From: Jesse Brandeburg Date: Thu, 21 Apr 2016 02:43:37 +0000 (-0700) Subject: i40e: Refactor receive routine X-Git-Url: https://git.stricted.de/?a=commitdiff_plain;h=1a557afc4dd59b85a5cae2be6d351eaeb31d2664;p=GitHub%2Fmoto-9609%2Fandroid_kernel_motorola_exynos9610.git i40e: Refactor receive routine This is part 1 of the Rx refactor series, just including changes to i40e. This refactor aligns the receive routine with the one in ixgbe which was highly optimized. This reduces the code we have to maintain and allows for (hopefully) more readable and maintainable RX hot path. In order to do this: - consolidate the receive path into a single function that doesn't use packet split but *does* use pages for Rx buffers. - remove the old _1buf routine - consolidate several routines into helper functions - remove ethtool control over packet split Change-ID: I5ca100721de65992aa0114f8b4bac844b84758e0 Signed-off-by: Jesse Brandeburg Tested-by: Andrew Bowers Signed-off-by: Jeff Kirsher --- diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h index ea6a69a1f1d7..ebf423bd5ccb 100644 --- a/drivers/net/ethernet/intel/i40e/i40e.h +++ b/drivers/net/ethernet/intel/i40e/i40e.h @@ -531,7 +531,6 @@ struct i40e_vsi { u8 *rss_lut_user; /* User configured lookup table entries */ u16 max_frame; - u16 rx_hdr_len; u16 rx_buf_len; u8 dtype; diff --git a/drivers/net/ethernet/intel/i40e/i40e_debugfs.c b/drivers/net/ethernet/intel/i40e/i40e_debugfs.c index c0a01e0eb181..8ae30f7f1839 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_debugfs.c +++ b/drivers/net/ethernet/intel/i40e/i40e_debugfs.c @@ -268,9 +268,8 @@ static void i40e_dbg_dump_vsi_seid(struct i40e_pf *pf, int seid) rx_ring->queue_index, rx_ring->reg_idx); dev_info(&pf->pdev->dev, - " rx_rings[%i]: rx_hdr_len = %d, rx_buf_len = %d\n", - i, rx_ring->rx_hdr_len, - rx_ring->rx_buf_len); + " rx_rings[%i]: rx_buf_len = %d\n", + i, rx_ring->rx_buf_len); dev_info(&pf->pdev->dev, " rx_rings[%i]: next_to_use = %d, next_to_clean = %d, ring_active = %i\n", i, @@ -361,8 +360,8 @@ static void i40e_dbg_dump_vsi_seid(struct i40e_pf *pf, int seid) " work_limit = %d\n", vsi->work_limit); dev_info(&pf->pdev->dev, - " max_frame = %d, rx_hdr_len = %d, rx_buf_len = %d dtype = %d\n", - vsi->max_frame, vsi->rx_hdr_len, vsi->rx_buf_len, vsi->dtype); + " max_frame = %d, rx_buf_len = %d dtype = %d\n", + vsi->max_frame, vsi->rx_buf_len, vsi->dtype); dev_info(&pf->pdev->dev, " num_q_vectors = %i, base_vector = %i\n", vsi->num_q_vectors, vsi->base_vector); diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c index 858e1699b87c..6fd730ac23a1 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c +++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c @@ -235,7 +235,6 @@ static const char i40e_priv_flags_strings[][ETH_GSTRING_LEN] = { "LinkPolling", "flow-director-atr", "veb-stats", - "packet-split", "hw-atr-eviction", }; diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index 84e8d4e05924..e46611122179 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -2855,10 +2855,8 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring) memset(&rx_ctx, 0, sizeof(rx_ctx)); ring->rx_buf_len = vsi->rx_buf_len; - ring->rx_hdr_len = vsi->rx_hdr_len; rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT; - rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT; rx_ctx.base = (ring->dma / 128); rx_ctx.qlen = ring->count; @@ -2910,7 +2908,7 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring) ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q); writel(0, ring->tail); - i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring)); + i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring)); return 0; } @@ -2949,15 +2947,13 @@ static int i40e_vsi_configure_rx(struct i40e_vsi *vsi) else vsi->max_frame = I40E_RXBUFFER_2048; - vsi->rx_hdr_len = 0; - vsi->rx_buf_len = vsi->max_frame; + vsi->rx_buf_len = I40E_RXBUFFER_2048; vsi->dtype = I40E_RX_DTYPE_NO_SPLIT; #ifdef I40E_FCOE /* setup rx buffer for FCoE */ if ((vsi->type == I40E_VSI_FCOE) && (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) { - vsi->rx_hdr_len = 0; vsi->rx_buf_len = I40E_RXBUFFER_3072; vsi->max_frame = I40E_RXBUFFER_3072; vsi->dtype = I40E_RX_DTYPE_NO_SPLIT; @@ -2965,8 +2961,6 @@ static int i40e_vsi_configure_rx(struct i40e_vsi *vsi) #endif /* I40E_FCOE */ /* round up for the chip's needs */ - vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len, - BIT_ULL(I40E_RXQ_CTX_HBUFF_SHIFT)); vsi->rx_buf_len = ALIGN(vsi->rx_buf_len, BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT)); @@ -10661,11 +10655,9 @@ static void i40e_print_features(struct i40e_pf *pf) #ifdef CONFIG_PCI_IOV i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs); #endif - i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d RX: %s", + i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d", pf->hw.func_caps.num_vsis, - pf->vsi[pf->lan_vsi]->num_queue_pairs, - "1BUF"); - + pf->vsi[pf->lan_vsi]->num_queue_pairs); if (pf->flags & I40E_FLAG_RSS_ENABLED) i += snprintf(&buf[i], REMAIN(i), " RSS"); if (pf->flags & I40E_FLAG_FD_ATR_ENABLED) diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index 450ecdd589db..b0edffe88492 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -1024,7 +1024,6 @@ err: void i40e_clean_rx_ring(struct i40e_ring *rx_ring) { struct device *dev = rx_ring->dev; - struct i40e_rx_buffer *rx_bi; unsigned long bi_size; u16 i; @@ -1034,30 +1033,20 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring) /* Free all the Rx ring sk_buffs */ for (i = 0; i < rx_ring->count; i++) { - rx_bi = &rx_ring->rx_bi[i]; - if (rx_bi->dma) { - dma_unmap_single(dev, - rx_bi->dma, - rx_ring->rx_buf_len, - DMA_FROM_DEVICE); - rx_bi->dma = 0; - } + 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) { - if (rx_bi->page_dma) { - dma_unmap_page(dev, - rx_bi->page_dma, - PAGE_SIZE, - DMA_FROM_DEVICE); - rx_bi->page_dma = 0; - } - __free_page(rx_bi->page); - rx_bi->page = NULL; - rx_bi->page_offset = 0; - } + if (!rx_bi->page) + continue; + + dma_unmap_page(dev, rx_bi->dma, PAGE_SIZE, DMA_FROM_DEVICE); + __free_pages(rx_bi->page, 0); + + rx_bi->page = NULL; + rx_bi->page_offset = 0; } bi_size = sizeof(struct i40e_rx_buffer) * rx_ring->count; @@ -1066,6 +1055,7 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring) /* Zero out the descriptor ring */ memset(rx_ring->desc, 0, rx_ring->size); + rx_ring->next_to_alloc = 0; rx_ring->next_to_clean = 0; rx_ring->next_to_use = 0; } @@ -1089,37 +1079,6 @@ void i40e_free_rx_resources(struct i40e_ring *rx_ring) } } -/** - * i40e_alloc_rx_headers - allocate rx header buffers - * @rx_ring: ring to alloc buffers - * - * Allocate rx header buffers for the entire ring. As these are static, - * this is only called when setting up a new ring. - **/ -void i40e_alloc_rx_headers(struct i40e_ring *rx_ring) -{ - struct device *dev = rx_ring->dev; - struct i40e_rx_buffer *rx_bi; - dma_addr_t dma; - void *buffer; - int buf_size; - int i; - - if (rx_ring->rx_bi[0].hdr_buf) - return; - /* Make sure the buffers don't cross cache line boundaries. */ - buf_size = ALIGN(rx_ring->rx_hdr_len, 256); - buffer = dma_alloc_coherent(dev, buf_size * rx_ring->count, - &dma, GFP_KERNEL); - if (!buffer) - return; - for (i = 0; i < rx_ring->count; i++) { - rx_bi = &rx_ring->rx_bi[i]; - rx_bi->dma = dma + (i * buf_size); - rx_bi->hdr_buf = buffer + (i * buf_size); - } -} - /** * i40e_setup_rx_descriptors - Allocate Rx descriptors * @rx_ring: Rx descriptor ring (for a specific queue) to setup @@ -1141,9 +1100,7 @@ int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring) u64_stats_init(&rx_ring->syncp); /* Round up to nearest 4K */ - rx_ring->size = ring_is_16byte_desc_enabled(rx_ring) - ? rx_ring->count * sizeof(union i40e_16byte_rx_desc) - : rx_ring->count * sizeof(union i40e_32byte_rx_desc); + rx_ring->size = rx_ring->count * sizeof(union i40e_32byte_rx_desc); rx_ring->size = ALIGN(rx_ring->size, 4096); rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size, &rx_ring->dma, GFP_KERNEL); @@ -1154,6 +1111,7 @@ int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring) goto err; } + rx_ring->next_to_alloc = 0; rx_ring->next_to_clean = 0; rx_ring->next_to_use = 0; @@ -1172,6 +1130,10 @@ err: static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val) { rx_ring->next_to_use = val; + + /* update next to alloc since we have filled the ring */ + rx_ring->next_to_alloc = val; + /* Force memory writes to complete before letting h/w * know there are new descriptors to fetch. (Only * applicable for weak-ordered memory model archs, @@ -1182,160 +1144,122 @@ static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val) } /** - * i40e_alloc_rx_buffers_ps - Replace used receive buffers; packet split - * @rx_ring: ring to place buffers on - * @cleaned_count: number of buffers to replace + * i40e_alloc_mapped_page - recycle or make a new page + * @rx_ring: ring to use + * @bi: rx_buffer struct to modify * - * Returns true if any errors on allocation + * Returns true if the page was successfully