Re: [PATCH v1] shebang: restrict python interactive prompt/interpreter

[Mimi Zohar <zohar@linux.vnet.ibm.com>]

On Sun, 2017-06-11 at 13:44 +0200, Mickaël Salaün wrote:
> On 10/06/2017 07:27, Tetsuo Handa wrote:
> > Kees Cook wrote:
> >> On Fri, Jun 9, 2017 at 10:23 AM, Matt Brown <matt@nmatt.com> wrote:
> >>> what does everyone thing about a envp_blacklist option that is a list of
> >>> environmental variables that will be stripped from exec calls. This can
> >>> be done in the LSM hook bprm_check_security.
> >>>
> >>> Is there any reason on a hardened system why you would need the
> >>> PYTHONINSPECT environmental variable?
> >>
> >> As part of shebang, it likely makes sense to whitelist (rather than
> >> blacklist) the env of the restricted interpreters. Though this is
> >> starting to get complex. :P
> > 
> > Blacklisting environment variables is dangerous. I think that
> > administrators can afford whitelisting environment variable names.
> > I think that implementing whitelist of environment variable names
> > as an independent LSM module would be fine.
> > 
> > While it is true that things starts getting complex if we check environment
> > variables, shebang will already become complex if it starts worrying about
> > updating inode number list in order to close the race window between doing
> > creat()+write()+close()+chmod()+rename() by the package manager and teaching
> > the kernel the new inode number determined by creat(). We will need an
> > interface for allowing the package manager to teach the kernel the new inode
> > number and modification of the package manager, for the kernel side is doing
> > inode number based blacklisting while user side can execute it before rename().

I don't think we're trying to protect against executing the
interpreter prior to the rename.  Rename, itself, would trigger
associating the interpreter name with an inode number.

> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-security-module" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> 
> Using filesystem xattr seems like a good idea for this kind of
> exceptions and instead of a hardcoded interpreter path. Something like
> "security.tpe.interpreter=1|2" (bitmask for interpreter-only and/or CLI)
> and "security.tpe.environment=HOME,LOGNAME" would be quite flexible to
> configure a security policy for some binaries. This could also be
> protected by IMA/EVM, if needed.

Checking for the existence of an xattr without caching is relatively
slow.  I'm not sure that we would want to go this route.

> This kind of xattr should be writable by the owner of the file. The TPE
> LSM [1] could then take these xattr into account according to the TPE
> policy.

Security xattrs are only writable by root.

Mimi

> The "security.tpe.environment" could also be set on a script file to be
> part of the union with the interpreter's environment whitelist. This may
> be needed to be able to use environment variables as configuration in a
> script.
> 
> In the future, a "security.tpe.memory" could contain a set of flags as
> PaX uses for mprotect-like exceptions (user.pax.flags).
> 
> Userland daemons such as paxctld or paxrat could be used (with some
> tweaks) to keep a consistent TPE policy over time.
> 
>  Mickaël
> 
> 
> [1] https://lkml.kernel.org/r/1497015878.21594.201.camel@linux.vnet.ibm.com
>