Merge branch 'mr_fix' into git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma for-next

Update mlx4 to support user MR creation against read-only memory, previously it required the memory to be writable. Based on rdma for-rc due to dependencies. * mr_fix: (2 commits) IB/mlx4: Mark user MR as writable if actual virtual memory is writable IB/core: Make testing MR flags for writability a static inline function
2018-05-28 11:44:35 -06:00 · 2018-05-28 11:44:35 -06:00 · 0394808d9e
parent bb42f87e29 d8f9cc328c
commit 0394808d9e
31 changed files with 174 additions and 117 deletions
--- a/6
+++ b/6
@ -5391,7 +5391,6 @@ S:	Maintained
 F:	drivers/iommu/exynos-iommu.c

 EZchip NPS platform support
-M:	Elad Kanfi <eladkan@mellanox.com>
 M:	Vineet Gupta <vgupta@synopsys.com>
 S:	Supported
 F:	arch/arc/plat-eznps
@ -9012,7 +9011,6 @@ Q:	http://patchwork.ozlabs.org/project/netdev/list/
 F:	drivers/net/ethernet/mellanox/mlx5/core/en_*

 MELLANOX ETHERNET INNOVA DRIVER
-M:	Ilan Tayari <ilant@mellanox.com>
 R:	Boris Pismenny <borisp@mellanox.com>
 L:	netdev@vger.kernel.org
 S:	Supported
@ -9022,7 +9020,6 @@ F:	drivers/net/ethernet/mellanox/mlx5/core/fpga/*
 F:	include/linux/mlx5/mlx5_ifc_fpga.h

 MELLANOX ETHERNET INNOVA IPSEC DRIVER
-M:	Ilan Tayari <ilant@mellanox.com>
 R:	Boris Pismenny <borisp@mellanox.com>
 L:	netdev@vger.kernel.org
 S:	Supported
@ -9078,7 +9075,6 @@ F:	include/uapi/rdma/mlx4-abi.h

 MELLANOX MLX5 core VPI driver
 M:	Saeed Mahameed <saeedm@mellanox.com>
-M:	Matan Barak <matanb@mellanox.com>
 M:	Leon Romanovsky <leonro@mellanox.com>
 L:	netdev@vger.kernel.org
 L:	linux-rdma@vger.kernel.org
@ -9089,7 +9085,6 @@ F:	drivers/net/ethernet/mellanox/mlx5/core/
 F:	include/linux/mlx5/

 MELLANOX MLX5 IB driver
-M:	Matan Barak <matanb@mellanox.com>
 M:	Leon Romanovsky <leonro@mellanox.com>
 L:	linux-rdma@vger.kernel.org
 W:	http://www.mellanox.com
@ -9821,7 +9816,6 @@ F:	net/netfilter/xt_CONNSECMARK.c
 F:	net/netfilter/xt_SECMARK.c

 NETWORKING [TLS]
-M:	Ilya Lesokhin <ilyal@mellanox.com>
 M:	Aviad Yehezkel <aviadye@mellanox.com>
 M:	Dave Watson <davejwatson@fb.com>
 L:	netdev@vger.kernel.org
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@ -117,20 +117,9 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 	umem->length     = size;
 	umem->address    = addr;
 	umem->page_shift = PAGE_SHIFT;
-	umem->pid	 = get_task_pid(current, PIDTYPE_PID);
-	/*
-	 * We ask for writable memory if any of the following
-	 * access flags are set.  "Local write" and "remote write"
-	 * obviously require write access.  "Remote atomic" can do
-	 * things like fetch and add, which will modify memory, and
-	 * "MW bind" can change permissions by binding a window.
-	 */
-	umem->writable  = !!(access &
-		(IB_ACCESS_LOCAL_WRITE   | IB_ACCESS_REMOTE_WRITE |
-		 IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_MW_BIND));
+	umem->writable   = ib_access_writable(access);

 	if (access & IB_ACCESS_ON_DEMAND) {
-		put_pid(umem->pid);
 		ret = ib_umem_odp_get(context, umem, access);
 		if (ret) {
 			kfree(umem);
@ -146,7 +135,6 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,

 	page_list = (struct page **) __get_free_page(GFP_KERNEL);
 	if (!page_list) {
-		put_pid(umem->pid);
 		kfree(umem);
 		return ERR_PTR(-ENOMEM);
 	}
@ -229,7 +217,6 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 	if (ret < 0) {
 		if (need_release)
 			__ib_umem_release(context->device, umem, 0);
-		put_pid(umem->pid);
 		kfree(umem);
 	} else
 		current->mm->pinned_vm = locked;
@ -272,8 +259,7 @@ void ib_umem_release(struct ib_umem *umem)

