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

Pull rdma fixes from Jason Gunthorpe:
 "Still not much going on, the usual set of oops and driver fixes this
  time:

   - Fix two uapi breakage regressions in mlx5 drivers

   - Various oops fixes in hfi1, mlx4, umem, uverbs, and ipoib

   - A protocol bug fix for hfi1 preventing it from implementing the
     verbs API properly, and a compatability fix for EXEC STACK user
     programs

   - Fix missed refcounting in the 'advise_mr' patches merged this
     cycle.

   - Fix wrong use of the uABI in the hns SRQ patches merged this cycle"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  IB/uverbs: Fix OOPs in uverbs_user_mmap_disassociate
  IB/ipoib: Fix for use-after-free in ipoib_cm_tx_start
  IB/uverbs: Fix ioctl query port to consider device disassociation
  RDMA/mlx5: Fix flow creation on representors
  IB/uverbs: Fix OOPs upon device disassociation
  RDMA/umem: Add missing initialization of owning_mm
  RDMA/hns: Update the kernel header file of hns
  IB/mlx5: Fix how advise_mr() launches async work
  RDMA/device: Expose ib_device_try_get(()
  IB/hfi1: Add limit test for RC/UC send via loopback
  IB/hfi1: Remove overly conservative VM_EXEC flag check
  IB/{hfi1, qib}: Fix WC.byte_len calculation for UD_SEND_WITH_IMM
  IB/mlx4: Fix using wrong function to destroy sqp AHs under SRIOV
  RDMA/mlx5: Fix check for supported user flags when creating a QP

drivers/infiniband/core/core_priv.h

@@ -267,7 +267,6 @@ static inline int ib_mad_enforce_security(struct ib_mad_agent_private *map,
 #endif
 
 struct ib_device *ib_device_get_by_index(u32 ifindex);
-void ib_device_put(struct ib_device *device);
 /* RDMA device netlink */
 void nldev_init(void);
 void nldev_exit(void);

drivers/infiniband/core/device.c

@@ -156,19 +156,26 @@ struct ib_device *ib_device_get_by_index(u32 index)
 	down_read(&lists_rwsem);
 	device = __ib_device_get_by_index(index);
 	if (device) {
-		/* Do not return a device if unregistration has started. */
-		if (!refcount_inc_not_zero(&device->refcount))
+		if (!ib_device_try_get(device))
 			device = NULL;
 	}
 	up_read(&lists_rwsem);
 	return device;
 }
 
+/**
+ * ib_device_put - Release IB device reference
+ * @device: device whose reference to be released
+ *
+ * ib_device_put() releases reference to the IB device to allow it to be
+ * unregistered and eventually free.
+ */
 void ib_device_put(struct ib_device *device)
 {
 	if (refcount_dec_and_test(&device->refcount))
 		complete(&device->unreg_completion);
 }
+EXPORT_SYMBOL(ib_device_put);
 
 static struct ib_device *__ib_device_get_by_name(const char *name)
 {
@@ -303,7 +310,6 @@ struct ib_device *ib_alloc_device(size_t size)
 	rwlock_init(&device->client_data_lock);
 	INIT_LIST_HEAD(&device->client_data_list);
 	INIT_LIST_HEAD(&device->port_list);
-	refcount_set(&device->refcount, 1);
 	init_completion(&device->unreg_completion);
 
 	return device;
@@ -620,6 +626,7 @@ int ib_register_device(struct ib_device *device, const char *name,
 		goto cg_cleanup;
 	}
 
