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, 200 insertions, 0 deletions

diff --git a/drivers/mtd/maps/physmap-gemini.c b/drivers/mtd/maps/physmap-gemini.c
new file mode 100644
index 000000000..1c34b4ef7
--- /dev/null
+++ b/drivers/mtd/maps/physmap-gemini.c
@@ -0,0 +1,200 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Cortina Systems Gemini OF physmap add-on
+ * Copyright (C) 2017 Linus Walleij <linus.walleij@linaro.org>
+ *
+ * This SoC has an elaborate flash control register, so we need to
+ * detect and set it up when booting on this platform.
+ */
+#include <linux/export.h>
+#include <linux/of.h>
+#include <linux/mtd/map.h>
+#include <linux/mtd/xip.h>
+#include <linux/mfd/syscon.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+#include <linux/bitops.h>
+#include <linux/pinctrl/consumer.h>
+#include "physmap-gemini.h"
+
+/*
+ * The Flash-relevant parts of the global status register
+ * These would also be relevant for a NAND driver.
+ */
+#define GLOBAL_STATUS			0x04
+#define FLASH_TYPE_MASK			(0x3 << 24)
+#define FLASH_TYPE_NAND_2K		(0x3 << 24)
+#define FLASH_TYPE_NAND_512		(0x2 << 24)
+#define FLASH_TYPE_PARALLEL		(0x1 << 24)
+#define FLASH_TYPE_SERIAL		(0x0 << 24)
+/* if parallel */
+#define FLASH_WIDTH_16BIT		(1 << 23)	/* else 8 bit */
+/* if serial */
+#define FLASH_ATMEL			(1 << 23)	/* else STM */
+
+#define FLASH_SIZE_MASK			(0x3 << 21)
+#define NAND_256M			(0x3 << 21)	/* and more */
+#define NAND_128M			(0x2 << 21)
+#define NAND_64M			(0x1 << 21)
+#define NAND_32M			(0x0 << 21)
+#define ATMEL_16M			(0x3 << 21)	/* and more */
+#define ATMEL_8M			(0x2 << 21)
+#define ATMEL_4M_2M			(0x1 << 21)
+#define ATMEL_1M			(0x0 << 21)	/* and less */
+#define STM_32M				(1 << 22)	/* and more */
+#define STM_16M				(0 << 22)	/* and less */
+
+#define FLASH_PARALLEL_HIGH_PIN_CNT	(1 << 20)	/* else low pin cnt */
+
+struct gemini_flash {
+	struct device *dev;
+	struct pinctrl *p;
+	struct pinctrl_state *enabled_state;
+	struct pinctrl_state *disabled_state;
+};
+
+/* Static local state */
+static struct gemini_flash *gf;
+
+static void gemini_flash_enable_pins(void)
+{
+	int ret;
+
+	if (IS_ERR(gf->enabled_state))
+		return;
+	ret = pinctrl_select_state(gf->p, gf->enabled_state);
+	if (ret)
+		dev_err(gf->dev, "failed to enable pins\n");
+}
+
+static void gemini_flash_disable_pins(void)
+{
+	int ret;
+
+	if (IS_ERR(gf->disabled_state))
+		return;
+	ret = pinctrl_select_state(gf->p, gf->disabled_state);
+	if (ret)
+		dev_err(gf->dev, "failed to disable pins\n");
+}
+
+static map_word __xipram gemini_flash_map_read(struct map_info *map,
+					       unsigned long ofs)
+{
+	map_word ret;
+
+	gemini_flash_enable_pins();
+	ret = inline_map_read(map, ofs);
+	gemini_flash_disable_pins();
+
+	return ret;
+}
+
+static void __xipram gemini_flash_map_write(struct map_info *map,
+					    const map_word datum,
+					    unsigned long ofs)
+{
+	gemini_flash_enable_pins();
+	inline_map_write(map, datum, ofs);
+	gemini_flash_disable_pins();
+}
+
+static void __xipram gemini_flash_map_copy_from(struct map_info *map,
+						void *to, unsigned long from,
+						ssize_t len)
+{
+	gemini_flash_enable_pins();
+	inline_map_copy_from(map, to, from, len);
+	gemini_flash_disable_pins();
+}
+
+static void __xipram gemini_flash_map_copy_to(struct map_info *map,
+					      unsigned long to,
+					      const void *from, ssize_t len)
+{
+	gemini_flash_enable_pins();
+	inline_map_copy_to(map, to, from, len);
+	gemini_flash_disable_pins();
+}
+
+int of_flash_probe_gemini(struct platform_device *pdev,
+			  struct device_node *np,
+			  struct map_info *map)
+{
+	struct regmap *rmap;
+	struct device *dev = &pdev->dev;
+	u32 val;
+	int ret;
+
+	/* Multiplatform guard */
+	if (!of_device_is_compatible(np, "cortina,gemini-flash"))
+		return 0;
+
+	gf = devm_kzalloc(dev, sizeof(*gf), GFP_KERNEL);
+	if (!gf)
+		return -ENOMEM;
+	gf->dev = dev;
+
+	rmap = syscon_regmap_lookup_by_phandle(np, "syscon");
+	if (IS_ERR(rmap)) {
+		dev_err(dev, "no syscon\n");
+		return PTR_ERR(rmap);
+	}
+
+	ret = regmap_read(rmap, GLOBAL_STATUS, &val);
+	if (ret) {
+		dev_err(dev, "failed to read global status register\n");
+		return -ENODEV;
+	}
+	dev_dbg(dev, "global status reg: %08x\n", val);
+
+	/*
+	 * It would be contradictory if a physmap flash was NOT parallel.
+	 */
+	if ((val & FLASH_TYPE_MASK) != FLASH_TYPE_PARALLEL) {
+		dev_err(dev, "flash is not parallel\n");
+		return -ENODEV;
+	}
+
+	/*
+	 * Complain if DT data and hardware definition is different.
+	 */
+	if (val & FLASH_WIDTH_16BIT) {
+		if (map->bankwidth != 2)
+			dev_warn(dev, "flash hardware say flash is 16 bit wide but DT says it is %d bits wide\n",
+				 map->bankwidth * 8);
+	} else {
+		if (map->bankwidth != 1)
+			dev_warn(dev, "flash hardware say flash is 8 bit wide but DT says it is %d bits wide\n",
+				 map->bankwidth * 8);
+	}
+
+	gf->p = devm_pinctrl_get(dev);
+	if (IS_ERR(gf->p)) {
+		dev_err(dev, "no pinctrl handle\n");
+		ret = PTR_ERR(gf->p);
+		return ret;
+	}
+
+	gf->enabled_state = pinctrl_lookup_state(gf->p, "enabled");
+	if (IS_ERR(gf->enabled_state))
+		dev_err(dev, "no enabled pin control state\n");
+
+	gf->disabled_state = pinctrl_lookup_state(gf->p, "disabled");
+	if (IS_ERR(gf->disabled_state)) {
+		dev_err(dev, "no disabled pin control state\n");
+	} else {
+		ret = pinctrl_select_state(gf->p, gf->disabled_state);
+		if (ret)
+			dev_err(gf->dev, "failed to disable pins\n");
+	}
+
+	map->read = gemini_flash_map_read;
+	map->write = gemini_flash_map_write;
+	map->copy_from = gemini_flash_map_copy_from;
+	map->copy_to = gemini_flash_map_copy_to;
+
+	dev_info(dev, "initialized Gemini-specific physmap control\n");
+
+	return 0;
+}