]> asedeno.scripts.mit.edu Git - linux.git/commitdiff
net: stmmac: Send TSO packets always from Queue 0
authorJose Abreu <jose.abreu@synopsys.com>
Wed, 30 Jan 2019 14:54:20 +0000 (15:54 +0100)
committerDavid S. Miller <davem@davemloft.net>
Thu, 31 Jan 2019 06:24:49 +0000 (22:24 -0800)
The number of TSO enabled channels in HW can be different than the
number of total channels. There is no way to determined, at runtime, the
number of TSO capable channels and its safe to assume that if TSO is
enabled then at least channel 0 will be TSO capable.

Lets always send TSO packets from Queue 0.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c

index 5afba69981cfcf87b98bf5a98f3a85305e5cfd99..6656008068de36485a0f9819228ff561408b1fc2 100644 (file)
@@ -3025,8 +3025,17 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
 
        /* Manage oversized TCP frames for GMAC4 device */
        if (skb_is_gso(skb) && priv->tso) {
-               if (skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))
+               if (skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)) {
+                       /*
+                        * There is no way to determine the number of TSO
+                        * capable Queues. Let's use always the Queue 0
+                        * because if TSO is supported then at least this
+                        * one will be capable.
+                        */
+                       skb_set_queue_mapping(skb, 0);
+
                        return stmmac_tso_xmit(skb, dev);
+               }
        }
 
        if (unlikely(stmmac_tx_avail(priv, queue) < nfrags + 1)) {