Merge branch 'devx-async' into k.o/for-next

Yishai Hadas says:

Enable DEVX asynchronous query commands

This series enables querying a DEVX object in an asynchronous mode.

The userspace application won't block when calling the firmware and it will be
able to get the response back once that it will be ready.

To enable the above functionality:

- DEVX asynchronous command completion FD object was introduced.
- The applicable file operations were implemented to enable using it by
  the user application.
- Query asynchronous method was added to the DEVX object, it will call the
  firmware asynchronously and manages the response on the given input FD.
- Hot unplug support was added for the FD to work properly upon
  unbind/disassociate.
- mlx5 core fence for asynchronous commands was implemented and used to
  prevent racing upon unbind/disassociate.

This branch is based on mlx5-next & v5.0-rc2 due to dependencies, from
git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux

* branch 'devx-async':
  IB/mlx5: Implement DEVX hot unplug for async command FD
  IB/mlx5: Implement the file ops of DEVX async command FD
  IB/mlx5: Introduce async DEVX obj query API
  IB/mlx5: Introduce MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD

Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

22 files changed, 443 insertions, 119 deletions

diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
index 742346ea5b0d..65e17de220f6 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
@@ -765,8 +765,8 @@ struct bnxt_qplib_rcfw_sbuf *bnxt_qplib_rcfw_alloc_sbuf(
 		return NULL;
 
 	sbuf->size = size;
-	sbuf->sb = dma_zalloc_coherent(&rcfw->pdev->dev, sbuf->size,
-				       &sbuf->dma_addr, GFP_ATOMIC);
+	sbuf->sb = dma_alloc_coherent(&rcfw->pdev->dev, sbuf->size,
+				      &sbuf->dma_addr, GFP_ATOMIC);
 	if (!sbuf->sb)
 		goto bail;
 
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_res.c b/drivers/infiniband/hw/bnxt_re/qplib_res.c
index 59eeac55626f..57d4951679cb 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_res.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_res.c
@@ -105,10 +105,10 @@ static int __alloc_pbl(struct pci_dev *pdev, struct bnxt_qplib_pbl *pbl,
 
 	if (!sghead) {
 		for (i = 0; i < pages; i++) {
-			pbl->pg_arr[i] = dma_zalloc_coherent(&pdev->dev,
-							     pbl->pg_size,
-							     &pbl->pg_map_arr[i],
-							     GFP_KERNEL);
+			pbl->pg_arr[i] = dma_alloc_coherent(&pdev->dev,
+							    pbl->pg_size,
+							    &pbl->pg_map_arr[i],
+							    GFP_KERNEL);
 			if (!pbl->pg_arr[i])
 				goto fail;
 			pbl->pg_count++;
diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c
index df4f7a3f043d..8ac72ac7cbac 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.c
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c
@@ -291,9 +291,9 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain,
 	if (!wq->sq)
 		goto err3;
 
-	wq->queue = dma_zalloc_coherent(&(rdev_p->rnic_info.pdev->dev),
-					     depth * sizeof(union t3_wr),
-					     &(wq->dma_addr), GFP_KERNEL);
+	wq->queue = dma_alloc_coherent(&(rdev_p->rnic_info.pdev->dev),
+				       depth * sizeof(union t3_wr),
+				       &(wq->dma_addr), GFP_KERNEL);
 	if (!wq->queue)
 		goto err4;
 
diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c
index c00a41144126..0fe87b9c1e10 100644
--- a/drivers/infiniband/hw/cxgb4/qp.c
+++ b/drivers/infiniband/hw/cxgb4/qp.c
@@ -2571,9 +2571,8 @@ static int alloc_srq_queue(struct c4iw_srq *srq, struct c4iw_dev_ucontext *uctx,
 	wq->rqt_abs_idx = (wq->rqt_hwaddr - rdev->lldi.vr->rq.start) >>
 		T4_RQT_ENTRY_SHIFT;
 
-	wq->queue = dma_zalloc_coherent(&rdev->lldi.pdev->dev,
-				       wq->memsize, &wq->dma_addr,
-			GFP_KERNEL);
+	wq->queue = dma_alloc_coherent(&rdev->lldi.pdev->dev, wq->memsize,
+				       &wq->dma_addr, GFP_KERNEL);
 	if (!wq->queue)
 		goto err_free_rqtpool;
 
diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c
index 09044905284f..7835eb52e7c5 100644
--- a/drivers/infiniband/hw/hfi1/init.c
+++ b/drivers/infiniband/hw/hfi1/init.c
@@ -899,10 +899,10 @@ int hfi1_init(struct hfi1_devdata *dd, int reinit)
 		goto done;
 
 	/* allocate dummy tail memory for all receive contexts */
-	dd->rcvhdrtail_dummy_kvaddr = dma_zalloc_coherent(
-		&dd->pcidev->dev, sizeof(u64),
-		&dd->rcvhdrtail_dummy_dma,
-		GFP_KERNEL);
+	dd->rcvhdrtail_dummy_kvaddr = dma_alloc_coherent(&dd->pcidev->dev,
+							 sizeof(u64),
+							 &dd->rcvhdrtail_dummy_dma,
+							 GFP_KERNEL);
 
 	if (!dd->rcvhdrtail_dummy_kvaddr) {
 		dd_dev_err(dd, "cannot allocate dummy tail memory\n");
@@ -1863,9 +1863,9 @@ int hfi1_create_rcvhdrq(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
 			gfp_flags = GFP_KERNEL;
 		else
 			gfp_flags = GFP_USER;
-		rcd->rcvhdrq = dma_zalloc_coherent(
-			&dd->pcidev->dev, amt, &rcd->rcvhdrq_dma,
-			gfp_flags | __GFP_COMP);
+		rcd->rcvhdrq = dma_alloc_coherent(&dd->pcidev->dev, amt,
+						  &rcd->rcvhdrq_dma,
+						  gfp_flags | __GFP_COMP);
 
