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

21 files changed, 117 insertions, 0 deletions

diff --git a/scripts/atomic/fallbacks/acquire b/scripts/atomic/fallbacks/acquire
new file mode 100755
index 000000000..4da0cab36
--- /dev/null
+++ b/scripts/atomic/fallbacks/acquire
@@ -0,0 +1,5 @@
+cat <<EOF
+	${ret} ret = arch_${atomic}_${pfx}${name}${sfx}_relaxed(${args});
+	__atomic_acquire_fence();
+	return ret;
+EOF
diff --git a/scripts/atomic/fallbacks/add_negative b/scripts/atomic/fallbacks/add_negative
new file mode 100755
index 000000000..1d3d4ab3a
--- /dev/null
+++ b/scripts/atomic/fallbacks/add_negative
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_${atomic}_add_return${order}(i, v) < 0;
+EOF
diff --git a/scripts/atomic/fallbacks/add_unless b/scripts/atomic/fallbacks/add_unless
new file mode 100755
index 000000000..95ecb2b74
--- /dev/null
+++ b/scripts/atomic/fallbacks/add_unless
@@ -0,0 +1,3 @@
+cat << EOF
+	return raw_${atomic}_fetch_add_unless(v, a, u) != u;
+EOF
diff --git a/scripts/atomic/fallbacks/andnot b/scripts/atomic/fallbacks/andnot
new file mode 100755
index 000000000..66760457e
--- /dev/null
+++ b/scripts/atomic/fallbacks/andnot
@@ -0,0 +1,3 @@
+cat <<EOF
+	${retstmt}raw_${atomic}_${pfx}and${sfx}${order}(~i, v);
+EOF
diff --git a/scripts/atomic/fallbacks/cmpxchg b/scripts/atomic/fallbacks/cmpxchg
new file mode 100644
index 000000000..1c8507f62
--- /dev/null
+++ b/scripts/atomic/fallbacks/cmpxchg
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_cmpxchg${order}(&v->counter, old, new);
+EOF
diff --git a/scripts/atomic/fallbacks/dec b/scripts/atomic/fallbacks/dec
new file mode 100755
index 000000000..60d286d40
--- /dev/null
+++ b/scripts/atomic/fallbacks/dec
@@ -0,0 +1,3 @@
+cat <<EOF
+	${retstmt}raw_${atomic}_${pfx}sub${sfx}${order}(1, v);
+EOF
diff --git a/scripts/atomic/fallbacks/dec_and_test b/scripts/atomic/fallbacks/dec_and_test
new file mode 100755
index 000000000..3a0278e0d
--- /dev/null
+++ b/scripts/atomic/fallbacks/dec_and_test
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_${atomic}_dec_return(v) == 0;
+EOF
diff --git a/scripts/atomic/fallbacks/dec_if_positive b/scripts/atomic/fallbacks/dec_if_positive
new file mode 100755
index 000000000..f65c11b4b
--- /dev/null
+++ b/scripts/atomic/fallbacks/dec_if_positive
@@ -0,0 +1,11 @@
+cat <<EOF
+	${int} dec, c = raw_${atomic}_read(v);
+
+	do {
+		dec = c - 1;
+		if (unlikely(dec < 0))
+			break;
+	} while (!raw_${atomic}_try_cmpxchg(v, &c, dec));
+
+	return dec;
+EOF
diff --git a/scripts/atomic/fallbacks/dec_unless_positive b/scripts/atomic/fallbacks/dec_unless_positive
new file mode 100755
index 000000000..d025361d7
--- /dev/null
+++ b/scripts/atomic/fallbacks/dec_unless_positive
@@ -0,0 +1,10 @@
+cat <<EOF
+	${int} c = raw_${atomic}_read(v);
+
+	do {
+		if (unlikely(c > 0))
+			return false;
+	} while (!raw_${atomic}_try_cmpxchg(v, &c, c - 1));
+
+	return true;
+EOF
diff --git a/scripts/atomic/fallbacks/fence b/scripts/atomic/fallbacks/fence
new file mode 100755
index 000000000..40d5b3976
--- /dev/null
+++ b/scripts/atomic/fallbacks/fence
@@ -0,0 +1,7 @@
+cat <<EOF
+	${ret} ret;
+	__atomic_pre_full_fence();
+	ret = arch_${atomic}_${pfx}${name}${sfx}_relaxed(${args});
+	__atomic_post_full_fence();
+	return ret;
+EOF
diff --git a/scripts/atomic/fallbacks/fetch_add_unless b/scripts/atomic/fallbacks/fetch_add_unless
new file mode 100755
index 000000000..8db7e9e17
--- /dev/null
