From eb4a7cbf27082bea34764bab3bc85595683f967b Mon Sep 17 00:00:00 2001
From: "John L. Burr" <jlburr@cadence.com>
Date: Tue, 11 Jan 2011 20:39:46 -0800
Subject: IB/mthca: Fix driver when sizeof (phys_addr_t) > sizeof (long)

Some systems have PCI addresses that don't fit in unsigned long (eg some
32-bit PowerPC 440 systems have 36-bit bus addresses).  Fix up the driver
by using phys_addr_t where appropriate, so we don't truncate any PCI
resource addresses before ioremapping them.

Signed-off-by: John L. Burr <jlburr@cadence.com>

[ Update to apply to current driver source.  - Roland ]

Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/mthca/mthca_catas.c b/drivers/infiniband/hw/mthca/mthca_catas.c
index 0aa0110..e4a08c2 100644
--- a/drivers/infiniband/hw/mthca/mthca_catas.c
+++ b/drivers/infiniband/hw/mthca/mthca_catas.c
@@ -146,7 +146,7 @@ static void poll_catas(unsigned long dev_ptr)
 
 void mthca_start_catas_poll(struct mthca_dev *dev)
 {
-	unsigned long addr;
+	phys_addr_t addr;
 
 	init_timer(&dev->catas_err.timer);
 	dev->catas_err.map  = NULL;
@@ -158,7 +158,8 @@ void mthca_start_catas_poll(struct mthca_dev *dev)
 	dev->catas_err.map = ioremap(addr, dev->catas_err.size * 4);
 	if (!dev->catas_err.map) {
 		mthca_warn(dev, "couldn't map catastrophic error region "
-			   "at 0x%lx/0x%x\n", addr, dev->catas_err.size * 4);
+			   "at 0x%llx/0x%x\n", (unsigned long long) addr,
+			   dev->catas_err.size * 4);
 		return;
 	}
 
diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c
index f4ceecd..7bfa2a1 100644
--- a/drivers/infiniband/hw/mthca/mthca_cmd.c
+++ b/drivers/infiniband/hw/mthca/mthca_cmd.c
@@ -713,7 +713,7 @@ int mthca_RUN_FW(struct mthca_dev *dev, u8 *status)
 
 static void mthca_setup_cmd_doorbells(struct mthca_dev *dev, u64 base)
 {
-	unsigned long addr;
+	phys_addr_t addr;
 	u16 max_off = 0;
 	int i;
 
diff --git a/drivers/infiniband/hw/mthca/mthca_eq.c b/drivers/infiniband/hw/mthca/mthca_eq.c
index 8e8c728..76785c6 100644
--- a/drivers/infiniband/hw/mthca/mthca_eq.c
+++ b/drivers/infiniband/hw/mthca/mthca_eq.c
@@ -653,7 +653,7 @@ static int mthca_map_reg(struct mthca_dev *dev,
 			 unsigned long offset, unsigned long size,
 			 void __iomem **map)
 {
-	unsigned long base = pci_resource_start(dev->pdev, 0);
+	phys_addr_t base = pci_resource_start(dev->pdev, 0);
 
 	*map = ioremap(base + offset, size);
 	if (!*map)
diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c
index 5eee666..8a40cd5 100644
--- a/drivers/infiniband/hw/mthca/mthca_main.c
+++ b/drivers/infiniband/hw/mthca/mthca_main.c
@@ -790,7 +790,7 @@ static int mthca_setup_hca(struct mthca_dev *dev)
 		goto err_uar_table_free;
 	}
 
-	dev->kar = ioremap(dev->driver_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
+	dev->kar = ioremap((phys_addr_t) dev->driver_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
 	if (!dev->kar) {
 		mthca_err(dev, "Couldn't map kernel access region, "
 			  "aborting.\n");
diff --git a/drivers/infiniband/hw/mthca/mthca_mr.c b/drivers/infiniband/hw/mthca/mthca_mr.c
index 065b208..44045c8 100644
--- a/drivers/infiniband/hw/mthca/mthca_mr.c
+++ b/drivers/infiniband/hw/mthca/mthca_mr.c
@@ -853,7 +853,7 @@ void mthca_arbel_fmr_unmap(struct mthca_dev *dev, struct mthca_fmr *fmr)
 
 int mthca_init_mr_table(struct mthca_dev *dev)
 {
-	unsigned long addr;
+	phys_addr_t addr;
 	int mpts, mtts, err, i;
 
 	err = mthca_alloc_init(&dev->mr_table.mpt_alloc,
-- 
cgit v0.10.2


From 4979d18fe105297f8f065743f31f8f735da8df2d Mon Sep 17 00:00:00 2001
From: Roland Dreier <rolandd@cisco.com>
Date: Wed, 12 Jan 2011 09:50:36 -0800
Subject: mlx4_{core, ib, en}: Fix driver when sizeof (phys_addr_t) > sizeof
 (long)

Some systems have PCI addresses that don't fit in unsigned long (eg some
32-bit PowerPC 440 systems have 36-bit bus addresses).  Fix up mlx4 drivers
by using phys_addr_t where appropriate, so we don't truncate any PCI
resource addresses before ioremapping them.

Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 4c85224..d68d849 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -1005,7 +1005,8 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	if (mlx4_uar_alloc(dev, &ibdev->priv_uar))
 		goto err_pd;
 
-	ibdev->uar_map = ioremap(ibdev->priv_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
+	ibdev->uar_map = ioremap((phys_addr_t) ibdev->priv_uar.pfn << PAGE_SHIFT,
+				 PAGE_SIZE);
 	if (!ibdev->uar_map)
 		goto err_uar;
 	MLX4_INIT_DOORBELL_LOCK(&ibdev->uar_lock);
diff --git a/drivers/net/mlx4/catas.c b/drivers/net/mlx4/catas.c
index 68aaa42..32f9471 100644
--- a/drivers/net/mlx4/catas.c
+++ b/drivers/net/mlx4/catas.c
@@ -113,7 +113,7 @@ static void catas_reset(struct work_struct *work)
 void mlx4_start_catas_poll(struct mlx4_dev *dev)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
-	unsigned long addr;
+	phys_addr_t addr;
 
 	INIT_LIST_HEAD(&priv->catas_err.list);
 	init_timer(&priv->catas_err.timer);
@@ -124,8 +124,8 @@ void mlx4_start_catas_poll(struct mlx4_dev *dev)
 
 	priv->catas_err.map = ioremap(addr, priv->fw.catas_size * 4);
 	if (!priv->catas_err.map) {
-		mlx4_warn(dev, "Failed to map internal error buffer at 0x%lx\n",
-			  addr);
+		mlx4_warn(dev, "Failed to map internal error buffer at 0x%llx\n",
+			  (unsigned long long) addr);
 		return;
 	}
 
diff --git a/drivers/net/mlx4/en_main.c b/drivers/net/mlx4/en_main.c
index f6e0d40..1ff6ca6 100644
--- a/drivers/net/mlx4/en_main.c
+++ b/drivers/net/mlx4/en_main.c
@@ -202,7 +202,8 @@ static void *mlx4_en_add(struct mlx4_dev *dev)
 	if (mlx4_uar_alloc(dev, &mdev->priv_uar))
 		goto err_pd;
 
-	mdev->uar_map = ioremap(mdev->priv_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
+	mdev->uar_map = ioremap((phys_addr_t) mdev->priv_uar.pfn << PAGE_SHIFT,
+				PAGE_SIZE);
 	if (!mdev->uar_map)
 		goto err_uar;
 	spin_lock_init(&mdev->uar_lock);
diff --git a/drivers/net/mlx4/main.c b/drivers/net/mlx4/main.c
index 782f11d..4ffdc18 100644
--- a/drivers/net/mlx4/main.c
+++ b/drivers/net/mlx4/main.c
@@ -829,7 +829,7 @@ static int mlx4_setup_hca(struct mlx4_dev *dev)
 		goto err_uar_table_free;
 	}
 
-	priv->kar = ioremap(priv->driver_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
+	priv->kar = ioremap((phys_addr_t) priv->driver_uar.pfn << PAGE_SHIFT, PAGE_SIZE);
 	if (!priv->kar) {
 		mlx4_err(dev, "Couldn't map kernel access region, "
 			 "aborting.\n");
-- 
cgit v0.10.2


From 948579cd8c6ea7c8c98c52b79f4470952e182ebd Mon Sep 17 00:00:00 2001
From: Joe Perches <joe@perches.com>
Date: Fri, 5 Nov 2010 03:07:36 +0000
Subject: RDMA: Use vzalloc() to replace vmalloc()+memset(0)

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/amso1100/c2_rnic.c b/drivers/infiniband/hw/amso1100/c2_rnic.c
index 85cfae4..8c81992 100644
--- a/drivers/infiniband/hw/amso1100/c2_rnic.c
+++ b/drivers/infiniband/hw/amso1100/c2_rnic.c
@@ -459,13 +459,12 @@ int __devinit c2_rnic_init(struct c2_dev *c2dev)
 	     IB_DEVICE_MEM_WINDOW);
 
