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, 223 insertions, 0 deletions

diff --git a/drivers/net/wireguard/ratelimiter.c b/drivers/net/wireguard/ratelimiter.c
new file mode 100644
index 000000000..2b3867f69
--- /dev/null
+++ b/drivers/net/wireguard/ratelimiter.c
@@ -0,0 +1,223 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+ */
+
+#include "ratelimiter.h"
+#include <linux/siphash.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <net/ip.h>
+
+static struct kmem_cache *entry_cache;
+static hsiphash_key_t key;
+static spinlock_t table_lock = __SPIN_LOCK_UNLOCKED("ratelimiter_table_lock");
+static DEFINE_MUTEX(init_lock);
+static u64 init_refcnt; /* Protected by init_lock, hence not atomic. */
+static atomic_t total_entries = ATOMIC_INIT(0);
+static unsigned int max_entries, table_size;
+static void wg_ratelimiter_gc_entries(struct work_struct *);
+static DECLARE_DEFERRABLE_WORK(gc_work, wg_ratelimiter_gc_entries);
+static struct hlist_head *table_v4;
+#if IS_ENABLED(CONFIG_IPV6)
+static struct hlist_head *table_v6;
+#endif
+
+struct ratelimiter_entry {
+	u64 last_time_ns, tokens, ip;
+	void *net;
+	spinlock_t lock;
+	struct hlist_node hash;
+	struct rcu_head rcu;
+};
+
+enum {
+	PACKETS_PER_SECOND = 20,
+	PACKETS_BURSTABLE = 5,
+	PACKET_COST = NSEC_PER_SEC / PACKETS_PER_SECOND,
+	TOKEN_MAX = PACKET_COST * PACKETS_BURSTABLE
+};
+
+static void entry_free(struct rcu_head *rcu)
+{
+	kmem_cache_free(entry_cache,
+			container_of(rcu, struct ratelimiter_entry, rcu));
+	atomic_dec(&total_entries);
+}
+
+static void entry_uninit(struct ratelimiter_entry *entry)
+{
+	hlist_del_rcu(&entry->hash);
+	call_rcu(&entry->rcu, entry_free);
+}
+
+/* Calling this function with a NULL work uninits all entries. */
+static void wg_ratelimiter_gc_entries(struct work_struct *work)
+{
+	const u64 now = ktime_get_coarse_boottime_ns();
+	struct ratelimiter_entry *entry;
+	struct hlist_node *temp;
+	unsigned int i;
+
+	for (i = 0; i < table_size; ++i) {
+		spin_lock(&table_lock);
+		hlist_for_each_entry_safe(entry, temp, &table_v4[i], hash) {
+			if (unlikely(!work) ||
+			    now - entry->last_time_ns > NSEC_PER_SEC)
+				entry_uninit(entry);
+		}
+#if IS_ENABLED(CONFIG_IPV6)
+		hlist_for_each_entry_safe(entry, temp, &table_v6[i], hash) {
+			if (unlikely(!work) ||
+			    now - entry->last_time_ns > NSEC_PER_SEC)
+				entry_uninit(entry);
+		}
+#endif
+		spin_unlock(&table_lock);
+		if (likely(work))
+			cond_resched();
+	}
+	if (likely(work))
+		queue_delayed_work(system_power_efficient_wq, &gc_work, HZ);
+}
+
+bool wg_ratelimiter_allow(struct sk_buff *skb, struct net *net)
+{
+	/* We only take the bottom half of the net pointer, so that we can hash
+	 * 3 words in the end. This way, siphash's len param fits into the final
+	 * u32, and we don't incur an extra round.
+	 */
+	const u32 net_word = (unsigned long)net;
+	struct ratelimiter_entry *entry;
+	struct hlist_head *bucket;
+	u64 ip;
+
+	if (skb->protocol == htons(ETH_P_IP)) {
+		ip = (u64 __force)ip_hdr(skb)->saddr;
+		bucket = &table_v4[hsiphash_2u32(net_word, ip, &key) &
+				   (table_size - 1)];
+	}
+#if IS_ENABLED(CONFIG_IPV6)
+	else if (skb->protocol == htons(ETH_P_IPV6)) {
+		/* Only use 64 bits, so as to ratelimit the whole /64. */
+		memcpy(&ip, &ipv6_hdr(skb)->saddr, sizeof(ip));
+		bucket = &table_v6[hsiphash_3u32(net_word, ip >> 32, ip, &key) &
+				   (table_size - 1)];
+	}
+#endif
+	else
+		return false;
+	rcu_read_lock();
+	hlist_for_each_entry_rcu(entry, bucket, hash) {
+		if (entry->net == net && entry->ip == ip) {
+			u64 now, tokens;
+			bool ret;
+			/* Quasi-inspired by nft_limit.c, but this is actually a
+			 * slightly different algorithm. Namely, we incorporate
+			 * the burst as part of the maximum tokens, rather than
+			 * as part of the rate.
