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, 228 insertions, 0 deletions

diff --git a/block/blk-stat.c b/block/blk-stat.c
new file mode 100644
index 000000000..d57c2fc6b
--- /dev/null
+++ b/block/blk-stat.c
@@ -0,0 +1,228 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Block stat tracking code
+ *
+ * Copyright (C) 2016 Jens Axboe
+ */
+#include <linux/kernel.h>
+#include <linux/rculist.h>
+
+#include "blk-stat.h"
+#include "blk-mq.h"
+#include "blk.h"
+
+struct blk_queue_stats {
+	struct list_head callbacks;
+	spinlock_t lock;
+	int accounting;
+};
+
+void blk_rq_stat_init(struct blk_rq_stat *stat)
+{
+	stat->min = -1ULL;
+	stat->max = stat->nr_samples = stat->mean = 0;
+	stat->batch = 0;
+}
+
+/* src is a per-cpu stat, mean isn't initialized */
+void blk_rq_stat_sum(struct blk_rq_stat *dst, struct blk_rq_stat *src)
+{
+	if (dst->nr_samples + src->nr_samples <= dst->nr_samples)
+		return;
+
+	dst->min = min(dst->min, src->min);
+	dst->max = max(dst->max, src->max);
+
+	dst->mean = div_u64(src->batch + dst->mean * dst->nr_samples,
+				dst->nr_samples + src->nr_samples);
+
+	dst->nr_samples += src->nr_samples;
+}
+
+void blk_rq_stat_add(struct blk_rq_stat *stat, u64 value)
+{
+	stat->min = min(stat->min, value);
+	stat->max = max(stat->max, value);
+	stat->batch += value;
+	stat->nr_samples++;
+}
+
+void blk_stat_add(struct request *rq, u64 now)
+{
+	struct request_queue *q = rq->q;
+	struct blk_stat_callback *cb;
+	struct blk_rq_stat *stat;
+	int bucket, cpu;
+	u64 value;
+
+	value = (now >= rq->io_start_time_ns) ? now - rq->io_start_time_ns : 0;
+
+	rcu_read_lock();
+	cpu = get_cpu();
+	list_for_each_entry_rcu(cb, &q->stats->callbacks, list) {
+		if (!blk_stat_is_active(cb))
+			continue;
+
+		bucket = cb->bucket_fn(rq);
+		if (bucket < 0)
+			continue;
+
+		stat = &per_cpu_ptr(cb->cpu_stat, cpu)[bucket];
+		blk_rq_stat_add(stat, value);
+	}
+	put_cpu();
+	rcu_read_unlock();
+}
+
+static void blk_stat_timer_fn(struct timer_list *t)
+{
+	struct blk_stat_callback *cb = timer_container_of(cb, t, timer);
+	unsigned int bucket;
+	int cpu;
+
+	for (bucket = 0; bucket < cb->buckets; bucket++)
+		blk_rq_stat_init(&cb->stat[bucket]);
+
+	for_each_possible_cpu(cpu) {
+		struct blk_rq_stat *cpu_stat;
+
+		cpu_stat = per_cpu_ptr(cb->cpu_stat, cpu);
+		for (bucket = 0; bucket < cb->buckets; bucket++) {
+			blk_rq_stat_sum(&cb->stat[bucket], &cpu_stat[bucket]);
+			blk_rq_stat_init(&cpu_stat[bucket]);
+		}
+	}
+
+	cb->timer_fn(cb);
+}
+
+struct blk_stat_callback *
+blk_stat_alloc_callback(void (*timer_fn)(struct blk_stat_callback *),
+			int (*bucket_fn)(const struct request *),
+			unsigned int buckets, void *data)
+{
+	struct blk_stat_callback *cb;
+
+	cb = kmalloc_obj(*cb);
+	if (!cb)
+		return NULL;
+
+	cb->stat = kmalloc_objs(struct blk_rq_stat, buckets);
+	if (!cb->stat) {
