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

Pull rdma fixes from Jason Gunthorpe:
 "This is the next batch of for-rc patches from RDMA. It includes the
  fix for the ipoib regression I mentioned last time, and the result of
  a fairly major debugging effort to get iser working reliably on cxgb4
  hardware - it turns out the cxgb4 driver was not handling QP error
  flushing properly causing iser to fail.

   - cxgb4 fix for an iser testing failure as debugged by Steve and
     Sagi. The problem was a driver bug in the handling of shutting down
     a QP.

   - Various vmw_pvrdma fixes for bogus WARN_ON, missed resource free on
     error unwind and a use after free bug

   - Improper congestion counter values on mlx5 when link aggregation is
     enabled

   - ipoib lockdep regression introduced in this merge window

   - hfi1 regression supporting the device in a VM introduced in a
     recent patch

   - Typo that breaks future uAPI compatibility in the verbs core

   - More SELinux related oops fixing

   - Fix an oops during error unwind in mlx5"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  IB/mlx5: Fix mlx5_ib_alloc_mr error flow
  IB/core: Verify that QP is security enabled in create and destroy
  IB/uverbs: Fix command checking as part of ib_uverbs_ex_modify_qp()
  IB/mlx5: Serialize access to the VMA list
  IB/hfi: Only read capability registers if the capability exists
  IB/ipoib: Fix lockdep issue found on ipoib_ib_dev_heavy_flush
  IB/mlx5: Fix congestion counters in LAG mode
  RDMA/vmw_pvrdma: Avoid use after free due to QP/CQ/SRQ destroy
  RDMA/vmw_pvrdma: Use refcount_dec_and_test to avoid warning
  RDMA/vmw_pvrdma: Call ib_umem_release on destroy QP path
  iw_cxgb4: when flushing, complete all wrs in a chain
  iw_cxgb4: reflect the original WR opcode in drain cqes
  iw_cxgb4: Only validate the MSN for successful completions

drivers/infiniband/core/security.c

@@ -386,6 +386,9 @@ int ib_open_shared_qp_security(struct ib_qp *qp, struct ib_device *dev)
 	if (ret)
 		return ret;
 
+	if (!qp->qp_sec)
+		return 0;
+
 	mutex_lock(&real_qp->qp_sec->mutex);
 	ret = check_qp_port_pkey_settings(real_qp->qp_sec->ports_pkeys,
 					  qp->qp_sec);

drivers/infiniband/core/uverbs_cmd.c

@@ -2074,8 +2074,8 @@ int ib_uverbs_ex_modify_qp(struct ib_uverbs_file *file,
 		return -EOPNOTSUPP;
 
 	if (ucore->inlen > sizeof(cmd)) {
-		if (ib_is_udata_cleared(ucore, sizeof(cmd),
-					ucore->inlen - sizeof(cmd)))
+		if (!ib_is_udata_cleared(ucore, sizeof(cmd),
+					 ucore->inlen - sizeof(cmd)))
 			return -EOPNOTSUPP;
 	}
 

drivers/infiniband/core/verbs.c

@@ -1438,7 +1438,8 @@ int ib_close_qp(struct ib_qp *qp)
 	spin_unlock_irqrestore(&real_qp->device->event_handler_lock, flags);
 
 	atomic_dec(&real_qp->usecnt);
-	ib_close_shared_qp_security(qp->qp_sec);
+	if (qp->qp_sec)
+		ib_close_shared_qp_security(qp->qp_sec);
 	kfree(qp);
 
 	return 0;

drivers/infiniband/hw/cxgb4/cq.c

@@ -395,7 +395,7 @@ next_cqe:
 