 	__ib_umem_release(umem->context->device, umem, 1);

-	task = get_pid_task(umem->pid, PIDTYPE_PID);
-	put_pid(umem->pid);
+	task = get_pid_task(umem->context->tgid, PIDTYPE_PID);
 	if (!task)
 		goto out;
 	mm = get_task_mm(task);
--- a/drivers/infiniband/hw/cxgb4/mem.c
+++ b/drivers/infiniband/hw/cxgb4/mem.c
@ -489,10 +489,10 @@ struct ib_mr *c4iw_get_dma_mr(struct ib_pd *pd, int acc)
 err_dereg_mem:
 	dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size,
 		  mhp->attr.pbl_addr, mhp->dereg_skb, mhp->wr_waitp);
-err_free_wr_wait:
-	c4iw_put_wr_wait(mhp->wr_waitp);
 err_free_skb:
 	kfree_skb(mhp->dereg_skb);
+err_free_wr_wait:
+	c4iw_put_wr_wait(mhp->wr_waitp);
 err_free_mhp:
 	kfree(mhp);
 	return ERR_PTR(ret);
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@ -5945,6 +5945,7 @@ static void is_sendctxt_err_int(struct hfi1_devdata *dd,
 	u64 status;
 	u32 sw_index;
 	int i = 0;
+	unsigned long irq_flags;

 	sw_index = dd->hw_to_sw[hw_context];
 	if (sw_index >= dd->num_send_contexts) {
@ -5954,10 +5955,12 @@ static void is_sendctxt_err_int(struct hfi1_devdata *dd,
 		return;
 	}
 	sci = &dd->send_contexts[sw_index];
+	spin_lock_irqsave(&dd->sc_lock, irq_flags);
 	sc = sci->sc;
 	if (!sc) {
 		dd_dev_err(dd, "%s: context %u(%u): no sc?\n", __func__,
 			   sw_index, hw_context);
+		spin_unlock_irqrestore(&dd->sc_lock, irq_flags);
 		return;
 	}

@ -5979,6 +5982,7 @@ static void is_sendctxt_err_int(struct hfi1_devdata *dd,
 	 */
 	if (sc->type != SC_USER)
 		queue_work(dd->pport->hfi1_wq, &sc->halt_work);
+	spin_unlock_irqrestore(&dd->sc_lock, irq_flags);

 	/*
 	 * Update the counters for the corresponding status bits.
--- a/drivers/infiniband/hw/hns/hns_roce_cq.c
+++ b/drivers/infiniband/hw/hns/hns_roce_cq.c
@ -377,6 +377,7 @@ struct ib_cq *hns_roce_ib_create_cq(struct ib_device *ib_dev,

 			hr_cq->set_ci_db = hr_cq->db.db_record;
 			*hr_cq->set_ci_db = 0;
+			hr_cq->db_en = 1;
 		}

 		/* Init mmt table and write buff address to mtt table */
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@ -722,6 +722,7 @@ static int hns_roce_v1_rsv_lp_qp(struct hns_roce_dev *hr_dev)
 	free_mr->mr_free_pd = to_hr_pd(pd);
 	free_mr->mr_free_pd->ibpd.device  = &hr_dev->ib_dev;
 	free_mr->mr_free_pd->ibpd.uobject = NULL;
+	free_mr->mr_free_pd->ibpd.__internal_mr = NULL;
 	atomic_set(&free_mr->mr_free_pd->ibpd.usecnt, 0);

 	attr.qp_access_flags	= IB_ACCESS_REMOTE_WRITE;
@ -1036,7 +1037,7 @@ static void hns_roce_v1_mr_free_work_fn(struct work_struct *work)

 	do {
 		ret = hns_roce_v1_poll_cq(&mr_free_cq->ib_cq, ne, wc);
-		if (ret < 0) {
+		if (ret < 0 && hr_qp) {
 			dev_err(dev,
 			   "(qp:0x%lx) starts, Poll cqe failed(%d) for mr 0x%x free! Remain %d cqe\n",
 			   hr_qp->qpn, ret, hr_mr->key, ne);
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@ -142,8 +142,8 @@ static int hns_roce_v2_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	unsigned long flags;
 	unsigned int ind;
 	void *wqe = NULL;
-	u32 tmp_len = 0;
 	bool loopback;
+	u32 tmp_len;
 	int ret = 0;
 	u8 *smac;
 	int nreq;
@ -189,6 +189,7 @@ static int hns_roce_v2_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,

 		owner_bit =
 		       ~(((qp->sq.head + nreq) >> ilog2(qp->sq.wqe_cnt)) & 0x1);
+		tmp_len = 0;

 		/* Corresponding to the QP type, wqe process separately */
 		if (ibqp->qp_type == IB_QPT_GSI) {
@ -547,16 +548,20 @@ static int hns_roce_v2_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 		}

 		if (i < hr_qp->rq.max_gs) {
-			dseg[i].lkey = cpu_to_le32(HNS_ROCE_INVALID_LKEY);
-			dseg[i].addr = 0;
+			dseg->lkey = cpu_to_le32(HNS_ROCE_INVALID_LKEY);
+			dseg->addr = 0;
 		}