+		kfree(cb);
+		return NULL;
+	}
+	cb->cpu_stat = __alloc_percpu(buckets * sizeof(struct blk_rq_stat),
+				      __alignof__(struct blk_rq_stat));
+	if (!cb->cpu_stat) {
+		kfree(cb->stat);
+		kfree(cb);
+		return NULL;
+	}
+
+	cb->timer_fn = timer_fn;
+	cb->bucket_fn = bucket_fn;
+	cb->data = data;
+	cb->buckets = buckets;
+	timer_setup(&cb->timer, blk_stat_timer_fn, 0);
+
+	return cb;
+}
+
+void blk_stat_add_callback(struct request_queue *q,
+			   struct blk_stat_callback *cb)
+{
+	unsigned int bucket;
+	unsigned long flags;
+	int cpu;
+
+	for_each_possible_cpu(cpu) {
+		struct blk_rq_stat *cpu_stat;
+
+		cpu_stat = per_cpu_ptr(cb->cpu_stat, cpu);
+		for (bucket = 0; bucket < cb->buckets; bucket++)
+			blk_rq_stat_init(&cpu_stat[bucket]);
+	}
+
+	spin_lock_irqsave(&q->stats->lock, flags);
+	list_add_tail_rcu(&cb->list, &q->stats->callbacks);
+	blk_queue_flag_set(QUEUE_FLAG_STATS, q);
+	spin_unlock_irqrestore(&q->stats->lock, flags);
+}
+
+void blk_stat_remove_callback(struct request_queue *q,
+			      struct blk_stat_callback *cb)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&q->stats->lock, flags);
+	list_del_rcu(&cb->list);
+	if (list_empty(&q->stats->callbacks) && !q->stats->accounting)
+		blk_queue_flag_clear(QUEUE_FLAG_STATS, q);
+	spin_unlock_irqrestore(&q->stats->lock, flags);
+
+	timer_delete_sync(&cb->timer);
+}
+
+static void blk_stat_free_callback_rcu(struct rcu_head *head)
+{
+	struct blk_stat_callback *cb;
+
+	cb = container_of(head, struct blk_stat_callback, rcu);
+	free_percpu(cb->cpu_stat);
+	kfree(cb->stat);
+	kfree(cb);
+}
+
+void blk_stat_free_callback(struct blk_stat_callback *cb)
+{
+	if (cb)
+		call_rcu(&cb->rcu, blk_stat_free_callback_rcu);
+}
+
+void blk_stat_disable_accounting(struct request_queue *q)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&q->stats->lock, flags);
+	if (!--q->stats->accounting && list_empty(&q->stats->callbacks))
+		blk_queue_flag_clear(QUEUE_FLAG_STATS, q);
+	spin_unlock_irqrestore(&q->stats->lock, flags);
+}
+EXPORT_SYMBOL_GPL(blk_stat_disable_accounting);
+
+void blk_stat_enable_accounting(struct request_queue *q)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&q->stats->lock, flags);
+	if (!q->stats->accounting++ && list_empty(&q->stats->callbacks))
+		blk_queue_flag_set(QUEUE_FLAG_STATS, q);
+	spin_unlock_irqrestore(&q->stats->lock, flags);
+}
+EXPORT_SYMBOL_GPL(blk_stat_enable_accounting);
+
+struct blk_queue_stats *blk_alloc_queue_stats(void)
+{
+	struct blk_queue_stats *stats;
+
+	stats = kmalloc_obj(*stats);
+	if (!stats)
+		return NULL;
+
+	INIT_LIST_HEAD(&stats->callbacks);
+	spin_lock_init(&stats->lock);
+	stats->accounting = 0;
+
+	return stats;
+}
+
+void blk_free_queue_stats(struct blk_queue_stats *stats)
+{
+	if (!stats)
+		return;
+
+	WARN_ON(!list_empty(&stats->callbacks));
+
+	kfree(stats);
+}