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, 154 insertions, 0 deletions

diff --git a/arch/m68k/virt/ints.c b/arch/m68k/virt/ints.c
new file mode 100644
index 000000000..896aa6eb8
--- /dev/null
+++ b/arch/m68k/virt/ints.c
@@ -0,0 +1,154 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/delay.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/sched/debug.h>
+#include <linux/types.h>
+#include <linux/ioport.h>
+
+#include <asm/hwtest.h>
+#include <asm/irq.h>
+#include <asm/irq_regs.h>
+#include <asm/processor.h>
+#include <asm/virt.h>
+
+#define GFPIC_REG_IRQ_PENDING           0x04
+#define GFPIC_REG_IRQ_DISABLE_ALL       0x08
+#define GFPIC_REG_IRQ_DISABLE           0x0c
+#define GFPIC_REG_IRQ_ENABLE            0x10
+
+static struct resource picres[6];
+static const char *picname[6] = {
+	"goldfish_pic.0",
+	"goldfish_pic.1",
+	"goldfish_pic.2",
+	"goldfish_pic.3",
+	"goldfish_pic.4",
+	"goldfish_pic.5"
+};
+
+/*
+ * 6 goldfish-pic for CPU IRQ #1 to IRQ #6
+ * CPU IRQ #1 -> PIC #1
+ *               IRQ #1 to IRQ #31 -> unused
+ *               IRQ #32 -> goldfish-tty
+ * CPU IRQ #2 -> PIC #2
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 1 to 32
+ * CPU IRQ #3 -> PIC #3
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 33 to 64
+ * CPU IRQ #4 -> PIC #4
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 65 to 96
+ * CPU IRQ #5 -> PIC #5
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 97 to 128
+ * CPU IRQ #6 -> PIC #6
+ *               IRQ #1 -> goldfish-timer
+ *               IRQ #2 -> goldfish-rtc
+ *               IRQ #3 to IRQ #32 -> unused
+ * CPU IRQ #7 -> NMI
+ */
+
+static u32 gfpic_read(int pic, int reg)
+{
+	void __iomem *base = (void __iomem *)(virt_bi_data.pic.mmio +
+					      pic * 0x1000);
+
+	return ioread32be(base + reg);
+}
+
+static void gfpic_write(u32 value, int pic, int reg)
+{
+	void __iomem *base = (void __iomem *)(virt_bi_data.pic.mmio +
+					      pic * 0x1000);
+
+	iowrite32be(value, base + reg);
+}
+
+#define GF_PIC(irq) ((irq - IRQ_USER) / 32)
+#define GF_IRQ(irq) ((irq - IRQ_USER) % 32)
+
+static void virt_irq_enable(struct irq_data *data)
+{
+	gfpic_write(BIT(GF_IRQ(data->irq)), GF_PIC(data->irq),
+		    GFPIC_REG_IRQ_ENABLE);
+}
+
+static void virt_irq_disable(struct irq_data *data)
+{
+	gfpic_write(BIT(GF_IRQ(data->irq)), GF_PIC(data->irq),
+		    GFPIC_REG_IRQ_DISABLE);
+}
+
+static unsigned int virt_irq_startup(struct irq_data *data)
+{
+	virt_irq_enable(data);
+	return 0;
+}
+
+static irqreturn_t virt_nmi_handler(int irq, void *dev_id)
+{
+	static int in_nmi;
+
+	if (READ_ONCE(in_nmi))
+		return IRQ_HANDLED;
+	WRITE_ONCE(in_nmi, 1);
+
+	pr_warn("Non-Maskable Interrupt\n");
+	show_registers(get_irq_regs());
+
+	WRITE_ONCE(in_nmi, 0);
+	return IRQ_HANDLED;
+}
+
+static struct irq_chip virt_irq_chip = {
+	.name		= "virt",
+	.irq_enable	= virt_irq_enable,
+	.irq_disable	= virt_irq_disable,
+	.irq_startup	= virt_irq_startup,
+	.irq_shutdown	= virt_irq_disable,
+};
+
+static void goldfish_pic_irq(struct irq_desc *desc)
+{
+	u32 irq_pending;
+	unsigned int irq_num;
+	unsigned int pic = desc->irq_data.irq - 1;
+
+	irq_pending = gfpic_read(pic, GFPIC_REG_IRQ_PENDING);
+	irq_num = IRQ_USER + pic * 32;
+
+	do {
+		if (irq_pending & 1)
+			generic_handle_irq(irq_num);
+		++irq_num;
+		irq_pending >>= 1;
+	} while (irq_pending);
+}
+
+void __init virt_init_IRQ(void)
+{
+	unsigned int i;
+
+	m68k_setup_irq_controller(&virt_irq_chip, handle_simple_irq, IRQ_USER,
+				  NUM_VIRT_SOURCES - IRQ_USER);
+
+	for (i = 0; i < 6; i++) {
+
+		picres[i] = (struct resource)
+		    DEFINE_RES_MEM_NAMED(virt_bi_data.pic.mmio + i * 0x1000,
+					 0x1000, picname[i]);
+		if (request_resource(&iomem_resource, &picres[i])) {
+			pr_err("Cannot allocate %s resource\n", picname[i]);
+			return;
+		}
+
+		irq_set_chained_handler(virt_bi_data.pic.irq + i,
+					goldfish_pic_irq);
+	}
+
+	if (request_irq(IRQ_AUTO_7, virt_nmi_handler, 0, "NMI",
+			virt_nmi_handler))
+		pr_err("Couldn't register NMI\n");
+}