[PATCH rdma-rc v1 0/4] Remove mlx5 support for IB_DEVICE_LOCAL_DMA_LKEY

[PATCH rdma-rc v1 0/4] Remove mlx5 support for IB_DEVICE_LOCAL_DMA_LKEY

[Sagi Grimberg]

The Connect-IB device has a specific issue with memory registration using
the reserved lkey (device global_dma_lkey). This caused user-space memory
registration which usually uses cached pre-registered memory keys to fail
due to a device access error during registration. kernel-space memory
registrations used an internal instance a physical memory key allocated with
the private pd context, so this error didn't happen there.

Since the reserved lkey is not fully functional, remove the support for
it altogether. Jason's patches commit 96249d70dd70 ("IB/core: Guarantee
that a local_dma_lkey is available") address consumers by allocating a physical
lkey per PD in the core layer. It also allows mlx5 driver to get rid of it's
private physical lkey (patch #2).

The ConnectX-4 device should have this issue fixed so the capability will be
restored depending on a FW query information.

Also, fix NFS client to use the PD local_dma_lkey instead of the device
local_dma_lkey (which requires a missing check of the device capability in
frwr mode). And, fix iser initiator which encountered some issues when
registering a signature capable memory region with an indirect dma_lkey
area. The fix covers a larger ground as it does not allow using a global MR
with remote access (long standing issue) but on the way makes the specific
registration issue go away.

Thanks to Haggai for catching this early enough.

Changes from v0:
- Replace xprtrdma patch to Chuck's one
- Fixed typo in iser modparam description

Chuck Lever (1):
  xprtrdma: Replace global lkey with lkey local to PD

Sagi Grimberg (3):
  IB/iser: Add module parameter for always register memory
  IB/mlx5: Remove support for IB_DEVICE_LOCAL_DMA_LKEY
  IB/mlx5: Remove pa_lkey usages

 drivers/infiniband/hw/mlx5/main.c            | 67 +---------------------------
 drivers/infiniband/hw/mlx5/mlx5_ib.h         |  2 -
 drivers/infiniband/hw/mlx5/qp.c              |  4 +-
 drivers/infiniband/ulp/iser/iscsi_iser.c     |  5 +++
 drivers/infiniband/ulp/iser/iscsi_iser.h     |  1 +
 drivers/infiniband/ulp/iser/iser_memory.c    | 18 +++++---
 drivers/infiniband/ulp/iser/iser_verbs.c     | 21 +++++----
 drivers/net/ethernet/mellanox/mlx5/core/fw.c | 22 ---------
 include/linux/mlx5/device.h                  | 11 -----
 include/linux/mlx5/driver.h                  |  1 -
 net/sunrpc/xprtrdma/fmr_ops.c                | 19 --------
 net/sunrpc/xprtrdma/frwr_ops.c               |  5 ---
 net/sunrpc/xprtrdma/physical_ops.c           | 10 +----
 net/sunrpc/xprtrdma/verbs.c                  |  2 +-
 net/sunrpc/xprtrdma/xprt_rdma.h              |  1 -
 15 files changed, 35 insertions(+), 154 deletions(-)

-- 
1.8.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-rc v1 1/4] xprtrdma: Replace global lkey with lkey local to PD

[Sagi Grimberg]

From: Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>

The core API has changed so that devices that do not have a global
DMA lkey automatically create an mr, per-PD, and make that lkey
available. The global DMA lkey interface is going away in favor of
the per-PD DMA lkey.

The per-PD DMA lkey is always available. Convert xprtrdma to use the
device's per-PD DMA lkey for regbufs, no matter which memory
registration scheme is in use.

Signed-off-by: Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
Signed-off-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 net/sunrpc/xprtrdma/fmr_ops.c      | 19 -------------------
 net/sunrpc/xprtrdma/frwr_ops.c     |  5 -----
 net/sunrpc/xprtrdma/physical_ops.c | 10 +---------
 net/sunrpc/xprtrdma/verbs.c        |  2 +-
 net/sunrpc/xprtrdma/xprt_rdma.h    |  1 -
 5 files changed, 2 insertions(+), 35 deletions(-)

diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c
index cb25c89da623..f1e8dafbd507 100644
--- a/net/sunrpc/xprtrdma/fmr_ops.c
+++ b/net/sunrpc/xprtrdma/fmr_ops.c
@@ -39,25 +39,6 @@ static int
 fmr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 	    struct rpcrdma_create_data_internal *cdata)
 {
-	struct ib_device_attr *devattr = &ia->ri_devattr;
-	struct ib_mr *mr;
-
-	/* Obtain an lkey to use for the regbufs, which are
-	 * protected from remote access.
-	 */
-	if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) {
-		ia->ri_dma_lkey = ia->ri_device->local_dma_lkey;
-	} else {
-		mr = ib_get_dma_mr(ia->ri_pd, IB_ACCESS_LOCAL_WRITE);
-		if (IS_ERR(mr)) {
-			pr_err("%s: ib_get_dma_mr for failed with %lX\n",
-			       __func__, PTR_ERR(mr));
-			return -ENOMEM;
-		}
-		ia->ri_dma_lkey = ia->ri_dma_mr->lkey;
-		ia->ri_dma_mr = mr;
-	}
-
 	return 0;
 }
 
diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
index d6653f5d0830..5318951b3b53 100644
--- a/net/sunrpc/xprtrdma/frwr_ops.c
+++ b/net/sunrpc/xprtrdma/frwr_ops.c
@@ -189,11 +189,6 @@ frwr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 	struct ib_device_attr *devattr = &ia->ri_devattr;
 	int depth, delta;
 
-	/* Obtain an lkey to use for the regbufs, which are
-	 * protected from remote access.
-	 */
-	ia->ri_dma_lkey = ia->ri_device->local_dma_lkey;
-
 	ia->ri_max_frmr_depth =
 			min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS,
 			      devattr->max_fast_reg_page_list_len);
diff --git a/net/sunrpc/xprtrdma/physical_ops.c b/net/sunrpc/xprtrdma/physical_ops.c
index 72cf8b15bbb4..617b76f22154 100644
--- a/net/sunrpc/xprtrdma/physical_ops.c
+++ b/net/sunrpc/xprtrdma/physical_ops.c
@@ -23,7 +23,6 @@ static int
 physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 		 struct rpcrdma_create_data_internal *cdata)
 {
-	struct ib_device_attr *devattr = &ia->ri_devattr;
 	struct ib_mr *mr;
 
 	/* Obtain an rkey to use for RPC data payloads.
@@ -37,15 +36,8 @@ physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 		       __func__, PTR_ERR(mr));
 		return -ENOMEM;
 	}
-	ia->ri_dma_mr = mr;
-
-	/* Obtain an lkey to use for regbufs.
-	 */
-	if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)
-		ia->ri_dma_lkey = ia->ri_device->local_dma_lkey;
-	else
-		ia->ri_dma_lkey = ia->ri_dma_mr->lkey;
 
+	ia->ri_dma_mr = mr;
 	return 0;
 }
 
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 682996779970..eb081ad05e33 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -1252,7 +1252,7 @@ rpcrdma_alloc_regbuf(struct rpcrdma_ia *ia, size_t size, gfp_t flags)
 		goto out_free;
 
 	iov->length = size;
-	iov->lkey = ia->ri_dma_lkey;
+	iov->lkey = ia->ri_pd->local_dma_lkey;
 	rb->rg_size = size;
 	rb->rg_owner = NULL;
 	return rb;
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index d252457ff21a..df5ad4e15702 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -66,7 +66,6 @@ struct rpcrdma_ia {
 	struct rdma_cm_id 	*ri_id;
 	struct ib_pd		*ri_pd;
 	struct ib_mr		*ri_dma_mr;
-	u32			ri_dma_lkey;
 	struct completion	ri_done;
 	int			ri_async_rc;
 	unsigned int		ri_max_frmr_depth;
-- 
1.8.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-rc v1 2/4] IB/iser: Add module parameter for always register memory

[Sagi Grimberg]

This module parameter forces memory registration even for
a continuous memory region. It is true by default as sending
an all-physical rkey with remote permissions might be insecure.