 		if (!rcd->rcvhdrq) {
 			dd_dev_err(dd,
@@ -1876,9 +1876,10 @@ int hfi1_create_rcvhdrq(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
 
 		if (HFI1_CAP_KGET_MASK(rcd->flags, DMA_RTAIL) ||
 		    HFI1_CAP_UGET_MASK(rcd->flags, DMA_RTAIL)) {
-			rcd->rcvhdrtail_kvaddr = dma_zalloc_coherent(
-				&dd->pcidev->dev, PAGE_SIZE,
-				&rcd->rcvhdrqtailaddr_dma, gfp_flags);
+			rcd->rcvhdrtail_kvaddr = dma_alloc_coherent(&dd->pcidev->dev,
+								    PAGE_SIZE,
+								    &rcd->rcvhdrqtailaddr_dma,
+								    gfp_flags);
 			if (!rcd->rcvhdrtail_kvaddr)
 				goto bail_free;
 		}
@@ -1974,10 +1975,10 @@ int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *rcd)
 	while (alloced_bytes < rcd->egrbufs.size &&
 	       rcd->egrbufs.alloced < rcd->egrbufs.count) {
 		rcd->egrbufs.buffers[idx].addr =
-			dma_zalloc_coherent(&dd->pcidev->dev,
-					    rcd->egrbufs.rcvtid_size,
-					    &rcd->egrbufs.buffers[idx].dma,
-					    gfp_flags);
+			dma_alloc_coherent(&dd->pcidev->dev,
+					   rcd->egrbufs.rcvtid_size,
+					   &rcd->egrbufs.buffers[idx].dma,
+					   gfp_flags);
 		if (rcd->egrbufs.buffers[idx].addr) {
 			rcd->egrbufs.buffers[idx].len =
 				rcd->egrbufs.rcvtid_size;
diff --git a/drivers/infiniband/hw/hfi1/pio.c b/drivers/infiniband/hw/hfi1/pio.c
index dd5a5c030066..04126d7e318d 100644
--- a/drivers/infiniband/hw/hfi1/pio.c
+++ b/drivers/infiniband/hw/hfi1/pio.c
@@ -2098,11 +2098,10 @@ int init_credit_return(struct hfi1_devdata *dd)
 		int bytes = TXE_NUM_CONTEXTS * sizeof(struct credit_return);
 
 		set_dev_node(&dd->pcidev->dev, i);
-		dd->cr_base[i].va = dma_zalloc_coherent(
-					&dd->pcidev->dev,
-					bytes,
-					&dd->cr_base[i].dma,
-					GFP_KERNEL);
+		dd->cr_base[i].va = dma_alloc_coherent(&dd->pcidev->dev,
+						       bytes,
+						       &dd->cr_base[i].dma,
+						       GFP_KERNEL);
 		if (!dd->cr_base[i].va) {
 			set_dev_node(&dd->pcidev->dev, dd->node);
 			dd_dev_err(dd,
diff --git a/drivers/infiniband/hw/hfi1/sdma.c b/drivers/infiniband/hw/hfi1/sdma.c
index b84356e1a4c1..96897a91fb0a 100644
--- a/drivers/infiniband/hw/hfi1/sdma.c
+++ b/drivers/infiniband/hw/hfi1/sdma.c
@@ -1453,12 +1453,9 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 		timer_setup(&sde->err_progress_check_timer,
 			    sdma_err_progress_check, 0);
 
-		sde->descq = dma_zalloc_coherent(
-			&dd->pcidev->dev,
-			descq_cnt * sizeof(u64[2]),
-			&sde->descq_phys,
-			GFP_KERNEL
-		);
+		sde->descq = dma_alloc_coherent(&dd->pcidev->dev,
+						descq_cnt * sizeof(u64[2]),
+						&sde->descq_phys, GFP_KERNEL);
 		if (!sde->descq)
 			goto bail;
 		sde->tx_ring =
@@ -1471,24 +1468,18 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 
 	dd->sdma_heads_size = L1_CACHE_BYTES * num_engines;
 	/* Allocate memory for DMA of head registers to memory */
-	dd->sdma_heads_dma = dma_zalloc_coherent(
-		&dd->pcidev->dev,
-		dd->sdma_heads_size,
-		&dd->sdma_heads_phys,
-		GFP_KERNEL
-	);
+	dd->sdma_heads_dma = dma_alloc_coherent(&dd->pcidev->dev,
+						dd->sdma_heads_size,
+						&dd->sdma_heads_phys,
+						GFP_KERNEL);
 	if (!dd->sdma_heads_dma) {
 		dd_dev_err(dd, "failed to allocate SendDMA head memory\n");
 		goto bail;
 	}
 
 	/* Allocate memory for pad */
-	dd->sdma_pad_dma = dma_zalloc_coherent(
-		&dd->pcidev->dev,
-		sizeof(u32),
-		&dd->sdma_pad_phys,
-		GFP_KERNEL
-	);
+	dd->sdma_pad_dma = dma_alloc_coherent(&dd->pcidev->dev, sizeof(u32),
+					      &dd->sdma_pad_phys, GFP_KERNEL);
 	if (!dd->sdma_pad_dma) {
 		dd_dev_err(dd, "failed to allocate SendDMA pad memory\n");
 		goto bail;
diff --git a/drivers/infiniband/hw/hns/hns_roce_alloc.c b/drivers/infiniband/hw/hns/hns_roce_alloc.c
index 6300033a448f..dac058d3df53 100644
--- a/drivers/infiniband/hw/hns/hns_roce_alloc.c
+++ b/drivers/infiniband/hw/hns/hns_roce_alloc.c
@@ -197,8 +197,8 @@ int hns_roce_buf_alloc(struct hns_roce_dev *hr_dev, u32 size, u32 max_direct,
 		buf->npages = 1 << order;
 		buf->page_shift = page_shift;
 		/* MTT PA must be recorded in 4k alignment, t is 4k aligned */
-		buf->direct.buf = dma_zalloc_coherent(dev,
-						      size, &t, GFP_KERNEL);
+		buf->direct.buf = dma_alloc_coherent(dev, size, &t,
+						     GFP_KERNEL);
 		if (!buf->direct.buf)
 			return -ENOMEM;
 
