From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bimmy Pujari Date: Mon, 28 Nov 2016 16:06:08 -0800 Subject: [Intel-wired-lan] [next PATCH S55 11/14] i40e/i40evf: Moves skb from i40e_rx_buffer to i40e_ring In-Reply-To: <1480377971-23412-1-git-send-email-bimmy.pujari@intel.com> References: <1480377971-23412-1-git-send-email-bimmy.pujari@intel.com> Message-ID: <1480377971-23412-12-git-send-email-bimmy.pujari@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: "Scott Peterson" This patch reduces the size of struct i40e_rx_buffer by one pointer, and makes the i40e driver a little more consistent with the igb driver in terms of packets that span buffers. We do this by moving the skb field from struct i40e_rx_buffer to struct i40e_ring. We pass the skb we already have (or NULL if we don't) to i40e_fetch_rx_buffer(), which skips the skb allocation if we already have one for this packet. Signed-off-by: Scott Peterson Change-ID: I4ad48a531844494ba0c5d8e1a62209a057f661b0 --- Testing Hints: Generate traffic with jumbo frames (e.g. netperf), and observe "non_eop_segs" stat increases on receiver. That confirms the path changed here was taken. drivers/net/ethernet/intel/i40e/i40e_txrx.c | 31 +++++++++++++-------------- drivers/net/ethernet/intel/i40e/i40e_txrx.h | 9 +++++++- drivers/net/ethernet/intel/i40evf/i40e_txrx.c | 30 +++++++++++++------------- drivers/net/ethernet/intel/i40evf/i40e_txrx.h | 9 +++++++- 4 files changed, 46 insertions(+), 33 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index 7b96ccb..43de4c4 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -1013,14 +1013,15 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring) if (!rx_ring->rx_bi) return; + if (rx_ring->skb) { + dev_kfree_skb(rx_ring->skb); + rx_ring->skb = NULL; + } + /* Free all the Rx ring sk_buffs */ for (i = 0; i < rx_ring->count; i++) { struct i40e_rx_buffer *rx_bi = &rx_ring->rx_bi[i]; - if (rx_bi->skb) { - dev_kfree_skb(rx_bi->skb); - rx_bi->skb = NULL; - } if (!rx_bi->page) continue; @@ -1608,7 +1609,8 @@ static bool i40e_add_rx_frag(struct i40e_ring *rx_ring, */ static inline struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring, - union i40e_rx_desc *rx_desc) + union i40e_rx_desc *rx_desc, + struct sk_buff *skb) { u64 local_status_error_len = le64_to_cpu(rx_desc->wb.qword1.status_error_len); @@ -1616,15 +1618,12 @@ struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring, (local_status_error_len & I40E_RXD_QW1_LENGTH_PBUF_MASK) >> I40E_RXD_QW1_LENGTH_PBUF_SHIFT; struct i40e_rx_buffer *rx_buffer; - struct sk_buff *skb; struct page *page; rx_buffer = &rx_ring->rx_bi[rx_ring->next_to_clean]; page = rx_buffer->page; prefetchw(page); - skb = rx_buffer->skb; - if (likely(!skb)) { void *page_addr = page_address(page) + rx_buffer->page_offset; @@ -1648,8 +1647,6 @@ struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring, * it now to avoid a possible cache miss */ prefetchw(skb->data); - } else { - rx_buffer->skb = NULL; } /* we are reusing so sync this buffer for CPU use */ @@ -1702,7 +1699,6 @@ static bool i40e_is_non_eop(struct i40e_ring *rx_ring, #define staterrlen rx_desc->wb.qword1.status_error_len if (unlikely(i40e_rx_is_programming_status(le64_to_cpu(staterrlen)))) { i40e_clean_programming_status(rx_ring, rx_desc); - rx_ring->rx_bi[ntc].skb = skb; return true; } /* if we are the last buffer then there is nothing else to do */ @@ -1710,8 +1706,6 @@ static bool i40e_is_non_eop(struct i40e_ring *rx_ring, if (likely(i40e_test_staterr(rx_desc, I40E_RXD_EOF))) return false; - /* place skb in next buffer to be received */ - rx_ring->rx_bi[ntc].skb = skb; rx_ring->rx_stats.non_eop_descs++; return true; @@ -1732,12 +1726,12 @@ static bool i40e_is_non_eop(struct i40e_ring *rx_ring, static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) { unsigned int total_rx_bytes = 0, total_rx_packets = 0; + struct sk_buff *skb = rx_ring->skb; u16 cleaned_count = I40E_DESC_UNUSED(rx_ring); bool failure = false; while (likely(total_rx_packets < budget)) { union i40e_rx_desc *rx_desc; - struct sk_buff *skb; u16 vlan_tag; u8 rx_ptype; u64 qword; @@ -1766,7 +1760,7 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) */ dma_rmb(); - skb = i40e_fetch_rx_buffer(rx_ring, rx_desc); + skb = i40e_fetch_rx_buffer(rx_ring, rx_desc, skb); if (!skb) break; @@ -1785,8 +1779,10 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) continue; } - if (i40e_cleanup_headers(rx_ring, skb)) + if (i40e_cleanup_headers(rx_ring, skb)) { + skb = NULL; continue; + } /* probably a little skewed due to removing CRC */ total_rx_bytes += skb->len; @@ -1811,11 +1807,14 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) le16_to_cpu(rx_desc->wb.qword0.lo_dword.l2tag1) : 0; i40e_receive_skb(rx_ring, skb, vlan_tag); + skb = NULL; /* update budget accounting */ total_rx_packets++; } + rx_ring->skb = skb; + u64_stats_update_begin(&rx_ring->syncp); rx_ring->stats.packets += total_rx_packets; rx_ring->stats.bytes += total_rx_bytes; diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h index 1ea820e..f809790 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h @@ -253,7 +253,6 @@ struct i40e_tx_buffer { }; struct i40e_rx_buffer { - struct sk_buff *skb; dma_addr_t dma; struct page *page; unsigned int page_offset; @@ -354,6 +353,14 @@ struct i40e_ring { struct rcu_head rcu; /* to avoid race on free */ u16 next_to_alloc; + struct sk_buff *skb; /* When i40e_clean_rx_ring_irq() must + * return before it sees the EOP for + * the current packet, we save that skb + * here and resume receiving this + * packet the next time + * i40e_clean_rx_ring_irq() is called + * for this ring. + */ } ____cacheline_internodealigned_in_smp; enum i40e_latency_range { diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c index bb498c3..04c266c 100644 --- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c @@ -501,14 +501,15 @@ void i40evf_clean_rx_ring(struct i40e_ring *rx_ring) if (!rx_ring->rx_bi) return; + if (rx_ring->skb) { + dev_kfree_skb(rx_ring->skb); + rx_ring->skb = NULL; + } + /* Free all the Rx ring sk_buffs */ for (i = 0; i < rx_ring->count; i++) { struct i40e_rx_buffer *rx_bi = &rx_ring->rx_bi[i]; - if (rx_bi->skb) { - dev_kfree_skb(rx_bi->skb); - rx_bi->skb = NULL; - } if (!rx_bi->page) continue; @@ -1086,7 +1087,8 @@ static bool i40e_add_rx_frag(struct i40e_ring *rx_ring, */ static inline struct sk_buff *i40evf_fetch_rx_buffer(struct i40e_ring *rx_ring, - union i40e_rx_desc *rx_desc) + union i40e_rx_desc *rx_desc, + struct sk_buff *skb) { u64 local_status_error_len = le64_to_cpu(rx_desc->wb.qword1.status_error_len); @@ -1094,15 +1096,12 @@ struct sk_buff *i40evf_fetch_rx_buffer(struct i40e_ring *rx_ring, (local_status_error_len & I40E_RXD_QW1_LENGTH_PBUF_MASK) >> I40E_RXD_QW1_LENGTH_PBUF_SHIFT; struct i40e_rx_buffer *rx_buffer; - struct sk_buff *skb; struct page *page; rx_buffer = &rx_ring->rx_bi[rx_ring->next_to_clean]; page = rx_buffer->page; prefetchw(page); - skb = rx_buffer->skb; - if (likely(!skb)) { void *page_addr = page_address(page) + rx_buffer->page_offset; @@ -1126,8 +1125,6 @@ struct sk_buff *i40evf_fetch_rx_buffer(struct i40e_ring *rx_ring, * it now to avoid a possible cache miss */ prefetchw(skb->data); - } else { - rx_buffer->skb = NULL; } /* we are reusing so sync this buffer for CPU use */ @@ -1182,8 +1179,6 @@ static bool i40e_is_non_eop(struct i40e_ring *rx_ring, if (likely(i40e_test_staterr(rx_desc, I40E_RXD_EOF))) return false; - /* place skb in next buffer to be received */ - rx_ring->rx_bi[ntc].skb = skb; rx_ring->rx_stats.non_eop_descs++; return true; @@ -1204,12 +1199,12 @@ static bool i40e_is_non_eop(struct i40e_ring *rx_ring, static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) { unsigned int total_rx_bytes = 0, total_rx_packets = 0; + struct sk_buff *skb = rx_ring->skb; u16 cleaned_count = I40E_DESC_UNUSED(rx_ring); bool failure = false; while (likely(total_rx_packets < budget)) { union i40e_rx_desc *rx_desc; - struct sk_buff *skb; u16 vlan_tag; u8 rx_ptype; u64 qword; @@ -1238,7 +1233,7 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) */ dma_rmb(); - skb = i40evf_fetch_rx_buffer(rx_ring, rx_desc); + skb = i40evf_fetch_rx_buffer(rx_ring, rx_desc, skb); if (!skb) break; @@ -1257,8 +1252,10 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) continue; } - if (i40e_cleanup_headers(rx_ring, skb)) + if (i40e_cleanup_headers(rx_ring, skb)) { + skb = NULL; continue; + } /* probably a little skewed due to removing CRC */ total_rx_bytes += skb->len; @@ -1275,11 +1272,14 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) le16_to_cpu(rx_desc->wb.qword0.lo_dword.l2tag1) : 0; i40e_receive_skb(rx_ring, skb, vlan_tag); + skb = NULL; /* update budget accounting */ total_rx_packets++; } + rx_ring->skb = skb; + u64_stats_update_begin(&rx_ring->syncp); rx_ring->stats.packets += total_rx_packets; rx_ring->stats.bytes += total_rx_bytes; diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h index a5fc789..8274ba6 100644 --- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h @@ -239,7 +239,6 @@ struct i40e_tx_buffer { }; struct i40e_rx_buffer { - struct sk_buff *skb; dma_addr_t dma; struct page *page; unsigned int page_offset; @@ -340,6 +339,14 @@ struct i40e_ring { struct rcu_head rcu; /* to avoid race on free */ u16 next_to_alloc; + struct sk_buff *skb; /* When i40evf_clean_rx_ring_irq() must + * return before it sees the EOP for + * the current packet, we save that skb + * here and resume receiving this + * packet the next time + * i40evf_clean_rx_ring_irq() is called + * for this ring. + */ } ____cacheline_internodealigned_in_smp; enum i40e_latency_range { -- 2.4.11