 static int cqe_completes_wr(struct t4_cqe *cqe, struct t4_wq *wq)
 {
-	if (CQE_OPCODE(cqe) == C4IW_DRAIN_OPCODE) {
+	if (DRAIN_CQE(cqe)) {
 		WARN_ONCE(1, "Unexpected DRAIN CQE qp id %u!\n", wq->sq.qid);
 		return 0;
 	}
@@ -494,7 +494,7 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 	/*
 	 * Special cqe for drain WR completions...
 	 */
-	if (CQE_OPCODE(hw_cqe) == C4IW_DRAIN_OPCODE) {
+	if (DRAIN_CQE(hw_cqe)) {
 		*cookie = CQE_DRAIN_COOKIE(hw_cqe);
 		*cqe = *hw_cqe;
 		goto skip_cqe;
@@ -571,10 +571,10 @@ static int poll_cq(struct t4_wq *wq, struct t4_cq *cq, struct t4_cqe *cqe,
 			ret = -EAGAIN;
 			goto skip_cqe;
 		}
-		if (unlikely((CQE_WRID_MSN(hw_cqe) != (wq->rq.msn)))) {
+		if (unlikely(!CQE_STATUS(hw_cqe) &&
+			     CQE_WRID_MSN(hw_cqe) != wq->rq.msn)) {
 			t4_set_wq_in_error(wq);
-			hw_cqe->header |= htonl(CQE_STATUS_V(T4_ERR_MSN));
-			goto proc_cqe;
+			hw_cqe->header |= cpu_to_be32(CQE_STATUS_V(T4_ERR_MSN));
 		}
 		goto proc_cqe;
 	}
@@ -748,9 +748,6 @@ static int c4iw_poll_cq_one(struct c4iw_cq *chp, struct ib_wc *wc)
 				c4iw_invalidate_mr(qhp->rhp,
 						   CQE_WRID_FR_STAG(&cqe));
 			break;
-		case C4IW_DRAIN_OPCODE:
-			wc->opcode = IB_WC_SEND;
-			break;
 		default:
 			pr_err("Unexpected opcode %d in the CQE received for QPID=0x%0x\n",
 			       CQE_OPCODE(&cqe), CQE_QPID(&cqe));

drivers/infiniband/hw/cxgb4/iw_cxgb4.h

@@ -693,8 +693,6 @@ static inline int to_ib_qp_state(int c4iw_qp_state)
 	return IB_QPS_ERR;
 }
 
-#define C4IW_DRAIN_OPCODE FW_RI_SGE_EC_CR_RETURN
-
 static inline u32 c4iw_ib_to_tpt_access(int a)
 {
 	return (a & IB_ACCESS_REMOTE_WRITE ? FW_RI_MEM_ACCESS_REM_WRITE : 0) |

drivers/infiniband/hw/cxgb4/qp.c

@@ -790,21 +790,57 @@ static int ring_kernel_rq_db(struct c4iw_qp *qhp, u16 inc)
 	return 0;
 }
 
-static void complete_sq_drain_wr(struct c4iw_qp *qhp, struct ib_send_wr *wr)
+static int ib_to_fw_opcode(int ib_opcode)
+{
+	int opcode;
+
+	switch (ib_opcode) {
+	case IB_WR_SEND_WITH_INV:
+		opcode = FW_RI_SEND_WITH_INV;
+		break;
+	case IB_WR_SEND:
+		opcode = FW_RI_SEND;
+		break;
+	case IB_WR_RDMA_WRITE:
+		opcode = FW_RI_RDMA_WRITE;
+		break;
+	case IB_WR_RDMA_READ:
+	case IB_WR_RDMA_READ_WITH_INV:
+		opcode = FW_RI_READ_REQ;
+		break;
+	case IB_WR_REG_MR:
+		opcode = FW_RI_FAST_REGISTER;
+		break;
+	case IB_WR_LOCAL_INV:
+		opcode = FW_RI_LOCAL_INV;
+		break;
+	default:
+		opcode = -EINVAL;
+	}
+	return opcode;
+}
+
+static int complete_sq_drain_wr(struct c4iw_qp *qhp, struct ib_send_wr *wr)
 {
 	struct t4_cqe cqe = {};
 	struct c4iw_cq *schp;
 	unsigned long flag;
 	struct t4_cq *cq;
+	int opcode;
 
 	schp = to_c4iw_cq(qhp->ibqp.send_cq);
 	cq = &schp->cq;
 