@@ -219,9 +219,10 @@ int hns_roce_buf_alloc(struct hns_roce_dev *hr_dev, u32 size, u32 max_direct,
 			return -ENOMEM;
 
 		for (i = 0; i < buf->nbufs; ++i) {
-			buf->page_list[i].buf = dma_zalloc_coherent(dev,
-								  page_size, &t,
-								  GFP_KERNEL);
+			buf->page_list[i].buf = dma_alloc_coherent(dev,
+								   page_size,
+								   &t,
+								   GFP_KERNEL);
 
 			if (!buf->page_list[i].buf)
 				goto err_free;
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index d63dd155cd33..5c483b437bdd 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -5311,7 +5311,7 @@ static int hns_roce_mhop_alloc_eq(struct hns_roce_dev *hr_dev,
 				eqe_alloc = i * (buf_chk_sz / eq->eqe_size);
 				size = (eq->entries - eqe_alloc) * eq->eqe_size;
 			}
-			eq->buf[i] = dma_zalloc_coherent(dev, size,
+			eq->buf[i] = dma_alloc_coherent(dev, size,
 							&(eq->buf_dma[i]),
 							GFP_KERNEL);
 			if (!eq->buf[i])
@@ -5346,9 +5346,9 @@ static int hns_roce_mhop_alloc_eq(struct hns_roce_dev *hr_dev,
 					size = (eq->entries - eqe_alloc)
 						* eq->eqe_size;
 				}
-				eq->buf[idx] = dma_zalloc_coherent(dev, size,
-							    &(eq->buf_dma[idx]),
-							    GFP_KERNEL);
+				eq->buf[idx] = dma_alloc_coherent(dev, size,
+								  &(eq->buf_dma[idx]),
+								  GFP_KERNEL);
 				if (!eq->buf[idx])
 					goto err_dma_alloc_buf;
 
@@ -5461,7 +5461,7 @@ static int hns_roce_v2_create_eq(struct hns_roce_dev *hr_dev,
 			goto free_cmd_mbox;
 		}
 
-		eq->buf_list->buf = dma_zalloc_coherent(dev, buf_chk_sz,
+		eq->buf_list->buf = dma_alloc_coherent(dev, buf_chk_sz,
 						       &(eq->buf_list->map),
 						       GFP_KERNEL);
 		if (!eq->buf_list->buf) {
diff --git a/drivers/infiniband/hw/i40iw/i40iw_utils.c b/drivers/infiniband/hw/i40iw/i40iw_utils.c
index a9ea966877f2..59e978141ad4 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_utils.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_utils.c
@@ -745,8 +745,8 @@ enum i40iw_status_code i40iw_allocate_dma_mem(struct i40iw_hw *hw,
 	if (!mem)
 		return I40IW_ERR_PARAM;
 	mem->size = ALIGN(size, alignment);
-	mem->va = dma_zalloc_coherent(&pcidev->dev, mem->size,
-				      (dma_addr_t *)&mem->pa, GFP_KERNEL);
+	mem->va = dma_alloc_coherent(&pcidev->dev, mem->size,
+				     (dma_addr_t *)&mem->pa, GFP_KERNEL);
 	if (!mem->va)
 		return I40IW_ERR_NO_MEMORY;
 	return 0;
diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c
index 356bccc715ee..6bcc63aaa50b 100644
--- a/drivers/infiniband/hw/mlx5/cmd.c
+++ b/drivers/infiniband/hw/mlx5/cmd.c
@@ -345,3 +345,40 @@ int mlx5_cmd_alloc_q_counter(struct mlx5_core_dev *dev, u16 *counter_id,
 				       counter_set_id);
 	return err;
 }
+
+int mlx5_cmd_mad_ifc(struct mlx5_core_dev *dev, const void *inb, void *outb,
+		     u16 opmod, u8 port)
+{
+	int outlen = MLX5_ST_SZ_BYTES(mad_ifc_out);
+	int inlen = MLX5_ST_SZ_BYTES(mad_ifc_in);
+	int err = -ENOMEM;
+	void *data;
+	void *resp;
+	u32 *out;
+	u32 *in;
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	out = kzalloc(outlen, GFP_KERNEL);
+	if (!in || !out)
+		goto out;
+
+	MLX5_SET(mad_ifc_in, in, opcode, MLX5_CMD_OP_MAD_IFC);
+	MLX5_SET(mad_ifc_in, in, op_mod, opmod);
+	MLX5_SET(mad_ifc_in, in, port, port);
+
+	data = MLX5_ADDR_OF(mad_ifc_in, in, mad);
+	memcpy(data, inb, MLX5_FLD_SZ_BYTES(mad_ifc_in, mad));
+
+	err = mlx5_cmd_exec(dev, in, inlen, out, outlen);
+	if (err)
+		goto out;
+
+	resp = MLX5_ADDR_OF(mad_ifc_out, out, response_mad_packet);
+	memcpy(outb, resp,
+	       MLX5_FLD_SZ_BYTES(mad_ifc_out, response_mad_packet));
+
+out:
+	kfree(out);
+	kfree(in);
+	return err;
+}
diff --git a/drivers/infiniband/hw/mlx5/cmd.h b/drivers/infiniband/hw/mlx5/cmd.h
index 1e76dc67a369..923a7b93f507 100644
--- a/drivers/infiniband/hw/mlx5/cmd.h
+++ b/drivers/infiniband/hw/mlx5/cmd.h
@@ -63,4 +63,6 @@ int mlx5_cmd_xrcd_alloc(struct mlx5_core_dev *dev, u32 *xrcdn, u16 uid);
 int mlx5_cmd_xrcd_dealloc(struct mlx5_core_dev *dev, u32 xrcdn, u16 uid);
 int mlx5_cmd_alloc_q_counter(struct mlx5_core_dev *dev, u16 *counter_id,
 			     u16 uid);
+int mlx5_cmd_mad_ifc(struct mlx5_core_dev *dev, const void *inb, void *outb,
+		     u16 opmod, u8 port);
 #endif /* MLX5_IB_CMD_H */
diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
index bbf9a26d8fa6..12de63c63fa0 100644
--- a/drivers/infiniband/hw/mlx5/devx.c
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -8,6 +8,7 @@
 #include <rdma/uverbs_types.h>
 #include <rdma/uverbs_ioctl.h>
 #include <rdma/mlx5_user_ioctl_cmds.h>
+#include <rdma/mlx5_user_ioctl_verbs.h>
 #include <rdma/ib_umem.h>
 #include <rdma/uverbs_std_types.h>
 #include <linux/mlx5/driver.h>
@@ -21,6 +22,16 @@ enum devx_obj_flags {
 	DEVX_OBJ_FLAGS_INDIRECT_MKEY = 1 << 0,
 };
 
