OpenSSL 1.0.2 introduced the "multiblock" performance improvement. This
feature only applies on 64 bit x86 architecture platforms that support AES
NI instructions. A defect in the implementation of "multiblock" can cause
OpenSSL's internal write buffer to become incorrectly set to NULL when
using non-blocking IO. Typically, when the user application is using a
socket BIO for writing, this will only result in a failed connection.
However if some other BIO is used then it is likely that a segmentation
fault will be triggered, thus enabling a potential DoS attack.
CVE-2015-0290
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Andy Polyakov <appro@openssl.org>
Add support for ServerInfo SSL_CONF option and update documentation. This
was wrongly omitted from the 1.0.2 release.
Reviewed-by: Richard Levitte <levitte@openssl.org>
The function tls1_PRF counts the number of digests in use and partitions
security evenly between them. There always needs to be at least one digest
in use, otherwise this is an internal error. Add a sanity check for this.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit 668f6f08c6)
If SSL_check_chain is called with a NULL X509 object or a NULL EVP_PKEY
or the type of the public key is unrecognised then the local variable
|cpk| in tls1_check_chain does not get initialised. Subsequently an
attempt is made to deref it (after the "end" label), and a seg fault will
result.
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
(cherry picked from commit d813f9eb38)
Ensure all malloc failures return -1.
Reported by Adam Langley (Google).
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 06c6a2b4a3)
Fix security issue where under certain conditions a client can complete a
handshake with an unseeded PRNG. The conditions are:
- Client is on a platform where the PRNG has not been seeded, and the
user has not seeded manually
- A protocol specific client method version has been used (i.e. not
SSL_client_methodv23)
- A ciphersuite is used that does not require additional random data
from the PRNG beyond the initial ClientHello client random
(e.g. PSK-RC4-SHA)
If the handshake succeeds then the client random that has been used will
have been generated from a PRNG with insufficient entropy and therefore
the output may be predictable.
For example using the following command with an unseeded openssl will
succeed on an unpatched platform:
openssl s_client -psk 1a2b3c4d -tls1_2 -cipher PSK-RC4-SHA
CVE-2015-0285
Reviewed-by: Richard Levitte <levitte@openssl.org>
(cherry picked from commit e1b568dd24)
Since commit 741c9959 ("DTLS revision."), we put the wrong protocol
version into our ClientHello for DTLS1_BAD_VER. The old DTLS
code which used ssl->version was replaced by the more generic SSL3 code
which uses ssl->client_version. The Cisco ASA no longer likes our
ClientHello.
RT#3711
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit f7683aaf36)
Commit 9cf0f187 in HEAD, and 68039af3 in 1.0.2, removed a version check
from dtls1_buffer_message() which was needed to distinguish between DTLS
1.x and Cisco's pre-standard version of DTLS (DTLS1_BAD_VER).
Based on an original patch by David Woodhouse <dwmw2@infradead.org>
RT#3703
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 5178a16c43)
This patch uses warning/fatal constants instead of numbers with comments for
warning/alerts in d1_pkt.c and s3_pkt.c
RT#3725
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit fd865cadcb)
Miscellaneous unchecked malloc fixes. Also fixed some mem leaks on error
paths as I spotted them along the way.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 918bb86529)
Conflicts:
crypto/bio/bss_dgram.c
Some Cisco appliances use a pre-standard version number for DTLS. We support
this as DTLS1_BAD_VER within the code.
This change fixes d2i_SSL_SESSION for that DTLS version.
Based on an original patch by David Woodhouse <dwmw2@infradead.org>
RT#3704
Reviewed-by: Tim Hudson <tjh@openssl.org>
Conflicts:
ssl/ssl_asn1.c
Fixed various missing return value checks in ssl3_send_newsession_ticket.
Also a mem leak on error.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Conflicts:
ssl/s3_srvr.c
This should be a one off operation (subsequent invokation of the
script should not move them)
This commit is for the 1.0.2 changes
Reviewed-by: Tim Hudson <tjh@openssl.org>
Sometimes it fails to format them very well, and sometimes it corrupts them!
This commit moves some particularly problematic ones.
