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, 120 insertions, 0 deletions

diff --git a/drivers/reset/reset-socfpga.c b/drivers/reset/reset-socfpga.c
new file mode 100644
index 000000000..da208270b
--- /dev/null
+++ b/drivers/reset/reset-socfpga.c
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2018, Intel Corporation
+ * Copied from reset-sunxi.c
+ */
+
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/init.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/platform_device.h>
+#include <linux/reset-controller.h>
+#include <linux/reset/reset-simple.h>
+#include <linux/reset/socfpga.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+
+#define SOCFPGA_NR_BANKS	8
+
+static int a10_reset_init(struct device_node *np)
+{
+	struct reset_simple_data *data;
+	struct resource res;
+	resource_size_t size;
+	int ret;
+	u32 reg_offset = 0x10;
+
+	data = kzalloc_obj(*data);
+	if (!data)
+		return -ENOMEM;
+
+	ret = of_address_to_resource(np, 0, &res);
+	if (ret)
+		goto err_alloc;
+
+	size = resource_size(&res);
+	if (!request_mem_region(res.start, size, np->name)) {
+		ret = -EBUSY;
+		goto err_alloc;
+	}
+
+	data->membase = ioremap(res.start, size);
+	if (!data->membase) {
+		ret = -ENOMEM;
+		goto release_region;
+	}
+
+	if (of_property_read_u32(np, "altr,modrst-offset", &reg_offset))
+		pr_warn("missing altr,modrst-offset property, assuming 0x10\n");
+	data->membase += reg_offset;
+
+	spin_lock_init(&data->lock);
+
+	data->rcdev.owner = THIS_MODULE;
+	data->rcdev.nr_resets = SOCFPGA_NR_BANKS * 32;
+	data->rcdev.ops = &reset_simple_ops;
+	data->rcdev.of_node = np;
+	data->status_active_low = true;
+
+	ret = reset_controller_register(&data->rcdev);
+	if (ret)
+		pr_err("unable to register device\n");
+
+	return ret;
+
+release_region:
+	release_mem_region(res.start, size);
+
+err_alloc:
+	kfree(data);
+	return ret;
+};
+
+/*
+ * These are the reset controller we need to initialize early on in
+ * our system, before we can even think of using a regular device
+ * driver for it.
+ * The controllers that we can register through the regular device
+ * model are handled by the simple reset driver directly.
+ */
+static const struct of_device_id socfpga_early_reset_dt_ids[] __initconst = {
+	{ .compatible = "altr,rst-mgr", },
+	{ /* sentinel */ },
+};
+
+void __init socfpga_reset_init(void)
+{
+	struct device_node *np;
+
+	for_each_matching_node(np, socfpga_early_reset_dt_ids)
+		a10_reset_init(np);
+}
+
+/*
+ * The early driver is problematic, because it doesn't register
+ * itself as a driver. This causes certain device links to prevent
+ * consumer devices from probing. The hacky solution is to register
+ * an empty driver, whose only job is to attach itself to the reset
+ * manager and call probe.
+ */
+static const struct of_device_id socfpga_reset_dt_ids[] = {
+	{ .compatible = "altr,rst-mgr", },
+	{ /* sentinel */ },
+};
+
+static int reset_simple_probe(struct platform_device *pdev)
+{
+	return 0;
+}
+
+static struct platform_driver reset_socfpga_driver = {
+	.probe	= reset_simple_probe,
+	.driver = {
+		.name		= "socfpga-reset",
+		.of_match_table	= socfpga_reset_dt_ids,
+	},
+};
+builtin_platform_driver(reset_socfpga_driver);