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

3 files changed, 81 insertions, 0 deletions

diff --git a/drivers/reset/tegra/Kconfig b/drivers/reset/tegra/Kconfig
new file mode 100644
index 000000000..4a2d26d12
--- /dev/null
+++ b/drivers/reset/tegra/Kconfig
@@ -0,0 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config RESET_TEGRA_BPMP
+	bool "Tegra BPMP Reset Driver" if COMPILE_TEST
+	default TEGRA_BPMP
diff --git a/drivers/reset/tegra/Makefile b/drivers/reset/tegra/Makefile
new file mode 100644
index 000000000..eccba896a
--- /dev/null
+++ b/drivers/reset/tegra/Makefile
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+obj-$(CONFIG_RESET_TEGRA_BPMP) += reset-bpmp.o
diff --git a/drivers/reset/tegra/reset-bpmp.c b/drivers/reset/tegra/reset-bpmp.c
new file mode 100644
index 000000000..4c5bba52b
--- /dev/null
+++ b/drivers/reset/tegra/reset-bpmp.c
@@ -0,0 +1,75 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2016 NVIDIA Corporation
+ */
+
+#include <linux/reset-controller.h>
+
+#include <soc/tegra/bpmp.h>
+#include <soc/tegra/bpmp-abi.h>
+
+static struct tegra_bpmp *to_tegra_bpmp(struct reset_controller_dev *rstc)
+{
+	return container_of(rstc, struct tegra_bpmp, rstc);
+}
+
+static int tegra_bpmp_reset_common(struct reset_controller_dev *rstc,
+				   enum mrq_reset_commands command,
+				   unsigned int id)
+{
+	struct tegra_bpmp *bpmp = to_tegra_bpmp(rstc);
+	struct mrq_reset_request request;
+	struct tegra_bpmp_message msg;
+	int err;
+
+	memset(&request, 0, sizeof(request));
+	request.cmd = command;
+	request.reset_id = id;
+
+	memset(&msg, 0, sizeof(msg));
+	msg.mrq = MRQ_RESET;
+	msg.tx.data = &request;
+	msg.tx.size = sizeof(request);
+
+	err = tegra_bpmp_transfer(bpmp, &msg);
+	if (err)
+		return err;
+	if (msg.rx.ret)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int tegra_bpmp_reset_module(struct reset_controller_dev *rstc,
+				   unsigned long id)
+{
+	return tegra_bpmp_reset_common(rstc, CMD_RESET_MODULE, id);
+}
+
+static int tegra_bpmp_reset_assert(struct reset_controller_dev *rstc,
+				   unsigned long id)
+{
+	return tegra_bpmp_reset_common(rstc, CMD_RESET_ASSERT, id);
+}
+
+static int tegra_bpmp_reset_deassert(struct reset_controller_dev *rstc,
+				     unsigned long id)
+{
+	return tegra_bpmp_reset_common(rstc, CMD_RESET_DEASSERT, id);
+}
+
+static const struct reset_control_ops tegra_bpmp_reset_ops = {
+	.reset = tegra_bpmp_reset_module,
+	.assert = tegra_bpmp_reset_assert,
+	.deassert = tegra_bpmp_reset_deassert,
+};
+
+int tegra_bpmp_init_resets(struct tegra_bpmp *bpmp)
+{
+	bpmp->rstc.ops = &tegra_bpmp_reset_ops;
+	bpmp->rstc.owner = THIS_MODULE;
+	bpmp->rstc.of_node = bpmp->dev->of_node;
+	bpmp->rstc.nr_resets = bpmp->soc->num_resets;
+
+	return devm_reset_controller_register(bpmp->dev, &bpmp->rstc);
+}