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, 165 insertions, 0 deletions

diff --git a/arch/arm/mach-omap1/timer.c b/arch/arm/mach-omap1/timer.c
new file mode 100644
index 000000000..399e6d840
--- /dev/null
+++ b/arch/arm/mach-omap1/timer.c
@@ -0,0 +1,165 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * OMAP1 Dual-Mode Timers - platform device registration
+ *
+ * Contains first level initialization routines which internally
+ * generates timer device information and registers with linux
+ * device model. It also has a low level function to change the timer
+ * input clock source.
+ *
+ * Copyright (C) 2011 Texas Instruments Incorporated - https://www.ti.com/
+ * Tarun Kanti DebBarma <tarun.kanti@ti.com>
+ * Thara Gopinath <thara@ti.com>
+ */
+
+#include <linux/clk.h>
+#include <linux/io.h>
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/platform_data/dmtimer-omap.h>
+#include <linux/soc/ti/omap1-io.h>
+
+#include <clocksource/timer-ti-dm.h>
+
+#include "soc.h"
+
+#define OMAP1610_GPTIMER1_BASE		0xfffb1400
+#define OMAP1610_GPTIMER2_BASE		0xfffb1c00
+#define OMAP1610_GPTIMER3_BASE		0xfffb2400
+#define OMAP1610_GPTIMER4_BASE		0xfffb2c00
+#define OMAP1610_GPTIMER5_BASE		0xfffb3400
+#define OMAP1610_GPTIMER6_BASE		0xfffb3c00
+#define OMAP1610_GPTIMER7_BASE		0xfffb7400
+#define OMAP1610_GPTIMER8_BASE		0xfffbd400
+
+#define OMAP1_DM_TIMER_COUNT		8
+
+static int omap1_dm_timer_set_src(struct platform_device *pdev,
+				int source)
+{
+	int n = (pdev->id - 1) << 1;
+	u32 l;
+
+	l = omap_readl(MOD_CONF_CTRL_1) & ~(0x03 << n);
+	l |= source << n;
+	omap_writel(l, MOD_CONF_CTRL_1);
+
+	return 0;
+}
+
+static int __init omap1_dm_timer_init(void)
+{
+	int i;
+	int ret;
+	struct dmtimer_platform_data *pdata;
+	struct platform_device *pdev;
+
+	if (!cpu_is_omap16xx())
+		return 0;
+
+	for (i = 1; i <= OMAP1_DM_TIMER_COUNT; i++) {
+		struct resource res[2];
+		u32 base, irq;
+
+		switch (i) {
+		case 1:
+			base = OMAP1610_GPTIMER1_BASE;
+			irq = INT_1610_GPTIMER1;
+			break;
+		case 2:
+			base = OMAP1610_GPTIMER2_BASE;
+			irq = INT_1610_GPTIMER2;
+			break;
+		case 3:
+			base = OMAP1610_GPTIMER3_BASE;
+			irq = INT_1610_GPTIMER3;
+			break;
+		case 4:
+			base = OMAP1610_GPTIMER4_BASE;
+			irq = INT_1610_GPTIMER4;
+			break;
+		case 5:
+			base = OMAP1610_GPTIMER5_BASE;
+			irq = INT_1610_GPTIMER5;
+			break;
+		case 6:
+			base = OMAP1610_GPTIMER6_BASE;
+			irq = INT_1610_GPTIMER6;
+			break;
+		case 7:
+			base = OMAP1610_GPTIMER7_BASE;
+			irq = INT_1610_GPTIMER7;
+			break;
+		case 8:
+			base = OMAP1610_GPTIMER8_BASE;
+			irq = INT_1610_GPTIMER8;
+			break;
+		default:
+			/*
+			 * not supposed to reach here.
+			 * this is to remove warning.
+			 */
+			return -EINVAL;
+		}
+
+		pdev = platform_device_alloc("omap_timer", i);
+		if (!pdev) {
+			pr_err("%s: Failed to device alloc for dmtimer%d\n",
+				__func__, i);
+			return -ENOMEM;
+		}
+
+		memset(res, 0, 2 * sizeof(struct resource));
+		res[0].start = base;
+		res[0].end = base + 0x46;
+		res[0].flags = IORESOURCE_MEM;
+		res[1].start = irq;
+		res[1].end = irq;
+		res[1].flags = IORESOURCE_IRQ;
+		ret = platform_device_add_resources(pdev, res,
+				ARRAY_SIZE(res));
+		if (ret) {
+			dev_err(&pdev->dev, "%s: Failed to add resources.\n",
+				__func__);
+			goto err_free_pdev;
+		}
+
+		pdata = kzalloc_obj(*pdata);
+		if (!pdata) {
+			ret = -ENOMEM;
+			goto err_free_pdata;
+		}
+
+		pdata->set_timer_src = omap1_dm_timer_set_src;
+		pdata->timer_capability = OMAP_TIMER_ALWON |
+				OMAP_TIMER_NEEDS_RESET | OMAP_TIMER_HAS_DSP_IRQ;
+
+		ret = platform_device_add_data(pdev, pdata, sizeof(*pdata));
+		if (ret) {
+			dev_err(&pdev->dev, "%s: Failed to add platform data.\n",
+				__func__);
+			goto err_free_pdata;
+		}
+
+		ret = platform_device_add(pdev);
+		if (ret) {
+			dev_err(&pdev->dev, "%s: Failed to add platform device.\n",
+				__func__);
+			goto err_free_pdata;
+		}
+
+		dev_dbg(&pdev->dev, " Registered.\n");
+	}
+
+	return 0;
+
+err_free_pdata:
+	kfree(pdata);
+
+err_free_pdev:
+	platform_device_put(pdev);
+
+	return ret;
+}
+arch_initcall(omap1_dm_timer_init);