 	/* Allocate the qptr_array */
-	c2dev->qptr_array = vmalloc(C2_MAX_CQS * sizeof(void *));
+	c2dev->qptr_array = vzalloc(C2_MAX_CQS * sizeof(void *));
 	if (!c2dev->qptr_array) {
 		return -ENOMEM;
 	}
 
-	/* Inialize the qptr_array */
-	memset(c2dev->qptr_array, 0, C2_MAX_CQS * sizeof(void *));
+	/* Initialize the qptr_array */
 	c2dev->qptr_array[0] = (void *) &c2dev->req_vq;
 	c2dev->qptr_array[1] = (void *) &c2dev->rep_vq;
 	c2dev->qptr_array[2] = (void *) &c2dev->aeq;
diff --git a/drivers/infiniband/hw/ehca/ipz_pt_fn.c b/drivers/infiniband/hw/ehca/ipz_pt_fn.c
index 1596e30..1898d6e 100644
--- a/drivers/infiniband/hw/ehca/ipz_pt_fn.c
+++ b/drivers/infiniband/hw/ehca/ipz_pt_fn.c
@@ -222,15 +222,14 @@ int ipz_queue_ctor(struct ehca_pd *pd, struct ipz_queue *queue,
 	queue->small_page = NULL;
 
 	/* allocate queue page pointers */
-	queue->queue_pages = kmalloc(nr_of_pages * sizeof(void *), GFP_KERNEL);
+	queue->queue_pages = kzalloc(nr_of_pages * sizeof(void *), GFP_KERNEL);
 	if (!queue->queue_pages) {
-		queue->queue_pages = vmalloc(nr_of_pages * sizeof(void *));
+		queue->queue_pages = vzalloc(nr_of_pages * sizeof(void *));
 		if (!queue->queue_pages) {
 			ehca_gen_err("Couldn't allocate queue page list");
 			return 0;
 		}
 	}
-	memset(queue->queue_pages, 0, nr_of_pages * sizeof(void *));
 
 	/* allocate actual queue pages */
 	if (is_small) {
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
index b33f045..ae92da2 100644
--- a/drivers/infiniband/hw/ipath/ipath_driver.c
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c
@@ -199,12 +199,11 @@ static struct ipath_devdata *ipath_alloc_devdata(struct pci_dev *pdev)
 		goto bail;
 	}
 
-	dd = vmalloc(sizeof(*dd));
+	dd = vzalloc(sizeof(*dd));
 	if (!dd) {
 		dd = ERR_PTR(-ENOMEM);
 		goto bail;
 	}
-	memset(dd, 0, sizeof(*dd));
 	dd->ipath_unit = -1;
 
 	spin_lock_irqsave(&ipath_devs_lock, flags);
diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c
index 9292a15..6d4b29c 100644
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c
@@ -1530,7 +1530,7 @@ static int init_subports(struct ipath_devdata *dd,
 	}
 
 	num_subports = uinfo->spu_subport_cnt;
-	pd->subport_uregbase = vmalloc(PAGE_SIZE * num_subports);
+	pd->subport_uregbase = vzalloc(PAGE_SIZE * num_subports);
 	if (!pd->subport_uregbase) {
 		ret = -ENOMEM;
 		goto bail;
@@ -1538,13 +1538,13 @@ static int init_subports(struct ipath_devdata *dd,
 	/* Note: pd->port_rcvhdrq_size isn't initialized yet. */
 	size = ALIGN(dd->ipath_rcvhdrcnt * dd->ipath_rcvhdrentsize *
 		     sizeof(u32), PAGE_SIZE) * num_subports;
-	pd->subport_rcvhdr_base = vmalloc(size);
+	pd->subport_rcvhdr_base = vzalloc(size);
 	if (!pd->subport_rcvhdr_base) {
 		ret = -ENOMEM;
 		goto bail_ureg;
 	}
 
-	pd->subport_rcvegrbuf = vmalloc(pd->port_rcvegrbuf_chunks *
+	pd->subport_rcvegrbuf = vzalloc(pd->port_rcvegrbuf_chunks *
 					pd->port_rcvegrbuf_size *
 					num_subports);
 	if (!pd->subport_rcvegrbuf) {
@@ -1556,11 +1556,6 @@ static int init_subports(struct ipath_devdata *dd,
 	pd->port_subport_id = uinfo->spu_subport_id;
 	pd->active_slaves = 1;
 	set_bit(IPATH_PORT_MASTER_UNINIT, &pd->port_flag);
-	memset(pd->subport_uregbase, 0, PAGE_SIZE * num_subports);
-	memset(pd->subport_rcvhdr_base, 0, size);
-	memset(pd->subport_rcvegrbuf, 0, pd->port_rcvegrbuf_chunks *
-				         pd->port_rcvegrbuf_size *
-				         num_subports);
 	goto bail;
 
 bail_rhdr:
diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c
index 7769382..fef0f42 100644
--- a/drivers/infiniband/hw/ipath/ipath_init_chip.c
+++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c
@@ -442,7 +442,7 @@ static void init_shadow_tids(struct ipath_devdata *dd)
 	struct page **pages;
 	dma_addr_t *addrs;
 
-	pages = vmalloc(dd->ipath_cfgports * dd->ipath_rcvtidcnt *
+	pages = vzalloc(dd->ipath_cfgports * dd->ipath_rcvtidcnt *
 			sizeof(struct page *));
 	if (!pages) {
 		ipath_dev_err(dd, "failed to allocate shadow page * "
@@ -461,9 +461,6 @@ static void init_shadow_tids(struct ipath_devdata *dd)
 		return;
 	}
 
-	memset(pages, 0, dd->ipath_cfgports * dd->ipath_rcvtidcnt *
-	       sizeof(struct page *));
-
 	dd->ipath_pageshadow = pages;
 	dd->ipath_physshadow = addrs;
 }
diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c
index 7896afb..304bd80 100644
--- a/drivers/infiniband/hw/qib/qib_init.c
+++ b/drivers/infiniband/hw/qib/qib_init.c
@@ -270,23 +270,20 @@ static void init_shadow_tids(struct qib_devdata *dd)
 	struct page **pages;
 	dma_addr_t *addrs;
 
-	pages = vmalloc(dd->cfgctxts * dd->rcvtidcnt * sizeof(struct page *));
+	pages = vzalloc(dd->cfgctxts * dd->rcvtidcnt * sizeof(struct page *));
 	if (!pages) {
 		qib_dev_err(dd, "failed to allocate shadow page * "
 			    "array, no expected sends!\n");
 		goto bail;
 	}
 
-	addrs = vmalloc(dd->cfgctxts * dd->rcvtidcnt * sizeof(dma_addr_t));
+	addrs = vzalloc(dd->cfgctxts * dd->rcvtidcnt * sizeof(dma_addr_t));
 	if (!addrs) {
 		qib_dev_err(dd, "failed to allocate shadow dma handle "
 			    "array, no expected sends!\n");
 		goto bail_free;
 	}
 
-	memset(pages, 0, dd->cfgctxts * dd->rcvtidcnt * sizeof(struct page *));
-	memset(addrs, 0, dd->cfgctxts * dd->rcvtidcnt * sizeof(dma_addr_t));
-
 	dd->pageshadow = pages;
 	dd->physshadow = addrs;
 	return;
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index c1c49f2..93d5580 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -352,15 +352,13 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i
 	int ret;
 	int i;
 
-	rx->rx_ring = vmalloc(ipoib_recvq_size * sizeof *rx->rx_ring);
+	rx->rx_ring = vzalloc(ipoib_recvq_size * sizeof *rx->rx_ring);
 	if (!rx->rx_ring) {
 		printk(KERN_WARNING "%s: failed to allocate CM non-SRQ ring (%d entries)\n",
 		       priv->ca->name, ipoib_recvq_size);
 		return -ENOMEM;
 	}
 
-	memset(rx->rx_ring, 0, ipoib_recvq_size * sizeof *rx->rx_ring);
-
 	t = kmalloc(sizeof *t, GFP_KERNEL);
 	if (!t) {
 		ret = -ENOMEM;
@@ -1097,13 +1095,12 @@ static int ipoib_cm_tx_init(struct ipoib_cm_tx *p, u32 qpn,
 	struct ipoib_dev_priv *priv = netdev_priv(p->dev);
 	int ret;
 
-	p->tx_ring = vmalloc(ipoib_sendq_size * sizeof *p->tx_ring);
+	p->tx_ring = vzalloc(ipoib_sendq_size * sizeof *p->tx_ring);
 	if (!p->tx_ring) {
 		ipoib_warn(priv, "failed to allocate tx ring\n");
 		ret = -ENOMEM;
 		goto err_tx;
 	}
-	memset(p->tx_ring, 0, ipoib_sendq_size * sizeof *p->tx_ring);
 
 	p->qp = ipoib_cm_create_tx_qp(p->dev, p);
 	if (IS_ERR(p->qp)) {
@@ -1521,7 +1518,7 @@ static void ipoib_cm_create_srq(struct net_device *dev, int max_sge)
 		return;
 	}
 
