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

3 files changed, 199 insertions, 0 deletions

diff --git a/net/ife/Kconfig b/net/ife/Kconfig
new file mode 100644
index 000000000..de36a5b91
--- /dev/null
+++ b/net/ife/Kconfig
@@ -0,0 +1,16 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# IFE subsystem configuration
+#
+
+menuconfig NET_IFE
+	tristate "Inter-FE based on IETF ForCES InterFE LFB"
+	default n
+	help
+	  Say Y here to add support of IFE encapsulation protocol
+	  For details refer to netdev01 paper:
+	  "Distributing Linux Traffic Control Classifier-Action Subsystem"
+	   Authors: Jamal Hadi Salim and Damascene M. Joachimpillai
+
+	  To compile this support as a module, choose M here: the module will
+	  be called ife.
diff --git a/net/ife/Makefile b/net/ife/Makefile
new file mode 100644
index 000000000..1258fcb07
--- /dev/null
+++ b/net/ife/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Makefile for the IFE encapsulation protocol
+#
+
+obj-$(CONFIG_NET_IFE) += ife.o
diff --git a/net/ife/ife.c b/net/ife/ife.c
new file mode 100644
index 000000000..2ddf725d3
--- /dev/null
+++ b/net/ife/ife.c
@@ -0,0 +1,177 @@
+/*
+ * net/ife/ife.c - Inter-FE protocol based on ForCES WG InterFE LFB
+ * Copyright (c) 2015 Jamal Hadi Salim <jhs@mojatatu.com>
+ * Copyright (c) 2017 Yotam Gigi <yotamg@mellanox.com>
+ *
+ * Refer to: draft-ietf-forces-interfelfb-03 and netdev01 paper:
+ * "Distributing Linux Traffic Control Classifier-Action Subsystem"
+ * Authors: Jamal Hadi Salim and Damascene M. Joachimpillai
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation.
+ */
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/skbuff.h>
+#include <linux/rtnetlink.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <net/net_namespace.h>
+#include <net/netlink.h>
+#include <net/pkt_sched.h>
+#include <linux/etherdevice.h>
+#include <net/ife.h>
+
+struct ifeheadr {
+	__be16 metalen;
+	u8 tlv_data[];
+};
+
+void *ife_encode(struct sk_buff *skb, u16 metalen)
+{
+	/* OUTERHDR:TOTMETALEN:{TLVHDR:Metadatum:TLVHDR..}:ORIGDATA
+	 * where ORIGDATA = original ethernet header ...
+	 */
+	int hdrm = metalen + IFE_METAHDRLEN;
+	int total_push = hdrm + ETH_HLEN;
+	struct ifeheadr *ifehdr;
+	struct ethhdr *iethh;	/* inner ether header */
+	int skboff = 0;
+	int err;
+
+	err = skb_cow_head(skb, total_push);
+	if (unlikely(err))
+		return NULL;
+
+	iethh = (struct ethhdr *) skb->data;
+
+	__skb_push(skb, total_push);
+	memcpy(skb->data, iethh, ETH_HLEN);
+	skb_reset_mac_header(skb);
+	skboff += ETH_HLEN;
+
+	/* total metadata length */
+	ifehdr = (struct ifeheadr *) (skb->data + skboff);
+	metalen += IFE_METAHDRLEN;
+	ifehdr->metalen = htons(metalen);
+
+	return ifehdr->tlv_data;
+}
+EXPORT_SYMBOL_GPL(ife_encode);
+