+++ b/scripts/atomic/fallbacks/fetch_add_unless
@@ -0,0 +1,10 @@
+cat << EOF
+	${int} c = raw_${atomic}_read(v);
+
+	do {
+		if (unlikely(c == u))
+			break;
+	} while (!raw_${atomic}_try_cmpxchg(v, &c, c + a));
+
+	return c;
+EOF
diff --git a/scripts/atomic/fallbacks/inc b/scripts/atomic/fallbacks/inc
new file mode 100755
index 000000000..56c770f59
--- /dev/null
+++ b/scripts/atomic/fallbacks/inc
@@ -0,0 +1,3 @@
+cat <<EOF
+	${retstmt}raw_${atomic}_${pfx}add${sfx}${order}(1, v);
+EOF
diff --git a/scripts/atomic/fallbacks/inc_and_test b/scripts/atomic/fallbacks/inc_and_test
new file mode 100755
index 000000000..7d16a10f2
--- /dev/null
+++ b/scripts/atomic/fallbacks/inc_and_test
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_${atomic}_inc_return(v) == 0;
+EOF
diff --git a/scripts/atomic/fallbacks/inc_not_zero b/scripts/atomic/fallbacks/inc_not_zero
new file mode 100755
index 000000000..1fcef1e55
--- /dev/null
+++ b/scripts/atomic/fallbacks/inc_not_zero
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_${atomic}_add_unless(v, 1, 0);
+EOF
diff --git a/scripts/atomic/fallbacks/inc_unless_negative b/scripts/atomic/fallbacks/inc_unless_negative
new file mode 100755
index 000000000..7b4b09868
--- /dev/null
+++ b/scripts/atomic/fallbacks/inc_unless_negative
@@ -0,0 +1,10 @@
+cat <<EOF
+	${int} c = raw_${atomic}_read(v);
+
+	do {
+		if (unlikely(c < 0))
+			return false;
+	} while (!raw_${atomic}_try_cmpxchg(v, &c, c + 1));
+
+	return true;
+EOF
diff --git a/scripts/atomic/fallbacks/read_acquire b/scripts/atomic/fallbacks/read_acquire
new file mode 100755
index 000000000..e319862d2
--- /dev/null
+++ b/scripts/atomic/fallbacks/read_acquire
@@ -0,0 +1,12 @@
+cat <<EOF
+	${int} ret;
+
+	if (__native_word(${atomic}_t)) {
+		ret = smp_load_acquire(&(v)->counter);
+	} else {
+		ret = raw_${atomic}_read(v);
+		__atomic_acquire_fence();
+	}
+
+	return ret;
+EOF
diff --git a/scripts/atomic/fallbacks/release b/scripts/atomic/fallbacks/release
new file mode 100755
index 000000000..1e6daf57b
--- /dev/null
+++ b/scripts/atomic/fallbacks/release
@@ -0,0 +1,4 @@
+cat <<EOF
+	__atomic_release_fence();
+	${retstmt}arch_${atomic}_${pfx}${name}${sfx}_relaxed(${args});
+EOF
diff --git a/scripts/atomic/fallbacks/set_release b/scripts/atomic/fallbacks/set_release
new file mode 100755
index 000000000..16a374ae6
--- /dev/null
+++ b/scripts/atomic/fallbacks/set_release
@@ -0,0 +1,8 @@
+cat <<EOF
+	if (__native_word(${atomic}_t)) {
+		smp_store_release(&(v)->counter, i);
+	} else {
+		__atomic_release_fence();
+		raw_${atomic}_set(v, i);
+	}
+EOF
diff --git a/scripts/atomic/fallbacks/sub_and_test b/scripts/atomic/fallbacks/sub_and_test
new file mode 100755
index 000000000..d1f746fe0
--- /dev/null
+++ b/scripts/atomic/fallbacks/sub_and_test
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_${atomic}_sub_return(i, v) == 0;
+EOF
diff --git a/scripts/atomic/fallbacks/try_cmpxchg b/scripts/atomic/fallbacks/try_cmpxchg
new file mode 100755
index 000000000..d4da82092
--- /dev/null
+++ b/scripts/atomic/fallbacks/try_cmpxchg
@@ -0,0 +1,7 @@
+cat <<EOF
+	${int} r, o = *old;
+	r = raw_${atomic}_cmpxchg${order}(v, o, new);
+	if (unlikely(r != o))
+		*old = r;
+	return likely(r == o);
+EOF
diff --git a/scripts/atomic/fallbacks/xchg b/scripts/atomic/fallbacks/xchg
new file mode 100644
index 000000000..e4def1e0d
--- /dev/null
+++ b/scripts/atomic/fallbacks/xchg
@@ -0,0 +1,3 @@
+cat <<EOF
+	return raw_xchg${order}(&v->counter, new);
+EOF