[PATCH v8.75 0/4] cgroups: implement moving a threadgroup's threads atomically with cgroup.procs

[Ben Blum <bblum@andrew.cmu.edu>]

On Mon, Feb 07, 2011 at 08:35:42PM -0500, Ben Blum wrote:
> On Sun, Dec 26, 2010 at 07:09:19AM -0500, Ben Blum wrote:
> > On Fri, Dec 24, 2010 at 03:22:26AM -0500, Ben Blum wrote:
> > > On Wed, Aug 11, 2010 at 01:46:04AM -0400, Ben Blum wrote:
> > > > On Fri, Jul 30, 2010 at 07:56:49PM -0400, Ben Blum wrote:
> > > > > This patch series is a revision of http://lkml.org/lkml/2010/6/25/11 .
> > > > > 
> > > > > This patch series implements a write function for the 'cgroup.procs'
> > > > > per-cgroup file, which enables atomic movement of multithreaded
> > > > > applications between cgroups. Writing the thread-ID of any thread in a
> > > > > threadgroup to a cgroup's procs file causes all threads in the group to
> > > > > be moved to that cgroup safely with respect to threads forking/exiting.
> > > > > (Possible usage scenario: If running a multithreaded build system that
> > > > > sucks up system resources, this lets you restrict it all at once into a
> > > > > new cgroup to keep it under control.)
> > > > > 
> > > > > Example: Suppose pid 31337 clones new threads 31338 and 31339.
> > > > > 
> > > > > # cat /dev/cgroup/tasks
> > > > > ...
> > > > > 31337
> > > > > 31338
> > > > > 31339
> > > > > # mkdir /dev/cgroup/foo
> > > > > # echo 31337 > /dev/cgroup/foo/cgroup.procs
> > > > > # cat /dev/cgroup/foo/tasks
> > > > > 31337
> > > > > 31338
> > > > > 31339
> > > > > 
> > > > > A new lock, called threadgroup_fork_lock and living in signal_struct, is
> > > > > introduced to ensure atomicity when moving threads between cgroups. It's
> > > > > taken for writing during the operation, and taking for reading in fork()
> > > > > around the calls to cgroup_fork() and cgroup_post_fork().
> > 
> > Well this time everything here is actually safe and correct, as far as
> > my best efforts and keen eyes can tell. I dropped the per_thread call
> > from the last series in favour of revising the subsystem callback
> > interface. It now looks like this:
> > 
> > ss->can_attach()
> >  - Thread-independent, possibly expensive/sleeping.
> > 
> > ss->can_attach_task()
> >  - Called per-thread, run with rcu_read so must not sleep.
> > 
> > ss->pre_attach()
> >  - Thread independent, must be atomic, happens before attach_task.
> > 
> > ss->attach_task()
> >  - Called per-thread, run with tasklist_lock so must not sleep.
> > 
> > ss->attach()
> >  - Thread independent, possibly expensive/sleeping, called last.
> 
> Okay, so.
> 
> I've revamped the cgroup_attach_proc implementation a bunch and this
> version should be a lot easier on the eyes (and brains). Issues that are
> addressed:
> 
> 1) cgroup_attach_proc now iterates over leader->thread_group once, at
>    the very beginning, and puts each task_struct that we want to move
>    into an array, using get_task_struct to make sure they stick around.
>     - threadgroup_fork_lock ensures no threads not in the array can
>       appear, and allows us to use signal->nr_threads to determine the
>       size of the array when kmallocing it.
>     - This simplifies the rest of the function a bunch, since now we
>       never need to do rcu_read_lock after building the array. All the
>       subsystem callbacks are the same as described just above, but the
>       "can't sleep" restriction is gone, so it's nice and clean.
>     - Checking for a race with de_thread (the manoeuvre I refer to as
>       "double-double-toil-and-trouble-check locking") now needs to be
>       done only once, at the beginning (before building the array).
> 
> 2) The nodemask allocation problem in cpuset is fixed the same way as
>    before - the masks are shared between the three attach callbacks, so
>    are made as static global variables.
> 
> 3) The introduction of threadgroup_fork_lock in sched.h (specifically,
>    in signal_struct) requires rwsem.h; the new include appears in the
>    first patch. (An alternate plan would be to make it a struct pointer
>    with an incomplete forward declaration and kmalloc/kfree it during
>    housekeeping, but adding an include seems better than that particular
>    complication.) In light of this, the definitions for
>    threadgroup_fork_{read,write}_{un,}lock are also in sched.h.

Same as before; using flex_array in attach_proc (thanks Kame).

-- Ben

---
 Documentation/cgroups/cgroups.txt |   39 ++-
 block/blk-cgroup.c                |   18 -
 include/linux/cgroup.h            |   10 
 include/linux/init_task.h         |    9 
 include/linux/sched.h             |   36 ++
 kernel/cgroup.c                   |  489 +++++++++++++++++++++++++++++++++-----
 kernel/cgroup_freezer.c           |   26 --
 kernel/cpuset.c                   |   96 +++----
 kernel/fork.c                     |   10 
 kernel/sched.c                    |   38 --
 mm/memcontrol.c                   |   18 -
 security/device_cgroup.c          |    3 
 12 files changed, 594 insertions(+), 198 deletions(-)