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, 50 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/rtc-cdev b/Documentation/ABI/testing/rtc-cdev
new file mode 100644
index 000000000..cec099a27
--- /dev/null
+++ b/Documentation/ABI/testing/rtc-cdev
@@ -0,0 +1,50 @@
+What:		/dev/rtcX
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	linux-rtc@vger.kernel.org
+Description:
+		The ioctl interface to drivers for real-time clocks (RTCs).
+		Following actions are supported:
+
+		* RTC_RD_TIME, RTC_SET_TIME: Read or set the RTC time. Time
+		  format is a Gregorian calendar date and 24 hour wall clock
+		  time.
+
+		* RTC_AIE_ON, RTC_AIE_OFF: Enable or disable the alarm interrupt
+		  for RTCs that support alarms
+
+		* RTC_ALM_READ, RTC_ALM_SET: Read or set the alarm time for
+		  RTCs that support alarms. Can be set up to 24 hours in the
+		  future. Requires a separate RTC_AIE_ON call to enable the
+		  alarm interrupt. (Prefer to use RTC_WKALM_*)
+
+		* RTC_WKALM_RD, RTC_WKALM_SET: For RTCs that support a more
+		  powerful interface, which can issue alarms beyond 24 hours and
+		  enable IRQs in the same request.
+
+		* RTC_PIE_ON, RTC_PIE_OFF: Enable or disable the periodic
+		  interrupt for RTCs that support periodic interrupts.
+
+		* RTC_UIE_ON, RTC_UIE_OFF: Enable or disable the update
+		  interrupt for RTCs that support it.
+
+		* RTC_IRQP_READ, RTC_IRQP_SET: Read or set the frequency for
+		  periodic interrupts for RTCs that support periodic interrupts.
+		  Requires a separate RTC_PIE_ON call to enable the periodic
+		  interrupts.
+
+		* RTC_VL_READ: Read the voltage inputs status of the RTC when
+		  supported. The value is a bit field of RTC_VL_*, giving the
+		  status of the main and backup voltages.
+
+		* RTC_VL_CLEAR: Clear the voltage status of the RTC. Some RTCs
+		  need user interaction when the backup power provider is
+		  replaced or charged to be able to clear the status.
+
+		The ioctl() calls supported by the older /dev/rtc interface are
+		also supported by the newer RTC class framework. However,
+		because the chips and systems are not standardized, some PC/AT
+		functionality might not be provided. And in the same way, some
+		newer features -- including those enabled by ACPI -- are exposed
+		by the RTC class framework, but can't be supported by the older
+		driver.