Re: Non-root chroot
- Reply: Jason Bacon : "Re: Non-root chroot"
- In reply to: Jason Bacon : "Re: Non-root chroot"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Fri, 08 Aug 2025 00:43:03 UTC
Jason Bacon <bacon4000@gmail.com> wrote: > On 8/5/25 22:59, Jamie Landeg-Jones wrote: > > But just as it is if you're not using a chroot, your non-root user cannot > > create suid-root binaries, and when you're setting things up, you'd simply not > > use root to copy a suid-root 'su' (or anything else) into your chroot tree. > > Not quite true: > > FreeBSD moray.acadix bacon ~ 1018: sysctl vfs.usermount > vfs.usermount: 1 > FreeBSD moray.acadix bacon ~ 1019: whoami > bacon > FreeBSD moray.acadix bacon ~ 1020: mount -t nullfs /usr/bin > /home/bacon/new-root/usr/bin/ > FreeBSD moray.acadix bacon ~ 1021: ls -l new-root/usr/bin/su > -r-sr-xr-x 1 root wheel 17K Jul 28 17:11 new-root/usr/bin/su* > > Granted, usermount should generally not be set on a shared system, but > it could happen. But usermounts are mounted nosuid. > Or, as I mentioned in a prior response, there's always the possibility > of duping a root user into installing an suid binary. Not all of them > are highly qualified, and everyone gets tired or rushed sometimes. Yeah, possible, especially if they are recursively copying a system tree (/bin /usr/bin etc.) to the chroot, but the point I was replying to said "Important point is that the user is not obliged to hand in any particular "su" program. The user may hand in any "su"-like code suitable for escaping the chroot.", so that's not very likely, though I concede that this is an attack vector greater than not allowing user-chroot. Would forcing all user-chrooted trees to be no-suid-root solve this, do you think? Cheers, Jamie