+	opcode = ib_to_fw_opcode(wr->opcode);
+	if (opcode < 0)
+		return opcode;
+
 	cqe.u.drain_cookie = wr->wr_id;
 	cqe.header = cpu_to_be32(CQE_STATUS_V(T4_ERR_SWFLUSH) |
-				 CQE_OPCODE_V(C4IW_DRAIN_OPCODE) |
+				 CQE_OPCODE_V(opcode) |
 				 CQE_TYPE_V(1) |
 				 CQE_SWCQE_V(1) |
+				 CQE_DRAIN_V(1) |
 				 CQE_QPID_V(qhp->wq.sq.qid));
 
 	spin_lock_irqsave(&schp->lock, flag);
@@ -819,6 +855,23 @@ static void complete_sq_drain_wr(struct c4iw_qp *qhp, struct ib_send_wr *wr)
 					   schp->ibcq.cq_context);
 		spin_unlock_irqrestore(&schp->comp_handler_lock, flag);
 	}
+	return 0;
+}
+
+static int complete_sq_drain_wrs(struct c4iw_qp *qhp, struct ib_send_wr *wr,
+				struct ib_send_wr **bad_wr)
+{
+	int ret = 0;
+
+	while (wr) {
+		ret = complete_sq_drain_wr(qhp, wr);
+		if (ret) {
+			*bad_wr = wr;
+			break;
+		}
+		wr = wr->next;
+	}
+	return ret;
 }
 
 static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr)
@@ -833,9 +886,10 @@ static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr)
 
 	cqe.u.drain_cookie = wr->wr_id;
 	cqe.header = cpu_to_be32(CQE_STATUS_V(T4_ERR_SWFLUSH) |
-				 CQE_OPCODE_V(C4IW_DRAIN_OPCODE) |
+				 CQE_OPCODE_V(FW_RI_SEND) |
 				 CQE_TYPE_V(0) |
 				 CQE_SWCQE_V(1) |
+				 CQE_DRAIN_V(1) |
 				 CQE_QPID_V(qhp->wq.sq.qid));
 
 	spin_lock_irqsave(&rchp->lock, flag);
@@ -852,6 +906,14 @@ static void complete_rq_drain_wr(struct c4iw_qp *qhp, struct ib_recv_wr *wr)
 	}
 }
 
+static void complete_rq_drain_wrs(struct c4iw_qp *qhp, struct ib_recv_wr *wr)
+{
+	while (wr) {
+		complete_rq_drain_wr(qhp, wr);
+		wr = wr->next;
+	}
+}
+
 int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		   struct ib_send_wr **bad_wr)
 {
@@ -875,7 +937,7 @@ int c4iw_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	 */
 	if (qhp->wq.flushed) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
-		complete_sq_drain_wr(qhp, wr);
+		err = complete_sq_drain_wrs(qhp, wr, bad_wr);
 		return err;
 	}
 	num_wrs = t4_sq_avail(&qhp->wq);
@@ -1023,7 +1085,7 @@ int c4iw_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 	 */
 	if (qhp->wq.flushed) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
-		complete_rq_drain_wr(qhp, wr);
+		complete_rq_drain_wrs(qhp, wr);
 		return err;
 	}
 	num_wrs = t4_rq_avail(&qhp->wq);

drivers/infiniband/hw/cxgb4/t4.h

@@ -197,6 +197,11 @@ struct t4_cqe {
 #define CQE_SWCQE_G(x)    ((((x) >> CQE_SWCQE_S)) & CQE_SWCQE_M)
 #define CQE_SWCQE_V(x)	  ((x)<<CQE_SWCQE_S)
 
+#define CQE_DRAIN_S       10
+#define CQE_DRAIN_M       0x1
+#define CQE_DRAIN_G(x)    ((((x) >> CQE_DRAIN_S)) & CQE_DRAIN_M)
+#define CQE_DRAIN_V(x)	  ((x)<<CQE_DRAIN_S)
+
 #define CQE_STATUS_S      5
 #define CQE_STATUS_M      0x1F
 #define CQE_STATUS_G(x)   ((((x) >> CQE_STATUS_S)) & CQE_STATUS_M)
@@ -213,6 +218,7 @@ struct t4_cqe {
 #define CQE_OPCODE_V(x)   ((x)<<CQE_OPCODE_S)
 
