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, 227 insertions, 0 deletions

diff --git a/drivers/net/phy/nxp-cbtx.c b/drivers/net/phy/nxp-cbtx.c
new file mode 100644
index 000000000..3286fcb4f
--- /dev/null
+++ b/drivers/net/phy/nxp-cbtx.c
@@ -0,0 +1,227 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Driver for 100BASE-TX PHY embedded into NXP SJA1110 switch
+ *
+ * Copyright 2022-2023 NXP
+ */
+
+#include <linux/kernel.h>
+#include <linux/mii.h>
+#include <linux/module.h>
+#include <linux/phy.h>
+
+#define PHY_ID_CBTX_SJA1110			0x001bb020
+
+/* Registers */
+#define  CBTX_MODE_CTRL_STAT			0x11
+#define  CBTX_PDOWN_CTRL			0x18
+#define  CBTX_RX_ERR_COUNTER			0x1a
+#define  CBTX_IRQ_STAT				0x1d
+#define  CBTX_IRQ_ENABLE			0x1e
+
+/* Fields */
+#define CBTX_MODE_CTRL_STAT_AUTO_MDIX_EN	BIT(7)
+#define CBTX_MODE_CTRL_STAT_MDIX_MODE		BIT(6)
+
+#define CBTX_PDOWN_CTL_TRUE_PDOWN		BIT(0)
+
+#define CBTX_IRQ_ENERGYON			BIT(7)
+#define CBTX_IRQ_AN_COMPLETE			BIT(6)
+#define CBTX_IRQ_REM_FAULT			BIT(5)
+#define CBTX_IRQ_LINK_DOWN			BIT(4)
+#define CBTX_IRQ_AN_LP_ACK			BIT(3)
+#define CBTX_IRQ_PARALLEL_DETECT_FAULT		BIT(2)
+#define CBTX_IRQ_AN_PAGE_RECV			BIT(1)
+
+static int cbtx_soft_reset(struct phy_device *phydev)
+{
+	int ret;
+
+	/* Can't soft reset unless we remove PHY from true power down mode */
+	ret = phy_clear_bits(phydev, CBTX_PDOWN_CTRL,
+			     CBTX_PDOWN_CTL_TRUE_PDOWN);
+	if (ret)
+		return ret;
+
+	return genphy_soft_reset(phydev);
+}
+
+static int cbtx_config_init(struct phy_device *phydev)
+{
+	/* Wait for cbtx_config_aneg() to kick in and apply this */
+	phydev->mdix_ctrl = ETH_TP_MDI_AUTO;
+
+	return 0;
+}
+
+static int cbtx_mdix_status(struct phy_device *phydev)
+{
+	int ret;
+
+	ret = phy_read(phydev, CBTX_MODE_CTRL_STAT);
+	if (ret < 0)
+		return ret;
+
+	if (ret & CBTX_MODE_CTRL_STAT_MDIX_MODE)
+		phydev->mdix = ETH_TP_MDI_X;
+	else
+		phydev->mdix = ETH_TP_MDI;
+
+	return 0;
+}
+
+static int cbtx_read_status(struct phy_device *phydev)
+{
+	int ret;
+
+	ret = cbtx_mdix_status(phydev);
+	if (ret)
+		return ret;
+
+	return genphy_read_status(phydev);
+}
+
+static int cbtx_mdix_config(struct phy_device *phydev)
+{
+	int ret;
+
+	switch (phydev->mdix_ctrl) {
+	case ETH_TP_MDI_AUTO:
+		return phy_set_bits(phydev, CBTX_MODE_CTRL_STAT,
+				    CBTX_MODE_CTRL_STAT_AUTO_MDIX_EN);
+	case ETH_TP_MDI:
+		ret = phy_clear_bits(phydev, CBTX_MODE_CTRL_STAT,
+				     CBTX_MODE_CTRL_STAT_AUTO_MDIX_EN);
+		if (ret)
+			return ret;
+
+		return phy_clear_bits(phydev, CBTX_MODE_CTRL_STAT,
+				      CBTX_MODE_CTRL_STAT_MDIX_MODE);
+	case ETH_TP_MDI_X:
+		ret = phy_clear_bits(phydev, CBTX_MODE_CTRL_STAT,
+				     CBTX_MODE_CTRL_STAT_AUTO_MDIX_EN);
