Pull cgroup fixes from Tejun Heo:

 - After cgroup.kill was written to a cgroup, every child cloned into it
   with CLONE_INTO_CGROUP was spuriously killed because the fork path
   snapshotted the kill counter before resolving the target cgroup

 - Releasing an isolated cpuset partition dropped the isolation of CPUs
   isolated on the kernel command line

 - Selftest and documentation fixes

* tag 'cgroup-for-7.3-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  selftests/cgroup: test clone3() into a previously killed cgroup
  cgroup: fix spurious SIGKILL of CLONE_INTO_CGROUP children
  selftests/cgroup: Add test for preserving boot-isolated CPUs
  cgroup/cpuset: Preserve boot-isolated CPUs on partition release
  selftests/cgroup: Drop invalid boot isolation comparison
  docs: cgroup-v2: fix misc.events key format description
  selftests/cgroup: Fix cg_run_in_subcgroups ignoring arg parameter
  selftests/cgroup: set the test plan after the setup checks

Since commit b69bb476dee9 ("cgroup: fix race between fork and
cgroup.kill"), the fork path snapshots the kill_seq of the child's
future cgroup into kargs->kill_seq, and cgroup_post_fork() SIGKILLs
the child if that cgroup's kill_seq has changed in the meantime, to
catch forks racing with a cgroup.kill sweep.

For CLONE_INTO_CGROUP, however, the snapshot in cgroup_css_set_fork()
is taken before the target cgroup has been resolved: kargs->cgrp is
always NULL at this point (it is only set at the end of the function).
So the "if (kargs->cgrp)" branch is dead code and the snapshot always
records the kill_seq of the parent's cgroup. cgroup_post_fork() then
compares it with the kill_seq of the target cgroup, so the child gets
SIGKILLed whenever the two cgroups have been killed a different number
of times.

As a result, once cgroup.kill has been written to a cgroup, every
child subsequently cloned into it with clone3(CLONE_INTO_CGROUP) is
killed on the spot, for as long as the cgroup exists: kill_seq is not
exposed to userspace and never resets.

Re-snapshot kill_seq from the target cgroup once it has been resolved,
and drop the dead branch at the early snapshot site.

This does not reopen the race fixed by b69bb476dee9. For
CLONE_INTO_CGROUP, everything from the snapshot to the check in
cgroup_post_fork() runs with cgroup_mutex held, and kill_seq is
only ever incremented under cgroup_mutex.

tj: Updated the comment above kill_seq to reflect the new serialization
rules as suggested by Shakeel Butt.

Fixes: b69bb476dee9 ("cgroup: fix race between fork and cgroup.kill")
Cc: stable@vger.kernel.org
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Assisted-by: LLM
Signed-off-by: Etienne Perot <eperot@google.com>
Signed-off-by: Tejun Heo <tj@kernel.org>

Pull sched_ext updates from Tejun Heo:
 "Most of this cycle completes the enqueue-path support for hierarchical
  sub-scheduling, which makes sub-scheduler support feature complete: a
  root BPF scheduler can now hand a cgroup subtree over to a nested
  sub-scheduler together with revocable CPU grants, and the
  sub-scheduler owns all scheduling decisions for its tasks on those
  CPUs.

  Development volume was high and a number of changes plugging holes in
  the new support landed late in the cycle. Also included are core
  scheduling fixes that were completed too late for the v7.2 release and
  are routed through this pull request.

  Sub-scheduler CPU delegation:

   - Parent schedulers now grant and revoke per-CPU capabilities
     (enqueueing, preemption, CPU frequency control) on their children,
     enforced on every path a scheduler can reach a CPU through.
     Previously only dispatching could be delegated; this lets
     sub-schedulers fully schedule their CPUs.

   - Rescue execution: a task whose scheduler doesn't have access to the
     CPUs the task needs to run on starved until the watchdog ejected
     the whole scheduler. The kernel now runs such tasks directly on a
     small bandwidth budget, turning a scheduler-killing failure into
     bounded degradation.

   - Cgroup integration: tasks migrating across a sub-scheduler boundary
     weren't re-homed to the new owner, causing wrong-scheduler
     scheduling and a use-after-free. Sub-schedulers now take over their
     cgroup subtree and receive its cgroup callbacks.

   - Arena objects now cross the kernel/BPF boundary as typed pointer
     arguments, translated transparently by the BPF tree's new arena
     argument support, replacing untyped arguments with manual
     translation.