 #define SW_CQE(x)         (CQE_SWCQE_G(be32_to_cpu((x)->header)))
+#define DRAIN_CQE(x)      (CQE_DRAIN_G(be32_to_cpu((x)->header)))
 #define CQE_QPID(x)       (CQE_QPID_G(be32_to_cpu((x)->header)))
 #define CQE_TYPE(x)       (CQE_TYPE_G(be32_to_cpu((x)->header)))
 #define SQ_TYPE(x)	  (CQE_TYPE((x)))

drivers/infiniband/hw/hfi1/hfi.h

@@ -1131,7 +1131,6 @@ struct hfi1_devdata {
 	u16 pcie_lnkctl;
 	u16 pcie_devctl2;
 	u32 pci_msix0;
-	u32 pci_lnkctl3;
 	u32 pci_tph2;
 
 	/*

drivers/infiniband/hw/hfi1/pcie.c

@@ -411,15 +411,12 @@ int restore_pci_variables(struct hfi1_devdata *dd)
 	if (ret)
 		goto error;
 
-	ret = pci_write_config_dword(dd->pcidev, PCIE_CFG_SPCIE1,
-				     dd->pci_lnkctl3);
-	if (ret)
-		goto error;
-
-	ret = pci_write_config_dword(dd->pcidev, PCIE_CFG_TPH2, dd->pci_tph2);
-	if (ret)
-		goto error;
-
+	if (pci_find_ext_capability(dd->pcidev, PCI_EXT_CAP_ID_TPH)) {
+		ret = pci_write_config_dword(dd->pcidev, PCIE_CFG_TPH2,
+					     dd->pci_tph2);
+		if (ret)
+			goto error;
+	}
 	return 0;
 
 error:
@@ -469,15 +466,12 @@ int save_pci_variables(struct hfi1_devdata *dd)
 	if (ret)
 		goto error;
 
-	ret = pci_read_config_dword(dd->pcidev, PCIE_CFG_SPCIE1,
-				    &dd->pci_lnkctl3);
-	if (ret)
-		goto error;
-
-	ret = pci_read_config_dword(dd->pcidev, PCIE_CFG_TPH2, &dd->pci_tph2);
-	if (ret)
-		goto error;
-
+	if (pci_find_ext_capability(dd->pcidev, PCI_EXT_CAP_ID_TPH)) {
+		ret = pci_read_config_dword(dd->pcidev, PCIE_CFG_TPH2,
+					    &dd->pci_tph2);
+		if (ret)
+			goto error;
+	}
 	return 0;
 
 error:

drivers/infiniband/hw/mlx5/cmd.c

@@ -47,17 +47,6 @@ int mlx5_cmd_null_mkey(struct mlx5_core_dev *dev, u32 *null_mkey)
 	return err;
 }
 
-int mlx5_cmd_query_cong_counter(struct mlx5_core_dev *dev,
-				bool reset, void *out, int out_size)
-{
-	u32 in[MLX5_ST_SZ_DW(query_cong_statistics_in)] = { };
-
-	MLX5_SET(query_cong_statistics_in, in, opcode,
-		 MLX5_CMD_OP_QUERY_CONG_STATISTICS);
-	MLX5_SET(query_cong_statistics_in, in, clear, reset);
-	return mlx5_cmd_exec(dev, in, sizeof(in), out, out_size);
-}
-
 int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
 			       void *out, int out_size)
 {

drivers/infiniband/hw/mlx5/cmd.h

@@ -37,8 +37,6 @@
 #include <linux/mlx5/driver.h>
 
 int mlx5_cmd_null_mkey(struct mlx5_core_dev *dev, u32 *null_mkey);
-int mlx5_cmd_query_cong_counter(struct mlx5_core_dev *dev,
-				bool reset, void *out, int out_size);
 int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
 			       void *out, int out_size);
 int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *mdev,

drivers/infiniband/hw/mlx5/main.c

@@ -1463,6 +1463,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	}
 
 	INIT_LIST_HEAD(&context->vma_private_list);
+	mutex_init(&context->vma_private_list_mutex);
 	INIT_LIST_HEAD(&context->db_page_list);
 	mutex_init(&context->db_page_mutex);
 
