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

Pull infiniband updates from Roland Dreier:
 "First batch of InfiniBand/RDMA changes for the 3.7 merge window:
   - mlx4 IB support for SR-IOV
   - A couple of SRP initiator fixes
   - Batch of nes hardware driver fixes
   - Fix for long-standing use-after-free crash in IPoIB
   - Other miscellaneous fixes"

This merge also removes a new use of __cancel_delayed_work(), and
replaces it with the regular cancel_delayed_work() that is now irq-safe
thanks to the workqueue updates.

That said, I suspect the sequence in question should probably use
"mod_delayed_work()".  I just did the minimal "don't use deprecated
functions" fixup, though.

* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (45 commits)
  IB/qib: Fix local access validation for user MRs
  mlx4_core: Disable SENSE_PORT for multifunction devices
  mlx4_core: Clean up enabling of SENSE_PORT for older (ConnectX-1/-2) HCAs
  mlx4_core: Stash PCI ID driver_data in mlx4_priv structure
  IB/srp: Avoid having aborted requests hang
  IB/srp: Fix use-after-free in srp_reset_req()
  IB/qib: Add a qib driver version
  RDMA/nes: Fix compilation error when nes_debug is enabled
  RDMA/nes: Print hardware resource type
  RDMA/nes: Fix for crash when TX checksum offload is off
  RDMA/nes: Cosmetic changes
  RDMA/nes: Fix for incorrect MSS when TSO is on
  RDMA/nes: Fix incorrect resolving of the loopback MAC address
  mlx4_core: Fix crash on uninitialized priv->cmd.slave_sem
  mlx4_core: Trivial cleanups to driver log messages
  mlx4_core: Trivial readability fix: "0X30" -> "0x30"
  IB/mlx4: Create paravirt contexts for VFs when master IB driver initializes
  mlx4: Modify proxy/tunnel QP mechanism so that guests do no calculations
  mlx4: Paravirtualize Node Guids for slaves
  mlx4: Activate SR-IOV mode for IB
  ...

9 files changed, 1113 insertions, 193 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx4/cmd.c b/drivers/net/ethernet/mellanox/mlx4/cmd.c
index c8fef435302..3d1899ff107 100644
--- a/drivers/net/ethernet/mellanox/mlx4/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx4/cmd.c
@@ -40,6 +40,7 @@
 
 #include <linux/mlx4/cmd.h>
 #include <linux/semaphore.h>
+#include <rdma/ib_smi.h>
 
 #include <asm/io.h>
 
@@ -394,7 +395,8 @@ static int mlx4_slave_cmd(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
 	struct mlx4_vhcr_cmd *vhcr = priv->mfunc.vhcr;
 	int ret;
 
-	down(&priv->cmd.slave_sem);
+	mutex_lock(&priv->cmd.slave_cmd_mutex);
+
 	vhcr->in_param = cpu_to_be64(in_param);
 	vhcr->out_param = out_param ? cpu_to_be64(*out_param) : 0;
 	vhcr->in_modifier = cpu_to_be32(in_modifier);
@@ -402,6 +404,7 @@ static int mlx4_slave_cmd(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
 	vhcr->token = cpu_to_be16(CMD_POLL_TOKEN);
 	vhcr->status = 0;
 	vhcr->flags = !!(priv->cmd.use_events) << 6;
+
 	if (mlx4_is_master(dev)) {
 		ret = mlx4_master_process_vhcr(dev, dev->caps.function, vhcr);
 		if (!ret) {
@@ -438,7 +441,8 @@ static int mlx4_slave_cmd(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
 			mlx4_err(dev, "failed execution of VHCR_POST command"
 				 "opcode 0x%x\n", op);
 	}
-	up(&priv->cmd.slave_sem);
+
+	mutex_unlock(&priv->cmd.slave_cmd_mutex);
 	return ret;
 }
 
@@ -627,6 +631,162 @@ static int mlx4_ACCESS_MEM(struct mlx4_dev *dev, u64 master_addr,
 			    MLX4_CMD_TIME_CLASS_A, MLX4_CMD_NATIVE);
 }
 