+struct devx_async_data {
+	struct mlx5_ib_dev *mdev;
+	struct list_head list;
+	struct ib_uobject *fd_uobj;
+	struct mlx5_async_work cb_work;
+	u16 cmd_out_len;
+	/* must be last field in this structure */
+	struct mlx5_ib_uapi_devx_async_cmd_hdr hdr;
+};
+
 #define MLX5_MAX_DESTROY_INBOX_SIZE_DW MLX5_ST_SZ_DW(delete_fte_in)
 struct devx_obj {
 	struct mlx5_core_dev	*mdev;
@@ -1289,6 +1300,153 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_QUERY)(
 			      cmd_out, cmd_out_len);
 }
 
+struct devx_async_event_queue {
+	spinlock_t		lock;
+	wait_queue_head_t	poll_wait;
+	struct list_head	event_list;
+	atomic_t		bytes_in_use;
+	u8			is_destroyed:1;
+};
+
+struct devx_async_cmd_event_file {
+	struct ib_uobject		uobj;
+	struct devx_async_event_queue	ev_queue;
+	struct mlx5_async_ctx		async_ctx;
+};
+
+static void devx_init_event_queue(struct devx_async_event_queue *ev_queue)
+{
+	spin_lock_init(&ev_queue->lock);
+	INIT_LIST_HEAD(&ev_queue->event_list);
+	init_waitqueue_head(&ev_queue->poll_wait);
+	atomic_set(&ev_queue->bytes_in_use, 0);
+	ev_queue->is_destroyed = 0;
+}
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct devx_async_cmd_event_file *ev_file;
+
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(
+		attrs, MLX5_IB_ATTR_DEVX_ASYNC_CMD_FD_ALLOC_HANDLE);
+	struct mlx5_ib_dev *mdev = to_mdev(uobj->context->device);
+
+	ev_file = container_of(uobj, struct devx_async_cmd_event_file,
+			       uobj);
+	devx_init_event_queue(&ev_file->ev_queue);
+	mlx5_cmd_init_async_ctx(mdev->mdev, &ev_file->async_ctx);
+	return 0;
+}
+
+static void devx_query_callback(int status, struct mlx5_async_work *context)
+{
+	struct devx_async_data *async_data =
+		container_of(context, struct devx_async_data, cb_work);
+	struct ib_uobject *fd_uobj = async_data->fd_uobj;
+	struct devx_async_cmd_event_file *ev_file;
+	struct devx_async_event_queue *ev_queue;
+	unsigned long flags;
+
+	ev_file = container_of(fd_uobj, struct devx_async_cmd_event_file,
+			       uobj);
+	ev_queue = &ev_file->ev_queue;
+
+	spin_lock_irqsave(&ev_queue->lock, flags);
+	list_add_tail(&async_data->list, &ev_queue->event_list);
+	spin_unlock_irqrestore(&ev_queue->lock, flags);
+
+	wake_up_interruptible(&ev_queue->poll_wait);
+	fput(fd_uobj->object);
+}
+
+#define MAX_ASYNC_BYTES_IN_USE (1024 * 1024) /* 1MB */
+
+static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY)(
+	struct uverbs_attr_bundle *attrs)
+{
+	void *cmd_in = uverbs_attr_get_alloced_ptr(attrs,
+				MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_CMD_IN);
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(
+				attrs,
+				MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_HANDLE);
+	u16 cmd_out_len;
+	struct mlx5_ib_ucontext *c = to_mucontext(uobj->context);
+	struct ib_uobject *fd_uobj;
+	int err;
+	int uid;
+	struct mlx5_ib_dev *mdev = to_mdev(uobj->context->device);
+	struct devx_async_cmd_event_file *ev_file;
+	struct devx_async_data *async_data;
+
+	uid = devx_get_uid(c, cmd_in);
+	if (uid < 0)
+		return uid;
+
+	if (!devx_is_obj_query_cmd(cmd_in))
+		return -EINVAL;
+
+	err = uverbs_get_const(&cmd_out_len, attrs,
+			       MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_OUT_LEN);
+	if (err)
+		return err;
+
+	if (!devx_is_valid_obj_id(uobj, cmd_in))
+		return -EINVAL;
+
+	fd_uobj = uverbs_attr_get_uobject(attrs,
+				MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_FD);
+	if (IS_ERR(fd_uobj))
+		return PTR_ERR(fd_uobj);
+
+	ev_file = container_of(fd_uobj, struct devx_async_cmd_event_file,
+			       uobj);
+
+	if (atomic_add_return(cmd_out_len, &ev_file->ev_queue.bytes_in_use) >
+			MAX_ASYNC_BYTES_IN_USE) {
+		atomic_sub(cmd_out_len, &ev_file->ev_queue.bytes_in_use);
+		return -EAGAIN;
+	}
+
+	async_data = kvzalloc(struct_size(async_data, hdr.out_data,
+					  cmd_out_len), GFP_KERNEL);
+	if (!async_data) {
+		err = -ENOMEM;
+		goto sub_bytes;
+	}
+
+	err = uverbs_copy_from(&async_data->hdr.wr_id, attrs,
+			       MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_WR_ID);
+	if (err)
+		goto free_async;
+
+	async_data->cmd_out_len = cmd_out_len;
+	async_data->mdev = mdev;
+	async_data->fd_uobj = fd_uobj;
+
+	get_file(fd_uobj->object);
+	MLX5_SET(general_obj_in_cmd_hdr, cmd_in, uid, uid);
+	err = mlx5_cmd_exec_cb(&ev_file->async_ctx, cmd_in,
+		    uverbs_attr_get_len(attrs,
+				MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_CMD_IN),
+		    async_data->hdr.out_data,
+		    async_data->cmd_out_len,
+		    devx_query_callback, &async_data->cb_work);
+
+	if (err)
+		goto cb_err;
+
+	return 0;
+
+cb_err:
+	fput(fd_uobj->object);
+free_async:
+	kvfree(async_data);
+sub_bytes:
+	atomic_sub(cmd_out_len, &ev_file->ev_queue.bytes_in_use);
+	return err;
+}
+
 static int devx_umem_get(struct mlx5_ib_dev *dev, struct ib_ucontext *ucontext,
 			 struct uverbs_attr_bundle *attrs,
 			 struct devx_umem *obj)
@@ -1434,6 +1592,123 @@ static int devx_umem_cleanup(struct ib_uobject *uobject,
 	return 0;
 }
 
