/******************************************************************************* * * Intel Ethernet Controller XL710 Family Linux Virtual Function Driver * Copyright(c) 2013 - 2016 Intel Corporation. * * This program is free software; you can redistribute it and/or modify it * under the terms and conditions of the GNU General Public License, * version 2, as published by the Free Software Foundation. * * This program is distributed in the hope it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for * more details. * * You should have received a copy of the GNU General Public License along * with this program. If not, see . * * The full GNU General Public License is included in this distribution in * the file called "COPYING". * * Contact Information: * e1000-devel Mailing List * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497 * ******************************************************************************/ #include #include #include "i40evf.h" #include "i40e_prototype.h" static inline __le64 build_ctob(u32 td_cmd, u32 td_offset, unsigned int size, u32 td_tag) { return cpu_to_le64(I40E_TX_DESC_DTYPE_DATA | ((u64)td_cmd << I40E_TXD_QW1_CMD_SHIFT) | ((u64)td_offset << I40E_TXD_QW1_OFFSET_SHIFT) | ((u64)size << I40E_TXD_QW1_TX_BUF_SZ_SHIFT) | ((u64)td_tag << I40E_TXD_QW1_L2TAG1_SHIFT)); } #define I40E_TXD_CMD (I40E_TX_DESC_CMD_EOP | I40E_TX_DESC_CMD_RS) /** * i40e_unmap_and_free_tx_resource - Release a Tx buffer * @ring: the ring that owns the buffer * @tx_buffer: the buffer to free **/ static void i40e_unmap_and_free_tx_resource(struct i40e_ring *ring, struct i40e_tx_buffer *tx_buffer) { if (tx_buffer->skb) { dev_kfree_skb_any(tx_buffer->skb); if (dma_unmap_len(tx_buffer, len)) dma_unmap_single(ring->dev, dma_unmap_addr(tx_buffer, dma), dma_unmap_len(tx_buffer, len), DMA_TO_DEVICE); } else if (dma_unmap_len(tx_buffer, len)) { dma_unmap_page(ring->dev, dma_unmap_addr(tx_buffer, dma), dma_unmap_len(tx_buffer, len), DMA_TO_DEVICE); } if (tx_buffer->tx_flags & I40E_TX_FLAGS_FD_SB) kfree(tx_buffer->raw_buf); tx_buffer->next_to_watch = NULL; tx_buffer->skb = NULL; dma_unmap_len_set(tx_buffer, len, 0); /* tx_buffer must be completely set up in the transmit path */ } /** * i40evf_clean_tx_ring - Free any empty Tx buffers * @tx_ring: ring to be cleaned **/ void i40evf_clean_tx_ring(struct i40e_ring *tx_ring) { unsigned long bi_size; u16 i; /* ring already cleared, nothing to do */ if (!tx_ring->tx_bi) return; /* Free all the Tx ring sk_buffs */ for (i = 0; i < tx_ring->count; i++) i40e_unmap_and_free_tx_resource(tx_ring, &tx_ring->tx_bi[i]); bi_size = sizeof(struct i40e_tx_buffer) * tx_ring->count; memset(tx_ring->tx_bi, 0, bi_size); /* Zero out the descriptor ring */ memset(tx_ring->desc, 0, tx_ring->size); tx_ring->next_to_use = 0; tx_ring->next_to_clean = 0; if (!tx_ring->netdev) return; /* cleanup Tx queue statistics */ netdev_tx_reset_queue(netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index)); } /** * i40evf_free_tx_resources - Free Tx resources per queue * @tx_ring: Tx descriptor ring for a specific queue * * Free all transmit software resources **/ void i40evf_free_tx_resources(struct i40e_ring *tx_ring) { i40evf_clean_tx_ring(tx_ring); kfree(tx_ring->tx_bi); tx_ring->tx_bi = NULL; if (tx_ring->desc) { dma_free_coherent(tx_ring->dev, tx_ring->size, tx_ring->desc, tx_ring->dma); tx_ring->desc = NULL; } } /** * i40evf_get_tx_pending - how many Tx descriptors not processed * @tx_ring: the ring of descriptors * @in_sw: is tx_pending being checked in SW or HW * * Since there is no access to the ring head register * in XL710, we need to use our local copies **/ u32 i40evf_get_tx_pending(struct i40e_ring *ring, bool in_sw) { u32 head, tail; if (!in_sw) head = i40e_get_head(ring); else head = ring->next_to_clean; tail = readl(ring->tail); if (head != tail) return (head < tail) ? tail - head : (tail + ring->count - head); return 0; } #define WB_STRIDE 0x3 /** * i40e_clean_tx_irq - Reclaim resources after transmit completes * @tx_ring: tx ring to clean * @budget: how many cleans we're allowed * * Returns true if there's any budget left (e.g. the clean is finished) **/ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget) { u16 i = tx_ring->next_to_clean; struct i40e_tx_buffer *tx_buf; struct i40e_tx_desc *tx_head; struct i40e_tx_desc *tx_desc; unsigned int total_packets = 0; unsigned int total_bytes = 0; tx_buf = &tx_ring->tx_bi[i]; tx_desc = I40E_TX_DESC(tx_ring, i); i -= tx_ring->count; tx_head = I40E_TX_DESC(tx_ring, i40e_get_head(tx_ring)); do { struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch; /* if next_to_watch is not set then there is no work pending */ if (!eop_desc) break; /* prevent any other reads prior to eop_desc */ read_barrier_depends(); /* we have caught up to head, no work left to do */ if (tx_head == tx_desc) break; /* clear next_to_watch to prevent false hangs */ tx_buf->next_to_watch = NULL; /* update the statistics for this packet */ total_bytes += tx_buf->bytecount; total_packets += tx_buf->gso_segs; /* free the skb */ dev_kfree_skb_any(tx_buf->skb); /* unmap skb header data */ dma_unmap_single(tx_ring->dev, dma_unmap_addr(tx_buf, dma), dma_unmap_len(tx_buf, len), DMA_TO_DEVICE); /* clear tx_buffer data */ tx_buf->skb = NULL; dma_unmap_len_set(tx_buf, len, 0); /* unmap remaining buffers */ while (tx_desc != eop_desc) { tx_buf++; tx_desc++; i++; if (unlikely(!i)) { i -= tx_ring->count; tx_buf = tx_ring->tx_bi; tx_desc = I40E_TX_DESC(tx_ring, 0); } /* unmap any remaining paged data */ if (dma_unmap_len(tx_buf, len)) { dma_unmap_page(tx_ring->dev, dma_unmap_addr(tx_buf, dma), dma_unmap_len(tx_buf, len), DMA_TO_DEVICE); dma_unmap_len_set(tx_buf, len, 0); } } /* move us one more past the eop_desc for start of next pkt */ tx_buf++; tx_desc++; i++; if (unlikely(!i)) { i -= tx_ring->count; tx_buf = tx_ring->tx_bi; tx_desc = I40E_TX_DESC(tx_ring, 0); } prefetch(tx_desc); /* update budget accounting */ budget--; } while (likely(budget)); i += tx_ring->count; tx_ring->next_to_clean = i; u64_stats_update_begin(&tx_ring->syncp); tx_ring->stats.bytes += total_bytes; tx_ring->stats.packets += total_packets; u64_stats_update_end(&tx_ring->syncp); tx_ring->q_vector->tx.total_bytes += total_bytes; tx_ring->q_vector->tx.total_packets += total_packets; if (tx_ring->flags & I40E_TXR_FLAGS_WB_ON_ITR) { unsigned