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, 258 insertions, 0 deletions

diff --git a/drivers/phy/phy-can-transceiver.c b/drivers/phy/phy-can-transceiver.c
new file mode 100644
index 000000000..9aa306621
--- /dev/null
+++ b/drivers/phy/phy-can-transceiver.c
@@ -0,0 +1,258 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * phy-can-transceiver.c - phy driver for CAN transceivers
+ *
+ * Copyright (C) 2021 Texas Instruments Incorporated - https://www.ti.com
+ *
+ */
+#include <linux/gpio/consumer.h>
+#include <linux/phy/phy.h>
+#include <linux/platform_device.h>
+#include <linux/property.h>
+#include <linux/module.h>
+#include <linux/mux/consumer.h>
+
+struct can_transceiver_data {
+	u32 flags;
+#define CAN_TRANSCEIVER_STB_PRESENT	BIT(0)
+#define CAN_TRANSCEIVER_EN_PRESENT	BIT(1)
+#define CAN_TRANSCEIVER_DUAL_CH		BIT(2)
+#define CAN_TRANSCEIVER_SILENT_PRESENT	BIT(3)
+};
+
+struct can_transceiver_phy {
+	struct phy *generic_phy;
+	struct gpio_desc *silent_gpio;
+	struct gpio_desc *standby_gpio;
+	struct gpio_desc *enable_gpio;
+	struct can_transceiver_priv *priv;
+};
+
+struct can_transceiver_priv {
+	struct mux_state *mux_state;
+	int num_ch;
+	struct can_transceiver_phy can_transceiver_phy[] __counted_by(num_ch);
+};
+
+/* Power on function */
+static int can_transceiver_phy_power_on(struct phy *phy)
+{
+	struct can_transceiver_phy *can_transceiver_phy = phy_get_drvdata(phy);
+	struct can_transceiver_priv *priv = can_transceiver_phy->priv;
+	int ret;
+
+	if (priv->mux_state) {
+		ret = mux_state_select(priv->mux_state);
+		if (ret) {
+			dev_err(&phy->dev, "Failed to select CAN mux: %d\n", ret);
+			return ret;
+		}
+	}
+	gpiod_set_value_cansleep(can_transceiver_phy->silent_gpio, 0);
+	gpiod_set_value_cansleep(can_transceiver_phy->standby_gpio, 0);
+	gpiod_set_value_cansleep(can_transceiver_phy->enable_gpio, 1);
+
+	return 0;
+}
+
+/* Power off function */
+static int can_transceiver_phy_power_off(struct phy *phy)
+{
+	struct can_transceiver_phy *can_transceiver_phy = phy_get_drvdata(phy);
+	struct can_transceiver_priv *priv = can_transceiver_phy->priv;
+
+	gpiod_set_value_cansleep(can_transceiver_phy->silent_gpio, 1);
+	gpiod_set_value_cansleep(can_transceiver_phy->standby_gpio, 1);
+	gpiod_set_value_cansleep(can_transceiver_phy->enable_gpio, 0);
+	if (priv->mux_state)
+		mux_state_deselect(priv->mux_state);
+
+	return 0;
+}
+
+static const struct phy_ops can_transceiver_phy_ops = {
+	.power_on	= can_transceiver_phy_power_on,
+	.power_off	= can_transceiver_phy_power_off,
+	.owner		= THIS_MODULE,
+};
+
+static const struct can_transceiver_data tcan1042_drvdata = {
+	.flags = CAN_TRANSCEIVER_STB_PRESENT,
+};
+
+static const struct can_transceiver_data tcan1043_drvdata = {
+	.flags = CAN_TRANSCEIVER_STB_PRESENT | CAN_TRANSCEIVER_EN_PRESENT,
+};
+
+static const struct can_transceiver_data tja1048_drvdata = {
+	.flags = CAN_TRANSCEIVER_STB_PRESENT | CAN_TRANSCEIVER_DUAL_CH,
+};
+
+static const struct can_transceiver_data tja1051_drvdata = {
+	.flags = CAN_TRANSCEIVER_SILENT_PRESENT | CAN_TRANSCEIVER_EN_PRESENT,
+};
+
+static const struct can_transceiver_data tja1057_drvdata = {
+	.flags = CAN_TRANSCEIVER_SILENT_PRESENT,
+};
+
+static struct phy *can_transceiver_phy_xlate(struct device *dev,
+					     const struct of_phandle_args *args)
+{
+	struct can_transceiver_priv *priv = dev_get_drvdata(dev);
+	u32 idx;
+
+	if (priv->num_ch == 1)
+		return priv->can_transceiver_phy[0].generic_phy;
+
+	if (args->args_count != 1)
+		return ERR_PTR(-EINVAL);
+
+	idx = args->args[0];
+	if (idx >= priv->num_ch)
+		return ERR_PTR(-EINVAL);
+
+	return priv->can_transceiver_phy[idx].generic_phy;
+}
+
+static int can_transceiver_phy_probe(struct platform_device *pdev)
+{
