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, 82 insertions, 0 deletions

diff --git a/include/sound/ac97/controller.h b/include/sound/ac97/controller.h
new file mode 100644
index 000000000..d5895ea39
--- /dev/null
+++ b/include/sound/ac97/controller.h
@@ -0,0 +1,82 @@
+/* SPDX-License-Identifier: GPL-2.0
+ *
+ * Copyright (C) 2016 Robert Jarzmik <robert.jarzmik@free.fr>
+ */
+
+#ifndef AC97_CONTROLLER_H
+#define AC97_CONTROLLER_H
+
+#include <linux/device.h>
+#include <linux/list.h>
+
+#define AC97_BUS_MAX_CODECS 4
+#define AC97_SLOTS_AVAILABLE_ALL 0xf
+
+struct ac97_controller_ops;
+
+/**
+ * struct ac97_controller - The AC97 controller of the AC-Link
+ * @ops:		the AC97 operations.
+ * @controllers:	linked list of all existing controllers.
+ * @adap:		the shell device ac97-%d, ie. ac97 adapter
+ * @nr:			the number of the shell device
+ * @slots_available:	the mask of accessible/scanable codecs.
+ * @parent:		the device providing the AC97 controller.
+ * @codecs:		the 4 possible AC97 codecs (NULL if none found).
+ * @codecs_pdata:	platform_data for each codec (NULL if no pdata).
+ *
+ * This structure is internal to AC97 bus, and should not be used by the
+ * controllers themselves, excepting for using @dev.
+ */
+struct ac97_controller {
+	const struct ac97_controller_ops *ops;
+	struct list_head controllers;
+	struct device adap;
+	int nr;
+	unsigned short slots_available;
+	struct device *parent;
+	struct ac97_codec_device *codecs[AC97_BUS_MAX_CODECS];
+	void *codecs_pdata[AC97_BUS_MAX_CODECS];
+};
+
+/**
+ * struct ac97_controller_ops - The AC97 operations
+ * @reset:	Cold reset of the AC97 AC-Link.
+ * @warm_reset:	Warm reset of the AC97 AC-Link.
+ * @read:	Read of a single AC97 register.
+ *		Returns the register value or a negative error code.
+ * @write:	Write of a single AC97 register.
+ *
+ * These are the basic operation an AC97 controller must provide for an AC97
+ * access functions. Amongst these, all but the last 2 are mandatory.
+ * The slot number is also known as the AC97 codec number, between 0 and 3.
+ */
+struct ac97_controller_ops {
+	void (*reset)(struct ac97_controller *adrv);
+	void (*warm_reset)(struct ac97_controller *adrv);
+	int (*write)(struct ac97_controller *adrv, int slot,
+		     unsigned short reg, unsigned short val);
+	int (*read)(struct ac97_controller *adrv, int slot, unsigned short reg);
+};
+
+#if IS_ENABLED(CONFIG_AC97_BUS_NEW)
+struct ac97_controller *snd_ac97_controller_register(
+	const struct ac97_controller_ops *ops, struct device *dev,
+	unsigned short slots_available);
+void snd_ac97_controller_unregister(struct ac97_controller *ac97_ctrl);
+#else
+static inline struct ac97_controller *
+snd_ac97_controller_register(const struct ac97_controller_ops *ops,
+			     struct device *dev,
+			     unsigned short slots_available)
+{
+	return ERR_PTR(-ENODEV);
+}
+
+static inline void
+snd_ac97_controller_unregister(struct ac97_controller *ac97_ctrl)
+{
+}
+#endif
+
+#endif