From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alice Michael Date: Wed, 5 Apr 2017 07:51:01 -0400 Subject: [Intel-wired-lan] [next S66 v2 09/11] i40e/i40evf: Add support for using order 1 pages with a 3K buffer In-Reply-To: <20170405115103.67374-1-alice.michael@intel.com> References: <20170405115103.67374-1-alice.michael@intel.com> Message-ID: <20170405115103.67374-9-alice.michael@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: intel-wired-lan@osuosl.org List-ID: From: Alexander Duyck There are situations where adding padding to the front and back of an Rx buffer will require that we add additional padding. Specifically if NET_IP_ALIGN is non-zero, or the MTU size is larger than 7.5K we would need to use 2K buffers which leaves us with no room for the padding. To preemptively address these cases I am adding support for 3K buffers to the Rx path so that we can provide the additional padding needed in the event of NET_IP_ALIGN being non-zero or a cache line being greater than 64. Signed-off-by: Alexander Duyck Change-ID: I938bc1ba611285428df39a613cd66f98e60b55c7 --- drivers/net/ethernet/intel/i40e/i40e_main.c | 3 ++- drivers/net/ethernet/intel/i40e/i40e_txrx.c | 27 +++++++++++++------------ drivers/net/ethernet/intel/i40e/i40e_txrx.h | 12 +++++++++++ drivers/net/ethernet/intel/i40evf/i40e_txrx.c | 27 +++++++++++++------------ drivers/net/ethernet/intel/i40evf/i40e_txrx.h | 12 +++++++++++ drivers/net/ethernet/intel/i40evf/i40evf_main.c | 6 ++++++ 6 files changed, 60 insertions(+), 27 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index 89f3659..859c8e1 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -3085,7 +3085,8 @@ static int i40e_vsi_configure_rx(struct i40e_vsi *vsi) #endif } else { vsi->max_frame = I40E_MAX_RXBUFFER; - vsi->rx_buf_len = I40E_RXBUFFER_2048; + vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 : + I40E_RXBUFFER_2048; } /* set up individual rings */ diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index e95428c..bee1672 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -1138,14 +1138,15 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring) dma_sync_single_range_for_cpu(rx_ring->dev, rx_bi->dma, rx_bi->page_offset, - I40E_RXBUFFER_2048, + rx_ring->rx_buf_len, DMA_FROM_DEVICE); /* free resources associated with mapping */ dma_unmap_page_attrs(rx_ring->dev, rx_bi->dma, - PAGE_SIZE, + i40e_rx_pg_size(rx_ring), DMA_FROM_DEVICE, I40E_RX_DMA_ATTR); + __page_frag_cache_drain(rx_bi->page, rx_bi->pagecnt_bias); rx_bi->page = NULL; @@ -1267,7 +1268,7 @@ static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, } /* alloc new page for storage */ - page = dev_alloc_page(); + page = dev_alloc_pages(i40e_rx_pg_order(rx_ring)); if (unlikely(!page)) { rx_ring->rx_stats.alloc_page_failed++; return false; @@ -1275,7 +1276,7 @@ static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, /* map page for use */ dma = dma_map_page_attrs(rx_ring->dev, page, 0, - PAGE_SIZE, + i40e_rx_pg_size(rx_ring), DMA_FROM_DEVICE, I40E_RX_DMA_ATTR); @@ -1283,7 +1284,7 @@ static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, * there isn't much point in holding memory we can't use */ if (dma_mapping_error(rx_ring->dev, dma)) { - __free_pages(page, 0); + __free_pages(page, i40e_rx_pg_order(rx_ring)); rx_ring->rx_stats.alloc_page_failed++; return false; } @@ -1343,7 +1344,7 @@ bool i40e_alloc_rx_buffers(struct i40e_ring *rx_ring, u16 cleaned_count) /* sync the buffer for use by the device */ dma_sync_single_range_for_device(rx_ring->dev, bi->dma, bi->page_offset, - I40E_RXBUFFER_2048, + rx_ring->rx_buf_len, DMA_FROM_DEVICE); /* Refresh the desc even if buffer_addrs didn't change @@ -1645,9 +1646,6 @@ static inline bool i40e_page_is_reusable(struct page *page) **/ static bool i40e_can_reuse_rx_page(struct i40e_rx_buffer *rx_buffer) { -#if (PAGE_SIZE >= 8192) - unsigned int last_offset = PAGE_SIZE - I40E_RXBUFFER_2048; -#endif unsigned int pagecnt_bias = rx_buffer->pagecnt_bias; struct page *page = rx_buffer->page; @@ -1660,7 +1658,9 @@ static bool i40e_can_reuse_rx_page(struct i40e_rx_buffer *rx_buffer) if (unlikely((page_count(page) - pagecnt_bias) > 1)) return false; #else - if (rx_buffer->page_offset > last_offset) +#define I40E_LAST_OFFSET \ + (SKB_WITH_OVERHEAD(PAGE_SIZE) - I40E_RXBUFFER_2048) + if (rx_buffer->page_offset > I40E_LAST_OFFSET) return false; #endif @@ -1694,7 +1694,7 @@ static void i40e_add_rx_frag(struct i40e_ring *rx_ring, unsigned int size) { #if (PAGE_SIZE < 8192) - unsigned int truesize = I40E_RXBUFFER_2048; + unsigned int truesize = i40e_rx_pg_size(rx_ring) / 2; #else unsigned int truesize = SKB_DATA_ALIGN(size); #endif @@ -1755,7 +1755,7 @@ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, { void *va = page_address(rx_buffer->page) + rx_buffer->page_offset; #if (PAGE_SIZE < 8192) - unsigned int truesize = I40E_RXBUFFER_2048; + unsigned int truesize = i40e_rx_pg_size(rx_ring) / 2; #else unsigned int truesize = SKB_DATA_ALIGN(size); #endif @@ -1821,7 +1821,8 @@ static void i40e_put_rx_buffer(struct i40e_ring *rx_ring, rx_ring->rx_stats.page_reuse_count++; } else { /* we are not reusing the buffer so unmap it */ - dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, PAGE_SIZE, + dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, + i40e_rx_pg_size(rx_ring), DMA_FROM_DEVICE, I40E_RX_DMA_ATTR); __page_frag_cache_drain(rx_buffer->page, rx_buffer->pagecnt_bias); diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h index bc66ec4..2f61853 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h @@ -119,6 +119,7 @@ enum i40e_dyn_idx_t { #define I40E_RXBUFFER_256 256 #define I40E_RXBUFFER_1536 1536 /* 128B aligned standard Ethernet frame */ #define I40E_RXBUFFER_2048 2048 +#define I40E_RXBUFFER_3072 3072 /* Used for large frames w/ padding */ #define I40E_MAX_RXBUFFER 9728 /* largest size for single descriptor */ /* NOTE: netdev_alloc_skb reserves up to 64 bytes, NET_IP_ALIGN means we @@ -389,6 +390,17 @@ struct i40e_ring_container { #define i40e_for_each_ring(pos, head) \ for (pos = (head).ring; pos != NULL; pos = pos->next) +static inline unsigned int i40e_rx_pg_order(struct i40e_ring *ring) +{ +#if (PAGE_SIZE < 8192) + if (ring->rx_buf_len > (PAGE_SIZE / 2)) + return 1; +#endif + return 0; +} + +#define i40e_rx_pg_size(_ring) (PAGE_SIZE << i40e_rx_pg_order(_ring)) + bool i40e_alloc_rx_buffers(struct i40e_ring *rxr, u16 cleaned_count); netdev_tx_t i40e_lan_xmit_frame(struct sk_buff *skb, struct net_device *netdev); void i40e_clean_tx_ring(struct i40e_ring *tx_ring); diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c index 95e383a..6b60c19 100644 --- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c @@ -509,14 +509,15 @@ void i40evf_clean_rx_ring(struct i40e_ring *rx_ring) dma_sync_single_range_for_cpu(rx_ring->dev, rx_bi->dma, rx_bi->page_offset, - I40E_RXBUFFER_2048, + rx_ring->rx_buf_len, DMA_FROM_DEVICE); /* free resources associated with mapping */ dma_unmap_page_attrs(rx_ring->dev, rx_bi->dma, - PAGE_SIZE, + i40e_rx_pg_size(rx_ring), DMA_FROM_DEVICE, I40E_RX_DMA_ATTR); + __page_frag_cache_drain(rx_bi->page, rx_bi->pagecnt_bias); rx_bi->page = NULL; @@ -638,7 +639,7 @@ static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, } /* alloc new page for storage */ - page = dev_alloc_page(); + page = dev_alloc_pages(i40e_rx_pg_order(rx_ring)); if (unlikely(!page)) { rx_ring->rx_stats.alloc_page_failed++; return false; @@ -646,7 +647,7 @@ static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, /* map page for use */ dma = dma_map_page_attrs(rx_ring->dev, page, 0, - PAGE_SIZE, + i40e_rx_pg_size(rx_ring), DMA_FROM_DEVICE, I40E_RX_DMA_ATTR); @@ -654,7 +655,7 @@ static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, * there isn't much point in holding memory we can't use */ if (dma_mapping_error(rx_ring->dev, dma)) { - __free_pages(page, 0); + __free_pages(page, i40e_rx_pg_order(rx_ring)); rx_ring->rx_stats.alloc_page_failed++; return false; } @@ -714,7 +715,7 @@ bool i40evf_alloc_rx_buffers(struct i40e_ring *rx_ring, u16 cleaned_count) /* sync the buffer for use by the device */ dma_sync_single_range_for_device(rx_ring->dev, bi->dma, bi->page_offset, - I40E_RXBUFFER_2048, + rx_ring->rx_buf_len, DMA_FROM_DEVICE); /* Refresh the desc even if buffer_addrs didn't change @@ -1006,9 +1007,6 @@ static inline bool i40e_page_is_reusable(struct page *page) **/ static bool i40e_can_reuse_rx_page(struct i40e_rx_buffer *rx_buffer) { -#if (PAGE_SIZE >= 8192) - unsigned int last_offset = PAGE_SIZE - I40E_RXBUFFER_2048; -#endif unsigned int pagecnt_bias = rx_buffer->pagecnt_bias; struct page *page = rx_buffer->page; @@ -1021,7 +1019,9 @@ static bool i40e_can_reuse_rx_page(struct i40e_rx_buffer *rx_buffer) if (unlikely((page_count(page) - pagecnt_bias) > 1)) return false; #else - if (rx_buffer->page_offset > last_offset) +#define I40E_LAST_OFFSET \ + (SKB_WITH_OVERHEAD(PAGE_SIZE) - I40E_RXBUFFER_2048) + if (rx_buffer->page_offset > I40E_LAST_OFFSET) return false; #endif @@ -1055,7 +1055,7 @@ static void i40e_add_rx_frag(struct i40e_ring *rx_ring, unsigned int size) { #if (PAGE_SIZE < 8192) - unsigned int truesize = I40E_RXBUFFER_2048; + unsigned int truesize = i40e_rx_pg_size(rx_ring) / 2; #else unsigned int truesize = SKB_DATA_ALIGN(size); #endif @@ -1116,7 +1116,7 @@ static struct sk_buff *i40e_construct_skb(struct i40e_ring *rx_ring, { void *va = page_address(rx_buffer->page) + rx_buffer->page_offset; #if (PAGE_SIZE < 8192) - unsigned int truesize = I40E_RXBUFFER_2048; + unsigned int truesize = i40e_rx_pg_size(rx_ring) / 2; #else unsigned int truesize = SKB_DATA_ALIGN(size); #endif @@ -1182,7 +1182,8 @@ static void i40e_put_rx_buffer(struct i40e_ring *rx_ring, rx_ring->rx_stats.page_reuse_count++; } else { /* we are not reusing the buffer so unmap it */ - dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, PAGE_SIZE, + dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, + i40e_rx_pg_size(rx_ring), DMA_FROM_DEVICE, I40E_RX_DMA_ATTR); __page_frag_cache_drain(rx_buffer->page, rx_buffer->pagecnt_bias); diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h index 3bb4d73..dc82f65 100644 --- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h @@ -106,6 +106,7 @@ enum i40e_dyn_idx_t { #define I40E_RXBUFFER_256 256 #define I40E_RXBUFFER_1536 1536 /* 128B aligned standard Ethernet frame */ #define I40E_RXBUFFER_2048 2048 +#define I40E_RXBUFFER_3072 3072 /* Used for large frames w/ padding */ #define I40E_MAX_RXBUFFER 9728 /* largest size for single descriptor */ /* NOTE: netdev_alloc_skb reserves up to 64 bytes, NET_IP_ALIGN means we @@ -376,6 +377,17 @@ struct i40e_ring_container { #define i40e_for_each_ring(pos, head) \ for (pos = (head).ring; pos != NULL; pos = pos->next) +static inline unsigned int i40e_rx_pg_order(struct i40e_ring *ring) +{ +#if (PAGE_SIZE < 8192) + if (ring->rx_buf_len > (PAGE_SIZE / 2)) + return 1; +#endif + return 0; +} + +#define i40e_rx_pg_size(_ring) (PAGE_SIZE << i40e_rx_pg_order(_ring)) + bool i40evf_alloc_rx_buffers(struct i40e_ring *rxr, u16 cleaned_count); netdev_tx_t i40evf_xmit_frame(struct sk_buff *skb, struct net_device *netdev); void i40evf_clean_tx_ring(struct i40e_ring *tx_ring); diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_main.c b/drivers/net/ethernet/intel/i40evf/i40evf_main.c index 2f8f599..95d5c04 100644 --- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c +++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c @@ -694,6 +694,12 @@ static void i40evf_configure_rx(struct i40evf_adapter *adapter) /* Legacy Rx will always default to a 2048 buffer size. */ #if (PAGE_SIZE < 8192) if (!(adapter->flags & I40EVF_FLAG_LEGACY_RX)) { + /* For jumbo frames on systems with 4K pages we have to use + * an order 1 page, so we might as well increase the size + * of our Rx buffer to make better use of the available space + */ + rx_buf_len = I40E_RXBUFFER_3072; + /* We use a 1536 buffer size for configurations with * standard Ethernet mtu. On x86 this gives us enough room * for shared info and 192 bytes of padding. -- 2.9.3