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, 152 insertions, 0 deletions

diff --git a/lib/crypto/mpi/mpiutil.c b/lib/crypto/mpi/mpiutil.c
new file mode 100644
index 000000000..f4faf7c90
--- /dev/null
+++ b/lib/crypto/mpi/mpiutil.c
@@ -0,0 +1,152 @@
+/* mpiutil.ac  -  Utility functions for MPI
+ * Copyright (C) 1998, 1999 Free Software Foundation, Inc.
+ *
+ * This file is part of GnuPG.
+ *
+ * GnuPG is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * GnuPG is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
+ */
+
+#include <linux/export.h>
+
+#include "mpi-internal.h"
+
+/****************
+ * Note:  It was a bad idea to use the number of limbs to allocate
+ *	  because on a alpha the limbs are large but we normally need
+ *	  integers of n bits - So we should change this to bits (or bytes).
+ *
+ *	  But mpi_alloc is used in a lot of places :-)
+ */
+MPI mpi_alloc(unsigned nlimbs)
+{
+	MPI a;
+
+	a = kmalloc_obj(*a);
+	if (!a)
+		return a;
+
+	if (nlimbs) {
+		a->d = mpi_alloc_limb_space(nlimbs);
+		if (!a->d) {
+			kfree(a);
+			return NULL;
+		}
+	} else {
+		a->d = NULL;
+	}
+
+	a->alloced = nlimbs;
+	a->nlimbs = 0;
+	a->sign = 0;
+	a->flags = 0;
+	a->nbits = 0;
+	return a;
+}
+EXPORT_SYMBOL_GPL(mpi_alloc);
+
+mpi_ptr_t mpi_alloc_limb_space(unsigned nlimbs)
+{
+	size_t len = nlimbs * sizeof(mpi_limb_t);
+
+	if (!len)
+		return NULL;
+
+	return kmalloc(len, GFP_KERNEL);
+}
+
+void mpi_free_limb_space(mpi_ptr_t a)
+{
+	if (!a)
+		return;
+
+	kfree_sensitive(a);
+}
+
+void mpi_assign_limb_space(MPI a, mpi_ptr_t ap, unsigned nlimbs)
+{
+	mpi_free_limb_space(a->d);
+	a->d = ap;
+	a->alloced = nlimbs;
+}
+
+/****************
+ * Resize the array of A to NLIMBS. the additional space is cleared
+ * (set to 0) [done by m_realloc()]
+ */
+int mpi_resize(MPI a, unsigned nlimbs)
+{
+	void *p;
+
+	if (nlimbs <= a->alloced)
+		return 0;	/* no need to do it */
+
+	if (a->d) {
+		p = kzalloc_objs(mpi_limb_t, nlimbs);
+		if (!p)
+			return -ENOMEM;
+		memcpy(p, a->d, a->alloced * sizeof(mpi_limb_t));
+		kfree_sensitive(a->d);
+		a->d = p;
+	} else {
+		a->d = kzalloc_objs(mpi_limb_t, nlimbs);
+		if (!a->d)
+			return -ENOMEM;
+	}
+	a->alloced = nlimbs;
+	return 0;
+}
+
+void mpi_free(MPI a)
+{
+	if (!a)
+		return;
+
+	if (a->flags & 4)
+		kfree_sensitive(a->d);
+	else
+		mpi_free_limb_space(a->d);
+
+	if (a->flags & ~7)
+		pr_info("invalid flag value in mpi\n");
+	kfree(a);
+}
+EXPORT_SYMBOL_GPL(mpi_free);
+
+/****************
+ * Note: This copy function should not interpret the MPI
+ *	 but copy it transparently.
+ */
+MPI mpi_copy(MPI a)
+{
+	int i;
+	MPI b;
+
+	if (a) {
+		b = mpi_alloc(a->nlimbs);
+		if (!b)
+			return NULL;
+		b->nlimbs = a->nlimbs;
+		b->sign = a->sign;
+		b->flags = a->flags;
+		b->flags &= ~(16|32); /* Reset the immutable and constant flags. */
+		for (i = 0; i < b->nlimbs; i++)
+			b->d[i] = a->d[i];
+	} else
+		b = NULL;
+	return b;
+}
+
+MODULE_DESCRIPTION("Multiprecision maths library");
+MODULE_LICENSE("GPL");