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, 211 insertions, 0 deletions

diff --git a/net/l2tp/trace.h b/net/l2tp/trace.h
new file mode 100644
index 000000000..8596eaa12
--- /dev/null
+++ b/net/l2tp/trace.h
@@ -0,0 +1,211 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM l2tp
+
+#if !defined(_TRACE_L2TP_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_L2TP_H
+
+#include <linux/tracepoint.h>
+#include <linux/l2tp.h>
+#include "l2tp_core.h"
+
+#define encap_type_name(e) { L2TP_ENCAPTYPE_##e, #e }
+#define show_encap_type_name(val) \
+	__print_symbolic(val, \
+			encap_type_name(UDP), \
+			encap_type_name(IP))
+
+#define pw_type_name(p) { L2TP_PWTYPE_##p, #p }
+#define show_pw_type_name(val) \
+	__print_symbolic(val, \
+	pw_type_name(ETH_VLAN), \
+	pw_type_name(ETH), \
+	pw_type_name(PPP), \
+	pw_type_name(PPP_AC), \
+	pw_type_name(IP))
+
+DECLARE_EVENT_CLASS(tunnel_only_evt,
+	TP_PROTO(struct l2tp_tunnel *tunnel),
+	TP_ARGS(tunnel),
+	TP_STRUCT__entry(
+		__array(char, name, L2TP_TUNNEL_NAME_MAX)
+	),
+	TP_fast_assign(
+		memcpy(__entry->name, tunnel->name, L2TP_TUNNEL_NAME_MAX);
+	),
+	TP_printk("%s", __entry->name)
+);
+
+DECLARE_EVENT_CLASS(session_only_evt,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session),
+	TP_STRUCT__entry(
+		__array(char, name, L2TP_SESSION_NAME_MAX)
+	),
+	TP_fast_assign(
+		memcpy(__entry->name, session->name, L2TP_SESSION_NAME_MAX);
+	),
+	TP_printk("%s", __entry->name)
+);
+
+TRACE_EVENT(register_tunnel,
+	TP_PROTO(struct l2tp_tunnel *tunnel),
+	TP_ARGS(tunnel),
+	TP_STRUCT__entry(
+		__array(char, name, L2TP_TUNNEL_NAME_MAX)
+		__field(int, fd)
+		__field(u32, tid)
+		__field(u32, ptid)
+		__field(int, version)
+		__field(enum l2tp_encap_type, encap)
+	),
+	TP_fast_assign(
+		memcpy(__entry->name, tunnel->name, L2TP_TUNNEL_NAME_MAX);
+		__entry->fd = tunnel->fd;
+		__entry->tid = tunnel->tunnel_id;
+		__entry->ptid = tunnel->peer_tunnel_id;
+		__entry->version = tunnel->version;
+		__entry->encap = tunnel->encap;
+	),
+	TP_printk("%s: type=%s encap=%s version=L2TPv%d tid=%u ptid=%u fd=%d",
+		__entry->name,
+		__entry->fd > 0 ? "managed" : "unmanaged",
+		show_encap_type_name(__entry->encap),
+		__entry->version,
+		__entry->tid,
+		__entry->ptid,
+		__entry->fd)
+);
+
+DEFINE_EVENT(tunnel_only_evt, delete_tunnel,
+	TP_PROTO(struct l2tp_tunnel *tunnel),
+	TP_ARGS(tunnel)
+);
+
+DEFINE_EVENT(tunnel_only_evt, free_tunnel,
+	TP_PROTO(struct l2tp_tunnel *tunnel),
+	TP_ARGS(tunnel)
+);
+
+TRACE_EVENT(register_session,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session),
+	TP_STRUCT__entry(
+		__array(char, name, L2TP_SESSION_NAME_MAX)
+		__field(u32, tid)
+		__field(u32, ptid)
+		__field(u32, sid)
+		__field(u32, psid)
+		__field(enum l2tp_pwtype, pwtype)
+	),
+	TP_fast_assign(
+		memcpy(__entry->name, session->name, L2TP_SESSION_NAME_MAX);
+		__entry->tid = session->tunnel ? session->tunnel->tunnel_id : 0;
