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, 111 insertions, 0 deletions

diff --git a/lib/crypto/tests/test-utils.h b/lib/crypto/tests/test-utils.h
new file mode 100644
index 000000000..73ca21f91
--- /dev/null
+++ b/lib/crypto/tests/test-utils.h
@@ -0,0 +1,111 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Test utility functions shared by the crypto library tests.
+ *
+ * For now this is simply a header that's included into the KUnit test suites
+ * that need it.  If this gets too large it could be made its own translation
+ * unit and libcrypto_test_utils module, but that seems overkill for now.
+ */
+#ifndef LIB_CRYPTO_TEST_UTILS_H
+#define LIB_CRYPTO_TEST_UTILS_H
+
+#include <kunit/test.h>
+#include <linux/math.h>
+#include <linux/minmax.h>
+#include <linux/string.h>
+#include <linux/vmalloc.h>
+
+static u64 random_seed;
+
+static __maybe_unused void action_free_guarded_buf(void *buf)
+{
+	vfree(buf);
+}
+
+/*
+ * Allocate a KUnit-managed buffer that has length @size bytes (> 0) immediately
+ * followed by an unmapped page, and assert that the allocation succeeds.
+ */
+static __maybe_unused void *alloc_guarded_buf(struct kunit *test, size_t size)
+{
+	size_t full_size = round_up(size, PAGE_SIZE);
+	void *buf = vmalloc(full_size);
+
+	KUNIT_ASSERT_NOT_NULL(test, buf);
+	KUNIT_ASSERT_EQ(test, 0,
+			kunit_add_action_or_reset(test, action_free_guarded_buf,
+						  buf));
+	return buf + full_size - size;
+}
+
+static __maybe_unused void *alloc_buf(struct kunit *test, size_t size)
+{
+	void *buf = kunit_kmalloc(test, size, GFP_KERNEL);
+
+	KUNIT_ASSERT_NOT_NULL(test, buf);
+	return buf;
+}
+
+static __maybe_unused void *memdup_buf(struct kunit *test, const void *src,
+				       size_t size)
+{
+	void *dst = alloc_buf(test, size);
+
+	return memcpy(dst, src, size);
+}
+
+/*
+ * This is a simple linear congruential generator.  It is used only for testing,
+ * which does not require cryptographically secure random numbers.  A hard-coded
+ * algorithm is used instead of <linux/prandom.h> so that it matches the
+ * algorithm used by the test vector generation script.  This allows the input
+ * data in random test vectors to be concisely stored as just the seed.
+ */
+static __maybe_unused u32 rand32(void)
+{
+	random_seed = (random_seed * 25214903917 + 11) & ((1ULL << 48) - 1);
+	return random_seed >> 16;
+}
+
+static __maybe_unused void rand_bytes(u8 *out, size_t len)
+{
+	for (size_t i = 0; i < len; i++)
+		out[i] = rand32();
+}
+
+static __maybe_unused void rand_bytes_seeded_from_len(u8 *out, size_t len)
+{
+	random_seed = len;
+	rand_bytes(out, len);
+}
+
+static __maybe_unused bool rand_bool(void)
+{
+	return rand32() % 2;
+}
+
+/* Generate a random length, preferring small lengths. */
+static __maybe_unused size_t rand_length(size_t max_len)
+{
+	size_t len;
+
+	switch (rand32() % 3) {
+	case 0:
+		len = rand32() % 128;
+		break;
+	case 1:
+		len = rand32() % 3072;
+		break;
+	default:
+		len = rand32();
+		break;
+	}
+	return len % (max_len + 1);
+}
+
+static __maybe_unused size_t rand_offset(size_t max_offset)
+{
+	return min(rand32() % 128, max_offset);
+}
+
+#endif /* LIB_CRYPTO_TEST_UTILS_H */