   - scx_qmap now demonstrates full hierarchical sub-scheduling.

  Other fixes and updates:

   - Robustness improvements: the abort path is now NMI-safe, fixing
     deadlocks when errors are raised from NMI context and making
     hardlockup recovery direct. Reenqueue loops that could monopolize a
     CPU ahead of the watchdog now eject the offending scheduler, and
     stalls are blamed on the scheduler actually responsible.

   - Hardening: BPF-writable arena memory is validated before kernel
     use, and task slice and vtime writes got explicit synchronization
     rules, closing corruption vectors open to buggy or malicious
     schedulers.

   - Core scheduling: sched_ext dispatching can drop the rq lock inside
     the core-wide pick, which let interleaving selections corrupt each
     other's state and hard-hang the machine. The selection now restarts
     when the lock was released. The task ordering callback was also
     invoked with its arguments swapped, and the default ordering is
     updated to work across sub-scheduler boundaries. The fixes are
     marked for stable.

   - Other fixes headed for stable: a task init leak on fork failure
     during enable, tooling compat macros that silently failed to detect
     newer kernels, and a crash on reenqueueing against a destroyed
     dispatch queue.

   - Tooling: scx_pair moves off deprecated callbacks, and the
     deprecated scx_bpf_cpu_rq() kfunc is removed"

* tag 'sched_ext-for-7.3' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext: (144 commits)
  sched_ext: Drop the dead SCX_DEQ_CORE_SCHED_EXEC test in dequeue_task_scx()
  sched_ext: Make core-sched task ordering hierarchy-aware
  sched_ext: Use runnable_at for the default core-sched task ordering
  sched_ext: Fix inverted ops.core_sched_before() invocation
  sched_ext: Move the config-off sub-cap kfunc stubs into sub.c
  sched_ext: Rename balance-era identifiers to dispatch terms
  sched_ext: Drop the stale keep_prev fixup in dispatch_pick()
  sched_ext: Keep kick_sync waiting on the rq's own CPU
  sched_ext: Make SCHED_CLASS_EXT select GENERIC_ALLOCATOR
  sched_ext/scx_flatcg: Fix cvtime true-up on slice expiry
  sched_ext: Don't BUG_ON a destroyed DSQ in process_deferred_reenq_users
  sched_ext: Fix scx_bpf_dsq_move_to_local___v2 compat detection
  sched_ext: Make scx_bpf_events() read the calling scheduler's counters
  sched_ext: Drop unlocked scx_rq_clock_invalidate() from scx_root_disable()
  selftests/sched_ext: Fix flaky ddsp failure tests on busy systems
  selftests/sched_ext: Make numa idle validation race-free
  sched_ext: Fix scx_bpf_dsq_reenq___compat kfunc extern prototype
  sched_ext/scx_flatcg: expire cached hweights on weight changes
  sched_ext: Fix exit_task leak on fork failure during enable
  sched_ext: fix stale references in doc comments
  ...

Pull cgroup updates from Tejun Heo:

 - Attach path bug fixes: migrations spanning multiple source or
   destination cpusets were mishandled, most visibly leaving thread
   affinities stale when the controller is disabled in a threaded
   subtree. Configuration writes could also race an in-flight attach and
   apply stale state, and the deadline task count could get corrupted by
   concurrent updates, skewing SCHED_DEADLINE admission decisions.

 - Memory binding bug fixes: which node masks get applied differed
   between the binding update paths, and tasks cloned with
   CLONE_INTO_CGROUP skipped rebinding entirely. Rebinding also now runs
   once per process instead of repeating for every thread sharing the
   mm.

 - Overhead removals with no behavior change: CPU hotplug iterated tasks
   of cpusets that just inherit the parent's effective masks, and the
   slab-spreading task flag was still being maintained although the SLAB
   allocator that consumed it is long gone.

 - Data-race annotations for benign races so that KCSAN reports stay
   meaningful, selftest coverage for the fixes above along with
   flakiness and portability fixes, and documentation corrections.