+		__entry->ptid = session->tunnel ? session->tunnel->peer_tunnel_id : 0;
+		__entry->sid = session->session_id;
+		__entry->psid = session->peer_session_id;
+		__entry->pwtype = session->pwtype;
+	),
+	TP_printk("%s: pseudowire=%s sid=%u psid=%u tid=%u ptid=%u",
+		__entry->name,
+		show_pw_type_name(__entry->pwtype),
+		__entry->sid,
+		__entry->psid,
+		__entry->sid,
+		__entry->psid)
+);
+
+DEFINE_EVENT(session_only_evt, delete_session,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session)
+);
+
+DEFINE_EVENT(session_only_evt, free_session,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session)
+);
+
+DEFINE_EVENT(session_only_evt, session_seqnum_lns_enable,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session)
+);
+
+DEFINE_EVENT(session_only_evt, session_seqnum_lns_disable,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session)
+);
+
+DECLARE_EVENT_CLASS(session_seqnum_evt,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session),
+	TP_STRUCT__entry(
+		__array(char, name, L2TP_SESSION_NAME_MAX)
+		__field(u32, ns)
+		__field(u32, nr)
+	),
+	TP_fast_assign(
+		memcpy(__entry->name, session->name, L2TP_SESSION_NAME_MAX);
+		__entry->ns = session->ns;
+		__entry->nr = session->nr;
+	),
+	TP_printk("%s: ns=%u nr=%u",
+		__entry->name,
+		__entry->ns,
+		__entry->nr)
+);
+
+DEFINE_EVENT(session_seqnum_evt, session_seqnum_update,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session)
+);
+
+DEFINE_EVENT(session_seqnum_evt, session_seqnum_reset,
+	TP_PROTO(struct l2tp_session *session),
+	TP_ARGS(session)
+);
+
+DECLARE_EVENT_CLASS(session_pkt_discard_evt,
+	TP_PROTO(struct l2tp_session *session, u32 pkt_ns),
+	TP_ARGS(session, pkt_ns),
+	TP_STRUCT__entry(
+		__array(char, name, L2TP_SESSION_NAME_MAX)
+		__field(u32, pkt_ns)
+		__field(u32, my_nr)
+		__field(u32, reorder_q_len)
+	),
+	TP_fast_assign(
+		memcpy(__entry->name, session->name, L2TP_SESSION_NAME_MAX);
+		__entry->pkt_ns = pkt_ns,
+		__entry->my_nr = session->nr;
+		__entry->reorder_q_len = skb_queue_len(&session->reorder_q);
+	),
+	TP_printk("%s: pkt_ns=%u my_nr=%u reorder_q_len=%u",
+		__entry->name,
+		__entry->pkt_ns,
+		__entry->my_nr,
+		__entry->reorder_q_len)
+);
+
+DEFINE_EVENT(session_pkt_discard_evt, session_pkt_expired,
+	TP_PROTO(struct l2tp_session *session, u32 pkt_ns),
+	TP_ARGS(session, pkt_ns)
+);
+
+DEFINE_EVENT(session_pkt_discard_evt, session_pkt_outside_rx_window,
+	TP_PROTO(struct l2tp_session *session, u32 pkt_ns),
+	TP_ARGS(session, pkt_ns)
+);
+
+DEFINE_EVENT(session_pkt_discard_evt, session_pkt_oos,
+	TP_PROTO(struct l2tp_session *session, u32 pkt_ns),
+	TP_ARGS(session, pkt_ns)
+);
+
+#endif /* _TRACE_L2TP_H */
+
+/* This part must be outside protection */
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH .
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_FILE trace
+#include <trace/define_trace.h>