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, 364 insertions, 0 deletions

diff --git a/include/net/l3mdev.h b/include/net/l3mdev.h
new file mode 100644
index 000000000..710e98665
--- /dev/null
+++ b/include/net/l3mdev.h
@@ -0,0 +1,364 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * include/net/l3mdev.h - L3 master device API
+ * Copyright (c) 2015 Cumulus Networks
+ * Copyright (c) 2015 David Ahern <dsa@cumulusnetworks.com>
+ */
+#ifndef _NET_L3MDEV_H_
+#define _NET_L3MDEV_H_
+
+#include <net/dst.h>
+#include <net/fib_rules.h>
+
+enum l3mdev_type {
+	L3MDEV_TYPE_UNSPEC,
+	L3MDEV_TYPE_VRF,
+	__L3MDEV_TYPE_MAX
+};
+
+#define L3MDEV_TYPE_MAX (__L3MDEV_TYPE_MAX - 1)
+
+typedef int (*lookup_by_table_id_t)(struct net *net, u32 table_d);
+
+/**
+ * struct l3mdev_ops - l3mdev operations
+ *
+ * @l3mdev_fib_table: Get FIB table id to use for lookups
+ *
+ * @l3mdev_l3_rcv:    Hook in L3 receive path
+ *
+ * @l3mdev_l3_out:    Hook in L3 output path
+ *
+ * @l3mdev_link_scope_lookup: IPv6 lookup for linklocal and mcast destinations
+ */
+
+struct l3mdev_ops {
+	u32		(*l3mdev_fib_table)(const struct net_device *dev);
+	struct sk_buff * (*l3mdev_l3_rcv)(struct net_device *dev,
+					  struct sk_buff *skb, u16 proto);
+	struct sk_buff * (*l3mdev_l3_out)(struct net_device *dev,
+					  struct sock *sk, struct sk_buff *skb,
+					  u16 proto);
+
+	/* IPv6 ops */
+	struct dst_entry * (*l3mdev_link_scope_lookup)(const struct net_device *dev,
+						 struct flowi6 *fl6);
+};
+
+#ifdef CONFIG_NET_L3_MASTER_DEV
+
+int l3mdev_table_lookup_register(enum l3mdev_type l3type,
+				 lookup_by_table_id_t fn);
+
+void l3mdev_table_lookup_unregister(enum l3mdev_type l3type,
+				    lookup_by_table_id_t fn);
+
+int l3mdev_ifindex_lookup_by_table_id(enum l3mdev_type l3type, struct net *net,
+				      u32 table_id);
+
+int l3mdev_fib_rule_match(struct net *net, struct flowi *fl,
+			  struct fib_lookup_arg *arg);
+
+static inline
+bool l3mdev_fib_rule_iif_match(const struct flowi *fl, int iifindex)
+{
+	return !(fl->flowi_flags & FLOWI_FLAG_L3MDEV_OIF) &&
+	       fl->flowi_l3mdev == iifindex;
+}
+
+static inline
+bool l3mdev_fib_rule_oif_match(const struct flowi *fl, int oifindex)
+{
+	return fl->flowi_flags & FLOWI_FLAG_L3MDEV_OIF &&
+	       fl->flowi_l3mdev == oifindex;
+}
+
+void l3mdev_update_flow(struct net *net, struct flowi *fl);
+
+int l3mdev_master_ifindex_rcu(const struct net_device *dev);
+static inline int l3mdev_master_ifindex(struct net_device *dev)
+{
+	int ifindex;
+
+	rcu_read_lock();
+	ifindex = l3mdev_master_ifindex_rcu(dev);
+	rcu_read_unlock();
+
+	return ifindex;
+}
+
+static inline int l3mdev_master_ifindex_by_index(struct net *net, int ifindex)
+{
+	struct net_device *dev;
+	int rc = 0;
+
+	if (ifindex) {
+		rcu_read_lock();
+
+		dev = dev_get_by_index_rcu(net, ifindex);
+		if (dev)
+			rc = l3mdev_master_ifindex_rcu(dev);
+
+		rcu_read_unlock();
+	}
+
+	return rc;
+}
+
+static inline
+struct net_device *l3mdev_master_dev_rcu(const struct net_device *_dev)
+{
+	/* netdev_master_upper_dev_get_rcu calls
+	 * list_first_or_null_rcu to walk the upper dev list.
