[PATCH mlx5-next 0/7] Clean MR key use across mlx5_* modules

[PATCH mlx5-next 0/7] Clean MR key use across mlx5_* modules

[Leon Romanovsky]

From: Leon Romanovsky <leonro@nvidia.com>

Hi,

This is cleanup series of mlx5_* MR mkey management.

Thanks

Aharon Landau (7):
  RDMA/mlx5: Don't set esc_size in user mr
  RDMA/mlx5: Remove iova from struct mlx5_core_mkey
  RDMA/mlx5: Remove size from struct mlx5_core_mkey
  RDMA/mlx5: Remove pd from struct mlx5_core_mkey
  RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
  RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
  RDMA/mlx5: Attach ndescs to mlx5_ib_mkey

 drivers/infiniband/hw/mlx5/devx.c             | 13 +--
 drivers/infiniband/hw/mlx5/devx.h             |  2 +-
 drivers/infiniband/hw/mlx5/mlx5_ib.h          | 31 ++++---
 drivers/infiniband/hw/mlx5/mr.c               | 82 +++++++++----------
 drivers/infiniband/hw/mlx5/odp.c              | 38 +++------
 drivers/infiniband/hw/mlx5/wr.c               | 10 +--
 .../mellanox/mlx5/core/diag/fw_tracer.c       |  6 +-
 .../mellanox/mlx5/core/diag/fw_tracer.h       |  2 +-
 .../mellanox/mlx5/core/diag/rsc_dump.c        | 10 +--
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en/trap.c |  2 +-
 .../ethernet/mellanox/mlx5/core/en_common.c   |  6 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c | 13 ++-
 .../ethernet/mellanox/mlx5/core/fpga/conn.c   | 10 +--
 .../ethernet/mellanox/mlx5/core/fpga/core.h   |  2 +-
 drivers/net/ethernet/mellanox/mlx5/core/mr.c  | 27 +++---
 .../mellanox/mlx5/core/steering/dr_icm_pool.c | 10 +--
 .../mellanox/mlx5/core/steering/dr_send.c     | 11 ++-
 .../mellanox/mlx5/core/steering/dr_types.h    |  2 +-
 drivers/vdpa/mlx5/core/mlx5_vdpa.h            |  8 +-
 drivers/vdpa/mlx5/core/mr.c                   |  8 +-
 drivers/vdpa/mlx5/core/resources.c            | 13 +--
 drivers/vdpa/mlx5/net/mlx5_vnet.c             |  2 +-
 include/linux/mlx5/driver.h                   | 30 ++-----
 25 files changed, 147 insertions(+), 195 deletions(-)

-- 
2.31.1

[PATCH mlx5-next 1/7] RDMA/mlx5: Don't set esc_size in user mr

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

reg_create() is used for user MRs only and should not set desc_size at
all. Attempt to set it causes to the following trace:

BUG: unable to handle page fault for address: 0000000800000000
PGD 0 P4D 0
Oops: 0000 [#1] SMP PTI
CPU: 5 PID: 890 Comm: ib_write_bw Not tainted 5.15.0-rc4+ #47
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
RIP: 0010:mlx5_ib_dereg_mr+0x14/0x3b0 [mlx5_ib]
Code: 48 63 cd 4c 89 f7 48 89 0c 24 e8 37 30 03 e1 48 8b 0c 24 eb a0 90 0f 1f 44 00 00 41 56 41 55 41 54 55 53 48 89 fb 48 83 ec 30 <48> 8b 2f 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 8b 87 c8
RSP: 0018:ffff88811afa3a60 EFLAGS: 00010286
RAX: 000000000000001c RBX: 0000000800000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000800000000
RBP: 0000000800000000 R08: 0000000000000000 R09: c0000000fffff7ff
R10: ffff88811afa38f8 R11: ffff88811afa38f0 R12: ffffffffa02c7ac0
R13: 0000000000000000 R14: ffff88811afa3cd8 R15: ffff88810772fa00
FS:  00007f47b9080740(0000) GS:ffff88852cd40000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000800000000 CR3: 000000010761e003 CR4: 0000000000370ea0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 mlx5_ib_free_odp_mr+0x95/0xc0 [mlx5_ib]
 mlx5_ib_dereg_mr+0x128/0x3b0 [mlx5_ib]
 ib_dereg_mr_user+0x45/0xb0 [ib_core]
 ? xas_load+0x8/0x80
 destroy_hw_idr_uobject+0x1a/0x50 [ib_uverbs]
 uverbs_destroy_uobject+0x2f/0x150 [ib_uverbs]
 uobj_destroy+0x3c/0x70 [ib_uverbs]
 ib_uverbs_cmd_verbs+0x467/0xb00 [ib_uverbs]
 ? uverbs_finalize_object+0x60/0x60 [ib_uverbs]
 ? ttwu_queue_wakelist+0xa9/0xe0
 ? pty_write+0x85/0x90
 ? file_tty_write.isra.33+0x214/0x330
 ? process_echoes+0x60/0x60
 ib_uverbs_ioctl+0xa7/0x110 [ib_uverbs]
 __x64_sys_ioctl+0x10d/0x8e0
 ? vfs_write+0x17f/0x260
 do_syscall_64+0x3c/0x80
 entry_SYSCALL_64_after_hwframe+0x44/0xae

Fixes: a639e66703ee ("RDMA/mlx5: Zero out ODP related items in the mlx5_ib_mr")
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Maor Gottlieb <maorg@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/mr.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 9091de381e83..0ffe7f5b77b6 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -1335,7 +1335,6 @@ static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, struct ib_umem *umem,
 		goto err_2;
 	}
 	mr->mmkey.type = MLX5_MKEY_MR;
-	mr->desc_size = sizeof(struct mlx5_mtt);
 	mr->umem = umem;
 	set_mr_fields(dev, mr, umem->length, access_flags);
 	kvfree(in);
-- 
2.31.1

[PATCH mlx5-next 2/7] RDMA/mlx5: Remove iova from struct mlx5_core_mkey

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

iova is already stored in ibmr->iova, no need to store it here.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/devx.c            |  1 -
 drivers/infiniband/hw/mlx5/mr.c              | 16 ++++++++--------
 drivers/infiniband/hw/mlx5/odp.c             |  8 ++++----
 drivers/net/ethernet/mellanox/mlx5/core/mr.c |  1 -
 drivers/vdpa/mlx5/core/resources.c           |  1 -
 include/linux/mlx5/driver.h                  |  1 -
 6 files changed, 12 insertions(+), 16 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
index 7c920be2921d..439b62e23afb 100644
--- a/drivers/infiniband/hw/mlx5/devx.c
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -1303,7 +1303,6 @@ static int devx_handle_mkey_indirect(struct devx_obj *obj,
 	mkey->key = mlx5_idx_to_mkey(
 			MLX5_GET(create_mkey_out, out, mkey_index)) | key;
 	mkey->type = MLX5_MKEY_INDIRECT_DEVX;
-	mkey->iova = MLX5_GET64(mkc, mkc, start_addr);
 	mkey->size = MLX5_GET64(mkc, mkc, len);
 	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	devx_mr->ndescs = MLX5_GET(mkc, mkc, translations_octword_size);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 0ffe7f5b77b6..e264c6be38f8 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -904,12 +904,13 @@ static struct mlx5_cache_ent *mr_cache_ent_from_order(struct mlx5_ib_dev *dev,
 }
 
 static void set_mr_fields(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
-			  u64 length, int access_flags)
+			  u64 length, int access_flags, u64 iova)
 {
 	mr->ibmr.lkey = mr->mmkey.key;
 	mr->ibmr.rkey = mr->mmkey.key;
 	mr->ibmr.length = length;
 	mr->ibmr.device = &dev->ib_dev;
+	mr->ibmr.iova = iova;
 	mr->access_flags = access_flags;
 }
 
@@ -967,11 +968,10 @@ static struct mlx5_ib_mr *alloc_cacheable_mr(struct ib_pd *pd,
 
 	mr->ibmr.pd = pd;
 	mr->umem = umem;
-	mr->mmkey.iova = iova;
 	mr->mmkey.size = umem->length;
 	mr->mmkey.pd = to_mpd(pd)->pdn;
 	mr->page_shift = order_base_2(page_size);
-	set_mr_fields(dev, mr, umem->length, access_flags);
+	set_mr_fields(dev, mr, umem->length, access_flags, iova);
 
 	return mr;
 }
@@ -1081,7 +1081,7 @@ static void *mlx5_ib_create_xlt_wr(struct mlx5_ib_mr *mr,
 	wr->pd = mr->ibmr.pd;
 	wr->mkey = mr->mmkey.key;
 	wr->length = mr->mmkey.size;
-	wr->virt_addr = mr->mmkey.iova;
+	wr->virt_addr = mr->ibmr.iova;
 	wr->access_flags = mr->access_flags;
 	wr->page_shift = mr->page_shift;
 	wr->xlt_size = sg->length;
@@ -1336,7 +1336,7 @@ static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, struct ib_umem *umem,
 	}
 	mr->mmkey.type = MLX5_MKEY_MR;
 	mr->umem = umem;
-	set_mr_fields(dev, mr, umem->length, access_flags);
+	set_mr_fields(dev, mr, umem->length, access_flags, iova);
 	kvfree(in);
 
 	mlx5_ib_dbg(dev, "mkey = 0x%x\n", mr->mmkey.key);
@@ -1383,7 +1383,7 @@ static struct ib_mr *mlx5_ib_get_dm_mr(struct ib_pd *pd, u64 start_addr,
 
 	kfree(in);
 
-	set_mr_fields(dev, mr, length, acc);
+	set_mr_fields(dev, mr, length, acc, start_addr);
 
 	return &mr->ibmr;
 
@@ -1767,7 +1767,7 @@ static int umr_rereg_pas(struct mlx5_ib_mr *mr, struct ib_pd *pd,
 	}
 
 	mr->ibmr.length = new_umem->length;
-	mr->mmkey.iova = iova;
+	mr->ibmr.iova = iova;
 	mr->mmkey.size = new_umem->length;
 	mr->page_shift = order_base_2(page_size);
 	mr->umem = new_umem;
@@ -1847,7 +1847,7 @@ struct ib_mr *mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
 		mr->umem = NULL;
 		atomic_sub(ib_umem_num_pages(umem), &dev->mdev->priv.reg_pages);
 
-		return create_real_mr(new_pd, umem, mr->mmkey.iova,
+		return create_real_mr(new_pd, umem, mr->ibmr.iova,
 				      new_access_flags);
 	}
 
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 77890a85fc2d..3b24cfc96d36 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -430,7 +430,7 @@ static struct mlx5_ib_mr *implicit_get_child_mr(struct mlx5_ib_mr *imr,
 	mr->umem = &odp->umem;
 	mr->ibmr.lkey = mr->mmkey.key;
 	mr->ibmr.rkey = mr->mmkey.key;
-	mr->mmkey.iova = idx * MLX5_IMR_MTT_SIZE;
+	mr->ibmr.iova = idx * MLX5_IMR_MTT_SIZE;
 	mr->parent = imr;
 	odp->private = mr;
 
@@ -500,7 +500,7 @@ struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd,
 	}
 
 	imr->ibmr.pd = &pd->ibpd;
-	imr->mmkey.iova = 0;
+	imr->ibmr.iova = 0;
 	imr->umem = &umem_odp->umem;
 	imr->ibmr.lkey = imr->mmkey.key;
 	imr->ibmr.rkey = imr->mmkey.key;
@@ -738,7 +738,7 @@ static int pagefault_mr(struct mlx5_ib_mr *mr, u64 io_virt, size_t bcnt,
 {
 	struct ib_umem_odp *odp = to_ib_umem_odp(mr->umem);
 
-	if (unlikely(io_virt < mr->mmkey.iova))
+	if (unlikely(io_virt < mr->ibmr.iova))
 		return -EFAULT;
 
 	if (mr->umem->is_dmabuf)
@@ -747,7 +747,7 @@ static int pagefault_mr(struct mlx5_ib_mr *mr, u64 io_virt, size_t bcnt,
 	if (!odp->is_implicit_odp) {
 		u64 user_va;
 
-		if (check_add_overflow(io_virt - mr->mmkey.iova,
+		if (check_add_overflow(io_virt - mr->ibmr.iova,
 				       (u64)odp->umem.address, &user_va))
 			return -EFAULT;
 		if (unlikely(user_va >= ib_umem_end(odp) ||
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mr.c b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
index 174f71ed5280..d239d559994f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mr.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
@@ -52,7 +52,6 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
-	mkey->iova = MLX5_GET64(mkc, mkc, start_addr);
 	mkey->size = MLX5_GET64(mkc, mkc, len);
 	mkey->key = (u32)mlx5_mkey_variant(mkey->key) | mlx5_idx_to_mkey(mkey_index);
 	mkey->pd = MLX5_GET(mkc, mkc, pd);
diff --git a/drivers/vdpa/mlx5/core/resources.c b/drivers/vdpa/mlx5/core/resources.c
index 15e266d0e27a..14d4314cdc29 100644
--- a/drivers/vdpa/mlx5/core/resources.c
+++ b/drivers/vdpa/mlx5/core/resources.c
@@ -215,7 +215,6 @@ int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mk
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
-	mkey->iova = MLX5_GET64(mkc, mkc, start_addr);
 	mkey->size = MLX5_GET64(mkc, mkc, len);
 	mkey->key |= mlx5_idx_to_mkey(mkey_index);
 	mkey->pd = MLX5_GET(mkc, mkc, pd);
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 1b8bae246b28..573c0cbb0879 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -364,7 +364,6 @@ enum {
 };
 
 struct mlx5_core_mkey {
-	u64			iova;
 	u64			size;
 	u32			key;
 	u32			pd;
-- 
2.31.1

