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, 80 insertions, 0 deletions

diff --git a/include/kunit/device.h b/include/kunit/device.h
new file mode 100644
index 000000000..2450110ad
--- /dev/null
+++ b/include/kunit/device.h
@@ -0,0 +1,80 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * KUnit basic device implementation
+ *
+ * Helpers for creating and managing fake devices for KUnit tests.
+ *
+ * Copyright (C) 2023, Google LLC.
+ * Author: David Gow <davidgow@google.com>
+ */
+
+#ifndef _KUNIT_DEVICE_H
+#define _KUNIT_DEVICE_H
+
+#if IS_ENABLED(CONFIG_KUNIT)
+
+#include <kunit/test.h>
+
+struct device;
+struct device_driver;
+
+/**
+ * kunit_driver_create() - Create a struct device_driver attached to the kunit_bus
+ * @test: The test context object.
+ * @name: The name to give the created driver.
+ *
+ * Creates a struct device_driver attached to the kunit_bus, with the name @name.
+ * This driver will automatically be cleaned up on test exit.
+ *
+ * Return: a stub struct device_driver, managed by KUnit, with the name @name.
+ */
+struct device_driver *kunit_driver_create(struct kunit *test, const char *name);
+
+/**
+ * kunit_device_register() - Create a struct device for use in KUnit tests
+ * @test: The test context object.
+ * @name: The name to give the created device.
+ *
+ * Creates a struct kunit_device (which is a struct device) with the given name,
+ * and a corresponding driver. The device and driver will be cleaned up on test
+ * exit, or when kunit_device_unregister is called. See also
+ * kunit_device_register_with_driver, if you wish to provide your own
+ * struct device_driver.
+ *
+ * Return: a pointer to a struct device which will be cleaned up when the test
+ * exits, or an error pointer if the device could not be allocated or registered.
+ */
+struct device *kunit_device_register(struct kunit *test, const char *name);
+
+/**
+ * kunit_device_register_with_driver() - Create a struct device for use in KUnit tests
+ * @test: The test context object.
+ * @name: The name to give the created device.
+ * @drv: The struct device_driver to associate with the device.
+ *
+ * Creates a struct kunit_device (which is a struct device) with the given
+ * name, and driver. The device will be cleaned up on test exit, or when
+ * kunit_device_unregister is called. See also kunit_device_register, if you
+ * wish KUnit to create and manage a driver for you.
+ *
+ * Return: a pointer to a struct device which will be cleaned up when the test
+ * exits, or an error pointer if the device could not be allocated or registered.
+ */
+struct device *kunit_device_register_with_driver(struct kunit *test,
+						 const char *name,
+						 const struct device_driver *drv);
+
+/**
+ * kunit_device_unregister() - Unregister a KUnit-managed device
+ * @test: The test context object which created the device
+ * @dev: The device.
+ *
+ * Unregisters and destroys a struct device which was created with
+ * kunit_device_register or kunit_device_register_with_driver. If KUnit created
+ * a driver, cleans it up as well.
+ */
+void kunit_device_unregister(struct kunit *test, struct device *dev);
+
+#endif
+
+#endif