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, 175 insertions, 0 deletions

diff --git a/net/sunrpc/auth_tls.c b/net/sunrpc/auth_tls.c
new file mode 100644
index 000000000..87f570fd3
--- /dev/null
+++ b/net/sunrpc/auth_tls.c
@@ -0,0 +1,175 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2021, 2022 Oracle.  All rights reserved.
+ *
+ * The AUTH_TLS credential is used only to probe a remote peer
+ * for RPC-over-TLS support.
+ */
+
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/sunrpc/clnt.h>
+
+static const char *starttls_token = "STARTTLS";
+static const size_t starttls_len = 8;
+
+static struct rpc_auth tls_auth;
+static struct rpc_cred tls_cred;
+
+static void tls_encode_probe(struct rpc_rqst *rqstp, struct xdr_stream *xdr,
+			     const void *obj)
+{
+}
+
+static int tls_decode_probe(struct rpc_rqst *rqstp, struct xdr_stream *xdr,
+			    void *obj)
+{
+	return 0;
+}
+
+static const struct rpc_procinfo rpcproc_tls_probe = {
+	.p_encode	= tls_encode_probe,
+	.p_decode	= tls_decode_probe,
+};
+
+static void rpc_tls_probe_call_prepare(struct rpc_task *task, void *data)
+{
+	task->tk_flags &= ~RPC_TASK_NO_RETRANS_TIMEOUT;
+	rpc_call_start(task);
+}
+
+static void rpc_tls_probe_call_done(struct rpc_task *task, void *data)
+{
+}
+
+static const struct rpc_call_ops rpc_tls_probe_ops = {
+	.rpc_call_prepare	= rpc_tls_probe_call_prepare,
+	.rpc_call_done		= rpc_tls_probe_call_done,
+};
+
+static int tls_probe(struct rpc_clnt *clnt)
+{
+	struct rpc_message msg = {
+		.rpc_proc	= &rpcproc_tls_probe,
+	};
+	struct rpc_task_setup task_setup_data = {
+		.rpc_client	= clnt,
+		.rpc_message	= &msg,
+		.rpc_op_cred	= &tls_cred,
+		.callback_ops	= &rpc_tls_probe_ops,
+		.flags		= RPC_TASK_SOFT | RPC_TASK_SOFTCONN,
+	};
+	struct rpc_task	*task;
+	int status;
+
+	task = rpc_run_task(&task_setup_data);
+	if (IS_ERR(task))
+		return PTR_ERR(task);
+	status = task->tk_status;
+	rpc_put_task(task);
+	return status;
+}
+
+static struct rpc_auth *tls_create(const struct rpc_auth_create_args *args,
+				   struct rpc_clnt *clnt)
+{
+	refcount_inc(&tls_auth.au_count);
+	return &tls_auth;
+}
+
+static void tls_destroy(struct rpc_auth *auth)
+{
+}
+
+static struct rpc_cred *tls_lookup_cred(struct rpc_auth *auth,
+					struct auth_cred *acred, int flags)
+{
+	return get_rpccred(&tls_cred);
+}
+
+static void tls_destroy_cred(struct rpc_cred *cred)
+{
+}
+
+static int tls_match(struct auth_cred *acred, struct rpc_cred *cred, int taskflags)
+{
+	return 1;
+}
+
+static int tls_marshal(struct rpc_task *task, struct xdr_stream *xdr)
+{
+	__be32 *p;
+
+	p = xdr_reserve_space(xdr, 4 * XDR_UNIT);
+	if (!p)
+		return -EMSGSIZE;
+	/* Credential */
+	*p++ = rpc_auth_tls;
+	*p++ = xdr_zero;
+	/* Verifier */
+	*p++ = rpc_auth_null;
+	*p   = xdr_zero;
+	return 0;
+}
+
+static int tls_refresh(struct rpc_task *task)
+{
+	set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_rqstp->rq_cred->cr_flags);
+	return 0;
+}
+
+static int tls_validate(struct rpc_task *task, struct xdr_stream *xdr)
+{
+	__be32 *p;
+	void *str;
+
+	p = xdr_inline_decode(xdr, XDR_UNIT);
+	if (!p)
+		return -EIO;
+	if (*p != rpc_auth_null)
+		return -EIO;
+	if (xdr_stream_decode_opaque_inline(xdr, &str, starttls_len) != starttls_len)
+		return -EPROTONOSUPPORT;
+	if (memcmp(str, starttls_token, starttls_len))
+		return -EPROTONOSUPPORT;
+	return 0;
+}
+
+const struct rpc_authops authtls_ops = {
+	.owner		= THIS_MODULE,
+	.au_flavor	= RPC_AUTH_TLS,
+	.au_name	= "NULL",
+	.create		= tls_create,
+	.destroy	= tls_destroy,
+	.lookup_cred	= tls_lookup_cred,
+	.ping		= tls_probe,
+};
+
+static struct rpc_auth tls_auth = {
+	.au_cslack	= NUL_CALLSLACK,
+	.au_rslack	= NUL_REPLYSLACK,
+	.au_verfsize	= NUL_REPLYSLACK,
+	.au_ralign	= NUL_REPLYSLACK,
+	.au_ops		= &authtls_ops,
+	.au_flavor	= RPC_AUTH_TLS,
+	.au_count	= REFCOUNT_INIT(1),
+};
+
+static const struct rpc_credops tls_credops = {
+	.cr_name	= "AUTH_TLS",
+	.crdestroy	= tls_destroy_cred,
+	.crmatch	= tls_match,
+	.crmarshal	= tls_marshal,
+	.crwrap_req	= rpcauth_wrap_req_encode,
+	.crrefresh	= tls_refresh,
+	.crvalidate	= tls_validate,
+	.crunwrap_resp	= rpcauth_unwrap_resp_decode,
+};
+
+static struct rpc_cred tls_cred = {
+	.cr_lru		= LIST_HEAD_INIT(tls_cred.cr_lru),
+	.cr_auth	= &tls_auth,
+	.cr_ops		= &tls_credops,
+	.cr_count	= REFCOUNT_INIT(2),
+	.cr_flags	= 1UL << RPCAUTH_CRED_UPTODATE,
+};