mtime would be changed after forgetting
and re-adding a known network we will get a new UUID.
Now, whenever a known network is removed, we lookup the UUID we have
in network_info and remove that entry in the settings file and
sync the frequency file.
The UUID was being generated every time we synced which is wasteful.
Instead we can track the UUID inside network_info and only generate
it once when needed.
Two new network_info APIs were added:
network_info_set_uuid
network_info_get_uuid
The setter is used when the frequency file is loaded. If a valid UUID
is found in the frequency file this UUID is set and used.
network_info_get_uuid will not just get the UUID, but actually generate
it if one has not been set yet. This will allow other modules to
get/generate the UUID if one has no been loaded from the frequency
file.
The QoS Map can come in either as a management frame or via the
Associate Response. In either case this IE simply needs to be
forwarded back to the kernel.
The extended capability bits were not being set properly inside
wiphy. Since we build the IE after the wiphy dump the first 2
bytes are the IE type and length. The way we were setting the bits
did not take this into account and were actually setting the
completely wrong bits.
The known frequency file was being loaded at the end of the known
networks initialization routine. This allowed all known networks
to be properly loaded, but since hotspot depends on known networks,
its initalization would be run afterwards meaning the frequency
loading would not have been finding any hotspot networks.
To fix this a new module was added inside known networks which
depends on hotspot. This means that first known networks will
initialize, then hotspot, then the frequency file would be loaded.
The current format for the .known_networks.freq file had a hidden
limitation of not being able to handle SSID's with some special
characters. Since the provisioning file path was used as the
group name the filename was limited to only characters supported
by l_settings groups, which conflicted with allowable SSID
characters.
Instead we can generate a unique UUID for each network and use
this as the group. For this particular case the group does not
really matter, so long as its unique. But we can utilize this unique
UUID for other purposes, including using it as a seed for changing
the MAC address per-connection in the future.
The .known_networks.freq file will now have the following format:
[<UUID>]
name=/path/to/provisioning/file
list= XXXX YYYY ZZZZ
The existing frequency syncing was done when IWD closes. Instead we
can sync as networks are connected to or promoted to known which
will keep the FS more up to date. This also allows hotspot networks
to use the known frequency file.
This API will sync the known frequencies of a network_info object
to disk. This will allow network to sync known frequencies as
known networks are added, rather that when IWD closes.
Since this will result in more frequent syncing that before, the
known_freqs settings pointer was moved globally in knownnetworks.c
as to only parse the file one time rather than on every sync.
Some of the EAP-PEAP server implementations seem to require a
cleartext ACK for the tunneled EAP-Success message similar to EAP-TLS
specification, instead of simply shutting down the tunnel like
EAP-PEAPv1 requires.
ACKing the tunneled EAP-Success seems also to work for implementations
which were relying on the tunnel close event.
create_dirs was dependent on the path ending in '/' to create the
full path. The hotspot code did not include a '/' at the end so
it was not getting created, which prevented the hotspot module
from initializing.
Station was building up the HS20 elements manually. Now we can
use this new API and let network take care of the complexity
of building network specific vendor IEs.
This op builds up the vendor IEs required for hotspot 2.0. The
version, and optionally the RC are provided in order to correctly
build the HS20 Indication Element and RC Selection element.
The HS20 module had its own getter for returning the matched roaming
consortium. Since we already have the network_info op for matching
we might as well return the matched RC rather than just a bool. This
allows the RC to be included in (Re)Association without the need for
a specific getter.
When performing a fast transition to another OPEN network the RSN
element won't be there and therefore the bss->rsne is gonna be NULL.
Fix crash by not accessing the rsne member when performing a fast
transition to an AP that doe snot advertise any RSN IE.
Crash caught with gdb:
src/station.c:station_transition_start() 186, target 34:8f:27:2f:b8:fc
Program received signal SIGSEGV, Segmentation fault.
handshake_state_set_authenticator_ie (s=0x555555626eb0, ie=0x0) at src/handshake.c:163
163 s->authenticator_ie = l_memdup(ie, ie[1] + 2u);
(gdb) bt
#0 handshake_state_set_authenticator_ie (s=0x555555626eb0, ie=0x0) at src/handshake.c:163
#1 0x0000555555561a98 in fast_transition (netdev=0x55555562fbe0, target_bss=0x55555561f4a0,
over_air=over_air@entry=true, cb=0x55555556d5b0 <station_fast_transition_cb>) at src/netdev.c:3164
#2 0x0000555555565dfd in netdev_fast_transition (netdev=<optimized out>, target_bss=<optimized out>,
cb=<optimized out>) at src/netdev.c:3232
#3 0x000055555556ccbd in station_transition_start (bss=0x55555561f4a0, station=0x555555617da0)
at src/station.c:1261
#4 station_roam_scan_notify (err=<optimized out>, bss_list=<optimized out>, userdata=0x555555617da0)
at src/station.c:1444
#5 0x0000555555579560 in scan_finished (sc=0x55555562bf80, err=err@entry=0, bss_list=0x55555561bd90,
sr=0x555555626b30, wiphy=<optimized out>) at src/scan.c:1234
#6 0x0000555555579620 in get_scan_done (user=0x555555618920) at src/scan.c:1264
#7 0x00005555555abd23 in destroy_request (data=0x55555561b000) at ell/genl.c:673
#8 0x00005555555ac129 in process_unicast (nlmsg=0x7fffffffc310, genl=0x55555560b7a0) at ell/genl.c:940
#9 received_data (io=<optimized out>, user_data=0x55555560b7a0) at ell/genl.c:1039
#10 0x00005555555a8aa3 in io_callback (fd=<optimized out>, events=1, user_data=0x55555560b840)
at ell/io.c:126
#11 0x00005555555a7ccd in l_main_iterate (timeout=<optimized out>) at ell/main.c:473
#12 0x00005555555a7d9c in l_main_run () at ell/main.c:520
#13 l_main_run () at ell/main.c:502
#14 0x00005555555a7fac in l_main_run_with_signal (callback=<optimized out>, user_data=0x0)
at ell/main.c:642
#15 0x000055555555e5b8 in main (argc=<optimized out>, argv=<optimized out>) at src/main.c:519
After wsc_store_credentials, wsc_try_credentials is called which
sets the PSK obtained via the protocol. After the known network
refactor network_settings_load was changed to depend on the
network_info->open() call. Since there is no known network for
this initial WSC connection this always fails and the PSK is not
set into the network object (and the connection is failed).