+	refcount_set(&device->refcount, 1);
 	device->reg_state = IB_DEV_REGISTERED;
 
 	list_for_each_entry(client, &client_list, list)

drivers/infiniband/core/umem_odp.c

@@ -352,6 +352,8 @@ struct ib_umem_odp *ib_alloc_odp_umem(struct ib_ucontext_per_mm *per_mm,
 	umem->writable   = 1;
 	umem->is_odp = 1;
 	odp_data->per_mm = per_mm;
+	umem->owning_mm  = per_mm->mm;
+	mmgrab(umem->owning_mm);
 
 	mutex_init(&odp_data->umem_mutex);
 	init_completion(&odp_data->notifier_completion);
@@ -384,6 +386,7 @@ struct ib_umem_odp *ib_alloc_odp_umem(struct ib_ucontext_per_mm *per_mm,
 out_page_list:
 	vfree(odp_data->page_list);
 out_odp_data:
+	mmdrop(umem->owning_mm);
 	kfree(odp_data);
 	return ERR_PTR(ret);
 }

drivers/infiniband/core/uverbs_main.c

@@ -204,6 +204,9 @@ void ib_uverbs_release_file(struct kref *ref)
 	if (atomic_dec_and_test(&file->device->refcount))
 		ib_uverbs_comp_dev(file->device);
 
+	if (file->async_file)
+		kref_put(&file->async_file->ref,
+			 ib_uverbs_release_async_event_file);
 	put_device(&file->device->dev);
 	kfree(file);
 }
@@ -964,11 +967,19 @@ void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile)
 
 		/* Get an arbitrary mm pointer that hasn't been cleaned yet */
 		mutex_lock(&ufile->umap_lock);
-		if (!list_empty(&ufile->umaps)) {
-			mm = list_first_entry(&ufile->umaps,
-					      struct rdma_umap_priv, list)
-				     ->vma->vm_mm;
-			mmget(mm);
+		while (!list_empty(&ufile->umaps)) {
+			int ret;
+
+			priv = list_first_entry(&ufile->umaps,
+						struct rdma_umap_priv, list);
+			mm = priv->vma->vm_mm;
+			ret = mmget_not_zero(mm);
+			if (!ret) {
+				list_del_init(&priv->list);
+				mm = NULL;
+				continue;
+			}
+			break;
 		}
 		mutex_unlock(&ufile->umap_lock);
 		if (!mm)
@@ -1096,10 +1107,6 @@ static int ib_uverbs_close(struct inode *inode, struct file *filp)
 	list_del_init(&file->list);
 	mutex_unlock(&file->device->lists_mutex);
 
-	if (file->async_file)
-		kref_put(&file->async_file->ref,
-			 ib_uverbs_release_async_event_file);
-
 	kref_put(&file->ref, ib_uverbs_release_file);
 
 	return 0;

drivers/infiniband/core/uverbs_std_types_device.c

@@ -168,12 +168,18 @@ void copy_port_attr_to_resp(struct ib_port_attr *attr,
 static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_PORT)(
 	struct uverbs_attr_bundle *attrs)
 {
-	struct ib_device *ib_dev = attrs->ufile->device->ib_dev;
+	struct ib_device *ib_dev;
 	struct ib_port_attr attr = {};
 	struct ib_uverbs_query_port_resp_ex resp = {};
+	struct ib_ucontext *ucontext;
 	int ret;
 	u8 port_num;
 
+	ucontext = ib_uverbs_get_ucontext(attrs);
+	if (IS_ERR(ucontext))
+		return PTR_ERR(ucontext);
+	ib_dev = ucontext->device;
+
 	/* FIXME: Extend the UAPI_DEF_OBJ_NEEDS_FN stuff.. */
 	if (!ib_dev->ops.query_port)
 		return -EOPNOTSUPP;

drivers/infiniband/hw/hfi1/file_ops.c

@@ -488,7 +488,7 @@ static int hfi1_file_mmap(struct file *fp, struct vm_area_struct *vma)
 		vmf = 1;
 		break;
 	case STATUS:
-		if (flags & (unsigned long)(VM_WRITE | VM_EXEC)) {
+		if (flags & VM_WRITE) {
 			ret = -EPERM;
 			goto done;
 		}