-	priv->cm.srq_ring = vmalloc(ipoib_recvq_size * sizeof *priv->cm.srq_ring);
+	priv->cm.srq_ring = vzalloc(ipoib_recvq_size * sizeof *priv->cm.srq_ring);
 	if (!priv->cm.srq_ring) {
 		printk(KERN_WARNING "%s: failed to allocate CM SRQ ring (%d entries)\n",
 		       priv->ca->name, ipoib_recvq_size);
@@ -1530,7 +1527,6 @@ static void ipoib_cm_create_srq(struct net_device *dev, int max_sge)
 		return;
 	}
 
-	memset(priv->cm.srq_ring, 0, ipoib_recvq_size * sizeof *priv->cm.srq_ring);
 }
 
 int ipoib_cm_dev_init(struct net_device *dev)
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index 7a07a72..aca3b44 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -916,13 +916,12 @@ int ipoib_dev_init(struct net_device *dev, struct ib_device *ca, int port)
 		goto out;
 	}
 
-	priv->tx_ring = vmalloc(ipoib_sendq_size * sizeof *priv->tx_ring);
+	priv->tx_ring = vzalloc(ipoib_sendq_size * sizeof *priv->tx_ring);
 	if (!priv->tx_ring) {
 		printk(KERN_WARNING "%s: failed to allocate TX ring (%d entries)\n",
 		       ca->name, ipoib_sendq_size);
 		goto out_rx_ring_cleanup;
 	}
-	memset(priv->tx_ring, 0, ipoib_sendq_size * sizeof *priv->tx_ring);
 
 	/* priv->tx_head, tx_tail & tx_outstanding are already 0 */
 
-- 
cgit v0.10.2


From da995a8aee044bc5d0847e19e351cd48a2cb8bcc Mon Sep 17 00:00:00 2001
From: Aleksey Senin <alex@senin.name>
Date: Thu, 2 Dec 2010 11:44:49 +0000
Subject: IB/mlx4: Handle protocol field in multicast table

The newest device firmware stores IB vs. Ethernet protocol in two bits
in members_count field of multicast group table (0: Infiniband, 1:
Ethernet).  When changing the QP members count for a multicast group,
it important not to reset this information.  When calling multicast
attach first time, the protocol type should be specified.  In this
patch we always set it IB, but in the future we will handle Ethernet
too.  When looking for a QP, the protocol type shoud be checked too.

Signed-off-by: Aleksey Senin <alekseys@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index d68d849..c7a6213 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -623,8 +623,9 @@ static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	struct mlx4_ib_dev *mdev = to_mdev(ibqp->device);
 	struct mlx4_ib_qp *mqp = to_mqp(ibqp);
 
-	err = mlx4_multicast_attach(mdev->dev, &mqp->mqp, gid->raw, !!(mqp->flags &
-				    MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK));
+	err = mlx4_multicast_attach(mdev->dev, &mqp->mqp, gid->raw,
+				    !!(mqp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK),
+				    MLX4_PROTOCOL_IB);
 	if (err)
 		return err;
 
@@ -635,7 +636,7 @@ static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	return 0;
 
 err_add:
-	mlx4_multicast_detach(mdev->dev, &mqp->mqp, gid->raw);
+	mlx4_multicast_detach(mdev->dev, &mqp->mqp, gid->raw, MLX4_PROTOCOL_IB);
 	return err;
 }
 
@@ -665,7 +666,7 @@ static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 	struct mlx4_ib_gid_entry *ge;
 
 	err = mlx4_multicast_detach(mdev->dev,
-				    &mqp->mqp, gid->raw);
+				    &mqp->mqp, gid->raw, MLX4_PROTOCOL_IB);
 	if (err)
 		return err;
 
diff --git a/drivers/net/mlx4/mcg.c b/drivers/net/mlx4/mcg.c
index c4f88b7..79cf42d 100644
--- a/drivers/net/mlx4/mcg.c
+++ b/drivers/net/mlx4/mcg.c
@@ -95,7 +95,8 @@ static int mlx4_MGID_HASH(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *mailbox
  * entry in hash chain and *mgm holds end of hash chain.
  */
 static int find_mgm(struct mlx4_dev *dev,
-		    u8 *gid, struct mlx4_cmd_mailbox *mgm_mailbox,
+		    u8 *gid, enum mlx4_protocol protocol,
+		    struct mlx4_cmd_mailbox *mgm_mailbox,
 		    u16 *hash, int *prev, int *index)
 {
 	struct mlx4_cmd_mailbox *mailbox;
@@ -134,7 +135,8 @@ static int find_mgm(struct mlx4_dev *dev,
 			return err;
 		}
 
-		if (!memcmp(mgm->gid, gid, 16))
+		if (!memcmp(mgm->gid, gid, 16) &&
+		    be32_to_cpu(mgm->members_count) >> 30 == protocol)
 			return err;
 
 		*prev = *index;
@@ -146,7 +148,7 @@ static int find_mgm(struct mlx4_dev *dev,
 }
 
 int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
-			  int block_mcast_loopback)
+			  int block_mcast_loopback, enum mlx4_protocol protocol)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_cmd_mailbox *mailbox;
@@ -165,7 +167,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
 
 	mutex_lock(&priv->mcg_table.mutex);
 
-	err = find_mgm(dev, gid, mailbox, &hash, &prev, &index);
+	err = find_mgm(dev, gid, protocol, mailbox, &hash, &prev, &index);
 	if (err)
 		goto out;
 
@@ -187,7 +189,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
 		memcpy(mgm->gid, gid, 16);
 	}
 
-	members_count = be32_to_cpu(mgm->members_count);
+	members_count = be32_to_cpu(mgm->members_count) & 0xffffff;
 	if (members_count == MLX4_QP_PER_MGM) {
 		mlx4_err(dev, "MGM at index %x is full.\n", index);
 		err = -ENOMEM;
@@ -207,7 +209,7 @@ int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
 	else
 		mgm->qp[members_count++] = cpu_to_be32(qp->qpn & MGM_QPN_MASK);
 
-	mgm->members_count       = cpu_to_be32(members_count);
+	mgm->members_count = cpu_to_be32(members_count | (u32) protocol << 30);
 
 	err = mlx4_WRITE_MCG(dev, index, mailbox);
 	if (err)
@@ -242,7 +244,8 @@ out:
 }
 EXPORT_SYMBOL_GPL(mlx4_multicast_attach);
 
-int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
+int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
+			  enum mlx4_protocol protocol)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_cmd_mailbox *mailbox;
@@ -260,7 +263,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
 
 	mutex_lock(&priv->mcg_table.mutex);
 
-	err = find_mgm(dev, gid, mailbox, &hash, &prev, &index);
+	err = find_mgm(dev, gid, protocol, mailbox, &hash, &prev, &index);
 	if (err)
 		goto out;
 
@@ -270,7 +273,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
 		goto out;
 	}
 
-	members_count = be32_to_cpu(mgm->members_count);
+	members_count = be32_to_cpu(mgm->members_count) & 0xffffff;
 	for (loc = -1, i = 0; i < members_count; ++i)
 		if ((be32_to_cpu(mgm->qp[i]) & MGM_QPN_MASK) == qp->qpn)
 			loc = i;
@@ -282,7 +285,7 @@ int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16])
 	}
 
 
-	mgm->members_count = cpu_to_be32(--members_count);
+	mgm->members_count = cpu_to_be32(--members_count | (u32) protocol << 30);
 	mgm->qp[loc]       = mgm->qp[i - 1];
 	mgm->qp[i - 1]     = 0;
 
diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h
index a7b15bc..0492146 100644
--- a/include/linux/mlx4/device.h
+++ b/include/linux/mlx4/device.h
@@ -144,6 +144,11 @@ enum {
 	MLX4_STAT_RATE_OFFSET	= 5
 };
 
+enum mlx4_protocol {
+	MLX4_PROTOCOL_IB,
+	MLX4_PROTOCOL_EN,
+};
+
 enum {
 	MLX4_MTT_FLAG_PRESENT		= 1
 };
@@ -500,8 +505,9 @@ int mlx4_INIT_PORT(struct mlx4_dev *dev, int port);
 int mlx4_CLOSE_PORT(struct mlx4_dev *dev, int port);
 
 int mlx4_multicast_attach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
-			  int block_mcast_loopback);
-int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16]);
+			  int block_mcast_loopback, enum mlx4_protocol protocol);
+int mlx4_multicast_detach(struct mlx4_dev *dev, struct mlx4_qp *qp, u8 gid[16],
+			  enum mlx4_protocol protocol);
 
 int mlx4_register_mac(struct mlx4_dev *dev, u8 port, u64 mac, int *index);
 void mlx4_unregister_mac(struct mlx4_dev *dev, u8 port, int index);
diff --git a/include/linux/mlx4/driver.h b/include/linux/mlx4/driver.h
index f407cd4..e1eebf7 100644
--- a/include/linux/mlx4/driver.h
+++ b/include/linux/mlx4/driver.h
@@ -34,6 +34,7 @@
 #define MLX4_DRIVER_H
 