+static int query_pkey_block(struct mlx4_dev *dev, u8 port, u16 index, u16 *pkey,
+			       struct mlx4_cmd_mailbox *inbox,
+			       struct mlx4_cmd_mailbox *outbox)
+{
+	struct ib_smp *in_mad = (struct ib_smp *)(inbox->buf);
+	struct ib_smp *out_mad = (struct ib_smp *)(outbox->buf);
+	int err;
+	int i;
+
+	if (index & 0x1f)
+		return -EINVAL;
+
+	in_mad->attr_mod = cpu_to_be32(index / 32);
+
+	err = mlx4_cmd_box(dev, inbox->dma, outbox->dma, port, 3,
+			   MLX4_CMD_MAD_IFC, MLX4_CMD_TIME_CLASS_C,
+			   MLX4_CMD_NATIVE);
+	if (err)
+		return err;
+
+	for (i = 0; i < 32; ++i)
+		pkey[i] = be16_to_cpu(((__be16 *) out_mad->data)[i]);
+
+	return err;
+}
+
+static int get_full_pkey_table(struct mlx4_dev *dev, u8 port, u16 *table,
+			       struct mlx4_cmd_mailbox *inbox,
+			       struct mlx4_cmd_mailbox *outbox)
+{
+	int i;
+	int err;
+
+	for (i = 0; i < dev->caps.pkey_table_len[port]; i += 32) {
+		err = query_pkey_block(dev, port, i, table + i, inbox, outbox);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+#define PORT_CAPABILITY_LOCATION_IN_SMP 20
+#define PORT_STATE_OFFSET 32
+
+static enum ib_port_state vf_port_state(struct mlx4_dev *dev, int port, int vf)
+{
+	if (mlx4_get_slave_port_state(dev, vf, port) == SLAVE_PORT_UP)
+		return IB_PORT_ACTIVE;
+	else
+		return IB_PORT_DOWN;
+}
+
+static int mlx4_MAD_IFC_wrapper(struct mlx4_dev *dev, int slave,
+				struct mlx4_vhcr *vhcr,
+				struct mlx4_cmd_mailbox *inbox,
+				struct mlx4_cmd_mailbox *outbox,
+				struct mlx4_cmd_info *cmd)
+{
+	struct ib_smp *smp = inbox->buf;
+	u32 index;
+	u8 port;
+	u16 *table;
+	int err;
+	int vidx, pidx;
+	struct mlx4_priv *priv = mlx4_priv(dev);
+	struct ib_smp *outsmp = outbox->buf;
+	__be16 *outtab = (__be16 *)(outsmp->data);
+	__be32 slave_cap_mask;
+	__be64 slave_node_guid;
+	port = vhcr->in_modifier;
+
+	if (smp->base_version == 1 &&
+	    smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
+	    smp->class_version == 1) {
+		if (smp->method	== IB_MGMT_METHOD_GET) {
+			if (smp->attr_id == IB_SMP_ATTR_PKEY_TABLE) {
+				index = be32_to_cpu(smp->attr_mod);
+				if (port < 1 || port > dev->caps.num_ports)
+					return -EINVAL;
+				table = kcalloc(dev->caps.pkey_table_len[port], sizeof *table, GFP_KERNEL);
+				if (!table)
+					return -ENOMEM;
+				/* need to get the full pkey table because the paravirtualized
+				 * pkeys may be scattered among several pkey blocks.
+				 */
+				err = get_full_pkey_table(dev, port, table, inbox, outbox);
+				if (!err) {
+					for (vidx = index * 32; vidx < (index + 1) * 32; ++vidx) {
+						pidx = priv->virt2phys_pkey[slave][port - 1][vidx];
+						outtab[vidx % 32] = cpu_to_be16(table[pidx]);
+					}
+				}
+				kfree(table);
+				return err;
+			}
+			if (smp->attr_id == IB_SMP_ATTR_PORT_INFO) {
+				/*get the slave specific caps:*/
+				/*do the command */
+				err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
+					    vhcr->in_modifier, vhcr->op_modifier,
+					    vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
+				/* modify the response for slaves */
+				if (!err && slave != mlx4_master_func_num(dev)) {
+					u8 *state = outsmp->data + PORT_STATE_OFFSET;
+
+					*state = (*state & 0xf0) | vf_port_state(dev, port, slave);
+					slave_cap_mask = priv->mfunc.master.slave_state[slave].ib_cap_mask[port];
+					memcpy(outsmp->data + PORT_CAPABILITY_LOCATION_IN_SMP, &slave_cap_mask, 4);
+				}
+				return err;
+			}
+			if (smp->attr_id == IB_SMP_ATTR_GUID_INFO) {
+				/* compute slave's gid block */
+				smp->attr_mod = cpu_to_be32(slave / 8);
+				/* execute cmd */
+				err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
+					     vhcr->in_modifier, vhcr->op_modifier,
+					     vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
+				if (!err) {
+					/* if needed, move slave gid to index 0 */
+					if (slave % 8)
+						memcpy(outsmp->data,
+						       outsmp->data + (slave % 8) * 8, 8);
+					/* delete all other gids */
+					memset(outsmp->data + 8, 0, 56);
+				}
+				return err;
+			}
+			if (smp->attr_id == IB_SMP_ATTR_NODE_INFO) {
+				err = mlx4_cmd_box(dev, inbox->dma, outbox->dma,
+					     vhcr->in_modifier, vhcr->op_modifier,
+					     vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
+				if (!err) {
+					slave_node_guid =  mlx4_get_slave_node_guid(dev, slave);
+					memcpy(outsmp->data + 12, &slave_node_guid, 8);
+				}
+				return err;
+			}
+		}
+	}
+	if (slave != mlx4_master_func_num(dev) &&
+	    ((smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) ||
+	     (smp->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED &&
+	      smp->method == IB_MGMT_METHOD_SET))) {
+		mlx4_err(dev, "slave %d is trying to execute a Subnet MGMT MAD, "
+			 "class 0x%x, method 0x%x for attr 0x%x. Rejecting\n",
+			 slave, smp->method, smp->mgmt_class,
+			 be16_to_cpu(smp->attr_id));
+		return -EPERM;
+	}
+	/*default:*/
+	return mlx4_cmd_box(dev, inbox->dma, outbox->dma,
+				    vhcr->in_modifier, vhcr->op_modifier,
+				    vhcr->op, MLX4_CMD_TIME_CLASS_C, MLX4_CMD_NATIVE);
+}
+
 int mlx4_DMA_wrapper(struct mlx4_dev *dev, int slave,
 		     struct mlx4_vhcr *vhcr,
 		     struct mlx4_cmd_mailbox *inbox,
@@ -950,7 +1110,7 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.out_is_imm = false,
 		.encode_slave_id = false,
 		.verify = NULL,
-		.wrapper = mlx4_GEN_QP_wrapper
+		.wrapper = mlx4_INIT2INIT_QP_wrapper
 	},
 	{
 		.opcode = MLX4_CMD_INIT2RTR_QP,
@@ -968,7 +1128,7 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.out_is_imm = false,
 		.encode_slave_id = false,
 		.verify = NULL,
-		.wrapper = mlx4_GEN_QP_wrapper
+		.wrapper = mlx4_RTR2RTS_QP_wrapper
 	},
 	{
 		.opcode = MLX4_CMD_RTS2RTS_QP,
@@ -977,7 +1137,7 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.out_is_imm = false,
 		.encode_slave_id = false,
 		.verify = NULL,
-		.wrapper = mlx4_GEN_QP_wrapper
+		.wrapper = mlx4_RTS2RTS_QP_wrapper
 	},
 	{
 		.opcode = MLX4_CMD_SQERR2RTS_QP,
@@ -986,7 +1146,7 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.out_is_imm = false,
 		.encode_slave_id = false,
 		.verify = NULL,
-		.wrapper = mlx4_GEN_QP_wrapper
+		.wrapper = mlx4_SQERR2RTS_QP_wrapper
 	},
 	{
 		.opcode = MLX4_CMD_2ERR_QP,
@@ -1013,7 +1173,7 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.out_is_imm = false,
 		.encode_slave_id = false,
 		.verify = NULL,
-		.wrapper = mlx4_GEN_QP_wrapper
+		.wrapper = mlx4_SQD2SQD_QP_wrapper
 	},
 	{
 		.opcode = MLX4_CMD_SQD2RTS_QP,
@@ -1022,7 +1182,7 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.out_is_imm = false,
 		.encode_slave_id = false,
 		.verify = NULL,
-		.wrapper = mlx4_GEN_QP_wrapper
+		.wrapper = mlx4_SQD2RTS_QP_wrapper
 	},
 	{
 		.opcode = MLX4_CMD_2RST_QP,
@@ -1061,6 +1221,24 @@ static struct mlx4_cmd_info cmd_info[] = {
 		.wrapper = mlx4_GEN_QP_wrapper
 	},
 	{
+		.opcode = MLX4_CMD_CONF_SPECIAL_QP,
+		.has_inbox = false,
+		.has_outbox = false,
+		.out_is_imm = false,
+		.encode_slave_id = false,
+		.verify = NULL, /* XXX verify: only demux can do this */
+		.wrapper = NULL
+	},
+	{
+		.opcode = MLX4_CMD_MAD_IFC,
+		.has_inbox = true,
+		.has_outbox = true,
+		.out_is_imm = false,
+		.encode_slave_id = false,
+		.verify = NULL,
+		.wrapper = mlx4_MAD_IFC_wrapper
+	},
+	{
 		.opcode = MLX4_CMD_QUERY_IF_STAT,
 		.has_inbox = false,
 		.has_outbox = true,
@@ -1340,6 +1518,8 @@ static void mlx4_master_do_cmd(struct mlx4_dev *dev, int slave, u8 cmd,
 		if (MLX4_COMM_CMD_FLR == slave_state[slave].last_cmd)
 			goto inform_slave_state;
 
