Call for testing: OpenSSH 8.3

Tue May 12 13:24:48 AEST 2020

Hi,

OpenSSH 8.3p1 is almost ready for release, so we would appreciate testing
on as many platforms and systems as possible. This is a bugfix release.

Snapshot releases for portable OpenSSH are available from
http://www.mindrot.org/openssh_snap/

The OpenBSD version is available in CVS HEAD:
http://www.openbsd.org/anoncvs.html

Portable OpenSSH is also available via git using the
instructions at http://www.openssh.com/portable.html#cvs
At https://anongit.mindrot.org/openssh.git/ or via a mirror at Github:
https://github.com/openssh/openssh-portable

Running the regression tests supplied with Portable OpenSSH does not
require installation and is a simply:

$ ./configure && make tests

Live testing on suitable non-production systems is also appreciated.
Please send reports of success or failure to
openssh-unix-dev at mindrot.org. Security bugs should be reported
directly to openssh at openssh.com.

Below is a summary of changes. More detail may be found in the ChangeLog
in the portable OpenSSH tarballs.

Thanks to the many people who contributed to this release.

Security
========

 * scp(1): when receiving files, scp(1) could be become desynchronised
   if a utimes(2) system call failed. This could allow file contents
   to be interpreted as file metadata and thereby permit an adversary
   to craft a file system that, when copied with scp(1) in a
   configuration that caused utimes(2) to fail (e.g. under a SELinux
   policy or syscall sandbox), transferred different file names and
   contents to the actual file system layout.

   Exploitation of this is not likely as utimes(2) does not fail under
   normal circumstances. Successful exploitation is not silent - the
   output of scp(1) would show transfer errors followed by the actual
   file(s) that were received.

   Finally, filenames returned from the peer are (since openssh-8.0)
   matched against the user's requested destination, thereby
   disallowing a successful exploit from writing files outside the
   user's selected target glob (or directory, in the case of a
   recursive transfer). This ensures that this attack can achieve no
   more than a hostile peer is already able to do within the scp
   protocol.

Potentially-incompatible changes
================================

This release includes a number of changes that may affect existing
configurations:

 * sftp(1): reject an argument of "-1" in the same way as ssh(1) and
   scp(1) do instead of accepting and silently ignoring it.

Changes since OpenSSH 8.2
=========================

The focus of this release is bug fixing.

New Features
------------

 * sshd(8): make IgnoreRhosts a tri-state option: "yes" to ignore
   rhosts/shosts, "no" allow rhosts/shosts or (new) "shosts-only"
   to allow .shosts files but not .rhosts.

 * sshd(8): allow the IgnoreRhosts directive to appear anywhere in a
   sshd_config, not just before any Match blocks; bz3148

 * ssh(1): add %TOKEN percent expansion for the LocalFoward and
   RemoteForward keywords when used for Unix domain socket forwarding.
   bz#3014

 * all: allow loading public keys from the unencrypted envelope of a
   private key file if no corresponding public key file is present.

 * ssh(1), sshd(8): prefer to use chacha20 from libcrypto where
   possible instead of the (slower) portable C implementation included
   in OpenSSH.

 * ssh-keygen(1): add ability to dump the contents of a binary key
   revocation list via "ssh-keygen -lQf /path" bz#3132

Bugfixes
--------

 * ssh(1): fix IdentitiesOnly=yes to also apply to keys loaded from
   a PKCS11Provider; bz#3141

 * ssh-keygen(1): avoid NULL dereference when trying to convert an
   invalid RFC4716 private key.

 * scp(2): when performing remote-to-remote copies using "scp -3",
   start the second ssh(1) channel with BatchMode=yes enabled to
   avoid confusing and non-deterministic ordering of prompts.

 * ssh(1), ssh-keygen(1): when signing a challenge using a FIDO token,
   perform hashing of the message to be signed in the middleware layer
   rather than in OpenSSH code. This permits the use of security key
   middlewares that perform the hashing implicitly, such as Windows
   Hello.

 * ssh(1): fix incorrect error message for "too many known hosts
   files." bz#3149

 * ssh(1): make failures when establishing "Tunnel" forwarding
   terminate the connection when ExitOnForwardFailure is enabled;
   bz#3116

 * ssh-keygen(1): fix printing of fingerprints on private keys and add
   a regression test for same.

 * sshd(8): document order of checking AuthorizedKeysFile (first) and
   AuthorizedKeysCommand (subsequently, if the file doesn't match);
   bz#3134

 * sshd(8): document that /etc/hosts.equiv and /etc/shosts.equiv are
   not considered for HostbasedAuthentication when the target user is
   root; bz#3148

 * ssh(1), ssh-keygen(1): fix NULL dereference in private certificate
   key parsing (oss-fuzz #20074).

 * ssh(1), sshd(8): more consistency between sets of %TOKENS are
   accepted in various configuration options.

 * ssh(1), ssh-keygen(1): improve error messages for some common
   PKCS#11 C_Login failure cases; bz#3130

 * ssh(1), sshd(8): make error messages for problems during SSH banner
   exchange consistent with other SSH transport-layer error messages
   and ensure they include the relevant IP addresses bz#3129

 * various: fix a number of spelling errors in comments and debug/error
   messages

 * ssh-keygen(1), ssh-add(1): when downloading FIDO2 resident keys
   from a token, don't prompt for a PIN until the token has told us
   that it needs one. Avoids double-prompting on devices that
   implement on-device authentication.

 * sshd(8), ssh-keygen(1): no-touch-required FIDO certificate option
   should be an extension, not a critical option.

 * ssh(1), ssh-keygen(1), ssh-add(1): offer a better error message
   when trying to use a FIDO key function and SecurityKeyProvider is
   empty.

 * ssh-add(1), ssh-agent(8): ensure that a key lifetime fits within
   the values allowed by the wire format (u32). Prevents integer
   wraparound of the timeout values. bz#3119

 * ssh(1): detect and prevent trivial configuration loops when using
    ProxyJump. bz#3057.

Portability
-----------

 * Detect systems where signals flagged with SA_RESTART will interrupt
   select(2). POSIX permits implementations to choose whether
   select(2) will return when interrupted with a SA_RESTART-flagged
   signal, but OpenSSH requires interrupting behaviour.

 * Several compilation fixes for HP/UX and AIX.

 * On platforms that do not support setting process-wide routing
   domains (all excepting OpenBSD at present), fail to accept a
   configuration attempts to set one at process start time rather than
   fatally erroring at run time. bz#3126

 * Improve detection of egrep (used in regression tests) on platforms
   that offer a poor default one (e.g. Solaris).

 * A number of shell portability fixes for the regression tests.

 * Fix theoretical infinite loop in the glob(3) replacement
   implementation.

 * Fix seccomp sandbox compilation problems for some Linux
   configurations bz#3085

 * Improved detection of libfido2 and some compilation fixes for some
   configurations when --with-security-key-builtin is selected.

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.