Re: [RFC PATCH 00/14] cgroup: Implement cgroup v2 thread mode & CPU controller

[Waiman Long]

On 04/21/2017 10:03 AM, Waiman Long wrote:
> This patchset incorporates the following 2 patchsets from Tejun Heo:
>
>  1) cgroup v2 thread mode patchset (5 patches)
>     https://lkml.org/lkml/2017/2/2/592
>  2) CPU Controller on Control Group v2 (2 patches)
>     https://lkml.org/lkml/2016/8/5/368
>
> Additional patches are then layered on top to implement the following
> new features:
>
>  1) An enhanced v2 thread mode where a thread root (root of a threaded
>     subtree) can have non-threaded children with non-threaded
>     controllers enabled and no internal process constraint does
>     not apply.
>  2) An enhanced debug controller which dumps out more information
>     relevant to the debugging and testing of cgroup v2 in general.
>  3) Separate control knobs for resource domain controllers can be
>     enabled in a thread root to manage all the internal processes in
>     the threaded subtree.
>
> Patches 1-5 are Tejun's cgroup v2 thread mode patchset.
>
> Patch 6 fixes a task_struct reference counting bug introduced in
> patch 1.
>
> Patch 7 moves the debug cgroup out from cgroup_v1.c into its own
> file.
>
> Patch 8 keeps more accurate counts of the number of tasks associated
> with each css_set.
>
> Patch 9 enhances the debug controller to provide more information
> relevant to the cgroup v2 thread mode to ease debugging effort.
>
> Patch 10 implements the enhanced cgroup v2 thread mode with the
> following enhancements:
>
>  1) Thread roots are treated differently from threaded cgroups.
>  2) Thread root can now have non-threaded controllers enabled as well
>     as non-threaded children.
>
> Patches 11-12 are Tejun's CPU controller on control group v2 patchset.
>
> Patch 13 makes both cpu and cpuacct controllers threaded.
>
> Patch 14 enables the creation of a special "cgroup.self" directory
> under the thread root to hold resource control knobs for controllers
> that don't want resource competiton between internal processes and
> non-threaded child cgroups.
>
> Preliminary testing was done with the debug controller enabled. Things
> seemed to work fine so far. More rigorous testing will be needed, and
> any suggestions are welcome.
>
> Tejun Heo (7):
>   cgroup: reorganize cgroup.procs / task write path
>   cgroup: add @flags to css_task_iter_start() and implement
>     CSS_TASK_ITER_PROCS
>   cgroup: introduce cgroup->proc_cgrp and threaded css_set handling
>   cgroup: implement CSS_TASK_ITER_THREADED
>   cgroup: implement cgroup v2 thread support
>   sched: Misc preps for cgroup unified hierarchy interface
>   sched: Implement interface for cgroup unified hierarchy
>
> Waiman Long (7):
>   cgroup: Fix reference counting bug in cgroup_procs_write()
>   cgroup: Move debug cgroup to its own file
>   cgroup: Keep accurate count of tasks in each css_set
>   cgroup: Make debug cgroup support v2 and thread mode
>   cgroup: Implement new thread mode semantics
>   sched: Make cpu/cpuacct threaded controllers
>   cgroup: Enable separate control knobs for thread root internal
>     processes
>
>  Documentation/cgroup-v2.txt     | 114 +++++-
>  include/linux/cgroup-defs.h     |  56 +++
>  include/linux/cgroup.h          |  12 +-
>  kernel/cgroup/Makefile          |   1 +
>  kernel/cgroup/cgroup-internal.h |  18 +-
>  kernel/cgroup/cgroup-v1.c       | 217 +++-------
>  kernel/cgroup/cgroup.c          | 862 ++++++++++++++++++++++++++++++++++------
>  kernel/cgroup/cpuset.c          |   6 +-
>  kernel/cgroup/debug.c           | 284 +++++++++++++
>  kernel/cgroup/freezer.c         |   6 +-
>  kernel/cgroup/pids.c            |   1 +
>  kernel/events/core.c            |   1 +
>  kernel/sched/core.c             | 150 ++++++-
>  kernel/sched/cpuacct.c          |  55 ++-
>  kernel/sched/cpuacct.h          |   5 +
>  mm/memcontrol.c                 |   3 +-
>  net/core/netclassid_cgroup.c    |   2 +-
>  17 files changed, 1478 insertions(+), 315 deletions(-)
>  create mode 100644 kernel/cgroup/debug.c
>
Does anyone has time to take a look at these patches?

As the merge window is going to open up next week, I am not going to
bother you guys when the merge window opens.

Cheers,
Longman