[PATCH for-next 0/3] Add memory window user-space support to mlx5

[PATCH for-next 0/3] Add memory window user-space support to mlx5

[Matan Barak]

Hi Doug,

This series adds memory window user-space support to the mlx5 driver.
Memory windows in mlx5 based hardware consists of a Mkey pointing to
another set of MKeys by using KLMs as pointers.

The required kernel part for user-space memory windows consists of
allocating and deallocating these MKeys and exposing the supported
functionality via query_device.
Currently, we support Mkeys that either check the QP or doesn't
check it. This correspond to memory window of type 2B and 1
(respectively).

The user-space driver could pass the required number of KLMs in order
to create more sophisticated Mkeys.

Patch 0001 refactores the mlx5_core_mr to mlx5_core_mkey, as mr
doesn't represent accurately enough the usage of mkey to memory
windows.
Patch 0002 adds vendor specific data to ib_uverbs_alloc_mw.
Patch 0003 adds the actual memory window support to the mlx5 driver.

This series depends on
"net/mlx5: Use offset based reserved field names in the IFC header file"
posted and accepted to net.

This series conflicts (trivially) with memory re-registeration for mlx5
posted on linux-rdma.

Thanks,
Matan

Matan Barak (3):
  net/mlx5: Refactor mlx5_core_mr to mkey
  IB/core: Add vendor's specific data to alloc mw
  IB/mlx5: Add memory windows allocation support

 drivers/infiniband/core/uverbs_cmd.c              |   8 +-
 drivers/infiniband/hw/cxgb3/iwch_provider.c       |   3 +-
 drivers/infiniband/hw/cxgb4/iw_cxgb4.h            |   3 +-
 drivers/infiniband/hw/cxgb4/mem.c                 |   4 +-
 drivers/infiniband/hw/mlx4/mlx4_ib.h              |   3 +-
 drivers/infiniband/hw/mlx4/mr.c                   |   4 +-
 drivers/infiniband/hw/mlx5/cq.c                   |  16 +--
 drivers/infiniband/hw/mlx5/main.c                 |  13 +++
 drivers/infiniband/hw/mlx5/mlx5_ib.h              |  20 +++-
 drivers/infiniband/hw/mlx5/mr.c                   | 133 ++++++++++++++++++----
 drivers/infiniband/hw/mlx5/odp.c                  |  10 +-
 drivers/infiniband/hw/mlx5/user.h                 |   7 ++
 drivers/infiniband/hw/nes/nes_verbs.c             |   3 +-
 drivers/net/ethernet/mellanox/mlx5/core/en.h      |   2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c |  12 +-
 drivers/net/ethernet/mellanox/mlx5/core/main.c    |   6 +-
 drivers/net/ethernet/mellanox/mlx5/core/mr.c      |  54 ++++-----
 include/linux/mlx5/driver.h                       |  24 ++--
 include/linux/mlx5/mlx5_ifc.h                     |   4 +-
 include/linux/mlx5/qp.h                           |   4 +-
 include/rdma/ib_verbs.h                           |   3 +-
 21 files changed, 237 insertions(+), 99 deletions(-)

-- 
2.5.0

--
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 for-next 1/3] net/mlx5: Refactor mlx5_core_mr to mkey

[Matan Barak]

Mlx5's mkey mechanism is also used for memory windows.
The current code base uses MR (memory region) naming, which is
inaccurate. Changing MR to mkey in order to represent its different
usages more accurately.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx5/cq.c                   | 16 +++----
 drivers/infiniband/hw/mlx5/mlx5_ib.h              |  6 +--
 drivers/infiniband/hw/mlx5/mr.c                   | 50 ++++++++++-----------
 drivers/infiniband/hw/mlx5/odp.c                  | 10 ++---
 drivers/net/ethernet/mellanox/mlx5/core/en.h      |  2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 12 ++---
 drivers/net/ethernet/mellanox/mlx5/core/main.c    |  6 +--
 drivers/net/ethernet/mellanox/mlx5/core/mr.c      | 54 ++++++++++++-----------
 include/linux/mlx5/driver.h                       | 24 +++++-----
 include/linux/mlx5/qp.h                           |  4 +-
 10 files changed, 94 insertions(+), 90 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index fd1de31..28b69b4 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -431,7 +431,7 @@ static int mlx5_poll_one(struct mlx5_ib_cq *cq,
 	struct mlx5_core_qp *mqp;
 	struct mlx5_ib_wq *wq;
 	struct mlx5_sig_err_cqe *sig_err_cqe;
-	struct mlx5_core_mr *mmr;
+	struct mlx5_core_mkey *mmkey;
 	struct mlx5_ib_mr *mr;
 	uint8_t opcode;
 	uint32_t qpn;
@@ -536,17 +536,17 @@ repoll:
 	case MLX5_CQE_SIG_ERR:
 		sig_err_cqe = (struct mlx5_sig_err_cqe *)cqe64;
 
-		read_lock(&dev->mdev->priv.mr_table.lock);
-		mmr = __mlx5_mr_lookup(dev->mdev,
-				       mlx5_base_mkey(be32_to_cpu(sig_err_cqe->mkey)));
-		if (unlikely(!mmr)) {
-			read_unlock(&dev->mdev->priv.mr_table.lock);
+		read_lock(&dev->mdev->priv.mkey_table.lock);
+		mmkey = __mlx5_mr_lookup(dev->mdev,
+					 mlx5_base_mkey(be32_to_cpu(sig_err_cqe->mkey)));
+		if (unlikely(!mmkey)) {
+			read_unlock(&dev->mdev->priv.mkey_table.lock);
 			mlx5_ib_warn(dev, "CQE@CQ %06x for unknown MR %6x\n",
 				     cq->mcq.cqn, be32_to_cpu(sig_err_cqe->mkey));
 			return -EINVAL;
 		}
 
