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, 51 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-uevent b/Documentation/ABI/testing/sysfs-uevent
new file mode 100644
index 000000000..c15e18c47
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-uevent
@@ -0,0 +1,51 @@
+What:           /sys/.../uevent
+Date:           May 2017
+KernelVersion:  4.13
+Contact:        Linux kernel mailing list <linux-kernel@vger.kernel.org>
+Description:
+                Enable passing additional variables for synthetic uevents that
+                are generated by writing /sys/.../uevent file.
+
+                Recognized extended format is::
+
+			ACTION [UUID [KEY=VALUE ...]]
+
+                The ACTION is compulsory - it is the name of the uevent
+                action (``add``, ``change``, ``remove``). There is no change
+                compared to previous functionality here. The rest of the
+                extended format is optional.
+
+                You need to pass UUID first before any KEY=VALUE pairs.
+                The UUID must be in ``xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx``
+                format where 'x' is a hex digit. The UUID is considered to be
+                a transaction identifier so it's possible to use the same UUID
+                value for one or more synthetic uevents in which case we
+                logically group these uevents together for any userspace
+                listeners. The UUID value appears in uevent as
+                ``SYNTH_UUID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx`` environment
+                variable.
+
+                If UUID is not passed in, the generated synthetic uevent gains
+                ``SYNTH_UUID=0`` environment variable automatically.
+
+                The KEY=VALUE pairs can contain alphanumeric characters only.
+
+                It's possible to define zero or more pairs - each pair is then
+                delimited by a space character ' '. Each pair appears in
+                synthetic uevent as ``SYNTH_ARG_KEY=VALUE``. That means the KEY
+                name gains ``SYNTH_ARG_`` prefix to avoid possible collisions
+                with existing variables.
+
+                Example of valid sequence written to the uevent file::
+
+                    add fe4d7c9d-b8c6-4a70-9ef1-3d8a58d18eed A=1 B=abc
+
+                This generates synthetic uevent including these variables::
+
+                    ACTION=add
+                    SYNTH_ARG_A=1
+                    SYNTH_ARG_B=abc
+                    SYNTH_UUID=fe4d7c9d-b8c6-4a70-9ef1-3d8a58d18eed
+
+Users:
+                udev, userspace tools generating synthetic uevents