* tag 'cgroup-for-7.3' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: (34 commits)
  selftests/cgroup: Remove redundant chown in test_cgcore_lesser_ns_open
  selftests/cgroup: Preserve CPU hotplug write errors
  cgroup/cpuset: Add test for partition root invalidation returning wrong CPUs
  cgroup/cpuset: Remove obsolete PFA_SPREAD_SLAB task flag
  docs: cgroup-v2: fix stale "io" controller introduction
  selftests/cgroup: Avoid awk -e in cpuset tests
  cgroup/cpuset: Use WRITE_ONCE() for shared prs_err updates
  selftests/cgroup: add user_usec sanity check in test_cpucg_nice
  cgroup: drop unneeded semicolon
  docs: cgroup-v2: mark memory.pressure and io.pressure as read-write
  selftests/cgroup: Fix minor defects in test_cpuset
  Docs/admin-guide/cgroup-v2: fix delay_nsec unit in io.latency doc
  selftests/cgroup: Remove redundant cg_enter_current() call in test_core
  selftests/cgroup: Add test for cpuset affinity on controller disable
  cgroup/cpuset: Handle the special case of non-moving tasks in cpuset_can_attach()
  cgroup/cpuset: Support multiple destination cpusets for cpuset_*attach()
  selftests/cgroup: fix missing TAP output in test_hugetlb_memcg
  cgroup/cpuset: Support multiple source cpusets for cpuset_*attach()
  cgroup/cpuset: Move mpol_rebind_mm/cpuset_migrate_mm() calls inside cpuset_attach_task()
  cgroup/cpuset: Make attach_ctx.old_cs track task group leader
  ...

Pull bpf-next d114bb989367 ("Merge branch
'add-arena-argument-support-to-kfuncs-and-struct_ops'") to make the __arena
and __arena__nullable kfunc and struct_ops argument suffixes available. The
suffixed arguments will be used to convert sched_ext kfuncs and struct_ops
callbacks that currently pass arena pointers as scalars and rebase them by
hand.

The trailing semicolon belongs at the point of use, not in the macro
definition. All uses have been verified to have their own semicolons.

This was found using the following Coccinelle semantic patch:

@r@
identifier i : script:ocaml() { String.lowercase_ascii i = i };
expression e;
@@

*#define i(...) e;

Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Signed-off-by: Tejun Heo <tj@kernel.org>

A subsystem can attach to the cgroup hierarchy itself, independent of which
controllers are enabled where - BPF hooks already behave this way and
sched_ext sub-schedulers do too. Controller callbacks can't track task
migrations for them: sched_ext must re-home a task whose migration crosses a
sub-scheduler boundary, but the cpu controller's attach callbacks fire only
when the task_group changes and miss moves whenever the controller topology
is coarser than the sub-scheduler topology.

Add cgroup_task_notifier with per-task migration events mirroring the
can_attach/attach/cancel_attach phases so that a consumer which prepares
per-task state can also veto a migration: CGROUP_TASK_MIGRATING fires
pre-commit, CGROUP_TASK_MIGRATED post-commit and
CGROUP_TASK_MIGRATE_CANCELED unwinds a failed migration. Only migrations
that change a task's dfl cgroup are reported.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>

a5b98009f16d ("sched/psi: fix race between file release and pressure write")
made pressure_write() hold cgroup_mutex across psi_trigger_create(), which
forks the psimon kthread for the first rtpoll trigger. As kthread creation
depends on the whole fork path, the commit inadvertently created a lot of
unwanted locking dependencies from cgroup_mutex.

sched_ext got hit by one: its enable path blocks forks and then grabs
cgroup_mutex, so a pressure write racing a scheduler enable deadlocks, with
every other fork piling up behind.

Fix it by splitting trigger creation so that the worker is forked with
cgroup_mutex dropped and the kernfs active reference left broken. The latter
matters because rmdir and cgroup.pressure writes drain active references
under cgroup_mutex. Publishing the trigger last keeps error reporting
synchronous and preserves the of->priv lifetime rules.

The trigger registered in the first stage pins the group's rtpoll machinery
across the unlocked window, leaving only creation races to resolve. The
catch-up poll on installation covers scheduling attempts dropped while there
was no worker.

v2: Retagged sched/psi (was cgroup).