-		mr = to_mibmr(mmr);
+		mr = to_mibmr(mmkey);
 		get_sig_err_item(sig_err_cqe, &mr->sig->err_item);
 		mr->sig->sig_err_exists = true;
 		mr->sig->sigerr_count++;
@@ -558,7 +558,7 @@ repoll:
 			     mr->sig->err_item.expected,
 			     mr->sig->err_item.actual);
 
-		read_unlock(&dev->mdev->priv.mr_table.lock);
+		read_unlock(&dev->mdev->priv.mkey_table.lock);
 		goto repoll;
 	}
 
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index d2b9737..b1cad93 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -413,7 +413,7 @@ struct mlx5_ib_mr {
 	int			ndescs;
 	int			max_descs;
 	int			desc_size;
-	struct mlx5_core_mr	mmr;
+	struct mlx5_core_mkey	mmkey;
 	struct ib_umem	       *umem;
 	struct mlx5_shared_mr_info	*smr_info;
 	struct list_head	list;
@@ -558,9 +558,9 @@ static inline struct mlx5_ib_qp *to_mibqp(struct mlx5_core_qp *mqp)
 	return container_of(mqp, struct mlx5_ib_qp_base, mqp)->container_mibqp;
 }
 
-static inline struct mlx5_ib_mr *to_mibmr(struct mlx5_core_mr *mmr)
+static inline struct mlx5_ib_mr *to_mibmr(struct mlx5_core_mkey *mmkey)
 {
-	return container_of(mmr, struct mlx5_ib_mr, mmr);
+	return container_of(mmkey, struct mlx5_ib_mr, mmkey);
 }
 
 static inline struct mlx5_ib_pd *to_mpd(struct ib_pd *ibpd)
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 6000f7a..fbcccc8 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -57,7 +57,7 @@ static int clean_mr(struct mlx5_ib_mr *mr);
 
 static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
-	int err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmr);
+	int err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
 
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 	/* Wait until all page fault handlers using the mr complete. */
@@ -86,7 +86,7 @@ static void reg_mr_callback(int status, void *context)
 	struct mlx5_cache_ent *ent = &cache->ent[c];
 	u8 key;
 	unsigned long flags;
-	struct mlx5_mr_table *table = &dev->mdev->priv.mr_table;
+	struct mlx5_mkey_table *table = &dev->mdev->priv.mkey_table;
 	int err;
 
 	spin_lock_irqsave(&ent->lock, flags);
@@ -113,7 +113,7 @@ static void reg_mr_callback(int status, void *context)
 	spin_lock_irqsave(&dev->mdev->priv.mkey_lock, flags);
 	key = dev->mdev->priv.mkey_key++;
 	spin_unlock_irqrestore(&dev->mdev->priv.mkey_lock, flags);
-	mr->mmr.key = mlx5_idx_to_mkey(be32_to_cpu(mr->out.mkey) & 0xffffff) | key;
+	mr->mmkey.key = mlx5_idx_to_mkey(be32_to_cpu(mr->out.mkey) & 0xffffff) | key;
 
 	cache->last_add = jiffies;
 
@@ -124,10 +124,10 @@ static void reg_mr_callback(int status, void *context)
 	spin_unlock_irqrestore(&ent->lock, flags);
 
 	write_lock_irqsave(&table->lock, flags);
-	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mr->mmr.key),
-				&mr->mmr);
+	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mr->mmkey.key),
+				&mr->mmkey);
 	if (err)
-		pr_err("Error inserting to mr tree. 0x%x\n", -err);
+		pr_err("Error inserting to mkey tree. 0x%x\n", -err);
 	write_unlock_irqrestore(&table->lock, flags);
 }
 
@@ -168,7 +168,7 @@ static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
 		spin_lock_irq(&ent->lock);
 		ent->pending++;
 		spin_unlock_irq(&ent->lock);
-		err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in,
+		err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in,
 					    sizeof(*in), reg_mr_callback,
 					    mr, &mr->out);
 		if (err) {
@@ -657,14 +657,14 @@ struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc)
 	seg->qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
 	seg->start_addr = 0;
 
-	err = mlx5_core_create_mkey(mdev, &mr->mmr, in, sizeof(*in), NULL, NULL,
+	err = mlx5_core_create_mkey(mdev, &mr->mmkey, in, sizeof(*in), NULL, NULL,
 				    NULL);
 	if (err)
 		goto err_in;
 
 	kfree(in);
-	mr->ibmr.lkey = mr->mmr.key;
-	mr->ibmr.rkey = mr->mmr.key;
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
 	mr->umem = NULL;
 
 	return &mr->ibmr;
@@ -813,7 +813,7 @@ static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
 
 	memset(&umrwr, 0, sizeof(umrwr));
 	umrwr.wr.wr_id = (u64)(unsigned long)&umr_context;
-	prep_umr_reg_wqe(pd, &umrwr.wr, &sg, dma, npages, mr->mmr.key,
+	prep_umr_reg_wqe(pd, &umrwr.wr, &sg, dma, npages, mr->mmkey.key,
 			 page_shift, virt_addr, len, access_flags);
 
 	mlx5_ib_init_umr_context(&umr_context);
@@ -830,9 +830,9 @@ static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
 		}
 	}
 
-	mr->mmr.iova = virt_addr;
-	mr->mmr.size = len;
-	mr->mmr.pd = to_mpd(pd)->pdn;
+	mr->mmkey.iova = virt_addr;
+	mr->mmkey.size = len;
+	mr->mmkey.pd = to_mpd(pd)->pdn;
 
 	mr->live = 1;
 
