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, 148 insertions, 0 deletions

diff --git a/include/net/mrp.h b/include/net/mrp.h
new file mode 100644
index 000000000..b28915ffe
--- /dev/null
+++ b/include/net/mrp.h
@@ -0,0 +1,148 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _NET_MRP_H
+#define _NET_MRP_H
+
+#include <linux/netdevice.h>
+#include <linux/skbuff.h>
+#include <linux/types.h>
+
+#define MRP_END_MARK		0x0
+
+struct mrp_pdu_hdr {
+	u8	version;
+};
+
+struct mrp_msg_hdr {
+	u8	attrtype;
+	u8	attrlen;
+};
+
+struct mrp_vecattr_hdr {
+	__be16	lenflags;
+	unsigned char	firstattrvalue[];
+#define MRP_VECATTR_HDR_LEN_MASK cpu_to_be16(0x1FFF)
+#define MRP_VECATTR_HDR_FLAG_LA cpu_to_be16(0x2000)
+};
+
+enum mrp_vecattr_event {
+	MRP_VECATTR_EVENT_NEW,
+	MRP_VECATTR_EVENT_JOIN_IN,
+	MRP_VECATTR_EVENT_IN,
+	MRP_VECATTR_EVENT_JOIN_MT,
+	MRP_VECATTR_EVENT_MT,
+	MRP_VECATTR_EVENT_LV,
+	__MRP_VECATTR_EVENT_MAX
+};
+
+struct mrp_skb_cb {
+	struct mrp_msg_hdr	*mh;
+	struct mrp_vecattr_hdr	*vah;
+	unsigned char		attrvalue[];
+};
+
+static inline struct mrp_skb_cb *mrp_cb(struct sk_buff *skb)
+{
+	BUILD_BUG_ON(sizeof(struct mrp_skb_cb) >
+		     sizeof_field(struct sk_buff, cb));
+	return (struct mrp_skb_cb *)skb->cb;
+}
+
+enum mrp_applicant_state {
+	MRP_APPLICANT_INVALID,
+	MRP_APPLICANT_VO,
+	MRP_APPLICANT_VP,
+	MRP_APPLICANT_VN,
+	MRP_APPLICANT_AN,
+	MRP_APPLICANT_AA,
+	MRP_APPLICANT_QA,
+	MRP_APPLICANT_LA,
+	MRP_APPLICANT_AO,
+	MRP_APPLICANT_QO,
+	MRP_APPLICANT_AP,
+	MRP_APPLICANT_QP,
+	__MRP_APPLICANT_MAX
+};
+#define MRP_APPLICANT_MAX	(__MRP_APPLICANT_MAX - 1)
+
+enum mrp_event {
+	MRP_EVENT_NEW,
+	MRP_EVENT_JOIN,
+	MRP_EVENT_LV,
+	MRP_EVENT_TX,
+	MRP_EVENT_R_NEW,
+	MRP_EVENT_R_JOIN_IN,
+	MRP_EVENT_R_IN,
+	MRP_EVENT_R_JOIN_MT,
+	MRP_EVENT_R_MT,
+	MRP_EVENT_R_LV,
+	MRP_EVENT_R_LA,
+	MRP_EVENT_REDECLARE,
+	MRP_EVENT_PERIODIC,
+	__MRP_EVENT_MAX
+};
+#define MRP_EVENT_MAX		(__MRP_EVENT_MAX - 1)
+
+enum mrp_tx_action {
+	MRP_TX_ACTION_NONE,
+	MRP_TX_ACTION_S_NEW,
+	MRP_TX_ACTION_S_JOIN_IN,
+	MRP_TX_ACTION_S_JOIN_IN_OPTIONAL,
+	MRP_TX_ACTION_S_IN_OPTIONAL,
+	MRP_TX_ACTION_S_LV,
+};
+
+struct mrp_attr {
+	struct rb_node			node;
+	enum mrp_applicant_state	state;
+	u8				type;
+	u8				len;
+	unsigned char			value[];
+};
+
+enum mrp_applications {
+	MRP_APPLICATION_MVRP,
+	__MRP_APPLICATION_MAX
+};
+#define MRP_APPLICATION_MAX	(__MRP_APPLICATION_MAX - 1)
+
+struct mrp_application {
+	enum mrp_applications	type;
+	unsigned int		maxattr;
+	struct packet_type	pkttype;
+	unsigned char		group_address[ETH_ALEN];
+	u8			version;
+};
+
+struct mrp_applicant {
+	struct mrp_application	*app;
+	struct net_device	*dev;
+	struct timer_list	join_timer;
+	struct timer_list	periodic_timer;
+
+	spinlock_t		lock;
+	struct sk_buff_head	queue;
+	struct sk_buff		*pdu;
+	struct rb_root		mad;
+	struct rcu_head		rcu;
+	bool			active;
+};
+
+struct mrp_port {
+	struct mrp_applicant __rcu	*applicants[MRP_APPLICATION_MAX + 1];
+	struct rcu_head			rcu;
+};
+
+int mrp_register_application(struct mrp_application *app);
+void mrp_unregister_application(struct mrp_application *app);
+
+int mrp_init_applicant(struct net_device *dev, struct mrp_application *app);
+void mrp_uninit_applicant(struct net_device *dev, struct mrp_application *app);
+
+int mrp_request_join(const struct net_device *dev,
+		     const struct mrp_application *app,
+		     const void *value, u8 len, u8 type);
+void mrp_request_leave(const struct net_device *dev,
+		       const struct mrp_application *app,
+		       const void *value, u8 len, u8 type);
+
+#endif /* _NET_MRP_H */