In this case if network_settings_load fails we can just create
an empty settings object to be filled later.
known_network_update was being used to both update and create known
networks as they appeared on the file system. Hotspot needs updating
capabilities so known_network_update was exposed and updated with
one major difference; it no longer can be used to create new known
networks. For creation, a new API was added (known_network_new)
which will create and add to the queue.
Since hotspot networks may require ANQP the autoconnect loop needed to
be delayed until after the ANQP results came back and the network
objects were updated. If there are hotspot networks in range ANQP will
be performed and once complete autoconnect will begin for all networks
including hotspots. If no hotspots are in range autoconnect will
proceed as it always has.
Note: Assuming hotspots are in range this will introduce some delay
in autoconnecting to any network since ANQP must come back. The full
plan is to intellegently decide when and when not to do ANQP in order
to minimize delays but since ANQP is disabled by default the behavior
introduced with this patch is acceptable.
The remove op was being called inside known_networks_remove, which only
gets called from L_DIR_WATCH events. In this case the actual provisioning
has already been removed. Calling remove() again causes the op
implementation to then try and remove the file that no longer exists.
Valgrind does not like uninitialized bytes used in a syscall. In this
case the buffer is an out buffer but since valgrind doesn't know that
it complains. Initializing to zero fixes the warning:
Syscall param socketcall.sendto(msg) points to uninitialised byte(s)
at 0x5162C4D: send (send.c:28)
by 0x457AF4: l_checksum_update (checksum.c:319)
by 0x43C03C: eap_wsc_handle_m2 (eap-wsc.c:842)
by 0x43CD33: eap_wsc_handle_request (eap-wsc.c:1048)
by 0x43A3A7: __eap_handle_request.part.0 (eap.c:266)
by 0x41A426: eapol_rx_packet.part.12 (eapol.c:2262)
by 0x41B536: __eapol_rx_packet (eapol.c:2650)
by 0x407C80: netdev_control_port_frame_event (netdev.c:3542)
by 0x407C80: netdev_unicast_notify (netdev.c:3684)
by 0x4598C5: dispatch_unicast_watches (genl.c:899)
by 0x4598C5: process_unicast (genl.c:918)
by 0x4598C5: received_data (genl.c:1039)
by 0x456452: io_callback (io.c:126)
by 0x45569D: l_main_iterate (main.c:473)
by 0x45576B: l_main_run (main.c:520)
Address 0x1ffeffe290 is on thread 1's stack
in frame #2, created by eap_wsc_handle_m2 (eap-wsc.c:797)
We were not using or taking into account the noencrypt flag obtained
from the kernel via CONTROL_PORT events. For the most part this still
worked as the kernel would never include NO_ENCRYPT flag (due to a bug).
However, this was actually incorrect and led to loss of synchronization
between the AP and STA 4-Way handshake state machines when certain
packets were lost and had to be re-transmitted.
Allow users to provide a glob string that the contents of the server
certificate's subject DN should be matched against as a primitive
protection against rogue APs using certificates purchased from
commercial CAs trusted by the client. If the network uses an AP
certificate emitted by a commerical CA and the clients are configured
to trust those CAs so that the client configurations don't have to be
updated when the AP renews its certificate, this new option can be used
to check if the CN in the AP certificate's DN matches the known domain
name. This logic assumes that the commercial CAs provide enough
assurance that only the owner of the domain can buy a certificate with
that domain in the CN field.
The format of this option is similar to apple's TLSTrustedServerNames
and wpa_supplicant's domain_match/domain_suffix_match format, the exact
syntax is documented in ell/tls.c.
Some capability bits are required by the spec to be set for
probe requests for certain features (HS20, FILS, FT). Currently
these features work as-is, but depending on the hardware we may
be in violation of the spec if we assume the correct bits are
set when we get the wiphy dump.
Just to be safe we can explicity set these capability bits.
There are also two ways the kernel exposes these capabilities.
Per-type or globally. The hardware may expose one, or both of
these capability arrays. To combat this we are now always
creating a per-type capability array for stations. If the
wiphy dump has not produced a per-type capability array we
now create one based off the global capability array. That
way we can always assume there is a capability array for a
station iftype.
This will be seen in Probe Requests. More IEs can and should
be added here depending on the support in IWD. E.g. HS20 indication,
Interworking, HT/VHT IE's etc.
In order to implement get_name/get_type we need some value for the name
of the hotspot network. For simplicity we now require a 'Name' value
be provided inside the hotspot provisioning file. Eventually this may
change (e.g. obtained via ANQP).
Rather than using timespec directly, ELL has a convenient API
to get the elapsed microseconds as a uint64_t. This can then
be used with the other l_time_ APIs for comparison.
This patch removes timespec from network_info and updates
to use l_time_* API's for sorting.
These operations will allow the hotspot module to implement
matching HESSID, Roaming Consortium, and NAI realms. This offloads
the matching details into the hotspot module.
This way resolve_remove can be called once per interface and
remove IPv4 and IPv6 addresses at once.
In addition, this allows to remove the IP addresses and DNS
servers within the same main loop cycle. This will allow iwd
to make an attempt to remove the DNS servers on shutdown of iwd.
