RDMA/mlx5: Use ib_umem_find_best_pgoff() for SRQ

SRQ uses a quantized and scaled page_offset, which is another variation of
ib_umem_find_best_pgsz(). Add mlx5_umem_find_best_quantized_pgoff() to
perform this calculation for each mailbox. A macro shows how the
calculation is directly connected to the mailbox format.

This new routine replaces the limited mlx5_ib_cont_pages() and
mlx5_ib_get_buf_offset() pairing which would reject valid configurations
rather than adjust the page_size to make it work.

In turn this is much more aggressive about choosing large page sizes for
these objects and when THP is enabled it will now often find a single page
solution.

Link: https://lore.kernel.org/r/20201115114311.136250-2-leon@kernel.org
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

1 files changed, 10 insertions, 10 deletions

diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c
index cb5ad045da9a..7dfdc9e54866 100644
--- a/drivers/infiniband/hw/mlx5/srq.c
+++ b/drivers/infiniband/hw/mlx5/srq.c
@@ -51,8 +51,8 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq,
 		udata, struct mlx5_ib_ucontext, ibucontext);
 	size_t ucmdlen;
 	int err;
-	int page_shift;
-	u32 offset;
+	unsigned int page_offset_quantized;
+	unsigned int page_size;
 	u32 uidx = MLX5_IB_DEFAULT_UIDX;
 
 	ucmdlen = min(udata->inlen, sizeof(ucmd));
@@ -85,22 +85,22 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq,
 		return err;
 	}
 
-	mlx5_ib_cont_pages(srq->umem, ucmd.buf_addr, 0, &page_shift);
-	err = mlx5_ib_get_buf_offset(ucmd.buf_addr, page_shift,
-				     &offset);
-	if (err) {
+	page_size = mlx5_umem_find_best_quantized_pgoff(
+		srq->umem, srqc, log_page_size, MLX5_ADAPTER_PAGE_SHIFT,
+		page_offset, 64, &page_offset_quantized);
+	if (!page_size) {
 		mlx5_ib_warn(dev, "bad offset\n");
 		goto err_umem;
 	}
 
-	in->pas = kvcalloc(ib_umem_num_dma_blocks(srq->umem, 1UL << page_shift),
+	in->pas = kvcalloc(ib_umem_num_dma_blocks(srq->umem, page_size),
 			   sizeof(*in->pas), GFP_KERNEL);
 	if (!in->pas) {
 		err = -ENOMEM;
 		goto err_umem;
 	}
 
-	mlx5_ib_populate_pas(srq->umem, 1UL << page_shift, in->pas, 0);
+	mlx5_ib_populate_pas(srq->umem, page_size, in->pas, 0);
 
 	err = mlx5_ib_db_map_user(ucontext, udata, ucmd.db_addr, &srq->db);
 	if (err) {
@@ -108,8 +108,8 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq,
 		goto err_in;
 	}
 
-	in->log_page_size = page_shift - MLX5_ADAPTER_PAGE_SHIFT;
-	in->page_offset = offset;
+	in->log_page_size = order_base_2(page_size) - MLX5_ADAPTER_PAGE_SHIFT;
+	in->page_offset = page_offset_quantized;
 	in->uid = (in->type != IB_SRQT_XRC) ?  to_mpd(pd)->uid : 0;
 	if (MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1 &&
 	    in->type != IB_SRQT_BASIC)