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, 112 insertions, 0 deletions

diff --git a/net/mac80211/rate.h b/net/mac80211/rate.h
new file mode 100644
index 000000000..5e4bde598
--- /dev/null
+++ b/net/mac80211/rate.h
@@ -0,0 +1,112 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright 2002-2005, Instant802 Networks, Inc.
+ * Copyright 2005, Devicescape Software, Inc.
+ * Copyright (c) 2006 Jiri Benc <jbenc@suse.cz>
+ * Copyright (C) 2022, 2024 Intel Corporation
+ */
+
+#ifndef IEEE80211_RATE_H
+#define IEEE80211_RATE_H
+
+#include <linux/netdevice.h>
+#include <linux/skbuff.h>
+#include <linux/types.h>
+#include <net/mac80211.h>
+#include "ieee80211_i.h"
+#include "sta_info.h"
+#include "driver-ops.h"
+
+struct rate_control_ref {
+	const struct rate_control_ops *ops;
+	void *priv;
+};
+
+void rate_control_get_rate(struct ieee80211_sub_if_data *sdata,
+			   struct sta_info *sta,
+			   struct ieee80211_tx_rate_control *txrc);
+
+void rate_control_tx_status(struct ieee80211_local *local,
+			    struct ieee80211_tx_status *st);
+
+void rate_control_rate_init(struct link_sta_info *link_sta);
+void rate_control_rate_init_all_links(struct sta_info *sta);
+void rate_control_rate_update(struct ieee80211_local *local,
+			      struct ieee80211_supported_band *sband,
+			      struct link_sta_info *link_sta,
+			      u32 changed);
+
+static inline void *rate_control_alloc_sta(struct rate_control_ref *ref,
+					   struct sta_info *sta, gfp_t gfp)
+{
+	spin_lock_init(&sta->rate_ctrl_lock);
+	return ref->ops->alloc_sta(ref->priv, &sta->sta, gfp);
+}
+
+static inline void rate_control_free_sta(struct sta_info *sta)
+{
+	struct rate_control_ref *ref = sta->rate_ctrl;
+	struct ieee80211_sta *ista = &sta->sta;
+	void *priv_sta = sta->rate_ctrl_priv;
+
+	ref->ops->free_sta(ref->priv, ista, priv_sta);
+}
+
+static inline void rate_control_add_sta_debugfs(struct sta_info *sta)
+{
+#ifdef CONFIG_MAC80211_DEBUGFS
+	struct rate_control_ref *ref = sta->rate_ctrl;
+	if (ref && sta->debugfs_dir && ref->ops->add_sta_debugfs)
+		ref->ops->add_sta_debugfs(ref->priv, sta->rate_ctrl_priv,
+					  sta->debugfs_dir);
+#endif
+}
+
+extern const struct debugfs_short_fops rcname_ops;
+
+static inline void rate_control_add_debugfs(struct ieee80211_local *local)
+{
+#ifdef CONFIG_MAC80211_DEBUGFS
+	struct dentry *debugfsdir;
+
+	if (!local->rate_ctrl)
+		return;
+
+	if (!local->rate_ctrl->ops->add_debugfs)
+		return;
+
+	debugfsdir = debugfs_create_dir("rc", local->hw.wiphy->debugfsdir);
+	local->debugfs.rcdir = debugfsdir;
+	debugfs_create_file("name", 0400, debugfsdir,
+			    local->rate_ctrl, &rcname_ops);
+
+	local->rate_ctrl->ops->add_debugfs(&local->hw, local->rate_ctrl->priv,
+					   debugfsdir);
+#endif
+}
+
+void ieee80211_check_rate_mask(struct ieee80211_link_data *link);
+
+/* Get a reference to the rate control algorithm. If `name' is NULL, get the
+ * first available algorithm. */
+int ieee80211_init_rate_ctrl_alg(struct ieee80211_local *local,
+				 const char *name);
+void rate_control_deinitialize(struct ieee80211_local *local);
+
+
+/* Rate control algorithms */
+#ifdef CONFIG_MAC80211_RC_MINSTREL
+int rc80211_minstrel_init(void);
+void rc80211_minstrel_exit(void);
+#else
+static inline int rc80211_minstrel_init(void)
+{
+	return 0;
+}
+static inline void rc80211_minstrel_exit(void)
+{
+}
+#endif
+
+
+#endif /* IEEE80211_RATE_H */