These two API's have been added to allow hotspot to add its
own networks to the known network list. This will allow any
added networks to behave exactly like they do now, including
all the dbus and watchlist functionality.
The known network APIs all revolved around the ssid/security matching
to do any operations on the provisioning file. In the near future
hotspot provisioning files (managed by hotspot.c) will be incorporated
into the known network list. Since these hotspot files do not use the
ssid as the file name hotspot.c will need other ways of matching.
This patch adds network_info_ops to the network object. This ops
structure will hold function pointers which operate on network_info
rather than ssid/security. This will allow hotspot and known networks
to both register their own operation routines.
For now open, touch, sync, remove, free, and get_path were added.
Wrappers were added for accessing these operations outside of
knownnetworks.c.
Isolate the known_frequency queue management to a function and place
that function in knownnetworks.c where it now belongs. Since we no
longer have network_info objects for unknown networks, only frequencies
for known networks are tracked
networks queue was intended to share basic network information between
multiple adapters running simultaneously. The network_info object was
also serving double duty to carry known network information. This made
things overly complicated and really didn't result in much savings.
This setup also made managing hotspot networks challenging as we would
have ended up with multiple network_info objects for each known hotspot
network.
So get rid of the networks queue and the is_known bit from the
network_info structure.
network_find_rank_index was used to find the offset of the selected
network_info among known networks so as to compute a modifier based on
the rankmod table. Instead of using known_networks_foreach for this,
moove it to knownnetworks.c where it can be coded and optimized
separately.
For now provide a simple for loop implementation.
Previously, the option PrivateDevices=true disabled access to
/dev/rfkill, which lead to:
'iwctl adapter phy0 set-property Powered {off|on}'
to fail.
This patch explicitly allows access to /dev/rfkill
src/rtnlutil.c: In function ‘rtnl_route_add’:
./ell/util.h:248:2: error: ‘rtmmsg’ may be used uninitialized in
this function [-Werror=maybe-uninitialized]
Instead of using a flag ipv4_static, just store the value of the rtm
protocol directly inside netconfig object. This allows us to simplify
the logic quite significantly and avoid repeating the conditional
expression needlessly
The routes are installed as a result of a successful installation
of the IP addresses. The gateway is fetched with netconfig_ipv4_get_gateway
helper function according to the origin of the installed IP address.
The route priority offset can be set in main.conf. The default value
of 300 is used if the offset isn’t set.
The API allows to add connected and gateway routes to the main
routing table.
rtnl_route_ipv4_add_gateway() is equivalent to the following
example 'ip route' command:
ip route add default via 10.0.0.1 dev wlan0 proto dhcp src 10.0.0.2 metric 339
rtnl_route_ipv4_add_connected() is equivalent to the following
example 'ip route' command:
sudo ip route add 10.0.0.0/24 dev wlan0 proto dhcp src 10.0.0.2 scope link
The 'ip route' output from the above commands looks as follows:
rtnl_route_ipv4_add_connected():
10.0.0.0/24 dev wlan0 proto dhcp scope link src 10.0.0.2
rtnl_route_ipv4_add_gateway():
default via 10.0.0.1 dev wlan0 proto dhcp src 10.0.0.2 metric 339
The DNS addresses are installed as a result of a successful
installation of the IP addresses. The DNS lists are fetched
with netconfig_ipv4_get_dns helper function according to the
origin of the installed IP address.
iwd reconfigures the wireless interfaces with respective
connection events. Each supported network protocol is
reconfigured. The address for each protocol is
selected as static or dynamic based on availability.
netconfig_ipv4_get_ifaddr helper function allows to fetch IPv4
addresses from static or dynamic sources. The origin of the addresses
is noted in 'ipv4_is_static' flag.
For (Re)Association the HS20 indication element was passed exactly as
it was found in the scan results. The spec defines what bits can be
set and what cannot when this IE is used in (Re)Association. Instead
of assuming the AP's IE conforms to the spec, we now parse the IE and
re-build it for use with (Re)Association.
Since the full IE is no longer used, it was removed from scan_bss, and
replaced with a bit for HS20 support (hs20_capable). This member is
now used the same as hs20_ie was.
The version parsed during scan results is now used when building the
(Re)Association IE.
The parser fully parses the IE and returns the version, Domain ID,
and PPS MO ID. This is meant to be used with an IE in scan results.
The builder only takes the version number, and assumes DGAF disabled,
and no Domain ID or PPS MO ID.
Previously, iwd used to throw net.connman.iwd.Busy when connection
attempt was made while connected. The new behavior allows iwd to
seamlessly disconnect from the connected network and attempt a new
connection.
Since NAI realms, Roaming Consortium and HESSID are defined in 802.11,
they are not a guarentee that the network is Hotspot 2.0. The indication
element in addition to these IE's gives a better idea of Hotspot 2.0
support. Now, when a BSS is added this is_hs20 boolean will get set to
true if the HS20 IE was found in the BSS.
Now, if is_hs20 is set AND one of NAI realms, roaming consortium, or
HESSID is set we know this is a hotspot 2.0 network.
It is possible for a zero-length anti-clogging token payload to cause
IWD to abort. If the length passed into sae_process_anti_clogging was
1, l_memdup would be called with a size of -1. This will cause malloc
to abort.
Fix this by checking for a minimum packet length and dropping the
packet if the length is too small.
The HS20 indication element should always be included during
(Re)Association per the spec. This removes the need for a
dedicated boolean, and now the hs20_ie can be used instead.
The hotspot spec specifically mentions the roaming consortium OI be
3 or 5 bytes long. This requirement also prevents potential buffer
overflows if the user were to configure a long roaming consortium OI.
