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, 52 insertions, 0 deletions

diff --git a/include/vdso/futex.h b/include/vdso/futex.h
new file mode 100644
index 000000000..3cd175eef
--- /dev/null
+++ b/include/vdso/futex.h
@@ -0,0 +1,52 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _VDSO_FUTEX_H
+#define _VDSO_FUTEX_H
+
+#include <uapi/linux/types.h>
+
+/**
+ * __vdso_futex_robust_list64_try_unlock - Try to unlock an uncontended robust futex
+ *					   with a 64-bit pending op pointer
+ * @lock:	Pointer to the futex lock object
+ * @tid:	The TID of the calling task
+ * @pop:	Pointer to the task's robust_list_head::list_pending_op
+ *
+ * Return: The content of *@lock. On success this is the same as @tid.
+ *
+ * The function implements:
+ *	if (atomic_try_cmpxchg(lock, &tid, 0))
+ *		*op = NULL;
+ *	return tid;
+ *
+ * There is a race between a successful unlock and clearing the pending op
+ * pointer in the robust list head. If the calling task is interrupted in the
+ * race window and has to handle a (fatal) signal on return to user space then
+ * the kernel handles the clearing of @pending_op before attempting to deliver
+ * the signal. That ensures that a task cannot exit with a potentially invalid
+ * pending op pointer.
+ *
+ * User space uses it in the following way:
+ *
+ * if (__vdso_futex_robust_list64_try_unlock(lock, tid, &pending_op) != tid)
+ *	err = sys_futex($OP | FUTEX_ROBUST_UNLOCK,....);
+ *
+ * If the unlock attempt fails due to the FUTEX_WAITERS bit set in the lock,
+ * then the syscall does the unlock, clears the pending op pointer and wakes the
+ * requested number of waiters.
+ */
+__u32 __vdso_futex_robust_list64_try_unlock(__u32 *lock, __u32 tid, __u64 *pop);
+
+/**
+ * __vdso_futex_robust_list32_try_unlock - Try to unlock an uncontended robust futex
+ *					   with a 32-bit pending op pointer
+ * @lock:	Pointer to the futex lock object
+ * @tid:	The TID of the calling task
+ * @pop:	Pointer to the task's robust_list_head::list_pending_op
+ *
+ * Return: The content of *@lock. On success this is the same as @tid.
+ *
+ * Same as __vdso_futex_robust_list64_try_unlock() just with a 32-bit @pop pointer.
+ */
+__u32 __vdso_futex_robust_list32_try_unlock(__u32 *lock, __u32 tid, __u32 *pop);
+
+#endif