allocated or + * reused. **/ -bool i40e_alloc_rx_buffers_ps(struct i40e_ring *rx_ring, u16 cleaned_count) +static bool i40e_alloc_mapped_page(struct i40e_ring *rx_ring, + struct i40e_rx_buffer *bi) { - u16 i = rx_ring->next_to_use; - union i40e_rx_desc *rx_desc; - struct i40e_rx_buffer *bi; - const int current_node = numa_node_id(); + struct page *page = bi->page; + dma_addr_t dma; - /* do nothing if no valid netdev defined */ - if (!rx_ring->netdev || !cleaned_count) - return false; + /* since we are recycling buffers we should seldom need to alloc */ + if (likely(page)) { + rx_ring->rx_stats.page_reuse_count++; + return true; + } - while (cleaned_count--) { - rx_desc = I40E_RX_DESC(rx_ring, i); - bi = &rx_ring->rx_bi[i]; + /* alloc new page for storage */ + page = dev_alloc_page(); + if (unlikely(!page)) { + rx_ring->rx_stats.alloc_page_failed++; + return false; + } - if (bi->skb) /* desc is in use */ - goto no_buffers; + /* map page for use */ + dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE); - /* If we've been moved to a different NUMA node, release the - * page so we can get a new one on the current node. + /* if mapping failed free memory back to system since + * there isn't much point in holding memory we can't use */ - if (bi->page && page_to_nid(bi->page) != current_node) { - dma_unmap_page(rx_ring->dev, - bi->page_dma, - PAGE_SIZE, - DMA_FROM_DEVICE); - __free_page(bi->page); - bi->page = NULL; - bi->page_dma = 0; - rx_ring->rx_stats.realloc_count++; - } else if (bi->page) { - rx_ring->rx_stats.page_reuse_count++; - } - - if (!bi->page) { - bi->page = alloc_page(GFP_ATOMIC); - if (!bi->page) { - rx_ring->rx_stats.alloc_page_failed++; - goto no_buffers; - } - bi->page_dma = dma_map_page(rx_ring->dev, - bi->page, - 0, - PAGE_SIZE, - DMA_FROM_DEVICE); - if (dma_mapping_error(rx_ring->dev, bi->page_dma)) { - rx_ring->rx_stats.alloc_page_failed++; - __free_page(bi->page); - bi->page = NULL; - bi->page_dma = 0; - bi->page_offset = 0; - goto no_buffers; - } - bi->page_offset = 0; - } - - /* Refresh the desc even if buffer_addrs didn't change - * because each write-back erases this info. - */ - rx_desc->read.pkt_addr = - cpu_to_le64(bi->page_dma + bi->page_offset); - rx_desc->read.hdr_addr = cpu_to_le64(bi->dma); - i++; - if (i == rx_ring->count) - i = 0; + if (dma_mapping_error(rx_ring->dev, dma)) { + __free_pages(page, 0); + rx_ring->rx_stats.alloc_page_failed++; + return false; } - if (rx_ring->next_to_use != i) - i40e_release_rx_desc(rx_ring, i); + bi->dma = dma; + bi->page = page; + bi->page_offset = 0; - return false; + return true; +} -no_buffers: - if (rx_ring->next_to_use != i) - i40e_release_rx_desc(rx_ring, i); +/** + * i40e_receive_skb - Send a completed packet up the stack + * @rx_ring: rx ring in play + * @skb: packet to send up + * @vlan_tag: vlan tag for packet + **/ +static void i40e_receive_skb(struct i40e_ring *rx_ring, + struct sk_buff *skb, u16 vlan_tag) +{ + struct i40e_q_vector *q_vector = rx_ring->q_vector; - /* make sure to come back via polling to try again after - * allocation failure - */ - return true; + if ((rx_ring->netdev->features & NETIF_F_HW_VLAN_CTAG_RX) && + (vlan_tag & VLAN_VID_MASK)) + __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_tag); + + napi_gro_receive(&q_vector->napi, skb); } /** - * i40e_alloc_rx_buffers_1buf - Replace used receive buffers; single buffer + * i40e_alloc_rx_buffers - Replace used receive buffers * @rx_ring: ring to place buffers on * @cleaned_count: number of buffers to replace * - * Returns true if any errors on allocation + * Returns false if all allocations were successful, true if any fail **/ -bool i40e_alloc_rx_buffers_1buf(struct i40e_ring *rx_ring, u16 cleaned_count) +bool i40e_alloc_rx_buffers(struct i40e_ring *rx_ring, u16 cleaned_count) { - u16 i = rx_ring->next_to_use; + u16 ntu = rx_ring->next_to_use; union i40e_rx_desc *rx_desc; struct i40e_rx_buffer *bi; - struct sk_buff *skb; /* do nothing if no valid netdev defined */ if (!rx_ring->netdev || !cleaned_count) return false; - while (cleaned_count--) { - rx_desc = I40E_RX_DESC(rx_ring, i); - bi = &rx_ring->rx_bi[i]; - skb = bi->skb; - - if (!skb) { - skb = __netdev_alloc_skb_ip_align(rx_ring->netdev, - rx_ring->rx_buf_len, - GFP_ATOMIC | - __GFP_NOWARN); - if (!skb) { - rx_ring->rx_stats.alloc_buff_failed++; - goto no_buffers; - } - /* initialize queue mapping */ - skb_record_rx_queue(skb, rx_ring->queue_index); - bi->skb = skb; - } + rx_desc = I40E_RX_DESC(rx_ring, ntu); + bi = &rx_ring->rx_bi[ntu]; - if (!bi->dma) { - bi->dma = dma_map_single(rx_ring->dev, - skb->data, - rx_ring->rx_buf_len, - DMA_FROM_DEVICE); - if (dma_mapping_error(rx_ring->dev, bi->dma)) { - rx_ring->rx_stats.alloc_buff_failed++; - bi->dma = 0; - dev_kfree_skb(bi->skb); - bi->skb = NULL; - goto no_buffers; - } - } + do { + if (!i40e_alloc_mapped_page(rx_ring, bi)) + goto no_buffers; - rx_desc->read.pkt_addr = cpu_to_le64(bi->dma); + /* Refresh the desc even if buffer_addrs didn't change + * because each write-back erases this info. + */ + rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset); rx_desc->read.hdr_addr = 0; - i++; - if (i == rx_ring->count) - i = 0; - } - if (rx_ring->next_to_use != i) - i40e_release_rx_desc(rx_ring, i); + rx_desc++; + bi++; + ntu++; + if (unlikely(ntu == rx_ring->count)) { + rx_desc = I40E_RX_DESC(rx_ring, 0); + bi = rx_ring->rx_bi; + ntu = 0; + } + + /* clear the status bits for the next_to_use descriptor */ + rx_desc->wb.qword1.status_error_len = 0; + + cleaned_count--; + } while (cleaned_count); + + if (rx_ring->next_to_use != ntu) + i40e_release_rx_desc(rx_ring, ntu); return false; no_buffers: - if (rx_ring->next_to_use != i) - i40e_release_rx_desc(rx_ring, i); + if (rx_ring->next_to_use != ntu) + i40e_release_rx_desc(rx_ring, ntu); /* make sure to come back via polling to try again after * allocation failure @@ -1343,43 +1267,36 @@ no_buffers: return true; } -/** - * i40e_receive_skb - Send a completed packet up the stack - * @rx_ring: rx ring in play - * @skb: packet to send up - * @vlan_tag: vlan tag for packet - **/ -static void i40e_receive_skb(struct i40e_ring *rx_ring, - struct sk_buff *skb, u16 vlan_tag) -{ - struct i40e_q_vector *q_vector = rx_ring->q_vector; - - if ((rx_ring->netdev->features & NETIF_F_HW_VLAN_CTAG_RX) && - (vlan_tag & VLAN_VID_MASK)) - __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_tag); - - napi_gro_receive(&q_vector->napi, skb); -} - /** * i40e_rx_checksum - Indicate in skb if hw indicated a good cksum * @vsi: the VSI we care about * @skb: skb currently being received and modified - * @rx_status: status value of last descriptor in packet - * @rx_error: error value of last descriptor in packet - * @rx_ptype: ptype value of last descriptor in packet + * @rx_desc: the receive descriptor + * + * skb->protocol must be set before this function is called **/ static inline void i40e_rx_checksum(struct i40e_vsi *vsi, struct sk_buff *skb, - u32 rx_status, - u32 rx_error, - u16 rx_ptype) + union i40e_rx_desc *rx_desc) { - struct i40e_rx_ptype_decoded decoded = decode_rx_desc_ptype(rx_ptype); + struct i40e_rx_ptype_decoded decoded; bool ipv4, ipv6, tunnel = false; + u32 rx_error, rx_status; + u8 ptype; + u64 qword; + + qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len); + ptype = (qword & I40E_RXD_QW1_PTYPE_MASK) >> I40E_RXD_QW1_PTYPE_SHIFT; + rx_error = (qword & I40E_RXD_QW1_ERROR_MASK) >> + I40E_RXD_QW1_ERROR_SHIFT; + rx_status = (qword & I40E_RXD_QW1_STATUS_MASK) >> + I40E_RXD_QW1_STATUS_SHIFT; + decoded = decode_rx_desc_ptype(ptype); skb->ip_summed = CHECKSUM_NONE; + skb_checksum_none_assert(skb); + /* Rx csum enabled and ip headers found? */ if (!(vsi->netdev->features & NETIF_F_RXCSUM)) return; @@ -1445,7 +1362,7 @@ checksum_fail: * * Returns a hash type to be used by skb_set_hash **/ -static inline enum pkt_hash_types i40e_ptype_to_htype(u8 ptype) +static inline int i40e_ptype_to_htype(u8 ptype) { struct i40e_rx_ptype_decoded decoded = decode_rx_desc_ptype(ptype); @@ -1473,7 +1390,7 @@ static inline void i40e_rx_hash(struct i40e_ring *ring, u8 rx_ptype) { u32 hash; - const __le64 rss_mask = + const __le64 rss_mask = cpu_to_le64((u64)I40E_RX_DESC_FLTSTAT_RSS_HASH << I40E_RX_DESC_STATUS_FLTSTAT_SHIFT); @@ -1487,114 +1404,419 @@ static inline void i40e_rx_hash(struct i40e_ring *ring, } /** - * i40e_clean_rx_irq_1buf - Reclaim resources after receive; single buffer - * @rx_ring: rx ring to clean - * @budget: how many cleans we're allowed + * i40e_process_skb_fields - Populate skb header fields from Rx descriptor + * @rx_ring: rx descriptor ring packet is being transacted on + * @rx_desc: pointer to the EOP Rx descriptor + * @skb: pointer to current skb being populated + * @rx_ptype: the packet type decoded by hardware * - * Returns number of packets cleaned + * This function checks the ring, descriptor, and packet information in + * order to populate the hash, checksum, VLAN, protocol, and + * other fields within the skb. **/ -static int i40e_clean_rx_irq_1buf(struct i40e_ring *rx_ring, int budget) +static inline +void i40e_process_skb_fields(struct i40e_ring *rx_ring, + union i40e_rx_desc *rx_desc, struct sk_buff *skb, + u8 rx_ptype) +{ + u64 qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len); + u32 rx_status = (qword & I40E_RXD_QW1_STATUS_MASK) >> + I40E_RXD_QW1_STATUS_SHIFT; + u32 rsyn = (rx_status & I40E_RXD_QW1_STATUS_TSYNINDX_MASK) >> + I40E_RXD_QW1_STATUS_TSYNINDX_SHIFT; + + if (unlikely(rsyn)) { + i40e_ptp_rx_hwtstamp(rx_ring->vsi->back, skb, rsyn); + rx_ring->last_rx_timestamp = jiffies; + } + + i40e_rx_hash(rx_ring, rx_desc, skb, rx_ptype); + + /* modifies the skb - consumes the enet header */ + skb->protocol = eth_type_trans(skb, rx_ring->netdev); + + i40e_rx_checksum(rx_ring->vsi, skb, rx_desc); + + skb_record_rx_queue(skb, rx_ring->queue_index); +} + +/** + * i40e_pull_tail - i40e specific version of skb_pull_tail + * @rx_ring: rx descriptor ring packet is being transacted on + * @skb: pointer to current skb being adjusted + * + * This function is an i40e specific version of __pskb_pull_tail. The + * main difference between this version and the original function is that + * this function can make several assumptions about the state of things + * that allow for significant optimizations versus the standard function. + * As a result we can do things like drop a frag and maintain an accurate + * truesize for the skb. + */ +static void i40e_pull_tail(struct i40e_ring *rx_ring, struct sk_buff *skb) +{ + struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0]; + unsigned char *va; + unsigned int pull_len; + + /* it is valid to use page_address instead of kmap since we are + * working with pages allocated out of the lomem pool per + * alloc_page(GFP_ATOMIC) + */ + va = skb_frag_address(frag); + + /* we need the header to contain the greater of either ETH_HLEN or + * 60 bytes if the skb->len is less than 60 for skb_pad. + */ + pull_len = eth_get_headlen(va, I40E_RX_HDR_SIZE); + + /* align pull length to size of long to optimize memcpy performance */ + skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long))); + + /* update all of the pointers */ + skb_frag_size_sub(frag, pull_len); + frag->page_offset += pull_len; + skb->data_len -= pull_len; + skb->tail += pull_len; +} + +/** + * i40e_cleanup_headers - Correct empty headers + * @rx_ring: rx descriptor ring packet is being transacted on + * @skb: pointer to current skb being fixed + * + * Also address the case where we are pulling data in on pages only + * and as such no data is present in the skb header. + * + * In addition if skb is not at least 60 bytes we need to pad it so that + * it is large enough to qualify as a valid Ethernet frame. + * + * Returns true if an error was encountered and skb was freed. + **/ +static bool i40e_cleanup_headers(struct i40e_ring *rx_ring, struct sk_buff *skb) +{ + /* place header in linear portion of buffer */ + if (skb_is_nonlinear(skb)) + i40e_pull_tail(rx_ring, skb); + + /* if eth_skb_pad returns an error the skb was freed */ + if (eth_skb_pad(skb)) + return true; + + return false; +} + +/** + * i40e_reuse_rx_page - page flip buffer and store it back on the ring + * @rx_ring: rx descriptor ring to store buffers on + * @old_buff: donor buffer to have page reused + * + * Synchronizes page for reuse by the adapter + **/ +static void i40e_reuse_rx_page(struct i40e_ring *rx_ring, + struct i40e_rx_buffer *old_buff) +{ + struct i40e_rx_buffer *new_buff; + u16 nta = rx_ring->next_to_alloc; + + new_buff = &rx_ring->rx_bi[nta]; + + /* update, and store next to alloc */ + nta++; + rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0; + + /* transfer page from old buffer to new buffer */ + *new_buff = *old_buff; +} + +/** + * i40e_page_is_reserved - check if reuse is possible + * @page: page struct to check + */ +static inline bool i40e_page_is_reserved(struct page *page) +{ + return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page); +} + +/** + * i40e_add_rx_frag - Add contents of Rx buffer to sk_buff + * @rx_ring: rx descriptor ring to transact packets on + * @rx_buffer: buffer containing page to add + * @rx_desc: descriptor containing length of buffer written by hardware + * @skb: sk_buff to place the data into + * + * This function will add the data contained in rx_buffer->page to the skb. + * This is done either through a direct copy if the data in the buffer is + * less than the skb header size, otherwise it will just attach the page as + * a frag to the skb. + * + * The function will then