qeth: enable scatter/gather in layer 2 mode
authorEugene Crosser <Eugene.Crosser@ru.ibm.com>
Thu, 16 Jun 2016 14:18:55 +0000 (16:18 +0200)
committerDavid S. Miller <davem@davemloft.net>
Fri, 17 Jun 2016 05:16:12 +0000 (22:16 -0700)
The patch enables NETIF_F_SG flag for OSA in layer 2 mode.
It also adds performance accounting for fragmented sends,
adds a conditional skb_linearize() attempt if the skb had
too many fragments for QDIO SBAL, and fills netdevice->gso_*
attributes.

Signed-off-by: Eugene Crosser <Eugene.Crosser@ru.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Reviewed-by: Lakhvich Dmitriy <ldmitriy@ru.ibm.com>
Reviewed-by: Thomas Richter <tmricht@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/s390/net/qeth_l2_main.c

index faab0b669ca46a1debd25f5fe18121b69ef6fa8b..ec163e4ededb064e15e6fba237b138248ec3f122 100644 (file)
@@ -869,6 +869,7 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
        int data_offset = -1;
        int elements_needed = 0;
        int hd_len = 0;
+       int nr_frags;
 
        if (card->qdio.do_prio_queueing || (cast_type &&
                                        card->info.is_multicast_different))
@@ -892,6 +893,17 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
        }
        netif_stop_queue(dev);
 
+       /* fix hardware limitation: as long as we do not have sbal
+        * chaining we can not send long frag lists
+        */
+       if ((card->info.type != QETH_CARD_TYPE_IQD) &&
+           !qeth_get_elements_no(card, new_skb, 0)) {
+               if (skb_linearize(new_skb))
+                       goto tx_drop;
+               if (card->options.performance_stats)
+                       card->perf_stats.tx_lin++;
+       }
+
        if (card->info.type == QETH_CARD_TYPE_OSN)
                hdr = (struct qeth_hdr *)skb->data;
        else {
@@ -943,6 +955,14 @@ static int qeth_l2_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
        if (!rc) {
                card->stats.tx_packets++;
                card->stats.tx_bytes += tx_bytes;
+               if (card->options.performance_stats) {
+                       nr_frags = skb_shinfo(new_skb)->nr_frags;
+                       if (nr_frags) {
+                               card->perf_stats.sg_skbs_sent++;
+                               /* nr_frags + skb->data */
+                               card->perf_stats.sg_frags_sent += nr_frags + 1;
+                       }
+               }
                if (new_skb != skb)
                        dev_kfree_skb_any(skb);
                rc = NETDEV_TX_OK;
@@ -1118,12 +1138,16 @@ static int qeth_l2_setup_netdev(struct qeth_card *card)
                &qeth_l2_ethtool_ops : &qeth_l2_osn_ops;
        card->dev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
        if (card->info.type == QETH_CARD_TYPE_OSD && !card->info.guestlan) {
-               card->dev->hw_features = NETIF_F_IP_CSUM | NETIF_F_RXCSUM;
+               card->dev->hw_features = NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
+                                        NETIF_F_SG;
                /* Turn on RX offloading per default */
                card->dev->features |= NETIF_F_RXCSUM;
        }
        card->info.broadcast_capable = 1;
        qeth_l2_request_initial_mac(card);
+       card->dev->gso_max_size = (QETH_MAX_BUFFER_ELEMENTS(card) - 1) *
+                                 PAGE_SIZE;
+       card->dev->gso_max_segs = (QETH_MAX_BUFFER_ELEMENTS(card) - 1);
        SET_NETDEV_DEV(card->dev, &card->gdev->dev);
        netif_napi_add(card->dev, &card->napi, qeth_l2_poll, QETH_NAPI_WEIGHT);
        netif_carrier_off(card->dev);