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, 278 insertions, 0 deletions

diff --git a/net/6lowpan/debugfs.c b/net/6lowpan/debugfs.c
new file mode 100644
index 000000000..600b9563b
--- /dev/null
+++ b/net/6lowpan/debugfs.c
@@ -0,0 +1,278 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ *
+ * Authors:
+ * (C) 2015 Pengutronix, Alexander Aring <aar@pengutronix.de>
+ * Copyright (c)  2015 Nordic Semiconductor. All Rights Reserved.
+ */
+
+#include <net/6lowpan.h>
+
+#include "6lowpan_i.h"
+
+#define LOWPAN_DEBUGFS_CTX_PFX_NUM_ARGS	8
+
+static struct dentry *lowpan_debugfs;
+
+static int lowpan_ctx_flag_active_set(void *data, u64 val)
+{
+	struct lowpan_iphc_ctx *ctx = data;
+
+	if (val != 0 && val != 1)
+		return -EINVAL;
+
+	if (val)
+		set_bit(LOWPAN_IPHC_CTX_FLAG_ACTIVE, &ctx->flags);
+	else
+		clear_bit(LOWPAN_IPHC_CTX_FLAG_ACTIVE, &ctx->flags);
+
+	return 0;
+}
+
+static int lowpan_ctx_flag_active_get(void *data, u64 *val)
+{
+	*val = lowpan_iphc_ctx_is_active(data);
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(lowpan_ctx_flag_active_fops,
+			 lowpan_ctx_flag_active_get,
+			 lowpan_ctx_flag_active_set, "%llu\n");
+
+static int lowpan_ctx_flag_c_set(void *data, u64 val)
+{
+	struct lowpan_iphc_ctx *ctx = data;
+
+	if (val != 0 && val != 1)
+		return -EINVAL;
+
+	if (val)
+		set_bit(LOWPAN_IPHC_CTX_FLAG_COMPRESSION, &ctx->flags);
+	else
+		clear_bit(LOWPAN_IPHC_CTX_FLAG_COMPRESSION, &ctx->flags);
+
+	return 0;
+}
+
+static int lowpan_ctx_flag_c_get(void *data, u64 *val)
+{
+	*val = lowpan_iphc_ctx_is_compression(data);
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(lowpan_ctx_flag_c_fops, lowpan_ctx_flag_c_get,
+			 lowpan_ctx_flag_c_set, "%llu\n");
+
+static int lowpan_ctx_plen_set(void *data, u64 val)
+{
+	struct lowpan_iphc_ctx *ctx = data;
+	struct lowpan_iphc_ctx_table *t =
+		container_of(ctx, struct lowpan_iphc_ctx_table, table[ctx->id]);
+
+	if (val > 128)
+		return -EINVAL;
+
+	spin_lock_bh(&t->lock);
+	ctx->plen = val;
+	spin_unlock_bh(&t->lock);
+
+	return 0;
+}
+
+static int lowpan_ctx_plen_get(void *data, u64 *val)
+{
+	struct lowpan_iphc_ctx *ctx = data;
+	struct lowpan_iphc_ctx_table *t =
+		container_of(ctx, struct lowpan_iphc_ctx_table, table[ctx->id]);
+
+	spin_lock_bh(&t->lock);
+	*val = ctx->plen;
+	spin_unlock_bh(&t->lock);
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(lowpan_ctx_plen_fops, lowpan_ctx_plen_get,
+			 lowpan_ctx_plen_set, "%llu\n");
+
+static int lowpan_ctx_pfx_show(struct seq_file *file, void *offset)
+{
+	struct lowpan_iphc_ctx *ctx = file->private;
+	struct lowpan_iphc_ctx_table *t =
+		container_of(ctx, struct lowpan_iphc_ctx_table, table[ctx->id]);
+
+	spin_lock_bh(&t->lock);
+	seq_printf(file, "%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n",
+		   be16_to_cpu(ctx->pfx.s6_addr16[0]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[1]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[2]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[3]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[4]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[5]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[6]),
+		   be16_to_cpu(ctx->pfx.s6_addr16[7]));
+	spin_unlock_bh(&t->lock);
+
+	return 0;
+}
+
+static int lowpan_ctx_pfx_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, lowpan_ctx_pfx_show, inode->i_private);
+}
+
+static ssize_t lowpan_ctx_pfx_write(struct file *fp,
+				    const char __user *user_buf, size_t count,
+				    loff_t *ppos)
+{
+	char buf[128] = {};
+	struct seq_file *file = fp->private_data;
+	struct lowpan_iphc_ctx *ctx = file->private;
+	struct lowpan_iphc_ctx_table *t =
+		container_of(ctx, struct lowpan_iphc_ctx_table, table[ctx->id]);
+	int status = count, n, i;
