cgroup: Use CAP_SYS_RESOURCE to allow a process to migrate other tasks between cgroups

This patch adds logic to allows a process to migrate other tasks
between cgroups if they have CAP_SYS_RESOURCE.

In Android (where this feature originated), the ActivityManager tracks
various application states (TOP_APP, FOREGROUND, BACKGROUND, SYSTEM,
etc), and then as applications change states, the SchedPolicy logic
will migrate the application tasks between different cgroups used
to control the different application states (for example, there is a
background cpuset cgroup which can limit background tasks to stay
on one low-power cpu, and the bg_non_interactive cpuctrl cgroup can
then further limit those background tasks to a small percentage of
that one cpu's cpu time).

However, for security reasons, Android doesn't want to make the
system_server (the process that runs the ActivityManager and
SchedPolicy logic), run as root. So in the Android common.git
kernel, they have some logic to allow cgroups to loosen their
permissions so CAP_SYS_NICE tasks can migrate other tasks between
cgroups.

I feel the approach taken there overloads CAP_SYS_NICE a bit much,
and is maybe more complicated then needed.

So this patch, as suggested by Michael Kerrisk, simply adds a
check for CAP_SYS_RESOURCE.

I've tested this with AOSP master, and this seems to work well
as Zygote and system_server already use CAP_SYS_RESOURCE. I've
also submitted patches against the android-4.4 kernel to change
it to use CAP_SYS_RESOURCE, and the Android developers have
seemed ok with this change.

Thouhts and feedback would be appreciated!

Cc: Tejun Heo <redacted>
Cc: Li Zefan <redacted>
Cc: Jonathan Corbet <redacted>
Cc: <redacted>
Cc: Android Kernel Team <redacted>
Cc: Rom Lemarchand <redacted>

17 more lines

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 09f84d2..866059a 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c

@@ -2857,7 +2857,7 @@ static int cgroup_procs_write_permission(struct task_struct *task,
 	if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
 	    !uid_eq(cred->euid, tcred->uid) &&
 	    !uid_eq(cred->euid, tcred->suid) &&
-	    !ns_capable(tcred->user_ns, CAP_CGROUP_MIGRATE))
+	    !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
 		ret = -EACCES;
 
 	if (!ret && cgroup_on_dfl(dst_cgrp)) {

-- 
2.7.4

Hi John,

On 10/21/2016 03:24 AM, John Stultz wrote:

This patch adds logic to allows a process to migrate other tasks
between cgroups if they have CAP_SYS_RESOURCE.

This appears to be a patch against your previous patch, 
rather than against mainline. Was that intended?

Cheers,

Michael

In Android (where this feature originated), the ActivityManager tracks
various application states (TOP_APP, FOREGROUND, BACKGROUND, SYSTEM,
etc), and then as applications change states, the SchedPolicy logic
will migrate the application tasks between different cgroups used
to control the different application states (for example, there is a
background cpuset cgroup which can limit background tasks to stay
on one low-power cpu, and the bg_non_interactive cpuctrl cgroup can
then further limit those background tasks to a small percentage of
that one cpu's cpu time).

However, for security reasons, Android doesn't want to make the
system_server (the process that runs the ActivityManager and
SchedPolicy logic), run as root. So in the Android common.git
kernel, they have some logic to allow cgroups to loosen their
permissions so CAP_SYS_NICE tasks can migrate other tasks between
cgroups.

I feel the approach taken there overloads CAP_SYS_NICE a bit much,
and is maybe more complicated then needed.

So this patch, as suggested by Michael Kerrisk, simply adds a
check for CAP_SYS_RESOURCE.

I've tested this with AOSP master, and this seems to work well
as Zygote and system_server already use CAP_SYS_RESOURCE. I've
also submitted patches against the android-4.4 kernel to change
it to use CAP_SYS_RESOURCE, and the Android developers have
seemed ok with this change.

Thouhts and feedback would be appreciated!

Cc: Tejun Heo <redacted>
Cc: Li Zefan <redacted>
Cc: Jonathan Corbet <redacted>
Cc: <redacted>
Cc: Android Kernel Team <redacted>
Cc: Rom Lemarchand <redacted>
Cc: Colin Cross <redacted>
Cc: Dmitry Shmidt <redacted>
Cc: Todd Kjos <redacted>

28 more quoted lines

-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/

Quoting John Stultz (john.stultz-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org):

This patch adds logic to allows a process to migrate other tasks
between cgroups if they have CAP_SYS_RESOURCE.

Hi,

fwiw this seems to me a reasonable choice, that is, giving a task
CAP_SYS_RESOURCE to be allow it to write to the procs file doesn't
seem like giving it too much.  +1 from me.  So long as the Android
folks don't have a reason why this won't work for them,

Acked-by: Serge Hallyn <redacted>

-serge

In Android (where this feature originated), the ActivityManager tracks
various application states (TOP_APP, FOREGROUND, BACKGROUND, SYSTEM,
etc), and then as applications change states, the SchedPolicy logic
will migrate the application tasks between different cgroups used
to control the different application states (for example, there is a
background cpuset cgroup which can limit background tasks to stay
on one low-power cpu, and the bg_non_interactive cpuctrl cgroup can
then further limit those background tasks to a small percentage of
that one cpu's cpu time).

However, for security reasons, Android doesn't want to make the
system_server (the process that runs the ActivityManager and
SchedPolicy logic), run as root. So in the Android common.git
kernel, they have some logic to allow cgroups to loosen their
permissions so CAP_SYS_NICE tasks can migrate other tasks between
cgroups.

I feel the approach taken there overloads CAP_SYS_NICE a bit much,
and is maybe more complicated then needed.

So this patch, as suggested by Michael Kerrisk, simply adds a
check for CAP_SYS_RESOURCE.

I've tested this with AOSP master, and this seems to work well
as Zygote and system_server already use CAP_SYS_RESOURCE. I've
also submitted patches against the android-4.4 kernel to change
it to use CAP_SYS_RESOURCE, and the Android developers have
seemed ok with this change.

Thouhts and feedback would be appreciated!

Cc: Tejun Heo <redacted>
Cc: Li Zefan <redacted>
Cc: Jonathan Corbet <redacted>
Cc: <redacted>
Cc: Android Kernel Team <redacted>
Cc: Rom Lemarchand <redacted>
Cc: Colin Cross <redacted>
Cc: Dmitry Shmidt <redacted>
Cc: Todd Kjos <redacted>

29 more quoted lines

On Thu, Oct 20, 2016 at 11:42 PM, Michael Kerrisk (man-pages)
[off-list ref] wrote:

Hi John,

On 10/21/2016 03:24 AM, John Stultz wrote:

quoted

This patch adds logic to allows a process to migrate other tasks
between cgroups if they have CAP_SYS_RESOURCE.

This appears to be a patch against your previous patch,
rather than against mainline. Was that intended?

Heh. Definitely not. Apologies. I thought I had folded things down
(and the real patch is a similar looking one-liner).

I'll re-send the proper patch on monday. :)

thanks
-john