 		/* rq support inline data */
-		sge_list = hr_qp->rq_inl_buf.wqe_list[ind].sg_list;
-		hr_qp->rq_inl_buf.wqe_list[ind].sge_cnt = (u32)wr->num_sge;
-		for (i = 0; i < wr->num_sge; i++) {
-			sge_list[i].addr = (void *)(u64)wr->sg_list[i].addr;
-			sge_list[i].len = wr->sg_list[i].length;
+		if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) {
+			sge_list = hr_qp->rq_inl_buf.wqe_list[ind].sg_list;
+			hr_qp->rq_inl_buf.wqe_list[ind].sge_cnt =
+							       (u32)wr->num_sge;
+			for (i = 0; i < wr->num_sge; i++) {
+				sge_list[i].addr =
+					       (void *)(u64)wr->sg_list[i].addr;
+				sge_list[i].len = wr->sg_list[i].length;
+			}
 		}

 		hr_qp->rq.wrid[ind] = wr->wr_id;
@ -613,6 +618,8 @@ static void hns_roce_free_cmq_desc(struct hns_roce_dev *hr_dev,
 	dma_unmap_single(hr_dev->dev, ring->desc_dma_addr,
 			 ring->desc_num * sizeof(struct hns_roce_cmq_desc),
 			 DMA_BIDIRECTIONAL);
+
+	ring->desc_dma_addr = 0;
 	kfree(ring->desc);
 }

@ -1081,6 +1088,7 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 	if (ret) {
 		dev_err(hr_dev->dev, "Configure global param fail, ret = %d.\n",
 			ret);
+		return ret;
 	}

 	/* Get pf resource owned by every pf */
@ -1373,6 +1381,8 @@ static int hns_roce_v2_write_mtpt(void *mb_buf, struct hns_roce_mr *mr,

 	roce_set_bit(mpt_entry->byte_12_mw_pa, V2_MPT_BYTE_12_PA_S,
 		     mr->type == MR_TYPE_MR ? 0 : 1);
+	roce_set_bit(mpt_entry->byte_12_mw_pa, V2_MPT_BYTE_12_INNER_PA_VLD_S,
+		     1);
 	mpt_entry->byte_12_mw_pa = cpu_to_le32(mpt_entry->byte_12_mw_pa);

 	mpt_entry->len_l = cpu_to_le32(lower_32_bits(mr->size));
@ -2171,6 +2181,7 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp,
 				    struct hns_roce_v2_qp_context *context,
 				    struct hns_roce_v2_qp_context *qpc_mask)
 {
+	struct hns_roce_dev *hr_dev = to_hr_dev(ibqp->device);
 	struct hns_roce_qp *hr_qp = to_hr_qp(ibqp);

 	/*
@ -2283,7 +2294,8 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp,
 	context->rq_db_record_addr = hr_qp->rdb.dma >> 32;
 	qpc_mask->rq_db_record_addr = 0;

-	roce_set_bit(context->byte_76_srqn_op_en, V2_QPC_BYTE_76_RQIE_S, 1);
+	roce_set_bit(context->byte_76_srqn_op_en, V2_QPC_BYTE_76_RQIE_S,
+		    (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_RQ_INLINE) ? 1 : 0);
 	roce_set_bit(qpc_mask->byte_76_srqn_op_en, V2_QPC_BYTE_76_RQIE_S, 0);

 	roce_set_field(context->byte_80_rnr_rx_cqn, V2_QPC_BYTE_80_RX_CQN_M,
@ -4707,6 +4719,8 @@ static const struct pci_device_id hns_roce_hw_v2_pci_tbl[] = {
 	{0, }
 };

+MODULE_DEVICE_TABLE(pci, hns_roce_hw_v2_pci_tbl);
+
 static int hns_roce_hw_v2_get_cfg(struct hns_roce_dev *hr_dev,
 				  struct hnae3_handle *handle)
 {
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@ -198,7 +198,7 @@ static int hns_roce_query_device(struct ib_device *ib_dev,

 	memset(props, 0, sizeof(*props));

-	props->sys_image_guid = cpu_to_be32(hr_dev->sys_image_guid);
+	props->sys_image_guid = cpu_to_be64(hr_dev->sys_image_guid);
 	props->max_mr_size = (u64)(~(0ULL));
 	props->page_size_cap = hr_dev->caps.page_size_cap;
 	props->vendor_id = hr_dev->vendor_id;
--- a/drivers/infiniband/hw/hns/hns_roce_qp.c
+++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
@ -660,6 +660,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 				goto err_rq_sge_list;
 			}
 			*hr_qp->rdb.db_record = 0;
+			hr_qp->rdb_en = 1;
 		}

