sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted

alb_break_llc() decides whether to break LLC preference during active
load balance. It does so by testing that every runnable fair task on the
source rq prefers its LLC:

	env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable

But the two counters cover different sets. nr_pref_llc_running is updated
in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued,
so it follows queued tasks. h_nr_runnable is updated in set_delayed()/
clear_delayed() and drops delay-dequeued tasks.

So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted
in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks,
alb_break_llc() returns false, and active balance is free to pull a task
off its preferred LLC. Active balance only moves runnable tasks, and this
is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB
skips the per-task test in can_migrate_task(). The runnable set is the one
we want.

Fix it on the counter side. A task should be counted in
nr_pref_llc_running exactly while it is both queued on its preferred LLC
(pref_llc_queued) and runnable (!sched_delayed). Define that membership
once in task_pref_llc_runnable(), and adjust the counter only through
pref_llc_running_inc()/pref_llc_running_dec() from the four sites that
change either input: account_llc_enqueue(), account_llc_dequeue(),
set_delayed() and clear_delayed(). Gating every update on the same
predicate keeps the delay, wake and dequeue paths from double-counting
or underflowing; see the comments at those sites for the ordering.

nr_llc_running and sd->llc_counts are not touched and stay on queued
semantics.

Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC")
Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/
Reported-by: Zhan Xusheng <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com

1 files changed, 84 insertions, 0 deletions

diff --git a/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml b/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml
new file mode 100644
index 000000000..99e2865f0
--- /dev/null
+++ b/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml
@@ -0,0 +1,84 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/access-controllers/access-controllers.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Generic Domain Access Controllers
+
+maintainers:
+  - Oleksii Moisieiev <oleksii_moisieiev@epam.com>
+
+description: |+
+  Common access controllers properties
+
+  Access controllers are in charge of stating which of the hardware blocks under
+  their responsibility (their domain) can be accesssed by which compartment. A
+  compartment can be a cluster of CPUs (or coprocessors), a range of addresses
+  or a group of hardware blocks. An access controller's domain is the set of
+  resources covered by the access controller.
+
+  This device tree binding can be used to bind devices to their access
+  controller provided by access-controllers property. In this case, the device
+  is a consumer and the access controller is the provider.
+
+  An access controller can be represented by any node in the device tree and
+  can provide one or more configuration parameters, needed to control parameters
+  of the consumer device. A consumer node can refer to the provider by phandle
+  and a set of phandle arguments, specified by '#access-controller-cells'
+  property in the access controller node.
+
+  Access controllers are typically used to set/read the permissions of a
+  hardware block and grant access to it. Any of which depends on the access
+  controller. The capabilities of each access controller are defined by the
+  binding of the access controller device.
+
+  Each node can be a consumer for the several access controllers.
+
+# always select the core schema
+select: true
+
+properties:
+  "#access-controller-cells":
+    description:
+      Number of cells in an access-controllers specifier;
+      Can be any value as specified by device tree binding documentation
+      of a particular provider. The node is an access controller.
+
+  access-controller-names:
+    $ref: /schemas/types.yaml#/definitions/string-array
+    description:
+      A list of access-controllers names, sorted in the same order as
+      access-controllers entries. Consumer drivers will use
+      access-controller-names to match with existing access-controllers entries.
+
+  access-controllers:
+    $ref: /schemas/types.yaml#/definitions/phandle-array
+    description:
+      A list of access controller specifiers, as defined by the
+      bindings of the access-controllers provider.
+
+additionalProperties: true
+
+examples:
+  - |
+    clock_controller: access-controllers@50000 {
+        reg = <0x50000 0x400>;
+        #access-controller-cells = <2>;
+    };
+
+    bus_controller: bus@60000 {
+        reg = <0x60000 0x10000>;
+        #address-cells = <1>;
+        #size-cells = <1>;
+        ranges;
+        #access-controller-cells = <3>;
+
+        uart4: serial@60100 {
+            reg = <0x60100 0x400>;
+            clocks = <&clk_serial>;
+            access-controllers = <&clock_controller 1 2>,
+                                 <&bus_controller 1 3 5>;
+            access-controller-names = "clock", "bus";
+        };
+    };