int j = 0; /* check to see if there are < 4 descriptors * waiting to be written back, then kick the hardware to force * them to be written back in case we stay in NAPI. * In this mode on X722 we do not enable Interrupt. */ j = i40evf_get_tx_pending(tx_ring, false); if (budget && ((j / (WB_STRIDE + 1)) == 0) && (j > 0) && !test_bit(__I40E_DOWN, &tx_ring->vsi->state) && (I40E_DESC_UNUSED(tx_ring) != tx_ring->count)) tx_ring->arm_wb = true; } netdev_tx_completed_queue(netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index), total_packets, total_bytes); #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2) if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) && (I40E_DESC_UNUSED(tx_ring) >= TX_WAKE_THRESHOLD))) { /* Make sure that anybody stopping the queue after this * sees the new next_to_clean. */ smp_mb(); if (__netif_subqueue_stopped(tx_ring->netdev, tx_ring->queue_index) && !test_bit(__I40E_DOWN, &tx_ring->vsi->state)) { netif_wake_subqueue(tx_ring->netdev, tx_ring->queue_index); ++tx_ring->tx_stats.restart_queue; } } return !!budget; } /** * i40evf_enable_wb_on_itr - Arm hardware to do a wb, interrupts are not enabled * @vsi: the VSI we care about * @q_vector: the vector on which to enable writeback * **/ static void i40e_enable_wb_on_itr(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector) { u16 flags = q_vector->tx.ring[0].flags; u32 val; if (!(flags & I40E_TXR_FLAGS_WB_ON_ITR)) return; if (q_vector->arm_wb_state) return; val = I40E_VFINT_DYN_CTLN1_WB_ON_ITR_MASK | I40E_VFINT_DYN_CTLN1_ITR_INDX_MASK; /* set noitr */ wr32(&vsi->back->hw, I40E_VFINT_DYN_CTLN1(q_vector->v_idx + vsi->base_vector - 1), val); q_vector->arm_wb_state = true; } /** * i40evf_force_wb - Issue SW Interrupt so HW does a wb * @vsi: the VSI we care about * @q_vector: the vector on which to force writeback * **/ void i40evf_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector) { u32 val = I40E_VFINT_DYN_CTLN1_INTENA_MASK | I40E_VFINT_DYN_CTLN1_ITR_INDX_MASK | /* set noitr */ I40E_VFINT_DYN_CTLN1_SWINT_TRIG_MASK | I40E_VFINT_DYN_CTLN1_SW_ITR_INDX_ENA_MASK /* allow 00 to be written to the index */; wr32(&vsi->back->hw, I40E_VFINT_DYN_CTLN1(q_vector->v_idx + vsi->base_vector - 1), val); } /** * i40e_set_new_dynamic_itr - Find new ITR level * @rc: structure containing ring performance data * * Returns true if ITR changed, false if not * * Stores a new ITR value based on packets and byte counts during * the last interrupt. The advantage of per interrupt computation * is faster updates and more accurate ITR for the current traffic * pattern. Constants in this function were computed based on * theoretical maximum wire speed and thresholds were set based on * testing data as well as attempting to minimize response time * while increasing bulk throughput. **/ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc) { enum i40e_latency_range new_latency_range = rc->latency_range; struct i40e_q_vector *qv = rc->ring->q_vector; u32 new_itr = rc->itr; int bytes_per_int; int usecs; if (rc->total_packets == 0 || !rc->itr) return false; /* simple throttlerate management * 0-10MB/s lowest (50000 ints/s) * 10-20MB/s low (20000 ints/s) * 20-1249MB/s bulk (18000 ints/s) * > 40000 Rx packets per second (8000 ints/s) * * The math works out because the divisor is in 10^(-6) which * turns the bytes/us input value into MB/s values, but * make sure to use usecs, as the register values written * are in 2 usec increments in the ITR registers, and make sure * to use the smoothed values that the countdown timer gives us. */ usecs = (rc->itr << 1) * ITR_COUNTDOWN_START; bytes_per_int = rc->total_bytes / usecs; switch (new_latency_range) { case I40E_LOWEST_LATENCY: if (bytes_per_int > 10) new_latency_range = I40E_LOW_LATENCY; break; case I40E_LOW_LATENCY: if (bytes_per_int > 20) new_latency_range = I40E_BULK_LATENCY; else if (bytes_per_int <= 10) new_latency_range = I40E_LOWEST_LATENCY; break; case I40E_BULK_LATENCY: case I40E_ULTRA_LATENCY: default: if (bytes_per_int <= 20) new_latency_range = I40E_LOW_LATENCY; break; } /* this is to adjust RX more aggressively when streaming small * packets. The value of 40000 was picked as it is just beyond * what the hardware can receive per second if in low latency * mode. */ #define RX_ULTRA_PACKET_RATE 40000 if ((((rc->total_packets * 1000000) / usecs) > RX_ULTRA_PACKET_RATE) && (&qv->rx == rc)) new_latency_range = I40E_ULTRA_LATENCY; rc->latency_range = new_latency_range; switch (new_latency_range) { case I40E_LOWEST_LATENCY: new_itr = I40E_ITR_50K; break; case I40E_LOW_LATENCY: new_itr = I40E_ITR_20K; break; case I40E_BULK_LATENCY: new_itr = I40E_ITR_18K; break; case I40E_ULTRA_LATENCY: new_itr = I40E_ITR_8K; break; default: break; } rc->total_bytes = 0; rc->total_packets = 0; if (new_itr != rc->itr) { rc->itr = new_itr; return true; } return false; } /** * i40evf_setup_tx_descriptors - Allocate the Tx descriptors * @tx_ring: the tx ring to set up * * Return 0 on success, negative on error **/ int i40evf_setup_tx_descriptors(struct i40e_ring *tx_ring) { struct device *dev = tx_ring->dev; int bi_size; if (!dev) return -ENOMEM; /* warn if we are about to overwrite the pointer */ WARN_ON(tx_ring->tx_bi); bi_size = sizeof(struct i40e_tx_buffer) * tx_ring->count; tx_ring->tx_bi = kzalloc(bi_size, GFP_KERNEL); if (!tx_ring->tx_bi) goto err; /* round up to nearest 4K */ tx_ring->size = tx_ring->count * sizeof(struct i40e_tx_desc); /* add u32 for head writeback, align after this takes care of * guaranteeing this is at least one cache line in size */ tx_ring->size += sizeof(u32); tx_ring->size = ALIGN(tx_ring->size, 4096); tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size, &tx_ring->dma, GFP_KERNEL); if (!tx_ring->desc) { dev_info(dev, "Unable to allocate memory for the Tx descriptor ring, size=%d\n", tx_ring->size); goto err; } tx_ring->next_to_use = 0; tx_ring->next_to_clean = 0; return 0; err: kfree(tx_ring->tx_bi); tx_ring->tx_bi = NULL; return -ENOMEM; } /** * i40evf_clean_rx_ring - Free Rx buffers * @rx_ring: ring to be cleaned **/ void i40evf_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; /* ring already cleared, nothing to do */ if (!rx_ring->rx_bi) return; if (ring_is_ps_enabled(rx_ring)) { int bufsz = ALIGN(rx_ring->rx_hdr_len, 256) * rx_ring->count; rx_bi = &rx_ring->rx_bi[0]; if (rx_bi->hdr_buf) { dma_free_coherent(dev, bufsz, rx_bi->hdr_buf, rx_bi->dma); for (i = 0; i < rx_ring->count; i++) { rx_bi = &rx_ring->rx_bi[i]; rx_bi->dma = 0; rx_bi->hdr_buf = NULL; } } } /* 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; } 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; } } bi_size = sizeof(struct i40e_rx_buffer) * rx_ring->count; memset(rx_ring->rx_bi, 0, bi_size); /* Zero out the descriptor ring */ memset(rx_ring->desc, 0, rx_ring->size); rx_ring->next_to_clean = 0; rx_ring->next_to_use = 0; } /** * i40evf_free_rx_resources - Free Rx resources * @rx_ring: ring to clean the resources from * * Free all receive software resources **/ void i40evf_free_rx_resources(struct i40e_ring *rx_ring) { i40evf_clean_rx_ring(rx_ring); kfree(rx_ring->rx_bi); rx_ring->rx_bi = NULL; if (rx_ring->desc) { dma_free_coherent(rx_ring->dev, rx_ring->size, rx_ring->desc, rx_ring->dma); rx_ring->desc = NULL; } } /** * i40evf_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 i40evf_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); } } /** * i40evf_setup_rx_descriptors - Allocate Rx descriptors * @rx_ring: Rx descriptor ring (for a specific queue) to setup * * Returns 0 on success, negative on failure **/ int i40evf_setup_rx_descriptors(struct i40e_ring *rx_ring) { struct device *dev = rx_ring->dev; int bi_size; /* warn if we are about to overwrite the pointer */ WARN_ON(rx_ring->rx_bi); bi_size = sizeof(struct i40e_rx_buffer) * rx_ring->count; rx_ring->rx_bi = kzalloc(bi_size, GFP_KERNEL); if (!rx_ring->rx_bi) goto err; 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 = ALIGN(rx_ring->size, 4096); rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size, &rx_ring->dma, GFP_KERNEL); if (!rx_ring->desc) { dev_info(dev, "Unable to allocate memory for the Rx descriptor ring, size=%d\n", rx_ring->size); goto err; } rx_ring->next_to_clean = 0; rx_ring->next_to_use = 0; return 0; err: kfree(rx_ring->rx_bi); rx_ring->rx_bi = NULL; return -ENOMEM; } /** * i40e_release_rx_desc - Store the new tail and head values * @rx_ring: ring to bump * @val: new head index **/ static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val) { rx_ring->next_to_use = 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, * such as IA-64). */ wmb(); writel(val, rx_ring->tail); } /** * i40evf_alloc_rx_buffers_ps - Replace used receive buffers; packet split * @rx_ring: ring to place buffers on * @cleaned_count: number of buffers to replace * * Returns true if any errors on allocation **/ bool i40evf_alloc_rx_buffers_ps(struct i40e_ring *rx_ring, u16 cleaned_count) { 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(); /* 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]; if (bi->skb) /* desc is in use */ goto no_buffers; /* 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 (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 (rx_ring->next_to_use != i) i40e_release_rx_desc(rx_ring, i); return false; no_buffers: if (rx_ring->next_to_use != i) i40e_release_rx_desc(rx_ring, i); /* make sure to come back via polling to try again after * allocation failure */ return true; } /** * i40evf_alloc_rx_buffers_1buf - Replace used receive buffers; single buffer * @rx_ring: ring to place buffers on * @cleaned_count: number of buffers to replace * * Returns true if any errors on allocation **/ bool i40evf_alloc_rx_buffers_1buf(struct i40e_ring *rx_ring, u16 cleaned_count) { u16 i = 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; } 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; } } rx_desc->read.pkt_addr = cpu_to_le64(bi->dma); 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); return false; no_buffers: if (rx_ring->next_to_use != i) i40e_release_rx_desc(rx_ring, i); /* make sure to come back via polling to try again after * allocation failure */ 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 (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 **/ static inline void i40e_rx_checksum(struct i40e_vsi *vsi, struct sk_buff *skb, u32 rx_status, u32 rx_error, u16 rx_ptype) { struct i40e_rx_ptype_decoded decoded = decode_rx_desc_ptype(rx_ptype); bool ipv4, ipv6, ipv4_tunnel, ipv6_tunnel; skb->ip_summed = CHECKSUM_NONE; /* Rx csum enabled and ip headers found? */ if (!(vsi->netdev->features & NETIF_F_RXCSUM)) return; /* did the hardware decode the packet and checksum? */ if (!(rx_status & BIT(I40E_RX_DESC_STATUS_L3L4P_SHIFT))) return; /* both known and outer_ip must be set for the below code to work */ if (!(decoded.known && decoded.outer_ip)) return; ipv4 = (decoded.outer_ip == I40E_RX_PTYPE_OUTER_IP) && (decoded.outer_ip_ver == I40E_RX_PTYPE_OUTER_IPV4); ipv6 = (decoded.outer_ip == I40E_RX_PTYPE_OUTER_IP) && (decoded.outer_ip_ver == I40E_RX_PTYPE_OUTER_IPV6); if (ipv4 && (rx_error & (BIT(I40E_RX_DESC_ERROR_IPE_SHIFT) | BIT(I40E_RX_DESC_ERROR_EIPE_SHIFT)))) goto checksum_fail; /* likely incorrect csum if alternate IP extension headers found */ if (ipv6 && rx_status & BIT(I40E_RX_DESC_STATUS_IPV6EXADD_SHIFT)) /* don't increment checksum err here, non-fatal err */ return; /* there was some L4 error, count error and punt packet to the stack */ if (rx_error & BIT(I40E_RX_DESC_ERROR_L4E_SHIFT)) goto checksum_fail; /* handle packets that were not able to be checksummed due * to arrival speed, in this case the stack can compute * the csum. */ if (rx_error & BIT(I40E_RX_DESC_ERROR_PPRS_SHIFT)) return; /* The hardware supported by this driver does not validate outer * checksums for tunneled VXLAN or GENEVE frames. I don't agree * with it but the specification states that you "MAY validate", it * doesn't make it a hard requirement so if we have validated the * inner checksum report CHECKSUM_UNNECESSARY. */ ipv4_tunnel = (rx_ptype >= I40E_RX_PTYPE_GRENAT4_MAC_PAY3) && (rx_ptype <= I40E_RX_PTYPE_GRENAT4_MACVLAN_IPV6_ICMP_PAY4); ipv6_tunnel = (rx_ptype >= I40E_RX_PTYPE_GRENAT6_MAC_PAY3) && (rx_ptype <= I40E_RX_PTYPE_GRENAT6_MACVLAN_IPV6_ICMP_PAY4); skb->ip_summed = CHECKSUM_UNNECESSARY; skb->csum_level = ipv4_tunnel || ipv6_tunnel; return; checksum_fail: vsi->back->hw_csum_rx_error++; } /** * i40e_ptype_to_htype - get a hash type * @ptype: the ptype value from the descriptor * * Returns a hash type to be used by skb_set_hash **/ static inline enum pkt_hash_types i40e_ptype_to_htype(u8 ptype) { struct