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, 102 insertions, 0 deletions

diff --git a/drivers/mtd/chips/chipreg.c b/drivers/mtd/chips/chipreg.c
new file mode 100644
index 000000000..a05e10368
--- /dev/null
+++ b/drivers/mtd/chips/chipreg.c
@@ -0,0 +1,102 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Registration for chip drivers
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/kmod.h>
+#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <linux/mtd/map.h>
+#include <linux/mtd/mtd.h>
+
+static DEFINE_SPINLOCK(chip_drvs_lock);
+static LIST_HEAD(chip_drvs_list);
+
+void register_mtd_chip_driver(struct mtd_chip_driver *drv)
+{
+	spin_lock(&chip_drvs_lock);
+	list_add(&drv->list, &chip_drvs_list);
+	spin_unlock(&chip_drvs_lock);
+}
+
+void unregister_mtd_chip_driver(struct mtd_chip_driver *drv)
+{
+	spin_lock(&chip_drvs_lock);
+	list_del(&drv->list);
+	spin_unlock(&chip_drvs_lock);
+}
+
+static struct mtd_chip_driver *get_mtd_chip_driver (const char *name)
+{
+	struct mtd_chip_driver *ret = NULL, *this;
+
+	spin_lock(&chip_drvs_lock);
+
+	list_for_each_entry(this, &chip_drvs_list, list) {
+		if (!strcmp(this->name, name)) {
+			ret = this;
+			break;
+		}
+	}
+	if (ret && !try_module_get(ret->module))
+		ret = NULL;
+
+	spin_unlock(&chip_drvs_lock);
+
+	return ret;
+}
+
+	/* Hide all the horrid details, like some silly person taking
+	   get_module_symbol() away from us, from the caller. */
+
+struct mtd_info *do_map_probe(const char *name, struct map_info *map)
+{
+	struct mtd_chip_driver *drv;
+	struct mtd_info *ret;
+
+	drv = get_mtd_chip_driver(name);
+
+	if (!drv && !request_module("%s", name))
+		drv = get_mtd_chip_driver(name);
+
+	if (!drv)
+		return NULL;
+
+	ret = drv->probe(map);
+
+	/* We decrease the use count here. It may have been a
+	   probe-only module, which is no longer required from this
+	   point, having given us a handle on (and increased the use
+	   count of) the actual driver code.
+	*/
+	module_put(drv->module);
+
+	return ret;
+}
+/*
+ * Destroy an MTD device which was created for a map device.
+ * Make sure the MTD device is already unregistered before calling this
+ */
+void map_destroy(struct mtd_info *mtd)
+{
+	struct map_info *map = mtd->priv;
+
+	if (map->fldrv->destroy)
+		map->fldrv->destroy(mtd);
+
+	module_put(map->fldrv->module);
+
+	kfree(mtd);
+}
+
+EXPORT_SYMBOL(register_mtd_chip_driver);
+EXPORT_SYMBOL(unregister_mtd_chip_driver);
+EXPORT_SYMBOL(do_map_probe);
+EXPORT_SYMBOL(map_destroy);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("David Woodhouse <dwmw2@infradead.org>");
+MODULE_DESCRIPTION("Core routines for registering and invoking MTD chip drivers");