Conflicts:
crypto/bn/bn.h
crypto/ec/ec_lcl.h
crypto/rsa/rsa.h
demos/engines/ibmca/hw_ibmca.c
ssl/ssl.h
ssl/ssl3.h
Reviewed-by: Tim Hudson <tjh@openssl.org>
indent will not alter them when reformatting comments
(cherry picked from commit 1d97c84351)
Conflicts:
crypto/bn/bn_lcl.h
crypto/bn/bn_prime.c
crypto/engine/eng_all.c
crypto/rc4/rc4_utl.c
crypto/sha/sha.h
ssl/kssl.c
ssl/t1_lib.c
Reviewed-by: Tim Hudson <tjh@openssl.org>
ssl3_setup_buffers or pqueue_insert fail. The former will fail if there is a
malloc failure, whilst the latter will fail if attempting to add a duplicate
record to the queue. This should never happen because duplicate records should
be detected and dropped before any attempt to add them to the queue.
Unfortunately records that arrive that are for the next epoch are not being
recorded correctly, and therefore replays are not being detected.
Additionally, these "should not happen" failures that can occur in
dtls1_buffer_record are not being treated as fatal and therefore an attacker
could exploit this by sending repeated replay records for the next epoch,
eventually causing a DoS through memory exhaustion.
Thanks to Chris Mueller for reporting this issue and providing initial
analysis and a patch. Further analysis and the final patch was performed by
Matt Caswell from the OpenSSL development team.
CVE-2015-0206
Reviewed-by: Dr Stephen Henson <steve@openssl.org>
(cherry picked from commit 652ff0f4796eecd8729b4690f2076d1c7ccb2862)
Fix to prevent use of DH client certificates without sending
certificate verify message.
If we've used a client certificate to generate the premaster secret
ssl3_get_client_key_exchange returns 2 and ssl3_get_cert_verify is
never called.
We can only skip the certificate verify message in
ssl3_get_cert_verify if the client didn't send a certificate.
Thanks to Karthikeyan Bhargavan for reporting this issue.
CVE-2015-0205
Reviewed-by: Matt Caswell <matt@openssl.org>
of the crash due to p being NULL. Steve's fix prevents this situation from
occuring - however this is by no means obvious by looking at the code for
dtls1_get_record. This fix just makes things look a bit more sane.
Reviewed-by: Dr Stephen Henson <steve@openssl.org>
OpenSSL clients would tolerate temporary RSA keys in non-export
ciphersuites. It also had an option SSL_OP_EPHEMERAL_RSA which
enabled this server side. Remove both options as they are a
protocol violation.
Thanks to Karthikeyan Bhargavan for reporting this issue.
(CVE-2015-0204)
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Fix bug where an OpenSSL client would accept a handshake using an
ephemeral ECDH ciphersuites with the server key exchange message omitted.
Thanks to Karthikeyan Bhargavan for reporting this issue.
CVE-2014-3572
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit b15f876964)
Since SGC has been removed from OpenSSL 1.0.2 the
SSL3_FLAGS_SGC_RESTART_DONE is no longer used. However the #define is
retained for compatibility.
Reviewed-by: Matt Caswell <matt@openssl.org>
MS Server gated cryptography is obsolete and dates from the time of export
restrictions on strong encryption and is only used by ancient versions of
MSIE.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 63eab8a620)
When parsing ClientHello clear any existing extension state from
SRP login and SRTP profile.
Thanks to Karthikeyan Bhargavan for reporting this issue.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 4f605ccb77)
Conflicts:
ssl/t1_lib.c
From BoringSSL
- Send an alert when the client key exchange isn't correctly formatted.
- Reject overly short RSA ciphertexts to avoid a (benign) out-of-bounds memory access.
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
(cherry picked from commit 4aecfd4d9f)
The client_version needs to be preserved for the RSA key exchange.
This change also means that renegotiation will, like TLS, repeat the old
client_version rather than advertise only the final version. (Either way,
version change on renego is not allowed.) This is necessary in TLS to work
around an SChannel bug, but it's not strictly necessary in DTLS.
(From BoringSSL)
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit ec1af3c419)
Odd-length lists should be rejected everywhere upon parsing. Nevertheless,
be extra careful and add guards against off-by-one reads.
Also, drive-by replace inexplicable double-negation with an explicit comparison.
Reviewed-by: Matt Caswell <matt@openssl.org>
The Supported Elliptic Curves extension contains a vector of NamedCurves
of 2 bytes each, so the total length must be even. Accepting odd-length
lists was observed to lead to a non-exploitable one-byte out-of-bounds
read in the latest development branches (1.0.2 and master). Released
versions of OpenSSL are not affected.