Signed-off-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/ulp/iser/iscsi_iser.c  |  5 +++++
 drivers/infiniband/ulp/iser/iscsi_iser.h  |  1 +
 drivers/infiniband/ulp/iser/iser_memory.c | 18 ++++++++++++------
 drivers/infiniband/ulp/iser/iser_verbs.c  | 21 +++++++++++++--------
 4 files changed, 31 insertions(+), 14 deletions(-)

diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c
index 1ace5d83a4d7..f58ff96b6cbb 100644
--- a/drivers/infiniband/ulp/iser/iscsi_iser.c
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.c
@@ -97,6 +97,11 @@ unsigned int iser_max_sectors = ISER_DEF_MAX_SECTORS;
 module_param_named(max_sectors, iser_max_sectors, uint, S_IRUGO | S_IWUSR);
 MODULE_PARM_DESC(max_sectors, "Max number of sectors in a single scsi command (default:1024");
 
+bool iser_always_reg = true;
+module_param_named(always_register, iser_always_reg, bool, S_IRUGO);
+MODULE_PARM_DESC(always_register,
+		 "Always register memory, even for continuous memory regions (default:true)");
+
 bool iser_pi_enable = false;
 module_param_named(pi_enable, iser_pi_enable, bool, S_IRUGO);
 MODULE_PARM_DESC(pi_enable, "Enable T10-PI offload support (default:disabled)");
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h
index 86f6583485ef..a5edd6ede692 100644
--- a/drivers/infiniband/ulp/iser/iscsi_iser.h
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.h
@@ -611,6 +611,7 @@ extern int iser_debug_level;
 extern bool iser_pi_enable;
 extern int iser_pi_guard;
 extern unsigned int iser_max_sectors;
+extern bool iser_always_reg;
 
 int iser_assign_reg_ops(struct iser_device *device);
 
diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c
index 2493cc748db8..4c46d67d37a1 100644
--- a/drivers/infiniband/ulp/iser/iser_memory.c
+++ b/drivers/infiniband/ulp/iser/iser_memory.c
@@ -803,11 +803,12 @@ static int
 iser_reg_prot_sg(struct iscsi_iser_task *task,
 		 struct iser_data_buf *mem,
 		 struct iser_fr_desc *desc,
+		 bool use_dma_key,
 		 struct iser_mem_reg *reg)
 {
 	struct iser_device *device = task->iser_conn->ib_conn.device;
 
-	if (mem->dma_nents == 1)
+	if (use_dma_key)
 		return iser_reg_dma(device, mem, reg);
 
 	return device->reg_ops->reg_mem(task, mem, &desc->pi_ctx->rsc, reg);
@@ -817,11 +818,12 @@ static int
 iser_reg_data_sg(struct iscsi_iser_task *task,
 		 struct iser_data_buf *mem,
 		 struct iser_fr_desc *desc,
+		 bool use_dma_key,
 		 struct iser_mem_reg *reg)
 {
 	struct iser_device *device = task->iser_conn->ib_conn.device;
 
-	if (mem->dma_nents == 1)
+	if (use_dma_key)
 		return iser_reg_dma(device, mem, reg);
 
 	return device->reg_ops->reg_mem(task, mem, &desc->rsc, reg);
@@ -836,14 +838,17 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task,
 	struct iser_mem_reg *reg = &task->rdma_reg[dir];
 	struct iser_mem_reg *data_reg;
 	struct iser_fr_desc *desc = NULL;
+	bool use_dma_key;
 	int err;
 
 	err = iser_handle_unaligned_buf(task, mem, dir);
 	if (unlikely(err))
 		return err;
 
-	if (mem->dma_nents != 1 ||
-	    scsi_get_prot_op(task->sc) != SCSI_PROT_NORMAL) {
+	use_dma_key = (mem->dma_nents == 1 && !iser_always_reg &&
+		       scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL);
+
+	if (!use_dma_key) {
 		desc = device->reg_ops->reg_desc_get(ib_conn);
 		reg->mem_h = desc;
 	}
@@ -853,7 +858,7 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task,
 	else
 		data_reg = &task->desc.data_reg;
 
-	err = iser_reg_data_sg(task, mem, desc, data_reg);
+	err = iser_reg_data_sg(task, mem, desc, use_dma_key, data_reg);
 	if (unlikely(err))
 		goto err_reg;
 
@@ -866,7 +871,8 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task,
 			if (unlikely(err))
 				goto err_reg;
 