+		if (ret)
+			return ret;
+
+		return phy_set_bits(phydev, CBTX_MODE_CTRL_STAT,
+				    CBTX_MODE_CTRL_STAT_MDIX_MODE);
+	}
+
+	return 0;
+}
+
+static int cbtx_config_aneg(struct phy_device *phydev)
+{
+	int ret;
+
+	ret = cbtx_mdix_config(phydev);
+	if (ret)
+		return ret;
+
+	return genphy_config_aneg(phydev);
+}
+
+static int cbtx_ack_interrupts(struct phy_device *phydev)
+{
+	return phy_read(phydev, CBTX_IRQ_STAT);
+}
+
+static int cbtx_config_intr(struct phy_device *phydev)
+{
+	int ret;
+
+	if (phydev->interrupts == PHY_INTERRUPT_ENABLED) {
+		ret = cbtx_ack_interrupts(phydev);
+		if (ret < 0)
+			return ret;
+
+		ret = phy_write(phydev, CBTX_IRQ_ENABLE, CBTX_IRQ_LINK_DOWN |
+				CBTX_IRQ_AN_COMPLETE | CBTX_IRQ_ENERGYON);
+		if (ret)
+			return ret;
+	} else {
+		ret = phy_write(phydev, CBTX_IRQ_ENABLE, 0);
+		if (ret)
+			return ret;
+
+		ret = cbtx_ack_interrupts(phydev);
+		if (ret < 0)
+			return ret;
+	}
+
+	return 0;
+}
+
+static irqreturn_t cbtx_handle_interrupt(struct phy_device *phydev)
+{
+	int irq_stat, irq_enabled;
+
+	irq_stat = cbtx_ack_interrupts(phydev);
+	if (irq_stat < 0) {
+		phy_error(phydev);
+		return IRQ_NONE;
+	}
+
+	irq_enabled = phy_read(phydev, CBTX_IRQ_ENABLE);
+	if (irq_enabled < 0) {
+		phy_error(phydev);
+		return IRQ_NONE;
+	}
+
+	if (!(irq_enabled & irq_stat))
+		return IRQ_NONE;
+
+	phy_trigger_machine(phydev);
+
+	return IRQ_HANDLED;
+}
+
+static int cbtx_get_sset_count(struct phy_device *phydev)
+{
+	return 1;
+}
+
+static void cbtx_get_strings(struct phy_device *phydev, u8 *data)
+{
+	ethtool_puts(&data, "100btx_rx_err");
+}
+
+static void cbtx_get_stats(struct phy_device *phydev,
+			   struct ethtool_stats *stats, u64 *data)
+{
+	int ret;
+
+	ret = phy_read(phydev, CBTX_RX_ERR_COUNTER);
+	data[0] = (ret < 0) ? U64_MAX : ret;
+}
+
+static struct phy_driver cbtx_driver[] = {
+	{
+		PHY_ID_MATCH_MODEL(PHY_ID_CBTX_SJA1110),
+		.name			= "NXP CBTX (SJA1110)",
+		/* PHY_BASIC_FEATURES */
+		.soft_reset		= cbtx_soft_reset,
+		.config_init		= cbtx_config_init,
+		.suspend		= genphy_suspend,
+		.resume			= genphy_resume,
+		.config_intr		= cbtx_config_intr,
+		.handle_interrupt	= cbtx_handle_interrupt,
+		.read_status		= cbtx_read_status,
+		.config_aneg		= cbtx_config_aneg,
+		.get_sset_count		= cbtx_get_sset_count,
+		.get_strings		= cbtx_get_strings,
+		.get_stats		= cbtx_get_stats,
+	},
+};
+
+module_phy_driver(cbtx_driver);
+
+static const struct mdio_device_id __maybe_unused cbtx_tbl[] = {
+	{ PHY_ID_MATCH_MODEL(PHY_ID_CBTX_SJA1110) },
+	{ },
+};
+
+MODULE_DEVICE_TABLE(mdio, cbtx_tbl);
+
+MODULE_AUTHOR("Vladimir Oltean <vladimir.oltean@nxp.com>");
+MODULE_DESCRIPTION("NXP CBTX PHY driver");
+MODULE_LICENSE("GPL");