Commit c08fbdc5 authored by Jason Gunthorpe's avatar Jason Gunthorpe
Browse files

RDMA/mlx5: mlx5_umem_find_best_quantized_pgoff() for CQ 

This fixes a bug where the page_offset was not being considered when
building a CQ. The HW specification says it 'must be zero', so use
a variant of mlx5_umem_find_best_quantized_pgoff() with a 0 pgoff_bitmask
to force this result.

Fixes: e126ba97 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Link: https://lore.kernel.org/r/20201115114311.136250-6-leon@kernel.org


Signed-off-by: default avatarLeon Romanovsky <leonro@nvidia.com>
Signed-off-by: default avatarJason Gunthorpe <jgg@nvidia.com>
parent a59b7b05

drivers/infiniband/hw/mlx5/cq.c

+34 −14
Original line number Diff line number Diff line
@@ -707,8 +707,9 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata, 
			  int *cqe_size, int *index, int *inlen)

{

	struct mlx5_ib_create_cq ucmd = {};

	unsigned long page_size;

	unsigned int page_offset_quantized;

	size_t ucmdlen;

	int page_shift;

	__be64 *pas;

	int ncont;

	void *cqc;
@@ -741,17 +742,24 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata, 
		return err;

	}



	page_size = mlx5_umem_find_best_cq_quantized_pgoff(

		cq->buf.umem, cqc, log_page_size, MLX5_ADAPTER_PAGE_SHIFT,

		page_offset, 64, &page_offset_quantized);

	if (!page_size) {

		err = -EINVAL;

		goto err_umem;

	}



	err = mlx5_ib_db_map_user(context, udata, ucmd.db_addr, &cq->db);

	if (err)

		goto err_umem;



	mlx5_ib_cont_pages(cq->buf.umem, ucmd.buf_addr, 0, &page_shift);

	ncont = ib_umem_num_dma_blocks(cq->buf.umem, 1UL << page_shift);

	ncont = ib_umem_num_dma_blocks(cq->buf.umem, page_size);

	mlx5_ib_dbg(

		dev,

		"addr 0x%llx, size %u, npages %zu, page_shift %d, ncont %d\n",

		"addr 0x%llx, size %u, npages %zu, page_size %lu, ncont %d\n",

		ucmd.buf_addr, entries * ucmd.cqe_size,

		ib_umem_num_pages(cq->buf.umem), page_shift, ncont);

		ib_umem_num_pages(cq->buf.umem), page_size, ncont);



	*inlen = MLX5_ST_SZ_BYTES(create_cq_in) +

		 MLX5_FLD_SZ_BYTES(create_cq_in, pas[0]) * ncont;
@@ -762,11 +770,12 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata, 
	}



	pas = (__be64 *)MLX5_ADDR_OF(create_cq_in, *cqb, pas);

	mlx5_ib_populate_pas(cq->buf.umem, 1UL << page_shift, pas, 0);

	mlx5_ib_populate_pas(cq->buf.umem, page_size, pas, 0);



	cqc = MLX5_ADDR_OF(create_cq_in, *cqb, cq_context);

	MLX5_SET(cqc, cqc, log_page_size,

		 page_shift - MLX5_ADAPTER_PAGE_SHIFT);

		 order_base_2(page_size) - MLX5_ADAPTER_PAGE_SHIFT);

	MLX5_SET(cqc, cqc, page_offset, page_offset_quantized);



	if (ucmd.flags & MLX5_IB_CREATE_CQ_FLAGS_UAR_PAGE_INDEX) {

		*index = ucmd.uar_page_index;
@@ -1131,7 +1140,7 @@ int mlx5_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period) 


static int resize_user(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,

		       int entries, struct ib_udata *udata,

		       int *page_shift, int *cqe_size)

		       int *cqe_size)

{

	struct mlx5_ib_resize_cq ucmd;

	struct ib_umem *umem;
@@ -1156,8 +1165,6 @@ static int resize_user(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq, 
		return err;

	}



	mlx5_ib_cont_pages(umem, ucmd.buf_addr, 0, page_shift);



	cq->resize_umem = umem;

	*cqe_size = ucmd.cqe_size;
@@ -1250,7 +1257,8 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) 
	int err;

	int npas;

	__be64 *pas;

	int page_shift;

	unsigned int page_offset_quantized = 0;

	unsigned int page_shift;

	int inlen;

	int cqe_size;

	unsigned long flags;
@@ -1277,11 +1285,22 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) 


	mutex_lock(&cq->resize_mutex);

	if (udata) {

		err = resize_user(dev, cq, entries, udata, &page_shift,

				  &cqe_size);

		unsigned long page_size;



		err = resize_user(dev, cq, entries, udata, &cqe_size);

		if (err)

			goto ex;

		npas = ib_umem_num_dma_blocks(cq->resize_umem, 1UL << page_shift);



		page_size = mlx5_umem_find_best_cq_quantized_pgoff(

			cq->resize_umem, cqc, log_page_size,

			MLX5_ADAPTER_PAGE_SHIFT, page_offset, 64,

			&page_offset_quantized);

		if (!page_size) {

			err = -EINVAL;

			goto ex_resize;

		}

		npas = ib_umem_num_dma_blocks(cq->resize_umem, page_size);

		page_shift = order_base_2(page_size);

	} else {

		struct mlx5_frag_buf *frag_buf;
@@ -1320,6 +1339,7 @@ int mlx5_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) 


	MLX5_SET(cqc, cqc, log_page_size,

		 page_shift - MLX5_ADAPTER_PAGE_SHIFT);

	MLX5_SET(cqc, cqc, page_offset, page_offset_quantized);

	MLX5_SET(cqc, cqc, cqe_sz,

		 cqe_sz_to_mlx_sz(cqe_size,

				  cq->private_flags &

drivers/infiniband/hw/mlx5/mlx5_ib.h

+10 −0
Original line number Diff line number Diff line
@@ -100,6 +100,16 @@ unsigned long __mlx5_umem_find_best_quantized_pgoff( 
		GENMASK(31, order_base_2(scale)), scale,                       \

		page_offset_quantized)



#define mlx5_umem_find_best_cq_quantized_pgoff(umem, typ, log_pgsz_fld,        \

					       pgsz_shift, page_offset_fld,    \

					       scale, page_offset_quantized)   \

	__mlx5_umem_find_best_quantized_pgoff(                                 \

		umem,                                                          \

		__mlx5_log_page_size_to_bitmap(                                \

			__mlx5_bit_sz(typ, log_pgsz_fld), pgsz_shift),         \

		__mlx5_bit_sz(typ, page_offset_fld), 0, scale,                 \

		page_offset_quantized)



enum {

	MLX5_IB_MMAP_OFFSET_START = 9,

	MLX5_IB_MMAP_OFFSET_END = 255,