i40e_rx_ptype_decoded decoded = decode_rx_desc_ptype(ptype); if (!decoded.known) return PKT_HASH_TYPE_NONE; if (decoded.outer_ip == I40E_RX_PTYPE_OUTER_IP && decoded.payload_layer == I40E_RX_PTYPE_PAYLOAD_LAYER_PAY4) return PKT_HASH_TYPE_L4; else if (decoded.outer_ip == I40E_RX_PTYPE_OUTER_IP && decoded.payload_layer == I40E_RX_PTYPE_PAYLOAD_LAYER_PAY3) return PKT_HASH_TYPE_L3; else return PKT_HASH_TYPE_L2; } /** * i40e_rx_hash - set the hash value in the skb * @ring: descriptor ring * @rx_desc: specific descriptor **/ static inline void i40e_rx_hash(struct i40e_ring *ring, union i40e_rx_desc *rx_desc, struct sk_buff *skb, u8 rx_ptype) { u32 hash; const __le64 rss_mask = cpu_to_le64((u64)I40E_RX_DESC_FLTSTAT_RSS_HASH << I40E_RX_DESC_STATUS_FLTSTAT_SHIFT); if (ring->netdev->features & NETIF_F_RXHASH) return; if ((rx_desc->wb.qword1.status_error_len & rss_mask) == rss_mask) { hash = le32_to_cpu(rx_desc->wb.qword0.hi_dword.rss); skb_set_hash(skb, hash, i40e_ptype_to_htype(rx_ptype)); } } /** * i40e_clean_rx_irq_ps - Reclaim resources after receive; packet split * @rx_ring: rx ring to clean * @budget: how many cleans we're allowed * * Returns true if there's any budget left (e.g. the clean is finished) **/ static int i40e_clean_rx_irq_ps(struct i40e_ring *rx_ring, const int budget) { unsigned int total_rx_bytes = 0, total_rx_packets = 0; u16 rx_packet_len, rx_header_len, rx_sph, rx_hbo; u16 cleaned_count = I40E_DESC_UNUSED(rx_ring); struct i40e_vsi *vsi = rx_ring->vsi; u16 i = rx_ring->next_to_clean; union i40e_rx_desc *rx_desc; u32 rx_error, rx_status; bool failure = false; u8 rx_ptype; u64 qword; u32 copysize; do { struct i40e_rx_buffer *rx_bi; struct sk_buff *skb; u16 vlan_tag; /* return some buffers to hardware, one at a time is too slow */ if (cleaned_count >= I40E_RX_BUFFER_WRITE) { failure = failure || i40evf_alloc_rx_buffers_ps(rx_ring, cleaned_count); cleaned_count = 0; } i = rx_ring->next_to_clean; rx_desc = I40E_RX_DESC(rx_ring, i); qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len); rx_status = (qword & I40E_RXD_QW1_STATUS_MASK) >> I40E_RXD_QW1_STATUS_SHIFT; if (!(rx_status & BIT(I40E_RX_DESC_STATUS_DD_SHIFT))) 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(); /* sync header buffer for reading */ dma_sync_single_range_for_cpu(rx_ring->dev, rx_ring->rx_bi[0].dma, i * rx_ring->rx_hdr_len, rx_ring->rx_hdr_len, DMA_FROM_DEVICE); rx_bi = &rx_ring->rx_bi[i]; skb = rx_bi->skb; if (likely(!skb)) { skb = __netdev_alloc_skb_ip_align(rx_ring->netdev, rx_ring->rx_hdr_len, GFP_ATOMIC | __GFP_NOWARN); if (!skb) { rx_ring->rx_stats.alloc_buff_failed++; failure = true; break; } /* initialize queue mapping */ skb_record_rx_queue(skb, rx_ring->queue_index); /* we are reusing so sync this buffer for CPU use */ dma_sync_single_range_for_cpu(rx_ring->dev, rx_ring->rx_bi[0].dma, i * rx_ring->rx_hdr_len, rx_ring->rx_hdr_len, DMA_FROM_DEVICE); } rx_packet_len = (qword & I40E_RXD_QW1_LENGTH_PBUF_MASK) >> I40E_RXD_QW1_LENGTH_PBUF_SHIFT; rx_header_len = (qword & I40E_RXD_QW1_LENGTH_HBUF_MASK) >> I40E_RXD_QW1_LENGTH_HBUF_SHIFT; rx_sph = (qword & I40E_RXD_QW1_LENGTH_SPH_MASK) >> I40E_RXD_QW1_LENGTH_SPH_SHIFT; rx_error = (qword & I40E_RXD_QW1_ERROR_MASK) >> I40E_RXD_QW1_ERROR_SHIFT; rx_hbo = rx_error & BIT(I40E_RX_DESC_ERROR_HBO_SHIFT); rx_error &= ~BIT(I40E_RX_DESC_ERROR_HBO_SHIFT); rx_ptype = (qword & I40E_RXD_QW1_PTYPE_MASK) >> I40E_RXD_QW1_PTYPE_SHIFT; /* sync half-page for reading */ dma_sync_single_range_for_cpu(rx_ring->dev, rx_bi->page_dma, rx_bi->page_offset, PAGE_SIZE / 2, DMA_FROM_DEVICE); prefetch(page_address(rx_bi->page) + rx_bi->page_offset); rx_bi->skb = NULL; cleaned_count++; copysize = 0; if (rx_hbo || rx_sph) { int len; if (rx_hbo) len = I40E_RX_HDR_SIZE; else len = rx_header_len; memcpy(__skb_put(skb, len), rx_bi->hdr_buf, len); } else if (skb->len == 0) { int len; unsigned char *va = page_address(rx_bi->page) + rx_bi->page_offset; len = min(rx_packet_len, rx_ring->rx_hdr_len); memcpy(__skb_put(skb, len), va, len); copysize = len; rx_packet_len -= len; } /* Get the rest of the data if this was a header split */ if (rx_packet_len) { skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_bi->page, rx_bi->page_offset + copysize, rx_packet_len, I40E_RXBUFFER_2048); /* If the page count is more than 2, then both halves * of the page are used and we need to free it. Do it * here instead of in the alloc code. Otherwise one * of the half-pages might be released between now and * then, and we wouldn't know which one to use. * Don't call get_page and free_page since those are * both expensive atomic operations that just change * the refcount in opposite directions. Just give the * page to the stack; he can have our refcount. */ if (page_count(rx_bi->page) > 2) { dma_unmap_page(rx_ring->dev, rx_bi->page_dma, PAGE_SIZE, DMA_FROM_DEVICE); rx_bi->page = NULL; rx_bi->page_dma = 0; rx_ring->rx_stats.realloc_count++; } else { get_page(rx_bi->page); /* switch to the other half-page here; the * allocation code programs the right addr * into HW. If we haven't used this half-page, * the address won't be changed, and HW can * just use it next time through. */ rx_bi->page_offset ^= PAGE_SIZE / 2; } } I40E_RX_INCREMENT(rx_ring, i); if (unlikely( !(rx_status & BIT(I40E_RX_DESC_STATUS_EOF_SHIFT)))) { struct i40e_rx_buffer *next_buffer; next_buffer = &rx_ring->rx_bi[i]; next_buffer->skb = skb; rx_ring->rx_stats.non_eop_descs++; continue; } /* ERR_MASK will only have valid bits if EOP set */ if (unlikely(rx_error & BIT(I40E_RX_DESC_ERROR_RXE_SHIFT))) { dev_kfree_skb_any(skb); continue; } i40e_rx_hash(rx_ring, rx_desc, skb, rx_ptype); /* 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); 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 (!i40e_fcoe_handle_offload(rx_ring, rx_desc, skb)) { dev_kfree_skb_any(skb); continue; } #endif i40e_receive_skb(rx_ring, skb, vlan_tag); rx_desc->wb.qword1.status_error_len = 0; } while (likely(total_rx_packets < budget)); u64_stats_update_begin(&rx_ring->syncp); rx_ring->stats.packets += total_rx_packets; rx_ring->stats.bytes += total_rx_bytes; u64_stats_update_end(&rx_ring->syncp); rx_ring->q_vector->rx.total_packets += total_rx_packets; rx_ring->q_vector->rx.total_bytes += total_rx_bytes; return failure ? budget : total_rx_packets; } /** * i40e_clean_rx_irq_1buf - Reclaim resources after receive; single buffer * @rx_ring: rx ring to clean * @budget: how many cleans we're allowed * * Returns number of packets cleaned **/ static int i40e_clean_rx_irq_1buf(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; struct sk_buff *skb; u16 vlan_tag; /* return some buffers to hardware, one at a time is too slow */ if (cleaned_count >= I40E_RX_BUFFER_WRITE) { failure = failure || i40evf_alloc_rx_buffers_1buf(rx_ring, cleaned_count); cleaned_count = 0; } i = rx_ring->next_to_clean; rx_desc = I40E_RX_DESC(rx_ring, i); qword = le64_to_cpu(rx_desc->wb.qword1.status_error_len); rx_status = (qword & I40E_RXD_QW1_STATUS_MASK) >> I40E_RXD_QW1_STATUS_SHIFT; if (!