In addition, start ordering header file includes to be in more
consistent order: system header files, src/utils, src/*, same
directory as the *.c file.
It looks like GnuTLS (at least newer versions) is using random padding
on the application data and the previously used 100 byte extra buffer
for tls_connection_encrypt() calls was not enough to handle all cases.
This resulted in semi-random authentication failures with EAP-PEAP and
EAP-TTLS during Phase 2.
Increase the extra space for encryption from 100 to 300 bytes and add an
error message into tls_gnutls.c to make it easier to notice this issue
should it ever show up again even with the larger buffer.
Previous version assumed that the Flags field is always present and
ended up reading one octet past the end of the buffer should the Flags
field be missing. The message length would also be set incorrectly
(size_t)-1 or (size_t)-5, but it looks like reassembly code ended up
failing in malloc before actually using this huge length to read data.
RFC 2716 uses a somewhat unclear description on what exactly is included
in the TLS Ack message ("no data" can refer to either Data field in 4.1
or TLS Data field in 4.2), so in theory, it would be possible for some
implementations to not include Flags field. However,
EAP-{PEAP,TTLS,FAST} need the Flags field in Ack messages, too, for
indicating the used version.
The EAP peer code will now accept the no-Flags case as an Ack message if
EAP workarounds are enabled (which is the default behavior). If
workarounds are disabled, the message without Flags field will be
rejected.
[Bug 292]
Windows Server 2008 NPS gets very confused if the TLS Message Length is
not included in the Phase 1 messages even if fragmentation is not used.
If the TLS Message Length field is not included in ClientHello message,
NPS seems to decide to use the ClientHello data (excluding first six
octets, i.e., EAP header, type, Flags) as the OuterTLVs data in
Cryptobinding Compound_MAC calculation (per PEAPv2; not MS-PEAP)..
Lets add the TLS Message Length to PEAPv0 Phase 1 messages to get rid of
this issue. This seems to fix Cryptobinding issues with NPS and PEAPv0
is now using optional Cryptobinding by default (again) since there are
no known interop issues with it anymore.
I fixed the engine issue in phase2 of EAP-TTLS. The problem was that you
only defined one engine variable, which was read already in phase1. I
defined some new variables:
engine2
engine2_id
pin2
and added support to read those in phase2 wheres all the engine
variables without number are only read in phase1. That solved it and I
am now able to use an engine also in EAP-TTLS phase2.
Even though we try to disable TLS compression, it is possible that this
cannot be done with all TLS libraries. For example, OpenSSL 0.9.8 does not
seem to have a configuration item for disabling all compression (0.9.9 has
such an option). If compression is used, Phase 2 decryption may end up
producing more data than the input buffer due to compressed data. This
shows up especially with EAP-TNC that uses very compressible data format.
As a workaround, increase the decryption buffer length to (orig_len+500)*3.
This is a hack, but at least it handles most cases. TLS compression should
really be disabled for EAP use of TLS, but since this can show up with
common setups, it is better to handle this case.
