Re: [PATCH 0/24] kernel: add a netlink interface to get information about processes (v2)

[Andrew Vagin <avagin@virtuozzo.com>]

On Mon, Dec 14, 2015 at 02:38:06PM -0800, Andy Lutomirski wrote:
> On Dec 13, 2015 11:52 PM, "Andrew Vagin" <avagin@virtuozzo.com> wrote:
> >
> > On Thu, Dec 03, 2015 at 03:20:30PM -0800, Andy Lutomirski wrote:
> > > On Tue, Nov 24, 2015 at 7:18 AM, Andrew Vagin <avagin@virtuozzo.com> wrote:
> > > > Hello Everybody,
> > > >
> > > > Sorry for the long delay. I wanted to resurrect this thread.
> > > >
> > > > Andy suggested to create a new syscall instead of using netlink
> > > > interface.
> > > >> Would it make more sense to have a new syscall instead?  You could
> > > >> even still use nlattr formatting for the syscall results.
> > > >
> > > > I tried to implement it to understand how it looks like. Here is my
> > > > version:
> > > > https://github.com/avagin/linux-task-diag/blob/task_diag_syscall/kernel/task_diag.c#L665
> > > > I could not invent a better interfaces for it than using netlink
> > > > messages as arguments. I know it looks weird.
> > > >
> > > > I could not say that I understood why a new system call is better
> > > > than using a netlink socket, so I tried to solve the problem which
> > > > were mentioned for the netlink interface.
> > > >
> > > > The magor question was how to support pid and user namespaces in task_diag.
> > > > I think I found a good and logical solution.
> > > >
> > > > As for pidns, we can use scm credentials, which is connected to each
> > > > socket message. They contain requestor’s pid and we can get a pid
> > > > namespace from it. In this case, we get a good feature to specify a pid
> > > > namespace without entering into it. For that, an user need to specify
> > > > any process from this pidns in an scm message.
> > >
> > > That seems a little messy.  A process can't currently move into
> > > another pidns, but how do you make sure you have any pid at all that
> > > belongs to the reference pidns?  You can, of course, always use your
> > > own pid, but that still seems odd to me.
> >
> > There is your pid by default, you need to do nothing for that.
> > If we look at containers or sandboxes, we ussualy know PID of
> > the init process.
> >
> >
> > >
> > > >
> > > > As for credentials, we can get them from file->f_cred. In this case we
> > > > are able to create a socket and decrease permissions of the current
> > > > process, but the socket will work as before. It’s the common behaviour for
> > > > file descriptors.
> > >
> > > Slightly off-topic, but this netlink is really rather bad as an
> > > example of how fds can be used as capabilities (in the real capability
> > > sense, not the Linux capabilities sense).  You call socket and get a
> > > socket.  That socket captures f_cred.  Then you drop privs, and you
> > > assume that the socket you're holding on to retains the right to do
> > > certain things.
> > >
> > > This breaks pretty badly when, through things such as this patch set,
> > > existing code that creates netlink sockets suddenly starts capturing
> > > brand-new rights that didn't exist as part of a netlink socket before.
> >
> > Sorry, I don't understand this part. Could you eloborate? Maybe give an
> > example.
> >
> > I always think that it's a feature, that we can create a descriptor and
> > drop capabilities of the process or send this descriptor to an
> > unprivilieged process.
> 
> Suppose there's an existing program that likes this feature.  It
> creates a netlink socket, optionally calls connect(2), and then drop
> privileges.  It expects to retain some subset of its privileges.
> 
> The problem is that by increasing the power of a netlink socket
> created with higher-than-current privilege, you've just increased the
> privilege retained by the old app.  In this particular case, it's
> especially odd because it retains privilege over the old pidns,
> whereas the old program (in theory -- probably no one does this) could
> have created a netlink socket, unshared pidns, and forked, and it
> would have expected to retain no privilege over the old pidns.

Thank you for the explanation.  If I understand you correctly, the
problem is that we can use an arbitrary netlink socket to use task_diag.

It can be a reason to not use netlink interface for task diag.

What do you think about the idea to add a a transaction file in
procfs? We will open it, send a request and get required information.

I want to have a file descriptor to transfer data between kernel and
userspace, because a size of response can be too big to receive it for
one call. If we use a file descriptor, we can divide a response into
parts.

Thanks,
Andrew

> --Andy