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, 76 insertions, 0 deletions

diff --git a/Documentation/admin-guide/media/remote-controller.rst b/Documentation/admin-guide/media/remote-controller.rst
new file mode 100644
index 000000000..188944b00
--- /dev/null
+++ b/Documentation/admin-guide/media/remote-controller.rst
@@ -0,0 +1,76 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+======================================================
+Infrared remote control support in video4linux drivers
+======================================================
+
+Authors: Gerd Hoffmann, Mauro Carvalho Chehab
+
+Basics
+======
+
+Most analog and digital TV boards support remote controllers. Several of
+them have a microprocessor that receives the IR carriers, convert into
+pulse/space sequences and then to scan codes, returning such codes to
+userspace ("scancode mode"). Other boards return just the pulse/space
+sequences ("raw mode").
+
+The support for remote controller in scancode mode is provided by the
+standard Linux input layer. The support for raw mode is provided via LIRC.
+
+In order to check the support and test it, it is suggested to download
+the `v4l-utils <https://git.linuxtv.org/v4l-utils.git/>`_. It provides
+two tools to handle remote controllers:
+
+- ir-keytable: provides a way to query the remote controller, list the
+  protocols it supports, enable in-kernel support for IR decoder or
+  switch the protocol and to test the reception of scan codes;
+
+- ir-ctl: provide tools to handle remote controllers that support raw mode
+  via LIRC interface.
+
+Usually, the remote controller module is auto-loaded when the TV card is
+detected. However, for a few devices, you need to manually load the
+ir-kbd-i2c module.
+
+How it works
+============
+
+The modules register the remote as keyboard within the linux input
+layer, i.e. you'll see the keys of the remote as normal key strokes
+(if CONFIG_INPUT_KEYBOARD is enabled).
+
+Using the event devices (CONFIG_INPUT_EVDEV) it is possible for
+applications to access the remote via /dev/input/event<n> devices.
+The udev/systemd will automatically create the devices. If you install
+the `v4l-utils <https://git.linuxtv.org/v4l-utils.git/>`_, it may also
+automatically load a different keytable than the default one. Please see
+`v4l-utils <https://git.linuxtv.org/v4l-utils.git/>`_ ir-keytable.1
+man page for details.
+
+The ir-keytable tool is nice for trouble shooting, i.e. to check
+whenever the input device is really present, which of the devices it
+is, check whenever pressing keys on the remote actually generates
+events and the like.  You can also use any other input utility that changes
+the keymaps, like the input kbd utility.
+
+
+Using with lircd
+----------------
+
+The latest versions of the lircd daemon supports reading events from the
+linux input layer (via event device). It also supports receiving IR codes
+in lirc mode.
+
+
+Using without lircd
+-------------------
+
+Xorg recognizes several IR keycodes that have its numerical value lower
+than 247. With the advent of Wayland, the input driver got updated too,
+and should now accept all keycodes. Yet, you may want to just reassign
+the keycodes to something that your favorite media application likes.
+
+This can be done by setting
+`v4l-utils <https://git.linuxtv.org/v4l-utils.git/>`_ to load your own
+keytable in runtime. Please read  ir-keytable.1 man page for details.