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, 123 insertions, 0 deletions

diff --git a/net/hsr/hsr_debugfs.c b/net/hsr/hsr_debugfs.c
new file mode 100644
index 000000000..5b2cfac3b
--- /dev/null
+++ b/net/hsr/hsr_debugfs.c
@@ -0,0 +1,123 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * debugfs code for HSR & PRP
+ * Copyright (C) 2019 Texas Instruments Incorporated
+ *
+ * Author(s):
+ *	Murali Karicheri <m-karicheri2@ti.com>
+ */
+#include <linux/module.h>
+#include <linux/errno.h>
+#include <linux/debugfs.h>
+#include "hsr_main.h"
+#include "hsr_framereg.h"
+
+static struct dentry *hsr_debugfs_root_dir;
+
+/* hsr_node_table_show - Formats and prints node_table entries */
+static int
+hsr_node_table_show(struct seq_file *sfp, void *data)
+{
+	struct hsr_priv *priv = (struct hsr_priv *)sfp->private;
+	struct hsr_node *node;
+
+	seq_printf(sfp, "Node Table entries for (%s) device\n",
+		   (priv->prot_version == PRP_V1 ? "PRP" : "HSR"));
+	seq_puts(sfp, "MAC-Address-A,    MAC-Address-B,    time_in[A], ");
+	seq_puts(sfp, "time_in[B], Address-B port, ");
+	if (priv->prot_version == PRP_V1)
+		seq_puts(sfp, "SAN-A, SAN-B, DAN-P\n");
+	else
+		seq_puts(sfp, "DAN-H\n");
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(node, &priv->node_db, mac_list) {
+		/* skip self node */
+		if (hsr_addr_is_self(priv, node->macaddress_A))
+			continue;
+		seq_printf(sfp, "%pM ", &node->macaddress_A[0]);
+		seq_printf(sfp, "%pM ", &node->macaddress_B[0]);
+		seq_printf(sfp, "%10lx, ", node->time_in[HSR_PT_SLAVE_A]);
+		seq_printf(sfp, "%10lx, ", node->time_in[HSR_PT_SLAVE_B]);
+		seq_printf(sfp, "%14x, ", node->addr_B_port);
+
+		if (priv->prot_version == PRP_V1)
+			seq_printf(sfp, "%5x, %5x, %5x\n",
+				   node->san_a, node->san_b,
+				   (node->san_a == 0 && node->san_b == 0));
+		else
+			seq_printf(sfp, "%5x\n", 1);
+	}
+	rcu_read_unlock();
+	return 0;
+}
+
+DEFINE_SHOW_ATTRIBUTE(hsr_node_table);
+
+void hsr_debugfs_rename(struct net_device *dev)
+{
+	struct hsr_priv *priv = netdev_priv(dev);
+	int err;
+
+	err = debugfs_change_name(priv->node_tbl_root, "%s", dev->name);
+	if (err)
+		netdev_warn(dev, "failed to rename\n");
+}
+
+/* hsr_debugfs_init - create hsr node_table file for dumping
+ * the node table
+ *
+ * Description:
+ * When debugfs is configured this routine sets up the node_table file per
+ * hsr device for dumping the node_table entries
+ */
+void hsr_debugfs_init(struct hsr_priv *priv, struct net_device *hsr_dev)
+{
+	struct dentry *de = NULL;
+
+	de = debugfs_create_dir(hsr_dev->name, hsr_debugfs_root_dir);
+	if (IS_ERR(de)) {
+		pr_err("Cannot create hsr debugfs directory\n");
+		return;
+	}
+
+	priv->node_tbl_root = de;
+
+	de = debugfs_create_file("node_table", S_IFREG | 0444,
+				 priv->node_tbl_root, priv,
+				 &hsr_node_table_fops);
+	if (IS_ERR(de)) {
+		pr_err("Cannot create hsr node_table file\n");
+		debugfs_remove(priv->node_tbl_root);
+		priv->node_tbl_root = NULL;
+		return;
+	}
+}
+
+/* hsr_debugfs_term - Tear down debugfs intrastructure
+ *
+ * Description:
+ * When Debugfs is configured this routine removes debugfs file system
+ * elements that are specific to hsr
+ */
+void
+hsr_debugfs_term(struct hsr_priv *priv)
+{
+	debugfs_remove_recursive(priv->node_tbl_root);
+	priv->node_tbl_root = NULL;
+}
+
+void hsr_debugfs_create_root(void)
+{
+	hsr_debugfs_root_dir = debugfs_create_dir("hsr", NULL);
+	if (IS_ERR(hsr_debugfs_root_dir)) {
+		pr_err("Cannot create hsr debugfs root directory\n");
+		hsr_debugfs_root_dir = NULL;
+	}
+}
+
+void hsr_debugfs_remove_root(void)
+{
+	/* debugfs_remove() internally checks NULL and ERROR */
+	debugfs_remove(hsr_debugfs_root_dir);
+}