Fixes: a5b98009f16d ("sched/psi: fix race between file release and pressure write")
Cc: stable@vger.kernel.org
Cc: Edward Adam Davis <eadavis@qq.com>
Cc: Chen Ridong <chenridong@huaweicloud.com>
Reported-by: Matt Fleming <mfleming@cloudflare.com>
Closes: https://lore.kernel.org/all/20260710100441.2653477-1-matt@readmodwrite.com/
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: Matt Fleming <mfleming@cloudflare.com>
Acked-by: Suren Baghdasaryan <surenb@google.com>

Pull cgroup updates from Tejun Heo:

 - Last cycle deferred css teardown on cgroup removal until the cgroup
   depopulated, so a css is not taken offline while tasks can still
   reference it. Disabling a controller through cgroup.subtree_control
   still had the same problem. This reworks the deferral from per-cgroup
   to per-css so that path is covered too.

 - New RDMA controller monitoring files: rdma.peak for per-device peak
   usage and rdma.events / rdma.events.local for resource-limit
   exhaustion. The max-limit parser was rewritten, fixing two input
   parsing bugs.

 - cpuset: fix a sched-domain leak on the domain-rebuild failure path
   and skip a redundant hardwall ancestor scan on v2.

 - Misc: pair the remaining lockless cgroup.max.* reads with WRITE_ONCE,
   assorted selftest robustness fixes, and doc path corrections.

* tag 'cgroup-for-7.2' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: (22 commits)
  cgroup: Migrate tasks to the root css when a controller is rebound
  docs: cgroup: Fix stale source file paths
  cgroup/cpuset: Free sched domains on rebuild guard failure
  cgroup: pair max limit READ_ONCE() with WRITE_ONCE()
  selftests/cgroup: enable memory controller in hugetlb memcg test
  cgroup/rdma: Drop unnecessary READ_ONCE() on event counters
  cgroup: Defer kill_css_finish() in cgroup_apply_control_disable()
  cgroup: Add per-subsys-css kill_css_finish deferral
  cgroup: Move populated counters to cgroup_subsys_state
  cgroup: Annotate unlocked nr_populated_* accesses with READ_ONCE/WRITE_ONCE
  cgroup: Inline cgroup_has_tasks() in cgroup.h
  cgroup/rdma: document rdma.peak, rdma.events and rdma.events.local
  cgroup/rdma: add rdma.events.local for per-cgroup allocation failure attribution
  cgroup/rdma: add rdma.events to track resource limit exhaustion
  cgroup/rdma: add rdma.peak for per-device peak usage tracking
  selftests/cgroup: check malloc return value in alloc_anon functions
  cgroup/cpuset: Skip hardwall ancestor scan in cpuset v2 in cpuset_current_node_allowed()
  selftests/cgroup: fix misleading debug message in test_cgfreezer_time_child
  selftests/cgroup: fix child process escaping to parent cleanup in test_cpucg_nice
  selftests/cgroup: Add NULL check after malloc in cgroup_util.c
  ...

cgroup_apply_control_disable() defers kill_css_finish() while a css is
still populated, relying on css_update_populated() to fire the deferred
kill once the populated count reaches zero.

This deadlocks when a controller is rebound out of a hierarchy. Mounting
an implicit_on_dfl controller such as perf_event as a v1 hierarchy steals
it off the default hierarchy, and rebind_subsystems() kills its
per-cgroup csses while they are still populated. The migration run in the
same step keeps the old css for a controller no longer in the hierarchy's
mask, so no task is migrated off the dying csses. Their populated count
never reaches zero, the deferred kill_css_finish() never fires, and the
next cgroup_lock_and_drain_offline() hangs forever under cgroup_mutex.

That migration is already a no-op pass over the rebound subtree. Add
cgroup_rebind_ss_mask so find_existing_css_set() resolves the leaving
controllers to the root css. Their tasks are migrated there, the
per-cgroup csses depopulate, and cgroup_apply_control_disable() kills
them synchronously. The deferral stays correct for the rmdir and
controller-disable paths it was meant for.

Fixes: 1dffd95575eb ("cgroup: Defer kill_css_finish() in cgroup_apply_control_disable()")
Reported-by: Mark Brown <broonie@kernel.org>
Closes: https://lore.kernel.org/all/41cd159c-54e5-45e0-81df-eaf36a6c028e@sirena.org.uk/
Reported-by: Bert Karwatzki <spasswolf@web.de>
Closes: https://lore.kernel.org/all/4e986b4ed7e16547805d54b6e67d09120bc4d2f2.camel@web.de/
Tested-by: Mark Brown <broonie@kernel.org>
Tested-by: Bert Karwatzki <spasswolf@web.de>
Signed-off-by: Tejun Heo <tj@kernel.org>