| Message ID | 20170308161718.2867-7-edumazet@google.com |
|---|---|
| State | Accepted, archived |
| Delegated to: | David Miller |
| Headers | show |
On 08/03/2017 6:17 PM, Eric Dumazet wrote: > We will soon use order-0 pages, and frag truesize will more precisely > match real sizes. > > In the new model, we prefer to use <= 2048 bytes fragments, so that > we can use page-recycle technique on PAGE_SIZE=4096 arches. > > We will still pack as much frames as possible on arches with big > pages, like PowerPC. > > Signed-off-by: Eric Dumazet <edumazet@google.com> > --- > drivers/net/ethernet/mellanox/mlx4/en_rx.c | 24 ++++++++++-------------- > drivers/net/ethernet/mellanox/mlx4/mlx4_en.h | 8 -------- > 2 files changed, 10 insertions(+), 22 deletions(-) > Acked-by: Tariq Toukan <tariqt@mellanox.com> Thanks.
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_rx.c b/drivers/net/ethernet/mellanox/mlx4/en_rx.c index 453313d404e3698b0d41a8220005b3834c9d68a1..0c61c1200f2aec4c74d7403d9ec3ed06821c1bac 100644 --- a/drivers/net/ethernet/mellanox/mlx4/en_rx.c +++ b/drivers/net/ethernet/mellanox/mlx4/en_rx.c @@ -1181,13 +1181,6 @@ int mlx4_en_poll_rx_cq(struct napi_struct *napi, int budget) return done; } -static const int frag_sizes[] = { - FRAG_SZ0, - FRAG_SZ1, - FRAG_SZ2, - FRAG_SZ3 -}; - void mlx4_en_calc_rx_buf(struct net_device *dev) { struct mlx4_en_priv *priv = netdev_priv(dev); @@ -1211,13 +1204,16 @@ void mlx4_en_calc_rx_buf(struct net_device *dev) int buf_size = 0; while (buf_size < eff_mtu) { - priv->frag_info[i].frag_size = - (eff_mtu > buf_size + frag_sizes[i]) ? - frag_sizes[i] : eff_mtu - buf_size; - priv->frag_info[i].frag_stride = - ALIGN(priv->frag_info[i].frag_size, - SMP_CACHE_BYTES); - buf_size += priv->frag_info[i].frag_size; + int frag_size = eff_mtu - buf_size; + + if (i < MLX4_EN_MAX_RX_FRAGS - 1) + frag_size = min(frag_size, 2048); + + priv->frag_info[i].frag_size = frag_size; + + priv->frag_info[i].frag_stride = ALIGN(frag_size, + SMP_CACHE_BYTES); + buf_size += frag_size; i++; } priv->rx_page_order = MLX4_EN_ALLOC_PREFER_ORDER; diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h index c98940c6acd02c1bd8c4f82107d82979ead051d1..608396dc6d95a05cd88bcf054df990298365ecfa 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h +++ b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h @@ -105,14 +105,6 @@ #define MLX4_EN_ALLOC_PREFER_ORDER min_t(int, get_order(32768), \ PAGE_ALLOC_COSTLY_ORDER) -/* Receive fragment sizes; we use at most 3 fragments (for 9600 byte MTU - * and 4K allocations) */ -enum { - FRAG_SZ0 = 1536 - NET_IP_ALIGN, - FRAG_SZ1 = 4096, - FRAG_SZ2 = 4096, - FRAG_SZ3 = MLX4_EN_ALLOC_SIZE -}; #define MLX4_EN_MAX_RX_FRAGS 4 /* Maximum ring sizes */
We will soon use order-0 pages, and frag truesize will more precisely match real sizes. In the new model, we prefer to use <= 2048 bytes fragments, so that we can use page-recycle technique on PAGE_SIZE=4096 arches. We will still pack as much frames as possible on arches with big pages, like PowerPC. Signed-off-by: Eric Dumazet <edumazet@google.com> --- drivers/net/ethernet/mellanox/mlx4/en_rx.c | 24 ++++++++++-------------- drivers/net/ethernet/mellanox/mlx4/mlx4_en.h | 8 -------- 2 files changed, 10 insertions(+), 22 deletions(-)