+static ssize_t devx_async_cmd_event_read(struct file *filp, char __user *buf,
+					 size_t count, loff_t *pos)
+{
+	struct devx_async_cmd_event_file *comp_ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &comp_ev_file->ev_queue;
+	struct devx_async_data *event;
+	int ret = 0;
+	size_t eventsz;
+
+	spin_lock_irq(&ev_queue->lock);
+
+	while (list_empty(&ev_queue->event_list)) {
+		spin_unlock_irq(&ev_queue->lock);
+
+		if (filp->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		if (wait_event_interruptible(
+			    ev_queue->poll_wait,
+			    (!list_empty(&ev_queue->event_list) ||
+			     ev_queue->is_destroyed))) {
+			return -ERESTARTSYS;
+		}
+
+		if (list_empty(&ev_queue->event_list) &&
+		    ev_queue->is_destroyed)
+			return -EIO;
+
+		spin_lock_irq(&ev_queue->lock);
+	}
+
+	event = list_entry(ev_queue->event_list.next,
+			   struct devx_async_data, list);
+	eventsz = event->cmd_out_len +
+			sizeof(struct mlx5_ib_uapi_devx_async_cmd_hdr);
+
+	if (eventsz > count) {
+		spin_unlock_irq(&ev_queue->lock);
+		return -ENOSPC;
+	}
+
+	list_del(ev_queue->event_list.next);
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (copy_to_user(buf, &event->hdr, eventsz))
+		ret = -EFAULT;
+	else
+		ret = eventsz;
+
+	atomic_sub(event->cmd_out_len, &ev_queue->bytes_in_use);
+	kvfree(event);
+	return ret;
+}
+
+static int devx_async_cmd_event_close(struct inode *inode, struct file *filp)
+{
+	struct ib_uobject *uobj = filp->private_data;
+	struct devx_async_cmd_event_file *comp_ev_file = container_of(
+		uobj, struct devx_async_cmd_event_file, uobj);
+	struct devx_async_data *entry, *tmp;
+
+	spin_lock_irq(&comp_ev_file->ev_queue.lock);
+	list_for_each_entry_safe(entry, tmp,
+				 &comp_ev_file->ev_queue.event_list, list)
+		kvfree(entry);
+	spin_unlock_irq(&comp_ev_file->ev_queue.lock);
+
+	uverbs_close_fd(filp);
+	return 0;
+}
+
+static __poll_t devx_async_cmd_event_poll(struct file *filp,
+					      struct poll_table_struct *wait)
+{
+	struct devx_async_cmd_event_file *comp_ev_file = filp->private_data;
+	struct devx_async_event_queue *ev_queue = &comp_ev_file->ev_queue;
+	__poll_t pollflags = 0;
+
+	poll_wait(filp, &ev_queue->poll_wait, wait);
+
+	spin_lock_irq(&ev_queue->lock);
+	if (ev_queue->is_destroyed)
+		pollflags = EPOLLIN | EPOLLRDNORM | EPOLLRDHUP;
+	else if (!list_empty(&ev_queue->event_list))
+		pollflags = EPOLLIN | EPOLLRDNORM;
+	spin_unlock_irq(&ev_queue->lock);
+
+	return pollflags;
+}
+
+const struct file_operations devx_async_cmd_event_fops = {
+	.owner	 = THIS_MODULE,
+	.read	 = devx_async_cmd_event_read,
+	.poll    = devx_async_cmd_event_poll,
+	.release = devx_async_cmd_event_close,
+	.llseek	 = no_llseek,
+};
+
+static int devx_hot_unplug_async_cmd_event_file(struct ib_uobject *uobj,
+						   enum rdma_remove_reason why)
+{
+	struct devx_async_cmd_event_file *comp_ev_file =
+		container_of(uobj, struct devx_async_cmd_event_file,
+			     uobj);
+	struct devx_async_event_queue *ev_queue = &comp_ev_file->ev_queue;
+
+	spin_lock_irq(&ev_queue->lock);
+	ev_queue->is_destroyed = 1;
+	spin_unlock_irq(&ev_queue->lock);
+
+	if (why == RDMA_REMOVE_DRIVER_REMOVE)
+		wake_up_interruptible(&ev_queue->poll_wait);
+
+	mlx5_cmd_cleanup_async_ctx(&comp_ev_file->async_ctx);
+	return 0;
+};
+
 DECLARE_UVERBS_NAMED_METHOD(
 	MLX5_IB_METHOD_DEVX_UMEM_REG,
 	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_UMEM_REG_HANDLE,
@@ -1544,6 +1819,27 @@ DECLARE_UVERBS_NAMED_METHOD(
 		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_out_cmd_hdr)),
 		UA_MANDATORY));
 
