IB/ipoib: Add handling for sending of skb with many frags

IPoIB converts skb-fragments to sge adding 1 extra sge when SG is enabled.
Current codepath assumes that the max number of sge a device support
is at least MAX_SKB_FRAGS+1, there is no interaction with upper layers
to limit number of fragments in an skb if a device suports fewer
sges. The assumptions also lead to requesting a fixed number of sge
when IPoIB creates queue-pairs with SG enabled.

A fallback/slowpath is implemented using skb_linearize to
handle cases where the conversion would result in more sges than supported.

Signed-off-by: Hans Westgaard Ry <hans.westgaard.ry@oracle.com>
Reviewed-by: Håkon Bugge <haakon.bugge@oracle.com>
Reviewed-by: Wei Lin Guay <wei.lin.guay@oracle.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

4 files changed, 45 insertions, 3 deletions

diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h
index a6f3eab0f350..85be0de3ab26 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib.h
+++ b/drivers/infiniband/ulp/ipoib/ipoib.h
@@ -244,6 +244,7 @@ struct ipoib_cm_tx {
         unsigned             tx_tail;
         unsigned long        flags;
         u32                  mtu;
+        unsigned             max_send_sge;
 };
 
 struct ipoib_cm_rx_buf {
@@ -390,6 +391,7 @@ struct ipoib_dev_priv {
         int     hca_caps;
         struct ipoib_ethtool_st ethtool;
         struct timer_list poll_timer;
+        unsigned max_send_sge;
 };
 
 struct ipoib_ah {
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index 917e46ea3bf6..c8ed53562c9b 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -710,6 +710,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
         struct ipoib_dev_priv *priv = netdev_priv(dev);
         struct ipoib_tx_buf *tx_req;
         int rc;
+        unsigned usable_sge = tx->max_send_sge - !!skb_headlen(skb);
 
         if (unlikely(skb->len > tx->mtu)) {
                 ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n",
@@ -719,7 +720,23 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
                 ipoib_cm_skb_too_long(dev, skb, tx->mtu - IPOIB_ENCAP_LEN);
                 return;
         }
-
+        if (skb_shinfo(skb)->nr_frags > usable_sge) {
+                if (skb_linearize(skb) < 0) {
+                        ipoib_warn(priv, "skb could not be linearized\n");
+                        ++dev->stats.tx_dropped;
+                        ++dev->stats.tx_errors;
+                        dev_kfree_skb_any(skb);
+                        return;
+                }
+                /* Does skb_linearize return ok without reducing nr_frags? */
+                if (skb_shinfo(skb)->nr_frags > usable_sge) {
+                        ipoib_warn(priv, "too many frags after skb linearize\n");
+                        ++dev->stats.tx_dropped;
+                        ++dev->stats.tx_errors;
+                        dev_kfree_skb_any(skb);
+                        return;
+                }
+        }
         ipoib_dbg_data(priv, "sending packet: head 0x%x length %d connection 0x%x\n",
                        tx->tx_head, skb->len, tx->qp->qp_num);
 
@@ -1031,7 +1048,8 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_
         struct ib_qp *tx_qp;
 
         if (dev->features & NETIF_F_SG)
-                attr.cap.max_send_sge = MAX_SKB_FRAGS + 1;
+                attr.cap.max_send_sge =
+                        min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1);
 
         tx_qp = ib_create_qp(priv->pd, &attr);
         if (PTR_ERR(tx_qp) == -EINVAL) {
@@ -1040,6 +1058,7 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_
                 attr.create_flags &= ~IB_QP_CREATE_USE_GFP_NOIO;
                 tx_qp = ib_create_qp(priv->pd, &attr);
         }
+        tx->max_send_sge = attr.cap.max_send_sge;
         return tx_qp;
 }
 
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
index fa9c42ff1fb0..899e6b7fb8a5 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
@@ -538,6 +538,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
         struct ipoib_tx_buf *tx_req;
         int hlen, rc;
         void *phead;
+        unsigned usable_sge = priv->max_send_sge - !!skb_headlen(skb);
 
         if (skb_is_gso(skb)) {
                 hlen = skb_transport_offset(skb) + tcp_hdrlen(skb);
@@ -561,6 +562,23 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
                 phead = NULL;
                 hlen  = 0;
         }
+        if (skb_shinfo(skb)->nr_frags > usable_sge) {
+                if (skb_linearize(skb) < 0) {
+                        ipoib_warn(priv, "skb could not be linearized\n");
+                        ++dev->stats.tx_dropped;
+                        ++dev->stats.tx_errors;
+                        dev_kfree_skb_any(skb);
+                        return;
+                }
+                /* Does skb_linearize return ok without reducing nr_frags? */
+                if (skb_shinfo(skb)->nr_frags > usable_sge) {
+                        ipoib_warn(priv, "too many frags after skb linearize\n");
+                        ++dev->stats.tx_dropped;
+                        ++dev->stats.tx_errors;
+                        dev_kfree_skb_any(skb);
+                        return;
+                }
+        }
 
         ipoib_dbg_data(priv, "sending packet, length=%d address=%p qpn=0x%06x\n",
                        skb->len, address, qpn);
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c
index d48c5bae7877..b809c373e40e 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c
@@ -206,7 +206,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
                 init_attr.create_flags |= IB_QP_CREATE_NETIF_QP;
 
         if (dev->features & NETIF_F_SG)
-                init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1;
+                init_attr.cap.max_send_sge =
+                        min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1);
 
         priv->qp = ib_create_qp(priv->pd, &init_attr);
         if (IS_ERR(priv->qp)) {
@@ -233,6 +234,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
         priv->rx_wr.next = NULL;
         priv->rx_wr.sg_list = priv->rx_sge;
 
+        priv->max_send_sge = init_attr.cap.max_send_sge;
+
         return 0;
 
 out_free_send_cq: