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, 202 insertions, 0 deletions

diff --git a/net/tls/trace.h b/net/tls/trace.h
new file mode 100644
index 000000000..2d8ce4ff3
--- /dev/null
+++ b/net/tls/trace.h
@@ -0,0 +1,202 @@
+/* SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) */
+/* Copyright (C) 2019 Netronome Systems, Inc. */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM tls
+
+#if !defined(_TLS_TRACE_H_) || defined(TRACE_HEADER_MULTI_READ)
+#define _TLS_TRACE_H_
+
+#include <linux/unaligned.h>
+#include <linux/tracepoint.h>
+
+struct sock;
+
+TRACE_EVENT(tls_device_offload_set,
+
+	TP_PROTO(struct sock *sk, int dir, u32 tcp_seq, u8 *rec_no, int ret),
+
+	TP_ARGS(sk, dir, tcp_seq, rec_no, ret),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+		__field(	u64,		rec_no		)
+		__field(	int,		dir		)
+		__field(	u32,		tcp_seq		)
+		__field(	int,		ret		)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+		__entry->rec_no = get_unaligned_be64(rec_no);
+		__entry->dir = dir;
+		__entry->tcp_seq = tcp_seq;
+		__entry->ret = ret;
+	),
+
+	TP_printk(
+		"sk=%p direction=%d tcp_seq=%u rec_no=%llu ret=%d",
+		__entry->sk, __entry->dir, __entry->tcp_seq, __entry->rec_no,
+		__entry->ret
+	)
+);
+
+TRACE_EVENT(tls_device_decrypted,
+
+	TP_PROTO(struct sock *sk, u32 tcp_seq, u8 *rec_no, u32 rec_len,
+		 bool encrypted, bool decrypted),
+
+	TP_ARGS(sk, tcp_seq, rec_no, rec_len, encrypted, decrypted),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+		__field(	u64,		rec_no		)
+		__field(	u32,		tcp_seq		)
+		__field(	u32,		rec_len		)
+		__field(	bool,		encrypted	)
+		__field(	bool,		decrypted	)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+		__entry->rec_no = get_unaligned_be64(rec_no);
+		__entry->tcp_seq = tcp_seq;
+		__entry->rec_len = rec_len;
+		__entry->encrypted = encrypted;
+		__entry->decrypted = decrypted;
+	),
+
+	TP_printk(
+		"sk=%p tcp_seq=%u rec_no=%llu len=%u encrypted=%d decrypted=%d",
+		__entry->sk, __entry->tcp_seq,
+		__entry->rec_no, __entry->rec_len,
+		__entry->encrypted, __entry->decrypted
+	)
+);
+
+TRACE_EVENT(tls_device_rx_resync_send,
+
+	TP_PROTO(struct sock *sk, u32 tcp_seq, u8 *rec_no, int sync_type),
+
+	TP_ARGS(sk, tcp_seq, rec_no, sync_type),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+		__field(	u64,		rec_no		)
+		__field(	u32,		tcp_seq		)
+		__field(	int,		sync_type	)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+		__entry->rec_no = get_unaligned_be64(rec_no);
+		__entry->tcp_seq = tcp_seq;
+		__entry->sync_type = sync_type;
+	),
+
+	TP_printk(
+		"sk=%p tcp_seq=%u rec_no=%llu sync_type=%d",
+		__entry->sk, __entry->tcp_seq, __entry->rec_no,
+		__entry->sync_type
+	)
+);
+
+TRACE_EVENT(tls_device_rx_resync_nh_schedule,
+
+	TP_PROTO(struct sock *sk),
+
+	TP_ARGS(sk),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+	),
+
+	TP_printk(
+		"sk=%p", __entry->sk
+	)
+);
+
+TRACE_EVENT(tls_device_rx_resync_nh_delay,
+
+	TP_PROTO(struct sock *sk, u32 sock_data, u32 rec_len),
+
+	TP_ARGS(sk, sock_data, rec_len),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+		__field(	u32,		sock_data	)
+		__field(	u32,		rec_len		)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+		__entry->sock_data = sock_data;
+		__entry->rec_len = rec_len;
+	),
+
+	TP_printk(
+		"sk=%p sock_data=%u rec_len=%u",
+		__entry->sk, __entry->sock_data, __entry->rec_len
+	)
+);
+
+TRACE_EVENT(tls_device_tx_resync_req,
+
+	TP_PROTO(struct sock *sk, u32 tcp_seq, u32 exp_tcp_seq),
+
+	TP_ARGS(sk, tcp_seq, exp_tcp_seq),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+		__field(	u32,		tcp_seq		)
+		__field(	u32,		exp_tcp_seq	)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+		__entry->tcp_seq = tcp_seq;
+		__entry->exp_tcp_seq = exp_tcp_seq;
+	),
+
+	TP_printk(
+		"sk=%p tcp_seq=%u exp_tcp_seq=%u",
+		__entry->sk, __entry->tcp_seq, __entry->exp_tcp_seq
+	)
+);
+
+TRACE_EVENT(tls_device_tx_resync_send,
+
+	TP_PROTO(struct sock *sk, u32 tcp_seq, u8 *rec_no),
+
+	TP_ARGS(sk, tcp_seq, rec_no),
+
+	TP_STRUCT__entry(
+		__field(	struct sock *,	sk		)
+		__field(	u64,		rec_no		)
+		__field(	u32,		tcp_seq		)
+	),
+
+	TP_fast_assign(
+		__entry->sk = sk;
+		__entry->rec_no = get_unaligned_be64(rec_no);
+		__entry->tcp_seq = tcp_seq;
+	),
+
+	TP_printk(
+		"sk=%p tcp_seq=%u rec_no=%llu",
+		__entry->sk, __entry->tcp_seq, __entry->rec_no
+	)
+);
+
+#endif /* _TLS_TRACE_H_ */
+
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH .
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_FILE trace
+
+#include <trace/define_trace.h>