-			err = iser_reg_prot_sg(task, mem, desc, prot_reg);
+			err = iser_reg_prot_sg(task, mem, desc,
+					       use_dma_key, prot_reg);
 			if (unlikely(err))
 				goto err_reg;
 		}
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c
index ae70cc1463ac..85132d867bc8 100644
--- a/drivers/infiniband/ulp/iser/iser_verbs.c
+++ b/drivers/infiniband/ulp/iser/iser_verbs.c
@@ -133,11 +133,15 @@ static int iser_create_device_ib_res(struct iser_device *device)
 			     (unsigned long)comp);
 	}
 
-	device->mr = ib_get_dma_mr(device->pd, IB_ACCESS_LOCAL_WRITE |
-				   IB_ACCESS_REMOTE_WRITE |
-				   IB_ACCESS_REMOTE_READ);
-	if (IS_ERR(device->mr))
-		goto dma_mr_err;
+	if (!iser_always_reg) {
+		int access = IB_ACCESS_LOCAL_WRITE |
+			     IB_ACCESS_REMOTE_WRITE |
+			     IB_ACCESS_REMOTE_READ;
+
+		device->mr = ib_get_dma_mr(device->pd, access);
+		if (IS_ERR(device->mr))
+			goto dma_mr_err;
+	}
 
 	INIT_IB_EVENT_HANDLER(&device->event_handler, device->ib_device,
 				iser_event_handler);
@@ -147,7 +151,8 @@ static int iser_create_device_ib_res(struct iser_device *device)
 	return 0;
 
 handler_err:
-	ib_dereg_mr(device->mr);
+	if (device->mr)
+		ib_dereg_mr(device->mr);
 dma_mr_err:
 	for (i = 0; i < device->comps_used; i++)
 		tasklet_kill(&device->comps[i].tasklet);
@@ -173,7 +178,6 @@ comps_err:
 static void iser_free_device_ib_res(struct iser_device *device)
 {
 	int i;
-	BUG_ON(device->mr == NULL);
 
 	for (i = 0; i < device->comps_used; i++) {
 		struct iser_comp *comp = &device->comps[i];
@@ -184,7 +188,8 @@ static void iser_free_device_ib_res(struct iser_device *device)
 	}
 
 	(void)ib_unregister_event_handler(&device->event_handler);
-	(void)ib_dereg_mr(device->mr);
+	if (device->mr)
+		(void)ib_dereg_mr(device->mr);
 	ib_dealloc_pd(device->pd);
 
 	kfree(device->comps);
-- 
1.8.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-rc v1 3/4] IB/mlx5: Remove support for IB_DEVICE_LOCAL_DMA_LKEY

[Sagi Grimberg]

ConnectIB has some known issues with memory registration
using the local_dma_lkey (SEND, RDMA, RECV seems to work ok).
Thus don't expose support for it (and remove device->local_dma_lkey
setting).

since commit 96249d70dd70 ("IB/core: Guarantee that a local_dma_lkey
is available") addressed that by allocating a DMA MR with local permissions
and converted the consumers to use the MR associated with the PD rather
then device->local_dma_lkey.

The local_dma_lkey support will be restored in CX4 depending on FW
capability query.

Signed-off-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx5/main.c            | 10 +---------
 drivers/net/ethernet/mellanox/mlx5/core/fw.c | 22 ----------------------
 include/linux/mlx5/device.h                  | 11 -----------
 include/linux/mlx5/driver.h                  |  1 -
 4 files changed, 1 insertion(+), 43 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 41d6911e244e..0ab9625911a1 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -245,7 +245,6 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		props->device_cap_flags |= IB_DEVICE_BAD_QKEY_CNTR;
 	if (MLX5_CAP_GEN(mdev, apm))
 		props->device_cap_flags |= IB_DEVICE_AUTO_PATH_MIG;
-	props->device_cap_flags |= IB_DEVICE_LOCAL_DMA_LKEY;
 	if (MLX5_CAP_GEN(mdev, xrc))
 		props->device_cap_flags |= IB_DEVICE_XRC;
 	props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
@@ -1245,18 +1244,10 @@ static int create_dev_resources(struct mlx5_ib_resources *devr)
 	struct ib_srq_init_attr attr;
 	struct mlx5_ib_dev *dev;
 	struct ib_cq_init_attr cq_attr = {.cqe = 1};
-	u32 rsvd_lkey;
 	int ret = 0;
 
 	dev = container_of(devr, struct mlx5_ib_dev, devr);
 