@@ -1624,7 +1625,9 @@ static void  mlx5_ib_vma_close(struct vm_area_struct *area)
 	 * mlx5_ib_disassociate_ucontext().
 	 */
 	mlx5_ib_vma_priv_data->vma = NULL;
+	mutex_lock(mlx5_ib_vma_priv_data->vma_private_list_mutex);
 	list_del(&mlx5_ib_vma_priv_data->list);
+	mutex_unlock(mlx5_ib_vma_priv_data->vma_private_list_mutex);
 	kfree(mlx5_ib_vma_priv_data);
 }
 
@@ -1644,10 +1647,13 @@ static int mlx5_ib_set_vma_data(struct vm_area_struct *vma,
 		return -ENOMEM;
 
 	vma_prv->vma = vma;
+	vma_prv->vma_private_list_mutex = &ctx->vma_private_list_mutex;
 	vma->vm_private_data = vma_prv;
 	vma->vm_ops =  &mlx5_ib_vm_ops;
 
+	mutex_lock(&ctx->vma_private_list_mutex);
 	list_add(&vma_prv->list, vma_head);
+	mutex_unlock(&ctx->vma_private_list_mutex);
 
 	return 0;
 }
@@ -1690,6 +1696,7 @@ static void mlx5_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
 	 * mlx5_ib_vma_close.
 	 */
 	down_write(&owning_mm->mmap_sem);
+	mutex_lock(&context->vma_private_list_mutex);
 	list_for_each_entry_safe(vma_private, n, &context->vma_private_list,
 				 list) {
 		vma = vma_private->vma;
@@ -1704,6 +1711,7 @@ static void mlx5_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
 		list_del(&vma_private->list);
 		kfree(vma_private);
 	}
+	mutex_unlock(&context->vma_private_list_mutex);
 	up_write(&owning_mm->mmap_sem);
 	mmput(owning_mm);
 	put_task_struct(owning_process);
@@ -3737,34 +3745,6 @@ free:
 	return ret;
 }
 
-static int mlx5_ib_query_cong_counters(struct mlx5_ib_dev *dev,
-				       struct mlx5_ib_port *port,
-				       struct rdma_hw_stats *stats)
-{
-	int outlen = MLX5_ST_SZ_BYTES(query_cong_statistics_out);
-	void *out;
-	int ret, i;
-	int offset = port->cnts.num_q_counters;
-
-	out = kvzalloc(outlen, GFP_KERNEL);
-	if (!out)
-		return -ENOMEM;
-
-	ret = mlx5_cmd_query_cong_counter(dev->mdev, false, out, outlen);
-	if (ret)
-		goto free;
-
-	for (i = 0; i < port->cnts.num_cong_counters; i++) {
-		stats->value[i + offset] =
-			be64_to_cpup((__be64 *)(out +
-				     port->cnts.offsets[i + offset]));
-	}
-
-free:
-	kvfree(out);
-	return ret;
-}
-
 static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
 				struct rdma_hw_stats *stats,
 				u8 port_num, int index)
@@ -3782,7 +3762,12 @@ static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
 	num_counters = port->cnts.num_q_counters;
 
 	if (MLX5_CAP_GEN(dev->mdev, cc_query_allowed)) {
-		ret = mlx5_ib_query_cong_counters(dev, port, stats);
+		ret = mlx5_lag_query_cong_counters(dev->mdev,
+						   stats->value +
+						   port->cnts.num_q_counters,
+						   port->cnts.num_cong_counters,
+						   port->cnts.offsets +
+						   port->cnts.num_q_counters);
 		if (ret)
 			return ret;
 		num_counters += port->cnts.num_cong_counters;