 		/* Allocate QP buf */
@ -955,7 +956,14 @@ int hns_roce_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	}

 	if (cur_state == new_state && cur_state == IB_QPS_RESET) {
-		ret = 0;
+		if (hr_dev->caps.min_wqes) {
+			ret = -EPERM;
+			dev_err(dev, "cur_state=%d new_state=%d\n", cur_state,
+				new_state);
+		} else {
+			ret = 0;
+		}
+
 		goto out;
 	}

--- a/drivers/infiniband/hw/i40iw/i40iw.h
+++ b/drivers/infiniband/hw/i40iw/i40iw.h
@ -207,6 +207,7 @@ struct i40iw_msix_vector {
 	u32 irq;
 	u32 cpu_affinity;
 	u32 ceq_id;
+	cpumask_t mask;
 };

 struct l2params_work {
--- a/drivers/infiniband/hw/i40iw/i40iw_cm.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_cm.c
@ -2085,7 +2085,7 @@ static int i40iw_addr_resolve_neigh_ipv6(struct i40iw_device *iwdev,
 	if (netif_is_bond_slave(netdev))
 		netdev = netdev_master_upper_dev_get(netdev);

-	neigh = dst_neigh_lookup(dst, &dst_addr);
+	neigh = dst_neigh_lookup(dst, dst_addr.sin6_addr.in6_u.u6_addr32);

 	rcu_read_lock();
 	if (neigh) {
--- a/drivers/infiniband/hw/i40iw/i40iw_hw.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_hw.c
@ -331,7 +331,7 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 		switch (info->ae_id) {
 		case I40IW_AE_LLP_FIN_RECEIVED:
 			if (qp->term_flags)
-				continue;
+				break;
 			if (atomic_inc_return(&iwqp->close_timer_started) == 1) {
 				iwqp->hw_tcp_state = I40IW_TCP_STATE_CLOSE_WAIT;
 				if ((iwqp->hw_tcp_state == I40IW_TCP_STATE_CLOSE_WAIT) &&
@ -360,7 +360,7 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 			break;
 		case I40IW_AE_LLP_CONNECTION_RESET:
 			if (atomic_read(&iwqp->close_timer_started))
-				continue;
+				break;
 			i40iw_cm_disconn(iwqp);
 			break;
 		case I40IW_AE_QP_SUSPEND_COMPLETE:
--- a/drivers/infiniband/hw/i40iw/i40iw_main.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_main.c
@ -687,7 +687,6 @@ static enum i40iw_status_code i40iw_configure_ceq_vector(struct i40iw_device *iw
 							 struct i40iw_msix_vector *msix_vec)
 {
 	enum i40iw_status_code status;
-	cpumask_t mask;

 	if (iwdev->msix_shared && !ceq_id) {
 		tasklet_init(&iwdev->dpc_tasklet, i40iw_dpc, (unsigned long)iwdev);
@ -697,9 +696,9 @@ static enum i40iw_status_code i40iw_configure_ceq_vector(struct i40iw_device *iw
 		status = request_irq(msix_vec->irq, i40iw_ceq_handler, 0, "CEQ", iwceq);
 	}

-	cpumask_clear(&mask);
-	cpumask_set_cpu(msix_vec->cpu_affinity, &mask);
-	irq_set_affinity_hint(msix_vec->irq, &mask);
+	cpumask_clear(&msix_vec->mask);
+	cpumask_set_cpu(msix_vec->cpu_affinity, &msix_vec->mask);
+	irq_set_affinity_hint(msix_vec->irq, &msix_vec->mask);

 	if (status) {
 		i40iw_pr_err("ceq irq config fail\n");
--- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
@ -394,6 +394,7 @@ static struct i40iw_pbl *i40iw_get_pbl(unsigned long va,

 	list_for_each_entry(iwpbl, pbl_list, list) {
 		if (iwpbl->user_base == va) {
+			iwpbl->on_list = false;
 			list_del(&iwpbl->list);
 			return iwpbl;
 		}
@ -614,6 +615,7 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 		return ERR_PTR(-ENOMEM);

 	iwqp = (struct i40iw_qp *)mem;
+	iwqp->allocated_buffer = mem;
 	qp = &iwqp->sc_qp;
 	qp->back_qp = (void *)iwqp;
 	qp->push_idx = I40IW_INVALID_PUSH_PAGE_INDEX;
@ -642,7 +644,6 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 		goto error;
 	}

