Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB/core: Remove unused struct ib_device.flags member
  IB/core: Add IP checksum offload support
  IPoIB: Add send gather support
  IPoIB: Add high DMA feature flag
  IB/mlx4: Use multiple WQ blocks to post smaller send WQEs
  mlx4_core: Clean up struct mlx4_buf
  mlx4_core: For 64-bit systems, vmap() kernel queue buffers
  IB/mlx4: Consolidate code to get an entry from a struct mlx4_buf

drivers/infiniband/hw/mlx4/cq.c

@@ -64,13 +64,7 @@ static void mlx4_ib_cq_event(struct mlx4_cq *cq, enum mlx4_event type)
 
 static void *get_cqe_from_buf(struct mlx4_ib_cq_buf *buf, int n)
 {
-	int offset = n * sizeof (struct mlx4_cqe);
-
-	if (buf->buf.nbufs == 1)
-		return buf->buf.u.direct.buf + offset;
-	else
-		return buf->buf.u.page_list[offset >> PAGE_SHIFT].buf +
-			(offset & (PAGE_SIZE - 1));
+	return mlx4_buf_offset(&buf->buf, n * sizeof (struct mlx4_cqe));
 }
 
 static void *get_cqe(struct mlx4_ib_cq *cq, int n)
@@ -332,6 +326,12 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 	is_error = (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
 		MLX4_CQE_OPCODE_ERROR;
 
+	if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) == MLX4_OPCODE_NOP &&
+		     is_send)) {
+		printk(KERN_WARNING "Completion for NOP opcode detected!\n");
+		return -EINVAL;
+	}
+
 	if (!*cur_qp ||
 	    (be32_to_cpu(cqe->my_qpn) & 0xffffff) != (*cur_qp)->mqp.qpn) {
 		/*
@@ -354,8 +354,10 @@ static int mlx4_ib_poll_one(struct mlx4_ib_cq *cq,
 
 	if (is_send) {
 		wq = &(*cur_qp)->sq;
+		if (!(*cur_qp)->sq_signal_bits) {
 			wqe_ctr = be16_to_cpu(cqe->wqe_index);
 			wq->tail += (u16) (wqe_ctr - (u16) wq->tail);
+		}
 		wc->wr_id = wq->wrid[wq->tail & (wq->wqe_cnt - 1)];
 		++wq->tail;
 	} else if ((*cur_qp)->ibqp.srq) {

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -120,6 +120,8 @@ struct mlx4_ib_qp {
 
 	u32			doorbell_qpn;
 	__be32			sq_signal_bits;
+	unsigned		sq_next_wqe;
+	int			sq_max_wqes_per_wr;
 	int			sq_spare_wqes;
 	struct mlx4_ib_wq	sq;
 

drivers/infiniband/hw/mlx4/srq.c

@@ -38,13 +38,7 @@
 
 static void *get_wqe(struct mlx4_ib_srq *srq, int n)
 {
-	int offset = n << srq->msrq.wqe_shift;
-
-	if (srq->buf.nbufs == 1)
-		return srq->buf.u.direct.buf + offset;
-	else
-		return srq->buf.u.page_list[offset >> PAGE_SHIFT].buf +
-			(offset & (PAGE_SIZE - 1));
+	return mlx4_buf_offset(&srq->buf, n << srq->msrq.wqe_shift);
 }
 
 static void mlx4_ib_srq_event(struct mlx4_srq *srq, enum mlx4_event type)

drivers/infiniband/ulp/ipoib/ipoib.h

@@ -143,7 +143,7 @@ struct ipoib_rx_buf {
 
 struct ipoib_tx_buf {
 	struct sk_buff *skb;
-	u64		mapping;
+	u64		mapping[MAX_SKB_FRAGS + 1];
 };
 
 struct ib_cm_id;
@@ -296,7 +296,7 @@ struct ipoib_dev_priv {
 	struct ipoib_tx_buf *tx_ring;
 	unsigned	     tx_head;
 	unsigned	     tx_tail;
-	struct ib_sge	     tx_sge;
+	struct ib_sge	     tx_sge[MAX_SKB_FRAGS + 1];
 	struct ib_send_wr    tx_wr;
 	unsigned	     tx_outstanding;
 

drivers/infiniband/ulp/ipoib/ipoib_cm.c

@@ -634,8 +634,8 @@ static inline int post_send(struct ipoib_dev_priv *priv,
 {
 	struct ib_send_wr *bad_wr;
 