If the scan was triggered and later aborted, make sure to reset the
triggered value when the CMD_NEW_SCAN_RESULTS event comes in.
src/station.c:station_enter_state() Old State: disconnected, new state: connecting
src/scan.c:scan_notify() Scan notification 33
src/station.c:station_netdev_event() Associating
src/scan.c:scan_notify() Scan notification 34
Aborting (signal 11) [/home/denkenz/iwd-master/src/iwd]
++++++++ backtrace ++++++++
#0 0x7efd4d6a2ef0 in /lib64/libc.so.6
#1 0x42b20d in scan_notify() at src/scan.c:1383
In the same fashion as the WSC WFA OUI, ie.[ch] will now expose the
other vendor OUIs to avoid duplication across multiple files in IWD
as well as used in iwmon.
P2P probe requests are to be sent at min 6.0 Mb/s using OFDM,
specifically the 802.11b rates are prohibited (section 2.4.1 in Wi-Fi
P2p Technical Spec v1.7), some of which use CCK modulation. This is
already the default for 5G but for 2.4G the drivers generally do this
if we set the NL80211_ATTR_TX_NO_CCK_RATE flags with
NL80211_CMD_TRIGGER_SCAN.
The length check was incorrectly assuming that PPS MO ID or
ANQP Domain ID would be present in the IE. Both these are optional
and without then the minimum length is 5 bytes, not 7.
Per the hotspot 2.0 spec, if a matching roaming consortium OI is
found it should be added to the (Re)Association request. vendor_ies
can now be provided to netdev_connect, which get appended to the IE
attribute.
This API will attempt to find a matching roaming consortium OI
if present in the config file. A single matching OI is returned
or NULL if one was not found.
Hotspot 2.0 network providers allow 'roaming' between a users home
network and other providers networks, assuming they are part of the
same roaming consortium. The roaming consortium is advertised as an
IE in beacon/probe frames.
In terms of the hotspot config files this is similar to HESSID, where
if the AP advertises the roaming consortium IE, and the config file
matches we do not need to do ANQP in order to connect.
This is duplicated when the first scan_bss is added to a network
object that contains the IE. Any future BSS's added will not re-add
the IE. Its assumed that all BSS's under a network will contain the
same roaming consortium OIs.
Parses up to 3 (the max) roaming consortium OIs out of the roaming
consortium IE. If more OIs are available via ANQP the 'num_anqp_out'
value will be set to indicate how many more OIs are available.
Builds according to the hotspot 2.0 spec using the vendor specific
IE.
Declare structures to hold the parsed contents of the P2P IEs and WSC
IEs in P2P-related frames and add functions to free memory used by
those structures.
Define structs and types for most P2P attributes and p2p_parse_attrs
similar to wsc_parse_attrs -- a generic parser for attributes in a P2P
IE payload. This parser may write into the provided buffer even on
error but it's private to p2putil.c. The local callers will take care
of keeping the user-provided buffers untouched on error.
Add a utility for building the simplified WSC IEs used in P2P action
frames and public action frames. Only three types of WSC attributes are
mandatory in those frames (but different subsets are needed by different
frame types) so add a single utility for building those IEs. We may
need to add some more optional attributes to those IEs later.
The ifindex is used to index the netdevs in the system (wlan, ethernet,
etc.) but we can also do wifi scanning on interfaces that have no
corresponding netdev object, like the P2P-device virtual interfaces.
Use the wdev id's to reference interfaces, the nl80211 api doesn't care
whether we use a NL80211_ATTR_IFINDEX or NL80211_ATTR_WDEV. Only
wireless interfaces have a wdev id.
Save the actual cmd_id returned from l_genl_family_dump and zero it in
the get_scan_done. There's no need to zero it in scan_cancel because
get_scan_done gets called automatically.
Store the scan_context pointer in scan_results directly instead of
storing the ifindex. We now cancel ongoing GET_SCAN commands when the
scan_context is being freed so there's no point going through the extra
step of looking up the scan_context by ifindex inside the command
callback to guard against non-existent scan_contexts.
method.ops is NULL, which causes method.ops->exit to crash. This
adds a check that method.ops is not NULL before dereferencing.
Fixes:
Aborting (signal 11) [/home/jprestwo/iwd/src/iwd]
++++++++ backtrace ++++++++
0 0x7f016b59cf20 in /lib/x86_64-linux-gnu/libc.so.6
1 0x432057 in resolve_exit() at /home/jprestwo/iwd/src/resolve.c:295
2 0x403b61 in iwd_modules_exit() at /home/jprestwo/iwd/src/main.c:195
3 0x7f016b57fb97 in /lib/x86_64-linux-gnu/libc.so.6
+++++++++++++++++++++++++++
The original idea was to allow the provisioning file to include HESSID
without the NAIRealmNames. Configuring this way would allow for ANQP
to be skipped completely, assuming the AP advertises its HESSID.
The way the code was written still required NAIRealmNames to be
provided in the provisioning file.
The framework enables the service specific implementations
to provide its own variations for the DNS installation tasks.
The selection of the address resolution service can be done
through dns_resolve_method setting.
The module is responsible for the configuration of the address
resolution services. It will consist of the multiple service
specific plugins such as: systemd-resolved plugin, dnsmasq
plugin, etc.
If supported by the driver, we can create an interface directly with a
random MAC if configured to do so. If the driver does not have this
capability, then tell netdev to perform the necessary logic as part of
the interface initialization procedure.
#0 0x7f5e25e71930 in /lib64/libc.so.6
#1 0x446faa in hs20_config_free() at src/hotspot.c:63
#2 0x469542 in l_queue_clear() at ell/queue.c:109
#3 0x4694e7 in l_queue_destroy() at ell/queue.c:83
#4 0x4475c1 in hotspot_exit() at src/hotspot.c:273
#5 0x403170 in iwd_modules_exit() at src/main.c:195
#6 0x404085 in main() at src/main.c:531
#7 0x7f5e25e5cbde in /lib64/libc.so.6
Regulatory domain management is now completely handled by the kernel, so
iwd doesn't really need to query or be aware of changes to this. This
may change in the future, but for now this code has not been used and
can be safely gotten rid of.