[PATCH mlx5-next 3/7] RDMA/mlx5: Remove size from struct mlx5_core_mkey

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

mkey->size is already stored in ibmr->length, no need to store it here.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/devx.c            | 1 -
 drivers/infiniband/hw/mlx5/mr.c              | 5 ++---
 drivers/net/ethernet/mellanox/mlx5/core/mr.c | 1 -
 drivers/vdpa/mlx5/core/resources.c           | 1 -
 include/linux/mlx5/driver.h                  | 1 -
 5 files changed, 2 insertions(+), 7 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
index 439b62e23afb..3d416850bba8 100644
--- a/drivers/infiniband/hw/mlx5/devx.c
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -1303,7 +1303,6 @@ static int devx_handle_mkey_indirect(struct devx_obj *obj,
 	mkey->key = mlx5_idx_to_mkey(
 			MLX5_GET(create_mkey_out, out, mkey_index)) | key;
 	mkey->type = MLX5_MKEY_INDIRECT_DEVX;
-	mkey->size = MLX5_GET64(mkc, mkc, len);
 	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	devx_mr->ndescs = MLX5_GET(mkc, mkc, translations_octword_size);
 	init_waitqueue_head(&mkey->wait);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index e264c6be38f8..6f831598b987 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -968,7 +968,6 @@ static struct mlx5_ib_mr *alloc_cacheable_mr(struct ib_pd *pd,
 
 	mr->ibmr.pd = pd;
 	mr->umem = umem;
-	mr->mmkey.size = umem->length;
 	mr->mmkey.pd = to_mpd(pd)->pdn;
 	mr->page_shift = order_base_2(page_size);
 	set_mr_fields(dev, mr, umem->length, access_flags, iova);
@@ -1080,7 +1079,7 @@ static void *mlx5_ib_create_xlt_wr(struct mlx5_ib_mr *mr,
 	wr->wr.opcode = MLX5_IB_WR_UMR;
 	wr->pd = mr->ibmr.pd;
 	wr->mkey = mr->mmkey.key;
-	wr->length = mr->mmkey.size;
+	wr->length = mr->ibmr.length;
 	wr->virt_addr = mr->ibmr.iova;
 	wr->access_flags = mr->access_flags;
 	wr->page_shift = mr->page_shift;
@@ -1768,7 +1767,7 @@ static int umr_rereg_pas(struct mlx5_ib_mr *mr, struct ib_pd *pd,
 
 	mr->ibmr.length = new_umem->length;
 	mr->ibmr.iova = iova;
-	mr->mmkey.size = new_umem->length;
+	mr->ibmr.length = new_umem->length;
 	mr->page_shift = order_base_2(page_size);
 	mr->umem = new_umem;
 	err = mlx5_ib_update_mr_pas(mr, upd_flags);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mr.c b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
index d239d559994f..b5dd44944265 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mr.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
@@ -52,7 +52,6 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
-	mkey->size = MLX5_GET64(mkc, mkc, len);
 	mkey->key = (u32)mlx5_mkey_variant(mkey->key) | mlx5_idx_to_mkey(mkey_index);
 	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	init_waitqueue_head(&mkey->wait);
diff --git a/drivers/vdpa/mlx5/core/resources.c b/drivers/vdpa/mlx5/core/resources.c
index 14d4314cdc29..d3d8b8b4e377 100644
--- a/drivers/vdpa/mlx5/core/resources.c
+++ b/drivers/vdpa/mlx5/core/resources.c
@@ -215,7 +215,6 @@ int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mk
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
-	mkey->size = MLX5_GET64(mkc, mkc, len);
 	mkey->key |= mlx5_idx_to_mkey(mkey_index);
 	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	return 0;
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 573c0cbb0879..cba157689ca8 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -364,7 +364,6 @@ enum {
 };
 
 struct mlx5_core_mkey {
-	u64			size;
 	u32			key;
 	u32			pd;
 	u32			type;
-- 
2.31.1

[PATCH mlx5-next 4/7] RDMA/mlx5: Remove pd from struct mlx5_core_mkey

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

There is no read of mkey->pd, only write. Remove it.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/devx.c            | 1 -
 drivers/infiniband/hw/mlx5/mr.c              | 3 ---
 drivers/net/ethernet/mellanox/mlx5/core/mr.c | 3 ---
 drivers/vdpa/mlx5/core/resources.c           | 3 ---
 include/linux/mlx5/driver.h                  | 1 -
 5 files changed, 11 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
index 3d416850bba8..465ea835f854 100644
--- a/drivers/infiniband/hw/mlx5/devx.c
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -1303,7 +1303,6 @@ static int devx_handle_mkey_indirect(struct devx_obj *obj,
 	mkey->key = mlx5_idx_to_mkey(
 			MLX5_GET(create_mkey_out, out, mkey_index)) | key;
 	mkey->type = MLX5_MKEY_INDIRECT_DEVX;
-	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	devx_mr->ndescs = MLX5_GET(mkc, mkc, translations_octword_size);
 	init_waitqueue_head(&mkey->wait);
 
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 6f831598b987..1450c2f0da00 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -968,7 +968,6 @@ static struct mlx5_ib_mr *alloc_cacheable_mr(struct ib_pd *pd,
 
 	mr->ibmr.pd = pd;
 	mr->umem = umem;
-	mr->mmkey.pd = to_mpd(pd)->pdn;
 	mr->page_shift = order_base_2(page_size);
 	set_mr_fields(dev, mr, umem->length, access_flags, iova);
 
@@ -1712,7 +1711,6 @@ static int umr_rereg_pd_access(struct mlx5_ib_mr *mr, struct ib_pd *pd,
 		return err;
 
 	mr->access_flags = access_flags;
-	mr->mmkey.pd = to_mpd(pd)->pdn;
 	return 0;
 }
 
@@ -1757,7 +1755,6 @@ static int umr_rereg_pas(struct mlx5_ib_mr *mr, struct ib_pd *pd,
 
 	if (flags & IB_MR_REREG_PD) {
 		mr->ibmr.pd = pd;
-		mr->mmkey.pd = to_mpd(pd)->pdn;
 		upd_flags |= MLX5_IB_UPD_XLT_PD;
 	}
 	if (flags & IB_MR_REREG_ACCESS) {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mr.c b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
index b5dd44944265..6e99fd166f98 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mr.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
@@ -41,7 +41,6 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 {
 	u32 lout[MLX5_ST_SZ_DW(create_mkey_out)] = {};
 	u32 mkey_index;
-	void *mkc;
 	int err;
 
 	MLX5_SET(create_mkey_in, in, opcode, MLX5_CMD_OP_CREATE_MKEY);
@@ -50,10 +49,8 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 	if (err)
 		return err;
 
-	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
 	mkey->key = (u32)mlx5_mkey_variant(mkey->key) | mlx5_idx_to_mkey(mkey_index);
-	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	init_waitqueue_head(&mkey->wait);
 
 	mlx5_core_dbg(dev, "out 0x%x, mkey 0x%x\n", mkey_index, mkey->key);
diff --git a/drivers/vdpa/mlx5/core/resources.c b/drivers/vdpa/mlx5/core/resources.c
index d3d8b8b4e377..72b2d80e75b0 100644
--- a/drivers/vdpa/mlx5/core/resources.c
+++ b/drivers/vdpa/mlx5/core/resources.c
@@ -203,7 +203,6 @@ int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mk
 {
 	u32 lout[MLX5_ST_SZ_DW(create_mkey_out)] = {};
 	u32 mkey_index;
-	void *mkc;
 	int err;
 
 	MLX5_SET(create_mkey_in, in, opcode, MLX5_CMD_OP_CREATE_MKEY);
@@ -213,10 +212,8 @@ int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mk
 	if (err)
 		return err;
 
-	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
 	mkey->key |= mlx5_idx_to_mkey(mkey_index);
-	mkey->pd = MLX5_GET(mkc, mkc, pd);
 	return 0;
 }
 
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index cba157689ca8..31e19f06bc3a 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -365,7 +365,6 @@ enum {
 
 struct mlx5_core_mkey {
 	u32			key;
-	u32			pd;
 	u32			type;
 	struct wait_queue_head wait;
 	refcount_t usecount;
-- 
2.31.1

[PATCH mlx5-next 5/7] RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

In mlx5_core and vdpa there is no use of mlx5_core_mkey members except
for the key itself.

As preparation for moving mlx5_core_mkey to mlx5_ib, the occurrences of
struct mlx5_core_mkey in all modules except for mlx5_ib are replaced by
a u32 key.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/mr.c               | 21 ++++++++++++------
 drivers/infiniband/hw/mlx5/odp.c              |  2 +-
 .../mellanox/mlx5/core/diag/fw_tracer.c       |  6 ++---
 .../mellanox/mlx5/core/diag/fw_tracer.h       |  2 +-
 .../mellanox/mlx5/core/diag/rsc_dump.c        | 10 ++++-----
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en/trap.c |  2 +-
 .../ethernet/mellanox/mlx5/core/en_common.c   |  6 ++---
 .../net/ethernet/mellanox/mlx5/core/en_main.c | 13 +++++------
 .../ethernet/mellanox/mlx5/core/fpga/conn.c   | 10 ++++-----
 .../ethernet/mellanox/mlx5/core/fpga/core.h   |  2 +-
 drivers/net/ethernet/mellanox/mlx5/core/mr.c  | 22 +++++++++----------
 .../mellanox/mlx5/core/steering/dr_icm_pool.c | 10 ++++-----
 .../mellanox/mlx5/core/steering/dr_send.c     | 11 +++++-----
 .../mellanox/mlx5/core/steering/dr_types.h    |  2 +-
 drivers/vdpa/mlx5/core/mlx5_vdpa.h            |  8 +++----
 drivers/vdpa/mlx5/core/mr.c                   |  8 +++----
 drivers/vdpa/mlx5/core/resources.c            |  8 +++----
 drivers/vdpa/mlx5/net/mlx5_vnet.c             |  2 +-
 include/linux/mlx5/driver.h                   | 14 +++++-------
 21 files changed, 82 insertions(+), 81 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 1450c2f0da00..2260b0298965 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -105,8 +105,14 @@ static int
 mlx5_ib_create_mkey(struct mlx5_ib_dev *dev, struct mlx5_core_mkey *mkey,
 		    u32 *in, int inlen)
 {
+	int ret;
+
 	assign_mkey_variant(dev, mkey, in);
-	return mlx5_core_create_mkey(dev->mdev, mkey, in, inlen);
+	ret = mlx5_core_create_mkey(dev->mdev, &mkey->key, in, inlen);
+	if (!ret)
+		init_waitqueue_head(&mkey->wait);
+
+	return ret;
 }
 
 static int
@@ -134,7 +140,7 @@ static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
 	WARN_ON(xa_load(&dev->odp_mkeys, mlx5_base_mkey(mr->mmkey.key)));
 
-	return mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+	return mlx5_core_destroy_mkey(dev->mdev, mr->mmkey.key);
 }
 
 static void create_mkey_callback(int status, struct mlx5_async_work *context)
@@ -261,10 +267,11 @@ static struct mlx5_ib_mr *create_cache_mr(struct mlx5_cache_ent *ent)
 		goto free_in;
 	}
 
-	err = mlx5_core_create_mkey(ent->dev->mdev, &mr->mmkey, in, inlen);
+	err = mlx5_core_create_mkey(ent->dev->mdev, &mr->mmkey.key, in, inlen);
 	if (err)
 		goto free_mr;
 
+	init_waitqueue_head(&mr->mmkey.wait);
 	mr->mmkey.type = MLX5_MKEY_MR;
 	WRITE_ONCE(ent->dev->cache.last_add, jiffies);
 	spin_lock_irq(&ent->lock);
@@ -291,7 +298,7 @@ static void remove_cache_mr_locked(struct mlx5_cache_ent *ent)
 	ent->available_mrs--;
 	ent->total_mrs--;
 	spin_unlock_irq(&ent->lock);
-	mlx5_core_destroy_mkey(ent->dev->mdev, &mr->mmkey);
+	mlx5_core_destroy_mkey(ent->dev->mdev, mr->mmkey.key);
 	kfree(mr);
 	spin_lock_irq(&ent->lock);
 }
@@ -651,7 +658,7 @@ static void clean_keys(struct mlx5_ib_dev *dev, int c)
 		ent->available_mrs--;
 		ent->total_mrs--;
 		spin_unlock_irq(&ent->lock);
-		mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
+		mlx5_core_destroy_mkey(dev->mdev, mr->mmkey.key);
 	}
 
 	list_for_each_entry_safe(mr, tmp_mr, &del_list, list) {
@@ -2350,7 +2357,7 @@ int mlx5_ib_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
 	return 0;
 
 free_mkey:
-	mlx5_core_destroy_mkey(dev->mdev, &mw->mmkey);
+	mlx5_core_destroy_mkey(dev->mdev, mw->mmkey.key);
 free:
 	kfree(in);
 	return err;
@@ -2369,7 +2376,7 @@ int mlx5_ib_dealloc_mw(struct ib_mw *mw)
 		 */
 		mlx5r_deref_wait_odp_mkey(&mmw->mmkey);
 
-	return mlx5_core_destroy_mkey(dev->mdev, &mmw->mmkey);
+	return mlx5_core_destroy_mkey(dev->mdev, mmw->mmkey.key);
 }
 
 int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 3b24cfc96d36..8ffd3ea8db7c 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -909,7 +909,7 @@ static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
 		pklm = (struct mlx5_klm *)MLX5_ADDR_OF(query_mkey_out, out,
 						       bsf0_klm0_pas_mtt0_1);
 
-		ret = mlx5_core_query_mkey(dev->mdev, mmkey, out, outlen);
+		ret = mlx5_core_query_mkey(dev->mdev, mmkey->key, out, outlen);
 		if (ret)
 			goto end;
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.c b/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.c
index f9cf9fb31547..02db3148240a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.c
@@ -745,7 +745,7 @@ static int mlx5_fw_tracer_set_mtrc_conf(struct mlx5_fw_tracer *tracer)
 	MLX5_SET(mtrc_conf, in, trace_mode, TRACE_TO_MEMORY);
 	MLX5_SET(mtrc_conf, in, log_trace_buffer_size,
 		 ilog2(TRACER_BUFFER_PAGE_NUM));
-	MLX5_SET(mtrc_conf, in, trace_mkey, tracer->buff.mkey.key);
+	MLX5_SET(mtrc_conf, in, trace_mkey, tracer->buff.mkey);
 
 	err = mlx5_core_access_reg(dev, in, sizeof(in), out, sizeof(out),
 				   MLX5_REG_MTRC_CONF, 0, 1);
@@ -1028,7 +1028,7 @@ int mlx5_fw_tracer_init(struct mlx5_fw_tracer *tracer)
 
 err_notifier_unregister:
 	mlx5_eq_notifier_unregister(dev, &tracer->nb);
-	mlx5_core_destroy_mkey(dev, &tracer->buff.mkey);
+	mlx5_core_destroy_mkey(dev, tracer->buff.mkey);
 err_dealloc_pd:
 	mlx5_core_dealloc_pd(dev, tracer->buff.pdn);
 err_cancel_work:
@@ -1051,7 +1051,7 @@ void mlx5_fw_tracer_cleanup(struct mlx5_fw_tracer *tracer)
 	if (tracer->owner)
 		mlx5_fw_tracer_ownership_release(tracer);
 
