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, 156 insertions, 0 deletions

diff --git a/block/holder.c b/block/holder.c
new file mode 100644
index 000000000..267beb3dc
--- /dev/null
+++ b/block/holder.c
@@ -0,0 +1,156 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/blkdev.h>
+#include <linux/slab.h>
+
+struct bd_holder_disk {
+	struct list_head	list;
+	struct kobject		*holder_dir;
+	int			refcnt;
+};
+
+static DEFINE_MUTEX(blk_holder_mutex);
+
+static struct bd_holder_disk *bd_find_holder_disk(struct block_device *bdev,
+						  struct gendisk *disk)
+{
+	struct bd_holder_disk *holder;
+
+	list_for_each_entry(holder, &disk->slave_bdevs, list)
+		if (holder->holder_dir == bdev->bd_holder_dir)
+			return holder;
+	return NULL;
+}
+
+static int add_symlink(struct kobject *from, struct kobject *to)
+{
+	return sysfs_create_link(from, to, kobject_name(to));
+}
+
+static void del_symlink(struct kobject *from, struct kobject *to)
+{
+	sysfs_remove_link(from, kobject_name(to));
+}
+
+/**
+ * bd_link_disk_holder - create symlinks between holding disk and slave bdev
+ * @bdev: the claimed slave bdev
+ * @disk: the holding disk
+ *
+ * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
+ *
+ * This functions creates the following sysfs symlinks.
+ *
+ * - from "slaves" directory of the holder @disk to the claimed @bdev
+ * - from "holders" directory of the @bdev to the holder @disk
+ *
+ * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is
+ * passed to bd_link_disk_holder(), then:
+ *
+ *   /sys/block/dm-0/slaves/sda --> /sys/block/sda
+ *   /sys/block/sda/holders/dm-0 --> /sys/block/dm-0
+ *
+ * The caller must have claimed @bdev before calling this function and
+ * ensure that both @bdev and @disk are valid during the creation and
+ * lifetime of these symlinks.
+ *
+ * CONTEXT:
+ * Might sleep.
+ *
+ * RETURNS:
+ * 0 on success, -errno on failure.
+ */
+int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk)
+{
+	struct bd_holder_disk *holder;
+	int ret = 0;
+
+	if (WARN_ON_ONCE(!disk->slave_dir))
+		return -EINVAL;
+
+	if (bdev->bd_disk == disk)
+		return -EINVAL;
+
+	/*
+	 * del_gendisk drops the initial reference to bd_holder_dir, so we
+	 * need to keep our own here to allow for cleanup past that point.
+	 */
+	mutex_lock(&bdev->bd_disk->open_mutex);
+	if (!disk_live(bdev->bd_disk)) {
+		mutex_unlock(&bdev->bd_disk->open_mutex);
+		return -ENODEV;
+	}
+	kobject_get(bdev->bd_holder_dir);
+	mutex_unlock(&bdev->bd_disk->open_mutex);
+
+	mutex_lock(&blk_holder_mutex);
+	WARN_ON_ONCE(!bdev->bd_holder);
+
+	holder = bd_find_holder_disk(bdev, disk);
+	if (holder) {
+		kobject_put(bdev->bd_holder_dir);
+		holder->refcnt++;
+		goto out_unlock;
+	}
+
+	holder = kzalloc_obj(*holder);
+	if (!holder) {
+		ret = -ENOMEM;
+		goto out_unlock;
+	}
+
+	INIT_LIST_HEAD(&holder->list);
+	holder->refcnt = 1;
+	holder->holder_dir = bdev->bd_holder_dir;
+
+	ret = add_symlink(disk->slave_dir, bdev_kobj(bdev));
+	if (ret)
+		goto out_free_holder;
+	ret = add_symlink(bdev->bd_holder_dir, &disk_to_dev(disk)->kobj);
+	if (ret)
+		goto out_del_symlink;
+	list_add(&holder->list, &disk->slave_bdevs);
+
+	mutex_unlock(&blk_holder_mutex);
+	return 0;
+
+out_del_symlink:
+	del_symlink(disk->slave_dir, bdev_kobj(bdev));
+out_free_holder:
+	kfree(holder);
+out_unlock:
+	mutex_unlock(&blk_holder_mutex);
+	if (ret)
+		kobject_put(bdev->bd_holder_dir);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(bd_link_disk_holder);
+
+/**
+ * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder()
+ * @bdev: the calimed slave bdev
+ * @disk: the holding disk
+ *
+ * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT.
+ *
+ * CONTEXT:
+ * Might sleep.
+ */
+void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk)
+{
+	struct bd_holder_disk *holder;
+
+	if (WARN_ON_ONCE(!disk->slave_dir))
+		return;
+
+	mutex_lock(&blk_holder_mutex);
+	holder = bd_find_holder_disk(bdev, disk);
+	if (!WARN_ON_ONCE(holder == NULL) && !--holder->refcnt) {
+		del_symlink(disk->slave_dir, bdev_kobj(bdev));
+		del_symlink(holder->holder_dir, &disk_to_dev(disk)->kobj);
+		kobject_put(holder->holder_dir);
+		list_del_init(&holder->list);
+		kfree(holder);
+	}
+	mutex_unlock(&blk_holder_mutex);
+}
+EXPORT_SYMBOL_GPL(bd_unlink_disk_holder);