-	ret = mlx5_core_query_special_context(dev->mdev, &rsvd_lkey);
-	if (ret) {
-		pr_err("Failed to query special context %d\n", ret);
-		return ret;
-	}
-	dev->ib_dev.local_dma_lkey = rsvd_lkey;
-
 	devr->p0 = mlx5_ib_alloc_pd(&dev->ib_dev, NULL, NULL);
 	if (IS_ERR(devr->p0)) {
 		ret = PTR_ERR(devr->p0);
@@ -1418,6 +1409,7 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 	strlcpy(dev->ib_dev.name, "mlx5_%d", IB_DEVICE_NAME_MAX);
 	dev->ib_dev.owner		= THIS_MODULE;
 	dev->ib_dev.node_type		= RDMA_NODE_IB_CA;
+	dev->ib_dev.local_dma_lkey	= 0 /* not supported for now */;
 	dev->num_ports		= MLX5_CAP_GEN(mdev, num_ports);
 	dev->ib_dev.phys_port_cnt     = dev->num_ports;
 	dev->ib_dev.num_comp_vectors    =
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw.c b/drivers/net/ethernet/mellanox/mlx5/core/fw.c
index aa0d5ffe92d8..9335e5ae18cc 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw.c
@@ -200,25 +200,3 @@ int mlx5_cmd_teardown_hca(struct mlx5_core_dev *dev)
 
 	return err;
 }
-
-int mlx5_core_query_special_context(struct mlx5_core_dev *dev, u32 *rsvd_lkey)
-{
-	struct mlx5_cmd_query_special_contexts_mbox_in in;
-	struct mlx5_cmd_query_special_contexts_mbox_out out;
-	int err;
-
-	memset(&in, 0, sizeof(in));
-	memset(&out, 0, sizeof(out));
-	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS);
-	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
-	if (err)
-		return err;
-
-	if (out.hdr.status)
-		err = mlx5_cmd_status_to_err(&out.hdr);
-
-	*rsvd_lkey = be32_to_cpu(out.resd_lkey);
-
-	return err;
-}
-EXPORT_SYMBOL(mlx5_core_query_special_context);
diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h
index 6e4169c5ad78..b943cd9e2097 100644
--- a/include/linux/mlx5/device.h
+++ b/include/linux/mlx5/device.h
@@ -402,17 +402,6 @@ struct mlx5_cmd_teardown_hca_mbox_out {
 	u8			rsvd[8];
 };
 
-struct mlx5_cmd_query_special_contexts_mbox_in {
-	struct mlx5_inbox_hdr	hdr;
-	u8			rsvd[8];
-};
-
-struct mlx5_cmd_query_special_contexts_mbox_out {
-	struct mlx5_outbox_hdr	hdr;
-	__be32                  dump_fill_mkey;
-	__be32                  resd_lkey;
-};
-
 struct mlx5_cmd_layout {
 	u8		type;
 	u8		rsvd0[3];
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 1e2e48ccb3fd..5722d88c2429 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -828,7 +828,6 @@ void *mlx5_get_protocol_dev(struct mlx5_core_dev *mdev, int protocol);
 int mlx5_register_interface(struct mlx5_interface *intf);
 void mlx5_unregister_interface(struct mlx5_interface *intf);
 int mlx5_core_query_vendor_id(struct mlx5_core_dev *mdev, u32 *vendor_id);
-int mlx5_core_query_special_context(struct mlx5_core_dev *dev, u32 *rsvd_lkey);
 
 struct mlx5_profile {
 	u64	mask;
-- 
1.8.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH rdma-rc v1 4/4] IB/mlx5: Remove pa_lkey usages

[Sagi Grimberg]

Since mlx5 driver cannot rely on registration using the
reserved lkey (global_dma_lkey) it used to allocate a private
physical address lkey for each allocated pd.
Commit 96249d70dd70 ("IB/core: Guarantee that a local_dma_lkey
is available") just does it in the core layer so we can go ahead
and use that.