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY,
+	UVERBS_ATTR_IDR(MLX5_IB_ATTR_DEVX_OBJ_QUERY_HANDLE,
+			UVERBS_IDR_ANY_OBJECT,
+			UVERBS_ACCESS_READ,
+			UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(
+		MLX5_IB_ATTR_DEVX_OBJ_QUERY_CMD_IN,
+		UVERBS_ATTR_MIN_SIZE(MLX5_ST_SZ_BYTES(general_obj_in_cmd_hdr)),
+		UA_MANDATORY,
+		UA_ALLOC_AND_COPY),
+	UVERBS_ATTR_CONST_IN(MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_OUT_LEN,
+		u16, UA_MANDATORY),
+	UVERBS_ATTR_FD(MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_FD,
+		MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD,
+		UVERBS_ACCESS_READ,
+		UA_MANDATORY),
+	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_DEVX_OBJ_QUERY_ASYNC_WR_ID,
+		UVERBS_ATTR_TYPE(u64),
+		UA_MANDATORY));
+
 DECLARE_UVERBS_GLOBAL_METHODS(MLX5_IB_OBJECT_DEVX,
 			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OTHER),
 			      &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_QUERY_UAR),
@@ -1554,13 +1850,30 @@ DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_OBJ,
 			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_CREATE),
 			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_DESTROY),
 			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_MODIFY),
-			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_QUERY));
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_QUERY),
+			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_OBJ_ASYNC_QUERY));
 
 DECLARE_UVERBS_NAMED_OBJECT(MLX5_IB_OBJECT_DEVX_UMEM,
 			    UVERBS_TYPE_ALLOC_IDR(devx_umem_cleanup),
 			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_REG),
 			    &UVERBS_METHOD(MLX5_IB_METHOD_DEVX_UMEM_DEREG));
 
+
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC,
+	UVERBS_ATTR_FD(MLX5_IB_ATTR_DEVX_ASYNC_CMD_FD_ALLOC_HANDLE,
+			MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD,
+			UVERBS_ACCESS_NEW,
+			UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_OBJECT(
+	MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD,
+	UVERBS_TYPE_ALLOC_FD(sizeof(struct devx_async_cmd_event_file),
+			     devx_hot_unplug_async_cmd_event_file,
+			     &devx_async_cmd_event_fops, "[devx_async_cmd]",
+			     O_RDONLY),
+	&UVERBS_METHOD(MLX5_IB_METHOD_DEVX_ASYNC_CMD_FD_ALLOC));
+
 static bool devx_is_supported(struct ib_device *device)
 {
 	struct mlx5_ib_dev *dev = to_mdev(device);
@@ -1578,5 +1891,8 @@ const struct uapi_definition mlx5_ib_devx_defs[] = {
 	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(
 		MLX5_IB_OBJECT_DEVX_UMEM,
 		UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)),
+	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(
+		MLX5_IB_OBJECT_DEVX_ASYNC_CMD_FD,
+		UAPI_DEF_IS_OBJ_SUPPORTED(devx_is_supported)),
 	{},
 };
diff --git a/drivers/infiniband/hw/mlx5/mad.c b/drivers/infiniband/hw/mlx5/mad.c
index 558638468edb..6c529e6f3a01 100644
--- a/drivers/infiniband/hw/mlx5/mad.c
+++ b/drivers/infiniband/hw/mlx5/mad.c
@@ -36,6 +36,7 @@
 #include <rdma/ib_smi.h>
 #include <rdma/ib_pma.h>
 #include "mlx5_ib.h"
+#include "cmd.h"
 
 enum {
 	MLX5_IB_VENDOR_CLASS1 = 0x9,
@@ -51,9 +52,10 @@ static bool can_do_mad_ifc(struct mlx5_ib_dev *dev, u8 port_num,
 	return dev->mdev->port_caps[port_num - 1].has_smi;
 }
 
-int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey,
-		 u8 port, const struct ib_wc *in_wc, const struct ib_grh *in_grh,
-		 const void *in_mad, void *response_mad)
+static int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey,
+			int ignore_bkey, u8 port, const struct ib_wc *in_wc,
+			const struct ib_grh *in_grh, const void *in_mad,
+			void *response_mad)
 {
 	u8 op_modifier = 0;
 
@@ -68,7 +70,8 @@ int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey,
 	if (ignore_bkey || !in_wc)
 		op_modifier |= 0x2;
 
