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, 96 insertions, 0 deletions

diff --git a/net/rxrpc/txbuf.c b/net/rxrpc/txbuf.c
new file mode 100644
index 000000000..55ef7a048
--- /dev/null
+++ b/net/rxrpc/txbuf.c
@@ -0,0 +1,96 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* RxRPC Tx data buffering.
+ *
+ * Copyright (C) 2022 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/slab.h>
+#include "ar-internal.h"
+
+static atomic_t rxrpc_txbuf_debug_ids;
+atomic_t rxrpc_nr_txbuf;
+
+/*
+ * Allocate and partially initialise a data transmission buffer.
+ */
+struct rxrpc_txbuf *rxrpc_alloc_data_txbuf(struct rxrpc_call *call, size_t data_size,
+					   size_t data_align, gfp_t gfp)
+{
+	struct rxrpc_txbuf *txb;
+	size_t total, doff, jsize = sizeof(struct rxrpc_jumbo_header);
+	void *buf;
+
+	txb = kzalloc_obj(*txb, gfp);
+	if (!txb)
+		return NULL;
+
+	/* We put a jumbo header in the buffer, but not a full wire header to
+	 * avoid delayed-corruption problems with zerocopy.
+	 */
+	doff = round_up(jsize, data_align);
+	total = doff + data_size;
+
+	data_align = umax(data_align, L1_CACHE_BYTES);
+	mutex_lock(&call->conn->tx_data_alloc_lock);
+	buf = page_frag_alloc_align(&call->conn->tx_data_alloc, total, gfp,
+				    data_align);
+	mutex_unlock(&call->conn->tx_data_alloc_lock);
+	if (!buf) {
+		kfree(txb);
+		return NULL;
+	}
+
+	refcount_set(&txb->ref, 1);
+	txb->call_debug_id	= call->debug_id;
+	txb->debug_id		= atomic_inc_return(&rxrpc_txbuf_debug_ids);
+	txb->alloc_size		= data_size;
+	txb->space		= data_size;
+	txb->offset		= 0;
+	txb->flags		= call->conn->out_clientflag;
+	txb->seq		= call->send_top + 1;
+	txb->data		= buf + doff;
+
+	trace_rxrpc_txbuf(txb->debug_id, txb->call_debug_id, txb->seq, 1,
+			  rxrpc_txbuf_alloc_data);
+
+	atomic_inc(&rxrpc_nr_txbuf);
+	return txb;
+}
+
+void rxrpc_see_txbuf(struct rxrpc_txbuf *txb, enum rxrpc_txbuf_trace what)
+{
+	int r = refcount_read(&txb->ref);
+
+	trace_rxrpc_txbuf(txb->debug_id, txb->call_debug_id, txb->seq, r, what);
+}
+
+static void rxrpc_free_txbuf(struct rxrpc_txbuf *txb)
+{
+	trace_rxrpc_txbuf(txb->debug_id, txb->call_debug_id, txb->seq, 0,
+			  rxrpc_txbuf_free);
+	if (txb->data)
+		page_frag_free(txb->data);
+	kfree(txb);
+	atomic_dec(&rxrpc_nr_txbuf);
+}
+
+void rxrpc_put_txbuf(struct rxrpc_txbuf *txb, enum rxrpc_txbuf_trace what)
+{
+	unsigned int debug_id, call_debug_id;
+	rxrpc_seq_t seq;
+	bool dead;
+	int r;
+
+	if (txb) {
+		debug_id = txb->debug_id;
+		call_debug_id = txb->call_debug_id;
+		seq = txb->seq;
+		dead = __refcount_dec_and_test(&txb->ref, &r);
+		trace_rxrpc_txbuf(debug_id, call_debug_id, seq, r - 1, what);
+		if (dead)
+			rxrpc_free_txbuf(txb);
+	}
+}