-	mlx5_core_destroy_mkey(tracer->dev, &tracer->buff.mkey);
+	mlx5_core_destroy_mkey(tracer->dev, tracer->buff.mkey);
 	mlx5_core_dealloc_pd(tracer->dev, tracer->buff.pdn);
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.h b/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.h
index 97252a85d65e..4762b55b0b0e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.h
@@ -89,7 +89,7 @@ struct mlx5_fw_tracer {
 		void *log_buf;
 		dma_addr_t dma;
 		u32 size;
-		struct mlx5_core_mkey mkey;
+		u32 mkey;
 		u32 consumer_index;
 	} buff;
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/diag/rsc_dump.c b/drivers/net/ethernet/mellanox/mlx5/core/diag/rsc_dump.c
index ed4fb79b4db7..538adab6878b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/diag/rsc_dump.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/diag/rsc_dump.c
@@ -30,7 +30,7 @@ static const char *const mlx5_rsc_sgmt_name[] = {
 
 struct mlx5_rsc_dump {
 	u32 pdn;
-	struct mlx5_core_mkey mkey;
+	u32 mkey;
 	u16 fw_segment_type[MLX5_SGMT_TYPE_NUM];
 };
 
@@ -89,7 +89,7 @@ static int mlx5_rsc_dump_trigger(struct mlx5_core_dev *dev, struct mlx5_rsc_dump
 		return -ENOMEM;
 
 	in_seq_num = MLX5_GET(resource_dump, cmd->cmd, seq_num);
-	MLX5_SET(resource_dump, cmd->cmd, mkey, rsc_dump->mkey.key);
+	MLX5_SET(resource_dump, cmd->cmd, mkey, rsc_dump->mkey);
 	MLX5_SET64(resource_dump, cmd->cmd, address, dma);
 
 	err = mlx5_core_access_reg(dev, cmd->cmd, sizeof(cmd->cmd), cmd->cmd,
@@ -202,7 +202,7 @@ static int mlx5_rsc_dump_menu(struct mlx5_core_dev *dev)
 }
 
 static int mlx5_rsc_dump_create_mkey(struct mlx5_core_dev *mdev, u32 pdn,
-				     struct mlx5_core_mkey *mkey)
+				     u32 *mkey)
 {
 	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
 	void *mkc;
@@ -276,7 +276,7 @@ int mlx5_rsc_dump_init(struct mlx5_core_dev *dev)
 	return err;
 
 destroy_mkey:
-	mlx5_core_destroy_mkey(dev, &rsc_dump->mkey);
+	mlx5_core_destroy_mkey(dev, rsc_dump->mkey);
 free_pd:
 	mlx5_core_dealloc_pd(dev, rsc_dump->pdn);
 	return err;
@@ -287,6 +287,6 @@ void mlx5_rsc_dump_cleanup(struct mlx5_core_dev *dev)
 	if (IS_ERR_OR_NULL(dev->rsc_dump))
 		return;
 
-	mlx5_core_destroy_mkey(dev, &dev->rsc_dump->mkey);
+	mlx5_core_destroy_mkey(dev, dev->rsc_dump->mkey);
 	mlx5_core_dealloc_pd(dev, dev->rsc_dump->pdn);
 }
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 7b8c8187543a..21244a4ec4b0 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -664,7 +664,7 @@ struct mlx5e_rq {
 	u8                     wq_type;
 	u32                    rqn;
 	struct mlx5_core_dev  *mdev;
-	struct mlx5_core_mkey  umr_mkey;
+	u32  umr_mkey;
 	struct mlx5e_dma_info  wqe_overflow;
 
 	/* XDP read-mostly */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c
index ee688dec67a9..6ebca117e5ff 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/ptp.c
@@ -683,7 +683,7 @@ int mlx5e_ptp_open(struct mlx5e_priv *priv, struct mlx5e_params *params,
 	c->tstamp   = &priv->tstamp;
 	c->pdev     = mlx5_core_dma_dev(priv->mdev);
 	c->netdev   = priv->netdev;
-	c->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey.key);
+	c->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey);
 	c->num_tc   = mlx5e_get_dcb_num_tc(params);
 	c->stats    = &priv->ptp_stats.ch;
 	c->lag_port = lag_port;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/trap.c b/drivers/net/ethernet/mellanox/mlx5/core/en/trap.c
index d54607a42740..a55b066746cb 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/trap.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/trap.c
@@ -137,7 +137,7 @@ static struct mlx5e_trap *mlx5e_open_trap(struct mlx5e_priv *priv)
 	t->tstamp   = &priv->tstamp;
 	t->pdev     = mlx5_core_dma_dev(priv->mdev);
 	t->netdev   = priv->netdev;
-	t->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey.key);
+	t->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey);
 	t->stats    = &priv->trap_stats.ch;
 
 	netif_napi_add(netdev, &t->napi, mlx5e_trap_napi_poll, 64);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_common.c b/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
index 84eb7201c142..c0f409c195bf 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
@@ -47,7 +47,7 @@ void mlx5e_mkey_set_relaxed_ordering(struct mlx5_core_dev *mdev, void *mkc)
 }
 
 static int mlx5e_create_mkey(struct mlx5_core_dev *mdev, u32 pdn,
-			     struct mlx5_core_mkey *mkey)
+			     u32 *mkey)
 {
 	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
 	void *mkc;
@@ -108,7 +108,7 @@ int mlx5e_create_mdev_resources(struct mlx5_core_dev *mdev)
 	return 0;
 
 err_destroy_mkey:
-	mlx5_core_destroy_mkey(mdev, &res->mkey);
+	mlx5_core_destroy_mkey(mdev, res->mkey);
 err_dealloc_transport_domain:
 	mlx5_core_dealloc_transport_domain(mdev, res->td.tdn);
 err_dealloc_pd:
@@ -121,7 +121,7 @@ void mlx5e_destroy_mdev_resources(struct mlx5_core_dev *mdev)
 	struct mlx5e_hw_objs *res = &mdev->mlx5e_res.hw_objs;
 
 	mlx5_free_bfreg(mdev, &res->bfreg);
-	mlx5_core_destroy_mkey(mdev, &res->mkey);
+	mlx5_core_destroy_mkey(mdev, res->mkey);
 	mlx5_core_dealloc_transport_domain(mdev, res->td.tdn);
 	mlx5_core_dealloc_pd(mdev, res->pdn);
 	memset(res, 0, sizeof(*res));
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 3fd515e7bf30..ea5b5b40e6cf 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -234,8 +234,7 @@ static int mlx5e_rq_alloc_mpwqe_info(struct mlx5e_rq *rq, int node)
 }
 
 static int mlx5e_create_umr_mkey(struct mlx5_core_dev *mdev,
-				 u64 npages, u8 page_shift,
-				 struct mlx5_core_mkey *umr_mkey,
+				 u64 npages, u8 page_shift, u32 *umr_mkey,
 				 dma_addr_t filler_addr)
 {
 	struct mlx5_mtt *mtt;
@@ -455,7 +454,7 @@ static int mlx5e_alloc_rq(struct mlx5e_params *params,
 		err = mlx5e_create_rq_umr_mkey(mdev, rq);
 		if (err)
 			goto err_rq_drop_page;
-		rq->mkey_be = cpu_to_be32(rq->umr_mkey.key);
+		rq->mkey_be = cpu_to_be32(rq->umr_mkey);
 
 		err = mlx5e_rq_alloc_mpwqe_info(rq, node);
 		if (err)
@@ -487,7 +486,7 @@ static int mlx5e_alloc_rq(struct mlx5e_params *params,
 		if (err)
 			goto err_rq_frags;
 
-		rq->mkey_be = cpu_to_be32(mdev->mlx5e_res.hw_objs.mkey.key);
+		rq->mkey_be = cpu_to_be32(mdev->mlx5e_res.hw_objs.mkey);
 	}
 
 	if (xsk) {
@@ -574,7 +573,7 @@ static int mlx5e_alloc_rq(struct mlx5e_params *params,
 	case MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ:
 		kvfree(rq->mpwqe.info);
 err_rq_mkey:
-		mlx5_core_destroy_mkey(mdev, &rq->umr_mkey);
+		mlx5_core_destroy_mkey(mdev, rq->umr_mkey);
 err_rq_drop_page:
 		mlx5e_free_mpwqe_rq_drop_page(rq);
 		break;
@@ -607,7 +606,7 @@ static void mlx5e_free_rq(struct mlx5e_rq *rq)
 	switch (rq->wq_type) {
 	case MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ:
 		kvfree(rq->mpwqe.info);
-		mlx5_core_destroy_mkey(rq->mdev, &rq->umr_mkey);
+		mlx5_core_destroy_mkey(rq->mdev, rq->umr_mkey);
 		mlx5e_free_mpwqe_rq_drop_page(rq);
 		break;
 	default: /* MLX5_WQ_TYPE_CYCLIC */
@@ -1991,7 +1990,7 @@ static int mlx5e_open_channel(struct mlx5e_priv *priv, int ix,
 	c->cpu      = cpu;
 	c->pdev     = mlx5_core_dma_dev(priv->mdev);
 	c->netdev   = priv->netdev;
-	c->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey.key);
+	c->mkey_be  = cpu_to_be32(priv->mdev->mlx5e_res.hw_objs.mkey);
 	c->num_tc   = mlx5e_get_dcb_num_tc(params);
 	c->xdp      = !!params->xdp_prog;
 	c->stats    = &priv->channel_stats[ix].ch;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c b/drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c
index 306279b7f9e7..12abe991583a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fpga/conn.c
@@ -115,7 +115,7 @@ static int mlx5_fpga_conn_post_recv(struct mlx5_fpga_conn *conn,
 	ix = conn->qp.rq.pc & (conn->qp.rq.size - 1);
 	data = mlx5_wq_cyc_get_wqe(&conn->qp.wq.rq, ix);
 	data->byte_count = cpu_to_be32(buf->sg[0].size);
-	data->lkey = cpu_to_be32(conn->fdev->conn_res.mkey.key);
+	data->lkey = cpu_to_be32(conn->fdev->conn_res.mkey);
 	data->addr = cpu_to_be64(buf->sg[0].dma_addr);
 
 	conn->qp.rq.pc++;
@@ -155,7 +155,7 @@ static void mlx5_fpga_conn_post_send(struct mlx5_fpga_conn *conn,
 		if (!buf->sg[sgi].data)
 			break;
 		data->byte_count = cpu_to_be32(buf->sg[sgi].size);
-		data->lkey = cpu_to_be32(conn->fdev->conn_res.mkey.key);
+		data->lkey = cpu_to_be32(conn->fdev->conn_res.mkey);
 		data->addr = cpu_to_be64(buf->sg[sgi].dma_addr);
 		data++;
 		size++;
@@ -221,7 +221,7 @@ static int mlx5_fpga_conn_post_recv_buf(struct mlx5_fpga_conn *conn)
 }
 
 static int mlx5_fpga_conn_create_mkey(struct mlx5_core_dev *mdev, u32 pdn,
-				      struct mlx5_core_mkey *mkey)
+				      u32 *mkey)
 {
 	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
 	void *mkc;
@@ -978,7 +978,7 @@ int mlx5_fpga_conn_device_init(struct mlx5_fpga_device *fdev)
 		mlx5_fpga_err(fdev, "create mkey failed, %d\n", err);
 		goto err_dealloc_pd;
 	}
-	mlx5_fpga_dbg(fdev, "Created mkey 0x%x\n", fdev->conn_res.mkey.key);
+	mlx5_fpga_dbg(fdev, "Created mkey 0x%x\n", fdev->conn_res.mkey);
 
 	return 0;
 
@@ -994,7 +994,7 @@ int mlx5_fpga_conn_device_init(struct mlx5_fpga_device *fdev)
 
 void mlx5_fpga_conn_device_cleanup(struct mlx5_fpga_device *fdev)
 {
-	mlx5_core_destroy_mkey(fdev->mdev, &fdev->conn_res.mkey);
+	mlx5_core_destroy_mkey(fdev->mdev, fdev->conn_res.mkey);
 	mlx5_core_dealloc_pd(fdev->mdev, fdev->conn_res.pdn);
 	mlx5_put_uars_page(fdev->mdev, fdev->conn_res.uar);
 	mlx5_nic_vport_disable_roce(fdev->mdev);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fpga/core.h b/drivers/net/ethernet/mellanox/mlx5/core/fpga/core.h
index 52c9dee91ea4..2a984e82ae16 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fpga/core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fpga/core.h
@@ -54,7 +54,7 @@ struct mlx5_fpga_device {
 	/* QP Connection resources */
 	struct {
 		u32 pdn;
-		struct mlx5_core_mkey mkey;
+		u32 mkey;
 		struct mlx5_uars_page *uar;
 	} conn_res;
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mr.c b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
index 6e99fd166f98..f099a087400e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mr.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mr.c
@@ -35,9 +35,8 @@
 #include <linux/mlx5/driver.h>
 #include "mlx5_core.h"
 
-int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
-			  struct mlx5_core_mkey *mkey,
-			  u32 *in, int inlen)
+int mlx5_core_create_mkey(struct mlx5_core_dev *dev, u32 *mkey, u32 *in,
+			  int inlen)
 {
 	u32 lout[MLX5_ST_SZ_DW(create_mkey_out)] = {};
 	u32 mkey_index;
@@ -50,33 +49,32 @@ int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
 		return err;
 
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
-	mkey->key = (u32)mlx5_mkey_variant(mkey->key) | mlx5_idx_to_mkey(mkey_index);
-	init_waitqueue_head(&mkey->wait);
+	*mkey = MLX5_GET(create_mkey_in, in, memory_key_mkey_entry.mkey_7_0) |
+		mlx5_idx_to_mkey(mkey_index);
 