-	return mlx5_core_mad_ifc(dev->mdev, in_mad, response_mad, op_modifier, port);
+	return mlx5_cmd_mad_ifc(dev->mdev, in_mad, response_mad, op_modifier,
+				port);
 }
 
 static int process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index ea124af56dce..7fcc3f095371 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -588,6 +588,7 @@ struct mlx5_ib_mr {
 	struct mlx5_ib_mr      *parent;
 	atomic_t		num_leaf_free;
 	wait_queue_head_t       q_leaf_free;
+	struct mlx5_async_work  cb_work;
 };
 
 static inline bool is_odp_mr(struct mlx5_ib_mr *mr)
@@ -949,6 +950,7 @@ struct mlx5_ib_dev {
 	struct mlx5_memic	memic;
 	u16			devx_whitelist_uid;
 	struct mlx5_srq_table   srq_table;
+	struct mlx5_async_ctx   async_ctx;
 };
 
 static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
@@ -1044,9 +1046,6 @@ void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db)
 void __mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq);
 void mlx5_ib_cq_clean(struct mlx5_ib_cq *cq, u32 qpn, struct mlx5_ib_srq *srq);
 void mlx5_ib_free_srq_wqe(struct mlx5_ib_srq *srq, int wqe_index);
-int mlx5_MAD_IFC(struct mlx5_ib_dev *dev, int ignore_mkey, int ignore_bkey,
-		 u8 port, const struct ib_wc *in_wc, const struct ib_grh *in_grh,
-		 const void *in_mad, void *response_mad);
 struct ib_ah *mlx5_ib_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr,
 				u32 flags, struct ib_udata *udata);
 int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index d78a94803e8c..705a79cd21da 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -120,9 +120,10 @@ static void update_odp_mr(struct mlx5_ib_mr *mr)
 	}
 }
 
-static void reg_mr_callback(int status, void *context)
+static void reg_mr_callback(int status, struct mlx5_async_work *context)
 {
-	struct mlx5_ib_mr *mr = context;
+	struct mlx5_ib_mr *mr =
+		container_of(context, struct mlx5_ib_mr, cb_work);
 	struct mlx5_ib_dev *dev = mr->dev;
 	struct mlx5_mr_cache *cache = &dev->cache;
 	int c = order2idx(dev, mr->order);
@@ -213,9 +214,9 @@ static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
 		ent->pending++;
 		spin_unlock_irq(&ent->lock);
 		err = mlx5_core_create_mkey_cb(dev->mdev, &mr->mmkey,
-					       in, inlen,
+					       &dev->async_ctx, in, inlen,
 					       mr->out, sizeof(mr->out),
-					       reg_mr_callback, mr);
+					       reg_mr_callback, &mr->cb_work);
 		if (err) {
 			spin_lock_irq(&ent->lock);
 			ent->pending--;
@@ -649,6 +650,7 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 		return -ENOMEM;
 	}
 
+	mlx5_cmd_init_async_ctx(dev->mdev, &dev->async_ctx);
 	timer_setup(&dev->delay_timer, delay_time_func, 0);
 	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
 		ent = &cache->ent[i];
@@ -688,33 +690,6 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 	return 0;
 }
 
-static void wait_for_async_commands(struct mlx5_ib_dev *dev)
-{
-	struct mlx5_mr_cache *cache = &dev->cache;
-	struct mlx5_cache_ent *ent;
-	int total = 0;
-	int i;
-	int j;
-
-	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
-		ent = &cache->ent[i];
-		for (j = 0 ; j < 1000; j++) {
-			if (!ent->pending)
-				break;
-			msleep(50);
-		}
-	}
-	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
-		ent = &cache->ent[i];
-		total += ent->pending;
-	}
-
-	if (total)
-		mlx5_ib_warn(dev, "aborted while there are %d pending mr requests\n", total);
-	else
-		mlx5_ib_warn(dev, "done with all pending requests\n");
-}
-
 int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
 {
 	int i;
@@ -726,12 +701,12 @@ int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
 	flush_workqueue(dev->cache.wq);
 
 	mlx5_mr_cache_debugfs_cleanup(dev);
+	mlx5_cmd_cleanup_async_ctx(&dev->async_ctx);
 
 	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++)
 		clean_keys(dev, i);
 
 	destroy_workqueue(dev->cache.wq);
-	wait_for_async_commands(dev);
 	del_timer_sync(&dev->delay_timer);
 
 	return 0;
diff --git a/drivers/infiniband/hw/mthca/mthca_memfree.c b/drivers/infiniband/hw/mthca/mthca_memfree.c
index cc9c0c8ccba3..112d2f38e0de 100644
--- a/drivers/infiniband/hw/mthca/mthca_memfree.c
+++ b/drivers/infiniband/hw/mthca/mthca_memfree.c
@@ -623,8 +623,9 @@ int mthca_alloc_db(struct mthca_dev *dev, enum mthca_db_type type,
 	page = dev->db_tab->page + end;
 
 alloc:
-	page->db_rec = dma_zalloc_coherent(&dev->pdev->dev, MTHCA_ICM_PAGE_SIZE,
-					   &page->mapping, GFP_KERNEL);
+	page->db_rec = dma_alloc_coherent(&dev->pdev->dev,
+					  MTHCA_ICM_PAGE_SIZE, &page->mapping,
+					  GFP_KERNEL);
 	if (!page->db_rec) {
 		ret = -ENOMEM;
 		goto out;
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
index 241a57a07485..097e5ab2a19f 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
@@ -380,8 +380,8 @@ static int ocrdma_alloc_q(struct ocrdma_dev *dev,
 	q->len = len;
 	q->entry_size = entry_size;
 	q->size = len * entry_size;
-	q->va = dma_zalloc_coherent(&dev->nic_info.pdev->dev, q->size,
-				    &q->dma, GFP_KERNEL);
+	q->va = dma_alloc_coherent(&dev->nic_info.pdev->dev, q->size, &q->dma,
+				   GFP_KERNEL);
 	if (!q->va)
 		return -ENOMEM;
 	return 0;
@@ -1819,7 +1819,7 @@ int ocrdma_mbx_create_cq(struct ocrdma_dev *dev, struct ocrdma_cq *cq,
 		return -ENOMEM;
 	ocrdma_init_mch(&cmd->cmd.req, OCRDMA_CMD_CREATE_CQ,
 			OCRDMA_SUBSYS_COMMON, sizeof(*cmd));
-	cq->va = dma_zalloc_coherent(&pdev->dev, cq->len, &cq->pa, GFP_KERNEL);
+	cq->va = dma_alloc_coherent(&pdev->dev, cq->len, &cq->pa, GFP_KERNEL);
 	if (!cq->va) {
 		status = -ENOMEM;
 		goto mem_err;
@@ -2209,7 +2209,7 @@ static int ocrdma_set_create_qp_sq_cmd(struct ocrdma_create_qp_req *cmd,
 	qp->sq.max_cnt = max_wqe_allocated;
 	len = (hw_pages * hw_page_size);
 
