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, 135 insertions, 0 deletions

diff --git a/drivers/clk/sprd/gate.c b/drivers/clk/sprd/gate.c
new file mode 100644
index 000000000..56e1714b5
--- /dev/null
+++ b/drivers/clk/sprd/gate.c
@@ -0,0 +1,135 @@
+// SPDX-License-Identifier: GPL-2.0
+//
+// Spreadtrum gate clock driver
+//
+// Copyright (C) 2017 Spreadtrum, Inc.
+// Author: Chunyan Zhang <chunyan.zhang@spreadtrum.com>
+
+#include <linux/clk-provider.h>
+#include <linux/regmap.h>
+
+#include "gate.h"
+
+static void clk_gate_toggle(const struct sprd_gate *sg, bool en)
+{
+	const struct sprd_clk_common *common = &sg->common;
+	unsigned int reg;
+	bool set = sg->flags & CLK_GATE_SET_TO_DISABLE ? true : false;
+
+	set ^= en;
+
+	regmap_read(common->regmap, common->reg, &reg);
+
+	if (set)
+		reg |= sg->enable_mask;
+	else
+		reg &= ~sg->enable_mask;
+
+	regmap_write(common->regmap, common->reg, reg);
+}
+
+static void clk_sc_gate_toggle(const struct sprd_gate *sg, bool en)
+{
+	const struct sprd_clk_common *common = &sg->common;
+	bool set = sg->flags & CLK_GATE_SET_TO_DISABLE ? 1 : 0;
+	unsigned int offset;
+
+	set ^= en;
+
+	/*
+	 * Each set/clear gate clock has three registers:
+	 * common->reg			- base register
+	 * common->reg + offset		- set register
+	 * common->reg + 2 * offset	- clear register
+	 */
+	offset = set ? sg->sc_offset : sg->sc_offset * 2;
+
+	regmap_write(common->regmap, common->reg + offset,
+			  sg->enable_mask);
+}
+
+static void sprd_gate_disable(struct clk_hw *hw)
+{
+	struct sprd_gate *sg = hw_to_sprd_gate(hw);
+
+	clk_gate_toggle(sg, false);
+}
+
+static int sprd_gate_enable(struct clk_hw *hw)
+{
+	struct sprd_gate *sg = hw_to_sprd_gate(hw);
+
+	clk_gate_toggle(sg, true);
+
+	return 0;
+}
+
+static void sprd_sc_gate_disable(struct clk_hw *hw)
+{
+	struct sprd_gate *sg = hw_to_sprd_gate(hw);
+
+	clk_sc_gate_toggle(sg, false);
+}
+
+static int sprd_sc_gate_enable(struct clk_hw *hw)
+{
+	struct sprd_gate *sg = hw_to_sprd_gate(hw);
+
+	clk_sc_gate_toggle(sg, true);
+
+	return 0;
+}
+
+static int sprd_pll_sc_gate_prepare(struct clk_hw *hw)
+{
+	struct sprd_gate *sg = hw_to_sprd_gate(hw);
+
+	clk_sc_gate_toggle(sg, true);
+	udelay(sg->udelay);
+
+	return 0;
+}
+
+static int sprd_gate_is_enabled(struct clk_hw *hw)
+{
+	struct sprd_gate *sg = hw_to_sprd_gate(hw);
+	struct sprd_clk_common *common = &sg->common;
+	struct clk_hw *parent;
+	unsigned int reg;
+
+	if (sg->flags & SPRD_GATE_NON_AON) {
+		parent = clk_hw_get_parent(hw);
+		if (!parent || !clk_hw_is_enabled(parent))
+			return 0;
+	}
+
+	regmap_read(common->regmap, common->reg, &reg);
+
+	if (sg->flags & CLK_GATE_SET_TO_DISABLE)
+		reg ^= sg->enable_mask;
+
+	reg &= sg->enable_mask;
+
+	return reg ? 1 : 0;
+}
+
+const struct clk_ops sprd_gate_ops = {
+	.disable	= sprd_gate_disable,
+	.enable		= sprd_gate_enable,
+	.is_enabled	= sprd_gate_is_enabled,
+};
+EXPORT_SYMBOL_GPL(sprd_gate_ops);
+
+const struct clk_ops sprd_sc_gate_ops = {
+	.disable	= sprd_sc_gate_disable,
+	.enable		= sprd_sc_gate_enable,
+	.is_enabled	= sprd_gate_is_enabled,
+};
+EXPORT_SYMBOL_GPL(sprd_sc_gate_ops);
+
+const struct clk_ops sprd_pll_sc_gate_ops = {
+	.unprepare	= sprd_sc_gate_disable,
+	.prepare	= sprd_pll_sc_gate_prepare,
+	.is_enabled	= sprd_gate_is_enabled,
+};
+EXPORT_SYMBOL_GPL(sprd_pll_sc_gate_ops);