(rx_status & BIT(I40E_RX_DESC_STATUS_DD_SHIFT))) 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(); 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); 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++; continue; } /* ERR_MASK will only have valid bits if EOP set */ if (unlikely(rx_error & BIT(I40E_RX_DESC_ERROR_RXE_SHIFT))) { dev_kfree_skb_any(skb); continue; } i40e_rx_hash(rx_ring, rx_desc, skb, rx_ptype); /* 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); 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; i40e_receive_skb(rx_ring, skb, vlan_tag); rx_desc->wb.qword1.status_error_len = 0; } while (likely(total_rx_packets < budget)); u64_stats_update_begin(&rx_ring->syncp); rx_ring->stats.packets += total_rx_packets; rx_ring->stats.bytes += total_rx_bytes; u64_stats_update_end(&rx_ring->syncp); rx_ring->q_vector->rx.total_packets += total_rx_packets; rx_ring->q_vector->rx.total_bytes += total_rx_bytes; return failure ? budget : total_rx_packets; } static u32 i40e_buildreg_itr(const int type, const u16 itr) { u32 val; val = I40E_VFINT_DYN_CTLN1_INTENA_MASK | /* Don't clear PBA because that can cause lost interrupts that * came in while we were cleaning/polling */ (type << I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) | (itr << I40E_VFINT_DYN_CTLN1_INTERVAL_SHIFT); return val; } /* a small macro to shorten up some long lines */ #define INTREG I40E_VFINT_DYN_CTLN1 /** * i40e_update_enable_itr - Update itr and re-enable MSIX interrupt * @vsi: the VSI we care about * @q_vector: q_vector for which itr is being updated and interrupt enabled * **/ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector) { struct i40e_hw *hw = &vsi->back->hw; bool rx = false, tx = false; u32 rxval, txval; int vector; vector = (q_vector->v_idx + vsi->base_vector); /* avoid dynamic calculation if in countdown mode OR if * all dynamic is disabled */ rxval = txval = i40e_buildreg_itr(I40E_ITR_NONE, 0); if (q_vector->itr_countdown > 0 || (!ITR_IS_DYNAMIC(vsi->rx_itr_setting) && !ITR_IS_DYNAMIC(vsi->tx_itr_setting))) { goto enable_int; } if (ITR_IS_DYNAMIC(vsi->rx_itr_setting)) { rx = i40e_set_new_dynamic_itr(&q_vector->rx); rxval = i40e_buildreg_itr(I40E_RX_ITR, q_vector->rx.itr); } if (ITR_IS_DYNAMIC(vsi->tx_itr_setting)) { tx = i40e_set_new_dynamic_itr(&q_vector->tx); txval = i40e_buildreg_itr(I40E_TX_ITR, q_vector->tx.itr); } if (rx || tx) { /* get the higher of the two ITR adjustments and * use the same value for both ITR registers * when in adaptive mode (Rx and/or Tx) */ u16 itr = max(q_vector->tx.itr, q_vector->rx.itr); q_vector->tx.itr = q_vector->rx.itr = itr; txval = i40e_buildreg_itr(I40E_TX_ITR, itr); tx = true; rxval = i40e_buildreg_itr(I40E_RX_ITR, itr); rx = true; } /* only need to enable the interrupt once, but need * to possibly update both ITR values */ if (rx) { /* set the INTENA_MSK_MASK so that this first write * won't actually enable the interrupt, instead just * updating the ITR (it's bit 31 PF and VF) */ rxval |= BIT(31); /* don't check _DOWN because interrupt isn't being enabled */ wr32(hw, INTREG(vector - 1), rxval); } enable_int: if (!test_bit(__I40E_DOWN, &vsi->state)) wr32(hw, INTREG(vector - 1), txval); if (q_vector->itr_countdown) q_vector->itr_countdown--; else q_vector->itr_countdown = ITR_COUNTDOWN_START; } /** * i40evf_napi_poll - NAPI polling Rx/Tx cleanup routine * @napi: napi struct with our devices info in it * @budget: amount of work driver is allowed to do this pass, in packets * * This function will clean all queues associated with a q_vector. * * Returns the amount of work done **/ int i40evf_napi_poll(struct napi_struct *napi, int budget) { struct i40e_q_vector *q_vector = container_of(napi, struct i40e_q_vector, napi); struct i40e_vsi *vsi = q_vector->vsi; struct i40e_ring *ring; bool clean_complete = true; bool arm_wb = false; int budget_per_ring; int work_done = 0; if (test_bit(__I40E_DOWN, &vsi->state)) { napi_complete(napi); return 0; } /* Since the actual Tx work is minimal, we can give the Tx a larger * budget and be more aggressive about cleaning up the Tx descriptors. */ i40e_for_each_ring(ring, q_vector->tx) { clean_complete = clean_complete && i40e_clean_tx_irq(ring, vsi->work_limit); arm_wb = arm_wb || ring->arm_wb; ring->arm_wb = false; } /* Handle case where we are called by netpoll with a budget of 0 */ if (budget <= 0) goto tx_only; /* We attempt to distribute budget to each Rx queue fairly, but don't * allow the budget to go below 1 because that would exit polling early. */ budget_per_ring = max(budget/q_vector->num_ringpairs, 1); i40e_for_each_ring(ring, q_vector->rx) { int cleaned; if (ring_is_ps_enabled(ring)) cleaned = i40e_clean_rx_irq_ps(ring, budget_per_ring); else cleaned = i40e_clean_rx_irq_1buf(ring, budget_per_ring); work_done += cleaned; /* if we didn't clean as many as budgeted, we must be done */ clean_complete = clean_complete && (budget_per_ring > cleaned); } /* If work not completed, return budget and polling will return */ if (!clean_complete) { tx_only: if (arm_wb) { q_vector->tx.ring[0].tx_stats.tx_force_wb++; i40e_enable_wb_on_itr(vsi, q_vector); } return budget; } if (vsi->back->flags & I40E_TXR_FLAGS_WB_ON_ITR) q_vector->arm_wb_state = false; /* Work is done so exit the polling mode and re-enable the interrupt */ napi_complete_done(napi, work_done); i40e_update_enable_itr(vsi, q_vector); return 0; } /** * i40evf_tx_prepare_vlan_flags - prepare generic TX VLAN tagging flags for HW * @skb: send buffer * @tx_ring: ring to send buffer on * @flags: the tx flags to be set * * Checks the skb and set up correspondingly several generic transmit flags * related to VLAN tagging for the HW, such as VLAN, DCB, etc. * * Returns error code indicate the frame should be dropped upon error and the * otherwise returns 0 to indicate the flags has been set properly. **/ static inline int i40evf_tx_prepare_vlan_flags(struct sk_buff *skb, struct i40e_ring *tx_ring, u32 *flags) { __be16 protocol = skb->protocol; u32 tx_flags = 0; if (protocol == htons(ETH_P_8021Q) && !