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, 253 insertions, 0 deletions

diff --git a/include/net/request_sock.h b/include/net/request_sock.h
new file mode 100644
index 000000000..5a9c826a7
--- /dev/null
+++ b/include/net/request_sock.h
@@ -0,0 +1,253 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * NET		Generic infrastructure for Network protocols.
+ *
+ *		Definitions for request_sock
+ *
+ * Authors:	Arnaldo Carvalho de Melo <acme@conectiva.com.br>
+ *
+ * 		From code originally in include/net/tcp.h
+ */
+#ifndef _REQUEST_SOCK_H
+#define _REQUEST_SOCK_H
+
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+#include <linux/bug.h>
+#include <linux/refcount.h>
+
+#include <net/sock.h>
+#include <net/rstreason.h>
+
+struct request_sock;
+struct sk_buff;
+struct dst_entry;
+struct proto;
+
+struct request_sock_ops {
+	int		family;
+	unsigned int	obj_size;
+	struct kmem_cache	*slab;
+	char		*slab_name;
+	void		(*send_ack)(const struct sock *sk, struct sk_buff *skb,
+				    struct request_sock *req);
+	void		(*send_reset)(const struct sock *sk,
+				      struct sk_buff *skb,
+				      enum sk_rst_reason reason);
+	void		(*destructor)(struct request_sock *req);
+};
+
+struct saved_syn {
+	u32 mac_hdrlen;
+	u32 network_hdrlen;
+	u32 tcp_hdrlen;
+	u8 data[];
+};
+
+/* struct request_sock - mini sock to represent a connection request
+ */
+struct request_sock {
+	struct sock_common		__req_common;
+#define rsk_refcnt			__req_common.skc_refcnt
+#define rsk_hash			__req_common.skc_hash
+#define rsk_listener			__req_common.skc_listener
+#define rsk_window_clamp		__req_common.skc_window_clamp
+#define rsk_rcv_wnd			__req_common.skc_rcv_wnd
+
+	struct request_sock		*dl_next;
+	u16				mss;
+	u8				num_retrans; /* number of retransmits */
+	u8				syncookie:1; /* True if
+						      * 1) tcpopts needs to be encoded in
+						      *    TS of SYN+ACK
+						      * 2) ACK is validated by BPF kfunc.
+						      */
+	u8				num_timeout:7; /* number of timeouts */
+	u32				ts_recent;
+	struct timer_list		rsk_timer;
+	const struct request_sock_ops	*rsk_ops;
+	struct sock			*sk;
+	struct saved_syn		*saved_syn;
+	u32				secid;
+	u32				peer_secid;
+	u32				timeout;
+};
+
+static inline struct request_sock *inet_reqsk(const struct sock *sk)
+{
+	return (struct request_sock *)sk;
+}
+
+static inline struct sock *req_to_sk(struct request_sock *req)
+{
+	return (struct sock *)req;
+}
+
+/**
+ * skb_steal_sock - steal a socket from an sk_buff
+ * @skb: sk_buff to steal the socket from
+ * @refcounted: is set to true if the socket is reference-counted
+ * @prefetched: is set to true if the socket was assigned from bpf
+ */
+static inline struct sock *skb_steal_sock(struct sk_buff *skb,
+					  bool *refcounted, bool *prefetched)
+{
+	struct sock *sk = skb->sk;
+
+	if (!sk) {
+		*prefetched = false;
+		*refcounted = false;
+		return NULL;
+	}
+
+	*prefetched = skb_sk_is_prefetched(skb);
+	if (*prefetched) {
+#if IS_ENABLED(CONFIG_SYN_COOKIES)
+		if (sk->sk_state == TCP_NEW_SYN_RECV && inet_reqsk(sk)->syncookie) {
+			struct request_sock *req = inet_reqsk(sk);
+
+			*refcounted = false;
+			sk = req->rsk_listener;
+			req->rsk_listener = NULL;
+			return sk;
+		}
+#endif
+		*refcounted = sk_is_refcounted(sk);
+	} else {
+		*refcounted = true;
+	}
+
+	skb->destructor = NULL;
+	skb->sk = NULL;
+	return sk;
+}
+
+void __reqsk_free(struct request_sock *req);
+
+static inline void reqsk_free(struct request_sock *req)
+{
+	DEBUG_NET_WARN_ON_ONCE(refcount_read(&req->rsk_refcnt) != 0);
+	__reqsk_free(req);
+}
+
+static inline void reqsk_put(struct request_sock *req)
+{
+	if (refcount_dec_and_test(&req->rsk_refcnt))
+		__reqsk_free(req);
+}
+
+/*
+ * For a TCP Fast Open listener -
+ *	lock - protects the access to all the reqsk, which is co-owned by
+ *		the listener and the child socket.
