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, 185 insertions, 0 deletions

diff --git a/include/net/ip6_tunnel.h b/include/net/ip6_tunnel.h
new file mode 100644
index 000000000..b99805ee2
--- /dev/null
+++ b/include/net/ip6_tunnel.h
@@ -0,0 +1,185 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _NET_IP6_TUNNEL_H
+#define _NET_IP6_TUNNEL_H
+
+#include <linux/ipv6.h>
+#include <linux/netdevice.h>
+#include <linux/if_tunnel.h>
+#include <linux/ip6_tunnel.h>
+#include <net/ip_tunnels.h>
+#include <net/dst_cache.h>
+
+#define IP6TUNNEL_ERR_TIMEO (30*HZ)
+
+/* capable of sending packets */
+#define IP6_TNL_F_CAP_XMIT 0x10000
+/* capable of receiving packets */
+#define IP6_TNL_F_CAP_RCV 0x20000
+/* determine capability on a per-packet basis */
+#define IP6_TNL_F_CAP_PER_PACKET 0x40000
+
+struct __ip6_tnl_parm {
+	char name[IFNAMSIZ];	/* name of tunnel device */
+	int link;		/* ifindex of underlying L2 interface */
+	__u8 proto;		/* tunnel protocol */
+	__u8 encap_limit;	/* encapsulation limit for tunnel */
+	__u8 hop_limit;		/* hop limit for tunnel */
+	bool collect_md;
+	__be32 flowinfo;	/* traffic class and flowlabel for tunnel */
+	__u32 flags;		/* tunnel flags */
+	struct in6_addr laddr;	/* local tunnel end-point address */
+	struct in6_addr raddr;	/* remote tunnel end-point address */
+
+	IP_TUNNEL_DECLARE_FLAGS(i_flags);
+	IP_TUNNEL_DECLARE_FLAGS(o_flags);
+	__be32			i_key;
+	__be32			o_key;
+
+	__u32			fwmark;
+	__u32			index;	/* ERSPAN type II index */
+	__u8			erspan_ver;	/* ERSPAN version */
+	__u8			dir;	/* direction */
+	__u16			hwid;	/* hwid */
+};
+
+/* IPv6 tunnel */
+struct ip6_tnl {
+	struct ip6_tnl __rcu *next;	/* next tunnel in list */
+	struct net_device *dev;	/* virtual device associated with tunnel */
+	netdevice_tracker dev_tracker;
+	struct net *net;	/* netns for packet i/o */
+	struct __ip6_tnl_parm parms;	/* tunnel configuration parameters */
+	struct flowi fl;	/* flowi template for xmit */
+	struct dst_cache dst_cache;	/* cached dst */
+	struct gro_cells gro_cells;
+
+	int err_count;
+	unsigned long err_time;
+
+	/* These fields used only by GRE */
+	__u32 i_seqno;	/* The last seen seqno	*/
+	atomic_t o_seqno;	/* The last output seqno */
+	int hlen;       /* tun_hlen + encap_hlen */
+	int tun_hlen;	/* Precalculated header length */
+	int encap_hlen; /* Encap header length (FOU,GUE) */
+	struct ip_tunnel_encap encap;
+	int mlink;
+};
+
+struct ip6_tnl_encap_ops {
+	size_t (*encap_hlen)(struct ip_tunnel_encap *e);
+	int (*build_header)(struct sk_buff *skb, struct ip_tunnel_encap *e,
+			    u8 *protocol, struct flowi6 *fl6);
+	int (*err_handler)(struct sk_buff *skb, struct inet6_skb_parm *opt,
+			   u8 type, u8 code, int offset, __be32 info);
+};
+
+#ifdef CONFIG_INET
+
+extern const struct ip6_tnl_encap_ops __rcu *
+		ip6tun_encaps[MAX_IPTUN_ENCAP_OPS];
+
+int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
+			  unsigned int num);
+int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
+			  unsigned int num);
+int ip6_tnl_encap_setup(struct ip6_tnl *t,
+			struct ip_tunnel_encap *ipencap);
+
+static inline int ip6_encap_hlen(struct ip_tunnel_encap *e)
+{
+	const struct ip6_tnl_encap_ops *ops;
+	int hlen = -EINVAL;
+
+	if (e->type == TUNNEL_ENCAP_NONE)
+		return 0;
+
+	if (e->type >= MAX_IPTUN_ENCAP_OPS)
+		return -EINVAL;
+
+	rcu_read_lock();
+	ops = rcu_dereference(ip6tun_encaps[e->type]);
+	if (likely(ops && ops->encap_hlen))
+		hlen = ops->encap_hlen(e);
+	rcu_read_unlock();
+
+	return hlen;
+}
+
+static inline int ip6_tnl_encap(struct sk_buff *skb, struct ip6_tnl *t,
+				u8 *protocol, struct flowi6 *fl6)
+{
+	const struct ip6_tnl_encap_ops *ops;
+	int ret = -EINVAL;
+
+	if (t->encap.type == TUNNEL_ENCAP_NONE)
+		return 0;
+
+	if (t->encap.type >= MAX_IPTUN_ENCAP_OPS)
+		return -EINVAL;
+
+	rcu_read_lock();
+	ops = rcu_dereference(ip6tun_encaps[t->encap.type]);
+	if (likely(ops && ops->build_header))
+		ret = ops->build_header(skb, &t->encap, protocol, fl6);
+	rcu_read_unlock();
+
+	return ret;
+}
+
+/* Tunnel encapsulation limit destination sub-option */
+
+struct ipv6_tlv_tnl_enc_lim {
+	__u8 type;		/* type-code for option         */
+	__u8 length;		/* option length                */
+	__u8 encap_limit;	/* tunnel encapsulation limit   */
+} __packed;
+
+int ip6_tnl_rcv_ctl(struct ip6_tnl *t, const struct in6_addr *laddr,
+		const struct in6_addr *raddr);
+int ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
+		const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
+		bool log_ecn_error);
+int ip6_tnl_xmit_ctl(struct ip6_tnl *t, const struct in6_addr *laddr,
+		     const struct in6_addr *raddr);
+int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
+		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, __u8 proto);
+__u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw);
+__u32 ip6_tnl_get_cap(struct ip6_tnl *t, const struct in6_addr *laddr,
+			     const struct in6_addr *raddr);
+struct net *ip6_tnl_get_link_net(const struct net_device *dev);
+int ip6_tnl_get_iflink(const struct net_device *dev);
+int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu);
+
+static inline void ip6tunnel_xmit(struct sock *sk, struct sk_buff *skb,
+				  struct net_device *dev, u16 ip6cb_flags)
+{
+	int pkt_len, err;
+
+	if (unlikely(dev_recursion_level() > IP_TUNNEL_RECURSION_LIMIT)) {
+		if (dev) {
+			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
+					     dev->name);
+			DEV_STATS_INC(dev, tx_errors);
+		}
+		kfree_skb_reason(skb, SKB_DROP_REASON_RECURSION_LIMIT);
+		return;
+	}
+
+	dev_xmit_recursion_inc();
+
+	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
+	IP6CB(skb)->flags = ip6cb_flags;
+	pkt_len = skb->len - skb_inner_network_offset(skb);
+	err = ip6_local_out(skb_dst_dev_net(skb), sk, skb);
+
+	if (dev) {
+		if (unlikely(net_xmit_eval(err)))
+			pkt_len = -1;
+		iptunnel_xmit_stats(dev, pkt_len);
+	}
+
+	dev_xmit_recursion_dec();
+}
+#endif
+#endif