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, 172 insertions, 0 deletions

diff --git a/net/ipv4/tcp_ulp.c b/net/ipv4/tcp_ulp.c
new file mode 100644
index 000000000..b58045df1
--- /dev/null
+++ b/net/ipv4/tcp_ulp.c
@@ -0,0 +1,172 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Pluggable TCP upper layer protocol support.
+ *
+ * Copyright (c) 2016-2017, Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2016-2017, Dave Watson <davejwatson@fb.com>. All rights reserved.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/mm.h>
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/gfp.h>
+#include <net/tcp.h>
+
+static DEFINE_SPINLOCK(tcp_ulp_list_lock);
+static LIST_HEAD(tcp_ulp_list);
+
+/* Simple linear search, don't expect many entries! */
+static struct tcp_ulp_ops *tcp_ulp_find(const char *name)
+{
+	struct tcp_ulp_ops *e;
+
+	list_for_each_entry_rcu(e, &tcp_ulp_list, list,
+				lockdep_is_held(&tcp_ulp_list_lock)) {
+		if (strcmp(e->name, name) == 0)
+			return e;
+	}
+
+	return NULL;
+}
+
+static const struct tcp_ulp_ops *__tcp_ulp_find_autoload(const char *name)
+{
+	const struct tcp_ulp_ops *ulp = NULL;
+
+	rcu_read_lock();
+	ulp = tcp_ulp_find(name);
+
+#ifdef CONFIG_MODULES
+	if (!ulp && capable(CAP_NET_ADMIN)) {
+		rcu_read_unlock();
+		request_module("tcp-ulp-%s", name);
+		rcu_read_lock();
+		ulp = tcp_ulp_find(name);
+	}
+#endif
+	if (!ulp || !try_module_get(ulp->owner))
+		ulp = NULL;
+
+	rcu_read_unlock();
+	return ulp;
+}
+
+/* Attach new upper layer protocol to the list
+ * of available protocols.
+ */
+int tcp_register_ulp(struct tcp_ulp_ops *ulp)
+{
+	int ret = 0;
+
+	spin_lock(&tcp_ulp_list_lock);
+	if (tcp_ulp_find(ulp->name))
+		ret = -EEXIST;
+	else
+		list_add_tail_rcu(&ulp->list, &tcp_ulp_list);
+	spin_unlock(&tcp_ulp_list_lock);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(tcp_register_ulp);
+
+void tcp_unregister_ulp(struct tcp_ulp_ops *ulp)
+{
+	spin_lock(&tcp_ulp_list_lock);
+	list_del_rcu(&ulp->list);
+	spin_unlock(&tcp_ulp_list_lock);
+
+	synchronize_rcu();
+}
+EXPORT_SYMBOL_GPL(tcp_unregister_ulp);
+
+/* Build string with list of available upper layer protocl values */
+void tcp_get_available_ulp(char *buf, size_t maxlen)
+{
+	struct tcp_ulp_ops *ulp_ops;
+	size_t offs = 0;
+
+	*buf = '\0';
+	rcu_read_lock();
+	list_for_each_entry_rcu(ulp_ops, &tcp_ulp_list, list) {
+		offs += snprintf(buf + offs, maxlen - offs,
+				 "%s%s",
+				 offs == 0 ? "" : " ", ulp_ops->name);
+
+		if (WARN_ON_ONCE(offs >= maxlen))
+			break;
+	}
+	rcu_read_unlock();
+}
+
+void tcp_update_ulp(struct sock *sk, struct proto *proto,
+		    void (*write_space)(struct sock *sk))
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	if (icsk->icsk_ulp_ops->update)
+		icsk->icsk_ulp_ops->update(sk, proto, write_space);
+}
+
+void tcp_cleanup_ulp(struct sock *sk)
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+
+	/* No sock_owned_by_me() check here as at the time the
+	 * stack calls this function, the socket is dead and
+	 * about to be destroyed.
+	 */
+	if (!icsk->icsk_ulp_ops)
+		return;
+
+	if (icsk->icsk_ulp_ops->release)
+		icsk->icsk_ulp_ops->release(sk);
+	module_put(icsk->icsk_ulp_ops->owner);
+
+	icsk->icsk_ulp_ops = NULL;
+}
+
+static int __tcp_set_ulp(struct sock *sk, const struct tcp_ulp_ops *ulp_ops)
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	int err;
+
+	err = -EEXIST;
+	if (icsk->icsk_ulp_ops)
+		goto out_err;
+
+	err = -EINVAL;
+	if (sk_has_decrypt_user(sk))
+		goto out_err;
+
+	if (sk->sk_socket)
+		clear_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags);
+
+	err = -ENOTCONN;
+	if (!ulp_ops->clone && sk->sk_state == TCP_LISTEN)
+		goto out_err;
+
+	err = ulp_ops->init(sk);
+	if (err)
+		goto out_err;
+
+	icsk->icsk_ulp_ops = ulp_ops;
+	return 0;
+out_err:
+	module_put(ulp_ops->owner);
+	return err;
+}
+
+int tcp_set_ulp(struct sock *sk, const char *name)
+{
+	const struct tcp_ulp_ops *ulp_ops;
+
+	sock_owned_by_me(sk);
+
+	ulp_ops = __tcp_ulp_find_autoload(name);
+	if (!ulp_ops)
+		return -ENOENT;
+
+	return __tcp_set_ulp(sk, ulp_ops);
+}