RDMA fourth 5.8 rc pull request

Two more merge window regressions, a corruption bug in hfi1 and a few
 other small fixes.
 
 - Missing user input validation regression in ucma
 
 - Disallowing a previously allowed user combination regression in mlx5
 
 - ODP prefetch memory leaking triggerable by userspace
 
 - Memory corruption in hf1 due to faulty ring buffer logic
 
 - Missed mutex initialization crash in mlx5
 
 - Two small defects with RDMA DIM
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEEfB7FMLh+8QxL+6i3OG33FX4gmxoFAl8kNQUACgkQOG33FX4g
 mxr+hg/7Bpr+77pMHGdVdEDLK+niN03W83fIMulONBpFo0/uuMFwuLdYapjBndro
 zhx9mCJXIBWNtEqxi82mq4BQEAH4MS6Jb9oZqSu77d6TvJg08fiBCFY7UojAoaKs
 pXT3Cp+GAu6WrDWc65MYnePYnt/4PB/bw2lPhHRjntsBhTN1xj65VUkHwOCIpKdQ
 ESuWk1Pc0OlLBcPjK1WKNfRbB0xCrvMZLVK7NVDBj/XshdsBZFJvByrL7L10c/zs
 qIrsrIjLQW7X2xT+av5sX3TfUzBb6SPcfjIKveC6onctman/oS++ZwHcnKxkYjlp
 SPOYCTaUVFOVwXGz17nuiCul9rkq6FJm8xa08+rJ4MoDxMBLiWvaqn6Not+8uJF/
 sO/wU/H+mgPkYwmpt0JNblGkgbs+a3Gs+PE8NM5a9euOTSNV21el/x4rv/QSAyqK
 KxS9b9/vkSZzmYQucds9EM2R70PKGr2VtRdL7f4nONCNoPL0No2eMrIR2QzXjhsC
 yydOIn5cUpvClLg30Bs1njkp7YtO/2wZRGy8RpI/9oX4/tqWiQcw+FK0BGA3lzVl
 kE6wrFAwnnhqSYjFimTMlV/2GA7wlkewuDGBHcqaGEcCVw/d7juzVaan5ciA8l4C
 WRIBo+DQYNkGOzBRJBs65JTMKfkHJrqevmDor3m15JB7Uk4Qacc=
 =0t3P
 -----END PGP SIGNATURE-----

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

Pull rdma fixes from Jason Gunthorpe:
 "Two more merge window regressions, a corruption bug in hfi1 and a few
  other small fixes.

   - Missing user input validation regression in ucma

   - Disallowing a previously allowed user combination regression in
     mlx5

   - ODP prefetch memory leaking triggerable by userspace

   - Memory corruption in hf1 due to faulty ring buffer logic

   - Missed mutex initialization crash in mlx5

   - Two small defects with RDMA DIM"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/core: Free DIM memory in error unwind
  RDMA/core: Stop DIM before destroying CQ
  RDMA/mlx5: Initialize QP mutex for the debug kernels
  IB/rdmavt: Fix RQ counting issues causing use of an invalid RWQE
  RDMA/mlx5: Allow providing extra scatter CQE QP flag
  RDMA/mlx5: Fix prefetch memory leak if get_prefetchable_mr fails
  RDMA/cm: Add min length checks to user structure copies

drivers/infiniband/core/cq.c

@@ -72,6 +72,15 @@ static void rdma_dim_init(struct ib_cq *cq)
 	INIT_WORK(&dim->work, ib_cq_rdma_dim_work);
 }
 
+static void rdma_dim_destroy(struct ib_cq *cq)
+{
+	if (!cq->dim)
+		return;
+
+	cancel_work_sync(&cq->dim->work);
+	kfree(cq->dim);
+}
+
 static int __poll_cq(struct ib_cq *cq, int num_entries, struct ib_wc *wc)
 {
 	int rc;
@@ -266,6 +275,7 @@ struct ib_cq *__ib_alloc_cq_user(struct ib_device *dev, void *private,
 	return cq;
 
 out_destroy_cq:
+	rdma_dim_destroy(cq);
 	rdma_restrack_del(&cq->res);
 	cq->device->ops.destroy_cq(cq, udata);
 out_free_wc:
@@ -331,12 +341,10 @@ void ib_free_cq_user(struct ib_cq *cq, struct ib_udata *udata)
 		WARN_ON_ONCE(1);
 	}
 
+	rdma_dim_destroy(cq);
 	trace_cq_free(cq);
 	rdma_restrack_del(&cq->res);
 	cq->device->ops.destroy_cq(cq, udata);
-	if (cq->dim)
-		cancel_work_sync(&cq->dim->work);
-	kfree(cq->dim);
 	kfree(cq->wc);
 	kfree(cq);
 }

drivers/infiniband/core/ucma.c

@@ -1084,6 +1084,8 @@ static ssize_t ucma_connect(struct ucma_file *file, const char __user *inbuf,
 	size_t in_size;
 	int ret;
 