After a scan, station can now pause future scans and start ANQP requests
to discover Hotspot's NAI realm. This lets us check if the AP's NAI realm
matches any stored hotspot configuration files. If so we can connect to
this network. If the network provides an HESSID and a matching one is
found in a hotspot provisioning file we can skip ANQP and directly connect
as this is expected to be our 'home network'
The actual ANQP request was handled by netdev, but in the case of P2P
their may be no netdev. For this reason all functionality needed for
an ANQP request has been moved into anqp.c. There are still a few netdev
references, which need to be removed when P2P is introduced. Leaving them
in for now as its still going to work as a first pass implementation
The initial ANQP parser design did not work well with how the hotspot
implementation was turning out. For one, much care was taken into parsing
the EAP credentials which are not really required. The assumption is
that any hotspot network will already be provisioned, so checking that
the EAP parameters match is a bit overkill. Instead only the NAI Realms
will be checked. This greatly simplifies the NAI realm parser, as now it
can just return a string list of realms instead of the full EAP
credential info.
This module will be in charge of managing Hotspot provisioning files
stored under the .hotspot/ directory. This includes a dir watch to
handle file changes/removal as well as an API to match a network
object to a hotspot provisioning file.
Hotspot networks are supposed to include an HESSID in the scan
results. This is more or less an identifier for the overall
network. In addition, the NAI Realms can be obtained via ANQP
and should be the same for each BSS. Since both HESSID and NAI
realms should be the same for a given network in range we can
store these values in the network object itself. This also allows
us to easily find hotspot configuration files by looking at
the HESSID/NAI Realms directly in the network object as opposed
to individual scan_bss's.
In order to do ANQP efficiently IWD needs the ability to suspend scanning
temporarily. This is because both scanning and ANQP go offchannel and must
remain off channel for some amount of time. This cannot be done
simultaneously and if e.g. ANQP is requested after a scan is already
pending, the kernel will wait till that scan finishes before sending out
the frame.
Use memset instead. explicit_bzero should only be used when we're
wiping a secret just prior to the encopassing storage being freed. The
compiler would usually optimize away the memset, leaving the secrets
around.
In rtnlutil we're simply zeroing the structure prior to filling it, so
the use of explicit_bzero is not needed and brings confusion to the
reader since no secrets are being wiped.
netconfig is interested in three station states: connected,
disconnected and connected after it has roamed. On connected
it tries to obtain a new DHCP lease, on disconnected it stops
the DHCP client and discards all addresses from interface, on
connected after roaming it will try to request a previously
issued address.
iwd keeps track of the addresses assigned to the managed
interfaces. The list of assigned IPv4/IPv6 addresses is stored
in ifaddr_list inside of netconfig. The tracking of the IP
addresses will help to remove them from an interface once they
are no longer valid.
netconfig module will be responsible for the orchestration
of the network configuration with the IP addresses.
iwd creates one netconfig structure per interface index.
The purpose of this struct is to hold all of the interface
related addressing states such as: assigned dhcp
clients, known addresses, routes, etc.
A not-yet-merged kernel patch will enable the FRAME_WAIT_CANCEL
event to be emitted when a CMD_FRAME duration expires. This can
shortcut the ridiculously long timeout that is required making
GAS requests with no response drastically quicker to handle.
This adds a new API netdev_anqp_request which will send out a GAS
request, parses the GAS portion of the response and forwards the
ANQP response to the callers callback.
This IE tells us what Advertisement Protocols the AP supports. This
is only here to look for ANQP support, so all this does is iterate
through all other Advertisement Protocol tuples looking for ANQP.
If found, anqp_capable is set in the scan_bss
Currently these are geared to support the WiFi Alliance Hotspot 2.0
ANQP elements, which all fall under the vendor specific ANQP element.
anqp_iter_next behaves similar to the genl parsers, where the id, length
and data will be returned as out parameters. Currently there is only
vendor support for Hotspot 2.0. anqp_iter_is_hs20 can be used to setup
the subtype, length, and data pointer to parse any Hotspot 2.0 ANQP
elements. From here the subtype can be checked and a vendor specific
parser for that subtype can be used to parse the data, e.g.
hs20_parse_osu_provider_nai.
The vendor specific IE was being parsed only to check if the AP supported
WPA, which used a Microsoft OUI. Hotspot/OSEN uses neither WPA or RSN
(although its nearly identical to RSN) so the we also need to check for
this Wifi-Alliance OUI and set bss->osen (new) if found.
The OSEN AKM uses the vendor specific IE, so when finding the RSNE
element we need to handle it specially to ensure that its both
a vendor specific element and it matches the WFA OUI since other
vendor specific elements may be included.
The OSEN AKM is nearly identical to the RSN IE, but differs slightly.
For one, OSEN is encapsulated into the vendor specific IE, and includes
the WFA OUI before the 'normal' RSN elements. OSEN also does not include
a WPA version, since its not technically WPA/WPA2.
Some of the RSN parsing was made common so both RSN/OSEN parsing could
use it.
The handshake object had 4 setters for authenticator/supplicant IE.
Since the IE ultimately gets put into the same buffer, there really
only needs to be a single setter for authenticator/supplicant. The
handshake object can deal with parsing to decide what kind of IE it
is (WPA or RSN).
The Hotspot 2.0 spec introduces 'Anonymous EAP-TLS' as a new EAP method
to be used with OSEN/Hotspot. The protocol details of this aren't
relevant to this patch, but one major difference is that it uses the
expanded EAP type rather than the TLS type. Since the common TLS code
was written with only EAP_TYPE_TLS in mind the vendor ID/type cause the
EAP packet to be malformed when using the expanded EAP type.
