Re: Relax the prohibition of usage fchdir(2) to quit a chroot
[email protected] (Christos Zoulas)
| Newsgroups | gmane.os.netbsd.devel.security |
|---|---|
| Message-ID | <[email protected]> |
In article <trinity-8dd4abc6-77a8-4d9d-889b-54aa9678554f-1411294647535@3capp-mailcom-bs16>, Kamil Rytarowski <[email protected]> wrote: >Thanks Marc, > >Well, right the better word is 'add feature' and we would change the >discussion from 'don't touch anything' to 'let's add chroot restrictions >swappable in runtime, that's a cool feature opening opportunities' and >then to mind-storm ideas what and how to restrict. > >In some other way we might end up with floating patches around ;-) > >With kind regards. I think that while I agree with the answer given (adding sysctls like that break the security model) there are alternatives which might help you to achieve your goal: - create a hard link to the file you want to change inside the chroot - give access to the subtree using a loopback mount. - create a named pipe that can be used from a process outside the chroot to exchange data. - if you want to get adventurous you can write a tiny LKM that lets you open a device node that points to the file descriptor you want. christos