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, 90 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-bus-siox b/Documentation/ABI/testing/sysfs-bus-siox
new file mode 100644
index 000000000..50e80238f
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-bus-siox
@@ -0,0 +1,90 @@
+What:		/sys/bus/siox/devices/siox-X/active
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		On reading represents the current state of the bus. If it
+		contains a "0" the bus is stopped and connected devices are
+		expected to not do anything because their watchdog triggered.
+		When the file contains a "1" the bus is operated and periodically
+		does a push-pull cycle to write and read data from the
+		connected devices.
+
+		When writing a "0" or "1" the bus moves to the described state.
+
+What:		/sys/bus/siox/devices/siox-X/device_add
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Write-only file. Write
+
+			<type> <inbytes> <outbytes> <statustype>
+
+		to add a new device dynamically. <type> is the name that is used to match
+		to a driver (similar to the platform bus). <inbytes> and <outbytes> define
+		the length of the input and output shift register in bytes respectively.
+
+		<statustype> defines the 4 bit device type that is check to identify connection
+		problems.
+
+		The new device is added to the end of the existing chain.
+
+What:		/sys/bus/siox/devices/siox-X/device_remove
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Write-only file. A single write removes the last device in the siox chain.
+
+What:		/sys/bus/siox/devices/siox-X/poll_interval_ns
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Defines the interval between two poll cycles in nano seconds.
+		Note this is rounded to jiffies on writing. On reading the current value
+		is returned.
+
+What:		/sys/bus/siox/devices/siox-X-Y/connected
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Read-only value. "0" means the Yth device on siox bus X isn't "connected" i.e.
+		communication with it is not ensured. "1" signals a working connection.
+
+What:		/sys/bus/siox/devices/siox-X-Y/inbytes
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Read-only value reporting the inbytes value provided to siox-X/device_add
+
+What:		/sys/bus/siox/devices/siox-X-Y/status_errors
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Counts the number of time intervals when the read status byte doesn't yield the
+		expected value.
+
+What:		/sys/bus/siox/devices/siox-X-Y/type
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Read-only value reporting the type value provided to siox-X/device_add.
+
+What:		/sys/bus/siox/devices/siox-X-Y/watchdog
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Read-only value reporting if the watchdog of the siox device is
+		active. "0" means the watchdog is not active and the device is expected to
+		be operational. "1" means the watchdog keeps the device in reset.
+
+What:		/sys/bus/siox/devices/siox-X-Y/watchdog_errors
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Read-only value reporting the number to time intervals when the
+		watchdog was active.
+
+What:		/sys/bus/siox/devices/siox-X-Y/outbytes
+KernelVersion:	4.16
+Contact:	Thorsten Scherer <t.scherer@eckelmann.de>, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
+Description:
+		Read-only value reporting the outbytes value provided to siox-X/device_add.