drivers/infiniband/hw/hfi1/ud.c

@@ -987,7 +987,6 @@ void hfi1_ud_rcv(struct hfi1_packet *packet)
 	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
 		wc.ex.imm_data = packet->ohdr->u.ud.imm_data;
 		wc.wc_flags = IB_WC_WITH_IMM;
-		tlen -= sizeof(u32);
 	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
 		wc.ex.imm_data = 0;
 		wc.wc_flags = 0;

drivers/infiniband/hw/hns/hns_roce_srq.c

@@ -210,6 +210,7 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd,
 				   struct ib_udata *udata)
 {
 	struct hns_roce_dev *hr_dev = to_hr_dev(pd->device);
+	struct hns_roce_ib_create_srq_resp resp = {};
 	struct hns_roce_srq *srq;
 	int srq_desc_size;
 	int srq_buf_size;
@@ -378,16 +379,21 @@ struct ib_srq *hns_roce_create_srq(struct ib_pd *pd,
 
 	srq->event = hns_roce_ib_srq_event;
 	srq->ibsrq.ext.xrc.srq_num = srq->srqn;
+	resp.srqn = srq->srqn;
 
 	if (udata) {
-		if (ib_copy_to_udata(udata, &srq->srqn, sizeof(__u32))) {
+		if (ib_copy_to_udata(udata, &resp,
+				     min(udata->outlen, sizeof(resp)))) {
 			ret = -EFAULT;
-			goto err_wrid;
+			goto err_srqc_alloc;
 		}
 	}
 
 	return &srq->ibsrq;
 
+err_srqc_alloc:
+	hns_roce_srq_free(hr_dev, srq);
+
 err_wrid:
 	kvfree(srq->wrid);
 

drivers/infiniband/hw/mlx4/mad.c

@@ -1411,7 +1411,7 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
 
 	sqp_mad = (struct mlx4_mad_snd_buf *) (sqp->tx_ring[wire_tx_ix].buf.addr);
 	if (sqp->tx_ring[wire_tx_ix].ah)
-		rdma_destroy_ah(sqp->tx_ring[wire_tx_ix].ah, 0);
+		mlx4_ib_destroy_ah(sqp->tx_ring[wire_tx_ix].ah, 0);
 	sqp->tx_ring[wire_tx_ix].ah = ah;
 	ib_dma_sync_single_for_cpu(&dev->ib_dev,
 				   sqp->tx_ring[wire_tx_ix].buf.map,
@@ -1902,7 +1902,7 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work)
 		if (wc.status == IB_WC_SUCCESS) {
 			switch (wc.opcode) {
 			case IB_WC_SEND:
-				rdma_destroy_ah(sqp->tx_ring[wc.wr_id &
+				mlx4_ib_destroy_ah(sqp->tx_ring[wc.wr_id &
 					      (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0);
 				sqp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah
 					= NULL;
@@ -1931,7 +1931,7 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work)
 				 " status = %d, wrid = 0x%llx\n",
 				 ctx->slave, wc.status, wc.wr_id);
 			if (!MLX4_TUN_IS_RECV(wc.wr_id)) {
-				rdma_destroy_ah(sqp->tx_ring[wc.wr_id &
+				mlx4_ib_destroy_ah(sqp->tx_ring[wc.wr_id &
 					      (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0);
 				sqp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah
 					= NULL;

drivers/infiniband/hw/mlx5/flow.c

@@ -630,8 +630,7 @@ const struct uapi_definition mlx5_ib_flow_defs[] = {
 		UAPI_DEF_IS_OBJ_SUPPORTED(flow_is_supported)),
 	UAPI_DEF_CHAIN_OBJ_TREE(
 		UVERBS_OBJECT_FLOW,
-		&mlx5_ib_fs,
-		UAPI_DEF_IS_OBJ_SUPPORTED(flow_is_supported)),
+		&mlx5_ib_fs),
 	UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_FLOW_ACTION,
 				&mlx5_ib_flow_actions),
 	{},

drivers/infiniband/hw/mlx5/odp.c

@@ -1595,10 +1595,12 @@ static void mlx5_ib_prefetch_mr_work(struct work_struct *work)
 	struct prefetch_mr_work *w =
 		container_of(work, struct prefetch_mr_work, work);
 