 #include <linux/device.h>
+#include <linux/mlx4/device.h>
 
 struct mlx4_dev;
 
@@ -44,11 +45,6 @@ enum mlx4_dev_event {
 	MLX4_DEV_EVENT_PORT_REINIT,
 };
 
-enum mlx4_protocol {
-	MLX4_PROTOCOL_IB,
-	MLX4_PROTOCOL_EN,
-};
-
 struct mlx4_interface {
 	void *			(*add)	 (struct mlx4_dev *dev);
 	void			(*remove)(struct mlx4_dev *dev, void *context);
-- 
cgit v0.10.2


From 695b83495e2fba9d3a883193cfc9d5eefa96a911 Mon Sep 17 00:00:00 2001
From: Bart Van Assche <bvanassche@acm.org>
Date: Thu, 13 Jan 2011 19:02:25 +0000
Subject: IB/srp: Test only once whether iu allocation succeeded

Merge the two tests in srp_queuecommand() of whether information unit
allocation succeeded into one.  An intended side effect of this change
is that we fix the warning:

    drivers/infiniband/ulp/srp/ib_srp.c: In function 'srp_queuecommand':
    drivers/infiniband/ulp/srp/ib_srp.c:1116: warning: 'req' may be used uninitialized in this function

(seen with CONFIG_CC_OPTIMIZE_FOR_SIZE=y at least with gcc 4.4.4)

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Acked-by: David Dillow <dillowda@ornl.gov>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index 4b62105..b28c2b9 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -1132,15 +1132,12 @@ static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
 
 	spin_lock_irqsave(&target->lock, flags);
 	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
-	if (iu) {
-		req = list_first_entry(&target->free_reqs, struct srp_request,
-				      list);
-		list_del(&req->list);
-	}
-	spin_unlock_irqrestore(&target->lock, flags);
-
 	if (!iu)
-		goto err;
+		goto err_unlock;
+
+	req = list_first_entry(&target->free_reqs, struct srp_request, list);
+	list_del(&req->list);
+	spin_unlock_irqrestore(&target->lock, flags);
 
 	dev = target->srp_host->srp_dev->dev;
 	ib_dma_sync_single_for_cpu(dev, iu->dma, srp_max_iu_len,
@@ -1185,6 +1182,8 @@ err_iu:
 
 	spin_lock_irqsave(&target->lock, flags);
 	list_add(&req->list, &target->free_reqs);
+
+err_unlock:
 	spin_unlock_irqrestore(&target->lock, flags);
 
 err:
-- 
cgit v0.10.2


From 2a4c97ead4b375a64063523210939b87ad225b85 Mon Sep 17 00:00:00 2001
From: Maciej Sosnowski <maciej.sosnowski@intel.com>
Date: Wed, 24 Nov 2010 17:29:30 +0000
Subject: RDMA/nes: Fix bonding on iw_nes

Enable configuring bonds on nes devices by adding missing support for
master net_device to the driver.

Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/nes/nes.c b/drivers/infiniband/hw/nes/nes.c
index 0c9f0aa..e17f52c 100644
--- a/drivers/infiniband/hw/nes/nes.c
+++ b/drivers/infiniband/hw/nes/nes.c
@@ -144,6 +144,7 @@ static int nes_inetaddr_event(struct notifier_block *notifier,
 	struct nes_device *nesdev;
 	struct net_device *netdev;
 	struct nes_vnic *nesvnic;
+	unsigned int is_bonded;
 
 	nes_debug(NES_DBG_NETDEV, "nes_inetaddr_event: ip address %pI4, netmask %pI4.\n",
 		  &ifa->ifa_address, &ifa->ifa_mask);
@@ -152,7 +153,8 @@ static int nes_inetaddr_event(struct notifier_block *notifier,
 				nesdev, nesdev->netdev[0]->name);
 		netdev = nesdev->netdev[0];
 		nesvnic = netdev_priv(netdev);
-		if (netdev == event_netdev) {
+		is_bonded = (netdev->master == event_netdev);
+		if ((netdev == event_netdev) || is_bonded) {
 			if (nesvnic->rdma_enabled == 0) {
 				nes_debug(NES_DBG_NETDEV, "Returning without processing event for %s since"
 						" RDMA is not enabled.\n",
@@ -169,7 +171,10 @@ static int nes_inetaddr_event(struct notifier_block *notifier,
 					nes_manage_arp_cache(netdev, netdev->dev_addr,
 							ntohl(nesvnic->local_ipaddr), NES_ARP_DELETE);
 					nesvnic->local_ipaddr = 0;
-					return NOTIFY_OK;
+					if (is_bonded)
+						continue;
+					else
+						return NOTIFY_OK;
 					break;
 				case NETDEV_UP:
 					nes_debug(NES_DBG_NETDEV, "event:UP\n");
@@ -178,15 +183,24 @@ static int nes_inetaddr_event(struct notifier_block *notifier,
 						nes_debug(NES_DBG_NETDEV, "Interface already has local_ipaddr\n");
 						return NOTIFY_OK;
 					}
+					/* fall through */
+				case NETDEV_CHANGEADDR:
 					/* Add the address to the IP table */
-					nesvnic->local_ipaddr = ifa->ifa_address;
+					if (netdev->master)
+						nesvnic->local_ipaddr =
+							((struct in_device *)netdev->master->ip_ptr)->ifa_list->ifa_address;
+					else
+						nesvnic->local_ipaddr = ifa->ifa_address;
 
 					nes_write_indexed(nesdev,
 							NES_IDX_DST_IP_ADDR+(0x10*PCI_FUNC(nesdev->pcidev->devfn)),
-							ntohl(ifa->ifa_address));
+							ntohl(nesvnic->local_ipaddr));
 					nes_manage_arp_cache(netdev, netdev->dev_addr,
 							ntohl(nesvnic->local_ipaddr), NES_ARP_ADD);
-					return NOTIFY_OK;
+					if (is_bonded)
+						continue;
+					else
+						return NOTIFY_OK;
 					break;
 				default:
 					break;
diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c
index 25ad0f9..009ec81 100644
--- a/drivers/infiniband/hw/nes/nes_cm.c
+++ b/drivers/infiniband/hw/nes/nes_cm.c
@@ -1107,6 +1107,7 @@ static int nes_addr_resolve_neigh(struct nes_vnic *nesvnic, u32 dst_ip, int arpi
 	struct flowi fl;
 	struct neighbour *neigh;
 	int rc = arpindex;
+	struct net_device *netdev;
 	struct nes_adapter *nesadapter = nesvnic->nesdev->nesadapter;
 
 	memset(&fl, 0, sizeof fl);
@@ -1117,7 +1118,12 @@ static int nes_addr_resolve_neigh(struct nes_vnic *nesvnic, u32 dst_ip, int arpi
 		return rc;
 	}
 
-	neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, nesvnic->netdev);
+	if (nesvnic->netdev->master)
+		netdev = nesvnic->netdev->master;
+	else
+		netdev = nesvnic->netdev;
+
+	neigh = neigh_lookup(&arp_tbl, &rt->rt_gateway, netdev);
 	if (neigh) {
 		if (neigh->nud_state & NUD_VALID) {
 			nes_debug(NES_DBG_CM, "Neighbor MAC address for 0x%08X"
-- 
cgit v0.10.2


From ea623455b736d82f476460647e8b5fe5dc36f4f2 Mon Sep 17 00:00:00 2001
From: Maciej Sosnowski <maciej.sosnowski@intel.com>
Date: Wed, 24 Nov 2010 17:29:38 +0000
Subject: RDMA/nes: Generate IB_EVENT_PORT_ERR/PORT_ACTIVE events

Depending on link state change, IB_EVENT_PORT_ERR or
IB_EVENT_PORT_ACTIVE should be generated when handling MAC interrupts.

Plugging in a cable happens to result in series of interrupts changing
driver's link state a number of times before finally staying at link
up (e.g. link up, link down, link up, link down, ..., link up).  To
prevent sending series of redundant IB_EVENT_PORT_ACTIVE and
IB_EVENT_PORT_ERR events, we use a timer to debounce them in
nes_port_ibevent().

Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index 1980a46..2b89b06 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -2608,6 +2608,13 @@ static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number)
 						netif_start_queue(nesvnic->netdev);
 					nesvnic->linkup = 1;
 					netif_carrier_on(nesvnic->netdev);
+
+					spin_lock(&nesvnic->port_ibevent_lock);
+					if (nesdev->iw_status == 0) {
+						nesdev->iw_status = 1;
+						nes_port_ibevent(nesvnic);
+					}
+					spin_unlock(&nesvnic->port_ibevent_lock);
 				}
 			}
 		} else {
@@ -2633,6 +2640,13 @@ static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number)
 						netif_stop_queue(nesvnic->netdev);
 					nesvnic->linkup = 0;
 					netif_carrier_off(nesvnic->netdev);
+
+					spin_lock(&nesvnic->port_ibevent_lock);
+					if (nesdev->iw_status == 1) {
+						nesdev->iw_status = 0;
+						nes_port_ibevent(nesvnic);
+					}
+					spin_unlock(&nesvnic->port_ibevent_lock);
 				}
 			}
 		}
diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h
index 1204c34..8a9ea9a 100644
--- a/drivers/infiniband/hw/nes/nes_hw.h
+++ b/drivers/infiniband/hw/nes/nes_hw.h
@@ -1193,6 +1193,8 @@ struct nes_listener {
 
 struct nes_ib_device;
 
+#define NES_EVENT_DELAY msecs_to_jiffies(100)
+
 struct nes_vnic {
 	struct nes_ib_device *nesibdev;
 	u64 sq_full;
@@ -1247,6 +1249,10 @@ struct nes_vnic {
 	u32 lro_max_aggr;
 	struct net_lro_mgr lro_mgr;
 	struct net_lro_desc lro_desc[NES_MAX_LRO_DESCRIPTORS];
+	struct timer_list event_timer;
+	enum ib_event_type delayed_event;
+	enum ib_event_type last_dispatched_event;
+	spinlock_t port_ibevent_lock;
 };
 
 struct nes_ib_device {
diff --git a/drivers/infiniband/hw/nes/nes_nic.c b/drivers/infiniband/hw/nes/nes_nic.c
index 5a4c364..81052fb 100644
--- a/drivers/infiniband/hw/nes/nes_nic.c
+++ b/drivers/infiniband/hw/nes/nes_nic.c
@@ -144,6 +144,7 @@ static int nes_netdev_open(struct net_device *netdev)
 	u32 nic_active_bit;
 	u32 nic_active;
 	struct list_head *list_pos, *list_temp;
+	unsigned long flags;
 
 	assert(nesdev != NULL);
 
@@ -233,18 +234,27 @@ static int nes_netdev_open(struct net_device *netdev)
 		first_nesvnic = nesvnic;
 	}
 
-	if (nesvnic->of_device_registered) {
-		nesdev->iw_status = 1;
-		nesdev->nesadapter->send_term_ok = 1;
-		nes_port_ibevent(nesvnic);
-	}
-
 	if (first_nesvnic->linkup) {
 		/* Enable network packets */
 		nesvnic->linkup = 1;
 		netif_start_queue(netdev);
 		netif_carrier_on(netdev);
 	}
+
+	spin_lock_irqsave(&nesvnic->port_ibevent_lock, flags);
+	if (nesvnic->of_device_registered) {
+		nesdev->nesadapter->send_term_ok = 1;
+		if (nesvnic->linkup == 1) {
+			if (nesdev->iw_status == 0) {
+				nesdev->iw_status = 1;
+				nes_port_ibevent(nesvnic);
+			}
+		} else {
+			nesdev->iw_status = 0;
+		}
+	}
+	spin_unlock_irqrestore(&nesvnic->port_ibevent_lock, flags);
+
 	napi_enable(&nesvnic->napi);
 	nesvnic->netdev_open = 1;
 
@@ -263,6 +273,7 @@ static int nes_netdev_stop(struct net_device *netdev)
 	u32 nic_active;
 	struct nes_vnic *first_nesvnic = NULL;
 	struct list_head *list_pos, *list_temp;
+	unsigned long flags;
 
 	nes_debug(NES_DBG_SHUTDOWN, "nesvnic=%p, nesdev=%p, netdev=%p %s\n",
 			nesvnic, nesdev, netdev, netdev->name);
@@ -315,12 +326,17 @@ static int nes_netdev_stop(struct net_device *netdev)
 	nic_active &= nic_active_mask;
 	nes_write_indexed(nesdev, NES_IDX_NIC_BROADCAST_ON, nic_active);
 
-
+	spin_lock_irqsave(&nesvnic->port_ibevent_lock, flags);
 	if (nesvnic->of_device_registered) {
 		nesdev->nesadapter->send_term_ok = 0;
 		nesdev->iw_status = 0;
-		nes_port_ibevent(nesvnic);
+		if (nesvnic->linkup == 1)
+			nes_port_ibevent(nesvnic);
 	}
+	del_timer_sync(&nesvnic->event_timer);
+	nesvnic->event_timer.function = NULL;
+	spin_unlock_irqrestore(&nesvnic->port_ibevent_lock, flags);
+
 	nes_destroy_nic_qp(nesvnic);
 
 	nesvnic->netdev_open = 0;
@@ -1750,7 +1766,10 @@ struct net_device *nes_netdev_init(struct nes_device *nesdev,
 		nesvnic->rdma_enabled = 0;
 	}
 	nesvnic->nic_cq.cq_number = nesvnic->nic.qp_id;
+	init_timer(&nesvnic->event_timer);
+	nesvnic->event_timer.function = NULL;
 	spin_lock_init(&nesvnic->tx_lock);
+	spin_lock_init(&nesvnic->port_ibevent_lock);
 	nesdev->netdev[nesdev->netdev_count] = netdev;
 
 	nes_debug(NES_DBG_INIT, "Adding nesvnic (%p) to the adapters nesvnic_list for MAC%d.\n",
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 99933e4..26d8018 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -3936,6 +3936,30 @@ struct nes_ib_device *nes_init_ofa_device(struct net_device *netdev)
 	return nesibdev;
 }
 
+
+/**
+ * nes_handle_delayed_event
+ */
+static void nes_handle_delayed_event(unsigned long data)
+{
+	struct nes_vnic *nesvnic = (void *) data;
+
+	if (nesvnic->delayed_event != nesvnic->last_dispatched_event) {
+		struct ib_event event;
+
+		event.device = &nesvnic->nesibdev->ibdev;
+		if (!event.device)
+			goto stop_timer;
+		event.event = nesvnic->delayed_event;
+		event.element.port_num = nesvnic->logical_port + 1;
+		ib_dispatch_event(&event);
+	}
+
+stop_timer:
+	nesvnic->event_timer.function = NULL;
+}
+
+
 void  nes_port_ibevent(struct nes_vnic *nesvnic)
 {
 	struct nes_ib_device *nesibdev = nesvnic->nesibdev;
@@ -3944,7 +3968,18 @@ void  nes_port_ibevent(struct nes_vnic *nesvnic)
 	event.device = &nesibdev->ibdev;
 	event.element.port_num = nesvnic->logical_port + 1;
 	event.event = nesdev->iw_status ? IB_EVENT_PORT_ACTIVE : IB_EVENT_PORT_ERR;
-	ib_dispatch_event(&event);
+
+	if (!nesvnic->event_timer.function) {
+		ib_dispatch_event(&event);
+		nesvnic->last_dispatched_event = event.event;
+		nesvnic->event_timer.function = nes_handle_delayed_event;
+		nesvnic->event_timer.data = (unsigned long) nesvnic;
+		nesvnic->event_timer.expires = jiffies + NES_EVENT_DELAY;
+		add_timer(&nesvnic->event_timer);
+	} else {
+		mod_timer(&nesvnic->event_timer, jiffies + NES_EVENT_DELAY);
+	}
+	nesvnic->delayed_event = event.event;
 }
 
 
-- 
cgit v0.10.2


From 5f61b2c6939bb6d26393df15765bc3cb260db063 Mon Sep 17 00:00:00 2001
From: Maciej Sosnowski <maciej.sosnowski@intel.com>
Date: Wed, 24 Nov 2010 17:29:46 +0000
Subject: RDMA/nes: Fix SFP+ link down detection issue with switch port disable

In case of SFP+ PHY, link status check at interrupt processing can
give false results.  For proper link status change detection a delayed
recheck is needed to give nes registers time to settle.  Add a
periodic link status recheck scheduled at interrupt to detect
potential delayed registers state changes.

