Changes since OpenSSH 6.3
=========================
This release fixes a security bug:
* sshd(8): fix a memory corruption problem triggered during rekeying
when an AES-GCM cipher is selected. Full details of the vulnerability
are available at: http://www.openssh.com/txt/gcmrekey.adv
Checksums:
==========
- SHA1 (openssh-6.4.tar.gz) = 4caf1a50eb3a3da821c16298c4aaa576fe24210c
- SHA1 (openssh-6.4p1.tar.gz) = cf5fe0eb118d7e4f9296fbc5d6884965885fc55d
Reporting Bugs:
===============
- Please read http://www.openssh.com/report.html
Security bugs should be reported directly to openssh@openssh.com
OpenSSH is brought to you by Markus Friedl, Niels Provos, Theo de Raadt,
Kevin Steves, Damien Miller, Darren Tucker, Jason McIntyre, Tim Rice and
Ben Lindstrom.
under NetBSD.org domain.
Multiple TNF hosts have an up-to-date SSHFP record inside the DNS.
This offers a second channel verification for host key fingerprints
(weaker than known_hosts, but spoofing a host on first connect would
also require DNS forgery).
This can provide a trusted second channel (like DANE TLSA records) once
DNSSEC gets more widely used, but for now it is purely informational.
No regression expected, except that the ssh client will print a message
upon first connect to confirm/infirm that it got a correct SSHFP record
from DNS.
Only done for NetBSD.org domain, SSHFP are sadly more an exception than
the rule.
Notified on netbsd-users@, no objection after a week -- committed.
consttime_memequal is the same as the old consttime_bcmp.
explicit_memset is to memset as explicit_bzero was to bcmp.
Passes amd64 release and i386/ALL, but I'm sure I missed some spots,
so please let me know.
OpenSSL now supports AES-NI in evp, not in an engine. We can now get
rid of the no longer maintained aesni engine, which was broken last
summer. Not only can OpenSSL now use AES-NI for everything it did
before we broke it last summer, but it can also use AES-NI for more
encryption modes than before, such as CTR.
Tested on amd64, both vanilla and in an i386 chroot.
ok christos
*) Correct fix for CVE-2013-0169. The original didn't work on AES-NI
supporting platforms or when small records were transferred.
[Andy Polyakov, Steve Henson]
This is primarily a bugfix release.
Features:
* sshd(8): This release turns on pre-auth sandboxing sshd by default for
new installs, by setting UsePrivilegeSeparation=sandbox in sshd_config.
* ssh-keygen(1): Add options to specify starting line number and number of
lines to process when screening moduli candidates, allowing processing
of different parts of a candidate moduli file in parallel
* sshd(8): The Match directive now supports matching on the local (listen)
address and port upon which the incoming connection was received via
LocalAddress and LocalPort clauses.
* sshd(8): Extend sshd_config Match directive to allow setting AcceptEnv
and {Allow,Deny}{Users,Groups}
* Add support for RFC6594 SSHFP DNS records for ECDSA key types. bz#1978
* ssh-keygen(1): Allow conversion of RSA1 keys to public PEM and PKCS8
* sshd(8): Allow the sshd_config PermitOpen directive to accept "none" as
an argument to refuse all port-forwarding requests.
* sshd(8): Support "none" as an argument for AuthorizedPrincipalsFile
* ssh-keyscan(1): Look for ECDSA keys by default. bz#1971
* sshd(8): Add "VersionAddendum" to sshd_config to allow server operators
to append some arbitrary text to the server SSH protocol banner.
Bugfixes:
* ssh(1)/sshd(8): Don't spin in accept() in situations of file
descriptor exhaustion. Instead back off for a while.
* ssh(1)/sshd(8): Remove hmac-sha2-256-96 and hmac-sha2-512-96 MACs as
they were removed from the specification. bz#2023,
* sshd(8): Handle long comments in config files better. bz#2025
* ssh(1): Delay setting tty_flag so RequestTTY options are correctly
picked up. bz#1995
* sshd(8): Fix handling of /etc/nologin incorrectly being applied to root
on platforms that use login_cap.
LP64.
