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, 154 insertions, 0 deletions

diff --git a/drivers/base/map.c b/drivers/base/map.c
new file mode 100644
index 000000000..bde415f93
--- /dev/null
+++ b/drivers/base/map.c
@@ -0,0 +1,154 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  linux/drivers/base/map.c
+ *
+ * (C) Copyright Al Viro 2002,2003
+ *
+ * NOTE: data structure needs to be changed.  It works, but for large dev_t
+ * it will be too slow.  It is isolated, though, so these changes will be
+ * local to that file.
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/mutex.h>
+#include <linux/kdev_t.h>
+#include <linux/kobject.h>
+#include <linux/kobj_map.h>
+
+struct kobj_map {
+	struct probe {
+		struct probe *next;
+		dev_t dev;
+		unsigned long range;
+		struct module *owner;
+		kobj_probe_t *get;
+		int (*lock)(dev_t, void *);
+		void *data;
+	} *probes[255];
+	struct mutex *lock;
+};
+
+int kobj_map(struct kobj_map *domain, dev_t dev, unsigned long range,
+	     struct module *module, kobj_probe_t *probe,
+	     int (*lock)(dev_t, void *), void *data)
+{
+	unsigned int n = MAJOR(dev + range - 1) - MAJOR(dev) + 1;
+	unsigned int index = MAJOR(dev);
+	unsigned int i;
+	struct probe *p;
+
+	if (n > 255)
+		n = 255;
+
+	p = kmalloc_objs(struct probe, n);
+	if (p == NULL)
+		return -ENOMEM;
+
+	for (i = 0; i < n; i++, p++) {
+		p->owner = module;
+		p->get = probe;
+		p->lock = lock;
+		p->dev = dev;
+		p->range = range;
+		p->data = data;
+	}
+	mutex_lock(domain->lock);
+	for (i = 0, p -= n; i < n; i++, p++, index++) {
+		struct probe **s = &domain->probes[index % 255];
+		while (*s && (*s)->range < range)
+			s = &(*s)->next;
+		p->next = *s;
+		*s = p;
+	}
+	mutex_unlock(domain->lock);
+	return 0;
+}
+
+void kobj_unmap(struct kobj_map *domain, dev_t dev, unsigned long range)
+{
+	unsigned int n = MAJOR(dev + range - 1) - MAJOR(dev) + 1;
+	unsigned int index = MAJOR(dev);
+	unsigned int i;
+	struct probe *found = NULL;
+
+	if (n > 255)
+		n = 255;
+
+	mutex_lock(domain->lock);
+	for (i = 0; i < n; i++, index++) {
+		struct probe **s;
+		for (s = &domain->probes[index % 255]; *s; s = &(*s)->next) {
+			struct probe *p = *s;
+			if (p->dev == dev && p->range == range) {
+				*s = p->next;
+				if (!found)
+					found = p;
+				break;
+			}
+		}
+	}
+	mutex_unlock(domain->lock);
+	kfree(found);
+}
+
+struct kobject *kobj_lookup(struct kobj_map *domain, dev_t dev, int *index)
+{
+	struct kobject *kobj;
+	struct probe *p;
+	unsigned long best = ~0UL;
+
+retry:
+	mutex_lock(domain->lock);
+	for (p = domain->probes[MAJOR(dev) % 255]; p; p = p->next) {
+		struct kobject *(*probe)(dev_t, int *, void *);
+		struct module *owner;
+		void *data;
+
+		if (p->dev > dev || p->dev + p->range - 1 < dev)
+			continue;
+		if (p->range - 1 >= best)
+			break;
+		if (!try_module_get(p->owner))
+			continue;
+		owner = p->owner;
+		data = p->data;
+		probe = p->get;
+		best = p->range - 1;
+		*index = dev - p->dev;
+		if (p->lock && p->lock(dev, data) < 0) {
+			module_put(owner);
+			continue;
+		}
+		mutex_unlock(domain->lock);
+		kobj = probe(dev, index, data);
+		/* Currently ->owner protects _only_ ->probe() itself. */
+		module_put(owner);
+		if (kobj)
+			return kobj;
+		goto retry;
+	}
+	mutex_unlock(domain->lock);
+	return NULL;
+}
+
+struct kobj_map *kobj_map_init(kobj_probe_t *base_probe, struct mutex *lock)
+{
+	struct kobj_map *p = kmalloc_obj(struct kobj_map);
+	struct probe *base = kzalloc_obj(*base);
+	int i;
+
+	if ((p == NULL) || (base == NULL)) {
+		kfree(p);
+		kfree(base);
+		return NULL;
+	}
+
+	base->dev = 1;
+	base->range = ~0;
+	base->get = base_probe;
+	for (i = 0; i < 255; i++)
+		p->probes[i] = base;
+	p->lock = lock;
+	return p;
+}