-	if (w->dev->ib_dev.reg_state == IB_DEV_REGISTERED)
+	if (ib_device_try_get(&w->dev->ib_dev)) {
 		mlx5_ib_prefetch_sg_list(w->dev, w->pf_flags, w->sg_list,
 					 w->num_sge);
-
+		ib_device_put(&w->dev->ib_dev);
+	}
+	put_device(&w->dev->ib_dev.dev);
 	kfree(w);
 }
 
@@ -1617,15 +1619,13 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
 		return mlx5_ib_prefetch_sg_list(dev, pf_flags, sg_list,
 						num_sge);
 
-	if (dev->ib_dev.reg_state != IB_DEV_REGISTERED)
-		return -ENODEV;
-
 	work = kvzalloc(struct_size(work, sg_list, num_sge), GFP_KERNEL);
 	if (!work)
 		return -ENOMEM;
 
 	memcpy(work->sg_list, sg_list, num_sge * sizeof(struct ib_sge));
 
+	get_device(&dev->ib_dev.dev);
 	work->dev = dev;
 	work->pf_flags = pf_flags;
 	work->num_sge = num_sge;

drivers/infiniband/hw/mlx5/qp.c

@@ -1912,14 +1912,16 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 		}
 
 		if (!check_flags_mask(ucmd.flags,
+				      MLX5_QP_FLAG_ALLOW_SCATTER_CQE |
+				      MLX5_QP_FLAG_BFREG_INDEX |
+				      MLX5_QP_FLAG_PACKET_BASED_CREDIT_MODE |
+				      MLX5_QP_FLAG_SCATTER_CQE |
 				      MLX5_QP_FLAG_SIGNATURE |
-					      MLX5_QP_FLAG_SCATTER_CQE |
-					      MLX5_QP_FLAG_TUNNEL_OFFLOADS |
-					      MLX5_QP_FLAG_BFREG_INDEX |
-					      MLX5_QP_FLAG_TYPE_DCT |
-					      MLX5_QP_FLAG_TYPE_DCI |
-					      MLX5_QP_FLAG_ALLOW_SCATTER_CQE |
-					      MLX5_QP_FLAG_PACKET_BASED_CREDIT_MODE))
+				      MLX5_QP_FLAG_TIR_ALLOW_SELF_LB_MC |
+				      MLX5_QP_FLAG_TIR_ALLOW_SELF_LB_UC |
+				      MLX5_QP_FLAG_TUNNEL_OFFLOADS |
+				      MLX5_QP_FLAG_TYPE_DCI |
+				      MLX5_QP_FLAG_TYPE_DCT))
 			return -EINVAL;
 
 		err = get_qp_user_index(to_mucontext(pd->uobject->context),

drivers/infiniband/hw/qib/qib_ud.c

@@ -512,7 +512,6 @@ void qib_ud_rcv(struct qib_ibport *ibp, struct ib_header *hdr,
 	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
 		wc.ex.imm_data = ohdr->u.ud.imm_data;
 		wc.wc_flags = IB_WC_WITH_IMM;
-		tlen -= sizeof(u32);
 	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
 		wc.ex.imm_data = 0;
 		wc.wc_flags = 0;

drivers/infiniband/sw/rdmavt/qp.c

@@ -2910,6 +2910,8 @@ void rvt_ruc_loopback(struct rvt_qp *sqp)
 			goto op_err;
 		if (!ret)
 			goto rnr_nak;
+		if (wqe->length > qp->r_len)
+			goto inv_err;
 		break;
 
