From ab6bf42e2339580b5d87746d0ff4da4b1578b03e Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Wed, 27 May 2009 14:38:34 -0700 Subject: mlx4_core: Add module parameter for number of MTTs per segment The current MTT allocator uses kmalloc() to allocate a buffer for its buddy allocator, and thus is limited in the amount of MTT segments that it can control. As a result, the size of memory that can be registered is limited too. This patch uses a module parameter to control the number of MTT entries that each segment represents, allowing more memory to be registered with the same number of segments. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/net/mlx4/mr.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/net/mlx4/mr.c') diff --git a/drivers/net/mlx4/mr.c b/drivers/net/mlx4/mr.c index 0caf74cae8b..3b8973d1993 100644 --- a/drivers/net/mlx4/mr.c +++ b/drivers/net/mlx4/mr.c @@ -209,7 +209,7 @@ int mlx4_mtt_init(struct mlx4_dev *dev, int npages, int page_shift, } else mtt->page_shift = page_shift; - for (mtt->order = 0, i = MLX4_MTT_ENTRY_PER_SEG; i < npages; i <<= 1) + for (mtt->order = 0, i = dev->caps.mtts_per_seg; i < npages; i <<= 1) ++mtt->order; mtt->first_seg = mlx4_alloc_mtt_range(dev, mtt->order); @@ -350,7 +350,7 @@ int mlx4_mr_enable(struct mlx4_dev *dev, struct mlx4_mr *mr) mpt_entry->pd_flags |= cpu_to_be32(MLX4_MPT_PD_FLAG_FAST_REG | MLX4_MPT_PD_FLAG_RAE); mpt_entry->mtt_sz = cpu_to_be32((1 << mr->mtt.order) * - MLX4_MTT_ENTRY_PER_SEG); + dev->caps.mtts_per_seg); } else { mpt_entry->flags |= cpu_to_be32(MLX4_MPT_FLAG_SW_OWNS); } @@ -391,7 +391,7 @@ static int mlx4_write_mtt_chunk(struct mlx4_dev *dev, struct mlx4_mtt *mtt, (start_index + npages - 1) / (PAGE_SIZE / sizeof (u64))) return -EINVAL; - if (start_index & (MLX4_MTT_ENTRY_PER_SEG - 1)) + if (start_index & (dev->caps.mtts_per_seg - 1)) return -EINVAL; mtts = mlx4_table_find(&priv->mr_table.mtt_table, mtt->first_seg + -- cgit v1.2.3-70-g09d2 From 5d23a1d2a317e272f171d7818f600fde324eaceb Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Wed, 27 May 2009 15:10:42 +0000 Subject: net: replace dma_sync_single with dma_sync_single_for_cpu This replaces dma_sync_single() with dma_sync_single_for_cpu() because dma_sync_single() is an obsolete API; include/linux/dma-mapping.h says: /* Backwards compat, remove in 2.7.x */ #define dma_sync_single dma_sync_single_for_cpu #define dma_sync_sg dma_sync_sg_for_cpu Signed-off-by: FUJITA Tomonori Signed-off-by: David S. Miller --- drivers/net/arm/ep93xx_eth.c | 4 ++-- drivers/net/arm/ixp4xx_eth.c | 4 ++-- drivers/net/mlx4/mr.c | 7 ++++--- drivers/net/wan/ixp4xx_hss.c | 4 ++-- 4 files changed, 10 insertions(+), 9 deletions(-) (limited to 'drivers/net/mlx4/mr.c') diff --git a/drivers/net/arm/ep93xx_eth.c b/drivers/net/arm/ep93xx_eth.c index b72b3d639f6..fbf4645417d 100644 --- a/drivers/net/arm/ep93xx_eth.c +++ b/drivers/net/arm/ep93xx_eth.c @@ -253,7 +253,7 @@ static int ep93xx_rx(struct net_device *dev, int processed, int budget) skb = dev_alloc_skb(length + 2); if (likely(skb != NULL)) { skb_reserve(skb, 2); - dma_sync_single(NULL, ep->descs->rdesc[entry].buf_addr, + dma_sync_single_for_cpu(NULL, ep->descs->rdesc[entry].buf_addr, length, DMA_FROM_DEVICE); skb_copy_to_linear_data(skb, ep->rx_buf[entry], length); skb_put(skb, length); @@ -331,7 +331,7 @@ static int ep93xx_xmit(struct sk_buff *skb, struct net_device *dev) ep->descs->tdesc[entry].tdesc1 = TDESC1_EOF | (entry << 16) | (skb->len & 0xfff); skb_copy_and_csum_dev(skb, ep->tx_buf[entry]); - dma_sync_single(NULL, ep->descs->tdesc[entry].buf_addr, + dma_sync_single_for_cpu(NULL, ep->descs->tdesc[entry].buf_addr, skb->len, DMA_TO_DEVICE); dev_kfree_skb(skb); diff --git a/drivers/net/arm/ixp4xx_eth.c b/drivers/net/arm/ixp4xx_eth.c index 322c49b908d..1fcf8388b1c 100644 --- a/drivers/net/arm/ixp4xx_eth.c +++ b/drivers/net/arm/ixp4xx_eth.c @@ -561,8 +561,8 @@ static int eth_poll(struct napi_struct *napi, int