Thanks to Felix Groebert of the Google Security Team for reporting this issue.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 33d5ba8629)
we will support then dtls1_do_write can go into an infinite loop. This commit
fixes that.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit d3d9eef316)
at least the minimum or it will fail.
There were some instances in dtls1_query_mtu where the final mtu can end up
being less than the minimum, i.e. where the user has set an mtu manually. This
shouldn't be allowed. Also remove dtls1_guess_mtu that, despite having
logic for guessing an mtu, was actually only ever used to work out the minimum
mtu to use.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 1620a2e49c)
and instead use the value provided by the underlying BIO. Also provide some
new DTLS_CTRLs so that the library user can set the mtu without needing to
know this constant. These new DTLS_CTRLs provide the capability to set the
link level mtu to be used (i.e. including this IP/UDP overhead). The previous
DTLS_CTRLs required the library user to subtract this overhead first.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 59669b6abf)
Conflicts:
ssl/d1_both.c
mtu that we have received is not less than the minimum. If its less it uses the
minimum instead. The second call to query the mtu does not do that, but
instead uses whatever comes back. We have seen an instance in RT#3592 where we
have got an unreasonably small mtu come back. This commit makes both query
checks consistent.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 6abb0d1f8e)
automatically updated, and we should use the one provided instead.
Unfortunately there are a couple of locations where this is not respected.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 001235778a)
RT#3592 provides an instance where the OPENSSL_assert that this commit
replaces can be hit. I was able to recreate this issue by forcing the
underlying BIO to misbehave and come back with very small mtu values. This
happens the second time around the while loop after we have detected that the
MTU has been exceeded following the call to dtls1_write_bytes.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit cf75017bfd)
Previously, state variant was not advanced, which resulted in state
being stuck in the st1 variant (usually "_A").
This broke certificate callback retry logic when accepting connections
that were using SSLv2 ClientHello (hence reusing the message), because
their state never advanced to SSL3_ST_SR_CLNT_HELLO_C variant required
for the retry code path.
Reported by Yichun Zhang (agentzh).
Signed-off-by: Piotr Sikora <piotr@cloudflare.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
the session's version (server).
See also BoringSSL's commit bdf5e72f50e25f0e45e825c156168766d8442dde.
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
(cherry picked from commit 9e189b9dc1)
once the ChangeCipherSpec message is received. Previously, the server would
set the flag once at SSL3_ST_SR_CERT_VRFY and again at SSL3_ST_SR_FINISHED.
This would allow a second CCS to arrive and would corrupt the server state.
(Because the first CCS would latch the correct keys and subsequent CCS
messages would have to be encrypted, a MitM attacker cannot exploit this,
though.)
Thanks to Joeri de Ruiter for reporting this issue.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit e94a6c0ede)
The server must send a NewSessionTicket message if it advertised one
in the ServerHello, so make a missing ticket message an alert
in the client.
An equivalent change was independently made in BoringSSL, see commit
6444287806d801b9a45baf1f6f02a0e3a16e144c.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit de2c7504eb)
The client sends a session ID with the session ticket, and uses
the returned ID to detect resumption, so we do not need to peek
at handshake messages: s->hit tells us explicitly if we're resuming.
An equivalent change was independently made in BoringSSL, see commit
407886f589cf2dbaed82db0a44173036c3bc3317.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 980bc1ec61)
Conflicts:
ssl/s3_clnt.c
The same change was independently made in BoringSSL, see commit
9eaeef81fa2d4fd6246dc02b6203fa936a5eaf67
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 7b3ba508af)
This ensures that it's zeroed even if the SSL object is reused
(as in ssltest.c). It also ensures that it applies to DTLS, too.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit a06cd5d056)
When no-ssl3 is set only make SSLv3 disabled by default. Retain -ssl3
options for s_client/s_server/ssltest.
When no-ssl3-method is set SSLv3_*method() is removed and all -ssl3
options.
We should document this somewhere, e.g. wiki, FAQ or manual page.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 3881d8106d)
The supported signature algorithms extension needs to be processed before
the certificate to use is decided and before a cipher is selected (as the
set of shared signature algorithms supported may impact the choice).