+void *ife_decode(struct sk_buff *skb, u16 *metalen)
+{
+	struct ifeheadr *ifehdr;
+	int total_pull;
+	u16 ifehdrln;
+
+	if (!pskb_may_pull(skb, ETH_HLEN + IFE_METAHDRLEN))
+		return NULL;
+
+	ifehdr = (struct ifeheadr *)(skb->data + ETH_HLEN);
+	ifehdrln = ntohs(ifehdr->metalen);
+	total_pull = ETH_HLEN + ifehdrln;
+
+	if (unlikely(ifehdrln < 2))
+		return NULL;
+
+	if (unlikely(!pskb_may_pull(skb, total_pull + ETH_HLEN)))
+		return NULL;
+
+	ifehdr = (struct ifeheadr *)(skb->data + ETH_HLEN);
+	skb_set_mac_header(skb, total_pull);
+	__skb_pull(skb, total_pull);
+	*metalen = ifehdrln - IFE_METAHDRLEN;
+
+	return &ifehdr->tlv_data;
+}
+EXPORT_SYMBOL_GPL(ife_decode);
+
+struct meta_tlvhdr {
+	__be16 type;
+	__be16 len;
+};
+
+static bool __ife_tlv_meta_valid(const unsigned char *skbdata,
+				 const unsigned char *ifehdr_end)
+{
+	const struct meta_tlvhdr *tlv;
+	u16 tlvlen;
+
+	if (unlikely(skbdata + sizeof(*tlv) > ifehdr_end))
+		return false;
+
+	tlv = (const struct meta_tlvhdr *)skbdata;
+	tlvlen = ntohs(tlv->len);
+
+	/* tlv length field is inc header, check on minimum */
+	if (tlvlen < NLA_HDRLEN)
+		return false;
+
+	/* overflow by NLA_ALIGN check */
+	if (NLA_ALIGN(tlvlen) < tlvlen)
+		return false;
+
+	if (unlikely(skbdata + NLA_ALIGN(tlvlen) > ifehdr_end))
+		return false;
+
+	return true;
+}
+
+/* Caller takes care of presenting data in network order
+ */
+void *ife_tlv_meta_decode(void *skbdata, const void *ifehdr_end, u16 *attrtype,
+			  u16 *dlen, u16 *totlen)
+{
+	struct meta_tlvhdr *tlv;
+
+	if (!__ife_tlv_meta_valid(skbdata, ifehdr_end))
+		return NULL;
+
+	tlv = (struct meta_tlvhdr *)skbdata;
+	*dlen = ntohs(tlv->len) - NLA_HDRLEN;
+	*attrtype = ntohs(tlv->type);
+
+	if (totlen)
+		*totlen = nla_total_size(*dlen);
+
+	return skbdata + sizeof(struct meta_tlvhdr);
+}
+EXPORT_SYMBOL_GPL(ife_tlv_meta_decode);
+
+void *ife_tlv_meta_next(void *skbdata)
+{
+	struct meta_tlvhdr *tlv = (struct meta_tlvhdr *) skbdata;
+	u16 tlvlen = ntohs(tlv->len);
+
+	tlvlen = NLA_ALIGN(tlvlen);
+
+	return skbdata + tlvlen;
+}
+EXPORT_SYMBOL_GPL(ife_tlv_meta_next);
+
+/* Caller takes care of presenting data in network order
+ */
+int ife_tlv_meta_encode(void *skbdata, u16 attrtype, u16 dlen, const void *dval)
+{
+	__be32 *tlv = (__be32 *) (skbdata);
+	u16 totlen = nla_total_size(dlen);	/*alignment + hdr */
+	char *dptr = (char *) tlv + NLA_HDRLEN;
+	u32 htlv = attrtype << 16 | (dlen + NLA_HDRLEN);
+
+	*tlv = htonl(htlv);
+	memset(dptr, 0, totlen - NLA_HDRLEN);
+	memcpy(dptr, dval, dlen);
+
+	return totlen;
+}
+EXPORT_SYMBOL_GPL(ife_tlv_meta_encode);
+
+MODULE_AUTHOR("Jamal Hadi Salim <jhs@mojatatu.com>");
+MODULE_AUTHOR("Yotam Gigi <yotam.gi@gmail.com>");
+MODULE_DESCRIPTION("Inter-FE LFB action");
+MODULE_LICENSE("GPL");