-	iwqp->allocated_buffer = mem;
 	iwqp->iwdev = iwdev;
 	iwqp->iwpd = iwpd;
 	iwqp->ibqp.qp_num = qp_num;
@ -1898,6 +1899,7 @@ static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,
 			goto error;
 		spin_lock_irqsave(&ucontext->qp_reg_mem_list_lock, flags);
 		list_add_tail(&iwpbl->list, &ucontext->qp_reg_mem_list);
+		iwpbl->on_list = true;
 		spin_unlock_irqrestore(&ucontext->qp_reg_mem_list_lock, flags);
 		break;
 	case IW_MEMREG_TYPE_CQ:
@ -1908,6 +1910,7 @@ static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,

 		spin_lock_irqsave(&ucontext->cq_reg_mem_list_lock, flags);
 		list_add_tail(&iwpbl->list, &ucontext->cq_reg_mem_list);
+		iwpbl->on_list = true;
 		spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
 		break;
 	case IW_MEMREG_TYPE_MEM:
@ -2045,14 +2048,18 @@ static void i40iw_del_memlist(struct i40iw_mr *iwmr,
 	switch (iwmr->type) {
 	case IW_MEMREG_TYPE_CQ:
 		spin_lock_irqsave(&ucontext->cq_reg_mem_list_lock, flags);
-		if (!list_empty(&ucontext->cq_reg_mem_list))
+		if (iwpbl->on_list) {
+			iwpbl->on_list = false;
 			list_del(&iwpbl->list);
+		}
 		spin_unlock_irqrestore(&ucontext->cq_reg_mem_list_lock, flags);
 		break;
 	case IW_MEMREG_TYPE_QP:
 		spin_lock_irqsave(&ucontext->qp_reg_mem_list_lock, flags);
-		if (!list_empty(&ucontext->qp_reg_mem_list))
+		if (iwpbl->on_list) {
+			iwpbl->on_list = false;
 			list_del(&iwpbl->list);
+		}
 		spin_unlock_irqrestore(&ucontext->qp_reg_mem_list_lock, flags);
 		break;
 	default:
--- a/drivers/infiniband/hw/i40iw/i40iw_verbs.h
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.h
@ -78,6 +78,7 @@ struct i40iw_pbl {
 	};

 	bool pbl_allocated;
+	bool on_list;
 	u64 user_base;
 	struct i40iw_pble_alloc pble_alloc;
 	struct i40iw_mr *iwmr;
--- a/drivers/infiniband/hw/mlx4/mr.c
+++ b/drivers/infiniband/hw/mlx4/mr.c
@ -367,6 +367,40 @@ int mlx4_ib_umem_calc_optimal_mtt_size(struct ib_umem *umem, u64 start_va,
 	return block_shift;
 }

+static struct ib_umem *mlx4_get_umem_mr(struct ib_ucontext *context, u64 start,
+					u64 length, u64 virt_addr,
+					int access_flags)
+{
+	/*
+	 * Force registering the memory as writable if the underlying pages
+	 * are writable.  This is so rereg can change the access permissions
+	 * from readable to writable without having to run through ib_umem_get
+	 * again
+	 */
+	if (!ib_access_writable(access_flags)) {
+		struct vm_area_struct *vma;
+
+		down_read(&current->mm->mmap_sem);
+		/*
+		 * FIXME: Ideally this would iterate over all the vmas that
+		 * cover the memory, but for now it requires a single vma to
+		 * entirely cover the MR to support RO mappings.
+		 */
+		vma = find_vma(current->mm, start);
+		if (vma && vma->vm_end >= start + length &&
+		    vma->vm_start <= start) {
+			if (vma->vm_flags & VM_WRITE)
+				access_flags |= IB_ACCESS_LOCAL_WRITE;
+		} else {
+			access_flags |= IB_ACCESS_LOCAL_WRITE;
+		}
+
+		up_read(&current->mm->mmap_sem);
+	}
+
+	return ib_umem_get(context, start, length, access_flags, 0);
+}
+
 struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata)
@ -381,10 +415,8 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	if (!mr)
 		return ERR_PTR(-ENOMEM);

-	/* Force registering the memory as writable. */
-	/* Used for memory re-registeration. HCA protects the access */
-	mr->umem = ib_umem_get(pd->uobject->context, start, length,
-			       access_flags | IB_ACCESS_LOCAL_WRITE, 0);
+	mr->umem = mlx4_get_umem_mr(pd->uobject->context, start, length,
+				    virt_addr, access_flags);
 	if (IS_ERR(mr->umem)) {
 		err = PTR_ERR(mr->umem);
 		goto err_free;
@ -454,6 +486,9 @@ int mlx4_ib_rereg_user_mr(struct ib_mr *mr, int flags,
 	}

 	if (flags & IB_MR_REREG_ACCESS) {
+		if (ib_access_writable(mr_access_flags) && !mmr->umem->writable)
+			return -EPERM;
+
 		err = mlx4_mr_hw_change_access(dev->dev, *pmpt_entry,
 					       convert_access(mr_access_flags));

