diff options
Diffstat (limited to 'drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c')
| -rw-r--r-- | drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c | 1214 |
1 files changed, 837 insertions, 377 deletions
diff --git a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c index 9b3d43d28106..3d9033f26eff 100644 --- a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c +++ b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c @@ -1,7 +1,7 @@ /******************************************************************************* Intel 82599 Virtual Function driver - Copyright(c) 1999 - 2015 Intel Corporation. + Copyright(c) 1999 - 2018 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, @@ -50,6 +50,9 @@ #include <linux/if_vlan.h> #include <linux/prefetch.h> #include <net/mpls.h> +#include <linux/bpf.h> +#include <linux/bpf_trace.h> +#include <linux/atomic.h> #include "ixgbevf.h" @@ -130,6 +133,9 @@ static void ixgbevf_service_event_complete(struct ixgbevf_adapter *adapter) static void ixgbevf_queue_reset_subtask(struct ixgbevf_adapter *adapter); static void ixgbevf_set_itr(struct ixgbevf_q_vector *q_vector); static void ixgbevf_free_all_rx_resources(struct ixgbevf_adapter *adapter); +static bool ixgbevf_can_reuse_rx_page(struct ixgbevf_rx_buffer *rx_buffer); +static void ixgbevf_reuse_rx_page(struct ixgbevf_ring *rx_ring, + struct ixgbevf_rx_buffer *old_buff); static void ixgbevf_remove_adapter(struct ixgbe_hw *hw) { @@ -318,7 +324,10 @@ static bool ixgbevf_clean_tx_irq(struct ixgbevf_q_vector *q_vector, total_packets += tx_buffer->gso_segs; /* free the skb */ - napi_consume_skb(tx_buffer->skb, napi_budget); + if (ring_is_xdp(tx_ring)) + page_frag_free(tx_buffer->data); + else + napi_consume_skb(tx_buffer->skb, napi_budget); /* unmap skb header data */ dma_unmap_single(tx_ring->dev, @@ -382,7 +391,7 @@ static bool ixgbevf_clean_tx_irq(struct ixgbevf_q_vector *q_vector, eop_desc = tx_ring->tx_buffer_info[i].next_to_watch; - pr_err("Detected Tx Unit Hang\n" + pr_err("Detected Tx Unit Hang%s\n" " Tx Queue <%d>\n" " TDH, TDT <%x>, <%x>\n" " next_to_use <%x>\n" @@ -392,6 +401,7 @@ static bool ixgbevf_clean_tx_irq(struct ixgbevf_q_vector *q_vector, " eop_desc->wb.status <%x>\n" " time_stamp <%lx>\n" " jiffies <%lx>\n", + ring_is_xdp(tx_ring) ? " XDP" : "", tx_ring->queue_index, IXGBE_READ_REG(hw, IXGBE_VFTDH(tx_ring->reg_idx)), IXGBE_READ_REG(hw, IXGBE_VFTDT(tx_ring->reg_idx)), @@ -399,7 +409,9 @@ static bool ixgbevf_clean_tx_irq(struct ixgbevf_q_vector *q_vector, eop_desc, (eop_desc ? eop_desc->wb.status : 0), tx_ring->tx_buffer_info[i].time_stamp, jiffies); - netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index); + if (!ring_is_xdp(tx_ring)) + netif_stop_subqueue(tx_ring->netdev, + tx_ring->queue_index); /* schedule immediate reset if we believe we hung */ ixgbevf_tx_timeout_reset(adapter); @@ -407,6 +419,9 @@ static bool ixgbevf_clean_tx_irq(struct ixgbevf_q_vector *q_vector, return true; } + if (ring_is_xdp(tx_ring)) + return !!budget; + #define TX_WAKE_THRESHOLD (DESC_NEEDED * 2) if (unlikely(total_packets && netif_carrier_ok(tx_ring->netdev) && (ixgbevf_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD))) { @@ -527,6 +542,51 @@ static void ixgbevf_process_skb_fields(struct ixgbevf_ring *rx_ring, skb->protocol = eth_type_trans(skb, rx_ring->netdev); } +static +struct ixgbevf_rx_buffer *ixgbevf_get_rx_buffer(struct ixgbevf_ring *rx_ring, + const unsigned int size) +{ + struct ixgbevf_rx_buffer *rx_buffer; + + rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean]; + prefetchw(rx_buffer->page); + + /* 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, + size, + DMA_FROM_DEVICE); + + rx_buffer->pagecnt_bias--; + + return rx_buffer; +} + +static void ixgbevf_put_rx_buffer(struct ixgbevf_ring *rx_ring, + struct ixgbevf_rx_buffer *rx_buffer, + struct sk_buff *skb) +{ + if (ixgbevf_can_reuse_rx_page(rx_buffer)) { + /* hand second half of page back to the ring */ + ixgbevf_reuse_rx_page(rx_ring, rx_buffer); + } else { + if (IS_ERR(skb)) + /* We are not reusing the buffer so unmap it and free + * any references we are holding to it + */ + dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, + ixgbevf_rx_pg_size(rx_ring), + DMA_FROM_DEVICE, + IXGBEVF_RX_DMA_ATTR); + __page_frag_cache_drain(rx_buffer->page, + rx_buffer->pagecnt_bias); + } + + /* clear contents of rx_buffer */ + rx_buffer->page = NULL; +} + /** * ixgbevf_is_non_eop - process handling of non-EOP buffers * @rx_ring: Rx ring being processed @@ -554,32 +614,38 @@ static bool ixgbevf_is_non_eop(struct ixgbevf_ring *rx_ring, return true; } +static inline unsigned int ixgbevf_rx_offset(struct ixgbevf_ring *rx_ring) +{ + return ring_uses_build_skb(rx_ring) ? IXGBEVF_SKB_PAD : 0; +} + static bool ixgbevf_alloc_mapped_page(struct ixgbevf_ring *rx_ring, struct ixgbevf_rx_buffer *bi) { struct page *page = bi->page; - dma_addr_t dma = bi->dma; + dma_addr_t dma; /* since we are recycling buffers we should seldom need to alloc */ if (likely(page)) return true; /* alloc new page for storage */ - page = dev_alloc_page(); + page = dev_alloc_pages(ixgbevf_rx_pg_order(rx_ring)); if (unlikely(!page)) { rx_ring->rx_stats.alloc_rx_page_failed++; return false; } /* map page for use */ - dma = dma_map_page_attrs(rx_ring->dev, page, 0, PAGE_SIZE, + dma = dma_map_page_attrs(rx_ring->dev, page, 0, + ixgbevf_rx_pg_size(rx_ring), DMA_FROM_DEVICE, IXGBEVF_RX_DMA_ATTR); /* if mapping failed free memory back to system since * there isn't much point in holding memory we can't use */ if (dma_mapping_error(rx_ring->dev, dma)) { - __free_page(page); + __free_pages(page, ixgbevf_rx_pg_order(rx_ring)); rx_ring->rx_stats.alloc_rx_page_failed++; return false; @@ -587,7 +653,7 @@ static bool ixgbevf_alloc_mapped_page(struct ixgbevf_ring *rx_ring, bi->dma = dma; bi->page = page; - bi->page_offset = 0; + bi->page_offset = ixgbevf_rx_offset(rx_ring); bi->pagecnt_bias = 1; rx_ring->rx_stats.alloc_rx_page++; @@ -621,7 +687,7 @@ static void ixgbevf_alloc_rx_buffers(struct ixgbevf_ring *rx_ring, /* sync the buffer for use by the device */ dma_sync_single_range_for_device(rx_ring->dev, bi->dma, bi->page_offset, - IXGBEVF_RX_BUFSZ, + ixgbevf_rx_bufsz(rx_ring), DMA_FROM_DEVICE); /* Refresh the desc even if pkt_addr didn't change @@ -685,6 +751,10 @@ static bool ixgbevf_cleanup_headers(struct ixgbevf_ring *rx_ring, union ixgbe_adv_rx_desc *rx_desc, struct sk_buff *skb) { + /* XDP packets use error pointer so abort at this point */ + if (IS_ERR(skb)) + return true; + /* verify that the packet does not