@@ -944,7 +944,7 @@ int mlx5_ib_update_mtt(struct mlx5_ib_mr *mr, u64 start_page_index, int npages,
 		wr.wr.opcode = MLX5_IB_WR_UMR;
 		wr.npages = sg.length / sizeof(u64);
 		wr.page_shift = PAGE_SHIFT;
-		wr.mkey = mr->mmr.key;
+		wr.mkey = mr->mmkey.key;
 		wr.target.offset = start_page_index;
 
 		mlx5_ib_init_umr_context(&umr_context);
@@ -1013,7 +1013,7 @@ static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr,
 	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
 	in->xlat_oct_act_size = cpu_to_be32(get_octo_len(virt_addr, length,
 							 1 << page_shift));
-	err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in, inlen, NULL,
+	err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, inlen, NULL,
 				    NULL, NULL);
 	if (err) {
 		mlx5_ib_warn(dev, "create mkey failed\n");
@@ -1024,7 +1024,7 @@ static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr,
 	mr->live = 1;
 	kvfree(in);
 
-	mlx5_ib_dbg(dev, "mkey = 0x%x\n", mr->mmr.key);
+	mlx5_ib_dbg(dev, "mkey = 0x%x\n", mr->mmkey.key);
 
 	return mr;
 
@@ -1091,13 +1091,13 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 		goto error;
 	}
 
-	mlx5_ib_dbg(dev, "mkey 0x%x\n", mr->mmr.key);
+	mlx5_ib_dbg(dev, "mkey 0x%x\n", mr->mmkey.key);
 
 	mr->umem = umem;
 	mr->npages = npages;
 	atomic_add(npages, &dev->mdev->priv.reg_pages);
-	mr->ibmr.lkey = mr->mmr.key;
-	mr->ibmr.rkey = mr->mmr.key;
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
 
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 	if (umem->odp_data) {
@@ -1141,7 +1141,7 @@ static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 
 	memset(&umrwr.wr, 0, sizeof(umrwr));
 	umrwr.wr.wr_id = (u64)(unsigned long)&umr_context;
-	prep_umr_unreg_wqe(dev, &umrwr.wr, mr->mmr.key);
+	prep_umr_unreg_wqe(dev, &umrwr.wr, mr->mmkey.key);
 
 	mlx5_ib_init_umr_context(&umr_context);
 	down(&umrc->sem);
@@ -1236,7 +1236,7 @@ static int clean_mr(struct mlx5_ib_mr *mr)
 		err = destroy_mkey(dev, mr);
 		if (err) {
 			mlx5_ib_warn(dev, "failed to destroy mkey 0x%x (%d)\n",
-				     mr->mmr.key, err);
+				     mr->mmkey.key, err);
 			return err;
 		}
 	} else {
@@ -1362,13 +1362,13 @@ struct ib_mr *mlx5_ib_alloc_mr(struct ib_pd *pd,
 	}
 
 	in->seg.flags = MLX5_PERM_UMR_EN | access_mode;
-	err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in, sizeof(*in),
+	err = mlx5_core_create_mkey(dev->mdev, &mr->mmkey, in, sizeof(*in),
 				    NULL, NULL, NULL);
 	if (err)
 		goto err_destroy_psv;
 
-	mr->ibmr.lkey = mr->mmr.key;
-	mr->ibmr.rkey = mr->mmr.key;
+	mr->ibmr.lkey = mr->mmkey.key;
+	mr->ibmr.rkey = mr->mmkey.key;
 	mr->umem = NULL;
 	kfree(in);
 
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index b8d7636..34e79e7 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -142,13 +142,13 @@ static struct mlx5_ib_mr *mlx5_ib_odp_find_mr_lkey(struct mlx5_ib_dev *dev,
 						   u32 key)
 {
 	u32 base_key = mlx5_base_mkey(key);
-	struct mlx5_core_mr *mmr = __mlx5_mr_lookup(dev->mdev, base_key);
-	struct mlx5_ib_mr *mr = container_of(mmr, struct mlx5_ib_mr, mmr);
+	struct mlx5_core_mkey *mmkey = __mlx5_mr_lookup(dev->mdev, base_key);
+	struct mlx5_ib_mr *mr = container_of(mmkey, struct mlx5_ib_mr, mmkey);
 
-	if (!mmr || mmr->key != key || !mr->live)
+	if (!mmkey || mmkey->key != key || !mr->live)
 		return NULL;
 
-	return container_of(mmr, struct mlx5_ib_mr, mmr);
+	return container_of(mmkey, struct mlx5_ib_mr, mmkey);
 }
 
 static void mlx5_ib_page_fault_resume(struct mlx5_ib_qp *qp,
@@ -232,7 +232,7 @@ static int pagefault_single_data_segment(struct mlx5_ib_qp *qp,
 	io_virt += pfault->mpfault.bytes_committed;
 	bcnt -= pfault->mpfault.bytes_committed;
 
-	start_idx = (io_virt - (mr->mmr.iova & PAGE_MASK)) >> PAGE_SHIFT;
+	start_idx = (io_virt - (mr->mmkey.iova & PAGE_MASK)) >> PAGE_SHIFT;
 
 	if (mr->umem->writable)
 		access_mask |= ODP_WRITE_ALLOWED_BIT;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index aac071a..6ef0bfd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -515,7 +515,7 @@ struct mlx5e_priv {
 	struct mlx5_uar            cq_uar;
 	u32                        pdn;
 	u32                        tdn;
-	struct mlx5_core_mr        mr;
+	struct mlx5_core_mkey      mkey;
 	struct mlx5e_rq            drop_rq;
 
 	struct mlx5e_channel     **channel;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 6a3e430..82570ea 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -982,7 +982,7 @@ static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
 	c->cpu      = cpu;
 	c->pdev     = &priv->mdev->pdev->dev;
 	c->netdev   = priv->netdev;
-	c->mkey_be  = cpu_to_be32(priv->mr.key);
+	c->mkey_be  = cpu_to_be32(priv->mkey.key);
 	c->num_tc   = priv->params.num_tc;
 
 	mlx5e_build_channeltc_to_txq_map(priv, ix);
@@ -2182,7 +2182,7 @@ static void mlx5e_build_netdev(struct net_device *netdev)
 }
 
 static int mlx5e_create_mkey(struct mlx5e_priv *priv, u32 pdn,
-			     struct mlx5_core_mr *mr)
+			     struct mlx5_core_mkey *mkey)
 {
 	struct mlx5_core_dev *mdev = priv->mdev;
 	struct mlx5_create_mkey_mbox_in *in;
@@ -2198,7 +2198,7 @@ static int mlx5e_create_mkey(struct mlx5e_priv *priv, u32 pdn,
 	in->seg.flags_pd = cpu_to_be32(pdn | MLX5_MKEY_LEN64);
 	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
 