+		mlx4_dispatch_event(dev, MLX4_DEV_EVENT_SLAVE_SHUTDOWN, slave);
+
 		/* write the version in the event field */
 		reply |= mlx4_comm_get_version();
 
@@ -1376,19 +1556,21 @@ static void mlx4_master_do_cmd(struct mlx4_dev *dev, int slave, u8 cmd,
 			goto reset_slave;
 		slave_state[slave].vhcr_dma |= param;
 		slave_state[slave].active = true;
+		mlx4_dispatch_event(dev, MLX4_DEV_EVENT_SLAVE_INIT, slave);
 		break;
 	case MLX4_COMM_CMD_VHCR_POST:
 		if ((slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR_EN) &&
 		    (slave_state[slave].last_cmd != MLX4_COMM_CMD_VHCR_POST))
 			goto reset_slave;
-		down(&priv->cmd.slave_sem);
+
+		mutex_lock(&priv->cmd.slave_cmd_mutex);
 		if (mlx4_master_process_vhcr(dev, slave, NULL)) {
 			mlx4_err(dev, "Failed processing vhcr for slave:%d,"
 				 " resetting slave.\n", slave);
-			up(&priv->cmd.slave_sem);
+			mutex_unlock(&priv->cmd.slave_cmd_mutex);
 			goto reset_slave;
 		}
-		up(&priv->cmd.slave_sem);
+		mutex_unlock(&priv->cmd.slave_cmd_mutex);
 		break;
 	default:
 		mlx4_warn(dev, "Bad comm cmd:%d from slave:%d\n", cmd, slave);
@@ -1529,14 +1711,6 @@ int mlx4_multi_func_init(struct mlx4_dev *dev)
 	struct mlx4_slave_state *s_state;
 	int i, j, err, port;
 
-	priv->mfunc.vhcr = dma_alloc_coherent(&(dev->pdev->dev), PAGE_SIZE,
-					    &priv->mfunc.vhcr_dma,
-					    GFP_KERNEL);
-	if (!priv->mfunc.vhcr) {
-		mlx4_err(dev, "Couldn't allocate vhcr.\n");
-		return -ENOMEM;
-	}
-
 	if (mlx4_is_master(dev))
 		priv->mfunc.comm =
 		ioremap(pci_resource_start(dev->pdev, priv->fw.comm_bar) +
@@ -1590,6 +1764,7 @@ int mlx4_multi_func_init(struct mlx4_dev *dev)
 		INIT_WORK(&priv->mfunc.master.slave_flr_event_work,
 			  mlx4_master_handle_slave_flr);
 		spin_lock_init(&priv->mfunc.master.slave_state_lock);
+		spin_lock_init(&priv->mfunc.master.slave_eq.event_lock);
 		priv->mfunc.master.comm_wq =
 			create_singlethread_workqueue("mlx4_comm");
 		if (!priv->mfunc.master.comm_wq)
@@ -1598,7 +1773,6 @@ int mlx4_multi_func_init(struct mlx4_dev *dev)
 		if (mlx4_init_resource_tracker(dev))
 			goto err_thread;
 
