> Can somebody who know internals of Al Viro's thinking help here?
There's only one person... :)
> > Please explain why you think it's wrong to be able to bind mount from
> > a different namespace?
>
>
> If It is allowed, the concept of namespaces itself becomes
> nebulous. one could bind mount the root vfsmount of all the other
> namespace in their own namespace and then it all becomes one big tree
> with all the other namespaces as a subtree.
1) you need not recursively bind the whole tree of the private
namespace. In fact you can only do that by hand, since the kernel
won't do it (!recurse || check_mnt(old_nd.mnt) in do_loopback).
2) you won't see changes made in other namespace, they are still
separate, they are just sharing some filesystems, just as after
clone, or just as after propagation within a shared subtree.
3) this is not automatic, so no filesystems in the other namespace are
visible unless there's an explicit action by processes both in the
originating namespace and the destination namespace.
4) in fact, the process in the originating namespace can single out a
mount and just send a file descriptor refering to that mount
(e.g. by binding it to a temporary directory, opening the root,
detaching from the mountpoint, and then sending the file descriptor
to the receiving process). This way the receiving process will see
no other mounts in the originating namespace, and can only bind
from that single mount.
> why would we need this feature? what extra advantage would this feature
> provide us? Is the advantage of this feature already discussed in this
> thread? (maybe i missed it).
http://lkml.org/lkml/2005/4/25/47
It was suggested as a way of sharing mounts between sessions (as
opposed to shared subtrees, which shares mounts between parent/child
process).
I'm not saying that that this is the way to do it. But it does seem
to me a useful feature.
And since it doesn't add _any_ complexity to the kernel, I think it
would be rather stupid to remove it.
> > What do you mean by cross contamination?
>
> A vfsmount in one namespace bound to a mountpoint in another
> namespace.
A vfsmount can only be in a single namespace at a time, since each
mount tree is rooted in a single namespace. So what you are saying is
impossible.
This feature is about binding a mount (copying a vfsmount in
clone_mnt()), which happens to be in a different namespace. After the
vfsmount is cloned, and is attached to the process's native namespace
it's in that namespace solely.
The bug being discussed, is an administrative error, of setting
mnt_namespace to the wrong value, which causes the mount to be
un-removable. So there was never a question of "cross contamination".
Miklos
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]