-	err = mlx5_core_create_mkey(mdev, mr, in, sizeof(*in), NULL, NULL,
+	err = mlx5_core_create_mkey(mdev, mkey, in, sizeof(*in), NULL, NULL,
 				    NULL);
 
 	kvfree(in);
@@ -2247,7 +2247,7 @@ static void *mlx5e_create_netdev(struct mlx5_core_dev *mdev)
 		goto err_dealloc_pd;
 	}
 
-	err = mlx5e_create_mkey(priv, priv->pdn, &priv->mr);
+	err = mlx5e_create_mkey(priv, priv->pdn, &priv->mkey);
 	if (err) {
 		mlx5_core_err(mdev, "create mkey failed, %d\n", err);
 		goto err_dealloc_transport_domain;
@@ -2321,7 +2321,7 @@ err_destroy_tises:
 	mlx5e_destroy_tises(priv);
 
 err_destroy_mkey:
-	mlx5_core_destroy_mkey(mdev, &priv->mr);
+	mlx5_core_destroy_mkey(mdev, &priv->mkey);
 
 err_dealloc_transport_domain:
 	mlx5_core_dealloc_transport_domain(mdev, priv->tdn);
@@ -2355,7 +2355,7 @@ static void mlx5e_destroy_netdev(struct mlx5_core_dev *mdev, void *vpriv)
 	mlx5e_destroy_rqt(priv, MLX5E_INDIRECTION_RQT);
 	mlx5e_close_drop_rq(priv);
 	mlx5e_destroy_tises(priv);
-	mlx5_core_destroy_mkey(priv->mdev, &priv->mr);
+	mlx5_core_destroy_mkey(priv->mdev, &priv->mkey);
 	mlx5_core_dealloc_transport_domain(priv->mdev, priv->tdn);
 	mlx5_core_dealloc_pd(priv->mdev, priv->pdn);
 	mlx5_unmap_free_uar(priv->mdev, &priv->cq_uar);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 1545a94..0916bbc 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -1117,7 +1117,7 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 	mlx5_init_cq_table(dev);
 	mlx5_init_qp_table(dev);
 	mlx5_init_srq_table(dev);
-	mlx5_init_mr_table(dev);
+	mlx5_init_mkey_table(dev);
 
 	err = mlx5_init_fs(dev);
 	if (err) {
@@ -1164,7 +1164,7 @@ err_sriov:
 err_reg_dev:
 	mlx5_cleanup_fs(dev);
 err_fs:
-	mlx5_cleanup_mr_table(dev);
+	mlx5_cleanup_mkey_table(dev);
 	mlx5_cleanup_srq_table(dev);
 	mlx5_cleanup_qp_table(dev);
 	mlx5_cleanup_cq_table(dev);
@@ -1237,7 +1237,7 @@ static int mlx5_unload_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
 #endif
 
 	mlx5_cleanup_fs(dev);
-	mlx5_cleanup_mr_table(dev);
+	mlx5_cleanup_mkey_table(dev);
 	mlx5_cleanup_srq_table(dev);
 	mlx5_cleanup_qp_table(dev);
 	mlx5_cleanup_cq_table(dev);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mr.c b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
index 6fa22b5..77a72939 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mr.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
@@ -36,25 +36,26 @@
 #include <linux/mlx5/cmd.h>
 #include "mlx5_core.h"
 
-void mlx5_init_mr_table(struct mlx5_core_dev *dev)
+void mlx5_init_mkey_table(struct mlx5_core_dev *dev)
 {
-	struct mlx5_mr_table *table = &dev->priv.mr_table;
+	struct mlx5_mkey_table *table = &dev->priv.mkey_table;
 
 	memset(table, 0, sizeof(*table));
 	rwlock_init(&table->lock);
 	INIT_RADIX_TREE(&table->tree, GFP_ATOMIC);
 }
 
-void mlx5_cleanup_mr_table(struct mlx5_core_dev *dev)
+void mlx5_cleanup_mkey_table(struct mlx5_core_dev *dev)
 {
 }
 
-int mlx5_core_create_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
+			  struct mlx5_core_mkey *mkey,
 			  struct mlx5_create_mkey_mbox_in *in, int inlen,
 			  mlx5_cmd_cbk_t callback, void *context,
 			  struct mlx5_create_mkey_mbox_out *out)
 {
-	struct mlx5_mr_table *table = &dev->priv.mr_table;
+	struct mlx5_mkey_table *table = &dev->priv.mkey_table;
 	struct mlx5_create_mkey_mbox_out lout;
 	int err;
 	u8 key;
@@ -83,34 +84,35 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
 		return mlx5_cmd_status_to_err(&lout.hdr);
 	}
 
-	mr->iova = be64_to_cpu(in->seg.start_addr);
-	mr->size = be64_to_cpu(in->seg.len);
-	mr->key = mlx5_idx_to_mkey(be32_to_cpu(lout.mkey) & 0xffffff) | key;
-	mr->pd = be32_to_cpu(in->seg.flags_pd) & 0xffffff;
+	mkey->iova = be64_to_cpu(in->seg.start_addr);
+	mkey->size = be64_to_cpu(in->seg.len);
+	mkey->key = mlx5_idx_to_mkey(be32_to_cpu(lout.mkey) & 0xffffff) | key;
+	mkey->pd = be32_to_cpu(in->seg.flags_pd) & 0xffffff;
 
 	mlx5_core_dbg(dev, "out 0x%x, key 0x%x, mkey 0x%x\n",
-		      be32_to_cpu(lout.mkey), key, mr->key);
+		      be32_to_cpu(lout.mkey), key, mkey->key);
 
