From a7559bc8c17c3f9a91dcbeefe8642ba757fd09e8 Mon Sep 17 00:00:00 2001 From: Andy Gospodarek Date: Fri, 8 Apr 2022 03:59:05 -0400 Subject: [PATCH] bnxt: support transmit and free of aggregation buffers This patch adds the following features: - Support for XDP_TX and XDP_DROP action when using xdp_buff with frags - Support for freeing all frags attached to an xdp_buff - Cleanup of TX ring buffers after transmits complete - Slight change in definition of bnxt_sw_tx_bd since nr_frags and RX producer may both need to be used - Clear out skb_shared_info at the end of the buffer v2: Fix uninitialized variable warning in bnxt_xdp_buff_frags_free(). Signed-off-by: Andy Gospodarek Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 18 ++- drivers/net/ethernet/broadcom/bnxt/bnxt.h | 7 +- .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 2 +- drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c | 117 ++++++++++++++++-- drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h | 5 +- 5 files changed, 126 insertions(+), 23 deletions(-) diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index f96f41c7927e..25d74c9030fd 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -1951,9 +1951,13 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr, skb = bnxt_copy_skb(bnapi, data_ptr, len, dma_addr); bnxt_reuse_rx_data(rxr, cons, data); if (!skb) { - if (agg_bufs) - bnxt_reuse_rx_agg_bufs(cpr, cp_cons, 0, - agg_bufs, false); + if (agg_bufs) { + if (!xdp_active) + bnxt_reuse_rx_agg_bufs(cpr, cp_cons, 0, + agg_bufs, false); + else + bnxt_xdp_buff_frags_free(rxr, &xdp); + } cpr->sw_stats.rx.rx_oom_discards += 1; rc = -ENOMEM; goto next_rx; @@ -1986,6 +1990,7 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr, skb = bnxt_xdp_build_skb(bp, skb, agg_bufs, rxr->page_pool, &xdp, rxcmp1); if (!skb) { /* we should be able to free the old skb here */ + bnxt_xdp_buff_frags_free(rxr, &xdp); cpr->sw_stats.rx.rx_oom_discards += 1; rc = -ENOMEM; goto next_rx; @@ -2605,10 +2610,13 @@ static void __bnxt_poll_work_done(struct bnxt *bp, struct bnxt_napi *bnapi) if ((bnapi->events & BNXT_RX_EVENT) && !(bnapi->in_reset)) { struct bnxt_rx_ring_info *rxr = bnapi->rx_ring; - if (bnapi->events & BNXT_AGG_EVENT) - bnxt_db_write(bp, &rxr->rx_agg_db, rxr->rx_agg_prod); bnxt_db_write(bp, &rxr->rx_db, rxr->rx_prod); } + if (bnapi->events & BNXT_AGG_EVENT) { + struct bnxt_rx_ring_info *rxr = bnapi->rx_ring; + + bnxt_db_write(bp, &rxr->rx_agg_db, rxr->rx_agg_prod); + } bnapi->events = 0; } diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h index 319d6851eecc..a498ee297946 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h @@ -701,13 +701,12 @@ struct bnxt_sw_tx_bd { }; DEFINE_DMA_UNMAP_ADDR(mapping); DEFINE_DMA_UNMAP_LEN(len); + struct page *page; u8 is_gso; u8 is_push; u8 action; - union { - unsigned short nr_frags; - u16 rx_prod; - }; + unsigned short nr_frags; + u16 rx_prod; }; struct bnxt_sw_rx_bd { diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c index 22e965e18fbc..b3a48d6675fe 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c @@ -3491,7 +3491,7 @@ static int bnxt_run_loopback(struct bnxt *bp) dev_kfree_skb(skb); return -EIO; } - bnxt_xmit_bd(bp, txr, map, pkt_size); + bnxt_xmit_bd(bp, txr, map, pkt_size, NULL); /* Sync BD data before updating doorbell */ wmb(); diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c index 5183357ca11c..c2905f0a8c6c 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c @@ -24,36 +24,91 @@ DEFINE_STATIC_KEY_FALSE(bnxt_xdp_locking_key); struct bnxt_sw_tx_bd *bnxt_xmit_bd(struct bnxt *bp, struct