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, 62 insertions, 0 deletions

diff --git a/rust/syn/thread.rs b/rust/syn/thread.rs
new file mode 100644
index 000000000..4165a173d
--- /dev/null
+++ b/rust/syn/thread.rs
@@ -0,0 +1,62 @@
+// SPDX-License-Identifier: Apache-2.0 OR MIT
+
+use std::fmt::{self, Debug};
+use std::thread::{self, ThreadId};
+
+/// ThreadBound is a Sync-maker and Send-maker that allows accessing a value
+/// of type T only from the original thread on which the ThreadBound was
+/// constructed.
+pub(crate) struct ThreadBound<T> {
+    value: T,
+    thread_id: ThreadId,
+}
+
+unsafe impl<T> Sync for ThreadBound<T> {}
+
+// Send bound requires Copy, as otherwise Drop could run in the wrong place.
+//
+// Today Copy and Drop are mutually exclusive so `T: Copy` implies `T: !Drop`.
+// This impl needs to be revisited if that restriction is relaxed in the future.
+unsafe impl<T: Copy> Send for ThreadBound<T> {}
+
+impl<T> ThreadBound<T> {
+    pub(crate) fn new(value: T) -> Self {
+        ThreadBound {
+            value,
+            thread_id: thread::current().id(),
+        }
+    }
+
+    pub(crate) fn get(&self) -> Option<&T> {
+        if thread::current().id() == self.thread_id {
+            Some(&self.value)
+        } else {
+            None
+        }
+    }
+}
+
+impl<T: Debug> Debug for ThreadBound<T> {
+    fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
+        match self.get() {
+            Some(value) => Debug::fmt(value, formatter),
+            None => formatter.write_str("unknown"),
+        }
+    }
+}
+
+// Copy the bytes of T, even if the currently running thread is the "wrong"
+// thread. This is fine as long as the original thread is not simultaneously
+// mutating this value via interior mutability, which would be a data race.
+//
+// Currently `T: Copy` is sufficient to guarantee that T contains no interior
+// mutability, because _all_ interior mutability in Rust is built on
+// std::cell::UnsafeCell, which has no Copy impl. This impl needs to be
+// revisited if that restriction is relaxed in the future.
+impl<T: Copy> Copy for ThreadBound<T> {}
+
+impl<T: Copy> Clone for ThreadBound<T> {
+    fn clone(&self) -> Self {
+        *self
+    }
+}