have any known errors */ if (unlikely(ixgbevf_test_staterr(rx_desc, IXGBE_RXDADV_ERR_FRAME_ERR_MASK))) { @@ -734,11 +804,10 @@ static inline bool ixgbevf_page_is_reserved(struct page *page) return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page); } -static bool ixgbevf_can_reuse_rx_page(struct ixgbevf_rx_buffer *rx_buffer, - struct page *page, - const unsigned int truesize) +static bool ixgbevf_can_reuse_rx_page(struct ixgbevf_rx_buffer *rx_buffer) { - unsigned int pagecnt_bias = rx_buffer->pagecnt_bias--; + unsigned int pagecnt_bias = rx_buffer->pagecnt_bias; + struct page *page = rx_buffer->page; /* avoid re-using remote pages */ if (unlikely(ixgbevf_page_is_reserved(page))) @@ -746,17 +815,13 @@ static bool ixgbevf_can_reuse_rx_page(struct ixgbevf_rx_buffer *rx_buffer, #if (PAGE_SIZE < 8192) /* if we are only owner of page we can reuse it */ - if (unlikely(page_ref_count(page) != pagecnt_bias)) + if (unlikely((page_ref_count(page) - pagecnt_bias) > 1)) return false; - - /* flip page offset to other buffer */ - rx_buffer->page_offset ^= IXGBEVF_RX_BUFSZ; - #else - /* move offset up to the next cache line */ - rx_buffer->page_offset += truesize; +#define IXGBEVF_LAST_OFFSET \ + (SKB_WITH_OVERHEAD(PAGE_SIZE) - IXGBEVF_RXBUFFER_2048) - if (rx_buffer->page_offset > (PAGE_SIZE - IXGBEVF_RX_BUFSZ)) + if (rx_buffer->page_offset > IXGBEVF_LAST_OFFSET) return false; #endif @@ -765,7 +830,7 @@ static bool ixgbevf_can_reuse_rx_page(struct ixgbevf_rx_buffer *rx_buffer, * the pagecnt_bias and page count so that we fully restock the * number of references the driver holds. */ - if (unlikely(pagecnt_bias == 1)) { + if (unlikely(!pagecnt_bias)) { page_ref_add(page, USHRT_MAX); rx_buffer->pagecnt_bias = USHRT_MAX; } @@ -777,136 +842,268 @@ static bool ixgbevf_can_reuse_rx_page(struct ixgbevf_rx_buffer *rx_buffer, * ixgbevf_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 + * @size: size of buffer to be added * * 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 ixgbevf_add_rx_frag(struct ixgbevf_ring *rx_ring, +static void ixgbevf_add_rx_frag(struct ixgbevf_ring *rx_ring, struct ixgbevf_rx_buffer *rx_buffer, - u16 size, - union ixgbe_adv_rx_desc *rx_desc, - struct sk_buff *skb) + struct sk_buff *skb, + unsigned int size) { - struct page *page = rx_buffer->page; - unsigned char *va = page_address(page) + rx_buffer->page_offset; #if (PAGE_SIZE < 8192) - unsigned int truesize = IXGBEVF_RX_BUFSZ; + unsigned int truesize = ixgbevf_rx_pg_size(rx_ring) / 2; #else - unsigned int truesize = ALIGN(size, L1_CACHE_BYTES); + unsigned int truesize = ring_uses_build_skb(rx_ring) ? + SKB_DATA_ALIGN(IXGBEVF_SKB_PAD + size) : + SKB_DATA_ALIGN(size); #endif - unsigned int pull_len; + skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page, + rx_buffer->page_offset, size, truesize); +#if (PAGE_SIZE < 8192) + rx_buffer->page_offset ^= truesize; +#else + rx_buffer->page_offset += truesize; +#endif +} - if (unlikely(skb_is_nonlinear(skb))) - goto add_tail_frag; +static +struct sk_buff *ixgbevf_construct_skb(struct ixgbevf_ring *rx_ring, + struct ixgbevf_rx_buffer *rx_buffer, + struct xdp_buff *xdp, + union ixgbe_adv_rx_desc *rx_desc) +{ + unsigned int size = xdp->data_end - xdp->data; +#if (PAGE_SIZE < 8192) + unsigned int truesize = ixgbevf_rx_pg_size(rx_ring) / 2; +#else + unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end - + xdp->data_hard_start); +#endif + unsigned int headlen; + struct sk_buff *skb; - if (likely(size <= IXGBEVF_RX_HDR_SIZE)) { - memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long))); + /* prefetch first cache line of first page */ + prefetch(xdp->data); +#if L1_CACHE_BYTES < 128 + prefetch(xdp->data + L1_CACHE_BYTES); +#endif + /* Note, we get here by enabling legacy-rx via: + * + * ethtool --set-priv-flags <dev> legacy-rx on + * + * In this mode, we currently get 0 extra XDP headroom as + * opposed to having legacy-rx off, where we process XDP + * packets going to stack via ixgbevf_build_skb(). + * + * For ixgbevf_construct_skb() mode it means that the + * xdp->data_meta will always point to xdp->data, since + * the helper cannot expand the head. Should this ever + * changed in future for legacy-rx mode on, then lets also + * add xdp->data_meta handling here. + */ - /* page is not reserved, we can reuse buffer as is */ - if (likely(!ixgbevf_page_is_reserved(page))) - return true; + /* allocate a skb to store the frags */ + skb = napi_alloc_skb(&rx_ring->q_vector->napi, IXGBEVF_RX_HDR_SIZE); + if (unlikely(!skb)) + return NULL; - /* this page cannot be reused so discard it */ - return false; - } - - /* 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, IXGBEVF_RX_HDR_SIZE); + /* Determine available headroom for copy */ + headlen = size; + if (headlen > IXGBEVF_RX_HDR_SIZE) + headlen = eth_get_headlen(xdp->data, IXGBEVF_RX_HDR_SIZE); /* align pull length to size of long to optimize memcpy performance */ - memcpy(__skb_put(skb, pull_len), va, ALIGN(pull_len, sizeof(long))); + memcpy(__skb_put(skb, headlen), xdp->data, + ALIGN(headlen, sizeof(long))); /* update all of the pointers */ - va += pull_len; - size -= pull_len; + size -= headlen; + if (size) { + skb_add_rx_frag(skb, 0, rx_buffer->page, + (xdp->data + headlen) - + page_address(rx_buffer->page), + size, truesize); +#if (PAGE_SIZE < 8192) + rx_buffer->page_offset ^= truesize; +#else + rx_buffer->page_offset += truesize; +#endif + } else { + rx_buffer->pagecnt_bias++; + } -add_tail_frag: - skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page, - (unsigned long)va & ~PAGE_MASK, size, truesize); + return skb; +} + +static inline void ixgbevf_irq_enable_queues(struct ixgbevf_adapter *adapter, + u32 qmask) +{ + struct ixgbe_hw *hw = &adapter->hw; - return ixgbevf_can_reuse_rx_page(rx_buffer, page, truesize); + IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, qmask); } -static struct sk_buff *ixgbevf_fetch_rx_buffer(struct ixgbevf_ring *rx_ring, - union ixgbe_adv_rx_desc *rx_desc, - struct sk_buff *skb) +static struct sk_buff *ixgbevf_build_skb(struct ixgbevf_ring *rx_ring, + struct ixgbevf_rx_buffer *rx_buffer, + struct xdp_buff *xdp, + union ixgbe_adv_rx_desc *rx_desc) { - struct ixgbevf_rx_buffer *rx_buffer; - struct page *page; - u16 size = le16_to_cpu(rx_desc->wb.upper.length); + unsigned int metasize = xdp->data - xdp->data_meta; +#if (PAGE_SIZE < 8192) + unsigned int truesize = ixgbevf_rx_pg_size(rx_ring) / 2; +#else + unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) + + SKB_DATA_ALIGN(xdp->data_end - + xdp->data_hard_start); +#endif + struct