update the page offset if necessary and return + * true if the buffer can be reused by the adapter. + **/ +static bool i40e_add_rx_frag(struct i40e_ring *rx_ring, + struct i40e_rx_buffer *rx_buffer, + union i40e_rx_desc *rx_desc, + struct sk_buff *skb) +{ + struct page *page = rx_buffer->page; + u64 qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len); + unsigned int size = (qword & I40E_RXD_QW1_LENGTH_PBUF_MASK) >> + I40E_RXD_QW1_LENGTH_PBUF_SHIFT; +#if (PAGE_SIZE < 8192) + unsigned int truesize = I40E_RXBUFFER_2048; +#else + unsigned int truesize = ALIGN(size, L1_CACHE_BYTES); + unsigned int last_offset = PAGE_SIZE - I40E_RXBUFFER_2048; +#endif + + /* will the data fit in the skb we allocated? if so, just + * copy it as it is pretty small anyway + */ + if ((size <= I40E_RX_HDR_SIZE) && !skb_is_nonlinear(skb)) { + unsigned char *va = page_address(page) + rx_buffer->page_offset; + + memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long))); + + /* page is not reserved, we can reuse buffer as-is */ + if (likely(!i40e_page_is_reserved(page))) + return true; + + /* this page cannot be reused so discard it */ + __free_pages(page, 0); + return false; + } + + skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page, + rx_buffer->page_offset, size, truesize); + + /* avoid re-using remote pages */ + if (unlikely(i40e_page_is_reserved(page))) + return false; + +#if (PAGE_SIZE < 8192) + /* if we are only owner of page we can reuse it */ + if (unlikely(page_count(page) != 1)) + return false; + + /* flip page offset to other buffer */ + rx_buffer->page_offset ^= truesize; +#else + /* move offset up to the next cache line */ + rx_buffer->page_offset += truesize; + + if (rx_buffer->page_offset > last_offset) + return false; +#endif + + /* Even if we own the page, we are not allowed to use atomic_set() + * This would break get_page_unless_zero() users. + */ + get_page(rx_buffer->page); + + return true; +} + +/** + * i40e_fetch_rx_buffer - Allocate skb and populate it + * @rx_ring: rx descriptor ring to transact packets on + * @rx_desc: descriptor containing info written by hardware + * + * This function allocates an skb on the fly, and populates it with the page + * data from the current receive descriptor, taking care to set up the skb + * correctly, as well as handling calling the page recycle function if + * necessary. + */ +static inline +struct sk_buff *i40e_fetch_rx_buffer(struct i40e_ring *rx_ring, + union i40e_rx_desc *rx_desc) +{ + 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; + + /* prefetch first cache line of first page */ + prefetch(page_addr); +#if L1_CACHE_BYTES < 128 + prefetch(page_addr + L1_CACHE_BYTES); +#endif + + /* allocate a skb to store the frags */ + skb = __napi_alloc_skb(&rx_ring->q_vector->napi, + I40E_RX_HDR_SIZE, + GFP_ATOMIC | __GFP_NOWARN); + if (unlikely(!skb)) { + rx_ring->rx_stats.alloc_buff_failed++; + return NULL; + } + + /* we will be copying header into skb->data in + * pskb_may_pull so it is in our interest to prefetch + * 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 */ + dma_sync_single_range_for_cpu(rx_ring->dev, + rx_buffer->dma, + rx_buffer->page_offset, + I40E_RXBUFFER_2048, + DMA_FROM_DEVICE); + + /* pull page into skb */ + if (i40e_add_rx_frag(rx_ring, rx_buffer, rx_desc, skb)) { + /* hand second half of page back to the ring */ + i40e_reuse_rx_page(rx_ring, rx_buffer); + rx_ring->rx_stats.page_reuse_count++; + } else { + /* we are not reusing the buffer so unmap it */ + dma_unmap_page(rx_ring->dev, rx_buffer->dma, PAGE_SIZE, + DMA_FROM_DEVICE); + } + + /* clear contents of buffer_info */ + rx_buffer->page = NULL; + + return skb; +} + +/** + * i40e_is_non_eop - process handling of non-EOP buffers + * @rx_ring: Rx ring being processed + * @rx_desc: Rx descriptor for current buffer + * @skb: Current socket buffer containing buffer in progress + * + * This function updates next to clean. If the buffer is an EOP buffer + * this function exits returning false, otherwise it will place the + * sk_buff in the next buffer to be chained and return true indicating + * that this is in fact a non-EOP buffer. + **/ +static bool i40e_is_non_eop(struct i40e_ring *rx_ring, + union i40e_rx_desc *rx_desc, + struct sk_buff *skb) +{ + u32 ntc = rx_ring->next_to_clean + 1; + + /* fetch, update, and store next to clean */ + ntc = (ntc < rx_ring->count) ? ntc : 0; + rx_ring->next_to_clean = ntc; + + prefetch(I40E_RX_DESC(rx_ring, ntc)); + +#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 */ +#define I40E_RXD_EOF BIT(I40E_RX_DESC_STATUS_EOF_SHIFT) + 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; +} + +/** + * i40e_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf + * @rx_ring: rx descriptor ring to transact packets on + * @budget: Total limit on number of packets to process + * + * This function provides a "bounce buffer" approach to Rx interrupt + * processing. The advantage to this is that on systems that have + * expensive overhead for IOMMU access this provides a means of avoiding + * it by maintaining the mapping of the page to the system. + * + * Returns amount of work completed + **/ +static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget) { unsigned int total_rx_bytes = 0, total_rx_packets = 0; u16 cleaned_count = I40E_DESC_UNUSED(rx_ring); - struct i40e_vsi *vsi = rx_ring->vsi; - union i40e_rx_desc *rx_desc; - u32 rx_error, rx_status; - u16 rx_packet_len; bool failure = false; - u8 rx_ptype; - u64 qword; - u16 i; - do { - struct i40e_rx_buffer *rx_bi; + while (likely(total_rx_packets < budget)) { + union i40e_rx_desc *rx_desc; struct sk_buff *skb; + u32 rx_status; u16 vlan_tag; + u8 rx_ptype; + u64 qword; + /* return some buffers to hardware, one at a time is too slow */ if (cleaned_count >= I40E_RX_BUFFER_WRITE) { failure = failure || - i40e_alloc_rx_buffers_1buf(rx_ring, - cleaned_count); + i40e_alloc_rx_buffers(rx_ring, cleaned_count); cleaned_count = 0; } - i = rx_ring->next_to_clean; - rx_desc = I40E_RX_DESC(rx_ring, i); + rx_desc = I40E_RX_DESC(rx_ring, rx_ring->next_to_clean); + qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len); + rx_ptype = (qword & I40E_RXD_QW1_PTYPE_MASK) >> + I40E_RXD_QW1_PTYPE_SHIFT; rx_status = (qword & I40E_RXD_QW1_STATUS_MASK) >> - I40E_RXD_QW1_STATUS_SHIFT; + I40E_RXD_QW1_STATUS_SHIFT; if (!(rx_status & BIT(I40E_RX_DESC_STATUS_DD_SHIFT))) break; + /* status_error_len will always be zero for unused descriptors + * because it's cleared in cleanup, and overlaps with hdr_addr + * which is always zero because packet split isn't used, if the + * hardware wrote DD then it will be non-zero + */ + if (!rx_desc->wb.qword1.status_error_len) + break; + /* This memory barrier is needed to keep us from reading * any other fields out of the rx_desc until we know the * DD bit is set. */ dma_rmb(); - if (i40e_rx_is_programming_status(qword)) { - i40e_clean_programming_status(rx_ring, rx_desc); - I40E_RX_INCREMENT(rx_ring, i); - continue; - } - rx_bi = &rx_ring->rx_bi[i]; - skb = rx_bi->skb; - prefetch(skb->data); - - rx_packet_len = (qword & I40E_RXD_QW1_LENGTH_PBUF_MASK) >> - I40E_RXD_QW1_LENGTH_PBUF_SHIFT; - - rx_error = (qword & I40E_RXD_QW1_ERROR_MASK) >> - I40E_RXD_QW1_ERROR_SHIFT; - rx_error &= ~BIT(I40E_RX_DESC_ERROR_HBO_SHIFT); + skb = i40e_fetch_rx_buffer(rx_ring, rx_desc); + if (!skb) + break; - rx_ptype = (qword & I40E_RXD_QW1_PTYPE_MASK) >> - I40E_RXD_QW1_PTYPE_SHIFT; - rx_bi->skb = NULL; cleaned_count++; - /* Get the header and possibly the whole packet - * If this is an skb from previous receive dma will be 0 - */ - skb_put(skb, rx_packet_len); - dma_unmap_single(rx_ring->dev, rx_bi->dma, rx_ring->rx_buf_len, - DMA_FROM_DEVICE); - rx_bi->dma = 0; - - I40E_RX_INCREMENT(rx_ring, i); - - if (unlikely( - !(rx_status & BIT(I40E_RX_DESC_STATUS_EOF_SHIFT)))) { - rx_ring->rx_stats.non_eop_descs++; + if (i40e_is_non_eop(rx_ring, rx_desc, skb)) continue; - } - /* ERR_MASK will only have valid bits if EOP set */ - if (unlikely(rx_error & BIT(I40E_RX_DESC_ERROR_RXE_SHIFT))) { + /* ERR_MASK will only have valid bits if EOP set, and + * what we are doing here is actually checking + * I40E_RX_DESC_ERROR_RXE_SHIFT, since it is the zeroth bit in + * the error field + */ + if (unlikely(i40e_test_staterr(rx_desc, BIT(I40E_RXD_QW1_ERROR_SHIFT)))) { dev_kfree_skb_any(skb); continue; } - i40e_rx_hash(rx_ring, rx_desc, skb, rx_ptype); - if (unlikely(rx_status & I40E_RXD_QW1_STATUS_TSYNVALID_MASK)) { - i40e_ptp_rx_hwtstamp(vsi->back, skb, (rx_status & - I40E_RXD_QW1_STATUS_TSYNINDX_MASK) >> - I40E_RXD_QW1_STATUS_TSYNINDX_SHIFT); - rx_ring->last_rx_timestamp = jiffies; - } + if (i40e_cleanup_headers(rx_ring, skb)) + continue; /* probably a little skewed due to removing CRC */ total_rx_bytes += skb->len; - total_rx_packets++; - skb->protocol = eth_type_trans(skb, rx_ring->netdev); + /* populate checksum, VLAN, and protocol */ + i40e_process_skb_fields(rx_ring, rx_desc, skb, rx_ptype); - i40e_rx_checksum(vsi, skb, rx_status, rx_error, rx_ptype); - - vlan_tag = rx_status & BIT(I40E_RX_DESC_STATUS_L2TAG1P_SHIFT) - ? le16_to_cpu(rx_desc->wb.qword0.lo_dword.l2tag1) - : 0; #ifdef I40E_FCOE if (unlikely( i40e_rx_is_fcoe(rx_ptype) && @@ -1603,10 +1825,15 @@ static int i40e_clean_rx_irq_1buf(struct i40e_ring *rx_ring, int budget) continue; } #endif + + vlan_tag = (qword & BIT(I40E_RX_DESC_STATUS_L2TAG1P_SHIFT)) ? + le16_to_cpu(rx_desc->wb.qword0.lo_dword.l2tag1) : 0; + i40e_receive_skb(rx_ring, skb, vlan_tag); - rx_desc->wb.qword1.status_error_len = 0; - } while (likely(total_rx_packets < budget)); + /* update budget accounting */ + total_rx_packets++; + } u64_stats_update_begin(&rx_ring->syncp); rx_ring->stats.packets += total_rx_packets; @@ -1615,6 +1842,7 @@ static int i40e_clean_rx_irq_1buf(struct i40e_ring *rx_ring, int budget) rx_ring->q_vector->rx.total_packets += total_rx_packets; rx_ring->q_vector->rx.total_bytes += total_rx_bytes; + /* guarantee a trip back through this routine if there was a failure */ return failure ? budget : total_rx_packets; } @@ -1759,9 +1987,7 @@ int i40e_napi_poll(struct napi_struct *napi, int budget) budget_per_ring = max(budget/q_vector->num_ringpairs, 1); i40e_for_each_ring(ring, q_vector->rx) { - int cleaned; - - cleaned = i40e_clean_rx_irq_1buf(ring, budget_per_ring); + int cleaned = i40e_clean_rx_irq(ring, budget_per_ring); work_done += cleaned; /* if we clean as many as budgeted, we must not be done */ diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h index 03e21d95c4f1..37643e6b3afd 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h @@ -102,8 +102,8 @@ enum i40e_dyn_idx_t { (((pf)->flags & I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE) ? \ I40E_DEFAULT_RSS_HENA_EXPANDED : I40E_DEFAULT_RSS_HENA) -/* Supported Rx Buffer Sizes */ -#define I40E_RXBUFFER_512 512 /* Used for packet split */ +/* Supported Rx Buffer Sizes (a multiple of 128) */ +#define I40E_RXBUFFER_256 256 #define I40E_RXBUFFER_2048 2048 #define I40E_RXBUFFER_3072 3072 /* For FCoE MTU of 2158 */ #define I40E_RXBUFFER_4096 4096 @@ -114,9 +114,28 @@ enum i40e_dyn_idx_t { * reserve 2 more, and skb_shared_info adds an additional 384 bytes more, * this adds up to 512 bytes of extra data meaning the smallest allocation * we could have is 1K. - * i.e. RXBUFFER_512 --> size-1024 slab + * i.e. RXBUFFER_256 --> 960 byte skb (size-1024 slab) + * i.e. RXBUFFER_512 --> 1216 byte skb (size-2048 slab) */ -#define I40E_RX_HDR_SIZE I40E_RXBUFFER_512 +#define I40E_RX_HDR_SIZE I40E_RXBUFFER_256 +#define i40e_rx_desc i40e_32byte_rx_desc + +/** + * i40e_test_staterr - tests bits in Rx descriptor status and error fields + * @rx_desc: pointer to receive descriptor (in le64 format) + * @stat_err_bits: value to mask + * + * This function does some fast chicanery in order to return the + * value of the mask which is really only used for boolean tests. + * The status_error_len doesn't need to be shifted because it begins + * at offset zero. + */ +static inline bool i40e_test_staterr(union i40e_rx_desc *rx_desc, + const u64 stat_err_bits) +{ + return !!(rx_desc->wb.qword1.status_error_len & + cpu_to_le64(stat_err_bits)); +} /* How many Rx Buffers do we bundle into one write to the hardware ? */ #define I40E_RX_BUFFER_WRITE 16 /* Must be power of 2 */ @@ -142,8 +161,6 @@ enum i40e_dyn_idx_t { prefetch((n)); \ } while (0) -#define i40e_rx_desc i40e_32byte_rx_desc - #define I40E_MAX_BUFFER_TXD 8 #define I40E_MIN_TX_LEN 17 @@ -213,10 +230,8 @@ struct i40e_tx_buffer { struct i40e_rx_buffer { struct sk_buff *skb; - void *hdr_buf; dma_addr_t dma; struct page *page; - dma_addr_t page_dma; unsigned int page_offset; }; @@ -280,7 +295,6 @@ struct i40e_ring { u16 count; /* Number of descriptors */ u16 reg_idx; /* HW register index of the ring */ - u16 rx_hdr_len; u16 rx_buf_len; #define I40E_RX_DTYPE_NO_SPLIT 0 #define I40E_RX_DTYPE_HEADER_SPLIT 1 @@ -322,6 +336,7 @@ struct i40e_ring { struct i40e_q_vector *q_vector; /* Backreference to associated vector */ struct rcu_head rcu; /* to avoid race on free */ + u16 next_to_alloc; } ____cacheline_internodealigned_in_smp; enum i40e_latency_range { @@ -345,9 +360,7 @@ struct i40e_ring_container { #define i40e_for_each_ring(pos, head) \ for (pos = (head).ring; pos != NULL; pos = pos->next) -bool i40e_alloc_rx_buffers_ps(struct i40e_ring *rxr, u16 cleaned_count); -bool i40e_alloc_rx_buffers_1buf(struct i40e_ring *rxr, u16 cleaned_count); -void i40e_alloc_rx_headers(struct i40e_ring *rxr); +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); void i40e_clean_rx_ring(struct i40e_ring *rx_ring);