 	case IB_WR_RDMA_WRITE_WITH_IMM:
@@ -3078,7 +3080,10 @@ void rvt_ruc_loopback(struct rvt_qp *sqp)
 	goto err;
 
 inv_err:
-	send_status = IB_WC_REM_INV_REQ_ERR;
+	send_status =
+		sqp->ibqp.qp_type == IB_QPT_RC ?
+			IB_WC_REM_INV_REQ_ERR :
+			IB_WC_SUCCESS;
 	wc.status = IB_WC_LOC_QP_OP_ERR;
 	goto err;
 

drivers/infiniband/ulp/ipoib/ipoib.h

@@ -248,7 +248,6 @@ struct ipoib_cm_tx {
 	struct list_head     list;
 	struct net_device   *dev;
 	struct ipoib_neigh  *neigh;
-	struct ipoib_path   *path;
 	struct ipoib_tx_buf *tx_ring;
 	unsigned int	     tx_head;
 	unsigned int	     tx_tail;

drivers/infiniband/ulp/ipoib/ipoib_cm.c

@@ -1312,7 +1312,6 @@ struct ipoib_cm_tx *ipoib_cm_create_tx(struct net_device *dev, struct ipoib_path
 
 	neigh->cm = tx;
 	tx->neigh = neigh;
-	tx->path = path;
 	tx->dev = dev;
 	list_add(&tx->list, &priv->cm.start_list);
 	set_bit(IPOIB_FLAG_INITIALIZED, &tx->flags);
@@ -1371,7 +1370,7 @@ static void ipoib_cm_tx_start(struct work_struct *work)
 				neigh->daddr + QPN_AND_OPTIONS_OFFSET);
 			goto free_neigh;
 		}
-		memcpy(&pathrec, &p->path->pathrec, sizeof(pathrec));
+		memcpy(&pathrec, &path->pathrec, sizeof(pathrec));
 
 		spin_unlock_irqrestore(&priv->lock, flags);
 		netif_tx_unlock_bh(dev);

include/rdma/ib_verbs.h

@@ -2579,9 +2579,10 @@ struct ib_device {
 
 	const struct uapi_definition   *driver_def;
 	enum rdma_driver_id		driver_id;
+
 	/*
-	 * Provides synchronization between device unregistration and netlink
-	 * commands on a device. To be used only by core.
+	 * Positive refcount indicates that the device is currently
+	 * registered and cannot be unregistered.
 	 */
 	refcount_t refcount;
 	struct completion unreg_completion;
@@ -3926,6 +3927,25 @@ static inline bool ib_access_writable(int access_flags)
 int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
 		       struct ib_mr_status *mr_status);
 
+/**
+ * ib_device_try_get: Hold a registration lock
+ * device: The device to lock
+ *
+ * A device under an active registration lock cannot become unregistered. It
+ * is only possible to obtain a registration lock on a device that is fully
+ * registered, otherwise this function returns false.
+ *
+ * The registration lock is only necessary for actions which require the
+ * device to still be registered. Uses that only require the device pointer to
+ * be valid should use get_device(&ibdev->dev) to hold the memory.
+ *
+ */
+static inline bool ib_device_try_get(struct ib_device *dev)
+{
+	return refcount_inc_not_zero(&dev->refcount);
+}
+
+void ib_device_put(struct ib_device *device);
 struct net_device *ib_get_net_dev_by_params(struct ib_device *dev, u8 port,
 					    u16 pkey, const union ib_gid *gid,
 					    const struct sockaddr *addr);

include/uapi/rdma/hns-abi.h

@@ -52,6 +52,11 @@ struct hns_roce_ib_create_srq {
 	__aligned_u64 que_addr;
 };
 
+struct hns_roce_ib_create_srq_resp {
+	__u32	srqn;
+	__u32	reserved;
+};
+
 struct hns_roce_ib_create_qp {
 	__aligned_u64 buf_addr;
 	__aligned_u64 db_addr;