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, 173 insertions, 0 deletions

diff --git a/samples/livepatch/livepatch-shadow-fix1.c b/samples/livepatch/livepatch-shadow-fix1.c
new file mode 100644
index 000000000..cbf68ca40
--- /dev/null
+++ b/samples/livepatch/livepatch-shadow-fix1.c
@@ -0,0 +1,173 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2017 Joe Lawrence <joe.lawrence@redhat.com>
+ */
+
+/*
+ * livepatch-shadow-fix1.c - Shadow variables, livepatch demo
+ *
+ * Purpose
+ * -------
+ *
+ * Fixes the memory leak introduced in livepatch-shadow-mod through the
+ * use of a shadow variable.  This fix demonstrates the "extending" of
+ * short-lived data structures by patching its allocation and release
+ * functions.
+ *
+ *
+ * Usage
+ * -----
+ *
+ * This module is not intended to be standalone.  See the "Usage"
+ * section of livepatch-shadow-mod.c.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/livepatch.h>
+#include <linux/slab.h>
+
+/* Shadow variable enums */
+#define SV_LEAK		1
+
+/* Allocate new dummies every second */
+#define ALLOC_PERIOD	1
+/* Check for expired dummies after a few new ones have been allocated */
+#define CLEANUP_PERIOD	(3 * ALLOC_PERIOD)
+/* Dummies expire after a few cleanup instances */
+#define EXPIRE_PERIOD	(4 * CLEANUP_PERIOD)
+
+struct dummy {
+	struct list_head list;
+	unsigned long jiffies_expire;
+};
+
+/*
+ * The constructor makes more sense together with klp_shadow_get_or_alloc().
+ * In this example, it would be safe to assign the pointer also to the shadow
+ * variable returned by klp_shadow_alloc().  But we wanted to show the more
+ * complicated use of the API.
+ */
+static int shadow_leak_ctor(void *obj, void *shadow_data, void *ctor_data)
+{
+	int **shadow_leak = shadow_data;
+	int **leak = ctor_data;
+
+	if (!ctor_data)
+		return -EINVAL;
+
+	*shadow_leak = *leak;
+	return 0;
+}
+
+static struct dummy *livepatch_fix1_dummy_alloc(void)
+{
+	struct dummy *d;
+	int *leak;
+	int **shadow_leak;
+
+	d = kzalloc(sizeof(*d), GFP_KERNEL);
+	if (!d)
+		return NULL;
+
+	d->jiffies_expire = jiffies + secs_to_jiffies(EXPIRE_PERIOD);
+
+	/*
+	 * Patch: save the extra memory location into a SV_LEAK shadow
+	 * variable.  A patched dummy_free routine can later fetch this
+	 * pointer to handle resource release.
+	 */
+	leak = kzalloc(sizeof(*leak), GFP_KERNEL);
+	if (!leak)
+		goto err_leak;
+
+	shadow_leak = klp_shadow_alloc(d, SV_LEAK, sizeof(leak), GFP_KERNEL,
+				       shadow_leak_ctor, &leak);
+	if (!shadow_leak) {
+		pr_err("%s: failed to allocate shadow variable for the leaking pointer: dummy @ %p, leak @ %p\n",
+		       __func__, d, leak);
+		goto err_shadow;
+	}
+
+	pr_info("%s: dummy @ %p, expires @ %lx\n",
+		__func__, d, d->jiffies_expire);
+
+	return d;
+
+err_shadow:
+	kfree(leak);
+err_leak:
+	kfree(d);
+	return NULL;
+}
+
+static void livepatch_fix1_dummy_leak_dtor(void *obj, void *shadow_data)
+{
+	void *d = obj;
+	int **shadow_leak = shadow_data;
+
+	pr_info("%s: dummy @ %p, prevented leak @ %p\n",
+			 __func__, d, *shadow_leak);
+	kfree(*shadow_leak);
+}
+
+static void livepatch_fix1_dummy_free(struct dummy *d)
+{
+	int **shadow_leak;
+
+	/*
+	 * Patch: fetch the saved SV_LEAK shadow variable, detach and
+	 * free it.  Note: handle cases where this shadow variable does
+	 * not exist (ie, dummy structures allocated before this livepatch
+	 * was loaded.)
+	 */
+	shadow_leak = klp_shadow_get(d, SV_LEAK);
+	if (shadow_leak)
+		klp_shadow_free(d, SV_LEAK, livepatch_fix1_dummy_leak_dtor);
+	else
+		pr_info("%s: dummy @ %p leaked!\n", __func__, d);
+
+	kfree(d);
+}
+
+static struct klp_func funcs[] = {
+	{
+		.old_name = "dummy_alloc",
+		.new_func = livepatch_fix1_dummy_alloc,
+	},
+	{
+		.old_name = "dummy_free",
+		.new_func = livepatch_fix1_dummy_free,
+	}, { }
+};
+
+static struct klp_object objs[] = {
+	{
+		.name = "livepatch_shadow_mod",
+		.funcs = funcs,
+	}, { }
+};
+
+static struct klp_patch patch = {
+	.mod = THIS_MODULE,
+	.objs = objs,
+};
+
+static int livepatch_shadow_fix1_init(void)
+{
+	return klp_enable_patch(&patch);
+}
+
+static void livepatch_shadow_fix1_exit(void)
+{
+	/* Cleanup any existing SV_LEAK shadow variables */
+	klp_shadow_free_all(SV_LEAK, livepatch_fix1_dummy_leak_dtor);
+}
+
+module_init(livepatch_shadow_fix1_init);
+module_exit(livepatch_shadow_fix1_exit);
+MODULE_DESCRIPTION("Live patching demo for shadow variables");
+MODULE_LICENSE("GPL");
+MODULE_INFO(livepatch, "Y");