+	struct phy_provider *phy_provider;
+	struct device *dev = &pdev->dev;
+	struct can_transceiver_phy *can_transceiver_phy;
+	struct can_transceiver_priv *priv;
+	const struct can_transceiver_data *drvdata;
+	struct phy *phy;
+	struct gpio_desc *silent_gpio;
+	struct gpio_desc *standby_gpio;
+	struct gpio_desc *enable_gpio;
+	struct mux_state *mux_state;
+	const char *propname;
+	int err, i, num_ch;
+	u32 max_bitrate;
+
+	drvdata = device_get_match_data(dev);
+	if (!drvdata)
+		return -ENODEV;
+
+	if (drvdata->flags & CAN_TRANSCEIVER_DUAL_CH)
+		num_ch = 2;
+	else
+		num_ch = 1;
+
+	priv = devm_kzalloc(dev, struct_size(priv, can_transceiver_phy, num_ch), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
+	priv->num_ch = num_ch;
+	platform_set_drvdata(pdev, priv);
+
+	mux_state = devm_mux_state_get_optional(dev, NULL);
+	if (IS_ERR(mux_state))
+		return PTR_ERR(mux_state);
+
+	priv->mux_state = mux_state;
+
+	propname = "max-bitrate";
+	if (device_property_present(dev, propname)) {
+		err = device_property_read_u32(dev, propname, &max_bitrate);
+		if (err)
+			return dev_err_probe(dev, err, "failed to parse %s\n", propname);
+
+		if (max_bitrate == 0)
+			dev_warn(dev, "Invalid value for transceiver max bitrate. Ignoring bitrate limit\n");
+	} else {
+		max_bitrate = 0;
+	}
+
+	for (i = 0; i < num_ch; i++) {
+		can_transceiver_phy = &priv->can_transceiver_phy[i];
+		can_transceiver_phy->priv = priv;
+
+		phy = devm_phy_create(dev, NULL, &can_transceiver_phy_ops);
+		if (IS_ERR(phy)) {
+			dev_err(dev, "failed to create can transceiver phy\n");
+			return PTR_ERR(phy);
+		}
+
+		phy->attrs.max_link_rate = max_bitrate;
+
+		can_transceiver_phy->generic_phy = phy;
+		can_transceiver_phy->priv = priv;
+
+		if (drvdata->flags & CAN_TRANSCEIVER_STB_PRESENT) {
+			standby_gpio = devm_gpiod_get_index_optional(dev, "standby", i,
+								     GPIOD_OUT_HIGH);
+			if (IS_ERR(standby_gpio))
+				return PTR_ERR(standby_gpio);
+			can_transceiver_phy->standby_gpio = standby_gpio;
+		}
+
+		if (drvdata->flags & CAN_TRANSCEIVER_EN_PRESENT) {
+			enable_gpio = devm_gpiod_get_index_optional(dev, "enable", i,
+								    GPIOD_OUT_LOW);
+			if (IS_ERR(enable_gpio))
+				return PTR_ERR(enable_gpio);
+			can_transceiver_phy->enable_gpio = enable_gpio;
+		}
+
+		if (drvdata->flags & CAN_TRANSCEIVER_SILENT_PRESENT) {
+			silent_gpio = devm_gpiod_get_index_optional(dev, "silent", i,
+								    GPIOD_OUT_HIGH);
+			if (IS_ERR(silent_gpio))
+				return PTR_ERR(silent_gpio);
+			can_transceiver_phy->silent_gpio = silent_gpio;
+		}
+
+		phy_set_drvdata(can_transceiver_phy->generic_phy, can_transceiver_phy);
+
+	}
+
+	phy_provider = devm_of_phy_provider_register(dev, can_transceiver_phy_xlate);
+
+	return PTR_ERR_OR_ZERO(phy_provider);
+}
+
+static const struct of_device_id can_transceiver_phy_ids[] = {
+	{
+		.compatible = "ti,tcan1042",
+		.data = &tcan1042_drvdata
+	},
+	{
+		.compatible = "ti,tcan1043",
+		.data = &tcan1043_drvdata
+	},
+	{
+		.compatible = "nxp,tja1048",
+		.data = &tja1048_drvdata
+	},
+	{
+		.compatible = "nxp,tja1051",
+		.data = &tja1051_drvdata
+	},
+	{
+		.compatible = "nxp,tja1057",
+		.data = &tja1057_drvdata
+	},
+	{
+		.compatible = "nxp,tjr1443",
+		.data = &tcan1043_drvdata
+	},
+	{ }
+};
+MODULE_DEVICE_TABLE(of, can_transceiver_phy_ids);
+
+static struct platform_driver can_transceiver_phy_driver = {
+	.probe = can_transceiver_phy_probe,
+	.driver = {
+		.name = "can-transceiver-phy",
+		.of_match_table = can_transceiver_phy_ids,
+	},
+};
+
+module_platform_driver(can_transceiver_phy_driver);
+
+MODULE_AUTHOR("Faiz Abbas <faiz_abbas@ti.com>");
+MODULE_AUTHOR("Aswath Govindraju <a-govindraju@ti.com>");
+MODULE_DESCRIPTION("CAN TRANSCEIVER PHY driver");
+MODULE_LICENSE("GPL v2");