-	priv->tx_sge.addr	= addr;
-	priv->tx_sge.length	= len;
+	priv->tx_sge[0].addr          = addr;
+	priv->tx_sge[0].length        = len;
 
 	priv->tx_wr.wr_id	= wr_id | IPOIB_OP_CM;
 
@@ -676,7 +676,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
 		return;
 	}
 
-	tx_req->mapping = addr;
+	tx_req->mapping[0] = addr;
 
 	if (unlikely(post_send(priv, tx, tx->tx_head & (ipoib_sendq_size - 1),
 			       addr, skb->len))) {
@@ -715,7 +715,7 @@ void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 
 	tx_req = &tx->tx_ring[wr_id];
 
-	ib_dma_unmap_single(priv->ca, tx_req->mapping, tx_req->skb->len, DMA_TO_DEVICE);
+	ib_dma_unmap_single(priv->ca, tx_req->mapping[0], tx_req->skb->len, DMA_TO_DEVICE);
 
 	/* FIXME: is this right? Shouldn't we only increment on success? */
 	++dev->stats.tx_packets;
@@ -1110,7 +1110,7 @@ static void ipoib_cm_tx_destroy(struct ipoib_cm_tx *p)
 
 	while ((int) p->tx_tail - (int) p->tx_head < 0) {
 		tx_req = &p->tx_ring[p->tx_tail & (ipoib_sendq_size - 1)];
-		ib_dma_unmap_single(priv->ca, tx_req->mapping, tx_req->skb->len,
+		ib_dma_unmap_single(priv->ca, tx_req->mapping[0], tx_req->skb->len,
 				    DMA_TO_DEVICE);
 		dev_kfree_skb_any(tx_req->skb);
 		++p->tx_tail;

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -239,6 +239,54 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 			   "for buf %d\n", wr_id);
 }
 
+static int ipoib_dma_map_tx(struct ib_device *ca,
+			    struct ipoib_tx_buf *tx_req)
+{
+	struct sk_buff *skb = tx_req->skb;
+	u64 *mapping = tx_req->mapping;
+	int i;
+
+	mapping[0] = ib_dma_map_single(ca, skb->data, skb_headlen(skb),
+				       DMA_TO_DEVICE);
+	if (unlikely(ib_dma_mapping_error(ca, mapping[0])))
+		return -EIO;
+
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; ++i) {
+		skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+		mapping[i + 1] = ib_dma_map_page(ca, frag->page,
+						 frag->page_offset, frag->size,
+						 DMA_TO_DEVICE);
+		if (unlikely(ib_dma_mapping_error(ca, mapping[i + 1])))
+			goto partial_error;
+	}
+	return 0;
+
+partial_error:
+	ib_dma_unmap_single(ca, mapping[0], skb_headlen(skb), DMA_TO_DEVICE);
+
+	for (; i > 0; --i) {
+		skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
+		ib_dma_unmap_page(ca, mapping[i], frag->size, DMA_TO_DEVICE);
+	}
+	return -EIO;
+}
+
+static void ipoib_dma_unmap_tx(struct ib_device *ca,
+			       struct ipoib_tx_buf *tx_req)
+{
+	struct sk_buff *skb = tx_req->skb;
+	u64 *mapping = tx_req->mapping;
+	int i;
+
+	ib_dma_unmap_single(ca, mapping[0], skb_headlen(skb), DMA_TO_DEVICE);
+
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; ++i) {
+		skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+		ib_dma_unmap_page(ca, mapping[i + 1], frag->size,
+				  DMA_TO_DEVICE);
+	}
+}
+
 static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
@@ -257,8 +305,7 @@ static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 
 	tx_req = &priv->tx_ring[wr_id];
 
-	ib_dma_unmap_single(priv->ca, tx_req->mapping,
-			    tx_req->skb->len, DMA_TO_DEVICE);
+	ipoib_dma_unmap_tx(priv->ca, tx_req);
 
 	++dev->stats.tx_packets;
 	dev->stats.tx_bytes += tx_req->skb->len;
@@ -341,13 +388,20 @@ void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr)
 static inline int post_send(struct ipoib_dev_priv *priv,
 			    unsigned int wr_id,
 			    struct ib_ah *address, u32 qpn,
-			    u64 addr, int len)
+			    u64 *mapping, int headlen,
+			    skb_frag_t *frags,
+			    int nr_frags)
 {
 	struct ib_send_wr *bad_wr;
+	int i;
 