@ -467,10 +502,9 @@ int mlx4_ib_rereg_user_mr(struct ib_mr *mr, int flags,

 		mlx4_mr_rereg_mem_cleanup(dev->dev, &mmr->mmr);
 		ib_umem_release(mmr->umem);
-		mmr->umem = ib_umem_get(mr->uobject->context, start, length,
-					mr_access_flags |
-					IB_ACCESS_LOCAL_WRITE,
-					0);
+		mmr->umem =
+			mlx4_get_umem_mr(mr->uobject->context, start, length,
+					 virt_addr, mr_access_flags);
 		if (IS_ERR(mmr->umem)) {
 			err = PTR_ERR(mmr->umem);
 			/* Prevent mlx4_ib_dereg_mr from free'ing invalid pointer */
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@ -2419,7 +2419,7 @@ static void set_proto(void *outer_c, void *outer_v, u8 mask, u8 val)
 	MLX5_SET(fte_match_set_lyr_2_4, outer_v, ip_protocol, val);
 }

-static void set_flow_label(void *misc_c, void *misc_v, u8 mask, u8 val,
+static void set_flow_label(void *misc_c, void *misc_v, u32 mask, u32 val,
 			   bool inner)
 {
 	if (inner) {
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@ -487,11 +487,6 @@ static int qp_has_rq(struct ib_qp_init_attr *attr)
 	return 1;
 }

-static int first_med_bfreg(void)
-{
-	return 1;
-}
-
 enum {
 	/* this is the first blue flame register in the array of bfregs assigned
 	 * to a processes. Since we do not use it for blue flame but rather
@ -517,6 +512,12 @@ static int num_med_bfreg(struct mlx5_ib_dev *dev,
 	return n >= 0 ? n : 0;
 }

+static int first_med_bfreg(struct mlx5_ib_dev *dev,
+			   struct mlx5_bfreg_info *bfregi)
+{
+	return num_med_bfreg(dev, bfregi) ? 1 : -ENOMEM;
+}
+
 static int first_hi_bfreg(struct mlx5_ib_dev *dev,
 			  struct mlx5_bfreg_info *bfregi)
 {
@ -544,10 +545,13 @@ static int alloc_high_class_bfreg(struct mlx5_ib_dev *dev,
 static int alloc_med_class_bfreg(struct mlx5_ib_dev *dev,
 				 struct mlx5_bfreg_info *bfregi)
 {
-	int minidx = first_med_bfreg();
+	int minidx = first_med_bfreg(dev, bfregi);
 	int i;

-	for (i = first_med_bfreg(); i < first_hi_bfreg(dev, bfregi); i++) {
+	if (minidx < 0)
+		return minidx;
+
+	for (i = minidx; i < first_hi_bfreg(dev, bfregi); i++) {
 		if (bfregi->count[i] < bfregi->count[minidx])
 			minidx = i;
 		if (!bfregi->count[minidx])
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@ -401,49 +401,47 @@ int qedr_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 {
 	struct qedr_ucontext *ucontext = get_qedr_ucontext(context);
 	struct qedr_dev *dev = get_qedr_dev(context->device);
-	unsigned long vm_page = vma->vm_pgoff << PAGE_SHIFT;
-	u64 unmapped_db = dev->db_phys_addr;
+	unsigned long phys_addr = vma->vm_pgoff << PAGE_SHIFT;
 	unsigned long len = (vma->vm_end - vma->vm_start);
-	int rc = 0;
-	bool found;
+	unsigned long dpi_start;
+
+	dpi_start = dev->db_phys_addr + (ucontext->dpi * ucontext->dpi_size);

 	DP_DEBUG(dev, QEDR_MSG_INIT,
-		 "qedr_mmap called vm_page=0x%lx vm_pgoff=0x%lx unmapped_db=0x%llx db_size=%x, len=%lx\n",
-		 vm_page, vma->vm_pgoff, unmapped_db, dev->db_size, len);
-	if (vma->vm_start & (PAGE_SIZE - 1)) {
-		DP_ERR(dev, "Vma_start not page aligned = %ld\n",
-		       vma->vm_start);
+		 "mmap invoked with vm_start=0x%pK, vm_end=0x%pK,vm_pgoff=0x%pK; dpi_start=0x%pK dpi_size=0x%x\n",
+		 (void *)vma->vm_start, (void *)vma->vm_end,
+		 (void *)vma->vm_pgoff, (void *)dpi_start, ucontext->dpi_size);
+
+	if ((vma->vm_start & (PAGE_SIZE - 1)) || (len & (PAGE_SIZE - 1))) {
+		DP_ERR(dev,
+		       "failed mmap, adrresses must be page aligned: start=0x%pK, end=0x%pK\n",
+		       (void *)vma->vm_start, (void *)vma->vm_end);
 		return -EINVAL;
 	}

