net/mlx5e: RX, Remove mlx5e_alloc_unit argument in page allocation
authorDragos Tatulea <dtatulea@nvidia.com>
Fri, 27 Jan 2023 15:06:12 +0000 (17:06 +0200)
committerSaeed Mahameed <saeedm@nvidia.com>
Tue, 28 Mar 2023 20:43:56 +0000 (13:43 -0700)
Change internal page cache and page pool api to use a struct page **
instead of a mlx5e_alloc_unit *.

This is the first change in a series which is meant to remove the
mlx5e_alloc_unit altogether.

Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c

index 3f7b63d6616bab70b87eee7448b84ca46ded6ed6..36300118b6e4edb1a5bcf4a4f7ac6bcf03f4e91d 100644 (file)
@@ -292,7 +292,7 @@ static inline bool mlx5e_rx_cache_put(struct mlx5e_rq *rq, struct page *page)
        return true;
 }
 
-static inline bool mlx5e_rx_cache_get(struct mlx5e_rq *rq, union mlx5e_alloc_unit *au)
+static inline bool mlx5e_rx_cache_get(struct mlx5e_rq *rq, struct page **pagep)
 {
        struct mlx5e_page_cache *cache = &rq->page_cache;
        struct mlx5e_rq_stats *stats = rq->stats;
@@ -308,35 +308,35 @@ static inline bool mlx5e_rx_cache_get(struct mlx5e_rq *rq, union mlx5e_alloc_uni
                return false;
        }
 
-       au->page = cache->page_cache[cache->head];
+       *pagep = cache->page_cache[cache->head];
        cache->head = (cache->head + 1) & (MLX5E_CACHE_SIZE - 1);
        stats->cache_reuse++;
 
-       addr = page_pool_get_dma_addr(au->page);
+       addr = page_pool_get_dma_addr(*pagep);
        /* Non-XSK always uses PAGE_SIZE. */
        dma_sync_single_for_device(rq->pdev, addr, PAGE_SIZE, rq->buff.map_dir);
        return true;
 }
 
-static inline int mlx5e_page_alloc_pool(struct mlx5e_rq *rq, union mlx5e_alloc_unit *au)
+static inline int mlx5e_page_alloc_pool(struct mlx5e_rq *rq, struct page **pagep)
 {
        dma_addr_t addr;
 
-       if (mlx5e_rx_cache_get(rq, au))
+       if (mlx5e_rx_cache_get(rq, pagep))
                return 0;
 
-       au->page = page_pool_dev_alloc_pages(rq->page_pool);
-       if (unlikely(!au->page))
+       *pagep = page_pool_dev_alloc_pages(rq->page_pool);
+       if (unlikely(!*pagep))
                return -ENOMEM;
 
        /* Non-XSK always uses PAGE_SIZE. */
-       addr = dma_map_page(rq->pdev, au->page, 0, PAGE_SIZE, rq->buff.map_dir);
+       addr = dma_map_page(rq->pdev, *pagep, 0, PAGE_SIZE, rq->buff.map_dir);
        if (unlikely(dma_mapping_error(rq->pdev, addr))) {
-               page_pool_recycle_direct(rq->page_pool, au->page);
-               au->page = NULL;
+               page_pool_recycle_direct(rq->page_pool, *pagep);
+               *pagep = NULL;
                return -ENOMEM;
        }
-       page_pool_set_dma_addr(au->page, addr);
+       page_pool_set_dma_addr(*pagep, addr);
 
        return 0;
 }
@@ -376,7 +376,7 @@ static inline int mlx5e_get_rx_frag(struct mlx5e_rq *rq,
                 * offset) should just use the new one without replenishing again
                 * by themselves.
                 */
-               err = mlx5e_page_alloc_pool(rq, frag->au);
+               err = mlx5e_page_alloc_pool(rq, &frag->au->page);
 
        return err;
 }
@@ -605,13 +605,14 @@ static int mlx5e_build_shampo_hd_umr(struct mlx5e_rq *rq,
                header_offset = (index & (MLX5E_SHAMPO_WQ_HEADER_PER_PAGE - 1)) <<
                        MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE;
                if (!(header_offset & (PAGE_SIZE - 1))) {
-                       union mlx5e_alloc_unit au;
 
-                       err = mlx5e_page_alloc_pool(rq, &au);
+                       err = mlx5e_page_alloc_pool(rq, &page);
                        if (unlikely(err))
                                goto err_unmap;
-                       page = dma_info->page = au.page;
-                       addr = dma_info->addr = page_pool_get_dma_addr(au.page);
+
+                       addr = page_pool_get_dma_addr(page);
+                       dma_info->addr = addr;
+                       dma_info->page = page;
                } else {
                        dma_info->addr = addr + header_offset;
                        dma_info->page = page;
@@ -715,7 +716,7 @@ static int mlx5e_alloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix)
        for (i = 0; i < rq->mpwqe.pages_per_wqe; i++, au++) {
                dma_addr_t addr;
 
-               err = mlx5e_page_alloc_pool(rq, au);
+               err = mlx5e_page_alloc_pool(rq, &au->page);
                if (unlikely(err))
                        goto err_unmap;
                addr = page_pool_get_dma_addr(au->page);