Reviewed-by: Matt Caswell <matt@openssl.org>
see RT #3203
Future versions of OpenSSL use the canonical terms "ECDHE" and "DHE"
as configuration strings and compilation constants. This patch
introduces aliases so that the stable 1.0.2 branch can be
forward-compatible with code and configuration scripts that use the
normalized terms, while avoiding changing any library output for
stable users.
Signed-off-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Matt Caswell <matt@openssl.org>
Tighten client-side session ticket handling during renegotiation:
ensure that the client only accepts a session ticket if the server sends
the extension anew in the ServerHello. Previously, a TLS client would
reuse the old extension state and thus accept a session ticket if one was
announced in the initial ServerHello.
Reviewed-by: Bodo Moeller <bodo@openssl.org>
(cherry picked from commit d663df2399)
SSL_set_SSL_CTX is used to change the SSL_CTX for SNI, keep the
supported signature algorithms and raw cipherlist.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(cherry picked from commit 14e14bf696)
Since SNI will typically switch the SSL_CTX structure to the one
corresponding to the appopriate server we need to parse custom
extensions using the switched SSL_CTX not the original one. This
is done by parsing custom extensions *after* SNI.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
When we're configured with no-ssl3 and we receive an SSL v3 Client Hello, we set
the method to NULL. We didn't used to do that, and it breaks things. This is a
regression introduced in 62f45cc27d. Keep the old
method since the code is not able to deal with a NULL method at this time.
CVE-2014-3569, PR#3571
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 392fa7a952)
Related to CVE-2014-3513
This fix was developed by the OpenSSL Team
Reviewed-by: Tim Hudson <tjh@openssl.org>
Conflicts:
util/mkdef.pl
util/ssleay.num
Conflicts:
util/mkdef.pl
CVE-2014-3513
This issue was reported to OpenSSL on 26th September 2014, based on an original
issue and patch developed by the LibreSSL project. Further analysis of the issue
was performed by the OpenSSL team.
The fix was developed by the OpenSSL team.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Original commit adb46dbc6d)
Use the new constant-time methods consistently in s3_srvr.c
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
(cherry picked from commit 455b65dfab)
that bad encryptions are treated like random session keys in constant
time.
(cherry picked from commit adb46dbc6d)
Reviewed-by: Rich Salz <rsalz@openssl.org>
Also tweak s3_cbc.c to use new constant-time methods.
Also fix memory leaks from internal errors in RSA_padding_check_PKCS1_OAEP_mgf1
This patch is based on the original RT submission by Adam Langley <agl@chromium.org>,
as well as code from BoringSSL and OpenSSL.
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Conflicts:
crypto/rsa/rsa_oaep.c
crypto/rsa/rsa_pk1.c
ssl/s3_cbc.c
that fixed PR#3450 where an existing cast masked an issue when i was changed
from int to long in that commit
Picked up on z/linux (s390) where sizeof(int)!=sizeof(long)
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit b5ff559ff9)
Fix a bug in handling of 128 byte long PSK identity in
psk_client_callback.
OpenSSL supports PSK identities of up to (and including) 128 bytes in
length. PSK identity is obtained via the psk_client_callback,
implementors of which are expected to provide a NULL-terminated
identity. However, the callback is invoked with only 128 bytes of
storage thus making it impossible to return a 128 byte long identity and
the required additional NULL byte.
This CL fixes the issue by passing in a 129 byte long buffer into the
psk_client_callback. As a safety precaution, this CL also zeroes out the
buffer before passing it into the callback, uses strnlen for obtaining
the length of the identity returned by the callback, and aborts the
handshake if the identity (without the NULL terminator) is longer than
128 bytes.
(Original patch amended to achieve strnlen in a different way.)
Reviewed-by: Rich Salz <rsalz@openssl.org>
(cherry picked from commit be0d851732)
Fix comments in ssltest.c: return value of 0 now means extension is
omitted and add_cb is not called for servers if the corresponding
extension is absent in ClientHello.
Test add_cb is not called if extension is not received.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit f47e203975)
Instead of SSL_CTX_set_custom_cli_ext and SSL_CTX_set_custom_srv_ext
use SSL_CTX_add_client_custom_ext and SSL_CTX_add_server_custom_ext.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 8cafe9e8bf)
Support separate parse and add callback arguments.
Add new callback so an application can free extension data.
Change return value for send functions so < 0 is an error 0
omits extension and > 0 includes it. This is more consistent
with the behaviour of other functions in OpenSSL.
