svn commit: r224462 - stable/8/usr.sbin/jail

Ben Kaduk minimarmot at gmail.com
Thu Jul 28 03:08:34 UTC 2011


On Wed, Jul 27, 2011 at 10:19 PM, Jason Hellenthal <jhell at dataix.net> wrote:
>
>
> On Wed, Jul 27, 2011 at 01:56:52AM +0000, Glen Barber wrote:
>> Author: gjb (doc committer)
>> Date: Wed Jul 27 01:56:52 2011
>> New Revision: 224462
>> URL: http://svn.freebsd.org/changeset/base/224462
>>
>> Log:
>>   MFC 224286:
>>
>>   Document the potential for jail escape.
>>
>>   PR:         142341
>>
>> Modified:
>>   stable/8/usr.sbin/jail/jail.8
>> Directory Properties:
>>   stable/8/usr.sbin/jail/   (props changed)
>>
>> Modified: stable/8/usr.sbin/jail/jail.8
>> ==============================================================================
>> --- stable/8/usr.sbin/jail/jail.8     Tue Jul 26 20:51:58 2011        (r224461)
>> +++ stable/8/usr.sbin/jail/jail.8     Wed Jul 27 01:56:52 2011        (r224462)
>> @@ -34,7 +34,7 @@
>>  .\"
>>  .\" $FreeBSD$
>>  .\"
>> -.Dd January 17, 2010
>> +.Dd July 23, 2011
>>  .Dt JAIL 8
>>  .Os
>>  .Sh NAME
>> @@ -913,3 +913,10 @@ Currently, the simplest answer is to min
>>  offered on the host, possibly limiting it to services offered from
>>  .Xr inetd 8
>>  which is easily configurable.
>> +.Sh NOTES
>> +Great care should be taken when managing directories visible within the jail.
>> +For example, if a jailed process has its current working directory set to a
>> +directory that is moved out of the jail's chroot, then the process may gain
>> +access to the file space outside of the jail.
>> +It is recommended that directories always be copied, rather than moved, out
>> +of a jail.
>
> How is either one of these different ?
>
> All mv(1) is doing is a cp(1) & rm(1). In either case the filehandle is

This is not always true when the source and destination live on the
same filesystem.  See rename(2).
Via VOP_RENAME, individual filesystems can override this behavior if
needed (e.g. for AFS where permissions are per-directory, so a
cross-directory copy would return EXDEV).

-Ben Kaduk

> still broken and a process is not going to just get up and move with it.
> On the other side though if you copied a pipe or socket or something
> similiar for example into a jail then it might make whatever is outside
> available to the jailed environment.
>
> Is there something I am misunderstanding about this ? has the way cp(1),
> rm(1) & mv(1) been changed recently ? or is this wording a little off ?
>


More information about the svn-src-all mailing list