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, 126 insertions, 0 deletions

diff --git a/block/blk-cgroup-rwstat.c b/block/blk-cgroup-rwstat.c
new file mode 100644
index 000000000..2e9bf081c
--- /dev/null
+++ b/block/blk-cgroup-rwstat.c
@@ -0,0 +1,126 @@
+/* SPDX-License-Identifier: GPL-2.0
+ *
+ * Legacy blkg rwstat helpers enabled by CONFIG_BLK_CGROUP_RWSTAT.
+ * Do not use in new code.
+ */
+#include "blk-cgroup-rwstat.h"
+
+int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp)
+{
+	int i, ret;
+
+	ret = percpu_counter_init_many(rwstat->cpu_cnt, 0, gfp, BLKG_RWSTAT_NR);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < BLKG_RWSTAT_NR; i++)
+		atomic64_set(&rwstat->aux_cnt[i], 0);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(blkg_rwstat_init);
+
+void blkg_rwstat_exit(struct blkg_rwstat *rwstat)
+{
+	percpu_counter_destroy_many(rwstat->cpu_cnt, BLKG_RWSTAT_NR);
+}
+EXPORT_SYMBOL_GPL(blkg_rwstat_exit);
+
+/**
+ * __blkg_prfill_rwstat - prfill helper for a blkg_rwstat
+ * @sf: seq_file to print to
+ * @pd: policy private data of interest
+ * @rwstat: rwstat to print
+ *
+ * Print @rwstat to @sf for the device assocaited with @pd.
+ */
+u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
+			 const struct blkg_rwstat_sample *rwstat)
+{
+	static const char *rwstr[] = {
+		[BLKG_RWSTAT_READ]	= "Read",
+		[BLKG_RWSTAT_WRITE]	= "Write",
+		[BLKG_RWSTAT_SYNC]	= "Sync",
+		[BLKG_RWSTAT_ASYNC]	= "Async",
+		[BLKG_RWSTAT_DISCARD]	= "Discard",
+	};
+	const char *dname = blkg_dev_name(pd->blkg);
+	u64 v;
+	int i;
+
+	if (!dname)
+		return 0;
+
+	for (i = 0; i < BLKG_RWSTAT_NR; i++)
+		seq_printf(sf, "%s %s %llu\n", dname, rwstr[i],
+			   rwstat->cnt[i]);
+
+	v = rwstat->cnt[BLKG_RWSTAT_READ] +
+		rwstat->cnt[BLKG_RWSTAT_WRITE] +
+		rwstat->cnt[BLKG_RWSTAT_DISCARD];
+	seq_printf(sf, "%s Total %llu\n", dname, v);
+	return v;
+}
+EXPORT_SYMBOL_GPL(__blkg_prfill_rwstat);
+
+/**
+ * blkg_prfill_rwstat - prfill callback for blkg_rwstat
+ * @sf: seq_file to print to
+ * @pd: policy private data of interest
+ * @off: offset to the blkg_rwstat in @pd
+ *
+ * prfill callback for printing a blkg_rwstat.
+ */
+u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
+		       int off)
+{
+	struct blkg_rwstat_sample rwstat = { };
+
+	blkg_rwstat_read((void *)pd + off, &rwstat);
+	return __blkg_prfill_rwstat(sf, pd, &rwstat);
+}
+EXPORT_SYMBOL_GPL(blkg_prfill_rwstat);
+
+/**
+ * blkg_rwstat_recursive_sum - collect hierarchical blkg_rwstat
+ * @blkg: blkg of interest
+ * @pol: blkcg_policy which contains the blkg_rwstat
+ * @off: offset to the blkg_rwstat in blkg_policy_data or @blkg
+ * @sum: blkg_rwstat_sample structure containing the results
+ *
+ * Collect the blkg_rwstat specified by @blkg, @pol and @off and all its
+ * online descendants and their aux counts.
+ *
+ * If @pol is NULL, blkg_rwstat is at @off bytes into @blkg; otherwise, it
+ * is at @off bytes into @blkg's blkg_policy_data of the policy.
+ */
+void blkg_rwstat_recursive_sum(struct blkcg_gq *blkg, struct blkcg_policy *pol,
+		int off, struct blkg_rwstat_sample *sum)
+{
+	struct blkcg_gq *pos_blkg;
+	struct cgroup_subsys_state *pos_css;
+	unsigned int i;
+
+	WARN_ON_ONCE(!rcu_read_lock_held());
+
+	memset(sum, 0, sizeof(*sum));
+	blkg_for_each_descendant_pre(pos_blkg, pos_css, blkg) {
+		struct blkg_rwstat *rwstat;
+
+		if (!data_race(pos_blkg->online))
+			continue;
+
+		if (pol) {
+			struct blkg_policy_data *pd = blkg_to_pd(pos_blkg, pol);
+
+			if (!pd)
+				continue;
+			rwstat = (void *)pd + off;
+		} else {
+			rwstat = (void *)pos_blkg + off;
+		}
+
+		for (i = 0; i < BLKG_RWSTAT_NR; i++)
+			sum->cnt[i] += blkg_rwstat_read_counter(rwstat, i);
+	}
+}
+EXPORT_SYMBOL_GPL(blkg_rwstat_recursive_sum);