Modify parse_cb handling so <= 0 is an error.
Make SSL_CTX_set_custom_cli_ext and SSL_CTX_set_custom_cli_ext argument
order consistent.
NOTE: these changes WILL break existing code.
Remove (now inaccurate) in line documentation.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 33f653adf3)
Use "parse" and "add" for function and callback names instead of
"first" and "second".
Change arguments to callback so the extension type is unsigned int
and the buffer length is size_t. Note: this *will* break existing code.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit de2a9e38f3)
Since sanity checks are performed for all custom extensions the
serverinfo checks are no longer needed.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 707b026d78)
Conflicts:
ssl/ssl3.h
ssl/t1_lib.c
Reject attempts to use extensions handled internally.
Add flags to each extension structure to indicate if an extension
has been sent or received. Enforce RFC5246 compliance by rejecting
duplicate extensions and unsolicited extensions and only send a
server extension if we have sent the corresponding client extension.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit 28ea0a0c6a)
Use the same structure for client and server custom extensions.
Add utility functions in new file t1_ext.c.
Use new utility functions to handle custom server and client extensions
and remove a lot of code duplication.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit ecf4d66090)
Conflicts:
ssl/ssl_lib.c
ssl/ssl_locl.h
ssl/t1_lib.c
Move custom extension structures from SSL_CTX to CERT structure.
This change means the form can be revised in future without binary
compatibility issues. Also since CERT is part of SSL structures
so per-SSL custom extensions could be supported in future as well as
per SSL_CTX.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>
(cherry picked from commit b83294fe30)
Conflicts:
ssl/ssl.h
ssl/ssl_cert.c
ssl/ssl_locl.h
Pull constant-time methods out to a separate header, add tests.
Reviewed-by: Bodo Moeller <bodo@openssl.org>
(cherry picked from commit 5a3d21c058)
Conflicts:
ssl/s3_cbc.c
test/Makefile
Extra SSL_get_selected_srtp_profile() declaration in ssl/srtp.h
causes -Werror builds to fail.
Cherry-picked from 3609b02305
Reviewed-by: Tim Hudson <tjh@openssl.org>
Limit the number of empty records that will be processed consecutively
in order to prevent ssl3_get_record from never returning.
Reported by "oftc_must_be_destroyed" and George Kadianakis.
Reviewed-by: Bodo Moeller <bodo@openssl.org>
The addition of SRP authentication needs to be checked in various places
to work properly. Specifically:
A certificate is not sent.
A certificate request must not be sent.
Server key exchange message must not contain a signature.
If appropriate SRP authentication ciphersuites should be chosen.
Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 8f5a8805b82d1ae81168b11b7f1506db9e047dec)
If a client attempted to use an SRP ciphersuite and it had not been
set up correctly it would crash with a null pointer read. A malicious
server could exploit this in a DoS attack.
Thanks to Joonas Kuorilehto and Riku Hietamäki from Codenomicon
for reporting this issue.
CVE-2014-2970
Reviewed-by: Tim Hudson <tjh@openssl.org>
In a couple of functions, a sequence number would be calculated twice.
Additionally, in |dtls1_process_out_of_seq_message|, we know that
|frag_len| <= |msg_hdr->msg_len| so the later tests for |frag_len <
msg_hdr->msg_len| can be more clearly written as |frag_len !=
msg_hdr->msg_len|, since that's the only remaining case.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>
Previously, a truncated DTLS fragment in
|dtls1_process_out_of_seq_message| would cause *ok to be cleared, but
the return value would still be the number of bytes read. This would
cause |dtls1_get_message| not to consider it an error and it would
continue processing as normal until the calling function noticed that
*ok was zero.
I can't see an exploit here because |dtls1_get_message| uses
|s->init_num| as the length, which will always be zero from what I can
see.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>
The |pqueue_insert| function can fail if one attempts to insert a
duplicate sequence number. When handling a fragment of an out of
sequence message, |dtls1_process_out_of_seq_message| would not call
|dtls1_reassemble_fragment| if the fragment's length was zero. It would
then allocate a fresh fragment and attempt to insert it, but ignore the
return value, leaking the fragment.
This allows an attacker to exhaust the memory of a DTLS peer.
Fixes CVE-2014-3507
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Emilia Käsper <emilia@openssl.org>