-	found = qedr_search_mmap(ucontext, vm_page, len);
-	if (!found) {
-		DP_ERR(dev, "Vma_pgoff not found in mapped array = %ld\n",
+	if (!qedr_search_mmap(ucontext, phys_addr, len)) {
+		DP_ERR(dev, "failed mmap, vm_pgoff=0x%lx is not authorized\n",
 		       vma->vm_pgoff);
 		return -EINVAL;
 	}

-	DP_DEBUG(dev, QEDR_MSG_INIT, "Mapping doorbell bar\n");
-
-	if ((vm_page >= unmapped_db) && (vm_page <= (unmapped_db +
-						     dev->db_size))) {
-		DP_DEBUG(dev, QEDR_MSG_INIT, "Mapping doorbell bar\n");
-		if (vma->vm_flags & VM_READ) {
-			DP_ERR(dev, "Trying to map doorbell bar for read\n");
-			return -EPERM;
-		}
-
-		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
-
-		rc = io_remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff,
-					PAGE_SIZE, vma->vm_page_prot);
-	} else {
-		DP_DEBUG(dev, QEDR_MSG_INIT, "Mapping chains\n");
-		rc = remap_pfn_range(vma, vma->vm_start,
-				     vma->vm_pgoff, len, vma->vm_page_prot);
+	if (phys_addr < dpi_start ||
+	    ((phys_addr + len) > (dpi_start + ucontext->dpi_size))) {
+		DP_ERR(dev,
+		       "failed mmap, pages are outside of dpi; page address=0x%pK, dpi_start=0x%pK, dpi_size=0x%x\n",
+		       (void *)phys_addr, (void *)dpi_start,
+		       ucontext->dpi_size);
+		return -EINVAL;
 	}
-	DP_DEBUG(dev, QEDR_MSG_INIT, "qedr_mmap return code: %d\n", rc);
-	return rc;
+
+	if (vma->vm_flags & VM_READ) {
+		DP_ERR(dev, "failed mmap, cannot map doorbell bar for read\n");
+		return -EINVAL;
+	}
+
+	vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
+	return io_remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff, len,
+				  vma->vm_page_prot);
 }

 struct ib_pd *qedr_alloc_pd(struct ib_device *ibdev,
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@ -761,7 +761,6 @@ static int rxe_post_send_kernel(struct rxe_qp *qp, struct ib_send_wr *wr,
 	unsigned int mask;
 	unsigned int length = 0;
 	int i;
-	int must_sched;

 	while (wr) {
 		mask = wr_opcode_mask(wr->opcode, qp);
@ -791,14 +790,7 @@ static int rxe_post_send_kernel(struct rxe_qp *qp, struct ib_send_wr *wr,
 		wr = wr->next;
 	}

-	/*
-	 * Must sched in case of GSI QP because ib_send_mad() hold irq lock,
-	 * and the requester call ip_local_out_sk() that takes spin_lock_bh.
-	 */
-	must_sched = (qp_type(qp) == IB_QPT_GSI) ||
-			(queue_count(qp->sq.queue) > 1);
-
-	rxe_run_task(&qp->req.task, must_sched);
+	rxe_run_task(&qp->req.task, 1);
 	if (unlikely(qp->req.state == QP_STATE_ERROR))
 		rxe_run_task(&qp->comp.task, 1);

--- a/drivers/infiniband/ulp/srpt/Kconfig
+++ b/drivers/infiniband/ulp/srpt/Kconfig
@ -1,6 +1,6 @@
 config INFINIBAND_SRPT
 	tristate "InfiniBand SCSI RDMA Protocol target support"
-	depends on INFINIBAND && INFINIBAND_ADDR_TRANS && TARGET_CORE
+	depends on INFINIBAND_ADDR_TRANS && TARGET_CORE
 	---help---

 	  Support for the SCSI RDMA Protocol (SRP) Target driver. The
--- a/drivers/nvme/host/Kconfig
+++ b/drivers/nvme/host/Kconfig
@ -27,7 +27,7 @@ config NVME_FABRICS

 config NVME_RDMA
 	tristate "NVM Express over Fabrics RDMA host driver"
-	depends on INFINIBAND && INFINIBAND_ADDR_TRANS && BLOCK
+	depends on INFINIBAND_ADDR_TRANS && BLOCK
 	select NVME_CORE
 	select NVME_FABRICS
 	select SG_POOL
--- a/drivers/nvme/target/Kconfig
+++ b/drivers/nvme/target/Kconfig
@ -27,7 +27,7 @@ config NVME_TARGET_LOOP

 config NVME_TARGET_RDMA
 	tristate "NVMe over Fabrics RDMA target support"
-	depends on INFINIBAND && INFINIBAND_ADDR_TRANS
+	depends on INFINIBAND_ADDR_TRANS
 	depends on NVME_TARGET
 	select SGL_ALLOC
 	help
--- a/drivers/staging/lustre/lnet/Kconfig
+++ b/drivers/staging/lustre/lnet/Kconfig
@ -34,7 +34,7 @@ config LNET_SELFTEST