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -115,6 +115,8 @@ enum {
 struct mlx5_ib_vma_private_data {
 	struct list_head list;
 	struct vm_area_struct *vma;
+	/* protect vma_private_list add/del */
+	struct mutex *vma_private_list_mutex;
 };
 
 struct mlx5_ib_ucontext {
@@ -129,6 +131,8 @@ struct mlx5_ib_ucontext {
 	/* Transport Domain number */
 	u32			tdn;
 	struct list_head	vma_private_list;
+	/* protect vma_private_list add/del */
+	struct mutex		vma_private_list_mutex;
 
 	unsigned long		upd_xlt_page;
 	/* protect ODP/KSM */

drivers/infiniband/hw/mlx5/mr.c

@@ -1637,6 +1637,7 @@ struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
 	MLX5_SET(mkc, mkc, access_mode, mr->access_mode);
 	MLX5_SET(mkc, mkc, umr_en, 1);
 
+	mr->ibmr.device = pd->device;
 	err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, inlen);
 	if (err)
 		goto err_destroy_psv;

drivers/infiniband/hw/vmw_pvrdma/pvrdma.h

@@ -94,7 +94,7 @@ struct pvrdma_cq {
 	u32 cq_handle;
 	bool is_kernel;
 	atomic_t refcnt;
-	wait_queue_head_t wait;
+	struct completion free;
 };
 
 struct pvrdma_id_table {
@@ -175,7 +175,7 @@ struct pvrdma_srq {
 	u32 srq_handle;
 	int npages;
 	refcount_t refcnt;
-	wait_queue_head_t wait;
+	struct completion free;
 };
 
 struct pvrdma_qp {
@@ -197,7 +197,7 @@ struct pvrdma_qp {
 	bool is_kernel;
 	struct mutex mutex; /* QP state mutex. */
 	atomic_t refcnt;
-	wait_queue_head_t wait;
+	struct completion free;
 };
 
 struct pvrdma_dev {

drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c

@@ -179,7 +179,7 @@ struct ib_cq *pvrdma_create_cq(struct ib_device *ibdev,
 		pvrdma_page_dir_insert_umem(&cq->pdir, cq->umem, 0);
 
 	atomic_set(&cq->refcnt, 1);
-	init_waitqueue_head(&cq->wait);
+	init_completion(&cq->free);
 	spin_lock_init(&cq->cq_lock);
 
 	memset(cmd, 0, sizeof(*cmd));
@@ -230,8 +230,9 @@ err_cq:
 
 static void pvrdma_free_cq(struct pvrdma_dev *dev, struct pvrdma_cq *cq)
 {
-	atomic_dec(&cq->refcnt);
-	wait_event(cq->wait, !atomic_read(&cq->refcnt));
+	if (atomic_dec_and_test(&cq->refcnt))
+		complete(&cq->free);
+	wait_for_completion(&cq->free);
 
 	if (!cq->is_kernel)
 		ib_umem_release(cq->umem);

drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c

@@ -346,9 +346,8 @@ static void pvrdma_qp_event(struct pvrdma_dev *dev, u32 qpn, int type)
 		ibqp->event_handler(&e, ibqp->qp_context);
 	}
 	if (qp) {
-		atomic_dec(&qp->refcnt);
-		if (atomic_read(&qp->refcnt) == 0)
-			wake_up(&qp->wait);
+		if (atomic_dec_and_test(&qp->refcnt))
+			complete(&qp->free);
 	}
 }
 
@@ -373,9 +372,8 @@ static void pvrdma_cq_event(struct pvrdma_dev *dev, u32 cqn, int type)
 		ibcq->event_handler(&e, ibcq->cq_context);
 	}
 	if (cq) {
-		atomic_dec(&cq->refcnt);
-		if (atomic_read(&cq->refcnt) == 0)
-			wake_up(&cq->wait);
+		if (atomic_dec_and_test(&cq->refcnt))
+			complete(&cq->free);
 	}
 }
 