To handle this the common TLS code now checks the EAP type, and if its
expanded we shift the payload 7 bytes further to account for the extra
header data.
802.11 defines GAS (generic advertisement service) which can be used
to query supported advertisement protocols from an AP before
authentication/association. Hotspot/OSEN only care about the ANQP
protocol, but the way the IE is structured potentially requires
iterating through several tuples before you reach the ANQP protocol
identifier. Because of this we define all protocol identifiers.
This adds some checks for the FT_OVER_FILS AKMs in station and netdev
allowing the FILS-FT AKMs to be selected during a connection.
Inside netdev_connect_event we actually have to skip parsing the IEs
because FILS itself takes care of this (needs to handle them specially)
FILS unfortunately is a special case when it comes to fast transition.
We have to process the FT IEs internally since we cannot trigger the
same initial mobility association code path (via netdev).
FT over FILS-SHA384 uses a 24 byte FT MIC rather than the 16 byte MIC
used for all other AKMs. This change allows both the FT builder/parser
to handle both lengths of MIC. The mic length is now passed directly
into ie_parse_fast_bss_transition and ie_build_fast_bss_transition
FILS-FT is a special case with respect to the PTK keys. The KCK getter
was updated to handle both FT-FILS AKMs, by returning the offset in
the PTK to the special KCK generated during FILS. A getter for the KCK
length was added, which handles the SHA384 variant. The PTK size was
also updated since FILS-FT can generate an additional 56 bytes of PTK
ifaddr is not guaranteed to be initialized, I'm not sure why there was
no compiler warning. Also replace a | with a || for boolean conditions
and merge the wiphy check with that line.
When handling a scan finished event for a scan we haven't started check
that we were not halfway through a scan request that would have its
results flushed by the external scan.
FT-over-DS is a way to do a Fast BSS Transition using action frames for
the authenticate step. This allows a station to start a fast transition
to a target AP while still being connected to the original AP. This,
in theory, can result in less carrier downtime.
The existing ft_sm_new was removed, and two new constructors were added;
one for over-air, and another for over-ds. The internals of ft.c mostly
remain the same. A flag to distinguish between air/ds was added along
with a new parser to parse the action frames rather than authenticate
frames. The IE parsing is identical.
Netdev now just initializes the auth-proto differently depending on if
its doing over-air or over-ds. A new TX authenticate function was added
and used for over-ds. This will send out the IEs from ft.c with an
FT Request action frame.
The FT Response action frame is then recieved from the AP and fed into
the auth-proto state machine. After this point ft-over-ds behaves the
same as ft-over-air (associate to the target AP).
Some simple code was added in station.c to determine if over-air or
over-ds should be used. FT-over-DS can be beneficial in cases where the
AP is directing us to roam, or if the RSSI falls below a threshold.
It should not be used if we have lost communication to the AP all
(beacon lost) as it only works while we can still talk to the original
AP.
To support FT-over-DS this API needed some slight modifications:
- Instead of setting the DA to netdev->handshake->aa, it is just set to
the same address as the 'to' parameter. The kernel actually requires
and checks for these addresses to match. All occurences were passing
the handshake->aa anyways so this change should have no adverse
affects; and its actually required by ft-over-ds to pass in the
previous BSSID, so hard coding handshake->aa will not work.
- The frequency is is also passed in now, as ft-over-ds needs to use
the frequency of the currently connected AP (netdev->frequency get
set to the new target in netdev_fast_transition. Previous frequency
is also saved now).
- A new vector variant (netdev_send_action_framev) was added as well
to support sending out the FT Request action frame since the FT
TX authenticate function provides an iovec of the IEs. The existing
function was already having to prepend the action frame header to
the body, so its not any more or less copying to do the same thing
with an iovec instead.
Since FT already handles processing the FT IE's (and building for
associate) it didn't make sense to have all the IE building inside
netdev_build_cmd_ft_authenticate. Instead this logic was moved into
ft.c, and an iovec is now passed from FT into
netdev_ft_tx_authenticate. This leaves the netdev command builder
unburdened by the details of FT, as well as prepares for FT-over-DS.
Blacklist some drivers known to crash when interfaces are deleted or
created so that we don't even attempt that before falling back to using
the default interface.
Read the driver name for each wiphy from sysfs if available. I didn't
find a better way to obtain the driver name for a phy than by reading
the dir name that the "driver" symlink points at. For an existing
netdev this can be done using the SIOCETHTOOL ioctl.
manager_interface_dump_done would use manager_create_interfaces() at the
end of the loop iterating over pending_wiphys. To prevent it from
crashing make sure manager_create_interfaces never frees the pending
wiphy state and instead make the caller check whether it needs to be
freed so it can be done safely inside loops.
Instead of having two separate types of scans make the periodic scan
logic a layer on top of the one-off scan requests, with minimum code to
account for the lower priority of those scans and the fact that periodic
scans also receive results from external scans. Also try to simplify
the code for both the periodic and one-off scans. In the SCAN_RESULTS
and SCAN_ABORT add more complete checks of the current request's state
so we avoid some existing crashes related to external scans.
scan_send_next_cmd and start_next_scan_request are now just one function
since their funcionality was similar and start_next_scan_request is used
everywhere. Also the state after the trigger command receives an EBUSY
is now the same as when a new scan is on top of the queue so we have
fewer situations to consider.
This code still does not account for fragmented scans where an external
scan between two or our fragments flushes the results and we lose some
of the results, or for fragmented scans that take over 30s and the
kernel expires some results (both situations are unlikely.)
