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, 39 insertions, 0 deletions

diff --git a/drivers/net/plip/Kconfig b/drivers/net/plip/Kconfig
new file mode 100644
index 000000000..67bfb01de
--- /dev/null
+++ b/drivers/net/plip/Kconfig
@@ -0,0 +1,39 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Parallel Line Internet Protocol (PLIP) network device configuration
+#
+
+config PLIP
+	tristate "PLIP (parallel port) support"
+	depends on PARPORT
+	help
+	  PLIP (Parallel Line Internet Protocol) is used to create a
+	  reasonably fast mini network consisting of two (or, rarely, more)
+	  local machines.  A PLIP link from a Linux box is a popular means to
+	  install a Linux distribution on a machine which doesn't have a
+	  CD-ROM drive (a minimal system has to be transferred with floppies
+	  first). The kernels on both machines need to have this PLIP option
+	  enabled for this to work.
+
+	  The PLIP driver has two modes, mode 0 and mode 1.  The parallel
+	  ports (the connectors at the computers with 25 holes) are connected
+	  with "null printer" or "Turbo Laplink" cables which can transmit 4
+	  bits at a time (mode 0) or with special PLIP cables, to be used on
+	  bidirectional parallel ports only, which can transmit 8 bits at a
+	  time (mode 1); you can find the wiring of these cables in
+	  <file:Documentation/networking/plip.rst>.  The cables can be up to
+	  15m long.  Mode 0 works also if one of the machines runs DOS/Windows
+	  and has some PLIP software installed, e.g. the Crynwr PLIP packet
+	  driver (<http://oak.oakland.edu/simtel.net/msdos/pktdrvr-pre.html>)
+	  and winsock or NCSA's telnet.
+
+	  If you want to use PLIP, say Y and read the PLIP mini-HOWTO as well
+	  as the NET-3-HOWTO, both available from
+	  <http://www.tldp.org/docs.html#howto>.  Note that the PLIP
+	  protocol has been changed and this PLIP driver won't work together
+	  with the PLIP support in Linux versions 1.0.x.  This option enlarges
+	  your kernel by about 8 KB.
+
+	  To compile this driver as a module, choose M here. The module
+	  will be called plip. If unsure, say Y or M, in case you buy
+	  a laptop later.