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, 105 insertions, 0 deletions

diff --git a/include/video/display_timing.h b/include/video/display_timing.h
new file mode 100644
index 000000000..37a017607
--- /dev/null
+++ b/include/video/display_timing.h
@@ -0,0 +1,105 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright 2012 Steffen Trumtrar <s.trumtrar@pengutronix.de>
+ *
+ * description of display timings
+ */
+
+#ifndef __LINUX_DISPLAY_TIMING_H
+#define __LINUX_DISPLAY_TIMING_H
+
+#include <linux/bitops.h>
+#include <linux/types.h>
+
+enum display_flags {
+	DISPLAY_FLAGS_HSYNC_LOW		= BIT(0),
+	DISPLAY_FLAGS_HSYNC_HIGH	= BIT(1),
+	DISPLAY_FLAGS_VSYNC_LOW		= BIT(2),
+	DISPLAY_FLAGS_VSYNC_HIGH	= BIT(3),
+
+	/* data enable flag */
+	DISPLAY_FLAGS_DE_LOW		= BIT(4),
+	DISPLAY_FLAGS_DE_HIGH		= BIT(5),
+	/* drive data on pos. edge */
+	DISPLAY_FLAGS_PIXDATA_POSEDGE	= BIT(6),
+	/* drive data on neg. edge */
+	DISPLAY_FLAGS_PIXDATA_NEGEDGE	= BIT(7),
+	DISPLAY_FLAGS_INTERLACED	= BIT(8),
+	DISPLAY_FLAGS_DOUBLESCAN	= BIT(9),
+	DISPLAY_FLAGS_DOUBLECLK		= BIT(10),
+	/* drive sync on pos. edge */
+	DISPLAY_FLAGS_SYNC_POSEDGE	= BIT(11),
+	/* drive sync on neg. edge */
+	DISPLAY_FLAGS_SYNC_NEGEDGE	= BIT(12),
+};
+
+/*
+ * A single signal can be specified via a range of minimal and maximal values
+ * with a typical value, that lies somewhere inbetween.
+ */
+struct timing_entry {
+	u32 min;
+	u32 typ;
+	u32 max;
+};
+
+/*
+ * Single "mode" entry. This describes one set of signal timings a display can
+ * have in one setting. This struct can later be converted to struct videomode
+ * (see include/video/videomode.h). As each timing_entry can be defined as a
+ * range, one struct display_timing may become multiple struct videomodes.
+ *
+ * Example: hsync active high, vsync active low
+ *
+ *				    Active Video
+ * Video  ______________________XXXXXXXXXXXXXXXXXXXXXX_____________________
+ *	  |<- sync ->|<- back ->|<----- active ----->|<- front ->|<- sync..
+ *	  |	     |	 porch  |		     |	 porch	 |
+ *
+ * HSync _|¯¯¯¯¯¯¯¯¯¯|___________________________________________|¯¯¯¯¯¯¯¯¯
+ *
+ * VSync ¯|__________|¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯|_________
+ */
+struct display_timing {
+	struct timing_entry pixelclock;
+
+	struct timing_entry hactive;		/* hor. active video */
+	struct timing_entry hfront_porch;	/* hor. front porch */
+	struct timing_entry hback_porch;	/* hor. back porch */
+	struct timing_entry hsync_len;		/* hor. sync len */
+
+	struct timing_entry vactive;		/* ver. active video */
+	struct timing_entry vfront_porch;	/* ver. front porch */
+	struct timing_entry vback_porch;	/* ver. back porch */
+	struct timing_entry vsync_len;		/* ver. sync len */
+
+	enum display_flags flags;		/* display flags */
+};
+
+/*
+ * This describes all timing settings a display provides.
+ * The native_mode is the default setting for this display.
+ * Drivers that can handle multiple videomodes should work with this struct and
+ * convert each entry to the desired end result.
+ */
+struct display_timings {
+	unsigned int num_timings;
+	unsigned int native_mode;
+
+	struct display_timing **timings;
+};
+
+/* get one entry from struct display_timings */
+static inline struct display_timing *display_timings_get(const struct
+							 display_timings *disp,
+							 unsigned int index)
+{
+	if (disp->num_timings > index)
+		return disp->timings[index];
+	else
+		return NULL;
+}
+
+void display_timings_release(struct display_timings *disp);
+
+#endif