Signed-off-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx5/main.c    | 57 ------------------------------------
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  2 --
 drivers/infiniband/hw/mlx5/qp.c      |  4 +--
 3 files changed, 1 insertion(+), 62 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 0ab9625911a1..f1ccd40beae9 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -794,53 +794,6 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
 	return 0;
 }
 
-static int alloc_pa_mkey(struct mlx5_ib_dev *dev, u32 *key, u32 pdn)
-{
-	struct mlx5_create_mkey_mbox_in *in;
-	struct mlx5_mkey_seg *seg;
-	struct mlx5_core_mr mr;
-	int err;
-
-	in = kzalloc(sizeof(*in), GFP_KERNEL);
-	if (!in)
-		return -ENOMEM;
-
-	seg = &in->seg;
-	seg->flags = MLX5_PERM_LOCAL_READ | MLX5_ACCESS_MODE_PA;
-	seg->flags_pd = cpu_to_be32(pdn | MLX5_MKEY_LEN64);
-	seg->qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
-	seg->start_addr = 0;
-
-	err = mlx5_core_create_mkey(dev->mdev, &mr, in, sizeof(*in),
-				    NULL, NULL, NULL);
-	if (err) {
-		mlx5_ib_warn(dev, "failed to create mkey, %d\n", err);
-		goto err_in;
-	}
-
-	kfree(in);
-	*key = mr.key;
-
-	return 0;
-
-err_in:
-	kfree(in);
-
-	return err;
-}
-
-static void free_pa_mkey(struct mlx5_ib_dev *dev, u32 key)
-{
-	struct mlx5_core_mr mr;
-	int err;
-
-	memset(&mr, 0, sizeof(mr));
-	mr.key = key;
-	err = mlx5_core_destroy_mkey(dev->mdev, &mr);
-	if (err)
-		mlx5_ib_warn(dev, "failed to destroy mkey 0x%x\n", key);
-}
-
 static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev,
 				      struct ib_ucontext *context,
 				      struct ib_udata *udata)
@@ -866,13 +819,6 @@ static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev,
 			kfree(pd);
 			return ERR_PTR(-EFAULT);
 		}
-	} else {
-		err = alloc_pa_mkey(to_mdev(ibdev), &pd->pa_lkey, pd->pdn);
-		if (err) {
-			mlx5_core_dealloc_pd(to_mdev(ibdev)->mdev, pd->pdn);
-			kfree(pd);
-			return ERR_PTR(err);
-		}
 	}
 
 	return &pd->ibpd;
@@ -883,9 +829,6 @@ static int mlx5_ib_dealloc_pd(struct ib_pd *pd)
 	struct mlx5_ib_dev *mdev = to_mdev(pd->device);
 	struct mlx5_ib_pd *mpd = to_mpd(pd);
 
-	if (!pd->uobject)
-		free_pa_mkey(mdev, mpd->pa_lkey);
-
 	mlx5_core_dealloc_pd(mdev->mdev, mpd->pdn);
 	kfree(mpd);
 
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index a5fa0b9c7580..b0a22fea76f1 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -103,7 +103,6 @@ static inline struct mlx5_ib_ucontext *to_mucontext(struct ib_ucontext *ibuconte
 struct mlx5_ib_pd {
 	struct ib_pd		ibpd;
 	u32			pdn;
-	u32			pa_lkey;
 };
 
 /* Use macros here so that don't have to duplicate
@@ -213,7 +212,6 @@ struct mlx5_ib_qp {
 	int			uuarn;
 
 	int			create_type;
-	u32			pa_lkey;
 
 	/* Store signature errors */
 	bool			signature_en;
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 203c8a45e095..45722a4fa99d 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -930,8 +930,6 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 			err = create_kernel_qp(dev, init_attr, qp, &in, &inlen);
 			if (err)
 				mlx5_ib_dbg(dev, "err %d\n", err);
-			else
-				qp->pa_lkey = to_mpd(pd)->pa_lkey;
 		}
 
 		if (err)
@@ -2050,7 +2048,7 @@ static void set_frwr_pages(struct mlx5_wqe_data_seg *dseg,
 		mfrpl->mapped_page_list[i] = cpu_to_be64(page_list[i] | perm);
 	dseg->addr = cpu_to_be64(mfrpl->map);
 	dseg->byte_count = cpu_to_be32(ALIGN(sizeof(u64) * wr->wr.fast_reg.page_list_len, 64));
-	dseg->lkey = cpu_to_be32(pd->pa_lkey);
+	dseg->lkey = cpu_to_be32(pd->ibpd.local_dma_lkey);
 }
 
 static __be32 send_ieth(struct ib_send_wr *wr)
-- 
1.8.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-rc v1 1/4] xprtrdma: Replace global lkey with lkey local to PD

[Chuck Lever]

> On Sep 21, 2015, at 9:41 AM, Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org> wrote:
> 
> From: Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
> 
> The core API has changed so that devices that do not have a global
> DMA lkey automatically create an mr, per-PD, and make that lkey
> available. The global DMA lkey interface is going away in favor of
> the per-PD DMA lkey.
> 
> The per-PD DMA lkey is always available. Convert xprtrdma to use the
> device's per-PD DMA lkey for regbufs, no matter which memory
> registration scheme is in use.
> 
> Signed-off-by: Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
> Signed-off-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

You’ll probably need an Acked-by from Anna or Trond.