In both netdev_{authenticate,associate}_event there is no need to check
for in_ft at the start since netdev->ap will always be set if in_ft is
set.
There was also no need to set eapol_sm_set_use_eapol_start, as setting
require_handshake implies this and achieves the same result when starting
the SM.
Since FT operates over Authenticate/Associate, it makes the most sense
for it to behave like the other auth-protos.
This change moves all the FT specific processing out of netdev and into
ft.c. The bulk of the changes were strait copy-pastes from netdev into
ft.c with minor API changes (e.g. remove struct netdev).
The 'in_ft' boolean unforunately is still required for a few reasons:
- netdev_disconnect_event relies on this flag so it can ignore the
disconnect which comes in when doing a fast transition. We cannot
simply check netdev->ap because this would cause the other auth-protos
to not handle a disconnect correctly.
- netdev_associate_event needs to correctly setup the eapol_sm when
in FT mode by setting require_handshake and use_eapol_start to false.
This cannot be handled inside eapol by checking the AKM because an AP
may only advertise a FT AKM, and the initial mobility association
does require the 4-way handshake.
Now the 'ft' module, previously ftutil, will be used to drive FT via
the auth-proto virtual class. This renaming is in preparation as
ftutil will become obsolete since all the IE building/processing is
going to be moved out of netdev. The new ft.c module will utilize
the existing ftutil functionality, but since this is now a full blown
auth protocol naming it 'ft' is better suited.
The duplicate/similar code in netdev_associate_event and
netdev_connect_event leads to very hard to follow code, especially
when you throw OWE/SAE/FILS or full mac cards into the mix.
Currently these protocols finish the connection inside
netdev_associate_event, and set ignore_connect_event. But for full
mac cards we must finish the connection in netdev_connect_event.
In attempt to simplify this, all connections will be completed
and/or the 4-way started in netdev_connect_event. This satisfies
both soft/full mac cards as well as simplifies the FT processing
in netdev_associate_event. Since the FT IEs can be processed in
netdev_connect_event (as they already are to support full mac)
we can assume that any FT processing inside netdev_associate_event
is for a fast transition, not initial mobility association. This
simplifies netdev_ft_process_associate by removing all the blocks
that would get hit if transition == false.
Handling FT this way also fixes FT-SAE which was broken after the
auth-proto changes since the initial mobility association was
never processed if there was an auth-proto running.
SAE was a bit trickier than OWE/FILS because the initial implementation
for SAE did not include parsing raw authenticate frames (netdev skipped
the header and passed just the authentication data). OWE/FILS did not
do this and parse the entire frame in the RX callbacks. Because of this
it was not as simple as just setting some RX callbacks. In addition,
the TX functions include some of the authentication header/data, but
not all (thanks NL80211), so this will require an overhaul to test-sae
since the unit test passes frames from one SM to another to test the
protocol end-to-end (essentially the header needs to be prepended to
any data coming from the TX functions for the end-to-end tests).
Since ERP is only used for FILS and not behaving in the 'normal' ERP
fashion (dealing with actual EAP data, timeouts etc.) we can structure
ERP as a more synchronous protocol, removing the need for a complete
callback.
Now, erp_rx_packet returns a status, so FILS can decide how to handle
any failures. The complete callback was also removed in favor of a
getter for the RMSK (erp_get_rmsk). This allows FILS to syncronously
handle ERP, and potentially fail directly in fils_rx_authenticate.
A new eapol API was added specifically for FILS (eapol_set_started). Since
either way is special cased for FILS, its a bit cleaner to just check the
AKM inside eapol_start and, if FILS, dont start any timeouts or start the
handshake (effectively what eapol_set_started was doing).
This is a new concept applying to any protocol working over authenticate
and/or associate frames (OWE/SAE/FILS). All these protocols behave
similarly enough that they can be unified into a handshake driver
structure.
Now, each protocol will initialize this auth_proto structure inside
their own internal data. The auth_proto will be returned from
the initializer which netdev can then use to manage the protocol by
forwarding authenticate/associate frames into the individual drivers.
The auth_proto consists only of function pointers:
start - starts the protocol
free - frees the driver data
rx_authenticate - receive authenticate frame
rx_associate - receive associate frame
auth_timeout - authenticate frame timed out
assoc_timeout - associate frame timed out
If the setting is true we'll not attempt to remove or create
interfaces on any wiphys and will only use the default interface
(if it exists). If false, force us managing the interfaces. Both
values override the auto logic.
An unexpected Associate event would cause iwd to crash when accessing
netdev->handshake->mde. netdev->handshake is only set if we're
attempting to connect or connected somewhere so check netdev->connected
first.
FILS needs to allocate an extra 16 bytes of key data for the AES-SIV
vector. Instead of leaving it up to the caller to figure this out (as
was done with the GTK builder) eapol_create_common can allocate the
extra space since it knows the MIC length.
This also updates _create_gtk_2_of_2 as it no longer needs to create
an extra data array.
Since FILS does not use a MIC, the 1/4 handler would always get called
for FILS PTK rekeys. We can use the fact that message 1/4 has no MIC as
well as no encrypted data to determine which packet it is. Both no MIC
and no encrypted data means its message 1/4. Anything else is 3/4.
For FILS rekeys, we still derive the PTK using the 4-way handshake.
And for FILS-SHA384 we need the SHA384 KDF variant when deriving.
This change adds both FILS-SHA256 and FILS-SHA384 to the checks
for determining the SHA variant.
crypto_derive_pairwise_ptk was taking a boolean to decide whether to
use SHA1 or SHA256, but for FILS SHA384 may also be required for
rekeys depending on the AKM.
crypto_derive_pairwise_ptk was changed to take l_checksum_type instead
of a boolean to allow for all 3 SHA types.
AP still relies on the get_data/set_length semantics. Its more convenient
to still use these since it avoids the need for extra temporary buffers
when building the rates IE.