-		sema_init(&priv->cmd.slave_sem, 1);
 		err = mlx4_ARM_COMM_CHANNEL(dev);
 		if (err) {
 			mlx4_err(dev, " Failed to arm comm channel eq: %x\n",
@@ -1612,8 +1786,6 @@ int mlx4_multi_func_init(struct mlx4_dev *dev)
 			mlx4_err(dev, "Couldn't sync toggles\n");
 			goto err_comm;
 		}
-
-		sema_init(&priv->cmd.slave_sem, 1);
 	}
 	return 0;
 
@@ -1643,6 +1815,7 @@ int mlx4_cmd_init(struct mlx4_dev *dev)
 	struct mlx4_priv *priv = mlx4_priv(dev);
 
 	mutex_init(&priv->cmd.hcr_mutex);
+	mutex_init(&priv->cmd.slave_cmd_mutex);
 	sema_init(&priv->cmd.poll_sem, 1);
 	priv->cmd.use_events = 0;
 	priv->cmd.toggle     = 1;
@@ -1659,14 +1832,30 @@ int mlx4_cmd_init(struct mlx4_dev *dev)
 		}
 	}
 
+	if (mlx4_is_mfunc(dev)) {
+		priv->mfunc.vhcr = dma_alloc_coherent(&(dev->pdev->dev), PAGE_SIZE,
+						      &priv->mfunc.vhcr_dma,
+						      GFP_KERNEL);
+		if (!priv->mfunc.vhcr) {
+			mlx4_err(dev, "Couldn't allocate VHCR.\n");
+			goto err_hcr;
+		}
+	}
+
 	priv->cmd.pool = pci_pool_create("mlx4_cmd", dev->pdev,
 					 MLX4_MAILBOX_SIZE,
 					 MLX4_MAILBOX_SIZE, 0);
 	if (!priv->cmd.pool)
-		goto err_hcr;
+		goto err_vhcr;
 
 	return 0;
 
+err_vhcr:
+	if (mlx4_is_mfunc(dev))
+		dma_free_coherent(&(dev->pdev->dev), PAGE_SIZE,
+				  priv->mfunc.vhcr, priv->mfunc.vhcr_dma);
+	priv->mfunc.vhcr = NULL;
+
 err_hcr:
 	if (!mlx4_is_slave(dev))
 		iounmap(priv->cmd.hcr);
@@ -1689,9 +1878,6 @@ void mlx4_multi_func_cleanup(struct mlx4_dev *dev)
 	}
 
 	iounmap(priv->mfunc.comm);
-	dma_free_coherent(&(dev->pdev->dev), PAGE_SIZE,
-		     priv->mfunc.vhcr, priv->mfunc.vhcr_dma);
-	priv->mfunc.vhcr = NULL;
 }
 
 void mlx4_cmd_cleanup(struct mlx4_dev *dev)
@@ -1702,6 +1888,10 @@ void mlx4_cmd_cleanup(struct mlx4_dev *dev)
 
 	if (!mlx4_is_slave(dev))
 		iounmap(priv->cmd.hcr);
+	if (mlx4_is_mfunc(dev))
+		dma_free_coherent(&(dev->pdev->dev), PAGE_SIZE,
+				  priv->mfunc.vhcr, priv->mfunc.vhcr_dma);
+	priv->mfunc.vhcr = NULL;
 }
 
 /*
diff --git a/drivers/net/ethernet/mellanox/mlx4/eq.c b/drivers/net/ethernet/mellanox/mlx4/eq.c
index 99a04648fab..51c764901ad 100644
--- a/drivers/net/ethernet/mellanox/mlx4/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/eq.c
@@ -164,13 +164,16 @@ static void slave_event(struct mlx4_dev *dev, u8 slave, struct mlx4_eqe *eqe)
 {
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	struct mlx4_slave_event_eq *slave_eq = &priv->mfunc.master.slave_eq;
-	struct mlx4_eqe *s_eqe =
-		&slave_eq->event_eqe[slave_eq->prod & (SLAVE_EVENT_EQ_SIZE - 1)];
+	struct mlx4_eqe *s_eqe;
+	unsigned long flags;
 
+	spin_lock_irqsave(&slave_eq->event_lock, flags);
+	s_eqe = &slave_eq->event_eqe[slave_eq->prod & (SLAVE_EVENT_EQ_SIZE - 1)];
 	if ((!!(s_eqe->owner & 0x80)) ^
 	    (!!(slave_eq->prod & SLAVE_EVENT_EQ_SIZE))) {
 		mlx4_warn(dev, "Master failed to generate an EQE for slave: %d. "
 			  "No free EQE on slave events queue\n", slave);
+		spin_unlock_irqrestore(&slave_eq->event_lock, flags);
 		return;
 	}
 
@@ -183,6 +186,7 @@ static void slave_event(struct mlx4_dev *dev, u8 slave, struct mlx4_eqe *eqe)
 
 	queue_work(priv->mfunc.master.comm_wq,
 		   &priv->mfunc.master.slave_event_work);
+	spin_unlock_irqrestore(&slave_eq->event_lock, flags);
 }
 
 static void mlx4_slave_event(struct mlx4_dev *dev, int slave,
@@ -200,6 +204,196 @@ static void mlx4_slave_event(struct mlx4_dev *dev, int slave,
 	slave_event(dev, slave, eqe);
 }
 