-	qp->sq.va = dma_zalloc_coherent(&pdev->dev, len, &pa, GFP_KERNEL);
+	qp->sq.va = dma_alloc_coherent(&pdev->dev, len, &pa, GFP_KERNEL);
 	if (!qp->sq.va)
 		return -EINVAL;
 	qp->sq.len = len;
@@ -2259,7 +2259,7 @@ static int ocrdma_set_create_qp_rq_cmd(struct ocrdma_create_qp_req *cmd,
 	qp->rq.max_cnt = max_rqe_allocated;
 	len = (hw_pages * hw_page_size);
 
-	qp->rq.va = dma_zalloc_coherent(&pdev->dev, len, &pa, GFP_KERNEL);
+	qp->rq.va = dma_alloc_coherent(&pdev->dev, len, &pa, GFP_KERNEL);
 	if (!qp->rq.va)
 		return -ENOMEM;
 	qp->rq.pa = pa;
@@ -2315,8 +2315,8 @@ static int ocrdma_set_create_qp_ird_cmd(struct ocrdma_create_qp_req *cmd,
 	if (dev->attr.ird == 0)
 		return 0;
 
-	qp->ird_q_va = dma_zalloc_coherent(&pdev->dev, ird_q_len, &pa,
-					   GFP_KERNEL);
+	qp->ird_q_va = dma_alloc_coherent(&pdev->dev, ird_q_len, &pa,
+					  GFP_KERNEL);
 	if (!qp->ird_q_va)
 		return -ENOMEM;
 	ocrdma_build_q_pages(&cmd->ird_addr[0], dev->attr.num_ird_pages,
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c
index d8ab396593b7..a902942adb5d 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c
@@ -73,8 +73,8 @@ bool ocrdma_alloc_stats_resources(struct ocrdma_dev *dev)
 	mem->size = max_t(u32, sizeof(struct ocrdma_rdma_stats_req),
 			sizeof(struct ocrdma_rdma_stats_resp));
 
-	mem->va = dma_zalloc_coherent(&dev->nic_info.pdev->dev, mem->size,
-				      &mem->pa, GFP_KERNEL);
+	mem->va = dma_alloc_coherent(&dev->nic_info.pdev->dev, mem->size,
+				     &mem->pa, GFP_KERNEL);
 	if (!mem->va) {
 		pr_err("%s: stats mbox allocation failed\n", __func__);
 		return false;
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
index eb69e71c2053..2a62936bef4d 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c
@@ -504,8 +504,8 @@ struct ib_ucontext *ocrdma_alloc_ucontext(struct ib_device *ibdev,
 	INIT_LIST_HEAD(&ctx->mm_head);
 	mutex_init(&ctx->mm_list_lock);
 
-	ctx->ah_tbl.va = dma_zalloc_coherent(&pdev->dev, map_len,
-					     &ctx->ah_tbl.pa, GFP_KERNEL);
+	ctx->ah_tbl.va = dma_alloc_coherent(&pdev->dev, map_len,
+					    &ctx->ah_tbl.pa, GFP_KERNEL);
 	if (!ctx->ah_tbl.va) {
 		kfree(ctx);
 		return ERR_PTR(-ENOMEM);
@@ -838,7 +838,7 @@ static int ocrdma_build_pbl_tbl(struct ocrdma_dev *dev, struct ocrdma_hw_mr *mr)
 		return -ENOMEM;
 
 	for (i = 0; i < mr->num_pbls; i++) {
-		va = dma_zalloc_coherent(&pdev->dev, dma_len, &pa, GFP_KERNEL);
+		va = dma_alloc_coherent(&pdev->dev, dma_len, &pa, GFP_KERNEL);
 		if (!va) {
 			ocrdma_free_mr_pbl_tbl(dev, mr);
 			status = -ENOMEM;
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index 265a2646ad12..989f08633fbe 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -556,8 +556,8 @@ static struct qedr_pbl *qedr_alloc_pbl_tbl(struct qedr_dev *dev,
 		return ERR_PTR(-ENOMEM);
 
 	for (i = 0; i < pbl_info->num_pbls; i++) {
-		va = dma_zalloc_coherent(&pdev->dev, pbl_info->pbl_size,
-					 &pa, flags);
+		va = dma_alloc_coherent(&pdev->dev, pbl_info->pbl_size, &pa,
+					flags);
 		if (!va)
 			goto err;
 
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
index 1bc415483d9b..b5864750fb6e 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
@@ -890,8 +890,8 @@ static int pvrdma_pci_probe(struct pci_dev *pdev,
 	dev_info(&pdev->dev, "device version %d, driver version %d\n",
 		 dev->dsr_version, PVRDMA_VERSION);
 
-	dev->dsr = dma_zalloc_coherent(&pdev->dev, sizeof(*dev->dsr),
-				       &dev->dsrbase, GFP_KERNEL);
+	dev->dsr = dma_alloc_coherent(&pdev->dev, sizeof(*dev->dsr),
+				      &dev->dsrbase, GFP_KERNEL);
 	if (!dev->dsr) {
 		dev_err(&pdev->dev, "failed to allocate shared region\n");
 		ret = -ENOMEM;