+	 * list_first_or_null_rcu does not handle a const arg. We aren't
+	 * making changes, just want the master device from that list so
+	 * typecast to remove the const
+	 */
+	struct net_device *dev = (struct net_device *)_dev;
+	struct net_device *master;
+
+	if (!dev)
+		return NULL;
+
+	if (netif_is_l3_master(dev))
+		master = dev;
+	else if (netif_is_l3_slave(dev))
+		master = netdev_master_upper_dev_get_rcu(dev);
+	else
+		master = NULL;
+
+	return master;
+}
+
+int l3mdev_master_upper_ifindex_by_index_rcu(struct net *net, int ifindex);
+static inline
+int l3mdev_master_upper_ifindex_by_index(struct net *net, int ifindex)
+{
+	rcu_read_lock();
+	ifindex = l3mdev_master_upper_ifindex_by_index_rcu(net, ifindex);
+	rcu_read_unlock();
+
+	return ifindex;
+}
+
+u32 l3mdev_fib_table_rcu(const struct net_device *dev);
+u32 l3mdev_fib_table_by_index(struct net *net, int ifindex);
+static inline u32 l3mdev_fib_table(const struct net_device *dev)
+{
+	u32 tb_id;
+
+	rcu_read_lock();
+	tb_id = l3mdev_fib_table_rcu(dev);
+	rcu_read_unlock();
+
+	return tb_id;
+}
+
+static inline bool netif_index_is_l3_master(struct net *net, int ifindex)
+{
+	struct net_device *dev;
+	bool rc = false;
+
+	if (ifindex == 0)
+		return false;
+
+	rcu_read_lock();
+
+	dev = dev_get_by_index_rcu(net, ifindex);
+	if (dev)
+		rc = netif_is_l3_master(dev);
+
+	rcu_read_unlock();
+
+	return rc;
+}
+
+struct dst_entry *l3mdev_link_scope_lookup(struct net *net, struct flowi6 *fl6);
+
+static inline
+struct sk_buff *l3mdev_l3_rcv(struct sk_buff *skb, u16 proto)
+{
+	struct net_device *master = NULL;
+
+	if (netif_is_l3_slave(skb->dev))
+		master = netdev_master_upper_dev_get_rcu(skb->dev);
+	else if (netif_is_l3_master(skb->dev) ||
+		 netif_has_l3_rx_handler(skb->dev))
+		master = skb->dev;
+
+	if (master && master->l3mdev_ops->l3mdev_l3_rcv)
+		skb = master->l3mdev_ops->l3mdev_l3_rcv(master, skb, proto);
+
+	return skb;
+}
+
+static inline
+struct sk_buff *l3mdev_ip_rcv(struct sk_buff *skb)
+{
+	return l3mdev_l3_rcv(skb, AF_INET);
+}
+
+static inline
+struct sk_buff *l3mdev_ip6_rcv(struct sk_buff *skb)
+{
+	return l3mdev_l3_rcv(skb, AF_INET6);
+}
+
+static inline
+struct sk_buff *l3mdev_l3_out(struct sock *sk, struct sk_buff *skb, u16 proto)
+{
+	struct net_device *dev;
+
+	rcu_read_lock();
+	dev = skb_dst_dev_rcu(skb);
+	if (netif_is_l3_slave(dev)) {
+		struct net_device *master;
+
+		master = netdev_master_upper_dev_get_rcu(dev);
+		if (master && master->l3mdev_ops->l3mdev_l3_out)
+			skb = master->l3mdev_ops->l3mdev_l3_out(master, sk,
+								skb, proto);
+	}
+	rcu_read_unlock();
+
+	return skb;
+}
+
+static inline
+struct sk_buff *l3mdev_ip_out(struct sock *sk, struct sk_buff *skb)
+{
+	return l3mdev_l3_out(sk, skb, AF_INET);
+}
+
+static inline
+struct sk_buff *l3mdev_ip6_out(struct sock *sk, struct sk_buff *skb)
+{
+	return l3mdev_l3_out(sk, skb, AF_INET6);
+}
+#else
+
+static inline int l3mdev_master_ifindex_rcu(const struct net_device *dev)
+{
+	return 0;
+}
+static inline int l3mdev_master_ifindex(struct net_device *dev)
+{
+	return 0;
+}
+
+static inline int l3mdev_master_ifindex_by_index(struct net *net, int ifindex)
+{
+	return 0;
+}
+
+static inline
+int l3mdev_master_upper_ifindex_by_index_rcu(struct net *net, int ifindex)
+{
+	return 0;
+}
+static inline
+int l3mdev_master_upper_ifindex_by_index(struct net *net, int ifindex)
+{
+	return 0;
+}
+
+static inline
+struct net_device *l3mdev_master_dev_rcu(const struct net_device *dev)
+{
+	return NULL;
+}
+
+static inline u32 l3mdev_fib_table_rcu(const struct net_device *dev)
+{
+	return 0;
+}
+static inline u32 l3mdev_fib_table(const struct net_device *dev)
+{
+	return 0;
+}
+static inline u32 l3mdev_fib_table_by_index(struct net *net, int ifindex)
+{
+	return 0;
+}
+
+static inline bool netif_index_is_l3_master(struct net *net, int ifindex)
+{
+	return false;
+}
+
+static inline
+struct dst_entry *l3mdev_link_scope_lookup(struct net *net, struct flowi6 *fl6)
+{
+	return NULL;
+}
+
+static inline
+struct sk_buff *l3mdev_ip_rcv(struct sk_buff *skb)
+{
+	return skb;
+}
+
+static inline
+struct sk_buff *l3mdev_ip6_rcv(struct sk_buff *skb)
+{
+	return skb;
+}
+
+static inline
+struct sk_buff *l3mdev_ip_out(struct sock *sk, struct sk_buff *skb)
+{
+	return skb;
+}
+
+static inline
+struct sk_buff *l3mdev_ip6_out(struct sock *sk, struct sk_buff *skb)
+{
+	return skb;
+}
+
+static inline
+int l3mdev_table_lookup_register(enum l3mdev_type l3type,
+				 lookup_by_table_id_t fn)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline
+void l3mdev_table_lookup_unregister(enum l3mdev_type l3type,
+				    lookup_by_table_id_t fn)
+{
+}
+
+static inline
+int l3mdev_ifindex_lookup_by_table_id(enum l3mdev_type l3type, struct net *net,
+				      u32 table_id)
+{
+	return -ENODEV;
+}
+
+static inline
+int l3mdev_fib_rule_match(struct net *net, struct flowi *fl,
+			  struct fib_lookup_arg *arg)
+{
+	return 1;
+}
+
+static inline
+bool l3mdev_fib_rule_iif_match(const struct flowi *fl, int iifindex)
+{
+	return false;
+}
+
+static inline
+bool l3mdev_fib_rule_oif_match(const struct flowi *fl, int oifindex)
+{
+	return false;
+}
+
+static inline
+void l3mdev_update_flow(struct net *net, struct flowi *fl)
+{
+}
+#endif
+
+#endif /* _NET_L3MDEV_H_ */