+int mlx4_gen_pkey_eqe(struct mlx4_dev *dev, int slave, u8 port)
+{
+	struct mlx4_eqe eqe;
+
+	struct mlx4_priv *priv = mlx4_priv(dev);
+	struct mlx4_slave_state *s_slave = &priv->mfunc.master.slave_state[slave];
+
+	if (!s_slave->active)
+		return 0;
+
+	memset(&eqe, 0, sizeof eqe);
+
+	eqe.type = MLX4_EVENT_TYPE_PORT_MNG_CHG_EVENT;
+	eqe.subtype = MLX4_DEV_PMC_SUBTYPE_PKEY_TABLE;
+	eqe.event.port_mgmt_change.port = port;
+
+	return mlx4_GEN_EQE(dev, slave, &eqe);
+}
+EXPORT_SYMBOL(mlx4_gen_pkey_eqe);
+
+int mlx4_gen_guid_change_eqe(struct mlx4_dev *dev, int slave, u8 port)
+{
+	struct mlx4_eqe eqe;
+
+	/*don't send if we don't have the that slave */
+	if (dev->num_vfs < slave)
+		return 0;
+	memset(&eqe, 0, sizeof eqe);
+
+	eqe.type = MLX4_EVENT_TYPE_PORT_MNG_CHG_EVENT;
+	eqe.subtype = MLX4_DEV_PMC_SUBTYPE_GUID_INFO;
+	eqe.event.port_mgmt_change.port = port;
+
+	return mlx4_GEN_EQE(dev, slave, &eqe);
+}
+EXPORT_SYMBOL(mlx4_gen_guid_change_eqe);
+
+int mlx4_gen_port_state_change_eqe(struct mlx4_dev *dev, int slave, u8 port,
+				   u8 port_subtype_change)
+{
+	struct mlx4_eqe eqe;
+
+	/*don't send if we don't have the that slave */
+	if (dev->num_vfs < slave)
+		return 0;
+	memset(&eqe, 0, sizeof eqe);
+
+	eqe.type = MLX4_EVENT_TYPE_PORT_CHANGE;
+	eqe.subtype = port_subtype_change;
+	eqe.event.port_change.port = cpu_to_be32(port << 28);
+
+	mlx4_dbg(dev, "%s: sending: %d to slave: %d on port: %d\n", __func__,
+		 port_subtype_change, slave, port);
+	return mlx4_GEN_EQE(dev, slave, &eqe);
+}
+EXPORT_SYMBOL(mlx4_gen_port_state_change_eqe);
+
+enum slave_port_state mlx4_get_slave_port_state(struct mlx4_dev *dev, int slave, u8 port)
+{
+	struct mlx4_priv *priv = mlx4_priv(dev);
+	struct mlx4_slave_state *s_state = priv->mfunc.master.slave_state;
+	if (slave >= dev->num_slaves || port > MLX4_MAX_PORTS) {
+		pr_err("%s: Error: asking for slave:%d, port:%d\n",
+		       __func__, slave, port);
+		return SLAVE_PORT_DOWN;
+	}
+	return s_state[slave].port_state[port];
+}
+EXPORT_SYMBOL(mlx4_get_slave_port_state);
+
+static int mlx4_set_slave_port_state(struct mlx4_dev *dev, int slave, u8 port,
+				     enum slave_port_state state)
+{
+	struct mlx4_priv *priv = mlx4_priv(dev);
+	struct mlx4_slave_state *s_state = priv->mfunc.master.slave_state;
+
+	if (slave >= dev->num_slaves || port > MLX4_MAX_PORTS || port == 0) {
+		pr_err("%s: Error: asking for slave:%d, port:%d\n",
+		       __func__, slave, port);
+		return -1;
+	}
+	s_state[slave].port_state[port] = state;
+
+	return 0;
+}
+
+static void set_all_slave_state(struct mlx4_dev *dev, u8 port, int event)
+{
+	int i;
+	enum slave_port_gen_event gen_event;
+
+	for (i = 0; i < dev->num_slaves; i++)
+		set_and_calc_slave_port_state(dev, i, port, event, &gen_event);
+}
+/**************************************************************************
+	The function get as input the new event to that port,
+	and according to the prev state change the slave's port state.
+	The events are:
+		MLX4_PORT_STATE_DEV_EVENT_PORT_DOWN,
+		MLX4_PORT_STATE_DEV_EVENT_PORT_UP
+		MLX4_PORT_STATE_IB_EVENT_GID_VALID
+		MLX4_PORT_STATE_IB_EVENT_GID_INVALID
+***************************************************************************/
+int set_and_calc_slave_port_state(struct mlx4_dev *dev, int slave,
+				  u8 port, int event,
+				  enum slave_port_gen_event *gen_event)
+{
+	struct mlx4_priv *priv = mlx4_priv(dev);
+	struct mlx4_slave_state *ctx = NULL;
+	unsigned long flags;
+	int ret = -1;
+	enum slave_port_state cur_state =
+		mlx4_get_slave_port_state(dev, slave, port);
+
+	*gen_event = SLAVE_PORT_GEN_EVENT_NONE;
+
+	if (slave >= dev->num_slaves || port > MLX4_MAX_PORTS || port == 0) {
+		pr_err("%s: Error: asking for slave:%d, port:%d\n",
+		       __func__, slave, port);
+		return ret;
+	}
+
+	ctx = &priv->mfunc.master.slave_state[slave];
+	spin_lock_irqsave(&ctx->lock, flags);
+
+	mlx4_dbg(dev, "%s: slave: %d, current state: %d new event :%d\n",
+		 __func__, slave, cur_state, event);
+
+	switch (cur_state) {
+	case SLAVE_PORT_DOWN:
+		if (MLX4_PORT_STATE_DEV_EVENT_PORT_UP == event)
+			mlx4_set_slave_port_state(dev, slave, port,
+						  SLAVE_PENDING_UP);
+		break;
+	case SLAVE_PENDING_UP:
+		if (MLX4_PORT_STATE_DEV_EVENT_PORT_DOWN == event)
+			mlx4_set_slave_port_state(dev, slave, port,
+						  SLAVE_PORT_DOWN);
+		else if (MLX4_PORT_STATE_IB_PORT_STATE_EVENT_GID_VALID == event) {
+			mlx4_set_slave_port_state(dev, slave, port,
+						  SLAVE_PORT_UP);
+			*gen_event = SLAVE_PORT_GEN_EVENT_UP;
+		}
+		break;
+	case SLAVE_PORT_UP:
+		if (MLX4_PORT_STATE_DEV_EVENT_PORT_DOWN == event) {
+			mlx4_set_slave_port_state(dev, slave, port,
+						  SLAVE_PORT_DOWN);
+			*gen_event = SLAVE_PORT_GEN_EVENT_DOWN;
+		} else if (MLX4_PORT_STATE_IB_EVENT_GID_INVALID ==
+				event) {
+			mlx4_set_slave_port_state(dev, slave, port,
+						  SLAVE_PENDING_UP);
+			*gen_event = SLAVE_PORT_GEN_EVENT_DOWN;
+		}
+		break;
+	default:
+		pr_err("%s: BUG!!! UNKNOWN state: "
+		       "slave:%d, port:%d\n", __func__, slave, port);
+			goto out;
+	}
+	ret = mlx4_get_slave_port_state(dev, slave, port);
+	mlx4_dbg(dev, "%s: slave: %d, current state: %d new event"
+		 " :%d gen_event: %d\n",
+		 __func__, slave, cur_state, event, *gen_event);
+
+out:
+	spin_unlock_irqrestore(&ctx->lock, flags);
+	return ret;
+}
+
+EXPORT_SYMBOL(set_and_calc_slave_port_state);
+
+int mlx4_gen_slaves_port_mgt_ev(struct mlx4_dev *dev, u8 port, int attr)
+{
+	struct mlx4_eqe eqe;
+
+	memset(&eqe, 0, sizeof eqe);
+
+	eqe.type = MLX4_EVENT_TYPE_PORT_MNG_CHG_EVENT;
+	eqe.subtype = MLX4_DEV_PMC_SUBTYPE_PORT_INFO;
+	eqe.event.port_mgmt_change.port = port;
+	eqe.event.port_mgmt_change.params.port_info.changed_attr =
+		cpu_to_be32((u32) attr);
+
+	slave_event(dev, ALL_SLAVES, &eqe);
+	return 0;
+}
+EXPORT_SYMBOL(mlx4_gen_slaves_port_mgt_ev);
+
 void mlx4_master_handle_slave_flr(struct work_struct *work)
 {
 	struct mlx4_mfunc_master_ctx *master =
@@ -251,6 +445,7 @@ static int mlx4_eq_int(struct mlx4_dev *dev, struct mlx4_eq *eq)
 	u32 flr_slave;
 	u8 update_slave_state;
 	int i;
+	enum slave_port_gen_event gen_event;
 
