Tejun reported the following race between fork() and cgroup.kill at [1].

Tejun:
  I was looking at cgroup.kill implementation and wondering whether there
  could be a race window. So, __cgroup_kill() does the following:

   k1. Set CGRP_KILL.
   k2. Iterate tasks and deliver SIGKILL.
   k3. Clear CGRP_KILL.

  The copy_process() does the following:

   c1. Copy a bunch of stuff.
   c2. Grab siglock.
   c3. Check fatal_signal_pending().
   c4. Commit to forking.
   c5. Release siglock.
   c6. Call cgroup_post_fork() which puts the task on the css_set and tests
       CGRP_KILL.

  The intention seems to be that either a forking task gets SIGKILL and
  terminates on c3 or it sees CGRP_KILL on c6 and kills the child. However, I
  don't see what guarantees that k3 can't happen before c6. ie. After a
  forking task passes c5, k2 can take place and then before the forking task
  reaches c6, k3 can happen. Then, nobody would send SIGKILL to the child.
  What am I missing?

This is indeed a race. One way to fix this race is by taking
cgroup_threadgroup_rwsem in write mode in __cgroup_kill() as the fork()
side takes cgroup_threadgroup_rwsem in read mode from cgroup_can_fork()
to cgroup_post_fork(). However that would be heavy handed as this adds
one more potential stall scenario for cgroup.kill which is usually
called under extreme situation like memory pressure.

To fix this race, let's maintain a sequence number per cgroup which gets
incremented on __cgroup_kill() call. On the fork() side, the
cgroup_can_fork() will cache the sequence number locally and recheck it
against the cgroup's sequence number at cgroup_post_fork() site. If the
sequence numbers mismatch, it means __cgroup_kill() can been called and
we should send SIGKILL to the newly created task.

Reported-by: Tejun Heo <tj@kernel.org>
Closes: https://lore.kernel.org/all/Z5QHE2Qn-QZ6M-KW@slm.duckdns.org/ [1]
Fixes: 661ee6280931 ("cgroup: introduce cgroup.kill")
Cc: stable@vger.kernel.org # v5.14+
Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
Reviewed-by: Michal Koutný <mkoutny@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>

During scx_ops_enable(), SCX needs to invoke the sleepable ops.init_task()
on every task. To do this, it does get_task_struct() on each iterated task,
drop the lock and then call ops.init_task().

However, a TASK_DEAD task may already have lost all its usage count and be
waiting for RCU grace period to be freed. If get_task_struct() is called on
such task, use-after-free can happen. To avoid such situations,
scx_ops_enable() skips initialization of TASK_DEAD tasks, which seems safe
as they are never going to be scheduled again.

Unfortunately, a racing sched_setscheduler(2) can grab the task before the
task is unhashed and then continue to e.g. move the task from RT to SCX
after TASK_DEAD is set and ops_enable skipped the task. As the task hasn't
gone through scx_ops_init_task(), scx_ops_enable_task() called from
switching_to_scx() triggers the following warning:

  sched_ext: Invalid task state transition 0 -> 3 for stress-ng-race-[2872]
  WARNING: CPU: 6 PID: 2367 at kernel/sched/ext.c:3327 scx_ops_enable_task+0x18f/0x1f0
  ...
  RIP: 0010:scx_ops_enable_task+0x18f/0x1f0
  ...
   switching_to_scx+0x13/0xa0
   __sched_setscheduler+0x84e/0xa50
   do_sched_setscheduler+0x104/0x1c0
   __x64_sys_sched_setscheduler+0x18/0x30
   do_syscall_64+0x7b/0x140
   entry_SYSCALL_64_after_hwframe+0x76/0x7e

As in the ops_disable path, it just doesn't seem like a good idea to leave
any task in an inconsistent state, even when the task is dead. The root
cause is ops_enable not being able to tell reliably whether a task is truly
dead (no one else is looking at it and it's about to be freed) and was
testing TASK_DEAD instead. Fix it by testing the task's usage count
directly.

- ops_init no longer ignores TASK_DEAD tasks. As now all users iterate all
  tasks, @include_dead is removed from scx_task_iter_next_locked() along
  with dead task filtering.

- tryget_task_struct() is added. Tasks are skipped iff tryget_task_struct()
  fails.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: David Vernet <void@manifault.com>
Cc: Peter Zijlstra <peterz@infradead.org>

A new BPF extensible sched_class will need more control over the forking
process. It wants to be able to fail from sched_cgroup_fork() after the new
task's sched_task_group is initialized so that the loaded BPF program can
prepare the task with its cgroup association is established and reject fork
if e.g. allocation fails.

Allow sched_cgroup_fork() to fail by making it return int instead of void
and adding sched_cancel_fork() to undo sched_fork() in the error path.

sched_cgroup_fork() doesn't fail yet and this patch shouldn't cause any
behavior changes.

v2: Patch description updated to detail the expected use.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>

Pull header cleanups from Kent Overstreet:
 "The goal is to get sched.h down to a type only header, so the main
  thing happening in this patchset is splitting out various _types.h
  headers and dependency fixups, as well as moving some things out of
  sched.h to better locations.

