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, 111 insertions, 0 deletions

diff --git a/block/partitions/of.c b/block/partitions/of.c
new file mode 100644
index 000000000..53664ea06
--- /dev/null
+++ b/block/partitions/of.c
@@ -0,0 +1,111 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/blkdev.h>
+#include <linux/major.h>
+#include <linux/of.h>
+#include <linux/string.h>
+#include "check.h"
+
+static int validate_of_partition(struct device_node *np, int slot)
+{
+	u64 offset, size;
+	int len;
+
+	const __be32 *reg = of_get_property(np, "reg", &len);
+	int a_cells = of_n_addr_cells(np);
+	int s_cells = of_n_size_cells(np);
+
+	/* Make sure reg len match the expected addr and size cells */
+	if (len / sizeof(*reg) != a_cells + s_cells)
+		return -EINVAL;
+
+	/* Validate offset conversion from bytes to sectors */
+	offset = of_read_number(reg, a_cells);
+	if (offset % SECTOR_SIZE)
+		return -EINVAL;
+
+	/* Validate size conversion from bytes to sectors */
+	size = of_read_number(reg + a_cells, s_cells);
+	if (!size || size % SECTOR_SIZE)
+		return -EINVAL;
+
+	return 0;
+}
+
+static void add_of_partition(struct parsed_partitions *state, int slot,
+			     struct device_node *np)
+{
+	struct partition_meta_info *info;
+	const char *partname;
+	int len;
+
+	const __be32 *reg = of_get_property(np, "reg", &len);
+	int a_cells = of_n_addr_cells(np);
+	int s_cells = of_n_size_cells(np);
+
+	/* Convert bytes to sector size */
+	u64 offset = of_read_number(reg, a_cells) / SECTOR_SIZE;
+	u64 size = of_read_number(reg + a_cells, s_cells) / SECTOR_SIZE;
+
+	put_partition(state, slot, offset, size);
+
+	if (of_property_read_bool(np, "read-only"))
+		state->parts[slot].flags |= ADDPART_FLAG_READONLY;
+
+	/*
+	 * Follow MTD label logic, search for label property,
+	 * fallback to node name if not found.
+	 */
+	info = &state->parts[slot].info;
+	partname = of_get_property(np, "label", &len);
+	if (!partname)
+		partname = of_get_property(np, "name", &len);
+	strscpy(info->volname, partname, sizeof(info->volname));
+
+	seq_buf_printf(&state->pp_buf, "(%s)", info->volname);
+}
+
+int of_partition(struct parsed_partitions *state)
+{
+	struct device *ddev = disk_to_dev(state->disk);
+	struct device_node *np;
+	int slot;
+
+	struct device_node *partitions_np = of_node_get(ddev->of_node);
+
+	if (!partitions_np ||
+	    !of_device_is_compatible(partitions_np, "fixed-partitions")) {
+		of_node_put(partitions_np);
+		return 0;
+	}
+
+	slot = 1;
+	/* Validate parition offset and size */
+	for_each_child_of_node(partitions_np, np) {
+		if (validate_of_partition(np, slot)) {
+			of_node_put(np);
+			of_node_put(partitions_np);
+
+			return -1;
+		}
+
+		slot++;
+	}
+
+	slot = 1;
+	for_each_child_of_node(partitions_np, np) {
+		if (slot >= state->limit) {
+			of_node_put(np);
+			break;
+		}
+
+		add_of_partition(state, slot, np);
+
+		slot++;
+	}
+
+	seq_buf_puts(&state->pp_buf, "\n");
+
+	of_node_put(partitions_np);
+	return 1;
+}