 	while ((eqe = next_eqe_sw(eq))) {
 		/*
@@ -347,35 +542,49 @@ static int mlx4_eq_int(struct mlx4_dev *dev, struct mlx4_eq *eq)
 		case MLX4_EVENT_TYPE_PORT_CHANGE:
 			port = be32_to_cpu(eqe->event.port_change.port) >> 28;
 			if (eqe->subtype == MLX4_PORT_CHANGE_SUBTYPE_DOWN) {
-				mlx4_dispatch_event(dev,
-						    MLX4_DEV_EVENT_PORT_DOWN,
+				mlx4_dispatch_event(dev, MLX4_DEV_EVENT_PORT_DOWN,
 						    port);
 				mlx4_priv(dev)->sense.do_sense_port[port] = 1;
-				if (mlx4_is_master(dev))
-					/*change the state of all slave's port
-					* to down:*/
-					for (i = 0; i < dev->num_slaves; i++) {
-						mlx4_dbg(dev, "%s: Sending "
-							 "MLX4_PORT_CHANGE_SUBTYPE_DOWN"
+				if (!mlx4_is_master(dev))
+					break;
+				for (i = 0; i < dev->num_slaves; i++) {
+					if (dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH) {
+						if (i == mlx4_master_func_num(dev))
+							continue;
+						mlx4_dbg(dev, "%s: Sending MLX4_PORT_CHANGE_SUBTYPE_DOWN"
 							 " to slave: %d, port:%d\n",
 							 __func__, i, port);
-						if (i == dev->caps.function)
-							continue;
 						mlx4_slave_event(dev, i, eqe);
+					} else {  /* IB port */
+						set_and_calc_slave_port_state(dev, i, port,
+									      MLX4_PORT_STATE_DEV_EVENT_PORT_DOWN,
+									      &gen_event);
+						/*we can be in pending state, then do not send port_down event*/
+						if (SLAVE_PORT_GEN_EVENT_DOWN ==  gen_event) {
+							if (i == mlx4_master_func_num(dev))
+								continue;
+							mlx4_slave_event(dev, i, eqe);
+						}
 					}
+				}
 			} else {
-				mlx4_dispatch_event(dev,
-						    MLX4_DEV_EVENT_PORT_UP,
-						    port);
+				mlx4_dispatch_event(dev, MLX4_DEV_EVENT_PORT_UP, port);
+
 				mlx4_priv(dev)->sense.do_sense_port[port] = 0;
 
