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, 109 insertions, 0 deletions

diff --git a/net/l2tp/Kconfig b/net/l2tp/Kconfig
new file mode 100644
index 000000000..0de178d5b
--- /dev/null
+++ b/net/l2tp/Kconfig
@@ -0,0 +1,109 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Layer Two Tunneling Protocol (L2TP)
+#
+
+menuconfig L2TP
+	tristate "Layer Two Tunneling Protocol (L2TP)"
+	depends on INET
+	select NET_UDP_TUNNEL
+	help
+	  Layer Two Tunneling Protocol
+
+	  From RFC 2661 <http://www.ietf.org/rfc/rfc2661.txt>.
+
+	  L2TP facilitates the tunneling of packets across an
+	  intervening network in a way that is as transparent as
+	  possible to both end-users and applications.
+
+	  L2TP is often used to tunnel PPP traffic over IP
+	  tunnels. One IP tunnel may carry thousands of individual PPP
+	  connections. L2TP is also used as a VPN protocol, popular
+	  with home workers to connect to their offices.
+
+	  L2TPv3 allows other protocols as well as PPP to be carried
+	  over L2TP tunnels. L2TPv3 is defined in RFC 3931
+	  <http://www.ietf.org/rfc/rfc3931.txt>.
+
+	  The kernel component handles only L2TP data packets: a
+	  userland daemon handles L2TP the control protocol (tunnel
+	  and session setup). One such daemon is OpenL2TP
+	  (http://openl2tp.org/).
+
+	  If you don't need L2TP, say N. To compile all L2TP code as
+	  modules, choose M here.
+
+config L2TP_DEBUGFS
+	tristate "L2TP debugfs support"
+	depends on L2TP && DEBUG_FS
+	help
+	  Support for l2tp directory in debugfs filesystem. This may be
+	  used to dump internal state of the l2tp drivers for problem
+	  analysis.
+
+	  If unsure, say 'Y'.
+
+	  To compile this driver as a module, choose M here. The module
+	  will be called l2tp_debugfs.
+
+config L2TP_V3
+	bool "L2TPv3 support"
+	depends on L2TP
+	help
+	  Layer Two Tunneling Protocol Version 3
+
+	  From RFC 3931 <http://www.ietf.org/rfc/rfc3931.txt>.
+
+	  The Layer Two Tunneling Protocol (L2TP) provides a dynamic
+	  mechanism for tunneling Layer 2 (L2) "circuits" across a
+	  packet-oriented data network (e.g., over IP).  L2TP, as
+	  originally defined in RFC 2661, is a standard method for
+	  tunneling Point-to-Point Protocol (PPP) [RFC1661] sessions.
+	  L2TP has since been adopted for tunneling a number of other
+	  L2 protocols, including ATM, Frame Relay, HDLC and even raw
+	  ethernet frames.
+
+	  If you are connecting to L2TPv3 equipment, or you want to
+	  tunnel raw ethernet frames using L2TP, say Y here. If
+	  unsure, say N.
+
+config L2TP_IP
+	tristate "L2TP IP encapsulation for L2TPv3"
+	depends on L2TP_V3
+	help
+	  Support for L2TP-over-IP socket family.
+
+	  The L2TPv3 protocol defines two possible encapsulations for
+	  L2TP frames, namely UDP and plain IP (without UDP). This
+	  driver provides a new L2TPIP socket family with which
+	  userspace L2TPv3 daemons may create L2TP/IP tunnel sockets
+	  when UDP encapsulation is not required. When L2TP is carried
+	  in IP packets, it used IP protocol number 115, so this port
+	  must be enabled in firewalls.
+
+	  To compile this driver as a module, choose M here. The module
+	  will be called l2tp_ip.
+
+config L2TP_ETH
+	tristate "L2TP ethernet pseudowire support for L2TPv3"
+	depends on L2TP_V3
+	help
+	  Support for carrying raw ethernet frames over L2TPv3.
+
+	  From RFC 4719 <http://www.ietf.org/rfc/rfc4719.txt>.
+
+	  The Layer 2 Tunneling Protocol, Version 3 (L2TPv3) can be
+	  used as a control protocol and for data encapsulation to set
+	  up Pseudowires for transporting layer 2 Packet Data Units
+	  across an IP network [RFC3931].
+
+	  This driver provides an ethernet virtual interface for each
+	  L2TP ethernet pseudowire instance. Standard Linux tools may
+	  be used to assign an IP address to the local virtual
+	  interface, or add the interface to a bridge.
+
+	  If you are using L2TPv3, you will almost certainly want to
+	  enable this option.
+
+	  To compile this driver as a module, choose M here. The module
+	  will be called l2tp_eth.