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, 102 insertions, 0 deletions

diff --git a/net/802/stp.c b/net/802/stp.c
new file mode 100644
index 000000000..03c9f75e9
--- /dev/null
+++ b/net/802/stp.c
@@ -0,0 +1,102 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ *	STP SAP demux
+ *
+ *	Copyright (c) 2008 Patrick McHardy <kaber@trash.net>
+ */
+#include <linux/mutex.h>
+#include <linux/skbuff.h>
+#include <linux/etherdevice.h>
+#include <linux/llc.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <net/llc.h>
+#include <net/llc_pdu.h>
+#include <net/stp.h>
+
+/* 01:80:c2:00:00:20 - 01:80:c2:00:00:2F */
+#define GARP_ADDR_MIN	0x20
+#define GARP_ADDR_MAX	0x2F
+#define GARP_ADDR_RANGE	(GARP_ADDR_MAX - GARP_ADDR_MIN)
+
+static const struct stp_proto __rcu *garp_protos[GARP_ADDR_RANGE + 1] __read_mostly;
+static const struct stp_proto __rcu *stp_proto __read_mostly;
+
+static struct llc_sap *sap __read_mostly;
+static unsigned int sap_registered;
+static DEFINE_MUTEX(stp_proto_mutex);
+
+/* Called under rcu_read_lock from LLC */
+static int stp_pdu_rcv(struct sk_buff *skb, struct net_device *dev,
+		       struct packet_type *pt, struct net_device *orig_dev)
+{
+	const struct ethhdr *eh = eth_hdr(skb);
+	const struct llc_pdu_un *pdu = llc_pdu_un_hdr(skb);
+	const struct stp_proto *proto;
+
+	if (pdu->ssap != LLC_SAP_BSPAN ||
+	    pdu->dsap != LLC_SAP_BSPAN ||
+	    pdu->ctrl_1 != LLC_PDU_TYPE_U)
+		goto err;
+
+	if (eh->h_dest[5] >= GARP_ADDR_MIN && eh->h_dest[5] <= GARP_ADDR_MAX) {
+		proto = rcu_dereference(garp_protos[eh->h_dest[5] -
+						    GARP_ADDR_MIN]);
+		if (proto &&
+		    !ether_addr_equal(eh->h_dest, proto->group_address))
+			goto err;
+	} else
+		proto = rcu_dereference(stp_proto);
+
+	if (!proto)
+		goto err;
+
+	proto->rcv(proto, skb, dev);
+	return 0;
+
+err:
+	kfree_skb(skb);
+	return 0;
+}
+
+int stp_proto_register(const struct stp_proto *proto)
+{
+	int err = 0;
+
+	mutex_lock(&stp_proto_mutex);
+	if (sap_registered++ == 0) {
+		sap = llc_sap_open(LLC_SAP_BSPAN, stp_pdu_rcv);
+		if (!sap) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+	if (is_zero_ether_addr(proto->group_address))
+		rcu_assign_pointer(stp_proto, proto);
+	else
+		rcu_assign_pointer(garp_protos[proto->group_address[5] -
+					       GARP_ADDR_MIN], proto);
+out:
+	mutex_unlock(&stp_proto_mutex);
+	return err;
+}
+EXPORT_SYMBOL_GPL(stp_proto_register);
+
+void stp_proto_unregister(const struct stp_proto *proto)
+{
+	mutex_lock(&stp_proto_mutex);
+	if (is_zero_ether_addr(proto->group_address))
+		RCU_INIT_POINTER(stp_proto, NULL);
+	else
+		RCU_INIT_POINTER(garp_protos[proto->group_address[5] -
+					       GARP_ADDR_MIN], NULL);
+	synchronize_rcu();
+
+	if (--sap_registered == 0)
+		llc_sap_put(sap);
+	mutex_unlock(&stp_proto_mutex);
+}
+EXPORT_SYMBOL_GPL(stp_proto_unregister);
+
+MODULE_DESCRIPTION("SAP demux for IEEE 802.1D Spanning Tree Protocol (STP)");
+MODULE_LICENSE("GPL");