-				if (mlx4_is_master(dev)) {
+				if (!mlx4_is_master(dev))
+					break;
+				if (dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH)
 					for (i = 0; i < dev->num_slaves; i++) {
-						if (i == dev->caps.function)
+						if (i == mlx4_master_func_num(dev))
 							continue;
 						mlx4_slave_event(dev, i, eqe);
 					}
-				}
+				else /* IB port */
+					/* port-up event will be sent to a slave when the
+					 * slave's alias-guid is set. This is done in alias_GUID.c
+					 */
+					set_all_slave_state(dev, port, MLX4_DEV_EVENT_PORT_UP);
 			}
 			break;
 
diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c
index c6964848732..4f30b99324c 100644
--- a/drivers/net/ethernet/mellanox/mlx4/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx4/fw.c
@@ -183,7 +183,7 @@ int mlx4_QUERY_FUNC_CAP_wrapper(struct mlx4_dev *dev, int slave,
 #define QUERY_FUNC_CAP_MTT_QUOTA_OFFSET		0x24
 #define QUERY_FUNC_CAP_MCG_QUOTA_OFFSET		0x28
 #define QUERY_FUNC_CAP_MAX_EQ_OFFSET		0x2c
-#define QUERY_FUNC_CAP_RESERVED_EQ_OFFSET	0X30
+#define QUERY_FUNC_CAP_RESERVED_EQ_OFFSET	0x30
 
 #define QUERY_FUNC_CAP_FMR_FLAG			0x80
 #define QUERY_FUNC_CAP_FLAG_RDMA		0x40
@@ -194,21 +194,39 @@ int mlx4_QUERY_FUNC_CAP_wrapper(struct mlx4_dev *dev, int slave,
 #define QUERY_FUNC_CAP_RDMA_PROPS_OFFSET	0x8
 #define QUERY_FUNC_CAP_ETH_PROPS_OFFSET		0xc
 
+#define QUERY_FUNC_CAP_QP0_TUNNEL		0x10
+#define QUERY_FUNC_CAP_QP0_PROXY		0x14
+#define QUERY_FUNC_CAP_QP1_TUNNEL		0x18
+#define QUERY_FUNC_CAP_QP1_PROXY		0x1c
+
 #define QUERY_FUNC_CAP_ETH_PROPS_FORCE_MAC	0x40
 #define QUERY_FUNC_CAP_ETH_PROPS_FORCE_VLAN	0x80
 
 #define QUERY_FUNC_CAP_RDMA_PROPS_FORCE_PHY_WQE_GID 0x80
 
 	if (vhcr->op_modifier == 1) {
-		field = vhcr->in_modifier;
-		MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_PHYS_PORT_OFFSET);
-
 		field = 0;
 		/* ensure force vlan and force mac bits are not set */
 		MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_ETH_PROPS_OFFSET);
 		/* ensure that phy_wqe_gid bit is not set */
 		MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_RDMA_PROPS_OFFSET);
 
+		field = vhcr->in_modifier; /* phys-port = logical-port */
+		MLX4_PUT(outbox->buf, field, QUERY_FUNC_CAP_PHYS_PORT_OFFSET);
+
+		/* size is now the QP number */
+		size = dev->phys_caps.base_tunnel_sqpn + 8 * slave + field - 1;
+		MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP0_TUNNEL);
+
+		size += 2;
+		MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP1_TUNNEL);
+
+		size = dev->phys_caps.base_proxy_sqpn + 8 * slave + field - 1;
+		MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP0_PROXY);
+
+		size += 2;
+		MLX4_PUT(outbox->buf, size, QUERY_FUNC_CAP_QP1_PROXY);
+
 	} else if (vhcr->op_modifier == 0) {
 		/* enable rdma and ethernet interfaces */
 		field = (QUERY_FUNC_CAP_FLAG_ETH | QUERY_FUNC_CAP_FLAG_RDMA);
@@ -253,99 +271,118 @@ int mlx4_QUERY_FUNC_CAP_wrapper(struct mlx4_dev *dev, int slave,
 	return err;
 }
 
-int mlx4_QUERY_FUNC_CAP(struct mlx4_dev *dev, struct mlx4_func_cap *func_cap)
+int mlx4_QUERY_FUNC_CAP(struct mlx4_dev *dev, u32 gen_or_port,
+			struct mlx4_func_cap *func_cap)
 {
 	struct mlx4_cmd_mailbox *mailbox;
 	u32			*outbox;
-	u8			field;
+	u8			field, op_modifier;
 	u32			size;
-	int			i;
 	int			err = 0;
 
+	op_modifier = !!gen_or_port; /* 0 = general, 1 = logical port */
 
 	mailbox = mlx4_alloc_cmd_mailbox(dev);
 	if (IS_ERR(mailbox))
 		return PTR_ERR(mailbox);
 
-	err = mlx4_cmd_box(dev, 0, mailbox->dma, 0, 0, MLX4_CMD_QUERY_FUNC_CAP,
+	err = mlx4_cmd_box(dev, 0, mailbox->dma, gen_or_port, op_modifier,
+			   MLX4_CMD_QUERY_FUNC_CAP,
 			   MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
 	if (err)
 		goto out;
 
 	outbox = mailbox->buf;
 
-	MLX4_GET(field, outbox, QUERY_FUNC_CAP_FLAGS_OFFSET);
-	if (!(field & (QUERY_FUNC_CAP_FLAG_ETH | QUERY_FUNC_CAP_FLAG_RDMA))) {
-		mlx4_err(dev, "The host supports neither eth nor rdma interfaces\n");
-		err = -EPROTONOSUPPORT;
-		goto out;
-	}
-	func_cap->flags = field;
+	if (!op_modifier) {
+		MLX4_GET(field, outbox, QUERY_FUNC_CAP_FLAGS_OFFSET);
+		if (!(field & (QUERY_FUNC_CAP_FLAG_ETH | QUERY_FUNC_CAP_FLAG_RDMA))) {
+			mlx4_err(dev, "The host supports neither eth nor rdma interfaces\n");
+			err = -EPROTONOSUPPORT;
+			goto out;
+		}
+		func_cap->flags = field;
+
+		MLX4_GET(field, outbox, QUERY_FUNC_CAP_NUM_PORTS_OFFSET);
+		func_cap->num_ports = field;
 
