Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Maxime Chevallier <maxime.chevallier@bootlin.com>
To:     davem@davemloft.net
Cc:     Maxime Chevallier <maxime.chevallier@bootlin.com>,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        thomas.petazzoni@bootlin.com, gregory.clement@bootlin.com
Subject: [PATCH net-next 1/2] net: mvneta: Remove per-cpu queue mapping for Armada 3700
Date:   Fri, 12 Feb 2021 16:12:19 +0100
Message-Id: <20210212151220.84106-2-maxime.chevallier@bootlin.com>
In-Reply-To: <20210212151220.84106-1-maxime.chevallier@bootlin.com>
References: <20210212151220.84106-1-maxime.chevallier@bootlin.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

According to Errata #23 "The per-CPU GbE interrupt is limited to Core
0", we can't use the per-cpu interrupt mechanism on the Armada 3700
familly.

This is correctly checked for RSS configuration, but the initial queue
mapping is still done by having the queues spread across all the CPUs in
the system, both in the init path and in the cpu_hotplug path.

Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
---
 drivers/net/ethernet/marvell/mvneta.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 6290bfb6494e..8e410fafff8d 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -3428,7 +3428,9 @@ static int mvneta_txq_sw_init(struct mvneta_port *pp,
 		return -ENOMEM;
 
 	/* Setup XPS mapping */
-	if (txq_number > 1)
+	if (pp->neta_armada3700)
+		cpu = 0;
+	else if (txq_number > 1)
 		cpu = txq->id % num_present_cpus();
 	else
 		cpu = pp->rxq_def % num_present_cpus();
@@ -4206,6 +4208,11 @@ static int mvneta_cpu_online(unsigned int cpu, struct hlist_node *node)
 						  node_online);
 	struct mvneta_pcpu_port *port = per_cpu_ptr(pp->ports, cpu);
 
+	/* Armada 3700's per-cpu interrupt for mvneta is broken, all interrupts
+	 * are routed to CPU 0, so we don't need all the cpu-hotplug support
+	 */
+	if (pp->neta_armada3700)
+		return 0;
 
 	spin_lock(&pp->lock);
 	/*
-- 
2.25.4