(tx_ring->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)) { /* When HW VLAN acceleration is turned off by the user the * stack sets the protocol to 8021q so that the driver * can take any steps required to support the SW only * VLAN handling. In our case the driver doesn't need * to take any further steps so just set the protocol * to the encapsulated ethertype. */ skb->protocol = vlan_get_protocol(skb); goto out; } /* if we have a HW VLAN tag being added, default to the HW one */ if (skb_vlan_tag_present(skb)) { tx_flags |= skb_vlan_tag_get(skb) << I40E_TX_FLAGS_VLAN_SHIFT; tx_flags |= I40E_TX_FLAGS_HW_VLAN; /* else if it is a SW VLAN, check the next protocol and store the tag */ } else if (protocol == htons(ETH_P_8021Q)) { struct vlan_hdr *vhdr, _vhdr; vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr); if (!vhdr) return -EINVAL; protocol = vhdr->h_vlan_encapsulated_proto; tx_flags |= ntohs(vhdr->h_vlan_TCI) << I40E_TX_FLAGS_VLAN_SHIFT; tx_flags |= I40E_TX_FLAGS_SW_VLAN; } out: *flags = tx_flags; return 0; } /** * i40e_tso - set up the tso context descriptor * @tx_ring: ptr to the ring to send * @skb: ptr to the skb we're sending * @hdr_len: ptr to the size of the packet header * @cd_type_cmd_tso_mss: Quad Word 1 * * Returns 0 if no TSO can happen, 1 if tso is going, or error **/ static int i40e_tso(struct i40e_ring *tx_ring, struct sk_buff *skb, u8 *hdr_len, u64 *cd_type_cmd_tso_mss) { u64 cd_cmd, cd_tso_len, cd_mss; union { struct iphdr *v4; struct ipv6hdr *v6; unsigned char *hdr; } ip; union { struct tcphdr *tcp; struct udphdr *udp; unsigned char *hdr; } l4; u32 paylen, l4_offset; int err; if (skb->ip_summed != CHECKSUM_PARTIAL) return 0; if (!skb_is_gso(skb)) return 0; err = skb_cow_head(skb, 0); if (err < 0) return err; ip.hdr = skb_network_header(skb); l4.hdr = skb_transport_header(skb); /* initialize outer IP header fields */ if (ip.v4->version == 4) { ip.v4->tot_len = 0; ip.v4->check = 0; } else { ip.v6->payload_len = 0; } if (skb_shinfo(skb)->gso_type & (SKB_GSO_UDP_TUNNEL | SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL_CSUM)) { if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM) { /* determine offset of outer transport header */ l4_offset = l4.hdr - skb->data; /* remove payload length from outer checksum */ paylen = (__force u16)l4.udp->check; paylen += ntohs(1) * (u16)~(skb->len - l4_offset); l4.udp->check = ~csum_fold((__force __wsum)paylen); } /* reset pointers to inner headers */ ip.hdr = skb_inner_network_header(skb); l4.hdr = skb_inner_transport_header(skb); /* initialize inner IP header fields */ if (ip.v4->version == 4) { ip.v4->tot_len = 0; ip.v4->check = 0; } else { ip.v6->payload_len = 0; } } /* determine offset of inner transport header */ l4_offset = l4.hdr - skb->data; /* remove payload length from inner checksum */ paylen = (__force u16)l4.tcp->check; paylen += ntohs(1) * (u16)~(skb->len - l4_offset); l4.tcp->check = ~csum_fold((__force __wsum)paylen); /* compute length of segmentation header */ *hdr_len = (l4.tcp->doff * 4) + l4_offset; /* find the field values */ cd_cmd = I40E_TX_CTX_DESC_TSO; cd_tso_len = skb->len - *hdr_len; cd_mss = skb_shinfo(skb)->gso_size; *cd_type_cmd_tso_mss |= (cd_cmd << I40E_TXD_CTX_QW1_CMD_SHIFT) | (cd_tso_len << I40E_TXD_CTX_QW1_TSO_LEN_SHIFT) | (cd_mss << I40E_TXD_CTX_QW1_MSS_SHIFT); return 1; } /** * i40e_tx_enable_csum - Enable Tx checksum offloads * @skb: send buffer * @tx_flags: pointer to Tx flags currently set * @td_cmd: Tx descriptor command bits to set * @td_offset: Tx descriptor header offsets to set * @tx_ring: Tx descriptor ring * @cd_tunneling: ptr to context desc bits **/ static int i40e_tx_enable_csum(struct sk_buff *skb, u32 *tx_flags, u32 *td_cmd, u32 *td_offset, struct i40e_ring *tx_ring, u32 *cd_tunneling) { union { struct iphdr *v4; struct ipv6hdr *v6; unsigned char *hdr; } ip; union { struct tcphdr *tcp; struct udphdr *udp; unsigned char *hdr; } l4; unsigned char *exthdr; u32 offset, cmd = 0, tunnel = 0; __be16 frag_off; u8 l4_proto = 0; if (skb->ip_summed != CHECKSUM_PARTIAL) return 0; ip.hdr = skb_network_header(skb); l4.hdr = skb_transport_header(skb); /* compute outer L2 header size */ offset = ((ip.hdr - skb->data) / 2) << I40E_TX_DESC_LENGTH_MACLEN_SHIFT; if (skb->encapsulation) { /* define outer network header type */ if (*tx_flags & I40E_TX_FLAGS_IPV4) { tunnel |= (*tx_flags & I40E_TX_FLAGS_TSO) ? I40E_TX_CTX_EXT_IP_IPV4 : I40E_TX_CTX_EXT_IP_IPV4_NO_CSUM; l4_proto = ip.v4->protocol; } else if (*tx_flags & I40E_TX_FLAGS_IPV6) { tunnel |= I40E_TX_CTX_EXT_IP_IPV6; exthdr = ip.hdr + sizeof(*ip.v6); l4_proto = ip.v6->nexthdr; if (l4.hdr != exthdr) ipv6_skip_exthdr(skb, exthdr - skb->data, &l4_proto, &frag_off); } /* compute outer L3 header size */ tunnel |= ((l4.hdr - ip.hdr) / 4) << I40E_TXD_CTX_QW0_EXT_IPLEN_SHIFT; /* switch IP header pointer from outer to inner header */ ip.hdr = skb_inner_network_header(skb); /* define outer transport */ switch (l4_proto) { case IPPROTO_UDP: tunnel |= I40E_TXD_CTX_UDP_TUNNELING; *tx_flags |= I40E_TX_FLAGS_VXLAN_TUNNEL; break; case IPPROTO_GRE: tunnel |= I40E_TXD_CTX_GRE_TUNNELING; *tx_flags |= I40E_TX_FLAGS_VXLAN_TUNNEL; break; default: if (*tx_flags & I40E_TX_FLAGS_TSO) return -1; skb_checksum_help(skb); return 0; } /* compute tunnel header size */ tunnel |= ((ip.hdr - l4.hdr) / 2) << I40E_TXD_CTX_QW0_NATLEN_SHIFT; /* indicate if we need to offload outer UDP header */ if ((*tx_flags & I40E_TX_FLAGS_TSO) && (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)) tunnel |= I40E_TXD_CTX_QW0_L4T_CS_MASK; /* record tunnel offload values */ *cd_tunneling |= tunnel; /* switch L4 header pointer from outer to inner */ l4.hdr = skb_inner_transport_header(skb); l4_proto = 0; /* reset type as we transition from outer to inner headers */ *tx_flags &= ~(I40E_TX_FLAGS_IPV4 | I40E_TX_FLAGS_IPV6); if (ip.v4->version == 4) *tx_flags |= I40E_TX_FLAGS_IPV4; if (ip.v6->version == 6) *tx_flags |= I40E_TX_FLAGS_IPV6; } /* Enable