Thread (23 messages) flat view 23 messages, 3 authors, 2021-02-17

Re: [PATCH net-next v4 8/8] net: dsa: felix: Add support for MRP

From: Vladimir Oltean <vladimir.oltean@nxp.com>
Date: 2021-02-17 10:26:03
Also in: bridge, lkml

On Tue, Feb 16, 2021 at 10:42:05PM +0100, Horatiu Vultur wrote:
quoted hunk ↗ jump to hunk
@@ -112,6 +113,7 @@ static struct sk_buff *ocelot_rcv(struct sk_buff *skb,
 	ocelot_xfh_get_qos_class(extraction, &qos_class);
 	ocelot_xfh_get_tag_type(extraction, &tag_type);
 	ocelot_xfh_get_vlan_tci(extraction, &vlan_tci);
+	ocelot_xfh_get_cpuq(extraction, &cpuq);
 
 	skb->dev = dsa_master_find_slave(netdev, 0, src_port);
 	if (!skb->dev)
@@ -126,6 +128,12 @@ static struct sk_buff *ocelot_rcv(struct sk_buff *skb,
 	skb->offload_fwd_mark = 1;
 	skb->priority = qos_class;
 
+#if IS_ENABLED(CONFIG_BRIDGE_MRP)
+	if (eth_hdr(skb)->h_proto == cpu_to_be16(ETH_P_MRP) &&
+	    cpuq & BIT(OCELOT_MRP_CPUQ))
Checking the EtherType seems redundant, since those are the only frames
trapped to the MRP CPU queue.

Also, the cpuq variable is potentially unused when CONFIG_BRIDGE_MRP is
unset. I'm concerned that static analysis people may come in and try to
fix it up with even more ifdeffery, which is definitely not what I would
like to go for.

How about just the following, which is not conditionally compiled:
	if (!(cpuq & BIT(OCELOT_MRP_CPUQ)))
		skb->offload_fwd_mark = 1;
+		skb->offload_fwd_mark = 0;
+#endif
+
 	/* Ocelot switches copy frames unmodified to the CPU. However, it is
 	 * possible for the user to request a VLAN modification through
 	 * VCAP_IS1_ACT_VID_REPLACE_ENA. In this case, what will happen is that
-- 
2.27.0
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help