+	unsigned int addr[8];
+
+	if (copy_from_user(&buf, user_buf, min_t(size_t, sizeof(buf) - 1,
+						 count))) {
+		status = -EFAULT;
+		goto out;
+	}
+
+	n = sscanf(buf, "%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x",
+		   &addr[0], &addr[1], &addr[2], &addr[3], &addr[4],
+		   &addr[5], &addr[6], &addr[7]);
+	if (n != LOWPAN_DEBUGFS_CTX_PFX_NUM_ARGS) {
+		status = -EINVAL;
+		goto out;
+	}
+
+	spin_lock_bh(&t->lock);
+	for (i = 0; i < 8; i++)
+		ctx->pfx.s6_addr16[i] = cpu_to_be16(addr[i] & 0xffff);
+	spin_unlock_bh(&t->lock);
+
+out:
+	return status;
+}
+
+static const struct file_operations lowpan_ctx_pfx_fops = {
+	.open		= lowpan_ctx_pfx_open,
+	.read		= seq_read,
+	.write		= lowpan_ctx_pfx_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static void lowpan_dev_debugfs_ctx_init(struct net_device *dev,
+					struct dentry *ctx, u8 id)
+{
+	struct lowpan_dev *ldev = lowpan_dev(dev);
+	struct dentry *root;
+	char buf[32];
+
+	if (WARN_ON_ONCE(id >= LOWPAN_IPHC_CTX_TABLE_SIZE))
+		return;
+
+	sprintf(buf, "%d", id);
+
+	root = debugfs_create_dir(buf, ctx);
+
+	debugfs_create_file("active", 0644, root, &ldev->ctx.table[id],
+			    &lowpan_ctx_flag_active_fops);
+
+	debugfs_create_file("compression", 0644, root, &ldev->ctx.table[id],
+			    &lowpan_ctx_flag_c_fops);
+
+	debugfs_create_file("prefix", 0644, root, &ldev->ctx.table[id],
+			    &lowpan_ctx_pfx_fops);
+
+	debugfs_create_file("prefix_len", 0644, root, &ldev->ctx.table[id],
+			    &lowpan_ctx_plen_fops);
+}
+
+static int lowpan_context_show(struct seq_file *file, void *offset)
+{
+	struct lowpan_iphc_ctx_table *t = file->private;
+	int i;
+
+	seq_printf(file, "%3s|%-43s|%c\n", "cid", "prefix", 'C');
+	seq_puts(file, "-------------------------------------------------\n");
+
+	spin_lock_bh(&t->lock);
+	for (i = 0; i < LOWPAN_IPHC_CTX_TABLE_SIZE; i++) {
+		if (!lowpan_iphc_ctx_is_active(&t->table[i]))
+			continue;
+
+		seq_printf(file, "%3d|%39pI6c/%-3d|%d\n", t->table[i].id,
+			   &t->table[i].pfx, t->table[i].plen,
+			   lowpan_iphc_ctx_is_compression(&t->table[i]));
+	}
+	spin_unlock_bh(&t->lock);
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(lowpan_context);
+
+static int lowpan_short_addr_get(void *data, u64 *val)
+{
+	struct wpan_dev *wdev = data;
+
+	rtnl_lock();
+	*val = le16_to_cpu(wdev->short_addr);
+	rtnl_unlock();
+
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(lowpan_short_addr_fops, lowpan_short_addr_get, NULL,
+			 "0x%04llx\n");
+
+static void lowpan_dev_debugfs_802154_init(const struct net_device *dev,
+					  struct lowpan_dev *ldev)
+{
+	struct dentry *root;
+
+	if (!lowpan_is_ll(dev, LOWPAN_LLTYPE_IEEE802154))
+		return;
+
+	root = debugfs_create_dir("ieee802154", ldev->iface_debugfs);
+
+	debugfs_create_file("short_addr", 0444, root,
+			    lowpan_802154_dev(dev)->wdev->ieee802154_ptr,
+			    &lowpan_short_addr_fops);
+}
+
+void lowpan_dev_debugfs_init(struct net_device *dev)
+{
+	struct lowpan_dev *ldev = lowpan_dev(dev);
+	struct dentry *contexts;
+	int i;
+
+	/* creating the root */
+	ldev->iface_debugfs = debugfs_create_dir(dev->name, lowpan_debugfs);
+
+	contexts = debugfs_create_dir("contexts", ldev->iface_debugfs);
+
+	debugfs_create_file("show", 0644, contexts, &lowpan_dev(dev)->ctx,
+			    &lowpan_context_fops);
+
+	for (i = 0; i < LOWPAN_IPHC_CTX_TABLE_SIZE; i++)
+		lowpan_dev_debugfs_ctx_init(dev, contexts, i);
+
+	lowpan_dev_debugfs_802154_init(dev, ldev);
+}
+
+void lowpan_dev_debugfs_exit(struct net_device *dev)
+{
+	debugfs_remove_recursive(lowpan_dev(dev)->iface_debugfs);
+}
+
+void __init lowpan_debugfs_init(void)
+{
+	lowpan_debugfs = debugfs_create_dir("6lowpan", NULL);
+}
+
+void lowpan_debugfs_exit(void)
+{
+	debugfs_remove_recursive(lowpan_debugfs);
+}