+			 */
+			spin_lock(&entry->lock);
+			now = ktime_get_coarse_boottime_ns();
+			tokens = min_t(u64, TOKEN_MAX,
+				       entry->tokens + now -
+					       entry->last_time_ns);
+			entry->last_time_ns = now;
+			ret = tokens >= PACKET_COST;
+			entry->tokens = ret ? tokens - PACKET_COST : tokens;
+			spin_unlock(&entry->lock);
+			rcu_read_unlock();
+			return ret;
+		}
+	}
+	rcu_read_unlock();
+
+	if (atomic_inc_return(&total_entries) > max_entries)
+		goto err_oom;
+
+	entry = kmem_cache_alloc(entry_cache, GFP_KERNEL);
+	if (unlikely(!entry))
+		goto err_oom;
+
+	entry->net = net;
+	entry->ip = ip;
+	INIT_HLIST_NODE(&entry->hash);
+	spin_lock_init(&entry->lock);
+	entry->last_time_ns = ktime_get_coarse_boottime_ns();
+	entry->tokens = TOKEN_MAX - PACKET_COST;
+	spin_lock(&table_lock);
+	hlist_add_head_rcu(&entry->hash, bucket);
+	spin_unlock(&table_lock);
+	return true;
+
+err_oom:
+	atomic_dec(&total_entries);
+	return false;
+}
+
+int wg_ratelimiter_init(void)
+{
+	mutex_lock(&init_lock);
+	if (++init_refcnt != 1)
+		goto out;
+
+	entry_cache = KMEM_CACHE(ratelimiter_entry, 0);
+	if (!entry_cache)
+		goto err;
+
+	/* xt_hashlimit.c uses a slightly different algorithm for ratelimiting,
+	 * but what it shares in common is that it uses a massive hashtable. So,
+	 * we borrow their wisdom about good table sizes on different systems
+	 * dependent on RAM. This calculation here comes from there.
+	 */
+	table_size = (totalram_pages() > (1U << 30) / PAGE_SIZE) ? 8192 :
+		max_t(unsigned long, 16, roundup_pow_of_two(
+			(totalram_pages() << PAGE_SHIFT) /
+			(1U << 14) / sizeof(struct hlist_head)));
+	max_entries = table_size * 8;
+
+	table_v4 = kvzalloc_objs(*table_v4, table_size);
+	if (unlikely(!table_v4))
+		goto err_kmemcache;
+
+#if IS_ENABLED(CONFIG_IPV6)
+	table_v6 = kvzalloc_objs(*table_v6, table_size);
+	if (unlikely(!table_v6)) {
+		kvfree(table_v4);
+		goto err_kmemcache;
+	}
+#endif
+
+	queue_delayed_work(system_power_efficient_wq, &gc_work, HZ);
+	get_random_bytes(&key, sizeof(key));
+out:
+	mutex_unlock(&init_lock);
+	return 0;
+
+err_kmemcache:
+	kmem_cache_destroy(entry_cache);
+err:
+	--init_refcnt;
+	mutex_unlock(&init_lock);
+	return -ENOMEM;
+}
+
+void wg_ratelimiter_uninit(void)
+{
+	mutex_lock(&init_lock);
+	if (!init_refcnt || --init_refcnt)
+		goto out;
+
+	cancel_delayed_work_sync(&gc_work);
+	wg_ratelimiter_gc_entries(NULL);
+	rcu_barrier();
+	kvfree(table_v4);
+#if IS_ENABLED(CONFIG_IPV6)
+	kvfree(table_v6);
+#endif
+	kmem_cache_destroy(entry_cache);
+out:
+	mutex_unlock(&init_lock);
+}
+
+#include "selftest/ratelimiter.c"