+ *	qlen - pending TFO requests (still in TCP_SYN_RECV).
+ *	max_qlen - max TFO reqs allowed before TFO is disabled.
+ *
+ *	XXX (TFO) - ideally these fields can be made as part of "listen_sock"
+ *	structure above. But there is some implementation difficulty due to
+ *	listen_sock being part of request_sock_queue hence will be freed when
+ *	a listener is stopped. But TFO related fields may continue to be
+ *	accessed even after a listener is closed, until its sk_refcnt drops
+ *	to 0 implying no more outstanding TFO reqs. One solution is to keep
+ *	listen_opt around until	sk_refcnt drops to 0. But there is some other
+ *	complexity that needs to be resolved. E.g., a listener can be disabled
+ *	temporarily through shutdown()->tcp_disconnect(), and re-enabled later.
+ */
+struct fastopen_queue {
+	struct request_sock	*rskq_rst_head; /* Keep track of past TFO */
+	struct request_sock	*rskq_rst_tail; /* requests that caused RST.
+						 * This is part of the defense
+						 * against spoofing attack.
+						 */
+	spinlock_t	lock;
+	int		qlen;		/* # of pending (TCP_SYN_RECV) reqs */
+	int		max_qlen;	/* != 0 iff TFO is currently enabled */
+
+	struct tcp_fastopen_context __rcu *ctx; /* cipher context for cookie */
+};
+
+/** struct request_sock_queue - queue of request_socks
+ *
+ * @rskq_accept_head - FIFO head of established children
+ * @rskq_accept_tail - FIFO tail of established children
+ * @rskq_defer_accept - User waits for some data after accept()
+ *
+ */
+struct request_sock_queue {
+	spinlock_t		rskq_lock;
+	u8			rskq_defer_accept;
+	u8			synflood_warned;
+
+	atomic_t		qlen;
+	atomic_t		young;
+
+	struct request_sock	*rskq_accept_head;
+	struct request_sock	*rskq_accept_tail;
+	struct fastopen_queue	fastopenq;  /* Check max_qlen != 0 to determine
+					     * if TFO is enabled.
+					     */
+};
+
+void reqsk_fastopen_remove(struct sock *sk, struct request_sock *req,
+			   bool reset);
+
+static inline bool reqsk_queue_empty(const struct request_sock_queue *queue)
+{
+	return READ_ONCE(queue->rskq_accept_head) == NULL;
+}
+
+static inline struct request_sock *reqsk_queue_remove(struct request_sock_queue *queue,
+						      struct sock *parent)
+{
+	struct request_sock *req;
+
+	spin_lock_bh(&queue->rskq_lock);
+	req = queue->rskq_accept_head;
+	if (req) {
+		sk_acceptq_removed(parent);
+		WRITE_ONCE(queue->rskq_accept_head, req->dl_next);
+		if (queue->rskq_accept_head == NULL)
+			queue->rskq_accept_tail = NULL;
+	}
+	spin_unlock_bh(&queue->rskq_lock);
+	return req;
+}
+
+static inline void reqsk_queue_removed(struct request_sock_queue *queue,
+				       const struct request_sock *req)
+{
+	if (req->num_timeout == 0)
+		atomic_dec(&queue->young);
+	atomic_dec(&queue->qlen);
+}
+
+static inline void reqsk_queue_added(struct request_sock_queue *queue)
+{
+	atomic_inc(&queue->young);
+	atomic_inc(&queue->qlen);
+}
+
+static inline int reqsk_queue_len(const struct request_sock_queue *queue)
+{
+	return atomic_read(&queue->qlen);
+}
+
+static inline int reqsk_queue_len_young(const struct request_sock_queue *queue)
+{
+	return atomic_read(&queue->young);
+}
+
+/* RFC 7323 2.3 Using the Window Scale Option
+ *  The window field (SEG.WND) of every outgoing segment, with the
+ *  exception of <SYN> segments, MUST be right-shifted by
+ *  Rcv.Wind.Shift bits.
+ *
+ * This means the SEG.WND carried in SYNACK can not exceed 65535.
+ * We use this property to harden TCP stack while in NEW_SYN_RECV state.
+ */
+static inline u32 tcp_synack_window(const struct request_sock *req)
+{
+	return min(req->rsk_rcv_wnd, 65535U);
+}
+#endif /* _REQUEST_SOCK_H */