@@ -404,7 +402,7 @@ static void pvrdma_srq_event(struct pvrdma_dev *dev, u32 srqn, int type)
 	}
 	if (srq) {
 		if (refcount_dec_and_test(&srq->refcnt))
-			wake_up(&srq->wait);
+			complete(&srq->free);
 	}
 }
 
@@ -539,9 +537,8 @@ static irqreturn_t pvrdma_intrx_handler(int irq, void *dev_id)
 		if (cq && cq->ibcq.comp_handler)
 			cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
 		if (cq) {
-			atomic_dec(&cq->refcnt);
-			if (atomic_read(&cq->refcnt))
-				wake_up(&cq->wait);
+			if (atomic_dec_and_test(&cq->refcnt))
+				complete(&cq->free);
 		}
 		pvrdma_idx_ring_inc(&ring->cons_head, ring_slots);
 	}

drivers/infiniband/hw/vmw_pvrdma/pvrdma_qp.c

@@ -246,7 +246,7 @@ struct ib_qp *pvrdma_create_qp(struct ib_pd *pd,
 		spin_lock_init(&qp->rq.lock);
 		mutex_init(&qp->mutex);
 		atomic_set(&qp->refcnt, 1);
-		init_waitqueue_head(&qp->wait);
+		init_completion(&qp->free);
 
 		qp->state = IB_QPS_RESET;
 
@@ -428,8 +428,16 @@ static void pvrdma_free_qp(struct pvrdma_qp *qp)
 
 	pvrdma_unlock_cqs(scq, rcq, &scq_flags, &rcq_flags);
 
-	atomic_dec(&qp->refcnt);
-	wait_event(qp->wait, !atomic_read(&qp->refcnt));
+	if (atomic_dec_and_test(&qp->refcnt))
+		complete(&qp->free);
+	wait_for_completion(&qp->free);
+
+	if (!qp->is_kernel) {
+		if (qp->rumem)
+			ib_umem_release(qp->rumem);
+		if (qp->sumem)
+			ib_umem_release(qp->sumem);
+	}
 
 	pvrdma_page_dir_cleanup(dev, &qp->pdir);
 

drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c

@@ -149,7 +149,7 @@ struct ib_srq *pvrdma_create_srq(struct ib_pd *pd,
 
 	spin_lock_init(&srq->lock);
 	refcount_set(&srq->refcnt, 1);
-	init_waitqueue_head(&srq->wait);
+	init_completion(&srq->free);
 
 	dev_dbg(&dev->pdev->dev,
 		"create shared receive queue from user space\n");
@@ -236,8 +236,9 @@ static void pvrdma_free_srq(struct pvrdma_dev *dev, struct pvrdma_srq *srq)
 	dev->srq_tbl[srq->srq_handle] = NULL;
 	spin_unlock_irqrestore(&dev->srq_tbl_lock, flags);
 
-	refcount_dec(&srq->refcnt);
-	wait_event(srq->wait, !refcount_read(&srq->refcnt));
+	if (refcount_dec_and_test(&srq->refcnt))
+		complete(&srq->free);
+	wait_for_completion(&srq->free);
 
 	/* There is no support for kernel clients, so this is safe. */
 	ib_umem_release(srq->umem);

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -1236,13 +1236,10 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv,
 		ipoib_ib_dev_down(dev);
 
 	if (level == IPOIB_FLUSH_HEAVY) {
-		rtnl_lock();
 		if (test_bit(IPOIB_FLAG_INITIALIZED, &priv->flags))
 			ipoib_ib_dev_stop(dev);
 
-		result = ipoib_ib_dev_open(dev);
-		rtnl_unlock();
-		if (result)
+		if (ipoib_ib_dev_open(dev))
 			return;
 
 		if (netif_queue_stopped(dev))
@@ -1282,7 +1279,9 @@ void ipoib_ib_dev_flush_heavy(struct work_struct *work)
 	struct ipoib_dev_priv *priv =
 		container_of(work, struct ipoib_dev_priv, flush_heavy);
 