The TLV builder APIs were not very intuative, and in some (or all)
cases required access to the builder structure directly, either to
set the TLV buffer or to get the buffer at the end.
This change adds a new API, ie_tlv_builder_set_data, which both sets
the length for the current TLV and copies the TLV data in one go.
This will avoid the need for memcpy(ie_tlv_builder_get_data(...),...)
ie_tlv_builder_finalize was also changed to return a pointer to the
start of the build buffer. This will eliminate the need to access
builder.tlv after building the TLVs.
ie_tlv_builder_init was changed to take an optional buffer to hold
the TLV data. Passing NULL/0 will build the TLV in the internal
buffer. Passing in a pointer and length will build into the passed
in buffer.
Let manager.c signal to wiphy.c when the wiphy parsing from the genl
messages is complete. When we query for existing wiphy using the
GET_WIPHY dump command we get many genl messages per wiphy, on a
notification we only get one message. So after wiphy_create there may
be one or many calls to wiphy_update_from_genl. wiphy_create_complete
is called after all of them, so wiphy.c can be sure it's done with
parsing the wiphy attributes when in prints the new wiphy summary log
message, like it did before manager.c was added.
I had wrongly assumed that all the important wiphy attributes were in
the first message in the dump, but NL80211_ATTR_EXT_FEATURES was not and
wasn't being parsed which was breaking at least testRSSIAgent.
SAE was behaving inconsitently with respect to freeing the state.
It was freeing the SM internally on failure, but requiring netdev
free it on success.
This removes the call to sae_sm_free in sae.c upon failure, and
instead netdev frees the SM in the complete callback in all cases
regardless of success or failure.
From netdev's prospective FILS works the same as OWE/SAE where we create
a fils_sm and forward all auth/assoc frames into the FILS module. The
only real difference is we do not start EAPoL once FILS completes.
FILS (Fast Initial Link Setup) allows a station to negotiate a PTK during
authentication and association. This allows for a faster connection as
opposed to doing full EAP and the 4-way. FILS uses ERP (EAP Reauth Protocol)
to achieve this, but encapsulates the ERP data into an IE inside
authenticate frames. Association is then used to verify both sides have
valid keys, as well as delivering the GTK/IGTK.
FILS will work similar to SAE/OWE/FT where netdev registers a fils_sm, and
then forwards all Auth/Assoc frame data to and from the FILS module.
Keeping the ERP cache on the handshake object allows station.c to
handle all the ERP details and encapsulate them into a handshake.
FILS can then use the ERP cache right from the handshake rather
than getting it itself.
wiphy_select_akm needed to be updated to take a flag, which can be
set to true if there are known reauth keys for this connection. If
we have reauth keys, and FILS is available we will choose it.
If the AP send an associate with an unsupported group status, OWE
was completely starting over and sending out an authenticate frame
when it could instead just resend the associate frame with a
different group.
With FILS support coming there needs to be a way to set the PTK directly.
Other AKMs derive the PTK via the 4-way handshake, but FILS computes the
PTK on its own.
This reverts commit 1e337259ce.
Using util_get_username was wrong in this context. MSCHAPv2 expects us
to only strip the domain name from identities of the form
domain\identity. util_get_username would also strip identities of the
form username@domain.com.
FILS-SHA384 got overlooked and the kek length was being hard coded
to 32 bytes when encrypting the key data. There was also one occurence
where the kek_len was just being set incorrectly.
If the input length is 16 bytes, this means aes_siv_decrypt should
only be verifying the 16 byte SIV and not decrypting any data. If
this is the case, we can skip over the whole AES-CTR portion of
AES-SIV and only verify the SIV.
In eapol_key_handle, 'have_snonce' is checked before decrypting the
key data. For FILS, there will be no snonce so this check can be
skipped if mic_len == 0.
The GTK handshake for FILS uses AES-SIV to encrypt the key data, and
does away with the MIC completely. Now, when finalizing the 2/2 GTK
packet we check the MIC length, and if zero we assume FILS is being
used and we use AES-SIV to encrypt the key data.
For FILS, there is no actual data being encrypted for GTK 2/2 (hence
why the input data length is zero). This results in only the SIV
being generated, which essentially serves the same purpose as a MIC.
FILS does not use a MIC, as well as requires encrypted data on GTK 2/2.
This updates eapol_create_gtk_2_of_2 to pass in extra data to
eapol_create_common, which will reserve room for this encrypted data.
Extra data is only reserved if mic_len == 0.
FILS does not use a MIC in EAPoL frames and also requires encrypted
data on all EAPoL frames. In the common builder the mic_len is now
checked and the flags are set appropriately.
FILS authentication does away with the MIC, so checking for key_mic
in the eapol key frame does not allow FILS to work. Now we pass in
the mic_len to eapol_verify_gtk_1_of_2, and if it is non-zero we can
check that the MIC is present in the frame.
FILS does not require an eapol_sm for authentication, but rekeys
are still performed using the 4-way handshake. Because of this
FILS needs to create a eapol_sm in a 'started' state, but without
calling eapol_start as this will initialize EAP and create handshake
timeouts.
This allows EAPoL to wait for any 4-way packets, and handle them
as rekeys.
ERP (EAP Reauthentication Protocol) allows a station to quickly
reauthenticate using keys from a previous EAP authentication.
This change both implements ERP as well as moves the key cache into
the ERP module.
ERP in its current form is here to only support FILS. ERP is likely not
widespread and there is no easy way to determine if an AP supports ERP
without trying it. Attempting ERP with a non-ERP enabled AP will actually
result in longer connection times since ERP must fail and then full EAP
is done afterwards. For this reason ERP was separated from EAP and a
separate ERP state machine must be created. As it stands now, ERP cannot
be used on its own, only with FILS.
