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, 62 insertions, 0 deletions

diff --git a/net/rxrpc/misc.c b/net/rxrpc/misc.c
new file mode 100644
index 000000000..8fcc8139d
--- /dev/null
+++ b/net/rxrpc/misc.c
@@ -0,0 +1,62 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* Miscellaneous bits
+ *
+ * Copyright (C) 2016 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#include <linux/kernel.h>
+#include <net/sock.h>
+#include <net/af_rxrpc.h>
+#include "ar-internal.h"
+
+/*
+ * The maximum listening backlog queue size that may be set on a socket by
+ * listen().
+ */
+unsigned int rxrpc_max_backlog __read_mostly = 10;
+
+/*
+ * How long to wait before scheduling an ACK with subtype DELAY (in ms).
+ *
+ * We use this when we've received new data packets.  If those packets aren't
+ * all consumed within this time we will send a DELAY ACK if an ACK was not
+ * requested to let the sender know it doesn't need to resend.
+ */
+unsigned long rxrpc_soft_ack_delay = 1000;
+
+/*
+ * How long to wait before scheduling an ACK with subtype IDLE (in ms).
+ *
+ * We use this when we've consumed some previously soft-ACK'd packets when
+ * further packets aren't immediately received to decide when to send an IDLE
+ * ACK let the other end know that it can free up its Tx buffer space.
+ */
+unsigned long rxrpc_idle_ack_delay = 500;
+
+/*
+ * Receive window size in packets.  This indicates the maximum number of
+ * unconsumed received packets we're willing to retain in memory.  Once this
+ * limit is hit, we should generate an EXCEEDS_WINDOW ACK and discard further
+ * packets.
+ */
+unsigned int rxrpc_rx_window_size = 255;
+
+/*
+ * Maximum Rx MTU size.  This indicates to the sender the size of jumbo packet
+ * made by gluing normal packets together that we're willing to handle.
+ */
+unsigned int rxrpc_rx_mtu = RXRPC_JUMBO(46);
+
+/*
+ * The maximum number of fragments in a received jumbo packet that we tell the
+ * sender that we're willing to handle.
+ */
+unsigned int rxrpc_rx_jumbo_max = 46;
+
+#ifdef CONFIG_AF_RXRPC_INJECT_RX_DELAY
+/*
+ * The delay to inject into packet reception.
+ */
+unsigned long rxrpc_inject_rx_delay;
+#endif