+	if (in_len < offsetofend(typeof(cmd), reserved))
+		return -EINVAL;
 	in_size = min_t(size_t, in_len, sizeof(cmd));
 	if (copy_from_user(&cmd, inbuf, in_size))
 		return -EFAULT;
@@ -1141,6 +1143,8 @@ static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
 	size_t in_size;
 	int ret;
 
+	if (in_len < offsetofend(typeof(cmd), reserved))
+		return -EINVAL;
 	in_size = min_t(size_t, in_len, sizeof(cmd));
 	if (copy_from_user(&cmd, inbuf, in_size))
 		return -EFAULT;

drivers/infiniband/hw/mlx5/odp.c

@@ -1797,9 +1797,7 @@ static bool init_prefetch_work(struct ib_pd *pd,
 		work->frags[i].mr =
 			get_prefetchable_mr(pd, advice, sg_list[i].lkey);
 		if (!work->frags[i].mr) {
-			work->num_sge = i - 1;
-			if (i)
-				destroy_prefetch_work(work);
+			work->num_sge = i;
 			return false;
 		}
 
@@ -1865,6 +1863,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
 	srcu_key = srcu_read_lock(&dev->odp_srcu);
 	if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
 		srcu_read_unlock(&dev->odp_srcu, srcu_key);
+		destroy_prefetch_work(work);
 		return -EINVAL;
 	}
 	queue_work(system_unbound_wq, &work->work);

drivers/infiniband/hw/mlx5/qp.c

@@ -1766,15 +1766,14 @@ err:
 }
 
 static void configure_requester_scat_cqe(struct mlx5_ib_dev *dev,
+					 struct mlx5_ib_qp *qp,
 					 struct ib_qp_init_attr *init_attr,
-					 struct mlx5_ib_create_qp *ucmd,
 					 void *qpc)
 {
 	int scqe_sz;
 	bool allow_scat_cqe = false;
 
-	if (ucmd)
-		allow_scat_cqe = ucmd->flags & MLX5_QP_FLAG_ALLOW_SCATTER_CQE;
+	allow_scat_cqe = qp->flags_en & MLX5_QP_FLAG_ALLOW_SCATTER_CQE;
 
 	if (!allow_scat_cqe && init_attr->sq_sig_type != IB_SIGNAL_ALL_WR)
 		return;
@@ -1853,8 +1852,6 @@ static int create_xrc_tgt_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	u32 *in;
 	int err;
 
-	mutex_init(&qp->mutex);
-
 	if (attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		qp->sq_signal_bits = MLX5_WQE_CTRL_CQ_UPDATE;
 
@@ -1938,7 +1935,6 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	u32 *in;
 	int err;
 
-	mutex_init(&qp->mutex);
 	spin_lock_init(&qp->sq.lock);
 	spin_lock_init(&qp->rq.lock);
 
@@ -2012,7 +2008,7 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	}
 	if ((qp->flags_en & MLX5_QP_FLAG_SCATTER_CQE) &&
 	    (qp->type == MLX5_IB_QPT_DCI || qp->type == IB_QPT_RC))
-		configure_requester_scat_cqe(dev, init_attr, ucmd, qpc);
+		configure_requester_scat_cqe(dev, qp, init_attr, qpc);
 
 	if (qp->rq.wqe_cnt) {
 		MLX5_SET(qpc, qpc, log_rq_stride, qp->rq.wqe_shift - 4);
@@ -2129,7 +2125,6 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	u32 *in;
 	int err;
 
-	mutex_init(&qp->mutex);
 	spin_lock_init(&qp->sq.lock);
 	spin_lock_init(&qp->rq.lock);
 
@@ -2543,13 +2538,18 @@ static void process_vendor_flag(struct mlx5_ib_dev *dev, int *flags, int flag,
 		return;
 	}
 
-	if (flag == MLX5_QP_FLAG_SCATTER_CQE) {
+	switch (flag) {
+	case MLX5_QP_FLAG_SCATTER_CQE:
+	case MLX5_QP_FLAG_ALLOW_SCATTER_CQE:
 		/*
-		 * We don't return error if this flag was provided,
-		 * and mlx5 doesn't have right capability.
-		 */
-		*flags &= ~MLX5_QP_FLAG_SCATTER_CQE;
+			 * We don't return error if these flags were provided,
+			 * and mlx5 doesn't have right capability.
+			 */
+		*flags &= ~(MLX5_QP_FLAG_SCATTER_CQE |
+			    MLX5_QP_FLAG_ALLOW_SCATTER_CQE);
 		return;
+	default:
+		break;
 	}
 	mlx5_ib_dbg(dev, "Vendor create QP flag 0x%X is not supported\n", flag);
 }
@@ -2589,6 +2589,8 @@ static int process_vendor_flags(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	process_vendor_flag(dev, &flags, MLX5_QP_FLAG_SIGNATURE, true, qp);
 	process_vendor_flag(dev, &flags, MLX5_QP_FLAG_SCATTER_CQE,
 			    MLX5_CAP_GEN(mdev, sctr_data_cqe), qp);
+	process_vendor_flag(dev, &flags, MLX5_QP_FLAG_ALLOW_SCATTER_CQE,
+			    MLX5_CAP_GEN(mdev, sctr_data_cqe), qp);
 