sk_buff *skb; - rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean]; - page = rx_buffer->page; - prefetchw(page); + /* Prefetch first cache line of first page. If xdp->data_meta + * is unused, this points to xdp->data, otherwise, we likely + * have a consumer accessing first few bytes of meta data, + * and then actual data. + */ + prefetch(xdp->data_meta); +#if L1_CACHE_BYTES < 128 + prefetch(xdp->data_meta + L1_CACHE_BYTES); +#endif - /* 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, - size, - DMA_FROM_DEVICE); + /* build an skb around the page buffer */ + skb = build_skb(xdp->data_hard_start, truesize); + if (unlikely(!skb)) + return NULL; - if (likely(!skb)) { - void *page_addr = page_address(page) + - rx_buffer->page_offset; + /* update pointers within the skb to store the data */ + skb_reserve(skb, xdp->data - xdp->data_hard_start); + __skb_put(skb, xdp->data_end - xdp->data); + if (metasize) + skb_metadata_set(skb, metasize); - /* prefetch first cache line of first page */ - prefetch(page_addr); -#if L1_CACHE_BYTES < 128 - prefetch(page_addr + L1_CACHE_BYTES); + /* update buffer offset */ +#if (PAGE_SIZE < 8192) + rx_buffer->page_offset ^= truesize; +#else + rx_buffer->page_offset += truesize; #endif - /* allocate a skb to store the frags */ - skb = netdev_alloc_skb_ip_align(rx_ring->netdev, - IXGBEVF_RX_HDR_SIZE); - if (unlikely(!skb)) { - rx_ring->rx_stats.alloc_rx_buff_failed++; - return NULL; - } + return skb; +} - /* 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); - } +#define IXGBEVF_XDP_PASS 0 +#define IXGBEVF_XDP_CONSUMED 1 +#define IXGBEVF_XDP_TX 2 - /* pull page into skb */ - if (ixgbevf_add_rx_frag(rx_ring, rx_buffer, size, rx_desc, skb)) { - /* hand second half of page back to the ring */ - ixgbevf_reuse_rx_page(rx_ring, rx_buffer); - } else { - /* We are not reusing the buffer so unmap it and free - * any references we are holding to it - */ - dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, - PAGE_SIZE, DMA_FROM_DEVICE, - IXGBEVF_RX_DMA_ATTR); - __page_frag_cache_drain(page, rx_buffer->pagecnt_bias); - } +static int ixgbevf_xmit_xdp_ring(struct ixgbevf_ring *ring, + struct xdp_buff *xdp) +{ + struct ixgbevf_tx_buffer *tx_buffer; + union ixgbe_adv_tx_desc *tx_desc; + u32 len, cmd_type; + dma_addr_t dma; + u16 i; - /* clear contents of buffer_info */ - rx_buffer->dma = 0; - rx_buffer->page = NULL; + len = xdp->data_end - xdp->data; - return skb; + if (unlikely(!ixgbevf_desc_unused(ring))) + return IXGBEVF_XDP_CONSUMED; + + dma = dma_map_single(ring->dev, xdp->data, len, DMA_TO_DEVICE); + if (dma_mapping_error(ring->dev, dma)) + return IXGBEVF_XDP_CONSUMED; + + /* record the location of the first descriptor for this packet */ + tx_buffer = &ring->tx_buffer_info[ring->next_to_use]; + tx_buffer->bytecount = len; + tx_buffer->gso_segs = 1; + tx_buffer->protocol = 0; + + i = ring->next_to_use; + tx_desc = IXGBEVF_TX_DESC(ring, i); + + dma_unmap_len_set(tx_buffer, len, len); + dma_unmap_addr_set(tx_buffer, dma, dma); + tx_buffer->data = xdp->data; + tx_desc->read.buffer_addr = cpu_to_le64(dma); + + /* put descriptor type bits */ + cmd_type = IXGBE_ADVTXD_DTYP_DATA | + IXGBE_ADVTXD_DCMD_DEXT | + IXGBE_ADVTXD_DCMD_IFCS; + cmd_type |= len | IXGBE_TXD_CMD; + tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type); + tx_desc->read.olinfo_status = + cpu_to_le32((len << IXGBE_ADVTXD_PAYLEN_SHIFT) | + IXGBE_ADVTXD_CC); + + /* Avoid any potential race with cleanup */ + smp_wmb(); + + /* set next_to_watch value indicating a packet is present */ + i++; + if (i == ring->count) + i = 0; + + tx_buffer->next_to_watch = tx_desc; + ring->next_to_use = i; + + return IXGBEVF_XDP_TX; } -static inline void ixgbevf_irq_enable_queues(struct ixgbevf_adapter *adapter, - u32 qmask) +static struct sk_buff *ixgbevf_run_xdp(struct ixgbevf_adapter *adapter, + struct ixgbevf_ring *rx_ring, + struct xdp_buff *xdp) { - struct ixgbe_hw *hw = &adapter->hw; + int result = IXGBEVF_XDP_PASS; + struct ixgbevf_ring *xdp_ring; + struct bpf_prog *xdp_prog; + u32 act; - IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, qmask); + rcu_read_lock(); + xdp_prog = READ_ONCE(rx_ring->xdp_prog); + + if (!xdp_prog) + goto xdp_out; + + act = bpf_prog_run_xdp(xdp_prog, xdp); + switch (act) { + case XDP_PASS: + break; + case XDP_TX: + xdp_ring = adapter->xdp_ring[rx_ring->queue_index]; + result = ixgbevf_xmit_xdp_ring(xdp_ring, xdp); + break; + default: + bpf_warn_invalid_xdp_action(act); + /* fallthrough */ + case XDP_ABORTED: + trace_xdp_exception(rx_ring->netdev, xdp_prog, act); + /* fallthrough -- handle aborts by dropping packet */ + case XDP_DROP: + result = IXGBEVF_XDP_CONSUMED; + break; + } +xdp_out: + rcu_read_unlock(); + return ERR_PTR(-result); +} + +static void ixgbevf_rx_buffer_flip(struct ixgbevf_ring *rx_ring, + struct ixgbevf_rx_buffer *rx_buffer, + unsigned int size) +{ +#if (PAGE_SIZE < 8192) + unsigned int truesize = ixgbevf_rx_pg_size(rx_ring) / 2; + + rx_buffer->page_offset ^= truesize; +#else + unsigned int truesize = ring_uses_build_skb(rx_ring) ? + SKB_DATA_ALIGN(IXGBEVF_SKB_PAD + size) : + SKB_DATA_ALIGN(size); + + rx_buffer->page_offset += truesize; +#endif } static int ixgbevf_clean_rx_irq(struct ixgbevf_q_vector *q_vector, @@ -914,11 +1111,18 @@ static int ixgbevf_clean_rx_irq(struct ixgbevf_q_vector *q_vector, int budget) { unsigned int total_rx_bytes = 0, total_rx_packets = 0; + struct ixgbevf_adapter *adapter = q_vector->adapter; u16 cleaned_count = ixgbevf_desc_unused(rx_ring); struct sk_buff *skb = rx_ring->skb; + bool xdp_xmit = false; + struct xdp_buff xdp; + + xdp.rxq = &rx_ring->xdp_rxq; while (likely(total_rx_packets < budget)) { + struct ixgbevf_rx_buffer *rx_buffer; union ixgbe_adv_rx_desc *rx_desc; + unsigned int size; /* return some buffers to hardware, one at a time is too slow */ if (cleaned_count >= IXGBEVF_RX_BUFFER_WRITE) { @@ -927,8 +1131,8 @@ static int ixgbevf_clean_rx_irq(struct ixgbevf_q_vector *q_vector, } rx_desc = IXGBEVF_RX_DESC(rx_ring, rx_ring->next_to_clean); - - if (!rx_desc->wb.upper.length) + size = le16_to_cpu(rx_desc->wb.upper.length); + if (!size) break; /* This memory barrier is needed to keep us from reading @@ -937,15 +1141,48 @@ static int ixgbevf_clean_rx_irq(struct ixgbevf_q_vector *q_vector, */ rmb(); + rx_buffer = ixgbevf_get_rx_buffer(rx_ring, size); + /* retrieve a buffer from the ring */ - skb = ixgbevf_fetch_rx_buffer(rx_ring, rx_desc, skb); + if (!skb) { + xdp.data = page_address(rx_buffer->page) + + rx_buffer->page_offset; + xdp.data_meta = xdp.data; + xdp.data_hard_start = xdp.data - + ixgbevf_rx_offset(rx_ring); + xdp.data_end = xdp.data + size; + + skb = ixgbevf_run_xdp(adapter, rx_ring, &xdp); + } + + if (IS_ERR(skb)) { + if (PTR_ERR(skb) == -IXGBEVF_XDP_TX) { + xdp_xmit = true; + ixgbevf_rx_buffer_flip(rx_ring, rx_buffer, + size); + } else { + rx_buffer->pagecnt_bias++; + } + total_rx_packets++; + total_rx_bytes += size; + } else if (skb) { + ixgbevf_add_rx_frag(rx_ring, rx_buffer, skb, size); + } else if (ring_uses_build_skb(rx_ring)) { + skb = ixgbevf_build_skb(rx_ring, rx_buffer, + &xdp, rx_desc); + } else { + skb = ixgbevf_construct_skb(rx_ring, rx_buffer, + &xdp, rx_desc); + } /* exit if we failed to retrieve a buffer */ if (!skb) { rx_ring->rx_stats.alloc_rx_buff_failed++; + rx_buffer->pagecnt_bias++; break; } + ixgbevf_put_rx_buffer(rx_ring, rx_buffer, skb); cleaned_count++; /* fetch next buffer in frame if non-eop */ @@ -987,6 +1224,17 @@ static int ixgbevf_clean_rx_irq(struct ixgbevf_q_vector *q_vector, /* place incomplete frames back on ring for completion */ rx_ring->skb = skb; + if (xdp_xmit) { + struct ixgbevf_ring *xdp_ring = + adapter->xdp_ring[rx_ring->queue_index]; + + /* Force memory writes to complete before letting h/w + * know there are new descriptors to fetch. + */ + wmb(); + ixgbevf_write_tail(xdp_ring, xdp_ring->next_to_use); + } + u64_stats_update_begin(&rx_ring->syncp); rx_ring->stats.packets += total_rx_packets; rx_ring->stats.bytes += total_rx_bytes; @@ -1260,85 +1508,6 @@ static irqreturn_t ixgbevf_msix_clean_rings(int irq, void *data) return IRQ_HANDLED; } -static inline void map_vector_to_rxq(struct ixgbevf_adapter *a, int v_idx, - int r_idx) -{ - struct ixgbevf_q_vector *q_vector = a->q_vector[v_idx]; - - a->rx_ring[r_idx]->next = q_vector->rx.ring; - q_vector->rx.ring = a->rx_ring[r_idx]; - q_vector->rx.count++; -} - -static inline void map_vector_to_txq(struct ixgbevf_adapter *a, int v_idx, - int t_idx) -{ - struct ixgbevf_q_vector *q_vector = a->q_vector[v_idx]; - - a->tx_ring[t_idx]->next = q_vector->tx.ring; - q_vector->tx.ring = a->tx_ring[t_idx]; - q_vector->tx.count++; -} - -/** - * ixgbevf_map_rings_to_vectors - Maps descriptor rings to vectors - * @adapter: board private structure to initialize - * - * This function maps descriptor rings to the queue-specific vectors - * we were allotted through the MSI-X enabling code. Ideally, we'd have - * one vector per ring/queue, but on a constrained vector budget, we - * group the rings as "efficiently" as possible. You would add new - * mapping configurations in here. - **/ -static int ixgbevf_map_rings_to_vectors(struct ixgbevf_adapter *adapter) -{ - int q_vectors; - int v_start = 0; - int rxr_idx = 0, txr_idx = 0; - int rxr_remaining = adapter->num_rx_queues; - int txr_remaining = adapter->num_tx_queues; - int i, j; - int rqpv, tqpv; - - q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS; - - /* The ideal configuration... - * We have enough vectors to map one per queue. - */ - if (q_vectors == adapter->num_rx_queues + adapter->num_tx_queues) { - for (; rxr_idx < rxr_remaining; v_start++, rxr_idx++) - map_vector_to_rxq(adapter, v_start, rxr_idx); - - for (; txr_idx < txr_remaining; v_start++, txr_idx++) - map_vector_to_txq(adapter, v_start, txr_idx); - return 0; - } - - /* If we don't have enough vectors for a 1-to-1 - * mapping, we'll have to group them so there are - * multiple queues per vector. - */ - /* Re-adjusting *qpv takes care of the remainder. */ - for (i = v_start; i < q_vectors; i++) { - rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - i); - for (j = 0; j < rqpv; j++) { - map_vector_to_rxq(adapter, i, rxr_idx); - rxr_idx++; - rxr_remaining--; - } - } - for (i = v_start; i < q_vectors; i++) { - tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - i); - for (j = 0; j < tqpv; j++) { - map_vector_to_txq(adapter, i, txr_idx); - txr_idx++; - txr_remaining--; - } - } - - return 0; -} - /** * ixgbevf_request_msix_irqs - Initialize MSI-X interrupts * @adapter: board private structure @@ -1411,20 +1580,6 @@ free_queue_irqs: return err; } -static inline void ixgbevf_reset_q_vectors(struct ixgbevf_adapter *adapter) -{ - int i, q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS; - - for (i = 0; i < q_vectors; i++) { - struct ixgbevf_q_vector *q_vector = adapter->q_vector[i]; - - q_vector->rx.ring = NULL; - q_vector->tx.ring = NULL; - q_vector->rx.count = 0; - q_vector->tx.count = 0; - } -} - /** * ixgbevf_request_irq - initialize interrupts * @adapter: board private structure @@ -1464,8 +1619,6 @@ static void ixgbevf_free_irq(struct ixgbevf_adapter *adapter) free_irq(adapter->msix_entries[i].vector, adapter->q_vector[i]); } - - ixgbevf_reset_q_vectors(adapter); } /** @@ -1583,11 +1736,14 @@ static void ixgbevf_configure_tx(struct ixgbevf_adapter *adapter) /* Setup the HW Tx Head and Tail descriptor pointers */ for (i = 0; i < adapter->num_tx_queues; i++) ixgbevf_configure_tx_ring(adapter, adapter->tx_ring[i]); + for (i = 0; i < adapter->num_xdp_queues; i++) + ixgbevf_configure_tx_ring(adapter, adapter->xdp_ring[i]); } #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2 -static void ixgbevf_configure_srrctl(struct ixgbevf_adapter *adapter, int index) +static void ixgbevf_configure_srrctl(struct ixgbevf_adapter *adapter, + struct ixgbevf_ring *ring, int index) { struct ixgbe_hw *hw = &adapter->hw; u32 srrctl; @@ -1595,7 +1751,10 @@ static void ixgbevf_configure_srrctl(struct ixgbevf_adapter *adapter, int index) srrctl = IXGBE_SRRCTL_DROP_EN; srrctl |= IXGBEVF_RX_HDR_SIZE << IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT; - srrctl |= IXGBEVF_RX_BUFSZ >> IXGBE_SRRCTL_BSIZEPKT_SHIFT; + if (ring_uses_large_buffer(ring)) + srrctl |= IXGBEVF_RXBUFFER_3072 >> IXGBE_SRRCTL_BSIZEPKT_SHIFT; + else + srrctl |= IXGBEVF_RXBUFFER_2048 >> IXGBE_SRRCTL_BSIZEPKT_SHIFT; srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF; IXGBE_WRITE_REG(hw, IXGBE_VFSRRCTL(index), srrctl); @@ -1767,10 +1926,21 @@ static void ixgbevf_configure_rx_ring(struct ixgbevf_adapter *adapter, ring->next_to_use = 0; ring->next_to_alloc = 0; - ixgbevf_configure_srrctl(adapter, reg_idx); + ixgbevf_configure_srrctl(adapter, ring, reg_idx); + + /* RXDCTL.RLPML does not work on 82599 */ + if (adapter->hw.mac.type != ixgbe_mac_82599_vf) { + rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK | + IXGBE_RXDCTL_RLPML_EN); - /* allow any size packet since we can handle overflow */ - rxdctl &= ~IXGBE_RXDCTL_RLPML_EN; +#if (PAGE_SIZE < 8192) + /* Limit the maximum frame size so we don't overrun the skb */ + if (ring_uses_build_skb(ring) && + !ring_uses_large_buffer(ring)) + rxdctl |= IXGBEVF_MAX_FRAME_BUILD_SKB | + IXGBE_RXDCTL_RLPML_EN; +#endif + } rxdctl |= IXGBE_RXDCTL_ENABLE | IXGBE_RXDCTL_VME; IXGBE_WRITE_REG(hw, IXGBE_VFRXDCTL(reg_idx), rxdctl); @@ -1779,6 +1949,29 @@ static void ixgbevf_configure_rx_ring(struct ixgbevf_adapter *adapter, ixgbevf_alloc_rx_buffers(ring, ixgbevf_desc_unused(ring)); } +static void ixgbevf_set_rx_buffer_len(struct ixgbevf_adapter *adapter, + struct ixgbevf_ring *rx_ring) +{ + struct net_device *netdev = adapter->netdev; + unsigned int max_frame = netdev->mtu + ETH_HLEN + ETH_FCS_LEN; + + /* set build_skb and buffer size flags */ + clear_ring_build_skb_enabled(rx_ring); + clear_ring_uses_large_buffer(rx_ring); + + if (adapter->flags & IXGBEVF_FLAGS_LEGACY_RX) + return; + + set_ring_build_skb_enabled(rx_ring); + + if (PAGE_SIZE < 8192) { + if (max_frame <= IXGBEVF_MAX_FRAME_BUILD_SKB) + return; + + set_ring_uses_large_buffer(rx_ring); + } +} + /** * ixgbevf_configure_rx - Configure 82599 VF Receive Unit after Reset * @adapter: board private structure @@ -1806,8 +1999,12 @@ static void ixgbevf_configure_rx(struct ixgbevf_adapter *adapter) /* Setup the HW Rx Head and Tail Descriptor Pointers and * the Base and Length of the Rx Descriptor Ring */ - for (i = 0; i < adapter->num_rx_queues; i++) - ixgbevf_configure_rx_ring(adapter, adapter->rx_ring[i]); + for (i = 0; i < adapter->num_rx_queues; i++) { + struct ixgbevf_ring *rx_ring = adapter->rx_ring[i]; + + ixgbevf_set_rx_buffer_len(adapter, rx_ring); + ixgbevf_configure_rx_ring(adapter, rx_ring); + } } static int ixgbevf_vlan_rx_add_vid(struct net_device *netdev, @@ -2136,13 +2333,13 @@ static void ixgbevf_clean_rx_ring(struct ixgbevf_ring *rx_ring) dma_sync_single_range_for_cpu(rx_ring->dev, rx_buffer->dma, rx_buffer->page_offset, - IXGBEVF_RX_BUFSZ, + ixgbevf_rx_bufsz(rx_ring), DMA_FROM_DEVICE); /* free resources associated with mapping */ dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma, - PAGE_SIZE, + ixgbevf_rx_pg_size(rx_ring), DMA_FROM_DEVICE, IXGBEVF_RX_DMA_ATTR); @@ -2172,7 +2369,10 @@ static void ixgbevf_clean_tx_ring(struct ixgbevf_ring *tx_ring) union ixgbe_adv_tx_desc *eop_desc, *tx_desc; /* Free all the Tx ring sk_buffs */ - dev_kfree_skb_any(tx_buffer->skb); + if (ring_is_xdp(tx_ring)) + page_frag_free(tx_buffer->data); + else + dev_kfree_skb_any(tx_buffer->skb); /* unmap skb header data */ dma_unmap_single(tx_ring->dev, @@ -2240,6 +2440,8 @@ static void ixgbevf_clean_all_tx_rings(struct ixgbevf_adapter *adapter) for (i = 0; i < adapter->num_tx_queues; i++) ixgbevf_clean_tx_ring(adapter->tx_ring[i]); + for (i = 0; i < adapter->num_xdp_queues; i++) + ixgbevf_clean_tx_ring(adapter->xdp_ring[i]); } void ixgbevf_down(struct ixgbevf_adapter *adapter) @@ -2278,6 +2480,13 @@ void ixgbevf_down(struct ixgbevf_adapter *adapter) IXGBE_TXDCTL_SWFLSH); } + for (i = 0; i < adapter->num_xdp_queues; i++) { + u8 reg_idx = adapter->xdp_ring[i]->reg_idx; + + IXGBE_WRITE_REG(hw, IXGBE_VFTXDCTL(reg_idx), + IXGBE_TXDCTL_SWFLSH); + } + if (!pci_channel_offline(adapter->pdev)) ixgbevf_reset(adapter); @@ -2375,6 +2584,7 @@ static void ixgbevf_set_num_queues(struct ixgbevf_adapter *adapter) /* Start with base case */ adapter->num_rx_queues = 1; adapter->num_tx_queues = 1; + adapter->num_xdp_queues = 0; spin_lock_bh(&adapter->mbx_lock); @@ -2396,8 +2606,13 @@ static void ixgbevf_set_num_queues(struct ixgbevf_adapter *adapter) case ixgbe_mbox_api_11: case ixgbe_mbox_api_12: case ixgbe_mbox_api_13: + if (adapter->xdp_prog && + hw->mac.max_tx_queues == rss) + rss = rss > 3 ? 2 : 1; + adapter->num_rx_queues = rss; adapter->num_tx_queues = rss; + adapter->num_xdp_queues = adapter->xdp_prog ? rss : 0; default: break; } @@ -2405,105 +2620,209 @@ static void ixgbevf_set_num_queues(struct ixgbevf_adapter *adapter) } /** - * ixgbevf_alloc_queues - Allocate memory for all rings + * ixgbevf_set_interrupt_capability - set MSI-X or FAIL if not supported * @adapter: board private structure to initialize * - * We allocate one ring per queue at run-time since we don't know the - * number of queues at compile-time. The polling_netdev array is - * intended for Multiqueue, but should work fine with a single queue. + * Attempt to configure the interrupts using the best available + * capabilities of the hardware and the kernel. **/ -static int ixgbevf_alloc_queues(struct ixgbevf_adapter *adapter) +static int ixgbevf_set_interrupt_capability(struct ixgbevf_adapter *adapter) { + int vector, v_budget; + + /* It's easy to be greedy for MSI-X vectors, but it really + * doesn't do us much good if we have a lot more vectors + * than CPU's. So let's be conservative and only ask for + * (roughly) the same number of vectors as there are CPU's. + * The default is to use pairs of vectors. + */ + v_budget = max(adapter->num_rx_queues, adapter->num_tx_queues); + v_budget = min_t(int, v_budget, num_online_cpus()); + v_budget += NON_Q_VECTORS; + + adapter->msix_entries = kcalloc(v_budget, + sizeof(struct msix_entry), GFP_KERNEL); + if (!adapter->msix_entries) + return -ENOMEM; + + for (vector = 0; vector < v_budget; vector++) + adapter->msix_entries[vector].entry = vector; + + /* A failure in MSI-X entry allocation isn't fatal, but the VF driver + * does not support any other modes, so we will simply fail here. Note + * that we clean up the msix_entries pointer else-where. + */ + return ixgbevf_acquire_msix_vectors(adapter, v_budget); +} + +static void ixgbevf_add_ring(struct ixgbevf_ring *ring, + struct ixgbevf_ring_container *head) +{ + ring->next = head->ring; + head->ring = ring; + head->count++; +} + +/** + * ixgbevf_alloc_q_vector - Allocate memory for a single interrupt vector + * @adapter: board private structure to initialize + * @v_idx: index of vector in adapter struct + * @txr_count: number of Tx rings for q vector + * @txr_idx: index of first Tx ring to assign + * @xdp_count: total number of XDP rings to allocate + * @xdp_idx: index of first XDP ring to allocate + * @rxr_count: number of Rx rings for q vector + * @rxr_idx: index of first Rx ring to assign + * + * We allocate one q_vector. If allocation fails we return -ENOMEM. + **/ +static int ixgbevf_alloc_q_vector(struct ixgbevf_adapter *adapter, int v_idx, + int txr_count, int txr_idx, + int xdp_count, int xdp_idx, + int rxr_count, int rxr_idx) +{ + struct ixgbevf_q_vector *q_vector; + int reg_idx = txr_idx + xdp_idx; struct ixgbevf_ring *ring; - int rx = 0, tx = 0; + int ring_count, size; + + ring_count = txr_count + xdp_count + rxr_count; + size = sizeof(*q_vector) + (sizeof(*ring) * ring_count); + + /* allocate q_vector and rings */ + q_vector = kzalloc(size, GFP_KERNEL); + if (!q_vector) + return -ENOMEM; + + /* initialize NAPI */ + netif_napi_add(adapter->netdev, &q_vector->napi, ixgbevf_poll, 64); - for (; tx < adapter->num_tx_queues; tx++) { - ring = kzalloc(sizeof(*ring), GFP_KERNEL); - if (!ring) - goto err_allocation; + /* tie q_vector and adapter together */ + adapter->q_vector[v_idx] = q_vector; + q_vector->adapter = adapter; + q_vector->v_idx = v_idx; + /* initialize pointer to rings */ + ring = q_vector->ring; + + while (txr_count) { + /* assign generic ring traits */ ring->dev = &adapter->pdev->dev; ring->netdev = adapter->netdev; + + /* configure backlink on ring */ + ring->q_vector = q_vector; + + /* update q_vector Tx values */ + ixgbevf_add_ring(ring, &q_vector->tx); + + /* apply Tx specific ring traits */ ring->count = adapter->tx_ring_count; - ring->queue_index = tx; - ring->reg_idx = tx; + ring->queue_index = txr_idx; + ring->reg_idx = reg_idx; - adapter->tx_ring[tx] = ring; + /* assign ring to adapter */ + adapter->tx_ring[txr_idx] = ring; + + /* update count and index */ + txr_count--; + txr_idx++; + reg_idx++; + + /* push pointer to next ring */ + ring++; } - for (; rx < adapter->num_rx_queues; rx++) { - ring = kzalloc(sizeof(*ring), GFP_KERNEL); - if (!ring) - goto err_allocation; + while (xdp_count) { + /* assign generic ring traits */ + ring->dev = &adapter->pdev->dev; + ring->netdev = adapter->netdev; + /* configure backlink on ring */ + ring->q_vector = q_vector; + + /* update q_vector Tx values */ + ixgbevf_add_ring(ring, &q_vector->tx); + + /* apply Tx specific ring traits */ + ring->count = adapter->tx_ring_count; + ring->queue_index = xdp_idx; + ring->reg_idx = reg_idx; + set_ring_xdp(ring); + + /* assign ring to adapter */ + adapter->xdp_ring[xdp_idx] = ring; + + /* update count and index */ + xdp_count--; + xdp_idx++; + reg_idx++; + + /* push pointer to next ring */ + ring++; + } + + while (rxr_count) { + /* assign generic ring traits */ ring->dev = &adapter->pdev->dev; ring->netdev = adapter->netdev; + /* configure backlink on ring */ + ring->q_vector = q_vector; + + /* update q_vector Rx values */ + ixgbevf_add_ring(ring, &q_vector->rx); + + /* apply Rx specific ring traits */ ring->count = adapter->rx_ring_count; - ring->queue_index = rx; - ring->reg_idx = rx; + ring->queue_index = rxr_idx; + ring->reg_idx = rxr_idx; - adapter->rx_ring[rx] = ring; - } + /* assign ring to adapter */ + adapter->rx_ring[rxr_idx] = ring; - return 0; + /* update count and index */ + rxr_count--; + rxr_idx++; -err_allocation: - while (tx) { - kfree(adapter->tx_ring[--tx]); - adapter->tx_ring[tx] = NULL; + /* push pointer to next ring */ + ring++; } - while (rx) { - kfree(adapter->rx_ring[--rx]); - adapter->rx_ring[rx] = NULL; - } - return -ENOMEM; + return 0; } /** - * ixgbevf_set_interrupt_capability - set MSI-X or FAIL if not supported + * ixgbevf_free_q_vector - Free memory allocated for specific interrupt vector * @adapter: board private structure to initialize + * @v_idx: index of vector in adapter struct * - * Attempt to configure the interrupts using the best available - * capabilities of the hardware and the kernel. + * This function frees the memory allocated to the q_vector. In addition if + * NAPI is enabled it will delete any references to the NAPI struct prior + * to freeing the q_vector. **/ -static int ixgbevf_set_interrupt_capability(struct ixgbevf_adapter *adapter) +static void ixgbevf_free_q_vector(struct ixgbevf_adapter *adapter, int v_idx) { - struct net_device *netdev = adapter->netdev; - int err; - int vector, v_budget; - - /* It's easy to be greedy for MSI-X vectors, but it really - * doesn't do us much good if we have a lot more vectors - * than CPU's. So let's be conservative and only ask for - * (roughly) the same number of vectors as there are CPU's. - * The default is to use pairs of vectors. - */ - v_budget = max(adapter->num_rx_queues, adapter->num_tx_queues); - v_budget = min_t(int, v_budget, num_online_cpus()); - v_budget += NON_Q_VECTORS; - - /* A failure in MSI-X entry allocation isn't fatal, but it does - * mean we disable MSI-X capabilities of the adapter. - */ - adapter->msix_entries = kcalloc(v_budget, - sizeof(struct msix_entry), GFP_KERNEL); - if (!adapter->msix_entries) - return -ENOMEM; + struct ixgbevf_q_vector *q_vector = adapter->q_vector[v_idx]; + struct ixgbevf_ring *ring; - for (vector = 0; vector < v_budget; vector++) - adapter->msix_entries[vector].entry = vector; + ixgbevf_for_each_ring(ring, q_vector->tx) { + if (ring_is_xdp(ring)) + adapter->xdp_ring[ring->queue_index] = NULL; + else + adapter->tx_ring[ring->queue_index] = NULL; + } - err = ixgbevf_acquire_msix_vectors(adapter, v_budget); - if (err) - return err; + ixgbevf_for_each_ring(ring, q_vector->rx) + adapter->rx_ring[ring->queue_index] = NULL; - err = netif_set_real_num_tx_queues(netdev, adapter->num_tx_queues); - if (err) - return err; + adapter->q_vector[v_idx] = NULL; + netif_napi_del(&q_vector->napi); - return netif_set_real_num_rx_queues(netdev, adapter->num_rx_queues); + /* ixgbevf_get_stats() might access the rings on this vector, + * we must wait a grace period before freeing it. + */ + kfree_rcu(q_vector, rcu); } /** @@ -2515,35 +2834,58 @@ static int ixgbevf_set_interrupt_capability(struct ixgbevf_adapter *adapter) **/ static int ixgbevf_alloc_q_vectors(struct ixgbevf_adapter *adapter) { - int q_idx, num_q_vectors; - struct ixgbevf_q_vector *q_vector; + int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS; + int rxr_remaining = adapter->num_rx_queues; + int txr_remaining = adapter->num_tx_queues; + int xdp_remaining = adapter->num_xdp_queues; + int rxr_idx = 0, txr_idx = 0, xdp_idx = 0, v_idx = 0; + int err; + + if (q_vectors >= (rxr_remaining + txr_remaining + xdp_remaining)) { + for (; rxr_remaining; v_idx++, q_vectors--) { + int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors); - num_q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS; + err = ixgbevf_alloc_q_vector(adapter, v_idx, + 0, 0, 0, 0, rqpv, rxr_idx); + if (err) + goto err_out; - for (q_idx = 0; q_idx < num_q_vectors; q_idx++) { - q_vector = kzalloc(sizeof(struct ixgbevf_q_vector), GFP_KERNEL); - if (!q_vector) + /* update counts and index */ + rxr_remaining -= rqpv; + rxr_idx += rqpv; + } + } + + for (; q_vectors; v_idx++, q_vectors--) { + int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors); + int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors); + int xqpv = DIV_ROUND_UP(xdp_remaining, q_vectors); + + err = ixgbevf_alloc_q_vector(adapter, v_idx, + tqpv, txr_idx, + xqpv, xdp_idx, + rqpv, rxr_idx); + + if (err) goto err_out; - q_vector->adapter = adapter; - q_vector->v_idx = q_idx; - netif_napi_add(adapter->netdev, &q_vector->napi, - ixgbevf_poll, 64); - adapter->q_vector[q_idx] = q_vector; + + /* update counts and index */ + rxr_remaining -= rqpv; + rxr_idx += rqpv; + txr_remaining -= tqpv; + txr_idx += tqpv; + xdp_remaining -= xqpv; + xdp_idx += xqpv; } return 0; err_out: - while (q_idx) { - q_idx--; - q_vector = adapter->q_vector[q_idx]; -#ifdef CONFIG_NET_RX_BUSY_POLL - napi_hash_del(&q_vector->napi); -#endif - netif_napi_del(&q_vector->napi); - kfree(q_vector); - adapter->q_vector[q_idx] = NULL; + while (v_idx) { + v_idx--; + ixgbevf_free_q_vector(adapter, v_idx); } + return -ENOMEM; } @@ -2557,17 +2899,11 @@ err_out: **/ static void ixgbevf_free_q_vectors(struct ixgbevf_adapter *adapter) { - int q_idx, num_q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS; - - for (q_idx = 0; q_idx < num_q_vectors; q_idx++) { - struct ixgbevf_q_vector *q_vector = adapter->q_vector[q_idx]; + int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS; - adapter->q_vector[q_idx] = NULL; -#ifdef CONFIG_NET_RX_BUSY_POLL - napi_hash_del(&q_vector->napi); -#endif - netif_napi_del(&q_vector->napi); - kfree(q_vector); + while (q_vectors) { + q_vectors--; + ixgbevf_free_q_vector(adapter, q_vectors); } } @@ -2611,21 +2947,14 @@ static int ixgbevf_init_interrupt_scheme(struct ixgbevf_adapter *adapter) goto err_alloc_q_vectors; } - err = ixgbevf_alloc_queues(adapter); - if (err) { - pr_err("Unable to allocate memory for queues\n"); - goto err_alloc_queues; - } - - hw_dbg(&adapter->hw, "Multiqueue %s: Rx Queue count = %u, Tx Queue count = %u\n", - (adapter->num_rx_queues > 1) ? "Enabled" : - "Disabled", adapter->num_rx_queues, adapter->num_tx_queues); + hw_dbg(&adapter->hw, "Multiqueue %s: Rx Queue count = %u, Tx Queue count = %u XDP Queue count %u\n", + (adapter->num_rx_queues > 1) ? "Enabled" : "Disabled", + adapter->num_rx_queues, adapter->num_tx_queues, + adapter->num_xdp_queues); set_bit(__IXGBEVF_DOWN, &adapter->state); return 0; -err_alloc_queues: - ixgbevf_free_q_vectors(adapter); err_alloc_q_vectors: ixgbevf_reset_interrupt_capability(adapter); err_set_interrupt: @@ -2641,18 +2970,8 @@ err_set_interrupt: **/ static void ixgbevf_clear_interrupt_scheme(struct ixgbevf_adapter *adapter) { - int i; - - for (i = 0; i < adapter->num_tx_queues; i++) { - kfree(adapter->tx_ring[i]); - adapter->tx_ring[i] = NULL; - } - for (i = 0; i < adapter->num_rx_queues; i++) { - kfree(adapter->rx_ring[i]); - adapter->rx_ring[i] = NULL; - } - adapter->num_tx_queues = 0; + adapter->num_xdp_queues = 0; adapter->num_rx_queues = 0; ixgbevf_free_q_vectors(adapter); @@ -2860,6 +3179,8 @@ static void ixgbevf_check_hang_subtask(struct ixgbevf_adapter *adapter) if (netif_carrier_ok(adapter->netdev)) { for (i = 0; i < adapter->num_tx_queues; i++) set_check_for_tx_hang(adapter->tx_ring[i]); + for (i = 0; i < adapter->num_xdp_queues; i++) + set_check_for_tx_hang(adapter->xdp_ring[i]); } /* get one bit for every active Tx/Rx interrupt vector */ @@ -3031,6 +3352,9 @@ static void ixgbevf_free_all_tx_resources(struct ixgbevf_adapter *adapter) for (i = 0; i < adapter->num_tx_queues; i++) if (adapter->tx_ring[i]->desc) ixgbevf_free_tx_resources(adapter->tx_ring[i]); + for (i = 0; i < adapter->num_xdp_queues; i++) + if (adapter->xdp_ring[i]->desc) + ixgbevf_free_tx_resources(adapter->xdp_ring[i]); } /** @@ -3081,26 +3405,44 @@ err: **/ static int ixgbevf_setup_all_tx_resources(struct ixgbevf_adapter *adapter) { - int i, err = 0; + int i, j = 0, err = 0; for (i = 0; i < adapter->num_tx_queues; i++) { err = ixgbevf_setup_tx_resources(adapter->tx_ring[i]); if (!err) continue; hw_dbg(&adapter->hw, "Allocation for Tx Queue %u failed\n", i); + goto err_setup_tx; + } + + for (j = 0; j < adapter->num_xdp_queues; j++) { + err = ixgbevf_setup_tx_resources(adapter->xdp_ring[j]); + if (!err) + continue; + hw_dbg(&adapter->hw, "Allocation for XDP Queue %u failed\n", j); break; } + return 0; +err_setup_tx: + /* rewind the index freeing the rings as we go */ + while (j--) + ixgbevf_free_tx_resources(adapter->xdp_ring[j]); + while (i--) + ixgbevf_free_tx_resources(adapter->tx_ring[i]); + return err; } /** * ixgbevf_setup_rx_resources - allocate Rx resources (Descriptors) + * @adapter: board private structure * @rx_ring: Rx descriptor ring (for a specific queue) to setup * * Returns 0 on success, negative on failure **/ -int ixgbevf_setup_rx_resources(struct ixgbevf_ring *rx_ring) +int ixgbevf_setup_rx_resources(struct ixgbevf_adapter *adapter, + struct ixgbevf_ring *rx_ring) { int size; @@ -3121,6 +3463,13 @@ int ixgbevf_setup_rx_resources(struct ixgbevf_ring *rx_ring) if (!rx_ring->desc) goto err; + /* XDP RX-queue info */ + if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, adapter->netdev, + rx_ring->queue_index) < 0) + goto err; + + rx_ring->xdp_prog = adapter->xdp_prog; + return 0; err: vfree(rx_ring->rx_buffer_info); @@ -3144,12 +3493,18 @@ static int ixgbevf_setup_all_rx_resources(struct ixgbevf_adapter *adapter) int i, err = 0; for (i = 0; i < adapter->num_rx_queues; i++) { - err = ixgbevf_setup_rx_resources(adapter->rx_ring[i]); + err = ixgbevf_setup_rx_resources(adapter, adapter->rx_ring[i]); if (!err) continue; hw_dbg(&adapter->hw, "Allocation for Rx Queue %u failed\n", i); - break; + goto err_setup_rx; } + + return 0; +err_setup_rx: + /* rewind the index freeing the rings as we go */ + while (i--) + ixgbevf_free_rx_resources(adapter->rx_ring[i]); return err; } @@ -3163,6 +3518,8 @@ void ixgbevf_free_rx_resources(struct ixgbevf_ring *rx_ring) { ixgbevf_clean_rx_ring(rx_ring); + rx_ring->xdp_prog = NULL; + xdp_rxq_info_unreg(&rx_ring->xdp_rxq); vfree(rx_ring->rx_buffer_info); rx_ring->rx_buffer_info = NULL; @@ -3244,28 +3601,31 @@ int ixgbevf_open(struct net_device *netdev) ixgbevf_configure(adapter); - /* Map the Tx/Rx rings to the vectors we were allotted. - * if request_irq will be called in this function map_rings - * must be called *before* up_complete - */ - ixgbevf_map_rings_to_vectors(adapter); - err = ixgbevf_request_irq(adapter); if (err) goto err_req_irq; + /* Notify the stack of the actual queue counts. */ + err = netif_set_real_num_tx_queues(netdev, adapter->num_tx_queues); + if (err) + goto err_set_queues; + + err = netif_set_real_num_rx_queues(netdev, adapter->num_rx_queues); + if (err) + goto err_set_queues; + ixgbevf_up_complete(adapter); return 0; +err_set_queues: + ixgbevf_free_irq(adapter); err_req_irq: - ixgbevf_down(adapter); -err_setup_rx: ixgbevf_free_all_rx_resources(adapter); -err_setup_tx: +err_setup_rx: ixgbevf_free_all_tx_resources(adapter); +err_setup_tx: ixgbevf_reset(adapter); - err_setup_reset: return err; @@ -3707,11 +4067,10 @@ static int ixgbevf_maybe_stop_tx(struct ixgbevf_ring *tx_ring, int size) return __ixgbevf_maybe_stop_tx(tx_ring, size); } -static int ixgbevf_xmit_frame(struct sk_buff *skb, struct net_device *netdev) +static int ixgbevf_xmit_frame_ring(struct sk_buff *skb, + struct ixgbevf_ring *tx_ring) { - struct ixgbevf_adapter *adapter = netdev_priv(netdev); struct ixgbevf_tx_buffer *first; - struct ixgbevf_ring *tx_ring; int tso; u32 tx_flags = 0; u16 count = TXD_USE_COUNT(skb_headlen(skb)); @@ -3726,8 +4085,6 @@ static int ixgbevf_xmit_frame(struct sk_buff *skb, struct net_device *netdev) return NETDEV_TX_OK; } - tx_ring = adapter->tx_ring[skb->queue_mapping]; - /* need: 1 descriptor per page * PAGE_SIZE/IXGBE_MAX_DATA_PER_TXD, * + 1 desc for skb_headlen/IXGBE_MAX_DATA_PER_TXD, * + 2 desc gap to keep tail from touching head, @@ -3780,6 +4137,29 @@ out_drop: return NETDEV_TX_OK; } +static int ixgbevf_xmit_frame(struct sk_buff *skb, struct net_device *netdev) +{ + struct ixgbevf_adapter *adapter = netdev_priv(netdev); + struct ixgbevf_ring *tx_ring; + + if (skb->len <= 0) { + dev_kfree_skb_any(skb); + return NETDEV_TX_OK; + } + + /* The minimum packet size for olinfo paylen is 17 so pad the skb + * in order to meet this minimum size requirement. + */ + if (skb->len < 17) { + if (skb_padto(skb, 17)) + return NETDEV_TX_OK; + skb->len = 17; + } + + tx_ring = adapter->tx_ring[skb->queue_mapping]; + return ixgbevf_xmit_frame_ring(skb, tx_ring); +} + /** * ixgbevf_set_mac - Change the Ethernet Address of the NIC * @netdev: network interface device structure @@ -3826,6 +4206,12 @@ static int ixgbevf_change_mtu(struct net_device *netdev, int new_mtu) int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN; int ret; + /* prevent MTU being changed to a size unsupported by XDP */ + if (adapter->xdp_prog) { + dev_warn(&adapter->pdev->dev, "MTU cannot be changed while XDP program is loaded\n"); + return -EPERM; + } + spin_lock_bh(&adapter->mbx_lock); /* notify the PF of our intent to use this size of frame */ ret = hw->mac.ops.set_rlpml(hw, max_frame); @@ -3839,6 +4225,9 @@ static int ixgbevf_change_mtu(struct net_device *netdev, int new_mtu) /* must set new MTU before calling down or up */ netdev->mtu = new_mtu; + if (netif_running(netdev)) + ixgbevf_reinit_locked(adapter); + return 0; } @@ -3917,17 +4306,11 @@ static int ixgbevf_resume(struct pci_dev *pdev) rtnl_lock(); err = ixgbevf_init_interrupt_scheme(adapter); + if (!err && netif_running(netdev)) + err = ixgbevf_open(netdev); rtnl_unlock(); - if (err) { - dev_err(&pdev->dev, "Cannot initialize interrupts\n"); + if (err) return err; - } - - if (netif_running(netdev)) { - err = ixgbevf_open(netdev); - if (err) - return err; - } netif_device_attach(netdev); @@ -3940,6 +4323,23 @@ static void ixgbevf_shutdown(struct pci_dev *pdev) ixgbevf_suspend(pdev, PMSG_SUSPEND); } +static void ixgbevf_get_tx_ring_stats(struct rtnl_link_stats64 *stats, + const struct ixgbevf_ring *ring) +{ + u64 bytes, packets; + unsigned int start; + + if (ring) { + do { + start = u64_stats_fetch_begin_irq(&ring->syncp); + bytes = ring->stats.bytes; + packets = ring->stats.packets; + } while (u64_stats_fetch_retry_irq(&ring->syncp, start)); + stats->tx_bytes += bytes; + stats->tx_packets += packets; + } +} + static void ixgbevf_get_stats(struct net_device *netdev, struct rtnl_link_stats64 *stats) { @@ -3953,6 +4353,7 @@ static void ixgbevf_get_stats(struct net_device *netdev, stats->multicast = adapter->stats.vfmprc - adapter->stats.base_vfmprc; + rcu_read_lock(); for (i = 0; i < adapter->num_rx_queues; i++) { ring = adapter->rx_ring[i]; do { @@ -3966,14 +4367,14 @@ static void ixgbevf_get_stats(struct net_device *netdev, for (i = 0; i < adapter->num_tx_queues; i++) { ring = adapter->tx_ring[i]; - do { - start = u64_stats_fetch_begin_irq(&ring->syncp); - bytes = ring->stats.bytes; - packets = ring->stats.packets; - } while (u64_stats_fetch_retry_irq(&ring->syncp, start)); - stats->tx_bytes += bytes; - stats->tx_packets += packets; + ixgbevf_get_tx_ring_stats(stats, ring); + } + + for (i = 0; i < adapter->num_xdp_queues; i++) { + ring = adapter->xdp_ring[i]; + ixgbevf_get_tx_ring_stats(stats, ring); } + rcu_read_unlock(); } #define IXGBEVF_MAX_MAC_HDR_LEN 127 @@ -4010,6 +4411,64 @@ ixgbevf_features_check(struct sk_buff *skb, struct net_device *dev, return features; } +static int ixgbevf_xdp_setup(struct net_device *dev, struct bpf_prog *prog) +{ + int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN; + struct ixgbevf_adapter *adapter = netdev_priv(dev); + struct bpf_prog *old_prog; + + /* verify ixgbevf ring attributes are sufficient for XDP */ + for (i = 0; i < adapter->num_rx_queues; i++) { + struct ixgbevf_ring *ring = adapter->rx_ring[i]; + + if (frame_size > ixgbevf_rx_bufsz(ring)) + return -EINVAL; + } + + old_prog = xchg(&adapter->xdp_prog, prog); + + /* If transitioning XDP modes reconfigure rings */ + if (!!prog != !!old_prog) { + /* Hardware has to reinitialize queues and interrupts to + * match packet buffer alignment. Unfortunately, the + * hardware is not flexible enough to do this dynamically. + */ + if (netif_running(dev)) + ixgbevf_close(dev); + + ixgbevf_clear_interrupt_scheme(adapter); + ixgbevf_init_interrupt_scheme(adapter); + + if (netif_running(dev)) + ixgbevf_open(dev); + } else { + for (i = 0; i < adapter->num_rx_queues; i++) + xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog); + } + + if (old_prog) + bpf_prog_put(old_prog); + + return 0; +} + +static int ixgbevf_xdp(struct net_device *dev, struct netdev_bpf *xdp) +{ + struct ixgbevf_adapter *adapter = netdev_priv(dev); + + switch (xdp->command) { + case XDP_SETUP_PROG: + return ixgbevf_xdp_setup(dev, xdp->prog); + case XDP_QUERY_PROG: + xdp->prog_attached = !!(adapter->xdp_prog); + xdp->prog_id = adapter->xdp_prog ? + adapter->xdp_prog->aux->id : 0; + return 0; + default: + return -EINVAL; + } +} + static const struct net_device_ops ixgbevf_netdev_ops = { .ndo_open = ixgbevf_open, .ndo_stop = ixgbevf_close, @@ -4026,6 +4485,7 @@ static const struct net_device_ops ixgbevf_netdev_ops = { .ndo_poll_controller = ixgbevf_netpoll, #endif .ndo_features_check = ixgbevf_features_check, + .ndo_bpf = ixgbevf_xdp, }; static void ixgbevf_assign_netdev_ops(struct net_device *dev) |