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, 105 insertions, 0 deletions

diff --git a/net/sctp/objcnt.c b/net/sctp/objcnt.c
new file mode 100644
index 000000000..0400c964e
--- /dev/null
+++ b/net/sctp/objcnt.c
@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* SCTP kernel implementation
+ * (C) Copyright IBM Corp. 2001, 2004
+ *
+ * This file is part of the SCTP kernel implementation
+ *
+ * Support for memory object debugging.  This allows one to monitor the
+ * object allocations/deallocations for types instrumented for this
+ * via the proc fs.
+ *
+ * Please send any bug reports or fixes you make to the
+ * email address(es):
+ *    lksctp developers <linux-sctp@vger.kernel.org>
+ *
+ * Written or modified by:
+ *    Jon Grimm             <jgrimm@us.ibm.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <net/sctp/sctp.h>
+
+/*
+ * Global counters to count raw object allocation counts.
+ * To add new counters, choose a unique suffix for the variable
+ * name as the helper macros key off this suffix to make
+ * life easier for the programmer.
+ */
+
+SCTP_DBG_OBJCNT(sock);
+SCTP_DBG_OBJCNT(ep);
+SCTP_DBG_OBJCNT(transport);
+SCTP_DBG_OBJCNT(assoc);
+SCTP_DBG_OBJCNT(bind_addr);
+SCTP_DBG_OBJCNT(bind_bucket);
+SCTP_DBG_OBJCNT(chunk);
+SCTP_DBG_OBJCNT(addr);
+SCTP_DBG_OBJCNT(datamsg);
+SCTP_DBG_OBJCNT(keys);
+
+/* An array to make it easy to pretty print the debug information
+ * to the proc fs.
+ */
+static struct sctp_dbg_objcnt_entry sctp_dbg_objcnt[] = {
+	SCTP_DBG_OBJCNT_ENTRY(sock),
+	SCTP_DBG_OBJCNT_ENTRY(ep),
+	SCTP_DBG_OBJCNT_ENTRY(assoc),
+	SCTP_DBG_OBJCNT_ENTRY(transport),
+	SCTP_DBG_OBJCNT_ENTRY(chunk),
+	SCTP_DBG_OBJCNT_ENTRY(bind_addr),
+	SCTP_DBG_OBJCNT_ENTRY(bind_bucket),
+	SCTP_DBG_OBJCNT_ENTRY(addr),
+	SCTP_DBG_OBJCNT_ENTRY(datamsg),
+	SCTP_DBG_OBJCNT_ENTRY(keys),
+};
+
+/* Callback from procfs to read out objcount information.
+ * Walk through the entries in the sctp_dbg_objcnt array, dumping
+ * the raw object counts for each monitored type.
+ */
+static int sctp_objcnt_seq_show(struct seq_file *seq, void *v)
+{
+	int i;
+
+	i = (int)*(loff_t *)v;
+	seq_setwidth(seq, 127);
+	seq_printf(seq, "%s: %d", sctp_dbg_objcnt[i].label,
+				atomic_read(sctp_dbg_objcnt[i].counter));
+	seq_pad(seq, '\n');
+	return 0;
+}
+
+static void *sctp_objcnt_seq_start(struct seq_file *seq, loff_t *pos)
+{
+	return (*pos >= ARRAY_SIZE(sctp_dbg_objcnt)) ? NULL : (void *)pos;
+}
+
+static void sctp_objcnt_seq_stop(struct seq_file *seq, void *v)
+{
+}
+
+static void *sctp_objcnt_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+	++*pos;
+	return (*pos >= ARRAY_SIZE(sctp_dbg_objcnt)) ? NULL : (void *)pos;
+}
+
+static const struct seq_operations sctp_objcnt_seq_ops = {
+	.start = sctp_objcnt_seq_start,
+	.next  = sctp_objcnt_seq_next,
+	.stop  = sctp_objcnt_seq_stop,
+	.show  = sctp_objcnt_seq_show,
+};
+
+/* Initialize the objcount in the proc filesystem.  */
+void sctp_dbg_objcnt_init(struct net *net)
+{
+	struct proc_dir_entry *ent;
+
+	ent = proc_create_seq("sctp_dbg_objcnt", 0,
+			  net->sctp.proc_net_sctp, &sctp_objcnt_seq_ops);
+	if (!ent)
+		pr_warn("sctp_dbg_objcnt: Unable to create /proc entry.\n");
+}