 config LNET_XPRT_IB
 	tristate "LNET infiniband support"
-	depends on LNET && PCI && INFINIBAND && INFINIBAND_ADDR_TRANS
+	depends on LNET && PCI && INFINIBAND_ADDR_TRANS
 	default LNET && INFINIBAND
 	help
 	  This option allows the LNET users to use infiniband as an
--- a/fs/cifs/Kconfig
+++ b/fs/cifs/Kconfig
@ -197,7 +197,7 @@ config CIFS_SMB311

 config CIFS_SMB_DIRECT
 	bool "SMB Direct support (Experimental)"
-	depends on CIFS=m && INFINIBAND && INFINIBAND_ADDR_TRANS || CIFS=y && INFINIBAND=y && INFINIBAND_ADDR_TRANS=y
+	depends on CIFS=m && INFINIBAND_ADDR_TRANS || CIFS=y && INFINIBAND_ADDR_TRANS=y
 	help
 	  Enables SMB Direct experimental support for SMB 3.0, 3.02 and 3.1.1.
 	  SMB Direct allows transferring SMB packets over RDMA. If unsure,
--- a/include/rdma/ib_umem.h
+++ b/include/rdma/ib_umem.h
@ -48,7 +48,6 @@ struct ib_umem {
 	int                     writable;
 	int                     hugetlb;
 	struct work_struct	work;
-	struct pid             *pid;
 	struct mm_struct       *mm;
 	unsigned long		diff;
 	struct ib_umem_odp     *odp_data;
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@ -3766,6 +3766,20 @@ static inline int ib_check_mr_access(int flags)
 	return 0;
 }

+static inline bool ib_access_writable(int access_flags)
+{
+	/*
+	 * We have writable memory backing the MR if any of the following
+	 * access flags are set.  "Local write" and "remote write" obviously
+	 * require write access.  "Remote atomic" can do things like fetch and
+	 * add, which will modify memory, and "MW bind" can change permissions
+	 * by binding a window.
+	 */
+	return access_flags &
+		(IB_ACCESS_LOCAL_WRITE   | IB_ACCESS_REMOTE_WRITE |
+		 IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_MW_BIND);
+}
+
 /**
 * ib_check_mr_status: lightweight check of MR status.
 *     This routine may provide status checks on a selected
--- a/include/rdma/uverbs_ioctl.h
+++ b/include/rdma/uverbs_ioctl.h
@ -411,13 +411,13 @@ static inline int uverbs_attr_get_enum_id(const struct uverbs_attr_bundle *attrs
 static inline void *uverbs_attr_get_obj(const struct uverbs_attr_bundle *attrs_bundle,
 					u16 idx)
 {
-	struct ib_uobject *uobj =
-		uverbs_attr_get(attrs_bundle, idx)->obj_attr.uobject;
+	const struct uverbs_attr *attr;

-	if (IS_ERR(uobj))
-		return uobj;
+	attr = uverbs_attr_get(attrs_bundle, idx);
+	if (IS_ERR(attr))
+		return ERR_CAST(attr);

-	return uobj->object;
+	return attr->obj_attr.uobject->object;
 }

 static inline int uverbs_copy_to(const struct uverbs_attr_bundle *attrs_bundle,
--- a/net/9p/Kconfig
+++ b/net/9p/Kconfig
@ -32,7 +32,7 @@ config NET_9P_XEN


 config NET_9P_RDMA
-	depends on INET && INFINIBAND && INFINIBAND_ADDR_TRANS
+	depends on INET && INFINIBAND_ADDR_TRANS
 	tristate "9P RDMA Transport (Experimental)"
 	help
 	  This builds support for an RDMA transport.
--- a/net/rds/Kconfig
+++ b/net/rds/Kconfig
@ -8,7 +8,7 @@ config RDS

 config RDS_RDMA
 	tristate "RDS over Infiniband"
-	depends on RDS && INFINIBAND && INFINIBAND_ADDR_TRANS
+	depends on RDS && INFINIBAND_ADDR_TRANS
 	---help---
 	  Allow RDS to use Infiniband as a transport.
 	  This transport supports RDMA operations.
--- a/net/sunrpc/Kconfig
+++ b/net/sunrpc/Kconfig
@ -50,7 +50,7 @@ config SUNRPC_DEBUG

 config SUNRPC_XPRT_RDMA
 	tristate "RPC-over-RDMA transport"
-	depends on SUNRPC && INFINIBAND && INFINIBAND_ADDR_TRANS
+	depends on SUNRPC && INFINIBAND_ADDR_TRANS
 	default SUNRPC && INFINIBAND
 	select SG_POOL
 	help