  This is prep work for the memory allocation profiling patchset which
  adds new sched.h interdepencencies"

* tag 'header_cleanup-2024-01-10' of https://evilpiepirate.org/git/bcachefs: (51 commits)
  Kill sched.h dependency on rcupdate.h
  kill unnecessary thread_info.h include
  Kill unnecessary kernel.h include
  preempt.h: Kill dependency on list.h
  rseq: Split out rseq.h from sched.h
  LoongArch: signal.c: add header file to fix build error
  restart_block: Trim includes
  lockdep: move held_lock to lockdep_types.h
  sem: Split out sem_types.h
  uidgid: Split out uidgid_types.h
  seccomp: Split out seccomp_types.h
  refcount: Split out refcount_types.h
  uapi/linux/resource.h: fix include
  x86/signal: kill dependency on time.h
  syscall_user_dispatch.h: split out *_types.h
  mm_types_task.h: Trim dependencies
  Split out irqflags_types.h
  ipc: Kill bogus dependency on spinlock.h
  shm: Slim down dependencies
  workqueue: Split out workqueue_types.h
  ...

by moving cond_resched_rcu() to rcupdate_wait.h, we can kill another big
sched.h dependency.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

More trimming of sched.h dependencies.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Created as testing for the conditional guard infrastructure.
Specifically this makes use of the following form:

  scoped_cond_guard (mutex_intr, return -ERESTARTNOINTR,
		     &task->signal->cred_guard_mutex) {
    ...
  }
  ...
  return 0;

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Link: https://lkml.kernel.org/r/20231102110706.568467727%40infradead.org

In put_task_struct(), a spin_lock is indirectly acquired under the kernel
stock. When running the kernel in real-time (RT) configuration, the
operation is dispatched to a preemptible context call to ensure
guaranteed preemption. However, if PROVE_RAW_LOCK_NESTING is enabled
and __put_task_struct() is called while holding a raw_spinlock, lockdep
incorrectly reports an "Invalid lock context" in the stock kernel.

This false splat occurs because lockdep is unaware of the different
route taken under RT. To address this issue, override the inner wait
type to prevent the false lockdep splat.

Suggested-by: Oleg Nesterov <oleg@redhat.com>
Suggested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20230614122323.37957-3-wander@redhat.com

Under PREEMPT_RT, __put_task_struct() indirectly acquires sleeping
locks. Therefore, it can't be called from an non-preemptible context.

One practical example is splat inside inactive_task_timer(), which is
called in a interrupt context:

  CPU: 1 PID: 2848 Comm: life Kdump: loaded Tainted: G W ---------
   Hardware name: HP ProLiant DL388p Gen8, BIOS P70 07/15/2012
   Call Trace:
   dump_stack_lvl+0x57/0x7d
   mark_lock_irq.cold+0x33/0xba
   mark_lock+0x1e7/0x400
   mark_usage+0x11d/0x140
   __lock_acquire+0x30d/0x930
   lock_acquire.part.0+0x9c/0x210
   rt_spin_lock+0x27/0xe0
   refill_obj_stock+0x3d/0x3a0
   kmem_cache_free+0x357/0x560
   inactive_task_timer+0x1ad/0x340
   __run_hrtimer+0x8a/0x1a0
   __hrtimer_run_queues+0x91/0x130
   hrtimer_interrupt+0x10f/0x220
   __sysvec_apic_timer_interrupt+0x7b/0xd0
   sysvec_apic_timer_interrupt+0x4f/0xd0
   asm_sysvec_apic_timer_interrupt+0x12/0x20
   RIP: 0033:0x7fff196bf6f5

Instead of calling __put_task_struct() directly, we defer it using
call_rcu(). A more natural approach would use a workqueue, but since
in PREEMPT_RT, we can't allocate dynamic memory from atomic context,
the code would become more complex because we would need to put the
work_struct instance in the task_struct and initialize it when we
allocate a new task_struct.

The issue is reproducible with stress-ng:

  while true; do
      stress-ng --sched deadline --sched-period 1000000000 \
	      --sched-runtime 800000000 --sched-deadline \
	      1000000000 --mmapfork 23 -t 20
  done

Reported-by: Hu Chunyu <chuhu@redhat.com>
Suggested-by: Oleg Nesterov <oleg@redhat.com>
Suggested-by: Valentin Schneider <vschneid@redhat.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20230614122323.37957-2-wander@redhat.com

Use __attribute__((__cleanup__(func))) to build:

 - simple auto-release pointers using __free()

 - 'classes' with constructor and destructor semantics for
   scope-based resource management.

 - lock guards based on the above classes.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20230612093537.614161713%40infradead.org