> ---
> net/sunrpc/xprtrdma/fmr_ops.c      | 19 -------------------
> net/sunrpc/xprtrdma/frwr_ops.c     |  5 -----
> net/sunrpc/xprtrdma/physical_ops.c | 10 +---------
> net/sunrpc/xprtrdma/verbs.c        |  2 +-
> net/sunrpc/xprtrdma/xprt_rdma.h    |  1 -
> 5 files changed, 2 insertions(+), 35 deletions(-)
> 
> diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c
> index cb25c89da623..f1e8dafbd507 100644
> --- a/net/sunrpc/xprtrdma/fmr_ops.c
> +++ b/net/sunrpc/xprtrdma/fmr_ops.c
> @@ -39,25 +39,6 @@ static int
> fmr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
> 	    struct rpcrdma_create_data_internal *cdata)
> {
> -	struct ib_device_attr *devattr = &ia->ri_devattr;
> -	struct ib_mr *mr;
> -
> -	/* Obtain an lkey to use for the regbufs, which are
> -	 * protected from remote access.
> -	 */
> -	if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) {
> -		ia->ri_dma_lkey = ia->ri_device->local_dma_lkey;
> -	} else {
> -		mr = ib_get_dma_mr(ia->ri_pd, IB_ACCESS_LOCAL_WRITE);
> -		if (IS_ERR(mr)) {
> -			pr_err("%s: ib_get_dma_mr for failed with %lX\n",
> -			       __func__, PTR_ERR(mr));
> -			return -ENOMEM;
> -		}
> -		ia->ri_dma_lkey = ia->ri_dma_mr->lkey;
> -		ia->ri_dma_mr = mr;
> -	}
> -
> 	return 0;
> }
> 
> diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
> index d6653f5d0830..5318951b3b53 100644
> --- a/net/sunrpc/xprtrdma/frwr_ops.c
> +++ b/net/sunrpc/xprtrdma/frwr_ops.c
> @@ -189,11 +189,6 @@ frwr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
> 	struct ib_device_attr *devattr = &ia->ri_devattr;
> 	int depth, delta;
> 
> -	/* Obtain an lkey to use for the regbufs, which are
> -	 * protected from remote access.
> -	 */
> -	ia->ri_dma_lkey = ia->ri_device->local_dma_lkey;
> -
> 	ia->ri_max_frmr_depth =
> 			min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS,
> 			      devattr->max_fast_reg_page_list_len);
> diff --git a/net/sunrpc/xprtrdma/physical_ops.c b/net/sunrpc/xprtrdma/physical_ops.c
> index 72cf8b15bbb4..617b76f22154 100644
> --- a/net/sunrpc/xprtrdma/physical_ops.c
> +++ b/net/sunrpc/xprtrdma/physical_ops.c
> @@ -23,7 +23,6 @@ static int
> physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
> 		 struct rpcrdma_create_data_internal *cdata)
> {
> -	struct ib_device_attr *devattr = &ia->ri_devattr;
> 	struct ib_mr *mr;
> 
> 	/* Obtain an rkey to use for RPC data payloads.
> @@ -37,15 +36,8 @@ physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
> 		       __func__, PTR_ERR(mr));
> 		return -ENOMEM;
> 	}
> -	ia->ri_dma_mr = mr;
> -
> -	/* Obtain an lkey to use for regbufs.
> -	 */
> -	if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)
> -		ia->ri_dma_lkey = ia->ri_device->local_dma_lkey;
> -	else
> -		ia->ri_dma_lkey = ia->ri_dma_mr->lkey;
> 
> +	ia->ri_dma_mr = mr;
> 	return 0;
> }
> 
> diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
> index 682996779970..eb081ad05e33 100644
> --- a/net/sunrpc/xprtrdma/verbs.c
> +++ b/net/sunrpc/xprtrdma/verbs.c
> @@ -1252,7 +1252,7 @@ rpcrdma_alloc_regbuf(struct rpcrdma_ia *ia, size_t size, gfp_t flags)
> 		goto out_free;
> 
> 	iov->length = size;
> -	iov->lkey = ia->ri_dma_lkey;
> +	iov->lkey = ia->ri_pd->local_dma_lkey;
> 	rb->rg_size = size;
> 	rb->rg_owner = NULL;
> 	return rb;
> diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
> index d252457ff21a..df5ad4e15702 100644
> --- a/net/sunrpc/xprtrdma/xprt_rdma.h
> +++ b/net/sunrpc/xprtrdma/xprt_rdma.h
> @@ -66,7 +66,6 @@ struct rpcrdma_ia {
> 	struct rdma_cm_id 	*ri_id;
> 	struct ib_pd		*ri_pd;
> 	struct ib_mr		*ri_dma_mr;
> -	u32			ri_dma_lkey;
> 	struct completion	ri_done;
> 	int			ri_async_rc;
> 	unsigned int		ri_max_frmr_depth;
> -- 
> 1.8.4.3
> 