-	mlx5_core_dbg(dev, "out 0x%x, mkey 0x%x\n", mkey_index, mkey->key);
+	mlx5_core_dbg(dev, "out 0x%x, mkey 0x%x\n", mkey_index, *mkey);
 	return 0;
 }
 EXPORT_SYMBOL(mlx5_core_create_mkey);
 
-int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev,
-			   struct mlx5_core_mkey *mkey)
+int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, u32 mkey)
 {
 	u32 in[MLX5_ST_SZ_DW(destroy_mkey_in)] = {};
 
 	MLX5_SET(destroy_mkey_in, in, opcode, MLX5_CMD_OP_DESTROY_MKEY);
-	MLX5_SET(destroy_mkey_in, in, mkey_index, mlx5_mkey_to_idx(mkey->key));
+	MLX5_SET(destroy_mkey_in, in, mkey_index, mlx5_mkey_to_idx(mkey));
 	return mlx5_cmd_exec_in(dev, destroy_mkey, in);
 }
 EXPORT_SYMBOL(mlx5_core_destroy_mkey);
 
-int mlx5_core_query_mkey(struct mlx5_core_dev *dev, struct mlx5_core_mkey *mkey,
-			 u32 *out, int outlen)
+int mlx5_core_query_mkey(struct mlx5_core_dev *dev, u32 mkey, u32 *out,
+			 int outlen)
 {
 	u32 in[MLX5_ST_SZ_DW(query_mkey_in)] = {};
 
 	memset(out, 0, outlen);
 	MLX5_SET(query_mkey_in, in, opcode, MLX5_CMD_OP_QUERY_MKEY);
-	MLX5_SET(query_mkey_in, in, mkey_index, mlx5_mkey_to_idx(mkey->key));
+	MLX5_SET(query_mkey_in, in, mkey_index, mlx5_mkey_to_idx(mkey));
 	return mlx5_cmd_exec(dev, in, sizeof(in), out, outlen);
 }
 EXPORT_SYMBOL(mlx5_core_query_mkey);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_icm_pool.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_icm_pool.c
index 66c24767e3b0..7f6fd9c5e371 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_icm_pool.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_icm_pool.c
@@ -24,7 +24,7 @@ struct mlx5dr_icm_dm {
 };
 
 struct mlx5dr_icm_mr {
-	struct mlx5_core_mkey mkey;
+	u32 mkey;
 	struct mlx5dr_icm_dm dm;
 	struct mlx5dr_domain *dmn;
 	size_t length;
@@ -33,7 +33,7 @@ struct mlx5dr_icm_mr {
 
 static int dr_icm_create_dm_mkey(struct mlx5_core_dev *mdev,
 				 u32 pd, u64 length, u64 start_addr, int mode,
-				 struct mlx5_core_mkey *mkey)
+				 u32 *mkey)
 {
 	u32 inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
 	u32 in[MLX5_ST_SZ_DW(create_mkey_in)] = {};
@@ -116,7 +116,7 @@ dr_icm_pool_mr_create(struct mlx5dr_icm_pool *pool)
 	return icm_mr;
 
 free_mkey:
-	mlx5_core_destroy_mkey(mdev, &icm_mr->mkey);
+	mlx5_core_destroy_mkey(mdev, icm_mr->mkey);
 free_dm:
 	mlx5_dm_sw_icm_dealloc(mdev, icm_mr->dm.type, icm_mr->dm.length, 0,
 			       icm_mr->dm.addr, icm_mr->dm.obj_id);
@@ -130,7 +130,7 @@ static void dr_icm_pool_mr_destroy(struct mlx5dr_icm_mr *icm_mr)
 	struct mlx5_core_dev *mdev = icm_mr->dmn->mdev;
 	struct mlx5dr_icm_dm *dm = &icm_mr->dm;
 
-	mlx5_core_destroy_mkey(mdev, &icm_mr->mkey);
+	mlx5_core_destroy_mkey(mdev, icm_mr->mkey);
 	mlx5_dm_sw_icm_dealloc(mdev, dm->type, dm->length, 0,
 			       dm->addr, dm->obj_id);
 	kvfree(icm_mr);
@@ -252,7 +252,7 @@ dr_icm_chunk_create(struct mlx5dr_icm_pool *pool,
 
 	offset = mlx5dr_icm_pool_dm_type_to_entry_size(pool->icm_type) * seg;
 
-	chunk->rkey = buddy_mem_pool->icm_mr->mkey.key;
+	chunk->rkey = buddy_mem_pool->icm_mr->mkey;
 	chunk->mr_addr = offset;
 	chunk->icm_addr =
 		(uintptr_t)buddy_mem_pool->icm_mr->icm_start_addr + offset;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c
index bfb14b4b1906..00aef47d7682 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_send.c
@@ -350,7 +350,7 @@ static void dr_fill_data_segs(struct mlx5dr_send_ring *send_ring,
 	send_info->read.length = send_info->write.length;
 	/* Read into the same write area */
 	send_info->read.addr = (uintptr_t)send_info->write.addr;
-	send_info->read.lkey = send_ring->mr->mkey.key;
+	send_info->read.lkey = send_ring->mr->mkey;
 
 	if (send_ring->pending_wqe % send_ring->signal_th == 0)
 		send_info->read.send_flags = IB_SEND_SIGNALED;
@@ -388,7 +388,7 @@ static int dr_postsend_icm_data(struct mlx5dr_domain *dmn,
 		       (void *)(uintptr_t)send_info->write.addr,
 		       send_info->write.length);
 		send_info->write.addr = (uintptr_t)send_ring->mr->dma_addr + buff_offset;
-		send_info->write.lkey = send_ring->mr->mkey.key;
+		send_info->write.lkey = send_ring->mr->mkey;
 	}
 
 	send_ring->tx_head++;
@@ -848,8 +848,7 @@ static void dr_destroy_cq(struct mlx5_core_dev *mdev, struct mlx5dr_cq *cq)
 	kfree(cq);
 }
 
-static int
-dr_create_mkey(struct mlx5_core_dev *mdev, u32 pdn, struct mlx5_core_mkey *mkey)
+static int dr_create_mkey(struct mlx5_core_dev *mdev, u32 pdn, u32 *mkey)
 {
 	u32 in[MLX5_ST_SZ_DW(create_mkey_in)] = {};
 	void *mkc;
@@ -908,7 +907,7 @@ static struct mlx5dr_mr *dr_reg_mr(struct mlx5_core_dev *mdev,
 
 static void dr_dereg_mr(struct mlx5_core_dev *mdev, struct mlx5dr_mr *mr)
 {
-	mlx5_core_destroy_mkey(mdev, &mr->mkey);
+	mlx5_core_destroy_mkey(mdev, mr->mkey);
 	dma_unmap_single(mlx5_core_dma_dev(mdev), mr->dma_addr, mr->size,
 			 DMA_BIDIRECTIONAL);
 	kfree(mr);
@@ -1039,7 +1038,7 @@ int mlx5dr_send_ring_force_drain(struct mlx5dr_domain *dmn)
 	send_info.write.lkey = 0;
 	/* Using the sync_mr in order to write/read */
 	send_info.remote_addr = (uintptr_t)send_ring->sync_mr->addr;
-	send_info.rkey = send_ring->sync_mr->mkey.key;
+	send_info.rkey = send_ring->sync_mr->mkey;
 
 	for (i = 0; i < num_of_sends_req; i++) {
 		ret = dr_postsend_icm_data(dmn, &send_info);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_types.h b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_types.h
index b20e8aabb861..6c67185c05c1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_types.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/dr_types.h
@@ -1275,7 +1275,7 @@ struct mlx5dr_cq {
 
 struct mlx5dr_mr {
 	struct mlx5_core_dev *mdev;
-	struct mlx5_core_mkey mkey;
+	u32 mkey;
 	dma_addr_t dma_addr;
 	void *addr;
 	size_t size;
diff --git a/drivers/vdpa/mlx5/core/mlx5_vdpa.h b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
index 01a848adf590..3163b313a470 100644
--- a/drivers/vdpa/mlx5/core/mlx5_vdpa.h
+++ b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
@@ -15,7 +15,7 @@ struct mlx5_vdpa_direct_mr {
 	u64 start;
 	u64 end;
 	u32 perm;
-	struct mlx5_core_mkey mr;
+	u32 mr;
 	struct sg_table sg_head;
 	int log_size;
 	int nsg;
@@ -25,7 +25,7 @@ struct mlx5_vdpa_direct_mr {
 };
 
 struct mlx5_vdpa_mr {
-	struct mlx5_core_mkey mkey;
+	u32 mkey;
 
 	/* list of direct MRs descendants of this indirect mr */
 	struct list_head head;
@@ -99,9 +99,9 @@ int mlx5_vdpa_alloc_transport_domain(struct mlx5_vdpa_dev *mvdev, u32 *tdn);
 void mlx5_vdpa_dealloc_transport_domain(struct mlx5_vdpa_dev *mvdev, u32 tdn);
 int mlx5_vdpa_alloc_resources(struct mlx5_vdpa_dev *mvdev);
 void mlx5_vdpa_free_resources(struct mlx5_vdpa_dev *mvdev);
-int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mkey, u32 *in,
+int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, u32 *mkey, u32 *in,
 			  int inlen);
-int mlx5_vdpa_destroy_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mkey);
+int mlx5_vdpa_destroy_mkey(struct mlx5_vdpa_dev *mvdev, u32 mkey);
 int mlx5_vdpa_handle_set_map(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb,
 			     bool *change_map);
 int mlx5_vdpa_create_mr(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *iotlb);
diff --git a/drivers/vdpa/mlx5/core/mr.c b/drivers/vdpa/mlx5/core/mr.c
index ff010c6d0cd3..a639b9208d41 100644
--- a/drivers/vdpa/mlx5/core/mr.c
+++ b/drivers/vdpa/mlx5/core/mr.c
@@ -88,7 +88,7 @@ static int create_direct_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_direct
 
 static void destroy_direct_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_direct_mr *mr)
 {
-	mlx5_vdpa_destroy_mkey(mvdev, &mr->mr);
+	mlx5_vdpa_destroy_mkey(mvdev, mr->mr);
 }
 
 static u64 map_start(struct vhost_iotlb_map *map, struct mlx5_vdpa_direct_mr *mr)
@@ -162,7 +162,7 @@ static void fill_indir(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mkey, v
 		}
 
 		if (preve == dmr->start) {
-			klm->key = cpu_to_be32(dmr->mr.key);
+			klm->key = cpu_to_be32(dmr->mr);
 			klm->bcount = cpu_to_be32(klm_bcount(dmr->end - dmr->start));
 			preve = dmr->end;
 		} else {
@@ -217,7 +217,7 @@ static int create_indirect_key(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr
 
 static void destroy_indirect_key(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mkey)
 {
-	mlx5_vdpa_destroy_mkey(mvdev, &mkey->mkey);
+	mlx5_vdpa_destroy_mkey(mvdev, mkey->mkey);
 }
 
 static int map_direct_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_direct_mr *mr,
@@ -449,7 +449,7 @@ static int create_dma_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mr)
 
 static void destroy_dma_mr(struct mlx5_vdpa_dev *mvdev, struct mlx5_vdpa_mr *mr)
 {
-	mlx5_vdpa_destroy_mkey(mvdev, &mr->mkey);
+	mlx5_vdpa_destroy_mkey(mvdev, mr->mkey);
 }
 
 static int dup_iotlb(struct mlx5_vdpa_dev *mvdev, struct vhost_iotlb *src)
diff --git a/drivers/vdpa/mlx5/core/resources.c b/drivers/vdpa/mlx5/core/resources.c
index 72b2d80e75b0..9800f9bec225 100644
--- a/drivers/vdpa/mlx5/core/resources.c
+++ b/drivers/vdpa/mlx5/core/resources.c
@@ -198,7 +198,7 @@ void mlx5_vdpa_dealloc_transport_domain(struct mlx5_vdpa_dev *mvdev, u32 tdn)
 	mlx5_cmd_exec_in(mvdev->mdev, dealloc_transport_domain, in);
 }
 
-int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mkey, u32 *in,
+int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, u32 *mkey, u32 *in,
 			  int inlen)
 {
 	u32 lout[MLX5_ST_SZ_DW(create_mkey_out)] = {};
@@ -213,17 +213,17 @@ int mlx5_vdpa_create_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mk
 		return err;
 
 	mkey_index = MLX5_GET(create_mkey_out, lout, mkey_index);
-	mkey->key |= mlx5_idx_to_mkey(mkey_index);
+	*mkey |= mlx5_idx_to_mkey(mkey_index);
 	return 0;
 }
 
-int mlx5_vdpa_destroy_mkey(struct mlx5_vdpa_dev *mvdev, struct mlx5_core_mkey *mkey)
+int mlx5_vdpa_destroy_mkey(struct mlx5_vdpa_dev *mvdev, u32 mkey)
 {
 	u32 in[MLX5_ST_SZ_DW(destroy_mkey_in)] = {};
 
 	MLX5_SET(destroy_mkey_in, in, uid, mvdev->res.uid);
 	MLX5_SET(destroy_mkey_in, in, opcode, MLX5_CMD_OP_DESTROY_MKEY);
-	MLX5_SET(destroy_mkey_in, in, mkey_index, mlx5_mkey_to_idx(mkey->key));
+	MLX5_SET(destroy_mkey_in, in, mkey_index, mlx5_mkey_to_idx(mkey));
 	return mlx5_cmd_exec_in(mvdev->mdev, destroy_mkey, in);
 }
 