-	/* connect to MR tree */
+	/* connect to mkey tree */
 	write_lock_irq(&table->lock);
-	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mr->key), mr);
+	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mkey->key), mkey);
 	write_unlock_irq(&table->lock);
 	if (err) {
-		mlx5_core_warn(dev, "failed radix tree insert of mr 0x%x, %d\n",
-			       mlx5_base_mkey(mr->key), err);
-		mlx5_core_destroy_mkey(dev, mr);
+		mlx5_core_warn(dev, "failed radix tree insert of mkey 0x%x, %d\n",
+			       mlx5_base_mkey(mkey->key), err);
+		mlx5_core_destroy_mkey(dev, mkey);
 	}
 
 	return err;
 }
 EXPORT_SYMBOL(mlx5_core_create_mkey);
 
-int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr)
+int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev,
+			   struct mlx5_core_mkey *mkey)
 {
-	struct mlx5_mr_table *table = &dev->priv.mr_table;
+	struct mlx5_mkey_table *table = &dev->priv.mkey_table;
 	struct mlx5_destroy_mkey_mbox_in in;
 	struct mlx5_destroy_mkey_mbox_out out;
-	struct mlx5_core_mr *deleted_mr;
+	struct mlx5_core_mkey *deleted_mkey;
 	unsigned long flags;
 	int err;
 
@@ -118,16 +120,16 @@ int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr)
 	memset(&out, 0, sizeof(out));
 
 	write_lock_irqsave(&table->lock, flags);
-	deleted_mr = radix_tree_delete(&table->tree, mlx5_base_mkey(mr->key));
+	deleted_mkey = radix_tree_delete(&table->tree, mlx5_base_mkey(mkey->key));
 	write_unlock_irqrestore(&table->lock, flags);
-	if (!deleted_mr) {
-		mlx5_core_warn(dev, "failed radix tree delete of mr 0x%x\n",
-			       mlx5_base_mkey(mr->key));
+	if (!deleted_mkey) {
+		mlx5_core_warn(dev, "failed radix tree delete of mkey 0x%x\n",
+			       mlx5_base_mkey(mkey->key));
 		return -ENOENT;
 	}
 
 	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_DESTROY_MKEY);
-	in.mkey = cpu_to_be32(mlx5_mkey_to_idx(mr->key));
+	in.mkey = cpu_to_be32(mlx5_mkey_to_idx(mkey->key));
 	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
 	if (err)
 		return err;
@@ -139,7 +141,7 @@ int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr)
 }
 EXPORT_SYMBOL(mlx5_core_destroy_mkey);
 
-int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mkey *mkey,
 			 struct mlx5_query_mkey_mbox_out *out, int outlen)
 {
 	struct mlx5_query_mkey_mbox_in in;
@@ -149,7 +151,7 @@ int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
 	memset(out, 0, outlen);
 
 	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_MKEY);
-	in.mkey = cpu_to_be32(mlx5_mkey_to_idx(mr->key));
+	in.mkey = cpu_to_be32(mlx5_mkey_to_idx(mkey->key));
 	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
 	if (err)
 		return err;
@@ -161,7 +163,7 @@ int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
 }
 EXPORT_SYMBOL(mlx5_core_query_mkey);
 
-int mlx5_core_dump_fill_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+int mlx5_core_dump_fill_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mkey *_mkey,
 			     u32 *mkey)
 {
 	struct mlx5_query_special_ctxs_mbox_in in;
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 1e3006d..8cb3c97 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -338,7 +338,7 @@ struct mlx5_core_sig_ctx {
 	u32			sigerr_count;
 };
 
-struct mlx5_core_mr {
+struct mlx5_core_mkey {
 	u64			iova;
 	u64			size;
 	u32			key;
@@ -426,7 +426,7 @@ struct mlx5_srq_table {
 	struct radix_tree_root	tree;
 };
 
-struct mlx5_mr_table {
+struct mlx5_mkey_table {
 	/* protect radix tree
 	 */
 	rwlock_t		lock;
@@ -484,9 +484,9 @@ struct mlx5_priv {
 	struct mlx5_cq_table	cq_table;
 	/* end: cq staff */
 
-	/* start: mr staff */
-	struct mlx5_mr_table	mr_table;
-	/* end: mr staff */
+	/* start: mkey staff */
+	struct mlx5_mkey_table	mkey_table;
+	/* end: mkey staff */
 
 	/* start: alloc staff */
 	/* protect buffer alocation according to numa node */
@@ -739,16 +739,18 @@ int mlx5_core_query_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
 			struct mlx5_query_srq_mbox_out *out);
 int mlx5_core_arm_srq(struct mlx5_core_dev *dev, struct mlx5_core_srq *srq,
 		      u16 lwm, int is_srq);
-void mlx5_init_mr_table(struct mlx5_core_dev *dev);
-void mlx5_cleanup_mr_table(struct mlx5_core_dev *dev);
-int mlx5_core_create_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+void mlx5_init_mkey_table(struct mlx5_core_dev *dev);
+void mlx5_cleanup_mkey_table(struct mlx5_core_dev *dev);
+int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
+			  struct mlx5_core_mkey *mkey,
 			  struct mlx5_create_mkey_mbox_in *in, int inlen,
 			  mlx5_cmd_cbk_t callback, void *context,
 			  struct mlx5_create_mkey_mbox_out *out);
-int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr);
-int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev,
+			   struct mlx5_core_mkey *mkey);
+int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mkey *mkey,
 			 struct mlx5_query_mkey_mbox_out *out, int outlen);
-int mlx5_core_dump_fill_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mr *mr,
+int mlx5_core_dump_fill_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mkey *_mkey,
 			     u32 *mkey);
 int mlx5_core_alloc_pd(struct mlx5_core_dev *dev, u32 *pdn);
 int mlx5_core_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn);
diff --git a/include/linux/mlx5/qp.h b/include/linux/mlx5/qp.h
index 5b8c89f..f46324e 100644
--- a/include/linux/mlx5/qp.h
+++ b/include/linux/mlx5/qp.h
@@ -621,9 +621,9 @@ static inline struct mlx5_core_qp *__mlx5_qp_lookup(struct mlx5_core_dev *dev, u
 	return radix_tree_lookup(&dev->priv.qp_table.tree, qpn);
 }
 
