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, 132 insertions, 0 deletions

diff --git a/drivers/net/arcnet/com20020.h b/drivers/net/arcnet/com20020.h
new file mode 100644
index 000000000..68f49fdc2
--- /dev/null
+++ b/drivers/net/arcnet/com20020.h
@@ -0,0 +1,132 @@
+/*
+ * Linux ARCnet driver - COM20020 chipset support - function declarations
+ *
+ * Written 1997 by David Woodhouse.
+ * Written 1994-1999 by Avery Pennarun.
+ * Derived from skeleton.c by Donald Becker.
+ *
+ * Special thanks to Contemporary Controls, Inc. (www.ccontrols.com)
+ *  for sponsoring the further development of this driver.
+ *
+ * **********************
+ *
+ * The original copyright of skeleton.c was as follows:
+ *
+ * skeleton.c Written 1993 by Donald Becker.
+ * Copyright 1993 United States Government as represented by the
+ * Director, National Security Agency.  This software may only be used
+ * and distributed according to the terms of the GNU General Public License as
+ * modified by SRC, incorporated herein by reference.
+ *
+ * **********************
+ *
+ * For more details, see drivers/net/arcnet.c
+ *
+ * **********************
+ */
+#ifndef __COM20020_H
+#define __COM20020_H
+#include <linux/leds.h>
+
+int com20020_check(struct net_device *dev);
+int com20020_found(struct net_device *dev, int shared);
+extern const struct net_device_ops com20020_netdev_ops;
+
+/* The number of low I/O ports used by the card. */
+#define ARCNET_TOTAL_SIZE 8
+
+#define PLX_PCI_MAX_CARDS 2
+
+struct ledoffsets {
+	int green;
+	int red;
+};
+
+struct com20020_pci_channel_map {
+	u32 bar;
+	u32 offset;
+	u32 size;               /* 0x00 - auto, e.g. length of entire bar */
+};
+
+struct com20020_pci_card_info {
+	const char *name;
+	u8 devcount;
+
+	struct com20020_pci_channel_map chan_map_tbl[PLX_PCI_MAX_CARDS];
+	struct com20020_pci_channel_map misc_map;
+
+	struct ledoffsets leds[PLX_PCI_MAX_CARDS];
+	int rotary;
+
+	unsigned int flags;
+};
+
+struct com20020_priv {
+	struct com20020_pci_card_info *ci;
+	struct list_head list_dev;
+	resource_size_t misc;
+};
+
+struct com20020_dev {
+	struct list_head list;
+	struct net_device *dev;
+
+	struct led_classdev tx_led;
+	struct led_classdev recon_led;
+
+	struct com20020_priv *pci_priv;
+	int index;
+};
+
+#define COM20020_REG_W_INTMASK	0	/* writable */
+#define COM20020_REG_R_STATUS	0	/* readable */
+#define COM20020_REG_W_COMMAND	1	/* standard arcnet commands */
+#define COM20020_REG_R_DIAGSTAT	1	/* diagnostic status */
+#define COM20020_REG_W_ADDR_HI	2	/* control for IO-mapped memory */
+#define COM20020_REG_W_ADDR_LO	3
+#define COM20020_REG_RW_MEMDATA	4	/* data port for IO-mapped memory */
+#define COM20020_REG_W_SUBADR	5	/* the extended port _XREG refers to */
+#define COM20020_REG_W_CONFIG	6	/* configuration */
+#define COM20020_REG_W_XREG	7	/* extra
+					 * (indexed by _CONFIG or _SUBADDR)
+					 */
+
+/* in the ADDR_HI register */
+#define RDDATAflag	0x80	/* next access is a read (not a write) */
+
+/* in the DIAGSTAT register */
+#define NEWNXTIDflag	0x02	/* ID to which token is passed has changed */
+
+/* in the CONFIG register */
+#define RESETcfg	0x80	/* put card in reset state */
+#define TXENcfg		0x20	/* enable TX */
+#define XTOcfg(x)	((x) << 3)	/* extended timeout */
+
+/* in SETUP register */
+#define PROMISCset	0x10	/* enable RCV_ALL */
+#define P1MODE		0x80    /* enable P1-MODE for Backplane */
+#define SLOWARB		0x01    /* enable Slow Arbitration for >=5Mbps */
+
+/* COM2002x */
+#define SUB_TENTATIVE	0	/* tentative node ID */
+#define SUB_NODE	1	/* node ID */
+#define SUB_SETUP1	2	/* various options */
+#define SUB_TEST	3	/* test/diag register */
+
+/* COM20022 only */
+#define SUB_SETUP2	4	/* sundry options */
+#define SUB_BUSCTL	5	/* bus control options */
+#define SUB_DMACOUNT	6	/* DMA count options */
+
+static inline void com20020_set_subaddress(struct arcnet_local *lp,
+					   int ioaddr, int val)
+{
+	if (val < 4) {
+		lp->config = (lp->config & ~0x03) | val;
+		outb(lp->config, ioaddr + COM20020_REG_W_CONFIG);
+	} else {
+		outb(val, ioaddr + COM20020_REG_W_SUBADR);
+	}
+}
+
+#endif /* __COM20020_H */