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, 234 insertions, 0 deletions

diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c
new file mode 100644
index 000000000..38c17846d
--- /dev/null
+++ b/drivers/md/dm-linear.c
@@ -0,0 +1,234 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2001-2003 Sistina Software (UK) Limited.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm.h"
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/blkdev.h>
+#include <linux/bio.h>
+#include <linux/dax.h>
+#include <linux/slab.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "linear"
+
+/*
+ * Linear: maps a linear range of a device.
+ */
+struct linear_c {
+	struct dm_dev *dev;
+	sector_t start;
+};
+
+/*
+ * Construct a linear mapping: <dev_path> <offset>
+ */
+static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv)
+{
+	struct linear_c *lc;
+	unsigned long long tmp;
+	char dummy;
+	int ret;
+
+	if (argc != 2) {
+		ti->error = "Invalid argument count";
+		return -EINVAL;
+	}
+
+	lc = kmalloc_obj(*lc);
+	if (lc == NULL) {
+		ti->error = "Cannot allocate linear context";
+		return -ENOMEM;
+	}
+
+	ret = -EINVAL;
+	if (sscanf(argv[1], "%llu%c", &tmp, &dummy) != 1 || tmp != (sector_t)tmp) {
+		ti->error = "Invalid device sector";
+		goto bad;
+	}
+	lc->start = tmp;
+
+	ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev);
+	if (ret) {
+		ti->error = "Device lookup failed";
+		goto bad;
+	}
+
+	ti->num_flush_bios = 1;
+	ti->num_discard_bios = 1;
+	ti->num_secure_erase_bios = 1;
+	ti->num_write_zeroes_bios = 1;
+	ti->flush_bypasses_map = true;
+	ti->private = lc;
+	return 0;
+
+bad:
+	kfree(lc);
+	return ret;
+}
+
+static void linear_dtr(struct dm_target *ti)
+{
+	struct linear_c *lc = ti->private;
+
+	dm_put_device(ti, lc->dev);
+	kfree(lc);
+}
+
+static sector_t linear_map_sector(struct dm_target *ti, sector_t bi_sector)
+{
+	struct linear_c *lc = ti->private;
+
+	return lc->start + dm_target_offset(ti, bi_sector);
+}
+
+int linear_map(struct dm_target *ti, struct bio *bio)
+{
+	struct linear_c *lc = ti->private;
+
+	bio_set_dev(bio, lc->dev->bdev);
+	bio->bi_iter.bi_sector = linear_map_sector(ti, bio->bi_iter.bi_sector);
+
+	return DM_MAPIO_REMAPPED;
+}
+
+static void linear_status(struct dm_target *ti, status_type_t type,
+			  unsigned int status_flags, char *result, unsigned int maxlen)
+{
+	struct linear_c *lc = ti->private;
+	size_t sz = 0;
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		result[0] = '\0';
+		break;
+
+	case STATUSTYPE_TABLE:
+		DMEMIT("%s %llu", lc->dev->name, (unsigned long long)lc->start);
+		break;
+
+	case STATUSTYPE_IMA:
+		DMEMIT_TARGET_NAME_VERSION(ti->type);
+		DMEMIT(",device_name=%s,start=%llu;", lc->dev->name,
+		       (unsigned long long)lc->start);
+		break;
+	}
+}
+
+static int linear_prepare_ioctl(struct dm_target *ti, struct block_device **bdev,
+				unsigned int cmd, unsigned long arg,
+				bool *forward)
+{
+	struct linear_c *lc = ti->private;
+	struct dm_dev *dev = lc->dev;
+
+	*bdev = dev->bdev;
+
+	/*
+	 * Only pass ioctls through if the device sizes match exactly.
+	 */
+	if (lc->start || ti->len != bdev_nr_sectors(dev->bdev))
+		return 1;
+	return 0;
+}
+
+#ifdef CONFIG_BLK_DEV_ZONED
+static int linear_report_zones(struct dm_target *ti,
+		struct dm_report_zones_args *args, unsigned int nr_zones)
+{
+	struct linear_c *lc = ti->private;
+
+	return dm_report_zones(lc->dev->bdev, lc->start,
+			       linear_map_sector(ti, args->next_sector),
+			       args, nr_zones);
+}
+#else
+#define linear_report_zones NULL
+#endif
+
+static int linear_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct linear_c *lc = ti->private;
+
+	return fn(ti, lc->dev, lc->start, ti->len, data);
+}
+
+#if IS_ENABLED(CONFIG_FS_DAX)
+static struct dax_device *linear_dax_pgoff(struct dm_target *ti, pgoff_t *pgoff)
+{
+	struct linear_c *lc = ti->private;
+	sector_t sector = linear_map_sector(ti, *pgoff << PAGE_SECTORS_SHIFT);
+
+	*pgoff = (get_start_sect(lc->dev->bdev) + sector) >> PAGE_SECTORS_SHIFT;
+	return lc->dev->dax_dev;
+}
+
+static long linear_dax_direct_access(struct dm_target *ti, pgoff_t pgoff,
+		long nr_pages, enum dax_access_mode mode, void **kaddr,
+		unsigned long *pfn)
+{
+	struct dax_device *dax_dev = linear_dax_pgoff(ti, &pgoff);
+
+	return dax_direct_access(dax_dev, pgoff, nr_pages, mode, kaddr, pfn);
+}
+
+static int linear_dax_zero_page_range(struct dm_target *ti, pgoff_t pgoff,
+				      size_t nr_pages)
+{
+	struct dax_device *dax_dev = linear_dax_pgoff(ti, &pgoff);
+
+	return dax_zero_page_range(dax_dev, pgoff, nr_pages);
+}
+
+static size_t linear_dax_recovery_write(struct dm_target *ti, pgoff_t pgoff,
+		void *addr, size_t bytes, struct iov_iter *i)
+{
+	struct dax_device *dax_dev = linear_dax_pgoff(ti, &pgoff);
+
+	return dax_recovery_write(dax_dev, pgoff, addr, bytes, i);
+}
+
+#else
+#define linear_dax_direct_access NULL
+#define linear_dax_zero_page_range NULL
+#define linear_dax_recovery_write NULL
+#endif
+
+static struct target_type linear_target = {
+	.name   = "linear",
+	.version = {1, 5, 0},
+	.features = DM_TARGET_PASSES_INTEGRITY | DM_TARGET_NOWAIT |
+		    DM_TARGET_ZONED_HM | DM_TARGET_PASSES_CRYPTO |
+		    DM_TARGET_ATOMIC_WRITES,
+	.report_zones = linear_report_zones,
+	.module = THIS_MODULE,
+	.ctr    = linear_ctr,
+	.dtr    = linear_dtr,
+	.map    = linear_map,
+	.status = linear_status,
+	.prepare_ioctl = linear_prepare_ioctl,
+	.iterate_devices = linear_iterate_devices,
+	.direct_access = linear_dax_direct_access,
+	.dax_zero_page_range = linear_dax_zero_page_range,
+	.dax_recovery_write = linear_dax_recovery_write,
+};
+
+int __init dm_linear_init(void)
+{
+	int r = dm_register_target(&linear_target);
+
+	if (r < 0)
+		DMERR("register failed %d", r);
+
+	return r;
+}
+
+void dm_linear_exit(void)
+{
+	dm_unregister_target(&linear_target);
+}