-static inline struct mlx5_core_mr *__mlx5_mr_lookup(struct mlx5_core_dev *dev, u32 key)
+static inline struct mlx5_core_mkey *__mlx5_mr_lookup(struct mlx5_core_dev *dev, u32 key)
 {
-	return radix_tree_lookup(&dev->priv.mr_table.tree, key);
+	return radix_tree_lookup(&dev->priv.mkey_table.tree, key);
 }
 
 struct mlx5_page_fault_resume_mbox_in {
-- 
2.5.0

--
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 for-next 2/3] IB/core: Add vendor's specific data to alloc mw

[Matan Barak]

Passing udata to the vendor's driver in order to pass data from the
user-space driver to the kernel-space driver. This data will be
used in downstream patches.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_cmd.c        | 8 +++++++-
 drivers/infiniband/hw/cxgb3/iwch_provider.c | 3 ++-
 drivers/infiniband/hw/cxgb4/iw_cxgb4.h      | 3 ++-
 drivers/infiniband/hw/cxgb4/mem.c           | 4 +++-
 drivers/infiniband/hw/mlx4/mlx4_ib.h        | 3 ++-
 drivers/infiniband/hw/mlx4/mr.c             | 4 +++-
 drivers/infiniband/hw/nes/nes_verbs.c       | 3 ++-
 include/rdma/ib_verbs.h                     | 3 ++-
 8 files changed, 23 insertions(+), 8 deletions(-)

diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 6ffc9c4..2bf751e 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -1174,6 +1174,7 @@ ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file,
 	struct ib_uobject             *uobj;
 	struct ib_pd                  *pd;
 	struct ib_mw                  *mw;
+	struct ib_udata		       udata;
 	int                            ret;
 
 	if (out_len < sizeof(resp))
@@ -1195,7 +1196,12 @@ ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file,
 		goto err_free;
 	}
 
-	mw = pd->device->alloc_mw(pd, cmd.mw_type);
+	INIT_UDATA(&udata, buf + sizeof(cmd),
+		   (unsigned long)cmd.response + sizeof(resp),
+		   in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr),
+		   out_len - sizeof(resp));
+
+	mw = pd->device->alloc_mw(pd, cmd.mw_type, &udata);
 	if (IS_ERR(mw)) {
 		ret = PTR_ERR(mw);
 		goto err_put;
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 2734820..42a7b89 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -657,7 +657,8 @@ err:
 	return ERR_PTR(err);
 }
 
-static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
+static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+				   struct ib_udata *udata)
 {
 	struct iwch_dev *rhp;
 	struct iwch_pd *php;
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
index fb2de75..423a3a9 100644
--- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
+++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
@@ -961,7 +961,8 @@ int c4iw_map_mr_sg(struct ib_mr *ibmr,
 		   struct scatterlist *sg,
 		   int sg_nents);
 int c4iw_dealloc_mw(struct ib_mw *mw);
-struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
+struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			    struct ib_udata *udata);
 struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start,
 					   u64 length, u64 virt, int acc,
 					   struct ib_udata *udata);
diff --git a/drivers/infiniband/hw/cxgb4/mem.c b/drivers/infiniband/hw/cxgb4/mem.c
index 7849890..766d39c 100644
--- a/drivers/infiniband/hw/cxgb4/mem.c
+++ b/drivers/infiniband/hw/cxgb4/mem.c
@@ -34,6 +34,7 @@
 #include <linux/moduleparam.h>
 #include <rdma/ib_umem.h>
 #include <linux/atomic.h>
+#include <rdma/ib_user_verbs.h>
 
 #include "iw_cxgb4.h"
 
@@ -552,7 +553,8 @@ err:
 	return ERR_PTR(err);
 }
 
-struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
+struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			    struct ib_udata *udata)
 {
 	struct c4iw_dev *rhp;
 	struct c4iw_pd *php;
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index 52ce7b0..1eca01c 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -711,7 +711,8 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata);
 int mlx4_ib_dereg_mr(struct ib_mr *mr);
-struct ib_mw *mlx4_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
+struct ib_mw *mlx4_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			       struct ib_udata *udata);
 int mlx4_ib_dealloc_mw(struct ib_mw *mw);
 struct ib_mr *mlx4_ib_alloc_mr(struct ib_pd *pd,
 			       enum ib_mr_type mr_type,
diff --git a/drivers/infiniband/hw/mlx4/mr.c b/drivers/infiniband/hw/mlx4/mr.c
index 242b94e..ce0b5aa 100644
--- a/drivers/infiniband/hw/mlx4/mr.c
+++ b/drivers/infiniband/hw/mlx4/mr.c
@@ -32,6 +32,7 @@
  */
 
 #include <linux/slab.h>
+#include <rdma/ib_user_verbs.h>
 
 #include "mlx4_ib.h"
 
