"unix_random.c in lshd for lsh 2.0.1 leaks file descriptors related
to the randomness generator, which allows local users to cause a denial
of service by truncating the seed file, which prevents the server from
starting, or obtain sensitive seed information that could be used to
crack keys."
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-0353
developer is officially maintaining the package.
The rationale for changing this from "tech-pkg" to "pkgsrc-users" is
that it implies that any user can try to maintain the package (by
submitting patches to the mailing list). Since the folks most likely
to care about the package are the folks that want to use it or are
already using it, this would leverage the energy of users who aren't
developers.
security/lsh at 1.4.3.
lsh-2.0.1 has interoperability problems with openssh servers
(always gets "Invalid server signature" errors).
lsh-1.4.3 is not affected by CAN-2003-0826. Add a patch to address
CAN-2005-0814 and bump PKGREVISION.
News for the 2.0.1 release
Fixed denial of service bug in lshd.
Fixed a bug in lsh-make-seed, which could make the program go
into an infinite loop on read errors.
lsh now asks for passwords also in quite (-q) mode, as
described in the manual.
Control character filtering used to sometimes consider newline
as a dangerous control character. Now newlines should be
displayed normally.
Removed support for the non-standard alias
"diffie-hellman-group2-sha1". The standardized name is for
this key exchange method is "diffie-hellman-group14-sha1".
News for the 2.0 release
Several programs have new default behaviour:
* lshd enables X11 forwarding by default (lsh still does not).
* lsh-keygen generates RSA rather than DSA keys by default.
* lsh-writekey encrypts the private key by default, using
aes256-cbc. Unless the --server flag is used.
Improved the lcp script. It is now installed by default.
Implemented the client side of "keyboard-interactive" user
authentication.
Support keyexchange with
diffie-hellman-group14-sha1/diffie-hellman-group2-sha1 (the
standardized name is at the moment not decided).
Fixes to the utf8 encoder, and in particular interactions
between utf8 and control character filtering.
News for the 1.5.5 release
Added SOCKS-style proxying to lsh and lshg. See the new -D
command line option. Supports both SOCKS-4 and SOCKS-5.
The lsh client no longer sets its stdio file descriptors into
non-blocking mode, which should avoid a bunch of problems. As
a consequence, the --cvs-workaround command line option has
been deleted.
In the user lookup code, lshd now ignores the shadow database
if getspnam returns NULL.
In the server pty setup code, use the group "system" as a
fallback if the group "tty" doesn't exist. This is the case on
AIX. (There are however more problems on AIX, which makes it
uncertain that lshd will work out of the box).
Deleted the --ssh1-fallback option for lshd. I hope ssh1 is
dead by now; if it isn't, you have to run ssh1d and lshd on
different ports.
Deleted code for bug-compatibility with ancient versions of
Datafellow's SSH2. There are zero bug-compatibility hacks in
this version.
News for the 1.5.4 release
Added logging of tcpip-forward requests.
Includes nettle-1.9, which have had some portability fixes and
optimizations. In particular, arcfour on x86 should be much
faster.
Implemented flow control on the raw ssh connection. Enforce
limits on the amount of buffered data waiting to be written to
the socket.
Moved all destructive string operations to a separate file
lsh_string.c, which has exclusive rights of accessing string
internals. Should make the code more robust, as buffer size
and index calculations elsewhere in the code should hit an
assert in lsh_string.c before doing damage.
Some general simplification and cleanup of the code.
News for the 1.5.3 release
Fixed heap buffer overrun with potential remote root
compromise. Initial bug report by Bennett Todd.
Fixed a similar bug in the check for channel number allocation
failure in the handling of channel_open, and in the
experimental client SRP code.
lshd now has an experimental mode similar to telnet, where it
accepts the 'none' authentication method and automatically
disables services such as X and TCP forwarding. This can be
useful in environment where it's required that /bin/login or
some other program handle authentication and session setup
(e.g. handle security contexts and so on).
News for the 1.5.2 release
Encrypted private keys works again.
New client escape sequence RET ~ ?, which lists all available
escape sequences. Also fixed the werror functions so that they
use \r\n to terminate lines when writing to a tty in raw mode.
Implemented handling of multiple --interface options to lshd.
As a side effect, The -p option must now be given before
--interface to have any effect.
Connecting to machines with multiple IP-adresses is smarter,
it connects to a few addresses at a time, in parallel.
Fixed a file descriptor leak in the server tcpip forwarding
code.
Lots of portability fixes.
News for the 1.5.1 release
Incompatible change to key format, to comply with the current
spki structure draft. You can use the script lsh-upgrade to
copy and convert the information in the old .lsh/known-hosts
to the new file .lsh/host-acls. The new code uses libspki.
Fixed IPv6 bug reported by Simon Kowallik.
lshd now does the equivalence of ulimit -n unlimited, this is
inherited by processes started upon client requests. If you
don't want this, you should use /etc/{profile,login,whatever}
to set limits for your users. Do note that PAM-based solutions
will NOT work as PAM is used from a separate process that
terminates as soon as the authentication is finished (this of
course goes for environment variables too).
lsh and and lshg now parses options from LSHFLAGS and
LSHGFLAGS, these are parsed before and can be overridden by
the command line.
News for the 1.5 release
Implemented the server side of X11 forwarding. Try lshd
--x11-forward. There's one known bug: The server may start
sending data on the session channel (typically your first
shell prompt) before it has sent the reply to the client's
"shell" or "exec" request. lsh will complain about, and ignore
that data.
As part of the X11 hacking, the socket code have been
reorganized.
Deleted one of the ipv6 configure tests. Now lsh will happily
build ipv6 support even if ipv6 is not available at run-time
on the build machine.
Fixed bug preventing -c none from working.
Another bug fix, call setsid even in the non-pty case.
Various bug fixes.
in the process. (More information on tech-pkg.)
Bump PKGREVISION and BUILDLINK_DEPENDS of all packages using libtool and
installing .la files.
Bump PKGREVISION (only) of all packages depending directly on the above
via a buildlink3 include.
