Bug 1949 - PermitOpen none option
Summary: PermitOpen none option
Status: CLOSED FIXED
Alias: None
Product: Portable OpenSSH
Classification: Unclassified
Component: sshd (show other bugs)
Version: 5.9p1
Hardware: All OpenBSD
: P2 enhancement
Assignee: Assigned to nobody
URL:
Keywords:
Depends on:
Blocks: V_6_1
  Show dependency treegraph
 
Reported: 2011-11-06 19:51 AEDT by Loganaden Velvindron
Modified: 2016-08-02 10:41 AEST (History)
1 user (show)

See Also:


Attachments
permitopen_none option diff (2.01 KB, application/octet-stream)
2011-11-06 19:51 AEDT, Loganaden Velvindron
no flags Details
permitOpen none with a single socket (1.98 KB, patch)
2011-11-20 03:32 AEDT, Loganaden Velvindron
no flags Details | Diff
permitopen none with sshd -T support (2.40 KB, patch)
2011-12-02 11:59 AEDT, Darren Tucker
djm: ok+
Details | Diff
OpenBSD sshd permitopen diff (2.50 KB, patch)
2011-12-02 19:32 AEDT, Loganaden Velvindron
no flags Details | Diff
PermitOpen None diff for native OpenSSH (2.50 KB, patch)
2011-12-08 06:03 AEDT, Loganaden Velvindron
no flags Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Loganaden Velvindron 2011-11-06 19:51:23 AEDT
Created attachment 2104 [details]
permitopen_none option diff

From debian bug tracker:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=543683

Package: openssh-server
Version: 1:5.1p1-7
Severity: wishlist

I'm trying to setup a reverse SSH box (i.e. one where people stuck
behind NAT can SSH in and initiate a tunnel back to their machine).
They use this something like this:

  ssh login@box -R 2000:localhost:22

I'm trying to lock this down as far as possible - in particular I'd
like to disable AllowTcpForwarding, however if I do this it prevents
both local _and_ remote tunnels.

Leaving AllowTcpForwarding open and setting "PermitOpen
127.0.0.1:65535" gets close - all the reverse tunnels work, but the
only local tunnel that will work is "ssh login@box -L
xxxx:localhost:65535".   

I'd like to use "PermitOpen none" (or just blank) however sshd doesn't
allow this (just checked the source code).

Thanks,

Adrian
-- 
Email: adrian@smop.co.uk  -*-  GPG key available on public key servers
Debian GNU/Linux - the maintainable distribution   -*-  www.debian.org

I thought I'd give it a try.

I added a new function that populates list of allowed sockets
with NULL, and also added the permitopen none option.

Any feedback on how to improve the code would be nice :-)

//Logan
C-x-C-c
Comment 1 Loganaden Velvindron 2011-11-20 03:32:20 AEDT
Created attachment 2108 [details]
permitOpen none with a single socket
Comment 2 Loganaden Velvindron 2011-11-20 04:55:06 AEDT
Instead of creating a bunch of sockets with hosttoconnect to as NULL,
It's simpler to create only one.
Comment 3 Damien Miller 2011-12-02 10:59:23 AEDT
Comment on attachment 2108 [details]
permitOpen none with a single socket

Darren is more familiar with this code than I am. The patch looks sane to me though.
Comment 4 Darren Tucker 2011-12-02 11:21:08 AEDT
Looks OK, but I think we need to add the equivalent code to channel_print_adm_permitted_opens() so that it'll output "permitopen none" when it sees the NULL in host_to_connect.
Comment 5 Darren Tucker 2011-12-02 11:59:16 AEDT
Created attachment 2111 [details]
permitopen none with sshd -T support
Comment 6 Loganaden Velvindron 2011-12-02 19:32:41 AEDT
Created attachment 2112 [details]
OpenBSD sshd permitopen diff

Port of dtucker's patch for openbsd
Comment 7 Loganaden Velvindron 2011-12-08 06:03:12 AEDT
Created attachment 2116 [details]
PermitOpen None diff for native OpenSSH

Remove a whitespace in channel_disable_adm_local_opens(void)

& add a space before none in printf() to make it more consistent.

Any comments ?
Comment 8 Loganaden Velvindron 2012-01-05 18:30:30 AEDT
Are there other issues that need fixing ?
Comment 9 Loganaden Velvindron 2012-01-20 05:34:24 AEDT
ping ?
Comment 10 Loganaden Velvindron 2012-02-14 20:14:20 AEDT
Now that the tree is unlocked, any chance this could make it to OpenSSH 6.1 ?

Patching each machine is a pain...
Comment 11 Darren Tucker 2012-03-30 10:55:34 AEDT
thanks for the patch (and patience).  this has been committed and will be in the 6.1 release.
Comment 12 Loganaden Velvindron 2012-04-01 01:50:13 AEDT
Awesome :-)
Thanks for finding time to looking at it !
Comment 13 Damien Miller 2016-08-02 10:41:46 AEST
Close all resolved bugs after 7.3p1 release