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, 184 insertions, 0 deletions

diff --git a/net/802/fddi.c b/net/802/fddi.c
new file mode 100644
index 000000000..e26f4549e
--- /dev/null
+++ b/net/802/fddi.c
@@ -0,0 +1,184 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * INET		An implementation of the TCP/IP protocol suite for the LINUX
+ *		operating system.  INET is implemented using the BSD Socket
+ *		interface as the means of communication with the user level.
+ *
+ *		FDDI-type device handling.
+ *
+ * Version:	@(#)fddi.c	1.0.0	08/12/96
+ *
+ * Authors:	Lawrence V. Stefani, <stefani@lkg.dec.com>
+ *
+ *		fddi.c is based on previous eth.c and tr.c work by
+ *			Ross Biro
+ *			Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
+ *			Mark Evans, <evansmp@uhura.aston.ac.uk>
+ *			Florian La Roche, <rzsfl@rz.uni-sb.de>
+ *			Alan Cox, <gw4pts@gw4pts.ampr.org>
+ *
+ *	Changes
+ *		Alan Cox		:	New arp/rebuild header
+ *		Maciej W. Rozycki	:	IPv6 support
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/socket.h>
+#include <linux/in.h>
+#include <linux/inet.h>
+#include <linux/netdevice.h>
+#include <linux/fddidevice.h>
+#include <linux/if_ether.h>
+#include <linux/skbuff.h>
+#include <linux/errno.h>
+#include <net/arp.h>
+#include <net/sock.h>
+
+/*
+ * Create the FDDI MAC header for an arbitrary protocol layer
+ *
+ * saddr=NULL	means use device source address
+ * daddr=NULL	means leave destination address (eg unresolved arp)
+ */
+
+static int fddi_header(struct sk_buff *skb, struct net_device *dev,
+		       unsigned short type,
+		       const void *daddr, const void *saddr, unsigned int len)
+{
+	int hl = FDDI_K_SNAP_HLEN;
+	struct fddihdr *fddi;
+
+	if(type != ETH_P_IP && type != ETH_P_IPV6 && type != ETH_P_ARP)
+		hl=FDDI_K_8022_HLEN-3;
+	fddi = skb_push(skb, hl);
+	fddi->fc			 = FDDI_FC_K_ASYNC_LLC_DEF;
+	if(type == ETH_P_IP || type == ETH_P_IPV6 || type == ETH_P_ARP)
+	{
+		fddi->hdr.llc_snap.dsap		 = FDDI_EXTENDED_SAP;
+		fddi->hdr.llc_snap.ssap		 = FDDI_EXTENDED_SAP;
+		fddi->hdr.llc_snap.ctrl		 = FDDI_UI_CMD;
+		fddi->hdr.llc_snap.oui[0]	 = 0x00;
+		fddi->hdr.llc_snap.oui[1]	 = 0x00;
+		fddi->hdr.llc_snap.oui[2]	 = 0x00;
+		fddi->hdr.llc_snap.ethertype	 = htons(type);
+	}
+
+	/* Set the source and destination hardware addresses */
+
+	if (saddr != NULL)
+		memcpy(fddi->saddr, saddr, dev->addr_len);
+	else
+		memcpy(fddi->saddr, dev->dev_addr, dev->addr_len);
+
+	if (daddr != NULL)
+	{
+		memcpy(fddi->daddr, daddr, dev->addr_len);
+		return hl;
+	}
+
+	return -hl;
+}
+
+/*
+ * Determine the packet's protocol ID and fill in skb fields.
+ * This routine is called before an incoming packet is passed
+ * up.  It's used to fill in specific skb fields and to set
+ * the proper pointer to the start of packet data (skb->data).
+ */
+
+__be16 fddi_type_trans(struct sk_buff *skb, struct net_device *dev)
+{
+	struct fddihdr *fddi = (struct fddihdr *)skb->data;
+	__be16 type;
+
+	/*
+	 * Set mac.raw field to point to FC byte, set data field to point
+	 * to start of packet data.  Assume 802.2 SNAP frames for now.
+	 */
+
+	skb->dev = dev;
+	skb_reset_mac_header(skb);	/* point to frame control (FC) */
+
+	if (skb->len < FDDI_K_8022_HLEN)
+		return htons(0);
+
+	if(fddi->hdr.llc_8022_1.dsap==0xe0)
+	{
+		skb_pull(skb, FDDI_K_8022_HLEN-3);
+		type = htons(ETH_P_802_2);
+	}
+	else
+	{
+		if (skb->len < FDDI_K_SNAP_HLEN)
+			return htons(0);
+		skb_pull(skb, FDDI_K_SNAP_HLEN);		/* adjust for 21 byte header */
+		type=fddi->hdr.llc_snap.ethertype;
+	}
+
+	/* Set packet type based on destination address and flag settings */
+
+	if (*fddi->daddr & 0x01)
+	{
+		if (memcmp(fddi->daddr, dev->broadcast, FDDI_K_ALEN) == 0)
+			skb->pkt_type = PACKET_BROADCAST;
+		else
+			skb->pkt_type = PACKET_MULTICAST;
+	}
+
+	else if (dev->flags & IFF_PROMISC)
+	{
+		if (memcmp(fddi->daddr, dev->dev_addr, FDDI_K_ALEN))
+			skb->pkt_type = PACKET_OTHERHOST;
+	}
+
+	/* Assume 802.2 SNAP frames, for now */
+
+	return type;
+}
+
+EXPORT_SYMBOL(fddi_type_trans);
+
+static const struct header_ops fddi_header_ops = {
+	.create		= fddi_header,
+};
+
+
+static void fddi_setup(struct net_device *dev)
+{
+	dev->header_ops		= &fddi_header_ops;
+	dev->type		= ARPHRD_FDDI;
+	dev->hard_header_len	= FDDI_K_SNAP_HLEN+3;	/* Assume 802.2 SNAP hdr len + 3 pad bytes */
+	dev->mtu		= FDDI_K_SNAP_DLEN;	/* Assume max payload of 802.2 SNAP frame */
+	dev->min_mtu		= FDDI_K_SNAP_HLEN;
+	dev->max_mtu		= FDDI_K_SNAP_DLEN;
+	dev->addr_len		= FDDI_K_ALEN;
+	dev->tx_queue_len	= 100;			/* Long queues on FDDI */
+	dev->flags		= IFF_BROADCAST | IFF_MULTICAST;
+
+	memset(dev->broadcast, 0xFF, FDDI_K_ALEN);
+}
+
+/**
+ * alloc_fddidev - Register FDDI device
+ * @sizeof_priv: Size of additional driver-private structure to be allocated
+ *	for this FDDI device
+ *
+ * Fill in the fields of the device structure with FDDI-generic values.
+ *
+ * Constructs a new net device, complete with a private data area of
+ * size @sizeof_priv.  A 32-byte (not bit) alignment is enforced for
+ * this private data area.
+ */
+struct net_device *alloc_fddidev(int sizeof_priv)
+{
+	return alloc_netdev(sizeof_priv, "fddi%d", NET_NAME_UNKNOWN,
+			    fddi_setup);
+}
+EXPORT_SYMBOL(alloc_fddidev);
+
+MODULE_DESCRIPTION("Core routines for FDDI network devices");
+MODULE_LICENSE("GPL");