—
Chuck Lever



--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-rc v1 0/4] Remove mlx5 support for IB_DEVICE_LOCAL_DMA_LKEY

[Or Gerlitz]

On Mon, Sep 21, 2015 at 7:41 PM, Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org> wrote:
> Changes from v0:
> - Replace xprtrdma patch to Chuck's one
> - Fixed typo in iser modparam description

The change log of patch #3 needs isn't phrased clearly, this is quick
feedback I provided you for V0 and you didn't apply here, please do.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-rc v1 0/4] Remove mlx5 support for IB_DEVICE_LOCAL_DMA_LKEY

[Or Gerlitz]

On Mon, Sep 21, 2015 at 11:03 PM, Or Gerlitz <gerlitz.or-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> On Mon, Sep 21, 2015 at 7:41 PM, Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org> wrote:
>> Changes from v0:
>> - Replace xprtrdma patch to Chuck's one
>> - Fixed typo in iser modparam description
>
> The change log of patch #3 needs isn't phrased clearly, this is quick
> feedback I provided you for V0 and you didn't apply here, please do.

Again, the change log of patch #3 isn't phrased clearly, this is quick
feedback I provided you for V0 and you didn't apply here, please do.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-rc v1 1/4] xprtrdma: Replace global lkey with lkey local to PD

[Sagi Grimberg]

On 9/21/2015 8:03 PM, Chuck Lever wrote:
>
>> On Sep 21, 2015, at 9:41 AM, Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org> wrote:
>>
>> From: Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
>>
>> The core API has changed so that devices that do not have a global
>> DMA lkey automatically create an mr, per-PD, and make that lkey
>> available. The global DMA lkey interface is going away in favor of
>> the per-PD DMA lkey.
>>
>> The per-PD DMA lkey is always available. Convert xprtrdma to use the
>> device's per-PD DMA lkey for regbufs, no matter which memory
>> registration scheme is in use.
>>
>> Signed-off-by: Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
>> Signed-off-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
>
> You’ll probably need an Acked-by from Anna or Trond.

Yea, I have a v2 as I forgot to address a comment from Or.

I'll CC linux-nfs on this patch. Sorry for forgetting.

Thanks Chuck.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH rdma-rc v1 0/4] Remove mlx5 support for IB_DEVICE_LOCAL_DMA_LKEY

[Sagi Grimberg]

On 9/21/2015 11:03 PM, Or Gerlitz wrote:
> On Mon, Sep 21, 2015 at 11:03 PM, Or Gerlitz <gerlitz.or-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
>> On Mon, Sep 21, 2015 at 7:41 PM, Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org> wrote:
>>> Changes from v0:
>>> - Replace xprtrdma patch to Chuck's one
>>> - Fixed typo in iser modparam description
>>
>> The change log of patch #3 needs isn't phrased clearly, this is quick
>> feedback I provided you for V0 and you didn't apply here, please do.
>
> Again, the change log of patch #3 isn't phrased clearly, this is quick
> feedback I provided you for V0 and you didn't apply here, please do.

You're right. Sorry.

I'll send out v2 tomorrow morning.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html