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/arch/mips/bcm63xx/timer.c b/arch/mips/bcm63xx/timer.c
new file mode 100644
index 000000000..74b83807d
--- /dev/null
+++ b/arch/mips/bcm63xx/timer.c
@@ -0,0 +1,206 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2008 Maxime Bizon <mbizon@freebox.fr>
+ */
+
+#include <linux/kernel.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/export.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <bcm63xx_cpu.h>
+#include <bcm63xx_io.h>
+#include <bcm63xx_timer.h>
+#include <bcm63xx_regs.h>
+
+static DEFINE_RAW_SPINLOCK(timer_reg_lock);
+static DEFINE_RAW_SPINLOCK(timer_data_lock);
+static struct clk *periph_clk;
+
+static struct timer_data {
+	void	(*cb)(void *);
+	void	*data;
+} timer_data[BCM63XX_TIMER_COUNT];
+
+static irqreturn_t timer_interrupt(int irq, void *dev_id)
+{
+	u32 stat;
+	int i;
+
+	raw_spin_lock(&timer_reg_lock);
+	stat = bcm_timer_readl(TIMER_IRQSTAT_REG);
+	bcm_timer_writel(stat, TIMER_IRQSTAT_REG);
+	raw_spin_unlock(&timer_reg_lock);
+
+	for (i = 0; i < BCM63XX_TIMER_COUNT; i++) {
+		if (!(stat & TIMER_IRQSTAT_TIMER_CAUSE(i)))
+			continue;
+
+		raw_spin_lock(&timer_data_lock);
+		if (!timer_data[i].cb) {
+			raw_spin_unlock(&timer_data_lock);
+			continue;
+		}
+
+		timer_data[i].cb(timer_data[i].data);
+		raw_spin_unlock(&timer_data_lock);
+	}
+
+	return IRQ_HANDLED;
+}
+
+int bcm63xx_timer_enable(int id)
+{
+	u32 reg;
+	unsigned long flags;
+
+	if (id >= BCM63XX_TIMER_COUNT)
+		return -EINVAL;
+
+	raw_spin_lock_irqsave(&timer_reg_lock, flags);
+
+	reg = bcm_timer_readl(TIMER_CTLx_REG(id));
+	reg |= TIMER_CTL_ENABLE_MASK;
+	bcm_timer_writel(reg, TIMER_CTLx_REG(id));
+
+	reg = bcm_timer_readl(TIMER_IRQSTAT_REG);
+	reg |= TIMER_IRQSTAT_TIMER_IR_EN(id);
+	bcm_timer_writel(reg, TIMER_IRQSTAT_REG);
+
+	raw_spin_unlock_irqrestore(&timer_reg_lock, flags);
+	return 0;
+}
+
+EXPORT_SYMBOL(bcm63xx_timer_enable);
+
+int bcm63xx_timer_disable(int id)
+{
+	u32 reg;
+	unsigned long flags;
+
+	if (id >= BCM63XX_TIMER_COUNT)
+		return -EINVAL;
+
+	raw_spin_lock_irqsave(&timer_reg_lock, flags);
+
+	reg = bcm_timer_readl(TIMER_CTLx_REG(id));
+	reg &= ~TIMER_CTL_ENABLE_MASK;
+	bcm_timer_writel(reg, TIMER_CTLx_REG(id));
+
+	reg = bcm_timer_readl(TIMER_IRQSTAT_REG);
+	reg &= ~TIMER_IRQSTAT_TIMER_IR_EN(id);
+	bcm_timer_writel(reg, TIMER_IRQSTAT_REG);
+
+	raw_spin_unlock_irqrestore(&timer_reg_lock, flags);
+	return 0;
+}
+
+EXPORT_SYMBOL(bcm63xx_timer_disable);
+
+int bcm63xx_timer_register(int id, void (*callback)(void *data), void *data)
+{
+	unsigned long flags;
+	int ret;
+
+	if (id >= BCM63XX_TIMER_COUNT || !callback)
+		return -EINVAL;
+
+	ret = 0;
+	raw_spin_lock_irqsave(&timer_data_lock, flags);
+	if (timer_data[id].cb) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	timer_data[id].cb = callback;
+	timer_data[id].data = data;
+
+out:
+	raw_spin_unlock_irqrestore(&timer_data_lock, flags);
+	return ret;
+}
+
+EXPORT_SYMBOL(bcm63xx_timer_register);
+
+void bcm63xx_timer_unregister(int id)
+{
+	unsigned long flags;
+
+	if (id >= BCM63XX_TIMER_COUNT)
+		return;
+
+	raw_spin_lock_irqsave(&timer_data_lock, flags);
+	timer_data[id].cb = NULL;
+	raw_spin_unlock_irqrestore(&timer_data_lock, flags);
+}
+
+EXPORT_SYMBOL(bcm63xx_timer_unregister);
+
+unsigned int bcm63xx_timer_countdown(unsigned int countdown_us)
+{
+	return (clk_get_rate(periph_clk) / (1000 * 1000)) * countdown_us;
+}
+
+EXPORT_SYMBOL(bcm63xx_timer_countdown);
+
+int bcm63xx_timer_set(int id, int monotonic, unsigned int countdown_us)
+{
+	u32 reg, countdown;
+	unsigned long flags;
+
+	if (id >= BCM63XX_TIMER_COUNT)
+		return -EINVAL;
+
+	countdown = bcm63xx_timer_countdown(countdown_us);
+	if (countdown & ~TIMER_CTL_COUNTDOWN_MASK)
+		return -EINVAL;
+
+	raw_spin_lock_irqsave(&timer_reg_lock, flags);
+	reg = bcm_timer_readl(TIMER_CTLx_REG(id));
+
+	if (monotonic)
+		reg &= ~TIMER_CTL_MONOTONIC_MASK;
+	else
+		reg |= TIMER_CTL_MONOTONIC_MASK;
+
+	reg &= ~TIMER_CTL_COUNTDOWN_MASK;
+	reg |= countdown;
+	bcm_timer_writel(reg, TIMER_CTLx_REG(id));
+
+	raw_spin_unlock_irqrestore(&timer_reg_lock, flags);
+	return 0;
+}
+
+EXPORT_SYMBOL(bcm63xx_timer_set);
+
+static int bcm63xx_timer_init(void)
+{
+	int ret, irq;
+	u32 reg;
+
+	reg = bcm_timer_readl(TIMER_IRQSTAT_REG);
+	reg &= ~TIMER_IRQSTAT_TIMER0_IR_EN;
+	reg &= ~TIMER_IRQSTAT_TIMER1_IR_EN;
+	reg &= ~TIMER_IRQSTAT_TIMER2_IR_EN;
+	bcm_timer_writel(reg, TIMER_IRQSTAT_REG);
+
+	periph_clk = clk_get(NULL, "periph");
+	if (IS_ERR(periph_clk))
+		return -ENODEV;
+
+	irq = bcm63xx_get_irq_number(IRQ_TIMER);
+	ret = request_irq(irq, timer_interrupt, 0, "bcm63xx_timer", NULL);
+	if (ret) {
+		pr_err("%s: failed to register irq\n", __func__);
+		return ret;
+	}
+
+	return 0;
+}
+
+arch_initcall(bcm63xx_timer_init);