Update security/openssl to 1.0.2k.

Changes between 1.0.2j and 1.0.2k [26 Jan 2017]

  *) Truncated packet could crash via OOB read

    If one side of an SSL/TLS path is running on a 32-bit host and a specific
    cipher is being used, then a truncated packet can cause that host to
    perform an out-of-bounds read, usually resulting in a crash.

    This issue was reported to OpenSSL by Robert ��wi��cki of Google.
    (CVE-2017-3731)
    [Andy Polyakov]

  *) BN_mod_exp may produce incorrect results on x86_64

    There is a carry propagating bug in the x86_64 Montgomery squaring
    procedure. No EC algorithms are affected. Analysis suggests that attacks
    against RSA and DSA as a result of this defect would be very difficult to
    perform and are not believed likely. Attacks against DH are considered just
    feasible (although very difficult) because most of the work necessary to
    deduce information about a private key may be performed offline. The amount
    of resources required for such an attack would be very significant and
    likely only accessible to a limited number of attackers. An attacker would
    additionally need online access to an unpatched system using the target
    private key in a scenario with persistent DH parameters and a private
    key that is shared between multiple clients. For example this can occur by
    default in OpenSSL DHE based SSL/TLS ciphersuites. Note: This issue is very
    similar to CVE-2015-3193 but must be treated as a separate problem.

    This issue was reported to OpenSSL by the OSS-Fuzz project.
    (CVE-2017-3732)
    [Andy Polyakov]

  *) Montgomery multiplication may produce incorrect results

    There is a carry propagating bug in the Broadwell-specific Montgomery
    multiplication procedure that handles input lengths divisible by, but
    longer than 256 bits. Analysis suggests that attacks against RSA, DSA
    and DH private keys are impossible. This is because the subroutine in
    question is not used in operations with the private key itself and an input
    of the attacker's direct choice. Otherwise the bug can manifest itself as
    transient authentication and key negotiation failures or reproducible
    erroneous outcome of public-key operations with specially crafted input.
    Among EC algorithms only Brainpool P-512 curves are affected and one
    presumably can attack ECDH key negotiation. Impact was not analyzed in
    detail, because pre-requisites for attack are considered unlikely. Namely
    multiple clients have to choose the curve in question and the server has to
    share the private key among them, neither of which is default behaviour.
    Even then only clients that chose the curve will be affected.

    This issue was publicly reported as transient failures and was not
    initially recognized as a security issue. Thanks to Richard Morgan for
    providing reproducible case.
    (CVE-2016-7055)
    [Andy Polyakov]

  *) OpenSSL now fails if it receives an unrecognised record type in TLS1.0
    or TLS1.1. Previously this only happened in SSLv3 and TLS1.2. This is to
    prevent issues where no progress is being made and the peer continually
    sends unrecognised record types, using up resources processing them.
    [Matt Caswell]

(jperkin)