budget) dma_unmap_single(&dev->dev, desc->data - NET_IP_ALIGN, RX_BUFF_SIZE, DMA_FROM_DEVICE); #else - dma_sync_single(&dev->dev, desc->data - NET_IP_ALIGN, - RX_BUFF_SIZE, DMA_FROM_DEVICE); + dma_sync_single_for_cpu(&dev->dev, desc->data - NET_IP_ALIGN, + RX_BUFF_SIZE, DMA_FROM_DEVICE); memcpy_swab32((u32 *)skb->data, (u32 *)port->rx_buff_tab[n], ALIGN(NET_IP_ALIGN + desc->pkt_len, 4) / 4); #endif diff --git a/drivers/net/mlx4/mr.c b/drivers/net/mlx4/mr.c index 0caf74cae8b..0a467785f06 100644 --- a/drivers/net/mlx4/mr.c +++ b/drivers/net/mlx4/mr.c @@ -402,7 +402,8 @@ static int mlx4_write_mtt_chunk(struct mlx4_dev *dev, struct mlx4_mtt *mtt, for (i = 0; i < npages; ++i) mtts[i] = cpu_to_be64(page_list[i] | MLX4_MTT_FLAG_PRESENT); - dma_sync_single(&dev->pdev->dev, dma_handle, npages * sizeof (u64), DMA_TO_DEVICE); + dma_sync_single_for_cpu(&dev->pdev->dev, dma_handle, + npages * sizeof (u64), DMA_TO_DEVICE); return 0; } @@ -549,8 +550,8 @@ int mlx4_map_phys_fmr(struct mlx4_dev *dev, struct mlx4_fmr *fmr, u64 *page_list for (i = 0; i < npages; ++i) fmr->mtts[i] = cpu_to_be64(page_list[i] | MLX4_MTT_FLAG_PRESENT); - dma_sync_single(&dev->pdev->dev, fmr->dma_handle, - npages * sizeof(u64), DMA_TO_DEVICE); + dma_sync_single_for_cpu(&dev->pdev->dev, fmr->dma_handle, + npages * sizeof(u64), DMA_TO_DEVICE); fmr->mpt->key = cpu_to_be32(key); fmr->mpt->lkey = cpu_to_be32(key); diff --git a/drivers/net/wan/ixp4xx_hss.c b/drivers/net/wan/ixp4xx_hss.c index 765a7f5d6aa..08b1a284b69 100644 --- a/drivers/net/wan/ixp4xx_hss.c +++ b/drivers/net/wan/ixp4xx_hss.c @@ -731,8 +731,8 @@ static int hss_hdlc_poll(struct napi_struct *napi, int budget) dma_unmap_single(&dev->dev, desc->data, RX_SIZE, DMA_FROM_DEVICE); #else - dma_sync_single(&dev->dev, desc->data, - RX_SIZE, DMA_FROM_DEVICE); + dma_sync_single_for_cpu(&dev->dev, desc->data, + RX_SIZE, DMA_FROM_DEVICE); memcpy_swab32((u32 *)skb->data, (u32 *)port->rx_buff_tab[n], ALIGN(desc->pkt_len, 4) / 4); #endif -- cgit v1.2.3-70-g09d2 From e727f5cde90d5a8b92a1ffa49c636a3790301469 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 22 Jun 2009 23:07:56 -0700 Subject: mlx4_core: Fix dma_sync_single_for_cpu() with matching for_device() calls Commit 5d23a1d2 ("net: replace dma_sync_single with dma_sync_single_for_cpu") replaced uses of the deprectated function dma_sync_single() with calls to dma_sync_single_for_cpu(). However, to be correct, the code should do a sync for_cpu() before touching the memory and for_device() after it's done. Signed-off-by: Roland Dreier --- drivers/net/mlx4/mr.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'drivers/net/mlx4/mr.c') diff --git a/drivers/net/mlx4/mr.c b/drivers/net/mlx4/mr.c index 5887e4764d2..f96948be0a4 100644 --- a/drivers/net/mlx4/mr.c +++ b/drivers/net/mlx4/mr.c @@ -399,11 +399,14 @@ static int mlx4_write_mtt_chunk(struct mlx4_dev *dev, struct mlx4_mtt *mtt, if (!mtts) return -ENOMEM; + dma_sync_single_for_cpu(&dev->pdev->dev, dma_handle, + npages * sizeof (u64), DMA_TO_DEVICE); + for (i = 0; i < npages; ++i) mtts[i] = cpu_to_be64(page_list[i] | MLX4_MTT_FLAG_PRESENT); - dma_sync_single_for_cpu(&dev->pdev->dev, dma_handle, - npages * sizeof (u64), DMA_TO_DEVICE); + dma_sync_single_for_device(&dev->pdev->dev, dma_handle, + npages * sizeof (u64), DMA_TO_DEVICE); return 0; } @@ -547,11 +550,14 @@ int mlx4_map_phys_fmr(struct mlx4_dev *dev, struct mlx4_fmr *fmr, u64 *page_list /* Make sure MPT status is visible before writing MTT entries */ wmb(); + dma_sync_single_for_cpu(&dev->pdev->dev, fmr->dma_handle, + npages * sizeof(u64), DMA_TO_DEVICE); + for (i = 0; i < npages; ++i) fmr->mtts[i] = cpu_to_be64(page_list[i] | MLX4_MTT_FLAG_PRESENT); - dma_sync_single_for_cpu(&dev->pdev->dev, fmr->dma_handle, - npages * sizeof(u64), DMA_TO_DEVICE); + dma_sync_single_for_device(&dev->pdev->dev, fmr->dma_handle, + npages * sizeof(u64), DMA_TO_DEVICE); fmr->mpt->key = cpu_to_be32(key); fmr->mpt->lkey = cpu_to_be32(key); -- cgit v1.2.3-70-g09d2