Re: [PATCH v2 1/3] capabilities: Introduce CAP_CHECKPOINT_RESTORE

[Christian Brauner]

On Tue, Jun 09, 2020 at 09:06:27AM -0700, Andrei Vagin wrote:
> On Tue, Jun 09, 2020 at 09:44:22AM +0200, Christian Brauner wrote:
> > On Mon, Jun 08, 2020 at 08:42:21PM -0700, Andrei Vagin wrote:
> > > On Wed, Jun 03, 2020 at 06:23:26PM +0200, Adrian Reber wrote:
> > > > This patch introduces CAP_CHECKPOINT_RESTORE, a new capability facilitating
> > > > checkpoint/restore for non-root users.
> > > > 
> > > > Over the last years, The CRIU (Checkpoint/Restore In Userspace) team has been
> > > > asked numerous times if it is possible to checkpoint/restore a process as
> > > > non-root. The answer usually was: 'almost'.
> > > > 
> > > > The main blocker to restore a process as non-root was to control the PID of the
> > > > restored process. This feature available via the clone3 system call, or via
> > > > /proc/sys/kernel/ns_last_pid is unfortunately guarded by CAP_SYS_ADMIN.
> > > > 
> > > > In the past two years, requests for non-root checkpoint/restore have increased
> > > > due to the following use cases:
> > > > * Checkpoint/Restore in an HPC environment in combination with a resource
> > > >   manager distributing jobs where users are always running as non-root.
> > > >   There is a desire to provide a way to checkpoint and restore long running
> > > >   jobs.
> > > > * Container migration as non-root
> > > > * We have been in contact with JVM developers who are integrating
> > > >   CRIU into a Java VM to decrease the startup time. These checkpoint/restore
> > > >   applications are not meant to be running with CAP_SYS_ADMIN.
> > > > 
> > > ...
> > > > 
> > > > The introduced capability allows to:
> > > > * Control PIDs when the current user is CAP_CHECKPOINT_RESTORE capable
> > > >   for the corresponding PID namespace via ns_last_pid/clone3.
> > > > * Open files in /proc/pid/map_files when the current user is
> > > >   CAP_CHECKPOINT_RESTORE capable in the root namespace, useful for recovering
> > > >   files that are unreachable via the file system such as deleted files, or memfd
> > > >   files.
> > > 
> > > PTRACE_O_SUSPEND_SECCOMP is needed for C/R and it is protected by
> > > CAP_SYS_ADMIN too.
> > 
> > This is currently capable(CAP_SYS_ADMIN) (init_ns capable) why is it
> > safe to allow unprivileged users to suspend security policies? That
> > sounds like a bad idea.
> 
> Why do you think so bad about me;). I don't suggest to remove or

Andrei, nothing could be further from me than to think bad about you!
You've done way too much excellent work. ;)

> downgrade this capability check. The patch allows all c/r related
> operations if the current has CAP_CHECKPOINT_RESTORE.
> 
> So in this case the check:
>      if (!capable(CAP_SYS_ADMIN))
>              return -EPERM;
> 
> will be converted in:
>      if (!capable(CAP_SYS_ADMIN) && !capable(CAP_CHECKPOINT_RESTORE))
>              return -EPERM;

Yeah, I got that but what's the goal here? Isn't it that you want to
make it safe to install the criu binary with the CAP_CHECKPOINT_RESTORE
fscap set so that unprivileged users can restore their own processes
without creating a new user namespace or am I missing something? The
use-cases in the cover-letter make it sound like that's what this is
leading up to:

> > > > * Checkpoint/Restore in an HPC environment in combination with a resource
> > > >   manager distributing jobs where users are always running as non-root.
> > > >   There is a desire to provide a way to checkpoint and restore long running
> > > >   jobs.
> > > > * Container migration as non-root
> > > > * We have been in contact with JVM developers who are integrating
> > > >   CRIU into a Java VM to decrease the startup time. These checkpoint/restore
> > > >   applications are not meant to be running with CAP_SYS_ADMIN.

But maybe I'm just misunderstanding crucial bits (likely (TM)).

Christian