@@ -334,7 +335,8 @@ int mlx4_ib_dereg_mr(struct ib_mr *ibmr)
 	return 0;
 }
 
-struct ib_mw *mlx4_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type)
+struct ib_mw *mlx4_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			       struct ib_udata *udata)
 {
 	struct mlx4_ib_dev *dev = to_mdev(pd->device);
 	struct mlx4_ib_mw *mw;
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 8c4daf7..5af19b4 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -56,7 +56,8 @@ static int nes_dereg_mr(struct ib_mr *ib_mr);
 /**
  * nes_alloc_mw
  */
-static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd, enum ib_mw_type type)
+static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd, enum ib_mw_type type,
+				  struct ib_udata *udata)
 {
 	struct nes_pd *nespd = to_nespd(ibpd);
 	struct nes_vnic *nesvnic = to_nesvnic(ibpd->device);
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 284b00c..3f79070 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -1808,7 +1808,8 @@ struct ib_device {
 						struct scatterlist *sg,
 						int sg_nents);
 	struct ib_mw *             (*alloc_mw)(struct ib_pd *pd,
-					       enum ib_mw_type type);
+					       enum ib_mw_type type,
+					       struct ib_udata *udata);
 	int                        (*dealloc_mw)(struct ib_mw *mw);
 	struct ib_fmr *	           (*alloc_fmr)(struct ib_pd *pd,
 						int mr_access_flags,
-- 
2.5.0

--
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 for-next 3/3] IB/mlx5: Add memory windows allocation support

[Matan Barak]

This patch adds user-space support for memory windows allocation and
deallocation. It also exposes the supported types via
query_device_caps verb.

Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx5/main.c    | 13 ++++++
 drivers/infiniband/hw/mlx5/mlx5_ib.h | 14 ++++++
 drivers/infiniband/hw/mlx5/mr.c      | 83 ++++++++++++++++++++++++++++++++++++
 drivers/infiniband/hw/mlx5/user.h    |  7 +++
 include/linux/mlx5/mlx5_ifc.h        |  4 +-
 5 files changed, 120 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 03c418c..bccd118 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -487,6 +487,11 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		props->device_cap_flags |= IB_DEVICE_AUTO_PATH_MIG;
 	if (MLX5_CAP_GEN(mdev, xrc))
 		props->device_cap_flags |= IB_DEVICE_XRC;
+	if (MLX5_CAP_GEN(mdev, imaicl)) {
+		props->device_cap_flags |= IB_DEVICE_MEM_WINDOW |
+					   IB_DEVICE_MEM_WINDOW_TYPE_2B;
+		props->max_mw = 1 << MLX5_CAP_GEN(mdev, log_max_mkey);
+	}
 	props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
 	if (MLX5_CAP_GEN(mdev, sho)) {
 		props->device_cap_flags |= IB_DEVICE_SIGNATURE_HANDOVER;
@@ -2269,6 +2274,14 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 
 	mlx5_ib_internal_fill_odp_caps(dev);
 
+	if (MLX5_CAP_GEN(mdev, imaicl)) {
+		dev->ib_dev.alloc_mw		= mlx5_ib_alloc_mw;
+		dev->ib_dev.dealloc_mw		= mlx5_ib_dealloc_mw;
+		dev->ib_dev.uverbs_cmd_mask |=
+			(1ull << IB_USER_VERBS_CMD_ALLOC_MW)	|
+			(1ull << IB_USER_VERBS_CMD_DEALLOC_MW);
+	}
+
 	if (MLX5_CAP_GEN(mdev, xrc)) {
 		dev->ib_dev.alloc_xrcd = mlx5_ib_alloc_xrcd;
 		dev->ib_dev.dealloc_xrcd = mlx5_ib_dealloc_xrcd;
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index b1cad93..cfd006d 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -43,6 +43,7 @@
 #include <linux/mlx5/srq.h>
 #include <linux/types.h>
 #include <linux/mlx5/transobj.h>
+#include <rdma/ib_user_verbs.h>
 
 #define mlx5_ib_dbg(dev, format, arg...)				\
 pr_debug("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
@@ -427,6 +428,11 @@ struct mlx5_ib_mr {
 	void			*descs_alloc;
 };
 
+struct mlx5_ib_mw {
+	struct ib_mw		ibmw;
+	struct mlx5_core_mkey	mmkey;
+};
+
 struct mlx5_ib_umr_context {
 	enum ib_wc_status	status;
 	struct completion	done;
@@ -588,6 +594,11 @@ static inline struct mlx5_ib_mr *to_mmr(struct ib_mr *ibmr)
 	return container_of(ibmr, struct mlx5_ib_mr, ibmr);
 }
 
+static inline struct mlx5_ib_mw *to_mmw(struct ib_mw *ibmw)
+{
+	return container_of(ibmw, struct mlx5_ib_mw, ibmw);
+}
+
 struct mlx5_ib_ah {
 	struct ib_ah		ibah;
 	struct mlx5_av		av;
@@ -648,6 +659,9 @@ struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc);
 struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata);
+struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			       struct ib_udata *udata);
+int mlx5_ib_dealloc_mw(struct ib_mw *mw);
 int mlx5_ib_update_mtt(struct mlx5_ib_mr *mr, u64 start_page_index,
 		       int npages, int zap);
 int mlx5_ib_dereg_mr(struct ib_mr *ibmr);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index fbcccc8..4ffb22b 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -40,6 +40,7 @@
 #include <rdma/ib_umem_odp.h>
 #include <rdma/ib_verbs.h>
 #include "mlx5_ib.h"
+#include "user.h"
 
 enum {
 	MAX_PENDING_REG_MR = 8,
@@ -1395,6 +1396,88 @@ err_free:
 	return ERR_PTR(err);
 }
 
+struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+			       struct ib_udata *udata)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_create_mkey_mbox_in *in = NULL;
+	struct mlx5_ib_mw *mw = NULL;
+	int ndescs;
+	int err;
+	struct mlx5_ib_alloc_mw req = {};
+	struct {
+		__u32	comp_mask;
+		__u32	response_length;
+	} resp = {};
+
+	err = ib_copy_from_udata(&req, udata, min(udata->inlen, sizeof(req)));
+	if (err)
+		return ERR_PTR(err);
+
+	if (req.comp_mask || req.reserved1 || req.reserved2)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (udata->inlen > sizeof(req) &&
+	    !ib_is_udata_cleared(udata, sizeof(req),
+				 udata->inlen - sizeof(req)))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	ndescs = req.num_klms ? roundup(req.num_klms, 4) : roundup(1, 4);
+
+	mw = kzalloc(sizeof(*mw), GFP_KERNEL);
+	in = kzalloc(sizeof(*in), GFP_KERNEL);
+	if (!mw || !in) {
+		err = -ENOMEM;
+		goto free;
+	}
+
+	in->seg.status = MLX5_MKEY_STATUS_FREE;
+	in->seg.xlt_oct_size = cpu_to_be32(ndescs);
+	in->seg.flags_pd = cpu_to_be32(to_mpd(pd)->pdn);
+	in->seg.flags = MLX5_PERM_UMR_EN | MLX5_ACCESS_MODE_KLM |
+		MLX5_PERM_LOCAL_READ;
+	if (type == IB_MW_TYPE_2)
+		in->seg.flags_pd |= cpu_to_be32(MLX5_MKEY_REMOTE_INVAL);
+	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
+
+	err = mlx5_core_create_mkey(dev->mdev, &mw->mmkey, in, sizeof(*in),
+				    NULL, NULL, NULL);
+	if (err)
+		goto free;
+
+	mw->ibmw.rkey = mw->mmkey.key;
+
+	resp.response_length = min(offsetof(typeof(resp), response_length) +
+				   sizeof(resp.response_length), udata->outlen);
+	if (resp.response_length) {
+		err = ib_copy_to_udata(udata, &resp, resp.response_length);
+		if (err) {
+			mlx5_core_destroy_mkey(dev->mdev, &mw->mmkey);
+			goto free;
+		}
+	}
+
+	kfree(in);
+	return &mw->ibmw;
+
+free:
+	kfree(mw);
+	kfree(in);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_dealloc_mw(struct ib_mw *mw)
+{
+	struct mlx5_ib_mw *mmw = to_mmw(mw);
+	int err;
+
+	err =  mlx5_core_destroy_mkey((to_mdev(mw->device))->mdev,
+				      &mmw->mmkey);
+	if (!err)
+		kfree(mmw);
+	return err;
+}
+
 int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
 			    struct ib_mr_status *mr_status)
 {
diff --git a/drivers/infiniband/hw/mlx5/user.h b/drivers/infiniband/hw/mlx5/user.h
index b94a554..61bc308 100644
--- a/drivers/infiniband/hw/mlx5/user.h
+++ b/drivers/infiniband/hw/mlx5/user.h
@@ -152,6 +152,13 @@ struct mlx5_ib_create_qp_resp {
 	__u32	uuar_index;
 };
 
+struct mlx5_ib_alloc_mw {
+	__u32	comp_mask;
+	__u8	num_klms;
+	__u8	reserved1;
+	__u16	reserved2;
+};
+
 static inline int get_qp_user_index(struct mlx5_ib_ucontext *ucontext,
 				    struct mlx5_ib_create_qp *ucmd,
 				    int inlen,
diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index 51f1e54..4fe9c6b 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -767,7 +767,9 @@ struct mlx5_ifc_cmd_hca_cap_bits {
 	u8         cd[0x1];
 	u8         reserved_at_22c[0x1];
 	u8         apm[0x1];
-	u8         reserved_at_22e[0x7];
+	u8         reserved_at_22e[0x2];
+	u8	   imaicl[0x1];
+	u8         reserved_at_231[0x4];
 	u8         qkv[0x1];
 	u8         pkv[0x1];
 	u8         reserved_at_237[0x4];
-- 
2.5.0

--
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 for-next 2/3] IB/core: Add vendor's specific data to alloc mw

[Max Gurtovoy]

On 2/29/2016 6:05 PM, Matan Barak wrote:
> Passing udata to the vendor's driver in order to pass data from the
> user-space driver to the kernel-space driver. This data will be
> used in downstream patches.
>
> Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> ---
>   drivers/infiniband/core/uverbs_cmd.c        | 8 +++++++-
>   drivers/infiniband/hw/cxgb3/iwch_provider.c | 3 ++-
>   drivers/infiniband/hw/cxgb4/iw_cxgb4.h      | 3 ++-
>   drivers/infiniband/hw/cxgb4/mem.c           | 4 +++-
>   drivers/infiniband/hw/mlx4/mlx4_ib.h        | 3 ++-
>   drivers/infiniband/hw/mlx4/mr.c             | 4 +++-
>   drivers/infiniband/hw/nes/nes_verbs.c       | 3 ++-
>   include/rdma/ib_verbs.h                     | 3 ++-
>   8 files changed, 23 insertions(+), 8 deletions(-)
>

Looks good,
Max.
--
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 for-next 3/3] IB/mlx5: Add memory windows allocation support

[Max Gurtovoy]

On 2/29/2016 6:05 PM, Matan Barak wrote:
> This patch adds user-space support for memory windows allocation and
> deallocation. It also exposes the supported types via
> query_device_caps verb.
>
> Signed-off-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> Reviewed-by: Yishai Hadas <yishaih-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> ---
>   drivers/infiniband/hw/mlx5/main.c    | 13 ++++++
>   drivers/infiniband/hw/mlx5/mlx5_ib.h | 14 ++++++
>   drivers/infiniband/hw/mlx5/mr.c      | 83 ++++++++++++++++++++++++++++++++++++
>   drivers/infiniband/hw/mlx5/user.h    |  7 +++
>   include/linux/mlx5/mlx5_ifc.h        |  4 +-
>   5 files changed, 120 insertions(+), 1 deletion(-)
>

Looks good., I've tested the alloc func with num_klms = 16 for the 
erasure coding.

Tested-by: Max Gurtovoy <maxg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

--
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