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, 95 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-kernel-livepatch b/Documentation/ABI/testing/sysfs-kernel-livepatch
new file mode 100644
index 000000000..3c3f36b32
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-kernel-livepatch
@@ -0,0 +1,95 @@
+What:		/sys/kernel/livepatch
+Date:		Nov 2014
+KernelVersion:	3.19.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		Interface for kernel live patching
+
+		The /sys/kernel/livepatch directory contains subdirectories for
+		each loaded live patch module.
+
+What:		/sys/kernel/livepatch/<patch>
+Date:		Nov 2014
+KernelVersion:	3.19.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		The patch directory contains subdirectories for each kernel
+		object (vmlinux or a module) in which it patched functions.
+
+What:		/sys/kernel/livepatch/<patch>/enabled
+Date:		Nov 2014
+KernelVersion:	3.19.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		A writable attribute that indicates whether the patched
+		code is currently applied.  Writing 0 will disable the patch
+		while writing 1 will re-enable the patch.
+
+What:		/sys/kernel/livepatch/<patch>/transition
+Date:		Feb 2017
+KernelVersion:	4.12.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		An attribute which indicates whether the patch is currently in
+		transition.
+
+What:		/sys/kernel/livepatch/<patch>/force
+Date:		Nov 2017
+KernelVersion:	4.15.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		A writable attribute that allows administrator to affect the
+		course of an existing transition. Writing 1 clears
+		TIF_PATCH_PENDING flag of all tasks and thus forces the tasks to
+		the patched or unpatched state. Administrator should not
+		use this feature without a clearance from a patch
+		distributor. Removal (rmmod) of patch modules is permanently
+		disabled when the feature is used. See
+		Documentation/livepatch/livepatch.rst for more information.
+
+What:		/sys/kernel/livepatch/<patch>/replace
+Date:		Jun 2024
+KernelVersion:	6.11.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		An attribute which indicates whether the patch supports
+		atomic-replace.
+
+What:		/sys/kernel/livepatch/<patch>/stack_order
+Date:		Jan 2025
+KernelVersion:	6.14.0
+Description:
+		This attribute specifies the sequence in which live patch modules
+		are applied to the system. If multiple live patches modify the same
+		function, the implementation with the biggest 'stack_order' number
+		is used, unless a transition is currently in progress.
+
+What:		/sys/kernel/livepatch/<patch>/<object>
+Date:		Nov 2014
+KernelVersion:	3.19.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		The object directory contains subdirectories for each function
+		that is patched within the object.
+
+What:		/sys/kernel/livepatch/<patch>/<object>/patched
+Date:		August 2022
+KernelVersion:	6.1.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		An attribute which indicates whether the object is currently
+		patched.
+
+What:		/sys/kernel/livepatch/<patch>/<object>/<function,sympos>
+Date:		Nov 2014
+KernelVersion:	3.19.0
+Contact:	live-patching@vger.kernel.org
+Description:
+		The function directory contains attributes regarding the
+		properties and state of the patched function.
+
+		The directory name contains the patched function name and a
+		sympos number corresponding to the nth occurrence of the symbol
+		name in kallsyms for the patched object.
+
+		There are currently no such attributes.