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, 215 insertions, 0 deletions

diff --git a/net/rxrpc/server_key.c b/net/rxrpc/server_key.c
new file mode 100644
index 000000000..3efe104b1
--- /dev/null
+++ b/net/rxrpc/server_key.c
@@ -0,0 +1,215 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* RxRPC key management
+ *
+ * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ *
+ * RxRPC keys should have a description of describing their purpose:
+ *	"afs@CAMBRIDGE.REDHAT.COM>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/net.h>
+#include <linux/skbuff.h>
+#include <linux/key-type.h>
+#include <linux/ctype.h>
+#include <linux/slab.h>
+#include <net/sock.h>
+#include <net/af_rxrpc.h>
+#include <keys/rxrpc-type.h>
+#include <keys/user-type.h>
+#include "ar-internal.h"
+
+static int rxrpc_vet_description_s(const char *);
+static int rxrpc_preparse_s(struct key_preparsed_payload *);
+static void rxrpc_free_preparse_s(struct key_preparsed_payload *);
+static void rxrpc_destroy_s(struct key *);
+static void rxrpc_describe_s(const struct key *, struct seq_file *);
+
+/*
+ * rxrpc server keys take "<serviceId>:<securityIndex>[:<sec-specific>]" as the
+ * description and the key material as the payload.
+ */
+struct key_type key_type_rxrpc_s = {
+	.name		= "rxrpc_s",
+	.flags		= KEY_TYPE_NET_DOMAIN,
+	.vet_description = rxrpc_vet_description_s,
+	.preparse	= rxrpc_preparse_s,
+	.free_preparse	= rxrpc_free_preparse_s,
+	.instantiate	= generic_key_instantiate,
+	.destroy	= rxrpc_destroy_s,
+	.describe	= rxrpc_describe_s,
+};
+
+/*
+ * Vet the description for an RxRPC server key.
+ */
+static int rxrpc_vet_description_s(const char *desc)
+{
+	unsigned long service, sec_class;
+	char *p;
+
+	service = simple_strtoul(desc, &p, 10);
+	if (*p != ':' || service > 65535)
+		return -EINVAL;
+	sec_class = simple_strtoul(p + 1, &p, 10);
+	if ((*p && *p != ':') || sec_class < 1 || sec_class > 255)
+		return -EINVAL;
+	return 0;
+}
+
+/*
+ * Preparse a server secret key.
+ */
+static int rxrpc_preparse_s(struct key_preparsed_payload *prep)
+{
+	const struct rxrpc_security *sec;
+	unsigned int service, sec_class;
+	int n;
+
+	_enter("%zu", prep->datalen);
+
+	if (!prep->orig_description)
+		return -EINVAL;
+
+	if (sscanf(prep->orig_description, "%u:%u%n", &service, &sec_class, &n) != 2)
+		return -EINVAL;
+
+	sec = rxrpc_security_lookup(sec_class);
+	if (!sec)
+		return -ENOPKG;
+
+	prep->payload.data[1] = (struct rxrpc_security *)sec;
+
+	if (!sec->preparse_server_key)
+		return -EINVAL;
+
+	return sec->preparse_server_key(prep);
+}
+
+static void rxrpc_free_preparse_s(struct key_preparsed_payload *prep)
+{
+	const struct rxrpc_security *sec = prep->payload.data[1];
+
+	if (sec && sec->free_preparse_server_key)
+		sec->free_preparse_server_key(prep);
+}
+
+static void rxrpc_destroy_s(struct key *key)
+{
+	const struct rxrpc_security *sec = key->payload.data[1];
+
+	if (sec && sec->destroy_server_key)
+		sec->destroy_server_key(key);
+}
+
+static void rxrpc_describe_s(const struct key *key, struct seq_file *m)
+{
+	const struct rxrpc_security *sec = key->payload.data[1];
+
+	seq_puts(m, key->description);
+	if (sec && sec->describe_server_key)
+		sec->describe_server_key(key, m);
+}
+
+/*
+ * grab the security keyring for a server socket
+ */
+int rxrpc_server_keyring(struct rxrpc_sock *rx, sockptr_t optval, int optlen)
+{
+	struct key *key;
+	char *description;
+
+	_enter("");
+
+	if (rx->securities)
+		return -EINVAL;
+
+	if (optlen <= 0 || optlen > PAGE_SIZE - 1)
+		return -EINVAL;
+
+	description = memdup_sockptr_nul(optval, optlen);
+	if (IS_ERR(description))
+		return PTR_ERR(description);
+
+	key = request_key(&key_type_keyring, description, NULL);
+	if (IS_ERR(key)) {
+		kfree(description);
+		_leave(" = %ld", PTR_ERR(key));
+		return PTR_ERR(key);
+	}
+
+	rx->securities = key;
+	kfree(description);
+	_leave(" = 0 [key %x]", key->serial);
+	return 0;
+}
+
+/**
+ * rxrpc_sock_set_security_keyring - Set the security keyring for a kernel service
+ * @sk: The socket to set the keyring on
+ * @keyring: The keyring to set
+ *
+ * Set the server security keyring on an rxrpc socket.  This is used to provide
+ * the encryption keys for a kernel service.
+ *
+ * Return: %0 if successful and a negative error code otherwise.
+ */
+int rxrpc_sock_set_security_keyring(struct sock *sk, struct key *keyring)
+{
+	struct rxrpc_sock *rx = rxrpc_sk(sk);
+	int ret = 0;
+
+	lock_sock(sk);
+	if (rx->securities)
+		ret = -EINVAL;
+	else if (rx->sk.sk_state != RXRPC_UNBOUND)
+		ret = -EISCONN;
+	else
+		rx->securities = key_get(keyring);
+	release_sock(sk);
+	return ret;
+}
+EXPORT_SYMBOL(rxrpc_sock_set_security_keyring);
+
+/**
+ * rxrpc_sock_set_manage_response - Set the manage-response flag for a kernel service
+ * @sk: The socket to set the keyring on
+ * @set: True to set, false to clear the flag
+ *
+ * Set the flag on an rxrpc socket to say that the caller wants to manage the
+ * RESPONSE packet and the user-defined data it may contain.  Setting this
+ * means that recvmsg() will return messages with RXRPC_CHALLENGED in the
+ * control message buffer containing information about the challenge.
+ *
+ * The user should respond to the challenge by passing RXRPC_RESPOND or
+ * RXRPC_RESPOND_ABORT control messages with sendmsg() to the same call.
+ * Supplementary control messages, such as RXRPC_RESP_RXGK_APPDATA, may be
+ * included to indicate the parts the user wants to supply.
+ *
+ * The server will be passed the response data with a RXRPC_RESPONDED control
+ * message when it gets the first data from each call.
+ *
+ * Note that this is only honoured by security classes that need auxiliary data
+ * (e.g. RxGK).  Those that don't offer the facility (e.g. RxKAD) respond
+ * without consulting userspace.
+ *
+ * Return: The previous setting.
+ */
+int rxrpc_sock_set_manage_response(struct sock *sk, bool set)
+{
+	struct rxrpc_sock *rx = rxrpc_sk(sk);
+	int ret;
+
+	lock_sock(sk);
+	ret = !!test_bit(RXRPC_SOCK_MANAGE_RESPONSE, &rx->flags);
+	if (set)
+		set_bit(RXRPC_SOCK_MANAGE_RESPONSE, &rx->flags);
+	else
+		clear_bit(RXRPC_SOCK_MANAGE_RESPONSE, &rx->flags);
+	release_sock(sk);
+	return ret;
+}
+EXPORT_SYMBOL(rxrpc_sock_set_manage_response);