Addresses: http://bugs.openfabrics.org/bugzilla/show_bug.cgi?id=2117
Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/nes/nes.c b/drivers/infiniband/hw/nes/nes.c
index e17f52c..3b4ec32 100644
--- a/drivers/infiniband/hw/nes/nes.c
+++ b/drivers/infiniband/hw/nes/nes.c
@@ -674,6 +674,8 @@ static int __devinit nes_probe(struct pci_dev *pcidev, const struct pci_device_i
 	}
 	nes_notifiers_registered++;
 
+	INIT_DELAYED_WORK(&nesdev->work, nes_recheck_link_status);
+
 	/* Initialize network devices */
 	if ((netdev = nes_netdev_init(nesdev, mmio_regs)) == NULL)
 		goto bail7;
@@ -756,6 +758,7 @@ static void __devexit nes_remove(struct pci_dev *pcidev)
 	struct nes_device *nesdev = pci_get_drvdata(pcidev);
 	struct net_device *netdev;
 	int netdev_index = 0;
+	unsigned long flags;
 
 		if (nesdev->netdev_count) {
 			netdev = nesdev->netdev[netdev_index];
@@ -782,6 +785,14 @@ static void __devexit nes_remove(struct pci_dev *pcidev)
 	free_irq(pcidev->irq, nesdev);
 	tasklet_kill(&nesdev->dpc_tasklet);
 
+	spin_lock_irqsave(&nesdev->nesadapter->phy_lock, flags);
+	if (nesdev->link_recheck) {
+		spin_unlock_irqrestore(&nesdev->nesadapter->phy_lock, flags);
+		cancel_delayed_work_sync(&nesdev->work);
+	} else {
+		spin_unlock_irqrestore(&nesdev->nesadapter->phy_lock, flags);
+	}
+
 	/* Deallocate the Adapter Structure */
 	nes_destroy_adapter(nesdev->nesadapter);
 
diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h
index b3d145e..6fe7987 100644
--- a/drivers/infiniband/hw/nes/nes.h
+++ b/drivers/infiniband/hw/nes/nes.h
@@ -268,6 +268,9 @@ struct nes_device {
 	u8                     napi_isr_ran;
 	u8                     disable_rx_flow_control;
 	u8                     disable_tx_flow_control;
+
+	struct delayed_work    work;
+	u8                     link_recheck;
 };
 
 
@@ -507,6 +510,7 @@ void nes_nic_ce_handler(struct nes_device *, struct nes_hw_nic_cq *);
 void nes_iwarp_ce_handler(struct nes_device *, struct nes_hw_cq *);
 int nes_destroy_cqp(struct nes_device *);
 int nes_nic_cm_xmit(struct sk_buff *, struct net_device *);
+void nes_recheck_link_status(struct work_struct *work);
 
 /* nes_nic.c */
 struct net_device *nes_netdev_init(struct nes_device *, void __iomem *);
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index 2b89b06..8b606fd 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -2650,6 +2650,13 @@ static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number)
 				}
 			}
 		}
+		if (nesadapter->phy_type[mac_index] == NES_PHY_TYPE_SFP_D) {
+			if (nesdev->link_recheck)
+				cancel_delayed_work(&nesdev->work);
+			nesdev->link_recheck = 1;
+			schedule_delayed_work(&nesdev->work,
+					      NES_LINK_RECHECK_DELAY);
+		}
 	}
 
 	spin_unlock_irqrestore(&nesadapter->phy_lock, flags);
@@ -2657,6 +2664,80 @@ static void nes_process_mac_intr(struct nes_device *nesdev, u32 mac_number)
 	nesadapter->mac_sw_state[mac_number] = NES_MAC_SW_IDLE;
 }
 
+void nes_recheck_link_status(struct work_struct *work)
+{
+	unsigned long flags;
+	struct nes_device *nesdev = container_of(work, struct nes_device, work.work);
+	struct nes_adapter *nesadapter = nesdev->nesadapter;
+	struct nes_vnic *nesvnic;
+	u32 mac_index = nesdev->mac_index;
+	u16 phy_data;
+	u16 temp_phy_data;
+
+	spin_lock_irqsave(&nesadapter->phy_lock, flags);
+
+	/* check link status */
+	nes_read_10G_phy_reg(nesdev, nesadapter->phy_index[mac_index], 1, 0x9003);
+	temp_phy_data = (u16)nes_read_indexed(nesdev, NES_IDX_MAC_MDIO_CONTROL);
+
+	nes_read_10G_phy_reg(nesdev, nesadapter->phy_index[mac_index], 3, 0x0021);
+	nes_read_indexed(nesdev, NES_IDX_MAC_MDIO_CONTROL);
+	nes_read_10G_phy_reg(nesdev, nesadapter->phy_index[mac_index], 3, 0x0021);
+	phy_data = (u16)nes_read_indexed(nesdev, NES_IDX_MAC_MDIO_CONTROL);
+
+	phy_data = (!temp_phy_data && (phy_data == 0x8000)) ? 0x4 : 0x0;
+
+	nes_debug(NES_DBG_PHY, "%s: Phy data = 0x%04X, link was %s.\n",
+		__func__, phy_data,
+		nesadapter->mac_link_down[mac_index] ? "DOWN" : "UP");
+
+	if (phy_data & 0x0004) {
+		nesadapter->mac_link_down[mac_index] = 0;
+		list_for_each_entry(nesvnic, &nesadapter->nesvnic_list[mac_index], list) {
+			if (nesvnic->linkup == 0) {
+				printk(PFX "The Link is now up for port %s, netdev %p.\n",
+						nesvnic->netdev->name, nesvnic->netdev);
+				if (netif_queue_stopped(nesvnic->netdev))
+					netif_start_queue(nesvnic->netdev);
+				nesvnic->linkup = 1;
+				netif_carrier_on(nesvnic->netdev);
+
+				spin_lock(&nesvnic->port_ibevent_lock);
+				if (nesdev->iw_status == 0) {
+					nesdev->iw_status = 1;
+					nes_port_ibevent(nesvnic);
+				}
+				spin_unlock(&nesvnic->port_ibevent_lock);
+			}
+		}
+
+	} else {
+		nesadapter->mac_link_down[mac_index] = 1;
+		list_for_each_entry(nesvnic, &nesadapter->nesvnic_list[mac_index], list) {
+			if (nesvnic->linkup == 1) {
+				printk(PFX "The Link is now down for port %s, netdev %p.\n",
+						nesvnic->netdev->name, nesvnic->netdev);
+				if (!(netif_queue_stopped(nesvnic->netdev)))
+					netif_stop_queue(nesvnic->netdev);
+				nesvnic->linkup = 0;
+				netif_carrier_off(nesvnic->netdev);
+
+				spin_lock(&nesvnic->port_ibevent_lock);
+				if (nesdev->iw_status == 1) {
+					nesdev->iw_status = 0;
+					nes_port_ibevent(nesvnic);
+				}
+				spin_unlock(&nesvnic->port_ibevent_lock);
+			}
+		}
+	}
+	if (nesdev->link_recheck++ < NES_LINK_RECHECK_MAX)
+		schedule_delayed_work(&nesdev->work, NES_LINK_RECHECK_DELAY);
+	else
+		nesdev->link_recheck = 0;
+
+	spin_unlock_irqrestore(&nesadapter->phy_lock, flags);
+}
 
 
 static void nes_nic_napi_ce_handler(struct nes_device *nesdev, struct nes_hw_nic_cq *cq)
diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h
index 8a9ea9a..d2abe07 100644
--- a/drivers/infiniband/hw/nes/nes_hw.h
+++ b/drivers/infiniband/hw/nes/nes_hw.h
@@ -1354,6 +1354,10 @@ struct nes_terminate_hdr {
 #define BAD_FRAME_OFFSET	64
 #define CQE_MAJOR_DRV		0x8000
 
+/* Used for link status recheck after interrupt processing */
+#define NES_LINK_RECHECK_DELAY	msecs_to_jiffies(50)
+#define NES_LINK_RECHECK_MAX	60
+
 #define nes_vlan_rx vlan_hwaccel_receive_skb
 #define nes_netif_rx netif_receive_skb
 
-- 
cgit v0.10.2


From 843276ad985cb59212ceb70d989474521ff516f6 Mon Sep 17 00:00:00 2001
From: Maciej Sosnowski <maciej.sosnowski@intel.com>
Date: Wed, 24 Nov 2010 17:29:54 +0000
Subject: RDMA/nes: Fix incorrect SFP+ link status detection on driver init

During iw_nes initialization the link status for SFP+ PHY is always
detected as "up" regardless of real state (cable either connected or
disconnected).  Add SFP+ PHY specific link status detection to the
iw_nes initialization procedure.  Use link status recheck for
netdev_open to detect delayed state updates.

Signed-off-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/nes/nes_nic.c b/drivers/infiniband/hw/nes/nes_nic.c
index 81052fb..2c9c193 100644
--- a/drivers/infiniband/hw/nes/nes_nic.c
+++ b/drivers/infiniband/hw/nes/nes_nic.c
@@ -241,6 +241,15 @@ static int nes_netdev_open(struct net_device *netdev)
 		netif_carrier_on(netdev);
 	}
 