-	priv->tx_sge.addr	      = addr;
-	priv->tx_sge.length	      = len;
-
+	priv->tx_sge[0].addr	     = mapping[0];
+	priv->tx_sge[0].length       = headlen;
+	for (i = 0; i < nr_frags; ++i) {
+		priv->tx_sge[i + 1].addr = mapping[i + 1];
+		priv->tx_sge[i + 1].length = frags[i].size;
+	}
+	priv->tx_wr.num_sge	     = nr_frags + 1;
 	priv->tx_wr.wr_id 	     = wr_id;
 	priv->tx_wr.wr.ud.remote_qpn = qpn;
 	priv->tx_wr.wr.ud.ah 	     = address;
@@ -360,7 +414,6 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_tx_buf *tx_req;
-	u64 addr;
 
 	if (unlikely(skb->len > priv->mcast_mtu + IPOIB_ENCAP_LEN)) {
 		ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n",
@@ -383,20 +436,19 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 	 */
 	tx_req = &priv->tx_ring[priv->tx_head & (ipoib_sendq_size - 1)];
 	tx_req->skb = skb;
-	addr = ib_dma_map_single(priv->ca, skb->data, skb->len,
-				 DMA_TO_DEVICE);
-	if (unlikely(ib_dma_mapping_error(priv->ca, addr))) {
+	if (unlikely(ipoib_dma_map_tx(priv->ca, tx_req))) {
 		++dev->stats.tx_errors;
 		dev_kfree_skb_any(skb);
 		return;
 	}
-	tx_req->mapping = addr;
 
 	if (unlikely(post_send(priv, priv->tx_head & (ipoib_sendq_size - 1),
-			       address->ah, qpn, addr, skb->len))) {
+			       address->ah, qpn,
+			       tx_req->mapping, skb_headlen(skb),
+			       skb_shinfo(skb)->frags, skb_shinfo(skb)->nr_frags))) {
 		ipoib_warn(priv, "post_send failed\n");
 		++dev->stats.tx_errors;
-		ib_dma_unmap_single(priv->ca, addr, skb->len, DMA_TO_DEVICE);
+		ipoib_dma_unmap_tx(priv->ca, tx_req);
 		dev_kfree_skb_any(skb);
 	} else {
 		dev->trans_start = jiffies;
@@ -615,10 +667,7 @@ int ipoib_ib_dev_stop(struct net_device *dev, int flush)
 			while ((int) priv->tx_tail - (int) priv->tx_head < 0) {
 				tx_req = &priv->tx_ring[priv->tx_tail &
 							(ipoib_sendq_size - 1)];
-				ib_dma_unmap_single(priv->ca,
-						    tx_req->mapping,
-						    tx_req->skb->len,
-						    DMA_TO_DEVICE);
+				ipoib_dma_unmap_tx(priv->ca, tx_req);
 				dev_kfree_skb_any(tx_req->skb);
 				++priv->tx_tail;
 				--priv->tx_outstanding;

drivers/infiniband/ulp/ipoib/ipoib_main.c

@@ -965,7 +965,9 @@ static void ipoib_setup(struct net_device *dev)
 	dev->addr_len		 = INFINIBAND_ALEN;
 	dev->type		 = ARPHRD_INFINIBAND;
 	dev->tx_queue_len	 = ipoib_sendq_size * 2;
-	dev->features		 = NETIF_F_VLAN_CHALLENGED | NETIF_F_LLTX;
+	dev->features		 = (NETIF_F_VLAN_CHALLENGED	|
+				    NETIF_F_LLTX		|
+				    NETIF_F_HIGHDMA);
 
 	/* MTU will be reset when mcast join happens */
 	dev->mtu		 = IPOIB_PACKET_SIZE - IPOIB_ENCAP_LEN;

drivers/infiniband/ulp/ipoib/ipoib_verbs.c

@@ -157,6 +157,7 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	};
 
 	int ret, size;
+	int i;
 
 	priv->pd = ib_alloc_pd(priv->ca);
 	if (IS_ERR(priv->pd)) {
@@ -191,6 +192,9 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	init_attr.send_cq = priv->cq;
 	init_attr.recv_cq = priv->cq;
 
+	if (dev->features & NETIF_F_SG)
+		init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1;
+
 	priv->qp = ib_create_qp(priv->pd, &init_attr);
 	if (IS_ERR(priv->qp)) {
 		printk(KERN_WARNING "%s: failed to create QP\n", ca->name);
@@ -201,11 +205,11 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	priv->dev->dev_addr[2] = (priv->qp->qp_num >>  8) & 0xff;
 	priv->dev->dev_addr[3] = (priv->qp->qp_num      ) & 0xff;
 