Fixes problems showing up on regression tests on i386 (which work fine on
amd64) i.e. turn:
t_netpgpverify (1/1): 2 test cases
netpgpverify_dsa: [0.309746s] Failed: atf-check failed; see the output of the test for details
netpgpverify_rsa: [0.183148s] Passed.
[0.495102s]
Failed test cases:
t_netpgpverify:netpgpverify_dsa
Summary for 1 test programs:
1 passed test cases.
1 failed test cases.
0 expected failed test cases.
0 skipped test cases.
into:
t_netpgpverify (1/1): 2 test cases
netpgpverify_dsa: [0.236076s] Passed.
netpgpverify_rsa: [0.154680s] Passed.
[0.393034s]
Summary for 1 test programs:
2 passed test cases.
0 failed test cases.
0 expected failed test cases.
0 skipped test cases.
agc-netpgp-standalone branch.
Rewrite the netpgpverify(1) functionality from RFC4880 up. This is a
completely new implementation, and uses its own bignum library derived
from libtommath. Apart from libz and libbz2, it just uses its own
library and is self-contained - this makes it easier to embed, and to
use from scripting languages.
netpgpverify(1) now verifies all the signed files i've thrown at it,
and the added bonus of using no functionality from libcrypto - all of
its bignum functionality comes from its own libnetpgpverify.so.
netpgpverify(1) now verifies not only signatures on binary files, but
also signatures on text documents. This fixes PR/46930. Please don't
start me on the hoops I had to jump through to calculate the digests
on text files; trust me, you will regret it.
% supersize `which netpgpverify`
text data bss dec hex filename
4452 860 72 5384 1508 /usr/bin/netpgpverify
79542 1408 0 80950 13c36 /usr/lib/libz.so.1
43994 984 488 45466 b19a /usr/lib/libgcc_s.so.1
1318116 49644 69272 1437032 15ed68 /usr/lib/libc.so.12
57253 4184 0 61437 effd /usr/lib/libbz2.so.1
108726 1712 0 110438 1af66 /usr/lib/libnetpgpverify.so.4
1612083 58792 69832 1740707 0x1a8fa3 total
%
% make t
env LD_LIBRARY_PATH=/usr/src/crypto/external/bsd/netpgp-standalone/lib/verify ./netpgpverify -c verify b.gpg > output16
diff expected16 output16
rm -f output16
env LD_LIBRARY_PATH=/usr/src/crypto/external/bsd/netpgp-standalone/lib/verify ./netpgpverify -c verify a.gpg > output17
diff expected17 output17
rm -f output17
env LD_LIBRARY_PATH=/usr/src/crypto/external/bsd/netpgp-standalone/lib/verify ./netpgpverify -c verify gpgsigned-a.gpg > output18
diff expected18 output18
rm -f output18
env LD_LIBRARY_PATH=/usr/src/crypto/external/bsd/netpgp-standalone/lib/verify ./netpgpverify -c verify NetBSD-6.0_RC2_hashes.asc > output19
diff expected19 output19
rm -f output19
...
env LD_LIBRARY_PATH=/usr/src/crypto/external/bsd/netpgp-standalone/lib/verify ./netpgpverify -k dsa-pubring.gpg in2.asc > output45
diff expected45 output45
rm -f output45
env LD_LIBRARY_PATH=/usr/src/crypto/external/bsd/netpgp-standalone/lib/verify ./netpgpverify -k problem-pubring.gpg NetBSD-6.0_hashes.asc > output46
diff expected46 output46
rm -f output46
cd tests/netpgpverify && make && atf-run
atf2kyua: I: Removing stale Kyuafiles from /tmp/.XXXXXX.004966aa
atf2kyua: I: Converting /usr/src/crypto/external/bsd/netpgp-standalone/tests/netpgpverify/Atffile -> /tmp/.XXXXXX.004966aa/Kyuafile
t_netpgpverify:netpgpverify_rsa -> passed [0.221s]
t_netpgpverify:netpgpverify_dsa -> passed [0.117s]
2/2 passed (0 failed)
Committed action 19
%
Features:
* ssh-keygen(1): Add optional checkpoints for moduli screening
* ssh-add(1): new -k option to load plain keys (skipping certificates)
* sshd(8): Add wildcard support to PermitOpen, allowing things like
"PermitOpen localhost:*". bz #1857
* ssh(1): support for cancelling local and remote port forwards via the
multiplex socket. Use ssh -O cancel -L xx:xx:xx -R yy:yy:yy user@host"
to request the cancellation of the specified forwardings
* support cancellation of local/dynamic forwardings from ~C commandline
Bugfixes:
* ssh(1): ensure that $DISPLAY contains only valid characters before
using it to extract xauth data so that it can't be used to play local
shell metacharacter games.
