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, 221 insertions, 0 deletions

diff --git a/drivers/input/input-poller.c b/drivers/input/input-poller.c
new file mode 100644
index 000000000..54dc07fca
--- /dev/null
+++ b/drivers/input/input-poller.c
@@ -0,0 +1,221 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Support for polling mode for input devices.
+ */
+
+#include <linux/device.h>
+#include <linux/export.h>
+#include <linux/input.h>
+#include <linux/jiffies.h>
+#include <linux/mutex.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+#include "input-poller.h"
+
+struct input_dev_poller {
+	void (*poll)(struct input_dev *dev);
+
+	unsigned int poll_interval; /* msec */
+	unsigned int poll_interval_max; /* msec */
+	unsigned int poll_interval_min; /* msec */
+
+	struct input_dev *input;
+	struct delayed_work work;
+};
+
+static void input_dev_poller_queue_work(struct input_dev_poller *poller)
+{
+	unsigned long delay;
+
+	delay = msecs_to_jiffies(poller->poll_interval);
+	if (delay >= HZ)
+		delay = round_jiffies_relative(delay);
+
+	queue_delayed_work(system_freezable_wq, &poller->work, delay);
+}
+
+static void input_dev_poller_work(struct work_struct *work)
+{
+	struct input_dev_poller *poller =
+		container_of(work, struct input_dev_poller, work.work);
+
+	poller->poll(poller->input);
+	input_dev_poller_queue_work(poller);
+}
+
+void input_dev_poller_finalize(struct input_dev_poller *poller)
+{
+	if (!poller->poll_interval)
+		poller->poll_interval = 500;
+	if (!poller->poll_interval_max)
+		poller->poll_interval_max = poller->poll_interval;
+}
+
+void input_dev_poller_start(struct input_dev_poller *poller)
+{
+	/* Only start polling if polling is enabled */
+	if (poller->poll_interval > 0) {
+		poller->poll(poller->input);
+		input_dev_poller_queue_work(poller);
+	}
+}
+
+void input_dev_poller_stop(struct input_dev_poller *poller)
+{
+	cancel_delayed_work_sync(&poller->work);
+}
+
+int input_setup_polling(struct input_dev *dev,
+			void (*poll_fn)(struct input_dev *dev))
+{
+	struct input_dev_poller *poller;
+
+	poller = kzalloc_obj(*poller);
+	if (!poller) {
+		/*
+		 * We want to show message even though kzalloc() may have
+		 * printed backtrace as knowing what instance of input
+		 * device we were dealing with is helpful.
+		 */
+		dev_err(dev->dev.parent ?: &dev->dev,
+			"%s: unable to allocate poller structure\n", __func__);
+		return -ENOMEM;
+	}
+
+	INIT_DELAYED_WORK(&poller->work, input_dev_poller_work);
+	poller->input = dev;
+	poller->poll = poll_fn;
+
+	dev->poller = poller;
+	return 0;
+}
+EXPORT_SYMBOL(input_setup_polling);
+
+static bool input_dev_ensure_poller(struct input_dev *dev)
+{
+	if (!dev->poller) {
+		dev_err(dev->dev.parent ?: &dev->dev,
+			"poller structure has not been set up\n");
+		return false;
+	}
+
+	return true;
+}
+
+void input_set_poll_interval(struct input_dev *dev, unsigned int interval)
+{
+	if (input_dev_ensure_poller(dev))
+		dev->poller->poll_interval = interval;
+}
+EXPORT_SYMBOL(input_set_poll_interval);
+
+void input_set_min_poll_interval(struct input_dev *dev, unsigned int interval)
+{
+	if (input_dev_ensure_poller(dev))
+		dev->poller->poll_interval_min = interval;
+}
+EXPORT_SYMBOL(input_set_min_poll_interval);
+
+void input_set_max_poll_interval(struct input_dev *dev, unsigned int interval)
+{
+	if (input_dev_ensure_poller(dev))
+		dev->poller->poll_interval_max = interval;
+}
+EXPORT_SYMBOL(input_set_max_poll_interval);
+
+int input_get_poll_interval(struct input_dev *dev)
+{
+	if (!dev->poller)
+		return -EINVAL;
+
+	return dev->poller->poll_interval;
+}
+EXPORT_SYMBOL(input_get_poll_interval);
+
+/* SYSFS interface */
+
+static ssize_t input_dev_get_poll_interval(struct device *dev,
+					   struct device_attribute *attr,
+					   char *buf)
+{
+	struct input_dev *input = to_input_dev(dev);
+
+	return sprintf(buf, "%d\n", input->poller->poll_interval);
+}
+
+static ssize_t input_dev_set_poll_interval(struct device *dev,
+					   struct device_attribute *attr,
+					   const char *buf, size_t count)
+{
+	struct input_dev *input = to_input_dev(dev);
+	struct input_dev_poller *poller = input->poller;
+	unsigned int interval;
+	int err;
+
+	err = kstrtouint(buf, 0, &interval);
+	if (err)
+		return err;
+
+	if (interval < poller->poll_interval_min)
+		return -EINVAL;
+
+	if (interval > poller->poll_interval_max)
+		return -EINVAL;
+
+	guard(mutex)(&input->mutex);
+
+	poller->poll_interval = interval;
+
+	if (input_device_enabled(input)) {
+		cancel_delayed_work_sync(&poller->work);
+		if (poller->poll_interval > 0)
+			input_dev_poller_queue_work(poller);
+	}
+
+	return count;
+}
+
+static DEVICE_ATTR(poll, 0644,
+		   input_dev_get_poll_interval, input_dev_set_poll_interval);
+
+static ssize_t input_dev_get_poll_max(struct device *dev,
+				      struct device_attribute *attr, char *buf)
+{
+	struct input_dev *input = to_input_dev(dev);
+
+	return sprintf(buf, "%d\n", input->poller->poll_interval_max);
+}
+
+static DEVICE_ATTR(max, 0444, input_dev_get_poll_max, NULL);
+
+static ssize_t input_dev_get_poll_min(struct device *dev,
+				     struct device_attribute *attr, char *buf)
+{
+	struct input_dev *input = to_input_dev(dev);
+
+	return sprintf(buf, "%d\n", input->poller->poll_interval_min);
+}
+
+static DEVICE_ATTR(min, 0444, input_dev_get_poll_min, NULL);
+
+static umode_t input_poller_attrs_visible(struct kobject *kobj,
+					  struct attribute *attr, int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct input_dev *input = to_input_dev(dev);
+
+	return input->poller ? attr->mode : 0;
+}
+
+static struct attribute *input_poller_attrs[] = {
+	&dev_attr_poll.attr,
+	&dev_attr_max.attr,
+	&dev_attr_min.attr,
+	NULL
+};
+
+struct attribute_group input_poller_attribute_group = {
+	.is_visible	= input_poller_attrs_visible,
+	.attrs		= input_poller_attrs,
+};