+	spin_lock_irqsave(&nesdev->nesadapter->phy_lock, flags);
+	if (nesdev->nesadapter->phy_type[nesdev->mac_index] == NES_PHY_TYPE_SFP_D) {
+		if (nesdev->link_recheck)
+			cancel_delayed_work(&nesdev->work);
+		nesdev->link_recheck = 1;
+		schedule_delayed_work(&nesdev->work, NES_LINK_RECHECK_DELAY);
+	}
+	spin_unlock_irqrestore(&nesdev->nesadapter->phy_lock, flags);
+
 	spin_lock_irqsave(&nesvnic->port_ibevent_lock, flags);
 	if (nesvnic->of_device_registered) {
 		nesdev->nesadapter->send_term_ok = 1;
@@ -1782,8 +1791,11 @@ struct net_device *nes_netdev_init(struct nes_device *nesdev,
 	      (((PCI_FUNC(nesdev->pcidev->devfn) == 1) && (nesdev->mac_index == 2)) ||
 	       ((PCI_FUNC(nesdev->pcidev->devfn) == 2) && (nesdev->mac_index == 1)))))) {
 		u32 u32temp;
-		u32 link_mask;
-		u32 link_val;
+		u32 link_mask = 0;
+		u32 link_val = 0;
+		u16 temp_phy_data;
+		u16 phy_data = 0;
+		unsigned long flags;
 
 		u32temp = nes_read_indexed(nesdev, NES_IDX_PHY_PCS_CONTROL_STATUS0 +
 				(0x200 * (nesdev->mac_index & 1)));
@@ -1805,6 +1817,23 @@ struct net_device *nes_netdev_init(struct nes_device *nesdev,
 				link_val = 0x02020000;
 			}
 			break;
+		case NES_PHY_TYPE_SFP_D:
+			spin_lock_irqsave(&nesdev->nesadapter->phy_lock, flags);
+			nes_read_10G_phy_reg(nesdev,
+					     nesdev->nesadapter->phy_index[nesdev->mac_index],
+					     1, 0x9003);
+			temp_phy_data = (u16)nes_read_indexed(nesdev, NES_IDX_MAC_MDIO_CONTROL);
+			nes_read_10G_phy_reg(nesdev,
+					     nesdev->nesadapter->phy_index[nesdev->mac_index],
+					     3, 0x0021);
+			nes_read_indexed(nesdev, NES_IDX_MAC_MDIO_CONTROL);
+			nes_read_10G_phy_reg(nesdev,
+					     nesdev->nesadapter->phy_index[nesdev->mac_index],
+					     3, 0x0021);
+			phy_data = (u16)nes_read_indexed(nesdev, NES_IDX_MAC_MDIO_CONTROL);
+			spin_unlock_irqrestore(&nesdev->nesadapter->phy_lock, flags);
+			phy_data = (!temp_phy_data && (phy_data == 0x8000)) ? 0x4 : 0x0;
+			break;
 		default:
 			link_mask = 0x0f1f0000;
 			link_val = 0x0f0f0000;
@@ -1814,8 +1843,14 @@ struct net_device *nes_netdev_init(struct nes_device *nesdev,
 		u32temp = nes_read_indexed(nesdev,
 					   NES_IDX_PHY_PCS_CONTROL_STATUS0 +
 					   (0x200 * (nesdev->mac_index & 1)));
-		if ((u32temp & link_mask) == link_val)
-			nesvnic->linkup = 1;
+
+		if (phy_type == NES_PHY_TYPE_SFP_D) {
+			if (phy_data & 0x0004)
+				nesvnic->linkup = 1;
+		} else {
+			if ((u32temp & link_mask) == link_val)
+				nesvnic->linkup = 1;
+		}
 
 		/* clear the MAC interrupt status, assumes direct logical to physical mapping */
 		u32temp = nes_read_indexed(nesdev, NES_IDX_MAC_INT_STATUS + (0x200 * nesdev->mac_index));
-- 
cgit v0.10.2


From f06267104dd9112f11586830d22501d0e26245ea Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Tue, 19 Oct 2010 15:24:36 +0000
Subject: RDMA: Update workqueue usage

* ib_wq is added, which is used as the common workqueue for infiniband
  instead of the system workqueue.  All system workqueue usages
  including flush_scheduled_work() callers are converted to use and
  flush ib_wq.

* cancel_delayed_work() + flush_scheduled_work() converted to
  cancel_delayed_work_sync().

* qib_wq is removed and ib_wq is used instead.

This is to prepare for deprecation of flush_scheduled_work().

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c
index 6888356..f9ba7d7 100644
--- a/drivers/infiniband/core/cache.c
+++ b/drivers/infiniband/core/cache.c
@@ -308,7 +308,7 @@ static void ib_cache_event(struct ib_event_handler *handler,
 			INIT_WORK(&work->work, ib_cache_task);
 			work->device   = event->device;
 			work->port_num = event->element.port_num;
-			schedule_work(&work->work);
+			queue_work(ib_wq, &work->work);
 		}
 	}
 }
@@ -368,7 +368,7 @@ static void ib_cache_cleanup_one(struct ib_device *device)
 	int p;
 
 	ib_unregister_event_handler(&device->cache.event_handler);
-	flush_scheduled_work();
+	flush_workqueue(ib_wq);
 
 	for (p = 0; p <= end_port(device) - start_port(device); ++p) {
 		kfree(device->cache.pkey_cache[p]);
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index a19effa..f793bf2 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -38,7 +38,6 @@
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <linux/mutex.h>
-#include <linux/workqueue.h>
 
 #include "core_priv.h"
 
@@ -52,6 +51,9 @@ struct ib_client_data {
 	void *            data;
 };
 
+struct workqueue_struct *ib_wq;
+EXPORT_SYMBOL_GPL(ib_wq);
+
 static LIST_HEAD(device_list);
 static LIST_HEAD(client_list);
 
@@ -718,6 +720,10 @@ static int __init ib_core_init(void)
 {
 	int ret;
 
+	ib_wq = alloc_workqueue("infiniband", 0, 0);
+	if (!ib_wq)
+		return -ENOMEM;
+
 	ret = ib_sysfs_setup();
 	if (ret)
 		printk(KERN_WARNING "Couldn't create InfiniBand device class\n");
@@ -726,6 +732,7 @@ static int __init ib_core_init(void)
 	if (ret) {
 		printk(KERN_WARNING "Couldn't set up InfiniBand P_Key/GID cache\n");
 		ib_sysfs_cleanup();
+		destroy_workqueue(ib_wq);
 	}
 
 	return ret;
@@ -736,7 +743,7 @@ static void __exit ib_core_cleanup(void)
 	ib_cache_cleanup();
 	ib_sysfs_cleanup();
 	/* Make sure that any pending umem accounting work is done. */
-	flush_scheduled_work();
+	destroy_workqueue(ib_wq);
 }
 
 module_init(ib_core_init);
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index 91a6603..e38be1b 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -425,7 +425,7 @@ static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event
 		port->sm_ah = NULL;
 		spin_unlock_irqrestore(&port->ah_lock, flags);
 
-		schedule_work(&sa_dev->port[event->element.port_num -
+		queue_work(ib_wq, &sa_dev->port[event->element.port_num -
 					    sa_dev->start_port].update_task);
 	}
 }
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index 415e186..b645e55 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -262,7 +262,7 @@ void ib_umem_release(struct ib_umem *umem)
 			umem->mm   = mm;
 			umem->diff = diff;
 
-			schedule_work(&umem->work);
+			queue_work(ib_wq, &umem->work);
 			return;
 		}
 	} else
diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
index ae92da2..47db4bf 100644
--- a/drivers/infiniband/hw/ipath/ipath_driver.c
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c
@@ -755,7 +755,7 @@ static void __devexit ipath_remove_one(struct pci_dev *pdev)
 	 */
 	ipath_shutdown_device(dd);
 
-	flush_scheduled_work();
+	flush_workqueue(ib_wq);
 
 	if (dd->verbs_dev)
 		ipath_unregister_ib_device(dd->verbs_dev);
diff --git a/drivers/infiniband/hw/ipath/ipath_user_pages.c b/drivers/infiniband/hw/ipath/ipath_user_pages.c
index 5e86d73..bab9f74 100644
--- a/drivers/infiniband/hw/ipath/ipath_user_pages.c
+++ b/drivers/infiniband/hw/ipath/ipath_user_pages.c
@@ -220,7 +220,7 @@ void ipath_release_user_pages_on_close(struct page **p, size_t num_pages)
 	work->mm = mm;
 	work->num_pages = num_pages;
 
-	schedule_work(&work->work);
+	queue_work(ib_wq, &work->work);
 	return;
 
 bail_mm:
diff --git a/drivers/infiniband/hw/qib/qib_iba7220.c b/drivers/infiniband/hw/qib/qib_iba7220.c
index 127a0d5..de799f1 100644
--- a/drivers/infiniband/hw/qib/qib_iba7220.c
+++ b/drivers/infiniband/hw/qib/qib_iba7220.c
@@ -1692,8 +1692,7 @@ static void qib_7220_quiet_serdes(struct qib_pportdata *ppd)
 	ppd->lflags &= ~QIBL_IB_AUTONEG_INPROG;
 	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
 	wake_up(&ppd->cpspec->autoneg_wait);
-	cancel_delayed_work(&ppd->cpspec->autoneg_work);
-	flush_scheduled_work();
+	cancel_delayed_work_sync(&ppd->cpspec->autoneg_work);
 
 	shutdown_7220_relock_poll(ppd->dd);
 	val = qib_read_kreg64(ppd->dd, kr_xgxs_cfg);
@@ -3515,8 +3514,8 @@ static void try_7220_autoneg(struct qib_pportdata *ppd)
 
 	toggle_7220_rclkrls(ppd->dd);
 	/* 2 msec is minimum length of a poll cycle */
-	schedule_delayed_work(&ppd->cpspec->autoneg_work,
-			      msecs_to_jiffies(2));
+	queue_delayed_work(ib_wq, &ppd->cpspec->autoneg_work,
+			   msecs_to_jiffies(2));
 }
 
