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, 101 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-class-extcon b/Documentation/ABI/testing/sysfs-class-extcon
new file mode 100644
index 000000000..f8e705375
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-class-extcon
@@ -0,0 +1,101 @@
+What:		/sys/class/extcon/.../
+Date:		February 2012
+Contact:	MyungJoo Ham <myungjoo.ham@samsung.com>
+Description:
+		Provide a place in sysfs for the extcon objects.
+		This allows accessing extcon specific variables.
+		The name of extcon object denoted as ... is the name given
+		with extcon_dev_register.
+
+		One extcon device denotes a single external connector
+		port. An external connector may have multiple cables
+		attached simultaneously. Many of docks, cradles, and
+		accessory cables have such capability. For example,
+		the 30-pin port of Nuri board (/arch/arm/mach-exynos)
+		may have both HDMI and Charger attached, or analog audio,
+		video, and USB cables attached simultaneously.
+
+		If there are cables mutually exclusive with each other,
+		such binary relations may be expressed with extcon_dev's
+		mutually_exclusive array.
+
+What:		/sys/class/extcon/.../name
+Date:		February 2012
+Contact:	MyungJoo Ham <myungjoo.ham@samsung.com>
+Description:
+		The /sys/class/extcon/.../name shows the name of the extcon
+		object. If the extcon object has an optional callback
+		"show_name" defined, the callback will provide the name with
+		this sysfs node.
+
+What:		/sys/class/extcon/.../state
+Date:		February 2012
+Contact:	MyungJoo Ham <myungjoo.ham@samsung.com>
+Description:
+		The /sys/class/extcon/.../state shows and stores the cable
+		attach/detach information of the corresponding extcon object.
+		If the extcon object has an optional callback "show_state"
+		defined, the showing function is overridden with the optional
+		callback.
+
+		If the default callback for showing function is used, the
+		format is like this::
+
+		    # cat state
+		    USB_OTG=1
+		    HDMI=0
+		    TA=1
+		    EAR_JACK=0
+		    #
+
+		In this example, the extcon device has USB_OTG and TA
+		cables attached and HDMI and EAR_JACK cables detached.
+
+		In order to update the state of an extcon device, enter a hex
+		state number starting with 0x::
+
+		    # echo 0xHEX > state
+
+		This updates the whole state of the extcon device.
+		Inputs of all the methods are required to meet the
+		mutually_exclusive conditions if they exist.
+
+		It is recommended to use this "global" state interface if
+		you need to set the value atomically. The later state
+		interface associated with each cable cannot update
+		multiple cable states of an extcon device simultaneously.
+
+What:		/sys/class/extcon/.../cable.X/name
+Date:		February 2012
+Contact:	MyungJoo Ham <myungjoo.ham@samsung.com>
+Description:
+		The /sys/class/extcon/.../cable.X/name shows the name of cable
+		"X" (integer between 0 and 31) of an extcon device.
+
+What:		/sys/class/extcon/.../cable.X/state
+Date:		February 2012
+Contact:	MyungJoo Ham <myungjoo.ham@samsung.com>
+Description:
+		The /sys/class/extcon/.../cable.X/state shows and stores the
+		state of cable "X" (integer between 0 and 31) of an extcon
+		device. The state value is either 0 (detached) or 1
+		(attached).
+
+What:		/sys/class/extcon/.../mutually_exclusive/...
+Date:		December 2011
+Contact:	MyungJoo Ham <myungjoo.ham@samsung.com>
+Description:
+		Shows the relations of mutually exclusiveness. For example,
+		if the mutually_exclusive array of extcon device is
+		{0x3, 0x5, 0xC, 0x0}, then the output is::
+
+		    # ls mutually_exclusive/
+		    0x3
+		    0x5
+		    0xc
+		    #
+
+		Note that mutually_exclusive is a sub-directory of the extcon
+		device and the file names under the mutually_exclusive
+		directory show the mutually-exclusive sets, not the contents
+		of the files.