 	if (qp->type == IB_QPT_RAW_PACKET) {
 		cond = MLX5_CAP_ETH(mdev, tunnel_stateless_vxlan) ||
@@ -2963,6 +2965,7 @@ struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attr,
 		goto free_ucmd;
 	}
 
+	mutex_init(&qp->mutex);
 	qp->type = type;
 	if (udata) {
 		err = process_vendor_flags(dev, qp, params.ucmd, attr);

drivers/infiniband/sw/rdmavt/qp.c

@@ -901,8 +901,6 @@ static void rvt_init_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp,
 	qp->s_tail_ack_queue = 0;
 	qp->s_acked_ack_queue = 0;
 	qp->s_num_rd_atomic = 0;
-	if (qp->r_rq.kwq)
-		qp->r_rq.kwq->count = qp->r_rq.size;
 	qp->r_sge.num_sge = 0;
 	atomic_set(&qp->s_reserved_used, 0);
 }
@@ -2366,31 +2364,6 @@ bad_lkey:
 	return 0;
 }
 
-/**
- * get_count - count numbers of request work queue entries
- * in circular buffer
- * @rq: data structure for request queue entry
- * @tail: tail indices of the circular buffer
- * @head: head indices of the circular buffer
- *
- * Return - total number of entries in the circular buffer
- */
-static u32 get_count(struct rvt_rq *rq, u32 tail, u32 head)
-{
-	u32 count;
-
-	count = head;
-
-	if (count >= rq->size)
-		count = 0;
-	if (count < tail)
-		count += rq->size - tail;
-	else
-		count -= tail;
-
-	return count;
-}
-
 /**
  * get_rvt_head - get head indices of the circular buffer
  * @rq: data structure for request queue entry
@@ -2465,7 +2438,7 @@ int rvt_get_rwqe(struct rvt_qp *qp, bool wr_id_only)
 
 	if (kwq->count < RVT_RWQ_COUNT_THRESHOLD) {
 		head = get_rvt_head(rq, ip);
-		kwq->count = get_count(rq, tail, head);
+		kwq->count = rvt_get_rq_count(rq, head, tail);
 	}
 	if (unlikely(kwq->count == 0)) {
 		ret = 0;
@@ -2500,7 +2473,9 @@ int rvt_get_rwqe(struct rvt_qp *qp, bool wr_id_only)
 		 * the number of remaining WQEs.
 		 */
 		if (kwq->count < srq->limit) {
-			kwq->count = get_count(rq, tail, get_rvt_head(rq, ip));
+			kwq->count =
+				rvt_get_rq_count(rq,
+						 get_rvt_head(rq, ip), tail);
 			if (kwq->count < srq->limit) {
 				struct ib_event ev;
 

drivers/infiniband/sw/rdmavt/rc.c

@@ -127,9 +127,7 @@ __be32 rvt_compute_aeth(struct rvt_qp *qp)
 			 * not atomic, which is OK, since the fuzziness is
 			 * resolved as further ACKs go out.
 			 */
-			credits = head - tail;
-			if ((int)credits < 0)
-				credits += qp->r_rq.size;
+			credits = rvt_get_rq_count(&qp->r_rq, head, tail);
 		}
 		/*
 		 * Binary search the credit table to find the code to

include/rdma/rdmavt_qp.h

@@ -305,6 +305,25 @@ struct rvt_rq {
 	spinlock_t lock ____cacheline_aligned_in_smp;
 };
 
+/**
+ * rvt_get_rq_count - count numbers of request work queue entries
+ * in circular buffer
+ * @rq: data structure for request queue entry
+ * @head: head indices of the circular buffer
+ * @tail: tail indices of the circular buffer
+ *
+ * Return - total number of entries in the Receive Queue
+ */
+
+static inline u32 rvt_get_rq_count(struct rvt_rq *rq, u32 head, u32 tail)
+{
+	u32 count = head - tail;
+
+	if ((s32)count < 0)
+		count += rq->size;
+	return count;
+}
+
 /*
  * This structure holds the information that the send tasklet needs
  * to send a RDMA read response or atomic operation.