+	rtnl_lock();
 	__ipoib_ib_dev_flush(priv, IPOIB_FLUSH_HEAVY, 0);
+	rtnl_unlock();
 }
 
 void ipoib_ib_dev_cleanup(struct net_device *dev)

drivers/net/ethernet/mellanox/mlx5/core/lag.c

@@ -137,6 +137,17 @@ int mlx5_cmd_destroy_vport_lag(struct mlx5_core_dev *dev)
 }
 EXPORT_SYMBOL(mlx5_cmd_destroy_vport_lag);
 
+static int mlx5_cmd_query_cong_counter(struct mlx5_core_dev *dev,
+				       bool reset, void *out, int out_size)
+{
+	u32 in[MLX5_ST_SZ_DW(query_cong_statistics_in)] = { };
+
+	MLX5_SET(query_cong_statistics_in, in, opcode,
+		 MLX5_CMD_OP_QUERY_CONG_STATISTICS);
+	MLX5_SET(query_cong_statistics_in, in, clear, reset);
+	return mlx5_cmd_exec(dev, in, sizeof(in), out, out_size);
+}
+
 static struct mlx5_lag *mlx5_lag_dev_get(struct mlx5_core_dev *dev)
 {
 	return dev->priv.lag;
@@ -633,3 +644,48 @@ bool mlx5_lag_intf_add(struct mlx5_interface *intf, struct mlx5_priv *priv)
 	/* If bonded, we do not add an IB device for PF1. */
 	return false;
 }
+
+int mlx5_lag_query_cong_counters(struct mlx5_core_dev *dev,
+				 u64 *values,
+				 int num_counters,
+				 size_t *offsets)
+{
+	int outlen = MLX5_ST_SZ_BYTES(query_cong_statistics_out);
+	struct mlx5_core_dev *mdev[MLX5_MAX_PORTS];
+	struct mlx5_lag *ldev;
+	int num_ports;
+	int ret, i, j;
+	void *out;
+
+	out = kvzalloc(outlen, GFP_KERNEL);
+	if (!out)
+		return -ENOMEM;
+
+	memset(values, 0, sizeof(*values) * num_counters);
+
+	mutex_lock(&lag_mutex);
+	ldev = mlx5_lag_dev_get(dev);
+	if (ldev && mlx5_lag_is_bonded(ldev)) {
+		num_ports = MLX5_MAX_PORTS;
+		mdev[0] = ldev->pf[0].dev;
+		mdev[1] = ldev->pf[1].dev;
+	} else {
+		num_ports = 1;
+		mdev[0] = dev;
+	}
+
+	for (i = 0; i < num_ports; ++i) {
+		ret = mlx5_cmd_query_cong_counter(mdev[i], false, out, outlen);
+		if (ret)
+			goto unlock;
+
+		for (j = 0; j < num_counters; ++j)
+			values[j] += be64_to_cpup((__be64 *)(out + offsets[j]));
+	}
+
+unlock:
+	mutex_unlock(&lag_mutex);
+	kvfree(out);
+	return ret;
+}
+EXPORT_SYMBOL(mlx5_lag_query_cong_counters);

include/linux/mlx5/driver.h

@@ -1165,6 +1165,10 @@ int mlx5_cmd_create_vport_lag(struct mlx5_core_dev *dev);
 int mlx5_cmd_destroy_vport_lag(struct mlx5_core_dev *dev);
 bool mlx5_lag_is_active(struct mlx5_core_dev *dev);
 struct net_device *mlx5_lag_get_roce_netdev(struct mlx5_core_dev *dev);
+int mlx5_lag_query_cong_counters(struct mlx5_core_dev *dev,
+				 u64 *values,
+				 int num_counters,
+				 size_t *offsets);
 struct mlx5_uars_page *mlx5_get_uars_page(struct mlx5_core_dev *mdev);
 void mlx5_put_uars_page(struct mlx5_core_dev *mdev, struct mlx5_uars_page *up);