* ssh(1): unbreak remote portforwarding with dynamic allocated listen ports
* scp(1): uppress adding '--' to remote commandlines when the first
argument does not start with '-'. saves breakage on some
difficult-to-upgrade embedded/router platforms
* ssh(1)/sshd(8): fix typo in IPQoS parsing: there is no "AF14" class,
but there is an "AF21" class
* ssh(1)/sshd(8): do not permit SSH2_MSG_SERVICE_REQUEST/ACCEPT during
rekeying
* ssh(1): skip attempting to create ~/.ssh when -F is passed
* sshd(8): unbreak stdio forwarding when ControlPersist is in use; bz#1943
* sshd(1): send tty break to pty master instead of (probably already
closed) slave side; bz#1859
* sftp(1): silence error spam for "ls */foo" in directory with files;
bz#1683
* Fixed a number of memory and file descriptor leaks
the client send its version string first if it is configured to speak
v2 only (the old hack of waiting to see the server version is only
really useful if you might be speaking v1 to some servers). The option
is on by default but can be disabled from the config file.
This aligns the OpenSSH client behavior with most other implementations
and eliminates a major source of connection delays and failures when
speaking SSH through particularly stupid proxies, of which, sadly, there
are many.
This change has also been submitted to OpenSSH as their bug #1999.
check for potentially exploitable overflows in asn1_d2i_read_bio
BUF_mem_grow and BUF_mem_grow_clean. Refuse attempts to shrink buffer
in CRYPTO_realloc_clean. (CVE-2012-2110)
explicitly seed the OpenSSL RNG in each new process rather than letting
it repeatedly open /dev/urandom to reseed, which depletes entropy severely.
Note that the OpenSSH part of this fix works better on NetBSD than it would
on many other platforms because on NetBSD, if you don't reopen /dev/urandom,
repeated reads don't deplete entropy. On other platforms, some other
approach might be required.
Note also that this problem does not arise on OpenBSD because OpenBSD seems
to have patched OpenSSL to seed the RAND functions from arc4random()! That
seems dangerous, so I am not taking that approach here.
userland part of PR kern/44952 by Wolfgang Stukenbrock, just changed
the "frag" option name to "esp_frag", for consistency to the existing
option of similar effect in racoon(8)
-rev.21964 for DTLS Plaintext Recovery Attack (CVE-2011-4108)
-rev.21961 for Uninitialized SSL 3.0 Padding (CVE-2011-4576)
-rev.21456+21954 for Malformed RFC 3779 Data Can Cause Assertion Failures
(CVE-2011-4577)
(rev.21456 is not mentioned in the advisory, but there is code overlap)
-rev.21958 for SGC Restart DoS Attack (CVE-2011-4619)
-rev.21956 for Invalid GOST parameters DoS Attack (CVE-2012-0027)
ipsec_*_policy() functions, as it was documented and used by clients
-remove "ipsec_policy_t" which was undocumented and only present
in the KAME version of the ipsec.h header
-misc cleanup of historical artefacts, and to remove unnecessary
differences between KAME ans FAST_IPSEC
resultant key listing
when using json to format keys returned from libnetpgp, also prepare for
machine-readable format ("mr") as well as human ("human"), even though
it's not yet used.
our version of GNU as decides to silently ignore it. Never versions on
other systems and Clang actually keep it around and create bad object
files. The object files are identical after the changes modulo the GOT
entry in the symbol table.
the "rep ret" trick is recommended by AMD as a branch prediction
optimization in certain circumstances (quoting their manual):
- any kind of branch (either conditional or unconditional) that has the
single-byte near-return RET instruction as its target
- a conditional branch that occurs in the code directly before the
single-byte near-return RET instruction.
