All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH rdma-rc] IB/mlx5: Fix MR registration flow to use UMR properly
@ 2019-07-31  8:19 Leon Romanovsky
  2019-08-01 15:50 ` Doug Ledford
  0 siblings, 1 reply; 2+ messages in thread
From: Leon Romanovsky @ 2019-07-31  8:19 UTC (permalink / raw)
  To: Doug Ledford, Jason Gunthorpe
  Cc: Guy Levi, RDMA mailing list, Majd Dibbiny, Moni Shoua, Leon Romanovsky

From: Guy Levi <guyle@mellanox.com>

Driver shouldn't allow to use UMR to register a MR when
umr_modify_atomic_disabled is set. Otherwise it will always end up with a
failure in the post send flow which sets the UMR WQE to modify atomic access
right.

Fixes: c8d75a980fab ("IB/mlx5: Respect new UMR capabilities")
Signed-off-by: Guy Levi <guyle@mellanox.com>
Reviewed-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
---
 drivers/infiniband/hw/mlx5/mr.c | 27 +++++++++------------------
 1 file changed, 9 insertions(+), 18 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 2c77456f359f..b74fad08412f 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -51,22 +51,12 @@ static void clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
 static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
 static int mr_cache_max_order(struct mlx5_ib_dev *dev);
 static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
-static bool umr_can_modify_entity_size(struct mlx5_ib_dev *dev)
-{
-	return !MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled);
-}
 
 static bool umr_can_use_indirect_mkey(struct mlx5_ib_dev *dev)
 {
 	return !MLX5_CAP_GEN(dev->mdev, umr_indirect_mkey_disabled);
 }
 
-static bool use_umr(struct mlx5_ib_dev *dev, int order)
-{
-	return order <= mr_cache_max_order(dev) &&
-		umr_can_modify_entity_size(dev);
-}
-
 static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 {
 	int err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
@@ -1271,7 +1261,7 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
 	struct mlx5_ib_mr *mr = NULL;
-	bool populate_mtts = false;
+	bool use_umr;
 	struct ib_umem *umem;
 	int page_shift;
 	int npages;
@@ -1303,29 +1293,30 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	if (err < 0)
 		return ERR_PTR(err);
 
-	if (use_umr(dev, order)) {
+	use_umr = !MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled) &&
+		  (!MLX5_CAP_GEN(dev->mdev, umr_modify_atomic_disabled) ||
+		   !MLX5_CAP_GEN(dev->mdev, atomic));
+
+	if (order <= mr_cache_max_order(dev) && use_umr) {
 		mr = alloc_mr_from_cache(pd, umem, virt_addr, length, ncont,
 					 page_shift, order, access_flags);
 		if (PTR_ERR(mr) == -EAGAIN) {
 			mlx5_ib_dbg(dev, "cache empty for order %d\n", order);
 			mr = NULL;
 		}
-		populate_mtts = false;
 	} else if (!MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset)) {
 		if (access_flags & IB_ACCESS_ON_DEMAND) {
 			err = -EINVAL;
 			pr_err("Got MR registration for ODP MR > 512MB, not supported for Connect-IB\n");
 			goto error;
 		}
-		populate_mtts = true;
+		use_umr = false;
 	}
 
 	if (!mr) {
-		if (!umr_can_modify_entity_size(dev))
-			populate_mtts = true;
 		mutex_lock(&dev->slow_path_mutex);
 		mr = reg_create(NULL, pd, virt_addr, length, umem, ncont,
-				page_shift, access_flags, populate_mtts);
+				page_shift, access_flags, !use_umr);
 		mutex_unlock(&dev->slow_path_mutex);
 	}
 
@@ -1341,7 +1332,7 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 
 	update_odp_mr(mr);
 
-	if (!populate_mtts) {
+	if (use_umr) {
 		int update_xlt_flags = MLX5_IB_UPD_XLT_ENABLE;
 
 		if (access_flags & IB_ACCESS_ON_DEMAND)
-- 
2.20.1


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH rdma-rc] IB/mlx5: Fix MR registration flow to use UMR properly
  2019-07-31  8:19 [PATCH rdma-rc] IB/mlx5: Fix MR registration flow to use UMR properly Leon Romanovsky
@ 2019-08-01 15:50 ` Doug Ledford
  0 siblings, 0 replies; 2+ messages in thread
From: Doug Ledford @ 2019-08-01 15:50 UTC (permalink / raw)
  To: Leon Romanovsky, Jason Gunthorpe
  Cc: Guy Levi, RDMA mailing list, Majd Dibbiny, Moni Shoua, Leon Romanovsky

[-- Attachment #1: Type: text/plain, Size: 731 bytes --]

On Wed, 2019-07-31 at 11:19 +0300, Leon Romanovsky wrote:
> From: Guy Levi <guyle@mellanox.com>
> 
> Driver shouldn't allow to use UMR to register a MR when
> umr_modify_atomic_disabled is set. Otherwise it will always end up
> with a
> failure in the post send flow which sets the UMR WQE to modify atomic
> access
> right.
> 
> Fixes: c8d75a980fab ("IB/mlx5: Respect new UMR capabilities")
> Signed-off-by: Guy Levi <guyle@mellanox.com>
> Reviewed-by: Moni Shoua <monis@mellanox.com>
> Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

Thanks, applied to for-rc.

-- 
Doug Ledford <dledford@redhat.com>
    GPG KeyID: B826A3330E572FDD
    Fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2019-08-01 15:50 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-07-31  8:19 [PATCH rdma-rc] IB/mlx5: Fix MR registration flow to use UMR properly Leon Romanovsky
2019-08-01 15:50 ` Doug Ledford

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.