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, 86 insertions, 0 deletions

diff --git a/drivers/net/arcnet/Kconfig b/drivers/net/arcnet/Kconfig
new file mode 100644
index 000000000..4611a3716
--- /dev/null
+++ b/drivers/net/arcnet/Kconfig
@@ -0,0 +1,86 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Arcnet configuration
+#
+
+menuconfig ARCNET
+	depends on NETDEVICES && PCI && HAS_IOPORT
+	tristate "ARCnet support"
+	help
+	  If you have a network card of this type, say Y and check out the
+	  (arguably) beautiful poetry in
+	  <file:Documentation/networking/arcnet.rst>.
+
+	  You need both this driver, and the driver for the particular ARCnet
+	  chipset of your card.
+
+	  To compile this driver as a module, choose M here. The module will
+	  be called arcnet.
+
+if ARCNET
+
+config ARCNET_1201
+	tristate "Enable standard ARCNet packet format (RFC 1201)"
+	help
+	  This allows you to use RFC1201 with your ARCnet card via the virtual
+	  arc0 device.  You need to say Y here to communicate with
+	  industry-standard RFC1201 implementations, like the arcether.com
+	  packet driver or most DOS/Windows ODI drivers.  Please read the
+	  ARCnet documentation in <file:Documentation/networking/arcnet.rst>
+	  for more information about using arc0.
+
+config ARCNET_1051
+	tristate "Enable old ARCNet packet format (RFC 1051)"
+	help
+	  This allows you to use RFC1051 with your ARCnet card via the virtual
+	  arc0s device. You only need arc0s if you want to talk to ARCnet
+	  software complying with the "old" standard, specifically, the DOS
+	  arcnet.com packet driver, Amigas running AmiTCP, and some variants
+	  of NetBSD. You do not need to say Y here to communicate with
+	  industry-standard RFC1201 implementations, like the arcether.com
+	  packet driver or most DOS/Windows ODI drivers. RFC1201 is included
+	  automatically as the arc0 device. Please read the ARCnet
+	  documentation in <file:Documentation/networking/arcnet.rst> for more
+	  information about using arc0e and arc0s.
+
+config ARCNET_RAW
+	tristate "Enable raw mode packet interface"
+	help
+	  ARCnet "raw mode" packet encapsulation, no soft headers.  Unlikely
+	  to work unless talking to a copy of the same Linux arcnet driver,
+	  but perhaps marginally faster in that case.
+
+config ARCNET_CAP
+	tristate "Enable CAP mode packet interface"
+	help
+	  ARCnet "cap mode" packet encapsulation. Used to get the hardware
+	  acknowledge back to userspace. After the initial protocol byte every
+	  packet is stuffed with an extra 4 byte "cookie" which doesn't
+	  actually appear on the network. After transmit the driver will send
+	  back a packet with protocol byte 0 containing the status of the
+	  transmission:
+	     0=no hardware acknowledge
+	     1=excessive nak
+	     2=transmission accepted by the receiver hardware
+
+	  Received packets are also stuffed with the extra 4 bytes but it will
+	  be random data.
+
+	  Cap only listens to protocol 1-8.
+
+config ARCNET_COM20020
+	tristate "ARCnet COM20020 chipset driver"
+	depends on LEDS_CLASS
+	help
+	  This is the driver for the new COM20020 chipset. It supports such
+	  things as promiscuous mode, so packet sniffing is possible, and
+	  extra diagnostic information.
+
+	  To compile this driver as a module, choose M here. The module will
+	  be called com20020.
+
+config ARCNET_COM20020_PCI
+	tristate "Support for COM20020 on PCI"
+	depends on ARCNET_COM20020 && PCI
+
+endif # ARCNET