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index bd56de7484dc..5c7d2a953dbd 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
@@ -865,7 +865,7 @@ static int create_virtqueue(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtque
 	MLX5_SET64(virtio_q, vq_ctx, desc_addr, mvq->desc_addr);
 	MLX5_SET64(virtio_q, vq_ctx, used_addr, mvq->device_addr);
 	MLX5_SET64(virtio_q, vq_ctx, available_addr, mvq->driver_addr);
-	MLX5_SET(virtio_q, vq_ctx, virtio_q_mkey, ndev->mvdev.mr.mkey.key);
+	MLX5_SET(virtio_q, vq_ctx, virtio_q_mkey, ndev->mvdev.mr.mkey);
 	MLX5_SET(virtio_q, vq_ctx, umem_1_id, mvq->umem1.id);
 	MLX5_SET(virtio_q, vq_ctx, umem_1_size, mvq->umem1.size);
 	MLX5_SET(virtio_q, vq_ctx, umem_2_id, mvq->umem2.id);
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 31e19f06bc3a..42f653c70e8c 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -650,7 +650,7 @@ struct mlx5e_resources {
 	struct mlx5e_hw_objs {
 		u32                        pdn;
 		struct mlx5_td             td;
-		struct mlx5_core_mkey      mkey;
+		u32			   mkey;
 		struct mlx5_sq_bfreg       bfreg;
 	} hw_objs;
 	struct devlink_port dl_port;
@@ -1019,13 +1019,11 @@ struct mlx5_cmd_mailbox *mlx5_alloc_cmd_mailbox_chain(struct mlx5_core_dev *dev,
 						      gfp_t flags, int npages);
 void mlx5_free_cmd_mailbox_chain(struct mlx5_core_dev *dev,
 				 struct mlx5_cmd_mailbox *head);
-int mlx5_core_create_mkey(struct mlx5_core_dev *dev,
-			  struct mlx5_core_mkey *mkey,
-			  u32 *in, int inlen);
-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,
-			 u32 *out, int outlen);
+int mlx5_core_create_mkey(struct mlx5_core_dev *dev, u32 *mkey, u32 *in,
+			  int inlen);
+int mlx5_core_destroy_mkey(struct mlx5_core_dev *dev, u32 mkey);
+int mlx5_core_query_mkey(struct mlx5_core_dev *dev, u32 mkey, u32 *out,
+			 int outlen);
 int mlx5_core_alloc_pd(struct mlx5_core_dev *dev, u32 *pdn);
 int mlx5_core_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn);
 int mlx5_pagealloc_init(struct mlx5_core_dev *dev);
-- 
2.31.1

[PATCH mlx5-next 6/7] RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

Move mlx5_core_mkey struct to mlx5_ib, as the mlx5_core doesn't use it
at this point.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/devx.c    |  2 +-
 drivers/infiniband/hw/mlx5/mlx5_ib.h | 25 +++++++++++++++++++------
 drivers/infiniband/hw/mlx5/mr.c      | 12 +++++-------
 drivers/infiniband/hw/mlx5/odp.c     |  8 ++++----
 include/linux/mlx5/driver.h          | 13 -------------
 5 files changed, 29 insertions(+), 31 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
index 465ea835f854..2778b10ffd48 100644
--- a/drivers/infiniband/hw/mlx5/devx.c
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -1293,7 +1293,7 @@ static int devx_handle_mkey_indirect(struct devx_obj *obj,
 				     void *in, void *out)
 {
 	struct mlx5_ib_devx_mr *devx_mr = &obj->devx_mr;
-	struct mlx5_core_mkey *mkey;
+	struct mlx5_ib_mkey *mkey;
 	void *mkc;
 	u8 key;
 
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index cf8b0653f0ce..ef6087a9f93b 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -628,6 +628,19 @@ struct mlx5_user_mmap_entry {
 	u32 page_idx;
 };
 
+enum mlx5_mkey_type {
+	MLX5_MKEY_MR = 1,
+	MLX5_MKEY_MW,
+	MLX5_MKEY_INDIRECT_DEVX,
+};
+
+struct mlx5_ib_mkey {
+	u32			key;
+	enum mlx5_mkey_type	type;
+	struct wait_queue_head wait;
+	refcount_t usecount;
+};
+
 #define MLX5_IB_MTT_PRESENT (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)
 
 #define MLX5_IB_DM_MEMIC_ALLOWED_ACCESS (IB_ACCESS_LOCAL_WRITE   |\
@@ -646,7 +659,7 @@ struct mlx5_user_mmap_entry {
 
 struct mlx5_ib_mr {
 	struct ib_mr ibmr;
-	struct mlx5_core_mkey mmkey;
+	struct mlx5_ib_mkey mmkey;
 
 	/* User MR data */
 	struct mlx5_cache_ent *cache_ent;
@@ -722,12 +735,12 @@ static inline bool is_dmabuf_mr(struct mlx5_ib_mr *mr)
 
 struct mlx5_ib_mw {
 	struct ib_mw		ibmw;
-	struct mlx5_core_mkey	mmkey;
+	struct mlx5_ib_mkey	mmkey;
 	int			ndescs;
 };
 
 struct mlx5_ib_devx_mr {
-	struct mlx5_core_mkey	mmkey;
+	struct mlx5_ib_mkey	mmkey;
 	int			ndescs;
 };
 
@@ -1605,7 +1618,7 @@ static inline bool mlx5_ib_can_reconfig_with_umr(struct mlx5_ib_dev *dev,
 }
 
 static inline int mlx5r_store_odp_mkey(struct mlx5_ib_dev *dev,
-				       struct mlx5_core_mkey *mmkey)
+				       struct mlx5_ib_mkey *mmkey)
 {
 	refcount_set(&mmkey->usecount, 1);
 
@@ -1614,14 +1627,14 @@ static inline int mlx5r_store_odp_mkey(struct mlx5_ib_dev *dev,
 }
 
 /* deref an mkey that can participate in ODP flow */
-static inline void mlx5r_deref_odp_mkey(struct mlx5_core_mkey *mmkey)
+static inline void mlx5r_deref_odp_mkey(struct mlx5_ib_mkey *mmkey)
 {
 	if (refcount_dec_and_test(&mmkey->usecount))
 		wake_up(&mmkey->wait);
 }
 
 /* deref an mkey that can participate in ODP flow and wait for relese */
-static inline void mlx5r_deref_wait_odp_mkey(struct mlx5_core_mkey *mmkey)
+static inline void mlx5r_deref_wait_odp_mkey(struct mlx5_ib_mkey *mmkey)
 {
 	mlx5r_deref_odp_mkey(mmkey);
 	wait_event(mmkey->wait, refcount_read(&mmkey->usecount) == 0);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 2260b0298965..675be5b1de9c 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -89,9 +89,8 @@ static void set_mkc_access_pd_addr_fields(void *mkc, int acc, u64 start_addr,
 	MLX5_SET64(mkc, mkc, start_addr, start_addr);
 }
 
-static void
-assign_mkey_variant(struct mlx5_ib_dev *dev, struct mlx5_core_mkey *mkey,
-		    u32 *in)
+static void assign_mkey_variant(struct mlx5_ib_dev *dev,
+				struct mlx5_ib_mkey *mkey, u32 *in)
 {
 	u8 key = atomic_inc_return(&dev->mkey_var);
 	void *mkc;
@@ -101,9 +100,8 @@ assign_mkey_variant(struct mlx5_ib_dev *dev, struct mlx5_core_mkey *mkey,
 	mkey->key = key;
 }
 
-static int
-mlx5_ib_create_mkey(struct mlx5_ib_dev *dev, struct mlx5_core_mkey *mkey,
-		    u32 *in, int inlen)
+static int mlx5_ib_create_mkey(struct mlx5_ib_dev *dev,
+			       struct mlx5_ib_mkey *mkey, u32 *in, int inlen)
 {
 	int ret;
 
@@ -117,7 +115,7 @@ mlx5_ib_create_mkey(struct mlx5_ib_dev *dev, struct mlx5_core_mkey *mkey,
 
 static int
 mlx5_ib_create_mkey_cb(struct mlx5_ib_dev *dev,
-		       struct mlx5_core_mkey *mkey,
+		       struct mlx5_ib_mkey *mkey,
 		       struct mlx5_async_ctx *async_ctx,
 		       u32 *in, int inlen, u32 *out, int outlen,
 		       struct mlx5_async_work *context)
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 8ffd3ea8db7c..8dd9d8457767 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -788,7 +788,7 @@ struct pf_frame {
 	int depth;
 };
 
-static bool mkey_is_eq(struct mlx5_core_mkey *mmkey, u32 key)
+static bool mkey_is_eq(struct mlx5_ib_mkey *mmkey, u32 key)
 {
 	if (!mmkey)
 		return false;
@@ -797,7 +797,7 @@ static bool mkey_is_eq(struct mlx5_core_mkey *mmkey, u32 key)
 	return mmkey->key == key;
 }
 
-static int get_indirect_num_descs(struct mlx5_core_mkey *mmkey)
+static int get_indirect_num_descs(struct mlx5_ib_mkey *mmkey)
 {
 	struct mlx5_ib_mw *mw;
 	struct mlx5_ib_devx_mr *devx_mr;
@@ -831,7 +831,7 @@ static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
 {
 	int npages = 0, ret, i, outlen, cur_outlen = 0, depth = 0;
 	struct pf_frame *head = NULL, *frame;
-	struct mlx5_core_mkey *mmkey;
+	struct mlx5_ib_mkey *mmkey;
 	struct mlx5_ib_mr *mr;
 	struct mlx5_klm *pklm;
 	u32 *out = NULL;
@@ -1703,8 +1703,8 @@ get_prefetchable_mr(struct ib_pd *pd, enum ib_uverbs_advise_mr_advice advice,
 		    u32 lkey)
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
-	struct mlx5_core_mkey *mmkey;
 	struct mlx5_ib_mr *mr = NULL;
+	struct mlx5_ib_mkey *mmkey;
 
 	xa_lock(&dev->odp_mkeys);
 	mmkey = xa_load(&dev->odp_mkeys, mlx5_base_mkey(lkey));
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 42f653c70e8c..441a2f8715f8 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -357,19 +357,6 @@ struct mlx5_core_sig_ctx {
 	u32			sigerr_count;
 };
 
-enum {
-	MLX5_MKEY_MR = 1,
-	MLX5_MKEY_MW,
-	MLX5_MKEY_INDIRECT_DEVX,
-};
-
-struct mlx5_core_mkey {
-	u32			key;
-	u32			type;
-	struct wait_queue_head wait;
-	refcount_t usecount;
-};
-
 #define MLX5_24BIT_MASK		((1 << 24) - 1)
 
 enum mlx5_res_type {
-- 
2.31.1

[PATCH mlx5-next 7/7] RDMA/mlx5: Attach ndescs to mlx5_ib_mkey

[Leon Romanovsky]

From: Aharon Landau <aharonl@nvidia.com>

Generalize the use of ndescs by adding it to mlx5_ib_mkey.

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/devx.c    | 10 ++++------
 drivers/infiniband/hw/mlx5/devx.h    |  2 +-
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  8 +-------
 drivers/infiniband/hw/mlx5/mr.c      | 24 ++++++++++++------------
 drivers/infiniband/hw/mlx5/odp.c     | 22 ++--------------------
 drivers/infiniband/hw/mlx5/wr.c      | 10 +++++-----
 6 files changed, 25 insertions(+), 51 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
index 2778b10ffd48..2baecf0367e5 100644
--- a/drivers/infiniband/hw/mlx5/devx.c
+++ b/drivers/infiniband/hw/mlx5/devx.c
@@ -1292,18 +1292,16 @@ static int devx_handle_mkey_indirect(struct devx_obj *obj,
 				     struct mlx5_ib_dev *dev,
 				     void *in, void *out)
 {
-	struct mlx5_ib_devx_mr *devx_mr = &obj->devx_mr;
-	struct mlx5_ib_mkey *mkey;
+	struct mlx5_ib_mkey *mkey = &obj->mkey;
 	void *mkc;
 	u8 key;
 
-	mkey = &devx_mr->mmkey;
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 	key = MLX5_GET(mkc, mkc, mkey_7_0);
 	mkey->key = mlx5_idx_to_mkey(
 			MLX5_GET(create_mkey_out, out, mkey_index)) | key;
 	mkey->type = MLX5_MKEY_INDIRECT_DEVX;
-	devx_mr->ndescs = MLX5_GET(mkc, mkc, translations_octword_size);
+	mkey->ndescs = MLX5_GET(mkc, mkc, translations_octword_size);
 	init_waitqueue_head(&mkey->wait);
 
 	return mlx5r_store_odp_mkey(dev, mkey);
@@ -1381,13 +1379,13 @@ static int devx_obj_cleanup(struct ib_uobject *uobject,
 	dev = mlx5_udata_to_mdev(&attrs->driver_udata);
 	if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY &&
 	    xa_erase(&obj->ib_dev->odp_mkeys,
-		     mlx5_base_mkey(obj->devx_mr.mmkey.key)))
+		     mlx5_base_mkey(obj->mkey.key)))
 		/*
 		 * The pagefault_single_data_segment() does commands against
 		 * the mmkey, we must wait for that to stop before freeing the
 		 * mkey, as another allocation could get the same mkey #.
 		 */
-		mlx5r_deref_wait_odp_mkey(&obj->devx_mr.mmkey);
+		mlx5r_deref_wait_odp_mkey(&obj->mkey);
 
 	if (obj->flags & DEVX_OBJ_FLAGS_DCT)
 		ret = mlx5_core_destroy_dct(obj->ib_dev, &obj->core_dct);
diff --git a/drivers/infiniband/hw/mlx5/devx.h b/drivers/infiniband/hw/mlx5/devx.h
index 1f69866aed16..ee2213275fd6 100644
--- a/drivers/infiniband/hw/mlx5/devx.h
+++ b/drivers/infiniband/hw/mlx5/devx.h
@@ -16,7 +16,7 @@ struct devx_obj {
 	u32			dinbox[MLX5_MAX_DESTROY_INBOX_SIZE_DW];
 	u32			flags;
 	union {
-		struct mlx5_ib_devx_mr	devx_mr;
+		struct mlx5_ib_mkey	mkey;
 		struct mlx5_core_dct	core_dct;
 		struct mlx5_core_cq	core_cq;
 		u32			flow_counter_bulk_size;
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index ef6087a9f93b..ed173af8ae75 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -637,6 +637,7 @@ enum mlx5_mkey_type {
 struct mlx5_ib_mkey {
 	u32			key;
 	enum mlx5_mkey_type	type;
+	int			ndescs;
 	struct wait_queue_head wait;
 	refcount_t usecount;
 };
@@ -681,7 +682,6 @@ struct mlx5_ib_mr {
 			void *descs_alloc;
 			dma_addr_t desc_map;
 			int max_descs;
-			int ndescs;
 			int desc_size;
 			int access_mode;
 