IP checksum offloads */ if (*tx_flags & I40E_TX_FLAGS_IPV4) { l4_proto = ip.v4->protocol; /* the stack computes the IP header already, the only time we * need the hardware to recompute it is in the case of TSO. */ cmd |= (*tx_flags & I40E_TX_FLAGS_TSO) ? I40E_TX_DESC_CMD_IIPT_IPV4_CSUM : I40E_TX_DESC_CMD_IIPT_IPV4; } else if (*tx_flags & I40E_TX_FLAGS_IPV6) { cmd |= I40E_TX_DESC_CMD_IIPT_IPV6; exthdr = ip.hdr + sizeof(*ip.v6); l4_proto = ip.v6->nexthdr; if (l4.hdr != exthdr) ipv6_skip_exthdr(skb, exthdr - skb->data, &l4_proto, &frag_off); } /* compute inner L3 header size */ offset |= ((l4.hdr - ip.hdr) / 4) << I40E_TX_DESC_LENGTH_IPLEN_SHIFT; /* Enable L4 checksum offloads */ switch (l4_proto) { case IPPROTO_TCP: /* enable checksum offloads */ cmd |= I40E_TX_DESC_CMD_L4T_EOFT_TCP; offset |= l4.tcp->doff << I40E_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; break; case IPPROTO_SCTP: /* enable SCTP checksum offload */ cmd |= I40E_TX_DESC_CMD_L4T_EOFT_SCTP; offset |= (sizeof(struct sctphdr) >> 2) << I40E_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; break; case IPPROTO_UDP: /* enable UDP checksum offload */ cmd |= I40E_TX_DESC_CMD_L4T_EOFT_UDP; offset |= (sizeof(struct udphdr) >> 2) << I40E_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; break; default: if (*tx_flags & I40E_TX_FLAGS_TSO) return -1; skb_checksum_help(skb); return 0; } *td_cmd |= cmd; *td_offset |= offset; return 1; } /** * i40e_create_tx_ctx Build the Tx context descriptor * @tx_ring: ring to create the descriptor on * @cd_type_cmd_tso_mss: Quad Word 1 * @cd_tunneling: Quad Word 0 - bits 0-31 * @cd_l2tag2: Quad Word 0 - bits 32-63 **/ static void i40e_create_tx_ctx(struct i40e_ring *tx_ring, const u64 cd_type_cmd_tso_mss, const u32 cd_tunneling, const u32 cd_l2tag2) { struct i40e_tx_context_desc *context_desc; int i = tx_ring->next_to_use; if ((cd_type_cmd_tso_mss == I40E_TX_DESC_DTYPE_CONTEXT) && !cd_tunneling && !cd_l2tag2) return; /* grab the next descriptor */ context_desc = I40E_TX_CTXTDESC(tx_ring, i); i++; tx_ring->next_to_use = (i < tx_ring->count) ? i : 0; /* cpu_to_le32 and assign to struct fields */ context_desc->tunneling_params = cpu_to_le32(cd_tunneling); context_desc->l2tag2 = cpu_to_le16(cd_l2tag2); context_desc->rsvd = cpu_to_le16(0); context_desc->type_cmd_tso_mss = cpu_to_le64(cd_type_cmd_tso_mss); } /** * __i40evf_chk_linearize - Check if there are more than 8 fragments per packet * @skb: send buffer * * Note: Our HW can't scatter-gather more than 8 fragments to build * a packet on the wire and so we need to figure out the cases where we * need to linearize the skb. **/ bool __i40evf_chk_linearize(struct sk_buff *skb) { const struct skb_frag_struct *frag, *stale; int gso_size, nr_frags, sum; /* check to see if TSO is enabled, if so we may get a repreive */ gso_size = skb_shinfo(skb)->gso_size; if (unlikely(!gso_size)) return true; /* no need to check if number of frags is less than 8 */ nr_frags = skb_shinfo(skb)->nr_frags; if (nr_frags < I40E_MAX_BUFFER_TXD) return false; /* We need to walk through the list and validate that each group * of 6 fragments totals at least gso_size. However we don't need * to perform such validation on the first or last 6 since the first * 6 cannot inherit any data from a descriptor before them, and the * last 6 cannot inherit any data from a descriptor after them. */ nr_frags -= I40E_MAX_BUFFER_TXD - 1; frag = &skb_shinfo(skb)->frags[0]; /* Initialize size to the negative value of gso_size minus 1. We * use this as the worst case scenerio in which the frag ahead * of us only provides one byte which is why we are limited to 6 * descriptors for a single transmit as the header and previous * fragment are already consuming 2 descriptors. */ sum = 1 - gso_size; /* Add size of frags 1 through 5 to create our initial sum */ sum += skb_frag_size(++frag); sum += skb_frag_size(++frag); sum += skb_frag_size(++frag); sum += skb_frag_size(++frag); sum += skb_frag_size(++frag); /* Walk through fragments adding latest fragment, testing it, and * then removing stale fragments from the sum. */ stale = &skb_shinfo(skb)->frags[0]; for (;;) { sum += skb_frag_size(++frag); /* if sum is negative we failed to make sufficient progress */ if (sum < 0) return true; /* use pre-decrement to avoid processing last fragment */ if (!--nr_frags) break; sum -= skb_frag_size(++stale); } return false; } /** * __i40evf_maybe_stop_tx - 2nd level check for tx stop conditions * @tx_ring: the ring to be checked * @size: the size buffer we want to assure is available * * Returns -EBUSY if a stop is needed, else 0 **/ int __i40evf_maybe_stop_tx(struct i40e_ring *tx_ring, int size) { netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index); /* Memory barrier before checking head and tail */ smp_mb(); /* Check again in a case another CPU has just made room available. */ if (likely(I40E_DESC_UNUSED(tx_ring) < size)) return -EBUSY; /* A reprieve! - use start_queue because it doesn't call schedule */ netif_start_subqueue(tx_ring->netdev, tx_ring->queue_index); ++tx_ring->tx_stats.restart_queue; return 0; } /** * i40evf_tx_map - Build the Tx descriptor * @tx_ring: ring to send buffer on * @skb: send buffer * @first: first buffer info buffer to use * @tx_flags: collected send information * @hdr_len: size of the packet header * @td_cmd: the command field in the descriptor * @td_offset: offset for checksum or crc **/ static inline void i40evf_tx_map(struct i40e_ring *tx_ring, struct sk_buff *skb, struct i40e_tx_buffer *first, u32 tx_flags, const u8 hdr_len, u32 td_cmd, u32 td_offset) { unsigned int data_len = skb->data_len; unsigned int size = skb_headlen(skb); struct skb_frag_struct *frag; struct i40e_tx_buffer *tx_bi; struct i40e_tx_desc *tx_desc; u16 i = tx_ring->next_to_use; u32 td_tag = 0; dma_addr_t dma; u16 gso_segs; u16 desc_count = 0; bool tail_bump = true; bool do_rs = false; if (tx_flags & I40E_TX_FLAGS_HW_VLAN) { td_cmd |= I40E_TX_DESC_CMD_IL2TAG1; td_tag = (tx_flags & I40E_TX_FLAGS_VLAN_MASK) >> I40E_TX_FLAGS_VLAN_SHIFT; } if (tx_flags & (I40E_TX_FLAGS_TSO | I40E_TX_FLAGS_FSO)) gso_segs = skb_shinfo(skb)->gso_segs; else gso_segs = 1; /* multiply data chunks by size of headers */ first->bytecount = skb->len - hdr_len + (gso_segs * hdr_len); first->gso_segs = gso_segs; first->skb = skb; first->tx_flags = tx_flags; dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE); tx_desc = I40E_TX_DESC(tx_ring, i); tx_bi = first; for (frag = &skb_shinfo(skb)->frags[0];; frag++) { if (dma_mapping_error(tx_ring->dev, dma)) goto dma_error; /* record length, and DMA address */ dma_unmap_len_set(tx_bi, len, size); dma_unmap_addr_set(tx_bi, dma, dma); tx_desc->buffer_addr = cpu_to_le64(dma); while (unlikely(size > I40E_MAX_DATA_PER_TXD)) { tx_desc->cmd_type_offset_bsz = build_ctob(td_cmd, td_offset, I40E_MAX_DATA_PER_TXD, td_tag); tx_desc++; i++; desc_count++; if (i == tx_ring->count) { tx_desc = I40E_TX_DESC(tx_ring, 0); i = 0; } dma += I40E_MAX_DATA_PER_TXD; size -= I40E_MAX_DATA_PER_TXD; tx_desc->buffer_addr = cpu_to_le64(dma); } if (likely(!data_len)) break; tx_desc->cmd_type_offset_bsz = build_ctob(td_cmd, td_offset, size, td_tag); tx_desc++; i++; desc_count++; if (i == tx_ring->count) { tx_desc = I40E_TX_DESC(tx_ring, 0); i = 0; } size = skb_frag_size(frag); data_len -= size; dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size, DMA_TO_DEVICE); tx_bi = &tx_ring->tx_bi[i]; } /* set next_to_watch value indicating a packet is present */ first->next_to_watch = tx_desc; i++; if (i == tx_ring->count) i = 0; tx_ring->next_to_use = i; netdev_tx_sent_queue(netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index), first->bytecount); i40e_maybe_stop_tx(tx_ring, DESC_NEEDED); /* Algorithm to optimize tail and RS bit setting: * if xmit_more is supported * if xmit_more is true * do not update tail and do not mark RS bit. * if xmit_more is false and last xmit_more was false * if every packet spanned less than 4 desc * then set RS bit on 4th packet and update tail * on every packet * else * update tail and set RS bit on every packet. * if xmit_more is false and last_xmit_more was true * update tail and set RS bit. * * Optimization: wmb to be issued only in case of tail update. * Also optimize the Descriptor WB path for RS bit with the same * algorithm. * * Note: If there are less than 4 packets * pending and interrupts were disabled the service task will * trigger a force WB. */ if (skb->xmit_more && !netif_xmit_stopped(netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index))) { tx_ring->flags |= I40E_TXR_FLAGS_LAST_XMIT_MORE_SET; tail_bump = false; } else if (!skb->xmit_more && !netif_xmit_stopped(netdev_get_tx_queue(tx_ring->netdev, tx_ring->queue_index)) && (!(tx_ring->flags & I40E_TXR_FLAGS_LAST_XMIT_MORE_SET)) && (tx_ring->packet_stride < WB_STRIDE) && (desc_count < WB_STRIDE)) { tx_ring->packet_stride++; } else { tx_ring->packet_stride = 0; tx_ring->flags &= ~I40E_TXR_FLAGS_LAST_XMIT_MORE_SET; do_rs = true; } if (do_rs) tx_ring->packet_stride = 0; tx_desc->cmd_type_offset_bsz = build_ctob(td_cmd, td_offset, size, td_tag) | cpu_to_le64((u64)(do_rs ? I40E_TXD_CMD : I40E_TX_DESC_CMD_EOP) << I40E_TXD_QW1_CMD_SHIFT); /* notify HW of packet */ if (!tail_bump) prefetchw(tx_desc + 1); if (tail_bump) { /* Force memory writes to complete before letting h/w * know there are new descriptors to fetch. (Only * applicable for weak-ordered memory model archs, * such as IA-64). */ wmb(); writel(i, tx_ring->tail); } return; dma_error: dev_info(tx_ring->dev, "TX DMA map failed\n"); /* clear dma mappings for failed tx_bi map */ for (;;) { tx_bi = &tx_ring->tx_bi[i]; i40e_unmap_and_free_tx_resource(tx_ring, tx_bi); if (tx_bi == first) break; if (i == 0) i = tx_ring->count; i--; } tx_ring->next_to_use = i; } /** * i40e_xmit_frame_ring - Sends buffer on Tx ring * @skb: send buffer * @tx_ring: ring to send buffer on * * Returns NETDEV_TX_OK if sent, else an error code **/ static netdev_tx_t i40e_xmit_frame_ring(struct sk_buff *skb, struct i40e_ring *tx_ring) { u64 cd_type_cmd_tso_mss = I40E_TX_DESC_DTYPE_CONTEXT; u32 cd_tunneling = 0, cd_l2tag2 = 0; struct i40e_tx_buffer *first; u32 td_offset = 0; u32 tx_flags = 0; __be16 protocol; u32 td_cmd = 0; u8 hdr_len = 0; int tso, count; /* prefetch the data, we'll need it later */ prefetch(skb->data); count = i40e_xmit_descriptor_count(skb); if (i40e_chk_linearize(skb, count)) { if (__skb_linearize(skb)) goto out_drop; count = TXD_USE_COUNT(skb->len); tx_ring->tx_stats.tx_linearize++; } /* need: 1 descriptor per page * PAGE_SIZE/I40E_MAX_DATA_PER_TXD, * + 1 desc for skb_head_len/I40E_MAX_DATA_PER_TXD, * + 4 desc gap to avoid the cache line where head is, * + 1 desc for context descriptor, * otherwise try next time */ if (i40e_maybe_stop_tx(tx_ring, count + 4 + 1)) { tx_ring->tx_stats.tx_busy++; return NETDEV_TX_BUSY; } /* prepare the xmit flags */ if (i40evf_tx_prepare_vlan_flags(skb, tx_ring, &tx_flags)) goto out_drop; /* obtain protocol of skb */ protocol = vlan_get_protocol(skb); /* record the location of the first descriptor for this packet */ first = &tx_ring->tx_bi[tx_ring->next_to_use]; /* setup IPv4/IPv6 offloads */ if (protocol == htons(ETH_P_IP)) tx_flags |= I40E_TX_FLAGS_IPV4; else if (protocol == htons(ETH_P_IPV6)) tx_flags |= I40E_TX_FLAGS_IPV6; tso = i40e_tso(tx_ring, skb, &hdr_len, &cd_type_cmd_tso_mss); if (tso < 0) goto out_drop; else if (tso) tx_flags |= I40E_TX_FLAGS_TSO; /* Always offload the checksum, since it's in the data descriptor */ tso = i40e_tx_enable_csum(skb, &tx_flags, &td_cmd, &td_offset, tx_ring, &cd_tunneling); if (tso < 0) goto out_drop; skb_tx_timestamp(skb); /* always enable CRC insertion offload */ td_cmd |= I40E_TX_DESC_CMD_ICRC; i40e_create_tx_ctx(tx_ring, cd_type_cmd_tso_mss, cd_tunneling, cd_l2tag2); i40evf_tx_map(tx_ring, skb, first, tx_flags, hdr_len, td_cmd, td_offset); return NETDEV_TX_OK; out_drop: dev_kfree_skb_any(skb); return NETDEV_TX_OK; } /** * i40evf_xmit_frame - Selects the correct VSI and Tx queue to send buffer * @skb: send buffer * @netdev: network interface device structure * * Returns NETDEV_TX_OK if sent, else an error code **/ netdev_tx_t i40evf_xmit_frame(struct sk_buff *skb, struct net_device *netdev) { struct i40evf_adapter *adapter = netdev_priv(netdev); struct i40e_ring *tx_ring = &adapter->tx_rings[skb->queue_mapping]; /* hardware can't handle really short frames, hardware padding works * beyond this point */ if (unlikely(skb->len < I40E_MIN_TX_LEN)) { if (skb_pad(skb, I40E_MIN_TX_LEN - skb->len)) return NETDEV_TX_OK; skb->len = I40E_MIN_TX_LEN; skb_set_tail_pointer(skb, I40E_MIN_TX_LEN); } return i40e_xmit_frame_ring(skb, tx_ring); }