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, 135 insertions, 0 deletions

diff --git a/drivers/video/cmdline.c b/drivers/video/cmdline.c
new file mode 100644
index 000000000..49ee3fb49
--- /dev/null
+++ b/drivers/video/cmdline.c
@@ -0,0 +1,135 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Based on the fbdev code in drivers/video/fbdev/core/fb_cmdline:
+ *
+ *  Copyright (C) 2014 Intel Corp
+ *  Copyright (C) 1994 Martin Schaller
+ *
+ *	2001 - Documented with DocBook
+ *	- Brad Douglas <brad@neruo.com>
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file COPYING in the main directory of this archive
+ * for more details.
+ *
+ * Authors:
+ *    Daniel Vetter <daniel.vetter@ffwll.ch>
+ */
+
+#include <linux/fb.h> /* for FB_MAX */
+#include <linux/init.h>
+
+#include <video/cmdline.h>
+
+/*
+ * FB_MAX is the maximum number of framebuffer devices and also
+ * the maximum number of video= parameters. Although not directly
+ * related to each other, it makes sense to keep it that way.
+ */
+static const char *video_options[FB_MAX] __read_mostly;
+static const char *video_option __read_mostly;
+static int video_of_only __read_mostly;
+
+static const char *__video_get_option_string(const char *name)
+{
+	const char *options = NULL;
+	size_t name_len = 0;
+
+	if (name)
+		name_len = strlen(name);
+
+	if (name_len) {
+		unsigned int i;
+		const char *opt;
+
+		for (i = 0; i < ARRAY_SIZE(video_options); ++i) {
+			if (!video_options[i])
+				continue;
+			if (video_options[i][0] == '\0')
+				continue;
+			opt = video_options[i];
+			if (!strncmp(opt, name, name_len) && opt[name_len] == ':')
+				options = opt + name_len + 1;
+		}
+	}
+
+	/* No match, return global options */
+	if (!options)
+		options = video_option;
+
+	return options;
+}
+
+/**
+ * video_get_options - get kernel boot parameters
+ * @name:	name of the output as it would appear in the boot parameter
+ *		line (video=<name>:<options>)
+ *
+ * Looks up the video= options for the given name. Names are connector
+ * names with DRM, or driver names with fbdev. If no video option for
+ * the name has been specified, the function returns the global video=
+ * setting. A @name of NULL always returns the global video setting.
+ *
+ * Returns:
+ * The string of video options for the given name, or NULL if no video
+ * option has been specified.
+ */
+const char *video_get_options(const char *name)
+{
+	return __video_get_option_string(name);
+}
+EXPORT_SYMBOL(video_get_options);
+
+#if IS_ENABLED(CONFIG_FB_CORE)
+bool __video_get_options(const char *name, const char **options, bool is_of)
+{
+	bool enabled = true;
+	const char *opt = NULL;
+
+	if (video_of_only && !is_of)
+		enabled = false;
+
+	opt = __video_get_option_string(name);
+
+	if (options)
+		*options = opt;
+
+	return enabled;
+}
+EXPORT_SYMBOL(__video_get_options);
+#endif
+
+/*
+ * Process command line options for video adapters. This function is
+ * a __setup and __init function. It only stores the options. Drivers
+ * have to call video_get_options() as necessary.
+ */
+static int __init video_setup(char *options)
+{
+	if (!options || !*options)
+		goto out;
+
+	if (!strncmp(options, "ofonly", 6)) {
+		video_of_only = true;
+		goto out;
+	}
+
+	if (strchr(options, ':')) {
+		/* named */
+		size_t i;
+
+		for (i = 0; i < ARRAY_SIZE(video_options); i++) {
+			if (!video_options[i]) {
+				video_options[i] = options;
+				break;
+			}
+		}
+	} else {
+		/* global */
+		video_option = options;
+	}
+
+out:
+	return 1;
+}
+__setup("video=", video_setup);