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, 103 insertions, 0 deletions

diff --git a/net/rxrpc/Kconfig b/net/rxrpc/Kconfig
new file mode 100644
index 000000000..2cebb666d
--- /dev/null
+++ b/net/rxrpc/Kconfig
@@ -0,0 +1,103 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# RxRPC session sockets
+#
+
+config AF_RXRPC
+	tristate "RxRPC session sockets"
+	depends on INET
+	select CRYPTO
+	select CRYPTO_LIB_DES if RXKAD
+	select KEYS
+	select NET_UDP_TUNNEL
+	help
+	  Say Y or M here to include support for RxRPC session sockets (just
+	  the transport part, not the presentation part: (un)marshalling is
+	  left to the application).
+
+	  These are used for AFS kernel filesystem and userspace utilities.
+
+	  This module at the moment only supports client operations and is
+	  currently incomplete.
+
+	  See Documentation/networking/rxrpc.rst.
+
+if AF_RXRPC
+
+config AF_RXRPC_IPV6
+	bool "IPv6 support for RxRPC"
+	depends on IPV6
+	help
+	  Say Y here to allow AF_RXRPC to use IPV6 UDP as well as IPV4 UDP as
+	  its network transport.
+
+config AF_RXRPC_INJECT_LOSS
+	bool "Inject packet loss into RxRPC packet stream"
+	help
+	  Say Y here to inject packet loss by discarding some received and some
+	  transmitted packets.
+
+config AF_RXRPC_INJECT_RX_DELAY
+	bool "Inject delay into packet reception"
+	depends on SYSCTL
+	help
+	  Say Y here to inject a delay into packet reception, allowing an
+	  extended RTT time to be modelled.  The delay can be configured using
+	  /proc/sys/net/rxrpc/rxrpc_inject_rx_delay, setting a number of
+	  milliseconds up to 0.5s (note that the granularity is actually in
+	  jiffies).
+
+config AF_RXRPC_DEBUG
+	bool "RxRPC dynamic debugging"
+	help
+	  Say Y here to make runtime controllable debugging messages appear.
+
+	  See Documentation/networking/rxrpc.rst.
+
+
+config RXKAD
+	bool "RxRPC Kerberos security"
+	help
+	  Provide kerberos 4 and AFS kaserver security handling for AF_RXRPC
+	  through the use of the key retention service.
+
+	  See Documentation/networking/rxrpc.rst.
+
+config RXGK
+	bool "RxRPC GSSAPI security"
+	select CRYPTO_KRB5
+	select CRYPTO_MANAGER
+	select CRYPTO_KRB5ENC
+	select CRYPTO_AUTHENC
+	select CRYPTO_SKCIPHER
+	select CRYPTO_HASH_INFO
+	select CRYPTO_HMAC
+	select CRYPTO_CMAC
+	select CRYPTO_SHA1
+	select CRYPTO_SHA256
+	select CRYPTO_SHA512
+	select CRYPTO_CBC
+	select CRYPTO_CTS
+	select CRYPTO_AES
+	select CRYPTO_CAMELLIA
+	help
+	  Provide the GSSAPI-based RxGK security class for AFS.  Keys are added
+	  with add_key().
+
+	  See Documentation/networking/rxrpc.rst.
+
+config RXPERF
+	tristate "RxRPC test service"
+	help
+	  Provide an rxperf service tester.  This listens on UDP port 7009 for
+	  incoming calls from the rxperf program (an example of which can be
+	  found in OpenAFS).
+
+config AF_RXRPC_KUNIT_TEST
+	tristate "RxRPC crypto KUnit test" if !KUNIT_ALL_TESTS
+	depends on KUNIT && RXKAD
+	default KUNIT_ALL_TESTS
+	help
+	  Enable the KUnit test suite for RxRPC's crypto routines.
+
+endif