bnxt_tx_ring_info *txr, - dma_addr_t mapping, u32 len) + dma_addr_t mapping, u32 len, + struct xdp_buff *xdp) { - struct bnxt_sw_tx_bd *tx_buf; + struct skb_shared_info *sinfo; + struct bnxt_sw_tx_bd *tx_buf, *first_buf; struct tx_bd *txbd; + int num_frags = 0; u32 flags; u16 prod; + int i; + if (xdp && xdp_buff_has_frags(xdp)) { + sinfo = xdp_get_shared_info_from_buff(xdp); + num_frags = sinfo->nr_frags; + } + + /* fill up the first buffer */ prod = txr->tx_prod; tx_buf = &txr->tx_buf_ring[prod]; + first_buf = tx_buf; + tx_buf->nr_frags = num_frags; + if (xdp) + tx_buf->page = virt_to_head_page(xdp->data); txbd = &txr->tx_desc_ring[TX_RING(prod)][TX_IDX(prod)]; - flags = (len << TX_BD_LEN_SHIFT) | (1 << TX_BD_FLAGS_BD_CNT_SHIFT) | - TX_BD_FLAGS_PACKET_END | bnxt_lhint_arr[len >> 9]; + flags = ((len) << TX_BD_LEN_SHIFT) | ((num_frags + 1) << TX_BD_FLAGS_BD_CNT_SHIFT); txbd->tx_bd_len_flags_type = cpu_to_le32(flags); txbd->tx_bd_opaque = prod; txbd->tx_bd_haddr = cpu_to_le64(mapping); + /* now let us fill up the frags into the next buffers */ + for (i = 0; i < num_frags ; i++) { + skb_frag_t *frag = &sinfo->frags[i]; + struct bnxt_sw_tx_bd *frag_tx_buf; + struct pci_dev *pdev = bp->pdev; + dma_addr_t frag_mapping; + int frag_len; + + prod = NEXT_TX(prod); + txr->tx_prod = prod; + + /* first fill up the first buffer */ + frag_tx_buf = &txr->tx_buf_ring[prod]; + frag_tx_buf->page = skb_frag_page(frag); + + txbd = &txr->tx_desc_ring[TX_RING(prod)][TX_IDX(prod)]; + + frag_len = skb_frag_size(frag); + frag_mapping = skb_frag_dma_map(&pdev->dev, frag, 0, + frag_len, DMA_TO_DEVICE); + + if (unlikely(dma_mapping_error(&pdev->dev, frag_mapping))) + return NULL; + + dma_unmap_addr_set(frag_tx_buf, mapping, frag_mapping); + + flags = frag_len << TX_BD_LEN_SHIFT; + txbd->tx_bd_len_flags_type = cpu_to_le32(flags); + txbd->tx_bd_opaque = prod; + txbd->tx_bd_haddr = cpu_to_le64(frag_mapping); + + len = frag_len; + } + + flags &= ~TX_BD_LEN; + txbd->tx_bd_len_flags_type = cpu_to_le32(((len) << TX_BD_LEN_SHIFT) | flags | + TX_BD_FLAGS_PACKET_END); + /* Sync TX BD */ + wmb(); prod = NEXT_TX(prod); txr->tx_prod = prod; - return tx_buf; + + return first_buf; } static void __bnxt_xmit_xdp(struct bnxt *bp, struct bnxt_tx_ring_info *txr, - dma_addr_t mapping, u32 len, u16 rx_prod) + dma_addr_t mapping, u32 len, u16 rx_prod, + struct xdp_buff *xdp) { struct bnxt_sw_tx_bd *tx_buf; - tx_buf = bnxt_xmit_bd(bp, txr, mapping, len); + tx_buf = bnxt_xmit_bd(bp, txr, mapping, len, xdp); tx_buf->rx_prod = rx_prod; tx_buf->action = XDP_TX; + } static void __bnxt_xmit_xdp_redirect(struct bnxt *bp, @@ -63,7 +118,7 @@ static void __bnxt_xmit_xdp_redirect(struct bnxt *bp, { struct bnxt_sw_tx_bd *tx_buf; - tx_buf = bnxt_xmit_bd(bp, txr, mapping, len); + tx_buf = bnxt_xmit_bd(bp, txr, mapping, len, NULL); tx_buf->action = XDP_REDIRECT; tx_buf->xdpf = xdpf; dma_unmap_addr_set(tx_buf, mapping, mapping); @@ -78,7 +133,7 @@ void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int nr_pkts) struct bnxt_sw_tx_bd *tx_buf; u16 tx_cons = txr->tx_cons; u16 last_tx_cons = tx_cons; - int i; + int i, j, frags; for (i = 0; i < nr_pkts; i++) { tx_buf = &txr->tx_buf_ring[tx_cons]; @@ -96,6 +151,13 @@ void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int nr_pkts) } else if (tx_buf->action == XDP_TX) { rx_doorbell_needed = true; last_tx_cons = tx_cons; + + frags = tx_buf->nr_frags; + for (j = 0; j < frags; j++) { + tx_cons = NEXT_TX(tx_cons); + tx_buf = &txr->tx_buf_ring[tx_cons]; + page_pool_recycle_direct(rxr->page_pool, tx_buf->page); + } } tx_cons = NEXT_TX(tx_cons); } @@ -103,6 +165,7 @@ void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int nr_pkts) if (rx_doorbell_needed) { tx_buf = &txr->tx_buf_ring[last_tx_cons]; bnxt_db_write(bp, &rxr->rx_db, tx_buf->rx_prod); + } } @@ -133,6 +196,23 @@ void bnxt_xdp_buff_init(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, xdp_prepare_buff(xdp, *data_ptr - offset, offset, *len, false); } +void bnxt_xdp_buff_frags_free(struct bnxt_rx_ring_info *rxr, + struct xdp_buff *xdp) +{ + struct skb_shared_info *shinfo; + int i; + + if (!xdp || !xdp_buff_has_frags(xdp)) + return; + shinfo = xdp_get_shared_info_from_buff(xdp); + for (i = 0; i < shinfo->nr_frags; i++) { + struct page *page = skb_frag_page(&shinfo->frags[i]); + + page_pool_recycle_direct(rxr->page_pool, page); + } + shinfo->nr_frags = 0; +} + /* returns the following: * true - packet consumed by XDP and new buffer is allocated. * false - packet should be passed to the stack. @@ -145,6 +225,7 @@ bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons, struct bnxt_sw_rx_bd *rx_buf; struct pci_dev *pdev; dma_addr_t mapping; + u32 tx_needed = 1; void *orig_data; u32 tx_avail; u32 offset; @@ -180,18 +261,28 @@ bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons, case XDP_TX: rx_buf = &rxr->rx_buf_ring[cons]; mapping = rx_buf->mapping - bp->rx_dma_offset; + *event = 0; - if (tx_avail < 1) { + if (unlikely(xdp_buff_has_frags(&xdp))) { + struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(&xdp); + + tx_needed += sinfo->nr_frags; + *event = BNXT_AGG_EVENT; + } + + if (tx_avail < tx_needed) { trace_xdp_exception(bp->dev, xdp_prog, act); + bnxt_xdp_buff_frags_free(rxr, &xdp); bnxt_reuse_rx_data(rxr, cons, page); return true; } - *event = BNXT_TX_EVENT; dma_sync_single_for_device(&pdev->dev, mapping + offset, *len, bp->rx_dir); + + *event |= BNXT_TX_EVENT; __bnxt_xmit_xdp(bp, txr, mapping + offset, *len, - NEXT_RX(rxr->rx_prod)); + NEXT_RX(rxr->rx_prod), &xdp); bnxt_reuse_rx_data(rxr, cons, page); return true; case XDP_REDIRECT: @@ -208,6 +299,7 @@ bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons, /* if we are unable to allocate a new buffer, abort and reuse */ if (bnxt_alloc_rx_data(bp, rxr, rxr->rx_prod, GFP_ATOMIC)) { trace_xdp_exception(bp->dev, xdp_prog, act); + bnxt_xdp_buff_frags_free(rxr, &xdp); bnxt_reuse_rx_data(rxr, cons, page); return true; } @@ -227,6 +319,7 @@ bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons, trace_xdp_exception(bp->dev, xdp_prog, act); fallthrough; case XDP_DROP: + bnxt_xdp_buff_frags_free(rxr, &xdp); bnxt_reuse_rx_data(rxr, cons, page); break; } diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h index 27290f649be3..505911ae095d 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h @@ -14,7 +14,8 @@ DECLARE_STATIC_KEY_FALSE(bnxt_xdp_locking_key); struct bnxt_sw_tx_bd *bnxt_xmit_bd(struct bnxt *bp, struct bnxt_tx_ring_info *txr, - dma_addr_t mapping, u32 len); + dma_addr_t mapping, u32 len, + struct xdp_buff *xdp); void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int nr_pkts); bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons, struct xdp_buff xdp, struct page *page, unsigned int *len, @@ -28,6 +29,8 @@ bool bnxt_xdp_attached(struct bnxt *bp, struct bnxt_rx_ring_info *rxr); void bnxt_xdp_buff_init(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons, u8 **data_ptr, unsigned int *len, struct xdp_buff *xdp); +void bnxt_xdp_buff_frags_free(struct bnxt_rx_ring_info *rxr, + struct xdp_buff *xdp); struct sk_buff *bnxt_xdp_build_skb(struct bnxt *bp, struct sk_buff *skb, u8 num_frags, struct page_pool *pool, struct xdp_buff *xdp,