@@ -736,12 +736,6 @@ static inline bool is_dmabuf_mr(struct mlx5_ib_mr *mr)
 struct mlx5_ib_mw {
 	struct ib_mw		ibmw;
 	struct mlx5_ib_mkey	mmkey;
-	int			ndescs;
-};
-
-struct mlx5_ib_devx_mr {
-	struct mlx5_ib_mkey	mmkey;
-	int			ndescs;
 };
 
 struct mlx5_ib_umr_context {
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 675be5b1de9c..0c4a5ac2544c 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -2335,7 +2335,7 @@ int mlx5_ib_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
 
 	mw->mmkey.type = MLX5_MKEY_MW;
 	ibmw->rkey = mw->mmkey.key;
-	mw->ndescs = ndescs;
+	mw->mmkey.ndescs = ndescs;
 
 	resp.response_length =
 		min(offsetofend(typeof(resp), response_length), udata->outlen);
@@ -2431,7 +2431,7 @@ mlx5_ib_map_pa_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
 	mr->meta_length = 0;
 	if (data_sg_nents == 1) {
 		n++;
-		mr->ndescs = 1;
+		mr->mmkey.ndescs = 1;
 		if (data_sg_offset)
 			sg_offset = *data_sg_offset;
 		mr->data_length = sg_dma_len(data_sg) - sg_offset;
@@ -2484,7 +2484,7 @@ mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
 	if (sg_offset_p)
 		*sg_offset_p = sg_offset;
 
-	mr->ndescs = i;
+	mr->mmkey.ndescs = i;
 	mr->data_length = mr->ibmr.length;
 
 	if (meta_sg_nents) {
@@ -2517,11 +2517,11 @@ static int mlx5_set_page(struct ib_mr *ibmr, u64 addr)
 	struct mlx5_ib_mr *mr = to_mmr(ibmr);
 	__be64 *descs;
 
-	if (unlikely(mr->ndescs == mr->max_descs))
+	if (unlikely(mr->mmkey.ndescs == mr->max_descs))
 		return -ENOMEM;
 
 	descs = mr->descs;
-	descs[mr->ndescs++] = cpu_to_be64(addr | MLX5_EN_RD | MLX5_EN_WR);
+	descs[mr->mmkey.ndescs++] = cpu_to_be64(addr | MLX5_EN_RD | MLX5_EN_WR);
 
 	return 0;
 }
@@ -2531,11 +2531,11 @@ static int mlx5_set_page_pi(struct ib_mr *ibmr, u64 addr)
 	struct mlx5_ib_mr *mr = to_mmr(ibmr);
 	__be64 *descs;
 
-	if (unlikely(mr->ndescs + mr->meta_ndescs == mr->max_descs))
+	if (unlikely(mr->mmkey.ndescs + mr->meta_ndescs == mr->max_descs))
 		return -ENOMEM;
 
 	descs = mr->descs;
-	descs[mr->ndescs + mr->meta_ndescs++] =
+	descs[mr->mmkey.ndescs + mr->meta_ndescs++] =
 		cpu_to_be64(addr | MLX5_EN_RD | MLX5_EN_WR);
 
 	return 0;
@@ -2551,7 +2551,7 @@ mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
 	struct mlx5_ib_mr *pi_mr = mr->mtt_mr;
 	int n;
 
-	pi_mr->ndescs = 0;
+	pi_mr->mmkey.ndescs = 0;
 	pi_mr->meta_ndescs = 0;
 	pi_mr->meta_length = 0;
 
@@ -2585,7 +2585,7 @@ mlx5_ib_map_mtt_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
 		 * metadata offset at the first metadata page
 		 */
 		pi_mr->pi_iova = (iova & page_mask) +
-				 pi_mr->ndescs * ibmr->page_size +
+				 pi_mr->mmkey.ndescs * ibmr->page_size +
 				 (pi_mr->ibmr.iova & ~page_mask);
 		/*
 		 * In order to use one MTT MR for data and metadata, we register
@@ -2616,7 +2616,7 @@ mlx5_ib_map_klm_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
 	struct mlx5_ib_mr *pi_mr = mr->klm_mr;
 	int n;
 
-	pi_mr->ndescs = 0;
+	pi_mr->mmkey.ndescs = 0;
 	pi_mr->meta_ndescs = 0;
 	pi_mr->meta_length = 0;
 
@@ -2651,7 +2651,7 @@ int mlx5_ib_map_mr_sg_pi(struct ib_mr *ibmr, struct scatterlist *data_sg,
 
 	WARN_ON(ibmr->type != IB_MR_TYPE_INTEGRITY);
 
-	mr->ndescs = 0;
+	mr->mmkey.ndescs = 0;
 	mr->data_length = 0;
 	mr->data_iova = 0;
 	mr->meta_ndescs = 0;
@@ -2707,7 +2707,7 @@ int mlx5_ib_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
 	struct mlx5_ib_mr *mr = to_mmr(ibmr);
 	int n;
 
-	mr->ndescs = 0;
+	mr->mmkey.ndescs = 0;
 
 	ib_dma_sync_single_for_cpu(ibmr->device, mr->desc_map,
 				   mr->desc_size * mr->max_descs,
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 8dd9d8457767..31596d8c5212 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -797,21 +797,6 @@ static bool mkey_is_eq(struct mlx5_ib_mkey *mmkey, u32 key)
 	return mmkey->key == key;
 }
 
-static int get_indirect_num_descs(struct mlx5_ib_mkey *mmkey)
-{
-	struct mlx5_ib_mw *mw;
-	struct mlx5_ib_devx_mr *devx_mr;
-
-	if (mmkey->type == MLX5_MKEY_MW) {
-		mw = container_of(mmkey, struct mlx5_ib_mw, mmkey);
-		return mw->ndescs;
-	}
-
-	devx_mr = container_of(mmkey, struct mlx5_ib_devx_mr,
-			       mmkey);
-	return devx_mr->ndescs;
-}
-
 /*
  * Handle a single data segment in a page-fault WQE or RDMA region.
  *
@@ -836,7 +821,6 @@ static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
 	struct mlx5_klm *pklm;
 	u32 *out = NULL;
 	size_t offset;
-	int ndescs;
 
 	io_virt += *bytes_committed;
 	bcnt -= *bytes_committed;
@@ -885,8 +869,6 @@ static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
 
 	case MLX5_MKEY_MW:
 	case MLX5_MKEY_INDIRECT_DEVX:
-		ndescs = get_indirect_num_descs(mmkey);
-
 		if (depth >= MLX5_CAP_GEN(dev->mdev, max_indirection)) {
 			mlx5_ib_dbg(dev, "indirection level exceeded\n");
 			ret = -EFAULT;
@@ -894,7 +876,7 @@ static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
 		}
 
 		outlen = MLX5_ST_SZ_BYTES(query_mkey_out) +
-			sizeof(*pklm) * (ndescs - 2);
+			sizeof(*pklm) * (mmkey->ndescs - 2);
 
 		if (outlen > cur_outlen) {
 			kfree(out);
@@ -916,7 +898,7 @@ static int pagefault_single_data_segment(struct mlx5_ib_dev *dev,
 		offset = io_virt - MLX5_GET64(query_mkey_out, out,
 					      memory_key_mkey_entry.start_addr);
 
-		for (i = 0; bcnt && i < ndescs; i++, pklm++) {
+		for (i = 0; bcnt && i < mmkey->ndescs; i++, pklm++) {
 			if (offset >= be32_to_cpu(pklm->bcount)) {
 				offset -= be32_to_cpu(pklm->bcount);
 				continue;
diff --git a/drivers/infiniband/hw/mlx5/wr.c b/drivers/infiniband/hw/mlx5/wr.c
index 8841620af82f..51e48ca9016e 100644
--- a/drivers/infiniband/hw/mlx5/wr.c
+++ b/drivers/infiniband/hw/mlx5/wr.c
@@ -217,7 +217,7 @@ static __be64 sig_mkey_mask(void)
 static void set_reg_umr_seg(struct mlx5_wqe_umr_ctrl_seg *umr,
 			    struct mlx5_ib_mr *mr, u8 flags, bool atomic)
 {
-	int size = (mr->ndescs + mr->meta_ndescs) * mr->desc_size;
+	int size = (mr->mmkey.ndescs + mr->meta_ndescs) * mr->desc_size;
 
 	memset(umr, 0, sizeof(*umr));
 
@@ -374,7 +374,7 @@ static void set_reg_mkey_seg(struct mlx5_mkey_seg *seg,
 			     struct mlx5_ib_mr *mr,
 			     u32 key, int access)
 {
-	int ndescs = ALIGN(mr->ndescs + mr->meta_ndescs, 8) >> 1;
+	int ndescs = ALIGN(mr->mmkey.ndescs + mr->meta_ndescs, 8) >> 1;
 
 	memset(seg, 0, sizeof(*seg));
 
@@ -439,7 +439,7 @@ static void set_reg_data_seg(struct mlx5_wqe_data_seg *dseg,
 			     struct mlx5_ib_mr *mr,
 			     struct mlx5_ib_pd *pd)
 {
-	int bcount = mr->desc_size * (mr->ndescs + mr->meta_ndescs);
+	int bcount = mr->desc_size * (mr->mmkey.ndescs + mr->meta_ndescs);
 
 	dseg->addr = cpu_to_be64(mr->desc_map);
 	dseg->byte_count = cpu_to_be32(ALIGN(bcount, 64));
@@ -861,7 +861,7 @@ static int set_reg_wr(struct mlx5_ib_qp *qp,
 	struct mlx5_ib_mr *mr = to_mmr(wr->mr);
 	struct mlx5_ib_pd *pd = to_mpd(qp->ibqp.pd);
 	struct mlx5_ib_dev *dev = to_mdev(pd->ibpd.device);
-	int mr_list_size = (mr->ndescs + mr->meta_ndescs) * mr->desc_size;
+	int mr_list_size = (mr->mmkey.ndescs + mr->meta_ndescs) * mr->desc_size;
 	bool umr_inline = mr_list_size <= MLX5_IB_SQ_UMR_INLINE_THRESHOLD;
 	bool atomic = wr->access & IB_ACCESS_REMOTE_ATOMIC;
 	u8 flags = 0;
@@ -1111,7 +1111,7 @@ static int handle_reg_mr_integrity(struct mlx5_ib_dev *dev,
 		memset(&pa_pi_mr, 0, sizeof(struct mlx5_ib_mr));
 		/* No UMR, use local_dma_lkey */
 		pa_pi_mr.ibmr.lkey = mr->ibmr.pd->local_dma_lkey;
-		pa_pi_mr.ndescs = mr->ndescs;
+		pa_pi_mr.mmkey.ndescs = mr->mmkey.ndescs;
 		pa_pi_mr.data_length = mr->data_length;
 		pa_pi_mr.data_iova = mr->data_iova;
 		if (mr->meta_ndescs) {
-- 
2.31.1

Re: [PATCH mlx5-next 0/7] Clean MR key use across mlx5_* modules

[Michael S. Tsirkin]

On Tue, Oct 12, 2021 at 01:26:28PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> Hi,
> 
> This is cleanup series of mlx5_* MR mkey management.
> 
> Thanks


Looks fine superficially

Acked-by: Michael S. Tsirkin <mst@redhat.com>

> Aharon Landau (7):
>   RDMA/mlx5: Don't set esc_size in user mr
>   RDMA/mlx5: Remove iova from struct mlx5_core_mkey
>   RDMA/mlx5: Remove size from struct mlx5_core_mkey
>   RDMA/mlx5: Remove pd from struct mlx5_core_mkey
>   RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
>   RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
>   RDMA/mlx5: Attach ndescs to mlx5_ib_mkey
> 
>  drivers/infiniband/hw/mlx5/devx.c             | 13 +--
>  drivers/infiniband/hw/mlx5/devx.h             |  2 +-
>  drivers/infiniband/hw/mlx5/mlx5_ib.h          | 31 ++++---
>  drivers/infiniband/hw/mlx5/mr.c               | 82 +++++++++----------
>  drivers/infiniband/hw/mlx5/odp.c              | 38 +++------
>  drivers/infiniband/hw/mlx5/wr.c               | 10 +--
>  .../mellanox/mlx5/core/diag/fw_tracer.c       |  6 +-
>  .../mellanox/mlx5/core/diag/fw_tracer.h       |  2 +-
>  .../mellanox/mlx5/core/diag/rsc_dump.c        | 10 +--
>  drivers/net/ethernet/mellanox/mlx5/core/en.h  |  2 +-
>  .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |  2 +-
>  .../net/ethernet/mellanox/mlx5/core/en/trap.c |  2 +-
>  .../ethernet/mellanox/mlx5/core/en_common.c   |  6 +-
>  .../net/ethernet/mellanox/mlx5/core/en_main.c | 13 ++-
>  .../ethernet/mellanox/mlx5/core/fpga/conn.c   | 10 +--
>  .../ethernet/mellanox/mlx5/core/fpga/core.h   |  2 +-
>  drivers/net/ethernet/mellanox/mlx5/core/mr.c  | 27 +++---
>  .../mellanox/mlx5/core/steering/dr_icm_pool.c | 10 +--
>  .../mellanox/mlx5/core/steering/dr_send.c     | 11 ++-
>  .../mellanox/mlx5/core/steering/dr_types.h    |  2 +-
>  drivers/vdpa/mlx5/core/mlx5_vdpa.h            |  8 +-
>  drivers/vdpa/mlx5/core/mr.c                   |  8 +-
>  drivers/vdpa/mlx5/core/resources.c            | 13 +--
>  drivers/vdpa/mlx5/net/mlx5_vnet.c             |  2 +-
>  include/linux/mlx5/driver.h                   | 30 ++-----
>  25 files changed, 147 insertions(+), 195 deletions(-)
> 
> -- 
> 2.31.1

Re: [PATCH mlx5-next 0/7] Clean MR key use across mlx5_* modules

[Leon Romanovsky]

On Tue, Oct 12, 2021 at 01:26:28PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> Hi,
> 
> This is cleanup series of mlx5_* MR mkey management.
> 
> Thanks
> 
> Aharon Landau (7):
>   RDMA/mlx5: Don't set esc_size in user mr
                         ^^^^ this typo, I will fix when apply to mmlx5-next.

