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, 116 insertions, 0 deletions

diff --git a/drivers/net/nlmon.c b/drivers/net/nlmon.c
new file mode 100644
index 000000000..8bfd4ee5a
--- /dev/null
+++ b/drivers/net/nlmon.c
@@ -0,0 +1,116 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/ethtool.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/netlink.h>
+#include <net/net_namespace.h>
+#include <linux/if_arp.h>
+#include <net/rtnetlink.h>
+
+static netdev_tx_t nlmon_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	dev_lstats_add(dev, skb->len);
+
+	dev_kfree_skb(skb);
+
+	return NETDEV_TX_OK;
+}
+
+struct nlmon {
+	struct netlink_tap nt;
+};
+
+static int nlmon_open(struct net_device *dev)
+{
+	struct nlmon *nlmon = netdev_priv(dev);
+
+	nlmon->nt.dev = dev;
+	nlmon->nt.module = THIS_MODULE;
+	return netlink_add_tap(&nlmon->nt);
+}
+
+static int nlmon_close(struct net_device *dev)
+{
+	struct nlmon *nlmon = netdev_priv(dev);
+
+	return netlink_remove_tap(&nlmon->nt);
+}
+
+static void
+nlmon_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
+{
+	dev_lstats_read(dev, &stats->rx_packets, &stats->rx_bytes);
+}
+
+static u32 always_on(struct net_device *dev)
+{
+	return 1;
+}
+
+static const struct ethtool_ops nlmon_ethtool_ops = {
+	.get_link = always_on,
+};
+
+static const struct net_device_ops nlmon_ops = {
+	.ndo_open = nlmon_open,
+	.ndo_stop = nlmon_close,
+	.ndo_start_xmit = nlmon_xmit,
+	.ndo_get_stats64 = nlmon_get_stats64,
+};
+
+static void nlmon_setup(struct net_device *dev)
+{
+	dev->type = ARPHRD_NETLINK;
+	dev->priv_flags |= IFF_NO_QUEUE;
+	dev->lltx = true;
+
+	dev->netdev_ops	= &nlmon_ops;
+	dev->ethtool_ops = &nlmon_ethtool_ops;
+	dev->needs_free_netdev = true;
+
+	dev->features = NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA;
+	dev->flags = IFF_NOARP;
+	dev->pcpu_stat_type = NETDEV_PCPU_STAT_LSTATS;
+
+	/* That's rather a softlimit here, which, of course,
+	 * can be altered. Not a real MTU, but what is to be
+	 * expected in most cases.
+	 */
+	dev->mtu = NLMSG_GOODSIZE;
+	dev->min_mtu = sizeof(struct nlmsghdr);
+}
+
+static int nlmon_validate(struct nlattr *tb[], struct nlattr *data[],
+			  struct netlink_ext_ack *extack)
+{
+	if (tb[IFLA_ADDRESS])
+		return -EINVAL;
+	return 0;
+}
+
+static struct rtnl_link_ops nlmon_link_ops __read_mostly = {
+	.kind			= "nlmon",
+	.priv_size		= sizeof(struct nlmon),
+	.setup			= nlmon_setup,
+	.validate		= nlmon_validate,
+};
+
+static __init int nlmon_register(void)
+{
+	return rtnl_link_register(&nlmon_link_ops);
+}
+
+static __exit void nlmon_unregister(void)
+{
+	rtnl_link_unregister(&nlmon_link_ops);
+}
+
+module_init(nlmon_register);
+module_exit(nlmon_unregister);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Daniel Borkmann <dborkman@redhat.com>");
+MODULE_AUTHOR("Mathieu Geli <geli@enseirb.fr>");
+MODULE_DESCRIPTION("Netlink monitoring device");
+MODULE_ALIAS_RTNL_LINK("nlmon");