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, 114 insertions, 0 deletions

diff --git a/mm/msync.c b/mm/msync.c
new file mode 100644
index 000000000..90b491a27
--- /dev/null
+++ b/mm/msync.c
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *	linux/mm/msync.c
+ *
+ * Copyright (C) 1994-1999  Linus Torvalds
+ */
+
+/*
+ * The msync() system call.
+ */
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/mman.h>
+#include <linux/file.h>
+#include <linux/pagemap.h>
+#include <linux/syscalls.h>
+#include <linux/sched.h>
+
+/*
+ * MS_SYNC syncs the entire file - including mappings.
+ *
+ * MS_ASYNC does not start I/O (it used to, up to 2.5.67).
+ * Nor does it marks the relevant pages dirty (it used to up to 2.6.17).
+ * Now it doesn't do anything, since dirty pages are properly tracked.
+ *
+ * The application may now run fsync() to
+ * write out the dirty pages and wait on the writeout and check the result.
+ * Or the application may run fadvise(FADV_DONTNEED) against the fd to start
+ * async writeout immediately.
+ * So by _not_ starting I/O in MS_ASYNC we provide complete flexibility to
+ * applications.
+ */
+SYSCALL_DEFINE3(msync, unsigned long, start, size_t, len, int, flags)
+{
+	unsigned long end;
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+	int unmapped_error = 0;
+	int error = -EINVAL;
+
+	start = untagged_addr(start);
+
+	if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC))
+		goto out;
+	if (offset_in_page(start))
+		goto out;
+	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
+		goto out;
+	error = -ENOMEM;
+	len = (len + ~PAGE_MASK) & PAGE_MASK;
+	end = start + len;
+	if (end < start)
+		goto out;
+	error = 0;
+	if (end == start)
+		goto out;
+	/*
+	 * If the interval [start,end) covers some unmapped address ranges,
+	 * just ignore them, but return -ENOMEM at the end. Besides, if the
+	 * flag is MS_ASYNC (w/o MS_INVALIDATE) the result would be -ENOMEM
+	 * anyway and there is nothing left to do, so return immediately.
+	 */
+	mmap_read_lock(mm);
+	vma = find_vma(mm, start);
+	for (;;) {
+		struct file *file;
+		loff_t fstart, fend;
+
+		/* Still start < end. */
+		error = -ENOMEM;
+		if (!vma)
+			goto out_unlock;
+		/* Here start < vma->vm_end. */
+		if (start < vma->vm_start) {
+			if (flags == MS_ASYNC)
+				goto out_unlock;
+			start = vma->vm_start;
+			if (start >= end)
+				goto out_unlock;
+			unmapped_error = -ENOMEM;
+		}
+		/* Here vma->vm_start <= start < vma->vm_end. */
+		if ((flags & MS_INVALIDATE) &&
+				(vma->vm_flags & VM_LOCKED)) {
+			error = -EBUSY;
+			goto out_unlock;
+		}
+		file = vma->vm_file;
+		fstart = (loff_t)linear_page_index(vma, start) << PAGE_SHIFT;
+		fend = fstart + (min(end, vma->vm_end) - start) - 1;
+		start = vma->vm_end;
+		if ((flags & MS_SYNC) && file &&
+				(vma->vm_flags & VM_SHARED)) {
+			get_file(file);
+			mmap_read_unlock(mm);
+			error = vfs_fsync_range(file, fstart, fend, 1);
+			fput(file);
+			if (error || start >= end)
+				goto out;
+			mmap_read_lock(mm);
+			vma = find_vma(mm, start);
+		} else {
+			if (start >= end) {
+				error = 0;
+				goto out_unlock;
+			}
+			vma = find_vma(mm, vma->vm_end);
+		}
+	}
+out_unlock:
+	mmap_read_unlock(mm);
+out:
+	return error ? : unmapped_error;
+}