-	MLX4_GET(field, outbox, QUERY_FUNC_CAP_NUM_PORTS_OFFSET);
-	func_cap->num_ports = field;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_PF_BHVR_OFFSET);
+		func_cap->pf_context_behaviour = size;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_PF_BHVR_OFFSET);
-	func_cap->pf_context_behaviour = size;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP_QUOTA_OFFSET);
+		func_cap->qp_quota = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP_QUOTA_OFFSET);
-	func_cap->qp_quota = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET);
+		func_cap->srq_quota = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_SRQ_QUOTA_OFFSET);
-	func_cap->srq_quota = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_CQ_QUOTA_OFFSET);
+		func_cap->cq_quota = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_CQ_QUOTA_OFFSET);
-	func_cap->cq_quota = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_MAX_EQ_OFFSET);
+		func_cap->max_eq = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_MAX_EQ_OFFSET);
-	func_cap->max_eq = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_RESERVED_EQ_OFFSET);
+		func_cap->reserved_eq = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_RESERVED_EQ_OFFSET);
-	func_cap->reserved_eq = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_MPT_QUOTA_OFFSET);
+		func_cap->mpt_quota = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_MPT_QUOTA_OFFSET);
-	func_cap->mpt_quota = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_MTT_QUOTA_OFFSET);
+		func_cap->mtt_quota = size & 0xFFFFFF;
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_MTT_QUOTA_OFFSET);
-	func_cap->mtt_quota = size & 0xFFFFFF;
+		MLX4_GET(size, outbox, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET);
+		func_cap->mcg_quota = size & 0xFFFFFF;
+		goto out;
+	}
 
-	MLX4_GET(size, outbox, QUERY_FUNC_CAP_MCG_QUOTA_OFFSET);
-	func_cap->mcg_quota = size & 0xFFFFFF;
+	/* logical port query */
+	if (gen_or_port > dev->caps.num_ports) {
+		err = -EINVAL;
+		goto out;
+	}
 
-	for (i = 1; i <= func_cap->num_ports; ++i) {
-		err = mlx4_cmd_box(dev, 0, mailbox->dma, i, 1,
-				   MLX4_CMD_QUERY_FUNC_CAP,
-				   MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
-		if (err)
+	if (dev->caps.port_type[gen_or_port] == MLX4_PORT_TYPE_ETH) {
+		MLX4_GET(field, outbox, QUERY_FUNC_CAP_ETH_PROPS_OFFSET);
+		if (field & QUERY_FUNC_CAP_ETH_PROPS_FORCE_VLAN) {
+			mlx4_err(dev, "VLAN is enforced on this port\n");
+			err = -EPROTONOSUPPORT;
 			goto out;
+		}
 
-		if (dev->caps.port_type[i] == MLX4_PORT_TYPE_ETH) {
-			MLX4_GET(field, outbox, QUERY_FUNC_CAP_ETH_PROPS_OFFSET);
-			if (field & QUERY_FUNC_CAP_ETH_PROPS_FORCE_VLAN) {
-				mlx4_err(dev, "VLAN is enforced on this port\n");
-				err = -EPROTONOSUPPORT;
-				goto out;
-			}
-
-			if (field & QUERY_FUNC_CAP_ETH_PROPS_FORCE_MAC) {
-				mlx4_err(dev, "Force mac is enabled on this port\n");
-				err = -EPROTONOSUPPORT;
-				goto out;
-			}
-		} else if (dev->caps.port_type[i] == MLX4_PORT_TYPE_IB) {
-			MLX4_GET(field, outbox, QUERY_FUNC_CAP_RDMA_PROPS_OFFSET);
-			if (field & QUERY_FUNC_CAP_RDMA_PROPS_FORCE_PHY_WQE_GID) {
-				mlx4_err(dev, "phy_wqe_gid is "
-					 "enforced on this ib port\n");
-				err = -EPROTONOSUPPORT;
-				goto out;
-			}
+		if (field & QUERY_FUNC_CAP_ETH_PROPS_FORCE_MAC) {
+			mlx4_err(dev, "Force mac is enabled on this port\n");
+			err = -EPROTONOSUPPORT;
+			goto out;
 		}
+	} else if (dev->caps.port_type[gen_or_port] == MLX4_PORT_TYPE_IB) {
+		MLX4_GET(field, outbox, QUERY_FUNC_CAP_RDMA_PROPS_OFFSET);
+		if (field & QUERY_FUNC_CAP_RDMA_PROPS_FORCE_PHY_WQE_GID) {
+			mlx4_err(dev, "phy_wqe_gid is "
+				 "enforced on this ib port\n");
+			err = -EPROTONOSUPPORT;
+			goto out;
+		}
+	}
 
-		MLX4_GET(field, outbox, QUERY_FUNC_CAP_PHYS_PORT_OFFSET);
-		func_cap->physical_port[i] = field;
+	MLX4_GET(field, outbox, QUERY_FUNC_CAP_PHYS_PORT_OFFSET);
+	func_cap->physical_port = field;
+	if (func_cap->physical_port != gen_or_port) {
+		err = -ENOSYS;
+		goto out;
 	}
 
+	MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP0_TUNNEL);
+	func_cap->qp0_tunnel_qpn = size & 0xFFFFFF;
+
+	MLX4_GET(size, outbox, QUERY_FUNC_CAP_QP0_PROXY);
+	func_cap->qp0_proxy_qpn