[CFT][PATCH 0/10] Making new mounts of proc and sysfs as safe as bind mounts

Eric W. Biederman ebiederm at xmission.com
Fri May 15 06:55:38 UTC 2015


Andy Lutomirski <luto at amacapital.net> writes:

> Can we please just get rid of this implicit nodev thing once and for all?  If it
> breaks some really weird /proc use case, then I think the right fix is to
> stop enforcing the nodev lock for the proc fully visible check.  After
> all, /proc doesn't contain useful device nodes anyway.

On second look I don't think that will actually cause issues in this
case.

I actually have a fix for the implicit nodev weirdness in my development
qeueue but it requires figuring out how to add s_user_ns to superblocks.
My last round of testing told me I was doing that wrong.

But if the implicit nodev is actually a problem I will definitely delay
this until I have that change ready to go as well.

> Other than that, the code here looks okay to me on brief inspection.

At a practical level I am concerned that enforcing things like noexec
and nosuid from the original normal global proc might cause problems for
things like sandstorm, lxc, and possibly libvirt-lxc.  So I would really
appreciate if people associated with those projects could test this and
tell me if I break things.

Other than my stupid refactor in my code for /proc/fs/nfsd that causes
the kernel to oops :(  Doh!

Eric



More information about the Containers mailing list