 /*
diff --git a/drivers/infiniband/hw/qib/qib_iba7322.c b/drivers/infiniband/hw/qib/qib_iba7322.c
index dbbb0e8..ea46fbc 100644
--- a/drivers/infiniband/hw/qib/qib_iba7322.c
+++ b/drivers/infiniband/hw/qib/qib_iba7322.c
@@ -2406,10 +2406,9 @@ static void qib_7322_mini_quiet_serdes(struct qib_pportdata *ppd)
 	ppd->lflags &= ~QIBL_IB_AUTONEG_INPROG;
 	spin_unlock_irqrestore(&ppd->lflags_lock, flags);
 	wake_up(&ppd->cpspec->autoneg_wait);
-	cancel_delayed_work(&ppd->cpspec->autoneg_work);
+	cancel_delayed_work_sync(&ppd->cpspec->autoneg_work);
 	if (ppd->dd->cspec->r1)
-		cancel_delayed_work(&ppd->cpspec->ipg_work);
-	flush_scheduled_work();
+		cancel_delayed_work_sync(&ppd->cpspec->ipg_work);
 
 	ppd->cpspec->chase_end = 0;
 	if (ppd->cpspec->chase_timer.data) /* if initted */
@@ -2706,7 +2705,7 @@ static noinline void unknown_7322_gpio_intr(struct qib_devdata *dd)
 			if (!(pins & mask)) {
 				++handled;
 				qd->t_insert = get_jiffies_64();
-				schedule_work(&qd->work);
+				queue_work(ib_wq, &qd->work);
 			}
 		}
 	}
@@ -4990,8 +4989,8 @@ static void try_7322_autoneg(struct qib_pportdata *ppd)
 	set_7322_ibspeed_fast(ppd, QIB_IB_DDR);
 	qib_7322_mini_pcs_reset(ppd);
 	/* 2 msec is minimum length of a poll cycle */
-	schedule_delayed_work(&ppd->cpspec->autoneg_work,
-			      msecs_to_jiffies(2));
+	queue_delayed_work(ib_wq, &ppd->cpspec->autoneg_work,
+			   msecs_to_jiffies(2));
 }
 
 /*
@@ -5121,7 +5120,8 @@ static void try_7322_ipg(struct qib_pportdata *ppd)
 		ib_free_send_mad(send_buf);
 retry:
 	delay = 2 << ppd->cpspec->ipg_tries;
-	schedule_delayed_work(&ppd->cpspec->ipg_work, msecs_to_jiffies(delay));
+	queue_delayed_work(ib_wq, &ppd->cpspec->ipg_work,
+			   msecs_to_jiffies(delay));
 }
 
 /*
diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c
index 304bd80..ffefb78 100644
--- a/drivers/infiniband/hw/qib/qib_init.c
+++ b/drivers/infiniband/hw/qib/qib_init.c
@@ -80,7 +80,6 @@ unsigned qib_wc_pat = 1; /* default (1) is to use PAT, not MTRR */
 module_param_named(wc_pat, qib_wc_pat, uint, S_IRUGO);
 MODULE_PARM_DESC(wc_pat, "enable write-combining via PAT mechanism");
 
-struct workqueue_struct *qib_wq;
 struct workqueue_struct *qib_cq_wq;
 
 static void verify_interrupt(unsigned long);
@@ -1044,24 +1043,10 @@ static int __init qlogic_ib_init(void)
 	if (ret)
 		goto bail;
 
-	/*
-	 * We create our own workqueue mainly because we want to be
-	 * able to flush it when devices are being removed.  We can't
-	 * use schedule_work()/flush_scheduled_work() because both
-	 * unregister_netdev() and linkwatch_event take the rtnl lock,
-	 * so flush_scheduled_work() can deadlock during device
-	 * removal.
-	 */
-	qib_wq = create_workqueue("qib");
-	if (!qib_wq) {
-		ret = -ENOMEM;
-		goto bail_dev;
-	}
-
 	qib_cq_wq = create_singlethread_workqueue("qib_cq");
 	if (!qib_cq_wq) {
 		ret = -ENOMEM;
-		goto bail_wq;
+		goto bail_dev;
 	}
 
 	/*
@@ -1091,8 +1076,6 @@ bail_unit:
 	idr_destroy(&qib_unit_table);
 bail_cq_wq:
 	destroy_workqueue(qib_cq_wq);
-bail_wq:
-	destroy_workqueue(qib_wq);
 bail_dev:
 	qib_dev_cleanup();
 bail:
@@ -1116,7 +1099,6 @@ static void __exit qlogic_ib_cleanup(void)
 
 	pci_unregister_driver(&qib_driver);
 
-	destroy_workqueue(qib_wq);
 	destroy_workqueue(qib_cq_wq);
 
 	qib_cpulist_count = 0;
@@ -1289,7 +1271,7 @@ static int __devinit qib_init_one(struct pci_dev *pdev,
 
 	if (qib_mini_init || initfail || ret) {
 		qib_stop_timers(dd);
-		flush_scheduled_work();
+		flush_workqueue(ib_wq);
 		for (pidx = 0; pidx < dd->num_pports; ++pidx)
 			dd->f_quiet_serdes(dd->pport + pidx);
 		if (qib_mini_init)
@@ -1338,8 +1320,8 @@ static void __devexit qib_remove_one(struct pci_dev *pdev)
 
 	qib_stop_timers(dd);
 
-	/* wait until all of our (qsfp) schedule_work() calls complete */
-	flush_scheduled_work();
+	/* wait until all of our (qsfp) queue_work() calls complete */
+	flush_workqueue(ib_wq);
 
 	ret = qibfs_remove(dd);
 	if (ret)
diff --git a/drivers/infiniband/hw/qib/qib_qsfp.c b/drivers/infiniband/hw/qib/qib_qsfp.c
index 35b3604..3374a52 100644
--- a/drivers/infiniband/hw/qib/qib_qsfp.c
+++ b/drivers/infiniband/hw/qib/qib_qsfp.c
@@ -485,7 +485,7 @@ void qib_qsfp_init(struct qib_qsfp_data *qd,
 		goto bail;
 	/* We see a module, but it may be unwise to look yet. Just schedule */
 	qd->t_insert = get_jiffies_64();
-	schedule_work(&qd->work);
+	queue_work(ib_wq, &qd->work);
 bail:
 	return;
 }
@@ -493,10 +493,9 @@ bail:
 void qib_qsfp_deinit(struct qib_qsfp_data *qd)
 {
 	/*
-	 * There is nothing to do here for now.  our
-	 * work is scheduled with schedule_work(), and
-	 * flush_scheduled_work() from remove_one will
-	 * block until all work ssetup with schedule_work()
+	 * There is nothing to do here for now.  our work is scheduled
+	 * with queue_work(), and flush_workqueue() from remove_one
+	 * will block until all work setup with queue_work()
 	 * completes.
 	 */
 }
diff --git a/drivers/infiniband/hw/qib/qib_verbs.h b/drivers/infiniband/hw/qib/qib_verbs.h
index 63b22a9..95e5b47 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.h
+++ b/drivers/infiniband/hw/qib/qib_verbs.h
@@ -805,7 +805,6 @@ static inline int qib_send_ok(struct qib_qp *qp)
 		 !(qp->s_flags & QIB_S_ANY_WAIT_SEND));
 }
 
-extern struct workqueue_struct *qib_wq;
 extern struct workqueue_struct *qib_cq_wq;
 
 /*
@@ -814,7 +813,7 @@ extern struct workqueue_struct *qib_cq_wq;
 static inline void qib_schedule_send(struct qib_qp *qp)
 {
 	if (qib_send_ok(qp))
-		queue_work(qib_wq, &qp->s_work);
+		queue_work(ib_wq, &qp->s_work);
 }
 
 static inline int qib_pkey_ok(u16 pkey1, u16 pkey2)
diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index 4b62105..70ecb94 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -638,7 +638,7 @@ err:
 	if (target->state == SRP_TARGET_CONNECTING) {
 		target->state = SRP_TARGET_DEAD;
 		INIT_WORK(&target->work, srp_remove_work);
-		schedule_work(&target->work);
+		queue_work(ib_wq, &target->work);
 	}
 	spin_unlock_irq(&target->lock);
 
@@ -2199,7 +2199,7 @@ static void srp_remove_one(struct ib_device *device)
 		 * started before we marked our target ports as
 		 * removed, and any target port removal tasks.
 		 */
-		flush_scheduled_work();
+		flush_workqueue(ib_wq);
 
 		list_for_each_entry_safe(target, tmp_target,
 					 &host->target_list, list) {
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index e04c488..55cd0a0 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -47,10 +47,13 @@
 #include <linux/list.h>
 #include <linux/rwsem.h>
 #include <linux/scatterlist.h>
+#include <linux/workqueue.h>
 
 #include <asm/atomic.h>
 #include <asm/uaccess.h>
 
+extern struct workqueue_struct *ib_wq;
+
 union ib_gid {
 	u8	raw[16];
 	struct {
-- 
cgit v0.10.2