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, 206 insertions, 0 deletions

diff --git a/drivers/reset/reset-sunplus.c b/drivers/reset/reset-sunplus.c
new file mode 100644
index 000000000..c136a1f07
--- /dev/null
+++ b/drivers/reset/reset-sunplus.c
@@ -0,0 +1,206 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+/*
+ * SP7021 reset driver
+ *
+ * Copyright (C) Sunplus Technology Co., Ltd.
+ *       All rights reserved.
+ */
+
+#include <linux/io.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+#include <linux/reset-controller.h>
+#include <linux/reboot.h>
+
+/* HIWORD_MASK_REG BITS */
+#define BITS_PER_HWM_REG	16
+
+/* resets HW info: reg_index_shift */
+static const u32 sp_resets[] = {
+/* SP7021: mo_reset0 ~ mo_reset9 */
+	0x00,
+	0x02,
+	0x03,
+	0x04,
+	0x05,
+	0x06,
+	0x07,
+	0x08,
+	0x09,
+	0x0a,
+	0x0b,
+	0x0d,
+	0x0e,
+	0x0f,
+	0x10,
+	0x12,
+	0x14,
+	0x15,
+	0x16,
+	0x17,
+	0x18,
+	0x19,
+	0x1a,
+	0x1b,
+	0x1c,
+	0x1d,
+	0x1e,
+	0x1f,
+	0x20,
+	0x21,
+	0x22,
+	0x23,
+	0x24,
+	0x25,
+	0x26,
+	0x2a,
+	0x2b,
+	0x2d,
+	0x2e,
+	0x30,
+	0x31,
+	0x32,
+	0x33,
+	0x3d,
+	0x3e,
+	0x3f,
+	0x42,
+	0x44,
+	0x4b,
+	0x4c,
+	0x4d,
+	0x4e,
+	0x4f,
+	0x50,
+	0x55,
+	0x60,
+	0x61,
+	0x6a,
+	0x6f,
+	0x70,
+	0x73,
+	0x74,
+	0x86,
+	0x8a,
+	0x8b,
+	0x8d,
+	0x8e,
+	0x8f,
+	0x90,
+	0x92,
+	0x93,
+	0x94,
+	0x95,
+	0x96,
+	0x97,
+	0x98,
+	0x99,
+};
+
+struct sp_reset {
+	struct reset_controller_dev rcdev;
+	void __iomem *base;
+};
+
+static inline struct sp_reset *to_sp_reset(struct reset_controller_dev *rcdev)
+{
+	return container_of(rcdev, struct sp_reset, rcdev);
+}
+
+static int sp_reset_update(struct reset_controller_dev *rcdev,
+			   unsigned long id, bool assert)
+{
+	struct sp_reset *reset = to_sp_reset(rcdev);
+	int index = sp_resets[id] / BITS_PER_HWM_REG;
+	int shift = sp_resets[id] % BITS_PER_HWM_REG;
+	u32 val;
+
+	val = (1 << (16 + shift)) | (assert << shift);
+	writel(val, reset->base + (index * 4));
+
+	return 0;
+}
+
+static int sp_reset_assert(struct reset_controller_dev *rcdev,
+			   unsigned long id)
+{
+	return sp_reset_update(rcdev, id, true);
+}
+
+static int sp_reset_deassert(struct reset_controller_dev *rcdev,
+			     unsigned long id)
+{
+	return sp_reset_update(rcdev, id, false);
+}
+
+static int sp_reset_status(struct reset_controller_dev *rcdev,
+			   unsigned long id)
+{
+	struct sp_reset *reset = to_sp_reset(rcdev);
+	int index = sp_resets[id] / BITS_PER_HWM_REG;
+	int shift = sp_resets[id] % BITS_PER_HWM_REG;
+	u32 reg;
+
+	reg = readl(reset->base + (index * 4));
+
+	return !!(reg & BIT(shift));
+}
+
+static const struct reset_control_ops sp_reset_ops = {
+	.assert   = sp_reset_assert,
+	.deassert = sp_reset_deassert,
+	.status   = sp_reset_status,
+};
+
+static int sp_restart(struct sys_off_data *data)
+{
+	struct sp_reset *reset = data->cb_data;
+
+	sp_reset_assert(&reset->rcdev, 0);
+	sp_reset_deassert(&reset->rcdev, 0);
+
+	return NOTIFY_DONE;
+}
+
+static int sp_reset_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct sp_reset *reset;
+	struct resource *res;
+	int ret;
+
+	reset = devm_kzalloc(dev, sizeof(*reset), GFP_KERNEL);
+	if (!reset)
+		return -ENOMEM;
+
+	reset->base = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
+	if (IS_ERR(reset->base))
+		return PTR_ERR(reset->base);
+
+	reset->rcdev.ops = &sp_reset_ops;
+	reset->rcdev.owner = THIS_MODULE;
+	reset->rcdev.of_node = dev->of_node;
+	reset->rcdev.nr_resets = resource_size(res) / 4 * BITS_PER_HWM_REG;
+
+	ret = devm_reset_controller_register(dev, &reset->rcdev);
+	if (ret)
+		return ret;
+
+	return devm_register_sys_off_handler(&pdev->dev, SYS_OFF_MODE_RESTART,
+					     192, sp_restart, reset);
+}
+
+static const struct of_device_id sp_reset_dt_ids[] = {
+	{.compatible = "sunplus,sp7021-reset",},
+	{ /* sentinel */ },
+};
+
+static struct platform_driver sp_reset_driver = {
+	.probe = sp_reset_probe,
+	.driver = {
+		.name			= "sunplus-reset",
+		.of_match_table		= sp_reset_dt_ids,
+		.suppress_bind_attrs	= true,
+	},
+};
+builtin_platform_driver(sp_reset_driver);