Thanks


>   RDMA/mlx5: Remove iova from struct mlx5_core_mkey
>   RDMA/mlx5: Remove size from struct mlx5_core_mkey
>   RDMA/mlx5: Remove pd from struct mlx5_core_mkey
>   RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
>   RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
>   RDMA/mlx5: Attach ndescs to mlx5_ib_mkey
> 
>  drivers/infiniband/hw/mlx5/devx.c             | 13 +--
>  drivers/infiniband/hw/mlx5/devx.h             |  2 +-
>  drivers/infiniband/hw/mlx5/mlx5_ib.h          | 31 ++++---
>  drivers/infiniband/hw/mlx5/mr.c               | 82 +++++++++----------
>  drivers/infiniband/hw/mlx5/odp.c              | 38 +++------
>  drivers/infiniband/hw/mlx5/wr.c               | 10 +--
>  .../mellanox/mlx5/core/diag/fw_tracer.c       |  6 +-
>  .../mellanox/mlx5/core/diag/fw_tracer.h       |  2 +-
>  .../mellanox/mlx5/core/diag/rsc_dump.c        | 10 +--
>  drivers/net/ethernet/mellanox/mlx5/core/en.h  |  2 +-
>  .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |  2 +-
>  .../net/ethernet/mellanox/mlx5/core/en/trap.c |  2 +-
>  .../ethernet/mellanox/mlx5/core/en_common.c   |  6 +-
>  .../net/ethernet/mellanox/mlx5/core/en_main.c | 13 ++-
>  .../ethernet/mellanox/mlx5/core/fpga/conn.c   | 10 +--
>  .../ethernet/mellanox/mlx5/core/fpga/core.h   |  2 +-
>  drivers/net/ethernet/mellanox/mlx5/core/mr.c  | 27 +++---
>  .../mellanox/mlx5/core/steering/dr_icm_pool.c | 10 +--
>  .../mellanox/mlx5/core/steering/dr_send.c     | 11 ++-
>  .../mellanox/mlx5/core/steering/dr_types.h    |  2 +-
>  drivers/vdpa/mlx5/core/mlx5_vdpa.h            |  8 +-
>  drivers/vdpa/mlx5/core/mr.c                   |  8 +-
>  drivers/vdpa/mlx5/core/resources.c            | 13 +--
>  drivers/vdpa/mlx5/net/mlx5_vnet.c             |  2 +-
>  include/linux/mlx5/driver.h                   | 30 ++-----
>  25 files changed, 147 insertions(+), 195 deletions(-)
> 
> -- 
> 2.31.1
> 
> _______________________________________________
> Virtualization mailing list
> Virtualization@lists.linux-foundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [PATCH mlx5-next 1/7] RDMA/mlx5: Don't set esc_size in user mr

[Jason Gunthorpe]

On Tue, Oct 12, 2021 at 01:26:29PM +0300, Leon Romanovsky wrote:
> From: Aharon Landau <aharonl@nvidia.com>
> 
> reg_create() is used for user MRs only and should not set desc_size at
> all. Attempt to set it causes to the following trace:
> 
> BUG: unable to handle page fault for address: 0000000800000000
> PGD 0 P4D 0
> Oops: 0000 [#1] SMP PTI
> CPU: 5 PID: 890 Comm: ib_write_bw Not tainted 5.15.0-rc4+ #47
> Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
> RIP: 0010:mlx5_ib_dereg_mr+0x14/0x3b0 [mlx5_ib]
> Code: 48 63 cd 4c 89 f7 48 89 0c 24 e8 37 30 03 e1 48 8b 0c 24 eb a0 90 0f 1f 44 00 00 41 56 41 55 41 54 55 53 48 89 fb 48 83 ec 30 <48> 8b 2f 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 8b 87 c8
> RSP: 0018:ffff88811afa3a60 EFLAGS: 00010286
> RAX: 000000000000001c RBX: 0000000800000000 RCX: 0000000000000000
> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000800000000
> RBP: 0000000800000000 R08: 0000000000000000 R09: c0000000fffff7ff
> R10: ffff88811afa38f8 R11: ffff88811afa38f0 R12: ffffffffa02c7ac0
> R13: 0000000000000000 R14: ffff88811afa3cd8 R15: ffff88810772fa00
> FS:  00007f47b9080740(0000) GS:ffff88852cd40000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 0000000800000000 CR3: 000000010761e003 CR4: 0000000000370ea0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Call Trace:
>  mlx5_ib_free_odp_mr+0x95/0xc0 [mlx5_ib]
>  mlx5_ib_dereg_mr+0x128/0x3b0 [mlx5_ib]
>  ib_dereg_mr_user+0x45/0xb0 [ib_core]
>  ? xas_load+0x8/0x80
>  destroy_hw_idr_uobject+0x1a/0x50 [ib_uverbs]
>  uverbs_destroy_uobject+0x2f/0x150 [ib_uverbs]
>  uobj_destroy+0x3c/0x70 [ib_uverbs]
>  ib_uverbs_cmd_verbs+0x467/0xb00 [ib_uverbs]
>  ? uverbs_finalize_object+0x60/0x60 [ib_uverbs]
>  ? ttwu_queue_wakelist+0xa9/0xe0
>  ? pty_write+0x85/0x90
>  ? file_tty_write.isra.33+0x214/0x330
>  ? process_echoes+0x60/0x60
>  ib_uverbs_ioctl+0xa7/0x110 [ib_uverbs]
>  __x64_sys_ioctl+0x10d/0x8e0
>  ? vfs_write+0x17f/0x260
>  do_syscall_64+0x3c/0x80
>  entry_SYSCALL_64_after_hwframe+0x44/0xae
> 
> Fixes: a639e66703ee ("RDMA/mlx5: Zero out ODP related items in the mlx5_ib_mr")

Can you explain why this is crashing?

reg_create isn't used on the ODP implicit children path.

Jason

Re: [PATCH mlx5-next 1/7] RDMA/mlx5: Don't set esc_size in user mr

[Aharon Landau]

On 10/12/2021 3:52 PM, Jason Gunthorpe wrote:
> On Tue, Oct 12, 2021 at 01:26:29PM +0300, Leon Romanovsky wrote:
>> From: Aharon Landau <aharonl@nvidia.com>
>>
>> reg_create() is used for user MRs only and should not set desc_size at
>> all. Attempt to set it causes to the following trace:
>>
>> BUG: unable to handle page fault for address: 0000000800000000
>> PGD 0 P4D 0
>> Oops: 0000 [#1] SMP PTI
>> CPU: 5 PID: 890 Comm: ib_write_bw Not tainted 5.15.0-rc4+ #47
>> Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
>> RIP: 0010:mlx5_ib_dereg_mr+0x14/0x3b0 [mlx5_ib]
>> Code: 48 63 cd 4c 89 f7 48 89 0c 24 e8 37 30 03 e1 48 8b 0c 24 eb a0 90 0f 1f 44 00 00 41 56 41 55 41 54 55 53 48 89 fb 48 83 ec 30 <48> 8b 2f 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 8b 87 c8
>> RSP: 0018:ffff88811afa3a60 EFLAGS: 00010286
>> RAX: 000000000000001c RBX: 0000000800000000 RCX: 0000000000000000
>> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000800000000
>> RBP: 0000000800000000 R08: 0000000000000000 R09: c0000000fffff7ff
>> R10: ffff88811afa38f8 R11: ffff88811afa38f0 R12: ffffffffa02c7ac0
>> R13: 0000000000000000 R14: ffff88811afa3cd8 R15: ffff88810772fa00
>> FS:  00007f47b9080740(0000) GS:ffff88852cd40000(0000) knlGS:0000000000000000
>> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>> CR2: 0000000800000000 CR3: 000000010761e003 CR4: 0000000000370ea0
>> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
>> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
>> Call Trace:
>>   mlx5_ib_free_odp_mr+0x95/0xc0 [mlx5_ib]
>>   mlx5_ib_dereg_mr+0x128/0x3b0 [mlx5_ib]
>>   ib_dereg_mr_user+0x45/0xb0 [ib_core]
>>   ? xas_load+0x8/0x80
>>   destroy_hw_idr_uobject+0x1a/0x50 [ib_uverbs]
>>   uverbs_destroy_uobject+0x2f/0x150 [ib_uverbs]
>>   uobj_destroy+0x3c/0x70 [ib_uverbs]
>>   ib_uverbs_cmd_verbs+0x467/0xb00 [ib_uverbs]
>>   ? uverbs_finalize_object+0x60/0x60 [ib_uverbs]
>>   ? ttwu_queue_wakelist+0xa9/0xe0
>>   ? pty_write+0x85/0x90
>>   ? file_tty_write.isra.33+0x214/0x330
>>   ? process_echoes+0x60/0x60
>>   ib_uverbs_ioctl+0xa7/0x110 [ib_uverbs]
>>   __x64_sys_ioctl+0x10d/0x8e0
>>   ? vfs_write+0x17f/0x260
>>   do_syscall_64+0x3c/0x80
>>   entry_SYSCALL_64_after_hwframe+0x44/0xae
>>
>> Fixes: a639e66703ee ("RDMA/mlx5: Zero out ODP related items in the mlx5_ib_mr")
> Can you explain why this is crashing?
>
> reg_create isn't used on the ODP implicit children path.
>
> Jason
It is not implicit ODP flow, therefore, mr->implicit_children shouldn't 
be set.

desc_size shares the same union with implicit_children, and setting it 
leads to a wrong iteration.

Aharon

Re: [PATCH mlx5-next 1/7] RDMA/mlx5: Don't set esc_size in user mr

[Jason Gunthorpe]

On Tue, Oct 12, 2021 at 04:57:16PM +0300, Aharon Landau wrote:
> 
> On 10/12/2021 3:52 PM, Jason Gunthorpe wrote:
> > On Tue, Oct 12, 2021 at 01:26:29PM +0300, Leon Romanovsky wrote:
> > > From: Aharon Landau <aharonl@nvidia.com>
> > > 
> > > reg_create() is used for user MRs only and should not set desc_size at
> > > all. Attempt to set it causes to the following trace:
> > > 
> > > BUG: unable to handle page fault for address: 0000000800000000
> > > PGD 0 P4D 0
> > > Oops: 0000 [#1] SMP PTI
> > > CPU: 5 PID: 890 Comm: ib_write_bw Not tainted 5.15.0-rc4+ #47
> > > Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
> > > RIP: 0010:mlx5_ib_dereg_mr+0x14/0x3b0 [mlx5_ib]
> > > Code: 48 63 cd 4c 89 f7 48 89 0c 24 e8 37 30 03 e1 48 8b 0c 24 eb a0 90 0f 1f 44 00 00 41 56 41 55 41 54 55 53 48 89 fb 48 83 ec 30 <48> 8b 2f 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 8b 87 c8
> > > RSP: 0018:ffff88811afa3a60 EFLAGS: 00010286
> > > RAX: 000000000000001c RBX: 0000000800000000 RCX: 0000000000000000
> > > RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000800000000
> > > RBP: 0000000800000000 R08: 0000000000000000 R09: c0000000fffff7ff
> > > R10: ffff88811afa38f8 R11: ffff88811afa38f0 R12: ffffffffa02c7ac0
> > > R13: 0000000000000000 R14: ffff88811afa3cd8 R15: ffff88810772fa00
> > > FS:  00007f47b9080740(0000) GS:ffff88852cd40000(0000) knlGS:0000000000000000
> > > CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > CR2: 0000000800000000 CR3: 000000010761e003 CR4: 0000000000370ea0
> > > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> > > Call Trace:
> > >   mlx5_ib_free_odp_mr+0x95/0xc0 [mlx5_ib]
> > >   mlx5_ib_dereg_mr+0x128/0x3b0 [mlx5_ib]
> > >   ib_dereg_mr_user+0x45/0xb0 [ib_core]
> > >   ? xas_load+0x8/0x80
> > >   destroy_hw_idr_uobject+0x1a/0x50 [ib_uverbs]
> > >   uverbs_destroy_uobject+0x2f/0x150 [ib_uverbs]
> > >   uobj_destroy+0x3c/0x70 [ib_uverbs]
> > >   ib_uverbs_cmd_verbs+0x467/0xb00 [ib_uverbs]
> > >   ? uverbs_finalize_object+0x60/0x60 [ib_uverbs]
> > >   ? ttwu_queue_wakelist+0xa9/0xe0
> > >   ? pty_write+0x85/0x90
> > >   ? file_tty_write.isra.33+0x214/0x330
> > >   ? process_echoes+0x60/0x60
> > >   ib_uverbs_ioctl+0xa7/0x110 [ib_uverbs]
> > >   __x64_sys_ioctl+0x10d/0x8e0
> > >   ? vfs_write+0x17f/0x260
> > >   do_syscall_64+0x3c/0x80
> > >   entry_SYSCALL_64_after_hwframe+0x44/0xae
> > > 
> > > Fixes: a639e66703ee ("RDMA/mlx5: Zero out ODP related items in the mlx5_ib_mr")
> > Can you explain why this is crashing?
> > 
> > reg_create isn't used on the ODP implicit children path.
> > 
> > Jason
> It is not implicit ODP flow, therefore, mr->implicit_children shouldn't be
> set.