-	priv->tx_sge.lkey	= priv->mr->lkey;
+	for (i = 0; i < MAX_SKB_FRAGS + 1; ++i)
+		priv->tx_sge[i].lkey = priv->mr->lkey;
 
 	priv->tx_wr.opcode	= IB_WR_SEND;
-	priv->tx_wr.sg_list	= &priv->tx_sge;
-	priv->tx_wr.num_sge	= 1;
+	priv->tx_wr.sg_list	= priv->tx_sge;
 	priv->tx_wr.send_flags	= IB_SEND_SIGNALED;
 
 	return 0;

drivers/net/mlx4/alloc.c

@@ -116,40 +116,53 @@ int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
 		buf->nbufs        = 1;
 		buf->npages       = 1;
 		buf->page_shift   = get_order(size) + PAGE_SHIFT;
-		buf->u.direct.buf = dma_alloc_coherent(&dev->pdev->dev,
+		buf->direct.buf   = dma_alloc_coherent(&dev->pdev->dev,
 						       size, &t, GFP_KERNEL);
-		if (!buf->u.direct.buf)
+		if (!buf->direct.buf)
 			return -ENOMEM;
 
-		buf->u.direct.map = t;
+		buf->direct.map = t;
 
 		while (t & ((1 << buf->page_shift) - 1)) {
 			--buf->page_shift;
 			buf->npages *= 2;
 		}
 
-		memset(buf->u.direct.buf, 0, size);
+		memset(buf->direct.buf, 0, size);
 	} else {
 		int i;
 
 		buf->nbufs       = (size + PAGE_SIZE - 1) / PAGE_SIZE;
 		buf->npages      = buf->nbufs;
 		buf->page_shift  = PAGE_SHIFT;
-		buf->u.page_list = kzalloc(buf->nbufs * sizeof *buf->u.page_list,
+		buf->page_list   = kzalloc(buf->nbufs * sizeof *buf->page_list,
 					   GFP_KERNEL);
-		if (!buf->u.page_list)
+		if (!buf->page_list)
 			return -ENOMEM;
 
 		for (i = 0; i < buf->nbufs; ++i) {
-			buf->u.page_list[i].buf =
+			buf->page_list[i].buf =
 				dma_alloc_coherent(&dev->pdev->dev, PAGE_SIZE,
 						   &t, GFP_KERNEL);
-			if (!buf->u.page_list[i].buf)
+			if (!buf->page_list[i].buf)
 				goto err_free;
 
-			buf->u.page_list[i].map = t;
+			buf->page_list[i].map = t;
 
-			memset(buf->u.page_list[i].buf, 0, PAGE_SIZE);
+			memset(buf->page_list[i].buf, 0, PAGE_SIZE);
+		}
+
+		if (BITS_PER_LONG == 64) {
+			struct page **pages;
+			pages = kmalloc(sizeof *pages * buf->nbufs, GFP_KERNEL);
+			if (!pages)
+				goto err_free;
+			for (i = 0; i < buf->nbufs; ++i)
+				pages[i] = virt_to_page(buf->page_list[i].buf);
+			buf->direct.buf = vmap(pages, buf->nbufs, VM_MAP, PAGE_KERNEL);
+			kfree(pages);
+			if (!buf->direct.buf)
+				goto err_free;
 		}
 	}
 
@@ -167,15 +180,18 @@ void mlx4_buf_free(struct mlx4_dev *dev, int size, struct mlx4_buf *buf)
 	int i;
 
 	if (buf->nbufs == 1)
-		dma_free_coherent(&dev->pdev->dev, size, buf->u.direct.buf,
-				  buf->u.direct.map);
+		dma_free_coherent(&dev->pdev->dev, size, buf->direct.buf,
+				  buf->direct.map);
 	else {
+		if (BITS_PER_LONG == 64)
+			vunmap(buf->direct.buf);
+
 		for (i = 0; i < buf->nbufs; ++i)
-			if (buf->u.page_list[i].buf)
+			if (buf->page_list[i].buf)
 				dma_free_coherent(&dev->pdev->dev, PAGE_SIZE,
-						  buf->u.page_list[i].buf,
-						  buf->u.page_list[i].map);
-		kfree(buf->u.page_list);
+						  buf->page_list[i].buf,
+						  buf->page_list[i].map);
+		kfree(buf->page_list);
 	}
 }
 EXPORT_SYMBOL_GPL(mlx4_buf_free);

drivers/net/mlx4/mr.c

@@ -419,9 +419,9 @@ int mlx4_buf_write_mtt(struct mlx4_dev *dev, struct mlx4_mtt *mtt,
 
 	for (i = 0; i < buf->npages; ++i)
 		if (buf->nbufs == 1)
-			page_list[i] = buf->u.direct.map + (i << buf->page_shift);
+			page_list[i] = buf->direct.map + (i << buf->page_shift);
 		else
-			page_list[i] = buf->u.page_list[i].map;
+			page_list[i] = buf->page_list[i].map;
 
 	err = mlx4_write_mtt(dev, mtt, 0, buf->npages, page_list);
 

include/linux/mlx4/device.h

@@ -133,6 +133,11 @@ enum {
 	MLX4_STAT_RATE_OFFSET	= 5
 };
 