It should always be initialized. That seems to be the bug here, add the
missing xa_init as well as delete the extra desc_size set:

diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index b4d2322e9ca564..46626e0fe08905 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -1525,6 +1525,7 @@ static struct ib_mr *create_user_odp_mr(struct ib_pd *pd, u64 start, u64 length,
                ib_umem_release(&odp->umem);
                return ERR_CAST(mr);
        }
+       xa_init(&mr->implicit_children);
 
        odp->private = mr;
        err = mlx5r_store_odp_mkey(dev, &mr->mmkey);

Jason

Re: [PATCH mlx5-next 1/7] RDMA/mlx5: Don't set esc_size in user mr

[Leon Romanovsky]

On Tue, Oct 12, 2021 at 11:04:33AM -0300, Jason Gunthorpe wrote:
> On Tue, Oct 12, 2021 at 04:57:16PM +0300, Aharon Landau wrote:
> > 
> > On 10/12/2021 3:52 PM, Jason Gunthorpe wrote:
> > > On Tue, Oct 12, 2021 at 01:26:29PM +0300, Leon Romanovsky wrote:
> > > > From: Aharon Landau <aharonl@nvidia.com>
> > > > 
> > > > reg_create() is used for user MRs only and should not set desc_size at
> > > > all. Attempt to set it causes to the following trace:
> > > > 
> > > > BUG: unable to handle page fault for address: 0000000800000000
> > > > PGD 0 P4D 0
> > > > Oops: 0000 [#1] SMP PTI
> > > > CPU: 5 PID: 890 Comm: ib_write_bw Not tainted 5.15.0-rc4+ #47
> > > > Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
> > > > RIP: 0010:mlx5_ib_dereg_mr+0x14/0x3b0 [mlx5_ib]
> > > > Code: 48 63 cd 4c 89 f7 48 89 0c 24 e8 37 30 03 e1 48 8b 0c 24 eb a0 90 0f 1f 44 00 00 41 56 41 55 41 54 55 53 48 89 fb 48 83 ec 30 <48> 8b 2f 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 8b 87 c8
> > > > RSP: 0018:ffff88811afa3a60 EFLAGS: 00010286
> > > > RAX: 000000000000001c RBX: 0000000800000000 RCX: 0000000000000000
> > > > RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000800000000
> > > > RBP: 0000000800000000 R08: 0000000000000000 R09: c0000000fffff7ff
> > > > R10: ffff88811afa38f8 R11: ffff88811afa38f0 R12: ffffffffa02c7ac0
> > > > R13: 0000000000000000 R14: ffff88811afa3cd8 R15: ffff88810772fa00
> > > > FS:  00007f47b9080740(0000) GS:ffff88852cd40000(0000) knlGS:0000000000000000
> > > > CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > CR2: 0000000800000000 CR3: 000000010761e003 CR4: 0000000000370ea0
> > > > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > > > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> > > > Call Trace:
> > > >   mlx5_ib_free_odp_mr+0x95/0xc0 [mlx5_ib]
> > > >   mlx5_ib_dereg_mr+0x128/0x3b0 [mlx5_ib]
> > > >   ib_dereg_mr_user+0x45/0xb0 [ib_core]
> > > >   ? xas_load+0x8/0x80
> > > >   destroy_hw_idr_uobject+0x1a/0x50 [ib_uverbs]
> > > >   uverbs_destroy_uobject+0x2f/0x150 [ib_uverbs]
> > > >   uobj_destroy+0x3c/0x70 [ib_uverbs]
> > > >   ib_uverbs_cmd_verbs+0x467/0xb00 [ib_uverbs]
> > > >   ? uverbs_finalize_object+0x60/0x60 [ib_uverbs]
> > > >   ? ttwu_queue_wakelist+0xa9/0xe0
> > > >   ? pty_write+0x85/0x90
> > > >   ? file_tty_write.isra.33+0x214/0x330
> > > >   ? process_echoes+0x60/0x60
> > > >   ib_uverbs_ioctl+0xa7/0x110 [ib_uverbs]
> > > >   __x64_sys_ioctl+0x10d/0x8e0
> > > >   ? vfs_write+0x17f/0x260
> > > >   do_syscall_64+0x3c/0x80
> > > >   entry_SYSCALL_64_after_hwframe+0x44/0xae
> > > > 
> > > > Fixes: a639e66703ee ("RDMA/mlx5: Zero out ODP related items in the mlx5_ib_mr")
> > > Can you explain why this is crashing?
> > > 
> > > reg_create isn't used on the ODP implicit children path.
> > > 
> > > Jason
> > It is not implicit ODP flow, therefore, mr->implicit_children shouldn't be
> > set.
> 
> It should always be initialized. That seems to be the bug here, add the
> missing xa_init as well as delete the extra desc_size set:

I would expect such change in mlx5_ib_init_odp_mr().

> 
> diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
> index b4d2322e9ca564..46626e0fe08905 100644
> --- a/drivers/infiniband/hw/mlx5/mr.c
> +++ b/drivers/infiniband/hw/mlx5/mr.c
> @@ -1525,6 +1525,7 @@ static struct ib_mr *create_user_odp_mr(struct ib_pd *pd, u64 start, u64 length,
>                 ib_umem_release(&odp->umem);
>                 return ERR_CAST(mr);
>         }
> +       xa_init(&mr->implicit_children);
>  
>         odp->private = mr;
>         err = mlx5r_store_odp_mkey(dev, &mr->mmkey);
> 
> Jason

Re: [PATCH mlx5-next 1/7] RDMA/mlx5: Don't set esc_size in user mr

[Jason Gunthorpe]

On Tue, Oct 12, 2021 at 05:14:20PM +0300, Leon Romanovsky wrote:
> On Tue, Oct 12, 2021 at 11:04:33AM -0300, Jason Gunthorpe wrote:
> > On Tue, Oct 12, 2021 at 04:57:16PM +0300, Aharon Landau wrote:
> > > 
> > > On 10/12/2021 3:52 PM, Jason Gunthorpe wrote:
> > > > On Tue, Oct 12, 2021 at 01:26:29PM +0300, Leon Romanovsky wrote:
> > > > > From: Aharon Landau <aharonl@nvidia.com>
> > > > > 
> > > > > reg_create() is used for user MRs only and should not set desc_size at
> > > > > all. Attempt to set it causes to the following trace:
> > > > > 
> > > > > BUG: unable to handle page fault for address: 0000000800000000
> > > > > PGD 0 P4D 0
> > > > > Oops: 0000 [#1] SMP PTI
> > > > > CPU: 5 PID: 890 Comm: ib_write_bw Not tainted 5.15.0-rc4+ #47
> > > > > Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
> > > > > RIP: 0010:mlx5_ib_dereg_mr+0x14/0x3b0 [mlx5_ib]
> > > > > Code: 48 63 cd 4c 89 f7 48 89 0c 24 e8 37 30 03 e1 48 8b 0c 24 eb a0 90 0f 1f 44 00 00 41 56 41 55 41 54 55 53 48 89 fb 48 83 ec 30 <48> 8b 2f 65 48 8b 04 25 28 00 00 00 48 89 44 24 28 31 c0 8b 87 c8
> > > > > RSP: 0018:ffff88811afa3a60 EFLAGS: 00010286
> > > > > RAX: 000000000000001c RBX: 0000000800000000 RCX: 0000000000000000
> > > > > RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000800000000
> > > > > RBP: 0000000800000000 R08: 0000000000000000 R09: c0000000fffff7ff
> > > > > R10: ffff88811afa38f8 R11: ffff88811afa38f0 R12: ffffffffa02c7ac0
> > > > > R13: 0000000000000000 R14: ffff88811afa3cd8 R15: ffff88810772fa00
> > > > > FS:  00007f47b9080740(0000) GS:ffff88852cd40000(0000) knlGS:0000000000000000
> > > > > CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > > CR2: 0000000800000000 CR3: 000000010761e003 CR4: 0000000000370ea0
> > > > > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > > > > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> > > > > Call Trace:
> > > > >   mlx5_ib_free_odp_mr+0x95/0xc0 [mlx5_ib]
> > > > >   mlx5_ib_dereg_mr+0x128/0x3b0 [mlx5_ib]
> > > > >   ib_dereg_mr_user+0x45/0xb0 [ib_core]
> > > > >   ? xas_load+0x8/0x80
> > > > >   destroy_hw_idr_uobject+0x1a/0x50 [ib_uverbs]
> > > > >   uverbs_destroy_uobject+0x2f/0x150 [ib_uverbs]
> > > > >   uobj_destroy+0x3c/0x70 [ib_uverbs]
> > > > >   ib_uverbs_cmd_verbs+0x467/0xb00 [ib_uverbs]
> > > > >   ? uverbs_finalize_object+0x60/0x60 [ib_uverbs]
> > > > >   ? ttwu_queue_wakelist+0xa9/0xe0
> > > > >   ? pty_write+0x85/0x90
> > > > >   ? file_tty_write.isra.33+0x214/0x330
> > > > >   ? process_echoes+0x60/0x60
> > > > >   ib_uverbs_ioctl+0xa7/0x110 [ib_uverbs]
> > > > >   __x64_sys_ioctl+0x10d/0x8e0
> > > > >   ? vfs_write+0x17f/0x260
> > > > >   do_syscall_64+0x3c/0x80
> > > > >   entry_SYSCALL_64_after_hwframe+0x44/0xae
> > > > > 
> > > > > Fixes: a639e66703ee ("RDMA/mlx5: Zero out ODP related items in the mlx5_ib_mr")
> > > > Can you explain why this is crashing?
> > > > 
> > > > reg_create isn't used on the ODP implicit children path.
> > > > 
> > > > Jason
> > > It is not implicit ODP flow, therefore, mr->implicit_children shouldn't be
> > > set.
> > 
> > It should always be initialized. That seems to be the bug here, add the
> > missing xa_init as well as delete the extra desc_size set:
> 
> I would expect such change in mlx5_ib_init_odp_mr().

It is too late

Jason

Re: [PATCH mlx5-next 6/7] RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib

[Jason Gunthorpe]

On Tue, Oct 12, 2021 at 01:26:34PM +0300, Leon Romanovsky wrote:
> From: Aharon Landau <aharonl@nvidia.com>
> 
> Move mlx5_core_mkey struct to mlx5_ib, as the mlx5_core doesn't use it
> at this point.
> 
> Signed-off-by: Aharon Landau <aharonl@nvidia.com>
> Reviewed-by: Shay Drory <shayd@nvidia.com>
> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
>  drivers/infiniband/hw/mlx5/devx.c    |  2 +-
>  drivers/infiniband/hw/mlx5/mlx5_ib.h | 25 +++++++++++++++++++------
>  drivers/infiniband/hw/mlx5/mr.c      | 12 +++++-------
>  drivers/infiniband/hw/mlx5/odp.c     |  8 ++++----
>  include/linux/mlx5/driver.h          | 13 -------------
>  5 files changed, 29 insertions(+), 31 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c
> index 465ea835f854..2778b10ffd48 100644
> +++ b/drivers/infiniband/hw/mlx5/devx.c
> @@ -1293,7 +1293,7 @@ static int devx_handle_mkey_indirect(struct devx_obj *obj,
>  				     void *in, void *out)
>  {
>  	struct mlx5_ib_devx_mr *devx_mr = &obj->devx_mr;
> -	struct mlx5_core_mkey *mkey;
> +	struct mlx5_ib_mkey *mkey;
>  	void *mkc;
>  	u8 key;
>  
> diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
> index cf8b0653f0ce..ef6087a9f93b 100644
> +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
> @@ -628,6 +628,19 @@ struct mlx5_user_mmap_entry {
>  	u32 page_idx;
>  };
>  
> +enum mlx5_mkey_type {
> +	MLX5_MKEY_MR = 1,
> +	MLX5_MKEY_MW,
> +	MLX5_MKEY_INDIRECT_DEVX,
> +};
> +
> +struct mlx5_ib_mkey {
> +	u32			key;
> +	enum mlx5_mkey_type	type;
> +	struct wait_queue_head wait;
> +	refcount_t usecount;
> +};

Please drop the horizontal whitespace when you move the struct

Jason

Re: [PATCH mlx5-next 7/7] RDMA/mlx5: Attach ndescs to mlx5_ib_mkey

[Jason Gunthorpe]

On Tue, Oct 12, 2021 at 01:26:35PM +0300, Leon Romanovsky wrote:

> diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
> index ef6087a9f93b..ed173af8ae75 100644
> +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
> @@ -637,6 +637,7 @@ enum mlx5_mkey_type {
>  struct mlx5_ib_mkey {
>  	u32			key;
>  	enum mlx5_mkey_type	type;
> +	int			ndescs;

unsigned int

Jason

Re: [PATCH mlx5-next 0/7] Clean MR key use across mlx5_* modules

[Jason Gunthorpe]

On Tue, Oct 12, 2021 at 01:26:28PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> Hi,
> 
> This is cleanup series of mlx5_* MR mkey management.
> 
> Thanks
> 
> Aharon Landau (7):
>   RDMA/mlx5: Don't set esc_size in user mr

Please sent just this patch to -rc after modifying as I noted, don't
put this in mlx5-next

>   RDMA/mlx5: Remove iova from struct mlx5_core_mkey
>   RDMA/mlx5: Remove size from struct mlx5_core_mkey
>   RDMA/mlx5: Remove pd from struct mlx5_core_mkey
>   RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
>   RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
>   RDMA/mlx5: Attach ndescs to mlx5_ib_mkey

It seems fine to me, other than the little notes, a V2 can go to
mlx5-next

Jason

Re: [PATCH mlx5-next 0/7] Clean MR key use across mlx5_* modules

[Leon Romanovsky]

On Wed, Oct 13, 2021 at 11:43:03AM -0300, Jason Gunthorpe wrote:
> On Tue, Oct 12, 2021 at 01:26:28PM +0300, Leon Romanovsky wrote:
> > From: Leon Romanovsky <leonro@nvidia.com>
> > 
> > Hi,
> > 
> > This is cleanup series of mlx5_* MR mkey management.
> > 
> > Thanks
> > 
> > Aharon Landau (7):
> >   RDMA/mlx5: Don't set esc_size in user mr
> 
> Please sent just this patch to -rc after modifying as I noted, don't
> put this in mlx5-next
> 
> >   RDMA/mlx5: Remove iova from struct mlx5_core_mkey
> >   RDMA/mlx5: Remove size from struct mlx5_core_mkey
> >   RDMA/mlx5: Remove pd from struct mlx5_core_mkey
> >   RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
> >   RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
> >   RDMA/mlx5: Attach ndescs to mlx5_ib_mkey
> 
> It seems fine to me, other than the little notes, a V2 can go to
> mlx5-next

Applied to vfio-next, to serve as a basis for live migration patches.
https://lore.kernel.org/kvm/20211019105838.227569-1-yishaih@nvidia.com/T/#m89f4f0ec4baddeb2828a2b38dcbbd6900009fb83

ae0579acde81 RDMA/mlx5: Attach ndescs to mlx5_ib_mkey
4123bfb0b28b RDMA/mlx5: Move struct mlx5_core_mkey to mlx5_ib
83fec3f12a59 RDMA/mlx5: Replace struct mlx5_core_mkey by u32 key
c64674168b6a RDMA/mlx5: Remove pd from struct mlx5_core_mkey
062fd731e51e RDMA/mlx5: Remove size from struct mlx5_core_mkey
cf6a8b1b24d6 RDMA/mlx5: Remove iova from struct mlx5_core_mkey

https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux.git/log/?h=vfio-next

Thanks

> 
> Jason