+static inline u64 mlx4_fw_ver(u64 major, u64 minor, u64 subminor)
+{
+	return (major << 32) | (minor << 16) | subminor;
+}
+
 struct mlx4_caps {
 	u64			fw_ver;
 	int			num_ports;
@@ -189,10 +194,8 @@ struct mlx4_buf_list {
 };
 
 struct mlx4_buf {
-	union {
 	struct mlx4_buf_list	direct;
 	struct mlx4_buf_list   *page_list;
-	} u;
 	int			nbufs;
 	int			npages;
 	int			page_shift;
@@ -308,6 +311,14 @@ struct mlx4_init_port_param {
 int mlx4_buf_alloc(struct mlx4_dev *dev, int size, int max_direct,
 		   struct mlx4_buf *buf);
 void mlx4_buf_free(struct mlx4_dev *dev, int size, struct mlx4_buf *buf);
+static inline void *mlx4_buf_offset(struct mlx4_buf *buf, int offset)
+{
+	if (BITS_PER_LONG == 64 || buf->nbufs == 1)
+		return buf->direct.buf + offset;
+	else
+		return buf->page_list[offset >> PAGE_SHIFT].buf +
+			(offset & (PAGE_SIZE - 1));
+}
 
 int mlx4_pd_alloc(struct mlx4_dev *dev, u32 *pdn);
 void mlx4_pd_free(struct mlx4_dev *dev, u32 pdn);

include/linux/mlx4/qp.h

@@ -154,7 +154,11 @@ struct mlx4_qp_context {
 	u32			reserved5[10];
 };
 
+/* Which firmware version adds support for NEC (NoErrorCompletion) bit */
+#define MLX4_FW_VER_WQE_CTRL_NEC mlx4_fw_ver(2, 2, 232)
+
 enum {
+	MLX4_WQE_CTRL_NEC	= 1 << 29,
 	MLX4_WQE_CTRL_FENCE	= 1 << 6,
 	MLX4_WQE_CTRL_CQ_UPDATE	= 3 << 2,
 	MLX4_WQE_CTRL_SOLICITED	= 1 << 1,

include/rdma/ib_verbs.h

@@ -95,7 +95,15 @@ enum ib_device_cap_flags {
 	IB_DEVICE_N_NOTIFY_CQ		= (1<<14),
 	IB_DEVICE_ZERO_STAG		= (1<<15),
 	IB_DEVICE_SEND_W_INV		= (1<<16),
-	IB_DEVICE_MEM_WINDOW		= (1<<17)
+	IB_DEVICE_MEM_WINDOW		= (1<<17),
+	/*
+	 * Devices should set IB_DEVICE_UD_IP_SUM if they support
+	 * insertion of UDP and TCP checksum on outgoing UD IPoIB
+	 * messages and can verify the validity of checksum for
+	 * incoming messages.  Setting this flag implies that the
+	 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
+	 */
+	IB_DEVICE_UD_IP_CSUM		= (1<<18),
 };
 
 enum ib_atomic_cap {
@@ -431,6 +439,7 @@ struct ib_wc {
 	u8			sl;
 	u8			dlid_path_bits;
 	u8			port_num;	/* valid only for DR SMPs on switches */
+	int			csum_ok;
 };
 
 enum ib_cq_notify_flags {
@@ -615,7 +624,8 @@ enum ib_send_flags {
 	IB_SEND_FENCE		= 1,
 	IB_SEND_SIGNALED	= (1<<1),
 	IB_SEND_SOLICITED	= (1<<2),
-	IB_SEND_INLINE		= (1<<3)
+	IB_SEND_INLINE		= (1<<3),
+	IB_SEND_IP_CSUM		= (1<<4)
 };
 
 struct ib_sge {
@@ -890,8 +900,6 @@ struct ib_device {
 	int                          *pkey_tbl_len;
 	int                          *gid_tbl_len;
 
-	u32                           flags;
-
 	int			      num_comp_vectors;
 
 	struct iw_cm_verbs	     *iwcm;